Strategic adjustment of parental care: life-history trade-offs and the role of glucocorticoids

Life history theory predicts that optimal strategies of parental investment will depend on ecological and social factors such as current brood value and offspring need. Parental care strategies are also likely to be mediated in part by the hypothalamic-pituitary-adrenal (HPA) axis and glucocorticoid hormones. Here we present an experiment in tree swallows (Tachycineta bicolor), a biparental songbird with wide geographic distribution, asking whether parental care is strategically adjusted in response to signals of offspring need and brood value and whether glucocorticoids are involved in these adjustments. Using an automated playback system, we carried out playbacks of nestling begging calls specifically to females in two populations differing in their brood value: a northern population in Ontario, Canada (relatively high brood value) and a southern population in North Carolina, USA (lower brood value). We quantified female offspring provisioning rates before and during playbacks and plasma corticosterone levels (cort) once during late incubation and once immediately after playbacks. Females in both populations increased feeding rates temporarily during the first two hours of playback but the increase was not sustained for the entire duration of playback (six hours). Cort levels from samples at the end of the playback did not differ between control females and females that received playbacks. However, females that had higher increases in cort between the incubation and nestling period had greater fledging success. These results suggest that females are able to strategically respond to offspring need, although the role of glucocorticoids in this strategic adjustment remains unclear.


Introduction 43
Parental investment comprises costly behaviors that can improve the survival of current 44 offspring at the expense of future reproduction (Trivers, 1972). Life history theory predicts that animals 45 will adopt strategies that optimize the level of parental investment in a given reproductive bout. To 46 determine the optimal level of investment, parents must incorporate several potential cues about the 47 2003; Wingfield et al., 1998). Conversely, experimental studies showed that individuals with high current 66 reproductive effort have a reduced glucocorticoid response to stressors (Lendvai and Chastel, 2008; 67 Lendvai et al., 2007). Finally, glucocorticoid levels often covary negatively with measures of individual 68 condition or habitat quality (Moore et al., 2000). Based on these lines of evidence, levels of 69 glucocorticoids are often expected to be negatively correlated with reproductive investment (Bonier et 70 al., 2009a). 71 More recently, evidence has started to accumulate that corticosterone (the main avian 72 glucocorticoid, henceforth referred to as "cort") at baseline levels may support reproductive investment 73 such that cort levels actually increase with some metrics of reproduction (Bonier et  Evidence for the Cort-Adaptation hypothesis also comes from studies of variation in parental 82 investment and cort within species. For instance in a recent study of tree swallows (Tachycineta bicolor), 83 females with higher baseline cort during the offspring provisioning stage fledged more young than 84 females with lower cort (Bonier et al., 2009b). Female swallows also increased baseline cort levels when 85 they were caring for an experimentally enlarged brood, as compared to females caring for 86 experimentally reduced broods (Bonier et al., 2011). Similarly in house sparrows (Passer domesticus), 87 the number of fledglings a female was able to raise was positively correlated with the change in the females' baseline cort levels from pre-laying to the nestling feeding period (Ouyang et al., 2011). In 89 another study on macaroni penguins (Eudyptes chrysolophus), experimentally increased cort levels 90 within the normal range of baseline caused increased foraging activity of females who subsequently 91 raised heavier chicks than control implanted females (Crossin et al. (2012). 92 Here, we ask two related questions regarding the strategic adjustment of parental care in 93 response to offspring signals in two box-nesting populations of tree swallows in Ontario, Canada and 94 North Carolina, USA. We manipulated offspring demand perceived by females through the use of an 95 automated playback system (Lendvai et al., 2015b) where nestling begging calls were directed 96 specifically to females for six hours when nestlings were six days of age. In addition, we measured 97 baseline cort levels in the same females once during the incubation period and once immediately after 98 the playback period (9-13 days after the first sample, depending on the hatch date). 99 Tree swallows in our southern population in North Carolina have higher annual return rates -a 100 robust proxy for annual survival in this highly philopatric species (Winkler et al., 2004) -and a longer 101 breeding season that in some instances even allows birds to raise a second brood (MS & ÇA, 102 unpublished data). In contrast, tree swallows in Ontario have lower annual return rates and a shorter 103 breeding season, with only one brood raised per pair each year. Greater annual survival in the North 104 Carolina population means that, on average, these birds have more opportunities for future 105 reproduction than birds in Ontario. Higher potential for future reproduction will tend to decrease the 106 value of the current brood (Ardia, 2005), such that current brood value will be lower for NC tree 107 swallows compared to Ontario tree swallows. Thus, the use of these two populations allows us to 108 compare responsiveness of parental investment to cues of offspring demand across populations with 109 different brood values (Bokony et al., 2009;Silverin et al., 1997;Sol et al., 2012). 110 The first question we ask is whether adjustment of parental care effort to offspring begging calls 111 differs between the two populations. This question relates to the Brood Value Hypothesis which predicts 112 that adjustment of parental investment in response to cues from offspring will depend on the value of 113 current reproduction vs. future reproduction (Silverin et al., 1997). Thus, this hypothesis predicts that 114 females should increase parental care in response to experimentally increased offspring demand more 115 (or only) in the northern population with higher brood value compared to the southern population. 116 The second question relates to the role of cort in mediating strategic adjustments in parental 117 effort. The Cort-Adaptation Hypothesis predicts that increases in parental investment should be 118 positively correlated with increases in cort levels (Bonier et al., 2009a). As such, females that received 119 playbacks should show greater increases cort levels than control females, which did not receive 120 playbacks. We also test the prediction from the Cort-Adaptation hypothesis that higher increases in cort 121 during nestling period will predict higher fledging success. 122

