|Home | About | Journals | Submit | Contact Us | Français|
The homeobox gene Gsx2 (formerly Gsh2) is known to be required for striatal and olfactory bulb neurogenesis, however, its specific role in the specification of these two neuronal subtypes remains unclear. To address this, we have employed a temporally-regulated gain-of-function approach in transgenic mice and found that misexpression of Gsx2 at early stages of telencephalic neurogenesis favors the specification of striatal projection neuron identity over that of olfactory bulb interneurons. In contrast, delayed activation of the Gsx2 transgene until later stages exclusively promotes olfactory bulb interneuron identity. In a complementary approach, we have conditionally inactivated Gsx2 in a temporally progressive manner. Unlike germline Gsx2 mutants, which exhibit severe alterations in both striatal and olfactory bulb neurogenesis at birth, the conditional mutants exhibited defects restricted to olfactory bulb interneurons. These results demonstrate that Gsx2 specifies striatal projection neuron and olfactory bulb interneuron identity at distinct time points during telencephalic neurogenesis.
The telencephalon represents the largest and most complex region of the mammalian brain. This region is charged with the task of complex neural processing that controls all cognitive processes and purposeful actions. Accordingly, the telencephalon exhibits the greatest amount of neuronal diversity of any portion of the CNS. Previously, many groups have focused on the generation of neuronal diversity within the telencephalon (reviewed in Marin and Rubenstein, 2003). While neuronal progenitors in the dorsal telencephalon (also termed the pallium) are thought to give rise to the excitatory cortical projection neurons, the vast majority of cortical interneurons originate from progenitor domains located in the ventral telencephalon. Although limited pallial contributions to ventral telencephalic neuronal subtypes have recently been proposed (Willaime-Morowak et al., 2006; Kohwi et al., 2007; Young et al., 2007; Williame-Morawek and van der Kooy, 2008), most of the neuronal diversity found in the mature telencephalon appears to derive from progenitor cells positioned in the ventral telencephalon during embryogenesis (Rallu et al., 2002; Campbell, 2003; Marin and Rubenstein, 2003).
The lateral ganglionic eminence (LGE) represents one such ventral telencephalic progenitor region, which has been shown to generate the projection neurons of the striatum as well as interneurons in the olfactory bulb (Deacon et al., 1993; Olsson et al., 1995, 1998; Wichterle et al., 2001). Despite that both striatal projection neurons and olfactory bulb interneurons derive from the LGE, they exhibit different temporal profiles of neurogenesis in the rodent; with the striatal neurons being generated almost exclusively at embryonic time points while the olfactory bulb interneurons begin their genesis at later embryonic time points and continue into the early postnatal period, when the vast majority are born (Hinds, 1968; Bayer and Altman, 2004; Batista-Brito et al., 2008). Recent studies have suggested that these two neuronal subtypes derive from separate progenitors located in distinct regions within the LGE termed the ventral (v)LGE and dorsal (d)LGE, respectively (Yun et al., 2001, 2003; Stenman et al., 2003; Waclaw et al., 2006). Yun et al. (2001) first described these two LGE subdivisions based on gene expression patterns at midgestation stages. The dLGE was characterized by high levels of Gsx2 and Er81 in progenitors of the ventricular zone (VZ), while the vLGE lacks Er81 expression and displays lower levels of Gsx2. These compartments can also be identified in the subventricular zone (SVZ) and mantle regions of the LGE. Isl1 (Isl1) is expressed in the vLGE SVZ and its striatal projection neuron derivatives whereas Er81 and Sp8 mark the dLGE SVZ and remain expressed in distinct subtypes of olfactory bulb interneurons (Stenman et al., 2003a; Waclaw et al., 2006; Allen et al., 2007; Saino-Saito et al., 2007). These two LGE progenitor domains are bordered dorsally by the ventral pallium (marked by Dbx1) and ventrally by the interganglionic sulcus, which is marked by Nkx6.2 expression (Stenman et al., 2003b). These progenitor domains have been shown to contribute to amygdalar projection neurons and cortical interneurons, respectively (Hirata et al., 2009; Sousa et al., 2009).
Correct patterning of the vLGE and dLGE requires Gsx2 and Pax6 gene function. The loss of the pallial regulator Pax6 results in a dorsal expansion of dLGE markers (Toresson et al., 2000; Stoykova et al., 2000; Yun et al., 2001; Stenman et al., 2003a; Kroll and O'Leary, 2005; Waclaw et al., 2006), suggesting a role for the paired homeodomain factor in repressing dLGE identity within pallial progenitors. In the absence of Gsx2, the vLGE and dLGE as well as their derivatives, the striatal projection neurons and olfactory bulb interneurons, are severely reduced (Corbin et al., 2000; Toresson et al., 2000; Toresson and Campbell, 2001; Yun et al., 2001, 2003; Stenman et al., 2003a; Waclaw et al., 2004, 2006). The specific role of Gsx2 in patterning and specification of the vLGE and dLGE, however, remains somewhat unclear. So far, no evidence has been provided to support a role for Gsx2 in directly ventralizing telencephalic progenitors (Corbin et al., 2000). Loss-of-function studies, however, suggested that Gsx2 indirectly controls LGE specification by repressing the expression of dorsal telencephalic regulators such as Pax6 in LGE progenitors (Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001). This appears to be a conserved function of Gsx2 since the Drosophila homologue Ind (intermediate neuroblasts defective) has also been shown to repress eyeless, the Pax6 homologue, in fly CNS development (Von Ohlen et al., 2007).
We have re-examined the role of Gsx2 in LGE specification using conditional gain-of-function and loss-of-function approaches in mice. These models afforded the analysis of temporally distinct roles for Gsx2 in the specification of vLGE and dLGE. Our results demonstrate that Gsx2 can directly ventralize pallial progenitors and depending on the developmental stage, specifies different neuronal fates. In particular, at early stages of telencephalic development Gsx2 is necessary and sufficient to correctly specify the vLGE and its major derivatives the striatal projection neurons, however, at later stages, high levels of Gsx2 specifies LGE progenitors towards dLGE fates including olfactory bulb interneurons.
