When juvenile interneurons arrive at the cortical environment following tangential migration, they are faced with the task of positioning themselves in cortical space in preparation for local circuit wiring. This includes integration into different cortical layers and cessation of migration at various positions to ensure adequate coverage. Little is known about the signals or mechanisms that initiate a conversion from the migratory phenotype to the arborization phenotype. This study looks at the immediate changes in interneuron morphology after culturing for 24 h in a three-dimensional collagen gel. Immature interneurons taken from different stages of corticogenesis showed increased neurite branching and outgrowth after interneuronal contacts were made. These responses were suppressed in the presence of Slit and brain-derived neurotrophic factor (BDNF) if the interneurons were sourced from early to mid-stages of corticogenesis. However, interneurons taken from the late period of corticogenesis responded to Slit and BDNF by increasing branching and neurite outgrowth. These results suggest an initial interneuronal cell contact as a stimulus for propagating neuronal arborization that may lead to the formation of inhibitory neuronal circuits. In addition, we have identified the late corticogenetic period when interneurons are most sensitive to the neurite promoting effects of Slit and BDNF.
It is now clear that the developing neocortical space is shared by two different and quite independent programs of neurogenesis. These processes result in the production of neurons with contrasting phenotypes. Excitatory neurons are generated locally in the neocortical neuroepithelium and they comprise the dominant phenotype with pyramidal, or stellate shaped cell bodies, and spine-bearing dendrites (Conti et al., 1989; Dori et al., 1989; DeFelipe and Farinas, 1992). Excitation of these neurons lead to glutamate release at asymmetric synaptic contacts on their postsynaptic targets, either locally or in distant cortical and subcortical areas (Streit, 1984; Carder and Hendry, 1994). On the other hand, inhibitory interneurons are generated remotely in the subcortical ganglionic eminence (Marin and Rubenstein, 2001), from where they migrate tangentially to contribute up to 20% of all neocortical neurons (Fitzpatrick et al., 1987; Ren et al., 1992; Micheva and Beaulieu, 1995). Interneurons have smooth or sparsely spiny dendrites that form symmetric synapses on their target cells and they utilize γ-aminobutyric acid (GABA) as the neurotransmitter (Fagg and Foster, 1983; Jones et al., 1987; Lund and Lewis, 1993).
GABAergic interneurons are distributed in all layers of the gray matter where they form intricate networks of synaptic contacts with both pyramidal and nonpyramidal neurons (Somogyi et al., 1983, 1998; Meskenaite, 1997). These local neuronal circuits are important for the proper function of mature cortical neurons. For instance, GABAergic inhibition of pyramidal neurons serves to prevent excessive firing in response to peripheral stimuli (Benardo and Wong, 1995). However, nonpyramidal innervation of other nonpyramidal neurons plays an important role to inhibit inhibitory neurons (Tamas et al., 1998) and the effect of this disinhibition is thought to synchronize pyramidal cell activity (Jefferys et al., 1996). Indeed, interconnected GABAergic neurons have been shown to exert synchronous output activity in modulating the operation of pyramidal neuron networks (Traub et al., 1996; Galarreta and Hestrin, 2001). The generation of synchronicity requires the formation of GABAergic networks between developing interneurons and the types of GABAergic interneurons involved have been defined in vivo and in vitro (Tamas et al., 1998; Beierlein et al., 2000; Voigt et al., 2001; Amitai et al., 2002). However, little is known concerning the generic and specific mechanisms that promote arborization of immature interneurons after their migration into the neocortex. For example, at what stages in development are GABAergic interneurons most actively involved in neurite outgrowth and branching? What is the influence of interneuron cell to cell contact on these processes? What chemotrophic factors may influence these activities?
