Can increased prenatal exposure to thyroid hormones alter physiology and behaviour in the long-term? Insights from an experimental study in Japanese quails

Materials

This study was conducted in accordance with the Finnish legislation and approved by the Finnish Animal Ethics Committee (Eläinkoelautakunta; ESAVI/1018/04.10.07/2016). Japanese quail (Coturnix japonica), a commonly used laboratory model with a reference genome, was used. This study was done with same individuals previously used in Sarraude et al. (2020).

Parental generation and egg collection

The parental generation was composed of adult quails that were supplied by private breeders in Finland and kept in two acclimatised rooms. Twenty-four breeding pairs were formed by pairing birds from different breeders. Individuals were identified using metal leg rings. The floor was covered with 3–5 cm of sawdust bedding, and a hiding place, sand and calcium grit were provided. Each pair was housed in an indoor aviary divided into pens with a floor area of 1 m². The ambient temperature was maintained at 20 °C, and the photoperiod was set to 16 light hours (06:00 am–22:00 pm, +3UCT) and 8 dark hours. The birds were fed ad libitum (Poultry Complete Feed, “Kanan Paras Täysrehu”, Hankkija, Finland), and their water was replaced daily. Pairs were monitored every morning to collect eggs for 7 days (longer storage time is known to decrease hatchability). The eggs were marked with a non-toxic marker, weighed, and stored in a climate-controlled chamber at 15 °C and 50% relative humidity. On the final day of the collection period, a solution was injected into a total of 4–8 eggs per pair (the variation was due to not all females producing an egg every day, average = 7, standard deviation = 1).

Treatment of eggs, rearing of chicks and collection of samples

We manipulated egg TH levels by injecting hormones directly into the eggs prior to embryonic development at a dose that mimicked increased maternal transfer (see details in Sarraude et al., 2020). Eggs from each breeding pair were randomly assigned to four treatment groups: T3, T4, T3 + T4 and control (0.9% saline), 24 clutches for all groups except T3 that had 22 clutches. We used saline control following previous studies on injecting THs into eggs (Van Herck et al., 2015), and because THs are not very well soluble in oil. Eggs laid by the same pair were always evenly distributed among all treatment groups. The injected hormone solutions were prepared from crystallised powder (T3: CAS number 6893-02-3, Sigma-Aldrich, St. Louis, MO, USA and T4: CAS number 51-48-9, Sigma-Aldrich, St. Louis, MO, USA). The amount of hormone injected (T4: 8.9 ng/egg, equivalent to 1.79 pg/mg yolk; T3: 4.7 ng/egg, equivalent to 1.24 pg/mg yolk) was based on previous measurements in the parental generation (mean ± SD: T4, 15.3 ± 4.43 ng/egg; T3, 7.58 ± 2.32 ng/egg) and the aim was to increase hormone contents in the yolk by two standard deviations (as recommended by Podmokla, Drobniak & Rutkowska, 2018). We were unsure how much of T4 is converted to T3 and therefore used 2SD of both hormones for the T3T4 group. The injection protocol has been described in detail in Sarraude et al. (2020). Briefly, each egg was injected with 50 µl of either the respective hormone solution (T3, T4 or T3+T4, in 0.9% saline) or 0.9% saline (control, CO), delivered by using 27G insulin syringes (BD Microlance™). After sealing the hole with sterile plaster (OPSITE Flexigrid, Smith & Nephew), the eggs were incubated artificially (37.8 °C and 55% relative humidity). A total of 158 eggs were treated (39 T3, 39 T4, 40 T3+T4 and 40 CO), of which 66 chicks hatched (15 T3, 20 T4, 21 T3 + T4 and 10 CO), with sex distributions of chicks surviving to adulthood (females/males): 7/4 for T3, 10/8 for T4, 9/12 for T3 + T4 and 3/4 for CO. Breeding pairs/pairs of siblings per treatment: CO 7/0, T3 9/2, T3T4 15/4 and T4 13/2. Sex was determined at maturity by examining the presence of the cloacal gland. The overall hatching rate (∼40%) was consistent with previous studies in domestic quail (Okuliarová, Škrobánek & Zeman, 2007).

