Distinct cortical interneuron (CIN) subtypes have unique circuit functions; dysfunction in specific subtypes is implicated in neuropsychiatric disorders. Somatostatin- and parvalbumin-expressing (SST+ and PV+) interneurons are the two major subtypes generated by medial ganglionic eminence (MGE) progenitors. Spatial and temporal mechanisms governing their cell-fate specification and differential integration into cortical layers are largely unknown. We provide evidence that Coup-TF1 and Coup-TF2 (Nr2f1 and Nr2f2) transcription factor expression in an arc-shaped progenitor domain within the MGE promotes time-dependent survival of this neuroepithelium and the time-dependent specification of layer V SST+ CINs. Coup-TF1 and Coup-TF2 autonomously repress PV+ fate in MGE progenitors, in part through directly driving Sox6 expression. These results have identified, in mouse, a transcriptional pathway that controls SST-PV fate.
Cortical interneuron (CIN) subtypes play different roles in cortical computations through their distinct molecular properties, physiology and connectivity (Kepecs and Fishell, 2014). For example, parvalbumin-expressing (PV+) fast-spiking, soma-innervating interneurons drive gamma oscillations, while somatostatin-expressing (SST+) regular-spiking interneurons target pyramidal neuron dendrites, where they are crucial for sharpening stimuli intensity (Sohal et al., 2009; Murayama et al., 2009). Defects in interneuron subtypes may underlie some phenotypes of neuropsychiatric disorders, such as autism and schizophrenia (Marin, 2012).
A fundamental issue is how correct numbers (and ratios) of interneuron subtypes are generated and integrated into appropriate cortical layers. The mouse medial ganglionic eminence (MGE) produces most of the PV and SST interneurons, which represent about 40% and 30% of CINs, respectively (Xu et al., 2008). These numbers are controlled by regional, temporal and cell-type specification, differentiation and survival, and are regulated by cascades of transcription factors (TFs), including Dlx1, Dlx2, Dlx5, Dlx6, Lhx6, Lhx8, Nkx2.1, Nkx6.2, Npas1, Npas3, Olig1, Satb1 and Sox6 (Azim et al., 2009; Batista-Brito et al., 2009; Cobos et al., 2005; Colasante et al., 2008; Liodis et al., 2007; Sussel et al., 1999; Zhao et al., 2008). Non-transcription factors, such as Ccnd2 (cyclin D2) and Pten, also control interneuron numbers (Glickstein et al., 2007; Vogt et al., 2015).
We focused on Coup-TF2 (Nr2f2) function (steroid/thyroid hormone receptor superfamily member) because it, and its paralogue Coup-TF1 (Nr2f1), are expressed in the MGE, and are largely confined to layer V SST+ CINs (Cai et al., 2013; Lodato et al., 2011). Previous work on Coup-TF1 and Coup-TF2 in the embryonic basal ganglia focused on their function in the caudal ganglionic eminence (CGE) and amygdala (Lodato et al., 2011; Tang et al., 2012), although loss of Coup-TF1 decreased SST/PV balance (Lodato et al., 2011). Here, through conditional mutagenesis, forced expression and transplantation we demonstrate that Coup-TF2, in conjunction with Coup-TF1, regulate cell cycle dynamics and cell fate decisions, including apoptosis, in MGE progenitors, but not in neurons. We show that Coup-TF1 and Coup-TF2 (Coup-TF1/2) function in a spatial- and temporal-specific manner to promote SST and repress PV interneurons. Furthermore, we provide evidence that Coup-TF2 directs the development of layer V SST+ interneurons, in part through driving Sox6 expression via an enhancer element near the Sox6 locus. Finally, we show that the caudal MGE is also a major source of SST+ CINs.
Coup-TF2 loss in the Nkx2.1-Cre lineage reduces the neocortical SST+/PV+ interneuron ratio
COUP-TF2 expression in the MGE at E12.5 is highly graded, with prominent levels caudoventrally, extending towards the CGE and into the preoptic area (POA) where sonic hedgehog (Shh) expression in the VZ is present (Fig. 1A; Fig. S1H,I). Rostrally, COUP-TF2 is restricted to the rostral dorsal MGE (rdMGE), where it overlaps with Nkx6.2 expression (Fig. S1G,H). Its graded expression is opposite to that of Otx2, which has a rostral-caudal gradient (Fig. S1B-E). High COUP-TF2 levels are also present in the CGE, caudoventral cortex and amygdala (Tang et al., 2012; Kanatani et al., 2008; Tripodi et al., 2004) (Fig. S15).
Coup-TF2 expression is complementary to Nkx2.1 in the rdMGE and caudal MGE (cMGE) (Fig. S1H,J). We therefore hypothesized that Nkx2.1 restricts Coup-TF2 expression within the MGE. To test this, we compared Coup-TF2 expression in E12.5 Nkx2.1 conditional knockout mutants (Olig2-Cre+; Nkx2.1lox/lox) with littermate controls (Olig2-Cre+; Nkx2.1lox/+). Coup-TF2 expanded ventrally in the rostral MGE (Fig. S1K,K′) and its expression increased in the medial MGE (mMGE) (Fig. S1L,L′). These results indicate that Nkx2.1 regionalizes Coup-TF2 MGE expression.
We investigated Coup-TF2 function in MGE- and POA-derived cells using conditional mutagenesis (Fig. S14A). Because Coup-TF2 has overlapping functions with Coup-TF1 (Tang et al., 2010), we sensitized the Coup-TF2 mutants using a Coup-TF1− allele. Mutants were generated using the following cross: Coup-TF1+/−; Coup-TF2 lox/lox; Ai14lox/lox females×BAC-Nkx2.1-Cre+; Coup-TF2 lox/+ males. Henceforth, we refer to Nkx2.1-Cre+; Coup-TF1+/−; Coup-TF2lox/lox; Ai14lox/+ as the Coup-TF2-sensitized mutants, unless specified otherwise. Controls were Nkx2.1-Cre+; Coup-TF1+/−; Coup-TF2lox/+; Ai14lox/+. The sensitized mutants were viable postnatally. Qualitatively, non-sensitized and sensitized Coup-TF2 mutants had the same phenotypes; reducing Coup-TF1 dose increased phenotype severity.
