Generation and transplantation of reprogrammed human neurons in the brain using 3D microtopographic scaffolds

Cell replacement therapy with human pluripotent stem cell-derived neurons has the potential to ameliorate neurodegenerative dysfunction and central nervous system injuries, but reprogrammed neurons are dissociated and spatially disorganized during transplantation, rendering poor cell survival, functionality and engraftment in vivo. Here, we present the design of three-dimensional (3D) microtopographic scaffolds, using tunable electrospun microfibrous polymeric substrates that promote in situ stem cell neuronal reprogramming, neural network establishment and support neuronal engraftment into the brain. Scaffold-supported, reprogrammed neuronal networks were successfully grafted into organotypic hippocampal brain slices, showing an ∼3.5-fold improvement in neurite outgrowth and increased action potential firing relative to injected isolated cells. Transplantation of scaffold-supported neuronal networks into mouse brain striatum improved survival ∼38-fold at the injection site relative to injected isolated cells, and allowed delivery of multiple neuronal subtypes. Thus, 3D microscale biomaterials represent a promising platform for the transplantation of therapeutic human neurons with broad neuro-regenerative relevance.

N eurodegenerative diseases and traumatic brain injuries result in a loss of functional neurons in the central nervous system (CNS) and are responsible for substantial deterioration in quality of life. Although cell transplantation therapies have shown some promise towards functional recovery in animal models, the efficacy of these therapies has been limited by poor cell survival rates [1][2][3] . Human induced pluripotent stem (iPS) cells have recently emerged as a promising renewable source of expandable patient-specific cells that can be used to generate human neurons [4][5][6] . These iPS cell-derived neurons are a potentially invaluable tool as a cell source for the in vivo treatment of neurodegenerative diseases and traumatic CNS injury 7,8 .
While many neuronal differentiation protocols have been established, a robust protocol was recently advanced for the accelerated production of human neuronal cells from iPS cells, called induced neuronal (iN) cells, by the direct conversion of iPS cells using ectopic expression of sets (Brn2, Ascl1, Myt1l) or single transcription factors (Ascl1, NeuroD1 or Ngn2) (refs 9-12). Direct conversion yields improved conversion efficiency and accelerates maturation relative to standard differentiation protocols. However, effective approaches to generate these cells in three-dimensional (3D) configurations and to deliver these cells therapeutically in vivo have yet to be established. Cell replacement therapies of matured neurons in the brain have conventionally been limited to injection of dissociated cells 2,13 . An alternative approach using biomaterial scaffolds can provide structural support to cells during transplantation, which could improve cell engraftment and survival. In addition, many types of cells behave differently when cultured in two-dimensional (2D) versus 3D substrates 14,15 , leading to the development of 3D biomaterials that better mimic aspects of the in vivo cellular microenvironment 16 . In particular, microscale fibrous substrates enhance several neural cell behaviours in vitro, including neurite outgrowth 17 , neuronal maturation 18 and neuronal differentiation. These results motivate the interfacing of microscale fibrous scaffolds with human induced neuronal cells for in vitro and in vivo applications. The concept of using 3D biomaterials to support transplantation has been used with synthetic hydrogels 19 , microparticles 20 and natural protein matrices 21,22 to transplant neural progenitor cells into the brain, however, these sorts of scaffolds are not conducive to extended culture or maturation of cells in vitro, and yield limited control of the cell microenvironment once transplanted.
In this study, we advance the concept of designing electrospun synthetic polymer fibres to support the neuronal reprogramming of iPS cells within a 3D environment via the ectopic expression of NeuroD1. We demonstrated that fine-tuning the fibrous architecture enhances neuronal maturation and functionality in vitro, and also reduces the population of residual unconverted cells. Finally, we investigated the role of micron-scale fibrous scaffolds as injectable transplantation vehicles for adherent networks of iNs, and their ability to enhance the survival and engraftment of these neurons in murine brain tissue ex vivo and in vivo.

Induction of neurons via single defined transcription factors.
