Time course of neuropathological events in hyperhomocysteinemic amyloid depositing mice reveals early neuroinflammatory changes that precede amyloid changes and cerebrovascular events

Vascular contributions to cognitive impairment and dementia (VCID) are increasingly recognized as a significant cause of dementia, behind only Alzheimer’s disease (AD) [1]. Furthermore, VCID co-morbid with AD is extremely common, estimated to occur in at least 60% of AD cases [2]. Clinical-neuropathological correlation data has revealed that cerebrovascular pathologies increase the likelihood of dementia for a given amount of AD pathology. In early-onset familial AD, cerebrovascular abnormalities as detected by neuroimaging appear to precede detectable AD pathology. Despite our increasing understanding of VCID as a contributing factor to clinical dementia, our appreciation for the mechanistic underpinnings of VCID, and also identification of potential therapeutic targets to treat VCID, has been lacking, in part due to a lack of relevant animal models.

One of the challenges of developing models for VCID is that VCID is an umbrella term for a variety of cerebrovascular pathologies including micro- and macro-infarcts, micro- and macro-hemorrhages, cerebral hypoperfusion, periventricular and deep white matter hyperintensities, and stroke. We have been working on a mouse model of VCID characterized by inflammation, microhemorrhages, and cerebral hypoperfusion. The model is induced through diet modification to elevate plasma homocysteine, a non-protein forming amino acid (Hcy), resulting in hyperhomocysteinemia (HHcy). Over 20 years ago, HHcy (elevated plasma homocysteine) was identified as an independent risk factor for stroke and vascular disease [3]. HHcy is associated with confirmed VCID and AD cases [4] and is recognized as an important risk factor for AD [5]. HHcy is associated with accelerated hippocampal atrophy [4] and cognitive loss in AD patients [6]. The association with hippocampal atrophy appears to be independent of Aβ pathology [7]. HHcy [8‐11] and/or low B12 [12] and folate levels [13, 14] are also accompanied by white matter lesions, indicative of vascular damage. Genetic mutations in cystathionine β synthase, a key enzyme in the homocysteine metabolism pathway, lead to HHcy and are recognized as a cause of stroke in children and young adults [15].

We have previously shown that induction of HHcy in wild-type (WT) mice leads to neuroinflammation, microhemorrhages., and cognitive impairment, producing a clinically relevant mouse model of VCID [16]. We have also shown that HHcy leads to astrocytic end foot disruption beginning between 6 and 10 weeks on diet [17]. In our co-morbidity model, APP/PS1 mice placed on the HHcy diet show a switch to a pro-inflammatory phenotype, microhemorrhages, additive cognitive deficits, and a redistribution of amyloid from the parenchyma to the vasculature [18]. While we have previously described the time course of neuropathological changes in our WT mice on the HHcy diet, the effect of amyloid deposition on this time course remains unknown. In this study, we examined neuroinflammation, microhemorrhages, amyloid deposition, and cognition along a time course of 2, 6, 10, 14, and 18 weeks in our co-morbidity model. Similar to our WT HHcy model, neuroinflammation appears first at 6 weeks, followed by significant cognitive deficits at 10 weeks, and finally, microhemorrhages and redistribution of amyloid from the parenchyma to the vasculature occurred after 14 weeks on diet.

Neuroinflammation is the first pathological abnormality that we observe in our APP/PS1-HHcy model. Immunohistochemistry for CD11b, which stains for both activated and resting microglia, showed a similar trend as the WT mice on the HHcy diet (previously published [17]). By 6 weeks on diet, the APP/PS1 mice on the HHcy diet have significantly more CD11b staining compared to APP/PS1 mice on the control diet (Fig. 1a, b, i). This increase in staining is sustained at 10 and 14 weeks on the HHcy diet but is reduced to control levels by 18 weeks (Fig. 1c–i).

