Gsk3β/PKA and Gli1 regulate the maintenance of neural progenitors at the midbrain-hindbrain boundary in concert with E(Spl) factor activity

Neurogenesis in the embryonic vertebrate neural tube is spatio-temporally controlled to coordinate neuronal production and progenitor cell maintenance. Neural progenitors are crucial for late events of brain maturation, but the mechanisms underlying their maintenance, recruitment and fate are only partially understood (Bally-Cuif and Hammerschmidt, 2003; Kageyama et al., 2005; Panchision and McKay, 2002). To approach these mechanisms, we are focusing on an evolutionarily conserved progenitor pool, the `intervening zone' (IZ), which adjoins the midbrain-hindbrain boundary (MHB) in the neural tube of all vertebrate embryos (Bally-Cuif et al.,1993; Geling et al.,2003; Stigloher et al.,2008; Vaage, 1969)(Fig. 1A-C). Lineage tracing of the IZ pool in the zebrafish embryo demonstrates that it progressively contributes neurons to the entire midbrain-hindbrain domain (MH)(Tallafuss and Bally-Cuif,2003), a territory involved in a number of physiological and integrative functions, including sensory processing, motor control and social behavior.

The multiple functional outputs of the MH require a diversity of neuronal subtypes organized in a precise neuroanatomical pattern. Two properties of the IZ demonstrate its importance in organizing MH maturation. Along the anteroposterior (AP) axis, the temporal order with which cells leave the IZ to populate the differentiating MH correlates with the future spatial organization and subtype of the neurons that they generate(Tallafuss and Bally-Cuif,2003). Hence, the mechanisms controlling maintenance of the IZ along the AP axis are likely to influence the generation of MH neuronal subtypes. Along the dorsoventral (DV, initially mediolateral) axis, IZ cells exhibit differences in their propensity to undergo neurogenesis, which correlate with the earlier maturation of the basal plate as compared with the alar plate (Easter et al.,1994): at early neurogenesis stages, medial (future ventral) IZ(MIZ) cells are more prone to undergo neurogenesis than lateral (future dorsal) IZ (LIZ) cells (Ninkovic et al.,2005).

The maintenance of IZ progenitors is controlled by E(Spl) bHLH transcription factors, such as mouse Hes1 and Hes3 or zebrafish Her5 and Her11, which inhibit expression of proneural genes, such as neurogenin 1 (neurog1) (Hatakeyama et al., 2004; Kageyama et al.,2005; Stigloher et al.,2008). The compound genetic ablation of mouse Hes1 and Hes3 leads to the premature differentiation of the IZ and loss of MH neuronal identities (Hatakeyama et al.,2004; Hirata et al.,2001). her5 and her11 are expressed across the IZ and play redundant roles in IZ maintenance(Ninkovic et al., 2005). Specifically, the IZ is sensitive to a total level of Her5+Her11, with differences along the mediolateral axis: three copies of her5 and/or her11 are sufficient to maintain the IZ, two copies are enough to maintain the LIZ but not the MIZ (which transforms into a neurog1-positive zone) (Fig. 1B), and a lower amount leads to ectopic expression of neurog1 across the LIZ as well(Ninkovic et al., 2005)(Fig. 1C). Hence, E(Spl)activity controls IZ maintenance and is sensed differently by MIZ and LIZ cells.

Unlike classical E(Spl) genes, her5 and her11 are not activated by Notch signaling (Bae et al.,2005; Geling et al.,2004; Hans et al.,2004; Hatakeyama and Kageyama,2006; Stigloher et al.,2008), and the molecular cascades involving E(Spl) activity during IZ formation remain unknown. Similarly, the mechanisms rendering MIZ cells less sensitive to E(Spl) activity than LIZ cells need to be discovered. To reveal these mechanisms, we used sensitized conditions in which the total level of E(Spl) activity is reduced, and we tested the influence of several signaling cascades and signal transduction pathways active in the anterior neural plate. We show that the kinases Gsk3β and PKA are novel determinants of IZ formation in zebrafish, and demonstrate that Gsk3β/PKA cooperate with E(Spl) activity in a dose-dependent manner throughout the IZ. Similar to E(Spl) factors, Gsk3β and PKA activities are sensed differentially along the mediolateral axis. We demonstrate that the transcription factor Gli1, expressed within the MIZ and displaying neurogenic activity, accounts for this differential response. Surprisingly, we also show that this activity of Gli1 is independent of Hedgehog (Hh) signaling. Our results provide a molecular framework to help understand IZ maintenance and its properties along the AP and DV axes.

Embryos from wild-type (AB) or transgenic[hsp70:tcf3-GFP^w26(Lewis et al., 2004) and Tg(her5PAC:EGFP)^ne1939(Tallafuss and Bally-Cuif,2003)] fish were staged according to Kimmel et al.(Kimmel et al., 1995). smoothened (smu^b641)(Varga et al., 2001) and Df^w5/w5 (Lekven et al., 2003) mutants were obtained by pairwise mating of heterozygous adult carriers.

