Leucine-Rich α2-Glycoprotein Is a Novel Biomarker of Neurodegenerative Disease in Human Cerebrospinal Fluid and Causes Neurodegeneration in Mouse Cerebral Cortex

Leucine-rich α2-glycoprotein (LRG) is a protein induced by inflammation. It contains a leucine-rich repeat (LRR) structure and easily binds with other molecules. However, the function of LRG in the brain during aging and neurodegenerative diseases has not been investigated. Here, we measured human LRG (hLRG) concentration in the cerebrospinal fluid (CSF) and observed hLRG expression in post-mortem human cerebral cortex. We then generated transgenic (Tg) mice that over-expressed mouse LRG (mLRG) in the brain to examine the effects of mLRG accumulation. Finally, we examined protein-protein interactions using a protein microarray method to screen proteins with a high affinity for hLRG. The CSF concentration of hLRG increases with age and is significantly higher in patients with Parkinson’s disease with dementia (PDD) and progressive supranuclear palsy (PSP) than in healthy elderly people, idiopathic normal pressure hydrocephalus (iNPH) patients, and individuals with Alzheimer’s disease (AD). Tg mice exhibited neuronal degeneration and neuronal decline. Accumulation of LRG in the brains of PDD and PSP patients is not a primary etiological factor, but it is thought to be one of the causes of neurodegeneration. It is anticipated that hLRG CSF levels will be a useful biomarker for the early diagnosis of PDD and PSP.


Introduction
The activation of inflammatory pathways in the brain is increasingly emphasized as a major risk factor for the development and progression of neurodegenerative disease [1][2][3]. Long-term use of anti-inflammatory drugs reduces the risk of Alzheimer's disease (AD) and Parkinson's disease (PD) [4][5][6], and animal studies have provided additional support for this hypothesis [7][8][9]. Furthermore, Berchtold et al. [10] reported that neuroinflammation is present in the brain long before cognitive decline, and immune activation is a highly prominent feature of normal brain aging.
Leucine-rich α2-glycoprotein (LRG) was first identified as a trace protein in human serum in 1977 [11]. This 24-amino acid consensus sequence, termed the leucine-rich repeat (LRR), has since been identified in a large family of proteins [12]. The LRR family is diverse, and its members have been shown to be involved in protein-protein interaction, signal transduction, and cell adhesion and development [13]. Shirai et al. [14] reported that LRG expression was enhanced by acute inflammation induction in mice. Therefore, like C-reactive protein (CRP), LRG might serve as a diagnostic biomarker in some inflammatory conditions.
Gait disturbance and dementia are present in numerous neurodegenerative disorders, including idiopathic normal pressure hydrocephalus (iNPH), PD with dementia (PDD), dementia with Lewy bodies (DLB), progressive nuclear palsy (PSP), and AD. Sometimes it is difficult to correctly diagnose these diseases, but cerebrospinal fluid (CSF)-based diagnosis of AD using the biomarkers amyloid β42 (Aβ42), total tau protein (t-tau), and hyperphosphorylated tau (p-tau) has been reported as useful [15]. However, the other neurodegenerative diseases described above lack disease-specific CSF biomarkers. Of these, iNPH is the only treatable dementia; symptoms can remit following a shunt procedure. iNPH is often comorbid with PD, PSP, or AD, but early diagnosis is difficult based on neurological symptoms and neuroimaging. Furthermore, even if neurological symptoms show improvement following a shunt procedure, a gradual decline in neurological symptoms has been observed with clear onset of co-morbid neurodegenerative diseases after several years [16]. As such, the development of CSF biomarkers to identify iNPH and other neurodegenerative diseases is an extremely important issue.
To examine the function of LRG in the brain in the context of aging and neurodegenerative diseases, we first measured human hLRG concentration in the CSF and observed hLRG expression in the cerebral cortex of autopsied brains. We then generated conditional Tg mice that overexpressed mouse mLRG in the brain to examine the effects of mLRG accumulation in the cerebral cortex. Finally, to examine proteinprotein interactions, we used a protein microarray method to screen proteins with a high affinity for hLRG.

