An evolutionarily conserved Lhx2-Ldb1 interaction regulates the acquisition of hippocampal cell fate and regional identity

Transcription factors act in macromolecular complexes comprising multiple interacting factors wherein individual components play unique structural and biochemical roles to locally modify the chromatin landscape. Although great progress has been made in identifying individual molecules that participate in a given complex, few complexes have been directly tested for identified molecular interactions and functional consequences. One elegant and well-characterized molecular assembly is the ‘tetrameric model’ for LIM-homeodomain (LIM-HD) transcription factor function, which was first demonstrated for LIM-HD protein Apterous (Ap) in the context of its role as a ‘dorsal selector’ in the Drosophila wing blade (van Meyel et al., 1999). This study demonstrated that it is necessary for Ap to bind with co-factor Chip/Ldb1 and for Chip to dimerize, thus bringing together two Ap molecules. This active tetramer could be replaced by a chimeric dimer that contained the HD of Ap and the dimerization domain (DD) of Chip, and this was sufficient to restore the normal patterning of the wing blade (Milan and Cohen, 1999; van Meyel et al., 1999).

The interaction between LDB and LIM-HD proteins seen in Drosophila is evolutionarily conserved (Thaler et al., 2002), but has thus far not been examined in the mammalian brain. The vertebrate ortholog of Ap, Lhx2, is the only LIM-HD gene to be expressed in the proliferating progenitors of the entire cortical neuroepithelium from the earliest stages of corticogenesis (Retaux et al., 1999; Bulchand et al., 2001). Lhx2 functions as a ‘cortical selector’, such that constitutive loss of Lhx2 causes loss of the entire cortical primordium and expansion of two flanking structures: the hem and the antihem (Bulchand et al., 2001; Mangale et al., 2008). In mosaic experiments, patches of Lhx2-null cells interspersed in wild-type medial cortical primordium differentiate into hem (Mangale et al., 2008). Each patch of hem functions as a ‘hippocampal organizer’, and induces hippocampal fate in adjacent Lhx2-expressing tissue (Mangale et al., 2008; Godbole et al., 2018). These studies demonstrate that Lhx2 represses hem fate in the medial telencephalon and, consistent with this, the wild-type embryonic hippocampal primordium expresses a high level of Lhx2, whereas the adjacent hem shows no detectable expression (Bulchand et al., 2001). This raises an important issue about whether the requirement for Lhx2 in the hippocampal primordium is limited to preventing this tissue from itself differentiating into hem. It remains untested whether Lhx2 functions within presumptive hippocampal cells to execute the instructive cues and direct hippocampal fate in this tissue. At later embryonic stages, when hippocampal neurogenesis is under way, conditional loss of Lhx2 in hippocampal progenitors induces premature astrogliogenesis (Subramanian et al., 2011). The molecular mechanism of Lhx2 function in this process remains unclear and the tetrameric model has never been examined with respect to these functions.

In vertebrates, Chip has two homologs, both of which are expressed in the developing mammalian cerebral cortex. Of these, Ldb1 is ubiquitously expressed, and Ldb2 has a limited expression and is absent from the telencephalic ventricular zone where progenitors reside (Bach et al., 1997; Bulchand et al., 2003). Cortex-specific loss of Ldb1 results in an extremely shrunken hippocampus, whereas loss of Ldb2 has no detectable effect on hippocampal development (Leone et al., 2017). In the current study, we found that loss of Ldb1 from embryonic day (E) 10.5 results in a profound loss of the hippocampal primordium from the earliest stages of development, and also an upregulation of GFAP, suggesting enhanced gliogenesis. We tested whether an Ldb1:Lhx2 complex is crucial for these processes by designing a chimeric molecule that comprises the HD of Lhx2 and the DD of Ldb1, modeling our construct based on the seminal studies performed in the Drosophila wing and the vertebrate spinal cord (van Meyel et al., 1999; Thaler et al., 2002). Using in utero electroporation, we demonstrate that this chimeric molecule not only rescues the appearance of hippocampal tissue that expresses field-specific markers, but is also sufficient to suppress astrogliogenesis and promote neurogenesis in an Ldb1 loss of function background. Although Ldb1 binding partners are not limited to the LIM-HD family (Ramain et al., 2000; Bronstein et al., 2010), this construct selectively restores Lhx2-Ldb1 function in the Ldb1 mutant. Our results demonstrate an elegant conservation of the Lhx2:Ldb1 molecular toolkit that is required for fundamental aspects of development in the brain, and provides a function for the intense expression of both genes in the early hippocampal primordium.