We first studied the effect of the Coup-TF2 mutation on MGE patterning by examining the expression at E12.5 of Nkx6.2, Otx2 and Shh. Nkx6.2 expression, a marker of the rdMGE [which is where some SST+ CINs originate (Fogarty et al., 2007; Sousa et al., 2009)], was unchanged (Fig. S1N,N′). Otx2 and Shh expression, markers of the rostral MGE and ventral MGE, respectively, also did not change (Fig. S1M,M′,O,O′). Therefore, no defect in MGE patterning was found in the Coup-TF2-sensitized mutant.
We then followed the fate of MGE-POA cells using Nkx2.1-Cre lineage analysis and the Ai14 (tdTomato) Cre recombination reporter. At E12.5, we detected a reduction in tdTomato+ cells coming out of the cMGE but not in the mMGE (Fig. 1B-C′). By E15.5, there were more tdTomato+ cells in the POA, and fewer cells migrating through the lateral ganglionic eminence (LGE) and CGE from the mMGE and cMGE, and into the piriform cortex and globus pallidus (GP) (Fig. 1D-E′). The caudal neocortex had 30% fewer tdTomato+ cells (Fig. 1F,H), whereas the rostral neocortex appeared normal. By P20, neocortical fate map analysis showed no obvious phenotype (Fig. 1G,I,J; Fig. S12A), including in layer V, where most MGE-derived COUP-TF2+ cells are located (Cai et al., 2013). Thus, Coup-TF2-sensitized mutants had a transient decrement of Nkx2.1-Cre lineage cells migrating to the neocortex.
CINs in the Coup-TF2-sensitized mutants had a decreased SST+/PV+ ratio at P20 (Fig. 2A,B). This was due to a 22.0% reduction in SST+ CIN density. This decrease mainly occurred in the lower part of layer V (Fig. 2C). In contrast, the PV+ CIN density was increased, with changes occurring in layers II/III, IV and VI (Fig. 2A-C; Fig. S12A).
Next, we examined whether SST+ CIN numbers and laminar distribution were affected at P0 in Coup-TF1/Coup-TF2 double mutants (2% of Coup-TF1−/− survive after P0; Qiu et al., 1997) (Nkx2.1-Cre+; Coup-TF1−/−; Coup-TF2lox/lox). SST+ CIN density was reduced by ∼30% (Fig. S2A,B). The ratio of SST+ CINs in deep layers (V, VI) versus cortical plate (CP) decreased by 85% (Fig. S2A,C; Fig. S12B). Thus, the SST+ CIN number and laminar positioning defects occurred as early as P0.
We examined SST+ cell numbers in Coup-TF2-sensitized mutants at prenatal ages to determine the onset of SST reduction. There was a reduction in SST+ cells migrating into the neocortex at E12.5, an age when Sst expression was first observed, and at E17.5 (Fig. 2D-F, Fig. S3A-C). At E12.5, the reduction was clearest in caudal regions (Fig. 2E,E′,G), consistent with Coup-TF2 expression in the cMGE (Fig. 1A).
We examined prenatal expression of known regulators of MGE-derived CIN development. At E12.5, Nkx2.1 expression was unchanged in the basal ganglia (Fig. S3D,E). Lhx6 and Sox6 expression, TFs that promote SST+ fate (Azim et al., 2009; Batista-Brito et al., 2009; Zhao et al., 2008; Vogt et al., 2014), was decreased in tangentially migrating interneurons (Fig. S3F,G), and in the SVZ and MZ of the caudal MGE (Fig. 7A,B), respectively. At E17.5, Npas1 expression was reduced, a TF that marks SST+ but not PV+ interneurons (Fig. S3H,I) (Stanco et al., 2014). Therefore, at E12.5 Coup-TF2-sensitized mutants generated fewer immature interneurons expressing Sst, Lhx6 and Sox6.
To determine whether the SST phenotype is also present when Coup-TF2 function is removed from MGE SVZ progenitors, we used the Dlx1-CreERT allele. Embryos exposed to tamoxifen at E10.5 (harvested at E13.5) and at E12.5 (harvested at E15.5) showed decreased Sst expression in the neocortex (Fig. 3A,A′) and decreased numbers of cells migrating through the LGE (Fig. 3B,B′), respectively.
To test Coup-TF2 function in postmitotic MGE-derived neurons, we used the Somatostatin-IRES-Cre allele, the expression of which begins at ∼E12.5 (Taniguchi et al., 2011) (Fig. S14B). Analysis comparing Somatostatin-IRES-Cre+; Coup-TF1+/−; Coup-TF2lox/lox; Ai14lox/+ (mutant) with Somatostatin-IRES-Cre+; Coup-TF1+/−; Coup-TF2lox/+; Ai14lox/+ (control) at P20 in the somatosensory cortex found no change in the number of tdTomato+, SST+ or PV+ cells, and the lamination of SST+ CINs (Fig. 3C-J). Thus, Coup-TF2 regulates the SST+/PV+ ratio and laminar identity in MGE/POA progenitors, but not in neurons.
Coup-TF2 regulates basal ganglia histogenesis and cholinergic neuron development
As shown in the Ai14 fate-mapping experiments, the Coup-TF2-sensitized mutation disrupted basal ganglia histogenesis (Fig. 1D,E). Gene expression analyses at E12.5 of TFs that regulate these structures showed no changes in Nkx2.1 and Lhx6 expression (Fig. S3D-G). However, by E15.5, Nkx2.1, Lhx6 and Lhx8 expression provided evidence for a hypoplastic GP and excessive ventral MGE/POA cells, consistent with the E15.5 Ai14 fate mapping (Fig. S4B-E). Because of the ventral MGE/POA defect, we examined differentiation of choline acetyltransferase (ChAT+) neurons that are derived from this region (Hoch et al., 2015a,b), by assessing Gbx2 and Lhx8 expression, which are TF regulators of these neurons (Chen et al., 2010; Zhao et al., 2003). Striatal interneuron expression of Gbx2 and Lhx8 was reduced at E15.5 and E17.5 (Fig. S4D-K), as was P0 expression of the ChAT-neuron marker TrkA (Ntrk1) (Fig. S4L,M) and P20 expression of ChAT (Fig. S4N,O). Thus, Coup-TF2 was required for the expression of Sox6, Lhx6 and Npas1 (regulators of MGE/POA-derived cortical GABAergic interneurons), and Gbx2, Lhx8, TrkA and ChAT (regulators of striatal cholinergic interneurons).
