Microglia-derived APOE2 improves remyelination even in the presence of endogenous APOE4

Abstract

Demyelination occurs with aging and is exacerbated in neurodegenerative diseases. During demyelination, microglia upregulate expression of APOE, the gene encoding for the brain’s primary lipid transport protein apolipoprotein E (ApoE), which also mediates microglial engulfment and elimination of myelin debris. Compared to the E3 allele of APOE, the E2 allele decreases risk for Alzheimer’s disease (AD), while the E4 allele increases AD risk and is associated with an increased severity and progression of multiple sclerosis. Previous work shows that mice expressing E2 exhibit improved microglial function and remyelination compared to mice expressing E4. However, whether microglial-derived APOE is responsible for driving these differences following demyelination, and if microglia-selective expression of E2 is sufficient to provide protection, is unknown. We sought to determine if microglia-specific replacement of the E4 allele with E2 can rescue myelin loss and promote remyelination, even in the presence of continued E4 expression by other central nervous system (CNS) cells. Using a novel APOE allelic “switch” model in which we can induce a replacement of E4 with E2 exclusively in microglia, we characterize the glial cell response and lipid profile of mice that underwent either lysophosphatidylcholine (LPC) or cuprizone (CPZ)-induced demyelination and subsequent remyelination. We found that although alterations to the brain lipid profile were subtle, microglial E2 replacement significantly improved remyelination, lessened microgliosis, and decreased astrocytic lipid droplet load following CPZ-remyelination. Our results indicate that microglia-specific E2 expression, in the presence of continued E4 expression, may provide protection against myelin loss via both cell-autonomous and non-autonomous immunometabolic mechanisms.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-025-03639-5.

Introduction

Myelin is the lipid-rich axonal insulator important for facilitating action potential conduction and maintaining neuronal integrity within the CNS [1]. Disease states such as Alzheimer’s disease (AD) and multiple sclerosis (MS) exacerbate age-related demyelination, thereby promoting cognitive impairment [2–4]. Demyelination results in the accumulation of myelin debris, composed primarily of cholesterol, phospholipids, and glycolipids [1], which if not cleared effectively is neurotoxic [5].

Microglia, the resident immune cells of the brain, are responsible for maintaining myelin structure and are critical in clearing debris following demyelination to promote remyelination [6–9]. Following demyelination in mice, microglia proliferate and adopt a unique transcriptional state characterized by an increased expression of phagocytic, inflammatory, cholesterol and lipid metabolism genes [6]. One of the genes most heavily upregulated by microglia following demyelination is APOE, a key lipid-regulating gene [6, 10]. APOE codes for the protein apolipoprotein E (ApoE), the primary lipid transport protein in the CNS. Unsurprisingly given its role in lipid metabolism, ApoE plays a critical role in myelination processes [11–14]. Although several cell types synthesize ApoE in the CNS, the contribution of microglial ApoE in de/remyelination is not well understood [15].

Three common polymorphic alleles of APOE exist in humans: E2, E3, E4. Compared to the most commonly expressed E3 allele, carriage of the E4 allele increases risk and progression of both AD and MS [16–22]. Correspondingly, E4 is associated with lower levels of myelin and impaired remyelination, which is attributed to decreased lipid efflux and transport capacity in astrocytes and microglia compared to E3 and E2 [23–28]. Carriage of the E2 allele, however, decreases risk for both AD and MS by 50 and 60%, respectively [17, 22]. Additionally, E2 expression has been shown to mitigate myelin loss and microglial lipid accumulation as well as increase microglial phagocytic function following demyelination compared to E3 and E4 expression in mice [29].

However, it has not been determined if microglia-derived E2 is sufficient to modulate E4-associated deficits in myelination. This is a critical therapeutic knowledge gap as microglial reprogramming and repopulating strategies continue to gain traction for the treatment of neurodegenerative disease, especially in high-risk E4 carriers. Here, we sought to determine if microglia-specific replacement of E4 with E2 can prevent myelin loss and promote remyelination, even in the presence of continued E4 expression from other CNS and peripheral sources. Using a novel APOE “switch” model that allows for the inducible and cell-specific replacement of E4 with E2, along with two models of demyelination, we characterized myelin loss, glial cell activation, and lipid metabolism. We found that while a microglia replacement with E2 did not affect the degree of demyelination, it significantly aided in remyelination, likely via mechanisms related to microglial reactivity and glial-cell lipid accumulation.

Results

Microglia selectively and efficiently transition from E4 to E2

To characterize the in vivo dynamics of microglia-specific APOE expression in the context of demyelination, we employed a novel APOE “switch” model to allow for a conditional, cell-type specific APOE4 to APOE2 allelic switch in microglia. We generated APOE4s2^flox/flox xTmem119-CreER^T2 APOE switch mice (4s2^M) that express APOE4 in all cell types except microglia, which selectively transition to APOE2 expression upon tamoxifen injection (Fig. 1A). Utilizing a 4s2^M strain crossed to an ai9 (tdTomato) Cre-reporter, immunofluorescent analysis of microglia expressing both tdTomato and IBA1 indicate efficient and specific activation of the Tmem119 Cre recombinase in microglia across several regions of the brain (74, 57.7, 61.5, and 77.4% of cells in cortex, hippocampus, corpus callosum, and thalamus, respectively) in mice that received tamoxifen, but not in mice that received vehicle (oil) treatment (Fig. 1B-C; Supplemental Fig. 1). Cell-type specificity and efficiency of the E4 to E2 switch was further confirmed using cell sorting and qPCR. Microglia in which Cre recombinase was active (tdTomato+) showed a clear shift toward APOE2 expression, while other CNS cell types (tdTomato-) (i.e. astrocytes, oligodendrocytes, neurons, etc.) continued to express APOE4 (Fig. 1D).

Fig. 1.

Microglia selectively and efficiently transition from E4 to E2. A, Graphic depicting the design of the tamoxifen inducible microglia specific APOE switch mice (APOE4s2^M) that undergo cell-specific replacement of E4 with E2. APOE4s2^flox/flox mice are crossed to microglia specific Tmem119-CreER^T2 mice to generate a model capable of an inducible, microglia specific APOE4 to APOE2 switch (4s2^M). B, Immunofluorescent staining of IBA1 + microglia and ai9 Cre-reporter + microglia in 4 different brain regions (cortex (CX), hippocampus (HIPP), corpus callosum (CC), and thalamus (THAL)) in a 4s2^M mouse that received tamoxifen. C, Quantification of the percent of cells positive for both tdTomato and IBA1 immunoreactivity in each region in 4s2^M (N = 3) compared to oil treated controls (N = 3); each dot is the average of 6 images per mouse; average manual quantification of 2 blinded individuals; red: vehicle (oil) treated 4s2^M mice, blue: 4s2^M mice injected with tamoxifen to induce allelic switching. D, Allelic discrimination plots depicting a shift from APOE4 to APOE2 mRNA expression in Cre-positive microglia sorted from other Cre-negative cells in the brain expressing APOE4 using an ai9 Cre-reporter. E, Schematic depicting the experimental design. 6-week old mice containing the APOE4s2 flox/flox with and without the Tmem119-CreER^T2 were injected with tamoxifen for 5 consecutive days to allow for allelic switching in Tmem119-CreER^T2 mice. At 8 weeks old, mice received either intracranial LPC injections or began the CPZ diet

To assess if microglial E2 modulates myelin loss in the presence of continued E4 expression, we employed two distinct models of demyelination. For both paradigms, 6-week-old 4s2^M mice received tamoxifen to induce an E4 to E2 switch selectively in microglia, with Cre-negative 4s2- littermates receiving tamoxifen (continued E4 expression in all cell types, including microglia) used as controls. After 2 weeks of recombination, mice underwent one of two demyelination paradigms – lysophosphatidylcholine (LPC) injections or cuprizone (CPZ) diet [30] (Fig. 1E).

