Comparison of the generic neuronal differentiation and neuron subtype specification functions of mammalian achaete-scute and atonal homologs in cultured neural progenitor cells

Proneural genes encode basic helix-loop-helix (bHLH) transcription factors that are essential for early steps in neurogenesis, in both Drosophila (Campuzano and Modolell, 1992) and vertebrates (reviewed by Guillemot, 1999). In Drosophila, proneural genes such as achaete-scute (ac-sc) and atonal (ato) play at least two roles. They select neural or neuronal precursors from a field of equivalent neuroepithelial cells, a function sometimes equated with promoting generic neuronal differentiation (Jarman and Ahmed, 1998), and they also specify neuronal subtype identity (Anderson and Jan, 1997; Hassan and Bellen, 2000). Thus, for example, in the fly peripheral nervous system (PNS), ac-sc specifies external sensory (ES) organ identity, while ato primarily specifies chordotonal (CD) organ identity (Jarman et al., 1993; Chien et al., 1996). However ATO can specify ES as well as CD identities, depending on context, while AC-SC exclusively promotes ES identity (Jarman et al., 1993; Jarman and Ahmed, 1998).

Vertebrate homologs of proneural genes, such as the ato-related neurogenins (NGNs) (Gradwohl et al., 1996; Ma et al., 1996; Sommer et al., 1996) and the ac-sc related Mash1 (Ascl1 – Mouse Genome Informatics) (Johnson et al., 1990; Guillemot and Joyner, 1993), are required for neurogenesis in distinct lineages (Guillemot et al., 1993; Fode et al., 1998; Ma et al., 1998) and are sufficient to promote generic neuronal differentiation both in vivo (Zimmerman et al., 1993; Turner and Weintraub, 1994; Ma et al., 1996; Blader et al., 1997; Perez et al., 1999) and in vitro (Lo et al., 1998; Farah et al., 2000; Sun et al., 2001). Whether these bHLH factors are sufficient to promote distinct neuronal subtype identities, however, is less clear.

In Drosophila, establishment of the identity-specification functions of the proneural genes has crucially depended on comparative gain-of-function (GOF) studies (Jarman et al., 1993; Jarman and Ahmed, 1998). By contrast, there is no reported GOF study in which the specific neuronal subtypes promoted by mis-expression of Mash1 and NGNs have been quantitatively compared side-by-side in the same vertebrate system. Although retroviral mis-expression of these bHLH factors has been performed in rodent cortex (Cai et al., 2000), no detailed information was reported on the relative ratios of different neuronal subtypes generated by the ac-sc and ato homologs. The ability of Xenopus ac-sc and ato homologs to promote pan-neuronal marker expression has been directly compared (Chitnis and Kintner, 1996), but specific neuronal subtypes have not been examined.

Many non-comparative studies of the neuronal subtype specification function of vertebrate proneural genes have been performed. There is evidence that constitutive or ectopic expression of NGNs, and other ato homologs, yields specific neuronal subtypes in vivo (Blader et al., 1997; Kanekar et al., 1997; Olson et al., 1998; Perez et al., 1999; Gowan et al., 2001). There is also evidence that MASH1 contributes to the specification of neuronal subtype identity in the CNS (Fode et al., 2000). However, in the absence of comparative data, it remains unclear whether ato and ac-sc homologs would both promote the same neuronal subtype(s) in a given cellular context. Moreover, because it is difficult to control the extracellular environment in such in vivo experiments, it is not clear whether these proneural genes autonomously determine specific neuronal subtypes, or rather simply provide a permissive context that allows identity specification by local cell-extrinsic signals (Scardigli et al., 2001). This issue can be addressed by performing comparative GOF experiments with NGNs and Mash1 in vitro, where the extracellular environment can be manipulated independently of the expression of these bHLH factors.

In the vertebrate PNS, NGNs are required for sensory neuron development (Fode et al., 1998; Ma et al., 1998; Ma et al., 1999), while Mash1 is, conversely, required for autonomic neurons (Guillemot et al., 1993). We have compared the subtype(s) of PNS neurons promoted by NGNs and Mash1 when these genes are misexpressed in primitive PNS progenitors derived from the neural tube, or in neural crest stem cells (NCSCs) (Stemple and Anderson, 1992). In neural tube progenitors dorsalized by low concentrations of BMP2, NGNs promote exclusively sensory neurogenesis, while MASH1 promotes only autonomic neurogenesis. However at higher concentrations of BMP2, or in NCSCs, NGNs promote only autonomic neurogenesis. These results parallel those obtained for ac-sc and ato in the Drosophila PNS (Jarman et al., 1993; Jarman et al., 1995; Jarman and Ahmed, 1998), and suggest that MASH1 and the NGNs, like their Drosophila counterparts, both have precursor-selection and subtype-specification functions, despite recent arguments to the contrary (Brunet and Ghysen, 1999; Hassan and Bellen, 2000). However, the subtype-specification function of NGNs is more sensitive to context, like that of ATO (Jarman et al., 1993; Jarman and Ahmed, 1998). Our experiments also reveal differences in the relative sensitivity of MASH1 and NGNs to Notch-mediated inhibition of neurogenesis and cell cycle exit. These differences may explain cellular differences between MASH1- and NGN-dependent lineages in the timing of neuronal differentiation and cell cycle exit, in both the PNS and CNS.

