General and Comparative Endocrinology 163 (2009) 184–192 Contents lists available at ScienceDirect General and Comparative Endocrinology journal homepage: www.elsevier.com/locate/ygcen Testosterone has a long-term effect on primary sex ratio of first eggs in pigeons—in search of a mechanism V.C. Goerlich *, C. Dijkstra, S.M. Schaafsma, T.G.G. Groothuis Behavioral Biology, University of Groningen, Kerklaan 30, P.O. Box 14, 9750 AA Haren, The Netherlands a r t i c l e i n f o Article history: Received 13 October 2008 Revised 18 December 2008 Accepted 13 January 2009 Available online 22 January 2009 Keywords: Avian Primary sex ratio Mechanism Follicle abortion Testosterone Maternal hormones Yolk hormones Pigeon a b s t r a c t Despite accumulating evidence that birds, in which females are the heterogametic sex, are able to manipulate primary offspring sex ratio, the underlying mechanism remains elusive. Steroid hormones, which govern female reproduction and are also accumulated by the developing follicle could potentially affect primary sex ratio by differential follicle development in relation to future sex and meiotic drive, or by sex specific influence on oocyte abortion or fertilization. So far, experimental results on the involvement of maternal testosterone (T) in offspring sex manipulation are ambiguous. To investigate the effect of T on primary sex ratio and elucidate underlying mechanisms, we elevated circulating T levels in female homing pigeons (Columba livia). During the course of the experiment females produced three clutches—before and during T implantation, and one year after implant removal. Intriguingly, first eggs, but not second eggs of T females were significantly male biased relative to sham-implanted controls. One year after cessation of the treatment the male bias was still present, indicating long-term effects on female reproductive physiology. T treatment did not affect maternal body condition, nor was body condition correlated with offspring sex ratio. Our data on timing of oviposition, lack of infertile eggs, and yolk weight indicate a possible role for sex specific follicle abortion, perhaps in combination with meiotic drive. However, despite T treatment elevating maternal plasma levels, egg yolk T concentrations did not differ between treatment groups and did not vary with embryo sex, suggesting that yolk T is not involved in meiotic drive. Ó 2009 Elsevier Inc. All rights reserved. 1. Introduction Various correlative and experimental studies have shown that birds and mammals manipulate their primary offspring sex ratio (proportion of sons at oviposition/conception) according to environmental circumstances (Pike and Petrie, 2003; Cameron, 2004; Alonso-Alvarez, 2006; Grant, 2007). Unlike mammals, in birds the female is the heterogametic sex bearing Z and W gametes and males are homologous for Z. Therefore, avian females could potentially bias the sex of their offspring at meiosis, i.e. before fertilization. Several possible mechanisms of embryo sex manipulation are currently under discussion and experimental evidence indicates that maternal hormones are likely to be involved in the process (Veiga et al., 2004; Correa et al., 2005; Pike and Petrie, 2006; Rutkowska and Cichon, 2006; Bonier et al., 2007). During egg formation steroid hormones from maternal circulation accumulate in the yolk and albumen, whether by active means or passive diffusion is yet to be determined (Groothuis and Schwabl, 2008). Maternal hormones in plasma or yolk could affect the primary sex ratio by differential follicle development and atresia * Corresponding author. Fax: +31 50 363 5205. E-mail address: V.Goerlich@rug.nl (V.C. Goerlich). 0016-6480/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.ygcen.2009.01.004 affecting future sex, possibly by skewed chromosome segregation during meiosis (meiotic drive), sex specific abortion of the oocyte (after meiosis), and sex specific fertilization (after ovulation). After oviposition, maternal hormones in the egg could affect the secondary sex ratio by sex specific allocation of these hormones influencing embryo or nestling survival (reviews: Pike and Petrie, 2003; Alonso-Alvarez, 2006). In this paper we focus on testosterone, which is one of the sex steroids that govern female reproductive physiology and shows a distinctive pattern with the ovulatory cycle (Doi et al., 1980; Jensen and Durrant, 2006). Furthermore, testosterone levels in the female and in her eggs both respond to changes in the females’ environment (review: Groothuis et al., 2005). Testosterone might therefore be the mechanism translating environmental factors to an adaptive adjustment of the sex ratio. Indeed, in some avian species, yolk androgen concentrations are correlated with embryo sex (Müller et al., 2002; Groothuis and Von Engelhardt, 2005; Badyaev et al., 2008) and in another species high plasma levels of maternal testosterone correlated positively with the proportions of sons produced (Pike and Petrie, 2005). Recently, three studies took an experimental approach by elevating maternal plasma testosterone concentrations by exogenous hormone administration but findings are not consistent. Female starlings (Sturnus unicolor) in the wild V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 produced more sons after testosterone implantation. Because the male bias persisted for three subsequent years the authors suggested that the effect of testosterone on offspring sex ratio might have been mediated by indirect effects on female social status (Veiga et al., 2004). In an experimentally controlled setting, zebra finches (Taeniopygia guttata) injected with testosterone enanthate (an ester of testosterone designed for pharmaceutical usage) produced eggs containing a higher proportion of male embryos compared to controls (Rutkowska and Cichon, 2006). In quail testosterone implantation did not affect offspring sex ratio. Instead, females skewed offspring sex ratio towards daughters in response to corticosterone manipulation (Pike and Petrie, 2006), an effect also found in white-crowned sparrows (Zonotrichia leucophrys; Bonier et al., 2007). The adrenal steroid corticosterone and the gonadal steroid testosterone have been shown to have an opposing relationship with offspring sex ratio (Pike and Petrie, 2006), whereas both hormones are thought to be negatively correlated (Moore et al., 1991). Hence, the involvement of testosterone in avian sex determination needs further investigation. Another question to solve is whether circulating levels of maternal