Introduction

Microglia, the brain-resident parenchymal macrophages, account for up to 12% of brain cells, with a particularly high density in the hippocampus [38]. They control multiple processes including neurogenesis, synapse monitoring and pruning, myelination, vasculogenesis, blood–brain barrier integrity, inflammation, and phagocytosis [56].

Microglial surveillance of the brain parenchyma, chemotactic movement of microglial processes and phagocytosis all depend on actin polymerization and subsequent cytoskeletal rearrangements. These, as well as activation, proliferation and production of inflammatory cytokines, are controlled, at least in part, by Ca2+ acting as a second messenger, and rises in [Ca2+]i have been reported to be associated with microglial functions including damage-induced chemotactic and phagocytic events [46, 50, 51, 53, 55]. In general, phagocytic cells exhibit periodic spontaneous Ca2+ transients as well as stimulus-evoked [Ca2+]i rises [9], and microglial Ca2+ activity is important for the regulation of lysosomes, which process incorporated material [63]. While ‘resting’ microglia rarely exhibit spontaneous Ca2+ transients in vivo, external pathological stimuli such as damage to nearby neurons or mechanical distortions rapidly raise their [Ca2+]i [30, 55].

Previous studies using mainly cultured microglia identified a number of ligands evoking Ca2+ responses in microglia, including ATP, ADP and UDP as endogenous stimuli [34, 40, 49] and pathology-related substances such as lipopolysaccharide (LPS) and amyloid beta (Aβ) [9]. However, Ca2+ responses are significantly different for cultured microglia compared to their counterparts embedded in the native CNS environment [9], as microglia partly lose their characteristic genetic profile under more artificial cell culture conditions [8]. In transgenic mice overexpressing Aβ, accumulation of which is a hallmark of Alzheimer’s disease (AD), altered Ca2+ signalling was detected in astrocytes and microglia electroporated with Ca2+-sensing dyes [10].

While previous research has relied mainly on Ca2+-sensitive dyes, we also took advantage of Cx3cr1CreER × GCaMP5g-IRES-tdTomato mice expressing an inducible genetically-encoded Ca2+ indicator (GECI) cross-bred with AppNL−G−F knock-in AD mice [59] to examine microglial Ca2+ activity, both spontaneous and upon inducing brain damage, in the presence and absence of AD-related amyloid plaques. This strategy allows all cells in the field of view to be captured spatially in relation to plaque pathology and abrogates the need for external manipulation to insert the Ca2+ sensor. As similar mechanisms are thought to occur during brain ageing and brain disease [3, 16, 28, 65, 67, 69], we also assessed whether the Ca2+ activity of microglia was affected differently by AD and normal ageing.

Materials and Methods

Animal procedures

Procedures were performed in accordance with the UK Animals (Scientific Procedures) Act 1986 (Home Office License 70/8976). Rats (postnatal day 12–14 for work with patch-clamp loading of Ca2+-sensitive dye) and mice (postnatal day 120–130 and 300–310 for work with the GECI investigating effects of AD and ageing, respectively) were housed in open-shelf units and individually ventilated cages, respectively, with food and water ad libitum. Animals of both sexes were sacrificed by cervical dislocation followed by decapitation (for live imaging), or by an intraperitoneal overdose of pentobarbital sodium (Euthatal, 200 µg/g body weight) followed by transcardial perfusion-fixation with 4% paraformaldehyde (for immunohistochemistry).

For experiments in rats, loading of the Ca2+ indicator Fluo-4 was achieved by whole-cell patch-clamping of hippocampal microglia (see below). To study microglial properties in the initial stages of AD-related pathology, 4-month-old homozygous AppNL−G−F knock-in AD mice or wild-type (WT) littermates were used, where Aβ plaque deposition starts in the AD mice from 2 months of age [59]. For Ca2+ imaging, AppNL−G−F mice were crossed with transgenic mice expressing GCaMP5g-IRES-tdTomato [21] and Cx3cr1CreER [70], to generate offspring where microglia exhibit expression of GCaMP5g and tdTomato following tamoxifen gavage (Sigma T5648; 120 µg/g body weight for four consecutive days). Mice that were homozygous for GCaMP5g, heterozygous for Cx3cr1CreER and either WT or homozygous for AppNL−G−F were used for imaging ≥ 21 days after the first tamoxifen dose (given at P85–P105).

Genotyping

Genotyping PCR primers and their final concentrations were as follows:

App WT (5’-tgtagatgagaacttaac-3’, 0.2 µM; 5’-atctcggaagtgaagatg-3’, 0.2 µM),

App mutated (5’-atctcggaagtgaatcta-3’, 0.4 µM; 5’-cgtataatgtatgctatacgaag-3’, 0.4 µM),

Polr2a GCaMP (5’-tagacacatgccaccaaacc-3’, 0.2 µM; 5’-tctctccagcaccataactcc-3’, 0.2 µM; 5’-gatcgataaaacacatgcgtca-3’, 0.2 µM),

Cx3cr1 Cre (5’-acgcccagactaatggtgac-3’, 0.2 µM; 5’-gttaatgacctgcagccaag-3’, 0.2 µM; 5’-agctcacgactgccttcttc-3’, 0.1 µM).

