Bapineuzumab Alters Aβ Composition: Implications for the Amyloid Cascade Hypothesis and Anti-Amyloid Immunotherapy

The characteristic neuropathological changes associated with Alzheimer’s disease (AD) and other lines of evidence support the amyloid cascade hypothesis. Viewing amyloid deposits as the prime instigator of dementia has now led to clinical trials of multiple strategies to remove or prevent their formation. We performed neuropathological and biochemical assessments of 3 subjects treated with bapineuzumab infusions. Histological analyses were conducted to quantify amyloid plaque densities, Braak stages and the extent of cerebral amyloid angiopathy (CAA). Amyloid-β (Aβ) species in frontal and temporal lobe samples were quantified by ELISA. Western blots of amyloid-β precursor protein (AβPP) and its C-terminal (CT) fragments as well as tau species were performed. Bapineuzumab-treated (Bapi-AD) subjects were compared to non-immunized age-matched subjects with AD (NI-AD) and non-demented control (NDC) cases. Our study revealed that Bapi-AD subjects exhibited overall amyloid plaque densities similar to those of NI-AD cases. In addition, CAA was moderate to severe in NI-AD and Bapi-AD patients. Although histologically-demonstrable leptomeningeal, cerebrovascular and neuroparenchymal-amyloid densities all appeared unaffected by treatment, Aβ peptide profiles were significantly altered in Bapi-AD subjects. There was a trend for reduction in total Aβ42 levels as well as an increase in Aβ40 which led to a corresponding significant decrease in Aβ42:Aβ40 ratio in comparison to NI-AD subjects. There were no differences in the levels of AβPP, CT99 and CT83 or tau species between Bapi-AD and NI-AD subjects. The remarkable alteration in Aβ profiles reveals a dynamic amyloid production in which removal and depositional processes were apparently perturbed by bapineuzumab therapy. Despite the alteration in biochemical composition, all 3 immunized subjects exhibited continued cognitive decline.


Introduction
The most common form of dementia is Alzheimer's disease (AD), currently affecting about 24 million people worldwide with expected doubling of incidence every 20 years [1]. The disease is neuropathologically characterized by the profuse deposition of fibrillar amyloid-b (Ab) peptides, amyloid plaques (AP) and cerebral amyloid angiopathy (CAA) as well as the intraneuronal accumulation of neurofibrillary tangles (NFT) mainly composed of tau protein. The abundance of these lesions has lent support to the amyloid cascade hypothesis as the fundamental causative incident in the pathogenesis of AD. This model has been reinforced by the fact that AP, CAA and NFT are also present in familial AD (FAD) due to presenilin (PS) and amyloid-b precursor protein (AbPP) mutations and are recapitulated in genetically-engineered transgenic (Tg) mice bearing mutated forms of AbPP, PS and tau. Furthermore, individuals with Down's syndrome carrying 3 copies of chromosome 21, the location of the AbPP gene, also develop the neuropathology of AD.
At present, there is no effective disease-modifying treatment for AD. However, the abundant deposits of insoluble Ab and elevated values of soluble oligomeric Ab have prompted the design of multiple active and passive immunotherapies aimed at removing the toxic forms of these peptides [2,3]. An example of one of the passive immunization treatments is bapineuzumab, a humanized monoclonal antibody (3D6) directed specifically against the Nterminal region of Ab (residues 1-5) [4]. Although passive and active immunotherapies have been effective in the removal of Ab in AbPP and PS Tg mice models, the application of these interventions to AD patients has been only partially successful from a neuropathological perspective [5][6][7][8][9], while no clear beneficial modification of the disease course has been observed in clinical trials [7,10,11].
In this study we assess the clinical history, neuropathological and biochemical outcomes in 3 individuals that participated in clinical trials evaluating bapineuzumab [12] (ClinicalTrials.gov Identifier NCT00112073) and who received 2, 6 or 20 immunotherapy doses. We quantified and characterized the levels of soluble and insoluble Ab peptides that remained in the frontal and temporal lobes of bapineuzumab-immunized AD (Bapi-AD) patients and compared them to 4 age-matched non-immunized AD (NI-AD) and 4 age-matched non-demented control (NDC) individuals. AbPP and its C-terminal (CT) peptides as well as the cytokine tumor necrosis factor-a (TNF-a) were also quantified. We also compared the bapineuzumab observations with those obtained from subjects who received AN-1792 immunotherapy.

Clinical and Neuropathological Reports and Treatment Synopsis
The present report deals with the neuropathological and biochemical observations made on 3 Bapi-AD patients and their comparison to NI-AD and NDC individuals. Brain samples from cases #1 and #2 were obtained from the Banner Sun Health Research Institute (BSHRI) Brain and Body Donation Program [13] whose operations have been approved by the Banner Health Institutional Review Board. All subjects enrolled in the Brain and Body Donation Program sign an informed consent approved by the Banner Health Institutional Review Board. The brain tissue for case #3 was provided by the Institute for Memory Impairments and Neurological Disorders and the University of California Alzheimer's Disease Research Center (UCI-ADRC). Participants enrolled in the UCI-ADRC provide informed consent approved by the UCI Institutional Review Board. A summary of subject demographics and neuropathology is presented in Table 1.

