Downregulation of MicroRNA-9 in iPSC-Derived Neurons of FTD/ALS Patients with TDP-43 Mutations

Transactive response DNA-binding protein 43 (TDP-43) is a major pathological protein in frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS). There are many disease-associated mutations in TDP-43, and several cellular and animal models with ectopic overexpression of mutant TDP-43 have been established. Here we sought to study altered molecular events in FTD and ALS by using induced pluripotent stem cell (iPSC) derived patient neurons. We generated multiple iPSC lines from an FTD/ALS patient with the TARDBP A90V mutation and from an unaffected family member who lacked the mutation. After extensive characterization, two to three iPSC lines from each subject were selected, differentiated into postmitotic neurons, and screened for relevant cell-autonomous phenotypes. Patient-derived neurons were more sensitive than control neurons to 100 nM straurosporine but not to other inducers of cellular stress. Three disease-relevant cellular phenotypes were revealed under staurosporine-induced stress. First, TDP-43 was localized in the cytoplasm of a higher percentage of patient neurons than control neurons. Second, the total TDP-43 level was lower in patient neurons with the A90V mutation. Third, the levels of microRNA-9 (miR-9) and its precursor pri-miR-9-2 decreased in patient neurons but not in control neurons. The latter is likely because of reduced TDP-43, as shRNA-mediated TDP-43 knockdown in rodent primary neurons also decreased the pri-miR-9-2 level. The reduction in miR-9 expression was confirmed in human neurons derived from iPSC lines containing the more pathogenic TARDBP M337V mutation, suggesting miR-9 downregulation might be a common pathogenic event in FTD/ALS. These results show that iPSC models of FTD/ALS are useful for revealing stress-dependent cellular defects of human patient neurons containing rare TDP-43 mutations in their native genetic contexts.

TDP-43 is a disease-related protein that links FTD and ALS. TDP-43 has two RNA recognition motifs and a glycine-rich region and is implicated in multiple aspects of RNA metabolism [14,15]. Among these, TDP-43 has been linked to the processing of microRNAs (miRNAs) [16][17][18]. MiRNAs are small noncoding RNAs that regulate gene expression post-transcriptionally by degrading their target mRNAs or repressing their translation [19,20]. Although potential involvement of miRNAs in neurodegeneration has been increasingly appreciated, information about their regulation and function in diseased human neurons remains scarce [21].
Modeling neurodegenerative diseases with TDP-43 pathology in animal models has been challenging. Although loss of nuclear TDP-43 and intracellular TDP-43 aggregates are common pathological features in neurons of FTD and ALS patients, disease-causing mutations in the gene are rare. Thus, in most patients, other contributing genetic factors appear to function upstream of TDP-43 during disease initiation and progression, which is difficult to model in animals. Moreover, the proper expression level of TDP-43 is critically important for its function. Ectopic expression of wildtype TDP-43 alone can be toxic, and TDP-43 itself is under complex autoregulation [22][23][24][25]. Deletion of TDP-43 in mice is lethal [26][27][28], highlighting its importance, but its conditional knockout shows disease-relevant phenotypes [29][30][31].
The ability to generate human induced pluripotent stem cells (iPSCs) from somatic cells [32] allows disease phenotypes of human neurons to be analyzed in the context of the unique genetic background of individual patients. Indeed, iPSCs have been generated from patients with ALS and other neurodegenerative diseases [33][34][35][36][37][38][39][40]. In this study, we generated iPSC lines from an FTD/ALS patient with TARDBP A90V mutation and from a healthy family member with no known mutations. We characterized neurons from this mutation and identified cell-specific phenotypes. Among these, we identified reduced miR-9 expression as a molecular defect that is downstream of TDP-43 and common to human neurons with TARDBP A90V or M337V mutations.

