Microglial-associated responses to comorbid amyloid pathology and hyperhomocysteinemia in an aged knock-in mouse model of Alzheimer’s disease

All mice were housed 1–5 per cage (503.22 usable cm²) in a room at 23 °C ± 2 °C, under a 14/10-h light/dark cycle beginning at 6:00 AM, with ad libitum access to water and chow. The APP^NLh/NLh × PS1^P264L/P264L double knock-in (KI) mouse model expresses humanized amyloid precursor protein with the Swedish mutation (K670N/M671L), along with a P264L point mutation in the mouse presenilin 1 gene [24]. The homozygous KI mice were maintained on a combined CD-1/129 background, with wildtype (WT) controls derived separately from matings of heterozygous KI animals. The experiment was carried out in a 2 × 2 diet by genotype design, where half the mice of each genotype were randomized to receive 8 weeks of control (Envigo, #TD.01636) or HHcy diet (Envigo, #TD.97345), beginning between 52 and 54 weeks of age (M = 53.4 weeks, SD = 0.36). The HHcy diet was deficient in vitamins B6, B9, and B12 and supplemented with excess methionine, while the control diet was nutritionally matched with normal methionine and B-vitamin levels [25]. Humane-endpoint criteria were 20% weight loss across the study or 10% weight loss within two weeks, along with detrimental changes in body condition (e.g., lethargy, hunched posture, dehydration, dermatitis). None of the mice reached humane-endpoint criteria; however, three mice died during the study and were excluded from all analyses: one KI female receiving HHcy diet, one WT male receiving control diet, and one WT male receiving HHcy diet. Importantly, mice were randomized irrespective of sex, resulting in an imbalance in the final sex ratios. The final numbers of mice for each group were 8 WT on control diet (5 female and 3 male), 10 WT on HHcy diet (5 female and 5 male), 11 KI on control diet (4 female and 7 male), and 10 KI on HHcy diet (8 female and 2 male). As the experiment was not powered to detect sex differences, it should be noted that the imbalance may have increased experimental variability and thus reduced statistical power, or otherwise influenced the data. The experiment was performed in compliance with the Institutional Animal Care and Use Committee of the University of Kentucky.

All mice subsequently underwent transcardial perfusion with 50 ml ice-cold phosphate-buffered saline (PBS) at a flow rate of 10 ml/min before decapitation and brain removal and dissection. The right hemisphere was post-fixed in 4% paraformaldehyde for 24 h at 4 °C and cryo-protected in 30% sucrose for at least 48 h at 4 °C. Samples were subsequently cut into 30 μm sections with a sliding microtome and stored in cryoprotectant solution at − 20 °C prior to immunofluorescent staining. The hippocampus and overlying cortex were dissected from the left hemisphere, flash frozen in liquid nitrogen, and stored at − 80 °C until processing for biochemical endpoints described below. The cortex was divided in two: a more caudal piece directly overlying the hippocampus was used for protein analysis. An adjacent rostral piece was used for mRNA expression analysis. A portion of dorsal hippocampus was used for protein analysis. See supplement 1 for an illustration of brain regions dissected and outlined for analysis.

Staining was performed on free-floating sections spaced approximately 300 apart through the dorsal hippocampus and overlying cortex, for a total of 6–8 sections per animal. Blocking was performed with 10% normal goat serum (Lampire Biological Laboratories, #7332500) and 0.2% Triton X-100 in PBS. All antibodies were diluted in PBS with 3% normal goat serum and 0.2% Triton X-100. Sections were incubated overnight at 4 °C with rabbit anti-P2ry12 (1:500, Anaspec #AS-55043A) and mouse anti-Aβ 6E10 conjugated to Alexa 647 (1:200, BioLegend #803021). Samples were subsequently incubated at room temperature for 2 h in 1:500 secondary antibody solution with Alexa 488 goat anti-rabbit (Invitrogen, #A-11034). Sections were mounted and treated for 5 min with 1x TrueBlack (VWR, #10119-144) in 70% ethanol to reduce autofluorescence before drying and coverslipping in Vectashield mounting medium with DAPI (Vector Laboratories, #H-1200). Entire slides were imaged on a Zeiss Axio Scan Z1 digital slide scanner at × 20 magnification.

