Gene expression analysis of human induced pluripotent stem cell-derived neurons carrying copy number variants of chromosome 15q11-q13.1

Idic(15) fibroblasts (catalog ID: GM07992) were obtained from the Coriell Institute for Medical Research Cell Repository. Idic(15) umbilical cord blood cells were donated by the patient’s family through the Dup15q Alliance and were exempted from consideration as human subject research by the University of Connecticut Health Center Institutional Review Board (IRB). Fibroblasts from an individual with inherited paternal interstitial 15q11-q13.1 duplication (patient 801-015) and an individual mosaic for maternal interstitial 15q11-q13.1 duplication (mother of patient 801-018) were obtained by Dr. Lawrence T. Reiter under IRB approval number 11-01350-FB from the University of Tennessee Health Science Center. All patient samples were obtained from subjects after they had given informed consent and were subsequently de-identified.

iPSCs were generated with Institutional Biosafety Committee approval number IBC08-005 from the University of Connecticut Health Center by the University of Connecticut - Wesleyan University Stem Cell Core as previously reported [18]. iPSCs were maintained on irradiated mouse embryonic fibroblasts (MEFs) in human embryonic stem cell medium which consists of DMEM/F12, 20% knockout serum replacement, 1 mM L-glutamine, 1X nonessential amino acids, 100 mM β-mercaptoethanol (all Gibco products through Life Technologies, Grand Island, NY, USA), and 4 ng/mL basic fibroblast growth factor (bFGF, Millipore, Billerica, MA, USA). iPSCs were manually passaged every 6 or 7 days.

Cytogenetic analysis of Dup15q iPSCs was performed by the Genetics and Genomics Division of the University of Connecticut - Wesleyan University Stem Cell Core. Twenty G-banded metaphase cells from each iPSC line were examined to generate a karyotype for each line. DNA fluorescence in situ hybridization (FISH) was performed on both metaphase and interphase cells using a dual-labeled probe containing the small nuclear ribonucleoprotein polypeptide N (SNRPN) gene and a control locus at 15qter (Cytocell Aquarius LPU005-A-034359, Cytocell, Cambridge, UK). Whole genome copy number analysis was performed on genomic DNA isolated from AS del 1-0 and Idic1-8 iPSCs using the Affymetrix CytoScan HD Array (Affymetrix, Santa Clara, CA, USA) to determine deletion/duplication breakpoints. Duplication breakpoints for the maternal and paternal Int dup(15) patient samples were previously published (as patients 801-018 and 801-015, respectively) [8] and for IdicCB were provided in an array CGH report accompanying the patient sample.

iPSC-derived neural progenitors were generated by either embryoid body (EB)-based or monolayer differentiation according to established protocols [19, 20] with minor modifications. For EB differentiation, iPSC colonies were manually detached from MEF feeders and put into suspension culture rather than enzymatic dissociation. After three weeks of neural differentiation (in both protocols), neural progenitors were plated on poly-ornithine/laminin coated substrates in neural differentiation medium consisting of Neurobasal Medium, B-27 supplement, nonessential amino acids, and L-glutamine (all Gibco products through Life Technologies, Grand Island, NY, USA) supplemented with 1 μM ascorbic acid, 200 μM cyclic adenosine monophosphate (cAMP), 10 ng/mL brain-derived neurotrophic factor (BDNF, Peprotech, Rocky Hill, NJ, USA), and 10 ng/mL glial-derived neurotrophic factor (GDNF, Peprotech, Rocky Hill, NJ, USA). All experiments were conducted on neural cultures that were at least 10 weeks old.

For mithramycin experiments, mithramycin A (Sigma-Aldrich, St. Louis, MO, USA) was prepared as a 1 mM concentrated stock in dimethyl sulfoxide (DMSO, Sigma-Aldrich, St. Louis, MO, USA). Working stocks were diluted in DMSO and added to neural differentiation medium immediately before addition to neural cultures.