We initially identified that of the four transcription factors (TFs), that is, Brn2, Ascl1, Myt1L and NeuroD1, used by Pang et al. 12 to reprogram human fibroblasts to neurons, only Ascl1 and NeuroD1 could individually induce neuronal conversion of iPS cells, as also reported elsewhere 10 . Early neurons induced by NeuroD1 expression exhibited complex neuronal morphologies and express mature neuronal markers, so all subsequent studies were done using NeuroD1 as the single TF for generating iN cells. We transduced human iPS cells with lentiviruses encoding tetracycline-inducible, tetOn-NeuroD1, rtTA, and, in selected cases, tetOn-EGFP, and maintained them as undifferentiated iPS cells in mTeSR medium for multiple passages in the absence of doxycycline (dox; Supplementary Fig. 1a). We refer to these cells as 'RN-iPS' cells, since these cells were transduced with rtTA and NeuroD1. RN-iPS cells maintain expression of pluripotency markers SSEA-4 and Oct-4 over multiple passages in the absence of dox, indicating that they remain undifferentiated after transduction ( Supplementary Fig. 1a).
To assess neuronal induction from RN-iPS cells, dense cultures of RN-iPS cells were seeded and dox was added 24 h after plating ( Supplementary Fig. 1b). Early stage iN cells expressed neuronal marker bIII-tubulin and residual undifferentiated iPS cells expressed Oct-4 ( Supplementary Fig. 1c). Dox addition rapidly induces the loss of undifferentiated morphology and the acquisition of bipolar, early neuronal morphologies in a subset of cells within 48 h ( Supplementary Fig. 1c-e). Dox also rapidly induces EGFP expression in control cells transduced with tetOn-EGFP. By contrast, cells transduced with only tetOn-EGFP and rtTA (lacking tetOn-NeuroD1) show strong GFP fluorescence but maintain undifferentiated hPSC morphology upon dox addition ( Supplementary Fig. 1e).
Despite the rapid, dox-induced neuronal conversion of a subset of RN-iPS cells, the remaining undifferentiated cells proliferated and rapidly overtook iN cultures. To address this, cells were replated 4 days post dox treatment, which established more uniform human neuronal cultures and eliminated many undifferentiated cells (Fig. 1a). The replating process may disrupt the cell-cell contacts necessary for undifferentiated iPS cell survival. After replating, enriched iN cells could be maintained for 4 weeks or longer before characterization.
As expected, human iN cells robustly express bIII-tubulin and microtubule-associated protein 2 (MAP2) by day 12-14 after dox addition (Fig. 1b). To identify the neuronal subtypes generated by this procedure, iN cells were replated 4 days after dox addition onto glial cell monolayers as reported elsewhere 23 . Immunocytochemistry on 28 day iNs post replating revealed that most cells expressed glutamate vesicular transporter VGLUT1, indicating that these cells are predominantly excitatory glutamatergic neurons (Fig. 1c). Occasional cells expressing markers of other neuronal subtypes were also observed, including inhibitory GABAergic neurons (expressing vesicular GABA transporter) and dopaminergic neurons (expressing tyrosine hydroxylase; Fig. 1c). Most cells also robustly expressed the pre-synaptic protein synaptophysin (Fig. 1d). Finally, patchclamp electrophysiology demonstrated that iN cells are electrically active and express functional voltage-dependent Na þ channels as well as voltage-dependent K þ channels as revealed by whole-cell current recordings (Fig. 1e). Strikingly, these human iN cells were capable of firing repetitive action potentials (Fig. 1f). Taken together, these data indicate the robust generation of functional neuronal cells via NeuroD1 overexpression.
Generation of functional neurons on 3D electrospun fibres. Next, we investigated whether NeuroD1 expression would similarly induce neuronal conversion and maturation within model 3D electrospun substrates. We constructed fibrous substrates by electrospinning poly(desaminotyrosyl tyrosine ethyl ester carbonate) (pDTEc), into two architectures, which will be referred to as 'thin' and 'thick' fibre substrates with average fibre diameters of 1.25 ± 0.05 mm and 3.23 ± 0.06 mm, respectively 24 ( Fig. 2a-d). pDTEc is the lead candidate polymer from a combinatorial library of tyrosine-derived polycarbonates 25 , as it effectively supports pluripotent stem cell culture when fabricated into microscale fibrous substrates 24 , and is biocompatible 26 . Similarly, our results with 3D polymeric substrates compared with 2D polymeric substrates suggest that the fibrous architecture governs the longer term cellular behaviours observed in contrast to the polymer composition that plays a role on early interfacial phenomena. The thick fibre scaffolds are volumetrically permeable to cellular infiltration, whereas the thin fibre scaffolds are relatively impermeable, due to decreased void space between fibres. Without additional material modifications, it is difficult to produce equal fibre sizes with variable porosity, as both properties are simultaneously modulated when altering electrospinning parameters. We hypothesize that cell permeable, thick fibre substrates will support improved iN maturation and functionality by promoting enhanced 3D organization and cellcell contacts relative to less permeable, thin fibre substrates and 2D controls.
