Genetic and pharmacological models of microglia ablation led to efficient microglia ablation
Given the recently reported important roles of microglia in multiple aspects of CNS function, we investigated chemical and genetic microglia ablation models, as described in previous studies, to evaluate the role of microglia under physiological and pathological conditions [15, 16]. We first examined the efficiency of microglia ablation in the Cx3cr1-iDTR model. Cx3cr1-iDTR mice were generated by breeding between Cx3cr1CreERT2 mice and ROSA26iDTR mice and treated with tamoxifen 4–6 weeks ahead of Dtx treatment to specifically target microglia, but not peripheral macrophages. Cx3cr1CreERT2 mice that do not have the ROSA26idTR transgene (termed Cx3cr1-Cre mice in this study) were used as control mice. ROSA26iDTR mice that do not have the Cre transgene (termed iDTR in this study) were used and separately analyzed to compared with control mice. As illustrated in Fig. 1A, B, Cx3cr1-Cre mice (n = 7), iDTR mice (n = 3) and Cx3cr1-iDTR (n = 7) mice were subjected to TAM administration to induce the expression of the diphtheria toxin receptor (DTR) in microglia. Diphtheria toxin (Dtx) was then administered 4–6 weeks after TAM-induced DTR expression to specifically target CNS microglia (peripheral macrophages are replenished from Cx3cr1 negative bone marrow progenitors in 6 weeks and, therefore, are no longer DTR +) [15, 16]. This allows for specific CNS microglia ablation, as previously reported . We adopted a widely used treatment dosage/paradigm for Dtx administration (20 ng/g/day × 3 days) and our data demonstrates that after 3 days of Dtx administration, we were able to achieve over 90% microglia ablation (Fig. 1F, n = 7) in the Cx3cr1-iDTR mice compared to the two different control mouse groups (Cx3cr1-cre only, n = 7 or iDTR only, n = 3). Data shown in Fig. 1C–E are representative images from brain cortex and are consistent across different brain regions (data not shown). These results confirm previous findings indicating that the Cx3cr1-iDTR mouse is efficient in ablating microglia within a short time window (3 days) after an initial Dtx treatment. We then assessed the pharmacological ablation model using the specific CSF1R kinase inhibitor, PLX5622. Figure 1G outlines administration of either PLX5622 or a control diet over the course of 7 days to wildtype mice on a C57BL/6 background. Following 7 days of PLX5622, up to 90% microglia were ablated as shown in Fig. 1J. Representative cortex images of IBA1 staining in Fig. 1H, I show efficient microglia ablation in the cortex, which was consistent across brain areas (data not shown). These results support previous reports showing efficient microglia ablation following 7 days of PLX5622 diet [44, 45].
Genetic microglia ablation results in loss of CSF/ventricular spaces in the brain that is due to the ROSA26iDTR allele but not microglia ablation
To our surprise, we observed a very robust brain phenotype in the genetic ablation model in T2-weighted MRI. MRI and subsequent ventricular size analysis demonstrated that following Dtx administration, 100% of the microglia-ablated (Cx3cr1-iDTR TAM treated) mice (n > 35) showed a substantial loss of CSF and ventricles (both lateral ventricles and the 3rd ventricle, red and green arrows, Fig. 2B) in both female and male Cx3cr1-iDTR mice. Cx3cr1CreER positive but iDTR negative control mice (Cx3cr1-Cre mice) subjected to the same TAM and Dtx treatment were unaffected (n > 39). However, ROSA26iDTR only mice (iDTR mice) that do not have any cre transgene showed the same phenotype of CSF and ventricle loss (n > 30). Indeed, Cx3cr1-iDTR and iDTR mice who received Dtx without TAM also showed the same phenotype (Fig. 2B, n > 4 for each group), confirming that iDTR allele is solely responsible for this phenotype. To ensure this is not due to potential mutation in our in-house breeding colony, we obtained the independently housed iDTR line from JAX and confirmed the same phenotype in there. Representative MRI for both the Cx3cr1-cre only or the iDTR mice (receiving only Dtx) as well as Cx3cr1-iDTR mice (after TAM treatment) before and after Dtx injection are shown in Fig. 2 (female and male Cx3cr1-iDTR or iDTR only mice demonstrated similar results). A cohort of microglia ablated mice (Cx3cr1-iDTR + TAM and Dtx, n = 7) and control mice (Cx3cr1-Cre only + TAM and Dtx n = 7, or iDTR only + Dtx n = 7) were used to quantify ventricular volume from the