While immunohistochemistry for CD11b showed an increase in microglial staining, we were interested in the specific cytokines expressed. The pro-inflammatory markers TNFα and IL-1β were significantly increased in the APP/PS1 mice on the HHcy diet starting at 6 weeks on diet and remained elevated through 18 weeks on diet (Fig. 2a). IL-12a and IL-6, anther two pro-inflammatory markers, were significantly elevated at 6, 10, and 14 weeks but returned to control levels by 18 weeks on the HHcy diet (Fig. 2a). The anti-inflammatory marker IL-10 was significantly elevated after 6 weeks on the HHcy diet but decreased by 10 weeks and was significantly decreased compared to controls by 18 weeks (Fig. 2b). Another anti-inflammatory marker, YM1, was significantly decreased at 6 and 14 weeks on the HHcy diet (Fig. 2b). Similar to the CD11b staining, the increases in pro-inflammatory cytokines starting at 6 weeks were also seen in the WT mice on the HHcy diet [17].

We tested spatial memory changes in the APP/PS1 mice on the HHcy and control diets using the 2-day radial arm water maze. After 6 weeks on the HHcy diet, the APP/PS1 mice were not significantly cognitively impaired compared to the APP/PS1 mice on control diet (Fig. 3a). By 10 weeks on diet, the APP/PS1 mice on the HHcy diet made significantly more errors only in the last block of trials on the second day (Fig. 3b), similar to WT mice on the HHcy diet for 10 weeks [17]. The APP/PS1 mice on the HHcy diet made significantly more errors on the last three blocks on the second day at 14 weeks compared to the APP/PS1 mice on control diet (Fig. 3c). After 18 weeks on either the HHcy or the control diet, both groups of mice made a similar number of errors and did not learn the task at all (Fig. 3d). This does not mean that one group was more impaired than the other; the mice have reached a ceiling effect with this particular behavior task and further cognitive decline cannot be seen with this task. However, neither group showed any learning and are therefore still considered impaired.

Using both T2* MRI and Prussian blue staining, we determined the time course of microhemorrhages due to the HHcy diet in our co-morbidity model. Both methods showed a significant increase in the number of microhemorrhages in the APP/PS1 mice beginning at 14 weeks on the HHcy diet compared to the APP/PS1 mice on the control diet (Fig. 4a–c). Again, in the WT mice on the HHcy diet, we see a significant increase in microhemorrhages at 14 weeks [17].

To determine brain beta-amyloid (Aβ) load, we performed biochemical assessment of soluble and insoluble Aβ_1–38, Aβ_1–40, and Aβ_1–42 in the Meso-Scale Discoveries V-Plex Aβ assay. We found no significant effect of HHcy induction in Aβ levels at any of the time points examined (Table 3). To determine changes in amyloid distribution, we performed Congo red staining, which enables differentiation of vascular and parenchymal amyloid. At each time point, HHcy did not alter total levels of Congo red staining in both the frontal cortex and hippocampus (Fig. 5g, h). However, when parenchymal plaques and vascular amyloid are separated from total Congo red staining, it is revealed that HHcy leads to a redistribution of amyloid towards the vasculature (Fig. 5g, h). This redistribution occurs at 14 weeks on diet and continues to 18 weeks on diet (Fig. 5c–f).

VCID is the second most common form of dementia yet it is commonly seen as a co-morbidity with AD [25‐28]. While VCID is being increasingly recognized in the clinic, there are still a limited number of animal models to study mechanisms of VCID alone and when it is co-morbid with AD. Previously, we characterized a mouse model of VCID and a co-morbidity model by inducing HHcy in WT and APP/PS1 mice, respectively [16‐18]. We have shown the time course of neuropathological events in our VCID model alone [17]; however, we do not know the effect of amyloid deposition on this time course. Understanding and comparing the time course of these pathological changes could provide critical information for the mechanism of VCID alone and when it is co-morbid with AD. In this study, we show that neuroinflammation is the first pathology to occur, followed by cognitive changes and finally microhemorrhages and amyloid redistribution.

