Correspondence to: K.S. O'Shea, Department of Cell and Developmental Biology, University of Michigan Medical School, 109 Zina Pitcher Place, 3051 BSRB, Ann Arbor, MI 48109-2200. E-mail: firstname.lastname@example.org
Vertebrate neurogenesis is a stepwise process in which the primitive ectoderm is first induced to form the neural ectoderm, which folds forming the neural tube. The neural tube is then patterned along the anterior–posterior (AP) and dorsal–ventral (DV) axes by morphogens that coordinate the expression of transcription factors that confer positional identity and/or promote neural differentiation. Proneural basic helix–loop–helix (bHLH) proteins comprise one such class of transcription factor essential during early neurogenesis (review, Bertrand et al., 2002). In Drosophila, the bHLH factors achaete-scute and atonal act as generic promoters of neuronal differentiation and neuronal subtype specification (Chien et al., 1996; Jarman and Ahmed, 1998). Vertebrate atonal homologs such as Neurogenin 1 (Neurog1), Neurogenin 2 (Neurog2; Gradwohl et al., 1996; Ma et al., 1996; Sommer et al., 1996) and achaete-scute homologs such as Atoh1 (Mash1; Johnson et al., 1990) and Ascl1 (Math1; Guillemot and Joyner, 1993) also promote neuronal differentiation both in vivo (Turner and Weintraub, 1994; Lee et al., 1995; Ma et al., 1996; Chung et al., 2002; Kim et al., 2004) and in vitro (Lo et al., 1998; Farah et al., 2000; Sun et al., 2001; Kanda et al., 2004; Satoh et al., 2010).
The expression of mammalian achaete-scute and atonal homologues within specific—largely nonoverlapping—regions of the developing central and peripheral nervous system (CNS and PNS) suggests roles in neuronal subtype specification that have been confirmed by loss- and gain-of-function studies. For example, Neurog1 is expressed in the dorsal telecephalon where it appears to promote glutaminergic neuronal fates, Atoh1 is expressed in the ventral telencephalon specifying γ-aminobutyric acid (GABA)ergic neurons (Fode et al., 2000; Parras et al., 2002; Kim et al., 2011), while Neurog1 is expressed in the caudal ventricular zone of the rhombic lip, where it defines multiple GABAergic lineages (Dalgard et al., 2011). In the spinal cord, Ascl1 is expressed in a dorsal stripe near the roof plate (Gowan et al., 2001), Neurog1 is expressed in the ventral half and in a small region just below the roof plate, whereas Atoh1 is found in the intervening domain (Sommer et al., 1996; Ma et al., 1997), where these transcription factors are thought to regulate neuronal phenotype by cross-inhibition (Briscoe et al., 2000; Gowan et al., 2001; Helms et al., 2005). Loss-of-function studies have shown that Neurog1 is required for the development of dI2 dorsal spinal neurons, trigeminal and otic cranial sensory ganglia, and TrkA neurons of dorsal root ganglia (DRG) (Ma et al., 1997; Fode et al., 1998; Gowan et al., 2001). Gain-of-function studies have demonstrated that over-expression of Neurog1 biases the migration of neural crest stem cells toward dorsal root sensory ganglia in vivo (Perez et al., 1999), whereas forced expression of Neurog1 in dorsal neural tube progenitors and neural crest cells promotes their differentiation into sensory lineages (Lo et al., 2002). These data indicate that Neurog1 is required for the development of sensory neuronal lineages in both the PNS and CNS; however, it is not clear whether Neurog1 is itself sufficient to induce these lineages because the gain-of-function studies were conducted either in the embryo or in neural progenitors where the effects of morphogens and other instructive signals cannot be separated. While mis-expression of proneural genes can produce ectopic neurogenesis in a variety of species (Quan and Hassan, 2005), relatively little is known regarding the molecular mechanisms involved or down-stream gene expression following bHLH gene expression. Because bHLH transcription factor expression is strongly affected by spatial and temporal context (Powell and Jarman, 2008), we used a gain-of-function approach in pluripotent embryonic stem cells (ESC) to determine the role of Neurog1 in cell fate specification. ESC may be a particularly informative starting material because they have a bivalent chromatin structure with promoters poised for both lineage differentiation as well as for self-renewal (e.g., Boyer et al., 2006). Lineage specifying genes such as bHLH and paired-box family members may therefore control differentiation programs by directly affecting transcription and by narrowing differentiation choices by controlling chromatin.
The current investigation identifies potential down-stream targets of Neurog1 including genes involved in cell cycle, cell migration and process outgrowth, and provides a source of neuronal precursor cells that remain sensitive to patterning molecules. Consistent with observations that Neurog1 is present in cells about to withdraw from cycle and differentiate into layer-specific neurons (Kim et al., 2011), forced expression of Neurog1 in ESC alters their cell cycle characteristics and is sufficient to initiate neuronal differentiation in the absence of other inducing factors. In fact, Neurog1 expression was sufficient to overcome the inhibitory effects of LIF and serum proteins on ESC differentiation (Williams et al., 1988). In addition, Neurog1 expression was also sufficient to generate both CNS and PNS neuronal subtypes typical of those dependent on Neurog1 in vivo. Yet, the positional identity and neuronal differentiation potential of induced cells could be influenced by early acting patterning factors, suggesting that Neurog1 promotes differentiation of neuronal precursors that can be influenced by the local microenvironment to subsequent regional and/or subtype-specific differentiation.
Inducible Expression of Neurog1 in ESC
In the current investigation, we used the Ainv15 ESC line (Kyba et al., 2002) that expresses a “Tet-on” reverse tetracycline transactivator (rtTA) from the constitutively active ROSA26 locus, and a tet-inducible element with a LoxP targeting site and a truncated neomycin resistance cassette placed upstream of the HPRT locus. Site-specific integration of the targeting vector carrying cDNA for a gene of interest places the transgene downstream of the tet-inducible element and confers neomycin resistance, thereby allowing efficient selection of targeted cell lines. Once selected, cells are maintained in neomycin to ensure that the transgene does not undergo methylation or rearrangement, although this is unlikely at the HPRT locus (Wutz et al., 2002).
We targeted Neurog1 to the inducible locus and obtained several G418 resistant colonies which were expanded to cell lines. Based on the reliability of this inducible system (Kyba et al., 2002; M. Kyba and G. Keller, personal communication) and our own analyses in which we observed no differences between the ESC lines in their differentiation capacity or response to the tetracycline analog doxycycline (data not shown), one line (N7) was selected for most subsequent experiments. As indicated in Figure 1A, robust Neurog1 expression in N7 cells was induced within 12 hr of addition of doxycycline (dox), to monolayer cultures grown in either complete ES medium, which inhibits spontaneous differentiation (Williams et al., 1988) or in a defined neural medium, which supports neural differentiation of ESC in monolayer cultures. We did not observe any evidence of transgene silencing as Neurog1 expression was maintained throughout the culture period. N7 cells cultured in complete (ES) medium without dox did not express Neurog1 at 12, 24, 48, or 72 hr (Fig. 1A). N7 cells cultured in defined neural medium without dox were intolerant of these culture conditions often undergoing cell death by 72 hr (not shown). However, in cells that survived to 72 hr, it was possible to detect slight expression of Neurog1 (Fig. 1A), likely due to stochastic neural differentiation and activation of endogenous Neurog1 elicited by culture in these conditions. When cultured in a different serum-free medium containing 5% knock-out serum replacer (Fig. 1B), we observed no expression of Neurog1 in the absence of dox, but, in the dox-treated groups, we could consistently induce expression of the Neurog1 transgene in a dose-dependent manner. Overall, these data suggest that the N7 cell line provides a consistent and tightly regulatable means to assess the sequelae of Neurog1 expression in ESC.
