Modeling Down Syndrome with Patient iPSCs Reveals Cellular and Migration Deficits of GABAergic Neurons

DS GABAergic Interneurons Exhibit Less Complexity in Morphology In Vitro

We first differentiated forebrain GABAergic interneurons from iPSCs derived from two DS patients (DS1 and 2DS3), euploid iPSCs (DS2U and IMR90-4), and euploid H9 hESCs (Figure 1A and Table S1). The euploid iPSCs (DS2U) were cloned from the same DS patient cells as DS1 in which about 10% of the cells exhibit the normal karyotype (Weick et al., 2013). Following our established protocol (Liu et al., 2013a), both the trisomy and euploid iPSCs differentiated to GABAergic neurons with a high efficiency (>80% of TUJ1⁺ neurons at day 35) (Figures 1B and 1C), in the presence of purmorphamine (Pur), a small-molecular sonic hedgehog (SHH) agonist. The remaining cells were positive for tyrosine hydroxylase (TH, <5%) and choline acetyltransferase (ChAT, <1%) (Figure 1C). The results suggest that trisomy iPSCs have a similar potential with euploid iPSCs to generate GABAergic interneurons.

Morphometric analysis indicated that the DS GABAergic interneurons (DS1 and 2DS3) exhibited a reduced soma size in comparison with euploid cells (IMR90-4 and H9) and isogenic euploid control (DS2U) at days 3, 7, and 10 after plating day-28 MGE-like progenitors (Figures 1D and 1E). Measurement of neurite growth revealed that trisomy GABAergic interneurons (DS1 and 2DS3) displayed a shortened neurite length and diminished branch complexity as compared with euploid neurons (DS2U, IMR90-4, and H9) at day 3, 7, and10 after plating (Figure 1E). These cellular deficits were replicated by high-content measurement (Figure 1F) whereby neurons were fixed 7 days after plating and labeled with TUJ1 (Figures 1F and 1G). Thus, trisomy GABAergic interneurons were generated from iPSCs in a similar efficiency as their normal counterparts but displayed less complexity in morphology.

DS GABAergic Interneurons Exhibit Altered Subtypes and Impaired Migration In Vitro

There are numerous types of GABAergic interneurons in the cerebral cortex. They may be subdivided into multiple subtypes according to the expression of calbindin (CB), SST, CR, and parvalbumin (PV) (Kawaguchi and Kubota, 1997, Wonders and Anderson, 2006). Dysfunction of any subtype may be associated with neurological disorders (Lawrence et al., 2010). The altered morphometric properties of DS neurons suggest a possibility of altered subtypes of GABAergic interneurons. Indeed, the percentage of CR⁺ interneurons in DS1 and 2DS3 was 23.1% ± 2.1% and 27.9% ± 1.7% at day 80, respectively, which is significantly lower than H9 (47.7% ± 4.7%) and DS2U (32.4% ± 2.1%) (Figures 2A and 2B). Similarly, the percentage of CB⁺ neurons in DS1 (2.2% ± 0.5%) and 2DS3 (2.0% ± 0.3%) is significantly lower than in H9 (7.4% ± 0.6%) and DS2U (5.6% ± 0.4%) at day 40 (Figures 2C and 2D). The SST⁺ or PV⁺ interneurons were rarely found at day 80 (less than 5 positive cells over one coverslip). Thus, the proportion of CR⁺ and CB⁺ subtype neurons in DS is reduced compared with the euploid neurons.

During development, GABAergic interneurons originate from ganglionic eminences and tangentially migrate to the cortex (Ma et al., 2013). We asked whether the migration of GABAergic interneurons was affected in DS. The cell migration was assayed in two ways. First, GABAergic progenitor clusters with similar diameter (Figure 2E) (about 180 μm) were plated on glass coverslips for 24 hr and the number of cells and the distance of cell migrated away from the sphere were measured. There was a significant decrease in the number of cells and distance that cells migrated out of the sphere relative to euploid control (Figure 2F). Second, we used the scratch assay (Fattahi et al., 2016) and found that DS1 and 2DS3 GABAergic interneurons showed a notable migration deficit at day 3 after scratching (Figure 2G). There were 442 ± 63 cells that migrated to the scratched area after 3 days in H9, and 330 ± 80 cells in DS2U, compared with 150 ± 40 cells in DS1 and 160 ± 35 cells in 2DS3 (Figure 2H). Therefore, DS GABAergic interneurons exhibit decreased migration in vitro.