Rat E10.5 neural tubes were dissected as described (Lo et al., 1998). About 40-60 pooled neural tubes were trypsinized, washed thoroughly and replated at 15,000 cells per 4 mm diameter cloning ring (an area of 0.126 cm²), in a 35 mm fibronectin-coated dish. Cells were allowed to settle down for 1 hour before the rings were removed. Cells were cultured in a defined medium as described elsewhere (Greenwood et al., 1999), except that basic fibroblast growth factor (bFGF, 12 ng/ml; UBI) was included at all times. Where indicated, the culture medium was supplemented with BMP2 (gift from Genetics Institute). The neural tube cells reaggregated after 24 hours and, depending on the concentration of BMP2, were surrounded by many dispersed cells after 48 hours. Retroviral infection was performed on 24 hour cultures, and viruses were added at an m.o.i. (multiplicity of infection) of one viral particle per cell for 3-4 hours at 37°C. Infected cells were washed once and cultured for an additional 3 days before fixation and staining.

The full-length coding sequences of MASH1, NGN1 (NEUROD3 – Mouse Genome Informatics) and NGN2 (ATOH4 – Mouse Genome Informatics) were amplified from plasmids (Perez et al., 1999) using PFU-polymerase (Stratagene) as described (Lo et al., 1998). These genes were cloned into a pLZRS vector (Kinsella and Nolan, 1996) which was modified to contain a cassette encoding IRES-nuclear super GFP with five Myc epitope tags (super GFP was a gift from Barbara Wold). The pLZRS retroviral vector used the nuclear replication and retention functions of the human Epstein-Barr virus (EBV) to maintain the retroviral constructs as episomally replicated plasmids in 293-T-based retroviral packing cells (Ory et al., 1996). Production of high titer virus was modified slightly from elsewhere (Burns et al., 1993; Okada et al., 1999); details can be provided upon request.

DNA sequences encoding the N terminus of the bHLH domain of NGN2 and the full-length coding sequence of SOX10 were cloned into pGEX-4T3 (Pharmacia). After IPTG induction, the GST fusion proteins were extracted from bacteria and purified by GST column. The immunization protocol was as described elsewhere (Lo et al., 1991). Initial screening of hybridoma supernatants was performed in pLZRS-NGN2 or pLZRS-SOX10 retroviral infected chicken embryonic fibroblast (CEF) cells, followed by screening on rat E12 to E13 sections. For anti-NGN2, out of 545 clones, two mouse IgG1- (5C6 and 8B1) and one mouse IgG2a-(7G4) – secreting hybridomas were selected for their specificity to NGN2. The specific staining of all three antibodies was blocked by the NGN2-GST fusion, but not by an NGN1-fusion protein. For anti-SOX10, out of more than 800 clones, three mouse IgG1-secreting hybridomas, 20B7,20A7 and 13C9, were selected. Epitope-mapping experiments indicated that 20B7 recognizes a determinant in the first 65 amino acids, which contains sequences unique to SOX10. Anti-NGN2 and anti-SOX10 hybridoma supernatants were used at a 1:20 dilution.

Rat embryos or cells were fixed in freshly prepared 4% paraformaldehyde in PBS. After sucrose sinking, the embryos were embedded and stored at –80°C. Only freshly sectioned embryos were used for staining. Antibodies were applied for 1-2 hours at room temperature for cell culture and overnight at 4°C for sections. Monoclonal antibodies were obtained and used as follows: BRN3A (POU4F1 – Mouse Genome Informatics) (1:50 hybridoma supernatants, mouse IgG1, Chemicon), GFP (1:200, mouse IgG2a, clone 3E6, Molecular Probe), TUJ1 (1:1000, mouse IgG2a, Covance), Myc tag (1:4 hybridoma supernatant, mouse IgG1, clone 9E10,ATCC), NF160 (1:200, mouse IgG1, clone NN18, Sigma), RET [1:4 hybridoma supernatant (Lo and Anderson, 1995)] and NeuN (1:200, Chemicon), and anti-BrdU (1:100, Caltag). Rabbit polyclonal antibodies were obtained and used as follows: BRN3A (1:1000, gift from Eric Turner), PHOX2B (PMX2B – Mouse Genome Informatics) (1:1000, gift from Goridis and Brunet), Myc tag (1:500, Santa Cruz), GFP (1:500, Molecular Probe), anti-NF-M (1:500, Chemicon) and peripherin (1:1000, Chemicon). Donkey anti-mouse or rabbit secondary antibodies conjugated to FITC, PE, Cy3 and Cy5 were obtained from Jackson ImmunoResearch). Goat anti-mouse IgG1 or IgG2a conjugated with PE or FITC were obtained from Southern Biotechnology. Goat anti-mouse IgG1 or IgG2a conjugated with Alexa 488 or Alexa 568 were obtained from Molecular Probes. Secondary antibodies were used at 1:200 to 1:300 dilution and applied for 1 hour at room temperature. For double- or triple labeling, a cocktail containing the required antibodies was mixed, filtered, and centrifuged to remove aggregates before use.

