Cytoplasmic TDP43 Binds microRNAs: New Disease Targets in Amyotrophic Lateral Sclerosis

Amyotrophic lateral sclerosis (ALS) is a progressive, fatal, and incurable neurodegenerative disease. Recent studies suggest that dysregulation of gene expression by microRNAs (miRNAs) may play an important role in ALS pathogenesis. The reversible nature of this dysregulation makes miRNAs attractive pharmacological targets and a potential therapeutic avenue. Under physiological conditions, miRNA biogenesis, which begins in the nucleus and includes further maturation in the cytoplasm, involves trans-activation response element DNA/RNA-binding protein of 43 kDa (TDP43). However, TDP43 mutations or stress trigger TDP43 mislocalization and inclusion formation, a hallmark of most ALS cases, that may lead to aberrant protein/miRNA interactions in the cytoplasm. Herein, we demonstrated that TDP43 exhibits differential binding affinity for select miRNAs, which prompted us to profile miRNAs that preferentially bind cytoplasmic TDP43. Using cellular models expressing TDP43 variants and miRNA profiling analyses, we identified differential levels of 65 cytoplasmic TDP43-associated miRNAs. Of these, approximately 30% exhibited levels that differed by more than 3-fold in the cytoplasmic TDP43 models relative to our control model. The hits included both novel miRNAs and miRNAs previously associated with ALS that potentially regulate several predicted genes and pathways that may be important for pathogenesis. Accordingly, these findings highlight specific miRNAs that may shed light on relevant disease pathways and could represent potential biomarkers and reversible treatment targets for ALS.

Amyotrophic lateral sclerosis (ALS) is a progressive, fatal, and incurable neurodegenerative disease. Recent studies suggest that dysregulation of gene expression by microRNAs (miRNAs) may play an important role in ALS pathogenesis. The reversible nature of this dysregulation makes miRNAs attractive pharmacological targets and a potential therapeutic avenue. Under physiological conditions, miRNA biogenesis, which begins in the nucleus and includes further maturation in the cytoplasm, involves trans-activation response element DNA/RNA-binding protein of 43 kDa (TDP43). However, TDP43 mutations or stress trigger TDP43 mislocalization and inclusion formation, a hallmark of most ALS cases, that may lead to aberrant protein/miRNA interactions in the cytoplasm. Herein, we demonstrated that TDP43 exhibits differential binding affinity for select miRNAs, which prompted us to profile miRNAs that preferentially bind cytoplasmic TDP43. Using cellular models expressing TDP43 variants and miRNA profiling analyses, we identified differential levels of 65 cytoplasmic TDP43-associated miRNAs. Of these, approximately 30% exhibited levels that differed by more than 3-fold in the cytoplasmic TDP43 models relative to our control model. The hits included both novel miRNAs and miRNAs previously associated with ALS that potentially regulate several predicted genes and pathways that may be important for pathogenesis. Accordingly, these findings highlight specific miRNAs that may shed light on relevant disease pathways and could represent potential biomarkers and reversible treatment targets for ALS.

INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterized by cortical, brainstem, and spinal cord motor neuron loss that results in progressive skeletal muscle weakness and atrophy (Brown and Al-Chalabi, 2017). There is currently no effective cure and most patients die within 3-5 years of diagnosis. The disease can occur as familial ALS (fALS), which constitutes around 15% of ALS incidence, or the more frequent sporadic ALS (sALS; Brown and Al-Chalabi, 2017;Chia et al., 2018;Goutman et al., 2018;Oskarsson et al., 2018).
While suggestive of a causal relationship, however, our current understanding lacks insight into specific mechanisms whereby cyTDP43 inclusions may interfere with miRNA bioavailability and/or function. To address this gap in our knowledge, we used a high-throughput assessment of miRNA levels associated with multiple TDP43 pathological states to identify specific changes in miRNA/protein binding linked with cyTDP43. Our results expand the understanding of how TDP43 mislocalization affects specific miRNA levels in ALS, and thereby provide a foundation to ultimately gain new mechanistic insights and uncover untapped diagnostic and therapeutic avenues in ALS.

Plasmids
To generate stable doxycycline-inducible cell lines expressing eGFP-His-tagged TDP43 variants that primarily localized to the nucleus or cytoplasm, we first amplified eGFP-His from pGW1-T202-TDP43-eGFP-His  using the primers 5 -ATA AGA ATG CGG CCG CAA CTA GAG CTG TTT GGG ACG-3 and 5 -CGG ACG CGT TTT TAG TGA TGG TGA TGG TGA TG-3 . The eGFP-His fragment was subcloned directionally into the NotI-MluI restriction sites (underlined in bold in the primer sequence) of the doxycycline-inducible pLVX-TRE3G lentiviral vector (Clontech, Takara Bio USA, Mountain View, CA, USA) to generate the pLVX-TRE3G-eGFP-His construct.