Twelve hours post-hatching, the chicks were marked with a unique combination of coloured rings and nail coding, and transferred to two cages (each with a floor area of 1 m² and a height of 30 cm). The cages contained approximately 30 chicks, sex and treatments mixed. The chicks were provided with heating mats and lamps as supplementary heat sources for the first two weeks. The chicks were fed with sieved commercial poultry feed (“Punaheltta paras poikanen”, Hankkija, Finland) and provided with calcium and bathing sand. Two weeks after hatching, the chicks were separated into four 1 m² cages (ca. 30 cm high) of approximately 16 individuals. Approximately three weeks after hatching, the coloured rings were replaced with unique metal rings. At week 4 after hatching, the birds were transferred to eight pens of 1 m² floor area (average of 7.1 birds/pen, range = 4–9), under the same conditions as the parents. At around the age of sexual maturity (ca. 6–8 weeks after hatching), the birds were separated by sex in twelve 1 m² pens (average of 4.8 birds/pen, range = 4–5). Photoperiod was 16L:8D for the first 7 months and 8L:16D thereafter. At approximately 4.5 months of age, we collected blood samples (200 µl) from the brachial vein, separated plasma, and stored it at −80 °C. As Japanese quails have reached adulthood at 4.5 months, this time point would have already shown if there are long-lasting, organizational effects of prenatal THs effects into adulthood. All birds were euthanised by decapitation at 10 months of age. After decapitation, brain tissue was collected for the gene expression analysis from the pallium (average 44.0 mg, std 2.8 mg, taken at the surface of the region) in PBS buffer, snap frozen in liquid nitrogen and stored at −80 °C. While differences in photoperiod limit comparisons between gene expression results and other results, our aim was to study whether there are persistent effects of prenatal THs independent of any environmental factors that individuals may face later in life, for which our experimental design is sufficient.

Behavioural experiments

To assess the potential effects of maternal THs on behaviour, we conducted a series of behavioural tests (open arena, emergence and tonic immobility tests) at different life stages. We followed previously published protocols that have been used extensively with many different model organisms. The tests were conducted in separate rooms where the birds had no visual or auditory contact with other individuals. The exception was the open arena test, where the subject was able to hear the vocalisations of another bird being tested at the same time in another room. All experiments were recorded, and the videos were analysed blindly of the treatment. All videos of the same test at both ages (chick/adult) were analysed by one person, except for the open arena test, where one person analysed chick videos and another person analysed adult videos. This may have confounded the age difference in behavior scores with the observer effect. Since the videos from different treatments at each age were analysed by the same person, the fact that chick and adult videos were analyzed by different people should not confound the effect of prenatal TH exposure.

The open arena test measured social motivation/stress and boldnes/fearfulness and it followed a setup developed by Forkman et al. (2007). The test was performed once on both sexes at three days (F/M/undetermined: CO 3/3/1, T3 6/3/2, T3T4 8/12, T4 8/8) and once approximately 4.5 months of age (F/M: CO 3/4, T3 8/3, T3T4 9/12, T4 9/8). In the test, the bird was placed in a circular illuminated arena and the movement duration and the number of escape attempts (attempt to jump/fly over the wall) were measured from videos for 10 min (from the time the researcher left the room). Only 10 of the 56 adults (one CO, two T3, two T4 and five T3T4) attempted to escape, so adult escape attempts were not analysed. Subjects who moved less and made fewer escape attempts were interpreted as less bold and less socially motivated.

The emergence test measured social motivation and boldness/fearfulness (especially the timidity aspect, see e.g., Davis et al., 2008). The test was performed on both sexes, twice on 6–9 d old chicks (as there is less data on repeatability, Forkman et al., 2007; F/M/u ndetermined: CO 3/4/1, T3 7/4/1 T3T4 9/12, T4 10/8) and once on 4.5 months old adults (F/M: CO 3/4, T3 7/4, T3T4 8/11, T4 9/7). The test followed a setup developed by Forkman et al. (2007), in which a bird was placed in a dark chamber connected to another, bright chamber. One minute after the subject was placed in the dark chamber, a door connecting the chambers was removed. After removing the door, the researcher left the observation room and the time it took for the subject to completely exit the dark chamber was measured. Subjects who left the dark chamber more quickly were interpreted as bolder and more socially motivated. The test lasted five minutes. Whether the subject remained in the dark chamber (0) or left it (1) during the test was also recorded.

The tonic immobility test (Mills & Faure, 1991) measured fearfulness to predators (Forkman et al., 2007). The test was performed on once both sexes at 12 days (F/M/undetermined: CO 3/4, T3 7/4/1, T3T4 9/12, T4 10/8) and once approximately 4.5 months of age (F/M: CO 3/4, T3 7/4, T3T4 8/12, T4 9/7). To induce immobility, the bird was placed on its back in the centre of a V-shaped test apparatus and restrained by hand for 15 s, after which the hand was removed, and the time taken for the bird to resume a standing posture was measured. If the subject resumed the standing posture in less than 10 s, this was recorded as a failed induction and the induction procedure was repeated. Four adult subjects who failed five times were excluded. If the subject remained stationary for more than 10 s, then the time to recovery was measured. The longer a subject took to recover, the more fearful it was interpreted. The test lasted 10 min. Five subjects (one chick and four adults) did not recover for 10 min and were excluded from the analysis (approximately 4% of all tests).

