The Thalamus Regulates Retinoic Acid Signaling and Development of Parvalbumin Interneurons in Postnatal Mouse Prefrontal Cortex

Abstract

GABAergic inhibitory neurons in the prefrontal cortex (PFC) play crucial roles in higher cognitive functions. Despite the link between aberrant development of PFC interneurons and a number of psychiatric disorders, mechanisms underlying the development of these neurons are poorly understood. Here we show that the retinoic acid (RA)-degrading enzyme CYP26B1 (cytochrome P450 family 26, subfamily B, member 1) is transiently expressed in the mouse frontal cortex during postnatal development, and that medial ganglionic eminence (MGE)-derived interneurons, particularly in parvalbumin (PV)-expressing neurons, are the main cell type that has active RA signaling during this period. We found that frontal cortex-specific Cyp26b1 knock-out mice had an increased density of PV-expressing, but not somatostatin-expressing, interneurons in medial PFC, indicating a novel role of RA signaling in controlling PV neuron development. The initiation of Cyp26b1 expression in neonatal PFC coincides with the establishment of connections between the thalamus and the PFC. We found that these connections are required for the postnatal expression of Cyp26b1 in medial PFC. In addition to this region-specific role in postnatal PFC that regulates RA signaling and PV neuron development, the thalamocortical connectivity had an earlier role in controlling radial dispersion of MGE-derived interneurons throughout embryonic neocortex. In summary, our results suggest that the thalamus plays multiple, temporally separate roles in interneuron development in the PFC.

Keywords: Cyp26b1; interneurons; parvalbumin; prefrontal cortex; retinoic acid; thalamocortical.

Graphical abstract

Figure 1.

Spatiotemporally regulated expression of Cyp26b1 and Aldh1a3 in the PFC and other forebrain regions. In situ hybridization of frontal sections through PFC at various stages is shown. A–E, Cyp26b1 expression in frontal cortex. At P0, only a weak expression is seen in medial PFC (A, arrowhead). B, At P2, Cyp26b1 starts to be detected clearly in medial PFC (B, arrowhead); expression in lateral cortex, especially agranular insula in more superficial layer (A, arrow) is strong, which continues into later stages (B–E, double arrowheads). At P8, expression in medial (C, arrowhead) and ventral (C, arrow) PFC is strong. At P21, the expression of Cyp26b1 is reduced in medial PFC (D, arrowhead) and is almost undetectable by P35 (E, arrowhead). F, Syt6, a layer 6 marker, is expressed in the same layer as Cyp26b1 in medial PFC at P8 (arrowhead). G–K, Aldh1a3 expression in frontal cortex. At P0, expression is not detected in PFC (G). At P2, clear expression is detected in the superficial layer of medial PFC (H, arrow). The expression continues into P8, P21, and P35 (I–K, arrow). L, Schematic summary of the spatial expression patterns of Cyp26b1 and Aldh1a3 in medial PFC of early postnatal mouse brains. M, Expression of Cyp26b1 is not detected in LGE or MGE at E14.5, but is already found in the hippocampus (arrowhead), septum (arrow), globus pallidus (data not shown), and amygdala (data not shown). N, At E16.5, Cyp26b1 is detected in hippocampus (arrowhead), piriform cortex (Pir), globus pallidus (GP), and amygdala (Amy). O, P, This pattern continues into P4 and adulthood (data not shown). P is at a more caudal level than O. Expression in the hippocampus is strongest in CA3 and hilus, whereas multiple nuclei in amygdala show strong expression of Cyp26b1 (P). Scale bars: A–K, N–P, 1 mm; M, 500 μm.

Figure 2.

