Diverse dependencies of developing Merkel innervation on the trkA and both full-length and truncated isoforms of trkC

During development of the peripheral nervous system, neurotrophins contribute to such diverse functions as cell proliferation, neuronal lineage differentiation, neuronal survival, neurite outgrowth and synapse formation. In mammals, nerve growth factor (NGF) and three related neurotrophins [brain-derived neurotrophic factor (BDNF), neurotrophin 3 (NT3) and neurotrophin 4 (NT4)] signal through three related tyrosine kinase receptors (trkA, trkB and trkC) that can exist in multiple isoforms (reviewed by Reichardt and Fariñas, 1997; Huang and Reichardt, 2001). Full-length kinase isoforms (trkA_K, trkB_K and trkC_K) have tyrosine kinase signaling domains, while truncated isoforms of trkB and trkC (trkB_Tr and trkC_Tr) lack kinase domains. All four neurotrophins also bind to a non-kinase receptor, p75, that has a small signaling domain similar to that of tumor necrosis factor receptors.

Phenotypes of mice with targeted null mutations or transgene enhancements indicate that development and maintenance of all cutaneous innervation is dependent on one or more of the four neurotrophins and four receptors (Ip et al., 1993; Albers et al., 1996; Pinon et al., 1996; Davies, 1997; Fundin et al., 1997; Rice et al., 1998; LeMaster et al., 1999). Dependency of some innervation on more than one neurotrophin or receptor (Fundin et al., 1997; Rice et al., 1998) is consistent with observations that many neurons in dorsal root and trigeminal ganglia can express more than one receptor either simultaneously or sequentially during development [reviewed by Reichardt and Fariñas (Reichardt and Fariñas, 1997)] (Buchman et al., 1994; Snider and Wright, 1996; Huang et al., 1999). Dependence on more than one neurotrophin-receptor interaction is supported by the observation that knockouts of the kinase domain in either trkA_K or trkC_K (trkA_K^–/– or trkC_K^–/–; Ntrk1^–/– and Ntrk3^–/– – Mouse Genome Informatics) each result in a loss of substantially more than half of the neurons in dorsal root or trigeminal ganglia. In addition, absence of NT3 (Ntf3^–/–) results in a greater loss of neurons than absence of trkC_K whose only known ligand is NT3. Presumably, the excess loss is due to possible interactions of NT3 with trkA_K and trkB_K.

Based on such results, hypotheses have been proposed about the types and sequences of receptors that should be expressed during the development of specific types of cutaneous innervation (Fundin et al., 1997; Rice et al., 1998). However, direct evidence for expression of these receptors on specific types of definitive cutaneous sensory endings has not been shown during their development. An especially interesting model for exploring complexities of neurotrophin and related receptor dynamics during development is the Merkel innervation located at the mid-follicle level of rodent mystacial vibrissae (Fundin et al., 1997; Rice et al., 1997). This innervation consists of large caliber myelinated axons, whose endings penetrate the basement membrane of the follicles and terminate on Merkel cells located among the keratinocytes in the outer root sheath. As shown schematically in Fig. 1, either trkA_K^–/– or trkC_K^–/– substantially reduced the number of Merkel endings and cells present at birth (P0). However, many endings and cells remain for at least four weeks postnatally in trkA_K^–/– mice, whereas in trkC_K^–/– virtually all of the Merkel endings disappear by P7 and all of the Merkel endings and cells are absent by P14. By contrast, Ntf3^–/– is relatively more severe with only a few Merkel endings and cells present at P0. All of the Merkel endings and nearly all of the Merkel cells disappear by P7. Ngf^–/– also had a partial reduction of Merkel innervation that was less severe than that in trkA_K^–/– mice. These observations led to the hypothesis that the mid-follicle Merkel innervation initially depends upon trkA_K and trkC_K prenatally, but becomes entirely dependent on trkC_K postnatally (Fundin et al., 1997). NT3 appears to be the primary ligand involved in the development and maintenance. Although the Merkel innervation appeared to be intact at P0 in p75^–/– (Ntfr^–/– – Mouse Genome Informatics) mice, a gradual loss over several weeks suggested that p75 contributes to long term maintenance perhaps by facilitating NT3 signaling through trkC_K.

