The authors have declared that no competing interests exist.
Conceived and designed the experiments: DP JMA AP FR CZM MMV. Performed the experiments: DP JMA FR. Analyzed the data: DP AP LS. Contributed reagents/materials/analysis tools: DP JMA AP LS MMV. Wrote the paper: DP JMA AP LS FR CZM MMV.
Chemical defences against predators are widespread in the animal kingdom although have been seldom reported in birds. Here, we investigate the possibility that the orange liquid that nestlings of an insectivorous bird, the Eurasian roller (
Chemical defence is one of the mechanisms that organisms use to enhance their survival prospects. Several different animal taxa from arthropods
The full understanding of the functioning of a chemical defence needs to address the three following general issues
Here, we aim to investigate in depth the possibility that the odorous orange substance that Eurasian roller (
Many herbivorous insects such as grasshoppers regurgitate when disturbed
In this context, we combine detailed diet and behavioural analysis of roller nestlings, experimental approaches and chemical analyses with high performance liquid chromatography–mass spectrometry (LC-MS) to address the following four objectives: 1) whether the substance that nestling rollers expel has a dietary origin; 2) the type of stimulus (mobile, visual, auditive or tactile) that triggers its expelling; 3) the composition of nestling oral secretion; and, finally 4) whether it is effective eliciting aversion by generalist predators.
The roller is a migratory socially monogamous bird that lays one clutch per year of about 5 eggs (mean ± s.e. = 5.20±0.12, N = 60 nests in the study period).
The study was carried out from mid-May to July of 2008 to 2012 in a nest-box breeding population in south-eastern Spain (37°18′N, 3°11′W) (see
From 2008 to 2011 we collected data on nestling diet by identifying prey (up to order level) offered by parents to nestlings from video recordings. In all years we recorded parental provisioning behaviour at nests with 10 day-old nestlings and in 2008 and 2009 also with 18 day-old nestlings.
In 2011 and 2012, we recorded whether each individual nestling vomited or not and weighed them to later associate with age and size of nestlings with vomit production. At each nest we wrote down the type of stimulus that induced vomiting. For that purpose when we arrived to a nest, we opened the nestbox and then followed the next sequence of actions: 1) to speak loudly to nestlings, 2) to show our face to them, 3) to gently touch them, and, finally, 4) to take them in the hand one by one and gently shake them. Actions were separated by ten-second periods. This sequence of actions allowed us to test whether vomiting was in response to an auditive, visual, tactile or mobile stimulus.
In 2012, we performed an experiment using neck collars to deprive nestlings of food and thus test food as the source for vomit production. At each nest with 7 to 20 day-old nestlings (age at which the vomit is expelled, see below), we took all nestlings and assigned them randomly to one of the following two treatments: with or without neck collar. Collars were gently applied to the neck of chicks in such a way that they prevented the transit of prey to the bird’s digestive while allowing birds to breath and expel out vomit. We are certain that collars do not restrict nestlings’ ability to vomit because none of the nestlings that stopped vomiting after collar application vomited after collar removal. Furthermore, many of the nestlings that vomited at the beginning of the experiment reduced their vomit production after collar application but still continued on vomiting. This approach has been widely used to study the diet of insectivorous birds and proved to be innocuous for nestlings
We restricted our analyses to the following compounds that were known to be present in chemically defended plants against herbivorous arthropods or in chemically defended arthropods against predators: L-hyoscyamine
A sample of homogenised vomit (100 µL) was measured with an automatic pipette and passed to a 15-mm glass tube to which 2.5 mL MilliQ water and 240 µL glacial acetic acid was added. The sample was stabilized for 5 min, added with 2.5 mL diethyl ether and vortexed at the highest velocity for 1 min. The mixture was then centrifuged (4000 rpm, 5°C, 5 min) and the organic layer transferred to another tube. The remaining aqueous phase was extracted twice again with 2.5 mL diethyl ether. The combined organic phases were evaporated in RapidVap (Speed: 76, 60°C, 4 min) to almost dryness and then to dryness under a gentle stream of N2 (20–30 min approximately). The dried extract was dissolved in 150 µL of acetonitrile (LC MS Grade, Fisher): MilliQ water before injection.
