DLX genes and proteins in mammalian forebrain development

The Dlx story began for J.L.R. in 1988 when, as a postdoc in Roland Ciaranello's lab, he launched a screen for developmental regulators in the mouse telencephalon. They devised a phagemid subtractive hybridization method to enrich for cDNAs preferentially expressed in the embryonic telencephalon, the most anterior part of the forebrain, compared with the adult telencephalon (Rubenstein et al., 1990). Called telencephalon-enriched genes (Tes) genes (Rubenstein et al., 1994; Bulfone et al., 2005), the first gene picked at random was called Tes1. J.L.R.’s student Matt Porteus also discovered Tes1 using low stringency hybridization to a cDNA library with a homeobox probe (Porteus et al., 1991). Because the Tes1 mouse homeobox gene was related to Distal-less in Drosophila, Tes1 was renamed Dlx2. In 1991, two other groups identified Dlx1 and Dlx2 (Price et al., 1991; Robinson et al., 1991). Soon thereafter, the full vertebrate Dlx family was discovered in zebrafish, Xenopus, chicken, mouse and human (Panganiban and Rubenstein, 2002).

There are six Dlx genes in mammals (Dlx1, Dlx2, Dlx3, Dlx4, Dlx5 and Dlx6) and more in other vertebrates. The vertebrate Dlx family comprises sets of bi-gene pairs (Fig. 1A). It is likely that a single distal-less (Dlx) gene was part of an old NKL cluster that also contained ancestors for genes of the Msx and Nkx families, among others (Pollard and Holland, 2000). The single Dlx gene would have undergone a duplication accompanied by an inversion to generate an ancestral bi-gene. This might have happened before the appearance of vertebrates, as such an arrangement is found in the urochordate Ciona intestinalis (Di Gregorio et al., 1995). An ancestral Dlx bi-gene then went through two rounds of genome duplication, followed by the loss of one bi-gene, resulting in the three bi-genes now found in mammals. Dlx bi-genes are often structurally linked to the Hox clusters, with Dlx1 and Dlx2 linked to HoxD, Dlx5 and Dlx6 linked to HoxA, and Dlx3 and Dlx4 linked to HoxB (Amores et al., 1998; Zerucha and Ekker, 2000).

Dlx nucleotide and DLX protein sequences are highly conserved across vertebrates, including conserved homeodomain DNA-binding domains (Fig. 1B). Each Dlx gene pair has three exons (all with protein-coding domains), with the homeodomain split between exons 2 and 3 (Fig. 1A). The Drosophila Dll gene has an intron at the identical location within the homeobox (Ellies et al., 1997a,b; Vachon et al., 1992). The Dlx2, Dlx3 and Dlx5 paralogues (defined here as DlxA genes) share a similar sequence, whereas the Dlx1, Dlx4 and Dlx6 paralogues (DlxB genes) are similar to each other. The pairs of Dlx genes have converging transcription (indicated by arrows in Fig. 1A). Dlx genes can have multiple transcripts, either due to alternative transcription initiation [e.g. Dlx1 (McGuinness et al., 1996)] or due to alternative splicing [e.g. Dlx4 (previously called Dlx7) and Dlx5 (Liu et al., 1997; Nakagawa et al., 1996; Yang et al., 1998)]. Alternative Dlx5 transcripts encode proteins both with and without the homeodomain and nuclear localization signal (Liu et al., 1997; Yang et al., 1998). Dlx bi-gene clusters also express antisense transcripts (Fig. 1A), the functions of which are mentioned below (Liu et al., 1997; Feng et al., 2006). The Dlx5/6 locus antisense RNA (known alternatively as Evf2, Dlx6os1 or Dlx6as) is an ultraconserved long non-coding RNA (lncRNA; not shown in Fig. 1A) that binds DLX proteins, forming a repressive complex (containing MeCP2) when interacting with specific enhancers (Feng et al., 2006; Bond et al., 2009).