Coup-TF2 loss in the Nkx2.1-Cre lineage results in an increase in apoptosis
Apoptosis could be responsible for reduced numbers of SST+ CINs and ChAT+ striatal interneurons in Coup-TF2-sensitized mutants. As these decreases were first detected at E12.5, we assayed for cleaved-caspase 3 (cC3) at E11.5, E12.5 and E13.5. At E11.5, we detected cC3+ cells in the VZ, SVZ1 and SVZ2 of the rdMGE and cMGE, where COUP-TF2 is expressed (Fig. 4A; Fig. S5A). At E12.5, cC3+ cells persisted in the rdMGE VZ, SVZ1 and SVZ2, but not in the cMGE (Fig. 4A; Fig. S5A). At E13.5, cC3+ cells were not detected in either MGE region. Moreover, few cC3+ cells were observed in the COUP-TF2-negative rostral ventral MGE (rvMGE)(Fig. 4A; Fig. S5A). Thus, cells from Coup-TF2-sensitized mutant rdMGE and cMGE (Coup-TF2+ zones) are susceptible to apoptosis within a restricted time window.
Coup-TF2 loss in the Nkx2.1-Cre lineage results in premature neurogenesis
We next determined whether impaired neurogenesis also contributed to decreased SST+ interneuron numbers in Coup-TF2-sensitized mutants. To measure neurogenesis, we counted the fraction (Q fraction) of newly postmitotic EdU+ cells (five hours after EdU administration) that were β-III-tubulin (TuJ1+, pan-neuronal marker). At E11.5, the Q fraction increased in Coup-TF2- sensitized mutants in the VZ and SVZ1 of the rdMGE and cMGE (Fig. 4B; Fig. S5B). At E12.5, the Q fraction remained higher in the sensitized mutant than control in rdMGE SVZ1 but decreased in SVZ1 and SVZ2 of the cMGE. The Q fraction was not altered in the Coup-TF2− rvMGE at E11.5 and E12.5 (Fig. 4B; Fig. S5B).
The increased Q fraction seen at E11.5 in the rdMGE and cMGE could be artificially due to progenitors undergoing apoptosis. We therefore assessed if the cC3+ cells in Coup-TF2-sensitized mutants were progenitors or differentiated cells. We found that ∼70% of cC3+ cells were TuJ1−; therefore, most of these dying cells were progenitors (Fig. 4C). While this can inflate the Q fraction, the significant increase in TuJ1+ neurons in the VZ and SVZ1 of the rdMGE and cMGE (13.5±1.44 in control versus 25.0±0.82 in Coup-TF2-sensitized mutant) provides evidence for precocious neurogenesis.
Coup-TF2 loss in the Nkx2.1-Cre lineage results in increased proliferation
Precocious neurogenesis and apoptosis could be due to abnormal cell cycle dynamics. To explore this possibility, we counted the number of S- and M-phase cells (EdU+ following a 5 h pulse and pH3+, respectively). There were no changes in EdU+ and pH3+ cell numbers at E11.5, the age with precocious differentiation and apoptosis. Changes in proliferation were first detected at E12.5, with increases in EdU+ cells in the rdMGE and cMGE SVZ1 (Fig. 4D; Fig. S5C), and in pH3+ cells in the rdMGE and cMGE SVZ1 and SVZ2 (Fig. 4E; Fig. S5D). These increases persisted at E13.5 in the rdMGE and cMGE SVZ1. In the VZ, pH3+ cell numbers only changed at E12.5, with decreases in the rdMGE and cMGE, whereas EdU+ cell numbers did not change at any age.
We next assessed expression of Ccnd2, a promoter of G1 progression (Glickstein et al., 2007). Coup-TF2-sensitized mutants had increased Ccnd2 expression in the SVZ of the rdMGE and cMGE (Fig. 4F,G). These results indicate an increase number of rdMGE and cMGE progenitors remaining in cell cycle in Coup-TF2-sensitized mutants. Although we did not find evidence for this based on EdU-labeling P-fraction analyses (Fig. S11), the increase in E12.5 pH3+ and EdU+ in SVZ1 (described above) is consistent with increased SVZ1 proliferation that supports more neurogenesis of later-born MGE-derived neurons. Indeed, in the mutants we observed increased numbers of CINs that are born at E14.5 (Fig. 5D), and a recovery in the number of Lhx6+ CINs at E17.5 (Fig. S6). Furthermore, the mutants had a normal number of neocortical Nkx2.1-Cre lineage cells by P20 (Fig. 1B-J), providing additional evidence for a recovery of CIN output.
Coup-TF2 loss in the Nkx2.1-Cre lineage has distinct temporal effects on PV+ and SST+ neurons
Because Coup-TF2 loss in MGE/POA progenitors has temporally specific alterations in cell cycle dynamics (Fig. 4), we next studied the effect of the mutation on neuronal output from different ages. We used EdU incorporation to follow the fate and location of the tdTomato+ MGE/POA-derived cells that had become postmitotic (EdU+) at different times. We introduced EdU into pregnant mice at E11.5, E12.5 or E14.5, and determined the density of cells double-labeled with tdTomato+ (driven by Nkx2.1Cre) and EdU+ in different telencephalic regions at P20. We also used immunohistochemistry to examine PV+ and SST+ neuron expression in the EdU+ cells, to assess whether specific subtypes of MGE/POA-derived neurons, with distinct birthdates, were differentially affected in the Coup-TF2-sensitized mutants.
Mutants had an increased density of MGE/POA (tdTomato+)-derived cells born at E11.5 (EdU+) in the somatosensory cortex, piriform cortex, globus pallidus and magnocellular preoptic (MCPO) (Fig. 5D-H, Fig. S7). This was associated with an increased number of PV+ neurons (except in the MCPO). An E11.5 birthdate corresponds to the time when the mutant had precocious neurogenesis (Fig. 4B). The numbers of SST+ neurons born at E11.5 did not change.
The density of MGE/POA-derived mutant cells born at E12.5 increased in the globus pallidus (Fig. 5H; Fig. S7), but decreased in somatosensory cortex and striatum (Fig. 5D,F; Fig. S7). Decreased SST+ somatosensory cortex interneuron numbers was associated with this reduction (Fig. 5D; Fig. S7). Although SST+ and PV+ neuron numbers did not change in the striatum, the reduction in striatal MGE/POA-derived cells may be accounted for by the decline in cholinergic striatal interneurons (Fig. S4).