Immunocytochemistry was performed as previously described [18] on iPSCs or 10-week old iPSC-derived neurons that had been cultured on glass coverslips. The following antibodies and concentrations were used: mouse anti-Tra-1-60 (1:200, Santa Cruz Biotechnology, Inc, Dallas, TX, USA), Nanog (1:200, Abcam, Cambridge, UK), mouse anti-SSEA-4 (1:20, Developmental Studies Hybridoma Bank, Iowa City, IA, USA), chicken anti-MAP2 (1:10,000, Abcam, Cambridge, UK), rabbit anti-MAP2 (1:500, Millipore, Billerica, MA, USA), rabbit anti-Synapsin I (1:400, Millipore, Billerica, MA, USA), mouse anti-PSD-95 (1:100, NeuroMab, Davis, CA, USA), rabbit anti-S100β (1:200, Abcam, Cambridge, UK), mouse anti-VGlut1 (1:100, Synaptic Systems, Gottingen, Germany), and rabbit anti-Gad65 (1:500, Sigma-Aldrich, St. Loius, MO, USA). All AlexaFluor fluorochrome conjugated (488, 594, and 647) secondary antibodies (Life Technologies, Grand Island, NY, USA) were used at 1:400. A goat anti-chicken IgY-650 secondary antibody (Abcam, Cambridge, UK) was used at 1:250. Nuclei were counterstained with DAPI and coverslips were mounted on slides with Vectashield (Vector Laboratories, Burlingame, CA, USA). Slides were imaged using a Zeiss Axiovision microscope (Carl Zeiss, Germany) at 20X, 40X, and 63X magnification.

Chromatin immunoprecipitation (ChIP) was performed according to the EZ-Magna ChIP protocol (Millipore, Billerica, MA, USA) according to manufacturer’s instructions with minor modifications. DMSO or mithramycin-treated 10-week old idic(15) neurons were treated with 1% formaldehyde to crosslink DNA/protein. Instead of a two-step cell lysis, cells were lysed once for 15 minutes in an SDS Cell Lysis Buffer (Millipore, Billerica, MA, USA) before sonication. A rabbit polyclonal anti-YY-1 (sc-281, Santa Cruz Biotechnology, Inc, Dallas, TX, USA) was used at 5 μg per immunoprecipitation reaction. For ChIP analysis of Sp1 binding, the following antibodies were used: mouse monoclonal anti-Sp1 (sc-17824, Santa Cruz Biotechnology, Inc, Dallas, TX, USA) and anti-Sp1 (#9389, Cell Signaling Technology, Beverly, MA, USA). Novex Protein A DynaBeads magnetic beads (Life Technologies, Grand Island, NY, USA) were used during immunoprecipitation. Immunoprecipitated DNA was purified by phenol-chloroform extraction and used for quantitative polymerase chain reaction (qPCR) with SYBR-Green reagents (Life Technologies, Grand Island, NY, USA). The following primers were used for qPCR: UBE3A-exon1 forward GGC AGA GGT GAA GCG TAA GT, UBE3A-exon1 reverse AGA TCC GTG TGT CTC CCA AG, UBE3A-upstream forward TCT GTG ACC CGA AAG AAT AAA CC, UBE3A-upstream reverse TTC CTC TGC TGG GTA CAC CAA. Sp1 promoter forward TGC CCG CCT GAT TTC TGA, Sp1 promoter reverse GGA TAT GCT TGG GCA AAA TCC, DHFR promoter forward TCG CCT GCA CAA ATA GGG AC, DHFR promoter reverse AGA ACG CGC GGT CAA GTT TG. ChIP was performed with triplicate independent batches of neurons. qPCR was performed in triplicate for each DNA sample and Ct values were used to calculate percent of input. Percent of input values for YY-1 binding were normalized by subtracting the background binding of normal rabbit IgG (Millipore, Billerica, MA, USA). Fold enrichment was then calculated by dividing percent input in mithramycin-treated samples by percent input of DMSO-treated controls.

Total RNA was isolated from iPSCs, EBs, or iPSC-derived neurons using RNA-Bee (AMS Biotechnology, Lake Forest, CA, USA) according to the manufacturer’s protocol. cDNA was produced using the High Capacity cDNA Reverse Transcription Kit (Life Technologies, Grand Island, NY, USA).

Analysis of iPSC pluripotency genes was performed on independent cultures of each iPSC line in triplicate with a TaqMan Human Stem Cell Pluripotency Array (Life Technologies, Grand Island, NY, USA). Analysis of multipotency and neural differentiation capacity was performed with a custom TaqMan Gene Signature Array Card (Life Technologies, Grand Island, NY, USA) as previously described [21] on three independent batches of EBs derived from each iPSC line. This custom array includes representative genes from all three germ lineages as well as pluripotency genes. A full list of gene assays included in the custom array is available in Martins-Taylor et al. [21]. Ct values for all genes were first normalized to 18S rRNA and then set relative to the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH).

