Vaccination with Aβ-Displaying Virus-Like Particles Reduces Soluble and Insoluble Cerebral Aβ and Lowers Plaque Burden in APP Transgenic Mice¹

Alzheimer’s disease (AD)³ is the most common form of senile dementia and histologically characterized by neurofibrillary tangles and extracellular amyloid plaques, predominantly composed of the β-amyloid peptide (Aβ). Several currently explored therapeutic strategies aim at reducing Aβ plaque deposition and/or reducing the soluble and oligomeric forms of Aβ that have been implicated in AD pathogenesis (1, 2). Immunotherapy with active vaccination against Aβ has been shown to reduce cerebral plaque load and improved deficits in learning and memory in several independent transgenic animal models of AD (3, 4, 5). Humoral immunity directed against Aβ was shown to play a key role, as passive immunization with monoclonal, Aβ-directed Abs had similar effects (6, 7, 8, 9). Initial clinical trials, however, which used full-length aggregated Aβ1–42 as Ag (AN-1792) and QS-21 adjuvant (10, 11, 12), had to be stopped in phase II due to aseptic meningoencephalitis in 6% of the treated patients; this was attributed to T cell-mediated adverse reactions (13, 14). As a consequence, more recent studies now aim for the induction of strong Ab responses without triggering inflammatory and cytotoxic T cell responses. In the absence of Th cell activation, however, Ab responses are dominated by IgM Igs and thus are short-lived of low binding affinity and lack B cell memory. As Aβ is an ubiquitously expressed self-Ag to which the immune system is normally tolerant (to suppress the induction of autoantibodies), vaccination needs to overcome B cell tolerance to induce high Ab titers (15). One strategy used to break self-tolerance is to repeatedly administer high doses of the Ag together with strong adjuvants. However, vaccination with AN-1792, even in the presence of a potent Th1 adjuvant such as QS-21, induced only a low/moderate immune response in a small subset of AD patients (19.7%), whereas ∼80% of immunized subjects were “nonresponders” (10, 14). Moreover, in this case, Th cell activation was triggered by self-epitopes within Aβ. This strongly suggests that alternative strategies need to be pursued toward a safe and efficient active immunotherapy for AD.

We thus aimed at developing a vaccine that is very immunogenic in the absence of any adjuvant and in which Th cell determinants are provided not by Aβ itself but in the context of a carrier system. Previously, we have shown that noninfectious recombinant retrovirus-like particles (retroparticles)—that do not fall under contained use regulations—can be used to display Ags at high density and that vaccination with nonreplicating retroparticles displaying either self-Ags (e.g., cellular prion protein (16)) or non-self-Ags (e.g., the glycoprotein (G) of vesicular stomatitis virus (VSV-G) (17)) induce high IgG Ab titers in the absence of adjuvant.

In this study, we adopted this strategy as an active immunization against AD and displayed the 15 N-terminal amino acids of human Aβ on the retroparticle surface (Aβ retroparticles). Adjuvant-free single immunization with Aβ retroparticles induced high and long-lived Aβ-specific IgG titers of Th2 isotypes (IgG1 and IgG2b) that bound to aggregated Aβ in plaques of mouse brain sections. Immunization of transgenic APP23 model mice not only led to prominent reductions of Aβ plaque load but also grossly lowered cerebral insoluble as well as soluble Aβ40/42 pools. In summary, Aβ retroparticle vaccination may thus hold promise as a future safe and efficient immunotherapy for AD.

Aβ retroparticles were produced by cotransfection of HEK-293T cells with the plasmid pHIT60 encoding the murine leukemia virus (MLV) gag and pol genes and the plasmid pDisplay-Aβ. Transfection was performed with Lipofectamine 2000 according to the manufacturer’s protocol (Invitrogen). Forty-eight hours after transfection, the cell supernatant containing the Aβ retroparticles were harvested by centrifugation at 4°C for at least 16 h for concentration. The pellet was resuspended in PBS Dulbecco/1% FCS. For titration of the particle stocks, ELISA plates were coated with particles, and the amount of Aβ displayed by the particles was quantified using an Aβ-specific mAb (clone 6E10; Signetlabs). The number of particles was quantified by reverse transcriptase assay as described previously (18).

The APP23 transgenic mouse line used in this study expresses human APP751 with the Swedish double mutation (K670N, M671L) under control of the neuron-specific Thy-1 promoter (19) and was bred at the University of Frankfurt under SPF conditions. Wild-type (WT) (C57BL/6) mice were purchased from Charles River Laboratories or were bred at the Paul- Ehrlich-Institute. Mouse experimental work was conducted using 6–12-wk-old mice in compliance with regulations of German animal welfare.