plasma testosterone or, alternatively, testosterone concentration in the yolk influence the sex of the embryo, since testosterone treatment of egg laying females elevates the concentration of the hormone in the eggs, too (Groothuis and Schwabl, 2008). Moreover, the extent to which hormone treatment asserts its effect on primary sex ratio by sex specific effects on follicle growth, atresia, meiotic drive, abortion or fertilization remains elusive. In order to study the potential effects of testosterone on avian primary sex ratio in more detail and explore its underlying pathways we implanted female homing pigeons (Columba livia) with testosterone and measured maternal plasma and yolk hormones levels, the primary sex ratio and, with respect to potential mechanisms of sex manipulation, timing of oviposition, interval between egg oviposition, and yolk weight. We chose pigeons as study species for three reasons. First, they are determinate layers, typically producing clutches of only two eggs with fixed intervals, facilitating the analyzes of potential underlying mechanisms of sex ratio adjustment. Second, they have been shown to bias primary offspring sex ratio of especially first eggs, both in response to season (Dijkstra, in preperation), and under an experimental treatment consisting of food deprivation combined with continuous egg removal (Pike, 2005). Third, Pike (2005) hypothesized for this species a process of selective follicle abortion and ovulation of replacement follicles as the underlying mechanism for skewed offspring sex ratios. We also aimed at testing the potential long-lasting effect of testosterone on offspring sex ratio, as reported by Veiga et al. (2004), by analyzing the primary sex ratio produced by the same experimental females one year after cessation of the hormonal treatment. In order to disentangle indirect effects on social hierarchy and direct hormonal effects we kept the birds separated in breeding pairs during pair formation and egg production. 2. Methods 2.1. Animals and housing In August 2005 we obtained 58 pairs of adult homing pigeons from various breeders across The Netherlands. We matched the females over the treatment groups according to age (ranged from 1 to 12 years) and previous breeding experience. Experimental treatment alternated across the cage numbers. During the experiment pairs were housed in individual cages and were kept in a light dark schedule of LD14:10h. Cages were divided over three rooms (24, 8 and 26 cages, respectively); temper- 185 ature was set constant at 22 °C. Birds had ad libitum access to a mix of pigeon feed (Beyers Belgium (Beduco NV), Supralith mineral supplementary (Teurlings), grit (Beduco NV), P40 Pigeons grain (Kasper Faunafood)) and water. 2.2. Experimental design Experimental clutches were synchronized between treatment groups to control for seasonal effects on sex ratio (Dijkstra et al., 1990). To obtain a baseline sex ratio for every female, each pair was allowed to produce one un-manipulated clutch before testosterone (T) implantation. Starting 6 days after pair formation we checked the nests hourly to ascertain time of oviposition and the time interval between eggs. Fresh eggs were removed from the nests and replaced with artificial ones. Collected eggs were placed into an incubator at 38 °C for 72 h to obtain sufficient embryonic tissue for DNA sex determination. After incubation the eggs were frozen at 20 °C for further analyzes. After 16 days, when all except one female had completed their clutch, we collected the artificial eggs and removed the males from the breeding cages. Between the experimental clutches males were housed in outside aviaries. Four days after separation, we implanted half of the females with T filled tubes (T females), the other half with empty tubes (controls, C females; see below for details). Four days after implantation, we expected plasma testosterone levels to be elevated and re-introduced the males into the breeding cages. Each female was paired with the same male as during the first clutch. We collected, treated, and replaced eggs in the same way as in the previous batch. Three weeks after renewed completion of this second clutch we removed the implants from all females and moved the birds to unisex outdoor aviaries. To determine possible long-term effects of T elevation on offspring sex ratio we continued the experiment one year later in August 2006. We moved females to the breeding cages again and paired them with the same males as in previous clutches. Clutch completion of all females took 20 days. The eggs of this third clutch were collected, incubated and stored just as in the earlier clutches. For sample sizes see Table 1. 2.3. Implantation and blood sampling Females were implanted with 14 mm Silastic tubes (RaumedicÒ, inner diameter 1 mm, outer diameter 3 mm), sealed at both ends with silicon glue (BisonÒ). This tube sizing has been successfully used to elevated plasma hormone levels also in previous studies (Duttmann et al., 1999; Ros, 1999) and caused an elevation within the physiological range in a pilot study comparing 3 implant sizes in female homing pigeons (Goerlich et al., unpublished data). In the experiment, T females received tubes filled with 10 mm of crystalline testosterone (Sigma Chemical Co., St. Louis, MO, catalog No. 86500), C females received empty tubes. After applying local anesthetics (XylocaineÒ, 10%), a small incision was made on the right flank under the wing and tubes were inserted subcutaneously. The incision was sealed with wound glue (Liquid Protect, HansaplastÒ). The next day we checked wounds for proper healing and presence of the implant. We excluded one bird from the analyzes as the implant could not be felt and was also not found at removal of the implants. Females were blood sampled three times during the course of the experiment to determine plasma T concentrations. During the first clutch we took blood on day 9 after pair formation, but as at that time some females had already laid their first eggs we decided to blood sample on day 5 after pair formation during the second and third clutch. We also chose this point in time as in pigeons the phase of rapid yolk deposition takes approximately 6.5 186 V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 Table 1 Overview of number of experimental pairs and fertile eggs produced during the three clutches. 