Analysis of 15q11-q13.1 genes and selected autism candidate genes in iPSCs and iPSC-derived neurons was performed in duplicate or triplicate from independent cultures. All qPCR assays used were TaqMan Gene Expression Assays (Life Technologies, Grand Island, NY, USA). Ct values for each gene were normalized to the house keeping gene GAPDH. Relative expression was quantified as 2^^-ΔΔCt using normal (Nml 1-0) iPSCs or normal iPSC-derived neurons as the calibrator sample.

Analysis of methylation at the Prader-Willi syndrome imprinting center (PWS-IC) was performed using the Methyl-Profiler DNA Methylation qPCR Assay kit (SA Biosciences, Valencia, CA, USA) and the EpiTect Methyl qPCR Assay for SNRPN (Qiagen, Valencia, CA, USA) as previously described [18].

The UCSC Genome Browser (http://​genome.​ucsc.​edu/​) was used to identify a single nucleotide polymorphism (SNP), rs691, which is located in the last exon of the imprinted in Prader- Willi syndrome (IPW) gene. Genomic DNA from mat. int dup(15), pat. int dup(15), and idic(15) fibroblasts, and IdicCB-09 iPSCs was used to determine heterozygosity for rs691 by PCR amplification across the SNP followed by sequencing by GENEWIZ. Total RNA was isolated from fibroblasts (mat. int dup(15) and idic(15)), iPSCs, and iPSC-derived neurons (mat. int dup(15)-02, pat. int dup(15)-04, Idic1-8, and IdicCB-09) using RNA-Bee (AMS Biotechnology, Lake Forest, CA, USA) and DNase treated using Turbo DNA-free Kit (Life Technologies, Grand Island, NY, USA). cDNA was generated using the High Capacity cDNA Reverse Transcription Kit (Life Technologies, Grand Island, NY, USA) and used for PCR amplification across rs691 followed by sequencing by GENEWIZ. The following primers were used to amplify across rs691: forward ATG CCC TCC TCT CTT CCA AT and reverse ATA GGG AGG TTC ATT GCA CA.

Individual coverslips were transferred to a recording chamber (room temperature) fixed to the stage of an Olympus BX51WI microscope (Olympus, Tokyo, Japan) fitted with a 40x water-immersion lens. The recording chamber was continuously perfused at 2 ml/min with oxygenated artificial cerebrospinal fluid (125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 1 mM MgCl2-6H2O, 25 mM NaHCO3, 2 mM CaCl2, and 25 mM dextrose). Cells were selected for recording based on neuronal morphology. Whole-cell voltage clamp (holding potential = -70 mV) and current clamp recordings were conducted as previously described [18, 22]. Series resistance (Rs) was compensated to 70% or greater at 10 to 100 μs lag. During the course of the experiments, input resistance (Ri) was continuously monitored with 5 mV hyperpolarizing voltage steps (50 ms). Neurons were rejected from analyses if (1) Rs was >25 MOhms at the time of break-in, if (2) Ri changed by >15% during the course of an experiment, or if (3) Ri fell below 100 MOhms.

Total RNA was isolated from two biological replicates of 10-week-old neurons derived from AS (AS del 1- 0), normal (Nml 1- 0), and idic(15) (Idic1-8) iPSCs. mRNA-Seq libraries were prepared using 5 μg of total RNA according to manufacturer’s specifications (Illumina, San Diego, CA, USA) using the Paired-end Library kit as described in [23]. Libraries were multiplexed and sequenced on Illumina GAIIx and HiSeq 2000 sequencers (Illumina, San Diego, CA, USA). The reads were aligned to the human genome (hg19) using Bowtie (version 0.12.0-beta1) and Tophat (version 1.3.1) [24]. At least 20 million mapped reads were generated for each sample. Cufflinks (version 0.9.3) was used to quantitate expression levels for all hg19 UCSC genes and transcripts as fragments per kilobase gene model per million base pairs (FPKM) [25]. All raw sequencing data has been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (http://​www.​ncbi.​nlm.​nih.​gov/​sra) under the accession number SRP044749.

The copy number of UBE3A in each iPSC line was analyzed by qPCR using genomic DNA purified from two biological replicates and TaqMan Copy Number Assays (Hs01665678_cn and Hs03908756_cn, Life Technologies, Grand Island, NY, USA). The RNase P Copy Number Reference Assay was used as an endogenous reference gene to allow for quantification of UBE3A copy number. Data were analyzed using the CopyCaller v2.0 software from Applied Biosystems (Life Technologies, Grand Island, NY, USA).

iPSCs grown on glass coverslips were processed for RNA-FISH as previously described [21] with the following modifications: UBE3A BAC probe RP11-1081A4 was labeled by nick translation and 500 ng of labeled probe was used for hybridization.