Generation of patient-specific iPSC lines
Multiple iPSC lines derived from two patients were examined in this study. Two lines were previously generated from a 56-year old male with the TARDBP M337V mutation [35]. In this study, we also generated iPSC lines from a subject with FTD/ALS who was a patient at the Memory and Aging Center, University of California, San Francisco. The patient had an A90V mutation in TARDBP, and his skin biopsy was obtained at the age 75 years. To minimize the difference in genetic background, a skin biopsy was also obtained from a 65-year-old family member who had no TDP-43 mutations or clinical symptoms of FTD and lacked the A90V mutation. The patient and control were tested negative for the common FTD genes including GRN, MAPT and C9ORF72. The patient's clinical symptoms and family history are described in detail in File S1.
Fibroblasts were derived from skin biopsies of both subjects. Genomic DNA was isolated from cultured fibroblasts, and the region flanking the expected TARDBP A90V mutation was PCRamplified and sequenced to confirm the presence of the c.269C.T (A90V) mutation (Fig. 1A). The fibroblasts were reprogrammed with retrovirus expressing OCT4, SOX2, KLF4, and cMYC as described (Fig. 1B) [41]. After 4 to 5 weeks, iPSC colonies were manually picked and transferred to a Matrigelcoated plate. At least 20 clones from each individual were obtained for further characterization.
To identify fully reprogrammed iPSC lines, we performed several assays. First, to confirm that ectopically expressed reprogramming factors were effectively silenced; we used quantitative RT-PCR (qRT-PCR) to measure the endogenous and total levels (expressed from both the endogenous locus and the retroviral vector) of each reprogramming factor. The total level was comparable with that expressed from the endogenous locus, consistent with silencing of the transgene (Fig. 1C). Next, we selected 2 iPSC lines from the family member and 3 lines from the patient. All these lines maintained a normal karyotype ( Fig. 1D and Fig. S1) and all continuously expressed pluripotent stem cell markers including NANOG, SSEA4, OCT4, TRA-1-60, and TRA-1-81 ( Fig. 2A). We then examined the differentiation capability of the iPSCs in vitro through the formation of EBs. Spontaneous differentiation into derivatives of the three embryonic germ layers was confirmed after 1 week of cell culture (Fig. 2B). Altogether, these results indicate that the reprogramming was successful. The five selected iPSC lines were used for further studies below.

Differentiation of iPSCs with TDP-43 mutations into functional neurons
The most prominent clinical phenotypes of the patient during 20 years of slow disease progression after the first diagnosis were all typical symptoms of FTD. To establish a human neuronal model of TDP-43-related FTD, we differentiated iPSCs with the TARDBP A90V and M337V mutations into postmitotic human neurons using a protocol for differentiating human embryonic stem cells into neural progenitor cells expressing the telencephalon marker BF1 [42,43]. Briefly, EBs were formed and cultured in suspension before the induction of rosette formation with basic fibroblast growth factor 2 in serum-free medium containing N2 supplement for 12 days. Rosette cells were isolated and expanded in suspension culture for 21 days to form neurospheres. To differentiate dissociated neural progenitor cells into mostly postmitotic neurons, brain-derived neurotropic factor and glial cell line-derived neurotropic factor were included in the culture medium for 14 days (Fig. 3A).
After neuronal differentiation, about 65% of cells differentiated from control or patient iPSC lines with the TARDBP A90V mutation were immunopositive for the neuronal marker TUJ1 (Fig. 3B, C). A similar percentage of TUJ1-positive cells was obtained from neuronal differentiation of M337V and its control iPSC lines (data not shown). Moreover, the percentages of excitatory and inhibitory neurons were indistinguishable (Fig.  S2), indicating the reproducibility of each differentiation process and the A90V mutation did not affect neuronal differentiation. The morphology and number of neurites were similar in neurons from the FTD/ALS patient and the family member ( Fig. 3D). At 4 to 5 weeks after induction, all recorded neurons differentiated from control and patient iPSCs generated normal action potentials (Fig. 3E). We also determined whether control and patient iPSC neurons could form functional synapses. After terminal differentiation for 4 to 5 weeks, 80% of recorded control neurons, but only about 50% of patient neurons, showed spontaneous synaptic currents (Fig. 3F). Although the amplitude of spontaneous synaptic currents was similar (control neurons: 27.1061.95 pA; patient neurons: 31.5862.48 pA), the frequency of postsynaptic currents was lower in the patient neurons (0.0960.01 versus 0.2660.05 events/second) (Fig. 3F). These results suggest there might be a significant functional difference between control and patient neurons, which should be further investigated in the future.