WT and female APP23 mice were immunized i.v. with ∼10¹⁰ Aβ retroparticles (equals ∼150 ng of Aβ) in 200 μl of PBS. Booster injections were performed monthly over a period of 6 mo starting at 4 mo of age. To monitor the Ab reactivity, blood was taken at different time points after immunization. APP23 mice were sacrificed at an age of 49 wk, brain hemispheres were prepared, and one-half each was used for histological and for biochemical analysis, respectively.

To assess T cell responses, WT mice were immunized with Aβ1–15 retroparticles or with synthetic peptide Aβ1–42 (Bachem). Briefly, lyophilized Aβ1–42 was resuspended in DMSO at a concentration of 1 mg/ml. Before immunization, one-half of Aβ1–42 was incubated overnight at 37°C to form amyloid aggregates and mixed 1:1 with nonaggregated peptide. Next, Aβ1–42 was mixed 1:1 with Freund’s complete adjuvant (Sigma-Aldrich), and 100 μg of Aβ1–42 was injected s.c. into each mouse. For the following vaccinations, the Ag suspension was mixed 1:1 with Freund’s incomplete adjuvant (Sigma-Aldrich). After immunization, five boosts were performed at 3-wk intervals. All groups of mice were bled 7–9 days after each injection, and sera were prepared for ELISA as described below. In addition, 9 days after the last boost, animals were sacrificed, and spleen cells were tested for IFN-γ expression by FACS analysis.

For immunonegative staining, 10 μl of virus-like particle (VLP) suspension was adsorbed to glow discharged carbon-coated Formvar grids for 2 min. After rinsing in PBS, grids were incubated with a monoclonal anti-Aβ1–16 Ab (clone 6E10; Signetlabs/Covance Research Laboratories) in a 1/5000 dilution for 15 min. After washing, grids were incubated with a 10 nm in diameter gold particle-labeled goat anti-mouse IgG Ab (Biocell Laboratories) at 1/50 dilution for 15 min. Finally, immunolabeled samples were negatively stained with 2% uranylacetate for 10 s. Electron microscopy preparations were examined by a Zeiss EM 109 or 902 electron microscope.

For Western blot analysis, 5 μl of a 1/10 dilution of retroparticle suspension was boiled for 5 min in a standard SDS sample buffer, separated on a 16% Tris-Tricine PAGE, and blotted on nitrocellulose membranes (Whatman). For detection with the anti-Aβ Ab (6E10, Signetlabs/ Research Laboratories; 1/2000 in PBS/0.05% Tween 20), the membrane was boiled for 10 min in PBS to re-expose the Aβ epitope. Anti-HA Ab (12CA5, Roche; 1/2000; PBS/0.05% Tween 20) and anti-p30 (American Type Culture Collection; 1/20,000; PBS/0.05% Tween 20) were used to detect the HA-tag and the MLV capsid protein p30, respectively. Synthetic Aβ1–40 (Bachem; 5 μg/ml in PBS) served as a positive control for the 6E10 Ab.

Human Aβ peptide (Bachem) was added in a concentration of 8.3 μg/ml in PBS per well on 96-well ELISA plates (Nunc). After blocking with 5% BSA, 20-fold prediluted serum was serially 3-fold diluted (20 × log₃) in PBS, 0.1% Tween 20, and 1% BSA and added for 2 h at room temperature to the ELISA plates. Upon thorough washing, the HRP-conjugated polyclonal rabbit Ab directed against mouse Ig of the M or G subclasses (anti-mouse IgM, anti-mouse IgG2a, anti-mouse IgG2b from Zymed Laboratories; anti-mouse IgG1, anti-mouse IgG3 from Southern Biotechnology Associates) was added in a 1/500 or 1/1000 dilution. After 1 h of incubation at room temperature, plates were washed, and for the detection of bound HRP-coupled, ABTS substrate was added (0.5 mg/ml 2.2-azino-di-ethyl-benzo-thiazolinsulfonat (Roche) in 0.1 M NaH₂PO₄ (pH 4) and 30% H₂O₂). ELISA titer was determined by endpoint titration, and the OD was measured at a wavelength of 405 nm.