Study site and species 124
The tree swallow is a widespread secondary cavity nesting species that breeds across a wide 125 range of latitudes from Alaska and Northern Canada to the southern USA. We studied tree swallows at 126 two field sites where they nest in artificial nest boxes: Queens's University Biological Station, Ontario, survival (Winkler et al., 2004). In our NC population, return rates are around 50% for females (51% in 132 been found in other studies comparing southern and northern populations of tree swallows (Ardia, 134 2005). The procedures used in the study followed the guidelines for animal care outlined by Animal 135 Behavior Society and Association for the Study of Animal Behavior, and were approved by approved by 136 the Institutional Animal Care and Use Committee at Virginia Tech (#12-020) and the Canadian Wildlife 137 Service (#10771). 138

Nest monitoring 139
We monitored the nests by visiting each nest box weekly until the parents started nest 140 construction, after which point we visited the nest box every three days until an egg was detected. We 141 checked the nest box every day until no new eggs were laid for two days in a row, which indicated that 142 incubation had started. Female tree swallows typically lay 1 egg per day, and begin incubation on the 143 day of laying of the last egg (Winkler et al., 2011). The date of laying of the last egg was considered day 0 144 of the incubation period. The incubation period typically lasts 14 days, so we checked each incubating 145 nest daily starting from day 12 of the incubation period until all chicks hatched to determine the date of 146 hatching, which was defined as the day when the first chick hatched. Day of hatching was considered 147 day 0 of the nestling period. Throughout the nestling period, we checked the nest at least every 3 days 148 until day 16, at which point we stopped disturbing the nest until day 22 to determine fledging success. 149

Banding parents and nestlings 150
We captured females using box traps at their nest on day 10 of the incubation period to record 151 body measurements (tarsus, wing chord, weight, skull size), collect a blood sample for cort analysis 152 (within 3 minutes of capture), and mark birds with a numbered metal band and a unique passive 153 integrated transponder (PIT) tag that was integrated into a plastic colored leg band (EM4102 tags from 154 except that we did not collect blood samples to minimize handling time and capture stress for males. 157 Males were also tagged with a numbered metal band and a blue PIT tag. We report a detailed analysis of 158 male parental behavior as a function of treatment and female behavior elsewhere. We measured tarsus 159 length and weighed nestlings on the afternoon of day 6 and again on day 12 when each nestling 160 received a numbered metal band. 161

