Tau depletion prevents progressive blood-brain barrier damage in a mouse model of tauopathy

The first indication that there could be BBB damage in the rTg4510 mouse model was the presence of intense background staining when using anti-mouse IgG secondary antibodies on aged tissue, similar to what has previously been shown following AAV delivery of P301L tau in wild-type mice [19,20]. This anti-IgG reactivity was only observed in aged rTg4510 and not in age-matched wild-type littermates or younger rTg4510 mice. To investigate the progression and severity of the IgG infiltration, a more thorough examination of IgG immunoreactivity was then performed on rTg4510 and wild-type mice at 1-, 3-, 6-, 9-, and 12-months old. We found that rTg4510 mice displayed progressively higher visible levels of anti-mouse IgG immunoreactivity than age-matched wild-type littermate (Figure 1a). The most visibly apparent regions affected by IgG accumulation were the hippocampus, stemming from the fimbria of the hippocampus, and the frontal cortex, most notably along the edges of the tissue. Both the hippocampus and frontal cortex have been characterized to have extensive tau accumulation and severe neuron loss in rTg4510 model [24]. In fact, the hippocampus has been shown in wild-type rodents to be susceptible to BBB impairment [35,36], so this natural predisposition combined with the intense pathology which is found throughout the hippocampus made this region particularly interesting. Anti-mouse IgG quantification in the hippocampus revealed a significant increase in 12-month old rTg4510 mice (Figure 1b) compared to age-matched wild-type mice. This same phenomenon was mirrored in previous studies using hippocampal tissue of AD cases [5]. Upon closer visual examination of the hippocampus in rTg4510 mice, the highest accumulation of IgG was found in the CA3 (Figure 1c), however the dentate gyrus and the CA1 region were also noticeably darker than the wild-type littermates. Although a similar pattern of staining were seen in the hippocampus of the wild-type mice, the overall staining was much lighter than that found in the rTg4510 model. Since regions adjacent to periventricular areas are most prone to BBB permeability, it was not surprising the CA3 region had the most IgG immunoreactivity [37]. In the frontal cortex, another region that has significant tau accumulation and atrophy [38], IgG immunoreactivity was also significantly increased in 12-month old rTg4510 mice compared to wild-type littermates (Figure 1d and e). This immunoreactivity was most marked on the edge of the tissue with a gradient to lighter IgG reactivity radiating laterally inward. Based on these findings, we speculated aging, combined with chronic tau overexpression, could lead to BBB disruption.

To more thoroughly investigate BBB permeability in these mice, three cohorts of rTg4510 mice and wild-type littermates aged 6-, 9-, and 12-months old were intracardially injected with EB, a dye that is typically occluded from the brain except when the BBB is damaged. EB is commonly used for BBB studies and emits fluorescence when bound to albumin in the blood [39-41]. Mice were perfused to remove all blood from the brain. Thus any fluorescent signal in the brain would indicate EB bound to albumin in the parenchyma and not the surrounding vasculature, indicative of BBB breakdown. Whole brains from mice that received intracardiac EB injections were imaged following sacrifice (Figure 2a). Blue staining was visible in each of the 12-month old rTg4510 tau mice predominantly in the cortical layers superior to the hippocampus, similar to the pattern of accumulation of IgG on the surface of the cortex (Figure 1e). This blue staining was not visible in 6-month old rTg4510 mice or 12-month old wild-type littermates, but was seen to a lesser extent in some 9-month old rTg4510 mice.

