mKlf7, a potential transcriptional regulator of TrkA nerve growth factor receptor expression in sensory and sympathetic neurons

Neural development is a complex process that involves the generation of diverse cell types assembled into functional circuits. Neuronal cell fate specification is determined by positional information, cell-cell interaction, extracellular and intracellular signaling events, transcription factor cascades, and ultimately, differential expression of neuronal genes. Different types of transcription factors such as basic helix-loop-helix (bHLH) proteins, homeodomain proteins and zinc-finger proteins have been identified to play important roles in neuronal determination and differentiation (Anderson, 1994; Anderson and Jan, 1997; Jan and Jan, 1990; Jan and Jan, 1993; Nolo et al., 2000; Pfaff and Kintner, 1998). Despite advances in understanding of transcriptional regulation in neural development, it is likely that additional, as yet unidentified transcription factors participate in this complex and highly regulated process. A challenge remains to ascribe physiologically relevant target genes for many of the identified regulatory transcription factors.

The nerve growth factor (NGF) family of neurotrophins (NTs) and their receptors play fundamental roles in the developing nervous system (Levi-Montalcini, 1987; Lewin and Barde, 1996; Snider, 1994). Besides NGF, this neurotrophin family also includes brain-derived neurotrophic factor (BDNF), NT3, and NT4/5. These NTs interact specifically with the Trk family of receptor tyrosine kinases (Bothwell, 1995; Parada et al., 1992). NGF binds TrkA, NT3 preferentially binds TrkC, while BDNF and NT4/5 bind TrkB. As target-derived trophic factors, neurotrophins are broadly expressed in both neural and non-neural innervation target tissues while Trk receptors are primarily expressed in neurons (Klein et al., 1990; Martin-Zanca et al., 1990; Tessarollo et al., 1993). TrkA (Ntrk1 – Mouse Genome Informatics) is specifically expressed in neural crest-derived sensory and sympathetic neurons in the peripheral nervous system (PNS), including dorsal root ganglia (DRG), trigeminal ganglia (TG), superior and jugular ganglia, and sympathetic ganglia (Martin-Zanca et al., 1990; Schecterson and Bothwell, 1992). In the central nervous system (CNS), TrkA expression is limited to a few populations of cells including cholinergic neurons in the basal forebrain and striatum, and some interneurons in the spinal cord (Holtzman et al., 1992; Liebl et al., 2001; Sobreviela et al., 1994). Consistent with cell culture and pharmacological experiments, the generation of TrkA and NGF knockout mice directly demonstrates the requirement of NGF- and TrkA-mediated signaling in the development of peripheral sensory and sympathetic neurons (Snider, 1994). In either NGF- or TrkA-null mice, about 70-80% of dorsal root ganglion (DRG) neurons, 70% of trigeminal neurons and 99% of superior cervical ganglion (SCG) neurons are lost through apoptosis (Crowley et al., 1994; Smeyne et al., 1994). In the CNS, TrkA appears to be required for the normal maturation of basal forebrain and striatal cholinergic neurons although most of these neurons survive in TrkA-null mice for the lifespan of the mutant animals (Fagan et al., 1997).

Although TrkA is expressed in specific areas of the nervous system, little is known about the forces that control such stereotypic expression. To date, the only known gene implicated in regulating TrkA expression encodes the POU domain transcription factor Brn3a (Pou4f1 – Mouse Genome Informatics; Huang et al., 1999). In Brn3a-null mice, the onset of TrkA expression in trigeminal ganglia at embryonic day (E) 11.5 is normal, but no TrkA expression is detectable after E17.5. Therefore, although Brn3a may be important for the maintenance of TrkA expression, it apparently is not required for initiation of TrkA expression in sensory ganglia (Huang et al., 1999). It remains to be tested whether Brn3a regulates TrkA expression directly during late development.

To better understand TrkA expression and to identify transcription factors that are critical for sensory and sympathetic neuron specification, we isolated a mouse TrkA enhancer and functionally characterized the cis regulatory elements within this genomic fragment using transgenic mouse embryos (Ma et al., 2000). Functional analysis revealed that TrkA expression is coordinately controlled by multiple cis elements including consensus DNA-binding core sequences defined for the transcription factor families DELTAEF, HAND, MZF, Ikaros, E box-binding protein, ETS and AP1. Therefore, in vivo, multiple transcription factors binding to these cis elements cooperate to regulate TrkA expression (Ma et al., 2000).

In this study, we identify a novel gene encoding a Kruppel-like zinc-finger protein isolated through expression cloning from a mouse E13.5 DRG expression library. We have named this gene mKlf7 (murine Kruppel-like factor 7) based on its significant sequence homology to human UKLF (ubiquitous Kruppel-like factor), also called KLF7 by the Human Gene Nomenclature Committee (Matsumoto et al., 1998; Turner and Crossley, 1999). We provide evidence that mKlf7 is a transcription factor that may directly regulate TrkA expression in vivo. Because mKlf7 gene expression precedes that of TrkA and continues into adulthood, it is further possible that mKlf7 is required for both initiation and maintenance of TrkA gene expression.