1 2 3 Clutch sequence Treatment group No. of pairs pre-implant pre-implant post-implant post-implant post-implant post-implant Control Testosterone Control Testosterone Control Testosterone 27 27 26 27 21 22 days (Birrenkott et al., 1988) and blood sampling females during the course of follicle development would be the most appropriate for testing whether yolk T concentrations reflect maternal plasma levels. We took approximately 300 ll blood from the ulnar vein using a 25 G needle and heparinized syringe. Immediately afterwards, blood samples were centrifuged for 10 min at 4000 rpm, the plasma was separated from the pellet and frozen at 20 °C until hormone analysis (see below). No. of first eggs No. of second eggs laid sexed infertile laid sexed infertile 26 27 25 25 21 22 19 25 20 23 11 14 6 1 4 2 8 6 24 26 24 21 21 22 18 20 20 21 17 19 6 6 3 0 3 3 (2000 rpm, 4 °C) the tubes subsequently. Then samples were snap frozen by placing the tubes into a mixture of ethanol and dry ice and the organic phase was decanted into fresh tubes. After drying under a stream of nitrogen we added 1 ml 70% Methanol, vortexed the tubes and stored them overnight at 20 °C. The next morning samples were centrifuged for 5 min, 2000 rpm at 4 °C. The Methanol phase was decanted into fresh tubes and dried off under nitrogen. We re-suspended the pellet in PBS buffer 1:1 according to the original volume of the plasma sample. 2.4. Other measurements We took measurements of body mass (to the nearest 0.1 g) and tarsus length (to the nearest 0.1 mm) of all birds before and after the second clutch and before the third clutch. We used the standardized residuals of the linear regression of female body mass on tarsus length (3 measurements of 50 females, Wald v21 = 7.340, p = 0.0067) as an index for body condition. Freshly laid eggs were weighed (to the nearest 0.01 g), measured (width and length to the nearest 0.01 mm) and subsequently placed into an incubator at 38 °C for 72 h, after which eggs were stored at 20 °C for analyzes of sex, yolk weight and yolk T concentrations. All experimental procedures were carried out under approval of the animal experimentation committee of the University of Groningen (license DEC 4347A). 2.5. Molecular sexing Eggs were allowed to thaw for ca. 3 min. After removal of the shell, we collected the embryonic disc and the yolk, which was still clearly separated from the albumen after the 3 days incubation period. Yolk was weighed to the nearest 0.001 g and frozen until hormone analysis (see below). Embryos were sexed following the methods of Walsh et al. (1991) and Griffiths et al. (1998). Embryonic DNA was extracted with Chelex and the two homologous genes CHD-W (females) and CHD-Z (males and females) on the sex chromosomes were amplified by polymerase chain reaction (PCR) using the primers P2 (50 TCTGCATCGCTAAATCCTTT-30 ) and P8 (50 -CTCCCAAGGATGAGRAA YTG-30 ). PCR products were run on a 2.5% agarose gel containing 0.005% ethidiumbromide. Control samples of known females (twobands) and males (one band) were incorporated in every PCR and on every gel. Eggs which had not developed a visible embryo after 3 days of incubation were considered as infertile (Table 1). 2.6. Hormone analysis 2.6.1. Plasma testosterone extraction After measuring plasma volume of the samples and two internal controls, we added ca. 5000 cpm radioactive labeled T (PerkinElmer Life and Analytical Sciences BV) to all samples to calculate for losses due to the extraction procedure (recovery). After 1 h incubation we added 2.5 ml of diethyl ether/petroleum benzine, 70:30 (vol/vol) to all samples, vortexed and centrifuged 2.6.2. Yolk testosterone extraction All yolks were homogenized with 2.5 ml water. We used 400 mg of the mixture for the extraction of T. We incorporated two controlyolk pools in every extraction to establish inter-assay variation. The same extraction protocol was followed as for plasma samples, except the step of adding diethyl ether/petroleum benzine and snap freezing was repeated. Yolk samples from first eggs were re-suspended in 250 ll PBS buffer. Due to higher T concentrations we added 500 ll PBS to yolk samples from second eggs. RIA values (T/ ml) were back-calculated per mg pure yolk used in the homogenate and corrected for initial yolk mass. Recoveries for plasma and yolk samples were on average 70.3% and 72.0%, respectively, samples with a recovery lower than 60% were extracted again. 2.6.3. Radioimmunoassay T levels in plasma and yolk samples were measured by radioimmunoassay (RIA). We used a commercial kit (ActiveÒTestosterone Coated-Tube RIA DSL-4000 kit, Diagnostic Systems Laboratories) with a sensitivity of 0.08 ng/ml T. The standard curve was prepared by diluting a home made T standard (40 ng/ml) with PBS buffer to obtain following concentrations: 20, 10, 5, 2.5, etc., down to 0.156 ng/ml for yolk and 10 to 0.078 ng/ml for plasma RIAs. We validated the assay for our experimental species by measuring serial dilutions of several plasma and yolk samples (both first and second eggs). Dilution curves ran parallel to the standard curve, indicating that the kits are suitable for measuring T in pigeon samples without disturbance by other substances. Two controls containing known low and high levels of T were incorporated in all RIAs to establish inter-assay variation. The average inter-assay coefficient of variation (CV, (stddev(avrg)/avrg)100) was 12.6% for yolk, and 15.7% for plasma. Average intra-assay CV (based on own control duplicates) was 2.57% for yolk, and 2.65% for plasma. Yolk and plasma samples were split over three RIAs each. All yolk samples were assayed in duplicates. Based on the high assay precision and the low intra CV we feel confident measuring plasma samples of low volume in single. 