The homeobox gene Gsx2 is first expressed in progenitors of the presumptive LGE between embryonic days (E)9 and E10 (Toresson et al., 2000; Corbin et al., 2003). Initially, this factor is expressed at high levels in cells throughout the presumptive LGE, with an apparent ventral to dorsal gradient in Gsx2-positive cell numbers (Figure 1A). Gsx2-positive cells are also scattered somewhat uniformly throughout the adjacent medial ganglionic eminence (MGE), as marked by Nkx2.1 expression (Figure 1A). At early stages (e.g. E11), Isl1-expressing cells are found in both the MGE and LGE. Within the LGE, Isl1 cells appear to emerge from a broad portion of the presumptive LGE (Figure 1B), however, Sp8 is seen in scattered cells adjacent to the dorsal-most portion of the LGE (Figure 1B). At later embryonic stages (e.g. E12.5 onward), Gsx2 exhibits a clear graded pattern of expression, with low levels found in cells located ventrally and the highest levels confined to those in the most dorsal portion of the LGE (Figure 1C and H). Previously, Yun et al. (2001) used this high level of Gsx2 along with other markers to define the dLGE. Indeed, Sp8 positive cells which mark the dLGE SVZ adjoin the high level Gsx2 expressing cells in the VZ (Figure 1D), while the Isl1 (vLGE) SVZ cells are positioned adjacent to the low level Gsx2-expressing cells of the VZ (Figure 1E). This dynamic expression pattern might suggest that high levels of Gsx2, at early stages of forebrain development (i.e. E9-11), are capable of specifying both vLGE and dLGE identity with a predominance for vLGE and thus striatal projection neuron fate. At later stages (i.e. E12 onward), however, high levels of Gsx2 in LGE progenitors may primarily promote dLGE fates, including olfactory bulb interneurons.
In order to test this temporal specification model, we developed a doxycyclin (Dox)-regulated binary transgenic system for the spatially and temporally controlled misexpression of Gsx2 in the embryonic mouse telencephalon (Figure 1F). We generated tetO-Gsx2-IRES-EGFP (IE) mice to use with Foxg1tTA/+ mice (Hanashima et al., 2002) to drive expression of Gsx2 in the embryonic telencephalon. Foxg1tTA/+;tetO-Gsx2-IE embryos expressed EGFP throughout the telencephalon (Figure 1G) and this expression was detected as early as E9.5 (data not shown). In addition to the EGFP, Gsx2 was also expressed in telencephalic progenitors rather uniformly both in the ventral and dorsal telencephalon (Figure 1I and J). Multiple lines of tetO-Gsx2-IE mice were identified that were responsive to the Foxg1tTA/+ with broad telencephalic expression of the transgenes. While most of these lines responded robustly to the tTA, the tetO-Gsx2-IE line showing the most reliable and robust response with no leaky expression of the transgenes, was used in the present study. All controls shown are tetO-Gsx2-IE single transgenic embryos, however, the Foxg1tTA/+ single transgenic embryos were identical with respect to the markers examined despite the fact that they only have one functional allele of Foxg1 (data not shown).
As mentioned above, EGFP could be detected as early as E9.5 in the Foxg1tTA/+;tetO-Gsx2-IE embryos and thus these animals represented an early misexpression of Gsx2 throughout the telencephalon. To determine whether ectopic Gsx2 could alter dorsal-ventral patterning, we examined the expression of the ventral telencephalic regulators Ascl1 (Mash1) and Dlx proteins at E12.5 (Anderson et al., 1997; Casarosa et al., 1999; Horton et al., 1999). Both Ascl1 (Figure 2B) and Dlx proteins (Figure 2E) were found ectopically expressed (as marked by EGFP) within the dorsal telencephalon (Figure 2C and F). Ascl1 was upregulated in the VZ of the pallium while the Dlx proteins were most highly expressed in the SVZ of the pallium, similar to their patterns of expression in the ventral telencephalon (Figure 2A and D). In contrast, the pallial regulators Pax6 (Figure 2H and I) and Tbr1 (Figure 2K and L) were both reduced in the dorsal telencephalon as compared to the control embryos (Figure 2G and J). All Foxg1tTA/+;tetO-Gsx2-IE embryos examined showed similar expression patterns of Ascl1, Dlx, Pax6 and Tbr1 proteins even at later stages (i.e. E14.5 and E15.5), however, analysis of the double transgenic brains after E15.5 was not performed as the morphology became very disrupted (Figure S1). We never observed ectopic expression of the MGE marker Nkx2.1 and, in fact, found that it was reduced or missing in all the double transgenic embryos examined (Figure S1H). These findings therefore suggest that early misexpression of Gsx2 specifically promotes LGE identity within pallial progenitors.
To study the role of Gsx2 in the specification of the vLGE versus the dLGE we examined Isl1 and Sp8 expression, respectively. Interestingly, the Foxg1tTA/+;tet-O-Gsx2-IE embryos were found to express Isl1 throughout the dorsal-ventral aspect of the telencephalon both at E14.5 (Figure 2N) and at E15.5 (Figure S1F). Conversely, only scattered Sp8 cells were detected in the pallium at either stage (Figures 2P and S1G). In all cases, the normal dLGE expression domain of Sp8 in the ventral telencephalon of double transgenic embryos was severely reduced. Taken together with the findings above, our data indicate that early (i.e. from E9.5 onward) misexpression of Gsx2 favors vLGE specification within the majority of telencephalic progenitors.
The widespread over-expression of Gsx2 (and EGFP) observed in the Foxg1tTA/+;tet-O-Gsx2-IE embryos shown in Figures 1--22 does not allow us to conclude whether Gsx2 regulates telencephalic gene expression in a cell autonomous manner or if it induces extrinsic factors, which subsequently control the patterns of gene expression in adjacent progenitors. In this respect, on very rare occasions, we obtained double transgenic embryos that displayed mosaic transgene expression (Figure S2A and B). EGFP and Gsx2 were expressed in what appeared as radial clones and specifically within these clones Pax6 was repressed (Figure S2E and F) while Ascl1 and Dlx were cell autonomously induced (Figure S2C and D). As was the case with the uniformly over-expressing embryos (see Figure 2), the mosaic embryos also showed a preferential induction of Isl1 over Sp8 in the EGFP-positive cells (Figure S2G and H), indicating that early misexpression of Gsx2 favors vLGE specification within telencephalic progenitors in a cell autonomous manner.
