Phosphorylation of the Amyloid-Beta Peptide Inhibits Zinc-Dependent Aggregation, Prevents Na,K-ATPase Inhibition, and Reduces Cerebral Plaque Deposition

The triggers of late-onset sporadic Alzheimer’s disease (AD) are still poorly understood. Impairment of protein phosphorylation with age is well-known; however, the role of the phosphorylation in β-amyloid peptide (Aβ) is not studied sufficiently. Zinc-induced oligomerization of Aβ represents a potential seeding mechanism for the formation of neurotoxic Aβ oligomers and aggregates. Phosphorylation of Aβ by Ser8 (pS8-Aβ), localized inside the zinc-binding domain of the peptide, may significantly alter its zinc-induced oligomerization. Indeed, using dynamic light scattering, we have shown that phosphorylation by Ser8 dramatically reduces zinc-induced aggregation of Aβ, and moreover pS8-Aβ suppresses zinc-driven aggregation of non-modified Aβ in an equimolar mixture. We have further analyzed the effect of pS8-Aβ on the progression of cerebral amyloidosis with serial retro-orbital injections of the peptide in APPSwe/PSEN1dE9 murine model of AD, followed by histological analysis of amyloid burden in hippocampus. Unlike the non-modified Aβ that has no influence on the amyloidosis progression in murine models of AD, pS8-Aβ injections reduced the number of amyloid plaques in the hippocampus of mice by one-third. Recently shown inhibition of Na+,K+-ATPase activity by Aβ, which is thought to be a major contributor to neuronal dysfunction in AD, is completely reversed by phosphorylation of the peptide. Thus, several AD-associated pathogenic properties of Aβ are neutralized by its phosphorylation.


INTRODUCTION
The amyloid-beta peptide (Aβ) is a normal subnanomolar component of biological fluids (Masters and Selkoe, 2012); however, its deposition in the form of amyloid plaques is one of the hallmarks of AD (Selkoe and Hardy, 2016). Amyloid plaques are associated with neuronal loss and cognitive impairment (Musiek and Holtzman, 2015), and they enhance the tau pathology (He et al., 2017). The trigger of the pathologic Aβ aggregation in AD is unknown (Musiek and Holtzman, 2015); however, in animal models of AD, the conversion of a monomeric Aβ into fibrillar aggregates, through neurotoxic oligomers, is triggered by chemically and structurally modified Aβ species (Meyer-Luehmann et al., 2006;Prusiner, 2012;Jucker and Walker, 2013). Amyloid plaques are abnormally rich in Fe, Cu, and Zn ions (Cummings, 2004), and data from animal models suggest that the formation of amyloid plaques is zinc dependent (Friedlich et al., 2004;Frederickson et al., 2005;DeGrado et al., 2016;James et al., 2017). It has been assumed that zinc ions promote Aβ aggregation via a population shift of polymorphic states (Miller et al., 2010). Zinc-induced aggregation of Aβ is governed by its metal-binding domain (Aβ 16 ), which comprises the N-terminal region (residues 1-16) of Aβ (Istrate et al., 2016). It has recently been shown that the metal-binding domain of Aβ containing isomerized Asp7 (isoD7-Aβ 16 ) is more prone to zinc-induced oligomerization (Istrate et al., 2016), suggesting a role for Asp7 isomerization in the initiation of the pathological aggregation process. Indeed, a full-length isoD7-Aβ 42 peptide is more neurotoxic than the unmodified peptide (Mitkevich et al., 2013) and is able to trigger cerebral amyloidosis in vivo (Kozin et al., 2013). Thus, the ability of Aβ to form zinc-induced aggregates may correlate with its amyloidogenic and neurotoxic properties.
Recently, it has been shown that Aβ undergoes phosphorylation at Ser8 both in vitro and in vivo (Kumar et al., 2011(Kumar et al., , 2013. In the absence of zinc ions, phosphorylated Aβ (pS8-Aβ) was found to be a fast-aggregating peptide species, producing stable fibrillar aggregates and neurotoxic oligomers in vitro (Rezaei-Ghaleh et al., 2016;Jamasbi et al., 2017). In the presence of Zn 2+ , the metal-binding domain of pS8-Aβ (pS8-Aβ 16 ) forms stable dimers (Kulikova et al., 2014); however, in contrast to the zinc-mediated behavior of similar domains in several other Aβ isoforms (including native Aβ), the zinc-bound dimers of pS8-Aβ 16 do not give rise to larger oligomers and aggregates (Istrate et al., 2016). Moreover, the zinc-driven heterodimers formed between pS8-Aβ 16 and isoD7-Aβ 16 or Aβ 16 also cannot oligomerize. We hypothesized that pS8-Aβ 42 does not aggregate in the presence of zinc ions and that pS8-Aβ 42 can prevent the aggregation of native Aβ through the formation of non-propagating heterodimers between pS8-Aβ 42 and Aβ (Mezentsev et al., 2016). By contrast, phosphorylation at Ser8 may change the interaction of Aβ with other proteins, such as Na + ,K + -ATPase. Previously, it was shown that Na + ,K + -ATPase activity was inhibited in post-mortem tissues of AD patients and in amyloid-containing hippocampi of transgenic mice (but not in the amyloid-free cerebellum) (Dickey et al., 2005;Kreutz et al., 2013;Zhang et al., 2013). The latest studies demonstrated that Aβ 42 in form of monomers or oligomers directly binds to Na + ,K + -ATPase, which results in the inhibition of the enzyme as well as the triggering of intracellular signaling cascades (Ohnishi et al., 2015;Petrushanko et al., 2016). Therefore, phosphorylation at Ser8 might neutralize some pathogenic properties of Aβ. Furthermore, the concentration of pS8-Aβ is very likely to change with age since brain aging is associated with changes in the activity of kinases and phosphatases in nerve tissue (Jin and Saitoh, 1995;Norris et al., 1998). An age-related shift in the neuronal protein phosphorylation has recently been shown in Drosophila models (Thomas and Haberman, 2016). Thus, phosphorylated Aβ species could be significant for the development of late-onset AD. However, there are no published data on the zinc-dependent oligomerization and related properties of a full-length pS8-Aβ peptide. In the present study, we investigated the role of pS8-Aβ 42 as a potential quencher of zinc-induced oligomerization of endogenous Aβ species and of pathological effects associated with this process, such as the inhibition of Na + ,K + -ATPase and induction of cerebral amyloidosis, in AD model mice.