Finally, there were increased densities of cells born at E14.5 in the somatosensory cortex, piriform cortex and striatum (Fig. 5D-F; Fig. S7). These increases were associated with the increased numbers of PV+ cells born at E14.5. SST+ neuron numbers did not change (Fig. 5D-F; Fig. S7).
Thus, EdU fate-mapping analyses found: (1) that increases in cells born at E11.5 contributed to pallidal projection neurons [globus pallidus (PV+) and MCPO], and PV+ interneurons in the piriform cortex and somatosensory cortex; (2) that decreases in cell numbers were found only if they became postmitotic at ∼E12.5; these cells were SST+ interneurons in the somatosensory cortex and piriform cortex; and (3) increases in the number of cells born at ∼E14.5 in the somatosensory cortex, piriform cortex and striatum; these cells were PV+. Thus, Coup-TF2 differentially regulates subtype identities of MGE/POA-derived neurons based on their birthdates.
Rostral and caudal MGEs generate inverse SST+/PV+ ratios of CINs
While premature neurogenesis and apoptosis may contribute to the decreased SST+/PV+ ratio in Coup-TF2-sensitized mutants, it does not rule out the possibility that Coup-TF2 functions cell-autonomously as a regulator of SST versus PV fate. We used the MGE-to-cortex transplantation assay to study Coup-TF2 function in CIN fate, as we had done to study Lhx6 function (Vogt et al., 2014). Because Coup-TF2 is preferentially expressed in the cMGE (Fig. 1), we compared the SST+/PV+ ratio from E12.5 transplanted cMGE and rostral MGE.
Transplanted interneurons were tdTomato+ (driven by Nkx2.1-Cre). At 27 days post-transplantation (DPT), we assayed tdTomato co-expression with SST and PV. The SST+/PV+ ratio was higher (5.5±0.6) from the cMGE than the rostral MGE (1.4±0.2) (blue, Fig. 6D). Correspondingly, cMGE transplants had more SST and less PV interneurons than rostral MGE transplants (blue, Fig. 6B,C; Fig. S8A,B). Thus, E12.5 cMGE preferentially generated SST interneurons, consistent with our genetic evidence, supporting the idea that Coup-TF2 promotes the generation of SST+ and represses PV+ interneuron fate (Fig. 2).
Coup-TF2 loss in cMGE, but not in rostral MGE, decreased the SST+/PV+ ratio
Next, we used the MGE transplantation assay to compare the effect of the Coup-TF2-sensitized mutation on the SST+/PV+ ratio from the rostral and caudal MGE. Although loss of Coup-TF2 did not affect the SST+/PV+ ratio from the rostral MGE, the Coup-TF2-sensitized mutation reduced the SST+/PV+ ratio by 85% from the cMGE (Fig. 6D; Fig. S8A-D). The transplantation assay provided evidence that Coup-TF2 cell-autonomously regulated the SST+/PV+ ratio. Moreover, because the SST+/PV+ ratio in the transplanted rostral and caudal Coup-TF2-sensitized mutant MGEs were similar, this suggests that Coup-TF2 is a key mechanism for ‘caudalizing’ the MGE.
Coup-TF2 overexpression in MGE cells increases the SST+/PV+ ratio
To further test the hypothesis that Coup-TF2 autonomously regulates the SST+/PV+ fate choice, we used lentiviral transduction (DlxI12b-Cre-T2a-Coup-TF2) to ectopically express Coup-TF2 in the MGE. The DlxI12b enhancer drove expression of Cre and Coup-TF2 (Vogt et al., 2014; Potter et al., 2009). In the control virus, DlxI12b drove expression of only Cre (DlxI12b-Cre). E12.5 Ai14+ MGE cells were dissociated and grown in vitro for ∼12 h, lentivirus-transduced, grown for 3 days, and then analyzed for in vitro expression of tdTomato (induced by Cre), COUP-TF2 and SST. We counted the fraction of tdTomato+ cells (virally transduced) that were COUP-TF2+ and/or SST+ (Fig. S9A,B). Transduction with the control virus produced 27±3% COUP-TF2+; tdTomato+ cells (Fig. S9C); this reflects the fact that dissected MGE had both rostral and caudal parts; only the cMGE has large numbers of COUP-TF2+ cells (Fig. 1A). Transduction with the Coup-TF2 virus resulted in 2.3-fold more (63±3%) of COUP-TF2+; tdTomato+ cells (Fig. S9C). There was a parallel increase in SST+; tdTomato+ cells and COUP-TF2+; SST+; tdTomato+ cells (Fig. S9C). Thus, Coup-TF2 transduction was sufficient to induce SST+ fate within 3 days in vitro.
To assess whether this fate change persisted in vivo, and to examine the effect of Coup-TF2 expression on the SST+/PV+ ratio, we transplanted dissociated E12.5 MGE cells that had been exposed to either the Coup-TF2 or control lentiviruses into wild-type P1 neocortex. At P28, we found that Coup-TF2 transduction increased the SST+/PV+ ratio twofold (Fig. 6H; Fig. S8E-J). In addition, there were increased tdTomato+ cells in layer V (Fig. 6K). Thus Coup-TF2 was sufficient to increase SST and decrease PV expression cell autonomously.
Together, the results from the Coup-TF2 in vivo mutant analysis, the MGE transplantation assay and the Coup-TF2 transduction experiment show that Coup-TF2 cell-autonomously promotes the generation of SST+ interneurons and represses the PV+ fate. Furthermore, both assays provide evidence that Coup-TF2 promotes layer V identity in SST+ CINs.
Coup-TF2 directly binds and activates expression from the Sox6 locus
As Coup-TF2 loss in vivo led to reduced Npas1 and Sox6 expression (Fig. 7A,B; Fig. S3H,I), we tested whether Coup-TF2-transduced and -transplanted MGE cells had increased numbers of Npas1+ and Sox6+ cells. Indeed, Coup-TF2 transduction increased the numbers of cells expressing Npas1 (2.1-fold) and Sox6 (1.4-fold) (Fig. 6I,J).