Human iN cells were generated within 3D constructs by treating RN-iPS cells, cultured on 2D tissue-culture plates, with dox for 4 days, followed by replating onto 3D electrospun substrates or 2D controls, according to the time course schematic shown in Fig. 2e. Human iN cells on 3D electrospun fibres showed complex morphology with extensive neurite outgrowth and expressed bIII-tubulin, MAP2 and synaptophysin after 12 days of reprogramming, similarly to 2D cultures (Fig. 2f,g). In addition, electrophysiological recordings revealed iN cells in electrospun substrates fired action potentials, demonstrating the derivation of functional iN cells (Fig. 2h,i).
Effect of 3D scaffold architecture on human iN conversion. Next, we examined whether the fibre architecture could be tuned to enhance human iN maturation, as we and others have shown that geometric cues can influence both human iPS cell and neural cell behaviours 24,27 . RN-iPS cells were treated with dox for 4 days, then replated onto 3D fibrous substrates or 2D controls, including 2D polymer-coated controls, for an additional 8 days of culture. Immunocytochemistry for proliferation marker Ki67 and Oct-4 revealed that significantly more proliferative and pluripotent cells were retained when iN cells were replated onto 2D substrates compared with 3D fibrous substrates (Po0.05, one-way ANOVA; Fig. 3a,c, Supplementary Figs 2,3). This suggests that the fibrous architectures may be able to selectively reduce the presence of residual proliferative and pluripotent iPS cells.
Human iN cells in all conditions expressed extensive bIII-tubulin-positive processes, along with robust MAP2 expression (Fig. 3a, Supplementary Fig. 2). Significantly greater numbers of human iN cells expressed MAP2 in thick fibre substrates relative to 2D controls (Po0.0001, one-way ANOVA) and thin fibre substrates (Po0.05, one-way ANOVA), indicating accelerated maturation (Fig. 3b). qRT-PCR also revealed increasing trends in expression of several neuronal genes in 3D substrates relative to 2D controls, including bIII-tubulin, MAP2, synapsin 1 and VGLUT1, though not statistically significant ( Supplementary Fig. 4). Most importantly, calcium imaging to identify the fraction of cells that respond to a field electrical stimulation indicated that thick fibre substrates yielded a high degree of activity, namely, 470% electrically active cells by day 12 of culture ( Fig. 3d-e), which was significantly greater than those on the thin fibre substrates (Po0.05, one-way ANOVA). Inhibition of E-cadherin-dependent cell-cell contacts markedly reduced neuronal outgrowth in 2D, and decreased activity measured by calcium imaging for iNs on 2D and 3D thick fibre substrates but not thin fibre substrates ( Supplementary Fig. 5).   This indicates that 3D microfibrous architectures establish neuronal networks with enhanced cell-cell contacts and influence both functional and phenotypic maturity of iN cellular networks.
Microscale scaffolds support outgrowth and survival in brain. Next, transplantable constructs were designed to deliver human iN cells into the brain for regenerative therapies. The large mats of electrospun fibres (0.3-2 cm diameter discs) conventionally fabricated for in vitro studies cannot be easily transplanted into the CNS. To allow for injection in vivo, 100 mm square 'microscale scaffolds' that could be injected through a 21-gauge needle were created by downscaling thick fibre electrospun substrates with a Vibratome. Human iN cells were seeded in suspension onto microscale scaffolds analogous to our previous studies, which resulted in efficient population of microscale scaffolds with iN cells. Cells in microscale scaffolds matured into bIII-tubulin and MAP2-expressing neuronal cells, similarly to cells on macroscale fibrous substrates ( Supplementary Fig. 6). The average number of live human iN cells in each scaffold was 83±13 (n ¼ 19; Supplementary Fig. 7).