Expression of Neurog1 in ESC Induces Neuronal Differentiation
Within 24 hr of inducing the Neurog1 transgene, N7 cells exhibited overt neuronal morphologies including round, phase-bright somata and one or more long cytoplasmic processes when viewed with phase contrast microscopy. Cells cultured in complete (ES) medium or in defined neural medium (DM) with dox were fixed after 12, 24, 48, and 72 hr in culture followed by immunohistochemical localization of Neurog1 and neuronal tubulin (TuJ1 antibody). Virtually all of the cells cultured in the presence of dox were Neurog1+ at each time point surveyed (Fig. 2), whereas those cultured without dox were largely Neurog1− (not shown), confirming the integrity of the tet-inducible locus. TuJ1 immunostaining demonstrated weakly reactive cells in both culture conditions as early as 12 hr following transgene induction, indicating the initiation of neuronal differentiation (Fig. 2A,B,I,J), although in defined medium there were occasionally cells that exhibited a more mature neuronal phenotype (Fig. 2A,B). In general, neuronal differentiation was robust in either culture condition after 72 hr of transgene induction, but differentiation progressed more rapidly in defined medium. After 72 hr, cell density was also considerably greater in cultures grown in complete medium. Overall, forced expression of Neurog1 was sufficient to induce widespread neuronal differentiation.
Neurog1 Expression Reduces Proliferation and Increases G1 Phase of the Cell Cycle
Neurog1 has previously been reported to promote cell cycle exit (Farah et al., 2000), to increase G1/G0 and decrease S phase (Piao et al., 2012), but the kinetics and target genes are not known. To examine this directly, we first determined the growth characteristics of control Ainv15 and N7 cells in the presence and absence of doxycycline at 24-hr intervals. There was no significant difference in cell number between control Ainv15 ESC exposed to doxycycline and N7 cells grown without dox at any time point (Fig. 3A). Although there was no significant change in cell number at 24 hr of culture, by 48 and 72 hr of transgene induction, there was a significant decrease in the number of Neurog1-expressing N7 cells compared with uninduced N7 or control Ainv15 ESC (P≤0.001, Student's t-test).
To determine where in the cell cycle Neurog1 expression affects proliferation, we cultured control Ainv15 and N7 cells±dox for 72 hr, labeled the cells with propidium iodide and used fluorescence activated cell sorting (FACS) to analyze cell cycle characteristics. Consistent with the proliferation data, there was little difference in the cycle characteristics of control Ainv15 ESC grown in the presence of doxycycline and N7 cells grown without dox: 36.8 and 38.1% of the cells were in G1, 50.8 and 50% in S, and 12.4 vs. 11.9% in G2/M in control+dox, and N7−dox cells, respectively (Fig. 3B). Induction of Neurog1 expression had a significant effect on the number of cells in G1 phase; with 66.2% in G1, 22.9% in S, and 10.9% in G2/M. Cell death was similar in N7 ESC exposed to doxycycline (8.8±2.7%) and N7 not exposed to dox (8.2±0.5%), but only 4.0±0.8% in control Ainv15 ESC grown with doxycycline. These data demonstrate that induction of Neurog1 expression in ES cells slows proliferation by significantly increasing G1 and decreasing the proportion of cells in the S phase of the mitotic cycle.
Treatment With Retinoic Acid Posteriorizes Induced N7 Cells
To determine the anterior–posterior characteristics of the differentiating cells, and to determine their responsiveness to patterning factors, cells were grown in defined neural medium with and without 1 μm retinoic acid (RA). Expression of Otx2, a forebrain marker (Mallamaci et al., 1996), and Hoxc6, expressed in the rostral cervical spinal cord, were examined. In both culture media, cells expressed Otx2 but not Hoxc6 at all time points assayed, indicative of a forebrain phenotype (Fig. 4A). When N7 cells were cultured for 3 days in defined medium supplemented with 1 μM RA and dox, there was a significant reduction in Otx2 expression and concomitant up-regulation of Hoxc6 (Fig. 4A). Like previous studies of ESC (Wichterle et al., 2002), these findings suggest that N7 cells neuralized by forced Neurog1 expression have an initial anterior (forebrain) phenotype that can be posteriorized by exposure to RA.
Expression of Dorsal and Ventral Markers Reflects Neurog1 Expression In Vivo
Molecular markers of dorsal–ventral identity in the posterior neural tube have been well-characterized (reviewed in Jessell, 2000; Helms and Johnson, 2003; Helms et al., 2005; Fig. 4B); Neurog1 is expressed in the ventral half and near the roof plate of the developing posterior neural tube (Sommer et al., 1996; Ma et al., 1997, Fig. 4B). We therefore sought to determine if induction of the Neurog1 transgene could produce a similarly mixed population of dorsal and ventral subtypes following treatment with retinoic acid. We examined induced cell cultures grown in ESC medium, in which an indeterminate amount of RA was likely present in the fetal bovine serum constituent (Napoli, 1986), and in defined neural medium supplemented with 1 μM RA. When assayed by reverse transcriptase-polymerase chain reaction (RT-PCR), induced cultures expressed markers associated with the dorsal (Pax3, Pax7), ventral (Nkx2.2, Nkx6.1), and middle (Dbx2) thirds of the developing neural tube (Fig. 4C), which generally correspond to regions of Neurog1 expression in vivo. Differences in the timing of expression were observed between the proteins assayed; e.g., Dbx2 expression was evident within 12 hr of inducing the Neurog1 transgene, although it was not expressed by undifferentiated ESC, while robust Pax7 expression was not observed until 48 hr after induction. Of interest, in subsequent experiments, Pax7 and Nkx6.1 were rarely co-expressed (Fig. 7), indicating the presence of distinct dorsal and ventral subtypes. Thus, forced expression of Neurog1 in ESC produces a population of precursors that approximately recapitulates the in vivo pattern of Neurog1 expression.
Forced Expression of Neurog1 Produces CNS and PNS Neurons
The distribution of CNS and PNS phenotypes was determined after 5 days of culture with and without the addition of RA. Cells were grown in defined medium plus dox to induce the Neurog1 transgene for 72 hr then switched to a medium supplemented with 5% knockout serum replacer (SRM5) without dox for 48 hr, as these conditions increased cell survival. After 5 days, there was extensive neuronal differentiation as evidenced by TuJ1 immunoreactivity (Fig. 5A); a subset of these cells (Fig. 5B) was also positive for peripherin, with a Tuj1+:peripherin+ ratio of approximately 21:1 (Table 1). To obtain quantitative data, we scored only peripherin+ somata, so these data likely underestimate the total, as there were many peripherin+ fibers whose cell bodies were outside the field of view. Some neurons expressed Islet1 (Fig. 5C), and, invariably, these cells were also Brn3a+ (Fig. 5D), an expression pattern typical of sensory ganglia (SG) of the PNS as well as of dI3 dorsal interneurons in the CNS (Gowan et al., 2001). There were also many Brn3a+/Isl1− cells (Fig. 5C,D); an expression pattern that characterizes dI-1, 2, or 5 interneurons in the CNS (reviewed in Helms and Johnson, 2003). Nearly all (approximately 97%) of the peripherin+ cells were also Brn3a+ and had long, well-developed processes and unipolar or bipolar morphologies (Fig. 5E,F)—a pattern most consistent with a SG neuronal phenotype. Of interest, there was little statistically significant difference in marker expression in groups exposed to RA vs. those not exposed to RA, although the overall number of neurons present as evidenced by TuJ1 immunoreactivity was slightly higher in the RA-treated cultures as was the number of peripherin+ neurons (Table 1).