Case #1
For a detailed clinico-pathological description of case #1 the reader is referred to our recent publication [14]. In brief, a 79year-old female was diagnosed with AD about 8 years prior to death. The patient received 4 doses of bapineuzumab (each dose 0.5 mg/kg) over a period of 39 weeks in the extension portion of the clinical trial. About 1 month after the last dose, the patient showed symptoms and signs of vasogenic edema on an MRI scan that was cleared prior to her death. There were no apparent signs of disease modification that could be attributed to the immunotherapy treatment. Neuropathological analysis revealed frequent AP and NFT and a concurrent diagnosis of both AD and Binswanger's type of vascular dementia. Case #1 was incorpo-rated in the present biochemical evaluations of cases #2 and #3 to increase the sample size. Furthermore, in the previous analysis of case #1, we only characterized the frontal region, while in this communication we added the temporal region.

Case #2: Clinical History
An 89-year-old man died with a diagnosis of AD about 12 years after symptom onset. The available private medical records include his first presentation, about 11 years prior to death, when he was seen by a neurologist for a several month history of cognitive deterioration following biopsy for temporal arteritis. He had been getting lost even when going for a walk close to home and had been repeating himself frequently. Additionally, he was being treated for depression. On examination he was not oriented to place. He scored 23/30 on the Mini Mental State Examination (MMSE), losing points on delayed recall and orientation. Gait and posture were normal with no focal neurological signs. Brain MRI showed only ectasia of the intracranial segment of the left vertebral artery. An EEG was interpreted as within normal limits. He was started on Aricept for a presumptive diagnosis of early Alzheimer's dementia, but this was quickly switched to Exelon because of gastrointestinal side effects. Over the next 10 years, his cognitive status declined only very slowly as documented by MMSE scores of 20/30, 19/30 and a final score of 21/30 about 1 month prior to death. Over this time period, he developed aggressive behavior and was treated with an anti-psychotic agent. He developed startle myoclonus, a stooped posture and a tremor. An EEG showed only slowing and disorganization. About 1 year prior to death, he had orthostatic hypotension with syncopal episodes and few months prior to death was hospitalized for bradycardia. He received annual standardized neurological and neuropsychological assessments at BSHRI between 2003 and 2011. The diagnostic impression was of dementia due to probable AD, with gait ataxia and essential tremor. His past medical history is otherwise notable for hyperlipidemia, mitral valve prolapse, first degree atrioventricular block, chronic obstructive pulmonary disease, glucose intolerance, thyroid nodules, bilateral cataract extractions, glaucoma and benign prostatic hypertrophy. The family history is notable for late-onset dementia in his mother.
Between Case #2: Neuropathology Report Gross examination. The brain weight at autopsy was 1132 g. The dura was normal and the leptomeninges showed moderate fibrosis. The convexities were symmetrical and showed moderate gyral atrophy of the frontal lobes, moderate to severe gyral atrophy of the parietal lobes and no gyral atrophy of the occipital lobes. No focal lesions were present on the convexities or base of the brain. The circle of Willis showed moderate patchy atherosclerosis. The mammillary bodies were normal in shape, color and size. The temporal lobes and unci showed mild to moderate gyral atrophy. The cerebellum and brainstem were externally normal. Cerebral slices showed moderate to marked enlargement of the posterior horns of the lateral ventricles and mild dilatation elsewhere. The basal ganglia, thalamus and subthalamic nucleus were unremarkable. The amygdala showed moderate atrophy in the left hemisphere and severe atrophy in the right hemisphere with compensatory enlargement of the temporal horns. The head of the hippocampus was mildly atrophied in both hemispheres. The body of the hippocampus and the parahippocampal gyrus were both mildly atrophied. The substantia nigra showed mild depigmentation bilaterally. Respective axial and parasagittal slices of the brainstem and cerebellum were normal.
Microscopic examination. Paraffin sections of the left hemibrain stained with hematoxylin and eosin (H&E) showed, in sections of cerebral cortex, mild to moderate upper layer gliosis. The amygdala and entorhinal cortex showed moderate to marked gliosis. Area CA1 of the hippocampus showed mild gliosis. The basal ganglia were unremarkable. There was mild to moderate gliosis of the hypothalamus. Subthalamic regions including the subthalamic nucleus and mammillary body were unremarkable. The substantia nigra showed no apparent depletion of pigmented neurons while the locus ceruleus was moderately to markedly depleted; there were no Lewy bodies present in either region. The cerebellar superior vermis showed moderate to marked patchy loss of Purkinje cells. Remaining sections of the cerebellum, brainstem and major levels of spinal cord were unremarkable. Large sections stained with H&E showed no significant cerebral white matter rarefaction and no infarcts. There were several mineralized blood vessels in the globus pallidus. Sections stained with Gallyas, Campbell-Switzer and Thioflavine-S methods showed, in neocortical regions, frequent senile plaques of the diffuse type while neuritic and cored plaques had a patchy distribution, ranging from sparse to frequent in the frontal, parietal and occipital lobes, with moderate to frequent densities in the temporal lobe. Neurofibrillary tangles were also patchily distributed, ranging from sparse to moderate to focally frequent in neocortical areas. Tangles were frequent in the amygdala, entorhinal cortex and hippocampal CA1 region. Argyrophilic grains were present at frequent densities in the amygdala, entorhinal cortex and area CA1 of the hippocampus. There were frequent Gallyas-positive glial cells around the circumference of the amygdala; these resembled small astrocytes with spiky processes. Cerebral amyloid angiopathy was present at sparse to moderate to focally frequent densities in most cerebral cortex regions while there were focally moderate densities in the cerebellar leptomeninges. Immunohistochemical staining for phosphorylated a-synuclein showed no evidence of immunoreactive inclusions or associated neurites in the olfactory bulb, brainstem, amygdala or cerebral cortex. Diagnosis: Alzheimer's disease; argyrophilic grains and non-specific glial tauopathy, mesial temporal lobe. Comment: This microscopic examination confirms the clinical diagnosis of AD. Argyrophilic grains are a microscopic finding of uncertain significance; they occur in approximately 25% of cognitively normal older people as well as a similar fraction of those with AD and other aging brain disorders. They are often, as in this case, accompanied by a nonspecific glial tauopathy.

