|Home | About | Journals | Submit | Contact Us | Français|
Human pluripotent stem cells are a promising source of differentiated cells for developmental studies, cell transplantation, disease modeling, and drug testing. However, their widespread use even for intensely studied cell types like spinal motor neurons is hindered by the long duration and low yields of existing protocols for in vitro differentiation and by the molecular heterogeneity of the populations generated. We report a combination of small molecules that within 3 weeks induce motor neurons at up to 50% abundance and with defined subtype identities of relevance to neurodegenerative disease. Despite their accelerated differentiation, motor neurons expressed combinations of HB9, ISL1 and column-specific markers that mirror those observed in vivo in human fetal spinal cord. They also exhibited spontaneous and induced activity, and projected axons towards muscles when grafted into developing chick spinal cord. Strikingly, this novel protocol preferentially generates motor neurons expressing markers of limb-innervating lateral motor column motor neurons (FOXP1+/LHX3−). Access to high-yield cultures of human limb-innervating motor neuron subtypes will facilitate in-depth study of motor neuron subtype-specific properties, disease modeling, and development of large-scale cell-based screening assays.
Generation of specific cell types from human pluripotent stem cells in vitro has the potential to yield novel insights into human development and disease. Directed differentiation of stem cells into spinal motor neurons allows access to cells that cannot be sampled in living human subjects and paves the way to their use for molecular studies, drug testing, or for targeted cell replacement therapies for neurological diseases and injuries. However, if stem cell-derived populations are to become usable in large-scale in vitro studies, robust methods for their production need to be optimized for high yield and, crucially, for generation of cell types with molecular and functional characteristics matching those of in vivo cells.
While all spinal motor neurons innervate skeletal muscles, different subpopulations vary widely in their functional properties and in their resistance to disease. For example, motor neurons that innervate limb muscles selectively express genes such as FOXP1 that determine their identity (Dasen et al., 2008; Rousso et al., 2008) and are among the earliest affected in many patients with amyotrophic lateral sclerosis (ALS) (Kanning et al., 2010). Our insights into motor neuron diversity have so far been primarily based on mouse genetic studies in vivo (Dasen et al., 2009). The recent advent of induced pluripotent stem cell (iPSC) technology combined with directed differentiation of pluripotent stem cells to motor neurons (Dimos et al., 2008; Ebert et al., 2009; Takahashi et al., 2007) opened the way to modeling motor neuron degenerative diseases and to developing therapeutic strategies in vitro (Dimos et al., 2008; Ebert et al., 2009; Hu and Zhang, 2009). However, the subtype identity of motor neurons produced by current protocols remains poorly defined, making functional studies and disease modeling difficult to interpret.
Here, we first characterized and validated a set of established markers of motor neuron diversity in the fetal human spinal cord. Next, by building on earlier reports, (Chambers et al., 2009; Lee et al., 2007b; Li et al., 2005; Li et al., 2008; Patani et al., 2011) we developed an accelerated and highly efficient protocol for directed differentiation of human embryonic stem cell (ESC) and iPSCs into motor neurons. Although the differentiation is accelerated, motor neurons generated under this protocol express canonical markers matching those in fetal human spinal cord in vivo, are responsive to glutamate agonists and show an ability to project axons when grafted into the developing chicken spinal cord. Most importantly, this method results in a novel preferential generation of motor neurons expressing markers of limb muscle-innervating lateral motor column (LMC) neurons. Thus, motor neurons produced under the new protocol might become an important and versatile tool to study subtype-specific differences in motor neuron function and susceptibility to degeneration in motor diseases such as ALS.
All cell cultures were maintained at 37 °C, 5% CO2. hES and iPS cells (HUES3 (control), male; H9 (control), female; HS001 (ALS-SOD1 N139K), male; LWM002 (ALS-SOD1 A4V), female; MBN007 (ALS-SOD1 A4V), female; TM008 (ALS-SOD1 A4V), female; DCM009 (ALS-SOD1 V148G), male; 10013.13 (control), male) were maintained on gelatinized tissue-culture plastic on a monolayer of irradiated CF-1 mouse embryonic fibroblasts (MEFs; GlobalStem), in hESC media, consisting of Dulbecco’s Modified Eagle Medium: nutrient mixture F-12 (DMEM/ F:12, Invitrogen) with 20% Knockout Serum Replacer (KSR; Invitrogen), 110 µM β-mercaptoethanol (BME; Sigma), L-Glutamine and non essential amino acids (NEAA; Invitrogen), and 20 ng/ml basic fibroblast growth factor (bFGF; Invitrogen) (Cowan et al., 2004). Media was changed every 24 hours and lines were passaged with dispase (Gibco, 1 mg/mL in hES media for 15–30min at 37 °C).