We recently identified an enhancer-like element near Sox6, named hs692 (chr7:122,274,934-122,276,253), that has activity in the mantle zone of the MGE and POA (Fig. 7D) (Nord et al., 2013). To test whether COUP-TF2 directly binds to hs692, we performed a ChIP-qPCR from E12.5 MGE, CGE and POA, and found an ∼8-fold enrichment compared with control (mouse IgG) (Fig. 7C). Next, we generated a stable transgenic line in which hs692 drives expression of GFP (Silberberg et al., 2016), and crossed it to the Coup-TF2-sensitized mutant. In the mutant, there was reduced GFP expression in the MGE and POA MZ (Fig. 7D′). Together, these data provide evidence that COUP-TF2 directly binds to an enhancer near Sox6, where it promotes expression in the MGE and POA.
Sox6 transduction rescues the SST+/PV+ ratio phenotype in Coup-TF2 mutants
As Coup-TF2 is necessary and sufficient to promote Sox6 expression, we examined whether Sox6 could rescue the SST+/PV+ ratio phenotype in Coup-TF2-sensitized mutants. We transduced E12.5 caudal control or mutant dissociated MGE cells with a DlxI12b-GFP-T2a-Sox6 lentivirus (Fig. 7E). Transplanted interneurons transduced with Sox6 were tdTomato+ (driven by Nkx2.1-Cre) and GFP+. Interneurons that were not transduced with Sox6 (tdTomato+, but GFP−) were used as internal controls.
Sox6 transduction in control cMGE cells increased the SST+/PV+ ratio 2.7-fold, by increasing the number of SST+ interneurons and decreasing the number of PV+ cells (Fig. 7F-H; Fig. S10). In Coup-TF2-sensitized mutant cMGE cells, Sox6 transduction increased the SST+/PV+ ratio 5.1-fold, also by increasing the number of SST+ interneurons and reducing the number of PV+ cells. This provides evidence that Coup-TF2 regulates the SST+/PV+ ratio through Sox6.
Regional and cell-type specification during embryogenesis are often coupled to expression of TFs that are restricted to specific progenitor domains. Here, we found that Nkx2-1 restricts the expression of Coup-TF1/2 in the MGE to an arc that extended from a small rostrodorsal domain to a larger caudoventral domain (Fig. 1; Fig. S1B,D,H,K,L). The Coup-TF1/2+ MGE domain was complementary to an Otx2+ domain (Fig. S1B-E), where Otx2 represses Coup-TF1 MGE expression (Hoch et al., 2015a,b).
Our data support a model in which the Coup-TF1/2+ MGE domain is biased towards generating SST+ CINs (Fig. 2; Fig. S3A-C), as well as cholinergic and pallidal neurons (Fig. S4). These conclusions differ from other publications regarding the major location for the origin of SST+ interneurons, and provide evidence for a Coup-TF1/2-dependent activation of Sox6 expression that promotes SST+ interneuron development.
Multiple lines of evidence show that Coup-TF1/2 have key roles in promoting SST+ interneuron development. Previously, layer V SST+ CINs were shown to express COUP-TF2 (Cai et al., 2013). Here we demonstrate that Coup-TF2-sensitized mutants (in vivo genetics and MGE-transplant experiments) have a reduced neocortical SST+/PV+ ratio (Fig. 2A-C). Coup-TF1/2-mediated regulation of SST fate operates in MGE progenitors, as we only see the reduced SST+/PV+ ratio using Nkx2.1-Cre and not SST-Cre (Fig. 3), the latter is only active postmitotically (Taniguchi et al., 2011). Coup-TF2 sensitized mutants have reduced SST expression in the MGE at E12.5, supporting the early function of these TFs in this cell type (Fig. 2D-G).
We propose that Coup-TF1/2 promote SST+ interneuron development based on three lines of evidence. First, they promote the survival of SST+ interneuron progenitors, as apoptosis occurs in E11.5 and E12.5 MGE progenitors in Coup-TF2-sensitized mutants (Fig. 4A,C; Fig. S5A). Second, Coup-TF1/2 regulates MGE differentiation and cell cycle dynamics; the mutant MGE has precocious neurogenesis and increased proliferation (Fig. 4B,D-F; Fig. S5B-D). Third, Coup-TF1/2 specifies SST+ and represses PV+ interneuron fates, as lentiviral transduction of Coup-TF2 into MGE cells increased the SST+/PV+ ratio (Fig. 6F-H; Figs S8 and 9). It does this by repressing Ccnd2, and by promoting Sox6, Lhx6 and Npas1 expression (Fig. 4F-G′; Fig. S3).
SST+ interneurons are present in cortical layers VI-II. We show that Coup-TF2-sensitized mutants had reduced SST+ CINs specifically in layer V (Fig. 2C). Furthermore, in the mutants, there was a reduction in the number of SST+ CIN labeled with EdU at E12.5, a peak age in the generation of layer V CINs (Fig. 5; Fig. S7). On the other hand, the numbers of SST+ CINs labeled with EdU at E11.5 or E14.5 did not change. Lentiviral transduction of Coup-TF2 in MGE cells, followed by their transplantation, biased the interneurons to acquire layer V identity (Fig. 6K). Together, these data support our hypothesis that Coup-TF2 contributes to the specification of layer V SST+ CINs. Moreover, the presence of a CIN lamination phenotype in mutants that used Nkx2.1-Cre but not with those using Somatostatin-IRES-Cre (Fig. 3C-E,J) shows that Coup-TF1/2-mediated regulation of layer identity occurs in MGE progenitors and not in immature neurons.
We show that the cMGE is a major contributor of cortical SST+ interneurons. This area had more expansive Coup-TF1/2 expression than the rostral MGE, and it produced more SST+ than PV+ interneurons in MGE transplantation experiments (Fig. 6B-D). By contrast, the rostral MGE produced more PV+ than SST+ interneurons than the cMGE (Fig. 6B-D); this property was reversed in the Coup-TF2-sensitized mutants (Fig. 6B-D). Furthermore, the Coup-TF2-sensitized mutants had stronger phenotypes in the caudal basal ganglia, including reduced numbers of Nkx2.1-Cre lineage cells (Ai14+) leaving the cMGE (Fig. 1D,E,H), and reduced expression of Sst, Lhx6, and Sox6 in the cMGE and CGE (Fig. 2D-G; Fig. S3). By 17.5, mutants showed reduced expression of Npas1 (Fig. S3H,I), a TF that marks SST+ and not PV+ interneurons, and that positively regulates SST+ interneuron numbers (Stanco et al., 2014).