Subjects and study design
This study was conducted over a period of 6 years between January 2006 and September 2012 on consecutive, unselected admissions to our program for 100 patients with clinically suspected iNPH (63 men, 37 women; mean age 75.8 ± 6.7 years) [17,18]. In addition, 26 elderly individuals (13 men, 13 women; mean age, 68.3 ± 9.8 years) without subjective cognitive impairment or known brain disease, with MMSE scores >25 and who consented to lumbar puncture at the Juntendo University Hospital were included as non-demented NCs. We also included 43 patients (24 men, 19 women; mean age, 73.9 ± 9.7 years) with probable AD in this study. The diagnostic criteria of the National Institute of Neurological Disease and Communicative Disorders and the Stroke/ Alzheimer's Disease and Related Disorders Association (NINCDS/ADRDA) were used to identify AD [19]. For PSP (possible PSP, n = 18), diagnoses were made according to the NINDS-SPSP criteria [20] for PSP. Among the iNPH, AD, and PSP patients, dementia severity was evaluated with the MMSE.

Ethics statement
All the patients (or their next of kin) provided written informed consent to participate in the study, including CSF protein measurements. Approval for this research was granted by the ethics committee of Juntendo University, Japan. All experiments were carried out with the approval of the institutional animal care committee of Juntendo University.

CSF samples and biomarker assay
Lumbar puncture was performed in the L3-L4 or L4-L5 interspace. A 10-30 ml CSF sample was collected and gently mixed to avoid gradient effects. CSF samples with >500 erythrocytes/μl were excluded. All CSF samples were centrifuged to remove cells and debris, aliquoted, and stored in polypropylene tubes at -80°C until biochemical analysis. The CSF biomarkers total tau (Tau) and tau phosphorylated at threonine 181 (p-tau) were determined with commercially available ELISAs (Innotest hTau-Ag and Innotest Phosphotau [181P], Innogenetics, Ghent, Belgium). hLRG and Abeta42 levels were also measured using commercially available ELISAs (Immuno Biological Laboratories [IBL], Gunma, Japan). All assays were performed according to the manufacturers' protocols.

Human LRG antibody for immunostaining
We produced a custom anti-human hLRG (329) rabbit immunoglobulin (Ig) G Affinity Purify (IBL) custom antibody for immunostaining of the brain.
Transfected COS cells, CHO-K1 cells, and HEK293 cells were fixed in 10% formalin/PBS for 10 min at 4°C, followed by acetone/methanol (1:1, v/v) for 10 min at 4°C. For blocking, cells were incubated with 5% normal goat serum/BSA/PBS for 20 min at 4°C. Anti-hLRG (329) rabbit IgG (5 µg/ml) was applied for 120 min at 4°C for the primary antibody reaction, and Alexa Fluor 488 goat anti-rabbit IgG (H+L) was used for the secondary antibody reaction.

Immunohistochemistry
Paraffin sections from each area of the autopsied 19 human and 31 mouse brains were immunostained with anti-hLRG(329)

Generation of CAG-EGFP/mLRG LoxP Tg mice
To examine whether LRG overexpression would induce neurodegeneration, B6D2F1 mice (C57BL/6N×DBA/2N; Charles River Laboratories Japan, Kanagawa, Japan) were used as oocyte and sperm donors for intracytoplasmic sperm injection-mediated transgenesis. Mice were fed ad libitum with a standard diet and maintained in an air-conditioned and lightcontrolled animal room (23 ± 1°C, 55 ± 5%, 12-h light/12-h dark). All animal experiments were performed in accordance with the guidelines of the Laboratory Animal Experimentation at Juntendo University School of Medicine.
Mouse LRG (mLRG) cDNA (BC030733, Invitrogen) was subcloned into a LoxP-GFP-pA-LoxP, which was excised with KpnI and NotI. This KpnI/NotI restriction fragment was inserted into the EcoRI site of the third exon of the rabbit-globin gene in the expression plasmid pCAGGS, which contains cytomegalovirus enhancer, chicken-actin promoter, β-actin intron, and rabbit -globin poly-A signal [21]. The resulting construct (designated CAG-LoxP-GFP-pA-LoxP-mLRG, 4.72 kbp), isolated from the plasmid backbone by digestion with ScaI/BamHI, was dissolved in 10 mM Tris-HCl (pH 7.6)/0.1 mM EDTA and then stored at 4°C until use.

