Background
Alzheimer’s disease (AD) is a progressive, incurable neurodegenerative disorder, occurring in mid to late life. The consequences of AD always lead to death, usually 7 to 10 years after diagnosis [
1]. Currently, 40 million people are affected, which makes AD the most common neurodegenerative disorder worldwide [
2].
Two main histological alterations can be discerned in the post mortem brain of AD patients: extracellular senile plaques [
3] and intracellular neurofibrillary tangles [
4]. The former are basically composed of fibrillar Aβ. Aβ peptides are formed by endoproteolytic cleavage of amyloid precursor protein (APP) [
5]. A number of specific mutations in genes related to Aβ production ultimately result in the development of AD. This provides strong support for the amyloid hypothesis of AD, stating that accumulating Aβ represents the central trigger for a cascade of pathological brain changes, eliciting tau hyperphosphorylation, neuronal damage, synapse and cell loss, and dementia [
6]. However, it remains unresolved how Aβ exerts its toxic effects and numerous drug failures in the past, among those small molecule inhibitors of Aβ production and monoclonal antibodies, call the amyloid hypothesis into question. Compelling evidence now also suggests that post-translational modifications of Aβ peptides accelerate their aggregation and induce toxicity, in turn driving disease progression [
7‐
11]. One of these post-translational modifications is the formation of isoaspartate (isoD). The generation of isoD from
l-asparaginyl or
l-aspartyl residues was extensively described earlier [
12,
13]. Both residues can isomerize spontaneously via an
l-succinimidyl intermediate, which either hydrolyzes into
l-isoaspartate (in 2/3 of the cases) or
l-aspartate (in 1/3 of the cases). A minor proportion of
l-succinimide might also epimerize to
d-succinimide at a slow rate, leading to the formation of
d-isoaspartate and
d-aspartate residues, respectively. The isomerization of asparagine and aspartate residues is a spontaneous post-translational modification, which is considered to determine the half-life of proteins [
14‐
16]. Moreover, isoD formation introduces an additional methylene group into the backbone of the protein or peptide [
17,
18], consequently altering its structure. Hence, this modification may also change the properties of proteins like solubility, conformation, and function. The presence of isoD7-Aβ variants in brain of AD patients was first described in 1993 [
12]. By using polyclonal anti-isoD7-Aβ antibodies, it was shown that isoD7-Aβ is present in extracellular deposits in AD brain as well as amyloid-bearing vessels and serves as an indicator of plaque age [
13,
19].
The influence of isoD7-Aβ on amyloid plaque formation is controversially discussed. Despite the fact that isoD7 modification might not influence aggregation of the Aβ peptide [
10,
20], it may be involved in the onset of AD. This modification contributes to the insolubility and stability of Aβ [
21], is located within the zinc-binding site of Aβ [
22], and was described to influence zinc-dependent oligomerization of Aβ (1–16) monomers [
23] as well as hydrolysis of Aβ by the angiotensin-converting enzyme [
24]. Furthermore, isoD7-Aβ was shown to be an exogenous trigger of extensive amyloid plaque formation in AD models [
25,
26] and isoD7-Aβ (1–42) is more toxic for neuronal cells than non-modified Aβ (1–42) [
27]. Finally, further evidence for an involvement of isoD7-Aβ in AD pathology results from an inherited form of AD called Japanese-Tottori FAD. In the affected members of this family, a missense mutation within APP (D678N) replaces the aspartate 7 of Aβ with asparagine [
28]. Asparagine residues undergo about 10 times more rapidly isomerization than aspartate [
29,
30]. Manifestation of AD symptoms in this pedigree may not be due to N7-Aβ, but to the enhanced formation of isoD7-Aβ.
On our quest to decipher the role of posttranslational modifications of peptides and proteins in protein misfolding disorders, we here aimed at investigating the isoD7-Aβ modification. To achieve this goal, we generated specific antibodies recognizing isoD7-Aβ. These antibodies were then used to study the formation of isoD7 in Aβ in vitro and in vivo. Furthermore, we applied one of the antibodies to 5xFAD mice to study a potential therapeutic effect of removing isoD7-Aβ from transgenic mouse brain. The results strongly imply an accumulation of isoD7-Aβ with progression of pathology. A specific targeting of isoD7-Aβ might thus represent a therapeutic strategy for treatment of AD.
