Astroglial NF-kB contributes to white matter damage and cognitive impairment in a mouse model of vascular dementia

All experiments were approved by a local animal welfare committee. We used 10–14 week-old male GFAP-IkBα-dn mice [14] and their wildtype age-matched male littermates. Animals were housed in groups (two to five per cage) on a 12 h light/dark cycle with food and water available ad libitum.

Gold plated stainless steel microcoils (Sawane) were wrapped around both common carotid arteries in isoflurane-anaesthetised mice [13]. Informed by earlier studies that revealed either subtle [15, 16] or very pronounced changes [17], we modified the existing model by using custom-made coils with an internal diameter of 170 μm. Animals were anaesthetised with isoflurane (1.5 % v/v) and maintained on a heating pad (37 °C). The common carotid arteries were exposed through a midline cervical incision, and the microcoils were wrapped around both common carotid arteries. A 30 min interval was allowed between each microcoil insertion. Sham-treated animals underwent identical procedures except that coils were not implanted. Local lidocaine gel and intraperitoneal buprenorphine were applied as analgesics. Mice from each litter were randomised to either microcoil insertion or sham surgery. All experiments and subsequent data analysis were conducted in a blinded manner, with respect to genotype and treatment. Mice were excluded if visual damage to the carotid arteries or bleeding during surgery occurred. In addition, carotid arteries were inspected for patency at the end of experiments, and animals with occluded carotid arteries were also excluded.

Animals were anaesthetised with isoflurane (1.5 % v/v), and the skull was exposed through a midline incision. A cranial window was created and closed by gluing a glass coverslip onto the skull and adhering the skin to the coverslip rim. For imaging, mice were anaesthetised with isoflurane (1 % v/v) and fixed within a stereotaxic frame (Narishige). Brain perfusion was determined through a closed cranial window in anaesthetised mice using PeriCam PSI Laser Speckle Imager and PIMsoft software (Perimed). PcomA development was determined by transcardial perfusion with India ink (Pelikan; 10 % in water containing 10 % gelatine), followed by decapitation and fixation in 4 % paraformaldehype (PFA). PcomA was scored as: 0, absent; 1, hypoplastic; 2, truncal. A single PcomA score was calculated by averaging left and right scores.

MRI of isoflurane-anaesthetised mice was performed on an 11.7 T horizontal small-bore magnet (Biospec 117/16, Bruker) using a two-element transmit/receive proton (¹H) cryocoil (Bruker Biospin). Body temperature was maintained at 37 °C via an integrated water heating system. Respiratory rate was maintained at 70-80 cycles/min. Anatomical images were acquired using a rapid acquisition relaxation enhancement (RARE) T₂-weighted (T₂-w) sequence [echo time (TE) = 0.01 s; repetition time (TR) = 2.5 s; in-plane resolution 45 μm²], and a RARE T₁-weighted (T₁-w) sequence (TE/TR = 0.01 s/1 s; in-plane resolution 45 μm²). White matter integrity was assessed using an EPI DTI sequence (TE/TR 26/4000 ms; b value 1000 s/mm²; 60 diffusion directions; 5 b0 images; in-plane resolution 100 μm²).

Coronal sections (20 μm) were obtained from PFA-fixed hemispheres, blocked with 10 % normal goat serum (Vector Labs) and 0.3 % Triton-X100 (Sigma) in PBS for 1 h, and incubated with primary antibodies in PBS containing 5 % goat serum and 0.05 % Triton overnight at 4 °C. The following primary antibodies were used: rabbit anti-GFAP (1:1000; Z0334, Dako), rat anti-GFAP (1:1000; 13-0300, Invitrogen), rabbit anti-Iba1 (1:500; #019-19741, Wako), mouse anti-SMI312 (1:500, smi-312r, Covance), rabbit anti-p65 (1:100; sc-372, Santa Cruz; or #3033, Cell Signaling). Stainings were visualised using secondary antibodies from goat conjugated with Alexa Fluor 488, 596 or 647 (1:1000; Invitrogen; 2 h at room temperature). Nuclei were stained with Hoechst 33258 (1:1000; Invitrogen). Omission of primary antibodies served as negative controls. Confocal images were acquired using a confocal laser scanning microscope (LSM 700; Zeiss).

