Age and Aβ stimulation synergize to increase microglial Slc11a2 and iron loading markers in primary microglia.
To assess a potential role for microglial iron and Slc11a2 in aging and amyloid-related pathology, we isolated glia from young and aged mice for primary cell in vitro experiments. We first observed significant ferritin (FtL) protein deposits in cells isolated from aged compared to young mice (Fig. 1A), demonstrating, as others have shown, a key iron-loading microglial phenotype in aging (38, 75). To determine whether Slc11a2 contributes to this age-associated increase in iron and whether the transporter gene plays a role in amyloid-related disease conditions, isolated microglia from young and aged mice were treated in vitro with an acute stimulus of 1 µM oligomeric Aβ for 24 h and gene expression of Slc11a2 was measured. As others have also shown (24, 45), there was a significant increase in microglial Slc11a2 in response to acute Aβ exposure (Fig. 1B, Treatment, F(1, 35) = 48.91, p < 0.0001). Additionally, glia from the two-year-old aged mice exhibited an augmented Aβ-induced Slc11a2 response, which was significantly greater than the response observed in the cells from young mice (Age, F(1, 35) = 11.21, p = 0.002; young vs. old Aβ, p = 0.005). In addition, there was a robust increase in pro-inflammatory cytokines Tnfα, Il1β, and Il6 in response to Aβ (Fig. 1C-E, Il6: Treatment, F(1, 32) = 41.20, p < 0.0001; Il1β: F(1, 34) = 24.23, p < 0.0001; Tnfα: F(1, 34) = 77.83, p < 0.0001), which was even greater in the glia from the aged mice compared to those isolated from the young mice (significant for Tnfα: Age, F(1, 34) = 6.52, p = 0.015, Interaction F(1, 34) = 5.57, p = 0.024; young vs. aged Aβ p = 0.005). Along with differences in Aβ-induced Slc11a2 gene levels in the glia from the aged mice, there was a significant increase in iron-storage genes Ftl and Fth1 in response to Aβ only in the cells from the aged animals (Fig. 1F-G). Specifically, Aβ induced an increase in Fth1 in the aged glia (Age, F(1, 29) = 12.46, p = 0.001, Treatment, F(1, 29) = 13.67, p = 0.0009), and Fth1 and Ftl were significantly higher in response to Aβ in the aged cells when compared to the young cells (Fth1, young vs. aged, p = 0.01; Ftl: Age, F(1, 34) = 7.92, p = 0.008, young vs. aged, p = 0.02). There were no differences in Tfrc gene expression – another main iron importer gene – due to age or Aβ treatment (Fig. 1H, p > 0.05), suggesting that a specific gene expression increase in Slc11a2 may accompany age- and Aβ-related changes in cellular iron and inflammatory status. Aβ also decreased Slc40a1 levels (gene for ferroportin, main iron exporter) to a similar degree in the cells from the young and aged animals (Fig. 1I, Treatment, F(1, 30) = 23.40, p < 0.0001), further suggesting that a specific alteration in Slc11a2 in response to age and amyloid may be involved in the progression of disease.
DMT1 inhibition in vitro significantly blunts Aβ-induced inflammation and decreases cellular iron levels in immortalized microglia.
Based on the purported roles for DMT1 in Aβ stimulation and iron load, we characterized the effect of inhibiting DMT1 on Aβ and iron-induced inflammation in an in vitro system. Cells from the murine immortalized microglial cell line, “IMG” cells (62), were treated with ebselen, a pharmacological inhibitor of DMT1 (63), before subsequent treatment with scrambled Aβ, oligomeric Aβ1−42 alone, or iron (50 µM FAC) + Aβ1−42. Aβ stimulation leads to a robust increase in microglial pro-inflammatory Il1β, Il6, Tnfα, and Nos2 transcription, as expected (Fig. 2A-D, Il1β: Treatment, F(3, 22) = 16.78, p < 0.0001; Il6: Treatment, F(3, 23) = 5.28, p = 0.006; Tnfα: Treatment, F(3, 23) = 10.89, p = 0.0001; Nos2: Treatment, F(3, 23) = 16.31, p < 0.0001). Addition of 50 µM FAC did not have a significant effect on Aβ-induced inflammatory markers. Ebselen profoundly decreased the Aβ-induced pro-inflammatory cytokine response for all three cytokines assayed along with Nos2, even in the absence of excess iron added to the media (Aβ alone condition) (Fig. 2A-D, Il1β: Interaction, F(3, 22) = 15.76, p < 0.0001; Il6: Interaction, F(3, 23) = 4.81, p = 0.0096; Tnfα: Interaction, F(3, 23) = 6.89, p = 0.0018; Nos2: Interaction, F(3, 23) = 13.18, p < 0.0001). Aβ induced a significant upregulation in Slc11a2 and ebselen inhibited this increase when a bolus of FAC was added as well (Fig. 2E, Treatment, F(3, 23) = 5.07, p = 0.008, Interaction, F(3, 23) = 3.49, p = 0.032). This was paralleled by a change in total intracellular iron levels as measured via ICP-MS, where ebselen significantly decreased cellular iron levels in the FAC + Aβ1−42 condition (Fig. 2F, Treatment, F(3, 16) = 72.53, p < 0.0001, Ebselen, F(1, 16) = 4.15, p = 0.058, Interaction, F(3, 16) = 3.52, p < 0.05). These data demonstrate associations between DMT1 inhibition, decreases in cellular iron levels, and blunted Aβ-induced pro-inflammatory responses in IMG cells.
