Of Vocal Loss, Songbirds, and the Recovering Brain*
Eric T. Vu, PhD
Division of Neurobiology, Barrow Neurological Institute, St. Joseph’s Hospital and Medical Center, Phoenix, Arizona
*Courtesy of Eric T. Vu, PhD
The Vu Laboratory
Eric Vu received his PhD in neuroscience from the UCLA School of Medicine. His postdoctoral work in the laboratory of Dr. Masakazu Konishi at the California Institute of Technology was awarded the 1995 Capranica Foundation Prize in Neuroethology. While at Barrow, he has received an Alfred P. Sloan Research Fellowship and has been supported by a grant from the National Institute of Mental Health. He is also an Adjunct Professor in the Harrington Department of Bioengineering at Arizona State University. Gustavo Arriaga obtained his BS in psychology from Duke University and began his graduate studies in neurobiology at Duke University in the Fall of 2005. Tuan-Anh Nguyen received his BS in physiology at the University of Arizona. Current student researchers in the laboratory include Amy Vo, Alicia Ortega, Pryianka Sarihan, Albert Hsia, and Elizabeth Carney.
One of the most traumatic and debilitating consequences of a head injury or a stroke is a loss of speech or of the capacity to understand speech. Currently, treatment for such aphasias and related developmental speech disorders is limited to reeducation. To establish a theoretical basis for developing new, physiologically based methods of treatment for impaired speech, our laboratory has studied the neural mechanisms mediating vocal learning and vocal recovery in the songbird. Young songbirds require an adult vocal model, learn and transmit vocal dialects, and exhibit a sensitive period for vocal learning just as young humans do, making songbirds a very promising model for vocal learning and vocal recovery by humans.
Key Words: aphasia, interhemispheric coordination, sensitive period, vocal learning, vocal recovery
Abbreviations used: HVC, acronym used as proper name of nucleus, Uva, nucleus uvaeformis
Loss of vocal control can occur after a brain injury such as a stroke or traumatic head injury. In the United States alone, about 300,000 people each year suffer from a disturbance in language function (aphasia) as a result of a stroke or brain injury. A percentage of these new cases recovers adequately with speech therapy, but little is known about the brain basis for this recovery. Even though intriguing discoveries are being made about the brain mechanisms underlying language and speech production and perception, we are years from a sufficiently detailed understanding of these mechanisms to significantly improve recovery rates in aphasic patients.
Although language is one of the most complex cognitive functions in humans, it is mediated by less complex processes that could be studied more efficiently in animal models. Understanding the neural mechanisms for the key process of vocal learning could lead to new therapeutic strategies for helping patients with speech impairments. Using an animal model permits a large arsenal of neuroscientific research techniques to be applied toward this goal, which has been a major focus of our laboratory.
Acquisition of Song
Besides humans, few other animal species actually learn to vocalize. The vocalizations of most animals are innate. In 1954 William Thorpe demonstrated that some songbirds, such as the chaffinch (Fig. 1), learn to sing, first by memorizing a tutor song and then by practicing until they can reproduce it. Subsequent work showed that the process by which young songbirds learn to sing is remarkably similar to the way human infants learn to speak.
Young songbirds first listen to adult birds sing during the sensory learning phase, and they memorize a tutor song (Fig. 1). Later, during the sensorimotor learning phase, the young birds try to reproduce the tutor song by singing it themselves, at first poorly. They then slowly correct their vocal output by comparing its auditory feedback to the memory of the tutor song. Finally, after much practice, their songs become stable and stereotyped (crystallized song). During the sensorimotor phase, young birds initially emit soft indistinct vocalizations called subsong, which is similar to the babbling of human babies. From subsong they progress to plastic song and then to crystallized song.
Figure 2. Diagram of the song system (Courtesy of Dr. Georg Striedter). The “song system,”the interconnected neural network that mediates birdsong learning. Brain areas in the diagram are color coded according to function. Green areas convey auditory feedback information to certain red areas, which are necessary for song production. Yellow areas are not needed for singing by adult songbirds but are important for song learning by young birds and are also necessary for song recovery after brain injury. Some of these yellow-colored brain areas are homologous to nuclei of the basal ganglia in humans, which have been implicated in motor learning and are adversely affected in Parkinson’s disease. lman = lateral magnocellular nucleus of the anterior nidopallium, nIf = nucleus interface of the nidopallium, hvc = acronym used as proper name of nucleus, ra = robust nucleus of the arcopallium, uva = nucleus uvaeformis, dlm = medial nucleus of the dorsolateral thalamus, dm = dorsomedial intercollicular nucleus, ram = retroambigualis, Xllts = tracheosyringeal part of the hypoglossal nucleus.
Similarities Between Speech and Birdsong
Striking similarities between the acquisition of speech and birdsong and the results of this learning suggest that the brains of humans and songbirds might have solved the goal of vocal learning in fundamentally similar ways. For example, songbirds of the same species learn and transmit different song dialects, just as humans exhibit regional dialects or “accents.” Learning both speech and birdsong requires an adult vocal model to develop normally, and they both involve a practice stage that requires auditory feedback. Just as with songbirds, human babies passively learn the sounds specific to their language well before they begin to babble or to comprehend any of the sounds.
Another key similarity is that learning both speech and birdsong is associated with a sensitive or “critical” period. The waning of the sensitive period is why most adults find it difficult to learn a language fluently. In contrast, if provided an appropriate model, young children require much less effort and time to learn a language. Similarly, songbirds that have been isolated beyond the time of their sensitive period later fail to learn from a model. These similarities suggest that birdsong could be a useful model for understanding the brain mechanisms underlying vocal learning at a very young age in humans and the relearning sometimes required after a brain injury. A considerable advantage of using birdsong learning as a model for speech acquisition is that considerably more is known about the specific neural circuits involved in singing and song learning (Fig. 2) and about the functional roles of the various portions of these circuits than is known about the brain circuits underlying speech learning.
