Introduction

Axonal connectivity is a critical foundation for CNS function. During CNS development, axons of projection neurons extend from cell bodies (somata), undergo long-distance growth, fasciculate, defasciculate, and collateralize, forming specific circuits in distant target areas.¹ Projection neuron axons navigate through a constantly changing in vivo environment. Correct targeting of axons and their collaterals to form functional circuitry is a complex, multistep, multi-component process integrating often interdependent mechanisms: axon growth; attractive and repulsive contact-mediated and diffusible axon guidance at major and minor trajectory choice points; rate of progression across targets while sensing, testing, and rejecting a myriad of alternative but inappropriate targets; tissue and ECM penetration; and often fasciculation within large axon tracts such as the corpus callosum either with sister axons with the same overall projection trajectory or/and with complementary axons of neurons with the opposite projection trajectory.^2-4 These steps eventually guide axons to desired targets for further circuit refinement. In the cerebral cortex, axons originating from projection neurons with distinct subtype specificity and areal identity often exhibit distinct responses to the same environment with its mixture of cues regulating growth, attraction, repulsion, and target specificity. Thus, axon guidance is controlled by an interplay between intrinsic neuronal subtype-specific molecular programs, and extrinsic environmental cues.

Corticofugal projection neurons (CFuPN) are a broad class of cortical output neurons that extend their axons away from the cortex, through the internal capsule in the striatum, to targets in subcortical (e.g., thalamus) and subcerebral (e.g., brainstem and spinal cord) regions. Two dominant CFuPN subtypes are corticothalamic projection neurons (CThPN), which reside primarily in layer VI and project axons to specific nuclei of the thalamus, and subcerebral projection neurons (SCPN), which reside in layer V, and extend primary axons to targets caudal to the cerebrum.^5,6 SCPN include the major and highly clinically relevant subset of corticospinal neurons (CSN).^7-10 After exiting the cortex, CFuPN axons enter the striatum, fasciculating in the internal capsule via interaction with striatal medium spiny neurons (MSN; also known as striatal projection neurons). Caudal to the striatum, CFuPN axons diverge into thalamic-projecting CThPN axons and subcerebral-projecting SCPN axons projecting to brainstem and/or spinal cord. CThPN axons reciprocally interact with, and are thus guided by, thalamocortical axons to innervate thalamus.^11-13 A recent study identified that SCPN axons exhibit convergent projection with pioneering MSN axons in the forebrain.¹⁴ SCPN axons traverse a substantial further distance caudally, forming the corticospinal tract (CST), to reach the brainstem and spinal cord.¹⁵ Very recent work regarding developing inter-hemispheric callosal axons in the corpus callosum reveals that axonal adhesion molecule levels instruct topographic axon guidance and targeting.² These examples highlight the importance of coordinated regulation between intrinsic and extrinsic molecular elements that together guide axons of distinct neuron populations and subtypes, but little is known about how this molecular coordination is achieved.

The zinc finger transcription factor Bcl11b/Ctip2 is expressed by SCPN at high level postmitotically, and it functions centrally in controlling their differentiation.¹⁶ Many other molecular-genetic controls over neocortical projection neuron differentiation operate at least in part by regulating Bcl11b expression either positively or negatively.^17-20 Bcl11b regulates transcription via both DNA binding and nucleosome remodeling.^21,22 In constitutive Bcl11b knockout (Bcl11b^-/-) mice, SCPN are born, express SCPN upstream control genes such as Fezf2, and migrate to layer V, but they exhibit remarkable defects in axon pathfinding, fasciculation, and outgrowth.¹⁶ SCPN axons normally fasciculate in the internal capsule, the myelinated forebrain tract through the striatum containing many fascicles of axons that project from cortex to subcortical targets. In Bcl11b^-/- mice, descending axons become misrouted and defasciculated as they pass through the forebrain. While SCPN axons efficiently project to the spinal cord in wild-type mice, SCPN in Bcl11b^-/- mice extend axons to ectopic targets.¹⁶ These mutant axons are clearly dysmorphic, and often possess bulbous structures suggestive of dysfunctional growth cones (GCs).²³ Most critically, SCPN axons in Bcl11b^-/- mice do not successfully reach the spinal cord; the projections that exit the forebrain mostly terminate in the midbrain, with only rare axons reaching pons, and none reaching the pyramidal decussation.¹⁶ These findings strongly indicate SCPN-autonomous Bcl11b functions in their axon development.

Intriguingly, Bcl11b is also expressed at high levels by MSN in the striatum, and Bcl11b^-/- MSN lose their characteristic patch-matrix organization and mis-express key axon guidance molecules.²⁴ Because MSN surround SCPN axons as they penetrate through the internal capsule, potentially providing important guidance cues to SCPN, and because the pallial-subpallial boundary (the embryonic border between cortex and striatum) is an important decision point for axons traveling through the internal capsule,²⁵ we hypothesized that Bcl11b might also contribute to SCPN development non-SCPN-autonomously via functions in MSN.

While a number of transcriptional controls regulating cortical projection neuron subtype specification and areal development have been identified, downstream subcellular mechanisms and molecular machinery that implement subtype-specific circuitry remain largely unknown. Importantly, previous studies provide evidence that complex subcellular molecular networks are axon GC-localized, including axon guidance ligands, receptors, and adhesion molecules,¹ and that tightly regulated local translation of select mRNA transcripts directly controls GC pathfinding.^26-28 Recent in vivo work from our lab identifies GC-specific enrichment of hundreds of transcripts and proteins compared with parent somata of developing interhemispheric callosal projection neurons (CPN) and CThPN,^2,29-32 and demonstrates that subcellular GC molecular machinery directly implements subtype-specific developmental programs to build circuitry, connect to appropriate targets, and select correct synaptic partners. However, subcellular GC molecular machinery underlying SCPN circuit formation remains unknown.

Here, we identify that Bcl11b contributes to SCPN axon development both SCPN-autonomously and SCPN-non-autonomously (via functions in MSN) by employing cortex-specific (Emx1-Cre) and/or striatum-specific (Gsx2-Cre) Bcl11b conditional deletion. The results indicate that Bcl11b orchestrates SCPN axon outgrowth, fasciculation, and pathfinding through complementary functions in two distinct but physically interacting neuron populations (SCPN and MSN). Further, to directly investigate subcellular molecular mechanisms by which Bcl11b cell-intrinsically controls SCPN axon projection targeting, we purified SCPN axonal GCs and somata in vivo from developing wild-type and cortex-specific Bcl11b knockout mice, and investigated their local subcellular molecular machinery. We identify that Cadherin 13 (Cdh13) is a key downstream target of Bcl11b in SCPN for their axon outgrowth. Cdh13 is expressed by SCPN under control of Bcl11b; is subcellularly localized along SCPN axons and on SCPN GC surfaces in vivo; and promotes SCPN axon extension. Together, these results demonstrate that Bcll1b controls multiple aspects of precise SCPN axon development by coordinated intrinsic and extrinsic mechanisms.

Results

Emx1-IRES-Cre and Gsx2-iCre efficiently delete Bcl11b from neocortical neurons and MSN, respectively

In developing forebrain, Bcl11b is expressed at high level by postmitotic SCPN in dorsal cortex, and by MSN in striatum (Figure 1A).^16,24 To delete Bcl11b specifically from glutamatergic cortical projection neurons, we crossed a loxP-flanked (floxed) Bcl11b mouse line³³ with an Emx1^IRES-Cre knock-in mouse line.³⁴ These Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice are born in expected Mendelian ratios, and survive until adulthood, unlike Bcl11b^-/- mice, which die at postnatal day (P)0.^16,35 Bcl11b^flox/flox;Emx1^IRES-Cre/+ tissue does not contain detectable levels of Bcl11b in layer V pyramidal neurons, while Bcl11b expression by MSN is preserved (Figure 1B).

Figure 1.

To directly examine MSN-specific Bcl11b functions in SCPN axon development, we crossed the Bcl11b^flox/flox line with a Gsx2-iCre transgenic mouse line (Gsx2 was formerly called Gsh2).^36,37 When Gsx2-iCre is crossed with a Rosa26^LSL-LacZ reporter line,³⁸ no Bcl11b-positive pyramidal neurons in cortex express β-galactosidase (Supplementary Figure 1A). Interneurons derived from medial ganglionic eminence, identified by immunocytochemistry for parvalbumin, somatostatin, and neuropeptide Y,³⁹ are the only β-galactosidase-positive neurons in cortex (Supplementary Figure 1B–D). Bcl11b^flox/flox;Gsx2-iCre mice are born in expected Mendelian ratios, and survive until adulthood. Bcl11b expression is normal in Bcl11b^flox/flox;Gsx2-iCre cortex, but is abolished in striatum (Figure 1C). Expression of the MSN marker DARPP-32 is reduced in Bcl11b^flox/flox;Gsx2-iCre striatum, as reported for Bcl11b^-/- striatum (Supplementary Figure 1E).²⁴

When Bcl11b^flox/flox mice are crossed with mice carrying both Emx1^IRES-Cre and Gsx2-iCre alleles, the vast majority of Bcl11b expression in the forebrain is abolished (Figure 1D), indicating that Bcl11b is predominantly expressed by Emx1-expressing and Gsx2-expressing cell lineages. We next used these single- and double-conditional Bcl11b mutants to investigate Bcl11b functions in SCPN axon development.

