Ser46-Phosphorylated MARCKS Is a Marker of Neurite Degeneration at the Pre-aggregation Stage in PD/DLB Pathology

Visual Abstract


Introduction
Protein aggregation is a widely accepted hallmark of neurodegenerative disorders, including Alzheimer's disease (AD), Parkinson's disease (PD), dementia with Lewy bodies (DLB), frontotemporal lobar degeneration (FTLD), Huntington's disease (HD), spinocerebellar ataxia (SCA), and amyotrophic lateral sclerosis (ALS). In general, the disease-related proteins are assumed to adopt misfolded structures that are subsequently converted to aggregation-prone structures, including ␤-sheets. However, the details of chronological and/or stochastic changes in these structures remain largely unknown, triggering debates about the identities of the true toxic species, as well as whether aggregated or soluble proteins are toxic.
Clinical trials of AD therapeutics have had a considerable impact on these discussions. Passive immunization with anti-A␤ antibodies has been largely successful in decreasing the abundance of extracellular A␤ plaques in the brain. However, those trials reported discrepancies between improvements in A␤-PET and amelioration of clinical symptoms (https://www.alzforum.org/news/researchnews/pib-pet-biomarker-study-confirms-bapineuzumablowers-amyloid). Therefore, elucidation of the early-stage pathologies is now an urgent issue for understanding pathogenesis and developing therapeutics of AD, and the situation is largely similar in the other neurodegenerative diseases including PD/DLB.
Comprehensive phosphoproteomic analysis of brain samples from mouse AD models and human AD patients revealed that changes in the phosphorylation of some proteins are initiated before A␤ plaques are histologically detectable (Tagawa et al., 2015). In one such protein, MARCKS, phosphorylation at Ser46 can be detected in degenerated neurites of mice at the age of 1 month, before cognitive symptoms and A␤ plaques arise (Tagawa et al., 2015). Phosphorylation of MARCKS at Ser46 is triggered by damage-associated molecular patterns (DAMPs), especially by HMGB1, via TLR4 (Fujita et al., 2016).
In this study, we detected similar phosphorylation of MARCKS at Ser46 in BAC-Tg mice overexpressing human normal ␣-synuclein (␣-Syn) in the glucocerebrosidase (GBA)-heterozygous-knockout (KO) background (human ␣-Syn-BAC-Tg/GBA-hetero-KO mice), as well as in human DLB patients. The histologic features of phosphorylation of MARCKS on serine 46 (pSer46-MARCKS) and chronological progression in the brain were similar in AD and PD/DLB. Intriguingly, this marker of neurodegeneration became positive before formation of histologically detectable ␣-Syn aggregates. These results cast light on the relationship between neuronal degeneration and protein aggregation, and reveal that the initiation of neurite degeneration precedes formation of protein aggregates.

Mouse PD/DLB model
Normal human ␣-Syn-BAC-Tg mice were created as previously described (Yamakado et al., 2012). Briefly, the BAC-Tg construct (PAC AF163864 and BAC AC097478, containing 28 kb of the 5=-flanking sequences and 50 kb of the 3=-flanking sequences, in addition to the entire human gene) was microinjected into C57BL6/J ova, yielding homozygous ␣-Syn-Tg mice. GBA-hetero-KO mice were purchased from the Jackson Laboratory (B6.129S6-Gbatm1Nsb/J, stock no. 003321) and mated with human ␣-Syn-BAC-Tg mice. The resultant normal human ␣-Syn-BAC-Tg/GBA-hetero-KO (homo/hetero) mice were maintained as a line. The ␣-Syn-BAC-Tg/GBA-hetero-KO mice were crossed for Ͼ10 generations, and 3 male ␣-Syn-BAC-Tg/GBA-hetero-KO mice were used at the ages of 1, 6, and 24 months for immunohistochemistry and biological analysis. Analysis of detailed phenotypes of ␣-Syn-BAC-Tg/GBA-hetero-KO mice will be reported elsewhere by one of the authors (R.T.).

