Generation of Human Striatal Neurons by MicroRNA-Dependent Direct Conversion of Fibroblasts

Enhancement of miR-9/9*-124-Mediated Reprogramming

We previously noticed that a large fraction of cells underwent cell death while human fibroblasts were transduced to express miR-9/9*-124 (Yoo et al., 2011). In an effort to optimize the miR-9/9*-124-mediated neuronal reprogramming, we tested if co-expression of an anti-apoptotic gene would reduce the number of cells deaths during neuronal reprogramming. Previous studies have shown that abrogation of apoptosis could enhance neurogenesis (Sahay et al., 2011; Zhang et al., 2006) and that overexpression of an anti-apoptotic gene BCL2L1 (also known as Bcl-xL) suppressed programmed cell death (Alavian et al., 2011; Blomer et al., 1998). When we incorporated Bcl-xL into our lentiviral vector to be expressed with miR-9/9*-124 (Supplemental Experimental Procedures), we detected an increased number of surviving cells by over 40% and significantly improved the reprogramming efficiency (Figure S1). We utilized a doxycycline (Dox)-inducible promoter to temporarily regulate the expression of miR-9/9*-124 and Bcl-xL. Our characterization of the Dox-inducible promoter indicated that the transgene readily became inactive upon the removal of Dox within three days (Figure S2). We found that the continuous expression of miR-9/9*-124 for approximately 30 days was necessary for stable neuronal conversion as interrupting the transgene expression at day 15 resulted in a significant reduction in the neuronal conversion (Figure S2). Accordingly, in tissue culture as well as in experiments in vivo, we routinely withdrew Dox treatment after 30 days into reprogramming. Interestingly, we found that endogenous miR-9/9* and miR-124 were expressed after Dox removal by day 30 as the cells adopted a neuronal fate (Figure S2).

Identification of Transcription Factors to Specify Striatal Neurons

We focused on transcription factors expressed in the GABAergic medium spiny neurons (MSNs) of the striatum, a clinically relevant neuronal subtype affected in Huntington’s disease (Albin et al., 1989). Striatum-enriched transcription factors were selected based on translational ribosome affinity purification data for DRD1 or DRD2-positive medium spiny neurons from the mouse striatum (Dougherty et al., 2010) and gene expression databases from Brainspan (brainspan.org) and Allen Brain Atlas (human-brain-map.org). In order to identify transcription factors that would favor the MSN fate, we transduced human postnatal fibroblasts with lentivirus to express miR-9/9*-124 and sixteen selected transcription factors individually, and examined by immunostaining the number of MAP2-positive cells that were also positive for the neurotransmitter GABA and the dopamine- and cAMP-regulated neuronal phosphoprotein (DARPP-32), a well-documented marker of MSNs (Arlotta et al., 2008; Lobo et al., 2006; Ouimet and Greengard, 1990) (Figure 1A and Figure S3). BCL11B (also known as CTIP2), a transcription factor critical for the differentiation of MSNs in vivo (Arlotta et al., 2008), was the only factor tested with miR-9/9*-124 to yield DARPP-32-positive neurons (Figure 1A). Furthermore, when miR-9/9*-124 were combined with DLX1 and DLX2, previously shown to be important for terminal differentiation of MSNs (Anderson et al., 1997), we detected a large percentage of GABAergic neurons (72.3% of MAP2-positive cells) (Figure 1A). It is interesting to note that MYT1L, which has previously been used by our group and others to enhance direct neuronal reprogramming with other factors (Pang et al., 2011; Yoo et al., 2011), significantly increased the number of MAP2-positive cells when tested alone with miR-9/9*-124 (Figure 1A and Figure S3). Importantly, single transcription factors tested without miR-9/9*-124 did not induce MAP2-positive cells (data not shown).

