Amyloid-β Immunotherapy Reduces Amyloid Plaques and Astroglial Reaction in Aged Domestic Dogs

Background: Alzheimer's disease (AD) is characterized by the dynamic accumulation of extracellular amyloid deposits from the interplay between amyloid-β (Aβ) plaques, reactive astrocytes and activated microglia. Several immunotherapies against Aβ have been shown to reduce amyloid neuropathology. However, the role of the associated glia in the recovery process requires clarification. Previously, we described the safety and effectiveness in aged domestic canine with cognitive dysfunction syndrome of a new active vaccine candidate for the treatment of AD in humans. Objective: The aim of this article is to gain a better understanding of how immunotherapy modifies the amyloid burden and its effects on astroglial and microglial reactivity in immunized dogs. Methods: In order to achieve this, we compared and quantified amyloid plaques and astroglial and microglial reactions in the frontal cortex of unimmunized and immunized aged domestic dogs. Results: We found amyloid plaques from immunized dogs to be smaller and more compact than those from unimmunized dogs. In these new plaques, the associated astrocytes were closer and less immunoreactive to the β subunit of S100 protein (S100B). We also found no modification in the microglial reaction associated with immunization. Conclusion: The anti-Aβ immunotherapy developed in our laboratory modifies the equilibrium between soluble and insoluble Aβ in aged dogs in close correlation with S100B-negative astrocytosis and microglial reaction.


Introduction
Alzheimer's disease (AD) is characterized by the accumulation of intracellular neurofibrillary tangles and different forms of extracellular amyloid-β (Aβ; monomers, oligomers, fibrils and plaques) resulting from amyloid precursor protein processing [1,2] . These soluble oligomers, fibrils and amyloid plaques are usually in equilibrium between brain tissue, cerebrospinal fluid and blood compartments. However, the imbalance between Aβ production and clearance in AD leads to the accumulation of these amyloid components [3] . Also, changes in the biochemical composition of Aβ aggregates (oligomers, protofibrils, and fibrils) are specifically associated with the occurrence of symptomatic AD [4] . Thus, the imbalance between soluble Aβ production and clearance modifies the amyloid plaque features [5] and the glial response [6] . The resulting abundance of amyloid deposits is an interplay between plaques of different biochemical composition, reactive astrocytes and activated microglia [6][7][8] .
One of the animals considered to be a good model for research into mild cognitive impairment and AD is the aged domestic dog (over 8 years old) with cognitive dysfunction syndrome (CDS) [9][10][11][12] . Since this animal spontaneously develops geriatric behavioral changes related to several markers of sporadic AD (soluble Aβ, Aβ diffuse plaques and brain atrophy), early AD physiopathology mechanisms and new therapeutic targets for innovative therapies can be researched through studies performed using CDS dogs [12][13][14] . Beginning around age 8, the formation and maturation of diffuse Aβ deposits are observed by immunostaining in all layers of canine cortical gray matter in a characteristic four-stage distribution that correlates with CDS severity and astroglial reaction [15][16][17][18][19] . Despite the human studies that suggest an association between amyloid plaque stages and the degree of microglial activation [20,21] , there is still controversy surrounding the activation of microglia by diffuse plaques. The contradictory results are probably due to differences in the antibodies and lectins used for detection [6,22] .
Astrocyte overexpression of the β subunit of S100 protein (S100B) plays a role in the pathogenesis and progression of AD and other neurodegenerative diseases through autocrine and paracrine effects on astrocytes, neurons and microglia [23] . In AD, most of these S100B-positive astrocytes are closely associated with both diffuse and neuritic plaques [24] , and their distribution throughout the brain regions mirrors the known distribution patterns for Aβ deposits [25] . In CDS dogs, the astrocyte proliferation around diffuse plaques was found to be S100Bnegative, which together with the absence of neuritic plaques in the dog brain would suggest an important pathogenic role for S100B in the genesis and evolution of AD plaques [20,24] .
Many anti-Aβ immunotherapies targeting soluble and/or insoluble Aβ [26,27] have been tested as ways of preventing or treating AD in animals and humans [28][29][30][31][32] . Due to the initially severe adverse effects observed in humans [33,34] , further studies focused on the design of effective and safe immunotherapy [35,36] . One of these immunotherapies, based on a mixture of fibrillar Aβ components, has been developed in our laboratory with great effectiveness and complete safety in domestic dogs with CDS [37] . In that study, the vaccine led to a very rapid cognitive improvement in all treated animals, with no side effects.