Electrophysiology data for current-voltage relationship and average frequency and amplitude of synaptic currents are presented as the mean plus or minus the standard error of the mean. All qPCR data were analyzed using Microsoft Excel and are presented as the mean relative expression plus or minus the standard error of the mean. Differential gene expression was analyzed for statistical significance using either one-way ANOVA followed by Tukey’s multiple comparisons test or an unpaired two-tailed t-test. ChIP data are presented as the mean fold enrichment plus or minus the standard error of the mean. Differences in fold enrichment were analyzed for statistical significance using an unpaired two-tailed t-test.

iPSC lines were established using fibroblasts from one idic(15) individual, one individual with a paternally-inherited duplication of chromosome 15q11-q13.1, and one individual who was mosaic for a maternally-inherited interstitial duplication of chromosome 15q11-q13.1 (Figure 1a). iPSC lines were also established from a cord blood sample from one individual with idic(15) (Figure 1a). iPSCs were generated using either retroviral, lentiviral, or episomal vectors encoding POU class 5 homeobox 1 (OCT4), SRY-box 2 (SOX2), Kruppel-like factor 4 (KLF4), v-myc avian myelocytomatosis viral oncogene homolog (MYC), and lin28 homolog A (LIN28) [see Additional file 1: Table S1] [26‐28]. Reprogrammed colonies were initially identified morphologically, and were subsequently validated using qRT-PCR and/or immunocytochemistry to verify the expression of the pluripotency markers, NANOG, stage specific embryonic antigen 4 (SSEA4), and TRA1-60 (Figure 1b-d) and pluripotency genes [see Additional file 2: Figure S1A]. iPSCs were shown to have the expected karyotypes of 46, XX (normal); 46,XX.ish dup(15)(q11.2q11.2)(SNRPN++); or 47,XX,+idic(15).ish15q12 SNRPN x 4, 15qter X2 (Figure 1e). Interstitial duplication of 15q11-q13.1 or the presence of the isodicentric chromosome 15 was confirmed in representative iPSC clones by DNA FISH with a probe for the SNRPN gene and a control probe in the distal long arm of chromosome 15 (Figure 1f and Figure 1g). Breakpoints involved in the duplications in the two int dup(15) samples were previously identified by array comparative genomic hybridization (CGH) [8]. Breakpoints in the cord blood idic(15) sample were determined by genetic report obtained during diagnosis, and the breakpoints in the fibroblast idic(15) and AS samples were determined using a high density SNP array as part of the current study. Gene expression arrays were used to analyze gene expression in iPSC-derived EBs after 16 days of spontaneous differentiation as previously described [18, 21]. These data demonstrate that all iPSCs used are capable of multi-lineage differentiation, since early lineage markers representing each of the three embryonic germ layers as well as the trophectoderm layer are expressed [see Additional file 2: Figure S1B].

We previously demonstrated that the methylation imprint at the PWS-IC was maintained during reprogramming in AS, PWS and normal iPSCs [18]. We assessed the methylation imprint at the PWS-IC in normal and Dup15q iPSCs using methylation-specific qPCR [29, 30]. We expected that iPSCs from idic(15) iPSCs would have approximately 75% methylation due to the presence of 3 maternal and 1 paternal alleles. We expected that paternal int dup(15) iPSCs would have approximately 33% methylation resulting from one maternal and two paternal alleles. We expected that some of the iPSCs from the individual mosaic for a maternal int dup(15) would arise from cells with the duplication and thus would have 66% methylation resulting from 2 maternal and one paternal alleles, while others would be derived from the normal cells and have 50% methylation. Consistent with this prediction, idic(15) iPSCs had methylation levels ranging from 71% to 84% and paternal int dup(15) iPSCs had methylation levels ranging from 23% to 27% (Table 1, Additional file 3: Table S2). Two iPSC lines from the mosaic int dup(15) individual had methylation levels of 54% and 57% [see Additional file 3: Table S2]. One line showed normal diploid cells by karyotype analysis (mat. int dup(15)-12), while the other was a mixed culture, having both duplication and normal cells (mat. int dup(15)-18). The remaining iPSC clones from the mosaic individual had methylation levels ranging from 65% to 68% (Table 1). Thus, DNA methylation at the PWS-IC is consistent with the predicted copy number variations of the Dup15q cell lines and seems to be maintained during reprogramming.