Vocal Loss and Recovery
To model vocal loss from a central injury, we induced severe impairments in singing without causing other gross behavioral impairments. Because lesions were targeted exclusively at parts of the song system, brain injury was minimal. A brain area of particular interest is the forebrain nucleus called HVC. Evidence indicates that neurons in the HVC help encode higher level features of birdsong, such as the complex sequence of vocalizations that make up a song. When HVC is ablated completely on one side (it exists in both hemispheres), song is impaired severely and permanently. However, even adult songbirds can recover their song if lesions of HVC were incomplete. This finding suggests that plasticity in the remaining portion of the lesioned HVC might mediate the recovery of the song.
To identify and distinguish the neural mechanisms involved with this induced plasticity in HVC from other cellular mechanisms that are also induced by injury to a brain area (e.g., those mediating cell death, gliosis, inflammatory responses), we avoided injuring HVC directly by removing a major input to HVC, the thalamic nucleus known as Uva (Fig. 2).
The typical song of an adult zebra finch is a repeated phrase preceded by a series of introductory notes (Fig. 3). Soon after unilateral lesioning of Uva, the song is severely impaired. The song pattern then begins to recover until it is almost the same as before the lesion after about 3 weeks. Thus, the adult songbird brain can adjust to complete removal of one Uva nucleus and recover from the initial singing deficits produced by the injury. The brain, however, cannot adjust to removal of bothUva nuclei. This finding indicates that song recovery after unilateral injury partially depends on adjustments made in the uninjured hemisphere. Similarly, some human case studies have suggested a role for the so-called “silent” right hemisphere in speech recovery after the left hemisphere is injured.
To understand why vocal loss occurs after injury to one or both Uva and to discover the mechanisms by which the brain subsequently recovers song, we can employ neurophysiological techniques that are more precise spatially and temporally than the noninvasive techniques used on humans. Neural activity was recorded simultaneously in the left and right HVC during an impaired song attempt soon after a unilateral Uva lesion. The electrical activity of functionally identified groups of neurons was tracked on a millisecond time scale during singing. Song impairment was associated with a loss in precise coordination of activity between equivalent groups of neurons in the two hemispheres (Fig. 4). Our preliminary data indicate that this interhemispheric coordination of premotor neural activity returns to normal after song is recovered. The finding strongly suggests that the initial impairment in song is caused by a loss of coordination between the two hemispheres.
The discrete song system (Fig. 2) provides a finite and localized set of brain areas to search for the mechanisms of neural plasticity that mediate the recovery of song after unilateral brain injury. A useful strategy is to look for cellular changes that occur at the same rate after brain injury as the rate at which song is recovered. This strategy would be even more powerful if the rate of song recovery could be manipulated because the plastic mechanisms that mediate the recovery most directly would be manipulated in parallel to the recovery. In fact, we have found that song recovery is a fragile process that can be prolonged or sidetracked by preventing auditory feedback, by placing small lesions in other nuclei of the song system, and even by a lack of song practice after the initial brain injury. The latter is reminiscent of the demonstrated benefits of speech therapy for aphasic patients soon after their brain injury.
Reorganization of axon terminals is a major candidate mechanism by which the brain can express plastic change to learn new skills or to recover previously learned ones. In contrast, the adult brain has only a limited capacity to make and incorporate new neurons to replace those that die. We have been quantifying the branching patterns of axons terminals in HVC from one of its sources of input to determine whether the patterns change as song recovers. These changes can be tracked by selectively labeling the axon terminals innervating HVC from only this source, without labeling neuron cell bodies or dendrites at the same time (Fig. 5).
Ultimately, plastic changes that affect song recovery require changes in the expression levels of specific genes in specific groups of neurons. One intriguing possibility is that some of the same genes that regulate the waning of the sensitive period for song learning in young birds are again regulated to support the recovery of song in adult birds. Thus, it is important to identify the sets of genes that are expressed at different levels in the song circuit of birds before and after the sensitive period for song learning. A nationwide group of laboratories studying birdsong, including ours, has formed a consortium to accomplish this goal by using gene microarray technology. We will determine whether the levels of any of these differentially expressed genes change during song recovery, particularly if their levels revert to those seen before the end of the sensitive period. Such a gene would be a candidate for manipulation either to enhance or impede the recovery of song. Perhaps one day, the same strategy might be used on people with developmental or strokeinduced speech disorders to revert specific brain circuits to a more plastic state that would help patients relearn to speak.
- Coleman MJ, Vu ET: Recovery of impaired songs following unilateral but not bilateral lesions of nucleus uvaeformis of adult zebra finches. J Neurobiol 63:70-89, 2005
- Doupe AJ, Kuhl PK: Birdsong and human speech: Common themes and mechanisms. Annu Rev Neurosci 22:567-631, 1999
- Schmidt MF, Ashmore RC, Vu ET: Bilateral control and interhemispheric coordination in the avian song motor system. Ann N Y Acad Sci 1016:171-186, 2004
- Striedter GF, Vu ET: Bilateral feedback projections to the forebrain in the premotor network for singing in zebra finches. J Neurobiol 34:27-40, 1998
- Thorpe WH: The process of song learning in the chaffinch as studied by means of the sound spectrograph. Nature 173:465, 1954
- Vu ET, Mazurek ME, Kuo YC: Identification of a forebrain motor programming network for the learned song of zebra finches. J Neurosci 14:6924-6934, 1994
- Vu ET, Schmidt MF, Mazurek ME: Interhemispheric coordination of premotor neural activity during singing in adult zebra finches. J Neurosci 18:9088-9098, 1998