Introduction

Amyotrophic lateral sclerosis (Lou Gehrig disease, ALS) is an age-dependent, neurodegenerative disorder of motor neurons with both sporadic and familial forms. It is a rapidly progressive disease that results in progressive paralysis until death occurs, usually within 5 years of onset. The cause of most types of ALS is unknown and the disease is untreatable. ALS arises as a dominantly inherited trait in approximately 10% of ALS cases. Approximately 25% of autosomal dominant ALS is caused by mutations in the Cu/Zn superoxide dismutase (SOD1) gene [1, 2]. A recessive form of juvenile-onset ALS has been associated with ALS in chromosome 2q33 [3, 4]. Additional chromosomal regions have been identified in chromosomes 9q [5], 15q [6], 16q [7, 8, 9], 18q [10], 20 [8], and X [11] for familial ALS. However, the underlying genetic causes of the remaining forms of ALS are still unknown.

The apolipoprotein E ( APOE) gene is a confirmed genetic risk factor that influences both risk and age at onset in Alzheimer disease [12, 13] and Parkinson disease [14, 15]. APOE functional polymorphisms have been examined for association with different phenotypes of ALS, including the risk of developing ALS, age at onset, site of onset, and duration of the disease [16, 17, 18, 19]. Most studies found no association between APOE and the risk of developing ALS [16, 17, 19]. However, for other phenotypes, no consensus has been made [17, 19, 20, 21, 22]. In particular, the studies of APOE affecting age at onset and site of onset in ALS have been the most inconsistent. Moulard et al. [19] found that bulbar-onset ALS patients with an APOE-4 allele showed an average 6 years earlier age at onset than those without APOE-4. Conversely, two later studies [17, 20] reported no significant difference in age at onset between patients with and without an APOE-4 allele in either the overall ALS dataset or the bulbar-onset ALS subset. These previous studies all used unrelated case-control samples to test allelic association between the groups of interest. No family-based association tests have been performed. In addition, most studies examining age at onset of ALS compared the average ag at onset between different APOE genotypes. This approach does not extract the full amount of information from the data.

In this study, we used family-based association approaches for quantitative and qualitative traits to assess the relationship of APOE with age at onset of ALS and the risk of developing ALS, respectively.

Materials and methods

Family data

The dataset consists of 446 singleton families with only 1 affected in each family (238 parent-child trio and 208 discordant sibpairs) and 62 multiplex families with at least 2 affected relatives (Table 1). All families have at least 1 affected individual with age at onset information available. The family ascertainment was conducted by two clinical sites, Northwestern University Medical School (396 families) and Massachussetts General Hospital (112 families). All patients diagnosed with ALS were examined by a neurologist (T.S, R.B. and colleagues) and met the El Escorial criteria [23] for the diagnosis of ALS. Clinical information, including patient’s age at onset (defined as the time at which symptoms first occurred), was obtained at the time of examination. Because ALS is characterized by sudden onset and rapid progression, age at onset is essentially equivalent to age at examination. All participants or their legal representatives gave informed consent prior to joining the study, and data were collected according to protocols approved by each contributing group’s institutional review board.

Table 1 Summary of the family data, age at onset, and APOE allele frequencies for the overall amyotrophic lateral sclerosis (ALS), early onset ALS (EOALS), and late-onset ALS (LOALS) datasets
Full size table

Genotyping

Genomic DNA was extracted from whole blood according to established protocols [24]. We genotyped APOE functional polymorphisms in a total of 1,462 samples. APOE genotyping was performed as previously described [25]. Substantial quality control procedures were followed [26]. Genotype data were uploaded into the PEDIGENE® database and merged into the LAPIS management system for data analysis [27].

Statistical analysis

All genotype data were examined for mendelian consistency, using the PedCheck program [28]. APOE functional polymorphisms were tested for Hardy-Weinberg equilibrium (HWE) in the affected group (1 affected from each family) and the unaffected group (1 unaffected from each family). An exact test implemented in Genetic Data Analysis (GDA) [29] was used to test HWE, in which 3,200 replicate samples were simulated for estimating the empirical P value.

To investigate the effect of APOE on age at onset of ALS, we treated age at onset as a quantitative trait. We used the TDT Q5 method [30], which utilizes parent-child triad data, implemented in the QTDT program [31] (http://www.sph.umich.edu/csg/abecasis/QTDT/) to test the allelic association between APOE and age at onset of ALS. The method proposed by Rabinowitz [32], which uses parents and multiple siblings data, served as a second method to confirm the results. To test whether ascertainment center has an effect on the results, we incorporated center as a covariate in both tests.

The pedigree disequilibrium test (PDT) [33] was employed to test APOE association with ALS risk. Two PDT statistics were used: the PDT-sum statistic [34], which examines allelic effects, and the genotype-PDT, which examines genotypic effects [35]. We performed the PDT analysis on the full dataset as well as stratified datasets based on an age at onset cutoff of 40 years. The late-onset subset (LOALS) consists of families with all affecteds having age at onset greater than or equal to 40 years. The early onset subset (EOALS) consists of families with at least 1 affected with age at onset less than 40 years.

Results

For each APOE genotype, the number of unaffected and affected individuals of all family members is summarized in Table 2. Since these counts were taken from family data, one should take into account familial correlation in interpretation of the data. Overall, APOE-33 is the most-frequent genotype. The allele frequencies in our overall ALS family dataset were 8.7% for APOE-2, 75.9% for APOE-3, and 15.4% for APOE-4. A similar range of APOE allele frequencies was observed in the overall ALS, EOALS, and LOALS subsets (Table 1).

