Animals

Adult male and female zebra finches (Taeniopygia guttata) from our breeding colony in Seewiesen, Germany, were used for all experiments. Animals were reared socially in mixed-sex colonies. For the experiments, birds were kept in pairs in a 14/10 (1200 lx/<0.0001 lx) light/dark (LD) cycle in sound- and light-proof boxes (cage size: 120 × 50 × 50 cm) each equipped with microphone, antenna and loudspeaker (24uC). The male birds and their female company birds were given at least a three-day adjustment period before surgery to get accustomed to the new housing situation. All housing conditions were in accordance with EU regulations and experiments including usage of transmitters and surgical 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).

Surgery

Detailed procedure of the surgery was described earlier (Ter Maat et al. 2014). In short, male birds were anesthetized using isoflurane inhalation (1.0– 2.0% at 0.5 L O2/min) and head-fixed in a custom-made stereotactic device. A reference electrode (platinum wire 0,025 mm, Goodfellow) was fixed close to the bifurcation between lower cranium layer and dura mater and a 2 MΩ tungsten recording electrode (FHC, Bowdoin, USA) was lowered towards the target region. During placement of the electrode in RA, electrical activity was amplified using a DAM 80 amplifier (WPI, AC Differential Amplifier) and monitored online using a continuous update of the ISI (inter spike intervals) of Schmitt triggered spikes, a reliable method to identify RA projection neurons (Hahnloser et al. 2006). After a successful surgery the transmitter was placed on the implanted connectors and the bird was allowed to recover in the cage together with the female company animal. Recordings of neuronal and audio signals were started immediately and maintenance of neuronal unit isolation was monitored throughout the whole experiment by audiovisual inspection.

To rule out a general effect of melatonin on neuronal activity in the whole song control-auditory system, we furthermore recorded from single neurons in Field L of two adult male zebra finches, again using the same experimental procedure described for RA. Field L, part of the auditory pathway providing auditory input to the song control areas, is sensitive to BOS playback but does not express melatonin receptors (Fusani and Gahr 2015). Since the location of the recording electrode in Field L could not be verified by unique neuronal activity patterns as found in RA, birds were sacrificed after the experiment and the brain was removed for histological examination of recording sites. The correct placement of the recording electrode in Field L was verified in Nissl sections.

Neuronal recordings

Neuronal activity was transmitted wirelessly with a lightweight (~ 1.0 g) telemetry device (Ter Maat et al. 2014; Schregardus et al. 2006) and recorded via a regular telescopic whip antenna (Nagoya Antenna, Taiwan) connected to an AOR5000 communication receiver (AOR, Ltd., Japan) with the upper audio frequency set at 12 kHz (–3dB). The signal was decoded as FM with intermediate frequency bandwidth set at 110 kHz. Signals were fed into an 8-channel audio A/D converter (M-Audio 1010; 22050 Hz) and recorded using custom written software.

Recordings of vocal and locomotor activity

Male and female vocalizations, movements (perch hopping, flying) and playback of BOS as well as other ambient sounds were recorded via in-box microphones simultaneously and synchronized with the neuronal recordings. Video recordings were used to confirm behavioral cues from audio recordings.

Bird’s own song (BOS) playback

BOS of the male birds was recorded during the habituation phase of the experiments before surgery. Recordings of BOS were presented randomly (custom written software) at 30–180 s intervals during playback sessions via built-in loudspeakers in recording boxes. The repeated playback sessions covered several hours of all phases of the light/dark cycle as well as the transitions from day to night (lights off in the evening), spontaneous lights off events during the day, and periods surrounding melatonin treatment in a constant light situation.

Melatonin manipulation

For manipulating the natural melatonin production of our animals, we used a protocol developed in our group and published previously (Seltmann et al. 2016). In short, to prevent naturally occurring melatonin production, animals were kept in constant light (LL) starting the day before surgery. For simulating nocturnal melatonin levels and their circadian fluctuation, animals were treated with a melatonin containing cream (approximately 0.65 µg melatonin per treatment) applied to the skin between the dorsal feather ridge, wing and neck of the birds. This treatment has been shown to successfully and reliably simulate the natural increase of melatonin levels in the transition from day to night as well as their natural peaks during the night while being minimally invasive (Seltmann et al. 2016).

