An Early Cortical Progenitor-Specific Mechanism Regulates Thalamocortical Innervation

Abstract

The cortical subplate is critical in regulating the entry of thalamocortical sensory afferents into the cortex. These afferents reach the subplate at embryonic day (E)15.5 in the mouse, but “wait” for several days, entering the cortical plate postnatally. We report that when transcription factor LHX2 is lost in E11.5 cortical progenitors, which give rise to subplate neurons, thalamocortical afferents display premature, exuberant ingrowth into the E15.5 cortex. Embryonic mutant subplate neurons are correctly positioned below the cortical plate, but they display an altered transcriptome and immature electrophysiological properties during the waiting period. The sensory thalamus in these cortex-specific Lhx2 mutants displays atrophy and by postnatal day (P) 7, sensory innervation to the cortex is nearly eliminated leading to a loss of the somatosensory barrels. Strikingly, these phenotypes do not manifest if LHX2 is lost in postmitotic subplate neurons, and the transcriptomic dysregulation in the subplate resulting from postmitotic loss of LHX2 is vastly distinct from that seen when LHX2 is lost in progenitors. These results demonstrate a mechanism operating in subplate progenitors that has profound consequences on the growth of thalamocortical axons into the cortex.

SIGNIFICANCE STATEMENT Thalamocortical nerves carry sensory information from the periphery to the cortex. When they first grow into the embryonic cortex, they “wait” at the subplate, a structure critical for the guidance and eventual connectivity of thalamic axons with their cortical targets. How the properties of subplate neurons are regulated is unclear. We report that transcription factor LHX2 is required in the progenitor “mother” cells of the cortical primordium when they are producing their “daughter” subplate neurons, in order for the thalamocortical pathway to wait at the subplate. Without LHX2 function in subplate progenitors, thalamocortical axons grow past the subplate, entering the cortical plate prematurely. This is followed by their eventual attrition and, consequently, a profound loss of sensory innervation of the mature cortex.

Introduction

Sensory input reaches the neocortex via thalamocortical tract, and the precise topography and connectivity of this tract are critical to sensory processing. A key feature of cortical innervation by thalamocortical axons is a “waiting period” during which thalamic afferents enter the dorsal telencephalon while cortical neurogenesis is still underway, but “wait” within the cortical subplate for a period lasting several days in rodents, before innervating layer 4 of the cortex (Lund and Mustari, 1977; Rakić, 1977; Shatz and Luskin, 1986; Ghosh and Shatz, 1992; Herrmann et al., 1994; Auladell et al., 2000; Hevner, 2000). The importance of the subplate as an intermediate target of the thalamocortical afferents is well established from studies in which subplate neurons were either experimentally ablated or mispositioned/reduced because of genetic perturbations (for review, see Hoerder-Suabedissen and Molnár, 2015). Thalamocortical afferents make synaptic contacts with the subplate (Herrmann et al., 1994), and electrical activity in the subplate is critical for correct thalamocortical innervation of the cortical plate at later stages (Ghosh et al., 1990; Catalano and Shatz, 1998; Kanold and Shatz, 2006; Tolner et al., 2012). Subplate neurons are born at embryonic day (E) 11.5 in the mouse from common progenitors that will later produce the neurons of the cortical plate (Molyneaux et al., 2007). However, the mechanisms that regulate the properties of the subplate neurons that are necessary for the waiting period are unclear because the regulatory processes in cortical progenitors that give rise to subplate neurons are poorly understood.

Previously, we reported a loss of the somatosensory barrels upon cortex-specific deletion of Lhx2 (Shetty et al., 2013). Barrels are a prominent feature of the rodent somatosensory cortex, formed by local aggregates of layer 4 cortical neurons when thalamocortical afferents innervate the cortex (Woolsey and van der Loos, 1970). We found that the barrels are absent in postnatal brains when Lhx2 is disrupted using an Emx1Cre line, which acts in cortical progenitors from E11.5 (Emx1Cre^11.5) (Shetty et al., 2013). In contrast, disruption of Lhx2 using the postmitotic neuron-specific NexCre permits somatosensory barrels to be formed postnatally in layer 4, albeit with some deficits (Zembrzycki et al., 2015; Wang et al., 2017). Based on these two distinct phenotypes, we hypothesized that LHX2 function in cortical progenitors is required for the key aspects of thalamocortical axon development.

Here, we analyze thalamocortical axon projection defects when Lhx2 is disrupted in cortical progenitors using Emx1Cre^11.5, focusing on the somatosensory projections. We find that, although there is extremely limited thalamocortical innervation detectable in the postnatal cortex of Emx1Cre^11.5::Lhx2^lox/lox mice, there is a premature exuberant ingrowth of thalamocortical fibers into the cortical plate in the embryo at E15.5. These fibers extend through the cortical plate and reach the marginal zone. This is followed by atrophy of the prominent sensory ventrobasal nucleus (VB) of the thalamus, and eventual loss of thalamocortical innervation of the cortex. Using CreER-driven tamoxifen-inducible Lhx2 deletion, we identify that loss of LHX2 from E11.5, but not from E12.5 or E13.5, recapitulates the premature ingrowth phenotype, implicating the subplate in this process, since it is born at E11.5. The subplate in Emx1Cre^11.5::Lhx2^lox/lox brains expresses specific markers and occupies its appropriate position under the cortical plate but exhibits severe dysregulation of its transcriptome. In addition, the electrophysiological properties of these subplate neurons are perturbed compared with controls. Our results offer novel insights into mechanisms that operate in cortical progenitors that are necessary for proper subplate differentiation and for thalamocortical axon development and circuit assembly.

Materials and Methods

Mice

All animal protocols were approved by the Institutional Animal Ethics Committee of Tata Institute of Fundamental Research according to regulations devised by the Committee for the Purpose of Control and Supervision of Experiments on Animals. Noon of the day of vaginal plug was designated as embryonic day 0.5 (E0.5). Mouse embryos of either sex were harvested at E15.5, E17.5, E18.5, P0, P2, and P7. The following lines were obtained from JAX labs: tdTomato reporter Ai9 (strain name: B6. CgGt (ROSA)26Sortm9 (CAG-tdTomato) Hze/J; stock #007909), tamoxifen-inducible CreERT2 line (strain name: B6; 129- Gt (ROSA)26Sortm1 (Cre/ERT)Nat/J; stock #004847). The following lines were obtained from: Emx1Cre^11.5 [Yuqing Li, University of Florida (Jin et al., 2000)]; Lhx2^lox/lox [Edwin Monuki, University of California (Mangale et al., 2008)]; NexCre [Klaus Nave, Max Planck Institute for Experimental Medicine (Goebbels et al., 2007)]; TCA-GFP or 5HTT-GFP/SERT-GFP, a BAC-transgenic line [Takuji Iwasato, National Institute of Genetics (Mizuno et al., 2014)].

The Emx1Cre^11.5 line (referred to as Emx1Cre^YL in Shetty et al., 2013) acts a day later than a commonly used Emx1Cre line that acts from E10.5 (Emx1Cre^10.5, referred to as Emx1Cre^KJ in Shetty et al., 2013). Using Emx1Cre^11.5 to disrupt Lhx2 permits the neocortex to be specified (Shetty et al., 2013) instead of being transformed into an ectopic paleocortex, which is seen with Emx1Cre^10.5 (Chou et al., 2009). Although reduced in extent compared with controls, the Emx1Cre^11.5::Lhx2^lox/lox cortex expresses molecular markers of all neocortical layers in the correct order (Shetty et al., 2013). Since Emx1Cre action is specific to the dorsal telencephalon, the Emx1Cre^11.5 permits the examination of disruption of Lhx2 in neocortical progenitors from E11.5 without simultaneously affecting Lhx2 expression in the thalamus, making it an ideal reagent to examine cortex-specific effects on thalamocortical innervation. Lhx2 is expressed in progenitors and also in newly postmitotic cells (Telley et al., 2016). We confirmed that NexCre-mediated recombination of an Ai9 reporter labels the postmitotic subplate (see Fig. 5D), and Lhx2 expression is barely detectable in the NexCre::Lhx2 ^lox/lox brain (Zembrzycki et al., 2015; Wang et al., 2017).

