Functional Neuronal Cells Generated by Human Parthenogenetic Stem Cells

Parent of origin imprints on the genome have been implicated in the regulation of neural cell type differentiation. The ability of human parthenogenetic (PG) embryonic stem cells (hpESCs) to undergo neural lineage and cell type-specific differentiation is undefined. We determined the potential of hpESCs to differentiate into various neural subtypes. Concurrently, we examined DNA methylation and expression status of imprinted genes. Under culture conditions promoting neural differentiation, hpESC-derived neural stem cells (hpNSCs) gave rise to glia and neuron-like cells that expressed subtype-specific markers and generated action potentials. Analysis of imprinting in hpESCs and in hpNSCs revealed that maternal-specific gene expression patterns and imprinting marks were generally maintained in PG cells upon differentiation. Our results demonstrate that despite the lack of a paternal genome, hpESCs generate proliferating NSCs that are capable of differentiation into physiologically functional neuron-like cells and maintain allele-specific expression of imprinted genes. Thus, hpESCs can serve as a model to study the role of maternal and paternal genomes in neural development and to better understand imprinting-associated brain diseases.


Introduction
Due to their unlimited self-renewal and multilineage differentiation potential, human pluripotent stem cells have become key cell sources for cell differentiation research, disease modeling, drug discovery, and have potential for cell replacement strategies [1]. In human, pluripotent stem cell lines can be derived from in vitro cultured inner cell mass (ICM) cells of blastocyst stage embryos, producing embryonic stem cell lines (hESC) [2]. More recently, factor-driven reprogramming of somatic cells has provided a longsought strategy to generate patient-and disease-specific pluripotent stem cells, termed induced pluripotent (iPS) cells [3]. Pluripotent stem cell types, i.e. ES and iPS cells, and different lines of the same type exhibit considerable variations in respect to epigenetic status, gene expression profiles, and differentiation propensity, preventing generalized approaches but allowing for the correlation of gene expression patterns with differentiation propensities [4]. One specific ESC type that is unique in this respect is parthenogenetic (PG) ESCs that are derived from blastocysts resulting from the activation and subsequent development of an unfertilized oocyte. While asexual development of offspring from an oocyte without male genetic contribution (parthenogenesis) occurs naturally in various invertebrate and some vertebrate species [5], mammalian uniparental (PG, gynogenetic: GG, or androgenetic: AG, with only paternally derived genomes) embryos do not develop to term as a consequence of imbalanced expression of imprinted genes with parent of origin-dependent allele-specific expression patterns [6]. Despite this developmental limitation, stable ESC lines can be isolated from uniparental blastocysts of several species including human [7][8][9][10]. The in vitro differentiation capacity of murine uniparental ESC into various cell lineages, including neural and transplantable hematopoietic progenitors [11][12][13][14][15][16] indicates that these cells represent a unique model system to study the role of maternal and paternal genomes in normal development and the contribution of imprinting in disease development.
Paternally and maternally inherited alleles play non-redundant and reciprocal roles in brain development and plasticity [17].
Studies of the developmental capacity of murine PG and AG ICM cells following aggregation with biparental embryos revealed that PG cells preferentially seeded to the neocortex, striatum and hippocampus while AG cells contributed to hypothalamus and septum but were not found in the cortex [18]. Recent highresolution screens in the mouse suggest that the developing and adult brain may be subject to complex effects of imprinting, including cell type and subregion specific effects, and temporal bias, with maternal-derived gene expression at earlier stages in the developing embryonic day 15 brain and paternal gene expression bias in both the prefrontal cortex and the hypothalamus of the adult brain [19].
In vitro differentiation studies have shown that hpESCs are capable of generating multiple cell lineages including mesenchymal stem cells, hepatocytes, pancreatic endocrine cells, retinal pigmented epithelial and neural progenitor cells [20][21][22][23][24]. However, more detailed investigation is required to verify the differentiation capability of hpESCs, particularly the potential for neurogenesis and further differentiation into functional neural subtypes. The apparent contribution bias of PG and AG ICM cells to different structures of the developing brain, the large number of imprinted brain genes, and the existence of imprinting-associated neuropsychiatric diseases [17,25] could indicate that hpESCs have limited neural potential. Here, we establish that hpESCs can differentiate in vitro via NSCs into functional neuronal cells without apparent changes in imprinting status.

