Structural Insights Into TDP-43 and Effects of Post-translational Modifications

Abstract

Transactive response DNA binding protein (TDP-43) is a key player in neurodegenerative diseases. In this review, we have gathered and presented structural information on the different regions of TDP-43 with high resolution structures available. A thorough understanding of TDP-43 structure, effect of modifications, aggregation and sites of localization is necessary as we develop therapeutic strategies targeting TDP-43 for neurodegenerative diseases. We discuss how different domains as well as post-translational modification may influence TDP-43 overall structure, aggregation and droplet formation. The primary aim of the review is to utilize structural insights as we develop an understanding of the deleterious behavior of TDP-43 and highlight locations of established and proposed post-translation modifications. TDP-43 structure and effect on localization is paralleled by many RNA-binding proteins and this review serves as an example of how structure may be modulated by numerous compounding elements.

Keywords: RRM domain; TDP-43 = TAR DNA–binding protein 43; post-translational modification; structure; subdomains.

FIGURE 1

TDP-43 Structure and Sequence Features. (A) Domain map showing relative sizes of the N-terminal domain (NTD, orange), linker containing nuclear localization signal (NLS, blue), RNA-recognition motif 1 (RRM1, yellow), RNA-recognition motif 2 (RRM2, green), and the C-terminal subdomains identified by Mompeán et al. (2016a): Gly-aromatic-Ser-rich regions (GaroS, light gray), hydrophobic region (Φ, dark gray), glutamine-arginine-rich region (Q/N, orange gray). (B) Representative structures with variant sites shown as spheres and colored as in panels A and C. NTD (5mrg), RRM1-RRM2 with RNA as gray sticks (4bs2), CTD fragments i [6n3c (Cao et al., 2019)], ii [2n3x (Jiang et al., 2016)], iii [6n3a (Guenther et al., 2018a)], iv [5wia (Guenther et al., 2018a)], v [5wiq (Guenther et al., 2018a)]. Asterisk marks LARKS/Omega Loop. Arrows indicate residues that correspond among three shown polymorphs. Ellipses indicate tendency to form fibrils (i, iii) or steric zipper structures (iv, v). (C) Annotated primary sequence. Post-translational modifications shown in circles above sequence (P, phosphorylation site PhosphoSitePlus (Hornbeck et al., 2015)]; S, SUMOylation sites identified by high-throughput studies (Hendriks et al., 2015, 2017); CH3, monomethylation site (Guo et al., 2014). Alternative start site at M85 [alt (Nishimoto et al., 2010)]. Caspase digestion sites [lightning bolts (Zhang et al., 2007; Rohn, 2008; Li et al., 2015; Chiang et al., 2016)]. Proposed mitochondrial targeting sequences (M1, M3, and M5) (Wang et al., 2016). Bipartite nuclear localization signal (NLS) (Winton et al., 2008) overlapping with the Poly-ADP-ribose (PAR) binding motif (PARBM) (McGurk et al., 2018). Conserved ribonucleotide interacting motifs (RNP1, RNP2) (Buratti et al., 2001). Amyloidogenic regions (amylo) (Shodai et al., 2012, 2013). Non-functional (former) nuclear export signal [f-nes (Archbold et al., 2018; Ederle et al., 2018; Pinarbasi et al., 2018)]. Sequence variants compiled using the ClinVar database (April 2019) (Landrum et al., 2016) and a review by Buratti et al. (Buratti, 2015), shown in red boxes if linked to disease, orange if clinical significance is not clear. Site of insertion/deletion [triangle (Solski et al., 2012)] and premature stop (X) also shown. Dots above sequence mark every 10 residues.

FIGURE 2

NTD Structures. (A) Monomeric structure shown in ribbon representation with location of mitochondrial targeting sequence 1 in black, residues involved in loop-sheet transition during dimer formation in yellow, Ser 48 in sticks representation, residues 97–102 in blue with A90 as a small orange sphere. (B) Dimeric X-ray structure with features as in A. (C) Overlay of monomer (gray) and subunit A of dimeric structure (orange) showing shift of S48 Cα and loop to β-strand transition (yellow). (D) Inter-subunit interactions of Ser 48 in dimeric X-ray (left) and NMR (right) structures. Dashes show contacts within hydrogen bonding distance. (E) NMR ensemble of NTD monomer showing varied position of the NLS. Coloring as in panel (A). Monomeric NMR structure PDB ID 5mrg (Mompeán et al., 2017), dimeric X-ray structure PDB ID 5mdi (Afroz et al., 2017) and dimeric NMR structure PDB ID 6b1g (Wang A. et al., 2018).

FIGURE 3

Structure of TDP-43 RRM domains. (A) Overview showing RRM1 in yellow with RNP-1 and RNP-2 in orange, flexible linker in dark gray, RRM2 in green with RNP-1 and RNP-2 in cyan and UG-rich RNA in light gray. (B) Rotated view of amyloidogenic core residues 166–173. (C) Overlap between amyloidogenic core residues 246–255 (red in view A) with offset showing former NES in purple. (D) Rotated view of cleavage site D219 [PDB ID 4bs2 (Lukavsky et al., 2013)].