Both Trisomy and Euploid Progenitors Differentiate to GABAergic Interneurons in the Mouse Brain

To determine whether the cellular phenotypes observed in vitro are intrinsic to DS GABAergic interneurons, we transplanted 50,000 7-week-old GABAergic progenitors, which were generated from trisomy and euploid control, into the medial septum (Figure 3A) in SCID mice (9 for DS1, 6 for 2DS3, 6 for H9, and 8 for DS2U). Transplanted human neural progenitors usually mature and form synaptic connections after 4–6 months (Liu et al., 2013b, Weick et al., 2011). When the grafts were analyzed by stereology 6 months after transplantation, we found that around 75,000 human nuclei (HN)-positive cells in the medial septum, and no obvious difference was discerned between the brains transplanted with trisomy and euploid cells (Figures 3B and 3C), suggesting that trisomy and euploid GABAergic progenitors survive in the brain in a similar manner.

Analysis of the grafted cells indicated that around 1% of the human cells were positive for NESTIN (Figure S1A) and hardly any were positive for Ki67 (Figure S1B), suggesting that the vast majority of the grafted cells become postmitotic. Indeed, 88.43% ± 5.34% of DS cells and 86.59% ± 2.64% of euploid cells expressed the neuronal marker TUJ1 (Figures S1D and S1E), 7.78% ± 5.48% of DS cells and 7.96% ± 0.91% of euploid cells were positive for an astrocyte marker glial fibrillary acidic protein (GFAP) (Figures S1D and S1E), and few were positive for an oligodendrocyte marker myelin basic protein (Figures S1C and S1E). In addition, about 40% HN⁺ cells expressed doublecortin (DCX) in both groups (Figures S1D and S1E). These results indicate that the majority of cells become postmitotic neurons and that some of them are migrating immature neurons.

Among the human TUJ1⁺ neurons in the medial septum, 77.34% ± 2.51% of the cells in DS and 78.35% ± 1.34% of those in euploid were positive for GABA, <2% for choline acetyltransferase (ChAT), and few (<1%) for TH (Figures S1F and S1G). No obvious difference in cell-type-specific marker expression was observed between DS and euploid cell transplant groups. Together, these results indicate that the transplanted trisomy and euploid cells generate GABAergic interneurons similarly in vivo.

DS GABAergic Interneurons Are Smaller with Shorter Neurites and Fewer Branches

We then asked whether the DS cells exhibit a similar phenotype in vivo. Morphometric analysis showed a wide range of soma size for the grafted GABAergic neurons. The average soma area for DS GABAergic interneurons was 38.76 ± 1.09 μm², as compared with 52.18 ± 2.15 μm² for euploid cells (Figures 3D and 3E). More DS GABAergic neurons were smaller than 30 μm² than euploid cells. Less than 5% of the DS GABAergic neurons were larger than 90 μm² compared with 30% for euploid cells (Figure 3E). Thus, more DS GABAergic neurons are in the range of smaller cell size.

Measurement of neurite branching revealed that DS GABAergic interneurons possessed fewer primary and secondary branches (Figures 3D and 3E), indicating reduced neurite arborization in DS neurons. Furthermore, neurite length, indicated by the longest constant neurites, was shorter in DS neurons compared with that in euploid cells (Figure 3E). These results were consistent with the in vitro properties.

Morphologically, the grafted neurons exhibited various shapes based on their soma size, neurite length, and branches, and we arbitrarily classified them into five categories (Figures 3F and 3G). The majority cells were type-A neurons, which exhibit short neurites and few branches. The DS groups had significantly more type-A neurons than the euploid group. In contrast, DS cells had fewer type-E neurons, those with rich branches and long neurites. Thus, the DS GABAergic interneurons are smaller with shorter neurites and fewer branches.

