EphB1 controls long-range cortical axon guidance through a cell non-autonomous role in GABAergic cells

In the developing nervous system, binding of EphB tyrosine kinase receptors to their cell surface-localized ephrin ‘ligands’ triggers bidirectional, intracellular signaling events that regulate proper axon guidance, cell migration, synapse formation and synapse plasticity (Kania and Klein, 2016). Previous studies have revealed deep-layer cortical axon guidance defects in the ventral telencephalon (VTel) of EphB1^−/− mice (Robichaux et al., 2014, 2016; Lodato et al., 2014), and these guidance errors were exacerbated in the EphB1^−/−;EphB2^−/− mice, suggesting partial compensation by EphB2 receptors (Robichaux et al., 2014). EphB1 shows peak expression in cortical layers V and VI at mouse embryonic day (E) 15.5 (Robichaux et al., 2014; Lodato et al., 2014) – a time when long-range cortical axons are navigating toward their subcortical target zones (Grant et al., 2012). However, at this developmental stage, EphB1 is also expressed in the developing epithalamus (Robichaux et al., 2014), GABAergic cell progenitor structures (i.e. ganglionic eminences and the preoptic area), GABAergic interneurons migrating toward the developing cortex (Rudolph et al., 2014) and GABAergic spiny projection neurons (SPNs) of the developing striatum. Developing deep-layer cortical axons also express ephrin-Bs and navigate through EphB1-expressing brain regions in the VTel. As such, it was unclear whether EphB1 regulates cortical axon guidance in a cortical cell autonomous manner or whether it regulates long-range cortical glutamatergic projections indirectly via a key role in other cell populations.

To investigate the cell populations in which EphB1 regulates proper cortical axon guidance, we generated a new floxed EphB1 mouse to allow for Cre-dependent EphB1 loss-of-function analysis. Our findings reveal that EphB1 is required within vesicular GABA transporter (Vgat; also known as Slc32a1)-positive cells, but surprisingly not in glutamatergic cortical neurons, to control long-range cortical axon guidance in the VTel. Moreover, in EphB1 knockout (KO) mice, we observed numerous aberrant subcortical axon bundles comprising both GABAergic striatal axons and glutamatergic cortical axons, suggesting that a subset of long-range cortical axons normally fasciculate along a subpopulation of long-range GABAergic axons to reach their proper target(s). Surprisingly, the misguided axons preferentially grew along major blood vessels in the VTel, suggesting that axonal EphB1 functions to repel navigating axons away from the developing blood vessels. These effects were not produced by EphB1 loss-of-function in D1 or D2 dopamine receptor-expressing SPNs, Tie2-expressing vascular endothelial cells, suggesting that EphB1 functions in a subpopulation of Vgat-positive neurons to indirectly produce cortical axon guidance defects in apposition to developing striatal vasculature.

As the background strain can influence phenotypes in mutant mice, we backcrossed the EphB1^−/− mice (on a CD-1 background; Williams et al., 2003) to C57BL/6 background strain (>8 generations). Similar to the EphB1^−/− on the CD-1 strain (Robichaux et al., 2014), the BL/6-backcrossed EphB1^−/− mice showed multiple axon guidance defects at postnatal day (P) 0 (Fig. 1), compared with wild-type mice (Fig. 1A), including disorganized axons and ectopic axon bundles within the dorsal striatum (Fig. 1B,C; see arrows), descending ectopic axon bundles within the internal capsule and ectopic axon projections descending from the external capsule and terminating near the brain floor (Fig. 1B,D; see arrows). Previous studies have also shown that axons from the somatosensory cortex aberrantly grow in the posterior branch of the anterior commissure in EphB1^−/− mice (Lodato et al., 2014).

We previously reported subtle deficits in thalamic axon guidance in EphB1^−/−;EphB2^−/− mice (Robichaux et al., 2014). As misguided VTel axons in EphB1^−/− mice originate largely from the cortex (Robichaux et al., 2014; Lodato et al., 2014) and ascending thalamocortical projections are thought to influence corticothalamic axon guidance (Molnar et al., 2012), we analyzed Gbx2-expressing thalamocortical projections in the EphB1^−/− mice. In the EphB1^−/− mice, the Gbx-GFP-labeled axon projections appeared to be normal (Fig. S1A,B), the L1CAM-positive misguided axons did not co-localize with GFP-positive thalamic fibers (Fig. S1C,C′) and the cortical barrel fields were similar to wild-type controls (Fig. S1D), suggesting that the cortical axon guidance errors in the VTel of EphB1^−/− mice are unlikely to be caused by aberrant thalamocortical axon navigation.

