Elevated exposure to prenatal thyroid hormones affects embryonic mortality but has no effects into adulthood

Maternal thyroid hormones (THs) are known to be crucial in embryonic development in humans, but their influence on other, especially wild, animals remains poorly understood. So far, the studies that experimentally investigated the consequences of maternal THs focused on short-term effects, while long-term organisational effects, as shown for other prenatal hormones, could also be expected. In this study, we aimed at investigating both the short- and long-term effects of prenatal THs in a bird species, the Japanese quail Coturnix japonica. We experimentally elevated yolk TH content (the prohormone T4, and its active metabolite T3, as well as a combination of both hormone) We analysed hatching success, embryonic development, offspring growth and oxidative stress as well as their potential organisational effects on reproduction, moult, and oxidative stress in adulthood. We found that eggs injected with both hormones had a higher hatching success compared with control eggs, suggesting conversion of T4 into T3 by the embryo. We detected no other short-term or organisational effects of yolk THs. Unfortunately, sex-specific responses could not be properly tested due to low sample sizes. These results suggest that yolk thyroid hormones are important in the embryonic stage of precocial birds, but may not have other short and long-term consequences, at least not in captivity. Research on maternal thyroid hormones will greatly benefit from studies investigating how embryos use and respond to this maternal signalling. Long-term studies on prenatal THs in other taxa in the wild are needed for a better understanding of this hormone-mediated maternal pathway. Summary statement Thyroid hormones are important hormones in all vertebrates, although overlooked in the context of maternal effect. We found short-term effects of prenatal THs but no evidence of programming effects.


Introduction 39
Maternal effects represent all the non-genetic influences of a mother on her offspring and 40 have received increasing attention in evolutionary and behavioural ecology. Through maternal 41 effects, mothers can influence the fitness of their progeny by adapting their phenotype to 42 expected environmental conditions ("adaptive maternal effects" in Marshall and Uller, 2007;43 Mousseau and Fox, 1998), and this view is now also incorporated in the human disease 44 literature (Gluckman et al., 2005). Maternal hormones transferred to the offspring can mediate 45 important maternal effects. Historically, research on maternal hormones has mostly focused 46 on steroid hormones (Groothuis et al., 2005;von Engelhardt and Groothuis, 2011). Recently, 47

Preparation of the solution, injection procedure and incubation 136
The preparation of hormone solution and the procedure of injection were based on previous 137 studies (Hsu et al., 2017;Ruuskanen et al., 2016). In brief, crystal T 4 (L-thyroxine, ≥ 98% 138 HPCL, CAS number 51-48-9, Sigma-Aldrich) and T 3 (3,3',5-triiodo-L-thyronine, > 95% 139 HPCL, CAS number 6893-02-3, Sigma-Aldrich) were first dissolved in 0.1M NaOH and then 140 diluted in 0.9% NaCl. The injection of thyroid hormones resulted in an increase of two 141 standard deviations (T 4 = 8.9 ng/egg; T 3 = 4.7 ng/egg), a recommended procedure for 142 hormone manipulation within the natural range (Hsu et al., 2017;Podmokła et al., 2018;143 Ruuskanen et al., 2016). The control solution (CO) was a saline solution (0.9% NaCl). The Hormone injections were performed at room temperature in a laminar hood. Eggs 148 were put sideways, allowing yolks to float up to the middle position. Before injection, the 149 shell was disinfected with a cotton pad dipped in 70% EtOH. We used a 27G needle (BD 150 Microlance ™) to pierce the eggshell and then used a 0.3 ml syringe to deliver 50 µl of the 151 respective hormone solution or control. After injection, the hole was sealed with a sterile 152 plaster (OPSITE Flexigrid, Smith&Nephew). 153 In total, 158 eggs were injected and divided as follows over the treatments: T 3 154 treatment (N = 39); T 4 treatment (N = 39); T 3 +T 4 treatment (N = 40); and control, CO (N = 155 40). To balance the genetic background of the parents and the effect of storage, each egg laid 156 by the same female was sequentially assigned to a different treatment and the order of 157 treatments was rotated among females. After injection, eggs were placed in an incubator at 158 37.8°C and 55% relative humidity. Until day 14 after starting incubation, eggs were 159 automatically tilted every hour by 90°. On day 14, tilting was halted and each egg was 160 transferred to an individual container to monitor which chick hatched from which egg. On day 161 16 after injection, (normal incubation time = 17 days), the temperature was set to 37.5°C and 162 the relative humidity to 70%. Eggs were checked for hatching every 4 hours from day 16 163 onwards. Four days after the first egg hatched, all unhatched eggs were stored in a freezer and 164 dissected to determine the presence of an embryo. The age of developed embryos was 165 assessed according to Ainsworth et al. (2010). 166

