The Association between C9orf72 Repeats and Risk of Alzheimer's Disease and Amyotrophic Lateral Sclerosis: A Meta-Analysis

C9orf72 is the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) in Caucasian populations. However, the relationship between C9orf72 repeats and Alzheimer's disease (AD) was not clear. Additionally, there were few articles assessing C9orf72 in other ethnicities with ALS. In this meta-analysis, we aimed to investigate the relationship between C9orf72 repeat expansions (≥30 repeats) and intermediate repeat copies (20–29 repeats) and AD or ALS. The results suggested positive correlations between C9orf72 repeat expansions and the risk of Alzheimer's disease (OR = 6.36, 95% CI = 3.13–12.92, and p < 0.00001), while intermediate repeat copies of C9orf72 gene were not associated with the risk of the disease. C9orf72 repeat expansions were positively correlated with the risk of familial and sporadic ALS (OR = 293.25, 95% CI = 148.17–580.38, and p < 0.00001; OR = 35.57, 95% CI = 19.61–64.51, and p < 0.00001). There was a positive correlation between the gene variations and ALS risk among Caucasians and Asians (OR = 57.56, 95% CI = 36.73–90.22, and p < 0.00001; OR = 6.35, 95% CI = 1.39–29.02, and p = 0.02).


Introduction
In 2006, two researchers reported a locus on the short arm of chromosome 9 which could be a causative for FTD, ALS, and ALS/FTD [1]. In 2011, hexanucleotide (GGGGCC) repeat expansions in the noncoding region of C9orf72 were confirmed to be the most common mutation of FTD, ALS, and ALS/FTD [2]. The frequencies of the mutation account for 29%, 50%, and 88% of the patients, respectively [3].
In human genome, C9orf72 gene is located on chromosome 9p21, spanning 27,546,543-27,573,864 base pairs. It encodes 11 exons and GGGGCC (G4C2) exists between noncoding exons 1a and 1b [4]. Some researchers suggested that G4C2 ≥ 30 repeats were pathological repeat expansions, while <20 repeat units were normal [5]. In neurodegenerative diseases such as ALS-FTD, the G4C2 repeat copies of C9orf72 could reach 700-1600 units [4]. In addition, some researchers demonstrated that G4C2 intermediate copies (20-29 repeats) could also contribute to the risk of neurodegenerative diseases [6]. The disease mechanisms of how C9orf72 expanded repeats lead to these diseases are still unknown. The possible hypotheses may be the loss of function of C9orf72 protein, the accumulation of toxic RNA foci, and the Repeat Associated Non-ATG Initiated Translation (RAN-Translation) [7][8][9][10].
Substantial clinical and pathological characteristics overlap among the common neurodegenerative diseases, FTD (frontotemporal dementia), ALS (amyotrophic lateral sclerosis), and AD (Alzheimer's disease). For example, ALS patients with C9orf72 repeat expansions can present with dementia which were common in FTD and AD patients [11,12]. Tau positive pathology which is typical for AD can be found in FTD patients [13]. Patients presenting with behavior symptoms may have pathological features of AD. However, the relationship between C9orf72 and the risk of AD remains controversial. Kohli [5,15].
C9orf72 repeat expansions vary strongly between different geographic regions. In countries such as Italy, United States, and Germany, the prevalence can be as high as 47% for familial ALS and 21% for sporadic ALS [2,5,16]. However, in some Asian countries such as China and Korea, the mutation cannot be found in ALS patients [17]. Although C9orf72 repeat expansions were considered pathological mutations of European ALS patients recently, the mutations were rare in Asians.
For better understanding the widening disease spectrum of C9orf72 repeat copies, we performed a meta-analysis to clarify the association between C9orf72 variations and AD or ALS.  May 20, 2015). We used search terms as follows: ("AD," "Alzheime * /$," or "Alzheimer's disease")/("ALS," "amyotrophic lateral sclerosis," "motor neuron disease") and ("C9orf72"). The languages of the articles were limited to Chinese and English. The strategies were made by two researchers (L. Shu and QY Sun). If there were any disputes, we consulted another researcher (JF Guo).

Selection of Studies.
Studies were selected when they met the following criteria: (1) original articles (when there were secondary articles, we added the useful original articles from them); (2) association studies between C9orf72 and AD or ALS; (3) sufficient data to calculate odds ratio (OR) and 95% confidence interval (95% CI); (4) study types being observational studies such as case-control studies and cohort studies; and (5) languages being limited to Chinese and English.
Studies were excluded when the following existed: (1) replicated data (if there were overlapped samples, we chose the largest samples); (2) incomplete data (no control group); (3) study types being reviews, case series, letters, editorials, and so forth; and (4)

Quality Assessment.
The Newcastle-Ottawa Scale (NOS) was used to assess the quality of the selected original articles. The scale included three aspects: (1) subjects selection (0-4 points); (2) comparability between groups (0-2 points); and (3) exposure (0-3 points). The total points range from 0 to 9 [18]. The studies scored higher than 5 were considered of "high quality." The assessments were conducted by two reviewers (L. Shu and QY Sun). The third author was consulted when agreements cannot be reached.

