In our initial studies we determined that both auditory and vocal nuclei in the X. laevis brain accumulate androgens (Kelley, 1980) and that these hormones control whether or not an adult male frog will sing (Wetzel and Kelley, 1983). The steroid hormone target auditory nucleus that we discovered (nucleus laminaris of the torus semicircularis or frog inferior colliculus in the midbrain) is a primary processing area for acoustic information in these frogs (Edwards and Kelley, 2001). It is in this nucleus that the temporal features of frog songs are first decoded (Elliott, Christiansen-Dalsgaard and Kelley, 2008). The vocal nuclei that accumulate androgens also include the laryngeal motor neurons and their major afferent nucleus, DTAM. Biochemical studies revealed that the vocal organ itself, the larynx, also expresses androgen receptor at high levels (Segil et al., 1987) and we became interested in exploring how androgen occupancy of its receptor directs masculine vocal function.

It proved considerably more difficult to induce females to sing using hormone implants (Hannigan and Kelley, 1986) than it had been to induce clasping. Producing a male-like advertisement call was most readily achieved by transplanting a testis into an ovariectomized female (Watson and Kelley, 1992). The testis transplant masculinized the female's larynx and her song circuitry. While it was possible, using this method, to induce advertisement calling even in adult females, the required period of exposure to the transplant was long and the songs produced by juvenile females were more masculinized than those of adult females. It thus appeared that, unlike clasping, the production of male songs might require a sexually differentiated program of development and we set out to determine what this program was.

The vox in vitro- Two new results made this job much easier. We discovered that it is possible to remove the larynx from the frog and produce click trains that closely resemble actual songs by stimulating the attached laryngeal nerve with the appropriate stimulus pattern (Tobias and Kelley, 1987). This was an accidental discovery (see Kelley and Tobias, 1989 for details) made while we were trying to record neural correlates of vocal behavior by stimulating various brain regions (to no avail). We realized that the preparation (informally known as vox in vitro or song in a dish) would be very valuable in determining how cellular and molecular properties of the larynx control the sounds that can actually be produced. Our nerve stimulation mimicked the pattern of neural activity actually sent to the larynx from the brain (Yamaguchi and Kelley, 2000). Nonetheless the cellular composition of the larynx, determined by developmental history of exposure to hormones, produces stringent constraints on song production. This developmental history results in different numbers and types of cells in the male and female larynx (see Kelley, 1996 and Kelley, 2002 for reviews) and these differences account for differences in male and female songs. For example, the female larynx cannot produce rapid click trains; her muscle fiber complement is mostly slow twitch muscle fibers and this cellular characteristic is determined by exposure to gonadal androgens during development (Sassoon et al., 1987; Tobias et al., 1991a,b).

Figure 4. The vox in vitro preparation. The larynx is isolated and the laryngeal nerves stimulated electrically. From this preparation we can record the electromyogram (EMG), the tension transients produced when muscle contractions pull on the arytenoids disks (the sound producing elements) and the actual sounds themselves.

Androgen-induced myogenesis- The second discovery was also an accident. Fernando Nottebohm and his colleagues had demonstrated that forebrain song nuclei were much larger in male canaries and zebra finches than in female brains (Nottebohm and Arnold, 1976). Furthermore, new neurons were added to forebrain song nuclei even in adult animals; John Paton was able to show that some of these new neurons were auditory and thus presumably functional in song control. We knew that there are more vocal motor neurons in male than in female X. laevis brains and set out to determine whether some of these might be generated in adult animals. We used the tritiated thymidine method (the radioactive tracer is incorporated into DNA of actively dividing cells) which was new to us and so needed a positive control to make sure the method was working. The juvenile larynx seemed like a reasonable choice; it grows tremendously when exposed to androgen. The end result was that the control became the experiment; we did not observe any new laryngeal motor neurons in females treated with androgen but we did manage to produce massive cell proliferation in muscle and cartilage precursor cells within the larynx specifically (Sassoon et al., 1986). The complement of vocal motor neurons is achieved during tadpole development, very slowly; the last new neurons in the nucleus are born just as metamorphosis is beginning, before androgen can influence their generation (Gorlick and Kelley, 1987). Androgen does act on laryngeal motor neurons to prevent ontogenetic (developmental or programmed) cell death (Kay et al., 1999) and this process accounts for sex differences in motor neuron numbers.

Myogenesis and fiber type switching - Laryngeal myogenesis and chondrogenesis (formation of muscle and cartilage) are prolonged. New muscle fibers are added throughout development and sub-populations of precursor cells can be induced to divide when exposed to androgen (Fischer et al., 1995) although, again, the testicular transplant is the only treatment that works in adult females (Watson et al., 1992). The number of muscle fibers in the larynx of males depends entirely on exposure to gonadal androgens (Marin et al., 1990; Tobias et al., 1993). We do not know how mitosis of myoblasts and chondroblasts is actually triggered by androgen. Reecently we have been investigating the link between androgen-induced myogenesis and androgen-induced muscle fiber type switching. The underlying hypothesis is that the developing larynx contains a specific stem cell population of muscle precursor cells (myoblasts) that will express the laryngeal myosin heavy chain gene (LM) after fusion (either with each other or with existing myofibers). If so, blocking myogenesis should block the ability of androgen to induce fiber type switching and this is the case (Nasipak and Kelley, 2008).

Androgen competence- The ability of the larynx to respond to androgen with cell proliferation and changes in cell type (slow to fast twitch muscle fibers) depends on the secretion of thyroid hormone and the consequent synthesis of prolactin from the pituitary (Robertson and Kelley, 1996; Cohen and Kelley, 1996; Edwards et al., 1998). We believe that these hormones (TH and prolactin) establish androgen competence or the ability of androgen to direct developmental programs. Androgen competence is the major factor opening the critical or sensitive period for hormone-directed sexual differentiation of the vocal system.