Following initial denaturation of tail snip DNA for 2 min at 95°C, amplification was performed for:

36 cycles of 30 s at 95°C, 40 s at 48°C, and 45 s at 72°C (App WT),

38 cycles of 30 s at 95°C, 40 s at 53°C, and 45 s at 72°C (App mutated),

36 cycles of 30 s at 95°C, 40 s at 58°C, and 45 s at 72°C (Polr2a GCaMP),

or 36 cycles of 30 s at 95°C, 40 s at 60°C, and 45 s at 72°C (Cx3cr1 Cre).

All PCR reactions were then completed with 4 min at 72°C. All PCR products were visualized on 2.5% Tris–borate EDTA gels, and product sizes for each allele in base pairs (bp) were as follows: 500 bp (App WT), 700 bp (App mutated), 395 bp (Pol2a GCaMP WT), 272 bp (Pol2a GCaMP mutated), 151 bp (Cx3cr1 WT), and 230 bp (Cx3cr1 mutated).

Solutions

Acute brain slices (250 µm, parasagittal) containing dorsal hippocampi were prepared on a Leica VT1200S vibratome with ice-cold slicing solution containing (in mM): 124 NaCl, 2.5 KCl, 26 NaHCO3, 1 NaH2PO4, 10 glucose, 1 CaCl2, 2 MgCl2 and 1 kynurenic acid. Osmolarity was adjusted to ~ 295 mOsm/kg and pH set to 7.4 when bubbled with 5% CO2/95% O2.

For live imaging, brain slices were incubated in artificial cerebrospinal fluid (aCSF) containing (in mM): 124 NaCl, 2.5 KCl, 26 NaHCO3, 1 NaH2PO4, 10 glucose, 2 CaCl2, 1 MgCl, and 0.1 Na-ascorbate. Osmolarity was adjusted to ~ 295 mOsm/kg and pH was set to 7.4 with NaOH. Solutions were oxygenated with 20% O2/5% CO2/75% N2.

Immunohistochemistry

Perfusion-fixed brain slices (250 µm, parasagittal) were prepared on a Leica VT1200S vibratome, and were permeabilised and blocked for 2 h at room temperature in blocking buffer (10% normal horse serum and 0.02% Triton X-100 in phosphate-buffered saline, PBS), followed by incubation with primary antibodies (mouse anti-human Aβ 82E1 [1:500, IBL 27725], rat anti-CD68 [1:250, BioRad MCA1957], and rabbit anti-Iba1 [1:500, Synaptic Systems 234003]) in blocking buffer for 12 h at 4°C with agitation. Following PBS washes, Alexa-conjugated secondary antibodies (Invitrogen) diluted 1:1000 in blocking buffer were applied for 4 h at room temperature with agitation. Brain slices were free-floating in the solution for all steps to allow antibody penetration from both sides of the slice. Lastly, slices were washed in PBS, incubated with DAPI (Invitrogen D1306) diluted 1:50,000 in PBS, and mounted.

Analysis of microglial morphology

Confocal z-stacks of Iba1-labelled CA1 stratum radiatum microglia encompassing the entire brain slice in 0.34 µm steps were imaged on a Zeiss LSM700 microscope with a Plan-Apochromat 63×/1.4 lens. Individual cells were 3D-reconstructed using the cell tracing tool Neuron Tracing v2.0 on Vaa3D (vaa3d.org), adjusting the background threshold for each image to obtain the optimal reconstruction. Only microglial cells with somata at ~ 50 µm from the slice surface were selected for analysis to avoid cells at depths without sufficient antibody penetration (~ 100 µm in our experiments) and cells too close to the surface (which are likely to have cut-off processes). After checking the reconstructions against the raw images, they were analysed using custom-written MATLAB software (github.com/AttwellLab/Microglia). Briefly, based on Sholl analysis, concentric spheres were drawn at 5 µm intervals from an analyst-established cell centre (the soma is assumed to have a 5 µm radius to avoid misassigning differences in Iba1 signal within the soma as representing processes [41]). Cell process branching was profiled based on distance from the soma to assess ramification. For analysis of the effects of amyloid β deposition, in AppNL−G−F brain slices, both cells at 82E1-labelled Aβ plaques and > 50 µm away from them were imaged. Analysis was performed with the researcher blind to genotype and Aβ signal.

Analysis of Aβ and lysosomal burden

Immunolabelled brain slices were imaged on a Zeiss AxioScan.Z1 scanner with a Plan-Apochromat 20×/0.8 M27 lens. After background subtraction of the images by channel (using a 10-pixel rolling ball average for CD68 and Iba1 and an 80-pixel rolling ball average for Aβ) and thresholding, masks for CA1 stratum radiatum microglia at and > 50 µm away from Aβ plaques (which were Iba1+/Aβ+ and Iba1+/Aβ–, respectively) were obtained and transposed to the binarised CD68 channel to calculate the percentage of the Iba1+ area that was also labelled for CD68 in microglia located at Aβ plaques and away from them.

