FBXO7 Y52C Polymorphism as a Potential Protective Factor in Parkinson's Disease

Mutations in the F-box only protein 7 gene (FBXO7), the substrate-specifying subunit of SCF E3 ubiquitin ligase complex, cause Parkinson's disease (PD)-15 (PARK15). To identify new variants, we sequenced FBXO7 cDNA in 80 Taiwanese early onset PD patients (age at onset ≤50) and only two known variants, Y52C (c.155A>G) and M115I (c.345G>A), were found. To assess the association of Y52C and M115I with the risk of PD, we conducted a case–control study in a cohort of PD and ethnically matched controls. There was a nominal difference in the Y52C G allele frequency between PD and controls (p = 0.045). After combining data from China [1], significant difference in the Y52C G allele frequency between PD and controls (p = 0.012) and significant association of G allele with decreased PD risk (p = 0.017) can be demonstrated. Upon expressing EGFP-tagged Cys52 FBXO7 in cells, a significantly reduced rate of FBXO7 protein decay was observed when compared with cells expressing Tyr52 FBXO7. In silico modeling of Cys52 exhibited a more stable feature than Tyr52. In cells expressing Cys52 FBXO7, the level of TNF receptor-associated factor 2 (TRAF2) was significantly reduced. Moreover, Cys52 FBXO7 showed stronger interaction with TRAF2 and promoted TRAF2 ubiquitination, which may be responsible for the reduced TRAF2 expression in Cys52 cells. After induced differentiation, SH-SY5Y cells expressing Cys52 FBXO7 displayed increased neuronal outgrowth. We therefore hypothesize that Cys52 variant of FBXO7 may contribute to reduced PD susceptibility in Chinese.

Recently, the F-box only protein 7 (FBXO7) mutations have been identified in several families with early-onset parkinsonism and pyramidal tract signs. Homozygous R378G missense mutation in an Iranian kindred, homozygous nonsense mutation (R498X) in an Italian family, a Pakistan family and a Turkey family, and compound heterozygous mutations (IVS7+1G/T and T22M) in a Dutch family are unambiguously responsible for the autosomal recessive, early-onset, parkinsonian-pyramidal syndrome [4,5,6]. The presynaptic nature of the parkinsonism in the families reported by Di Fonzo et al. is shown by the dramatic abnormality of DaTSCAN SPECT, the beneficial effect of levodopa, and the presence of levodopa-induced dyskinesias, suggesting that FBXO7 mutations may be the potential genetic causes of early-onset Parkinson's disease (EOPD) or familial PD [5]. Two missense substitutions, p.Ile87Thr and p.Asp328Arg, in a single heterozygous state, were found in two EOPD patients in Taiwan [7]. Although no pathogenetic mutations in the FBXO7 gene were detected in 135 Chinese early-onset parkinsonism patients, the PD patients showed a trend toward decrease in Y52C G allele frequency compared with the controls [1].
FBXO7, a member of the F-box-containing protein (FBP) family, encodes a protein of 522 amino acids consisting of several discrete domains: the ubiquitin-like domain, cyclin-dependent protein kinase 6 (CDK6) binding site, FBXO7/PI31 domain, Fbox motif, proline-rich region, and R(ar)DP motif [5,8]. Through the interaction between the F-box and the Skp1 protein, FBPs become part of SCF (Skp1-Cullin1-F-box protein) ubiquitin ligase complexes, and play roles in ubiquitin-mediated proteasomal degradation (review in [9]). F-box proteins recruit a large number of diverse substrates to SCF complexes and allow for their ubiquitination [10]. FBXO7 promotes ubiquitin conjugation to TRAF2 (a member of the tumor necrosis factor receptor associated factor protein family with ubiquitin ligase activity) and cIAP1 (an apoptosis inhibitor possessing ubiquitin ligase activity), resulting in decreased receptor-interacting protein 1 (RIP1) ubiquitination and lowered NF-kB signaling activity [11,12]. Strong evidence has shown that NF-kB induced neuroinflammation may be involved in development of PD [13,14]. Therefore, FBXO7 may play a role in protecting neurons from PD process. In contrast, the mutations or variations that change the function, expression, or stability of FBXO7 may confer PD risk to the subjects.