To gain a better understanding of how immunotherapy modifies the amyloid burden and its effects on astroglial and microglial reactivity, we quantified amyloid plaques and astroglial and microglial reactions in brain samples from unimmunized and immunized male and female domestic dogs of different ages and breeds and with different cognitive deficits. To avoid the controversy surrounding canine microglia quantification with antibodies and lectins, we estimated the microglial reaction by in vitro autoradiography of [ 3 H]-PK11195 binding to the peripheral benzodiazepine receptor (PBR) [38][39][40][41] .

Vaccine
The fibrillar Aβ vaccine was initially prepared with 100 μg of Aβ 40 combined with the Aβ x-40 peptide KLH conjugated. Aβ 40 and Aβ x-40 peptides were initially dissolved with dimethyl sulfoxide (1 mg/ml), aliquoted and stored at -20 ° C. Once thawed, each Aβ peptide was mixed with 1 mg of alum hydroxide suspension (Sigma-Aldrich, Mo., USA). After overnight incubation at 4 ° C, a saline solution was added up to a volume of 1.5 ml prior to the subcutaneous (s.c.) immunization of the dogs.

Animals
Twenty-three male and female domestic dogs of different ages and breeds and with different cognitive deficits were included in the study ( table 1 ). All dogs were previously classified as nondemented dogs (n = 12) or CDS dogs (n = 11), depending on the score obtained using a validated cognitive test of nine items adapted from the Mini-Mental State Examination (MMSE) and the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) for AD [10,12] . Five of these 23 dogs had been included in a previous immunization study performed by our group [37] in which they received six s.c. injections of the fibrillar Aβ vaccine over a period of 51 days and were sacrificed at the end of the immunization study after blood extraction ( fig. 1 ; table 1 ). One of the immunized dogs presented CDS symptoms at the beginning of the immunization period.
All animals were treated according to European legislation on animal handling and experimentation (86/609/EU). Procedures were approved by the University of Barcelona Ethics Committee, Barcelona, Spain. All efforts were made to minimize animal suffering and to use no more than the number of animals required to obtain reliable scientific data.

Brain Tissue Preparation
Animals were euthanatized with an i.v. overdose of sodium thiopental (75 mg/kg, Thiobarbital, Braun Medical S.A., Spain) and tissue preparation was adapted from human brain bank methods [42] . In all cases, donation was formally approved by the owner and euthanasia justified for medical reasons, none of which were related to immunotherapy. The dog brains were removed immediately following death. Both hemispheres were separated and cut into 1-cm-thick coronal sections. One hemisphere was frozen in powdered dry ice and stored at -80 ° C. The other hemisphere was fixed by immersion in cold 4% formaldehyde diluted in 10 m M phosphate buffer (pH 7.4) for 4 days. After 3 days of cryoprotection, they were frozen on powdered dry ice and kept at -40 ° C until use. Some liver and spleen samples were also collected as tissue control for histological procedures.
Anti-Aβ 40 IgG Purification pH gradient elution by immunoaffinity column was used for anti-Aβ 40 purification. For this, Sepharose powder (Pharmacia, Barcelona, Spain) activated with cyanogen bromide and coupled to the Aβ 40 peptide (40 amino acid length) was placed into a chromatography column (Bio-Rad, Calif., USA). After being washed with 10 ml of 100 m M cold phosphate-buffered saline (PBS, pH 7.4), 1 ml dog serum was loaded into the column 3 times. After being washed with 10 ml cold PBS, anti-Aβ 40 Igs were eluted by pH gradient using firstly 100 m M glycine, pH 2.5 and secondly 100 m M glycine +4 M urea, pH 2.5; 200 μl of antibody fractions was collected in Eppendorf tubes containing 1 M Tris, pH 8. Anti-Aβ 40 IgGs were detected in each Eppendorf tube using a dot blot method. Briefly, antibody samples were loaded on a PVDF membrane (Bio-Rad, Madrid, Spain) and incubated for 1 h with biotinylated anti-dog IgG (LifeSpan Biosciences, Derio, Spain). Detection was carried out by incubation in 20 m M Tris-HCl and 137 m M NaCl (TBS, pH 7.4) containing 0.03% 3,3 ′ -diaminobenzidine (DAB) and 0.006% H 2 O 2 . Finally, purified antibodies were dialyzed, concentrated to half of the volume in 50 m M Tris-HCl, pH 7.4 and stored at -20 ° C until use.

