A neural network model for generating complex birdsong syntax

The singing behavior of songbirds has been investigated as a model of sequence learning and production. The song of the Bengalese finch, Lonchura striata var. domestica, is well described by a finite state automaton including a stochastic transition of the note sequence, which can be regarded as a higher-order Markov process. Focusing on the neural structure of songbirds, we propose a neural network model that generates higher-order Markov processes. The neurons in the robust nucleus of the archistriatum (RA) encode each note; they are activated by RA-projecting neurons in the HVC (used as a proper name). We hypothesize that the same note included in different chunks is encoded by distinct RA-projecting neuron groups. From this assumption, the output sequence of RA is a higher-order Markov process, even though the RA-projecting neurons in the HVC fire on first-order Markov processes. We developed a neural network model of the local circuits in the HVC that explains the mechanism by which RA-projecting neurons transit stochastically on first-order Markov processes. Numerical simulation showed that this model can generate first-order Markov process song sequences.     

Differential estrogen accumulation among populations of projection neurons in the higher vocal center of male canaries

The higher vocal center (HVC) of adult male canries undergoes a seasonal change in volume that corresponds to seasonal modifications of vocal behavior: HVC is large when birds produce stereotyped song (spring) and is small when birds produce plastic song and add new song syllables into their vocal repertoires (fall). We reported previously that systemic exposure to testosterone (T) produces an increase in the volume of HVC similar to that observed with long-day photoperiods. T-induced growth of HVC occured regardless of wheter the borders of HVC were defined by Nissl-staining, the distribution of androgen-concentrating cells, or the distribution of projection neurons [separate neuronal populations within HVC project to the robust nucleus of the archistriatum (RA) and to Area X of the avian striatum (X)]. In the present study we used steroid autoradiography to determine whether T can influence the distribution of HVC cells that bind estrogen, and we combined estrogen autoradiography with retrograde labeling to determine whether HVC neurons that project to RA versus X differ in their ability to accumulate estrogen. Results showed that T increased the volume of Nissl-defined HVC and although HVC contained a low density of estrogen-concentrating cells, T increased the spatial distribution of these cells to match the Nissl borders of HVC. We also identified a region containing a high density of estrogenconcentrating cells located medial to HVC [we call this region paraHVC (pHVC)], and T also increased the volume of pHVC. pHVC also contained numerous X-projecting neurons, but few if any RA-projecting neurons. Double-labeling analysis revealed the RA-projecting neurons did not accumulate estrogen, a small percentage of X-projecting neurons in HVC accumulated estrogen, and the majority of X-projecting neurons in pHVC showed heavy accumulation of estrogen. The data reported here and in our previous article suggest distinct roles for gonadal steroids within the HVC-pHVC complex: estrogens are concentrated by neurons that project to a striatal region that influences vocal production during song learning (X), whereas androgens are concentrated primarily by neurons that project to a motor region that is involved in vocal production during both song learning and the recitation of already-learned song (RA). © 1995 John Wiley & Sons, Inc.     

Birth,migration, incorporation,and death of vocal control neurons in adult songbirds

Neurogenesis continues in the brain of adult birds. These cells are born in the ventricular zone of the lateral ventricles. Young neurons then migrate long distances guided, in part, by radial cell processes and become incorporated throughout most of the telencephalon. In songbirds, the high vocal center (HVC), which is important for the production of learned song, receives many of its neurons after hatching. HVC neurons which project to the robust nucleus of the archistriatum to form part of the efferent pathway for song production, and HVC interneurons continue to be added throughout life. In contrast, Area X-projecting HVC cells, thought to be part of a circuit necessary for song learning but not essential for adult song production, are only born in the embryo. New neurons in HVC of juvenile and adult birds replace older cells that die. There is a correlation between seasonal cell turnover rates (addition and loss) and testosterone levels in adult male canaries. Available evidence suggests that steroid hormones control the recruitment and/or survival of new HVC neurons, but not their production. The functions of neuronal replacement in adult birds remain unclear. However, rates of HVC neuron turnover are highest at times of year when canaries modify their songs. Replaceable HVC neurons may participate in the modification of perceptual memories or motor programs for song production. In contrast, permanent HVC neurons could hold long-lasting song-related information. The unexpected large-scale production of neurons in the adult brain holds important clues about brain function and, in particular, about the neural control of a learned behavior—birdsong. © 1997 John Wiley & Sons, Inc. J Neurobiol 33: 585–601, 1997     