A cDNA expression library was constructed using mRNA isolated from mouse E13.5 dorsal root ganglia and directionally cloned into the EcoRI/XhoI sites of the lambda ZAP Express vector (Stratagene) (J. Merenmies, K. Vogel and L. F. P., unpublished). This expression library was screened using oligonucleotide probes as described (Singh et al., 1988). The probes contained three copies of the wild-type Ikaros2, MZF2 or HAND sites from the mouse TrkA minimal enhancer. The sequences are Ikaros2(3×), 5´-AATTGAA AAA TAG TGG GAG AGA AGA GTC GAA AAA TAG TGG GAG AGA AGA GTC GAA AAA TAG TGG GAG AGA AGA GTC-3´; MZF2(3×), 5´-AATTCAG AAC CTG GGGAGA AAAA CAG AAC CTG GGG AGA AAA ACA GAA CCT GGG GAG AAA AA-3´; and HAND(3×), 5´-AATTAAC GCT CTC CAGACC CTA GT AAC GCT CTC CAG ACC CTA GTA ACG CTC TCC AGA CCC TAGT-3´. One copy of each binding site is underlined and the core sequence for each binding site is highlighted in bold. The 5´ overhang of each oligonucleotide (italics) was added to facilitate labeling with radioactive nucleotides and cloning into the pHisi-1 vector (Clontech, see below), respectively. Each probe was labeled with [α -³²P]-dATP and [α -³²P]-dCTP using the Klenow enzyme in the presence of cold dGTP and dTTP. Probes were then purified using Sephadex G-25 spin column (Boehringer Mannheim). After three rounds of screening, positive lambda phages were purified and the cDNA inserts were excised in vivo using helper phages (Stratagene) and sequenced.

The coding region of mKlf7 was amplified by PCR using primers to generate an EcoRI site at the N terminus and an XhoI site at the C terminus. The PCR product was cloned into the pGemTeasy vector (Promega) and sequenced. The mKlf7 fragment was then released by EcoRI/XhoI digestion and cloned into EcoRI/XhoI sites of the vector pEGFP-C2 (Clontech) to generate the construct pEGFP-Klf7 expressing GFP-Klf7 fusion proteins. Human embryonic kidney 293 (HEK293) cells and mouse neuroblastoma-rat glioma hybrid 108 (NG108) cells were plated on poly-D-lysine coated coverslips and subsequently transfected with 1 µ g of pEGFP-Klf7 or pEGFP-C2 plasmid using Fugene 6 reagents (Boehringer Mannheim). 48 hours after transfection, cells were fixed with 4% paraformaldehyde (PFA) for 30 minutes at 4°C, washed with PBS twice and incubated with 1 µ g/ml 4,6-diamidino-2 phenylinodole (DAPI) for 30 minutes at 4°C. Cells were then washed with PBS three times and coverslips were mounted onto glass slides. Cells were viewed using a fluorescence microscope.

The cDNA fragment encoding the C-terminal DNA-binding domain (amino acid 213-301) of mKlf7 was amplified by PCR using primers to generate an EcoRI site at the N terminus and an XhoI site at the C terminus. The PCR product was cloned into pGemTeasy and sequenced. The mKlf7 fragment was released and subcloned into the EcoRI/XhoI sites of pGEX4T1 vector (Amersham) to generate the construct pGEX-Klf7C expressing GST-Klf7(213-301) fusion proteins. pGEX-Klf7C was transformed into Escherichia coli BL21 cells. Cells were grown at 37°C in LB/ampicilin until OD600 was about 0.6 and isopropylthio-β -galactoside (IPTG) was added at a final concentration of 1 mM. Cells were further incubated for 2 hours, harvested and GST-Klf7(213-301) proteins were purified using Glutathione-Sepharose 4B beads (Pharmacia). To use as controls, plasmids encoding GST, GST-p75 intracellular domain fusion protein (GST-p75ICD), or GST-Brn3a were used to transform BL21 cells; recombinant proteins were purified using Glutathione-Sepharose 4B beads. The GST-p75ICD plasmid and the GST-Brn3a plasmid were kindly provided by Drs Moses Chao and Mengqing Xiang, respectively. p75 is the low-affinity receptor for NGF.

The EMSA protocol was as described (Lei et al., 1998). Each probe contains one copy of wild-type Ikaros2 (5´-GAA AAA TAG TGG GAG AGA AGA GTC-3´), MZF2 (5´-CAG AAC CTG GGG AGA AAA A-3´) or HAND (5´-AAC GCT CTC CAG ACC CTA GT-3´) site from the TrkA minimal enhancer. The core sequence of each binding site is highlighted in bold. The sequence of the Ikaros2 mutant oligonucleotide is 5´-GAA AAA TAG TAA AGG AGA AGA GTC-3´ in which the core sequence was changed to AAAG. Probes were end-labeled with [γ-³²P]-ATP using T4 polynucleotide kinase and purified using Sephadex G25 spin columns. In a 20 µ l reaction, GST-Klf7(213-301) or GST-p75ICD proteins were incubated with 6.7 µ l Buffer S (75 mM Hepes, pH 7.9, 37.5 mM MgCl₂, 30 µ M ZnSO₄, 60% glycerol and 0.3% Tween 20), 1 µ g poly(dI-dC)-poly(dI-dC), 1 µ g bovine serum albumin (BSA) and 20,000 cpm of an appropriate probe. The salt concentration was adjusted to 100 mM NaCl for each reaction. Samples were incubated at room temperature for 20 minutes and loaded onto a 4.5% non-denaturing polyacrylamide gel, separated in 0.5× TBE at 120 V for 3 hours. The gel was dried, exposed to Kodak X-OMAT AR films and analyzed by autoradiograph. For competition experiments, cold oligonucleotides were added 15 minutes prior to the addition of the labeled probe.

