Neurotransmitter Switching? No Surprise

Abstract

Among the many forms of brain plasticity, changes in synaptic strength and changes in synapse number are particularly prominent. However, evidence for neurotransmitter respecification or switching has been accumulating steadily, both in the developing nervous system and in the adult brain, with observations of transmitter addition, loss, or replacement of one transmitter with another. Natural stimuli can drive these changes in transmitter identity, with matching changes in postsynaptic transmitter receptors. Strikingly, they often convert the synapse from excitatory to inhibitory or vice versa, providing a basis for changes in behavior in those cases in which it has been examined. Progress has been made in identifying the factors that induce transmitter switching and in understanding the molecular mechanisms by which it is achieved. There are many intriguing questions to be addressed.

The identity of the neurotransmitters synthesized and released by neurons has long been considered a core feature of neuronal phenotype and a critical aspect of a neuron’s stable differentiated fate. Genetic programs specify the initial expression of transmitters (Thor and Thomas, 1997; Tanabe et al., 1998; Pierani et al., 2001; Mo et al., 2004; Mizuguchi et al., 2006; Pillai et al., 2007), but evidence for subsequent transmitter switching has an extensive history. The understanding of transmitter respecification begins with consideration of the influential studies of the differentiation of neurons derived from the neural crest. I then review studies of transmitter reassignment during the development of the CNS, extending the early findings in the neural crest. Finally, the process of transmitter switching in the mature nervous system is examined and perspectives for future investigation are developed.

Transmitter Switching during Development: Neural Crest-Derived Neurons

Studies of Switching in Culture

Early work by Patterson and colleagues, based on studies by Furshpan and Potter, aimed at identifying the nutritional requirements and role of cell interactions in neuronal development (Mains and Patterson 1973a, 1973b, 1973c; Patterson and Chun, 1974). These studies demonstrated that culture conditions regulate the biosynthesis of acetylcholine versus norepinephrine in neonatal rat superior cervical ganglion neurons. The presence of non-neuronal cells or medium conditioned by these cells favored cholinergic differentiation, while their relative absence led to noradrenergic development. Electrophysiological recordings by Furshpan, Potter, and colleagues showed that neurons cultured under these different conditions can make excitatory noradrenergic, inhibitory cholinergic, or dual-function synapses (O’Lague et al., 1974; Furshpan et al., 1976). Acetylcholine or noradrenalin synthesis from ³H-choline or ³H-tyrosine by single neurons in microwells demonstrated that cholinergic or adrenergic differentiation in single cells depended on culture conditions (Reichardt and Patterson, 1977). Culturing single neurons on micro-islands of small numbers of cardiac myocytes (Potter et al., 1980) established dense innervation facilitating identification of the pharmacology of synaptic potentials. Repeated recordings from the same pair of single neurons and innervated cardiac myocytes revealed neuronal transitions from noradrenergic to cholinergic status (Figure 1) (Furshpan et al., 1976; Potter et al., 1986).

Figure 1. Microcultures Enabled Serial Assays of Single Neurons during Transition from Adrenergic to Cholinergic Status.

Top: a solitary neuron from the superior cervical ganglion of a newborn rat embryo; 19 days in vitro. The arrow at H points to a cluster of cardiac myocytes. Inset shows an impulse in this neuron; scales are 20 ms and 50 mV. After Furshpan et al. (1976). Bottom: assay of a solitary neonate-derived rat sympathetic ganglion neuron that underwent a transition from adrenergic to cholinergic phenotype.

(A–F) At 17 days in vitro, intracellular recording revealed no autaptic effect of a single action potential (A), and 2 s of 20 Hz stimulation exerted an excitatory effect on cardiac myocytes (B) that was blocked by 1 μM propranolol (C). At 62 days in vitro, a single action potential in the same neuron generated a pronounced autaptic effect (D), and the effect of the same stimulus train on cardiac myocytes was inhibitory (E) and blocked by 0.2 μM atropine (F). Vertical scale: 80 mV for (A), (B), and (F); 40 mV for other traces. Horizontal scale: 40 ms for (A) and (D); 20 s for other traces. After Potter et al. (1986).

In associated studies, Landis found that the ultrastructure of synaptic vesicle populations matched these biochemical and physiological findings: more dense core vesicles were observed in presynaptic terminals of neurons in the noradrenergic condition and more clear core vesicles were observed in the cholinergic condition (Figure 2); both populations were observed in terminals of dual-function neurons (Landis, 1976; Johnson et al., 1976). Remarkably, adult superior cervical ganglion neurons appeared to retain some neurotransmitter plasticity when grown under appropriate culture conditions (Wakshull et al., 1979), which seemed to be lost in neurons from aged rats (Adler and Black, 1984).

Figure 2.

Ultrastructural evidence for transmitter switching provided by electron micrographs of axonal synapses (autapses, arrows; top two panels) and varicosities (bottom two panels) of neonatal rat superior cervical ganglion neurons grown in microcultures after transmitter properties had been electrophysiologically identified. Synaptic vesicles are larger and rounded in cholinergic neurons (two left panels) and smaller and more pleomorphic with dense cores in adrenergic neurons (two right panels). 14, 10, 19, and 21 days in vitro. After Landis (1976).