EphB1 is highly expressed in a dorsal region of the early developing thalamus (Robichaux et al., 2014). Careful analysis of EphB1 expression (using XGal staining) in the Gbx-GFP;EphB1-lacZ mice showed that EphB1 is highly expressed in the epithalamus (Fig. S2B), but largely undetectable in the developing thalamus (Fig. S2A). Indeed, the habenula-specific marker Brn3a (Pou4f1; Quina et al., 2009) revealed strong co-localization with EphB1 (X-gal) at E14.5, and EphB1 was highly expressed in the adult habenula (X-gal; Fig. S2B,C, respectively). Despite its strong expression in the habenula, we detected no axon guidance defects in the habenular commissure, the fasciculus retroflexus or the habenular axon tract in the interpeduncular nucleus in P0 EphB1^−/− pups (Fig. S2D). Together, these findings suggest that EphB1 is expressed highly in the developing epithalamus/habenula, but it is not required for normal habenular or thalamic axon guidance. The deficits in thalamic axon guidance in EphB1^−/−;EphB2^−/− mice (Robichaux et al., 2014) might be due to the loss of EphB2 or to a synergistic effect of the loss of both EphB1 and EphB2.

To determine the cell population(s) in which EphB1 functions to regulate proper cortical axon guidance, we generated a mutant mouse with loxP sites flanking EphB1 exon 3 (EphB1^lox/lox; Fig. S3) using traditional homologous recombination. To confirm that Cre-dependent loss of EphB1 exon 3 reproduced the original EphB1^−/− phenotype, we generated germline transmission of the Δexon 3 allele by crossing EphB1^fl/fl mice with Prm1-Cre mice. After extensive backcrossing to C57BL/6, we confirmed the loss of EphB1 expression in the brain of EphB1^ΔEx3/ΔEx3 mice (Fig. S4A) and, importantly, we observed the same VTel axon guidance defects observed in the original EphB1^−/− mice (Fig. S4B).

As EphB1 is expressed in the deep layers of the developing cortex (Robichaux et al., 2014; Lodato et al., 2014), we crossed EphB1^lox/lox mice with Emx1-Cre mice to generate conditional EphB1 KO (EphB1 cKO^Emx1) in most cortical and hippocampal glutamatergic pyramidal neurons starting at ∼E11.5 (Liang et al., 2012). Despite the loss of cortical EphB1 expression in the EphB1 cKO^Emx1 mice (Fig. S5A), we observed no cortical axon guidance defects (Fig. 2A,B). EphB1 is also highly expressed in multiple GABAergic neuron populations (Robichaux et al., 2014; Rudolph et al., 2014), including those within the developing VTel. To test EphB1 function in GABAergic neurons, we generated an EphB1 cKO in virtually all GABAergic populations using Vgat-ires-Cre mice (Vong et al., 2011) (Fig. S5B-D). Interestingly, P0 EphB1 cKO^Vgat pups completely phenocopied the cortical axon guidance defects found in the global EphB1^−/− pups (Fig. 2C,D). Together, these data revealed that EphB1 functions in GABAergic cells to influence cortical axon guidance.

A subset of cortical GABAergic neurons [e.g. some parvalbumin (PV)- or somatostatin (Sst)-positive neurons] project long range to subcortical regions (Melzer and Monyer, 2020), suggesting a role of EphB1 in long-range projecting cortical GABAergic axon guidance. Using compound mutant mice EphB1 cKO^Vgat;tdTomato^Vgat (EphB1^lox/lox×Vgat-Cre×Ai14), we indeed observed tdTomato-positive GABAergic long-range projecting axons located within the L1CAM-positive ectopic axon bundles at both P0 (Fig. 2E,F) and E15.5 (Fig. S6), suggesting that the GABAergic axon guidance deficits emerge as early as the misguided cortical axons. However, using a Cre-dependent virus approach in EphB1 cKO^Vgat mice, we did not detect Vgat-positive axons from the somatosensory cortex (i.e. the origin of many of the misprojected VTel axons in EphB1^−/−; EphB2^−/− mice; Robichaux et al., 2014) within the ectopic axon bundles of EphB1 cKO^Vgat mice (Fig. S7). These data suggest that the misguided GABAergic axons do not originate from long-range projecting cortical GABAergic neurons in the somatosensory cortex. GABAergic striatal SPNs in the striatum also project long range, and EphB1 is expressed in GABAergic cells of the developing and adult striatum (Rudolph et al., 2014; Fig. S5D). Using the same Cre-dependent virus approach in EphB1 cKO^Vgat mice, we clearly detected bundles of ectopic Vgat-positive cell axons originating from the dorsal striatum in the EphB1 cKO^Vgat mice (Fig. 3D), compared with none seen in the control mice (Fig. 3A). To verify that misguided VTel axons in EphB1 cKO^Vgat mice also originate from cortical glutamatergic neurons, we used a virus labeling approach to label CaMKIIα-positive glutamatergic neurons in the somatosensory cortex. Indeed, in EphB1 cKO^Vgat mice, we observed misguided, CamKIIα-positive cortical glutamatergic axons in ectopic VTel axon bundles (Fig. 3E), compared with none seen in control mice (Fig. 3B). Importantly, in EphB1 cKO^Vgat mice, we detected co-fasciculation of ectopic Vgat-positive dorsal striatum axons and somatosensory cortex glutamatergic cell axons within the VTel ectopic axon bundles and in the posterior branch of the anterior commissure (Fig. 3F), compared with none seen in control mice (Fig. 3C). Together, these data revealed that EphB1 controls proper long-range glutamatergic cortical axon guidance through a cell non-autonomous role in GABAergic cells. Interestingly, in control mice, we also observed cortical glutamatergic axons co-mingled with striatal GABAergic axons along the rostro-caudal axis – specifically in the tracts projecting to the globus pallidus (Fig. S8A-D) and to the substantia nigra (Fig. S8E-H), two major targets of SPNs. These data strongly suggest that cortical glutamatergic axons traveling through the striatum fasciculate with striatal GABAergic axons to reach their proper targets.

