Just as birds of a feather flock together, genes involved in common biological pathways often soar and fall in unison, at least when it comes to their expression. If that is so, then shouldn’t a genetic variant that affects one of those co-expressed genes affect them all? This idea led Philip De Jager, Columbia University, New York, and colleagues to search the human genome for such proposed “module quantitative trait loci” or mQTLs. And sure enough, they found some.
- A search for “module quantitative trait loci” nets TMEM106B haplotype.
- The haplotype tracks expression of four gene-expression modules.
- One of them controls lysosomes, myelination, TDP-43 aggregation.
As reported in the June 11 Neuron, mQTLs near the TMEM106B gene seem to alter transcription of four modules of other genes, including one module needed for lysosomes and myelination to work properly. What’s more, the expression changes wrought by these mQTLs ultimately drive accumulation of TDP-43 in the aging brain, the authors conclude.
“This is a very interesting study, which may provide a direct link between TMEM106B/GRN dysfunction and TDP-43 pathology,” wrote Christian Haass, Ludwig-Maximilians University, Munich. “We have made a somewhat similar observation recently, by combining TMEM106B and GRN knockouts. The double knockout led to severe motor defects, microglial activation, lysosomal dysfunction and most importantly to a profound TDP-43 pathology. Therefore, dysregulation of TMEM106B in both directions (up- and downregulation) may cause similar pathological defects,” he wrote.
Rosa Rademakers, University of Antwerp, Belgium, praised the study’s approach but noted that some of its findings are difficult to reconcile with prior work. “This is a very nice study that confirms TMEM106B has a huge impact on the aging brain, but the mechanistic implications need to be tested empirically,” she told Alzforum
TMEM106B and progranulin, encoded by the GRN gene, are intimately intertwined, with protective mutations in the former being able to suppress risk alleles of the latter (Dec 2018 conference news).
In general, geneticists believe that disease-linked risk variants may influence the expression of many distant genes. Alas, finding such remote-acting variants, or trans expression quantitative trait loci, has proven difficult. Potentially, millions of such loci exist in a person’s genome, and scientists must match each against subtle changes in thousands of transcripts. Given the limited availability of cell-specific transcriptomes from aging human brains, linking a trans-eQTL to a neurodegenerative disease becomes a tall task.
First author Hyun-Sik Yang, Brigham and Women’s Hospital, Boston, and colleagues took a reductionist approach, building on some of de Jager’s previous work on gene modules. Using tissue samples from the Religious Orders Study and Memory and Aging Project (ROSMAP) run by co-authors David Bennett and Julie Schneider at Rush University in Chicago, De Jager had previously linked a network of gene co-expression modules to disease pathology and cognitive decline (Jun 2018 news).
In the new work, Yang and colleagues ran genome-wide association analyses to discover variants in the genome that affect expression of those 47 distinct modules. They then repeated the analysis using two replication datasets. One came from single-nucleus RNA-Seq of samples from the Mayo Clinic and Banner Sun Health brain banks, the other from an RNA-Seq analysis of samples from ROSMAP that were not in the primary dataset.
Module QTLs. Circular Manhattan plot shows summary data of genome-wide association with expression of five modules. Expression of four modules—m16 (red); m17 (green); m18 (gray); m110 (blue)—associated with single-nucleotide polymorphisms in the TMEM106B locus. A fifth—m234 (turquoise)—associated with an SNP in the RBFOX1 gene. Outer red circle denotes statistical threshold corrected for 47 multiple module comparisons. Blue circle represents significance threshold for a single GWAS. [Courtesy of Yang et al., Neuron, 2020.]
The analysis identified five mQTLs affecting expression of five modules. One mQTL turned up in the RBFOX1 gene. There are hints that expression of this RNA-binding protein falls in AD (Oct 2017 news). De Jager told Alzforum that, coincidentally, researchers in Richard Mayeux and Christiane Reitz’s labs at Columbia have linked variants in this gene to amyloid PET signals and to AD in African Americans, respectively. For this study, however, De Jager and colleagues the researchers focused on the other four mQTLs.