Ikaros2(3×), MZF2(3×) and HAND(3×) oligonucleotides were cloned into the EcoRI/XbaI sites of the pHisi-1 vector, respectively. The pHisi-1 vector contains the HIS3 reporter gene downstream of a HIS3 minimal promoter (Clontech). Each construct was confirmed by DNA sequencing and integrated into the HIS3 locus of yeast strain YM4271 (Clontech) to generate the reporter strains YM4271-IK, YM4271-MZF or YM4271-HAND. The coding region of mKlf7 was cloned into the vector pGADT7 (Clontech) to generate the construct pGAD-Klf7 encoding the Gal4 activation domain-Klf7 fusion protein. pGADT7 and pGAD-Klf7 were transformed into YM4271, YM4271-IK, YM4271-MZF or YM4271-HAND and selected on SD-Leu plates. Single colonies from each transformed strain were streaked onto SD-His-Leu plates to test the expression of the HIS3 reporter gene.

Total RNA was isolated from adult mouse tissues using TRIZOL (Gibco BRL) following the manufacturer’s instruction. An estimated 10 µ g of denatured total RNA from each sample was electrophoresed on a 1.5% formaldehyde/agarose gel and transferred to a Hybond N+ nylon membrane (Amersham). Hybridization was performed with the NorthernMax prehybridization/hybridization buffer (Ambion). Probes were generated by random priming with [α -³²P]-dCTP (RediPrime kit, Amersham) using as templates cDNAs for mKlf7 and G3PDH (Gapd – Mouse Genome Informatics) After the initial hybridization with the mKlf7-specific probe and autoradiograph, the blots were stripped of the remaining radioactivity by rinsing the membrane twice with a boiling 0.5% sodium dodecyl sulfate (v/w) solution and rehybridized with a G3PDH-specific probe to compare the quantity of RNA loaded.

For RT-PCR, total RNAs isolated from E8.5 embryos or E13.5 DRGs were used as templates to synthesize cDNAs using the Superscript II reverse transcriptase and oligo-dT primers (Gibco-BRL). The PCR program was as follows 1 minute at 94°C, 1 minute at 59°C, 2 minutes at 72°C; repeated for 25 cycles with a final extension of 5 minutes at 72°C. The gene-specific primers used in the PCR reactions were TrkA, sense 5´-TCA GCA CCG AGA GTG ATG TGT GGA GCTT-3´, antisense 5´-GGA TCC TAG CCC AGA ACG TCC AGG TAA CTC GGT-3´; mKlf7, sense 5´-TTT CCT GGC AGT CAT CTG CAC-3´, antisense 5´-GGG TCT GTT TGT TTG TCA GTC TGTC-3´; and G3PDH, sense 5´-ACC ACA GTC CAT GCC ATC AC-3´, antisense 5´-TCC ACC ACC CTG TTG CTG TA-3´. The PCR product of mKlf7, from − 301 to − 52 relative to the start codon, is highly divergent among Klf family members.

E13.5 DRG neurons were dissected and plated as described (Vogel et al., 1995). Neurons were kept in culture for 2 days in media containing either NGF (10 ng/ml) or NT3 (10 pg/ml). The DRG neuronal plates were washed with PBS without Ca²⁺/Mg²⁺ and single neurons were picked under a microscope and injected into 500 µ l PCR tubes each containing 4 µ l of ice-cold cDNA lysis buffer. Single-cell cDNA synthesis and amplification were performed and checked as described (Dulac and Axel, 1995). To determine the pattern of gene expression in NGF- or NT3-dependent single neurons, 1 µ l of amplified cDNA from each single cell was used as the template for PCR with primers specific for the genes of interest. The PCR program was the same as used in the RT-PCR. PCR products were analyzed on a 1% agarose gel. Sequences for the gene-specific primers were as follows: TrkB, sense 5´-GGA TGG AGA TCA CAG AGG GT-3´, antisense 5´-AGA GGC AAA TGG GTG ACT TG-3´; TrkC, sense 5´-ATG GTG TGA GGT GGG AGG AC-3´, antisense 5´-TTG TAT GTG TAG CAG GCA CT-3´; CGRP (calcitonin gene related peptide), sense 5´-GCA GCC TCC AGG CAG-3´, antisense 5´-GAA GGT CCC TGC GGC G-3´; mKlf7, sense 5´-GCA CAG TGA CGT TGA AAC TGG TG-3´, antisense 5´-TGG TCA GAC CTG GAG AAA CAC CTG-3´; and NCAM (neural cell adhesion molecule), sense 5´-GAA GGA GGG ATG GAC TCC AC-3´, antisense 5´-TTG AAC ACA AGT ATT CTG AC-3´. The primers for TrkA and G3PDH were the same as described previously.