These studies revealed the release of multiple transmitters by single neurons, consistent with investigations revealing anatomical colocalization (Hökfelt et al., 1977, 1983; Fried et al., 1986). In addition to noradrenalin and acetylcholine, superior cervical ganglion neurons were also capable of releasing purines that hyperpolarized cardiac myocytes in culture (Wolinsky and Patterson, 1985; Furshpan et al., 1986b) and serotonin and an un-identified substance or substances (X) that depolarized these target cells (Matsumoto et al., 1987; Sah and Matsumoto, 1987). Single neonate-derived neurons released as many as four of the five transmitters expressed in this population of neurons (norepinephrine, acetylcholine, purines, serotonin, and X), and single adult-derived superior cervical ganglion neurons were observed to release three. It was not clear that there were any real differences in the transmitter repertoire between neonatally- and adult-derived neurons. These findings raised the question of the relationship of multiple transmitters to transmitter switching.

Changes in Levels of Transmitters

These investigations alsoraised the question of the extent to which the switch between cholinergic and adrenergic phenotype is complete. Analysis of the specific activity and immuncytochemical staining of tyrosine hydroxylase (TH), the rate-limiting enzyme for catecholamine synthesis, as well as other markers and measurements of transmitter synthesis, showed substantial differences between cholinergic and adrenergic conditions for cultured rat superior cervical ganglion neurons. However, predominantly cholinergic neurons still displayed low levels of TH (Higgins et al., 1981; Iacovitti et al., 1981). The extent of reduction of the adrenergic phenotype depended on the strength of cholinergic induction (Wolinsky and Patterson, 1983). Neurons that had become cholinergic retained the ability to take up catecholamines in vitro (Landis, 1976; Reichardt and Patterson, 1977) and in vivo (Landis and Keefe, 1983). Importantly, transmitter function depended on a high level of expression of the principal transmitter. The possibility that the low levels of secondary, tertiary, and additional transmitters influence function in a subtle manner was difficult to address.

Black and colleagues discovered examples of transmitter gain and loss in other neural crest derivatives. Adrenomedullary chromaffin cells convert from the noradrenergic to strictly adrenergic phenotype with the appearance of the transcripts and protein of the enzyme that methylates norepinephrine (phenylethanolamine N-methyltransferase; Bohn et al., 1981; Sabban et al., 1982). Transient noradrenergic differentiation of neurons in the developing gut where enteric ganglia are forming involves coordinate expression of the noradrenalin synthetic enzymes and uptake machinery that disappears by 2 weeks of age (Cochard et al., 1979; Jonakait et al., 1979; Teitelman et al., 1979). The persistence of transmitter uptake enabled identification of the same neurons following disappearance of synthetic enzymes and catecholamine. Cranial nerve and dorsal root sensory ganglia also transiently express a noradrenergic phenotype (Jonakait et al., 1984). These neurons presumably express another transmitter or transmitters to enable their network function after the disappearance of noradrenalin, but this was not investigated.

Mechanisms of Transmitter Induction

Efforts to identify soluble factors in conditioned medium that induced the cholinergic phenotype in rat superior cervical ganglion neurons led to purification of a protein (Fukada, 1985). Cloning of the gene revealed that it is leukemia inhibitory factor (LIF; Yamamori et al., 1989), previously identified as an inhibitor of hematopoietic differentiation. Recombinant LIF induced the cholinergic phenotype and suppressed the noradrenergic phenotype of cultured sympathetic superior cervical ganglion neurons. LIF also altered neuropeptide expression, increasing VIP, substance P, and somatostatin levels while decreasing the levels of neuropeptide Y (NPY) and TH. LIF induced ChAT activity and VIP but decreased the expression of substance P in dorsal root ganglion neurons (Nawa and Patterson, 1990; Freidin and Kessler, 1991; Nawa et al., 1991; Mulderry, 1994). Thus, LIF by itself is not a specific differentiation factor for the cholinergic phenotype. Further, it is not unique. Ciliary neurotrophic factor (CNTF), originally identified as a survival factor for ciliary ganglion neurons, also induced the cholinergic phenotype, suppressed the noradrenergic phenotype of sympathetic neurons (Saadat et al., 1989; Rao et al., 1990), and induced the same profile of altered neuropeptide expression (Ernsberger et al., 1989; Rao et al., 1992). Substantial progress has been made in determining the genes that regulate the noradrenergic and cholinergic phenotypes (Ernsberger et al., 1995; Lo et al., 1999; Stanke et al., 1999; Brosenitsch and Katz, 2002; Hendershot et al., 2007). More rapid conversion of sympathetic ganglion neurons from adrenergic to cholinergic status was achieved by BDNF, which appeared to stimulate preferential presynaptic release of acetylcholine over norepinephrine within 15 min of application (Yang et al., 2002).

A further aspect of regulation of transmitter phenotype was revealed when Patterson and colleagues demonstrated that depolarization with high potassium, veratridine, or electrical current reduced acetylcholine synthesis in response to exposure to conditioned medium and left neonatal superior cervical ganglion neurons primarily noradrenergic (Walicke et al., 1977; Walicke and Patterson, 1981). Blockade of calcium entry or buffering of intracellular calcium prevented this effect, arguing for a role of intracellular calcium in noradrenergic stabilization (Table 1). It would have been interesting to determine whether these cultured neurons exhibit greater spontaneous activity in the absence of nonneuronal cells or conditioned medium that favor noradrenergic differentiation, but the tools for calcium imaging were rudimentary at that time. These authors presciently concluded “calcium appears to play a major role in the developmental effects of depolarization and in the intracellular events governing transmitter choice.” Depolarization elicited a different spectrum of changes in neuropeptides, increasing VIP and NPY, but not affecting levels of somatostatin or substance P (Rao et al., 1992; see also Kessler et al., 1981). However, depolarization did not suppress induction of the cholinergic phenotype or changes in neuropeptides by CNTF, in contrast to its antagonism of the effects of LIF, indicating that the two differentiation factors engage different mechanisms.