The dorsal hippocampus and overlying cortex were manually outlined in the HALO analysis suite (Indica Labs, version 2.3.2089.34) by an investigator blinded to experimental groups. The algorithm minimum intensity settings for all analyses were manually thresholded based upon negative control (no primary antibody). For cortical and hippocampal analyses of P2ry12 staining, and P2ry12 and 6E10 co-localization, the positive pixel algorithm (Area Quantification FL v1.2) was applied to the traced region across all sections per animal to give a single value of total percent area stained per region per mouse. For cortical and hippocampal analyses of 6E10 staining, the object counter algorithm (Object Colocalization FL v1.0) was applied to the traced region across all sections to give a single average count per square millimeter of tissue per region. Minimum object size was set to 10 μm². Plaque size distributions were analyzed in two ways. First, plaques were binned into small (< 300 μm²), medium (301 to 900 μm²) and large (901+ μm²) size ranges. Arbitrary cutoff points were based on a previous study of this model [26], and the average number of plaques/mm² within each bin were determined per animal and compared between diet groups. Second, cumulative size distribution curves were generated for each diet by treating each plaque as an individual data point within each dietary condition, and testing the curves for difference with the Kolmogorov-Smirnov test [27]. For the spatial analysis of microglia around amyloid plaques, 1–4 individual plaques were manually circled per section within the outlined cortical region, totaling 15 plaques per animal or 120 per diet condition. This number was based upon the plaque sampling rate used in the previous reference [24] of about 100 per condition. The circle annotation tool was used to manually circle the entirety of the plaque, and 3 outward concentric partitions added with a fixed width of 30 μm. The positive pixel algorithm was applied within each ring (inner, middle, and outer). To be selected, plaques had to be far enough away from tissue edges or other plaques to enable non-overlapping ROIs completely contained within the tissue. The semi-randomly selected plaques had a significantly larger median size (215.6 μm²) than the full distribution (51.5 μm²), likely because the smallest plaques tended to be more clustered and therefore not amenable to this analysis. Nonetheless, the median size of the plaques analyzed did not differ between the control (293.7 μm²) or HHcy (219.8 μm²) dietary groups (two-tailed Mann-Whitney U, p = 0.235). The cumulative size distributions of the circled plaques were also not significantly different between the two groups (Kolmogorov-Smirnov, D = 0.099, p = 0.482).

Pieces of the dissected hippocampus and cortex from the left hemisphere were used that approximate the regions outlined for the immunofluorescence analyses in the right hemisphere. Tissues from each mouse were homogenized using an Omni Bead Ruptor 24 (Omni International) at a 1:20 weight to volume ratio in lysis buffer: PBS with 1 mM PMSF, 0.5 mM EDTA and 0.2X Halt Protease Inhibitor Cocktail (Thermo Scientific, #87786). Homogenates were centrifuged at 12,000×g for 20 min at 4 °C. Supernatants were collected for cytokine or Aβ_1–40 (Aβ40)/Aβ_1–42 (Aβ42) measurement using MSD custom mouse V-Plex ELISA kits and a human 6E10 Aβ kit (K15200E), respectively. IL-1β, IL-6, TNFα, and CXCL1 were run multiplexed as part of the Proinflammatory Panel 1 Mouse Kit (K15048). CCL3 and IL-33 were separately run multiplexed as part of the Cytokine Panel 1 Mouse Kit (K15245). All samples were run undiluted. For the cytokines, samples were incubated overnight at 4 °C. Samples were incubated for 2 h at room temperature for the Aβ kit. Cytokine levels and Aβ peptide levels were normalized to the total milligrams (mg) of protein loaded in the sample as determined by BCA Protein Assay (ThermoFisher #23225).

RNA was isolated from the cortical tissue piece just rostral to that used for protein extraction with the RNeasy Plus Mini Kit (Qiagen, #74136) according to the manufacturer’s instructions. Tissue was weighed and homogenized in an appropriate volume of buffer, and genomic DNA removed with the gDNA eliminator column. The samples were mixed with 70% ethanol, run through the RNeasy column, washed, and eluted in RNase-free water. Quantity and quality of RNA and 260/280 absorbance ratios were assessed using a NanoDrop spectrophotometer (ThermoFisher Scientific). Reverse transcription was performed with the High Capacity cDNA Reverse Transcription kit (Applied Biosystems, #4368814) according to the manufacturer’s protocol. Real-time PCR was performed on a ViiA 7 Real-Time PCR System (Applied Biosystems) using individual TaqMan® probes for P2ry12, Clec7a, and Itgax and custom TaqMan®Array Cards (ThermoFisher Scientific, #4342253) with TaqMan Fast Advanced Master Mix (ThermoFisher Scientific, #4444557). See Table 1 for a complete list of genes and probes included on the array. Established microglial markers affected by amyloid pathology were chosen based on recent studies [15, 16]. Relative gene expression was calculated using the 2^−∆∆CT method and log2 normalized. HPRT and 18S were used as housekeeping genes for the individual probes and the custom array, respectively.