Figure 5.

Subplate neurons are specified and positioned below the cortical plate upon Emx1Cre-mediated loss of Lhx2. A-C, Expression of subplate markers in control, Emx1Cre^11.5::Lhx2^lox/lox, and NexCre::Lhx2^lox/lox brains (n = 3 brains per condition). At P7, CTGF expression and CPLX3 immunoreactivity reveal a tightly packed subplate in controls (A) and NexCre::Lhx2^lox/lox brains (C), but a less compact subplate in Emx1Cre^11.5::Lhx2^lox/lox brains (B), located below the cortical plate (CP). D, NexCre drives the expression of an Ai9 reporter in the cortical plate and also the subplate, which contains Ai9 and CPLX3-coexpressing cells in P7 NexCre::Ai9::Lhx2^lox/lox brains. E, F, At E15.5, Nurr1 is expressed in both control and Emx1Cre^11.5::Lhx2^lox/lox subplate neurons (arrowheads). Lhx2 expression is lost in the cortex of Emx1Cre^11.5::Lhx2^lox/lox brains (white asterisk), whereas expression in the thalamus is maintained (black asterisk). The CPLX3 immunohistochemistry panels in A, B and the low-magnification image in D are composites of multiple confocal images. Scale bars: A-C, E, F, 200 µm; D, low-magnification image, 100 µm; high-magnification images, 10 µm.

Sample preparation

For histological procedures, embryonic brains were harvested in prechilled 1× PBS. Postnatal pups were anesthetized on ice before transcardial perfusion with 4% (wt/vol) PFA (Sigma Millipore) made in 0.1 M PB, pH 7.4. Brains were fixed in 4% PFA overnight at 4°C and equilibrated in 30% sucrose before sectioning.

Administration of tamoxifen

Tamoxifen (Sigma) dissolved in corn oil (20 mg/ml) was administered by oral gavage to pregnant Lhx2^lox/lox dams at different time points as mentioned in the text, and embryos were harvested at E15.5. For each embryo, the extent of recombination was examined in one series of sections. Control embryos were littermates with one WT copy of the relevant gene. For Lhx2^lox/lox mice, the tamoxifen dosage administered was 75 µg/g body weight.

In situ Hybridization (ISH)

ISH was performed as follows: the sections were fixed in 4% PFA, washed in 1 × PBS, and treated with Proteinase K (1 µg/ml). Hybridization was performed overnight at 70°C in hybridization buffer (4× SSC, 50% formamide, and 10% SDS) containing different antisense RNA probes. Post-hybridization washes were performed at 70°C in Solution X (2× SSC, 50% formamide, and 1% SDS). These were followed by washes in 2× SSC, 0.2× SSC, and then TBS–1% Tween 20 (TBST). The sections were incubated in anti-digoxigenin Fab fragments (Roche) at 1:5000 in TBST overnight at 4°C. The color reaction was performed using NBT/BCIP (Roche) in NTMT (100 mm NaCl, 100 mm Tris-pH 9.5, 50 mm MgCl₂, and 1% Tween-20) according to the manufacturer's instructions.

Probe preparation

All probes were prepared by in vitro transcription using a kit from Roche as per the manufacturer's instructions. Templates for Calb2, Chst2, CTGF, Hpca, Lhx2 exon2–3, Nrp2, Nurr1, Prox1, and SERT were generated by PCR using specific primers from E15.5 (for Lhx2 exon 2-3, Nurr1), P0 (for Hpca, Nrp2), and P7 (for Calb2, Chst2, CTGF, Prox1, SERT) mouse brain cDNA (T7 polymerase promoter sequence was added to the reverse primer sequence). Template for Cad6 and Lhx9 was generated from plasmid DNA by restriction enzyme digestion. Calb2-forward, AGCACTTTGATGCTGACGGA, Calb2-reverse, AGGTGGTGAGCTGTTGGATG; Chst2-forward, CATCTTTGGGGCAGCCACTA, Chst2-reverse, CGAAAGGCTTGGAGGAGGAG; CTGF-forward, AGCGGTGAGTCCTTCCAAAG, CTGF-reverse, GTAATGGCAGGCACAGGTCT; Hpca-forward, CAGGACCTGCGAGAGAACAC, Hpca-reverse, CAGTCCTCTTTTCCGGGGTC; Lhx2 exon2-3-forward, CGCGGATCCACCATGCCGTCCATCAGC, Lhx2 exon2-3-reverse, TAATACGACTCACTATAGGG; Nrp2-forward, AGAAGCCAGCAAGATCCACC, Nrp2-reverse, GGCCAGACTCCATTCCCAAA; Nurr1-forward, CAGTCCGAGGAGATGATGCC, Nurr1-reverse, AACCATCCCAACAGCTAGGC; Prox1-forward, 5′-GCAGGCCTACTATGAGCCAG-3′, Prox1-reverse, 5′-TTGACCACCGTGTCCACAA-3′; SERT-forward, CAAAACGTCTGGCAAGGTGG, SERT-reverse, CATACGCCCCTCCTGATGTC.

Immunohistochemistry

Brains were sectioned (30 µm) using a freezing microtome, and sections were mounted on Superfrost Plus slides (Electron Microscopy Sciences). Sections were quenched with 0.1 m PB, pH 7.4, containing 50 mm NH₄Cl for 10 min followed by washing with 0.1 m PB containing 0.1% (v/v) Triton X-100 for 10 min. Sections were incubated in blocking solution containing 10% (v/v) horse serum (Invitrogen) in 0.1 m PB with 0.3% (v/v) Triton X-100 (Sigma) for 45 min at room temperature. The sections were then incubated overnight at 4°C in primary antibodies: biotinylated goat anti-GFP (1:400; Abcam, catalog #ab6658), mouse anti-RFP (1:200; Allele Biotech, catalog #ABP-MAB-RT008), rabbit anti-CPLX3 (1:200; Synaptic Systems, catalog #122302), and rabbit anti-cleaved Caspase-3 (1:200; Cell Signaling, catalog #9664S). The sections were then washed in 0.1 m PB, followed by incubation in secondary antibody for 1.5 h at room temperature. Secondary antibodies used were as follows: streptavidin Alexa-488 (1:800; Invitrogen, catalog #S32354) for GFP, goat anti-mouse antibody conjugated to Alexa-568 (1:400, Invitrogen, catalog #A11004) for RFP, and donkey anti-rabbit antibody conjugated to Alexa-647 (1:400, Invitrogen, catalog #A31573) for Complexin3 and cleaved Caspase-3. The slides were again washed with 0.1 m PB and mounted in Fluoroshield (Sigma). For cytochrome oxidase staining, free-floating sections (50 µm vibratome sections, Leica Microsystems, VT1000S) were placed in a solution of 0.6 mg/ml DAB, 0.3 mg/ml Cytochrome C (Sigma), 45 mg/ml sucrose in 0.1 m PBS at 37°C until staining appeared.

In utero electroporation

This procedure was used to visualize the thalamocortical tract at embryonic stages since the TCA-GFP line does not display GFP expression in the thalamocortical axons until postnatal day (P) 2. In utero electroporation was performed to transfect a GFP-expressing plasmid into the diencephalon of control and mutant embryos at E11.5, when the neurons of the VB nucleus are born.

All procedures conducted followed the guidelines prescribed by the Institutional Animal Ethics Committee. Timed pregnant Lhx2^lox/lox dams with E11.5 embryos were anesthetized with isoflurane (4% induction, 3% during the surgery) followed by oral administration of Meloxicam (Melonex, United Pharmacies). Uterine horns were successively exposed after a midline laparotomy of 1 cm. Embryos were injected with 1-2 µl plasmid DNA solution dissolved in nuclease-free water with 0.1% Fast Green, into the third ventricle (for targeting the dorsal thalamus) through the uterine wall using a fine-glass microcapillary. pCAGG-IRES-eGFP (2 µg/µl, a gift from G. Fishell, Harvard Medical School) was injected, and plasmid DNA was prepared using Macherey Nagel Nucleobond Xtra MaxiPrep Kit (catalog #740414). Embryos were electroporated by holding their head between tweezers-style circular electrodes (3 mm diameter, BTX Harvard Apparatus, catalog #450487) across the uterus wall by delivering three pulses (34 V, 40 ms duration with 999 ms intervals) with a square-wave electroporator (Nepagene CUY21). The uterine horns were returned into the abdominal cavity, the incision was sutured, and the embryos were allowed to continue their normal development until E15.5. The uterine horns and the abdominal cavity were kept moist throughout the surgery (20-25 min) by flushing 0.9% NaCl (prewarmed to 37°C). Animals were kept on a 37°C warm plate for 30 min for postsurgical recovery. An oral suspension of Meloxicam (Melonex, United Pharmacies) was mixed with the water in the feeding bottles of the dams (0.6 µl/ml) as an analgesic and given to the animals until 2 d after surgery.