In situ hybridization experiments were performed as described(Hammerschmidt et al., 1996b; Ninkovic et al., 2005) with the following probes: her5(Müller et al., 1996), neurog1 (Korzh et al.,1998), pax2.1 (pax2a - ZFIN)(Lun and Brand, 1998), gli1 [recloned using the primers recommended by Thisse and Thisse(Thisse and Thisse, 2005) for probe n°eu934], gli2(Karlstrom et al., 2003), gli3 (Tyurina et al.,2005), wnt1 (Molven et al., 1991), myoD (myod1 - ZFIN)(Weinberg et al., 1996), shh (Krauss et al.,1993) and EGFP (Clontech). Primary antibodies for immunohistochemistry were rabbit anti-GFP (AMS Biotechnology Europe, TP401,1/500), mouse anti-human neural protein HUC/HUD (ELAVL3/4 - HUGO) (MoBiTec,A-21271, 1/300), rat anti-BrdU (Abcam, 1/200) and rabbit anti-phospho-histone H3 (Upstate Biotechnology, 1/200), revealed using FITC-, AF488- or Cy3-conjugated secondary antibodies (Invitrogen). BrdU immunohistochemistry involved pretreatment with 3.3 M HCl for 30 minutes at room temperature,followed with two 15-minute washings in sodium tetraborate buffer (0.1 M, pH 8.5). Embryos were scored and photographed under a Zeiss Axioplan photomicroscope.

Capped RNAs were synthesized using the Ambion mMessage mMachine Kit following the supplier's instructions and were injected at the one-cell stage:a constitutively active form of protein kinase A, PKA^*(Hammerschmidt et al., 1996a)at 20 ng/μl; a dominant-negative form of the PKA regulatory subunit, dnReg (Strähle et al.,1997) at 50 ng/μl; and truncated Patched(patchedΔloop2)(Briscoe et al., 2001) at 50-150 ng/μl. Morpholino antisense oligonucleotides (gli1MO)(Gene-Tools, OR) and gripNA antisense oligonucleotides (her11GripNAand her5GripNA) (Active Motif, Belgium) were injected into one-cell stage embryos at 0.5 mM. Sequences of antisense oligonucleotides are: her5GripNA, 5′-GGTTCGCTCATTTTGTG-3′; her11GripNA^ATG, 5′-ATTCGGTGTGCTCTTCAT-3′(Ninkovic et al., 2005); and gli1MO, 5′-CCGACACACCCGCTACACCCACAGT-3′(Karlstrom et al., 2003).

Glycogen synthase kinase 3 β (Gsk3β) inhibition was achieved by applying LiCl or Gsk3β inhibitor III(2,4-dibenzyl-5-oxothiadiazolidine-3-thione, OTDZT) (Calbiochem, Germany). Lithium (0.3 M LiCl) treatments were performed at 28°C in embryo medium for 15 minutes, followed by three washes in embryo medium and further development. OTDZT was diluted to 1 mM in embryo medium and applied as with LiCl. Cyclopamine (BIOMOL, Germany) was applied at 100 μM to dechorionated embryos from the stages indicated until fixation for in situ hybridization. Controls were soaked in carrier only.

Tg(her5PAC:EGFP)^ne1939 transgenic zebrafish embryos(Tallafuss and Bally-Cuif,2003) were injected with 50 ng/μl 2xlckmRFP capped RNA(Megason and Fraser, 2003) in two central cells at the eight-cell stage. Developing embryos were imaged live using an inverted laser-scanning confocal microscope (LSM510Meta, Zeiss) as described (Distel and Köster,2007; Köster and Fraser,2004). Development was followed at 28°C using a 20× 0.5 Plan-Apochromat objective from shield until tailbud stage, when distinct expression of eGFP was visible. Three-dimensional stacks were collected every 5 minutes using excitation at 561 nm to record 2xlckmRFP protein. The movies were stopped when the embryos reached 90% epiboly and the final six scans were continued manually, including excitation at 488 nm to record GFP protein and locate the IZ. Subsequent image processing was performed using LSM software(version 3.2 SP1.1, Zeiss) and Photoshop (version 9.0, Adobe Systems). Tracking of cells was performed using the manual tracking function in ImageJ(version 1.37a, Wayne Rasband, National Institutes of Health USA, http://rsb.info.nih.gov/ij/)with the plug-in from Fabrice Cordeli (Institut Curie, Orsay, France).

Wild-type embryos were dechorionated and soaked in 10 mM BrdU, 15% DMSO in embryo medium for 10 minutes on ice. Embryos were then washed three times on ice for 10 minutes each with washing buffer containing 15% DMSO, fixed immediately and processed for BrdU immunohistochemistry.

Signaling pathways active along the IZ at the onset of neurogenesis might influence the neurogenic potential of MIZ and LIZ cells. At late gastrulation,expression of wnt1, wnt3a (wnt3l - ZFIN) and wnt10bare confined to the IZ area, at high levels within the LIZ and low or undetectable levels medially (Buckles et al., 2004; Lekven et al.,2003). Because Wnt signaling has been implicated in controlling neurogenesis in other instances (e.g. Megason and McMahon, 2002; Panhuysen et al., 2004; Zechner et al., 2003), in particular by supporting the proliferating state, we tested whether it influenced MIZ and LIZ cells. We expected that Wnt would favor the progenitor state and that higher Wnt signaling laterally would account for the higher sensitivity of LIZ cells to the inhibition provided by Her5/Her11.

