Generating Inhibitory Neuron Diversity through Morphogenic Patterning: From in vivo Studies to New in vitro Models Open Access

Inhibitory GABAergic neurons (INs) play a crucial role in maintaining the excitatory-inhibitory balance in neural circuits and are present in different areas of the brain, including the cerebral cortex and basal ganglia [1]. INs are highly heterogeneous in morphology, molecular identity, and firing properties [2]. They include local neurons that form circuits with neighboring cells and long-range projection neurons connecting distinct brain regions [3]. This diversity enables INs to integrate and fine-tune activity within neuronal networks [4]. Understanding mechanisms regulating the generation of such incomparable cellular diversity from a limited number of progenitors is a crucial challenge in neurodevelopmental neurosciences [5, 6]. In particular, the relative contributions of intrinsic and extrinsic cues need to be disentangled [7‒10]. Recent technical advances, such as single-cell transcriptomics and organoid cultures, have started to tackle these questions and allowed us to gain greater insight into the cellular and molecular mechanisms involved [5, 11, 12].

During embryonic development, INs arise from distinct progenitor domains in the ventral telencephalon, also known as the subpallium. These domains form transient structures that can be broadly categorized into four main regions: the lateral ganglionic eminence (LGE), the medial ganglionic eminence (MGE), the preoptic area (POA) and the caudal ganglionic eminence (CGE). These domains are defined by the expression of specific transcription factors and show, to some extent, a predictive value for the types of INs they generate [13]. For instance, interneurons expressing somatostatin (SST), parvalbumin (PV), and neuropeptide Y (NPY) are born in the MGE/POA while interneurons containing the 5-HT3A receptor, such as those expressing calbindin, vasoactive intestinal peptide, and reelin, come mainly from the CGE [9, 14‒17]. In contrast, the LGE is the primary source of projection INs including striatal medium spiny neurons (MSNs) of types D1 and D2 as well as projection neurons populating the globus pallidus and amygdala [18, 19]. Additionally, it contributes to the generation of olfactory bulb interneurons [20].

In light of the pivotal role of accurate domain formation in maintaining IN diversity, it is crucial to understand the developmental processes that govern domain identities during the early stages of development. The telencephalon starts as a simple sheet of neuroepithelial cells and progressively undergoes regionalization, notably relying on extrinsic factors. Cell to cell communication is indeed a cornerstone during the development of the nervous system. For instance, morphogens, which are signaling molecules secreted from organizing centers, establish concentration gradients that define body axes and play a pivotal role in directing cell fate decisions [21]. Notably, sonic hedgehog (Shh) coming from the ventral midline and bone morphogenetic proteins (Bmps)/Wnt signaling from the dorsal midline contribute to the establishment of the dorsoventral (DV) axis [8].

This review describes the genetic programs that drive DV patterning in the telencephalon and more specifically the mechanisms regulated by morphogens that subdivide the subpallium into distinct progenitor subdomains. We summarize recent work using single-cell transcriptomics and lineage tracing that provide insight on the sequence of events that leads to the emergence of IN diversity. Finally, we discuss the relevance of emerging technologies in addressing these questions as well as revisiting the role of morphogen signaling on subpallial development and IN diversity.

In mice, the specification of the telencephalon primordium and the expression of transcription factor Foxg1 occurs as early as embryonic day (E) 8.5. While dorsal identity requires the presence of Bmps and Wnts secreted by the cortical hem at the dorsal midline, ventral identity is promoted by the combined action of Shh and fibroblast growth factors (Fgfs) signaling. Shh is a morphogen with wide pleiotropic functions during development [22‒24]. Evidence highlights the crucial role of Shh signaling in proper DV patterning as its complete loss function results in the absence of ventral telencephalic derivatives [25]. DV patterning begins with the induction of Shh expression in the prospective subpallium, driven by Shh secretion from the underlying mesoendoderm [26] (shown in Fig. 1a). This signaling cascade directly modulates the activity of the Gli family of transcription factors, the primary effectors of Shh signaling. Specifically, Shh inhibits the repressive form of Gli3 that is widely expressed throughout the developing telencephalon. This inhibition triggers the expression of Fgfs, including Fgf8, in the ventral region, which in turn promote the activation of transcription factors that are crucial for ventral fate specification [27]. For example, studies have demonstrated that Fgf8 regulates the expression of Nkx2-1, a gene frequently used as a marker for the MGE, the first eminence to emerge from the ventral telencephalon, and the POA. In line with this, Fgf expression is absent in the Shh^−/− mutant resulting in the loss of Nkx2-1 expression due to the unrestrained repressive activity of Gli3. However, in the Shh^−/−Gli3^−/− double mutant, Fgf expression remains intact, and DV patterning is preserved, further confirming that Shh ventralizing effect is indirect and mediated through the modulation of Gli3 [28]. Moreover, during neurogenesis Nkx2-1 expression remains sensitive to changes in Shh signaling levels, even in the absence of Gli3 expression [29]. This indicates the presence of additional Gli3-independent mechanisms contributing to the regulation of Nkx2-1 expression.