Table 1.

Calcium Regulates the Developmental Effects of K⁺ Depolarization on Transmitter Switching

Additives	Neuronal Number	ACh/CA	ACh (fmole/cell)	CA (fmole/cell)
None	3345 ± 375	0.04 ± 0.02a	0.1 ± 0.0a	2.6 ± 0.1
CM	3820 ± 240	5.68 ± 0.13	8.9 ± 0.8	1.6 ± 0.1
K+ + CM	4220 ± 960	0.20 ± 0.06a	0.5 ± 0.2a	2.4 ± 0.4
K+ + CM + Ca2+	1025 ± 106a	0.07 ± 0.07a	0.1 ± 0.1a	1.8 ± 0.7
K+ + CM + Mg2+	1840 ± 520a	3.97 ± 0.64a	9.5 ± 0.4b	2.5 ± 0.3
K+ + CM + D600	5170 ± 594	2.54 ± 0.42a,b	4.6 ± 0.9a,b	1.8 ± 0.2

Cultures of neonatal rat superior cervical ganglion neurons were grown for 2 days. Conditioned medium (CM) was added where indicated on day 9; other additions were made on day 2. Neurons were counted on day 16, and assays were run on day 17. [CaCl₂] was 5 mM, [MgCl₂] was 20 mM, and [D600] was 1 μM. After Walicke and Patterson (1981).

^ap < 0.05 compared to CM (line 2)

^bp < 0.05 compared to K⁺ + CM (line 3)

The olfactory epithelium is derived in part from the neural crest (Katoh et al., 2011), and Baker found that unilateral, olfactory deprivation in neonatal rats reduces the number of neurons expressing TH, but not aromatic amino acid decarboxylase or GABA immunoreactivity, demonstrating selective experience-dependent regulation of transmitter expression (Baker, 1990). The initial expression of TH in the rat main olfactory bulb was closely associated with afferent innervation, leading to the suggestion that it may be required (Baker and Farbman, 1993). Application of GABA, which induces proliferation and migration of olfactory bulb progenitors, induces expression of TH in additional periglomerular neurons in neonatal mouse brain slices via activation of GABA receptors and voltage-gated calcium channels (Akiba et al., 2009).

In addition to the large and apparently qualitative changes in transmitter expression, smaller activity-dependent quantitative changes in levels of transmitter expression have been described in the adrenergic system (Mueller et al., 1969; Chalazonitis and Zigmond, 1980; Dreyfus et al., 1986). They were associated with the appearance of TH transcripts and blocked by inhibitors of mRNA or protein synthesis (Black et al., 1987) and are likely to be reduced versions of those that become functionally qualitative.

Developmental Transmitter Switching In Vivo

In work that was a major milestone, moving out of culture and into the animal, Landis and colleagues investigated the switching of transmitter phenotype in vivo by following the development of cholinergic sympathetic innervation of sweat glands in the rat footpad. Catecholamine fluorescence and small dense core vesicles were observed during the innervation of the glands, both of which decreased with age and were no longer detected at 3 weeks (Figure 3), although uptake of exogenous catecholamines persisted (Landis and Keefe, 1983). Significantly, these results established transmitter switching as a normal aspect of development and not solely a feature of cell culture. Transplantation of sweat glands enabled sympathetic neurons normally innervating noradrenergic target organs to innervate these cholinergic target organs. Under these conditions, the neurons developed a cholinergic phenotype, demonstrating that target organs can induce the proper neurotransmitter trait in innervating neurons (Schotzinger and Landis, 1988). Sympathetic sweat gland axons that innervate noradrenergic targets fail to become cholinergic and retain noradrenergic markers (Schotzinger and Landis, 1990). Interestingly, adrenergic innervation of sweat glands induces their secretion of a sweat gland factor that drives cholinergic conversion (Habecker and Landis, 1994; Habecker et al., 1995). Although transgenic expression of LIF can induce transmitter switching of sympathetic neurons in vivo (Bamber et al., 1994), neither LIF nor CNTF mediate the normal transmitter switch in footpad sweat gland innervation (Rao et al., 1993; Francis et al., 1997).

Figure 3. Demonstrating the Adrenergic to Cholinergic Switch In Vivo, the Number of Small Granular Vesicles in Axons of Sympathetic Neurons Associated with Developing Sweat Glands Decreased with Rat Postnatal Age.

(A–E) The percentage declined sharply from 52% at 7 days to 0% at 21 days (A). Small granular vesicles (arrows) are evident at 7 days (B) and 10 days (C), rare at 14 days (D), and absent at 21 days (E). After Landis and Keefe (1983).

These results were consistent with elegant studies of Le Douarin investigating differentiation of avian neurons from neural crest cells after their migration to different locations, using heterotopic transplantation from quail embryo to chick embryo to identify the cells of origin by the morphology of interphase nuclei. Quail neural tube and neural crest were transplanted from the adrenomedullary region of the thoracic neural crest, normally giving rise to sympathetic noradrenergic neurons and adrenal medulla, to the cranial region of the crest from which parasympathetic cholinergic neurons arise. Following crest migration, subsequent innervation of the duodenum lacked catecholamine fluorescence and stimulation elicited atropine-sensitive mechanical responses that were not distinguishable from the responses to vagal stimulation of unoperated controls or isotopic transplants from the cranial region. Similarly, quail transplants from the cranial region to the adrenomedullary region yielded adrenal glands populated with quail cells with catecholamine fluorescence and dense core secretory granules (Le Douarin and Teillet, 1974; Le Douarin et al., 1975. Transmitter phenotype depended on the environment in which the neurons differentiated and not on their site of origin. The fate of differentiated quail cholinergic ganglion cells could also be reversed by grafting them to the thoracic region of the neural axis of a younger chick embryo; quail cells that had already expressed choline acetyltransferase now demonstrated catecholamine fluorescence (Le Douarin et al., 1978).