Plasma thyroid hormone levels

To test whether elevated maternal THs affect plasma TH levels later in life (4.5 months of age), we extracted THs from plasma samples and measured them using a previously validated LC-MS/MS protocol (Ruuskanen et al., 2018), where CV <10%). Due to practical limitations (i.e., volume of plasma available and haemolysis in some samples), the sample size was reduced to 48 individuals (F/M: T3 6/4, T4 8/8, T3T4 6/9 and 3/4 in the control). Hormone concentrations are expressed in pmol/ml plasma. The repeatability of the TH measurements, was validated in Ruuskanen et al. (2018) with CV% less than 10% (3.5−9.7%, depending on the hormone and sample type).

Gene expression

To test for potential long-term changes in the expression of TH related genes (DIO2, DIO3, THRA, and NCOA1), we measured gene expression from pallium samples (10 months old). RNA was extracted from the homogenised pallium samples using the NucleoSpin RNA Plus kit (MACHEREY-NAGEL). Since the sample size of the control group was the main factor limiting statistical power, we decided to analyse all control individuals and only an equivalent number of individuals (randomly selected) in the three treatment groups (N = 31, 3F/4M in control and 4F/4M per other treatment groups). Extraction was done from 50 µl of the homogenised brain sample according to the manufacturer’s instructions, except that the gDNA removal column was centrifuged longer (1 min, 11,000 g) and the RNA was eluted once with 40 µl of Rnase-free water. RNA concentration and purity were quantified by optical density, and all samples not meeting our quality criteria (i.e., RNA concentration >30 ng/µl and 260/280 >1.80) were excluded from further analysis. RNA integrity was checked using an E-Gel 2% electrophoresis system (Invitrogen), and the ribosomal RNA 18S vs. 28S band intensity and was deemed satisfactory (the visual inspection revealed the expected two bands of rRNA with 28S band being >2 times more intense than 18S band). Samples were stored at −80 °C for 2 weeks prior to cDNA synthesis. cDNA synthesis was performed using the SensiFAST ™ cDNA Synthesis Kit (Bioline) according to the manufacturer’s instructions. The reaction was performed using 500 ng of RNA as the starting material. A control (noRT) was prepared without reverse transcriptase enzyme. The cDNA was diluted with water to a concentration of 1.2 ng/µl and stored at −20 °C. The 1.2 ng/uL was calculated from RNA pre-RT, and slight variations in cDNA starting quantity (due to differences in RT efficiency) are controlled for using the reference genes when calculating relative expression.

GAPDH and PGK were used as control genes, using primers previously designed and validated in Japanese quail (Vitorino Carvalho et al., 2019). Stability of expression between experimental groups was further verified using geNorm software (geNorm M < 0.5, geNorm V < 0.15), and thus the geometric mean of these two genes was used as our normalization factor. Primers for the target genes were designed using the Primer-BLAST tool (NCBI). The primer pairs (Table 1) were designed to have at least one intron between them. The suitability of the primers in qPCR was tested prior to the actual analysis using different primer concentrations and different amounts of cDNA. The tests confirmed the sufficient amplification efficiency of the primers (between 90 and 110% based on the standard curve) and the formation of only one product of the expected size based on both the presence of a single narrow peak in the melting curve, and the presence of a single band of the expected size on the agarose gel. DIO3 had extremely poor amplification, meaning that we could not accurately quantify it, and was excluded from the analysis.

qPCR assays were performed on a Mic qPCR (Bio Molecular Systems) in a final volume of 12 µL containing 6 ng of cDNA, 800 nM of forward and reverse primers and SensiFAST SYBR Lo-ROX (Bioline). We ran gene-specific qPCR plates and each sample was analysed in duplicate. Two plates were run for each gene and experimental groups and sex were always balanced within each plate. We used a cDNA reference sample (i.e., ratio = 1), which was a pool of 10 different individuals on each plate. We used a two-step cycling with the following conditions: 2 min at 95 °C; then 45 cycles of 5s at 95 °C followed by 20s at 60 °C (fluorescence reading) for all reactions, as recommended by manufacturer. The expression of each gene was calculated as (1+Ef_Target)^ΔCq(Target)/geometric mean ((1+Ef_GAPDH)^ΔCqGAPDH; (1+Ef_PGK1)^ΔCqPGK1), where Ef is the amplification’s efficiency and ΔCq is the difference between the Cq-values of the reference sample and the sample of interest. Amplification in the no-RT controls never occurred until at least eight cycles after the lower Cq sample, and thus contamination by genomic DNA could not interfere with our results. Technical precision and repeatability were generally good (Table 1), but a few samples with a CV% of duplicates (based on the final ratio) greater than 15% were excluded from the analysis (final sample size: N_DIO2 = 27, N_NCOA1 = 28, N_THRa = 26).