Methods
Generation of isoD7-Aβ peptides
The synthesis of the peptides was performed according to standard Fmoc solid phase peptide synthesis on a Tetras peptide synthesizer (Advanced ChemTech, Louisville, USA). The C- and N-truncated Aβ-peptides were synthesized at 60-μmol scale as C-terminal amides on Rink amide resin (Iris Biotech) using standard Fmoc/tBu-protected amino acids (Iris Biotech). Non-canonical amino acids were incorporated using Fmoc-L-Asp-OtBu (isoD), Fmoc-D-Asp-OtBu (isod), Fmoc-Tyr (3NO2)-OH (3NY), Fmoc-Ser (PO (OBzl)OH)-OH (phosphoSer), and Boc-Pyr-OH (pE) (Merck Millipore, Iris Biotech, Bachem). The full-length Aβ1–40 peptides were synthesized at 60-μmol scale as C-terminal acids on Fmoc-Val-Novasyn TGA resin (Merck Millipore). Coupling was performed using O-(benzotriazol-1-yl)-
N,
N,
N′,
N′-tetramethyluronium tetrafluoroborate (TBTU) and
N-methylmorpholine (NMM). Fmoc-deprotection was carried out using 20% piperidine in DMF. Final cleavage and deprotection of the peptides was performed using TFA:DOTA (or EDT):H
2O:TIS (30:2:2:1 v/v). After precipitation with cold diethylether, the peptides were purified by preparative RP-HPLC (Phenomenex Luna C18 (2) column, eluents: water and acetonitrile containing 0.04% TFA). Purity and identity were assessed by analytical RP-HPLC, MALDI-TOF MS or ESI MS. However, during the synthesis of isoD7-Aβ (1–40) predominant formation of succinic imide side product was detected by MALDI-MS, which was assigned to a cyclization of the isoD side chain. Hence, in this case, the isoD-residue was introduced via a Fmoc-isoD-Ser-OH pseudoproline dipeptide building block [
31], which was synthesized prior to peptide synthesis by coupling Fmoc-Asp-OtBu and H-Ser-OH using NHS/EDC followed by subsequent cyclization by means of 2,2-dimethoxypropane and p-toluenesulfonic acid.
Antibody derivation, generation, and biophysical characterization
The antibody 3D6 with IgG2b subtype was obtained from the murine Hybridoma Cell Line RB96 3D6.32.2.4 (ATCC). Purified monoclonal 6E10, 4G8, and 4G8-HRP antibodies were obtained from Biolegend, San Diego. In preparation for antibody application to 5xFAD mice, 3D6 and K11 were recombinantly expressed with an IgG2a subtype in Freestyle 293-F cells (Thermo Fisher Scientific) by using the bicistronic vector pVITRO1-neo-mcs (InvivoGen). The isotype control antibody originally possesses an IgG2a subtype and was expressed in hybridoma cells. Antibody purifications from hybridoma or Freestyle 293-F supernatants have been done by Protein G affinity chromatography. Bound antibodies were eluted using 100 mM Glycine-HCl, pH 2.7, and dialyzed twice against PBS (138 mM NaCl, 8 mM Na2HPO4, 1.5 mM KH2PO4, 3 mM KCl, pH 7.1) overnight at 4 °C.
Generation of anti-isoD7-Aβ antibody expressing hybridoma cells
We aimed at generating monoclonal antibodies, which bind to isoD7-Aβ, but not to D7-Aβ. For immunization, the peptide isoD7-Aβ (1–12)-Cys was used. The sulfhydryl group of terminal cysteine residue was used to conjugate the peptide to Bacterial Transglutaminase (BTG) as carrier. For generation of monoclonal antibodies, 8-week-old female BALB/c mice were immunized with the peptide-BTG-conjugates. Mice were immunized intraperitoneally with a water-in-oil emulsion that was prepared by emulsifying both antigens in equal volumes of Freund’s complete adjuvant (priming) or incomplete adjuvant (boosting). After mice showed sufficient antibody titer in serum, they were sacrificed by cervical dislocation. Spleens were aseptically removed, pooled, homogenized, and immortalized by cell fusion using myeloma cell line SP2/0-Agl4 purchased from the German Collection of Microorganisms and Cell Culture (DSMZ GmbH, Braunschweig). The resulting hybridoma clones were screened according their ability to bind isoD7-Aβ (1–18), but not the wild-type peptide Aβ (1–18). Screening of antigen binding occurred via direct enzyme-linked immunosorbent assay (ELISA) and surface plasmon resonance (SPR), immobilizing the antigen (~ 200 RU) onto a streptavidin sensor chip (GE Healthcare). Stable antibody-producing hybridomas have been selected and subsequently cloned for a second time by limited dilution in order to ensure the monoclonality of the hybridomas.
