Neural Pathways for Bilateral Vocal Control in Songbirds

Transcription

1 THE JOURNAL OF COMPARATIVE NEUROLOGY 423: (2000) Neural Pathways for Bilateral Vocal Control in Songbirds J. MARTIN WILD, 1 * MATTHEW N. WILLIAMS, 1 AND RODERICK A. SUTHERS 2 1 Department of Anatomy, School of Medicine and Health Science, University of Auckland, Auckland 92019, New Zealand 2 Section of Physiology, Medical Sciences, Indiana University, Bloomington, Indiana ABSTRACT Ipsilateral and contralateral projections of nucleus robustus archistriatalis (RA), a telencephalic vocal premotor nucleus, to respiratory-vocal nuclei in the brainstem were defined in adult male Wasserschlager canaries, grey catbirds, and zebra finches, three songbird species that appear to differ in the degree of lateralized syringeal dominance. In all three species, ipsilateral projections of RA to the medulla included the tracheosyringeal part of the hypoglossal nucleus (XIIts), that innervates the syrinx, the bird s vocal organ, the suprahypoglossal area (SH), and two respiratory-related nuclei, retroambigualis (RAm) and parambigualis (PAm; Reinke and Wild [1998] J Comp Neurol 391: ). Projections of RA to the contralateral XIIts, SH and RAm, were substantial in canaries, which use the left side of the syrinx predominantly during singing; less pronounced in catbirds, which have no lateral dominance for song control; and least pronounced in zebra finches, in which there is a right-sided dominance for song control. There were no obvious differences in the number of crossed projections in birds injected in the left or right RA. Local sources of inputs to XIIts and RAm were defined anatomically in zebra finches and canaries. RAm, including neurons in close proximity to XIIts, was found to project to XIIts and the suprahypoglossal area bilaterally but predominantly ipsilaterally. RAm also had reciprocal connections with its contralateral homologue. These results suggest a pattern of connections between premotor and motor respiratory-vocal nuclei that may be involved in bilateral control of vocal output at medullary levels, a control that involves a high degree of coordination between vocal and respiratory structures on both sides of the body. J. Comp. Neurol. 423: , Wiley-Liss, Inc. Indexing terms: nucleus robustus; nucleus retroambigualis; zebra finch; canary; catbird Song in songbirds, as in humans, is produced by the coordinated action of vocal, supravocal, and respiratory muscles (Wild, 1997). Unlike the larynx of humans, however, the sound source of songbirds is a duplicated structure, with the muscles of each syringeal half receiving independent neural control from the ipsilateral hypoglossal nucleus. This configuration allows songbirds to produce song components using the two sides of the syrinx in either sequential or simultaneous fashion, depending on the species (Suthers, 1997). An early finding in canary, chaffinch, white-throated sparrow, white-crowned sparrow, and java sparrow was that song production is asymmetric, in the sense that the majority of syllables in the songs of these species is produced using the left side of the syrinx, as usually measured by the number of syllables lost or retained following unilateral syringeal denervation (Nottebohm, 1970, 1971, 1972, 1977; Lemon, 1973; Nottebohm and Nottebohm, 1976; Nottebohm et al., 1976; Seller, 1979; Hartley and Suthers, 1989; Suthers, 1992). In the zebra finch, however, it was later found that syringeal denervation did not affect the number of syllables retained postoperatively so much as their acoustic structure, this being more affected following denervation of the right than of the left side of the syrinx (Williams et al., 1992; Floody and Arnold, 1997). Furthermore, with gray catbirds and brown thrashers, Suthers (1990) and Goller and Suthers (1996a) found that neither side of the syrinx was dominant for syllabic production. In these mimic thrushes, different notes or syllables can be produced either on the left side or on the right side of the syrinx simultaneously, thus producing two harmonically unre- *Correspondence to: Dr. J.M. Wild, Department of Anatomy, School of Medicine and Health Science, University of Auckland, Private Bag, Auckland 92019, New Zealand. jm.wild@auckland.ac.nz Received 2 December 1999; Revised 15 March 2000; Accepted 23 March WILEY-LISS, INC.

