TSE diseases or prion diseases are a rare group of slow neurodegenerative brain diseases which affect humans as well as a variety of domestic and wild mammals. In prion diseases normal host-encoded cell surface-anchored prion protein (also known as PrP^C or PrPsen due to its sensitivity to protease digestion) undergoes misfolding and aggregation to generate a partially protease-resistant form known as PrPres or PrP^Sc, which is detectable by immunoblot and is an important biochemical marker for the disease (for reviews see [1, 2]). PrPres may also be a causal factor in the associated brain damage observed. Neuropathology of prion diseases in humans and in animals, such as sheep, cattle and rodents, is typically characterized by gray matter vacuolation and gliosis, as well as deposition of PrPres in either a diffuse non-amyloid form or a more dense amyloid form. Diffuse PrPres is the more common form and is seen in human sporadic CJD as well as in most types of animal prion diseases [3–6]. The amyloid form of PrPres is seen in many familial human prion diseases associated with PrP mutations as well as in certain animal prion disease models [7–9].

In mammals PrP is mainly expressed as a glycoprotein anchored to the cell surface by a glycophosphatidylinositol (GPI) linkage. Previously we generated transgenic Tg44 mice, which express only the anchorless form of PrP, and in these mice scrapie-infection results in an unusual type of slow fatal prion brain disease distinguished by widespread deposition of PrPres amyloid in the central nervous system [10] and in extraneural sites such as heart, brown fat, white fat and colon [11, 12]. In the CNS of infected Tg44 mice the gray matter vacuolation typical of prion diseases is minimal, and PrPres is primarily deposited as perivascular and intravascular amyloid which is associated with extensive non-vacuolar neuronal damage as well as astrogliosis and microgliosis [10].

The perivascular PrPres amyloid deposition and associated pathology seen in scrapie-infected Tg44 mice is similar to cerebral amyloid angiopathy (CAA) seen in several familial brain amyloid diseases involving mutant proteins or peptides (for review see [13]). In the case of prion diseases, PrP mutations including Y145X, Q160X, Y163X and Y226X [13–16] have been associated with development of CAA with perivascular PrPres amyloid, and all these mutations give rise to truncated PrP molecules which also lack the glycophosphatidylinositol (GPI) anchor. Interestingly, one other human patient expressing PrP Q227X which lacks the GPI anchor had disease with multicentric plaques and no CAA [15]. In humans CAA is most commonly observed with Alzheimer’s disease (AD) where Aβ amyloid is deposited in CNS as both vascular and parenchymal plaques. In CAA associated with Alzheimer’s disease, small arteries and arterioles are the most common sites of Aβ amyloid deposition [13, 17, 18], but amyloid is also detected in capillaries and in veins at a lower incidence [18–20]. In leptomeningeal vessels of AD patients the ratio of Aβ-positive arteries to veins was 4.5 to 1 [21]. This Aβ amyloid distribution mirrored the pattern of the bulk flow of brain interstitial fluid (ISF) tracers which appear to drain from the brain in perivascular spaces of arteries more than veins [22, 23], suggesting that ISF flow might play a role in the vascular distribution of amyloid in AD [24].

In our work with Tg44+/+ mice, we previously observed PrPres amyloid plaques associated with various brain blood vessels which appeared to include arteries, capillaries and veins based on morphological criteria [10]. In the present work using dual staining with anti-PrP monoclonal antibody plus antibody reactive with alpha smooth muscle actin (ASMA) in arteries and arterioles, we studied the vascular specificity of PrPres amyloid distribution in scrapie-infected Tg44+/+ mice. The distribution pattern of PrPres amyloid on blood vessels in this model appeared to be different than that of Aβ amyloid observed in humans with AD, and suggested that PrPres vascular deposition and subsequent CAA might involve a different pathogenic process than that seen in AD.

Scrapie-infected Tg44+/+ mice expressing only the anchorless form of PrP were found previously to deposit PrPres in the CNS as amyloid predominantly in a perivascular and intravascular distribution [10]. This disease required scrapie infection, as it did not occur spontaneously in uninfected Tg44+/+ mice. The lack of the GPI anchor group and the minimal amount of glycan on the anchorless PrP are likely to contribute to the tendency of anchorless PrP to form amyloid, but the reasons for the vascular distribution are not known. In the current studies of the vascular specificity of PrPres amyloid in brain of Tg44 mice, amyloid accumulated mostly on capillaries (73-78%), but also was found on both small arteries (11-14%) and veins (11-13%) (Tables 1 and 3). The association of PrPres with all these blood vessel types could be explained either by hematogenous re-distribution of intracerebrally inoculated scrapie PrPres or, alternatively by dissemination of PrPres via the flow of ISF towards capillaries, small arteries and veins. Our results on studies of mice after intravenous scrapie inoculation did not support the suggestion of hematogenous spread of scrapie infectivity and PrPres (Table 4). Thus, the most likely explanation for the brain perivascular PrPres distribution was dissemination of small diffusible PrPres aggregates by the ISF flow.