Timed pregnant ICR female mice were sacrificed at various stages to obtain embryos. Gestational age was calculated by taking the morning of the appearance of the mother’s vaginal plug as E0.5. Embryos of E8.5 to E15.5 were dissected and fixed in 4% paraformaldehyde (PFA) overnight, dehydrated in 30% sucrose and 10 µ m-thick sections were cut in a cryostat. For neonatal trunk sections, postpartum (P)0 pups were perfused with 4% PFA, fixed in 4% PFA overnight, dehydrated in 30% sucrose and 10 µ m sections were collected. Adult brains were quickly dissected out after the animals were perfused with 4% PFA, fixed overnight with 4% PFA, dehydrated in 30% sucrose and 10 µ m sections were collected. Adult trigeminal ganglia were dissected out, fixed overnight with 4% PFA, processed and embedded in paraffin, and 5 µ m serial sections were collected using a microtome. Mice with a targeted deletion in TrkA were maintained as described (Liebl et al., 1997). To obtain DRG sections, neonatal pups from the mating of TrkA heterozygous mice were perfused with 4% PFA, their spinal columns dissected out, fixed in Bouin’s solution overnight, washed extensively with 70% ethanol, processed and embedded in paraffin. Sections (10 µ m) were collected using a microtome. The tail DNA from each pup was used for genotyping by Southern blots (data not shown).

Cryostat or paraffin sections were processed for in situ hybridization as described (Martin-Zanca et al., 1990; Nef et al., 1996). [α -³⁵S]-CTP-labeled radioactive antisense cRNAs were produced by in vitro transcription with the T7 RNA polymerase using the Riboprobe Combination System (Promega). mKlf7 antisense cRNA probe was synthesized using as a template the 1.3 kb fragment comprising the entire cDNA shown in Fig. 1A. TrkA antisense cRNA probe was synthesized using a 454 bp fragment encoding the mouse TrkA extracellular domain as the template. After in situ hybridization, slides were counterstained with Hematoxylin for bright field views. Both bright-field and dark-field micrographs were taken using a digital camera. In Fig. 7, the sections were digitally photographed using phase-contrast and dark-field microscopy sequentially, and the dark field images were processed in Adobe Photoshop to display the in situ signals as green particles. The dark field images were then superimposed onto the phase-contrast images to show the cellular morphology and the localization of the in situ signals.

Several consensus cis regulatory elements within the mouse TrkA enhancer are crucial for its specific activity in sensory and sympathetic neurons during mouse embryogenesis (Ma et al., 2000). These cis elements include, among others, two Ikaros sites (Ikaros1 and Ikaros2), two MZF sites (MZF1 and MZF2) and a HAND site. To identify transcription factors present in relevant tissues that may directly bind these DNA elements and potentially regulate endogenous TrkA expression, we screened a mouse E13.5 DRG cDNA expression library using a mixture of three [³²P]-labeled oligonucleotide probes, each containing three copies of the Ikaros2, MZF2 or HAND consensus binding sites. After three rounds of hybridization, six positive lambda phage clones were identified. Sequence analysis of the cDNA inserts indicated two identical clones containing a 903 base pair open reading frame (ORF) preceded by termination codons. This ORF encodes a novel murine Kruppel-like zinc-finger transcription factor (Fig. 1A). The overall amino acid sequence identity between this protein and human UKLF, also known as KLF7, is over 97% (Matsumoto et al., 1998; Turner and Crossley, 1999). Therefore, we named this murine gene mKlf7, based on the significant proteins including GST and GST-Brn3a, fail to bind these probes (data not shown). Thus, mKlf7 exhibits specificity among the three functional sites in the TrkA enhancer, even though both Ikaros2 and MZF2 sites are conserved binding sites for transcription factors containing Kruppel-type zinc fingers in sequence conservation.

mKlf7 protein has a calculated molecular weight of 33.3 kDa and an estimated isoelectic point (pI) of 8.82. Similar to UKLF, the primary structure of mKlf7 protein consists of three distinct domains (Fig. 1B). The N-terminal region (amino acid residues 1-76) is negatively charged and enriched in serines and glutamic acids. This region is identical to the N-terminal region of human UKLF, which was shown to be a transactivation domain (Matsumoto et al., 1998). The central region of mKlf7 (residues 77-211) is enriched in hydrophobic residues and serines, which may serve as a protein interaction domain. The amino acid identity between mKlf7 and UKLF in this region is 94.8%. A putative leucine-zipper motif (residues 98-119) is conserved in both UKLF and mKlf7. The C-terminal region of mKlf7 (residues 212-301) is a potential DNA-binding domain consisting of three zinc fingers of the Cys₂-His₂ type. This region is identical in both mKlf7 and UKLF. The nuclear localization signal KKRVHR identified in human UKLF is also conserved in mKlf7 (residues 214-219), just upstream of the first zinc finger (Matsumoto et al., 1998).