2.7. Statistical analyzes Offspring sex ratio was analyzed using multilevel random intercept models for binary response data with a logit link function (generalized linear model (GLM); MlwiN 2.02). We applied hierarchical models to prevent pseudo-replication when analyzing V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 intra-individual changes and nested egg sequence within clutch sequence within female. Estimation procedure was specified as restricted iterative generalized least squares and second-order penalized quasi-likelihood approximation (Rasbash et al., 2004). We tested the significance of the treatment effect using the Wald statistic, which follows a v2 distribution. Starting with a full model, non-significant interactions and terms were sequentially eliminated. Wald v2 and p values are given before removal of the factor. In all tests significance levels were set to an a < 0.05 (two-tailed). Based on the hypotheses of Emlen (1997) and Pike (2005), and the available data in pigeons (Riddle, 1917; Pike, 2005, Dijkstra (in preperation)) we expected a priori specifically first eggs to show a biased offspring sex ratio and, following Pike (2005), analyzed first and second eggs separately. In the first model we included both the pre- and post-implant clutch (1&2) to allow a within female comparison, and tested the interaction of treatment (control or testosterone) and clutch sequence as a predictor of offspring sex ratio. Next we analyzed only post-implant clutches (2&3) to test whether the treatment effect on offspring sex ratio persisted until one year after removal of the implant. Due to the lower sample size during clutch 3, testing the change in sex ratio between clutch 1 and 3 lacked statistical power. Therefore we pooled data from both post-implant clutches and, incorporating all clutches and accounting for the within female comparison, tested the interaction of treatment and clutch phase (preimplantation (1), versus post-implantation (2&3)). Female ID and clutch sequence were set as hierarchical level. Measurements of yolk weight and yolk T were taken only during the second clutch. Both variables were distributed normally and analyzed using a model with female ID and egg sequence as hierarchical levels, and treatment, embryo sex and egg sequence and their interactions as categorical predictors. The same analyzes were performed on egg weight and maternal condition (with clutch sequence as additional level), which was measured during all three clutches. Non-normally distributed variables (plasma T, latency to oviposition, and egg interval) were analyzed using Mann–Whitney-U tests (SPSS 14.0). 187 p = 0.89, Fig. 1). During the second clutch, T implants effectively elevated plasma T levels within the physiological range (MWUTest, U = 191, p = 0.007, Fig. 1). One year after removal of the implants we did not find a statistical difference in plasma T levels between experimental groups (MWU-Test, U = 177, p = 0.27, Fig. 1). 3.2. Offspring sex ratio We found no causal relationship between natural variation in plasma testosterone levels and offspring sex (clutch 1, model corrected for latency to oviposition: egg 1, Wald v21 = 0.088, p = 0.77; egg 2, Wald v21 = 0.76, p = 0.38). T treatment during the second clutch resulted in male biased first eggs of T females compared to the pre-implantation clutch and C females (interaction treatmentclutch sequence: p = 0.027, Table 2 a; Fig. 2). As predicted, the sex ratio of second eggs did not significantly differ between experimental groups (p = 0.856). During both post-implantation clutches (2&3) T females produced a higher proportion of male embryos in their first eggs than C females (treatment: p = 0.047, Table 2b; Fig. 2), despite cessation of treatment almost one year before clutch 3. Again this was not the case in second eggs (p = 0.501). Since offspring sex ratio of clutches 2 and 3 did not significantly differ (Table 2b, treatmentclutch sequence: p = 0.84/p = 0.869) we pooled the data from both post-implant clutches. When comparing them to the pre-implant clutch we found a significant interaction between T treatment and clutch phase (pre-/post-implantation: p < 0.039, Table 2c; Fig. 2) indicating again a long-lasting effect of T on the male biased sex ratio in first eggs only (second eggs p = 0.66). 3.3. Egg fertility T treatment did not affect clutch size (MWU-Test, U = 301, p = 0.153), nor the proportion of infertile eggs (first eggs Wald v21 = 0.176, p = 0.68, second eggs Wald v21 = 0.069, p = 0.79, Table 1), indicating that these variables cannot account for the shift in sex ratio due to T treatment. 3. Results 3.4. Yolk testosterone 3.1. Plasma testosterone levels T levels did not statistically differ between experimental groups during the first clutch (pre-implantation) (MWU-Test, U = 156, Regarding the potential role of yolk hormones in sex determination and the observed male bias in first eggs of T females, we tested for sex and treatment related patterns of hormone allocation. In spite of the difference in plasma T levels, yolk T concentrations did not significantly differ between treatment groups, (Wald v21 = 0.034, p = 0.85, Fig. 3), nor did they differ between male and female eggs (Wald v21 = 0.212, p = 0.65, Fig. 3). Also the interactions with egg sequence and treatment did not explain a significant part of the variation in yolk T levels (Fig. 3). Intriguingly all clutches showed the same pattern of T levels with second egg yolks containing on average 2.7 times higher T levels than first egg yolks (egg sequence: Wald v21 = 174.262, p < 0.001, Fig. 3). Yolk T concentrations of first and second eggs within a clutch were not correlated (Pearson correlation, n = 36, r = 0.123, p = 0.48). Female plasma T levels (log transformed) were not correlated with yolk T levels (Pearson correlation, first egg n = 41, r = 0.025, p = 0.88; second eggs, n = 37, r = 0.52, p = 0.76). 3.5. Yolk weight Fig. 1. Plasma testosterone levels of control (white) and testosterone (grey) females (box plots showing the median, interquartile range, and top/bottom quartile) during the three experimental clutches. We analyzed yolk weight based on the hypothesis that pigeons abort follicles and ovulate premature replacement follicles. Embryo sex did not predict yolk weight (Wald v21 = 0.862, p = 0.35, Fig. 4), but yolk weight varied significantly with the interaction of treatmentegg sequence (Wald v21 = 4.015, p = 0.045). Post hoc analysis showed that in C-clutches first egg yolks were signifi- 188 V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 Table 2 Overview of GLMs testing offspring sex ratio as dependant variable with female treatment (T or C), clutch sequence (1–3), and clutch phase (pre-/post-implantation) as predictors. Results of significant predictors are indicated in bold. Dataset clutch sequence First eggs B Second eggs SEM Wald v21 p B SEM Wald v21 p (a) 1 vs. 2 Treatmentclutch seq. Treatment Clutch seq. 2.09 0.78 0.68 0.95 0.74 0.71 4.88 1.10 0.92 0.027 0.295 0.337 0.23 0.19 2.05 1.28 1.15 0.61 0.03 0.03 11.48 0.856 0.869 0.001 (b) 2 vs. 3 Treatmentclutch seq. Treatment Clutch seq. 0.22 1.05 0.05 1.07 5.29 0.53 0.04 3.94 0.01 0.840 0.047 0.933 0.16 0.34 0.90 0.96 0.50 0.47 0.03 0.45 3.62 0.869 0.501 0.057 (c) 1 vs. 2&3 (pre- vs. post-) Treatmentpre-/post-implant Treatment Pre-/post-implant 1.65 0.59 0.59 0.80 0.65 0.60 4.22 0.81 0.99 0.039 0.368 0.321 0.36 0.26 0.56 0.81 0.46 0.40 0.19 0.32 1.97 0.660 0.575 0.160 Fig. 2. Primary offspring sex ratio (proportion of male embryos, mean ± SE) for first and second eggs produced by control (white) and testosterone (black) females during the three experimental clutches. Fig. 4. Egg and yolk weight (mean ± SE) of first and second eggs bearing female or male embryos produced by control (white) and testosterone (black) females. Yolk weight was collected during the experimental phase of hormone treatment (clutch 2); egg weight is pooled from all three clutches. not significantly differ between treatment groups (Wald v21 = 1.553, p = 0.21). 