Non-neuronal and neuronal BACE1 elevation in association with angiopathic and leptomeningeal β-amyloid deposition in the human brain

Increased BACE1 immunoreactivity (IR) relative to parenchymal amyloid pathology has been reported in animal and human brains [32,35,36]. Consisted with the existing data, increased BACE1 and Aβ IR were found in the cases with brain amyloid pathology as compared to controls (Figure 1). Briefly, in the cases of the amyloidosis group, BACE1 IR was increased at local sites against an overall neuropil-like background (Figure 1A,B,D,E). In contrast, BACE1 IR in control cases exhibited essentially a diffuse neuropil-like pattern across the temporal lobe cortical structures (Figure 1C,F). The distinct BACE1 labeling normally present at the mossy fiber terminal field was otherwise comparable between cases with and without cerebral amyloid pathology (Figure 1A-C).

Importantly, while essentially no vascular BACE1 IR was identifiable in sections from the control group (Figure 1C,G,H), increased BACE1 IR was seen at some vascular profiles in the cortex and hippocampal formation in the cases with neuritic amyloid pathology (Figure 1B,E,I,J). In general, the increased vascular BACE1 IR could occur around the pia (will be detailed in following section), in any cortical layer or in the white matter. The labeled vessels varied in size, but appeared to be mostly capillaries and arterioles as judged on the basis of their size, histological configuration and laminar distribution. This site-specific increase of BACE1 IR at capillary and arteriole-like profiles was confirmed by double immunofluorescence for BACE1/6E10 (Figure 1K-N) and BACE1/collagen IV (Figure 2A, B). At higher magnifications, BACE1 IR at blood vessels exhibited variable patterns and intensities in DAB (Figure 2C,D) or DAB/H.E. dual-stain (Figure 2E-H) preparations, better appreciated in the latter wherein the vascular lamination was displayed. Specially, BACE1 IR at arteriole-like profiles showed a differential laminar distribution that was related to the overall amount of labeling at individual vessels. Thus, when comparing intracortical arterioles without (Figure 2E) to those with BACE1 IR (as assessed in the same sections), the labeling appeared to occur initially then intensify in the tunica intima (TI) or endothelial layer (Figure 2F). As the endothelial BACE1 IR became evident, labeled elements often occurred in the perivascular area, with the tunica media (TM) or the smooth muscle layer spared (Figure 2C). As the overall BACE1 IR became more abundant, the labeling occurred also in the TM (Figure 2G,H). BACE1 IR in and surrounding the vascular wall might appear segmentally along the vessels cut longitudinally, giving the profile a feather-like appearance (Figure 2D). It should be noted that BACE1 labeled elements in the (TM) appeared as fine processes, some had local swellings (Figure 2G,H), suggestive of a labeling to neuronal terminals than smooth muscle cells.

The vascular Aβ IR exhibited a differential labeling pattern in ABC-DAB immunohistochemical preparations. As defined by some neuropathologists [13], vascular Aβ deposition could be staged according to the pattern and amount of Aβ IR between individual vessels, which was applicable to the intracerebral arterioles exhibiting 6E10 IR in this study (Figure 2I-M). Thus, light 6E10 IR appeared to initially occur at the TI or endothelial layer (Figure 2I), followed by the emergence and intensification of the labeling in the TM (Figure 2J-L). However, we observed perivascular 6E10 IR at arterioles with fairly light or heavy Aβ IR in the vascular wall (Figure 2J-K). The profiles containing vascular and perivascular Aβ IR resembled the so-called “dysphoric or dyshoric CAA” defined in previous studies [13,37,38].

In double immunofluorescence, a distinct endothelial BACE1/6E10 colabeling was identified at intracerebral arterioles, with some exhibited 6E10 IR in the TM arranged as a ring-like pattern in the absence of BACE1 IR (Figure 3A-C). Among the intracortical arterioles displaying BACE1/6E10 IR in the middle wall, a pattern of partial colocalization was recognized (Figure 3D-F). At some intracortical vessels with heavy 6E10 IR, the intensity of BACE1 IR was otherwise at the background levels (Figure 3G-I). As a reference, typical neuritic plaques showed a partial colocalization pattern as well, with BACE1/6E10 IR overlapped only in the dystrophic neurites that were surrounded by 6E10 reactive Aβ deposits (Figure 3G-I). The differential colocalization of BACE1 relative to intracellular vs. extracellular 6E10 IR might be related to a cross-reactivity of 6E10 to intraneuronal APP and/or APP β-C terminal fragments, although the 6E10 IR inside the dystrophic neurites could reflect the existence of intracellular Aβ [25].

