Maternal glucocorticoids do not predict reproductive investment nor breeding success in a free-ranging bird

4 negatively associated with habitat quality, body condition, hatching success, and fledging 64 success across various species, e.g. blue tits Cyanistes caeruleus (Henderson et al., 2017); tree 65 swallows Tachycineta bicolor (Patterson, Winkler & Breuner, 2011); barn swallows Hirundo 66 rustica (Saino et al., 2005); Adelie penguins Pygoscelis adeliae (Thierry, Ropert-coudert & 67 Raclot, 2013). 68 In contrast, a positive relationship is predicted by “the CORT-adaptation hypothesis.” 69 This hypothesis suggests that because CORT can mediate the mobilization of fuels, causing 70 changes in behaviour or physiology that can increase investment in reproduction, elevated CORT 71 will lead to higher fitness during energetically demanding times (Wingfield & Sapolsky, 2003; 72 Bonier et al., 2009a). Indeed, across a variety of species and life history strategies individuals 73 with higher reproductive success have been reported to have higher CORT levels, e.g. eastern 74 bluebirds Sialia sialis (Burtka, Lovern & Grindstaff, 2016); black-legged kittiwakes Rissa 75 tridactlya (Chastel, Weimerskirch & Wing, 2005); petrels Macronectes spp (Crossin et al., 76 2013); western bluebirds S. mexicana (Kleist et al., 2018); mourning doves Zenaida macroura 77 (Miller, Vleck & Otis, 2009). 78 In fact, there may exist no consistent relationship between CORT and fitness, due to a 79 variety of factors masking directionality (Madliger & Love, 2016a). For example, a lack of 80 relationship could be due to different functions of CORT; when resources are plentiful, elevated 81 CORT could stimulate energy mobilization and parental provisioning; however, CORT could 82 also be elevated in parents experiencing stressors (Vitousek, Jenkins & Safran, 2014). Even 83 within a breeding season, different stages can have differing parental energetic requirements 84 presumably requiring different levels of GC-mediated energy mobilization (Humphreys, Wanless 85 & Bryant, 2006; Nager, 2006; Tulp et al., 2009; Sakaluk, Thompson & Bowers, 2018). 86

Nestling measurements 156 Beginning on day 12 of incubation (incubation day 0 = first day no new eggs were laid, and eggs 157 were warm to the touch), nest boxes were checked twice daily. The hatch day of the first nestling 158 was defined as day 0 for that nest. It was not possible to match nestlings to egg identity. We  Blood sampling procedure 169 We recaptured adult females in nest boxes between day 2 and 5 of incubation (both years) and 170 between day 3 and 6 post-hatch during chick rearing (in 2016 only) between 0600 and 1200 171 hours. Upon capture, we collected a 100µl blood sample from the brachial vein using a micro-172 capillary tube within three minutes of the female entering the nest box. The mean time taken to 173 draw blood (±SE) was 118 ± 5s (N=57) during incubation, and 115 ± 8s (N=33) during the 174 nestling stage. Blood samples were kept on ice for up to eight hours. Samples were then 175 centrifuged for four minutes at 19,200 x g (Thermo IEC Micro-MB) before plasma and red blood 176 cells were frozen separately at -80 o C. Prior to release, we recorded female body mass and 177 marked the tail feathers and right primaries with a spot of white acrylic paint to distinguish females from males during subsequent behavioural observations (Whittingham, Dunn & 179 Clotfelter, 2003;Bonier, Moore & Robertson, 2011). If the female had not been captured 180 previously (i.e., she was to become a Control female), upon first capture during incubation we 181 recorded her head-bill length and wing length, and banded her. 182 We collected nestling blood samples (50µl) from the brachial vein on days 7 or 8, and 13 183 post-hatch. Samples taken on day 7 or 8 post-hatch were for molecular sexing and were added to 184 1ml of lysis buffer in the field and subsequently stored at -20 o C. Samples collected on day 13 185 were centrifuged and plasma was stored at -80 o C (as part of a separate study).

