Cortical interneurons are inhibitory, locally projecting cells that form a distributed network of repetitive circuits throughout the cortex (Kepecs and Fishell, 2014; Tremblay et al., 2016). To establish this distributed network, interneurons migrate from the ventral subpallium to arealize throughout the cortex in order to form connections with layered pyramidal neurons and other interneurons. The timescale for interneuron migration varies widely across species. The process lasts days in mice (Lavdas et al., 1999; Marín and Rubenstein, 2001) and several months in humans (Arshad et al., 2016; Hansen et al., 2013; Ma et al., 2013). This species difference extends to stem cell-derived interneurons where mouse ES-derived interneurons develop rapidly (Au et al., 2013; Maroof et al., 2010; McKenzie et al., 2019), while human stem cell-derived interneuron (hSC-interneuron) development is protracted (Nicholas et al., 2013; Shao et al., 2019). Indeed, this limitation has restricted the scope of analysis on hSC-interneurons, hampering detailed functional studies.

In this study, we sought to identify factors that regulate the timing of interneuron migration across species. Using embryonic mouse development as a model system, we found that interneuron migration coincides with vascularization of the medial ganglionic eminence (MGE), and that by manipulating the degree of MGE vascularization in vivo, it regulates the degree of interneuron migration into the cortex. Using an in vitro approach, we identified paracrine factors, SPARC and SerpinE1, produced by endothelial cells that enhance interneuron migration in mouse MGE explants and organotypic cultures. Given the slow developmental rate of hSC-interneurons, we tested whether SPARC and SerpinE1 treatment similarly induced migration in human cells. We found that treated hSC-interneurons xenografted into neonatal mouse cortex exhibited enhanced migration. Further, transplanted hSC-interneurons also showed greater morphological complexity, and more mature electrophysiological characteristics compared to controls. These data suggest that interneurons in mice and humans share a common vascular-based mechanism that regulates their developmental timing. And, by characterizing this process in mice, we were able to reverse-engineer our findings to accelerate functional maturation in human interneurons.

Cortical interneurons are generated from a distal source (primarily the MGE) and undergo long-distance migration to their final destination in the neocortex. In the embryonic mouse, this developmental process is rapid: interneurons are generated starting ~e11 and robustly migrate into the cortex by e15 (Figure 1A; Lavdas et al., 1999; Marín and Rubenstein, 2001). As a result, it is tempting to assume that newly born interneurons automatically transition to a migratory state. As a counterexample, however, in human fetal development interneuron migration into the cortex is highly protracted. Previous studies have found that postmitotic interneurons slowly transition to a migratory state to populate the neocortex over the course of many weeks (Arshad et al., 2016; Hansen et al., 2013; Ma et al., 2013). To confirm this, we examined interneurons in fetal cortical sections by immunohistochemistry for Dlx2 (an interneuron marker). Consistent with prior reports, we observed an increase in the density of Dlx2⁺ interneurons in the cortex over time (15 post-conception weeks [pcw] to 22 pcw) (Figure 1B, Figure 1—figure supplement 1A,B). We hypothesized that the discrepancy between human and mouse interneuron development may be due to an external cue delivered to the MGE to promote interneuron migration. We therefore examined the developing mouse MGE (from e10.5 to e15.5) by histology and observed a striking increase in vascularization of the MGE and the underlying mantle region during the time period when mouse interneuron migration initiates (Figure 1C,D; Daneman et al., 2009; Paredes et al., 2018). Similarly, we found that vascularization of human fetal MGE slowly increased with developmental age, with robust vascularization not occurring until after 20 pcw (Figure 1E, Figure 1—figure supplement 1C).

To determine whether there is a causal relationship between MGE vascularization and interneuron migration, we examined Apcdd1 mutant mice (Mazzoni et al., 2017). Apcdd1, a negative regulator of Wnt/β-catenin signaling, is critical for CNS vascular development and blood-brain barrier maturation (Shimomura et al., 2010). We examined MGE vessel density and average vessel size in Apcdd1 GOF (an endothelial cell-specific transgenic gain-of-function) and LOF (whole animal null) mutants at e14.5 and found that, similar to the retina and cerebellum (Mazzoni et al., 2017), vascularization in the MGE is decreased in Apcdd1 GOF and increased in Apcdd1 LOF mice versus wildtype controls (Figure 1F–H). To confirm that a whole animal Apcdd1 null was appropriate for analysis, we performed in situ hybridization for Apcdd1 and found that its expression is confined to vasculature in the embryonic mouse brain (Figure 1—figure supplement 3). Importantly, fewer interneurons migrated into the cortex in Apcdd1 GOF mutants, in which vascularization was less extensive compared to wildtype controls. Conversely, a trend toward more calbindin + migrating interneurons were present in Apcdd1 LOF mutants (p=0.236, paired t-test) in which vessel size and density was greater (Figure 1I). These results suggest that the density of endothelial cells within neural tissue is a critical regulator of interneuron migration.