To address this issue, we first tested whether the Wnt signaling activator LiCl (Hedgepeth et al., 1997; Klein and Melton, 1996)affected IZ formation. Wild-type embryos were incubated in 0.3 M LiCl for 15 minutes at late gastrulation, then washed and allowed to develop until the 3-to 5-somite stage, when they were analyzed for neurog1 expression. When LiCl was applied at 60% epiboly, the anterior neural plate was posteriorized, as in mutants with hyperactivated Wnt(Dorsky et al., 2003; Heisenberg et al., 2001; Kim et al., 2000; van de Water et al., 2001)(Fig. 2A). The IZ was enlarged along the AP axis, but neurogenesis inhibition was not affected (90% of cases, n=50) (Fig. 2A). By contrast, when LiCl was applied at 80% epiboly, prominent expression of neurog1 was observed ectopically within the MIZ (83% of cases, n=60) (Fig. 2B,E,E′, arrows, compare with Fig. 2D,F,F′), with only a moderate effect on AP patterning. This neurogenic phenotype was crucially dependent on the stage of LiCl application: incubations at later times did not affect IZ formation (88% of cases, n=56)(Fig. 2C). Thus, LiCl applied at 80% epiboly induces the premature commitment of MIZ progenitors towards differentiation.

Because LIZ cells are less prone to undergo neurogenesis than MIZ cells, it is possible that LiCl modulates neurogenesis throughout the IZ, but that its effects are only visible medially in the wild-type embryo. To test this, we further applied LiCl to embryos in which Her5 activity was decreased by injection of her5 antisense oligonucleotides (her5GripNA). Here, neurog1 expression was induced by LiCl throughout the IZ (86.5%of cases, n=52) (not shown), a phenotype never obtained with her5 knockdown alone (Fig. 1B) (Ninkovic et al.,2005). Thus, in the presence of LiCl, IZ cells are globally rendered more susceptible to undergo neurogenesis. her5 or her11 expression was not modified by LiCl treatment (96% of cases, n=30) (not shown). LiCl might rather modulate Her5/Her11 protein activity, or act downstream or in parallel to these factors.

Importantly, the effect of LiCl is opposite to that expected if endogenous Wnt signaling controlled the behavior of MIZ and LIZ cells: the expression of Wnt ligands in the LIZ would predict that activated Wnt decreases the tendency of a cell to differentiation. Hence, the neurogenic effects of LiCl might not be linked to Wnt signaling activation. To directly test this interpretation,we assessed the effects of LiCl on blockade of canonical Wnt downstream of LiCl activity. We used hsp70l:tcf3-GFP^w26 transgenic embryos, which express a truncated form of Tcf3a (Tcf7l1a - ZFIN) that acts as a dominant repressor of canonical Wnt upon heat shock(Lewis et al., 2004). We submitted these embryos to a heat-shock pulse at 50% epiboly, followed 2 hours later by LiCl treatment (at 80% epiboly), and analyzed neurog1expression at the 3- to 5-somite stage. A 2-hour delay after the heat-shock pulse is sufficient to significantly antagonize Wnt activity at gastrulation(Lewis et al., 2004); thus,Gsk3β inhibition by LiCl should occur while downstream components of canonical Wnt signaling were blocked. neurog1 expression was still strongly induced by LiCl across the MIZ (83% of cases, n=42)(Fig. 2, compare G with H),showing that LiCl is unlikely to influence IZ neurogenesis via activation of Tcf-mediated Wnt signaling.

Further arguments speak against a role for endogenous Wnt signaling in attributing IZ mediolateral differences. Firstly, in Tg(TOP:GFP)^w25 transgenic embryos in which GFP expression is under the control of Tcf-binding sites(Dorsky et al., 2002), we were unable to reveal active canonical Wnt signaling in the anterior neural plate at the onset of neurogenesis (98% of cases, n≥100), whereas GFP expression was obvious in active Wnt signaling domains such as the embryonic margin (not shown). Secondly, lowering Wnt signaling (e.g. in hsp70l:tcf3-GFP^w26 transgenic embryos heat-shocked at 50%epiboly) did not change neurog1 induction, neither by itself nor upon Her5 blockade (not shown). We conclude that canonical Wnt signaling is not involved in modulating the tendency of IZ cells to undergo neurogenesis, and that LiCl affects IZ neurogenesis independently of canonical Wnt.

