In Huntington’s disease (HD), expansion of CAG codons in the huntingtin gene (HTT) leads to the aberrant formation of protein aggregates and the differential degeneration of striatal medium spiny neurons (MSNs). Modeling HD using patient-specific MSNs has been challenging, as neurons differentiated from induced pluripotent stem cells are free of aggregates and lack an overt cell death phenotype. Here we generated MSNs from HD patient fibroblasts through microRNA-based direct neuronal conversion, bypassing the induction of pluripotency and retaining age signatures of the original fibroblasts. We found that patient MSNs consistently exhibited mutant HTT (mHTT) aggregates, mHTT-dependent DNA damage, mitochondrial dysfunction and spontaneous degeneration in culture over time. We further provide evidence that erasure of age stored in starting fibroblasts or neuronal conversion of presymptomatic HD patient fibroblasts results in differential manifestation of cellular phenotypes associated with HD, highlighting the importance of age in modeling late-onset neurological disorders.
HD is a progressive neurodegenerative disorder caused by an abnormal expansion of CAG codons in the huntingtin (HTT) gene1,2. HD symptoms typically manifest in midlife and may include motor deficits, psychiatric symptoms and cognitive decline3. While healthy individuals have an average HTT CAG tract size of 17–20 repeats, HD patients have an expansion of 36 or more CAGs4. Moreover, CAG repeat length is directly correlated to the severity of the disease and inversely related to the age of onset5. Expanded CAG trinucleotides encode a polyglutamine stretch that can accumulate into proteinaceous cytoplasmic and intranuclear aggregates that are thought to be neurotoxic3, although the formation of inclusion bodies has also been suggested to be a neuroprotective mechanism6. HD pathology is characterized by the selective degeneration of MSNs while other neuronal subpopulations are relatively spared7.
Because of the clinical importance of MSNs in HD, differentiation protocols have been developed to generate MSNs from induced pluripotent stem cells (iPSCs)8,9. However, modeling HD with iPSC-derived neurons often requires additional cellular insults to detect HD-relevant phenotypes10,11,12,13. For example, neurons differentiated from patient iPSCs showed elevated levels of caspase activity only upon trophic factor withdrawal, treatment with hydrogen peroxide, or high levels of glutamate, but otherwise display no overt cell death phenotype9,10,13. Additionally, iPSC-derived neurons do not exhibit mHTT aggregates even after the addition of cellular stressors11, and other studies required culturing cells for at least 6–8 months and treatment with proteasome inhibitors before aggregates could be detected12,14. Therefore, an alternative reprogramming approach that generates an enriched population of patient-derived MSNs that more robustly display HD phenotypes will greatly facilitate the modeling of HD.
Ectopic expression of the brain-enriched microRNAs (miRNAs) miR-9/9* and miR-124 (miR-9/9*-124) in human adult fibroblasts has been shown to directly convert fibroblasts to neurons through extensive chromatin reconfigurations. The miRNA-9/9*-124-induced neuronal state, generated by miRNA instruction in switching the activities of chromatin remodeling complexes15,16, allows terminal selector genes to guide neuronal conversion to produce a highly enriched population of specific neuronal subtypes16, specifically MSNs with CTIP2, DLX1, DLX2 and MYT1L (CDM)17. In contrast to neurons differentiated from iPSCs, in which the age stored in original fibroblasts is erased during the induction of pluripotency18,19, directly converted neurons have been shown to retain age-associated marks of starting adult human fibroblasts, including the epigenetic age (also known as the epigenetic clock19), oxidative stress, DNA damage, miRNAome, telomere lengths and transcriptome20,21. This feature offers potential advantages in modeling adult-onset disorders; however, the value of MSNs directly converted from HD patient fibroblasts in modeling HD remains to be determined.
Here we propose the generation of HD patient-derived MSNs (HD-MSNs) through miR-9/9*-124-CDM-based conversion of fibroblasts as a cellar model of HD. We focused on HD samples with 40–50 CAG repeats, which represent the majority of HD cases4. HD-MSNs recapitulated essential HD-associated phenotypes, including the formation of mHTT aggregates, DNA damage, spontaneous neuronal death in culture, and a decline in mitochondrial function. We further provide evidence that cellular age is an essential component underlying the manifestation of HD phenotypes. By inducing HD-fibroblasts into iPSCs and redifferentiating them back into embryonic fibroblasts for neuronal conversion, we discovered that age-associated reduction in protein homeostasis levels was primarily responsible for mHTT aggregation in adult HD-MSNs. Furthermore, MSNs reprogrammed from presymptomatic HD patients were less vulnerable to mHTT-induced toxicity than MSNs reprogrammed from symptomatic patients, despite comparable levels of mHTT aggregates. Finally, modifying the terminal neuronal cell fate to cortical neurons alleviated mHTT-induced toxicity. These results underscore the importance of direct neuronal conversion for modeling age-related phenotypes of late-onset diseases with specific neuronal subtypes.
Generation of MSNs from HD patient fibroblasts
We first tested the efficacy of miR-9/9*-124-CDM-based neuronal conversion in fibroblasts from ten symptomatic HD patients, including males and females ranging from 6 to 71 years of age with various CAG repeat expansions (Supplementary Table 1). HD-fibroblasts could be directly reprogrammed to MSNs regardless of age or CAG repeat number (Fig. 1 and Supplementary Fig. 1), and we focused our analyses on patient samples with CAG repeats lower than 50, as this range reflects most adult-onset cases3 but remains understudied. We validated MSN conversion in three independent HD patient fibroblast samples containing 40, 43 or 44 CAG repeats (HD.40, HD.43 and HD.44) and their respective age- and sex-matched healthy controls (Ctrls) with 19, 17 or 18 CAG repeats (Ctrl.19, Ctrl.17 and Ctrl.18) (Fig. 1). At post-induction day 30 (PID 30), HD-MSNs expressed the neuronal markers TUBB3, NeuN and MAP2, the GABAergic neuron marker GABA, and the MSN marker DARPP-32 (Fig. 1a and Supplementary Fig. 1). We found no significant differences in the reprogramming efficiency between HD and control samples, generating approximately 90% MAP2- and 70–80% GABA- and DARPP-32-positive neurons (Fig. 1b,c). Furthermore, CAG repeat lengths remained stable after neuronal conversion (Supplementary Fig. 2).