To identify Aβ plaques in living brain tissue, a number of stains have been developed [6, 14, 32]. However, insufficient penetration into slices, labelling of only some Aβ plaque types or non-selective binding to other proteins can hinder the interpretation of results [18, 62, 71]. Thus, in this study we relied on microglial clustering around Aβ plaques (see Supplementary Fig. 1), reported previously in AppNL−G−F mice [13, 59], as an intrinsic cue for identification of Aβ plaques (and thus of microglia at and away from them).

Fluo-4 loading of microglia by patch-clamp electrophysiology

CA1 microglia in rat hippocampal brain slices were identified by acute labelling with isolectin-B4 coupled to Alexa-594 [41] and patch-clamped using a KCl-based intracellular solution containing (mM): 130 KCl, 4 NaCl, 10 HEPES, 0.01 BAPTA, 0.01 CaCl2, 10 Na-phosphocreatine, 2 MgATP, 0.5 Na2GTP, 0.1 Fluo-4 pentapotassium salt, adjusted to a final osmolarity of 285 ± 5 mOsm/kg and a pH of 7.2. Whole-cell recordings from microglia were obtained at a depth of > 40 μm below the slice surface using borosilicate glass pipettes with a tip resistance of 3.5–5 MΩ, resulting in access resistances of < 20 MΩ that were not compensated. The average resting membrane potential of cells was −35.7 ± 2.4 mV (n = 22). Throughout imaging, cells were voltage-clamped at −30 mV.

Microglial Ca2+ imaging

CA1 stratum radiatum microglia from Cx3cr1CreER × GCaMP5g-IRES-tdTomato and Cx3cr1CreER × GCaMP5g-IRES-tdTomato × AppNL−G−F mice, or patch-clamped microglia filled with Fluo-4, were imaged on a Zeiss LSM780 or LSM710 two-photon microscope with a Plan-Apochromat 20×/1.0 lens and a Spectraphysics Mai Tai DeepSee eHP Ti:Sapphire infrared laser tuned to 920 nm at < 2% of its maximum power, which corresponded to a maximum power under the objective of 5.5 mW. Fields of view of 106 µm × 106 µm were scanned with an overall acquisition time of 25 ms/frame (pixel size 0.21 µm, 1 µs pixel dwell time). Prior to longitudinal single-plane imaging (to reduce bleaching), a z-stack (encompassing the entire cell at 1 µm steps) was acquired to confirm the morphology of the imaged microglia as visualised by tdTomato. Acquisition rates for Fluo-4 and GCaMP5g-based Ca2+ imaging were 0.2 Hz and 1 Hz, respectively.

To assess spontaneous [Ca2+]i changes, microglial cells were imaged for 5 min (1 frame/s). Recordings were re-registered with StackReg in ImageJ/FIJI and ROIs were drawn around microglial somata or individual processes; only processes ≥ 1 µm in diameter were analysed to exclude filopodia [7]. In each ROI, Ca2+ transients were analysed with custom-written MATLAB code (github.com/AttwellLab/MyelinCalcium). A locally time-smoothened baseline (100 frames of smoothing time) was generated using a piecewise cubic Hermite interpolating polynomial fit, to account for potential drifts in baseline that may arise from subtle movements or volume changes of the specimen during the time course of recording. Ca2+ transients were then defined by a detection threshold for the fractional change of fluorescence (ΔF/F > 2.25 × standard deviation of baseline points that had ΔF/F < 0.10, to exclude contributions of Ca2+ transients to the baseline), confirmed using a minimal area threshold (∫ΔF/F dt > 0.15), and manually checked to exclude false positives [37]. An example of the baseline subtraction is shown in Supplementary Fig. 2.

To trigger damage-evoked [Ca2+]i changes, following a baseline recording of 20 s (1 frame/s), a focal laser lesion (6 µm radius, 177.3 µs pixel dwell time) was performed in an area > 30 µm from the cell of interest, with the laser tuned to 920 nm at 80% of its maximum power. For analysis, recordings were re-registered with StackReg in ImageJ/FIJI and ΔF/F for ROIs drawn around microglial somata were calculated as (Ft − Fo)/Fo, where Ft is the fluorescence intensity at each timepoint t and Fo is the average fluorescence of the pre-lesion baseline. Peak ΔF/F values were compared for statistical analysis. To help distinguish between intensity values, a multicolour (“Fire”) lookup table was applied in ImageJ/FIJI to generate the representative images provided.

