Distinct Cis Regulatory Elements Govern the Expression of TAG1 in Embryonic Sensory Ganglia and Spinal Cord

Cell fate commitment of spinal progenitor neurons is initiated by long-range, midline-derived, morphogens that regulate an array of transcription factors that, in turn, act sequentially or in parallel to control neuronal differentiation. Included among these are transcription factors that regulate the expression of receptors for guidance cues, thereby determining axonal trajectories. The Ig/FNIII superfamily molecules TAG1/Axonin1/CNTN2 (TAG1) and Neurofascin (Nfasc) are co-expressed in numerous neuronal cell types in the CNS and PNS – for example motor, DRG and interneurons - both promote neurite outgrowth and both are required for the architecture and function of nodes of Ranvier. The genes encoding TAG1 and Nfasc are adjacent in the genome, an arrangement which is evolutionarily conserved. To study the transcriptional network that governs TAG1 and Nfasc expression in spinal motor and commissural neurons, we set out to identify cis elements that regulate their expression. Two evolutionarily conserved DNA modules, one located between the Nfasc and TAG1 genes and the second directly 5′ to the first exon and encompassing the first intron of TAG1, were identified that direct complementary expression to the CNS and PNS, respectively, of the embryonic hindbrain and spinal cord. Sequential deletions and point mutations of the CNS enhancer element revealed a 130bp element containing three conserved E-boxes required for motor neuron expression. In combination, these two elements appear to recapitulate a major part of the pattern of TAG1 expression in the embryonic nervous system.


Introduction
Neurons extend their axons toward intermediate and final targets with remarkable precision. The neuronal fate of neurons and their subsequent axonal trajectory are governed by transcription factors (TFs) [1,2]. Gene networks that specify neuronal wiring in the vertebrate CNS are being actively unraveled, notably in spinal motor and interneurons [3,4,5]. In the spinal cord, classical surgical manipulations of the early embryo demonstrated that motor neurons are programmed to innervate their corresponding muscular target as they differentiate [6]. In the lumbar and brachial motor neurons, the choice to innervate the dorsal vs. the ventral limb muscles is encoded by an Lim-HD code: Lhx1 and Isl1 are expressed in LMCl and LMCm neurons, respectively, and are required to direct ventral versus dorsal limb innervation. This target selection by LMC neurons is manifest by the downstream activation of Eph receptors: EphA4 by Lhx1 and EphB receptors by Isl1 [3,4,7]. However, it is not known whether either Lhx1 or Isl1 activates Eph receptor expression directly. Subsequent innervation of specific muscles is similarly governed by the Hox code [8], but the downstream targets of the Hox TFs, presumably genes encoding receptors that decode the musclederived guidance cues, remain thus far elusive.
The choice of spinal interneurons, whether to project axons ipsilaterally or contralaterally, is regulated on a cellular level at several choice points. Ipsilaterally projecting neurons turn longitudinally at several dorso-ventral levels: dI3d in the dorsal funiculus (DF), dI1i in the lateral funiculus (LF), dI3v in the ventral lateral funiculus (VLF) and V2 in the ventral funiculus (VF) [9,10,11]. The axons of commissural interneurons are guided towards the floor plate (FP) by FP-derived attractants: Netrin, Shh and VEGF, and are desensitized to floor plate-derived repulsive molecules (Slits) via the expression of Robo3 [12]. In vitro perturbation and loss of function experiments in the chick embryo support a role for the interactions between axonal TAG1/Axonin1 and midline-expressed NrCAM in regulating the entry into the floor plate [13,14]. Slits and their axonal receptors, Robo1 and Robo2, regulate the exit from the floor plate [15,16].