LiCl effects are broad-ranged, but primarily inhibit Gsk3β activity(Berridge et al., 1989; Hedgepeth et al., 1997; Klein and Melton, 1996). To confirm that LiCl-induced neurogenesis across the IZ was mediated by Gsk3β blockade, we tested the effects of the selective Gsk3βinhibitor OTDZT (Martinez et al.,2002). Similar to LiCl treatment, wild-type embryos were incubated in OTDZT at 80% epiboly for 15 minutes, and processed for neurog1 in situ hybridization at the 3- to 5-somite stage. OTDZT induced ectopic neurog1 expression across the MIZ, mimicking the LiCl effect (87% of cases, n=38) (Fig. 2I-J′). Furthermore, as with LiCl, lowering Her5/Her11 dosage revealed that Gsk3β is active throughout the IZ (85% of cases, n=40) (Fig. 2K,L). Therefore, Gsk3β is a crucial element modulating neurogenesis at the IZ and is required in vivo for the formation and early maintenance of the MIZ.

In addition to targeting canonical Wnt, Gsk3β is involved in a number of signaling pathways, where it triggers enhanced phosphorylation of target proteins after these have been primed by phosphorylation via cAMP-dependent protein kinase A (PKA) (Jia et al.,2002; Price and Kalderon,2002; Zhang et al.,2005). To assess whether such a process might be at play during IZ formation, we tested whether PKA activation influences the neurogenic effect of inhibiting Gsk3β. As reported, only a few embryos (50% in our case)injected with capped mRNA encoding a constitutively active catalytic subunit of PKA (PKA^*) (Hammerschmidt et al., 1996a) develop a normal neural plate(Blader et al., 1997). Among these, all formed a normal IZ (Fig. 3, compare A,F with D,H). However, PKA^* inhibited the neurogenic effect of LiCl, restoring the MIZ when LiCl was applied to PKA^*-injected embryos (88% of cases, n=61)(Fig. 3, compare C with B). Therefore, general activation of PKA does not in itself expand the IZ, but it can compensate for the loss of Gsk3β function to maintain neurogenesis inhibition in this location.

We next addressed whether PKA activation could also compensate for a downregulation of E(Spl) function at the MHB, by testing the effects of co-injecting her5GripNA and PKA^* capped RNA into wild-type embryos. Whereas blocking Her5 function with gripNAs lead to a loss of the MIZ in the vast majority of cases (92% of cases, n=26)(Fig. 3E), the co-expression of PKA^* efficiently rescued this phenotype and restored MIZ formation(66% of cases, n=50) (Fig. 3, compare G with H). Thus, general activation of PKA also promotes neurogenesis inhibition downstream of, or in parallel to, Her5 function.

To determine whether PKA is in itself required for IZ formation, we blocked its activity by injection of capped RNA encoding a dominant-negative form of the PKA regulatory subunit, dnReg(Strähle et al., 1997). dnReg robustly prevents the transcription of downstream targets of the Hh pathway, such as spalt (sall1a - ZFIN), that require PKA for their expression (88% of cases, n=72) (not shown). In embryos expressing dnReg, the MIZ was replaced by ectopic neurog1 expression(84% of cases, n=50) (Fig. 4, compare C with A). When Her5 activity was also inhibited by the co-injection of her5GripNA, ectopic neurog1 expression was also detectable within the LIZ (86% of cases, n=35)(Fig. 4, compare E with A-C). To demonstrate that these cells were truly induced within the LIZ and did not originate from the neurog1-positive r2L domain (see Fig. 1A), we counted the number of cells in r2L upon co-injection of her5GripNA and dnReg. This was unchanged as compared with control embryos(Fig. 4I). Cell proliferation in this domain, measured by the number of cells expressing the M-phase marker phospho-histone H3 (PH3), also remained comparable to control embryos(Fig. 4J). Hence, migration of precursors from the r2L area, possibly compensated for by increased proliferation in r2L, is unlikely to account for the ectopic neurog1-positive cells found within the LIZ. Rather, we conclude that neurog1 expression is ectopically induced in this location upon blocking Her5 and PKA activities. Thus, blocking PKA increases the propensity of all IZ cells to undergo neurogenesis, and PKA is crucially required in vivo for MIZ formation.

Because blocking PKA or Gsk3β triggers identical phenotypes, we wondered whether these two factors act cooperatively, and so we concomitantly inhibited both activities by treating dnReg-injected embryos with OTDZT. As described above, when PKA and Gsk3β are manipulated separately, neurog1 expression is induced in place of the MIZ, but the LIZ is unaffected (Fig. 2K, Fig. 4C,D). By contrast,co-inhibition of PKA and Gsk3β generated ectopic neurog1-positive cells within the LIZ (75% of cases, n=62)(Fig. 4F). Again, cell counts revealed no change in the number of neurog1-positive cells within the r2L domain (Fig. 4I), strongly arguing for a bona fide ectopic induction of neurog1 expression within the LIZ. These results suggest a dose-dependent process co-regulated by PKA and Gsk3β.