Immunohistochemistry
To detect amyloid plaques and immunoreactive (IR) astrocytes on dog sections, immunohistochemistry with specific antibodies against Aβ peptide (Dako, Glostrup, Denmark), S100B (Dako) and glial fibrillary acidic protein (GFAP) (Dako) was carried out on 12-μm-thick serial sections obtained from the formaldehyde-fixed  frontal cortex. To detect amyloid plaques on human sections, purified dog anti-Aβ 40 antibodies from immunized CDS dogs or commercial mouse anti-Aβ antibodies (Dako) were used for the immunohistochemistry. For comparisons, immunohistochemistry was carried out in 5-μm-thick sections from frontal cortex samples of AD human brain. Paraffin-fixed brain tissues from AD patients at stages II and VI according to the Braak and Braak scale [43] were obtained from the Neurological Tissue Bank of the Universitat de Barcelona (Barcelona, Spain). Some sections were stained with Perl's Prussian blue counterstained with neutral red to identify hemosiderin deposits resulting from putative microhemorrhages [44] . For immunohistochemistry, slices were deparaffinized or defrosted and pretreated in formic acid for 2 min. After endogenous peroxidase inhibition for 30 min, the slices were incubated with PBS 100 m M , pH 7.4, containing 0.5% Triton X-100, 5% normal goat serum and 5% bovine serum albumin as a blocking agent for 2 h at room temperature. Purified primary antibodies or commercial mouse anti-Aβ antibodies were diluted (1/100) with immunobuffer (PBS, 5% Triton X-100, 1% normal goat serum, 1% bovine serum albumin) and incubated overnight at 4 ° C. After being washed with PBS, the samples were incubated with 1/4,000 diluted biotinylated anti-dog IgG or 1/100 diluted biotinylated anti-mouse IgG for 2 h. After washing, the slices were incubated for 1 h with ExtrAvidin (1/250) and developed with 0.03% DAB-0.006% H 2 O 2 in PBS. Some sections were also incubated with anti-CD3 (1: 300; AbD Serotec, Oxford, UK) followed by biotinylated anti-mouse IgG, for immunodetection of infiltrated lymphocytes.
Double detection of amyloid plaques/astrocytes was performed by sequential immunohistochemistry. Amyloid plaques were first stained in black with anti-Aβ antibody (diluted 1/100) as described above, and developed by adding NiCl to the developing solution to stain astrocytes brown on the same slices. Sections were incubated with 1/250 diluted rabbit anti-GFAP antibody or with diluted 1/500 rabbit anti-S100B antibody detected with biotinylated anti-rabbit IgG (1/500) and developed with 0.03% DAB-0.006% H 2 O 2 in PBS. To reduce the variability due to immunolabeling procedures, for each specific antibody all slices were processed at the same time, using the same antibody solutions, and incubation/ reaction and developing times.
Plaques and astrocytes were visualized by optical microscopy (Olympus America Inc., N.Y., USA) and micrographs were taken at 5× and 40× following the same image thresholding and acquisition protocol. The number and reactivity of plaques and GFAP-S100B IR astrocytes were quantified in dog samples, as previously described [20] . For this, five semirandomized micrographs were taken and analyzed (Axio-Vision, Carl Zeiss, Germany) in three conditions: (1) to measure the number, area and gray color intensity of plaques, we obtained photomicrographs, at a magnification of 5×, of the whole frontal cortex area; (2) to measure the number and gray color intensity of IR astrocytes, we obtained photomicrographs, at a magnification of 40×, from the frontal cortex, and (3) to estimate the number and gray color intensity of IR astrocytes found within 200 μm of the most external part of the plaque, we obtained photomicrographs, at a magnification of 10×, from the same five frontal cortex plaques. To improve data reliability, GFAP and S100B immunohistochemistry were performed on serial slices and the same semirandomized areas in white matter, and the same semirandomized plaques were used for all parameters analyzed.