For BrdU analysis, GFP or NF-M positive cells were detected using mouse anti-GFP or rabbit anti-NF-M, followed by Alexa 488-conjugated goat anti-mouse IgG2a or Cy5-conjugated donkey anti-rabbit secondary antibodies. BrdU incorporation was detected as described previously (Novitch et al., 1996) using a mouse anti-BrdU antibody and Alexa 568-conjugated goat anti-mouse IgG1 secondary antibody (Molecular Probes)

Positive cells were visualized using a Olympus microscope or Leica TCS SP confocal Microscope. Randomly selected 10× or 20× fields were sequentially scanned with different laser beams for FITC, Cy3 (or PE) and Cy5. Data were collected by directly counting the positive cells from the pictures. In some cases, Topro-3 (nucleic acid dye, Molecular Probes) were used to reveal the total cells in a given field.

To compare the PNS neuronal subtypes generated by constitutive expression of NGNs and Mash1 in vitro, we first sought to identify and characterize a cultured neural progenitor population that is competent to generate both sensory and autonomic neurons de novo. NCSCs isolated from E10.5 rat neural tube explants can generate autonomic neurons (Shah et al., 1996), but these stem cells have never been observed to generate sensory neurons, either in vitro or in vivo (White et al., 2001). Neural tube explant cultures also contain a subpopulation of sensory neuron precursors; however, such precursors appear determined for a sensory fate and lineally distinct from NCSCs (Greenwood et al., 1999). We therefore reasoned that dissociated E10.5 neural tube cultures might contain an earlier population of premigratory neural crest cells, which are unspecified with respect to sensory and autonomic neuronal subtypes (Bronner-Fraser and Fraser, 1988; Bronner-Fraser and Fraser, 1989).

The formation of the neural crest from the neural tube is known to require a dorsalization function that is provided by BMP family members expressed in the dorsal neural tube and/or dorsal epidermis, such as BMP2, BMP4, BMP7 and GDF7 (Liem et al., 1995; Liem et al., 1997; Lee and Jessell, 1999). Consistent with this idea, BMP2 (and BMP4) has been shown to promote the appearance of cells expressing the neural crest marker p75^LNTR in dissociated cultures of E10.5 neural tubes (Mujtaba et al., 1998). These cells can differentiate to autonomic-like neurons. However those studies did not establish whether these p75⁺ crest-like cells also generate sensory neurons. To address this question, we examined the expression of markers of sensory precursors and sensory neurons in such dissociated neural tube (dNT) cultures.

We first characterized the formation and differentiation of neural crest cells in such cultures using a new monoclonal antibody (mAb) to the HMG-box protein SOX10 (Kuhlbrodt et al., 1998; Southard-Smith et al., 1998). Unlike p75, which is expressed not only by neural crest cells, but also eventually by neural tube-derived cells (Kalyani et al., 1997), SOX10 is neural crest-specific (Britsch et al., 2001), as confirmed by staining sections of E10.5 rat embryos with the anti-SOX10 mAb (Fig. 1F, arrowheads).

We next examined the development of SOX10⁺ cells in dNT cultures grown in the absence or presence of different concentrations of BMP2. In the absence of exogenously added BMP2, growth of dissociated neural tube cells for 24 hours resulted in the formation of dense cellular reaggregates that were predominantly SOX10^– (Fig. 1A, arrows). At the margins of these reaggregates, however, more dispersed cells were visible that were often SOX10⁺ (Fig. 1A,B; arrowheads). As the concentration of BMP2 was increased, the proportion of SOX10⁺ cells increased (Fig. 1A,C,E). The location of the relatively more dispersed SOX10⁺ cells at the margins of the dense SOX10^– clusters suggested that these two cell populations probably represent neural crest and re-aggregated neural tube cells, respectively.

The presence of neural crest cells in dNT cultures raised the question of whether these cells might include sensory neuron precursors. To investigate this, we first examined the expression of NGN2, which marks a subset of such precursors in vivo (Ma et al., 1999). In sections of E10.5 rat embryos, NGN2 is expressed by a small subset of cells at the dorsal margin of the neural tube (Fig. 1G, arrows), as well as in more numerous SOX10⁺ cells in the neural crest migration pathway (Fig. 1G,H; arrowheads). However the majority of migrating SOX10⁺ crest cells are Ngn2^–. At later stages, very few Ngn2⁺ neural crest cells co-express SOX10 (not shown), suggesting that the HMG-box factor is rapidly downregulated in sensory precursors. In addition to neural crest-derived cells, the anti-NGN2 antibody stained many cells in the ventral neural tube, which include motoneuron precursors (Scardigli et al., 2001) (Fig. 1G).

In dNT cultures grown for 48 hours in 10 ng/ml BMP2, and then double-labeled with anti-NGN2 and anti-SOX10, small clusters of SOX10⁺ cells containing a few NGN2⁺ cells could be observed (Fig. 1I-K, arrows). In addition, there were NGN2⁺ cells that did not co-express SOX10 in dense aggregates, as well as more dispersed, SOX10⁺, NGN2^– cells (Fig. 1K). Approximately 8% of NGN2⁺ cells co-expressed SOX10. These data paralleled the distribution of SOX10-expressing and NGN2-expressing cells in the neural crest in vivo, and suggest that the NGN2⁺, SOX10⁺ cells in dNT cultures grown in 10 ng/ml BMP2 may be sensory neuron precursors. Similar results were obtained in cultures grown in 1 ng/ml BMP2 (not shown). However, the frequency of SOX10⁺, NGN2⁺ cells increased fourfold between 1 and 10 ng/ml BMP2 (from an average of 11 cells per 20× field to 47 cells per 20× field; n=8 and 10 fields counted, respectively).

