A Trem2R47H mouse model without cryptic splicing drives age- and disease-dependent tissue damage and synaptic loss in response to plaques

The Trem2 R47H NSS mutation promotes loss of oligodendrocyte gene expression in response to cuprizone treatment.

Results of previous studies of mice with the Trem2R47H missense mutation introduced via CRISPR suggested that it acts as a near-complete loss of function, recapitulating phenotypes seen in Trem2 knock-out (KO) mice [34, 36]. However, subsequent analyses of Trem2 expression and splicing in these models identified the unexpected generation of a cryptic splice site and subsequent reduction of Trem2 expression due to the synonymous codon changes introduced as part of the CRISPR repair template [35]. Given the importance of a Trem2R47H mouse model that more accurately recapitulates human cases, we designed alternative CRISPR repair templates, guided in part by Cheng et al., to introduce the R47H mutation into C57BL/6 J mice [44]. Confirmation of the desired sequence change (CGC – > CAT; arginine – > histidine) and synonymous silent codon changes are shown in Fig. 1a. To analyze expression and isoform usage of Trem2 from the Trem2R47H NSS allele, we extracted RNA from whole brains of 15 week old wild-type and Trem2R47H NSS mice, as well as from a cohort of Trem2R47H CSS mice with the identified Cryptic Splice Site (CSS) and reduced expression [35] and conducted Nanopore long-read RNA-seq to examine the association of exons within individual transcripts. We identified three known annotated Trem2 isoforms – Trem2-201 encoding transmembrane TREM2, Trem2-202 encoding secreted TREM2, and Trem2-203 which contains retained introns (Fig. 1b, Supplemental Fig. 1). However, a novel truncated exon 2 containing isoform was abundant in the Trem2R47H CSS brain and was not present in Trem2R47H NSS mice. We designated our model as Trem2R47H NSS (Normal Splice Site, NSS; available at The Jackson Laboratory—stock #034,036). Furthermore, reads of the major isoform, Trem2-201, are reduced in the Trem2R47H CSS brains (> 50% reduction—wild-type mean TPM = 43.7, NSS = 33.8, CSS = 11.03; Fig. 1c).

Fig. 1figure 1

Cuprizone model of demyelination using wild-type, Trem2R47H NSS, Trem2R47H CSS, and Trem2 KO mice. a Amino acid coding sequence within exon 2 of mouse wild type Trem2 and Trem2 R47H NSS alleles. The triplet codon for arginine (R, green), the G to A transition that encodes histidine (H, red), and ten silent DNA mutation (tan) are shown in the Trem2R47H NSS allele. b Transcript models from long-read RNA-seq analysis showing Trem2 isoforms 201, 202, 203, and the truncated exon 2 from cryptic splicing reported in Trem2R47H CSS. The green asterisk denotes the truncated exon 2. c Trem2 isoform TPM values of wild-type, Trem2R47H NSS and Trem2R47H CSS. d Cuprizone (CPZ) feeding scheme of wild-type, Trem2R47H NSS, Trem2R47H CSS and Trem2 KO mice. e Trem2 TPM values of wild-type, Trem2R47H NSS, Trem2R47H CSS and Trem2 KO males examined from whole-brain bulk RNA-seq data showed similar increase in Trem2 expression in response to CPZ treatment in wild-type and Trem2R47H NSS that was attenuated in Trem2R47H CSS and Trem2 KO mice. f Representative whole-brain stitched images of wild-type, Trem2R47H NSS, Trem2R47H CSS and Trem2 KO males on control or CPZ diet stained for myelinated fibers with Luxol Fast Blue. Yellow dotted lines indicate region of demyelination. Scale bar = 1000 µm. g Quantification of demyelination reveals no change between genotypes. h Plasma neurofilament light-chain (NfL) quantified via Meso Scale Discovery suggests exacerbated axonal damage in Trem2 KO mice. i Representative images of corpus collosum from wild-type, Trem2R47H NSS, Trem2R47H CSS and Trem2 KO mice on control vs CPZ diet stained for myelin basic protein (blue), microglia (IBA1, red), and DAM gene marker (CD11c, green). j Quantification of IBA1+ cells per mm2 revealed expected increase in microgliosis in response to demyelination in CPZ-treated mice across all groups with Trem2 KO having fewer microglia than wild-type. k Quantification of CD11c+ microglia cell volume normalized to microglia volume. n = 4–8. Data are represented as mean ± SEM. Two-way ANOVA followed by Tukey’s post-hoc tests to examine biologically relevant interactions. Statistical significance is denoted by *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