2 414 J.M. WILD ET AL. lated voices, or on different sides in sequential fashion, the bird rapidly switching back and forth from one side to the other, even within the short time constraints of a single syllable. Frequently in these birds, and in northern cardinals, the left and right sides of the syrinx coordinate to produce a frequency-modulated syllable, in which lower frequency components are produced on the left side of the syrinx and higher frequency components on the right (Nottebohm, 1977; Allan and Suthers, 1994; Suthers and Goller, 1996; reviewed by Suthers, 1997, and by Suthers and Goller, 1997). In mimic thrushes and brown-headed cowbirds, the gating of phonation and the determination of which side of the syrinx produces sound at any particular moment is accomplished by the dorsal syringeal muscles, the motoneurons of which are concentrated caudally in XIIts (Vicario and Nottebohm, 1988; Suthers et al., 1994; Ruan and Suthers, 1996; Goller and Suthers, 1996b; Goller and Larsen, 1997). The control of fundamental frequency, on the other hand, appears to be by the action of ventral syringeal muscles, the motoneurons of which are concentrated rostrally in XIIts (Vicario and Nottebohm, 1988; Ruan and Suthers, 1996). The ventral syringeal muscles, like the expiratory muscles that provide the pressure head, are active on both sides, irrespective of which side of the syrinx is being used to produce the sound (Goller and Suthers, 1996a,b). This seems to imply that fundamental frequency is controlled by the same motor program being sent to both sides of rostral XIIts and that phonation using one side of the syrinx is controlled by different motor programs being sent to the two sides of caudal XIIts. During two-voiced syllables, however, in which air is flowing through both sides of the syrinx, either the same motor program or different motor programs could be sent to the two sides of caudal XIIts. However, whether in any particular species there is or is not a lateral dominance for song production, the question arises of how the bird coordinates the left and right sides of the syrinx and respiratory apparatus to produce the species-typical song pattern, which includes syllables produced on both the left and the right sides. This question is all the more pertinent in the context of the ipsilateral innervation of each syringeal half by the hypoglossal motor nucleus (XIIts) and nerve (ts) in songbirds (Nottebohm and Nottebohm 1976; Nottebohm, 1977; Vicario and Nottebohm, 1988), the absence of any known morphological differences between the song control nuclei on the left and right sides of the brain rostral to the hypoglossal nucleus itself (DeVoogd and Nottebohm, 1981; Nottebohm et al., 1982; DeVoogd et al., 1991), and the absence of any callosal connections between song control nuclei on opposite sides of the brain. However, several neural pathways have recently been described that may function to transfer information derived from the vocal control circuitry in one cerebral hemisphere, or on one side of the brainstem, to that in the other, in a feedback manner (Vates et al., 1997; Striedter and Vu, 1997; Reinke and Wild, 1998). One of these, which originates in the rostroventrolateral medulla and projects bilaterally upon the thalamic nucleus uvaeformis (Uva; Okuhata and Nottebohm, 1992; Striedter and Vu, 1997; Reinke and Wild, 1998), may be particularly concerned with the coordination of the outputs of the left and right premotor vocal control nuclei in the two hemispheres, so that the temporal sequencing of syllables follows the learned, species-typical pattern (Williams and Vicario, 1993; Coleman et al., 1999; see also Vu et al., 1994, 1996). It has not yet been established, however, whether this pathway actually feeds back information that is derived directly from descending motor commands. Abbreviations A Ai Cb Cbd cc CI Cu DIP DLL DM DMP FA FLM FPL GCt HIP HL HV ICo IOS Ipc La LAD MLd MV N nbor NC NIII NVIII OI archistriatum archistriatum intermedium cerebellum tractus spinocerebellaris dorsalis canalis centralis commissura infima nucleus cuneatus nucleus dorsalis intermedius posterior thalami nucleus dorsolateralis anterior thalami, pars lateralis dorsomedial nucleus of the intercollicular complex nucleus dorsomedialis posterior thalami tractus frontoarchistriatalis fasciculus longitudinalis medialis fasciculus prosencephali lateralis substantia grisea centralis tractus habenulointerpeduncularis nucleus habenularis lateralis hyperstriatum ventrale nucleus intercollicularis nucleus infraolivaris superior nucleis isthmi, pars parvocellularis nucleus laminaris lamina archistriatalis dorsalis nucleus mesencephalicus lateralis, pars dorsalis nucleus motorius nervi trigemini neostriatum nucleus of the basal optic root neostriatum caudale nervus oculomotorius nervus octavus nucleus olivaris inferior OM tractus occipitomesencephalicus OS nucleus olivaris superior PA paleostriatum augmentatum PAm nucleus parambigualis (Reinke and Wild, 1998) PM nucleus pontis medialis PP paleostriatum primitivum PrV nucleus sensorius principalis nervi trigemini RA nucleus robustus archistriatalis RAm nucleus retroambigualis (Wild, 1993a) Rt nucleus rotundus Ru nucleus ruber RVL ventrolateral nucleus of the rostral medualla (Wild, 1993b) SCE stratum cellulare externum SGC stratum griseum centrale SGF stratum griseum et fibrosum superficiale SH suprahypoglossal area (Wild, 1993b) SHL nucleus subhabenularis lateralis SpM nucleus spiriformis medialis SSp nucleus supraspinalis Tn nucleus taeniae TPc nucleus tegmenti pedunculopontinus, pars compacta TTD nucleus et tractus descendens nervi trigemini Uva nucleus uvaeformis v ventriculus Vc nucleus descendens nervi trigemini, pars caudalis VI nucleus nervi abducentis VIId nucleus nervi facialis, pars dorsalis VIIv nucleus nervi facialis, pars ventralis X nucleus motorius dorsalis nervi vagi XIIts nucleus nervi hypoglossi, pars tracheosyringealis

3 BILATERAL VOCAL CONTROL PATHWAYS IN SONGBIRDS 415 An alternative to contralateral feedback control of vocal output is contralateral feedforward control, by which motor commands from RA are routed to brainstem nuclei on the opposite side. In zebra finches and green finches, RA has a minor contralateral projection directly upon XIIts and RAm, but it also has more numerous indirect projections, by way of a largely ipsilateral projection upon the pontine nucleus infraolivaris superior (IOS) and the ventrolateral nucleus of the rostral medulla (RVL), which then project upon the vocal motoneurons and respiratory premotor neurons bilaterally (Wild, 1993a,b, 1994; Wild et al., 1997; Reinke and Wild, 1997, 1998). In the present study, we have again focussed our attention on the descending projections of RA, with a view to determining whether there was an obvious difference in the presence or extent of contralateral projections to XIIts and RAm between species that appear to vary in the degree of lateralized syringeal dominance. Thus, a comparison was made between the descending projections of RA in male zebra finches (Taeniopygia guttata; Wild, 