Molecular characterization of the TrkA/NGF receptor minimal enhancer reveals regulation by multiple cis elements to drive embryonic neuron expression

Neural development is controlled by accurate temporal and spatial gene expression. Among the many critical genes involved in this process are those encoding the NGF family of neurotrophins and their cognate Trk family receptor tyrosine kinases (Levi-Montalcini, 1987; Snider, 1994). The neurotrophins are widely distributed during development with expression found in neural tissues but also in diverse non-neural innervation target tissues where they fulfill their role in the ‘Neurotrophic Hypothesis’. In contrast, Trk family receptors are largely expressed in neurons and, in the peripheral nervous system (PNS), are confined to complementary restricted subsets of neuronal subpopulations (Martin-Zanca et al., 1990; Tessarollo et al., 1993). Among them, the receptor for NGF, TrkA, is the most spatially restricted with embryonic expression confined to subsets of sensory neural crest-derived neurons and sympathetic trunk neurons. Gene knockout experiments have demonstrated the requirement for accurate TrkA expression during development (Liebl et al., 2000; Smeyne et al., 1994). In the absence of TrkA, sensory ganglia lose a significant proportion of their neurons via apoptosis and studies have demonstrated that these are largely pain transmitting (nociceptive) neurons (Liebl et al., 2000, 1997; Smeyne et al., 1994). In addition, absence of TrkA function results in loss of virtually all sympathetic neurons (Smeyne et al., 1994). Thus, in the embryonic PNS, TrkA receptor expression defines a subpopulation of neurons that rely on its function for survival.

The study of organisms ranging from Drosophila and C. elegans to mammals has resulted in the identification of transcription factor encoding gene families that are critical for specifying domains of cellular differentiation and specialization. For example, the myoD gene family controls early myogenic differentiation (Olson, 1990; Weintraub et al., 1991). In the nervous system, the achaete-scute related gene family governs many aspects of neurogenesis in Drosophila and in the mammalian peripheral nervous system (Anderson, 1994; Jan and Jan, 1990, 1993). The current challenge is to determine the physiologically relevant target genes through which these and other important transcription factors mediate gene regulation.

Despite the importance of Trk family receptor expression during peripheral neurogenesis, relatively little is known about the underlying mechanisms that direct their restricted and complementary expression patterns. The neurogenin family of transcription factors appears to be required for the emergence of peripheral neurons (Anderson et al., 1997; Lee, 1997; Sommer et al., 1996), but current understanding indicates a role for these transcription factors in neuronal precursor expansion rather than directly on Trk gene expression. In a series of elegant studies, the Pou related Brn3a transcription factor has recently been identified as a likely regulator of Trk family gene expression (Huang et al., 1999; McEvilly et al., 1996; Xiang et al., 1996, 1995). The mechanism whereby this regulation takes place remains to be determined.

In the present study, we have undertaken to identify the cis genomic elements that define murine TrkA gene expression as a means of identifying the critical transcription factors that target upon TrkA. We have chosen this neurotrophin receptor because of its particularly restricted gene expression profile during embryogenesis (Martin-Zanca et al., 1990; Tessarollo et al., 1993). Through transgenic analysis using a β-galactosidase reporter assay, we have defined a 457 base pair sequence that is sufficient to drive accurate TrkA expression in embryogenesis. Survey of the human and chick TrkA genes has confirmed the conservation of this enhancer region in approximately the same domain. Mutational dissection of this minimal enhancer further reveals the importance of previously identified regulatory elements as essential for TrkA expression in trigeminal, dorsal root and sympathetic neurons. These studies provide the first information about essential transcriptional target domains for the neurotrophin receptors and form a basis for identification of the factors whose in vivo targets are the Trk family genes.

Constructs in Fig. 1 were made as follows: plasmid pBS-Hsp68-lacZ-pA (Kothary et al., 1989; Verma-Kurvari et al., 1998) was digested with SmaI and treated with calf alkaline phosphatase (GibcoBRL). The linearized plasmid was ligated to different TrkA enhancer DNA fragments. Constructs with inserts were sequenced.