Dlx genes play dual regulatory roles, activating transcriptional programs necessary for specific developmental programs in the forebrain and other organ systems (Table 1), while also repressing programs associated with alternative fates. The in vivo DLX motif bound by DLX1, DLX2 and DLX5 has been defined using chromatin immunoprecipitation (ChIP) sequencing (ChIP-seq) from nuclei of developing mouse forebrain (Lindtner et al., 2019). Although their consensus TAATTA motif is nearly identical, there is sequence divergence at the start and end of the core region and in flanking sequences. The non-homeodomain sequences of DLXA and DLXB proteins differ (Liu et al., 1997; Panganiban and Rubenstein, 2002).

Here, we first describe the expression and role of Dlx genes during mammalian forebrain development. We then discuss how Dlx genes are regulated at a transcriptional and functional level through their enhancers, by cofactor protein-protein interactions, and through the regulation of target gene transcriptional activity. We integrate these data to propose a gene regulatory network downstream of the Dlx genes in the developing basal ganglia. Finally, we address many of the areas where more work is needed to fully elucidate their story in driving development, evolution and in maintaining health.

A fundamental principle is that the Dlx genes are expressed in forebrain GABAergic regions and cell types. The forebrain has multiple subdivisions, including the cortex, basal ganglia and hypothalamus, which have roles in cognition, sensory perception, movement and the homeostatic control of body and brain physiology (Rubenstein and Puelles, 2004; Rubenstein and Rakic, 2020). Unlike earlier discovered families of homeobox transcription factors (TFs) (e.g. Hox and Engrailed), which are expressed in the spinal cord, hindbrain and midbrain, the expression of the Dlx genes in the central nervous system (CNS) is restricted to the forebrain. Dlx genes are also expressed outside of the CNS in other tissues, summarized in Table 1. In situ RNA hybridization, immunohistochemistry and Dlx enhancer activity analyses have shown that Dlx expression is restricted to two forebrain domains: one telencephalic [subcortical (subpallial)] and one extending from the prethalamus to the rostral end of the hypothalamus (Porteus et al., 1991; Bulfone et al., 1993; Puelles et al., 2021, 2023) (Fig. 2A).

Deployment of Dlx forebrain expression follows a strict temporal pattern of Dlx2, Dlx1, Dlx5 and then Dlx6 (Liu et al., 1997; Eisenstat et al., 1999; Su-Feher et al., 2022). Forebrain expression begins at ∼E9.5: Dlx2 and Dlx1 are expressed in neural progenitors scattered in the ventricular zone (VZ), becoming denser over development. Then, Dlx1, Dlx2, Dlx5 and Dlx6 are co-expressed in most secondary progenitors of subventricular zones (SVZ) 1 and 2. Dlx1- and Dlx6-antisense transcripts are found in the SVZ and neurons in the mantle zone (MZ) (Fig. 2A). Dlx3 and Dlx4 have not been detected in the developing forebrain.

As neurons are generated, migrate and mature, Dlx1, Dlx2, Dlx5 and Dlx6 continue to be expressed in the MZ in many forebrain GABA cell types, including in the striatum (medium spiny neurons and interneurons), the pallidum (globus pallidus), the preoptic area, the septum, and GABAergic nuclei of the amygdala, hypothalamus and prethalamus (Wang et al., 2011a,b; Puelles et al., 2021, 2023) (Fig. 2A). GABAergic neurons are the principal source of neural inhibition and thus play a central role in regulating virtually all brain circuits. Importantly, the Dlx genes are also expressed in GABAergic local circuit neurons in the pallium (olfactory bulb/cortex, neocortex and hippocampus) (Anderson et al., 1997a,b; Stühmer et al., 2002a,b; Cobos et al., 2005; Long et al., 2007). Whereas some cells express all four forebrain Dlx genes, some have cell-type-specific patterns. For example, Dlx5 and Dlx6 are expressed in the central nucleus of the amygdala, whereas Dlx1 and Dlx2 are expressed in the intercalated amygdalae nuclei (Wang et al., 2011a,b). Dlx1 is expressed early in the lineage-generating parvalbumin⁺ cortical interneurons (CINs) and is then downregulated in these cells but maintained in other CINs (Cobos et al., 2005; Pla et al., 2018) (Fig. 2A,B).