A species-specific view of song representation in a sensorimotor nucleus

Songbirds constitute a powerful model system for the investigation of how complex vocal communication sounds are represented and generated, offering a neural system in which the brain areas involved in auditory, motor and auditory–motor integration are well known. One brain area of considerable interest is the nucleus HVC. Neurons in the HVC respond vigorously to the presentation of the bird’s own song and display song-related motor activity. In the present paper, we present a synthesis of neurophysiological studies performed in the HVC of one songbird species, the canary (Serinus canaria). These studies, by taking advantage of the singing behavior and song characteristics of the canary, have examined the neuronal representation of the bird’s own song in the HVC. They suggest that breeding cues influence the degree of auditory selectivity of HVC neurons for the bird’s own song over its time-reversed version, without affecting the contribution of spike timing to the information carried by these two song stimuli. Also, while HVC neurons are collectively more responsive to forward playback of the bird’s own song than to its temporally or spectrally modified versions, some are more broadly tuned, with an auditory responsiveness that extends beyond the bird’s own song. Lastly, because the HVC is also involved in song production, we discuss the peripheral control of song production, and suggest that interspecific variations in song production mechanisms could be exploited to improve our understanding of the functional role of the HVC in respiratory–vocal coordination.     

Site-specific retinoic acid production in the brain of adult songbirds

The song system of songbirds, a set of brain nuclei necessary for song learning and production, has distinctive morphological and functional properties. Utilizing differential display, we searched for molecular components involved in song system regulation. We identified a cDNA (zRalDH) that encodes a class 1 aldehyde dehydrogenase. zRalDH was highly expressed in various song nuclei and synthesized retinoic acid efficiently. Brain areas expressing zRalDH generated retinoic acid. Within song nucleus HVC, only projection neurons not undergoing adult neurogenesis expressed zRalDH. Blocking zRalDH activity in the HVC of juveniles interfered with normal song development. Our results provide conclusive evidence for localized retinoic acid synthesis in an adult vertebrate brain and indicate that the retinoic acid-generating system plays a significant role in the maturation of a learned behavior.     

Inhibitory and excitatory mechanisms underlying auditory responses to learned vocalizations in the songbird nucleus HVC

Speech and birdsong require auditory feedback for their development and maintenance, necessitating precise auditory encoding of vocal sounds. In songbirds, the telencephalic song premotor nucleus HVC contains neurons that respond highly selectively to the bird's own song (BOS), a property distinguishing HVC from its auditory afferents. We examined the contribution of inhibitory and excitatory synaptic inputs to BOS-evoked firing in those HVC neurons innervating a pathway essential for audition-dependent vocal plasticity. Using in vivo intracellular techniques, we found that G protein-coupled, potassium-mediated inhibition, tuned to the BOS, interacts with BOS-tuned excitation through several mechanisms to shape neuronal firing patterns. Furthermore, in the absence of this inhibition, the response bias to the BOS increases, reminiscent of cancellation mechanisms in other sensorimotor systems.     