Playback experiment 162
We recorded begging calls from 10 nests on the afternoon of day 6 by pointing a Sennheiser 163 ME66/K6 directional microphone attached to a Marantz PMD 660 Solid State recorder into the nest. To 164 initiate nestling begging, we tapped at the nest entrance, which is a similar sound to what the parents 165 make as they land on the nest box. We used the software Syrinx (John Burt, Seattle, WA; 166 www.syrinxpc.com) to create 30 second stimulus files from the recorded begging calls, with a standard 167 call rate that initially was 14 begs/sec that gradually decreased to a constant 4 begs/sec, simulating a 168 natural pattern of begging in which the nestlings beg vigorously immediately upon arrival of the parent 169 and then tail off gradually. The 10 stimulus files were randomly allocated to the treatment nests. We 170 used a radio-frequency identification (RFID) reader (an upgrade of the model described in Bridge and 171 Bonter, 2011), that was obtained from Cellular Tracking Technology, PA, USA) attached to a micro-172 computer (Raspberry PI) to carry out the playbacks automatically every time the female (but not the 173 male) perched at the nest box entrance. The playback set-up is described in detail elsewhere (Lendvai et 174 al., 2015b). Briefly, we attached an antenna around the entrance hole of the nest box that was 175 connected to an RFID reader. The RFID reader in turn was connected to a Raspberry PI computer which 176 was running a Python script that played back the begging calls for 30 seconds every time the RFID reader 177 detected the female's PIT tag, with the exception of a refractory period of 2 minutes from the start of 178 each playback (to avoid situations where the playback would be triggered when the female left the nest). The playback apparatus was also installed for control nests, but no sound was played. Treatments 180 were allocated to the nests using a randomized block design, to control for seasonal differences. 181 The playback system was set up in the morning around 7am on day 6 post hatching and stopped 182 approximately 6 hours later when we captured the females in their nest box and obtained a second 183 blood sample for cort analysis. We had 21 control and 15 playback nests in NC and 19 control and 18 184 playback nests in Ontario. In NC, three of the nests that were intended to be playback nests never 185 received any playbacks due to the failure of the system, and as such they were included in the analysis 186 as control nests, which caused the uneven sample sizes. One additional nest in NC was excluded from 187 analyses because it only received 2.7 hr of playback due to equipment failure halfway through the 188 experiment, making it intermediate to control and playback conditions. 189 Blood sampling and hormone assay 190 We obtained blood samples (approximately 120 μl) by puncturing the brachial vein within 3 191 minutes of capturing the females to minimize the influence of the stress of capture on measured cort 192 levels (Romero and Reed, 2005). Blood was stored on ice in the field, and centrifuged in the laboratory 193 within 6 hours to separate the plasma. The plasma was then stored at -20°C until taken to Virginia Tech 194 for hormone assay. Hills, CA 91301, Product number: B3-163). We then added dextran-coated charcoal to separate cort 201 bound to antibodies. Intra-assay variation of known concentration standards was 3.93%.

Quantifying parental effort 203
We quantified parental visit rates in two ways: first, we carried out 1-hour feeding watches on 204 day 5 and day 6 (the day before and the day of the treatments) where an observer sat 30 m from the 205 nest and noted every visit of the male and female using a spotting scope and a voice recorder. We also 206 quantified visit rates from the RFID records as described in detail in Lendvai et al. (2015a). We checked 207 the visit rates from 1-hour nest watches against the visit rates calculated from RFID logs of the same 208 time periods. There was a high correspondence between the two (r= 0.68, p=0.2 × 10 -7 , for females and 209 r=0. 67, p=0.4 × 10 -7 for males, n= 43). Because the RFID observations spanned the entire duration of the 210 experiment we used these as the main measure of parental visit rates. Visit rates are an excellent 211 measure of the feeding rates in the tree swallows, as most visits are for feeding (McCarty, 2002). 212