One objective of our study was to determine if trkA, trkC and p75 receptors are expressed on this mid-follicle Merkel innervation in mice in a sequence consistent with these hypotheses and if development of this Merkel innervation is indeed totally dependent on both trkA_K and trkC_K. As rats are more amenable for experimental studies involving nerve manipulations, behavioral assessments and neurophysiology, expression of receptors was also assessed in rats to see if their normal receptor expressions are comparable with those in mice where the genetic manipulations have been performed.

An especially puzzling aspect of neurotrophin research is the functional role of the truncated isoforms (Tr) of the trk receptors (Klein et al., 1990; Middlemas et al., 1991; Valenzuela et al., 1993; Tsouflas et al., 1993; Garner and Large, 1994). Mice containing null mutations targeted at the ligand-binding extracellular domain locus (trkC_E^–/–) appeared to eliminate both the trkC_K and trkC_Tr isoforms, and resulted in a more severe neuronal losses than trkC_K^–/– (Tessarollo et al., 1997). However, the specific impact of the trkC_E^–/– on cutaneous innervation was not explored in detail. Interestingly, in situ hybridization results indicate that trkC_Tr isoforms are expressed primarily on non-neuronal tissues of fetal mystacial pads instead of trigeminal neurons (Wheeler et al., 1998). These observations indicate that expression of trkC_Tr isoforms in non-neuronal tissues may contribute to development of the innervation. Therefore, the second objective of our study was to determine whether trkC_Tr also contributes to development of the Merkel innervation.

Normal fetal and postnatal Swiss-Webster mice were obtained after overnight matings (Taconic Farms). At least five specimens from three different pregnancies were obtained at consecutive daily intervals ranging from embryonic day E9.5 to postnatal day P14. E0.5 was regarded as noon on the first day after the mating, P0 was the morning of the day of birth. Fetal and postnatal Sprague Dawley rats were also obtained from overnight matings (Taconic Farms). At least five specimens from three different pregnancies were obtained at consecutive daily intervals ranging from embryonic day E10.5 to postnatal day P14.

Fetal and newborn trkC_K^–/– or trkC_E^–/– mice, as well as comparable age wild-type mice were obtained from overnight matings of adult male and female trkC_K^+/– (129/c57bl/6) (Klein et al., 1994) or trkC_E^+/– (129/c57bl/6) (Tessarollo et al., 1997). For each type of mutation, a minimum of five homozygous knockouts and five wild-type mice were obtained at E14, E15 and P0. At least two P0 trkA_K^–/–/trkC_K^–/–, trkA_K^+/+/trkC_K^–/–, trkA_K^–/–/trkC_K^+/+ and trkA_K^+/+/trkC_K^+/+ hybrids were obtained from matings of male and female trkA_K^+/–/trkC_K^+/– (129/c57bl/6) (Smeyne et al., 1994; Klein et al., 1994). The fetal and postnatal offspring were genotyped by PCR. As the original observation (Fundin et al., 1997) of the impact of Ntf3^–/– was based on mice with a different background from the trkC-related knockouts used in this study, P0 and P7 Ntf3^+/+ and Ntf3^–/– mice were obtained from matings of Ntf3^+/– parents with the same background (i.e. 129/c57bl/6) (L. T., unpublished). Mice with a heterozygous mutation in which the coding region of the lacZ gene replaced a coding exon for NT3 and bred on a c57bl/6 background were mated to obtain E14, E15 and E16 embryos that were Ntf3^+/+, Ntf3^+/lacZ and Ntf3^lacZ/lacZ (Fariñas et al., 1994). Animals were genotyped by DNA blot analysis as described (Fariñas et al., 1994).

