Uropygial gland and bib colouration in the house sparrow

General procedure

The study was conducted during March 2011 with 21 adult male house sparrows captured with mist-nets on a farm in Padul (SE Spain, 37°01′N, 3°37′W) and transported to an outdoor aviary located in Moraleda de Zafayona (37°11′N, 3°57′W). No bird suffered any damage during capture, transport, maintenance in the aviary, or as a consequence of the experiment. The aviary structure followed the recommendations of the European directive as well as national legislation. Measuring about 20 m³, the aviary was built with bricks at the base (1-m height) and a complete wall, the remaining being covered with a mesh supported by a metal framework. The structure was designed to avoid injuring the birds. A roof was provided to protect birds from rainfall and direct sunlight. All birds were individually marked with colour rings, and were supplied with food (seed mixture, fruit, and different vitamins and minerals) and water with ad libitum access, as well as diverse perches, and trays with water and powder for bathing and dust bathing. The aviary, and especially the food and water containers, were carefully cleaned and disinfected before the capture of the birds. Confinement lasted for a week, and when the study ended, the sparrows were released in the same place where they had been captured. The study was performed with the permission of the Andalusian government.

On 03 March 2011, 12 sparrows were subcutaneously injected in the patagium with 0.1 mg of LPS (serotype 055:B5, L-2880, Sigma Aldrich), diluted in 0.01 ml of isotonic phosphate-buffered saline. LPS acts as an antigen, provoking a humoral immune reaction that mimics an infection, and diverts energy from other functions to the immune system. Consequently, inoculating LPS usually lowers body mass in the house sparrow (Bonneaud et al., 2003; Moreno-Rueda, 2011). Another nine sparrows were injected with 0.01 ml of PBS as a sham control. To determine whether the antigen effectively stimulated the immune system, I measured the thickness of the patagium where the substances were inoculated with a pressure-sensitive micrometre (Mitutoyo; accuracy 0.01 mm). Measurements were taken before injecting the substance and four hours afterwards, when an immune response to LPS is already detectable (Parmentier, De Vries Reilingh & Nieul, 1998). Then, I tested whether the patagium was significantly swelled in LPS-inoculated birds, which indicates an immune response to the antigen.

I took a number of measurements just prior the experiment and just when the experiment was ended (on 10 March). Firstly, I measured the length, width, and height (from the base of the gland to the base of the papilla) of the uropygial gland (three times each) with a digital calliper (accuracy 0.01 mm), and estimated its size by multiplying the three measurements, which is a good estimator of gland volume and preen-oil production in house sparrows (Pap et al., 2010). I estimated the repeatability of uropygial gland size by measuring twice the gland in 21 individuals randomly selected (following Lessells & Boag, 1987). Repeatability was 0.76 (F_1,19 = 72.7, p < 0.001). Also, I measured body mass with a digital balance (accuracy 0.1 g), and wing length with a ruler (accuracy 0.5 mm). Body condition was estimated as the residuals of the regression of body mass (log-transformed) against the wing length (log-transformed) as a measurement of skeletal body size (review in Green, 2001). Furthermore, I measured bib size by means of photography on gridded paper with a digital camera (Fujifilm 10.2 megapixels; following Figuerola & Senar, 2000). The camera was mounted on a tripod, consistently at the same distance from birds, which were held with the breast plumage combed in order to ensure a normal position. Afterwards, patch surface areas were measured with the program Image J (Abramoff, Magelhaes & Ram, 2004). Photos were scaled using the gridded paper as reference. Then, I adjusted the area using the “colour threshold” tool, and measured the area of bib of each bird with the “analyse particles” tool.

Plus, the colouration of the bib was measured with a spectrophotometer (Minolta CM-2600d). After reference calibration of white, the spectrophotometer was placed over the bib and three beams of light were projected through a hole of 3 mm in diameter. As a result, three reflectance measurements were taken and automatically averaged (Andersson & Prager, 2006). The spectrophotometer did not measure the ultraviolet spectrum, whose measurement is unnecessary given that house sparrows do not reflect ultraviolet radiation in the bib (Václav, 2006). Bib colouration was measured in the L*a*b* colour-space of the Commission Internationale d’Eclairage (CIE; Montgomerie, 2006a). L*a*b* is a three-dimensional rectangular colour space. L* axis represents lightness (0 is completely black, 100 is completely white); a* axis represents red-green gradient (positive values are red, negative values are green); b* axis represents blue-yellow gradient (positive values are yellow, negative values are blue). From L*a*b* values I determined the saturation, i.e., the radiance in a specific part of the spectrum in relation to the radiance from the whole visible spectrum. Saturation was calculated as C* = [(a*)² +(b*)²]^1∕2, measured as the percentage distance from the centre of the colour space to its circumference, where pure spectral colours are represented. Hue angle (the “colour” in common parlance) was calculated as H* = tan⁻¹(b*/a*) (Endler, 1990).