To further validate the Trem2R47H NSS model we conducted an off-target analyses of any other putative cut sites that might have been targeted by CRISPR-Cas9 during generation of the Trem2R47H NSS allele. Trem2R47H NSS founder mice were backcrossed with wild type C57BL/6 J animals for three generations before use to generate animals for this study, making it unlikely that a mutation caused by an off-target effect of CRISPR/Cas9 would be present on a chromosome other than chromosome 17, i.e., the location of Trem2 (C57BL/6 J; Chr17; 48.6 Mb, GRCm39, Ensembl release 108). Potential CRISPR/Cas9 off-target sites with up to four mismatches using crRNA TMF1342 were screened for using Cas-OFFinder (http://www.rgenome.net/cas-offinder/; [45]). Eleven potential off-target sites on mouse chromosome 17 were identified (Supplemental Table 2). Two potential off-target sites were within coding exons of genes, six were within introns, while the remaining three were within intergenic regions. To screen for evidence of CRISPR/Cas9 RNP activity at each site, DNA from a wild-type and homozygous Trem2R47H NSS mouse was amplified by PCR using the primers listed (Supplementary Table 1) then sequenced across the potential off-target region at each locus (Supplemental Fig. 2a-k). None of the 11 potential off-target sites showed a difference in sequence between WT and homozygous Trem2R47H NSS mice. Moreover, no significant change in gene expression was seen in any of these 11 genes via RNA-seq from whole brain extracted RNA (Supplemental Table 2). We also explored transcript isoform production from these genes through long-read RNA-seq and found no change (e.g. data for Zpf945 shown in Supplemental Fig. 2l, m).

Having confirmed correct expression, isoform usage, and an absence of off-target genomic effects, we next assessed the impact of the Trem2R47H NSS variant on microglial function and tested if it acts as a loss of function allele using a cuprizone model of demyelination [46]. Both myelin and fibrillar Aβ act as TREM2 ligands and induce a distinctive gene expression profile in microglia known as a Disease Associated Microglia (DAM) signature, which can be explored with cuprizone. For comparison, we included cohorts of Trem2R47H CSS mice, as well as Trem2 KO mice. These 3 groups and a wild-type group, were treated with cuprizone (0.3%) or control chow for 6 weeks (Fig. 1d), then were euthanized at 15 weeks and brains examined by histology and bulk RNA-seq. Only male mice were used for the cuprizone study due to the influence of estrogen and progesterone on myelination, as well as evidence of the treatment disrupting estrous cycle in mice [47, 48]. RNA was extracted from half brains, and bulk RNA-seq conducted, with a principal component analysis (PCA) plot for all samples shown in Supplemental Fig. 3a. Trem2 expression values were plotted, showing that Trem2R47H NSS mice have similar Trem2 expression levels to wild-type mice consistent with results of the long-read RNA-seq analysis, and that both Trem2R47H CSS and Trem2 KO mice have reduced expression (Fig. 1e). Notably, with cuprizone treatment, Trem2 levels increased similarly in wild-type and Trem2R47H NSS mice, but to a lesser extent in Trem2R47H CSS mice.

Luxol Fast Blue staining for myelin confirmed demyelination in all cuprizone treated mice with no overt difference in the extent of demyelination between the four groups (Fig. 1f, g). Plasma neurofilament light-chain (NfL) reflects axonal damage in the brain [49]. Interestingly, cuprizone treatment/demyelination did not increase plasma NfL in either WT mice, or the Trem2 variants/KO, although a substantial elevation of plasma NfL was present in Trem2 KO mice independent of cuprizone treatment (Fig. 1h). Immunofluorescence analysis for microglia (IBA1) and DAM marker CD11c (encoded by Itgax) reveals extensive microgliosis in the corpus callosum of cuprizone treated wild-type, Trem2R47H NSS, and Trem2R47H CSS mice, and to a lesser extent Trem2 KO mice (Fig. 1i, j). While CD11c expression is induced in cuprizone treated wild-type, Trem2R47H NSS, and Trem2R47H CSS mice, no induction is observed in the Trem2 KO mice (Fig. 1k).

