Microglia limit the expansion of β-amyloid plaques in a mouse model of Alzheimer’s disease

APPswe/PSEN1dE9 (APP/PS1) were purchased from Guangdong Medical Laboratory Animal Center, China. C57BL/6 mice expressing YFP in layer-V pyramidal neurons (Thy1 H-line) and Rosa26-stop-iDTR (DTR) mice were purchased from the Jackson Laboratories, US. CX ₃ CR1 ^CreER mice were generated in New York University School of Medicine [40]. These mice were crossbred and litters were genotyped by PCR using the following primers: for APP/PS1 mice, 5′- TCATGACTATCCTCCTGGTGG-3′ and 5′- CGTTATAGGTTTTAAACACTTCCCC-3′; for CX ₃ CR1 ^CreER mice, 5′- AAGACTCACGTGGACCTGCT-3′, 5′- CGGTTATTCAACTTGCACCA-3′ and 5′- AGGATGTTG ACTTCCGAGTTG-3′; for DTR mice, 5′- CTGGCTTCTGAGGACCG-3′ and 5′-CGAAGAGTTTGTCCTCAACCG-3′. 12–24 months old quadruple transgenic mice were used for microglia depletion experiment and in vivo imaging. Age-matched animals with APP/PS1 mutation but no CreER or DTR gene were used as controls.

For studies of dendritic abnormalities near plaques, mutant human amyloid precursor protein (FAD APP _670/671 line Tg2576) and mutant human Presenilin 1 (PS1 _M146L line 6.2) mice were obtained from Taconic and the University of South Florida, respectively. All experiments were done in accordance with the institutional guidelines.

Mice were anesthetized with pentobarbital sodium (100 mg/kg) 24–48 h before imaging. Congo Red was injected into the subarachnoid space using a fine glass electrode guided by a micromanipulator. The electrode was backfilled with Congo Red (3 μl, 0.5% in artificial cerebral spinal fluid (ACSF), filtered through a 0.20 μm syringe filter before use, Sigma) and inserted through the thinned skull. After subarachnoid space was reached (as evidenced by free dye diffusion), Congo Red was then pressure injected with a picospritzer (20 p.s.i., 20 ms, 0.8 Hz) over 20–30 min. Dye labeling of amyloid plaques was typically observed within 30 min of injection.

The procedure of transcranial two-photon imaging was described previously [7, 41]. Briefly, after exposing the skull surface and gluing the skull to a custom-made steel plate, a small region ~1 mm in diameter was thinned with a highspeed drill, and a microsurgical blade was then used to continue thinning until the skull area ~500 μm in diameter was ~20 μm in thickness. A picture of the brain vasculature in the thinned region was taken with a CCD camera and used as a landmark for future relocation. Animals were placed under a two-photon microscope, and image stacks of plaques within a depth of 300 μm from the pial surface were obtained in the step size of 1 μm with a 1.05 N.A. 25× water-immersion objective and with two-photon laser tuned to 880 nm. For each frame, a 512 × 512 pixel, zoom 1× image was taken from a 508 × 508 μm ROI. Image stacks from motor, somatosensory and visual cortices yielded a full three-dimensional data set of Aβ plaques labeled with Congo Red in these cortical regions. After imaging, the plate was gently detached from the skull and the scalp was sutured, and the animals were returned to their home cages until the next viewing.

Depletion of microglia was performed according to previously published studies [40]. Animals received two doses of 10 mg of tamoxifen (20 mg/ml, dissolved in corn oil, Sigma) to induce Cre-mediated recombination and DTR expression, with a separation of 48 h between doses. Diphtheria toxin (Sigma) was diluted in PBS (50 μg/ml) and 1 μg of toxin was given i.p. for three consecutive days for depletion of microglia.

