Microglial motility in Alzheimer’s disease and after Aβ42 immunotherapy: a human post-mortem study

Brain tissue from 76 donors was sourced from the South West Dementia Brain Bank comprising 44 AD cases and 32 controls. Alzheimer’s cases had a clinical diagnosis of sporadic AD made during life and satisfied post-mortem neuropathological consensus criteria for AD [20]. Cases with any other significant brain pathologies such as stroke, tumour, or traumatic brain injury were excluded from the study. Controls were aged-matched cases, with no history of neurological or psychiatric disease or symptoms of cognitive impairment. Additionally, 16 immunised AD subjects (iAD) who participated in the AN1792 clinical trial were also included in the study. The characteristics of the groups are presented in Table 1 and individual details in the Additional file 1: Tables S1, S2 and S3.

Cerebral cortex from the inferior parietal lobule was investigated in all cases. This is one of the areas recommended for neuropathological assessment by the CERAD (Consortium to Establish a registry for Alzheimer’s Disease) [11] as it is significantly affected by AD pathology. Formalin-fixed paraffin embedded tissue was used for the immunodetection of neuropathological and neuroinflammatory markers. Frozen tissue, available for 31 controls, 35 AD cases and 11 iAD cases was used for detection of inflammation-related proteins by Meso Scale Discovery multiplex assays.

Immunohistochemistry was performed on 4 μm paraffin-embedded sections, targeting the markers of AD pathology pan-Aβ (clone 4G8, BioLegend), Aβ42 (clone 21F12, Elan Pharmaceuticals Ltd) and the phosphorylated (p) tau (clone AT8, ThermoScientific) protein, as well as the microglial motility-related proteins Iba1 (rabbit polyclonal, Wako Chemicals), CFL1 (polyclonal rabbit, ThermoScientific), CORO1A (rabbit polyclonal, LifeSpan Biosciences) and P2RY12 (rabbit polyclonal, Sigma Aldrich). Bound antibodies were visualized using the avidin–biotin–peroxidase complex method (Vectastain Elite, Vector Laboratories) with 3,3′-diaminobenzidine as chromogen and 0.05% hydrogen peroxide as substrate (Vector Laboratories). All sections were counterstained with haematoxylin, then dehydrated and mounted in Pertex (Histolab Products AB). The staining was performed in two batches with each batch containing cases from all groups (Control, AD, iAD). All experiments included a negative control slide incubated in buffer with no primary antibody and a positive control slide containing a specific tissue type known to express the protein of interest (e.g. tonsil).

Quantification was blinded to the case designation and performed on the grey matter in the same sulcus of the inferior parietal lobule for all cases. For each antibody, the slides were scanned together at a magnification of × 20 in an automated slide scanner microscope Olympus VS110 (Olympus America Inc.) and visualised with Olympus VS-Desktop software using a specialised add-in that allows the extraction of the region of interest (ROI) of 0.25mm². For each slide, 30 ROIs were selected in a zig-zag pattern within the grey matter, ensuring coverage of all cortical layers. Extracted ROIs were analysed with the ImageJ software (version 1.49u, Wayne Rasband, NIH, USA) using an automated macro specific to each antibody. The area fraction labelled by the antibody in each ROI was obtained by quantifying the presence or absence of the staining in each pixel, expressed as protein load (%).

Inflammatory proteins were measured on the V-Plex Meso Scale Discovery (MSD) electrochemiluminescence multi-spot assay platform (MesoScale Diagnostics, Rockville USA). 100 mg of frozen grey matter were homogenised at a tissue concentration of 20% w/v in RIPA lysis buffer (Thermo Fisher Scientific), supplemented with protease inhibitors (Sigma Aldrich) and phosphatase inhibitors (Thermo Fisher Scientific). Total protein concentration in the supernatant was measured by BCA Protein Assay Kit (Thermo Fisher Scientific). 25 μl of brain homogenate (1:2 dilution) were used for the V-PLEX human proinflammatory panel 1 and cytokine panel 1, and each plate was read on a Meso Quickplex SQ120 (Meso Scale Discovery) according to manufacturer’s instructions. Absolute target protein levels (pg/ml) were obtained and normalised with respect to the total protein concentration, as obtained by the BCA assay. Three frozen homogenates from a multiple sclerosis brain containing either chronic inactive, acute or chronic active lesions were included as positive controls.