Bcl11b deletion from MSN causes abnormal CFuPN axon fasciculation in the internal capsule

To investigate abnormalities in axon tract morphology and cellular organization within the forebrain due to loss of Bcl11b expression by SCPN and/or MSN, we performed Nissl stains (Figure 2A–E) and GAP43 immunocytochemistry (Figure 2F–J; Supplementary Figure 1E) on P0 forebrain tissue sections from wild-type (Bcl11b^flox/flox), Bcl11b^flox/flox;Emx1^IRES-Cre/+, Bcl11b^flox/flox;Gsx2-iCre, Bcl11b^flox/flox;Emx1^IRES-Cre/+;Gsx2-iCre, and Bcl11b^-/- mice.

Figure 2.

Cortical projection neurons are located within the cortical plate, and their efferent axons project through the cortical white matter. Wild-type cortex exhibits clear separation of the cortical plate and white matter (Figure 2A’’’). While this cortical structure is unchanged when Bcl11b is deleted only from striatum (Figure 2C’’’), the separation becomes less clear when Bcl11b is deleted from cortex (Figure 2B’’’, D’’’, E’’’), indicating cell-autonomous Bcl11b function in cortical projection neurons for their early axon development within cortex, and for their initial organization toward the internal capsule.

In wild-type mice, CFuPN axons then penetrate into striatum, and fasciculate to form the internal capsule (Figure 2A’, A”, F; Supplementary Figure 1E). When Bcl11b is deleted from cortex, those CFuPN axon fascicles that succeed in entering the striatal internal capsule appear normal (Figure 2B’, G; Supplementary Figure 1E), indicating that Bcl11b expression in cortex is largely dispensable for CFuPN axons to fasciculate in striatum. However, organization and entry into the internal capsule is affected: an ectopic external capsule emerges at the lateral border of striatum (Figure 2B”), indicating that cortical Bcl11b is necessary for CFuPN axons to optimally penetrate into the striatum. This ectopic external capsule emerges when Bcl11b is deleted from cortex (Figure 2B”, D”, E”), but not when it is deleted from striatum (Figure 2C”). Overall, Bcl11b expressed in cortex is dispensable for fasciculation of CFuPN axons that successfully enter striatum.

We previously reported that cellular architecture in striatum is substantially disrupted in Bcl11b^-/- mice.²⁴ Indeed, deletion of Bcl11b specifically from striatum leads to severely disrupted patch-matrix organization (Figure 2C’–E’). We find disorganized internal capsule structure (Figure 2C’–E’), and impeded CFuPN fasiculation (Figure 2H–J; Supplementary Figure 1E). These results demonstrate that Bcl11b expression in striatum is required for CFuPN axon fasciculation in the internal capsule, indicating that Bcl11b expressed by MSN importantly regulates CFuPN axon fasciculation non-cell-autonomously.

Bcl11b deletion from SCPN causes reduced SCPN axon outgrowth in brainstem, and Bcl11b deletion from MSN causes ectopic entry of SCPN axons into subthalamic nucleus and dorsocaudal midbrain

To investigate how Bcl11b deletion from distinct neuronal populations affects SCPN axon development caudal to striatum, we anterogradely labeled SCPN axons using a genetic reporter. We crossed Bcl11b wild-type or mutant mice with Emx1^IRES-Flpo;Rosa26^{frt-LacZ-frt-EGFP} mice^8,40 to broadly label cortical projection neurons by EGFP, independent of Cre expression status (Figure 3A). At P1, EGFP expression is indistinguishable between wild-type and any of the Bcl11b mutant cortices (Supplementary Figure 2A–D). EGFP expression is mosaic with a dorsal-to-ventral gradient, reflecting a graded expression pattern of Emx1.⁴¹ CFuPN axons form a large bundle of axons after exiting striatum, then a subset (CThPN axons) innervate thalamus, while the rest (SCPN axons) extend caudal to thalamus (Supplementary Figure 2E–H). Despite low-level Bcl11b expression by CThPN,¹⁶ innervation of CThPN axons into thalamus appears essentially normal in all mutant mice compared to wild-type mice (Supplementary Figure 2E–H). We next focused on and deeply investigated SCPN axon trajectory.

Figure 3.

In wild-type mice, SCPN axons fasciculate tightly and form the cerebral peduncle in the midbrain, projecting laterally to the subthalamic nucleus (STN; STN neurons also express Bcl11b), and extending caudally (Figure 3B). These SCPN axons eventually enter the pons (Figure 3F, F’), and no SCPN axons project toward dorsocaudal midbrain (Figure 3F”). SCPN axons continue extending caudally to reach the medulla (Figure 3J; Supplementary Figure 2I); many form the pyramidal decussation (Supplementary Figure 2M).

SCPN axons in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice fasciculate normally in the cerebral peduncle and project laterally to STN (Figure 3C). No SCPN axons target dorsocaudal midbrain, but very few SCPN axons reach pons and medulla, resulting in a substantially underdeveloped pyramidal decussation (Figure 3G, G’, G”, K; Supplementary Figure 2J, N; see also Supplementary Figure 5). These results indicate that Bcl11b expressed by SCPN is required cell-autonomously for SCPN axon outgrowth in midbrain and hindbrain (and thus further caudally).

We next find that SCPN axon targeting is regulated non-cell-autonomously by Bcl11b expressed by MSN. While SCPN axons in Bcl11b^flox/flox;Gsx2-iCre mice exhibit fasciculation defects within the internal capsule (Figure 2C, C’, C”), the majority of these axons ultimately project through the cerebral peduncle and pons (Figure 3D, H, L; Supplementary Figure 2K, O). However, a small subset of SCPN axons penetrate through the STN, even though Bcl11b expression in the STN remains intact (Figure 3D). Further, a subset of SCPN axons mistargets dorsocaudal midbrain (Figure 3H”), presumably leading to a corresponding reduction of SCPN axons reaching the pons (Figure 3H’, L; Supplementary Figure 2K, O). Quite intriguingly, SCPN axon mistargeting toward dorsocaudal midbrain is not always symmetric between hemispheres: some mice show unilateral mistargeting (1/5 mice), while others show bilateral (4/5 mice), and the severity of bilateral mistargeting is often variable between hemispheres.

In Bcl11b^flox/flox;Emx1^IRES-Cre/+;Gsx2-iCre mice (Figure 3E, I, M; Supplementary Figure 2L, P), defects in SCPN axon pathfinding and outgrowth recapitulate those observed in Bcl11b^flox/flox;Emx1^IRES-Cre/+ and Bcl11b^flox/flox;Gsx2-iCre mice, resulting in even fewer SCPN axons extending in the brainstem (Figure 3M; Supplementary Figure 2L, P). These results further reinforce our findings that Bcl11b expressed by SCPN themselves cell-autonomously controls SCPN axon outgrowth, and that Bcl11b expressed by MSN is non-cell-autonomously required for optimal SCPN axon pathfinding.

We further investigated SCPN axon development using an orthogonal anterograde labeling approach: focal DiI injection. We placed DiI crystals in sensorimotor cortex of fixed P0 brain tissue from wild-type, Bcl11b^flox/flox;Emx1^IRES-Cre/+, Bcl11b^flox/flox;Gsx2-iCre, and Bcl11b^flox/flox;Emx1^IRES-Cre/+;Gsx2-iCre mice, then visualized anterogradely spread DiI along axons in sagittal sections. We focused on DiI labeled axons projecting more caudal than thalamus, which are unequivocally SCPN axons. We observed essentially identical defects in SCPN axon development (Supplementary Figure 2Q–T) that are identified by our genetic labeling approach (Figure 3A–L), although mistargeting of SCPN axons into STN was not readily noticeable due to sectioning orientation. We additionally identified that Bcl11b expressed by SCPN is required for corticotectal branching of SCPN axons toward the optic tectum in the midbrain, and for fasciculation of SCPN axons in the cerebral peduncle (Supplementary Figure 2Q’–T’). DiI stains cell membranes brightly, enabling identification of this SCPN axon defasciculation in the cerebral peduncle.