Mice were anesthetized and perfused with 0.9% PBS. CNS tissue was removed and fixed in 4% PFA, rinsed with PBS and sectioned at 150 μm with a vibratome. Sections were permeabilized in 1% Titron X-100 in PBS for 3 h and blocked with 5% normal goat serum for 1 h. Sections were incubated overnight with primary antibodies against Iba1 (Wako, 1:500). Sections were then washed with PBS/0.05% Tween-20, and then incubated with Alexafluor-conjugated goat anti-rabbit IgG secondary antibodies (Life Technologies, 1:500) for 2 h. Sections were washed as before and mounted for imaging. Confocal images were obtained on a Biorad Radiance 2000 confocal microscope.

Labeled brain slices were imaged by laser scanning confocal microscopy using either a N.A. 1.25 40× or a N.A. 1.3 60× oil-immersion objective. Neuronal structures labeled with YFP and fibrillar amyloid deposits with Congo Red were scanned sequentially using 488 nm and 568 nm laser excitation, respectively. Image stacks at 0.3–1.0 μm steps were acquired to generate three-dimensional data sets of neuronal structures and amyloid plaques.

To analyze the in vivo images of Aβ plaques, all images were processed using a custom-written Matlab algorithm. For each plaque, the frame with highest mean fluorescence intensity out of the entire image stack was identified as the center frame of this plaque. 2 additional frames above and below the center (5 frames combined) were used to analyze plaque size. In each image frame, plaque border was determined by setting a threshold of 3 times of standard deviation of the background fluorescence (mean value of the lowest 10% fluorescence intensity of the entire image) of the current frame. The averaged plaque area with the border was used as the plaque size.

To quantify fluorescence surrounding a plaque, fluorescence intensity across a radial line from the center point towards the outer area of a plaque was extracted. This radial line was drawn in a randomized direction and was kept in the same direction in analyzing the same plaque between different imaging sessions. This radial line was separated into the “inside” part and the “outside” part by the border of plaque. Fluorescence intensity across the line was normalized to a 0–100% scale with the highest value across the line as 100%. Normalized fluorescence intensity within 5 μm surrounding plaque border in the “outside” part was used as the measurement of fluorescence intensity surrounding the plaque.

To quantify changes of dendritic structures near Aβ deposits in fixed brain slices, dendritic spine densities and shaft diameters were measured with Metamorph software. Spine density and shaft diameter were measured for the proximal or distal dendritic segment (proximal or distal dendritic segments had comparable length, ranging from 20 to 80 μm), depending on the region in which the segment was located. Segments close to the soma and before the plaques were defined as proximal segments. Segments far away from the soma and after plaques were defined as distal segments. The border of Aβ deposits was determined as the position in which a sharp increase in fluorescence intensity was observed.

Previous studies have reported that microglia depletion by activating suicide gene HSVTK or by inhibiting CSFR1 had no significant effect on the number and size of amyloid plaques in mouse models of AD [38, 39]. To better understand the role of microglia in amyloid deposition, we took advantage of a recently-developed method to deplete microglia by administrating diphtheria toxin (DT) into CX ₃ CR1-iDTR mice [40]. Two-month-old CX ₃ CR1-iDTR mice first received 10 mg of tamoxifen by gavage both on Day 1 and Day 3 to induce Cre-mediated recombination and DTR expression under CX ₃ CR1 promoter. 1 μg DT was then administrated by intraperitoneal injections for three consecutive days on Day 10–12. The effect of DT administration on microglia depletion was examined 1 to 14 days after DT administration (Fig. 1a). We found that CX ₃ CR1-iDTR mice had a marked reduction of microglia 1 day after DT administration (Day 13) and the surviving microglia number was 0.8 ± 0.8% compared with control mice without DTR expression. Microglia remained largely absent in the cortex 7 days after DT administration (Day 20, 13.1 ± 1.3%). 2 weeks after DT administration, 57.4 ± 4.6% of microglia were found in the cortex of CX ₃ CR1-iDTR mice, indicating microglia repopulation occurred at this stage (Fig. 1b, c). Similar dynamics of microglia depletion and repopulation following DT administration were observed in the cortex of APP/PS1/CX ₃ CR1-iDTR mice at 15 month of age (Additional file 1: Figure S1). Thus, consistent with previous findings [40], these results show that after DT administration, microglia are efficiently depleted from the cortex of CX ₃ CR1-iDTR mice or APP/PS1/CX ₃ CR1-iDTR mice within one week, but start to repopulate during the second week.