Case #3: Clinical History
This patient was an 86-year-old man with 12 years of education, who presented with a 4-year history of memory impairment of sudden onset prior to his initial evaluation in 1999 at the UCI-ADRC. He was followed longitudinally by the ADRC Clinical Core with slow progressive decline. The patient was screened for participation in the bapineuzumab clinical trial in August of 2006, and enrolled in this double blind randomized multicenter study with a moderate level of cognitive impairment (MMSE 17/30). This patient received only 2 doses of 2.0 mg/kg of bapineuzumab. The first infusion was given in September of 2006, followed by the second infusion 13 weeks later. MRI scans were completed for safety assessment 6 weeks after each infusion and acute vasogenic edema was observed on the safety scan at 6 weeks following the second infusion. Progressive cognitive decline occurred with a MMSE of 8/30 measured just 6 months prior to death. At autopsy, sufficient neuropathology for a diagnosis of AD with Lewy body disease was observed with a Braak stage VI. The APOE genotype of this individual was e3/4.

Case #3: Neuropathology Report
Gross examination. The brain weight at autopsy was 1170 g. On external examination the cerebral gyri and sulci were of normal gross appearance. The cerebral arteries and their major branches at the base of the brain showed moderate focal atherosclerosis, with up to 40% narrowing of the lumen. Coronal sections of the right cerebrum showed the presence of mild gyral narrowing and sulcal widening in the sylvian region. The lateral ventricle did not appear to be enlarged, although the hippocampus was judged to be small. Horizontal sections of the brainstem and cerebellum revealed normal intensity of neuromelanin pigmentation within the substantia nigra and markedly reduced intensity of pigmentation within the locus ceruleus.
Microscopic examination. Blocks of tissue were examined from middle frontal, superior temporal, inferior parietal, calcarine/pericalcarine, and rostral and caudal cingulate cortex and white matter, as well as from hippocampus, amygdala, corpus striatum, thalamus, midbrain, pons, medulla, and cerebellum. Both diffuse and neuritic (Braak and Braak stage C) plaque formation were intense within frontal, temporal, parietal, occipital, and rostral and caudal cingulate cortex and within hippocampal CA1, subiculum, entorhinal-transentorhinal region, and amygdala. Within the cerebral neocortex neuritic plaques were more prominent and densely stained (in both modified Bielschowsky and Ab-stained sections) within the superficial layer than within deeper layers while the number of plaques appeared to be similar, but lighter in stain intensity. These findings were also observed within the hippocampal formation, particularly within the subiculum and, even more strikingly, within the amygdala, where the intensity of Ab immunostaining within many of the neuritic plaques was markedly reduced. Vascular intramural Ab deposition was also observed in moderate degree within frontal and occipital  leptomeningeal and intracortical vessels and, in very mild degree, within amygdala capillaries. Iron deposition, as seen with the aid of Perl's iron stains, was minimal. Neurofibrillary degeneration, which was judged overall Braak stage VI was severe at all sites examined except for the calcarine cortex, where it was absent. Granulovacuolar degeneration and Hirano body formation were readily observed within hippocampal CA1. The concentration of neuromelanin-bearing neurons was normal within substantia nigra and moderately severely reduced within locus ceruleus. Alphasynuclein-immunoreactive intracytoplasmic Lewy inclusion bodies were observed within substantia nigra, amygdala, subiculum, entorhinal cortex and rostral cingulate cortex.

Assessment of Cerebrovascular Amyloid Angiopathy (CAA)
To investigate the degree of vascular amyloidosis in the leptomeninges, these membranes were carefully removed from the convexities and medial aspect of the cerebral hemispheres, prior to the brain coronal sectioning. The leptomeninges were rinsed in phosphate buffered saline (PBS) and frozen at 280uC until the moment of utilization. To remove entrapped intravascular blood, the leptomeninges were rinsed 6 times each with 2 L of cold PBS, and a final rinse with 2 L of deionized water (DW).