To generate motor neurons, undifferentiated hESCs were passaged using dispase (1 mg/mL) and triturated into small, 50- to 100-cell clumps and placed into ultra-low adherent culture dishes (Corning). For the first three days, cells were kept in suspension in hESC medium, supplemented with 10 µM Rho-associated kinase inhibitor Y27632 (Ascent Scientific) to enhance single cell survival (Watanabe et al., 2007), 20 ng/mL bFGF (Invitrogen) to enhance growth and 10 µM SB435142 (SB, Sigma) and 0.2 µM LDN193189 (LDN, Stemgent) for neuralization. At day 3, eymbroid bodies (EBs) were switched to neural induction medium (DMEM/F:12 with L-glutamine, NEAA, penicillin/streptomycin, heparin (2 µg/ml), N2 supplement (Invitrogen). At day 5, all-trans retinoic acid (RA; 0.1 or 1 µM, Sigma), ascorbic acid (0.4 µg/ml, Sigma), and BDNF (10 ng/mL, R&D) were added. Dual ALK inhibition (SB+LDN) was pursued until day 7. Hedgehog signaling was initiated on day 7 by application of either C25II modified SHH (R&D), at the standard concentration of 200 ng/ml, a human Smo agonist (HAG, 1 µM, gift from Lee Rubin (Boulting et al., 2011; Dimos et al., 2008)), mouse Smo agonist 1.3 (SAG, 1 µM, (Boulting et al., 2011; Frank-Kamenetsky et al., 2002; Wada et al., 2009; Wichterle et al., 2002)), or purmorphamine (PUR, 1 µM, (Li et al., 2008; Sinha and Chen, 2006), Stemgent). At day 17, basal medium was changed to Neurobasal (Invitrogen), containing all previous factors and with the addition of 10 ng/mL each of IGF-1, GDNF, and CNTF (R&D), plus B27 (Invitrogen). At day 20 or 30, EBs were dissociated with 0.05% trypsin (Invitrogen), and plated onto poly-lysine/laminin-coated 8-well chamber slides (BD Biosciences) at 0.2–0.5.106 cells/well, and/or 15-mm coverslips at 0.5.106. Plated neurons were cultured in the same medium with the addition of 25 µM BME, and 25 µM glutamic acid (Sigma), and fixed 1 day later.
For immunocytochemistry assays, cultures were fixed for 30 minutes with 4% PFA in phosphate buffered saline (PBS) at 4 °C, washed 3 times for 5 min in PBS, quenched and permeabilized in wash buffer (PBS, 0.1% Triton X-100) plus 50 mM glycine for 15 min. For the EB outgrowth RALDH2 staining, samples were fixed for 10 minutes at room temperature with 4% PFA/10% sucrose pre-warmed to 37°C. Samples were blocked with wash buffer plus 10% normal donkey serum for 1 hr and incubated with primary antibody (Table 1) overnight. Cells were then washed, incubated with DyLight coupled donkey primary anti secondary antibodies (Jackson Immunoresearch, 1:1,000). Finally, cells were washed and counterstained with DAPI (Invitrogen).
Quantitative image analysis of differentiated neuronal cultures was performed using the Multi-Wavelength Cell Scoring module in MetaMorph© software (Molecular Devices). Briefly, EBs were dissociated enzymatically and plated in the presence of neurotrophic factors at densities for which cell overlap was minimal. Following immunostaining, images of at least 9 randomly selected fields (>15,000 cells in total) for each condition were captured using a pre-programmed automated microscope stage. Images were analyzed using the “Multi-Wavelength Cell Scoring” module of the MetaMorph© software, using parameters pre-defined to count only unambiguous bright labeling for each antigen. Intensity thresholds were set while blinded to sample identity, to selectively identify positive cells that displayed unambiguous signal intensity above local background. These parameters were used on all samples in a given experiment, and only minimally adjusted for different staining batches as necessary. Script and Parameter files are available upon request (typically, a cell was ~5,000 grey levels above background to be called positive for any nuclear marker, and was ~10,000 for cytoplasmic markers). A minimum of 15,000 cells per sample was analyzed. All samples were imaged using 10× or 20× objectives on a Zeiss AxioObserver with a Coolsnap HQ2 camera (Photometrics). Some images were acquired using a structured illumination technique using an Apotome module (Zeiss) to achieve 1.9 µm optical sections to ensure co-localization of labeling. For the figures, the brightness and contrast of each channel of an image were adjusted in an appropriate manner to improve clarity.
For Ca2+ imaging experiments utilizing the Hb9::GFP reporter, stem cells were differentiated under the motor neuron differentiation protocol described above, dissociated at day 21 or day 31 and FACS-sorted based on GFP intensity with a 5 laser ARIA-IIu ROU Cell Sorter configured with a 100 µm ceramic nozzle and operating at 20 psi, BD BioSciences. The H9 assays were comprised of mixed neuronal cultures, which a parallel coverslip was stained and quantified to have 53% HB9/ISL1+ motor neurons. All cultures were plated onto 15–25 mm diameter coverslips at a density of 100,000–150,000 cells per coverslip in day 17+ neurobasal media with factors described above with the addition of 0.5 µM EdU, and matured 6 days prior to Ca2+ imaging. Cells were loaded with 3 μM Fluo-4 AM (Invitrogen, Carlsbad, CA) dissolved in 0.2% dimethyl sulfoxide/0.04% pluronic acid (Sigma) in HEPES-buffered physiological salt solution (PSS) for 1 hour at room temperature. PSS contained (mM): NaCl 145, KCl 5, HEPES 10, CaCl2 2, MgCl2 2 and glucose 5.5, pH 7.4. Cultures were continuously superfused with PSS at a rate of approximately 0.5 ml/minute. The cultures were imaged using a 10× objective on an inverted epi-fluorescent Zeiss AxioObserver microscope, equipped with a Coolsnap HQ2 camera (Photometrics). For imaging spontaneous Ca2+ transients, single sets of 200–300 images were acquired at a rate of approximately 2 Hz from each coverslip. For the kainate experiments, 36 images were acquired at a rate of 0.033 Hz and the superfusing PSS was replaced with PSS containing kainate (100 μM) for 60 seconds. Image analysis was performed using ImageJ (http://rsb.info.nih.gov/ij/) or AxioVision 4.7 (Zeiss). Ca2+ transients were determined from regions of interest encompassing the soma of individual cells. A minimum of two cultures obtained from a single differentiation of each cell line and each time point were used for the kainate and all Ca2+ imaging experiments.