Analyses and figure generation were performed with JMP Pro 14 (SAS) and Prism 8.3.0 (GraphPad). Two-way analysis of variance (ANOVA) with Sidak’s or Dunnett’s post hoc testing, student’s t tests, or non-parametric Mann-Whitney U tests as indicated in the figure legends or text. For plaque size analysis, Kolmogorov-Smirnov tests were used to determine whether the cumulative plaque size distributions differed between dietary groups. For the gene expression data, three group comparisons were made: WT HHcy versus WT control, KI control vs WT control, and KI HHcy versus KI control. To generate the heatmap, group scores were averaged and transformed into z-scores using JMP; then, the variable clustering script was run to group genes by similarity of expression pattern prior to visualization in Prism. Where reported in the text, group means are followed by the standard deviation (SD).

The average plasma homocysteine levels after HHcy diet were similar between genotypes, with WT mice achieving an average concentration of 19.0 μM (SD = 6.8 μM) versus 14.7 μM (SD = 2.8 μM) in the KI animals (for full dietary validation data see supplement 2). The plasma samples from all mice on control diet were below the limit of detection for the test after dilution—meaning the undiluted samples were less than 5 μM—consistent with what has been reported in the literature by ourselves and others [25, 28]. Additionally, we found that in the KI mice HHcy was associated with greater amyloid plaque burden in the cortex and hippocampus, as measured by 6E10 percent area positivity, and corresponding with an increase in the total number of plaques (Fig. 1a–c). In the cortex, average plaque number increased from 13.3 per mm² (SD = 6.1) in KI mice on control diet to 29.4 mm² (SD = 17.3) in those on HHcy diet. Hippocampal plaques increased from 3.9 (SD = 3.3) to 10.1 (SD = 8.7) per mm² in mice on control or HHcy diet, respectively. No changes were detected in hippocampal or cortical levels of Aβ40, Aβ42, nor the Aβ42/40 ratio as measured by ELISA (Fig. 1d–i). Notably, no evidence of cerebral amyloid angiopathy (CAA) was found.

Next, amyloid plaque size analysis was performed in cortex and hippocampus in two ways. First, plaques were binned into small (10–300 μm²), medium (301–900 μm²), and large sizes (901+ μm²), with average plaques/mm² in each bin compared across dietary conditions (Fig. 2a) [26]. This analysis showed an increase in smaller plaques in the cortex, consistent with increased plaque deposition (Fig. 2), as well as an increase in the medium-sized plaques possibly representing growth of extant plaques. The increase of plaque number in these bins did not reach statistical significance in the hippocampus. Next, cumulative probability distribution histograms were generated for plaque size in each dietary condition (Fig. 2b) [27]. In accordance with the binned data analysis, the distribution was significantly shifted leftward in the cortex of the HHcy versus the control KI group, indicative of a larger proportion of smaller plaques (Kolmogorov-Smirnov test, p = .0078). This was not significant in the hippocampus (Kolmogorov-Smirnov test, p = 0.78).

To assess whether the increased amyloid plaque burden resulted in an exacerbation of pro-inflammatory responses, we measured levels of the inflammation-associated molecules CXCL1, CCL3, IL-33, TNFα, IL-6, and IL-1β by multiplexed MSD ELISA (Fig. 3). The overall changes in the cytokines measured were modest with respect to either strain or diet. There was a significant main effect of genotype on CCL3 expression in the cortex (F(1, 35) = 37, p < .0001) and hippocampus (F(1, 35) = 45.04, p < .0001), with significant differences detected by Sidak’s post hoc tests between KI and WT groups within both dietary conditions. This was also true of cortical IL-1β (F(1, 35) = 22.52, p < .0001). In the hippocampus, there was a significant main effect of genotype on IL-1β (F(1, 35) = 8.45, p = .0063) and also a significant main effect of diet (F(1, 35) = 7.26, p = .0108). Sidak’s post-hoc tests showed significant differences between the KI HHcy and WT HHcy groups as well as the KI HHcy and KI control groups (Fig. 3b). No significant differences were detected in levels of CXCL1, TNFα, or IL-6 in either region as a function of genotype or diet, but there was a significant diet by genotype interaction detected in the hippocampal IL-33 data (F(1, 35) = 5.08, p = .0306). Although the interaction was significant, post hoc tests for differences between KI and WT control or KI HHcy and KI control groups did not reach statistical significance.