WT TDP43 from pGW1-TDP43WT-eGFP (Barmada et al., 2014) was next amplified using the primers 5 -CGG GAT CCA TGT CTG AAT ATA TTC GGG TAA CCG-3 and 5 -ATA GTT TAG CGG CCG CCA TTC CCC AGC CAG AAG A-3 , and the WT TDP43 fragment was ligated directionally into the BamHI-NotI restriction sites (underlined in bold in the primer sequence) of the pLVX-TRE3G-eGFP-His construct to create the final pLVX-TRE3G-WT-TDP43-eGFP-His plasmid. PCR reactions were performed with Q5 High-Fidelity DNA Polymerase (New England BioLabs) and transformations with Stellar competent cells (Clontech, Takara Bio USA Mountain View, CA, USA), both according to the manufacturer's instructions. Missense mutations disrupting the two nuclear localization signals (∆NLS) in TDP43 to force cytoplasmic localization were next introduced into the pLVX-TRE3G-TDP43WT-eGFP-His plasmid using site-directed mutagenesis as previously described (Winton et al., 2008; to generate pLVX-TRE3G-∆NLS-TDP43-eGFP-His. All sequences were verified by the University of Michigan Sequencing Core 1 .

Generation of Doxycycline-Inducible Stable HeLa Cell Lines
HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM; 11965-092, Gibco, Thermo Fisher Scientific, Gaithersburg, MD, USA) supplemented with 10% fetal bovine serum (A3160401, Thermo Fisher Scientific, Gaithersburg, MD, USA) at 37 • C in 10% CO 2 . Confluent cells were transiently transfected with 100 ng of the pLVX-TRE3G-eGFP-His, pLVX-TRE3G-WT-TDP43-eGFP-His, or pLVX-TRE3G-∆NLS-TDP43-eGFP-His constructs along with pLVX-EF1α-Tet3G (Clontech, Takara Bio USA) using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. Following a 6 h incubation, wells were rinsed and replenished with growth media containing the selection antibiotics puromycin (1 µg/ml; #P8833, Sigma, St. Louis, MO, USA) and geneticin (1 µg/ml; G418; #10131-035, Gibco) for 48 h. Doxycycline (1 µg/ml; #PHR1145, Sigma) was then added to the media to induce protein expression and confirm transfection. Transfection efficiency was determined to be about 20%, and stable lines were generated using a serial dilution protocol in 96-well plates (Corning Inc., Corning, NY, USA; Ryan, 2008). HeLa cells were monitored daily and scored for wells that contained a single colony, which was grown to confluence. Induction with doxycycline was performed to select eGFP-positive clones for expansion. A confluent sample from each expanded eGFP-expressing clone was fixed with 4% paraformaldehyde and stained with Hoechst reagent to visualize the nuclei and localize eGFP (i.e., nuclear vs. cytoplasmic). Multiple clones were generated that varied modestly in the level of WT-TDP43-eGFP-His, ∆NLS-TDP43-eGFP-His, or eGFP-His expression, and those with the most optimal eGFP level (high expression with normal cellular morphology and survival) were selected for further experiments.

WT-TDP43-eGFP-His and ∆NLS-TDP43-eGFP-His Pull-Down
To further evaluate TDP43 variant expression patterns, WT-TDP43-eGFP-His, ∆NLS-TDP43-eGFP-His, and control eGFP-His were pulled down via their His tags from the subcellular fractions using nickel-agarose (Ni 2+ ) beads (Qiagen; Figueroa-Romero et al., 2009). Briefly, 40 µl Ni 2+ -beads prewashed in CHAPS buffer were added to each 600 µl sample and rotated overnight at 4 • C. Beads were then washed three times with CHAPS buffer and divided into two aliquots, one for protein and one for RNA extraction. A bead-only sample was included throughout the rest of the protocol as a negative control.
For extracting miRNAs, bead aliquots from the cytoplasmic fractions were washed twice with 1× PXL (0.1% SDS, 0.5% Triton X-100, 0.5% deoxycholate in 1× PBS) and twice with 5× PXL. Samples were then boiled for 3 min to reverse crosslinking (King et al., 2014), and QIAzol (Qiagen) was added and samples were stored at −80 • C. Next, a miRNeasy Mini Kit (cat# 217004, Qiagen) was used to extract total RNA, enriched in cyTDP43-bound miRNAs. Purified RNA was treated with RNase-free DNase and stored at −80 • C until required for cDNA synthesis or miRNA profiling.

