The influence of melatonin on birdsong and its underlying neuronal correlates : integrating technical advances to address melatonin's role as an endocrine switch

(1)

The influence of melatonin on birdsong and its underlying neuronal correlates

Integrating technical advances to address melatonin’s role as an endocrine switch

Dissertation submitted for the degree of Doctor of Natural Sciences

Presented by

Susanne Christine Seltmann

at the

Faculty of Science Department of Biology

Date of the oral examination: October 24, 2016 First supervisor: Giovanni Galizia

(2)

(3)

(4)

(5)

for my dad

(6)

(7)

Summary

Studying how the nervous system mediates behavior is a central theme of behavioral neurobiology. Investigating the neuronal correlates of birdsong is especially attractive because the ethological aspects as well as the organization of the songbird’s brain are already exceptionally well understood. Particularly fascinating is the process of learning and memorizing birdsong as well as the maintenance of this memory, not least due to its similarities to human language. Sleep seems to play an important role for offline processes involved in song learning in juveniles as well as memory consolidation and song maintenance in adult birds, but the exact processes relevant for switching between motor output during the day and memory consolidation and neuronal rehearsal processes during night are yet to be described. Several facts hint towards melatonin as an important factor in this context. In zebra finches, juveniles have difficulties adjusting their plastic song to their tutor’s song and crystallized song in adult birds deteriorates in absence of natural melatonin production. Furthermore, the presence of melatonin binding sites expressed in several nuclei of the song control system, a neuronal network responsible for learning and producing birdsong, hints towards the importance of melatonin for this system. To investigate effects of melatonin on neuronal processes in the song control system disentangled from the confounding factors of sleep was the main goal of this thesis. By adapting and combining available methods to manipulate natural melatonin levels of the zebra finch (chapter 1) we prepared the ground for focusing on the effects of melatonin on neuronal activity in the song control system. As the sole cortical output of the song system and an important connection between auditory, motor and learning pathways, the nucleus robustus of the arcopallium (RA) provided an ideal target area to record neuronal activity underlying birdsong. By utilizing a

(12)

not only in song production but also in communication via unlearned vocalizations. Finally, by recording single neurons in RA for several consecutive days and at the same time manipulating melatonin levels (chapter 3) we were able to show circadian changes in the activity of single neurons and to describe the influence melatonin might have on this system.

Taken together, we suggest melatonin as an endocrine switch directly influencing crucial parts of the song control system and therefore being involved in the regulation of neuronal activity underlying birdsong.

(13)

Zusammenfassung

Verhalten wird von komplexen neuronalen Systemen erzeugt und gesteuert – diesen Zusammenhang zu verstehen ist Ziel der Verhaltensneurobiologie. Ein besonders gut geeignetes Modell für die Erklärung solcher Zusammenhänge ist der Vogelgesang. So ist nicht nur das Gesangsverhalten an sich bereits ausführlich beschrieben, sondern auch die diesem Verhalten zugrunde liegenden neuronalen Strukturen im Vogelhirn. Besondere Aufmerksamkeit in diesem Kontext verdienen die mit dem Gesang zusammenhängenden Lernprozesse sowie die für seine dauerhafte Erhaltung nötige Gedächtnisleistung, nicht zuletzt aufgrund der Parallelen zur menschlichen Sprache. Sowohl für Jungtiere, die im Begriff sind ihren artspezifischen Gesang zu erlernen, als auch für adulte Tiere mit bereits ausgereiftem Gesang scheint Schlaf eine besonders wichtige Rolle zu spielen und sogenannte “off-line Prozesse“ zu ermöglichen. So werden tagsüber gesammelte Informationen nachts verarbeitet und langfristig gespeichert, um das Erlernen und Beibehalten eines individuellen, stabilen Gesangsmusters zu ermöglichen. Die genauen Abläufe, die diesen Vorgängen zugrunde liegen sind allerdings bislang weitgehend unbekannt. Nervenzellen in bestimmten Bereichen des neuronalen Gesangssystems zeigen beispielsweise stereotypische Aktivitätsmuster, sobald ein Tier beginnt zu singen. Sehr ähnliche Aktivitätsmuster treten auch spontan in den gleichen Gehirnregionen schlafender Tiere auf, führen dann allerdings nicht zur Produktion von Lautäußerungen. Diese sogenannten „Replays“ werden mit der neuronalen Informationsverarbeitung und Speicherung in Zusammenhang gebracht. Sowohl für juvenile als auch für adulte Tiere scheinen diese off-line Prozesse wichtig für die Modulation und Anpassung des Gesangs zu sein. Etliche Studien weisen bereits auf die Funktion von Melatonin in diesem Kontext hin. So können bei juvenilen Zebrafinken

(14)

Melatonin beobachtet werden. Neben diesen offensichtlichen Auswirkungen ist besonders auffällig, dass mehrere wichtige Bereiche des Gesangssystems Melatonin Rezeptoren aufweisen. Dort kann Melatonin gezielt auf die das Gesangslernen und das Singen an sich steuernden neuronalen Strukturen einwirken.

Um die genauen Auswirkungen von Melatonin auf das Gesangssystem zu beschreiben ist es notwendig, diese vom generellen Einfluss von Schlaf zu unterscheiden. Diese konzeptionelle Trennung sowie die Beschreibung der jeweiligen Auswirkungen war das Hauptziel meiner Arbeit.

Das Anpassen und Kombinieren bereits verfügbarer Methoden zur Manipulation sowie die pharmakologische Imitation natürlicher Melatonin Werte von Zebrafinken und ihrer circadianen Rhythmik (Kapitel 1) schufen die Grundlage, um den Einfluss von Melatonin auf das Gesangssystem isoliert zu betrachten und zu bewerten. Im Gesangssystem selbst erwies sich der Nucleus robustus im Arcopallium (RA) als eine gut für die Aufzeichnung neuronaler Aktivität geeignete Region. RA leitet gesammelt Informationen aus verschiedenen Teilen des Gesangssystems an die Motoneuronen, welche die Atmung sowie das Stimmorgan des Vogels kontrollieren, weiter. Gleichzeitig ist er auch in die spontanen nächtlichen Wiederholungen neuronaler Aktivitätsmuster involviert. Mit Hilfe einer kabellosen Aufzeichnungsmethode (Kapitel 2) gelang es uns, neuronale Aktivität in RA mit unterschiedlichen Aspekten der vokalen Kommunikation in Gruppen von Zebrafinken in Zusammenhang zu bringen. Dies erlaubte Einblicke in die Rolle des Gesangssystems für die Produktion von erlerntem Gesang sowie für die Kommunikation mit Hilfe von angeborenen Lauten. Durch die mehrtägige

(15)

Zusammengenommen deuten die hier vorgestellten Ergebnisse darauf hin, dass Melatonin als eine Art hormoneller Regler direkten Einfluss auf wichtige Bereiche des Gesangssystems ausübt und somit potentiell eine wichtige Rolle in den dem Vogelgesang zugrunde liegenden Lernprozessen und der Gedächtniserhaltung spielt.