The transcriptional programs that direct the ipsi-versus contralateral axonal projection are starting to unravel. Ipsilateral projection of retinal ganglion cells requires the expression of the zinc finger transcription factor Zic2 [17]. In the spinal cord the ipsilateral projection of dI1i spinal interneurons requires the expression of BarhL2 [18], and Isl1 is implicated in the ipsilateral turn of dI3v axons [19]. For the commissural projection, the Lim- Figure 1. Proximal TAG1 enhancer drives expression in embryonic peripheral nervous system. A. Scheme showing organization of mouse TAG1 gene 7 kb upstream and 8.5 kb downstream of exon 1, which is non-coding (green triangle). Includes exon2, which encodes the start methionine and leader sequence (green and black box) and exons 3-5 (black boxes) which are separated from exon 2 by a 4 kb intron, only part of which is shown. Below are indicated constructs used in this study, which include human TAG1 genomic DNA (blue line) fused in-frame at the ATG codon to a LacZ reporter gene (light blue). B. Expression of pTGM-3M in line TAG/LacZii [26] at ages indicated, as detected by X-gal staining. Expression is first evident at E9.5 in the 5 th (V) and 7 th (VII) cranial ganglia and rostral migrating neural crest (arrow). In addition, at E11.5 staining can be seen in the ganglia of several cranial nerves, including the 10 th (arrow), and in spinal dorsal root ganglia (arrowheads), but not in the 12 th nerve (black asterisk; inset shows staining with anti-TAG1 (green) at an equivalent stage in wild type embryo, white asterisk marks 12 th nerve in ventral neural tube). By E13.5, most sensory subsets are seen to express the transgene. C. Section of E11.5 TAG/LacZii embryo confirms X-gal staining in CNS is due to afferents entering hindbrain from trigeminal (arrowheads) and that there are no cells expressing the transgene within the hindbrain neuroepithelium. X-gal staining is also seen in axons extending to jaw (arrow). D,E. Similarly, X-gal staining in spinal cord is due to axons entering the dorsal funiculus (DF) from the DRG and no staining can be seen in the spinal neuroepithelium, either in a dorsal view of a wholemount stained E11.5 embryo (D), or in a transverse section at the same age (E). Note particularly the absence of staining in the ventral root (small arrows), in the commissural axon pathway (large arrow) or under the floor plate (fp). Arrowheads in D indicate peripheral branches of sensory axons. F. A transient transgenic embryo at E11.5 carrying the pTGM-prom construct. Expression overall is weaker, though consistent with that of pTGM-3M transgenicsshowing for example expression in the trigeminal (tg) -but complicated by frequent, inconsistent ectopic expression, present in this example in the limb (ec). White arrow shows X-gal staining in axons of trigeminal. doi:10.1371/journal.pone.0057960.g001 HD transcription factors Lhx2 and Lhx9 were demonstrated to be required for floor plate crossing of dI1c axons [5].
How do TFs regulate the responsiveness to midline cues? Ipsilaterally projecting retinal ganglion cells are repelled by optic chiasm-derived EphrinB and Zic2 positively regulates the expression of the EphB1 receptor [17]. In dI1c neurons, Lhx2 and Lhx9 are required for the expression of Robo3, which in turn is required to suppress activation of Robo1 and Robo2 by Slit, which otherwise repel these axons from midline crossing [5]. The regulatory network that instructs commissural projection of dI1c axons is probably mediated by direct activation since binding sites for Lhx2 and Lhx9 are located in the presumed regulatory elements of Robo1, 2, and 3 [5,20]. The ipsilateral projection of dI1i axons requires BarHL2 to downregulate Robo3 expression and in its absence dI1i axons ectopically express Robo3 and aberrantly cross the midline [18].
In the present study we have begun to elucidate the genetic networks that control the expression of TAG1 in the embryonic spinal cord and hindbrain levels. We have identified two evolutionarily conserved DNA enhancer elements, one located between the Nfasc and TAG1 genes and the second directly 59 to the first exon and encompassing the first intron of TAG1. These elements directed complementary expression in the CNS and PNS, respectively, of the embryonic hindbrain and spinal cord. The CNS enhancer was further characterized utilizing genome alignment and sequential deletion. An evolutionarily conserved minimal, 130 bp DNA element upstream to the TAG1 gene, that is sufficient to drive expression in motor and subset of commissural neurons was identified. Mutagenesis analysis of putative binding sites for bHLH and Lim-HD TFs, revealed that bHLH binding sites -3 E-boxes harbored by the 130 bp element -are required for its expression in motor neurons.