In order to identify a mechanism by which this occurs, we used an MGE explant culture approach from mouse embryos to quantify the number of migrating neurons at various developmental stages (Figure 2A, Figure 1—figure supplement 2A–D). Given that the MGE is progressively vascularized from e10 to e15, we reasoned that MGE explant migration would increase with developmental age. We further reasoned that in early MGE explants that were not extensively vascularized, that mouse interneurons would be similarly immobile like human interneurons. To normalize for differences in MGE size at different ages, whole MGE tissue was dissected and sectioned at 250 μm. Then, the number of DAPI + migratory cells was normalized to the surface area of the MGE explant. Using this approach, we found a strong linear correlation between the number of migratory cells and MGE explant surface area (r² = 0.6829) (Figure 1—figure supplement 2B). Further, we found that nearly all migratory DAPI + cells were interneurons using the Dlx6a^Cre driver line crossed to Ai9 in order to fate-map the migratory lineage (Monory et al., 2006; Figure 1—figure supplement 2C,D). Classic studies have shown that MGE explants at e14.5 and e15.5 exhibit robust interneuron migration within hours of initial plating (Bellion et al., 2005; Polleux et al., 2002; Wichterle et al., 1999). Consistently, we found that interneurons later timepoint MGE explants exhibited robust migration, whereas explants from earlier timepoints (e10.5–e12.5) had a limited capacity to migrate (Figure 2B, Figure 1—figure supplement 2E). One possibility is that interneurons possess an intrinsic timer such that, given sufficient time in culture, early timepoint explants would migrate to the same extent as older MGE explants. However, even when cultured for up to a week, early explants did not significantly migrate more after the first 48 hr (data not shown).

Given the diffuse distribution of blood vessels in the MGE, the most likely mechanism is that endothelial cells produce a paracrine signal to induce interneuron migration. To test this directly, we prepared conditioned medium from primary cultures of embryonic brain ECs (primary culture endothelial cell conditioned medium [p-EndoCM]) and added it to e11.5 MGE explants. Consistent with our hypothesis, addition of p-EndoCM resulted in a robust increase in interneuron migration (Figure 2C). We found, however, that primary embryonic brain EC cultures introduced unwanted variability, and for subsequent experiments, we obtained EndoCM from an immortalized human EC line, HBEC5i (Wassmer et al., 2006). To further reduce inter-experimental variability, interneuron migration counts were normalized to within-experiment negative (untreated) controls. Hence, subsequent data is presented as fold-change in interneuron migration over controls. Similar to p-EndoCM, conditioned medium from HBEC-5i (EndoCM) also robustly increased e11.5 MGE explant migration (Figure 2C). Interestingly, e14.5 MGE migration was also significantly increased by EndoCM (Figure 2C). We next tested the effect of EndoCM on interneuron migration in an organotypic slice culture. Here, we used Dlx6a^Cre; Ai9 + e11.5 embryos in order to visualize interneurons as they migrate within a coronal slice of telencephalon from the MGE into the cortex (Figure 2A,E). As with MGE explants, EndoCM also significantly increased the rate and overall number of interneurons that migrated into the cortex (Figure 2F). As a negative control, we tested the biological activity of conditioned medium from HEK 293 cells (HEK CM). HEK CM did not increase MGE explant migration at either age (Figure 2D). Moreover, we found that pre-boiling EndoCM eliminated its biological activity, suggesting a protein source as a regulator of interneuron migration (Figure 2D). Finally, we size-fractionated EndoCM and found that biological activity was strongly reduced between 30 and 100 kDa (Figure 2G).

In order to identify candidate proteins, we performed bulk RNA sequence analysis on HBEC5i and HEK cells. We screened the dataset for genes with the greatest differential expression, enriched in HBEC5i that produced proteins between 30 and 100 kDa, which were also secreted (Gene Ontology [GO] term: extracellular space) (Figure 3—figure supplement 1). After further curation to eliminate membrane-tethered molecules, we obtained a short list of 24 candidates (Supplementary file 1). We functionally tested a number of candidates, including VEGF-A (Barber et al., 2018) and follistatin; however, two proteins, SPARC and SerpinE1, exhibited the most robust biological activity in their ability to increase e11 MGE migration (Figure 3).