To test whether the mKlf7 gene encodes a nuclear protein, the full-length mKlf7 cDNA was fused in frame with a DNA fragment encoding the green fluorescent protein (GFP). Cellular localization of fluorescence and DAPI staining of transfected HEK293 and NG108 cells (Fig. 2). GFP-Klf7 proteins accumulated exclusively in the nuclei of both HEK293 and NG108 cells after transfection, while control GFP proteins were detected in the entire cell bodies of transfected cells. These data confirm that mKlf7 is a nuclear protein.

mKlf7 was initially cloned by screening an embryonic DRG expression library using a mixture of three probes including Ikaros2, MZF2 and HAND binding sites. To determine which cis element binds mKlf7, GST-Klf7C, which contains the C-terminal DNA-binding domain (amino acid residues 213-301) of mKlf7 fused in frame with GST (glutathione-S-transferase), was expressed and purified from E. coli (data not shown). Recombinant proteins GST-Klf7C or GST-p75ICD (GST-p75 intracellular domain) as a control, were incubated with Ikaros2, MZF2 or HAND probes in gel shift reactions (Fig. 3A). As shown, GST-KLF7C binds specifically to the Ikaros2 site but not to either the MZF2 site or the HAND site. Similar to GST-p75ICD, several additional their DNA-binding domains (Georgopoulos et al., 1992; Georgopoulos et al., 1997; Hromas et al., 1991). Furthermore, unlabeled oligonucleotides containing a wild-type Ikaros2 site, but not an MZF2 site, effectively competes with the Ikaros2 probe for binding GST-Klf7C (Fig. 3B). In contrast, Ikaros2 oligonucleotides harboring mutations in the Ikaros2 core sequence that has previously been shown to disrupt in vivo function of the TrkA enhancer (Ma et al., 2000), failed to compete for binding GST-Klf7C (Fig. 3B). These results indicate that the zinc-finger domain of mKlf7 is capable of binding the wild-type Ikaros2 site in vitro. Moreover, the decreased affinity between mKlf7 and the mutant Ikaros site probably accounts for the reduced activity observed for a TrkA enhancer bearing the same mutated Ikaros2 site in vivo.

To further establish the interaction of mKlf7 with the Ikaros2 consensus sequence, a yeast one-hybrid approach was used (Fig. 3C). Triple copies of either Ikaros2, MZF2 or HAND sequences were cloned upstream of a minimal promoter-driven HIS3 reporter gene and the resulting reporter plasmids were integrated into the host strain YM4271. Plasmids encoding either the GAL4 activation domain (Gal4AD) alone or the Gal4AD-Klf7 fusion protein were transformed into these reporter strains and activation of the HIS3 gene was analyzed by growth on SD-Leu-His plates. Gal4AD-Klf7 was able only to specifically activate expression of HIS3 from the Ikaros2 site. Both the MZF2-driven reporter and the HAND-driven reporter exhibited ‘leaky’ expression in the presence of Gal4AD. However, no additional activation was detected in the presence of Gal4AD-Klf7. These results are consistent with the gel shift data and further support that mKlf7 binds the Ikaros2 site of the TrkA enhancer both in vitro and in vivo.

TrkA is specifically expressed in neural crest-derived sensory and sympathetic ganglia during development (Martin-Zanca et al., 1990; Phillips and Armanini, 1996; Schecterson and Bothwell, 1992; Wright and Snider, 1995). Expression of TrkA in DRG begins at E9.5, peaks at E13.5, and remains constant during late development (Martin-Zanca et al., 1990). Transcription activators of the TrkA gene must be expressed in these tissues with appropriate timing. To examine whether mKlf7 could be a physiologically relevant transcription factor for regulating TrkA expression in vivo, expression of mKlf7 was analyzed by in situ hybridization (Fig. 4). Similar to TrkA at E13.5, mKlf7 was strongly and specifically expressed in dorsal root, trigeminal and superior XI/X complex ganglia (superior and jugular ganglia). Expression of mKlf7 at this stage was also apparent in the CNS including developing brain and spinal cord. One key feature of a putative TrkA regulatory gene would be that its expression precedes the onset of TrkA expression. We therefore compared the expression patterns of mKlf7 and TrkA during early embryogenesis (Fig. 5). The earliest expression of TrkA has been reported at E9.5 (Martin-Zanca et al., 1990). We observed expression of mKlf7 at E8.5, preceding that of TrkA by in situ hybridization (data not shown) and by RT-PCR (Fig. 5A). mKlf7 transcripts were present at E8.5 when TrkA transcripts could not be detected. Fig. 5B-I shows oblique transverse sections of an E10.5 embryo and sagittal sections of an E11.5 embryo. At E10.5, mKlf7 is expressed in trigeminal ganglion, the VII-VIII neural crest complex, and the subventricular neuroepithelium of both forebrain and hindbrain regions. TrkA expression is also apparent in trigeminal ganglion, the VII-VIII neural crest complex, and the subventricular neuroepithelium of the hindbrain. At E11.5, the expression of TrkA and mKlf7 is stabilized in the neural crest derived sensory nervous system (compare Fig. 4 with Fig. 5). Thus mKlf7 gene expression coincides with that of TrkA in neural crest-derived structures as well as in the neuroepithelium of the hindbrain, an area of TrkA expression that has not been previously described. With progressing development, TrkA expression becomes restricted with no detectable expression in neuroepithelium after E10.5 and reduced expression in maturing cranial Ganglia VII and VIII (Fig. 6A,B). At E15.5, mKlf7 exhibits continued coexpression with TrkA but also retains expression in additional neural regions including brain, spinal cord, retinal neuroepithelium and the inferior XI/X complex (nodose and petrosal ganglia) (Fig. 6C,D).