Replacement of E4 with E2 in microglia decreases astrogliosis but not myelin debris or microgliosis following LPC-induced demyelination

To understand how microglial E2 expression may mediate the response to focal demyelination, we utilized the LPC model of demyelination [31–34]. 8-week-old 4s2- and 4s2^M mice were intracranially injected with LPC at two locations (anterior and posterior) targeting the corpus callosum (CC) within the ipsilateral hemisphere. The contralateral hemisphere served as an internal control for each mouse (Fig. 2A). Ten days after LPC injections, whole brains were collected for immunofluorescent analysis of gliosis and degraded myelin basic protein (dMBP). As expected, LPC injections into the corpus callosum of 4s2^M and 4s2- mice resulted in the accumulation of myelin debris (Fig. 2B-D), GFAP + reactive astrocytes (Fig. 2E-G), IBA1+ (Fig. 2H-M), and CD68+ (phago-lysosomal marker of activated phagocytic microglia [35, 36]) microglia in the ipsilateral corpus callosum compared to the contralateral corpus callosum. Anterior corpus callosum LPC injection analyses mirrored the results from the posterior CC (Supplemental Fig. 2).

Fig. 2.

Microglial replacement with E2 decreases astrogliosis following LPC-induced demyelination. A, Graphic showing 2-point LPC intracranial injections targeting the CC in 4s2- (n = 7) and 4s2^M (n = 7) mice. Tamoxifen was administered at 6 weeks to induce allelic switching, LPC injections were performed at 8 weeks, and tissue was collected 10 days post-LPC injections. B-C, Representative images of dMBP staining in the whole brain (left hemi: contralateral, right hemi: ipsilateral) (B) and in the CC (C) of LPC injected 4s2- and 4s2^M mice. D, Percent area of dMBP + coverage in the CC of 4s2- and 4s2^M mice (D). E-F, Representative images of GFAP staining in the whole brain (E) and in the CC of LPC injected 4s2- and 4s2^M mice (F). G, Quantification of percent area coverage of GFAP in the CC of 4s2- and 4s2^M mice. H-I, Whole brain representative image of IBA1 staining in an LPC-injected mouse brain (H). IBA1 staining in the CC of 4s2- and 4s2^M mice (I). J, Quantification of percent area coverage of IBA1 in the CC of 4s2- and 4s2^M mice.) K-L, Whole brain representative image of CD68 staining in an LPC-injected mouse brain (K). CD68 staining in the CC of 4s2- and 4s2^M mice (L). M, Quantification of percent area coverage of CD68 in the CC of 4s2- and 4s2^M mice. D, G, J, M, Circles (male), squares (female). (2way ANOVA with Fisher’s LSD multiple comparisons; *p < 0.05; ***p < 0.0005; ****p < 0.0001). Anterior LPC-injection quantification can be found in Supplemental Fig. 2

Interestingly, selective expression of E2 in microglia did not affect the accumulation of either dMBP, CD68 + or IBA1 + microglia within the CC, but did decrease GFAP + reactive astrocyte area compared to 4s2- controls in the ipsilateral corpus callosum (Fig. 2G). This suggests that while selective microglial expression of E2 was not sufficient to mitigate myelin debris accumulation or mediate microgliosis/phagocytic activity, microglial E2 expression does decrease astrogliosis following focal demyelination.

Microglial replacement of E4 with E2 drives increased remyelination

To determine the effects of microglial E2 expression in a global model of demyelination, 4s2- and 4s2^M mice were placed on either a cuprizone (CPZ) or a standard chow control diet for 5 weeks. To assess processes related to remyelination, a separate cohort of mice received CPZ for 5 weeks and were then placed back on standard chow for 1 additional week (Fig. 3A) [30]. All mice maintained similar body weights regardless of treatment group or genotype (Fig. 3B).

Fig. 3.

Microglial replacement of E4 with E2 drives increased remyelination. A, Timeline of cuprizone-mediated demyelination and remyelination following microglial-specific allelic switching induced at 6 weeks. At 8 weeks, mice received CPZ diet (normal chow for controls) for 5 weeks before tissue collection (Demyelination/DM) or for 5 weeks before returning to Chow diet for 1 additional week before tissue collection (Remyelination/RM). B, Normalized weekly body weight (normalized to Control) for DM and RM mice of each genotype throughout the duration of the experiment. C-F, Fluorescent staining using FluoroMyelin (C) and dMBP (E) in the CC of 4s2- and 4s2M mice from Control, DM, and RM paradigms. Quantification of the percent area coverage of FluoroMyelin (D) and dMBP (F) in the CC of 4s2- and 4s2M mice in each treatment group. Circles (male), squares (female). A significant Genotype x Treatment interaction effect was observed (p < 0.0091) for FluoroMyelin quantification. (*p < 0.05; ***p < 0.0005; ****p < 0.0001)

FluoroMyelin is a lipophilic dye that preferentially binds to lipids found in the myelin sheath [37]. Percent of FluoroMyelin area coverage in the CC was measured to determine if microglial E2 expression limits myelin loss during demyelination and/or induces myelin repair in the remyelination period (Fig. 3C-D). CPZ induced a similar degree of demyelination, as measured by lower FluoroMyelin percent area coverage, in both 4s2- and 4s2^M mice. Following one week of remyelination, the FluoroMyelin area in 4s2- mice, whose microglia continued to express E4, decreased even further. Interestingly, in 4s2^M mice, whose microglia express E2, this decrease was not observed (Fig. 3D). These results suggest that the presence of microglial E4 promotes further myelin loss throughout the remyelination period, while replacing E4 microglia with E2 allows for myelin repair (or synthesis) during remyelination.

Degraded Myelin Basic Protein (dMBP) within the corpus callosum was measured to quantify the amount of myelin debris in the CC following de/remyelination in 4s2- and 4s2^M (Fig. 3E-F). As expected, control mice had the lowest area coverage of dMBP, with no significant difference between 4s2- and 4s2^M. dMBP + area was significantly, and similarly, increased following CPZ demyelination regardless of microglial APOE genotype. Following remyelination, dMBP + area was not significantly reduced in either 4s2- or 4s2^M compared to demyelination (Fig. 3F). Together, these results indicate that – relative to E4 – microglial E2 expression promotes restored myelin even while dMBP + fragments remain similarly detectable after a short remyelination period.