Mice were sacrificed, and the corpus callosum was dissected and stored in liquid nitrogen. Samples were homogenised in PBS containing protease/phosphatase inhibitor (Thermo Scientific) using a Precellys 24 homogeniser (Peqlab), and lysed in RIPA buffer (25 mM TRIS, 150 mM NaCl, 1 % NP-40, 0.5 % Na-deoxycholate, 0.1 % SDS, pH 7.2). Quantitative cytokine determination was performed using an electrochemiluminescence ELISA (Mouse ProInflammatory Ultra-Sensitive Kit, Meso Scale Discovery). Signals were measured on a SECTOR Imager 2400 reader (Meso Scale Discovery). For NF-kB measurements, nuclear extracts were created from whole brain homogenates (Nuclear Extraction Kit, #2900, Millipore), and NF-kB activity in nuclear extracts was determined with a commercial assay (EZ-TFA assay, #70-610, Millipore) according to the manufacturer’s instructions using a microplate reader (FLUOstar Omega, BMG Labtech). All measurements were performed in duplicates.

Male GFAP-IkBα-dn and wildtype mice have similar phenotypes in behavioural paradigms, including spatial learning and memory [18]. All tests were recorded and analysed using Ethovision XT9 (Noldus). The Y maze consisted of three arms (length, 34.5 cm; width, 8 cm; height, 12.5 cm) extruding at equal angles from a central platform. The Y Maze test was modified from previous designs [19]. During the acquisition trial, one arm was closed and mice were placed at the end of an open arm (chosen at random) and allowed to explore the two open arms for 5 min. After 24 h, mice were returned to the maze with all 3 arms open; the start point corresponding to the individual start point of each mouse in the acquisition trial, and allowed to explore for 5 min. Arm preference was quantified as the ratio between the time spent in one arm and the time spent in all arms (preferency index). For open field exploration, mice were placed in the centre of a dimly lit (20–30 lx) arena. The area was virtually divided into a centre and an area ≤5 cm to the walls (surround). General locomotor activity was quantified as total distance moved and velocity. The level of anxiety was estimated by the percentage of time spent in the surround.

Immunohistochemical images were imported into ImageJ 1.50 (W. Rasband, NIH), smoothed and despeckled using a median filter, and converted to an 8-bit grey level image. Following contrast enhancement (0.4 % saturated pixels), images were binarised with a threshold defined as the mean background intensity plus the standard deviation of background intensity multiplied by 3. Astrocyte area coverage and diameter, and micoglial cell count were determined automatically by ImageJ. For astrocyte process length, binarised images were skeletonised and analysed using the Analyze Skeleton (2D/3D) plugin for ImageJ. These semi-automatic quantifications were compared with a subset of data (n = 3 from each group) that was quantified by manual analysis; no significant difference was detected between the analysis methods (data not shown). GFAP-positive astrocytes were considered positive for p65 when they showed a spherical somatic p65 staining that was smaller than the GFAP-positive cell soma and overlapped with the Hoechst 33258 nuclear counterstain. Mean pixel intensity of SMI312 staining in 8-bit grey level images was taken as a measure of axonal integrity. Demyelination was analysed in LFB-stained sections using a semi-quantitative scoring system (0 = no demyelination; 1 = visible fiber disarrangement; 2 = vacuoles and focal demyelination; 3 = widespread/continuous demyelination). Examples are provided in Additional file 1: Figure S1. Analysis of the corpus callosum was performed within pre-defined areas (0.75 mm and 2.0 mm from the midline). All measurements were taken in triplicates, and the values averaged. Laser speckle contrast data were down-sampled, exported into Excel (Microsoft), and normalised to perfusion levels before microcoil implantation. Paravision 5.1 (Bruker) was used to generate the parametric images used for white matter analysis. Fractional anisotropy (FA), apparent diffusion coefficient (ADC), axial diffusivity (AD) and radial diffusivity (RD) were measured. Direction-encoded colour (DEC) maps based on the primary eigenvectors were generated using in-house software. The colour determines the principal direction (X, Y or Z) of the primary eigenvector exhibiting the largest diffusion coefficient from the measured diffusion tensor: red designates medial to lateral orientation, blue represents dorsal to ventral orientation and green exhibits the rostral to caudal orientation. Regions of interest (ROIs) including the corpus callosum, external capsule and internal capsule were manually drawn onto the diffusion tensor images, and verified by comparison with the corresponding T₂-w image and a stereotaxic mouse brain atlas [20]. Heatmaps were generated by a) registering the FA images to a reference image within the respective cohort and b) setting landmarks on trace-weighted diffusion images and transferring them onto their respective FA images. Registration was performed using the bUnwarp plugin (version 2.6.3) for ImageJ.