RA signaling in early postnatal PFC. In all sections, the right hemisphere is shown, and the white dashed line marks the medial surface of the frontal cortex. A–E, H–J, Immunostaining for β-gal on frontal sections of P0 (A) and P14 (B–E, H–J) brains of RARE-LacZ transgenic mice. A, At P0, β-gal expression is found only in the radial glial fibers (arrow). B–J, At P14, β-gal-positive cells are abundant in medial PFC and they are SOX6 positive. B is a 10× image including the medial surface of the brain, and C–E are 40× confocal images of the same region outlined in the square in B. E is the merged image of C and D. Yellow arrows in C and D show SOX6/β-gal-double-positive cells, and green arrow in C shows a rare, β-gal-positive, SOX6-negative cell. F, Average number of β-gal-positive cells per section by layers (mean ± SEM). L1, Layer 1 as marked by sparse labeling in DAPI staining; L6, layer 6 as marked by TBR1 staining. The two middle columns represent equal-width bins between layer 1 and layer 6, and approximately corresponds to layers 2/3 and layer 4/5, respectively. Because most β-gal-positive cells are immediately above layer 6, and layer 4 is thin in medial PFC (Fig. 6G), the highest peak in the third column likely represents layer 5. G, The ratios of SOX6; β-gal-double-positive cells among β-gal-positive cells are shown by layers (mean ± SEM). H–J, β-gal does not overlap with SP8, CTIP2, or TBR1. Scale bars: A, 200 μm; B, 100 μm; C–E, H–J, 50 μm.

Figure 3.

Overlapping expression of RARE-LacZ transgene and PV in early postnatal PFC. A–C, In situ hybridization for Pvalb mRNA on frontal sections of control mice at P8 (A), P14 (B), and P21 (C). In medial PFC, Pvalb is undetectable at P8, but is robustly expressed at P14, which further increases by P21 (A–C, single arrow). Pvalb mRNA is already expressed in many cells in dorsolateral frontal cortex (A, double arrow). D–G, Double immunostaining for β-gal and PV on frontal sections of P14 brains of RARE-LacZ transgenic mice. D is a 10× image including the medial surface of the brain, and E–G are 40× confocal images of the same region outlined in the square in D. G is a merged image of E and F. Note the heavy overlap between β-gal and PV. H, I, Timecourse of Sst expression in medial PFC of postnatal mice. In H, Sst mRNA was detected by in situ hybridization using a Tyramide Signal Amplification system, followed by immunostaining with anti-SOX6 antibody. The section is from a control P14 PFC and left is to the medial surface. In I, the ratio of Sst-positive, SOX6-positive cells to SOX6-positive cells in medial PFC is shown for P0, P4, P7, P14, and P59. A plateau value of ∼0.4 is reached by P7. At P0, a much smaller portion of SOX6-positive cells expressed Sst mRNA. Each dot indicates an average number of cells in medial PFC of at least three sections of a wild-type brain. J, Double immunostaining for β-gal and SST on frontal sections of P14 brains of RARE-LacZ transgenic mice. Note the little or no overlap between β-gal and SST. K, The ratios of PV; β-gal-double-positive cells among β-gal-positive cells in P14 medial PFC are shown by layers (mean ± SEM). L, The ratios of PV/β-gal-double-positive cells among PV-positive cells in P14 medial PFC are shown by layers (mean ± SEM). M, A schematic summary of the interneuron populations in medial PFC. Based on the results of this study, a subpopulation of PV interneurons responds to RA via RAR/RXR receptor complex. Scale bars: A–C, 1 mm; D, 100 μm; E–G, J, 50 μm; H, 200 μm.

Figure 4.