1993b; Vicario, 1993), which are said to have a right-sided dominance for song production (Price, 1977; Williams et al., 1992); male Wasserschlager canaries (Serinus canaria), which are strongly left-side-dominant (Nottebohm and Nottebohm, 1976), and whose RA projections to the brainstem, originally thought to be entirely ipsilateral, have not been reexamined in detail since the original lesion/silver degeneration study of Nottebohm et al. (1976); and grey catbirds (Dumatella carolinensis), a species of mimic thrush having no lateralized syringeal dominance (Goller and Suthers, 1996a,b) and whose RA projections have not previously been charted. In addition, because RAm projects bilaterally to spinal motoneurons that innervate expiratory muscles (Wild, 1993a), and because it has also been suggested to project upon XIIts, where it may impose an expiratory-related rhythm (Vicario, 1993; Wild, 1993a), a further objective of the present study was to determine whether RAm projects upon XIIts contralaterally as well as ipsilaterally and whether it has reciprocal projections with its contralateral homologue. Such bilateral and reciprocal projections of RAm could be involved in ensuring that syringeal and expiratory muscle activities are temporally coordinated on both sides of the body during phonation. These local projections were examined in zebra finches and canaries. MATERIALS AND METHODS Surgical procedures were approved by and performed according to the guidelines of the Animal Ethics Committee of the University of Auckland. Four male zebra finches (Taeniopygia guttata), four male Wasserschlager canaries (Serinus canaria), and four male grey catbirds (Dumatella carolinensis) were each anesthetized either with isoflurane gas or with an intramuscular injection of ketamine (50 mg/kg) and xylazine (20 mg/kg) and the heads fixed in a stereotaxic apparatus with the beak tilted downward at 45 to the horizontal. Nucleus RA was identified either using the stereotaxic atlas of Stokes et al. (1974) and/or electrophysiologically by recording multiunit activity with tungsten microelectrodes, the dorsal border of RA being readily identified by a marked increase in neuronal activity and by the fact that, under ketamine/xylazine anesthesia, RA neurons have a characteristic irregular and bursty pattern of discharge. Injections of biotinylated dextran amine [BDA; 10,000 mw; Molecular Probes, Eugene, OR; 10% in 0.1 M phosphate-buffered saline (PBS; ph 7.4); Veenman et al., 1992] were then made in RA via glass micropipettes (outside diameter 30 m) attached to a picospritzer (General Valve, Fairfield, NJ). BDA was the tracer of choice because of its proved ability to label the RA pathway with clarity and completeness (Wild, 1993b). Two birds of each species were injected on the left and two on the right. The injections were deliberately large (30 40 nl) in order to maximize the anterograde labeling of descending axons. In two other male canaries, and in 10 more male zebra finches, unilateral iontophoretic injections of BDA (4 A positive current, 7 seconds on, 7 seconds off, for minutes) were made into either the left or the right XIIts or RAm using glass micropipettes, following identification of the nuclei by a combination of stereotaxis and electrophysiological recording of expiratory-related unit activity (Wild, 1993a), using tungsten microelectrodes (Frederick Haer and Company, 3 5 M ). Again, BDA was the tracer of choice, because of the high reliability with which relatively small, iontophoretic injections could be made in this confined region of the caudal medulla and because of the detail with which individual fibers could be traced to and from the injection site. The birds were allowed to survive 4 or 5 days before they were anesthetized with an intramuscular injection of ketamine (50 mg/kg) and xylazine (20 mg/kg) and perfused transcardially with 50 ml of saline followed by 100 ml 4% paraformaldehyde in 0.1 M phosphate buffer, ph7.4. The calvaria were removed and the head placed in fixative for 3 4 hours before blocking the brain in the transverse plane, extracting it from the skull, and postfixing it overnight. Following equilibration in PBS plus 30% sucrose for 24 hours, the brains were sectioned on a feezing microtome at 40 m in the transverse plane, and the sections were collected in three series. These were treated with 50% methanol and 1% H 2 O 2 to enhance the penetration of secondary reagents and to quench endogenous peroxidase activity and then were incubated at room temperature for hours in streptavidin-peroxidase conjugate (Molecular Probes) at 1:1,000 dilution in PBS, ph 7.4, containing 0.4% Triton X-100. Following washing in PBS, the sections were reacted with 0.025% diaminobenzidine in PBS and H 2 O 2. Sections were mounted on subbed slides, and at least one series was counterstained with Giemsa for the identification of nuclear groups. RESULTS Ipsilateral projections resulting from archistriatal injections As anticipated, none of the injections in RA was confined to the nucleus, there being spread of tracer to the archistriatum surrounding RA and up the pipette track into the overlying caudal neostriatum (Fig. 1D). In no case, however, was there a part of RA that appeared to be free of reaction product. The result of the extranuclear spread of deposited tracer was that, in addition to the diencephalic and mesencephalic terminal fields that are known to originate from RA (e.g., certain dorsal thalamic nuclei and the dorsomedial nucleus (DM) of the intercollicular complex (Nottebohm et al., 1976; Gurney, 1981; Wild, 1993b; Vates et al., 1997), there were terminal fields

4 416 J.M. WILD ET AL. Fig. 1. A F: Schematic drawings of a rostrocaudal series of transverse sections through the brain of a canary, summarizing the projections of the left RA (BDA injection shown as solid black in D) and surrounding archistriatal regions involved by spread from the center of the injection, indicated by hatching in D. Note the BDA-labeled fibers (short wavy lines) in the medial part of the occipitomesencephalic tract (OM) at the base of the hemisphere (B,C), which then enters the dorsal diencephalon (A), where the major terminal fields are in SHL, dorsal DIP, and region ventral to DMP (B). Another dense terminal field in the thalamus is found in the dorsal cap of SpM (B). Major trajectories into the hypothalamus, deep tectum, intercollicular region, and ventral tegmentum are shown as thicker dashed lines terminating in inverted arrowheads. Contralateral projections, which are relatively minor versions of the ipsilateral projections, are indicated by thinner dashed lines. Similar projections were found in catbirds and zebra finches. The continuing trajectory and terminations of OM are summarized in Figure 6A. in various other diencephalic and mesencephalic regions [e.g., other diencephalic nuclei, including the hypothalamus, nucleus spiriformis medialis (SpM), periventricular and deep layers of the optic tectum, nucleus intercollicularis, isthmic regions, and medial tegmentum; Fig. 1A C]. These regions have either not been shown to receive projections from RA or have been shown to receive projections from those archistriatal regions surrounding RA that were inadvertently labeled by spread from the present injections (Wild, 1993b; Vicario, 1993; Mello et al., 1998; Wild and Williams, 2000). For these reasons, and because the primary interest of the present study is in the RA projections to the respiratory-vocal nuclei of the pons and medulla, the present description of the descending projec-