Volcano plots of brain gene expression in cuprizone treated vs. control mice across the 4 groups reveal that animals in wild-type, Trem2R47H NSS, and Trem2R47H CSS groups all show clear upregulation of inflammatory genes in response to cuprizone that is markedly suppressed in the Trem2 KO group (Fig. 2a). We selected differentially expressed genes (DEGs) from the Trem2R47H NSS mice (FDR < 0.05 for cuprizone vs. control) and created a heatmap to compare the response to that in the other 3 groups (Fig. 2b). Upregulated genes are all associated with inflammation and are mostly shared with the wild-type and Trem2R47H CSS groups and include DAM genes such as Apoe, Clec7a, and Itgax (Fig. 2c), consistent with CD11c histology (Fig. 1k). Some inflammatory genes are also significantly upregulated in Trem2 KO mice suggesting their induction is Trem2 independent and includes more classical inflammation related genes such as C1qa, Hmox1, Tyrobp, and Trem2 itself (Fig. 2d and Supplemental Fig. 3b). Downregulated genes in Trem2R47H NSS include a unique set not altered in wild-type, Trem2R47H CSS, or Trem2 KO groups, and are associated with dopaminergic signaling in the striatum (Fig. 2e). Shared downregulated genes are associated with myelin and oligodendrocytes, including Cldn11, Mal, Mobp, Opalin, and Plp1 (Fig. 2f), suggesting that demyelination induced by cuprizone is not dependent upon the Trem2-dependent inflammation. Comparisons of DEGs between the three genotypes compared to wild-type in the absence of cuprizone treatment are shown as volcano plots (Supplemental Fig. 3c). Notably, Trem2R47H CSS brains show more DEGs compared to wild-type mice than Trem2R47H NSS (Supplemental Fig. 3d). Likewise, exploration of the number of DEGs induced by cuprizone in each of the four genotypes further revealed an extensive number of genes induced in the Trem2R47H CSS brains that are not induced by cuprizone in wild-type, Trem2R47H NSS or Trem2 KO brains (Supplemental Fig. 3e; shown as a heatmap in Supplemental Fig. 3f).

Fig. 2figure 2

Trem2R47H NSS induces increased inflammatory response but reduces myelination gene expression in response to cuprizone. a Volcano plot of differentially expressed genes (DEG), displaying fold change of genes (log2 scale) and P values (− log10 scale) between cuprizone (CPZ) treatment vs. control across 4 groups; wild-type, Trem2R47H NSS, Trem2R47H CSS, and Trem2 KO (FC = 0.5, FDR < 0.05). b Heatmap of selected DEG from Trem2R47H NSS (FDR < 0.05 for CPZ vs. control) compared across mouse models (see color scheme in b). Asterisk denotes the group of interest, Trem2R47H NSS on CPZ. c, d List of inflammation DEG upregulated in CPZ compared to control diet that are found to be either (c) Trem2-dependent (upregulated in all groups but Trem2 KO) or (d) Trem2-independent (upregulated in all groups). e List of uniquely downregulated DEG only found in Trem2R47H NSS. f List of myelination-related genes seen down-regulated in CPZ-treated mice across all groups. g Subset of modulerait relationship heatmap by PyWGCNA on wild-type, Trem2R47H NSS, Trem2R47H CSS, and Trem2 KO associated with CPZ treatment. Color corresponding to correlation (red denotes positive correlation; blue denotes negative correlation) and the number in parenthesis shows relative significance of each correlation. Two modules (white and lightcoral modules) were chosen based on their significant correlation with CPZ treatment. h, j Barplots for the eigengene of the genes in the white and lightcoral modules, respectively. i k Gene ontology analysis of the genes in the white and lightcoral modules, respectively. n = 4–8. Data are represented as mean ± SEM. Two-way ANOVA followed by Tukey’s post hoc tests to examine biologically relevant interactions. Statistical significance is denoted by *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