The purpose of the current study is to follow the early events of interneuron neurite outgrowth and branching in a three-dimensional collagen gel environment. The collagen gel system provides neurite outgrowth conditions that mostly close mimics conditions in vivo (Harris et al., 1985). Previous in vitro studies have established that the differentiation of GABAergic neurons closely resembles the development of their in vivo counterparts, suggesting that tangential migration per se is not absolutely necessary and the in vitro system provides an accessible model for studying interneuron arborization (de Lima and Voight, 1997). By combining a low cell density culture environment with the collagen gel matrix, it is possible to directly visualize individual neurites (axons and dendrites) from the cell body to the growth cone, permitting individual branches to be correctly assigned to the appropriate cell body (Wang et al., 1999). The current study covers the entire neurogenetic period when immature GABAergic interneurons are known to be entering the cerebral wall. Using the three-dimensional gel system, we report the development of branch points and neurite outgrowth between interneurons of various developmental stages. In addition, we compare the effects of interneuron cell–cell contact with single interneurons that did not make contact. Finally, we report on the effects of Slit protein and brain-derived neurotrophic factor (BDNF), two known chemobranching molecules, on these interneurons.
Materials and Methods
Timed pregnant mice (C57BL/6J) were anesthetized with pentobarbitone sodium (Nembutal 60 mg/kg, Rhone Merieux, Melbourne, Australia) and brains from different embryonic and postnatal stages (E13.5, 15.5, 17.5, P1.5) were dissected out. The neocortical wall was dissected from the underlying basal telencephalon and following washing in phosphate-buffered saline, the tissue was triturated and dissociated in 0.1% trypsin for 30 min. This was followed by centrifugation and the cells resuspended (at 5 × 106 cells/ml) in growth media before plating onto collagen-coated dishes at 105 cells per cm2. Collagen-coated culture dishes were prepared by restoring acidic rat tail collagen solution (4 mg/ml, Roche Diagnostic, Mannheim, Germany) with 5X DMEM (Gibco, Life Technologies, Australia) and 0.2 M NaOH. The gel was then diluted to 1 mg/ml with the growth media. Growth media consist of Dulbecco’s modified Eagle’s medium (DMEM, Gibco) containing 10% fetal bovine serum (FBS), 100 unit/ml penicillin and streptomycin.
In experiments to test the effects of Slit protein, growth medium was used to culture human embryonic kidney (HEK) 293 cells that had been stably transfected with Xenopus Slit (xSlit) carrying the six-myc epitope tag or with control vector plasmid (Li et al., 1999). Conditioned medium was collected after 72 h from cells expressing Slit or control plasmid, and used to reconstitute the collagen gel cultures. In testing for the effects of BDNF, a recombinant protein (gift from Dr T. Hughes, University of Melbourne) was added to growth medium at a final concentration of 5 ng/ml.
Cultures were incubated at 37°C in a CO2 atmosphere for 24 h before fixation and processing for GABA immunocytochemistry. Primary antibody against GABA was raised in rabbit (Sigma, Australia) and used at a 1:200 dilution. GABA immunoreactivity was revealed by an Alexa-conjugated donkey anti-rabbit IgG (1:200, Molecular Probes, Eugene, OR). To reveal cell nuclei, cells were stained with bisbenzamide.
Fluorescent images of individual neurons were captured (×40 objective) with a Spot Digital Camera (Diagnostic Instrument Inc., Sterling Heights, MI) and the magnified images used for analysis of neurite length and branching. Only neurons with unambiguous extension of neurites from the cell soma were included in the analysis (Image-Pro-Plus software; Media Cybernetics, Silver Spring, MD). The length of the longest neurite, presumed to be the axon, was calculated from the captured image. Neurites projecting from the soma were individually counted as branches, as were neurites found extending from other shafts. In each category, 120 GABAergic neurons were randomly selected for analysis until this number was reached. For studying the effects of interneuron cell to cell contact, 60 pairs of GABA+ cells were randomly chosen and the same measurements applied. Statistical comparisons between different age groups were performed using the one-way ANOVA test. To compare neurite length and branching between contacting and non-contacting GABAergic interneurons, the Student’s t-test was used.
GABA immunocytochemistry has been shown to be a sensitive and reliable method for studying the development of GABAergic interneurons in vitro (Stichel and Muller, 1991; de Lima and Voight, 1997). The use of a three-dimensional collagen gel culture overcomes the disadvantages that are associated with two-dimensional substrates such as the non-physiological induction of multipolar neurons and growth cone bifurcations (Harris et al., 1985). Using this system, we have examined the short-term (24 h) behavior of cultured GABAergic interneurons isolated from the developing neocortex at different stages.