The samples were analysed using a HPLC separation module (Allience 2695, Waters) with a Quattro Micro triple quadrupole mass spectrometer detector (Waters, Milford, MA). Instrument control, data collection, analysis, and management were controlled by MassLynx 4.0 and Quanlynx V4.1 software packages. Separation was performed using an Atlantis T3 column (2.1×100 mm, 3 µm, Waters) connected to an Atlantis precolumn (2.1×10 mm, 3 µm, Waters) with a flow of 0.3 mL/min. The mobile phase consisted in acetonitrile and MilliQ water, both added with formic acid at 0.1%. The gradient started at 30% of acetonitrile, changed to 40% in 4 min and then changed to 30% of acetonitrile in 6 min and these conditions were held for 6 min. Retention times of the compounds are shown in
The effluents from the HPLC were introduced into the mass spectrometer using an orthogonal Z-spray electrospray interface (Micromass, Manchester, U.K.). The ionization source temperature was 120°C and the desolvatation gas temperature 350°C. The cone gas and desolvation gas-flow rates were 600 and 0 L/h, respectively. The capillary voltage was 3.0 kV and the cone voltage 15 V. Argon gas (2.83 10−3 mbar) was in the collision cell. We optimized the mass spectrometric parameters by continuous infusion of individual solutions of each compound at 10 ppm in methanol:water (1∶1). Detection of the compounds was performed in the positive and negative ionization modes. The quantification of the compounds was based on appropriate Multiple Reaction Monitoring of ion pairs (
Calibration plots were constructed at two different concentration ranges (high and low) (
We assessed the deterrent effects of vomit to predators in July 2010 using dogs
Before the daily feeding, two Petri dishes (50 cm apart), one containing chicken meat smeared with water and the other one containing chicken smeared with roller nestling vomit, were presented to dogs in isolation. We balanced the side (right or left) where each treatment was located across trials. Each dog and vomit was tested only once. Vomit samples used in the experiments came from different nests. Dogs’ behaviour was observed until they ate both pieces of meat or a maximum time of 10 minutes. After that time we considered dogs were non-responsive to the test. From a vantage point we recorded the option each tested dog ate first as a measure of the interest for the stimuli. In addition, we recorded whether each dog ate or not the meat smeared with vomit during the observation period irrespective of which option was taken the first and the time spent to do so.
This study was conducted under licenses of the Junta de Andalucía (Spain) to make the fieldwork with rollers and the Ayuntamiento de Almería (Spain) to perform tests of deterrence of vomit to dogs. Hence, all necessary permits were obtained for the study, which complied the national legislation of Spain concerning animal handling. Study areas are privately owned and permission to use the areas was acquired from the land owners.
We performed a General Linear Mixed Model (MIXED SAS procedure) to test for the effects of the neck collar experiment on nestling weight variation. The effect of the experiment of neck collars on vomit production (decrease
We used a Chi-squared goodness of fit test (FREQ SAS procedure) to compare the observed frequencies in the deterrence test with dogs with the expected frequencies under a scenario of random distribution of choices (i.e. 50% prefer meat with vomit and 50% prefer meat with water).
We identified at least one prey item provided by parents in 34 video recordings (36.2% of total recordings) from 32 different nests (50% of the observed nests). From these 34 video recordings, we identified 112 items, all of them arthropods, mainly belonging to the order Orthoptera (N = 103, 92%), but also some Coleoptera (N = 2, 1.8%), Lepidoptera (N = 3, 2.7%) and centipedes Scolopendromorpha (N = 4, 3.6%) (
Source of prey identification | ||
Prey type | Video recordings | Neck collars |
103 (92.0%) | 19 (90.5%) | |
- 15 (15 |
||
- 4 (2 |
||
2 (1.8%) | 1 (4.8%) | |
- 1 ( |
||
4 (3.6%) | 1 (4.8%) ( |
|
3 (2.7%) | 0 | |
When species identification was possible the latin name of the species is specified in brackets.
We also collected some prey items from neck collars sporadically applied to nestlings in 2008 and in the experiment of food deprivation in 2012. Specifically, we collected 21 arthropods from 14 different nests, 19 belonged to the order Orthoptera (90.5%), 1 to the order Coleoptera (4.8%) and 1 to the order Scolopendromorpha (4.8%) (
All nestlings (N = 43) expelled out the vomit when they were moved but not in response to the other stimuli (auditive, visual or tactile). Furthermore, most nestlings began to vomit when they still were blind, indicating that at that age regurgitation cannot be a response to a visual stimulus.
The vomiting behaviour was initiated when nestlings were 6.7±0.7 days old (mean ± s.e., N = 43 chicks from 11 nests) and weighed 57.2±6.8 g (mean ± s.e., N = 34 chicks from 9 nests). Nestlings lost this behaviour when they were 19.6±0.4 days old (mean ± s.e., N = 37 chicks from 11 nests), which is around fledging time.