We carried out whole-cell recordings to determine functional properties of HD-MSNs in comparison to Ctrl-MSNs. All recorded cells displayed multiple action potentials and robust inward and outward currents upon stimulation (Fig. 2). HD- and Ctrl-MSNs displayed spontaneous action potentials at similar frequencies and had similar action potential thresholds (Fig. 2a–c). Notably, more HD-MSNs than Ctrl-MSNs fired multiple action potentials (Fig. 2d). However, all other passive membrane properties recorded did not differ significantly between HD- and Ctrl-MSNs (Fig. 2c). To further access electrophysiological properties under the same recording condition, HD-MSNs (HD.40, HD.43 and HD.44) and Ctrl-MSNs (Ctrl.17, Ctrl.18 and Ctrl.19) were cocultured and recorded on the same coverslip (Supplementary Fig. 3a). At PID 35, all six reprogrammed lines fired action potentials, but, as in cells that were cultured separately, a higher percentage of HD-MSNs fired multiple action potentials than Ctrl-MSNs upon current injection (Supplementary Fig. 3b,c) and displayed spontaneous generation of action potentials (Supplementary Fig. 3d). Whereas passive membrane properties measured remained similar between HD- and Ctrl-MSNs, we detected increased current responses with high-voltage stimulus in HD-MSNs (Supplementary Fig. 3e–g), reflecting increased excitability and firing complexity in HD-MSNs.
To further analyze the acquisition of MSN fate, we performed RNA sequencing (RNA-seq) at PID 32 and compared the gene expression profiles between fibroblasts and converted neurons in HD and control samples. Analysis of 15 representative fibroblast-associated genes and 53 genes highly enriched in the striatum revealed the successful acquisition of MSN fate in neurons converted from HD and control samples (Fig. 3a). To identify genes potentially dysregulated in HD-MSNs, we carried out transcriptional analysis of seven independent HD-MSN and five Ctrl-MSN samples (Supplementary Table 1). Principal component analysis indicated sample segregation based on the genotype (mHTT vs. healthy control) as well as the sex of sample donors (Fig. 3b). Analysis of protein-coding genes revealed 1,127 differentially expressed genes (DEGs) between sex-matched HD-MSNs and Ctrl-MSNs (false-discovery rate (FDR) ≤ 0.01 and log2(fold change) (LFC) ≥ 0.5 or ≤ –0.5) (Fig. 3c). Gene ontology analysis showed DEGs in HD-MSNs to be significantly enriched for genetic networks associated with cell differentiation (P = 1.51 × 10−10), neurotransmission (P = 1.31 × 10−8), calcium signaling (P = 5.31 × 10−6), HD (P = 7.22 × 10−4) and apoptosis (P = 1.19 × 10−2) (Fig. 3d and Supplementary Table 2). Several DEGs identified in HD-MSNs have been previously implicated in HD. For example, we detected the upregulation of matrix metalloproteinase 9 (MMP9), which has been shown to be increased in postmortem human HD-affected brains22 and to significantly decrease the survival of striatal neurons23. Moreover, our analysis revealed the downregulation of huntingtin-associated protein-1 (HAP1) in HD-MSNs (LFC –0.55 and FDR 5.04 × 10−4; Fig. 3c), which has previously been shown to antagonize mHTT-mediated cytotoxicity and enhance cell viability24. Additionally, we detected the downregulation of 7-dehydrocholesterol reductase (DHCR7) in HD-MSNs (LFC –0.71 and FDR 2.8 × 10−5; Fig. 3c), an enzyme previously shown to have reduced expression in patients and mouse models of HD and thought to be involved in HD-specific metabolic pathway alterations25,26. Notably, many DEGs upregulated in HD-MSNs were associated with neurophysiological processes, such as the voltage-gated potassium channel subunit KCNA4 (LFC 1.15 and FDR 1.1 × 10−4) (Fig. 3c) and several subunits of GABA type-A receptors and AMPA receptors, suggesting increased neurotransmission in HD-MSNs (Fig. 3c,d and Supplementary Fig. 4). We also detected upregulation of α-synuclein (SNCA) (LFC 1.15 and FDR 1.1 × 10−4), an aggregation-prone protein shown to accumulate in mHTT polyglutamine inclusions27. Overexpression of α-synuclein has been reported to accelerate the onset of HD symptoms in multiple mouse models28. Further, we found NTRK2 (also known as TRKB), the main receptor for brain-derived neurotrophic factor (BDNF), to be downregulated in HD-MSNs (LFC –0.77 and FDR 7.44 × 10−3) (Fig. 3c). The loss of BDNF in HD pathology has been investigated extensively and proposed to be critical in the degeneration of MSNs29,30; our results indicate that mHTT may induce downregulation of BDNF signaling at the receptor level in HD-MSNs. The impairment of TRKB receptor was suggested to mediate postsynaptic dysfunction of MSNs in mouse models of HD, although changes in NTRK2 mRNA levels were not detected in HD mouse models31. Our analysis also uncovered DEGs with no previous association with HD. For instance, SP9, a zinc finger transcription factor recently shown to be necessary for the maintenance and survival of striatopallidal MSNs32, was significantly downregulated (LFC –1.7 and FDR 1.49 × 10−8) in HD-MSNs. Several of these genes, including SP9, HAP1 and NTRK2, were further validated by quantitative PCR (qPCR) in reprogrammed MSNs at PID 35 (Supplementary Fig. 4).