Interneuron Branching and Neurite Outgrowth Is Promoted by Cell–Cell Contact
GABA immunoreactivity was detected in immature neurons from as early as E13.5 onwards following 24 h culture (Fig. 1A). By using a low to medium plating density, the morphology of individual GABA immunoreactive neurons could be easily observed. Interneurons were frequently present as single cells but in many cases, pairs of contacting GABAergic neurons were also seen (Fig. 1A–D). Their cell bodies were invariably ovoid-shaped or fusiform, and the overall morphology suggests a migrating phenotype with dissimilar neurites resembling leading and trailing processes (Fig. 1A–D). Overall, there was no overt differences in the appearance of contacting (Fig. 1E, white cell bodies), compared to non-contacting neurons (Fig. 1E, dark cell bodies). For non-contacting GABA neurons, the number of branch points per neuron did not change appreciably between the various time-points examined, varying between two and three branch points per neuron (Fig. 1F, bottom curve). Similarly, non-contacting neurons showed an average longest neurite lengths of ∼30 μm although for the age group P1.5 this had increased to ∼40 μm (Fig. 1G, bottom curve).
A different picture was presented by contacting GABAergic interneurons. Except for the E15.5 cohort, all other contacting interneurons displayed an increased number of branch points per cell that was statistically significant compared with non-contacting cells (Fig. 1F, upper curve with asterisks; P < 0.05–0.001). It was unclear why the E15.5 group of interneurons, although increased in branching, was not statistically significant. With respect to neurite length, the longest neurite in a contacting GABA cell was significantly greater than that of its non-contacting counterpart (Fig. 1G, upper curve with asterisks, P < 0.05–0.001). Although not the case with neurite branching, neurite length was increased with greater developmental stages (Fig. 1G). Taken together, these observations suggest that cultured interneurons respond during the first 24 h by increasing neurite length and branching but these effects were most overt among contacting interneurons. No efforts were made to distinguish between the effects of cell body contact or neurite contact. However, to rule out the possibility that neurons with longer processes may be more likely to contact with other long process neurons, we compared pairs (n = 60) of contacting long neurons (process > 50 μm) with pairs of short neurons (process < 50 μm). If process length was the determining factor, then the associated phenotype of increased branching should only be found among long process neurons because short neurons would make contact less frequently in our culture system. This analysis showed that the number of branches per neuron did not vary between contacting long neurons (mean 2.9 branches) versus contacting short neurons (mean 2.7 branches) (P > 0.1). In contrast, comparison of contact short neurons (process < 50 μm) with non-contacting short neurons (n = 60) showed a significant difference in the number of branches per neuron (P < 0.01). These results indicate that the observed increases in neurite length and branching are secondary to cell–cell contact.
The Effects of Exogenous Slit Protein on Neurite Branching and Outgrowth
In a previous study, we demonstrated that old but not young interneurons in vitro respond to Slit by increased branching and neurite lengths (Sang et al., 2002). Young interneurons, isolated from E13.5 embryos displayed suppressed neurite lengths and branching in respond to Slit protein present in conditioned media. In contrast, old interneurons (in vitro equivalent of E18.5 cells) showed increased neurite outgrowth and branching following exposure to Slit protein for 24 h. This Slit-mediated activity was attenuated in the presence of excess quantity of the Slit-receptor, RoboN. We concluded that Slit protein is an important accessory for interneuron arborization among mature, but not juvenile, GABAergic neurons.
In the present work, we extended our study by examining cultured interneurons taken from four different stages, and also compared the effects of Slit protein on contacting versus non-contacting interneurons. As before, contacting interneurons increased their number of branch points and have greater neurite lengths (Figs 2A,D,E and 3A,D,E), but these effects in the E13.5, E15.5 and P1.5 cohorts were counteracted in the presence of Slit protein (Figs 2B,D,E and 3B,D,E). In contrast, the E17.5 cohort showed a totally different picture in the presence of Slit (Fig. 3B,D). In this age group, Slit protein promoted neurite branching and neurite outgrowth in both contacting and non-contacting interneurons (Figs 3D and 4A,B). Thus it would appear that a positive response to Slit protein was only found within interneurons isolated from E17.5 embryos, not younger or older stages. In the cortical plate, Slit1 is expressed from E15.5 onwards and is present until birth (Whitford et al., 2002).
