Avian eggshell colouration might function as a post-mating sexually selected signal of female quality, influencing male parental investment and, hence, reproductive success. This hypothesis has been tested for intrinsic eggshell pigments as biliverdin (blue-green colouration) and/or protoporphyrin (brown coloured spots), but never for colourations applied post-laying. Post-laying staining colouration due to, for instance, uropygial secretion of the female could reflect its phenotypic properties and, thus, might be a cue for male investment in reproduction. In hoopoes, the uropygial gland of incubating females hosts symbiotic bacteria that are responsible for the brown colour of their uropygial secretion and of the eggshells, as they cover their bluish-grey eggshells with gland secretion after laying. The secretion protects embryos from pathogenic trans-shell infections and, thus, egg colouration may function as a cue or even as a post-mating sexually selected signal of antimicrobial potential in hoopoes. In a wild hoopoe population breeding in nest boxes in Spain, we test this hypothesis by exploring whether egg colour predicts male parental investment. In accordance with the hypothesis, we found that the amount of food provided by males to incubating females was higher in nests with less saturated eggshell colours. This relationship was affected by female body condition. High quality females in terms of body condition and/or in secretion colour obtained better males in terms of provisioning effort during incubation. Given that eggshell saturation is negatively related to density of bacterial symbionts in uropygial secretions, one possibility is that males may regulate their parental investment in accordance to the expected characteristics of mutualistic bacteria hosted in uropygial glands and deposited on eggshells. We discuss alternative explanations for our results, concluding that the post-mating sexual selection hypothesis is the most likely but experimental modification of egg colour is needed to test it further.

MATERIAL AND METHODS

Study Species

The hoopoe is a migratory bird distributed throughout Europe, Asia and Africa. This species is a cavity nester that readily nests in nest boxes. From February to July, females lay one or two clutches with usually between 6 and 8 eggs (Martín-Vivaldi et al. 1999, Plard et al. 2018). They start to incubate, and paint eggshells with secretion, with the first or second egg in the laying sequence. Just before laying, the uropygial gland of incubating females dramatically increases in size, producing a great amount of secretion (Martín-Vivaldi et al. 2009a) that allows them to cover the entire egg surface of the whole clutch. In addition, the external structure of hoopoe eggshells is different from those of other birds (Martín-Vivaldi et al. 2014a), full of small crypts helping to retain secretion. Asynchronous hatching, due to females starting incubating with the first or second egg, generates a nestling body size hierarchy within the brood (Cramp 1998). Eight days after the first nestling hatched, the size of the female gland decreases and the secretion characteristics returns to prelaying conditions (white and odourless secretions, with reduced symbiotic bacteria load) (Martín-Vivaldi et al. 2009a). Thus, the amount of symbiotic bacteria seems to be consistent across the breeding season with the secretion colour being brown, and after the eighth day of the nestling phase, bacteria progressively decline in numbers until disappear. Only the female incubates, but both sexes take care for offspring. Females stay in the nest during the first 8 days after the first day of hatching and males carry to the nest the food that females use to feed nestlings. Afterwards, both parents provide food to developing nestlings until they fledge, with 24 to 30 days (Martín-Vivaldi et al. 2014b). Given that females uncover the clutch when being fed by their males during incubation, males have a chance to see the eggs. Variation in the colour of the uropygial secretion and eggshells is noticeable between individuals within our study population (Soler et al. 2014).

Study Area and General Procedures

The fieldwork was carried out in the Hoya de Guadix (37 ºC 18´N, 11´W), Granada (southern Spain) during the 2015 breeding season. This hoopoe population has been studied throughout the last 25 years. Hoopoes breed in this area in natural cavities and in nest boxes situated in trees. Nest boxes were made of cork with the following dimensions: 5.5 cm (entrance diameter), 24 cm (bottom-to-hole height), 35 × 18 × 21 cm (internal height × width × depth).

            Nest boxes were inspected from early March to the end of July every five days. Assuming that one egg was laid daily (Cramp 1998), laying date was considered the day when female laid her first egg, and hatching date as the day on which the first nestling hatched. Females were captured by hand inside the nest boxes twice, 15 days after laying date and 5 days after hatching of the first egg. Nestlings were sampled 19-20 days after the start of hatching. Bill and tarsus lengths were measured with a calliper (accuracy 1mm) and body mass with a hanging scale (Pesola 0-100 g, accuracy 1 g). The uropygial gland secretion was extracted by means of automatic 1-10 μl micropipettes (see below for more details). Finally, all individuals were marked with numbered aluminium rings (Spanish Institute for Nature Conservation, ICONA). Adult individuals were provided with unique colour ring combinations. After sampling, they were released into their nest box. To prevent contamination among nests and among nestlings, all the manipulations were made using disposable latex gloves previously cleaned with 96% ethanol.   