For whole cell patch clamp recordings, S+P differentiated HUES3 Hb9::GFP cells were plated on polyornithine/laminin-coated 25 mm diameter coverglass at density of 50,000 per coverslip and cultured for 7 days in the presence of 0.5 µM EdU prior to recording (i.e. DIV 21+7). Current clamp recordings were carried out using an Axopatch 2B amplifier. Data were digitized using a Digidata 1322A digital to analogue converter and were recorded at a 10 KHz sample rate using pClamp 10 software (all equipment from Molecular Devices). Patch pipettes were fabricated using a P-97 pipette puller (Sutter Instruments). The external recording solution contained (in mM), 145 NaCl, 5 KCl, 10 HEPES, 10 glucose, 2 CaCl2, 2 MgCl2. The pH was adjusted to 7.3 using NaOH and the osmolality adjusted to 325 mOsm with sucrose. The pipette solution contained (in mM): 130 CH3KO3S, 10 CH3NaO3S, 1 CaCl2, 10 EGTA, 10 HEPES, 5 MgATP, 0.5 Na2GTP, pH 7.3, 305 mOsm. Experiments were carried out at room temperature (21 – 23 °C). During recordings, current was injected to hold the cells at approximately −60 mV. Action potentials were evoked using incrementally increasing current steps 1 s in duration. The maximum amplitude of the current step (20 – 50 pA) and the size of the increment was calculated based on the input resistance of the cell.
To perform xenotransplantations day 21 EBs from HUES3 Hb9::GFP under the ventralization with SAG+PUR were collected and placed into L-15 media (Invitrogen) containing penicillin/streptomycin (GIBCO). Transplantation was performed as previously described (Wichterle et al., 2002). Briefly, after a small suction lesion at the prospective intraspinal site was created in a chick embryo at stage 15–18 at somites 15–20, lightly triturated EBs were loaded into a handheld micro-injector. The EBs was placed into the lesion. After 48 hours, the chicks were sacrificed, fixed with 4% PFA for 2 hours at 4°C, and neurite outgrowth and cell body placement was accessed by cutting 200 µm vibratome sections (n = 2), and by cutting 30 µm sections along the spinal cord (n = 5).
Human fetal spinal cords were collected in accordance with the national guidelines of the United States (NIH, FDA) and the State of New York and under Columbia University institutionally approved ethical guidelines relating to anonymous tissue. The fetal material was obtained after elective abortions, and was classified on the basis of external morphology according to the Carnegie stages. Gestational age was determined by last menstrual period of the patient or by ultrasound, if the ultrasound estimate differed by more than one week as indicated by the obstetrician. The spinal cord was removed as intact as possible prior to fixation with fresh, cold 4% PFA for 1.5 hours on ice. Post fixation, the cord was measured and cut into 3 anatomical sections to accommodate embedding in OCT Compound (Tissue-Tek, Redding, CA) and stored at −80 °C prior to cutting on a microtome. 12µm sections were cut along the full length of the cord, taking care to have all 3 sections on each slide in 7 independent sections. This allowed for full analysis and internal staining controls since each slide had cervical, brachial, thoracic and lumbar sections that clearly showed staining within the various motor columns present at different rostal-caudal levels of the spinal cord.
cDNA was obtained from 50,000 FACS purified MN’s from either day 21 S+P (methods described above), or from RA/SHH MN’s at day 31. cDNA preparation was carried out using commercially available kits following the manufacturer’s instructions: RNA isolation (Trizol LS; Invitrogen), cDNA by Brilliant II SYBR green (Stratagene) without amplification. All samples were processed in parallel on the same qPCR plate.
|STD qPCR amplification: 95°- 30”, 55°-60”, 72°-45”|
For paired-end RNA-Seq experiments, 400 ng of total RNA was prepared after FACS purification of 500,000 GFP+ or GFP− cells. The RNA samples were then amplified using a NuGEN RNA kit for genomic sample amplification, and sequenced to a depth of 21 (S+P) and 35 (SHH) million paired-end reads on an Illumina HiSeq instrument at the HudsonAlpha Institute of Biotechnology. The reads were aligned to the reference transcriptome as well as a library of exon junctions using Bowtie (Version 1) (Langmead et al., 2009). Data was analyzed using Expression Plot (Friedman and Maniatis, 2011) using a P value of 0.001 and a 2 fold change threshold. Gene ontology was performed using DAVID (Huang et al., 2008, 2009) with enrichment sets from Expression Plot. The RNA-seq data is available in the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE41795.
All quantitative data was analyzed using Sigma Plot 11 or Microsoft Excel. Sample groups were subject to Student’s t-test or where appropriate a One-Way ANOVA with Holm-Sidak post hoc pair-wise comparisons was performed. All experimental data passed an equal variance and normality test (Shapiro-Wilk).
Specification of motor neuron identity depends on three critical steps - neuralization, caudalization and ventralization of precursors (Wichterle et al., 2002). Previous reports have shown that dual inhibition of SMAD signaling by a combination of recombinant noggin protein, or a small-molecule substitute, LDN193189, and SB431542 neuralizes hESCs with very high efficiency (Boulting et al., 2011; Chambers et al., 2009; Kriks et al., 2011), bypassing the need for manual rosette picking. Thus, we sought to determine the effects of optimal caudalization and ventralization on motor neuron yields following neuralization. To assess motor neuron numbers we relied initially on the HUES3 Hb9::GFP reporter line (Di Giorgio et al., 2008), which contains a transgene expressing GFP under the control of the motor neuron-specific Hb9 promoter.