T2-weighted MRI. Our results show that TAM and Dtx treatment does not lead to ventricular volume changes in Cx3cr1-Cre control mice that do not carry the iDTR allele; however, TAM and Dtx treatment in Cx3cr1-iDTR mice or Dtx only treatment in iDTR mice led to a substantial decrease of CSF/ventricular volume (Fig. 2C, 70% loss, p < 0.001, ANOVA) measured at 1 day after 3 days of Dtx treatment. To confirm the loss of ventricular space in T2-weighted MRI, this phenotype was also validated by histological analysis of Cx3cr1-Cre or iDTR only groups and genetic microglia-ablated (Cx3cr1-iDTR) brain sections (Fig. 2D–F). Histological quantification of the brain section areas demonstrates that at all three representative coronal positions (+ 0.85 mm, − 0.65 mm and − 1.55 to − 2.25 from Bregma), ventricle areas were significantly decreased in the microglia ablated mouse brains or iDTR+Dtx group as compared to Cx3cr1-Cre control littermates (n = 4–5 per group and p < 0.01 or p < 0.001, Fig. 2G). DAPI stained mouse brain sections at the equivalent forebrain position containing the lateral ventricles are presented to clearly show a loss of ventricular space in Cx3cr1-iDTR and the iDTR only groups (Fig. 2D–F). Importantly, the loss of ventricle spaces is not due to either TAM or Dtx administration alone, as Cx3cr1Cre:iDTRwt/wt control mice (Cx3cr1-Cre) that also received TAM and Dtx treatment did not show this phenotype (by MRI or histology). Furthermore, Cx3cr1-iDTR mice expressing DTR on their microglia (that received TAM administration) but were not ablated (i.e., before Dtx injection, or that received vehicle instead of Dtx injection) did not develop this phenotype (Fig. 2B for pre Dtx injection and data not shown for vehicle injection). Since both Cx3cr1-iDTR and iDTR mice receiving just Dtx presented this phenotype, this suggests that the observed loss of ventricular space is specifically due to the iDTR allele and Dtx treatment in this genetic model.
Pharmacological depletion of microglia does not result in loss of ventricular spaces in the brain
After observing the robust phenotype in the genetic ablation model, we wanted to investigate whether this pathological condition is due to the loss of microglia, per se, or is specific to the genetic ablation model. Using MRI to compare the ventricular spaces before and after 7 days of PLX5622 or control diet (Fig. 3), CSF/ventricular volumetric changes were not observed after pharmacological microglia ablation (Fig. 3C, n = 3 per group, p = 0.536 for diet and p = 0.616 for pre vs post diet, Two-way ANOVA). Figure 3B shows representative scans from before and after PLX5622 diet, including data from both males and females, showing similar results. This supports the notion that depleting the microglia population alone is not sufficient to induce the ventricle loss phenotype. Additionally, the mechanism through which PLX5622 depletes the microglia population does not induce the same CSF/ventricle loss phenotype seen in the genetic ablation model.
Microglia ablation in the Cx3cr1Cre:iDTR model but not the PLX5622 model leads to astrocyte activation and upregulation of multiple cytokines in brain
After confirming that the pharmacological ablation method did not show the same phenotype as the genetic model, we next investigated whether additional cytokines are altered upon acute microglia ablation in the genetic model (either Cx3cr1-iDTR ablation or just iDTR mice with Dtx treatment) vs. pharmacological ablation model. Utilizing immunohistology, we first confirmed successful ablation of microglia and the activation of astrocytes, as established by GFAP upregulation within astrocytes, to be consistent with the prior study  in the Cx3cr1-iDTR microglia ablated brains (Fig. 4A, B). qRT-PCR demonstrated that Iba1 mRNA levels are significantly decreased in Cx3cr1-iDTR microglia ablated mice, consistent with our immunohistology results (Fig. 4E, n = 5–7 per group, p < 0.001, Student’s t-test). In contrast, Gfap mRNA levels are significantly upregulated (Fig. 4B and E, n = 5–7 per group, p < 0.001, Student’s t-test), which is consistent with the activation of astrocytes (revealed by an increased GFAP immunoreactivity) in the genetic microglia ablation model. To determine whether cytokines are upregulated in the microglia ablation model, we utilized a multiplex ELISA based approach for targeted discovery using the validated V-Plex inflammatory cytokine panel (Meso Scale Discovery). This panel allows