Microglial Cx3cr1 knockout prevents neuron loss in a mouse model of Alzheimer’s disease

It has long been thought that microglia, the immune effector cells of the brain, stay in a resting state before being activated by different insults. This view has changed, as it has been shown that microglia actively survey their surrounding brain parenchyma by protracting and retracting their fine processes under healthy conditions¹. In Alzheimer’s disease, microglia represent a double-edged sword. On the one hand, microglia can have a beneficial effect by secreting neurotrophic factors and phagocytosing amyloid beta (Aβ)², the latter of which remains controversial³. On the other hand, microglia may also be neurotoxic⁴. Little is known about the neurotoxic role of microglia in Alzheimer’s disease. Human peripheral blood monocytes that are stimulated with Aβ induce neuron loss in vitro⁵. Neurons cultured without microglia are resistant to Aβ-induced neurotoxicity⁶.

To shed more light on the neurotoxic role of microglia in Alzheimer’s disease, we simultaneously monitored the fate of individual microglia and neurons in the intact brain of a triple-transgenic Alzheimer’s disease mouse model⁷ (3xTg-AD: PS1_M146V knockin, transgenic APP_Swe and tau_P301L) over a period of 1 month by two-photon in vivo imaging⁸ (Supplementary Fig. 1 and Supplementary Video 1). To fluorescently label neurons with yellow fluorescent protein (YFP) and microglia with green fluorescent protein (GFP), we intercrossed these mice with two additional transgenic mice to yield a quintuple-transgenic (5xTg) mouse model with a PS1_M146V knockin, gfp inserted into Cx3cr1, transgenic APP_Swe, tau_P301L and yfp (Supplementary Methods and Supplementary Fig. 2). The transgenic mice with GFP-labeled microglia also enabled us to interfere with neuron-microglia communication, as gfp was inserted into the chemokine receptor Cx3cr1 locus, leading to CX₃CR1 deficiency in homozygous 5xTg-Cx3cr1^−/− mice. CX₃CR1 is the unique receptor for fractalkine/CX₃CL1, which is expressed in neurons and presumably acts as a membrane-bound adhesion molecule and/or cleaved chemoattractant and is important for recruiting CX₃CR1-expressing microglia to injured neurons^9,10.

We were able to detect and observe neuron loss simultaneously with surrounding microglia in living mice (Fig. 1a). A significant loss of 9.6 ± 2.9 YFP-positive layer III neurons per mm₃, representing 1.8% (n = 626, P < 0.01), was observed in 5xTg-Cx3cr1**^+/− mice at 4–6 months of age (Fig. 1a,b). None of the control mice, Cx3cr1^+/− or Cx3cr1^−/−, exhibited neuron loss over the 28-d imaging period (Fig. 1b). Knockout of Cx3cr1 rescued the neuron loss in 5xTg-Cx3cr1^−/− mice, indicating that microglia are necessary for neuron elimination (Fig. 1b). At this age, layer III neurons of 5xTg-Cx3cr1 mice were immunoreactive for intracellular Aβ but lacked hyperphosphorylated tau and extracellular amyloid plaques (Supplementary Fig. 3). Intact neurons that received the identical laser power were detected next to lost neurons, ruling out the possibility of phototoxicity (Fig. 1a). Moreover, we used postmortem Nissl and DAPI staining on previously imaged neurons and found that the elimination of the neuron was the cause of the neuronal signal loss in 100% of the analyzed cases (n = 4 mice) rather than just a loss of YFP expression (Supplementary Fig. 4).