5 BILATERAL VOCAL CONTROL PATHWAYS IN SONGBIRDS 417 tions is largely limited to these levels, where, for the most part, the trajectory follows that previously described specifically for the RA pathway in zebra finches and green finches, excepting contralateral terminations (Vicario, 1993; Wild, 1993b). In each case in each of the three species, labeled fibers coursed rostrally from the injection site, collected in the medial part of the occipitomesencephalic tract (OM) at the base of the hemisphere (Fig. 1A C), and passed through the dorsal diencephalon and into the brainstem. En route, terminal fields were primarily in the nucleus subhabenularis lateralis (SHL), dorsal cap of nucleus spiriformis medialis (SpM), deep and periventricular layers of the optic tectum, and DM of the intercollicular complex (Fig. 1B,C). In the pons, fibers left OM to pass ventrally around and through the subnuclei of the trigeminal motor complex (Fig. 1D) and terminated in the nucleus infraolivaris superior (IOS; Figs. 1E, 2A,C) and, more caudally, in the ventrolateral nucleus of the rostral medulla (RVL; Figs. 1F, 2D,E,H). In all three species at middle and caudal levels of the medulla, fibers left OM medially to terminate within XIIts and the suprahypoglossal area (SH), a relatively cell-poor region that is most evident dorsal and medial to XIIts (Figs. 3A C, 4B D). In catbirds, the terminal field in and around XIIts caudal to the obex formed a characteristic pattern, in which the terminations in SH and the fibers leaving OM to cross to the opposite side were closely applied to the lateral aspect of the caudal pole of the dorsal motor nucleus of the vagus (NX), grasping it in pincer-like fashion (Fig. 4B). Fibers also left OM ventrolaterally to terminate throughout the arc that extends from XIIts and the suprahypoglossal area to the ventrolateral periphery of the medulla (Figs. 3A, 4A). This arc includes the inspiratory premotor nucleus parambigualis (PAm) at periobex levels and the expiratory premotor nucleus retroambigualis more caudally, as far as the spinomedullary junction (Wild, 1993a; Reinke and Wild, 1998). Contralateral projections resulting from archistriatal injections In the diencephalon and mesencephalon, contralateral projections were a minor version of the ipsilateral projections, although projections to the contralateral tectum were limited to periventricular layers (Fig. 1A C). In the pons and medulla of canaries, and to a lesser extent in zebra finches, but not in catbirds, a few fibers were seen to extend into the contralateral pontine tegmentum, where they passed ventrally and caudally to terminate sparsely in IOS and RVL (Figs. 1E,F, 2B,F,G). More caudally in the medulla of canaries, a substantial contingent of fibers left the ipsilateral OM medially and crossed to the opposite side, primarily through the commissura infima (Fig. 3A,B). Some of these fibers extended into and terminated throughout the contralateral RAm and PAm, particularly the former (Fig. 3A). Others terminated within the contralateral XIIts, and particularly within SH (Fig. 3A,B). This pattern of crossed projections was also evident in catbirds (Fig. 4B D), but to a lesser extent than in canaries, and was least evident in zebra finches (Fig. 3C), in which only a few fibers and terminations were observed in the contralateral XIIts and RAm, confirming previous observations (Wild, 1993b). Local sources of projections to XIIts and RAm In canaries and zebra finches, the iontophoretic injections of BDA that were centered on XIIts always included parts of the relatively cell-poor region that surrounds XI- Its, including the suprahypoglossal area. These injections gave rise to a characteristic pattern of anterograde and retrograde labeling within the medulla. Invariably the injections resulted in the labeling of fibers and terminations, but not cell bodies, in the contralateral XIIts (Fig. 5A,F H). In addition, when the injection was centered on XIIts caudal to the obex (see, e.g., Fig. 5F), there were many labeled cell bodies, fibers, and terminations scattered throughout RAm bilaterally, but predominantly ipsilaterally (Fig. 5A D). If, however, the injection was centered on XIIts rostral to the obex (see, e.g., Fig. 5E), the labeled cell bodies, fibers, and terminations tended to cluster in the ipsilateral nucleus parambigualis (Fig. 5E), which extends from the rostral pole of RAm caudal to the obex, to a position rostral to the obex (Reinke and Wild, 1998). Some of the nonmotoneurons that were labeled bilaterally in the medulla as a result of injections centered on XIIts and its surroundings were situated in close proximity to XIIts. These cells and their processes could more easily be identified on the contralateral side, where they were not obscured by the fringes of the injection itself (Fig. 5H). Occasionally, such cells were located between XIIts and the suprajacent dorsal motor nucleus of the vagus, i.e., within SH, but more frequently they were located laterally adjacent to XIIts, from which their labeled processes could be followed into the adjacent XIIts itself (Fig. 5H). Some of these processes were clearly dendrites, but others could have been axons. Retrogradely labeled cells located at a greater distance from XIIts, e.g., those in more ventrolateral parts of RAm on both sides of the brain, sent their axons across the midline through the central medulla and/or through the commissura infima to ramify within the opposite RAm and XIIts. This pattern of bilateral projections of retrogradely labeled neurons located in more or less close proximity to XIIts is consistent with a process of somatopetal and somatofugal transport and was substantiated by the injections in RAm. These gave rise to dense projections to the ipsilateral XIIts and less dense, but nonetheless substantial, projections to the contralateral XIIts and RAm (Fig. 6B D). A summary diagram that is consistent with the local medullary circuitry delineated in the present study and with the descending projections of RA is shown in Figure 6A. The XIIts injections also retrogradely labeled cells in PAm, RVL, IOS, and dorsomedial nucleus of the intercollicular complex (DM), predominantly ipsilaterally, and also supplied projections to these nuclei, the specific origin of which has not been determined but which may include cells in the region surrounding XIIts that were included in the injection. The XIIts injections also retrogradely labeled substantial numbers of cells in all but the dorsal part of the ipsilateral RA. Only in canaries were a very few labeled cells observed in the contralateral RA.

