Microvascular pathology in the aging human brain: Evidence that senile plaques are sites of microhaemorrhages

Microvascular pathology in the aging human brain: Evidence that senile plaques are sites of microhaemorrhages

Neurobiology of Aging 27 (2006) 1786–1796 Microvascular pathology in the aging human brain: Evidence that senile plaques are sites of microhaemorrhag...

2MB Sizes 0 Downloads 50 Views

Neurobiology of Aging 27 (2006) 1786–1796

Microvascular pathology in the aging human brain: Evidence that senile plaques are sites of microhaemorrhages Karen M. Cullen a,b,∗ , Zolt´an K´ocsi c , Jonathan Stone b a

Anatomy and Histology, School of Medical Sciences, Institute for Biomedical Research, The University of Sydney, Australia b Research School of Biological Sciences, Australian National University, Canberra, ACT, Australia c Bendor Research Pty, Ltd, Sydney, NSW, Australia Received 8 December 2004; received in revised form 13 October 2005; accepted 18 October 2005 Available online 18 January 2006

Abstract Amyloid-rich plaques are a feature of the aging human cerebral cortex. We have recently described another feature of aging human cortex, microhaemorrhages, identified by their content of haem, red blood cells, collagen and clotting factors, and their spatial relationship to capillaries. Here we relate microhaemorrhages to amyloid deposits. Observations were made in three groups: patients with no history of dementia, patients with Alzheimer’s disease (AD) and patients with Down’s syndrome (DS) and dementia. Amyloid deposits and microhaemorrhages were labelled in adjacent sections, amyloid deposits with antibodies to ␤-amyloid (␤A), and microhaemorrhages by Prussian blue histochemistry for haem. The densities and sizes of ␤A deposits and haem-rich deposits (HRDs), and their relationship to blood vessels, were surveyed in temporal, cingulate and superior frontal cortex. Our results suggest that HRDs and ␤A deposits are the same sites of pathology. Their densities in the cortex and white matter of the regions surveyed varied markedly between cases, particularly between demented and non-demented cases, but they always co-varied; where haem deposits were sparse or numerous, so were ␤A deposits. Both HRDs and ␤A deposits formed adjacent to or encircling small vessels, often at branch points, and a spatial proximity analysis confirmed that both were found close to or colocalising with microvessels. Both HRDs and ␤A deposits were associated with blood- or vessel-derived proteins (fibrinogen, von Willebrand factor and collagen VI). Since haem is an established marker of cerebral bleeding, and amyloid is a marker of senile plaques, our results indicate that senile plaques are sites of microhaemorrhages. This colocalisation raises the very testable questions of whether microhaemorrhages are early events in plaque formation and whether therapies which stabilise cerebral microvessels can prevent the onset or slow the progress of dementias associated with plaque formation. © 2005 Elsevier Inc. All rights reserved. Keywords: Alzheimer’s disease; Microvasculature; Senile plaques; Microhaemorrhage; Haem; Collagen IV; Fibrinogen; Capillary; ␤-Amyloid; von Willebrand factor

1. Introduction The ageing of the vasculature presents a major health challenge, with several leading causes of death related to vascular disease. There is a growing literature on the frequent occurrence of vascular disease in Alzheimer’s disease (AD), with epidemiological evidence of overlapping risk factors [20,26,32,45,46]. Some authors estimate that cerebral vessel pathology (atherosclerosis, cerebral amyloid angiopathy)



Corresponding author. Tel.: +61 2 9351 2696; fax: +61 2 9351 2813. E-mail address: [email protected] (K.M. Cullen).

0197-4580/$ – see front matter © 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.neurobiolaging.2005.10.016

occurs in up to 90% of AD cases [26]. Most of these studies have emphasised pathology of larger vessels, assessed from macroscopically visible evidence of haemorrhagic and ischaemic stroke. However, pathology of cerebral capillaries has also been described in AD, with evidence of endothelial degeneration [28], cerebral hypoperfusion [17], collagenosis [5], tortuosity and vessel density changes [6]. Recently we reported evidence that haem-rich deposits (HRDs) are a common feature of the aging cortex, in cases both with and without AD, and in Down’s syndrome [15]. Given that haem is a stable marker of intracerebral bleeding, and because the HRDs form around small vessels (capillaries and small arterioles) and are associated with other

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

vessel and blood-borne components, we suggest that these deposits represent sites of intracerebral bleeding. In nondemented patients, the densities of HRDs are low. However, these deposits are numerous in AD and DS patients, and we therefore suggest that the damage caused by the microhaemorrhages is cumulative, and a direct cause of neurodegeneration and AD. The present study tests the relationship between HRDs and deposits of ␤-amyloid (␤A), a marker of senile plaques, and tests their proximity to nearby microvessels, to deposits of blood- or vessel-derived proteins, and to each other.

1787

2. Materials and methods

from Stages I to VI in the Braak and Braak paradigm [4]. No attempt was made to correlate dementia severity to this ranking scale: our rationale for using these stages was to test the hypothesis that ␤A and haem colocalise independent of disease severity. For this we required an independent marker for severity, i.e. tangle distribution. Cases with AD showed degrees of neurofibrillary tangle (NFT) and plaque densities equivalent to Braak Stages IV–VI and had marked cortical atrophy and dense gliosis. Control patients fell into of two groups: (1) cases without any evidence of plaque and tangle pathology and (2) cases with small numbers of plaques and tangles in the amygdala, hippocampus and entorhinal cortex insufficient for a diagnosis of AD (equivalent to Braak and Braak Stages I and II).