Replacing E4 with E2 in microglia decreases microgliosis following a short remyelination period

To investigate potential cell-specific contributions to myelination processes in E4 mice with either E4 or E2 expressing microglia, we examined the glial cell response following the de/remyelination periods. First, we measured the microglial response to de/remyelination by quantifying IBA1 and CD68 via immunofluorescence and Trem2 via qPCR. Total IBA1+, CD68+, and CD68 + IBA1 + area similarly increased in the CC following demyelination for both 4s2- and 4s2^M mice (Fig. 4A-E). Following the remyelination period, CD68 + area and CD68 + IBA1 + microglia significantly and similarly decreased for both 4s2- and 4s2^M compared to demyelination (Fig. 4D, E). Interestingly, the decrease in IBA1 + area following the remyelination period was significantly lower in 4s2^M mice compared to the 4s2- controls (Fig. 4B). These results were localized to the CC, as no difference was observed in microgliosis in the hippocampus or cortex between 4s2- and 4s2^M (Supplemental Fig. 3 A, C, E). As expected, Trem2 levels increased during demyelination in both genotypes; however, while the remyelination period resulted in decreases in Trem2 expression in both groups, this lowering only reached significance in the 4s2^M mice (Fig. 4F). In sum, these results demonstrate that microglial E2 expression lowers IBA1 + microglial reactivity within the CC without affecting the phago-lysosomal CD68 + microglial population or Trem2 expression following remyelination.

Fig. 4.

Microglial E2 replacement decreases microgliosis following a short remyelination period. A-H, Immunofluorescent staining of IBA1 (A), CD68 (C), and GFAP (G) in the hippocampus and CC of 4s2- and 4s2^M mice in each treatment group. Quantification of percent area coverage of IBA1 (B), CD68 (D), CD68 + IBA1+ (E), and GFAP (H), in the CC of 4s2- and 4s2^M mice in each treatment group. Regional quantification can be found in Supplemental Fig. 4. Log2-transformed Trem2 fold change in the brains of 4s2- and 4s2^M mice from each treatment group (F). Circles (male), squares (female). (2way ANOVA with Tukey’s multiple comparisons; *p < 0.05; ***p < 0.0005; ****p < 0.0001)

Next, we wanted to know if selective replacement of microglial APOE could promote downstream changes in other glial cell populations. As expected, reactive GFAP + astrocyte area increased in both the de/remyelination groups relative to the control diet, but did not significantly differ between 4s2- and 4s2^M mice (Fig. 4G-H; See Supplemental Fig. 3D, F).

Microglia expression of E2 subtly alters oligodendrocyte lineage marker expression

Oligodendrocytes, the glial cell responsible for myelination during development and myelin regeneration, arise from oligodendrocyte progenitor cells (OPCs). OPC differentiation into mature, myelinating oligodendrocytes is essential for myelin formation [38, 39]. To determine if expression of microglial E2 is sufficient to alter oligodendrocyte differentiation and mature oligodendrocyte populations, potentially lending to the increased myelin repair seen in 4s2^M RM mice, we stained for Olig2 (all oligodendrocyte cell lineage marker), PDGFRα (OPC marker), and CC1 (mature myelinating oligodendrocyte marker) in the CC (Fig. 5A-B). The percentage of Olig2 + cells increased following demyelination and remyelination compared to the control diet, with no difference between 4s2- and 4s2^M (Fig. 5C). Similarly, OPCs (PDGFRα+) and mature (CC1+) oligodendrocytes increased in both 4s2- and 4s2^M mice following demyelination. However, following a remyelination period, OPCs (PDGFRα+) and mature (CC1+) oligodendrocytes were only significantly decreased in 4s2^M, not 4s2- mice (Fig. 5D-E). These results demonstrate that oligodendrocyte populations may be modulated by microglial E2 expression following remyelination, but not during demyelination.

Fig. 5.

Microglia expression of E2 subtly alters oligodendrocyte lineage marker expression. A-B, Representative images of immunofluorescent staining of Olig2, PDGFRα, and CC1 from the CC of 4s2- and 4s2^M following Control, DM, RM. C-E, Quantification of percent area coverage of Olig2 (C), PDGFRα (D), and CC1 (E) in the CC of 4s2- and 4s2^M mice. Circles (male), squares (female). (2way ANOVA with Tukey’s multiple comparisons; *p < 0.05; ***p < 0.0005; ****p < 0.0001)

Microglial E2 replacement has a modest effect on the E4 brain lipidome

The major lipid constituents of myelin are cholesterol, phospholipids, and glycolipids [1]. During demyelination, myelin-derived lipids are released into the extracellular space and microglia mediate clearance of these lipids to allow for remyelination [15, 40]. To investigate if E2 or E4 microglial APOE expression affects the lipidome during a de/remyelination challenge, we performed a targeted lipidomic analysis (401 identified lipids) on brain tissue from 4s2- and 4s2^M mice following the de/remyelination paradigms. To visualize the overall variation in lipidomic profiles across our experimental groups, we first performed a principal component analysis (PCA), which reveals a clear separation of the de/remyelination groups relative to controls, but minimal separation based on genotype (Fig. 6A). Heatmap analysis of the top 25 most differentially regulated lipids – the majority of which were ceramides and cholesterol esters – displays lipid abundance patterns across the six experimental conditions (2 genotypes × 3 treatments). Samples were grouped by experimental condition to enable direct comparison of lipid profiles between genotypes and de/remyelination treatments (Fig. 6B). Volcano plots illustrating differences between 4s2- and 4s2^M brains showed that E2 microglia-specific replacement promoted an upregulation of LPC(20:2) and AC(14:0)-OH following demyelination and PC(O-36:0) and LPC(18:1) after remyelination (Fig. 6C-E). Together, these data suggest that microglial E2 expression subtly alters the lipidome of an E4 brain following de/remyelination.

Fig. 6.

Microglial E2 replacement has a modest effect on the E4 brain lipidome following de/remyelination. A, PCA plot including all identified lipids depicting separation between Control and DM/RM treatment groups. B, Heatmap displaying lipid abundance profiles grouped into six clusters corresponding to the experimental conditions (2 genotypes × 3 treatments). Clustering was based on predefined experimental grouping rather than unsupervised data-driven analysis. Colors indicate relative (log₂-scaled) lipid abundances. C-E, Volcano plots for lipid species differentially expressed between 4s2- and 4s2^M following Control (C), DM (D), and RM (E) treatments. (2way ANOVA with FDR-corrected multiple comparisons (Benjamini-Hochberg); *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001)

Microglial APOE genotype modulates ApoE protein expression and lipid droplet distribution

To assess if lipid processing during de/remyelination was differentially affected following a microglial allelic switch, we stained for lipid droplet surface protein Plin2 within microglia and astrocytes (Fig. 7A-B). While both 4s2- and 4s2^M showed an overall increase in the percentage of Plin2 + staining following demyelination, the effect was only significant in 4s2^M mice (Fig. 7C). Microglial Plin2 immunostaining (Plin2 + IBA1+) increased substantially following the demyelination period but did not differ significantly between the 4s2- or 4s2^M groups regardless of treatment (Fig. 7D). Interestingly, 4s2^M mice had significantly less astrocytic Plin2 (Plin2 + GFAP+) following remyelination compared to 4s2-, potentially suggesting an indirect effect of microglia-derived E2 in promoting lipid efflux from astrocytes (Fig. 7E).