Immunostaining Analysis of Neuronal Markers

Based on our initial screening of individual transcription factors, we asked if the combination of CTIP2, DLX1, DLX2 (DLX1/2) and MYT1L (collectively termed CDM) would synergize with miR-9/9*-124 to generate MSN-like cells (Figure 1A). We first determined if miR-9/9*-124 with CDM (miR-9/9*-124-CDM) would robustly generate neurons by examining the expression of general neuronal markers, MAP2, TUBB3 (also known as β Figure 1B). Counting random fields of view revealed that approximately 90% of DAPI-positive cells were MAP2-positive (n=207), 87% TUBB3-positive (n=328) and 84% NeuN-positive (n=328) at five weeks post-transduction (Figure 1C). The converted cells also expressed proteins important for neuronal function including voltage-gated sodium channels (Figure 1D, left) and Synapsin 1 (Figure 1D, right). The neuronal conversion was dependent on miR-9/9*-124 as transducing human fibroblasts with CDM factors alone was ineffective in generating neurons (0.3% MAP2-positive, n=343) (Figure 1E). Moreover, expression of CDM in human fibroblasts did not induce the expression of endogenous neural microRNAs such as miR-9/9*, miR-124 or miR-132 (Figure S4).

Detection of Markers of Medium Spiny Neurons

We found that the majority of the converted cells were GABAergic neurons (90% of MAP2-positive, n=474) assayed by immunostaining for GABA (Figure 1F, top left and Figure S5, top for larger overview) and GAD67 (Figure 1F, top right and Figure S5), without VGLUT1-positive (a marker for glutamatergic neurons) cells within the population of reprogrammed neurons (0.75% of MAP2-positive cells, n=400; Figure S6). Furthermore, we found that a large fraction of converted neurons (70% of MAP2-positive, n=474) expressed DARPP-32 (Figure 1G, Figure S7 for view of GABAergic DARPP-32-positive cells) and expressed FOXP1 and DLX5 (Figure 1F, bottom left and right, respectively, and Figure 1G for quantification, Figure S7 for larger fields of view) which are also highly expressed in MSNs (Desplats et al., 2006; Ferland et al., 2003; Stenman et al., 2003; Tamura et al., 2004). In addition, we found that the endogenous DLX1 was also expressed in cells converted by miR-9/9*-124-CDM (Figure S8). Reprogramming human fibroblasts using miR-9/9*-124 with DLX1/2 and MYT1L but without CTIP2 yielded highly GABAergic neurons that were completely devoid of DARPP-32 expression (Figure S9).

Interestingly, overexpression of CTIP2 (BCL11B) has been shown to protect hematopoietic progenitor cells from apoptosis, partially through the activity of members of the BCL2 family, including Bcl-xL (Albu et al., 2007). We therefore tested if ectopically expressing CTIP2 would induce the expression of Bcl-xL, but found no significant changes (Figure S4).

Single Cell Gene Expression Profiling

To further characterize the converted cells, we performed multiplex gene expression analyses in single cells to determine the enrichment of neuronal subtypes. Single cells induced from postnatal human fibroblasts were collected at five weeks post-transduction and identified based on the expression of housekeeping genes, GAPDH, RSP18 and HPRT1 (Figure 2A, left marked by green dendrogram; n=76 cells). Consistent with our immunostaining data, we found that the majority of the induced cells were positive for gene products expressed in neurons including MAP2, TUBB3 and MAPT as well as components of voltage-gated sodium channels (SCN2A and SCN3A), neural cell adhesion molecule (NCAM1), neuronal ankyrin (ANK2), brain-derived neurotrophic factor (BDNF), and the synapse-associated protein, PSD95 (also known as DLG4) (Figure 2A, left panel yellow dendrogram and Figure 2B, red bars). Interestingly, we did not detect the expression of doublecortin (DCX), a marker of immature, migratory neurons (Karl et al., 2005), suggestive of the maturity of the induced neurons. Importantly, miR-9/9*-124-CDM-induced cells were devoid of markers of dopaminergic neurons (DBH, DDC, DAT and TH), cholinergic neurons (CHAT), glutamatergic neurons (VGLUT2 and VGLUT3) and serotonergic neurons (SERT, TPH2, HTR2C and HTR3A) (Figure 2A, left, gray dendrogram and Figure 2B). Instead, the majority of converted cells were positive for GABAergic markers (GAD1/GAD2) and DARPP-32 (also known as PPP1R1B) in agreement with our immunostaining data. We also detected expression of genes associated with striatonigral (DRD1, CHRM4, TAC1) and striatopallidal (DRD2, PENK, GPR6) MSNs (Lobo et al., 2006) as well as transcription factors RARB and POU3F1 (Figure 2A, left, red dendrogram and Figure 2B). SHANK3, which has been shown to be important for the synaptic function of MSNs (Peca et al., 2011), was also expressed in a high fraction of the converted cells. Interestingly, we did not detect expression of genes enriched in striatal patch neurons (OPRM1 and PDYN) (Gong et al., 2003; Pert et al., 1976) in the converted cells. This suggests that the converted MSNs are likely to be of the striatal matrix (Figure 2B), which constitutes about 85% of the volume of the striatum (Johnston et al., 1990), consistent with the role of DLX1/2 in specifying matrix neurons (Anderson et al., 1997).