Table 2 Summary of APOE genotypes for the affecteds and unaffecteds in the overall ALS, multiplex, and singleton family datasets
Full size table

Age at onset ranged from 18 to 81 years, with average age at onset (±SD) being 48.1±13.1 years in the overall ALS dataset. Although the average age at onset tends to be higher in multiplex than singleton families in all three subsets, a large standard deviation of age at onset is found in the multiplex families due to their small sample sizes (Table 1). No deviation from HWE was found in either affected or unaffected samples for APOE.

In the TDT Q5 analysis, APOE-2 showed the strongest evidence of positive association with age at onset ( P =0.001) and APOE-3 showed a marginally significant P value ( P =0.058). Similar results were obtained with the Rabinowitz method ( P =0.008 for APOE-2 and P =0.054 for APOE-3). Furthermore, the association between APOE-2 and age at onset was not changed after taking into account the ascertainment effect as a covariate in the tests. The positive association between APOE-2 and age at onset indicates that age at onset increases as the number of APOE-2 alleles transmitted from parents increases. This is also seen in the mean age at onset in each APOE-2 carrier group, in which the average age at onset was 48±13.2 years for non- APOE-2 carriers (433 individuals), 51±11.3 years for one- APOE-2 carriers (81 individuals), and 64±3.5 years for two- APOE-2 carriers (2 individuals). Overall, we found approximately a 3-year difference between APOE-2 carrier and non- APOE-2 carrier groups (51 vs. 48 years) for age at onset.

To update our analyses [17] of APOE and the risk of ALS, we examined all 508 families for both allelic and genotypic effects. No significant results were observed (global P =0.276 for alleles and global P =0.491 for genotypes). For the association between APOE and the risk of ALS in the stratified subsets, we found that APOE-2 significantly reduced the risk of ALS in the EOALS subset ( P =0.012), but not in the LOALS subset.

In order to investigate the relationship of APOE-2 with the risk and age at onset of ALS, we also examined the pattern of APOE-2 carriers in each dataset. We randomly selected 1 affected and 1 unaffected with APOE genotype available from each family. The percentages of APOE-2 carriers in affected and unaffected groups, respectively, for each subset are depicted in Fig. 1. The frequencies of APOE-2 carriers in unaffecteds are similar (15.9%–17.6%) among EOALS, LOALS, and the overall ALS datasets. For the affecteds, the EOALS subset clearly consists of far fewer APOE-2 carriers than the other two datasets (8.7% vs. 18.6% and 15.5%) as well as all unaffecteds. The differences in the numbers of APOE-2 carriers between affected and unaffected groups in the EOALS subset may lead to significant association between APOE-2 and the risk of ALS.

Fig. 1
figure 1

Percentages of APOE-2 carriers for affecteds and unaffecteds in each amyotrophic lateral scleosis (ALS) subset. Early onset ALS ( EOALS) consists of families with at least 1 affected with age at onset <40 years and late-onset ALS ( LOALS) consists of families with all affected with age at onset ≥40 years. We selected 1 affected and 1 unaffected individual from each family for each subset for the calculation of percentages of APOE-2 carriers

Full size image

Discussion

APOE has been implicated in several neurodegenerative diseases [13, 15, 36]. To our knowledge, there are at least eight publications using unrelated samples with inconsistent conclusions on the role of APOE in ALS. Some datasets are larger than the others, for instance, no effect of APOE on age at onset of ALS was found using 360 ALS cases versus 351 controls [17] in either the overall data or bulbar-onset subset, while significant results were found for age at onset of bulbar-onset patients using 130 ALS cases versus 161 controls [19]. The current study is different from the previous studies in several respects. In particular, we used the family based association approach to dissect the effect of APOE on the age at onset of ALS. This approach protects against spurious association due to population substructure [37]. In this study, age at onset is treated as a quantitative trait, increasing the amount of information extracted from the data.

Overall, the present study showed a strong association of later onset with increasing number of APOE-2 alleles. The protective effect of APOE-2 on age at onset of ALS may have been because the early onset ALS patients tended to receive less APOE-2 from parents than the late-onset ALS patients. This leads to a plausible scenario that parents transmit APOE-2 less than expected to ALS affected individuals within the early onset group, as shown in Fig. 1. This is probably the main reason that we detected negative association between APOE-2 and ALS risk in the early onset group. We, therefore, concluded that the effect of APOE-2 on age at onset is the main cause of the APOE risk effect in early onset ALS.

Previous studies have focused on the risk effect of APOE-4 in ALS [17, 20, 21]. Clearly, the strong dose effect of APOE-4 in Alzheimer disease has impacted the direction of previous research on APOE in ALS. Overall, there are few studies of the risk of APOE-2 in ALS. Moulard et al. [19] showed a difference for APOE-23 genotype frequency between 130 cases and 540 controls (7.7% vs. 14%, respectively), but they did not stratify the data by age at onset. Our comprehensive analysis examined risk and age at onset effects of each allele in overall as well as stratified datasets, which provides the first in-depth analysis of APOE-2.

Unlike the dose effect of APOE-4 in Alzheimer disease [13], we did not find that APOE-4 affected age at onset of ALS ( P =0.75), which is consistent with our previous report of no significant differences in age at onset between patients with APOE-X/X and APOE-4/4 or APOE-4/X genotypes [17]. Furthermore, we did not detect any association between APOE and the risk of ALS, which is consistent with our previous report using sporadic ALS data [17] and several others [16]. Rather, we found that APOE affects the risk of ALS within age at onset-specific groups. The use of an age at onset cutoff of 40 years to stratify the dataset is rather arbitrary, because there is no clear definition for the early and late-onset ALS. We also investigated the subsets based on other cut-offs (e.g., 45 and 50 years), but similar results were generated. In conclusion, our findings suggest that APOE may express its strongest effect on the risk of ALS through age at onset.