The vast majority of striatal GABAergic cells are SPNs that express D1 or D2 dopamine receptors (Anderson et al., 2020). To test whether EphB1 might play a crucial role in D1- or D2-SPNs to produce the axon guidance defects, we first analyzed the effect of EphB1^−/− on D1- and D2-SPN projections in the Drd1-tdTomato or Drd2-GFP reporter mice. However, we failed to detect tdTomato-positive (D1) or GFP-positive (D2) axons within the VTel axon bundles in EphB1^−/− mice (Fig. S9A,B). Moreover, despite effective recombination in the Drd1-Cre and Drd2-Cre mice at E14.5 (Fig. S10A,B), conditional EphB1 KO in either of these transgenic lines (i.e. EphB1 cKO^D1 and EphB1 cKO^D2) failed to produce the axon guidance deficits (Fig. 4A,B), suggesting that the VTel axon guidance phenotype in EphB1^−/− mice is not caused by a key role in developing D1- or D2-SPNs.

A subpopulation of Sst and PV GABAergic neurons are known to project long range from the cortex (Melzer and Monyer, 2020; Fig. 4C). PV neurons also project long range from the globus pallidus and substantia nigra (Saunders et al., 2016; Rodriguez and Gonzalez-Hernandez, 1999). In EphB1^ΔEx3/ΔEx3;PV-tdTomato mice, we detected long-range PV-positive axons within the misguided axon bundles (Fig. 4D). However, due to the absence of a PV-specific Cre line that recombines embryonically, we were not able to test the role of EphB1 specifically in PV neurons.

In the EphB1^ΔEx3/ΔEx3 mice, we noticed that the VTel ectoptic axon fascicles were typically located in a similar anatomical location and pattern as VTel vasculature (Fig. 5A,B) (Dorr et al., 2007; Xiong et al., 2017; Kirst et al., 2020). Interestingly, the ectopic axon bundles in EphB1^ΔEx3/ΔEx3 mice were observed in close apposition to large CD31 (Pecam1)-positive blood vessels (adults; Fig. 5C) and elastin-positive arteries (P0; Fig. 5D). In addition, long-range GABAergic projections were observed in the ectopic axon bundles located beside the striatal blood vessels (Fig. 5E).

Vascular endothelial cells express many axon guidance molecules, including members of the Eph/ephrin family (Fig. S11; Walchli et al., 2015; Adams et al., 1999; La Manno et al., 2021), that are required for proper vasculature development. Moreover, a subpopulation of vascular endothelial cells also express Vgat (Baruah and Vasudevan, 2019; Li et al., 2018), and EphB1 is expressed in angioblasts (i.e. endothelial cells precursor cells) at E14.5 (Fig. S11). To test whether EphB1 is required in Vgat-positive endothelial cells (instead of Vgat-positive GABAergic neurons) to possibly prevent ephrin-B-expressing GABAergic and glutamatergic axons (Fig. S11) from growing along developing blood vessels, we generated vascular endothelial cell-specific EphB1 cKO^Tie2 mice using Tie2-Cre mice (Kisanuki et al., 2001), where Cre expression in endothelial cells begins by ∼E13 (Li et al., 2018). However, although the Tie2-Cre mice showed specific brain vasculature recombination (Fig. 6A), the EphB1 cKO^Tie2 mice displayed no VTel axon guidance phenotypes (Fig. 6B). Taken together, these data suggest that EphB1 does not cause the cortical axon guidance deficits via a role in vascular endothelial cells, but rather EphB1 in GABAergic axons promotes avoidance of striatal, and indirectly cortical, axons from the developing brain vasculature (Fig. S12 and Table S1).