Intracellular cytokine staining was performed according to the Cytofix/Cytoperm plus GolgiPlug kit protocol (BD Bioscience). Briefly, 1 × 10⁶ splenocytes of experimental and control mice were restimuated for 5 h with the different peptides (Aβ1–40 and Aβ1–16; Bachem) at a concentration of 20 μg/ml or with Con A (concentration 25 μg/ml; Roth). After 2 h of restimulation, GolgiPlug (brefeldin A) was added to block cytokine secretion, which increases intracellular accumulation. Then, cells were washed with FACS buffer (PBS supplemented with 1% BSA and NaN₃) and incubated with an Ab directed against CD16/CD32 (Caltag Laboratories) to block FcR. Surface staining was performed by using PE-labeled anti-mouse CD3 Ab (BD Biosciences) and FITC-labeled anti CD4 Ab (Caltag Laboratories). Cells were washed and fixed, and CD4⁺ T cell subsets producing IFN-γ were detected using an Alexa Fluor 647 labeled rat anti-mouse IFN-γ Ab (BD Biosciences). IFN-γ-producing CD3⁺/CD4⁺ T cells were analyzed on FACS BD LSRII (BD Biosciences) using BD FACSDiva software (BD Biosciences).

Sagittal 40-μm vibratome sections were stained with anti-Aβ rabbit antiserum NT12. Systematic-random series of brain sections at three different anatomical planes per animal were used for the analysis. Amyloid plaques were quantified using an MCID image analyzer (Program Version M7 elite; Imaging Research, Brock University). The microscopic image was digitized by use of a Roper black-and-white CCD TV camera and stored with 1124 × 1124 pixel resolution at 256 gray levels. Diffuse amyloid was captured by a detail enhancing function of the M7 software. The pixel size was calibrated using an object micrometer at ×5 magnification (Leica Neoplan Objective). Using a motor driven microscope stage for exact positioning of adjacent object fields the entire neocortex, hippocampus, thalamus, and caudate, putamen of each section was analyzed. For each object field, the anatomical area was defined by manual outline. For each individual section, the sample area was defined by manual threshold setting (gray level) between immunopositive amyloid plaques and tissue background. Isolated tissue artifacts were excluded by manual outline. Raw data were measured as individual counts (amyloid deposits) and proportional area (plaques) per square millimeter and total plaque area in percentage of the individual brain area. Data were transferred to Microsoft Excel sheets for amyloid plaque classification by size and statistical analysis (19, 20).

Histology was performed as described recently (21). In brief, brains were removed and fixed in 4% buffered formalin. Then, tissues were dissected and embedded in paraffin before staining with H&E, MAC-3 (BD Pharmingen) for macrophages/microglia, CD3 for T cells (Serotec), B220 for B cells (Serotec), and GFAP for astrocytes (DakoCytomation). To induce autoimmune encephalomyelitis female mice were immunized s.c. with 200 mg of MOG_35–55 peptide emulsified in CFA containing 1 mg of Mycobacterium tuberculosis (H37RA; Difco Laboratories) as described previously (21). CNS (spinal cord) sections of these animals with autoimmune CNS inflammation served as positive controls for staining.

Homogenates of one brain hemisphere have been processed using two different extraction methods to distinguish between soluble and insoluble, plaque-bound (aggregated) Aβ. Extraction procedures have been accomplished with modifications as previously described (22, 23) and have been repeated at least once to correct for experimental variations. Briefly, brain hemispheres were homogenized in tissue homogenization buffer (THB; 250 mM sucrose, 20 mM Tris-HCl (pH 7.4), and 1 mM EGTA supplemented with Complete (Roche)) at 1 ml per 100 mg of fresh weight using a Potter S (Sartorius) 20 times at 400 rpm. One milliliter of extract diluted 1/10 with THB was extracted with 1 ml of diethylamine/NaCl solution to extract soluble Aβ. The solution was homogenized using a Potter S 20 times at 1500 rpm. Insoluble, plaque-bound Aβ was extracted using formic acid (FA) by adding 1 ml of extract diluted 1/10 with THB to 2.2 ml of FA. The solution was sonified for 1 min (output 4, 40%) using a Branson 250-A sonifier (Branson). Both extracts were subject to ultracentrifugation (100,000 × g, 1 h at 4°C). To regain appropriate buffer conditions for ELISA, 900 μl of diethylamine/NaCl solution extracts was reconditioned with 100 μl of 0.5 M Tris HCl (pH 6.8), and 420 μl of FA extract was reneutralized with 8 ml of 1 M Tris HCl and 0.5 M Na₂HPO₄ (not pH adjusted). Aliquots were stored at −80°C.