An additional functionally prominent area of TrkA expression is in sympathetic ganglia that acquire NGF/TrkA dependence for survival in late development (Crowley et al., 1994; Levi-Montalcini, 1987; Smeyne et al., 1994). We found that mKlf7 was coexpressed with TrkA in sympathetic ganglia at P0 (Fig. 6E-H). TrkA expression is maintained in adult neural crest-derived sensory and sympathetic ganglia (Kaplan et al., 1991; Phillips and Armanini, 1996). To examine whether mKlf7 coexpression is also retained, serial sections from trigeminal ganglia of adult mice were hybridized with TrkA and mKlf7 probes, respectively (Fig. 7A,B). The expression patterns suggest that mKlf7 is coexpressed with TrkA in sensory nociceptive neurons. In addition, mKlf7 is apparently expressed in all other neurons that do not express TrkA, but not in Schwann cells. Indeed, similar experiments using TrkB-(Ntrk2 – Mouse Genome Informatics) and TrkC-(Ntrk3 – Mouse Genome Informatics) specific probes demonstrate that mKlf7 is also expressed in TrkB-positive neurons and TrkC-positive neurons (data not shown). An alternative way to examine the scope of mKlf7 expression in sensory ganglia was to generate single cell cDNA libraries from DRG neurons cultured in either NGF or NT3. In culture, NGF supports the survival of TrkA-expressing nociceptive neurons, while NT3 supports only TrkC-expressing proprioceptive neurons (Liebl et al., 1997). As shown in Fig. 7C, G3PDH and a panneuronal marker NCAM (Ncam – Mouse Genome Informatics) are detected in all neuronal libraries but not in the control libraries, while TrkA and TrkC expression display appropriate NGF or NT3 dependence, further validating the reliability and specificity of this approach. CGRP is a marker for TrkA-positive neurons. mKlf7 transcripts are present not only in NGF-dependent, TrkA- and CGRP-positive neurons, but also in NT3-dependent, TrkC-expressing neurons.

To gain clear confirmation about the scope of mKlf7 expression in sensory neurons, compared with TrkA, we made use of TrkA-null mice that lose all TrkA-expressing, NGF-dependent sensory neurons (Smeyne et al., 1994). TrkA expression is completely abolished in TrkA-null DRG whereas mKlf7 expression is retained at considerably reduced levels (Fig. 8). Consistent with Fig. 7, these data lead us to conclude that mKlf7 is expressed in all DRG and trigeminal sensory neurons including TrkA-positive neurons.

We next examined mKlf7 expression in adult tissues by northern blot analysis. mKlf7 transcripts are detected in brain, heart, spleen and lung but not in liver, skeletal muscle, kidney or testis (Fig. 9A, and data not shown). The size of the mKlf7 mRNA is approximately 8.5 kb, similar to that of human UKLF mRNA. Fig. 9B demonstrates that mKlf7 transcripts are present in various regions of adult brain and spinal cord. To gain a better understanding of mKlf7 expression in adult brain and to examine whether coexpression with TrkA is maintained in the CNS, we performed comparative in situ hybridization (Fig. 9C-F). As previously described (Holtzman et al., 1992; Sobreviela et al., 1994), cholinergic neurons of the basal forebrain express TrkA (Fig. 9C,E). The identity of these TrkA-expressing cells was confirmed by positive staining with anti-ChAT (choline acetyltransferase) antibodies (data not shown). However, mKlf7 expression is not detected in this region (Fig. 9D,F). The lack of mKlf7 expression in mature basal forebrain cholinergic neurons implies that mKlf7 may not be involved in the expression of TrkA in the CNS. However, it remains to be determined whether mKlf7 is present in basal forebrain cholinergic neurons during the generation and early maturation of these cells.

Consistent with the northern blot analysis (Fig. 9B), mKlf7 is detected in many areas of the CNS, including the cortex, hippocampus and cerebellum by in situ hybridization (Fig. 9G,H). As in the PNS, mKlf7 is mostly, if not exclusively, expressed in neurons in the CNS. For example, in cortex (Fig. 9G), mKlf7 is expressed strongly in neuronal layers II-VI but weakly in layer I, which contains mostly non-neuronal cells. In the cerebellum, the mKlf7 transcript is abundant in granular neurons but absent in the white matter (Fig. 9H).