The dMGE has been proposed to be a major source of SST+ interneurons (Flames et al., 2007; Wonders et al., 2008; Inan et al., 2012). Furthermore, Nkx6.2-Cre and Nkx6.2-CreER fate mapping and Nkx6.2 functional analyses support the observation that ∼45% of SST+ interneurons are generated by the Nkx6.2+ dMGE (Fogarty et al., 2007; Sousa et al., 2009). There is evidence that the dMGE promotes SST+ generation due to high SHH signaling levels (Xu et al., 2010). Our results do not contradict these ideas, as Coup-TF1/2 are expressed in the rdMGE. However, our transplantation analyses provide evidence for another major source of SST+ interneurons: the caudal Coup-TF1/2+ domain (Fig. 6B-D). Thus, based on Nkx6.2-Cre fate mapping (Fogarty et al., 2007) and our transplantation data, we propose that: (1) the cMGE generates at least 30% of SST+ interneurons; (2) PV+ interneurons are primarily generated from Coup-TF1/2-negative MGE progenitors, which is primarily the Otx2+ rostral MGE domain; and (3) Coup-TF1/2 promote layer V SST+ CINs.
We hypothesize that Coup-TF1/2 control the temporal generation of SST versus PV neurons by regulating cell cycle dynamics. At E11.5, Coup-TF1/2 repress both cell cycle exit and PV+ interneuron fate, based on the findings that: (1) E11.5 Coup-TF1/2-sensitized mutants exhibited precocious neurogenesis; and (2) P20 mutants had increased numbers of PV+ neurons in the neocortex, piriform cortex and globus pallidus that were EdU+ (labeled at E11.5) (Fig. 5; Fig. S7). At E12.5, Coup-TF1/2 promoted SST fate because: (1) E11.5 mutants had progenitor apoptosis and premature neurogenesis, which together could deplete the progenitor pool and neuronal generation at E12.5, including SST+ CINs (Fig. 5); and (2) more E12.5 mutant progenitors remained in the cell cycle (Fig. 4D,E), which subsequently (at E14.5) led to increased numbers of PV+ neurons in the neocortex, piriform cortex and striatum (Fig. 5).
Reduced Sox6 expression in Coup-TF2-sensitized mutants (Fig. 7A-B′) suggests that Sox6 is an essential component of the transcriptional program of COUP-TF1/2. ChIP-qPCR provided evidence that COUP-TF2 directly binds near the Sox6 locus (Fig. 7C). We generated a mouse line with a Sox6 candidate enhancer (hs692) transgene (https://enhancer.lbl.gov). The transcriptional activity of enhancer hs692 in the MGE MZ was reduced in Coup-TF2-sensitized mutants (Fig. 7D), supporting the evidence that Coup-TF1/2 directly regulates this enhancer. Furthermore, Sox6 transduction rescued the reduced SST+/PV+ ratio in Coup-TF2-sensitized mutants (Fig. 7F-H). Previous Sox6−/− mutant analysis has shown that it regulates the SST and PV interneuron maturation and does not regulate proliferation and apoptosis in MGE progenitors (Azim et al., 2009; Batista-Brito et al., 2009).
These studies add basic information for elucidating the transcriptional network orchestrating MGE development. Nkx2.1 and Otx2 regulate regional fate in the VZ (Corbin et al., 2003; Hoch et al., 2015a,b). Nkx2.1 represses LGE/CGE and POA fate (Sussel et al., 1999; Flandin et al., 2010), and initiates Lhx6 and Lhx8 expression (Sussel et al., 1999; Marin et al., 2000; Du et al., 2008; Flandin et al., 2010). Lhx6 and Lhx8 drive differentiation of SST+ and PV+ interneurons, GP and subpallial cholinergic neurons (Rubenstein and Campbell, 2013). Otx2 specifies ventral MGE identity by repressing POA fate, and promotes early neural and oligodendroctye differentiation (Hoch et al., 2015a,b). Coup-TF1/2 also function in MGE progenitors, where they specify SST+ and repress PV+ fate in CINs (Fig. 2); they perform this by promoting Sox6 expression. Coup-TF1/2 are upstream of Gbx2 and Lhx8, which drive ChAT+ neuron development (Fig. S4D-O) (Rubenstein and Campbell, 2013).
Coup-TF1/2 are the first TFs shown to specify SST+ versus PV+ identity. In addition, other mechanisms are known to control SST+/PV+ ratios (Table S1). For example, there is evidence that these interneuron subtypes can be generated from the same neuroepithelial stem cell, but that some stem cells may produce one subtype (Brown et al., 2011; Ciceri et al., 2013). There is also evidence that these subtypes are generated based on their neurogenic location, where apical progenitors give rise to SST+ and basal progenitors give rise to PV+ interneurons (Petros et al., 2015). Thus, it will be important to learn whether Coup-TF1/2 expression is enriched in SST+ interneuron-generating stem cells and apical progenitors. Ccnd2 mutants have an increased SST+/PV+ ratio due to a reduction in PV+ interneuron numbers; the expression of Ccnd2 is repressed by Coup-TF1/2 (Fig. 4F,G).
Alterations in CIN subtype numbers have been reported in post-mortem brains from individuals with holoprosencephaly (Fertuzinhos et al., 2009) and neuropsychiatric disorders, and in mouse models of autism spectrum disorder and schizophrenia (Marin, 2012). We suggest that the SST+/PV+ ratio should be measured in these individuals, as mutations in genes associated with neuropsychiatric disorders (NPAS3 and PTEN) alter the SST+/PV+ ratio in mice with mutations of these genes (Vogt et al., 2015; Stanco et al., 2014). Moreover, circuit-level changes, such as gamma oscillations, which is a feature of schizophrenia, have been observed in PTEN mouse mutants (Stanco et al., 2014).
MATERIALS AND METHODS
The following mouse strains have been previously reported: tdTomatolox/+ (Ai14) Cre-reporter (Madisen et al., 2010), Coup-TF1+/− (Qiu et al., 1997), Coup-TF2lox/+ (Takamoto et al., 2005), Nkx2.1lox/+ Dlx1-CreERT (Taniguchi et al., 2011), Olig2-Cre, Somatostatin-IRES-Cre (Taniguchi et al., 2011), Nkx2.1-Cre (Xu et al., 2008) and hs692-GFP (Silberberg et al., 2016). All strains were on a C57BL/6 background. P1 wild-type hosts (Figs 6 and 7C-F) were CD-1 (Charles River Laboratories). All animals were housed in a vivarium with a 12 h light/12 h dark cycle. Postnatal animals used for experiments were kept with their littermates. Males and females were both used in all experiments.