Patient Neurons with the TARDBP A90V Mutation Show Stress-Dependent Cellular and Molecular Phenotypes
One of the challenges in using iPSCs to model age-dependent neurodegenerative disease is whether one can observe any diseaserelevant phenotypes in cell culture. We hypothesized that under certain cellular stress, patient neurons would exhibit cellular and molecular phenotypes distinct from those of control neurons. First we examined the viability of neurons treated with different inducers of cellular stress. Patient and controls neurons were equally sensitive to sorbitol, an inducer of osmotic and oxidative stress (Fig. S3A) [44], sodium arsenite (Fig. S3B) and tBOOH ( Fig  S3C), both inducers of oxidative stress [45,46]. However, more patient neurons died after treatment with 100 nM staurosporine (STS) (Fig. 4A), an inducer of apoptotic cell death [47].
Redistribution of TDP-43 from the nucleus to the cytoplasm is a characteristic of neurons in brain and spinal cord sections from FTD and ALS patients [9]. Immunostaining to detect the subcellular distribution of TDP-43 in patient neurons revealed that, under normal conditions, TDP-43 was present in the nucleus in the majority of cells (Fig. 4B). However, in response to100 nM STS, TDP-43 was mislocalized from nucleus to cytoplasm in more patient neurons than control neurons (Fig. 4C, 4D). FUS, another RNA binding protein associated with FTD and ALS [46], remained in the nucleus in both control and patient neurons after STS treatment (Fig. 4E, 4F).
Increasing evidence suggests that either increased or decreased TDP-43 expression is associated with neurodegeneration [14,15,24,30]. To examine TDP-43 expression in patient neurons under cellular stress, we measured TDP-43 in the insoluble and soluble fractions of patient neurons with the TARDBP A90V mutation treated with STS. The cells were extracted first with RIPA buffer and then with urea buffer to recover insoluble TDP-43, and analyzed by western blot. The TDP-43 level in the insoluble fraction was similar in untreated control and patient neurons with the TARDBP A90V mutation, but levels were significantly lower in patient neurons after STS treatment (Fig. 5A, B). Similarly, in the soluble fraction, the TDP-43 level was comparable in untreated patient and control neurons (Fig. 5C). However, after STS treatment, TDP-43 ( Fig. 5D) but not FUS ( Fig. S4A) was significantly lower in patient neurons with the TARDBP A90V mutation, consistent with the notion that loss of TDP-43 activity is also associated with neurodegenerative processes. This decrease is likely in part due to transcriptional regulation since TDP-43 mRNA level is reduced in patient neurons after STS treatment (Fig. S4B). This decrease may not be contributed by the difference in protein stability between TDP-43 WT and TDP-43 A90V since STS treatment does not seem to affect the steady state level of transfected TDP-43 A90V compared TDP-43 has been implicated in the miRNA pathway [16], and alterations in the activity of several miRNAs have been associated with neurodegeneration [21]. Among them, miR-9, a brainspecific miRNA whose nucleotide sequence is highly conserved through evolution, is implicated in several neurodegenerative diseases and of great interest [49]. Indeed, the level of miR-9 ( Fig. 6A) was significantly lower in patient neurons with the TARDBP A90V mutation than control neurons only after STS treatment. This decrease did not seem to be due to a defect in the miRNA-processing pathway, since STS also reduced the levels of both pre-miR-9-2 and pri-miR-9-2 in patient neurons (Fig. 6B, C). Moreover, pri-miR-124-1, the most abundant brain-specific miRNA, was not affected (Fig. S6A). Pri-miR-9-2 and pri-miR-124-1 were analyzed here because they are the most abundant among three miR-9 or miR-124 alleles, separately [50,51]. In neurons with the TARDBP M337V mutation, the levels of pre-miR-9-2 and pri-miR-9-2 were also significantly lower after STS treatment than that in neurons derived from two control individuals (Fig. 6D). Interestingly, even without STS treatment, miR-9-2 expression was already lower in neurons with the TARDBP M337V mutation (Fig. 6D), consistent with the more pathogenic nature of the M337V mutation. To demonstrate that the decrease in miR-9a expression resulted directly from reduced TDP-43 function under cellular stress, we used mouse TDP-43specific shRNAs (Fig 6E). TDP-43 knockdown in mouse primary cortical neurons indeed reduced the expression of pri-miR-9-2 ( Fig. 6F) but not miR-124-1 (Fig. S6C).