187
Adult behavioural observation 188 On day 7 or 8 post-hatch between 0830 and 1400 hours, nest boxes were observed from a 189 distance of 10 m for 1 hr , during which we counted the number of visits 190 made by males and females to the nest box. This was the maximum distance at which it was still 191 possible to distinguish the sex of the adult entering the box through binoculars. Observations 192 made mid-day have been shown to provide the best estimates of feeding rate, although 1-hour 193 observations periods done at any time of day predict total daily feeding rates (Lendvai et al.,194 (Washburn et al., 2002). Plasma was diluted 1:25 (10µl of plasma plus 240µl of assay buffer).

201
Samples that were not detectable were set to the lowest point on the standard curve (3.125 202 ng/ml), following Hogle and Burness (2014). We did not extract plasma because a serial dilution 203 of non-extracted plasma pooled from five individuals was parallel to the standard curve. A total 204 of 23 individual assays was performed. The average inter-assay variation was 7.99%; the intra-205 assay variation was 10.7%. Nestling blood samples taken on day 7 or 8 post-hatch were used for genetic sexing using the 209 CHD1W and CHD1Z genes (Fridolfsson & Ellegren, 1999;Hogle & Burness, 2014 were not always able to collect all measurements from all individuals. We included 'year' as a 229 factor only in analyses of maternal CORT during incubation (CORTinc), because during the 230 nesting phase (CORTnest) we measured CORT in one year only (2016). 231 We constructed our statistical models including only main effects that were of likely 232 biological importance; as such, not all two-way interactions were included. We report outputs 233 from global statistical models. Because we had explicit hypotheses, and because none of our 234 response variables was correlated, we did not to use a post-hoc correction for the number of tests 235 performed (Perneger, 1998;Streiner, 2015).

237
Morphological and hormonal measures of adult females 238 We ran preliminary tests to determine whether females that had been assigned to CORT or Sham 239 treatment groups differed in pre-implant body mass (measured at time of implant; females in the 240 Control group were not captured prior to incubation and thus there was no pre-implant mass 241 measurement). To test for possible differences in body size among treatments, we compared a 242 female's wing length (measured pre-laying in the CORT and Sham treatments, and during early 243 incubation in the Control females). Finally, we tested for differences in clutch initiation date in 244 Julian days among treatments. Separate linear models (LM) were run with female pre-laying 245 mass, wing length, and clutch initiation date as the response variable, and treatment (CORT,246 Sham, Control), site (Nature areas, Sewage Lagoon), year (2015,2016), and age (Second year, SY; After second year, ASY) as the predictors. We did not include any interactions terms as they 248 were not of a priori interest.

249
Implanted females recaptured during incubation may have differed phenotypically from 250 individuals that abandoned their nests (and were never recaptured). To test this, we ran a 251 generalized linear model (GLM) with binomial errors, with recapture status (recaptured/non-252 recaptured) as the dependent variable, and treatment and year as the fixed effects. To test 253 whether the total number of individuals that subsequently laid eggs differed between the CORT 254 and Sham maternal treatment groups, we used a chi-square test (because Control females were 255 only captured post-egg laying, they were not included in this analysis). To test whether treatment affected maternal CORT levels within each breeding stage (incubation 259 and nestling), we used linear models (LM) with either CORT during incubation (from hereafter 260 CORTinc) or CORT during the nestling stage (CORTnest) as the response variable and maternal 261 treatment, age, site, sample time (time from initial contact with bird to end of blood sample), and 262 clutch initiation date (in Julian days) as fixed effects. We had no a priori predictions regarding 263 interactions, so none was included in the models. 264 We analyzed CORTinc and CORTnest separately because CORTnest was only measured 265 in 2016. Baseline CORTinc measurements (N=56) had one suspected outlier (121.22 ng/ml) 266 removed prior to analysis. This value was > 3 standard deviations from the mean; considerably 267 higher than the 0.5 to 14 ng/ml range reported for previously (Franceschini et al., 2008;Ouyang 268 et al., 2011;Patterson, Winkler & Breuner, 2011;Madliger et al., 2015). Preliminary analyses were run with and without this outlier, and although no difference was found in the pattern of 270 significance of parameters, we chose to exclude it. reproductive success. To test whether nestling mass differed with maternal CORT or treatment, 295 we used a LMM with nestling mass at day 14 as the response variable and CORTnest, maternal 296 treatment, maternal age, and site as fixed effects (year was not included because CORTnest was 297 measured in 2016 only), and Nest ID as a random effect. Finally, to test whether fledging success 298 differed with maternal CORT or treatment, fledging success (0 or 1 for each chick) was used as 299 the response variable in a GLMM with binomial errors with maternal treatment, maternal age, 300 site, and CORTnest as fixed effects, and Nest ID as a random effect. No interaction terms were 301 included in these analyses.