Rearing conditions of the experimental birds 167
In total, 66 chicks hatched (N = 10 CO, 15 T 3 , 20 T 4 and 21 T 3 T 4 ). The overall hatching success was rather low (ca. 40%), partly due to the injection procedure itself (Groothuis and  169 von Engelhardt, 2005), although low hatching success in quails has also been reported in 170 unmanipulated conditions previously (e.g. Okuliarová et al., 2007). Among the unhatched 171 eggs, 33.7% (31 out of 92) had no developed embryos. Twelve hours after hatching, the 172 chicks were marked by a unique combination of coloured rings and nail coding and 173 transferred to two 1 m² cages (ca. 30 chicks/cage, sex and treatments mixed together). The 174 chicks were provided with heating mats and lamps as extra heat sources for the first two 175 weeks. The chicks were fed with sieved commercial poultry feed ("Punaheltta paras 176 poikanen", Hankkija, Finland), and provided with Calcium and bathing sand. Two weeks after 177 hatching, the chicks were separated in four 1 m² cages of about 16 individuals. Around 3 178 weeks after hatching, coloured rings were replaced by unique metal rings. On week 4 after 179 hatching, birds were transferred to eight 1 m² pens (average of 7.1 birds/pen, range = 4-9), 180 under the same conditions as the parents. Around the age of sexual maturity (ca. 6-8 weeks 181 after hatching), the birds were separated by sex in twelve 1 m² pens (average of 4.8 birds/pen, 182 range = 4-5). 183

Monitoring of growth and reproductive maturation 184
Body mass and wing length were measured twelve hours after hatching. Tarsus was not 185 measured because it bends easily, resulting in inaccurate measures and potential harm for the 186 young. From day 3 to day 15, these three traits were monitored every 3 days. From day 15 to 187 day 78 (ca. 12 weeks), chicks were measured once a week. Body mass was recorded using a 188 digital balance to the nearest 0.1 g. Wing and tarsus lengths were respectively measured with 189 a ruler and a calliper to the nearest 0.5 mm and 0.1 mm. From week 6 to week 10, we 190 monitored cloacal gland development and foam production in 28 males. Cloacal glands were 191 measured every other day with a calliper to the nearest 0.1 mm as a proxy for testes 192 development and sexual maturation (Biswas et al., 2007). Foam production (by gently 193 squeezing the cloacal gland) was assessed at the same time and coded from 0 (no foam) to 3 194 (high production of foam), as a proxy of cloacal gland function (Cheng et al., 1989a;Cheng et 195 al., 1989b). The same observer performed all measurements. We collected eggs produced by 196 10-week-old females over a 6-day period, and measured the short and long axes of the eggs 197 with a calliper to the nearest 0.01 mm and record their mass to the nearest 0.1 g. We collected 198 on average 5.7 eggs (range = 4-7) per female from 28 females.