As development progresses, the telencephalon undergoes further DV regionalization and this is partly due to the joint action of a set of homeobox transcription factors, Pax6 and Gsx2 (previously Gsh2). Pax6 expression starts early on, in the telencephalon primordium, and is quickly restricted to the dorsal portion following Nkx2-1 activation. At this point, the border between the pallium and the subpallium is defined by the expression of Pax6 dorsally and Nkx2-1 ventrally [30]. Shortly after, these two regions become separated by a domain of high Gsx2 expression, a transcription factor involved in the specification of progenitor cells of the LGE and CGE [31, 32]. Moreover, Gsx2 and Pax6 functionally cross-repress each other and are involved in the correct positioning of the pallial-subpallial boundary (PSB) [33]. The PSB is located at the borderline between the ventral pallium and the dorsal LGE and is an important source of telencephalic cell diversity [34, 35]. It also acts as a secondary patterning center, releasing multiple morphogens, including Fgf7 and Sfrp2, a secreted Wnt signaling activator involved in establishing the dorsal-high to ventral-low gradient of Wnt signaling [36, 37] (shown in Fig. 1a).

As previously mentioned, diversity among subpallial progenitors is first apparent during early development with the distinct expression of a small subset of transcription factors that molecularly define each ganglionic eminence (GE) and the POA [13]. The predictive value for the types of INs produced is commonly attributed to the diversity of progenitors residing in these subpallial domains. For instance, SST- and PV-expressing interneurons originate predominantly from the MGE/POA, both defined by Nkx2-1 expression. This transcription factor plays a critical role by inducing target genes such as Lhx6 and Lhx8 [38‒40]. Ultimately, INs derived from the Nkx2-1 domain will populate the cortex, the globus pallidus, the striatum, the hippocampus, and the septum (shown in Fig. 1a) [16]. Conversely, development of striatal MSNs and olfactory bulb interneurons occurs in the LGE (shown in Fig. 1a) and involves a cascade of transcription factors, including Gsx2 [41]. Surprisingly, recent transcriptome analyses found that the diversity of progenitor cells within each GE was less extensive than previously anticipated. Although most studies can determine the broad subpallial domain from which cycling progenitors originated, directly linking these populations to mature IN subtypes remains challenging [42, 43]. Instead, overall cell fate bifurcations of INs seem to occur after progenitors become postmitotic. For instance, future PV- and SST-expressing cortical interneurons were only distinguishable in postmitotic neurons starting at E13. At this stage of development, precursor cells already express markers that continue to be present in mature cells such as Mef2c and Erbb4 for PV interneurons and Sst and Satb1 for SST interneurons [44‒48]. Furthermore, Lee and colleagues [43] were able to identify genes with restricted expression patterns within broad GEs domains, which is reminiscent of previously described spatial biases of certain INs subtypes [49, 50]. Notably, Maf expression is enriched in the dorsal MGE, which produces more SST-expressing interneurons than the ventral MGE, which predominantly generates PV interneurons. In fact, conditional Maf overexpression in MGE progenitors increases the production of SST interneurons at the expense of PV interneurons [45]. It is worth noting that this shift is relatively modest and that the SST:PV ratios remain similar when knocking out exclusively Maf and just slightly shifted in the Maf/Mafb double knockout, where another Maf family transcription factor is also lost [51]. This suggests that additional, yet-to-be-identified factors contribute to specifying interneuron subtype identities.

Similar restricted expression patterns were identified along the DV axis of the LGE. For example, the dorsal part of the LGE was shown to predominantly generate olfactory bulb interneurons, whereas the ventral region appears to primarily produce striatal MSNs [41, 50, 52]. In support of this, postmitotic cells in the ventral LGE were found to express genes such as Ebf1, Zfp503, and Zfhx3, which have been implicated in the specification of MSNs [53‒55]. Notably, Zfhx3 expression can be regulated by the CREB family of transcription factors, which are known for their ability to be activated by a diverse range of extracellular cues [56, 57]. Consistent with this, retinoic acid has been shown to modulate Zfhx3 expression within the context of neuronal differentiation [58]. Additionally, in different parts of the central nervous system Zfhx3 plays an essential role during neurogenesis and mutations in this gene have been linked to intellectual disabilities [59, 60].

