CSF1R is solely deleted in microglia.
CSF1R conditional knock-out (cKO) was made by a Cre/Lox system (Fig. 1B). Cre recombinase is under the CX3CR1 promoter. Lox sites are on each side of the CSF1R Exon 5. When tamoxifen is injected, the Cre recombinase complex translocates to the nucleus to interact with lox sites thereby CSF1R gene is non-functional. Mice were injected with tamoxifen at 3-month-old (Fig. 1A). To determine whether the KO was efficient, we used RosaredTm-CSF1R-lox/CX3CR1-Cre/ER mice. Mice express robust redtm fluorescence following Cre-mediated recombination in CX3CR1 cells in the brain, meaning the knock-out is effective (Fig. 1C, D). Quantification shows an endogenous activity of Cre-recombinase in mice without tamoxifen is around 20%, however, after induction the knock-out cells reach 89%. To confirm these results, we quantified by immunohistochemistry the percentage of CSF1R+ cells in parenchyma of cKO mice. Results shown a robust decrease of CSF1R expression in cKO mice. These data indicate that our model is reliable, and strongly efficient to induce the knock-out selectively in microglia.
The deletion of CSF1R does not affect microglia survival and delay cognitive decline in APP cKO mice.
As previously described, CSF1R is largely depleted in microglia. Here we show using an unbiased stereological analysis of Iba1-positive cells, no significative difference in microglia number into hippocampus and cortex at 6 and 8-month-old in APP cKO compared to APPSwe/PS1 (Fig. 2A-C). These data are corroborating our previous findings, suggesting that KO has no impact on microglia survival in a specific model of neuronal injury (33). To investigate the impact of the KO on cognition, the memory of APP cKO and their littermate control mice were tested at 3, 6 and 8-month-old using novel object recognition task (NOR). This test is a standard to evaluate the cognitive decline in mouse models of AD (34). The test measures the time spent on exploring the novel object compared to the common object. During the acquisition phase, every animal explored more than 10 seconds each object (Fig. 2E). The test phase analysis shown no differences between groups at 3-month-old. However, at 6-month-old APPSwe/PS1 mice present an expected sign of cognitive decline (***p = 0.0002), which was also confirmed at 8-month-old (***p = 0.0007). APPSwe/PS1 mice spent equal time at exploring both objects rather than exploring the novel one. Interestingly, APP cKO mice did not have the same cognitive decline at 6 or 8-month-old, they spent an equivalent time to explore the novel object as the wild-type group did (Fig. 2D). Our data suggest a protective effect of CSF1R gene deletion to prevent cognitive decline in APP cKO mice.
The nesting behavior was conducted to evaluate the effect of CSF1R knock-out on social withdrawal and apathy linked to AD. As depicted by the Fig. 2F, the nest scores were equivalent between groups at 3 and 6-month-old. However, 8-month-old APPSwe/PS1 mice had a lower score compared with both WT and APP cKO groups (Fig. 2E, 2.5, APPSwe/PS1 ***p = 0.0009, 4.5, WT and 4.3 APP cKO). These data suggest that APP cKO do not exhibit a significant decrease in apathy and social withdrawal.
Altogether these results indicated that APP cKO mice do not present memory and social behavior impairment associated with AD. NOR and nesting tests did not show a significative difference between APP cKO mice and WT, unlike APPSwe/PS1. as expected arbor a robust cognitive decline. Moreover, the number of microglia was equal between APP cKO and their control APPSwe/PS1, suggesting CSF1R is not the only receptor involved in microglia proliferation and survival in this model.
Long-term knock-out reduces volume and plaque number, with the onset of cerebral amyloid angiopathy (CAA).
Cognitive impairment is correlated with the onset of Aβ plaque formation in the cortex and hippocampus in APPSwe/PS1. In this animal model, plaques are established and observable at 6-month-old. Since CSF1R-deleted mice performed as well as WT mice, they should have fewer amyloid deposits in hippocampus and cortex compared to APP mice. Sections were stained with Iba1 and anti-Aβ (6E10) at 6 and 8-month-old (Fig. 3A). Here we observed a difference in plaque number and volume in the cortex area. Using unbiased stereology, we quantified the number of plaques per region and volume of Aβ deposit in APP cKO at 6 and 8-month-old and their littermate controls (Fig. 3B-E). We observed a diminution by 2.1-fold of plaque volume in APP cKO group at 6-month-old (*p = 0.0354 hippocampus, cortex *p = 0.0479). We also wanted to see if the number of plaques in each structure was changed. For this matter we counted every plaque in both hippocampus and cortex and then normalized the data with the volume of these structures. Results are expressed in number of plaques per mm3, the software gave an unbiased count on the whole brain. The relative number of plaques is significantly decreased by 1.8-fold in the cortex of the APP cKO 6-month-old group (*p = 0.0270). At 8-month-old, the volume of plaques is also reduced in hippocampus (*p = 0.0320) and cortex (*p = 0.0227) compared to their control littermates, respectively by 6.6-fold and 10-fold. Actually, the volume of plaque does not differ between 6 and 8 months in APP cKO. However, regarding the number of plaques in hippocampus or cortex, it remains similar in both groups at 8-month-old (Fig. 3E).