miRNA NanoString Profiling, Analysis, and Validation
To evaluate miRNA profiles associated with cyTDP43, total RNA from the cytoplasmic fractions pulled-down with Ni 2+beads for WT-TDP43-eGFP-His, ∆NLS-TDP43-eGFP-His, and eGFP-His lines (n = 4 for each line) was processed using the nCounter Human v3A miRNA Gene List (NanoString Technologies, Seattle, WA, USA) as published previously (Wohlfarth et al., 2017). A nCounter Digital Analyzer counted individual fluorescent barcodes that quantified and identified levels of up to 800 human-specific target miRNA molecules. nSolver Analysis Software 3.0 (NanoString Technologies) was used to background subtract and normalize the top 100 miRNA counts. Fold-change differences for WT-TDP43-or ∆NLS-TDP43-associated miRNAs relative to control (eGFP-His) were identified using P < 0.05 as the significance cutoff. Significant miRNAs were then analyzed using mirPath v.3 2 (Vlachos et al., 2015) using a false discovery rate corrected P-value threshold of 0.05 and a microT threshold of 0.8 for predicted miRNA targets in the microT-CDS database to identify Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways 3 that associate with the cyTDP43-interacting miRNAs.
To validate the NanoString dataset, quantitative real-time PCR (qPCR) of select cyTDP43-associated miRNAs was performed using TaqMan Universal PCR Master Mix, TaqMan MicroRNA Assays (Thermo Fisher Scientific), and a 1:100 dilution of the pre-amplified NanoString analysis template as previously described (Figueroa-Romero et al., 2016) using the standard TaqMan protocol on a StepOnePlus Real-Time PCR System (Applied Biosystems). The cycle threshold (CT) values representing miRNA expression in the samples were calculated using StepOnePlus software and normalized to mir-92a-3p, a miRNA with a low coefficient of variation across samples. Next, ∆CT and ∆∆CT were calculated relative to the control group. miRNA levels were expressed as the mean ± standard error of the mean (SEM) for triplicates and statistically significant differences between either WT-TDP43-eGFP-His or ∆NLS-TDP43-eGFP-His to control (eGFP-His) were evaluated with a two-sample equal-variance Student's t-test (Prism 5, GraphPad, San Diego, CA, USA); P < 0.05 was considered statistically significant.

TDP43 Binds Differentially to Distinct miRNAs
To gain insight into TDP43 s ability to interact with specific miRNAs, we first used a native gel mobility shift assay to evaluate binding of denatured TDP43 to miR-132, miR-143, and miR-574, miRNAs all previously reported to associate with TDP43 (Buratti et al., 2010;Kawahara and Mieda-Sato, 2012;Freischmidt et al., 2013). Results demonstrated a band shift only for miR-574 ( Figure 1A), with no observable shift for miR-132 or miR-143, suggesting that denatured TDP43 has a strong affinity only for miR-574. Since TDP43 binding often relies on associations with UG repeats (Ayala et al., 2005;Bhardwaj et al., 2013), the apparent binding affinity of TDP43 for miR-574 (9 UG repeats), miR-652 (3 UG repeats), and miR-204 (2 UG repeats) was next evaluated using a colorimetric ELISA to detect natively folded TDP43 bound to surface-captured biotinylated miRNAs (Rumora et al., 2010(Rumora et al., , 2013. EC50 values were determined for the binding of TDP43 to each miRNA from a titration of the biotinylated miRNAs. TDP43 had the highest apparent affinity for miR-574 (EC50 = 2.84 nM), a slightly lower binding affinity for miR-652 (EC50 = 6.41 nM), and the lowest binding affinity for miR-204 (EC50 = 37.7 nM; Figure 1B). These EC50s are proportional to the number of UG repeats in each miRNA sequence (miR-574 > miR-652 > miR-204) and support the contention that UG content dictates the relative binding affinity of TDP43 (Bhardwaj et al., 2013).