Microglial Slc11a2 knockdown results in a hyperactive phenotype in female APP/PS1 mice and worsens hyperactivity in male APP/PS1 mice at 12–15 months.
To determine the effects of knocking down Slc11a2 in vivo, we generated a transgenic mouse line allowing for inducible knockdown of Slc11a2 in microglia between 5–6 months of age. Between 7–9 months after tamoxifen treatment, when mice were 12–15 months of age, male and female control WT, control APP/PS1, Slc11a2KD WT, and Slc11a2KD APP/PS1 mice were run through a series of behavioral assays to assess the effect of microglial Slc11a2 knockdown on aspects of behavior and cognition.
First, to assess locomotor activity, mice were tested in elevated zero maze (EZ maze, 5 min), open field chambers (45 min), and one-trial spontaneous alternation Y-maze tests (6 min) and total distance traveled was measured in each. In females, control APP/PS1 mice did not exhibit differences in baseline locomotor activity compared to control WT female mice in any of the assays tested (Fig. 3A-F; p > 0.05). However, microglial Slc11a2KD female APP/PS1 animals exhibited a significant increase in distance traveled in all three activity measurement assays compared to their non-APP/PS1 counterparts (Fig. 3A, C, E, F; activity measurements, EZ maze: APP/PS1, F(1, 38) = 9.28, p = 0.004, Interaction effect, F(1, 38) = 12.29, p = 0.001; open field: Interaction, F(1, 39) = 5.36, p = 0.03; Y-maze activity: APP/PS1, F(1, 40) = 5.23, p = 0.03, Interaction, F(1, 40) = 5.92, p = 0.02; arm entries in Y-maze: APP/PS1, F(1, 40) = 5.76, p = 0.02, Interaction, F(1, 40) = 7.93, p = 0.008). As control measurements to assess for anxiety-like behavior, the amount of time spent in the open arms of the EZ maze (Fig. 3B, p > 0.05) or in the center area of the open field chambers were not significantly different (Fig. 3D, p > 0.05). Additionally, there were no significant differences in Y-maze spontaneous alternation capacity between any groups (Fig. 3G, p > 0.05).
Male APP/PS1 mice exhibited a significant increase in activity in the EZ maze compared to WT controls (Fig. 4A; EZ Maze: APP/PS1 effect, F(1, 48) = 22.61, p < 0.0001). There was a significant main effect of Slc11a2 knockdown on activity in the EZ maze in males (Knockdown effect, F(1, 48) = 8.18, p = 0.0063), and post-hoc analyses revealed that Slc11a2 knockdown had a greater effect on the hyperactive phenotypes observed in the APP/PS1 males compared to corresponding controls (Fig. 4A, EZ maze: Control vs. APP/PS1, p = 0.013, Slc11a2KD Control vs. Slc11a2KD APP/PS1, p = 0.0006). This was observed in the absence of a significant anxiety-like phenotype, with no difference in time spent in the open arms of the EZ maze (Fig. 4B). Male APP/PS1 mice did not show any significant differences in total distance traveled or anxiety-like behavior in the open field chambers over 45 min (Fig. 4C-D, p > 0.05). However, there was a significant APP/PS1-associated increase in activity in a 6 min Y-maze in the males (Fig. 4E; Y-maze: APP/PS1 effect, F(1, 46) = 8.40, p = 0.006), which was exacerbated in the Slc11a2 knockdown animals (Fig. 4E, Y-maze activity post-hoc comparisons: Control vs. APP/PS1, p = 0.39, Slc11a2KD Control vs. Slc11a2KD APP/PS1, p = 0.012; Fig. 4F, Y-maze arm entries: F(1, 46) = 5.65, p = 0.02; post-hoc comparisons: Control WT vs. APP/PS1, p = 0.61, Slc11a2KD WT vs. Slc11a2KD APP/PS1, p = 0.03). There were no significant differences in Y-maze spontaneous alternation capacity as a measure of working memory (Fig. 4G). Overall, these data suggest that microglial Slc11a2 knockdown is associated with an exaggerated hyperactive phenotype in the APP/PS1 animals, particularly in female mice.