In situ hybridization
Mouse LRG probes were designed and applied according to the QuantiGene ViewRNA protocol by using branched DNA technology (Panomics, Fremont, CA, USA). Five-micrometerthick paraffin-embedded tissue sections were fixed, permeabilized with protease, and hybridized with oligonucleotides conjugated to alkaline phosphatase. After incubation with FastRed substrate, slides were counterstained with hematoxylin and mounted with coverslips using Permount (Fisher Scientific, Pittsburgh, PA, USA). Images were acquired using a Zeiss AX10 Scope AI, ProgRes digital microscope camera and Mac Capture Pro 2.6.0 software (Jenopik, Jena, Germany).

Western blot analysis of mLRG expression in mouse brain
Levels of mLRG in the cerebral cortex of Tg and WT mice were determined with immunoblot analyses. The cerebral cortex was dissected from mice after they were lethally anesthetized in accordance with protocols approved by Laboratory Animal Research of Juntendo University. Cerebral tissues were sonicated in lysis buffer (150 mM NaCl, 1mM EDTA, 0.5% Triton X-100, 1% SDS, 10 mM Tris [pH 7.4]) containing 1 mM PMSF, 10 mM NaF, and 1% (v/v) protease and phosphatase inhibitor cocktails (Sigma-Aldrich). The samples were clarified through centrifugation at 20,000× g, and protein concentrations were determined using a BCA kit (Pierce/Thermo Scientific, Waltham, MA, USA). Equal amounts of protein were run on 10% Criterion-XT gels (Bio-Rad Laboratories, Hercules, CA, USA) and transferred to nitrocellulose membranes (GE Healthcare). Membranes were probed with mouse anti-LRG antibody followed by fluorophoreconjugated (anti-mouse-Cy3 and anti-rabbit-Cy5) goat secondary antibodies (GE Healthcare). The signals were acquired with a Typhoon fluorescent scanner (GE Healthcare), and the band intensities were quantified using ImageQuant software (GE Healthcare).

Production of DYKDDDDK-human LRG
The full-length cDNA of the human LRG coding region was inserted into the pcDNA3.1/Hygro (-) vector (Invitrogen), adding a DYKDDDDK-tag to its C-terminus and transfected into CHO-K1 cells.
The culture supernatant of stable transfectant was collected and assessed with affinity column chromatography using ANTI-DYKDDDDK M2 Affinity Gel (Sigma-Aldrich). The concentration of the purified protein was measured by Human LRG Assay Kit-IBL (IBL).

Protein-protein interactions using protein arrays
DYKDDDDK-human LRG was assayed at concentrations of 5 and 50 ng/µl on ProtoArray Human Protein Microarrays v5.0 using the manufacturer's protocol to identify interactions with human proteins present on the ProtoArray, which contains over 9,000 different human proteins. Interacting proteins were identified using mouse anti-DYKDDDDK antibody followed by Alexa Fluor 647-anti-mouse IgG. Arrays were then washed and scanned with a fluorescence microarray scanner.

Identification of protein interactors for DYKDDDDKhuman LRG
All possible protein interactors were evaluated by their signal characteristics within the array and compared to the negative control assay. A protein was defined as exhibiting a significant interaction if it met the following conditions: the Z-Factor, or signal-to-noise ratio, was >0.5, indicating a signal greater than 2-fold above the noise. The Z-Score, or normalized fluorescent signal, was >2.0 standard deviations above the mean human protein signal on the corresponding array. The Signal Used value was at least 3-fold higher than the corresponding Signal Used value in the negative control assay. The protein had a replicate spot coefficient value <50% on the corresponding array.