Statistical analysis

All statistical analyses were performed with R software (version 3.6.2; R Core Team, 2019), using the packages lme4 version 1.1-23 (Bates et al., 2015) and emmeans version 1.6.0 (Lenth, 2021). Normality and homoscedasticity of the residual distributions were checked visually (quantile–quantile and residuals against fitted values plots)with additionally confirmation by Shapiro–Wilk test of normality. For the behavioural analysis, recovery time (tonic immobility) and exit time (emergence test), were natural log-transformed to ensure that the residuals were normally distributed. Possible outliers were tested with R using the simulateResiduals function in DHARMa package (Hartig, 2024).

Linear mixed models (LMM, lmer function in lme4) were used to analyse time to exit in the emergence test, movement duration and number of escape attempts in the open arena test, recovery time in the tonic immobility test, plasma TH concentrations and qPCR results. Generalised linear mixed model (GLMM, glmer function in lme4) with a binomial error distribution was used to analyse the proportion of individuals leaving the dark chamber in the emergence test. Fixed effects were treatment, sex, and test number/age (in repeated tests). Interactions between sex and treatment, and between treatment and age (in repeated tests), were also analysed. Random effects were individual ID (in repeated tests) and parental ID (to account for relatedness between siblings in all experiments). F- and p-values for LMMs were obtained using the Kenward-Roger approximation on the denominator degree of freedom (ddf), implemented in the package lmerTest (anova function) (Kuznetsova, Brockhoff & Christensen, 2017). For binomial regression models, main effects and interactions were tested by comparing the model with and without the factor. For all models, we removed non-significant (p > 0.05) interactions to ensure correct interpretation on the main effects. Post-hoc tests were performed using Tukey tests (emmeans function from emmeans package; Lenth, 2021) where appropriate.

Effects of prenatal thyroid hormone supplementation on behaviour

In general, no statistically significant differences in chick or adult behavioural traits were observed between the egg TH manipulation groups (T3, T4, T3 + T4, control). Elevated THs had no significant effect on social motivation or boldness (Table 2, Figs. 1–2), nor on fearfulness to predators (Table 2, Fig. 3). A statistically significant interaction between treatment and age was found in the emergence test (F = 3.58, ddf = 67.2, p = 0.018, Fig. 2): In all treatments except the control group (Tukey t = 0.59, p = 0.56), chick exit times were significantly longer than those of adults (all t > 3.21, p < 0.002) while no differences between treatments were found within each age-category (all t < 2.48, all p > 0.07). There was also a statistically significant interaction between treatment and sex in the tonic immobility test (F = 2.92, ddf = 37.5, p = 0.047, Fig. 3): In males, the recovery time was longer in the T4-group (marginal mean ± SE: 177.1 ± 47.8 s) than in the T3-group (marginal mean ± SE: 56.1 ± 20.8 s, Tukey t = 2.71, p = 0.047), whereas no differences between treatments were observed in females (all t < 2.29, all p > 0.12). In the T4-group, the recovery time of males (marginal mean ± SE: 177.1 ± 47.8 s) was longer than that of females (marginal mean ± SE: 85.2 ± 19.9 s) (t = 2.16, p = 0.037), while no sex differences were observed within the other treatments (all t < 1.80, all p > 0.079).

Effects of prenatal thyroid hormone supplementation on plasma thyroid hormone levels

Increasing TH levels in eggs did not significantly affect plasma TH levels in 4.5-month-old birds (T3: F = 0.13, ddf = 37.7, p = 0.94, T4: F = 2.1, ddf = 41.5, p = 0.12, Fig. 4). Treatment also had no sex-dependent effect on plasma T3 or T4 (treatment-sex interactions, p > 0.36).

Effects of prenatal thyroid hormone supplementation on pallial gene expression

Increasing TH levels in eggs did not affect pallial gene expression of TH-related genes in 10-month-old birds (DIO2: F = 2.49, ddf = 17.3, p = 0.09, NCOA1: F = 0.092, ddf = 21.1, p = 0.96, THRA: F = 0.47, ddf = 18.9, p = 0.71, Fig. 5). Treatment also did not have sex-dependent effect on gene expression (treatment-sex interactions, p > 0.26). Females had higher expression of NCOA1 (F = 7.68, p = 0.012) and THRA (F = 5.17, p = 0.034) than males.