(16)

(17)

List of abbreviations

AFP anterior forebrain pathway MP vocal motor pathway

Avian brain regions

AIV ventral part of the arcopallium AV avalanche nucleus

CLM caudal lateral mesopallium CM caudal mesopallium

CMM caudal medial mesopallium CN Cochlear nuclei

DLM dorsolateral thalamic nucleus

DM dorsomedial nucleus of the intercollicular complex DMP posterior part of the dorsomedial thalamic nucleus HVC used as a proper name caudal nidopallium

HVC_RA RA-projecting HVC neurons HVC_X RA-projecting Area X neurons ICo intercollicular complex

LC locus coeruleus

LLV ventral portion of the lateral lemniscus MLd dorsal lateral nucleus of the mesencephalon

LMAN lateral magnocellular nucleus of the anterior nidopallium MMAN medial magnocellular nucleus of the anterior nidopallium NCM caudomedial nidopallium

NIf nucleus interfacialis

nXIIst tracheosyringeal portion of the hypoglossal motor nucleus OV nucleus ovoidalis

(18)

RA robust nucleus of the arcopallium RAm nucleus retroambigualis (expiration) SCN suprachiasmatic nuclei

SNc substantia nigra pars compacta VP ventral pallidum

UVA nucleus uvaformis VTA ventral tegmental area

Others

AMPA α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid ANAAT arylalkylamine N-acetyltransferase

BOS bird’s own song

cAMP cyclic adenosine monophosphate CV coefficient of variation

DA dopamine

Dph days post hatch

GABA gamma-Aminobutyric acid ISI inter spike intervals

L/D light/dark rhythm LL constant light NE Norepinephrine

NMDA N-Methyl-D-aspartic acid PSTH peristimulus time histogram RIA Radioimmunoassay

(19)

General introduction

Starting in the early 1950s with work done by William Thorpe and Peter Marler, a systematic analysis of birdsong and singing behavior as well as its learning process emerged. By linking ethology and neurophysiology, Masakazu Konishi’s early work in the 1960s on the neuronal template underlying birdsong set another milestone in this field of research. To date, detailed knowledge about birdsong, its learning process and its neuronal base is available and the underlying brain circuits in several model species like the zebra finch are well described. Nevertheless, the exact mechanisms controlling and regulating the underlying neuronal activity in different situations still need to be identified. Several lines of behavioral and neurobiological evidence suggest that sleep - or one of its confounding factors - have an important influence on both song learning and song maintenance. Various cues point towards the hormone Melatonin to play a key role, but its exact mode of action is difficult to disentangle from the overall effect of sleep. Moreover, its influence on neuronal activity on different parts of the avian song system remains to be investigated. The aim of this study is the integration of new technical advances to approach this topic and to address melatonin’s role in the neuronal control of singing related processes in the avian brain.

Learned birdsong and its significance as a model system

Vocalization is an essential part of communication found in many vertebrate species, with human language as its most sophisticated variant. What distinguishes speech or language from other types of vocal communication is the necessity to learn the required vocalizations as well as their meaning and how to produce them from conspecific tutors. This ability for vocal learning

(20)

Learned birdsong has thus far been documented in parrots (Pittaciformes), hummingbirds (Apodiformes) and songbirds (oscine Passeriformes). Its parallels to human speech and language have drawn scientific attention for a long time, starting with Aristotle more than two thousand years ago, who compared songbirds and children, both acquiring sophisticated, patterned vocalizations, “articulated voice”, from listening to “tutors” in his Historia Animalium (about 350 BCE).

Juvenile songbirds indeed have to learn their song from a tutor. If reared in isolation or only with non singing conspecifics and therefore without tutor input during the “critical period" (Nottebohm 1969, Leppelsack 1986), a highly abnormal song, the so-called isolate song, with some species-specific features but a relatively simple structure will be produced (Konishi 1965, Marler 1970, Feher et al. 2009). If reared socially, juvenile birds undergo two learning phases. During the sensory phase, auditory input is essential to form a neuronal template of the tutor’s song (Konishi 1965). During the sensorimotor phase, young birds start producing first vocalizations, comparable to the babbling of children (Doupe and Kuhl 1999). This

“subsong” then turns into plastic song when the juveniles start adjusting their vocalizations to the tutor song template acquired during the sensory phase and finally transforms into a variation of the tutor song, the “crystallized song” (Konishi 1965, Nottebohm 1968, Marler 1970) (figure 1d). The length of these phases, as well as their collocation, can vary from species to species and can be used to subdivide them into different learning types (Figure 1f).

Closed-ended learners like the zebra finch go through a single sensorimotor phase and maintain a relatively stable song for the rest of their life. Open- ended learners like the canary undergo those phases anew every breeding

(21)

Birdsong in general consists of ordered strings of sounds, so called motifs, with separate sound units, called syllables (Figure 1d). Birds often sing multiple motifs in a row, called a song bout. Some species, like the zebra finch, only repeat a single motif in a highly stereotyped manner, while other bird species such as the canary can produce several different motifs in various orders and thus have a more variable song (for review see ten Cate et al. (2013)). Furthermore, birds also produce several call subtypes, which can be learned or innate (Figure 1a).