Dlx enhancers flank and are embedded within the non-coding regions of the Dlx bi-gene pairs. They are described as ultraconserved because their nucleotide sequences have changed little during vertebrate evolution (Bejerano et al., 2004). For example, within the Dlx1-Dlx2 locus, there are at least four enhancers: URE1, URE2 (upstream regulator element) both 5′ of Dlx1, I12b (intergenic), and an enhancer just 5′ of Dlx2 (Thomas et al., 2000; Ghanem et al., 2007) (Fig. 1A). The Dlx5-Dlx6 locus also has intergenic enhancers (I56i and I56ii) (Zerucha et al., 2000; Stühmer et al., 2002a,b) (Fig. 1A).

In transgenic mice, these enhancers have robust and reproducible activity that matches endogenous Dlx RNA and protein expression in the forebrain, especially in the ganglionic eminences (Ghanem et al., 2007; Zerucha et al., 2000; Stühmer et al., 2002a,b). URE2 is remarkable for its activity in the VZ; I12b and I56i are striking because their activity is maintained in maturing cortical interneurons and can be used as enhancers for Cre-Drivers (Stenman et al., 2003; Potter et al., 2009), as well as for lentiviral and AAV vectors that drive expression in Dlx-expressing cell types (GABAergic interneurons) in the cerebral cortex of several mammalian species (Arguello et al., 2013; Lee et al., 2014; Dimidschstein et al., 2016; Mich et al., 2021).

Genome sequencing of autistic individuals has identified a point mutation in the I56i enhancer that alters a homeodomain motif (Hamilton et al., 2005) and decreases the enhancer's activity in a transgenic assay (Poitras et al., 2010). Such results reflect the importance of regulatory element mutations in human disease, as well as their known role in evolution.

Deleting Dlx1/2 (i.e. I12b) and Dlx5/6 (i.e. I56i, I56ii) mouse enhancers (Fig. 1) has identified specific and redundant functions in homozygotes and compound mutants. I56ii deletion reduces Dlx5/6 expression, as well as Ebf1, Isl1 and Meis2 expression between the lateral ganglionic eminence (LGE) and medial ganglionic eminence (MGE) (Fazel-Darbandi et al., 2016). I56i deletion, or mice with an I56i point mutation in a homeodomain motif identified in an individual with autism (Hamilton et al., 2005; Poitras et al., 2010), exhibit reduced Dlx5/6 expression and transient reduction of LGE Gad65 expression (Fazel-Darbandi et al., 2021). The deletion of I12b reduces Dlx1/2 expression but otherwise lacks a clear phenotype. On the other hand, I56i^–/–, I12b^–/– mutants show a >2-fold reduction in Dlx1, Dlx2, Dlx5 and Dlx6 RNA, have increased sociability and a reduced response in fear conditioning (Fazel-Darbandi et al., 2021). Thus, modest Dlx expression changes during the development of forebrain GABAergic neurons can have important ramifications for behavior.

DLX1, DLX2 and DLX5 all bind to the Dlx5/6 (I56i, I56ii) enhancers, consistent with Dlx1/2 regulation of Dlx5 and Dlx6 (Long et al., 2009a); it is not known whether DLX5 autoregulates its expression. Analysis of Ascl1, Gsx2 and Otx2 TF mouse mutants show that these TFs promote Dlx expression in the developing telencephalon (Long et al., 2009a; Wang et al., 2013; Hoch et al., 2015). Our TF ChIP-seq supports the hypothesis for direct regulation of GSX2 and OTX2 on Dlx1/2 (URE2, I12b) and ASCL1, GSX2 and OTX2 on Dlx5/6 (I56i) enhancers (Catta-Preta et al., 2023 preprint). Furthermore, analyses of protein-DNA interactions provide evidence of a direct role for ASCL1 in activating the I12b Dlx1/2 enhancer (Poitras et al., 2007).

