Dlx1 and Dlx2 Promote Interneuron GABA Synthesis, Synaptogenesis, and Dendritogenesis

Abstract

The postnatal functions of the Dlx1&2 transcription factors in cortical interneurons (CINs) are unknown. Here, using conditional Dlx1, Dlx2, and Dlx1&2 knockouts (CKOs), we defined their roles in specific CINs. The CKOs had dendritic, synaptic, and survival defects, affecting even PV+ CINs. We provide evidence that DLX2 directly drives Gad1, Gad2, and Vgat expression, and show that mutants had reduced mIPSC amplitude. In addition, the mutants formed fewer GABAergic synapses on excitatory neurons and had reduced mIPSC frequency. Furthermore, Dlx1/2 CKO had hypoplastic dendrites, fewer excitatory synapses, and reduced excitatory input. We provide evidence that some of these phenotypes were due to reduced expression of GRIN2B (a subunit of the NMDA receptor), a high confidence Autism gene. Thus, Dlx1&2 coordinate key components of CIN postnatal development by promoting their excitability, inhibitory output, and survival.

Figure 1.

Loss of Dlx1/2 function leads to a decrease of CINs only at late postnatal stages. (a, b, e, f) Coronal sections through the telencephalon of control (a, e) and Dlx1/2^f; GFP; I12b-Cre mutant (b, f) E15.5 embryos (a, b) or P30 (e, f) mice showing the distribution of GFP-expressing cells after GFP staining. (a) Numbers identify bins for quantification. (c, d, g, h) Quantification of GFP⁺ cells in cortex in E15.5 embryos (c, d) and P30 (g, h) by (c) (control: 2004.2 ± 151.8 cells/mm², n = 4; mutant: 1775.7 ± 128.5 cells/mm², n = 4; P = 0.2) (g) (control: 370.5 ± 17.01 cells/mm², n = 4; mutant: 269.9 ± 9.04 cells/mm², n = 4; P = 0.0014) cell density and (d, h) distribution along the cortex by bins (d) (n = 4, a total of 2141 counted cells, P = 0.4) or layers (h) (control; layer2/3:494.3 ± 45.2; layer4:401.6 ± 58.4; layer5/6:370.02 ± 26.3 cells/mm², n = 4; mutant; layer2/3:329.03 ± 23.3; layer4:232.2 ± 21.3; layer5/6:305.3 ± 13.3 cells/mm², n = 4; P layer2/3 = 0.01, layer4 = 0.03, layer5/6 = 0.07) of control (black bars) and Dlx1/2 mutant (white bars) mice. *P < 0.05, **P < 0.01 (t-test). Scale bar (in a) a–b, 100 μm and (in e) e–f, 200 μm. (i–l, m–p) Coronal sections through the cortex of P30 control (i–l) and Dlx1/2 mutant (m–p) mice showing the distribution of positive cells after DAB immunohistochemistry for PV (i, m), SOM (j, n), CR (k, o) and VIP (l, p). (q) Quantification of the density of PV (control: 248.06 ± 25.4 cells/mm2, n = 4; mutant: 132.2 ± 19.6 cells/mm2, n = 4; P = 0.011); SOM (control: 202,4 ± 3.7 cells/mm², n = 3; mutant: 121.7 ± 11.2 cells/mm², n = 3; P = 0.02); CR (control: 52.6 ± 1.4 cells/mm², n = 4; mutant: 35.4 ± 4.3 cells/mm², n = 4; P = 0.0096) and VIP (cells control: 50.6 ± 2.9 cells/mm², n = 3; mutant: 23.03 ± 1.8 cells/mm², n = 4; P = 0.0003) cells in the cortex. *P < 0.05, **P < 0.01, ***P < 0.001 (t-test). I–VI, cortical layers from I to VI. Histograms show average ± SEM. Scale bar (in m) i–l, m–p 200 μm.

Figure 2.

Dlx1 is expressed prenatally, but not neonatally, in cells that become PV⁺CINs; adult PV⁺ interneurons maintain Dlx2 expression. (a) Schematic diagram of the experimental design. Dlx1Cre-ER; AI14 pregnant females received a single injection of tamoxifen at E12 or P4. Embryos were allowed to be born and analyzed at P30. (b–e) Coronal sections through the somatosensory cortex of Dlx1Cre-ER; AI14 P30 mice showing the distribution after the immunohistochemistry of PV⁺ (green) and tdTomato⁺ (red) cells after E12 (b, c) or P4 (d, e) tamoxifen injection. (c’, c”, e’, e”) Magnification of the areas boxed in c and e, respectively. Scale bar (in d) b–c, d–e 200 μm and (in e”) c’, c”, e’, e” 50 μm. (f–i) Coronal sections through the cortex (f–g) and the hippocampus (h–i) P30 WT mice showing immunohistochemistry expression for PV⁺ (green) and DLX2⁺ (red) cells. (g’, g”, h, i) Confocal images of somatosensory cortex (g’–g”) and hippocampal pyramidal layer (h, i) showing immunohistochemistry expression for PV⁺ (green) and DLX2⁺ (red) cells. (j) Quantification of the DLX2/PV colocalization. Colocalized (arrowhead), non-colocalized (open arrowhead) cells are noted in the panels. Scale bar (in f) f, g 100 μm, (in g”) g’, g” 20 μm and (in h) h, i 20 μm.