Conditional deletion of Cyp26b1 using Synaptotagmin6-Cre (Syt6-Cre). A–E, Recombination in Syt6-Cre transgene mice. A–E, Expression of ZSGreen in Syt6-Cre/+; Ai6 (ZSGreen Cre reporter) mice at E12.5 (A), E14.5 (B), and P0 (C–E) are shown. All sections are coronal, and the midline is to the left. At E12.5, the expression of ZSGreen reporter is found in meninges (A, arrow) and preplate (A, arrowhead), but not in the rest of the cortex or MGE and LGE. At E14.5, a small number of cortical cells (B, double arrows) below the marginal zone (B, arrowhead) start to express ZSGreen. C–E, At P0, many layer 6 cells of frontal cortex express ZSGreen (C, arrow), but not in more caudal neocortex (D, Ncx), CA1, CA3, and hilus regions of the hippocampus (E, note that strong signal is found in the meninges of the hippocampus) or the amygdala (F, Amy). ic, Internal capsule. Scale bar, 200 μm. F–Q, Generation of conditional Cyp26b1 knock-out mice. F–Q, In situ hybridization of frontal sections of P8 (F, G, I, J, L–Q) or P14 (H, K) Cyp26b1 conditional knock-out (I–K, O–Q) and control littermate (F–H, L–N) brains. Cyp26b1 was conditionally knocked out using the Syt6-Cre transgene. Syt6 is expressed in layer 6 of both control (F) and Cyp26b1 knock-out (I) brains (arrow). The expression of Cyp26b1 in layer 6 of frontal cortex (arrow) is detected in control brains, but not in Cyp26b1 knock-out brains at P8 (G, J) and P14 (H, K). The expression of Cyp26b1 in agranular insula is unchanged in Cyp26b1 knockouts (G, H, J, K, arrowhead). The expression of Cyp26b1 in CA3 and hilus region of the hippocampus (L, O, arrowhead), globus pallidus (M, P, arrow), and amygdala (N, Q, arrowhead) is unchanged in Cyp26b1 knockouts. Scale bar, 1 mm.

Figure 5.

Increased Pvalb-expressing interneurons in medial PFC of Cyp26b1 knock-out mice. A, B, In situ hybridization of frontal sections of P14 Cyp26b1 conditional knock-out mice (B) and littermate controls (A). Expression of Pvalb mRNA is shown. See Materials and Methods on the binning of the medial PFC. Numbers of Pvalb-positive cells in the two superficial bins and two deep bins were added together and compared separately between Cyp26b1 mutants and littermate controls. C–E, Result of paired ratio t tests for cell counts in the medial PFC (C), motor cortex (D), and somatosensory cortex (E), all on the same frontal sections. Each line connecting red and blue dots represents a pair of brains analyzed in the same experiment (n = 5). The p values of the ratio of paired t tests for each layer (superficial, deep, total) are shown. In repeated-measures two-way ANOVA, the p values for layer (superficial versus deep), pair (between control and knockout), and interactions (between layer and pair) are 0.0017, 0.0325, and 0.1416 (P14 in PFC); 0.0441, 0.0961 and 0.7807 (P14 in motor cortex); 0.1771, 0.4751, and 0.5496 (P14 in somatosensory cortex), respectively. Scale bar, 1 mm. L1, layer 1. F, G, At P21, the density of Pvalb-expressing cells (F), but not Sst-expressing cells (G), is increased in Cyp26b1 conditional knock-out mice. In repeated-measures two-way ANOVA, the p values for layer (superficial versus deep), pair (between control and knockout), and interactions (between layer and pair) are 0.0047, 0.0287, and 0.0637 (Pvalb); 0.0065, 0.3621, and 0.7609 (Sst). H, I, No significant changes in the density of Pvalb- and Sst-expressing interneurons in medial PFC of adult (P56–P67) Cyp26b1 knock-out mice. Each line connecting red and blue dots represents a pair of brains analyzed in the same experiment (n = 4). The p values of paired t tests for individual layers are shown. J–R, No significant changes in the number of Sst-, Vip-, and Lhx6-expressing interneurons in medial PFC of Cyp26b1 knock-out mice at P14. J–O, In situ hybridization of frontal sections of P14 Cyp26b1 conditional knock-out mice (J, L, N) and littermate controls (K, M, O). Expression of Sst (J, K), Vip (L, M), and Lhx6 (N, O) is shown. Binning and cell counts were performed as shown in A and B. Scale bar, 1 mm. P–R, Result of statistical analysis. Each line connecting red and blue dots represents a pair of brains analyzed in the same experiment (n = 5). The p values of paired ratio t test for individual layer are shown.

Figure 6.