To identify the cell populations in which EphB1 is required to regulate cortical long-range axon guidance, we generated and validated a new floxed EphB1 mouse. This novel tool allowed us to show that, despite its expression in long-range glutamatergic cortical neurons, EphB1 functions in GABAergic cell populations, but surprisingly not in D1- or D2-receptor-expressing striatal SPNs, to influence correct cortical glutamatergic axon guidance in the developing striatum. We also detected striatal GABAergic axons co-fasciculated with cortical glutamatergic ectopic axon bundles, indicating that EphB1 is required for striatal cell long-range axon guidance, and suggesting that cortical glutamatergic axons are misrouted as an indirect consequence. The Eph/ephrin system is known to be involved in the proper migration of GABAergic neurons generated in the ganglionic eminences and preoptic area (Rudolph et al., 2014; Zimmer et al., 2008, 2011; Steinecke et al., 2014; Rudolph et al., 2010; Talebian et al., 2017); however, little was known about the role of the Eph/ephrin system in GABAergic axon guidance. Moreover, the absence of EphB1 in GABAergic cells causes the misguided VTel axons to navigate along the developing striatal vasculature, possibly owing to a failure of VTel axons to repel from ephrin-expressing vascular endothelial cells.

Based on the corticothalamic handshake hypothesis (Molnar et al., 2012), and our previous findings in EphB1^−/−;EphB2^−/− mice showing deficits in both cortical and thalamic axon guidance (Robichaux et al., 2014), we speculated that the cortical axon guidance errors were an indirect effect of a role for EphB1 and EphB2 in ascending thalamocortical axons in the developing VTel. In the current study, we focused on the VTel axon guidance phenotypes in the single EphB1^−/− mouse, which were less severe than those seen in the EphB1^−/−;EphB2^−/− mice (Robichaux et al., 2014). Using the Gbx-EGFP reporter mouse crossed to EphB1^−/− mice, we observed VTel axon guidance errors, but normal thalamocortical axon projections and no comingling of L1CAM-positive ectopic cortical axons with the GFP-positive thalamic axons. Careful examination revealed that EphB1 is highly expressed in the epithalamus (future habenula), but not detectably in the developing thalamus. However, no habenular axon guidance defects were observed in the EphB1^−/− mice. These findings suggest that the VTel axon guidance defects in EphB1^−/− mice are not likely to be caused by defects in thalamocortical axon guidance. However, EphB2 is expressed in the thalamus (Robichaux et al., 2014), so future studies in compound mutant mice will be required to investigate how EphB1 and EphB2 cooperate to influence proper thalamocortical axon guidance.

EphB1 mediates many of its biological functions through bidirectional signaling induced by contact-mediated binding to ephrins. EphrinB2 (Efnb2) is strongly expressed in the developing cortex and striatum (Robichaux et al., 2014), and we have previously shown that conditional deletion of Efnb2 in Nes-Cre mice, where recombination includes cortical glutamatergic neurons, did not phenocopy the axon guidance deficits found in EphB1^−/− mice (Robichaux et al., 2014). Conditional deletion of Efnb2 using the FoxG1-Cre line produced thalamo-cortical axon guidance deficits similar to those observed in EphB1^−/−;EphB2^−/− mice, but did not phenocopy the cortical axon guidance deficits found in EphB1^−/− mice and in EphB1^−/−;EphB2^−/− mice. EphrinB3, also expressed in the cortex and hippocampus, has been shown to control corticospinal tract axon guidance (Paixao et al., 2013), but future studies will be needed to assess its possible role in the EphB1 cKO^Vgat phenotype.

