Protective TMEM106B-rs3173615 delays age at onset in GRN mutation carriers Authors: Lorenzo Pagano, Claudia Saraceno, Andrea Geviti, Silvia Fostinelli, Alessandro Facconi, Valentina Laganà, Giulia Giacomucci, Maria Sofia Cotelli, Assunta Ingannato, Silvia Bagnoli, Antonio Longobardi, Alice Russotto, Sonia Bellini, Davide Lagrotteria, Ersilia Paparazzo, Giuliano Binetti, Alberto Montesanto, Benedetta Nacmias, Raffaele Maletta, Barbara Borroni, Roberta Ghidoni Published: 2026 ABSTRACT AbstractOne of the major causative genes involved in Frontotemporal dementia (FTD) is Granulin (GRN), encoding for Progranulin (PGRN). GRN mutation carriers show a substantial heterogeneity with high variability in age at onset and pathological presentation, even within the same family or identical mutations, suggesting the presence of additional genetic factors. Single nucleotide polymorphisms in the Transmembrane protein 106B (TMEM106B) locus were identified as a genetic risk-associated factor for FTD. The top variant identified was the non-coding rs1990622, with the major allele (T) associated with an increased risk to develop FTD, while subjects with the minor allele (C) were less likely to develop disease, suggesting a protective effect. In this study, we investigate in a large Italian cohort of GRN mutation carriers, how the coding variant TMEM106B-rs3173615, in linkage disequilibrium with rs1990622, modulates age at onset, survival, and PGRN levels, including, up to date, the highest sample size of homozygous protective allele carriers. Genetic screening for TMEM106B-rs3173615 was performed on a total of 187 GRN mutation carriers, comprising 131 FTD patients and 56 pre-symptomatic subjects. Individuals with the protective genotype (GG) had a risk of FTD onset reduced by 80%, with a median age at onset of 77 years compared to a median age at onset of 63 years for individuals without the protective genotype. TMEM106B-rs3173615 acts as a genetic modifier of age at onset in the presence of GRN mutations and could be considered in clinical practice to optimize risk stratification for FTD.Subject terms: Genetics, Predictive markers Introduction Frontotemporal dementia (FTD) is the second most common form of early-onset dementia, particularly affecting individuals under the age of 65 years old [1, 2]. In general, up to 40% of FTD cases report a family history of dementia, although only 10% show an autosomal dominant pattern [1–5]. One of the major causative genes involved in FTD is Granulin (GRN), which leads to intracellular transactive response DNA-binding protein (TDP)-43 positive inclusions (FTLD-TDP) [6–8]. GRN mutations are the most frequent genetic determinant of familial dementia in Northern Italy [9]. GRN encodes Progranulin (PGRN) a growth factor involved in multiple cellular processes including growth, lysosomal function, and neuroinflammation [10]. A decrease in circulating PGRN protein, easy to detect by plasma/serum dosage [11], is accounted by the majority of GRN stop codons mutations. GRN carriers show a substantial heterogeneity with a high variability in age at onset and clinicopathological presentation, even within the same family or with identical mutations. The penetrance of GRN mutations is not complete, even at old age, suggesting that environmental influences or additional genetic factors modify the disease manifestation [12–18]. Genetic studies have played a significant role in identifying risk factors for FTLD-TDP. A genome-wide association study (GWAS) identified single nucleotide polymorphisms (SNPs) at the Transmembrane protein 106B (TMEM106B) locus, on chromosome 7p21.3, as a genetic risk-associated factor for FTLD-TDP. The top variant identified was the non-coding rs1990622, with the major allele (T) associated with an increased risk to develop FTLD-TDP. On the contrary, subjects with the minor allele (C) were less likely to develop disease, suggesting a protective effect. Interestingly, the association and the effect of TMEM106B was most evident in the subgroup of FTLD-TDP patients carrying pathogenic variants in GRN gene [19]. Subsequently, several groups have replicated the GWAS with almost the same results [20–22]. Interestingly, the first independent replication study showed a highly significant decrease in the frequency of the minor/protective alleles of all TMEM106B SNPs in FTD cases compared to