We next investigated SCPN mistargeting after Bcl11b deletion from MSN. Because MSN axons pioneer SCPN axons,¹⁴ we hypothesized that Bcl11b deletion from MSN might perturb their pioneering functions for SCPN axons (while maintaining normal MSN projections), or might mistarget MSN axons, leading to secondary SCPN mistargeting. To distinguish these possibilities, we simultaneously visualized SCPN axons (Emx1^IRES-Flpo;Rosa26^{frt-LacZ-frt-EGFP}) and MSN axons (Gsx2-iCre;Rosa26^LSL-tdTomato) (Figure 3N). In control mice (Bcl11b^flox/+;Gsx2-iCre;Emx1^IRES-Flpo/+;Rosa26^{frt-LacZ-frt-EGFP/LSL-tdTomato}), tdTomato-positive MSN axons and EGFP-positive SCPN axons traverse laterally to STN, and no tdTomato- or EGFP-positive axons penetrate inside STN (Figure 3O). In Gsx2-iCre conditional knockout mice (Bcl11b^flox/flox;Gsx2-iCre;Emx1^IRES-Flpo/+; Rosa26^{frt-LacZ-frt-EGFP/LSL-tdTomato}), a small fraction of axons aberrantly pass through STN— all are positive for both tdTomato and EGFP (Figure 3P). Additionally, bundles of both tdTomato-positive and EGFP-positive axons undergo ectopic dorsal misrouting in midbrain of Gsx2-iCre conditional null mice (Figure 3Q, Q’, R, R’), which presumably contributes to fewer MSN axons reaching substantia nigra and fewer SCPN axons entering pons (Figure 3Q”, Q”). These findings indicate that mistargeting of Bcl11b-null MSN axons secondarily causes mistargeting of wild-type SCPN axons, while Bcl11b expressed by MSN is dispensable for the interaction between MSN and SCPN axons.

Although ectopic expression of Bcl11b is sufficient to direct some CPN axons into the internal capsule,⁴² and although several molecular controls over neocortical projection neuron development act at least in part by suppressing Bcl11b expression in non-SCPN populations,^17,19,20,43 it is not known whether Bcl11b expression by SCPN is necessary for a subset of SCPN axons (i.e., corticospinal axons) to innervate spinal cord, given the perinatal lethality of Bcl11b^-/- mice.^16,35 To investigate whether Bcl11b expressed in the cortex regulates SCPN axon extension into spinal cord, we retrogradely labeled corticospinal axons from cervical spinal cord at P21 using cholera toxin B subunit (CTB). At P25, substantially fewer somata are labeled by CTB in Bcl11b^flox/flox;Emx1^IRES-Cre/+ cortex than in wild-type cortex (Supplementary Figure 3), indicating that Bcl11b is necessary for optimal corticospinal projection.

Taken together, our results indicate that Bcl11b orchestrates SCPN axon development both cell-autonomously via expression by SCPN, and non-cell-autonomously via expression by MSN.

Bcl11b controls expression of key axon guidance genes in developing SCPN somata

To further investigate cell-autonomous functions of Bcl11b in SCPN axon growth and circuit development, we investigated in vivo SCPN transcriptomic changes induced by loss of Bcl11b function. We first employed SCPN somata purified by fluorescence-activated cell sorting (FACS).^16,44 These investigations identify changes in soma transcriptomes (Figure 4; Supplementary Figure 4), which predominantly reflect dysregulated transcription in the nucleus.

Figure 4.

To purify SCPN somata, we retrogradely labeled SCPN by injecting green fluorescent microspheres into rostral brainstem of P0 Bcl11b^+/+;Emx1^IRES-Cre/+;Rosa26^{LSL-tdTomato/+} and Bcl11b^flox/flox;Emx1^IRES-Cre/+;Rosa26^{LSL-tdTomato/+} mice (Supplementary Figure 4A), then isolated cells at P1 that are positive for both tdTomato (i.e., Emx1-expressing cortical cells) and green microspheres (i.e., retrogradely labeled neurons; SCPN) via FACS (Figure 4A). Although loss of Bcl11b function in the cortex causes substantial reduction of SCPN axons in brainstem (Figure 3; Supplementary Figure 2), the distribution of SCPN retrogradely labeled from rostral brainstem is not markedly different between Bcl11b^flox/+;Emx1^IRES-Crel+;Rosa26^{LSL-tdTomato/+} and Bcl11b^flox/flox;Emx1^IRES-Crell+;Rosa26^{LSL-tdTomato/+} mice (Supplementary Figure 4B). Retrogradely labeled SCPN in lateral cortex are moderately reduced, but the number of isolated tdTomato/microsphere double-positive neurons per hemisphere is only slightly (but not significantly) reduced by Bcl11b deletion from the cortex (Supplementary Figure 4C), indicating that Bcl11b is not required for SCPN axon extension into the midbrain/brainstem, despite its necessity for SCPN axon growth and extension after reaching brainstem.

We next performed RNA-seq on P1 FACS-purified SCPN somata to identify potential transcriptomic changes caused by Bcl11b deletion from the cortex (Supplementary Tables 1 and 2). Soma transcriptomes cluster by genotype (Supplementary Figure 4D), and demonstrate successful SCPN purification, since wild-type SCPN abundantly express SCPN marker and control genes, including Fezf2, Cntn6, Cdh13, Sox5, Ldb2, Grb14, Stk39, and Crym.¹⁶ Interestingly, while exon 2 of the Bcl11b gene, which is flanked by loxP cassettes, is not detected in SCPN from Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice, expression of the remaining Bcl11b gene is increased by ∼60% in these SCPN (q < 10^-6) (Supplementary Figure 4E). Since Bcl11b protein is not detected in Bcl11b^flox/flox;Emx1^IRES-Cre/+ cortex by antibodies against either the N- or C-terminus (Supplementary Figure 4B, F), excision of Bcl11b exon 2 efficiently ablates Bcl11b protein expression in the cortex. As expected, Bcl11a expression is increased by ∼50% in SCPN from Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice (q < 10^-6), since Bcl11a and Bcl11b are genetically cross-repressive.¹⁸ The expression of transcriptional regulators with central control over postmitotic cortical projection neuron subtype specification— Fezf2, Sox5, Tbr1, and Satb2⁵— is indistinguishable between wild-type and Bcl11b-null SCPN (Supplementary Table 3), indicating that initial delineation of overall SCPN subtype identity is superficially normal in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice. However, expression of other key transcriptional regulators is altered, including the upregulation in Bcl11b-null SCPN of CSN marker Etv1/ER81⁴⁵ and transcriptional regulators that regulate CThPN differentiation vs. alternate SCPN differentiation— Zfpm2/Fog2 and Tle4^46,47 (Supplementary Table 3). Corticospinal subpopulation-specific genes Crym and Cartpt⁸ are downregulated by Bcl11b deletion, consistent with their functions in corticospinal projection. These dysregulations indicate altered progressive refinement of SCPN subtype identity in the absence of Bcl11b.⁵

Overall, expression of 515 genes is significantly altered (q < 0.1) in SCPN by Bcl11b deletion from the cortex (Figure 4B; Supplementary Table 2). Of the 515 differentially expressed genes (DEGs), 299 are upregulated and 216 are downregulated (Figure 4B). 35% of previously identified / reported SCPN identity and development genes have significantly altered expression upon Bcl11b deletion, while genes known to be expressed by CPN and/or corticotectal neurons, but not by SCPN, remain unchanged (Supplementary Table 4),¹⁶ supporting Bcl11b’s specific role as a central regulator of SCPN axon development. To identify DEGs known in other systems to be involved in biological processes possibly related to SCPN axon development, we performed gene ontology (GO) term enrichment analysis (Supplementary Table 5). The 10 most significant terms are notably enriched for cell membrane-related biological processes (e.g. trans-synaptic signaling, membrane potential, ion transport, cell adhesion, and receptor protein signaling) (Figure 4C; Supplementary Table 5), indicating that Bcl11b regulates expression of intercellular communication molecules relevant to axon guidance. Therefore, we focused a next stage of investigation on the effect of Bcl11b deletion on well-known axon guidance molecules in SCPN. Expression of Efnb3, Epha4, Ephb1, Ephb6, Plxna1, Sema3C, Draxin, and Flrt3 change significantly in Bcl11b-null SCPN compared to wild-type SCPN (Supplementary Table 6), suggesting that disrupted transcriptional regulation of axon guidance molecules might contribute importantly to SCPN axon projection defects in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice. In addition to these canonical axon guidance molecules, expression of several cadherin genes, including Cdh13, reported as an SCPN “general identity” gene,¹⁶ are differentially expressed by Bcl11b-null SCPN (Supplementary Tables 1 and 4).