Human Striatum Laser Microdissection

To compare the gene expression profile of the converted cells to primary striatal neurons in the human brain, we analyzed human striatal cells collected by laser capture microdissection (LCM) from human postmortem adult striatal sections using the same set of assay primers. Consistent with the notion that the striatum is highly populated with MSNs (Gerfen, 1992), we detected that a large percentage of microdissected cells expressed DARPP-32. In addition, LCM samples expressed CTIP2, DLX1, DLX2 and MYT1L, and the overall gene expression profile was analogous to miR-9/9*-124-CDM-induced neurons (Figure 2A, right panel and Figure 2B, blue bars) (Figure S10 for pairwise comparison of LCM and reprogrammed MSNs). Importantly, miR-9/9*-124-CDM generated a homogenous population of MSNs, as we did not detect the expression of genes exclusive to striatal or cortical GABAergic interneurons (SST, PVALB, Calretinin, LHX6 and NKX2-1) (Marin et al., 2000; Tepper et al., 2010) (Figure 2). In both sets of data, genes enriched in the cerebellum (PCP2 and GRP), peripheral nervous system (PRPH), neural progenitors (EOMES, NES, ASCL1 and PAX6) and in cortical neurons (TBR1, NEUROD1, P75NTR, OTX1 and OTX2) were also largely undetectable (Figure 2). Finally, a pairwise comparison of the expression of genes tested between the reprogrammed cells and microdissected human MSN indicated a strong correlation with a coefficient of determination (R-squared) value of 0.77 (Figure S10). We also found miR-9/9* and miR-124 to be highly expressed in the human striatum (Figure S11), consistent with the expression of endogenous miR-9/9*-124 in converted MSNs (Figure S2).