Body condition of females was estimated as the residuals of body mass on tarsus length³ (Senar and Pascual 1997, Peig and Green 2010, Labocha and Hayes 2012).

Parental Investment

The visits to the nest by parents were video-recorded at three different stages: incubation (10 days after laying the first egg) and days 3rd and 11th of the nestling period (nestlings 1 and 2, respectively). Digital video cameras (Sony Handycam DCR-SR55 and DCR SR190 models) were placed several metres away from the nest, camouflaged by vegetation, stones or trunks. Each recording started around 1600 hours and registered periods of approximately 3 hours. The period considered valid to estimate feeding effort started when  males and females behaved normally and last one to two hours (average: 1 hour 86 minutes; range of video duration: 70 minutes - 2 hours). The length of the period was considered appropriate to represent the overall feeding effort. We identified individuals by their colour ring combinations and other characteristics of plumage or body cues. To estimate the amount of food provided by each individual, we counted the number of prey carried to the nest per hour (feeding rate) and average relative prey size. Relative size of prey was estimated at an ordinal scale. Value 1 was assigned to prey size less than a quarter of the beak length; value 2 for prey size between a quarter and a half of the beak length; and value 3 for prey size larger than the half beak length) (Martín-Vivaldi et al. 1999). Number of prey and prey size were multiplied to obtain an index of provisioning effort. When it was not possible to estimate the size of the prey, we assigned the estimated average size for the recorded individual. We successfully recorded 39 nests and visualized them through the VLC Media Player software 2.2.6. Only first clutches were considered in the analysis.

Egg Colour Measurements and Estimations

Egg colour was measured on the 15^th day of the incubation period, when all the eggs were completely covered by secretion. Eggshell colouration was measured using an Ocean Optics S2000 spectrometer connected to a deuterium-halogen light (D2-W, Mini). A black bag that wrapped the tip of the optical fibre and the egg was used to standardise ambient light conditions. Before the measurement of each clutch, the spectrometer was calibrated using a standard white and black reference. Reflectance spectra at 10nm intervals from 300-700nm was obtained for all eggs of the clutch. Eggshell colour was measured on five equidistant points on a random line along the long egg axis, from the apex to the base. To determine repeatability, each zone was measured three times. To prevent nest desertion by parents, all measures were performed close to the nest boxes as quick as possible (maximum of 15 minutes), and the nest box entrance was locked during handling time of eggs and the females.

             In a visual signalling context, it is desirable to analyse colour data in a way that is appropriate for the avian vision system (Endler, 1990; Renoult, Kelber, & Schaefer, 2017) given they have four types of cones in their retina. Thus, we estimated the visual parameters from a physiological model that account for the tetrachromatic violet vision (hereafter, VS) of hoopoes (Hart 2001, Ödeen and Håstad 2003) using Avicol V.6 software (Gómez 2006). Prior to all analyses, we applied the following corrections to our interpolated spectra data: negative values were set to zero and reflectance curves were corrected for noise using triangular smoothing (Gómez 2006). Specifically, we used the model proposed by Endler and Mielke (2005) with some modifications proposed by Stoddard and Prum (2008), such as, the non-logarithmic transformation of the photoreceptor response values and the correction for dark colours. Information on the spectral sensitivity of the hoopoe is not available.  Therefore,  we used data from another non-passerine species, the peafowl (Pavo cristatus) with VS (Hart 2002). Moreover, we considered ambient light conditions inside a nest box to calculate the colour visualisation by hoopoes. Finally, we obtained information of achromatic (qrQ) and chromatic components of colouration (i.e., the spherical coordinates tetha and phi, and “r” chroma value) by quantifying the four following colour components: (i) qrQ, which measure the quantum catch for the photoreceptor(s) responsible for brightness perception, adapted to the background (Gómez 2006). (ii) Tetha and (iii) phi, which inform about colour hue and, respectively measure the angle in the red-green-blue plane (between -180º and 180º), and in the ultraviolet/violet sensitive (UV/V) range (between -90º and 90º) (Endler & Mielke, 2005). Finally, r values inform on colour saturation, which measures the distance to the centre of the tetrahedron (Stoddard and Prum 2008, Saino et al. 2013). Since the colour space is a tetrahedral and not a sphere, maximum potential chroma (r_max) depends on hue values. For this reason, we estimated achieved chroma (rA, hereafter saturation) as rA = r / r_max (Stoddard and Prum 2008)_. The data obtained take into account the sensitivity and the stimulation of each cone of the visual system of birds, allowing us to measure the colour in ways relevant to birds (Endler, 1990; Saino et al., 2013). However little is known about neural processes, i.e. how the eyes and brain process colour patterns (Kemp et al. 2015, Renoult et al. 2017, Stoddard and Osorio 2019), therefore, even when using bird physiological visual models, we have to interpret these results with caution.