Based on our present finding of PrPres amyloid on capillaries, arteries and veins, amyloid polymerization might be initiated by some BM components, such as glucosaminoglycans (GAGs), found on all of these types of vessels [2, 31]. GAGs were previously shown to bind PrPsen [32] and block or potentiate PrP conversion in vitro [33]. Therefore, GAGs in BM might serve as scaffold structures to concentrate, bind and/or align diffusible PrPres aggregates which are too small to act alone as seeds for further PrPres conversion. Presumably after such re-alignment along a scaffold, these aggregates might function effectively as seeds for PrP conversion (Figure 4).

In contrast to the apparent involvement of blood vessel components in PrPres amyloid formation, it was surprising that amyloid plaques in this model often extended large distances (10 to 100 μm) away from the ablumenal blood vessel walls. Possibly, after establishment at the BM, PrPres amyloid might act without GAGs to self-scaffold further conversion of anchorless PrP at locations quite distant from the blood vessel BM. This conversion might also be increased by the abundance of diffusible anchorless PrP which is secreted into the extracellular space by many CNS cell types. In addition, anchorless glycan-negative free-floating PrP in extracellular spaces might have a greater tendency to form amyloid compared to normal membrane-anchored glycan-rich PrP [2].

Anchorless PrP, similar to that found in Tg44 mice, has been detected in association with several different human PrP mutations in patients with familial prion disease. In these cases, mutation of one PrP allele results in aberrant C-terminal PrP truncation (Y145X, Q160X, Y163X, Y226X, and Q227X) with production of shortened PrP lacking the GPI anchor moiety [13–16]. In these patients PrP amyloid with CAA is often a predominant neuropathological finding, and in the figures describing the Y145X and Y226X patients, the vascular specificity of the CAA appears to involve capillaries, arteries and veins. This was similar to the Tg44 model, except that in Tg44 mice capillary amyloid was more prominent (Tables 1 and 3). With all the above-mentioned human PrP mutations, the plaques appeared to be located in blood vessel walls. In addition, in the publications describing both the Y145X and Y226X mutations, the plaques also extended up to 15-35 μm into the parenchyma from the ablumenal vessel wall. This was similar to infected Tg44 mice where the PrPres amyloid plaques extended 10-100 μm into the surrounding brain tissue (Figure 1a-g and 2a-e). Thus, based on vascular specificity, plaque size and plaque location, neuropathology in humans with Y145X and Y226X was similar to that seen in scrapie-infected Tg44 mice.

CAA has also been extensively observed in association with Aβ amyloid deposits in Alzheimer’s disease patients, and CAA may be an important aspect of the neuropathogenesis of AD [17, 20, 24, 35–38]. Interestingly, CAA in these patients differs from infected Tg44 mice and from that seen in the Y145X and Y226X human PrP truncation mutations described above. First, the amyloid in Aβ CAA is mostly located within or very close to the vessel walls [13, 20], whereas PrPres amyloid appears to extend quite far away from the vessel wall [10, 14, 15]. Second, in Aβ CAA the vascular specificity of the amyloid deposition shows a preferential association with small arteries compared to veins which is not seen in PrPres CAA. For example, in the leptomeninges of AD patients, in one study small arteries and arterioles appeared to be the site of 80% of the amyloid, whereas only 20% was associated with veins [21]. However, in a different paper, the amyloid distribution in leptomeningeal vessels was arteries 56% and veins 45% [19]. Therefore, there may be variation in this ratio among different patients. In the cerebral cortex small arteries have been described as the main site of amyloid deposition, with lesser amounts of amyloid associated with veins, venules or capillaries [20]. However, in a subset of AD patients with CAA, designated Type 1-CAA, cortical Aβ amyloid was associated with capillaries, arteries and veins [19], and a role for capillary CAA in specific clinical dysfunction has been suggested [39, 40]. Thus, based on the vascular amyloid distribution, scrapie-infected Tg44 mice appear to be somewhat similar to the group of CAA-type 1 Aβ patients.

The reasons for the vascular specificity of CAA involving various different amyloid proteins are not known. Perhaps different proteins have differing requirements for BM-derived co-factors for amyloid formation. If Aβ amyloid were unable to self-scaffold efficiently in vivo so that a blood vessel-derived co-factor was always required, Aβ amyloid deposition might be more restricted to the blood vessel wall and less likely to extend out into the parenchyma. Similarly, if vascular smooth muscle cells, rather than BM, were the source of the essential co-factor(s), this might restrict deposition of Aβ amyloid to the walls of small arteries and arterioles which have abundant smooth muscle cells, as opposed to veins which have few smooth muscle cells. Perhaps capillary-derived co-factors are involved in Aβ CAA type 1 where capillary CAA is prominent [19].