Mutant HTT aggregates in MSNs directly converted from HD-fibroblasts
Because polyglutamine expansion in HTT leads to the formation of insoluble structures of aggregated mHTT, or inclusion bodies3, we performed immunocytochemistry and ultrastructural and biochemical analyses to assess whether HD-MSNs would display mHTT inclusions. Notably, HD-MSNs exhibited mHTT aggregates, in contrast to their corresponding fibroblasts or Ctrl-MSNs (Fig. 4a–c). Non-reprogrammed HD-fibroblasts were devoid of detectable aggregates even upon cellular insults, including the induction of oxidative stress with hydrogen peroxide or cellular senescence by serial passaging (Supplementary Fig. 5a,b). Furthermore, mimicking reprogramming using CDM factors with a nonspecific microRNA, a condition previously shown to be ineffective for neuronal conversion17, did not lead to detectable mHTT aggregates (Supplementary Fig. 5c), demonstrating the specificity of the aggregation phenotype to successfully reprogrammed neurons. Cytoplasmic (Fig. 4c,d) and intranuclear (Fig. 4d) mHTT aggregates were evident in HD-MSNs reprogrammed from all HD patient samples as early as PID 14 when analyzed with antibodies (MW8 and EM48) that selectively recognize aggregated mHTT inclusion bodies colocalized with ubiquitin (Fig. 4f and Supplementary Fig. 5d–f). HD models that have been engineered to overexpress mHTT with a large number of CAG repeats report high levels of cells with inclusions. However, studies analyzing postmortem HD patient brains found that only up to 10% of MSNs showed inclusion bodies33, similar to the levels we detected in HD-MSNs (Fig. 4e). Examining the ultrastructure of immunogold-labeled mHTT inclusions by transmission electron microscopy (TEM) in converted MSNs (HD.40 and Ctrl.19) plated in microdishes (Fig. 4g) revealed the presence of nanogold particles labeling mHTT aggregates, as well as structures of fibrillar morphology, found only in HD-MSNs (Fig. 4h and Supplementary Fig. 6a,d). We further confirmed the expression of mHTT by immunoblot analysis at PID 28 in three HD-MSN samples (HD.42, HD.46 and HD.47; Supplementary Fig. 7) using the monoclonal antibody MW1,which was shown to specifically detect the polyglutamine domain of HTT exon 1 while showing no detectable binding to normal HTT34. Additionally, insoluble aggregated HTT could be detected in all reprogrammed HD samples, but not in Ctrl-MSNs (Ctrl.16 MSNs) using the HTT aggregate-specific monoclonal antibody MW834 (Supplementary Fig. 7). In our TEM studies, we found immunogold particles compartmentalized inside autophagosomes, cytosolic double-membrane vesicles involved in macroautophagy (Fig. 4i). This suggests that autophagic vacuoles can recognize and trap cytosolic mHTT inclusions in HD-MSNs harboring low CAG repeats. In fact, we observed colocalization of mHTT and LC3-II, a well-established marker of autophagosomes, in HD-MSNs reprogrammed from three independent HD lines, which is similar to previously reported findings in a HD mouse model35 (Fig. 4i and Supplementary Fig. 6f).
Induction of pluripotency alters mHTT aggregation propensity
Because our findings contrasted with previous studies that report the lack of mHTT aggregates in iPSC-derived neurons from HD patients, we tested whether altering the cellular state of adult HD-fibroblasts to an embryonic-like stage18 would affect the aggregation propensity of mHTT in HD-MSNs. We derived HD-iPSCs from adult HD fibroblasts and differentiated these iPSCs back into fibroblasts to generate human embryonic fibroblasts (HEFs)18 (Fig. 5a and Supplementary Fig. 8a). Briefly, HD.40 fibroblasts were transduced with Sendai viral vectors to express the four reprogramming factors (OCT4, SOX2, KLF4 and c-MYC), which resulted in integration-free iPSCs that expressed markers of pluripotency and retained a normal karyotype and the same number of CAG repeats (Supplementary Fig. 8b–e). iPSC-derived HD-HEFs expressed fibroblast markers fibronectin and vimentin (Fig. 5b). We confirmed that HD-HEFs exhibited cellular markers associated with the re-induction of an embryonic state, including high expression of the nuclear lamina-associated protein 2α (LAP2α)18 (Supplementary Fig. 8f). Upon direct conversion of HD-HEFs to MSNs (human embryonic MSNs, heMSNs), little to no aggregated mHTT was detectable in HD-heMSNs at PID 21 (Fig. 5c,d). We further verified these results using another, independent iPSC line from a symptomatic 37-year-old HD patient with 50 CAG repeats in HTT (HD.50; Supplementary Fig. 8g,h). We then investigated differences between adult MSNs and heMSNs in mHTT aggregation propensity to elucidate the contribution of aging to protein aggregation in HD. We began by ectopically expressing EGFP fused to 23 or 74 polyglutamine repeats (GFP-23Q or GFP-74Q) in either HD adult fibroblasts or HD-HEFs to track protein aggregation by live imaging (Fig. 5e). While the expression of GFP-23Q stayed diffuse, we observed a rapid rate of GFP-74Q aggregate formation (Fig. 5e) in adult fibroblasts, with over 70% of HD-fibroblasts displaying GFP-74Q aggregates and increased fluorescence density indicating the formation of inclusion bodies after 24 h, whereas in HEFs aggregates were only visible in fewer than 10% of cells at each time point analyzed (Fig. 5f). Given that aggregates can be induced by treatment with proteasome inhibitors in iPSC-derived MSNs12, we postulated that a higher proteasome activity in HEFs likely prevented GFP-74Q from forming aggregates. We performed qPCR analysis for 17 genes associated with the ubiquitin–proteasome system (UPS), the main protein quality control machinery in the cell, in HD adult fibroblasts and two lines of HEFs differentiated from two independent HD-iPSC clones. We found eight genes consistently upregulated in HEFs (Supplementary Fig. 9). The upregulated UPS genes included the heat-shock transcription factor HSF1, a protein that regulates the expression of genes involved in protein homeostasis36 and is reduced in the striatum of HD patients37. To directly test whether proteasome activity was preventing the formation of inclusions in HEFs, we treated GFP-74Q-expressing HEFs with the proteasome inhibitor lactacystin. Lactacystin-treated HEFs had significantly more cells bearing inclusions than DMSO-treated HEFs (Fig. 5g).
Although iPSCs have been previously shown to possess higher proteasome activities than their originating fibroblasts, differentiation of iPSCs into neurons also was shown to reduce proteasome activity38. To determine whether changes in the proteasome activity could account for the detection of mHTT aggregates in MSNs but not in heMSNs, we assessed the functional activity of the proteasome in converted MSNs with the fluorogenic peptide LLVY-AMC. We discovered that proteostasis was collapsed in HD-MSNs in comparison to heMSNs, which retained the proteasome activity comparable to that in iPSCs (Fig. 5h,i). Aggregation propensity was not dependent on HTT mutation, as evidenced by similar levels of proteostasis between HD and control samples (Fig. 5f,i). To explore the age-dependent collapse in proteostasis, we analyzed the expression of 300 UPS-associated genes in a previously published dataset generated by the transcriptional profiling of MSNs converted from fibroblasts of young (3 d, 5 months and 1 year old) or old (aged 90, 92 and 92 years old) donors20. By comparing gene expression in young versus old fibroblasts and MSNs, we determined that fibroblasts did not display drastic changes in the expression of UPS-related genes with age, but MSNs from older individuals showed a dramatic increase in the number of downregulated UPS-related genes (Fig. 5j and Supplementary Fig. 9). In fact, gene ontology analysis of downregulated genes in old MSNs showed significant enrichment for the positive regulation of proteolysis (P = 2.01 × 10−2). These data suggest that the proteostasis collapse in MSNs, but not in originating fibroblasts or iPSC-derived neurons, depends on the cellular age of converted neurons.