Monitoring of cloacal gland regression and moult 200
In Japanese quails, exposure to short photoperiod and cold temperature triggers reproductive 201 inhibition and postnuptial moulting (Tsuyoshi and Wada, 1992). Thyroid hormones are known 202 to coordinate these two responses (see introduction). When the birds reached the age of ca. 7 203 months, we exposed birds to short photoperiod (8L:16D, i.e., light from 08.00 to 16.00) with a 204 12:12-h cycle of normal (20°C) and low (9°C) temperature (low temperature was effective 205 from 18.00 to 06.00). Cloacal gland regression (as a proxy for testes regression) was 206 monitored every other day for 2 weeks with a calliper by measuring the width and length to 207 obtain the area of the gland to the nearest 0.1 mm² (N = 26 males). Primary moult was 208 recorded from a single wing by giving a score to each primary from 0 (old feather) to 5 (new 209 fully-grown feather) following Ginn and Melville (1983) (N = 54 males and females). The 210 total score of moulting was obtained by adding the score of all feathers. 211

Oxidative status biomarker analyses 212
Two blood samples were drawn, when birds were 2 weeks (N = 51 chicks) and 4 months old 213 (N = 49 adults), respectively. 200 µl of blood was collected from the brachial vein in 214 heparinized capillaries and directly frozen in liquid nitrogen. Then, the samples were stored at 215 -80°C until analyses. We measured various biomarkers of antioxidant status; the antioxidant 216 glutathione (tGSH), the ratio of reduced and oxidised glutathione (GSH:GSSG) and activity 217 of the antioxidant enzymes glutathione peroxidase (GPx), catalase (CAT) and superoxide 218 dismutase (SOD) from the blood. It is important to measure multiple biomarkers of oxidative 219 and antioxidant status for a broader understanding of the mechanism. Also, the interpretation 220 of the results is more reliable if multiple markers show similar patterns. Of the measured 221 biomarkers, the ratio of GSH:GSSG represents the overall oxidative state of cells and 222 consequently, deviations in this ratio is often used as an indicator of oxidative stress 223 (Halliwell and Gutteridge, 2015;Hoffman, 2002;Isaksson et al., 2005;Lilley et al., 2013;224 Rainio et al., 2013). GPx enzymes catalyse the glutathione cycle, whereas CAT and SOD 225 directly regulate the level of reactive oxygen species (ROS) (Ercal et al., 2001;Halliwell and 226 Gutteridge, 2015). The methodology for measuring each biomarker is described in detail in 227 Rainio et al. (2015). All analyses were conducted blindly of the treatment. Briefly, the 228 samples were analyzed using a microplate reader (EnVision, PerkinElmer-Wallac, Finland). 229 All antioxidant and enzyme activities were measured in triplicate (intra-assay coefficient of 230 variability [CV] < 10% in all cases) using 96-(CAT) or 384-well (GPx, SOD, tGSH and 231 GSH:GSSG) microplates. Three control samples were used with each plate, to be able to 232 correct inter-assay precision with the ratio specific to the particular plate. Overall protein 233 concentration (mg/ml) was measured according to the Bradford method (Bradford, 1976) 234 using BioRad stock (BioRad, Finland) diluted with dH 2 O (1:5) and BSA (bovine serum 235 albumin, 1 mg/ml) (Sigma Chemicals, USA) as a standard. GPx-assay was conducted using 236 Sigma CGP1 kit, CAT-assay using SigmaCAT100 kit and SOD-assay using Fluka 19160 SOD 237 determination kit. Total GSH and the ratio of GSH:GSSG were measured with the ThioStar® 238 glutathione detection reagent (Arbor Assays, USA) according to kit instructions, using 239 reduced glutathione as a standard (Sigma Chemicals, USA). 240

Ethics 241
The study complied with Finnish regulation and was approved by the Finnish Animal 242 Experiment Board (ESAVI/1018/04.10.07/2016). 243

