Domain-specific regulation of foxP2 CNS expression by lef1
© Bonkowsky et al; licensee BioMed Central Ltd. 2008
Received: 07 August 2008
Accepted: 24 October 2008
Published: 24 October 2008
Skip to main content
© Bonkowsky et al; licensee BioMed Central Ltd. 2008
Received: 07 August 2008
Accepted: 24 October 2008
Published: 24 October 2008
FOXP2 is a forkhead transcription factor critical for normal development of language in humans, but little is known of its broader function and regulation during central nervous system (CNS) development. We report here that lef1, a member of the Lef/Tcf family of transcription factors activated by Wnt signaling, regulates foxP2 during embryogenesis, and we isolate novel foxP2 enhancers which are lef1-dependent.
Loss, knock down, or inhibition of lef1 led to loss of foxP2 expression. We isolated DNA fragments from the foxP2 genomic region that function as enhancers to drive GFP expression in the CNS during development, including in the telencephalon, diencephalon, eye, tectum, and hindbrain. Three of these enhancers, foxP2-enhancerA.1, foxP2-enhancerB, and foxP2-enhancerD, contain putative Lef1 binding sites, and are regulated by lef1. However, two other genomic fragments containing Lef1 sites failed to function in vivo as enhancers. Chromatin immunoprecipitation confirmed that Lef1 binds to sites in foxP2-enhancerA.1 and foxP2-enhancerB.
This work shows that lef1 is necessary for expression of foxP2 in the tectum, mid-hindbrain boundary, and hindbrain during CNS development, and is the first insight into the upstream regulation of foxP2 during development. We also demonstrate that in silico prediction of potential lef1 binding sites poorly predicts their ability to function in vivo as enhancers. The foxP2 enhancers we identified will allow dissection of foxP2's role during CNS development.
FOXP2 is a forkhead domain transcription factor whose mutation has been associated with severe deficits in language [1–6]. Its cloning and expression during development have been described in humans, mouse, songbird, frog, medaka, and zebrafish [7–16]. CNS expression during development in the different vertebrate species is remarkably similar, with conserved expression in the telencephalon, basal ganglia, thalamus, tectum, tegmentum, cerebellum, and hindbrain. However, the function of FOXP2 during CNS development is poorly understood. Homozygous knockout or point mutation of Foxp2 in mice leads to early postnatal death, with reports of disordered Purkinje cell layers [17–19] and smaller cerebellar size [18, 20]. In songbirds, FoxP2 appears to be necessary for vocal learning and is expressed in neurons during active song learning [9, 21, 22].
The neural circuits and genetic cascades in which FOXP2 participates remain uncharacterized. Chromatin immunoprecipitation methods have identified potential downstream targets of FOXP2 [23, 24], but in vivo function and importance of the identified targets is uncertain. To understand in greater detail the role of foxP2 in CNS development, we sought to identify how foxP2 expression is regulated. The conservation of six predicted lef1 binding sites between pufferfish, zebrafish, mouse, and human in the foxP2 genomic region (this study; , and the overlapping expression of lef1 and foxP2 in the zebrafish CNS during development, led us to consider whether lef1 might regulate foxP2. Lef1 is a transcription factor activated by the canonical Wnt/β-catenin signaling pathway, which has been shown to play a critical role in proliferation, tissue patterning, CNS neuronal cell fate specification, and axon pathfinding . We found that loss or knockdown of lef1 led to a loss of foxP2 expression in the tectum, mid-hindbrain boundary, and hindbrain. Of six conserved potential lef1 binding sites predicted in the foxP2 genomic region , we show that only three lie in genomic fragments that function in vivo as enhancers, underscoring the importance of in vivo testing of predicted enhancers. Using chromatin immunoprecipitation (ChIP), we demonstrated that lef1 can bind directly to the functional enhancer sites, and showed that in the absence of lef1 these enhancers fail to function. The foxP2 enhancers will be useful for dissection of foxP2 function by allowing detailed analysis of axon pathfinding and synaptogenesis in foxP2-expressing neurons.
To confirm that loss of foxP2 expression was due to knockdown of lef1 and not a non-specific morpholino effect, we utilized two alternative means to remove lef1 expression. First, we examined expression of foxP2 in Df(LG01)x8 mutant embryos. Df(LG01)x8 is a deletion on chromosome 1 which contains lef1 (but not foxP2); homozygous deficiency mutants do not express lef1 [27, 29]. We found that homozygous Df(LG01)x8 mutants do not express foxP2 in the tectum, MHB, or hindbrain (Figure 3C) (100%, n = 25 embryos).
