TDP-43 Epigenetic Facets and Their Neurodegenerative Implications

Abstract

Since its initial involvement in numerous neurodegenerative pathologies in 2006, either as a principal actor or as a cofactor, new pathologies implicating transactive response (TAR) DNA-binding protein 43 (TDP-43) are regularly emerging also beyond the neuronal system. This reflects the fact that TDP-43 functions are particularly complex and broad in a great variety of human cells. In neurodegenerative diseases, this protein is often pathologically delocalized to the cytoplasm, where it irreversibly aggregates and is subjected to various post-translational modifications such as phosphorylation, polyubiquitination, and cleavage. Until a few years ago, the research emphasis has been focused particularly on the impacts of this aggregation and/or on its widely described role in complex RNA splicing, whether related to loss- or gain-of-function mechanisms. Interestingly, recent studies have strengthened the knowledge of TDP-43 activity at the chromatin level and its implication in the regulation of DNA transcription and stability. These discoveries have highlighted new features regarding its own transcriptional regulation and suggested additional mechanistic and disease models for the effects of TPD-43. In this review, we aim to give a comprehensive view of the potential epigenetic (de)regulations driven by (and driving) this multitask DNA/RNA-binding protein.

Keywords: ALS; DNA repair; FTD/FTLD; TARDBP; TDP-43; chromatin; epigenetics; neurodegeneration; retrotransposon; transcriptional regulation.

Figure 3

TDP-43 connections to RTE detrimental action in ALS. (A) Overview of the general consequences of TDP-43 alterations in DNA stability and cellular homeostasis due to RTE inhibition failure. (B–E) TDP-43 impact on ALS through the misregulation of RTE. In the cartoons, regulatory sequences (promoter regions), i.e., LTRs in exogenous and endogenous ERVs, and 5′UTR in Lines, are in blue; retroviral gag, pro, and pol genes produced from the same (polycistronic) transcript are in white; and the env gene produced by an alternatively spliced transcript is in purple. ORFs for accessory proteins (Tat in HIV, rec and np9 or rec in HERV-K) are depicted. Specificities of each class/specific elements and impact on ALS are listed on the right of each cartoon. (B) TDP-43 is able to repress HIV-1 provirus activation. TDP-43 binds to the TAR binding site within the 5′LTR R region and represses transcription. This binding was shown to impede the binding of TAR RNA and Tat-activating protein. A reduction in TDP-43 binding (by ubiquitination and proteosomal degradation) can reverse HIV-1 provirus latency, potentially leading to the production of infectious viral particles (HIV virion). It is known that HIV-1 can promote ALS-like symptoms. HIV can also activate HERV-K elements, notably via Tat [155]. (C) TDP-43 overexpression binds to and activates specific HERV-K HML-2 provirus(es), producing the toxic env glycoprotein HERV(HML-2). The HERV-K proviruses found activated in ALS cases (ALS neurons shows immunoreactivity for HERV-K Env) could be HERV-K C7-C and HERV-K C10-A, which are polymorphic proviruses in the human population. TDP-43 binds to the LTR5Hs sequence at a polypyrimidine track in the U3 region (5′-CCCTCTCCC-3′) with high affinity and is able to activate the LTR-promoted transcription. Conversely, HERV-K Env potentially triggers TDP-43 upregulation. (D) In Drosophila, the failure of TDP-43 to indirectly repress Gypsy retrovirus, a family of endogenous LTR-retrotranposon, leads to cell autonomous and non-autonomous toxicity. The family contains copies with preserved ORFs, capable of retrotransposition and replication. TDP-43 alterations (hTDP-43 overexpression or fly TDP-43 homologue TDPH null) induce the activation, specifically in glial cells, of Gypsy copies. It is not clear if replication involves infectious or non-virus-like particles (VLP). Mechanistically, TDP-43 binds to and positively regulates Dicer-2 (Dicer in the human) mRNA and protein. Dicer-2 in the RISC complex controls Gypsy and other RTE activity via the endo-siRNA pathway, inhibiting or impeding transduction by promoting mRNA degradation. Lack of TDP-43 by reducing Dicer-2 levels impedes endo-siRNA-mediated control (see text for more detail). (E) TDP-43 nuclear loss in the neurons of ALS patients induces L1 expression. TDP-43 represses human L1 by at least two mechanisms: (i) by binding at the 5′UTR promotes and maintaining L1 heterochromatinization; and (ii) by interacting with the ORF2p in cases of L1 retrotransposition, inhibiting the pasting of new copies into the host genome.