controls. Moreover, in this study, the authors demonstrated that the protective allele of TMEM106B SNPs was nominally associated with a later onset of FTLD, in particular in a recessive model, assuming that only the presence of two protective alleles was required to confer protection against FTLD-TDP. They showed that the protective allele of TMEM106B SNPs was associated with a significant increase in PGRN levels in plasma of controls [21]. Among the associated TMEM106B SNPs, the rs3173615 was the only one coding variant that was found to be in perfect linkage disequilibrium (LD) with rs1990622 [21]. The major/risk allele (C) encodes a threonine at residue 185, whereas the minor/protective allele (G) encodes serine. This change, p.Thr185Ser, has been suggested to contribute to the disease-modifying effect, altering the protein stability and degradation rate [23]. Pottier et al. in 2018 estimated that GRN carriers (n = 382) of the TMEM106B protective allele (G) of rs3173615 (GG, n = 4), had 50% lower chance to develop FTLD symptoms than carriers of the risk one (C) [24]. Recently, Marks et al. demonstrated that the protective allele at rs3173615 is associated with longer survival after symptoms onset in FTLD-TDP patients, suggesting a less aggressive disease course for the protective allele carriers [25]. All these studies support the role of TMEM106B locus as an important genetic factor for disease modification, suggesting a protective effect of the homozygous genotype, although the majority of them shows a limited sample size of the homozygous genotype group. In this study, we investigate in a large Italian cohort of GRN mutation carriers, how the coding variant rs3173615 within the TMEM106B gene modulates age at onset, survival, and PGRN levels, including, up to date, the highest sample size of homozygous protective allele carriers. Methods Participants This study has been performed within the framework of the GARDENIA “Genetic and epigenetic modulAtors in Rare neurodegenerative disease with DEmentia: a National study on autosomal dominant Alzheimer disease and genetic frontotemporal degeneration with dementia” project funded by the European Union—Next Generation EU (PNRR-MR1-2022-12375654, [26]). In particular, the individuals investigated in this study were selected among FTD patients (n = 131) and pre-symptomatic subjects (n = 56) for which pathogenic mutations in GRN gene have been identified. The complete list of GRN mutations of all subjects is listed in Table S1. The pathogenicity of genetic variants was assessed according to the variant interpretation criteria established by the American College of Medical Genetics and Genomics (ACMG) [27]. Clinical significance data were retrieved from ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/), VarSome [28], the Human Gene Mutation Database (HGMD) [29]. Additional information on mutations associated with FTD was obtained from FTDtalk (https://www.ftdtalk.org/) and Genetic FTD Initiative database (GENFI; https://www.genfi.org/ftd-mutation-database/). Two Variants of Uncertain Significance (VUSs) were also included in the mutations list since their carriers presented reduced PGRN concentration under the cut-off of 74.8 ng/mL [30]. The clinical diagnosis of FTD was made according to current clinical criteria [31, 32]. DNA samples, belonging to Italian pedigrees, were already stored at the institutional biobank/biorepositories of the three Italian centers belonging to GARDENIA Consortium: IRCCS Centro San Giovanni di Dio Fatebenefratelli BioBank Brescia, Italy (bbmri-eric ID: IT_138442378660827 and Orphanet Biobank) (n = 168); Azienda Ospedaliero Universitaria Careggi in Florence, Tuscany (Orphanet Biorepository of neurodegenerative disorders) (n = 2); Azienda Sanitaria Provinciale di Catanzaro, Calabria (Orphanet Biobank CRN (Regional Neurogenetic Centre)) (n = 17). Informed consent was obtained from all subjects involved in the study. The study was conducted in accordance with the Declaration of Helsinki, and approved by the local Ethics Committee “Comitato Etico IRCCS San Giovanni di Dio Fatebenefratelli” of the IRCCS Centro San Giovanni di Dio Fatebenefratelli, Brescia (Prot. N. 63/2022; date of approval: 7 December 2022). Genetic analysis The whole exome sequencing (WES) was performed by means of bead-linked transposome and hybrid-capture chemistry on