Rs1990622 is associated with age at onset of frontotemporal lobar degeneration with TAR DNA-binding protein inclusions (FTLD-TDP) in granulin (GRN) mutation carriers and with GRN plasma levels in individuals without dementia. A, Age at onset was analyzed for association with rs1990622 in 50 GRN mutation carriers by the Kaplan-Meier method and tested for significant differences using a Cox proportional hazards model (proc PHREG; SAS Institute Inc, Cary, North Carolina). Family and sex were included in the model to take into account the relatedness between samples and the potential differences in age at onset between mutations. Healthy mutation carriers were included in the analyses as censored data. Homozygotes for the major allele had an earlier age at onset than the heterozygotes and homozygotes for the minor allele (mean age at onset, 58 vs 74 years; P = 9.9 × 10−7). B, The GRN plasma levels were measured in 79 healthy individuals and tested for association with rs1990622. Rs1990622 showed a significant association with GRN plasma levels (P = 4 × 10−4). C, TMEM106B frontal cortex messenger RNA (mRNA) levels measured in 40 healthy individuals by real-time polymerase chain reaction showed no association with rs1990622 (P = .78). *Levels corrected for sex and postmortem interval.
Cruchaga C, Graff C, Chiang H, Wang J, Hinrichs AL, Spiegel N, Bertelsen S, Mayo K, Norton JB, Morris JC, Goate A. Association of TMEM106B Gene Polymorphism With Age at Onset in Granulin Mutation Carriers and Plasma Granulin Protein Levels. Arch Neurol. 2011;68(5):581-586. doi:10.1001/archneurol.2010.350
Genome-wide association studies are a powerful tool to identify genetic variants associated with risk for disease.1- 4 In the last 2 years, genome-wide association studies for the most common neurodegenerative diseases, such as Alzheimer disease (AD), Parkinson disease, progressive supranuclear palsy, and corticobasal degeneration,1- 4 have found genetic variants associated with risk for disease. The main drawback of these studies is that the functional and biological mechanism that explains the association is often unknown. This is the case for a recent genome-wide association study by Van Deerlin et al5 for frontotemporal lobar degeneration with TAR DNA-binding protein inclusions (FTLD-TDP). Frontotemporal dementia is the third most common neurodegenerative disease after AD and Parkinson disease.6 Their study was performed in 515 FTLD-TDP cases, including 89 individuals carrying pathogenic mutations in the granulin (GRN) gene, a known cause of familial FTLD-TDP. The strongest association was observed with rs1990622, located 6.9 kilobases downstream of the TMEM106B gene (chromosome 7p21). Interestingly, this signal was stronger among the GRN mutation carriers than in the other FTLD-TDP samples. Van Deerlin et al suggested that rs1990622 or another single-nucleotide polymorphism (SNP) in linkage disequilibrium (LD) was modifying the risk for disease through modulation of TMEM106B expression levels because they found a strong association between this SNP and TMEM106B messenger RNA (mRNA) levels. These studies were performed in lymphoblast cell lines and replicated in a small series of human brain samples. However, rs1990622 is not located in an obvious promoter region or transcription factor–binding site for TMEM106B.
Individuals with pathogenic GRN mutations have very low mRNA and plasma levels of GRN,7- 13 but the relationship between TMEM106B, GRN levels, and/or FTLD-TDP pathology is unknown. In this study, we tested whether rs1990622 is a disease modifier in GRN mutation carriers and examined possible pathogenic mechanisms by which variation in TMEM106B influences risk for FTLD-TDP.
DNA was extracted from 50 individuals from families previously shown to have FTLD-TDP caused by mutation in GRN (6 samples from the HDDD1 family,7 13 from the FD1 family,8 18 from the HDDD2 family,9 and 13 from the Karolinska family10) (Table 1). All samples were sequenced and confirmed to have a pathogenic GRN mutation. All participants gave informed consent and the Washington University School of Medicine institutional review board and the Human Subjects Committee of the Karolinska Institute (Stockholm, Sweden) approved the study. Plasma from 73 healthy, elderly individuals and 6 GRN mutation carriers was obtained according to standard procedures.
Genomic DNA was isolated from blood using standard procedures. Rs1990622 was genotyped using KASPar (KBioscience, Hoddesdon, Hertfordshire, England) or TaqMan (Applied Biosystems, Carlsbad, California) technologies. Plasma levels of GRN were measured, in duplicate, using an enzyme-linked immunosorbent assay kit (Human Progranulin ELISA Kit; Adipo-Gen Inc, Seoul, Korea).