Host Mice
B6C3-Tg(APPswe,PSEN1dE9)85Dbo mice were used at the age of 2-8 months (weight of 24-26 g). Mice were housed in the Pushchino Animal Breeding Facility (branch of the Shemyakin and Ovchinnikov Institute of Bioorganic Chemistry, Russian Academy of Sciences), under specific pathogen-free conditions. Housing and use of laboratory animals were approved by the commission IACUC,protocol No. 479 The amino acid sequence of the peptide was confirmed with an ultra-high resolution Fourier transform ion cyclotron resonance mass-spectrometer 7T Apex Qe BRUKER (Bruker Daltonics, Bellerica, MA, United States) utilizing a de novo sequencing approach based on a CID fragmentation technique as described earlier .

Synthetic pS8-Aβ 42 Preparations for Injections
Two thousand micrograms of peptide pS8-Aβ 42 were dissolved in 2000 µL of sterile physiological saline (PS), then the prepared solution was filtered through a 0.22µm filter (Millex-GV, Millipore), aliquoted to 125 µL and frozen. For injection, one aliquot was thawed, sterile PS was added to obtain 1500 µL of solution with a peptide concentration of 0.08333 µg/µL ("administration solution"). Then, 150 µL of "administration solution" were withdrawn and 125 µL of this solution were injected into one animal. Thus, with a single injection, 10 µg of the peptide were ingested.
The aggregation state of the synthetic peptide pS8-Aβ 42 in the samples used for the injections, characterized by us using a standard test based on thioflavin T (Ban et al., 2003), did not change during the time of storage (1-8 months), and did not differ from that of the freshly prepared solutions of the corresponding peptides.
Synthetic Aβ 42 and pS8-Aβ 42 Preparations for Na,K-ATPase Studies To prepare working solutions of Aβ 42 and pS8-Aβ 42 peptides, cold hexafluoroisopropanol ("Fluka") was added to dry Aβ to a concentration of 1 mM and incubated for 60 min at room temperature. The resulting solution was then placed in ice for 10 min and transferred to non-siliconized microcentrifuge tubes (0.56 mg peptide per tube). The peptides in the tubes were vacuum-dried with Eppendorf Concentrator 5301. The resulting dry peptides were stored at −80 • C. A fresh 2.5 mM Aβ solution was prepared by adding 20 µl of 100% anhydrous dimethyl sulfoxide ("Sigma") to 0.56 mg of the peptide, followed by incubation for 1 h at room temperature.