Cells Expressing WT-TDP43-eGFP-His and ∆NLS-TDP43-eGFP-His Reflect Differential Subcellular TDP43 Localization
To provide a model to study the effect of altered TDP43 cellular localization on miRNAs, we established doxycycline-inducible HeLa cell lines that stably express eGFP-and His-tagged WT-TDP43 or ∆NLS-TDP43, along with eGPF-His-expressing controls. Upon doxycycline-induced expression, eGFP imaging revealed primarily nuclear localization of overexpressed WT-TDP43-eGFP-His vs. primarily cytoplasmic localization of ∆NLS-TDP43-eGFP-His (Figure 2A). This differential localization was further supported by Western blotting of eGFP in nuclear and cytoplasmic fractions. α-Tubulin, a cytoplasmic protein, was detected in whole-cell lysates and cytoplasmic fractions for all three cell lines, as expected ( Figure 2B, top panel). The nuclear marker H2B was present in whole-cell lysates and nuclear fractions of all three cell lines, and only weakly detected in their cytoplasmic fractions ( Figure 2B, middle panel). eGFP expression was also confirmed in whole-cell lysates of all three cell lines ( Figure 2B, bottom panel), although the expression levels were low in some fractions, so we enriched using a Ni 2+ -bead pull-down assay for His-tagged proteins to better evaluate subcellular localization in each fraction. The resulting data corroborated the primarily cytoplasmic localization of ∆NLS-TDP43-eGFP-His ( Figure 2C) and verified the largely nuclear localization of WT-TDP43-eGFP-His that was accompanied by some anticipated cytoplasmic localization from its natural nucleus-cytoplasm shuttling role (Li et al., 2015;Weskamp and Barmada, 2018; Figure 2C). Of note, the lower of the two TDP43 bands in the nuclear fractions likely reflects endogenous TDP43 associating with the eGFP-Histagged TDP43 aggregates, whereas the higher band represents the eGFP-His fusion protein. The third band in the ∆NLS-TDP43-eGFP-His cytoplasmic fraction most likely reflects a proteolytic TDP43 cleavage product (Li et al., 2015;Weskamp et al., 2020). Together, these data support our ability to efficiently separate the cytoplasmic and nuclear compartments for all three cell lines, with minimal inter-fraction contamination, and our ability to effectively pull down TDP43 via its His tag.

WT-TDP43-eGFP-His-and ∆NLS-TDP43-eGFP-His-Expressing Cells Exhibit Differential miRNA Profiles
To identify miRNAs that interact with cytoplasmic WT-TDP43-eGFP-His or ∆NLS-TDP43-eGFP-His, we performed NanoString miRNA profiling of TDP43-bound miRNAs isolated by Ni 2+ -pulldown from cytoplasmic fractions of the respective stable HeLa cell lines. Appreciable target detection occurred for most miRNAs across all samples, and assessment of fold-changes vs. the control eGFP-His line identified 65 significantly altered cyTDP43-associated miRNAs (Figure 3 and Table 1). For WT-TDP43-eGFP-His cells, 11 miRNAs decreased and 14 increased in level (Figure 3B). In the ∆NLS-TDP43-eGFP-His cell line with primarily cytoplasmic TDP43 localization, more miRNAs satisfied the criteria for significant fold-enrichment, with 29 decreased and 23 increased in levels ( Figure 3C). FIGURE 1 | Trans-activation response element DNA/RNA-binding protein of 43 kDa (TDP43) binding specificity and apparent binding affinities for select microRNAs (miRNAs). (A) A native gel mobility shift assay was used to evaluate the association between TDP43 and previously associated miRNAs: miR-132, miR-143, and miR-574. Trace amounts of 32 P-radiolabeled RNA (5 nM) were incubated with increasing concentrations of TDP43 (0, 0.05, 0.125, 0.25, 0.5, and 1.25 mM) and analyzed on a 12.5% native PAGE gel. A shift in band size was observed only for miR-574 (right). (B) Apparent binding affinities (EC50s) of TDP43 for select miRNAs with varying numbers of UG repeats were determined using a titration of biotinylated miRNAs in a colorimetric ELISA. EC50s are proportional to the number of UG repeats in each miRNA (miR-574, nine repeats > miR-652, three repeats > miR-204, two repeats).
Interestingly, nine miRNAs (four decreased and five increased) in the WT-TDP43-eGFP-His cells and 11 miRNAs (six decreased and five increased) in the ∆NLS-TDP43-eGFP-His cells were enriched by more than 3-fold. Moreover, 12 miRNAs were present in both cell lines and exhibited similar overor under-enrichment.