6 Figure 2

7 BILATERAL VOCAL CONTROL PATHWAYS IN SONGBIRDS 419 Fig. 3. A: Darkfield photomicrographic montage showing the bilateral projections of RA to nucleus retroambigualis (RAm), the tracheosyringeal motor nucleus (XIIts), and the suprahypoglossal area (SH) in a male canary. Ipsilateral projections are on the left. Note the labeled fibers crossing to the contralateral side through the commissura infima (CI). B: Darkfield photomicrograph showing more clearly the bilateral projections of RA to XIIts and SH in another canary. C: Darkfield photomicrograph showing the largely ipsilateral projections of RA to XIIts and SH in a zebra finch, with minor projections crossing the midline and sparse terminations in the contralateral XIIts. Scale bars 100 m. DISCUSSION Technical considerations The fact that none of the injections centered on RA was confined to the nucleus was an anticipated consequence of the attempt to label as much of RA as possible by making Fig. 2. Darkfield photomicrographs of transverse sections through the lower pons and rostral medulla, showing the projections of RA to the nucleus infraolivaris superior (IOS) and the ventrolateral nucleus of the rostral medulla (RVL), respectively. A: Ipsilateral IOS in a canary. B: Contralateral IOS in canary. C: Ipsilateral IOS in catbird. D: Ipsilateral RVL in catbird. E: Ipsilateral RVL in canary. F: Contralateral RVL in canary. G: Contralateral RVL in zebra finch. H: Ipsilateral RVL in zebra finch. Scale bars 100 m. large deposits of tracer. The implications of this extranuclear spread of tracer are that some of the projections to the diencephalon and upper brainstem cannot be assumed to originate from RA itself. Projections of RA to certain dorsal thalamic nuclei have been reported in zebra finches and canaries in previous studies (Wild, 1993b; Vates et al., 1997), although in the present study the main terminal field in the dorsal thalamus was in subhabenularis lateralis (SHL) rather than in nucleus dorsomedialis posterior (DMP), as reported by Vates et al. (1997) who used parasagittal sections. Projections to other diencephalic regions, however, including the hypothalamus, to ICo exclusive of DM, to the tectum, to isthmic regions medial to the lateral lemniscal nuclei, and to the medial tegmentum of the mesencephalon almost certainly arise from archistriatal

8 420 J.M. WILD ET AL. Fig. 4. Brightfield photomicrographs showing BDA-labeled terminal fields in the medulla of catbirds, following injections into the left RA (dashed lines in B D indicate the midline). A: Left RAm. B: XIIts and SH caudal to the obex, with crossing fibers and terminations in the contralateral XIIts. C: XIIts and SH caudal to the obex, in another catbird, with crossing fibers and terminations in the contralateral XIIts. D: XIIts at its caudal pole, with crossing fibers and terminations in the contralateral XIIts. Scale bars 200 m in A,B; 100 m in C,D. regions outside RA (Wild et al., 1993; Mello et al., 1998; Wild and Williams, 2000). Injections of tracer into intercollicular regions, or into deep tectal layers in the zebra finch, for instance, retrogradely label large numbers of cells surrounding RA, or on its dorsal and lateral aspects (Wild and Williams, 2000), and a projection from the medial archistriatum to a dorsal cap of SpM, similar to that noted in the present study, has been observed in the pigeon, in which no RA exists (Wild, 1992). In any case, only the projections to the pons and medulla are of central interest in the present context. However, the lateral archistriatum, as well as RA in the medial archistriatum, has major projections to these levels, but 1) the injections centered on RA did not encroach on the lateral archistria-

9 BILATERAL VOCAL CONTROL PATHWAYS IN SONGBIRDS 421 tum and 2) the targets of projections arising in the lateral archisriatum are for the most part different from those of RA and do not include the respiratory-vocal nuclei (Wild and Farabaugh, 1996). On the ipsilateral side of the medulla, RVL, XIIts, SH, RAm, and PAm were almost the sole targets of the projections labeled by the injections centered on RA, and, insofar as no extratelencephalic projections are known to arise from the neostriatum, to which tracer spread from the injections in RA, it is considered that the pontine and medullary projections described here arose from RA and not from surrounding archistriatal or neostriatal regions. It is also considered that this is as true for the contralateral projections as it is for the ipsilateral projections, because, in the pons and medulla, contralateral projections were entirely confined to the respiratory-vocal nuclei. Furthermore, injections in XIIts retrogradely label cells only within RA and not within other parts of the archistriatum (Vicario, 1991; Wild, 1993a,b; Reinke and Wild, 1998). Following XIIts injections, only a very few retrogradely labeled cells were observed in the contralateral RA in the canary and none in the zebra finch. These findings are consistent with the relative proportions of ipsi- and contralateral RA projections in the two species and the small size of the injections, which were delivered iontophoretically and were largely confined to XIIts and the suprahypoglosssal area. Retrogradely labeled cells have been noted in the contralateral RA following large pressure injections of cholera toxin B-chain into OM and RAm of zebra finches (Wild, 1993a). Despite the paucity of retrograde confirmation of contralateral RA projections in the present study, the evidence of contralateral projections resulting from the unilateral RA injections cannot be denied. The contralateral projections of RA also cannot be assumed to arise by way of collaterals of somatopetally labeled neurons in other nuclei, because the retrogradely labeled cells that were observed following RA injections, such as those in the high vocal center (HVc) and the lateral magnocellular nucleus of the anterior neostriatum (lman), known sources of RA afferents, do not project beyond the telencephalon (Nottebohm et al., 1982; Bottjer et al., 1989). Projections of nucleus robustus (RA) to the pons and medulla After