We examined the expression of SPARC and SerpinE1 in e14.5 MGE through publicly available expression databases (Allen Brain Atlas and GenePaint) and found that both were specifically expressed in the embryonic CNS vasculature (Figure 3A,B). We also confirmed by western blotting that SPARC and SerpinE1 were present in EndoCM at significantly higher levels compared to HEK292 CM (Figure 3C,D). We then added recombinant SPARC or SerpinE1 to MGE explants and found that cell migration was increased, in particular with SerpinE1 (Figure 3E). We then tested EndoCM in which SPARC and SerpinE1 activity were depleted. We used a function-blocking antibody for SPARC (Sweetwyne et al., 2004), whereas SerpinE1 activity was blocked with a small molecule inhibitor (SK216; Masuda et al., 2013). These reagents separately significantly reduced the capacity of EndoCM to increase e11.5 MGE migration (Figure 3F). However, inhibition of both SPARC and SerpinE1 does not reduce interneuron migration further, suggesting that the two molecules may act on intersecting pathways. Finally, we tested whether SPARC and SerpinE1 could increase interneuron migration in an organotypic slice over time. We found that both proteins significantly increased the rate and overall number of interneurons that migrated into the cortex over time (Figure 3G). Finally, we tested whether SPARC and SerpinE1 were active components in p-EndoCM. We found that p-EndoCM increased interneuron migration into the cortex in organotypic slice cultures over control and that the effect was abrogated by adding SPARC function-blocking antibody and SK216 to p-EndoCM (Figure 3H).

Previous studies have demonstrated that hSC-interneurons migrate and mature at a slow rate, reminiscent of the protracted time frame of interneuron development in the human fetus (Arshad et al., 2016; Hansen et al., 2013; Ma et al., 2013). Given that SPARC and SerpinE1 elicit interneuron migration in e11.5 MGE explants and organotypic slice cultures, we tested whether these factors might similarly accelerate the developmental time frame for hSC-interneurons. Using a pan-tdTomato expressing human iPSC line (CAG-tdTomato knocked into AAVS1 locus), we achieved efficient differentiation to ventral telencephalic identity using established protocols (Bagley et al., 2017; Figure 4—figure supplement 1A). hSC-interneuron differentiation was confirmed using AAV Dlx5/6-GFP (Dimidschstein et al., 2016; Figure 4—figure supplement 1B). At day 35 of differentiation (DIFF 35), ventral telencephalic organoids were treated with either SPARC, SerpinE1, or both for 14 days. At DIFF 49, we tested for hSC-interneuron migration by dissociating either untreated or SPARC/SerpinE1-treated organoids. Here, we further divided the groups: one group continued to be exposed to SPARC and SerpinE1 and the other was left untreated (Figure 4—figure supplement 2A). We observed a significant increase in migratory distance in the group treated both before and afterward with SPARC and SerpinE1. Importantly, we observed an even greater biological effect when SPARC and Serpin were added to dissociated cells after pre-treatment (Figure 4—figure supplement 2C). Thus, in subsequent xenograft experiments, SPARC and SerpinE1 were added as a pre-treatment for 14 days and also added to the cells at the time of transplantation.

Previous studies have xenografted hSC-interneurons into a more functionally relevant setting: neonatal mouse cortex. They found that functional maturation rate is prolonged, requiring ~7 months (Nicholas et al., 2013; Shao et al., 2019). We then tested the capacity of control and SPARC/SerpinE1-treated ventral organoids to integrate following xenotransplant into immune-compromised (NSG) mouse cortex. We first analyzed transplants 28 days post-engraftment. We confirmed that tdTomato + hSC interneurons were almost all Dlx2+ (Figure 4A) and also found that Combo-treated hSC-interneurons migrated significantly further than controls (Figure 4B,C). At 56 days post-transplantation, we traced and analyzed the morphologies of hSC-interneurons in 3D and found that they possessed longer processes and branched more extensively (Figure 4D, Figure 4—figure supplement 3).

Given the improvements in migration and morphology exhibited by treated hSC-interneurons, we tested their functional maturity by employing whole-cell recordings from tdTomato labeled transplanted cells in acute slices of frontal cortex 56 days post-transplant. Consistent with more mature neuronal function, recordings from treatment group neurons showed significantly reduced membrane input resistance, faster membrane time constants, more hyperpolarized action potential (AP) thresholds, and increased ratios of voltage-gated sodium currents versus voltage-gated potassium currents (Figure 4E–O). Additionally, other measures of neuronal maturity we tested appeared to collectively trend toward more mature functional phenotypes in the treatment group of neurons, albeit these were not statistically significant (Figure 4—figure supplement 4). Specifically, treated cells appear to have larger membrane capacitances (p=0.067), more hyperpolarized resting potentials (p=0.12), larger voltage-gated sodium currents (p=0.20), faster AP rise rates (p=0.21), narrower APs (p=0.22), and generated a greater number of APs (p=0.24) (Figure 4—figure supplement 4). Taken together, these electrophysiological profiles suggest that treated neurons may have an accelerated trajectory toward functional maturity.