Functional properties of reprogrammed human MSNs in tissue culture

The striatal circuitry is exceptionally intricate as it converges inputs from several brain regions with distinctive neurotransmitters, such as glutamatergic input from cortex and thalamus, dense dopaminergic innervation from substantia nigra, and local cholinergic and GABAergic inhibition (Graybiel et al., 1994). Many of these inputs have been shown to contribute to shaping the membrane properties and functional maturation of MSNs. For example, parvalbumin-expressing fast-spiking interneurons (PV-FSI) and neuropeptide-Y positive low-threshold spiking interneurons (NPY-LTS) have been shown to form synaptic connections with MSNs and regulate their firing activity (Do et al., 2012; Koos and Tepper, 1999). Therefore, MSNs can be distinguished in vivo not only due to their biochemical and anatomical differences, but also by their intrinsic physiological properties. MSNs have been shown to rest at a hyperpolarized membrane potential (−80 mV to −90 mV), have significant inwardly rectifying K⁺ currents at or below resting membrane potentials, and a firing pattern with long delays to the initial action potential (AP) with little or no spike frequency accommodation (Gertler et al., 2008; Venance and Glowinski, 2003). To examine the generic neuronal electrophysiological properties as well as MSN-specific properties in cells converted by miR-9/9*-124-CDM in tissue culture, we performed whole-cell patch-clamp recordings at six weeks post-transduction in converted cells cultured with rat primary glia. We found that all converted cells elicited multiple APs upon the injection of depolarizing currents (Figure 3A) and large fast-inactivating inward currents followed by outward potassium currents when evoked by a series of voltage steps (Figure 3B) (see Table S1 for intrinsic membrane properties) (n=12). Inward currents were ablated upon local perfusion of tetrodotoxin (TTX, 300 nM), indicating that they were mediated via TTX-sensitive voltage-gated sodium channels (Figure 3B). To test the ability of converted cells to become electrically active in the absence of primary glia, we decided to investigate the membrane properties of our cells cultured in isolation (monoculture). Surprisingly, converted neurons were also capable of firing APs, and displayed functional maturation over time in culture. Whereas the converted cells initially displayed modest action potentials at six weeks, a much more complex firing pattern with nearly 50% of the recorded cells firing repetitively was evident at 12 weeks post-transduction (shown in Figure 3D). In addition, converted cells displayed significantly larger peak amplitudes of inward currents at 12 weeks, correlated with increased excitability (see Figure S12 and Table S1 for additional traces and membrane properties). Since one of the major inputs impinging upon MSNs in vivo are excitatory glutamatergic afferents from the cerebral cortex (Kreitzer, 2009), we investigated if co-culturing the converted cells with rat cortical neurons would lead to the formation of functional synapses. As seen by the synaptic currents (Figure 3E) (8 out of 13 total cells recorded), reprogrammed MSNs successfully mature into synaptically active neurons. In addition to determining that the reprogrammed cells are excitable, we also detected long delays to initial spike (Figure 3F) (6 out of 12 total cells recorded), a typical firing pattern of MSNs, whereas we only detected modest inward-rectification (shown in Figure 3G and quantified in Figure 3H). We found that the degree of hyperpolarized resting membrane potential was minimal in tissue culture (−58.35 mV ± 3.08). The functional maturation of our cells from 6 to 12 weeks in monoculture (Figure S12) suggests that perhaps achieving a significantly hyperpolarized resting membrane potential requires longer culturing conditions. Consistently, MSNs from dissociated mouse striatal cultures have been shown to develop more negative membrane potentials with time in vitro (Lalchandani and Vicini, 2013).

Conversion of Adult Human Fibroblasts

Direct reprogramming of human postnatal and adult somatic cells into neurons has proved to be more challenging than reprogramming murine cells (Banito et al., 2009; Zhang et al., 2013). We asked whether the miR-9/9*-124-CDM-based approach could convert adult human fibroblasts into MSNs. We found that adult human dermal fibroblasts derived from a 42-year-old individual could be successfully converted into neurons with 82% of the cells being positive for MAP2 expression (n=405 total cells) (representative image in Figure 4A). Consistent with the postnatal cells, the majority of adult cells were also GABAergic (86%, n=147 total cells) and expressed DARPP-32 (77%, n=333 total cells) (Figure 4A). In fact, we found that miR-9/9*-124 could reprogram human fibroblasts from multiple individuals independent of the donor’s age with a similar efficiency (Figure S3).

The direct conversion of mouse fibroblasts into induced neurons has been reported with three transcription factors, BRN2, ASCL1 and MYT1L (Vierbuchen et al., 2010), while others have reprogrammed embryonic and fetal human fibroblasts with the same factors but supplemented with NeuroD1 (Pang et al., 2011; Son et al., 2011). Working solely with human postnatal fibroblasts, we showed that miR-9/9*-124 in combination with ASCL1, NEUROD2, and MYT1L successfully reprogramed fibroblasts into functional neurons (Yoo et al., 2011). The role for ASCL1 in direct neuronal reprogramming (Caiazzo et al., 2011; Pang et al., 2011; Yoo et al., 2011) has recently been reported to be due in part to its activity as a pioneer transcription factor (Wapinski et al., 2013), and we asked if miR-9/9*-124-mediated reprogramming involved the activity of ASCL1. We transduced human adult fibroblasts with miR-9/9*-124 to determine if ASCL1 expression would be induced, and found ASCL1 expression to be undetectable upon the expression of miR-9/9*-124 at post-transduction days 14, 21 and 30 (Figure 4B). We also attempted to reprogram human adult fibroblasts into MSNs with CDM factors in conjunction with ASCL1. In comparison to miR-9/9*-124-CDM, ASCL1-CDM generated very few MAP2 positive cells that lacked neuronal morphologies (Figure 4C). Collectively, these findings indicate that miR-9/9*-124-mediated neuronal reprogramming is independent of the activity of ASCL1.