The ability of microscale electrospun scaffolds to promote human iN cell survival and engraftment was first assessed using an ex vivo model consisting of organotypic hippocampal slice cultures from NOD-SCID IL2Rgc null mice (Fig. 4a). Human iN cells on 4-5 microscale scaffolds were injected into hippocampal slices, alongside equivalent numbers of dissociated cells on paired slices, and engraftment and functionality was assessed. Immunocytochemistry revealed that 3 days after transplantation, injected scaffold-supported iN cells had average neurite lengths of 831 ± 169 mm, which was significantly greater (Po0.0001, one-way ANOVA) than those of injected dissociated iNs, which had neurite lengths of 241 ± 42 mm (Fig. 4b-e). Electrophysiological recordings from human iN cells after 3 weeks indicated that both dissociated and scaffold-supported iN cells fire action potentials in response to current injection (Fig. 4f,g), with scaffold-seeded iN cells displaying enhanced excitability in response to injected currents (Fig. 4h). In addition, both modes of transplanted cells had mature Na þ channel expression ( Fig. 4i). This suggests that microscale scaffolds enhance engraftment and functionality of transplanted iN cells.
Next, human iN cell survival was assessed after transplantation of scaffold-supported or dissociated cells into the mouse striatum in vivo. Three weeks after transplantation (Fig. 5a), immunocytochemistry of a 1.8 mm Â 1.8 mm field including the injection site ( Fig. 5b-e) revealed an average survival rate of 5.74±3.16% based on injected scaffold-seeded iN cells, a 38-fold improvement (Po0.05, one-way ANOVA) compared with an average survival rate of 0.15 ± 0.15% of 100,000 injected dissociated iN cells (Fig. 5g). The survival rate of dissociated iN cell controls was comparable to that reported by Zhang et al. 10 When magnitude of injected cells were matched at B1,000 cells, quantification of surviving cells yielded an average of 7.58±4.60% of injected scaffold-seeded iN cells compared with 0 ± 0% out of 1,000 injected dissociated iN cells. The scaffold-seeded iNs maintained neurite length (35 ± 8 mm) comparable to those of viable, dissociated iNs (39±15 mm) in the experiment when number of injected cells was unmatched. No overt difference in inflammatory response was detected between injection modes, and some ingrowth of host tissue into scaffolds was observed ( Supplementary Fig. 8). Surviving transplanted iNs expressed neuronal cell-adhesion molecule CD56, bIII-tubulin and synaptophysin (Fig. 5f, Supplementary Fig. 9). Post-synaptic density protein 95 (PSD95, depicted in blue, with downwardpointing arrows) was detected adjacent or co-localized to transplanted GFP-labelled and synaptophysin-expressing iN neurite terminals, suggestive of synaptic integration with host tissue ( Fig. 5f; Supplementary Movie 1). Notably, when microscale scaffolds were used to transplant multiple subtypes of neurons, we observed retention of neurons of distinct specificity in close apposition in vivo (Fig. 5i,j). Similar differences in survival were also seen when glutamatergic and dopamine iNs (Supplementary Fig. 10  some human iNs were observed to migrate off scaffolds, and minimal scaffold degradation was observed in the 1-3 weeks post-transplantation.

Discussion
Cell therapies for treatment of CNS brain injury and disease are challenged by poor cell survival, engraftment and retention after transplantation both into the brain and spinal cord, with cell survival ofo1% routinely reported 28 . The objective of this study was to advance a new integrated biomaterials-based paradigm to both rapidly reprogram pluripotent stem cells to neurons and efficiently deliver enriched, organized neuronal networks to the brain. To achieve this, we engineered human iPS cells with the neuronal TF, NeuroD1, and demonstrated efficient and rapid neuronal conversion in 3D fibrous substrates fabricated from synthetic polymers that were geometrically tuned to accelerate neuronal maturation and network establishment. Utilizing this system, we report that 3D microfibrous substrates can guide many important characteristics of human iN cells derived from iPS cells that are relevant for in vivo applications, including (i) in vitro maturation, electrical activity and purification; (ii) neurite outgrowth, survival and electrical activity after transplantation into ex vivo brain tissue and (iii) survival and engraftment after transplantation in vivo into the striatum. These results highlight the promise of biomaterials-based delivery as a platform for reprogramming and regenerative cell sourcing in the CNS and as a potential model that can be further adapted and evaluated for therapeutic efficacy.