Statistical analysis 244
Data were analysed with the software R version 3.5.3 (R core team, 2019). In this study, two 245 different statistical approaches were used: null-hypothesis testing with Generalised Linear 246 Mixed Models (GLMMs) and Linear Mixed Models (LMMs), and multimodel inference with 247 Generalised Additive Mixed Models (GAMMs). GAMMs were used to analyse the data on 248 body and cloacal gland growth to account for its non-linear pattern (see Growth). In this 249 analysis, we preferred multimodel inference as GAMMs generate many candidate models that 250 cannot be directly compared (e.g., by the Kenward-Roger approach). Instead, candidate 251 models were ranked based on their Akaike Information Criterion (AIC) values. Models with a 252 Δ AIC ≤ 2 from the top-ranked model were retained in the set of best models. Akaike weights 253 of all models were calculated following (Burnham and Anderson, 2002), and evidence ratios 254 of the top-ranked models were calculated as the weight of a model divided by the weight of 255 the null model (Burnham et al., 2011). To estimate the effect of the predictors, we computed 256 the 95% confidence intervals from the best models using the nlme package (Pinheiro et al., 257 2018). GLMMs and LMMs were fitted using the R package lme4 (Bates et al., 2015), and 258 GAMMs were fitted using the package mgcv (Wood, 2017). P-values for GLMMs were 259 obtained by parametric bootstrapping with 1,000 simulations and p-values for LMMs were 260 calculated by model comparison using Kenward-Roger approximation, using the package 261 pbkrtest in both cases (Halekoh and Højsgaard, 2014). Post-hoc Tukey analyses were 262 conducted with the package multcomp (Hothorn et al., 2008). Due to our experiment design, 263 eggs injected with both hormones received a higher absolute amount of hormones than eggs 264 injected with T 4 or T 3 only. Therefore, we also tested a potential dose-dependent effect of the 265 treatment on the response variables when treatment groups showed significant differences. 266 Model residuals were checked visually for normality and homoscedasticity. Covariates and 267 interactions were removed when non-significant (α = 0.05). 268

Hatching success 269
To analyse hatching success, each egg was given a binary score: 0 for unhatched egg and 1 for 270 hatched egg. A series of GLMMs were fitted with a binomial error distribution (logit link) and 271 mother identity as a random intercept. The first model included the 4-level treatment 272 (treatments: CO, T 3 , T 4 and T 3 T 4 ) as the predictor, while in a second model treatment was 273 converted into an ordered variable, following the increasing levels (i.e. CO, T 3 , T 4 and T 3 T 4 ). 274 The second model was meant to test for a potential dose-dependent effect as the eggs received 275 an increasing amount of total THs as a potential source of T 3 , the most potent hormone. Egg 276 mass might affect hatchability and was therefore added as a covariate in both models. The 277 potential effect of storage duration on hatchability (Reis et al., 1997) was accounted for by 278 including laying order as a covariate in both models. 279

Duration of embryonic period, age at embryonic mortality and mass at hatching 280
Duration of embryonic period and mass at hatching were modelled with LMMs. Treatment, 281 sex of the individuals and egg mass were included as fixed factors. Laying order was added as 282 a covariate to account for potential effects of storage duration on hatching time and on chick 283 weight (Reis et al., 1997). Mother identity was included as a random intercept. In the model 284 for mass at hatching, duration of embryonic period was further added as a covariate. 285 The data for embryonic age had a skewed distribution and residuals were not normally 286 distributed and heterogenous, which violated LMM assumptions on residual distribution. We 287 therefore performed a simple Kruskal-Wallis test. 288 Growth 289 As growth curves typically reach an asymptote, we fitted non-linear GAMMs to these curves. 290 Growth in body mass, tarsus and wing length were analysed in separate GAMMs. Growth 291 was analysed until week 10 after hatching as all birds appeared to have reached their 292 maximum body mass and tarsus and wing length. The data are composed of repeated 293 measurements of the same individuals over time; therefore, we first corrected for temporal 294 autocorrelation between the measurements using an ARMA(1,1) model for the residuals (Zuur 295 et al., 2009). Second, as mothers produced several eggs, the models included nested random effects, with measured individuals nested into mother identity, allowing for random intercepts. 297 GAMMs allow modelling the vertical shift of the curves (i.e., changes in intercepts) and their 298 shape. Treatment and sex were included as predictors. A smoothing function for the age of the 299 birds was included to model the changes in the growth curves, and was allowed to vary by sex 300 or treatment only, or none of these predictors. The interaction between sex and treatment was 301 not analysed due to low statistical power. Additive effect of treatment and sex was tested for 302 the intercept but could not be computed for curve shape. All combinations of the relevant 303 predictors were tested for both shape parameters (i.e., intercept and curve shape). 304