Repeatability was calculated for each colour parameter obtained for each measured zone of the egg, and for each eggs within the same clutch. In all cases, we verified that eggshell colouration was more variable among than within items (Repeatability of each zone of the egg: GLM: qrQ: r = 0.64, F_{380, 1524} = 10.06, P < 0.001; Tetha: r = 0.60, F_380,1524 = 8.58, P < 0.001; Phi: r = 0.58, F_{380, 1524} = 7.8, P < 0.001; rA: r = 0.71, F_{380, 1524} = 12.94, P < 0.001. Repeatability among the eggs of each nest: GLM: qrQ: r = 0.87, F_51,1853 = 47.49, P < 0.001; Tetha: r = 0.82, F_{51, 1853} = 33.15, P < 0.001, Phi: r = 0.83, F_{51, 1853} = 34.7, P < 0.001; rA: r = 0.90, F_{51, 1853} = 64.43, P < 0.001) and, accordingly, average values of colour parameters for each egg and then for each nest were used. 

Uropygial Secretion Bacterial Loads

To obtain information about the bacterial load of the uropygial secretion, we sampled incubating females. Briefly, we first cleaned the surroundings of the gland with a cotton swab soaked in 96% ethanol. Second, we gently introduced a previously autoclaved tip of an automatic 1-10 μl micropipette, into the opening of the papilla of the gland and directly pipetted the secretion. Finally, the secretion was introduced into a sterile microcentrifuge tube, repeating this procedure until the papilla emptied. To estimate bacterial loads, the samples were processed within the following 24h. Briefly, 5 μl of the secretion were homogenised with 45 μl of PBS in a sterile microcentrifuge tube. Then, 5 μl of each tenfold dilutions to 10^-4 of this mixture was spread onto the following medium (Scharlau Chemie S.A., Barcelona): Tryptone Soya Agar (TSA), a broadly used general medium to grow mesophilic bacteria. The plates were incubated aerobically at 37 ºC for 24h before colony counting. Estimates of bacterial loads were standardised to number of colony forming units (CFU) per millilitre of secretion (Nº colonies * 10^{dilution factor}) / ml spread).

Statistical Analysis

To look for the best combination of eggshell colour variables that explained the bacterial load in the uropygial gland secretion of the female, we used a best subset General Regression Model (GRM) with qrQ, tetha, phi and rA as predictors and CFU/ml as dependent variable. The best subset GRM model were analysed by means of Mallow´s CP (Mallows 1973), equivalent to Akaike´s Information Criterion (AIC) (Boisbunon et al. 2013). Since tetha and phi values are located in the same quadrant (Range values: tetha: from 4.45 to 90.42; phi: from -86.65 to -66.27), we can use these variables in linear scales in the sense that increasing values always imply changes in the same direction. Moreover, rA, the variable that explains bacterial load and feeding effort (see result), has a linear nature.

To explore the effect of eggshell saturation (rA) on male provisioning effort, we performed a multivariate General Lineal Model (GLM) that also included clutch size, laying date and female body condition and its interaction with rA, as additional independent factors. We included this interaction to test if the relationship between rA and male provisioning effort changes depending on female body condition. To check for the effect of those predictors in each phase separately, we used the corresponding Univariate GLM models. To test if other colour variables besides rA had an effect on male provisioning effort (tetha, phi and/or qrQ), we performed best subset GRM in each phase separately, adding the rest of the colour variables to the initial model and forcing the presence of significant predictors of each phase.

Tolerance values were obtained for each variable to check possible collinearity problems between them. The high values of tolerance obtained (higher than 0.748) indicated no problem of colinearity in our models (tolerance > 0.1 and Variance Inflation Factor (VIF) < 10) (Quinn and Keough 2002). All the variables followed approximately a normal distribution except for CFU/ml of uropygial secretion in TSA medium that was log10+ 0.1 transformed.

All statistical tests were performed with Statistica 7 software (Statsoft 2006).