|Home | About | Journals | Submit | Contact Us | Français|
Neuronal differentiation is accomplished through cascades of intrinsic genetic factors initiated in neuronal progenitors by external gradients of morphogens. Activity was thought to be important only late in development, but recent evidence indicates that activity also regulates early neuronal differentiation. Activity in post-mitotic neurons prior to synapse formation can regulate phenotypic specification, including neurotransmitter choice, but the mechanisms are not clear. Here we identify a mechanism that links endogenous calcium spike activity with an intrinsic genetic pathway to specify neurotransmitter choice in neurons in the dorsal embryonic spinal cord of Xenopus tropicalis. Early activity modulates transcription of the GABAergic/glutamatergic selection gene tlx3 and requires a variant cAMP response element (CRE) in its promoter. The cJun transcription factor binds to this CRE site, modulates transcription, and regulates neurotransmitter phenotype through its transactivation domain. Calcium signals through cJun N-terminal phosphorylation, thus integrating activity-dependent and intrinsic neurotransmitter specification. This mechanism provides a basis for early activity to regulate genetic pathways at critical decision points, switching the phenotype of developing neurons.
Excitability regulates many aspects of neuronal development in addition to forming the basis for rapid signaling in the mature nervous system 1,2. Key features determining neuronal structure and function are regulated by calcium (Ca) signaling, including cell proliferation and migration, axon guidance, synapse refinement, cell survival and neurotransmitter specification 3–9. Much work has been done to elucidate the way in which synaptic activity regulates these processes through neuronal gene expression 10,11, but less is known about modulation of neuronal development by earlier forms of activity prior to synapse formation. Here we focus on understanding the signaling cascade that mediates Ca spike-dependent specification of neurotransmitter phenotype in embryonic Xenopus spinal neurons.
Specification of neurotransmitters is a crucial aspect of development because it determines the polarity of synapses and the function of microcircuits. Many studies of neurotransmitter specification have focused on transcription factor cascades and downstream gene regulation. Gradients of members of the Shh and Bmp signaling pathways define the dorsoventral axis in the spinal cord and the expression of a combinatorial code of transcription factors that determine cell identity in neuronal progenitors 12,13. As neuronal precursors exit the cell cycle this combinatorial code initiates a cascade of transcription factors that determine the expression of effector genes and launch various programs underlying a neuron’s functional identity, including axon trajectory, neurotransmitter specification and neurotransmitter receptor expression 14–18. Compelling evidence identifies roles for transcription factors in neurotransmitter choice. Ectopic expression of the homeobox gene MNR2 in neural cells can cause motor neuron differentiation, including the expression of acetylcholine 18. Dbx1 knockout mice exhibit an increase in GABA-positive interneurons 19. Gain- and loss-of-function experiments in mice show that transcription factors play a role in the expression of other classical neurotransmitters, glutamate and glycine 20–22. The central role played by transcription factors in neurotransmitter specification was demonstrated by the ability of Tlx3 and Lbx1 to act as switches in determining glutamatergic and GABAergic phenotypes in chick and mouse 23,24.
Activity also regulates neurotransmitter specification, including acetylcholine in mouse hypothalamic neurons 25 and dopamine in mouse petrosal ganglion neurons 26. In Xenopus, Ca spike activity regulates dopaminergic specification in the ventral suprachiasmatic nucleus and specification of four classical neurotransmitters, GABA, glutamate, glycine, and acetylcholine, in the spinal cord, via mechanisms that require transcription 9,27,28. Prior to synapse formation, the frequency of Ca spikes, which involve developmentally transient Ca-dependent action potentials and Ca-induced Ca release, homeostatically regulates specification of neurotransmitters in the spinal cord during a brief critical period 9,29. When Ca spikes are suppressed using molecular or pharmacological approaches, fewer neurons express the excitatory neurotransmitters glutamate and acetylcholine. In contrast, when Ca spiking is increased, more neurons express the inhibitory neurotransmitters GABA and glycine.