Statistical analysis
All analyses were performed with PASW Statistics software, version 18.0 (SPSS, Inc., Chicago, IL). To assess differences between two groups, Mann-Whitney U tests were used, and one-way analysis of variance (ANOVA) followed by Dunnett T3 post hoc tests were used for multiple comparisons. Correlations were evaluated using linear regression analysis (Pearson's correlation), as well as Spearman's nonparametric correlation to control for potential contributions secondary to outliers.
A receiver operating characteristic (ROC) curve was used to calculate the relationship between sensitivity and specificity for the disease group vs. healthy or disease controls, and to evaluate the diagnostic performance of the analytes. The optimum cutoff value from the ROC curve was determined by anchoring the sensitivity to 90% to 95%, except in a few cases where sensitivity was below 90% but the sum of sensitivity and specificity was maximal. Values of p < 0.05 were regarded as significant, and values of p < 0.001 were used for all our major findings to minimize errors related to multiple comparisons.
Diagnostic utility of individual biomarkers. The results are shown in Table 2.
For NC vs. PDD/DLB, the pattern of analytes was significantly different between NC and PDD/DLB. Receiveroperating characteristic (ROC) statistics showed that hLRG was superior to the other biomarkers in terms of area under the curve (AUC) ( Figure 1E). hLRG concentration was higher on average in the PDD/DLB groups than in the NC group with an excellent diagnostic sensitivity (92.9%) and specificity (96.4%).
For NC vs. PSP, the analyte pattern also differed between NC and PSP; ROC statistics demonstrated that hLRG was superior to any of these biomarkers alone in terms of AUC ( Figure 1F). hLRG concentration was higher on average in the PSP groups than in the NC group with an excellent diagnostic sensitivity (88.9%) and specificity (92.9%). For NC vs. iNPH, high sensitivity and specificity were achieved with hLRG (sensitivity, 76.1%; specificity, 90.9%) ( Figure 1G).
For iNPH vs. PDD/DLB, when comparing NC and PDD/DLB, high sensitivity and specificity could be achieved with hLRG (sensitivity, 85.7%; specificity, 79.8%) ( Figure 1H). For iNPH vs. PSP, as mentioned above, it is often difficult to clinically differentiate between iNPH and Parkinson's disorders, particularly at the early disease stages, and there is no established laboratory test or biomarker that can assist in differential diagnosis in a regular hospital setting. With CSF hLRG, iNPH patients were differentiated with a sensitivity of 68.4%, and specificity of 79.8% ( Figure 1I).  The analyte patterns differed among NC, iNPH, PDD/DLB, and PSP; the ROC statistics showed that LRG was superior to any of the assessed biomarkers alone in terms of AUC.

Immunostaining of autopsied brains
Western blotting of hLRG-transfected COS cell lysates showed strong expression of a 38-kDa band corresponding to hLRG (Figure 2A). Immunoreactivity in hLRG-transfected cells was confirmed with an anti-hLRG(329) rabbit IgG ( Figure 2B), and this antibody was used to immunostain cerebral cortex sections of autopsied brains.
With NCs (n = 11), we confirmed hLRG immunoreactivity in some glial cells and neurons in the deep layer of the cerebral cortex. Compared to the NC, hLRG-positive glial cells increased with iNPH (n = 2). Furthermore, with PDD (n = 2) and PSP (n = 2), there was a significant increase in the number of glial cells positive for hLRG, and we confirmed immunoreactivity in the protruding glial cell foot processes. We also confirmed hLRG immunoreactivity in neuronal cytoplasm. With AD (n = 2), we confirmed strong hLRG immunoreactivity in neurons rather than in glial cells ( Figure 2C). We quantified the number of cells demonstrating a positive immunoreactivity and observed a difference between NC and iNPH with regard to the number of LRG-positive cells (p < 0.05). We also noted a significant increase in the number of LRG-positive cells in PDD and PSP as compared to that in NC (p < 0.001) and iNPH (p < 0.001) ( Figure 2D). Importantly, hLRG immunoreactivity in post-mortem cerebral cortex corresponded to the hLRG concentration measured in the CSF.