To investigate whether microglia depletion affects Aβ deposition, we crossed APP/PS1 mice with CX ₃ CR1-iDTR mice and depleted microglia in mice more than 12 months old (17 ± 2 months old). We first examined the effect of microglia depletion on the number of amyloid plaques over 1–2 weeks after DT administration (Fig. 2a). To label Aβ plaques, Congo Red was injected into the subarachnoid space 24–48 h before each imaging session (see Methods). Congo Red-labeled Aβ plaques were imaged through thinned-skull window using two-photon microscopy. Time-lapse imaging of the same cortical regions was performed 1 day (Day 13), 1 week (Day 20) and 2 weeks (Day 27) after DT administration (Fig. 2b, d). We found that the average plaque density in APP/PS1/CX ₃ CR1-iDTR mice before DT administration was 515 ± 63 mm⁻³, comparable to that in APP/PS1 mice (574 ± 42 mm^-3, P > 0.05). Consistent with the previous study showing no formation of new plaques over weeks in APP/PS1 mice older than 9 months [42], we found that in >12 month-old APP/PS1 mice, no disappearance of existing plaques or appearance of new plaques was detected over 1–2 weeks (Fig. 2b, c). Notably, the number of Aβ plaques also remained the same over 1–2 weeks in mice when microglia were depleted by DT administration (Fig. 2d, e). No new plaques or disappearance of existing plaques was observed during this period. Thus, consistent with the previous study reporting microglia depletion has no significant effect on the number of plaques [38, 39], our findings suggest that ablation of microglia by DT administration does not affect the number of Aβ plaques over 1–2 weeks.

In addition to examining the effect of microglia depletion on plaque number, we also compared the size of plaques over time in mice with or without microglia by two-photon time-lapse imaging. The Congo Red-labeled area of plaque was used as the measurements of plaque size. The border of amyloid plaques was determined such that the fluorescence intensity of amyloid plaques was above 3 times of standard deviation of the background fluorescence (Fig. 3a, see Methods). The size of Aβ plaques identified in APP/PS1 (308.9 ± 27.6 μm²) and APP/PS1/CX ₃ CR1-iDTR mice (343.3 ± 35.5 μm²) were comparable before DT administration (P = 0.23, Fig. 3b). Consistent with previous studies [42‐44], we found that the size of plaques showed no significant change over 1–2 weeks in APP/PS1 mice older than 12 months (Fig. 3c–f). In contrast to the stable size of amyloid plaques in APP/PS1 mice, we found that after DT administration plaques showed a significant increase (12.8 ± 3.2%, P < 0.001) in size from Day 13 to Day 20 in APP/PS1/CX ₃ CR1-iDTR mice (Fig. 3d, e). Furthermore, during the second week after DT administration when microglia repopulated (Fig. 1b, c), we found that APP/PS1/CX ₃ CR1-iDTR mice showed no additional increase of plaque size from Day 20 to Day 27 (P = 0.32, Fig. 3d, f). Taken together, these results strongly suggest that microglia have an important role in limiting the growth of amyloid plaques.

To further understand how the absence of microglia affects amyloid deposition, we examined Congo Red fluorescence intensity surrounding plaques. Fluorescence intensity across a radial line from plaque center towards outer area was measured (Fig. 4a) and separated into the “inside” part and the “outside” part by the border of plaque (Fig. 4b, see Methods). By examining fluorescence intensity within 5 μm surrounding plaque border in the “outside” part, we found a significant increase of Congo Red fluorescence intensity from Day 13 to Day 20 (43.7%, P < 0.001, Fig. 4c). In the following second week when microglia repopulated (Day 20 to Day 27), fluorescence intensity in the surrounding area didn’t continue to increase (P = 0.84, Fig. 4d). The change and stabilization of Congo Red fluorescence surrounding plaques after microglia depletion and repopulation provide further evidence that microglia are involved in limiting plaque growth.