Slc11a2 knockdown worsens memory performance in Morris water maze and cued fear conditioning assay in APP/PS1 females.
To determine whether Slc11a2 knockdown affected measurements of well-being, cognition, and longer-term learning and memory, several behavioral tasks were utilized. An overnight nest building assay revealed a robust APP/PS1-associated deficit in nestlet amount shredded in the females; however, there was no additional effect of Slc11a2 knockdown on this measurement of cognition and well-being (Control WT mean, 4.3 g ± 0.28; Control APP/PS1 mean, 1.73 g ± 0.30; Slc11a2KD WT mean, 3.5 g ± 0.47; Slc11a2KD APP/PS1 mean, 1.48 g ± 0.35; APP/PS1 effect, F(1, 38) = 43.54, p < 0.0001). To assess learning and spatial memory, mice underwent five days of trials to find a hidden platform in Morris water maze (MWM), a widely used test for hippocampal-dependent spatial navigation and memory. Over the course of five days, all female mice (regardless of APP/PS1 genotype or Slc11a2 knockdown) effectively learned the location of the platform compared to their baseline on day one, exhibiting significantly shorter path lengths to find the platform by day five (Fig. 5A; Day effect, F(2.75, 110.1) = 11.38, p < 0.0001). Female APP/PS1 mice were not different than control WT females at finding the hidden platform during training days. However, microglial Slc11a2KD female APP/PS1 animals exhibited slightly longer path lengths to find the hidden platform, although this was not statistically significant (p = 0.1). Furthermore, in accordance with data from earlier tasks assessing locomotor activity, female Slc11a2KD APP/PS1 mice were significantly more hyperactive in the water maze (i.e., greater average swim speed) compared to all other groups (Fig. 5B; Knockdown x APP/PS1 Interaction, F(1, 40) = 5.45, p = 0.025). Mice underwent one 60 sec probe trial for memory of platform location 24 h after the last set of training trials, in which the platform was removed from the pool and mice were allowed to swim freely. There were no significant differences in time spent in the target quadrant where the platform location was previously (Fig. 5C, p > 0.05); however, female APP/PS1 mice overall exhibited a decrease in time spent around the exact platform location (exact platform location, plus 1.5 cm surrounding radius) compared to WT littermate controls (Fig. 5D; Females: APP/PS1 effect, F(1, 39) = 8.90, p = 0.005). Female Slc11a2KD APP/PS1 mice exhibited a significant further reduction in time spent around the platform location, suggesting an exacerbated loss of memory function in these animals (Females: post hoc analysis: Control WT vs. Control APP/PS1, p = 0.68; Slc11a2KD WT vs. Slc11a2KD APP/PS1, p = 0.004). To further assess the effects of Slc11a2 knockdown on memory function, we utilized a fear conditioning assay in which a tone was succeeded by a mild foot shock. During the initial training session, all groups significantly increased freezing by the third tone presentation, albeit APP/PS1 females overall froze less over the course of the 8 min training session (Fig. 5E, Time effect, F(6.9, 279.3) = 41.1, p < 0.0001; Interaction of Time x APP/PS1, F(15,600) = 4.91, p < 0.0001). In the contextual fear conditioning assay, female APP/PS1 mice exhibited a disease model-associated deficit in fear memory (Fig. 5F, APP/PS1 effect, F(1, 39) = 12.26, p = 0.0012); although, there was no additional effect of Slc11a2 knockdown. However, in the cued fear conditioning memory task, female Slc11a2KD APP/PS1 mice displayed a significant worsening in fear memory associated with presentation of a tone (Fig. 5G, Knockdown x APP/PS1 Interaction, F(1, 39) = 4.19, p = 0.047). Indeed, although all females exhibited an increase in freezing in response to the presentation of the tone (Tone, F(1, 39) = 145.2, p < 0.0001), female Slc11a2KD APP/PS1 mice were significantly less responsive compared to all other groups (Fig. 5H; Interaction of Knockdown x APP/PS1, F(1, 39) = 5.39, p = 0.026).