Statistical analysis was performed using the IBM SPSS v24 statistical software package (SPSS Inc. Chicago IL). The normality of distribution of each marker across the groups was assessed by the Shapiro-Wilk test. Data were not normally distributed, and thus non-parametric tests were used. Comparisons among study groups were performed using the Kruskal-Wallis test. The Spearman’s test was performed to assess correlations between variables – to explore the relationship (i) between microglial motility proteins and key features of AD pathology (Aβ and tau loads), and (ii) between the microglial-motility proteins. To account for multiple testing, the Benjamini-Hochberg procedure to control for the false discovery rate (FDR) was used as post-hoc correction. For all tests, an adjusted P value < 0.05 was considered significant.

Aβ and ptau immunoreactivity was quantified in order to characterise the cases with respect to AD pathology and to permit correlations with the microglial motility-related proteins. A significantly increased pan-Aβ load was observed in AD compared to control cases (P < 0.001), whereas it was significantly decreased in the iAD group compared to AD cases (P < 0.001), as expected. No significant difference was observed between control and iAD cases (Fig. 1a). Aβ42 load was significantly increased in AD vs control cases (P = 0.006) and decreased in iAD vs AD cases (P < 0.001), as expected. Interestingly, Aβ42 load, was also significantly decreased in iAD compared to control cases (P = 0.021) (Fig. 1b). Ptau was significantly increased in AD vs control cases (P < 0.001) and decreased in iAD vs AD cases (P = 0.026), with no difference between the iAD and control groups (Fig. 1c).

Immunohistochemistry for the four proteins related to microglial motility was investigated in the cortical grey matter. Iba1, CFL1 and CORO1A immunolabelled microglia and perivascular macrophages and in AD microglial clusters around amyloid plaques. While P2RY12 was specific to microglia with no labelling of perivascular macrophages (Fig. 2).

Quantification of these four proteins showed no significant differences between AD and the control group: Iba1: control = 0.34% (IQR: 0.08–0.63%) vs AD = 0.29%(IQR: 0.13–0.68%), P = 0.788; CFL1: control = 0.15% (IQR: 0.03–0.038%) vs AD = 0.08% (IQR: 0.04–0.022%), P = 0.314; P2RY12: control = 0.00% (IQR: 0.00–0.55%) vs AD = 0.09% (IQR: 0.00–1.09%), P = 0.290; CORO1A: control = 0.21% (IQR: 0.10–0.52%) vs AD = 0.17% (IQR: 0.03–0.33%), P = 0.406. However, there was a significant increase for Iba1 and P2RY12 in iAD compared to AD cases: Iba1: AD = 0.29% (IQR: 0.13–0.68%) vs iAD = 0.87% (IQR: 0.40–1.11%), P = 0.036; P2RY12: AD = 0.09% (IQR: 0.00–1.09%) vs iAD = 0.60% (IQR: 0.50–1.08%), P = 0.041 (Fig. 3). No significant difference was found in CFL1 and CORO1A loads between the AD and iAD groups: CFL1: AD = 0.08% (IQR: 0.04–0.022%) vs iAD 0.13% (IQR: 0.06–0.41%), P = 0.314; CORO1A: AD = 0.17% (IQR: 0.03–0.33%) vs iAD = 0.20% (IQR: 0.03–0.61%), P = 0.406).

We then explored whether the expression of the microglial motility-related proteins was correlated with the proteins that aggregate in AD. In the control group, correlations were detected for Iba1 and P2RY12 with (i) pan-Aβ (r_s = 0.440, P = 0.046 and r_s = 0.443, P = 0.042; respectively) and (ii) Aβ42 (r_s = 0.691, P < 0.001 and r_s = 0.629, P < 0.001; respectively) (Table 2). No correlations were observed with pan-Aβ and Aβ42 in the AD or iAD groups, and with ptau for the three groups.

We also assessed whether the different mechanisms involved in motility (actin-related or purinergic signalling) were related. In the control group, Iba1 was correlated with CFL1 (r_s = 0.502, P = 0.025) and P2RY12 (r_s = 0.572, P = 0.005). In the AD cases, Iba1 was related to CFL1 (r_s = 0.559, P < 0.001), CORO1A (r_s = 0.436, P = 0.014) and P2RY12 (r_s = 0.558, P < 0.001). CFL1 was also correlated with CORO1A (r_s = 0.530, P = 0.002) and P2RY12 (r_s = 0.542, P = 0.001), and CORO1A with P2RY12 (r_s = 0.357; P = 0.049). No correlation was detected in the iAD group (Table 3).