Discussion
To better understand the molecular pathogenic events in FTD, we used human neurons of telencephalon lineage differentiated from iPSCs with the TARDBP M337V mutation [35] or with the TARDBP A90V mutation as generated and characterized here.
Using this system, we found that patient neurons with the TARDBP A90V mutation are less viable under cellular stress with the kinase inhibitor staurosporine treatment than control neurons. In these patient neurons, the percentage of cells with mislocalized cytoplasmic TDP-43 was increased. The expression of TDP-43 was also significantly decreased under cellular stress induced by STS. Moreover, miR-9 expression is significantly reduced in patient neurons with the TARDBP M337V mutation or in neurons with A90V mutation after STS treatment, revealing another stressrelated molecular defect in patient neurons.
A major advantage of using iPSC-derived human neurons to study neurodegeneration is the presence of other patient-specific genetic factors in these cells. Several hundred genetic mutations may exist in each patient's genome and their collective contribution to disease process can be assessed. The A90V mutation is rare and considered to be a risk factor for FTD/ALS [52,53]. Thus, examining the A90V mutation alone in animal models without other genetic modifiers may not be sufficient to reveal disease mechanisms. The cellular and molecular phenotypes observed here in iPSC-derived patient neurons are likely due to the overall genetic and epigenetic makeup of the patient including contributions from the A90V mutation. Although we used a family member as negative control to minimize the difference in genetic background, we understand the results obtained here from one patient need to be interpreted cautiously and we are aware of the challenges of generating more iPSC lines from other patients with the same rare mutation. So it remains to be seen whether neurons derived from other patients with the same or different TDP-43 mutations exhibit these cellular phenotypes under stress.
It is interesting to note that the A90V patient's disease progression after the initial diagnosis was unusually slow and lasted over two decades, while the partial loss of TDP-43 expression in the patient's iPSC-derived neurons occurred only after the treatment with a high concentration of STS. This decrease in TDP-43 level could be due to the effects of A90V mutation on nuclear import or in combination with other genetic factors. These findings support the hypothesis that the A90V mutation is pathogenic but with a slower kinetics. This observation stands in contrast to human neurons differentiated from patients with the M337V mutation where increased TDP-43 aggregates were observed even without stress treatment [35,36], consistent with the notion that this mutation is more pathogenic.
Our results also indicate that miR-9 is specifically dysregulated in stressed patient neurons with the TARDBP A90V mutation, likely as a direct result of the reduced TDP-43 level observed in these neurons, as shRNA-mediated knockdown of TDP-43 in mouse primary cortical neurons also reduced miR-9 expression (Fig. 6). In human neurons containing the more pathogenic M337V mutation, miR-9 expression is also reduced even without STS treatment (Fig. 6). However, it is worth noting that human neurons in the normal culture conditions are already under cellular stress. Although total level of TDP-43 may increase in human neurons containing the TARDBP M337V mutation once the inter-patient variability is taken into account [36], a number of studies indicate that the M337V mutation results in the partial loss  [35,36]. Thus, normal function of TDP-43 is compromised in human neurons containing the M337V mutation. Taken together, these findings support the notion that partial loss of TDP-43 normal function is also associated with neurodegeneration in FTD/ALS.
A role for TDP-43 in miRNA regulation has been demonstrated in previous studies [16][17][18]56]. Similarly, partial loss of FUS function also results in a reduced miRNA biogenesis [57]. A prominent feature of TDP-43 pathology is the depletion of nuclear TDP-43 in many affected neurons, suggesting a possible significant pathogenic role for the loss of its normal nuclear function [9]. Moreover, mutations in another miRNA regulating protein FUS cause ALS [12,13] and wild-type FUS forms inclusions in 5% of FTD cases [48]. Thus, misregulation of miRNAs and miR-9 in particular may be a common downstream molecular event of both disorders. TDP-43 regulates hundreds of targets [25,58,59]. To what extent miR-9 and specific miR-9 targets contribute to TDP-43 mediated neurodegeneration remains to be determined. The availability of different FTD/ALS patients-specific iPSC lines generated here and from other labs will undoubtedly further facilitate studies on the miRNA pathway and other pathogenic mechanisms.