Western blotting
Mouse cerebral cortex and human temporal lobe tissues were dissolved in extraction buffer containing 2% SDS, 1 mM DTT, and 10 mM Tris-HCl (pH 7.5) and homogenized using 20 strokes of a Dounce glass homogenizer on ice. The crude extracts were centrifuged at 16,000 ϫ g at 4°C for 10 min, added to an equal volume of sample buffer (0.1 M Tris-HCl, pH 7.5, 4% SDS, 20% glycerol, 12% ␤-mercaptoethanol, and 1% bromophenol blue), and boiled at 95°C for 10 min.

Mass analysis
For phosphoproteomic analysis, phosphorylated peptides were prepared as described previously (Tagawa et al., 2015;Fujita et al., 2016). Briefly, cerebral cortex tissues were dissected from the temporal pole or occipital pole of postmortem brains of five females in each group of AD and DLB, their extracts were denatured by detergent and heat treatment, and cysteine residues were reduced and blocked by alkylation. Protein samples were digested with trypsin. Phosphopeptides were enriched using the Titansphere Phos-TiO Kit (GL Sciences), labeled using an iTRAQ Reagent multiplex kit (SCIEX), and fractionated by strong cation exchange chromatography. Each fraction was analyzed using a LC-MS/MS system (Triple TOF 5600 System, Eksigent LC system; SCIEX) and a C18 column (0.1 ϫ 100 mm; KYA Technologies Corporation). The ion spray voltage was 2.3 kV, and the information-dependent acquisition (IDA) setting was 400 -1250 m/z.

Data analysis
Mass spectrum data of peptides were acquired and analyzed by Analyst TF (version 1.6; AB SCIEX). Proteins corresponding to the results were identified by searching UniProtKB/Swiss-Prot (downloaded from http://www.uniprot.org on June 22, 2010) and identified by ProteinPilot (version 4; AB SCIEX), which employs the Paragon algorithm (Shilov et al., 2007). Tolerance for the search of peptides by ProteinPilot was set to 0.05 Da for MS and 0.10 Da for MS/MS analyses. Redundantly identified proteins were excluded using the Pro Group algorithm (AB SCIEX). The confidence scores of protein or peptide identification were calculated by ProteinPilot and used as the confidence threshold. The threshold for detection was set at 95% confidence, and peptides with Ͼ95% confidence were accepted as identified peptides. Comprehensive phosphoproteomic analyses of postmortem human AD and DLB brains. A, Occipital and temporal tips of pathologically pure AD (five females) and DLB (5 females) brains along with age-matched normal controls (5 females) were used for comprehensive phosphoproteome analyses as described previously (Tagawa et al., 2015). Independent phospho-sites were compared between AD and DLB. B, Comparison of changes at four different phosphorylation sites in MARCKS [Ser27/Ser27, Ser46/ Ser46, Ser145/Ser138, and Thr150/Thr143 (human/mouse)] between human DLB and AD patients. (See also Fig. 1-1.) Quantification of proteins was performed through analysis of iTRAQ reporter groups in the MS/MS spectrum that were generated on fragmentation in the mass spectrometer. For quantification of peptides and proteins, bias correction was employed to normalize signals among different iTRAQ reporters under the assumption that the total amount of signal from each iTRAQ should be equal. For quantification of peptides, the bias correction option was used to normalize different iTRAQ signals.
Peptide ratio was calculated as the ratio of reporter signals in patient versus disease control samples after bias correction. The details of this formulation are provided in the manual from AB SCIEX.
The results of the peptide summary from ProteinPilot were output as an Excel file for further data analysis.
Quantity of a peptide fragment was calculated as the geometric mean of signal intensities of multiple MS/MS fragments including the phosphorylation site. For every disease group, biological differences relative to the control group were evaluated by Welch's test. For multipletesting correction, p-values were adjusted using the Benjamini-Hochberg procedure. Phosphopeptides that changed in multiple disease groups were identified and selected for further analysis.