In ABC-DAB immunohistochemical preparations, cell loss was readily noticeable in the wall of the intracortical arterioles with relatively strong BACE1 IR (compare Figure 2G,H to E,F). We also noticed a loss of bisbenzimide labeled nuclei in the vascular wall of vessels with moderate to strong BACE1/6E10 IR in double immunofluorescence. In order to assess whether there existed a trend of relative changes in BACE1 IR, 6E10 IR and cell loss, we carried out correlative densitometry for the 3 variables among individual cortical and white matter arterioles. This was conducted in batch-processed immunofluorescent sections from 5 cases. While scanning over the cortex and white matter (following an alternating path), each 6E10 labeled arterioles encountered was imaged. A set of images [the red (BACE1), green (6E10) and blue (bisbenzimide) channel] from the same brain was then put together, converted into a black and white tiff file. Using the OptiQuant software, a measuring template was created in the original green image over the area occupied by 6E10 IR at the blood vessel. This same template was copied and pasted over the other two images and aligning to the same vessel (see Figure 3D/E and G/H for methodological illustration). The optic densities for the 3 types of labeling at the same vascular profile were thus obtained. For a given brain, 20 cross-sectioned intracerebral arteriolar profiles were analyzed by this method. To obtain the baseline densities, 10 cross-sectioned intracortical arterioles not labeled by BACE1 and 6E10 encountered in the same microscopic fields were also photographically documented in each brain. The “immunofluorescent” (actually the background) and bisbenzimide densities were obtained by the same method, except that the vascular measuring template was created in the bisbenzimide image (for a methodological demonstration see the framed artery in Figure 4A).

Among a total of 100 quantified amyloid arterioles, the mean optic densities were 4397.9 ± 298.4 DLU/mm² for BACE1 IR, 6328.6 ± 794.9 DLU/mm² for 6E10 IR and 222.5 ± 11.9 DLU/mm² for bisbenzimide fluorescence. For the 50 control arterioles (10 profiles/brain from the same 5 brains), the mean densities were 3298.3 ± 268.5, 3317.1 ± 237.0 and 253.4 ± 41.3 DLU/mm² for BACE1, 6E10 and bisbenzimide signals, respectively (Figure 3J). There was a significant difference between the amyloid and control arterioles for the mean density of BACE1 IR (P < 0.0001), 6E10 IR (P = 0.0003) and bisbenzimide (P = 0.0001) (two-tailed Student-t test). To explore a trend of change between the 3 variables, the density of each marker measured at a given vessel was normalized to the mean density of the corresponding marker from all control profiles (i.e., defined as 100% or baseline). The order of the vascular profiles was then resorted according to the normalized 6E10 densities (from low to high), with the 3 variables plotted as a function of the vascular order numbers (Figure 3K). BACE1 and 6E10 IR concurrently increased initially, with BACE1 IR further increased to approximately at a plateau as the levels of 6E10 IR continued to rise until about 2 fold of the baseline. While 6E10 IR increased further up to 4 fold of the baseline density, BACE1 IR did not continue to elevate but tended to drop towards the baseline among many profiles. Bisbenzimide density tended to decrease with the increase in 6E10 IR, especially among the vascular profiles with 6E10 density reached and above 2 fold of the baseline level (Figure 3K).

The aforementioned BACE1 immunoreactive elements in the perivascular area and TM were of particular interest as they could imply other cellular contributors than the endothelial cells (verifiable simply by regular histology) of Aβ to amyloidosis. Using double immunofluorescence, the process-like BACE1 immunoreactive elements were found to partially colocalize with synaptophysin (Figure 4A-C). These elements were not labeled with αSMA, a marker of vascular smooth muscle cells (data not shown). The incomplete BACE1/synaptophysin colocalization was considered to indicate that BACE1 elevation was not limited to (dystrophic) presynaptic terminals. As a cross-validation for this latter possibility, we used another neuronal marker, NADPH-d, which could display neuronal processes extensively [39-41]. Following an initial histochemical screening, cases with optimal neuronal NADPH-d reactivity were subjected to further BACE1 and 6E10 immunolabeling. NADPH-d reactive dystrophic neurites were observed in the vascular wall or the perivascular zone, coexisting with BACE1 (Figure 4D) or in mix with local Aβ deposition (Figure 4E). NADPH-d reactive dystrophic neurites were also present in some amyloid plaques (Figure 4F).

Considering the well-established link between CAA and microbleeding [10-12], we explored whether vascular profiles with BACE1 labeled neurites might be associated with hemosiderin leakage. In sections counterstained with Prussian blue, ferric iron deposition was seen in the wall as well as at perivascular area among the arterioles with BACE1 labeled neuritic profiles (Figure 4G-I). There were BACE1 immunoreactive neurites directly colabeled by Prussian blue, with some extended to near the vascular cavity, even in arterioles with a small amount of dystrophic neurites (Figure 4G,H). Iron deposition was also seen at subset of BACE1 labeled dystrophic neurites arranged as rosette-like clusters (Figure 4I), which would represent the sites of parenchymal neuritic plagues.

As with previous animal and human brain studies [35,36,42,43], BACE1 IR was essentially not detectable at cerebral vasculature in brain sections from the control cases or at normal vessels in brain sections with amyloid pathology. This would appear not surprising in general given that the differentiation of specific immunolabeling depends on a sufficient signal to background ratio. In other words, since BACE1 is more abundantly expressed in neuronal than vascular elements, neuronal BACE1 IR could mask vascular BACE1 IR, if any, in brain sections under normal conditions. Because “de novo” BACE1 IR was identified at selected vascular sites in postmortem human brain in the present study, we speculated that vascular cells should express BACE1 and perhaps other Aβ producing proteins endogenously.