To further investigate the maintenance of parental imprinting following reprogramming of patient cells into iPSCs and differentiation of iPSCs into neurons, we performed allele-specific RT-PCR for the paternally expressed imprinted in Prader- Willi syndrome (IPW) gene. We first identified a SNP (rs691) in the last exon of IPW and confirmed its heterozygosity in genomic DNA from the mat. int dup(15), pat. int dup(15), and idic(15) patient fibroblasts [see Additional file 4: Figure S2]. We were unable to investigate the SNP in genomic DNA from the umbilical cord blood idic(15) starting cell population; however, we did confirm heterozygosity in genomic DNA from the iPSCs derived from these cells. We then generated cDNA from the patient fibroblasts, iPSCs, and neural derivatives and, by sequencing across the SNP in IPW, confirmed that IPW is expressed monoallelically and from the same allele as in the patient cell sample [see Additional file 4: Figure S2]. Together with the DNA methylation pattern at the PWS-IC, this data suggests that parental imprinting at the 15q11-q13.1 locus is maintained following reprogramming to iPSCs.

To prioritize individual iPSC lines based on their ability to differentiate into neurons, all clones of the Dup15q iPSCs from each genotype were allowed to spontaneously differentiate in EB culture and were subjected to qRT-PCR using a custom PCR array containing a panel of neural genes and pluripotency genes [21]. Expression levels of all genes in the array were analyzed relative to those of EBs derived from AS iPSCs, which we previously reported to be capable of generating functional neurons [see Additional file 5: Figure S3A-D] [18]. Representative iPSC clones of each genotype were selected for future experiments based on low levels of expression of the pluripotency genes NANOG and zinc finger protein 42 (ZFP42) and expression levels of neural lineage genes, such as paired box 6 (PAX6), microtubule associated protein 2 (MAP2), and neural cell adhesion molecule 1 (NCAM1) that were similar to or higher than those of AS iPSC-derived EBs.

The selected iPSC clones from each Dup15q genotype were then differentiated using either an EB-based differentiation protocol that mimics embryonic development or a modified monolayer neural differentiation protocol [19, 20, 31]. Both protocols generated neural progenitors and subsequently post-mitotic neurons from all genotypes. Analysis of 10-week-old monolayer neuronal cultures by immunocytochemistry demonstrated the cultures largely consisted of MAP2-positive neurons (Figure 2a). Astrocytes, labeled by S100 calcium binding protein beta (S100β), were also observed (Figure 2b). The neuronal population contained both vesicular glutamate transporter 1 (VGlut1)-positive excitatory neurons and glutamate decarboxylate 65 (GAD65)-positive inhibitory neurons (Figure 2c and Figure 2d). Co-localization of the postsynaptic density protein PSD-95 and the presynaptic marker Synapsin I along MAP2-positive neurites indicates the formation of functional synapses in 10-week-old neuronal cultures (Figure 2e-e”).

In order to compare neural cultures derived from the different cell lines used in this study, with respect to the types of cells present, we analyzed expression of neuronal (βIII-tubulin, RNA binding protein fox-1 homolog 3 (RBFOX3), and T-box brain 1 (TBR1)) and glial (S100β) genes in three independent neural cultures from each cell line. This analysis revealed that there was no significant difference in mean expression levels between cell lines [see Additional file 6: Figure S4A] (one-way ANOVA; βIII-tubulin: F_6,14 = 0.9295, P = 0.5037; RBFOX3: F_6,14 = 1.938, P = 0.1443; TBR1: F_6,14 = 2.174, P = 0.1086; and S100β: F_6,14 = 2.786, P = 0.0535). We also determined that all cell lines yielded neural cultures expressing similar levels of genes specific to excitatory (VGLUT2) and inhibitory (GAD1) neurons (Additional file 6: Figure S4A, one-way ANOVA; VGLUT2: F_6,14 = 2.597, P = 0.0662 and GAD1: F_6,14 = 2.932, P = 0.0456). Though the levels of expression varied between cell lines, we observed that all cell lines generated cells which expressed markers of dorsal forebrain including forkhead box G1 (FOXG1), PAX6, and orthdenticle homeobox 2 (OTX2) [see Additional file 6: Figure S4A] (one-way ANOVA; FOXG1: F_6,14 = 11.75, P < 0.0001; PAX6: F_6,14 = 3.878, P = 0.0172; and OTX2: F_6,14 = 11.97, P < 0.0001). The midbrain specific gene engrailed 1 (EN1) was expressed in neural cultures from all cell lines but at noticeably lower levels than the forebrain-specific genes [See Additional file 6: Figure S4A] (one-way ANOVA, F_6,14 = 7.664, P = 0.0009). We also assayed for expression of NK2 homeobox 2 (NKX2.1), a marker of ventral forebrain progenitors, and found that its levels were very low or undetectable in most cultures. Importantly, the replicate neural cultures used in these experiments included those derived by both monolayer culture and EB differentiation, demonstrating that there were no significant differences in the types of cells generated between the two protocols.