Whole cell patch-clamp analysis
Cells grown on glass coverslips in differentiation media for 28 days were transferred into a recording chamber and continuously superfused with extracellular solution containing 125 mM NaCl, 25 mM NaHCO 3 , 25 mM glucose, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 2 mM CaCl 2 , 2 mM MgCl 2 purged by 95% CO 2 / 5% O 2 ). The ion channel antagonists tetraethylammonium chloride (TEA, Sigma-Aldrich, 30 mM) or tetrodotoxin (TTX, Sigma-Aldrich, 1 mM) were added to the extracellular solution when indicated. All experiments were performed at room temperature using an EPC 10 double patch clamp amplifier and pulse software (HEKA, Lambrecht, Germany). Electrodes were pulled from thick-walled borosilicate glass and filled with intracellular solution (140 mM KCl, 10 mM Hepes, 10 mM EGTA, 2 mM Na 2 ATP, 2 mM MgCl 2 ) and had a resistance between 3 and 4.5 MV. Cells were held in whole-cell configuration at 280 mV and were discarded if the series resistance was higher than 25 MV at the beginning of the measurements.

Analysis of gene expression using semi-quantitative RT-PCR
Feeder cells were depleted by repeated passages on Matrigelcoated plates. Total RNA was isolated from biparental hESCs (I3 and H9 cell lines), hESC-derived neural stem cells (hNSCs), hpESCs and hpNSCs using peqGOLD RNAPure TM (Peqlabs, Göttingen, Germany). Passage numbers of hpESCs that were used to generate hpNSCs were identical. Before cDNA generation, RNA preparations were treated with DNase I (Applied Biosystems, Darmstadt, Germany). 1 mg RNA was reverse transcribed into cDNA using M-MLV reverse transcriptase (Gibco Invitrogen). For control experiments cDNA was generated from human parthenogenetic neural crest stem cells (hpNCSCs) (isolated from differentiating hpESCs at the attached EB stage), human fetal brain (hFB, 18 weeks, female; Stratagene, Santa Clara, CA, USA) and from human adipose tissue-derived mesenchymal stromal cells (hMSCs). For amplification Taq Polymerase (HT biotechnology, Cambridge, UK) was used. PCR conditions were: 94uC, 1 min, 55uC to 62uC, 60 seconds according to the primers, 72uC, 1 min (35 to 40 cycles). GAPDH was used a control. All reactions were performed on a T3 thermocycler (Biometra, Göttingen, Germany). Primers (Eurofins MWG Operon, Ebersberg, Germany) used were: Acta1 forward . PCR was performed using a modified touchdown protocol that consisted of an initial denaturation step at 94uC for 3 min, followed by 4 cycles of 94uC for 40 seconds, 62uC for 40 seconds and 72uC for 45 second. After additional 6 cycles of the same length with 60uC annealing temperature, 20 cycles were performed with successive annealing temperature decrements of 0.5uC in every cycle, followed by 15 cycles with 52uC annealing temperature. The amplified DNA fragments were sub-cloned into pJet1.2/blunt (Fermentas, Glen Burnie, Mary-land, USA) for sequencing. Analysis of sequences and diagram generation was performed using BISMA [31].

Data analysis
Results were expressed as the mean 6 SEM. Statistical analysis was performed using the Student t-test.

Neural differentiation of hpESCs
Uniparental hpESCs (ESC lines LLC9P and LLC6P [10]) were cultured using a multi-step in vitro differentiation protocol that can produce NSCs from pluripotent stem cells [26]. Initial differentiation of hpESCs produced floating embryoid bodies (EB) that formed neural rosettes after attachment (Fig. 1A, Fig. S1 A). Isolated neural rosettes could be expanded as floating neurospheres that formed monolayers with NSC-like homogeneous morphology after plating onto polyornithine/laminin-coated plates (Fig. 1A right panel). The NSC identity of monolayer cells was confirmed by gene expression analysis revealing upregulation of NSC markers Sox1, Nestin, Pax6, and Musashi1 (MS1) (Fig. 1B,.  Fig. S1 B), silencing of pluripotency marker genes (Oct4 or Nanog), and absence of activation of markers of non-neural lineage commitment, including neural crest (Snai2, FoxD3) and mesoderm (Acta1). Expression of the neural stem cell markers Nestin, Sox1, Sox2 and Vimentin in hpNSC cultures was ubiquitous and not limited to subsets of cells ( Fig. 1C and in. Fig. S1 C). Upon differentiation, two 10 cm 2 plate dishes of LLC9P hpESCs yielded a mean of 29 (63.5) million hpNSCs whereas LLC6P hpESCs generated 11.8 (61.7) million cells. As a recent report described aberrant expression levels of molecules related to spindle formation and chromosome segregation in hpESCs [20], we verified expression of these markers in undifferentiated LLC6P and LLC9P hpESCs, and detected variations in gene expression not only between PG and N cells but also between individual hpESCs (Fig. S2 A). Additionally, reduced levels of extracellular matrix (ECM) transcripts had been detected in PG compared to N (biparental) ESCs [22]. We observed variation in ECM transcript levels of ECM molecules between individual PG and N cell lines, with lower expression in LLC6P hpESCs compared to LLC9P cells and to hESCs (Fig. S2 B). In conclusion, hpESCs can differentiate into hpNSCs that express neural stem cell markers in the absence of pluripotency or neural crest cell marker expression.