DS Progenitors Differentiate to More SST but Fewer CR GABAergic Neurons

Our in vitro analysis revealed altered subtypes in DS GABAergic neurons. We then asked whether this also happens in vivo. Immunostaining showed that the main types of GABAergic neurons were those positive for CR, CB, and SST (Figures 3H and 3J). Very few (<1%) were positive for PV, which were mostly on the edge or outside of the grafts (Figures 3I and 3J). The percentage and the soma size of CB⁺ neurons in DS were similar to those in euploid (Figures 3H, 3J, and 3K). However, the percentage of SST⁺ neurons increased but that of CR⁺ neurons decreased in DS (Figures 3H and 3J). Moreover, the soma size of SST⁺ and CR⁺ neurons in DS decreased compared with that in euploid (Figures 3L and 3M). Thus, the DS progenitors tend to generate more SST⁺ but fewer CR⁺ GABAergic interneurons at 6 months after transplantation.

DS GABAergic Neurons Exhibit Impaired Migration and Neurite Projection

GABAergic neurons in the medial septum migrate to the olfactory bulb by the rostral migratory stream and project primarily to the hippocampus via the septohippocampal pathway (Bianchi et al., 2014, Merkle et al., 2017, Vega-Flores et al., 2014, Qin et al., 2017). We therefore asked whether human GABAergic interneurons follow the same pathways for migration/projection and whether the DS and euploid cells follow the same pattern. Immunostaining of sagittal brain sections with a human specific antibody, Stem121, showed that significantly fewer DS cells (DS1 and 2DS3) than euploid cells (DS2U and H9) were present in the olfactory bulb (Figures 4A and 4B). The distance of migration to olfactory bulb in euploid was up to 4.22 ± 0.43 mm, while in DS it was up to 2.38 ± 0.21 mm (Figure 4C).

The majority of the human cells in the olfactory bulb were DCX-positive cells (Figure S2A), reflecting the immature nature of the human cells. Some of the human cells in olfactory bulb were GFAP positive (Figure S2B). Few grafted cells were found in the cortex with the exception of areas around the needle track in both groups. Thus, the human GABAergic interneurons follow the endogenous pathways, albeit with differential migration capacities.

Besides the olfactory bulb, we also observed human cells in the hippocampus. Significantly fewer cells were present in the hippocampus from the DS groups (DS1 and 2DS3) than from the euploid groups (DS2U and H9) (Figures 4D and 4E). The distance of the DS GABAergic neurons migrated from the injection site to hippocampus was 1.58 ± 0.31 mm, whereas that for euploid GABAergic neurons was 2.73 ± 0.13 mm (Figure 4F). In both DS and euploid groups, most of those individual (migrating) cells in the hippocampus were GABA⁺ neurons, and less than 5% were GFAP⁺ (Figure S2C). Among the GABA⁺ neurons, very few were positive for SST, CR, CB, and PV. Instead, they were largely positive for DCX (Figure S2C), suggesting that they remain immature by 6 months post transplantation.

We also found that euploid cells presented a long projection trail along the septohippocampal pathway to the posterior part of hippocampus (Figures 4D and 5A). In contrast, fewer DS neurites were present in the hippocampus (Figures 5A and 5B). Quantitative measurement indicated a 50.02% reduction in neurite length from DS neurons compared with euploid cells (Figure 5C). Projection area by DS neurites was also significantly smaller than that by euploid neurites (Figure 5D). The reduced neurite projection was also demonstrated by counting the number of neurites that cross the two artificial lines according to the hippocampus structure (Figures 5B and 5E). In the euploid group, 28 ± 4.96 neurites crossed line 1, which is in the anterior hippocampus, and 12 ± 1.83 neurites traversed line 2 which is in the posterior hippocampus (Figure 5D). In contrast, approximately 10 ± 1.64 neurites crossed line 1 and only 4 ± 1.03 neurites traversed line 2 in the DS group (Figure 5E). These Stem121⁺ neurites coexpressed GABA (Figure 5F), indicating their GABAergic neuron identity. In addition, most of the GABA⁺ neurites were DCX⁺ with some positive for CR but not CB, SST, or PV (Figures S3A and S3B), suggesting that the majority of the neurites are from immature GABAergic neurons. Thus, DS GABAergic neurons exhibit a decreased axonal projection to the hippocampus.