Cdh13 is localized to axons and GC surfaces of SCPN in vivo in a Bcl11b-dependent manner

Many genes dysregulated with loss of Bcl11b function are involved in intercellular communication, including axon guidance and cell adhesion, but little is known about how these genes, downstream of Bcl11b, function in SCPN axon development. Therefore, we focused investigation on Cdh13 as an exemplar DEG in SCPN axon development.

Cdh13, like other cadherin family members, enables intercellular adhesion via homophilic interactions, but, unlike typical cadherins, it is GPI-anchored, rather than membrane inserted, and does not tether cytoskeletal components or transmit signals on its own.^48,49 Cdh13 is highly expressed by wild-type SCPN (233 ± 50 transcripts per million (TPM), Mean ± SD, n = 5), and is reduced by ∼60% in Bcl11b-null SCPN (99 ± 20 TPM, n = 5) (Figure 4B). Cdh13 was previously reported by our lab as a general SCPN “general identity” gene,¹⁶ meaning in that report strikingly higher level expression by SCPN compared with CPN from early development through postnatal maturation, and suggesting potential importance for establishment and maintenance of identity. However, nothing further is known about its protein expression by SCPN, or its potential function(s) in SCPN.

To investigate Cdh13 protein expression, we immunolabeled Cdh13 at P1. In wild-type cortical plate, Cdh13 is detected along cell membranes, and is more abundant in deep layers than superficial layers (Figure 5A; Supplementary Figure 5). In striatum, Cdh13 signal is widely present and is prominently enriched along axon fascicles in the internal capsule (Figure 5A; Supplementary Figure 5). Cdh13 is abundant along CFuPN axons, including CThPN axons innervating thalamic nuclei, and SCPN axons projecting through the midbrain and hindbrain (Figure 5A, B; Supplementary Figure 5). Notably, Cdh13 immunolabeling along SCPN axon fascicles is stronger in more caudal locations along the corticospinal tract (Figure 5B’; Supplementary Figure 5), suggesting that Cdh13 might be particularly involved in distal SCPN axon growth and extension. In P1 Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice, Cdh13 immunolabeling in cortex itself is essentially indistinguishable from wild-type (Figure 5C; Supplementary Figure 5), and is unchanged along CFuPN axons in the forebrain, including CThPN axons innervating thalamus (Figure 5C, D; Supplementary Figure 5).

Figure 5.

Strikingly, however, Cdh13 immunolabeling along SCPN axons in the brainstem is substantially reduced in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice compared to wild-type mice (Figure 5D’; Supplementary Figure 5). To control for the reduction in the number of SCPN axons reaching the brainstem in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice, we immunolabeled in parallel for L1CAM, which is not transcriptionally regulated by Bcl11b in SCPN (Supplementary Table 1). Importantly, while L1CAM reduces in proportion to the reduction of SCPN axons in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice (Supplementary Figure 5), there is a substantially disproportional decrease of Cdh13. These results indicate that Cdh13 is expressed abundantly by SCPN under the control of Bcl11b, and is localized along SCPN axons in vivo, particularly in distal regions of the corticospinal tract.

To investigate whether Cdh13 is present in SCPN GCs in vivo, and whether its GC abundance is altered by Bcl11b deletion from SCPN, we developed a subtype-specific “surface labeling Fluorescent Small Particle Sorting” (slFSPS) approach that enables quantitative identification of surface antigens on subtype-specific GC membranes in vivo (Figure 5E; Supplementary Figure 6A), similar to approaches combining non-specific synaptosome immunolabeling and flow cytometry.^50,51 In this slFSPS approach, bulk GCs (“GC fraction”, isolated via subcellular fractionation⁵²) are incubated with a surface antigen primary antibody, then a fluorescent secondary antibody is added prior to analysis via FSPS (see Materials and Methods).^2,30 To determine the feasibility of this approach, we overexpressed GPI-anchored EGFP or cytosolic EGFP in CPN by unilateral in utero electroporation,⁵³ isolated the GC fraction from contralateral hemispheres,^30,31,52 then incubated this fraction with an anti-GFP antibody. EGFP-GPI GCs are successfully labeled by anti-GFP antibody, and anti-GFP signal is proportional to EGFP fluorescence (Supplementary Figure 6B), indicating that slFSPS is quantitative. Importantly, cytosolic EGFP GCs are not labeled by anti-GFP antibody (Supplementary Figure 6C), demonstrating that slFSPS detects only surface antigens. Additionally, CPN GCs are labeled by anti-NCAM1 antibody (Supplementary Figure 6B, C), consistent with previous work.^2,54 Thus, slFSPS enables quantitative investigation of subtype-specific GC-surface antigens in vivo.

Using slFSPS, we investigated potential presence and abundance of Cdh13 and L1CAM on SCPN GC surfaces in control Bcl11b^+/+;Emx1^IRES-Cre/+;Rosa26^{LSL-tdTomato/+} and loss-of-function Bcl11b^flox/flox;Emx1^IRES-Cre/+;Rosa26^{LSL-tdTomato/+} mice (Figure 5E–G). We isolated bulk GCs, including fluorescent SCPN GCs, from P1 brainstem and spinal cord. In Bcl11b^+/+;Emx1^IRES-Cre/+ mice, antibodies against Cdh13 and L1CAM both produce higher fluorescence intensities relative to a control IgG antibody, demonstrating that Cdh13 and L1CAM are present on SCPN GC surfaces. Importantly, Cdh13 abundance on SCPN GC surfaces is reduced in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice, while L1CAM abundance is unexpectedly increased on SCPN GC surfaces in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice. These findings demonstrate that Cdh13 is both present on normal SCPN GC surfaces in vivo, and locally reduced on SCPN GC surfaces upon Bcl11b deletion from SCPN.

Nuclear gene expression sets the foundation for cellular specification and function. Importantly, however, cells also transport transcripts and/or proteins to peripheral cellular compartments to tightly control local subcellular functions,^31,55-61 enabling prompt and targeted local responses to dynamic environments, and liberating distant subcellular compartments from reliance on transcriptional responses in the nucleus.

We therefore investigated potential dysregulation of subcellular GC RNA localization in Bcl11b-null SCPN, and asked whether subcellular localization of Cdh13 RNA is regulated by Bcl11b. To investigate GC transcriptomes, we purified fluorescent SCPN GCs at P1 from brainstem and spinal cord of Bcl11b wild-type;Emx1^IRES-Cre/+;Rosa26^{LSL-tdTomato/+} or Bcl11b^flox/flox;Emx1^IRES-Cre/+;Rosa26^{LSL-tdTomato/+} mice via established methods^29-32 for subcellular fractionation and FSPS (Figure 5E; Supplementary Figure 7A, B). Following GC purification, sequencing, and assessment by rigorous quality control metrics (Supplementary Figure 7C–F; Supplementary Tables 7–9; see Materials and Methods),^31,62 we identified a focused set of 88 genes as “differentially abundant genes” (DAGs; agnostic regarding potential mechanisms (e.g. expression, trafficking, stability etc.)) between Bcl11b wild-type and null GCs (q < 0.1; Supplementary Figure 7F, G; Supplementary Tables 8 and 10), among which 46 are upregulated and 42 are downregulated. GO term enrichment analysis of DAGs does not identify overrepresentation of any specific biological process across the sets, likely due to their small size and the particular of GO term statistical analysis. These results indicate that Bcl11b is required for appropriate localization of specific mRNAs to SCPN axonal GCs during development. It remains unknown to what extent Bcl11b control of Cdh13 abundance on SCPN GC surfaces is mediated by nuclear Cdh13 transcription vs. subcellular Cdh13 transcript localization and translation.

Together, these findings demonstrate that Cdh13 is localized along SCPN axons, with pronounced distal enrichment along brainstem and spinal cord corticospinal tract axons, and is extracellularly presented on GC surfaces of growing SCPN axons in a Bcl11b-dependent manner.

Cdh13 is a critical downstream target of Bcl11b in SCPN axon extension control

Cdh13 has been implicated in axonal development of the serotonergic system and is required for retinotectal synapse maintenance,^63,64 but its function in cortical development has not been investigated. To determine whether Cdh13 regulates SCPN axon development, we examined the effect of Cdh13 suppression on SCPN axon development. We designed 4 guide RNAs (gRNAs) targeting the Cdh13 transcription start site, and used CRISPR activation (CRISPRa) to screen functional gRNAs.⁶⁵ All 4 gRNAs induce Cdh13 expression (Supplementary Figure 8A). We then selected the most efficient Cdh13 gRNAs (#2 and #3) for further testing with CRISPR interference (CRISPRi) to inhibit Cdh13 expression.⁶⁶ This two-step screening identified that gRNA #3 efficiently knocks down Cdh13 expression in vivo (Supplementary Figure 8B, B’). In utero electroporation of control gRNA⁶⁷ at embryonic day (E) 12.5 leads to efficient SCPN axon extension in the cerebral peduncle and brainstem at E18.5 (Figure 6A).