These data provide clear evidence that deletion of CSF1R prevents the accumulation and/or induces a better clearance of Aβ in the brain.
The equilibrium between Aβ burden in the parenchymal and blood vessels is well described (35). AD patients present a diminution of the transporter ABCB1 that impairs the efflux of Aβ in blood vessels. According to previous data, we studied the vascular amyloid to see if this transport was maintained. We used Elisa kits to quantify Aβ40 levels in blood vessels. We did not observe any difference at 6 months in APPSwe/PS1 or APP cKO groups. However, at 8-month-old, we detected and significant augmentation of Aβ40 in APP cKO group (**p = 0.0098) (Fig. 3F) associated with a stable expression of ABCB1 at 8 months for APP cKO (*p = 0.023) (Fig. 3G). These data indicate that CSF1R could play a role in CAA onset or at least it may contribute to down-regulate ABCB1. These results also show that peripheral immune system is not adequately activated to clear vascular amyloid.
Altogether these data indicate that the KO has a beneficial effect in this model on cerebral amyloid load by decreasing the volume of senile plaques (Fig. 3B, E) and may accelerate vascular Aβ deposits and CAA (Fig. 3F, G, H).
TREM2/β-Catenin and IL-34 brain protein levels following CSF1R gene deletion.
We have previously demonstrated that a genetic ablation of CSF1R in a non-inflammatory model did not affect microglia proliferation and activation. Moreover, CSF1R knockout microglia overexpress TREM2 following nerve section. However, in acute inflammatory model such as cuprizone, microglia are unable to proliferate and activate properly (33), indicating that the role of CSF1R is dependent on the microglial environment. Here, we detected the same amount of microglia in APPSwe/PS1 and APP cKO and microgliosis around plaques was detected in both groups (Fig. 2A, 3A). This suggests that the genetic deletion in this model did not impair microglial proliferation and survival that may depend on other factors or compensatory mechanisms due to the CSF1R gene deletion.
We then studied the expression of TREM2, β-Catenin and IL-34 in young WT and cKO 10-week-old mice, because we postulated that these molecules can play an important role in AD by compensating for the KO. TREM2 protein levels increased by 2-fold in cKO mice compared to littermate controls (*p = 0.0384), which was associated with the stabilization of β-Catenin (*p = 0.0360) and the diminution of adenomatous polyposis coli (APC), a member of β-Catenin destruction complex (**p = 0.0023) (Fig. 4A).
We then looked at our both groups of 6 and 8-month-old mice and found a robust expression of TREM2, especially at 6-month-old APP cKO mice compared to APPSwe/PS1 age matched animals (Fig. 4B) (*p = 0.036). This was again associated with an augmentation of β-Catenin protein levels (*p = 0.0027). Such up-regulation of TREM2/β-Catenin may suggest that these pathways could compensate for the loss of CSF1R in order to keep microglia alive and functional. Interestingly, other very critical molecules increased in the 6 month-old cKO group, namely BDNF (**p = 0.0071), Syndecan-1 (*p = 0.0498), IL-34 (*p = 0.0274) and PSD-95 (*p = 0.0120) (Fig. 4B). These increases suggest a beneficial effect of CSF1R deletion on neurons, synapses and microglia in the onset of the disease. Similar profile was found at 8-month-old (Fig. 4C), at least for BDNF (*p = 0.0494) and syndecan-1 (*p = 0.0378), but not for PSD-95 and IL-34. It is noteworthy that BACE-1 significantly decreased (**p = 0.001) in the brain of 8-month-old cKO mice (Fig. 4C-D).
The number of microglia following knock-out induction remains unchanged.
Several studies have shown a robust microglia lethality following pharmacological inhibition of CSF1R (36, 37). We have previously demonstrated that number of microglia in hippocampus and cortex between APPSwe/PS1 and APP cKO is not statistically different (Fig. 2A). We aimed to show whether microglia following the KO died and repopulated the brain or signalling pathways previously described are sufficient to keep microglia alive. We have injected tamoxifen in WT and cKO mice at 3-month-old every day for 4 days. We then have sacrificed mice every 2 days after the last injection, until 20th days post-injection. The count of microglia over the time indicates that microglia survived even depleted from CSF1R (Fig. 5A). Interestingly, at D4 we can observe that almost all microglia are KO, unlike at D2 suggesting that tamoxifen induces KO within 4 days after the last injection (Fig. 5B). These data strongly suggest that compensatory mechanisms must take place immediately following the conditional gene deletion to allow such microglial survival and activation in the brain of APP mice or CSF1R is far from being the only receptor involved in such process, at least in this mouse model of AD.