Synthetic Aβ 42 and pS8-Aβ 42 Preparations for DLS Measurements
To measure the average diameter of Aβ aggregates, synthetic peptides Aβ 42 and pS8-Aβ 42 were treated with hexafluoroisopropanol, dried, and dissolved in 10 mM NaOH at concentration of 0.5 mM. Peptides were brought to required concentrations in 10 mM HEPES (pH 7.4) supplemented with 150 mM NaCl, using appropriate buffers.

Dynamic Light Scattering
Dynamic light scattering measurements were carried out using a Zetasizer Nano ZS apparatus (Malvern Instruments, Ltd., United Kingdom) at 25 • C in accordance with the manufacturer instruction. The instrument is equipped with a He-Ne laser source (λ = 632.8 nm) and operates in the back-scatting mode, measuring particle size in the range between 0.6 nm and 10 µm. The particle size distribution is estimated with a spherical approximation of particles, employing a CONTIN data analysis utility available as a part of the instrument software, and used to calculate the average particle diameter (D). The aggregation of Aβ 42 and pS8-Aβ 42 peptides was triggered by diluting peptide solutions with a ZnCl 2 -containing buffer so as to provide a twofold molar excess of zinc ions over peptides.

Turbidimetry
A peptide/zinc mixture was placed in a BRAND UV microcuvette (BRAND GMBH, Germany) and its turbidity was monitored in time (at room temperature) as optical density at 405 nm, using an Agilent 8453E spectrophotometer (Agilent Technologies, United States). Turbidity measurements were started in 0.5 min after triggering the zinc-induced Aβ aggregation as described for the DLS experiments. The initial (zero time) points were measured in the absence of zinc ions, using 25-µM Aβ solutions.

Intravenous Injections
Retro-orbital injections of the venous sinus in mice were performed according to Yardeni et al. (2011). Mice received one retro-orbital injection with 1-month intervals between injections. The contents of injections for each group of mice are described in Table 1. The mice were assigned to the various groups randomly.

Histology and Immunohistochemistry
Euthanasia procedure was applied to 8-month-old mice. Mouse euthanasia was carried out by CO 2 according to the IACUCapproved protocol with the use of automated CO 2 -box Bioscape (Germany). Mice were transcardially perfused with 50 mL of PBS, followed by 50 mL of 4% paraformaldehyde (PFA). Mouse brains were fixed in 10% formalin. Process for paraffin embedding was scheduled as follow: 75% ethanol overnight, 96% ethanol 5 min, 96% ethanol 10 min, 100% ethanol 10 min (two changes), ethanol-chloroform (1:1) 30 min, chloroform overnight. Paraffin embedding was performed at 60 • C for 3 h (three changes). The embedding of tissues into paraffin blocks was done using Leica EG1160 device. Serial brain sections (8 um) were cut using microtome Leica RM2265 mounted onto slides. For deparaffinization, hydration and staining of the sections the following steps were done: slides were consistently put in xylene three changes (10 min each), 96% ethanol (5 min), 90% ethanol (2 min), 75% ethanol (2 min), H 2 O three changes (5 min each), Congo Red solution (5 min), potassium alkali solution, and water. The Immu-Mount medium (Thermo Scientific) was used for mounting.
Immunostaining was carried out as described elsewhere (Kozin et al., 2013). Briefly, sections were deparaffinized and after antigen retrieval by microwaving in citrate buffer washed in PBS and blocked with 10% goat serum in 0.04% Tween20 in PBS (T-PBS). Sections were incubated with primary antibody for 2 h at room temperature, washed thrice in T-PBS, incubated with secondary antibody for 1 h at room temperature followed by washing in T-PBS. The mouse anti-human Aβ monoclonal antibodies 6E10 (Covance, Dallas, TX, United States) diluted in the block solution 1:1000 were used as the primary antibodies, Alexa Fluor 488 goat anti-mouse antibodies (Invitrogen, Grand Island, NY, United States) were used as the secondary antibodies for immunofluorescence staining (dilution 1:1000 in T-PBS). The images were captured with Leica DFS490 digital camera (Leica Microsystems, Wetzlar, Germany) at 100x magnification using Leica DMI 4000 fluorescent microscope (Leica Microsystems, Wetzlar, Germany).