While little is known about inflammation’s role in VCID, it is hypothesized that inflammation and oxidative stress play a key role in neurovascular dysfunction leading to cognitive decline. Chronic hypertension, a risk factor for VCID, leads to lumen narrowing and ultimately reduced blood flow. This decrease in blood flow induces expression of hypoxia inducible factor 1α which recruits macrophages from the systemic circulation as well as activation of endogenous microglia [29]. Activation of microglia leads to increases in proteases and free radicals that are released during remodeling of damaged vessels. Free radicals and proteases can break down the fibrotic basal lamina, disrupt tight junctions leading to vasogenic edema, and attack the myelinated fibers leading to demyelination [30, 31]. This increase in reactive oxygen species can also stimulate inflammatory pathways via toll-like receptors and lead to blood brain barrier breakdown. This vascular damage via inflammation most likely interferes with neurovascular coupling and the proliferation, migration, and differentiation of oligodendrocytes contributing to white matter damage and VCID [32]. In both our VCID and co-morbidity model, alterations in the inflammatory phenotype are the first pathology to appear along with an increase in microglial staining. In addition, both models show a significant increase in several pro-inflammatory markers starting at 6 weeks before any disruptive vessel changes or cognitive decline is seen. While the microglial staining returned to control levels by 18 weeks on diet, the increase in pro-inflammatory markers remains throughout the course of the diet, possibly continuing to contribute to the neuropathological changes seen later. Future studies will look more closely at the morphological phenotype of the microglia to determine the number of activated vs inactive microglia. Together, this suggests neuroinflammation is an initial, key player in homocysteine-induced VCID.

As mentioned above, the proteases and free radicals released by activated microglia can induce blood brain barrier breakdown and leakage. Matrix metalloproteinase 9 (MMP9) is a key protease shown to be activated by pro-inflammatory cytokines (IL-1β and TNFα) and has been shown to degrade tight junction proteins leading to leaky vessels and eventually microhemorrhages [33‐35]. We have previously shown a significant increase in MMP9 gene expression and protease activity in both our VCID and co-morbidity models, with gene expression of MMP9 being significantly elevated after 24 weeks on diet [16, 18, 36]. MMP9 gene expression is still significantly elevated in the co-morbidity mice after 24 weeks on diet, showing a continued expression even after inflammation is returned to control levels [37]. In the current study, we show a slight increase in microhemorrhages at 10 weeks, with a significant increase by 14 weeks on diet. The appearance in microhemorrhages also coincides with the cognitive changes as well in both our VCID and co-morbidity models. At 10 weeks, there is a slight decline in cognition that becomes more pronounced by 14 weeks. With neuroinflammatory changes occurring before the appearance of microhemorrhages, we hypothesize that neuroinflammation induces MMP9 activation leading to microhemorrhages and then cognitive decline.

Another target of MMP9 includes the dystroglycans, specifically β-dystroglycan, which forms a complex that anchors the astrocytic end-foot to the basement membrane surrounding arterioles and capillaries [38, 39]. As part of the neurovascular unit, astrocytic end-feet and their anchoring complexes function to maintain the ionic and osmotic homeostasis of the brain. Disruption of these homeostatic processes by breakdown of the anchoring complex and removal of the end-feet from the vasculature could lead to cognitive decline. Previously, we have shown a significant decrease in several astrocytic end-feet markers by 10 weeks on diet in our WT mice [17]. This coincides with the slight increase in microhemorrhages at 10 weeks as well as the slight behavioral deficits at 10 weeks seen in both our models. In addition, it has been shown that astrocytic end-foot disruption is associated with cerebral amyloid angiopathy in both AD and mouse models of cerebral amyloid angiopathy [20]. In our co-morbidity model, since we have shown a redistribution of Congophilic amyloid from the parenchyma to the vasculature that occurs at 14 weeks on diet, we hypothesize that, similar to the WT HHcy mice, the co-morbidity mice will also have astrocyte end-feet degeneration around a similar time point. If so, with microhemorrhages and astrocytic end-foot disruption occurring before the redistribution of amyloid, astrocytes could play a key role in clearing amyloid through the vasculature. Any disruptions in the astrocytes could then lead to hindered clearance and thus the redistribution seen in our co-morbidity model.

Taken together, we have shown that neuroinflammation, specifically, an elevation of pro-inflammatory cytokines, is the initial change in both our VCID and co-morbidity mouse models and that the presence of amyloid does not alter the time course of neuropathological events in our mouse model. Overall, this data provides several therapeutic targets and time points that can be tested in our mouse models as well as potential overlapping pathologies to look for in other VCID and AD models and human tissue.