The gleaming membranes were spread on the surface of 3 plastic Petri dishes (14 cm diameter) and allowed to dry and adhere to the plastic surface by heating in an oven at 65uC. The membranes were fixed with absolute ethanol for 1 h, rinsed with DW and stained with 1% aqueous Thioflavine-S for 15 min followed by 4 rinses with 70% ethanol to remove unbound fluorochrome and immediately observed in an epifluorescence microscope. For the appraisal of cortical vascular amyloid, 0.5 cm 3 of cerebral cortex were suspended in 600 ml of 50 mM Tris-HCl pH 7.5 buffer containing 5% SDS and 0.01% sodium azide with continuous stirring for 72 h until all tissue was lysed except the tufts of insoluble blood vessels. The blood vessels were rinsed with 4 L of DW, to remove excess SDS, and processed and stained as described above for the leptomeninges.

Immunohistochemistry
Coronal free-floating, formalin-fixed sections of 40 mm thickness were taken using a sliding freezing microtome from Bapi-AD cases  Quantification of Soluble and Insoluble Ab by ELISA A detailed description of ELISA methods has been previously published [23]. In brief, gray matter from the frontal and temporal lobes (100 mg) was gently homogenized in 800 ml of 20 mM Tris-HCl, 5 mM EDTA, pH 7.8, protease inhibitor cocktail (PIC, Roche Diagnostics, Mannheim, Germany), centrifuged at 435,0006g and the Tris-HCl-soluble supernatant collected. The pellet was reconstituted in 600 ml of 90% glass-distilled formic acid (GDFA), centrifuged at 435,0006g and the supernatant collected, dialyzed against deionized water followed by 0.1 M ammonium bicarbonate to remove the GDFA, then lyophilized. The lyophilized material was reconstituted in 500 ml 5 M guanidine hydrochloride (GHCl), 50 mM Tris-HCl, pH 8.0, PIC (Roche), centrifuged and the supernatant collected. Total protein was Figure 5. ELISA quantification of soluble and insoluble Ab in frontal and temporal lobes. A) Frontal cortex Tris-soluble Ab 40 . Notice the sharp and significance difference between the Ab levels of NI-AD and those of the Bapi-AD. B) Temporal cortex Tris-soluble Ab 40 . There was a large mean difference between NI-AD and Bapi-AD, however there was not a statistically significant difference due to the spread of values. Frontal (C) and temporal (D) cortices GDFA/GHCl-soluble Ab 40 . In both cases there were significant elevations in mean Ab 40 in Bapi-AD relative to NI-AD. Frontal (E) and temporal (F) cortices Tris-soluble Ab 42 . In both lobes there were no statistical differences between the NI-AD and Bapi-AD due to the spread in values. Frontal (G) and temporal (H) cortices GDFA/GHCl-soluble Ab 42 . There was a mean decrease in Bapi-AD Ab 42 relative to NI-AD more noticeable in the latter than in the former. The numbers in the Bapi-AD columns (1, 2 and 3), correspond to the cases #s given in Table 1. Statistical analysis was performed using a one-way ANOVA followed by Tukey's multiple comparison test (*p = 0.05-0.01; **p = 0.01-0.001; ***p,0.0001). Abbreviations: Tris, 20 mM Tris-HCl, 5 mM EDTA, pH 7.8, plus protease inhibitor cocktail; GDFA, glass-distilled formic acid; GHCl, 5 M guanidine hydrochloride, 50 mM Tris-HCl, pH 8.0, plus protease inhibitor cocktail; NDC, non-demented control; NI-AD, non-immunized Alzheimer's disease; Bapi-AD, bapineuzumab-immunized Alzheimer's disease. doi:10.1371/journal.pone.0059735.g005

Quantification of Tumor Necrosis Factor-a (TNF-a) by ELISA
Gray matter (100 mg) was homogenized in 10 volumes of 20 mM HEPES, 1.5 mM EDTA, pH 7.4, PIC (Roche) as previously communicated [23] and total protein concentrations determined (BCA protein assay, Pierce). Human TNF-a levels were quantified using an ELISA kit (PromoKine, Heidelberg, Germany) following the manufacturer's directions.

Statistical Analyses
Statistical calculations were performed with GraphPad Prism 5 software (La Jolla, CA) using a one-way ANOVA followed by Tukey's multiple comparison test. Statistical significance was determined by p values of #0.05. Table 1 shows a detailed account of the ages, gender, postmortem intervals, brain weights, last MMSE scores, APOE genotypes and neuropathological assessments of the NDC, NI-AD and Bapi-AD cases employed in this study. In each of the NDC and NI-AD groups there were 2 females and 2 males, while in the Bapi-AD group there was 1 female (case #1) and 2 males (cases #2 and #3). The mean ages of the individuals in these groups were 80, 91, and 85 years, respectively. The average brain weight of the Bapi-AD individuals was 1101 g which was very close to the average weight of the NI-AD cases (1059 g), suggestive of some degree of atrophy when compared to the mean weight of the NDC group (1281 g). The Bapi-AD subjects had a final MMSE mean value of 10 while the NI-AD and NDC MMSE values were 11 and 28, respectively. The APOE e4 allelic frequencies in the 3 groups under study were: NDC = 13%, NI-AD = 38% and Bapi-AD = 50%, respectively. The AP densities were zero to sparse in the NDC cases while they were frequent in the NI-AD and Bapi-AD cases (Table 1), as can be appreciated in Figure 1 and Figure 2. Amyloid plaques were elevated between NDC and NI-AD/Bapi-AD groups ( Table 1). Neurofibrillary tangle Braak stages were similar between NI-AD and Bapi-AD individuals ( Table 1). There was a wide variation in CAA content (Table 1): none to mild in NDC, none to mild in NI-AD (with severe in the occipital lobe of cases #12 and #13) and moderate to severe in the case of the Bapi-AD (severe in the occipital and parietal lobes of case #1). As shown in Table 1, there was a relatively wide variation in total WMR score among the 3 groups.