Male APP/PS1 animals displayed a significant deficit in nest building capacity compared to littermate WT control mice, with no additional effect due to Slc11a2KD (Control WT mean, 3.17 g ± 0.52; Control APP/PS1 mean, 2.03 g ± 0.41; Slc11a2KD WT mean, 3.82 g ± 0.34; Slc11a2KD APP/PS1 mean, 2.35 g ± 0.44; APP/PS1 effect, F(1, 47) = 9.15, p = 0.004). In the MWM, all males regardless of experimental group learned the location of the platform by the end of five training days, albeit APP/PS1 males exhibited longer path lengths over the course of the training compared to WT controls (Fig. 6A; Males: Day effect, F(2.891, 135.9) = 20.45, p < 0.0001; APP/PS1 effect, F(1, 47) = 5.99, p = 0.018). This behavioral phenotype was observed in the absence of differences in swim speeds between groups (Fig. 6B, p > 0.05), demonstrating a disease model-associated learning deficit in the males. In the MWM probe trial, there were no significant differences between groups in time spent in the target quadrant of the previous platform location (Fig. 6C, p > 0.05); however, male APP/PS1 mice overall spent significantly less time around the remembered platform location (platform location, including 1.5 cm surrounding radius) compared to WT controls (Fig. 6D; Males: APP/PS1 effect, F(1, 46) = 6.55, p = 0.01). There were no differences in male Slc11a2KD animals compared to Slc11a2-intact control animals in MWM. In the fear conditioning task, male APP/PS1 animals exhibited decreased freezing during the training session (Fig. 6E, Interaction of Time x APP/PS1, F(15, 705) = 2.25, p = 0.004). There were no significant differences between any groups of the males in the contextual fear conditioning assay (Fig. 6F, p > 0.05), although male APP/PS1 mice overall performed worse on the cued fear conditioning task for memory compared to WT controls (Fig. 6G-H, APP/PS1 effect, F(1, 46) = 4.15, p = 0.047). Slc11a2 knockdown had no effect on performance in these assays in the males. Overall, these data suggest that microglial Slc11a2 knockdown is associated with significant worsening of cognitive dysfunction in several tasks in a sex-specific manner, particularly in female APP/PS1 animals.
Hippocampal microglia from female Slc11a2KD APP/PS1 animals exhibit significant alterations in DAM-like inflammatory and oxidative gene markers.
Significant alterations in gene expression from isolated microglia have been shown in AD models and human patients (35, 36). Thus, to examine transcriptomic changes in microglia in our studies, we magnetically isolated CD11b + microglia from the bilateral hippocampus from female mice for bulk RNA-sequencing. Primarily, we sought to determine changes in hippocampal microglia that may underlie the behavioral and memory-associated deficits observed primarily in the female Slc11a2KD APP/PS1 animals. In the Slc11a2 knockdown microglia, we first confirmed abrogation of expression in the Slc11a2 gene between exons 6–8 (Additional File 4: Fig. S2A), similar to what has been shown by others in this mouse model used to knockdown Slc11a2 (76, 77). Principal component analysis revealed a primary effect of APP/PS1 genotype on overall gene expression in isolated cells (Fig. 7A). As expected, hippocampal microglia isolated from APP/PS1 control animals exhibited a significant and robust pattern of differential gene expression compared to microglia isolated from WT controls. We found 1,301 differentially expressed genes (DEG) that were elevated in microglia from APP/PS1 control animals and 1,236 genes that were significantly downregulated in APP/PS1 controls compared to WT controls (adjusted p-value < 0.05). In examining the top 50 DEG (by fold-change and adj. p-value) in hippocampal microglia isolated from APP/PS1 compared to WT control females, we observed changes in similar gene markers previously reported in AD-associated microglia. Specifically, there were robust increases in microglial phagocytic marker Cd68 (78), hypoxia-related gene Hif1α (79), aging-associated marker Clec7a (80), lipid-droplet-associated marker Plin2 (81), as well as Type I IFN-signaling gene, Mamdc2 (82) (Additional File 5: Fig. S3A). DEGs that were downregulated in APP/PS1 hippocampal microglia compared to WT controls included homeostatic microglial marker Tmem119, as well as iron export gene, Slc40a1 (ferroportin) (Additional File 5: Fig. S3B). Gene-set enrichment analysis (GSEA) revealed significant upregulations in genes involved in cholesterol homeostasis, cellular metabolism, and inflammatory activation in APP/PS1 microglia (Additional File 5: Fig. S3C), similar to what others have shown previously in AD models (83).