Data analyses 213
We used generalized linear mixed models (GLMM) to assess the effects of treatment and 214 population on cort and behavioral data. We entered the cort data into a GLMM with time period 215 (incubation or nestling), treatment (playback vs. control), and population (Ontario vs. NC) as fixed 216 factors. Playback stimulus and bird ID were included as random factors. 217 We analyzed the female feeding rates derived from RFID recordings with GLMMs using the fixed 218 factors treatment (playback vs. control), population (Ontario vs. NC), and time period. For the latter 219 factor, we used four levels: pre-treatment (day 5) feeding rates (6 hrs during the same time of day as the 220 experimental period on the next day) and feeding rates from the period while the playback or control 221 treatment was in effect in day 6, which we further divided into three two-hour periods to assess any 222 temporal changes in effects of playback on female behavior. We included playback stimulus and bird ID 223 as random factors and also included an offset variable for log of duration of playback to control for the 224 variation in how long the birds were exposed to the playbacks (mean= 6.24 ± 0.05 SE hours). Because of the large number of predictor variables (3 fixed factors and their interactions) in these mixed models, 226 we used a model averaging approach. We first ran a model with all predictor variables and their 227 interactions and subsequently used model averaging with R-package MuMIn (Bartoń, 2013). In the 228 averaged models, we included all models within 2 AICc of the best model (i.e. the model with the lowest 229 AICc). 230 We also examined whether change in cort (base-10 logarithm of ratio of post-treatment cort to 231 pre-treatment cort) between incubation and nestling periods predicted number of chicks fledged using a 232 Finally, we analyzed nestling growth (nestling mass on day 6 and day 12) as well as fledging 236 success using GLMMs. In these models, we examined the fixed factors population, treatment, and their 237 interactions and included relative lay date (number of days from the first egg of the respective 238 population) as a random factor as it has a strong effect on clutch size with later clutches containing 239 fewer eggs in both populations. 240

Results 241
Female feeding rates: Nestling playbacks had a transient effect on female feeding rates. Females 242 that received playbacks of nestling begging calls increased their feeding rates in the first two hours of 243 the playbacks on day 6, compared to their average feeding rates the previous day. No such increase was 244 observed in the control females (Figure 1, see model results in Table 1). No other main or interaction 245 effects were significant (Table 1) There was a positive correlation between female change in cort (ratio of post-to pre-treatment 264 cort) and fledging success. In the averaged model, females with greater increases in cort from the 265 incubation to the nestling stage fledged more offspring (Figure 3, see Table 4 for the averaged model). 266 The effects of treatment and population were not significant.

273
Mass: There was a significant effect of nesting stage on body mass: almost all females lost 274 weight from the first to second capture (see Table 5 for the best model). The main effects of population 275 and treatment were also significant. The main effects were modified by two significant interactions: 276 treatment by population (females in the playback treatment were lighter than the control females in the 277 NC but not in Ontario population), and population by stage (females in NC lost more weight between 278 the incubation and nestling periods as compared to females in Ontario). Finally, there was an effect of 279 relative laying date with females starting to lay eggs later in the season being lighter.
Clutch size, nestling mass, and fledging success: There were no significant effects of treatment 281 on clutch size, nestling mass at day 6 and day 12, or fledging success. Nestlings in Ontario were on 282 average significantly heavier at both day 6 and day 12 (Tables S6 and S7, Figures S2, S3), but clutch size 283 and number of nestlings fledged did not differ between the populations. There were no significant 284 interactions of population and treatment in any of the models (see supplementary information). 285

Discussion 286
Our aim was to manipulate nestling begging calls to study (1) whether a perceived increase in 287 offspring demand induces a change in parental effort of females in two different populations of tree 288 swallows with distinct brood values. If there was a change in parental effort, we further asked whether it 289 led to (2) an increase in baseline cort as predicted by the Cort-Adaptation Hypothesis and whether the 290 changes in baseline cort from incubation to nestling period was predictive of fledging success. We found 291 that (1) females increased parental effort in response to offspring begging call playback in both 292 populations, but the increase was transient and confined to the early hours of the playback treatment. 293 Consequently, (2) baseline cort levels obtained from blood samples at the end of the 6 hour period did 294 not differ between the control and playback groups. However, change in baseline cort from incubation 295 to nestling period did predict how many nestlings females fledged (with higher increases in cort 296 associated with greater fledgling success). 297