Surgically isolated blood vessels and primary cell culture, together with human cortical extracts as control, were used to explore the expression of amyloidogenic proteins in adult human vascular cells. In biopsied leptomeningeal and peripheral (image not shown) arteriolar, venular (image not shown) and capillary preparations, BACE1, APP and PS1 (not shown) IR were detectable by immunohistochemistry, localized largely to the endothelial cells (Figure 5A-C). It should be noted that, in relatively large arteries [i.e. with an internal elastic lamina (IEL)], strong and weak fluorescent signal existed in the IEL and TM, respectively (Figure 5B,C). By excluding the primary (Figure 5D,E) or secondary (image not shown) antibodies, the endothelial BACE1 and APP IR became eliminated, while the fluorescent signal in the IEL and TM largely remained, as compared with that in the normally processed sections (Figure 5B,C, inserts). Thus, the fluorescent signal in the endothelial layer appeared to be specific, whereas that in IEL and TM was largely autofluorescence.

Protein products of APP, BACE1 and APP C-terminal fragments were detected in the extracts of the leptomeningeal and peripheral vascular biopsies, at lower levels relative to the cortex (Figure 5F). Notably, vascular APP holo-proteins migrated at a slightly heavier molecular weight position relative to brain APP. Presenilin-1 N-terminal fragments (PS1-NTF), an active component of γ-secretase complex, were also detected in vascular lysates, although at similar levels between the cortical, arterial and venous samples. The vascular samples showed an expected minimal amount of the β-tubulin proteins as compared to cortical lysates (Figure 5F). Vascular extracts were also found to exhibit APP β-site cleavage enzymatic activity, and contain detectable amounts of Aβ42 (data not shown). Immunocytochemical characterization on the expression of amyloidogenic proteins was carried out in primary cell cultures of resected peripheral arteriolar samples maintained in vitro for 4-8 days. Cells expressing CD31, a marker of the endothelia, were found to co-express APP, BACE1 and PS1-NTF IR (Figure 5G-L). Cells expressing the smooth muscle actin ∀SMA, including some appearing in a fused configuration, exhibited minimal BACE1 IR (Figure 5M,N).

In addition to the above findings, we observed increased BACE1 IR on the pial surface in a subset of cases with parenchymal and vascular amyloid pathology (Figure 6). Specifically, increased BACE1 IR around the pia was distinct in sections from 5 cases of the group with prominent intracortical CAA (Figure 6A), whereas BACE1 IR at the same location was minimal in control cases (Figure 6C). Individual BACE1 immunoreactive cells in round or oval shape localized along the pia, apparently lacking processes typical of neuronal or glial cells (inserts in Figure 6A). In adjacent sections, amyloid deposition was present on the cortical surface and also in nearby vascular profiles in these same cases, in contrast to controls (Figure 6B,D).

The finding of concurrent increases of BACE1 and Aβ IR along the pia allowed an opportunity for biochemical verification of the protein change using isolated leptomeningeal samples (Figure 6E,F). Extracts of the leptomeningeal samples over the temporal neocortex of the anatomically verified amyloid and control cases were assayed (Figure 6G-I). BACE1 protein levels were elevated in the lysates of the amyloidosis group (134.2 ± 19.2% of GADPH levels, mean ± SD) relative to control (95.0 ± 11.1%) (P = 0.021, n = 5, Paired t test, same test below) (Figure 6G,H). APP β-cleavage activity was increased in the amyloidosis cases (174.9 ± 35.5%) relative to controls (100 ± 16.9%) (P = 0.001) (Figure 6H). Levels of APP C99 were also higher in the amyloidosis group (103.2 ± 32.3% of GADPH levels) than control (70.9 ± 18.3%) (P = 0.009) (Figure 6G,H). Moreover, Aβ42 levels were elevated for about 2 fold in the amyloidosis cases (221.6 ± 86.2 pM/ml) compared to controls (63.2 ± 8.1 pM/ml) (P = 0.013) (Figure 6I).

Immunocytochemistry was used to determine if BACE1 and other Aβ producing proteins could be endogenously expressed in meningeal cells. Meningeal biopsies (from 6 patients) were available for primary cell culture to address this question. In initial assays the cultured cells derived from 3 cases showed 70-90% purity of human meningeal cells, as characterized by a polygonal morphology (Figure JE) and the expression of fibronectin (FN), and a lack of labeling for GFAP (astrocytic marker), CD11 (endothelial marker) or MAP2 (neuronal marker) [44]. We then carried out double immunofluorescence to check the expression of amyloidogenic protein. These experiments revealed BACE1 (Figure 6K,L), APP (Figure 6M) and PS1 (data not shown) labeling in the meninge-derived cells.