Figure 6.

In contrast, reducing Cdh13 expression by SCPN via gRNA #3 causes a reduced number of axons reaching the cerebral peduncle and brainstem (Figure 6B). Normalizing axon abundance to the axon abundance in the cerebral peduncle demonstrates that Cdh13 knockdown significantly reduces SCPN axon extension caudal to the cerebral peduncle (Figure 6C). This indicates that Cdh13 importantly regulates SCPN axon extension in the brainstem.

To further investigate in a complementary manner whether Cdh13 downregulation by the loss of Bcl11b in SCPN underlies deficient SCPN axon extension, we overexpressed Cdh13 by Bcl11b-null SCPN (Figure 6D–G; Supplementary Figure 8C–E). We performed in utero electroporation at E12.5 in Bcl11b^flox/flox mice, using a recently developed Cre recombinase-based mosaic gene manipulation system (Binary Expression Aleatory Mosaic; BEAM),⁶⁸ to simultaneously produce interspersed green-fluorescent, Cre-negative control neurons and red-fluorescent, Cre-positive Bcl11b null neurons in the same region of the same cortex (Supplementary Figure 8D). In control BEAM (without Cdh13 overexpression), Cre-negative control SCPN efficiently extend axons into brainstem at P1 (Figure 6D, E). In striking contrast, while Cre-positive SCPN are present in forebrain (Supplementary Figure 8D), their axons are substantially reduced in brainstem (Figure 6D, D’, E), recapitulating the SCPN axon extension defects in Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice (Figure 3; Supplementary Figure 2). Strikingly, in this Cdh13 rescue experiment employing BEAM, overexpression of Cdh13 only by Cre-positive SCPN partially rescues extension of their axons relative to Cre-negative SCPN (Figure 6F, F’, G; Supplementary Figure 8C– E). These results demonstrate that SCPN axon extension defects caused by the loss of Bcl11b in SCPN is mediated, at least in part, by downregulation of Cdh13.

Discussion

The CST is the longest axon tract in the mammalian CNS, and SCPN axons, including corticospinal axons, execute a particularly complex trajectory through multiple brain regions to reach their final targets in the spinal cord.¹⁵ SCPN and other CFuPN axons first extend while ventrally repulsed from cortex via Semaphorin signaling,^69,70 and navigate the pallial-subpallial boundary.^25,37 Once they enter the internal capsule, CFuPN axons fasciculate with each other, thought to be important for further pathfinding,¹⁵ then pass the telencephalic-diencephalic boundary.⁷¹ SCPN axons continue traveling through the midbrain and hindbrain, finishing as axons of CSN reach the pyramidal decussation, cross to the contralateral side, and enter the spinal cord.⁷² The peak birthdate of SCPN is E12.5–13.5 in mouse, and SCPN axons spend the next week alternately growing and pausing until they reach the pyramidal decussation at ∼P0.^73-75 Because SCPN must interpret a wide range of extracellular signals to arrive at their brainstem and spinal cord targets, investigating interactions between SCPN-autonomous and SCPN-non-autonomous sources of guidance is critical to deepen understanding of SCPN axon development, and potentially to enable directed regeneration.

In this work, we demonstrate that Bcl11b/Ctip2 is not only a critical cell-autonomous regulator of SCPN axon development, but also a substantial and critical non-SCPN-autonomous regulator of SCPN axon projection and tract fasciculation via Bcl11b-expressing MSN, outside the cortex. Within SCPN, Bcl11b is necessary for efficient SCPN axon entry into the internal capsule, corticotectal branching past the internal capsule, and further outgrowth caudal to the midbrain. In contrast to Bcl11b-null mice, which have no SCPN axons reaching the spinal cord (though this absolute lack might be partially due to axon growth delay rather than total disruption of SCPN axon development, since Bcl11b-null mice die soon after birth), deleting Bcl11b from only cortex, with normal Bcl11b expression by MSN (Bcl11b^flox/flox;Emx1^IRES-Cre/+ mice), results in substantial but partial reduction of SCPN axons reaching the spinal cord at P25. Strikingly, deleting Bcl11b from only MSN, with normal Bcl11b expression by SCPN (Bcl11b^flox/flox;Gsx2-iCre mice), strikingly impairs SCPN fasciculation in the internal capsule, and results in other pathfinding defects. Additionally, Bcl11b deletion from MSN using Gsx2-iCre disorganizes striatal patch-matrix structure,²⁴ and alters striato-nigral MSN projection (Figure 3). This is intriguing, since SCPN axons and striato-nigral MSN axons converge and course alongside each other.¹⁴ Our results indicate that Bcl11b expressed by MSN regulates broad CFuPN axon fasciculation in striatum and continuing SCPN axon pathfinding in midbrain, primarily via control over multi-circuit MSN cellular functions, rather than by only affecting interactions between MSN and CFuPN/SCPN.

Bcl11b regulates transcription via both direct DNA binding and indirect chromatin remodeling, and modulates distinct sets of genes in distinct cell types and contexts.^21,22,76 To better understand how Bcl11b controls SCPN axon development cell-autonomously, we investigated transcriptomic changes caused by Bcl11b deletion from SCPN. We focused on Cdh13, an exemplar differentially expressed gene (DEG) that is continuously expressed by SCPN from early development into maturation¹⁶ and that has unique molecular features as an atypical cadherin.^48,49,77 Cdh13 expression and localization at P1 marks the CST. Cdh13 is more enriched in distal portions of SCPN axons. Now, our slFSPS subtype-specific subcellular approaches provide strong evidence that Cdh13 is present directly on SCPN GC surfaces.

Since Cdh13, which is GPI-anchored, is not expected to transmit signals intracellularly on its own, we speculate that Cdh13 serves as a tether to tightly fasciculate SCPN axons via physical forces, thereby enabling long-distance projections of SCPN axons. It is also possible that yet-to-be-discovered co-receptor(s) mediate Cdh13 functions.⁷⁸ Notably, the GPI-anchor biosynthesis gene Pigq is a differentially abundant gene (DAG) in null vs. wt SCPN GCs (Supplementary Figure 7G). Importantly, these data reveal that Cdh13 overexpression ameliorates SCPN axon extension defects by Bcl11b deletion. Since there might be parallels to the extremely limited axon regrowth of mature CSN/SCPN after CST injury, it would be interesting to investigate whether Cdh13 manipulation might partially promote axon regeneration. Further, Cdh13 has been implicated in multiple neurodevelopmental disorders, including attention-deficit/hyperactivity disorder and autism spectrum disorder,^79-81 so it would be interesting to investigate whether altered connectivity of cortical projection neurons contributes to pathogenesis and pathophysiology of such disorders.

The distinct contributions to SCPN axon development of Bcl11b expressed by multiple distinct but physically interacting neuron types highlight both the complexity and the precision of cortical output circuitry evolution. Throughout vertebrate evolution, the CNS has expanded and added more structural complexity rostrally than caudally. SCPN somata are located in cerebral cortex, an advanced brain system specific to mammals, and the CST is a notable feature of all mammals. Only minor projections from the telencephalon to the spinal cord are found in a limited number of non-mammalian vertebrates, including some sharks and birds,⁸² and avian pallium has a small population of neurons with characteristics similar to SCPN, such as expression of SCPN genes (including Fezf2 and Bcl11b), and descending axon projections.⁸³ Similar to its expression pattern in mammals, Bcl11b is expressed abundantly in chick striatum.⁸³ We speculate that the functions of Bcl11b we identify across distinct-but-interacting neuron types might have initially emerged in a common ancestor of birds and mammals, then evolutionarily enabled in mammals increasingly complex and interactive motor-sensory circuitry. In particular, we speculate that expanded use of these neuronal circuit functions in mammals, in large part via an expanded SCPN population in an enlarged cortex, has contributed substantially to the repertoire of skilled and dexterous voluntary movement in mammals. Interestingly, Bcl11b is also expressed in the subthalamic nucleus (STN), a structure located at the junction of the diencephalon and midbrain, and SCPN axons normally project along the STN border together with the striato-nigral MSN projection.¹⁴ This suggests that Bcl11b expression in other brain structures might also contribute to guidance of SCPN axons. This raises the more general speculation that single transcriptional regulators beyond Bcl11b might function centrally in distinct cellular populations to enable precise targeting and guidance of long-distance axon projections, thus circuit formation, function, precision, and complexity in the evolution of cortical output circuitry.