Illumina NextSeq 2000 system (Illumina, San Diego, CA, USA). The quality assessment of gDNA was performed on a 0.8% agarose gel, and gDNA was quantified with a Qubit dsDNA BR Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). A total of 250 ng of gDNA was used for library preparation with Illumina DNA Prep with Exome 2.5 Enrichment kit (Illumina, San Diego, CA, USA). gDNA was tagmented, amplified, and purified with Illumina Purification Beads (Illumina, San Diego, CA, USA). The size, quality, and quantity of libraries were assessed with a High Sensitivity DNA kit on a Bioanalyzer instrument (Agilent Technologies, Santa Clara, CA, USA). A 2000 pM sample of the pooled library was loaded with NextSeq 1000/2000 P3 reagents (Illumina, San Diego, CA, USA) and sequenced on an Illumina NextSeq 2000 system (Illumina, San Diego, CA, USA). Bioinformatic analysis: data pre-processing, mapping, and variant calling Exome sequences underwent an initial quality check using FastQC software (version 0.12.1) [http://www.bioinformatics.babraham.ac.uk/projects/fastqc/]. Low-quality reads and adapter sequences were removed with Trimmomatic (version 0.39) [33]. The remaining high-quality reads were then aligned to the hg38 reference genome using BWA-MEM (version 0.7.17-r1188) [34]. Afterwards, duplicate reads were marked using the MarkDuplicates module from Picard (version 3.0), while base quality scores were recalibrated with the BaseRecalibrator and ApplyBQSR modules of GATK (version 4.4.0.0). Single nucleotide variant (SNV) calling was performed using the HaplotypeCaller module of GATK (version 4.4.0.0), generating gVCF files that were subsequently merged with the GenotypeGVCFs module (GATK, version 4.4.0.0). To assess false-positive variants, Variant Quality Score Recalibration (VQSR) was executed with the VariantRecalibrator and ApplyVQSR modules of GATK (version 4.4.0.0). The coverage of the TMEM106B-rs3173615 variant was assessed, indicating that over 90% of the samples exhibited a depth of coverage (DP) exceeding 20X. In silico functional variant annotations were retrieved with ANNOVAR (version 2020-06-08) [35], complemented by manually curated annotations from various online databases. The following annotation sources were used for variant interpretation: ClinVar [36], CADD (version 1.7) [37], PolyPhen-2 (version 2.2.3) [38], SIFT [39], PROVEAN (v1.1) [40], FATHMM (version 2.3) [41], and GERP [42], all accessed on 3 February 2025. Population allele frequencies were retrieved from gnomAD (version 4.1.0) (accessed on 3 February 2025) [43]. Protein stability predictions were conducted using I-Mutant (version 2.0) [44], MuPro (version 1.0) [45], and Missense3D-DB (version 1.5.4) [46], all accessed on 4 February 2025. Statistical analysis Descriptive statistics were computed, with the continuous variables reported as mean (± Standard Deviation) and the categorical variables as n (%). The normality of continuous variables was assessed using Shapiro-Wilk test. Between-group comparisons of continuous variables were conducted using one-way ANOVA tests, while categorical data were analyzed using the Chi-squared test. Kaplan-Meier analysis was performed to estimate age at onset stratified according to the rs3173615 genotypes and a Cox proportional hazard regression model was used to assess the association between rs3173615 and disease onset risk, adjusting for covariates [47]. As a complementary analysis, the association between rs3173615 and age at onset was additionally evaluated among affected individuals only using a Linear Mixed Model (LMM). This approach was included to provide an estimate of genotype-related differences in mean onset age, without incorporating pre-symptomatic subjects. The LMM and Cox models were adjusted for sex, mutation, and year of birth. Given that the majority of subjects were mutated for Leu271LeufsX10, mutation was coded as a dichotomous variable assuming value 1 for Leu271LeufsX10 mutated individuals and 0 for the others. Year of birth was added to further reduce the possibility of confounding, as subjects belonged to different generations [48]. Furthermore, we added a genetic relationship matrix (GRM) and a family code random effect to account for the correlation between subjects. The proportional hazard assumption was checked. In the LMM model, we considered only the 131 GRN FTD patients, while the