RNA Profiling Reveals Altered Gene Expression in DS GABAergic Interneurons

To gain insights into the molecular underpinning of DS GABAergic interneuron behaviors, we performed RNA sequencing (RNA-seq) using day-45 GABA interneurons that were differentiated from three independent Down syndrome patients (DS1, 2DS3, and DSP) and three euploid iPSCs (DS2U, IMR90-4, and H9). Of all 21,720 genes profiled, we found that 119 genes were significantly upregulated while 260 were downregulated in trisomic GABAergic interneurons compared with euploid control by using Cufflinks software (Figure 6A and Table S3), with a false discovery rate (FDR) ≤ 0.05 and |log₂(fold change)| ≥ 1.

About 224 terms were significantly altered in DS GABAergic interneurons compared with the controls (p ≤ 0.05) by using gene ontology biological process (Table S4). Development-related terms were highly enriched in DS GABAergic interneurons, including system development, cell differentiation, and nervous system development, especially those relating to cell morphogenesis (Figure 6B). Lymphoid enhancer binding factor 1 (LEF1) and DCC netrin 1 receptor (DCC) were increased while lumican (LUM) and fibroblast growth factor receptor 2 (FGFR2) were decreased (Figure 6C and Table S5). Cell differentiation-related genes were also altered in DS GABAergic interneurons, such as achaete-scute family basic helix-loop-helix transcription factor 1 (ASCL1), Wnt family member 5A (WNT5A), nerve growth factor receptor (NGFR), and GLI family zinc finger 3 (GLI3) that are involved in basal forebrain development (Figure 6D and Table S6). This gene expression profile is consistent with our finding of impaired interneuron subtype differentiation in DS.

Pathway analysis using the DAVID Bioinformatics Resource revealed that146 pathways were significantly changed in DS GABAergic interneuron in comparison with controls (Table S7). Among them, cell motility-related pathways were prominent, including extracellular matrix (ECM)-receptor interaction, focal adhesion, hedgehog signaling pathways, cell adhesion molecules, and regulation of actin cytoskeleton (Figure 6E). Notably, the regulation of the actin cytoskeleton pathway is critical for neural migration (Solecki et al., 2009), hinting that the altered actin cytoskeleton pathway may play an important role in the defective migratory ability in DS. Protein-protein interaction network analysis revealed 212 genes that were involved in cytoskeleton pathway, including Rac family small guanosine triphosphatase 1 (Rac1), cell division cycle 42 (Cdc42), alkaline phosphatase (PhoA), and p21 activated kinase 1 (PAK1), which may contribute to the impaired migration of DS GABAergic interneurons (Figure 6F).

Correction of PAK1 Expression Rescues Migration Deficit of DS GABAergic Neurons

Our RNA-seq, confirmed by qPCR analysis, showed that the expression of PAK1 was increased in DS cells (Figures 7A and 7B). Western blotting revealed a substantial increase in PAK1 protein (Figure 7C). Its downstream gene cofilin was not altered. However, the phosphorylated cofilin (p-cofilin) was substantially increased in DS GABAergic neurons that were derived from DS1, 2DS3, and DSP iPSC lines compared with euploid controls (DS2U, IMR90-4, H9) (Figure 7D). PAK1 is important for neural migration during cortical genesis in mice (Causeret et al., 2008). Its downstream protein, cofilin, is an actin-binding protein that modulates cytoskeleton. We hypothesized that altered expression of PAK1 and cofilin is associated with the cell migration deficit seen in vitro and in vivo. The DS GABAergic progenitors were treated with FRAX486, an inhibitor of PAK1 with excellent kinase inhibition (Dolan et al., 2013), for 5 days. While the content of PAK1 was similar in the control and treated cells (Figure 7E), the expression of phosphorylated PAK1 (p-PAK1) and p-cofilin in DS GABAergic neurons was decreased after application of FRAX (Figures 7E–7G). Importantly, we found that there was a significant increase in migration distance of DS GABAergic interneurons with the treatment of FRAX486 (Figures 7H and 7I). These results suggest that an altered PAK1 pathway is at least partly responsible for DS interneuron migration defects.

