Vocal learning is the ability to imitate or modify vocalizations. Studying the mechanisms requires that we compare the genes, vocal behavior, and associated brain pathways of the few rare groups that have vocal learning with the vast majority of species that do not. Vocal learners include at least three groups of distantly related mammals (humans, cetaceans, and bats) and three groups of distantly related birds (parrots, hummingbirds, and songbirds). Remarkably, although vocal learning groups are distantly related to each other, they share a similar organization of brain vocal pathways: a premotor pathway necessary for vocal learning and a motor pathway necessary for production of learned vocalizations. These forebrain pathways are not found in vocal nonlearners, yet vocal learners and nonlearners possess similar brain pathways for learning and production of other motor behaviors.
We hypothesize that the fundamental difference between vocal learners and nonlearners is one or more genetic differences that control the connection of forebrain motor-learning pathways onto brainstem motor neurons that normally control production of innate vocalizations. Once a vocal-learning circuit is established, we believe that it uses the same molecular mechanisms to perform its functions as motor pathway circuits that we found are adjacent to vocal-learning circuits. We seek to determine the basic mechanisms of vocal learning that are shared across species and conserved over evolution within a broad framework of brain mechanisms of motor learning.
Constructing Vocal-Learning Circuits
We seek to discover the necessary and sufficient molecules that construct the circuits that make the critical differences between vocal-learning and vocal-nonlearning animals. Using behavioral molecular brain mapping, we have found that the three distantly related avian vocal learners possess seven comparable cerebral vocal nuclei active in the production of learned vocalizations—song. Three of the nuclei make up part of the anterior forebrain vocal pathway necessary for song learning. The other four make up part of the posterior vocal pathway that projects onto brainstem vocal nuclei and is necessary for the production of learned song. These nuclei are not present in vocal-nonlearning avian species. Instead, these species only have the brainstem vocal nuclei that are used for the production of innate vocalizations.
We have proposed that the vocal-learning brain systems in birds have parallels with brain pathways for speech production in humans. In particular, the avian anterior forebrain vocal pathway is analogous to cortical-basal-ganglia-thalamic loops of mammals. This may include Broca's area in humans. The avian posterior vocal pathway may be analogous to the face motor cortex in humans. The face motor cortex in humans sends direct projections to brainstem motor neurons involved in the production of learned speech. These direct projections are yet to be seen in vocal-nonlearning birds or mammals, including nonhuman primates.
We found that the seven cerebral nuclei of vocal-learning birds are embedded in or adjacent to brain regions active in the production of head and body movements. These same movement-associated brain regions are in vocal nonlearners and have connectivity that is similar to the cerebral vocal nuclei. A similar relationship appears to exist with the relative positioning of known motor pathways and the hypothesized vocal systems in humans. Based on these findings, we propose a motor theory of vocal-learning origin: Cerebral systems that control vocal learning in distantly related animals evolved out of a preexisting motor system that controls the production and learning of movements.
To test this idea, we have embarked on a study to identify and manipulate the genes that construct vocal-learning circuits. Because there is a robust difference in the presence of forebrain connections to brainstem vocal motor neurons in vocal learners versus nonlearners, we assume that this difference reflects a molecular difference that controls the circuit's construction.
Modifying and Maintaining Vocal-Learning Circuits
When a vocal-learning circuit is formed during development, the circuit must be modified and maintained in order for an individual to learn and control its learned motor behavior. We believe that these processes involve both constitutive and activity-dependent genes. We, along with others, have identified a cascade of gene regulation that is activated by production of learned vocalizations. We call this activation, motor-driven gene expression. To date, 37 activity-dependent genes have been identified that either are up- or down-regulated within the first three hours of producing song in songbirds. We estimate that at least several hundred genes are activated during singing. Each vocal nucleus and cell type has a different, but overlapping, set of activated genes. The mRNA syntheses of the early activated genes are responsive to electrophysiological signals between neurons. Many of the genes are involved in basic cellular processes and a subset is differentially expressed during vocal learning.
To decipher the interactions among the genes in the cascades, in collaboration with Alexander Hartemink (Duke University), we have developed and utilized computational inference algorithms to infer gene regulatory networks based on gene-expressed data from singing animals. The inferred networks serve as hypotheses to be tested with gene manipulation experiments. We have also used the inference algorithms on multielectrode array data from the songbird brain to infer neural flow networks along the anatomical connections within the auditory pathway when songbirds hear songs. The neural flow networks suggest that song is processed in a hierarchal manner, from simple to complex acoustic features in higher stations of the auditory pathway. By combining gene, neural flow, and anatomical networks, we hope to generate a systems view of how forebrain, motor-learning system works.
In related studies, we have found that remarkably subtle variations in behavior in different social contexts can lead to striking differences in brain gene regulation. Other laboratories, such as those of Allison Doupe (University of California, San Francisco), Michale Fee (Massachusetts Institute of Technology), and Gregory Ball (Johns Hopkins University), have studied this phenomenon at the electrophysiological and cellular levels. Collectively, these studies have yielded important insights into the relationships between gene regulation, neural activity, behavior, and learning.
Specifically, when birds direct their songs to other birds, portions of the anterior vocal pathway and one of its target nuclei of the vocal motor pathway, show dramatically less singing-driven gene regulation and less variable neural activity compared to when the birds produce their songs in an undirected manner. Directed song is used for courtship, whereas undirected song is thought to be used for practice or vocal exploration. In this manner, undirected song is delivered slightly slower, has fewer introductory notes, and is more variable in syllable structure and syntax. The gene regulation and neural activity differences in the basal ganglia nuclei of the anterior vocal pathway are thought to be modulated by the midbrain aminergic systems dopamine and norepinephrine. The basal ganglia pathway in turn regulates differences in the vocal motor pathway, which in turn produces the vocal variability differences in directed and undirected singing. The small acoustic differences are perceived as different by the listening birds; the females are more attracted to the acoustics of the male's stereotyped directed songs.
In summary, our research addresses basic questions of biomedical relevance pertaining to the molecular mechanisms that control vocal-learning circuits and behavior. These advancements will give us the ability to decipher the signal transduction and gene regulatory mechanisms of song, a quantifiable natural behavior with analogies to human speech. Since vocal pathways follow a connectivity design similar to other sensory and motor-learning neural systems, the study of vocal-learning systems yields insight into sensorimotor systems of the vertebrate brain.