Mutation/variant analysis of FBXO7
Since most of the mutations found in FBXO7 result in truncated FBXO7 protein, amino acid replacement or multiple aberrant frame-shift splice variants, we sequenced FBXO7 cDNA (Table 1) instead of genomic DNA in a cohort of ethnic Chinese patients with EOPD in Taiwan to identify previously undetected variants in EOPD cases of Chinese origin, followed by a case-control study for the identified variants. The cDNA samples contained only DNA sequences from genes that were transcribed into RNA. Thus FBXO7 cDNA fragments from 80 EOPD patients were amplified for sequence analysis. However, only two known substitutions that caused changes in the peptide sequence were identified: a c.155A.G substitution leading to an amino acid change from tyrosine to cysteine in position 52 (Y52C) in one EOPD patient (heterozygote) and a c.345G.A substitution resulting in a methionine to isoleucine change at amino acid position 115 (M115I) (rs11107) in 74 EOPD patients (40 homozygotes and 34 heterozygotes) (amino acid number according to NM_012179) (Fig. 1A). The two reported [1] variants were confirmed using PCR-restriction fragment length polymorphism (RFLP) method (Fig. 1B). Both Y52C and M115I are not evolutionarily conserved in the known mammalian homologues of the FBXO7 protein (Fig. 1C).

Case-control study of Y52C and M115I
A case-control study in a cohort of PD patients (n = 516, 80 EOPD patients included) and ethnically matched controls (n = 516) was conducted to assess the association of Y52C and M115I with the risk of PD (Table 2). All genotype frequencies were confirmed to be in the Hardy-Weinberg equilibrium. There was no statistically significant difference (p.0.025) in genotype or allele distribution between patients and controls for both single nucleotide polymorphisms (SNPs) examined, after correction of multiple SNP testing. However, for Y52C, the frequency of AG genotype (0.8% vs. 2.3%, p = 0.044) or G allele (0.4% vs. 1.2%, p = 0.045) was notably lower in PD patients than the controls. Y52C AG genotype or G allele demonstrated a trend toward decrease in risk of developing PD (odds ratio: 0.33, 95% confidence interval: 0.09-0.95, p = 0.055-0.056). Analysis combining our patient and control subjects as well as the population in Luo's study [1] yielded results of statistically significant difference in genotype (0.9% vs. 2.8%, p = 0.012) and allele (0.5% vs. 1.4%, p = 0.012) distribution between patients and controls. The negative association of the Y52C AG genotype or G allele with PD was significant (odds ratio: 0.32-0.33, 95% confidence interval: 0.12-0.77, p = 0.016-0.017). The identified one EOPD and three late-onset PD patients carrying Y52C presented with asymmetrical tremor, rigidity, and bradykinesia without pyramidal signs, all of whom had a good response to anti-parkinsonian medication.

FBXO7 expression analysis
Since the case-control study suggests that FBXO7 Y52C G allele might be a potential protective factor, we cloned the polymorphic FBXO7 cDNA, which was then expressed in HEK-293T cells to investigate the functional consequences.
The common cellular abnormality found in the PARK15 patients from the Dutch and Italian families is the depletion of the FBXO7 isoform 1 (NM_012179), which normally is located in the cell nucleus [15]. Forty-eight hours after transfection of EGFP tagged FBXO7 isoform 1 constructs, cells were analyzed by fluorescent microscopy. Although Cys52 FBXO7 protein displayed nuclear and cytosolic staining pattern similar to Tyr52, a significantly stronger green (FBXO7 fusion protein) relative to blue (nuclei staining) fluorescence signal was observed in Cys52 FBXO7 cells (1.60 vs. 2.57, p = 0.015; Fig. 2A). To further examine the transiently expressed FBXO7-EGFP fusion protein, protein blotted with FBXO7 antibody was performed. As shown in Fig. 2B, FBXO7-EGFP fusion protein in the expected size range for Tyr52 and Cys52 constructs was observed. However, the protein expression level of Cys52 FBXO7 was increased compared with the Tyr52 FBXO7 (211%, p = 0.016). The stability of Cys52 variant was further examined by a cycloheximide (200 mg/ml) chase experiment. While the Tyr52 protein was degraded to 68%, 18%, 11%, 9% and 7% left after 6, 12, 24, 36, and 48 hr of protein synthesis blocking, reduced rates of decay were observed for Cys52 variant (90%, 78%, 72%, 52%, and 21% remained, respectively) (Fig. 2C).