High-sensitivity, optimized anti-Aβ sandwich ELISA was performed as described before (24). Briefly, Aβ preparations were used in 1/5 dilutions for soluble and undiluted for aggregated Aβ. Capture Abs were specific for Aβ40 and for Aβ42, respectively, binding C-terminal epitopes of Aβ. Biotinylated detection Ab was generated against the Aβ N terminus detecting all Aβ specimens. ELISA development used poly-SA-HRP conjugate (GE Healthcare) and tetramethylbenzidine (Pierce) as substrate and 1 M H₂SO₄ as stop solution. Absorption was determined using a BioTek EL808 microplate reader, and results were analyzed using Gen5 software (both Bio-Tek Instruments). Measurements were conducted in triplicates to correct for assay variation, and extractions and determinations have been repeated at least twice to correct for experimental variations.

To assess binding of Abs produced in response to Aβ retroparticles to fibrillar Aβ, vibratome sections (50 μm) of APP23 mouse brain were incubated with crude antiserum of vaccinated mice (pre- and postimmunserum). In brief, after a blocking step, sections were incubated in pre- or postimmunserum at different dilutions (1/100, 1/1000, and 1/10,000; diluted in 0.1% Triton X-100 and 1% BSA in 0.05 M TBS) overnight at 4°C. Some sections were simultaneously incubated in the anti-Aβ rabbit antiserum NT12 (Novartis Pharmaceuticals). For the detection of tissue-bound Abs, the sections were washed in buffer and incubated for 2 h at room temperature in an appropriate secondary Ab (Alexa 568-conjugated anti-mouse IgG, 1/1000 or Alexa 488-conjugated anti-rabbit IgG, 1/1000; both from Molecular Probes). Finally, sections were mounted in Dako Fluorescent Mounting Medium (DakoCytomation) and photographed using an Olympus BX50 microscope. In control experiments, the primary antisera were omitted, and sections were incubated only in the secondary Abs. Under these conditions, no specific immunocytochemical staining was observed.

To develop an Aβ epitope-based vaccine, we engineered as a novel immunogen retrovirus-like particles derived from MLV that display the N-terminal amino acid residues of human Aβ (Aβ retroparticles). When expressed in mammalian cells, the viral gag-encoded proteins self-assemble into noninfectious VLPs, which bud from the cell surface in the absence of the viral envelope protein (25). Moreover, overexpressed foreign transmembrane proteins become incorporated into these retroparticles. To generate Aβ-displaying retroparticles, expression constructs encoding the Aβ1–15 peptide were generated, thus omitting Aβ-specific T cell epitopes (26, 27). For this, an Aβ1–15 encoding fragment was ligated into pDisplay, encoding a CMV promoter-driven murine Ig-κ chain signal peptide at the N terminus and the transmembrane domain of the platelet-derived growth factor receptor (PDGFR) at the C terminus (pDAβ1–15) (Fig. 1,A). For generation of Aβ retroparticles, HEK-293T cells were cotransfected with the plasmids pDAβ1–15 and pHIT60, the latter encoding retroviral gag and pol genes. The Aβ1–15 peptide fused to the PDGFR transmembrane domain was inserted into the plasma membrane (Fig. 1,A). Cotransfected cells thus released Aβ retroparticles, which can be harvested from the cell supernatant. To verify display of Aβ fusion protein on retroparticles, we used electron microscopy combined with immunogold labeling. Aβ retroparticles stained with an anti-Aβ Ab revealed strong and specific accumulation of gold particles on the retroparticle membrane, indicating that Aβ was efficiently displayed (Fig. 1,B). In contrast, no immunogold staining was found in experiments with retroparticles displaying epidermal growth factor (EGF) as a control (Fig. 1 B).

Supernatants of HEK-293T cells transfected with both pDEGF/pHit60 and pDAβ1–15/pHit60 were analyzed by Western blotting for particle incorporation of the Aβ fusion protein using the Aβ-specific mAb 6E10 (Fig. 1,C). As expected, the Aβ fusion protein was detectable in pDAβ1–15 particle preparations giving rise to a band corresponding to ∼14 kDa, indicating that Aβ protein was readily incorporated into retrovirus-like particles (Fig. 1,C, upper panel, lane 3). Both, the anti-Aβ Ab and the anti-HA Ab detect the HA-Aβ1–15–PDGFR fusion protein at the expected m.w. of 13.5 kDa (Fig. 1,C, open arrowheads). In contrast, no Aβ-specific signal was obtained in pDEGF control particle preparations (Fig. 1,C, lane 2). Synthetic Aβ1–40 peptide (5 ng) was loaded as a positive control (Fig. 1,C, lane 4) and is detected by the 6E10 Ab (Fig. 1,C, upper panel, filled arrowhead). To characterize particle preparations in more detail, supernatants of pDAβ1–15 and pDEGF-transfected cells were further analyzed using an anti-HA-specific Ab and an Ab directed against the MLV capsid protein p30. As expected, expression of the HA-tag was detected in both EGF and pDAβ1–15 particle preparations (Fig. 1,C, middle panel). The MLV capsid protein p30 was found to be present in all supernatants of pDAβ1–15/pHit60-transfected cells (Fig. 1 C, lower panel), confirming that similar amounts of retroparticles had been applied.