Our findings in APP/PS1/CX ₃ CR1-iDTR mice showed that microglia depletion resulted in ~13% increase in plaque size within one week. Based on the averaged plaque size in APP/PS1 mice (308.9 ± 27.6 μm², Fig. 3b), we estimated that for each plaque the cortex area covered by Aβ accumulation increase by ~40 μm² over one week in the absence of microglia. Previous studies have shown various abnormalities of axons and dendrites inside or near plaques [5, 7‐10]. It is possible that ~13% expansion of plaques after microglia depletion may lead to more extensive damage in neuronal circuits. To better understand neuronal damage associated with plaques, we examined spine density and shaft diameter of dendrites passing through and outside plaques in fixed brain slices. In this experiment, PS1 mutant mice (PS1 _M146L line 6.2) were first crossed with APP mutant mice (FAD APP _670/671 line Tg2576). This mouse model of AD (PSAPP mice) was further crossed with Thy1 YFP H-line mice to visualize dendrites and dendritic spines of pyramidal neurons in fixed brain slices from 6 to 10-month-old mice. Amyloid plaques were labeled with Congo Red (Fig. 5a). When dendritic segments passing through plaques were divided into “proximal” segments (segments between the cell body and the plaque) and “distal” segments (remaining segments after exiting the plaque, Fig. 5a, b), we found that distal segments of dendrites passing through plaques showed 19.4 ± 2.8% reduction in spine density and 11.1 ± 1.3% reduction in shaft diameter (P < 0.01) as compared to proximal segments (Fig. 5c). In contrast, dendrites that did not pass through plaques and were located > 30 μm away from plaque border showed no significant difference in spine density and shaft diameter between the corresponding distal and proximal segments (Fig. 5d). These findings show that amyloid plaque formation not only leads to dendritic abnormalities inside plaques, but also causes more global changes in distal segments of dendrites passing through plaques. The alterations of distal parts of dendrites passing through plaques suggest that ~13% expansion of amyloid plaques after microglia depletion could cause sizable damage in neuronal connectivity.

Mutations of CD33 and TREM2 genes expressed in microglia have been linked to increased risks of AD [26, 27, 37]. Altered Aβ load has been observed in CX ₃ CR1 ^−/− [29, 30], TREM2 ^−/− and CD33 ^−/− mice [27]. In vitro studies have shown that microglia phagocytose Aβ [33‐35]. Furthermore, in vivo studies have shown that Aβ is localized within microglial lysosomes and microglia volume surrounding plaques correlates with the reduction of plaque size over one month [25]. Together, these studies strongly suggest a role of microglia in Aβ clearance and/or regulating d Aβ deposition. However, previous studies of depleting microglia by introducing suicide gene HSVTK or inhibiting CSFR1 have shown that microglia depletion has no effect on both formation and maintenance of plaques [38, 39]. In the present study, using Cre-dependent microglia depletion and time-lapse imaging, we have now provided direct evidence that in the AD mouse model older than 12 months, microglia depletion over 1 week does not affect the formation or maintenance of amyloid plaques, but leads to a ~13% enlargement of plaque size in the cortex (Fig. 3d, e). Furthermore, 2 weeks after DT administration, microglia repopulation was associated with the stabilization of plaque size (Fig. 3d, f). Our findings strongly suggest the role of microglia in restricting the expansion of plaques.