We next determined whether overt sensory neurogenesis occurred in dNT cultures grown in 10 ng/ml BMP2. After 3 days in vitro, such cultures contained numerous neurons, as revealed by staining with antibodies to pan-neuronal markers such as NeuN (Mullen et al., 1992) and neuron-specific β-III tubulin (Geisert and Frankfurter, 1989) (Fig. 2B and data not shown). To determine whether any of these neurons might be sensory, we stained them with antibodies to the POU homeodomain transcription factor BRN3A, which, in the PNS, marks sensory neurons and their immediate precursors in vivo (Gerrero et al., 1993; Fedtsova and Turner, 1995) and in vitro (Greenwood et al., 1999). Such staining revealed that many BRN3A⁺ cells co-expressed NeuN or β-III tubulin (Fig. 2A-C, arrows and data not shown). Many of these neurons were located in cell clusters or aggregates resembling peripheral ganglia, and were closely associated with SOX10⁺ cells (Fig. 2G-I).

Since BRN3A is also expressed by some dorsal interneurons in the spinal cord (White and Anderson, 1999), we also counterstained for peripherin, which is expressed at hight levels in peripheral neurons compared with central neurons (Parysek et al., 1988; Parysek and Goldman, 1988), and for RET, which is expressed by motoneurons (which are BRN3A^–) and not by dorsal interneurons (Pachnis et al., 1993). Many of the BRN3A⁺ neurons co-expressed peripherin and/or RET (Fig. 2D-F, arrows and data not shown), suggesting that they are indeed peripheral sensory neurons. Taken together, these data suggest that dNT cultures grown in 1 or 10 ng/ml BMP2 contain progenitor cells that can differentiate into sensory neurons.

In low concentrations of BMP2 (1 or 10 ng/ml), very few autonomic neurons, identified by expression of the paired homeodomain protein PHOX2B (Pattyn et al., 1997; Pattyn et al., 1999), were observed to develop (Fig. 3B,G). However, as the concentration of BMP2 was raised above 10 ng/ml, increasing numbers of PHOX2B⁺ cells were observed, while the number of BRN3A⁺ cells concomitantly decreased (Fig. 3G). At 50 ng/ml BMP2, very few BRN3A⁺ cells were observed in comparison with PHOX2B⁺ cells (Fig. 3D-G). Co-expression of BRN3A and PHOX2B in the same cells was never observed (Fig. 3F). Like the BRN3A⁺ cells, these PHOX2B⁺ cells co-expressed pan-neuronal markers such as NeuN, and peripheral neuron markers such as RET or peripherin, indicating that they are likely to be autonomic neurons (not shown). Consistent with this identification, cultures in which such PHOX2B⁺ neurons developed also contained cells expressing MASH1 at earlier times (data not shown).

Taken together, the foregoing data suggest that dNT cultures contain progenitors to sensory and autonomic neurons. The fact that such PNS neurons appear to differentiate in a reciprocal manner at different concentrations of BMP2 is, furthermore, consistent with the idea that they may develop from a common progenitor. Attempts to test this hypothesis directly by clonal analysis or retroviral lineage tracing were unsuccessful, however, for technical reasons. Nevertheless, at the population level, the dNT cultures were competent for both sensory and autonomic neurogenesis, and thus provided a system in which to test the neuronal subtype(s) promoted by constitutive expression of NGNs and Mash1.

To mis-express NGNs and Mash1 in dNT cultures, we cloned their coding sequences into pseudotyped, replication-incompetent retroviruses, where their expression is under the control of the viral LTRs. These constructs also contained an IRES-GFP cassette to mark infected cells (see Materials and Methods). We first compared the neuronal subtypes promoted by these proneural genes at 10 ng/ml BMP2 (Fig. 4). Under these conditions, dNT cells infected with Ngn1- or Ngn2-expressing retroviruses predominantly expressed the sensory marker BRN3A (Fig. 4A,C,E), while those infected with the Mash1-expressing virus primarily expressed the autonomic marker PHOX2B (Fig. 4B,D,F). These BRN3A⁺ or PHOX2B⁺ cells co-expressed pan-neuronal markers such as βIII-tubulin (Fig. 5A-D,G-H, arrows), as well as RET (Fig. 5I-L; O-P, arrows). The latter marker confirms their identity as peripheral sensory and autonomic neurons.

Importantly, the proportion of neurons expressing BRN3A was more than twice as high among NGN1-infected cells as among control GFP-infected cells (Table 1, 10 ng/ml BMP2). By contrast, the overall extent of neuronal differentiation was only 1.5-fold higher in NGN1-infected cultures (Table 1). At 0 and 1 ng/ml BMP2, this difference was even more pronounced (Table 1 and Fig. 6C). These data suggest that the enhancement of BRN3A expression by NGN1 cannot be explained simply by its enhancement of generic neuronal differentiation. Consistent with this, MASH1 enhanced neuronal differentiation to a similar extent as NGN1, yet the percentage of BRN3A⁺ neurons among MASH1-infected cells was never significantly higher than among control GFP-infected cells (Table 1). Thus, in low concentrations of BMP2, NGNs enhanced sensory but not autonomic differentiation, while MASH1 induced only autonomic neurogenesis (but did not suppress sensory differentiation).