Association with expression was carried out using complementary DNA (cDNA) obtained from the frontal lobes of 40 individuals without clinical dementia (Clinical Dementia Rating14 = 0) obtained through the Washington University Charles F. and Joanne Knight Alzheimer's Disease Research Center (WU-ADRC) Neuropathology Core. We included samples for which the Clinical Dementia Rating assessment was done within 6 months of the date of death. The mean (SD) age at death of the included individuals was 85 (9) years (range, 64-107 years). Twenty-one percent of the individuals carried at least 1 APOE ε4 allele. Thirty-nine percent of the individuals were male. Twenty-four brains had a Braak and Braak score ranging from 1 to 4, indicating the presence of some tangle pathology. All the brains had a Consortium to Establish a Registry for Alzheimer's Disease score15 lower than B. Only samples with a postmortem interval lower than 24 hours were included. As explained in the “Statistical and Bioinformatic Analyses” subsection, postmortem interval was not significantly associated with GRN or TMEM106B mRNA levels. RNA integrity was also checked by agarose gel electrophoresis (data not shown).
Total RNA was extracted from the frontal lobe using the RNeasy Mini Kit (Qiagen, Hilden, Germany), following the manufacturer's protocol. The cDNA was prepared from the total RNA, using the High-Capacity cDNA Archive Kit (Applied Biosystems). Gene expression was analyzed by real-time polymerase chain reaction (PCR), using an ABI-7500 real-time PCR system (Applied Biosystems). Real-time PCR assays were used to quantify TMEM106B, GRN, GAPDH, and cyclophilin cDNA levels using TaqMan assays. Each real-time PCR run included within-plate duplicates. Real-time data were analyzed using the comparative Ct method. The Ct values of each sample were normalized with the Ct value for the housekeeping genes GADPH and cyclophilin and were corrected for the PCR efficiency of each assay,16 although the efficiency of all reactions was close to 100%. Only samples with a standard error less than 0.15% were analyzed.
We also used the GEO data set GSE891917 to analyze the association between rs1990622 and TMEM106B gene expression. In this data set, there are genotype and expression data from 486 late-onset AD cases and 279 neuropathologically clean individuals. We extracted the data for normalized TMEM106B mRNA levels measured in the parietal cortex from neuropathologically confirmed healthy individuals and the genotype data for rs1468804. Only mRNA levels from the parietal cortex were used to minimize heterogeneity and maximize sample size (n = 105). Genotypes for rs1468804 were used because rs1990622 was not included in this data set, and rs1468804 is in perfect LD with rs1990622 based on the HapMap database.
Association with age at onset (AAO) was carried out using the Kaplan-Meier method and tested for significant differences using a Cox proportional hazards model (proc PHREG; SAS Institute Inc, Cary, North Carolina), including sex in the model and family ID in the aggregate option (covs[aggregate]), which creates a robust sandwich estimate, to take into account the relatedness between samples and potential differences in AAO between mutations. This approach has been widely used to analyze the association of genetic variants with both risk for disease and AAO in family-based analyses in multiple studies.18- 22 The PHREG procedure in SAS can be used to fit the Cox proportional hazards model and produce a robust variance estimate that is valid for testing association in the presence of linkage within families of arbitrary size.18,19 The advantage of this method is that it is a valid test in the presence of linkage but does not require the correlation structure within families to be specified. This approach is more powerful than the sibship disequilibrium test of Horvath and Laird.20 This test is valid even under extreme residual familial correlation and with no cost in power at the causal locus.20 In our analyses, minor allele homozygotes (n = 2) were combined with heterozygotes because of the small number of individuals with this genotype (dominant model). Healthy mutation carriers were included in the analyses as censored data. The inclusion of these samples did not change the association (eFigure 1).
Association between GRN plasma levels and rs1990622 genotypes was carried out using analysis of covariance. Rs1990622 was tested using an additive model with minor allele homozygotes coded as 0, heterozygotes coded as 1, and major allele homozygotes coded as 2. The GRN plasma levels were not normally distributed and were log-transformed for analysis. We performed a stepwise discriminant analysis to test whether age, APOE, and/or sex affect GRN plasma levels, but none of these factors were associated in healthy individuals and were not included in the model. APOE was tested as a potential covariate because it has been reported that APOE is associated with early memory deficits in GRN mutation carriers.23 In GRN mutation carriers, age was significant and was therefore included in the analysis of covariance. In our study, GRN mutation carriers had GRN plasma levels between 40.5 and 70 ng/mL. The GRN gene was sequenced in the 13 healthy individual samples with GRN plasma levels less than 100 ng/mL, but no mutations were found in these samples.