Results
Sequences immediately upstream of the TAG1 gene harbor dorsal root and cranial ganglia-specific control elements of TAG1 expression TAG1/Axonin1/CNTN2 (here referred to as TAG1) is expressed in neurons of the central and peripheral nervous system [13,21,22]. During early embryonic development of rodents and aves, TAG1 is expressed in commissural neurons, motor neurons, cranial ganglia and DRG neurons [13,22,23,24]. However, the temporal profile of TAG1 expression in the spinal cord versus DRG neurons, both in vivo and in vitro, is dissimilar, suggesting differential regulation by central and peripheral neurons [25]. To identify the regulatory elements and decode the gene networks that govern TAG1 expression in the CNS and PNS, we aimed to identify the cis elements that control the spatial expression of TAG1 in the embryonic nervous system.
An ,15 kb genomic fragment extending from ,4 kb upstream of exon 1 (non-coding) through to the ATG start site in exon 2 of human TAG1, has been shown to drive expression of a LacZ reporter gene (fused in-frame immediately downstream of the ATG start site; Fig. 1A) in the postnatal cerebellum of transgenic mice in a manner similar to endogenous TAG1 [26], but its embryonic expression was not reported. Denaxa et al., 2003 [27] reported that a 4kb fragment of human genomic DNA, apparently lying directly upstream of exon 1, is sufficient to largely recapitulate the TAG1 expression pattern, including expression in the embryonic spinal cord, when placed next to a LacZ reporter gene. However, the reporter gene expression driven by this fragment was considerably weaker and more variable than endogenous TAG1 expression, and expression in motor neurons and dorsal root ganglia (DRG), among other tissues, was not seen, leading the authors to conclude that a number of regulatory elements were missing. Therefore, since the transgenic construct (pTGM-3M; Fig. 1A) used by Bizzoca et al (2003) utilizes both the natural transcription and translation start sites of TAG1, and encompasses a substantial proportion of sequences upstream to the transcription start site and all of intron 1, which is known in the genes of other members of TAG1 family to contain critical regulatory elements [28,29], we examined its expression in the embryonic spinal cord of our established transgenic line [26]. Strong reporter gene expression was detected as early as E9.5 in the 5 th (trigeminal) and 7 th (facial) cranial ganglia (Fig. 1B). By E11.5 expression was also evident in the ganglia of several additional cranial nerves, including the vagus (10 th ; arrow) -but not in the hypoglossal (12 th ; asterisk) or oculomotor nerves. Beta-galactosidase (ß-gal) from reporter expression was also evident in the axons of the trigeminal ganglia, including the mandibular, maxillary and ocular branches, and also in its projections into the hindbrain (Fig. 1B, C). More caudally, reporter expression was seen in neurons in the dorsal root ganglia and ß-gal was present in both their peripheral and centrally projecting axons (Fig. 1B, D, E). However, there was no evidence of reporter gene expression in neurons within the spinal cord, the only labelling within the CNS being of axons extending in from the periphery (Fig. 1D, E). Similar staining was found at the level of the trigeminal ganglia, with only incoming peripheral axons labelling within the hindbrain itself ( Fig. 1C arrowheads). Analysis of a second permanent line and of transient transgenic embryos also made with pTGM-3M, indicated that this expression pattern was typical.