Beginning in 1995, a series of papers described how loss-of-function deletions in the Dlx genes alter mouse development, including several general findings in the forebrain (for studies in other tissues see Table 1). First, Dlx1 and Dlx2 induce Dlx5 and Dlx6; once Dlx5 and Dlx6 expression is established, Dlx1 and Dlx2 are not required to maintain their expression (Anderson et al., 1997a; Pla et al., 2018). Second, there is partial functional redundancy between the Dlx genes in many cell types; Dlx single mutants have modest phenotypes (e.g. Dlx1^−/−, discussed below), whereas Dlx1/2^−/− and Dlx5/6^−/− mutants have strong forebrain phenotypes. For example, the germline (constitutive) Dlx1/2 mutant lacks Dlx5 and Dlx6 expression and partially blocks the differentiation of GABAergic neurons (in part owing to increased Notch signaling and increased oligodendrogenesis), leading to hypoplasia of the striatum and other parts of the basal ganglia (Anderson et al., 1997a; Yun et al., 2002; Long et al., 2009a,b; Petryniak et al., 2007). The constitutive Dlx1/2 mutant has revealed that the embryonic basal ganglia [e.g. MGE and caudal ganglionic eminence (CGE)] are the source of most, if not all, mouse GABAergic CINs, as well as hippocampal (HINs) and striatal interneurons (SINs). Most of these interneurons fail to migrate to the cortex in Dlx1/2^−/− mutants (Anderson et al., 1997a,b; Pleasure et al., 2000; Marín et al., 2000). In addition, mutations in Dlx genes impair the development of interneurons destined for the olfactory bulb (Bulfone et al., 1998; Long et al., 2007; Long et al., 2007; Guo et al., 2019).

RNA-seq analysis of Dlx1/2^−/− at embryonic day (E)15.5 has identified ∼60 TFs and non-TFs with either increased or decreased expression in the ganglionic eminences (Long et al., 2009a,b). In situ hybridization has been used to define cell types along the VZ-SVZ-MZ differentiation axis where these TFs have expression changes, setting the stage for deciphering Dlx-driven transcriptional pathways. More recently, the use of DLX-ChIP-seq has identified promoters and enhancers bound by DLX proteins (Lindtner et al., 2019; see below).

Dlx1 and Dlx2 strongly regulate a core set of genes (Fig. 3). They promote the expression of Arx, Cxcr4, Cxcr7 (Ackr3), Erbb4, ER81 (Etv1), GAD1, GAD2, Gucy1b3, Meis2, Slc32a1 (vGAT), Six3, Sp8, Tshz1 and Vax1, and repress Ascl1 (Mash1), Gsx2, Hes5, Nr2f1, Olig2 and Otx2 expression (Long et al., 2009a,b; Le et al., 2017). Mutations in many of these genes contribute to defects in GABAergic neuronal migration, differentiation and function (Colombo et al., 2007; Hoch et al., 2015; Flames et al., 2004; Long et al., 2009a,b; Mandal et al., 2013; Su et al., 2022; Wang et al., 2011a,b, 2013; Xu et al., 2018). Dlx1 and Dlx2 repression of Ascl1, Gsx1, Gsx2 and Olig2 has been further investigated by making triple mutants. Eliminating Ascl1 function from the Dlx1/2^−/− mutants amplifies the loss of GABAergic fate (Long et al., 2009a) and the loss of Gsx2 function in Dlx1/2^−/− mutants rescues increased Notch signaling and oligodendrogenesis (Wang et al., 2013). Likewise, eliminating Olig2 function from Dlx1/2^−/− mutants rescues the increased oligodendrogenesis observed in Dlx1/2^−/− double mutants (Petryniak et al., 2007). Furthermore, Dlx1 and Dlx2 directly promote Zfhx1b (Zeb2) TF expression; Zfhx1b mutants phenocopy the Dlx1/2^−/− CIN defect. Zfhx1b functions in part by repressing Nkx2-1 in immature CINs, which is required to direct their migration to the cortex and away from the striatum (McKinsey et al., 2013).

Postnatally, Dlx1 and Dlx1/2 promote GABA production and packaging (through Gad1, Gad2 and vGat), dendritogenesis and synaptogenesis. Dlx1/2^−/− mutants have electrophysiological defects including reduced excitatory drive (Pla et al., 2018). Perhaps secondarily to these problems, postnatal mutant mice have CIN apoptosis and develop epilepsy (Cobos et al., 2005; Jones et al., 2011). Further to the CIN dysfunction, there are also defects in circuit function in the auditory and visual cortex (Mao et al., 2012; Seybold et al., 2012). Remarkably, transplantation of wild-type CIN into Dlx1^−/− partially rescues their electrophysiology status (Howard et al., 2014), further strengthening the evidence that proper interneuron numbers and function are crucial to achieve an appropriate excitatory/inhibitory balance (Merzenich and Rubenstein, 2003; Sohal and Rubenstein, 2019).