Our primary hypothesis was that microfibrous substrates with thick fibres and interfibre spacing would provide a 3D microniche for rapid, high-purity conversion and accelerated maturation of iPS cell-derived iNs, and that such 3D substrate-supported neuronal networks would exhibit improved levels of retention and engraftment following transplantation into the brain. In vitro, thick fibre substrates enhanced human iNs maturation and excitability, while selectively excluding residual undifferentiated iPS cells. Both of these phenomena arise from controlled cell confinement and 3D juxtacrine signalling organization. While the enhanced MAP2 expression in iN cells in thick fibre substrates is consistent with previously reported results 18,[29][30][31] , the concurrent increase in excitability indicates active both phenotypic and functional maturation. We propose that the thick fibre substrates drive iN confinement 24 , leading to accelerated maturation due to increased engagement of neural cell-adhesion molecules such as L1 or N-cadherin, which have known roles in neural development [32][33][34] and neuritogenesis 29,31 . In contrast, the thin fibre substrates fail to support neuronal infiltration and aggregation, resulting in the diminished excitability relative to more 3D, thick fibre substrates.
For effective scaffolds that can form reprogrammed neuronal networks, a key feature is that the proliferation of undifferentiated iPS cells should be suppressed. The geometry we have identified for electrospun scaffolds limits the proportion of undifferentiated iPS cells. Undifferentiated human ES and iPS cells require media supplementation with the rho-associated kinase (ROCK) inhibitor Y-27632 to survive dissociation to single cells 35 . In the absence of Y-27632, the degree to which dissociated iPS cells can re-establish E-cadherin-mediated cell-cell contacts that have been demonstrated to be required for survival and proliferation 36 is severely diminished when seeded onto electrospun fibrous substrates. Isolated pluripotent stem cells have been observed to form viable colonies largely based on motility-induced aggregation rather than single cell clonal expansion when intercellular distances were less than 6.4 mm (ref. 37). The large surface area, added substrate dimension and constrained migration paths of the electrospun fibres effectively increase intercellular distances, which would limit motility-induced aggregation and could explain the observed increase in differentiation of iPS cells and loss of residual proliferative cells on microfibrous substrates. This reduction in residual undifferentiated iPS cells could be used to supplement other methods for eliminating the risk of teratoma formation in vivo due to the presence of undifferentiated cells, however, additional studies are required to assess the tumorigenic potential of these cells once delivered in vivo.
After establishing a substrate geometry that accelerated iN maturation, the ability of these substrates to promote engraftment of human iN cells into the CNS was first probed in an ex vivo organotypic brain slice model. Delivery of iN cells on microscale Transplantation of iN cells in scaffolds into the mouse striatum showed that the percentage of viable cells after 3 weeks was an order-of-magnitude greater than that relative to injection of isolated single cells. This increased survival is likely indicative of decreased anoikis or apoptosis that otherwise arises from enzymatic disruption of cell-cell and cell-matrix interactions 38,39 . These results are promising particularly in light of recent studies investigating the effects of human iPS cell-derived neurons on treating CNS disorders that have shown efficacy in vivo even with low levels of cell survival 1,10 , suggesting that increased cell survival could lead to amplified levels of efficacy of these cell therapeutics. In addition, host neuron post-synaptic markers were detected adjacent to pre-synaptic marker-expressing transplanted neurons, suggesting that single-factor reprogrammed neurons are capable of functionally integrating into host tissue. Interestingly, some authors have found that extensive lesions can actually enhance neuronal graft survival and integration, as denervated targets may provide a positive stimulatory effect to new neurons 40,41 , suggesting that even greater cell survival may be possible when applying this scaffold system in a neurodegenerative disease or injury context. Finally, the ability for 3D microscale scaffolds to support co-culture and concurrent transplantation of neuronal networks consisting of neurons of multiple subtypes was evaluated. Transplantation of intact networks of mixed subtype neuron populations could be an invaluable tool for treating complex brain disorders or injuries that affect multiple subtypes, or for priming transplanted neurons with pre-established synaptic inputs. For example, attempts for transplanting dopamine neurons derived from fetal brain tissue 42,43 or derived from human iPS cells 1,44 have been made and yield amelioration of Parkinson's Disease; with the paradigm we proposed one can envision a model to graft mini-neurocircuitry composed of excitatory-dopaminergic neurons into the host brain. We hypothesize that proper excitation from the mini-neurocircuitry could produce more profound amelioration of locomotive deficits in Parkinson's Disease.