Fig. 7.

Microglial APOE genotype modulates ApoE protein expression and lipid droplet distribution. A-B, Representative images from the CC of immunofluorescent staining of IBA1 and Plin2 (A) and GFAP and Plin2 (B). C-E, Quantification of percent area coverage of Plin2 (C), Plin2 + IBA1+ (D), and Plin2 + GFAP+ (E) within the CC of 4s2- and 4s2^M mice. F, ApoE immunofluorescent staining in the CC of 4s2- and 4s2^M mice. G, Quantification of percent area coverage of ApoE in the CC of 4s2- and 4s2^M mice following de/remyelination. H-I, Representative images from the CC of immunofluorescent staining of IBA1 and ApoE (H), and GFAP and ApoE (I). J-K, Quantification of percent area coverage of microglial ApoE (APOE + IBA1+) (J) and astrocytic ApoE (APOE + GFAP+) (K) within the CC of 4s2- and 4s2^M mice. Circles (male), squares (female). A significant Genotype x Treatment interaction (p < 0.0276) was observed for APOE + IBA1 + quantification. (2way ANOVA with Tukey’s multiple comparisons; *p < 0.05; ***p < 0.0005; ****p < 0.0001)

Next, we asked whether the APOE genotype of microglia would alter ApoE protein expression in that cell type – or indirectly in astrocytes (the primary producer of ApoE in the brain) – following de/remyelination. There was no difference in total ApoE across treatment or genotype outside of the white matter tracts (i.e. hippocampus or cortex, Supplemental Fig. 4). In contrast, total ApoE within the CC was significantly elevated during both the de/remyelination periods relative to control, regardless of genotype (Fig. 7F, G). These changes appeared to be driven – at least in part – by increases in microglial co-localized ApoE. Interestingly, in mice expressing E4 in all cell types, microglia ApoE expression (APOE + IBA1+) significantly increased during demyelination and then slightly declined following remyelination (Fig. 7H, J); however, there was no change across the de/remyelination paradigms in mice whose microglia express E2, as they showed sustained elevation of ApoE relative to 4s2- mice on the control diet (Fig. 7J). Immunofluorescent staining for APOE + GFAP + cells within the CC showed no changes in astrocytic ApoE abundance across the de/remyelination paradigms (Fig. 7I, K).

Finally, to understand if the changes we observed in 4s2^M mice correlate to the amount of E2 expression following microglial allelic switching, we performed RT-PCR to measure the relative E2 expression in the whole brain (that continues to express E4 by other cells) (Supplemental Fig. 5). We found that in normal chow-fed mice, microglial E2 replacement had a negligible effect on total E2 expression in the whole brain. However, following de/remyelination, there was a modest shift towards E2 expression in the whole brain of 4s2^M mice that was not observed in 4s2- controls (Supplemental Fig. 5) [41–44]. Despite this increase in E2 expression, we found that there was no significant correlation between total E2 expression and our aforementioned measures. This suggests that functional differences between the protein isoforms themselves (ex. lipidation, receptor binding, etc.), rather than the amount of APOE expression, may be driving the observed differences.

Discussion

In the current study, we sought to determine if the microglia-derived pool of ApoE is responsible for driving previously reported deficits in myelination and related glial processes in E4 versus E2 brains. By leveraging our novel mouse model in which we selectively transition microglia from expression of E4 to expression of E2 prior to de/remyelination, we specifically investigated how microglial ApoE affects myelin loss, the glial cell response, and lipid metabolism. We hypothesized that selective expression of microglial E2 would rescue the lipid metabolism dysregulation and glial reactivity previously observed in E4 microglia, potentially mitigating myelin loss during demyelination paradigms and augmenting myelin repair during remyelination. Our findings reveal that microglial E4 to E2 replacement attenuates astrogliosis following LPC-induced demyelination, while in the CPZ model it enhances remyelination and reduces both microgliosis and astrocytic lipid droplet accumulation. In the CPZ model, these results were accompanied by modest changes in the brain lipidome and appear to act independently of relative E2 expression by microglia.

It has been well established that myelin is lost with both aging and disease, specifically in AD and MS [4, 26, 27, 45–49]. Interestingly, APOE genotype plays a significant role in modulating risk and progression of both of these neurodegenerative disease states. Individuals with E4 have up to a 15-fold increased risk for AD and display accelerated disease progression in MS [16, 19, 50]. In contrast, individuals with the E2 allele are protected against AD and may have a lower risk for MS [20, 22, 23, 28, 51–53]. E4 carriers also exhibit lower levels of myelin compared to E3 or E2 carriers in both cognitively normal and AD individuals [53].

Microglia, the primary immune and phagocytic cells of the brain, upregulate expression of APOE in disease states, which is critical for engulfment and elimination of myelin debris [6, 10, 15]. E4 expression has been shown to impair microglia lipid metabolism and phagocytic capabilities in demyelination compared to E3 and E2 expression [29] but whether this impairment is a result of microglial ApoE remains unknown. Utilizing our APOE “switch” mouse model, we can selectively and efficiently switch from E4 to E2 expression in microglia while all other remaining APOE-expressing cells continue to express E4 in the CNS and periphery. This novel approach is highly relevant given the emergence of myeloid/microglial replacement therapies which seek to replace dysfunctional microglia to restore immune homeostasis and mitigate neuropathological features [54–56]. As these therapies continue to advance, understanding the contribution of microglial APOE isoforms in disease will offer mechanistic insight into more targeted approaches.

Following an early-life allelic switch at 6-weeks, mice underwent LPC-induced demyelination to investigate if microglial E2 expression can mediate effects on myelination and glial cell reactivity. LPC induces demyelination through oligodendrocyte cell membrane permeabilization resulting in oligodendrocyte cell death and subsequent loss of myelin [34]. As expected, myelin debris and reactive microglia accumulated within the corpus callosum of the ipsilateral hemisphere of both 4s2- and 4s2^M mice following LPC-mediated demyelination. E2 expression from microglia alone did not appear sufficient to modulate myelin debris clearance following a 10 d.p.i demyelination timepoint. However, the presence of E2 microglia did result in a decrease in astrogliosis in the ipsilateral CC following LPC. This suggests that E2 microglia can have non-cell-autonomous effects after demyelination via LPC.

Because the LPC model concomitantly induces an injury at the site of injection and only allows us to examine focal demyelination, we chose to also test a less-invasive, global model of de/remyelination – the cuprizone (CPZ) diet. CPZ, a copper chelator, induces demyelination through oligodendrocyte-specific cell death [57, 58]. We found that following a remyelination period, mice with E2-expressing microglia had significantly greater myelin area, even though dMBP accumulation did not differ, compared to mice expressing only E4. dMBP has been shown to be increased in E4 individuals with AD and in mice challenged with CPZ-induced demyelination compared to E3 and E2 [29, 53]. Our data suggest that the differences in dMBP observed in these studies may not be driven by microglial ApoE since microglial replacement with E2 did not mitigate dMBP accumulation following CPZ-induced demyelination or subsequent remyelination. Following remyelination, E2-expressing mice have been shown to have the greatest restoration of MBP compared to E3 and E4 mice [29]. This aligns with the increase we see in FluoroMyelin area after switching the microglia from E4 to E2 expression. This indicates that microglial E2 expression may aid in remyelination despite continued E4 expression by other CNS cells.