Together, these results place neurogenesis inhibition by Gsk3β/PKA downstream of, or in parallel to, E(Spl) activity. To resolve this issue, we tested whether increased E(Spl) activity would compensate for decreased Gsk3β/PKA. We blocked PKA in Tg(her5PAC:EGFP)^ne1939transgenic fish (Tallafuss and Bally-Cuif,2003), which carry one additional copy of her11 under the control of its own regulatory elements, and thereby express three doses of Her5/Her11 activity in a correct spatio-temporal manner(Ninkovic et al., 2005). This transgenic background permits normal MIZ formation in the absence of Her5(Ninkovic et al., 2005). However, it proved insufficient to block ectopic neurog1 activation and rescue the MIZ upon expression of dnReg (79% of cases, n=56)(Fig. 4, compare H with G). dnReg had no effect on the level of expression of her5 and her11, and we failed to detect putative PKA or Gsk3βphosphorylation sites on Her5 and Her11 (not shown). Although we cannot exclude the possibility that higher doses of Her5/Her11 could be effective,the most parsimonious explanation for these results is that Gsk3β/PKA act downstream of E(Spl) factors.

Cell-intrinsic components or local signaling cues could account for rendering the MIZ and LIZ different in their response to the Gsk3β/PKA and Her5/Her11 pathways. To test for the relevance of cell-intrinsic mechanisms, we first assessed whether MIZ and LIZ differ in lineage. Extensive cell exchanges across the midline of the zebrafish embryonic neural tube have been documented, as well as dorsoventral dispersion, after the 2-somite stage(Papan and Campos-Ortega,1997), but not before (Woo and Fraser, 1995). However, IZ precursors have not been specifically studied in this context. To address this, we time-lapsed MIZ and LIZ precursors from the shield stage onwards using confocal microscopy in Tg(her5PAC:EGFP)^ne1939 transgenic embryos, in which her5-expressing cells are also positive for GFP(Tallafuss and Bally-Cuif,2003). With the help of a lineage reporter (2xlckmRFP)(Megason and Fraser, 2003)injected as capped RNA into a subset of cells, by playing the movies backwards from the tailbud stage we could trace the origin of single cells that contribute to the MIZ and LIZ (as well as to the r2M and r2L domains). We observed that these precursor cell groups remained spatially segregated at all times between the shield and tailbud stages(Fig. 5A,B). In particular, we did not observe any contribution of MIZ precursors to the LIZ and vice versa. This raises the possibility that MIZ and LIZ precursors inherit different determinants that might influence their sensitivity to neurogenesis inhibitors.

These determinants might act on cell proliferation, on the degree of cellular commitment toward differentiation, or both. Because cell cycle kinetics influence the capacity of a cell to enter the neurogenesis process(Calegari and Huttner, 2003),we tested whether MIZ and LIZ cells differed in their proliferation characteristics. Markers for the S (BrdU incorporation) and M (PH3, and mitotic figures) phases were analyzed together with the IZ marker her5 in triple-labeled preparations at 75% epiboly to tailbud. At these stages, there is still a tendency for cells to divide synchronously(Kimmel et al., 1994) and thus the co-analysis of S- and M-phase markers on single specimens was important to distinguish between differences in cell cycle length and cell division timing. Cell cycle kinetics, as well as the expression of p27Xic(cdkn1c - ZFIN), did not significantly differ between the MIZ and LIZ in any of the more than 50 embryos analyzed(Fig. 5C, and data not shown). These observations suggest that the cell division characteristics of the MIZ and LIZ are comparable, and that these populations differ in their commitment in a manner independent of cell cycle control.

Among commitment factors expressed within the early neural plate,transcription factors of the Gli family have been implicated in neurogenesis control in many systems (reviewed by Agathocleous et al., 2007; Ruiz et al., 2002). Furthermore, Gli factors have been identified as intracellular targets for the PKA/Gsk3β pair (Huangfu and Anderson,2006; Riobo and Manning,2007). gli1 appears to be expressed differentially between the MIZ and LIZ: in agreement with published data(Karlstrom et al., 2003), we observed that gli1 is transcribed within the anterior neural plate following a clear mediolaterally decreasing gradient starting at 80% epiboly(Fig. 6A,C,E, and data not shown). Double staining for GFP, identifying the IZ in Tg(her5PAC:EGFP)^ne1939 embryos, further confirmed that this gradient is also found at the level of the IZ(Fig. 6B,D,F). Three other Gli genes have been identified in zebrafish(Karlstrom et al., 2003; Ke et al., 2005; Tyurina et al., 2005). gli2b is not expressed at detectable levels within the IZ area at these stages, whereas expression of gli2a and gli3 is ubiquitous (not shown).

