Patients with ALS and healthy volunteers participating in this study were recruited in the outpatient clinic for motor neuron diseases of the Utrecht University Medical Center, or were part of a population-based study on ALS in the Netherlands. This population has been described in detail elsewhere.10
Patients and controls participating in previous studies on SMN
were excluded from the copy number analyses. Patients with ALS had no family history of the disease and all fulfilled the 1994 El Escorial criteria for probable or definite ALS.11
All participants gave written informed consent. Genomic DNA of patients with ALS and controls was isolated in the same laboratory, using a salting-out procedure. In total we included 847 patients with ALS (57% male) and 984 controls (52% male) in the copy number analyses and 975 patients with ALS (60% male) and 1,044 controls (53% male) in the mutation screen ().
Multiplexed ligation-dependent probe amplification (MLPA) assays were run using standard protocols (www.mlpa.com
). We used the SALSA P060 MLPA kit (MRC Holland, the Netherlands), containing 2 probes specifically targeted to SMN1, 2 probes targeted to SMN2, and control probes targeted to other chromosomal loci for normalization and assay quality control. A total of 50–100 ng of genomic DNA was used in each MLPA assay. Data normalization and analysis were performed with GeneMarker software (SoftGenetics, State College, PA) using standard parameters.
To determine the reproducibility of our MLPA assay, we ran 90 samples twice, in separate reactions, and calculated the copy numbers for both replicates of each sample as described below. For the SMN1 probes, the percentage of agreement was 99% (1 of 90 samples had different copy numbers between the 2 replicates), and 98% for SMN2 (2 out of 90 samples showed different copy numbers between replicates).
For mutation screening, we used PCR and sequencing protocols described elsewhere.12
In short, we designed 2 nested primer pairs for each amplicon, amplified exonic sequences and intron-exon boundaries, and sequenced the amplicons using di-deoxy sequencing. Sequencing was done on ABI 3,730 capillary sequencers with Big Dye Terminator v3.1 chemistry (Applied Biosystems, Foster City, CA). Sequence data were imported in PolyPhred software,13
and sequences were visually inspected for heterozygous sites. All putative mutations were confirmed with an independent PCR and sequencing reaction. Functional impact of identified mutations was predicted using PolyPhen software (http://genetics.bwh.harvard.edu/pph/
). Primer sequences are available upon request. These primers are not specific to SMN1
, but amplify sequences from both genes. Identified mutations cannot, therefore, be mapped specifically to 1 of the 2 genes. We chose this method because approaches to specifically sequence either SMN1
would be extremely laborious, and would only be justified in the case of a suspected association. Two SMA patients with known SMN1
mutations (in the presence of normal SMN2
copy numbers) were used as positive controls. The software called both mutations, thus demonstrating that our method reliably detects mutant alleles at least in a 1:3 ratio.
All statistical procedures were carried out in R
2.10.1 statistical environment (http://www.r-project.org
). Because the quantitative measurement of copy number data is prone to systematic bias leading to false-positive associations,14
we used 2 different methods to test SMN1
copy number state for association with ALS susceptibility. First, we determined SMN1
copy number states for each individual using Gaussian mixture modeling with the CNVtools software package in R
The mean signal of the 2 probes for each gene was used as the input signal. Gaussian distributions were fitted on the signal intensity distributions and individuals were assigned to copy number states based on the highest a posteriori probability. For SMN1
a 3-component model was used (corresponding to 1, 2 and 3 copies) and for SMN2
a 5-component model was used (corresponding to 0, 1, 2, 3, and 4 copies). These copy number states were then used in a multivariate logistic regression model including SMN1
copy number state and with age at onset and gender as covariates. Secondly, we used a likelihood ratio association test employing CNVtools, using a linear trend model. This method was specifically designed to handle intensity data from quantitative measurements, and allows for differential bias due to possible differences in data quality between cases and controls, causing spurious associations.14
Cox regression was used to test for effect of SMN1
copy number on survival, using age at onset, gender, and site of onset as covariates. For the effect on age at onset, we used Cox regression with gender and site of onset as covariates. For the combined analysis of the different studies, we used the random-effects meta-analysis (DerSimonian-Laird) in the rmeta package in R
. We used the Woolf test to test for significant heterogeneity between different studies. In order to test for difference in frequency of SMN
mutations between patients and controls, the Fisher exact test (2-sided) was applied.