Thioflavin-T-Positive Detection of Amyloid Plaques by Purified Anti-Aβ 40 IgGs
Thioflavin-T-positive (ThT+) detection of diffuse amyloid plaques by purified anti-Aβ 40 IgGs was carried out on 5-μm-thick slices from the frontal cortex of AD patients at stage II according to the Braak and Braak scale [45] . First, Aβ plaques were detected following the immunohistochemistry procedure and using purified dog anti-Aβ 40 from immunized CDS dogs or commercial mouse anti-Aβ antibodies, as described above. The slices were then dried in the dark at 4 ° C overnight, rehydrated and immersed in a solution of 0.05% ThT-HCl 0.1 M . Afterwards, they were incubated with 1% acetic acid for 20 min, washed and mounted with Pro-Long antifade reagent (Sigma-Aldrich). Plaques were visualized by fluorescence microscopy (Carl Zeiss, Germany) and photomicrographs were taken at a magnification of 20×.

In vitro Autoradiography Binding of the [ 3 H]-PK11195 to the PBR
Autoradiographic labeling of PBR, a specific marker of microglia, was performed with [ 3 H]-PK11195 (NEN, Boston, Mass., USA) on 5-μm-thick serial sections of frozen entorhinal cortex, as described previously [38][39][40] . Sections were incubated in an ice-cold 50 m M Tris-HCl solution (pH 7.7) containing 1 n M [ 3 H]-PK11195 for 2 h at room temperature, followed by two rinses in 4 ° C Tris-HCl buffer for 5 min. Nonspecific binding was determined in the presence of 1 m M unlabeled PK11195. After rinsing in water, sections were dried under a stream of cold air and apposed to Amersham Hyperfilm (Amersham, Bucks, UK) for 2 weeks. Films were developed and analyzed densitometrically after calibration with gray scales ( 3 H-Microscales; Amersham) using the Image-Pro Plus program (Md., USA). The average brain protein content was 8%. For each brain, four sections were processed for total binding and two others for nonspecific binding. The values obtained for each brain after nonspecific binding subtraction were averaged and used for statistical comparisons.

Statistical Analysis
Kurtosis and skewness were calculated to verify the normal distribution of data. One-way ANOVA was performed to compare the plaque, astroglia and microglia parameters between the treatment and plaque stage, followed by Fisher's LSD and Bonferroni post hoc tests. To quantify the association between two variables, a Pearson correlation was performed. When normality was not achieved, the values of all groups were compared using nonparametric analysis. In these cases, the Kruskal-Wallis test followed by the Mann-Whitney test were used to compare dependent variable groups, and the Spearman correlation was used to quantify the association between two variables. In all cases, p < 0.05 was considered as significant. Results are expressed as a mean ± SEM. All analyses were performed with the programs SPSS 17 (IBM Spain, Madrid, Spain) and Statgraphics (STSC Inc., Rockville, Md., USA).

Increased Cortical Microglial Reaction in CDS Dogs
Initially, astroglial and microglial reactivity was compared between unimmunized control (n = 8) and CDS (n = 10) dogs. To this end, four parameters (number of S100B; number of GFAP-IR astrocytes; gray color intensity of S100B-positive astrocyte cytoplasm, and binding of [ 3 H]-PK11195) were measured in the cortex of both groups. In CDS dogs, the number of S100B-IR and GFAP-IR both showed a nonsignificant tendency to increase ( fig. 2 A, C), but we found no differences in the intensity of S100B labeling ( fig. 2 B).
To study the microglial reaction, [ 3 H]-PK11195 autoradiography was performed. [ 3 H]-PK11195-specific binding showed low levels (between 551 and 1,068 fmol/ mg protein) in the cortex of control dogs, with the lowest values in the internal layers of this area ( fig. 4H). Non-specific binding was homogeneous and very low, representing less than 10% of the total binding. In the cortex of CDS dogs, [ 3 H]-PK11195-specific binding was 29.40% higher when compared with controls (p = 0.034). This increase in specific labeling was homogeneously distributed in the tissue, and no cluster-associated labeling was observed in any layer of the prefrontal cortex ( fig. 5H).