Imaging

Bright-field images were taken using a Carl Zeiss Axioplan 2 + microscope, Nikon Digital Sight DS-F12 camera, and Nikon NIS 4.0 imaging software. Images of immunohistochemistry slides were obtained using a Carl Zeiss LSM 710 and LSM 880 imaging system. Image stacks were processed using ImageJ (National Institutes Health) (Schindelin et al., 2012) and ZEN 2.1 (black) software. Confocal images of multiple fields were acquired using tile-scanning mode using a motorized stage and were stitched using Zen2.1 software. Low-magnification images were acquired using 10× objective, and high-magnification images were acquired at 40× objective. Exposure, z-stack (20 µm) and gain settings were uniform across all tile images taken for a given composite image. Figure panels were prepared using Adobe Photoshop CS6 and Adobe Illustrator.

Microdissection

P0 brains were dissected in ice-cold HBSS, and 200-µm-thick sections were cut using a vibratome (VT1000S). The lowermost part of the cortical plate (presumptive subplate) was visually identified and micro-dissected under stereomicroscopic guidance in ice-cold HBSS under RNase-free conditions. Micro-dissected subplate tissue was stored in Buffer RLT Plus (QIAGEN) at −80°C, and RNA was extracted using RNeasy PLUS Micro kit (QIAGEN).

RNA sequencing

RNA sequencing was performed by Medgenome Technologies. mRNA was isolated from total RNA, cDNA was prepared, and libraries were prepared using the IlluminaTruSeq RNA Sample preparation V2. Libraries were multiplexed and sequenced as 100 bp pair-end reads using the Illumina HiSEQ2500 platform, at an expected depth of 60 M single reads. Sequenced reads were aligned on the latest mouse reference genome assembly (mm10) using HISAT2, SAM/BAM files were further processed using SAMtools, and the number of reads per transcript was calculated and normalized by DESeq2 using R software (Love et al., 2014). Genes were considered significantly differentially expressed for FDR of <0.1. For representation, genes with fold change cutoff of ±1.5 were used. Mutually exclusive and overlapping genes between datasets were identified using Venny (2.1.0). Gene ontology analysis was performed using Gene Ontology Resource (http://geneontology.org/). The RNA-seq data generated in this study have been deposited in the GEO database under accession number GSE159006.

Electrophysiology

Brain slices were prepared as described previously (Hirsch and Luhmann, 2008). Briefly, E15.5 embryos and P0-P1 Lhx2^lox/lox and Emx1Cre^11.5::Lhx2^lox/lox mouse pups were anesthetized by hypothermia and decapitated. Coronal slices (400 µm thick), including the primary somatosensory cortex, were cut in ice-cold aCSF (Hirsch and Luhmann, 2008) on a vibratome (VT1200S). Slices were allowed to equilibrate in oxygenated aCSF for 30 min in a storage chamber before being transferred to the submerged type recording chamber at 30°C-34°C. During preparation and recording procedures, slices were maintained in aCSF containing the following (in mm): 124 NaCl, 26 NaHCO₃, 3 KCl, 1.6 CaCl₂, 1.8 MgCl₂, 1.25 NaH₂PO₄, and 20 D-glucose, pH 7.4, after equilibration with 95% O₂−5% CO₂ (osmolarity 333 mOsm). The subplate neurons were visualized using infrared differential interference contrast optics using a 40× objective (water immersion lens, 0.9 NA) on an Olympus BX1WI microscope. They were identified by their location, morphology, and electrophysiological properties (Hirsch and Luhmann, 2008). Subplate neurons had typical horizontal bipolar morphologies and were located between the cell-dense cortical plate with radially oriented neurons and the cell-sparse white matter. Whole-cell patch-clamp recordings in current-clamp mode were obtained with an Axopatch 200B amplifier (Molecular Devices) and patch-pipettes (3-5 mΩ resistance) filled with the following (in mm): K-gluconate 155, MgCl₂ 2, NaHEPES 10, Na-PiCreatine 10, Mg₂-ATP₂ and Na₃-GTP 0.3, pH 7.3 (300 mOsm), and 2 mg/ml Biocytin (Sigma) for post hoc anatomical staining. Series resistance and input resistance were continuously monitored during the experiment, and the cell was discarded if these parameters changed by >25%. Slices were perfused at all times with aCSF at a flow rate of 2-3 ml/min. Electrophysiological recordings were digitized using Digidata 1400A. Current-clamp recordings were filtered at 10 kHz and sampled at 20 kHz. All analysis was done using custom-written code in MATLAB (R2013a) and Clampfit (10.7.0.3), and plots were made in GraphPad Prism software. Input resistance was measured by computing slope of the I-V curve obtained by current injections from –10 pA to 10 pA, in steps of 5 pA.

Experimental design and statistical analysis

The thalamocortical pathway phenotype we report in Emx1Cre^11.5::Lhx2^lox/lox mice is 100% penetrant. Together, 92 Emx1Cre^11.5::Lhx2^lox/lox brains, 12 CreERT2::Lhx2^lox/lox brains, 15 NexCre::Lhx2^lox/lox brains, and 87 control brains were analyzed for this study. Animals of both sexes were examined. The Emx1Cre^11.5 allele was always maintained in a hemizygous condition. Therefore, each litter produced ∼50% controls without the Cre transgene. These littermate controls were used wherever possible. We did not find any variation between litters at any given stage. The detailed breakup of embryos/pups for each experiment/quantification is below:

Figure 1:

Figure 1.

Emx1Cre-mediated loss of Lhx2 causes severe disruption of the somatosensory barrels. A, TCA-GFP reporter expression reveals the thalamocortical innervation and somatosensory barrels in flattened tangential sections and coronal sections at P7 in control brains. B, In Emx1Cre^11.5::Lhx2^lox/lox::TCA-GFP brains, aberrant thalamocortical innervation (arrowheads) is seen clustered in a limited region of the cortex at P7. C, NexCre::Lhx2^lox/lox::TCA-GFP brains reveal that postmitotic deletion of Lhx2 permits thalamocortical afferents to innervate layer 4 and barrels can be detected. D, Fluorescence intensity of the somatosensory cortex, measured in coronal sections, is significantly reduced in P7 Emx1Cre^11.5::Lhx2^lox/lox brains compared with controls. Unpaired two-tailed Student's t test, ***p = 0.0001; n = 5 for each genotype. Tangential and coronal TCA-GFP images in A-C are composites of multiple confocal images. Scale bars, 500 µm.

n = 5 for control and Emx1Cre^11.5::Lhx2^lox/lox.

n = 3 for NexCre::Lhx2^lox/lox. Our results were consistent with the phenotype for this genotype already published by other groups (Zembrzycki et al., 2015; Wang et al., 2017).

For quantification, three sections spanning the somatosensory cortex were scored in each brain. The fluorescent intensity of axonal terminals marked by TCA-GFP was measured in coronal sections using ImageJ in the area shown in the diagram (Fig. 1D). The fluorescence intensity was computed as the corrected pixel intensity difference between the “integrated density” of the selected area and the mean fluorescence of background readings from the same area, using functions available in ImageJ.

Figure 2:

Figure 2.