N7 cells were grown for three days in defined neural medium in the presence of Dox, followed by two days without Dox in defined neural medium±1 μM retinoic acid (RA). At least 1000 cells from 10 randomly selected fields in three replicate dishes were scored for immunoreactivity to the markers indicated and divided by the total cell number in each field determined by Hoechst nuclear staining. The data presented are the overall means for each treatment group±SEM across the three replicates. P values are based on results of analysis of variance, followed by pairwise Student's t-test.
% of Peri+ also Brn3a+
Induced Cells Respond to Dorsal–Ventral Patterning Factors
Because forced expression of Neurog1 in ESC is sufficient to promote both dorsal and ventral neural phenotypes, we wished to determine if the resulting neuronal cells could be patterned by exogenous factors. We cultured N7 cells in defined medium with dox and RA for the first 3 days and then changed the culture media to a serum replacement medium (SRM) with RA but not dox, to enhance cell survival. To test the effects of exogenous factors, media were supplemented with Sonic hedgehog (Shh), bone morphogenetic protein-4 (BMP4), or Noggin (Nog) recombinant proteins from the beginning of each experiment. RT-PCR analysis indicated that Shh treatment induced expression of the ventral marker, Nkx6.1, and suppressed expression of the dorsal marker, Pax7 (Fig. 6A). Exposure of cells to either high (25 nM) or low (2.5 nM) concentrations of Shh, resulted in a dose-dependent expression of ventral markers. On day 1, cultures grown in either high or low levels of Shh expressed higher levels of Nkx6.1 compared with those without Shh. At day 3, Nkx6.1 expression levels were similar between Shh-treated and untreated cultures, but by day 5, cells grown in the high Shh condition expressed higher levels of Nkx6.1. Expression of the intermediate dorsal–ventral marker, Dbx2, was reduced by treatment with Shh in a dose-dependent manner. Treatment with Shh also resulted in the expression of Islet1, which is associated with motor neurons in the ventral spinal cord, and of nestin, expressed by neural ectoderm and neural stem cells.
Treatment with BMP4 (5 nM) drastically reduced CNS neural differentiation as evidenced by down-regulation of all three CNS neural progenitor markers (Pax7, Dbx2, and Nkx6.1) (Fig. 6A). RT-PCR analysis confirmed expression in BMP4 exposed cultures of GATA-4, Brachyury, and Claudin6, markers of endoderm, mesoderm, and epidermal ectoderm respectively (not shown). BMP4-treated cultures also showed a marked increase in the expression of Snail, a marker of EMT and early neural crest cells and of peripherin, expressed in the PNS and motor neurons (Fig. 6A), although we did not observe many overtly neuronal cells in these cultures. Taken together, these data support the recognized role of BMPs as neural antagonists in the early gastrula stage embryo as well as during the early stages of mouse ESC differentiation. However, the expression of neural crest and PNS markers suggests that, in the context of forced Neurog1 expression, BMPs can act instructively to promote neural crest and PNS phenotypes (e.g., Aihara et al., 2010).
To address the possibility that endogenous BMPs might influence the neuronal differentiation observed in our cultures, we treated cell cultures with low (2 nM) or high (5 nM) levels of the BMP antagonist, Noggin (Nog), in combination with low or high Shh treatment. Consistent with previous observations, Shh-treated cultures expressed low levels of the dorsal marker, Pax7 (Fig. 6B). Of interest, Pax7 expression in Nog-treated cultures was higher than untreated controls. However, as we observed that Nog treatment appeared to enhance neuronal differentiation and/or survival such that there were many more neurons and primitive neural “rosettes” present in Nog-treated groups compared either with controls or cultures treated with Shh alone (not shown). Because the enhanced Pax7 signal could be due to the increase in neuronal differentiation, we examined the expression of Nkx6.1 and Pax7 using immunohistochemistry (Fig. 7) after 5 days of differentiation. Very few cells co-expressed Pax7 and Nkx6.1, indicating that the cultures contained a mixed population of precursor cells that could be characterized as distinctly "dorsal" (i.e., Pax7+) or "ventral" (i.e., Nkx6.1+). After quantifying the Pax7 and Nkx6.1-positive populations, we then applied the Tuj1 antibody to assess overall neuronal differentiation and found little difference between the treatment groups. Moreover, the Tuj1+ neurons were generally neither Pax7+ nor Nkx6.1+, further confirming that these markers labeled precursors rather than more mature neurons (Fig. 7, right panel). In cultures not treated with Shh or Nog, 46.1% (± 1.0) of cells were Pax7 positive, while 14.4% (± 0.4) were Nkx6.1 positive (Fig. 8). Treatment with either Shh or Nog at low or high levels resulted in a significant reduction in Pax7 immunoreactivity and a corresponding increase in the number of Nkx6.1-positive cells. We also observed more ventral phenotypes in Shh-treated groups compared with Nog treatment, and the ventralizing effect appeared to be dose-dependent. Of interest, co-application of Shh and Nog at either dose produced more ventral phenotypes than either factor alone, suggesting that these two molecules may act synergistically to promote ventral phenotypes. Altogether, the data suggest that forced expression of Neurog1 in ESC produces a population of neuronal precursors that remain responsive to extrinsic patterning signals.
Neural Induction by Means of Forced Neurog1 Expression Can Be Influenced by Fibroblast Growth Factor Signaling
The role of fibroblast growth factor (Fgf) signaling in neural induction and ESC differentiation is unresolved, with some studies suggesting that Fgf signaling is indispensable in neural induction both in vivo (Streit et al., 2000) and in ESC in vitro (Ying et al., 2003a). To assess the role of Fgf signaling during neuronal differentiation of ESC in the context of Neurog1 expression, we exposed N7 cells in defined medium to a 12-hr pulse of 5 nM SU5402, which targets the kinase domain of FGFR1 and therefore abrogates all signaling by this receptor (Mohammadi et al., 1997). We treated cells at three intervals: 12 hr before the addition of dox (pre-dox), concurrent with the addition of dox (simultaneous), and 12 hr following the addition of dox (post-dox); the control group received dox 12 hr after plating (Fig. 9A). Cells were then cultured for 72 hr in defined neural medium containing dox then assayed for the expression of Sox3, a marker of neural ectoderm and neuronal precursors, TuJ1 labeling early neurons, or Oct4, expressed in undifferentiated ESC. Media changes and monitoring by phase microscopy occurred at 12, 24, 36, and 48 hr at which time we observed no differences in cell death (as evidenced by nonadherent cells observed in the plates and trypan blue staining of withdrawn media) across any of the treatment groups compared with controls. Consistent with either a requirement for Fgf/ERK signaling for ESC differentiation (Kunath et al., 2007), or alternatively, that Fgf/ERK signaling prevents de-differentiation to a more primitive ES-like state (Greber et al., 2010), there were considerably more Oct4-expressing cells (53.4±5.4%) in the pre-dox group (Fig. 9B). Neuronal differentiation as evidenced by Tuj1 immunoreactivity was strikingly reduced in the pre-dox group (16.7±6.2%) and the number of Sox3+ precursors (66.3±7.4%) was higher compared with the other treatment groups. Simultaneous inhibition of Fgf signaling with Neurog1 induction resulted in many fewer Oct4+ ESC (18.4±2.3%), and slightly more TuJ1+ neurons (28.1±1.9%) compared with the pre-dox group. The number of Sox3+ precursors was slightly reduced in the simultaneous group (63.5±6.6%) compared with the pre-dox group, but the difference only approached statistical significance (P=0.06). In the post-dox group, very few Oct4+ ESC remained (5.5±3.5%). There was also a slight, but statistically significant (P≤0.05), reduction in the number of Sox3+ precursors (58.8±3.9%) in the post-dox group compared with either the pre-dox or simultaneous treatment groups, and a concomitant increase in the number of mature neurons (40.5±6.2%). Controls receiving no SU5402 showed a dramatic overall increase in differentiation as there were few Oct4+ cells remaining (0.7±1.0%), and a significant reduction in the number of Sox3+ precursors (35.9±6.4%). The number of TuJ1+ neurons in the control group was much higher than any of the treatment groups (64.6±4.7%), suggesting that the reduction in Sox3+ cell numbers was likely the result of increased differentiation of the Sox3+ precursor pool into (Tuj1+/Sox3−) neurons. Overall, these data indicate that the induction of Sox3+ neuronal precursors in the context of Neurog1 expression is largely Fgf-independent, but there is a temporal window within which neuronal differentiation can be held in check by Fgf-mediated persistence of the Oct4 pluripotency pathway.