In 2012 we applied neck collars to half of the nestlings (14 nestlings) from 9 nests. Collars were efficient because nestlings with neck collars lost more weight than nestlings without neck collars (General Lineal Mixed Model: F1,18 = 8.33, P = 0.0098. Mean weight loss = 4 g (N = 14 nestlings with collars)
The results show that all the vomit samples contained Hydroxybenzoic and Hydroxycinnamic acids although in 2 and 4 cases respectively out of 16 samples, there were only traces of the chemicals. Hydroxybenzoic acidconcentration was 481.2±77.6 ppb (mean ± s.e.) (min–max = 130.1–1139.1 (N = 14)). The content in Hydroxycinnamic acid was 150.0±26.1 ppb (mean ± s.e.) (min–max = 60.2–354.4 (N = 12)). In one sample Psoralen was found close to the Limit of Quantification (17.99 ppb) and in some samples (4 out of 16) traces of Psoralen were detected but could not be quantified since their amount was close to the Limit of Detection, but below the Limit of Quantification (<9 and >3 ppb). On the other hand, traces of Dihydronepetalactone, close to or well below the Limit of Detection (<10 ppb), were sporadically detected. Additionally, if Hyoscyamine was present it could not be detected as above indicated. Another processing system was also assayed, by using Ostro cartridges (Waters, Mildford), but the compound could not be recovered from any of the spiked vomit samples. P-Benzoquinone was also included as a candidate compound. However, this chemical showed a high resistance to be broken and did not produce any fragment under the MS conditions used here, preventing its determination with the mass spectrometric detector. The analysis by HPLC-UV at 290 nm did not lead to any positive conclusion either.
We performed the deterrence test to 25 dogs, 5 of which were not responsive. Before deciding whether eating or not the offered meat, dogs either smelt (most cases) or licked it. Most of the reactive dogs (18 out of 20) preferred as the first option the meat smeared with water instead of the meat smeared with roller vomit (Goodness of fit test: χ21 = 12.8, P = 0.0003). 12 out of 18 dogs (67%) that chose meat with water as the first option also consumed meat with vomit as the second option but they did that after 2 minutes in average (mean = 118.4 seconds). The remaining 6 dogs out of 18 only ate meat with water. Meanwhile, the 2 dogs that chose meat with vomit as the first option also ate the meat with water immediately after (mean = 31.0 seconds).
In this paper we first show that arthropods from the order Orthoptera are the main prey of roller nestlings in the study area. We also demonstrate that the vomit expelled by roller nestlings depends on food provided by parents and that vomiting is triggered by grasping and moving of nestlings. In addition, we have found that vomit samples contain variable concentrations of hydroxycinnamic and hidroxybenzoic acids, two phenolic acids, and that some of the vomit samples also have traces of psoralen, a furanocoumarin. Finally, we have shown that vomit of nestling rollers alone makes chicken meat unappealing for dogs. Below, we will critically assess these findings in the light of the hypothesis that nestling rollers regurgitate when disturbed, expelling an orange and odorous substance
We have found that the movement of nestlings by the investigator seemed to trigger vomit ejection. This fact suggests that the vomit might be produced in response to some kind of predators that actively grasp and move prey during the predation event such as snakes, rats and mustelids, which are common predators of hole-nesting species
Our results also indicate that the production of vomit depends directly on recently consumed food because when nestlings were food-deprived for 1 hour they reduced vomit production. This result suggests that the vomit has not an endogenous (i.e. glandular) but a dietary origin. The oral emissions of arthropods contain a blend of digestive enzymes, salivary secretions, and partially digested food as plant secondary compounds
In the study area roller nestlings are mainly fed with Orthoptera, which are relatively polyphagous species
It should be acknowledged here that despite the initial avoidance that dogs showed against meat with vomit, many dogs finally ate it. However, they did that after some minutes, perhaps after the volatilization of much of the smell of the vomit
To summarize, several lines of evidence support the idea that the vomit of nestling rollers might have a defensive function against predation: 1) It is expelled in response to a threat, our handling, at nests. 2) Vomit seems not to be produced
Retention time (tR) and optimised mass spectrometric parameters for the detection of the compounds under study.
(DOCX)
Assessment of the analytical parameters in chemical analyses.
(DOCX)
We are grateful to the staff of the Centro Zoosanitario Municipal in Almería, who kindly provided help with dogs, to the Instrumentation Service of the EEZ (CSIC) in Granada, where chemical analyses were done, and to Juan Rodríguez for his help in the field.