Association with mRNA levels was carried out using analysis of covariance. Rs1990622 was tested using an additive model with minor allele homozygotes coded as 0, heterozygotes coded as 1, and major allele homozygotes coded as 2. Stepwise discriminant analyses identified postmortem interval and sex as significant covariates for TMEM106B mRNA levels and APOE and age for GRN mRNA levels in the WU-ADRC samples and were included as covariates in the analyses. Stepwise discriminant analysis identified postmortem interval and site as significant covariates in the GEO data set and were included in the analyses.
The SNPs in LD with rs1990622 were identified using the CEU-HapMap data and the Haploview software.24 The LD between rs1990622 and rs3173615 was confirmed by direct genotyping in the WU-ADRC samples. The functional implication of each SNP was analyzed using PupaSuite software.25
To test whether rs1990622 modifies AAO of dementia in GRN mutation carriers, we genotyped rs1990622 in 50 GRN mutation carriers from 4 FTLD families (HDDD1, HDDD2, FD1, and Karolinska) (Table 1) associated with 3 different mutations (Ala9Asp, Gly35fs, and Ala237fs).7- 10 We observed a strong association between rs1990622 and AAO. Individuals homozygous for the risk allele (A allele) had a median AAO 13 years earlier than heterozygotes and homozygotes for the minor allele (G) (P = 9.9 × 10−7) (Figure, A). Family was included in these analyses to take into account the relatedness between samples and potential variation in AAO resulting from different mutations (see the “Methods” section). Although we have included families with different mutations, available evidence suggests that each mutation causes disease via haploinsufficiency. In the case of the Gly35fs and Ala237fs mutations, the haploinsufficiency is caused by nonsense-mediated decay of the mutant mRNA, while the A9D mutation interferes with protein secretion, leading to functional haploinsufficiency.7- 10,26
Most pathogenic mutations in GRN causing FTLD-TDP are null alleles and have been associated with lower GRN mRNA,7- 10 cerebrospinal fluid,13 and plasma levels11- 13 as a result of a haploinsufficiency.7- 10 Therefore, we hypothesized that rs1990622 modifies risk for sporadic FTLD-TDP and AAO in GRN mutation carriers by regulating GRN protein levels. In our series, GRN plasma levels are highly variable in healthy individuals, showing as much as a 4-fold difference between individuals (n = 73; mean [SD] level, 163  ng/μL; range, 76-314 ng/μL), while GRN mutation carriers have very low GRN plasma levels with lower interindividual variability (n = 6; mean [SD] level, 47  ng/μL; range, 42-70 ng/μL). We first tested whether rs1990622 was associated with GRN plasma levels in healthy older adults without dementia and observed a significant difference in mean GRN plasma levels for each genotype. Consistent with our hypothesis, homozygotes for the risk allele had the lowest GRN plasma levels while heterozygotes had intermediate levels and homozygotes for the protective allele had the highest GRN plasma levels (P = 4 × 10−4) (Table 2 and Figure, B). Among the GRN mutation carriers, we also found a significant difference in the GRN plasma levels between the different genotypes (mean [SD], AA, 49.6  ng/μL [range, 40-66 ng/μL] vs AG, 63  ng/μL [range, 56-71 ng/μL]; P = .003).
Van Deerlin et al5 found that rs1990622 is associated with TMEM106B gene expression in lymphoblastoid cell lines and in a small number for frontal cortex samples (7 samples from neurologically unaffected individuals and 18 with FTLD-TDP), suggesting that rs1990622 modifies risk for FTLD-TDP through modulation of TMEM106B mRNA expression. To follow up this result, we tested whether rs1990622 was associated with TMEM106B mRNA levels in the frontal cortex from 40 elderly individuals without dementia. Surprisingly, we found no association between rs1990622 with gene expression (P = .78) (Figure, C) in a sample size almost 2 times the size of the original study. We also tried to replicate the association of rs1990622 with TMEM106B gene expression by analyzing the publicly available GEO data set GSE8919.17 We analyzed TMEM106B mRNA levels in parietal cortex samples (n = 105) from individuals without dementia but did not detect evidence for association with rs1468804 (P = .56), a SNP in perfect LD with rs1990622 (D′ = 1; R2 = 1, based on the CEU-HapMap data). Despite our large data sets, we were unable to replicate the original observation. Because our tissue was from elderly adults without dementia and the original study examined tissue from FTLD-TDP brains, it is possible that the association with expression could represent a SNP × environment interaction, in which rs1990622 modulates TMEM106B mRNA levels in the presence of TDP pathology.