To further explore gene expression changes across the groups, we analyzed functional networks of correlated genes (weighted correlation gene network analysis (WGCNA)) and identified two modules (White and Lightcoral) associated with cuprizone treatment (Fig. 2g). White module eigengene values were increased with cuprizone to similar extents in Trem2R47H NSS and Trem2R47H CSS mice, but significantly decreased in Trem2 KO compared to wild-type mice (Fig. 2h), with gene ontology being associated with inflammation (Fig. 2i). Lightcoral module eigengene values were reduced in both Trem2R47H NSS and Trem2 KO mice with cuprizone but not in wild-type or Trem2R47H CSS mice (Fig. 2j), and gene ontology associated with myelination and lipid metabolism (Fig. 2k), suggesting that the presence of Trem2R47H NSS induces the appropriate inflammatory response to demyelination yet exacerbates its effects on myelination gene expression compared to wild-type Trem2. As myelin acts as a TREM2 ligand, inducing similar gene expression changes in microglia as exposure to Aβ plaques, we then compared all modules to AMP-AD identified modules that define gene expression changes in AD [50], for significant gene overlap (Supplemental Fig. 4). Notably, the White module (i.e., microglia/inflammation) strongly overlaps with the immune response and cytokine signaling AMP-AD modules, while the Lightcoral module (i.e., myelin and lipid metabolism related) strongly overlaps with the myelination AMP-AD modules. In addition to these two modules, the Lightgray module has significant overlap with the neuronal-related AMP-AD modules and is significantly inversely correlated to gene expression in Trem2R47H NSS, but not Trem2R47H CSS or Trem2 KO brains (Fig. 2g). Collectively, these results show that Trem2R47H NSS mice show appropriate expression of Trem2 transcripts, and that the Trem2R47H NSS allele does not function as a null allele as assessed by cuprizone challenge.

The Trem2 R47H NSS allele produces disease stage dependent effects on plaque density, size, and intensity

To investigate the contributions of Trem2R47H to the pathogenesis of AD, we bred Trem2R47H NSS mice with 5xFAD mice on a congenic B6J background to generate 4 groups: (i) wild-type, (ii) homozygous Trem2R47H NSS, (iii) hemizygous 5xFAD, and (iv) hemizygous 5xFAD; homozygous Trem2R47H NSS. We used Trem2R47H NSS homozygotes rather than the heterozygous TREM2R47H allele found in human AD individuals, to exacerbate phenotypes associated with the R47H mutation. Hereafter, for simplicity we refer to the Trem2R47H NSS genotype as Trem2R47H. Mice were aged to 4 and 12 months and analyzed. Four-month-old 5xFAD mice are in their rapid plaque growth stage, which then plateaus throughout the brain by ~ 8–12 months depending on brain region [37]. While plaque densities are similar for both 5xFAD and 5xFAD/Trem2R47H 4-month-old mice when collapsed cross the sexes (Fig. 3c, e), homozygosity for Trem2*R47H induces a robust sex difference in the manifestation of Thio-S+ plaques in both the cortex and subiculum, with lower plaque load in male 5xFAD/Trem2R47H vs. 5xFAD mice (Supplemental Fig. 5b, n). Similar sex differences for plaque densities have been reported for Trem2 KO mice crossed with 5xFAD mice [51]. Furthermore, female 5xFAD/Trem2R47H mice exhibit increased plaque volume in the cortex compared to the age-matched 5xFAD mice (Supplemental Fig. 5e). Plaques in the subiculum of 4-month-old 5xFAD/Trem2R47H mice are smaller with reduced mean plaque intensity compared to 5xFAD mice (Fig. 3f, g). Differences in the effects of Trem2R47H on plaque volume between the cortex (where they are larger) and subiculum (where they are smaller) may be disease stage dependent as pathology begins earlier in the subiculum than the cortex, where at this timepoint plaques are only just beginning to form. Immunostaining for Aβ fibrillary oligomers using the conformation specific antibody OC, revealed less OC+ material, which includes more diffuse material in addition to the dense cores, in 5xFAD/Trem2R47H compared to 5xFAD mice in the subiculum at 4 months (Fig. 3h, i). Quantification of Thio-S+ plaques in the visual cortex at the 12-month time point shows no differences (Fig. 3k-l), but there is a trend towards increased plaque volume in male compared to female 5xFAD/Trem2R47H mice (Supplemental Fig. 5h). In the subiculum, 5xFAD/Trem2R47H mice have higher plaque counts with comparable sizes (Fig. 3m-n). However, increased mean Thio-S+ plaque intensity in 5xFAD/Trem2R47H mice is seen, with no difference in OC+ plaques (Fig. 3o-q). In the cortex, however, there is no difference in either mean plaque intensity or total OC+ volume at both 4- and 12-months (Supplemental Fig. 6).