Early developmental stress negatively affects neuronal recruitment to avian song system nucleus HVC

Adverse environmental conditions can impact the life history trajectory of animals. Adaptive responses enable individuals to cope with unfavorable conditions, but altered metabolism and resource allocation can bear long‐term costs. In songbirds, early developmental stress can cause lifelong changes in learned song, a culturally transmitted trait, and nestlings experiencing developmental stress develop smaller song control nucleus HVCs. We investigated whether nutrition‐related developmental stress impacts neurogenesis in HVC, which may explain how poor nutrition leads to smaller HVC volume. We provided different quality diets (LOW and HIGH) by varying the husks‐to‐seeds ratio to zebra finch families for the first 35 days after the young hatched (PHD). At PHD14–18 and again at nutritional independence (PHD35), juveniles were injected with different cell division markers. To monitor growth, we took body measures at PHD10, 17, and 35. At PHD35 the number of newly recruited neurons in HVC and the rate of proliferation in the adjacent ventricular zone (VZ) were counted. Males raised on the LOW diet for their first weeks of life had significantly fewer new neurons in HVC than males raised on the HIGH diet. At the time when these new HVC neurons were born and labeled in the VZ (PHD17) the birds exposed to the LOW diet had significantly lower body mass. At PHD35 body mass or neuronal proliferation no longer differed. Our study shows that even transitory developmental stress can have negative consequences on the cellular processes underlying the development of neural circuits. © 2015 Wiley Periodicals, Inc. Develop Neurobiol 76: 107–118, 2016     

Aromatase inhibition affects testosterone-induced masculinization of song and the neural song system in female canaries

Songbirds have a specialized steroid-sensitive network of brain nuclei, the song system, for controlling song. Most nuclei of the song system express androgen receptors, and the sensory-motor integration nucleus High Vocal Center (HVC) alone also expresses estrogen receptors. Apart from expressing estrogen receptors in the vocal control system, songbirds are unique among birds because they have high concentrations of the estrogen-synthesizing enzyme aromatase in the neostriatum surrounding HVC. However, the role of estrogen in controlling the development of the song structure has been scarcely investigated. In this work, we show that blocking the production of estrogen during testosterone-induced song motor development in adult female canaries alters the song pattern compared to control females treated with testosterone only. These effects were correlated with inhibition of the expression of estrogen-sensitive genes, such as brain-derived nerve growth factor, in HVC. The expression of the ATP-synthase gene, an indicator of cell activity, in HVC, and the size of HVC, were not affected by the treatment. Our results provide the first example of estrogen-sensitive mechanisms controlling the structural features of adult birdsong.     

LMAN lesions prevent song degradation after deafening without reducing HVC neuron addition

In some songbirds perturbing auditory feedback can promote changes in song structure well beyond the end of song learning. One factor that may drive vocal change in such deafened birds is the ongoing addition of new vocal-motor neurons into the song system. Without auditory feedback to guide their incorporation, the addition of these new neurons could disrupt the established song pattern. To assess this hypothesis, the authors determined if neuronal recruitment into the vocal motor nucleus HVC is affected by neural signals that influence vocal change in adult deafened birds. Such signals appear to be conveyed via LMAN, a nucleus in the anterior forebrain that is necessary for vocal change after deafening. Here the authors tested whether LMAN lesions might restrict song degradation after deafening by reducing the addition or survival of new HVC neurons that would otherwise corrupt the ongoing song pattern. Using [3H]thymidine autoradiography to identify neurons generated in adult zebra finches, it was shown here that LMAN lesions do not reduce the number or percent of new HVC neurons surviving for either several weeks or months after [3H]thymidine labeling. However, the authors confirmed previous reports that LMAN lesions restrict vocal change after deafening. These data suggest that neurons incorporated into the adult HVC may form behaviorally adaptive connections without requiring auditory feedback, and that any role such neurons may play in promoting vocal change after adult deafening requires anterior forebrain pathway output.     