Here we provide evidence for a molecular mechanism by which Ca spikes during the critical period cause a change in the neuron’s intrinsic transcriptional cascade specifying GABA and glutamate. The end result is the mature neurotransmitter phenotype of the neuron, determined by both genetic factors and activity. This interplay of intrinsic cellular context and Ca signaling is likely to allow assembly of circuits with functions appropriate to an environment of changing activity.
To discover possible integration points between early activity and genetic pathways regulating neuronal differentiation we sought to identify genes with activity-dependent expression by in situ hybridization for transcription factors involved in neurotransmitter specification. Xenopus tropicalis (X. tropicalis) embryos (Supplementary Fig. 1) 30 were injected in one cell at the two-cell stage with mRNA encoding myc-tagged Kir to suppress Ca spike activity unilaterally 9. Overexpression of this inward rectifier potassium channel hyperpolarizes neurons and suppresses generation of Ca spikes. Ca spike suppression caused an increase in both the area and intensity of expression of the t-cell leukaemia homeobox-3 (tlx3) gene compared to the unmanipulated side of the embryo (Supplementary Fig. 2). tlx3 expression begins in post-mitotic lateral spinal neurons at the neural tube stage. These neurons migrate dorsally and medially, strongly suggesting they are glutamatergic Rohon-Beard sensory neurons and dorsolateral interneurons. tlx3 expression continues throughout the critical period for neurotransmitter specification when spontaneous Ca spike activity is present (Supplementary Fig. 3) 31. To manipulate tlx3 expression we generated splice- and translation-blocking morpholinos (MOs; Supplementary Fig. 4) and an overexpression construct. Knockdown of tlx3 function with MOs caused an increase in the number of GABA-immunoreactive (-IR) neurons and a decrease in the number of vesicular glutamate transporter (VGluT1)-IR neurons in the spinal cord, demonstrating that tlx3 functions as a switch specifying the glutamatergic over the GABAergic phenotype as in mouse and chick 23,24. Overexpression of tlx3 caused the opposite effect on neurotransmitter specification. In addition, the overexpression construct lacks the target sequence for the translation-blocking MO and reversed the MO phenotype when they were co-expressed (Supplementary Fig. 5). Antagonism of lbx1 by tlx3 determines the glutamatergic over GABAergic phenotype in mice 24. Although lbx1 is not co-expressed with tlx3 in Rohon-Beard cells, tlx3 MOs reduced VGluT1 expression in these sensory neurons, demonstrating that lbx1 is dispensable for the glutamatergic phenotype in these cells (Supplementary Fig. 5). tlx1 partially compensates for the loss of tlx3 in mice 24. Similar to mice, tlx1 is expressed in a subset of neurons expressing tlx3 (data not shown), which may account for the VGluT1-IR cells remaining after tlx3 knockdown.
To determine whether tlx3 is an integration point for activity and a genetic pathway specifying neurotransmitter expression, we performed simultaneous manipulations of Ca activity and tlx3 gene function. Overexpressing mRNA encoding Kir to suppress Ca activity produced a decrease in the number of GABA-IR neurons and an increase in the number of VGluT1-IR neurons. MO knockdown of tlx3 generated the opposite phenotype. Simultaneous suppression of Ca activity and knockdown of tlx3 gene function phenocopied the result obtained by tlx3 knockdown (Fig. 1a,b). Overexpressing mRNA encoding Nav to increase sodium channel expression and enhance Ca activity 9 caused an increase in the incidence of GABA-IR neurons and a decrease in the incidence of VGluT1-IR neurons, while overexpressing tlx3 produced the opposite phenotype. Simultaneous enhancement of Ca activity and overexpression of tlx3 phenocopied overexpression (Fig. 1c,d). These results demonstrate that tlx3 is required for activity-suppression-dependent respecification of GABA and glutamate and that overexpression occludes the switch caused by increased Ca spike activity. Furthermore, these data indicate that Ca activity regulates neurotransmitter specification upstream of a crucial genetic choice point.