Kaplan-Meier and Cox regression included also the pre-symptomatic subjects, right-censored to their age of last follow-up. Survival analysis was then performed, using Kaplan-Meier curves and a Cox model as well. Survival was calculated as time from symptom onset to time of death from any cause (outcome= 1) or censoring date (February 1, 2025, outcome= 0). The same adjustments and random effects were included in the Cox model. Finally, we investigated the association between rs3173615 and PGRN levels in a subgroup of GRN subjects (n = 100) with available data on PGRN plasma concentration, measured by AdipoGen assay (n = 66 FTD patients and n = 34 pre-symptomatic subjects), implementing a LMM. The estimates were adjusted for sex and mutation and we consistently included the GRM and the family code random effect. For all analyses (Kaplan–Meier, Cox proportional hazards, and LMM), recessive, dominant, and additive genetic models were tested. Accordingly, p-values were adjusted for multiple comparisons using Bonferroni correction by multiplying the original p-values by three. Survival analyses were carried out using the “survival”, “survminer”, and “coxme” packages of R [49]. All analyses were performed using R software (version 4.3.2). Results Subjects A total of n = 187 GRN mutations carriers (n = 131 FTD patients and n = 56 pre-symptomatic subjects), belonging to 113 genealogically unrelated pedigrees from Italy were screened for the presence of TMEM106B-rs3173615. We found a remarkably high number of homozygous protective allele carriers (GG, n = 18). In particular, n = 8 of them belong to the same family. The rs3173615 genotypes, demographic and clinical characteristics of subjects are summarized in Table 1. Table 1TMEM106B-rs3173615 genotypes, demographic and clinical characteristics of subjects included in the study.GRN FTD patients (n = 131)GRN pre-symptomatic (n = 56)All GRN carriers (n = 187)p ValueGenotypesCC (n = 61)CG (n = 64)GG (n = 6)CC (n = 10)CG (n = 34)GG (n = 12)CC (n = 71)CG (n = 98)GG (n = 18)Age69.4 (8.8)67.5 (10.1)73.8 (8.8)52.4 (12.6)46.7 (11.7)63.7 (14.1)67.0 (11.1)60.3 (14.6)67.1 (13.2)2.0 ×10−16 aSex, % female27.0 (44.3%)33.0 (51.6%)2.0 (33.3%)7.0 (70.0%)16.0 (47.1%)5.0 (41.7%)34.0 (47.9%)49.0 (50.0%)7.0 (38.9%)0.65 bAge at Onset63.4 (8.4)61.2 (8.3)64.3 (7.9)///63.4 (8.4)61.2 (8.3)64.3 (7.9)0.38 aSurvival (years)6.0 (4.8)6.3 (5.7)9.5 (6.5)///6.0 (4.8)6.3 (5.7)9.5 (6.5)0.28 aMean (± Standard Deviation) was computed for continuous variables, while n (%) was computed for categorical variables. p-value of the comparisons between FTD patients and pre-symptomatic genotypes.aOne-way ANOVA test for comparisons among six groups.bChi-squared test. TMEM106B-rs3173615 genotypes, demographic and clinical characteristics of subjects included in the study. Mean (± Standard Deviation) was computed for continuous variables, while n (%) was computed for categorical variables. p-value of the comparisons between FTD patients and pre-symptomatic genotypes. aOne-way ANOVA test for comparisons among six groups. bChi-squared test. TMEM106B-rs3173615 and disease onset - survival The distribution of age at onset of the GRN FTD patients classified by TMEM-rs3173615 genotype was shown in Fig. 1. Boxplot of age at onset. Boxplot of age at onset distribution according to TMEM106B-rs3173615 genotypes. Homozygous risk allele CC, heterozygotes CG and homozygous protective allele GG. The Kaplan–Meier curve for the recessive model (Fig. 2) revealed a significant effect of rs3173615 (p = 0.036). The median age at onset was 63 years for the CC + CG carriers (n = 169) (95% CI: 61 ÷ 65), and significantly higher for the homozygous GG (n = 18), with a median age at onset of 77 years (95% CI: 65 ÷ N.A.