providing a relatively small terminal field to the nucleus infraolivaris superior (IOS) in the pons, and a more substantial one to the ventrolateral nucleus of the rostral medulla (RVL), the projections of nucleus robustus (RA) terminate, as summarized in Figure 6, upon XIIts, upon the suprahypoglossal area (SH), and throughout the arc of RAm [and, at more rostral, periobex, levels of the medulla, throughout nucleus parambigualis (PAm); not shown in Fig. 6]. A variable and relatively small proportion of RA projections in different species leaves OM, crosses through the commissura infima (CI), and terminates in all the same nuclei on the contralateral side. RAm projects upon XIIts bilaterally, but again predominanltly ipsilaterally, and also has reciprocal connections with its contralateral homologue. The ipsilateral projections of RA to the pons and medulla in the three species examined were essentially similar and did not appear to differ depending on which side was injected. The projections in the zebra finch have been described previously (Gurney, 1981; Wild, 1993b; Vicario, 1993) and were completely confirmed in the present study using either left- or right-sided injections. Descending projections of RA to the brainstem in canaries have not been reported since Nottebohm et al. s (1976) original description, also in canaries of the Wasserschlager strain, based on silver impregnation of degenerating fibers subsequent to RA lesions. In that study, the RA projections were described as strictly ipsilateral and were predominantly to XIIts, although a relatively minor projection to the ventrolateral brainstem was also observed. In the present study, a pattern of projections resembling that in the zebra finch, with substantial ipsilateral terminations in IOS and RVL, and major ipsilateral terminations in PAm and RAm, as well as in XIIts and SH, is thus described here for the first time in this species. In addition, the RA projections were shown to have a contralateral component. The projections of RA in catbirds are also described for the first time and are generally very similar to those in the other two species, although the pattern of terminations in and around the catbird XIIts is slightly different, there being a characteristic pincer-like clustering of fibers and terminations lateral to the dorsal motor nucleus of the vagus caudal to the obex (see Fig. 4B). The contralateral projections of RA to the medulla do differ between species, however; those in the zebra finch being very sparse, as was noted previously (Wild, 1993b), whereas those in catbirds and canaries are more substantial, particulary in canaries. On the face of it, these results do not provide for a simple, consistent correlation with the degree of lateralized syringeal dominance reported in the three species. That is, if lateralized syringeal dominance is assumed to be greatest in canaries, least in zebra finches, and not present in catbirds, then a correlated pattern of contralateral RA projections might suggest that they are most evident in canaries, which is true, least evident in zebra finches, which is also true, and not present at all in catbirds, which is not true. However, what is meant by lateralized syringeal dominance differs depending on how dominance is measured. Zebra finches have been said to be right-side-dominant for song control, but this refers not so much to the loss of syllables following unilateral syringeal denervation, as it does in canaries, for instance, as to changes in the acoustic structure of syllables following section of the right vs. the left tracheosyringeal nerve or lesion of the high vocal center (Nottebohm and Nottebohm, 1976; Williams et al., 1992; Floody and Arnold, 1997). Syllables that are lost postoperatively in zebra finch song tend to be the high-frequency ones, especially following section of the right tracheosyringeal nerve, a finding that could be seen as consistent with the general tendency in passerine species for high notes to be produced on the right side (Suthers, 1997). It may be, therefore, that there is very little lateralized syringeal dominance for syllabic production in the zebra finch, in the sense in which this term has been applied to canary and sparrow song, and it is tempting to speculate that the paucity of contralateral RA projections to XIIts and RAm in this species reflects this situation. However, in catbirds, in which both sides of the syrinx frequently contribute to syllabic production, and in which there is no lateral dominance for syllabic production, there were considerably more RA projections to the contralateral XIIts than in the zebra finch. It must be concluded, therefore, that the relative number of contralateral RA projections in different species does not necessarily reflect the degree of lateral-

10 Figure 5

11 BILATERAL VOCAL CONTROL PATHWAYS IN SONGBIRDS 423 ized syringeal dominace. Furthermore, even in the strongly lateralized canary, if there is a difference in the extent of contralateral projections originating in the left vs. the right RA, it was not apparent in the small number of animals used in the present study. The pronounced behavioral left right syringeal differences in syllabic production cannot therefore readily be explained on the basis of an obvious anatomical asymmetry in the premotor control pathway. Although contralateral RA projections may not be related to syringeal dominance in a general way, the possibility remains that, within any one species, they are used for the coordination of left and right sides during vocal production. These contralateral inputs of RA to XIIts and RAm are also substantially reinforced by the bilateral projections of more rostral brainstem nuclei, such as IOS and RVL, to which RA also projects bilaterally in canaries and zebra finches (present study; Wild, 1993b; Vicario, 1993). DM also receives a bilateral input from RA in all three species, and this nucleus also projects bilaterally upon XIIts and RAm (Wild, 1993b; Wild et al., 1997). There are thus multiple ways for the RA of one side to influence the activity of XIIts and RAm on both sides of the medulla, one of which is via a direct