2.1. Case selection and classification

2.2. Tissue staining

Tissue from 20 individuals was used in this study: 12 with a diagnosis of AD, 2 cases with Down’s syndrome (DS) and AD, and 5 neurologically normal age-matched controls and 1 young control (age 27 years) (Table 1). The tissue was obtained through the New South Wales and the Victorian Brain Bank Networks supported by the National Health and Medical Research Council. All brains were fixed in formalin at autopsy. All AD cases were diagnosed on the basis of ante mortem evidence of dementia and absence of significant motor disturbance. All were diagnosed postmortem by presence of abundant senile plaques and tangles using Bielschowsky silver staining, tau and ␤A immunohistochemistry. For diagnosis, the CERAD criteria were applied [23,36,37]. To ensure that the tissue examined showed a broad range of severity of neuropathology, we studied tissue ranked

Blocks of tissue were taken from 3 to 5 mm coronal slices of the medial temporal lobe (including hippocampus, entorhinal cortex and fusiform gyrus), at the level of the lateral geniculate nucleus, and the anterior cingulate (Brodmann area 24) and superior frontal (Brodmann area 9) cortices at the first appearance of the anterior commissure. Blocks were sectioned (one in three series) at 45 ␮m on a CO2 Leica microtome and sections were collected into 0.1 M Tris buffer (pH 7.4). Matched serial sections (45 ␮m) were taken from each block; that is, alternating sections were collected in separate pots. An additional series was then taken from the remaining block, for general labelling, as required. From the matched (alternating) series, one set was mounted on gelatinised slides and stained with Prussian blue for haem and in some cases were counterstained with cresyl violet (0.5%). Our Prussian

Table 1 Case details and proximity analysis outcomes for the association between ␤-amyloid deposits and blood vessels Case

B&B*

Age (years)

Sex

PMD* (h)

p-Value (maximum from ∼30 images analysed for each region)

C1 C2 C3 C4 C5 AD1 AD2 AD3 AD4 AD5 AD6 AD7 AD8 AD9 AD10 AD11 AD12 DS1 DS2

0 0 II I I VI V VI VI VI V V IV VI VI IV VI V V

27 71 75 91 98 49 58 63 64 66 67 67 76 76 82 83 84 54 58

F M M M M M F F F M M M F M F F M F M

20 24 24 6 10 8 7 5.5 6 21 24 20 8 4 10 ∼24 ∼24 24 57

ND ND <0.001 <0.001 <0.001 <0.001 <0.001 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.001 <0.001 <0.01 <0.01 <0.001 <0.01

The probability cited is the likelihood that the observed association between HRDs and capillaries could be sampled from a randomised population of HRDs. For each brain we tested (and obtained a p-value) for approximately 30 images, and the p-value shown is the highest (least certain) value from the 30 images. B&B: Braak and Braak disease stage [4]; PMD: post-mortem delay; ND: not determinable, no HRDs detected.

1788

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

blue method used extended incubations (72 h) in acidic potassium ferrocyanide, as previously described [15]. The second series was labelled immunohisotchemically for ␤A and collagen IV. Immunohistochemical protocols were performed using antibodies against ␤A (monoclonal, Glostrup DK) and collagen IV (monoclonal, DAKO) for cerebral blood vessels, as well as clotting associated proteins, fibrinogen and von Willebrand factor. For ␤A and collagen immunohistochemistry, sections were pretreated in 90% formic acid for 3 min, then incubated in 0.04% pepsin (Sigma, St. Louis, MO) in 0.01N HCl to expose antigenic sites. Briefly, free-floating tissue sections were washed in 50% ethanol, endogenous peroxidase was quenched with 3% H2 O2 and non-specific binding blocked in 1% bovine serum albumen (Sigma) in Tris buffer, pH 7.4, containing 0.01% Triton-X. Sections were incubated in primary antibody for 48 h at 4 ◦ C. Bound antibody was visualised using biotinylated anti-mouse secondary antibody and avidin biotin kits (Vector Laboratories, Burlingame, CA) [13]. To visualise bound antibodies to collagen, we used peroxidase linked avidin biotin complex and DAB (Sigma) and for ␤-amyloid, we used alkaline phosphotase linked avidin biotin complex and Vector Red chromogen (Vector Laboratories). Double-immunolabelling was performed sequentially, with sections blocked in 0.1% BSA between primary antibody incubations. Additional series of sections were doubleimmunolabelled for ␤A and von Willebrand factor (DAKO, Glostrup DK) or fibrinogen (DAKO). All labelled sections were mounted on gelatinised slides, dehydrated and clear and coverslipped. We also attempted labelling some immunolabelled sections for haem with the Prussian blue technique. The harshness of the Prussian blue method, using the 72 h incubation needed to demonstrate the full population of HRDs [15] was such that the immunostaining was removed. When shorter incubation times (2 h) were used for the Prussian blue reaction, however, some HRDs (presumably newer deposits) were stained, with the immunoproduct preserved, enabling direct demonstration of colocalisation of haem and ␤-amyloid, at least for this subpopulation of deposits. 2.3. Proximity (spatial correlation) analysis of βA deposits and cerebral vessels To select areas for analysis of vessel/␤A association, strips of cortex ∼1 mm wide and orthogonal to the pial surface were outlined, by marking the coverslip with a marking pen. Marked strips were randomly chosen by picking a random number n between 1 and 6 and selecting each nth strip for photography [14]. Cortical strips were digitally photographed at 200× magnification. For each area sampled, 10 fields, each 0.7 mm × 0.7 mm, were photographed. For control cases, most microscope fields were free of plaques or HRDs, so sampling continued until at least five fields contained one or more amyloid deposits. For spatial correlation analysis, we used software module (Proxan) on a Linux platform. The software was devel-