Emx1Cre-mediated loss of Lhx2 causes shrinkage of the somatosensory VB nucleus from embryonic stages. A–H, A comparison of control and Emx1Cre^11.5::Lhx2^lox/lox mutant thalamus. A, Cytochrome oxidase staining at P7 reveals a greatly shrunken VB (white ovals) and dorsal thalamus (black outline) on cortex-specific loss of Lhx2. B, E, The VB shrinkage is apparent at P0 by SERT and Cad6 expression within this nucleus (black ovals, B) and Chst2, Prox1, Lhx9, and Calb2 expression along its borders and in other sensory nuclei (black ovals, E). C, D, G, Examination of SERT expression in control and mutant brains from E17.5- P7 brains reveals that the shrinkage of the VB is apparent from E17.5. Area measurements were quantified and compared in G using ordinary one-way ANOVA; E17.5, *p = 0.0153; E18.5, **p = 0.0015; P0, ****p < 0.0001; P2, ***p = 0.0004; P7, ****p < 0.0001. n = 6 for ages E17.5-E18.5; n = 5 for P0; n = 3 for ages P2-P7 for each genotype. F, Cleaved Caspase-3 immunostaining indicates increased cell death in the VB nucleus at E18.5 and P0 on cortex-specific loss of LHX2. H, An Ai9 reporter (red) reveals that corticothalamic axons in both control and Emx1Cre^11.5::Lhx2^lox/lox brains have reached the reticular thalamic nucleus at E18.5 (white solid lines), but have not yet innervated the VB nucleus. Scale bars: A-E, H, 500 µm; high-magnification images in F, 200 µm. I, J, Quantification of dLGN area (I) and the total dorsal thalamus area (J). A comparison of the average values from control and Emx1Cre^11.5::Lhx2^lox/lox brains reveals a shrinkage of these structures in Emx1Cre^11.5::Lhx2^lox/lox brains at P2 and P7 (dLGN) and P7 (dorsal thalamus). n = 3 brains for each age and genotype. I, Ordinary one-way ANOVA; E18.5, not significant, p = 0.3130; P0, not significant, p = 0.9271; P2, *p = 0.0248; P7, **p = 0.0017. J, Ordinary one-way ANOVA; E17.5, not significant, p = 0.2500; E18.5, not significant, p = 0.0972; P0, not significant, p = 0.5742; P2, not significant, p = 0.3111; P7, ***p = 0.0001.

Area measurements of the thalamus/thalamic nuclei: n = 6 for ages E17.5-E18.5; n = 5 for P0; n = 3 for ages P2-P7 for each genotype. For each embryo, VB/dLGN/dorsal thalamus area measurements were taken from three sections spanning the rostro-caudal extent of the structure using ImageJ software. The average mutant value was normalized to that of the average control value.

For thalamic nuclei markers/cytochrome oxidase/cCaspase3 staining, a total of 12 control and 12 mutant brains were examined for each genotype, and each marker was examined in at least three brains.

Figures 3 and 4:

Figure 3.

Embryonic thalamocortical axons are defasciculated and display premature growth into the cortical plate when Lhx2 is disrupted using Emx1Cre, but not NexCre. A, Electroporation of a plasmid encoding GFP into the diencephalon at E11.5 labels the thalamus and also thalamocortical axons by E15.5. B, Cortical thickness is unaffected at E15.5 in Emx1Cre^11.5::Lhx2^lox/lox and NexCre::Lhx2^lox/lox brains. The average cortical thickness from the mutant sections was normalized to that of the respective controls. Ordinary one-way ANOVA; Control versus Emx1Cre^11.5::Lhx2^lox/lox, not significant, p > 0.9999; Control versus NexCre::Lhx2^lox/lox, not significant, p = 0.8440; Emx1Cre^11.5::Lhx2^lox/lox versus NexCre::Lhx2^lox/lox, not significant, p = 0.8440. n = 3 brains for each genotype. C, D, Loss of Lhx2 using Emx1Cre causes thalamocortical afferents to defasciculate as they course through the dorsal telencephalon at E15.5, whereas they remain tightly fasciculated in controls. In Emx1Cre^11.5::Lhx2^lox/lox brains, thalamocortical axons enter the cortical plate prematurely and extend aberrantly up to the marginal zone (arrowheads). E, F, When Lhx2 is disrupted using NexCre, thalamocortical afferents appear indistinguishable from those in the controls in E15.5 brains. G, Sections of Emx1Cre^11.5::Lhx2^lox/lox, NexCre::Lhx2^lox/lox and control E15.5 brains were scored for aberrant axonal branching. Scatter dots in the bar plot represent aberrant branching value for each brain. Ordinary one-way ANOVA; Control versus Emx1Cre^11.5::Lhx2^lox/lox, ****p < 0.0001; Control versus NexCre::Lhx2^lox/lox, not significant, p = 0.9992; Emx1Cre^11.5::Lhx2^lox/lox versus NexCre::Lhx2^lox/lox, ****p < 0.0001. n = 3 brains for each genotype. All fluorescence images are composites of multiple confocal images. C, D, Boxes represent regions displayed at high magnification in the adjacent panels. Scale bars, 100 µm.

Figure 4.

Early, but not late, cortical progenitors require LHX2 for normal thalamocortical innervation. A-F, A′-F′, Sections of E15.5 Control (A,A′), Emx1Cre^11.5::Lhx2^lox/lox (B,B′), CreERT2::Lhx2^lox/lox (C-F,C′-F′) brains in which thalamocortical innervation is visualized by electroporation of a plasmid encoding GFP into the thalamus at E11.5. A, A′, Thalamocortical afferents in control brains course in a tight bundle below the cortical plate (CP). B, B′, Loss of Lhx2 using Emx1Cre causes premature ingrowth of thalamocortical afferents into the cortical plate that extend up to the marginal zone (arrowheads). C-F, C′-F′, This phenotype is recapitulated when tamoxifen is administered at E11.5 to CreERT2::Lhx2^lox/lox but not E12.5, E12.75, or E13.5 (D-F). G, G′, Aberrant axonal branching of TCAs seen in CreERT2::Lhx2^lox/lox Tam 11.5 brains is similar to Emx1Cre^11.5::Lhx2^lox/lox brains. The exuberant branching phenotype is significantly reduced when tamoxifen is administered at E12.5, E12.75, or E13.5. Scatter dots in the bar plot represent aberrant branching value for each brain. Ordinary one-way ANOVA; Control versus Emx1Cre^11.5::Lhx2^lox/lox, ****p < 0.0001; Control versus CreERT2::Lhx2^lox/lox Tam E11.5, ****p < 0.0001; Emx1Cre^11.5::Lhx2^lox/lox versus CreERT2::Lhx2^lox/lox Tam E11.5, not significant, p = 0.9969; n = 3 brains for each genotype. All panels are composites of multiple confocal images. A′-F′ are high-magnification images of A-F, respectively. Scale bars, 100 µm.

For each condition, three embryos with the most extensive thalamic electroporation and no mis-targeting in the cortex were selected for quantification. Examples of the extent of thalamic electroporation are in Figure 3A. In each embryo, electroporated fibers were seen in the cortex in 2 or 3 contiguous sections. To maintain consistency, 2 sections with the most thalamocortical fibers labeled were scored. The area occupied by aberrant axonal branches (pink solid lines in the diagram) was measured using ImageJ, and normalized to the area of the cortical plate (dark gray region in the diagram) in mutants and controls.

Cortical thickness (Fig. 3B) was measured in the same sections scored for thalamocortical afferent ingrowth (Fig. 3G). The mutant value was normalized to that of the control value.

Figure 5:

For each genotype, 8 (P7) and 5 (E15.5) brains were analyzed for subplate markers. Lhx2 expression was examined in serial sections.

Figure 6:

Figure 6.

Transcriptomic analysis of the subplate. A, Emx1Cre^11.5::Lhx2^lox/lox, NexCre::Lhx2^lox/lox, and control sections at P0 displaying CTGF expression in the subplate, and a control section in which the subplate was removed by microdissection. B, Bulk RNA-seq revealed 328 and 167 differentially expressed genes unique to the Emx1Cre^11.5::Lhx2^lox/lox, NexCre::Lhx2^lox/lox subplate, respectively, and 29 overlapping differentially expressed genes. C, D, GO analysis of the differentially expressed genes in the Emx1Cre^11.5::Lhx2^lox/lox subplate (C) and selected genes in different functional categories (D; for related data, see Extended Data Figs. 6-1, 6-2, and 6-3). E, Bar plot representing log₂fold change values of representative differentially expressed genes Nrp2 and Hpca. F, Nrp2 expression is decreased and Hpca expression is increased in the Emx1Cre^11.5::Lhx2^lox/lox subplate (arrowheads). SP, Subplate. Scale bars: A, F, 500 µm; high magnification images in F, 50 µm.