The zebra finch as a model organism

The zebra finch (Taeniopygia guttata) has emerged as one of the choice model organisms to investigate vocal learning, birdsong and its neuronal correlates. Zebra finches are gregarious, highly social songbirds, forming monogamous pairs that breed readily in captivity. Male juveniles learn from adult male tutors, often the social father (Immelmann 1969, Slater 1988), copying the highly stereotyped song. Since their initial use in experimental biology (Morris 1954, Immelmann 1969), a wide variation of tools and methodologies have been developed to characterize their song, analyze the degree of accurate copying during vocal learning, map the brain circuits that control singing and song learning, and investigate the physiology of these circuits (Mello 2014). As a result, a large base of knowledge on song production and song learning as well as on the underlying neural mechanisms became available, confirming the zebra finch as an optimal model species to study birdsong and its neuronal correlates.

(22)

(23)

Figure 1

a. In zebra finches, both males and females produce various call subtypes for communication in different situations. While all female and most male calls are innate and are highly stereotyped, male distance calls have learned components and can vary between individuals.

b. – e. Male zebra finches furthermore produce a highly stereotypic song, consisting of introductory elements and song motifs which are composed of syllables and are repeated several times during one song bout (d). To acquire their individual song, males have to be exposed to tutor song (e) during a “sensitive period” and memorize heard song elements as a neuronal template. This takes place during the sensory phase of the learning process. In the sensorimotor phase, the juvenile birds start vocalizing and initially produce a subsong (b, 25 days post hatching (dph)) to train their vocal organs. Consecutively they start producing a plastic song (c, 40 dph), which already shows typical elements of adult song but is still very variable. Around 90 dph the song crystallizes into the very stereotyped final version, the adult song (d, 120 dph). Once crystallized, the song structure is very consistent throughout life, making the zebra finch a “closed ended learner”.

f. Since the zebra finch is a non-seasonal, opportunistic breeder and therefore sings all year round, the learning phases are not coupled to seasons and overlap with each other. In general there are two learning strategies described in songbirds – closed ended learners as the zebra finch and open-ended learners, for example the canary (f). Closed ended seasonal birds go through the sensory learning phase in the breeding season they are born in but only start practicing and finally crystallize their song during the following breeding season. Open-ended learners, as the canary, undergo a sensorimotor phase with plastic song every year, with the crystallization of the individual song only lasting for one breeding season. If open ended learners also go through a sensory phase repeatedly and would therefor be able to acquire new song elements throughout life or if the song is always constructed from elements stored in a neuronal template acquired during a single sensory phase is still unclear to date.

(24)

Neuronal correlates of birdsong – the song system

In songbirds, a circuitry comparable to mammalian cortical- and basal- ganglia networks is involved in the acquisition and production of learned vocalizations (Jarvis 2004). Due to the non-layered, nuclear organization of the avian forebrain, areas that control birdsong are distinct (Nottebohm et al.

1976). This structure allows to study the connectivity of this system and to map the location of individual projection neurons and interneurons within this so-called “song system” (Figure 2a) whose discrete set of brain areas seems to be devoted to singing related behavior (Nottebohm et al. 1976) and might also be involved in reproductive behavior in general (Wild and Botelho 2015).

The zebra finch song system is dimorphic, with nuclei well established in males but very small in females (Nottebohm and Arnold 1976). This dimorphism is also mirrored in behavior as females produce unlearned (innate) calls but do not sing themselves (Vicario 2004). Besides an auditory pathway (Figure 3), two vocal pathways are formed within the song system (Figure 2). The anterior forebrain pathway (AFP) is active during song learning (Bottjer and Arnold 1984, Sohrabji et al. 1990, Scharff and Nottebohm 1991) and responsible for adult song plasticity (Williams and Mehta 1999, Brainard and Doupe 2002) while the posterior vocal pathway or vocal motor pathway (MP) controls song production and innervates vocal and respiratory control centers in the brainstem. Both vocal pathways rise from nucleus HVC (used as a proper name) in the caudal nidopallium and directly (MP) or indirectly (AFP) connect it to the robust nucleus of the arcopallium (RA) in the motor cortex-like avian forebrain. RA is considered the “output-nucleus” of the song control system and projects to the brainstem, more precisely to the

(25)

Auditory response, especially to playback of the bird’s own song (BOS), has been recorded throughout the motor pathway (Katz and Gurney 1981, Mccasland and Konishi 1981, Margoliash 1983), indicating that it might be important for sensory-motor integration and maintenance. Neurons in the AFP also often exhibit song-selective auditory response, suggesting an involvement in processing of song-related auditory information.

This distinct organization of the avian song system has facilitated the usage of methods for obtaining in vivo electrophysiological recordings from different target areas in the song system, increasing our understanding of how individual cells, embedded in a complex circuit, participate in the encoding of learned behaviors (Mello 2014). Both HVC and RA are prime targets for such recordings of pre-motor activity. HVC contains roughly 40,000 neurons projecting to RA (Wang et al. 2002), forwarding sensory input as well as initiating motor output. RA neurons generate complex sequences of spikes during singing (Yu and Margoliash 1996, Leonardo and Fee 2005).

During each song motif, RA projection neurons generate about 10 such high- frequency spike bursts, which are highly stereotyped and precisely timed to the ongoing vocalization (Chi et al. 2007). In contrast to RA neurons, RA- projecting HVC (HVC_RA) neurons generally only produce a single burst during a song motif (Hahnloser et al. 2002, Kozhevnikov and Fee 2007, Amador et al. 2013). Those single bursts are again temporally locked to the ongoing vocalization, seemingly coding for its temporal order. Taken together, HVC_RA neurons, as a population, are thought to burst sequentially throughout the song, with each neuron generating a single burst at one unique time point in the song. What exactly each of these single neuronal bursts represents is currently still under discussion (Amador et al. 2013, Lynch et al. 2016, Picardo et al. 2016). These bursts drive, with a short latency, burst trains in the subset of RA neurons they are connected with, interrupting the regular

(26)

the state of the animal (Dave et al. 1998). This tonic firing patter is intrinsic to RA and not triggered by activity of HVC or the lateral magnocellular nucleus of the anterior nidopallium (LMAN), RA’s two main sources of input (Hahnloser et al. 2006). However, baseline spiking will be interrupted by song related input driven by HVC_RA neurons, resulting in the typical spike burst trains described earlier and followed by a short inhibition before going back to regular firing. The information carried by those bursts is then forwarded downstream by roughly 8000 RA neurons. These are thought to be myotopically organized, projecting to the different brainstem motor neurons where the bursts are translated into the final motor output (Sturdy et al.

2003).