Less is known about the function of Dlx5 and Dlx6 in the forebrain compared with Dlx1 and Dlx2. Consistent with their later expression than Dlx1/2 (see above), Dlx5 and Dlx6 are implicated in promoting differentiation by repressing cell division at the G₁/S checkpoint (MacKenzie et al., 2019).

Dlx5/6^−/− constitutive mutants have exencephaly (due to their craniofacial functions), complicating the interpretation of their forebrain phenotypes. Although early basal ganglia differentiation showed no definitive phenotype, the tangential migration of MGE-derived CINs appeared to be attenuated, perhaps owing to reduced Cxcr4 expression (Wang et al., 2010). MGE transplantation suggested a reduction in parvalbumin⁺ CINs in Dlx5^−/−and Dlx5/6^−/− mutants. Consistent with this, Dlx5/6^+/− adult mice have defects in fast-spiking parvalbumin⁺ CINs, which are associated with defects in gamma rhythms and cognitive function (Cho et al., 2015). Dlx6^−/− mutants have subtle RNA expression defects in basal ganglia nuclei (striatum, nucleus accumbens and central nucleus of the amygdala) (Wang et al., 2011a,b).

Conditional deletion of Dlx5 and Dlx6 using Vgat-Cre (active in the entire CNS) results in a variety of behavioral phenotypes (e.g. hypervocalization and hypersocialization, anxiety, locomotion and nest building) and physiological phenotypes (reduced weight and extended lifespan) (de Lombares et al., 2019; Levi et al., 2021, 2022; Aouci et al., 2022). These are thought to reflect Dlx5 and Dlx6 forebrain GABAergic functions, including in cortical parvalbumin⁺ interneurons.

The interactome of DLX proteins is just beginning to be understood. Co-immunoprecipitation analyses have not provided evidence that the DLX paralogues bind each other. On the other hand, the DLX proteins interact with the nucleosome remodeling and deacetylase (NuRD) complex and the BAF complex (Cajigas et al., 2015, 2021; Price et al., 2022). A mass spectroscopy screen has provided evidence that DLX1 binds to RBBP4, a component of the NuRD chromatin remodeling complex (Price et al., 2022). In vitro and in vivo analyses have found that both RBBP4 and RBBP7 interact with DLX amino acids (AA) adjacent to the homeodomain that are part of the nuclear localization signal (Price et al., 2022). DLX1, RBBP4 and RBBP7, as well as other members of the NuRD complex (e.g. CHD4, HDAC1, HDAC2, MBD3), are found at an overlapping set of genomic loci using ChIP-seq in E13.5 mouse subpallium (Price et al., 2022). Furthermore, RBBP4 or RBBP7 shows the highest correlation with DLX1 and DLX2 bound loci, and they share the TAATT DLX homeodomain motif, providing evidence that DLX guides the DLX/RBBP complex to specific chromosomal loci. DLX-RBBP co-bound loci are largely intronic or intergenic and rarely at promoters. These loci are associated with genes with an altered expression in Dlx1/2^−/− mutants: increased (e.g. Otx2, Sall3) or decreased (e.g. Arx, Sp8). Rbbp7 conditional null mice (using Nkx2-1-Cre to delete in the MGE), in a Rbbp4^+/− background, phenocopy aspects of Dlx1/2^−/− mutants, including reduced output of CIN from the MGE (Price et al., 2022).

Mass spectrometry of proteins from the embryonic mouse forebrain and biochemical analyses have provided evidence that the Evf2-DLX1 ribonucleoprotein complex also contains the SWI/SNF-related chromatin remodelers Brahma-related gene 1 (BRG1; SMARCA4), Brahma-associated factor (BAF170; SMARCC2) and SOX2 (Cajigas et al., 2015, 2021). The N-terminal region of DLX5 interacts with the adaptor protein MAGE-D1 and with NDN (Ju et al., 2017). A yeast two-hybrid screen provided evidence that GRIP1, a PDZ cytoplasmic protein, interacts with DLX2 and promotes the activity of DLX2 and DLX5, but not DLX1 (Yu et al., 2001). Going forward, a more comprehensive understanding of the DLX protein-protein interactions and functions will be essential to fully understand how these TFs regulate transcription.