Micro-dissected tissue was pooled from 8 slices (four slices each from 2 brains) for each replicate sample. Two replicates each of control, Emx1Cre^11.5::Lhx2^lox/lox and NexCre::Lhx2^lox/lox, were processed for RNA sequencing after ascertaining that their RIN values were >8.

Figures 7 and 8:

Figure 7.

LHX2-deficient subplate neurons are electrically silent at E15.5 and electrophysiologically immature up to birth. A–F, Spontaneous responses from slices of E15.5 control and Emx1Cre^11.5::Lhx2^lox/lox brains. A, The majority of control neurons fire regularly in absence of any current input, whereas the majority of mutant neurons are electrically silent. B, Typical examples of mutant and control traces. C–F, RMP and IR of subplate neurons from E15.5 and P0 control and Emx1Cre^11.5::Lhx2^lox/lox slices. Both parameters are similar in mutant and control neurons at E15.5 but control neurons display a hyperpolarized RMP and lower IR by birth, which the mutant neurons do not (C,E). For E15.5, n = 14 cells were examined from 7 Control brains and n = 17 cells from 10 Emx1Cre^11.5::Lhx2^lox/lox brains. For P0-P1, n = 16 cells were examined from 5 Control brains and n = 17 cells from 5 Emx1Cre^11.5::Lhx2^lox/lox brains. Data are mean ± SEM. Individual values and statistical analyses are detailed in Extended Data Figure 7-1. **p ≤ 0.01. ****p ≤ 0.0001. D, F, Typical examples of mutant and control traces at P0.

Figure 8.

LHX2-deficient subplate neurons display attenuated evoked responses at E15.5 and birth. A–E, Evoked responses from slices of E15.5 Control and Emx1Cre^11.5::Lhx2^lox/lox brains. A, B, Mutant neurons have an attenuated spike frequency compared with controls at E15.5 and P0. Typical examples of mutant and control traces are shown adjacent to the line graphs. C, The proportion of single-spiking neurons is greater in mutants compared with controls at E15.5 and birth. D, Mutant neurons display decreased spike amplitudes at both stages. E, Current injection required to elicit the first action potential is higher for E15.5 mutant neurons. Data are mean ± SEM. Individual values and statistical analyses are detailed in Extended Data Figure 7-1. *p ≤ 0.05. **p ≤ 0.01. ***p ≤ 0.001. F, Biocytin fills of mutant and control subplate neurons at the end of the recording reveal horizontal bipolar morphologies at both ages. Top, E15.5. Bottom, P0. Scale bars, 10 µm.

For E15.5, 14 cells from 7 control brains and 17 cells from 10 mutant brains were examined. For P0-P1, 16 cells from 5 control brains and 17 cells from 5 mutant brains were examined.

Statistical analysis

Statistical tests were performed using GraphPad Prism software version 8.3.0 (538) for Windows, GraphPad Prism Software. All the data were plotted as mean ± SEM. Results are indicated as significant if the p value is p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), or p ≤ 0.0001 (****).

Figure 1D: Statistical test: Unpaired two-tailed Student's t test, ***p = 0.0001; n = 5 for both control and Emx1Cre^11.5::Lhx2^lox/lox brains.

Figure 2G: Ordinary one-way ANOVA with Fisher's LSD test, F = 12.13, ****p < 0.0001. E17.5 Control versus Mutant, *p = 0.0153; E18.5 Control versus Mutant, **p = 0.0015; P0 Control versus Mutant, ****p < 0.0001; P2 Control versus Mutant, ***p = 0.0004; P7 Control versus Mutant, ****p < 0.0001.

Figure 2I: Ordinary one-way ANOVA with Fisher's LSD test, F = 4.171, **p < 0.01. E18.5 Control versus Mutant, not significant, p = 0.3130; P0 Control versus Mutant, not significant, p = 0.9271; P2 Control versus Mutant, *p = 0.0248; P7 Control versus Mutant, **p = 0.0017.

Figure 2J: Ordinary one-way ANOVA with Fisher's LSD test, F = 3.812, **p < 0.01. E17.5 Control versus Mutant, not significant, p = 0.2500; E18.5 Control versus Mutant, not significant, p = 0.0972; P0 Control versus Mutant, not significant, p = 0.5742; P2 Control versus Mutant, not significant, p = 0.3111; P7 Control versus Mutant, ***p = 0.0001.

Figure 3B: Ordinary one-way ANOVA with Holm-Sidak's multiple comparisons test, F = 0.6267, not significant, p = 0.6007; Control versus Emx1Cre^11.5::Lhx2^lox/lox, not significant, p > 0.9999; Control versus NexCre::Lhx2^lox/lox, not significant, p = 0.8440; Emx1Cre^11.5::Lhx2^lox/lox versus NexCre::Lhx2^lox/lox, not significant, p = 0.8440.

Figure 3G: Ordinary one-way ANOVA with Holm-Sidak's multiple comparisons test, F = 67.14, ****p < 0.0001; Control versus Emx1Cre^11.5::Lhx2^lox/lox, ****p < 0.0001; Control versus NexCre::Lhx2^lox/lox, not significant, p = 0.9992; Emx1Cre^11.5::Lhx2^lox/lox versus NexCre::Lhx2^lox/lox, ****p < 0.0001.

Figure 4G: Ordinary one-way ANOVA with Holm-Sidak's multiple comparisons test, F = 67.14, ****p < 0.0001; Control versus Emx1Cre^11.5::Lhx2^lox/lox, ****p < 0.0001; Control versus CreERT2::Lhx2^lox/lox Tam E11.5, ****p < 0.0001; Emx1Cre^11.5::Lhx2^lox/lox versus CreERT2::Lhx2^lox/lox Tam E11.5, not significant, p = 0.9969; CreERT2::Lhx2^lox/lox Tam E11.5 versus CreERT2::Lhx2^lox/lox Tam E12.5, ****p < 0.0001; CreERT2::Lhx2^lox/lox Tam E11.5 versus CreERT2::Lhx2^lox/lox Tam E12.75, ****p < 0.0001; CreERT2::Lhx2^lox/lox Tam E11.5 versus CreERT2::Lhx2^lox/lox Tam E13.5, ****p < 0.0001; CreERT2::Lhx2^lox/lox Tam E12.5 versus CreERT2::Lhx2^lox/lox Tam E12.75, not significant, p = 0.4386; CreERT2::Lhx2^lox/lox Tam E12.5 versus CreERT2::Lhx2^lox/lox Tam E13.5, not significant, p = 0.3098; CreERT2::Lhx2^lox/lox Tam E12.75 versus CreERT2::Lhx2^lox/lox Tam E13.5, not significant, p = 0.9989.

Results

In order to visualize thalamocortical projections, we used a TCA-GFP reporter line in which the expression of membrane-bound enhanced GFP is driven in primary sensory thalamic nuclei under serotonin transporter (SERT) promoter (Mizuno et al., 2014). The TCA-GFP background allowed us to compare the two Lhx2 conditional mutants that were previously reported to have defects in thalamocortical innervation in postnatal brains: a complete loss of somatosensory barrels (Emx1Cre^11.5::Lhx2^lox/lox) (Shetty et al., 2013) or defects in polarized dendritic extension of layer 4 spiny stellate neurons (NexCre::Lhx2^lox/lox) (Wang et al., 2017).

Postnatal day (P) 7 brains were examined for GFP fluorescence indicative of thalamocortical innervation in either tangential slices of the cortex or coronal sections of hemispheres. The barrel map was visible in control brains (Fig. 1A), but was profoundly disrupted in Emx1Cre^11.5::Lhx2^lox/lox brains, in which tangential slices revealed extremely reduced innervation in unstructured clusters that did not form any recognizable barrel field pattern (Fig. 1B, arrowheads). In contrast, barrels were still apparent in NexCre::Lhx2^lox/lox brains (Fig. 1C). Coronal sections of Emx1Cre^11.5::Lhx2^lox/lox brains showed the GFP-expressing axons to be clustered at one location and projecting aberrantly up to the marginal zone of the cortex, instead of layer 4. Quantification of the fluorescence intensity in the somatosensory cortex revealed a significant reduction in Emx1Cre^11.5::Lhx2^lox/lox brains compared with controls (Fig. 1D).