mHTT-mediated DNA damage and spontaneous degeneration
Because aging contributes to the onset of HD, we tested whether direct conversion would allow detection of spontaneous neuronal death in HD-MSNs. We first measured DNA damage in HD-MSNs converted from three independent HD patients in comparison to starting fibroblasts and Ctrl-MSNs. At PID 30, HD-MSNs exhibited increased oxidative DNA damage as determined by levels of 8-hydroxy-2′-deoxyguanosine (8-OHdG) (Fig. 6a,b), as well as increased double-stranded breaks as assessed by the presence of nuclear 53BP1-positive foci (Fig. 6c,d). Analysis by single-cell gel electrophoresis that visualizes the migration of broken DNA strands from individual agarose-embedded cells (also known as the comet assay) showed a marked increase in comet tail lengths in comparison to Ctrl-MSNs at PID 30 while no significant difference was detected between HD and control fibroblasts (Fig. 6e,f). Next we quantified spontaneous cell death in three controlled pairs of HD- and Ctrl-MSNs using SYTOX green, a nucleic acid stain impermeable to live cells, at multiple time-points during reprogramming (Fig. 6g,h). Cell death levels were comparable at PID 30, but increased drastically for HD-MSNs in relation to their controls at PID 35 and 40 (Fig. 6h), further evidenced by a stark reduction of DARPP-32-positive HD-MSNs (Supplementary Fig. 10). The detected DNA damage depended on HTT, as AAV-shRNA-mediated reduction of HTT significantly reduced 8-OHdG levels and number of 53BP1 foci in HD-MSNs (Fig. 6i). Given that the neuronal death of HD-MSNs was preceded by extensive DNA damage, we tested whether HD-MSN cell death would be amenable to pharmacological intervention by treating HD-MSNs with KU60019, an inhibitor of ataxia-telangiectasia mutated (ATM) kinase. ATM is a central regulator of the DNA damage response activated upon DNA damage or oxidative stress to induce apoptosis, and KU60019 has been previously reported to reduce mHTT-induced cell death39. Consistent with these findings, we found that the treatment of HD-MSNs with 0.5 μM KU60019 significantly reduced levels of spontaneous and stress-induced neuronal death in HD-MSNs (Supplementary Fig. 10).
We also discovered that the neuronal death seen in HD-MSNs was responsive to genetic perturbations. For instance, we found SP9, a transcription factor necessary for the maintenance and survival of MSNs32, to be significantly downregulated in HD-MSNs by RNA-seq analysis (Fig. 6j) and further validated it by qPCR in five independent HD-MSN samples (Fig. 6k). We then cloned the cDNA of SP9 downstream of the ubiquitous EF1α promoter in a lentiviral vector to allow the consistent expression of SP9 in HD-MSNs. At PID 14 of neuronal conversion, three HD-MSN (HD.40, HD.42 and HD.46) and three Ctrl-MSN (Ctrl.16, Ctrl.17b and Ctrl.19) samples were transduced with lentivirus carrying the SP9 cDNA construct, cultured until PID 35, and then assayed for cell death with SYTOX green. We found that restoring SP9 expression in HD-MSNs reduced cell death to levels indistinguishable from those of controls (Fig. 6l). Although loss of SP9 has been previously shown to lead to apoptosis of MSNs in mice32, further studies are needed to probe the neuroprotective mechanism of this transcription factor and its potential role in HD pathogenesis. Our results show that directly converted HD-MSNs could potentially serve as a useful platform for identifying pharmacological and genetic factors that have therapeutic potential for treating HD.
Mitochondrial dysfunction, oxidative stress and metabolic deficits in HD-MSNs
Ultrastructural analysis in HD-MSNs and Ctrl-MSNs (HD.40 and Ctrl.19) revealed HD-MSNs to be enriched with lipofuscin granules, aging pigments that accumulate due to incomplete lysosomal degradation of damaged mitochondria, which are commonly detected in the brains of HD patients35,40 (Supplementary Fig. 6b,c). HD-MSNs also exhibited high levels of mitophagy, a selective degradation of dysfunctional mitochondria typical of apoptotic cells41 (Supplementary Fig. 6e). HD-MSNs further showed the accumulation of cytoplasmic lipid droplets, which are known to be caused by oxidative stress and mitochondrial dysfunction42 (Supplementary Fig. 6e). To gain insight into the mitochondrial and metabolic dysfunction present in HD-MSNs, we reprogrammed six lines (HD.42, HD.46, HD.47, Ctrl. 19, Ctrl.20, Ctrl.17c and Ctrl. 18b; Supplementary Fig. 11) to quantify mitochondrial functions. We first determined, using the mitochondrial indicator MitoTracker Red, that the total pool of mitochondria was unchanged between HD- and Ctrl-MSNs (Fig. 7a). We next assessed changes in the mitochondrial membrane potential with TMRE, an indicator of active and polarized mitochondria, and found significantly lower levels of TMRE signal in HD-MSNs, indicating decreased membrane potential of mitochondria (Fig. 7b). Increased production of reactive oxygen species by mitochondria is thought to be a major cause of oxidative stress in HD and a critical component in the progression of the disease43. Live imaging of HD-MSNs with the superoxide indicator MitoSOX Red revealed significantly higher levels of reactive oxygen species in HD-MSNs (Fig. 7c). HD-MSNs displayed significantly larger lipid droplets than controls as measured by the lipid dye BODIPY 498/503 (Fig. 7d). Since our results indicated impaired mitochondrial health, we measured levels of mitophagy in HD-MSNs (Supplementary Fig. 11a). Converted MSNs were labeled with MitoTracker and immunostained for the autophagosome marker LC3-II for colocalization analysis. Two out of three HD-MSNs derived from independent patients showed higher cytoplasmic LC3-II than controls, although the difference was not statistically significant (Supplementary Fig. 11b). However, we found a greater percentage of mitochondria and autophagosome colocalization in HD-MSNs (Supplementary Fig. 11c). Collectively, these results point to substantial mitochondrial dysfunction in HD-MSNs.