Pronuclear injection and oviduct transplantation of injected embryos were performed as described by Hogan et al. (1986) using either CD2 or B6D2F1 females. The DNA for pronulcear injection was first digested with SalI restriction enzyme and then purified from 1% agarose gels with QIAGEN gel purification kit (QIAGEN). After injection, the eggs were transferred back to the oviducts of pseudopregnant CD2 female mice for development. The recipient females were sacrificed around embryonic day (E)13 or on the day of birth for recovery of embryos or postpartum (P)0 pups. Embryos or P0 pups were genotyped (see below) and all transgene-positive embryos or P0 pups were used for β-galactosidase staining (see below). Three stable transgenic lines were established by standard procedures (Hogan et al., 1986).

Embryos or P0 pups were fixed in 4% paraformaldehyde for 30 minutes at 4°C. Staining for β-galactosidase was carried out in 1× PBS, 2 mM MgCl2, 50 mM potassium ferri/ferrocyanide, 1 mg/ml X-gal at 25°C for 48 hours or at 37°C for 24 hours. After whole-mount analysis, specimens were either embedded in 3% agar, chicken albumin or OCT for either vibratome sectioning at 100 to 200 μm (agar or chicken albumin embedded) or cryostat sectioning at 50 μm (OCT embedded). The sections were then counterstained with Eosin and embedded in DPX-mountain dew for long term storage.

EMSA was carried out as described by Edmondson et al. (1992). Briefly, 10000 to 40000 cts/minute 32P-labeled oligonucleotide probes were mixed with 1 μg DRG nuclear extract, 5× gel shift buffer, 1 μg poly(dI-dC) (Boehringer) and water to a final volume of 20 μl. The mix was incubated at 25°C for 30 minutes. For supershift assay with antibody (Wang et al., 1996; rabbit anti-mouse ikaros), the antibody was added 10 minutes after the mix was prepared. The mix was incubated for a further 20 minutes. Electrophoresis was done with 5% polyacrylamide gels in 0.5 ×TBE buffer at 25°C for 2 to 3 hours. The gel was then dried and exposed to Kodak X-ray film or to phosphoimager (BioRad).

Genotyping of embryos or adult mice for lacZ transgene was carried out using the following PCR conditions: 94°C, 30 seconds; 65°C, 30 seconds; 72°C, 30 seconds; 30 cycles. The primers for lacZ PCR are: 5′-AACTGGAAGTCGCCGCGCCACTGGTGTGGG-3′ and 5′-TGAACTGCCAGCTGGCGCAGGTAGCAGAGC-3′. lacZ-positive embryos or adults are considered as transgenic positive.

The TrkA gene maps to mouse chromosome 3 and human chromosome 1 (Tessarollo et al., 1993). In studying the TrkA genomic structure, we uncovered the presence of a gene encoding the insulin receptor-related receptor (IRR) located in the opposite transcriptional orientation with its start site approximately 2 kb upstream of the first TrkA coding exon (Bates et al., 1997; Kelly-Spratt et al., 1999). Analysis of human, rat and chicken genomic DNA indicated conservation of linkage for the two receptor encoding genes (Fig. 7). Previous studies had demonstrated coexpression of IRR and TrkA in vivo (Reinhardt et al., 1993, 1994). These investigators demonstrated a per cell coexpression for the two genes in the PNS and basal forebrain, whereas the gene encoding IRR has additional sites of expression (Bates et al., 1997; Reinhardt et al., 1993, 1994). We reasoned it likely that TrkA and IRR share enhancer elements directing neuronal coexpression and that such an enhancer would lie proximal to both genes. We therefore began analysis using a 9 kb mouse genomic DNA fragment that spanned the first exon and part of the first intron of the TrkA gene and the first four exons and three introns of the IRR gene (Fig. 1; and not shown). To facilitate the identification of enhancer activity, we used a β-galactosidase (lacZ) reporter gene linked to an Hsp68 minimal promoter (Kothary et al., 1989; Verma-Kurvari et al., 1998). In the present study we performed enhancer analysis using transgenic animals as we were unable to identify cell lines that expressed both the TrkA and IRR genes leading to concerns about the potential lack of promoter/enhancer fidelity in available cell lines.