FIGURE 4

Prediction of amyloidogenicity of TDP-43 and TDP-43-”ΔNES.” Those mutations reduce the aggregation prone characteristic of the 244–255 region.

FIGURE 5

Structures of the C-terminal domain. Structures shown with mutation sites as sticks and phosphorylation sites with dots at the SG atom. Simplified primary sequence annotations as in Figure 1. Structure 2n2c (Lim et al., 2016) is highly similar to that of 2n3x (Jiang et al., 2016), thus only a portion is shown here, highlighting the non-overlapping residues M307-G310. Representatives of the short segments that form steric zippers or LARKS (5wkd, 5whn, 6cew, 6cb9, 6cfh, 5wia, 5wiq) (Guenther et al., 2018a) are shown in darker gray with neighboring strands as determined by symmetry in lighter gray. The R-shaped polymorph (6n3c) (Cao et al., 2019) with chain A in darker gray and only three filaments and immediate adjacent layers shown for clarity. A representative dagger-shaped fibril-forming polymorph (6n3a) (Cao et al., 2019) is shown with chain B in darker gray with two filaments and immediate adjacent layers. Note different shapes of LARKS segments within 6n3c and 6n3b (thin boxes). For clarity, mutation or phosphorylation sites shown only on the primary strand in all images.

FIGURE 6

Oxidation of Cysteine residues in TDP-43. (A) Distance and localization of C39 and C50 in TDP-43 NTD (PDB code: 5mrg). (B) Distance between Cys173 and Cys175 and between Cys198 and Cys244 (C) in RRM1-2 (PDB code: 4bs2). (D) Table showing accessibility of cysteine residues (Cys39 and Cys50 were calculated using 5mrg, Cys173, 175, 198, and 244 were calculated using 4bs2). The prediction, using the online tool RSCP, validated a residue as likely to be oxidized when the decision value was above the threshold (0.7). (E) Accessibility of the cysteine residues calculated with Areaimol as implemented in the CCP4 suite (Otwinowski and Minor, 1997).

FIGURE 7

Acetylation of TDP-43 Lysines. Lys 145 in RRM1 (orange) (A) and Lys 192 in RRM2 (green) (B) shown in Cohen et al. (2015) to be acetylated are located at the interface with nucleic acid (RNA, gray) PDB ID 4bs2 (Lukavsky et al., 2013). (C) Accessibility of the lysine residues shown to be acetylated calculated using Areaimol as implemented in the CCP4 suite (Otwinowski and Minor, 1997). (D) Prediction of acetylated lysines by different acetylation systems, using ASEB (http://bioinfo.bjmu.edu.cn/huac/; Li et al., 2012, 2013; Zhai et al., 2017) shown as a heat map of p-values. For each query, a p-value is assigned based on its similarity to known sites. The smaller the p-value of the residue, the closer it is to known acetylated sites.

FIGURE 8

SUMOylation of lysine residues in TDP-43. (A) TDP-43 sites that conform to SUMOylation consensus motifs (ranked by Abgent SUMOplot^TM Analysis Program). Higher scores indicate greater similarity to positively identified SUMOylation motifs and higher predicted likelihood of endogenous SUMOylation. (B) Accessibility of the SUMO motifs at the TDP-43 surface were calculated using Areaimol as implemented in the CCP4 suite (Otwinowski and Minor, 1997) on the NTD structure [PDB ID 5mrg (Mompeán et al., 2017), K84 motif] or the free form of the tethered RRM domains structure [PDB ID 4bs2 (Lukavsky et al., 2013), K136, K140, K192, and K224 motifs]. To note, there is no significant difference in accessibility values when RNA is present. Mapping of the SUMOylation motifs on the surface of TDP-43 NTD (C) or RRM1-2 with and without RNA (D).

FIGURE 9

Ubiquitination of lysines in TDP-43. (A) Accessibility of the known ubiquitinated Lysine residues were calculated using Areaimol as implemented in the CCP4 suite (Otwinowski and Minor, 1997) on the NTD structure [PDB ID 5mrg (Mompeán et al., 2017), K84 and K95] or the tethered RRM domains structure [PDB ID 4bs2 (Lukavsky et al., 2013), K160, K181, and K263]. Mapping of the SUMOylation motifs on the surface of TDP-43 NTD (B) or RRM1-2 with and without RNA (C,D).

FIGURE 10

Mitochondrial targeting sequences in TDP-43. (A) M1 in the NTD monomeric structure [PDB ID 5mrg (Mompeán et al., 2017)]. (B) M1 in the dimeric structure [PDB ID 5mdi (Afroz et al., 2017)] (C) M3 in RRM1, overlapping with RNP-1 (orange) [PDB ID 4bs2 (Lukavsky et al., 2013)]. Location of mitochondrial targeting signals shown in black. RNA is in light gray.