The VB nucleus of the thalamus projects to the barrel cortex, and its neurons require the establishment of stable synapses with their layer 4 targets for survival (Zembrzycki et al., 2013; Greig et al., 2016). Therefore, we examined cytochrome oxidase staining in the P7 thalamus. The sensory thalamus as a whole and the VB nucleus in particular displayed a marked shrinkage in size in Emx1Cre^11.5::Lhx2^lox/lox mutant brains (Fig. 2A). The VB nucleus, identified by the expression of SERT and Cad6, appeared shrunk at P0 as well (Fig. 2B) (Lebrand et al., 1996). To ascertain the time course of this shrinkage, we examined a range of ages and measured the VB area at three rostro-caudal levels. These area quantifications revealed that the VB is smaller in Emx1Cre^11.5::Lhx2^lox/lox mutant brains compared with controls as early as E17.5 (Fig. 2C–G). Other markers that delineate the borders of the VB Chst2, Lhx9, and Prox1 also reveal this shrinkage at P0, whereas Calb2 expression reveals no expansion of higher order thalamic nuclei (Fig. 2E) (Yuge et al., 2011; Frangeul et al., 2016; Kinare et al., 2019). The presence of cleaved Caspase3 indicates apoptosis in the VB at E18.5 and P0 (Fig. 2F). The dLGN appears unaffected by P0, but displays shrinkage by P7, together with the entire dorsal thalamus in mutant brains (Fig. 2I,J).

The VB shrinkage being apparent from E17.5 suggests that this is not because of any disruptions in corticothalamic input, since it is well established that projections from layer 6 do not enter VB until after E18.5 (Auladell et al., 2000; Jacobs et al., 2007). When we examined corticothalamic projections labeled by the Ai9 reporter in Emx1Cre^11.5 control and Lhx2^lox/lox brains, the corticothalamic axons were detected in the reticular nucleus in both conditions and had not entered the dorsal thalamus (Fig. 2H).

The finding that the VB appears to be affected as early as E17.5 in Emx1Cre^11.5::Lhx2^lox/lox mutants motivated the hypothesis that thalamocortical axons may display deficits in their development during the waiting period. Since the TCA-GFP line does not display detectable fluorescence at embryonic stages, we used in utero electroporation to label embryonic VB projections by transfecting a GFP-expressing plasmid into the diencephalon of control and mutant embryos at E11.5, when VB neurons are born. Control, Emx1Cre^11.5::Lhx2^lox/lox, and NexCre::Lhx2^lox/lox brains were examined at E15.5, by which stage the dorsal thalamus expressed robust GFP expression (Fig. 3A). Cortical thickness at E15.5 was similar in control, Emx1Cre^11.5::Lhx2^lox/lox, and NexCre::Lhx2^lox/lox brains (Fig. 3B). At E15.5, the thalamocortical tract coursed below the cortical plate in a tight bundle in control embryos, but in Emx1Cre^11.5::Lhx2^lox/lox brains it displayed an exuberant ingrowth of thalamocortical fibers into the cortical plate at E15.5 (Fig. 3C,D). Furthermore, thalamocortical axons appeared defasciculated, extended several branches prematurely into the cortical plate, and extended aberrantly up to the marginal zone in Emx1Cre^11.5::Lhx2^lox/lox brains (Fig. 3C, arrowheads). Such exuberant branching was not seen in E15.5 NexCre::Lhx2^lox/lox brains, which displayed a well-ordered, bundled thalamocortical tract that coursed below the cortical plate, similar to controls (Fig. 3E,F). The area of the cortical plate in which green fibers were present was scored in each condition and found to be significantly increased in E15.5 Emx1Cre^11.5::Lhx2^lox/lox brains compared with NexCre::Lhx2^lox/lox and control brains (Fig. 3G).

The thalamocortical exuberant ingrowth phenotype was only seen when Lhx2 is disrupted using Emx1Cre. Therefore, we reasoned that the critical requirement for this transcription factor must be in progenitors, not postmitotic neurons. Since essentially all cortical layers are produced sequentially from progenitors that express Emx1Cre, we sought to narrow down the stage at which LHX2 function is relevant for the proper guidance of thalamocortical axons, using tamoxifen-inducible CreERT2. Embryos of the genotype CreERT2::Lhx2^lox/lox were electroporated with a GFP-expressing plasmid at E11.5 in the diencephalon to label the VB thalamocortical fibers, tamoxifen was administered at different stages to disrupt Lhx2 (Fig. 4C–F), and the embryos were harvested at E15.5. Whereas tamoxifen administration at E11.5 closely recapitulated the defasciculation and exuberant ingrowth of thalamocortical axons seen in Emx1Cre^11.5::Lhx2^lox/lox embryos (Fig. 4B,C,G), tamoxifen administration at E12.75 and at E13.5 resulted in a phenotype similar to control brains (Fig. 4A,E,F). Tamoxifen administration at E12.5 produced a phenotype mostly resembling that of control brains, except for a few fibers at dorsomedial levels that prematurely entered the cortex (Fig. 4D). Birthdating studies have established that subplate neuron production peaks at E11.5, but medially, the subplate also contains neurons born at E12.5 (Price et al., 1997; Hoerder-Suabedisen and Molnár, 2013). Therefore, loss of Lhx2 at the time of birth of subplate neurons correlates with the thalamocortical overshooting phenotype. LHX2 has multiple roles in this system: Lhx2 mutant ventral telencephalic explants do not support normal topographic sorting of wild-type thalamocortical axons (Lakhina et al., 2007); loss of Lhx2 using thalamus-specific driver Gbx2-CreERT2 (tamoxifen at E10.5) disrupts the topographic sorting of thalamocortical axons originating from specific thalamic nuclei, and results in occasional fibers overshooting the subplate (Marcos-Mondéjar et al., 2012). These results are distinct from our findings using the cortex-specific Emx1Cre^11.5, which indicate that LHX2 is necessary in early (E11.5) cortical progenitors at the time the subplate is being generated in order for normal thalamocortical axon projection to occur. In summary, our results suggest that daughter subplate neurons may develop aberrantly when LHX2 is absent in their mother cells, leading to abnormal subplate properties.

The phenotype of premature thalamocortical afferent ingrowth into the cortical plate with fibers reaching the marginal zone in the Emx1Cre^11.5::Lhx2^lox/lox mutant was reminiscent of the thalamocortical projection phenotype described in reeler mutants, in which the subplate is mislocalized above the cortical plate (Molnár et al., 1998). Therefore, we examined whether the cortical subplate was specified and present in its correct location in mutant brains. We first examined the expression of CTGF (Connective Tissue Growth Factor) and CPLX3 (Complexin 3), established markers of the postnatal subplate (Hoerder-Suabedissen and Molnár et al., 2013). At P7, control brains displayed a tightly packed subplate expressing both markers (Fig. 5A). In Emx1Cre^11.5::Lhx2^lox/lox brains, the subplate also expressed both markers (Fig. 5B). Furthermore, subplate cells were positioned normally, below the cortical plate, although they were not as tightly packed as in the controls. NexCre::Lhx2^lox/lox brains also displayed a tightly packed subplate (Fig. 5C). In NexCre::Ai9 brains, the Ai9 reporter was coexpressed with CPLX3, indicating that NexCre is indeed active in subplate cells (Fig. 5D). Neither CTGF nor CPLX3 are expressed in the embryonic subplate, so we used Nurr1 (Nr4a2; Nuclear Receptor Subfamily 4 Group A Member 2), which labels the subplate at E15.5 (Fig. 5E,F). Nurr1 expression reveals a less compact subplate in E15.5 Emx1Cre^11.5::Lhx2^lox/lox brains than in controls; nonetheless, it is correctly positioned below the cortical plate when the thalamocortical axons arrive. We examined serial sections for Lhx2 expression using a probe designed specifically against the floxed exon and confirmed that Lhx2 expression was indeed lost from the entire cortex as expected for Emx1Cre-driven recombination (Fig. 5E,F). In summary, the premature thalamocortical ingrowth phenotype, with axons extending up to the marginal zone, cannot be explained by a mispositioning of subplate neurons in Emx1Cre^11.5::Lhx2^lox/lox brains.