Given the strong association of TMEM106B expression with rs1990622 in GRN carriers in the original study, we tested whether rs1990622 was associated with GRN mRNA levels but failed to detect an association (P = .35; data not shown). We also found no correlation between TMEM106B and GRN mRNA levels (P = .80; Pearson correlation R2 = 0.04). Together these results indicate that the association of rs1990622 with GRN plasma levels is not likely to be driven by modulation of GRN expression levels.
We used the CEU-HapMap data and bioinformatics tools to identify all putative functional SNPs in high LD (r2>0.9) with rs1990622 that could explain the association of rs1990622 with risk for FTLD-TDP, AAO in GRN mutation carriers, and GRN plasma levels. Thirty-two SNPs within 10 kilobases of the TMEM106B gene region showed high LD with rs1990622 (eTable). Six SNPs are located in transcription factor binding sites that could modify gene expression and a seventh is located in a sequence susceptible to the formation of a DNA triplex, which has also been suggested to regulate gene expression.27 Analysis of the CEU-HapMap data revealed that the nonsynonymous SNP in TMEM106B, rs3173615 (Thr-185-Ser), is in perfect LD with rs1990622 (D′ = 1; R2 = 1). Direct genotyping confirmed that these 2 SNPs were also in very high LD in our population (D′ = 1; R2 = 0.927). Threonine 185 and the entire protein are very highly conserved between species (eFigure 2), indicating that this variant may have important functional implications. Polyphen228 predicts that rs3173615 results in a possibly damaging amino-acid change, suggesting that rs3173615 may affect protein function. Our bioinformatics analysis also identified another SNP, rs1042949, in perfect LD with rs1990622, which disrupts an exonic splicing enhancer that may affect alternative splicing, thereby changing protein function.
Our data provide additional support for the role of rs1990622 (chromosome 7p21) or variants in LD with this SNP and genetic risk for FTLD-TDP, especially in GRN mutation carriers. We found that rs1990622 is strongly associated with AAO in GRN mutation carriers. Individuals carrying the risk allele of rs1990622 have a mean AAO 13 years earlier than GRN mutation carriers without an rs1990622 risk allele (P = 9.9 × 10−7). Furthermore, the association of rs1990622 with GRN plasma levels in both normal individuals and GRN mutation carriers suggests that TMEM106B may influence risk for FTLD-TDP by modulating GRN protein levels. We were not able to test directly the association between GRN plasma levels and AAO, but we found that unaffected GRN mutation carriers have higher GRN plasma levels than affected individuals. This analysis was done in a small sample and should be interpreted with caution. There are several reports in which GRN plasma levels have been measured in affected and unaffected GRN mutation carriers as well as in healthy elderly individuals or in individuals with other neurodegenerative diseases.11,13,29,30 In all of these reports, it was found that GRN mutation carriers, regardless of their disease status, have very low GRN plasma levels compared with other individuals. However, there are inconsistent results regarding GRN levels in affected GRN mutation carriers compared with the unaffected mutation carriers. In Ghidoni et al30 and Sleegers et al,29 unaffected mutation carriers had slightly lower GRN plasma levels compared with the affected individuals. On the other hand, our results and Finch et al13 show unaffected carriers with higher GRN levels than the affected carriers. These contradictory results, and the fact that all of these studies were carried out in small series, indicate that more studies are necessary. These studies should be designed to address directly whether GRN plasma levels are associated with the onset of disease. There are several factors that should be taken into account in the design of these experiments. Our data and others indicate that GRN plasma levels in GRN mutation carriers are very low and show much lower interindividual variability compared with healthy individuals11,13,29,30; therefore, small differences are expected to be found and very sensitive assays should be used. It will be necessary to avoid external sources of variability that could mask the difference in GRN plasma levels between the affected and unaffected carriers. For example, we found that the number of freeze-thaw cycles and the time that the plasma has been frozen decrease the GRN plasma level measured by enzyme-linked immunosorbent assay (data no shown). Obviously, it is necessary to include the maximum number of samples to have enough power to find a significant difference.