Fetuses were obtained from dams by cesarean section after lethal anesthetic overdose (100 mg/kg sodium pentobarbital). Embryos were removed and whole embryos or heads were fixed by immersion for 4 hours in 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS) at pH 7.4 and 4°C. Postnatal specimens were anesthetized by hypothermia (P0-P5) or anesthetic overdose (after P5) and were perfused transcardially and postfixed for 4 hours in the above fixative. After fixation, specimens were rinsed and stored in PBS at 4°C. The relative status of mystacial pad development was used to correct for variations in the rate of development between comparable age specimens from different pregnancies and genetic backgrounds.

Fixed whole fetuses, heads or mystacial pads were cryoprotected by overnight infiltration in 30% sucrose in PBS and frozen sectioned (14 μm) in a plane parallel to the central rows of vibrissa follicles and perpendicular to the skin surface. Sections were directly thawed onto slides coated with chrome-alum gelatin and air dried overnight.

Sections were prepared for immunofluorescence analyses (Rice et al., 1997) by incubation overnight at 4°C under primary antibodies against pan-neuronal protein gene product 9.5 (PGP) (rabbit polyclonal, 1:1000, UltraClone); calcitonin gene-related peptide (CGRP) (rabbit polyclonal, 1:1000; Peninsula); neurofilament 200 (NF200) (rabbit polyclonal, 1:400; Chemicon International), keratin 20 (mouse monoclonal, 1:50; Chemicon International), trkA (rabbit polyclonal, 1:5000) (Huang et al., 1999); trkC (goat polyclonal, 1:500) (Huang et al., 1999) and p75 (rabbit polyclonal, 1:5000) (Weskamp and Reichardt, 1991). The trkA and trkC antibodies were generated against extracellular domains and label both kinase and truncated isoforms. After rinses, sections were incubated for 1 hour at room temperature in appropriate species of secondary antibodies raised in goat or donkey and conjugated with either Cy3 or Cy2 fluorophores (1:250-500: Jackson Immuno Research Laboratories). All incubating and rinse solutions consisted of 1% bovine serum albumin and 0.3% Triton X-100 in PBS. Sections were examined with an Olympus Provis AX70 epifluorescence microscope and digital images were captured with a Sony DKC-ST5 digital photo camera interfaced with Northern Eclipse imaging and deconvolution software (Empix Imaging).

Presence of mRNAs for all isoforms of trkC or just trkC_K was determined by in situ hybridization in sections of fetal mouse and rat mystacial pads (Wheeler et al., 1998). Rat trkC probes were prepared from a pBlue-script vector containing the full-length cDNA sequence [nucleotide 1 to 2003: gift of L. Parada (Tsouflas et al., 1993)]. This antisense trkC probes and encodes trkC_K and recognizes mRNAs for both trkC_K and trkC_Tr isoforms. TrkC riboprobes that recognize only trkC_K mRNAs were generated from the 125 bp fragment encoded within the kinase domain downstream from sequences that encode the alternatively spliced isoforms. The gene sequence runs from nucleotide 1908 to a site located in the 3′ untranslated region at nucleotide 2033. The construct was made by subcloning an AccIII-AvaI restriction fragment into the pGEM 3Z in vitro transcription vector (Promega, Madison, WI).

The RNA probes were labeled with [³⁵S]-UTP by using an in vitro transcription system and polymerases (Promega, Madison, WI). Each template (500 ng) was transcribed. Reactions were incubated at 37°C for 2 hours. Resulting RNA transcripts were hydrolyzed to an average length of 150 bp. Hybridization reactions and washes were carried out as described previously (Wheeler et al., 1998). Processed slides were dipped in photographic emulsion (NTB from Eastman Kodak, Rochester, NY), exposed at 4°C for 10 days to 3 weeks, developed, fixed and counterstained with Methyl Green. Sections were examined under both bright-field and dark-field illumination.