TrkA is primarily expressed in neural crest derived sensory neurons during embryogenesis (Martin-Zanca et al., 1990). After E17, sympathetic neurons also express TrkA (Tessarollo et al., 1993; Wyatt and Davies, 1995). As described in Fig. 1, the Hsp68/lacZ reporter gene was placed on either end of the 9 kb fragment to test the directional specificity of the fragment (constructs no. 1 and 2). These linearized constructs were microinjected into fertilized eggs and the recipient pseudopregnant females were sacrificed around E13 to analyze sensory expression by X-gal staining. As outlined in Fig. 1, out of a total of 16 transgenic embryos obtained with these two constructs, 12 expressed the transgene (Fig. 1, Blue) and 7 exhibited strong and specific expression of the lacZ reporter gene in trigeminal (TG) and dorsal root ganglion (DRG). The remaining 5 embryos showing β-gal activity varied in their expression probably resulting from either positional effects of chromosomal integration or rearrangement of the TrkA sequences upon integration. Fig. 2A shows a representative transverse section from a construct 2 derived lacZ-positive transgenic embryo. The major sites of expression are the prominent TG and in the upper region, the DRG (arrows). At the apex (roof plate) of the spinal cord, X-gal staining reflects DRG afferent innervation into superficial laminae of the spinal cord (see below). Expression in the retinal epithelium is not reproducible and therefore likely a site of ectopic transgene expression (Fig. 2A). From this initial analysis, we concluded that the 9 kb region contained sufficient information to drive TrkA expression at the appropriate sites in sensory neurons. In addition, the 9 kb fragment fulfills the requirement of an ‘enhancer’ by retaining reporter gene activity when placed in either orientation.

On the basis of the preceding data, we adopted a strategy of microinjecting progressively smaller genomic fragments together with the minimal promoter and reporter gene. As outlined in Fig. 1, transgenic constructs 3, 4, 6, 7 and 8 each provided specific and reliable lacZ transgene expression in TG and DRG at E13.5. Construct 5, which contained about 5.5 kb of the original 9 kb enhancer, did not produce any TrkA-like activity (data not shown). The blue embryos derived from construct 5 exhibited random and varying lacZ expression. Taken together, the data derived from constructs 1-8 (Fig. 1) indicated that the enhancer activity was likely contained within a 740 bp DNA sequence shared by the active fragments, upstream of the first TrkA exon and excluding a role for intronic sequences within the TrkA and IRR genes (Fig. 1).

To further delineate the TrkA specific enhancer, we tested construct 9, which retained approximately 740 bp of sequence in common with the preceding active fragments. Construct 9 again provided sufficient TrkA enhancer information with 2 of 2 lacZ-expressing embryos showing both TG and DRG expression (Fig. 1). As indicated in Fig. 1, constructs 9-12 further refine the boundaries of the enhancer to a minimal fragment of 457 base pairs that has TrkA-specific expression activity in either orientation. Fig. 2B shows a representative transverse section through embryonic brain showing expression in DRG and TG generated with construct 9. Fig. 2C-G illustrates the lacZ expression pattern driven by the minimal 457 bp fragment (construct 11) in the TG and DRG, including central and peripheral projections in an E13.5 embryo. Fidelity of TrkA expression is further exemplified by X-gal labeling of all DRG along the embryonic spinal cord. Afferent projection into the superficial laminae of the dorsal horn spinal cord (Fig. 2E) is consistent with appropriate nociceptive afferent innervation, where TrkA-expressing neurons are normally found (Liebl et al., 1997). This is further supported by the punctate expression pattern seen in DRG (Fig. 2G), reflecting TrkA expression only in a subset of DRG neurons, similar to the expression pattern documented for embryonic TrkA expression (Fig. 2G; Martin-Zanca, 1990; Tessarollo, 1993). Thus, a 457 bp cis genomic fragment exhibits the hallmarks of a minimal enhancer capable of driving TrkA expression in DRG and TG neurons in vivo.

Attempts to further refine the size of the minimal enhancer were unsuccessful (Fig. 1, constructs 13-15). Constructs 13 and 14, lacking progressive sequences of the 3′ enhancer region did not exhibit lacZ expression in DRG or TG (Fig. 1; Fig. 2H for construct 14). Construct 15 included the 3′-most 250 bp of the enhancer but excluded the 5′ 200 bp. This construct was insufficient for appropriate expression, however it did produce weak staining in DRG but not in TG (Fig. 2I). Thus, sequences in the 3′ half of the enhancer are absolutely required for minimal expression and may contain regulatory sequences that are sufficient to drive DRG but not TG expression.