(27)

Figure 2

The vocal motor pathway (MP, purple) controls vocal activity in the brain. It arises in HVC (proper name) and directly projects to the robust nucleus of the arcopallium (RA). RA neurons then connect to the dorsomedial intercollicular nucleus (DM) in the midbrain, involved in call generation, as well as to motor neurons in the tracheosyringeal portion of the hypoglossal motor nucleus (XIIts, controlling the Syrinx) and to respiratory premotor neurons in the ventral respiratory group (VRG). VRG comprises the nucleus retroambigualis (RAm, expiration) and the nucleus parambigualis (PAm, inspiration).

The auditory forebrain pathway (AFP, green) is essential for learning in juvenile birds and responsible for song plasticity in adult birds. It receives input from HVC neurons projecting to Area X, and from the midbrain ventral tegmental area (VTA).

Area X provides inhibitory (GABAergic) input to the medial nucleus of the dorsolateral thalamus (DLM), which in turn provides excitatory input to the lateral portion of the magnocellular nucleus of the anterior nidopallium (LMAN). LMAN

(28)

Figure 3

Two different pathway types convey auditory (red) and recurrent song motor information (blue/orange/pink) to HVC. Auditory information is received by the inner ear (hair cells) and forwarded via the cochlear nuclei (CN) to HVC through two “sub-pathways”. One includes the ventral portion of the lateral lemniscus (LLv) and the thalamic nucleus Uvaformis (UVA). The other one leads to the thalamic nucleus ovoidalis (Ov). Axons from Ov terminate in the telencephalic area Field L, an analog of the mammalian auditory cortex. From Field L, activity is relayed through an interconnected network comprising the caudal medial nidopallium (NCM) and the caudal mesopallium (CM), which in turn projects to HVC, directly as well as indirectly via the nucleus interfacialis (NIf). Singing related feedback of motor activity and possibly also respiratory-related feedback from the brainstem reaches HVC furthermore through a recurrent circuit that includes PAm as part of the ventral respiratory group (VRG), forwarding

(29)

Even though HVC receives input from other brain areas including the forebrain nucleus interface of the nidopallium (NIf), the medial magnocellular nucleus of the anterior nidopallium (MMAN) and the thalamic nucleus Uvaeformis (UVA), HVC seems to be the most upstream site in the song system presenting an explicit song motor representation (Williams and Vicario 1993, Cardin et al. 2005, Coleman and Vu 2005). Since getting auditory feedback – in terms of both listening to tutor song as a juvenile and hearing one’s BOS throughout life – is essential for song learning and song maintenance alike, auditory information must influence the motor circuit (Konishi 1965, Nottebohm 1968, Nordeen and Nordeen 1992, 1993, Okanoya and Yamaguchi 1997, Woolley and Rubel 1997, 1999, Scott et al. 2000, Woolley and Rubel 2002, Liu et al. 2013). Indeed, HVC not only shows distinct, singing related neuronal activity, but is also strongly excitable by presentation of tutor song in awake juvenile birds (Vallentin et al. 2016) and highly selective for BOS in sleeping or anaesthetized adult birds (Margoliash 1983, 1986, Doupe and Konishi 1991, Vicario and Yohay 1993, Dave et al.

1998, Schmidt and Konishi 1998, Theunissen and Doupe 1998, Cardin and Schmidt 2003, Theunissen et al. 2004, Nick and Konishi 2005, Roy and Mooney 2009, Vallentin et al. 2016). Depending on input from HVC, downstream neurons in RA as well as in several other sites of the motor pathway and the AFP can be excited by presentation of BOS playback as well.

The RA of adult birds in particular generates stereotyped sequences of spike bursts driven by combined activity of HVC_RA neurons during singing (Yu and Margoliash 1996, Hahnloser et al. 2002) which can also be elicited by playback of BOS in the sleeping or anaesthetized bird (Dave et al. 1998).

These stereotyped bursts, which are temporally locked to BOS, furthermore occur spontaneously during the night in sleeping birds, matching the

(30)

Sleep as a factor

In reinforcement learning, systems learn through interaction with the environment by trying to optimize some measure of performance (Dave and Margoliash 2000). In the case of birdsong, the tutor song has to be memorized and stored as a neuronal template during the sensory phase, which the juvenile bird can then revert to during the sensorimotor phase and use to improve its plastic song by matching neuronal template and auditory feedback. The striking detail of this process is the timing of learning related song-improvement. Sensorimotor song learning seems to result in part from

“online” mechanisms, whereby during singing, auditory feedback is used to calibrate the song according to the stored neuronal template right away.

Intriguingly, Derégnaucourt and colleagues (2005) found a sleep-associated oscillation in song performance and therefore a delayed learning success.

The biggest changes in song structure did not occur after practicing sessions during the day, but after an initial decrease of syllable structure in the morning during intense morning practice after a night of sleep. The decrease in song quality in the morning – which can be delayed by prevention of song practicing - was furthermore correlated to the quality of the final copy of the tutor song. They therefore speculate that this initial decrease might be linked to the spontaneous nightly replays along with a lack of auditory feedback, confirming the prediction previously made by Dave and colleagues (2000) that birdsong learning depends on “offline processes” occurring during sleep.

Further studies showed that exposure to tutor song during the day is followed by profound and tutor song-specific changes in bursting activity of

(31)

shaped by the interaction of sensory template and auditory feedback during the night (Shank and Margoliash 2009).

Interestingly, such overnight changes in spike bursts of RA neurons are not only observable in juvenile birds. In adult zebra finches, periods of nighttime sleep are commonly associated with small but stable changes in spike burst structure, surprisingly not influencing syllables or song structure (Rauske et al. 2010). Together with the fact that auditory feedback is still necessary for adult birds to sustain a stereotyped song (Nordeen and Nordeen 1992), these changes in spiking activity hint towards the importance of nighttime sleep for song maintenance in adult birds, where spontaneous neuronal replay may be used for fine-tuning the neuronal network. This process may be furthermore important in the context of adult neurogenesis. In the adult songbird brain, new HVC_RA neurons are continuously integrated into the descending motor pathway (Alvarez-Buylla et al. 1990, Kirn et al. 1991) where they have been shown to encode the typical BOS spiking pattern (Hahnloser et al. 2002). In zebra finches, these new neurons do not replace older neurons (as in the canary (Kirn and Nottebohm 1993), but are added, resulting in an age related net growth of the amount of neurons found in RA (Walton et al. 2012). This could explain the increase of stereotypy of zebra finch song with age (Kao and Brainard 2006, Pytte et al. 2007) since new projection neurons, possibly

“programmed” by nightly replays, strengthen the influence of HVC on RA in comparison to the connection between LMAN and RA, which is thought to be responsible for variability in song throughout life (Kao and Brainard 2006).