Fig. 3figure 3

Initial sex-dependent amyloid plaque development in 5xFAD/Trem2R47H mice. a Schematic showing mouse groups and study design. b, j Representative hemispheric coronal brain images of (b) 4-month-old and (j) 12-month-old 5xFAD and 5xFAD/Trem2R47H stained for dense-core plaques using Thioflavin-S (green) with insets of higher magnification images of the (c, k) visual cortex and (e, m) subiculum. Scale bar = 500 µm. c, d In 4-month-old mice, there is a sex-dependent difference in density of Thio-S+ dense-core plaque burden within 5xFAD/Trem2R47H in the visual cortex (c, plaque density; d, plaque volume; vertical significance bars). d, f Average plaque volumes showed no difference in (d) the visual cortex but decreased in (f) the subiculum of 5xFAD/Trem2R47H compared to 5xFAD. g Reduced mean plaque intensity in 5xFAD/Trem2R47H compared to 5xFAD at 4-months of age. h Quantification of total volume of OC+ diffused plaques in the subiculum shows less diffused plaques in 5xFAD/Trem2R47H compared to 5xFAD at 4 months. i, q Representative confocal images of immunostained subiculum for OC (red) and Thio-S (green) at (i) 4-months and (q)12-months. k, m In 12-month old mice, there was no difference in density of dense-core plaques in the visual cortex of 5xFAD and 5xFAD/Trem2R47H mice (k) but an increase in the subiculum (m). l, n No difference was observed in average plaque volume in (l) the cortex and (n) subiculum although there was a trend towards male 5xFAD/Trem2R47H mice having larger plaques than females in (l) the cortex. o At 12 months, total mean plaque intensity is greater in 5xFAD/Trem2R47H compared to 5xFAD. p Total volume of OC+ diffused plaques in the subiculum show lesson difference in diffuse plaques in 5xFAD/Trem2R47H. n = 10–12. Data are represented as mean ± SEM. Student’s t-test. Statistical significance is denoted by *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Statistical trends are given by # 0.05 < p < 0.1

We quantified Aβ40 and Aβ42 in detergent soluble and insoluble fractions from microdissected hippocampi and cortices. At 4 months of age, increased soluble Aβ40 and 42 are found in the 5xFAD/Trem2R47H hippocampus (Fig. 4h) and a trending reduction in soluble Aβ42 in the cortex (Fig. 4f), but no difference in the insoluble fraction in either brain region (Fig. 4a-d), compared to 5xFAD mice. By 12 months of age, Aβ is increased in both brain regions and fractions, with increased soluble Aβ42 in the hippocampus and increased insoluble Aβ40 and 42 in the cortex of 5xFAD/Trem2R47H mice (Fig. 4i-p) compared to 5xFAD mice. Collectively, these results show that Trem2R47H impacts the level of both plaque and Aβ in a brain region and age-specific manner.

Fig. 4figure 4

Quantification of insoluble and soluble Aβ in micro-dissected hippocampi and cortices. a-d No difference in insoluble Aβ40 and Aβ42 in hippocampus in 4-month-old 5xFAD/ Trem2R47H vs 5xFAD mice. eh In the soluble fraction, there is no difference in cortical Aβ40 (e), but a trending decrease in Aβ42 level was observed in the cortical fraction of 4-month-old 5xFAD/ Trem2R47H vs 5xFAD animals (f, p = 0.0997). Increases in (g) Aβ40 and (h) Aβ42 are observed in hippocampal fraction of 5xFAD/ Trem2R47H compared to 5xFAD. i-l At 12-months, insoluble cortical (i) Aβ40 and (j) Aβ42 are increased in 5xFAD/Trem2R47H compare to 5xFAD while no difference was observed in hippocampal fraction (k, l). m-p No difference in soluble fraction except for an increase in hippocampal Aβ42 in 5xFAD/Trem2R47H (p). Data are represented as mean ± SEM. Student’s t-test. Statistical significance is denoted by *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Statistical trends are given by # 0.05 < p < 0.1