Furthermore, we used ELISA analysis for titration of individual particle stocks. To this end, ELISA plates were coated with serial dilutions of two independent particle preparations (prep1 and prep2) and the amount of Aβ peptide expressed on the surface of retroparticles was quantified relative to a synthetic Aβ peptide standard using the Aβ-specific mAb 6E10. ELISA curves of both particle preparations were comparable, thus revealing similar incorporation rates of Aβ (Fig. 1 D). In parallel, particle numbers were determined by reverse transcriptase assays. On average, stocks contained ∼1 × 10¹² and 5 × 10¹² particles per ml and between 10 and 20 μg/ml Aβ. Thus, ∼10,000 molecules of Aβ per particle are being displayed.

To analyze immunogenicity of Aβ retroparticles, WT mice were immunized once with ∼10¹⁰ Aβ retroparticles (corresponding to 150 ng of Aβ1–15) applied i.v. without additional adjuvant. This dose had been determined in preliminary immunizations as being the minimal dose still inducing a similar humoral immune response (for particle dosedependency of Ab titers, see Fig. 2,A). In comparison to Aβ-specific mAb (6E10), serum of Aβ retroparticle-immunized WT mice showed high Aβ-specific Ab titers 14–28 days after vaccination that slightly declined over time (Fig. 2B). However, even at day 230 postimmunization low but still significant Aβ-specific titers were detectable when compared with preimmune control serum. After reboosting with a single Aβ retroparticle injection (at day 250), Aβ-specific Ab titers increased up to initial immunization levels (Fig. 2 C). Thus, memory B cells were generated that can rapidly induce high Aβ-specific IgG titers after boosting.

Furthermore, we tested the subclasses of Aβ-specific Abs by ELISA. Serum of Aβ retroparticle-immunized WT mice were taken at days 7 and 21 after first immunization and analyzed for IgM or IgG Abs. The analysis of IgM Abs showed high Aβ-specific binding 7 days after immunization (Fig. 3,A). Most notably, after 2 wk, IgM titers decreased, whereas IgG levels increased (Fig. 3 B). Among IgG subclasses, mainly IgG1 and IgG2b were detected in the serum 21 days after single vaccination, indicating that predominantly a Th2-dependent Ab response was induced. Thus, Aβ retroparticles are highly immunogenic and induce long-lived IgG Ab titers in WT mice.

We next assessed potential adverse effects of Aβ retroparticle immunization due to autoreactive T cells and/or inflammatory responses. To this end, WT mice were immunized six times (with 3-wk intervals) either with Aβ retroparticles or fibrillar Aβ1–42 emulsified in Freund’s adjuvant, known to induce Th1-type responses. As previously observed, Aβ retroparticles induced robust Ab titers already after the first immunization that were much higher in magnitude compared with titers obtained after repetitive immunization with Aβ1–42/CFA (Fig. 4,A). To examine the activation of autoreactive anti-Aβ Th cells, we analyzed CD3⁺/CD4⁺ splenocytes for the production of IFN-γ by FACS assay (Fig. 4,B). In these experiments, splenocytes of naive or immunized mice were either restimulated with the peptide Aβ1–16 (assumed not to contain a T cell epitope) or Aβ1–40 (containing known T cell epitopes), respectively. Interestingly, IFN-γ was only induced by Aβ1–40 but not by Aβ1–16 in Aβ1–42/CFA immunized mice, consistent with the lack of a T cell epitope in the shorter peptide. Of note, IFN-γ levels detected in CD4⁺ splenocytes of mice immunized with Aβ retroparticles (Fig. 4 B, red symbols) were not significantly different from background values when stimulated with either peptide.

Thorough histopathological examinations of the CNS revealed no obvious infiltration by mononuclear cells such as CD3⁺ T cells, B220⁺ B cells or MAC-3⁺ macrophages in the hippocampus, frontal cortex, brainstem, and cerebellum (Fig. 4,C and data not shown). Notably, control CNS tissue from animals suffering from autoimmune encephalomyelitis showed a plethora of infiltrating cells (Fig. 4 D). Furthermore, no apparent activation of astrocytes or microglia as unspecific signs of brain damage (28) was detectable (data not shown).