The interaction between activity and tlx3 in specifying neurotransmitter phenotype makes the regulation of this gene attractive for analysis of the integration of genetic factors and early activity-dependent processes controlling neuronal differentiation. Aligning human and mouse genomic sequences with more distantly related vertebrates has proven successful for identifying important cis-regulatory elements 32,33. We adopted this approach to identify elements that may be important for imparting activity dependence. Comparative genomics revealed that only a single region of the gene is conserved among X. tropicalis, mouse, and human for ~15 kb upstream and ~40 kb downstream of the gene (see Online Methods). 384 bp of the promoter region upstream of the start ATG is 65% conserved including a 152 bp subregion that is 81% conserved (Supplementary Fig. 6a). This 384 bp region drives eGFP expression in the dorsal spinal cord and sensory ganglia, similar to the endogenous expression pattern, demonstrating that it is sufficient to recapitulate native expression of tlx3 (Supplementary Fig. 6b). In addition, expression is observed in dorsal muscle, indicating that the 384 bp region is likely missing an additional negative regulatory element that normally represses expression in this tissue. Because this small region is highly conserved among frog, mouse, and human, we surmised that it may contain an activity-responsive element.
To test this hypothesis, we generated a wildtype reporter construct to express the firefly luciferase gene under control of the conserved 384 bp of the promoter region. This reporter was co-injected with a normalization vector (see Online Methods) and mRNA encoding Kir or Nav to suppress or enhance Ca-activity, respectively. Suppressing Ca-activity increased expression of the luciferase reporter while enhancing Ca-activity decreased reporter expression (Fig. 2a). The smaller change observed when enhancing activity may result from asymmetric effects of spike suppression and enhancement on downstream regulatory elements. No changes in reporter expression were observed when assays were performed on larvae from which the neural tube had been dissected (Supplementary Fig. 6c), indicating that expression of luciferase in other embryonic tissue is not activity-dependent and that the activity-dependent changes observed are occurring in the neural tube. In addition, no changes in reporter expression were observed when luciferase assays were performed on embryos prior to the period of Ca activity responsible for regulating neurotransmitter specification, demonstrating that activity present during earlier periods of development does not influence reporter expression (Supplementary Fig. 6d).
These data demonstrate that transcriptional regulation of the genetic pathway by Ca spike activity could determine neurotransmitter specification and suggest that there may be activity-responsive elements within this promoter. Analysis of the 384 bp promoter region for conserved transcription factor binding sites identified a variant form of the cAMP response element (CRE; TGATGTCA), an activating protein-1 (AP1) site, a signal transducer and activator of transcription (STAT) site, in addition to two nuclear factor Y (NFY) loci previously shown to be important for basal transcription of tlx3 (Supplementary Fig. 6a) 34. The high conservation of this region between X. tropicalis, mouse, and human suggests the importance of these sites for regulation of tlx3 transcription.
CRE and AP1 have both been implicated in activity-dependent transcriptional regulation 11,35. To determine the requirement of these sites for tlx3 regulation by Ca activity, we performed additional luciferase assays with reporters containing either a mutated AP1 binding site (AP1mut reporter; Fig. 2b) or a mutated CRE (CREmut reporter; Fig. 2c). Mutating the AP1 site had no effect, while mutating the CRE abolished the activity-dependence of reporter expression. These data indicate that the CRE is required for activity-dependent regulation of tlx3 and suggest that this site could integrate genetic and activity-dependent specification of glutamatergic and GABAergic phenotypes.
We tested the ability of candidate transcription factors to bind the CRE sequence in the tlx3 promoter using electrophoretic mobility shift assays (EMSAs). Previous work indicates that cJun and ATF2 preferentially bind the variant CRE sequence TGATGTCA 36. Incubating probe containing the CRE and flanking sequence from the tlx3 promoter with recombinant cJun or ATF2 caused a shift in migration of the probe that was blocked with excess unlabeled probe, but not by excess unlabeled probe containing a mutant CRE or a non-specific probe (Fig. 3a,b). Luciferase assays demonstrated that constitutively overexpressing a dominant negative mutant of cJun lacking the transactivation domain (cJunTAM) 37, but not a similar dominant-negative ATF2TAM, caused a CRE-dependent increase in the reporter (Fig. 3c). We confirmed the presence of cJun in the embryonic spinal cord by RT-PCR (Fig. 3d) and immunocytochemistry (Fig. 3e). These data suggest that cJun can act through the variant CRE site to regulate transcription of tlx3.