oped by one of the authors (ZK) and is freely available at www.bendor.com.au/Proxan; the algorithm has been described previously [15]. Briefly, Proxan can, in digital images, outline two separate sets of objects (e.g. immunolabelled ␤A deposits and blood vessels; or HRDs and blood vessels). Proxan selects and separates the two sets by exploiting the colour and saturation differences between their chromogens. Proxan then outlines each set of objects, using user-adjustable algorithms that allow optimal matching of object boundaries to the observed structure of the tissue. Proxan, then, within each image, pairs each deposit with the nearest vessel, and identifies and measures the shortest distance between them, generating a set of deposit–vessel distances. Next, Proxan redistributes the deposits randomly, re-identifies deposit–vessel pairings and measures the new set of distance. This randomisation/remeasure step was repeated 500 times, yielding 500 sets of distances between randomised deposits and the unchanged vessel set. Proxan then calculated the ‘density distribution’ of distances for the 500 random trials. The probability that the observed distribution could have occurred within the 500 random trials was then determined using a χ2 analysis. We have previously described a Proxanbased analysis of the relationship of HRDs to vessels [15]. We here report the corresponding analysis for ␤-amyloid deposits and vessels. Specifically the tendency of deposits to be close to or overlap with vessels is expressed (Table 1) as the probability that the observed set of deposit–vessel distances could be sampled, if deposit–vessel relationships were random. Since we analysed approximately 30 images from each cortical area tested, this analysis yielded many p-values for each area; Table 1 shows the maximum p-value obtained for each cortical area. 2.4. Comparison of βA and HRD distribution The prolonged acidic steps of the haem reaction made it difficult (but not impossible, Fig. 6E and F) to successfully label ␤A deposits in the same section. HRDs and ␤A deposits were therefore studied in adjacent sections. To achieve accurate maps at high resolution, photomontages were prepared from digital images taken at 500× magnification. The images were collected by stepping the microscope stage by 700 ␮m in the X- and Y-axes, overlapping fields to facilitate assembly. For each section, approximately 100–180 fields were imaged, depending on the size of the tissue section. Using Photoshop (CS for Macintosh OSX) and the ‘automate > photomerge’ facility, a montage was assembled. The montage was checked for accurate matching of fields. Then, using the Photoshop colour selection tool to exploit the colour differential between the deposits (HRDs or ␤A deposits) and the background, deposits were selected and filled with contrasting colour. The selection of deposits was then manually edited to include overlooked deposits or to remove mistakenly selected artifacts. The mapped deposits were then superimposed on the section outline, with the grey/white matter boundary traced.

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

3. Results 3.1. Maps of haem and βA deposits Fig. 1 summarises the process of preparing maps of haem and ␤A deposits. High-resolution digital photographs were taken at 50× magnification across the tissue section (in this case the temporal lobe and hippocampus in Case AD3, see Table 1) with approximately 10% overlap. The ‘Photomerge’ tool in Photoshop was then used to automatically assem-

1789

ble the montage (Fig. 1A), with only minor manual editing being necessary. The montage was then adjusted for contrast and colour balance to maximise the differential between stained deposit (HRD or amyloid) and background. The deposits were then selected with the Photoshop colour tool and recoloured in a separate layer. The quality and detail of the selection of deposits that was possible with these techniques are shown in Fig. 1B. Manual editing was necessary to optimise the completeness and specificity of the selections. Once edited, for the whole montage, the deposits were plotted on an outline of the section (Fig. 1C). 3.2. Mapping of HRDs and βA deposits on adjacent sections Fig. 2 shows photomontages of adjacent sections of the temporal lobe (hippocampus, entorhinal cortex and fusiform gyrus from Case AD3, Table 1). One section is labelled for

Fig. 1. Illustration of mapping lesions from photomontages of stained 45␮m sections. (A) Digital photomontage of ␤A immunostained section from medial temporal lobe (Case AD3). The rectangular patch outlines an area shown at higher magnification in (B). This montage consists of 117 images of this size. (B) The area delineated in (A), at higher power. The dark blobs are ␤A deposits. Each has been outlined in blue, and coloured, using Photoshop colour selection tool. (C) Distribution of ␤A deposits in Case AD3, constructed from the 173 images which make up the image in (A), each processed as in (B).

Fig. 2. Maps of HRDs and ␤A deposits in entorhinal cortex and hippocampus for Case AD3. Original digital photomontages are shown for haem (A) and ␤A deposits (B). Maps are shown for HRDs (C) and ␤A deposits (D). Both forms of deposit are profuse in grey matter and sparse in white matter.

1790

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

Fig. 3. Maps of entorhinal cortex in control case (C4) showing HRDs (A) and ␤A (B) deposits. In control (dementia-free) cases, such as this, the densities of both forms of deposit were lower than in AD or DS/AD cases, and there was a tendency for HRDs to be more numerous than ␤A deposits. The similarity in the distributions of the two forms of deposit is shown in (C), in which the maps in (A) and (B) are superimposed.

␤A (Fig. 2A), the other for haem (Fig. 2B). The distributions of ␤A deposits and HRDs in these sections were then mapped (Fig. 2C and D). Similarities between the distributions are apparent upon inspection. Both forms of deposit were dense throughout the cortical laminae and sparse in subcortical white matter, with the grey/white boundary distinct. In several control cases, typically those with a Braak and Braak Stage II ranking, only small numbers of ␤A and haem deposits were found, most in entorhinal and hippocampal areas. Even in the tissue from the youngest case (Case C1, aged 27 years), two to three deposits were detected in all sections examined (data not shown). In some older nondemented patients, such as Case C4 (Fig. 3), HRDs and ␤A deposits were quite frequent in some regions of cortex, in this case in entorhinal cortex. Again both forms of deposit were mostly confined to the cortex and both were more frequent in grey than in white matter. Comparison of HRD and ␤A maps (Fig. 3) shows the degree of similarity between the two distributions. In severe AD cases with and without DS, the distributions of HRDs and ␤A deposits could be analysed in superior frontal cortex (Fig. 4). In non-DS AD (Fig. 4A and B), the sizes and distributions of the two forms of deposit were very similar. Both were dense in the cortex and very sparse in the white matter. In DS with AD (Fig. 4C and D), both forms of deposit were dense in the frontal cortex, and both were relatively sparse in white matter, though more common than in the non-DS AD brain.