We carried out electrophysiological recordings from differentiated neurons derived from the idic(15) iPSC line (Idic1-8) during in vitro development using the EB differentiation protocol. In 3-week-old cultures, most neurons (21/24 cells) fired a single mature action potential in response to depolarizing current injection. In 6-week-old cultures, trains of action potentials were induced in more than half of the neurons recorded (29/51 cells). At weeks 12 to 14, a similar percentage of cells fired trains of mature action potentials and displayed large (>1 nA) inward and outward currents in response to depolarization (Figure 3a-c). Inward sodium currents and action potentials were blocked by the sodium channel blocker tetrodotoxin (1 μM; data not shown). Spontaneous excitatory synaptic events were seen in about half of the neurons in 6-week-old cultures (28/52 cells), with an average frequency of 0.11 ± 0.03 Hz (n = 28). In 14-week-old cultures, as shown in Figure 3d-e, isolated synaptic currents and occasional bursts of activity were seen in the majority of cells at a slightly higher frequency (0.22 ± 0.05 Hz; n = 20). Spontaneous excitatory synaptic currents were blocked by the AMPA receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) (10 μM; data not shown). These data indicate that Dup15q neurons show in vitro maturation of intrinsic excitability and evidence of functional synapses, laying the groundwork for a detailed comparison of functional properties among different genotypes.

The 15q11-q13.1 region includes several genes involved in neuronal structure and function (Figure 1a). Imprinting in the 15q11-q13.1 region results in allele-specific expression of several genes such that in non-neuronal cells (including pluripotent cells), SNRPN is expressed from the paternal allele and UBE3A is expressed from both parental alleles. In neurons, however, SNRPN is still expressed solely from the paternal allele, but UBE3A becomes paternally silenced and is expressed only from the maternal allele [32]. While some studies suggest that ATP10A is maternally expressed in the brain, others suggest that it is biallelically expressed or that its imprinting status varies among individuals [33]. Other genes in the region, including TUBGCP5, CYFIP1, NIPA1, GABRB3, HERC2, and CHRNA7 are expressed biallelically in all cell types tested to date. Evidence from analysis of both Dup15q lymphoblast and postmortem brain tissue samples suggests that expression levels of these genes differ significantly from normal individuals [34, 35]. To investigate whether copy number variations of the 15q11-q13.1 region result in corresponding changes in the expression of genes located in the region, we performed qRT-PCR with iPSCs and iPSC-derived neurons (both monolayer and EB-derived) from all genotypes for TUBGCP5, CYFIP1, NIPA1, SNRPN, UBE3A, ATP10A, GABRB3, HERC2, and CHRNA7 [see Additional file 7: Figure S5A-B]. In each case, we compared the expression level of the gene in the disease-specific cells to the normal cells.

For all genotypes, expression of the 15q11-q13.1 genes in iPSCs generally corresponded to expressed copy number when compared to iPSCs derived from a control individual (Nml 1-0) (Figure 4a, top panel). The levels of expression of paternal-specific SNRPN in AS iPSCs and in the maternal int dup (15) and idic(15) iPSCs were similar to normal iPSCs because they only have a single active copy of the gene. The paternal int dup(15) iPSCs showed a twofold increase in SNRPN expression compared to normal iPSCs, reflecting the two active copies of the gene (Figure 4a, top panel; Additional file 7: Figure S5A). UBE3A, which is not imprinted in iPSCs was expressed at approximately half the level of normal in AS iPSCs and between two- and threefold of normal in idic(15) iPSCs. While UBE3A expression fell within the range of normal expression in both iPSC lines derived from the maternal int dup(15) sample and the paternal int dup(15) iPSCs, the average UBE3A expression level in all three of these iPSC lines was consistently higher than that of normal iPSCs.

The expression levels of HERC2 and NIPA1 were not as predicted from copy number in some of the iPSC lines. HERC2 was expressed at levels near or slightly less than those seen in normal iPSCs in both the paternal and maternal int dup(15) iPSCs (Figure 4b, top panel; Additional file 7: Figure S5A). Expression of NIPA1 was roughly half that of normal iPSCs in both clones of the maternal int dup(15) iPSC lines [see Additional file 7: Figure S5A].