Using a standard RA/SHH protocol involving all-trans retinoic acid (RA) and a modified sonic hedgehog (SHH-C25II) protein as a benchmark for differentiation of neuralized hESCs into motor neurons, we systematically compared the ventralizing activity of three Smoothened (Smo) agonists (Boulting et al., 2011; Frank-Kamenetsky et al., 2002; Li et al., 2008; Wada et al., 2009). The abundance of motor neurons was estimated on day 31, following dissociation and plating on day 30 (see Methods), by counting the percentage of all cells that expressed GFP (Figure 1A, n = 3 biological replicates for all experiments unless specified). Very few GFP-positive cells were observed in the absence of exogenous SHH agonists. Recombinant SHH and human-specific Smo agonist (HAG) each gave rise to <10% GFP+ cells, as previously reported (Boulting et al., 2011; Di Giorgio et al., 2008; Lee et al., 2007b). In contrast, the Smo agonist SAG alone gave rise to 16 ± 4% GFP+ cells and purmorphamine (PUR) alone induced 22 ± 6%. To optimize caudalization of the cultures, we then tested a higher concentration of RA (1 µM, Figure 1B). This led to an additional increase in the abundance of GFP+ cells: 27 ± 4% of cells expressed GFP at day 31 when ventralized with SAG (1 µM) alone or 24 ± 6% with PUR (1 µM) alone (Figure 1C). When cultures were exposed to a combination of 1 µM SAG and 1 µM PUR 29 ± 4% cells were GFP+ (Figure 1C) on day 31. Systematic titration of PUR and SAG alone or in combination did not further increase motor neuron abundance on day 21, indicating that the combination of 1 μM SAG and 1 μM PUR is close to optimal (data not shown). Thus, for subsequent experiments we treated cells with a combination of RA, SAG and PUR (we refer to this as the “S+P” method) or RA and SHH (“SHH”) as a control.
We noticed that with the S+P method GFP-positive cells appeared earlier in culture than with SHH. Using the Hb9::GFP reporter line we therefore examined the time-course of motor neuron generation from neuralized precursors using either SHH- or S+P-mediated ventralization (Figure 1D). By day 21 under S+P conditions, GFP+ cells had already reached a plateau at 29 ± 3% of all cells (Figure 1E, F), a value significantly higher than with SHH (12 ± 1% GFP+; n = 3–4, P = 0.001) at the same time (Figure 1F, G). These data indicate that treatment of neural precursors with a combination of S+P accelerates motor neuron differentiation, possibly through enhanced and more uniform ventralization.
To ensure that the S+P condition did not induce ectopic GFP reporter expression in non-motor neurons, we examined the expression of endogenous glial, neural crest, pan-neuronal and motor neuron specific markers (Hester et al., 2011; Hu and Zhang, 2009; Li et al., 2005; Wada et al., 2009). At day 21 there were no GFAP- or SOX10-positive glial cells (data not shown). Less than 1% of all cells expressed BRN3A, a marker of neural crest-derived sensory neurons and spinal interneurons (Lee et al., 2007a) and none of the BRN3A+ cells co-stained with GFP (Figure 2A). A majority of plated cells (83 ± 1% of DAPI+, n = 2) expressed the pan-neuronal marker βIII-tubulin (TUJ1), consistent with efficient neuralization (Figure 2B), and essentially all GFP+ cells (98 ± 0%; n = 2) exhibited neuronal morphology and were positive for TUJ1 (Figure 2B). Immunostaining with SMI-32 antibody recognizing a mature neuronal marker neurofilament heavy chain labeled most GFP-expressing cells in S+P cultures (Figure 2C). Quantitative real time PCR (qPCR) analysis of RNA isolated from FACS-sorted GFP positive cells from both S+P and SHH cultures revealed 700- to 800-fold induction of high-affinity choline transporter 1 (CHT1 or SLC5A7), a marker of cholinergic neurons, when compared to RNA isolated from undifferentiated ESCs. Thus, optimized differentiation of human ESCs results in efficient and accelerated production of cells expressing markers of mature spinal motor neuron identity.
To determine whether the accelerated differentiation was also applicable to other cell lines that do not carry a fluorescent motor neuron reporter we established parameters for scoring motor neuron induction immunocytochemically, using antibodies to the motor neuron markers HB9 and ISL1. Although nearly all GFP+ cells (83 ± 4%) expressed either HB9 and/or ISL1 (Figure 2D, E), many did so in a non-overlapping manner. Indeed, of all GFP-expressing cells 16 ± 5% were HB9-positive, 30 ± 6% were ISL1-positive, and 37 ± 2% were double-positive (Figure 2E). Since previous studies have often considered Hb9 and ISL1 to be overlapping markers for motor neurons, we asked whether this heterogeneity had an in vivo counterpart. In the developing mouse spinal expression of Hb9::GFP reporter is maintained in all postmitotic motor neurons (Arber et al., 1999; Thaler et al., 2004; Wichterle et al., 2002), whereas HB9 and ISL1 transcription factors are selectively downregulated in medial and lateral divisions of the lateral motor column (LMCm and LMCl), respectively (Thaler et al., 2004). Therefore, the observed non-overlapping pattern of HB9 and ISL1 expression may reflect acquisition of distinct motor neuron subtype identities by in vitro-generated motor neurons.