Conclusions from Studies of Developing Neural Crest

The studies of transmitter switching in neural crest derivatives established many of its core features. They demonstrated a switch between excitatory and inhibitory transmitters in single neurons, that transmitter identity can change in both developing and adult neurons in culture, and that protein factors and electrical activity can regulate this process. They revealed that neurons can make and release up to four transmitters at one time. They also showed that developmental changes in transmitter identity occur naturally in vivo and that this too depends on protein factors and enables mature function in the form of autonomic behavior. Changes in transmitter identity were found to involve changes in the proportion of different transmitters rather than complete switches.

The discovery of transmitter switching in neural crest derivatives might have been expected, given the plasticity of neural crest and the dependence of differentiated states on the environment of the migratory path that different cells traverse. However, these findings raised a number of questions. Does transmitter switching occur in neurons in the CNS during development? Are transmitter switches associated with corresponding switches in postsynaptic transmitter receptors in vivo? Can environmental stimuli, voluntary activity, and experience drive changes in transmitter identity? Do these forms of transmitter switching lead to changes in complex behaviors other than functions of the autonomic nervous system? What are the mechanisms of changes in transmitter identity in these neurons?

Transmitter Switching during Development: CNS Neurons

Studies in Culture

An early clue to transmitter switching in the CNS came from studies of neurons from the embryonic Xenopus spinal cord dissociated and grown in culture. These cells generated spontaneous elevations of intracellular calcium that were suppressed by blockers of voltage-gated calcium channels, but the function of these calcium transients was unknown. However, the number of GABAergic neurons was 3-fold higher when these cells were cultured in the presence of extracellular calcium compared to in its absence (Figure 4) (Spitzer et al., 1993; Gu and Spitzer, 1995). It seemed likely that the process involved an activity-dependent change in phenotype since no neuronal birth or death was detected. A requirement for RNA synthesis was consistent with the possibility that glutamic acid decarboxylase expression was regulated by activity. xGAD67 transcripts were subsequently found to be expressed in the embryonic spinal cord during the period in which calcium spikes are generated, and transcript levels were 3-fold higher when cells were cultured in the presence of calcium compared to in its absence (Watt et al., 2000). The lower expression of GABA and xGAD67 transcripts was rescued in the absence of extracellular calcium by reimposition of intracellular calcium transients at the frequency normally generated in cultured neurons (Figure 5) (Gu and Spitzer, 1995; Watt et al., 2000). In more recent work, blockade of calcium spikes increased the number of neurons expressing xVGluT1 and decreased the number of neurons expressing xGAD67 transcripts; enhancing calcium spiking reduced the number of neurons expressing xVGluT1 without detectable effect on the number of neurons expressing xGAD67 (Lewis et al., 2014).

Figure 4. Spontaneous Calcium Transients Are Both Necessary and Sufficient to Determine Normal Differentiation of the GABAergic Phenotype.

Top: development of GABA immunoreactivity of cultured Xenopus spinal neurons in the presence and in the absence of extracellular calcium. 40% of neurons in the cultures develop GABA immunoreactivity during the first day in culture. Acquisition of immunoreactivity is suppressed by growth in calcium-free medium in comparison with controls. Bottom: tuning curve reveals stimulation of GABA expression by imposition of natural frequencies of calcium transients during growth in calcium-free medium. Normal maturation of GABA expression is stimulated by calcium transients at a frequency of 3/hr. After Spitzer et al. (1993) and Gu and Spitzer (1995).

Figure 5. Suppression or Enhancement of Spike Activity In Vivo Causes Homeostatic Superposition or Replacement of One Transmitter with Another in the Xenopus Larval Neural Tube.

Left: controls doubly stained for VGluT (red, excitatory) plus glycine or GABA (purple, inhibitory) illustrate the normal distribution of immunoreactivity. Center: embryos in which spike activity was bilaterally suppressed by expression of hKir2.1, stained as at left. Pink indicates coexpression of excitatory and inhibitory transmitters. Right: embryos in which spike activity was bilaterally enhanced by expression of rNav2aαβ, stained for glutamate and a marker of sensory Rohon-Beard cell identity (HNK-1), plus glycine or GABA. Light blue denotes coexpression of marker and inhibitory transmitter; white shows coexpression of marker and both excitatory and inhibitory transmitters. After Borodinsky et al. (2004).

Developmental Transmitter Switching In Vivo

Do patterns of calcium transients regulate transmitter identity in vivo? Misexpression of voltage-gated sodium channels or inward rectifier potassium channels in Xenopus embryos enhanced or suppressed this activity and changed the transmitter that neurons expressed without affecting the expression of markers of cell identity (Borodinsky et al., 2004). Increasing or decreasing the incidence of calcium transients by increasing the dorsoventral gradients of sonic hedgehog (Shh) or bone morphogenetic protein (BMP) morphogens identified endogenous mechanisms of phenotypic regulation (Belgacem and Borodinsky, 2011; Swapna and Borodinsky, 2012). Transmitter switching appeared to be homeostatic since enhancing activity led to an increased number of neurons expressing inhibitory transmitters, GABA and glycine; decreasing activity produced an increased number of neurons expressing excitatory transmitters, acetylcholine and glutamate (Figure 5). These results suggested that transmitter plasticity provides feedback that maintains electrical activity in an operating range. Similarly, altering spike frequencies in vitro did not affect labels of cell identity but specified expression of transmitters that were inappropriate for the markers the neurons expressed. Spontaneous miniature excitatory post-synaptic currents were recorded from muscle cells innervated by cultured neurons expressing acetylcholine or glutamate, engaging endogenous ACh receptors or misexpressed glutamate receptors and indicating that the transmitters are functionally released. Transmitter switches were reversible during a brief early period in vitro (Borodinsky et al., 2004).