Brain tissue was fixed, sectioned, and mounted for brain region specific analysis of EB. EB fluorescence was found to be highest in 12-month old rTg4510 mice, compared to younger rTg4510 mice or age-matched wild-type mice (Figure 2b). Regions with high EB expression corresponded to those with greatest anti-mouse IgG immunoreactivity (Figure 2c), the highest being the hippocampus and the cortex, followed by the entorhinal cortex and hippocampal commissure. This again shows that periventricular regions are most susceptible to BBB dysfunction [37], and that this is severely exacerbated in the rTg4510 mouse model. Thus, we analyzed the extravasation of EB in the hippocampus and frontal cortex, since these areas showed prominent IgG infiltration. Significant EB fluorescence was observed in the hippocampus (Figure 2d) and frontal cortex (Figure 2e) of 12-month old rTg4510 mice but not in 6-month old rTg4510 mice or any wild-type mice. Similar to IgG, extravasation of EB was greatest in the CA3 region of the hippocampus (Additional file 1: Figure S1), compared to the CA1 and dentate gyrus, however both of these regions still showed significant EB accumulation. 9-month old rTg4510 mice showed high levels of EB fluorescence in both the hippocampus and frontal cortex which were significant by Student t-test, but not ANOVA. Thus rTg4510 mice can develop a major BBB defect, but this is only significant after 9 months of age, well after robust brain atrophy and tau accumulation begins in this model.

As previously reported [22,23,42], tau accumulation was significantly increased in an age-dependent manner in the hippocampus (Figure 3a) and frontal cortex (Figure 3b) of rTg4510 mice. Tau accumulation throughout the brain accompanied by visible brain atrophy, including the hippocampus (Additional file 1: Figure S2), was progressively increased in mice from 6 to 12 months of age (Figure 3c). Tau staining in the hippocampus (Figure 3d) and frontal cortex (Figure 3e) showed region-specific increases in tau accumulation with age.

It is well-documented that tau accumulation in the rTg4510 mouse model induces neuroinflammation, including microgliosis and astrocytosis [24,30,32,43]. In fact, there is a robust inflammatory response that begins somewhere between 2 and 4 months of age in these mice [43]. This inflammation is closely correlated with neurodegeneration in this model and can be reversed by DOX-treatment [20,38,43]. Here we focused on astrocytosis rather than microgliosis given the intimate role of astrocytes with BBB stability [44]. We investigated the levels and morphology of heat shock protein 27 kDa (Hsp27) in 6-, 9-, and 12-month old rTg4510 and age-matched wild-type mice. Hsp27 was selected because it has been shown to increase with age [22], is highly expressed in activated astrocytes [45], is found lining blood vessels [46], is constitutively expressed in vascular endothelial cells [47], and has been implicated in regulation of the BBB through its role in actin stabilization [48]. We found that Hsp27 levels in the hippocampus (Figure 4a) and frontal cortex (Figure 4b) were significantly increased in rTg4510 mice at 12 months of age. This progressive increase was clearly visible throughout entire brain sections (Figure 4c). Upon closer examination of the hippocampus (Figure 4d) and frontal cortex (Figure 4e), Hsp27 immunoreactivity appeared adjacent to blood vessels and with age these astrocytes appeared more activated (Figure 4f). The age-related increases in Hsp27 were significantly exacerbated by tau overexpression (Figure 4a and b).

To confirm the accumulation of astrocytes as detected by Hsp27 staining, we also stained tissue from 6-, 9-, and 12-month old rTg4510 and wild-type mice for GFAP, a more traditional marker of astrocytes [18,49,50]. Similar to the results from Hsp27, GFAP was significantly elevated in the hippocampus (Figure 5a) and frontal cortex (Figure 5b) of aged rTg4510 mice compared to wild-type controls. Whole brain sections show that accumulation of GFAP followed a similar pattern to that of tau in the rTg4510 mice (Figure 5c). Closer examination of the hippocampus (Figure 5d) showed that there was an age related increase in GFAP immunoreactivity that was exacerbated in the rTg4510 mice starting at 9 months of age, significantly higher than wild-type mice. While the rTg4510 cortex (Figure 5e) showed less of an age-related increase in GFAP immunoreactivity, there was still a clear elevation starting at 6 months of age that progressively increased.