After 24 hr of transgene induction, 1,384 genes were significantly differentially expressed; 951 up-regulated. At 48 hr, 5,894 genes were significantly altered (2,856 up-regulated); increasing to 7,205 (3,263 up-regulated) after 72 hr of doxycycline exposure. Gene ontology classification analysis identified significant alterations in the expression of genes associated with: nervous system development (48%), development (24%), metabolism (4%), cell cycle (4%), signaling (4%), and cellular organization (4%). Supplementary Table S1, which is available online, summarizes the results of functional annotation clustering of genes significantly up- and down-regulated at 24, 48, and 72 hr of transgene induction. There were significant changes in the expression of genes involved in neurogenesis, in process outgrowth and cell migration, those involved in the regional development of the nervous system, in cell cycle, as well as genes involved in ESC homeostasis.
We carried out KEGG pathway analysis to map alterations in signaling pathways and molecular functions and identified sequential induction of genes in pathways involved in axon guidance, cancer and melanogenesis (Supp. Table S2). By 72 hr of transgene induction, 19 pathways were activated, including: ErbB, Neurotrophin, Wnt, Mapk, Focal adhesion, Calcium, and Insulin signaling pathways. After 72 hr, only one pathway, cysteine and methionine metabolism, was significantly down-regulated.
Consistent with our RT-PCR results, Neurog1 was up-regulated 74-, 74-, and 42.7-fold at 24-hr intervals compared with uninduced cells. Other bHLH genes were also induced: Neurod4 increased by 33-, 206-, and 127.8-fold; Neurod1 increased 9.7-, 11.5-, and 10.2-fold, while Neurog2 was up-regulated by: 0-, 4.4-, and 3.8-fold (Table 2). Pluripotency factors expressed by undifferentiated ESC: Eed, Eras, Fbxo15, Foxd3, Klf2, Klf4, Nanog, Pou5f1 (Oct4), Sox2, Zic3, were significantly down-regulated with differentiation. Genes associated with neurogenesis including: Ebf2, Ebf3, Elavl3, Elavl4, Fabp7, Hes5, Lhx2, Ncam1, Nhlh1, Nhlh4, Pax3, Pax6, Reln, Zic1, were induced, while the negative regulator of neurogenesis Rest (RE1-silencing factor), was down-regulated. Many genes expressed in the ventricular zone (Fu et al., 2009) and by neural stem cells including Dll1, Dcx, Prom1, Reln, Coup-TFI and II (Naka et al., 2008), and Hes5 (Hatakeyama et al., 2004; Basak and Taylor, 2007) were expressed, as were genes associated with cell cycle exit and differentiation including Ebf2,3 (Pozzoli et al., 2001; Garcia-Dominguez et al., 2003). Id4, Pax3, Pax6, Sox3, Sox4, and Sox11 (Bergsland et al., 2006) were also increased. Factors involved in process outgrowth and cell migration including ephrins and their Eph receptors, semaphorins and plexins, were induced by Neurog1. Of interest, epithelial genes including junctional proteins and cell–cell adhesion molecules such as E-Cadherin (Cdh1) were down-regulated whereas N-Cadherin (Cdh2) expression significantly increased.
Fold changes in gene expression following 24 hr, 48 hr, and 72 hr of transgene induction.
There were also significant increases in the expression of genes associated with regional neuronal cell types, including telencephalon-associated genes Elavl4, Fabp7, Foxd1 and Foxg1. Lhx2, which plays a critical role in cortical patterning (Mangale et al., 2008; Chou et al., 2009) was particularly strongly induced by Neurogenin1. Several factors associated with ventral neural fates, e.g., striatal (Foxp2), midbrain (Pcdh8), and midbrain dopaminergic neurons (Foxa2, Ebf3) were also induced. Hindbrain markers Hoxa2 and Hoxb2 were significantly induced, as was Hoxd9, which is expressed in thoracic lateral motor neuron columns (Dasen et al., 2005). Other genes associated with a motor neuron phenotype such as Irex3 and Isl1 were also significantly increased. Genes expressed in cephalic placodes including Eya1, Ebf2, Fbxo2, Six1, Netrin1 and its receptors DCC and Unc5C were up-regulated as were genes associated with neural crest cells, including Ednrb and Ret. Few genes associated with glial differentiation (Fu et al., 2009) were identified, and genes that inhibit oligodendrocyte or astrocyte differentiation, e.g., Hes5 (Liu et al., 2006) and Id4 (Marin-Husstege et al., 2006) were induced.
Activation or silencing of lineage-specific sets of genes during development is controlled by transcription factors as well as by epigenetic regulators of chromatin structure. Because ESC have a chromatin configuration that is “poised” for lineage differentiation (e.g., Boyer et al., 2006), selective induction of chromatin regulators associated with other lineages, or removal of repressive marks on neuronal promoters/enhancers could drive neuronal differentiation. We examined the expression of transcripts encoding proteins involved in chromatin modification including SWI/SNF, polycomb family members, HDACs, etc., and found minimal (< three-fold) changes. There were three exceptions that are of potential interest. The first is in Forkhead box (Fox) gene expression. Foxa proteins are considered “pioneering” in that they open chromatin to allow modifiers access to chromatin, Foxp factors function as classic transcription factors recruiting enzymes to regulate gene expression, while Foxo proteins appear to do both (Lalmansingh et al., 2012). Following Neurog1 expression, Foxa2 was increased 21.2- and 13.7-fold at 48 and 72 hr, and Foxp2 was increased 62.4- and 53.8-fold at similar time points, suggesting a role in neuronal lineage differentiation. Second, Phc2, a polycomb group member expressed in germinal zones of the nervous system (Kim et al., 2005), was induced 7.1-, 8.9-, and 8.3-fold. Finally, the histone methyltransferase, Suv39h1 was down-regulated 35.7- and 17.6-fold at 48 and 72 hr. Suv39h proteins have been shown to oppose Ring1 to coordinate early lineage decisions in the blastocyst (Alder et al., 2010), and to interact with Smads to coordinate BMP-induced gene repression and lineage differentiation (Frontelo et al., 2004). Somewhat surprisingly, given a recent study demonstrating that Ezh2 is required to inhibit transcription of nonmuscle lineage genes in satellite cells (Juan et al., 2011), we did not observe alterations in either the ubiquitous Ezh1 or in Ezh2 associated with proliferating tissues (Margueron et al., 2008). Future interrogation using ChIP and knock-down will explore these possibilities more directly.