projection, whereas others are indirect. A possible function of the indirect projections (e.g., via IOS and RVL) might be to introduce temporal delays into the descending motor commands that eventually reach XIIts and RAm, delays that could be important in the temporal patterning of the species-typical song. Local projections of nucleus retroambigualis (RAm) In contrast to the marked differences in the contralateral projections of RA to the medulla in canaries and zebra finches, no such differences were observed in the projections of RAm in these species. In both there was a strong projection of RAm to the ipsilateral XIIts and a comparatively weak, but nonetheless substantial, projection to the Fig. 5. Brightfield photomicrographs showing, in transverse sections, retrograde and/or anterograde labeling in the medulla of canaries and zebra finches, following iontophoretic injections of BDA in XIIts. A,B: Montage of a Nissl-counterstained section caudal to the obex in a canary. The injection of BDA is in the right XIIts, which produced dense retrograde labeling in the ipsilateral RAm, substantial retrograde labeling in the contralateral RAm (boxed area), and anterograde labeling in the contralateral XIIts (the cells in the left XIIts are not retrogradely labeled). C: Higher power view of the boxed area in A. D: Zebra finch, uncounterstained section. Retrograde and anterograde lebeling in the left RAm, and anterograde labeling in the left XIIts, following an injection of BDA in the left XIIts at a more caudal level. E: Zebra finch, Nissl-counterstained section. An injection of BDA in the ventromedial part of the left XIIts rostral to the obex and retrogradely labeled cells in the dorsomedial part of the ipsilateral nucleus parambigualis (PAm). F: Zebra finch, uncounterstained section. An injection in the right XIIts and SH caudal to the obex and anterograde labeling in the contralateral XIIts and SH. G: Zebra finch, Nissl-counterstained section. An injection in the right XIIts and SH rostral to the obex and anterograde labeling in the contralateral XIIts and SH (the cell bodies in the left XIIts are not retrogradely labeled). H: Zebra finch, Nissl-counterstained section. Left XIIts and surrounding regions, including SH, showing retrogradely labeled cells ventrolateral to XIIts, and anterograde labeling in XIIts and SH, following an injection of BDA centered on the right XIIts and SH (see text). Note the processes (dendrites) of the retrogradely labeled cells entering XIIts. Scale bars 200 m in A,B,D G; 100 m in C,H. contralateral XIIts and the opposite RAm, although, again, left right asymmetries in these projections were not obvious to the eye. As described here, RAm includes neurons throughout the whole arc that extends from the borders of XIIts to the ventrolateral periphery of the medulla. This arc is largely coextensive, on the one hand, with the terminal field of nucleus robustus in the caudal medulla (present study; Wild, 1993b) and, on the other, with the origins of bulbospinal neurons that may project upon expiratory motoneurons in the lower thoracic and upper lumbosacral spinal cord (Reinke and Wild, 1998). RAm may even include some cells that lie dorsal to XIIts, i.e., within the suprahypoglossal area (SH). Thus, the bilateral projections of RAm to XIIts may explain the findings of Vicario and Nottebohm (1988), who found that applying microstimulation unilaterally to XIIts in zebra finches resulted in a short-latency response in the ipsilateral syrinx and a longer latency response that included the contralateral syrinx. The contralateral responses were usually evoked from regions dorsal to those from which ipsilateral responses were evoked, an observation that may be consistent with the activation of suprahypoglossal (SH) and RAm neurons that in the present study were found to project to XIIts bilaterally. Similarly, these bilateral projections of RAm to XIIts, and of the relatively minor contralateral projections of nucleus robustus to RAm and XIIts, may also bear on the findings of Paton and Manogue (1982) in canaries and zebra finches to the effect that electrical stimulation of the high vocal center in the telencephalon could evoke a weak response in the contralateral tracheosyringeal nerve, in addition to a strong response in the ipsilateral nerve, only when high rates of stimulation were used. The suggestion that the vocal motoneurons themselves could mediate such a contralateral response by sending a collateral axon across the midline through the commissura infima (Nottebohm, 1980) can now be ruled out on the basis of the present results; no motoneuronal cell bodies were retrogradely labeled in XIIts following injections of tracer into the opposite XIIts. Nottebohm (1980) assumed correctly, however, that an alternative source of bilateral interactions between the two XIIts nuclei might be an interneuron, such as was identified in RAm in the present study. The present results also provide the anatomical substantiation of the suggestions of Manogue and Paton (1982) regarding the medullary source of respiratory gating of vocal output. These authors found that electrical stimulation of premotor vocal control nuclei in the telencephalon of budgerigars and zebra finches produced greater activity in nucleus XIIts or tracheosyringeal nerve during expiration than during inspiration, but RA excitability was not affected by antidromic stimulation of XIIts, suggesting that neither RA nor the high vocal center was the source of the respiratory gating. The present results strongly suggest that RAm and SH neurons that may project bilaterally to XIIts are the source of the increased excitability of XIIts neurons during expiration. It is also possible, however, that the projections to XIIts from inspiratory-related neurons in PAm (Reinke and Wild, 1998; present results) reduce the excitability of XIIts neurons during inspiration. In any case, the local projections to XIIts from RAm and PAm seem likely to be involved in respiratory-vocal coordination, by ensuring that the activity of syringeal motoneurons is timed with that of appropriate expiratory effort during phonation.