3.6. Egg weight We also analyzed egg weight of the second clutches for effects of T treatment or sex specific variation. Neither treatment (Wald v21 = 0.088, p = 0.77), nor its interactions with embryo sex (Wald v21 = 1.842, p = 0.17) or egg sequence (Wald v21 = 0.770, p = 0.38) predicted egg weight. Egg weight did vary significantly with egg sequence by embryo sex (Wald v21 = 5.206, p = 0.023) and this interaction was still significant when pooling data from all three clutches (egg sequenceembryo sex, Wald v21 = 7.975, p = 0.005). First eggs containing female embryos were heavier than first eggs containing male embryos, whereas the opposite pattern was present in the second eggs (Fig. 4). Fig. 3. Yolk testosterone levels (mean ± SE) measured in yolks of first and second eggs bearing female or male embryos produced by control (white) and testosterone (black) females during the experimental phase of hormone treatment (clutch 2). cantly heavier than second egg yolks (Wald v21 = 7.345, p = 0.007, Fig. 4) which was not the case in T-clutches (Wald v21 = 0.006, p = 0.94, Fig. 4). However, yolk weight of second eggs alone did 3.7. Timing of oviposition Since delayed oviposition or prolonged egg intervals may indicate follicle abortion, we analyzed the effect of treatment on timing of oviposition for the three clutches separately. Egg interval however was only available during the treatment phase (clutch 2). During the second clutch, T females initiated their clutches (calculated as days between pair formation and first egg) significantly later V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 Fig. 5. Latency to oviposition during clutch 2 calculated as the number of days between pair formation and oviposition of the first egg. Bars represent number of control (white) and testosterone (black) females for each latency. 189 We did not find a causal relationship between natural testosterone levels and offspring sex. Testosterone fluctuates during the ovulatory cycle and our measurement is a limited snapshot during the course of egg production. It is still unclear at which stage of egg formation/sex determination high T levels actually induce the shift in sex ratio. During our pilot study implants elevated plasma levels over at least two weeks (Goerlich et al., unpublished data), therefore T females very likely have had constantly elevated testosterone levels during the course of egg production. Based on this pilot study we carefully scaled testosterone manipulation to ensure elevated levels within physiological range and avoid negative effects on female fecundity (Rutkowska et al., 2005). Under the ad lib food conditions, testosterone treatment did not alter female body condition nor was maternal body condition correlated with embryo sex. Other studies have reported that maternal body condition does not only relate to primary offspring sex ratio (Nager et al., 1999; Clout et al., 2002; Pike, 2005; Pike and Petrie, 2005) but is also reflected by levels of circulating plasma testosterone (Verboven et al., 2003; Pike and Petrie, 2005). Our results indicate that circulating testosterone levels can act on sex determination independently of maternal condition. Additional results on timing of oviposition, yolk weight, and yolk testosterone levels now allow us to further explore the underlying mechanism of avian sex ratio manipulation. 4.1. Sex specific fertilization or embryo mortality than C females (MWU-Test, U = 130, p < 0.0001, Fig. 5) although latency to oviposition did not predict the observed bias in sex ratio in first eggs from T females (Wald v21 = 0.126, p = 0.72). The delayed clutch initiation was not apparent during the first (un-manipulated) clutch (MWU-Test, U = 317, p = 0.70) and also lacking one year later during the third male biased clutch (MWU-Test, U = 223.5, p = 0.85). The interval between egg one and two did not differ between treatment groups (clutch 2, C-clutches: n = 23, median 44.75 h, interquartile range 2 h; T-clutches: n = 20, median 44.88 h, interquartile range 1 h; MWU-Test, U = 196, p = 0.41). 3.8. Maternal condition Several studies have shown maternal condition to predict offspring sex ratio (Nager et al., 1999; Clout et al., 2002; Pike and Petrie, 2005). We found no effects of T treatment on female body condition (clutch 1&2, treatmentclutch sequence: Wald v21 = 0.165, p = 0.69; clutch 2&3, treatment: Wald v21 = 0.107, p = 0.74) nor did female body condition during our ad lib food conditions predict offspring sex ratio (clutch 1, first eggs Wald v21 = 2.035, p = 0.15, second eggs, Wald v21 = 01.586, p = 0.21; clutch 2, first eggs Wald v21 = 0.008, p = 0.93, second eggs, Wald v21 = 0.318, p = 0.57; clutch 3, first eggs Wald v21 = 1.058, p = 0.30, second eggs, Wald v21 = 0.046, p = 0.83). 4. Discussion Primary offspring sex ratio of first eggs was significantly shifted towards males after the female parent received a testosterone implant whereas sex ratio of second eggs was not affected. This effect was maintained even one year after cessation of treatment. Our results from the homing pigeon are consistent with two of the three studies in which experimental manipulation of testosterone in the breeding female skewed primary offspring sex ratio towards males as well, although under different experimental conditions (Starlings: Veiga et al., 2004; Zebra finches: Rutkowska and Cichon, 2006; but see Pike and Petrie, 2006). We cannot exclude testosterone manipulation affecting offspring sex ratio via influence on corticosterone levels (Bonier et al., 2007) and suggest future studies measuring plasma concentrations of both hormones. The proportion of fertile eggs was similar between treatment groups. It is therefore unlikely that the male bias in the testosterone implanted group was mediated by mechanisms such as sex specific fertilization of oocytes or increased embryo mortality (Pike and Petrie, 2003; Alonso-Alvarez, 2006). 