The absolute fraction of NGN-infected cells expressing BRN3A increased as a function of BMP2 concentration between 0 and 10 ng/ml (Fig. 6A, red dots). However the enhancement of BRN3A expression by NGNs over control decreased in parallel (Fig. 6C and Table 2). This decrease reflects an increasing frequency of BRN3A expression among control GFP-infected cells (Fig. 6A, blue dots). These data suggest that there is a limit to the ability of exogenous NGN1 to enhance BRN3A expression above control levels. This could reflect a limit in the number of precursors that can differentiate to BRN3A⁺ neurons, and/or in the amount of exogenous NGN expression needed to activate BRN3A expression. Consistent with the latter explanation, at 10 ng/ml BMP2 there was a fourfold increase in the number of cells expressing endogenous NGN2 (data not shown).

Above 10 ng/ml BMP2, the percentage of BRN3A⁺ cells among NGN1-infected cells decreased, while that of PHOX2B⁺ cells increased (Fig. 6A, compare red dots with red triangles), paralleling the trend in uninfected cultures (Fig. 4). Nevertheless, the proportion of NGN-infected cells expressing PHOX2B was always higher than among control GFP-infected cells at 25 and 50 ng/ml BMP2 (Fig. 6A, red versus blue triangles; Table 2). At 25 ng/ml BMP2, a similar percentage of NGN1-infected cells expressed either BRN3A or PHOX2B (Fig. 6A, red symbols). However, triple-labeling of NGN1-infected cultures with anti-BRN3A, anti-PHOX2B and anti-GFP antibodies indicated that the sensory and autonomic markers were never co-expressed by the same infected cells (Fig. 7, arrows versus arrowheads). At 50 ng/ml BMP2, NGN1 only promoted autonomic neurogenesis, and did not enhance BRN3A expression significantly above control (Fig. 6A, red symbols; Table 2).

In contrast to NGN1, MASH1 always enhanced autonomic and not sensory neurogenesis, irrespective of BMP2 concentration (Fig. 6B, red symbols). For example, at 10 ng/ml BMP2, MASH1 stimulated PHOX2B expression about 25-fold relative to control (Figs 4H, 6B, red and blue triangles; Table 2). As the concentration of BMP2 was raised to 25 ng/ml, the enhancement of PHOX2B expression by MASH1 was slightly diminished (Table 2), reflecting the increased frequency of PHOX2B expression among control GFP-expressing cells (Fig. 6B, blue triangles). Nevertheless, at 50 ng/ml BMP2, MASH1 still enhanced expression of the autonomic marker substantially above control levels (Fig. 6B, triangles and Table 2).

At 25 ng/ml BMP2, NGN1 enhanced autonomic differentiation to a lesser extent than MASH1, whereas at 50 ng/ml BMP2 it did so to about the same extent as did MASH1 (∼30-fold; Table 2). The enhancement of PHOX2B expression by NGN1 and MASH1 does not simply reflect the promotion of generic neuronal differentiation, as both bHLH factors enhanced neurogenesis only approx. twofold under these conditions. Furthermore, the promotion of autonomic neurogenesis by NGN1 does not appear to be mediated by induction of MASH1: for example, out of 336 NGN1-infected cells examined at 25 ng/ml BMP2, none co-expressed endogenous MASH1 2 days post-infection. Thus, at high concentrations of BMP2, NGNs are converted to inducers of autonomic neurogenesis in dNT cells.

The foregoing data indicated that NGNs can specifically promote sensory neurogenesis in dNT cells at 0-10 ng/ml BMP2, while under these same conditions MASH1 promotes only autonomic neurogenesis. We were therefore interested to know whether NGNs could similarly promote sensory neurogenesis in NCSCs. NCSCs are multipotent self-renewing stem cells isolated from E10.5 neural tube explants that can generate multiple classes of autonomic neurons, glia and smooth muscle cells (Stemple and Anderson, 1992; Shah et al., 1996). MASH1 promotes autonomic neurogenesis in these cells (Lo et al., 1998). However, conditions that promote sensory neurogenesis in dNT cultures (e.g. 10 ng/ml BMP2) do not promote sensory neurogenesis in NCSC cultures (not shown).

When NCSCs were infected with either the NGN1 or MASH1 retroviruses, both proneural genes strongly enhanced generic neuronal differentiation after 4 days, a time at which virtually no neuronal differentiation was observed in control GFP-infected cultures (Fig. 8A-D, arrow and data not shown). When such infected colonies were triple-labeled for GFP, PHOX2B and BRN3A expression, however, in both NGN1- and MASH1-infected colonies, the infected cells only expressed PHOX2B and never BRN3A (Fig. 8K-N). As was the case in dNT cultures, the extent of PHOX2B induction by both bHLH factors was greater than the extent of induction of neurogenesis (49% PHOX2B⁺ versus 37% TuJ1⁺ cells for MASH1-infected cultures (n=162 cells in 10 colonies); 70% PHOX2B⁺ versus 50% TuJ1⁺ cells for NGN1-infected cultures (n=76 cells in 12 colonies), consistent with previous studies of MASH1 in these cells (Lo et al., 1998). Similar results were obtained in the absence or presence of 10 ng/ml BMP2, or with an NGN2 virus (not shown). Thus, NGNs exclusively promote autonomic neurogenesis in NCSCs, even under culture conditions that favor sensory neurogenesis in dNT cultures.

