bioRxiv preprint doi: https://doi.org/10.1101/243014; this version posted January 4, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Mixed Charge RNA binding proteins aggregate in AD RNA-binding proteins with mixed charge domains self-assemble and aggregate in Alzheimer's Disease* Isaac Bishof1,4, Eric B. Dammer1,4, Duc M. Duong1,4, Marla Gearing3,4, James J. Lah2,4, Allan I. Levey2,4, and Nicholas T. Seyfried1,2,4# 1Department of Biochemistry, 2Department of Neurology, 3Department of Pathology and Laboratory Medicine, 4Center for Neurodegenerative Diseases, Emory University School of Medicine, Atlanta, GA, 30322 #Address correspondence to: Nicholas T. Seyfried, Department of Biochemistry, Emory University School of Medicine, 1510 Clifton Road, Atlanta, Georgia 30322, USA. Tel. 404.712.9783, Email: [email protected] *Running title: Mixed Charge RNA binding proteins aggregate in AD Keywords: protein-protein interaction, protein aggregation, systems biology, Tau protein, neurodegeneration, mass spectrometry, proteomics, RNA-binding protein, intrinsically disordered protein, RNA processing ABSTRACT on the U1-70K LC domain(s) for their U1 small nuclear ribonucleoprotein 70 kDa interaction. This included structurally similar (U1-70K) and other RNA binding proteins RBPs, such as LUC7L3 and RBM25, which (RBPs) are mislocalized to cytoplasmic require their respective mixed charge domains neurofibrillary Tau aggregates in Alzheimer’s for reciprocal interactions with U1-70K and disease (AD), yet understanding of the for participation in nuclear RNA granules. mechanisms that cause their aggregation is Strikingly, a significant proportion of RBPs limited. Many RBPs that aggregate in with mixed charge domains have elevated neurodegenerative diseases self-assemble into insolubility in AD brain proteome compared RNA granules through intrinsically to controls. Furthermore, we show that the disordered low complexity (LC) domains. We mixed charge LC domain of U1-70K can report here that a LC domain within U1-70K interact with Tau from AD brain. These of mixed charge, containing highly repetitive findings highlight mechanisms for mixed complementary repeats of basic (R/K) and charge domains in stabilizing RBP acidic (D/E) residues, shares many of the same interactions and in potentially mediating co- properties of the Q/N-rich LC domains found aggregation with pathological Tau isoforms in in the RBPs TDP-43 and FUS. These AD. properties include the ability to self-assemble INTRODUCTION into oligomers, and to form nuclear granules. To analyze the functional roles of the U1-70K The molecular processes that contribute LC domains, we performed co- to neurodegenerative diseases are not well immunoprecipitation and quantitative mass understood. Recent observations suggest that spectrometry analysis of recombinant U1-70K numerous neurodegenerative diseases are and deletions lacking the C-terminal LC promoted by the accumulation of RNA-binding domain(s). A network-driven approach protein (RBP) aggregates (1-3). This includes resolved functional classes of U1-70K Alzheimer’s disease (AD), where pathological interacting proteins that showed dependency RNA-protein aggregates are often, but not exclusively, associated with Tau neurofibrillary 1 bioRxiv preprint doi: https://doi.org/10.1101/243014; this version posted January 4, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Mixed Charge RNA binding proteins aggregate in AD tangles in brain (4). For example, U1 small was sufficient for robust aggregation, and nuclear ribonucleoprotein 70 kDa (U1-70K) and through cross-linking studies was found to other core components of the spliceosome directly interact with insoluble U1-70K in AD complex form detergent-insoluble aggregates in brain homogenates (5). Collectively, these both sporadic and familial human cases of AD (5- observations led to a hypothesis that pathological 7). Furthermore, RNA-seq analysis from AD and aggregation of RBPs in neurodegenerative control brains revealed a significant diseases, including U1-70K, is driven by LC accumulation of unspliced pre-mRNA disease domains. However, unlike the prion-like Q/N- related transcripts in AD consistent with a loss of rich LC domains of TDP-43 and FUS, the LC1 U1-spliceosome function (7,8). Currently, our domain of U1-70K contains highly repetitive knowledge of the specific mechanisms complementary basic (R/K) and acidic (D/E) underlying U1-70K aggregation is limited. This residues. These structurally unique motifs were has proved to be a barrier to developing cellular originally described by Perutz (27), who models that would further our understanding of proposed their ability to self-assemble and form U1-70K and related RBP aggregation events in higher-order structures termed polar zippers (27). the pathogenesis of AD. Currently the physiological