Homology modeling of Cys52 FBXO7
To understand the structure-based information of Cys52 variant in FBXO7, homology modeling of Tyr52 and Cys52 FBXO7 was performed. After energy minimization, the modeled structures for Tyr52 (WT) and Cys52 (Y52C) were shown in Fig. 3. The potential energy of Tyr52 and Cys52 mutant was 22463.854 and 22471.736 kcal/mol, indicating Cys52 FBXO7 exhibited a more stable feature than Tyr52 FBXO7. According to hydrogen-bond (H-bond) computing analysis, the H-bond interaction of Tyr and Cys52 was shown. In the Tyr52 model, Tyr52 did not form any H-bonds with adjacent residue. On the other hand, H-bond formed by the Cys52 with the Asp54 causes decrease of the local energy.

Discussion
Up to now, only four different types of FBXO7 mutations (T22M, R378G, R498X, IVS7+1G/T) have been reported to be responsible for parkinsonian-pyramidal disease, which has been designated as the cause of PARK15 [4,5,6]. Recently, two missense substitutions (p.Ile87Thr and p.Asp328Arg) were found in two EOPD patients in Taiwan [7]. We did not detect any mutation in our EOPD patients, which is compatible with the previous results of studies showing rare pathogenic mutations of the FBXO7 gene in typical PD patients of Chinese ethnicity [1,7]. Nevertheless, our method using cDNA sequencing may miss some mutations in the non-coding and regulatory elements of the FBXO7 gene. Also variants which result in a truncated protein or nonsense mediated decay would not be identified.
Although the role of FBXO7 in neurons is still not known, because it is a part of SCF ubiquitin ligase complex, its function in the ubiquitin-mediated protein degradation is implicated [9]. Impaired ubiquitin-mediated protein degradation has been found in sporadic PD [17], autosomal dominant PARK1 [18], and autosomal recessive PARK2 [19]. Therefore, it is postulated that FBXO7 mutations may compromise the ubiquitin-proteasome function and cause neuronal dysfunction in PD. In this study, we showed that the Y52C AG genotype or G allele of FBXO7 conferred a reduced susceptibility to Chinese PD when the data from our study and Luo's [1] were combined. However, as Y52C G allele is rare and the difference in frequency is small (0.4% in PD and 1.2% in controls), the genetic evidence is very limited and the findings may be due to chance alone. While FBXO7 mutations typically causes autosomal recessive parkinsonism with pyramidal tract signs, the Y52C AG genotype or G allele in this study appears to provide a protective effect in a dominant mode. T22M, R378G, R498X mutations resulted in decreased stability of FBXO7 protein and T22M caused loss of its nuclear activity, both of which may jeopardize the neuronal function [15]. Using the program SWISS-MODEL, we showed that the Cys52 formed H-bond with the Asp54 causing decreased local energy, which may increase the stability of the protein. The increased stability of FBXO7 protein consequent to Cys52 was further confirmed by cycloheximide chase experiment (Fig. 2). These results suggest that Cys52 may play a protective role in PD via increasing stability of FBXO7 protein, which is in contrast to the decreased stability and loss of function caused by T22M, R378G, and R498X [15]. Indeed, knock down of FBXO7 leading to dopaminergic neuronal dysfunction and cell loss and locomotor deficits that are improved by apomorphine, have been shown in zebrafish [20]. The study in zebrafish further suggests that FBXO7 plays an important role in the development of dopaminergic neurons and its loss of function caused by mutations may be responsible for the phenotype of PD.
FBXO7 can interact with three proteins including hepatoma upregulated protein, cIAP1, and the proteasome inhibitor protein PI31 [11,21,22]. FBXO7 was also reported to enhance activity of cyclin D/cdk6 that plays an important role in regulating neuronal death processes [23]. How these interactions may contribute to the pathogenesis of PD remains to be clarified. More emerging evidence has suggested that neuroinflammation is involved in the pathogenesis of PD through inflammatory mediators such as TNFa, nitric oxide (NO), IL-6, and IL-1b [13]. NF-kB activation is required for all of these inflammatory mediators to be produced by microglial cells [14]. Recently, few NF-kB inhibitors have been applied to the therapeutic approaches of several chronic inflammatory diseases including PD [13]. Interestingly, FBXO7 protein has been shown to be a negative regulator of NF-kB signaling pathway, through binding to and ubiquitinating TRAF2 and cIAP1, which would lead to decreased RIP1 ubiquitination and NF-kB activity [12]. To further investigate the role of FBXO7 in NF-kB signaling, Cys52 variant was expressed in HEK-293T and SH-SY5Y cells and TRAF2 (NF-kB signaling protein) was examined in this study. As shown in Fig. 4-6, structurally stable Cys52 FBXO7 facilitated the degradation of TRAF2 protein via increasing TRAF2 ubiquination, which may theoretically further lower the NF-kB activity. Although the NF-kB activity was not evaluated in this study, the results provide some evidence that Cys52 variant may play a protective role in PD pathogenesis. Whether other NF-kB signaling proteins such as RIP1 are also regulated by FBXO7 needs to be investigated in future studies.
In conclusion, gene mutations may be rare in Chinese earlyonset Parkinsonism patients. We have shown that Y52C polymorphism of FBXO7 may contribute to reduced PD susceptibility in Chinese population. However, additional casecontrol studies are needed to establish whether FBXO7 variants truly play a role in PD.