Fig. 4. Maps of the distributions of HRDs and ␤A deposits in superior frontal cortex in AD (Case AD2) and DS (Case DS1). (A) AD haem, (B) AD ␤amyloid, (C) DS haem and (D) DS ␤A. There was considerable variation between cases in the density of deposits in white matter (more in the DS case). In this feature also, HRDs and ␤A deposits co-varied.

The co-occurrence of HRDs and ␤A deposits was remarkably consistent, and was true independent of age, and of a history/non-history of dementia. Where HRDs occurred, ␤A deposits occurred; where one was sparse or dense the other sparse or dense. The most obvious difference in the distributions of the two forms of deposit is apparent in Fig. 3; where the densities of both forms of deposit were relatively low, HRDs were more common. 3.3. Proximity analysis of spatial relationship between βA deposits and vessels Proxan was used to outline blood vessels and ␤A deposits, keeping the populations separate. Proxan tests the hypothesis that one set of features (active set, in this case ␤A deposits) is non-randomly distributed compared to the other set (reference set, blood vessels). The quality of the outlining possible is shown in Fig. 5A and C. The deposit/vessel distances are then defined and measure by Proxan are shown in Fig. 5B and C (yellow). Many of these yellow ‘distances’ appear as blobs rather than lines. Each blob is an area of overlap of a deposit and a vessel; in these cases the distance was recorded as the negative of the diameter of a circle equal in area to the blob. Fig. 5D shows the outcome of Proxan’s

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

1791

Fig. 5. Proxan analysis of an image of entorhinal cortex in Case AD2 labelled for ␤A and blood vessels. (A) Digitally captured image from a section of entorhinal cortex (Layers V–VI) labelled to demonstrate ␤A deposits (red, antibody to ␤-amyloid) and blood vessels (brown, antibody to collagen IV). Scale bar also applies to (B). (B) Using the colour selection tool in Photoshop, the colour difference between ␤A deposits and vessels was intensified. Proxan determines the object edges, for the two populations independently, and discards the background. Proxan next identifies and measures the shortest distance from each deposit to the nearest vessel (yellow lines). Where a deposit overlaps a vessel Proxan plots the overlap, and records the ‘distance’ between deposit and vessel as a negative distance; see Section 2 for the algorithm used to record overlaps. (C) Magnified view of central region of (A), showing the ability of Proxan to outline amyloid deposits and vessels, shown here in their original colours. (D) Statistical report from Proxan for the images in (A–C). In the graph, the X-axis is a normalised distance measurement, and the Y-axis is the frequency of distance measurements; essentially, the graph represents a smoothed frequency/distance histogram. The red curve is the distribution obtained from the observed pattern of amyloid deposits, while the blue curve is the distribution obtained from the 500 random-set measurements. Where it is darkest, this blue curve shows the average frequencies obtained; the lighter regions extend to the maxima and minima. The vertical green line represents zero distance. Note that the major deviation of the observed curve from the random is at small negative values, indicating that overlaps of amyloid and vessels are significantly more common in the observed relationships. Proxan uses a χ2 test to assess the probability that the observed vessel/deposit distances could be sampled from a random distribution of lesions. When the probability is low, Proxan states a conclusion (‘objects are related’) and gives the confidence level of the conclusion. Effectively, this is the probability that the conclusion is wrong. The confidence level is reported to four decimal places.

analysis. The vertical green line represents zero distance. In the observed set of distances (red curve) small negative distances (i.e. overlaps) were more frequent. The 500 sets of distance calculated by Proxan after 500 random redistributions of the deposits (the vessels remaining fixed), yielded the distribution shown in blue. The frequency/probability with which the observed (red) distribution could occur in the random set was low (p < 0.00001). For all cases studied, non-demented and AD with and without DS, the observed distances between ␤A deposits and vessels were less than in randomised sets, with low probabilities that the observed distribution could be observed where deposits were in fact randomly related to vessels (p < 0.001; Table 1). We have previously shown that HRDs have a similarly close spatial relationship to blood vessels in the same cases [15].

3.4. Colocalisation of haem and βA deposits Maps of haem and ␤A deposits have similar distribution patterns (Figs. 2–4). Proximity analysis shows a consistent association of both lesions with blood vessels (Fig. 4). In favourable tissue sections, we were also able to show colocalisation of both lesions, in two ways. Firstly, we compared HRDs and ␤A deposits in matched areas of adjacent sections (Fig. 6A and B). The comparison shows the similarity between HRDs and ␤A deposits in their size, density and distribution. For example, both tended to be sparse in the most superficial layers of the cortex (Layer 1, top of images). On a few occasions it seemed possible to identify the same lesion in the two sections (asterisks in Fig. 6A and B). Secondly, in some sections we attempted to label the same section for both

1792

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

Fig. 6. Evidence of one-to-one colocalisation of HRDs and ␤A deposits. (A) Haem and (B) ␤-amyloid(B) labelling in adjacent sections of superior frontal cortex in Case AD5. The pial surface is to the upper left. Not every HRD can be matched with a ␤A deposit (due to geometric limitations of using parallel sections), but asterisks mark transected deposits that can be matched between the two sections. (C) ␤A labelled and (D) Prussian blue labelled plaques in Case AD5. (E) Examples of deposits double-labelled for haem (blue) and ␤A (red). This direct colocalisation (black) was possible only occasionally, because the haem reaction usually destroyed the amyloid antigen. Case AD5 entorhinal cortex. (F) The DAB reaction product in peroxidase linked immmunohistochemistry for ␤A (brown) is occasionally stable enough to withstand long Prussian blue incubations. This field shows the colocalisation of ␤A and haem (black) within a senile plaque. Case AD10 superior frontal cortex.

haem and ␤-amyloid. In a few cases the harsh Prussian blue processing did not completely eliminate the ␤A labelling. In those cases, colocalisation of HRDs and ␤A deposits was readily demonstrable (Fig. 6B–D). 3.5. Colocalisation of βA and collagen We have shown previously that HRDs correspond to areas where collagen debris and clotting-related factors, markers of blood or blood vessels, are found. In sections labelled for collagen, vessels are labelled clearly (brown in Fig. 7) and collagen debris can be seen. As already noted, ␤A deposits formed around or close to vessels, and the collagen of vessels was commonly observed within the deposits.