When data from immunized dogs were included in the study, the intensity of S100B-IR showed a nonsignificant tendency to decrease in all animals. However, no differences were found in the number of S100B-IR or GFAP-IR cells or in the [ 3 H]-PK11195-specific binding ( fig. 2 A-D). binding to the PBR in CDS than in control dogs, indicating more activated microglia in these dogs ( * p = 0.034). When immunized dogs were included in the comparison, only a nonsignificant tendency for the intensity of cytoplasmic S100B-IR to decrease was observed.
Aβ Immunotherapy in Aged Domestic Dogs

Anti-Aβ 40 IgGs from Immunized Dogs Recognize Diffuse Plaques
Previously, we showed that this vaccine in dogs generates anti-Aβ 40 IgGs that cross the blood-brain barrier and bind to human diffuse plaques [37] . To determine whether or not they bind similarly to human neuritic, diffuse and fibrillar amyloid plaques in AD stages II and VI, serum from dogs was taken on day 51 of the immunization period. Purified anti-Aβ 40 IgGs presented stronger immunoreactivity to dif-fuse plaques than neuritic ones. These antibodies were not IR to ThT+ plaques or blood vessels ( fig. 3 ). A similar study performed with commercial anti-Aβ 40 IgGs showed an increased immunoreactivity to neuritic plaques ( fig. 3 ).

Immunization Reduces the Size of Amyloid Plaques and Increases Their Compaction
Prussian blue histochemistry revealed no hemosiderin deposits ( fig. 4 A-C) in the prefrontal cortex of any of the Fig. 3. Detection of human amyloid plaques with purified anti-Aβ IgGs from immunized dogs. Immunohistochemistry using commercial anti-Aβ IgGs or purified anti-Aβ IgGs as a primary antibody were performed in samples from AD patients in stage II and stage VI of the Braak and Braak scale. A , C Commercial antibodies can detect neuritic plaques (black arrows) and leptomeningeal blood vessels (arrowheads) in AD stage VI samples more effectively. B , D Purified antibodies, however, can detect diffuse plaques (black arrows) more effectively, but not leptomeningeal blood vessels (arrowheads). E , F Double immunohistochemistry to detect anti-Aβ IgGs (red) and ThT+ (blue) showing that fibrillar amyloid plaques were not detected by purified anti-Aβ IgGs, while anti-Aβ labeling with commercial antibodies shows that they are colocalized with ThT+ plaques (pink; colors refer to the online version only). Scale bars = 200 μm ( A-D ), 50 μm (E, F, inset in C ). To determine whether or not the immunization process modifies them, we characterized amyloid plaques by immunohistochemistry in 3 immunized dogs and 9 control aged animals (older than 8, since young animals' brains do not form any plaques). We identified two groups of nonimmunized dogs, according to the size, density and cortical-layer distribution of amyloid plaques, as previously described [12] ; 5 dogs presented diffuse amyloid plaques (stage I-II) that had a 'cloud-like' aspect, covered a large area, were poorly compacted, and were located in layers V-VI of the prefrontal cortex ( fig. 5 A-C). Four dogs presented more compact amyloid plaques (stage III-IV) that were smaller in size, more aggregated and located in layers II-III or throughout all layers of the prefrontal cortex ( fig. 5 A-C). The 3 immunized dogs presented different specific plaques, which were the smallest in size and had a more dense aspect than the stage I-II and stage III-IV plaques (p = 0.036 and p = 0.033, respectively; fig. 5 A-C). The plaque distribution in these animals was similar to that of the stage III-IV plaques, since they were located in layers II-III or throughout all layers of the frontal cortex.

Color version available online
Immunization Decreases Immunoreactivity to S100B A comparison of the astroglial reaction of dogs classified by amyloid plaque morphology showed no difference in the number of S100B-IR and GFAP-IR astrocytes ( fig. 5 D, F). However, immunized dogs presented a lower intensity of S100B-positive astrocyte cytoplasm than the other groups (p = 0.016; fig. 5 E). Moreover, both GFAP-IR and S100B-IR astrocytes in the prefrontal cortex of immunized dogs presented a lower number of cell processes than astrocytes in the prefrontal cortex of nonimmunized animals' amyloid plaques ( fig. 6 A, B).