Functional Integration of Converted MSNs in vivo

Next, we tested the ability of adult-derived cells to survive and differentiate in vivo by transplanting reprogrammed human MSNs into the mouse striatum. Adult fibroblasts were transduced with miR-9/9*-124-CDM and labeled with EGFP. Cells were harvested two weeks post-transduction, injected unilaterally into the striatum of immunodeficient mice and analyzed at day 50 after transplantation (DAT) (Figure 5A) (n=14). EGFP-positive cells were detected in the dorsal striatum adjacent to the lateral ventricle (Figure 5B). Remarkably, we regularly detected EGFP cells in mice even after 3.5 and 6.5 months, demonstrating the stable conversion and long-term survival of the reprogrammed cells (Figure 5C). EGFP-marked cells were 93% MAP2-positive (n=203), 76% GABAergic (n=174), 91% DARPP-32-positive (n=173) and 91% FOXP1-positive (n=173) (Figure 5D, 5E and 5F). Moreover, transplanted cells exhibited dense dendritic spines (Figure 5F, right).

Membrane Properties of Converted MSNs in vivo

We assessed the functional properties of reprogrammed MSNs in vivo in comparison to native mouse MSNs using whole-cell recordings in acute striatal slices at 112 DAT (Figure 6A). A total of 18 cells were targeted for recordings, including 11 non-EGFP native mouse MSNs (mMSNs) and 7 EGFP-positive human MSNs (hMSNs). Native mMSNs were identified and distinguished from striatal interneurons based on their characteristic membrane properties and AP firing patterns (Tepper et al., 2010). We found that all cells we examined (hMSNs and mMSNs) were able to fire AP trains when evoked by current steps, and displayed long delays to initial AP spike (shown in the inset of Figure 6A and 6B, and quantified in Figure 6I). Strong inward rectification, also characteristic of MSNs, was observed by an injection of voltage-ramp (Figure 6C for hMSNs and Figure 6E for mMSNs). The corresponding current-voltage relationship (I–V curve) was also identical between hMSNs and mMSNs (Figure 6K). Furthermore, we examined the AP threshold using a current-ramp protocol (Figure 6D for hMSNs and Figure 6F for mMSNs) and found no significant difference between the two cell types (Figure 6I). We then assessed whether transplanted cells were able to incorporate into local circuits by examining spontaneous postsynaptic potentials (sPSPs), which represent the network activity of surrounding neurons that form functional synapses onto hMSNs. As shown by the representative traces in (Figure 6G, 6H), sPSPs were readily detected in all hMSNs (7 out of 7) and mMSNs (11 out of 11) with no significant difference in the sPSP frequency (Figure 6I), providing further evidence that transplanted cells could integrate into functional circuits in vivo. Additionally, we found that the resting membrane potentials of hMSNs were hyperpolarized (−83.29 mV ± 2.40) in hMSNs, similar to native mMSNs (−85.20 mV ± 1.67) (Figure 6I). We quantified the spike frequency accommodation (SFA) ratio by calculating the frequency of the first spike (F_initial) in relation to the frequency of the average of the last two spike intervals (F_final), as previously reported (Venance and Glowinski, 2003). We found that both hMSNs and mMSNs displayed little to no spike accommodation with SFA values close to 1 (Figure 6J). Essentially, all membrane properties characteristic of MSNs that we analyzed were similar between transplanted and native MSNs during ex vivo recordings (Figure 6I).