For timed pregnancies, noon on the day of the vaginal plug was counted as embryonic day (E) 0.5. For EdU experiments, pregnant mice were injected with a single dose of 50 mg/kg body weight (from a 10 mg/ml stock) intraperitoneally, and brains were harvested 1 (Fig. S13), 5 hours (Fig. 4; Fig. S5) or 24 h (Fig. S11) later or at P20 (Fig. 5; Fig. S7). For tamoxifen experiments, pregnant mice were administered 2 mg/40 kg body weight (from a 10 mg/ml stock dissolved in corn oil) by oral gavaging with silicon-protected needles. All animal care and procedures were performed according to the University of California at San Francisco Laboratory Animal Research Center guidelines. Mice were euthanized with CO2 inhalation followed by cervical dislocation. Isolated brains that were used for in situ hybridization and immunohistochemistry were fixed overnight in 4% paraformaldehyde (PFA), transferred to 30% sucrose overnight for cryoprotection, and then OCT embedded for cryosectioning. Sections were cut coronally at 20 μm. For postnatal experiments, animals were anesthetized with intraperitoneal avertin (0.015 ml/g of a 2.5% solution) and perfused transcardially with PBS and then with 4% PFA, followed by brain isolation, overnight fixation, 2 days of cryoprotection and microtome sectioning (coronal, 40 μm).
Antibodies and reagents
Antibodies used were rabbit cC3 (1:250, 9661S, Cell Signaling), mouse Coup-TF2 (1:350, PP-H7147-00, Perseus Proteomics), rabbit DsRed (1:500, 632496, Clontech), chicken GFP (1:500, GFP1020, Aves), rabbit ki67 (1:250, 16667, Abcam), mouse parvalbumin (1:300, MAB1572, Swant Swiss Abs), rabbit pH3 (1:500, 06-750, Upstate), rat somatostatin (1:200, MAB354, Chemicon), goat somatostatin (1:100, SC-7819, Santa Cruz), mouse TuJ1 (1:1000, MMS-435P, Covance), goat ChAT (1:100, AB144P, Millipore) and Alexa-conjugated secondary antibodies (1:300, Molecular Probes). Sections were cover slipped with Vectashield containing DAPI (Vector labs). For antibody validation references, see supplementary Materials and Methods.
In situ hybridization
We performed in situ hybridization on a minimum of n=3 biological replicates for each control and mutant. In each case, a rostrocaudal series of at least ten sections was examined. In situ hybridizations were performed using digoxigenin-labeled riboprobes as described previously (Hoch et al., 2015a,b).
cC3, dsRed, GFP, ki67, pH3 and TuJ1 immunohistochemistry on embryonic brains was performed on slide-mounted cryosections. These were rinsed in phosphate-buffered saline (PBS), blocked in 10% normal serum/PBST (PBS with 0.1% Triton X-100), incubated in primary antibody overnight (4°C), washed in PBST, incubated with secondary antibody for 1-2 h (at room temperature) and washed in PBS. For COUP-TF2, immunohistochemistry was performed as described above but with antigen retrieval [20 min boiling in 10 mM sodium citrate and 0.05% Tween 20 (pH 6), cooled at room temperature for 20 min] and immunoperoxidase staining with ImmPRESS reagent kit (Vector Laboratories). For EdU IHC, EdU was labeled with a Click-iT EdU Alexa Fluor Imaging kit (Catalog #C10337, Molecular Probes).
ChAT IHC on postnatal brains was performed as described (Hoch et al., 2015a,b). For all immunohistochemistry on postnatal brains, staining was carried out on free-floating sections as described previously (Stanco et al., 2014). We performed all immunohistochemistry on n≥3 biological replicates for each control and mutant, except for Fig. 7D (n=2 biological replicates).
Whole-mount X-gal staining
We performed whole-mount X-gal staining on E12.5 brains from Nkx2.1-Cre+; Coup-TF2lox/+ embryos to report expression of Coup-TF2 in MGE/POA. A Cre recombination reporter was used: lacZ (encoding βgal) located 3′ of the floxed region in the Coup-TF2lox allele. Its expression is activated upon Cre-mediated recombination. Brains were isolated from E12.5 embryos and rinsed in PBS, fixed in 2% PFA and 0.2% glutaraldehyde for 1 h, rinsed again in PBS and stained overnight at 37°C with 1 mg/ml X-gal, 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCl2, 0.01% sodium deoxycholate, and 0.02% nonidet P-40 in PBS. The reaction was stopped with a PBS wash, 4% PFA postfix and another PBS wash.
MGE transplantation and transduction, and lentiviral production
For experiments involving Coup-TF2-sensitized mutant MGE transplants (Figs 6A-D and 7C-F; Figs S8A-D and S10), MGEs were split into rostral and caudal halves. For experiments involving COUP-TF2-transduced MGE transplants (Fig. 6E-K; Fig. S8E-J), whole MGEs were used. Transplantation, transduction and lentiviral production were performed as described previously (Vogt et al., 2014).
Lentiviral vector generation
Generation of the DlxI12b-GFP-T2a-Sox6 lentiviral vector has been reported previously (Vogt et al., 2014). For the DlxI12b-Cre-T2a-CoupTF2 lentiviral vector, a human full-length COUP-TF2 was PCR amplified with EcoRI and AgeI sites introduced on the 5′ and 3′ ends, respectively, from a pcDNA5/FRT/TO vector (5′GAGAGAATTCAAGCTTGCCACCATGGCC, 3′GAGAACCGGTTTATTGAATTGCCATATA) (the vector was a gift from Ming-Jer Tsai and Sophia Tsai, Baylor College of Medicine, TX, USA). The PCR product was cloned in frame, 3′ to the T2a sequence of a CMV-GFP-T2a-mcs vector. Next, the T2a-CoupTF2 was isolated from this vector by digestion with BsrGI and SacII; it was then ligated into a DlxI12b-Cre-T2a-mcs vector (also digested with BsrGI and SacII to remove the T2a-mcs site) 3′ to the Cre gene. All vectors were verified by sequencing.