Changes in Postsynaptic Receptors

What happens to postsynaptic transmitter receptors when the identity of the presynaptic transmitter changes? If transmitter switching is a physiologically regulated process, it seemed probable that there would be corresponding changes in postsynaptic receptor populations in order to ensure continuity of synaptic function. Examination of embryonic Xenopus striated muscle cells in vivo revealed that they normally express transcripts and protein for glutamate, GABA, and glycine receptors in addition to those for acetylcholine (Figure 6). Receptor expression was altered following ion channel misexpression to drive changes in transmitter identity in motor neurons. The appearance of glutamate in the neurons was accompanied by the expression of more AMPA and NMDA receptor subunit transcripts and protein in the muscle; the appearance of GABA or glycine was accompanied by the expression of mRNA and protein for subunits of the corresponding receptors (Borodinsky and Spitzer, 2007). Glutamatergic, GABAergic, and glycinergic miniature synaptic currents were recorded from muscle cells under these conditions, leading to the conclusion that transmitter-receptor matching can be regulated by activity in the assembly of functional synapses. These changes were also observed in uninnervated muscles in neuron-muscle cocultures, implying that neurons signal to muscle cells by diffusible factors, potentially including the transmitters themselves, which specify expression of appropriate receptors.

Figure 6. Activity-Dependent Transmitter Specification Drives Selection of Receptors at the Embryonic Xenopus Neuromuscular Junction.

Top: expression of nAChR, NMDAR, AMPAR, GABA_AR, and GlyR transcripts in skeletal muscle during normal development. RT-PCR was used for detection of subunit transcripts of five neurotransmitter receptors in muscle, notochord, and neural tube at three stages of development. Tissue-specific RNA was analyzed from embryos at 1 day and at 1.3 days and from larvae at 3 days. Bottom: whole-cell recordings from a larval Xenopus muscle cell of the axial musculature of 3-day spike-suppressed larva in which motor neurons express glutamate. Example of pharmacological identification of glutamatergic mpscs. A single mpsc is shown on an expanded time base to illustrate its kinetics. Performed in 2 mM Ca²⁺, Mg²⁺-free saline and 3 μM TTX; V_h = −80 mV. After Borodinsky and Spitzer (2007).

Physiological Transmitter Switching Regulates Behavior

Can natural environmental stimuli respecify transmitter expression in a neuronal population? The circuit regulating adaptation of skin pigmentation to background and the camouflage behavior of adult Xenopus lead from the retina to the ventral suprachiasmatic nucleus (VSC) to melanotrope cells that release melanocyte-stimulating hormone. In Xenopus larvae, briefly altering light exposure and changing the sensory input to the circuit controlling adaptation of skin pigmentation to background changed the number of VSC neurons expressing dopamine in a circuit-specific and activity-dependent manner. Physiologically stimulated calcium spike activity increased the numbers of behaviorally relevant dopaminergic neurons during the course of several hours at an early stage of development. Tyrosine hydroxylase transcripts, dopamine transporters, and dopamine appeared in neurons that already expressed NPY (Figure 7) (Dulcis and Spitzer, 2008). No changes in axon terminal branching were observed, but D2 dopamine receptors appeared on the postsynaptic melanotrope target cells. The newly dopaminergic neurons, identified by the absence of Pax6 and the expression of NPY and LIM1/2, enhanced the speed of camouflage behavior. Neurons in different dopaminergic nuclei of Xenopus larvae expressing different combinations of transcription factors responded differently to manipulation of early electrical activity by ion channel misexpression. Specific nuclei displayed different increases and decreases in numbers of dopaminergic neurons, consistent with the dependence of activity-dependent differentiation on molecular context (Velázquez-Ulloa et al., 2011). The discovery that a natural stimulus can change the transmitter and receptors that are functionally expressed in a circuit, leading to altered behavior, enhanced the relevance of transmitter switching and raised the question of the basis by which it is achieved.

Figure 7. Illumination Changes the Number of Dopaminergic Neurons in the Ventral Suprachiasmatic Nucleus of the Larval Xenopus Hypothalamus.

Top: transverse sections show the core (dashed inner circle) and annular neurons (between dashed circles) in 2-hr black- and white-adapted larvae triply labeled with anti-sense to tyrosine hydroxylase (TH) transcripts and antibodies to TH and NPY. Bottom: quantification of changes in TH immunoreactivity in response to either background illumination or incident light. Scale bar, 60 μm. **p < 0.001. After Dulcis and Spitzer (2008).

Mechanisms of Transmitter Switching: Respecification

Transmitter switching involves respecification after the appearance of spontaneous activity in Xenopus embryos. Examination of transmitter expression at neural plate and early neural tube stages showed that neurons initially exhibited glutamate and GABA promiscuously and often coexpressed the two; acetylcholine and glycine were expressed later in the appropriate neurons as glutamate and GABA became restricted to the neurons that classically express them (Root et al., 2008). The initial expression of glutamate and GABA was necessary to drive expression of acetylcholine and glycine, reminiscent of the action of serotonin driving the differentiation of glutamatergic neurons in organotypic slice cultures of embryonic rat cerebral cortex (Lavdas et al., 1997). A similar transmitter switch, from GABA to glycine, occurs during postnatal development of inhibitory inputs from the medial nucleus of the trapezoid body (MNTB) to the rat lateral superior olive (LSO), potentially providing longer duration depolarization at early stages to enable calcium influx that drives differentiation (Kotak et al., 1998; Nabekura et al., 2004). Glutamate co-release at these GABA/glycinergic synapses is important for the topographic specification of the inhibitory circuit (Noh et al., 2010).