Figure 1

TDP-43-mediated transcriptional regulation. Genes for which TDP-43 has been shown to regulate the transcription by acting at the promoter level are illustrated in their context. (a–e) Transcriptional repression, involving the direct binding of TDP-43 to the target DNA regulatory region. (a) TDP-43 binding on Acrv1 promoter via two GTGTGT-motifs controls the production of Sp-10 protein during mouse spermatogenesis. TDP-43 at Acrv1 promoter is still observed when histones acquire activating modifications (H3K9Ac, H3K4me3, increases in RNA-pol II) and transcription starts in spermatids. In the liver, TDP-43 binding and inactive chromatin mark H3K9me2 associates with Acrv1 inhibition (adapted from [40]). (b) Repressive potential of TDP-43 on the c-fos promoter. Tethering of TDP-43 to a reporter plasmid using Gal4 DNA Binding Domain (DBD), fused to TDP-43 at Gal4 binding sequences (blue boxes), upstream of the c-fos promoter, represses the promoter-induced luciferase expression (adapted from [40]). (c) In neurons, TDP-43 represses the promoter of VSP4B, ensuring recycling endosome transport. The repression occurs via the binding of TDP-43 at a GT-rich region less than 1 kb before VPS4B TSS. Loss of TDP-43 derepresses the VPS4B promoter, leading to loss of dendrites and dendritic spines (adapted from [41]). (d) TDP-43 contributes to the supplementary X inactivation (Xi) and X-linked genes in females. TDP-43 interacts with Xist RNA in female cells together with other Xist RNA binding proteins: PTBP1, MATR3, or CELF1. The TDP-43 strongest binding within Xist occurs at the 3′ end of the E-repeat containing multiple (GU)n tracts and persists after completion of X inactivation. Depletion of TDP-43 induces significant nuclear dispersal of Xist and defects in DNA compaction (adapted from [44]). (e) TDP-43 binds to a short 40 bp region located from −200 to −160 of Cyp8b1 promoter in liver and represses its expression. The decrease in Cyp8b1 results in the activation of FXR and an increase in apoC2 levels and diffusion, resulting in enhanced triglyceride (TG) clearance in several mice tissues (muscle, heart, and adipose cells). lncLSTR, a liver-specific lncRNA, binds TDP-43 protein and impedes its binding onto Cyp8b1 promoter, consequently counteracting TG clearance (adapted from [45]). (f–h) Transcriptional activation. (f) TDP-43 binds to and activates the TNF-alpha promoter at an LPS-sensitive binding site, located −550 to −487, and mediates the activation of Thd1 macrophage-like. siRNA against TDP-43 reduces the LPS induction of TNF-alpha by 50% (adapted from [46]). (g) TDP-43 is a direct transcriptional activator of the CHOP/GADD153 promoter in SH-SY5Y, provoking cell death. Binding within the CHOP promoter potentially occurs in a region comprised within the bp −300 and −30 from the TSS. TDP-43 also increases CHOP mRNA stability. Acetylation of TDP-43 at lysine 145 and 192 impedes TDP-43 activation of the CHOP promoter (adapted from [48]). (h) During C2C12 differentiation, TDP-43 is tethered by the muscle-enriched lncRNA Myolinc to the promoter of several genes linked to the differentiation of myoblasts into myocytes, such as Acta1, MyoD1, Filip1, and others (adapted from [42]). (i) Circplot-like summary of the different modalities by which TDP-43 regulates gene expression. TDP-43 can act at “single” or “multiple” targets “functionally” (e.g., the myogenesis pathway) or “spatially” (chromosome X) related. It can be “repressive” or “activating”, involving lncRNAs acting either by “evicting” TDP-43 or tethering it, thus acting as a “scaffold”. Generally, direct binding of TDP-43 on its target’s promoter has been demonstrated. The dependence for DNA binding on GT-rich sequences (“GT-rich”) or not (“Not GT”), when known, is shown, but is has not always been specified (“?”).

