As a graduate student I was interested in the development of behavior, its neural basis and the role of gonadal steroid hormones in sex differences (see One of the reasons for beginning studies on Xenopus laevis was the ability to sex reverse genetic males and to determine whether the differentiation of the testes in response to genetic instructions is the only controlling factor in masculinization of behavior or whether gene expression plays a role in neural masculinization that is independent of hormonal control. In Drosophila, for example, cell-autonomous genetic programs direct the sex of brain cells without a globally acting, hormonal intermediary (reviewed in Hodgkin, 1990). The masculine behavior that I chose to study was clasping; in Xenopus, the male clasps the female just in front of her legs (an inguinal clasp) while they are mating (while they are in amplexus). The neural basis of clasping behavior had been explored by John Hutchison and by Lester Aronson (at the American Museum of Natural History) and it seemed a reasonable candidate for study. Clasping in males relies entirely on production of circulating androgens from the gonads (Kelley and Pfaff, 1976).

While I was sex reversing a batch of tadpoles in estrogen I also ran studies on adult female frogs and found that simply removing the ovaries and providing a continuous source of androgen (a compressed pellet of testosterone) sufficed to produce clasping of another female that could not be distinguished from male clasping. If hormones were so preeminent in controlling behaviors there seemed little room for genetic factors and so I stopped pursuing the question for the time being. In the meantime, part of the genetic cascade that underlies sexual differentiation (in particular the SRY gene on the mammalian Y chromosome) has been worked out (reviewed in Ramkisson and Goodfellow, 1996) and expression of sry in rodent brain has reopened this issue, thus far without conclusive results (Xu et al., 2002). We have returned to the role of estrogen in sex reversal in our studies of Xenopus tropicalis.

Figure 3. An androgen treated, genetic female frog clasping another female frog.

When I carried out the experiments for my thesis, it was possible to identify hormone target cells in the brain and spinal cord by injecting a frog with radioactive sex steroids, waiting several hours, and determining which neurons accumulated the radioactive hormones (methods pioneered by Walter Stumpf and Don Pfaff). Accumulation is due to the expression of steroid receptors, a member of a large family of nuclear proteins that can act as transcription factors when complexed with the appropriate hormone. How this class of molecules has evolved is an interesting question (Thornton, 2001; 2003); we have analyzed the Xenopus androgen and estrogen receptors using molecular phylogenetic methods (Thornton and Kelley, 1998; Wu et al., 2003a,b).

A striking feature of androgen receptor expression in the Xenopus central nervous system is that the hormone accumulates in neurons implicated in the control of vocal behaviors. For example, the laryngeal motor neurons and their major input nucleus express very high levels of androgen receptor mRNA (Perez et al., 1996) and accumulate radioactive testosterone (Kelley et al., 1975) and dihydrotestosterbone (Kelley, 1981). These are the two major androgenic steroids in Xenopus laevis (Kang et al., 1995); DHT appears to be the more active androgen (reviewed in Kelley, 1996).

Influenced by this observation and by a stint as a post-doctoral fellow on the neural basis of bird song (Kelley and Nottebohm, 1979; Nottebohm et al., 1982), I decided to focus on vocal behavior instead of clasping when I set up my laboratory at Princeton. I didn't know, at that point, what a complex vocal repertoire the species has (see Kelley and Tobias, 1999) nor did I know that the development of masculine properties in the vocal system would be so informative as to the mechanisms of hormone action on the brain (see Kelley, 1996 and Kelley, 2002 for recent reviews). I did know that vocal behaviors were likely to be more entertaining to study than amplexus and so they have proved.