(32)

Figure 4

The avian circadian pacemaking system comprises three oscillatory components.

The retina and the pineal gland, which are photoreceptive (orange border), and the medial suprachiasmatic nuclei (mSCN) and visual suprachiasmatic nuclei (vSCN), visualized here as area SCN. Pineal and retina, both synchronized to the circadian rhythm by endogenous photo pigments, release melatonin during the night and inhibit activity in SCN (blue arrows). As soon as melatonin production wanes in the morning, oscillators in the SCN become active, rhythmically regulating sympathetic activity and releasing norepinephrine, influencing many peripheral targets. Among these is the pineal gland, where norepinephrine inhibits melatonin synthesis and release (red arrow).

When released during the night, melatonin binds to its receptors, which are expressed in several target areas in the song system. The presence of those

(33)

Hormonal influences on birdsong

Singing in birds is a circadian or seasonal behavior, widely dependent on gonadal steroid hormones which control most of the ontogenetic and adult plasticity of both the song pattern and the vocal control network (Gurney and Konishi 1980, Konishi and Akutagawa 1985, Devoogd 1986, Nordeen et al.

1986, Nottebohm et al. 1986, Nottebohm et al. 1987, Guttinger et al. 1993, Gahr 1994, Gahr and Kosar 1996). Seasonal as well as circadian rhythms are frequently regulated by the hormone melatonin (Gwinner 1981, Cassone 1990, Gwinner et al. 1997, Bentley 2001, Gwinner and Brandstatter 2001), which therefore indirectly influences the regulation of singing initiated by seasonal changes in steroid hormones. The profound knowledge of song learning and song production and their underlying neuronal mechanisms furthermore clearly indicate distinct diurnal differences in neuronal activity found in different parts of the song system. Since differences in overall brain activity as well as on the neuronal level can be observed comparing awake and sleeping or anaesthetized animals, sleep in general appears to be crucial. However, the exact way in which those changes are regulated and triggered is poorly understood so far. Diurnal rhythmicity in birds, including the wake-sleep cycle, is regulated by a pace-making system comparable to the mammalian hypothalamic suprachiasmatic nucleus (SCN), consisting of at least three autonomous circadian oscillators (Brandstaetter 2002) (see Figure 3). One major part of this system, the pineal gland, rhythmically releases the hormone melatonin into the blood stream (Gwinner et al. 2000).

Melatonin (N-acetyl-5-methoxytryptamine) is an indoleamine, whose two functional groups are decisive for specificity of receptor binding (Hardeland et al. 2006). Due to the light-sensitivity of the melatonin synthesis pathway (Klein 2006, 2007), melatonin levels increase soon after the onset of night and remain high for several hours without significant fluctuation during the

(34)

from serotonin (Klein 2006, 2007). The light-insensitive basal expression of the AANAT mRNA and protein as well as the light-dependent rapid changes of cAMP levels lead to the destabilization of the AANAT complex, which is followed by proteolytic destruction of AANAT (Klein 2007). Thus, the absence of light accompanied by low cAMP levels allows a fast increase in active AANAT protein even without up-regulation of AANAT gene expression and therefore regulates the release of melatonin into the system. Since suppressing the natural melatonin production also abolishes circadian rhythms of locomotor activity (Gaston and Menaker 1968), body temperature (Binkley and Menaker 1971), and feeding (Heigl and Gwinner 1994), melatonin seems to also be a likely candidate for a “switch” in other systems including the song control system initiating diurnal changes.

Specific membrane receptors that belong to the G-protein-coupled receptor superfamily mediate melatonin action on a central level. Three different subtypes of melatonin receptors, Mel1a, Mel1b and Mel1c, are expressed in various combinations and densities in different regions of the avian brain (Fusani and Gahr 2015). After reporting of such melatonin binding sites in auditory areas of the avian brain in general (Cassone et al. 1995), the existence of melatonin receptors in the avian song system was shown for HVC, describing a high density of high-affinity Melatonin binding sites which increase in number during song development and maturation of the song system (Gahr and Kosar 1996). With the presence of melatonin binding sites in the vocal control network, a direct role of melatonin for singing and song development not mediated through gonadal steroid production became likely. Other studies found melatonin-binding sites in further parts of the male song system, including RA (Whitfield-Rucker and Cassone 1996), which were

(35)

Since the spontaneous replays observed in RA neurons occur during night in sleeping birds when Melatonin levels are high (Van't Hof and Gwinner 1996, Dave et al. 1998, Dave and Margoliash 2000, Hahnloser et al. 2002), and since melatonin is directly influencing baseline firing of RA neurons (Jansen et al. 2005), this strongly hints at a direct involvement of melatonin on song learning and maintenance. Indeed, males prevented from producing melatonin by pinealectomy or housing in constant light (Deregnaucourt et al.

2012) failed to copy tutor’s song as juveniles and by blocking Mel_1breceptors in adults lead to a significant decrease in song and motif length as well as a higher variability in syllable duration and even transiently lost whole syllables (Jansen et al. 2005).

Research aims

Despite the detailed knowledge about song learning and song production, both, on a behavioral and on a neuronal level, the “missing link” explaining changes in the underlying neuronal properties in different situations still needs to be identified. In this thesis I propose melatonin as this “missing link”. The main aim of this study was to disentangle the effect of melatonin from the overall effect of sleep to investigate its role as an endocrine switch between motor- and auditory activity and possibly from motor production to memory consolidation. In particular, changes in the baseline activity of single RA neurons and a possible connection of those changes with the onset of BOS-sensitivity of the same neurons in adult male birds seemed interesting.

To look at the exact way melatonin could act as such a switch, a technique to permanently record the activity of single neurons in RA over consecutive days while simultaneously manipulating the natural melatonin production in a minimally invasive way became necessary.

Methodological challenges

(36)

recording techniques strongly impair experimental animals and their behavior, one of the main goals of this thesis was to continuously document vocal activity as well as song related neuronal activity in the song system of the zebra finch with as little influence on the bird’s behavior as possible.