for the evaluation of multiple cytokines in the same sample. Our data showed that of the 10 inflammatory cytokines examined, three of them were significantly increased in microglia ablated (Cx3cr1-iDTR) mouse brains: TNF-α, IL1β and the KC/Gro (Fig. 4D, p < 0.05 or p < 0.01, n = 4 per group). To investigate whether the increase in these cytokines occurs at a transcriptional level and to additionally assess whether iDTR mice similarly exhibit this increase, we measured mRNA levels in the cortex of all three groups of mice (Cx3cr1-Cre, Cx3cr1-iDTR, and the iDTR mice). Notably, TNF-α, IL-1β and KC/GRO cytokine mRNA levels demonstrated a significant increase in genetic microglia ablated (Cx3cr1-iDTR) mice (Fig. 4E, p < 0.01, Student’s t-test, n = 5–7 per group), consistent with the changes in the cytokine protein levels. Thereafter, we examined whether the iDTR mice that were subjected to Dtx treatment had cytokine upregulation and astrocyte activation, as observed in the Cx3cr1-iDTR ablated brains. Our data show that although we observed the same CSF/ventricle loss in the iDTR mice subjected to Dtx treatment, we do not observe any of the cytokine gene upregulation or astrocytes activation at 1 day after the 3 day Dtx treatment (Fig. 4C, E). We next investigated whether these mRNA and protein level changes in inflammatory cytokines and reactive astrocytes were specific in the microglia genetic ablation model. First, although the microglia ablation efficiency is similar to the genetic ablation model, in the PLX5622 model an increase in GFAP expression (as assessed by immunohistochemistry) was not observed (Fig. 4F–G). Additionally, only a decrease in Iba1 mRNA levels was evident without changes in the mRNA of Gfap or inflammatory cytokines (Fig. 4H). Furthermore, multiplex ELISA assay found no changes in corresponding protein levels (Fig. 4I, p > 0.05 for all cytokines, Student’s t-test). This suggests that the loss of CSF/ventricular spaces is not due to loss of microglia per se, and that the cytokine upregulation in the Cx3cr1-iDTR microglia ablated brains is not a driving factor accounting for the loss of CSF /ventricle phenotype in the genetic ablation model.
Loss of ventricular spaces is not a result of increased parenchymal volume or altered brain water content
Two potential alternative mechanisms may contribute to the observed decrease in CSF/ventricular volume in the iDTR genetic model. Parenchyma swelling could lead to expansion of brain parenchyma to, thereby, reduce ventricular spaces (a phenomenon that can be observed in edema caused by stroke, for example). Alternatively, the loss of ventricular spaces could be caused by decreased CSF production/circulation. After establishing the decreased CSF/ventricle size in the genetic ablation model, we sought to investigate whether parenchymal swelling is present in the microglia ablation model that could potentially contribute to the reduced ventricle size. First, we measured the total parenchyma volume (excluding ventricular volume) of control or microglia ablated mouse brains to investigate whether there is brain tissue swelling and a potential increase in brain parenchyma volume. Overall parenchymal volume was measured using 3D MRI images, and no difference was evident in brain parenchyma volume after administration of Dtx (Fig. 5A, n = 3–7, p > 0.05, ANOVA) among the three groups (Cx3cr1-Cre or iDTR or Cx3cr1-iDTR mice). As an alternative method to measure potential brain edema/swelling, brain water content was measured in control or microglia genetically ablated mouse brains using methods previously described . As shown in Fig. 5B, there is no difference in total brain water content percentages among the Cx3cr1-Cre, iDTR and the microglia ablation (Cx3cr1-iDTR) groups (Fig. 5B, n = 6–9, p > 0.05, ANOVA) at 1 day after the last day of Dtx treatment, a time point when ventricular space loss can be readily observed. We additionally used diffusion tensor imaging (DTI) to evaluate water diffusion in the cortex and striatum as well as directionality in the corpus callosum. Similarly, we did not observe differences in diffusion metrics in any location (Fig. 5C–H, p > 0.05 for all parameters examined, ANOVA). In summary, different methods of measuring brain tissue swelling or water content and diffusion suggest that the decrease in CSF/ventricular spaces in the iDTR or the Cx3cr1-iDTR microglia ablated mice is not likely due to tissue swelling in these mouse brains.