Cell Culture and Generation of Human iPSCs
Skin biopsies were collected from a 75-year-old man who had a TARDBP missense mutation (A90V) and from an unaffected 65year-old family member who lacked the mutation. The study was approved by the Institutional Review Board and Ethics Committees at the University of California, San Francisco (UCSF), and written informed consent was obtained in all cases. To protect patient family privacy, no additional information is presented here. Primary dermal fibroblasts derived from the explants were cultured in DMEM supplemented with 10% FBS, nonessential amino acids, and penicillin/streptomycin and reprogrammed into iPSCs using four transcription factors (OCT3/4, SOX2, KLF4, and c-MYC) as described (41). Five weeks after viral transduction, colonies were transferred to 12-well plates coated with Matrigel (BD Biosciences) and cultured in mTeSR1 medium (StemCell Technologies). For expansion, colonies were dissociated with Accutase and plated on 6-well plates. We also used iPSC lines from a 56-year-old man with the TARDBP M337V mutation and from two healthy controls [35].

Karyotyping
Karyotyping analysis was carried out by the Cytogenetic Laboratory at the University of Massachusetts Memorial Medical Center (Worcester, MA). Cells were analyzed by standard Gbanding technique, and results were interpreted by specialists in the Cytogenetic Laboratory.

Neuronal Differentiation
iPSCs with either TARDBP A90V mutation generated in this study or the M337V mutation [35] were differentiated as described [37,42] with some modifications. Embryoid bodies (EBs) were cultured in suspension for 6 days and attached to dishes coated with poly-L-ornithine/laminin. 12-14 days later, rosettes were collected and grown in suspension as neurospheres. After 3 weeks in culture, the neurospheres were dissociated with Accutase, and the neurons were plated on glass coverslips coated with poly-L-lysine/laminin. In most experiments, neurons 2-weeks after the neurosphere stage were used. For spontaneous differentiation in vitro, EBs were transferred to Matrigel-coated plates after 6-8 days as a floating culture for 8 days. Migrating cells from EBs were stained for markers of the three germ layers and analyzed by immunofluorescence microscopy. For the analysis on M337V neurons, two independent differentiation experiments were

Immunocytochemistry
Cells were fixed in 4% paraformaldehyde for 10 minutes, permeabilized with 0.3% Triton X-100 for 10 minutes, and blocked with 10% BSA for 1 hour at room temperature. Cultures were then stained overnight at 4uC with goat anti-NANOG

MTT Assay
After 14 days in culture, human neurons were treated for 24 hours with the following stressors: staurosporine, t-butyl hydroperoxide, sodium arsenite, and DMSO. The sorbitol treatment was for 30 minutes. MTT assays were performed with CellTiter 96 Non-Radioactive Cell Proliferation Assay kit

Quantitative RT-PCR
Total RNA was extracted and purified with the miRNeasy kit (Qiagen) and the Turbo DNA-free kit (Ambion). RNA (300 ng) was used for reverse transcription reaction with TaqMan reverse transcription reagent kits (Applied Biosystems). mRNA and miRNA expression levels were quantified with SYBR Green Master Mix, TaqMan microRNA assays, or Taqman Universal PCR master mix (Applied Biosystems). Ct values for each sample and gene were normalized to GAPDH (mRNA) or U6 (miRNA). Primer sequences are shown in Table S1.