4. Discussion The present study examines the relationship between ␤A deposits and HRDs (sites of microhaemorrhages in the cerebral cortex) and the relationship between ␤A deposits and cortical vessels. Our results indicate that the distribution of ␤A deposits closely resembles that of HRDs in the cerebral cortex of all cases examined. Independent of age or diagnosis of AD or DS, where HRDs are profuse (or sparse), ␤A deposits are profuse (or sparse). Further, ␤A deposits appear

to form around small blood vessels (capillaries, arterioles, venules), and this close apposition has been confirmed by a proximity analysis with a high statistical significance. Where we were able to label the same section for both haem and ␤A, colocalisation of HRDs and ␤A deposits was demonstrable. Finally, ␤A deposits colocalise with the collagen of blood vessels, an association previously demonstrated for HRDs [15]. Although incidence estimates vary, vascular pathology is common in AD, with some autopsy series showing up to 90% of confirmed AD cases with significant vascular lesions, including small haemorrhages [17,27,28,45]. These reported haemorrhages are considerably larger than described here; even ‘microinfarcts’ resolvable in the living brain by MRI are at least 1 mm in diameter [51]. The present study documents an abundance of much smaller (<200 ␮m diameter) haemorrhages in the AD and AD/DS brain as well as in cases with minimal plaque load. These microhaemorrhages have a similar size and distribution to, and colocalise with, ␤A deposits. It is unclear whether larger vascular lesions simply add to, magnify or hasten disease progression/cognitive decline in AD, however, there are several lines of evidence that support a central role for cerebral microvascular dysfunction in the pathophysiology of AD. Abnormalities in capillary morphology have been described in AD, leading some to propose that capillary dystrophy (such as basement membrane

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

1793

Fig. 7. Colocalisation of ␤A and collagen. The collagen label shows vessels, but also shows less structured labelling within ␤A deposits. This colocalisation is of interest because collagen is present in the brain only in association with endothelial basement membrane. Scale bar = 50 ␮m for (A–F). (A) Vessels passing through a ␤A deposit (Case AD10). (B) Vessels within a ␤A deposit (Case DS2). (C and D) Vessels circle and converge on these ␤A deposits. Diffuse collagen labelling is apparent within the deposit in (D) (superior frontal cortex, AD1 and AD9). (E) This deposit is adjacent to a capillary branch point and contains distinct collagen-labelled debris (Case AD6 subiculum).

thickening and collagen accumulation) contributes directly to the progression of the disease [17,20] and that chronic hypoperfusion may promote ␤A deposition [5]. Endothelial abnormalities are also present in AD [28] and we suggest that the likely vascular weakening leaves the capillary prone to rupture. In addition, markers of clotting and fibrinolysis abnormalities (e.g. increased thrombomodulin, von Willebrand factor and fibrinogen) can be detected in plasma in very early stages of AD [3,34]. The present study builds on this prior evidence of capillary abnormalities in AD, and shows

a striking correlation between one of the classic diagnostic features of the disease (␤A deposits) and sites of capillary bleeding. We suggest that that ␤A deposits, the classic markers of AD, appear at sites of microhaemorrhages. Haem is a potent neurotoxin [41]. Within 1 day following intracerebral haemorrhage, blood-borne and resident microglia begin to metabolise and compartmentalise free haem [31,52]. Haemosiderin, a highly stable iron-binding protein, remains in microglia and is an enduring marker of previous bleeds [31,52]. The intracellular location of haem in

1794

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

the perivascular deposits is not demonstrated here; however, microglia are the likely site of the majority of the perivascular iron. 4.1. Disease severity and microhaemorrhage/βA colocalisation Sampling of tissue with both severe (Braak [4] Stage VI) and clinically silent (Braak [4] Stage I) tangle burden, and from cases with Down syndrome and confirmed AD, strengthens the conclusions of this investigation. We find that the colocalisation of ␤A and haem is independent of dementia and of the severity of AD neuropathology. Staging the neuropathology in this manner provides a means of categorisation of severity of AD neuropathology. We recognise that there are limitations in extrapolating the degree of neuronal pathology to cognitive status. Discrepancies may arise due to the inhomogeneous distribution of lesions in tissue samples [24] as well as potential inter-individual differences in the impact of tangle burden determined post-mortem on measures of cognitive status obtained ante mortem. Nonetheless, the staging protocol provides a simple, reliable, if low resolution, means to categorise the severity of neuropathology in order to ensure a breadth of patient sampling and provides a starting point for studying microhaemorrhages across a range of neurofibrillary tangle loads. The impact of accumulated microhaemorrhages on cognitive erosion in AD is an important question. In our sampled cases, the numbers and distribution of HRDs are very similar to the numbers and distribution of plaques. A strong correlation between plaque density, tangle density and cognitive loss has been established in many studies [1,19,24]. It is likely therefore that a strong correlation will emerge between the density of HRDs and cognitive loss. It will be interesting to learn whether HRDs are more (or less) effective predictors of cognitive loss than tangles and plaques. A major focus for further study will be to test the possibility that microhaemorrhages contribute to NFT formation, neuronal death and cognitive dysfunction. 4.2. Disease progression and microhaemorrhages There is a lively debate on the spatial/temporal link between ␤A deposits and NFTs and the progression of neuropathology over neuroanatomical pathways (see Ref. [44] and commentary). The colocalisation of microhaemorrhages and ␤A deposits presents a number of mechanisms by which neurons might be damaged both locally and distally and could provide some insight into this discussion. In terms of the present findings, there are several variables to be assessed in rationalising the possible progression of pathology across neurocircuitry: microhaemorrhagic damage to neuronal terminals, damage to soma and dendrites, and damage to fibres of passage, and finally transynaptic damage (e.g. by excitotoxicity). A few key pieces of evidence may be significant in this debate. Neurons with terminals on or near cortical