Since microglia are responsible for myelin debris clearance, we aimed to assess how allelic switching from E4 to E2 exclusively in microglia altered the microglial response to de/remyelination. Microglial replacement of E2 did not affect CD68 immunostaining but did result in lower IBA1 + area in the CC following remyelination. In contrast, Wang et. al. found that following demyelination, E4 mice had decreased microgliosis and less CD68 + microglia compared to both E3 and E2 [29]. Additionally, E4 microglia have reduced phagocytic activity compared to E3 and E2 [59, 60]. These studies indicate a blunted phago-lysosomal response of E4 microglia compared to E3 and E2. Here, we show that while microglial E2 may lower the overall reactivity of microglia following remyelination, it was not sufficient to alter the microglial phago-lysosomal burden, as measured by CD68 immunostaining, following de/remyelination.

To investigate potential non-cell-autonomous effects of microglial APOE switching in de/remyelination, we assessed astrogliosis and oligodendrocyte populations. Similar to Wang et.al., who reported that astrogliosis was not affected by APOE genotype following demyelination, we found that microglial expression of E2 did not change the reactive astrocyte response in either demyelination or remyelination. Our LPC demyelination paradigm, however, demonstrated a slight decrease in astrogliosis in 4s2^M mice. This suggests that microglial E2 modulates the astrocytic response differentially following LPC-induced demyelination but not CPZ-mediated de/remyelination. This differential response could be due to the injection-related injury (stereotaxic needle) and concomitant blood-brain-barrier/vascular disruption and glial cell activation induced by the LPC model; however, the mechanisms underlying this response require further investigation. While less is known about how microglial E2 may impact astrocytes, it has been shown that selective removal of E4 from microglia in an amyloid overexpression mouse model (APP/PS1) increases astrogliosis compared to APP/PS1 mice that globally express E4 [61]. Conversely, another study using the APP/PS1 model found that microglia-specific expression of E4 increased astrogliosis compared to microglia-specific expression of E3 [60]. Further, removal of astrocytic E4 promotes more homeostatic astrocyte and microglial signatures in amyloid and tau mouse models [62, 63]. These results, in combination with our current findings, suggest a dynamic relationship between microglial APOE expression and astrocyte reactivity in different disease/injury conditions.

Oligodendrocytes are the myelinating cells of the CNS and differentiation of OPCs into mature myelinating oligodendrocytes is critical for remyelination [38, 39]. Mature CC1 + oligodendrocytes and OPCs have been shown to initially decrease upon CPZ diet followed by an increase after 5 weeks of CPZ diet [58]. Compared to E3, aged E4 mice were reported to have fewer Olig2 + cells, no difference in CC1 + mature oligodendrocytes, and a greater abundance of Nkx2.2 + OPCs [51]. We found that there was only a significant decrease in mature (CC1+) oligodendrocytes and OPCs (PDGFRα+) after a remyelination period in mice that express E2 in their microglia (compared to 4s2^M DM). This decrease was not observed in 4s2- mice suggesting that E2 microglia may promote oligodendrocyte turnover or differentiation following remyelination compared to mice expressing E4 in all cell types. This modulation of oligodendrocyte populations by E2 microglial replacement may somehow promote more efficient myelinating oligodendrocytes and potentially explain the increased remyelination (via FluoroMyelin staining) we observe in 4s2^M mice.

Since ApoE is the primary lipid transporter in the brain, we suspected that the changes we observed may be a result of differential lipid transport between E4 and microglial E2 expression. Studies have profiled the lipidomic alterations between ApoE isoforms in humans and mice, revealing several lipid species that are differentially regulated in E4s in a region-dependent manner compared to E3s and E2s [23, 52, 64, 65]. Additionally, microglia from E4 mice demonstrate a significantly different lipid profile compared to microglia from E3 mice, indicating that microglial APOE expression is sufficient to alter the lipidomic profile [61]. In our current study, targeted lipidomic analysis indicated that treatment was the predominant factor shaping lipid profiles in 4s2⁻ and 4s2^M mice, consistent with lipid remodeling that accompanies de/remyelination [66, 67]. In line with this, our lipidomic results parallel prior reports of myelin-associated lipid class changes during demyelination, including alterations in ceramides, sphingomyelins, cholesterol esters, and triglycerides [67–71]. Notably, several ceramide species were increased in whole-brain homogenates after demyelination, likely reflecting accumulation of myelin-derived debris and lipid storage within glia rather than new myelin synthesis. Such elevations are consistent with microglial lipid droplet formation and impaired lipid efflux during myelin clearance, processes that precede remyelination and may influence the efficiency of repair [68, 72]. Together, these findings suggest that demyelination-driven lipid remodeling dominates the global brain lipid landscape, and that microglia-specific E2 expression does not substantially alter the overall lipid profile in the presence of continued E4 expression by other CNS cell types.

In aging and disease, lipid metabolism and lipid accumulation are dysregulated in microglia and contribute to an impaired phagocytic state [73–75]. ApoE is essential for the recycling and transport of lipids in the brain, a process crucial for myelin repair and maintenance [15]. E4 microglia have been shown to accumulate lipids, promoting a pro-inflammatory and metabolically dysfunctional state compared to E3 and E2 [25, 73, 74]. To investigate if transitioning from E4 to E2 expression in microglia is sufficient to alter lipid accumulation following de/remyelination, we quantified ApoE and Plin2 in both astrocytes and microglia. We found that microglial ApoE in E2 expressing microglia was higher at baseline (control-fed mice) compared to E4 microglia. This corresponds with previous literature showing an ApoE isoform dependence on ApoE concentration in the brain (E2 >E3 >E4) [41, 76, 77]. Interestingly, there was a decrease in microglial ApoE and Plin2 in 4s2- mice following remyelination that was not observed in 4s2^M mice, suggesting that E4 microglia rapidly turn over accumulated ApoE/Plin2, while E2 microglia retain or produce more. Sustained E2 expression may result from greater synthesis or reduced clearance of more highly lipidated E2 lipoproteins, in contrast to the poorly lipidated, aggregation-prone E4. This sustained E2 microglial expression could potentially enhance microglial lipid handling (i.e. myelin debris clearance) while lack of E4 accumulation could indicate impaired lipid processing. Astrocytic Plin2 was decreased in mice with E2 microglial expression following remyelination. Previous literature in mouse models of Alzheimer’s disease has found that microglia-specific E4 expression increases microglial Plin2 accumulation compared to microglia-specific E3 expression and that deletion of microglial E4 mitigates this accumulation [60, 61]. Wang et. al. found that microglial Plin2 accumulation was significantly increased in E4 mice compared to E3 and E2 following demyelination [29]. Our data indicate that microglial E2 expression mediates ApoE and Plin2 accumulation differently during remyelination, potentially through retention or increased production of ApoE/Plin2. Moreover, microglial E2 may reduce astrocytic Plin2 expression through direct and/or indirect lipid-signaling mechanisms, such as E2-containing lipoproteins engaging astrocytic LDLR/LRP1 receptors to activate LXR/PPARγ pathways that suppress lipid droplet formation, or via E2-dependent changes in microglial lipid metabolism – possibly involving TREM2 signaling, oxysterols, or extracellular vesicles – that influence astrocytic lipid homeostasis [66, 78–84]. Together, these data suggest that microglial E2 modulates both microglial and astrocytic lipid handling during remyelination.