Induced human neurons offer a rich cell source for a variety of in vitro and in vivo applications in modelling and treating CNS diseases and injuries. Our prototype epitomizes a biomaterial device for subtype specific neuronal reprogramming and (a) Hundred-micrometre edge scaffolds were cut from electrospun fibres before seeding with iNs, followed by injection onto ex vivo cultured mouse pup brain slices. (b-e) Neurite length was found to be significantly enhanced in transplanted scaffold-supported iNs when comparing dissociated GFP-labelled iNs (b) with iNs seeded on scaffolds (c) injected onto mouse ex vivo brain slices (n ¼ 8 brain slices for each transplantation mode). (f-h) Both scaffold-supported iNs (n ¼ 6) and dissociated iNs (n ¼ 5) fire action potentials in response to current injection 14 days after transplantation onto mouse ex vivo brain slices, however, scaffold-supported iNs displayed enhanced excitability relative to dissociated iNs. Both modes of transplanted iNs were observed to have mature sodium channel expression (i). ***Po0.0001 by one-way ANOVA. Scale bar, 20 mm. All error bars presented as mean±1 s.d.  . The following day, medium was replaced with mTeSR-1 and cells were returned to normal culture for up to eight additional passages. To induce neuronal conversion, infected iPSCs (termed iPSC-RNs) were passaged with Accutase and Y-27632, and the following day switched to an N2-based conversion medium (N2M) consisting of DMEM/F12, 2 mM L-glutamine, 1 Â N2 supplement, 1 Â non-essential amino acids, 1% penicillin/streptomycin and 2 mg ml À 1 doxycycline, which induces expression of the NeuroD1 and EGFP constructs. Medium was replaced daily, and cells were replated on day 3-5 onto 2D or 3D substrates. One day after replating, medium was switched to NBM, consisting of neurobasal medium, 2 mM L-glutamine, 1 Â B27 supplement without Vitamin A (Life Technologies), 0.2 mM ascorbic acid, 1 mM cAMP (Sigma-Aldrich), 2 mg ml À 1 doxycycline and 10 ng ml À 1 each of brain-derived neurotrophic factor, Glial cell-line derived neurotrophic factor and neurotrophin-3 (Peprotech). Medium was replaced every 2-3 days for the duration of the experiments. To generate dopamine iNs for co-culture studies, 2.5 Â 10 5 cells per six-well plate were plated, and the following day medium was replaced with mTeSR-1 with 5 mM Y-27632, 2 mM polybrene, and concentrated virus was added (rtTA-FUW, Ascl1, Nurr1, Lmx1a and, in some experiments, Tet-O-FUW RFP). After three days, medium was replaced with Neurobasal media (Life Technologies) and concentrated virus (Pitx3, EN1, Foxa2). Four days after initial infection, the dopamine iNs would be co-seeded onto microscale scaffolds.
Electrospun fibrous substrate fabrication. Fibrous substrates and 2D polymer film controls were fabricated from tyrosine-derived polycarbonate pDTEc via electrospinning and spin-coating, respectively 24 . These polymers come from a combinatorial library of polymers with tunable hydrophobicity and degradation properties. pDTEc was dissolved in 1,1,1,3,3,3-Hexafluoro-2-propanol (Sigma-Aldrich) at 9%, or 18% weight by volume for thin and thick fibre configurations respectively. For the 9% weight by volume solution, 5% N,Ndimethylformamide was added to prevent beading of fibres. Polymer solutions were electrospun from a þ 18 kV spinneret to a À 6 kV flat plate collector at 3 ml h À 1 over a distance of 30 cm, resulting in fibre mats 250-500-mm thick. Surface morphology was observed on an AMRAY 1830 I scanning electron microscope and fibre diameter was quantified by measuring 100 individual fibres from each fibrous substrate using NIH-ImageJ software (http://rsb.info.nih.gov/ij/). Quantitative data are presented as mean±95% confidence interval. Control 2D polymer films were prepared using a 1% polymer solution dissolved in tetrahydrofuran, which was then spin-coated onto glass coverslips at 4,000 r.p.m. for 30 s (ref. 47). For in vitro studies, electrospun fibres were sterilized by ultraviolet treatment, oxygen plasma treated for hydration, and adsorbed with 10 mg ml À 1 poly-D-lysine in HEPES buffer (pH 8.4) and 4 mg ml À 1 laminin in PBS before seeding dissociated iNs 4 days after initiating neuronal conversion with N2M þ dox.