*; * the number of observations is too small to obtain an estimate for the upper limit of the confidence interval). Kaplan-Meier analyses for the additive and dominant models showed non-significant results (p = 0.06 and p = 1 after Bonferroni adjustment, respectively). Kaplan–Meier curves of age at onset. Kaplan–Meier curves for TMEM106B-rs3173615 genotypes. Kaplan–Meier curve showing age at onset. Homozygous and heterozygotes C-carriers (CC + CG) and homozygous (GG), recessive model, p = 0.036. Accordingly, the Cox regression analysis for the recessive model showed that rs3173615 was significantly associated with age at onset (p = 0.007) with a hazard ratio of 0.20 (95% CI: 0.07 ÷ 0.57), suggesting that the relative risk of FTD onset for GG carriers (n = 18) was reduced by approximately 80% (1-0.20) compared to CC + CG carriers (n = 169). Sex and mutation type were not associated with onset. The proportional hazards assumption was satisfied (p = 0.73). Cox regression analyses for the additive and dominant models did not yield significant results for rs3173615. The secondary analysis evaluating the association between rs3173615 genotypes and age at onset among 131 affected GRN-FTD patients (LMM) yielded non-significant results under recessive, dominant and additive genetic models. The association between the rs3173615 genotypes and survival from disease onset was evaluated with the three genetic configurations, as well. Both Kaplan-Meier and Cox regression analyses did not reveal significant longer survival in patients carrying rs3173615. We present the results for the recessive model (p = 0.96 and p = 1, respectively) (Fig. 3). Kaplan–Meier curves of survival. Kaplan–Meier curves for TMEM106B-rs3173615 genotypes. Kaplan–Meier curve showing survival after onset. Homozygous and heterozygotes C-carriers (CC + CG) and homozygous (GG), recessive model, p = 0.96. TMEM106B-rs3173615 and PGRN plasma levels We investigated the association between rs3173615 and PGRN levels in a subgroup of GRN subjects (n = 100) with available data on PGRN plasma concentration (n = 66 FTD patients and n = 34 pre-symptomatic subjects). No significant differences among genotypes were observed under the recessive or additive models. The dominant model showed a nominally significant association (p = 0.039), which did not remain significant after Bonferroni correction (p = 0.117) (Figure S1). These results are based on models adjusted for sex and mutation, with the GRM and family code included as random effects. Variant interpretation The rs3173615 variant exhibits a high population frequency in gnomAD datasets, reaching 41% in the Non-Finnish European (NFE) population and 45% in the general population. Given this high prevalence, it can be unequivocally classified as benign according to the American College of Medical Genetics and Genomics (ACMG) variant interpretation guidelines [27]. Online databases, including ClinVar and the Human Gene Mutation Database (HGMD), also describe rs3173615 as a benign polymorphism, despite reported associations with certain neurological disorders, particularly FTD. In silico prediction tools, such as CADD, PolyPhen-2, SIFT, PROVEAN, and FATHMM, consistently classify the variant as benign or tolerated. Notably, GERP predicts the variant site to be highly conserved, with a positive score of 5.36 (maximum = 6.18). Regarding protein stability predictions, Missense3D-DB classifies the protective variant as structurally neutral, whereas MUPro and I-Mutant predict a decrease in TMEM106B protein stability (ΔΔG = −1.02 and ΔΔG = −3.03, respectively). Discussion In this study, we aimed to investigate in a well-characterized Italian cohort of GRN mutation carriers (n = 187) mostly carrying the Leu271LeufsX10 mutation (n = 142), how the coding variant TMEM106B-rs3173615, in LD with the top one rs1990622 identified in the first GWAS, modulates the disease status and progression. Our data further support the protective effect of the minor allele by modulating age at onset. This study is strengthened by a large sample size of GRN mutation carriers and to the best of our knowledge, this is the first to include a remarkable number of individuals homozygous for the protective allele (GG, n = 18), even compared to the biggest FTD Consortia studying TMEM106B [24, 50–52]. We showed that rs3173615 was significantly associated with a reduced risk of FTD onset, with GG carriers exhibiting approximately a 80% lower relative risk compared to CC + CG carriers. Notably, one individual homozygous for the protective allele died at the age of 86 years old without any neurological symptoms. However, the relationship between rs3173615 and survival after symptoms onset did not indicate a prolonged survival for GRN patients carrying the protective genotype. These findings suggest that the protective effect significantly delays symptoms onset in FTD but does not extend survival following disease onset. This could imply that the protective effect may be limited to early