Using viral-mediated axon tracing tools, we detected GABAergic axons intermingled with the misguided cortical axon bundles in the VTel and located along large blood vessels. As EphB1 is not required within excitatory cortical neurons for the axon guidance phenotypes but is necessary in Vgat, our findings suggest that EphB1 functions in a subset of GABAergic neurons to control proper long-range GABAergic axon guidance and avoidance of the developing VTel vasculature. Moreover, the cortical glutamatergic axon guidance defect is likely caused by co-fasciculation of the descending cortical glutamatergic neuron axons along the misrouted GABAergic fibers. As such, this suggested that a subset of long-range cortical axons might normally fasciculate along a subpopulation of long-range GABAergic axons to reach their proper target. Indeed, we found that, in control mice, somatosensory cortical glutamatergic axons co-fasciculated with striatal GABAergic axons along the rostro-caudal axis, specifically in the axon tracts projecting to the globus pallidus and substantia nigra. A recent study also showed that corticofugal axons fasciculate with striatal axons (Ehrman et al., 2022). In addition, a subpopulation of axons from the somatosensory and motor cortex projects to the globus pallidus (Karube et al., 2019), which is the predominant target of striatal SPNs. Moreover, the EphB1^−/− mice display a reduction in corticospinal tract axons (Lodato et al., 2014), suggesting that the misguided axons are largely corticofugal axons. In control mice, there is a subset of corticofugal axons that project to both the spinal cord and to striatal neurons via axon collaterals (Grillner and Robertson, 2016). Together, these observations strongly suggest that, in our model, a subset of cortical axons are misguided by following misrouted EphB1-null GABAergic striatal axons.

Although D1- and D2-SPNs are the predominant, long-range projecting GABAergic cell type in the striatum, conditional loss of EphB1 in D1- or D2-dopamine-receptor-expressing cells failed to produce axon guidance defects, and no D1- or D2-positive axons were clearly observed in the ectopic bundles. We confirmed Drd1-Cre and Drd2-Cre recombination efficiency using a reporter mouse line (i.e. Ai14), so we cannot completely rule out the possibility that these Cre lines are less efficient at recombining the floxed EphB1 allele, but that explanation of our negative findings seems unlikely. It is also possible that a very small immature population of Drd1/Drd2-negative SPNs (Anderson et al., 2020) might produce the misrouted striatal axons labeled by the viral approach. Unpublished single-nuclei RNA-seq data in the mature striatum from our laboratory revealed two small subpopulations of Grm8-expressing SPNs that express very low or undetectable levels of D1 or D2 dopamine receptor mRNA (B. Hughes, S.B. and C.W.C., unpublished), so it is possible that these emerging ‘unconventional’ SPNs require EphB1 to control their axon guidance and give rise to the EphB1 loss-of-function phenotypes. Sst-positive interneurons in the striatum can project over long distances within the striatum (Straub et al., 2016), but EphB1 cKO^Sst failed to produce the EphB1^−/− phenotypes (A.A. and C.W.C., unpublished). Future studies validating EphB1 recombination in Sst-Cre mice will be necessary to confirm this negative result. In normally developing striatum, the PV-expressing GABAergic interneurons project at short-ranges (Straub et al., 2016), but it remains possible that loss of EphB1 causes striatal PV-positive cells to misproject. In addition, a subpopulation of PV-positive neurons (∼17%) in the globus pallidus projects to the striatum (Saunders et al., 2016), and loss of EphB1 in this population could misroute their axons within the striatum. Future studies examining EphB1 loss-of-function in various GABAergic subpopulations will be crucial for identifying the key GABAergic cell population(s) in which EphB1 influences our observed axon guidance deficits in the EphB1 KO^Vgat mice.

Of note, we found that a subset of striatal GABAergic and cortical glutamatergic axons in EphB1^ΔEx3/ΔEx3 and EphB1 cKO^Vgat are in close apposition to large blood vessels in the VTel. However, conditional loss of EphB1 in Tie2-expressing vascular endothelial cells did not produce the noted axon guidance deficits, suggesting a role of EphB1 in GABAergic neurons. The developing vasculature and navigating axons express many of the same guidance molecules, including ephrins and Ephs (Walchli et al., 2015; Adams and Eichmann, 2010), suggesting that both systems can influence each other. However, the interplay between axon navigation and the developing vasculature has only begun to be examined in the central nervous system (Carmeliet and Tessier-Lavigne, 2005; Mukouyama et al., 2002; Mondo et al., 2020; Andreone et al., 2015). Several studies showed that some peripheral nerves (e.g. autonomic sympathetic axons) navigate along the vasculature via attractive cues expressed on, or secreted by, blood vessels (Mukouyama, 2014). Conversely, certain sensory nerves can provide a template to guide arterial patterning (Adams and Eichmann, 2010). Here, we add to this literature by showing that blood vessels also appear to influence striatal and cortical axon guidance. Except in the external capsule, cortical and striatal axons are generally not found in close apposition to the major blood vessels in the VTel, suggesting that EphB1 likely mediates the repulsion of these VTel axons from developing blood vessels. Using a single-cell RNA-seq dataset from whole embryonic brains at E14.5, we found that ephrinBs are expressed in numerous brain regions and cell types, including neurons, but also angioblasts, endothelial cells, mural cells and perivascular fibroblast-like cells. The arteriole endothelium strongly expresses Efnb2 (Adams et al., 1999; Wang et al., 1998; Gerety et al., 1999), which could promote EphB1-dependent repulsion of navigating axons in the VTel. Unfortunately, loss of Efnb2 from endothelial cells leads to early embryonic lethality (by E11.5) due to angiogenic defects (Adams et al., 1999), precluding our ability to examine whether loss of vascular Efnb2 produces the VTel axon guidance phenotypes seen in the EphB1^−/− mice. Moreover, we cannot exclude the possibility that in the EphB1 mutant mice, the colocalization of axons along the VTel blood vessels might be unrelated to the ability of EphB1 to mediate repulsive axon guidance. Together, our findings here reveal that EphB1 controls long-range cortical axon guidance through a cell non-autonomous role in one or more GABAergic cell populations during early brain development, and that the misguided cortical and striatal axons fascicles are observed in close apposition to the developing VTel vasculature.