To examine, in greater detail, the properties of the TrkA enhancer, we generated three independent stable mouse transgenic strains using either constructs 10 or 12 (see Materials and Methods for details). All three transgenic strains exhibited specific reporter gene expression in TG and DRG at P0 (Figs 2J-L, 3). In addition, a degree of variable, non-specific expression was observed from strain to strain (data not shown).

TrkA is essential for the survival of nociceptive sensory neurons (Liebl et al., 2000; Smeyne et al., 1994). We therefore crossed the three stable TrkA enhancer transgenic lines into the TrkA null genetic background (Liebl et al., 2000) to determine whether lacZ expression was confined to TrkA-positive, NGF-crosses of the lacZ transgene into TrkA heterozygotes (+/–; left), or TrkA null P0 pups (–/–; right). Whereas the compound enhancer transgenic/heterozygotes express lacZ appropriately, in the null background, X-gal staining was abolished (compare also Fig. 3B-F). These data confirm the fidelity of the 457 bp enhancer in TrkA-expressing sensory neurons.

The highly stereotypic expression of the TrkA gene during development in subsets of sensory neurons and in sympathetic neurons in late development suggests that regulation may be mediated by diverse regulatory elements. Therefore, we surveyed the mouse sequence for known consensus transcription factor binding sites (Fig. 4A). This analysis identified 15 different types of highly conserved cis-elements within the minimal enhancer: E4 BP4 (Cowell et al., 1992); Nkx2.5 (Chen and Schwartz, 1995); DeltaEF (Funahashi et al., 1993; Remacle et al., 1999; Sekido et al., 1994, 1997); e/d Hand (Cserjesi et al., 1995; Hollenberg et al., 1995); Ets (Ahne and Stratling, 1994; Chen et al., 1996; Pastorcic and Das, 1999); MZF (Hromas et al., 1996; Hui et al., 1995; Morris et al., 1994); Ikaros (Molnar and Georgopoulos, 1994); Ebox (Olson, 1993; Weintraub et al., 1994); CACCC box (Cassard-Doulcier et al., 1994); AP1FJ1 (Kouzarides and Ziff, 1988; Neuberg et al., 1989; Rauscher et al., 1988); Sp1 (Dynan and Tjian, 1983a,b; Schmidt et al., 1989); NF1 (Osada et al., 1996, 1999); Cux (Andres et al., 1994; Harada et al., 1995); Gfi-1 (Zweidler-Mckay et al., 1996); and S8 (de Jong et al., 1993). These cis-elements are spread throughout the 457 bp minimal enhancer and form small clusters of 1 to 3 cis-elements with 0 to 10 intervening nucleotide spaces (Fig. 4A).

As a first approach toward determining whether the identified consensus binding sites are required for DRG TrkA expression, we generated 14 oligonucleotides that spanned the consensus binding domains (Fig. 4A), and subjected them to gel mobility shift assays with embryonic DRG nuclear extracts (Fig. 4B,C). Strong mobility shift was observed with five of the oligomers (Fig. 4B; oligos 4, 5, 9, 10, 11). One oligomer shifted moderately weakly (oligo 8; Fig. 4B), and weak or no shifts were observed with eight of the oligomers (oligo 14, Fig. 4C; and oligos 1, 2, 3, 6, 7, 12 and 13, not shown). The mobility shift data provided evidence for the presence of binding factors in DRG extracts. While the negative results do not exclude the presence of weakly binding elements, the binding data provided a basis to pursue a functional analysis of the minimal enhancer.

As illustrated in Fig. 5, eleven core transcription binding elements contained within gel-shifted oligonucleotides were selected for functional analysis, including an Ebox (Fig. 5, Ebox(1)) within oligo 3 that did not shift as a comparison to the selected Ebox (Fig. 5, Ebox(2)). On the basis of these criteria, the NF1, S8, Gfi-1, Cux, CACCC box, Nkx2.5, and E4BP4 sites were excluded from the present studies. The 457 bp fragment contained in construct 11 (Fig. 1) was subjected to in vitro mutagenesis designed to alter at least four of the consensus core nucleotides for each of the core consensus sequences (Fig. 5, ‘Mutation’). Replacement mutagenesis was selected to avoid disruption of possible spacing requirements for a functional enhancer. Each mutation described in Fig. 5 was employed to generate transgenic embryos. lacZ expression in TG and DRG was analyzed from E12.5 to E14.5. The results of this analysis are summarized in Fig. 5, illustrated in Fig. 6, and described below.