role of polar zipper Supporting evidence indicates that a motifs in U1-70K and other RNA binding select group of RBPs are poised for aggregation proteins is unclear, and understanding their role because they self-assemble to form structures, in protein-protein interactions may shed light on including RNA granules (9), which are the mechanisms underlying RBP aggregation and membrane-free organelles composed of RNA and their association with Tau in AD (7,28). RBPs (1,9-12). It has been proposed that RNA Here we report that the mixed charge granules form via liquid–liquid phase separation LC1 domain of U1-70K shares many of the same (LLPS), which is driven by a dynamic network of properties of the Q/N-rich LC domains found in multivalent interactions between structurally TDP-43 and FUS, despite having a vastly disordered low complexity (LC) domains (13-15) different amino acid composition. These that have limited diversity in their amino acid properties include the ability to self-assemble into composition (16). LLPS allows specific RBPs to high molecular weight oligomers and associate concentrate and separate, leading to the formation with nuclear granules in cells. To analyze the of higher-order structures including oligomers, functional roles for the LC domains in U1-70K, granules, and ultimately aggregates (16-18). we performed co-immunoprecipitation of Notably, several RBPs that aggregate in recombinant U1-70K and serial deletions lacking neurodegenerative disease contain LC domains, one or both LC domains followed by quantitative including TDP-43 and FUS (19-21). The LC proteomic analysis. Using a network-based domains found in TDP-43 and FUS mediate self- bioinformatic approach we mapped classes of association, are necessary for RNA granule U1-70K interacting proteins that showed a formation, and polymerize into amyloid-like dramatic reduction in their association with U1- aggregates (9,22,23). Mutations harbored within 70K in the absence of the LC1 domain. the LC domains of TDP-43 and FUS cause Remarkably, this revealed a group of functionally amyotrophic lateral sclerosis (ALS) and and structurally similar RBPs that also contained increased RNA granule stability, highlighting a mixed charge domains analogous to the LC1 critical role for LC domains in disease domain in U1-70K. These included LUC7L3 and pathogenesis (24-26). RBM25, which we confirm also require their We recently reported that human AD respective mixed charge domains for reciprocal brain homogenates induced the aggregation of interactions with U1-70K and for proper nuclear soluble U1-70K from control brain and RNA granule association in cells. Furthermore, recombinant U1-70K, rendering it detergent- global analysis of the AD detergent-insoluble insoluble (5). The C-terminus of U1-70K, which proteome revealed elevated levels of mixed harbors two LC domains (LC1 and LC2) was charge RBPs within AD brain compared to necessary for this aggregation (5). Furthermore, controls. Finally, we show that the LC1 domain the LC1 domain (residues 231-308) of U1-70K of U1-70K can interact with Tau from AD brain, 2 bioRxiv preprint doi: https://doi.org/10.1101/243014; this version posted January 4, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Mixed Charge RNA binding proteins aggregate in AD which supports a hypothesis that mixed charge U1-70K in cells, it was unclear whether this structural motifs on U1-70K and related RBPs interaction is direct or facilitated by indirect could mediate cooperative interactions with Tau interactions with additional RBPs. To in AD. determine if the LC1 of U1-70K can directly self-associate, we performed blue native gel RESULTS polyacrylamide gel electrophoresis (BN- PAGE) of the GST-purified LC1 (residues 231- The LC1 domain of U1-70K is necessary and 310) and N-terminal domain (residues 1-99) of sufficient for self-association in cells rU1-70K; the latter was unable to interact with The U1-70K LC1 domain (residues 231- native U1-70K in cells (Fig. 1B). In contrast to 308) is necessary and sufficient for robust SDS-PAGE, which resolves proteins under aggregation in AD brain homogenates (5), yet it denaturing conditions, BN-PAGE is used to remains unknown if this domain is required for determine native protein complex masses, endogenous U1-70K self-association under including high molecular weight oligomeric physiological conditions. To test this states and to identify physiological protein– hypothesis, we over-expressed full-length protein interactions (29). Under the denaturing recombinant GST-fused and Myc-tagged U1- conditions of SDS-PAGE (Fig. 2A), both the 70K (rU1-70K) in HEK293 cells with serial LC1 and N-terminal domain have equivalent deletions lacking one or both LC domains molecular weights (~65 kDa) compared to followed by co-immunoprecipitation (co-IP) purified GST (~20 kDa).