Quantitative Assessment of Cerebral β-Amyloidosis
The sections spanning brain from 0.48 to 1.92 mm relative to the midline in lateral stereotaxic coordinates were used to quantify congophilic amyloid plaques in the hippocampus. Every 15th section was analyzed, yielding 10 sections per animal. Amyloid plaques were identified by Congo Red staining and manually counted as described previously (Ninkina et al., 2009;Bachurin et al., 2012) using Zeiss Axiovert 200 M microscope with 10×, 20×, and 40× objectives (Carl Zeiss, Oberkochen, Germany), with examination under bright-field and between crossed polarizers. Amyloid plaques of all sizes were accepted for counting if they were visible and met the following requirements: red coloring under bright-field and the green birefringence in polarized light. Analyses were undertaken by two researchers independently (EB, SK). To determine the reproducibility of the plaques counts, an intra-class correlation (ICC) was calculated yielding good inter-rater reliability between the two researchers (ICC > 0.85). For each group of mice, the average values ( ± SD) of the plaques number per section were calculated.
Hydrolytic Activity of Na + /K + -ATPase The purified preparation of Na + /K + -ATPase (α1β1 isozyme) was obtained from duck salt glands as described elsewhere (Petrushanko et al., 2016). The purity grade of Na + /K + -ATPase was 99% and specific activity of the enzyme reached ∼2400 µmol of Pi (mg of protein × h) −1 at 37 • C.
Hydrolytic activity of Na + /K + -ATPase in the purified preparation was measured as ouabain-sensitive (1 mM) ATP cleavage in the reaction medium containing 130 mM NaCl, 20 mM KCl, 3 mM MgCl 2 , 3 mM ATP, and 30 mM imidazole, pH 7.4, 37 • C as elsewhere (Petrushanko et al., 2016). Stock solutions containing Aβ 42 and pS8-Aβ 42 peptides, prepared as described above, were added to the reaction medium (without ATP) up to a concentration of 40 µM. An equivalent amount of DMSO was added to the control samples. Following 3-60 min of preincubation with amyloid peptides, the enzymatic reaction was started by adding ATP.
Modeling of the Structure of pS8-Aβ42:Na +/ K + -ATPase Complex Model of the pS8-Aβ 42 peptide was constructed using as templates the Aβ 42 structure modeled in Petrushanko et al. (2016). The resulting model of pS8-Aβ 42 was minimized in the AMBER99 force field with the AutoDockTools program. Modeling of the pS8-Aβ 42 :Na + /K + -ATPase complex was performed using the structure of Na + /K + -ATPase from shark glands 2zxe (PDB id) solved at 2.4 Å resolution (Shinoda et al., 2009), and the modeled structure of pS8-Aβ 42 . Docking has been carried out with VinaAutoDock program (Trott and Olson, 2010), and the docking was constrained to cover only the extracellular part of the protein.

Isothermal Titration Calorimetry (ITC)
The thermodynamic parameters of amyloid peptides binding to Na + /K + -ATPase were measured using a MicroCal iTC200 instrument, as described elsewhere (Mitkevich et al., 2012;Petrushanko et al., 2014). Experiments were carried out at 25 • C in 10 mM imidazole buffer (pH 7.5), containing 130 mM NaCl, 30 mM KCl, 3 mM MgCl 2 . Aliquots (2.6 µl) of ligands were injected into a 0.2-ml cell containing protein solution to achieve a complete binding isotherm. Protein concentration in the cell ranged from 10 to 20 µM, and ligand concentration in the syringe ranged from 100 to 200 µM. The resulting titration curves were fitted using the MicroCal Origin software, assuming one set of binding sites. Affinity constants (Ka), enthalpy variations ( H) and stoichiometry of binding (N) were determined and the Gibbs energy ( G) and entropy variations ( S) were calculated from the equation: G = −RTlnKa = H-T S.

Statistical Methods Used for Data Analysis
Data are presented as means of at least three independent experiments ± SD. The Mann-Whitney test was used for pairwise comparison between examined groups of mice; P < 0.05 was considered significant. The comparison of Na + /K + -ATPase data groups was performed using one-way ANOVA with post hoc testing (using paired samples Student's t-test with Bonferroni correction); after a Bonferroni correction a P-value < 0.016 was considered as statistically significant. Statistical analysis was performed using STATISTICA 8.0 (StatSoft, Inc., Tulsa, OK, United States).