A similar pattern was obtained with a reporter construct lacking most of intron 1 (pTGM-prom), but expression was consistently weaker, sometimes labelling only the trigeminal, and ectopic expression was often found (Fig. 1F). A construct including just 1.5 kb upstream of exon 1 (pTGM-min), although partly recapitulating the overall pTGM-3M pattern, consistently revealed weaker and frequently ectopic expression (not shown). Notably, all the three constructs contain the 59 region of the 1 st intron, that is conserved between mammals and the marsupial -opossum (Fig. S1). However, with none of the three constructs studied was there evidence of expression by cells located within the spinal cord, suggesting that the element(s) controlling commissural and motor neuron transcription of TAG1 are positioned elsewhere in the genome TAG-1 and Nfasc genes are closely linked and overlap in their embryonic spinal cord expression To identify additional potential TAG1 cis regulatory elements, we focused on sequences located between the TGM-3M region and the 39 end of the gene lying centromeric to TAG1, Neurofascin (Nfasc). Neurofascin is a transmembrane-linked adhesion molecule closely related to TAG1 in structure [30]. The expression pattern in the spinal cord of both molecules is similar [31], However the relative expression of Nfasc and TAG1 has not been reported. Because of the proximity of the two genes (the distance between the 39 end of Nfasc and the 59 of TAG1 on mouse and human chromosomes 1 is 25 kb and 20 kb, respectively, and just 2.7 kb on chick chromosome 26), we speculated that they may share regulatory elements. Consistent with this, there is substantial overlap between the two proteins in the early chick spinal cord ( Fig. 2). At HH24 (Hamburger and Hamilton stages), TAG1 is expressed in DRG axons, and in subsets of motor axons and interneurons. Expression in commissural spinal interneurons is evident in axons that project diagonally toward the floor plate and in the axons that elongate at the midline (arrows, Fig. 2A). Nfasc is expressed in the same neurons, although its expression in motor and interneurons and their axons is more widespread (Fig. 2B). Interestingly, TAG1 expression is observed on pre-crossing commissural axons, while Nfasc expression is predominant on commissural axons that are crossing the floor plate ( Fig. 2A,B), suggesting that the translation or the membrane localization of the proteins is subjected to different modes of regulation. Similar expression patterns are maintained at HH26 and HH28, where TAG1 is clearly expressed in a subpopulation of Nfasc expressing neurons ( Fig. 2E,F,I,J). Notably, TAG1 expression is restricted to commissural axons while Nfasc is expressed also in ipsilaterally projecting axons (Fig. 2I,J). These observations are therefore consistent with the possibility that a shared regulatory element lies in the Nfasc-TAG1 intergenic region.
An evolutionarily conserved element between TAG1 and Nfasc directs expression to motor and commissural neurons Comparative genomics and complementary transgenic approaches are demonstrated to be a reliable method for identifying enhancer elements in various tissues [32,33,34] including embryonic spinal cord neurons [35,36]. Alignment of the genomic region between TAG1 and Nfasc, in various mammals, utilizing UCSC alignment tools, revealed three conserved regions designated E1, E2 and E3 (Fig. 3A). Each potential enhancer element was isolated from the mouse genome and cloned upstream of Cre recombinase. These Cre plasmids were electroporated into the chick neural tube along with a Cre-dependent GFP plasmid, using the experimental paradigm used previously to identify neuronal specific enhancer elements [9,10,19,36]. In contrast to the E1 and E2 elements, which yielded no expression of the reporter gene, E3 directed expression of GFP specifically in motor neurons, dorsal interneurons and roof plate cells (Fig. 2C,D,G,H,K,L, Fig 3D, table 1).
To further dissect the E3 enhancer element, we broadened our homology search: The homology of E3 between mammals is high along its entire 2.7 Kb. However, alignment with the genome of the marsupial Opossum, revealed two short regions of homology: E3.1 (450 bp) and E3.3 (380 bp) (Fig. 3B,C). Both E3.1 and E3.3 directed expression of GFP in motor and the dorsal interneurons, in a similar pattern to the E3 enhancer element (Fig. 3E,G, table 1). A marked difference between the long E3 and the short E3.1 and E3.3 elements is the labeling of progenitor motor neurons, residing at the ventricular zone and the medial ventral spinal cord, in the short enhancers (Fig. 3E,G), while E3 labeled the differentiated, laterally positioned motor neurons (Fig. 2K,L, Fig. 3D). Accordingly, E3-labeled motor neurons co-express TAG1 (Fig. 1G, S3), while pMN labeled with E3.1 and E3.3 have not initiated the expression of TAG1, but do express Isl1 (Fig. S3). Hence, it is likely that E3 cis elements that inhibit expression in progenitor neurons are absent in E3.1 and E3.3.