Expression levels of most 15q11-q13.1 genes in 10-week-old AS and idic(15) iPSC-derived neurons also largely correlated with their respective expressed copy numbers (Figure 4a, bottom panel), with one notable exception. HERC2 expression in one of the idic(15) cell lines (IdicCB-09) was nearly 4.5-fold higher than normal, which is a deviation from the expected twofold increase due to these cells having two additional copies of the gene (Figure 4b, bottom panel, Additional file 7: Figure S5B). UBE3A is imprinted in iPSC-derived neurons and, accordingly, UBE3A expression levels were less than 25% of normal levels in AS iPSC-derived neurons. In idic(15) iPSC-derived neurons, UBE3A levels were about 2.5-fold higher than normal neurons, as expected for cells with three active copies of the gene (Figure 4b, bottom panel; Additional file 7: Figure S5B).

In the maternal int dup(15) iPSC-derived neurons (mat. int dup(15)-02), we found that UBE3A was expressed at the predicted 1.5-fold over normal. However, UBE3A was also increased to a similar degree in the isogenic normal (mat. int dup(15)-12) neurons. In order to understand this apparent discrepancy in expected UBE3A expression, we analyzed UBE3A copy number in the maternal int dup(15) and isogenic normal clones using a qPCR CNV assay, and determined that both cell lines had three copies of UBE3A [see Additional file 8: Figure S6A]. In addition, we detected RNA expression from three UBE3A alleles by RNA-FISH with a UBE3A BAC probe in iPSCs of both cell lines [see Additional file 8: Figure S6B]. This suggests that while initial analysis of iPSC clones, using DNA FISH for SNRPN and the PWS-IC methylation assay, indicated that the mat. int dup(15)-12 clone did not contain the 15q11-q13.1 duplication, this clone was also mosaic and over approximately 25 passages resolved to consist mostly of duplication-carrying cells. Therefore, we can interpret data generated using this cell line as a second maternal interstitial duplication clone.

15q11-q13.1 gene expression in the maternal int dup(15) and paternal int dup(15) iPSC-derived neurons did not correlate well with expressed copy number for several genes (Figure 4a, bottom panel). HERC2 levels were threefold higher than normal in maternal int dup(15) neurons, as opposed to the predicted 1.5-fold increase from the three expressed copies of the gene in these cells (Figure 4b, bottom panel). We also observed unexpected increases in expression of CYFIP1 and CHRNA7 in both the maternal and paternal int dup(15) iPSC-derived neurons (Figure 4b, bottom panel). We did observe the expected increase in SNRPN expression in the paternal int dup(15) neurons, however, expression of NIPA1, UBE3A, HERC2, and CHRNA7 were also higher than expected based on expressed copy number.Together, these data suggest that gene expression level correlates well with expressed copy number in iPSCs but less so in iPSC-derived neurons (compare top and bottom panels of Figure 4a). Furthermore, interstitial duplication iPSCs and neurons are more likely to have gene expression levels different from expected levels based on expressed copy number.

We also sought to analyze gene expression genome-wide. To test the hypothesis that alterations in gene expression would reflect either 1) disrupted pathways caused by altered dosage of the duplicated 15q region or 2) the transcriptional responses to the physiological changes in the AS and idic(15) neurons, we performed strand-specific mRNA-Seq on 10-week old neurons generated from AS, normal, and idic(15) (Idic1-8) iPSCs. mRNA-Seq data were generated using neurons derived from two independent rounds of in vitro differentiation using the EB protocol from iPSCs of each of the three genotypes. Gene expression levels were compared between biological replicates and were found to be highly similar [see Additional file 9: Table S3, r² ≥ 0.97]. The gene expression levels for the replicates were averaged and the AS and idic(15) samples were each compared to the normal sample.