Although the position of the subplate in Emx1Cre^11.5::Lhx2^lox/lox brains is not altered, it may be molecularly disrupted in a manner that is critical for thalamocortical axon guidance. To get an overall picture of the transcriptomic changes within subplate neurons on loss of LHX2, we performed RNA-seq on micro-dissected subplate tissue. We hypothesized that loss of LHX2 from the progenitor stage versus the postmitotic stage may have distinct regulatory consequences on the transcriptome of subplate neurons, and these differences may underlie the differences in the thalamocortical axon phenotype in Emx1Cre^11.5::Lhx2^lox/lox and NexCre::Lhx2^lox/lox brains. Therefore, we compared the wild-type subplate transcriptome with that of the Emx1Cre^11.5::Lhx2^lox/lox and NexCre::Lhx2^lox/lox subplate. We selected P0 as the stage of analysis because the last wave of cortical neurons has migrated past the subplate by this stage and would not be present in the dissected sample. The lowermost part of the cortical plate was visually identified in slices and micro-dissected, thus enriching the subplate population. The micro-dissection was confirmed by processing some slides for ISH for subplate marker CTGF (Fig. 6A). As expected, the transcriptomes of the Emx1Cre^11.5::Lhx2^lox/lox and NexCre::Lhx2^lox/lox subplate were each dysregulated compared with that of their corresponding littermate controls. We examined the genes that displayed a fold change >1.5 in each condition and found that 357 genes were differentially expressed in the Emx1Cre^11.5::Lhx2^lox/lox subplate and 196 in the NexCre::Lhx2^lox/lox subplate (Fig. 6B). However, when we compared these two differentially expressed datasets, only 29 genes were in common between the two (Fig. 6B). This indicated that, although both Cre drivers resulted in a loss of LHX2 from the subplate, using a driver that acts in progenitors versus postmitotic cells has vastly different consequences on the transcriptome of the resulting subplate neurons.

Figure 6-1

Gene Ontology Analysis of Differentially Regulated Genes (from Fig. 6C) represented in a tabular format under the categories: Upregulated Genes and Downregulated Genes Download Figure 6-1, XLSX file ^{(10.1KB, xlsx)} .

Figure 6-2

Table of RNA-seq log₂fold change values (as mean± SEM) of representative differentially expressed genes (from Fig. 6D). Download Figure 6-2, XLSX file ^{(10.6KB, xlsx)} .

Figure 6-3

Bar plots representing log₂fold change values (from RNA-Seq) of selected genes in different functional categories (from Fig. 6D). Download Figure 6-3, TIF file ^{(56.1MB, tif)} .

We further analyzed the 328 differentially expressed genes that were specific to the Emx1Cre^11.5::Lhx2^lox/lox dataset. GO analysis revealed multiple developmentally significant gene clusters (Fig. 6C). We categorized these 328 dysregulated genes according to their known functions and found they encode molecules that mediate axon guidance, such as EphA4, Nrp2, PlxnC1, and Sema7a; cell signaling components; transcription factors; membrane-associated molecules, including protocadherin Pcdh10; as well as ion channels/transporters/synaptic molecules (Fig. 6D; Extended Data Fig. 6-3). These putative LHX2 targets suggest mechanisms by which the subplate could regulate the waiting of thalamocortical afferents. Several genes previously reported to be enriched in the subplate, such as Hpca, Syn2, Prdm8, Cplx3, Kcnab1, Pde1a, Sema7a, Kctd4, Adcyap1, and Pou3f1 (Hoerder-Suabedissen et al., 2009; Osheroff and Hatten, 2009; Oeschger et al., 2012), are dysregulated in the Emx1Cre^11.5::Lhx2^lox/lox subplate. We used ISH to validate the RNA-seq results for some of these genes (Fig. 6E,F).

The RNA-seq analysis indicated that several genes involved in the bioelectrical properties of neurons were dysregulated in the Emx1Cre^11.5::Lhx2^lox/lox subplate, for example, potassium channel subunits Kcnab1, Kcnh3, Kcnj10, and Kcnj16; a ligand-gated chloride channel subunit Glra2, an ionotropic glutamate receptor Grid2; and solute transporters Slc1a2, Slc1a3, and Slc6a1. This suggested to us that the electrophysiological properties of the subplate may be perturbed, and this may be relevant in the regulation of thalamocortical axon innervation of the cortex. A recent study demonstrates that subplate neurons are not only electrically active as early as E16, but they also form transient glutamatergic synapses on migrating multipolar neurons that will populate the cortical plate, which is critical to their multipolar-to-bipolar transition and subsequent migration (Ohtaka-Maruyama et al., 2018). These observations motivated an examination of whether the electrophysiological properties of subplate neurons were impaired on Emx1Cre-driven disruption of Lhx2. We prepared live slices from control and Emx1Cre^11.5::Lhx2^lox/lox littermate embryos at stages at the beginning and the end of the waiting period, E15.5 and P0/P1. Subplate neurons were identified by their location between the cell-dense cortical plate with radially oriented neurons and the cell-sparse white matter, and their typical horizontal bipolar morphologies (Hirsch and Luhmann, 2008). First, we examined the intrinsic electrical properties of control and mutant subplate cells. Most control E15.5 subplate cells displayed a characteristic spontaneous firing pattern with only 36% being silent. In contrast, a larger fraction of mutant subplate neurons were silent (88%; Fig. 7A). The resting membrane potentials (RMPs) of mutant and control subplate neurons were similar at E15.5. While control neurons became hyperpolarized over time, mutant neurons failed to do so; by birth, their RMP was more depolarized compared with controls (Fig. 7C). We estimated input resistance by measuring membrane potential changes in response to a series of current steps (−10 pA to 10 pA, steps of 5 pA) and computing the slope (Fig. 7E,F). While the input resistance (IR) was similar in control and mutant subplate neurons at E15.5, control IR decreased by birth whereas mutant subplate cells continued to display a higher IR (Fig. 7E,F). Since a depolarized RMP and high IR are features associated with immature neurons (Cepeda et al., 2007; Valiullina et al., 2016), the Lhx2 mutant subplate appeared to be electrically immature at birth.

Figure 7-1

Individual values for measurements of electrophysiological properties of subplate neurons from control and Emx1Cre^11.5::Lhx2^lox/lox brains subplate neurons from Figures 7 and 8. Download Figure 7-1, DOCX file ^{(13KB, docx)} .

Excitability, as assessed by measuring responses evoked by stepwise current injections, was reduced in Emx1Cre^11.5::Lhx2^lox/lox subplate neurons compared with controls. At E15.5, the mutant subplate neurons were mostly limited to single spikes for input currents from 5 to 60 pA, whereas wild-type cells displayed a regular spiking pattern (Fig. 8A). At P0, although mutant subplate neurons displayed an improved spike frequency, it was lower than that of control neurons at input currents >35 pA (Fig. 8B). At both ages, the mutants displayed a higher proportion of single spiking neurons compared with regular spiking neurons (81% vs 21% at E15.5 and 33% vs none at P0; Fig. 8C). Mutant subplate neurons also displayed consistently lower spike amplitudes than controls (Fig. 8D). The current required to elicit spiking (rheobase) was higher for E15.5 mutant cells (Fig. 8E). After recording, biocytin-fillings revealed typical horizontal bipolar morphologies of subplate neurons in mutants as well as controls, at each age (Fig. 8F). In summary, the majority of subplate neurons in Emx1Cre^11.5::Lhx2^lox/lox brains were electrically silent from E15.5, when the thalamocortical afferents first arrive. Their intrinsic bioelectrical properties failed to mature from E15 to P0, and their evoked responses were deficient, remaining typical of immature neurons at birth.