In APP23 mice, amyloid depositions appear starting at ∼6 mo of age, predominantly in the neocortex and hippocampus. We next examined whether Aβ retroparticle immunization would also efficiently induce a humoral immune response in APP23 mice that overexpress human Aβ ∼10-fold over endogenous murine Aβ levels. APP23 mice were monthly immunized over a period of 6 mo, starting at the age of 4 mo. As observed for WT mice, high levels of Aβ-specific IgM Abs were present 7 days after vaccination and a subclass switch to IgG Abs was induced at later time points (data not shown). High levels of Aβ-specific IgG Abs were obtained already 10 days after a single injection in the absence of any adjuvant (Fig. 5,A). Strikingly, levels of Aβ-specific Abs obtained after a single Aβ retroparticle immunization or by six consecutive booster injections were comparable in magnitude, suggesting that a single immunization was sufficient to induce a potent humoral immune response (Fig. 5, A and B). Overall, binding strength of immune sera to ELISA plates was comparable to the monoclonal 6E10 Ab used as a positive control, as evidenced by closely overlapping ELISA curves (Fig. 5,B). As determined by ELISA measurements, vaccination of both APP23 and WT mice resulted in comparable anti-Aβ Ab titers (Fig. 5 C).

To assess whether vaccination with Aβ retroparticles affected plaque deposition, vaccinated APP23 mice were sacrificed at 12 mo (49 wk) of age and saggital sections were stained with anti-Aβ rabbit antiserum NT12 (Fig. 6, A and B). The entire neocortex, hippocampus, thalamus, and caudate putamen of each section were evaluated, and quantitative analysis on series of brain sections was performed as described previously (20). Vaccination strongly inhibited plaque deposition as evidenced by a striking and significant reduction of plaque number to 58 ± 7% (SEM) (Fig. 6,C, right panel), as compared with nonvaccinated control mice. Plaque area was reduced to a similar extent (66 ± 14% (SEM) of control; Fig. 6 C, left panel).

To evaluate whether vaccination not only reduced plaque load but also affected cerebral total Aβ levels, brain homogenates were processed using two different extraction methods to distinguish between soluble and insoluble, plaque-bound (aggregated) Aβ pools. Subsequently, Aβ levels were quantified using a high-sensitivity sandwich ELISA measuring Aβx-40 and Aβx-42 separately (24). Consistent with reduced plaque loads the levels of aggregated, FA extracted Aβ40 and Aβ42 pools were both significantly reduced down to ∼60% of controls (Fig. 6,D; Aβ40: 45.2 ± 11.0 ng, in controls vs 21.4 ± 4.4 ng in vaccinated mice; Aβ42: 44.5 ± 8.5 ng in controls vs 26.1 ± 2.1 ng in vaccinated mice; all values per mg fresh weight brain; p ≤ 0.05, Student’s t test). Moreover, also soluble Aβ40 and Aβ42 were reduced to a similar extent (Fig. 6 D; Aβ40: 11.8 ± 1.8 ng in controls vs 6.6 ± 1.1 ng in vaccinated mice; Aβ42: 13.5 ± 1.9 ng in controls vs 9.1 ± 0.8 ng in vaccinated mice; all values per mg fresh weight brain; p ≤ 0.05, Student’s t test). Although the effect of vaccination reduced Aβ40 levels slightly stronger than Aβ42 levels, the overall Aβ40/42 ratio was not significantly changed by vaccination with Aβ retroparticles (data not shown). Thus, Aβ retroparticles are able to induce an Aβ-specific immune response that significantly inhibits plaque deposition and substantially lowers both insoluble and soluble cerebral Aβ40 and Aβ42.

To assess whether Abs produced in response to Aβ retroparticles can bind to fibrillar Aβ, sections from APP23 mice were incubated with crude antiserum of vaccinated mice and analyzed by immunofluorescence. Strong and specific staining of human Aβ plaques was found with Aβ-specific postimmunserum (Fig. 7,A), whereas no signal was detectable with either preimmune serum (Fig. 7,D) or secondary Ab alone (Fig. 7, E and F). Moreover, double-immunolabeling of sections with postimmunserum (Fig. 7,A) and the well-characterized polyclonal rabbit NT12 antiserum (Fig. 7,B) revealed prominent colocalization of immunofluorescence (Fig. 7 C). Thus, Aβ retroparticles induce Aβ-specific Abs in APP23 transgenic mice that recognize aggregated Aβ species.