We also quantified microglial activation by [ 3 H]-PK11195 autoradiography. In all animals presenting amyloid plaques, the levels of [ 3 H]-PK11195-specific binding sites in the cortex showed a marked increase when compared to young controls (mean 28.95% increase; p = 0.016; fig. 5 G, H). We found no immunization effects in that [ 3 H]-PK11195-specific binding increase.
We then considered amyloid plaques as a unique entity, divided them into three groups (stage I-II, stage III-IV and immunized dogs) and analyzed the astrocytes included within 200 μm of the plaque perimeter. In the stage III-IV plaque group, the number of S100B-IR (p = 0.012) and GFAP-IR cells (p = 0.031) and the intensity of S100B-IR (p < 0.001) were lower than those of the stage I-II plaques ( fig. 6 C-E). The plaques from immunized dogs presented a similar number of GFAP-IR cells to stage III-IV plaques ( fig. 6 C). However, they presented a lower number of associated S100B-positive astrocytes (p = 0.050), and the S100B-IR intensity in these cells was the lowest found (p = 0.001; fig. 6 A, B).
To gain a better understanding of the relationship between astrocytes and plaques, we subdivided the 200-μm distance into four 50-μm segments, and calculated the percentage of either S100B-IR or GFAP-IR present in each Fig. 4. Histological study of brain microhemorrhages or lymphocyte infiltration induced by immunization. A-C Perl's Prussian blue counterstained with neutral red identified hemosiderin deposits in dog liver slices (arrows in A ), but not in the perivascular prefrontal cortex of immunized control ( B ) and CDS ( C ) dogs. D-F Immunohistochemistry with specific anti-CD3 antibodies stained in dark brown lymphocytes in dog spleen slices ( D ), but not in the prefrontal cortex of immunized control ( E ) and CDS ( F ) dogs. Asterisks identify blood vessels. Scale bar = 50 μm.   . 6 F-H). In the stage I-II plaque group, the percentage of S100B-IR cells was similar in all four analyzed segments ( fig. 6 F), while the furthest-away segment (150-200 μm) presented the highest percentage of GFAP-IR cells (p χ 2 = 0.019). In the stage III-IV plaque group, the percentage of S100B-IR cells presented a nonsignificant tendency to increase in the 25-to 50-μm and 50-to 100μm segments when compared to the stage I-II plaques, and the same was true of the percentage of GFAP-IR cells ( fig. 6 G). In the plaques of immunized dogs this tendency to increase reached statistical significance in the percentage of GFAP-IR cells included in the 25-to 50-μm and 50-to 100-μm segments (p χ 2 = 0.002; fig. 6 H).

Discussion
This article presents evidence that the anti-Aβ immunotherapy developed in our laboratory [37] modifies the shape and density of Aβ plaques in aged dogs, in close correlation with S100B-negative astrocytosis and microglial reaction. We also provide data that reinforce the idea that glial cell reactivity is associated with amyloid pathology and cognitive dysfunction [17,19,20] . In a previous article, we showed that this vaccine led to a very rapid cognitive improvement in all treated CDS dogs, with no evident side effects [37] . We also found that plasma Aβ 40 was increased but remained constant   6. Immunoreactive GFAP and S100B astrocytes associated with the amyloid plaque. Illustrative photomicrographs of either GFAP-IR or S100B-IR cells in brain sections of dogs presenting stage I-II plaques ( A ) and immunized dogs ( B ). Quantification of the GFAP-IR and S100B-IR cells within 200 μm of the plaque. Histograms show the number of S100B-IR cells ( C , & p = 0.012; # p = 0.050), the intensity of S100B cytoplasmic staining ( D , * * p < 0.001, # p = 0.001) and the number of GFAP-IR cell astrocytes ( E , * p = 0.031). To estimate the differences in the astrocyte-plaque associa-tion, we then divided the distance of 200 μm into four 50-μm segments and calculated the percentage of both GFAP-IR and S100B-IR cells (arrows in pictures) located in each segment ( F-H ). GFAP-IR cells associated with plaques from immunized dogs ( H ) showed a nonsignificant tendency to be located in the closest segments when compared to stage I-II ( F ) and stage III-IV ( G ) plaques. This tendency was not observed for S100B-IR cells (p χ 2  in the cerebrospinal fluid of immunized dogs. These modifications were accompanied by a 10-fold increased serum anti-Aβ 40 IgG [37] . These results established