HVC microlesions do not destabilize the vocal patterns of adult male zebra finches with prior ablation of LMAN

The songs of adult male zebra finches (Taeniopygia guttata) arise by an integration of activity from two neural pathways that emanate from the telencephalic nucleus HVC (proper name). One pathway descends directly from HVC to the vocal premotor nucleus RA (the robust nucleus of the arcopallium) whereas a second pathway descends from HVC into a basal ganglia circuit (the anterior forebrain pathway, AFP) that also terminates in RA. Although HVC neurons that project directly to RA outnumber those that contribute to the AFP, both populations are distributed throughout HVC. Thus, partial ablation (microlesion) of HVC should damage both pathways in a proportional manner. We report here that bilateral HVC microlesions in adult male zebra finches produce an immediate loss of song stereotypy from which birds recover, in some cases within 3 days. The contribution of the AFP to the onset of song destabilization was tested by ablating the output nucleus of this circuit (LMAN, the lateral magnocellular nucleus of the anterior nidopallium) prior to bilateral HVC microlesions. Song stereotypy was largely unaffected. Together, our findings suggest that adult vocal production involves nonproportional integration of two streams of neural activity with opposing effects on song--HVC's direct projection to RA underlies production of stereotyped song whereas the AFP seems to facilitate vocal variation. However, the rapid recovery of song in birds with HVC microlesions alone suggests the presence of dynamic corrective mechanisms that favor vocal stereotypy.     

A reafferent and feed-forward model of song syntax generation in the Bengalese finch

Adult Bengalese finches generate a variable song that obeys a distinct and individual syntax. The syntax is gradually lost over a period of days after deafening and is recovered when hearing is restored. We present a spiking neuronal network model of the song syntax generation and its loss, based on the assumption that the syntax is stored in reafferent connections from the auditory to the motor control area. Propagating synfire activity in the HVC codes for individual syllables of the song and priming signals from the auditory network reduce the competition between syllables to allow only those transitions that are permitted by the syntax. Both imprinting of song syntax within HVC and the interaction of the reafferent signal with an efference copy of the motor command are sufficient to explain the gradual loss of syntax in the absence of auditory feedback. The model also reproduces for the first time experimental findings on the influence of altered auditory feedback on the song syntax generation, and predicts song- and species-specific low frequency components in the LFP. This study illustrates how sequential compositionality following a defined syntax can be realized in networks of spiking neurons.     

Hormone-induced changes in identified cell populations of the higher vocal center in male canaries

Male canaries revise their vocal repertoire every year. Early work indicated that the volume and neuron number of the song-control nucleus HVC (Higher Vocal Center) declined in late-summer/fall as birds added and deleted syllables from their repertoire, and increased in spring as the set of song syllables stabilized to a fixed number. Seasonal variation in serum testosterone levels suggested that these changes in brain and behavior were regulated by testosterone (T). However, although initial studies describing growth and regression of HVC used Nissl-staining to define its borders, recent experiments that have measured the distribution of identified populations of HVC cells (projection neurons, hormone target cells) suggest that there are no seasonal changes in HVC volume or neuron number. In order to clarify the role of T in the regulation of HVC morphology, we castrated male canaries, maintained them on short (fall-like) days, and treated them with either T, antisteroid drugs, or nothing. After 1 month of treatment, we used a double-labeling technique to characterize HVC projection neurons and androgen target cells. The results showed that hormonal manipulation influenced HVC volume, the density and size of HVC cells, and the absolute number and percentage of androgen target cells in HVC. Hormonal manipulation did not influence the absolute number of cells in HVC. Moreover, the distribution of projection neurons, androgen target cells, and the Nissl-defined borders of HVC were closely aligned in all experimental groups, indicating that exposure to T and/or its metabolites (estradiol and dihydrotestosterone) regulates the overall size of HVC by affecting the distributions of both projection neurons and androgen target cells. Analysis of double-labeling results suggests that T specifically influences both cell size and the ability to accumulate androgen among HVC neurons that project to the robust nucleus of the archistriatum (RA). The results of this study show that steroid hormones exert potent effects on HVC morphology in male canaries, but differences between our results and studies of seasonal males suggest there may be additional factors that can regulate HVC morphology. © 1993 John Wiley & Sons, Inc.     