Statistical analyses

In a first analysis, with data taken prior the inoculation of LPS and PBS, I used the t-test to check for differences between LPS- and PBS-inoculated sparrows for wing length (as indicator of structural body size), body condition, uropygial gland size, bib size, and bib colouration (bib lightness, saturation, and hue). In addition, I examined the relationships (by running Pearson correlations) among the different biometrical variables (bib size, wing length, body condition, and uropygial gland size) and bib colouration parameters (lightness, saturation, and hue). I corrected for multiple tests by applying sequential Bonferroni (Rice, 1989), but, given that the correction of Bonferroni increases the rate of statistical error type II (Moran, 2003), I show the uncorrected P-values, indicating if they remain significant when corrected. Given that bib saturation was correlated with body condition, uropygial gland size, and bib size (see ‘Results’), in order to ascertain the independent relationships among these variables with bib saturation, I carried out a multiple regression model (Linear Model) of type-III Sums of Squares (Quinn & Keough, 2002). Collinearity among continuous variables was checked by examining tolerance, which was always >0.90, indicating that collinearity was consistently very low (Quinn & Keough, 2002). Normality and homoscedasticity of residuals of the models were checked according to Shapiro–Wilks and Levene’s tests, respectively (Quinn & Keough, 2002), and, when necessary, the variables were log-transformed in order to improve the fit of the models. Similar multiple regression models with bib lightness and hue as dependent variables were not carried out (as unnecessary) because these variables did not correlate with more than one biometrical variable (see below).

In order to confirm that the inoculation of LPS effectively produced an immune response, I tested with paired t-test for differences in patagium swelling before and 4-h after the inoculation in both LPS- and PBS-inoculated sparrows. I estimated the change in patagium swelling (final minus initial patagium swelling) and used a t-test for unpaired samples to examine for differences in change in patagium swelling between treatments. Also, I used the t-test to check for differences between treatments in the changes (final value minus initial value) in uropygial gland size, body condition, and bib colouration (lightness, saturation, and hue), measured once the experiment ended (seven days after inoculations). Paired t-test were also used to examine the within-individuals change in uropygial gland size.

Lastly, in order to examine which variables were related to changes in bib colouration during the experiment, I carried out correlations among the change in bib colouration variables (change in lightness, saturation, and hue) and bib size, body condition, and both the change in uropygial gland size and in body condition. I identified changes in colouration correlated with initial body condition and change in uropygial gland size (but not with bib size or change in body condition; see ‘Results’). Therefore, in order to ascertain the independent effect of treatment, body condition, and change in uropygial gland size, I performed a set of Linear Models with the change in every component of bib colouration (lightness, saturation, and hue) as dependent variables, and body condition, change in uropygial gland size, treatment, and all the interactions among these variables as predictors. Moreover, although both bib size and change in body condition were uncorrelated with changes in colouration, in order to be sure that these variables did not bias the results, I repeated the models including these variables as predictors. Non-significant interactions (for which P > 0.05) were removed from models (following Engqvist, 2005). Moreover, models including all interactions did not significantly differ from models including only significant interactions (F-ratio for comparing models, in all cases P > 0.25). As described above, normality and homoscedasticity of residuals of the models were checked according to Shapiro–Wilks and Levene’s tests, respectively (Quinn & Keough, 2002), and, when necessary, the variables were log-transformed in order to improve the fit of the models.

The complete dataset is available in Table S1. All analyses were performed with STASTISTICA 8.0 (StatSoft, Tulsa, OK, USA).