Human brain
Temporal-and occipital-pole human brain samples used for proteome analysis were dissected from five female AD, five female DLB, and five female age-matched Low and high magnification of temporal lobes from non-neurologic disease control patients (5 females) and human DLB patients (5 females). Images were acquired on an Olympus FV1200 IX83 confocal microscopy. All bar graphs indicate averages and SEM. In each patient, quantitative analyses of signal intensities (mean pixel intensities) were performed in 10 visual fields (100 ϫ 100 m) randomly selected from the corresponding area. Statistical analyses were performed with Student's t test; ‫,ءء‬ p Ͻ 0.01. B, Staining of pSer129-␣-Syn revealed multiple cytoplasmic inclusions (Lewy bodies) in the same patient group. C, Costaining of ubiquitin with pSer129-␣-Syn. D, Costaining of pSer46-MARCKS with pSer129-␣-Syn. E, Western blotting analysis of temporal lobes from non-DLB control patients and DLB patients with antibody against pSer46-MARCKS, total MARCKS, and ␤-actin. Graph shows the quantitative result of pSer46-MARCKS from 5 patients and 5 non-neurologic disease controls. The band intensity was normalized against ␤-actin. Statistical analyses were performed with Student's t test; ‫,ء‬ p Ͻ 0.05. Figure 3. Chronological changes of pSer46-MARCKS in human ␣-Syn-BAC-Tg/GBA-hetero-KO mice. pSer46-MARCKS and pSer129-␣-Syn were costained in human normal ␣-Syn-BAC-Tg/glucocerebrosidase (GBA)-hetero-KO mice at 1, 6, and 24 months of age (3 male mice at each time point). Images were acquired by Olympus FV1200 IX83 confocal microscopy. All bar graphs show average and SEM. Three mice were used for each group, and quantitative analyses of signal intensities (mean pixel intensities) were performed in 10 visual fields (100 ϫ 100 m) of each mouse, randomly selected from the brain region. Statistical analyses were performed with two-way ANOVA followed by Student's t test; ‫,ء‬ p Ͻ 0.05; ‫,ءء‬ p Ͻ 0.01. The increase in phosphorylation was first detected in olfactory bulb and frontal cortex, followed by the temporal and occipital cortices. Signal intensities of pSer46-MARCKS increased rapidly in the temporal and occipital regions and became most prominent among multiple brain regions. (See also Figs. 3-1, 3-2, and 3-3.) normal control patients and deep-frozen (-80°C) within 1 h after death.

Ethics
The experiments with human samples were approved by the Ethics Committees of the Tokyo Medical and Dental University (O2014-005-03). All animal experiments were approved by the Institutional Animal Care and Use Committee of Tokyo Medical and Dental University (A2017-242A). The same experiments were also approved by the Committees on Gene Recombination Experiments of Tokyo Medical and Dental University (2016-007C4).

Elevated levels of pSer46-MARCKS in the phosphoproteomes of human AD and DLB
Following previous analyses of brain samples from AD mouse models and human AD patients (Tagawa et al., 2015), we performed comprehensive phosphoproteome analysis with postmortem human DLB patients. In this case, we selected postmortem brains with pure pathology, with intracellular ␣-Syn aggregates in neurons but without extracellular A␤ aggregates or cytoplasmic tau/ TDP43 pathology. The comparison revealed that 107 phosphorylation sites were shared between AD and DLB samples. One such modification, pSer46-MARCKS, was elevated in both AD mouse models and postmortem human AD brains (Fujita et al., 2016; Figs. 1A and 1-1).
Interestingly, the pSer46-MARCKS level was elevated in the temporal but not the occipital lobe in DLB, but in the opposite pattern (i.e., occipital but not temporal lobe) in AD (Fig. 1B). In 5xFAD mice, the increase in the pSer46-MARCKS level at the early stage of the disease (preclinical/pre-aggregation phase) had normalized by the late stage (Fujita et al., 2016). Therefore, the discrepancy in pSer46-MARCKS levels between the two lobes might reflect the lobar dominance of neurodegeneration of each disease. Mass spectrometry results suggesting that common changes occurred in AD and PD/DLB human brains prompted us to investigate whether pSer46-MARCKS also reflects the preclinical/pre-aggregation phase in PD/DLB.