Taken together, the present results identify mKlf7 as a TrkA enhancer-binding protein that is confined to the nervous system during development. In both developing and mature nervous systems, mKlf7 gene is expressed primarily in neurons, and in all TrkA-expressing neurons in the PNS.

Although the highly stereotypic expression of the TrkA-NGF receptor was reported more than a decade ago (Martin-Zanca et al., 1990), until recently, little was known about the molecular basis for regulation of its expression (Ma et al., 2000). Even less information has been available about the transcription factors that regulate such temporal- and spatial-specific expression. Brn3a is the only transcription factor previously identified to modulate TrkA expression (Huang et al., 1999). The available information implicates Brn3a in maintenance of TrkA gene expression in sensory neurons but not in initiation of expression, as TrkA expression initiates normally in Brn3a knockout mice (Huang et al., 1999). In addition, no consensus Brn3a-binding site is present in the TrkA minimal enhancer that is sufficient for driving specific expression of TrkA in sensory and sympathetic neurons (Ma et al., 2000).

Other candidate genes for regulating TrkA expression may include transcription factors that are important for the generation and differentiation of sensory and sympathetic neurons that eventually become NGF dependent for their survival. Several bHLH genes including Mash1 (Ascl1 – Mouse Genome Informatics) neurogenin 1 (Ngn1; Neurod3 – Mouse Genome Informatics), neurogenin 2 (Ngn2; Atoh4 – Mouse Genome Informatics), and the homeobox gene Phox2b (Pmx2b – Mouse Genome Informatics) are essential for the generation of sensory or sympathetic neurons (Fode et al., 1998; Guillemot et al., 1993; Ma et al., 1998; Ma et al., 1999; Pattyn et al., 1999). Mash1 is expressed in sympathetic precursors and no sympathetic neurons are generated in Mash1-null mice (Guillemot et al., 1993; Johnson et al., 1990; Lo et al., 1991). Sympathetic development in these mutant mice is arrested at embryonic day 10.5 (E10.5) (Guillemot et al., 1993). Similarly, Phox2b is expressed in sympathetic ganglia during early development and no sympathetic neurons are formed in Phox2b null mice at E13.5 (Pattyn et al., 1999). Because TrkA is not expressed in wild-type sympathetic neurons until E17, the failure of sympathetic neuron generation in Mash1- and Phox2b-null mice must occur in a TrkA-independent manner. Furthermore, neural crest-derived sensory neurons are normal in Mash1- and Phox2b-null mice (Guillemot et al., 1993; Pattyn et al., 1999). These results suggest that it is unlikely that either Mash1 or Phox2b regulates TrkA expression. Ngn1 and Ngn2 are expressed in sensory ganglia during early gangliogenesis (Ma et al., 1999). In DRG,

Ngn2 is transiently expressed from E8.75 to E10.5, while Ngn1 is expressed from E9 to E13, overlapping with the generation of TrkA-positive DRG neurons. Ngn1 is required for generation of TrkA-positive neurons, whereas Ngn2 is required for the generation of TrkB-positive and TrkC-positive neurons (Fode et al., 1998; Ma et al., 1998; Ma et al., 1999). It is possible that the loss of TrkA-positive DRG neurons in Ngn1-null mice occurs in a TrkA-independent fashion and therefore the apparent lack of TrkA expression in these mutant mice may be a consequence of general neuronal death rather than transcriptional downregulation of TrkA. To date, no compelling data exist to determine whether Ngn1 has a direct or indirect impact on TrkA expression.

The bHLH gene NeuroD (Neurod1 – Mouse Genome Informatics) is also expressed in DRG and trigeminal ganglia during early development (Lee et al., 1995). NeuroD is implicated in differentiation of neurons and pancreatic beta cells and NeuroD-null mice die shortly after birth, owing to severe neonatal diabetes (Lee et al., 1995; Naya et al., 1997). It is unknown whether sensory neurons depend on NeuroD for differentiation, although NeuroD is required for differentiation of cerebellar and hippocampal granular neurons (Miyata et al., 1999). Another bHLH gene dHAND (Hand2 – Mouse Genome Informatics) is expressed in sympathetic neurons during early development (Srivastava et al., 1995). dHAND is sufficient to induce formation of sympathetic neurons when it is expressed ectopically in chick embryos using retroviruses (Howard et al., 2000). However, it is unknown whether TrkA is regulated by dHAND, although our previous functional analysis indicates a role for HAND-related transcriptional factors in TrkA regulation (Ma et al., 2000).

In this study, mKlf7 was identified as a TrkA enhancer-binding protein. The N-terminal domain of human UKLF, identical to the same region of mKlf7, is a potent transactivation domain (Matsumoto et al., 1998). Although both Ikaros2 and MZF2 sites share similar core sequences for transcription factors containing Kruppel-type zinc fingers (Georgopoulos et al., 1992; Georgopoulos et al., 1997; Hromas et al., 1991), mKlf7 does not interact with the MZF2 site. This is probably due to the diverging sequences flanking the core binding sites. This result demonstrates the specificity of the interaction between mKlf7 and the Ikaros2 cis element. Moreover, the decreased interaction between mKlf7 protein and the mutant Ikaros2 site is consistent with the severely decreased activity observed for the TrkA enhancer bearing the same mutation in vivo (Ma et al., 2000). Apparently, mutations introduced into the core sequence of the Ikaros2 cis element did not completely disrupt the interaction between mKlf7 and the mutated binding site (Fig. 3B). Therefore, it is possible that the TrkA enhancer bearing mutations that abolish the interaction between the Ikaros2 cis element and mKlf7 may be completely inactive in vivo.