Lentiviral Production and Mouse Primary Neuronal Cultures
HEK293T cells were transfected with a mix of 2.6 mg plasmids expressing TDP-43-specific shRNA (Sigma), 26 ml lentiviral packaging mix (Sigma), and 16 ml FuGENE 6 (Promega). Thirty-six to forty-eight hours after transfection, medium containing viruses was filtered and centrifuged at 23,000 rpm for 2 hours at 4uC.The supernatant was removed, and the pellet containing the lentiviral particles was dissolved in 100 ml HBSS overnight and stored at 280uC. Primary neuronal cultures were obtained from C57/B6 mice (E18). Cells were plated at a density of 70,000 cells/ well in 12-well plates and transduced with lentiviral particles after 3 days in vitro.

Electrophysiology
Whole-cell perforated patch recordings were performed on FTD/ALS (n = 16) and control (n = 15) 4-to 5-week-old neurons grown in neuron medium containing 0.5% FBS on acid-etched coverslips. The recording micropipettes (tip resistance 3 to 6 MV) were tip-filled with internal solution composed of 115 mM Kgluconate, 4 mM NaCl, 1.5 mM MgCl 2 , 20 mM HEPES, and 0.5 mM EGTA (pH 7.4) and back-filled with the same internal solution containing 200 mg/ml amphotericin B (Calbiochem). Recordings were made with an Axopatch 200B amplifier (Axon Instruments). Signals were sampled at 10 kHz and filtered at 2 kHz. Whole-cell capacitance was fully compensated; series resistance was uncompensated but monitored during the experiment by the amplitude of the capacitive current in response to a 5-mV pulse. The bath was constantly perfused with fresh HEPESbuffered saline composed of 115 mM NaCl, 2 mM KCl, 10 mM HEPES, 3 mM CaCl 2 , 10 mM glucose, and 1.5 mM MgCl 2 (pH 7.4). For voltage-clamp recordings, cells were clamped at 260 to 280 mV; Na + currents and K + currents were stimulated by voltage step depolarization. Command voltage varied from 250 to 20 mV in 10-mV increments. For current-clamp recordings, induced action potentials were stimulated with current steps from 20.2 to 0.5 nA. All recordings were performed at room temperature.

Stable Cell Lines with Inducible TDP-43 Expression
Wildtype TDP-43 cDNA (HindIII-BamHI) was generated by PCR amplification and 2XFLAG tag was added at its C-terminal using primers (59-GCGAAGCTTATGTCTGAATATATTC-GGGTAAC-39 and 59-GCGGATCCTTACTTGTCATCGT-CGTCCTTGTAGTCCTTGTCATCGTCGTCCTTGTAGTC CATTCCCCAGCCAGAAGACTTAG-39). The TDP-43 PCR fragment was inserted into the polylinker of pcDNA5-TetO (Invitrogen) to make TDP-FLAG plasmid. A90V mutation was made by primer based Quick Change Method (Stratagene). 2 mg TDP-FLAG or A90V-FLAG constructs were transfected into HEK293 cells together with 0.2 mg of pOG44 using 7 ml of TransIT-293 (Mirus). Cell lines stably expressing TDP-FLAG or A90V-FLAG were generated by treating the cells with several rounds of Blasticidin (15 mg/ml) and Hygromycin (100 mg/ml) selection. To induce transgene expression, tetracycline (Tet; 500 ng/ml) (Sigma) was added at different time points in the absence or presence of STS.
To examine TDP-43 stability, 500 ng/ml of tetracycline was added to the stable cell lines to induce the expression of TDP-FLAG or A90V-FLAG. After 48 h of induction, cells were washed twice by 1XPBS and then replenished with fresh media without tetracycline. Cells were collected at different time points (4-48 hrs) after removing tetracycline. Harvested cells were lysed with RIPA buffer or trizol (Ambion) to get whole cell protein extract or total RNAs separately.

Statistical Analysis
Values are presented as mean 6 SEM. Statistical significance was assessed by ANOVA or t test. P,0.05 was considered significant. In this experiment, the sample collected at 48 hr after tetracycline induction appears to have higher protein level of TDP-43 WT than TDP-43 A90V, which is due to higher transfection since the wildtype mRNA is also higher (not shown).  File S1 Patient information. The patient's clinical symptoms and family history are described in detail. (DOCX)