Reproductive maturation, regression and investment 305
Due to low sample sizes in sex-specific responses, we could not perform robust statistical 306 analyses. We therefore present these analyses and results in the supplementary material and 307 only briefly discuss them. 308

Oxidative stress 309
A principal component analysis (PCA) was first performed on measured antioxidant markers 310 (SOD, CAT, GPx, tGSH and GST), to reduce the number of metrics for subsequent analyses. 311 The first and the second principal components (PCs) explained together 60.2% of the variance 312 (Table 1). PC1 and PC2 were then used as dependent variables in separate LMMs. LMMs 313 included the treatment, sex and age of individuals (2 weeks and 4 months old) as fixed factors 314 and the 2-way interactions between treatment and sex, and treatment and age. Mother and 315 individual identities, to account for repeated measures, were added as random intercepts. 316 Malondialdehyde (MDA) is a marker of oxidative damage, which is a different measure from 317 antioxidant activity, and was therefore analysed in a separate LMM using the same parameters 318 as for PC1 and PC2, adding the batch of the assay as an additional random intercept. The 319 marker of cell oxidative status (GSH:GSSG ratio) was analysed with the same model used for 320 PC1 and PC2. 321

Moult 322
Two parameters of moult were analysed in separate LMMs: the timing of moult (i.e., the 323 moult score after one week of short photoperiod), and the rate of moult (i.e., how fast birds 324 moulted). Both models included treatment and sex as fixed factors, and mother identity as a 325 random intercept. The rate of moult was tested by fitting an interaction between treatment and 326 age. This model also included the main effect of age and individual identity, nested within 327 mother identity, as a random intercept to account for repeated measures. Estimated marginal 328 means and standard errors (EMMs ± SE) were derived from the model using the package 329 emmeans (Lenth, 2019). 330

Effects of prenatal THs on hatching success and age of embryo mortality 332
There was a significant effect of elevated prenatal THs on hatching success (GLMM, p = 333 0.05, Fig. 1). Tukey post-hoc analysis revealed that hatching success in the T 3 T 4 group was 334 significantly higher than in the CO group (Estimate±SE = 1.24±0.50, Tukey z = 2.46, p = 335 0.05). The other groups (T 3 and T 4 ) were not different from the control group (all z < 2.22 and 336 p > 0.09). The data suggested a dose-dependent effect that we tested by changing the 337 treatment factor to an ordered variable, as the eggs received an increasing amount of TH from 338 Regarding body mass growth, the top-ranked model showed that the curve shape and 353 the intercept differ according to sex (Table 2). After 10 weeks, females had a larger body mass 354 than males (mean±SE females = 214.4±5.7 g, males = 172.4±4.5 g, Fig. 2), which was 355 supported by the 95% CIs (Table 3). Based on model selection we conclude that the treatment 356 had no effect on body mass growth (Table 2). 357

T 3 to T 3 T 4 injections (T 3 < T 4 < T 3 T 4 ). We found a linear
For wing length, the top-ranked model (ΔAIC ≤ 2) included sex in the intercept, while 358 treatment was not included in the best supported model (Table S1). The 95% CIs (Table 3) confirmed that males had a lower wing length than females (Fig. S3). 360 Concerning tarsus length, the models within Δ AIC ≤ 2 included no predictors for the 361 curve shape but included treatment for the intercept (Table S2). The 95% CIs of the parameter 362 estimates from these models suggested that there was a slight negative effect of T 3 T 4 363 treatment on tarsus growth (Table 3, Fig. S4). However, as the estimates were close to 0 364 (Table 3) and evidence ratios showed that the model with treatment as a predictor was only 365 3.5 times more supported than the null model (Table S2), we conclude that the effect of THs 366 on tarsus length is likely to be very small. Likewise, the second model for tarsus length 367 included sex as a predictor for the intercept, but its 95% CIs overlapped with 0 (Table 3). We 368 therefore conclude that sex had no effect on tarsus growth. 369