303
Measures of female survival 304 We estimated female survival by using the return rates of adult females to the study sites the 305 following spring and comparing this with CORTinc or CORTnest during the previous year in 306 separate models. Return rate (either 0 or 1) was the response variable in a general linear model 307 (GLM), with CORTinc (or CORTnest), treatment, year, age, site, and number of nestlings 308 fledged as main effects. In analyses of CORTnest, "year" was not included in the model because 309 CORTnest was only measured in a single year (2016). We also used a chi square test to 310 determine whether there was a difference in return rate by treatment or year.

313
Morphology and hormonal measures of adult females 314 We implanted 51 females with corticosterone-filled implants (CORT), and 48 with sham 315 implants (Sham); an additional 25 females were captured for the first time during incubation were allocated to the Control treatment (Table 1). There was no difference in pre-egg laying 317 body mass between females allocated to the CORT and Sham groups (treatment: p=0.108; Table   318 S1); females in the Control group were not captured prior to egg laying, so there was no pre-egg 319 laying mass. Focussing on individuals that retained their implants, wing length and clutch 320 initiation date did not differ significantly among treatments (Wing length, treatment: p = 0.238;

321
Clutch ignition date, treatment: p=0.859; Table S1). There was no significant difference between 322 the Sham and CORT treatments in the percentage of females that retained their implants and    Table 3). Although maternal treatment did not influence nestling growth 349 rate between days 3 and 7 (p = 0.730), there was a marginally significant negative relationship 350 between maternal CORTnest and nestling growth rate (p = 0.076, Table 3). Maternal age 351 influenced nestling growth rates, with nestlings from SY mothers having higher growth rates 352 than nestlings from ASY mothers (age: p<0.001, Table 3).  (Table 4). There was, however, a nonsignificant trend for mothers with 358 higher CORT during incubation to have higher hatching success (CORTinc: p=0.061), and for 359 older females to have higher hatching success that younger females (age: p=0.093; Table 4).

360
Nestling mass at day 14 post-hatch was not predicted by either CORTnest, nor maternal treatment (CORTnest: p=0.123; treatment p=0.372), although nestlings at the Nature Area site 362 tended to be heavier (Site: p = 0.092; Table 4).

363
The probability of a nestling fledging did not vary with maternal corticosterone levels

DISCUSSION
Our data do not convincingly support either the CORT-fitness or the CORT-adaptation 385 hypothesis. During egg incubation, corticosterone levels of female tree swallows were positively 386 related (albeit non-significantly) to one measure of reproductive success. However, during the 387 nestling stage, there was no relationship between corticosterone and indices of either 388 reproductive investment or reproductive success. During neither period did we detect a negative 389 relationship between CORT and fitness, as predicted by the CORT-fitness hypothesis.  During incubation, individuals may experience more unpredictable stressors than during the 421 nestling stage (Romero, 2002). For example, challenging environmental conditions such as lower 422 temperatures and scarcer food resources in early spring can cause a negative relationship 423 between both temperature and foraging success and baseline CORT levels, depending on the 424 fitness and environmental measure used (Angelier et al., 2007;Wingfield, Weimerskirch & 425 Chastel, 2010; Ouyang et al., 2015). Because higher baseline levels may prime the body to 426 perform better under stress, females with higher baseline CORT during incubation in our study 427 may have been better able to meet these challenges (Romero, 2002). 431 We predicted that if there were a relationship between CORT and reproductive investment and 432 success, it would most likely emerge post-hatch, given the higher maternal energy expenditure 433 required during chick rearing than during incubation (Nilsson & Raberg, 2001;Humphreys, 434 Wanless & Bryant, 2006;Sakaluk, Thompson & Bowers, 2018) but see (Williams, 2018) . reproductive success may be due to various fitness measures used, the relative importance of 444 paternal investment, or environmental variation.