Figure 2

TDP-43-mediated DNA repair: direct and indirect roles. Schematic diagrams of TPD-43 role in DSB repair. (A–C) Direct role of TDP-43 in DSB repair and its misregulation in ALS. (A) TDP-43 in DNA double-stranded break (DSB) repair. TDP-43 interaction with activated (phosphorylated) DNA damage repair (DDR) response factors (p-ATM, p-53BP1, p-H2AX = yH2AX) facilitates the NHEJ repair in neurons by supporting the recruitment and activity of the XLF/XRCC4/Lig4 complex [23]. (B) TDP-43 in transcription-coupled DSB (TC-DSB) repair. TDP-43 interacts with several key factors in the transcription-coupled repair (i.e., DHX9, COPS3/4, AQR, RFC, PARP1, XRCC1, TDP1, APEX1, Ku70/80, and condensin SMC3), and binds non-blocked dsDNA ends such as those created at DSB. It is also probably involved in the transcriptional silencing following DSB through its recruitment of SIRT-2 and subsequent H3K18 deacetylation, as evidenced in HeLa and MEFs cells [97] and in post-mitotic neuronal cells. The hypothetical recruitment of RNA helicase DDX5 with TDP-43 at R-loop by the Lnc530 to resolve their aberrant formation is still to be investigated in neurons and in the human. (C) TDP-43-related genome damage in ALS motor and cortical neurons and in differentiated neuronal SH-SY5Y cells. In presence of a mutant or mislocalized TDP-43, yH2AX levels are reduced, and the NHEJ complex (XLF/XRCC4/Lig4) is not recruited to damage sites for repair, resulting in an accumulation of damaged DNA and leading to neurodegeneration. (D) TDP-43 regulation of genes impacting DNA damage and repair. TDP-43 positively regulates Sirt1 and Poldip3 mRNA levels by binding to their 3′UTR, stabilizing them. Upon TPD-43 alterations, Sirt1 mRNA levels decrease and SIRT1-mediated deacetylation of Ku70 is reduced, lowering HR and NEHJ DSB repair. Likewise, a decrease in Poldip3 mRNA levels could reduce the DNA damage checkpoint and the resolution of R-loop. SIRT1 and POLDIP3 functions in the DSB response in mature neurons remain evasive. Different means by which functional TDP-43 acts against RTE activity at the chromatin level are listed, as well as the general negative consequences which RTE uncontrolled activity can have on the genome and transcriptome stability.

Figure 4

Short-TDP-43 isoforms: alternative splicing and context-specific fate. (A) Alternative splicing leading to short-TDP-43 proteins in brain. Transcripts resulting from various splicing of the alternative intron 6 (*) within TDP-43 pre-mRNA and translated into different short-TDP-43 proteins readily observed in mouse and human brains are depicted: ENST00000629725.2 encoding sTDP43-2 (also called hTDP-S7, mTDP-S7, TDP43-2, m2, or Cyte); ENST00000315091.7 encoding sTDP3-1 (also called hTDP-S6, mTDP-S6, TDP43-4, m1, or Tid). Both are highly conserved at the transcript and protein levels and the recently identified transcript producing TDP43C-spl. Note that the TSS and the TTS of the transcripts of these isoforms have not been validated, and it is not clear if they hold the 3′UTR TDPBR needed for autoregulation via TDP-43 FL. Short TDP43 isoforms-specific intron borders with splice donors (SD) and acceptors (SA) are indicated under each transcript in blue, with numbering given relative to the CDS +1. SD and SA are all located within TARDBP exon 6, eliminating the majority of exon 6. The sTDP43 protein isoforms contain at least the first 256AA and up to the first 280AA of TDP-43 (indicated in red). They contain the N-Term NLS, the RRM 1 and 2, and the NES, but not the C-terminal glycine-rich region in the TDP-43 protein. They gain an alternative 16 to 18AA, forming a unique C-end term VHLISNVYGRSTSLKVV, and sheltering a second nuclear export sequence (NSE) consisting in TSLKV. sTDP43-1, sTDP43-2, and TDP43C-spl have a mass of about 34 kDa, 32 kDa, and 30 kDa, respectively, and have been spotted in several vulnerable zones of the CNS both in normal and ALS patients, as well as in mouse male germ cells, as indicated between brackets. (B) Cell type sensitivity to high expression of short TDP proteins. Abnormal localization and ubiquitination of short TDP-43 aggregates leading to toxic inclusions are a pathological hallmark of neurons and glia in neurodegenerative diseases. Note the repressive potential of the mouse short TDP-43 isoforms Cyte and Tid (sTDP43-2 and sTDP43-1, respectively) on Acrv1 and c-fos promoters in GC-2 and Hela cells. The tethering of Cyte or Tid to a reporter plasmid using Gal4 DNA binding domain (DBD) fused to TDP-43 at Gal4 binding sequences (blue boxes) represses c-fos or Acrv1 promoters-induced luciferase expression, as does TDP-43 FL in the same conditions (up right frame). In neurons (left frames), human and mouse sTDP43-1 and 2 are present either in the nucleus or in the cytoplasm (soma and axons) or in both. They are upregulated by age and in response to increased neuronal activity (e.g., TEA in human iNeurons, or bicuculline in rodent primary mixed cortical neurons), and are conversely downregulated by TTX that abolishes neuronal activity. When overexpressed, they form insoluble aggregates in the cytoplasm able to sequester full-length TDP-43, leading to nuclear clearance of endogenous TDP-43 and neurotoxicity. TDP-43C-spl is observed in the cytoplasm of the human spinal cord, brain tissue, and dorsal root ganglia. TDP-43Cspl overexpression in neuronal cell lines convey their delocalization to the cytoplasm, where they form toxic ubiquitinated aggregates. In astrocytoma and microglia cell lines, TDP-43Cspl is not delocalized to the cytoplasm and localizes in interchromatin granule clusters (speckles) in the nucleus. TDP-43 is not recruited to the TDP43C-spl aggregates, but keeps its nuclear localization (bottom right frame).