To validate the NanoString profiling results, we performed qPCR on the same samples used for miRNA profiling for miR-204-5p and miR-129-5p. qPCR miRNA levels were normalized to miR-92a, which has a low coefficient of variation relative to other miRNAs in all cell lines (data not shown). Similar to the NanoString data, qPCR showed statistically significant increases in miR-204-5p and levels in both WT-TDP43-eGFP-His and ∆NLS-TDP43-eGFP-His HeLa cell lines (Figures 4A,C). Likewise, miR-129-5p increased in both WT-TDP43-eGFP-His and Immunoblotting of whole-cell lysates and cytoplasmic and nuclear fractions for α-tubulin (cytoplasmic marker; top panel), H2B (nuclear marker; middle panel), and eGFP (a marker for expression of the eGFP-His-tagged doxycycline-inducible plasmids) following subcellular fractionation of the control (C), WT-TDP43-eGFP-His (WT), and ∆NLS-TDP43-eGFP-His (∆) lines. eGFP immunoblotting demonstrates that eGFP-His (lower band) is expressed in the whole-cell lysates and cytoplasmic fraction of the control cell line, while eGFP-His-tagged TDP43 (upper band) is present in the whole-cell lysates and cytoplasmic fraction of the WT-TDP43-eGFP-His and ∆NLS-TDP43-eGFP-His cell lines, although only marginal expression seen in the WT-TDP43-eGFP-His cell line. (C) TDP43 immunoblotting following pull-down of the His-tagged proteins using Ni 2+ beads. ∆NLS-TDP43-eGFP-His is detected mainly in the cytoplasmic fraction, with some expression also seen in the nuclear fraction, while WT-TDP43-eGFP-His is more highly abundant in the nuclear fraction, though modest expression is also seen in the cytoplasmic fraction. The slightly higher bands in both TDP43 lines represent the TDP43-eGFP-His fusion proteins, while the lower bands represent endogenous TDP43. In the cytoplasmic fraction, ∆NLS-TDP43-eGFP-His also exhibits a third band (lowest) that likely reflects a proteolytic cleavage product of TDP43. C = eGFP-His; WT = WT-TDP43-eGFP-His; ∆ = ∆NLS-TDP43-eGFP-His; B = Ni 2+ bead only control; M = marker/ladder. ∆NLS-TDP43-eGFP-His cell lines but did not reach statistical significance (Figures 4B,C).
miRNA pathway analysis of the 65 miRNAs using mirPath v.3 (Vlachos et al., 2015) further revealed several predicted gene targets and associated KEGG pathways that offered insight into the relevance of the identified cyTDP43-interacting miRNAs. Pathways represented in both WT-TDP43-eGFP-His and ∆NLS-TDP43-eGFP-His included multiple cancer and neuronal function and health pathways, such as GABAergic and glutamatergic synapses, axon guidance, and neurotrophin signaling pathways (Figure 5; Supplementary Tables S1, S2).
Taken together, these data indicate that TDP43 expression and localization significantly affects the levels of multiple miRNAs, which are predicted to affect several genes and biological pathways that are relevant to nervous system health and disease.

DISCUSSION
TDP43 mutations are linked to ALS (Kapeli et al., 2017), and cyTDP43 is a universal histopathological hallmark of postmortem brain and spinal cord tissue from fALS and sALS subjects  (Neumann et al., 2006;Zhang et al., 2008;Liu-Yesucevitz et al., 2010). However, the implications of these cytoplasmic inclusions on key cellular processes, including gene expression, have not been fully elucidated. TDP43 is central to miRNA biogenesis, and its ability to directly interact with miRNAs regulates expression of numerous genes (Kawahara and Mieda-Sato, 2012;King et al., 2014;Emde et al., 2015;Loffreda et al., 2015;Paez-Colasante et al., 2015). We hypothesized that one consequence of TDP43 cytoplasmic translocation is aberrant protein/miRNA interactions that may ultimately affect cell function and viability through epigenetic mechanisms. To address this idea, we first showed that TDP43 exhibits differential binding affinity for specific miRNAs and that binding affinities correlate with the number of miRNA UG repeats. This suggests that cyTDP43 may preferentially affect certain miRNAs during disease pathogenesis. Next, using stable cell lines expressing inducible TDP43 variants with primarily nuclear (WT-TDP43-eGFP-His) or cytoplasmic (∆NLS-TDP43-eGFP-His) subcellular localization, we identified 65 differentially enriched miRNAs that associate with cyTDP43, including known and novel hits that potentially regulate multiple predicted genes and pathways. Together, these findings provide insight into the consequences of TDP43 mislocalization in ALS by identifying specific miRNAs that associate with cyTDP43. These data could ultimately shed light on pathogenic mechanisms, biomarkers, and reversible ALS treatment targets.