To establish whether a similarly complex pattern of ISL1 and HB9 expression extends to human cells in vivo, we examined the expression of these markers in human fetal spinal cord at the 7th week of development (Carnegie stage 19–22), a stage analogous to mouse E13–14 when motor columns are anatomically well demarcated (Figure 2F–H). At brachial and lumbar limb levels, motor columns corresponding to the LMCm and LMCl, and to the median motor column (MMC) were apparent. Consistent with the pattern of expression in chick and mouse, the MMC expressed both HB9 and ISL1, the LMCl was ISL1−/HB9+, and many LMCm neurons were ISL1+/HB9−. Interestingly, a group of motor neurons located within the LMCm division were double-positive for ISL1 and HB9, an expression pattern not reported in chick and mouse spinal cord (Figure 2G, and data not shown). At thoracic levels, putative MMC and hypaxial motor column (HMC) neurons co-expressed HB9 and ISL1, with a tendency for higher HB9 expression in the MMC and higher ISL1 expression in HMC neurons (Figure 2H). To estimate the relative abundance of cells expressing different combinations of HB9 and ISL1, motor neurons with bright HB9 or ISL1 labeling were evaluated in multiple brachial and thoracic segments. Overall, 44% of all motor neurons expressed ISL1 alone, 31% expressed HB9 alone, and 25% expressed both HB9 and ISL1 (Figure 2F–H). This distribution was remarkably similar to the distribution of single- and double-positive motor neuron populations observed in vitro (Figure 2F). Therefore, as in developing avian and rodent spinal cords, human motor neurons are characterized by variable patterns of HB9 and ISL1 expression.
Based on these observations we suggest that a combination of HB9 and ISL1 (referred to subsequently as “pan-MN”) is the most accurate measure of total motor neuron yield when human cell lines lacking a GFP reporter are differentiated into motor neurons. However, since not all ISL1 positive cells in the developing spinal cord are motor neurons, we considered potential sources of error in this parameter. For the Hb9::GFP line, the pan-MN value was 42 ± 1%, which is indeed similar to the total percentage of GFP-expressing cells (31 ± 2%, Figure 1F). Although >98% of all cells that express HB9 were GFP+, a smaller fraction, 71 ± 5%, of all ISL1+ cells co-express GFP. The remaining 29% that did not express GFP may correspond to non-motor neurons, meaning that of the pan-MN value (42%), the fraction that is ISL1+ only may be overestimated by 9%. However, this potential error is minor compared to that induced by counting HB9 or ISL1 labeled cells only. Although frequently used, such approaches potentially underestimate total motor neuron abundance by ~20%. Therefore, pan-MN staining provides a more reliable estimate of motor neuron number.
Using this criterion, we asked whether the S+P method, which had been optimized using a single hESC line, was applicable to other human stem cell lines. The H9 hESC line was recently reported to show stronger proneural characteristics than the HUES3 Hb9::GFP reporter line (Bock et al., 2011). Using the accelerated protocol, H9 was indeed more efficient than HUES3 at producing motor neurons: a total of 51% of all cells expressed the pan-MN signature (Figure 2I). We expanded our analysis to a set of 6 induced pluripotent stem (iPS) cell lines generated from healthy patients and patients with ALS. We found that all the lines generated motor neurons under S+P conditions with a similar efficiency to that of control human ESCs by day 21 of culture (27 ± 1% pan-MNs, 6 lines each differentiated in 2 biological replicates, Figure 2I). We conclude that the S+P protocol provides a robust and general approach for the generation of motor neurons from human stem cell lines.
To further characterize stem cell derived neurons in an unbiased manner, we performed whole-transcriptome sequencing of RNA (RNA-seq) from FACS-purified S+P-derived and SHH-derived motor neurons (Hb9::GFP reporter line) and compared these with the GFP-negative fraction of the S+P culture, which contains <4% of pan-MN+ neurons. RNA-seq reads were aligned to the reference human genome, normalized, and expression level of individual genes calculated as Reads Per Kilobase per Million mapped reads (RPKM). The profiles of housekeeping gene GAPDH, neuronal marker neurofilament light chain NEFL and motor neuron marker ISL1 corroborate the neuronal and motor neuron identity of sorted GFP-positive cells (Figure 3A). Global analysis revealed 145 genes enriched in S+P motor neurons compared to their negative fraction (Figure 3B, C). Markers of spinal motor neurons (ISL1, ISL2, HB9, RET) as well as key cholinergic genes (CHAT, CHT1, VACHT, CHRNA3, CHRNA4, CHRNB2) were enriched in GFP-positive cells (Figure 3D). For example, HB9 was enriched 21.6 fold and CHT1 13.7 fold. Unbiased gene ontology (GO) analysis revealed that transmission of nerve impulses, neuronal development and differentiation, synaptic transmission, and axon pathfinding were the top four sets of cellular processes defining S+P-derived GFP-positive cells (Figure 3E). Importantly, the cluster of genes associated with neuronal development and differentiation contained key motor neuron markers (MNX1 (HB9), RET, CHAT, ISL1) (Figure 3E), further supporting the conclusion that the S+P-differentiated cells acquired motor neuron identity.
A crucial step in motor neuron maturation is the acquisition of a defined columnar subtype identity. Based on the mutually exclusive expression of LHX3 and FOXP1 transcription factors, motor neurons in chick and mouse embryos can be subdivided into three major categories: median motor column (MMC) neurons that express LHX3 and innervate axial muscles, lateral motor column (LMC) neurons that innervate the limb muscles and express FOXP1, and finally the hypaxial motor column (HMC) neurons that express neither FOXP1 or LHX3 and innervate the hypaxial muscles (Dasen et al., 2008). We first asked whether differential expression of LHX3 and FOXP1 could be employed to determine motor neuron columnar identities in developing human spinal cord (Figure 4A). Consistent with the pattern of expression in the developing chick and mouse spinal cord (Dasen et al., 2008; Otaegi et al., 2011), strong FOXP1 immunoreactivity was detected in lateral HB9+ or ISL1+ motor neurons at limb levels as expected for LMC neurons (Figure 4A, B), and in a small number of scattered HB9−/ISL1− cells outside of the ventral horn, which are presumptive interneurons. LHX3 marked the most medial group of presumptive MMC motor neurons, and a more dorsally located group of interneurons (Figure 4A, B). Lateral to MMC neurons we observed a group of motor neurons that lacked expression of both FOXP1 and LHX3, which is consistent with an HMC identity (Figure 4A, right panel). Based on these studies we conclude that similar to the chick and mouse, expression of LHX3 and FOXP1 in human embryonic spinal cord distinguishes motor neurons of different columnar subtype identities.