EphB1 knockout mice (EphB1^−/−) on a C57BL/6 background were generated by crossing EphB1^−/− mice on a CD-1 background (donated by Dr Henkemeyer, Williams et al., 2003) to C57BL/6 background strain (>8 generations). EphB1-lacZ mice were donated by Dr Henkemeyer (Chenaux and Henkemeyer, 2011). EphB1^ΔEx3/ΔEx3 mice (EphB1 total loss of function) were generated by crossing floxed EphB1 mice (EphB1^lox/lox, described below) to Prm-Cre mice (The Jackson Laboratory, #003328) to produce germline recombination. The Prm-Cre allele was subsequently removed during repeated backcrossing to C57BL/6J wild-type mice. EphB1 cKO were generated by crossing EphB1^lox/lox mice with cell type-selective Cre-expressing transgenic mice (Emx1-Cre; The Jackson Laboratory, #005628), Vgat-Cre (The Jackson Laboratory, #028862), Drd1a-Cre (The Jackson Laboratory, #37156-JAX), Drd2-Cre (GENSAT; MMRRC, #036716-UCD), Tie2-Cre (The Jackson Laboratory, #008863), and they were compared with their Cre-negative littermate controls. The Vgat-Cre, Tie2-Cre, Drd1-Cre and Drd2-Cre lines were crossed to the Ai14 reporter mouse line (The Jackson Laboratory, #007914). To visualize long-range GABAergic projections, the Vgat-Cre line was crossed to both EphB1^lox/lox and Ai14 (EphB1 cKO^Vgat; tdTomato^Vgat) mice or to Ai14 (tdTomato^Vgat). To visualize thalamic and striatal axons in EphB1 mice, Gbx-CreER^T2-IRES-EGFP (The Jackson Laboratory, #022135), Drd1-tdTomato (The Jackson Laboratory, #016204) and Drd2-GFP (Mouse Genome Informatics, MGI:3843608) reporter mice were crossed to EphB1^−/− mice or to EphB1^ΔEx3/ΔEx3 mice. To visualize the projections from PV-positive neurons, the parvalbumin-Cre (PV-Cre, The Jackson Laboratory, #017320) was crossed to the Ai14 mice (tdTomato^PV), and the PV-tdTomato reporter line (The Jackson Laboratory, #027395) was crossed to EphB1^ΔEx3/ΔEx3 mice (EphB1^ΔEx3/ΔEx3;PV-tdTomato). All procedures were conducted in accordance with the Medical University of South Carolina Institutional Animal Care and Use Committee (IACUC) and National Institutes of Health guidelines.

A targeting vector was generated by using a pL452 based mini targeting vector (Liu et al., 2003), which was recombined into mouse 129 strain bacterial artificial chromosome clone BMQ422J21 (Adams et al., 2005). The mini targeting vector was cloned so that loxP site #1 was inserted 722 bp upstream of EphB1 exon 3 and loxP site #2 was 293 bp downstream of exon 3. In addition to this, an FRT site flanked insertion was cloned downstream of loxP site #1, which contained the following cassettes: an engrailed two slice acceptor, the internal ribosomal entry site (IRES) from encephalomyocarditis virus (EMCV) (Bochkov and Palmenberg, 2006), a bovine tau protein and enhanced green fluorescent protein (eGFP) expression sequence (Utton et al., 2002), an SV40 polyadenylation signal and a neomycin selection cassette driven by prokaryotic and eukaryotic constitutive promoters. A capture vector was used to retrieve 10,630 bp for the left homology arm and 10,146 bp for the right homology arm with flanking diphtheria toxin and thymidine-kinase-negative selection cassettes, respectively. Murine stem cells were targeted and screened by the UC Davis Mouse Biology Program. Mice were crossed with Flp recombinase germline expression mice to remove the FRT flanked knock-in cassettes to generate EphB1^lox/lox mice lacking the selection cassettes. Cells that expressed Cre recombinase deleted 1854 bp of the EphB1 genomic locus, which includes all of exon 3, to generate EphB1^Δexon3 mice. Exons 1 and 2 have the potential to generate a small, truncated protein. In the EphB1^Δexon3 mice, exon 2 splicing to exon 4 is predicted to produce either nonsense mediated decay or to code for the following protein sequence: MALDCLLLFLLASAVAAMEETLMDTRTATAELGWTANPASGPVLRGPSRPARKLKAAPTAPPTVAPLQRRLPSAPAGLAITELTLIHQRWRVLVSHRVLEMSSPS (bolded letters are produced by a frame-shift and represent the predicted non-EphB1 amino acids before the premature stop codon). The primers used to genotype floxed EphB1 mice were: forward primer 5′-GGGAGAAGAGAGAGCCTAC-3′; reverse primer 5′-CCAGAGGGCTTTGAGTTAAT-3′ (floxed band: 316 bp; wild-type band: 420 bp). See Fig. S3.