How function-specific cortical circuitry is regulated with specificity and precision beyond transcriptional controls is still an underexplored question. This is in large part due to prior inaccessibility of subtype- and stage-specific in vivo subcellular RNA and protein molecular machinery. The paired purifications of somata and GCs by subtype-specific FACS and FSPS fills this important gap. As an exemplar of likely combinatorial and multi-element mechanisms, we identify decreased Cdh13 protein on SCPN GC surfaces upon deletion of Bcl11b from SCPN (Figure 5F, G). We also unexpectedly identify increased L1CAM protein on Bcl11b-null SCPN GC surfaces (Figure 5F, G), even though L1CAM transcript abundance in both soma and GC subcellular compartments are unchanged after Bcl11b deletion (Supplementary Tables 1 and 8), suggesting that L1CAM protein trafficking to GCs is increased by Bcl11b deletion, potentially by altered protein trafficking machinery (see below) and/or by L1CAM substituting in trafficking pathways normally employed for Cdh13.

Further, we identify that deletion of Bcl11b causes disruption of subtype-specific axon GC transcriptomes (Supplementary Figure 7), with changes in developmentally and subcellularly functional transcripts likely contributing to disruption of SCPN axonal connectivity, targeting, and organization. Importantly, there is little overlap between DEGs and DAGs (Supplementary Tables 2 and 10). Changes in GC transcriptomes are dependent at the individual transcript level on combinations of not only transcriptional regulation but also or alternatively regulation of subcellular trafficking— GC transcriptomes are often not a direct consequence of nuclear gene expression changes.²⁹

Multiple potential molecular mechanisms might underlie altered subcellular transcript and protein abundance in axon GCs of Bcl11b-null SCPN. Notably, Bcl11b deletion upregulates Kif5c expression in SCPN somata (Figure 4B; Supplementary Table 2). Kif5 motor proteins regulate anterograde transport of RNAs, proteins, and organelles,^61,84 and the motor domain of Kif5c selectively translocates to axonal GCs,⁸⁵ indicating that local molecular machinery in Bcl11b-null SCPN GCs might be selectively altered, at least in part, by changes in selective anterograde trafficking. Dysregulation of RNA binding proteins (RBPs) that are critical for transcript-dependent RNA stability and trafficking to GCs might also play a key role — nine RBPs are differentially expressed in Bcl11b-null SCPN somata (Ncl, Csdc2, Zc3h13, Ralyl, Rbm24, Pcbp2, Rbms1, Raver2, Larp1b), while no RBP RNAs are differentially abundant in GCs, consistent with somatic translation of most RBPs. We did not identify enrichment of 3’ UTR motifs that are bound by RBPs, presumably due to both the small list of DAGs and the currently limited sets of RBP motifs and/or experimentally validated RBP binding sites reported in the literature. Further investigation offers substantial promise to elucidate mechanisms by which Bcl11b controls corticospinal GC subcellular transcriptomes and proteomes, thus corticospinal circuit development and function, and potentially to more broadly elucidate generalizable mechanisms of subtype- and function-specific circuit development, function, dysfunction, disease, and approaches to regeneration.

Experimental procedures

Mice

All mouse work was approved by the Harvard University Institutional Animal Care and Use Committee, and was performed in accordance with institutional and federal guidelines. The date of vaginal plug detection was designated E0.5, and the day of birth as P0. The genders of early postnatal mice were not determined. Mice were group housed with light on a 12:12 or 14:10 h cycle. Water and food were provided ad libitum.

Wild-type mice on a C57BL/6J or CD1 background were obtained from Charles River Laboratories (Wilmington, MA). Bcl11b^-/- and Bcl11b^flox/flox mice were the generous gift of Prof. Ryo Kominami.^33,35 Emx1^{IRES-Cre/IRES-Cre} (strain number 005628), Rosa26^{LSL-tdTomato/LSL-tdTomato} (strain number 007909), Ai65^{RCFL-tdT/RCFL-tdT} (strain number 021875), and Rosa26^{LSL-LacZ/LSL-LacZ} (strain number 003474) mice were purchased from Jackson Laboratories. Gsx2-iCre mice were generated by Prof. Nicoletta Kessaris and colleagues,³⁶ and were initially provided by Prof. Kessaris and Prof. David Price, then purchased from Jackson Laboratories (strain number 025806). Emx1^{IRES-FlpO/IRES-FlpO} mice were generated previously.^8,9

Immunocytochemistry

Mice were transcardially perfused with cold PBS, followed by 4% paraformaldehyde (PFA) in PBS, and brains and spinal cords were dissected and postfixed in 4% PFA/PBS at 4ºC overnight. Tissue was sectioned at 50 µm thickness on a vibrating microtome (Leica) or on a cryostat (Leica). Non-specific binding was blocked by incubating tissue and antibodies in 2% donkey serum (Millipore, S30-100ML)/0.3% BSA (Sigma, A3059-100G) in PBS with 0.3% Triton X-100, or 0.3% BSA in PBS with 0.3% Triton X-100. Tissue sections were incubated with primary antibodies at 4°C overnight. Secondary antibodies were chosen from the Alexa series (Invitrogen), and used at a dilution of 1:500 for 3–4 hours at room temperature (RT). For DAPI staining, tissue was mounted in DAPI-Fluoromount-G (SouthernBiotech, 0100-20).

Primary antibodies and dilutions used: rabbit anti-DARPP32 (Chemicon, AB1656, 1:250); mouse anti-GAP43 (Millipore, MAB347, 1:500); rat anti-Bcl11b (Abcam, ab18465, 1:2,000; detects N-terminus); rabbit anti-Bcl11b (Novus, NB100-2600, 1:250; detects C-terminus), rabbit anti-GFP (Invitrogen, A11122, 1:1,000); chicken anti-GFP (Invitrogen, A10262, 1:500); rabbit anti-RFP (Rockland, 600-401-379, 1:500); goat anti-Cdh13 (R&D, AF3264, 1:200); normal goat IgG control (R&D, AB-108-C; 1:1,000); rat anti-L1CAM (Millipore, MAB5272, 1:500); rat anti-Somatostatin (Chemicon, AB5494, 1:100); rabbit anti-NP-Y (Immunostar, 22940, 1:500); mouse anti-Parvalbumin (Sigma-Aldrich, P3088, 1:500); and rabbit anti-b-galactosidase (Rockland, 200-4136, 1:500).

Nissl staining

Tissue was initially prepared as for immunocytochemistry above, then mounted on gelatin-coated slides and allowed to dry. Tissue was wetted, then stained in 0.25% cresyl violet for 4 minutes. Tissue was rinsed and dehydrated through a series of ethanol and xylene baths (Two minutes each: 50% EtOH, 70% EtOH, 95% EtOH, 100% EtOH; 45 minutes: xylenes), and mounted in DPX.

Anterograde labeling

Projection neurons were labeled by insertion of crystalline 1,1′.dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) into sensorimotor cortex of P0 fixed tissue. Tissue was incubated in PBS/0.05% sodium azide at 37 °C for four weeks to allow DiI transport to occur. DiI-labeled brains were sectioned sagittally at 70 µm on a vibrating microtome. Images shown are a composite of multiple sagittal montages collapsed from 3-D into a single projected 2-D image of the corticospinal tract.

Retrograde labeling

Developmental SCPN at P0 or P1 were retrogradely labeled unilaterally from rostral brainstem by injecting 690 nl of the retrograde label Green RetroBeads^TM IX (Lumafluor) guided by ultrasound backscatter microscopy (VisualSonics, Vevo 3100) via a pulled glass micropipette (Drummond Scientific, 3-000-203-G/X) with a digitally-controlled, volume-displacement nanojector (Nanoject II, Drummond Scientific). For these neonatal injections, pups were deeply anesthetized by hypothermia under ice for 4 minutes. After injections, pups were placed on a heating pad at 37 ºC for recovery.

CSN at P21 were retrogradely labeled bilaterally from the cervical C1 segment by CTB-555 (2 mg/ml in PBS; Thermo Scientific, C22843). For these P21 injections, mice were anesthetized with isoflurane. After laminectomy at the C1 vertebral segment, 322 nl of CTB-555 solution was injected into each side of the midline via a pulled glass micropipette with a Nanoject II. The skin was sutured, and mice were placed on a heating pad at 37ºC for recovery. Mice were subsequently perfused at P25 for analysis of CTB-labeled CSN in cortex.