To quantify the microglia density and velocity around lost neurons, we tracked individual microglia in 5xTg and control mice over a 28-d period (Fig. 1c–f and Supplementary Video 2). In 5xTg-Cx3cr1^+/− mice, the density of microglia significantly increased (P < 0.001) over time around lost neurons from 1.3 ± 0.3 to 3.4 ± 0.2 × 10⁴ microglia per mm³ but not in areas more than 100 µm distant (Fig. 1g). In 5xTg-Cx3cr1^−/−, Cx3cr1^+/− and Cx3cr1^−/− mice, where layer III neuron loss was not detected, microglia density stayed at a constant level of 1.5 ±0.2 × 10⁴ microglia per mm³ (Fig. 1g). Notably, we found that the microglia density increased significantly (P < 0.001) before neuron elimination (Fig. 1g). In addition, microglia migration to the lost neuron stopped after neuron elimination (Supplementary Fig. 5). Moreover, the migration velocity of microglia around lost neurons in 5xTg-Cx3cr1^+/− mice was more than twofold greater (1.05 ± 0.03 µm d⁻¹) than in areas more than 100 µm distant (0.41 ± 0.07 µm d⁻¹), in Cx3cr1^+/− mice (0.22 ± 0.03 µm d⁻¹) and in Cx3cr1^−/− mice (0.36 ± 0.07 µm d⁻¹). This increase was rescued by Cx3cr1 knockout in 5xTg- Cx3cr1^−/− mice (0.4 ± 0.03 µm d⁻¹; Fig. 1h).

As microglia have been shown to phagocytose Aβ¹¹, we tested the effect of fractalkine receptor knockout on Aβ levels. The levels of insoluble Aβ40 were increased in both 5xTg-Cx3cr1^+/− and 5xTg-Cx3cr1^−/− mice, with no significant difference (P > 0.05) between the two groups (Fig. 1i). Therefore, Cx3cr1 knockout appears not to interfere with the Aβ-phagocytosing activity of microglia.

As a measure of the microglial surveying capacity, we analyzed the total fine process turnover rate (TOR; Fig. 1j, Supplementary Videos 3 and 4 and Supplementary Methods). Cx3cr1^+/− and 5xTg-Cx3cr1^+/− mice had a comparable TOR of 39 ± 1% versus 37 ± 1%. Cx3cr1 knockout in either wild-type (38 ± 1%) or 5xTg (34 ± 3%) mice did not change the TOR significantly (P > 0.05; Fig. 1k). However, there was a significant reduction (P < 0.001) in the TOR per cell of individual microglia from 5xTg-Cx3cr1^+/− mice, which was rescued by Cx3cr1 knockout in 5xTg-Cx3cr1^−/− mice (Fig. 1l). This finding is probably a result of the fact that the number of microglia increased around disappearing neurons in 5xTg-Cx3cr1^+/− mice (Fig. 1g), indicating that the screening of brain tissue is just carried out by more cells.

By simultaneously recording microglia and neurons in their intact environment using repetitive two-photon in vivo imaging, we found that microglia are involved in Alzheimer’s disease–associated cortical layer III neuron loss. Analyzing the kinetics of microglia revealed that they are recruited to the neuron before and not after the elimination of the neuron and with increased migration velocity. However, whether peripheral macrophages or proliferating microglia underlie increased numbers remains an open question. Notably, our results demonstrate for the first time, to the best of our knowledge, a role for the microglial chemokine receptor CX₃CR1 in Alzheimer’s disease– associated neuron loss. Obviously, neuron elimination depends on an intact communication between the microglia and neuron via fractalkine and CX₃CR1, as CX₃CR1 deficiency prevented layer III neuron loss in 5xTg-Cx3cr1^−/− mice. However, the effect of wild-type CX₃CR1 in 5xTg mice remains unknown. Under neuroinflammatory conditions other than Alzheimer’s disease, Cx3cr1 knockout has been shown to increase¹² or decrease¹³ neurotoxicity, as well as have no effect^{14, 15}. Therefore, various neuroinflammatory conditions affect neuronmicroglia communication via fractalkine and CX₃CR1 differently. Cx3cr1 knockout did not change the elevated Aβ levels in 5xTg mice, suggesting that the phagocytosing activity of microglia was not altered or was not involved³. As CX₃CR1 deficiency prevented neuron loss, CX₃CR1 may serve as a target for the development of new therapeutic strategies for the treatment of neuron loss in Alzheimer’s disease.