Initial impairment in microglial-plaque interactions is lost with age/disease progression

Given the expression of Trem2 by microglia in the brain, we looked for changes in microglial densities and morphologies in both a non-pathologic and early pathologic (i.e., cortical regions of 5xFAD mice without overt dense core plaque load) state. Morphological analyses of cortical IBA1+ microglia showed increased process length but decreased diameter in Trem2R47H compared to WT mice, while these differences were less apparent in the 5xFAD hemizygous background (Fig. 5a, b; images and sample analyses shown in Supplemental Fig. 7). We next imaged microglia and plaques in the visual cortex and subiculum (Fig. 5c, d). At 4-months old, as expected, the sex difference observed in plaque load of 5xFAD/Trem2R47H mice is reflected in cortical microglial density with females showing higher microglia and plaque densities (Fig. 5e; data separated by sex included in Supplemental Fig. 8). In the subiculum, where Thio-S+ plaques are abundant, IBA1+ microglial density and size are increased in 5xFAD and 5xFAD/Trem2R47H compared to wild-type mice (Fig. 5g, h). However, IBA1+ cell densities and volumes of 5xFAD/Trem2R47H are lower than 5xFAD mice (Fig. 5g, h). Notably, in both brain regions, Trem2R47H mice exhibit lower IBA1+ cell volume compared to WT, indicating that the Trem2R47H variant elicits a plaque-independent effect on microglia morphology (Fig. 5f, h). A lack of plaque-microglia interaction is found in 5xFAD/Trem2R47H mice as shown through quantification of Thio-S and IBA1 colocalization in the subiculum, with a further impairment in female 5xFAD/Trem2R47H mice compared to males (Fig. 5o). However, the decreased microglia volume in Trem2R47H mice and 5xFAD/Trem2R47H compared to WT and 5xFAD, respectively, is absent at 12-month (Fig. 5 j, l), as is the impairment in plaque-microglia interaction, suggesting age/disease-dependent changes in the Trem2R47H variant effect on microglia morphology and function (Fig. 5m-p).

Fig. 5figure 5

Age/disease-dependent impairment of plaque-microglia interaction driven by Trem2R47H. a, b Quantification of cortical microglial morphology of wild-type (WT), Trem2R47H, 5xFAD, and 5xFAD/Trem2R47H revealed (a) increased dendrite length per IBA1+ cell in Trem2R47H compared to WT but (b) decreased average dendrite diameter. c, d Subiculum—representative confocal images from wild-type, Trem2R47H, 5xFAD, and 5xFAD/Trem2R47H mice at (c) 4- and (d) 12-months old stained with Thio-S for dense-core plaques (green), immunolabeled with 6E10 for diffused plaque (blue), and IBA1 for microglia (red). eh Quantification of IBA1+ cell density and average volume in the (e, f) visual cortex and (g, h) subiculum at 4-months of age. In the cortex (e) a sex-dependent increase in microglia number and (f) a decrease in average microglial volume in the presence of Trem2R47H are found. g, h In the subiculum, a decrease in both microglial density (g) and volume (h) is observed in 5xFAD/Trem2R47H compared to 5xFAD. i -l IBA1+ cell density and average volume in the (i, j) visual cortex and (k, l) subiculum at 12-months-old. mn Representative images of Thio-S (green) and IBA1 (red) colocalization in the subiculum at (m) 4-months and (n) 12-months old. o, p Quantification of percent colocalized volume of Thio-S+ and IBA1+ cell normalized to total Thio-S volume per field of view in the subiculum revealed decreased plaque-microglia interaction in 5xFAD/Trem2R47H at (o) 4-months with sex-differences but not at (p) 12-months old. n = 10–12. Data are represented as mean ± SEM. Two-way ANOVA followed by Tukey’s post hoc tests to examine biologically relevant interactions. Statistical significance is denoted by *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

We quantified astrocytes in the visual cortex and subiculum and found that the Trem2R47H variant has minimal effects on the astrocytic response to plaques at both 4- and 12-months (Supplemental Fig. 9).