12 424 J.M. WILD ET AL. Figure 6

13 BILATERAL VOCAL CONTROL PATHWAYS IN SONGBIRDS 425 In the case of neither RAm nor PAm, however, is it currently known whether the neurons that project to XIIts are the same ones that project upon respiratory motoneurons in the cord (Reinke and Wild, 1998), nor do the present data allow us to determine with confidence whether the projections of RAm involve syringeal motoneurons innervating groups of both dorsal and ventral syringeal muscles or are restricted to only one of these groups. This will be important to determine in future studies (using intracellular labeling techniques) because of the possible role of these groups of motoneurons in lateralized phonation and bilateral frequency control, respectively (Goller and Suthers, 1996a; b). In addition to whatever role the RAm projections to XIIts may play in vocalization, they seem likely to subserve a more basic function related to syringeal muscle tone during respiration, because expiratory-related activity is present in XIIts even during quiet respiration, at least in zebra finches, budgerigars, and canaries (Manogue and Paton, 1982; Vicario and Nottebohm, 1988; Wild, unpublished observations). These findings are supported by the finding that cutting the motor roots of the tracheosyringeal nerve in zebra finches may result in a bird that phonates with every expiration, particularly if this is done on the right side and if the bird is stressed (Wild and Gahr, unublished observations). This is presumably because the syringeal membranes are thereby rendered flaccid and hence partially obstruct the airway (Manogue and Paton 1982). Comparison to mammals A neural system for the mediation of respiratory-vocal control, similar to that delineated in the brainstem of birds, is also present in the brainstem of mammals (Davis et al., 1996a,b; Reinke and Wild, 1997). In particular, RAm and PAm in birds are components of an avian ventral respiratory group (VRG), which, like nucleus retroambiguus (NRA) in the cat, project upon spinal motoneurons innervating respiratory muscles (Holstege, 1989; Wild, 1993a; Reinke and Wild, 1997, 1998). NRA in the cat was also thought to project upon many of the perioral, hypoglossal, and possibly laryngeal motor neurons and was therefore considered to be the nexus of a final common Fig. 6. A: Summary diagram (transverse section through the caudal medulla of a finch) depicting respiratory-vocal pathways consistent with the main neural circuitry delineated in the present study. Descending projections from RA terminate predominantly in the ipsilateral XIIts, in SH, and throughout the entire arc that includes the respiratory premotor nucleus Ram as well as PAm more rostrally. In the canary, in particular, the RA projections also terminate in the contralateral XIIts, SH, and RAm. Cell bodies (black dots) adjacent to XIIts, in more ventrolateral parts of RAm (and in PAm), project bilaterally upon XIIts, and cells in RAm also have reciprocal projections with the contralateral RAm. The extent to which single cells project to XIIts and RAm via branched axons is not known. B: An injection (at the arrow) in the right RAm of a canary that produced predominantly anterograde labeling in XIIts at that level. C: Montage showing the caudal end of the injection depicted in B and anterograde and retrograde labeling (small arrows) in the contralateral RAm. D: Montage of a Nissl-counterstained section through the caudal medulla of a zebra finch showing part of an injection in the right RAm, anterograde labeling in XIIts and SH bilaterally and anterograde labeling and retrogradely labeled cells in the left (contralateral) RAm. Scale bars 200 m. path for vocalization (Holstege, 1989; cf. Thoms and Jürgens, 1987, for the squirrel monkey). However, whether these brainstem projections actually originate from NRA or from neurons in close proximity to NRA is not clear. A significant descending input to NRA in the cat originates from the periaqueductal grey (PAG; Holstege, 1989; Jürgens, 1994; Gerrits and Holstege, 1996), which in mammals, including humans, is considered to be the major pattern generator involved in respiratory-vocal control (Davis et al., 1996a,b). A similar projection to RAm and PAm in birds originates, not in the central grey that surrounds the aqueduct (i.e., GCt), but in the dorsomedial nucleus of the intercollicular complex (DM), which, because of the massive lateral expansion of the tectum in birds, may be homologous with parts of the lateral PAG in mammals. In nonsongbirds, such as pigeons, DM is also likely to be a major pattern generator for respiratory-vocal movements involved in calling (Wild et al., 1997), but, in nonsongbirds and most mammals, apart from the human, there appears to be little or no projection originating in the telencephalon that terminates directly on vocal motoneurons, and the extent of cortical terminations on respiratory premotor neurons is unknown (Kuypers, 1958; Müller-Preuss and Jürgens, 1976; Sutton and Jürgens, 1988). In contrast, songbirds have a major telencephalic source of direct projections onto vocal motoneurons and respiratory premotor neurons, shown here and elsewhere to originate in nucleus robustus (RA; Nottebohm et al., 1976). A major reason for this difference between songbirds and nonsongbirds, and between songbirds and nonhuman primates and other mammals, appears to be that song in songbirds, as with speech in humans, is largely a learned phenomenon and hence requires a dedicated neural system within the telencephalon not only for vocal learning but also for vocal production (Doupe and Kuhl, 1999). As we have shown here and elsewhere, vocal production in birds in general and songbirds in particular is mediated via neural pathways that incorporate bilateral projections for the efficient coordination of structures on both sides of the body. ACKNOWLEDGMENTS This work was supported by Whitehall Foundation, Inc. grant M94RO7 to J.M.W. and National Institutes of Health grant 2 RO1 NS to R.A.S. LITERATURE CITED Allan SE, Suthers RA Lateralization and motor stereotopy of song production in the brown headed cowbird. J Neurobiol 25: Bottjer SW, Halsema KA, Brown SA, Meisner EA Axonal connections of a forebrain nucleus involved with vocal learning in zebra finches. J Comp Neurol 279: Colemen MJ, Sule PJ, Vu ET Recovery of impaired songs following unilateral but not bilateral lesions of nucleus uvaeformis of adult zebra finches. Soc Neurosci Abstr 25:1367. Davis PJ, Zhang SP, Winkworth A, Bandler R. 1996a. Neural control of vocalization: respiratory and emotional influences. J Voice 10: Davis PJ, Zhang SP, Bandler R. 1966b. Midbrain and medullary regulation of vocalization. In: Davis PJ, Fletcher NH, editors, Vocal fold physiology, controlling complexity and chaos. San Diego: Singular Publishing Group, Inc. p DeVoogd TJ, Nottebohm F Sex differences in dendritic morphology of a song control nucleus in the canary: a quantitative Golgi study. J Comp Neurol 196: DeVoogd TJ, Pyskaty DJ, Nottebohm F Lateral asynmmetries and