To explore the relationship between microglial motility and the neuroinflammatory environment, the MSD platform was used to measure the levels of IFNγ, IL1β, IL2, IL4, IL6, IL8, IL10, IL12p70, IL13, TNFα, IL1α, IL5, IL7, IL12/IL23p40, IL15, IL16, IL17A, GM-CSF, TNFβ and VEGF in the three groups. Significant increases in the AD group compared to the control group were observed for: IFNγ (P < 0.001), IL10 (P = 0.003), IL12p70 (P < 0.001), IL13 (P < 0.001), IL2 (P = 0.002), IL4 (P = 0.007), TNFα (P < 0.001), IL15 (P = 0.024), IL16 (P = 0.020) and TNFβ (P = 0.009). In the iAD group compared to AD, IL13 (P = 0.017), IL8 (P = 0.023) and VEGF (P < 0.001) were significantly increased; whereas IL7 was significantly decreased (P = 0.012). Of note, of the proteins that were significantly different between AD and iAD, IL13 and VEGF were also significantly increased with respect to the control levels (IL13: P < 0.001; VEGF: P = 0.012), while IL7 was significantly decreased (P = 0.003) (Table 4).

As expected, the expression of the key features of AD neuropathology, Aβ and ptau, was increased in AD. The effect of the immunotherapy on AD has been previously reported by us [4, 32, 33, 42]. However in this study, we were able to compare the effects of the treatment with both controls and unimmunised AD. Interestingly, we observed that Aβ42 immunotherapy reduced the overall Aβ levels to those of controls, indicating an effect not restricted to the Aβ42 form. This is consistent with our studies on the same immunised cohort using antibodies specific for the Aβ40, Aβ42 and Aβ43 species and reporting significant decrease of the three species [21, 32].

Of note, its effect on the Aβ42 form seems to be more pronounced, with a reduction to significantly lower levels than those of controls. This highlights a possible higher affinity of the immunisation towards the Aβ42 species, perhaps due to the fact that full-length Aβ42 was the immunogen originally used in the AN1792 trial, even though earlier studies showed that the antibodies produced by the patients were primarily N-terminal specific [24]. This implies that the antibodies are able to target all different species of Aβ. The ptau load was reduced to the level of the controls, confirming the effect of Aβ-immunotherapy on the tau pathology [4, 33]. Indeed, our previous studies on the immunised cohort showed that ptau load was reduced after Aβ immunisation, in association with a lower expression of the tau phosphorylating enzyme GSK-3b [2]. Ptau load is mainly constituted of tau-containing dystrophic neurite clusters and ptau in neuronal processes [4], rather than tangles. Nevertheless, despite the decrease in ptau load as assessed by quantification in specific localised neuroanatomical regions of cortex, most of the iAD subjects still had progressed to Braak stages V or VI, as evaluated by the anatomical distribution of neurofibrillary tangles throughout the brain areas. We interpret these findings as indicating that localised removal of Aβ following Aβ immunotherapy results in a corresponding removal of ptau in that location, consistent with the amyloid cascade hypothesis. However, it does not stop the spread of ptau through the brain, consistent with the evidence for prion-like spread of ptau [4, 33].

In AD, expression of all four motility-related proteins examined was not modified compared with controls, consistent with another post-mortem AD study [45]. In this paper, they evaluated expression of Iba1 and P2RY12 and also found no difference in the expression of both proteins between the control and AD groups. Interestingly, in that study, when the AD cases were split according to the Braak stage, a decrease for both Iba1 and P2RY12 proteins was observed in the late-stage of the disease (Braak stage V/VI) but only in the hippocampal area, a region heavily affected by neuronal loss and tau accumulation [50]. In our study, even though most of the AD cases correspond to Braak stages V and VI, we did not observe decrease in these proteins, which could be indicative of regional differences between the hippocampus and the cerebral cortex due to the exacerbated degree of neuronal death and tau aggregation in the hippocampal region in late stages of the disease. With respect to CFL1, our observation confirms previous reports of no difference in CFL1 expression in AD [39]. The presence of CORO1A has been observed in human microglia [1], but to our knowledge has never been assessed in AD.