The numbers of neurons expressing different transmitters in the Xenopus neural tube changed when GABA or glutamate signaling was blocked with morpholinos to knock down transmitter-synthetic enzymes or application of pharmacological receptor antagonists during a sensitive period of development somewhat longer than the critical period observed in vitro (Root et al., 2008). Suppression of metabotropic GABA and glutamate receptors reduced the incidence of calcium-dependent embryonic electrical activity and upregulated the number of neurons expressing ACh and glutamate while downregulating the number of neurons expressing GABA and glycine. These results revealed a novel role for embryonically expressed neurotransmitters, driving electrical activity that respecifies neuronal transmitters, consistent with studies demonstrating that initial transmitter expression is driven by expression of genetic programs (Thor and Thomas, 1997; Tanabe et al., 1998; Pierani et al., 2001; Mo et al., 2004; Mizuguchi et al., 2006; Pillai et al., 2007). These findings suggested that transmitter respecification should be considered a form of plasticity rather than a change in neuronal fate.

Mechanisms of Transmitter Switching: Activity-Dependent Transcription Factors

Regulation of transcription factors is a common basis of neuronal differentiation, and activity-dependent phosphorylation of the c-Jun transcription factor controls the switch between glutamate and GABA in developing Xenopus spinal neurons. Silencing activity in single neurons in vivo failed to change the identity of their transmitter, indicating that it is regulated non-cell autonomously by the activity of neurons that surround it. BDNF was a likely candidate to regulate the process and is present in the CNS and released by calcium spikes. Activation of the BDNF receptor, trkB, engaged a MAP kinase cascade that culminated in JNK-dependent activation of cJun (Güemez-Gamboa et al., 2014). Knocking down or overexpressing tlx3, a homeobox glutamate-GABA selector gene (Cheng et al., 2004, 2005), decreased or increased the number of glutamatergic neurons, with the opposite effect on the number of GABAergic neurons, and occluded the effects of altering activity (Marek et al., 2010). Phospho-cJun bound to a non-canonical cAMP response element (CRE) in the promoter of tlx3 repressed transcription and enabled expression of glutamic acid decarboxylase, promoting appearance of GABA. Decreased calcium spike activity reduced the level of P-cJun and enabled tlx3 expression that led to the appearance of glutamate. Regulation of transcription factor phosphorylation provides a way for early activity to regulate gene expression and change the phenotype of developing neurons.

The molecular mechanisms underlying transmitter specification in vivo in Xenopus embryos also entail activity-dependent appearance of transcription factors. The cascade leading to specification of the serotonergic phenotype includes Nkx2.2, Lmx1b, and Pet1, which are necessary for expression of tryptophan hydroxylase that catalyzes serotonin synthesis. Spontaneous calcium spike activity in the raphe acted downstream of Nkx2.2 and modulated the specification of serotonergic neurons by regulating expression of Lmx1b without affecting cell proliferation or patterning (Demarque and Spitzer, 2010). Decreased activity increased the number of neurons expressing Lmx1b, while increased activity decreased it, and overexpression or knockdown of Lmx1b occluded the effects of altering activity. Application of serotonin shortens swim episodes in Xenopus larvae. Reduction in the duration of swimming behavior by serotonin was phenocopied by suppressing activity, and overexpression or knockdown of xLmx1b reduced or increased swim durations. These results connect activity-dependent regulation of transcription factor expression to transmitter respecification and an affected behavior.

Conclusions from Studies of Developing CNS Neurons

These investigations extended observations of switching from excitatory to inhibitory transmitter or vice versa. They demonstrated that transmitter switching during development in vivo is associated with corresponding changes in postsynaptic receptors, providing recognition of the biological significance of the transmitter switch by the changes in receptor expression. They showed that natural environmental stimuli can drive transmitter switching that leads to behavioral change. The mechanism of induction of transmitter switching involves early transmitter release that regulates electrical activity. Expression of transmitter switching entails regulation of transcription factors by a signal conduction cascade triggered by this activity. Is any of this plasticity involved in the operation of the adult CNS?

Transmitter Switching in the Mature Nervous System

Evidence for transmitter switching in the mature brain has been available for almost three decades, although its consequences for behavior remained elusive. Early experiments demonstrated loss of function rather than gain of function or exchange of function. Adult naris closure (Kosaka et al., 1987; Baker et al., 1993) or deafferentation of the olfactory bulb (Baker et al., 1988; Stone et al., 1991) reduced expression of TH without effect on expression of GAD and GABA. These findings showed that activity-dependent regulation of transmitter specification can be specific to transmitter phenotype. Intraocular injection of tetrodotoxin or eyelid suture in young adult monkeys reduced expression of GABA and GAD in the visual cortex (Hendry and Jones, 1986) that was restored when activity was reintroduced (Hendry and Jones, 1988). A decreased number of neurons expressing GAD transcripts and protein without loss of neurons in the postmortem prefrontal cortex of patients with schizophrenia suggested that transmitter loss is related to the neurological disorder (Akbarian et al., 1995). Blockade of cholinergic input to the mature frog optic tectum decreased the number of neurons expressing Substance P (Tu et al., 2000).