Lhx2 and Ldb1 are both expressed in the dorsal telencephalon at E10.5 in neuroepithelium that will form the hippocampus and neocortex (Fig. 1A,B). Lhx2 expression excludes the hem, which forms at the telencephalic midline (arrowhead, Fig. 1A). This pattern of expression is maintained through subsequent stages, such that at E12.5, the entire dorsal telencephalic neuroepithelium except the hem expresses Lhx2 (Bulchand et al., 2003), and this expression persists upon cortex-specific deletion of Ldb1 using Emx1Cre (Fig. 1B, Fig. S1). Within this Lhx2-expressing territory, the neuroepithelium of the hippocampal primordium expresses Ephb1 in control brains (white asterisk, Fig. 1B). This expression is greatly reduced or undetectable when either Lhx2 or Ldb1 are deleted using Emx1Cre (arrowheads, Fig. 1B). As the hippocampus does not form in the absence of Wnt signaling from the hem (Lee et al., 2000), we examined the expression of Wnt3a and Wnt2b in Emx1Cre;Ldb1^lox/lox embryos. The expression of both these Wnt genes was apparently normal in the hem of Ldb1 mutant brains at E12.5 (Fig. 1C). Therefore, the near-complete lack of specification of the hippocampal primordium in the Ldb1 conditional mutants could not be attributed either to the lack of Wnt gene expression in the hem, or lack of Lhx2 expression in the hippocampal primordium.

The lack of hippocampal specification in both the Ldb1 and Lhx2 mutants, together with the ability of Lhx2 to bind to Ldb1 in vitro (Agulnick et al., 1996) and in vivo (Gueta et al., 2016; de Melo et al., 2018) is consistent with the well-established tetrameric model of LIM-HD (Lhx2/Ap) and LDB (Ldb1/Chip) protein interaction (Milan and Cohen, 1999; van Meyel et al., 1999) illustrated in Fig. 1E. Immunoprecipitation using anti-Lhx2 antibody pulls down Ldb1, and vice versa (Gueta et al., 2016; de Melo et al., 2018). Anti-Lhx2 antibody similarly pulls down Ldb1 from E15 hippocampal tissue (Fig. 1D).

Two constructs were used by Milan and Cohen (1999), and van Meyel et al. (1999) to provide key data for the tetrameric model; we used the mouse counterparts of these same constructs to test Lhx2:Ldb1 interactions in the hippocampus. The first was a truncated Ldb1 construct lacking the dimerization domain, which is capable of sequestering Lhx2 (Ldb1ΔDD; Fig. 1F). For the present study, we designed a second construct similar to the chimeric construct used by Milan and Cohen (1999), and van Meyel et al. (1999). This Lhx2ΔLIM-Ldb1ΔLID construct encodes a protein in which the LIM domains of Lhx2 are deleted, and the remaining Lhx2-HD containing fragment is fused to a region of the Ldb1 molecule containing the Ldb1-dimerization domain (Fig. 1F,G). We tested whether this chimeric construct can rescue the severely shrunken hippocampus in the cortex-specific Ldb1 conditional mutant (Emx1Cre; Ldb1^lox/lox).

The chimeric construct was electroporated into the medial telencephalon of Emx1Cre; Ldb1^lox/lox embryos at E13-E14, when hippocampal neurogenesis is under way. The embryos and the brains were examined at postnatal day 0 (P0; Fig. 2). At this stage, the pan-hippocampal marker Zbtb20, as well as hippocampal field-specific markers KA1 (GRIK4) (field CA3) and Neurod1 (CA3 and DG), are robustly expressed in control brains (Fig. 2A). In contrast, the non-electroporated hemisphere revealed barely detectable expression and an extremely reduced hippocampus (Fig. 2B). The electroporated half of the same sections displayed a somewhat larger hippocampus that expressed each marker in a patchy manner (Fig. 2C,D). When compared with the corresponding GFP fluorescence image of the same region, it was apparent that the patches of rescued expression indeed correspond well with regions that had been electroporated with the chimeric construct (Fig. 2D-F).