capillaries, such as the nucleus basalis [43], are vulnerable to tangle formation early in AD [14] and ipsilateral to large cortical stroke [29]. The vulnerability of the nucleus basalis in hypertension has been reported as a potential mechanism in AD pathophysiology [20]. Other cell groups with perivascular terminals, such as the dorsal raphe [42], are also susceptible in AD [25]. With both anterograde and retrograde degeneration emanating from sites of microhaemorrhage, the disease could progress along neurocircuitry, even if the primary event were non-neuronal. A greater number of terminals in particular cortical areas would increase the probability of damage following cortical microhaemorrhage, and as such, a widely broadcast output and large input catchment leaves the entorhinal hub vulnerable. Thus the neurocircuitry model of AD progression becomes more compelling with the initial trigger in the model modified to reflect the occurrence of microhaemorrhage at the site of plaque formation. Further analysis of correlations between microvascular anatomy and microhaemorrhage in vulnerable brain regions is warranted. 4.3. The source of the amyloid in senile plaques There are several potential sources of ␤A deposits within the brain parenchyma. Plaque-associated microglia and astrocytes are amyloid-rich, but it remains unclear whether they are sources [7,54] modifiers [38] or phagocytic sinks [16,38,53] for the protein. Dying neurons may release amyloid [11], and in humans and in amyloid precursor protein (APP) transgenic mice, ␤A is found intraneuronally [49]. Amyloid is also associated with smooth muscle cell basement membrane in cerebral amyloidosis [12]. Cerebral microvessel endothelial cells [33] and smooth muscle cells can produce APP [22]. Because of the abundance of amyloid and precursors in plasma [35] and blood cells, particularly platelets [8,10] some have suggested that parenchymal amyloid may be haematogenous [3]. Phagocytic macrophages can proteolytically process platelet APP to ␤A [18], providing means and mechanism for depositing intravascular ␤A. Routes by which circulating ␤A and precursors might reach the brain parenchyma are suggested by the colocalisation of ␤A and microhaemorrhage markers in our study: a plausible source of intraparenchymal amyloid is macrophage processing of platelet APP. We cannot reject an intracerebral origin for ␤A, nor can we rule out the possibility that blood-derived amyloid and its precursor may be supplemented or modified by brain-resident cells (neurons, astrocytes, microglia, endothelial cells), but we would widen the debate to include blood-borne sources. An important issue to be resolved is whether amyloid deposition precedes or follows microhaemorrhage. Excess amyloid may damage endothelium and occlude vessels. Intravascular ␤A can alter vasoactivity [50], compromise the blood brain barrier [47] and provoke blood vessel rupture [12]. In contrast, ␤A is also well positioned to interact with clotting cascades and seal vascular breaches [2,22]. APP transgenic mice immunised with antibodies against

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

various epitopes of ␤A show cerebral haemorrhages [39,40]. In human trials of ␤A immunotherapy, inflammation, cerebral haemorrhage and small vessel pathology have been noted [21], possibly accounting for the encephalitis that plagued these clinical investigations. 4.4. Discrepancies The present results demonstrate striking co-variance of HRDs and ␤A deposits, relative to each other and to blood vessels. The co-variance is not perfect. The most interesting discrepancy is shown in Fig. 3, from a case without dementia. Both HRDs and ␤A deposits were present, at lower density than in cases with dementia. The two deposits occurred in the same areas of cortex and white matter, however, HRDs tended to be more common. This difference could result from a technical factor (pre- or post-mortem dispersion of amyloid, for example), which led to incomplete labelling of ␤A deposits. Alternatively, the difference may be real, suggesting that microhaemorrhages form first, with amyloid depositing at the site of bleeding. This idea now deserves testing. A previous study [30] reported a 60–77% ‘association’ between senile plaque and vessels, and ∼8–13% penetration of plaques by vessels, yet reached a skeptical conclusion about the relationship. We have sought evidence that would allow greater confidence, by using thicker sections (45 ␮m rather than 7 ␮m), multiple labels (haem, amyloid, collagen) and a specially developed proximity analysis (Proxan, www.bendor.com.au/Proxan). This evidence suggests that it is time to move past earlier caution and explore the relationship between neuronal death, tangles, plaques, cognitive loss and the microvasculature. Recent work has shown that tangle stage may not be the best predictor of cognitive function if the non-homogeneously distributed amyloid deposits are thoroughly sampled [9]. We find that, independent of tangle staging, amyloid deposits and haemorrhages are perivascular. However, the discordance of cognitive stage and tangle stage requires now a direct assessment of the degree of neuronal degeneration associated with microhaemorrhage. The possibility that the protective effect of anti-inflammatory medication, evident from epidemiological studies (as assessed in Ref. [48]), is mediated by stabilisation of the microvasculature also deserves exploration. In conclusion, we find that microhaemorrhages co-vary with plaques and suggest they are the same sites of pathology. These findings suggest that microvascular breakdown begins early in the disease, and that the stabilisation of microvasculature will be valuable clinically in prevention as well as treatment. This conclusion raises the possibility that breakdown of the microvasculature is the proximal cause of AD. The present data show that breakdown of the cerebral microvasculature is an integral part of the pathogenesis of Alzheimer-like dementias, rather than a concomitant process. We propose that the present findings provide a new avenue for in vivo diagnosis, such as measures of microvascular breakdown in circulation.

1795

Acknowledgements This work has been supported by The Sir Zelman Cowen Universities Fund. The authors acknowledge the important tissue resource supported by The National Health and Medical Research Council Network for Brain Research into Mental Health. Brain tissue was obtained Victorian Brain Bank Network, The University of Melbourne, The Mental Health Research Institute of Victoria, Victorian Forensic Institute of Medicine and Neurosciences Australia and from The Department of Pathology, The University of Sydney.