4.2. Yolk testosterone and meiotic drive Non-random segregation of sex chromosomes during meiosis has been suggested as a mechanism underlying skewed offspring sex ratios (Howe, 1977; Krackow, 1995; Correa et al., 2005; Rutkowska and Cichon, 2006). Meiotic drive would be a very economic way to bias embryo sex, avoiding costs due to loss of investment in follicle growth and delayed egg production. Approximately 1 to 2 h before ovulation the first meiotic division is completed and either the W or the Z chromosome is retained in the ovum whereas the other sex chromosome is drawn into the inactive polar body (Rutkowska and Badyaev, 2008). The germinal disk, containing the nucleus of the oocyte, is located at the periphery of the follicle, embedded in the outermost yolk layer. Surrounding steroid hormone concentrations might influence the orientation of the meiotic spindle and drive chromosome segregation in a certain direction (Krackow, 1995; Rutkowska and Badyaev, 2008). We measured total yolk testosterone after three days of incubation and did not find a relationship with embryo sex. Our data therefore strongly suggest that sex ratio manipulation, although possibly caused by skewed chromosome segregation, is not directly mediated by total yolk testosterone concentration. It is however conceivable that the germinal disk is mainly exposed to yolk hormones in the outer yolk layer, which we did not analyze separately, or to actual plasma levels of the female, due to the strong vascularization of the follicular wall (Johnson and Whittow, 2000). Furthermore, elevated testosterone levels in maternal circulation may affect deposition of other yolk components that may affect meiotic drive. Therefore we cannot support but also not safely rule out meiotic drive as the mechanism underlying the male biased offspring sex ratio. Moreover, the potential mechanisms discussed below such as sex specific follicle growth or atresia, require that the female can recognize the sex of the follicle before meiosis. 190 V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 This paradox can be solved by assuming that both follicle growth and segregation of sex chromosomes are influenced by the same factor, suggesting an important role for meiotic drive. Still, it remains unclear why in the case of meiotic drive only the sex of the first egg was male biased. Interestingly, the similarity in concentration of yolk testosterone between clutches of both treatment groups despite the difference in plasma levels is in line with the suggestion that yolk concentrations of androgens can be regulated independently from circulating plasma levels (Groothuis and Schwabl, 2008). 4.3. Sex specific follicle growth By measuring the thickness of yolk layers, Badyaev et al. (2005) presented evidence that in House finches (Carpodacus mexicanus), male and female oocytes strongly differ in their rate of growth. Overcoming the ovarian hierarchy by changing the order of ovulation according to follicle sex might ultimately result in a sex biased pattern at oviposition which is observed in several species (Ankney, 1982; Dijkstra et al., 1990; Nager et al., 1999). However, differential follicle growth might only occur in species in which several follicles grow simultaneously. If the time lag between follicles is small and ovulation occurs daily (e.g. in the house finch (Badyaev et al., 2005)) follicles could overtake each other during the period of rapid growth. In pigeons the time lag between ovulation of the two follicles is 44 h. Birrenkott et al. (1988) used both the dichromate staining and Sudan feeding technique to visualize yolk layers in pigeon eggs and found no significant differences in the phase of rapid follicle growth (ca. 156 h) comparing first and second yolks within clutches. Nevertheless, incorporating embryo sex into the analyzes might have yielded a different result. To this day literature considers the family of Columbidae as determinate layers, i.e. independent of external cues (e.g. addition or removal of eggs) pigeons typically produce a clutch of two eggs, but not more (Blockstein, 1989; Haywood, 1993). It is thought that only two follicles are recruited into the rapid yolking phase for further maturation; autopsies of breeding ring doves (Streptopelia risoria) revealed only two large yolky follicles present in the ovary prior to laying (Cuthbert, 1945). If the male bias in first eggs was caused by selecting second follicles to grow faster and ovulate first in case the first one was of the female sex, the female sex should be over-represented in second eggs that are partly the delayed first follicles of the wrong sex. Although second eggs were slightly female biased, this was also the case for both first and second eggs of the control group, making this hypothesis unlikely. 4.4. Sex specific follicle atresia Ovarian dynamics are defined by follicle growth and atresia (Johnson, 1996). Cohorts of early stage follicles are selected to grow, and during this process some follicles become atretic and do not develop further. Specific atresia of follicles, in relation to future sex could result in an overrepresentation of follicles from a certain sex (males in case of T females) to reach ovulatory size. However, this would again not explain why the male bias was restricted to first eggs. hierarchy presumably would result in laying gaps and possible negative effects on nestling survival. Based on his study in homing pigeons, in which the combined treatment of food deprivation and continuous egg removal resulted in a female biased primary offspring sex ratio, Pike (2005) modified the ‘‘first follicle abortion” hypothesis. According to his selective abortion model, pigeons might also abort second follicles of the undesired sex and, despite being smaller, ovulate follicles further down in the size hierarchy, thereby avoiding laying gaps. It is however doubtful whether pigeons are manipulating embryo sex of second eggs. In our study primary sex ratio of second eggs did not differ between treatment groups during neither of the experimental clutches and also our data on yolk weight did not give indication for abortion of follicles destined for second eggs. The occurrence of skewed sex ratio was similar in the pigeon study by Pike (2005) in which the female bias was clearly present in first eggs and statistically significant, but less strong in second eggs and actually not significant (a one tailed p value of 0.031 was reported accidentally as two-tailed). Further support for biased sex ratio amongst first eggs comes from feral rock pigeons which show a distinct seasonal shift in primary offspring sex ratio especially