a relationship between the grade of CDS and the levels of Aβ peptide in cerebrospinal fluid and plasma. We now show that immunization modifies the shape and density of Aβ plaques, which argues for a modification of the equilibrium between soluble and insoluble Aβ in aged dogs. Due to the lack of reliable antibodies available to detect activated microglia in the dog brain by immunohistochemistry [46] (and pers. observations), the first prominent result of this study was the description of an effective method for measuring activated microglia in the dog brain. Quantification of PBR concentration by [ 3 H]-PK11195-specific binding to brain slices has been used extensively to estimate microglial reactivity in human and rodent brains [38,[47][48][49] . Similarly, quantification of in vivo radiolabeled PK11195 binding to the cerebral PBR is currently used to measure microglial reactivity in patients with various pathologies involving neuroinflammation, including AD patients, through modern imaging techniques [50][51][52] . When compared with controls, we herein found an increase in [ 3 H]-PK11195 binding in the cerebral cortex of CDS dogs, which indicates a relationship between microglial activation and canine cognitive deterioration. This relationship, also found in human AD [53] , is another of the several AD hallmarks present in the CDS dog that highlight dogs as a good model for studying the cascade of events that take place with brain amyloid deposition, aging and dementia [12] .
We previously established that S100B plays a role in the link between diffuse Aβ plaque maturation and astroglial reactivity in the aging dog brain [20] . This time, we found no clear association between cognitive decline and an increase in the total number of S100B-and GFAP-positive astrocytes. This aspect may represent acute events of the animal life that bear no relationship to the aging and cognitive process [18] . By including dogs with a diverse range of characteristics in the study we avoided estimating the possible influence of specific aspects like breed or sex on this result. However, our data confirm a relationship between Aβ deposition and canine cognitive deficit [15,18] , which may be related to alterations in brain function through common processes at the early stages of amyloid deposition, as has been suggested for AD [54,55] .
We have previously designed and validated in aged dogs the efficacy and safety of a new vaccine considered a good candidate for human AD prevention and treatment [37] . When we assessed the brains of some of those immunized animals, we were able to analyze the effects of the immunization in both the inflammation and the Aβ plaque shape and distribution in the brains. Our results argue for a lack of microhemorrhages or inflammation associated with this immunotherapy. We describe the absence of hemosiderin-positive deposits and of CD3-immunopositive cells in the perivascular parenchyma. We also found a lack of anti-Aβ antibody binding to blood vessels and a homogeneous widespread binding of [ 3 H]-PK11195 to the prefrontal cortex. These results are in line with the lack of side effects found in the clinical follow-up of these same immunized animals, and the normality of the plasma and cerebrospinal fluid analyses previously published [37] .
Aged dogs spontaneously develop diffuse plaques that do not fulfill the β-pleated sheet (equivalent to AD maturation stages I-IV) and are thioflavin-negative, with no apparent neuritic component [12] . In all of the immunized dogs, we identified a new group of smaller and more compact amyloid plaques that probably resulted from immunization. Aβ immunization induces plaque clearance in both 3×Tg-AD mice and humans (see [54] for a review). In dogs, the lack of booster doses in the immunization procedure may induce a slow course of Aβ clearance that enables the plaques to be completely removed in a few months but prevents side effects. These Aβ aggregates may also present differences in the biochemical composition with respect to those of nonimmunized animals. In human AD, the grade of Aβ phosphorylation is specifically associated with the occurrence of symptomatic AD [4] . These Aβ phosphorylation variations, also found in dispersible Aβ oligomers, protofibrils, and fibrils, may be modified by immunization and account, at least in part, for the cognitive improvement of immunized dogs.