The observation that MASH1 and the NGNs each promote autonomic neurogenesis in NCSCs raised the question of whether these neural bHLH factors exhibit any detectable differences in activity in this stem cell context. We noted a difference in the size and composition of colonies infected with the two proneural genes. NGN-infected colonies tended to be smaller than MASH1-infected colonies by about a factor of three, and contained a higher percentage of neurons, on average (Fig. 8C-F and Table 3). In addition, a higher proportion of NGN-infected colonies consisted only of neurons [49% versus 19% for MASH1; Fig. 8C,E,G; Table 4 (% Neuron-only clones – Control)]. Conversely, the proportion of mixed colonies was much higher in MASH1-infected cultures [Figs 8D,F,H, 9A,B; Table 4 (% Mixed clones, Control)], and the size of such colonies tended to be larger (by ∼1.5- to 2-fold) as well. These data suggested that NGNs might promote cell cycle exit and neuronal differentiation more effectively or more rapidly than MASH1, in NCSCs.

To determine whether the smaller size of NGN-infected colonies indeed reflected a difference in proliferation, we performed BrdU-labeling experiments. MASH1-infected colonies pulsed for 24 hours with BrdU 3 days after infection had more than twice as many labeled cells as did NGN1-infected colonies (Table 3; Fig. 9A-D). Interestingly, this difference in BrdU incorporation was smaller when labeling was performed 24 hrs after infection (74±4% versus 64±2% BrdU⁺ cells in MASH1 versus NGN1-infected colonies, respectively). These data suggest that NGN1-infected NCSCs may differentiate and drop out of division sooner than MASH1-infected NCSCs. Consistent with this, when BrdU labeling was performed 3 days after infection, in both NGN1- and MASH1-infected cultures, most of the BrdU⁺ cells were undifferentiated non-neuronal cells (Fig. 9B,D,H; arrowhead and data not shown), although there were many more such cells in MASH1-infected cultures. In addition, there was a ninefold higher frequency of BrdU⁺ neurons in MASH1- than in NGN1-infected colonies (26±6% versus 3±1%, respectively), suggesting that forced expression of MASH1 is more compatible with continued proliferation in differentiating neuroblasts, than is that of NGN1. Little or no apoptotic cell death was detected by DAPI staining, suggesting that differential apoptosis does not contribute to the observed difference in colony size (not shown).

While the average number of neurons per clone was similar in MASH1- and NGN1-infected colonies (2.2±0.5 versus 1.8±0.3, respectively), MASH1-infected colonies contained many more non-neuronal cells (Fig. 9A-D; Table 3). One explanation for this is that MASH1 function might be more susceptible to lateral inhibition mediated by local Notch signaling within such colonies (Kubu et al., 2002). To test this, we exposed MASH1- and NGN1-infected colonies to an Fc-fusion form of the Notch ligand Delta (Delta-Fc), which has previously been shown to inhibit neurogenesis in NCSCs (Morrison et al., 2000). The ligand was added ∼2 hours post-infection to ensure that Notch signaling would be activated before the cells had a chance to differentiate into neurons in response to the bHLH factors.

Strikingly, Delta-Fc inhibited MASH1-induced neurogenesis more strongly than NGN1-induced neurogenesis (Fig. 9I-P; Table 4). The Notch ligand virtually abolished neuron-only colonies in MASH1-infected cultures, whereas it reduced them in NGN1-infected cultures by sevenfold (Table 4). The differential effect of Delta-Fc was also evident in mixed colonies: in MASH1-infected cultures the percentage of such colonies was reduced by 3.5-fold, whereas there was no significant reduction in NGN1-infected cultures (Table 4). Together, these data suggest that MASH1 is more sensitive than NGN1 to Notch-mediated inhibition of neurogenesis in NCSCs.

It has been suggested that the NGNs, unlike their Drosophila homolog ato, do not play a role in specifying neuronal identity in the vertebrate PNS (Hassan and Bellen, 2000). We have directly compared the subtype(s) of peripheral neurons promoted by constitutive expression of NGNs and Mash1 in vertebrate PNS progenitor cells. In dNT cultures dorsalized by low concentrations of BMP2, NGNs promoted sensory neurogenesis while MASH1 promoted autonomic neurogenesis. Thus, in the appropriate progenitor cell context, the NGNs and Mash1 each promote distinct neuronal subtype identities. However, in other contexts, both proneural genes promote autonomic differentiation, analagous to the behavior of their Drosophila homologs ac-sc and ato in the fly PNS (Jarman et al., 1993; Jarman and Ahmed, 1998). Our data also reveal differences in Notch sensitivity and cell cycle arrest-promoting activity between these vertebrate proneural genes, which may explain why some neural lineages employ MASH1 and others the NGNs.