Increased pSer46-MARCKS in immunohistochemistry of human DLB
Next, we investigated whether pSer46-MARCKS was elevated in human postmortem DLB brains. For this purpose, we stained temporal lobe samples from DLB patients and found that apical dendrites marked with MAP2 were costained by pSer46-MARCKS ( Fig. 2A). Meanwhile, we confirmed that this brain region included pSer129-␣-Syn cytoplasmic inclusion bodies (Fig. 2B), which were mostly ubiquitinated (Fig. 2C). Because cytoplasmic staining of pSer46-MARCKS seemed like aggregates ( Fig.  2A), we tested whether they were ␣-Syn aggregates and found that half of the cytoplasmic pSer46-MARCKS-posi- tive structures in neurons were stained with pSer129-␣-Syn (Fig. 2D, yellow arrow); however, the other half were not stained strongly with pSer129-␣-Syn (Fig. 2D, red arrow). The pSer46-MARCKS-positive/pSer129-␣-Syn-negative cells may correspond to surrounding neurons indirectly affected by HMGB1 released from damaged neurons that accumulates pSer129-␣-Syn, suggesting that pathology continues to progress in human brain even at the terminal stage of DLB.
To support the above results of immunohistochemistry, we performed Western blotting of pSer46-MARCKS and total MARCKS with temporal lobe of human postmortem DLB patients and non-DLB control patients (Fig. 2E). The level of pSer46-MARCKS was increased in DLB patients.

Increased pSer46-MARCKS in humanized ␣-Syn-Tg mice
To determine the initial time point of the increase of pSer46-MARCKS in the brain, we analyzed human ␣-Syn-BAC-Tg/GBA-hetero-KO mice at 1, 6, and 24 months of age, when the animals did not exhibit motor dysfunction or behavioral abnormality. The mice had slightly elevated signals of pSer46-MARCKS in the olfactory bulb, frontal cortex, and parietal cortex at 1 month (Figs. 3 and 3-1). Subsequently, the area of pSer46-MARCKS-positive staining expanded to parietal and occipital cortex at 6 months ( Figs. 3 and 3-2), but did not increase in the hippocampus before 24 months of age (Figs. 3 and 3-3).
No definite cytoplasmic ␣-Syn aggregates stained by anti-phosphorylated ␣-Syn (pSer129-␣-Syn) antibody were detected in neurons of the internal plexiform layer and mitral cell layer of the olfactory bulb until 24 months of age (Fig. 4A-C), whereas ␣-Syn-skeinor Lewy neurite-like structures were present in these layers from 1 month of age, and increased in abundance over the course of aging. These structures were not costained with anti-ubiquitin antibody at 1 month, but costaining was evident from 6 months onward (Fig. 4D). Interestingly, pSer46-MARCKS-positive neurites were more frequent than p-␣-Syn-costained neurites. Together, these results indicated that neurite changes detected by pSer46-MARCKS preceded formation of ubiquitin-positive aggregates of ␣-Syn, at least in this PD/DLB mouse model (Table 1).
At 24 months of age, cytoplasmic ␣-Syn aggregates were also detected in parietal cortex (Fig. 5A). The relationships between pMARCKS and p-␣-Syn and between p-␣-Syn and ubiquitin in the cortical neurons during aging were similar to those in olfactory neurons (Fig. 5B).
To confirm these results, we performed Western blot analysis of pSer46-MARCKS, pSer129-␣-Syn, and ubiquitin with total cerebral cortex tissues at 1, 6, and 24 months of age (Fig. 6). The level of phosphorylated MARCKS was elevated at 1, 6, and 24 months of age, whereas phosphorylated ␣-Syn and ubiquitin arose and increased in abundance after 6 months of age (Fig. 6). We also tested the level of pSer46-MARCKS in peripheral blood cells and found it far lower than that in the brain (Fig. 6-1).
Ϯ, positive staining not observed in background mice was observed in Ͻ10% of cells; ϩ, positive staining was observed in 10%-50% of cells of Tg mice; ϩϩ, positive staining was observed in Ͼ50% of cells of Tg mice. ␣-Syn antibody also coprecipitated pSer46-MARCKS in both the mouse model and human patients (Fig. 7A, B), indicating biochemical interaction of the two phosphoproteins.