While reconstructing developmental trajectories has proven valuable for understanding the molecular pathways involved in INs diversification, it does not provide direct evidence of clonal relationships. To tackle this question, recent studies combined high-throughput single-cell RNA sequencing with clonal lineage tracing [61, 62] and showed that INs exhibit a particularly high degree of divergence, with many clonally related INs showing very different gene expression patterns. For instance, although striatal MSNs and olfactory bulb interneurons are molecularly and functionally very distinct, lineage reconstruction suggests they can be generated by one common progenitor cell [61]. The exact mechanisms involved in the sequential production of different types of INs from a single progenitor cell are currently poorly understood. However, morphogens have been shown to provide additional signals to progenitor cells already engaged in region-specific intrinsic genetic programs [63]. As in the context of spinal cord development, Shh signaling in the subpallium is also involved in INs fate determination after its initial role in DV patterning. Indeed, high levels of Shh favor the production of MGE-derived SST interneurons at the expense of PV interneurons [63, 64]. This is consistent with the expression pattern of several Shh target genes including Gli1, Hipp, and Nkx6-2 in the most dorsal part of the MGE, which as previously mentioned preferentially gives rise to SST interneurons [38].

With the development of tools based on stem cell engineering, known as organoids [65‒69], new strategies to investigate the impact of the extracellular environment upon nervous system development, patterning, and cell diversity has emerged. These organoids can be cultured under specific conditions to guide their fate toward neural tissue via modifications of extracellular cues [69‒71]. For instance, protocols have been developed to generate telencephalon organoids via WNT/BMP pathway inhibition, and refinement of these protocols allows to precisely generate dorsal telencephalon organoids [72‒74], or ventral telencephalon organoids via activation of SHH signaling [73, 75, 76] (shown in Fig. 1b). Recently, it has also been shown that FGF8 treatment can impact DV patterning in cortical brain organoids as it increases the proportion of LGE-derived markers [77]. However, all these protocols induce a homogeneous fate to the entire organoid structure and do not yet instruct a spatial patterning as is described in vivo.

Consequently, new techniques have been developed to create artificial signaling centers in organoids (known to secrete morphogens in vivo). This enhances internal organization and facilitates the investigation of the role of these factors during development (shown in Fig. 1b). Cederquist et al. [78] developed an approach to induce DV patterning in forebrain organoids by embedding inducible SHH-expressing cells at one pole of an organoid. They showed that organoids displayed a gradient-like SHH distribution and induced a spatially restricted pattern of expression of forebrain markers. For instance, similar to in vivo data, an SHH gradient induced a sequential expression of DV markers: NKX2.1, followed by GSX2, and PAX6. Similarly, a recent study used optogenetic stimulation to activate an SHH organizer locally and observed an in vivo-like DV topography and gene expression pattern (neural-tube-like) [79]. Other groups cultured stem cells in microfluidic device to recreate a DV morphogen gradient in their microenvironment and investigated how this approach impacted developmental processes [80, 81]. Sanchís-Calleja and colleagues [81] used a previously established device, called MiSTR [82], to induce a DV continuous gradient on neural organoids. The transcriptomes of organoids exposed to a combination of SHH signaling activator (SAG) and WNT pathway antagonist gradients were compared to those of nongraded organoids. Interestingly, their data revealed differences in cell fate along the DV axis between graded or nongraded organoids. In particular, nongraded SHH-derived organoids tended to adopt a more ventral fate, MGE followed by POA and hypothalamic identity when exposed to low and high concentrations of SHH, respectively. In contrast, organoids derived from graded SHH conditions showed a clear pattern in which low SHH concentration led to LGE, cortical, and PSB identity, whereas high SHH concentrations led to MGE fate, similar to low Shh concentration in nongraded organoids. Thus, organoids derived from the nongraded condition tended to adopt a more ventral fate compared to those derived from the graded condition. Similarly, another study investigated the effect of morphogen gradients on cell specification and organoid patterning [80]. They developed a microfluidic system that exposes microfluidic neural spheres to antiparallel gradients of SHH and BMP4 along the DV axis. They also successfully developed patterned forebrain spheroids (FOXG1+), with an organization that follows pallium or subpallium, including a clear separation between NKX2.1 (MGE marker) and CTIP2 (LGE and cortical markers) [64]. Overall, artificial organizers or microfluidic devices induced topographic DV organization and developed both dorsal (cortical) derivatives and subpallial (GE) derivatives. All of these studies highlight the importance of precisely controlling the formation of directed gradients in order to recreate a patterning process that more closely resembles that observed in vivo.