Dot blot analysis
1.5 μl of Aβ peptides (100 ng/μl) were spotted on a nitrocellulose membrane and blocked for 1 h in blocking solution (5% (w/v) milk powder in TBST (TBS + 0.05% Tween 20 (v/v)). Antibodies 4G8, K11, 6E10, and 3D6 were diluted to 1 μg/ml in blocking solution and incubated with the membrane for 1 h, followed by 3 × 5 min washing steps with TBST. Anti-mouse antibody conjugated to alkaline phosphatase (AP) was added and incubated for 1 h, followed by 3 × 5 min washing steps and subsequent colorimetric detection of AP activity by addition of substrates BCIP (5-bromo-4-chloro-3-indolyl-phosphate) and NBT (nitro blue tetrazolium).
Biophysical antibody characterization by SPR analysis
Binding kinetics (ka, kd, and KD) to different Aβ peptides were determined by using Biacore 3000 at a temperature of 25 °C. In order to capture the antibody of interest to a CM5 sensor Chip (GE Healthcare, Product code BR100012), approximately 15,000–20,000 RU of goat anti-mouse IgG (ThermoFisher Scientific, PA128555) were immobilized first. To immobilize the anti-mouse IgG, the carboxymethylated dextran surface of the sensor chip was activated by mixing 0.1 M N-hydroxysuccinimide (NHS) with 0.4 M N-ethyl-N′-(dimethylaminopropyl) carbodiimide hydrochloride (EDC) 1:1. EDC/NHS was applied to the sensor chip for 10 min with a flow rate of 10 μl/min. Goat anti-mouse IgG was diluted to 50 μg/ml in 10 mM sodium acetate, pH 5.5, and injected for 2 × 3 min contact time with a flow rate of 10 μl/min. After deactivation with 1 M ethanolamine, pH 8.5, for 2 × 7 min contact time with a flow rate of 10 μl/min, 0.1 M glycine, pH 1.7, was applied to the sensor chip with a flow rate of 30 μl/min for 3 min, followed by a washing step with HBS-EP buffer (GE Healthcare, Product code BR100188). Capturing of about 2000 RU anti-isoD7-Aβ antibodies occurred with a flow rate of 10 μl/min. To achieve this, antibodies were diluted to 25 μg/ml in HBS-EP buffer and applied to the sensor chip, followed by washing with HBS-EP until the RU signal remained constant. Binding to the clones K16, K23, K29, K119, K129, and K211 was determined by applying 1 nM to 1 μM of isoD7-Aβ (1–18) and 10 nM to 10 μM of Aβ (1–18), respectively, to the antibodies in a multi cycle kinetic analysis. The kinetic constants as well as the dissociation constant were calculated over all recorded sensorgrams using the 1:1 Langmuir binding model. The binding to antibody clone K11 was determined by applying 3–243 nM of isoD7-Aβ (1–18) and 10–810 nM of Aβ (1–18) in a single cycle kinetic analysis. The kinetic constants as well as the dissociation constant were calculated using the single cycle kinetic model. All evaluations were performed using the BIAevaluation 4.1.1 software.
Immunohistochemical analyses
Passive immunization of 5xFAD mice
Discussion
In spite of recent drawbacks in development of anti-Aβ vaccines, the general concept of Aβ oligomer and aggregate removal by antibodies still keeps significant promise. On our quest to develop more tailored drugs for Aβ immunotherapy, we here generated a set of highly specific monoclonal antibodies recognizing post-translational isoD7-modified Aβ. By the means of these antibodies, we confirmed the disease-related presence of isoD7-Aβ in the brain of AD patients as described before by using polyclonal isoD7-Aβ antisera [
12,
13,
19].