The fidelity and specificity of E3.1 enhancer was also assessed in the mouse. The enhancer was cloned upstream of the Hsp68 minimal promoter driving the expression of LacZ reporter gene. At E11.5, 9 out of 11 (82%) embryos transgenic for E3.1-Hsp68-LacZ construct revealed expression of LacZ marker restricted to spinal motor neurons, hindbrain and cervical dorsal commissural neurons (Fig. 3H,I). In contrast to the TGM constructs ( Fig. 1), the E3.1 enhancer did not label the DRG or any of the cranial nerves, with the exception of the hypoglossal ( Fig. 3I; asterisk). Hence, the 450 bp of the E3.1 element harbors an enhancer activity capable of driving expression in motor and commissural neurons, complementary to that seen with the TGM elements.

E3 directs expression predominately in dI1 neurons
By HH28, coincident with the expression of Nfasc (Fig. 2J), E3directed GFP expression (Fig. 2K, L) is found in both commissural (Fig. 2K, white arrows) and ipsilateral axons (Fig. 2K, green arrows), consistent with the idea that E3 is a 39 enhancer of Nfasc. However, an alternative possibility is that the E3 enhancer is active in a progenitor that is common to both contra and ipsi-projecting neurons, but is then silenced in cells that become ipsi-projecting, as is characteristic of dl1 progenitors [5]. In this case, because GFP reporter expression would be permanently activated by Cremediated recombination in the progenitor, it would remain on in both daughter lineages. To distinguish between these two models we set out to compare the subtype of interneurons that express E3 and their axonal projection pattern.
To determine whether the E3 enhancer is able to activate expression in dorsal interneurons with dl1 characteristics, GFP with a nuclear localization signal (nGFP) was electroporated into embryos under the control of E3 and cross sections were then examined with markers that identify different cell fates. Among the post mitotic neurons, 83% of the dorsal interneurons expressing the nGFP are dI1 (as defined by expression of Lhx2/9; 281 Lhx2/ 9 + neurons from 337 nGFP neurons. Neurons were counted from 5 HH26 embryos, 5 cross sections from each embryo), 9% are dI2 (31-Lhx1+/Brn3a+ neurons from 331 nGFP neurons), and 8% of the neurons were distributed between other interneurons (Fig. 4A,B). The preferential expression in dI1 neurons was tested by comparing the soma and axonal labeling of the E3 enhancer element with that of the dI1 specific enhancer element -EdI1. EdI1 directs reporter expression in the two dI1-derived subpopulation: dI1c and dI1i [10]. E3::Cre and EdI1::FLPo were simultaneously electroporated along with two corresponding reporter cassettes: Cre-dependent mCherry and FLP-dependent GFP, respectively (Fig. 4D). All the mCherry expressing dorsal interneurons co expressed GFP, while motor neurons expressed only mCherry (Fig. 4C). GFP+/mCherry-neurons are settled at the ventral lateral spinal cord (Fig. S2), position that is populated with the ipsi lateral dI1 subpopulation -dI1i [5,10]. Collectively these data provide additional evidence that E3 directs expression in the dorsal spinal cord primarily to dI1c neurons. Concomitantly, commissural axons co-expressing GFP and mCherry, and motor axons expressing only mCherry are evidenced (Fig. S2). Interestingly, many of the ipsilaterally axons are GFP+/mCherry- (Fig. S2), proving further support to the commissural expression of the E3 enhancer element.