Figure 3.

Synaptic transmission is altered in the cortex of adult mice after Dlx1/2 gene deletion. (a) Scheme of Dlx1/2 deletion in cortical interneurons and recording location (pipette) in the cortex. (b) Whole-cell recordings at P30 reveal that sIPSC frequency (control: 23.3 ± 3.4 Hz, n = 12; mutant: 18.5 ± 2.5 Hz, n = 17; P = 0.2) and amplitude (control: 32.9 ± 13.2 pA, n = 11; mutant: 26.2 ± 2.4 pA, n = 17; P = 0.1) are unchanged between WT and mutant mice. (c) Synaptic inhibition is reduced, with significant reductions in mIPSC frequency (control: 15.6 ± 1.9 Hz, n = 12; mutant: 10.2 ± 1.2 Hz, n = 15; P = 0.02) and amplitude (control: 21.3 ± 1.8 pA, n = 11; mutant: 13.5 ± 1.4 pA, n = 15; P = 0.002). Analysis of mIPSC kinetics show significant increases in both rise time (control: 1.59 ± 0.08 ms, n = 11; mutant: 2.00 ± 0.11 ms, n = 15; P = 0.01) and decay tau (control: 12.11 ± 0.84 ms, n = 10; mutant: 15.17 ± 0.88 ms, n = 15; P = 0.02). *P < 0.05 **P < 0.01, (t-test), as well as in the probability distributions (d) of inter-event intervals and amplitudes ***P < 0.001, (Kolmogorov–Smirnov test). (e) Scheme of Dlx1/2 deletion in Calretinin interneurons and recording location (pipette) in the cortex. Measures of excitatory synaptic transmission in cortical layer 2 on CR⁺ interneurons at P30, (f) show a significantly decreased sEPSC frequency (control: 13.1 ± 1.4 Hz, n = 7; mutant: 7.01 ± 0.7 Hz, n = 4; P = 0.01), but normal sEPSC amplitude (control: 12.5 ± 1.0 pA, n = 7; mutant: 13.2 ± 0.8 pA, n = 4; P = 0.6). (g) mEPSC frequency was significantly reduced (control: 10.5 ± 1.3 Hz, n = 5; mutant: 4.4 ± 0.8 Hz, n = 4; P = 0.009) but not the mEPSC amplitude (control: 15.6 ± 0.8 pA, n = 5; mutant: 13.9 ± 1.6 pA, n = 4; P = 0.3). Analysis of mEPSC kinetics show no significant differences in both rise time (control: 1.46 ± 0.12 ms, n = 5; mutant: 1.72 ± 0.13 ms, n = 4; P > 0.05), and decay tau (control: 5.72 ± 1.0 ms, n = 5; mutant: 5.46 ± 0.70 ms, n = 4; P > 0.05). *P < 0.05, **P < 0.01 (t-test), an effect also revealed by analysis of inter-event interval (h); ***P < 0.001, (Kolmogorov–Smirnov test). Histograms show average ± SEM. n, number of cells.

Figure 4.