pS8-Aβ 42 Suppresses Zinc-Dependent Aggregation of Aβ 42
Using DLS, we observed a time-dependent formation of Aβ 42 or pS8-Aβ 42 in the presence of Zn 2+ (Figure 1A). Prior to zinc addition, only oligomers 20-30 nm in size were detected. In the absence of zinc ions, there were no appreciable changes in the characteristic size of both Aβ 42 and pS8-Aβ 42 oligomers during the 90 min incubation period (at 25 • C and quiescent conditions). After 10 min of incubation with Zn 2+ , the characteristic diameter of Aβ 42 aggregates reached 700-800 nm, and it became more than 2,000 nm by the end of the observation period (100 min). In contrast, pS8-Aβ 42 did not form aggregates larger than 50 nm in the presence of zinc ions during the entire observation period. The results of the turbidity measurements support those of DLS experiments ( Figure 1B): Ser8 phosphorylation substantially suppressed the propensity of Aβ peptides for zinctriggered aggregation, which was manifested in the remarkably higher turbidity of the Aβ 42 /Zn 2+ mixture than that of the pS8-Aβ 42 /Zn 2+ mixture. The effect of Ser8 phosphorylation on zinc-induced Aβ aggregation was the opposite of that on spontaneous Aβ aggregation: as revealed in the ThT assay, in the absence of zinc ions, pS8-Aβ 42 aggregates much faster than Aβ 42 (Supplementary Figure 1). This result is consistent with those reported by Kumar et al. (2011Kumar et al. ( , 2013. To study a possible effect of pS8-Aβ 42 on the zinc-induced aggregation of Aβ 42 , we determined the characteristic size of zincinduced aggregates in a series of pS8-Aβ 42 /Aβ 42 mixtures with different molar ratios of the peptides (Figure 1C) at the 70-min time point (the time point at which D reaches a plateau at a zinc/Aβ 42 molar ratio of 1:3; Supplementary Figure 2). When equimolar mixtures of pS8-Aβ 42 and Aβ 42 (12.5 µM each) were tested in the presence of 50 µM Zn 2+ , the aggregate diameter decreased to (45 ± 10) nm. Thus, mixing of pS8-Aβ 42 peptides with the unmodified peptide strongly suppressed the zincdependent aggregation of the latter. The half-maximal inhibitory concentration (IC 50 ) for pS8-Aβ 42 was estimated, under the conditions tested, to be approximately 9 µM ( Figure 1C); this value corresponds to about one pS8-Aβ 42 peptide per three Aβ 42 peptides.

Injection of pS8-Aβ 42 Reduces the Amyloid Load in the Hippocampus of Transgenic AD Model Mice
We examined the ability of the synthetic pS8-Aβ 42 peptide to reduce the cerebral amyloidogenesis in an APP/PS1 doubly transgenic murine model of AD. These mice have cognitive features of an AD-like pathology and accumulate significant amounts of dense-core congophilic amyloid plaques starting from 4 to 6 month of age, regardless of the sex (Borchelt et al., 1997;Garcia-Alloza et al., 2006). The experimental groups, which included male and female animals, were subjected to retro-orbital injections of peptide pS8-Aβ 42 (10 µg in 125 µL of PS) starting from 2 months of age. After serial (at monthly intervals) inoculations with the peptide, the host  mice were sacrificed at the age of 8 months. The brains were extracted, and sagittal brain sections (8-µm thick) were analyzed histochemically using Congo Red staining. The hippocampus was chosen as the target region for manual counting of the stained congophilic amyloid plaques using bright-field microscopy in the sections representing the brain layer located between 0.48 and 1.92 mm relative to the midline in lateral stereotaxic coordinates. The congophilic plaques found in the brains of all experimental animals were similar in terms of their location and size distribution in the brain parenchyma (Figure 2). Additionally, immunohistochemical characterization of the congophilic amyloid plaques revealed the presence of Aβ (Supplementary Figure 3). Quantitative analysis revealed a significantly lower number of congophilic amyloid plaques per section in the pS8-Aβ 42 -inoculated 8-month-old transgenic mice (−36.3%, P < 0.05) than that in the untreated littermates (Figure 2 and Table 1).
pS8-Aβ 42 Binds to Na + ,K + -ATPase Without Inhibiting Its Hydrolytic Activity For the measurements, we used purified Na + ,K + -ATPase from duck salt glands, which is a homolog of the α1β1 human isozyme. Earlier, we have demonstrated that the unmodified Aβ 42 inhibits the hydrolytic activity of the enzyme (Petrushanko et al., 2016). In contrast to Aβ 42 , which inhibited Na + ,K + -ATPase after 60min incubation with 40 µM peptide, pS8-Aβ 42 had no effect on enzyme activity ( Figure 3A). The interaction of Na + ,K + -ATPase with the Aβ 42 and pS8-Aβ 42 peptides was measured by ITC (Figures 3B,C). The stoichiometry of binding to Na + ,K + -ATPase was equal to 1 for both peptides, demonstrating that the peptides were predominantly in a monomeric state, as we have shown previously for Aβ 42 (Petrushanko et al., 2016). The binding constants of Aβ 42 and pS8-Aβ 42 with the enzyme were close to each other, and the energy profiles [enthalpy ( H) and entropy (T S)] for binding of both peptides were practically the same ( Table 2). This indicates that phosphorylation of Aβ does not affect the peptide interaction with Na + ,K + -ATPase.