It is noteworthy that multiple region- and field- specific features of the hippocampus, which are lost in the absence of Ldb1, are rescued in a region-appropriate manner by the chimeric construct. Whereas expression of the pan-hippocampal marker Zbtb20 appears throughout the extent of the electroporated region (ovals in Fig. 2D,E), Neurod1 and KA1 expression remains undetectable in the CA1 region, even though electroporated cells are present (compare boxes in corresponding images of Fig. 2D,E). Therefore, the introduction of the chimeric construct does not induce widespread upregulation of hippocampal regional markers; rather, it appears to allow the electroporated cells to read out specification cues in a region-specific manner. This reinforces the interpretation that the chimeric construct substitutes for an endogenous mechanism rendered inoperable by the loss of Ldb1.

Previously, we demonstrated that Lhx2 regulates the neuron-glia cell fate switch in hippocampal progenitors. Loss of Lhx2 during hippocampal neurogenesis causes premature astrogliogenesis, and overexpression of Lhx2 during the period of astrogliogenesis prolongs neurogenesis (Subramanian et al., 2011). As the chimeric Lhx2ΔLIM-Ldb1ΔLID construct is capable of restoring the formation of cells with specific hippocampal field identities, we further sought to test whether it is also capable of restoring neurogenesis in the absence of either Lhx2 or Ldb1.

We used the dominant-negative construct Ldb1ΔDD, in which the dimerization domain of Ldb1 was deleted (Fig. 1H), to sequester Lhx2 in E14.5 hippocampal progenitors. Previously, we have shown that the protein encoded by this construct does indeed bind the LIM-domains of Lhx2 (Ldb1ΔDD is called ClimΔDD; Subramanian et al., 2011, Fig. S1). We further demonstrated that electroporation of this construct causes a premature gliogenic effect indistinguishable from that seen upon electroporation of Cre-GFP into Lhx2^lox/lox embryos (Subramanian et al., 2011). Thus, Ldb1ΔDD recapitulates the functions of its Drosophila counterpart ChipΔDD, which binds and sequesters Apterous, but cannot form a functionally active tetrameric complex (Bach et al., 1999; van Meyel et al., 1999; Becker et al., 2002). The Drosophila counterpart of the chimeric Lhx2ΔLIM-Ldb1ΔLID construct also effectively substitutes for loss of Ap and rescues Ap mutant phenotypes in Drosophila (van Meyel et al., 1999). We therefore used Ldb1ΔDD and the chimeric Lhx2ΔLIM-Ldb1ΔLID construct to further test the tetrameric model in the mouse hippocampus.

In our previous studies, we used GFAP as a marker of astrocytes derived from the hippocampal primordium, and showed that E15 hippocampal progenitors produce either β-tubulin-expressing neurons or GFAP-expressing astrocytes in the 5-7 day window after electroporation in vivo (Subramanian et al., 2011) and in vitro (Muralidharan et al., 2017a). We further demonstrated that the GFAP-expressing cells produced by loss of Lhx2 also express astrocyte marker AldoC, but do not express oligodendrocyte marker Olig2 (Subramanian et al., 2011). Therefore, in the present study, we proceeded with using GFAP as a validated marker for astrocytes.

We performed in utero electroporation in wild-type embryos at E14.5, examined the brains 7 days later at P2, and scored for the expression of GFAP in the electroporated cells (Fig. 3B). Electroporation of a control GFP construct revealed a baseline gliogenesis of 27%. Electroporation of the chimeric construct alone resulted in 20% gliogenesis, which is not significantly different from the baseline and was expected as endogenous Lhx2 expression is already very high at E14.5 (Bulchand et al., 2003; Subramanian et al., 2011); hence, a construct designed to mimic Lhx2 function would not be expected to have a significant additive effect. The dominant-negative Ldb1ΔDD construct caused significantly enhanced gliogenesis (61%), as seen by Subramanian et al. (2011). This was suppressed by co-electroporation of the chimeric construct to 32%, similar to baseline levels, thus effectively restoring the percentage gliogenesis to control levels (Fig. 3C). These results suggest that the chimeric Lhx2ΔLIM-Ldb1ΔLID construct can preserve Lhx2-like functionality even when Lhx2 is sequestered.