To examine a later role for Gsx2 in the specification of vLGE and dLGE we made use of the fact that doxycyclin (Dox) can repress transgene expression in the Foxg1tTA/+;tetO-Gsx2-IE embryos (Figure 1G). By adding Dox (0.02 mg/ml) in the drinking water of the pregnant dams from E7 to E9 we found that transgene expression was repressed until around E12.5 (Figure 3F) and not until E13.5 was the transgene expressed throughout the telencephalon (Figure 3I), in a manner similar to endogenous Foxg1. As mentioned above, when no Dox is administered the transgene is first detected at E9.5, with robust expression between E10.5 and E13.5 (Figure 3B, E and H). The tetO-Gsx2-IE embryos served as controls for these studies and do not show ectopic expression at any stage examined (Figure 3A, D and G). At E13.5, the level of Gsx2 and EGFP expression is considerably lower in the Dox-treated embryos (Figure 3I, L and M) than in the double transgenic embryos which were not administered Dox (Figure 3H, J and K). By E15.5, however, the Gsx2 transgene appears to be expressed at similar levels to that in non-Dox-treated embryos (Figure S1E and I). A dramatic improvement in the morphology of brains was observed at later stages in the Dox-treated double transgenic embryos (Figures 4 and S1I-L) as compared to the untreated double transgenic embryos (Figures 2N, P and Figure S1E-H).
Concomitant with the induction of Gsx2 and EGFP around E13.5 in the Dox-treated Foxg1tTA/+;tetO-Gsx2-IE embryos, we observed an upregulation of Ascl1 and Dlx proteins in the pallium (data not shown). Tbr1 expression in the Dox-treated double transgenics at E13.5 was only minimally reduced from the control embryos (Figure 4A and F), unlike the case in non-Dox-treated embryos (i.e. early misexpression of Gsx2) (Figure 2J and K). At E15.5, the Tbr1 expression domain in Dox-treated embryos is not increased from that seen two days earlier (compare Figure 4G with F) suggesting that the production of these cells ceases subsequent to Gsx2 misexpression. In fact, many breaks in the Tbr1 expression domain were evident in the Foxg1tTA/+;tetO-Gsx2-IE embryos and in each case, these Tbr1 negative areas were found to express the EGFP and Gsx2 transgenes at high levels (see inset in Figure 4G). Again, the areas misexpressing Gsx2 and EGFP were found to be Ascl1 and Dlx expressing (data not shown). As was the case in the early misexpression embryos, no ectopic Nkx2.1 was detected. However, in the Dox-treated double transgenic (i.e. delayed misexpression) embryos much more Nkx2.1 expression remained in the MGE region as compared to their early misexpression (non-Dox-treated) counterparts (Figure S1L). Thus it appears that even when Gsx2 is expressed in progenitors of the pallium from E13.5 it is capable of respecifying them towards LGE fates.
To determine if the delayed misexpression of Gsx2 differentially regulates the specification of the vLGE versus the dLGE, we examined the expression of Isl1 and Sp8, respectively. At E15.5, Dox-treated embryos showed a dramatic reduction in the Isl1 expression domain (Figure S1J) and concomitant expansion both ventrally and dorsally of Sp8 expression (Figure S1K). This was opposite to that observed in the double transgenic embryos that were not Dox-treated (i.e. early misexpression; Figure S1F and G), despite the fact that similar levels of Gsx2 expression were observed in both Dox-treated and non-Dox-treated embryos (Figure S1E and I). As mentioned above, the Dox-treated Foxg1tTA/+;tetO-Gsx2-IE embryos exhibit improved morphology at later stages and thus we were able to examine vLGE and dLGE development at E18.5. By this stage, Isl1 marks the major vLGE derivative, i.e. the forming striatal complex (Figure 4C). In the Dox-treated double transgenics, Isl1 expression and accordingly the size of the forming striatum is greatly reduced (Figure 4H). Conversely, Sp8 expression is up-regulated throughout the developing cerebral cortex suggestive of dLGE respecification (Figure 4I and J). Interestingly, as seen at E15.5 in delayed misexpression embryos, Sp8 expression is expanded ventrally in the LGE also at E18.5. This could indicate that vLGE cells have been respecified to dLGE identity by the delayed misexpression of Gsx2, however, it is possible that selective proliferation of dLGE progenitors accounts for the enlarged Sp8 expression domain. The Foxg1tTA/+;tetO-Gsx2-IE brains (with or without Dox treatment) do not form normal olfactory bulbs and therefore it is difficult to fully assess olfactory bulb interneuron development in these animals. We can conclude, however, that the specification of these dLGE derivatives is exclusively promoted by delayed misexpression of Gsx2. In summary, early misexpression (i.e. E9-10) of Gsx2 appears to favor vLGE specification while later misexpression (i.e. E13 and onward) promotes dLGE specification, apparently at the expense of the vLGE.
To complement the gain-of-function experiments described above, we have taken a conditional loss-of-function approach to study the temporal role of Gsx2 in the specification of LGE fates. We generated a conditional mutant allele (Gsx2flox) by engineering loxP sites flanking exon 2, which includes the entire homeodomain (Figures 5A and S3). Gsx2flox/flox mice appear normal and express wild type levels of Gsx2 protein (data not shown), suggesting the floxed allele acts in a similar manner as the wild type Gsx2 allele. To determine if our recombined conditional allele produced a loss of function mutation of Gsx2, we crossed Gsx2flox/+ mice with mice carrying an EIIA-cre transgene (Lasko et al., 1996), which recombines in the germline, to generate mice carrying a globally recombined allele (RA) of Gsx2 (Gsx2RA/+). Gsx2RA/RA mutant embryos do not express Gsx2 protein (Figure S3D). In addition, Gsx2RA/RA mutant embryos displayed an identical phenotype in the telencephalon (data not shown) as the previously described mutation of Gsx2 (Szucsik et al., 1997; Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001).