Results
As can be appreciated from Figure 1A, B, E, F and Figure 2A, B, E, F, based on semi-quantitative visual analyses, the Bapi-AD subjects harbored AP densities similar to those observed in NI-AD cases ( Figure 1C and D and Figure 2C and D), which were matched for APOE genotype. No histological evidence of extensive AP removal, patchy or otherwise, was visually apparent in the 3 Bapi-AD subjects. Furthermore, the density of CAA in the isolated vascular network of cerebral  Figure 1G). Our previously published report on Bapi-AD [14], which corresponds to case #1 of the present study, showed severe leptomeningeal CAA (see Figure 1D in reference [14]). Leptomeningeal Ab was not assessed for case #3.
The Bapi-AD cases #1 and #2 exhibited a dense distribution of microglia in the frontal and temporal cortices, as stained by the CD68 antibody, relative to the NDC cases; however the microglial density did not differ between Bapi-AD cases and NI-AD cases ( Figures 3A, B and C). This observation was confirmed by the HLA-DR antibody ( Figure 3D, E and F). HLA-DR immuno-reactive activated microglia were seen in abundance within cerebral cortex and white matter of Bapi-AD case #3, and were even more numerous within subiculum and, particularly, within the amygdala ( Figure 3G). Cases #1 and #2 showed no T-cell lymphocyte infiltration in either the frontal and temporal cortices. In contrast, the Bapi-AD case #3 showed sparse but unequivocal infiltration by CD3-immunoreactive T-lymphocytes, typically in perivascular location, within cerebral cortex and white matter. Tlymphocytes were observed in higher concentrations within subiculum and, even more strikingly, within the amygdala ( Figure 3H). Immunofluorescence staining with GFAP demonstrated fibrous astrocytes ( Figure 4A). The NI-AD cases ( Figure 4B) appeared to have more voluminous astrocytic cytoplasm and more frequent processes than the Bapi-AD group ( Figure 4C). Despite the higher levels of the proinflammatory cytokine TNF-a in the temporal lobe of the Bapi-AD cases relative to the NI-AD cohort, we did not observe commensurate changes in the temporal lobe microglia morphology or density.
Quantification of Ab 40 peptides by ELISA revealed that in the frontal lobe and temporal lobe (Figures 5A and 5B) the Trissoluble fraction was, on the average, strongly increased in the Bapi-AD cases relative to NI-AD by 18-fold and 32-fold, respectively, although only the frontal lobe levels ( Figure 5A) reached statistical significance (p,0.05). In the temporal lobe, the soluble Ab 40 was not statistically significant due to the spread of the values in the Bapi-AD ( Figure 5B). In comparison, the NI-AD group had only 2.0-fold (frontal lobe) and 2.7-fold (temporal lobe) increases in amounts of Tris-soluble Ab 40 as compared to the NDC cohort (Figures 5A and 5B). A similar, but not as dramatic, situation was observed for the Tris-extracted Ab 42 pool where the average levels were 2.3-fold and 1.4-fold increased in the frontal and temporal lobes respectively, between the NI-AD and Bapi-AD cases ( Figure 5E and 5F).
Comparisons between Bapi-AD and NI-AD groups revealed an apparently selective removal of GDFA/GHCl-solubilized fibrillar Ab 42 , (Figure 5G and 5H), coupled with a concomitant increase of insoluble fibrillar Ab 40 ( Figure 5C and 5D). In the frontal lobe, the mean GDFA/GHCl-soluble Ab 42 level in the Bapi-AD group was 429 ng/mg total protein, while the mean level in the NI-AD group as compared to 549 ng/mg total protein in the NI-AD group, a 1.3-fold decrease in immunized cases ( Figure 5G). In contrast, the mean GDFA/GHCl-soluble Ab 40 levels were elevated 18-fold in the Bapi-AD group compared to the NI-AD subjects (113 ng/mg and 6.2 ng/mg, respectively) ( Figure 5C). In the temporal lobe, a more dramatic apparent relative loss of GDFA/GHCl-soluble Ab 42 was evident, decreasing from 461 ng/ mg in the NI-AD cases to an average of 231 ng/mg in the Bapi-AD cases, a 2-fold reduction ( Figure 5H). Mean GDFA/GHClsoluble temporal Ab 40 levels increased from 5.0 ng/mg in the NI-AD group to a mean of 43 ng/mg in the Bapi-AD cases, an 8.6fold increase ( Figure 5D).