Statistics

Quantitative data are presented throughout as mean ± standard error of the mean. Normality was assessed using the D'Agostino-Pearson test. Statistical significance (defined as p < 0.05) was assessed using Mann–Whitney tests (Fig. 5B–D, Suppl. Figure 1, Suppl. Figure 4), two-tailed Wilcoxon matched-pairs signed rank test (Fig. 3E), one-way analysis of variance (ANOVA) followed by Dunn’s (Figs. 2D, 3B–D, Suppl. Figure 3) or two-way ANOVA followed by Sidak’s post-hoc tests for individual comparisons (Figs. 2E, 4B, 5F). All statistical analyses were performed in Microsoft Excel 2016 and GraphPad Prism 8.

Results

Microglia show spontaneous and damage-evoked Ca2+ rises in situ

Early studies on microglia mainly relied on cultured cells [22, 48], where gene expression is drastically changed and cells often fail to develop the complex process ramification and surveillance seen in intact microglia [8]. To overcome this limitation, while preserving the option to apply pharmacological agents in a controlled manner to modulate function, we studied microglia in situ in acute brain slices.

First, for an initial characterization of microglial Ca2+ activity in brain slices, we filled hippocampal microglia from young rats with the Ca2+ indicator Fluo-4 via the patch pipette to detect spontaneous Ca2+ transients. This experiment showed that spontaneous Ca2+ transients occur more frequently in single microglial processes than in cell somata, with individual somata and individual processes showing a spontaneous frequency of 0.04 ± 0.02 and 0.23 ± 0.04 transients/min, respectively (Fig. 1A–C; see also [64] and Discussion; comparable data recorded using GCaMP5g in WT mice are presented in Fig. 3E below). To study evoked Ca2+ responses in microglia, focal laser lesions were induced as a well-established proxy to simulate responses of microglia to brain injury, similar to focal ATP application mimicking neuronal damage [15, 25, 41]. Focal lesions lead to acute increases in microglial [Ca2+]i at short latencies (0–5 s [19]). We found that, in Fluo-4–filled microglia laser lesions triggered rapid, transient Ca2+ signals (Fig. 1D, E).

Fig. 1
figure 1

Microglia exhibit spontaneous and local damage-evoked Ca2+ activity in acute brain slices. (A) Image of hippocampal slice with microglia (and vascular basement membrane) labelled with Alexa 594-conjugated isolectin B4, showing a patch-clamped microglial cell filled with the Ca2+ indicator Fluo-4. Somatic and process regions of interest (ROIs; soma and processes 1–6) are shown. (B) Microglial [Ca2+]i time course in the ROIs shown in (A). Red circles denote peaks detected at ≥ 2.25 standard deviations from the baseline. (C) Quantification of spontaneous Ca2+ transient frequency, showing a higher rate in single processes vs. somata (n = 5 cells; see also Fig. 3E for comparable date acquired using GCaMP5g). (D) As in (A) showing nearby site of laser damage (red dotted circle). (E) A laser lesion at the site shown in (D) evokes a transient rise of [Ca2+]i. (F) Scheme depicting the genomic changes in Cx3cr1CreER × GCaMP5g-IRES-tdTomato mice to drive expression of GCaMP5g and the morphological marker tdTomato in microglia induced by tamoxifen gavage. (G) Microglia in brain slices from tamoxifen-induced Cx3cr1CreER × GCaMP5g-IRES-tdTomato mice express tdTomato (red, left panel) and show spontaneous Ca2+ activity in different regions of interest (ROIs) over time, as measured by GCaMP5g fluorescence changes (ΔF/F, “Fire” scale). White arrowheads indicate regions showing spontaneous Ca2+ activity at different timepoints (time stamps on the figures are from the time scale on panel (H); fluorescence brightness has been linearly adjusted to make the transients more visible; two persistent spots in the images represent constitutive cell fluorescence). (H) Microglial [Ca2+]i time course in the ROIs shown in (G). Red circles denote peaks detected at ≥ 2.25 standard deviations from the baseline. (I) Representative cell showing that local laser damage (red dotted circle) evokes a rapid, transient [Ca2+]i rise in tdTomato-labelled microglia. Times indicate seconds from laser lesion. (J) Quantification of somatic [Ca2+]i levels over time (ΔF/F) showing their increase upon laser lesion (vertical dashed line)

Full size image
Fig. 2
figure 2

Proximity to Aβ plaques alters microglial morphology in AppNL−G−F mice. (A) Representative sagittal brain slice from a 4-month-old AppNL−G−F mouse showing Aβ plaques (82E1 antibody, green) and cell nuclei (DAPI, white) for reference. (B Quantification of the area covered by Aβ plaques in the neocortex (layer 2/3), dorsal hippocampus (CA1), cerebellum (molecular layer) and dorsal striatum of 4 AppNL−G−F mice. C Representative 3D-reconstructed App WT microglia, and AppNL−G−F microglia at or > 50 µm away from Aβ plaques in 4-month-old mice. (D–E) Sholl analysis-derived (D) total process length and (E) number of process branches at 5 μm increment distances from the soma, showing a sharp deramification of microglia at plaques (n = 53 App WT cells, 28 AppNL−G−F cells away from plaques and 42 AppNL−G−F cells at plaques from 3 mice each)