References [1] Arriagada PV, Growden JH, Hedley-Whyte T, Hyman BT. Neurofibrillary tangles but not senile plaques parallel duration and severity of Alzheimer’s disease. Neurology 1992;42:631–9. [2] Atwood CS, Bowen RL, Smith MA, Perry G. Cerebrovascular requirement for sealant, anti-coagulant and remodeling molecules that allow for the maintenance of vascular integrity and blood supply. Brain Res Rev 2003;43(1):164–78. [3] Borroni B, Akkawi N, Martini G, Colciaghi F, Prometti P, Rozzini L, et al. Microvascular damage and platelet abnormalities in early Alzheimer’s disease. J Neurol Sci 2002;203–204:189–93. [4] Braak H, Braak E. Neuropathological staging of Alzheimer-related changes. Acta Neuropathol (Berl) 1991;82:239–59. [5] Brown WR, Moody DM, Thore CR, Challa VR. Cerebrovascular pathology in Alzheimer’s disease and leukoaraiosis. Ann N Y Acad Sci 2000;903:39–45. [6] Buee L, Hof PR, Bouras C, Delacourte A, Perl DP, Morrison JH, et al. Pathological alterations of the cerebral microvasculature in Alzheimer’s disease and related dementing disorders. Acta Neuropathol (Berl) 1994;87(5):469–80. [7] Busciglio J, Gabuzda DH, Matsudaira P, Yankner BA. Generation of beta-amyloid in the secretory pathway in neuronal and nonneuronal cells. Proc Natl Acad Sci USA 1993;50(5):2092–6. [8] Bush AI, Martins RN, Rumble B, Moir R, Fuller S, Milward E, et al. The amyloid precursor protein of Alzheimer’s disease is released by human platelets. J Biol Chem 1990;265(26):15977–83. [9] Bussiere T, Gold G, Kovari E, Giannakopoulos P, Bouras C, Perl DP, et al. Stereologic analysis of neurofibrillary tangle formation in prefrontal cortex area 9 in aging and Alzheimer’s disease. Neuroscience 2003;117(3):577–92. [10] Chen M, Inestrosa NC, Ross GS, Fernandez HL. Platelets are the primary source of amyloid beta-peptide in human blood. Biochem Biophys Res Commun 1995;213(1):96–103. [11] Chen XH, Siman R, Iwata A, Meaney DF, Trojanowski JQ, Smith DH. Long-term accumulation of amyloid-beta, beta-secretase, presenilin-1, and caspase-3 in damaged axons following brain trauma. Am J Pathol 2004;165(2):357–71. [12] Coria F, Rubio I. Cerebral amyloid angiopathies. Neuropathol Appl Neurobiol 1996;22:216–27. [13] Cullen KM. Perivascular astrocytes within Alzheimer’s disease plaques. Neuroreport 1997;8:1961–6. [14] Cullen KM, Halliday GM. Neurofibrillary degeneration and cell loss in the nucleus basalis in comparison to cortical Alzheimer pathology. Neurobiol Aging 1998;19:297–306. [15] Cullen KM, K´ocsi Z, Stone J. Vascular relationships of haem-rich deposits in the aging cerebral cortex. J Cereb Blood Flow Metab 2005;25(12):1656–67. [16] D’Andrea MR, Cole GM, Ard MD. The microglial phagocytic role with specific plaque types in the Alzheimer disease brain. Neurobiol Aging 2004;25(5):675–83.

1796

K.M. Cullen et al. / Neurobiology of Aging 27 (2006) 1786–1796

[17] de la Torre J. Impaired cerebromicrovascular perfusion. Summary of evidence in support of its causality in Alzheimer’s disease. Ann N Y Acad Sci 2000;924:136–52. [18] De Meyer G, De Cleen D, Cooper S, Knaapen M, Jans D, Martinet W, et al. Platelet phagocytosis and processing of beta-amyloid precursor protein as a mechanism of macrophage activation in atherosclerosis. Circ Res 2002;90(11):1197–204. [19] Duyckaerts C, Bennecib M, Grignon Y, Uchihara T, He Y, Piette F, et al. Modeling the relation between neurofibrillary tangles and intellectual status. Neurobiol Aging 1997;18(3):267–73. [20] Farkas E, De Vos RA, Jansen Steur EN, Luiten PG. Are Alzheimer’s disease, hypertension, and cerebrocapillary damage related? Neurobiol Aging 2000;21(2):235–43. [21] Ferrer I, Boada Rovira M, Sanchez Guerra ML, Rey MJ, CostaJussa F. Neuropathology and pathogenesis of encephalitis following amyloid-beta immunization in Alzheimer’s disease. Brain Pathol 2004;14(1):11–20. [22] Frackowiak J, Mazur-Kolecka B, Wisniewski HM, Potempska A, Carroll RT, Emmerling MR, et al. Secretion and accumulation of Alzheimer’s beta-protein by cultured vascular smooth muscle cells from old and young dogs. Brain Res 1995;676(1):225–30. [23] Gearing M, Mirra SS, Hedreen JC, Sumi SM, Hansen LA, Heyman A. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part X. Neuropathology confirmation of the clinical diagnosis of Alzheimer’s disease. Neurology 1995;45(3):461–6. [24] Giannakopoulos P, Herrmann FR, Bussiere T, Bouras C, Kovari E, Perl DP, et al. Tangle and neuron numbers, but not amyloid load, predict cognitive status in Alzheimer’s disease. Neurology 2003;60(9):1495–500. [25] Halliday GM, McCann HL, Pamphlett R, Brooks WS, Creasy H, McCusker E, et al. Brainstem serotonin-synthesizing neurons in Alzheimer’s disease: a clinicopathological correlation. Acta Neuropathol (Berl) 1992;84:638–50. [26] Jellinger KA. Alzheimer disease and cerebrovascular pathology: an update. J Neural Transm 2002;109(5–6):813–36. [27] Jellinger KA. The pathology of ischemic-vascular dementia: an update. J Neurol Sci 2002;203–204:153–7. [28] Kalaria RN, Hedera P. Differential degeneration of the cerebral microvasculature in Alzheimer’s disease. Neuroreport 1995;6(3):477–80. [29] Kato T, Hirano A, Katagiri T, Sasaki H, Yamada S. Neurofibrillary tangle formation in the nucleus basalis of Meynert ipsilateral to a massive cerebral infarct. Ann Neurol 1988;23:620–3. [30] Kawai M, Kalaria RN, Harik SI, Perry G. The relationship of amyloid plaques to cerebral capillaries in Alzheimer’s disease. Am J Pathol 1990;137(6):1435–46. [31] Koeppen AH. The history of iron in the brain. J Neurol Sci 1995;134S:1–9. [32] Kudo T, Imaizumi K, Tanimukai H, Katayama T, Sato N, Nakamura Y, et al. Are cerebrovascular factors involved in Alzheimer’s disease? Neurobiol Aging 2000;21(2):215–24. [33] Lang IM, Moser KM, Schleef RR. Expression of Kunitz Protease Inhibitor containing forms of amyloid fl-protein precursor within vascular thrombi. Circulation 1996;94(11):2728–34. [34] Mari D, Parnetti L, Coppola R, Bottasso B, Reboldi G, Senin U, et al. Hemostasis abnormalities in patients with vascular dementia and Alzheimer’s disease. Thromb Haemost 1996;75(2):216–8. [35] Mayeux R, Tang MX, Jacobs DM, Manly J, Bell K, Merchant C, et al. Plasma amyloid beta-peptide 1–42 and incipient Alzheimer’s disease. Ann Neurol 1999;46(3):412–6. [36] Mirra SS, Heyman A, McKeel D, Sumi SM, Crain BJ, Brownlee LM, et al. The Consortium to Establish a Registry for Alzheimer’s Disease