Remarkably, Ab 40 levels increased while Ab 42 was decreased in the Bapi-AD subjects relative to the NI-AD cohort, resulting in no significant net change of GDFA/GHCl-soluble total Ab levels in the Bapi-AD compared to NI-AD subjects ( Figure 6C and 6D). On the other hand, the levels of Tris-soluble total Ab were significantly increased in the frontal cortex of the Bapi-AD cohort compared to NI-AD ( Figure 6A). Only Bapi-AD case #3 had elevated Tris-soluble total Ab relative to the other 2 Bapi-AD subjects in the temporal lobe ( Figure 6B). This profound shift in Ab 40 accumulation in the Bapi-AD cohort was also evident in the GDFA/GHCl-soluble fractions from the frontal and temporal lobes in which Ab 42 :Ab 40 ratios shifted dramatically after immunotherapy. For the frontal and temporal lobes, the average Ab 42 :Ab 40 ratios in the GDFA/GHCl-soluble fraction were 107 and 112 in the NI-AD group, respectively. These ratios were significantly decreased to 5.0 and 5.5 in the Bapi-AD cases, respectively (Table 2). Similarly, as shown in Table 2, the Tris-soluble average Ab 42 :Ab 40 ratios were significantly reduced in the immunized cases relative to the NI-AD cases (Bapi-AD = 0.31 vs. NI-AD = 2.51; 8.1-fold) in the frontal lobe, and in the temporal lobe (Bapi-AD = 0.35 vs. NI-AD = 2.19; 6.3-fold).
In the frontal and temporal lobes the pro-inflammatory TNFa molecule was on the average increased in the Bapi-AD cases relative to the NI-AD cases, but this difference reached a level of statistical significance only in the temporal lobe (Figures 7A and 7B). As described above, removal of Ab in the Bapi-AD was more extensive in the temporal lobe in comparison to the NI-AD group which could explain the relative increase in this pro-inflammatory cytokine.
Evaluation of the AbPP content by Western blot using the CT20AbPP antibody revealed no statistically significant differences in the frontal and temporal lobes between the Bapi-AD and NI-AD cases (Figures 8A and 8B). The CT20AbPP antibody also detects the CT99 (,11 kDa) and CT83 (,9 kDa) ( Figures 8A and 8B), the peptides derived from the AbPP proteolysis by the band a-secretases, respectively. These peptides were significantly increased in the NI-AD frontal lobe when compared to the NDC ( Figure 8A), but showed no differences from the Bapi-AD group. Identification of total tau by the HT7 antibody revealed the expected multiplicity of isoforms ranging from ,58 to ,40 kDa. However, the overall quantities of these isoforms did not show statistical differences among the studied groups or brain regions (Figures 8C and 8D).

Discussion
Our investigation revealed: 1) On the basis of histology and number, APs were apparently undisturbed in the Bapi-AD individuals compared to age-and APOE genotype-matched NI-AD cases; 2) Microgliosis and astrogliosis in the Bapi-AD and NI-AD cases were increased relative to the NDC cohort while T-cell lymphocyte infiltration was only present in Bapi-AD case #3; 3) In the Bapi-AD subjects, the total amounts of Ab 42 were reduced relative to the Ab 40 levels, which were concomitantly elevated; 4) A remarkable reduction in the Ab 42 :Ab 40 ratios between the Bapi-AD and NI-AD cohorts; and 5) A lack of noticeable difference in the relative levels of AbPP, CT99 and CT83 between Bapi-AD and NI-AD despite the alterations in Ab 42 :Ab 40 ratios. It is unclear whether the shift in Ab peptide composition following immunotherapy represents a true compensatory physiological response to AP and CAA disaggregation or simply reflects a new equilibrium due to continuous AbPP processing under changed environmental circumstances. Interestingly, in the Bapi-AD patients there was an apparent increase in the Tris-soluble Ab pool relative to those levels observed in the NI-AD cases. The pathophysiological consequences of increased soluble Ab levels that are unable to exit the brain are unknown.
A detailed examination of the individual responses to bapineuzumab in terms of the amount of antibody per infusion and number of administered doses relative to Ab peptide levels did not show a clear pattern for the Tris-soluble Ab fractions in both the frontal and temporal lobes. However, in the Bapi-AD individuals the levels of the GDFA/GHCl-soluble Ab 40 and Ab 42 fractions had the tendency to cluster together.
Germane to the present study are our previous observations of similar alterations in the Ab 42 :Ab 40 ratio in the gray matter of the frontal lobes of 7 patients actively immunized with the AN-1792 (containing the Ab 42 peptide) vaccine which were compared to 6 NI-AD subjects [5]. ELISA of the GDFA/GHCl-extracted Ab also revealed an average relative loss of Ab 42 and a concomitant average increase in Ab 40 in the AN-1792 subjects. The average Ab 42 :Ab 40 ratios were 17.8 for the NI-AD and 0.59 for the AN-1792 subjects, thus demonstrating trends analogous to those exhibited in the current bapineuzumab study. As a group, the AN-1792-immunized subjects demonstrated an average GDFA/ GHCl-soluble total Ab loss of 9.3%. Furthermore, the total Trissoluble Ab was 19-fold elevated in the AN-1792 subjects relative to the NI-AD [5].