Full size image
Fig. 3
figure 3

Spontaneous Ca2+ activity in microglial somata is increased near Aβ plaques. (A) Representative traces showing spontaneous Ca2+ transients (red arrowheads) in microglial somata from App WT and AppNL−G−F mice at, or > 50 µm away from, Aβ plaques. (B) Quantification of spontaneous Ca2+ transient frequency, showing a higher rate near Aβ plaques (App WT: n = 27 somata from 5 mice; AppNL−G−F away from plaques: n = 56 somata from 8 mice; AppNL−G−F at plaques: n = 43 somata from 8 mice). (C) As (A), but for individual processes. (D) As (B), but for individual processes, showing no effect of Aβ status (App WT: n = 11 processes from 3 mice; AppNL−G−F away from plaques: n = 42 processes from 6 mice; AppNL−G−F at plaques: n = 34 processes from 6 mice). (E) Paired comparisons between individual process and somatic Ca2+ transient rate, showing that, for a given microglial cell, spontaneous Ca2+ activity per process was higher in processes in App WT (n = 9 cells from 3 mice; comparable data acquired with Fluo-4 are shown in Fig. 1C) and AppNL−G−F microglia away from plaques (n = 18 cells from 6 mice), but not in microglia at Aβ plaques (n = 19 cells from 6 mice) where the somatic frequency is similar to the frequency per process in WT mice

Full size image
Fig. 4
figure 4

Damage-evoked Ca2+ rises are reduced in microglia near Aβ plaques. (A) Representative microglia (tdTomato, red, lower panels) from App WT and AppNL−G−F mice at Aβ plaques or away (> 50 µm) from plaques. Laser damage-evoked Ca2+ rises (upper panels, shown on a “Fire” scale) are measured by GCaMP5g fluorescence changes (ΔF/F shown at peak). (B) Quantification of somatic [Ca2+]i levels over time (ΔF/F) showing their increase upon laser lesion (vertical dashed line). AppNL−G−F microglia, both at (n = 54 from 8 mice; p = 10–4) and away from Aβ plaques (n = 52 from 8 mice; p = 0.03), showed a significantly reduced damage-evoked [Ca2+]i rise compared with App WT microglia (n = 56 from 5 mice)

Full size image

To build on these initial findings, we employed mice that can be induced to express GCaMP5g under the control of the Cx3cr1 promoter, which enables detection of Ca2+ activity in microglia (Fig. 1F) with an affinity (Kd 0.46 μM, ~ 30-fold dynamic range [2]) similar to Fluo-4 (Kd 0.345 μM, ~ 100-fold dynamic range [20]), and allows the Ca2+ concentration to be monitored in several microglia simultaneously. While use of organic Ca2+ indicators is very common in previous studies, these may be harmful to cells, altering their membrane potential and metabolism and causing swelling [61]. By contrast, GECIs may cause less side effects [61] and allow the study of otherwise undisturbed microglia (although all Ca2+ indicators will intrinsically buffer [Ca2+]i to some extent [45]). As in young rats, we found that microglia in GCaMP5g-expressing mice showed detectable spontaneous Ca2+ activity (Fig. 1G, H) that was 3.5-fold higher in rate in individual cell processes than in somata (as quantified below) and also exhibited local laser lesion-evoked rapid (latency ~ 1 s), transient [Ca2+]i rises (Fig. 1I, J).

Proximity to Aβ plaques alters Ca2+ signalling in activated microglia

To test whether compartmentalised Ca2+ signalling is altered in brain pathology when microglia become activated, we used AppNL−G−F mice that recapitulate amyloid plaque deposition as a pathological hallmark in AD. Amyloid burden increases progressively with age in AppNL−G−F mice [13, 59]. For a reliable assessment of amyloid pathology (i.e., avoiding individual variability at onset but also allowing analysis of microglia both close to and away from plaques, which is difficult at late times because of the high Aβ plaque density), 4-month-old mice were used to study microglia in AD mice (a common age range used across studies with this model [1, 14, 52]). At this age, Aβ plaques can be detected in multiple brain regions apart from the cerebellum in perfusion-fixed slices from AppNL−G−F mice using an antibody against human Aβ (Fig. 2A, B).

We first assessed whether microglial morphology was affected by proximity to Aβ plaques, as has been shown previously in other non-knock-in models of AD, which have a differing development of disease parameters [4, 35, 54]. While microglia were highly ramified in WT mice, their overall process length and ramification were reduced in AppNL−G−F mice (Fig. 2C–E). The total process length per cell was reduced by 13.4% in AppNL−G−F microglia > 50 μm away from Aβ plaques compared with WT microglia (not significantly different, p = 0.07) and by 64.3% in AppNL−G−F microglia at Aβ plaques compared with WT microglia (p = 10–4; Fig. 2D). Similarly, process ramification was reduced closer to Aβ plaques, with AppNL−G−F microglia showing less branched processes and less branches near the soma compared with WT cells (Fig. 2E).