Descriptive relationships among variables prior the experiment

Prior to the experiment, no statistical differences were found between control (PBS-inoculated) and experimental (LPS-inoculated) house sparrows in wing length, body condition, uropygial gland size, and bib colouration (light, saturation, and hue) (Table 1). However, despite the randomization of the treatment assignment, bib size was significantly larger in PBS- than in LPS-inoculated birds (Table 1). Nevertheless, differences in bib size turned not significant when applying Bonferroni correction and, moreover, this possible differences did not seem to affect results of the experiment (see analyses below and Table 3C). Bib size was not significantly correlated with any colour parameter (Table 2), although there was an almost significant trend for sparrows with larger bibs to have more saturated bibs (r = 0.416, P = 0.061). Bib saturation, moreover, was negatively correlated with body condition, and uropygial gland size (Table 2). Notice that uropygial gland size was not correlated with body condition (r = 0.17, P = 0.45) or bib size (r = − 0.10, P = 0.66), and body condition was not correlated with bib size (r = 0.18, P = 0.43). When body condition, uropygial gland size, and bib size were included as predictors in a multiple regression model, a significant correlation between bib size and bib saturation emerged (partial r = 0.47, P < 0.01), meanwhile bib saturation remained significantly correlated with body condition (partial r = − 0.53, P < 0.01), and uropygial gland size (partial r = − 0.40, P = 0.01) (multiple R = 0.83, F_3,17 = 12.46, P < 0.001, adjusted R² = 0.63, tolerance > 0.92). On the other hand, bib hue was also positively correlated with uropygial gland size, but bib lightness was not correlated with any variable (Table 2). Saturation and hue were significantly correlated among themselves (r = 0.664, P = 0.001), but not with lightness (in both cases, r < 0.4, P > 0.10).

Effects of the experiment on immune system and uropygial gland size

The experimental treatment had a significant effect on the house sparrows’ immune system. Sparrows in the LPS-inoculated group showed a significant patagium swelling 4 h after the inoculation (average ± SD variation in patagium thickness: 0.37 ± 0.20 mm, t₁₁ = 6.29, P < 0.001), while the control group showed no swelling (change in thickness: −0.03 ± 0.06 mm, t₈ = 1.33, P = 0.22), difference in swelling between LPS- and PBS-inoculated sparrows being significant (t₁₉ = 6.16, P < 0.001). For all individuals considered together, the uropygial gland size decreased during the experiment (in average, −0.01 ± 0.02 mm³; paired t-test, t₂₀ = 2.77, P = 0.011); nevertheless, the change in uropygial gland size did not differ between treatments (t₁₉ = 1.55, P = 0.14).

Determinants of changes in bib colouration during the experiment

The change in uropygial gland size correlated positively with the change in hue (r = 0.61, P = 0.003). Initial body condition also was related to changes in colouration, as birds in better condition had a greater change in lightness (r = 0.44, P = 0.046) and trended (almost significantly) to have a greater change in saturation (r = 0.41, P = 0.066) and hue (r = − 0.40, P = 0.073) (Fig. 1). Bib size or change in body condition were not related to change in any colouration parameter (in all cases ∣r∣ < 0.4, P > 0.05).

In a more detailed analysis, I examined the effect of the treatment, change in uropygial gland size (hereafter, $\mathrm{\Delta }$UGS), and body condition on the change in colouration by using Linear Models in which, therefore, the effect of each variable was controlled for the effect of the other variables introduced in the model. After the experiment was carried out, I found a significant effect of body condition, treatment, and the interaction treatment × $\mathrm{\Delta }$UGS on change in bib saturation (Table 3). In this model, a significant effect of the treatment emerged: in sparrows inoculated with LPS, bib saturation did not change significantly (average change 0.45, with lower and upper 95% CI limits of −0.004 and 0.91). However, in the control group, the bib became less saturated, with an average change of −0.54 (95% CI limits: −1.06 and −0.026; significantly below zero (Nakagawa & Cuthill, 2007)). Moreover, there was a positive effect of body condition on change in saturation (β = 0.568). Regarding the interaction between treatment and change in uropygial gland size, experimental individuals showed no correlation between change in hue and $\mathrm{\Delta }$UGS (r = 0.25, P = 0.43), while in control birds there was an almost significant trend for a greater decrease in uropygial gland size accompanying greater change in saturation (r = − 0.66, P = 0.053; Fig. 2). These findings remained significant in models in which change in body condition or bib size were included (Tables 3B and 3C).

Change in hue was significantly determined by $\mathrm{\Delta }$UGS, with an almost significant effect of treatment, body condition, and the interaction treatment × $\mathrm{\Delta }$UGS (Table 3). More specifically, in the control group, there was a significant correlation between $\mathrm{\Delta }$UGS and change in hue (r = 0.72, P = 0.03). By contrast, in the experimental group, such a correlation was inexistent (r = 0.11, P = 0.73). However, these results were not robust when in the model I corrected for bib size (Table 3C). For the case of lightness, in a first analysis, I found a significant interaction between $\mathrm{\Delta }$UGS and treatment (Table S2), but this finding seemed to be caused by an outlier (Fig. S1). When the outlier was removed from the analyses, no significant result emerged (Table 3). Lastly, the change in lightness correlated with change in saturation (r = 0.44, P = 0.045), but not with the change in hue (r = − 0.38, P = 0.092). In turn, changes in saturation and hue were strongly correlated (r = − 0.70, P < 0.001).