DeltaEF: DeltaEF belongs to the homeobox-zinc finger family of transcription factors (Funahashi et al., 1993; Sekido et al., 1994; Sekido et al., 1997). Oligonucleotide 11 containing the DeltaEF site in the TrkA enhancer shifts strongly with DRG nuclear extracts (Fig. 4B, oligo 11). Mutation of the DeltaEF site specifically ablated TG expression while retaining normal DRG expression (Figs 5, 6A).

Hand: Oligonucleotides containing a highly conserved non-canonical e/d-Hand transcription factor binding site (Cserjesi et al., 1995; Hollenberg et al., 1995) shifted strongly with DRG nuclear extract (Figs 4A; 4B, oligo 11). The TrkA minimal enhancer with mutation of this site maintained the ability to drive lacZ reporter expression, albeit reduced, in both TG and DRG (Fig. 5; compare Fig. 6B and L). In addition, mutation of the e/d-Hand site ablated sympathetic chain expression (see below).

Ets: Two highly conserved Ets transcription factor binding sites were also found in the minimal enhancer (Fig. 4A; Ahne and Stratling, 1994; Chen et al., 1996; Pastorcic and Das, 1999). Oligonucleotides containing either Ets site were shifted with DRG nuclear extracts (Fig. 4B, oligos 8 and 11, respectively). Mutation of the second Ets site (Fig. 5, Ets(2)) did not affect minimal enhancer function (Figs 5, 6C). Two of three blue embryos generated with the Ets(2) mutation showed normal TG and DRG. However, mutation of the first Ets site (Fig. 5; Ets(1)) completely abolished lacZ expression in both TG and DRG (Figs 5, 6G and data not shown). Ectopic lacZ transgene expression was observed in several of the embryos.

Thus, one of two Ets sites appears required for TrkA expression in sensory neurons. MZF: The MZF family of mouse zinc finger transcription factors is homologous to Drosophila Kruppel (Hromas et al., 1996; Hui et al., 1995; Morris et al., 1994). Two highly conserved MZF binding cis-elements (Morris et al., 1994; Shields and Yang, 1998) are found in the minimal enhancer (Fig. 4A) and oligonucleotides including each MZF site shift strongly with DRG nuclear extracts (Fig. 4B, oligos 5 and 10). Mutation of the 5′ MZF(1) site, like the DeltaEF site, ablated expression in TG but retained expression in DRG (Fig. 5; Fig. 6H) while mutation of the 3′ MZF(2) site ablates lacZ reporter gene expression in both TG and DRG (Fig. 5; Fig. 6D),.

Ikaros: The Ikaros family of mouse zinc finger transcription factors is homologous to Drosophila Hunchback (Georgopoulos et al., 1992). Three highly conserved Ikaros binding cis-elements (Molnar and Georgopoulos, 1994) are found within the minimal TrkA enhancer (Fig. 4A), however we discounted the 3′-most site since it is contained within an oligomer that was not shifted by DRG extracts (Fig. 4C). To examine if any Ikaros-like transcription factors bind the second and third Ikaros cis-elements, we performed gel shift assays using an Ikaros polyclonal antibody (Wang et al., 1996) (Fig. 4C, oligos 4 and 9, second lane). Both the Ikaros(1) and Ikaros(2) cis-elements were ‘supershifted’ with addition of antibody, while neither the 3′-most Ikaros site (Fig. 4C, oligo 14) nor the MZF(1) cis-element containing oligomer (Fig. 4C, oligo 10) showed evidence of a supershifted band.

Mutation of the first Ikaros(1) site completely ablated expression in TG while retaining weak DRG expression (Figs 5, 6J). Mutation of the Ikaros(2) site attenuated lacZ expression in both TG and DRG compared with wild-type enhancer (Figs 5, 6E). Thus, Ikaros-mediated transcription is an important element for TrkA regulation and the Ikaros(1) site appears to primarily be required for TG expression as were the MZF(2), and DeltaEF sites described above.