Microarray analysis also identified a complex pattern of expression of genes involved in cell cycle control. At 24 and 48 hr, cluster analysis did not identify any significant clusters containing cell cycle related genes, but by 72 hr of transgene induction, three clusters identified down-regulated genes involved in cell cycle. Cluster 3 (Enrichment Score=7, P≤5.0 × 10−9, Benjamini-Hochbert corrected) contained genes involved in cell cycle progression, Cluster 4 (Enrichment Score=5.5, P≤3.1 × 10−9, corrected) identified genes involved in chromosome organization, sister chromatid, and centromere structure, while Cluster 7 (Enrichment Score=4.9, P≤2.3 × 10−6, corrected) identified genes involved in chromosome organization (Supp. Table S2).
Because the ESC population is relatively homogeneous (compared with brain tissue) it is possible to examine the sequential expression/repression of transcripts involved in cycle regulation (Table 2). Positive regulators of cell cycle including: Birc5, Id3, Skp2 were down-regulated, and negative regulators: Btg2, Ccng2, Ebf2,3, Gadd45g, Gspt1, Hipk2, Prmt2 were up-regulated with transgene expression. However, other positive regulators were stimulated, including: Ccnd1, Ccnd2, Cdc25b, Cyr61, Gpc1, Hmga2, Tead2. Others were strongly changed initially by transgene induction but then were reduced by 48 or 72 hr, including: Cdc25b, Gadd45g, and Id3. Of interest, the tumor suppressor Trp53/p53 was down-regulated significantly both at 48 and 72 hr of induction. P53 plays major roles in apoptosis, cell cycle arrest, and by means of its downstream targets Gadd45g and p21 (Cdkn1a) which is up-regulated at all time points, which functions as a G1 regulator to produce growth arrest.
We also compared gene expression in Neurog1-expressing cells at 24 vs. 48, 48 vs. 72 and 24 vs. 72 hr time points. Overall, microarray analysis of changes in gene expression between 24 and 48 hr identified sets of genes expressed in neural stem cells, neural crest and placodes. Comparing 24 and 72 hr patterns, increasing numbers of mature markers including channel genes, neurotransmitter/receptor genes were identified, consistent with our previous observations that they exhibit mature axon potentials (Reyes et al., 2008).
To validate the microarray data, we chose to examine Notch-Delta pathway members, as they have previously been suggested to be targets of Neurog1 (Ma et al., 1997). As early as 12 hr after doxycycline treatment, expression of Notch pathway members was strikingly up-regulated, remaining high during the 72 hr culture period. Over time in control cultures there was a gradual increase in expression of pathway members, Notch1, Jag1, and Hes1 as neuronal differentiation increased (Fig. 10). Consistent with these observations, at 24 hr, array analysis identified Notch2 (three-fold), Hes5 (21) and Hes6 (17.9) as up-regulated (Table 2). By 48 hr, many pathway members including Notch1 (2.8), Notch2 (2.7), Notch4 (-4), Jag1 (6.2), Hes5 (384), Hes6 (7.5), Dner (12) were identified. At 72 h, Jag1 (3.8) and Hes5 (203.2), Hes6 (2.6), and Dner (23) were up-regulated.
Microarray data have been deposited in GEO, accession numbers: GSE42883, GSM1052734-GSM1052745.
Inducible Expression of Neurog1 in ESC Provides a Novel Model of Neurogenesis
ESC provide a powerful model system to study the mechanisms of cell fate specification and lineage segregation at a stage in early development not readily accessible for manipulation. Studies of directed ESC differentiation typically rely on differentiation in embryoid bodies and/or exposure to stromal cell derived factors or to nonspecific morphogens such as RA. While these approaches can produce populations enriched with neurons (e.g., Zeng et al., 2011), undefined factors and endogenous cytokines present in EBs, media containing serum, or stromal cell co-culture, complicate efforts to understand the role of specific signaling pathways in lineage segregation. Differentiation of ESC in monolayer cultures in defined media minimizes these effects (e.g., Ying et al., 2003b) and, if combined with forced expression of a lineage restricted gene, presents a unique opportunity to tease out the interactions of particular genes and growth factors in early development. An additional advantage of the inducible gene expression paradigm is the ability to maintain transgene expression following implantation (Reyes et al., 2008). Here, we demonstrate that expression of Neurog1 in ESC is sufficient to induce neuronal differentiation even in conditions that normally maintain pluripotency (adherent culture at high density with serum and LIF), in defined media where the presence of morphogens can be more precisely controlled.
Neurog1 Expression in ESC Generates Representative Neuronal Subtypes and Neural Crest Derivatives
Consistent with the pattern of Neurog1 expression in defined domains of both the CNS and PNS (Ma et al., 1996; Sommer et al., 1996; Murray et al., 2000; Kim et al., 2011; Takano-Maruyama et al., 2012), expression of Neurog1 in ESC is sufficient to generate neuronal precursors and mature cell types representative of these expression domains (Figs. 4, 5). In fact, this paradigm may represent a reliable method to obtain sensory neurons, which have been difficult to generate from embryonic stem cells (Sasai, 2005). We observed a significant number of neurons that expressed Brn3a, a POU homeodomain transcription factor associated with sensory lineages in both the CNS and PNS. In the CNS, Brn3a+ sensory lineages arise throughout the anterior–posterior extent of the dorsal alar plate in the neural tube. In the posterior spinal cord, the alar plate generates Brn3a+ dorsal interneurons, of which the dI2 commissural lineage is specifically Neurog1-dependent (Gowan et al., 2001), whereas rostral domains of Brn3a and Isl1 expression are present in the tectum and tegmentum of the midbrain (Fedstova and Turner, 2001). A 4 kb enhancer has been shown to drive reporter gene expression to the midbrain, hindbrain and spinal cord (Nakada et al., 2004), while a second 0.8 kb enhancer drives expression to cells fated to become interneurons in the ventral spinal cord (Quinones et al., 2010). The presence of peripherin+ cells in our cultures presents the intriguing possibility of having generated motor neurons which express this marker (Escurat, et al., 1990). However, these cells almost invariably co-expressed Brn3a, which has not been reported in motor neurons, but is more suggestive of PNS lineages. Regarding glial differentiation, we did not observe any up-regulation in glial markers in either our microarray analyses (Table 2) or by RT-PCR or immunohistochemistry (not shown). Moreover, our previous work similarily failed to detect GFAP+ cells in vivo or in vitro (Reyes et al., 2008), consistent with observations that Neurog1 promotes neurogenesis while inhibiting gliogenesis (Sun et al., 2001) and detailed fate mapping studies in which Neurog1 was found to be exclusively associated with neuronal lineages (Kim et al., 2011).
In the PNS, Neurog1 has been shown to be instructive for sensory neurogenesis (Perez et al., 1999; Lo et al., 2002) and required for the generation of proximal cranial sensory ganglia and TrkA neurons of dorsal root ganglia (Ma et al., 1997), whereas Atoh1 is associated with development of adrenergic autonomic ganglia (Guillemot et al., 1993). Correspondingly, we failed to detect tyrosine hydroxylase, a marker of adrenergic neurons in any of our cultures (not shown) but instead found many peripherin+ neurons that invariably co-expressed Brn3a (Fig. 5E,F), a phenotype representative of both cranial and dorsal root sensory ganglia (Gowan et al., 2001). A possible caveat to this interpretation is that Brn3a (Nadal-Nicolas et al., 2009) and peripherin (Tajika et al., 2004) are also co-expressed in retinal ganglion cells, but other markers of retinal ganglion cells such as Chx10, Dlx1, Dlx2, and Brn3b (de Melo, et al., 2003) were not significantly up-regulated in our microarray analysis, making this phenotype less likely.