Loss of Lhx2 causes premature astrogliogenesis and hence upregulation of GFAP expression in the hippocampus (Subramanian et al., 2011). We examined GFAP expression in Emx1Cre; Ldb1^lox/lox brains at P0, and found apparently enhanced GFAP expression in the rudimentary hippocampus (Fig. 4A,B). We established a baseline for the percentage astrogliogenesis at E14.5 in these mutant brains by electroporating GFP at this age and scoring the fraction of GFP+GFAP co-expressing cells at P0. Upon electroporation of a control GFP construct, 43% of the GFP-positive cells in the shrunken Ldb1 mutant hippocampus were seen to express GFAP (Fig. 4C,D). Electroporation of the chimeric construct suppressed astrogliogenesis to 6% (Fig. 4D-F). Furthermore, electroporated cells were seen to produce axons that reach the rostral level of the hippocampal commissure, confirming that these cells are indeed neurons (arrowhead, Fig. 4F). Together, our results confirm a rescue of hippocampal neurogenesis and a suppression of astrogliogenesis in the cortex-specific Ldb1 mutant. In summary, these results demonstrate that the chimeric Lhx2ΔLIM-Ldb1ΔLID construct restores hippocampal neurogenesis in the absence of Ldb1.

The present study was prompted by the striking similarity in the hippocampal phenotype upon cortex-specific deletion of Ldb1 and Lhx2. In both mutants, the medial telencephalic neuroepithelium corresponding to the hippocampal primordium is shrunken, and displays limited expression of hippocampal markers, indicating a loss or lack of specification of this structure. The central finding of our study is that the introduction of a chimeric construct containing the Lhx2-HD and the Ldb1-DD is sufficient to restore field specification as well as neurogenesis in the developing hippocampus. These findings take on a particular significance in light of the fact that Lhx2 expression is not lacking in the Ldb1 mutants, supporting the interpretation that the transcription factor is unable to function in the absence of its co-factor Ldb1. Further studies will seek to identify other members of the multi-protein complex that are likely crucial for Lhx2-Ldb1 function (Muralidharan et al., 2017b), taking advantage of the fact that the chimeric construct, in effectively substituting for both Lhx2 and Ldb1, may help narrow down a limited set of key binding partners.

Lhx2 and Ldb1 are normally expressed in the entire hippocampal progenitor zone, but field-specific features arise in distinct regions of the hippocampal primordium, consistent with the interpretation that the electroporated progenitors may be responding to a morphogen gradient from the hippocampal organizer, the hem, which we found to express apparently normal patterns of Wnt2b and Wnt3a in the cortex-specific Ldb1 mutant. Together, the data support the interpretation that while the Lhx2:Ldb1 complex is not sufficient to impart specific hippocampal field identites, it is necessary to execute the inductive influence of the hippocampal organizer. This extends the established roles of Lhx2 and Ldb1 function.

The fact that Lhx2, a mammalian ortholog of Ap, functions in a molecular complex with Ldb1 in the brain in a manner that is indistinguishable from what has been established for Ap function in the Drosophila wing (Milan and Cohen, 1999; van Meyel et al., 1999), indicates an exquisite conservation of this evolutionarily ancient mechanism in the developing hippocampus. These results underscore the importance of this particular protein complex as a fundamental toolkit that has been reused in diverse organisms and systems ranging from invertebrates to mammals.

All animal protocols were approved by the Institutional Animal Ethics Committee (Tata Institute of Fundamental Research, Mumbai, India) according to regulations devised by the Committee for the Purpose of Control and Supervision of Experiments on Animals (CPCSEA), India. The mouse lines used in this study were kind gifts from Yangu Zhao (Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health; Ldb1^lox/lox; Zhao et al., 2007) and from Edwin Monuki (University of California, Irvine, USA; Lhx2^lox/lox; Mangale et al., 2008). The Emx1Cre line was from Jackson Laboratories [strain name: B6.Cg-Emx1tm1(cre)Krj/J; stock number: 005628]. Noon of the day of the vaginal plug was designated as embryonic day 0.5 (E0.5). Control embryos used were littermates that did not have the Cre allele.