The observation that NGNs can promote either sensory or autonomic neurogenesis in dNT cells, depending on the ambient level of BMP2, suggests that these bHLH factors collaborate with other dominant determinants of neuronal identity whose expression is promoted by different BMP2 concentrations (Fig. 10B – ID_Sn, ID_An, etc.). Thus, BMP2 appears to be the primary determinant of a sensory versus autonomic fate in dNT cells. However, the fact that NGNs can quantitatively enhance the expression of sensory-specific markers over and above their effect to promote neurogenesis, suggests that levels of Ngn1/2 expression may limit the proportion of dNT cells that can acquire a sensory fate (Fig. 10B). That this effect of NGNs is specific is supported by the fact that MASH1 does not promote a sensory identity under the same conditions. Consistent with this, substitution of the Mash1-coding sequence for that of the endogenous Ngn2 gene fails to rescue sensory neurogenesis in vivo (Parras et al., 2002). Thus, NGNs have a sensory-promoting function that MASH1 lacks.

At high concentrations of BMP2, NGNs promoted autonomic and not sensory neurogenesis, while at intermediate BMP2 concentrations Ngn-infected cells expressed either sensory or autonomic markers, but never both. The induction of autonomic neurogenesis by NGNs is unlikely to be mediated by endogenous MASH1, as an approximately equal proportion of NGN-infected cells expressed PHOX2B or BRN3A at 25 ng/ml BMP2 (Table 2) and yet endogenous MASH1 was not expressed in any NGN virus-infected cells at 2 days post-infection. The mutual exclusivity of sensory and autonomic marker expression suggests that there may be a reciprocal inhibition between the sensory and autonomic identity co-determinants with which NGNs interact (Fig. 10B, blunt arrows). Consistent with this notion, in the spinal cord, different homeodomain determinants of neuronal identity expressed in adjacent progenitor domains reciprocally inhibit each other’s expression (Muhr et al., 2001).

Unlike the NGNs, MASH1 promoted exclusively autonomic neurogenesis at virtually all concentrations of BMP2 tested. Such an observation might imply that MASH1 is sufficient to instruct an autonomic identity. However, MASH1 was unable to promote autonomic neurogenesis in the absence of BMP2, suggesting that it interacts with additional co-determinants of autonomic identity that are induced by BMP2 (Fig. 10B, ID_An). The existence of autonomic identity determinants independent of MASH1 is further suggested by the observation that Ngn2 can promote autonomic neurogenesis in the absence of endogenous Mash1 function in vivo (Parras et al., 2002).

Strong candidates for such autonomic identity co-determinants are PHOX2A (ARIX – Mouse Genome Informatics) and PHOX2B, paired homeodomain proteins that share identical DNA-binding domains (Pattyn et al., 1997) and that activate autonomic-specific genes in the PNS (Swanson et al., 1997; Lo et al., 1998; Yang et al., 1998; Lo et al., 1999; Stanke et al., 1999). Like Mash1 (Guillemot et al., 1993) Phox2b is essential for autonomic neurogenesis in vivo (Pattyn et al., 1999) but its expression is independent of Mash1 (Hirsch et al., 1998), and vice-versa (Pattyn et al., 1999). Another candidate for such a co-determinant is the bHLH factor eHAND, which also promotes autonomic neurogenesis when mis-expressed in chick embryos (Howard et al., 2000).

In complementary experiments, Guillemot and colleagues have shown that when Ngn2 is substituted for the Mash1-coding sequence by homologous recombination, it is able to rescue the sympathetic autonomic phenotype of Mash1^–/– mice, and does not induce sensory-specific markers in sympathetic ganglia (Parras et al., 2002). These results are consistent with our observation that NGNs promote autonomic and not sensory neurogenesis in dNT cultures at high concentrations of BMP2, or in NCSCs, and provide further evidence that this effect is unlikely to be mediated via induction of endogenous MASH1. Mash1 is first expressed when neural crest cells migrate to the dorsal aorta, which expresses BMPs (Reissman et al., 1996; Shah et al., 1996; Schneider et al., 1999). Our results suggest that Ngn2 does not induce sensory markers in autonomic ganglia when substituted for Mash1 (Parras et al., 2002), because the Mash1 transcriptional control elements constrain its expression to progenitors already restricted to an autonomic fate by autonomic identity co-determinants. However, our results also show that the inability of NGNs to override such autonomic determinants in vivo does not imply that these proneural genes play no role in promoting a sensory identity in the context where they normally function.

Comparative GOF studies in the Drosophila PNS first established that the proneural genes achaete-scute and atonal coupled neural precursor selection (analagous to generic neuronal differentiation) to neuronal subtype specification (Jarman et al., 1993; Chien et al., 1996; Jarman and Ahmed, 1998). It has recently been argued that in the vertebrate PNS, these two functions have become uncoupled (Hassan and Bellen, 2000; Brunet and Ghysen, 1999). According to this ‘uncoupling’ hypothesis, NGNs promote only precursor selection/generic neuronal differentiation and do not determine subtype identity; other, downstream-acting ato homologs are postulated to control identity specification (Bermingham et al., 2001; Gowan et al., 2001). Conversely, MASH1 is argued to specify subtype identity but not precursor selection in the PNS. The latter process is postulated to be controlled by other undiscovered proneural genes (Hassan and Bellen, 2000).