We modified a model of bilateral common carotid artery stenosis (BCAS) [13] by placing microcoils around both common carotid arteries to induce chronic cerebral hypoperfusion in mice (Fig. 1a). Longitudinal laser speckle imaging demonstrated a reduction in cerebral blood flow by 32 ± 9 % 24 h and 29 ± 8 % 7 days following BCAS induction, respectively, whereas blood flow remained unaltered in sham-treated mice (Fig. 1b-c). Separate cohorts of mice, perfused either 1 week or 6 weeks after BCAS induction, exhibited white matter demyelination (Fig. 1d), reactive astrogliosis (Fig. 1e-g), microglial activation (Fig. 1h), and delayed degradation of white matter axonal integrity (Fig. 1i). Reactive astrogliosis was evident by increased area coverage, cell size and process length, although the latter parameter is also influenced by GFAP redistribution induced by astrogliosis [12]. The number of GFAP-positive astrocytes in the corpus callosum did not differ between the groups, owing to the high fraction of GFAP-positive astrocytes already present in the white matter (148 ± 12/mm² and 159 ± 10/mm², respectively; n = 6 for each group; p > 0.05, Mann–Whitney test). No difference was seen in the number of NeuN-positive neuronal cell bodies in the cortex (Sham, 1987 ± 211/mm²; BCAS, 2027 ± 274; n = 6 for each group; p > 0.05, Mann–Whitney test). There was no change in the rostro-caudal diameter of the corpus callosum, measured in coronal sections, in mice subjected to BCAS compared with sham-treated mice (211 ± 26 μm vs. 201 ± 32 μm; n = 6 for each group; p > 0.05, Mann–Whitney test). Representative examples are shown in the Additional file 2: Figure S2. Changes comparable to those observed in the corpus callosum were detected in the internal capsule and external capsule (Additional file 3: Figure S3).

Moreover, we found that the concentrations of the pro-inflammatory cytokines, interleukin-1 and interleukin-6, were increased in homogenates of the corpus callosum isolated from mice subjected to BCAS (Fig. 1j).

To investigate spatial long-term memory, we used a two-trial Y maze spatial memory task [19]. In the first trial (acquisition), mice were allowed to visit two arms of the maze, while the third arm was blocked. In the second trial (retrieval) performed 24 h later, mice had access to all three arms. As expected, given the natural tendency of rodents to explore new environments, sham-treated mice showed a strong preference for the novel arm in the retrieval trial (Fig. 2a-b). By contrast, mice subjected to BCAS showed no preference for the novel arm (Fig. 2b), indicating decreased spatial memory function. Spontaneous locomotor activity and the level of anxiety, which were investigated in the open field test, did not differ between the groups (Fig. 2c-e) indicating that the deficits of hypoperfused mice in the Y maze test were due to impaired memory function and not to a decline in motor function or general behavioural alterations.

One consequence of the expression of pro-inflammatory cytokines such as IL-1 is the activation of the nuclear factor (NF)-kB signaling cascade in reactive astrocytes [21]. Under control conditions, inactive NF-kB is localised to the cytoplasm by its interaction with inhibitory IkB proteins such as IkBα. Dissociation from IkB and subsequent nuclear translocation of NF-kB is initiated by the degradation of IkB through cytokine-induced IkB kinase activation [22]. Once activated, NF-kB acts as an important regulator of reactive gliosis, inflammation, myelination and neuronal survival in different disease models [12, 14, 23].