The Effects of Exogenous BDNF on Neurite Branching and Outgrowth
BDNF has been shown in other studies to regulate GABAergic interneuron development, including the promotion of neurite outgrowth and their maturation, stimulation of GABA expression and other calcium-binding proteins (Huang et al., 1999; Mizuno et al., 1994; Widmer and Hefti, 1994). In the presence of BDNF, branching activity among E13.5 and E15.5 interneurons was suppressed. This was observed with both contacting and non-contacting interneurons in these two age groups (Figs 2C,D,E and 4A). Neurite length was slightly increased, but not significantly (Fig. 4B). In this respect, BDNF differed from Slit in promoting neurite extension despite inhibiting neurite branching (Fig. 4A,B). In contrast, the E17.5 cohort of interneurons showed a different response to BDNF. In the E17.5 group, branching activity in response to BDNF was significantly increased, exceeding the effects seen with Slit (Fig. 4A). After birth, BDNF did not elicit a similar branching activity from P1.5 interneurons (Fig. 4A). Paradoxically, BDNF had an opposite effect on neurite outgrowth of E17.5 neurons (Fig. 3D). So whilst E17.5 neurons displayed maximal branching in the presence of BDNF (Fig. 3C,D), they showed less neurite outgrowth compared with neurons cultured in the presence of Slit (Fig. 4A,B). A similar suppression of neurite extension was also seen with P1.5 cells (Fig. 4B). Thus, immature interneurons responded differently to BDNF depending upon their age of culture. Younger interneurons responded to BDNF by suppression of neurite branching whilst E17.5 interneurons responded by increasing branching activity but decreasing neurite outgrowth.
We have shown that single immature GABAergic interneurons, plated at low to medium density, did not arborize or grow their neurites appreciably at early to mid-stages of corticogenesis when they are known to be migrating into the neocortical environment. However, this was altered in situations where cultured interneurons were found in contact with one another. At all stages examined (with the possible exception of E15.5 cohort), the number of branch points and neurite length were significantly increased compared to their non-contacting counterparts during the 24 h culture period. It is not clear why values for the E15.5 population, although showing a slight increase, were not significant. In terms of the number of additional branches acquired by contacting neurons during the assay period, the embryonic age did not seem to make a big difference. On average, one additional branch per neuron was observed in all age groups. This is not to say that had the observation period been extended to days or even weeks, the number of additional neurites, or increases in neurite length, would still be the same (de Lima and Voight, 1997). What the current data suggest is that cell to cell contact among immature interneurons, after 24 h in vitro, is a stimulus for interneuron branching and neurite outgrowth. This observation is consistent with other reports that interneuronal contacts may lead to increased rates of neurite extension (Fletcher et al., 1994; van den Pol et al., 1998).
The question then follows regarding the nature of the stimulus and how this may lead to the back propagation of trophic signals to the cell machinery. One possibility is for interneuronal cell contact to establish gap junction coupling that may facilitate the passage of micromolecular signals between cells (Connors et al., 1983). These signals may initiate cellular changes leading to increased neuronal arborization. These may also represent the initial steps of establishing interneuronal circuits between GABAergic neurons (Tamas et al., 1998; Voigt et al., 2001). Another possibility is for interneuronal contact to trigger receptor–ligand complexes which in turn may lead to transcriptional changes favoring neurite outgrowth. A classic example of this is found in the neocortical pyramidal neurons grown at high density. Following neurite to neurite contact, there is activation of the Notch receptor–ligand complex leading to the generation of signals to inhibit further neurite outgrowth (Sestan et al., 1999).