Primary cell culture
MGE tissue was dissected from E12.5 Ai14lox/+ embryos and mechanically dissociated using a P1000 pipette tip. Two hundred thousand cells were seeded onto tissue-culture slides precoated with poly-L-lysine (10 mg/ml, Sigma) and laminin (5 mg/ml, Sigma), and grown in vitro with DMEM-H21 media with 10% fetal bovine serum for ∼12 h. MGE primary cultures were then transduced with lentivirus for 30 min at 37°C. Cultures were washed and grown for 3 days in Neurobasal medium containing B27 supplement, 25% glucose and glutamax. Cultures were fixed with 4% PFA for 10 min and processed for immunocytochemistry as follows: washed in PBS, blocked in 5% bovine serum albumin/PBST, incubated in primary antibody overnight (4°C), washed in PBS, incubated in secondary antibody for 1-2 h (at room temperature), washed in PBS and mounted.
Image acquisition and analysis
Fluorescent images were taken using a Coolsnap camera (Photometrics) mounted on a Nikon Eclipse 80i microscope using NIS Elements acquisition software (Nikon). Bright-field images were taken using a DP70 camera (Olympus) mounted on an Olympus SZX7 microscope. Brightness and contrast were adjusted and images merged using ImageJ software.
For assessing cell densities, 10× images were taken at the somatosensory cortex, piriform cortex, striatum, magnocellular preoptic nucleus or globus pallidus at embryonic or postnatal ages from two or three nonadjacent sections and from both hemispheres for each replicate. A box of a defined area was drawn over a region of interest. Cells were counted within that box and were divided by the box area. For lamination counts, we used DAPI to subdivide neocortical layers. Counting of transplanted MGE cells were performed as described previously (Vogt et al., 2014). See supplementary Materials and Methods for details.
For cC3, EdU, Ki67, pH3 and TuJ1 MGE counts, a 0.17 mm×0.25 mm box was drawn for all ages on the rostral dorsal, caudal and rostral ventral MGE regions. Boxes were placed based on COUP-TF2 expression in adjacent sections. The VZ and SVZ boundary was drawn based on DAPI staining. The SVZ1/SVZ2 boundary was drawn across half of the SVZ subregion. All counts were carried out in these three regions within the box. The Q fraction was the number of cells that were double positive for TuJ1 and EdU, divided by the total number of EdU+ cells. The P fraction was the numbers of cells that were double positive for Ki67 and EdU, divided by the total number of Ki67+ cells.
No statistical methods were used to predetermine sample size. Number of replicates was limited by the probability of generating mutants from a given cross. Sample sizes mentioned above were deemed adequate as reported phenotypes were clear and seen in all replicates. Moreover, most phenotypes were quantified and statistically tested to see if these were real.
All statistical analyses were carried out on SPSS15 software except for the χ2 test (Microsoft Excel). All data points except for ChIP-qPCR data (Fig. 7C) were found to lie within a normal distribution using a Shapiro-Wilk test, and were therefore suitable for parametric testing. All data groups were determined to have equal variance as analyzed by Levene's test. For cell density data, significance was determined using Student's t-test as groups lie within a normal distribution and have equal variance. For transplant data and layer counting, significance was determined by χ2 test. For determining significance in the ChIP-qPCR data (Fig. 7C), we used a Mann–Whitney test (nonparametric) as n=2. All data except for ChIP-qPCR (Fig. 7C) were collected and processed blindly during the data analysis. No data were randomized.
Chromatin immunoprecipitation and qPCR
Chromatin immunoprecipitation (ChIP) (according to Millipore-Upstate ChIP protocol) was performed on E12.5 wild-type CD1 MGE, CGE and POA tissues, which were fixed in 1.5% formaldehyde for 20 min at room temperature and then neutralized with glycine. The cross-linked chromatin was sheared using a Bioruptor UCD-200 for 40 cycles (1 round=30 s on/45 s off at high intensity). The sheared chromatin was incubated at 4°C overnight with 5 μg of mouse COUP-TF2 monoclonal antibody (Perseus Proteomics, PP-H7147-00) or 5 μg mouse IgG (Santa Cruz, SC-2025) as a negative control. Protein/antibody complexes were collected using Dynabeads (20 μl protein A+20 μl protein G, Invitrogen), washed and eluted.
ChIP-qPCR analysis was performed on a 7900HT Fast Real-Time PCR System (Applied Biosystems) using SYBR GreenER qPCR SuperMix (Invitrogen), and qPCR data were analyzed as described by Vokes et al. (2007). For each condition, n=2 biological replicates tested. The following qPCR primers were used to detect enhancer hs692 (near Sox6): forward, CACGGGGCCAGATCTAGTTA; reverse, CACCATTGTTACCCGCCTAT.
We are grateful to Ke Tang, Ming-Jer Tsai and Sophia Tsai for the Coup-TF1+/− and Coup-TF2lox/+ mice, and Coup-TF2 expression construct. We are also grateful to members of the Rubenstein Lab for insightful comments and feedback.
Conceptualization: J.S.H., D.V., S.L., J.L.R.R.; Methodology: J.S.H., D.V., S.L., M.S., S.N.S., J.L.R.R.; Validation: J.S.H., D.V., S.L., M.S., S.N.S., J.L.R.R.; Formal analysis: J.S.H., D.V., S.L., M.S., J.L.R.R.; Investigation: J.S.H., D.V., S.L., M.S., J.L.R.R.; Resources: J.S.H., D.V., S.L., M.S., S.N.S., J.L.R.R.; Writing - original draft: J.S.H., J.L.R.R.; Writing - review & editing: J.S.H., J.L.R.R.; Supervision: J.L.R.R.; Project administration: J.L.R.R.; Funding acquisition: J.S.H., J.L.R.R.
This work was supported by research grants from the National Institute of Mental Health (T32 MH089920 to J.S.H., T32 GM00744 to S.S., and R01 MH081880 and R37 MH049428 to J.L.R.R.) and from the Brain and Behavior Research Foundation (2013 NARSAD Young Investigator Grant to J.S.H.). Deposited in PMC for release after 12 months.
J.L.R.R. is a co-founder and stockholder, and is currently on the scientific board of Neurona, a company studying the potential therapeutic use of interneuron transplantation.