Decreased expression of Gad1, Gad2, and Vgat in the Dlx1/2;I12bCre CKO. (a–f) Coronal sections through the cortex of control (a–c) and Dlx1/2 mutant (d–f) mice at P0 showing the expression of Gad2 (a, d), Gad1 (b, e) and Vgat (c, f) mRNA. (g), Quantification of the expression of Gad2 (control: 283.5 ± 23.4 cells/mm², n = 3; mutant: 141.2 ± 6.5 cells/mm², n = 3; P = 0.004), Gad1 (control: 253.2 ± 4.7 cells/mm², n = 3; mutant: 189.3 ± 5.6 cells/mm², n = 3; P = 0.0009) and Vgat (control: 233.2 ± 14.3 cells/mm², n = 3; mutant: 156.6 ± 21.1 cells/mm², n = 3; P = 0.04) by cell density in the cortex from controls (black bars) and Dlx1/2 mutants (white bars) mice. (h) Quantitative RT-PCR measurement RNA levels of Gad1 (control: 102.4 ± 15.1%, n = 3; mutant: 54.9 ± 6.01%, n = 3; P = 0.04), Gad2 (control: 112.8 ± 17.8%, n = 3; mutant: 57.5 ± 8%, n = 3; P = 0.04), Vgat (control: 101.7 ± 14.3%, n = 3; mutant: 54.6 ± 2.8%, n = 3; P = 0.07), Dlx2 (control: 100.1 ± 4.4%, n = 3; mutant: 6.04 ± 0.4%, n = 3; P = 2.9e−005) and Dlx1 (control: 84.3 ± 5.9%, n = 3; mutant: 1.6 ± 0.2%, n = 3; P = 0.0001) on RNA from P2 cortical tissue from control (closed symbols) and Dlx1/2 mutant (open symbols) mice. Data are expressed as the percentage of GAPDH control RNA. *P < 0.05, **P < 0.01 and ***P < 0.001 (t-test). Histograms and scatter plots show average ± SEM. Scale bar (in d) a–f, 200 μm.

Figure 5.

Evidence for Direct Control of Gad1, Gad2, and Vgat by DLX2. (a–c) Diagram showing DLX2 binding peaks in E13.5 and E16.5 GE in the (a) Gad2/Myo3a locus (peaks #1 and 2), (b) in the Gad1/Myo3b locus, and (c) the Vgat (Slc32a1)/Arhgap40 locus. H3K27Ac ChIP-Seq peaks show candidate active regulatory elements in the E13.5 GE; these peaks are generally in the same position as the DLX2 peaks. The orange box in (a) shows putative Gad2 enhancer 1. Arrows shows the direction of transcription from the promoter. Thick lines underneath the peaks are the computationally “called” peaks. (d) Schematic of the luciferase reporter transcriptional assay in P19 cells. (e) Luciferase assay data, presented as fold change in activation with DLX2 over activation with GFP, and normalized to activation of the empty pGL4.23 control (pGL4.23: 1.0 ± 0.08, n = 3; Gad2 En1: 7.7 ± 0.54, n = 3, P = 0.0053). Histograms show average ± SEM. **P < 0.01, (t-test). (f–h”) Immunohistochemistry from the Gad2En1-GFP transgenic mouse. (f) GFP and DAPI labeling of an E12.5 telencephalic coronal section. (g) GFP and DAPI labeling of P30 neocortex (coronal section). (h–h”) Colocalization (arrows) of GFP (h, h’) and GABA (h, h”) in P30 neocortex. Scale bar in f, 250 μm, g, 250 μm, h, 50 μm. Cx, Cortex; LGE, Lateral Ganglionic Eminence; MGE, Medial Ganglionic Eminence.

Figure 6.

Loss of Dlx1/2 in CR⁺ CINs reduces synaptic boutons. (a, e) Schematic drawings of Dlx1/2 deletion in CR⁺ (green) CINs. The circles indicate the synapses were analyzed onto the axon (a) or the dendrites (e) of neurons in the somatosensory cortex. (b, c) Single confocal images of Dlx1/2^f; GFP; CR-Cre control (b), and mutant (c), showing the colocalization of Synaptophysin⁺ boutons and Gephryn⁺ clusters onto CR axon (arrowheads) at P30. (b’, c’) Binary images after ImageJ software processing on confocal images (b, c) for quantification. (d) Quantification of Synaptophysin⁺ boutons and Gephryn⁺ clusters colocalization onto the CR axon (control: 0.22 ± 0.02 puncta/μm, n = 17; mutant: 0.15 ± 0.009 puncta/μm, n = 19; average ± SEM, P = 0.006). *P < 0.05 (t-test). (f, g) Single confocal images of Dlx1/2^f; GFP; CR-Cre control (f) and mutant (g) showing the colocalization of VGlut1⁺ boutons and PSD95⁺ clusters onto CR dendrite (arrowheads) at P30. (f’, g’) ImageJ software was used to process confocal images (f, g) for quantification. (h) Quantification of VGlut1⁺ boutons and PSD95⁺ clusters colocalization onto the CR dendrite (control: 0.3 ± 0.05 puncta/μm number, n = 15; mutant: 0.2 ± 0.02 puncta/μm, n = 17; P = 0.04). *P < 0.05 (U of Mann–Whitney). Scale bar (in f) b, c, f, g 5 μm. n, number of cells.

Figure 7.