pathogenic mechanisms rather than influencing disease progression. However, the number of homozygous GG rs3173615 FTD patients was small (n = 6). Increasing the sample size may reveal a significant effect on survival as well. Age at onset was primarily evaluated using survival analyses, which account for censoring and familial relatedness. Although the complementary LMM in affected individuals only was not statistically significant, this approach does not include unaffected carriers and is less sensitive to differences in onset timing. The survival analysis results indicate that rs3173615 mainly influences the hazard and timing of disease onset rather than mean onset age among affected individuals. Interestingly, in contrast to our study, a GRN mutation carrier who was homozygous for the TMEM106B risk allele remained asymptomatic at the age of 80 [53], suggesting that the presence of the risk allele does not preclude a benign clinical course even at advanced age, probably due to protective genetic factors within TMEM106B or other genes. TMEM106B and PGRN play a critical role in lysosomal processes and they may have a strong functional relationship in the brain. TMEM106B is a 274 amino acid, single-pass, type II transmembrane protein, predominantly located in the membranes of endosomes and lysosomes and highly expressed in neurons, glia, and endothelial cells [23, 54–57]. TMEM106B levels are essential for lysosome size, acidification, function, and transport in different cell types [55, 58–61]. PGRN is a growth factor involved in multiple cellular processes including growth, lysosomal function, and neuroinflammation [10, 11]. In the central nervous system, PGRN modulates inflammation and acts as a neurotrophic and neuroprotective factor and its role is crucial in maintaining lysosomal health [10, 62–65]. We investigated how the coding variant rs3173615 modulates PGRN levels in a subgroup of GRN subjects with the plasma dosage available (n = 100). While the recessive model did not reveal significant differences in plasma PGRN levels, the dominant model showed a significant reduction in PGRN across genotypes; however, this effect did not remain significant after Bonferroni correction. These results are in contrast with the original study of Cruchaga et al., that showed an opposite trend with the lowest levels for the homozygous risk allele and the highest for the homozygous protective allele [20]. Recently, a systematic review including 56 publications consisting of PGRN levels detection using Adipogen assay, which is the same used for our analyses, assessed the association between rs1990622 and protein levels in 3301 individuals. They found no significant differences in PGRN levels among the rs1990622 genotypes in both GRN and GRN non-carriers [30]. Moreover, a study of Meeter et al. in 2016, showed a similar trend as us, with the protective allele C of rs1990622 associated with lower plasma PGRN levels in the entire group of subjects (n = 10 patients GRN carriers, n = 28 pre-symptomatic GRN carriers and n = 29 controls). This association was not maintained when considering only the GRN carrier group. Nevertheless, PGRN levels were measured with a different kit from BioVendor [66]. The interaction between TMEM106B and PGRN in the context of the protective variants remain unclear and more research is required for further elucidations. The presence of multiple SNPs in strong LD at the TMEM106B locus complicates the identification of the causal variant and its role in modulating disease risk. The non-coding variant, rs1990620 (in LD with the previous reported rs1990622), has been proposed to modulate TMEM106B expression through transcriptional activation, potentially through altered long-range chromatin-looping interactions [67]. Meanwhile, the coding variant rs3173615, located in TMEM106B’s fourth N-X-T/S glycosylation motif, may influence TMEM106B protein levels by affecting its stability and degradation rate due to differences in N-glycosylation [23]. A recent hypothesis suggests that the variants could protect against disease preventing or significantly reducing the generation of TMEM106B C-Terminal fragment (CTF) [68]. Consistently, cell biological studies showed that increased TMEM106B levels were cytotoxic and led to an increase in lysosomal size and reduced lysosomal acidification [69]. However, it remains under