To avoid early developmental defects caused by constitutive manipulation of cJun with morpholinos, we generated hormone-inducible overexpression constructs for wildtype cJun and the dominant negative mutant cJunTAM. These overexpressed fusion proteins remain inactive until injected embryos are incubated with dexamethasone 38. Inducing overexpressed wildtype cJun decreased wildtype reporter levels in the luciferase assay (Fig. 4a), while induction of cJunTAM increased wildtype luciferase reporter expression (Fig. 4b). The effects were abolished when the CRE site was mutated in the luciferase reporter (Fig. 4a,b). These experiments show that cJun can control tlx3 transcription via the CRE binding site.
We next directly assessed the role of cJun in neurotransmitter specification by overexpressing wildtype or dominant negative cJunTAM and immunostaining for GABA and VGluT1. Inducing overexpressed cJun increased the number of neurons expressing GABA and decreased the number of neurons expressing VGluT1 (Fig. 4c,d), similar to the results obtained for tlx3 knockdown. In contrast, inducing cJunTAM generated the opposite phenotype (Fig. 4c,d), similar to tlx3 overexpression. Thus, both Ca activity and cJun regulate intrinsic neurotransmitter specification by modulating tlx3 transcription via the CRE site located in its promoter.
To find out if cJun integrates genetic and activity-dependent neurotransmitter specification pathways we determined whether Ca spike activity regulates cJun function. The extent of cJun expression was unaffected by manipulation of Ca activity via unilateral injection of mRNA encoding Kir or Nav (Fig. 5a). Phosphorylation of S63, S73, T91 and T93 is the primary basis of regulating the transactivation domain of cJun 39. Because cJun regulation of tlx3 was altered by deletion of the transactivation domain, we ascertained whether Ca activity regulates phosphorylation of amino acids in this region. Immunocytochemistry using phosphospecific antibodies showed that phosphorylation of S73 and T91 was decreased in response to Ca spike suppression and increased in response to Ca spike enhancement (Fig. 5b,c).
Mutations converting all four residues to alanines (cJunmut) reproduced the CRE-dependent increase of tlx3 reporter expression observed for the cJun transactivation domain mutant in luciferase assays (Fig. 6a). In addition, induction of cJunmut increased tlx3 expression observed by in situ (Supplementary Fig. 7). Overexpressing this phosphorylation mutant caused a reduction in GABA-IR and an increase in VGlut1-IR (Fig. 6b,c). These experiments demonstrate the functional role of these phosphorylation sites and identify cJun phosphorylation as the Ca spike entry point in the genetic pathway.
Our results reveal a mechanism by which early activity and genetic factors interact to drive differentiation. The data support a model in which Ca spike activity prior to synapse formation modulates the genetic pathway for specification of neurotransmitter phenotype (Supplementary Fig. 8a). It is currently unclear whether Ca spike activity functions cell-autonomously. Ca spikes could cause phenotypic changes within the spiking cell or they could initiate a signaling cascade to neighbouring cells via diffusible factors. However, the downstream signaling cascade in dorsal neurons initiated by Ca activity is cell-autonomous -- it involves phosphorylation of cJun that regulates tlx3 transcription through a CRE site in its promoter (Supplementary Fig. 8b). cJun appears to function as a repressor of tlx3 transcription. Although cJun is often described as an activator, it functions as a repressor in some cases, operating by recruiting co-repressors to the promoter or sequestering other activation factors 40.
The conservation of the variant CRE site in Xenopus, mouse and human tlx3 suggests that activity-dependent regulation of tlx3 is present in mammals, as well as in X. tropicalis. However, there appear to be differences in the specification of glutamatergic and GABAergic neurons in Xenopus and mammals. In mice, tlx3 functions to antagonize lbx1: in the absence of tlx3 neurons default to a GABAergic fate under the control of lbx1, but when lbx1 is also knocked out, the glutamatergic fate is restored 24. tlx3 and lbx1 appear to be largely non-overlapping in Rohon-Beard cells of Xenopus tropicalis, so this mechanism is not conserved. Another transcription factor may fulfill the same role as lbx1 or tlx3 may act alone in Xenopus. Although there is strong conservation in the promoter sequence of tlx3 across species, other species-specific sites regulating activity-dependence are not excluded.