Electrophysiological analysis of PG neurons
We further investigated whether PG neurons can functionally mature in vitro. As shown in Table S1, electrophysiological properties of PG neurons at 28 days of differentiation were comparable to those reported in literature for human in vitro induced neuronal cells [32]. PG neurons showed typical neuronal Na + /K + currents in voltage clamp mode (vc stimulation pattern: 280 mV to +55 mV, step size 15 mV, stimulation time 20 ms) (Fig. 4A). Depolarizing step current injections over a 500 ms time period elicited multiple action potentials with a maximum frequency of 30 Hz (Fig. 4B). When maximum in-and outward currents were plotted against the corresponding stimulation voltage, PG neurons depicted a typical neuron-like current pattern (Fig. 4C) that responded to selective pharmacological blockers of sodium (tetrodotoxin) and potassium (tetraethylammonium) channels (Fig. 4D).

Analysis of imprinted genes
To assess the status of epigenetic marks involved in the control of imprinted gene expression during neural differentiation of hpESCs, we analyzed the methylation status of CpG islands of two differentially methylated regions, the 59 region of the long noncoding RNA Kcnq1ot1 (KvDMR1) and the H19 DMR1 (Fig. 5). Methylation of KvDMR1 on the maternal allele, acquired during germ cell development, is associated with silencing of Kcnq1ot1, whereas Kcnq1ot1 expression from the unmethylated paternal allele is involved in domain-wide chromatin repression of a cluster of genes including Cdkn1c and Kcnq1 [33]. Consistent with PG origin, CpGs of the KvDMR1 were mostly methylated in hpESCs and hpNSCs, while conventional hESCs and hNSCs exhibited 50% methylation, indicating the presence of maternal and paternal alleles (Fig. 5A). Quantitative RT-PCR analysis revealed absence of Kcnq1ot1 RNA in hpESCs and hpNSCs, and higher expression of Kcnq1 but not Cdkn1c in PG compared to N cells (Fig. 5B).
Differential germline-acquired methylation of the H19 DMR, a chromatin insulator, controls reciprocal allelic silencing of the Igf2 and H19 genes. On the unmethylated maternal allele, CTCF binding blocks enhancer-initiated transcription of Igf2, allowing H19 expression, while methylation on the paternal allele abolishes insulator function permitting Igf2 expression and leading to silencing of H19 in cis [34]. The majority of DMR1 CpGs were methylated in N cells, whereas PG cells exhibited partial or complete absence of CpG methylation (Fig. 5C). Parent of originspecific expression of Igf2 and H19 was maintained in PG cells with absence of Igf2 expression and overexpression of H19 in PG compared to N cells (Fig. 5D).
Expression analyses of additional imprinted genes revealed that silencing of the paternally expressed Snrpn and Nnat genes was preserved in hpESCs and hpNSCs. Levels of the maternally expressed Gtl2, Dlx5 and Kcnk9 genes were overall higher in hpESC and hpNSC compared to N cells. However, Igf2r expression was elevated only in hpESCs but not in hpNSCs ( Fig. 5E, Fig S5). Together, parent of origin-specific gene expression control appears to be largely maintained in hpESC lines LLC6P and LLC9P, and neural differentiation is not associated with a loss of silencing of paternally expressed genes that were analyzed.

