Thyroid hormones clearly play a role in the seasonal regulation of reproduction, but any role they might play in song behavior and the associated seasonal neuroplasticity in songbirds remains to be elucidated. To pursue this question, we first established seasonal patterns in the expression of thyroid hormone regulating genes in male European starlings employing in situ hybridization methods. Thyroid hormone transporter LAT1 expression in the song nucleus HVC was elevated during the photosensitive phase, pointing towards an active role of thyroid hormones during this window of possible neuroplasticity. In contrast, DIO3 expression was high in HVC during the photostimulated phase, limiting the possible effect of thyroid hormones to maintain song stability during the breeding season. Next, we studied the effect of hypothyroidism on song behavior and neuroplasticity using in vivo MRI. Both under natural conditions as with methimazole treatment, circulating thyroid hormone levels decreased during the photosensitive period, which coincided with the onset of neuroplasticity. This inverse relationship between thyroid hormones and neuroplasticity was further demonstrated by the negative correlation between plasma T3 and the microstructural changes in several song control nuclei and cerebellum. Furthermore, maintaining hypothyroidism during the photostimulated period inhibited the increase in testosterone, confirming the role of thyroid hormones in activating the hypothalamic–pituitary–gonadal (HPG) axis. The lack of high testosterone levels influenced the song behavior of hypothyroid starlings, while the lack of high plasma T4 during photostimulation affected the myelination of several tracts. Potentially, a global reduction of circulating thyroid hormones during the photosensitive period is necessary to lift the brake on neuroplasticity imposed by the photorefractory period, whereas local fine-tuning of thyroid hormone concentrations through LAT1 could activate underlying neuroplasticity mechanisms. Whereas an increase in circulating T4 during the photostimulated period potentially influences the myelination of several white matter tracts, which stabilizes the neuroplastic changes. Given the complexity of thyroid hormone effects, this study is a steppingstone to disentangle the influence of thyroid hormones on seasonal neuroplasticity.

1. Subjects and experimental design

Thirty male starlings (Sturnus vulgaris) were wild caught as adults in Normandy (France) in November 2014. All animals were housed in two large indoor aviaries (L x W x H: 2.2 x 1.4 x 2.1 m) at the University of Antwerp with food and water ad libitum with artificial light dark cycle. Starting from January 2013, all birds were kept in a long day photoperiod (16L/8D) in order to remain photorefractory. The housing and experimental procedures were performed in agreement with the Belgian and Flemish laws and were approved by the Committee on Animal Care and Use of the University of Antwerp, Belgium (2014-52).

Starlings were divided into two groups: a hypothyroid group (N=16) and a control group (N=14). The study was started when all birds were photorefractory. Then they were switched from long to short (8L:16D) days to induce the return to photosensitivity. Methimazole (MMI) treatment was started in one group to induce hypothyroidism. By supplementing the drinking water with 0.05% MMI, the endogenous stock of THs gradually decreased until it was fully depleted after 2-3 weeks. MMI treatment was continued for the remainder of the experiment. After 12 weeks of short days, the photosensitive starlings were switched back to long days (16L:8D) so that they became photostimulated. In parallel the control group was exposed to the same photoperiodic regime without receiving any hormone manipulation.

We monitored the neuroplasticity repeatedly at 6 different time points. The first time point was at the end of the photorefractory state (PR). After switching to short days we measured every 4 weeks to follow up the song control system plasticity during the photosensitive period (SD4, SD8, SD12). Additionally, we measured after 1 week of long days, as it is known that exposure to 1 long day can already affect TH. Finally, we measured after 4 weeks on long days (LD4), when starlings were fully photostimulated. At each time point, songs were recorded, blood samples were taken, and MRI (DTI and 3D) was acquired. In addition, body weight and beak color were registered.

2. MRI data acquisition

The birds were initially anesthetized using 2% Isoflurane (Isoflo ®, Abbot Laboratories Ltd.) in a mixture of 30% O₂ and 70% N₂ at a flow rate of 600 ml/min. Throughout the entire imaging procedure, respiration rate was monitored with a small pneumatic sensor (SA Instruments, NY, USA) positioned under the bird. Depending on the breathing rate, the anesthetic dose was lowered, ranging between 1% - 2% isoflurane. Body temperature was monitored with a cloacal temperature probe and kept within narrow physiological ranges (41.0 ± 0.2 °C) using a warm air system with a feedback unit (SA Instruments, NY, USA).

All MRI measurements were performed on a 7T horizontal MR system (Pharmascan 70/16 US, Bruker Biospin, Germany). Each imaging session started with a T2-weighted 3D anatomical RARE scan (TR: 2000 ms; TE: 11 ms; RARE factor: 8; zero-filled to a matrix of (256x92x64) with voxel resolution (0.089x0.25x0.25) mm³). Subsequently, a 4 shot SE-EPI DTI scan (TR: 7000 ms; TE: 23 ms; d 4ms, D 12ms; b-value 670 s/mm²; 60 diffusion gradient directions; spatial resolution: (0.179x0.179x0.35) mm³; 28 coronal slices) was acquired. After the imaging procedure, birds were left to recover in a warmed recovery box before returning to the aviary.

3. MRI data processing

Diffusion data were analyzed with MRtrix3 version 3.0 (Tournier et al., 2012) following the same processing steps as described in (Orije et al., 2021b). Preprocessing of the individual DW-images included following steps: denoising (Veraart et al., 2016), correction for Gibbs ringing (Kellner et al., 2016), motion and distortion correction using FSL (Jenkinson et al., 2012), bias field correction using ANTS (Advanced Normalization Tool; Avants et al. (2010)), whole brain extraction and upsampling to isotropic voxels of 1.75 mm. These preprocessed diffusion weighted images were used to calculate individual diffusion maps (fractional anisotropy (FA), mean, axial and radial diffusivity) and fiber orientation distribution (FOD) images. The neuroanatomical contrast of the individual FA maps allowed the delineation of different ROI’s (Area X and RA) to determine their volume changes using AMIRA software (De Groof et al., 2006). The calculation of FOD images requires global intensity normalization, fiber response function estimation using the unsupervised Dhollander algorithm (Dhollander et al., 2019) and spherical deconvolution (Jeurissen et al., 2014). These FOD images were normalized to create an unbiased study-based FOD template, which involves linear and non-linear registration (Raffelt et al., 2011). Fiber density (FD) and fiber-bundle cross-section (FC) were estimated from the normalized FOD images. The transformation parameters derived from building the FOD template were also applied to the diffusion maps to warp them into the template space to perform voxel-based analysis. Next, these images were smoothed to double voxel size (3.5 x 3.5 x 3.5 mm³). Finally, all normalized diffusion maps were averaged to create a FA template that is used as background to display the statistical results.

Subject 5 and subject 16 died after the second and fourth time point respectively, so only MRI data later time points could not be acquired. Scans of subject 7 at time point 4 failed due to excessive movement.