In summary, our results revealed dysregulation of diverse set of mechanisms in Emx1Cre^11.5::Lhx2^lox/lox subplate neurons that could underlie the phenotype of premature exuberant overgrowth by the thalamocortical axons into the cortical plate and an eventual attrition of thalamocortical innervation of the postnatal cortex.

Discussion

The mechanisms that produce normally functioning subplate neurons are poorly understood. Our study offers insights into this process by highlighting an essential transcriptional regulatory function of LHX2 that operates in cortical progenitors at E11.5, when the subplate is born.

In order to identify the precise stage of progenitor that requires LHX2 function, and also to discriminate between progenitors versus postmitotic neurons, we used different Cre lines and compared the resulting phenotype with that seen in Emx1Cre^11.5::Lhx2^lox/lox brains. NexCre-mediated Lhx2 recombination does not recapitulate the Emx1Cre-mediated phenotype, since the thalamocortical axons wait at the subplate without showing aberrant innervation at E15.5. Therefore, we have uncovered a novel LHX2 function that selectively operates in progenitors and not in postmitotic neurons. Furthermore, when Lhx2 recombination is induced via CreER, only tamoxifen administration at E11.5, but not E12.5- E13.5, recapitulates the Emx1Cre-mediated phenotype. Therefore, the previously reported phenotype of a reduced Tbr1-expressing layer 6, and an expanded Ctip2-expressing layer 5 in Emx1Cre^11.5::Lhx2^lox/lox brains (Muralidharan et al., 2017) may not be relevant to the premature ingrowth of thalamocortical axons, since these cells are born primarily at E12.5 and E13.5, respectively. We cannot, however, exclude a role of an E11.5 born cohort of layer 6 neurons that may participate in this phenotype.

Loss of LHX2 from E11.5 progenitors has profound consequences on the transcriptome and the bioelectrical properties of the subplate neurons which are born at E11.5. Since the subplate is the primary target of the thalamocortical pathway when it enters the dorsal telencephalon, we propose that loss of LHX2 in E11.5 progenitors results in a subplate with deficient molecular and electrophysiological properties such that the thalamocortical axons appear to overshoot it and enter the cortical plate prematurely. Alternatively, the subplate and E11.5-born layer 6 neurons may gain aberrant permissive properties on loss of LHX2 function, resulting in exuberant thalamocortical afferent ingrowth. Examples of dysregulated genes in the subplate that could affect thalamocortical ingrowth are molecules involved in axon guidance, membrane associated functions, such as cell adhesion, transcriptional control, and cell signaling, listed in Figure 6.

In this manner, molecular players acting downstream of LHX2 may control the temporary pausing of axonal growth and restarting subsequent reinnervation.

Interaction of thalamocortical and corticothalamic axons in the ventral telencephalon

The subplate plays a guidance role for the entry of the thalamocortical pathway into the dorsal telencephalon and for its initial trajectory below the cortical plate. In the Tbr1 and Gli3 mutants, the subplate appears to be absent altogether. The thalamocortical axons extend toward the pallial-subpallial boundary but fail to cross it, suggesting that interactions between subplate projections in the ventral telencephalon are required for thalamocortical afferents to enter the dorsal telencephalon (Molnár and Blakemore, 1995; Hevner et al., 2001; Bedogni et al., 2010; Magnani et al., 2013). Thalamocortical afferents target subplate neurons even when they are mislocalized. In the reeler mutant, cortical layering is inverted (Caviness and Sidman, 1973), and the subplate, identified by the expression of CTGF and CPLX3, is located above the cortical plate (a “superplate”). Thalamic axons extend through the cortical plate and reach the superplate and then turn downward to innervate layer 4 (Molnár et al,1998; Tissir and Goffinet, 2003). In the p35 mutant, subplate neurons are located in the middle of an inverted cortical plate, in which earlier-born deep layer neurons are positioned above the subplate and later-born superficial layer neurons are positioned below the subplate. The thalamic axons enter the cortex obliquely toward the mispositioned subplate and eventually grow toward the marginal zone (Kwon and Tsai, 1998; Rakić et al., 2006). Together, these studies demonstrate that thalamocortical axons cannot cross the pallial-subpallial boundary without the presence of the subplate and will first target the subplate, even when it is mislocalized, underscoring its importance as a primary target.

In contrast to the mutants described above, in which subplate neurons are either absent or mislocalized, the subplate is correctly positioned below the cortical plate in Emx1Cre^11.5::Lhx2^lox/lox brains. Corticothalamic projections reach the thalamus similar to wild-type brains, and thalamocortical fibers also navigate normally through the ventral telencephalon and cross the pallial-subpallial boundary. This suggests that the role of the subplate in guiding thalamocortical fibers into the dorsal telencephalon is unaffected. This mutant, therefore, serves as an excellent model to analyze mechanisms that control key subplate properties beyond its guidance function in terms of entry of thalamocortical axons into the dorsal telencephalon.

VB patterning and cortical area patterning

The defects we report in thalamocortical projections are not likely to be specific to the somatosensory modality, since the entire sensory cortex displays impoverished innervation from the thalamus in maturity (Shetty et al., 2013). In this study, we focused on the somatosensory thalamic nucleus and its projections to the cortex, but the role of the subplate is likely to be relevant to other thalamocortical projections as well.

Previous studies have shown that the VB is responsive to its connectivity with its target cortical area. When the primary somatosensory area (S1) is shifted to an extreme caudal location upon cortex-specific loss of COUP-TF1, the thalamocortical innervation targets the ectopic S1 and barrels are formed there (Armentano et al., 2007). When the S1 area is reduced as a consequence of cortex-specific loss of Pax6, or repatterned to a motor identity on cortex-specific loss of Ctip1, the VB undergoes selective repatterning and/or apoptosis (Zembrzycki et al., 2013; Greig et al., 2016). Similar mechanisms could explain why the VB continues to shrink postnatally in the Emx1Cre:: Lhx2^lox/lox mutant (Fig. 2). Our study extends this target-dependent survival model further, in that VB shrinkage appears to begin embryonically, as a result of the thalamocortical axons encountering a subplate born from LHX2-deficient progenitors.

Progenitor-specific mechanisms regulate the properties of the postmitotic subplate

The subplate is described as the most mature neuronal population in the cortex since it is born first and achieves properties associated with mature neurons well ahead of the other cortical neurons (for review, see Luhmann et al., 2009). Synaptic transmission from subplate neurons plays an instructive role in multipolar to bipolar transition of migrating neurons in the developing cortex as early as E16 (Ohtaka-Maruyama et al., 2018). Thalamocortical axons are also known to make synaptic contacts with the subplate, which serves as a critical and necessary primary target of the thalamocortical pathway (Rakić, 1983; Ghosh et al., 1990; Ghosh and Shatz, 1992; Herrmann et al., 1994; Kanold et al., 2003; Kanold and Shatz, 2006; Tolner et al., 2012). Subplate neurons play distinct roles at different developmental stages; however, very little is known about the molecular mechanisms that regulate the properties of this population. These neurons are morphologically heterogeneous, ranging from inverted pyramidal neuron-like, to horizontal monotufted, bitufted, or multipolar with extensive arborization (Hanganu et al., 2002). Gene expression profiling of the subplate at different ages indicates tight temporal regulation of genes expressed specifically in the subplate (Hoerder-Suabedissen et al., 2009; Osheroff and Hatten, 2009; Oeschger et al., 2012).

The temporally dynamic and heterogeneous nature of the subplate is poorly understood. Our study uncovers an additional feature, that severe electrophysiological deficits manifest in postmitotic subplate neurons that are born from LHX2-deficient cortical progenitors from E15.5, when thalamocortical axons contact them. Furthermore, the transcriptome of the subplate arising from Lhx2-null progenitors shows gene dysregulation that has little overlap with that seen when LHX2 is lost postmitotically in those very cells. It is also striking that loss of LHX2 results in thalamocortical axons displaying inappropriate exuberant overgrowth only in the former but not the latter case. Recent work has shown that transcriptional priming of daughter neurons by their mother cells plays a critical role in setting the developmental trajectories during differentiation (Zahr et al., 2018; Telley et al., 2019). Our study motivates a broader investigation of transcriptional and epigenetic controls in progenitors that control the properties of the resulting postmitotic progeny, eventually regulating how neurons participate in cortical circuitry.