The ostensible preferential depletion in Ab 42 by bapineuzumab immunotherapy may reflect the antibody avidity for a specific conformational domain of the Ab 42 peptide. Molecular modeling of Ab 42 suggests an ionic bond between the N-terminal amino group of 1Asp and the C-terminal carboxyl group of Ab 42 Ala [24] at the surface of dimeric Ab 42 . This intramolecular conformation may have a higher affinity for the antigen-antibody interaction. Removal of dimeric Ab from the filamentous structures is in all likelihood the way by which these structures are disassembled from AP by antibodies, since the intermolecular bonds within dimers are thermodynamically a billion times more stable than those within free metastable monomers in a misfolded state [25][26][27]. Furthermore, bapineuzumab immunotherapy may be limited in its effects because its specific epitope target is partially or totally absent in a high proportion of deposited Ab peptides due to the partial N-terminal degradation of these molecules [28,29] and to the post-translational modifications such as Asp-isomerization [28] and pyroglutamyl-cyclization [30], which enhance polymerization and dimer resistance to enzymatic degradation [31]. In contrast, the apparently more completely disaggregated 'collapsed' plaque skeletons observed in AN-1792-immunized individuals [9] could be a consequence of the polyclonal nature of this vaccine which may have efficiently eliminated Ab 42 as well as a wide range of other species including N-terminally-degraded Ab peptides. If it is true that a cascade of deleterious events leading to dementia follows amyloid deposition, it suggests protective elimination of both Ab 40 and Ab 42 must be preemptive and thorough.
It is important to bear in mind that although bapineuzumab immunotherapy impacted Ab 42 levels, it does not apparently suppress the primary impetus for amyloidogenic AbPP processing or deposition. The loss of Ab 42 and the noticeable elevation of Ab 40 suggest a possible compensatory mechanism in which one form of Ab is substituted for the other. For example, compensatory deposition of Ab 40 may explain the lack of remarkable differences between AP densities in Bapi-AD and NI-AD individuals despite Ab 42 depletion. The apparent compensatory production and the observed high levels of vascular amyloid in Ab-immunized Figure 8. Western blot analyses of AbPP and its C-terminal peptides CT99, CT83 and of the tau isoforms. Frontal (A) and temporal (B) cortices Western blots of CT20AbPP. The CT20AbPP antibody, raised against the last 20 amino acids of AbPP, was used to detect AbPP and its CT peptides. Frontal (C) and temporal (D) cortices Western blots of tau. The tau isoforms were probed using the HT7 clone made against amino acids 159-163 of the tau molecule. Molecular weights, in kDa, are given on the left of each blot. Actin re-probes are shown below as a total protein loading control. For comparison all Western blots were analyzed against NDC and NI-AD in the same gel. There were no statistical differences between the immunized and non-immunized groups. Statistical analysis and abbreviations are described in the legend to Figure 5. doi:10.1371/journal.pone.0059735.g008 individuals lends support to the hypothesis postulating that one of the functions of Ab peptides is to generate a protective hemostaticlike patch/scab to seal leaky blood vessels at the abluminal side of the brain microvessels (reviewed in [32][33][34][35]). Bapineuzumab immunotherapy induces brain effusive vasogenic edema defined as Amyloid Related Imaging Abnormalities (ARIA-E) and microhemorrhages (ARIA-H) in AD individuals [36]. Flair-MRI scans showed that 36 (17%) out of 210 bapineuzumab-treated AD patients developed vasogenic edema and that 17 (47%) of these patients developed microhemorrhages while only seven (4%) patients out of the remaining 177 without ARIA-E developed microhemorrhages [37]. Transient vasogenic edema has been observed in other studies of Bapi-AD subjects [12,38,39]. A study of 2762 AD patients at baseline in immunotherapy clinical trials found AD-related vasogenic edema to be rare, occurring in only 2 cases [40]. The generation of microhemorrhages in immunized AbPP Tg mice is a well documented observation [41][42][43][44][45][46]. It is possible that blood-brain barrier (BBB) breaches could have been caused by bapineuzumab or autoantibodies generated by the high levels of Ab in addition to a vascular inflammatory reaction. Taken together, these observations imply that partial removal of Ab 42 from the vascular walls by bapineuzumab immunotherapy leads to a cycle of increased production of Ab 40 in an attempt to maintain the integrity of the BBB in the face of immunotherapyinduced alterations.
Interestingly, in the Dominantly Inherited Alzheimer's Network (DIAN) study, 6% of young asymptomatic FAD mutation carriers showed brain microhemorrhages, and 25% of mildly symptomatic carriers also had these lesions of which 16% had 1-4 and 9% had .5 microhemorrhages [47]. In addition, pure PS Tg mice, without AbPP mutations and hence without AP, demonstrated widespread ultrastructural microvascular pathology [48]. Older AbPP/PS1 Tg mice (19-23 months of age), also developed spontaneous ARIA-E and ARIA-H in areas adjacent to vascular amyloid deposition in the absence of immunotherapy [49,50].