In our previous study, we detected anti-Aβ 40 IgGs in the cerebrospinal fluid of immunized dogs [37] , which demonstrates that these antibodies cross the blood-brain barrier. We herein found that these anti-Aβ 40 IgGs recognize diffuse nonfibrillar plaques much more easily than neuritic ones or blood vessels. Thus, antibodies may recognize diffuse plaques and activate a neuroinflammatory response that would lead to smaller and more compact amyloid plaques with consequences for the soluble-insoluble Aβ equilibrium between the brain and blood [37] . If this is true, immunized dogs should present changes in neuroinflammatory activity when compared with control animals.
In the aged dogs, we found hypertrophic astrocytes overexpressing GFAP and S100B generally located just outside the plaque boundary. This distribution is associated with plaque stage and proximity and is related to cognitive status [20] , as has also been described for AD [7,56,57] . In the brains of immunized aged dogs, we found a lower number of reactive astrocytes related to Aβ plaques. These astrocytes were less immunoreactive to S100B and GFAP and were grouped more closely to the plaques. Activated S100B-IR astrocytes are related to the induction and maintenance of dystrophic neurites in amyloid plaques [24,58] , due to amyloid toxicity. For that reason, S100B has been considered an important pathogenic factor in the genesis and evolution of plaques in AD. The reduced S100B-IR of astrocytes may therefore lead to the recovery of dystrophic neurites and the subsequent reduction in amyloid plaque processes. Thus, a decrease in S100B overexpression appears in the plaques of immunized dogs; these plaques correspond to the more compact stage of plaque maturation and present a shorter astrocyte-plaque distance. We previously described a correlation between canine graded cognitive deficit, diffuse plaque maturation, and S100B astrocytosis [20] . Thus, astroglial reactivity around the plaque varies according to plaque maturation, and S100B-positive astrocytes are involved in the more advanced stages of maturation and cognitive deficit. Based on that, the reduced intensity of S100B-IR, but not GFAP-IR, in astrocytes close to plaques of immunized dogs that we found in the present study may be related to the improved cognitive deficit of these immunized CDS animals [37] . Also, the correlation between S100B overexpression and the more significant impairments argues for astrocyte participation in autocrine and paracrine effects associated with the neurodegenerative process [20,23] . To approach this putative relationship, further experiments are needed with a higher number of immunized CDS dogs.
At physiological levels, S100B exerts a protective effect, but when it increases, its interactions with other cytokines and microglial cells potentiate the activation of microglial, neurodegenerative or apoptosis-inducing effects. Therefore, the absence of a microglial reaction specifically associated with diffuse plaques in our study may be related to lower plaque-astrocyte interactions and reactivity. Together with the absence of microglial cells in diffuse AD plaques [58,59] , our data suggest that the involvement of S100B and microglia in the progression of AD does not begin in the very early stages of plaque formation. However, we found an increase in the widespread homogeneous binding of [ 3 H]-PK11195 in the cerebral cortex of CDS dogs, which would suggest that microglia are involved in the initiation and maintenance of amyloid pathology [50] . Microglial activity in AD is a controversial subject. While numerous studies in animal models of AD have demonstrated that modulating microglial activation can be a powerful strategy for clearing Aβ plaques, it is also thought that Aβ might not adequately activate microglia effector functions to clear the accumulating amyloid burden from the brain [53] . Our results showed that Aβ immunization does not modify the distribution and increase of [ 3 H]-PK11195 in CDS dogs. However, other changes in microglial activity may occur. As explained for AD [60] , Aβ immunization specifically increases the phagocytic activity of microglia, which would contribute to fibrillar soluble Aβ clearance and lead to plaque compaction and decreased amyloid toxicity. Taken together, all these neuroinflammatory changes may contribute, at least in part, to the cognitive improvement of immunized CDS dogs observed after several weeks of treatment.
In conclusion, in this study we demonstrated that the anti-Aβ immunotherapy developed in our laboratory modifies the shape and density of plaques in aged dogs in close correlation with S100B-negative astrocytosis and microglial reaction. Together with our previous study in which we described a cognitive improvement and a shift in equilibrium between soluble and insoluble Aβ in immunized dogs, our overall results indicate that immunotherapy decreases Aβ toxicity and they reaffirm the interest in anti-Aβ immunotherapy as a possible treatment for canine CDS and human AD. Future studies would be necessary for a better characterization of the immunization effects on Aβ biochemical composition and the interplay between the S100B-positive astrocytes, microglia and Aβ burden.