Adult mice were anesthetized with ketamine/xylazine diluted in 0.9% saline (120 mg/kg and 16 mg/kg, respectively) and hypothermia anesthesia was used for pups by placing them in ice for 5-8 min. P0 pups and adult mice were perfused transcardially with 4% (w/v) paraformaldehyde (PFA) in phosphate buffered saline (PBS), the brains were post-fixed overnight in 4% PFA, cryoprotected in 30% sucrose and then sectioned at 40 µm (adults) and 70 µm (pups) using a sliding microtome (Leica). Sections were washed in PBS, incubated in blocking solution [5% (v/v) normal donkey serum, 1% (w/v) bovine serum albumin, 0.2% (v/v) glycine, 0.2% (w/v) lysine, 0.3% (v/v) Triton X-100 in PBS] for 1 h at room temperature (RT) with shaking, incubated with primary antibodies diluted in blocking solution overnight at 4°C under shaking and then washed three times for 10 min in PBS with gentle shaking at RT. Sections were then incubated with secondary antibodies diluted in blocking solution for 90 min at RT with shaking and protected from light, washed three times for 10 min in PBS and mounted in ProLong Gold Antifade Mountant (Invitrogen, #P36931). Antibodies used were: rat anti-L1CAM (1:1000, Millipore Sigma, #MAB5272); chicken anti-GFP (1:1000, Abcam, #13970); mouse anti-Brn3a (1:200, Thermo Fisher Scientific, #MAB1585); rabbit anti-DsRed for tdTomato staining (1:1000, Living Colors, #632496); rabbit anti-RFP for tdTomato staining (1:3000, Rockland, #600-401-379); goat anti-CD31 (1:800, Novus Biologicals, #AF3628); mouse anti-elastin (1:800, Sigma-Aldrich, #MAB2503); Cy3 donkey anti-rat (1:500, Thermo Fisher Scientific, #NC0236073); AlexaFluor 488 donkey anti-mouse (1:500, Thermo Fisher Scientific, #NC0192065); AlexaFluor 488 donkey anti-chicken (1:500, Thermo Fisher Scientific, #703-545-155).

Adult mice were perfused transcardially with 4%(w/v) PFA, and the brains were post-fixed overnight in 4% (w/v) PFA. Female mice were bred and vaginal plugs were assessed with the day of plug detection considered as E0.5. After live-decapitation, embryo brains were drop-fixed in 4% PFA overnight. Adult and embryo brains were then cryoprotected in 30% sucrose. Embryo brains were plunged into M1-embedding matrix (Thermo Fisher Scientific, #1310), flash-frozen for 1 min in isopentane between −20°C and −30°C, and stored at −80°C. Adult brains were cut at 40 µm using a sliding microtome (Leica) and embryo brains at 20 µm using a cryostat (Leica). XGal staining was performed using the beta-galactosidase staining kit (Mirus, MIR 2600), following the manufacturer's instructions. Briefly, the sections were washed in PBS, incubated in the Cell Staining Working Solution containing the X-Gal Reagent in a dark, humidified chamber at 37°C overnight, washed once in PBS and mounted in ProLong Gold Antifade Mountant (Invitrogen, #P36931).

Sections were stained for myelin using the BrainStain Imaging kit (Thermo Fisher Scientific, #B34650; FluoroMyelin, 1:300), following the manufacturer's instructions.