Second, we examined whether a dominant negative construct which inhibits lef1 function would cause a loss of foxP2 expression. We used embryos carrying the transgene hsp70l:Δtcf-GFP, which expresses an N-terminal deletion of Tcf3a fused to GFP. In the absence of its N-terminal DNA-binding domain, Tcf3a acts as a dominant repressor of Wnt-mediated transcription . Following heat-shock at 32 hpf for 1 hour, embryos were collected 3 hours post-heat shock. There was loss of foxP2 expression in the tectum, MHB, and hindbrain (Figure 3D) in 61% of transgenic embryos (n = 38), and in 0% of non-transgenic siblings (n = 82). Decrease of foxP2 telencephalic expression is presumably due to a dominant effect of the transgene, since the decrease was not seen using the morpholino or Df(LG01)x8.
The loss of foxP2 expression in the tectum and hindbrain is not secondary to an absence of cells, as expression of the tectal marker mbx  is indistinguishable between wild type and lef1 morphants (Figure 3E, F). Other markers for the tectum (emx2), and hindbrain (isl1, zash1a) of lef1 morphants also appear as wild type patterns and levels ; J.E.L. and R.I.D., unpublished data). Furthermore, the loss of foxP2 expression persists at 48 hpf (Additional File 1). These results show that loss of lef1 specifically causes loss of foxP2 expression in the tectum, MHB, and hindbrain.
Because the foxP2 genomic assembly is incomplete in the region upstream of the 2nd coding exon (J.L.B., unpublished data), we sought to identify the sixth predicted lef1 binding site. In human FOXP2 this site is 8.8 kb upstream of the first coding exon. We obtained the sequence for this region in zebrafish from the unassembled BAC DKEY-116L11. Using the enhancer element locator algorithm , we compared the 9 kb regions immediately upstream of the first coding exon of foxP2 from the human and zebrafish genomes. We found conservation of this same element (zebrafish: ttgtgggctGCTTTCATCtgtgggttaa; human: atgatcagtGCTTTCATCtttattttaa) located 8.5 kb upstream of the first coding exon in zebrafish (contained within foxP2-enhancerA).
foxP2-enhancerA drove expression in the eye, diencephalon, tectum, hindbrain, and telencephalon (Figure 5A, B). GFP expression was first noted in the telencephalon at 24 hpf, then at 48 hpf in the eye, tectum and hindbrain, becoming maximal at 72 hpf. By 80 hpf expression in the CNS diminished significantly, while jaw expression became visible. Subcloning of foxP2-enhancerA to yield enhancerA.1 and A.2 revealed very similar patterns to foxP2-enhancerA (Figure 5C–F). However, enhancerA.2 had significantly fewer labeled cells, suggesting that the more proximal region of enhancerA and enhancerA.1 contains necessary elements. foxP2-enhancerB expression in the telencephalon began at 24 hpf, with eye and hindbrain expression apparent by 72 hpf (Figure 5G, H). foxP2-enhancerD expression in the eye, dorsal diencephalon, and tectum began at 36 hpf, was maximal at 48 hpf, and decreased by 72 hpf (Figure 5I, J).
The enhancers we have identified partially mirror endogenous foxP2 expression . While the enhancer fragments were chosen based on in silico prediction of potential lef1 binding sites , our analysis shows the importance of in vivo validation, since only 3 of the 6 predicted binding sites appear to lie in functional enhancer elements for the stages analyzed.
To demonstrate that loss of GFP expression in the lef1 morphants was not simply due to absence of the GFP-expressing cells, we injected lef1 morpholino into transgenic fish lines Tg(elavl3:EGFP) zf8 (also known as HuC:GFP) and Tg(pax2a:GFP) e1 . Tg(elavl3:EGFP) zf8 expresses GFP in all post-mitotic neurons , while Tg(pax2a:GFP) e1 expresses GFP in a pattern mirroring pax2a expression, including in the MHB, cerebellum, and hindbrain . In both lines we found that lef1 morphants still expressed GFP in the tectum, MHB, and hindbrain (Figure 6G, H, and Additional File 2). In addition, using Tg(isl3:GFP)zc7 to label retinal axons (A. Pittman and C.B.C., unpublished), we observed that the pattern of retinotectal projections appeared normal in 48 hpf morphants (data not shown). Although overall numbers of GFP-expressing cells appeared reduced in lef1 morphants of Tg(elavl3:EGFP) zf8 and Tg(pax2a:GFP) e1 , in the foxP2 enhancer lines there was a complete lack of GFP-expressing cells in the tectum and hindbrain. These results show that loss of tectal and hindbrain expression observed with the lef1 morpholino in foxP2-enhancerA.1, foxP2-enhancerB, and foxP2-enhancerD is not simply due to a loss of neurons, but instead indicates a requirement for lef1 in these enhancers' function.