in first eggs (Dijkstra, in preperation). If T females aborted first female follicles, the second follicle would still need to complete 44 h of development before reaching ovulatory size. This time lag would explain the delayed oviposition of T females. However, this delay was not present in the third clutch while females still produced male biased first eggs, which is inconsistent with this hypothesis. Assuming that in homing pigeons indeed only two follicles mature at the beginning of the ovulatory cycle, the abortion of first follicles would have likely resulted in one of the following three scenarios: (1) T females mainly produce one egg clutches; we found no effect of testosterone treatment on clutch size. (2) The abortion of the first follicle induces the growth of a third replacement follicle that was not yet recruited and would have to undergo the whole phase of rapid yolk deposition (156 h); this would result in laying gaps between first and second eggs, which were absent in our experiment. (3) The replacement follicle would be ovulated prematurely after only 88 h of growth which would result in substantially lighter yolks in second eggs; yolk weight in second eggs did not differ between treatment groups. Thus, the principle of follicle abortion after meiosis requires that already in an early stage of the ovulatory cycle more than two follicles are selected to enter the rapid yolking phase. This seems to us a likely possibility and measurement of ovarian dynamics and follicle sizes in female homing pigeons, being in the situation which induced sex ratio biases, are currently in progress (Goerlich et al., in preperation). 4.6. Sex specific provisioning Our data on sex and clutch order specific variation in egg weight, but not yolk weight, suggest that, after meiosis, (i.e. after yolk provisioning), the mother is indeed able to detect the sex of her follicle. Given this, the mother could either abort the undesired sex or provision the developing embryo sex specifically with different amounts of albumen (Müller et al., 2004). 4.5. Sex specific follicle abortion after meiosis 4.7. Long-lasting effect If the most mature follicle has the ‘‘wrong” sex after meiosis it could be aborted, thereby allowing the next follicle in the ovarian hierarchy to become ovulated instead, increasing the chance on an oocyte of the desired sex. The process of follicle abortion would not require meiotic drive to result in a skewed offspring sex ratio. Emlen (1997) proposed follicle abortion underlying biased sex ratio of first eggs only, as aborting an intermediate follicle in the sequential Intriguingly, one year after testosterone implants were removed, first eggs of T females still contained a higher proportion of sons than C females, despite no significant differences in plasma testosterone concentrations between groups. Veiga et al. (2004) explained their long-term effect of testosterone on sex ratio by the persistent effect of the hormone on female V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 social dominance. We kept our breeding pairs in separate cages not only during the treatment but also during egg production almost 1 year later. Therefore birds could not establish rank orders and did not compete over mating partners or nest boxes during the production of all three clutches. After removal of the implants females were group-housed. But because they were kept in unisex groups under ad lib food conditions, we did not observe aggressive encounters. Nevertheless we cannot exclude the possibility that the testosterone treatment had a carry over effect on the third clutch by affecting female social status. It is conceivable that testosterone manipulation re-organized the females’ ovarian system in the long-term (Arnold and Breedlove, 1985). The density of ovarian or neuronal steroid receptors might have been up-regulated by the earlier hormone treatment, so that T females became more sensitized towards naturally occurring elevations of testosterone during the ovarian cycle (Drummond, 2006; Ball and Balthazart, 2008). To conclude, regarding potential mechanisms underlying avian sex ratio manipulation we found no indication that maternal testosterone deposition to the yolk affects embryonic sex. The male bias being restricted to only first eggs and the sex differences in egg weight support the idea that the mother is able to recognize the sex of the follicle after meiosis. Abortion of first follicles bearing the undesired sex is therefore a possible pathway to bias offspring sex. However, in the absence of laying gaps or one egg clutches abortion is only conceivable if pigeons develop more than two follicles per ovulatory cycle (Pike, 2005). To fully understand the dynamics of follicle abortion and maturation in the pigeon further investigation of follicle growth dynamics is necessary. We are confident that our results corroborate the involvement of maternal testosterone in avian sex determination and open a promising pathway for further research determining the underlying mechanism. Acknowledgments We are thankful for the valuable comments of Frances Bonier and one anonymous reviewer which improved this paper. We thank Bonnie de Vries and Maarten Lasthuizen for their help in the isotopes lab, Sjoerd Veenstra, Roelie Veenstra-Wiegman, Saskia Helder, and Monique Huizinga for support in taking care of the pigeon colony, and Gerard Overkamp for his practical help. References Alonso-Alvarez, C., 2006. Manipulation of primary sex ratio: an updated review. Avian Poult. Biol. Rev. 17, 1–20. Ankney, C.D., 1982. Sex-ratio varies with egg sequence in lesser snow geese. Auk 99, 662–666. Arnold, A.P., Breedlove, S.M., 1985. Organizational and activational effects of sex steroids on brain and behavior—a reanalysis. Horm. Behav. 19, 469–498. Badyaev, A.V., Schwabl, H., Young, R.L., Duckworth, R.A., Navara, K.J., Parlow, A.F., 2005. Adaptive sex differences in growth of pre-ovulation oocytes in a passerine bird. Proc. R. Soc. B 272, 2165–2172. Badyaev, A.V., Young, R.L., Hill, G.E., Duckworth, R.A., 2008. Evolution of sex biased maternal effects in birds. IV. Intra-ovarian growth dynamics can link sex determination and sex specific acquisition of resources. J. Evol. Biol. 21, 449– 460. Ball, G.F., Balthazart, J., 2008. Individual variation and the endocrine regulation of behaviour and physiology in birds: a cellular/molecular perspective. Philos. Trans. R. Soc. Lond. Biol. 363, 1699–1710. Birrenkott, G.P., Shoop, M.A., Cooper, K., Wiggins, M., 1988. Ovarian folliculargrowth and maturation in the domestic pigeon and Guinea fowl (NumidaMeleagris). Poult. Sci. 67, 1783–1786. Blockstein, D.E., 1989. Crop milk and clutch size in mourning doves. Wilson Bull. 101, 11–25. Bonier, F., Martin, P.R., Wingfield, J.C., 2007. Maternal corticosteroids influence primary offspring sex ratio in a free-ranging passerine bird. Behav. Ecol. 18, 1045–1050. Cameron, E.Z., 2004. Facultative adjustment of mammalian sex ratios in support of the Trivers–Willard hypothesis: evidence for a mechanism. Proc. R. Soc. B 271, 1723–1728. 