[37]

[38]

[39]

[40]

[41] [42]

[43]

[44]

[45]

[46] [47]

[48]

[49]

[50]

[51]

[52]

[53]

[54]

(CERAD). Part II. Standardization of the neuropathologic assessment of Alzheimer’s disease. Neurology 1991;41(4):479–86. Morris JC, Heyman A, Mohs RC, Hughes JP, van Belle G, Fillenbaum G, et al. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part I. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurology 1989;39(9):1159–65. Nagele RG, Wegiel J, Venkataraman V, Imaki H, Wang KC. Contribution of glial cells to the development of amyloid plaques in Alzheimer’s disease. Neurobiol Aging 2004;25(5):663–74. Pfeifer M, Boncristiano S, Bondolfi L, Stalder A, Deller T, Staufenbiel M, et al. Cerebral hemorrhage after passive anti-A-beta immunotherapy. Science 2002;298:1379. Racke MM, Boone LI, Hepburn DL, Parsadainian M, Bryan MT, Ness DK, et al. Exacerbation of cerebral amyloid angiopathy-associated microhemorrhage in amyloid precursor protein transgenic mice by immunotherapy is dependent on antibody recognition of deposited forms of amyloid. J Neurosci 2005;25:629–36. Regan RF, Rogers B. Delayed treatment of hemoglobin neurotoxicity. J Neurotrauma 2003;20(1):111–20. Reinhard Jr JF, Liebmann JE, Schlosberg AJ, Moskowitz MA. Serotonin neurons project to small blood vessels in the brain. Science 1979;206(4414):85–7. Sato A, Sato Y, Uchida S. Regulation of regional cerebral blood flow by cholinergic fibers originating in the basal forebrain. Int J Dev Neurosci 2001;19(3):327–37. Sch¨onheit B, Zarski R, Ohm TG. Spatial and temporal relationships between plaques and tangles in Alzheimer-pathology. Neurobiol Aging 2004;25(6):697–711. Shi J, Perry G, Smith MA, Friedland RP. Vascular abnormalities: the insidious pathogenesis of Alzheimer’s disease. Neurobiol Aging 2000;21(2):357–61. Snowdon DA. Aging and Alzheimer’s disease: lessons from the Nun Study. Gerontologist 1997;37(2):150–6. Su GC, Arendash GW, Kalaria RN, Bjugstad KB, Mullan M. Intravascular infusions of soluble beta-amyloid compromise the blood–brain barrier, activate CNS glial cells and induce peripheral hemorrhage. Brain Res 1999;818:105–17. Szekely CA, Thorne JE, Zandi PP, Ek M, Messias E, Breitner JC, et al. Nonsteroidal anti-inflammatory drugs for the prevention of Alzheimer’s disease: a systematic review. Neuroepidemiology 2004;23(4):159–69. Takahashi RH, Milner TA, Li F, Nam EE, Edgar MA, Yamaguchi H, et al. Intraneuronal Alzheimer abeta42 accumulates in multivesicular bodies and is associated with synaptic pathology. Am J Pathol 2002;161(5):1869–79. Thomas T, Thomas G, McLendon C, Sutton T, Mullan M. ␤Amyloid-mediated vasoactivity and vascular endothelial damage. Nature 1996;380:168–71. van Dijk EJ, Prins ND, Vermeer SE, Koudstaal PJ, Breteler MM. Frequency of white matter lesions and silent lacunar infarcts. J Neural Transm Suppl 2002;62:25–39. Wagner K, Sharp F, Ardizzone T, Lu A, Clark J. Heme and iron metabolism: role in cerebral hemorrhage. J Cereb Blood Flow Metab 2003;23(6):629–52. Wegiel J, Wisniewski HM, Muzylak M, Tarnawski M, Badmajew E, Nowakowski J, et al. Fibrillar amyloid-beta production, accumulation, and recycling in transgenic mice pancreatic acinar cells and macrophages. Amyloid 2000;7(2):95–104. Wisniewski HM, Wegiel J, Vorbrodt AW, Mazur-Kolecka B, Frackowiak J. Role of perivascular cells and myocytes in vascular amyloidosis. Ann N Y Acad Sci 2000;903:6–18.