Our data show that lef1 and foxP2 expression overlap in the developing CNS, and that lef1 is necessary for foxP2 expression in the tectum, MHB, and hindbrain. We identified Lef1binding sites in the foxP2 regulatory regions, based on the identification of foxP2 enhancers that depend on lef1 for expression, and ChIP analysis showing binding of Lef1 to these enhancers.
Since foxP2 expression in the telencephalon is not lef1-dependent, other factors must regulate foxP2. Conversely, in regions where lef1 is expressed but foxP2 is not, for example the dorsal diencephalon, lack of foxP2 expression could either reflect an absence of some necessary co-factors, or the presence of an inhibitor of expression, or both mechanisms. We have tried to globally activate foxP2 by heat-shock induction of a constitutively active form of tcf3 (RID, unpublished data), but did not observe any ectopic foxP2 expression (data not shown), implying stringent, multifactorial control of foxP2 expression.
Previous computer algorithm-based methods identified several other potential Lef1 binding sites in the foxP2 genomic region , but our in vivo analysis failed to support a role for them. Another recently described in silico method to identify transcription factor target genes would also fail to identify the lef1 enhancers for foxP2, as this method only included the 5 kb regions immediately upstream of transcriptional start sites . Our work shows the importance of actual in vivo testing for analysis of enhancer gene regulation.
Since we analyzed expression up to 96 hours after fertilization, enhancers responsible for controlling expression at later times might have been missed. Given that all of the CNS domains of expression for foxP2 are present by 72 hpf [7, 11], we feel that this is unlikely. However, we can not exclude the possibility that untested genomic regions might regulate expression, or that several independent regions might function in concert.
This study identified three regions functioning as enhancers for foxP2, specifically driving expression in the telencephalon, eye, diencephalon, tectum, and hindbrain. Their expression reveals that foxP2 is expressed in multiple brain regions, under strict temporal and spatial control, regulated through both lef1-dependent and -independent mechanisms. These transgenic lines label multiple distinct axon tracts (data not shown), and will allow analysis of subsets of foxP2 neurons and axons, in both wild type and mutant backgrounds, as well as misexpression and rescue experiments.
The role of foxP2 in CNS development has remained elusive, despite multiple genetic studies in humans demonstrating its necessity for normal language development [1–6]. In affected members of the human KE family, who carry a heterozygous point mutation which disrupts the function of one copy of FOXP2, voxel-based morphometric MRI analyses suggest disturbances in Broca's area, the basal ganglia, and the cerebellum [38, 39]. Studies of FoxP2 knockout or point mutation mice have consistently shown cerebellar involvement. Homozygotes have been reported to have disordered Purkinje cell layering [17–19, 40] and smaller cerebellar size [18, 20], while heterozygotes are noted to display impaired motor learning and altered Purkinje cell synaptic plasticity [19, 20]. Our identification of lef1 as a regulator of foxP2 expression in the MHB and hindbrain is a first step towards understanding the foxP2 genetic network involved in cerebellar development. The role of the cerebellum in language function is at least partially understood , but it is uncertain whether the oromotor apraxia of the KE family is due to defects of cerebellar or striatal pathways (or both) [40, 42, 43]. Importantly, other domains of foxP2 expression, for example in the telencephalon (including the subpallium), appear to be regulated independently of lef1.
Wnt signaling has multiple roles in neuronal specification and the development of connectivity in the CNS , as does lef1, which is activated by the canonical Wnt signaling cascade. lef1 is necessary for development of dentate gyrus neurons , for neurogenesis and specification of neuronal subsets in the hypothalamus , and for expression of the transcription factors zic2a and zic5 in the tectum . Here we show that lef1 regulates foxP2 expression in the tectum, MHB, and hindbrain. Interestingly, ChIP studies suggest that FOXP2 may in turn regulate components of the Wnt signaling pathway [23, 24]. The identification of components of the foxP2 signaling cascade, including upstream, interacting, and downstream members, will be important for understanding foxP2 function, and for elucidating the genes and neural circuits involved in language development.