Figure 5

Human TARDBP locus organization. Human TARDBP locus (black track; NCBI Gene ID: 23435), encoding full-length TDP-43 (FL) protein, is localized on chromosome 1p36.22 upstream of the MASP2 gene locus and counts up to 10 exons (obligatory and alternative exons). The blue tracks report the reference transcript coding for TDP-43 FL (NM_007375.4), as well as other predicted alternative transcripts isoforms (GENCODE. Version 42lift37 (Ensembl 108), ENST00000639083.1, ENST00000629725.2, ENST00000240185.8, ENST00000315091.7, ENST00000649624.1, ENST00000616545.4, and ENST00000621790.4,). NM_007375.4 is composed of six exons, flanked by a 5′ and a 3′ UTR. The 3′UTR embedded in exon 6 also contains two cryptic introns, intron 6* holding multiple alternative splicing sites and intron 7 (yellow frames, with splicing highlighted with blue dashed lines), as well as alternative polyadenylation signals (PAS) that have a fundamental importance for TDP-43 production control. The bottom dark green track shows the conservation within 100 vertebrates (“100 Vertebrates Conservation by PhastCons”). On the top of the TARDBP locus UCSC genome browser, tracks are reported. CpG score (red bars) shows different degrees of CpG methylation along the locus as found in the human cortex. H3K4me3 histone modification (green bars), the enrichment of which marks active promoter regions, is found upstream of exon 1 (TSS) down to exon 2. Common SNP (black and blue bars; from update V155) are found mainly within introns. OMIM variants (in green bars) are mainly in the exon 6, containing both the DNA and RNA signals for auto-regulation, and the region coding the C-term of the TDP-45 protein. All tracks are from the UCSC Genome Browser on Human (GRCh37/hg19).

Figure 6

Role of DNA CpG methylation in TDP-43 levels autoregulation. 1. Expected TDP-43 production. TDP-43 production is promoted by the transcription of an mRNA (blue, NM_007375.4) with 6 exons and usage of poly(A) signal (pA)1 located just after the TDPBR both within the alternative intron 7 (yellow box ≪7≫) or extending up to the pA4. When present at adequate levels, the mRNA is exported to the cytoplasm and is translated to produce TDP-43. 2. Normal autoregulation. When nuclear levels of TDP-43 increase over the expected threshold, TDP-43 binds co-transcriptionally to its own pre-mRNA at TDPBR and down to pA1, stalling RNA pol II at DNA in the correspondence of this region. This impedes pA1 signal usage and facilitates the splicing of the cryptic intron 7. At this stage, three scenario exists: (1) alternative use of pA4 generates an mRNA, mainly retained in the nucleus, with no additional splicing and holding a long 3′UTR; (2,3) pA2 or pA4 are used, and the alternative splicing of cryptic intron 7 and then 6 are triggered. The double-spliced isoforms are exported to the cytoplasm and become degraded via nonsense-mediated mRNA decay. These autoregulatory mechanisms, by resulting in a decrement in the cytoplasmic TDP-43 FL mRNA, allow for the adjusting of the TDP-43 protein levels. 3. Autoregulation failure and role of CpG methylation in the 3′UTR DNA region. The amount of TDP-43 in the nucleus determines the ratio of these isoforms, and when nuclear TDP-43 levels are reduced, the splicing is repressed. The absence of nuclear TDP-43 induces an abnormal autoregulation and increases the amount of TARDBP mRNA in the cytoplasm. The DNA region of the 3′UTR spanning the alternative exon 6-exon 7 junction in correspondence to the TDPBR on pre-mRNA bears a track of 15 CpGs, methylated in the human cortex. When these 15 CpGs are demethylated, TDP-43 autoregulation is negatively affected, suppressing the alternative exon 7 splicing and increasing canonical mRNA levels. The precise mechanisms are not known but possibly involve a reduction in RNA Pol II pausing or alteration in the recruitment of other factors. In vitro, this region is sensible to Tet1-induced demethylation and DNMT3b-induced re-methylation, both of which modulate intron splicing. In the human control motor cortex, DNA methylation at 3′UTR CpGs 10–15 is inversely correlated with age, and, generally, the CNS appears more affected than the liver, showing disparities among regions, with the motor cortex having the lowest CpG 10-15 methylation levels.