Blood component infiltration into the brains of aged rTg4510 mice suggested impaired BBB integrity. Therefore, we sought to determine if whole peripheral blood cells could also enter the parenchyma with age. Brain sections from 6-, 9-, and 12-month old rTg4510 mice were stained with hematoxylin and eosin (H&E) to visualize red blood cells (RBCs; erythrocytes) that remained in the brain parenchyma following perfusion [51]. RBCs, which are highly eosinophilic, stain bright red with circular morphology void of blue nuclear staining by H&E staining. These RBCs were manually counted throughout at least 6 sections from each mouse in the 6-, 9-, and 12-month old cohorts. Significantly more RBCs were found in the rTg4510 parenchyma at 12 months of age (Figure 6a, b). RBCs appeared to infiltrate the tissue near ventricles, the outer edges of the cortex, and along vasculature, particularly in the hippocampus.

As BBB defects are often correlated with entry of immune cells into the brain, staining for CD3⁺ T cells was performed. CD3⁺ T cells significantly infiltrated the brain of 12-month old rTg4510 mice (Figure 6c), especially near the longitudinal hippocampal blood vessels (internal transverse hippocampal artery and/or vein; Figure 6d) and periventricular to the lateral ventricles (Figure 6e), but were rarely observed in age-matched wild-type mice (Figure 6f). CD4⁺ T cells were also found to significantly infiltrate the brains of 12-month old rTg4510 (Figure 6g) mice compared to their younger counterparts and to age-matched wild-type mice. The infiltration of these CD4⁺ T cells were found in similar regions as CD3⁺ T cells, along the longitudinal hippocampal blood vessels (Figure 6h) and near periventricular regions (Figure 6i and j). Taken together, these data show that BBB disruption in 12-month old rTg4510 mice is sufficient to not only allow blood components into the brain, but also whole blood cells.

We next sought to test the resiliency of the BBB to damage caused by tau accumulation. We took advantage of the ability to suppress tau expression in the rTg4510 model using DOX feed. DOX was fed to 10.5- and 11-month old rTg4510 and wild-type mice for ~5 weeks. At 12 months of age, mice were injected with EB forty minutes before perfusion and tissue was collected following this in an identical manner to the cohorts described above. Immunostaining revealed that the levels of IgG were lower in 12-month old rTg4510 mice that received DOX treatment compared to the untreated group. In fact, the IgG immunoreactivity in the hippocampus (Figure 7a and b) and frontal cortex (Figure 7c and d) of mice treated with DOX beginning at ~11 months was indistinguishable from 9-month old rTg4510 mice, suggesting that tau suppression by DOX could slow and possibly reverse the BBB dysfunction in these mice. Superficial examination of EB staining also suggested that DOX-treated 12-month old rTg4510 mice resembled 9-month old rTg4510 mice with regard to EB extravasation (Figure 7e). Histochemical analysis of the area ratio of EB extravasation in these mice showed that the hippocampi of 12-month old DOX-treated rTg4510 mice were quantitatively indistinguishable from either 9-month old or 12-month old untreated rTg4510 (Figure 7f and g), and the frontal cortex had significantly lower levels of EB compared to 12-month old untreated rTg4510 mice (Figure 7h and i). Similar trends were observed for EB particle count, total area positive and average particle size (Additional file 1: Figure S3) Taken together, this suggests that BBB integrity was maintained and in part restored following tau suppression by DOX. As expected, hippocampal atrophy was also significantly slowed by DOX treatment, as measured by traces used for segmentation from NearCYTE software (Figure 7j). However, 12-month old DOX-treated rTg4510 mice still had significantly lower hippocampal volume than 9-month old untreated rTg4510 mice. Altogether, these data suggest that tau suppression can slow and possibly even rescue the progression of BBB dysfunction in this model of tauopathy.