To conditionally inactivate Gsx2, we crossed Gsx2flox/flox mice with mice that were double heterozygous for the floxed allele (Gsx2flox/+) and a cre knock-in to the Emx1 locus (Emx1cre/+) (Gorski et al., 2002). The Emx1cre mice are widely used to recombine floxed genes in the dorsal telencephalon, however, these mice also induce recombination in scattered cells of the LGE (see Figure 1F in Gorski et al., 2002). Our initial plan was to use the Emx1cre/+ mice to produce a mosaic inactivation of Gsx2 in the LGE in order to study cell autonomous versus non-cell autonomous requirements as a complement to our mosaic misexpression (Figure S2). Indeed, recombination of the ROSA26 reporter (R) (Mao et al., 1999) was detected as early as E10.5 in the presumptive LGE (insets in Figure 5B and C) and at that stage Gsx2 showed a slight mosaic recombination pattern in the conditional mutants (Emx1cre/+;Gsx2flox/flox;R26R) (compare Figure 5C to 5B). A clear mosaic inactivation of Gsx2 within the LGE was evident by E12.5 when nearly half of all Gsx2 staining was lost (compare Figure 5E to 5D, see also S4). Surprisingly, however, by E18.5, nearly 80% (430.3 ± 6.7 vs 96.3 ± 3.9 cells/section, p<0.001; n=3) of Gsx2 expressing cells were lost in the LGE of conditional mutants (Figure 5I) compared to controls (Figure 5H). This finding is quite significant because Emx1cre/+ mice have been largely characterized as pallial specific for cre recombination (Gorski et al., 2002; Willaime-Morawek et al., 2006; Willaime-Morawek and van der Kooy, 2008; Kohwi et al., 2007). Our data provide clear evidence that Emx1cre induces recombination of the Gsx2 floxed allele in the LGE beginning around E10.5. In addition, the recombination is progressively more severe in the LGE later in development, resulting in an extensive loss of Gsx2 protein expression at perinatal stages.
Germline Gsx2 mutants display abnormalities in dorsal-ventral patterning of the telencephalon as observed by a ventral expansion of pallial regulators into the LGE and a concomitant down-regulation of ventral telencephalic genes (Corbin et al., 2000; Toresson et al., 2000; Toresson and Campbell 2001; Yun et al., 2001, 2003). As mentioned above, conditional deletion of Gsx2 using Emx1cre/+ mice results in a mosaic loss of Gsx2 protein in the LGE at E12.5 (Figure 5E), which allows for the assessment of the Gsx2 mutant phenotype in a partial wild type environment. To compare mutant cells to wild type cells we crossed the R26R mouse onto our conditional mutant breeding scheme (Gsx2flox/+;Emx1cre/+× Gsx2flox/flox;R26R+/-). These experiments clearly revealed that β-Gal expression from the R26R mouse under-represents cre recombination in the LGE because the reduction in Gsx2 protein expression in conditional mutants is considerably more extensive (Figure 6B). Interestingly, the Gsx2 null regions of the LGE, particularly in the dorsal half, showed ectopic expression of pallial markers. As is the case in germline Gsx2 mutants (Yun et al., 2001, 2003; Stenman et al., 2003b), the expression of the ventral pallial marker Dbx1 (Figure 5F) was expanded throughout the LGE in Emx1cre recombined areas of Gsx2 conditional mutants at E12.5. Moreover, a number of the fate mapped cells colocalized Dbx1 (Figure 5G). High levels of Pax6 expression, normally observed at the pallio-subpallial boundary (Figure 5D) were expanded within the dorsal half of the LGE, specifically in the Gsx2 null areas (Figure 5E). In germline Gsx2 mutants, the pallial restricted markers Tbr2 and Tbr1 (Figure 6C and G) have both been shown to be ectopic within the mutant LGE (Yun et al., 2001). Accordingly, in the Gsx2 conditional mutant LGE, ectopic Tbr2 and Tbr1 expressing cells were observed in the SVZ and mantle zone of the dorsal half of the LGE, respectively, and many of these corresponded to Emx1cre fate mapped cells (Figure 6E,F and I,J). It should be noted that there were clear Gsx2 null regions in the vLGE (Figure 6B and S4), but only weak Tbr2 cells (Figure 6E and F) and no Tbr1 cells (Figure 6I and J) were observed in this region. These results indicate that Gsx2 is required cell autonomously to repress dorsal telencephalic identity in LGE progenitors (predominantly in the dLGE) until late embryonic stages.
Previous studies have shown that germline deletion of Gsx2 results in reductions in the expression of dLGE and vLGE genes that eventually leads to abnormalities in the production of olfactory bulb interneurons and striatal projection neurons (Corbin et al., 2000; Toresson et al., 2000; Toresson and Campbell 2001; Yun et al., 2001, 2003; Stenman et al., 2003a; Waclaw et al. 2004, 2006). Accordingly, dLGE identity as labeled by Sp8 expression (Figure 6K-L and S4C,I) is severely disrupted in the Gsx2 conditional mutants (Figure 6M-N and S4F,L). Unlike in the control embryos where Sp8 and β-gal (i.e. fate mapped cells) are broadly co-expressed (Figure 6L), little overlap is detected in Gsx2 conditional mutants (Figure 6N). In contrast to the dLGE, vLGE identity as labeled by Isl1 expression is not severely affected in the Gsx2 conditional mutant (compare Figure S4E,K with B,H). In fact, we were able to identify β-gal positive Emx1 fate mapped cells that also express Isl1 in the Gsx2 conditional mutant LGE (Figure 6R). These findings suggest that the vLGE may be specified by Gsx2 in LGE progenitors earlier than E10.5, since Emx1cre mediated recombination starts in the LGE around this time point. Taken together, these results suggest that Emx1cre induced mosaic recombination of Gsx2 in the LGE severely affects the establishment of dLGE identity while leaving vLGE specification largely intact.