While our findings provide important insights, there are limitations that if addressed in future experiments would help elucidate how microglial E2 expression can aid in myelin repair via non-cell and cell-autonomous mechanisms. The LPC model of demyelination induces an injury (via intracranial stereotaxic injection) that may have introduced additional variables (i.e. increased glial cell reactivity). Future studies looking at the remyelination response following LPC would bolster the findings from both our LPC demyelination and CPZ remyelination paradigms. In the CPZ model, we allowed 1 week for remyelination to capture an early/active remyelination period [30]. However, it is possible that a longer remyelination period would allow us to see more differences between 4s2- and 4s2^M. Our 5-week demyelination timeframe was chosen because this is the period of peak gliosis and myelin loss [30, 57]; however, more subtle/nuanced differences in gliosis and myelin loss may be observed at different timepoints. While we did not perform a priori power analyses, our post-hoc detectable-effect calculations indicate that the study was well powered to detect large biologically meaningful effects in the microglial response. Thus, although we interpret null genotype effects cautiously due to limited power for modest between-genotype differences, the absence of a detectable phenotype suggests that any ApoE-dependent contribution to our measures in these paradigms is likely to be modest. Future studies with larger cohorts will be valuable to further evaluate subtle genotype effects. Additionally, rapid repopulation/proliferation of microglia occurs following CPZ/LPC. While this repopulation is thought to largely be driven by resident microglia [85–88], we cannot rule out that microglial repopulation from non-resident – and/or non-recombined - cells could re-introduce E4 expression in microglia (of 4s2^M mice). Finally, we chose to investigate the effects of microglial E2 expression because of the protective effects of E2 in AD and MS [17, 19, 21, 22, 50]. However, it remains to be seen how selective expression of E3 or other risk-protective variants (Christchurch, Jacksonville) could modulate de/remyelination [89].

While microglial replacement with E2 appears to have an overall mild effect following de/remyelination, our current findings demonstrate that selective replacement of E4 with E2 in microglia promotes remyelination, reduces microgliosis, and alters glial-cell lipid droplet load, even in the presence of continued E4 expression in the CNS. These results suggest that microglia-derived E2 can exert both cell-autonomous and non-cell-autonomous protective effects during demyelination and repair. This work highlights microglia-specific APOE modulation as a potential therapeutic strategy for mitigating myelin loss in neurodegenerative disease.

Materials and methods

Animals

APOE switch mice (APOE4s2) were generated (on a C57BL/6 N background) at TransViragen Inc (Chapel Hill, NC). Mouse Apoe exons 2–4 were replaced by human APOE exons 2–4. The coding region of APOE (exon 4) was followed by a stop cassette (3x SV40 polyA), and flanked with loxP. Following the coding region for APOE4 was an additional exon 4 containing the sequence for APOE2. Tmem119-CreER^T2 mice were purchased from Jackson Laboratories (Strain #:031820, RRID: IMSR_JAX:031820) and crossed with homozygous APOE4s2^flox/flox mice to generate Tmem119^Cre/+/APOE4s2^flox/flox (4s2^M). 4s2^M mice were then crossed with an Ai9 Cre-reporter strain to induce fluorescent tdTomato expression upon tamoxifen administration. Cre-negative littermates (4s2-) were employed as controls across all experiments.

Genotyping and gene expression analysis

Animals were genotyped at 21 days for the APOE switch (4s2) transgene using PCR. Animals were bred to homozygosity for the 4s2 allele. Primers specific for the humanized target allele were used to identify whether animals were homozygous for 4s2 allele. Primers for the native and target alleles are listed as follows: WT-1 forward primer sequence (5’AGGGGTTACCTCCAGGAAAGG), WT-1 reverse primer sequence (5’TGGATCCGCTGCCAAAA), WT-2 forward primer sequence (5’CAAAGGCTGGAATTACAGG), WT-2 reverse primer (5’TTGGGCTCCATGAGCTCTGG), Tg-1 forward pimer sequence (5’AGGGGTTACCTCCAGGAAGG), Tg-1 reverse primer sequence (5’GCCGTGTTCCATTTATGAGC), Tg-2 forward primer sequence (5’TTCTCCCCGGCCTGCTTGTT), and Tg-2 reverse primer sequence (5’TTGGGCTCCATGAGCTCTGG). Mice were bred heterozygous for Tmem119-CreER^T2 and were genotyped for creER^T2 by a commercial vendor using real-time PCR (Transnetyx, Cordova, TN).

To determine APOE allele expression and cell-type specificity after tamoxifen, RNA was extracted from sorted cells from brain tissue using the RNeasy Micro Kit (Qiagen, cat. #74004) and converted to cDNA (Thermo Fisher Scientific, cat. #4388950). APOE genotype was determined using PCR with TaqMan assay primers for the two allele-determining SNPs of APOE: rs7412 and rs429358 (Thermo Fisher Scientific, cat. #4351379). Positive human DNA controls for APOE genotypes E2/E2 and E4/E4 were included with each assay. Allelic discrimination plots were generated to show the shift in APOE allele expression.

Tamoxifen administration

Tamoxifen (Sigma-Aldrich cat. #T5648) was dissolved overnight at 37 C with gentle rocking in sterile sunflower seed oil (Sigma-Aldrich cat. #S5007) at a concentration of 20 mg/mL and protected from light. Tamoxifen was administered to 6-week-old mice via intraperitoneal injection once every 24 h for 5 consecutive days. Injection dose of 75 mg/kg was determined by animal bodyweight. Injection site was sanitized with an alcohol wipe prior to injection. A minimum of six weeks was given between final injection and analysis.

LPC intracranial injections and tissue collection

Lysophosphatidylcholine (LPC, Sigma-Aldrich #L4129) was diluted to a concentration of 1% in 0.9% NaCl and loaded into a 10ul syringe (Hamilton #7653-01) connected to a 30G needle (Hamilton #7762-03). Mice were anesthetized in an induction chamber with 3% isoflurane and 100% oxygen at a rate of 0.3 L/min and maintained with 1.5–2% isoflurane through a nose cone. Prior to placing the mice in the stereotaxic frame, their heads were shaved, and lubricating eye cream was applied. Mice were securely fixed in the stereotaxic apparatus using ear bars and the skull was exposed by cutting a small incision along the midline of the scalp. Once properly positioned, coordinates were marked, and two small holes were created in the skull. Each injection site (1 mm lateral, 1.1 mm anterior, 1.3 mm depth; 1.5 mm lateral, 0.3 mm anterior, 2.0 mm depth) received 2uL of LPC at a rate of 0.4uL/min. The needle was kept in the site for an additional 3 min then slowly removed. Afterwards, the skin was sutured, and mice were euthanized 10 days post-injection. Blood was collected via retro-orbital injections and transcardial perfusions with paraformaldehyde (PFA) followed by sterile saline were performed. Whole brain was collected and fixed in 4% PFA for 24 h then stored in 30% sucrose in PBS.