Derivation of neural crest from ESC typically is a multistep process that requires sequential exposure to multiple growth and differentiation factors (Mizuseki et al., 2003; Aihara et al., 2010; Kawaguchi et al., 2010). Our findings suggest that Neurog1 expression can direct ESC toward Ret+, Ednrb+ neural crest fates, and induces expression of Sox4 and Sox11 required in early sympathetic ganglia (Potzer et al., 2010), in the absence of exogenous instructive signals. However, our cultures contained a mixture of CNS and PNS types, and it remains to be determined what factors in addition to Neurog1 expression mediate the choice between these lineages.
ESC Expressing Neurog1 Can be Patterned Along the Anterior–Posterior and Dorsal–Ventral Axes and Influenced by Fgf Signaling
Expression of Neurog1 prompted early expression of forebrain (Otx2, Hoxa1) as well as midbrain (Engrailed1, not shown) markers, strong induction of Hoxb2 but not of more posterior markers including Hoxc6. However, treatment of these cultures with RA promoted expression of Hoxc6 and down-regulation of anterior markers. These data suggest that these cells possess an intrinsic anterior character that can be posteriorized by RA- mediated signals (Irioka et al., 2005). Because these cultures lack other cell types, these results clearly demonstrate that Neurog1-expressing neuronal precursors possess an anterior character in the absence of confounding “mesendodermal” signals.
Based on previous gain-of-function studies, we expected that Neurog1 expression in ESC would produce specific subtypes of neurons independent of any endogenous patterning molecules. However, we were surprised to find that induced cells were quite responsive to dorsal–ventral (DV) patterning molecules (Wilson and Maden, 2005) such that more ventral phenotypes were observed in cultures treated with Shh, and this effect could be augmented by addition of Noggin protein. RA treatment alone also produced ventral phenotypes. In addition, noggin treatment accelerated and BMP treatment suppressed neurogenesis in our cultures. Together, these data indicate that ESC forced to express Neurog1 progress through a stage where they remain receptive to patterning cues.
Recent evidence indicates that FGFs may have an independent and indispensable role in the induction of neural fate in vivo (Streit et al., 2000; Wilson et al., 2000; Lanner and Rossant, 2010) and in the neuronal differentiation of ESC (Ying et al., 2003a). Fgf signaling can maintain self-renewal of ESC (Sterneckert et al., 2010; Staviridis et al., 2010), and is required for cell cycle exit and the transition to lineage differentiation (Kunath et al., 2007). Other studies have demonstrated that Fgf signaling is essential for the progression of ESC to an epiblast state (Stavridis et al., 2010); preventing de-differentiation to a more primitive ICM-like state (Greber et al., 2010) and inhibiting neural differentiation (Greber et al., 2010; Jaeger et al., 2011). Our data suggest that expression of Neurog1 in ESC is sufficient to generate Sox3+ progenitors independent of Fgf signaling (Fig. 9B). However, the reduction in the percentage of neurons in the preinduction and simultaneous induction treatment groups, and striking increase in the number of Oct4+ and Sox3+ cells in the earlier treatment groups, suggests that the initial specification of Sox3+ precursors in the context of Neurog1 expression does not require Fgf signaling. However, there is a window where inhibition of Fgf signaling interferes with the ability of these precursors to differentiate into mature neurons. It has been suggested that inhibition of Fgf signaling can revert epiblast stem cells to an Oct4+ ESC state (Greber et al., 2010). However, we failed to detect up-regulation of epiblast markers such as Fgf5, but we did see markers of neural ectoderm such as Pax 6 (Table 2), suggesting that neuronal differentiation by means of Neurog1 expression likely bypasses an epiblast state. Nonetheless, the Oct4 pathway appears to be initially intact in these cells and abrogation of Fgf signaling re-engages ESC programs that inhibit neuronal differentiation.
ESC Neuralized by Neurog1 Expression May Progress Through a Notch-Sensitive Neuronal Precursor Stage
Although our immunohistochemical analyses indicated that all of the cells in our cultures express Neurog1 (Fig. 2), not all matured into neurons after 3 or even 5 days in culture, similar to previous observations in embryonal carcinoma (EC) (Farah et al., 2000; Kim et al., 2004) and ESC (Kanda et al., 2004). Although primitive neural stem cells appear relatively insensitive to Notch signaling, definitive neural stem cells may be maintained in an undifferentiated state by activation of the Notch pathway (Tropepe et al., 2001), like neuronal precursors, whose differentiation into mature neurons can be held in check by the Notch-Delta pathway (reviewed in Louvi and Artavanis-Tsakonas, 2006). We observed that induction of Neurog1 results in transient expression of the Notch ligand, Jag1 (Fig. 10; Table 2), and of Dll1,3,4 (Table 2), as well as up-regulation of Hes1, Hes5, and Hes6, repressor-type bHLH factors that are downstream targets of the Notch signaling pathway. The Notch target gene Hes5, required for NSC maintenance (Basak and Taylor, 2007; Hatakeyama et al., 2004), was strongly induced by Neurog1. In fact, the up-regulation of these repressor-type bHLHs is more likely the effect of Jag1 expression in neighboring cells than a direct effect of Neurog1 expression. Thus, ESC expressing Neurog1 may progress through a Notch-sensitive stage in which cells at the edge of rosettes escape inhibition to differentiate into neurons, and then repress neuronal differentiation in neighboring cells.
Gene Expression Patterns Down-stream of Neurog1
Because Neurog1 is expressed early in the neurogenic cascade at the time of lineage commitment/determination (Blader et al., 1997), we used microarray analysis to identify transcripts expressed down-stream of Neurog1. Many of the genes strongly induced by Neurog1 are expressed in the stratifying neural ectoderm at the time of cell cycle exit (Fu et al., 2009); coupling Neurogenin1 expression with cell cycle exit and neuronal differentiation (Kim et al., 2011). Other candidate transcription factors such as Sox4 and Sox11, which are known to play a role in the downstream effects of proneural bHLH factors, were strongly induced, while genes associated with ESC maintenance, and inhibitors of neuronal differentiation were down-regulated. In addition, genes associated with astrocyte and oligodendrocyte differentiation were not identified or were down-regulated.
Genes associated with neuronal cell migration were strongly induced by Neurogenin1, including the microtubule associated protein doublecortin (Dcx), Cdn2, reelin, CoupTF factors (Nr2f1, Nr2f2; Tripodi et al., 2004), as well as EphA5 previously identified as down-regulated in Neurog2:Neurog1 double mutant mice (Mattar et al., 2004). There was also a striking switch in expression of genes associated with an epithelial morphology including junctional proteins, as well as E-Cadherin (by ESC) to N-Cadherin by differentiating neurons. These data suggest that Neurog1 expression activates a program of neurogenesis that includes cell surface changes, induction of genes involved in cell migrations, transcription factors and cell cycle modulators.
In a screen of Ngnr1 and NeuroD targets in Xenopus, Seo et al. (2007) identified similar sets of target genes: 47% of the genes induced by Ngnr1 over-expression in Xenopus were identified in our analysis, while 43% of the genes induced by NeuroD were present in our screen. Somewhat surprisingly, all of the seven core regulatory factors suggested to promote Neurogenin-mediated neurogenesis: Ebf2,3, Hes6, Myt1, Nhlh1, Neurod1, Neurod4, and Runx1t1 (Pang et al., 2011) were significantly increased in our cells. An in silico analysis to identify targets of Neurog2 (dorsal telencephalon) vs. Ascl1 (ventral telencephalon) (Gohlke et al., 2008) identified only Dcx, Elavl4, Nrarp, in common with our Neurog1 down-stream targets; and one, Fzd5, was down-regulated significantly in our analysis. Of interest, constitutive over-expression of NEUROG1 in a cell line derived from human fetal telencephalon identified 588 potential target genes; 27.7% of them were identified in this analysis at 72 hr. These represent a very different population as telencephalon progenitor cells expressed transcripts associated with glial cells as well as with neurons (Satoh et al., 2010). Similarly, expression of transcripts in the hindbrain neural ectoderm on E11.5—when Purkinje progenitors are developing—in Neurog1−/− embryos identified 31/117 transcripts also present in our analysis. These data suggest that ultimately, subtype specificity is likely controlled by combinations of bHLH proteins and region-restricted transcription factors (particularly homeodomain proteins), which in combination with noncoding RNAs, may regulate each other or stabilize a neurogenic program.