Microglia are known to readily phagocytose Aβ debris [39]. In our AppNL−G−F model, engulfment of Aβ inside microglial CD68-positive lysosomes was confirmed by confocal imaging (Supplementary Fig. 1A). CD68 coverage approximately doubled in hippocampal microglia at Aβ plaques compared with cells away from plaques (p = 10–4; Supplementary Fig. 1B, C), suggesting an increase in the phagocytic capacity of microglia in the presence of Aβ. This is consistent with the increase in CD68 expression reported in plaque-proximal microglia from other AD models and in microglia from patients with AD [26, 68, 72].

Together, these results suggest that microglial responses to Aβ depend considerably on the cells’ local environment, whereby the cells closer to Aβ plaques exhibit stronger changes in morphology and lysosomal content while those further away display an intermediate phenotype closer to WT microglia. Therefore, in subsequent experiments we analysed data from cells at Aβ plaques and away from plaques separately (using a distance threshold of > 50 µm from the nearest plaque edge as the defining criterion for plaque-distant microglia).

We next examined whether the microglial morphological changes in the AppNL−G−F mice are reflected in a specific Ca2+ activity pattern. We found that the frequency of spontaneous Ca2+ transients in microglia somata was dependent on Aβ pathology and proximity to plaques (Fig. 3A, B). While somata of WT microglia exhibited a frequency of 0.09 ± 0.03 transients/min, AppNL−G−F microglia located away from Aβ plaques and at Aβ plaques had somatic transient rates of 0.25 ± 0.05 transients/min (not significantly different from WT, p = 0.1) and 0.44 ± 0.06 transients/min (p = 10–4), respectively. Of note, this effect was not seen in microglial processes, where spontaneous Ca2+ events occurred independently of AppNL−G−F genotype; the rate was 0.55 ± 0.12 transients/min per process in WT microglia, 0.58 ± 0.07 transients/min in AppNL−G−F microglia away from Aβ plaques (p = 0.9) and 0.53 ± 0.05 transients/min in plaque-associated cells (p = 0.9; Fig. 3C, D). In all of these conditions the Ca2+ transient amplitudes showed no significant differences (Supplementary Fig. 3).

Because of the different Ca2+ activity in microglial somata vs. processes when averaged across cells, we also analysed a random subset of cells calculating their individual transient rates in the somata and processes in a paired manner (averaging over processes to obtain a single mean process value for each cell: Fig. 3E). In WT mice, microglia displayed a significantly higher rate of Ca2+ transients per process than per soma (3.5-fold higher; p = 0.04; note that GCaMP5g detects a ratio of the number of transients in single processes to that in somata similar to that for Fluo-4 in Fig. 1C). This was also seen in microglia away from Aβ plaques in AppNL−G−F mice (1.8-fold higher; p = 0.02). In contrast, microglia at Aβ plaques showed a similar rate of spontaneous Ca2+ events in their processes and somata (1.1-fold higher, p = 0.6).

We compared the surface area of a typical microglial soma in a WT mouse (assumed for simplicity to be spherical and with a radius r of 7.5 µm, giving a surface area of 4πr2 = 707 µm2), with that of the ROI used for a typical microglial process segment (assumed for simplicity to be cylindrical and typically with a length l of 15 µm and an approximate radius r of 0.75 µm, giving a surface area of 2πrl = 71 µm2). Thus, the process surface area is approximately tenfold smaller than that of the soma, and this fact cannot account for the higher transient frequency per process.

Next, we studied whether amyloid pathology altered lesion-evoked Ca2+ responses in microglia. While laser injury triggered a robust, rapid transient rise of [Ca2+]i in WT microglial somata, this response was impaired in the AppNL−G−F mice (Fig. 4). In AppNL−G−F microglia away from Aβ plaques, the peak Ca2+ response was reduced by 31.2% (p = 0.09), and in those at Aβ plaques the peak Ca2+ response was reduced by 77.9% (p = 10–4). Together, these data suggest that, while Aβ increases spontaneous Ca2+ activity in microglial somata, it renders microglia less able to respond to substances (such as ATP and ATP-derived metabolites) released acutely by laser-evoked brain damage. In both cases, microglia > 50 μm distant from Aβ plaques exhibited an intermediate phenotype between those at plaques and their WT counterparts.

Ageing alters microglial Ca2+ signalling differently from Aβ pathology

There is growing interest in the cellular processes of ageing and how these relate to neurodegenerative processes [3, 23, 31, 42]. Therefore, we analysed whether the changes in Ca2+ signalling that we found in microglia from AD mice may also occur in old (P300–310) WT mice.