To address the function of Gli1 during IZ formation, we blocked gli1 translation by injection of a gli1 morpholino (MO)(Karlstrom et al., 2003) into wild-type embryos. This did not alter the IZ area (82% of cases, n=71) (Fig. 6, compare H,L with J,N; Fig. 7, compare E with A), although it was efficient at blocking expression of the Hh target nkx2.1 along the ventral midline of the anterior neural tube (96% of cases, n=30) (not shown). However, blocking Gli1 totally abolished the neurogenic effect of LiCl: in gli1 MO-injected embryos subjected to LiCl treatment at 80% epiboly, the MIZ formed normally (82% of cases, n=52) (Fig. 6, compare I with G). Thus, blocking Gli1 function can increase neurogenesis inhibition and rescue IZ formation in the absence of Gsk3β activity. Using a similar approach, we tested whether blocking Gli1 could compensate for the lack of PKA activity, by co-injecting gli1 MO and dnReg capped RNA into wild-type embryos. The co-inhibition of Gli1 and PKA abolished the neurogenic effect of blocking PKA alone and rescued MIZ formation (84% of cases, n=61) (Fig. 6, compare M with K). These results demonstrate that Gli1 exerts a neurogenesis-promoting activity that opposes the activity of Gsk3β/PKA. Under normal conditions,the activity of Gli1 is sub-threshold, such that neurogenesis inhibition is obtained. It becomes visible in the absence of Gsk3β or PKA activity,when the repression of Gli1 alone suffices to restore neurogenesis inhibition.

The co-regulation of IZ formation by Gsk3β/PKA and E(Spl) factors described above suggests that Gli1 function should also influence E(Spl)function. To verify this point, we reduced E(Spl)-mediated neurogenesis inhibition by injection of her5GripNA or her11GripNA into wild-type embryos, and we simultaneously blocked Gli1 by the co-injection of gli1 MO. The ectopic neurogenesis triggered by Her5 blockade was abolished by the loss of Gli1 function, restoring the IZ (89% of cases, n=72) (Fig. 7, compare F with B). Likewise, blocking Gli1 together with Her11 rescued the MIZ as compared with the effect of blocking Her11 alone (74% of cases, n=67)(Fig. 7, compare G with C). Thus, Gli1 expression also acts antagonistically to E(Spl) activity by promoting IZ neurogenesis, an effect that becomes apparent when E(Spl) dosage is reduced.

The neurogenesis-promoting function of Gli1 and its MIZ-specific expression make it a good candidate to enhance the neurogenesis potential of the MIZ. As described, blocking Gli1 activity increases MIZ sensitivity to two copies (or fewer) of Her5/Her11 (Fig. 7F,G). In addition, in the absence of Gli1, we observed that both the MIZ and LIZ concomitantly lost their responsiveness to a further downregulation of Her5/Her11 to now upregulate neurog1 in an identical manner (90% of cases, n=56)(Fig. 7H,H′). These results identify Gli1 as a crucial element rendering MIZ and LIZ cells differentially sensitive to E(Spl) factors.

We next searched for mechanisms controlling Gli1 expression or activity within the MIZ. Hh signaling is active in the presumptive axial mesoderm from 50% epiboly onwards (Ertzer et al.,2007; Krauss et al.,1993), and Gli1 behaves as a classical positive activator of Hh targets. Thus, we studied whether IZ formation and Gli1 expression require Hh signaling. We blocked Hh signaling upstream of PKA/Gsk3β action by incubating wild-type embryos in cyclopamine. Cyclopamine blocks the transmembrane protein Smoothened (Chen et al., 2002), which normally initiates intracellular Hh signaling events and in particular the inhibition of Gsk3β and PKA(Huangfu and Anderson, 2006; Riobo and Manning, 2007). As expected, such treatment performed at 50% epiboly efficiently inhibited expression of myoD in adaxial cells (97% of cases, n=40)(Fig. 8, compare H with G)(Barresi et al., 2000; Chen et al., 2001) and of spalt in the anterior neural plate (not shown). Surprisingly,however, cyclopamine treatment did not affect gli1 expression in the IZ area at any stage (100% of cases, n>20 for each stage)(Fig. 8A-F, and data not shown). Whatever the stage of application, cyclopamine also did not affect IZ formation (Fig. 8, compare J-L with N). Because Hh signaling might promote Gli1 activity rather than expression, we assessed whether cyclopamine downregulated Gli1 function. For this, we first reduced E(Spl) activity using her5GripNA, and next applied cyclopamine at 50% epiboly. Whereas inhibiting Gli1 function in such cases led to a rescue of the MIZ (Fig. 7F), cyclopamine was without effect and was incapable of counteracting ectopic neurog1 induction (89% of cases, n=75)(Fig. 8, compare M with I). We conclude that cyclopamine treatment affects neither the expression nor the function of Gli1 at the IZ. Similar results were obtained in smu^b641 mutants, deficient in Smoothened function(Varga et al., 2001), analyzed for gli1 expression at 70% epiboly, tailbud and 3 somites(n>30 embryos in each case) (not shown). Finally, we also observed that gli1 expression and IZ formation were unaffected when transduction of Hh signaling was blocked upon overexpression of a Hh-insensitive form of the receptor Patched, PatchedΔloop2(Briscoe et al., 2001) (100% of cases, n>100) (Fig. 8, compare O with P), even in the absence of Her5 function (not shown). These observations strongly suggest that the neurogenic activity of Gli1 at the IZ does not require Hh signaling.