Transmitter gain of function has been demonstrated in more recent studies. Aged macaque monkeys injected intramuscularly with the neurotoxin MPTP and intraperitoneally with BrdU demonstrated a greater than 2-fold increase in the number of striatal interneurons expressing the dopamine transporter (DAT) and GAD, but not BrdU. The results suggested that newly dopaminergic neurons are derived from preexisting GABAergic neurons in Parkinsonian monkeys (Tandé et al., 2006).

Mice recovering from 6-OHDA injection into the substantia nigra pars compacta (SNc) demonstrated changes in the number of TH-IR neurons that depended on inhibition or facilitation of calcium-dependent potassium channels as well as on other manipulations of activity (Aumann et al., 2008; 2011). Mice allowed to mate for 1 week showed changes in the numbers of TH-IR neurons in the SNc. Male mice exposed to environmental enrichment for 2 weeks exhibited an increase in TH-IR neurons in the SNc that was abolished by concurrent local infusion of GABA_A receptor antagonists (Aumann et al., 2013).

Intriguingly, deep brain stimulation (DBS) of the rat anterior thalamic nucleus induced an increase in the number of TH-IR neurons in the ventral tegmental area (VTA); there was no change in the number of TH-IR neurons in the SNc or in response to sham stimulation. Stimulation of the mammillothalamic tract or entorhinal cortex was not effective in increasing the number of TH-IR neurons in the VTA (Figure 8) (Dela Cruz et al., 2014). These studies indicate that activity regulates the number of dopaminergic neurons in the adult brain and that DBS can cause a phenotypic switch that may provide a basis for treatment of multiple disorders.

Figure 8. Deep Brain Stimulation of Rat Anterior Thalamus Leads to Increased Numbers of Tyrosine Hydroxylase-Stained Neurons in the Ventral Tegmental Area.

(A–D) Top: there is no change following stimulation in the mammillothalamic tract (MTT) (A and B) in contrast to the increase after stimulation in the ATT (C and D). Bottom: stereological quantification of the numbers of TH-stained neurons in the VTA and substantia nigra pars compacta (SNc) following stimulation in different brain regions. EC, entorhinal cortex. Scale bar: (A) and (C), 500 μm; (B) and (D), 50 μm. *p < 0.05. After Dela Cruz et al. (2014).

Transmitter gain of function is not restricted to the dopaminergic phenotype. Induction of seizures or kindling in the dentate gyrus in adult rat hippocampal slices led to the appearance of GABAergic along with glutamatergic transmission by mossy fiber innervation of hippocampal CA3 pyramidal neurons by normally glutamatergic granule cells (Gutiérrez 2000, 2002). Sustained synaptic or direct activation of glutamate receptors led to the appearance of immunohistochemically detectable GAD67 that was prevented by protein synthesis inhibitors.

Simultaneous Gain and Loss of Transmitter

Transmitter exchange also appears to take place in the adult brain. Rats exposed to long- or short-day photoperiods showed transmitter switching in hypothalamic nuclei (Dulcis et al., 2013). Neurons in the paraventricular and periventricular nuclei (PaVN and PeVN) had a greater number of dopaminergic neurons following exposure to the short-day photoperiod and fewer dopaminergic neurons after exposure to the long-day photoperiod. A reciprocal relation was observed for neurons expressing somatostatin (SST), with a greater number of neurons expressing SST following exposure to long-day photoperiods and fewer neurons expressing SST following short-day exposure. No neurogenesis or apoptosis were detected, and the loss of one phenotype was matched quantitatively by the gain of the other. Neurons expressing both transmitters were prominent following exposure to a balanced day-night photo-period, as expected for transitional expression. These changes in dopamine and SST expression were reversible when rats were subsequently re-exposed to a control photoperiod. The change in transmitters was matched by changes in the level of expression of D2 dopamine receptors, while the level of expression of SST receptors was unchanged (Figure 9). The matching of dopamine receptor levels to the number of dopaminergic neurons was consistent with earlier studies showing that glutamatergic reinnervation of adult rat internal obliquus abdominis muscle led to the appearance of glutamate receptors on the muscle, at the sites of the original endplates, restoring motor function (Brunelli et al., 2005; Francolini et al., 2009).

Figure 9. Changes in Postsynaptic Receptor Populations Are Associated with Changes in Presynaptic Transmitter Expression in the Adult Rat Brain.

Changing exposure from long-day (19 hr light, 5 hr dark, 19L:5D) to balanced-day (12L:12D) or short-day (5L:19D) photoperiods switches transmitter expression in the PaVN and PeVN from somatostatin (SST) to dopamine (DA). Top: colocalization of D2R and SST2/4R on corticotrophin releasing factor (CRF) neurons lining the walls of the third ventricle (dotted lines), following exposure of rats to each of the photoperiods. Bottom: the number of labeled CRF neurons was averaged for 10 30-μm sections of the rostral hypothalamus. Scale bar, 40 μm. **p < 0.01. After Dulcis et al. (2013).

The molecular changes were associated with changes in behavior. Rats demonstrated increased anxiety and depression following the stressful long-day photoperiod, assessed with the elevated plus maze and the forced swim test, which decreased following exposure to the short-day photoperiod. The induction of newly dopaminergic neurons by short-day exposure rescued the anatomical and behavioral consequences of 6-OHDA treatment and loss of dopaminergic neurons, demonstrating a causal link of transmitter identity to behavior (Dulcis et al., 2013). These results suggest that the transmitter-switching capability first identified in the neural crest extends to the adult mammalian brain.