TDP43 is normally predominantly found in the nucleus; however, ALS-associated mutations and/or cellular stress can prompt hyper-phosphorylation and sequestration into insoluble, ubiquitin-positive, denatured cytoplasmic aggregates (Neumann et al., 2006(Neumann et al., , 2009. Previous studies also indicate that TDP43 knockdown is toxic (Sephton et al., 2010;Iguchi et al., 2013;Barmada et al., 2015), that TDP43 toxicity more closely correlates with increased cyTDP43 than with the level of nuclear TDP43 , and that cyTDP43 clearance can exert beneficial effects on neuronal survival (Barmada et al., 2014) and functional outcomes (Walker et al., 2015). cyTDP43 has thus been the focus of multiple in vitro and in vivo studies in recent years (Miguel et al., 2011;Walker et al., 2015;Birsa et al., 2020). Of particular relevance is the rNLS8 ALS mouse that exhibits inducible expression of ∆NLS-TDP43 in motor neurons. This transgenic mouse recapitulates many of the pathologic characteristics of ALS, including cyTDP43 aggregation, motor neuron FIGURE 4 | Confirmation of select enriched miRNAs. Levels of select miRNAs from the NanoString profiling results in cytoplasmic fractions of the control (eGFP-His), WT-TDP43-eGFP-His (WT-TDP43), and ∆NLS-TDP43-eGFP-His (∆NLS-TDP43) lines were confirmed using quantitative real-time PCR (qPCR). Results were normalized to miR-92a (∆CT) and then relative to the control group (∆∆CT) and expressed as mean fold-change ± standard error of the mean (SEM; n = 4 per line). (A) miR-204-5p was significantly upregulated relative to control in both TDP43 lines in a similar proportion to the NanoString results (*p < 0.05). (B) miR-129-5p levels in both TDP43 lines were upregulated at fold-changes equivalent to the NanoString results, but did not reach significance relative to the control line.
(C) Fold-change comparisons between the qPCR and NanoString profiling results.
death, increases in CSF neurofilament levels, neuromuscular junction loss, muscle atrophy, and abnormal compound muscle axon potentials measured by electromyogram (Spiller et al., 2016a(Spiller et al., ,b, 2018(Spiller et al., , 2019. However, a complete understanding of the effect TDP43 mislocalization has on specific miRNAs is lacking. This is important for neuronal health because cytoplasmic TDP43 aggregates can interact with and sequester key miRNAs, thus limiting their ability to perform their normal regulatory functions. Cytoplasmic TDP43 aggregates also impair normal miRNA maturation through a loss of nuclear localization and interactions with miRNA biosynthetic pathway components Kawahara and Mieda-Sato, 2012;Weskamp and Barmada, 2018). Herein, we specifically focused on understanding how cytoplasmic translocation of TDP43 affects miRNA dynamics as another potential pathomechanism.
Before examining the impact of cyTDP43 on miRNAs, we first assessed TDP43 interactions and binding affinities to three specific miRNAs known to associate with TDP43: miR-132, miR-143, and miR-574-5p (Buratti et al., 2010;Kawahara and Mieda-Sato, 2012;Freischmidt et al., 2013). We found that denatured TDP43 complexed with miR-574-5p, a miRNA first predicted to bind to TDP43 based on its UG-repeat regions in primary sequence (Buratti et al., 2010), but not miR-132 or miR-143, within the detection limit of our native gel mobility shift assay. We further verified that apparent binding affinities of TDP43 to miRNAs increased relative to the number of UG repeats; miR-574-5p with nine UG repeats had the highest affinity relative to miRNAs with only two or three UG repeats. Though all three miRNAs we selected to examine are known to exhibit altered levels in CSF and serum from ALS subjects (Freischmidt et al., 2013), we did not observe interactions between TDP43 and miR-132 or miR-143. This may be related to the use of denatured TDP43 in our assay, selected because of its biological representation of TDP43 aggregates, or due to differential regulation of the miRNAs in the CSF and serum vs. cells, suggesting that celltype-specific mechanisms may be involved (Freischmidt et al., 2013). Alternatively, TDP43 can bind UG-rich sequences as well as other secondary structures in pri-and pre-miRNAs (Kawahara and Mieda-Sato, 2012). Nonetheless, our data support the idea that TDP43 localization can differentially affect specific miRNAs.
To better understand the role of TDP43 mislocalization on miRNA dynamics, we created stable inducible cell lines that express predominantly nuclear (WT) or cytoplasmic (∆NLS) TDP43 variants in HeLa cells, which are a known and reliable model of TDP43 pathology (Ayala et al., 2008;Ling et al., 2010;Gu et al., 2017). cyTDP43 expression was regulated by a doxycycline-inducible promoter to prevent cytotoxicity from the overexpressed or mutant proteins (Walker et al., 2015), a GFP tag facilitated visualization, and a His-tag was present to leverage cyTDP43/miRNA profiling using a pulldown assay. We confirmed moderate WT-TDP43 overexpression that led to primarily nuclear expression and marginal cytoplasmic expression, similar to other WT-TDP43 overexpression models (Neumann et al., 2006(Neumann et al., , 2009Barmada et al., 2015;Li et al., 2015;Wang et al., 2015). We similarly verified that the mutant ∆NLS-TDP43 was localized to the cytoplasm, leading to diffuse or aggregated cyTDP43 as in previous in vitro cell models (Winton et al., 2008;Liu-Yesucevitz et al., 2010), in vivo animal models (Wils et al., 2010;Miguel et al., 2011;Walker et al., 2015), and human postmortem neuronal tissue
Among the most highly differentially represented miRNAs, we identified several novel miRNAs that have not previously been linked to ALS, such as miR-3065 and miR-129-5p. While little has been reported on the function of miR-3065, miR-129-5p is implicated in epilepsy (Sosanya et al., 2015;Liu et al., 2017;Rajman et al., 2017), neuroinflammation in ischemiareperfusion (Li et al., 2017), and Alzheimer's disease (Zeng et al., 2019). Our study suggests a potential neuroprotective role for miR-129-5p, which is blocked when miR-129-5p associates with cyTDP43; however, more research is needed to fully understand the meaning of the interaction of miR-129-5p with cyTDP43.