DNA constructs

The EGFP-GPI expression construct was derived by subcloning cDNA encoding Acrosin signal peptide:hemagglutinin-tag:EGFP:GPI attachment signal peptide into a pCBIG vector backbone, a gift from C. Lois. Cdh13 cDNA purchased from Sino Biological was subcloned into Cre-activated plasmid backbone (CB-FLEX). For construction of gRNA expression plasmids, the oligonucleotides were inserted into pSPgRNA plasmid, a gift from C. Gersbach (Addgene plasmid # 47108; RRID:Addgene_47108).⁸⁶ Target specific gRNA sequences are as follows: Cdh13 gRNA#1, 5’-AAAACGAGGGAGCGTTATGA-3’; Cdh13 gRNA#2, 5’-CAACTTCCCAAATAGATCAG-3’; Cdh13 gRNA#3, 5’-GGACCAATGGCTTTACAAGA-3’; Cdh13 gRNA#4, 5’-AATCCGTCTTGTAAAGCCAT-3’; and LacZ gRNA, 5’-TGCGAATACGCCCACGCGAT-3’. SP-dCas9-VPR was a gift from G. Church (Addgene plasmid #63798; RRID:Addgene_63798).⁶⁵ dCas9-KRAB-MeCP2 was a gift from A. Chavez & G. Church (Addgene plasmid #110821; RRID:Addgene_110821).⁶⁶

In utero electroporation

Following deep anesthesia of timed pregnant mice, the uterine horns were exposed, and the plasmids, mixed with Fast Green dye (Sigma), were microinjected into one of the lateral ventricles. For electroporation at E14.5, four 35 volt pulses (50 ms on and 950 ms off per pulse) were delivered in appropriate orientation across the embryonic head using 5 mm diameter platinum plate electrodes (CUY650-P5, Protech International Inc) and a CUY21EDIT square wave electroporator (Nepa Gene, Japan). For electroporation at E12.5, four 25 volt pulses (30 ms on and 970 ms off per pulse) were delivered. In the DNA mixture, the concentration was 2 or 4 µg/µl for plasmids expressing cytosolic EGFP or EGFP-GPI, respectively, for labeling CPN for slFSPS. For CRISPR experiments, the concentration was 1 µg/µl for cytosolic EGFP expression plasmid, 1 µg/µl for Cas9 variant expression plasmids, and 2 µg/µl for gRNA expression plasmids. For dual population mosaic labeling (BEAM experiments),⁶⁸ the plasmid concentrations were as follows: 50 ng/µl for the Cre-expression plasmid, 0.5 µg/µl for the CreM plasmid (Cre-inducible Cre recombinase), 2 µg/µl for the Cre-activated myristoylated tdTomato plasmid, 2 µg/µl for the Cre-activated Cdh13 plasmid, and 2 µg/µl for the Cre-silenced myristoylated mClover3 plasmid.

Soma purification (FACS)

Fluorescently labeled SCPN somata were purified according to established approaches.^16,30,44 Briefly, retrogradely labeled cortical hemispheres were collected in chilled dissociation solution, then micro-dissected in chilled HBSS using a fluorescence dissection microscope to remove meninges and ventricular zone, thus enrich for labeled SCPN. Cortical tissue was transferred to dissociation solution and enzymatically dissociated for FACS. Fluorescently labeled somata were purifeid from the bulk soma prep by sorting on a customized FACS SORP AriaII instrument equipped with an 85 µm nozzle operating at 45 p.s.i. Somata were sorted directly into the RLT plus buffer (QIAGEN RNeasy Plus Micro Kit) supplemented with 2-Mercaptoethanol for total RNA extraction.

GC purification (FSPS)

Fluorescently labeled SCPN or CPN axonal GCs were isolated and purified as previously described.^30,31 Briefly, brainstem and spinal cord were dissected rapidly for SCPN GCs, and cortical contralateral hemisphere was dissected rapidly for CPN GCs. Dissected tissues were homogenized in 0.32 M sucrose supplemented with 4 mM HEPES (Thermo Fisher, 15630106), Halt protease and phosphatase inhibitor cocktail (Thermo Fisher, 78442), and RNasin ribonuclease inhibitors (Promega, N2515) at 900 RPM in a glass-Teflon tissue homogenize tube with 10-12 strokes. Following a low speed centrifugation at 1,660 g for 15 min at 4 °C, “post-nuclear homogenate” (PNH) supernatant was collected, and layered onto a 0.83 M and 2.5 M sucrose gradient (Top-to-bottom: PNH, 0.83 M sucrose, 2.5 M sucrose). The discontinuous sucrose gradient was centrifuged at 242,000 g for 47 min at 4 °C in a vertical rotor (VTi50 rotor, Beckman Coulter). The bulk GC fraction (GCF) was extracted from the interface between 0.32 M and 0.83 M sucrose.

Labeled axonal GCs were purified from bulk GCF via fluorescent small particle sorting (FSPS), utilizing a customized BD SORP FACS AriaII instrument.^30,31 Bulk GCF was diluted 4- to 6-fold in chilled PBS prior to loading onto the sorter. GCs were sorted directly into the RLT plus buffer (QIAGEN RNeasy Plus Micro Kit) supplemented with 2-Mercaptoethanol for total RNA extraction.

GC surface labeling fluorescent small particle sorting (slFSPS)

200 µl CPN GCF was diluted 5-fold in chilled PBS with 3% BSA, then incubated with anti-GFP (Invitrogen, A31852, 1:500), anti-NCAM1 (R&D, FAB7820R, 1:500), or anti-rabbit IgG (1:500) antibodies conjugated to Alexa647 for 60 min on ice. 200 µl SCPN GCF was diluted 5-fold in chilled PBS with 3% BSA, then incubated with normal goat IgG control (R&D, AB-108-C; 1:500), anti-Cdh13 antibody (1:100), or anti-L1CAM antibody (1:500) for 60 min on ice, followed by incubation with a Alexa488-conjugated secondary antibody (1:1,000) for 20 min on ice. Following incubation, GCF was diluted 20-fold with 0.32 M sucrose, and spun at 15,000 g at 4ºC for 30 min. After carefully removing supernatant, pelleted GC particles were resuspended in 1 mL 0.32 M sucrose, and diluted 5-fold with chilled PBS prior to loading onto the customized BD SORP FACS AriaII instrument for fluorescent small particle analysis.

Genome-wide RNA-sequencing and bioinformatic analyses

Following total RNA extraction from purified SCPN GCs and somata, RNA was quality controlled and quantified using an Agilent 2100 Bioanalyzer. Purified RNA samples were converted to cDNA using the SMART-seq v4 Ultra Low Input RNA Kit (Takara Bio), and cDNA libraries were generated with the Nextera XT DNA Library Preparation Kit (Takara Bio) by the Harvard University Bauer Core Facility. High-throughput sequencing was performed by loading the pooled GC and soma cDNA libraries on an Illumina NovaSeq platform. The raw FASTQ files of 100 bp paired-end reads were collected for downstream analysis. Five Bcl11b WT and five KO biological replicates were analyzed in both soma and GC analyses.

Adapter sequences were clipped, and unpaired reads were trimmed with Trimmomatic version 0.39. Reads were aligned to the GRCm39 mouse genome with STAR,⁸⁷ and reads mapping to exons were quantified with featureCounts.⁸⁸ Based on read count distributions, genes that had a mean TPM greater than 3 across all replicates, and more than 3 TPM in at least 4 replicates of at least one genotype, were retained for downstream analyses in both soma and GC datasets. To identify “ambient” transcripts among GC transcripts that might arise from supplementing labeled B6 brainstems and spinal cords with unlabeled CD1 cortices during GC purification, variants were called on all samples with GATK HapotypeCaller.^29,89 3,492 genes were filtered from differential analysis, since their CD1 penetrance was significantly higher (q < 0.05, one-sided binomial test) than 30% in at least 4 of the 10 GC samples. 7,946 genes passed this quality control test, and were evaluated for differential abundance. Differentially expressed genes between Bcl11b WT and KO somata were identified with DESeq2, using an FDR of 0.1.⁹⁰ 10,975 genes had a normalized mean count of >= 23, the independent filtering threshold, and effect sizes were shrunken with the apeglm method.⁹¹ Since some GC samples clustered by replicate number (Supplementary Figure 5B), potentially due to batch effects, differentially abundant genes between Bcl11b WT and KO GCs were identified by DESeq2 pairwise comparison, using “batch” in the experiment design, and an FDR of 0.1. 7,794 genes had a normalized mean count of >= 19, the independent filtering threshold. GO enrichment analysis of biological processes was performed with ClusterProfiler,⁹² and terms were reduced based on semantic similarity with Revigo.⁹³ All data are deposited in NCBI’s Gene Expression Omnibus, and will be accessible at GEO Series accession number GSE260678 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE260678) upon publication.

Imaging

For epifluorescence microscopy, tissue sections and cells were imaged using either a Nikon Eclipse 90i or NiE microscope (Nikon Instruments) with a mounted CCD or sCMOS camera (ANDOR Technology), respectively. Z stacks were collapsed using the “Extended Depth of Focus” function on the NIS-Elements acquisition software (Nikon Instruments). Images were processed using Fiji.⁹⁴ For confocal imaging, samples were imaged on an LSM 780 (Zeiss).