Activation of the upstream kinases in DLB brains
Finally, we investigated the upstream kinases that phosphorylate MARCKS at Ser46. MARCKS is a representative continued Syn were costained in external and internal pyramidal cell layers at 24 months of age. pSer46-MARCKS was stained in both apical dendrites and cell bodies, whereas pSer129-␣-Syn was stained in cytoplasmic aggregates. B, Parietal cortex tissues from human ␣-Syn-BAC-Tg/GBA-hetero-KO mice were costained for pSer129-␣-Syn and pSer46-MARCKS or ubiquitin at 1, 6, and 24 months of age. Costaining patterns similar to those in the olfactory bulb were confirmed in the parietal lobe. Figure 6. Western blot analysis of ␣-Syn-BAC-Tg/GBA-hetero-KO mice at multiple time points. Left panels show Western blots of whole cortex (3 males at each time point) with antibodies against pSer46-MARCKS, pSer129-␣-Syn, and ubiquitin. Right graphs show quantitative analyses of three independent blots for pSer46-MARCKS, pSer129-␣-Syn, and ubiquitin; band intensities were normalized against ␤-actin. Statistical analyses were performed with two-way ANOVA followed by Student's t test; ‫,ء‬ p Ͻ 0.05; ‫,ءء‬ p Ͻ 0.01. (See also Fig. 6-1.) substrate of PKC (Aderem, 1992;Blackshear, 1993), as known from its name, myristoylated alanine-rich C kinase substrate. It was reported that PKC␣ phosphorylates Ser159, Ser163, and Ser170 (Calabrese and Halpain, 2005). Meanwhile, our previous experiments revealed that Erk1/ Erk2 (MAPK3/MAPK1) instead of PKC phosphorylates MARCKS at Ser46 (Fujita et al., 2016). Consistently, another group also showed that MAPK phosphorylates MARCKS in hippocampal neurons, although the exact phosphorylation sites by MAPK were determined in their study (Ohmitsu et al., 1999).
Therefore, we examined activation of Erk1/2 by immunohistochemistry and Western blot (Fig. 8) by using a rabbit monoclonal antibody detecting Erk1 phosphorylation at Thr202/Tyr204 and Erk2 phosphorylation at Thr185/Tyr187. As expected, immunostaining of pErk1/2 was increased in cerebral cortex of ␣-Syn-BAC-Tg/GBAhetero-KO mice (Fig. 8A). The cells were costained with MAP2 but not with GFAP and were shown to be neurons (Fig. 8B). The stains of pErk1/2 were colocalized with pSer46-MARCKS in the same neurons (Fig. 8A), consistent with their enzyme-substrate relationship. The similar costains of increased pSer46-MARCKS and pErk1/2 were confirmed in temporal lobe cortex of DLB patients (Fig. 8C). Western blot analyses also confirmed the increase of pErk1/2 in the cerebral cortex of ␣-Syn-BAC-Tg/GBA-hetero-KO mice (Fig. 8D) and DLB patients (Fig. 8E). These analyses collectively revealed abnormal increase of Erk1/2 phosphorylation and its age-dependent enhancement in cortical neuron under the PD/DLB pathology (Fig. 8).