iPSC Culture and GABAergic Interneuron Differentiation

iPSC lines were generated from two Down syndrome patients as reported by Weick et al. (2013). DS iPSC lines were trisomy 21 (DS1, 2DS3, and DSP) and euploid iPSC line (DS2U, the isogenic control of DS1 with identical background) (Weick et al., 2013), and wild-type iPSC line (IMR90-4) and hESC line (H9) were used as additional controls. DSP was trisomy 21 and generated from a separate mosaic DS individual. For pluripotency maintenance, iPSCs/hESCs were maintained on a feeder layer of irradiated embryonic mouse fibroblast as previously described (Liu et al., 2013b).

For neural differentiation, iPSCs were detached with Dispase (1 U/mL, Gibco) to form embryoid bodies (EBs) in suspension in neural induction medium (NIM) consisting of DMEM/F12 (Gibco), N2 supplement (1:100; Gibco), and non-essential amino acids (NEAA) (1:100; Gibco) for 7 days. EBs were attached on day 7, and formed neural tube-like rosettes at days 10–16. To induce MGE-like progenitors, we treated the neuroepithelial cells with Pur (1.5 μM) or together with SHH at days 10–25 as previously described (Liu et al., 2013a). Rosette colonies were detached by gently blowing the neural colonies with a pipette and suspended to form neurospheres on day 16. At day 24, neurospheres were dissociated to single cells with TrypLE and plated on polyornithine-coated coverslips for relevant experiments.

Immunostaining

Neurons were fixed with 4% paraformaldehyde for 30 min. Neurons were washed with PBS for 5 min three times, permeabilized in 0.2% Triton X-100 for 10 min, and blocked for 1 hr in 10% donkey serum before being incubated in the primary antibody in 5% serum and 0.1% Triton X-100 at 4°C overnight. Cells were subsequently washed and stained with Alexa Fluor-conjugated secondary antibodies and Hoechst in 5% donkey serum for 1 hr before being washed and mounted onto glass slides with Fluoromount-G mounting solution (SouthernBiotech). The primary antibodies are listed in Table S2. Images were taken using a Nikon 80i fluorescence microscope (Nikon Instruments).

High-Content Measurement

GABAergic progenitors (day 35) were dissociated and plated at a density of 2,000 neurons per well in 96-well flat-bottom plates that were pre-coated with Matrigel. Neurons were cultured with NIM as previously described. After 7 days of attachment, the neurons were immunostained for TUJ1. For high-content assays, image acquisition and analysis were performed using the Arrayscan VTI (Thermoscientifics, USA). The nuclei were detected based on the Hoechst staining, and soma size was detected based on TUJ1. A minimum of 500 cells per well were analyzed at a 20× objective, and at least five wells were analyzed for each iPSC-derived neuronal group. To quantify neurite outgrowth, we used the “neuronal profiling bioapplication” to track TUJ1-positive neurites. Each value was the average value of one well.

GABAergic Progenitor Migration Assay

Before assay, neurospheres were triturated by Pasteur pipette (Liu et al., 2013a). The narrow opening and bend in the pipette help to shear the clusters into smaller clusters with a roughly uniform size. Similar sizes of neurospheres were selected for the migration assay. Neurospheres were transferred carefully and plated dispersedly onto Matrigel-coated coverslips. After 24 hr, neurons migrated out from neurospheres and were assessed under a phase-contrast microscope (Nikon, TS100).