Loss of Dlx1/2 function (I12b-Cre) shortens dendrite length of CR⁺ and PV⁺ CINs. (a, b) Immunohistochemistry for CR on coronal sections through the cortices of control (a) and Dlx1/2 mutant (b) P30 mice. (c) Quantification of dendrite morphology of control (black bars) and Dlx1/2 mutants (white bars) (control: 140.2 ± 14.2 μm, n = 21; mutant: 49.9 ± 5.3 μm, n = 28; P = 3.6e−008) ***P < 0.001 (U of Mann–Whitney). Scale bar (in a) a, b 100 μm. (d) Schematic diagram of the experimental design. MGE from E13.5 Dlx1/2^f; GFP; I12b-Cre controls and mutants were dissected, dissociated and transplanted into the cortices of P1 WT mice. Host mice were analyzed 28 DPT (days post-transplant). (e, f) Coronal sections trough the cortex of transplanted mice showing the dendritic morphology of a Dlx1/2^f; GFP; I12b-Cre control (e) and mutant (f) PV⁺ cells after staining of GFP (green) and PV (red) and highlighted by imageJ software. (g, h) Quantification of dendrite length (g) (control: 1420.5 ± 124.7 μm, n = 11; mutant: 896.1 ± 103.1 μm, n = 12; P = 0.004) and dendrite number (h) (control: 15.6 ± 1.06 dendrite number, n = 11; mutant: 13.9 ± 1.7 dendrite number, n = 12; P = 0.2) in controls (black bars) and Dlx1/2 mutant (white bars) after 28 DAT. **P < 0.01 (t-test). Histograms show average ± SEM. n, number of cells. Scale bar (in e) e, f 100 μm.

Figure 8.

Grin2b rescues the dendritic phenotype of Dlx1/2 CKO CINs. (a–d) Single confocal images of Dlx1/2^f; GFP; CR-Cre (a, b) and Dlx1/2^f; GFP; I12b-Cre (c, d) control (a, c) and mutant (b, d) mice at P30 showing the colocalization of GRIN2B (magenta) boutons onto GFP⁺ dendrite (arrowheads) at P30. (e) Quantification of GRIN2B⁺ boutons density on control (black and dark gray bars) and mutant (white and light gray bars) Dlx1/2^f; GFP; CR-Cre (control: 0.67 ± 0.04 puncta/μm number, n = 23; mutant: 0.42 ± 0.03 puncta/μm, n = 24; P = 1.8e−005) or Dlx1/2^f; GFP; I12b-Cre (control: 0.64 ± 0.06 puncta/μm number, n = 23; mutant: 0.37 ± 0.03 puncta/μm, n = 29; P = 2.4e−004) P30 mice and P8 Dlx1/2^f; GFP; I12b-Cre (control: 0.76 ± 0.06 puncta/μm number, n = 25; mutant: 0.44 ± 0.03 puncta/μm, n = 29; P = 2.0e−005). ***P < 0.001 (t-test). Histograms show average ± SEM. n, number of cells. Scale bar (in b) a–d 10 μm. (f) Schematic diagram of the experimental assay. MGE from E13.5 Dlx1/2^f control and mutants were dissected and dissociated. Cells, transfected with a mixture of CMV-Cre⁺Flex-GFP or CMV-hGRIN2B-T2a-Cre⁺Flex-GFP, were transplanted into cortices of P1 WT mice. Mice were analyzed 28 DAT (days after transplant). (g–i) Images and drawings of PV cells transfected with CMV-Cre⁺Flex-GFP (g, h) or CMV-hGRIN2B-T2a-Cre⁺Flex-GFP (i) from Dlx1/2^f controls (g) or mutants (h, i). (j) Quantification of dendrite length of controls (black bar), mutants (white bar), and human GRIN2B rescued mutants (gray bar) 28 DAT (control: 2824.6 ± 237.3 μm, n = 10; mutant: 1550.1 ± 223.1 μm, n = 10; rescued: 2533.8 ± 163.2 P = 0.0005). ***P < 0.001, **P < 0.01 (ANOVA test, Tukey’s B post hoc). Histograms show average ± SEM. Scale bar (in g) G–I 100 μm. n, number of cells. (k) Diagram showing GE ChIP peaks in the Grin2b gene locus at E13.5 after H3K27Ac ChIP and at E13.5 and E16.5 after DLX2 ChIP. Three DLX2 peaks are identifiable at both ages (#1-3). (l) Luciferase reporter assay in P19 cells testing DLX2 activation of region #3 (orange box in k). Data are presented as fold change in activation with DLX2 over activation with GFP, and normalized to activation of the empty pGL4.23 control (pGL4.23: 1.0 ± 0.08, n = 3, same as control in Figure 5; Grin2b En3: 2.5 ± 0.20, n = 3, P = 0.01). Histograms show average ± SEM. *P < 0.05 (t-test).