Experimental design

Neuronal and auditory recordings were started as soon as the bird was released back into the cage to recover in presence of its female company bird (around noon). After three to five hours (depending on the recovery process) the first playback session was started. The male bird was subsequently treated with melatonin to imitate the development of natural melatonin levels in LL conditions while being presented with BOS playback. After one night in LL, the light regime returned to the birds’ normal 14/10 L/D schedule. A second playback session covered this period. A third playback session was held on the second day after surgery, covering either a second melatonin treatment or a spontaneous dark period during circadian daytime. To compare the data collected in RA with those observed in a song control area not expressing melatonin receptors, the same experimental setup was used for additional recordings in Field L.

RNAScope® in situ hybridization assay

RNAScope® fluorescent multiplex assay (Advanced Cell Diagnostics) was used with three birds (three 20-µm para-sagittal sections including HVC for each bird) to visualize mRNA expression of melatonin receptor 1B (MNTR1B), glutamate decarboxylase (GAD), and the vesicular glutamate transporter SLC17AG6 following the protocols provided by the manufacturer. Probes were conjugated to Atto 550 (MNTR1B), Alexa 488 (GAD), and Atto 647 (SLC17A6) fluorophores, respectively. Sections were counterstained with DAPI. The stained tissue sections were imaged using a Leica DM6000 B microscope. Quantification was performed using ImageJ2 (Fiji Distribution (Rueden et al. 2017; Schindelin et al. 2012). We counted the number of neurons labeled only with one probe, of double-labeled or triple-labeled neurons in RA sections.

Statistical analysis

Neuronal activity was processed using peak detection (Jansen and Ter Maat 1992). Spikes were then classified by waveform and sorted into clusters using a K-means sorting algorithm, an unsupervised algorithm that partitions spike waveforms into k groups by assigning each recorded waveform to the nearest centroid based on Euclidean distance and updating centroids as the mean of their assigned waveforms. Each resulting cluster is interpreted as the activity of a putative single neuron which then allows for further analysis of single unit activity. Further analysis was performed using custom software. The resulting single unit isolation was additionally confirmed by visual inspection of overlaid waveforms as well as ISI distribution for the RA recordings (see below). To compare the direct effect of BOS playback on neuronal activity under different conditions, the mean as well as the maximum firing rate as response to BOS playback in different conditions was calculated. For doing an inter individual comparison an index was calculated. Mean as well as maximum firing rate directly following the onset of a BOS playback were subtracted from that during an equally long period before the playback. To correct for differences in overall firing rate, this difference was divided by the sum of mean/maximum firing rate before and after playback onset, covering equally long periods. To compare the different situations, a restricted maximum likelihood (REML) procedure with treatment as factor and animal ID as random factor, followed by Tukey LSD with P = 0.01 was used.

For comparison of regular baseline firing under different conditions, the modal inter spike interval (ISI) was calculated per situation for every individual using recording sequences of at least 15 min up to several hours, depending on the duration of the situation and the recording quality. Stretches of failed or bad quality signal transmission as well as with direct disturbances (animal care, setup checks etc.) were excluded. We furthermore excluded stretches including singing activity. The modal inter-spike intervals were estimated from the interval histogram after removing the high-frequency activity (0-10ms). The bin with the maximum count was first determined and then the counts in the 10 bins before as well as the 10 bins after the maximum bin were used to estimate the peak of a 4th order polynomial. This peak is considered the modal ISI.

The effect of the experimental treatments on the modal inter-spike interval was tested with a REML procedure with treatment as factor and animal ID as random factor, followed by Tukey LSD with P = 0.01.

To compare the response of Field L neurons to BOS playback, we used probability density function and Kolmogorov-Smirnov comparison. For Field L, Z-scores were calculated based on the standard deviation of 4 to 3 s before stimulus onset. Calculations were done using custom-written software as described elsewhere (Ter Maat et al. 2014; Gill et al. 2015).