As the brain ages, inevitable microvascular decline and accumulating damage decrease BBB integrity [51]. Chronic diseases such as hypertension, hypotension, diabetes, atherosclerosis, arteriosclerosis [52,53] as well as traumatic brain injury [54,55] have a deleterious effect on brain microcirculation [56][57][58][59]. Partial ligation of the thoracic aorta in pigs produces both hypertension and a substantial increase in Ab 40, Ab 42 and p 412 -tau in the brains of these animals [60], suggesting general perfusion failure synergizes the critical pathological cascades associated with AD. The physical attributes of amyloid fibrils such as their chemical stability, insolubility, amphipathic structure, cement-like properties (reviewed in [32]), metal and heme binding capacity [24,[61][62][63][64] and their ability to interact with the extracellular matrix to generate resilient meshes, similar to those of the coagulation cascade, make amyloid an ideal brain microvascular repair/protective substance [32,33,65]. Moreover, Ab binds and sequesters plasma molecules that under normal circumstances would be excluded from the brain [66][67][68][69][70][71]. The Ab peptides also interact with key coagulation molecules including thrombin, fibrinogen and plasminogen [72][73][74][75][76][77][78] and in this context the AbPP 770 and AbPP 751 isoforms containing the Kunitz protease inhibitor domain play an important regulatory role in the coagulation cascade [79,80]. Moreover, in humans, the complex amyloid deposits appear to be protected against proteolytic degradation by a1-antichymotrypsin [81,82]. The continuous and excessive accretion of amyloid around the brain microvessels may terminate by constricting the lumen, creating blind capillary stumps and impaired perfusion [32]. It can be postulated that the brain of a patient with mild and moderate AD has attained or is approaching an adaptive equilibrium in which amyloid production has nearly reached a plateau [83]. Under these circumstances, the removal of Ab 42 may trigger a compensatory overproduction of Ab 40 .
Anti-amyloid immunotherapy, in spite of a reduction of AP in some areas of the brain [5][6][7][8][9]84] and/or dramatic alterations in amyloid composition, has consistently failed to yield commensurate changes in the clinical course of dementia [7,10,11]. Although clinical trials of ponezumab, bapineuzumab and solanezumab for mild and moderate AD were halted or did not achieve their primary end points [11,85,86], other clinical trials using a variety of immune approaches are still being evaluated. The observation that anti-Ab immunotherapies produce substantial increases in soluble Ab peptides provides a possible explanation for the weak impact of these interventions against dementia. Soluble/oligomeric Ab molecules have been isolated from AD brains [87] and these peptides are clearly neurotoxic [88,89]. In addition, these species have been largely incriminated as playing a central role in the pathogenesis and pathophysiological evolution of AD and, to a certain extent, displaced the insoluble plaque and vascular Ab as the primary targets for therapeutic intervention. If these tenets are correct, and in view of the present evidence, it would be expected that in immunized AD patients, an increased pool of soluble oligomeric Ab should aggravate the natural course of the disease. Larger studies enabling more precise evaluation of the role played by soluble/oligomeric Ab in immunotherapy recipients are awaited.
The Alzheimer's Prevention (Preclinical) Initiative (API) and the DIAN studies will involve the proactive administration of anti-Ab immunotherapy treatments to subjects harboring PS mutations years prior to the estimated time of cognitive impairment appearance. Clinical trials have confirmed that anti-amyloid immunotherapy produces a range of considerable positive and negative effects on critical biomarkers of sporadic AD [5][6][7][8][9]12,14,39,84,[90][91][92][93][94][95]. However, whether strategically timed presymptomatic treatments will produce analogous responses and/or preserve cognitive function in PS mutation carriers is uncertain. Biochemical studies of PS mutations have led to the widely held hypothesis that enriched production of Ab 42 underlies dementia [96,97], although rigorous quantitation of Ab peptides in humans with FAD has revealed that this rule is clearly not universal [98,99]. Whether or not crenezumab, solanezumab or gantenerumab treatments favor Ab 40 production and accumulation in FAD-PS mutation carriers, API and DIAN will offer an ideal opportunity to directly test the hypothesis that excess Ab 42 production alone specifically initiates dementia emergence. However, several lines of evidence, including the clinically disappointing, but highly informative, outcomes of immunotherapy trials suggest that dementia progression has an unappreciated biochemical complexity. Our study and previous investigations of the molecular aftermath of immunotherapy [5,9,14] suggest that Ab 40 may not be entirely benign, as elevated quantities of this species are strongly associated with increased vascular amyloidosis and continued decline into dementia. Again, the API and DIAN will provide an ideal prospect to unequivocally examine the potential association of these specific biomarkers with dementia onset and progression as well as a test for the validity of the amyloid cascade hypothesis.
In summary, our study and previous investigations confirm that Ab immunotherapy profoundly impacts the prime AP and Ab42 targets. In some cases, immunotherapy resulted in the physical disruption of AP. In the case of our study, although AP appeared unaltered physically, the amyloid profile in treated subjects was radically perturbed with levels of Ab 42 substantially decreased and Ab 40 levels increased. Despite these responses, immunotherapy did not produce any substantial impact on dementia evolution. The precise mechanism(s) underlying this equilibrium shift is unknown, but the net effect of immunotherapy may be a marked preferential accumulation of the shorter and more soluble Ab 40 species. Coupled with the recognized complexity of AD biochemistry and neuropathology, these observations suggest that amyloid plaque deposits either cannot be the sole instigator of cognitive breakdown or their precise molecular constitution is of little overall consequence to dementia appearance. If this is the case, it suggests that despite remarkable impacts on AD biomarkers, the consistent lack of success of immunotherapy against dementia reflects the general failure to understand the function of amyloid deposition and its dynamics comprehensively.