The DLX target homeobox motif features a core TAATT sequence flanked by additional preferred bases, which was identified via comparative genomics and functional studies (Feledy et al., 1999; Zerucha et al., 2000). DLX ChIP-seq has since identified in vivo DLX-genomic interactions (Zhou et al., 2004; Lindtner et al., 2019). DLX1, DLX2 and DLX5 ChIP-seq of the E11.5, E13.5 and E16.5 mouse ganglionic eminences have shown that the three proteins bound very similar loci with TAATTA core motifs (Lindtner et al., 2019). DLX TFs bind to both promoters and distal regions in embryonic mouse basal ganglia, with the majority (∼80%) of binding sites at presumed distal enhancer elements. Motifs for other TFs are also present at these loci and most correspond to TF families that function in the developing basal ganglia.

To build a detailed genomic perspective of how DLX drives forebrain development, we have applied DLX ChIP-seq paired with characterization of the epigenomic and transcriptional state in wild-type and Dlx1/2^−/− mutant mice (Lindtner et al., 2019; Catta-Preta et al., 2023 preprint). Dlx1/2^−/− exhibits both increased expression of some genes and decreased expression of others, based on RNA expression array, RNA-seq and histochemical analyses (Long et al., 2009a,b; Lindtner et al., 2019), consistent with DLX TFs having dual context-dependent functionality of driving transcriptional activation or repression. To explore genomic correlates of this dichotomy, we have performed histone ChIP-seq for histone modifications associated with gene activation (H3K27ac, H3K4me1, H3K4me3) or gene repression (H3K27me3) (Lindtner et al., 2019) in wild-type and Dlx1/2^−/− mice. This approach has revealed a subset of DLX binding events in which the absence of DLX binding results in the loss of activation or repression of the regulatory target genes. DLX-bound enhancers near downregulated genes are marked by a reduced ratio of activating:repressing histones, which have been termed activating regulator elements (a.REs). By contrast, those enhancers near upregulated genes and marked by an increased ratio of activating:repressing histones in the Dlx1/2^−/− mutant mice have been termed repressing regulator elements (r.REs).

Genes with strong expression changes in Dlx1/2^−/− are typically located in genomic intervals with one or several a.REs or r.REs. There is a close correlation of decreased RNA expression with a.REs, and increased RNA expression with r.REs. r.RE targets are enriched for TF genes (e.g. Gsx1, Otp, Pax7), whereas a.RE targets include a combination of TF genes (e.g. Dlx5, Sp8) and genes controlling GABAergic neuron maturation and function (e.g. Gad2, vGAT, Nrxn3). The Arx locus is paradigmatic; it has enhancers with either cortical [hs122 (Rr43), hs123 (Rr44)] or subcortical (hs119 (Rr41), hs121 (Rr3)] activity (Dickel et al., 2018). Although DLX1, DLX2 and DLX5 bind all four enhancers, they only activate transcription from the subcortical enhancers (Lindtner et al., 2019). Overall, a.RE enhancers are enriched for basic Helix-Loop-Helix (bHLH) and DLX-type homeodomain motifs, whereas r.RE enhancers are enriched for Forkhead and OTX/GSC-type homeodomain motifs (Lindtner et al., 2019), indicating that combinatorial binding of DLX and partner TFs encodes activation versus repressive control.