BDNF mediates the effects of testosterone on the survival of new neurons in an adult brain

New neurons are incorporated into the high vocal center (HVC), a nucleus of the adult canary (Serinus canaria) brain that plays a critical role in the acquisition and production of learned song. Recruitment of new neurons in the HVC is seasonally regulated and depends upon testosterone levels. We show here that brain-derived neurotrophic factor (BDNF) is present in the HVC of adult males but is not detectable in that of females, though the HVC of both sexes has BDNF receptors (TrkB). Testosterone treatment increases the levels of BDNF protein in the female HVC, and BDNF infused into the HVC of adult females triples the number of new neurons. Infusion of a neutralizing antibody to BDNF blocks the testosterone-induced increase in new neurons. Our results demonstrate that BDNF is involved in the regulation of neuronal replacement in the adult canary brain and suggest that the effects of testosterone are mediated through BDNF.     

Intrinsic bursting enhances the robustness of a neural network model of sequence generation by avian brain area HVC

Avian brain area HVC is known to be important for the production of birdsong. In zebra finches, each RA-projecting neuron in HVC emits a single burst of spikes during a song motif. The population of neurons is activated in a precisely timed, stereotyped sequence. We propose a model of these burst sequences that relies on two hypotheses. First, we hypothesize that the sequential order of bursting is reflected in the excitatory synaptic connections between neurons. Second, we propose that the neurons are intrinsically bursting, so that burst duration is set by cellular properties. Our model generates burst sequences similar to those observed in HVC. If intrinsic bursting is removed from the model, burst sequences can also be produced. However, they require more fine-tuning of synaptic strengths, and are therefore less robust. In our model, intrinsic bursting is caused by dendritic calcium spikes, and strong spike frequency adaptation in the soma contributes to burst termination.     

Aromatase inhibition affects testosterone‐induced masculinization of song and the neural song system in female canaries

Songbirds have a specialized steroid‐sensitive network of brain nuclei, the song system, for controlling song. Most nuclei of the song system express androgen receptors, and the sensory‐motor integration nucleus High Vocal Center (HVC) alone also expresses estrogen receptors. Apart from expressing estrogen receptors in the vocal control system, songbirds are unique among birds because they have high concentrations of the estrogen‐synthesizing enzyme aromatase in the neostriatum surrounding HVC. However, the role of estrogen in controlling the development of the song structure has been scarcely investigated. In this work, we show that blocking the production of estrogen during testosterone‐induced song motor development in adult female canaries alters the song pattern compared to control females treated with testosterone only. These effects were correlated with inhibition of the expression of estrogen‐sensitive genes, such as brain‐derived nerve growth factor, in HVC. The expression of the ATP‐synthase gene, an indicator of cell activity, in HVC, and the size of HVC, were not affected by the treatment. Our results provide the first example of estrogen‐sensitive mechanisms controlling the structural features of adult birdsong. © 2002 Wiley Periodicals, Inc. J Neurobiol 54: 370–379, 2003     

Increasing stereotypy in adult zebra finch song correlates with a declining rate of adult neurogenesis

Adult neurogenesis is often correlated with learning new tasks, suggesting that a function of incorporating new neurons is to permit new memory formation. However, in the zebra finch, neurons are added to the song motor pathway throughout life, long after the initial song motor pattern is acquired by about 3 months of age. To explore this paradox, we examined the relationship between adult song structure and neuron addition using sensitive measures of song acoustic structure. We report that between 4 and 15 months of age there was an increase in the stereotypy of fine-grained spectral and temporal features of syllable acoustic structure. These results indicate that the zebra finch continues to refine motor output, perhaps by practice, over a protracted period beyond the time when song is first learned. Over the same age range, there was a decrease in the addition of new neurons to HVC, a region necessary for song production, but not to Area X or the hippocampus, regions not essential for singing. We propose that age-related changes in the stereotypy of syllable acoustic structure and HVC neuron addition are functionally related.     