We examined expression levels of FOXP1 and LHX3 markers in the RNA-seq data. Comparison of profiles of GFP-positive cells revealed a ~1.3-fold increase in FOXP1 and a ~1.8-fold decrease in LHX3 RNA levels in S+P as compared to SHH motor neurons, suggesting that the two differentiation protocols might yield motor neurons of different columnar subtype identities. We confirmed the increase in FOXP1 and decrease in LHX3 RNA levels in S+P motor neurons by qPCR analysis in 3 independent biological samples of purified GFP+ motor neurons (Figure 4C). Immunostaining revealed that S+P-differentiated cultures contained significantly higher percentages of FOXP1+ motor neurons (68 ± 4%, n = 6 biological replicates, Figure 4D, E), and a corresponding decrease in the number of motor neurons expressing LHX3 (9 ± 3% of S+P motor neurons compared to 64 ± 2% of SHH motor neurons) (Figure 4D, E). Importantly, FOXP1 and LHX3 were expressed in a non-overlapping manner indicating proper segregation of columnar subtype identities in cultured human motor neurons (data not shown). FOXP1+ cells expressing GFP under S+P conditions exhibited neuronal morphology and stained for a non-phosphorylated form of heavy neurofilament (SMI-32) that, within the spinal cord in vivo, is selectively expressed in motor neurons (Clowry et al., 2005) (Figure 4F). Using co-labeling for FOXP1 and pan-MN as a criterion to identify LMC motor neurons we expanded our analysis to a set of 6 additional iPSC lines generated from healthy donors and patients with ALS. We quantified the percentage of FOXP1+ motor neurons and found that the S+P method generated on average 62 ± 3% LMC motor neurons, consistent with our hESC study (Figure 4G).
To further substantiate the LMC identity of S+P-generated motor neurons we examined the expression of another LMC marker, RALDH2 (Nedelec et al., 2012; Sockanathan and Jessell, 1998). In human spinal cords, RALDH2 immunoreactivity was largely restricted to the cytoplasm of FOXP1+ LMC neurons at limb levels (Figure 4H). Quantitative PCR analysis revealed an increase in LMC marker RALDH2 expression in S+P-differentiated motor neurons (Figure 4C), but RALDH2 staining of dissociated cultures produced only a weak signal that was difficult to interpret. However, when S+P-derived embryoid bodies were plated on laminin coated substrata we observed a subset of GFP+ motor axons that were co-labeled with anti-RALDH2 antibodies, further supporting the notion that motor neurons differentiated under the S+P condition acquired LMC columnar identity (Figure 4I).
Rodent, avian, and human LMC neurons are found in brachial and lumbar spinal domains harboring motor neurons innervating fore/upper and hind/lower limbs, respectively ((Dasen et al., 2008), Figure 4A, and data not shown). Brachial and lumbar LMC neurons can be distinguished by their respective expression of HOX6 and HOX10 transcription factors. Brachial LMC can be further subdivided into a rostral segment expressing HOXA5 transcription factor and caudal HOXC8 positive domain (Dasen et al., 2009). To refine the identity of these LMC neurons we examined expression of HOX genes in human ESCs differentiated under the S+P condition. We determined that differentiated cells exhibit elevated levels of HOXA5, HOXC6 (6.03 and 23.8, S+P fold change over ESCs respectively) and to a lesser extent HOXC8 expression compared to undifferentiated ESCs. In contrast we did not detect significant expression of HOXC9 or HOXC10 (Figure 5A, and data not shown), indicating that the motor neurons are primarily of rostral brachial identity.
To further characterize the LMC neurons generated using S+P we asked to which division of the LMC they belonged (Figure 5B). Motor neurons in the lateral division of the LMC (LMCl) express LHX1 and HB9 in vivo (Palmesino et al., 2010) (Figure 5C), while those in the medial LMC (LMCm) express FOXP1 and ISL1 while lacking LHX1 (Figure 5B) (Palmesino et al., 2010; Thaler et al., 2004). Interestingly, a substantial portion of LMCm neurons in the human spinal cord also co-express ISL1 and HB9 (Figure 2G). Analysis of the LMC population by HB9 and ISL1 expression in both SHH- and S+P-derived FOXP1 motor neurons revealed that under both conditions most FOXP1 cells expressed ISL1 (either alone or in combination with HB9), suggestive of an LMCm identity (Figure 5D). The remaining 20 ± 1% of the S+P FOXP1+ motor neurons expressed HB9 only, indicating that these might be of LMCl identity. However, very few of these ISL1− HB9+ FOXP1+ motor neurons expressed LHX1 (2 ± 1%) (Figure 5E), a definitive marker for LMCl motor neurons in vivo (Figure 5C), raising the possibility that in vitro conditions do not support full maturation of LMC subtype identity by 21 days in culture. The HB9+ FOXP1+ LHX1− cells might be LMCm neurons that failed to downregulate HB9 or LMCl neurons that failed to acquire LHX1.