Hydrophobic C-Terminal Domain of Aβ 42
Is Responsible for Its Binding to Na + ,K + -ATPase Structure analysis of the Aβ 42 :Na + ,K + -ATPase complex, which we performed earlier (Petrushanko et al., 2016), has shown that the Ser8 residue is located outside of the interaction site (Figure 4). Introduction of the phosphate group to Ser8 had no effect on the conformation of the Aβ 42 polypeptide chain and did not change the interaction interface (Figure 4).

Metal-Dependent Aggregation of Aβ 42 and pS8-Aβ 42 Correlates With Their
Ability to Inhibit Na + ,K + -ATPase Using DLS, we showed that Mg 2+ at concentrations of 3 mM and above induces the aggregation of Aβ 42 after 10 min of incubation (Supplementary Figure 4). By contrast, no oligomers of pS8-Aβ 42 were observed even after 20 min of incubation with 10 mM Mg 2+ . Since Na + ,K + -ATPase activity was measured in a buffer containing 3 mM Mg 2+ , we hypothesized that Mg 2+ -dependent oligomers of Aβ 42 are required to inhibit Na + ,K + -ATPase and the absence of such aggregates in pS8-Aβ 42 solution determines the absence of inhibition. To support this conclusion, we measured the Na + ,K + -ATPase inhibition by Aβ 42 at different time points and found that the degree of inhibition increases over time and plateaus after 30 min of incubation (Supplementary Figure 5). However, Aβ 42 incubated in Mg 2+ -containing solution for 30 min (in the absence of Na + ,K + -ATPase) did not inhibit Na + ,K + -ATPase if the activity was measured immediately after the addition of the oligomers to the enzyme-containing solution (Supplementary Figure 6). Since both Aβ 42 and pS8-Aβ 42 are able to bind to Na + ,K + -ATPase, we hypothesized that the initial binding of Aβ peptides to the enzyme acts as a seed for further aggregation of Aβ on the enzyme matrix and results in the inhibition of the enzyme. In the case of pS8-Aβ 42 , binding to the enzyme is not followed by oligomerization, which explains the absence of inhibition.