Several of the cyTDP43-associated miRNAs we identified herein are known to have a role in ALS pathogenesis. miR-577, which was highly increased in both cells lines in our study, is predicted to target eukaryotic translation initiation factor 2C, 2/argonaute 2 (EIF2C2/AGO2), a member of the RNA-induced silencing complex (RISC), which interacts with mature miRNAs to bind mRNAs and interfere with their translation (Ha and Kim, 2014;Kobayashi and Tomari, 2016). Interestingly, miR-577 is decreased in postmortem sALS spinal cord tissue with a concomitant increase in EIF2C2/AGO2 gene expression (Figueroa-Romero et al., 2016). In parallel, we found that miR-129-2-3p was highly associated with cyTDP43 in our ∆NLS-TDP43-eGFP-His line and that miR-204-5p was increased in both lines. However, both miR-129-2-3p and miR-204-5p are lower at the neuromuscular junction of ALS subjects (De Felice et al., 2018). Finally, miR-221-3p was among the more highly enriched cy-TDP43-associated miRNAs in our study but has been reported to be both increased and decreased in other ALS studies (D'Erchia et al., 2017;Di Pietro et al., 2018;Taguchi and Wang, 2018).
The discordance between reported studies and some of our top miRNAs likely reflects that the clinical profiling studies evaluated case/control differences in plasma, blood, CSF, muscle biopsies, or spinal cord tissue (Joilin et al., 2019), while we looked more directly at miRNAs that associate with cyTDP43. It is possible that miRNAs that more highly associate with cyTDP43 may be sequestered and do not reach peripheral biofluids or tissues, and thereby are reduced in studies profiling biofluids and tissue samples. It is also plausible that miRNAs that interact with cyTDP43 earlier in the course of the disease could lead to ultimate depletion and/or compensatory changes in postmortem assessment.
Another confounding factor is that TDP43 autoregulates its expression levels (Ayala et al., 2011;Prasad et al., 2019). TDP43 is known to localize to stress granules under stress conditions (Emde et al., 2015;Paez-Colasante et al., 2015;Weskamp and Barmada, 2018;Chen and Cohen, 2019), and cytoplasmic TDP43 redistribution may represent a physiologic response to stress that becomes toxic if the normal distribution is not reestablished . Thus, it is unclear if the association of certain miRNAs with cyTDP43 represents a possible preventive measure or rather promotes neuronal injury. In the case of miR-204, levels are increased in a cellular model of Parkinson's disease (Talepoor Ardakani et al., 2019), it is implicated in optic nerve injury , and miR-204 overexpression impairs neurite outgrowth (López-González et al., 2018) while inhibition provides neuroprotection (Yan et al., 2019). Ultimately, more research is needed to determine how miRNA profiles change over time, associate with TPD43, and impact ALS progression.
We also found an increased association of miR-15a-5p and miR-16-5p with cyTDP43. These miRNAs belong to a miR-15/16 cluster in an intronic region of the DLEU2 gene on chromosome 13 and are linked to ALS through profiling peripheral blood, neuron-derived extracellular vesicles, and muscle from ALS subjects (Liguori et al., 2018;Si et al., 2018;Katsu et al., 2019). Much of the published data on this cluster indicate that miR-15/16 members directly target the apoptosis regulator Bcl-2 to inhibit cellular proliferation, induce cancer cell apoptosis, and thereby reduce tumorigenicity (Aqeilan et al., 2010;Pekarsky et al., 2018). Since Bcl-2 is also tied to apoptotic regulation in neural cells (Azzouz et al., 2000;Soane and Fiskum, 2005;Hollville et al., 2019), similar mechanisms could be relevant to ALS. We speculate that the increased associations of miR-15/16 we observed with cyTDP43 could sequester these miRNAs, preventing Bcl-2 regulation, which in turn could promote cell death in ALS. Future studies are needed to confirm this possibility and understand the role of miR-15/16 in ALS pathogenesis.