Scratch Assay

Neurospheres were dissociated into single cells and plated onto Matrigel-coated 96-well culture plates (450,000/cm²) in a neural differentiation medium containing DMEM/F12, N2 (1:100), NEAA (1:100), and B27 (1:50). After 3 days, the adherent culture was scratched using a 200-μL pipette tip. The neurons were fixed for imaging and migration analysis after 3 days. The scratched area was defined immediately after scratching. The cells that were present on the scratched area after a 3-day migration were counted.

Cell Transplantation

The animal experiments were carried out following the protocols approved by the Animal Care and Use Committee at the University of Wisconsin-Madison and Nanjing Medical University. The mice, aged 8–10 weeks, were anesthetized with 1% isoflurane mixed with oxygen. GABAergic progenitors were differentiated from DS1, 2DS3, DS2U, and H9 human PSCs for 7 weeks before transplantation. A week before transplantation, a small portion of GABAergic progenitors was dissociated and plated onto coverslips for the maturity and purity analysis. If more than 90% of the cells were NKX2.1-positive at day 30, and more than 50% of the cells were GABA-positive around day 35, the progenitors were “good” for transplantation. GABAergic progenitor clusters, around 30 μm of diameter, were injected into the medial septum with a glass pipette at a 15° angle toward the midline at the following stereotactic coordinates: anterior-posterior = +0.38 mm; left lateral = +1 mm; and dorsoventral = −4.12 mm. About 50,000 cells were injected into medial septum in 1 μL of artificial cerebrospinal fluid over a 5-min period.

Tissue Preparation and Immunohistochemistry

At 6 months after transplantation, animals were perfused transcardially with 4% paraformaldehyde. The brain was serially sectioned coronally and sagittally to 35 μm in thickness and subsequently stored in the preservation solution at −20°C.

Immunohistochemistry was performed as previously described (Liu et al., 2013b). Sections were washed with PBS for 5 min three times, and permeabilized and blocked for 1 hr in 10% donkey serum and 0.2% Triton X-100 before being incubated in the primary antibody in 5% serum and 0.2% Triton X-100 at 4°C overnight. Sections were subsequently washed and stained with Alexa Fluor (Life) secondary antibodies and Hoechst in 5% donkey serum for 1 hr before being washed and mounted onto glass slides with Fluoromount-G mounting solution (SouthernBiotech). The primary antibodies used in this study are listed in Table S2. Images were visualized using a Nikon 80i fluorescence microscope (Nikon Instruments) and a Zeiss confocal microscope (LSM700, Zeiss instruments).

Real-Time PCR

Total RNA was isolated using the TRIzol kit (Invitrogen, USA) according to the manufacturer's manual. One microgram of total RNA from each sample was reverse transcribed into cDNA and subjected to real-time PCR using SuperScript III First-Strand (Invitrogen). Primers for real-time PCR were: PAK1 forward primer CAG CCC CTC CGA TGA GAA ATA; PAK1 reverse primer CAA AAC CGA CAT GAA TTG TGT GT.

Western Blotting

For analysis of cofilin and p-cofilin protein, neurons were lysed in RIPA buffer with protease and protease inhibitor cocktail (Roche). Proteins was separated by SDS-PAGE and transferred to polyvinylidene fluoride membranes. Anti-PAK1 (1:500 dilution, #SC-166887, Santa Cruz Biotechnology), anti-Phospho-PAK1 (1:500 dilution, #2605P, Cell Signaling Technology), and anti-Phospho-cofilin (1:1,000 dilution, #3311, Cell Signaling) antibodies were used. Western blotting was performed using horseradish peroxidase-conjugated immunoglobulin G as a secondary antibody and the ECL system for detection.

Data Analysis

Data are shown as mean ± SEM. Statistical comparisons between two groups were performed by Student's t test. Statistical analyses were performed using one-way ANOVA (with Tukey’s or Dunnett’s test for multiple comparisons). p Values of <0.05 were considered statistically significant.