A third possibility concerns the GABA neurotransmitter. There is evidence that GABA plays a trophic role in the immature brain (Lauder et al., 1986), including stimulation of cell migration (Behar et al., 1998), and regulation of its dendritic morphology via GABA transmission and release (Matsutani and Yamamoto, 1998). Activation of GABA receptors in migrating interneurons causes a rise in intracellular calcium (Soria and Valdeolmillos, 2002). The precise mechanisms are unclear but it has been suggested that GABAergic transmission causes BDNF release from target cells which have paracrine effects on interneuron soma size and dendritic arborization (Marty et al., 1996). Both BDNF and its receptor TrkB are expressed in the rodent cortical plate from the earliest phases of cortical neurogenesis (Fukumitsu et al., 1998). BDNF has been found to promote dendritic elongation of interneurons (Matsutani and Yamamoto, 1998), but the responses are not uniform, with increases in dendritic arbors among 1 week old interneurons but quite the reverse in older interneurons (Marty et al., 1996). In the current study, BDNF has the general property of increasing neurite length among interneurons at embryonic but not at postnatal stages. However, the standout effect of BDNF is found among E17.5 interneurons where the branching effect is maximal, exceeding the effect of Slit. By P1.5, the branch-promoting effects of BDNF was no longer present. The precise reason for the stage-specific effect of BDNF at E17.5 is unclear, but this stage is coincident with the invasion of thalamic axons (that lie waiting in the underlying subplate) into the cortical plate (Molnar and Blakemore, 1995). This leads us to propose the hypothesis that thalamic innervation of maturing pyramidal neurons in the cortical plate may create an environment whereby interneurons are stimulated to extend their arbors. Following contact with target pyramidal neurons, transient GABA transmission from interneurons may then elicit BDNF release from the target cells, providing further trophic stimuli for local circuit arborization.
Slit belongs to a family of secreted proteins with chemorepellant activity and is known to be expressed in the striatal– cortical regions through which interneurons migrate (Nguyen Ba-Charvet et al., 1999; Yuan et al., 1999; Whitford et al., 2002). In addition to a chemorepulsive role for interneuron migration (Zhu et al., 1999), Slit has been shown to repel extending GABAergic neurites (Sang et al., 2002). In addition, we and others have shown that Slit has a chemobranching effect on cortical interneurons (Sang et al., 2002; Whitford et al., 2002). In agreement with our previous study, the current results emphasize that not all interneuron cohorts respond in a similar fashion to Slit. Whereas interneuron cultures established from early stages (E13.5 and 15.5) showed suppressed neurite outgrowth, interneurons taken from E17.5 brains responded positively by increasing arborization. The effect was not seen at postnatal P1.5.
Taken together, the above results indicate increased neurite growth and arborization following contact between neighboring cells in culture. Interneurons from different developmental stages respond differently to BDNF and Slit, with repression of branching at earlier time-points but accelerated branching and neurite extension at E17.5. Translating these findings to the in vivo cortical environment would suggest a critical period during late corticogenesis when interneurons are most responsive to chemotrophic stimuli. This period corresponds with the near-completion of cortical layering and arrival of most interneurons following long distance tangential migration (Anderson et al., 2001). One may therefore hypothesize that the capacity to respond to local chemotrophic signals is delayed until E17.5 when the full complement of interneurons have filled the neocortical space and nearest-neighbor distances between interneurons are undergoing adjustment. Therefore inappropriate arborization before E17.5 would exclude late-arriving interneurons from tiling the cortical landscape. This developmental mechanism has been shown to occur for amacrine cells, another interneuron population in the retina known to respond to short-range interactions during the creation of neuronal mosaics (Galli-Resta et al., 1997). While there is still no evidence that cortical interneurons obey similar rules for mosaic arrangement, they are known regularly dispersed and the spatial rules for electrical coupling between interneurons are beginning to emerge (Amitai et al., 2002). In this context, neurite extension and branching from the immature interneuron may be mechanisms for positional adjustment and sampling of neuronal space.
We thank Tony Hughes for the gift of BDNF. Cells expressing Slit were kindly donated by Yi Rao and Jane Wu. This research was supported by the National Health and Medical Research Council (Institute Block Grant and Peter Doherty Fellowship to Q.S.) and a grant from The Myer Foundation.