Trem2 R47H increases neuronal injury in response to plaques

Microglia form a protective barrier around plaques while contributing to their compaction and growth [21]. Given the initial impairment in microglial-plaque interactions in 5xFAD/Trem2R47H mice we next explored the halo of dystrophic neurites that develops around dense-core plaques [52, 53] which can be visualized with lysosomal-associated membrane protein 1 (LAMP1, Fig. 6a, b). Normalization of LAMP1 volume to plaque volume reveals an increase in dystrophic neurites per plaque area in 5xFAD/Trem2R47H mice at 4 months of age (Fig. 6c). Consistent with the restoration of microglial-plaque interactions at 12 months of age, no difference in dystrophic neurites per plaque is seen between 5xFAD/Trem2R47H and 5xFAD mice at this age (Fig. 6d). Twelve-month-old 5xFAD females have more dystrophic neurites than male (Fig. 6d). Interestingly, LAMP1+ dystrophic neurites exhibit a more dissipated morphology at 12-month compared to 4-month, consistent with disrupted axonal transport at later age/disease stages [54].

Fig. 6figure 6

Trem2R47H induces age/disease-dependent dystrophic neurites and axonal damage. a, b Representative confocal images of subiculum in (a) 4- and (b) 12-month-old wild-type, Trem2R47H, 5xFAD, and 5xFAD/Trem2R47H mice stained with Amylo-Glo for dense-core plaques (blue) and immunolabeled for neurofilament light chain (NfL, green) and LAMP1 (red) for dystrophic neurites. c, d Quantification of subiculum LAMP1 volume normalized to Amylo-Glo volume shows increased dystrophic neurites at (c) 4-month but not at (d) 12-month with sex-difference in 5xFAD indicated. eg Measurement of NfL in (e) plasma, (f) soluble fraction reveals consistent increase in NfL level in 5xFAD/Trem2R47H compared to 5xFAD at both 4- and 12-months of age, and (g) cortical insoluble fraction. h, k Representative higher magnification images of immunolabeled NfL spheroids (green) colocalized with LAMP1 (red) in the subiculum of (h) 4-month-old and (k) 12-month-old 5xFAD and 5xFAD/Trem2R47H mice. i, j Reduced density of NfL+ spheroids (i) but no change in spheroid volume (j) in subiculum of 5xFAD/Trem2R47H compared to 5xFAD mice at 4-months of age. l, m No difference in number (l) or volume (m) of NfL+ spheroids between 5 and 5xFAD/Trem2R47H despite a sex-difference in 5xFAD at 12-months of age. n = 10–12. Data are represented as mean ± SEM. Two-way ANOVA followed by Tukey’s post hoc tests to examine biologically relevant interactions. Statistical significance is denoted by *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Neurofilament light chain (NfL) is emerging as a clinically useful plasma biomarker for damage occurring in the brain, including in AD where it tracks with cortical thinning and cognitive decline [55,56,57], while in mouse models of AD it correlates with plaque load [38]. We measured plasma NfL as a surrogate marker of brain damage and found it increased in 5xFAD mice compared to WT mice at 4 months of age, with further increase at 12-months of age, consistent with plaque load (Fig. 6e). The Trem2R47H variant increased plasma NfL in 5xFAD at both ages (Fig. 6e), in line with the hypothesis that TREM2 dysfunction exacerbates neuronal damage. We also measured NfL levels in the detergent soluble and insoluble fractions of microdissected cortices (Fig. 6f, g). NfL in the soluble fraction is not increased in 5xFAD mice compared to WT at either 4 or 12 months of age but is elevated by the presence of Trem2R47H (Fig. 6f). NfL in the insoluble fraction aligns with levels in the plasma, with increases seen in 5xFAD compared to wild-type mice at 4 months and further increases at 12 months (Fig. 6g). As with plasma, NfL levels further trend higher in 5xFAD/Trem2R47H mice. Thus, the presence of Trem2R47H induces changes in NfL, and further exacerbates plaque-induced increases in both detergent insoluble and plasma NfL. To identify the cellular source of NfL, we immunostained for dystrophic neurites (LAMP1) and NfL. Large spherical structures of NfL are seen in the vicinity of plaques and are absent from wild-type and Trem2R47H mice, where staining is observed only in axonal fibers (Fig. 6a, b). Similar bead-like NfL+ spheroids were reported in ischemia-affected human and mouse tissues as sign of axonal damage [58]. Notably, these NfL spheroids colocalize with dystrophic neurites associated with plaques (Fig. 

留言 (0)

沒有登入
gif