In this study, we have developed a highly immunogenic novel candidate vaccine based on retrovirus-like particles displaying N-terminal Aβ epitopes for active and adjuvant-free immunization against AD. Ag presentation in a repetitive and regularly spaced array is an important feature that enables B cells to discriminate between self-Ags vs non-self-Ags (29, 30). To overcome tolerance against the self-Ag Aβ in human APP-expressing transgenic mice, we thus used MLV-based retroviral particles as a multivalent and highly ordered platform to display the N terminus of human Aβ. Residues Aβ1–15 that comprise a B cell epitope, but lack known T cell epitopes (26, 27), were fused to the transmembrane domain of PDGFR and displayed on the surface of retroparticles. Using immunoelectron microscopic analysis, Western blot, and ELISA analyses, we demonstrate that Aβ epitopes are efficiently incorporated into retroparticles. Notably, a remarkably high density of on average several thousand Aβ1–15 epitopes per particle was obtained.

Without the use of adjuvant, Aβ retroparticles evoked a potent humoral immune response characterized by high and long-lasting levels of Aβ-specific IgG Abs.

In addition, restimulation induced memory Ab responses in WT mice. Ig subclass analysis revealed a Th-dependent switch from IgM to IgG of Th2 isotypes with high levels of IgG1, IgG2b in the absence of detectable IgG2a and IgG3 levels. Although there is some debate on the importance of IgG subclasses regarding their potential for plaque clearance (31), it is interesting that IgG1 and 2b subclass Abs are efficiently recognized by high-affinity FcγR expressed, e.g., on microglia that are involved in Aβ plaque removal. Of note, IgG2b (but not IgG1) isotypes can bind complement and complement C3 activation has recently been shown to play a crucial role in clearing plaques in APP transgenic animal models (32).

When APP23 transgenic mice were challenged with Aβ retroparticles, a single immunization was sufficient to induce high anti-Aβ titers, comparable in magnitude to titers obtained after several rounds of consecutive boosting, but by far above those observed in animals immunized with Aβ1–42, even in the presence of CFA. The vast amount of Aβ retroparticles will most likely accumulate in spleen and liver after i.v. injection as observed for retroviruses (33, 34). However, a small amount of particles may also be transported to lymph nodes via dendritic cells within minutes after injection (35), which may well contribute to their strong immunogenicity. Importantly, splenocytes of mice immunized with Aβ retroparticles did not react with IFN-γ production when restimulated with Aβ1–16 or full-length Aβ, indicating the potential for avoiding an autoreactive Aβ-specific T cell response. This suggests that other epitopes provided by the VLP induced the Th2 response. The potential of Aβ retroparticles as a safe alternative vaccine is further underscored by the absence of infiltrating lymphocytes and mononuclear cells in the brains of immunized WT mice.

Strikingly, Ab levels evoked by Aβ retroparticles in APP23 mice were similar to levels obtained in WT mice, suggesting that tolerance to Αβ was efficiently overcome. In previous studies with aggregated Aβ1–42 as an immunogen, a marked tolerance of APP transgenic mice had been observed in comparison to WT mice (26, 36, 37, 38). Thus, our data confirm and extend our previous success with retroparticles as an efficient presentation platform for generating autoreactive Abs, e.g., against Aβ (this study) or the cellular endogenous PrP protein as reported previously (16).

Most importantly, Aβ retroparticle vaccination was also highly efficient in significantly reducing Αβ plaque burden in APP23 mice to ∼60% of control levels. Biochemical analysis showed a concomitant similar reduction of insoluble cerebral Aβ. Notably, vaccination not only lowered Aβ40 but also affected Aβ42 to a similar extent. Aβ42 triggers plaque seeding and is a prominent component of compact plaques (39, 40). A serious concern in AD immunotherapy is that resolution of insoluble brain Αβ may lead to increased levels of soluble Αβ species (as recently observed during a follow-up analysis of the AN-1792 trial (41)). Even more important, these Aβ species may also contain Aβ oligomers that are particularly neurotoxic, confer synaptic dysfunction, and impair memory (1, 42). Naturally produced Aβ dimers and/or oligomers of either animal or human origin have been shown to disrupt long-term potentiation—a correlate of synaptic plasticity—when infused into the ventricles. Interestingly, this effect could be prevented by acute systemic infusion of N-terminal Aβ Abs (including 6E10) or diminished by active immunization with synthetic Aβ1–40/42 (42, 43, 44). In this regard, it is striking that Aβ retroparticle vaccination not only reduced insoluble Aβ pools but also potently and significantly lowered soluble Aβ40 and Aβ42 levels down to ∼60% of controls. In contrast to our findings, two recent studies using Aβ epitope vaccination either reported no effect on soluble Aβ (45) or even observed a significant increase in soluble cerebral Aβ (46) (although vaccination still induced subtle behavioral improvements).