Microscale scaffold preparation and neuronal seeding. Microscale fibrous scaffolds were prepared by cutting oxygen plasma treated and hydrated electrospun fibre mats into 100 mm squares with a McIlwain tissue chopper (Vibratome), and sterilized and coated with poly-D-lysine and laminin as described above. On day 4 of neuronal conversion, iN cells were dissociated with Accutase and resuspended in N2M with dox at 5 million cells per ml. The scaffolds and iN cell suspension were mixed in a 48-well non-TCPS plate for 1 h with occasional gentle agitation, followed by three washes to remove unattached cells. Scaffolds containing iN cells were cultured for 1 day in N2M, then cultured in NBM until used for ex vivo or in vivo experiments. For experiments examining co-cultures of dopamine iNs and NeuroD1 iNs, dissociated dopamine and NeuroD1 iNs were dissociated, resuspended and combined 1:1 ratios to a concentration of 5 million total cells per ml.
Ex vivo slice preparation and scaffold transplantation. Organotypic mouse brain slice cultures were prepared by isolating the cortico-striatal area of 6-day-old C57BL/ 6 mouse brains and cutting into 300 mm slices using a vibratome (Leica Microsystems) in ice-cold sucrose solution (204 mM sucrose, 26.2 mM NaHCO 3 , 11 mM glucose, 2.5 mM KCl, 2 mM MgSO 4 , 1 mM NaH 2 PO 4 , 0.5 mM CaCl 2 saturated with 95% O 2 and 5% CO 2 ) (ref. 48). Slices were transferred onto Millicell organotypic cell culture inserts (Millipore) in a six-well tissue-culture plate and cultured in Neurobasal A medium (Life Technologies) supplemented with 1 mg ml À 1 insulin, 0.5 mM ascorbic acid, 25% horse serum, and 1% penicillin/streptomycin. Mouse brain slices were cultured for 2 days before iN cell transplantation. Two days after preparing brain slices, dissociated and scaffold-seeded GFP þ iN cells (day 6 post-conversion) were transplanted onto the surface of the cortico-striatal slices using a pipette. Medium was changed every 2-3 days and slices were fixed in 4% paraformaldehyde after 7 days for immunohistochemistry to quantify neurite outgrowth, or cultured 14 days before electrophysiology studies.
In vivo cell injection and scaffold transplantation. All animal experiments were carried out according to the Rutgers University Policy on Animal Welfare and were approved by the Institutional Animal Care and Use Committee (IACUC) at Rutgers University Robert Wood Johnson Medical School. Male, 5-week-old NOD-SCID IL2Rgc null mice (20-35 g; Jackson Laboratory) were anaesthetised with isoflurane (induction at 4% and maintained at 0.5-1% inhalation) and injected with a total volume of 10 ml, containing either 1 Â 10 3 , 1 Â 10 5 dissociated GFP þ iN cells or B10 scaffolds seeded with GFP þ iN cells (B85 cells per scaffold) at day 6 after initiating neuronal conversion, resuspended in ice-cold MEM (Life Technologies). Cells or cells on scaffolds were injected stereotactically into the striatum using a 100 ml gastight Hamilton syringe and 21G Hamilton needle. Bilateral injections were made at the following coordinates (in mm): AP, 0.5 (from bregma); ML, À 2.0; DV, À 3.0 (from dura). Mice were killed and processed for immunohistochemistry 3 weeks after transplantation for quantification of human iN cells within the graft by visualization of the human-specific nuclei antibody HuNu (Millipore) GFP, and RFP in the case of multiple-subtype co-transplantation (see Supplementary Table 1 for a detailed list of primary antibodies). Neurite outgrowth was quantified by measurement of GFP þ neurites.