E3 is a potential target of bHLH transcription factors
Transcription factors that belong to the bHLH and the Lim-HD families determine the cell fate of spinal neurons [1,2,37]. The activity of the E3 enhancer might be stipulated by direct interactions with these transcription factors. The E3 element was screened for potential binding sequences of bHLH and Lim-HD proteins. Four E-boxes, the canonical binding site of bHLH TFs, three of them conserved in mammals and marsupials, and one Lim-HD binding site were found in E3.1 (Fig. 5A), and three Eboxes and one Lim-HD binding sites in E3.3 (not shown). The requirement of the E-boxes and Lim-HD cis element of E3.1 for the expression in the spinal cord was challenged by point mutating the three 39 conserved E-boxes (designated E3.1E3m), all the four E-boxes (E3.1E4 m) or the Lim-HD binding site (E3.1Lm) (Fig. 5A). E3.1Lm directed expression in an identical pattern to the native E3.1 (not shown), suggesting that the Lim-HD cis element is dispensable for the enhancer properties. Consistently, E3 deletion constructs E3.1B and E3.1BB, which do not contain the Lim-HD target, were sufficient to drive expression in motor and dI1 neurons (Fig. 3F, S4), whereas the complementary element E3.1a that contains only the 59 E-box and the Lim-HD cis elements yielded only sporadic, non-specific expression (Fig. S4). By contrast, E3.1E4m and E3.1E3m directed expression of a reporter gene to the dorsal interneurons in the same pattern as the native E3.1. Significantly, however, substantially reduced expression from these elements was detected in motor neurons (Fig. 5B,C, table 1, E3.1E3m not shown). The specificity of E3.1, E3.1B, E3.1BB, E3.1E4m and E3.1A was tested by utilizing the alternate mCherry/GFP cassette [10] that directs ubiquitous expression of mCherry and enhancer-mediated expression of GFP (Fig. 4B,C, S4), and immunohistochemistry for TAG1 (Fig. S3). Thus, the motor neuron expression module of the E3 enhancer appears to require bHLH transcription factor binding sites, whereas Lim-HD protein binding sites are dispensable.

A conserved bHLH target in chick directs expression in motor neurons
Alignment of the genomic region that spans TAG1 and Nfasc, utilizing UCSC alignment tools, revealed that the conservation of TGM-3M and E3 elements is restricted to mammals (Fig. 3A-C,  Fig. S1). However, the mouse E3 enhancer drives expression in motor and dorsal interneurons in chick and mouse. Therefore, the  trans-activators and the cis elements that mediate expression in MN and dI1 neurons should be conserved between mammals and aves. The 2.7Kb that separate TAG1 and Nfasc genes in the chick were screened for E-boxes. A 760 bp element (cE3 enhancer) that contains three E-boxes was utilized in ovo as a potential enhancer element (Fig. 6A). Two domains of expression are apparent in the ventral spinal cord, 80% of the GFP labeled cells are Isl1+ (Fig. 6B) and TAG1+ (Fig. 6C) indicating expression in motor neurons. In the dorsal spinal cord expression is confined to the dorsal midline. 3% of the GFP labeled cells are Brn3a+ indicating that they are dI1-3 neurons, while 97% of the GFP+/Brn3a-cells reside in the dorsal midline, and are likely to be roof plate cells. Hence, cE3 enhancer directed expression of GFP in motor neurons (Fig. 6A,B), but not in dI1 neurons, thus, providing further support to the hypothesized role of bHLH proteins in regulating the MN expression of the E3 enhancer.

Discussion
Motor neurons and interneurons emerge from distinct progenitor cell domains along the dorsoventral axis of the neural tube. Progenitor cells in each domain are specified by graded sonic hedgehog (Shh) and BMP signaling, leading to the expression of unique combinations of homeodomain and basic helix-loop-helix (bHLH) transcription factors. Subsequent differentiation and connectivity of spinal neurons is governed by an assortment of cell type specific transcription factors, defined as a transcriptional code, that controls the expression of receptors for guidance cues [1,2]. TAG1 and Nfasc are co-expressed in numerous neuronal cell types in the CNS and PNS; both promote neurite outgrowth [24,38,39] and both are required for the architecture and function of nodes of Ranvier [40,41]. In addition, TAG1 is required in vivo for axon guidance [13,42,43], neuronal migration [44,45], and modulates the responses of sensory axons to diffusible guidance signals by controlling the trafficking of their receptors [43]. In the current study, we have identified two complementary enhancer elements of the TAG1 gene that direct expression during the midembryonic period to either the PNS or the CNS. Further characterization of the CNS enhancer element, situated 39 to Nfasc and 59 to TAG1, demonstrated that it directs expression in dorsal interneurons and in motor neurons. Deletion and point mutation analysis reveal that a minimal element of 130 bp, that contains 3 E-boxes, is required for directing expression in motor neurons.