IsoD represents a common modification of the peptide backbone, occurring preferably at asparagine but also at aspartic acid residues. The peptide sequence and three-dimensional structure affects the rate of isoD formation; hot spots are typically found if the side chain of the C-terminally adjacent amino acid is relatively small and hydrophilic and is less likely to be formed where bulky or hydrophobic residues are in this position. The most favorable C-flanking amino acids are glycine, serine, and histidine [
45]. Strongly reduced rates of isoD formation are also observed at sites incorporated in rigid secondary structures. Although evidence was mentioned that the N-terminal part of Aβ is rather flexible even in the mature fibril [
46] which should enable isoD formation, a recent cryo-electron microscopic analysis suggested formation of an intermolecular β-sheet within the Aβ fibril [
47]. The rapid formation of isoD7 within preformed fibrils, as shown here for the first time, is rather consistent with an at least partially flexible N-terminal region of Aβ. Thus, it is very likely that isoD7-Aβ accumulates in the brain of AD patients within aggregates during peptide aging to the point of pathological brain examination post mortem. Hence, presence of the modification discerns aged Aβ from newly formed peptides, which presumably have physiological functions [
48].
Moreover, we collected evidence that a mutation causing FAD is associated with an induction of isoD7 formation, thus supporting speculations from previous examinations [
21‐
27]. To address this question, we replaced the aspartate residue in position 7 of Aβ by an asparagine. The resulting N7-Aβ peptide corresponds to the D678N-APP Tottori missense mutation leading to the development of FAD in a Japanese pedigree [
28]. Because the formation of the isoD from asparagine is 10 to 100 times more accelerated compared to aspartate [
29,
49], this experiment could deliver evidence for a possible function of rather isoD7-Aβ, instead of N7-Aβ, in the onset of AD. Transfection of D678N-APP in HEK293 cells leads to significantly enhanced formation of isoD7-Aβ in the cell culture supernatant. Moreover, isoD7-Aβ formation occurs very fast and can be tracked by using our ELISA systems within a few days. This observation is consistent with a putative causal function of isoD7-Aβ within the described Japanese-Tottori FAD pedigree.
In order to examine the efficacy of isoD7-Aβ antibodies in an AD mouse model, we first analyzed the presence of isoD7-Aβ in 5xFAD mice. These mice rapidly accumulate Aβ (42) in cerebrum, where amyloid plaque deposition begins at 2 months of age and increases with aging [
35]. In preparation of passive immunotherapy of 5xFAD mice with our anti-isoD7-Aβ antibody, we provided evidence for the presence of isoD7-Aβ in amyloid plaques in the brains of these mice, starting before the age of 6 months and further increasing with age (see Fig.
3). At every postnatal age, isoD7-Aβ represents a fraction of the total amyloid load. However, the proportional increase of isoD7-Aβ appeared higher with increasing age and pathology (data not shown). This suggests that the isoD7 modification accumulates specifically at ages, where the phenotype, e.g., disturbances in elevated plus maze, develops. Hence, testing of a therapeutic potential of isoD7-Aβ antibodies appeared reasonable.
In order to rank the efficacy of isoD7-Aβ targeting, we directly compared our anti-isoD7-Aβ K11_IgG2a antibody to antibody 3D6_IgG2a, an IgG2a isoform of the murine version of bapineuzumab. Both antibodies show similar binding affinities to the Aβ peptide: 3–5 nM for 3D6 [
43] and 4–6 nM for K11 (Table
1). Furthermore, both antibodies were recombinantly expressed as IgG2a isotype, using the same constant region. Therefore, they should have the same ability to bind and activate FcγRs. However, both antibodies differ in their epitopes: 3D6 recognizes the N-terminus of Aβ (1–40/42), whereas K11 exclusively binds the post-translationally modified variant isoD7-Aβ (Fig.
1a–c).