Discussion
In this study, we revealed that the level of MARCKS phosphorylated at pSer46, a hallmark of neurite degeneration at the pre-aggregation stage of AD pathology (Fujita et al., 2016), was also elevated in PD/DLB pathologies in both mouse models and human patients. The immunohistochemical staining patterns of pSer46-MARCKS were also similar in AD and PD/DLB patholo-gies. In a mouse model, neurite degeneration revealed by pSer46-MARCKS was initially detected in the olfactory bulb and subsequently became prominent in the occipital and temporal cortices (Table 1). This pattern was consistent with the proposed progression of neurodegeneration in human PD/DLB pathology, in which the earliest lesions appear in the olfactory bulb, and the occipital lobe is ultimately predominant (Barber et al., 2000;Lobotesis et al., 2001;Burton et al., 2002, ;Cousins et al., 2003;Ballmaier et al., 2004;Hanyu et al., 2005;Barkhof et al., 2007;Beyer et al., 2007;Whitwell et al., 2007;Perneczky et al., 2008;Klein et al., 2010;Lee et al., 2010;Sanchez-Castaneda et al., 2010;Galvin et al., 2011;Chow et al., 2012;Hayashi et al., 2012;Kantarci et al., 2012;Watson and O'Brien, 2012;Lebedev et al., 2013).
The second critical conclusion of this study is that neurite degeneration precedes disease-related protein aggregation at the histologic level. In the mouse model, the increase in pSer46-MARCKS preceded formation of ubiquitinated ␣-Syn aggregates ( Table 1). The chronological order in the PD/DLB mouse model, in which pSer46-MARCKS precedes histologic aggregate formation, was analogous to our previous observation in AD model mice, in which pSer46-MARCKS was detected in the cytoplasm and neurites of neurons before staining of extracellular A␤ aggregates was apparent (Fujita et al., 2016). Using this model, we confirmed the importance of pSer46-MARCKS at the biochemical level. These results strongly suggest that intracellular misfolded ␣-Syn in the monomeric or oligomeric form, rather than aggregates, plays the more critical role in initiation of neurite degeneration.
Interestingly, in human postmortem brains, pSer46-MARCKS was elevated in brain regions relatively unaffected by the disease (temporal lobe in DLB, occipital lobe in AD) rather than in severely affected regions (occipital lobe in DLB, temporal lobe in AD). The reason for this has not yet been determined. One possibility is that neurites in severely affected brain regions are metabolically "burned out" and cannot maintain high levels of pSer46-MARCKS. Interaction between MARCKS and ␣-Syn. A, Immunoprecipitates by anti-pSer46-MARCKS or anti-pSer129-␣-Syn antibody from brain samples of ␣-Syn-BAC-Tg/GBA-hetero-KO mice at 1, 6, and 24 months were blotted with anti-pSer129-␣-Syn or anti-pSer46-MARCKS antibody, respectively. B, The similar coprecipitations from temporal lobe cortexes of human DLB patients were examined. Figure 8. Activation of Erk1/2 in cortical neurons of mouse PD/DLB model and human DLB patient. A, Costaining of occipital cortex tissues from ␣-Syn-BAC-Tg/GBA-hetero-KO (Tg) or the nontransgenic sibling control (non-Tg) mice at 1, 6, and 24 months of age, by antibodies against Erk1/2 active forms (pThr202/Tyr204-Erk1 and pThr185/Tyr187-Erk2) and against pSer46-MARCKS. Right graph shows quantitative analysis of pErk1/2 signal intensities in 3 mice (10 visual fields for mean value of each mouse). Statistical analyses were performed with two-way ANOVA followed by Student's t test; ‫,ء‬ p Ͻ 0.05; ‫,ءء‬ p Ͻ 0.01. B, Costaining of pSer46-MARCKS with MAP2 or GFAP in mouse cortex. C, The similar costaining of pErk1/2 and pSer46-MARCKS in human temporal lobe from DLB patients and non-neurologic disease controls. D, Western blot analyses of whole cortex tissues from ␣-Syn-BAC-Tg/GBA-hetero-KO (Tg) or non-Tg mice at 1, 6, and 24 months of age with anti-pErk1/2 and -Erk1/2 antibodies. E, The similar Western blot analyses of temporal lobes from human DLB patients.
Based on these observations, it would be worthwhile to develop pSer46-MARCKS as a biomarker capable of detecting the molecular pathology in PD/DLB at the ultra-early (pre-aggregation/preclinical) phase. In addition, it would be useful to develop mass spectrometry-based and ELISAbased assays and/or positron emission tomography. If such a highly sensitive assay system were to become available, it would lead directly to detection of ultra-early-phase pathology in people at risk of neurodegenerative diseases. Preliminarily, we examined by Western blot whether peripheral blood cells (PBCs) such as red blood cells, granule cells, or lymphocytes express pSer46-MARCKS. If it is the case, discrimination of brain-derived from PBC-derived pSer46-MARCKS becomes difficult, and we need some additional tricks to use pSer46-MARCKS as a diagnostic biomarker outside of the brain. Fortunately PBCs did not express a detectable level of pSer46-MARCKS comparable to brain tissues of 5xFAD mice (Fig. 6-1), supporting the possibility to develop pSer46-MARCKS as a blood biomarker.
In this study, we focused on pSer46-MARCKS but did not extend our analyses to other molecules detected in the phosphoproteome analysis ( Fig. 1-1). Another potentially interesting commonality between AD and PD/DLB is phosphorylation of phosphoglucomutase-1 (PGM1) at Ser117 in the temporal and occipital lobes in patients with both diseases. This enzyme is essential for the conversion of glucose 1-phosphate (generated by glycogenolysis) to glucose 6-phosphate, which is used for energy production by the glycolytic pathway. Further studies are necessary to evaluate the pathologic significance of changes in the phosphorylation-state changes of such additional candidate proteins.