Adult fish were bred according to standard methods. Embryos were raised at 28.5°C in E3 embryo medium with 0.003% phenylthiourea to inhibit pigment formation and staged by time and morphology . For in situ staining, embryos were fixed in 4% paraformaldehyde (PFA) (in PBS) for 3 h at room temperature (RT) or overnight (O/N) at 4°C, washed briefly in PBS, dehydrated, and stored in 100% MeOH at -20°C until use.
Transgenic fish lines and alleles used were as follows: Df(LG01)x8 ; Tg(hsp70l:Δtcf-GFP) w26 ; Tg(foxP2-enhancerA:EGFP) zc42 ; Tg(foxP2-enhancerA.1:EGFP) zc44 ; Tg(foxP2-enhancerA.2:EGFP) zc46 ; Tg(foxP2-enhancerB:EGFP) zc41 ; Tg(foxP2-enhancerD:EGFP) zc47 ; Tg(pax2a:GFP) e1 ; Tg(elavl3:EGFP) zf8 . Df(LG01)x8 homozygotes were identified by their smaller forebrain and flattened hindbrain phenotype (J.E.L. and R.I.D., unpublished). Tg(hsp70l:Δtcf-GFP) w26 embryos were identified by GFP expression after heat shock.
Heat shock was performed by incubation of 32 hpf embryos for 1 hour at 37°C, then collecting 3 hours after the end of the heat shock.
Whole-mount in situ labeling for foxP2, gfp, and lef1 was performed as previously described [7, 27, 47]; the mbx RNA probe derived from full-length mbx cDNA cloned into pBluescript . Double in situs were performed using a DNP-labeled probe for foxP2 and digoxigenin-labeled probe for lef1. Following standard hybridization and washes, DNP probe was detected using an anti-DNP HRP conjugate diluted 1:200 in TNTB block at 4°C overnight (PerkinElmer; ), followed by detection using Alexa Fluor 488 Tyramide diluted 1:250 (Molecular Probes) with 0.0015% hydrogen peroxide for 1 hour. Embryos were washed in TNT, blocked in TNTB, incubated with anti-digoxigenin alkaline phosphatase diluted 1:5000 in TNTB at 4°C overnight, and then detected with a standard BM Purple color reaction.
PCR primers to clone foxP2 genomic fragments were as follows (forward and reverse primers, sequences 5' to 3'; size in kb listed immediately following the enhancer name): foxP2-enhancerA (9.7 kb): FP2.12L GTCGTAATTGCTCGGTGAC, FP2.3R GTGTGAATGCCAGCGATAGA; foxP2-enhancerA.1 (6.8 kb): FP2.12L, FP2.16R CGTCTCGACTGAGCAGAGTT; foxP2-enhancerA.2 (5.1 kb): FP2.12L, FP2.46R ACAACTGGCGTGTAAGGTGT; foxP2-fragment A.3 (4.7 kb): FP2.41L GACACCTTACACGCCAGTTG, FP2.3R; foxP2-fragment A.4 (2.3 kb): FP2.41L, FP2.40R CAGGGTGTGTTATAAACATGCAT; foxP2-enhancerB (4.6 kb): FP2.39L GACACTCTGGAGGAACTATG, FP2.38R GGAAACGGTGCAGTATGTGT; foxP2-fragment C (1.6 kb): F.FP1 GGCGGGTACCTGGTCATATT, F.RP2 TTTCCACCCAACCATAAATCA; foxP2-enhancerD (1.2 kb): H.FP1 CCAGCTATCCGAGAGGTTCA, H.RP2 CCGCCTGTTCAAATCAGAAT; foxP2-fragment E (0.69 kb): G.FP1 TGACCTCTGTGTAGCCTTGC, G.RP2 CATTGCTAGGGGAACGTGAT.
PCR was performed using standard conditions (TaKaRa LA PCR amplification kit 2.1, Millipore) from total genomic DNA, and PCR fragments gel purified prior to cloning (Qiagen gel purification kit). BP and LR reactions (Gateway cloning system, Invitrogen) were performed to clone the DNA fragments upstream of an adenovirus E1b minimal promoter and carp β-actin transcriptional start fused to EGFP (pENTRbasEgfp) in a Tol2 plasmid backbone (pTolR4-R2) [32, 33, 49]. BP reactions were performed by adding attB4 sequence to the 5' primer (5'-GGGGACAACTTTGTATAGAAAAGTTG-gene specific primer-3'), and attB1 sequence to the 3' primer (5'-GGGGACTGCTTTTTTGTACAAACTTG-gene specific primer-3'). The identity of the genomic fragments was confirmed by restriction enzyme digests and partial sequencing.