While the frequency of spontaneous Ca2+ transients in old microglial somata was not significantly changed with age (40.0% lower than in microglia from young adult (P120–130) mice but not statistically significantly different; p = 0.46), microglial processes from old mice exhibited a significantly reduced frequency of Ca2+ transients (70.7% lower; p = 10–3; Figure 5A–D). The amplitude of the Ca2+ transients was not significantly affected (Supplementary Fig. 4). We also analysed lesion-evoked somatic Ca2+ responses and found that they were largely abolished in microglia from old mice, with the peak Ca2+ response reduced by 77.4% (p = 10–4) compared with microglia from young adult animals (Fig. 5E, F). These data suggest that normal ageing affects microglial Ca2+ activity differently from the pathological conditions of amyloid plaque deposition, reducing not only brain damage-evoked but also spontaneous Ca2+ activity in the cell processes.

Fig. 5
figure 5

Microglia from old mice exhibit reduced spontaneous and damage-evoked Ca2+ activity compared with microglia from young adult mice. (A) Representative traces showing spontaneous Ca2+ transients (red arrowheads) in microglial somata from P120–130 (young adult) and P300–310 (old) mice. (B) Quantification of spontaneous Ca2+ transient frequency, showing no difference between somata of microglia from young adult and old mice (n = 27 somata from 5 young adult mice and 30 somata from 3 old mice). (C) As (A), but for processes. (D) As (B), but for processes, showing a statistically significant decrease in spontaneous Ca2+ transient frequency per process with age (n = 11 processes from 3 young adult mice and 25 processes from 3 old mice). Data from App WT microglia in Fig. 3 are included as young adult microglia for comparison. (E) Representative microglia (tdTomato, red) from P120–130 (young adult) and P300–310 (old) mice. Laser damage-evoked Ca2+ rises are measured by GCaMP5g fluorescence changes (ΔF/F shown at peak, “Fire” scale).(F) Quantification of somatic [Ca2+]i levels over time (ΔF/F) showing their increase upon laser lesion (vertical dashed line). Microglia from old mice (n = 32 cells from 3 mice) showed a significantly reduced damage-evoked [Ca2+]i rise compared with microglia from young adult mice (n = 56 from 5 mice; p = 10–4). Data from App WT microglia in Fig. 4 are included as young adult microglia for comparison

Full size image

Discussion

Intracellular Ca2+ elevations are commonly considered an indicator of cellular activity in neurons and astrocytes; however, knowledge of microglial Ca2+ signalling is still sparse despite its importance in controlling key cellular functions. While spontaneous Ca2+ transients are extremely infrequent in acutely isolated or cultured microglia from adult mice [34], in vivo experiments on OGB-1–electroporated microglia showed that ~ 20% of them displayed transients over 15 min [19], and a later study using GECI-expressing microglia found that ~ 4% of cells were active over 20 min [55]. This suggests that the environment, type of preparation and Ca2+ sensor, and the activation state may all affect spontaneous Ca2+ activity in microglia. We found that the great majority of Ca2+ transients in healthy rat or mice microglia occurred in the cells’ processes, whether studied with the soluble Ca2+ sensor Fluo-4 or with GCaMP5g. This finding, which is in agreement with previous work using GECI-expressing mice [64], implies that microglial Ca2+ signalling is finely regulated across cell compartments (as in astrocytes [60]). Ca2+ transients may fulfil different functions in the soma vs. processes, which may be related to differential expression and localisation of Ca2+ signal-transducing membrane receptors and intracellular Ca2+-dependent organelles, such as the endoplasmic reticulum or the endolysosomal compartment.

Although microglia rarely generate Ca2+ transients in the healthy brain, elevations in [Ca2+]i are much more prominent under pathological conditions such as epileptiform activity, acute brain injury, neuroinflammation, and neurodegeneration [10, 19, 55, 63, 64]. In AD, the comparatively well characterised changes in microglial morphology and biomarker expression contrast with the still poorly understood changes of Ca2+ signalling in these cells, and how these vary between plaque-proximal and plaque-remote regions and in different parts of each microglial cell. In AppNL−G−F knock-in mice, we observed that microglial deramification and increased lysosomal burden depend critically on the cells’ proximity to Aβ plaques. This is consistent with studies of transgenic models of AD, showing that changes in microglial gene expression, morphology and electrophysiological properties [13, 54, 66], as well as synapse loss [33], occur predominantly near plaques. In agreement with this, our study revealed significant differences in spontaneous and damage-evoked Ca2+ responses in the AppNL−G−F model of AD between Aβ plaque-proximal and plaque-distant microglia, with the latter presenting an intermediate phenotype closer to that of WT cells. Notably, and adding to previous work, we found that these changes were also dependent on the subcellular compartment, revealing differential effects of Aβ on Ca2+ activity in processes compared with somata. While the Ca2+ activity of microglia has been previously studied in mice modelling AD [11], this was investigated with two different transgenic mouse strains pooled together (both overexpressing App) and using a dye indicator, OGB-1, electroporated into cortical microglial cells to estimate changes in [Ca2+]i. Here, we took advantage of GCaMP5g to visualise changes in [Ca2+]i, because intrinsic GECI expression may affect cell function less than using patch-clamping or electroporation to introduce a dye into the cell (although the affinity and dynamic range of the indicator used will also affect its usefulness).

