a. Manhattan plot of −log10(observed P-value) across genome demonstrating region of genome-wide significant association on chromosome 7; b. Regional plot of the TMEM106B associated interval. Foreground plot: Scatter plot of the −log₁₀ P-values plotted against physical position (NCBI build 36). Background Plot: Estimated recombination rates (from phase 2 of the HapMap) plotted to reflect the local LD structure. The color of the dots represents the strength of LD between the top SNP rs1990622, and its proxies (red: r2 ≥ 0.8; orange 0.8 < r2 ≥ 0.4; blue < 0.4). Gene annotations were obtained from assembly 18 of the UCSC genome browser; c. Location of 3 highest associated SNPs (green arrows) relative to the gene structure of TMEM106B (blue bars, 3′ and 5′-untranslated regions; larger red bars, coding exons; thick gray line, intronic regions; gray dashed line, downstream chromosome sequence) and chromosome 7 location.

a. TMEM106B mRNA expression by QRT-PCR in frontal cortex differed significantly by genotype at rs1990622 (overall P=0.027, genotype TT vs. TC P=0.017, TT vs. CC P=0.03). Black circles, FTLD-TDP (n=18); open squares, normal (n=7); horizontal lines, group mean. Significance of P-values are denoted by the numbers of asterisks. b. TMEM106B mRNA expression in frontal cortex was significantly higher in samples from FTLD-TDP patients compared to normal controls (P=0.045). c. TMEM106B expression in frontal cortex samples in FTLD-TDP with (GRN pos, n=8) or without (GRN neg, n=10) GRN mutations compared to normals (n=7). GRN mutation carriers had significantly higher levels of TMEM106B expression (overall P =0.0009, GRN pos vs. controls P =0.0005, GRN pos vs. GRN neg P =0.002). d. When only cases heterozygous at rs1990622 (n=14) were evaluated, GRN mutations remained significantly associated with a higher level of TMEM106B expression (P =0.039) in frontal cortex. QRT-PCR was performed in triplicate for all expression studies. Expression values were normalized to the geometric mean of two housekeeping genes and are shown relative to a single reference normal control sample23. Error bars represent the standard error of the mean. Normalized gene expression data and sample genotype and gender data used for these analyses are provided online in Supplementary Material.

Manhattan plot of −log10(observed P-value) across genome in cases with (a) and without (b) GRN mutations. The subset of cases with GRN mutations demonstrates regions of genome-wide significant association on chromosomes 7 and 17. The chr 17 association is confirmed to be driven by a shared haplotype in c.1477C>T (p.R493X) GRN mutation carriers representing ~20% of mutation positive cases, however the chromosome 7 association is not related to any single GRN mutation and remains when the cases with c.1477C>T are removed (P=1.446×10⁻¹⁰). The same locus on chr 7 identified in the GRN mutation cases is also the strongest signal in the GRN negative cases, although it does not reach genome-wide significance. A list of the SNPs with the highest signals in b is given in Supplementary Table 8.