Mid-follicle Merkel endings are the first innervation to develop on the nascent vibrissa follicles. Paralleling a caudal to rostral sequence of follicle development (Fig. 2), arriving axons completely engulf the deep end of the developing follicles beginning at E12.5 in mice and E13.5 in rats (not shown). As follicles elongate, these axons shift to the dorsal side and become limited to the mid-follicle level (Fig. 2). Beginning caudally at E13.5 in mice and E14.5 in rats, axons penetrate the basement membrane and begin to form endings among adjacent follicle cells (Fig. 2). Within a day or two, the cells that the axons terminate among become immunoreactive for anti-PGP (Fig. 2B, Fig. 3A,E) and anti-keratin 20 (not shown) which are indicative of differentiated Merkel cells (Rice et al., 1997). By E14.5 in mouse and E15.5 in rat, Merkel endings and cells are well established in the most caudal follicles, whereas axons have only begun to engulf the most rostral follicles (Fig. 2).

We previously hypothesized that this developing Merkel innervation should initially co-express trkA_K and trkC_K but should only express trkC_K postnatally (Fundin et al., 1997). Immunolabeling revealed that axons growing towards, but not yet contacting nascent whisker follicles label with either anti-trkA or anti-trkC, not both (not shown). Detectable immunoreactivity for these two receptors remains on axons and on the earliest definitive Merkel endings for several days (Fig. 2C-F, Fig. 3B,C). Contrary to our hypothesis, double labeling revealed that trkA and trkC immunoreactivity is expressed on completely separate sets of axons and Merkel endings (Fig. 4A,B). The trkA labeling on the one set of Merkel innervation diminishes to barely detectable levels by E17.5 in the mouse (data not shown) and E19.5 in the rat (Fig. 2D) and is undetectable by birth in both species (E18.5 and E21.5 respectively). The trkC labeling on other Merkel axons and endings continues into the postnatal period and becomes barely detectable by the end of the first postnatal week.

Although our original knockout study only detected a detrimental impact of p75^–/– on the Merkel innervation well after birth (Fundin et al., 1997), anti-p75 labeling is robust on growing nerves before they reached nascent follicles and on all of the developing and definitive Merkel-related axons and endings throughout the prenatal and postnatal period (Fig. 2G,H, Fig. 3D). Whereas, trkA and trkC labeling was limited to axons and endings, p75 labeling is robust on accompanying Schwann cells at all prenatal and postnatal ages examined, and becomes expressed on Merkel cells soon after they differentiate.

The expression of anti-trkA and anti-trkC labeling on separate sets of Merkel innervation is consistent with the reduction of Merkel endings at birth in trkA_K^–/– and in trkC_K^–/– mice (Fig. 1, Fig. 4C-E). In order to determine if all Merkel innervation is dependent on either trkA_K or trkC_K, mystacial pads were examined from newborn offspring from crosses of trkA_K^+/–/trkC_K^+/–. Mystacial pads of newborn trkA_K^–/–/trkC_K^–/– mice completely lacked Merkel innervation; however, some Merkel cells were still present (Fig. 4F). These results indicate that trkA_K accounts for the development of one set of Merkel innervation and trkC_K accounts for the remaining Merkel innervation.

TrkC immunoreactivity was also present on all of the developing follicle cells on the day that the Merkel endings begin to form (E13.5 in the mouse and E14.5 in the rat) and diminishes substantially over the next 2 days (Fig. 2E, Fig. 3E, Fig. 4A,B). As seen along the caudal-to-rostral gradient within the same mystacial pads, the onset of trkC expression within the follicles slightly precedes detectable penetration of axons into the developing follicles (Fig. 2E). By the time trkC labeling disappears among the follicle cells, the Merkel endings are well established and Merkel cells are immunodetectable with anti-PGP (Fig. 2B, Fig. 3A) and anti-keratin 20 (not shown). By E16.5 in mouse and E17.5 in rat, non-neuronal trkC labeling has shifted to the upper dermis between the vibrissa follicles (e.g. E19.5 in Fig. 2F). Interestingly, at the time trkC is expressed within the follicle, an intense complimentary expression of p75 occurs in the dermal cells immediately surrounding the developing follicles (Fig. 2D,G).