Cuprizone diet

0.3% Cuprizone diet (Envigo #TD.140805) was given to 8-week-old mice in their cages ad libitum for 5 weeks for both demyelination and remyelination treatment groups. Mice in the remyelination group were placed back on standard chow diet for one additional week prior to euthanasia. Diet was replaced weekly. Animals were weighed weekly; mice that experienced a weight loss of 20% from baseline were removed from the study.

Brain extraction and preparation – CPZ paradigm

For all CPZ immunofluorescent staining experiments, mice were transcardially perfused with sterile saline and brains were extracted. Brains were cut down the mid-line to separate into hemispheres. The right hemisphere of the brain was post-fixed in 4% PFA (Thermo Fisher Scientific, cat. #047349.9 M) for 24 h at 4˚C then stored in a 30% Sucrose solution with 0.05% Sodium Azide (VWR, cat. #76102-628). Coronal cryosections (30 μm) were obtained using a Leica Cryostat (CM1520) and serially collected in a cryoprotectant solution (50% 0.1 M PB, 30% Ethylene glycol, 20% glycerol). Sections were stored in cryoprotectant solution at −20 °C until use.

Sectioning LPC brains

Whole brains were sectioned anterior to posterior across the coronal plane via cryostat at (−18 C) at 30 μm thickness. Prior to sectioning, a small hole was gently made in the thalamus of the left hemisphere (contralateral to injected hemisphere) for proper identification of the ipsilateral hemisphere during staining. Free floating sections were placed in cryoprotectant solution and kept at −20 C until staining.

Immunofluorescent staining

Hippocampal sections or anterior sections containing the corpus callosum were selected for all stains, which were performed free-floating. For immunofluorescent detection of ApoE (Cell Signaling Technologies, cat. #13366S, 1:500), GFAP (Invitrogen, cat. #13–0300,1:500), IBA1 (Novus, cat. 3NB100-1028, 1:500; Abcam, cat. #ab178846, 1:500), CD68 (Biorad cat. MCA1957; 1:250), dMBP (Milipore Sigma cat. #ab5864; 1:2000), PLIN2 (Progen cat. GP42; 1:500), OLIG2 (Abcam, cat. #ab109186; 1:500), CC1 (Abcam, cat #ab16794; 1:500), PDGFRa (R&D Systems, cat. #AF1062; 1:500), free-floating sections were washed three times for 10 min in 1xPBS and permeabilized in 1xPBS + 0.2% Triton-X for 10 min. Sections were blocked in 10% Goat Serum in 1xPBS + 0.2% Triton-X or 10% Donkey Serum in 1xPBS + 0.2% Triton-X (ApoE, GFAP, IBA1) for 1 h at room temperature. Primary antibody cocktails were diluted in 3% Goat or Donkey serum and sections were incubated overnight at 4˚C. Sections were washed three times for 10 min in 1xPBS before applying corresponding secondary antibodies (Goat anti-Rabbit AF647, Abcam, cat. #ab150079; Goat anti-Rabbit AF633, Invitrogen, cat. A-32795; Goat anti-Rat AF568, Invitrogen, cat. #A11077; Goat anti-Guinea Pig AF488, Invitrogen, cat. #A-11073; Goat anti-Chicken AF555, Invitrogen, cat. #A21437; Donkey anti-Goat AF555, Invitrogen, cat. #A32816; Donkey anti-Rabbit AF647, Invitrogen, cat. #A32795; Donkey anti-Mouse AF488, Invitrogen, cat. #A21202; Donkey anti-Rat AF488, Invitrogen, cat. #A-21208) at a dilution of 1:200. Sections were incubated in secondary antibodies for 2 h at room temperature, protected from light. For detection of myelin, sections were incubated in FluoroMyelin (Invitrogen, cat. #F34651; 1:300) for 20 min. Sections were briefly washed in 1xPBS and mounted and coverslipped using a mounting media with DAPI (Invitrogen, cat. #P36931).

Fluorescent microscopy and HALO quantification

Fluorescent imaging of dMBP, MBP, FluoroMyelin, IBA1 and CD68 was performed using the ZEISS AxioScan Z7 to capture full mouse brain coronal sections at 40X magnification. Images were uploaded to Indica Labs HALO^® image analysis platform. Entire hemisphere and corpus callosum regions of each image were manually annotated. All images were analyzed using Co-localization FL (v2.1.4) algorithms. A Nikon AX/AXR confocal microscope system was used for immunofluorescent imaging of ApoE (40X magnification), Plin2 (60X magnification), and Olig2 (40X magnification) staining. The superior half (including the Hippocampus, corpus callosum, and Iso-cortex) of whole mouse brain coronal sections were imaged and uploaded to Indica Labs HALO^® image analysis platform. The hippocampal, corpus callosum, and iso-cortex region of each image were then manually annotated. All images were analyzed using Co-localization FL (v2.1.4) algorithms.

qRT-PCR for Trem2

RNA was isolated from cyropulverized brain tissue from CPZ 4s2- and 4s2^M mice using the Qiagen RNAeasy Plus Mini Kit (Qiagen 74136). Briefly, ~ 25 mg of cyropulverized tissue was lysed with a guanidine-isothiocyanate containing buffer before addition to a gDNA Eliminator spin column. 85% ethanol was added to the remaining supernatant and the sample was added to a RNeasy spin column for specific binding of RNA. Once eluted, RNA concentration and quality were measured on a NanoDrop One Spectophotometer (ThermoFisher #13–400-525). To achieve a desired concentration of 2ug/20ul, RNA was diluted with nuclease free water (ThermoFisher #AM9937). The High-Capacity RNA-to-cDNA Buffer and Enzyme mix (Applied Biosystems #4387406) was added to each diluted RNA sample and samples were run in a SimpliAmp ThermoCycler (Applied Biosystems #A35422). The resulting cDNA was then used for qRT-PCR to measure Trem2 (ThermoFisher #4331182, Mm04209422_m1) expression with TaqMan Fast Advanced Master Mix (Applied Biosystems #4444557). A QuantStudio3 Real-Time PCR System (ThermoFisher A28567) was used for denaturation and amplification. Relative gene expression was calculated using the comparative Ct method with the housekeeping gene, 18 s, used for normalization. Statistical analysis was performed using GraphPad Prism and log2-transformed fold changes were reported relative to the 4s2- Control group.