DISCUSSION
Since it was first proposed by Hardy in 1992, the amyloid hypothesis of AD has undergone a number of changes (Selkoe and Hardy, 2016). The focus of research has shifted FIGURE 3 | Interaction of Na + /K + -ATPase with beta-amyloid peptides. (A) Hydrolytic activity of purified Na +/ K + -ATPase was measured after 60 min incubation with 40 µM of Aβ 42 or pS8-Aβ 42 . The histogram represents the enzyme activity in the presence or absence of amyloid peptides, the enzyme activity without Aβ 42 is accepted as 100%. Data are mean values for three independent experiments ± SD. Statistical analysis was performed using one-way ANOVA (F = 1894.8, degree of freedom 2, P < 0.00001) with post hoc testing (using paired samples Student's t-test with Bonferroni correction); after a Bonferroni correction, a P-value < 0.016 was considered as statistically significant; * P < 0.001. ITC titration curves (upper panels) and binding isotherms (lower panels) for Aβ 42 (B), pS8-Aβ 42 (C), and Aβ 17-42 (D) interaction with Na,K-ATPase at 25 • C.
from senile plaques to soluble toxic oligomers of Aβ, and the inherent role of the tau-protein in the AD pathogenesis has been elucidated. Accumulation and aggregation of Aβ are still considered triggers of the AD pathological cascade (Musiek and Holtzman, 2015); however, recent failures of monoclonal antibodies against Aβ in clinical trials call for reassessment of the role of Aβ in AD pathogenesis (Abbott and Dolgin, 2016). It is possible that the pivotal factor is not the total amount of Aβ in the blood but the range and relative quantities of modified Aβ species. A number of different modified Aβ forms, such as isomerized (Roher et al., 1993;Shimizu et al., 2000), pyroglutamylated (Wirths et al., 2009;Nussbaum et al., 2012), truncated (Kummer and Heneka, 2014), and other peptides, have been identified in senile plaques. Previously identified chemical modifications of Aβ seem to increase the pathogenic properties of the peptide. However, based on the present data, Ser8 phosphorylation could be the first identified modification that reverses some disease-associated properties of Aβ.
Protein phosphorylation is a ubiquitous modification, which tightly and precisely regulates the structural and functional characteristics of proteins (Hunter, 1995;Cohen, 2001). Aberrant protein phosphorylation is a disease-modifying factor, one of the most prominent examples of which is the hyperphosphorylation of tau in AD (Cohen, 2001;Ballatore et al., 2007). Recently, pS8-Aβ was obtained in vitro by Aβ phosphorylation with protein kinase A and was subsequently identified in vivo (Kumar et al., 2011(Kumar et al., , 2013. Based on the oligomeric state of pS8-Aβ derived from brain tissue of AD model mice (Kumar et al., 2013) and FIGURE 4 | The model of Na + /K + -ATPase:pS8-Aβ 42 peptide complex on the basis of 2zxe PDB structure. Na + /K + -ATPase is shown as a translucent gray molecular surface, C-terminal part (17-42) of pS8-Aβ 42 peptide is shown in green and the N-terminal metal-binding domains (1-16) is shown in yellow. The phosphorylated Ser8 residue is located outside of the interaction region (highlighted in red). The zinc-binding site 11-14 and pSer8 residue are shown by the ball and stick representation. accelerated aggregation of pS8-Aβ 40 in vitro, Kumar et al. (2011) have proposed that pS8-Aβ represents a potentially pathogenic agent in AD. However, the aggregation of pS8-Aβ was previously studied in the absence of divalent cations, particularly zinc ions, which are involved in both physiological processes and AD pathogenesis (Takeda, 2000;Frederickson et al., 2005;Kumar et al., 2011;Jamasbi et al., 2017). In a complex with Aβ, zinc ions can form seeds of pathogenic aggregation (Frederickson and Bush, 2001;Miller et al., 2010;Bush, 2012). This is especially likely to occur in synapses, where concentrations of Zn 2+ may reach 100s of micromoles per liter (Paoletti et al., 2009).
We have previously shown that phosphorylation of Aβ at Ser8 leads to an increase in the zinc-dependent dimerization of Aβ (Kulikova et al., 2014). Zinc-induced dimerization of the unmodified Aβ occurs through residues 11EVHH14 , and the His6 residue is recruited for further oligomerization (Istrate et al., 2016). However, in pS8-Aβ dimers, His6 forms an additional intramolecular Zn 2+ -binding site with the pSer8 residue, which thereby excludes the His6 residue from further oligomerization (Kulikova et al., 2014). Thus, phosphorylation should lead to a decrease in the ability of Aβ to form zinc-induced aggregates. We investigated the effect of Ser8 phosphorylation on the pathogenic properties of the Aβ species Aβ 42 , which is more prone to aggregation than Aβ 40 and seems to trigger the disease in a number of models (Jarrett et al., 1993;Johnson-Wood et al., 1997;Gouras et al., 2000). As expected, Ser8 phosphorylation reduced the zinc-driven aggregation of Aβ 42 . The formation of aggregates in an equimolar mixture of pS8-Aβ 42 and Aβ 42 was also dramatically suppressed. This may be due to the fact that pS8-Aβ 42 forms zinc-induced heterodimers with the unmodified Aβ 42 , which are not capable of further aggregation. Formation of heterodimers between the metal-binding domains of Aβ and pS8-Aβ was previously observed in vitro (Mezentsev et al., 2016). We hypothesized that in vitro inhibition of zinc-dependent aggregation by pS8-Aβ 42 will result in an anti-amyloidogenic effect in vivo. To test the hypothesis, we studied the effect of retro-orbital injections of pS8-Aβ 42 on the progression of cerebral amyloidosis in a murine model of AD, B6C3-Tg(APPswe,PSEN1dE9)85Dbo/j. We have previously found that retro-orbital injection of isoD7-Aβ 