All reside at the TMEM106B locus and are part of the same haplotype, tagged by the rs1990622A allele. “This was quite surprising,” said De Jager. This one region seems to affect expression of more than 1,000 genes. Two other modules also seemed affected by this haplotype but missed statistical significance. Other researchers had previously linked this haplotype to frontotemporal dementia and cognitive resilience; De Jager’s group had linked it to Limbic-predominant Age-related TDP-43 Encephalopathy, aka LATE (van Deerlin et al., 2010; White et al., 2017; Feb 2015 news). Researchers introduced the concept of LATE to account for the predominantly limbic TDP-43 pathology that is common among those older than 80. LATE shares clinical characteristics with AD and is proposed to account for 20 percent of AD diagnoses in that age group (May 2019 news).
How does the TMEM106B haplotype lead to disease? Researchers led Rademakers and Carlos Cruchaga at Washington University, St. Louis, had tied this haplotype to reduced numbers of neurons and increased numbers of microglia and other brain cell types based on RNA markers (Ren et al., 2018; Li et al., 2020).
De Jager believes the answer lies elsewhere; Yang did find RNA changes consistent with cell number changes in the ROSMAP data. However, taking advantage of detailed immunohistochemistry from this study, Yang was able to count neurons, microglia, and astrocytes directly and found no difference between carriers and noncarriers of this haplotype. The authors concluded that the expression changes do not capture cell numbers.
Instead, the haplotype seems to drive aggregation of TDP-43. In ROSMAP, one of the four modules regulated by TMEM106B—m110—associated with this pathology. It contains genes involved in lysosomal biology and myelination. Do module changes come before the pathology, or the other way around? A mediation analysis puts m110 expression first. Increased expression of the module explained the association of the TMEM106B haplotype with LATE pathology in ROSMAP, suggesting that TMEM drives changes in the module, which then promote TDP-43 aggregation.
Curiously, De Jager and others had linked ApoE4 to LATE TDP-43 pathology (Yang et al., 2018; Wennberg et al., 2018). In another mediation analysis, Yang found that ApoE4, by increasing amyloid deposition, drives m110 expression. This would imply that TMEM106B and ApoE4/Aβ similarly affect the module. “Our findings suggest that dysregulated myelination and lysosomal dysfunction are shared molecular events between Alzheimer’s and LATE neuropathological change, which could shed light on why these two common causes of dementia are often combined in individuals of advanced age,” Yang wrote to Alzforum.” De Jager thinks this was an important finding. “It suggests a key transit point between risk factors and pathology. Perhaps perturbing genes in this module or correcting their expression could be one way to tackle disease,” he said.
Where does progranulin fit into all this? The researchers found evidence that trans-eQTLs in the GRN locus affect expression of the same modules, albeit that data was not as robust. The TMEM haplotype also increased expression of progranulin, but the progranulin eQTL did not increase expression of TMEM106B. The researchers concluded that TMEM is mechanistically upstream of GRN in their shared pathological pathway.
Rademakers thought this was a curious finding. “Patients with progranulin mutations are known to have increased TMEM106B levels and mouse studies also reported an increase in TMEM106B levels upon progranulin loss,” she told Alzforum. “What they find might be based on a [GRN] variant that is not strong enough to evoke the kind of lysosomal stress we see in disease.”