Genes showing a twofold increase or decrease in expression compared to the normal sample were considered differentially expressed between those two samples. 5,369 genes were differentially expressed between idic(15) and normal iPSC-derived neurons (Figure 5a). A total of 1,667 genes were found to be differentially expressed between AS and normal iPSC-derived neurons (Figure 5a). Of these, 751 genes were differentially expressed in both samples compared to normal (Figure 5a). The number of genes found to be differentially expressed in both samples is greater than expected by chance (hypergeometric distribution with population size = 27,297; P = 1.88 X 10^-131). Moreover, 76% of the genes differentially expressed in both AS and idic15 neurons show the same tendency (that is, both upregulated or both downregulated). Figure 5b shows a heat map of the top 200 most differentially expressed genes in both samples. Gene Ontology analysis was performed on the lists of genes specifically up- or downregulated in AS and idic(15) neurons compared to normal neurons [see Additional file 10: Table S4, Additional file 11: Table S5, Additional file 12: Table S6, and Additional file 13: Table S7]. The gene ontology term ‘neuronal differentiation’ was the most significantly enriched category from the lists of genes downregulated in both AS and idic(15) neurons (P = 9.7 X 10^-10 and P = 4.31 X 10^-8 for AS and idic(15) neurons, respectively). Genes involved in cell cycle (P = 1.64 X 10^-23) and protein catabolic processes (P = 5.96 X 10^-19) were enriched among the list of genes upregulated in idic(15) neurons.

Interestingly, several known autism and epilepsy candidate genes, namely distal-less homebox 2 (DLX2), aristaless related homeobox (ARX), ISL LIM homeobox 1 (ISL1), neuroligin 1 (NLGN1), SH3 and multiple ankyrin repeat domains 1 (SHANK1), adenosine A2A receptor (ADORA2A), and distal-less homeobox 5 (DLX5), were found in the list of genes differentially expressed and downregulated in AS or idic(15). To validate the transcriptome data and further interrogate some of these genes, qRT-PCR was performed on RNAs isolated from normal, AS and idic(15) (Idic1-8) neurons (Figure 6a). All seven of the genes assayed were shown to be downregulated in both AS and idic(15) iPSC-derived neurons. qRT-PCR data from 19 different genes demonstrated agreement between qRT-PCR and mRNA-Seq data (Figure 6b). Although maternal int dup(15) iPSC-derived neurons were not included in the whole transcriptome analysis, we also assayed expression of the selected autism and epilepsy candidate genes in these neurons by qRT-PCR. We observed that, with the exception of ADORA2A, expression of these genes was also downregulated in maternal int dup(15) iPSC-derived neurons when compared to normal neurons [see Additional file 14: Figure S7].

To investigate whether the aberrant expression levels of UBE3A in idic(15) neurons could be pharmacologically manipulated, we treated 10-week-old idic(15) iPSC-derived monolayer neuron cultures with the DNA-binding compound mithramycin. This drug binds to GC-rich sequences in specific gene promoters and competes with transcription factors that regulate gene expression [36, 37]. We reasoned that since the UBE3A promoter is GC-rich and is known to be occupied by many transcription factors in various cell types [38], competition for binding by mithramycin may alter the levels of UBE3A transcription. Following 72 hours of drug treatment, we found that UBE3A mRNA levels in idic(15) neurons were reduced in a dose-dependent manner to levels near that of normal iPSC-derived neurons (Figure 7a). These data suggest that UBE3A can be pharmacologically regulated at the transcriptional level by mithramycin. However, competition for transcription factor binding by mithramycin is not specific for UBE3A and occurs at many gene promoters. In line with this, we detected changes in expression levels of several other genes following mithramycin treatment, including the other 15q11-q13.1 genes SNRPN, GABRB3, and CYFIP1 [see Additional file 15: Figure S8].

In an effort to identify possible mechanisms whereby mithramycin treatment rescues UBE3A expression levels, we assayed transcription factor binding to UBE3A by ChIP. It is well established that mithramycin inhibits binding of the Sp1 transcription factor and affects transcription of Sp1 target genes [37, 39]. We were unable to detect Sp1 enrichment at several sites analyzed within the UBE3A promoter [see Additional file 16: Figure S9A]. For this reason, we instead analyzed binding of the transcription factor yin yang 1 (YY-1), which we confirmed does bind within the UBE3A promoter in both iPSCs and iPSC-derived neurons [see Additional file 16: Figure S9B]. YY-1 can act both as a transcriptional activator and repressor [40]. Transcriptional repression by YY-1 is thought to be mediated by recruitment of histone deacetylases and histone acetyltransferases [41]. We assayed binding of YY-1 to UBE3A at two sites, one within exon 1 and one about 250 base pairs upstream of exon 1, following treatment of idic(15) iPSC-derived neurons for 72 hours with either 100 nM mithramycin or DMSO. While there was no significant difference between the treatments in binding of YY-1 at exon 1, binding at the upstream site was increased in mithramycin-treated neurons to roughly 1.4-fold that of DMSO-treated neurons (Figure 7b, P = 0.0571). These data suggest that mithramycin treatment may recruit YY-1 to the UBE3A promoter, where it can act in a repressive manner to decrease UBE3A expression levels.