42 , but not of Aβ 42 , promoted the amyloid plaque formation in mice of this line (Kozin et al., 2013). It was the first evidence of the facilitation of cerebral amyloidosis by a synthetic Aβ species injected into the bloodstream. The ability of blood-derived Aβ to induce AD-like pathology has recently been confirmed in a murine parabiosis model (Bu et al., 2017); however, the authors have not identified the Aβ species that triggered the pathogenesis. It is important to mention that isoD7-Aβ 42 exhibits increased zinc-dependent oligomerization in vitro (Istrate et al., 2016), and may serve as a seed of zinc-dependent aggregation of Aβ in brains of transgenic mice. In the case of pS8-Aβ 42 , which is unable to form zinc-dependent aggregates, we expected to observe an opposite effect. Indeed, the number of plaques in the hippocampus of the transgenic mice that received pS8-Aβ 42 injections was approximately two-third of that in the mice that received PS injections. Apparently, pS8-Aβ 42 is unable to serve as an aggregation seed in vivo and also partially prevents the amyloidogenic aggregation of the endogenous Aβ peptides. Based on zinc-dependent aggregation of the Aβ 42 /pS8-Aβ 42 mixture in vitro, it is likely that pS8-Aβ 42 forms heterodimers with Aβ 42 in the nerve tissue, which prevents the formation of senile plaques. This indicates that bloodstream-derived Aβ can serve not only as a trigger but also as an obstacle for cerebral amyloidosis progression, depending on the Aβ species composition. These findings further support the role of zinc-induced aggregation as an important event in the amyloidogenic process.
It is known that cognitive deficits in AD or in corresponding models do not always correlate with the appearance of amyloid plaques and often appear before Aβ aggregates can be detected (Musiek and Holtzman, 2015). Such effects are FIGURE 5 | The possible role of pS8-Aβ 42 in maintaining brain health. In healthy brain synapses (A), the release of Zn 2+ during synaptic transmission does not cause Aβ accumulation, as phosphorylated Aβ does not form zinc-induced aggregates and also prevents zinc-induced aggregation of the intact peptide through the formation of heterodimers. Phosphorylated Aβ does not inhibit Na + /K + -ATPase (NKA) and the heterodimers are easily cleared, therefore the synapse function is not impaired. If pS8-Aβ 42 is depleted (B), synaptic release of zinc ions promotes oligomerization of Aβ, inhibition of Na + ,K + -ATPase and, eventually, leads to the amyloid plaques formation, which impairs the synapse function. associated with soluble toxic oligomers of the Aβ peptide (Haass and Selkoe, 2007) or with receptor-mediated effects of Aβ (Dinamarca et al., 2012). Recently, we have shown that Aβ can bind to Na + ,K + -ATPase (Petrushanko et al., 2016), whose activity is critically important for maintaining electrogenic properties of neurons. Here, we found that binding of monomeric Aβ to the enzyme and subsequent oligomerization of the peptide on the Aβ:Na + ,K + -ATPase matrix leads to the inhibition of enzyme activity. This observation provides a possible explanation for the decrease in the activity of Na + ,K + -ATPase in brain tissues of AD patients (Zhang et al., 2013;Kairane et al., 2014;Petrushanko et al., 2016). Unlike the unmodified peptide, pS8-Aβ 42 does not show an inhibitory effect on Na + ,K + -ATPase. Surprisingly, the binding parameters of pS8-Aβ 42 to the enzyme were almost identical to those of the unmodified peptide; thus, it is not the initial binding that defines the inhibitory properties of Aβ 42 toward Na + ,K + -ATPase. We further showed that the N-terminal domain of Aβ 42 (including the pS8 residue) is not involved in the initial binding and is probably exposed to the solution. Since the N-terminal domain of the peptide governs its metal-dependent oligomerization, we suggest that the inhibition of Na + ,K + -ATPase is caused by the metaldependent formation of Aβ oligomers seeded by the solutionexposed N-terminal domain of the first Na,K-ATPase-bound peptide. Phosphorylation of the peptide at Ser8 seems to interfere with this process, which may correspond to its inability to form oligomers triggered by Mg 2+ (present in Na + ,K + -ATPase activity measurements buffer) or by Zn 2+ . Thus, a decrease in the level of phosphorylation of the peptide in the elderly may lead to the inhibition of Na + ,K + -ATPase, development of neurotoxic effects, and the disruption of nerve transduction long before the appearance of amyloid aggregates. The possible role of Aβ phosphorylation in the brain is presented at Figure 5.

CONCLUSION
In this study, we demonstrated that the phosphorylation of Aβ 42 at Ser8 changes its properties of zinc-driven aggregation, inhibition of Na + /K + -ATPase, and amyloidogenicity. The obtained data indicate that the phosphorylation of Aβ 42 neutralizes some of its AD-associated properties. Our findings provide the basis for discussion about the role of Ser8 phosphorylation in Aβ , which was previously considered only as a pathogenic modification. The anti-amyloidogenic properties of pS8-Aβ 42 in vivo support the intrinsic role of zinc-mediated aggregation in the formation of the senile plaques. Further studies addressing the role of pS8-Aβ in the human brain are needed for better clarification of both the significance of pS8-Aβ and the relevance of the obtained data to AD.

AUTHOR CONTRIBUTIONS
EB and IP performed most of the experiments with contributions from GT, AC, and SR. EB, IP, and VM drafted the paper. SK, OL, and AM contributed the conception and design of the study. All authors contributed to manuscript revision, read and approved the submitted version.