Our in situ hybridization analysis revealed that anti-trkC labeling on follicle cells in E14.5 mice was due to transient expression of trkC_TR (Wheeler et al., 1998). Riboprobe that detects all isoforms of trkC revealed high levels of hybridization within vibrissa follicles (Fig. 5B), whereas no trkC message was detected on follicle cells with the kinase probe (Fig. 5A). Similar results were obtained on E15 rat follicles, although low levels of transcripts encoding kinase isoforms were also present (data not shown).

The timing of trkC_Tr expression in developing follicles suggested that it may be involved in the development of the Merkel innervation. To test this possibility, the impact of trkC_K deletion was compared with trkC_E deletion, as trkC_K deletion removes only the tyrosine kinase domain of trkC_K, whereas the trkC_E deletion should eliminate all known isoforms of trkC (Tessarollo et al., 1997). Newborn trkC_E^–/– lacked any Merkel endings or Merkel cells (Fig. 6E,F), whereas many endings and cells were present in newborn trkC_K^–/– (Fig. 6C,D) (Fundin et al., 1997). Surprisingly, the impact of trkC_E deletion was not only more severe than trkC_K deletion but was also more severe than Ntf3 deletion (Fig. 6A,B). In Ntf3^–/– mice, a few Merkel cells and some innervation are present at birth. The original impact of Ntf3 deletion (Fundin et al., 1997) was assessed in a different strain of mice (outbred from 129SV/ter and BALB/c) from those used to assess trkC_K deletion and trkC_E deletion (129/c57bl/6). Our results revealed the impact of Ntf3 deletion was the same on the 129/c57bl/6 background. Thus, different phenotypes of Ntf3^–/–, trkC_K^–/– and trkC_E^–/– are not due to differences between mouse strains.

Examination of fetal E15.5 trkC_E^–/– (Fig. 3I,L) revealed that axons had grown to developing vibrissa follicles but, unlike in trkC_K^–/– or Ntf3^–/– fetuses (Fig. 3E, Fig. 7I), no axons penetrated the basement membrane and formed endings. Moreover, trkC_E^–/– fetuses lacked any evidence of Merkel cell differentiation in the follicles. Consistent with the fact that trkC_K^–/– is targeted at the kinase domains and may not eliminate truncated isoforms, immunolabeling was observed on the follicle cells of E14.5 and E15.5 trkC_K^–/– with trkC antibody (Fig. 3G), which had been generated against the extracellular domain of rat trkC. In addition, mRNA was detected in trkC_K^–/– with the riboprobe detecting all isoforms of trkC (Fig. 5C) but not with the kinase probe. Consistent with the fact that trkC_E^–/– should eliminate the extracellular domains of truncated and full-length isoforms, no obvious non-neuronal trkC immunolabeling was detected in trkC_E^–/– mutants (Fig. 3K) and no hybridization was obtained with the riboprobes for all isoforms or for only kinase isoforms (Fig. 5D). Both trkC_E^–/– and trkC_K^–/– fetuses had normal looking non-neuronal labeling with anti-p75 (Fig. 3H,L).

No trkC immunolabeling of any innervation was detected in trkC_E^–/– or trkC_K^–/– fetuses on E14.5-E16.5 (Fig. 3G,K) indicating that innervation at these ages lacks detectable levels of naturally occurring truncated isoforms and that the kinase mutations did not result in the formation of truncated isoforms. In these fetuses, axons projecting to the vibrissa follicles were labeled with anti-p75 and anti-trkA but Merkel endings were only seen in trkC_K^–/– follicles (Fig. 3F,G,J,K).