The genomic link between TAG1 and Nfasc is conserved in all vertebrates: mammals, birds and fish. Their spatial co-expression pattern, in DRG and in commissural and motor neurons, is also conserved (this study; [24,46,47,48]). The preferential E3-mediated expression of reporter genes in the commissural dI1 neurons -dI1c, versus in the ipsilateral subpopulation -dI1i, supports the theory that E3 is a TAG1, rather than Nfasc, enhancer element. However, we cannot exclude that E3 regulates both genes. Deletion of the E3 element in the context of a large BAC that contains both genes, or via gene targeting in mice, would be required to verify this issue.
Our conclusion that regulatory elements lying between 4kb upstream of exon 1 and the translation start site in exon 2 drive expression in the embryonic PNS, appears to be at odds with the report by Denaxa et al that a fragment lying 4 kb 59 to exon 1 directs expression to commissural axons [27]. However, although there is ambiguity as to exactly which elements were assayed in this study (the 4 kb sequence deposited in Genbank, X92681, appears to be compilation of sequences from different parts of the CNTN2 locus), an apparent difference is that their construct did not include sequences from intron 1, which is known to harbor key regulatory elements in the genes of other members of the TAG1 family [28,29]. Notably all of the constructs we assayed included at least the first 2 kb of intron 1, which is enriched in conserved sequences, yet in none of these constructs did we see expression within the neural tube, suggesting this region may contain elements that suppress CNS expression in the embryo. Moreover, the most robust expression was found when all of intron 1 was present, suggesting that further intronic elements contribute to the complete embryonic PNS expression pattern.
The separation of regulatory elements controlling PNS from CNS expression might suggest that these different systems may have evolved separately. However, this separation may be restricted to the mid-embryonic phase, since at later stages the TGM-3M enhancer also drives expression in the CNS, notably in the cerebellum [26], the cortex [49] and the retina (SVK unpublished).
The E3 element was isolated based on its homology across mammals. Yet, the mouse E3 element directs expression of reporter gene to motor neurons and dorsal interneurons in both mouse and chick. These observations are highly indicative that the transcriptional complexes directing the expression in these two neuronal populations are potentially conserved among vertebrates. It is also plausible that, through evolution, only the obligatory short cis elements, e.g. the E-boxes, remained conserved between mammals and avian.
bHLH and Lim-HD proteins are expressed in motor and dI1 neurons: Olig2, Ngn2, NeuroD and Ascl1 in motor neurons and Atoh1 in dI1 neurons; Isl1,Isl2 and Lhx3 in motor neurons and Lhx2 and Lhx9 in dI1 neurons. Ectopic expression of the dI1 transcription factors -Lhx2 and Lhx9, either separately or in combination, failed to elevate the expression of the endogenous TAG1 (not shown). It is likely, therefore, that other transcription factors play a role in regulating the expression of TAG1 in commissural interneurons. Two candidates were tested by us: BarHl1 (expressed in dI1 neurons) and UNC4 (expressed in all the commissural INs) -neither of these was sufficient to up-regulate the expression of TAG1 (data not shown).
Among the dorsal commissural interneurons, E3 directed expression only in dI1 neurons. Based on the expression of TAG1 and Robo3, dI2, dI4 and dI6 are also commissural neurons [5,10]. All these interneurons express bHLH and Lim-HD proteins: Atoh1+Lhx2/9 in dI1, Ngn1/2+Lhx1/5 in dI2,dI4 and dI6. The combined expression of bHLH and Lim-HD proteins is probably not the sole requirement for mediating the expression of commissural specific receptors for guidance cues, such as Robo3 and TAG1, since the ipsi-lateral projecting INs (dI1i and dI3) express a similar combination of bHLH and Lim-HD TFs (Lhx9+Atoh1 and Isl1+Ngn2, respectively), yet do not express either Robo3 or TAG1 [5]. Hence, a higher combinatorial order of transcription factors, presumably including additional proteins, is likely to control the expression of commissural neuron guidance cue receptors.