Differential vulnerability of neuronal subtypes to mHTT toxicity
Although HTT is ubiquitously expressed throughout the brain, mHTT leads to selective degeneration of MSNs and, to a lesser extent, cortical neurons as the disease progresses44. Human postmortem studies have shown that, at a stage when neuronal loss is low in the cortex but high in the striatum, mHTT aggregates are more common in the cortex than in the striatum33. We hypothesized that conversion of HD fibroblasts to cortical neurons (HD-CNs) could model the selective vulnerability of HD-MSNs to neurodegeneration. We transduced control and HD-patient fibroblasts either with miR-9/9*-124-CDM or with miR-9/9*-124 in conjunction with NeuroD2, ASCL1 and MYT1L (DAM) (miR-9/9*-124-DAM), a transcription factor cocktail shown to guide the neuronal conversion to cortical neurons45 (Supplementary Fig. 12a–c). Surprisingly, HD-CNs exhibited lower levels of DNA damage (Supplementary Fig. 12d,e) and cell death than HD-MSNs (Supplementary Fig. 12f) but a higher level of mHTT aggregation, (Supplementary Fig. 12g), suggesting that cellular properties intrinsic to MSNs render them differentially vulnerable to neurodegeneration.
Manifestation of HD cellular phenotypes is dependent on patient age
The maintenance of aging signatures upon neuronal conversion has long been postulated to be an important advantage of using directly converted patient neurons to model late-onset diseases. However, no studies have provided empirical evidence that age information stored in a donor’s somatic cells contributes to the manifestation of disease-related phenotypes in converted neurons. Even though our findings from HD-heMSNs were insightful (Fig. 5), to further evaluate the significance of cellular age in HD phenotype manifestation in HD-MSNs we investigated the properties of MSNs reprogrammed from HD-fibroblasts sampled before disease onset. We acquired six fibroblast lines from presymptomatic HD patients (pre-HD), sampled 13 to 17 years before the onset of clinical symptoms, with CAG tract sizes of 42–49 repeats (Supplementary Table 1). All six pre-HD fibroblasts were reprogrammed using miR-9/9*-124-CDM to generate MSNs (pre-HD-MSNs), alongside fibroblasts from three controls and three symptomatic HD patients (Fig. 8a,b). Pre-HD-MSNs were less vulnerable to mHTT-induced toxicity at PID 35, with lower levels of cell death and oxidative DNA damage (Fig. 8c,d). Notably, pre-HD-MSNs still contained mHTT aggregates at a level similar to that in symptomatic HD-MSNs (Fig. 8c,d). These results are noteworthy as they directly show that the age-dependent onset of HD can be modeled with directly converted HD-MSNs, which provide a human cellular model for examining the contributions of age and genetic factors to disease onset.
Since neurological disorders often affect distinct neuronal subpopulations, studies using generic protocols to induce unrestricted neuronal fates are likely only capturing a partial snapshot of factors that contribute to disease onset and progression. The direct conversion of fibroblasts of symptomatic HD patients generates MSNs that retain their cellular age status. To test the involvement of cellular age in the manifestation of disease-relevant phenotypes in HD-MSNs, we applied two distinct cellular reprogramming approaches that diverge in the maintenance of age signatures from donor cells. The induction of pluripotency has been well documented to erase age marks and reset the age of donor cells to an embryonic state18,19 while direct neuronal conversion has been shown to maintain age-related transcriptional, cellular and epigenetic signatures20,21. In this study, we demonstrate that age retention through direct neuronal conversion is crucial for modeling HD, exemplified by the detection of mHTT aggregates, a direct reflection of the age-associated decline in proteostasis that is absent in iPSC-derived neurons.
We found that mHTT-induced DNA damage contributed to the spontaneous degeneration of HD-MSNs, as treating the cells with an inhibitor of the DNA damage response protein ATM rescued the cell death phenotype, similarly to iPSC-derived neurons undergoing degeneration upon BDNF withdrawal39. We also provide evidence that controlling the specificity of MSN fate during neuronal conversion is critical for the manifestation of disease phenotypes, as altering the terminal neuronal fate of HD-fibroblasts to CNs drastically reduced the levels of DNA damage and cell death, despite the persistence of mHTT aggregates. Although cortical neurons are not completely spared in HD, they degenerate at a much slower rate during disease progression than MSNs, even though mHTT aggregates are more common in the cortex than in the striatum33. Accordingly, postmortem studies in HD patients have also shown significantly less DNA damage in the cortex than in the striatum46. The cellular properties that render MSNs selectively vulnerable to mHTT-induced toxicity are poorly understood, and subtype-specific neuronal conversion approaches may offer an experimental means to examine neuroprotective attributes in discrete neuronal subtypes. The ability to model the progression of HD in an age-dependent manner provides a patient-based platform for applications in human disease modeling and a means to gain mechanistic insight into the pathogenesis of HD.
Plasmids and lentiviral preparation
The construction of all plasmids used in this study has been previously described17,47, and they are publicly available at Addgene as pTight-9-124-BclxL (#60857), rtTA-N144 (#66810), pmCTIP2-N106 (#66808), phMYT1L-N174 (#66809), phDLX1-N174 (#66859), phDLX2-N174 (#66860), with the exception of hSP9-N174, which was cloned in-house and not previously published. Polyglutamine fusion protein constructs pEGFP-23Q and pEGFP-74Q were generated by David Rubinsztein’s lab and acquired from Addgene (#40261 and #40262), and transfected into human fibroblasts. Lentiviral production was carried out separately for each plasmid, but they were transduced together as a single cocktail as previously described47. Briefly, supernatant was collected 60–70 h after transfection of Lenti-X 293LE cells (Clontech) with each plasmid, in addition to psPAX2 and pMD2.G (Addgene), using polyethyleneimine (Polysciences). Collected lentiviruses were filtered through 0.45 µm PES membranes and concentrated at 70,000 g for 2 h at 4 °C. Viral pellets were resuspended in Dulbecco’s phosphate-buffered saline (DPBS, Gibco) and stored at –80 °C until transduction.