High vocal center growth and its relation to neurogenesis,neuronal replacement and song acquisition in juvenile canaries

It is generally thought that most circuits of the adult central nervous system (CNS) are sculpted, in part at least, by selective elimination of some of the neurons present in an initial overabundant set. In this scenario, the birth of neurons precedes the period when brain functions, such as learning, first occur. In contrast to this form of brainassembly, we describe here the delayed development of the high vocal center (HVC) and one of its efferent pathways in canaries. The retrograde tracer Fluoro-Gold (FG) was injected into one of HVC's two efferent targets, the nucleus robustus archistriatalis (RA), to define the boundaries of HVC. The HVC grows markedly between 1 and 4 months, invading neighboring territories of the caudal telencephalon. During this same period, 0.43%–0.64% of the HVC neurons present at 1 year of age are labeled per day of [³H]-thymidine injection. [³H]-Thymidine labeling is a marker of cell birth, and during the first 4 months HVC neuron number increases, probably accounting for part of the HVC growth observed. Thereafter, the number of HVC neurons remains constant, but neuronal birth persists. We infer from this that neuronal replacement starts as early as 4 months after hatching and perhaps before then. About half of the neurons born after posthatching day 10 grow an axon to RA to form the main efferent pathway exiting from HVC. HVC growth, neurogenesis, axogenesis, and the observed replacement of neurons happen during the period of juvenile vocal learning. However, the recruitment of neurons that are still present at 1 year shows no particular inflections corresponding to the various stages in song learning, and continues at essentially the same rate after the more stereotyped adult song has been acquired. We suggest that a combination of neurogenesis and neuronal replacement provides unique advantages for learning.     

Activity in a Premotor Cortical Nucleus of Zebra Finches Is Locally Organized and Exhibits Auditory Selectivity in Neurons but Not in Glia

Motor functions are often guided by sensory experience, most convincingly illustrated by complex learned behaviors. Key to sensory guidance in motor areas may be the structural and functional organization of sensory inputs and their evoked responses. We study sensory responses in large populations of neurons and neuron-assistive cells in the songbird motor area HVC, an auditory-vocal brain area involved in sensory learning and in adult song production. HVC spike responses to auditory stimulation display remarkable preference for the bird''s own song (BOS) compared to other stimuli. Using two-photon calcium imaging in anesthetized zebra finches we measure the spatio-temporal structure of baseline activity and of auditory evoked responses in identified populations of HVC cells. We find strong correlations between calcium signal fluctuations in nearby cells of a given type, both in identified neurons and in astroglia. In identified HVC neurons only, auditory stimulation decorrelates ongoing calcium signals, less for BOS than for other sound stimuli. Overall, calcium transients show strong preference for BOS in identified HVC neurons but not in astroglia, showing diversity in local functional organization among identified neuron and astroglia populations.     

The song must go on: resilience of the songbird vocal motor pathway

Stereotyped sequences of neural activity underlie learned vocal behavior in songbirds; principle neurons in the cortical motor nucleus HVC fire in stereotyped sequences with millisecond precision across multiple renditions of a song. The geometry of neural connections underlying these sequences is not known in detail though feed-forward chains are commonly assumed in theoretical models of sequential neural activity. In songbirds, a well-defined cortical-thalamic motor circuit exists but little is known the fine-grain structure of connections within each song nucleus. To examine whether the structure of song is critically dependent on long-range connections within HVC, we bilaterally transected the nucleus along the anterior-posterior axis in normal-hearing and deafened birds. The disruption leads to a slowing of song as well as an increase in acoustic variability. These effects are reversed on a time-scale of days even in deafened birds or in birds that are prevented from singing post-transection. The stereotyped song of zebra finches includes acoustic details that span from milliseconds to seconds--one of the most precise learned behaviors in the animal kingdom. This detailed motor pattern is resilient to disruption of connections at the cortical level, and the details of song variability and duration are maintained by offline homeostasis of the song circuit.