Figure 7

TARDBP locus promoter regulation. The main TSS identified for TARDBP and TDP-43 FL is from Seq NM_007375.4 (TSS1), starting at chr1; 11072711(+strand) (GRCh37/hg19). It is located in exon1 (Ex1), 102bp before intron1. Other confirmed (solid arrows, TSS2) or predicted (dashed arrows) TSSs lay upstream or downstream of TSS1 and are displayed. A minor TSS (TSS2) within exon2 (Ex2) has been identified in human fetal brain, without a precise localisation reported. Note that since the predicted TSSs at −32 served as a reference for start-numbering in different works, it is indicated here as ≪+1*≫, representing a shift of −32 bp relative to TDP−43 FL Seq transcript NM_007375.4 TSS (TSS1). The TARDBP promoter region is bipartite, with two core promoters (dark green boxes): one proximal TATA-less promoter located before the exon1 (−327–+1) and within (−451–−230); and one distal located at +972_+1094, i.e., within the intron1 position +850_+972. The proximal TATA-less promoter is necessary for the minimal promoter activity in all tested cell lines. Other regulatory features identified are positive (green boxes) and negative (red boxes) regulatory regions. A 58 bp region (−281–−223) in the proximal core is crucial for promoter activity. The upstream regulatory region (−927–−300), holding an iMotif (−371–−309), is important for maximal activity. Sequences in the +1–+123 region, encompassing exon1, positively enhance transcription. In the distal intron1 promoter, the region +788–+972 (+666–+850 of intron 1) is repressive. Epigenetic characteristics of TARDBP locus in the human cortex are displayed, as in Figure 5. H3K4me3 histone modification (green bars), the enrichment of which marks the active promoter regions and is found upstream of exon 1 (TSS) down to exon 2. The promoter region extending from −836 to + 1106 contains 125 CpG mostly within 3 CpG islands (CGI, dashed blue boxes) (−874 to + 1069). In the human cortex, along the promoter region, the CpG score (red bars) shows different degrees of CpG methylation. The region from CGI 2–3 down to exon 2 is not methylated. TF action and binding zone: factors shown to activate or repress TARDBP promoter are displayed on top of the locus by green or red zone, respectively. LPS and neuronal hyperactivity (triggered, e.g., by TEA, inhibited by TTX) positively regulate the TARDBP promoter, although the specific regions have not been defined yet. TDP-43 itself can repress its cognate promoter and, in particular, through the proximal upstream promoter. TDP-43mut: two TDP-43 mutants (G348C, A382T) activate the intron1 distal promoter. HNRNP-K binds to the iMotif in the upstream tropism. ALS-linked variants: position of variants identified in ALS patients, with their non-different higher frequency (f(ALS)) relative to frequency in HapMap controls (f(controls)), or found in ALS only, are depicted in white-to-purple circles within the TARDBP regulatory region. A (c.1-562t>c, rs9430335); B (c.1-100t>c; rs968545); C (c.13g>a); D (c.122+85c>t); E (c.122+95c>t); F (c.122+150delg); G (c.122+218c>t); H (c.122+284g>t); I (c.123-450a>c); J (c.123-262-263del); K (c.170c>t (p.N12N)); L (c.198t>c (p.A66A)) (see [175]). None of these variants were found to have a significant incidence on the TARDBP promoter in the tested conditions, and no OMIM variant is described in this region to date. All tracks are from UCSC Genome Browser on the human (GRCh37/hg19).