To examine dLGE and vLGE derivatives in Gsx2 conditional mutants, we analyzed markers of olfactory bulb interneurons and striatal projection neurons in E18.5 embryos. Germline Gsx2 mutants are known to have severe defects in striatal development, notably a greater than 50% reduction in striatal size as observed by Isl1 and FoxP1 expression (Toresson and Campbell 2001; Waclaw et al. 2004). Consistent with this result, germline deletion of our Gsx2 conditional allele Gsx2RA/RA (i.e. null allele) results in a 57% reduction in striatal volume, as marked by FoxP1 expression, compared to control embryos (p<0.01; n=3) (compare Figure 7B to 7A). On the contrary, striatal volume in the Emx1cre;Gsx2flox/flox mutants was not significantly different from that in controls (Compare Figure 7C to 7A). In fact, the striatal volume of Gsx2RA/RA mutants was 53% reduced from that in the Emx1cre conditional mutants (p<0.01; n=3) (compare Figure 7B and C). At birth, most striatal projection neurons have been born but are still undergoing maturation and thus most mature markers of these cells are not yet expressed. DARPP-32, which is ultimately expressed by all striatal projection neurons, only marks the early born neurons at birth, which largely belong to the patch compartment (Foster et al., 1987). The germline Gsx2 mutants show a disproportionate reduction in DARPP-32 neurons (Corbin et al., 2000; Toresson and Campbell, 2001; Waclaw et al., 2004). Emx1cre;Gsx2 conditional mutants (Figure 7C) exhibited considerably more DARPP-32 positive cells than the Gsx2RA/RA mutants (Figure 7B) and seems to be only slightly reduced from that seen in controls (compare Figure 7A and C). Together these data support the notion that the delayed recombination of Gsx2 that is induced by Emx1cre results in minimal alterations of vLGE specification or striatal development, at least with respect to what can be examined at perinatal stages.
As mentioned above, germline Gsx2 mutants exhibit defects in dLGE specification and an attendant reduction in the generation of embryonic olfactory bulb interneurons (Corbin et al., 2000; Toresson and Campbell 2001; Yun et al., 2001, 2003; Stenman et al., 2003; Waclaw et al. 2006). To determine if Emx1cre;Gsx2flox/flox mutants displayed defects in embryonic olfactory bulb neurogenesis, we analyzed the expression of the GABA synthesizing enzyme GAD67, an established marker of many embryonic olfactory bulb interneurons (Kohwi et al., 2007; Kosaka and Kosaka, 2007; Parrish-Aungst et al., 2007). Gsx2 conditional mutants exhibited an 81% reduction in GAD67 positive cells in the E18.5 olfactory bulb compared to controls (Figure 7F and D) (p<0.01, n=3). This reduction was similar to, but not as severe as, that observed in Gsx2RA/RA mutants that show a 97% reduction compared to controls (Figure 7E and D) (p<0.01, n=3). The difference between the Gsx2RA/RA and Emx1cre conditional mutants is likely due to the remaining Gsx2 expression in the dLGE (approx. 20% of control) that is not recombined by Emx1cre/+ during development (see Figure 5). In addition to GAD67 expression we analyzed Meis2 expression, which is a broad marker of olfactory bulb interneuron subtypes (Allen et al., 2007). Meis2 is also clearly reduced in the Gsx2 conditional mutant (Figure 7I) and Gsx2RA/RA mutant (Figure 7H) compared to controls (Figure 7G). Note that there is almost a complete loss of Meis2 in the forming glomerular layer of both Gsx2 conditional mutants (Figure 7I) and germline Gsx2 mutants (Figure 7H) compared to controls (Figure 7G). These results, combined with the analysis of dLGE identity (i.e. Sp8), indicate that the delayed inactivation of Gsx2 by Emx1cre significantly affects dLGE development and the generation of olfactory bulb interneurons in a manner similar to that observed in the Gsx2RA/RA mutants.
Previous studies have indicated that the dLGE contributes to the generation of olfactory bulb interneurons (Stenman et al., 2003; Yun et al., 2003; Waclaw et al., 2006). However, recent studies have suggested that multiple telencephalic regions, including the pallium, may also produce these interneurons (Kohwi et al., 2007; Merkle et al., 2007; Ventura and Goldman, 2007; Young et al., 2007). Some of the supporting data for the pallial contribution has come from fate mapping Emx1cre cells in the olfactory bulb. As would be expected, many of the Emx1cre fate mapped cells comprise projection neurons (i.e. mitral or tufted cells) (Gorski et al., 2002), however, a significant number of recombined cells possess interneuron phenotypes (Kohwi et al., 2007; Young et al., 2007). To determine the effect of Gsx2 conditional loss-of-function in dLGE cells destined for the olfactory bulb, we crossed the CC-EGFP recombination reporter mice (Nakamura et al., 2006) onto the Emx1cre;Gsx2flox/flox mutant background. Tbr1 can be used to mark projection neurons (mitral and tufted cells) as well as their progenitors in the olfactory bulb at perinatal time points (Bulfone et al., 1998). We found that 56.2 ± 6.1% of the Emx1cre fate mapped cells located near the progenitor regions of the olfactory bulb were Tbr1 expressing (i.e. projection neuron identity) (Figure 8B and B'). In contrast, Sp8 can be used to mark many olfactory bulb interneurons, particularly those expressing calretinin (Waclaw et al., 2006), which have also been shown to be Emx1cre derivatives (Kohwi et al., 2007; Young et al., 2007). Germline Gsx2 mutants show severe reductions in Sp8 positive cells within the olfactory bulb (Waclaw et al., 2006). We observed that 34.9 ± 3.1% of the Emx1cre fate mapped cells in the control olfactory bulbs were Sp8 positive (Figure 8F and F'). In the Emx1cre;Gsx2flox/flox mutant bulb we noticed a dramatic increase in Tbr1 staining at E18.5 together with a disorganization in the typical staining pattern for Tbr1 (Figure 8C). Since this result has not been reported in the germline Gsx2 mutants, we examined the Gsx2RA/RA mutant olfactory bulbs at E18.5 and found that they also show increased numbers of Tbr1 positive cells (Figure S5B). Quantification of the Emx1cre fate mapped cells in the conditional mutant olfactory bulb showed that 77.6 ± 1.2% of the EGFP cells expressed Tbr1, which was a 38% increase from the controls (p<0.05, n=3) (Figure 8D and D'). Conversely, Sp8 showed a dramatic reduction in the Gsx2 conditional mutant olfactory bulb (Figure 8G) and only 6.0 ± 0.2% of the fate mapped mutant cells contained Sp8 staining which was an 82% reduction from that observed in controls (p<0.001, n=3) (Figure 8H and H'). Therefore, it seems that the increase of Tbr1 positive cells in the Emx1cre;Gsx2flox/flox mutant olfactory bulb occurs at the expense of the normal generation of Sp8 positive interneurons from Gsx2 dependent dLGE cells. Moreover, these findings suggest that the respecified (i.e. ectopic Tbr1 and Tbr2) cells in the conditional mutant dLGE retain the ability to migrate rostrally and populate the olfactory bulb. Indeed, using Gsx2EGFP mice (Wang et al., 2009) as a short-term fate map of cells derived from Gsx2-expressing cells, we observed many EGFP and Tbr1 co-expressing cells in the olfactory bulb of Gsx2EGFP/RA (i.e. null) mutants (Figure S5D) unlike the case in Gsx2EGFP/+ embryos (Figure S5C). Finally, our data question the assumption that Emx1cre recombination indicates pallial ancestry since we find that this cre driver significantly recombines Gsx2 in the LGE in a temporally progressive manner. This leads to a loss of dLGE specification and ultimately a fate shift in dLGE cells (normally destined to become olfactory bulb interneurons) towards pallial fates.