Age-related changes in mouse cerebral cortex
Western blot analysis demonstrated that mLRG protein levels in 48-week-old mice (n = 5) were higher than those in 4-(n = 5) and 8-week-old mice (n = 5) ( Figure 3A,B). Comparison of mLRG immunostaining of mouse cerebral cortex at 4, 8, 48, and 108 (n = 3) weeks of age confirmed only weak mLRG immunoreactivity in the neurons and glial cells of 4-and 8week-old mice, while significant mLRG immunoreactivity was observed in the neurons and glial cells of 48-and 108-week-old mice ( Figure 3C, I). In addition, mLRG in situ hybridization demonstrated that the cerebral cortex of 48-week-old mice displayed slightly increased mLRG mRNA expression compared to that of 8-week-old mice ( Figure 3J). Thus, similar to human brain, mLRG expression was observed in the neurons and glial cells of mouse cerebral cortex and increased with age.

Conditional Tg mice
We generated Tg mice that overexpressed mLRG in the neurons and glial cells of the brain. In 8-week-old Tg mice, macroscopic comparison with wild type (WT) showed that although the brain was smaller, there were no noticeable morphological changes in the cerebellum or brain stem ( Figure  3D). Hematoxylin and eosin (HE) and Kluver-Barrera (KB) staining of coronal sections did not reveal any morphological abnormalities in the Tg brain ( Figure 3E). In 48-week-old mice, we noted more severe thinning of the cerebral cortex, fewer granule cells, and moderate ventricular dilatation ( Figure 3F). Western blot analysis confirmed that compared to WT, mLRG proteins were significantly higher in Tg mice ( Figure 3G,H). Histologically, the cerebral cortex of 8-(n = 5) and 48-week-old Tg mice (n = 3) presented much more significant mLRG immunoreactivity and higher mRNA levels in the cytoplasm of neurons and glial cells than 8-(n = 5) and 48-week-old WT mice (n = 5) ( Figure 3C,I, J).
We conducted immunostaining of cerebral cortex sections from 4-(WT, n = 5; Tg, n = 5), 8-(WT, n = 5; Tg, n = 5), and 48week-old mice (WT, n = 5; Tg, n = 3) using NeuN antibody, which specifically stains neuronal nuclei, and then measured the number of cells demonstrating positive immunoreactivity. We did not observe a difference between Tg and WT mice regarding the number of NeuN-positive cells in 4-week-old mice, but there was a significant decline in the number of NeuN-positive cells in 8-and 48-week-old Tg mice compared to that in WT ( Figure 4A,B).     Immunostaining for p-tau showed that the cerebral cortex of 8-week-old Tg mice demonstrated more positive neurons and glial cells than that of WT mice ( Figure 4C,D). Furthermore, NF-L immunostaining demonstrated that 8-week-old Tg mice had large, winding dendritic neurons, which suggested that neurodegeneration had occurred ( Figure 4E). In 8-week-old Tg mice, we confirmed neuronal decline and neurodegeneration seen in very old WT mice.

hLRG protein interactors
Finally, we carried out a protein array to conduct a comprehensive screening of proteins bound with hLRG proteins. Analysis of the data for the proteins being studied relative to the negative control identified four candidate interactors for DYKDDDDK-hLRG at the 5 ng/µl concentration and 17 candidate interactors for DYKDDDDK-hLRG at the 50 ng/µl concentration (Table 3). Concentration-dependent signals were also observed for most candidate interactors identified at the 5 ng/µl concentration. The results include proteins identified using the statistical thresholds as described in the methods. We identified 17 proteins with an affinity factor of 3-fold or more. Of these, the affinities of presynaptic cytomatrix protein (PCLO), xin actin-binding repeat containing 2 (CMYA3), cortactin (CTTN), and peroxisomal trans-2-enoyll-CoA reductase (PECR) were dose dependent.
The data presented include the background subtracted pixel intensity value (the "Signal Used" value) and the ratio of Signal Used/Negative control Signal Used, which was used to rank the proteins.
We identified four candidate interactors for Flag-hLRG at the 5 ng/µl concentration and seventeen candidate interactors for Flag-hLRG at the 50 ng/µl concentrations. Concentrationdependent signals were observed for most candidate interactors identified at the 5 ng/µl concentration. The results include proteins identified using the statistical thresholds as described in the methods.