RNA extraction was performed using the miRNeasy Mini kit (Qiagen, #1038703), following the manufacturer's instructions. Total RNA was reverse-transcribed using Superscript III (Invitrogen) with random hexamers, following the manufacturer's instructions. PCRs were performed using the complementary DNA to detect EphB1 expression (for EphB1 cKO^Emx: forward primer 5′-TACAGAGATGCGCTT-3′, reverse primer 5′-ACAGCGTGGCCTGCA-3′; for EphB1^ΔEx3/ΔEx3 and for EphB1 cKO^Vgat: forward primer 5′-AGACATTGATGGACACAAGG-3′, reverse primer 5′-TCAAAGTCAGCTCGGTAATA-3′). GAPDH was used as a control (forward primer 5′-TGAAGGTCGGTGTCAACGGATTTGGC-3′; reverse primer 5′-CATGTAGGCCATGAGGTCCACCAC-3′).

After live-decapitation, the brains were plunged in M1-embedding matrix (Thermo Fisher Scientific, #1310), flash-frozen for 1 min in isopentane between −20°C and −30°C, stored at −80°C, and then cut at 16 µm thick slices using a cryostat (Leica). Sections were kept at −20°C during the cutting process and stored at −80°C. RNAscope^® was performed using the RNAscope^® kit (ACD Bio, #323110) and following the ACD protocol provided by the manufacturer. Sections were immersed in 4% (w/v) PFA for 15 min, then in 50% (v/v) ethanol for 5 min, then in 70% (v/v) ethanol for 5 min, and then twice in 100% ethanol for 5 min. Sections were covered by RNAscope^® hydrogen peroxide for 10 min in a humidified chamber and washed three times in PBS, and the protease incubation step was omitted. Sections were placed in a humidified chamber in the HybEZ™ Oven for all the following steps at 40°C and were washed twice for 2 min in wash buffer after each of the following incubation steps. Sections were incubated with the probes (ACD EphB1 custom probe designed in EphB1 exon3 #541171-C2; Vgat probe #319191; tdTomato probe #317041-C3; C1 probes were used without dilution; C2 and C3 probes were diluted 50× in C1 probe or in diluent) and placed at 40°C for 2 h. AMP1, AMP2 and AMP3 were successively applied on the sections for 30 min each at 40°C. Sections were then covered by the appropriate horseradish peroxidase isoenzyme C and placed at 40°C for 15 min, covered by the appropriate fluorophore [PerkinElmer, #NEL744E001KT (Cyanine3), #NEL741E001KT (Fluorescein), #NEL745E001KT (Cyanine5); 1:2000 in TSA Buffer provided in the RNAScope^® kit], placed at 40°C for 30 min, covered by the HRP blocker and incubated at 40°C for 15 min. Finally, DAPI was applied for 1 min and sections were mounted in ProLong Gold Mountant (Invitrogen, #P36931).

Unilateral stereotaxic injections of AAV5-CaMKIIα-EGFP (Addgene plasmid #50469; virus titer ≥3×10¹² vg/ml) and AAV5-hSyn-DIO-hM4D(Gi)-mCherry (Addgene plasmid #44362; virus titer ≥7×10¹² vg/ml) viruses were performed on adult anesthetized (isoflurane) control (Vgat-Cre) and EphB1 cKO^Vgat (EphB1^lox/lox×Vgat-Cre) mice, into the dorsal striatum (DV: −2.8, ML: +1.6, AP: 0; 150 nl) and into two locations of the somatosensory cortex [(1) DV: −1.9, ML: +3.2, AP: −0.4; (2) DV: −1.4, ML: +2.7, AP: −1.7; 200 nl in each location], using a Nanoinjector (Thermo Fisher Scientific, #13-681-455; 50 nl/30 s). Placement was confirmed by immunohistochemistry (GFP and DsRed antibodies; see above).

Single-cell gene expression data were downloaded from http://mousebrain.org/ (La Manno et al., 2021). Briefly, loom files were converted into Seurat objects with UMI data and metadata. Downstream analysis was performed in R with Seurat (v4.1.0) (Hao et al., 2021) and customized R scripts. Data were preprocessed, normalized and filtered. The Seurat function DotPlot was used to visualize the gene expression and relative abundance of the genes of interest across the different cell subclasses. Cell information was stored in the metadata.

The authors thank Dr Mark Henkemeyer for sharing the EphB1-lacZ knock-in mice. We also thank Ben Zirlin, Brandon Hughes, Rachel Penrod-Martin, Nicolas Narboux-Neme and Nicolas Renier for invaluable assistance with the project. The Genomics and Bioinformatics Core, within the Center of Biomedical Research Excellence in Neurodevelopment and its Disorders (CNDD), provided analysis assistance with RNA-seq datasets. The MUSC Proteogenomics Facility provided equipment for quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR) analyses.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201439.reviewer-comments.pdf