These procedures were performed as described previously (Subramanian et al., 2011). For the experiments in Figs 3 and 4, sections were immunostained using anti-GFP antibody (Abcam, ab6658, 1:200) and anti-GFAP antibody (Sigma, G9269, 1:200). Probes used were kind gifts from Forbes D. Porter (Lhx2; Heritable Disorders Branch, NIH, USA), Elizabeth Grove (EphB1 and NeuroD; University of Chicago, IL, USA), Jim Boulter (KA1; University of California, USA), Clifton Ragsdale (Wnt3a and Wnt2b; University of Chicago, IL, USA), Jakob Nielsen (Zbtb20; University of Southern Denmark, Denmark) and Yangu Zhao (Ldb1). All in situ hybridization experiments shown in Fig. 1 were performed in three embryos of each genotype. For the data in Figs 2-4, three embryos were examined for each condition.

The immunoprecipitation was performed using E15.5 hippocampal tissue lysate. Goat anti-Lhx2 (Santa Cruz, SC 19344; 1 μg per 500 μg protein) was used for immunoprecipitation and western blot was performed using anti-Ldb1 antibody (Santa Cruz, SC 11198, 1:250).

Medial tissue from control and Emx1 Cre; Ldb1^lox/lox was harvested at E12.5 and RNA was extracted using the RNeasy Micro Kit (Qiagen, 74004). cDNA was prepared using SuperScript IV First-strand Synthesis System (Invitrogen, 18091050) followed by real time quantitative PCR using Kapa SYBR Fast qPCR kit (Kapa Biosystems, KR0389_S) and the Light Cycler 96 (Roche). The following primers were used to assess Lhx2 expression: primer set 1: fwd, 5′ TCTGACCGCTACTACCTGCT-3′, rev, 5′ GGGAGGGGCTGTAGTAGTCT-3′; and primer set 2: fwd, 5′ GCTGAACACCTGGATCGTGA-3′, rev, 5′ACCAGACCTGGAGGACTCTC-3′.

The chimeric construct Lhx2ΔLIM-Ldb1ΔLID was cloned into the pEF1-Myc-His A vector. This chimeric construct was designed based on (van Meyel et al., 1999). DNA-encoding amino acids 1-300 of Ldb1 (Ldb1/Clim2 variant3 {NM_010697}) and amino acids 173-406 of Lhx2 were used for the chimeric fusion protein. For electroporation, this pEF1-Lhx2ΔLIM-Ldb1ΔLID-Myc-His chimeric construct was mixed with one encoding pCAG-IRES2-EGFP in a 1:1 ratio for detection of the electroporated cells. The pCAG-Ldb1ΔDD-IRES2-EGFP construct (Ldb1ΔDD/ClimΔDD) was cloned as described in (Subramanian et al., 2011).

Confocal images were acquired using the Olympus FV1200, Zeiss 510 and Zeiss LSM 5 Exciter - AxioImager M1 imaging systems. For calculating the percentage gliogenesis, three different electroporated brains were scored for each condition. The­ numbers of GFP+GFAP double-positive cells were expressed as a percentage of the total GFP cells scored in each brain (Subramanian et al., 2011). Statistical analysis was performed using one-way ANOVA (Tukey's multiple comparison) for Fig. 3C. For Fig. 4D, unpaired two-tailed Student t-test was used. In Fig. 3, the total cells scored per condition were: 290 (GFP), 298 (chimeric construct+GFP), 285 (Ldb1ΔDD) and 305 (chimeric construct+Ldb1ΔDD). In Fig. 4, the total cells scored per condition were: 284 (GFP) and 412 (chimeric construct+GFP).

We thank Edwin S. Monuki for the kind gift of the Lhx2^lox/lox line, and Yangu Zhao, Paul Love and Sue McConnell for the Ldb1^lox/lox line. We also thank Elizabeth Grove (EphB1, Lhx9 and NeuroD), Yangu Zhao (Ldb1), Jim Boulter (KA1), Clifton Ragsdale (Wnt3a and Wnt2b), Jakob Nielsen (Zbtb20) and Forbes D. Porter (Lhx2) for gifts of plasmid DNA; Shital Suryavanshi and the animal house staff of the Tata Institute for Fundamental Research (TIFR) for excellent support; and Anindita Sarkar and Lakshmi Subramanian for critical input on the manuscript. We gratefully acknowledge the support and mentorship of Medha Rajadhyaksha (V.K.).