Abs binding to the Aβ N terminus have previously been shown to recognize monomeric, soluble oligomeric, and fibrillar forms of Aβ (31, 45, 47, 48, 49, 50, 51). There is evidence for three principal nonexclusive mechanisms how various Aβ Abs may function in Aβ clearance: 1) Abs may directly bind to Aβ and thereby either prevent oligomerization and fibril formation or dissolve Aβ aggregates; 2) Abs may evoke microglia-mediated removal of Aβ deposits (reviewed in Ref. 2); and 3) Abs may bind Aβ in the plasma and thereby may enhance efflux of soluble Aβ from the brain to the plasma (also termed “peripheral sink mechanism”). In this study, we have shown that Αβ retroparticle-induced sera bind to monomeric Aβ1–15 species on ELISA plates and also recognize Aβ deposits in plaques of brain sections from APP23 mice. Aβ retroparticle vaccination efficiently targeted both insoluble as well as soluble forms of Aβ.

During the past years, several studies described vaccination approaches directed against N-terminal Aβ epitopes. These strategies included Aβ epitope presentation in tandem and/or branched structures either alone or in conjunction with known strong Th cell epitopes (31, 45, 46, 49, 52). In some approaches, immunogenicity was further enhanced by coadministration of strong adjuvants (45, 46). However, adjuvants that potently stimulate innate immune responses may have unwanted effects in humans—even if considered safe in mouse models (10, 11, 12, 14, 53). Alternatively, Aβ display via Escherichia coli filamentous or Qβ phage (47, 54, 55) and Aβ presented on the surface of nonenveloped papilloma VLPs has been used (52, 55) as immunogens. Zamora et al. (51) fused Aβ1–9 to the major capsid protein L1 of Bovine papilloma virus. Although these Aβ1–9 VLP presented Aβ at a rather high density (360 copies per particle) and evoked Abs that inhibited fibril formation in vitro, no significant reduction of plaque burden or cerebral Aβ levels were obtained in vaccinated APP transgenic mice (51). Another study used papilloma VLP or Qβ phage that carried cross-linked Aβ epitopes of variable length (54). Interestingly, this led to high Ab titers (comparable to Aβ1–42/CFA immunization) and Th cell responses that were predominantly directed against VLP epitopes. In this study, however, potential effects on plaque load and cerebral Aβ were not addressed.

Thus, adjuvant-free Aβ retroparticle vaccination appears as a novel alternative strategy to previous and ongoing approaches. Possible advantages compared with passive immunization are a polyclonal and therefore more versatile immune response (including different IgG subclass Abs), absence of an inactivating Ab response to therapeutic Abs (56), considerably reduced costs, and less frequent dosing. Of note, in this study Aβ-specific Ab responses can be restimulated upon a single Aβ retroparticle boost, which suggests that retroparticle immunization may allow to considerably reduce the boost frequency to maintain sufficiently high, therapeutically effective Ab titers. In this regard, it is highly encouraging that in anti-VSV immunizations a single VSV-G retroparticle vaccination was sufficient to generate protective titers (17). However, it is also clear that in contrast to infectious diseases, anti-AD vaccination strategies will likely need to confer a certain constant minimal level of anti-Aβ Abs, as Aβ is continuously produced as a natural product of cell and brain metabolism. A further important aspect favoring retroparticle-based immunization strategies is that VLPs provide strong cross-species T cell epitopes that are sufficient to induce Th function in mice and humans without additional adjuvants. Finally, VLP-based immunizations have been proven safe and versatile in clinical studies with papilloma virus-based VLPs (57, 58, 59) and importantly also recently with retroviral HIV-based particles as novel AIDS vaccines (60, 61). In conclusion, Aβ retroparticles hold great promise to be further developed as a novel and safe active immunization strategy against AD.

We thank Franziska Stöcklin, Julia Gobbert, and Anke Biczysko for excellent technical assistance. We also thank Konrad Beyreuther and Stefan Kins for support in the ELISA measurements and Angela Dann for help in histological analysis.

The authors have no financial conflict of interest.