EBox: Three EBox cis-elements were found in the minimal enhancer (Fig. 4A; Olson, 1993; Weintraub et al., 1994). Oligonulceotides containing the 5′-most (Ebox(1)) and third EBox (3′-most) sites showed no shift in DRG nuclear extract assays (data not shown), while oligonucleotides containing the second Ebox (Ebox (2)) site were shifted in the assay (Fig. 4B, oligo 8). The Ebox(2) site is only one nucleotide removed from the Ets(1) site described above (Fig. 4A), suggesting that interaction could exist (Sieweke et al., 1998). As shown in Figs 5 and 6F, mutation of Ebox(2) resulted in complete ablation of enhancer function in TG while still maintaining weak activity in DRG. Mutation of the 5′-most Ebox(1) site contiguous with a consensus SP1 site had no apparent effect on sensory neuron expression (Figs 5, 6K), consistent with the gelshift data that oligonucleotide 11 containing this Ebox did not shift with DRG nuclear extract.

AP1FJ1: Two AP1 cis-elements are found in the TrkA minimal enhancer (Kouzarides and Ziff, 1988; Neuberg et al., 1989; Rauscher et al., 1988) (Fig. 4A). The 3′-most AP1FJ1 cis-element, present in oligo 12, was not shifted by DRG nuclear extracts (data not shown). The core sequence of the 5′-most AP1FJ1 cis-element is only 2 nucleotides removed from the core sequence of the Ikaros(1) site whose mutation ablated TG expression while retaining weak DRG expression (Fig. 4A, see above). Mutation of the 5′-most AP1FJ1 cis-element also ablated lacZ reporter gene in the TG while retaining weak DRG expression (Figs 5, 6I).

In summary, regulation of TrkA expression by the 457 bp minimal enhancer is complex and contains cis elements required for expression in both TG and DRG, while additional elements that are particularly important for TG expression.

The intervening genomic sequence between the TrkA and IRR genes from human and chick was sequenced and surveyed for regions of homology with the murine enhancer. As schematically represented in Fig. 7A, highly conserved domains were found in all three species. Overall nucleotide conservation was 65% between mouse and human and 42% between mouse and chick. Strikingly, the clusters of transcriptional binding sites that affected enhancer function (Fig. 7B, red, yellow and purple), are highly conserved. As noted in Fig. 7B, the MZF(1) site that is required for trigeminal expression appears absent in human or chick. However, inspection of the human and chick sequences reveals MZF binding sequences within a few nucleotides, just 3-prime (Fig. 7B, red boxes). The Ets(2) site (nucleotide 382) is not required for function although conserved and present in an oligomer that was shifted (Fig. 4, oligo 11). We provisionally attribute the shift to Hand and DeltaEF sites present in the same oligomer and required for global and transgenic expression respectively (Figs 5, 6).

The regions spanned by oligomers 1-3, and 12-14 (Fig. 4), were not shifted with DRG extracts but exhibit high sequence conservation (Fig. 7B). Accordingly, mutation of the conserved Ebox(1) domain within oligo 3 (Fig. 4) did not affect enhancer function in DRG. It is likely that these conserved sites, may serve to specify TrkA expression in locations other than DRG such as the superior ganglion, cholinergic basal forebrain neurons, and selected spinal cord neurons (Martin-Zanca et al., 1990; Liebl et al., unpublished). We did not prepare an oligomer to span nucleotides 140-180 (Fig. 7B) as we did not discern consensus transcriptional binding sites. However the high degree of evolutionary conservation suggests a possible functional domain (see Discussion).

In addition to expression in neural crest-derived sensory neurons, TrkA function is present and essential for the survival of late embryonic and postnatal sympathetic neurons. In mice, sympathetic expression of TrkA begins at E17.5 and reaches highest levels several days after birth (Dixon and McKinnon, 1994; Tessarollo et al., 1993). To examine whether the minimal enhancer determined for sensory neuron expression was also sufficient for sympathetic expression, we tested construct 11 (Fig. 1, far right column) by allowing the microinjected zygotes to develop to term prior to analysis. Of four transgenic pups, two expressed lacZ. Each embryo exhibited TG and DRG expression and one of two embryos also exhibited expression in neurons of the sympathetic trunk (Fig. 1 and not shown). These results indicate the presence of regulatory elements within the 457 bp minimal enhancer that specify sympathetic neuron expression.