Ethics statement
This study was performed according to a protocol approved by the Institutional Review Board of Chang Gung Memorial Hospital, and all examinations were performed after obtaining written informed consents. Subjects A total of 516 unrelated Taiwanese PD subjects (45.0% females) were recruited from the neurology clinics of Chang Gung Memorial Hospital (CGMH). All patients were diagnosed by two neurologists specialized in movement disorders (Y.-R. Wu and C.-M. Chen) with probable idiopathic PD according to the published criteria [24], which includes substantial and sustained response to levodopa or a dopamine agonist. Subjects with prior history of multiple cerebrovascular events or other causes of parkinsonian symptoms (e.g. brain injury or tumor, encephalitis, antipsychotic medication) were excluded. The mean age at onset (AAO) of PD was 62.0611.5 years, ranging between 19 and 93 years. For juvenile PD patients (AAO#50), mutations in the Parkin, PINK1, DJ-1, ATP13A2 and G2019S LRRK2 were excluded ( [25,26,27] and unpublished results). A group of 516 normal controls without neurodegenerative diseases were recruited from the same ethnic community. Control subjects (50.2% females) had mean age at examination of 60.9612.3 years, ranging between 20 and 92 years. All examinations were performed after obtaining written informed consent from patients and control individuals. This study was approved by the Institutional Review Board of CGMH.

Genetic analysis
Genomic DNA was extracted from peripheral blood leucocytes using the standard protocols. For PD patients with onset #50 (n = 80, mean age at onset 43.760.7 years, 33.7% females), RNA was extracted using PAXgene Blood RNA Kit (PreAnalytiX). The RNA was DNase (Stratagene) treated, quantified, and reversetranscribed to cDNA using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Using polymerase chain reaction (PCR) with designed primers and conditions (Table 1), the 1955-bp amplified FBXO7 cDNA was gel purified and sequenced directly using the ABI PRISM 3130 Genetic Analyzer (Applied Biosystems). The identified Y52C and M115I variants were verified by genomic DNA PCR and sequencing. For population screening, the Y52C and M115I were examined using the PstI and PaeI (gain of sites) restriction enzymes, respectively ( Table 1). The digested PCR products were visualized with ethidium bromide after electrophoresis in 2.2% or 1.6% agarose gel.

FBXO7 cDNA constructs
Using the designed primers to remove translation termination codon (Table 1), the full-length FBXO7 cDNA fragments from an individual heterozygous for Cys52 were cloned into pGEM-T Easy vector (Promega) and sequenced. The 1.7 kb HindIII (added in the forward primer)-AgeI (added in the reverse primer)  fragments were removed from pGEM-T Easy vector and ligated into the corresponding sites of pEGFP-N1 (Clontech) to generate Tyr52 and Cys52 FBXO7 cDNA in-frame fused to the EGFP gene. The resulting EGFP-tagged FBXO7 constructs were used in transient expression studies for confocal microscopy examination, FBXO7 stability and anti-TRAF2 co-immunoprecipitation. Additionally, the HindIII-XhoI fragments containing FBXO7 were ligated into pcDNA3.1/V5-His (Invitrogen) to generate Tyr52 and Cys52 FBXO7 cDNA in-frame fused to the V5-His for Western blot analysis of NF-kB signaling pathway protein TRAF2.