Wired recording techniques generally restrict the ability to move and strongly influence behavior – especially in birds. Recordings from head fixed animals are feasible for answering certain questions but certainly not practicable for an extended period of time. Of course it is possible to do recordings from the same animal in different situations during several recording sessions this way, but targeting the same neuron to reliably compare the activity of one distinct unit is not easily achieved. A similar problem arises with the use of electrodes adjustable with a wired micro drive currently used for most studies. The clear advantage is the possibility to adjust the position of the recording electrode to compensate electrode movements and remedy a possible loss of neuronal signal. Nevertheless, continuous recordings of one unit with this method seem to be very challenging and despite the advantage of a bigger sample size, the recording of several different units in one animal under different situations does not allow a direct comparison of individual neuronal activity. Those methods were therefore not well suited for answering the main research questions of this thesis.

Since chronic, extracellular recordings from one single neuron over several days without impairing normal behavior by using a wired connection would be ideal, a novel telemetric transmitter (Schregardus et al. 2006) to permanently record neuronal activity from RA was used for the experiments reported in this thesis. Because the experimental procedure necessary for such electrophysiological recordings in the song system is rather invasive

(37)

To simulate natural melatonin levels at any given time, a reliable method to treat animals with melatonin was necessary as well. Most available methods such as hormone implants or injections are not only invasive but also lead to very high initial peaks in hormone concentration. On the other hand, when administered via food or water, the amount of melatonin cannot be controlled sufficiently. A gentle method developed by Goymann and colleagues (2008) allows the administration of melatonin via a precisely applicable skin cream though. Since the cream was initially established for two other songbird species and used in the wild, the suitability for our experiments using zebra finches was yet to be tested.

Chapter overview

In Chapter 1 of this thesis I present the method to continuously record neuronal activity of a single neuron in a freely behaving bird. Using this method we described the involvement of the song control system in the call based male-female communication of zebra finches. Both male and female birds produce a variety of unlearned call types with a situation dependent call repertoire composition on the individual level (Gill et al. 2015). By using lightweight and wireless recording techniques we were able to document the neuronal correlates of call communication in the song system of adult male zebra finches as well as the individual vocal activity of the whole group. Each bird was equipped with a lightweight (0.6 g including batteries) telemetric microphone developed in house to selectively record vocal activity.

Furthermore, males were chronically implanted with a single channel electrode connected to a lightweight (1.0 g including batteries) telemetric device (Schregardus et al. 2006) to record neuronal activity in RA. This method made it possible to analyze the call communication of different group members and to investigate the involvement of the song system in this

(38)

the production of melatonin during night. Treatment with a melatonin containing skin cream, first described by Goymann et al. (2008) and optimized for the zebra finch, subsequently allowed us to simulate natural melatonin fluctuations during the transition from day to night in awake birds and to maintain high melatonin levels, comparable to those observed under natural conditions, throughout the subjective night.

The combination of wireless neuronal recordings from the song system described in chapter 1 and our method to manipulate natural melatonin levels described in chapter 2 consequently allowed me to investigate the influence of melatonin as well as the effect of its absence on song related activity in the nucleus RA – disentangled from the overall effect of sleep.

The results of the study using this experimental setup are reported in chapter 3 of this thesis. I chronically implanted adult male zebra finches, housed together with their female partners, with our wireless telemetric system and recorded neuronal activity in RA as well as the vocal and general activity of the couple for several days. By exposing the birds to different light situations including constant light (LL), spontaneous “lights off” and a normal LD 14/10 light/dark cycle as well as melatonin treatment, a documentation of changes in the tonic baseline activity of RA on a single-neuron level over time was possible. Furthermore, playback of BOS recorded earlier from the male bird was used to test the responsiveness of the recorded neuron to auditory stimulation during the different phases of the experiment.

The results of the experimental work described in this thesis and their implications will be discussed in the general Discussion in reference to the literature.

(39)

Chapter 1 Melatonin fluctuation and its minimally invasive simulation in the zebra finch

Susanne Seltmann, Lisa Trost, Andries Ter Maat, Manfred Gahr

Seltmann et al. (2016), Natural melatonin fluctuation and its minimally invasive simulation in the zebra finch. PeerJ 4:e1939; DOI

10.7717/peerj.1939

(40)

(41)

Abstract

Melatonin is a key hormone in the regulation of circadian rhythms of vertebrates including songbirds. Understanding diurnal melatonin fluctuations and being able to reverse or simulate natural melatonin levels are critical to investigate the influence of melatonin on various behaviors such as singing in birds. Here we give a detailed overview of natural fluctuations in plasma melatonin concentration throughout the night in the zebra finch. As shown in previous studies, we confirm that “lights off” initiates melatonin production at night in a natural situation. Notably, we find that melatonin levels return to daytime levels as early as two hours prior to the end of the dark-phase in some individuals and 30 minutes before “lights on” in all animals, suggesting that the presence of light in the morning is not essential for cessation of melatonin production in zebra finches. Thus, the duration of melatonin production seems not to be specified by the length of night and might therefore be less likely to directly couple circadian and annual rhythms.

Additionally, we show that natural melatonin levels can be successfully simulated through a combination of light-treatment (daytime levels during subjective night) and the application of melatonin containing skin-cream (nighttime levels during subjective day). Moreover, natural levels and their fluctuation in the transition from day to night can be imitated, enabling the decoupling of the effects of melatonin, for example on neuronal activity, from sleep and circadian rhythmicity. Taken together, our high-resolution profile of natural melatonin levels and manipulation techniques open up new possibilities to answer various melatonin related questions in songbirds.