We quantified cortical and hippocampal P2ry12 staining as a microglia-specific marker [29] (Fig. 4a) indicative of overall microgliosis without potential inadvertent measurement of infiltrating myeloid cells. There was a significant main effect of genotype on P2ry12 staining in the cortex (F(1, 29) = 9.973, p = 0.0037), but no differences found in the hippocampus (Fig. 4b). A modest decrease in P2ry12 staining was observed in the cortex of KI mice on HHcy versus control diet; however, the difference between KI mice on HHcy diet and KI mice on control diet was not statistically significant (p = 0.08).

Given the established downregulation of P2ry12 in plaque-associated microglia [15, 30], the subtle decrease in P2ry12 positivity in the cortex might be attributable either to a straightforward increase in plaque burden, to a further downregulation of P2ry12 in plaque-associated microglia when HHcy is present, or both. To clarify these possibilities, we performed spatial analyses of P2ry12 and 6E10 plaque staining. We found that the percent area double-positive for P2ry12 and Aβ 6E10 staining was unaltered within cortex between KI control (M = .018%, SD = .015) and KI HHcy mice (M = .018%, .012) (t(18) = 0.0897, p = 0.9295). This was similarly true in the hippocampus of the KI control (M = .013%, SD = .009) and HHcy mice (M = .014%, SD = .011) (t(18) = 0.1474, p = 0.8843). To examine the relationship more closely, a spatial proximity analysis was performed around a subset of plaques in the cortex (Fig. 5a). Results were consistent with a lack of dietary effect on the spatial distribution of P2ry12 staining in response to amyloid plaques. This was true when assessed either by total % positive area of P2ry12 staining (Fig. 5b) or average pixel intensity (Fig. 5c). The relative dearth of P2ry12 staining and intensity in the inner ring closest to the plaque is consistent with previous reports of P2ry12 downregulation near plaques [15, 30]. Taken together, these data indicate that the small decline in overall cortical P2ry12 staining may be primarily attributable to an increase in overall plaque number.

To begin exploring possible additive or interactive gene expression changes that occur in the comorbidity context, we measured whole-tissue cortical changes across a panel of 50 genes implicated in microglial responses to amyloid pathology (Table 1). A subset of 8 randomly selected mice per group (32 total samples) were used for the gene expression analysis, with data normalized to the expression level of WT mice. Chil3 is excluded from statistical analysis because two of the samples did not show any detectable level of this gene (see supplement 3 for full dataset). We first assessed whether our panel, despite not being run on RNA extracted from purified microglial populations, could detect any of the expected increase in DAM markers and decrease in homeostatic markers in the KI control vs WT control comparison. We found that 6 genes were detectably altered (Fig. 6a). In line with expectations, the DAM markers Itgax and Cst7 were upregulated and the homeostatic markers Golm1 and Csf1r were downregulated. Interestingly, the homeostatic markers Fcrls and P2ry12 were found to be upregulated, which could be indicative of an overall increase in the number of microglia.

When comparing WT HHcy and WT control groups, we found 4 significantly upregulated DAM markers: Itgax, Cst7, Clec7a, and Tpt1 (Fig. 6b). To determine whether there was an overall increase in the DAM signal or evidence of unique transcriptional changes, we next compared the KI HHcy and KI control groups. Seven genes were significant within this comparison, and the average transcript levels of these genes across all 4 groups are displayed in Fig. 7a. Interestingly, not only some damage-associated genes (Lag3, Clec7a, Lyz2) but also putative homeostatic markers (C1qb, C1qc, Sall1, Csf1r) were upregulated in the comorbid condition relative to the KI control group. A summary heatmap of z-scores with significant comparisons is shown in Fig. 7b. The full gene expression dataset is provided in supplement 3, and graphical representations of the processed data in supplement 4. These findings indicate that the comorbid pathologies have complex interactive effects in addition to straightforwardly additive ones. Replication and extension of these findings using unbiased whole-transcriptome analyses in isolated microglial (and other) cells will be an important next step in this line of research.