The differences in PrPres and Aβ plaque size and vascular distribution might also be influenced by numerous other factors, including the rates of association, refolding and dissociation of the amyloid complexes in different sites, and the ability of glial cells and macrophages to breakdown the amyloid. The large difference in size between Aβ (40–42 amino acids (aa)) versus anchorless PrP (210aa in Tg44, 204aa in Y226X, and 120aa in Y145X), might also act by influencing some of these mechanisms. Single amino acid sequence differences can also influence the vascular specificity of amyloid deposition, as seen by differences between the effects of Y226X and Q227X PrP mutations, where the former is associated with CAA and the latter has parenchymal multicentric plaques without CAA [15]. However, the mechanisms of most of these effects remain unknown.

Many previous studies have followed clearance of ISF tracers labeled with radioactive or fluorescent tags (for review see [22, 30, 41, 42]). In our previous work, tracer was detectable on blood vessels at 30 min, much reduced at 7 hours and almost completely gone by 24 hours [27]. The present study provides quantitative data on specificity of the ISF tracer FITC-OVA for capillaries, arteries and veins. At 30 min post-injection of Tg44 mice we found the majority of tracer was associated with capillaries (96%), whereas arteries and veins were much lower (3.2% and 0.8% respectively) (Tables 5 and 6). It was not surprising that capillaries were the primary site of tracer location at early times post-injection because capillaries are the most numerous blood vessels in the brain. ISF tracer association with capillaries and arteries has been described previously by several groups [22]. In contrast, tracer association with veins was inconsistent in different systems. Some reported no association of ovalbumin or 3 kD dextran with veins [22]; however, another paper by this same group using 10 kD dextran reported a significant association with veins [30]. In the present studies, we found an artery to vein (A/V) ratio of approximately 4 to 1 for non-capillary-associated tracer at 30 min post-injection (Table 6), and at 7 hours post-injection this ratio was 3.1 to 1, which was not significantly different from the 30 min time-point. Thus, approximately, 20-30% of the tracer associated with larger vessels was associated with veins, and this tracer appeared to be cleared from the CNS by a mechanism involving veins.

The explanation for the skewed A/V ratio in our tracer clearance results remains unclear at the present time. ISF flow is thought to be driven mainly by pressure due to production of new ISF by brain capillary endothelial cells [41]. Although some of the ISF appears to drain via diffusion into the CSF [42], it is believed that much of the perivascular ISF is collected by lymphatic vessels surrounding blood vessels as they emerge from the skull [22, 23]. This explanation for outflow and collection of brain ISF outflow could be applicable equally to arteries and veins. The 4 to 1 A/V ratio seen in our experiments may relate to the amount of BM space available around endothelial and smooth muscle cells in vessels where the ISF flows. The larger amount of BM associated with smooth muscle cells in walls of arteries compared to veins might provide an increased space for ISF flow out of the brain on arteries versus veins.

Another explanation for the prominent detection of ISF tracers on arteries versus veins suggests that the contrary (or reflection) wave traveling in the reverse direction from the pulse wave created by the heart beat might drive the ISF along the outside of arteries in the opposite direction to the blood flow [23, 43]. The absence of valves along the ISF path outside blood vessels may be a problem for this option. However, of equal importance, this proposal would not explain the clearance of a substantial amount of ISF tracer associated with veins.

The relative selectivity of ISF tracer for arteries versus veins of 4 to 1 observed in our studies is in close agreement with the 4.6 to 1 ratio for Aβ association with arteries versus veins in leptomeninges of humans with AD [21]. Therefore, our results support the previous conclusion by Weller and co-workers that the vascular specificity of ISF flow may be an important factor in the accumulation and distribution of vascular Aβ amyloid in AD [24, 38]. In contrast, our present data on vascular distribution of PrPres amyloid suggested that ISF flow might be one of several factors involved in perivascular PrPres distribution. Other factors, such as vascular BM-derived co-factors for amyloid scaffolding, might also be important in vascular specificity of PrP amyloid both in the Tg44 mouse model and in humans with familial prion disease related to expression of mutated truncated PrP.

Vascular distribution of PrPres amyloid in scrapie-infected Tg44+/+ mice expressing anchorless PrP differed from the vascular distribution of ISF tracer molecules followed in these same mice. Thus, although ISF flow appeared to mediate the transport of small molecules such as PrP oligomers towards blood vessels, the ISF flow alone did not completely determine the vascular deposition of PrPres amyloid. We hypothesize that other factors including basement membrane components such as glucosaminoglycans and laminin, might act as scaffolds to increase amyloid formation by PrP or other proteins starting at the vascular basement membrane (Figure 4). Such scaffolding molecules might differ on different blood vessel types, and the requirements for such factors might differ among different proteins forming amyloid. These factors might account for the varying vascular specificities observed in different CAA diseases in mice and in humans.