Our results show for the first time that Gsx2 is capable of directly ventralizing pallial progenitors and promoting LGE neuronal cell fate specification. Interestingly, this homeobox transcription factor appears to function within a temporal framework to specify predominantly striatal projection neuron (i.e. vLGE) fates at early stages of LGE neurogenesis and subsequently dLGE fates (i.e. olfactory bulb interneurons) at later stages of development. Accordingly, our conditional loss-of-function studies demonstrate that at late stages of development, Gsx2 is required for the specification of dLGE fates. Indeed, in the conditional Gsx2 mutants we found that many of the cells normally fated to become olfactory bulb interneurons appeared to be respecified towards pallial fates, as marked by Tbr1 expression. Taken together, our findings indicate that the dorsal-ventral patterning activities of Gsx2 function within an independently controlled temporal framework.
Gsx2 is known to be required for correct dorsal-ventral patterning in the telencephalon (Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001), however, it has been thought to play an indirect role in this process. In fact, viral misexpression of Gsx2, to examine its ability to impart ventral telencephalic identity within early pallial progenitors, did not support a ventralizing role (Corbin et al., 2000). These authors reported that neither Ascl1 or Dlx genes were induced nor was Pax6 expression in the dorsal telencephalon reduced by ectopic Gsx2. They did, however, find that the LGE-specific radial glial marker retinol binding protein (RBP-I) (Toresson et al., 1999) was moderately up-regulated in some of the infected pallial progenitors. These findings are in stark contrast to those presented here. It could be that widespread ectopic Gsx2 (as shown here) is required to ventralize pallial progenitors, essentially representing a “community effect”. However, on rare occasions, we obtained Foxg1tTA;tet-O-Gsx2-IE embryos that exhibited mosaic expression of Gsx2 which appeared similar to the pattern observed after viral transductions. The effects of Gsx2 misexpression in small clones of EGFP expressing cells were identical to that observed in the embryos showing widespread misexpression thus indicating a cell autonomous function for Gsx2 in repressing dorsal identity and promoting LGE specification. The discrepancy between the present results and those of Corbin et al. (2000), could be due to down-regulation of virally misexpressed Gsx2 protein prior to the assay points. If this was the case, the induced RBP-I that Corbin et al. (2000) observed may represent a very sensitive readout of transient Gsx2 activity, while the regulation of Pax6, Ascl1 and Dlx genes may require sustained Gsx2 expression.
Although viral misexpression of Gsx2 did not drastically alter the development of cortical progenitors (Corbin et al., 2000), it was found to repress Nkx2.1 in the MGE when delivered at E9.5 (Corbin et al., 2003). This was similar to our results from early misexpression of Gsx2 where Nkx2.1 was reduced or missing in the MGE. Interestingly, delayed misexpression of Gsx2 was not efficient at repressing Nkx2.1 in the MGE, indicating that by E13.5 MGE progenitors are resistant to Gsx2 influence. Despite this, the present findings clearly show that using the binary transgenic system described here, ectopic Gsx2 robustly and reproducibly promotes LGE specification in pallial progenitors at both early and late time points.
Gsx2 is expressed in a temporally dynamic fashion within the VZ progenitors of the LGE. At early stages, it is expressed at high levels in progenitors throughout the presumptive LGE, while at later stages high-level expression is confined to VZ cells in the dLGE. The early (E9-11) high-level expression of Gsx2 correlates with the appearance of Isl1 cells throughout most of the LGE with limited numbers of Sp8 expressing cells in the dorsal-most region. This is not the case at later stages (E12 onward), when only Sp8 cells are found in close association with high-level Gsx2 expressing LGE (i.e. dLGE) cells. This suggests that Gsx2 can specify both vLGE and dLGE at early stages of LGE neurogenesis, however, vLGE specification appears to be favored. Our early misexpression data support this notion. In contrast, high levels of Gsx2 at later stages (i.e. E12 onward) may exclusively promote dLGE fates. Indeed, when Gsx2 misexpression is delayed by Dox treatment to around E13, only dLGE specification is observed. Moreover, in the ventral telencephalon, this appears to be at the expense of the vLGE (at least its most dorsal portion). These results correspond well with the fact that striatal projection neurons commence neurogenesis at an earlier time point than the olfactory bulb interneurons (Bayer and Altman, 2004).