At the pathway level, our results overlap with reported studies. Examining predicted target pathways identified over 30 linked to both WT-TDP43-eGFP-His and ∆NLS-TDP43-eGFP-His lines, as well as several others that were represented in either cell line. While many were related to cancer signaling, pathways connected to neuronal function and health were also commonly reflected, including GABAergic and glutamatergic synapses, axon guidance, and neurotrophin signaling pathways. Moreover, our analysis identified several pathways linked to miRNA profiles identified in ALS subject muscle tissue and serum, including regulation of actin cytoskeleton, ubiquitin-mediated proteolysis, and transforming growth factor-β (TGF-β) signaling (Kovanda et al., 2018;Taguchi and Wang, 2018). The TGF-β signaling pathway is an established player in neuronal health Katsuno et al., 2011) and increased levels are associated with early disease stages in animal models and ALS subjects (Si et al., 2014(Si et al., , 2015Meroni et al., 2019). Thus, these and other related pathways may provide important insights into the mechanisms and therapeutic opportunities for ALS.
Although the current study identified several intriguing cyTDP43-associated miRNAs and predicted pathways, we acknowledge that these data are based on an in vitro system with mutant cyTDP43 expression. The ∆NLS-TDP43 mutation, however, offers an important well-established strategy to model ALS and examine both disease mechanisms and resultant phenotypic consequences of cyTDP43 expression (Winton et al., 2008;Miguel et al., 2011;Walker et al., 2015). Notably, in vitro studies using similar strategies have also verified that toxicity associated with ∆NLS-TDP43 parallels that of common TDP43 mutations seen in ALS subjects (Winton et al., 2008;Barmada et al., , 2014. The overlap between our ∆NLS-TDP43 results and those from the WT-TDP43 line exhibiting moderate overexpression further supports the relevance of our observations. Future studies examining miRNA profiles that associate with cyTDP43 in relevant ALS animal models (Walker et al., 2015) and postmortem tissue are needed to both confirm the in vitro data as well as assess potential interventions based on identified targets.
We also acknowledge that cyTDP43 translocation can have implications beyond direct interactions with mature miRNAs. Our initial assessments of pri-miRNA levels for two differentially expressed mature miRNAs observed in ALS subject postmortem spinal cord tissue indicates that pri-miRNA expression was not altered, but that alterations must emerge at the level of miRNA processing (Figueroa-Romero et al., 2016). Other studies have also verified that TDP43 directly interacts with pri-miRNA and pre-miRNA intermediates and miRNA biosynthetic pathway components (Buratti et al., 2010;Kawahara and Mieda-Sato, 2012). Thus, while in the current study we only assessed mature miRNAs that interacted with TDP43, it is possible the impact of TDP43 may be conferred on miRNA intermediates along the biogenesis pathway or via sequestration in inclusions like stress granules (Emde et al., 2015;Paez-Colasante et al., 2015;Weskamp and Barmada, 2018). Future assessment of these alternative roles of cyTDP43 is needed. Finally, the miRNAs we identified did not overlap entirely with other reported studies (Dardiotis et al., 2018). This may be due to the heterogeneity in sALS samples examined by other studies relative to our specific focus on miRNAs bound to cyTDP43 in HeLa cells.
Overall, the current study used cellular models to identify miRNAs that interact with cyTDP43. We identified known and novel dysregulated miRNAs, data that provide a foundation for inferred biological insight and potential pathomechanisms from predicted pathways. cyTDP43 inclusions, by sequestering miRNAs, may skew neuronal epigenetic homeostasis and result in motor neuron death. Our findings may therefore guide further research examining the molecular downstream events resulting from dysregulated miRNAs that influence neurodegeneration in ALS and identify new biomarkers and therapeutic targets for disease diagnosis and intervention.

DATA AVAILABILITY STATEMENT
The datasets generated for this study can be found in the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/ query/acc.cgi?acc=GSE145214).

AUTHOR CONTRIBUTIONS
XP-C, CF-R, JH, LH, NW, SB, and EF contributed to the conception and design of the study. XP-C, CF-R, AR, FM, JMH, CB, GT, and LH contributed to the acquisition of data. XP-C and JH performed the statistical analysis. XP-C, C-FR, AR, JH, LH, SB, SS, and EF interpreted the data. XP-C and SS wrote the first draft of the manuscript. CF-R and AR wrote sections of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

FUNDING
Funding support was provided by the NeuroNetwork for Emerging Therapies, the A. Alfred Taubman Medical Research Institute, the Sinai Medical Staff Foundation Neuroscience Scholar Fund, and the National Institutes of Health (F32 DK112642 and K99 DK119366 to AR; R01 GM081025 and GM062357 to NW; and R01 NS097542 to SB).