The IZ gives rise in spatio-temporal order to the large majority of midbrain-hindbrain neurons (Tallafuss and Bally-Cuif, 2003). We previously demonstrated that E(Spl) activity is a crucial mechanism inhibiting neurogenesis to permit IZ formation(Geling et al., 2003; Geling et al., 2004; Ninkovic et al., 2005). We identify here a collaborative mechanism, relying on the activity of the two kinases, PKA and Gsk3β. The E(Spl) and Gsk3β/PKA pathways act in a dose-dependent manner, and one of their activities is to oppose the neurogenic effect of the transcription factor Gli1. Gli1, expressed medially, enhances the tendency of MIZ cells to undergo neurogenesis, a mechanism that might account for the earlier differentiation of basal versus alar plate neurons during MH development. Together, these results help refine a molecular model for the sequential differentiation of midbrain-hindbrain neurons along the AP and mediolateral (DV) axes (Fig. 9).

An important finding of our work is the identification of Gsk3β and PKA as new mediators of IZ formation. Both enzymes often act sequentially on the same targets in vivo, PKA priming target proteins for a subsequent phosphorylation by Gsk3β (Price and Kalderon, 2002), but alternative models are emerging in which PKA and Gsk3β independently phosphorylate the same target(Taurin et al., 2006). At the IZ, the enhanced effect of lowering both Gsk3β and PKA, and the fact that constitutive PKA activation can compensate for reduced Gsk3β, suggest the involvement of an unconventional dose-dependent process incorporating the level of activity of both enzymes. PKA and Gsk3β might be rate-limiting for the full phosphorylation necessary to functionally modify the same target. Alternatively, PKA and Gsk3β might act on distinct molecular targets,cooperating in parallel pathways during IZ formation. Our results do not allow us to distinguish between these possibilities.

E(Spl) and Gsk3β/PKA act in an additive manner, and our observations support parallel activities of E(Spl) and Gsk3β/PKA on a dose-dependent process sensitive to the global level of these pathways. Gsk3β/PKA and E(Spl) inhibition might converge onto common targets promoting neurogenesis inhibition, or have parallel targets cooperating in the neurogenesis inhibition process. The direct targets of E(Spl) activity during IZ formation remain unknown, although it is known to downregulate expression of neurog1, coe2 and p27Xic (cdkn1c - ZFIN)(Geling et al., 2004). Many targets inhibited by Gsk3β/PKA phosphorylation have been identified,including factors controlling cell cycle (e.g. N-myc1; Mycn)(Kenney et al., 2004; Mill et al., 2005),neurogenesis and cell differentiation (e.g. Neurogenin 2, XNeuroD, Xash1 and Mash1) (Ma et al., 2008; Moore et al., 2002), or both(e.g. β-catenin or Gli proteins) (reviewed by Frame and Cohen, 2001; Huangfu and Anderson, 2006; Riobo and Manning, 2007). None of our observations suggests a major role for cell proliferation control in IZ formation, corroborating previous findings that neurogenesis does not systematically follow cell cycle exit at the early neural plate stage(Geling et al., 2003; Hardcastle and Papalopulu,2000; Harris and Hartenstein,1991). Rather, in this system, entry into the neurog1-positive state follows information unlinked to cell cycle characteristics and we favor commitment factors as the main targets for the Gsk3β/PKA pathway.

There are target sites for Gsk3β on zebrafish NeuroD(Moore et al., 2002) but not on Neurog1 (our observations) and, in the IZ, Gsk3β/PKA affects neurog1 transcription rather than its activity. We failed to identify Gsk3β or PKA target sites on Her5 and Her11 (not shown), but an indirect role of Gsk3β/PKA on the activity of these factors cannot be excluded. Gli1 is an obvious candidate target, as Gsk3β/PKA classically regulate the nuclear translocation (Gli1) or processing (Gli2, Gli3) of Gli factors in a manner antagonized by Hh signaling(Huangfu and Anderson, 2006; Riobo and Manning, 2007). However, Gsk3β/PKA also block neurogenesis in the LIZ, where gli1 is not expressed. Hence, other direct targets of Gsk3β/PKA remain to be identified that permit IZ formation.

Finally, as with E(Spl) factors, the upstream pathways involving Gsk3β/PKA at the IZ are currently unknown. Classically, Wnt and Hh inhibit Gsk3β and/or PKA to permit signal transduction (reviewed by Hooper and Scott, 2005; Huangfu and Anderson, 2006; Logan and Nusse, 2004; Riobo and Manning, 2007). However, Wnt reporters remained silent in the anterior neural plate at the stage of interest, and blocking Wnt or Hh activities or ubiquitously enhancing PKA activity did not expand the IZ. IZ formation might be achieved through a constitutive activation of Gsk3β/PKA in concert with local cofactors,and/or through signaling pathways other than Wnt and Hh. A number of upstream regulators have been identified for these kinases - in particular, classical mitogenic pathways driven by PI3 kinase and Akt(Dudek et al., 1997; Kenney et al., 2004). It will be important to test their influence during IZ formation.

gli1 expression is concentrated medially and has a neurogenesis-promoting effect, albeit at sub-threshold levels. Significantly,blocking Gli1 activity abolishes the difference between MIZ and LIZ sensitivity towards E(Spl)-mediated neurogenesis inhibition. Although multiple additional blocks might still exist between neurog1 expression and the completion of the neuronal differentiation program, the mechanism uncovered here is likely to facilitate the transition of MIZ cells towards neuronal production and might be relevant to the earlier maturation of the midbrain basal plate that is observed in all vertebrates during development(Easter et al., 1994; Puelles et al., 1987; Ross et al., 1992; Wilson et al., 1990). Although E(Spl) and Gli1 activities have opposing effects on neurogenesis at the IZ,our triple knockdown experiments of Gli1, Her5 and Her11 demonstrate that Gli1 itself is dispensable for neurog1 expression in the absence of the inhibitory activity of E(Spl) factors. Gli1 activity might, however, be required in this context for the progression of neuronal differentiation beyond neurog1 expression, an issue that remains to be tested.