After Aβ immunotherapy in AD, a 3-fold increase in Iba1 expression and a 6.6-fold increase in P2RY12 expression was observed. Data on anti-Aβ antibody titres measured in the blood during life some years earlier were available [57] and did not correlate with the microglial analyses. Iba1 is expressed by microglia in both physiological and disease conditions, suggesting that Iba1 participates in both baseline and directed motility. In vitro studies have characterised the role of Iba1 in relation to membrane ruffling, and the formation of phagocytic cups is Iba1-dependent [35]. Therefore, the higher Iba1 expression observed is in accordance with the reported increased phagocytic activity associated with Aβ clearance [58] as the consequence of immunotherapy stimulating microglial phagocytosis towards Aβ. Thus, after immunotherapy, Iba1 expression is likely to reflect the motility changes associated with phagocytosis. Similarly to Iba1, P2RY12 is detected on all microglial cells and a role of P2RY12 in directed motility is also well-documented. The release of ATP to the extracellular space, caused by cell death, and its subsequent hydrolysis to ADP activates P2RY12 [48]. Studies have shown that microglia from mice lacking P2RY12 expression show normal baseline motility but impaired directed motility [15, 26]. Moreover, it was observed that P2RY12 expression in ramified microglia was markedly decreased as cells turn to the reactive/amoeboid form [15, 38]. This suggests a role of P2RY12 in directed motility taking place at an early stage of the cell activation. Thus, P2RY12 expression after immunotherapy might suggest an increase both in baseline and directed motility. This is also consistent with the neuroinflammatory environment observed in the immunised group, which is slightly leaning towards a more anti-inflammatory profile, considering evidence that P2RY12 is downregulated under pro-inflammatory conditions and upregulated in an anti-inflammatory context [22, 30]. Of note, a previous study by our group in the immunised group showed that neuronal loss is correlated with Iba1-positive microglia and phagocytic CD68-positive microglia associated with pro-apoptotic neurons, but without accelerated cognitive decline compared to non-immunised AD patients [36]. This suggests that immunotherapy may enhance microglial elimination of damaged and dysfunctional neurons. Considering the role of P2RY12 as a purinergic receptor upregulated with the release of ATP that occurs during cell damage, it could be hypothesized that the increased P2RY12 expression in the immunised cohort reflects microglial increased response, as a result of the treatment, towards damaged neurons.

Relationships between the microglial motility-related proteins and Aβ and ptau, key pathological features of AD, were explored. In controls, Iba1 and P2RY12 were associated with Aβ, with these associations lost in AD and not restored by the immunotherapy. It is recognised that with ageing, Aβ accumulates independently of the dementia status [8]. Therefore, the relationship of Iba1 and P2RY12 with Aβ in the absence of dementia may reflect a healthy chemotactic microglial response towards Aβ, via the directed motility mechanism, as previously reported in the Cognitive Function and Ageing Studies group [28]. In AD, these correlations were lost, possibly due to either (i) Aβ over-accumulation to a level at which microglia are overwhelmed and no longer capable of responding accordingly, (ii) microglia attracted to plaques and been immobilised there [57], or (iii) microglia being “distracted” by the tau pathology [28]. However, the absence of correlation of the motility-related proteins with Aβ or ptau in the AD group is in favour of a microglial response towards other pathological changes, such as the synaptic/neuronal loss. Of note, the correlations were not restored after immunotherapy despite Aβ levels as lower as in the controls and this despite the increased Iba1 and P2RY12 expression. No relationship was observed with CFL1 or CORO1A.

Since no correlations were found between the microglial motility-related markers and ptau, this implies that the motility response driven by the proteins studied is independent of tau, or at least they do not directly interact with tau.

Correlations between the microglial motility-related proteins were also investigated. In controls, Iba1 correlated with CFL1 and P2RY12, reflecting an interaction between the actin-related motility mechanisms and the purinergic signalling pathway in physiological conditions. Interestingly, in AD all markers were related to each other, implying that the pathogenesis elicited a coordinated response of the microglial motility proteins as part of the pathogenesis spreading. In the immunised group, no relation was observed despite the increased expression of the homeostatic markers Iba1 and P2RY12. This implies that immunotherapy did not restore the physiological profile of microglial protein interactions.

Inflammation-related proteins were evaluated to characterise the microenvironment. In AD, 10 were increased, including the pro-inflammatory proteins IFNγ, TNFα, TNFβ, IL12p70, IL2, IL15 and IL16, and the anti-inflammatory proteins IL4, IL10 and IL13. This reflects that both pathways of activation coexist in human AD as previously reported [9, 55].

After Aβ immunotherapy, only four markers were modified when compared with unimmunised AD, with increased expression of the pro-inflammatory chemokine IL8, the anti-inflammatory cytokine IL13, the vascular endothelial growth factor (VEGF), and a decrease in the pro-inflammatory cytokine IL7. IL13 and IL7 expression tends to reflect a slightly less pro--inflammatory environment after the treatment. VEGF expression was the most affected by immunotherapy with a 5-fold increase. This might be the result of (i) increased soluble Aβ species following plaque solubilisation [25] leading to neurotoxicity and angiogenesis via a VEGF-mediated mechanism [7], or (ii) reflected the vascular side effects reported after immunotherapy, known as amyloid-related imaging abnormalities (ARIA) including microhaemorrhages and oedema [5, 43]. Interestingly, both VEGF and IL8 are involved in angiogenesis [27]. It is worth noting that, from the four markers that differed between AD and controls, IL13 and VEGF were also significantly increased with respect to controls, while IL7 was decreased with respect to control, emphasizing that Aβ-immunotherapy, even though inducing a slightly more anti-inflammatory environment, did not restore the normal state.