(42)

Introduction

The “night-hormone” melatonin plays an essential role in maintaining and synchronizing the circadian rhythmicity of various physiological processes such as sleep, locomotion, temperature regulation and singing behavior of birds (Cassone and Menaker 1984, Gwinner et al. 1997, Bentley 2001, Brandstatter 2003). In songbirds such as the zebra finch, melatonin directly affects brain areas involved in learning and producing birdsong (Gahr and Kosar 1996, Whitfield-Rucker and Cassone 1996, Bentley et al. 1999, Bentley and Ball 2000, Jansen et al. 2005, Deregnaucourt et al. 2012, Fusani and Gahr 2015). Mel 1B receptors, which bind melatonin, are expressed in brain nuclei HVC (formerly known as “high vocal center”, now used as a formal name; located in nidopallium) and RA (Nucleus robustus arcopallii) of the avian song control system (Fusani and Gahr 2015), a model system for understanding motor learning in general, including language learning in humans. Both HVC and RA are involved in song learning as well as song production (Nottebohm et al. 1976, Vu et al. 1994). To investigate the role of melatonin in these and many other processes, detailed knowledge about the circadian melatonin rhythm is necessary. Likewise, a minimally invasive method that manipulates melatonin availability is required, so that its effects can be disentangled from other covariates such as sleep.

Melatonin diel rhythmicity, with high blood plasma concentration during the night and low levels during the day, is driven by a circadian clock and fine- tuned by light (Ralph et al. 1967, Klein 2006, Tan et al. 2010). Melatonin synthesis and release seem to precisely follow the photoperiod and therefore couple circadian and seasonal rhythms (Ralph et al. 1967, Perreau-Lenz et al.

(43)

Follett 1993) still lack the resolution or – as well as in the mentioned songbird studies – a statistically validated tracking of changes in melatonin blood concentration. Thus, the first aim of this study is to precisely monitor and statistically validate circadian changes in melatonin levels in zebra finches with a focus on the transition between day and night.

The second aim of this study is to validate a combination of minimally invasive methods to successfully simulate natural melatonin levels as well as the circadian fluctuation of these levels. Conventional methods for increasing melatonin levels like direct hormone injections or implants are rather invasive and can therefore influence behavior. Moreover, these methods cause abnormally high peaks of hormone blood concentration after treatment and are therefore unsuitable for imitation of the natural melatonin production.

Additionally, implants cause a continuous release of hormones and therefore exclude the simulation of the natural production with its circadian rhythmicity (Goymann et al. 2008). Similarly, while oral administration of melatonin leads to a sufficiently elevated melatonin level to change behavior (Binkley and Mosher 1985, Heigl and Gwinner 1994, Pohl 1996, Neddegaard and Kennaway 1999), neither the exact amount of melatonin taken in by the animal nor the duration of the elevation can be controlled reliably. Recent advances make the reliable and non-invasive simulation of natural melatonin levels feasible. For simulating natural melatonin levels in birds, Goymann and colleagues introduced a minimally invasive method to raise melatonin levels by applying melatonin dissolved in Eucerinum anhydricum (Beiersdorf;

termed “melatonin cream” in the present paper) to the birds’ skin (Goymann et al. 2008). This method was originally developed for wild birds to ensure the least invasive treatment possible to increase natural melatonin levels at night, but it seems to be similarly ideal for manipulating hormone levels of birds in experimental setups under various situations in the laboratory.

(44)

method of choice for suppressing natural melatonin production (McMillan 1972, Binkley 1976, Van't Hof and Gwinner 1996). However, the extremely invasive nature of this procedure may have adverse effects on behavior and does certainly impair follow-up experiments as electrophysiological measurements of brain activity. Although keeping animals in constant light might affect other bodily functions as well as behavior, it is a powerful tool to temporary suppress melatonin production (Turek et al. 1976). Consequently, establishing a combination of minimally invasive methods that allow for successful emulation of natural melatonin levels in zebra finches under laboratory conditions could provide a powerful tool for answering various questions in songbirds.

(45)

Materials and Methods Animals

Adult male and female zebra finches from the breeding colony in Seewiesen, Germany were used for the experiment. Zebra finches were reared and subsequently housed socially in mixed-sex colonies in a 14/10 light/dark (LD) light cycle. For the experiment, birds were transferred and kept as couples or in mixed sex groups of five in a 14/10 (1200 lux/<0.0001 lux) LD cycle in sound- and light-proof boxes two weeks prior to, and during, the experiment.

This housing situation allowed us to minimize any outside influences as well as disturbances of birds kept in other boxes caused by the experiment itself such as melatonin treatment, catching individuals, or using night light for blood sampling at different time points during the experiment. Resampling was furthermore excluded by using individual ring numbers for each bird. All housing conditions were in accordance with EU regulations, and experimental procedures were performed in compliance with national legislation on animal experimentation (animal testing license of the government of Upper Bavaria 55.2-1-54-2531-108-10).

Melatonin application

Two dilutions of melatonin cream, 13 μg/ml (“low concentration”) and 130 μg/ml (“high concentration”, comparable to the lowest concentration used by Goymann et al. for a different songbirds species in their publication) were prepared according to Goymann et al. (Goymann et al. 2008) by dissolving melatonin (Sigma M 5250) in 200 μl ethanol and mixing the solution with Eucerin cream (Eucerinum anhydricum; Beiersdorf AG, Hamburg, Germany).

For each treatment 50 μl cream (corresponding to 0.65 μg melatonin in the low concentration cream and 6.5 μg melatonin in the high concentration

(46)

Blood sampling

150-200 μl of blood were collected from the alar vein into heparinized micro- capillaries and centrifuged immediately after collection for ten minutes at 2500 RPM (Sigma 1-14 centrifuge). Plasma was stored at -80°C until analysis.

Melatonin levels in different conditions

To characterize the difference in natural melatonin levels during day and night as well as the effects of our treatments on these levels, we repeatedly took blood samples of 19 zebra finches (11 male and 8 female). Every individual was sampled once in each of the five conditions. A two-week interval was maintained between subsequent sampling events to give the birds enough time to recover. All blood samples were analyzed in a single radioimmunoassay (for details see supplementary material).

First, to estimate the average melatonin level during different phases of the diurnal cycle, we took blood samples five hours after lights on as well as three hours after lights off during a normal 14/10 LD light cycle. For the night sampling, individual birds were removed from their lightproof boxes and prepared for blood collection in the dark. After reclosing the lightproof box, a night-light was turned on immediately prior to bleeding, thereby minimizing the disturbance and light exposure of all individuals including the ones bled at a single time point.

Subsequently, to evaluate the effects of constant light (LL) and melatonin treatment, we took samples of the same individuals during LL without melatonin treatment as well as during LL after treatment with the two

(47)

Natural nighttime profile and influence of light/melatonin treatment To obtain high-resolution data, we collected three series of samples over a 12-hour (normal night, no treatment) / 24-hour (LL / high concentration cream treatment) / 84-hour (LL / low concentration cream treatment) period to compare natural melatonin fluctuations over night with melatonin levels observed after treatment with both melatonin cream concentrations.