It is important to point out several differences between previous studies and our work on the effect of microglia depletion. One difference is that previous studies examined the number and size of amyloid plaques using fixed brain tissues from different animals with or without microglia depletion [38, 39]. The variability in plaque number and size between different animals may make it difficult to detect relatively small changes in plaque size after microglia depletion. Taking advantage of time-lapse imaging, we have been able to track the same plaques over time and reveal the relative small changes (~13% over one week) of plaque size in response to microglia depletion. It is also important to note that microglia depletion in previous studies may take longer to occur than in our studies. The prolonged process of microglia depletion might cause compensatory responses such as astrocyte activation, which could lead to degradation of Aβ [45, 46]. Further studies are needed to address these possibilities in order to better understand the role of microglia in amyloid plaque deposition.

In addition to microglia in the brain, peripheral myeloid cells also express CX₃CR1 and therefore could be depleted by DT administration in APP/PS1/CX ₃ CR1-iDTR mice. However, microglia and peripheral CX₃CR1⁺ cells have substantially different turnover rates and are derived from different precursor populations [40, 47‐50]. Microglia are long-lived population [47] and it has been shown that when tamoxifen is administrated ~30 days prior to the administration of DT, CX ₃ CR1-iDTR mice have a dramatic reduction of microglia within 1 day after DT administration [40]. On the other hand, CX₃CR1⁺ cells in the spleen or blood are not affected as these peripheral CX₃CR1⁺ myeloid cells in CX ₃ CR1-iDTR mice are replenished through a CX₃CR1⁻ bone marrow precursor [47‐50] and no longer express DT receptors 30 days after tamoxifen administration [40]. In our experiment, the time period between tamoxifen and DT administration was 7 days. Based on the rapid turnover of peripheral CX₃CR1⁺ cells (Fig. 2 in ref. [40]), we expect that over this 7 day period the majority of peripheral CX₃CR1⁺ cells would be replenished through a CX₃CR1⁻ bone marrow precursor and do not express DT receptors. A small fraction of peripheral myeloid cells would still express DT receptors and would be depleted following DT administration. Therefore, we could not completely rule out the possibility that in addition to microglia depletion, the depletion of some peripheral CX₃CR1⁺ myeloid cells may also affect the growth of amyloid plaques. Future studies to specifically deplete peripheral CX₃CR1⁺ myeloid cells using bone marrow transplantation are needed to determine the potential contribution of peripheral CX₃CR1⁺ cell population in amyloid plaque formation.

A variety of studies have shown that abnormal spine density, spine turnover and diameter of dendritic shaft are associated with Aβ plaques [5, 7‐9, 51]. In the current study, we found dendritic spine loss and shaft atrophy in the distal segments of dendrites passing through plaques when compared with proximal segments in the PSAPP mouse model of AD (Fig. 5a‐c). This distal effect was not found in the dendrites which did not pass through plaques and were 30 μm away from Aβ plaques (Fig 5d). These results suggest that the expansion of Aβ plaques could cause damage not only in the local position but also to the entire distal segments of dendrites passing through. In this way, amyloid deposits might disrupt the signal propagation and protein transportation generated from soma and further lead to the degeneration of distal segments. The distal effect in dendrites passing through plaques also implies that extensive damage of neurites could be caused by enlargement of Aβ plaques when microglia are absent or microglia function is altered. These findings underscore the role of microglia in restricting the enlargement of plaques and limiting the neuronal damage in Alzheimer’s disease. The expansion of amyloid plaques after microglia depletion and plaque-associated dendritic abnormalities likely contribute to memory loss and cognitive decline in AD [15‐19]. However, it is worth mentioning that microglia in AD mice also likely have impacts on the function of neuronal circuits and animal’s behaviors beyond their role in amyloid deposition and associated dystrophic neuritis [40, 52]. Thus, the depletion of microglia and resulting plaque expansion may or may not indicate disease worsening in AD mice. Future studies are needed to examine the impact of microglia depletion on neuronal network activity and plaque-associated synaptic abnormalities in the living mouse cortex in order to better understand microglia functions in AD. Furthermore, while plaque-associated dystrophic neurites have been shown in several AD mouse models and AD patients [4, 5, 7, 9, 51, 53], whether similar effects of amyloid plaques on distal dendrites observed in PSAPP mice also occur in other mouse models of AD remains to be investigated.