Change in parental care and baseline cort 298
Prior experiments on tree swallows and other species showed that glucocorticoids vary with 299 parental effort (Bonier et al., 2009a;Bonier et al., 2009b;Bonier et al., 2011). These studies generally 300 looked at changes in parental care over a longer period (e.g. through brood enlargement or reduction 301 throughout the nestling period). Our aim here was to extend these findings by asking whether cort 302 would dynamically co-vary with changes in cues of offspring demand in the short term. Our failure to detect a difference in cort between females in the playback and control groups does not support this 304 hypothesis. Given our relatively large sample size (n = 73 nests) we do not believe the absence of a 305 treatment effect is due to lack of power. We do however offer another caveat in that the effect of the 306 nestling playbacks on female parental behavior was transient and confined to the first few hours of the 307 playbacks whereas we obtained blood samples at about 6 hours after the start of the playback, and 308 compared cort in those samples to cort measured 9-10 days prior to the playback. By the time the post-309 experimental blood samples were collected, female nest visit rates had returned to baseline levels, and 310 were not significantly different between the playback and control group. Our experimental approach 311 may have therefore lacked the precision to detect a transient increase in cort corresponding to the 312 transient increase in female feeding rates. 313 The fact that playbacks had only a transient effect on female feeding behavior is somewhat 314 surprising given that previous studies manipulating female feeding behavior through automated or 315 manual playbacks of nestling begging calls used durations from 1 hour (in great tits, Parus major; Hinde, 316 2006; and in blue tits; Lucass et al., 2016) or several days (in pied flycatchers, Ficedula hypoleuca; 317 Ottosson et al., 1997), and in both cases found an effect of the playbacks on behavior. This could reflect 318 habituation to the stimulus, and/or an inability of the females to maintain a high rate of feeding, 319 although the latter possibility seems less plausible given the above studies found persistent effects in 320 different species. Whatever the cause, the transient effect of the playbacks means that any effect on 321 cort may also have been transient and therefore would only have been detected if we had captured the 322 females when the playbacks had their maximal behavioral effect. 323 The Cort-Adaption Hypothesis predicts that increases in baseline cort during the period of 324 parental care should increase fledging success. We found support for this prediction: females with the 325 greatest increases in cort from incubation to nestling period fledged more offspring. This finding is consistent with an earlier study in the Ontario population, in which experimentally increased broods 327 induced greater increases in cort through the breeding season, and changes in cort within females were 328 positively correlated with fledging success across both experimental and control groups (Bonier et al., 329 2011). The current results extend the previous findings by showing that the positive link between small 330 increases in baseline cort and fledging success can be detected in natural brood sizes. 331 The finding that females in both populations adjusted parental care in response to begging calls 332 is in contrast to an earlier study in this species that found effects of brood size manipulation on females 333 consistent with the brood value hypothesis (Ardia, 2005). In that study, broods were either enlarged or 334 reduced by 50% in two populations of tree swallows in Alaska and Tennessee. Females in Alaska 335 increased their nest visit rate in the enlarged condition to maintain the same level of nestling condition, 336 whereas females in Tennessee did not increase their visit rate and, subsequently, nestlings in enlarged 337 broods were of lower condition. These results were consistent with the life-history theory and the 338 brood-value hypothesis (Ghalambor and Martin, 2001;Martin et al., 2000). However, the present result 339 may suggest that females in southern populations are able to increase their parental effort in the short-340 term similar to their northern counterparts (the finding in the present study) but may be limited in the 341 long term due to lower food availability in the southern populations (the situation faced by females due 342 to permanent addition of extra nestlings in the study by Ardia, 2005). Indeed, Ardia (2005) found that 343 insect availability was lower in the southern population, making a chronic increase in offspring demand 344 harder to meet for the southern parents. 345 In summary, our results suggest that females are able to flexibly adjust their feeding rates in 346 response to simulated increased demand from their nestlings. Additionally, increases in baseline cort 347 levels of females from incubation to nestling period predicted fledging success across broods. Therefore, 348 although it is unclear whether glucocorticoids are involved in short-term strategic adjustment of parental care, the data suggest that longer-term changes in baseline cort levels are positively correlated 350 with fitness. We believe the data warrant further research into hormonal changes that may occur at 351 shorter time scales and that glucocorticoids may play a casual role in short-term adjustment of parental 352 effort. 353