As previously outlined in the second section of this review, the generation of inhibitory neurons introduces an additional layer of complexity beyond regional patterning. The INs composition within organoids was analyzed in several studies using different experimental paradigms. First, the diversity of INs produced by standard ventralization protocols has been extensively characterized. In particular, when combined into assembloids, several interneurons can be found. For example, Birey et al. [75], Bagley et al. [76], and Xiang et al. [73] described the presence of various interneurons subtypes in subpallial organoids, including NPY, REELIN, and PV interneurons, with the predominant subtype being the SST and calbindin MGE-derived interneurons (shown in Fig. 1b). Moreover, Xiang et al. [73] showed that SST interneurons are produced in a SHH dose-dependent manner. Given that all 3 studies used different methods, the subtypes developed, their frequency and their origin vary across the three teams. In these studies, authors fused ventral-derived organoids with dorsal-derived ones, effectively linking the neurons’ birthplace with their destination. Despite differences in protocols for organoid production and fusion, all three teams showed that migrating interneurons exhibited saltatory behavior and integrated into the dorsal region. In line with these studies, it has recently been shown that specific morphogen combinations, concentrations, and time window of exposure impact organoids cellular diversity [83] (shown in Fig. 1b). For instance, low SAG concentration or late application produces MEIS2-derived neurons, whereas greater concentration or early application orient the fate toward NKX2.1-derived INs. They also largely explore how morphogen combinations affect cellular diversity, showing, for example, that dopaminergic neurons (derived from the LGE) are produced in SAG+ and CHIR (WNT pathway activator) and FGF8 conditions.

However, the establishment of an endogenous gradient during the ventralization protocol induces a clear change in IN diversity. Indeed, another group generated GEs derivatives in neural organoids by submitting the organoids to an SHH gradient in a passive diffusion-based device and compared the cell composition of organoids generated from nongraded ventral organoids [12]. This study achieved a full ventral patterning, unlike previous microfluidic studies that also obtain cortical/dorsal-derived neurons. They highlighted that subpallial organoids were mostly enriched in MGE derivatives compared to those derived from gradient conditions, which display more diversity. Graded organoids comprise a broader range of INs subtypes, including those from the LGE and CGE, with a lesser representation of MGE interneurons. As previously mentioned, microfluidic-derived brain organoids have the potential to develop into all GE-derived derivatives and give rise to various INs populations, although their specificities are not discussed in detail in the papers [80, 81]. Finally, using artificial SHH signaling centers as a tool to induce a very localized gradient is also efficient to induce interneuronal diversity (SST, PV, and CR) and promote interneuronal migration in organoids [78].

Significant progress has been made in understanding the generation and diversification of INs. This review highlights key genetic pathways and extracellular cues involved in early patterning of the telencephalon. At an early embryonic stage, ventral progenitor cells from different subpallial domains already exhibit heterogeneity and bias toward the production of specific INs subtypes. However, recent single-cell omics studies have revealed that progenitor diversity within these broad domains is more limited than expected and insufficient to account for the full spectrum of mature INs. Instead, evidence suggests that cell fate commitment begins shortly after cell cycle exit, emphasizing the importance of the combined influence of intrinsic genetic programs and extracellular cues. Consistent with this, lineage reconstruction studies have shown that INs generally exhibit a high degree of divergence, meaning that transcriptomically distinct IN subtypes can arise from common progenitors.

Many questions remain regarding the role and mechanisms by which extracellular cues regulate INs development. Advances in organoid-based models provide a promising platform to study these cues and uncover human-specific mechanisms. While morphogens were initially used only to generate reproducible regionalized neural organoids, the development of new bioengineering methods will undoubtedly be useful to study their role beyond the establishment of DV patterning and to unravel the mechanisms underlying IN diversification. Such advances will not only improve our understanding of normal IN development but also provide insights into disease-related processes. Indeed, neurodevelopmental disorders including epilepsy, schizophrenia, and autism spectrum disorder have been associated with abnormal IN development [84, 85]. A recent study used assembloids of dorsal and ventral organoids to screen for genes associated with neurodevelopmental disorders. This approach already identified key genes critical for the proper migration of cortical interneurons to the developing pallium [86]. Therefore, with the new protocols and paradigms described in this review, showing how the establishment of precise gradients and patterns tends to create better conditions for generating in vivo-like IN diversity, we can only imagine the future discoveries that will be made using such techniques in the field of neurodevelopmental diseases research.

The authors thank Patricia Gaspar for critical reading of the manuscript.

The authors have no conflicts of interest to declare.

Our projects are supported by AVENIR/ATIP, Inserm, the ANR, and NARSAD Young Investigator Grant from the Brain & Behavior Research Foundation. J.F.’s salary is supported by Inserm. T.D.G. and C.B. are both supported by a PhD fellowship of the Ministère de la Recherche. Our institute is supported by Inserm and Sorbonne University.

T.D.G. wrote the introduction, the first two sections, and the conclusion and helped design the figure. C.B. wrote the last two sections and designed the figure. J.F. provided direction for the review, discussed ideas and concepts, and edited the text and figure.