445
While female tree swallows are solely responsible for egg incubation, nestling 446 provisioning is shared with the male . As a result, variation in paternal 447 quality may obscure relationships between maternal CORT and investment during the nestling 448 stage. A lack of relationship between CORTnest and female nest box visits has been found in 449 bluebirds (Davis & Guinan, 2014) and other populations of tree swallows (Patterson,Winkler & 450 Breuner, 2011), suggesting variation among females in their glucocorticoid levels may not 451 directly reflect maternal behaviour. In contrast, Madliger & Love (2016b) found that higher baseline CORTnest in female tree swallows correlated with lower rates of maternal provisioning; implantation), the baseline CORT levels of implanted birds did not differ from unmanipulated 499 birds. Across species, silastic implants have been successfully used to raise CORT levels for 500 anywhere from a few days (Astheimer, Buttemer & Wingfield, 2000;Hayward & Wingfield, 501 2004;Criscuolo et al., 2005;Martin et al., 2005;Angelier et al., 2007) to three weeks post-502 implantation in vivo (Ouyang et al., 2013) and in vitro (Newman et al., 2010). However, the use 503 of implants to raise CORT levels has not been consistently successful (Crossin et al., 2012;504 Ouyang et al., 2013;Hau & Goymann, 2015;Lattin, Breuner & Romero, 2016). Although the 505 implants used in our study may have failed to release CORT, this seems unlikely given that in 506 vitro studies have shown that CORT continues to be released across the membrane over 4 weeks 507 (Newman et al., 2010). More likely, the implants resulted in decreased secretion of endogenous 508 CORT via negative feedback, or increased clearance of CORT from the blood via increased 509 excretory activity (Newman et al., 2010;Henriksen, Groothuis & Rettenbacher, 2011;Robertson, 510 Newman & MacDougall-Shackleton, 2015).

511
Rather than experimentally manipulate CORT levels via implants, an alternative 512 approach may be to manipulate maternal condition, such as with feather clipping (Rivers et al., 513 2017), predator experiments (Clinchy et al., 2011Pitk et al., 2012), or density manipulations 514 (Bentz, Navara & Siefferman, 2013).Such an approach would encompass how maternal CORT 515 levels change based on how each female perceives her condition/ environment, how that is 516 reflected in blood CORT levels, and how those levels might influence the next generation   (Romero, 2002), weather (Pakkala et al., 2016) and habitat 524 variability (Madliger & Love, 2016b). If it can be reasonably assumed that these will always 525 differ among individuals, then perhaps there is no consistent relationship, and any relationship 526 detected will always be context-dependent (Madliger & Love, 2016a). A formal meta-analyses of 527 passerine birds, such as was done recently in seabirds (Sorenson et al., 2017), may help clarify 528 patterns.

529
The use of integrative measures of CORT may be an alternative way to improve our 530 understanding of the relationship between CORT and fitness. By measuring CORT deposited in 531 feathers during growth, or metabolites excreted in feces, it may be possible to infer CORT levels 532 over multiple days of the incubation or nestling stage (Lucas et al., 2006;Bortolotti et al., 2008;533 Romero & Fairhurst, 2016). For example, giant petrels that successfully bred had higher feather 534 CORT levels than failed breeders, but were less likely to breed the following year, a pattern 535 which was not observed using plasma CORT from these same individuals (Crossin et al., 2013).

536
Ideally, studies could be extended over the winter, as has been done recently in adult tree 537 swallows (Vitousek et al., 2018). This would help elucidate the longer-term effects of maternal 538 CORT on offspring and maternal and fitness.  Bracketed values represent total number of individuals handled/implanted; non-bracketed values indicate sample sizes of birds with implants that were still present when the bird was recaptured during incubation.

Table 2. Factors contributing to variation in corticosterone levels in female tree swallows during incubation (CORTinc) and the nestling stage (CORTnest).
Year was not included for CORTnest, because data were collected in a single year. Statistically significant main effects are in bold.  Each model includes Nest ID as a random effect; marginal (M) and conditional (C) R 2 values are provided. There were two analyses of Fledging success (A and B), with predictors including either CORTinc or CORTnest, respectively. There were two analyses of Return rate (A and B), with predictors including either CORTinc or CORTnest, respectively.

Response variable
Year was not included in analyses of Return rate B, because CORTnest was measure in one year only. Statistically significant main effects are in bold.