investigation whether the disease-modifying effect of the TMEM106B protective haplotype is solely due to a reduction of TMEM106B or if also involves a more direct interaction between TMEM106B and PGRN levels or functions, especially in GNR mutations carriers. We acknowledge that this study entails some limitations. First, some individuals homozygous for the protective allele G of the rs3173615 (n = 8) belong to the same family, and thus we accounted for relatedness in the analysis, without substantial changes in the significance. Second, we cannot draw clear-cut conclusions on the relationship between TMEM106B genotype and PGRN levels due to the inability to measure PGRN levels for the entire cohort. Third, unlike the studies of the biggest FTD Consortia studying TMEM106B [50–52], clinical variables beyond age at disease onset, as well as imaging data, were not evaluated. Future studies integrating longitudinal clinical and neuroimaging data will then be essential to further clarify the modifying role of TMEM106B in GRN-associated FTD. Fourth, although our cohort of homozygous carriers represents the largest reported to date, notably since it is derived from a single country (Italy), we acknowledge that an even larger and multinational cohort would further provide a more complete picture. In conclusion, our study confirms that TMEM106B protective variants act as a genetic modifier in the presence of GRN mutations. Due to age at onset modulation effect of the TMEM106B haplotype in GRN mutation carriers, genetic screening for TMEM10B should be included in clinical trials, as suggested by Mirza et al., 2025 [52], and could be considered in clinical practice to optimize risk stratification for FTD. The identification of additional genetic modifiers of GRN-associated FTD has the potential to result in enhanced genetic counselling, and to facilitate the identification of novel targets for disease-modifying treatments. Supplementary information Supplementary Acknowledgements The Molecular Markers Lab researchers and all the authors mourn the sudden loss of Luisa Benussi. Luisa Benussi passed away on 26 June 2023, at the age of 51. She was the PI of the project and we would like to dedicate this study to her. Dear friend and colleague, you are always with us, beside us. Author contributions conceptualization, funding acquisition, and project administration: RG; data curation: LP, CS, AG, AF, AM, BN, RM and RG; formal analysis, methodology, and visualization: LP, CS, AG, and AF; investigation: LP, CS, SF, VL, GG, MSC, AI, SB, AR and GB; resources: GB, BN, RM, BB, and RG; supervision: AM, BN, RM, and RG; writing−original draft preparation: LP, CS, AG, AF and RG; writing−review and editing: SF, VL, GG, MSC, AI, SB, AL, AR, SB, DL, EP, GB, AM, BN, RM and BB. Funding This research was funded by the European Union—Next Generation EU—NRRP M6C2—Investment 2.1 Enhancement and strengthening of biomedical research in the NHS—PNRR-MR1-2022-12375654—Cup Code: C83C22001300001. Title: Genetic and epigenetic modulAtors in Rare neurodegenerative diseases with DEmentia: a National study on autosomal dominant Alzheimer disease and genetic frontotemporal degeneration with dementIA (GARDENIA). Data availability The data presented in this study are available in the Zenodo Data Repository at https://doi.org/10.5281/zenodo.18016695. Competing interests The authors declare no competing interests. Ethics approval and consent to participate Informed consent was obtained from all subjects involved in this study. This study was conducted according to the guidelines of the Declaration of Helsinki and approved by the local Ethics Committee “Comitato Etico IRCCS San Giovanni di Dio Fatebenefratelli” of the IRCCS Centro San Giovanni di Dio Fatebenefratelli, Brescia (Prot. N. 63/2022; date of approval: 7 December 2022). Consent for publication Not applicable. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Lorenzo Pagano, Claudia Saraceno. Contributor Information Claudia Saraceno, Email: csaraceno@fatebenefratelli.eu. Roberta Ghidoni, Email: rghidoni@fatebenefratelli.eu. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41380-026-03653-w. References Associated Data Supplementary Materials Supplementary Data Availability Statement The data presented in this study are available in the Zenodo Data Repository at https://doi.org/10.5281/zenodo.18016695.