Discussion
In the initial proteome analysis of CSF, we focused on LRG as a protein that is known to increase during iNPH [23,24]. Based on our results regarding LRG concentration in the CSF and immunoreactivity in the cerebral cortex, it became clear that LRG expression increases as a function of age [24]. Furthermore, as LRG is induced by inflammation, the observed increase with aging may actually be due to accumulated inflammation. We found that CSF hLRG concentrations were negatively correlated with the MMSE score in the NC and iNPH groups. This indicates that patients with dementia had higher hLRG concentrations than the non-demented patients.
As the CSF from PDD/DLB and PSP patients demonstrated higher LRG concentrations than the CSF from NC, iNPH, and AD patients, LRG may be an effective biomarker for identifying these diseases. In particular, it has been reported that PD, PSP, and AD demonstrate frequent comorbidity with iNPH; neurodegenerative disease is suspected in cases of iNPH with extremely high CSF concentrations of LRG [16].
In examining LRG levels in mouse cerebral cortex, it was clear that it increased with aging, similar to that in humans. To elucidate tLRG function, we constructed a Tg mouse that overexpressed mLRG in glial cells and neurons. Compared to WT, the brains of 8-week-old Tg mice were small but did not show any morphological abnormalities of the cerebellum or brain stem. However, in 48-week-old Tg mice, the cerebral cortex was thinner and the number of neurons in the cerebrum was significantly decreased.
LRG's function may be related to its LRR structure. Leucinerich repeat kinase 2 (LRRK2) possesses the same LRR structure as LRG. At present, mutations in LRRK2 represent the most common cause of late-onset familial PD and may be central to pathogenetic pathways for PD [25][26][27][28][29][30][31][32][33]. Although the functions of many of the members of the LRR-containing superfamily are known, the functions of LRG and LRRK2 have not been elucidated. Recently, both LRRK2 and LRG have been linked to Crohn's disease (CD) [34,35]. Microglial activation is commonly observed in postmortem PD brain tissue, suggesting that inflammation may play a role in disease pathogenesis [36], although it is unclear whether this inflammation is causative or a secondary effect of earlier pathological events. Inflammation appears to be a common theme in PD and CD. Indeed, recent literature describes the roles for LRG and LRRK2 in immune response pathways [37][38][39][40][41][42][43]. Leucine-rich repeat transmembrane neuronal proteins (LRRTMs) also possess the same LRR structure as LRG. LRRTMs were recently found to guide presynaptic and mediate postsynaptic glutamatergic differentiation [44]. . Siddiqui [45] reported that LRRTMs bind neurexins with a differential code to cooperate in glutamate synapse development.
LRG function is thought to be mediated via binding with other proteins such as cytochrome c that restricts the functions of these proteins [14].
The protein microarray results indicated a strong affinity with Piccolo (PCLO), which is a novel component of the presynaptic cytoskeletal matrix (PCM) that is assembled at the active zone of neurotransmitter release. In addition, LRG demonstrates a strong affinity with cortactin, which exists in neuronal dendrites [46,47]. We speculated that strong binding of LRG with these proteins impedes synaptic function, reduces cognitive function, and is related to neuronal loss. Whether this observation is of potential physiological significance warrants further investigation.

Conclusion
Our results show that in both humans and mouse brains, LRG increases with age. In humans, cognitive function declines with greater CSF LRG concentration. High CSF LRG concentrations were observed in both PDD/DLB and PSP, and this would be particularly useful in identifying patients with iNPH. Increased LRG in neurons and glial cells was associated with mouse cerebral cortex neurodegeneration. Due to the LRR structure of LRG, there is a possibility that it binds with proteins involved in synaptic function. However, these results needed to be validated in a different, and hopefully even larger, cohort of patients, preferably from an independent group.