One possible role of transiently-expressed trkC_Tr within the developing follicles could be to sequester and increase the local concentration of NT3, thereby promoting the development of the Merkel endings. Examination in Ntf3^+/lacZ and Ntf3^lacZ/lacZ fetuses (Fig. 7) revealed intense β-galactosidase expression in several layers of dermal cells intimately surrounding all the nascent follicles on E13, prior to invasion of axons into caudal follicles and formation of Merkel endings (see rostral follicles at E14 in Fig. 7A,D). Concurrent with the caudal-to-rostral sequence of trkC immunolabeling throughout the follicles, a dense expression of the reporter construct also appears within the follicles along the same gradient. However, unlike trkC_Tr which is expressed throughout the follicle (Fig. 7A-C), the Ntf3 reporter construct is restricted to the mid-follicle level where the Merkel endings develop (Fig. 7B,C,E,F). Thus, the site of NT3 production is naturally more precisely restricted to the site of Merkel ending development than trkC_Tr. The presumptive production of NT3 in this mid-follicle location appears to be brief because the reporter gene expression is barely detectable by E15 (Fig. 7G-J), even before trkC_Tr labeling completely disappears within the follicles (Fig. 7H).

In contrast to the complete absence of Merkel ending and Merkel cell development in trkC_E^–/–, analysis of the Ntf3^lacZ/lacZ E14 and E15 specimens revealed that at least some Merkel endings and cells were able to develop without NT3 (Fig. 7I) as had been seen in the neonates. Some of this successful innervation labeled with anti-trkA (Fig. 7G) but more was revealed with anti-p75 and anti-PGP (Fig. 7D,E,I). Anti-trkC labeling of innervation was substantially reduced in Ntf3^+/lacZ fetuses (Fig. 7C) and undetectable in Ntf3^lacZ/lacZ (Fig. 7A,B). This suggests that many axons are present that lack either trkA or trkC immunoreactivity. Therefore, many axons that would normally express anti-trkC labeling may have downregulated trkC_K production, but are still present. The presence of p75 may be contributing to their development. As was also seen with trkC_E^–/– and trkC_K^–/– (Fig. 3), Ntf3^lacZ/lacZ did not have an obvious effect on patterns of non-neuronal labeling with anti-trkC (Fig. 7I) or anti-p75 (Fig. 7D-E). However, compared with heterozygotes, homozygous mutants have smaller follicles and the onset of vibrissa shaft differentiation was delayed (Fig. 7B,C,E,F).

Several important findings from this study are summarized in Fig. 8. First, some developing Merkel endings express and depend upon trkA_K, while others express and depend upon trkC_K. Thus, sensory neurons forming the same type of endings on the same target can express different receptors and depend upon a different combination of neurotrophins. Second, NT3 is briefly produced precisely at the time and place where Merkel endings develop. Thus, production of a neurotrophin can be regulated with a high degree of precision at the time and location where a set of innervation will terminate. Third, while some Merkel cells and endings develop in the absence of trkC_K, transient expression of trkC_Tr within follicles is essential for development of any Merkel cells and endings. This indicates that non-neuronal expression of trkC_Tr can play an essential developmental role for specific target cells and innervation. Fourth, the impact of trkC_E^–/– on Merkel innervation and Merkel cells was more severe than absence of NT3, the only known ligand for trkC. This suggests that the essential developmental role of trkC_Tr involves more than interactions with NT3. Finally, the same combinations of neuronal and non-neuronal receptors are expressed in mouse and rat, indicating that similar developmental mechanisms are occurring in both species for the comparable innervation.

As has been shown previously, Merkel innervation is partially reduced at mid-follicle levels in newborn trkA_K^–/– and trkC_K^–/– mice (Fundin et al., 1997). Complete absence of this Merkel innervation in newborn trkA_K^–/–/trkC_K^–/– hybrids indicates that each ending depends on either trkA_K or trkC_K for prenatal development or survival. Consistent with this dual impact, Merkel axons and endings expressed either trkA or trkC, but not both. Likewise, developing trigeminal ganglion neurons rarely if ever co-express trkA and trkC (Huang et al., 1999). Importantly, our results indicate that Merkel innervation can be supplied by neurons that have two different receptor and neurotrophin dependencies.

Presence of reduced Merkel innervation in trkA_K^–/– newborns and complete absence in trkA_K^–/–/trkC_K^–/– newborns indicates that some Merkel innervation never solely requires trkA_K, but does require trkC_K for its development or prenatal maintenance. This set of innervation presumably is included in the increased apoptosis that occurs in Ntf3^–/– or trkC_K^–/– mice. Likewise, other sets of Merkel innervation, which develop later at the mouths of hair follicles, also require trkC_K, but not trkA_K (Fundin et al., 1997).

The surviving mid-follicle Merkel innervation in trkA_K^–/– mutants presumably is the innervation that expresses trkC_K. In trkA_K^–/– mutants, surviving innervation is maintained through at least the fourth postnatal week (Fundin et al., 1997). Although trkC is immunodetectable postnatally on Merkel endings, it drops below detectable levels within a few weeks of birth. As trkC_K levels decline, prolonged survival of this innervation depends upon p75 (Fundin et al., 1997), which was detected throughout development of the Merkel innervation and at least four weeks postnatally.

The presence of a reduced set of Merkel endings in newborn trkC_K^–/– is consistent with our observation that some endings have detectable trkA immunoreactivity prenatally but not trkC. Presumably the trkA-expressing innervation is the set that is lost in trkA_K^–/– mutants. Thus, a unique set of Merkel innervation may initially develop in response to either NGF or NT3 signaling through trkA_K. Consistent with such a trkA_K dependency, a similar loss of Merkel innervation occurs in Ngf^–/– mutants (Fundin et al., 1997) although not quite as severe as in trkA_K^–/– (F. L. R., unpublished). This suggests that NT3 signaling through trkA_K could partially compensate for lack of NGF (White et al., 1996). The likelihood that NT3/trkA_K signaling normally does contribute to development or maintenance of some Merkel innervation is indicated by the more severe impact of Ntf3^–/– when compared with trkC_K^–/–. Nonetheless, even though expression of Ntf3/lacZ construct indicates that NT3 is intensely produced at mid-follicle levels where Merkel endings develop, clearly some of the innervation requires NGF at some point, possibly early in development.

Unlike innervation that survives without NGF or trkA_K, surviving Merkel innervation seen at birth in trkC_K^–/– mice deteriorates over the first 2 postnatal weeks. As trkA drops below immunodetectable levels just prior to birth, postnatal deterioration in trkC_K^–/– suggests that the trkA_K-dependent innervation may switch to a dependency on trkC_K for maintenance postnatally. Interestingly, neither Ntf3^–/–, trkC_K^–/– or trkC_E^–/– obviously increase the labeling intensity and proportion of axons expressing trkA immunoreactivity. In all three knockouts, trkA labeling remains relatively faint. Moreover, numerous other axons only label with anti-PGP and p75. Thus, trkA_K is limited to a subset set of Merkel innervation and other sets do not appear to upregulate trkA_K to compensate for the loss of trkC_K.

Surprisingly, during Merkel ending development in Ntf3^–/– fetuses, a substantial proportion of axons that are both trkA and trkC immunonegative still labels with anti-p75. However, non-neuronal trkC labeling within and around the follicle is not obviously affected. This suggests that absence of NT3 eliminates trkC_K expression on many Merkel axons but these axons are not initially lacking in the Ntf3 knockouts (Wyatt et al., 1999). However, most of the innervation disappears before birth. Thus, NT3 or trkC_K may not necessarily be essential factors for early survival of the Merkel innervation that will eventually become dependent on NT3 and trkC_K (Wright et al., 1992).