Electrophysiology. Electrophysiolocial recordings were performed at prescribed time intervals from iN cells in 2D or 3D configurations in whole-cell mode using a Multiclamp 700B amplifier 12,45 . The bath solution contained 140 mM NaCl, 5 mM KCl, 2 mM MgCl 1 , 2 mM CaCl 2 , 10 mM HEPES and 10 mM glucose, at pH 7.4. The pipette solution for whole-cell voltage-dependent current recordings and for current-clamp experiments contained 10 mM KCl, 123 mM K-gluconate, 1 mM MgCl 2 , 10 mM HEPES, 4 mM glucose, 1 mM EGTA, 0.1 mM CaCl 2 , 1 mM K 2 ATP, 0.2 mM Na 4 GTP, pH 7.2. Membrane resting potentials were kept in the range of À 65 to À 70 mV, with step currents injected to elicit action potentials.
Immunocytochemistry. Neuronal phenotypes were characterized by immunocytochemistry for 2D and 3D substrates. Cells were fixed with 4% paraformaldehyde for 30 min at room temperature, followed by three washes with PBS. Cells were simultaneously blocked and permeabilized in blocking buffer consisting of PBS supplemented with 5% normal goat serum (MP Biomedicals), 1% bovine serum albumin (Sigma-Aldrich), and 0.1% Triton X-100 (Sigma-Aldrich) for 1 h at room temperature. Primary antibodies were incubated in blocking buffer at 4°C overnight, followed by three 15-min PBS washes. Fluorophore-conjugated secondary antibodies (Alexa Fluor 488, 594, or 647, Life Technologies) were incubated in blocking buffer for 1 h at room temperature, followed by three 15-min washes and at least one wash of 1-2 h. Samples were counter-stained with 4 0 ,6-Diamidino-2-phenylindole dihydrochloride (Sigma-Aldrich) to visualize nuclei and mounted with Prolong Gold Anti-Fade reagent (Life Technologies) before imaging. Samples were imaged on a Leica SP2 laser scanning confocal microscope using 10 Â dry or 63 Â immersion objectives. Images of cells on 3D electrospun fibres are presented as maximum intensity projections of z-stacks, unless otherwise stated. Image quantification was performed via custom ImageJ macros and manual counting. A complete listing of the primary antibodies used in this study can be found in Supplementary Table 1.
Calcium imaging. iN cells were labelled with 3 mM Fluo-4 AM calcium indicator dye for 30 min at room temperature in extracellular bath solution (147 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 10 mM HEPES and 10 uM Glucose, pH 7.4 and osmolarity 290-300 mOsm) containing 0.02% pluronic F-127. After incubation, cells were rinsed twice with bath solution and incubated for an additional 30 min at room temperature to allow complete de-esterification of the Fluo-4 AM dye before imaging. Time-lapse imaging was performed on a Leica SP2 confocal microscope at a resolution of 512 Â 512 pixels and a time resolution of 1 Hz. Image analysis was performed using custom ImageJ macros and MATLAB routines for semi-automated ROI selection and peak detection to quantify the fraction of cells that were electrically active. Cells were considered to be electrically active if they showed a spike in fluorescence following application of an electrical stimulus. The electrical stimulus was applied using a function generator (Global Specialties) and consisted of a 5 s stimulus at 7.5 V cm À 1 with 5 ms square pulses at 40 Hz.
Quantitative real-time polymerase chain reaction. Total RNA was extracted from iPS cells cultured in 2D or 3D using the RNEasy kit (Qiagen, Valencia, CA) according to the manufacturer's instructions, including treatment with RNase-free DNase to remove genomic DNA. A high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA) containing random primers was used to reverse transcribe 200 ng total RNA from each sample to cDNA. Taqman gene 194 expression master mix and Taqman gene-expression assays (Applied Biosystems) were used for template amplification of 10 ng cDNA per reaction. Quantitative real-time polymerase chain reaction (qRT-PCR) was carried out on a 7500 Fast Real-Time PCR instrument (Applied Biosystems). Relative gene expression was calculated using the DDCT method, normalizing to GAPDH as the endogenous control and undifferentiated IPS cells in standard 2D culture (on Matrigel-coated dishes in mTeSR media) after 7 days as the reference sample. Taqman gene-expression assays were used in these studies, listed in Supplementary  Table 2.
Statistical analysis. All data are presented as mean±s.d. Statistical significance is evaluated by single-factor ANOVA and a Tukey's post hoc test, with Po0.05 considered statistically significant.