Injection of DNA constructs and raising of stable transgenic lines was performed essentially as described [33, 49]. 20 pg of each enhancer:EGFP-Tol2 construct was co-injected with 20 pg of Tol2 transposase RNA in a total volume of 1 nL at the 1-cell stage. Embryos were screened for GFP expression from 12 hpf through 96 hpf. Patterns of enhancer expression were confirmed by multiple independent transient injections of the plasmid (> 5 injections for all constructs, > 100 embryos per injection), as well as isolation of 2 or more independent stable transgenic lines (in cases where stable transgenics were isolated).
The lef1 and tcf3b splice-blocking morpholinos (MO) were synthesized by Gene Tools (Philomath, OR); efficacy and use were as previously described [27, 28, 50, 51]. For lef1, 2 ng of morpholino E7I7 was injected; for tcf3b, 5 ng was injected.
Whole-mount images were taken using brightfield microscopy with embryos mounted in 80% glycerol. Confocal microscope images of live embryos were taken after mounting embryos anesthetized using tricaine (0.004%) in 1.5% low melt agarose dissolved in 0.33× PBS/7.3% glycerol. Image acquisition and analysis were performed as described previously . Confocal images of double in situs were taken on an Olympus FV1000 after mounting embryos in 80% glycerol. BM Purple fluorescence was imaged with 633 nm excitation, collecting emission from 700–800 nm .
We used coordinate conversion ("convert" function) in the UCSC genome web server http://genome.ucsc.edu/goldenPath/help/hgTracksHelp.html#Convert to identify conserved lef1 binding sites in the foxP2 region of zebrafish, based on predicted Tcf4 binding sites positions in the human genome (since Tcf4 and Lef1 bind to a shared motif) . The genomic upstream sequence of foxP2 was obtained by performing a BLAST search against unfinished sequences in the Sanger Centre Danio rerio sequencing project (sequence for foxP2-enhancerA is contained in BAC DKEY-116L11; GenBank accession CU468733). To identify the sixth potential Lef1 binding site in the foxP2-enhancerA region predicted by Hallikas et al. , we used the enhancer element locator algorithm (available at http://www.cs.helsinki.fi/u/kpalin/EEL/, and compared the 9 kb regions upstream of the first coding exon of human and zebrafish FOXP2 using the Tcf4 consensus binding motif.
ChIP was performed as previously described [27, 55] with the following modifications. 80–120 embryos of each genotype (wild type or homozygous Df(LG01)x8) were collected between 28 hpf and 30 hpf, dechorionated, and fixed in 2.2% formaldehyde for 15 minutes at room temperature. Embryos were rinsed in 0.125 M glycine, followed by PBS, and then lysed. The Df(LG01)x8 lysate was checked by genomic PCR for the lef1 gene to confirm that no PCR product was obtained. PCR products were visualized on ethidium bromide-stained agarose gels. Primers used for ChIP analysis of the foxP2-enhancerA.1 genomic region were the following (forward and reverse primers, sequences 5' to 3'): FP300, CGACTCTCCGCGAAACAC, CATTGCCATTATCACCAGCA; FP900, CCTATCAAAGGCAGGCAGA, TTATCGCACTAAACAAGCTATTACAC; FP1800, CACAGAGTGAAGTCATGGAGAAA, AGCGCACGCACAACTAATC; FP2700, AGAGAAACGGATAAACAGTGAGAAG, CCCTCTCGAACCCTCAAAA; B3end, TGATTAACGCCGTCTTTTCC, AGCGCACGCACAATAAAATA. Primers used for ChIP analysis of the foxP2-enhancerB genomic region were (forward and reverse primers, sequences 5' to 3'): FPb200, AGCTGGAAGGAAGTGTCTGG, TTTTGGCATGTGCAAAGAAG; FPb1500, GCGTATGTATGCTTGTCAGGTT, TGCGTGGTGTTTTACTTGGA; FPb2600, CAGATCGACGGATGATACACA, TTCCGCCAATTATCATGTCA; FPb3800, GACCCCTTCGCAATGTCTAAT, TTCGAGAAATGCTTGCACAC.
central nervous system
We thank members of the Chien and Dorsky labs for their assistance. This work was supported by a PCMC Foundation grant and CHRC grant to JLB, NIH K12 HD001410 and K08 DA024753 to JLB, and NIH R01 NS053897 to RID.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.