191 Clout, M.N., Elliott, G.P., Robertson, B.C., 2002. Effects of supplementary feeding on the offspring sex ratio of kakapo: a dilemma for the conservation of a polygynous parrot. Biol. Conserv. 107, 13–18. Correa, S.M., Adkins-Regan, E., Johnson, P.A., 2005. High progesterone during avian meiosis biases sex ratios toward females. Biol. Lett. 1, 1744–9561. Cuthbert, N.L., 1945. The ovarian cycle of the ring dove (Streptopelia Risoria). J. Morphol. 77, 35. Dijkstra, C., in preperation. Seasonal variation of primary and secondary sex ratio in pigeons. Dijkstra, C., Daan, S., Buker, J.B., 1990. Adaptive seasonal-variation in the sex-ratio of kestrel broods. Funct. Ecol. 4, 143–147. Doi, O., Takai, T., Nakamura, T., Tanabe, Y., 1980. Changes in the pituitary and plasma LH, plasma and follicular progesterone and estradiol, and plasma testosterone and estrone concentrations during the ovulatory cycle of the quail (Coturnix coturnix japonica). Gen. Comp. Endocrinol. 41, 156–163. Duttmann, H., Dieleman, S., Groothuis, T.G.G., 1999. Timing of moult in male and female shelducks Tadorna tadorna: effects of androgens and mates. Ardea 87, 33–39. Drummond, A.E., 2006. The role of steroids in follicular-growth. Reprod. Biol. Endocrin. 4, 16. Emlen, S.T., 1997. When mothers prefer daughters over sons. Trends Ecol. & Evol. 12, 291–292. Goerlich, V.C., Dijkstra, C., Groothuis, T.G.G. in preperation. Follicle abortion – a mechanism for primary sex ratio adjustment in pigeons?. Grant, V.J., 2007. Could maternal testosterone levels govern mammalian sex ratio deviations? J. Theor. Biol. 246, 708–719. Griffiths, R., Double, M.C., Orr, K., Dawson, R.J.G., 1998. A DNA test to sex most birds. Mol. Ecol. 7, 1071–1075. Groothuis, T.G.G., von Engelhardt, N., 2005. Investigating maternal hormones in avian eggs: measurement, manipulation, and interpretation. Ann. N Y Acad. Sci. 1046, 168–180. Groothuis, T.G.G., Müller, W., von Engelhardt, N., Carere, C., Eising, C.M., 2005. Maternal hormones as a tool to adjust offspring phenotype in avian species. Neurosci. Biobehav. R. 29, 329–352. Groothuis, T.G.G., Schwabl, H., 2008. Hormone-mediated maternal effects in birds: mechanisms matter but what do we know of them? Philos. Trans. R. Soc. Lond. Biol. 363, 1647–1661. Haywood, S., 1993. Sensory and hormonal-control of clutch size in birds. Q. Rev. Biol. 68, 33–60. Howe, H.F., 1977. Sex-ratio adjustment in common grackle. Science 198, 744–746. Jensen, T., Durrant, B., 2006. Assessment of reproductive status and ovulation in female brown kiwi (Apteryx mantelli) using fecal steroids and ovarian follicle size. Zoo. Biol. 25, 25–34. Johnson, A.L., 1996. The avian ovarian hierarchy: a balance between follicle differentiation and atresia. Poult. Avian Biol. Rev. 7, 99–110. Johnson, A.L., Whittow, G.C., 2000. Reproduction in the female. In: Sturkie’s avian physiology, fifth ed. Academic Press, San Diego, pp. 569–596. Krackow, S., 1995. Potential mechanisms for sex ratio adjustment in mammals and birds. Biol. Rev. 70, 225–241. Moore, M.C., Thompson, C.W., Marler, C.A., 1991. Reciprocal changes in corticosterone and testosterone levels following acute and chronic handling stress in the tree lizard, Urosaurus-Ornatus. Gen. Comp. Endocrinol. 81 (2), 217– 226. Müller, W., Eising, C.M., Dijkstra, C., Groothuis, T.G.G., 2002. Sex differences in yolk hormones depend on maternal social status in Leghorn chickens (Gallus gallus domesticus). Proc. R. Soc. B 269, 2249–2255. Müller, W., Eising, C.M., Dijkstra, C., Groothuis, T.G.G., 2004. Within-clutch patterns of yolk testosterone vary with the onset of incubation in black-headed gulls. Behav. Ecol. 15, 893–897. Nager, R.G., Monaghan, P., Griffiths, R., Houston, D.C., Dawson, R., 1999. Experimental demonstration that offspring sex ratio varies with maternal condition. Proc. Natl. Acad. Sci. USA 96, 570–573. Pike, T.W., Petrie, M., 2003. Potential mechanisms of avian sex manipulation. Biol. Rev. Camb. Philos. Soc. 78, 553–574. Pike, T.W., 2005. Sex ratio manipulation in response to maternal condition in pigeons: evidence for pre-ovulatory follicle selection. Behav. Ecol. Sociobiol. 58, 407–413. Pike, T.W., Petrie, M., 2005. Maternal body condition and plasma hormones affect offspring sex ratio in peafowl. Anim. Behav. 70, 745–751. Pike, T.W., Petrie, M., 2006. Experimental evidence that corticosterone affects offspring sex ratios in quail. Proc. R. Soc. B 273, 1093–1098. Rasbash, J., Steele, F., Browne, W., Prosser, B., 2004. A User’s Guide to MLwiN Version 2.0. Centre for Multilevel Modelling, University of Bristol. Riddle, O., 1917. The control of sex ratio. J. Wash. Acad. Sci. 7, 319–356. Ros, A.F.H., 1999. Effects of testosterone on growth, plumage pigmentation, and mortality in black-headed gull chicks. Ibis 141, 451–459. Rutkowska, J., Cichon, M., Puerta, M., Gil, D., 2005. Negative effects of elevated testosterone on female fecundity in zebra finches. Horm. Behav. 47, 585– 591. Rutkowska, J., Cichon, M., 2006. Maternal testosterone affects the primary sex ratio and offspring survival in zebra finches. Anim. Behav. 71, 1283– 1288. Rutkowska, J., Badyaev, A.V., 2008. Review. Meiotic drive and sex determination: molecular and cytological mechanisms of sex ratio adjustment in birds. Philos. Trans. R. Soc. Lond. Biol. 363, 1675–1686. 192 V.C. Goerlich et al. / General and Comparative Endocrinology 163 (2009) 184–192 Veiga, J.P., Vinuela, J., Cordero, P.J., Aparicio, J.M., Polo, V., 2004. Experimentally increased testosterone affects social rank and primary sex ratio in the spotless starling. Horm. Behav. 46, 47–53. Verboven, N., Monaghan, P., Evans, D.M., Schwabl, H., Evans, N., Whitelaw, C., Nager, R.G., 2003. Maternal condition, yolk androgens and offspring performance. a supplemental feeding experiment in the lesser black-backed gull (Larus fuscus). Proc. R. Soc. B. 270, 2223–2232. Walsh, P.S., Metzger, D.A., Higuchi, R., 1991. Chelex-100 as a medium for simple extraction of DNA for PCR-based typing from forensic material. Biotechniques 10, 506–513.