Sympathetic neurons require NGF signaling through the TrkA receptor for survival (Smeyne et al., 1994; Snider, 1994). At P0, TrkA null pups have severe loss of sympathetic neurons and the remaining populations progressively disappear in the ensuing days. Two of the three derived stable TrkA minimal enhancer transgenic lines exhibited lacZ expression in sympathetic ganglia (SG) at P0 (Fig. 3G; and data not shown). When we examined the expression of lacZ in the compound TrkA heterozygotes (+/–) and homozygotes (–/–; Fig. 3G), in the homozygotes both the size of the ganglia and the degree of lacZ expression was considerably diminished. Comparison of Fig. 3H and I illustrates the reduction of lacZ expression in dying neurons. These results provide further evidence for the specific expression of the lacZ transgene in sympathetic neurons mediated by the 457 bp TrkA enhancer.

The eHand and dHand transcription factors are strongly expressed in sympathetic chain during embryonic development (Cserjesi et al., 1995; Hollenberg et al., 1995). We therefore examined whether the site encompassed by oligo 11 (Fig. 4) contributes to function of the minimal enhancer in sympathetic neurons. Of three lacZ-expressing transgenic mouse pups (P0 stage) generated with this site mutated in the minimal enhancer, none showed lacZ expression in sympathetic ganglion, while one showed weak expression in TG and DRG (data not shown). These data support the idea that the e/d Hand site is required for appropriate TrkA expression in sympathetic neurons. Thus, a 457 bp DNA fragment upstream of the TrkA gene is sufficient to support appropriate expression in neural crest derived sensory and sympathetic neurons.

The molecular forces that regulate development include a nexus of extracellular stimuli, intracellular signaling and transcriptional modulation, leading to coordinate expression of key target genes that deploy specialized functions. Multiple examples of identified transcription factor families that are essential for normal development exist, including mash (Anderson et al., 1997; Guillemot et al., 1993), the neurogenins (Anderson et al., 1997; Fode et al., 1998; Lee, 1997; Sommer et al., 1996) and otx (Shimamura et al., 1995). In each instance, genetic analysis has confirmed the requirement of these transcription factors for neural development as was initially suspected from stereotypic expression profiles and/or relatedness to gene products with important regulatory functions in other organisms. A major challenge remains in the understanding of the complex coordinate regulation by which transcription factors specify the networks of physiologically relevant target genes.

TrkA is an example of an essential developmental target gene that specifies survival of nociceptive sensory and sympathetic neurons (Liebl et al., 2000; Smeyne et al., 1994). The highly stereospecific profile of this gene has made it particularly suitable for seeking cis-transcriptional regulators. Molecular and transgenic dissection of the genomic region 5-prime to the first coding exon has uncovered a contiguous 457 base pair sequence that contains sufficient information to drive appropriate TrkA expression in sensory and sympathetic neurons. We have restricted the present analysis to consensus transcription factor binding sequences contained in oligomers that could be detected in gel shift assays using embryonic DRG nuclear extracts. Thus, one limitation of the present study is the requirement for strong binding under the conditions used, and the requirement that the transcription factors be expressed in embryonic DRG at E13.5. In consequence, our functional analysis would only be expected to identify enhancer sequences that are required for either global TrkA expression or those that impinge on DRG expression.

The results obtained by mutation analysis are consistent with the preceding deletion analysis of the minimal enhancer. Construct 14 (Fig. 1), lacking 200 bp of 3-prime sequence, lost all enhancer activity, whereas the 3-prime 250 bp region retained weak DRG activity (construct 15). Consistent with this observation, mutation of the MZF(1), AP1FJ1 and Ikaros(1) sites, all located in the 5-prime 200 bp region, ablated expression in TG while retaining weak expression in DRG. These data support the model that the three sites mentioned are crucial for the function of the 5-prime 200 bp region of the enhancer. Conversely, mutations of the MZF(2) and ETS(1) sites in the 3-prime 200 bp region ablated both TG and DRG expression, suggesting these two sites are absolutely required for the 3-prime 200 bp function. Of note, mutation of the overlapping consensus Ets(1) and Ebox(2) sites yielded very similar effects on enhancer function causing loss of TG expression. We cannot rule out that mutation of either core consensus sequence may in fact impair the whole region. However, it is interesting to note that possible interactions between these two different types of transcription factors have been proposed (Sieweke et al., 1998).