Cell lines and culture
Adult dermal fibroblasts from symptomatic HD patients (Coriell NINDS and NIGMS Repositories: ND33947, ND30013, GM02173, GM09197, GM04230, GM04194, GM04196, GM04198, GM02147, GM04687), presymptomatic HD patients (GM04717, GM04861, GM04855, GM04831, GM04853, GM04829) and healthy controls (Coriell NINDS, NIA and NIGMS Repositories: ND34769, AG04148, GM02171, GM05879, AG16409, AG11357, AG11483, GM05879, AG16409, AGO5265, AG09599, AG04062, AG04060) were acquired from the Coriell Institute for Medical Research. One additional healthy control adult dermal fibroblast line was acquired from the Washington University School of Medicine iPSC Core Facility (#F09-238). The International Cell Line Authentication Committee (ICLAC) lists none of these primary cells as commonly misidentified cell lines. In regards to deidentified skin fibroblasts samples and induced pluripotent stem cells (iPSCs) acquired from the Coriell Institute for Medical Research, we do not have access to the master list to reidentify subjects. This activity is not considered to meet federal definitions under the jurisdiction of an institutional review board, and is thus exempt from the definition of human subject. All fibroblasts were cultured in fibroblast medium (FM): Dulbecco’s Modified Eagle Medium (DMEM) with high glucose containing 15% FBS (Gibco), 0.01% β-mercaptoethanol (BME), 1% nonessential amino acids (NEAA), 1% sodium pyruvate, 1% GlutaMAX, 1% 1 M HEPES buffer solution and 1% penicillin/streptomycin solution (all from Invitrogen). We routinely check all our cell cultures and confirm them to be free of mycoplasma contamination. Our step-by-step MSN conversion protocol has been previously published47. Briefly, the lentiviral cocktail of rtTA, pTight-9-124-BclxL, CTIP2, MYT1L, DLX1 and DLX2 was added to fibroblasts for 16 h, then cells were washed and fed with FM containing 1 μg/mL doxycycline (DOX). Cells were fed at post-induction day (PID) 3 with FM + puromycin (3 μg/mL) + blasticidin (3 μg/mL) + DOX and replated at PID 5 onto polyornithine/fibronectin/laminin-coated glass coverslips in FM + DOX. Medium was switched on PID 6 to Reprogramming Neuronal Medium (RNM): Neuronal Medium (NM; ScienCell Research Laboratories) with 200 μM dibutyl cyclic AMP, 1 mM valproic acid, 10 ng/mL BDNF, 10 ng/mL NT-3 and 1 μM retinoic acid, supplemented with DOX. Half-volume medium changes with RNM were performed every 4 d with addition of DOX every 2 d thereafter until PID 30–35. Addition of puromycin and blasticidin was terminated after PID 14.
DNA extraction and CAG sizing
Fibroblasts were expanded in culture, collected by cell scraper, pelleted, and lysed for DNA extraction and ethanol precipitation following typical lab procedures with proteinase K (Roche). DNA samples were CAG sized by Laragen, Inc (Culver City, CA).
Cells were fixed using 4% paraformaldehyde (PFA) for 20 min and permeabilized using 0.2% Triton-X solution for 10 min following three phosphate-buffered saline (PBS) washes. Cells were blocked for 1 h at room temperature using 1% normal goat serum (NGS) and 5% bovine serum albumin (BSA) in PBS. Primary antibodies were added in the presence of blocking buffer overnight at 4 °C. Secondary antibodies were added following three PBS washes at 1:1,000 in blocking buffer at room temperature for 1 h. The following primary antibodies were used for the immunofluorescence studies: mouse anti-MAP2 (Sigma-Aldrich #M9942 clone HM2, 1:750), rabbit anti-β-III tubulin (BioLegend #MMS-435P, 1:2,000), chicken anti-NeuN (Aves, #NUN 1:500), rabbit anti-GABA (Sigma #A2052, 1:2,000), mouse anti-GABA (Sigma #A0310 clone GB-69, 1:500), rabbit anti-DARPP32 (Santa Cruz Biotechnology #sc-11365, 1:400), rabbit anti-S100A4 (FSP1) (Abcam #124805, 1:200), mouse anti-HTT (mEM48, Millipore #MAB5374, 1:50) (MW8, Developmental Studies Hybridoma Bank, 1:100), rabbit anti-ubiquitin (Abcam #ab7780, 1:50), mouse anti-vimentin (Sigma-Aldrich, #V6630 1:500), rabbit anti-fibronectin (Sigma-Aldrich, #F3648 1:500), mouse anti-phospho-histone H2A.X (Millipore #05-636-I, 1:200), rabbit anti-lap2-α (Abcam #ab5162, 1:500), rabbit anti-53BP1 (Abcam #ab21083, 1:200), mouse anti-8OHdG (Santa Cruz Biotechnology #sc-139586, 1:1,000), rabbit anti-LC3B (Sigma-Aldrich # L7543, 1:1,000). The secondary antibodies were goat anti-rabbit or mouse IgG conjugated with Alexa-488, Alexa-594 or Alexa-647 (Invitrogen). Images were captured using a Leica SP5X white light laser confocal system with Leica Application Suite (LAS) Advanced Fluorescence 188.8.131.5223. All staining quantification was performed by counting number of positive-stained cells over DAPI signal. Antibodies were validated by staining fibroblasts as negative controls, and they exhibited low background.
At PID 28, cells were lysed in SDS lysis buffer (1 M Tris-HCl pH 6.8, 2% SDS, 30% glycerol) supplemented with protease inhibitors (Roche, #04693132001). The concentrations of whole-cell lysates were measured using the Pierce BCA protein assay kit (Thermo Scientific, #23227). Equal amounts of whole cell lysates were resolved by SDS-PAGE and transferred to a nitrocellulose membrane (GE Healthcare Life Sciences, #10600006) using a transfer apparatus according to the manufacturer’s protocols (Bio-Rad). After incubation with 5% BSA in TBS containing 0.1% Tween-20 (TBST) for 30 min, the membrane was incubated with primary antibodies at 4 °C overnight: MW8 (Developmental Studies Hybridoma Bank, 1:500) and MW1 (Developmental Studies Hybridoma Bank, 1:500). Following incubation, membranes were incubated with a horseradish peroxidase–conjugated anti-mouse or anti-rabbit antibody for 1 h. Blots were developed with the ECL system (Thermo Scientific, #34080) according to the manufacturer’s protocols.
The cell-permeant mitochondrial indicator MitoTracker Red CMXRos (ThermoFisher Scientific #M7512) was added directly to live cells at final concentration of 50 nm in serum-free medium. After 20 min of incubation in 37 °C, cells were imaged with an epifluorescence microscope and then fixed and processed for immunostaining as described above. Analysis of colocalization of MitoTracker Red and LC3-II (Anti-LC3B antibody, Sigma-Aldrich # L7543) was performed using Metamorph bioimaging software after image acquisition using a Leica SP5X white light laser confocal system with Leica Application Suite (LAS). Mitochondrial membrane potential was assayed with the TMRE Mitochondrial Membrane Potential Assay Kit (Abcam #ab113852) following the manufacturer’s protocol. Briefly, TMRE was added to live cells at a final concentration of 20 nm in serum-free medium. After 15 min of incubation at 37 °C, coverslips were removed from medium and Vaseline was applied to edges of coverslips to create a rim for live mounting and microscopy and imaged using a Leica SP5X white light laser confocal system with Leica Application Suite (LAS). Lipid droplets were stained with BODIPY 493/503 (4,4-difluoro-1,3,5,7,8-pentamethyl-4-bora-3a,4a-diaza-s-indacene) (ThermoFisher Scientific #D3922) at a final concentration of 0.1 μm in serum-free medium. After 30 min of incubation in 37 °C, cells were imaged with an epifluorescence microscope and quantified with Leica Application Suite (LAS) quantification tools.