She also emphasized that the findings are based on transcript expression but not protein, which may fluctuate independently. De Jager said that looking into change in protein levels would be the next important step.—Tom Fagan
- 11th ICFTD Meeting in Sydney Sorts Out Clinical Subtypes
- Culling Connection From Chaos, Alzheimer’s Genetic Network Study Pins PLXNB1 and INPPL1
- Estimates of RNA Decay Hint at Destabilization in Alzheimer’s Brains
- FTLD Gene Bad Actor in Many TDP-43 Proteinopathies
- Introducing LATE—A Common TDP-43 Proteinopathy that Strikes After 80
- Van Deerlin VM, Sleiman PM, Martinez-Lage M, Chen-Plotkin A, Wang LS, Graff-Radford NR, Dickson DW, Rademakers R, Boeve BF, Grossman M, Arnold SE, Mann DM, Pickering-Brown SM, Seelaar H, Heutink P, van Swieten JC, Murrell JR, Ghetti B, Spina S, Grafman J, Hodges J, Spillantini MG, Gilman S, Lieberman AP, Kaye JA, Woltjer RL, Bigio EH, Mesulam M, Al-Sarraj S, Troakes C, Rosenberg RN, White CL, Ferrer I, Lladó A, Neumann M, Kretzschmar HA, Hulette CM, Welsh-Bohmer KA, Miller BL, Alzualde A, Lopez de Munain A, McKee AC, Gearing M, Levey AI, Lah JJ, Hardy J, Rohrer JD, Lashley T, Mackenzie IR, Feldman HH, Hamilton RL, Dekosky ST, van der Zee J, Kumar-Singh S, Van Broeckhoven C, Mayeux R, Vonsattel JP, Troncoso JC, Kril JJ, Kwok JB, Halliday GM, Bird TD, Ince PG, Shaw PJ, Cairns NJ, Morris JC, McLean CA, Decarli C, Ellis WG, Freeman SH, Frosch MP, Growdon JH, Perl DP, Sano M, Bennett DA, Schneider JA, Beach TG, Reiman EM, Woodruff BK, Cummings J, Vinters HV, Miller CA, Chui HC, Alafuzoff I, Hartikainen P, Seilhean D, Galasko D, Masliah E, Cotman CW, Tuñón MT, Martínez MC, Munoz DG, Carroll SL, Marson D, Riederer PF, Bogdanovic N, Schellenberg GD, Hakonarson H, Trojanowski JQ, Lee VM. Common variants at 7p21 are associated with frontotemporal lobar degeneration with TDP-43 inclusions. Nat Genet. 2010 Mar;42(3):234-9. PubMed.
- White CC, Yang HS, Yu L, Chibnik LB, Dawe RJ, Yang J, Klein HU, Felsky D, Ramos-Miguel A, Arfanakis K, Honer WG, Sperling RA, Schneider JA, Bennett DA, De Jager PL. Identification of genes associated with dissociation of cognitive performance and neuropathological burden: Multistep analysis of genetic, epigenetic, and transcriptional data. PLoS Med. 2017 Apr;14(4):e1002287. Epub 2017 Apr 25 PubMed.
- Ren Y, van Blitterswijk M, Allen M, Carrasquillo MM, Reddy JS, Wang X, Beach TG, Dickson DW, Ertekin-Taner N, Asmann YW, Rademakers R. TMEM106B haplotypes have distinct gene expression patterns in aged brain. Mol Neurodegener. 2018 Jul 3;13(1):35. PubMed.
- Li Z, Farias FH, Dube U, Del-Aguila JL, Mihindukulasuriya KA, Fernandez MV, Ibanez L, Budde JP, Wang F, Lake AM, Deming Y, Perez J, Yang C, Bahena JA, Qin W, Bradley JL, Davenport R, Bergmann K, Morris JC, Perrin RJ, Benitez BA, Dougherty JD, Harari O, Cruchaga C. The TMEM106B FTLD-protective variant, rs1990621, is also associated with increased neuronal proportion. Acta Neuropathol. 2020 Jan;139(1):45-61. Epub 2019 Aug 27 PubMed.
- Yang HS, Yu L, White CC, Chibnik LB, Chhatwal JP, Sperling RA, Bennett DA, Schneider JA, De Jager PL. Evaluation of TDP-43 proteinopathy and hippocampal sclerosis in relation to APOE ε4 haplotype status: a community-based cohort study. Lancet Neurol. 2018 Sep;17(9):773-781. Epub 2018 Aug 6 PubMed.
- Wennberg AM, Tosakulwong N, Lesnick TG, Murray ME, Whitwell JL, Liesinger AM, Petrucelli L, Boeve BF, Parisi JE, Knopman DS, Petersen RC, Dickson DW, Josephs KA. Association of Apolipoprotein E ε4 With Transactive Response DNA-Binding Protein 43. JAMA Neurol. 2018 Nov 1;75(11):1347-1354. PubMed.
- Yang HS, White CC, Klein HU, Yu L, Gaiteri C, Ma Y, Felsky D, Mostafavi S, Petyuk VA, Sperling RA, Ertekin-Taner N, Schneider JA, Bennett DA, De Jager PL. Genetics of Gene Expression in the Aging Human Brain Reveal TDP-43 Proteinopathy Pathophysiology. Neuron. 2020 Aug 5;107(3):496-508.e6. Epub 2020 Jun 10 PubMed.