Discussion
Our objective was to define the neural differentiation potential of hpESCs in vitro. In summary, we describe that hpESCs -despite having a maternal genome only -generate proliferating NSCs that are capable of differentiating towards neurons that express specific markers for neuronal transmitters and synaptic proteins and show electrical activity. PG cells maintain allele-specific expression of imprinted genes.
Our results confirm the neural differentiation potential of hpESCs. In particular, the results show that hpESCs can generate proliferating hpNSCs, which can further differentiate not only to early neural lineages as described earlier [8][9][10] but also into mature neural and glial cell types. hpNSCs respond to signals directing the derivation of ventral midbrain dopaminergic and motoneurons. Similar to earlier reports on conventional NSCs, we observed high frequencies of GABAergic neurons [26,35,36]. The reason for the preference of hpNSCs towards GABAergic differentiation is not clear, although the high concentrations of mitogens present during the expansion of hpNSCs are likely a contributing factor. The preference towards neuronal and less glial differentiation outcomes mimics the developmental potential of NSCs in the developing brain [37]. We further show that PG neuron-like cells were capable to generate action potentials and possessed membrane characteristics similar to newly formed neurons [26,35]. The unperturbed neural differentiation potential of hpESCs is consistent with earlier reports of successful murine AG ESC-derived neurogenesis [11,12,16]. Our analyses indicate that uniparental ESCs are less restricted in their neural developmental potential than predicted from in vivo studies [18,38].
Previous analyses of neural differentiation potential of hpESC via sphere formation suggested that hpESCs yielded low quantities and less mature neural cells compared to conventional ESCs [22]. Our results are in contrast to this report, with several factors likely to contribute to such a difference. Firstly, we subjected hpESC lines (a subset of those used by [22]) to an alternative differentiation protocol, which was optimized towards the derivation of a homogeneous population of NSCs, and minimized spontaneous differentiation and lineage restriction. Secondly, our results revealed differences in gene expression of extracellular matrix proteins not only between hpESCs and hESCs but also between individual hpESC cell lines. Therefore, low yields of ESderived NSCs from LLC6P compared to LLC9P hpESCs may be related to poor cell-cell interaction caused by low levels of ECM gene expression in LLC6P [22]. High hpNSC yields of LLC9P hpESCs could be caused by the elevated expression of the early neuroblast marker NCAM1 [39]. We also observed differences in the expression of mitotic checkpoint genes in hpESCs in comparison to hESCs as well as between the two hpESC cell lines. Possible explanations for these differences likely include cell line to cell line variation [40] and potentially an underlying genetic instability of uniparental ESCs [20]. A recent report suggested that PG ES cells of different species origin show centrosomal amplification and chromosomal instability [41]. Previous analyses of hpESC line LLC6P and LLC9P revealed a normal human 46,XX karyotype suggesting that the cells under study are chromosomally normal [10].
ESC lines can undergo epigenetic changes during in vitro culture [42][43][44]. Although hESC exhibit a substantial degree of epigenetic stability, despite differences in genetic background, derivation and expansion conditions [44], imprinted loci have been found to exhibit varying susceptibility to culture-induced epigenetic changes, with more stability at the Kcnq1ot1 locus and less at H19/Ifg2 [43]. Consistent with such observations, we detected conservation of maternal-specific CpG methylation at the KvDMR, low expression of Kcnq1ot1, and upregulation of maternally expressed Kcnq1 in PG cells, although Cdkn1c transcripts were only upregulated in one PG cell line (LLC6P). Our analyses of methylation of DMR1 of the H19/Igf2 locus agree with previous reports suggesting that late passage hESCs are prone to hypermethylation this region [43]. Here we observe hypermethylation in hESCs and hNSCs, and modest gain of CpG methylation in hpESCs. Despite these changes, Igf2 and H19 transcript levels in hpESCs and NSCs remained consistent with PG origin, indicating that regulatory mechanisms other than CpG methylation are involved in imprinting control of H19 and Igf2 [45]. Other paternal (Snrpn and Nnat) and maternal (Dlx5, Gtl2, Ube3a and Kcnk9) imprinted genes maintained their parent of origin-specific gene expression pattern. Igf2r expression was elevated only in hpESCs but not in hpNSCs. The molecular basis for this remains unclear. Increased methylation in the higher passage hESCs used in our study may be a consequence of the in vitro expansion of ESCs, however, overall, our analyses indicate that PG cells are epigenetically as stable as N cells.
Considering the putative prevalence of imprinted genes expressed in the developing mammalian brain [17] and the altered expression of ECM genes and molecules related to spindle formation and chromosome segregation [20,22], the capacity of hpESCs to undergo similar in vitro neural differentiation as hESCs seems surprising. This suggests either that there is a less stringent role for imprinted gene expression during neuronal in vitro differentiation, or that a requirement for balanced expression of imprinted genes is not required for differentiation to the stages analyzed. While we show the successful in vitro differentiation of hpESCs into neural subtypes and that PG neurons develop synaptic contacts and electrical activity, transplant models will ultimately be required to assess the broader neural differentiation potential of hpESCs.  Figure S5 RT-PCR analysis of imprinted genes in hpESCs and hpNSCs (LLC6P). Relative expression levels of the imprinted genes: Ifg2, Snrpn Nnat and Kcnq1ot1 (paternally expressed) and, Dlx5, H19, Ube3a, Igf2r, Kcnq1, Cdkn1c, Gtl2 and Kcnk9 (maternally expressed) were analyzed by RT-PCR in PG and N cells (I3 and H9). The 2 2DDCt method was used to calculate fold change in the expression of imprinted genes. Expression levels N cells were set to 1. GAPDH was used as a reference gene. n = 3, * p,0.05, ** p,0.01, *** p,0.001 by Student's t-test. (TIF)