The observations presented here do not support this ‘uncoupling’ hypothesis. In dNT cells dorsalized by low concentrations of BMP2, NGNs and MASH1 each promote both generic neuronal differentiation, and the expression of distinct neuronal subtype-specific markers. These data argue that MASH1 and the NGNs each play a role in both precursor selection and neuronal identity. However, NGNs can promote either sensory or autonomic neurogenesis, depending on cellular context, while MASH1 exclusively promotes autonomic neurogenesis. Similarly, in Drosophila, forced expression of ac-sc promotes exclusively ES organ formation in the PNS, while forced expression of ato can promote either ES organ or CD organ formation, depending on context (Brand et al., 1993; Jarman et al., 1995; Jarman and Ahmed, 1998) (A. Jarman, personal communication). Thus, in the fly and vertebrate PNS, the function of proneural genes in promoting both generic neuronal differentiation and neuronal subtype specification appears to be conserved.

Our data unexpectedly revealed functional differences between MASH1 and NGNs that are unrelated to identity specification. Specifically, we observed a greater sensitivity of MASH1 to the inhibitory effects of NOTCH signaling on neurogenesis, and less rapid promotion of cell cycle withdrawal. Interestingly, as Mash1 and the NGNs are driven by the same retroviral transcriptional control elements, these differences must reflect post-transcriptional rather than transcriptional mechanisms (Kageyama and Nakanishi, 1997; Artavanis-Tsakonas et al., 1999). Consistent with this notion, in Xenopus XASH3, when expressed from micro-injected mRNA, is more sensitive to inhibition by activated Notch than is NeuroD, another ato-related gene (Chitnis and Kintner, 1996).

Neural bHLH factors, like their myogenic counterparts (Crescenzi et al., 1990), are known to promote cell cycle-withdrawal in conjunction with differentiation (Farah et al., 2000; Novitch et al., 2001). The difference in proliferation between MASH1- and NGN-infected NCSCs may therefore reflect intrinsic differences in the cell cycle withdrawal-promoting activities of the two bHLH factors. Alternatively, it may be secondary to their different sensitivities to Notch-mediated inhibition of neurogenesis. As lateral inhibition is known to occur within differentiating NCSC colonies (Kubu et al., 2002), MASH1-infected colonies may contain more proliferating cells than do NGN-infected colonies simply because they are more easily inhibited from undergoing neurogenesis by endogenous Notch signaling. Consistent with this, exogenous activation of Notch signaling by treatment with Delta-Fc more than doubled the average size of NGN1-infected colonies (Fig. 9M,N). However, we also observed that the proportion of BrdU⁺ neurons was ninefold higher in MASH1-infected than in NGN-infected cultures. These latter data suggest that NGNs may have an intrinsically stronger ability than MASH1 to promote cell cycle withdrawal in differentiating peripheral neuroblasts. Consistent with this idea, there was reduced proliferation in the sympathetic ganglia of mouse embryos in which the Ngn2-coding sequence was substituted for that of Mash1 by homologous recombination (Parras et al., 200).

These observations may explain the paradox of why some neural lineages have evolved to use Mash1, when Ngn2 can functionally substitute for it in vivo (Parras et al., 2002). During normal development, sensory neurons withdraw from the cell cycle and differentiate more rapidly than do sympathetic neurons (Rohrer and Thoenen, 1987). Sympathetic neuroblasts also continue to proliferate for several days after extending processes and expressing neuronal genes (Anderson and Axel, 1986; Rohrer and Thoenen, 1987; DiCicco-Bloom et al., 1990). Similarly, in the forebrain NGNs are used by cortical progenitors (Fode et al., 2000) that exit the cell cycle and differentiate shortly after leaving the ventricular zone (Chen and McConnell, 1995; Noctor et al., 2001), while MASH1 is used by striatal interneuron progenitors (Fode et al., 2000) that migrate to the cortex and olfactory bulb before differentiating (reviewed by Wilson and Rubenstein, 2000). These in vivo observations, and the functional differences we observe between Mash1 and the NGNs in vitro, suggest that MASH1 is employed in lineages where an extended period of proliferative amplification of undifferentiated neuronal precursors is required, while NGNs are used where more rapid differentiation without intervening proliferation is needed. The mechanistic basis of these functional differences will be interesting to determine.

We thank Eric Turner for anti-BRN3A antibody, Christo Goridis and Jean-François Brunet for anti-PHOX2B antibody, Rusty Lansford for introducing us to the pLZRS vectors, Susan Ou for assistance with monoclonal antibody production, Barbara Wold and Jeong Kyo Yoon for super-GFP, Peter Snow for assistance with GST-fusion proteins, Jaesang Kim for pLZRS-SOX10 retrovirus, François Guillemot for communicating unpublished data and for helpful comments on the manuscript, and Gaby Mosconi for laboratory management. This work was supported in part by a grant from the March of Dimes Foundation. E. D. was supported by a fellowship from the Wellcome Trust. D. J. A. is an Investigator of the Howard Hughes Medical Institute.