**3.1 Replication study chr9p21 GWAS**

544 Amyotrophic Lateral Sclerosis

Project (http://www.1000genomes.org). Third, variants in nucleotide stretches were filtered out because they are known to be error-prone in NGS data. This resulted in a total of 189 variants, all located outside coding regions of known genes confirming genebased mutation analyses. These variants were genotyped in all 29 individuals of the DR14 family using Sanger sequencing and tested for segregation. 120 variants were located on the disease haplotype and were analyzed in a series of 300 neurologically healthy control individuals collected in Flanders, Belgium, i.e. the geographical region of which family DR14 originates, using multiplex Sequenom MassARRAY technology. 37 of these variants were completely absent in 300 control individuals and are all located in untranslated regions or introns of genes, or intergenic. We are currently prioritizing these variants based on evolutionary conservation, regulatory potential, location compatible with cis-acting function on functional candidate genes, etc. Also, we are determining the presence of these variants in a Belgian population of unrelated patients with ALS (N=124), ALS-FTLD (N=21) and FTLD (N=203), aiming to find a possible founder mutation. We already showed evidence for the presence of founder mutations in the Flanders-Belgian FTLD collection, by the *GRN* IVS1+5 G>C founder mutation identified in 19% of familial FTLD (Cruts et al., 2006). We have investigated the patient population for chromosome 9p STR markers and did not find evidence for haplotype sharing with family DR14; however, we cannot exclude the presence of a small,

**3. Population-based association for ALS and FTLD to chromosome 9p** 

In 2009, the first ALS GWAS showing association with a locus at chromosome 9p21 was reported by Van Es and colleagues. They identified genome-wide significance with two SNPs, rs2814707 and rs3849942, almost in complete linkage disequilibrium (LD) with each other and located in an LD block of ~80 kb. Also a third SNP in this LD block (rs774359) showed suggestive association (figure 1). This LD block is situated at the telomeric end of the minimally linked candidate region found in the ALS-FTLD families and contains only three genes: part of *MOBKL2B*, *IFNK* and *C9orf72* (figure 1). Next, data of the first GWAS in FTLD-TDP were suggestive for association of five SNPs (rs774352, rs774351, rs3849942, rs2814707, rs774359) on chromosome 9p21, in the same LD block (Van Deerlin et al., 2010). Subsequently, a Finnish and a British independent ALS GWAS identified genome-wide significance with SNPs rs3849942, rs2814707, rs774359, rs2225389 (Laaksovirta et al., 2010) and with SNPs rs3849942, rs2814707, rs903603 (Shatunov et al., 2010) respectively, all in the same locus at chromosome 9p21. The Finnish study defined a 42-SNP haplotype associated with increased risk of ALS in the Finnish population, located in a 232 kb LD block which overlaps with the previously reported 80 kb LD block (van Es et al., 2009) and the 106.5 kb LD block of the UK study (Shatunov et al., 2010). Because of the unique homogeneous genetic structure of the Finnish isolated population, the extent and structure of LD is different than in other European countries. To date, one study replicated the association of the chr9p21 locus in an ALS-FTLD subpopulation (Rollinson

To assess the contribution of the chr9p21 risk factor to disease etiology in Belgium, we replicated one of the top SNPs associated in all GWAS reports, rs2814707, in a Belgian population of ALS, ALS-FTLD and FTLD patients. In addition, we performed a meta-

analysis of the different published association studies with inclusion of our study.

previously undetected founder haplotype.

et al., 2011).

We investigated association of the most widely studied GWAS top SNP at chr9p21, rs2814707, in a Flanders-Belgian population of genealogically unrelated patients clinically diagnosed with ALS (N=124), ALS-FTLD (N=21) or FTLD (N=203) according to established consensus criteria (Brooks et al., 2000; Neary et al., 1998), compared to a group of 510 unrelated neurologically healthy control individuals from the same region in Belgium. We genotyped rs2814707 and showed that this SNP is in Hardy-Weinberg Equilibrium. Allelic and genotypic single SNP association was calculated using logistic regression analysis. The SNP showed significant allelic and genotypic association in the total population and highly significant association in the ALS and ALS-FTLD subpopulation reaching a maximal odds ratio of 3.27 in ALS patients homozygous for the minor allele (table 2). In the FTLD subpopulation no association was found, demonstrating that the effect in the total population can entirely be explained by the effect in patients with an ALS phenotype. When we include 21 ALS samples of Bulgarian origin, the relative risk became even higher, compared to Belgians only, indicating that the associated allele is the same between different populations.