Formation of axonal projections by transplanted MSNs

Striatal MSNs give rise to projections terminating in two distinct nuclei within the basal ganglia, the globus pallidus and the substatia nigra, and these pair of nuclei in turn forms the two major striatal output systems (Gerfen, 1992). We therefore investigated whether transplanted hMSNs could extend projections into these distal targets of endogenous MSNs. Remarkably, we found EGFP-positive projections extending from the injection site in the dorsal striatum to the substantia nigra (6 out of 6 brains analyzed at 56–64 DAT) with axon terminals also visible in the globus pallidus (Figure 7A). This finding signifies not only that reprogrammed human MSNs are capable of long distance axonal outgrowth, but also that they recognize and follow axonal guidance cues intrinsic to striatal projection neurons. Interestingly, although transplanted hMSNs are mostly confined within 1mm of the injection site (Figure 7B), we observed that a small number of cells (4.7%; n=2,770 cells) could be found beyond the striatal boundary (Figure 7C). Due to the required transduction of multiple lentiviral vectors, we suspected that cells adopting positions outside the striatum lacked the expression of one or more factors. We found that while the vast majority of our EGFP-positive cells residing within the striatum expressed CTIP2, the occasional EGFP-positive cells found outside the striatum mostly lacked the expression of CTIP2 (92.3% n=332) (Figure 7C). Since CTIP2 is highly expressed in the striatum as seen by immunostaining (Figure 7D), this finding may suggest that expression of CTIP2 may be important for determining migratory boundaries for cells destined to become MSNs within the developing striatum.

Cell Culture

Lentiviral preparation of a doxycycline-responsive synthetic cluster of miR-9/9* and miR-124 (Yoo et al., 2011) as well as transcription factors cloned downstream of the EF1α promoter were used to transduce human postnatal or adult fibroblasts (ATCC and ScienCell). Typically, infected human fibroblasts were maintained in fibroblast media for 2 days with doxycycline before re-plating onto coated coverslips. Cells were then selected with appropriate antibiotics in neuronal media (ScienCell) supplemented with valproic acid (1 mM), dibutyryl cAMP (200 µM), BDNF (10 ng/mL), NT-3 (10 ng/mL), and RA (1 µM). Doxycycline was replenished every two days and media were changed every 4 days. A detailed reprogramming protocol is available in Supplemental Experimental Procedures.

Immunofluorescence

Cells were fixed using 4% paraformaldehyde or combined with 0.2% glutaraldehyde (both from Electron Microscopy Sciences) for 20 min at room temperature (RT) followed by permeabilization with 0.2% Triton X-100 for 10 min. One percent goat or donkey serum was used for blocking followed by incubation of primary antibodies overnight at 4°C. Secondary antibodies conjugated to Alexa-488, −594, or −647 were applied for 1 hour at RT. Detailed list of antibodies used and procedure for immunohistochemistry analysis can be found in the Supplemental Experimental Procedures.

Single Cell Quantitative PCR

Reprogrammed postnatal cells were harvested and single cells were collected by FACS sorting 5 weeks post-transduction. De-identified human striatal sections from postmortem autopsy samples were obtained in accordance with the guideline of HRPO of Washington University School of Medicine. Post-mortem single cells stained with cresyl violet were laser-dissected from striatal putamen tissue sections of an 89-year-old individual. Samples were used for reverse-transcription and gene expression analysis using Fluidigm Dynamic arrays (see Supplemental Experimental Procedures).

Cell harvesting and transplantation

Reprogrammed cells were typically transduced with a Synapsin promoter-EGFP lentivirus at reprogramming day 2 and harvested for transplantation studies at reprogramming day 14. Adherent cell-cultures were mechanically dissociated and concentrated by centrifugation (4 minutes at 1000 RPM). Concentrated cell suspensions (~10⁴ cells/µL) were loaded into a 5 µl Hamilton syringe (26s/2/2 Gauge/Length/Point) and 2 µl injected into the right striatum of NSG mice (P0-P1).

Animals

Mice were maintained in standard housing conditions with food and water provided ad libitum. All procedures followed the guidelines of the Washington University's Institutional Animal Care and Use Committee (IACUC) and the Animal Studies Committee (ASC).

Electrophysiology, Slice Preparation and Recordings

Detailed methods and description of the composition of internal and external solutions used for ex vivo recording can be found in the Supplemental Experimental Procedures.