Effects of prenatal THs on oxidative stress 378
The elevation of yolk THs had no effect on PC1 or PC2 of antioxidants at either 2 weeks 379 ("chicks") or 4 months ("adults") old (LMM on PC1, F 3,40.3 = 2.40 , p = 0.08; LMM on PC2, 380 F 3,42.2 = 0.92, p = 0.44, treatment × age, F < 0.91, p > 0.44). The age of the birds had a highly 381 significant effect on PC1, with chicks generally having higher antioxidant capacities (CAT,382 GST and tGSH) than adults (LMM, Estimate±SE = -1.34±0.19, F 1,49.2 = 52.1, p < 0.0001). All 383 the other predictors had no effect on either PC1 or PC2 (all F < 2.93 and all p > 0.09). 384 The marker of oxidative damage, MDA, was affected by the elevation of yolk THs 385 (LMM, F 3,43.6 = 3.08, p = 0.04, Fig. 4). Tukey post-hoc analysis showed that the T4 group had The aim of this experimental study was to investigate the short-term and organisational effects 397 of maternal thyroid hormones (THs) in a precocial species, the Japanese quail, by 398 experimental elevation of THs in eggs. Our study is the first to investigate the effects of yolk 399 T 3 and T 4 separately, within the natural range of the study model. In addition we studied both 400 short and long term effects on embryonic development, growth, life stage transitions and 401 oxidative stress. We only detected a positive effect of yolk THs on hatching success. All other 402 response variables studied were not affected by elevated prenatal THs. 403

Effects of prenatal THs on hatching success and embryonic development 404
We found that hatching success increased when the eggs received an injection of both T 4 and 405 T 3 . Previous similar studies reported comparable effects of yolk THs in rock pigeons (Hsu et 406 al., 2017) and in collared flycatchers (Hsu et al., 2019). In these studies, injections consisted 407 of a mixture of both T 3 and T 4 . Our results point towards a dose-dependent effect of yolk THs, 408 as found previously of androgen hormones (e.g., Muriel et al., 2015). Importantly, given that 409 only T 3 binds to receptors, these results also suggest that embryos must express deiodinase 410 enzymes to convert T 4 to T 3 , and/or yolk may contain maternally derived deiodinase mRNA. 411 Indeed, precocial embryos start to produce endogenous T 3 and deiodinase expression has 412 previously been characterised in chicken embryos (Darras et al., 2009;Van Herck et al., 413 2012). In contrast with our study, a similar study in great tits detected no increased hatching 414 success due to the injection of THs (Ruuskanen et al., 2016). The dissimilarities between the 415 studies may come from inter-specific differences in terms of utilisation of yolk THs by the 416 embryos or from context-dependent effects (e.g. due to other egg components). Further 417 comparative and mechanistic studies could help understanding the dynamic of yolk THs 418 during incubation. 419 Increased yolk THs did not improve age of embryo mortality. Similar to our study, 420 Ruuskanen et al. (2016) did not find any difference in the timing of mortality in great tit 421 embryos. Conversely, the study on rock pigeons found that yolk THs increased the proportion 422 of well-developed embryos (Hsu et al., 2017). Similarly to our result on hatching success, yolk THs' effects on embryonic development may differ in a species-specific manner.          Table 2: Results of the Generalised Additive Mixed Models (GAMMs) on body mass growth, with sex and treatment fitted either as intercept, curve shape or both (all combinations tested). A total of 12 GAMMs were fitted and ranked based on their AIC, from the lowest to the highest. Weight: Akaike's weights.

List of symbols and abbreviations
Model Intercept Curve shape Table 3: 95% confidence intervals of the predictors in the top-ranked models according to AIC values (see Tables 2, S1 and S2). Predictors in bold have confidence intervals that do not overlap with 0. For the intercept, the reference groups are female and CO for the predictors sex and treatment, respectively.