Table 2. Allelic and genotypic association of a GWAS top SNP in the total population and the ALS/ALS-FTLD subpopulation. P-values are corrected for age at onset or inclusion and gender. (OR: odds ratio; CI: confidence interval)

### **3.2 Meta analysis on chromosome 9p21**

We combined the data from the different GWA studies and our study to determine the relative risk of carrying the risk allele on chromosome 9p21. A meta-analysis of the most widely studied SNP on chromosome 9p21 (rs2814707) underscores the presence of a genetic risk factor for ALS and/or FTLD at this locus. Carriers of the rs2814707 minor allele are at increased risk to develop ALS or FTLD (ORmeta 1.29 (95% CI 1.18-1.41), p-value 2.3\*10-8 (Figure 4)). When excluding the GWAS cohorts in which the association was first reported (van Es et al., 2009) to exclude bias because of winner's curse, the strength of the association remains similar (ORmeta 1.32 (95%CI 1.17-1.49; p-value 3.5\*10-6). Exclusion of three studies, including our own, which combine FTLD and ALS phenotypes would have resulted in an ORmeta 1.24 (95%CI 1.13-1.36); p-value 3.3\*10-6).

A Major Genetic Factor at Chromosome 9p Implicated in

Amyotrophic Lateral Sclerosis (ALS) and Frontotemporal Lobar Degeneration (FTLD) 547

context means the gene, regulatory element, conserved element or, in the absence of recognizable elements, 1 kb flanking each side of the putative mutation. This might identify additional mutations resulting in the same functional defect as the mutations detected in DR14 and further enhance the likelihood of the variant(s) to be disease-related. Finding such variants will provide strong genetic evidence of a disease causing effect of the variants. Alternatively, in case we do not find a mutation in this first selection of variants, we can use more relaxing filters. Taking into account that dbSNP may include rare clinical variants, rare or non-validated dbSNP SNPs will also be considered (N=91). Also the candidate region can be extended to the next recombinant or to the large DR14 candidate region. Further, regions that are not covered in more than one genome, will be completed using classical sanger sequencing. Finally, structural variants and copy number mutations will be investigated. More than five years of research in the ALSFTD2 locus in different ALS-FTLD families worldwide did not identify pathogenic mutations yet (table 1), although mutations in two different genes on chromosome 9 outside the minimal candidate region, *IFT74* and *SIGMAR1*, were suggested (Luty et al., 2010; Momeni et al., 2006) but without further confirmation in other families. The fact that the culprit gene is still not found may in part be explained by the fact that families linked with chromosome 9p21 do not all have the same disease haplotype so that different mutations, probably with the same effect on the same gene, are most likely involved. Also, the causal mutations are most likely unusual with respect to position or type. For example, deep intronic mutations or mutations in a distant regulatory element might cause the disease but assessing their effect is rather complicated.

Also, identification of small insertions/deletions or inversions is challenging.

chromosome 9p21 region.

In addition, we replicated association in a Belgian cohort of ALS, ALS-FTLD and FTLD patients of two major top SNPs on chromosome 9p21 previously associated in several ALS and FTLD GWA studies. More specifically, we found that the risk haplotype at chromosome 9p21 is most substantially increased in patients with ALS or ALS-FTLD compared to control individuals. The lack of association in the FTLD subpopulation is similar to what was observed in a previous replication study in which association was only found in ALS-FTLD patients (Rollinson et al., 2011). Also, the weakest association signal was found in the FTLD GWAS compared to ALS GWAS. This is the first time that a susceptibility locus for ALS is replicated in different GWA studies and replication studies, underlining the importance of the chromosome 9p21 locus harbouring a risk increasing factor for ALS (and ALS-FTLD) across multiple populations with a high relative risk of disease susceptibility. We are further characterizing this genetic association to reduce the associated region in the Belgian population. We are finemapping the chromosome 9p risk haplotype in great detail in our ALS, FTLD, ALS-FTLD patient cohorts by making a high density SNP map of the complete LD block and using extended association analyses of series of known and newly identified variants in the LD block. These variants were identified in previous publications, hapmap, 1000 Genomes Project and extended genomic sequencing efforts of the linkage disequilibrium block in a selection of ALS and ALS-FTLD patients carrying the associated allele of the GWAS SNPs in a homozygous or heterozygous state. This will finally result in the identification of the functional variant explaining the strong association in the

The observation that the chromosome 9p21 region is harboring both disease-causing variants and susceptibility factors with high penetrance, might suggest that different genetic variants with variable degree of biological consequences might be involved. Alternatively,

Fig. 4. Forest plot of a random effects meta-analysis of rs2814707. Meta-analysis was conducted in rmeta v2.16, and based on effect estimates and standard errors for the minor allele reported in each individual publication. Odds Ratios and 95% Confidence Intervals are given for each study separately along with a summary Odds Ratio, of the minor allele relative to the major allele. All 9p21 association studies on ALS, ALS-FTLD and FTLD published until July 2011 were included, in addition to our own unpublished data. For the study of Shatunov and colleagues we only included data on the independent UK cohort, to avoid overlap of datasets with previous studies. From Rollinson et al, only data on the Manchester ALS-FTLD cohort are included.