By applying K11_IgG2a as well as 3D6_IgG2a to 5xFAD mice, we could not only show the significant reduction of isoD7-Aβ and total Aβ level in the brain, but furthermore an improvement of cognitive deficits in different behavioral tests. Despite the fact that the epitope for the anti-isoD7-Aβ antibody K11 is much less abundant, treatment efficacy of K11_IgG2a is similar to or even better than the effects obtained by administration of 3D6_IgG2a. Moreover, in contrast to 3D6_IgG2a, K11_IgG2a treatment did not lead to an increase of plasma Aβ level. This might be due to either sequestration of 3D6_IgG2a by circulating Aβ peptides in the periphery or a general different mode of action. Because Aβ peptides are present in blood and plasma at picomolar concentrations [
50], the circulating Aβ peptides capture the peripherally injected antibodies, thereby potentially reducing the amount of active antibody available for passaging the blood-brain barrier (BBB). It was already shown in other studies that the peripheral administration of monoclonal antibodies directed against non-modified Aβ enhances plasma Aβ amounts [
36,
51]. In accordance with our findings, DeMattos et al. also showed that the Aβ peptides, arising in the plasma of animals after peripheral anti-Aβ antibody application, are completely bound to the administered antibodies and they hypothesized an underlying peripheral sink mechanism [
51]. Besides peripheral sequestration of 3D6_IgG2a, a general difference in the mode of action might underlie the efficacy of both antibodies. Several general mechanisms are proposed how antibodies remove Aβ from the brain, among those the peripheral sink hypothesis [
51,
52] and the initiation of microglial phagocytosis in the brain [
41,
53]. The latter would need the antibody to enter the brain. However, beside the fact that 3D6_IgG2a enhances plasma Aβ, our analysis of 3D6-concentration in the cerebellum of 3D6_IgG2a-treated mice provides strong support for the presence of the antibody in the brain (Additional file
12). Consequently, the applied antibody crossed the BBB, although to a lower extend in comparison to K11_IgG2a, potentially due to previous sequestration by peripheral Aβ peptides.
The current results observed with isoD7-Aβ are reminiscent of results obtained with pGlu3-Aβ antibodies in several terms. Also here, not only the amount of the post-translationally modified peptide is reduced but also total Aβ level [
54]. Moreover, treatment with these antibodies rescued cognitive deficits [
36]. Furthermore, also isoD7-Aβ epitopes reveal an age-dependent accumulation in amyloid plaques as already demonstrated for pGlu3-Aβ [
43]. In addition, the antibodies did not change the amount of Aβ in serum, whereas a strong increase in serum Aβ level was observed with antibodies binding the non-modified Aβ peptide [
36,
43].
Hence, we hypothesize that passive immunotherapy by using anti-isoD7-Aβ antibodies offers several advantages:
(1) Direct targeting of aged, “non-physiological and toxic Aβ varieties”. Newly formed monomeric Aβ-peptides, which presumably have physiological functions, stay untouched. (2) No peripheral sequestration of the anti-isoD7-Aβ antibody by circulating Aβ peptides deriving from non-neuronal tissues, because there are no aged Aβ variants indicated. This might allow the reduction of effective therapeutic antibody amount and thereby decrease of antibody related side effects. (3) Aβ aggregates located in the brain may contain a certain number of isoD7-Aβ peptides, depending on their age. These epitopes are recognized by our antibody and the entire aggregates are marked for microglial phagocytosis in this way. This leads to a reduction of epitope density to be targeted by the antibody molecules and therefore fewer antibodies have to cross the blood-brain barrier. (4) If the posttranslational modification possesses a causal function in the development of AD, as it might be the case for the Japanese-Tottori mutation, the modified peptide will directly be eliminated.
Thus, targeting posttranslational modifications is certainly a different—and even might be an improved—approach than targeting non-modified Aβ.
Limitations
One limitation should be considered. In order to achieve statistically valid results, especially for the behavioral analyses, the cohort of 5xFAD mice should have been larger than 12 animals per group. Furthermore, the results could be confirmed by its implementation in a different institution. Nevertheless, regarding the principles of the 3Rs (replacement, reduction, and refinement) in animal research, we decided to perform an exploratory treatment study with a minimal cohort size.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.