Previous studies using germline Gsx2 mutants have shown that this factor is required for the correct development of both the striatal projection neurons and olfactory bulb interneurons (Corbin et al., 2000; Toresson et al, 2000, Toresson and Campbell, 2001; Yun et al., 2001, 2003). The fact that delayed inactivation of Gsx2 using Emx1cre preserves striatal specification, indicates that correct striatal development depends largely on early (i.e. before E10.5) Gsx2 expression. Conversely, dLGE and olfactory bulb interneuron development requires Gsx2 function at least until birth. Taken together, our gain-of-function and loss-of-function results suggest that vLGE fate is specified earlier than dLGE fate and that this specification occurs in a sequential manner. This does not appear to be the case in the developing spinal cord, where Gsx1/2 are required for the normal production of both early and late generated excitatory interneurons that share similar transcription factor profiles (Mizuguchi et al., 2006). Thus, our results reveal a novel role for Gsx2 in the temporal production of molecularly distinct neuronal subtypes from LGE progenitors.
Given the protracted period of neurogenesis in the developing telencephalon, it is not surprising that temporally regulated mechanisms would also be involved in the generation of neuronal diversity within this brain region. In fact, previous transplantation studies have demonstrated a progressive restriction in the fate potential of cortical progenitors. Those transplanted from early stages are capable of generating neurons that populate both the deep and superficial cortical layers (McConnell, 1988), while the later stage progenitors are limited to populating the superfical layers, even if back transplanted to an earlier cortical environment (Frantz and McConnell, 1996; Desai and McConnell, 2000). A likely explanation for this finding is that the molecular nature of telencephalic progenitors changes over time and that this constrains fate specification accordingly. In this respect, Hanashima et al. (2004) previously investigated the role of Foxg1 in the temporal production of different cortical neuron subtypes. They found that Foxg1 suppresses the generation of Cajal-Retzius neurons, the earliest cell type generated in the cerebral cortex. Moreover, these authors demonstrated that conditional inactivation of Foxg1 in progenitors that normally produce deep layer cortical neurons results in the production of Cajal-Retzius neurons. A subsequent study by Muzio and Mallamaci (2005), suggested that Foxg1 may regulate areal patterning and through suppression of Cajal-Retzius neurogenesis in most of the pallium, restrict the production of these early-born neurons to the dorsal medial telencephalon. Cortical progenitors isolated in vitro have recently been shown to follow the same schedule of neuron generation as their in vivo counterparts, indicating that the factors responsible for progressive restriction in developmental potential are cell-intrinsic (Shen et al., 2006). Furthermore, these authors demonstrated that knock-down of Foxg1 in midgestation (e.g. E12) cortical progenitors actually appeared to reset the timing of neuron generation such that early fates (i.e. Cajal-Retzius neurons) were first generated followed by later cortical fates. Interestingly, late stage cortical progenitors (e.g. E15) were not reverted in their developmental potential by the knock-down of Foxg1 (Shen et al., 2006). Thus Foxg1 plays an important role in restricting the developmental potential of cortical progenitors, specifically by repressing the earliest cortical neuron fate. However, factors that restrict the fate of cortical progenitors at later embryonic stages have yet to be identified.
Although evidence for the temporal restriction of developmental potential has been demonstrated for dorsal telencephalic progenitors, our genetic data indicate, for the first time, that similar restrictions may also occur in the ventral telencephalon. Importantly, the response of telencephalic progenitors to misexpressed Gsx2, at early versus late developmental stages, was similar regardless of their location in the pallium or LGE. This suggests that similar temporal patterning mechanisms may regulate the progressive restriction of developmental potential in both pallial and LGE progenitors. It may be that in order to specify vLGE fates, Gsx2 needs to co-operate with a factor(s) that are restricted in their telencephalic expression to early time points. In this way, down-regulation of such a factor(s) at later stages would limit Gsx2 to specifying dLGE fates exclusively. At present, the identity of such a factor(s) is unknown. Therefore, it seems important to perform molecular profiling studies on telencephalic progenitors, over time, in order to identify candidate factors involved in regulating this process. Another mechanism that could regulate progressive restriction in developmental potential within pallial and LGE progenitors is cell cycle length, as suggested previously (Shen et al., 2006). In fact, it seems that the cell cycle length of dorsal telencephalic progenitors increases as corticogenesis proceeds and this increase is rather specific to G1 (Caviness et al., 2003; Calegari et al., 2005). Thus it could be that the duration of Gsx2 activity during G1 is fundamental to its role in specifying dLGE versus vLGE at different stages of development. Finally, it is possible that progressive restriction in developmental potential is also regulated in SVZ progenitors downstream of Gsx2 such that the induction of Sp8 leads to a repression of Isl1 and thus vLGE identity, or vice versa. We are currently performing gain-of-function studies to address this possibility.
The conditional mutant analysis shown here was facilitated by the Emx1cre mice (Gorski et al., 2002). These mice have been widely used to recombine floxed alleles within pallial progenitors, however, in their initial characterization the authors noted recombination also within scattered cells of the LGE (Gorski et al., 2002). Furthermore, they found recombined cells within interneuron layers of the olfactory bulb as well as dispersed cells in the striatum. The Emx1cre derived cells in the striatum expressed the calcium binding protein, calbindin (which marks the projection neurons of the matrix (Gerfen et al., 1987)), but not markers of striatal interneuron subtypes. The existence of Emx1cre recombined cells in the striatum and interneuron layers of the olfactory bulb has recently been used to argue for a pallial contribution to ventral telencephalic neuronal diversity, particularly within the striatal projection neuron and olfactory bulb interneuron populations (Willaime-Morowak et al., 2006; Kohwi et al., 2007; Young et al., 2007; Williame-Morawek and van der Kooy, 2008). While our data does not exclude this possibility, it supports a simpler explanation whereby Emx1cre is expressed in LGE progenitors and by virtue of this, labels significant numbers of striatal projection neurons and olfactory bulb interneurons. Indeed, we show recombination in the presumptive LGE beginning as early as E10.5 commensurate with a slight reduction of Gsx2. By E12.5, large portions of the Gsx2 expression domain in the LGE are recombined and surprisingly at late stages, nearly 80% of the Gsx2 cells are gone from the LGE. The kinetics of this recombination were well suited to our questions concerning temporal constraints on Gsx2 in LGE specification. Indeed, the delayed recombination of Gsx2 that was facilitated by the Emx1cre mice revealed that vLGE development was largely preserved but that the dLGE specification depends on late stage Gsx2 expression. Indeed, no significant changes were detected in the striatum of the Emx1cre conditional mutants at birth, while the reductions in olfactory bulb interneuron markers in these mutants were reduced to levels similar to that seen in the germline Gsx2 mutants. The dorsal septum has also been suggested to give rise to olfactory bulb interneurons at perinatal stages (Merkle et al., 2007). Interestingly, Gsx2 is expressed in the dorsal septum and was observed to be recombined in the Emx1cre conditional mutants (Figure 6B), suggesting that Gsx2 may also be required for septal-derived olfactory bulb interneuron generation.