Cell cultivation and transfection
Human embryonic kidney (HEK)-293T (ATCC No. CRL-11268) cells were cultivated in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (FBS) in a 37uC humidified incubator with a 5% CO 2 atmosphere. Cells were plated into 6-well (6610 5 /well) dishes, grown for 20 hr and transfected by the lipofection method (GibcoBRL) with EGFPtagged FBXO7 constructs (4 mg/well). The cells were grown for 48 hr for the protein studies. To evaluate the stability of FBXO7 protein, protein synthesis inhibitor cycloheximide (200 mg/ml) was added 24 hr after transfection for 0, 6, 12, 24, 36, and 48 hr before protein preparation. For immunoprecipitation studies, proteasome inhibitor MG-132 (5 mM) was added 24 hr after transfection for 24 hr before protein preparation.

Confocal microscopy examination
For visualizing intracellular FBXO7-EGFP protein, transfected cells on coverslips were stained with 49-6-diamidino-2-phenylindole (DAPI) to detect nuclei. The stained cells were examined for dual fluorescent imaging using a Leica TCS confocal laser scanning microscope.

FBXO7 SH-SY5Y cell lines generation
The SH-SY5Y-derived FIp-In host cells [28] and Flp-In T-REx System (Invitrogen) was used to generate stably induced SH-SY5Y cell lines exhibiting tetracycline-inducible expression of Tyr52 and Cys52 FBXO7. Briefly, the SH-SY5Y host cells were cotransfected with pOG44 plasmid (constitutively expressed the Flp recombinase) and pcDNA5/FRT/TO-FBXO7-EGFP plasmid according to the supplier's instructions. These cell lines were grown in medium containing 5 mg/ml blasticidin and 100 mg/ml hygromycin. Doxycycline (dox, 5 mg/ml) was added to induce EGFP-tagged FBXO7 expression for two days. The proteins were prepared for Western blotting using antibody to FBXO7 or actin as described. Neuronal phenotypes were examined after induced differentiation with retinoid acid (10 mM) and induced expression of FBXO7 for 7 to 21 days. The morphologic differentiation of Tyr52 and Cys52 SH-SY5Y cells including total outgrowth, processes, and branches was assessed by using Metamorph microscopy automation and image analysis software (Molecular Devices).

Statistical analysis
The genotype frequency data and the expected genotypic frequency under random mating were computed and Chi-square tested for Hardy-Weinberg equilibrium using standardized formula. The genotype and allele association analysis was carried out using the Chi-square test. Odds ratios with 95% confidence intervals (95% CI) were calculated to test association between genotype/allele and disease. Given the observed Y52C G allele frequency of 0.0078 (0.0095, in combined data from Taiwan and China) and a total of 1032 (1367, combined subjects from Taiwan and China) subjects in the present study, at significance level of 0.05, we had power greater than 0.8 to identify an association when the allele genetic effect size was greater than 3.1 (2.5). Given the observed M115I G allele frequency of 0.286 and 1032 subjects in the present study, at significance level of 0.05, we had power greater than 0.8 to identify an association when the allele genetic effect was greater than 1.5.
For statistical analysis of microscopy images, immunoblots and cell viability assays, data were expressed as the means 6 standard deviation (SD). Three independent experiments were performed and non-categorical variables were compared using the Student's t-test. All p-values were two-tailed, with values of p,0.05 being considered significant.

Homology modeling
We modeled the three dimensional structures of the Tyr52 and Cys52 FBXO7 proteins by comparative methods and energy minimization using the program SWISS-MODEL [29]. The 2.9-Å coordinate set for the crystal structure of human UBC protein (PDB code 2ZVO, chain A) served as the template for modeling the residue 1-79 of human FBXO7. The energy computation was done with the GROMOS96 [30] implementation of Swiss-PdbViewer. The resulting FBXO7 three-dimensional models were manipulated and rendered in PyMOL (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC).