We report that BDNF acts directly as a chemoattractant for geniculate neurites at embryonic stages corresponding to initial targeting, but by P1, this chemoattractive influence is lost. These results provide crucial support for the possibility that BDNF is an attractant for geniculate axons during initial innervation in vivo [4
]. We also conducted dose-response studies using BDNF. These studies demonstrated that there is a concentration optimum for stimulating neurite extension in the case of BDNF. In addition to expanding our previous studies on embryonic geniculate ganglia, this is the first study to report the growth-promoting effects of BDNF on postnatal geniculate neurites.
In the absence of added neurotrophins, E15 and E18 ganglia did not extend neurites and no or little neuronal staining was observed within explants, indicating that most neurons did not survive for 2 days. P1–7 ganglia extended thin neurites under these conditions. The ganglia from postnatal and older rats are larger than those at embryonic stages, owing largely to the proliferation of nonneuronal support cells, which may provide more trophic support than is available at embryonic stages. In contrast to infant ganglia, adult ganglia exhibited little neurite growth in the absence of added neurotrophic factors, even when Matrigel (which contains a variety of neurotrophic factors) was used as the matrix and the culture time was doubled. Our unpublished observations indicate that the decline in spontaneous outgrowth is apparent after the end of the first postnatal week. This correlates with the opening of taste pores in fungiform papillae [44
], but whether increased taste bud signaling induces this change in neurotrophin independence remains to be determined. The decline in neurite extension with adult explants may be due to a loss in intrinsic neuronal factors that are critical for regeneration, a change in substratum preference, or an alteration in trophic support provided by nonneuronal cells within ganglion explant [48
Dose-response studies with bath-applied BDNF indicated that neurite length and morphology are affected by the concentration of BDNF. Intermediate concentrations of BDNF (25–50 ng/ml) resulted in longer neurites than lower and higher concentrations. Inhibition of outgrowth by high concentrations of nerve growth factor (NGF) and neurotrophin-3 (NT-3) was reported for other populations of sensory neurons and was interpreted as possibly reflecting desensitization [49
]. Consistent with this possibility, continuous infusion of BDNF into the rat hippocampus decreased tyrosine kinase B (trkB) levels [52
]. Note that BDNF was infused at much higher levels (1 mg/ml) than those used in this study. A recent study examined the influence of applying 25 ng/ml BDNF acutely or introducing it stepwise slowly to hippocampal slice cultures [53
]. Total trkB levels were unaltered with both protocols, but acute application decreased surface trkB and triggered only a transient increase in trkB phosphorylation. Slow application did not decrease surface trkB levels and led to a maintained increase in phosphorylated BDNF. Although the collagen in which the ganglia are embedded has the indicated concentration of BDNF from the time of plating, whether peripheral neurons recovering from dissection respond similarly to hippocampal neurons that have been cultured for 2 weeks prior to BDNF treatment remains to be determined. High levels of BDNF may also disrupt motility-promoting interactions between the growth cone and other cells (including neurites) or the collagen I gel matrix [54
]. In support of the latter possibility, there is clearly a change in morphology of the outgrowth that presumably reflects altered interactions between the growth cones and their surroundings: at E15 and E18, outgrowth promoted by intermediate concentrations of BDNF was segregated into discrete fascicles, whereas outgrowth in the presence of high concentrations of BDNF emanated from the ganglion in a more continuous, web-like pattern (fig. ). Compared to embryonic ganglia, P1–7 ganglia were less affected by high concentrations of BDNF (100–200 ng/ml) both in terms of the morphology and the length of outgrowth. Outgrowth at postnatal stages was also significantly longer than that observed for embryonic cultures. It will be important to determine whether embryonic and postnatal neurons differ in their levels of trkB, p75, or BDNF.
We used in vitro assays to demonstrate that BDNF is a chemoattractant for embryonic geniculate neurites. When ganglia were dissected at E15, BDNF-soaked beads promoted outgrowth that was nearly exclusively directed toward the bead. The bias was also evident if 25 ng/ml of BDNF was present in the bath, indicating that neurites can distinguish gradients over a wide range of concentrations despite a high background level of BDNF. BDNF continued to act as an attractant at E18, although the outgrowth bias ratio was less than at E15. At both E15 and E18, neurite outgrowth toward the BDNF bead was faster than growth away from the bead, even when the same amount or less BDNF was encountered. Despite this robust bias, the majority of neurites do not make sharp turns toward the bead. A recent analysis of neurite extension in the presence of gradients of NGF that differ in slope suggests that persistent turning toward a source of a chemotropic factor is primarily observed when the slope of the gradient is high (enabling growth cones to reliably distinguish concentrations on each side of the growth cone) [55
]. Mortimer et al. [55
] proposed an alternative mechanism, “growth rate modulation”, to explain highly biased neurite outgrowth characterized by inconsistent turning responses distributed over gradients with low slopes. In this model, turning occurs randomly, but turns toward the source result in more rapid advance than turns away from the source. Our observations are consistent with this model. We also demonstrated that neurites do not grow beyond BDNF beads at these stages. These observations indicate that the tropic influence of BDNF dominates the trophic influence at initial targeting/target penetration stages. These results differ from those obtained in a study of the effects of NGF gradients on mouse trigeminal neurite outgrowth [49
]. That study showed that trigeminal neurite outgrowth was biased toward NGF beads (at low doses) only at the equivalent of E15; no bias was observed at the equivalent of E18, just after target contact. This difference could be attributable to species or neuronal type, but it may also be related to different functions of these neurotrophins. NGF is present along the pathway explored by many trigeminal afferents and is critical for maintaining axon growth (trophism) [56
], whereas BDNF is implicated as a target-derived attractant.
One cannot assume that in vitro results are relevant to the in vivo setting. All of our tropism data were gathered using collagen I gels and serum-containing medium, which is undoubtedly different from the in vivo setting. The responses of growth cones to attractants and repellents are influenced by other diffusible molecules and by the substratum [54
], so this is not a trivial concern. It is equally unreasonable to dismiss in vitro studies as artifacts with no relevance to the in vivo phenomenon. As discussed next, our current results support a straightforward interpretation of the in vivo neurotrophin perturbation studies. Substantial parallels between in vivo and in vitro observations were also found in our previous studies of other diffusible cues [58
]. Our in vitro findings combined with localization data would have predicted the results from in vivo studies.
The bias in outgrowth by BDNF gradients is highest just prior to the invasion of the target epithelium by taste afferents, and persists until shortly after target penetration. Combined with in vivo studies showing that geniculate axons stop extending prior to reaching gustatory papillae in BDNF–/– mice [12
] and misexpression studies that show that geniculate axons invade tissues that overexpress BDNF [4
], our findings support a model in which BDNF directly attracts axons to pregustatory epithelium during the final stage of targeting. None of these observations excludes roles for other, yet to be identified, attractants during intralingual targeting. The existence of other targeting molecules is suggested by the finding that a subset of papillae is properly innervated in mice lacking epithelial BDNF [12
]. However, improper targeting was also observed, so it is possible that proper ‘targeting’ in BDNF–/– mice was due primarily to selective stabilization of randomly targeting geniculate axons.
Two observations raise the possibility that BDNF could act as a stop signal during initial innervation, preventing the axons from continuing to grow once they penetrate the epithelium. E15 and E18 neurites that contacted beads appeared to stop advancing rather than encircling the bead or growing beyond it, and high concentrations (200 ng/ml) of BDNF reduced or eliminated geniculate neurite extension. It seems unlikely that such high concentrations would be encountered in vivo, but it is difficult to determine in vivo concentrations of secreted proteins. Neurotrophic factors may accumulate in the extracellular matrix or be held at high concentration on cell surfaces [25
]. Inhibition of neurite growth was also reported for NGF using mouse trigeminal neurons [49
At postnatal stages, gradients of BDNF do not significantly bias neurite outgrowth length, although it does bias the density of outgrowth. The greater density could result from more neurons extending neurites or the same number of neurons extending more neurites. Regardless, these results suggest that BDNF switches from acting as a chemoattractant to acting as a predominantly trophic factor. Neurotrophic factors promote regeneration following injury through adulthood [61
], so understanding how the response to neurotrophic factors is altered during postnatal development may have therapeutic applications.
We have stressed the correspondence between our findings and observations of the innervation of the tongue, but we note that the geniculate ganglion contains taste neurons that innervate taste buds in the palate, as well as somatosensory neurons that innervate the skin of the outer ear [62
]. Although the distribution of BDNF mRNA in palatal tissues has not been published, innervation of palatal taste buds in BDNF–/– animals is compromised similarly (albeit to a lesser extent) to innervation of lingual taste buds [12
]. It is not known if the geniculate somatosensory neurons respond in the same way to BDNF as the taste neurons, or if they are more similar to trigeminal somatosensory neurons. It seems unlikely that these neurons are closely similar to trigeminal neurons because NGF and NT-3 do not support neurite extension from geniculate explants at targeting stages, but both of these neurotrophins support neurite extension from trigeminal neurons [59
] and are present during and required for normal innervation [6
]. It will be helpful to develop markers for these populations of geniculate neurons so that differences in the responses of their neurites can be studied in vitro.
In summary, our data support a role for BDNF in attracting taste axons into pre-gustatory epithelium during initial innervation. Our data do not support a guidance role for BDNF after birth, nor do they exclude such a role. It will be important to determine whether somatosensory afferents that terminate adjacent to taste buds are also attracted toward the epithelium by neurotrophic factors such as NT-3, which becomes localized in nongustatory lingual epithelium at the time that trigeminal afferents begin to invade. It will also be important to determine whether NT-4, which acts through the same receptors as BDNF, is tropic for geniculate neurites during development. Although NT-4 does not appear to be concentrated in taste axon targets [47
] and innervation is normal for the surviving geniculate neurons in NT-4 knockout mice, overexpression of NT-4 in the lingual epithelium interferes with penetration by geniculate axons [12