We are interested in how gene expression contributes to courtship song for two reasons. First, given the conservation of genetic mechanisms across species and the utility of Xenopus as an engine for gene discovery, this model system may allow us to uncover key elements in the regulation of hindbrain vocal patterning, in the ability of auditory neurons to decode songs and that permit forebrain neurons to match hearing to utterance. Second, each species of Xenopus (and all sub-soecies of X. laevis, sing a distinct song and this ability is heritable. Thus genes that participate in producing and decoding songs are strong candidates for the genetic sunstrates of evolutionary change.

To date we have identified five genes that participate in the control of vocal behaviors used in social communication. Three are receptors for hormones (the androgen, estrogen and gonadotropin receptors), one is a hyperpolarization-activated potassium channel (HCN) and the fifth is a laryngeal specific, myosin heavy chain gene (LM). We have cloned and sequenced the receptors and the LM gene; candidate HCN channels are present in the X. tropicalis genome sequence. To understand how these genes participate in the sexual differentiation of the vocal circuit, we have begun to map their expression onto the developing and adult brain and larynx. The LM gene is expressed specifically in a population of muscle stem cells in the larynx and in adult male muscle fibers. Adult muscle also expresses androgen and estrogen receptors. For the neural circuit that controls vocal behavior, androgen receptor is expressed in the hindbrain (vocal motor neurons and nucleus DTAM).Iin the midbrain, both estrogen receptor and androgen receptor are expressed in the laminar nucleus of the torus (LTOR), a major auditory region that contains cells responsive to specific temporal features of songs. In the forebrain, nucleus VST (auditory and pre-motor) expresses estrogen receptor and gonadotropin receptor. We found that the entire vocal circuit receives input from a small serotinergic nucleus of the midbrain (rRpd) and Ayako Yamaguchi's lab has shown that serotonin can activate the hindbrain pattern generator that produces advertisement calling and ticking. Serotonin receptors are expressed in components of the hindbrain circuit. It is thus likely that the serotonin receptor is an important player in song generation and it will be interesting to determine how its expression is controlled.

Our original studies of androgen receptor expression were propelled by cloning of the X. laevis androgen receptor gene (He et al., 1990) and the development of methods (in situ hybridization; Dworkin-Rastl et al., 1986) to follow the expression of the receptor and of genes related to vocal function within various tissues and at key developmental stages. We detected two mRNA species that code for androgen receptors in the developing larynx. One, ARb, is expressed throughout development while the other, ARb, is expressed only in precursor cells for laryngeal muscle and cartilage (Fischer et al., 1993,5). The extremely high levels of ARb in developing larynx (due to the prevalence of stem cells for muscle and cartilage) accounts for the very high receptor levels that we detected biochemically in juvenile stages (Kelley et al., 1989). Why stem cells express ARb and its specific function (a role in androgen induced proliferation?) are mysterious.

With respect to the discovery of LM, we knew that male laryngeal muscle fibers are entirely fast twitch due to androgen secretion from the testes. We suspected (an educated guess) that androgen might control muscle fiber type via regulation of a myosin heavy chain (MHC) gene. To see if this was the case we screened a laryngeal cDNA library with a clone of an embryonic MHC gene and found an abundant transcript (which we called laryngeal myosin or LM). The expression of LM is androgen regulated and parallels the developmental trajectory of muscle fiber type expression in males and females (Catz et al., 1992, 1995). The ability of androgen to control fiber types in the male larynx depends on androgen-induced myogenesis (Nasipak, Ph.D. thesis, 2007).

More recently we have been examining the molecular basis of the sex difference in synaptic efficacy at the vocal neuromuscular junction (Wu et al., 2003a,b). A number of studies in the laboratory (e.g., Tobias and Kelley, 1995; Tobias et al., 1998) had established that the developmental default state for these synapses is masculine (weak) and that sexual differentiation occurs as a consequence of exposure to estrogen during development, a female-specific program. To understand the mechanisms by which estrogen produces these effects, we have characterized the ERs of X. laevis and their expression in laryngeal muscle and other tissues. We identified two distinct ERa genes, xlERa1 and xlERa2, which represent, to our knowledge, the first discovery of retained duplicates of the ERa gene in any species. These two genes are highly conserved at the amino acid level but have distinct nucleotide sequences; moreover,ERa2 has no N-terminal domain. Cloning of ERa and ERb in the related species Xenopus tropicalis and phylogenetic analysis indicate that the two xlERa loci were generated by a duplication specifc to the X. laevis lineage—most likely the genome duplication that led to a doubling of the X. laevis chromosome number about 30 million years ago. The primary ER expressed in X. laevis laryngeal muscle is the novel gene xlERa2; ERa1 is primarily expressed in liver, forebrain, and oviduct.

Both of these genes, the AR and LM, were cloned using the "candidate gene" approach. That is to say we made an educated guess as to which genes might be important, based on physiology, development and the scientific literature. I had chosen this tack over an alternative - looking at panels of laryngeal genes upregulated by androgen - because I knew (from 2D protein gels) that there would be many regulated genes and I would have no a priori way to figure out which are particularly important for development, for laryngeal function or for behavior. The more sophisticated current variant of this approach, gene chips, also suffers from similar drawbacks but does have the advantage of being able to follow, in time, programs of gene expression set into motion by hormone exposure. In the end, however, all of these methods are correlative: none reveal what role a specific gene plays in the process of interest. Does the AR actually permit androgen-evoked cell proliferation? Is LM expression essential for the conversion of slow to fast twitch muscles? Another drawback of the candidate gene approach is that one is always looking where the light is; there are no real surprises because the choice of a gene to study bears the weight of everything that is already known about how such a gene should function. Our studies of sexual differentiation had, I thought, hit a wall. We could demonstrate exquisite correlations but could not really probe function. In addition, we had no way to study the process by which gene expression was controlled (the upstream regulators and downstream consequences). The most effective way to get at these questions is to use genetics but Xenopus had no genetics.

So the news here is that several techniques have been developed that facilitate the genetic analysis of behavior in X. laevis. A major advance has been the ability to make transgenic frogs (pioneered by Kris Kroll and Enrique Amaya); frogs that express ectopic genes either universally or under the control of spatially- or temporally-specific promoters. In the vocal system, for example, the transgenic approach means that we can create a dominant-negavtive androgen receptor gene (coding for a receptor that will prevent hormone binding or transport to the nucleus) and knock-down expression of functional androgen receptor in laryngeal muscle (driven off the LM promoter) or in the CNS vocal circuit (driven off of the AR promoter).