Transient expression of Cyp26b1 in the PFC does not occur in the absence of thalamus–cortex interactions in Gbx2 mutant mice. A–F, Thalamus–PFC disconnection in Gbx2 mutant mice. A, B, Immunostaining for NetrinG1 at E16.5. In control mice, NetrinG1-labeled thalamocortical axons are visible in coronal sections of frontal cortex. Arrowhead in A shows the medial PFC, where robust labeling is detected. In contrast, NetrinG1 labeling is barely detectable in the frontal cortex of Gbx2 mutant mice, including the medial PFC (B, arrowhead). Scale bar, 200 µm. C–F, DiI labeling at P14. C, D, DiI placement in medial PFC retrogradely labels medial thalamic nuclei in the control brains. E, F, In Gbx2 mutants, the label is severely reduced, indicating the deficiency of both thalamocortical and corticothalamic projections. G–J, Expression of RORβ and Lmo4 is qualitatively normal in the PFC of Gbx2 mutant mice at P8. G, I, The expression of RORβ in layer 4 is comparable between control (G) and Gbx2 mutant (cko) mice (I, arrows). H, J, Laminar expression patterns of Lmo4 also appear unchanged in Gbx2 mutants. Scale bar, 1 mm. K–Z, Transient expression of Cyp26b1 in the PFC does not occur in the absence of thalamus–cortex interactions in Gbx2 mutant mice. K–T, in situ hybridization of frontal sections through PFC at various postnatal stages with a Cyp26b1 probe. K–O, In control mice (K–O), Cyp26b1 expression starts at P2 in medial (K, arrowhead) and ventral (K, single arrow) PFC, and continues until P14 (M). N, O, At P21, expression in medial PFC is reduced (N) and is no longer detectable at P35 (O). K–O, In addition to medial and ventral PFC, Cyp26b1 is also expressed in lateral frontal cortex, including the motor and somatosensory areas (double arrows) and agranular insula (double arrowheads). P–T, In Gbx2 mutant mice, the expression of Cyp26b1 is not induced in medial or ventral PFC at P2 as well as at later stages, although ventral PFC does not appear to be affected at P14 and later. P–T, Expression in more superficial layer of lateral cortex (double arrows and double arrowheads) is not affected in Gbx2 mutant mice. U, X, Expression of the layer 6 marker Syt6 is not affected in Gbx2 mutant mice. V, Y, Expression of Aldh1a3 in layer 2 of medial PFC and anterior cingulate cortex (arrow) is not affected in Gbx2 mutant mice. Scale bar, 1 mm. W, Z, Summary schematic for this figure.

Figure 7.

Abnormal radial distribution of MGE-derived interneurons in the medial PFC of neonatal and P21 Gbx2 mutant mice. A, B, Representative images of immunostaining for LHX6 in medial PFC of wild-type (A) and Gbx2 mutant (B) mice at P0. Binning is shown in yellow. Layer 1 (L1) was defined as the cell-sparse layer detected by DAPI staining. Layer 6 (L6) was defined as the layer with TBR1 staining on the same sections (data not shown). The intervening region was equally divided into three layers. Sublayer 6 was defined as the layer below layer 6. Scale bar, 200 μm. C, D, Comparison of LHX6-positive cells in Gbx2 mutants (red dots) and wild-type littermates (blue dots) in medial PFC at P0. Each line connecting red and blue dots represents a pair of brains analyzed in the same experiment (n = 5). C shows laminar distribution pattern. The p values of paired t test for individual layer are shown. In repeated-measures two-way ANOVA, the p values for layer, pair (between control and knockout), and interactions (between layer and pair) are 0.0001, 0.5950, and 0.0021, respectively. D shows the total number of Pvalb- and Sst-expressing neurons in all layers. The p values of paired t tests are shown. “Sub-L6” was defined as the region below the expression domain of TBR1, which was stained in all immunostaining slides for a reference. *p < 0.05, **p < 0.005, ***p < 0.0005. E–G, Comparison of Pvalb-positive and Sst-positive cells in Gbx2 mutants (red dots) and wild-type littermates (blue dots) in medial PFC at P21. Each line connecting red and blue dots represents a pair of brains analyzed in the same experiment (n = 5). E, F, Comparison of laminar distribution of Pvalb-expressing and Sst-expressing neurons, respectively, in Gbx2 mutant mice and littermate controls. Layer 1 was defined as the cell-sparse layer detected by DAPI staining. The remaining cortical wall was equally divided into three layers. The deepest layer (shown as “L5/6”) contains the entire layer 6 and the deep part of layer 5. G, The total number of Pvalb- and Sst-expressing neurons in all layers. The p values of a paired t test for an individual layer are shown. In repeated-measures two-way ANOVA, the p values for layer, pair (between control and knockout), and interactions (between layer and pair) are <0.0001, 0.0002, and 0.0001 (Pvalb); <0.0001, 0.0247, and 0.0001 (Sst), respectively. H, Comparison of the numbers of cleaved caspase 3-positive, LHX6-positive cells in Gbx2 mutants (red dots), and wild-type littermates (blue dots) in medial PFC at P8. Each line connecting red and blue dots represents a pair of brains analyzed in the same experiment (n = 4). Each value is the mean of 7–15 sections. The p values of ratio paired t tests are shown.