Lipidomics

Following euthanasia, the left hemisphere of brains from Control, DM and RM mice were flash frozen in liquid nitrogen and stored at −80 until cryopulverization. Once cryopulverized, lipid extraction from tissue samples was performed using a 1-butanol/methanol protocol optimized for LC-MS analysis. Frozen brain tissues were cryo-pulverized using a 6770 Freezer/Mill (Spex SamplePrep, cat #6770). Approximately 5.0 ± 0.5 mg of powdered tissue was weighed directly into pre-chilled 2 mL screw-cap tubes containing 150 µL of extraction solvent composed of 1:1 (v/v) 1-butanol: methanol with 10 mM ammonium formate. The total weight of each sample was recorded, and tubes were vortexed for approximately 30 s. After vortexing, samples were sonicated at room temperature for 60 min and centrifuged at 15,000 rpm for 10 min at 4 °C. The resulting supernatants were transferred into HPLC vials and stored at − 80 °C until LC-MS analysis.

Lipid extracts were analyzed using an Agilent 1290 Infinity II UPLC system coupled to an Agilent 6495 C TQ-MS mass spectrometer operating in dynamic multiple reaction monitoring (dMRM) mode. Chromatographic separation was achieved using an Eclipse Plus C18 column (2.1 × 100 mm, 1.8 μm particle size) at a flow rate of 0.4 mL/min. Mobile phase A consisted of 10 mM ammonium formate with 5 µM medronic acid in a 5:3:2 (v/v/v) mixture of water, acetonitrile, and 2-propanol. Mobile phase B consisted of 10 mM ammonium formate in a 1:9:90 (v/v/v) mixture of water, acetonitrile, and 2-propanol. The deactivator additive (medronic acid) was added only to mobile phase A. The following gradient program conditions were used: 0.0–2.5 min (15% B), 2.5–2.6 min (15 to 50% B), 2.6–9.0 min (50 to 70% B), 9.0–9.1 min (93 to 96% B), 9.1–11.0 min (96 to 100% B), and 11.1–12.0 min (100% B), 12.0–12.2 min (15% B), 12.2–16.0 min (15% B).

The following mass spectrometer settings were applied: gas temperature, 150 °C; drying gas flow, 17 L/min; nebulizer pressure, 20 psi; sheath gas temperature, 200 °C; and sheath gas flow, 10 L/min. The capillary voltage was set at + 3,500 V (positive mode) and − 3,000 V (negative mode), with nozzle voltages of + 1,000 V and − 1,500 V. The iFunnel high/low pressure RF settings were 200/110 V in positive mode and 150/60 V in negative mode. Data were acquired in dynamic MRM mode, with Q1/Q2 resolution set to unit (0.7 amu). Additional parameters included a Delta EMV of + 100 V (positive) and 0 V (negative), a cell acceleration voltage of 4–5 V, a cycle time of 650 ms, and a time filter width of 0.07 min. A total of 665 MRM transitions were monitored, including 646 in positive mode and 19 in negative mode. Data were analyzed using Agilent MassHunter Workstation Software for peak intergration, quantification, and quality control assessment. Instruments and column were purchased through Agilent. See “Statistics” section for statistical analysis performed.

Statistics

Lipidomics data were analyzed at both the individual lipid species and lipid subclass levels. For each sample, lipid abundances were log2-transformed to correct for right-skewed distributions and stabilize variance. Subclass-level abundances were calculated by summing the intensities of all lipid species within each subclass prior to transformation. All statistical analyses were performed in R (v4.3.0) using the tidyverse, car, broom, and emmeans packages. Two-way analysis of variance (ANOVA) was conducted for each lipid species and subclass, with treatment and genotype as fixed factors. Type II sums of squares were used via the Anova() function from the car package to account for unbalanced group sizes. Post hoc comparisons were performed using estimated marginal means calculated with the emmeans package. Pairwise contrasts were tested between genotypes within each treatment and across all treatment–genotype combinations. All p-values were corrected for multiple comparisons using false discovery rate control via the Benjamini-Hochberg) method. Results were used to identify significantly altered lipids and subclasses across experimental groups.

Principal component analysis (PCA) was performed on log₂-transformed, mean-centered lipid abundance data. All identified lipid species (n = 401) detected across all samples were included to provide an unbiased overview of lipidomic variation. Heatmaps were generated using log₂-transformed and scaled lipid abundance data. Samples were organized according to the six predefined experimental groups (2 genotypes × 3 treatments) rather than unsupervised clustering, to highlight group-specific lipid abundance patterns. The colored blocks correspond to the six experimental conditions.

All values were reported as ± SEM. All statistical analyses were conducted in Prism 10 (GraphPad) unless otherwise noted. Two-way ANOVAs were used to assess significance between more than two groups. Significance was determined by P values less than 0.05 for all tests unless otherwise stated. Significant P values, post-hoc analyses, and experimental groups can be found in figure legends.

Abbreviations

AD

Alzheimer’s disease

APOE

Apolipoprotein E

CC

Corpus Callosum

CC1

Anti-adenomatous polyposis coli antibody clone CC1

CD68

Cluster of Differentiation 68

CNS

Central Nervous System

CPZ

Cuprizone

CX

Cortex

D.P.I

Days Post-Injection

DAPI

4′,6-Diamidino-2-Phenylindole

DM

Demyelination

DNA

Deoxyribonucleic Acid

dMBP

Degraded Myelin Basic Protein

GFAP

Glial Fibrillary Acidic Protein

HIPP

Hippocampus

IBA1

Ionized Calcium-Binding Adapter Molecule 1

LPC

Lysophosphatidylcholine

MS

Multiple Sclerosis

OLIG2

Oligodendrocyte Transcription Factor 2

OPC

Oligodendrocyte Progenitor Cell

PB

Phosphate Buffer

PBS

Phosphate-Buffered Saline

PCR

Polymerase Chain Reaction

PDGFRA

Platelet-Derived Growth Factor Receptor Alpha

PFA

Paraformaldehyde

PLIN2

Perilipin 2

qRT-PCR

Quantitative Reverse Transcription Polymerase Chain Reaction

RM

Remyelination

RNA

Ribonucleic Acid

SEM

Standard Error of the Mean

SNP

Single Nucleotide Polymorphism

THAL

Thalamus

TREM2

Triggering Receptor Expressed on Myeloid Cells 2

Authors’ contributions

G.N., J.F., J.M., and L.J. contributed to conceptualization and manuscript writing. L.G. and L.J. designed and validated the mouse model utilized. G.N., G.H., S.M. contributed to mouse injections and euthanizations. G.N. performed all stereotaxic intracranial injections. D.A., G.N., J.F. contributed to brain tissue cryosectioning. G.N., S.T., C.L., A.P. contributed to the immunofluorescent staining. S.T. and G.N. imaged and annotated all immunofluorescent images. G.N. performed all immunofluorescent analyses. I.S., C.B., and H.W. contributed to lipidomic LCMS execution and data curation. All authors read and approved the final manuscript.

Funding

This work was supported by NIH R01AG081421 (L.A.J.), R01AG080589 (L.A.J.), CNS Metabolism COBRE P20 GM148326 (L.A.J.), NIH T32 AG078110 (L.A.J., S.M.M), and NIH RF1NS118558 (J.M.M.). This work was supported by a grant from the Alzheimer’s Association (L.A.J.), with support from the AD Strategic Fund and the WoodNext Foundation, administered by Greater Houston Community Foundation.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.