There is increasing evidence for remarkable conservation of enhancer sequences across species (Aparicio et al., 1995; Brenner et al., 1993; Elgar et al., 1996). For example, such homology has assisted in the identification and dissection of regulatory sequences for wnt genes, neurogenin, nestin and Ddx in the nervous system (Helms et al., 2000; Josephson et al., 1998; Rowitch et al., 1998; Zerucha et al., 2000). In the present study we find conservation of the 457 bp enhancer region in mouse, human and chick. This finding provides strong support for the physiological relevance of the domain and of the identified consensus binding sites within.

A likely component of TrkA regulation that we have not so far examined or experimentally addressed, is the activity of repressors that may act to confine its expression to neurons and, in DRG, to nociceptive neurons (Chen et al., 1998; Chong et al., 1995; Kraner et al., 1992). Of note, among the consensus sites not analyzed in this study because of absence of gel shift, two Cux consensus sites were identified. Products of this family are known to act as transcriptional repressors (Higgy et al., 1997). Activity of repressors would be anticipated in non-TrkA-expressing cells, thus, utilization of the Cux containing oligonucleotides (Fig. 4, oligos 1 and 2) in gel shifts assays employing extracts either from non-neural tissues, or neurons that do not express TrkA, may provide evidence for action of cux genes in suppression of TrkA expression outside its tightly restricted realm. Thus, continued analysis of the 457 bp minimal enhancer may yet reveal additional information about gene repression.

The Brn3a gene, a member of the POU domain subclass of transcription factors, is strongly expressed in sensory neural crest-derived neurons (Ninkina et al., 1993; Xiang et al., 1995). Genetic ablation of Brn3a function results in loss of sensory neurons, and recent work by Huang and colleagues has demonstrated a significant attenuation of Trk family gene expression specifically in trigeminal ganglion (Huang et al., 1999). We have scrutinized the 457 bp minimal TrkA enhancer and fail to detect well conserved Brn3a consensus binding sites (not shown). Thus, the action of this factor may lie outside the key determinants for TrkA expression but may rather exert its influence from a genomic location that can impinge on the activity of all three Trk family genes (Huang et al., 1999). Alternatively, Brn3a binding sites may lie within the 457 bp minimal enhancer, but of necessity would diverge from the reported consensus binding domains (Gruber et al., 1997; Xiang et al., 1995). We do not rule out this possibility since in this study, other identified domains, when mutated, either attenuate or abrogate TrkA expression in the trigeminal ganglion.

A recent study has examined the promoter elements that drive TrkA gene expression in neural cell lines (Sacristan et al., 1999). This work has defined a promoter region proximal to the 5′ exon for the TrkA gene that is sufficient to elicit reporter gene activity. This promoter activity lies downstream (3′) from the minimal enhancer described in the present study. Our analysis was designed to uncover enhancer sequences in the presence of a minimal (Hsp68) promoter in vivo and did not address minimal promoter function.

In conclusion, we identify an enhancer that contains key contiguous cis sequences required for in vivo, global, DRG-specific, TG-specific or sympathetic neuron expression. The data reveal a coordinated regulation of TrkA gene expression involving multiple regulatory sites. While this provides a complex picture, it is one that most likely conforms with physiological gene regulation, rather than the existence of a single ‘master regulator’. Moreover, this information, provides the first tools to identify the specific transcription factors whose physiological targets include the TrkA gene. In DRG, TrkA is only expressed in nociceptive neurons while TrkB and TrkC are found in mechanoreceptive and proprioceptive neurons respectively. It is therefore likely that common elements of regulation for the three genes may be shared while unique components may specify modality associated neuronal subsets. Complementary approaches to those described here including yeast 1-hybrid and expression cloning using the defined consensus sites in the minimal enhancer have yielded sensory neuron-specific transcription factors consistent with the present study (Lei Lei, L. M. and L. F. P., unpublished data). Thus, continued analysis of the

TrkA minimal enhancer may lead to identification of the key transcription factors that determine neurotrophin receptor expression. We thank Eric Olson and Jane Johnson for critical reading of the manuscript. We thank Lei Lei, Jimi Adu, Frank Namin, Mark Steciuk and all members of the Parada lab for helpful discussions and technical support. Dr Katia Georgopolous kindly provided us with the ikaros polyclonal antibody. This work was supported by NIH R01 NS33199 and R37NS331999 to L. F. P.