Our data support the hypothesis that rs1990622, rs3173615, or other SNPs in LD with these SNPs are associated with AAO, possibly through a modulation of GRN plasma levels. This mechanism shows some similarities with that proposed for APOE in AD. Although the pathogenic mechanism of APOE is not entirely understood, multiple studies have demonstrated that the APOE ε4 allele is associated with both increased risk for AD and earlier AAO in the general population1- 3 and earlier AAO in familial AD mutation carriers.31 We have reported in several articles that the APOE ε4 allele is strongly associated with lower cerebrospinal fluid β-amyloid 42 levels,32- 35 increasing the risk for disease and modifying the AAO. Similarly, the TMEM106B polymorphism increases risk for FTLD-TDP and decreases the AAO through modulation of GRN plasma levels, a protein known to play a causal role in FTLD. The absence of association between rs1990622 and TMEM106B or GRN mRNA levels suggests that TMEM106B affects GRN protein levels, but not through modulation of GRN or TMEM106B gene expression. TMEM106B is a transmembrane protein, conserved among mammals, with an unknown function. GRN encodes a secreted protein, progranulin, that can be cleaved into smaller peptides (GRNs). Our results suggest that TMEM106B may influence secretion, clearance, or cleavage of GRN protein and that genetic variants in TMEM106B could affect this process, modifying the AAO of dementia in individuals with GRN mutations, and be a risk factor for disease in the general population, similar to APOE and AD.
In our data set, carriers of the protective allele of rs1990622 had a mean AAO 13 years later than risk allele carriers. The effect of this polymorphism on the AAO is similar in magnitude to the APOE genotype on the onset of AD or the length of the CAG repeat in the huntingtin gene on the onset of Huntington disease.36 Given this observation, inclusion of the variant should be considered in clinical genetic testing for families with GRN mutations. The identification of the responsible genetic variant in TMEM106B that drives the association and the characterization of the functional mechanism implicated in this association could represent a critical step in the development of new therapies for FTLD-TDP. A therapy that could mimic the effect of the protective allele could significantly delay the AAO of symptomatic disease.
Correspondence: Alison Goate, DPhil, Department of Psychiatry, Washington University School of Medicine, 660 S Euclid Ave, B8134, St Louis, MO 63110 (email@example.com).
Accepted for Publication: November 10, 2010.
Published Online: January 10, 2011. doi:10.1001/archneurol.2010.350
Author Contributions:Study concept and design: Cruchaga and Goate. Acquisition of data: Graff, Chiang, Wang, Spiegel, Mayo, Norton, and Morris. Analysis and interpretation of data: Cruchaga, Hinrichs, Spiegel, Bertelsen, and Goate. Drafting of the manuscript: Cruchaga, Graff, Chiang, and Goate. Critical revision of the manuscript for important intellectual content: Graff, Wang, Hinrichs, Spiegel, Bertelsen, Mayo, Norton, Morris, and Goate. Statistical analysis: Cruchaga, Hinrichs, and Bertelsen. Obtained funding: Morris and Goate. Administrative, technical, and material support: Graff, Chiang, Spiegel, Mayo, Norton, and Morris. Study supervision: Graff and Goate.
Financial Disclosure: None reported.
Funding/Support: This work was supported by National Institutes of Health grants AG16208, P01AG03991, P50AG05681, and P01AG026276, the Alzheimer's Association, the Barnes–Jewish Hospital Foundation, Swedish Brain Power, Swedish Alzheimer Foundation, the Marianne and Marcus Wallenberg Foundation, the Gun and Bertil Stohnes Foundation, the Gamla tjänarinnor Foundation, and Karolinska Institutet KID funding. Dr Cruchaga has a fellowship from Fundacion Alfonso Martin Escudero. Dr Wang is supported by grant IIRG-07-60299 from the Alzheimer's Association.
Additional Contributions: We thank the Clinical Core of the ADRC for subject recruitment and assessment and the Genetics Core of the ADRC for APOE genotypes. We gratefully acknowledge the individuals who participated in this study. We also acknowledge the contributions of the Genetics and Clinical cores of the WU-ADRC.