The KC/GRO pathway is involved in brain edema formation in an experimental stroke model but does not resolve shrinking ventricles after genetic ablation of microglia
While analyzing brain swelling and water content/diffusion in the genetic microglia ablation model, we also explored the potential contribution of the less studied KC/Gro pathway in the observed CSF/ventricular shrinkage phenotype. Of the three cytokines that are elevated in the genetic microglia ablation model, TNF-α and IL1β have been shown to mediate brain inflammation and edema by us and others [46,47,48,49,50]. We therefore decided to focus on the third cytokine, KC/Gro, whose role is less well understood but is starting to gain interest in neurological diseases. The chemokine receptor CXCR2 and its ligands have been implicated in a variety of peripheral inflammatory diseases  and, more recently, have been linked to CNS disorders [52,53,54]. We examined the efficacy of blocking KC/GRO, the most upregulated candidate cytokine revealed by both RT-PCR and ELISA following genetic ablation of microglia (Fig. 4), to mitigate brain CSF/ventricular shrinkage. Specifically, an inhibitor (Repertaxin) to the receptor of the KC/Gro cytokine, CXCR2, was used. Repertaxin is a non-competitive allosteric inhibitor for CXCR2 and has been shown to be neuroprotective in rodent stroke models in two previous studies [43, 55]. We utilized a previously established Repertaxin dosage  to inhibit the CXCR2 signaling pathway during each day of Dtx administration (6 h after administration to avoid drug interactions) to assess whether blockade of KC/GRO signaling was sufficient to resolve the ventricle shrinking phenotype observed in the genetic microglia ablation model. We validated the efficacy of the Repertaxin dosage on a stroke model which showed decreased edema of the stroke mice treated with Repertaxin (Fig. 6A–D). As shown in Fig. 6, Repertaxin treatment in stroke mice decreased brain edema but did not prevent or improve the phenotype previously seen in the genetic ablation model (Fig. 6F–G, n = 3–4, p < 0.001 for control vs. genetic microglia ablation; p > 0.05 for vehicle vs. Repertaxin, ANOVA). This data supports the notion that the loss of ventricular space phenotype in the genetic microglia ablation model is not likely due to edema and cannot be reversed by blocking the KC/GRO pathway.
Loss of ventricular spaces in Cx3cr1-iDTR mice or iDTR mice does not lead to neurodegeneration in cortical layers at 10 days after the last Dtx injection
One previous study has reported that acute microglia ablation using the Cx3cr1-iDTR mouse model leads to neurodegeneration of cortical layers at 10 days after the last Dtx injection . To examine whether we observe similar neurodegeneration in our Cx3cr1-iDTR mouse model or in our iDTR mice that do not have microglia ablation but show ventricular space/CSF loss, we carried out unbiased stereology counts of all layers of somatosensory cortex in Cx3cr1-Cre, Cx3cr1-iDTR and iDTR mice subjected to Tamoxifen and Dtx treatment. At the same time point as previously reported (10 days after the last Dtx injection), we did not observe any differences in NeuN + neuronal density in the cortical layers (Fig. 7), suggesting that neither microglia ablation nor the loss of CSF/ventricular space at this time point leads to neurodegeneration.
Activated IBA1 positive cells are observed surrounding the ventricular spaces and within choroid plexus following genetic ablation of microglia
After establishing that brain swelling/edema is not likely a cause for the observed the ventricular space loss in the genetic microglia ablation model, we investigated other potential avenues that could explain the resulting phenotype. The fact that iDTR mice receiving Dtx also develop this loss of CSF/ventricle phenotype without upregulation of cytokines and reactive astrocytes further supports the notion that production/circulation of CSF might be contributing to this pathology rather than edema or brain swelling. In this light, the next region of interest was the choroid plexus (CP) as it is recognized to be a major source for CSF production in the brain . We examined the IBA1 + cells in CP during the period when ventricular spaces were substantially reduced in both iDTR and Cx3cr1-iDTR mice (d4 after the first Dtx injection). Immunostaining of IBA1 in CP revealed an increased number of IBA1 + cells in both the iDTR and Cx3cr1-iDTR microglia ablated animals. In the microglia ablated mice (Cx3cr1-iDTR), despite a decrease in IBA1 positive cells, validating successful ablation of microglia, in the parenchyma (Fig. 8C, positive cells in parenchyma highlighted with arrow heads), the CP and ventricular wall showed an enriched IBA1 positive population following microglia ablation (representative images in Fig. 8D, positive cells in the CP highlighted with arrows quantification in Fig. 8G, p < 0.001, ANOVA). In iDTR mice subjected to Dtx treatment, we did not observe any ablation of IBA1 + microglia in the parenchyma (Fig. 8E, positive cells highlighted with arrow head) but observed a similar increase in the IBA1 positive population in the CP following Dtx treatment (representative images in Fig. 8F, positive cells in the CP highlighted with arrows quantification in Fig. 8G, p < 0.01, ANOVA), Additionally, the IBA1 positive cells in the CP in microglia ablated brains appear less ramified than homeostatic microglia, which is suggestive of an activated microglia or macrophage origin. This data suggests that, instead of brain tissue swelling, potential pathology in the CP which is reflected by increased activated IBA1 + cells might contribute to the observed ventricular shrinkage phenotype observed in the iDTR and the Cx3cr1-iDTR microglia ablation model. The precise mechanism(s) of pathology in the CP and the effects on various CP cellular components is currently under further investigation.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.