Gli1 exerts positive and negative effects on neurogenesis depending on the cell type or state. Neurogenesis progression fails when Gli1 function is blocked in the Xenopus neural plate or zebrafish retina(Masai et al., 2005; Nguyen et al., 2005). By contrast, downregulation of Gli1 prevents re-entry into S phase in the chicken ventral neural tube (Cayuso et al.,2006). Overall, Gli1 might bring cells closer to differentiation,hence, pushing early progenitors to amplify, and later progenitors to a final cell cycle. At the IZ, however, where an amplifying population has not been observed, an effect on the cell cycle is improbable and we propose that Gli1 targets cell cycle-independent commitment genes. Further arguments supporting this hypothesis stem from previous analyses of zebrafish gli1(detour) mutants (Karlstrom et al., 1999). These mutants lack cranial motoneurons, including nerve III (Chandrasekhar et al.,1999) [which is likely to originate from the MIZ(Tallafuss and Bally-Cuif,2003) (K. Webb and C.S., unpublished)], but display no defects in neurogenesis at 18 hours, pointing to a differentiation rather than an induction defect (Chandrasekhar et al.,1999). However, the lack of cranial motoneurons of gli1mutants can be mimicked by loss of Smoothened function or by cyclopamine treatment (Vanderlaan et al.,2005), and hence might be a result of impaired Hh signaling, which contrasts with our interpretation of Gli1 regulation at the MIZ. In any case,both analyses point to a role for Gli1 in promoting differentiation in a manner independent of cell cycle control.

Given the importance of Gli1 function in neurogenesis progression, it remains crucial to uncover the pathway(s) controlling Gli1 expression in the MIZ. Gsk3β/PKA can downregulate Gli1(Huangfu and Anderson, 2006; Riobo and Manning, 2007) and might do so during IZ formation. This is likely to occur at the post-transcriptional level, as gli1 transcript levels were not modified by Gsk3β or PKA blockade (not shown). However, this downregulation of Gli1 by Gsk3β/PKA is only partial, because blocking Gli1 activity has further phenotypic consequences.

A most surprising aspect of our study is that Gli1 expression and activity at the IZ are not under the control of Hh signaling alone. Blocking Hh with cyclopamine, in smoothened (smu) mutants or by overexpressing a Hh-insensitive form of Patched, affected neither gli1 expression nor IZ formation. Although surprising, these findings are in keeping with previous observations by Karlstrom et al. who reported maintenance of gli1 expression in the anterior neural plate of smu mutants or cyclopamine-treated embryos(Karlstrom et al., 2003). Recently, a discrepancy between the effects of smu mutations or cyclopamine and the effects of forskolin, which activates PKA, were noted during retinal neurogenesis in zebrafish: whereas forskolin strongly impaired neurogenesis progression, smu mutant or cyclopamine-treated embryos exhibited only mild defects (Masai et al.,2005). Our results are comparable to these observations and, we believe, question the role of Hh, if any, in Gli1 regulation across the IZ. In a few instances, activation of Gli1 (or Gli2) has been reported to follow activation of the Erk pathway by Fgf signaling(Brewster et al., 2000; Riobo et al., 2006a; Riobo et al., 2006b) or of the PI3K and Akt pathways by mitogens (Riobo et al., 2006b). It is possible that such an atypical,Hh-independent mechanism is involved at the midline of the anterior neural plate. Unravelling this process will be important to our understanding of neurogenesis control and the cross-regulatory activities of Hh signaling pathway components.

We are grateful to Dr Kenji Imai and to members of the Bally-Cuif laboratory for support and discussions; to Silvia Krutsch, Birgit Tannhäuser and Stefanie Topp for technical support; to the R. Köster laboratory for help with time-lapse microscopy; to Dr William Norton for his careful and critical reading of the manuscript; and to Drs Rolf Karlstrom, Uwe Strähle and Masanari Takamiya for sharing knowledge and reagents on the Hedgehog pathway and neurogenesis. Work in the Bally-Cuif laboratory is funded by a junior group grant from the Volkswagen Association, the EU 6th Framework Integrated Project ZF-Models (LSHC-CT-2003-503466), the Life Science Association (GSF 2005/01), the Center for Protein Science-Munich (CIPSM), and a special research grant from the Institut du Cerveau et de la Moelleépinière (ICM, Paris).