Other studies have demonstrated that sensory neurons forming similar types of endings can depend upon different receptors and/or neurotrophins. For example, innervation of muscle spindles in limb musculature appears to have a different neurotrophin dependency from that of spindles in muscles of mastication (Tessarollo et al., 1994). In addition, BDNF and NT3 have both been shown to play a role in development of taste bud structure and innervation, but different components in different types of taste buds (Nosrat et al., 1997). Although, BDNF and NT3 are required for development of hair cell innervation at apical and basal ends of the cochlea, respectively (Fritzsch et al., 1999), recent evidence indicates that all of this innervation expresses trkB and trkC and that different neurotrophin survival dependencies are due to opposing gradients of BDNF and NT3 expression within the developing cochlea (Fariñas et al., 2001). By contrast, our results reveal that the same type of sensory endings (Merkel endings) terminating in the same target structure (a vibrissa follicle) can depend upon different trk receptors (trkA or trkC). Conceivably, this difference may be related to the fact that Merkel endings at the mid-follicle level of rat vibrissae can have two different types of terminal arborizations (Fundin et al., 1994; Ebara et al., 2002).

Our results revealed that no Merkel endings or cells begin to form in trkC_K^–/– fetuses although axons do converge at mid-follicle levels. Thus, development of Merkel endings as well as Merkel cells is completely dependent on trkC_Tr transiently expressed in all of the developing follicle cells. Importantly, at least some Merkel cells are present at birth, even though absence of trkA_K and trkC_K completely eliminates their innervation. Conceivably, some Merkel endings may still develop and disappear prenatally in trkA_K^–/–/trkC_K^–/–.

Surprisingly, the impact of losing both the truncated and kinase isoforms of trkC was more severe than absence of trkC_K alone or NT3, the only known trkC ligand. Thus, the trkC_Tr appear to play an active developmental role beyond that of trkC_K. This trkC_Tr function is at least partially independent of NT3. In this particular system, this role does not appear to be either a dominant-negative (Eide et al., 1996; Gonzalez et al., 1999; Eggert et al., 2001) or NT3 sequestering effect (Fryer et al., 1997; Das et al., 2000; Lin et al., 2000; Liebl et al., 2001). Conceivably trkC_Tr may have constitutively active signaling that is necessary for development of Merkel ending and cells, or may have a signaling capability mediated by an unknown ligand.

We initially hypothesized that non-neuronal trkC_Tr may be concentrating NT3 within follicles to induce formation of Merkel endings (Rice et al., 1998). Our results indicate that NT3 is abruptly and transiently produced within follicles precisely at the time and place where Merkel endings develop. Merkel cells become immunodetectable shortly thereafter. Conceivably, trkC_Tr expressed throughout follicles may be required to induce production of NT3 at this precise mid-follicle location. The loss of some Merkel innervation in Ngf^–/– mice, coupled with formation of some Merkel endings in the absence of NT3, indicates that NGF may also be expressed in this location and that complete failure of ending formation in trkC_E^–/– mutants may be due to a loss of both NGF and NT3 at this site. Importantly, the mid-follicle site of NT3 production is more precise than the distribution of trkC_Tr within the follicles. Thus, if trkC_Tr does play an essential role in the production of neurotrophins, another factor must be regulating the specific site of NT3 production.

We wish to thank Marilyn Dockum for help with tissue preparation, and Drs Sally Temple, Chris Ingraham and Ann Messer for critiques of the manuscript. This research was supported by NIH grants R01 NS34692 (F. L. R.) and P01 NS 16033 (L. F. R.), SCORE GM08194 (E. F. W.) and predoctoral training grant T32 EY 07120 (G. A. W.); by NSF IBN-9723990 (E. F. W.); and by the Howard Hughes Medical Institute (L. F. R.).