Other candidates that may act in concert with bHLH factors to restrict lineage progression are chromatin regulators. The ESC faces the unique problem of maintaining both self-renewal and pluripotency, which it may solve with a unique bivalent chromatin structure in which enhancers of lineage-specific genes are associated with specific methylation marks that repress lineage gene expression “poised chromatin”, while pluripotency genes maintain an “active” chromatin through successive rounds of division. These modifications are lost with differentiation (Boyer et al., 2006), but many are re-set with somatic cell reprogramming (Creyghton et al., 2010). In fact, it has been suggested that because many lineage-specific transcription factors belong to large families that recognize a common binding sequence, e.g., the E box, additional mechanisms, such as chromatin modification, must exist to limit the sites available for transcription factor binding (Conerly et al., 2011). Our microarray analysis identifies several transcripts encoding chromatin modifying factors as possible targets of Neurog1. These factors may function to restrict binding of non-neuronal lineage factors during Neurog1 driven differentiation, and may ultimately suggest methods for direct reprogramming of adult cells to neurons. Additional work will examine direct interactions and use shRNA to probe their roles in the context of Neurog1 expression.
Neurog1 Alters Cell Cycle Progression
Although the proneural bHLH genes Atoh1 (Math1), Asc1 (Mash1), Neurog1/2 promote cell cycle exit and expression of neuronal genes, the mechanisms involved and target genes are largely unknown. Recently two large-scale analyses of gene expression downstream of Ascl1 (Castro et al., 2011) and Atoh1 (Machold et al., 2011) identified transcripts involved in cell cycle exit and sequential phases of neural differentiation. While it is not surprising that proliferating granule cell precursors maintained Atoh1 and cell cycle genes, in the current investigation we also identified genes involved in positive regulation of the cell cycle. Of the 95 cycle-related genes expressed by Mash1+ rhombic lip precursors, 17 were also present in Neurog1 over-expressing ESC, including: Aurka, Birc5, Brca2, Bub1b, Ccnd1, Ccnd3, Cdc20, Cdca2, Cdca5, Cdt1, E2f8, Fbxo5, Gmnn, Hells, Jub, Nek2, Trpt53/p53. Downstream of Ascl1 (Castro et al., 2011) only five were genes were common to Neurog1 cells: Ccnd1, Ccd3, Cdkn1a/p21, Plk1, Gadd45g.
Gadd45g, a strong inhibitor of cell cycle progression (e.g., Mak and Kultz, 2004) was strongly induced by Neurog1 at 24 hr. Ebf3 was particularly strongly up-regulated by Neurog1 (48, 162, 182-fold at 24-hr intervals); it has previously been reported to be downstream of NeuroD (Pozzoli et al., 2001). Ebf genes inhibit cell cycle progression at G1/S, and couple cycle exit and neural differentiation (Garcia-Domingues, 2003); over-expression decreases proliferation consistent with our observations. Neuron-specific activators of the cell cycle progression inhibitor Cdk5 (Cdk5r1/p35 and Cdk5r2/p39), which are responsible for producing the unique neuronal cell cycle, were also increased. Other genes such as E2f8, which is induced at G1/S and acts as a repressor of cycle progression by means of p53 (Christensen et al., 2005), were down-regulated in Neurogenin1-expressing cells. Cdc20, which has previously been associated with increased neural plasticity (Conway et al., 2007), was also decreased in N7 cells following transgene induction. This work demonstrates that the neurogenic bHLH genes affect transcripts involved in cell cycle progression, neural differentiation, migration and apoptosis, and may reflect the presence of a population of proliferative progenitors in the Neurog1-expressing cells.
ESC, like the early epiblast, spend most of their time in S phase and have an attenuated G1 (Burdon et al., 2002; White et al., 2005), possibly to minimize exposure to differentiation factors present in the local microenvironment (Orford and Scadden, 2008). Lengthening of G1, which occurs with differentiation of the epiblast and ESC and is characteristic of somatic cells, was also observed following expression of Neurogenin1. Whether a cell exits cycle is determined at the G1 checkpoint, and exit from cycle is required for many cell fate decisions, consistent with the presence of a population of proliferative precursors in these cultures.
The ability to induce expression of Neurog1 in ESC constitutes a straightforward and powerful model system that provides insight into the complex interplay of signaling molecules and transcription factors that shape the two-cell embryo into the thousands of mature cell types present in the adult PNS and CNS, and increases our understanding and eventual control of normal development, birth defects and tumor formation. Ultimately, elucidation of the molecular histogenesis of the nervous system will also improve our understanding of the factors that promote ectopic neurogenesis in several neurodegenerative conditions (Curtis et al., 2003). In addition, primitive precursors such as those that can be generated by the N7 cell line may be useful for cell transplantation or as a resource to identify down-stream targets involved in activation of a program of neurogenesis.
Targeting Vector Construction and Generation of Inducible Neurog1 Cell Lines
A Neurog1 insert from pCS2-ND3 (M. McCormick) was ligated with an EcoRI-NotI fragment from pI2R (Clontech, Palo Alto, CA) containing an IRES-DsRed2 cassette into the pLox targeting vector to generate the targeting construct, pLox-N1-I2R. To generate targeted cell lines, 1.0 × 106 Ainv15 ESC (Kyba et al., 2002) were plated in a 60-mm dish and co-transfected 24 hr later with 1 μg each of pLox-N1-I2R and pSalk-Cre (M. Kyba) using Lipofectamine/Plus Reagent (Invitrogen) in serum-free DMEM. After 3 hr, the transfection medium was replaced with complete ES medium. The following day, the cells were split to a 150-mm dish in complete ES medium (see below) containing 350 μg/ml G418 (Invitrogen). After 10 days in selection, 12 colonies remained, from which we were able to establish 7 independent cell lines. Three of these lines were selected for additional analysis to confirm targeted integration of the Neurog1 transgene at the tet-inducible locus, stable expression of rtTA, doxycycline-inducible expression of Neurog1, and uniform neuronal differentiation after 1 and 3 days of culture in defined neural medium (described below) with 1 μg/ml doxycycline (Supp. Fig. S1). We observed no differences between the cell lines and we therefore selected one of the lines, N7, to carry out most subsequent experiments.