Strikingly, a strong cell-non-autonomous effect is also observed in both tlx3 loss-of-function and cJun gain-of-function embryos: the increase in GABA-IR cells is greater than the loss of VGluT1-IR cells and there are additional ventral GABA-IR neurons in the tlx3 morphants where tlx3 is not expressed. This result raises the possibility that other mechanisms exist to maintain circuit homeostasis in the developing spinal cord, perhaps by altering neurotransmitter specification in downstream neurons or generating compensatory synaptic connections.
Many other transcription factors are involved in neurotransmitter specification, but their relation to activity-dependent processes is unclear. However, several examples of activity-dependent neurotransmitter regulation involve transcription. Changes in Ca spike activity change the number of neurons expressing tyrosine hydroxylase transcripts in the postembryonic brain 27. Synaptic activation leads to appearance of GAD67 and vesicular GABA transporter transcripts in adult glutamatergic granule cells and GABAA receptor-mediated responses to granule cell stimulation 41. These results suggest that the interplay between activity and genetic pathways in neurotransmitter specification is extensive. Indeed, the difference between the changes in VGluT1-IR and GABA-IR that we observe when manipulating activity is likely caused by the involvement of other transcription factors, which may or may not be differentially regulated by activity. Future work will determine whether the mechanisms underlying activity-dependent specification of other neurotransmitters share common elements.
This integration of activity and genetic pathways may have important implications for development of the nervous system. The Ca activity-dependent switch from excitatory glutamatergic to inhibitory GABAergic phenotype through regulation of a binary selection gene could reverse the polarity of a developing neuronal circuit. Our findings raise the intriguing possibility that activity-dependent regulation occurs at other fundamental choice points during development. Activity-responsive genes may identify developmental switches where extrinsic factors impinge on intrinsic pathways.
Activity plays an important role in many aspects of neuronal development. The functions and mechanisms of synaptic activity have been extensively studied. For example, this form of activity regulates the number and strength of synapses formed 42,43 and cell survival 44. Less is known about the function of earlier forms of activity, although it has been identified throughout the nervous system in many organisms. Early activity regulates proliferation of neuronal precursors and neuronal migration as well as neuronal differentiation 2. Ca transients occur at many times and in many places during development, but the mechanisms by which their patterns regulate differentiation have not been fully elucidated. It is unclear how Ca transients generated at frequencies of 1–20 hr −1 initiate the signaling cascade regulating neurotransmitter phenotype, although bursts of these Ca transients stimulate generation of cAMP transients 45. Coding of transcription by the frequency of Ca transients has been reported in T lymphocytes 46 and in basophilic leukaemia cells 47. Stimulation of CaM kinase II, NFAT and OCT/OAP transcription factor activity depends on relatively high frequencies of Ca transients 46–48. In contrast, activation of the NF-kB transcription factor by Ca oscillations occurs at low frequencies 46 that are within the range of those generated by developing Xenopus spinal neurons. It will be interesting to assess the role of cJun and CRE in other organisms and to identify signaling components linking Ca spikes to changes in cJun activity. N-terminal Jun kinases regulate cJun activity via phosphorylation of S63/S73 and T91/T93 (39), making them attractive candidates.
We thank Sera Chung, Gabriela Monsalve and Armando de la Torre for technical assistance. We thank Darwin Berg, Anirvan Ghosh and Joseph Gleeson for critical comments on the manuscript. This work was supported by a Damon Runyon Cancer Research Foundation Postdoctoral Fellowship to K.W.M., NIH Predoctoral Fellowship T32 AG00216 to L.M.K. and by NIH grant MH 074702 to N.C.S.
Note: Supplementary information is available on the Nature Neuroscience website.
Competing interests statement The authors declare that they have no competing financial interests.