Whole-cell patch-clamp recordings were performed at PID 28–35 with miR-9/9*-124-CDM. At PID 14, cells undergoing reprogramming were transduced with pSYNAPSIN tRFP or GFP, and the next day they were trypsinized and plated together on top of rat primary neurons and glia isolated from perinatal pups, with the exception of recordings shown in Supplementary Fig. 3h, which were performed in monoculture in the absence of rat primary cells. Fluorescent reporter expression was visible within days and remained segregated for each population. Data were acquired using pCLAMP 10 software with MultiClamp 700B amplifier and Digidata 1550 digitizer (Molecular Devices). Electrode pipettes were pulled from borosilicate glass (World Precision Instruments) and typically ranged between 4 and 6 MΩ resistance. Solutions used to study intrinsic neuronal properties were the same as previously reported17. Postsynaptic potentials were detected spontaneously. Data were collected in Clampex and initially analyzed in Clampfit (Molecular Devices).
RNA extraction and gene expression profiling
Total RNA was extracted and isolated with TRIzol reagent (Thermo Fisher Scientific) according to manufacturer’s instructions. cDNA was generated from isolated RNA with Superscript III Reverse Transcriptase (Thermo Fisher Scientific) primed with random hexamers. qPCR was performed with the primer sets listed in Supplementary Table 3. For RNA-seq, reads were aligned to the human genome (assembly hg38) with STAR version 2.4.2a . Gene counts were derived from the number of uniquely aligned unambiguous reads by Subread:featureCount , version 1.4.6, with GENCODE gene annotation (V23) . All gene-level transcript counts were then imported into the R/Bioconductor package EdgeR  and TMM normalized to adjust for differences in library size. Genes not expressed in any sample were excluded from further analysis. The fits of the trended and tagwise dispersion estimates were then plotted to confirm proper fit of the observed mean-to-variance relationship, where the tagwise dispersions are equivalent to the biological coefficients of variation of each gene. Differentially expressed genes were then filtered for those having fold-changes (FC) > 1.5 together with false-discovery rate (FDR) adjusted P-values ≤ 0.05. Gene expression heat maps were generated using Z-scores for expression values of each gene among different samples (GENE-E Matrix Visualization and Analysis Platform, Broad Institute). MSN-specific genes were selected from previous studies that have profiled transcriptome profiles of isolated MSNs48. RNA-seq data are publicly available at GEO (accession code GSE84013).
SYTOX green nucleic acid staining (Thermo Fisher Scientific) was performed following manufacturer’s suggestions, adapted as follows: a final concentration of 0.1 μM SYTOX green was added directly to the medium of live cells. In addition, Hoechst 33342 solution (Thermo Fisher Scientific) was added as a counterstain to label all nuclei at a final concentration of 1 μg/ml in culture medium. Samples were incubated for at least 10 min in 37 °C. Images were captured using a Leica DMI 400B inverted microscope with Leica Application Suite (LAS) Advanced Fluorescence. Three images were taken from random areas of each coverslip for at least three biological replicates per experiment. Quantification performed by counting number of SYTOX-positive cells over total Hoechst signal.
DNA damage was assessed by using the CometAssay reagent kit for single-cell gel electrophoresis assay (Trevigen, MD USA), following the recommended protocol for neutral conditions and adapting the gel electrophoresis methods for use in the Sub-Cell GT electrophoresis system (Bio-Rad, CA USA). Briefly, cells were collected from coverslips by treatment with 0.25% trypsin, pelleted, resuspended at 100,000 cells/ml in DPBS (Ca2+ and Mg2+ free; Thermo Fisher Scientific) and verified to be greater than 95% viable by trypan blue exclusion using an automated cell counter before continuing analysis. Approximately 5,000 cells were embedded in low-melting agarose, plated on slides and lysed overnight. The next day, electrophoresis was run at 30 V for 30 min in 1× TBE (National Diagnostics). Samples were fixed in 70% ethanol for 5 min, and slides were immersed in TE buffer pH 8.0 (Ambion) with SYBR green nucleic acid stain (10,000×, Thermo Fisher Scientific). Fluorescence images were captured using a Leica DMI 400B inverted microscope for scoring.
Generation of iPSCs and derivation of HEFs
iPSC lines used in this study were either directly acquired from the Coriell Institute for Medical Research NINDS Biorepository (#ND42235) or derived from adult dermal fibroblast acquired from the Coriell NINDS Biorepository (#ND33947) with the assistance of the Washington University School of Medicine Genome Engineering and iPSC Center (GEiC). For the generation of ND33947 iPSCs, fibroblasts were transduced with integration-free Sendai reprogramming vectors for Oct3/4, Sox2, Klf4 and c-Myc and characterized by the expression of the pluripotency markers Oct4, SSEA4, SOX2 and TRA-1-60 (PSC 4-Marker Immunocytochemistry Kit, Molecular Probes). Cytogenetic analysis was performed on twenty G-banded metaphase cells from iPSC line at passage 5, and all 20 cells demonstrated an apparently normal karyotype (Cell Line Genetics, Madison WI). In addition, an embryoid body formation assay confirmed the potential for acquisition of all three germ layers. iPSCs were expanded on ES-grade Matrigel (Corning)-coated plates cultured in mTeSR medium (Stemcell Technologies) or DMEM/F-12 with 20% KnockOut Serum Replacement, 1% GlutaMAX, 0.1 mM NEAA, 10 ng/mL fibroblast growth factor-basic (bFGF) and 55 μM BME. To differentiate iPSCs into human embryonic fibroblasts (HEFs), culture medium was replaced with DMEM plus 20% FBS without bFGF for at least three passages. HEFs were transduced and reprogrammed to MSNs following our established protocol previously reported in detail47.