Figure 8.

Normal induction of Cyp26b1 in PFC in mice expressing tetanus toxin light chain in thalamocortical axons. A–D, Immunostaining for VAMP2 on frontal sections of somatosensory cortex at E16.5 control (A, B) and mutant mice with ectopic expression of TeNT in thalamic neurons (C, D). TeNT expression leads to the deletion of VAMP2, specifically in thalamocortical axons at E16.5. Thalamocortical axons are shown by NetrinG1 staining (B, D, green). In control brains, both thalamocortical (bracket, A–D) and corticofugal (asterisk, A–D) axons express VAMP2, whereas in TeNT-expressing mice, VAMP2 staining is specifically diminished in thalamocortical axons (C, D, bracket). Scale bar, 500 μm. E, F, Deletion of VAMP2 in thalamocortical axons results in the lack of the characteristic pattern of RORβ expression in the barrel field of primary somatosensory cortex at P8 (arrow), similar to the defect found in Gbx2 mutant mice (Vue et al., 2013). G–J, Immunostaining for VAMP2 on frontal sections of prefrontal cortex at P0 control (G, H) and mutant mice with ectopic expression of TeNT in thalamic neurons (I, J). Similar to the somatosensory cortex, VAMP2 staining in thalamocortical axons is diminished in TeNT-expressing mice (I, J, bracket). Scale bar, 200 μm. K, L, Expression of Cyp26b1 in medial (arrowhead) and ventral PFC is intact in TeNT-expressing mice (L), similar to control (K), at P8. Scale bars: G, H, I, J, 200 μm; E, F, K, L, 1 mm.

Figure 9.

Schematic diagrams of the current finding. A–C, Embryonic roles of thalamocortical axons as observed in neonatal mice. A, Thalamocortical axons reach the medial PFC (mPFC) by E16.5 and control migration of MGE-derived interneurons. B, In normal neonatal mice, MGE-derived interneurons have largely completed tangential migration to the mPFC and have taken proper laminar positioning by radial dispersion (arrows). C, In thalamus-specific Gbx2 mutant mice, radial positioning of MGE-derived interneurons are aberrant, resulting in their accumulation in layer 6 and below. D–G, Postnatal roles of thalamus–PFC interactions and RA-degrading enzyme CYP26B1 in the development of PV interneurons in the mPFC. D, Early postnatal mPFC is positioned between the source of RA synthesis (layer 2, by ALDH1A3) and the RA-degrading “sink” (layer 6, by CYP26B1). The expression of both enzymes is induced early postnatally, but only Cyp26b1 is dependent on the connections with the thalamus. The main cell population that responds to RA in postnatal mPFC is PV interneurons, and their development is controlled by CYP26B1. E, In normal postnatal mice, PV neurons mature and start to express Pvalb mRNA and PV protein mainly in deep layers of mPFC between P7 and P14. F, In thalamus-specific Gbx2 mutant mice, Cyp26b1 is not induced in mPFC. The number of both Pvalb and Sst-expressing neurons is reduced in the middle layers at least partly due to the earlier defects in radial dispersion (described in C). G, In frontal cortex-specific Cyp26b1 mutant mice, lack of the RA sink in mPFC leads to an increased number of neurons that express Pvalb mRNA or PV protein in deep layers.