Cell Culture and Neural Differentiation
Ainv15 ESC were cultured on 0.1% gelatin coated substrates in complete ES medium (ESM) consisting of high-glucose DMEM (Invitrogen, Carlsbad, CA) supplemented with 10% ES-tested fetal bovine serum (FBS) (Atlanta Biologicals, Norcross, GA, or Harlan Bioproducts, Indianapolis, IN), 10−4 M β-mercaptoethanol (Sigma, St. Louis, MO), 0.224 μg/ml L-glutamine (Invitrogen), 1.33 μg/ml HEPES (Invitrogen), and 1,000 units/ml human recombinant LIF (Oncogene Research Products, San Diego, CA). N7 cells were maintained in ESM plus 350 μg/ml G418 (Invitrogen) and 1.5 μg/ml Puromycin (Sigma). For neural differentiation, cells were plated at 5.0 × 104 cells/cm2 in gelatin-coated 6- or 12-well plates in ESM or in defined medium supplemented with 1 μg/ml doxycycline (Sigma), and media changed every day. Defined neural medium consisted of 80% F-12/DMEM and 20% Neurobasal media with 10 μl/ml MEM pyruvate, 8 μl/ml N2 supplement, and 4 μl/ml RA-free B27 supplement (all media components from Invitrogen). Where indicated, media were supplemented with 1 μM all-trans-retinoic acid (Sigma). For long-term culture, defined medium was replaced after 3 days with doxycycline-free serum replacement medium (SRM5) consisting of high glucose DMEM supplemented with 5% knockout serum replacer (Invitrogen), 10−4 M β-mercaptoethanol, 0.224 μg/ml L-glutamine, and 1.33 μg/ml HEPES to enhance cell survival. Where indicated, recombinant growth factors (all from R&D Systems, Minneapolis, MN) were added to freshly prepared media at the following concentrations: 25 nM (high) or 2.5 nM (low) mouse Shh; 5 nM human BMP4; and 5 nM (high) or 2 nM (low) Noggin-Fc chimera. To examine the role of FGF signaling, cultures were exposed at intervals to 5 nM SU5402 (Calbiochem) for 12 hr to block FGFR1 signaling.
Proliferation and Cell Cycle Analysis
To analyze the effects of transgene induction on proliferation, Ainv15 and N7 ESC were plated at 1 × 105 cells/well in 12-well plates in triplicate (with four biological replicates) in complete ES medium±2 μg/ml doxycycline. After 24, 48, and 72 hr, cells were disaggregated using trypsin and counted in a Coulter counter (Becton Dickinson).
For cell cycle analysis, cells were plated in triplicate at 2 × 105 in six-well plates in complete ES medium±2 μg/ml doxycycline. After 72 hr, cells were fixed in cold absolute ethanol, suspended in PBS and DNAse free RNase A (Sigma) at a final concentration of 20 μg/ml. Samples were incubated at 37° for 30 min and DNA labeled with 0.1 mg/ml propidium iodide (Sigma). Analysis was done in a FACS Calibur using Cell Quest Pro MAC 9.0 and cell cycle analysis done with Mod Fit LT Mac 3.1 SP3.
RNAs were extracted from lysates of cells treated with TRIzol reagent (Invitrogen), DNAsed, and quantified by spectrophotometry. Control RNA samples were obtained from N7 cells cultured in ES medium and from pooled samples of gestation day 9, 10, and 11 mouse embryos. RNAs (1 μg each) served as templates in reverse transcription reactions with oligo-dT primers, and 1/20 of the single-strand cDNA products were used in each PCR amplification. General PCR conditions were 94°C/3 min, 94°C/30 s, 53–63°C/1 min, 72°C/1 min for 25–40 cycles. Specific information regarding PCR primer sequences and reaction conditions is provided in the on-line supplementary material. The PCR products (10 μl each) were electrophoresed in 1.5% agarose gels in the presence of ethidium bromide, visualized on a ultraviolet transilluminator and digitally photographed using a Bio-Rad Gel Documentation system. Primers and conditions are summarized in Supplemental Table 3.
To examine cell type-specific and positional markers, cells were fixed in 2% paraformaldehyde for 15 min at 25°C, then stored in PBS at 4° before processing for immunohistochemistry (IHC). Briefly, cells were permeabilized and nonspecific antibody binding blocked using host serum before incubation with primary antibodies at 4° overnight. Cells were washed, then exposed to secondary antibody conjugated to Cy3 or FITC (Jackson ImmunoResearch Laboratories) for 30 min at 25°C. Nuclear staining for total cell number counts was achieved by incubating cells with 1 μM Hoechst 33285 (Sigma) for 2 min. Cells were washed, then mounted using Prolong Gold (Molecular Probes/Invitrogen) and 25 mm coverslips. Cells were examined, counted and photomicrographs imported into Photoshop using a Leica DMIRB inverted microscope and Olympus digital camera.
Primary antibodies included: Sox3 (Mike Klymkowsky; 1:1,000) expressed by in neural ectoderm and neuronal precursors (Zhang et al., 2003), Tuj1 (Covance MMS-435P; 1:250) which recognizes a neuron-specific class III β-tubulin expressed very early in neuronal differentiation and persisting in mature neurons (Moody et al., 1987), peripherin (7C5, AbCam ab4573, 1:1,000) which is expressed in the PNS, Brn3a (Eric Turner, 1:1,000) expressed in sensory neurons, tyrosine hydroxylase (Chemicon AB5968, 1:1,000) expressed in adrenergic neurons, Oct4 (Santa Cruz sc-8628, 1:100) expressed in undifferentiated ES cells. Dorsal–ventral markers included: Pax7 (Developmental Studies Hybridoma Bank supernatant, 1:20) labeling dorsal neural progenitors, Nkx6.1 (N15, Santa Cruz sc-15027, 1:100) expressed in ventral progenitors, and Islet 1 which recognizes Islet 1 and Islet 2 (Developmental Studies Hybridoma Bank, 39.4D5, 1:100) labeling ventral motor neurons and dorsal root (sensory) ganglia, or Dbx2 expressed in the middle zone of the neural tube (Agalliu and Schieren, 2009).
For quantitative analysis of the cell types present, cell counts were performed on 10 microscopic fields at ×20 from at least two replicate cultures, counting at least 1,000 cells per replicate. For each field, the number of a given cell type was divided by the total number of cells present as determined by Hoechst nuclear staining to calculate a percentage for each field which was then summed to obtain a mean percentage for each replicate. The data presented are the overall mean±SEM across all replicates. Data were analyzed using ANOVA, Students t-, or Chi-square test where appropriate (with Bonferroni correction as necessary for multiple comparisons); a P value of≤0.05 was considered significant.
RNAs were extracted as described above from triplicate samples (cells differentiated in three independent experiments) of control N7 cells (-Dox) at 24-hr culture intervals and N7 cells exposed to doxycycline for 24, 48, and 72 hr. RNA was hybridized to Affymetrix Mouse 430 2.0 arrays (Santa Clara, CA, http://www.affymetrix.com) by the NIH NINDS arraying core at TGen (http://www.tgen.org). The signal intensity of each array was normalized to a mean of 150 by robust multiarray averaging (RMA), then GeneSpring (Agilent) software was used to select genes with≥two-fold change between groups. At each time point, N7 cells (−Dox) were used as the reference for comparison with +Dox samples. Welch t-tests were then carried out to identify significant differences between controls and Neurogenin1-expressing cells. Gene expression was also compared in +Dox (and −Dox control) samples at sequential time points (e.g., 24 vs. 48 hr, 48 vs. 72 hr, and 24 vs. 72 hr of transgene induction). Benjamini-Hochbert false discovery corrections were performed on each comparison to reduce the number of false-positive results. Finally, Gene Ontology, Functional Annotation Clustering, and KEGG pathway analyses were carried out using DAVID V 6.7 (Huang et al., 2009).
The authors thank Drs. Sean Morrison, Ben Novitch, Pamela Raymond, and David Turner for advice with the work, Dr. Mary McCormick for the Neurog1 cDNA, Dr. Mike Klymkowski for the anti-Sox3 antibody, and Dr. Michael Kyba for constructs, Ainv15 cells and advice. The authors acknowledge the use of DAVID V. 6.7, the University of Michigan Comprehensive Cancer Center FACS Core facility, and antibodies from the Developmental Biology Hybridoma Bank. J.M.V. was funded by the NIH.