We found that proximity to Aβ plaques increased the rate of spontaneous Ca2+ transients in the somata of hippocampal microglia. This is in line with results from Brawek et al. (2014), noting an increase in the fraction of microglia with Ca2+ activity in older, transgenic amyloid-depositing mice (although that study, which analysed cortical rather than hippocampal microglia, did not observe any changes in the rate of spontaneous Ca2+ transients in the cells that exhibited them [11]). Increased Ca2+ activity in microglial cells exposed to Aβ may be due to Aβ-evoked UTP/UDP release by stressed neurons acting on microglial metabotropic P2Y6 receptors [57] or Aβ-mediated mechanotransduction mediated via PIEZO1 Ca2+-permeable ion channels involved in amyloid clearance [29, 30]. However, while the Ca2+ transient rate in microglial somata depended on their proximity to Aβ plaques, the higher rate in microglial processes did not. Given the decrease in ramification of microglia at Aβ plaques, the increased Ca2+ frequency in somata vs. processes could also reflect the membrane and Ca2+ stores of the latter being pulled back into the soma, or a redistribution of receptors underlying Ca2+ responses [11]. Of note, the higher Ca2+ activity in the soma of microglia at Aβ plaques correlates with a large increase in CD68 immunoreactivity in these cells, consistent with an increased phagocytic rate (which has been linked to microglial Ca2+ activity [63]). Indeed, in microglia this marker of mature phagocytic lysosomes is localised mainly in somatic regions [27], where it may support an increased Ca2+-dependent phagocytic activity in plaque-associated microglia.

Another key finding from the present study is that microglia at Aβ plaques generated reduced responses to laser lesions. This suggests that cells activated by Aβ might be less able to mount large Ca2+ responses even though they exhibited an increased rate of spontaneous transients in the absence of acute injury. In support of this hypothesis, basal Ca2+ levels are higher in microglia isolated from post-mortem AD brains than in cells from non-dementing donors, but their ATP-evoked responses are smaller [44]. Similarly, in App-overexpressing transgenic mice, chemotaxis towards local ATP is also impaired in microglia near Aβ plaques [11], suggesting that injury-evoked Ca2+ rises might also be reduced in these cells. The decreased response to laser damage–evoked nucleotide release may reflect reduced expression of P2Y12 receptors as a result of microglial activation [36, 47].

Finally, it is important to assess how changes in Ca2+ responses in AD compare to other conditions, as diverse stimuli inducing microglial activation might converge by regulating [Ca2+]i as a core signalling element. The microglial genetic signature is similar between AD and old age [3, 23, 31, 42], and accumulation of Aβ (and tau) in the disease are thought to exacerbate processes involved in ageing, such as cellular senescence, Ca2+ dyshomeostasis, and inflammation [12]. Of note, our study revealed that amyloid plaque deposition and ageing affected the Ca2+ activity of hippocampal microglia differently. In AD mice, spontaneous Ca2+ activity was increased in the somata of microglia but not in their processes, compared with age-matched, non-AD mice. By contrast, spontaneous Ca2+ activity in aged mice was unchanged in microglial somata, but decreased in processes, compared with younger adults. Interestingly and in line with the lack of observed changes in somatic Ca2+ activity with age, microglial CD68 expression is only mildly altered between young and old mice [24, 43]. Of note, Brawek et al. (2014) suggested that the fraction of microglia showing spontaneous Ca2+ activity may be increased in ageing [11], and a subsequent study revealed a bell-shaped pattern with mice of 9–11 months of age (i.e., what we refer to as old age) showing a higher rate of Ca2+ transients than both 2–4- and 18–21-month-old animals [17]. This apparent difference with our results may arise from differences between the AD mouse models used, between the brain regions studied (neocortex vs. hippocampus) or in the Ca2+ detection method used (dye electroporation vs. GECI). Although each method has its own limitations, the high sensitivity of microglia to mechanical influences mediated by channels such as PIEZO1 and TRPV4 [5, 29, 58], with differences reported even between in vivo preparations (acute vs. chronic window preparations [64]), may be a critical factor to consider along with an age-dependent increase in cellular vulnerability and thus a reduced experimental resilience to mechanical perturbations. While reduced lesion-evoked Ca2+ responses are common to microglia in both physiological ageing and AD, our study suggests that clear differences exist at the subcellular level for spontaneous Ca2+ events, the details of which should be clarified by future work. Overall, our work highlights the multifaceted functional states that microglia can adopt depending on their environment (i.e., in a healthy brain, in AD whether located at Aβ plaques or away from them, and in the aged brain), and shows that Ca2+ activity is differently regulated on a subcellular level in each case.