As expected, total tau levels in DOX-treated mice were significantly decreased in the hippocampus (Figure 8a and b) and frontal cortex (Figure 8c and d) compared to untreated rTg4510 mice. As depicted by the representative images in the hippocampus (Figure 8b) and frontal cortex (Figure 8d), there was a noticeable reduction of the dispersed tau-positive staining in the 12-month old DOX-treated and 9-month old rTg4510 mice compared to 12-month old untreated rTg4510 mice. In fact, the DOX treated mice were statistically indistinguishable from 9-month old untreated rTg4510 mice. This reduction in dispersed tau staining may be due to a decrease in more soluble tau species which are lowered by DOX, since previous studies showed that insoluble neurofibrillary tangles are not cleared by tau suppression [23]. Hsp27 accumulation was also significantly reduced by tau suppression in both the hippocampus (Figure 8e and f) and frontal cortex (Figure 8g and h), but was still significantly higher than 9-month old untreated rTg4510 mice in the hippocampus (Figure 8e). In addition to slowing the accumulation of the levels of Hsp27, the morphology of cells expressing Hsp27 appears to be slightly altered in both the hippocampus (Figure 8f) and frontal cortex (Figure 8h), suggesting that distinct activation states of astrocytes may also contribute to BBB disruption.

Confirmation of the reduction of activated astrocytes in DOX-treated rTg4510 mice was verified by GFAP staining, which was recently shown to be rescued by DOX in a separate study [43]. Consistent with Hsp27 staining, GFAP immunoreactivity was significantly lower in the hippocampi of 12-month old DOX-treated rTg4510 mice compared to their untreated counterparts (Figure 9a and b), but still significantly higher than 9-month old untreated rTg4510 mice. In the frontal cortex, GFAP immunoreactivity in 12-month old DOX-treated mice was significantly lower than 12-month old untreated rTg4510 mice, instead resembling 9-month old rTg4510 mice (Figure 9c and d).

T cell lymphocyte and RBC infiltration were also reversed by tau suppression. The RBC counts from 12-month old DOX-treated rTg4510 mice was significantly lower than 12-month old untreated rTg4510 mice and were indistinguishable from 9-month old untreated rTg4510 mice (Figure 9e). We also found that CD3⁺ (Figure 9f) and CD4⁺ (Figure 9g) T cells were significantly lower in the DOX-treated rTg4510 mice compared to untreated mice. Unlike CD3⁺ T cells, CD4⁺ T cells, while reduced, were still significantly higher in 12-month old DOX-treated rTg4510 mice compared to 9-month old rTg4510 mice.

Since the hippocampi of the rTg4510 model develop the most robust and aggressive tau pathology, we examined this area for any apparent differences in the morphology of tau staining. Through this process, we discovered tau immunoreactivity in and around the longitudinal blood vessels of the hippocampus (Figure 10a; additional examples are included in Additional file 1: Figure S4a) [52,53]. Quantification of this perivascular tau revealed that it was significantly more abundant in 12-month old rTg4510 mice than in 9-month old rTg4510 mice or 12-month old DOX-treated rTg4510 mice (Figure 10b). Tau was not localized to this region in 12-month old wild-type mice (Figure 10c) or in 6-month old rTg4510 mice (Figure 10d). It only emerged in rTg4510 mice 9 months of age or older (Figure 10e). We found that the accumulation of tau in and around these blood vessels was significant by 12-months of age (Figure 10f), but was markedly reduced in DOX-treated rTg4510 mice (Figure 10g). Lectin and tau co-staining (Figure 11) revealed that tau was found closely lining the endothelial cells of the blood vessels in the hippocampus in 12-month old rTg4510 mice. This was not found in age-matched WT mice and appeared less robust in 12-month old DOX-treated mice. These findings suggest that tau along these vessels is not within endothelial cells based on the lack of co-localization of tau with lectin staining. We then performed co-localization studies of Hsp27, tau and EB to better define the location of this perivascular tau. Tau was found co-localized with both EB and Hsp27, suggesting that tau and EB were within Hsp27 positive cells along the vasculature (Additional file 1: Figure S4b). Since tau did not appear to co-localize with endothelial cells, and since tau has not been found in neuroglia up to 13 months in the rTg4510 mice by electron microscopy [54], we can conclude that the perivascular tau is likely either within neuronal processes that could contain Hsp27 and EB, or it could be extracellular as both EB and Hsp27 are found outside of cells as well [55,56]. Regardless of the localization of this perivascular tau, its emergence coincides much more closely with BBB dysfunction than other pathologies in this model, such as neuronal loss and gliosis. Thus, these findings suggest an important relationship between this perivascular tau pathology and BBB integrity.