Kruppel-like factors constitute a family of transcription factors defined by their conserved C-terminal DNA-binding domains, each consisting of three Cys₂His₂ type of zinc fingers that are also present in the Drosophila melanogaster developmental regulator Kruppel (Schuh et al., 1986; Turner and Crossley, 1999). The GC-box-binding protein Sp1 was the first mammalian Kruppel-like factor identified (Kadonaga et al., 1987). So far, at least 18 different mammalian Kruppel-like factors have been reported (Turner and Crossley, 1999; Uchida et al., 2000). Related transcription factors have also been found in Saccharomyces cerevisiae and Caenorhabditis elegans (Turner and Crossley, 1999). Many transcription factors containing Kruppel-type zinc fingers have also been reported but not included in the Kruppel-like factor gene family because of the different number and location of the zinc fingers in these proteins.

Although similar in their DNA-binding domains, Kruppel-like factors are divergent in other regions. Furthermore, they differ in their expression patterns in vivo (Turner and Crossley, 1999). This family of transcription factors plays important roles in the control of tissue-specific genes as well as ubiquitously expressed ‘housekeeping’ genes. The physiological functions of several Kruppel-like factors have been directly demonstrated by gene knockout studies. Sp1 is widely expressed and Sp1-null mice are embryonic lethal with reduced expression of target genes encoding thymidine kinase and methyl-CpG-binding protein MeCP2 (Kadonaga et al., 1987; Marin et al., 1997). EKLF (erythroid Kruppel-like factor; Klf1 – Mouse Genome Informatics) is highly expressed in erythropoetic tissues and EKLF-null mice develop fatal anemia (Perkins et al., 1995). EKLF is an activator of the β -globin locus control region and essential for the erythropoietin-induced hemoglobin production in vivo (Gillemans et al., 1998; Spadaccini et al., 1998). LKLF (lung Kruppel-like factor; Klf2 – Mouse Genome Informatics) is expressed at high levels in lung but also abundant in many other tissues. Besides being required for lung development, LKLF is also important for T-cell survival and blood vessel formation (Kuo et al., 1997a; Kuo et al., 1997b; Wani et al., 1999). GKLF (gut-enriched Kruppel-like factor; Klf4 – Mouse Genome Informatics) is expressed in epithelial cells of the epidermis and essential for the barrier function of the skin (Segre et al., 1999; Shields et al., 1996). Therefore, the phenotypes of the Klf gene family knockout mice are consistent with the predominant expression sites of these genes in vivo.

Although UKLF, the human homolog of mKlf7, is widely expressed in all human tissues examined (Matsumoto et al., 1998), our study identifies mKlf7 as the first mammalian Kruppel-like factor that is specifically expressed in the nervous system during development.

Our results indicate mKlf7 expression precedes TrkA expression. From E10.5 to adulthood, mKlf7 is coexpressed in all sensory and sympathetic ganglia with TrkA. Of particular note, two novel areas that transiently express TrkA are identified in this study: the VII-VIII neural crest complex at E10.5 and E11.5, and the spinal neuroepithelium at E10.5. Strikingly, mKlf7 is coexpressed in these sites, again supporting the idea of a functional link between expression of this transcription factor and TrkA. The in situ data using serial sections of adult trigeminal ganglia and the single neuron expression study clearly demonstrate that mKlf7 is coexpressed with TrkA in nociceptive sensory neurons. It is also apparent that mKlf7 is expressed in other sensory neurons that do not express TrkA. This is supported by the reduced expression of mKlf7 in the TrkA-null DRG.

In summary, the specific interaction between mKlf7 protein and the Ikaros2 regulatory element of the TrkA enhancer, and the striking coexpression patterns of mKlf7 and TrkA in sensory and sympathetic neurons support the hypothesis that mKlf7 may directly regulate TrkA expression in vivo. Because mKlf7 is also expressed in areas that do not express TrkA, other potential target genes of mKlf7 remain to be identified. In particular, the presence of mKlf7 in TrkB- and TrkC-expressing sensory neurons may indicate a role for this transcription factor in the regulation of all Trk gene family members in sensory neurons. In vivo mutational, as well as molecular and genetic studies will determine the scope of the biological functions of mKlf7 during neural development.

Recent work by Laub et al. describes similar expression results to those found in this study (Laub et al., 2001).

We thank Drs Steven Kernie and Mario Romero, and other members of the Parada lab for helpful discussions. We thank Dr Francesco Ramirez and colleagues for sharing their unpublished results. This work was supported by NIH Grants R01NS33199 and R37NS331999 to L. F. P. L. L. is a recipient of a National Research Service Award from NIH.