Exposure of nascent LMC motor neurons to high concentrations of RA promotes specification of LMCl identity (Sockanathan and Jessell, 1998). We therefore tested whether culturing S+P generated LMC neurons in the presence of high concentrations of RA would increase the number of cells displaying LMCl character. Indeed, culturing day 21 motor neurons in the presence of 5 µM RA for 3 days post-plating significantly increased the fraction of LMCl neurons, as judged by the co-expression of LHX1 and HB9 within the FOXP1+ motor neuron population (1.3 ± 1% using 1 µM RA vs. 4.4 ± 0.2% using 5 µM RA, P < 0.05, Figure 5F). Taken together, these observations indicate that S+P motor neurons acquire a predominantly LMCm phenotype and that acquisition of LMCl identity can be potentiated by the treatment of differentiating cells with a high concentration of RA.
To determine whether the accelerated FOXP1+ S+P cultures yielded functional motor neurons, we investigated their behavior in vitro and in vivo. We previously reported that human ES/iPS-derived motor neurons display spontaneous activity and generate Ca2+ transients upon exposure to the glutamate receptor agonist kainate (KA) by day 35 of differentiation (Boulting et al., 2011). To determine whether functional maturation keeps pace with the accelerated appearance of molecular markers, day 21 S+P-generated LMC motor neurons were FACS-purified based on GFP intensity. We first confirmed that FACS-purified cells maintained their LMC identity in culture (Figure 6A, B). Motor neurons were cultured for 6 days then loaded with Fluo-4 AM, a Ca2+-sensitive indicator. Overall, 62 ± 5% (cells = 58) showed spontaneous variations in Ca2+ transients (Figure 6C) and all cells responded to a brief pulse of KA with a large increase in Fluo-4 fluorescence intensity that returned rapidly to baseline levels (cells = 17, Figure 6D, E). Similar results were obtained using the H9 hESC line (60 ± 1% spontaneous activity; cells = 63) or SHH motor neurons at day 31 + 6 (57 ± 3%; cells = 58). In order to determine whether S+P-derived motor neurons are capable of firing action potentials (APs), whole cell patch current clamp recordings were carried out using GFP-positive HUES3 Hb9::GFP cells. In all cells tested (n = 9), APs could be evoked by 20–50 pA, 1-second current injections (Figure 6F). These results suggest that motor neurons derived using the accelerated S+P protocol acquire physiological properties consistent with functional motor neurons, and are as mature as motor neurons derived under the standard 31 day SHH protocol.
Motor neurons are the only CNS neurons that innervate peripheral tissues. To study the ability of S+P motor neurons to project axons towards the muscle targets, differentiated cells were transplanted in ovo into the lesioned neural tube of chicken embryos at HH stage 15–16 (Figure 6G) (Lee et al., 2007b; Peljto et al., 2010; Wichterle et al., 2002). A majority (86%) of transplanted embryos showed successful engraftment of human motor neurons (as marked by GFP expression) into the ventral horn (n = 7). Importantly, we observed GFP-labeled axons of grafted motor neurons projecting through the ventral (and more rarely, dorsal) roots and along the peripheral nerves of the host in all of the successfully transplanted chicken embryos (n = 6 of 7) (Figure 6G). These data indicate that human motor neurons generated by the accelerated protocol exhibit proper ability to extend their axons outside of the CNS and follow typical motor nerve trajectories.
We describe a novel approach for high-yield, high-abundance production of motor neurons from human stem cells with a defined and controllable subtype identity. The protocol generates motor neurons within three weeks using only small-molecule compounds to drive neuralization, caudalization and ventralization of precursors. Despite the accelerated differentiation, the resulting motor neurons express HB9 and ISL1 in a manner that closely resembles the pattern observed in human motor neurons in vivo and exhibit functional properties similar to those of primary rodent neurons. They also have a clear motor neuron-like signature by immunocytochemistry, qPCR, and RNA-seq profiling. Unexpectedly, the new method also produced highly enriched populations of limb-innervating motor neurons, which are of fundamental interest to developmental studies and disease modeling.
Human ESCs and iPSCs represent a potentially valuable resource for modeling human developmental biology and neurodegenerative disease in vitro (Han et al., 2011; Kanning et al., 2010; Liu and Zhang, 2010). However, development of successful cell-based assays has been hindered by the long duration and low efficiency of many differentiation protocols. Optimization of motor neuron differentiation depends on the development of unbiased methods for motor neuron quantification in cultures. Our analysis of HB9 and ISL1 expression in the developing human spinal cord reveals that these markers are expressed only in subsets of spinal motor neurons. To prevent underestimating yields, or missing a subset of motor neurons, we therefore relied on a combined expression of the two markers (“pan-MN” staining). To further eliminate any bias in quantification resulting from analysis of select embryoid bodies or clumps of adherent cells in cultures, differentiated cells were dissociated followed by either flow cytometry (in the case of Hb9::GFP reporter) or semi-automated randomized sampling of fields of plated cells immunostained for HB9/ISL1 expression. The acquired images were analyzed using the Metamorph software package to further reduce potential experimenter bias. We propose that the adoption of this approach as a standard method for quantification of motor neuron yields, as it will allow future cross-comparisons of differentiation protocols between different investigators and labs.
Using this quantification method, we performed a systematic optimization of motor neuron differentiation using off-the-shelf small-molecule analogs of patterning factors. Since the initial neuralization step had already been optimized, (Boulting et al., 2011; Chambers et al., 2009) we focused on the subsequent steps of caudalization and ventralization, aiming to specify ventral spinal motor neuron progenitors with high efficiency. The resulting procedure, utilizing RA, SAG and PUR, is robust and efficient and reduces the time and cost of the procedure, making human stem cell-derived motor neurons a more widely accessible reagent.