The ATM kinase inhibitor KU-60019 was obtained from Abcam (ab144817), solubilized in DMSO and directly added to the cell culture medium for a final concentration of 0.5 μM at 30 d after miR-9/9*-124 induction, and then cell death was assessed by SYTOX at PID 35. Controls were treated with the same volume of DMSO but no drug. At day 35, cells treated with DMSO or KU-60019 also were treated with 1 mM H2O2 for 3 h. SYTOX green and Hoechst stain were added as already described and imaged for scoring.
20 S proteasome activity assay
Adherent cells were dissociated with 0.25% trypsin, pelleted by centrifugation and washed in cold PBS twice. Cell pellets were then resuspended in chilled cell lysis buffer (50 mM HEPES (pH 7.5), 5 mM EDTA, 150 mM NaCl, 1% Triton X-100 and 2 mM ATP) and incubated on ice for 30 min, vortexing every 10 min. Cell lysates were then centrifuged at 15,000g for 15 min at 4 °C. Lysate was then transferred to a microcentrifuge tube, and 10 μl of each sample was used to determine protein concentration with a BCA protein assay kit (Thermo Scientific, Prod. #23227) following the manufacturer’s recommendations. Proteasome activity was assayed using 10 μg of each lysate with a 20 S proteasome activity assay kit (Millipore, APT280). Fluorescence intensity was measured every 5 min for 1 h with a microplate reader. Data were analyzed following previously reported methods38.
Cells cultured in gridded glass-bottom μ-dishes (Ibidi, Madison, WI) were fixed with EM grade 4% PFA + 0.05% glutaraldehyde (GA) (Electron Microscopy Sciences) in PBS with 2 mM CaCl2 at 37 °C for 5 min (min) then transferred to ice for 1 h. Samples were then incubated for 5 min in 50 mM glycine in PBS and permeabilized with 0.05% saponin with 1% BSA in PBS for 30 min. Cells were blocked with 1% BSA in PBS for 15 min and incubated with primary antibodies (mouse anti-HTT (MW8), 1:100, and rabbit anti-β-III tubulin BioLegend, 1:2,000) at room temperature for 2 h with gentle agitation. After washing in PBS-BSA three times for 10 min each, cells were incubated for an additional 2 h with Alexa Fluor 594 fluoro-nanogold secondary antibody (Nanoprobes, Yaphank, NY #7301) at a 1:250 dilution in PBS and 1% BSA at room temperature with gentle agitation while wrapped in foil. After washing in PBS three times for 10 min each, cells were fixed with 1% GA for 5 min and labeled with DAPI (1:10,000) for 5 min. After fluorescence imaging, the samples were rinsed twice in ultrapure water for 1 min each and then rinsed in 0.02 M citrate buffer (pH 4.8) three times for 5 min each. The fluoro-nanogold label was silver enhanced using HQ Silver (Nanoprobes, Yaphank, NY) for 9–11 min and the samples immediately rinsed with ultrapure water twice for 5 min each. The culture dishes were then rinsed in PBS buffer three times for 10 min each and subjected to a secondary fixation step for 1 h in 1% osmium tetroxide, 0.3% potassium ferrocyanide in PBS on ice. The samples were then washed in ultrapure water three times for 10 min each and then stained en bloc for 1 h with 2% aqueous uranyl acetate. After staining was complete, samples were briefly washed in ultrapure water, dehydrated in a graded ethanol series (50%, 70%, 90%, 100% twice) for 10 min in each step, and infiltrated with microwave assistance (Pelco BioWave Pro, Redding, CA) into LX112 resin. Samples were cured in an oven at 60 °C for 48 h. Once the resin was cured, the gridded glass coverslips were etched away with concentrated hydrofluoric acid and the exposed cells were excised with a jeweler’s saw and mounted onto blank resin blocks with epoxy, oriented in the coverslip growing plane. Sections 70 nm thick were then taken and imaged on a TEM (JEOL JEM 1400 Plus, Tokyo, Japan) at 80 KeV.
For all quantified data, multiple cells were counted from at least three biological replicates from multiple independent experiments or multiple lines. Statistical analyses were performed in GraphPad Prism using a two-tailed Student’s t-test or a one-way ANOVA followed by a post hoc Tukey’s test with P < 0.05 considered significant. Multiple comparisons were corrected with the Bonferroni or Holm-Sidak method as described in the figure legends. Studies were performed blindly and automated whenever possible with the aid of ImageJ cell counting tools, and multiple investigators confirmed quantification results. Data distribution was assumed to be normal, but this was not formally tested. Data in graphs are expressed as mean and error bars represent s.e.m. unless noted otherwise. Outliers were detected and excluded with Grubbs’ test for α levels of 0.05. In total for this study, only two data points were excluded, from the Supplementary Fig. 10c,d control DMSO group (9 total data points), following pre-established criteria. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications11,12,13. Data collection was not randomized, but was always done in parallel with controls. Allocation of primary patient cells acquired from Coriell Biorepository into the HD group was done randomly. Samples were allocated into the Control group by age- and sex-matching healthy controls with HD samples acquired and available through Coriell Biorepository.
Life Sciences Reporting Summary
Further information on experimental design is available in the Life Sciences Reporting Summary.
Gene expression data generated for this study have been made public at NCBI’s Gene Expression Omnibus (GEO), accession GSE84013. All other data supporting the findings of this study are available from the corresponding author upon reasonable request.
A step-by-step protocol for this study is available at Nature Protocols47.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors thank A. Bowman and L. Solnica-Krezel for suggestions, B. Steger and J. Peyer for data quantification, the Washington University Center for Cellular Imaging (WUCCI) for their help in generating electron microscopy data, the Genome Technology Access Center (GTAC) for generating transcriptome datasets, and the Core Usage Funding Program from the Institute of Clinical and Translational Services (ICTS) and the Genome Engineering and iPSC Center (GEiC) at Washington University School of Medicine for their assistance in generating and characterizing iPSC lines. M.B.V. is supported by a National Science Foundation Graduate Research Fellowship (DGE-1143954) and a NIH/NIA dissertation award (1R36AG053444-01). A.S.Y. is supported by the Andrew B. and Virginia C. Craig Faculty Fellowship endowment, an NIH Director’s Innovator Award (DP2NS083372-01), a Seed Grant from Washington University Center of Regenerative Medicine, the Ellison Medical Foundation New Scholar in Aging Award, Cure Alzheimer’s Fund (CAF) and a Presidential Early Career Award for Scientists and Engineers (PECASE) (4DP2NS083372-02).