To trace the mutant cells in the LGE, striatum and olfactory bulb we used the recombination reporters R26R (Mao et al., 1999) or CC-EGFP (Nakamura et al., 2006) in combination with the Emx1cre;Gsx2flox/flox alleles. We consistently found that the fate mapped cells grossly under-represent the number of recombined LGE cells (as indicated by the loss of Gsx2 expression in the mutants). The mutant cells in the dLGE lost the expression of Sp8 and at the same time up-regulated the pallial markers Pax6, Tbr2 and Tbr1. Surprisingly, Emx1cre fate mapped, mutant cells in the vLGE did not appear to change fate, since they retained expression of Isl1, suggesting that vLGE fate was specified in these cells prior to cre activity (i.e. before E10.5). In the conditional mutant olfactory bulb, a dramatic decrease in Sp8 positive, Emx1cre fate mapped cells was observed coincident with a near doubling of the number of fate mapped mutant cells that express the pallial marker Tbr1. The expansion of pallial markers into the Gsx2 mutant LGE has previously been suggested to indicate a respecification towards ventral pallial fates (Yun et al., 2001, 2003; Stenman et al., 2003b). While ventral pallial progenitors have been proposed to give rise to projection neurons of the amygdala and pyramidal cortex (Fernandez et al., 1998; Puelles et al., 1999, 2000; Stenman et al., 2003c; Hirata et al., 2009), no reports to date suggest that these progenitors migrate rostrally to populate the olfactory bulb. Thus, it may be that the respecified (i.e. Tbr1-positive) cells in the dLGE of Gsx2 conditional mutants retain the ability to migrate rostrally as they would normally do despite their apparent pallial respecification.
In summary, we show here that Gsx2 is sufficient to specify LGE fates in pallial progenitors. Additionally, our results indicate that Gsx2 functions within an independently regulated temporal framework to specify vLGE and dLGE neuronal fates at distinct time points in telencephalic development.
TetO-Gsx2-IE mice were generated by pronuclear injection and characterized as described in the Supplemental Information. Doxycycline (Dox) was used to repress transgene expression in Foxg1tTA/+;tetO-Gsx2-IE embryos. Specifically, doxycycline hyclate (Sigma) given to pregnant mice in the drinking water at 0.02 mg/ml beginning at E7 (morning of vaginal plug defined as E0.5) and removed approximately 48 hours later (i.e. E9). This Dox dose was the lowest that we found to completely repress transgene expression in Foxg1tTA/+;tetO-Gsx2-IE embryos.
A conditional allele of Gsx2 was generated by homologous recombination in W4 ES cells (kindly provided by A. Joyner). Briefly, the floxed allele was made by flanking the second exon of Gsx2 with loxP sites as described in the Supplemental Information (Figure 5A and S3). Chimeric mice were generated from two correctly targeted clones (7-B11 and 7-F3) by the transgenic core at CCHMC. All experiments were performed on the 7-B11 line.
β-actin-FLPe (enhanced Flpase) mice, R26R mice, EIIA-cre and Emxcre mice were obtained from the Jackson Laboratory, Bar Harbor, ME and genotyped following protocols from the Jackson Laboratory. CAG-CAT(CC)-EGFP were genotyped as described in Nakamura et al., (2006). See Supplementary Information for genotyping details concerning the Foxg1tTA/+, tetO-Gsx2-IE and Gsx2flox mice and embryos. For staging of embryos, the morning of vaginal plug detection was designated embryonic day (E) 0.5. Embryos and postnatal brains were fixed overnight in 4% paraformaldehyde, rinsed thoroughly in PBS and cryoprotected in 30% sucrose in PBS before sectioning at 12μM on a cryostat.
Quantification for each experiment was done on three different embryos of each genotype. GAD67+ cells in the GCL were counted in three olfactory bulb sections at E18.5 in control embryos, Gsx2 null embryos, and Gsx2 conditional mutant embryos. Striatal size of control embryos, Gsx2 null embryos, and Gsx2 conditional mutants was determined by quantifying the FoxP1 expression area using the NIH ImageJ program. Statistics were performed between control embryos, Gsx2 null embryos, and Gsx2 conditional mutants using a one-way ANOVA with a Tukey post hoc test. Gsx2+ cells in the LGE were counted at 400× magnification. Each genotype was analyzed at identical rostral and caudal levels. For double staining quantification of Emx1cre/+ fate mapped cells (EGFP) and either Sp8+ or Tbr1+ cells, double and single labeled cells in the GCL were counted in two separate areas (400× magnification) of two olfactory bulb sections for each genotype (at least 100 EGFP+ cells per embryo). Statistics were performed between control and Gsx2 conditional mutants using a Student's unpaired t test.
We thank Masato Nakafuku for helpful suggestions on the manuscript and both Richard Converse and Claire Schreiner for excellent technical assistance as well as Stewart Anderson and Eseng Lai for providing the Foxg1tTA mice and Jeff Whitsett for providing the pTetO7 construct. We also wish to thank A. Buchberg, T. Edlund, J. Johnson, T Jessell, J. Kohtz. E.E. Morrisey and A. Pierani for providing antibodies. This work was supported by NIH grants NS044080 and MH069643. R.R.W. is supported by an NIH training grant (HD046387).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.