To dissect how combinatorial binding by DLX factors with other TFs drives the spatiotemporal specificity of telencephalic enhancer activity, we have generated TF in vivo binding data (ARX, ASCL1, DLX1, DLX2, DLX5, GSX2, LHX6, NKX2-1, OTX2, PBX1, SP9) and intersected it with in vivo enhancer activity at E11-E12 in transgenic mice. For this study, we have used previously reported subpallial enhancers (Ghanem et al., 2003, 2007; Visel et al., 2013) and novel enhancers that were chosen based on ChIP-seq experiments using the TFs listed above (Catta-Preta et al., 2023 preprint). We found that enhancers most likely to have specific subpallial activity had combinatorial binding of multiple of these TFs, are strongly evolutionarily conserved and are enriched for homeodomain motifs (TAATTA). We also identified several combinatorial TF-enhancer modules that were associated with different neurodevelopmental processes. For example, enhancers bound by DLX2, ARX and LHX6 are associated with GABAergic neuronal maturation genes, whereas DLX2, GSX2 and OTX2 combinatorial binding is found at enhancers that had activity in neural progenitors in the VZ. The molecular interactions that determine how combinatorial binding confers specificity remain unknown, but the DLX proteins have a central role in this process.

By integrating transcriptomic and epigenomic data from the Dlx mutants, and cis-regulatory data from enhancer assays and in vivo DLX genomic binding, we are making in-roads to elucidate a gene regulatory network upstream and downstream of the Dlx genes.

Genomic studies of DLX regulatory function revealed several generalized features: DLX proteins bind directly to a shared set of regulatory targets that feature TAATTA consensus or degenerative motifs; DLX regulatory target genes with altered expression in Dlx1/2^−/− mutants feature clustered DLX binding and deeply evolutionarily conserved enhancers with multiple TAATTA motifs; DLX TFs repress alternative developmental transcriptional programs and activate subpallial neurogenic fates via context-dependent combinatorial binding with other developmentally-active TFs. Given that the DLX proteins are expressed first in neuronal progenitors and then in postmitotic neurons, and that DLX interactions at a.RE and r.RE are crucial for activation and repression of fate-specification genes, DLX factors appear to play a uniquely central role in initiating and maintaining forebrain (GABAergic) neuronal fate properties from development through adulthood. Integrating the transcriptional changes in the Dlx mutants, with the identification of candidate enhancers and promoters for these genes, has enabled the postulation of Dlx-regulated gene regulatory networks in progenitors (VZ, SVZ) and neurons (MZ) of the developing basal ganglia (Lindtner et al., 2019). When coupled with similar approaches for the TFs upstream and downstream of Dlx, one can assemble a more complete view of the transcriptional circuits; these pathways are important for predicting the effect of disease-causing mutations.

Going forward, much more needs to be known about DLX biochemistry and the functions of the non-homeodomain regions, such as the DLX2 C-terminal poly-histidine tract (Porteus et al., 1991), and the DLX6 CAG/CCG (poly-glutamine/poly-proline) repeat region in exon 1 (Pfeffer et al., 2001) (Fig. 1B). The DLXA proteins (DLX2, DLX3, DLX5; Fig. 1B) also contain a region of about 20 AA that is well conserved across species (Akimenko et al., 1994; Liu et al., 1997). Given the importance of DLX combinatorial binding for driving enhancer specificity, structural analyses are needed to elucidate: (1) the organization of the TF complexes with DNA, transcriptional cofactors and perhaps RNAs; (2) how these relate to activation versus repression; (3) how these relate to the stabilization/formation of 3D interactions and their nuclear localization. A better understanding of the non-homeodomain protein structures of DLX TFs should also provide insights into the redundant and distinct roles of co-expressed DLX bi-gene pairs and the differences between the DlxA and DlxB paralog groups.

There is also a lack of understanding regarding the selectivity of DLX targets considering the common occurrence of the TAATTA motif in the genome and regarding the functional relevance of DLX TF binding events. We found that DLX genomic binding in the forebrain extends to many loci beyond Dlx1/2^−/− knockout-responsive a.RE/r.RE elements. Currently, nothing is known about the functions of loci bound by DLX but that show no phenotype in Dlx1/2^−/− mutants – perhaps they prefigure functions in other cell states or they are evolutionary vestiges/inventions. As work on the DLX TF genomic targets is currently limited to the forebrain, comparative studies of DLX genomic binding in other tissues are needed and will help answer questions regarding context-dependent DLX TF regulatory function and conserved or divergent regulatory programming across relevant mammalian organ systems and cell types.

This article is part of the collection ‘40 years of the homeobox’. See related articles in this collection at https://journals.biologists.com/dev/collection/10249/40-years-of-the-homeobox.