Conclusions from Studies of Mature CNS Neurons

Reports of transmitter switching in the mature brain focus on a relatively little-studied form of brain plasticity. It is not clear yet whether there are significant differences in transmitter switching between the developing and the mature nervous system. In each case, natural stimuli can drive changes in transmitter identity with associated changes in postsynaptic receptors that validate biological significance and can lead to changes in behavior. The process appears to be directed toward circuit homeostasis, as excitatory stimuli promote the appearance of inhibition or reduction of excitation and inhibitory stimuli stimulate the appearance of excitation or reduction of inhibition. It will be of interest to learn whether the molecular mechanisms underlying transmitter switching in the adult nervous system are the same as those involved during development. It seems unnecessary to refer to this as a change in neuronal fate, since fate involves much more than transmitter identity. When we learn to speak a second language we don’t change our character or personality, much less the direction of our lives, but we are able to communicate more flexibly with a wider range of partners.

Perspectives

Work on transmitter switching prompts a number of questions. What is the basic phenomenon? Does it come in one form or three? Perhaps both loss of transmitter and gain of transmitter are simply half of transmitter replacement, for which the other half has yet to be identified. Since neurons express more than one transmitter, do they switch singly or perhaps in sets, and are there preferred switch partners? A related issue concerns the relationship of transmitter switching to transmitter coexpression and corelease that have now been widely observed (Gillespie et al., 2005; Hnasko and Edwards, 2012; Shabel et al., 2014; Nelson et al., 2014; Saunders et al., 2015). Perhaps coexpression arises when the molecular machinery of switching stalls in midstride or has been arrested. Alternatively, since much of the evidence supporting a complete switch is based on immunohistochemical detection of markers that has an inherent detection threshold, switching could also result from an increased bias to a particular transmitter. On this view switching may be an extreme case of corelease with a strong tilt toward one transmitter.

The interpretation of the connectome is more challenging when the sign of a synapse and the function of a circuit may be able to switch from excitatory to inhibitory or vice versa in an experience-dependent manner. What is the prevalence of transmitter switching? Changes in synaptic strength or synapse number are ubiquitous in the nervous system. Is the ability to switch transmitters restricted to a particular set of neurons—the reserve pool—or do all neurons have the ability to switch at least one transmitter in or out? Presumably only one neuron in any circuit changes its transmitter in response to circuit activity, because a larger number could cancel out the impact of the change. How this is coordinated remains to be determined. This form of plasticity may have a long phylogenetic history, as suggested by the observation that hypoxia acts through HIF1α to upregulate serotonin expression in the ASG sensory neurons in C. elegans that normally express glutamate (Pocock and Hobert, 2010; Serrano-Saiz et al., 2013).

Do neurotransmitter switches contribute to neurological disorders? Mutations in genes involved in the plasticity of synaptic strength are associated with several illnesses (Chao et al., 2007; Fromer et al., 2014; La Montanara et al., 2015). Similarly, sustained exposure to stimuli that drive transmitter switching may contribute to these conditions. Seasonal fluctuations in transmitter expression in the human brain have been known for decades (Karson et al., 1984; Brewerton et al., 1988; Eisenberg et al., 2010). The highly effective treatment of seasonal affective disorder (SAD) with phototherapy (Lam et al., 2006; Pail et al., 2011) suggests that the photoperiod-dependent transmitter switching and behavioral changes observed in adult rats may be relevant in this context. PET scanning dopamine receptor expression in patients both with SAD and in remission could be revealing.

Future Directions

The development of new tools will be important in advancing understanding of transmitter switching. Future work will make use of the technique of single-cell RNA sequencing (RNA-seq) (Tang et al., 2009; Lovatt et al., 2014) to analyze transmitter specification and respecification. It is difficult to apply multiple in situ probes to single neurons. Single-cell transcriptome analysis promises to yield a clearer answer to the question of the number of transmitters a neuron expresses, at the transcriptional level, and the changes in the level of the transcripts for each one. Other sensitive tools will be needed to evaluate expression and function of multiple proteins regulating transmitter specification.

Determination of the patterns of activity that are associated with transmitter switching will be facilitated by chronic multi-photon calcium imaging (Peters et al., 2014) and by reporters driven by activity-dependent promoters (Kawashima et al., 2014). The necessity and sufficiency of these patterns will be tested by optogenetic silencing and stimulation (Tye et al., 2013). These tools should lead to understanding of the molecular mechanisms by which cumulative activity is integrated to modulate gene expression.

Other intriguing questions await investigation. Do embryonic stem cell- and induced pluripotent stem cell-derived neurons exhibit transmitter switching? Pathways to specification of neurons expressing particular transmitters have been identified (Lee et al., 2000; Bibel et al., 2004), mirroring the initial process of transmitter specification in early development. Whether these transmitter phenotypes can be altered by activity or protein factors can now be tested (Stroh et al., 2011; Sun et al., 2013). Is transmitter switching regulated epigenetically? Environmental enrichment (Fischer et al., 2007; Sweatt, 2009) and activity (Tang and Goldman, 2006; Guo et al., 2011; Michod et al., 2012) control the expression of epigenetic marks, and activity regulates long-term synaptic function (Guan et al., 2002), brain plasticity, and behavior (Feng et al., 2007; Fagiolini et al., 2009) via changes in histone acetylation and DNA methylation. Inhibition of histone deacetylases (HDACs) de-represses the expression of the dopaminergic phenotype in murine olfactory bulb and rostral migratory stream (Akiba et al., 2010), suggesting that HDAC function influences this form of plasticity. Is this form of plasticity age dependent? Whether transmitter switching is involved in learning and memory is a fascinating issue for future study.