CNS vascularization is a tightly regulated process that operates in tandem with neuronal and glial cell development (Paredes et al., 2018). Here, we demonstrate that not only does vascularization of mouse MGE coincide with interneuron migration, but it also plays a regulatory role since genetic manipulation of vascular pruning affects the number of interneurons migrating into the cortex. Further, two EC-derived paracrine factors SPARC and SerpinE1 induce increased interneuron migration as assessed by in vitro assays. Moreover, ablation of SPARC and SerpinE1 in immortalized or p-EndoCM reduced its capacity increase in interneuron migration. Having demonstrated a critical crosstalk between the neural and vascular compartments in mouse, we leveraged our findings to test if SPARC and SerpinE1 can accelerate hSC-interneuron maturation. We found that hSC-interneurons treated with SPARC and SerpinE1 also show a more robust migration both in vitro and upon xenotransplantation into host mouse cortex. Finally, xenografted hSC-interneurons also exhibit significantly more complex morphologies and more mature electrophysiological properties.

Our findings support previous studies that have linked angiogenesis to pathfinding during interneuron tangential migration (Barber et al., 2018; Li et al., 2018) and MGE mitosis (Tan et al., 2016). Further, radial glia in the MGE undergo a transition during embryonic development where their pial endfeet detach and connect to MGE vasculature (Tan et al., 2016). In light of our data, it is tempting to speculate that this rearrangement allows for a more direct communication between ECs and interneurons, thereby facilitating their tangential migration into the cortex. Single-cell RNA-sequencing (RNA-seq) in mouse indicates that SPARC expression is significantly enriched in brain endothelial cells versus other endothelial populations (log₂-fold increase 3.97) (Hupe et al., 2017). In the same study, SerpinE1 expression increases in mouse brain endothelial cells, peaking at e14.5, which coincides with the timepoint where we observe robust migration in MGE explants. Both SPARC (Butler et al., 2016; Girard and Springer, 1995; Sage et al., 1989) and SerpinE1 (Canfield et al., 1989) are expressed in human endothelial cells.

In humans, interneuron migration during fetal development is a protracted process (Arshad et al., 2016; Hansen et al., 2013; Ma et al., 2013). We found that hSC-interneurons exhibit greater migration and faster functional maturation when treated with vascular factors, SPARC and SerpinE1, suggesting that the maturation of CNS vasculature may similarly regulate the timing of human fetal interneurons migration. During human fetal development, both SPARC and SerpinE1 are expressed at low levels in subpallium at 8 and 9 pcw, and both peak in their expression in the cortex between 37 pcw and 1 year of age (Allen Institute BrainSpan Database, Developmental Transcriptome, https://www.brainspan.org/rnaseq/search/index.html). Further, the MGE (also known as the germinal matrix) is the primary site periventricular hemorrhages in prematurely born babies (Luo et al., 2019). It has been hypothesized that this is due to newly formed (and thus weaker) vasculature in the MGE during the third trimester (Ballabh et al., 2004; Ballabh et al., 2007). As such, it is an appealing possibility that the timing of vascularization, which we posit to be the source of SPARC and SerpinE1 in the MGE, helps to coordinate species-specific aspects of neuronal maturation. Here, delayed vascularization in humans may serve a regulatory role in delaying interneuron migration until pyramidal neurons have achieved sufficient maturity to allow for interneuron circuit integration.

We identified SPARC and SerpinE1 as important proteins that account for most of the EndoCM activity. This is consistent with the described roles for both proteins in other systems. SPARC has been implicated in cell differentiation (Hrabchak et al., 2008; Stary et al., 2005) and migration (Arnold and Brekken, 2009). SPARC is also correlated with increased cancer metastasis and its increased expression in cancer stromal cells results in epithelial-to-mesenchymal transition by downregulation of E-cadherin and upregulation of N-cadherin (Nagaraju et al., 2014). SPARC can influence migration through integrins and integrin-linked kinases (Thomas et al., 2010) and by modulating the activity of matrix metalloproteases. Indeed integrin signaling also regulates interneuron migration during development (Stanco et al., 2009). SerpinE1 (also known as plasminogen activator inhibitor-1) is also implicated in cell migration through the uPA/urokinase pathway (Grøndahl-Hansen et al., 1993; Mahmood et al., 2018). The urokinase pathway has been linked to interneuron tangential migration by acting through the c-Met receptor and hepatocyte growth factor (Powell et al., 2001).