Cell culture and Immunoblot analysis

HEK 293T cells were cultured at 37ºC with 5% CO² in Dulbecco’s modified Eagle medium (Gibco, 10566024) supplemented with 10% (v/v) fetal bovine serum, 100 U/ml penicillin, and 100 µg/ml streptomycin (Thermo Scientific, 15140122). Cells were transfected using FuGENE HD transfection reagent (Promega, E2311).

One day following transfection, cells were washed with PBS, and lysed with a cell lysis buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1.0% Triton X-100, 1 mM dithiothreitol, 1 mM EDTA, and Halt protease and phosphatase inhibitor cocktail (Thermo Scientific, 78440). The lysate was centrifuged at 20,000g at 4 ºC for 15 min to collect supernatant. Following standard Tris-glycine SDS–PAGE, resolved proteins were electroblotted onto PVDF membranes using semi-dry transfer. Membranes were incubated with primary antibodies diluted in 5% BSA in TBS with 0.1% Tween-20. The following primary antibodies were used for immunoblotting: mouse anti-β-actin (Santa Cruz, sc-47778, 1:1,000); goat anti-Cdh13 (R&D, AF3264, 1:200). HRP-conjugated secondary antibodies were used for ECL imaging. Immunoreactive bands were detected by chemiluminescence using SuperSignal West Pico PLUS (Thermo Scientific, 34580), which was visualized using a CCD camera imager (FluoroChemM, Protein Simple).

Statistics

Statistical analyses were performed using Graphpad Prism 9.0 software. Sample sizes were predetermined by consistency with established studies. A two-tailed t-test or Mann-Whitney test was used for comparison of two data sets. A two-way ANOVA followed by Bonferroni multiple comparison test was used for experiments to investigate progressive SCPN axon extension. Equal variances between groups and normal distribution of data were presumed, but not formally tested.

Supplementary figures

Supplementary Figure 1.

Supplementary Figure 2.

Supplementary Figure 3.

Supplementary Figure 4.

Supplementary Figure 5.

Supplementary Figure 6.

Supplementary Figure 7.

Supplementary Figure 8.

Acknowledgements

We thank J. Heo, K. Liu, H. McKee, M. Ross, B. Wall, and G. Wheeler for superb technical assistance; K. Ozkan for schematics; members of the Macklis laboratory for scientific discussions and helpful suggestions; N. Kheradmand and J. LaVecchio of the HSCRB-HSCI Flow Cytometry core; the Harvard Bauer RNA sequencing core; the Harvard Center for Biological Imaging for infrastructure and support; R. Kominami, Y. Katsuragi, N. Kessaris, and D. Price for generous sharing of mice. This work was supported by grants from the National Institutes of Health (R01s NS045523 and NS075672, with additional infrastructure supported by DP1 NS106665, NS049553, and NS104055), the Max and Anne Wien Professor of Life Sciences fund, an Allen Distinguished Investigator award from the Paul G. Allen Frontiers Group, and the Massachusetts Department of Public Health Spinal Cord Injury Fund to J.D.M.. Y.I. was partially supported by fellowships from the Uehara Memorial Foundation, Kanae Foundation for the Promotion of Medical Science, the Murata Overseas Scholarship Foundation, and by the DEARS Foundation (to J.D.M.). M.B.W. was partially supported by NIH individual predoctoral National Research Service Award NS064730 and the DEARS Foundation. L.C.G. was partially supported by the Harvard Medical Scientist Training Program, NIH individual predoctoral National Research Service Award NS080343, and the DEARS Foundation. A.K.E. was supported by a Swiss National Science Foundation Postdoctoral Fellowship and a Jean-Jacques et Felicia Lopez-Loreta Foundation Award. D.T. was partially supported by a 2022 NSF-Simon Center at Harvard University QBio Graduate Student Award and NIH NRSA F31 NS127518. J.J.H. was partially supported by NIH NRSA F31 NS103262 and NIH Training Grant T32 GM007226.

Additional information

Author contributions

Research designed by Y.I., M.B.W., L.C.G., and J.D.M.; research performed by Y.I., M.B.W., L.C.G., A.K.E., D.T., and J.J.H.; data analyzed by Y.I., M.B.W., L.C.G., D.T., and J.D.M.; and manuscript written by Y.I., M.B.W., and J.D.M.

Additional files

Supplementary Table 1. Comparison between Bcl11b^+/+;Emx1^IRES-Cre/+ and Bcl11b^flox/flox;Emx1^IRES-Cre/+ SCPN soma transcriptomes. List of all SCPN soma genes that were analyzed by DESeq2, and their expression changes after Bcl11b deletion. Columns correspond to gene names, means of normalized counts, log2-fold expression changes, standard error estimates (lfcSE), p-values, q-values, and ENSEMBL gene IDs. Positive log2-fold expression changes indicate increased abundance after Bcl11b deletion.

Supplementary Table 2. Differentially expressed genes in SCPN somata after Bcl11b deletion. List of 515 SCPN soma DEGs (q < 0.1) after Bcl11b deletion. Extracted from Supplementary Table 1.

Supplementary Table 3. Comparison of transcriptional regulators controlling cortical projection neuron development between Bcl11b^+/+;Emx1^IRES-Cre/+ and Bcl11b^flox/flox;Emx1^IRES-Cre/+ SCPN somata. List of transcriptional regulator genes expressed by P1 SCPN that play key roles in cortical projection neuron development, and their expression changes after Bcl11b deletion. Extracted from Supplementary Table 1.

Supplementary Table 4. Comparison of SCPN marker genes between Bcl11b^+/+;Emx1^IRES-Cre/+ and Bcl11b^flox/flox;Emx1^IRES-Cre/+ SCPN somata. List of SCPN marker genes expressed by P1 SCPN, and their expression changes after Bcl11b deletion. Extracted from Supplementary Table 1.

Supplementary Table 5. GO term enrichment analysis of SCPN soma DEGs after Bcl11b deletion from SCPN somata. List of GO terms that are significantly over-represented (q < 0.05) among 515 SCPN soma DEGs. All SCPN soma genes were used as background. Columns correspond to GO term IDs, descriptions, q-values, gene counts, gene ratio, and gene names.

Supplementary Table 6. Comparison of axon guidance genes between Bcl11b^+/+;Emx1^IRES-Cre/+ and Bcl11b^flox/flox;Emx1^IRES-Cre/+ SCPN somata and GCs. List of axon guidance genes expressed by P1 SCPN and/or abundant in P1 SCPN GCs, and their expression/abundance changes after Bcl11b deletion. Extracted from Supplementary Tables 1 and 8.

Supplementary Table 7. Comparison between Bcl11b^+/+;Emx1^IRES-Cre/+ SCPN soma and GC transcriptomes. List of all Bcl11b^+/+;Emx1^IRES-Cre/+ SCPN soma and GC genes that were analyzed by DESeq2, and their subcellular mRNA localization. Columns correspond to gene names, means of normalized counts, log2-fold transcript abundance changes, standard error estimates (lfcSE), p-values, q-values, and ENSEMBL gene IDs. Positive log2-fold transcript abundance changes indicate higher transcript subcellular localization in GC compartment over soma compartment.

Supplementary Table 8. SCPN GC transcriptome comparison between Bcl11b^+/+;Emx1^IRES-Cre/+ and Bcl11b^flox/flox;Emx1^IRES-Cre/+. List of all SCPN GC transcripts that were analyzed by DESeq2, and their abundance changes after Bcl11b deletion. Columns correspond to gene names, mean of normalized counts, log2-fold transcript abundance changes, standard error estimates (lfcSE), p-values, q-values, and ENSEMBL gene IDs. Positive log2-fold transcript abundance changes indicate increased abundance after Bcl11b deletion.

Supplementary Table 9. GO term enrichment analysis of GC-enriched and soma-enriched genes in Bcl11b^+/+;Emx1^IRES-Cre/+ SCPN. List of GO terms that are significantly over-represented (q < 0.05) among 2,049 GC-enriched and 1,938 soma-enriched genes in Bcl11b^+/+;Emx1^IRES-Cre/+ SCPN. All shared genes between SCPN GC and soma genes (6,396 genes) were used as background. Columns correspond to GO term IDs, descriptions, q-values, gene counts, gene ratio, enriched compartment, and gene names.

Supplementary Table 10. Differentially abundant genes in SCPN GCs in Bcl11b deletion. List of 88 SCPN GC DAGs (q < 0.1) after Bcl11b deletion. Extracted from Supplementary Table 8.