A total of 277 samples were collected from different adult male and female zebra finches and analyzed in four different radioimmunoassays (for a detailed overview of sampling see table S1). In total, seven samples had to be excluded (contaminated during sampling or lost during analysis in the lab).

The extension of the sampling period for the low and the high concentration cream from 12 to 24 and 84 hours respectively was a result of evaluating the samples collected in a first run, up to 12 hours after treatment with melatonin.

For the natural nighttime profile a total of 96 adult birds (48 male, 47 female) were sampled (once each), covering 12 hours, including the 10-hour dark period of a 14/10 LD cycle. Two series of samples were collected from birds treated with the two different melatonin cream concentrations covering a 24- hour (high concentration treatment) and an 84-hour period (low concentration treatment) respectively. For these two series, birds were shifted to LL three days prior to treatment, and 50 μl of either high or low concentration melatonin cream was applied at subjective lights-off time. For the high concentration melatonin cream series we collected blood samples from 81 adult zebra finches (39 male and 42 female) in total. For the low concentration melatonin cream series we sampled a total of 100 adult zebra finches (46 male and 54 female).

Tarsus length and body mass were measured after blood sampling in a

(48)

Radioimmunoassay

The concentration of melatonin in the blood plasma of all collected samples was determined by direct radioimmunoassay (RIA) following Fusani and Gwinner (2004) and Goymann (2007, 2008). The lower detection limit of the assays ranged from 4.3 to 15.6 pg/ml depending on the sampled volume and the recovery value, the upper detection limit was between 5846 and 8276 pg/ml (see table S1). Intra-assay coefficient of variation (CV) ranged from 0.9 to 9%, and intra-extraction CV between 0.4 and 5.8%. The inter-assay CV was 10.4% for the four assays used for the natural nighttime profile and the influence of light / melatonin treatment was 10.4% (for more details see table S1).

Statistics

Statistical analysis was done with JMP 10.0 (SAS Institute Inc.). To test for differences in the five treatment groups (Melatonin levels in different conditions) we used a REML (restricted maximum likelihood) analysis with treatment and sex as factors, and animal ID as a random factor. Melatonin levels were log₁₀-transformed to meet requirements for parametrical statistical testing. Significance of specific differences was determined posthoc using a Tukey LSD test at alpha 0.01 to minimize the chance of false positives. Raw data were also analyzed using Friedman randomized blocks followed by post-hoc testing (Fig. 1). This yielded the same result as the parametric test after log-transformation. Fluctuations of blood plasma melatonin concentrations under normal LD conditions, and after treatment with both cream concentrations under LL conditions were compared using a

(49)

plasma melatonin levels return to daytime levels after treatment, melatonin levels at all time points after treatment with the low concentration cream were compared to the control (0.5 h before treatment) using a Dunnett’s test (Fig. S2).

Results

Mean melatonin levels under different conditions

Samples of all 19 birds measured in five different conditions showed substantial variation in melatonin concentration (REML fixed effects test of treatment: numerator F_{4, 88} = 62.27; P <0.0001, followed by the Tukey test at P < 0.01; Fig. 1). Melatonin concentrations after five hours of daylight during a normal day (LD 14/10) were significantly lower than three hours after onset of night (152.75 ± 22.398 pg/ml and 716.68 ± 52.793 pg/ml respectively; P <

0.01, Tukey test). After keeping birds in LL for three days, samples collected three hours after the subjective onset of night showed melatonin concentrations comparable to those measured during daytime (203.4 ± 33.964 pg/ml; P > 0.01, Tukey test). Treated with high concentration melatonin cream and measured three hours after treatment, birds showed strongly elevated melatonin levels (3104.12 ± 782.122 pg/ml; P < 0.01, Tukey test), significantly higher than under normal nighttime conditions. Birds treated with the low concentration melatonin cream showed an average melatonin level (1539.06 ± 349 pg/ml) that was not significantly different from birds sampled under natural night conditions (P > 0.01, Tukey test). A significant difference between male and female birds could not be measured under any condition.

(50)

Figure 1. Plasma melatonin concentrations of 19 adult zebra finches measured during day (LD day), night (LD night), nighttime in constant light (LL night), with high concentration melatonin cream (LL night mel high) and with low concentration melatonin cream (LL night mel low); LD indicates a normal light cycle; LL indicates constant light. Groups with different letters (A/B/C) are significantly different (Friedman’s post-hoc test, p < 0.01).

Natural nighttime profile of plasma melatonin concentrations

To quantify how melatonin levels vary during a normal day/night cycle, we measured melatonin levels at 15 time points. Analysis of variance showed that levels differed significantly between sampling time points (F_14,81 = 24.21, P < 0.0001). Birds kept under natural conditions showed daytime melatonin

6070

10080 200 300 400 500600 700800

1000

2000 3000 4000 5000 60007000

100008000

melatonin conc. (pg/ml)

LD day LD night

LL night mel low LL night mel high LL night

A B A B C

(51)

nighttime levels already occurred as early as 2h before lights-on in the morning (Fig.2). Thirty minutes before lights-on, melatonin levels further decreased and reached average daytime levels (Fig. 2).

Figure 2. Melatonin plasma concentrations over time in normal LD conditions.

Onset of night is indicated by “lights off”, end of night by “lights on”. Melatonin concentrations under normal nighttime conditions decreased as early as two hours prior to lights on. Groups with different letters are significantly different (Tukey post- hoc test, p < 0.01).

Effects of melatonin treatment over time

Melatonin levels showed a significant increase one hour after treatment compared to the control group (low concentration: 995.41 ± 103.216 pg/ml;

high concentration: 6843.95 ± 4541.146 pg/ml; Dunnett’s test; P < 0.01 for both concentrations) and remained elevated for 23 hours after treatment with

time (h)

-2.5 0.0 2.5 5.0 7.5 10.0 12.5

10 100 1000 10000

C C C C CC

lights on lights off

A A A

B B B B

melatonin conc. (pg/ml)

C C

The influence of melatonin on birdsong and its underlying neuronal correlates : integrating technical advances to address melatonin's role as an endocrine switch