Accelerating the speed of motor neuron specification raised a concern as to whether the resulting motor neurons would acquire definitive motor neuron physiological properties. Previous reports using either Ca2+ imaging at day 35 in vitro (Boulting et al., 2011), or electrophysiological recording (Lee et al., 2007b; Li et al., 2008; Singh Roy et al., 2005; Takazawa et al., 2012) have demonstrated that stem cell-derived motor neurons become electrically active with prolonged time in culture. Despite their accelerated differentiation, the motor neurons derived by the S+P protocol exhibit spontaneous Ca2+ transients, fire APs, and project axons along major motor nerves upon transplantation to the developing chick spinal cord, paving the way for follow-up studies of the in vivo behavior of human motor neuron subtypes (Patani et al., 2011; Peljto et al., 2010).
We observed that the majority of S+P neurons acquired the identity of limb-innervating LMC motor neurons. This was a surprising finding, as most of the current differentiation protocols for mouse or human motor neurons yield predominantly the MMC motor neuron subtype (Patani et al., 2011; Wichterle et al., 2002). While most of the limb-innervating motor neurons (~80%) expressed markers consistent with LMCm divisional identity, retinoic acid treatment promoted specification of LMCl identity, demonstrating that in vitro-derived LMC motor neurons are responsive to developmental cues controlling motor neuron subtype differentiation. Mechanisms that underlie specification of columnar identity have been a focus of several recent studies. The expression of the LMC determinant FOXP1 in the developing mouse and chick spinal cord is controlled by a rostro-caudally restricted expression of brachial and lumbar Hox genes (Dasen et al., 2008). Establishment of MMC identity on the other hand relies on non-canonical Wnt signals expressed by ventrally located floor plate cells (Agalliu et al., 2009). Thus, the observed increase in the number of FOXP1-positive motor neurons might be a result of efficient ventralization of embryoid bodies expressing limb-level Hox genes in combination with an inhibition of non-canonical Wnt signaling. Alternatively, S+P conditions might accelerate maturation of LMC motor neurons present under all differentiation conditions presented in this study, leading to rapid down regulation of LHX3 and acquisition of FOXP1 expression. The in vitro system described in this study also provides a unique opportunity to characterize the molecular pathways controlled by SAG and PUR leading to the acquisition of LMC motor neuron identity.
In conclusion, we have established reference expression profiles of principal motor neuron markers in the human developing spinal cord, and relying on these maps we developed an efficient, rapid, and technically simple protocol for differentiation of human pluripotent stem cells into motor neurons acquiring preferentially either LMC or MMC columnar identity, without the use of recombinant proteins or viral factors (Hester et al., 2011; Son et al., 2011). Efficient specification of the LMC motor neuron fate from pluripotent stem cells will provide experimental tools for building more sophisticated in vitro models of neural development, motor circuit function, and more refined transplantation assays. Furthermore, in a disease context, motor neurons exhibit subtype-specific differences in their susceptibility to degeneration. Whereas MMC and HMC are affected predominantly in SMA, non-bulbar onset ALS is typically manifested by initial weakness in distal limb muscles innervated by LMC neurons (Kanning et al., 2010; Theys et al., 1999). Taken together, derivation of cultures in which motor neuron subtype identity can be controlled with great efficiency provides a potential basis for modeling selective disease sensitivity of motor neuron subtypes in vitro and for designing targeted cell-based screens for motor neuron disease therapeutics.
We thank Amy MacDermott for introducing calcium-imaging approaches into our work and Kevin Eggan for open sharing of reagents and ideas throughout this study. We are grateful to F.P. Di Giorgio for the HUES3 HB9 reporter line, Susan Brenner-Morton and Thomas M. Jessell for graciously providing ISL1, HB9, RALDH2 and FOXP1 antibodies and motor neuron expertise, Lee Rubin for the human Smo agonist, David J. Kalher and Mathew Darwin-Zimmer (NYSCF) for help with FACS, R.M. Myers and F. Pauli, for performing RNA sequencing, and the anonymous donors for tissue samples. We benefited from many helpful discussions with members of the Project A.L.S., Wichterle, Henderson and Eggan laboratories, and unwavering support from Valerie and Meredith Estess (Project A.L.S.). Project A.L.S., P2ALS, NIH GO grant 5 RC2 NS069395-02, NYSTEM grant C024415, the Dr. Leigh G. Cascarilla Post-Doctoral Fellowship in Stem Cell Research (L.R). National Institutes of Health (DP1OD003930 to T.M.), the ALS Association (to M.A.C.), and The ALS Therapy Alliance (S.O.) funded this work.
Conflict of interest: The authors declare no conflicts of interest.
Author ContributionsM.W.A., C.E.H, and H.W. conceived the experiments. M.W.A. performed ES and iPS motor neuron assays and optimization experiments. G.F.C. established standard motor neuron differentiation protocol and automated motor neuron quantification methods. G.F.C. and M.W.A. stained and analyzed human tissue sections. H.W. performed xenotransplantation. D.J.W. designed Ca2+ imaging experiments, performed and analyzed patch-clamping assays. S.O., M.A.C., and T.M. processed RNA-seq data. M.W.A. and H.W. analyzed the RNA-seq data. A.R.D. contributed materials. L.R. conceived data not shown on adherent neuralization. D.H.O. contributed preliminary data not shown on MN efficiency with iPSC lines. M.W.A. analyzed all other data. M.W.A., C.E.H. and H.W. wrote the manuscript. All authors provided input and approved the final manuscript.