Therefore, we investigated the occurrence of activated nuclear NF-kB by immunohistochemistry against the active p65 (RelA) subunit of NF-kB. We found that 27.2 ± 7.1 % of all GFAP-positive white matter reactive astrocytes in hypoperfused mice were positive for nuclear (i.e. active) p65 (n = 7 mice; Fig. 3a). By contrast, only sparse and cytosolic (i.e. inactive) p65 labeling was visible in astrocytes from sham-treated mice (3.8 ± 1.8 %, n = 6 mice; Fig. 3b). White matter microglia and neurons were only very rarely positive for nuclear p65 in chronic hypoxic mice (3.4 ± 0.9 % of Iba1-positive microglia and 0.8 ± 0.3 % of NeuN-positive neurons, respectively). Hoechst-positive and GFAP-negative punctuate p65 signals were rare (7.3 ± 2.4 % of all Hoechst-positive p65 signals). These data indicate that the majority of white matter NF-kB activation in our model occurs in astrocytes. To quantify NF-kB activity, we used a chemiluminescent assay measuring the amount of p65 bound to the flanked DNA binding consensus sequence for NF-kB in nuclear extracts from brain homogenates. Consistent with the immunohistochemistry, this assay demonstrated strong NF-kB activation in the brains of mice subjected to BCAS compared with sham (Fig. 3c). These data indicate that reactive astrogliosis and chronic inflammation of the white matter in hypoperfused mice induces NF-kB activation in reactive astrocytes.

To investigate the consequences of astroglial NF-kB activation, we used transgenic GFAP-IkBα-dn mice in which astroglial NF-kB activation is selectively inhibited by targeting the overexpression of a dominant negative form of IkBα to astrocytes [14]. Earlier studies have shown that the transgene is selectively active in astrocytes, while NF-kB activity remains unchanged in neurons [14]. Following BCAS, we found that nuclear p65 was absent from reactive white matter astrocytes (Fig. 3d) and that NF-kB activity was similar to sham-treated mice (Fig. 3c).

The level and time course of cerebral blood flow reduction (Fig. 4a) and the plasticity of the posterior communicating artery (PcomA; Fig. 4b) were similar in GFAP-IkBα-dn and wildtype mice (Laser speckle blood flow levels from each individual day of measurement were compared between GFAP-IkBα-dn and wildtype mice; p > 0.05 for each comparison, Mann–Whitney test). Thus, differences between these groups are unlikely to be due to variability in the degree of hypoperfusion or collateralisation between the anterior and posterior cerebral artery circulation.

Interestingly, astrogliosis was detectable in GFAP-IkBα-dn mice following BCAS but reactive microglial activation and demyelination in white matter structures were significantly attenuated (Fig. 4c-e). Furthermore, axonal integrity within the corpus callosum was preserved in GFAP-IkBα-dn mice compared with sham-treated controls (Fig. 4f). Importantly, GFAP-IkBα-dn mice subjected to hypoperfusion spent significantly more time in the novel arm in the retrieval trial of the Y maze test (Fig. 4g), indicating preserved spatial memory function as a consequence of decreased NF-kB signaling. The open field test revealed no differences in distance moved, velocity or time spent in the surround between hypoperfused GFAP-IkBα-dn, sham-treated GFAP-IkBα-dn and wildtype mice (Fig. 4h-j). Overall, these data show that NF-kB activation in reactive astrocytes contributes to neuropathological damage and behavioural changes in our VaD model.

Magnetic resonance diffusion tensor imaging (DTI) is a key technology that enables the non-invasive study of structural connectivity of white matter tracts, and is an invaluable tool for the diagnosis and longitudinal study of patients with vascular dementia [24, 25]. Therefore, we investigated whether the preservation of axonal integrity observed in GFAP-IkBα-dn mice was detectable by DTI MRI. DTI was performed on hypoperfused or sham-treated wildtype and GFAP-IkBα-dn mice 6 weeks post surgery. Coronal images from brain areas encompassing the major white matter tracts (corpus callosum, internal capsule and external capsule) were acquired, and fractional anisotropy (FA), apparent diffusion coefficient (ADC), axial (AD) and radial diffusivity (RD) were measured. Representative direction-encoded colour (DEC) maps are shown in Fig. 5a. We detected significant changes in white matter FA, ADC, AD and RD values in wildtype mice subjected to BCAS compared with sham-treated mice (Fig. 5b-e and Additional file 4: Figure S4). This indicates that hypoperfusion induced deterioration in white matter integrity consistent with the observed histopathological changes. By contrast, GFAP-IkBα-dn mice subjected to BCAS mice did not exhibit a difference in the diffusivity measures when compared with sham-treated GFAP-IkBα-dn mice, underlining the strong attenuation in pathology compared with wildtype animals (Fig. 5b-e). T₁- and T₂-weighted imaging revealed no signs of overt anatomical damage in any group (Additional file 5: Figure S5).