In addition to its effects on intracellular signaling, SPARC and SerpinE1 also modulate interactions between cells and extracellular matrix (ECM). As a matricellular protein, SPARC functions at the interface between cells and ECM. Reduced collagen levels are observed in SPARC null mice and there is compelling evidence that SPARC reduces cell adhesion to ECM, enabling an intermediate state of adhesion that is conducive to migration (DiMilla et al., 1991; Palecek et al., 1997). SerpinE1 also modulates ECM by binding to vitronectin. This interferes with its binding to integrin αV, resulting in greater integrin αV-fibronectin binding and reduced cell adhesion. In the context of neural stem cells, the balance between integrin αV interactions with vitronectin and fibronectin helps to regulate self-renewal versus differentiation (Dai et al., 2016; Varun et al., 2017). Vitronectin has a similar role in regulating differentiation in cerebellar granule precursors (Hashimoto et al., 2016).

Taken together, we hypothesize that SPARC and SerpinE1 act by tipping the balance toward cell differentiation and migration. SerpinE1 does so possibly by modulating uPA/urokinase signaling, which is active in the MGE (Powell et al., 2001) and possibly by interfering with integrin interactions with vitronectin. SPARC likely acts as an anti-adhesive factor that favors cell migration, as has been described for radially migrating cortical neurons with SPARC-like 1 (Gongidi et al., 2004). In future, it will be important to study the molecular basis for how SPARC and SerpinE1 function in regulating interneuron differentiation and migration.

Of note, we empirically determined that human ventral telencephalic organoids require ~14 days pre-treatment with SPARC and SerpinE1 before we could detect enhanced migration in vitro. This may reflect a human-specific difference in responding to endothelial cues compared with mouse interneurons that operate under a more compressed timescale. In fact, this may account for previous observations by Nicholas and colleagues that migration of hSC-interneurons into host mouse cortex occurs at a slow rate over 7 months (Nicholas et al., 2013). This may be explained by host vascularization of the xenograft, which has been shown to occur over a period of months (Mansour et al., 2018). We hypothesize that pre-treatment with SPARC and SerpinE1 allows our xenograft to bypass host vascularization as a means to trigger migration. Therefore, pre-treatment may prime hSC-interneurons to more rapidly integrate and mature into the host cortex. Of note, we found that levels of SPARC and SerpinE1 were detected a low levels in the mouse MGE at e14.5 (data not shown), consistent with in situ hybridization from Allen Institute and GenePaint (Figure 3A and B). This stood in contrast to the higher levels of SPARC and SerpinE1 used to elicit a biological effect in hSC-interneurons. We hypothesize that this is due to context. In culture, SPARC and SerpinE1 are added to the medium and must diffuse through Matrigel-embedded organoids to reach hSC-interneurons. This is in contrast to the embryonic microenvironment in which vasculature is in close contact with interneuron progenitors, and this configuration likely enhances the effects of SPARC and SerpinE1. A number of studies have utilized bioengineering to recapitulate the vascular microenvironment (Raghavendran et al., 2016; Shirure et al., 2017). Employing SPARC and SerpinE1 in such a context could potentially amplify the effects we demonstrated in this study.

Our findings demonstrate the utility of priming hSC-interneurons for transplantation by pre-inducing them to adopt a migratory state. Although there are likely other impediments to overcome before hSC-interneurons are fully synchronized with the developmental timescale of host mouse cortex, our findings represent an important step toward harnessing the full potential of hSC-interneurons. Cortical interneurons are critical modulators of brain function (Kepecs and Fishell, 2014; Tremblay et al., 2016) and it is important to develop human interneuron models of disease (Catterall, 2018; Inan et al., 2013; Marín, 2012; Rapanelli et al., 2017). Moreover, our pre-treated hSC-interneurons also exhibit more maturity in vitro, suggesting SPARC/SerpinE1 treatment may also be effective when applied to organoid and monolayer approaches. Finally, consistent with previous work (Karakatsani et al., 2019), paracrine cues derived from ECs may similarly regulate the development of other neuronal populations. As such, our study may well serve as a general strategy for inducing hSC-neuron functional maturation in other neuronal cell types such as pyramidal neurons.