The identity of motor neurons is traditionally monitored by the expression of transcription factors that are expressed along the differentiation cascade, such as Isl1 and Hb9 [50,51]. In such studies, various bHLH proteins have been found to be involved in the neurogenesis and differentiation of motor neurons. Olig2 is required to determine the progenitor motor neuron (pMN) fate. Functional studies indicate that Olig2 down-regulation is required to release pMN cells from an inhibitory block on post-mitotic motor neuron formation [52,53,54]. An Olig2 to Ngn2 switch is required for the differentiation of post-mitotic motor neurons [52]. In addition, the coordinated activity of Lim-HD proteins Isl1 and Lhx3, with bHLH proteins, either NeuroM or Ngn2, synchronizes the neurogenesis and specification of motor neurons [50]. In the current study, the expression of a known guidance-cue receptor -TAG1 -was studied. The sole requirement of the E-boxes to recapitulate the endogenous expression in the mid-gestation spinal cord, suggests that post-neurogenesis, a bHLH protein may function as a critical transcription factor regulating the expression of axon guidance receptors in motor neurons.

In ovo electroporations
Fertilized white Leghorn chicken eggs were incubated at 38.5 to 39uC. A DNA solution of 5 mg/ml was injected into the lumen of the neural tube at stage HH17 to HH18. Electroporation was performed using three 50 ms pulses at 25V, applied across the embryo using a 0.5 mm Tungsten wire and a BTX electroporator (ECM 830). Embryos were incubated for additional 1 to 3 days prior to analysis.

Transgenic mice
Elements of E3.1 enhancer fragment, or a mutant version thereof lacking E-box binding sites, have been cloned upstream of a minimal Hsp68 promoter driving expression of LacZ marker. Purified fragments have been injected in fertilized mouse oocytes (FVB strain) and transferred into pseudopregnant recipient females by standard procedures. Embryos were collected at appropriate gestation age (mostly, at E11.5 post fertilization), yolk sacs separated for subsequent genotyping to determine transgene integration while embryos were fixed in 1% formaline/0.25% glutaraldehyde. Staining in 1mg/ml X-gal solution, also containing 2 mM MgCl 2 , 0.05% Triton-X100 and 5 mM each of K 4 [Fe(CN) 6 ] and K 3 [Fe(CN) 6 ] at room temperature has been allowed to proceed overnight. Figure S1 The organization and conservation of the TAG1 gene. Scheme showing organization of mouse TAG1 gene 7 kb upstream and 8.5 kb downstream of exon 1. Blue boxes directly below indicate the sequences of the human TAG1 gene deposited in Genbank by Denaxa et al., 2003 (Accession Number X92681.1), from which substantial parts of intron 1 and intron 2 are missing (dashed lines). Below these are indicated constructs used in this study, which include human TAG1 genomic DNA (blue line) fused in-frame at the ATG codon to a LacZ reporter gene (light blue). Below that, an alignment of the mouse region to rat, rabbit, dog, elephant and opossum elements. The Alignment was done using BLAT alignment tool of UCSC genome bioinformatics (http://genome.ucsc.edu/cgi-bin/hgGateway). (TIF) Figure S2 E3 enhancer direct expression preferentially to dI1c neurons. Co expression of GFP driven by dI1 specific enhancer [10], and mCherry driven by E3 enhancer (E). A-D Images of the ventral lateral spinal cord. Yellow arrowheads point to the dorsal medial dI1 neurons (the position occupied by dI1c neurons) that are co-labeled by GFP and mcherry. Green arrowheads point to the ventral lateral dI1 neurons (the position occupied by dI1i neurons) that are labeled by EdI1-derived GFP, but not mcherry-derived E3. Yellow arrows point to commissural axons that project toward or at the floor plate. Green arrows point to ipsilaterally projecting axons. Red arrows point to motor axons. Scale Bar in D 100 mm.