Data Extraction.
Basic data were extracted for the statistical analysis including first author, publication year, ethnicity, country, age at onset, ages of cases and controls, genotyping methods, number of cases and controls, number of cases and controls who carried C9orf72 repeat expansions or intermediate repeat copies, diagnostic criteria, and NOS scores. Data were selected by two researchers (L. Shu and QY Sun). The third author was involved to solve the disagreements.

Statistical
Analysis. The correlation between C9orf72 and neurodegenerative diseases was analyzed by pooled OR with 95% CI. The heterogeneity among the included studies was estimated by -test or 2 statistic. If the -test showed value ≤ 0.1, the heterogeneity was considered significant. We chose fixed-effects model for statistical analysis with low heterogeneity while we chose random-effects model with moderate to high heterogeneity. test was conducted to measure the association between C9orf72 and neurodegenerative diseases.
value < 0.05 indicated statistically significant difference. Funnel plot was visually inspected to assess the possible publication bias [19]. Sensitivity analysis was conducted by sequentially removing one publication to evaluate the influence of single publication on the whole results. RevMan 5.2 software was used for all the statistical analyses and graphics.

Association between C9orf72 and AD
3.1.1. Eligible Studies. After searching of PubMed, EMBASE, and Cochrane databases, 299 articles were identified. 89 duplicated datasets were removed and a total of 210 articles were reviewed by title and abstract. 190 articles did not meet the inclusion criteria and only 20 of them were reviewed for full-test assessment. Ten articles were excluded: 5 conference abstracts, 1 not peripheral blood test, and 4 without controls. Thus, there were 10 articles left for the final statistical analysis [13][14][15][16][20][21][22][23][24][25][26] (the detailed flowchart was shown in Supplementary Figure 1). Table 2.  Table 2).

Cumulative Analysis.
The results of the meta-analysis were present in Figures 1(a) and 1(b). There was no significant heterogeneity; thus a fixed-effects model was chosen for the meta-analysis. The results indicated that C9orf72 repeat expansions were related to the risk of AD (OR = 6.36, 95% CI = 3.13-12.92, and < 0.00001, Figure 1   of AD (OR = 1.04, 95% CI = 0.32-3.43, and = 0.94, Figure 1(b)).

Sensitivity Analysis
We performed sensitivity analyses by omitting single articles to test the stability of the results. After sequentially omitting single studies about ALS and AD, the total results were similar.

Publication Bias
There were no asymmetries of the funnel plot below and no significant publication biases of the meta-analysis about C9orf72 repeat expansions and AD or ALS (Supplementary Figure 3). We failed to do publication bias analysis on the association between C9orf72 intermediate copies and the risk of AD because of the scarcities of the included articles.

Discussion
There were clinical, pathological, and hereditary overlap between FTD and AD; therefore some researchers regarded them as the diseases of the same disease spectrum [46]. After discovering the most common genetic cause of FTD-C9orf72 repeat expansions, many researchers began to focus on the association between C9orf72 and AD. However, the disputes have not been solved yet. The results of our metaanalyses demonstrated that C9orf72 repeat expansions were positively correlated with the risk of AD while C9orf72 intermediate repeat copies were not related to the risk of AD. However, the limited number of the original articles about the association between C9orf72 intermediate repeat copies and AD influenced the validity of the analysis. Additionally, the diagnostic standards of the patients in the original articles were based on the clinical diagnosis.
C9orf72 has been discovered as the most common causative gene for ALS in white populations, which accounts for 40% of familial cases and 20% of sporadic cases in Finland [47]. While there were a large number of studies focusing on the identification of the gene in European countries, there were few studies reporting C9orf72 mutation in Asian ALS patients. Our meta-analysis selected high-quality studies and demonstrated that C9orf72 repeat expansions were related to the risk of ALS in Asians and Caucasians. We also proved that C9orf72 repeat expansions were not only correlated with the risk of familial ALS but also related to the risk of sporadic ALS. However, there were evidences that anticipation played a role in C9orf72 families. Gijselinck et al. demonstrated that there was a decreasing onset of age in younger generation in ALS families carrying longer G4C2 expansion. The original researches on the association between C9orf72 repeat expansions and ALS can be invalid because of the genetic anticipation and the wide range of age of onset of C9orf72 diseases [48]. Therefore, the different age of onset can cause bias in the meta-analysis of familial ALS. Additionally, there were few case-control studies about the relationship between C9orf72 repeat expansions or intermediate repeat copies and ALS in other populations. Further large sample studies are essential to clarify the association.
Despite the limitations stated above, there were some other limitations. First, the C9orf72 repeat expansions were analyzed by conventional PCR-based methods. When C9orf72 repeat copies were more than 60 copies, the accurate size cannot be gotten by conventional PCR-based methods. Other measures such as Southern blot were needed to detect the true repeat size [3]. Second, although our meta-analysis incorporated all the published case-control cohorts, some negative unpublished results were possibly neglected.
In conclusion, C9orf72 repeat expansions were risk factors for AD while C9orf72 intermediate repeat copies were not associated with the risk of AD. C9orf72 repeat expansions were correlated with the risk of ALS. However, inevitable limitations existed in our meta-analysis such as the limited number of original articles and possible publication biases. Further assessments were needed with enough case-control samples.