To determine the effect of Slc11a2 knockdown on hippocampal microglia, we first compared microglial gene expression between Slc11a2KD and Control WT females. As a result of knockdown alone, we only found 7 DEGs (Additional File 6: Fig. S4A). Top genes altered included Ccr6 and Cd5 (Additional File 6: Fig. S4B-C). We then aimed to determine how Slc11a2 knockdown affects microglial gene expression in the APP/PS1 female animals. There were 150 genes significantly upregulated and 484 downregulated genes in microglia isolated from Slc11a2KD APP/PS1 animals compared to microglia from control APP/PS1 mice. Of these DEGs, Enpp2 and Ttr were robustly upregulated in knockdown cells compared to controls (Fig. 7B). Of the top 50 identified DEGs between Slc11a2KD APP/PS1 and control APP/PS1 females, Apoe (encoding apolipoprotein E), Cybb (gene for NOX2), and homeostatic marker Bin2 were also significantly downregulated in the knockdown cells compared to the control APP/PS1 cells (Fig. 7B-C). GSEA in the Slc11a2KD and control APP/PS1 microglia revealed significant increases in genes associated with cellular metabolism – in particular, oxidative phosphorylation and fatty acid metabolism – and reactive oxygen species (ROS) pathways (Fig. 7D). Slc11a2 knockdown cells also exhibited significant decreases in genes associated with TNF and NFκB inflammatory signaling and Wnt signaling. When comparing relative expression of specific genes in the sequencing dataset, we observed significant alterations in several genes involved in inflammatory and iron-related pathways in Slc11a2KD versus control cells from APP/PS1 females. Specifically, we observed a significant decrease in DAM markers Ctsb and Csf1 (84) in the knockdown cells, as well as a significant increase in Tgfbr1 (p < 0.05) and increase in Trem2 compared to control cells (although not statistically significant, p = 0.068) (Fig. 7E). In examining genes related to iron handling and redox status, we observed a significant increase in iron exporter gene Slc40a1 and antioxidant gene Gpx4 in Slc11a2KD APP/PS1 cells compared to control APP/PS1 microglia (Fig. 7F). Additionally, Slc11a2 knockdown cells exhibited decreases in pro-oxidant genes, such as Hif1α and Cybb, and a robust decrease in the iron-related gene encoding ceruloplasmin (Cp) (Fig. 7F). Although Slc11a2KD cells isolated from APP/PS1 mice exhibited significant differences in the expression of several DAM markers compared to control APP/PS1 microglia, Slc11a2KD APP/PS1 microglia displayed a transcriptional profile still distinct from control, non-APP/PS1 WT cells (black dotted line, Fig. 7E, F). In comparison to control WT cells, Slc11a2KD APP/PS1 microglia upregulated DAM and aging-related markers Csf1, Hif1α, Cybb, and Ctsb – albeit, to a lesser degree than control APP/PS1 microglia.
Initial assessment of overall gene expression via PCA and DEGs in these samples revealed significant variance in gene expression in one sample in the Slc11a2KD APP/PS1 group compared to the rest of the Slc11a2KD APP/PS1 biological replicates (sample labeled as -0004 in PCA plot shown in Additional File 4: Fig. S2B and in heat map Fig. 7B). Although this sample was not considered to be a statistical outlier, further RNA-seq analyses conducted following the removal of this sample are shown in Additional File 7: Fig. S5. In this analysis, there were 2,210 genes significantly upregulated and 2,230 significantly downregulated in the Slc11a2KD versus control APP/PS1 females (Additional File 7: Fig. S5B). The top DEGs revealed upregulations in genes including phagocytic-associated Igkc, along with Ttr and Enpp2 (Additional File 7: Fig. S5C and D). Slc11a2KD cells also exhibited significant downregulations in heat-shock-related genes Hspa1a and Hspa1b, as well as in inflammatory-related genes Lag3, Inpp5d, Erap1, and H2-Aa, and in LDAM (lipid-droplet-accumulating microglia) marker Ly9 (Additional File 7: Fig. S5C and D). Overall, these data suggest that microglial Slc11a2 knockdown in females decreases the DAM-like and aging-associated transcriptional signatures in the APP/PS1 model.