Darwin’s finches habitually anoint their feathers with leaves of the endemic tree Psidium galapageium during the non- breeding season

Birds host a wide range of ectoparasites and have developed behavioural strategies to combat them, such as preening, dust bathing and water bathing. In addition, a wide range of avian taxa anoint their feathers with insects or plants that have pharmaceutical properties, though most observations on anointing are anecdotal. Darwin's finches preen with leaves of an endemic tree ( Psidium galapageium ) and a previous laboratory study has shown that this plant has compounds that repel both mosquitoes and the invasive parasitic fly Philornis downsi , whose larvae suck blood from nestlings and incubating females and cause high nestling mortality. In the current study, we tested the hypothesis that preening with P . galapageium leaves serves to repel these parasites with an indirect approach. Mosquitoes and P . downsi affect their hosts mainly during the bird breeding season and P . downsi only affects breeding females, but not adult males. To test our hypothesis, we gathered quantitative data on leaf- preening behaviour in Darwin's finches during their breeding and non-breeding season and also investigated the influence of time of day and humidity, as humid conditions facilitate the release of volatile organic compounds. Contrary to our predictions, anointing occurred significantly more often during the non- breeding season when mosquito and P . downsi numbers are lower. Four Darwin's finch species anointed their feathers habitually, and during the non- breeding season, 56% of all preening events were with leaves. We found no effect of sex, but preening with leaves occurred predominately in the morning when leaves were wet. Our study is the first to provide quantitative data on anointing behaviour in birds and the high percentage of preening with leaves in the non- breeding season suggests that the behaviour has an adaptive value. However, further studies are needed to test whether it reduces the negative impact of parasites other than mosquitoes and P . downsi .


| INTRODUC TI ON
Parasites can have strong detrimental effects on host fitness and are therefore a major selective force on various host traits (Clayton & Moore, 1997;Grenfell et al., 1995;Loye & Zuk, 1991;Rätti et al., 1993;Toft et al., 1991). Hosts have developed a wide array of defence mechanisms to counteract the effects of parasitic pressure on their fitness (reviewed in Clayton & Wolfe, 1993). Birds are hosts to various types of ectoparasites including arthropods, bacteria and fungi (reviewed in Bush & Clayton, 2018). To combat them, they have developed a range of nest and body maintenance behaviours such as nest sanitation, preening, allopreening, scratching, water bathing, dust bathing and sunning (reviewed in Bush & Clayton, 2018). In addition, a wide range of avian taxa supplement their behavioural strategies with self-fumigation, which is the application of substances to the body or their nests for the treatment or control of parasites or parasite vectors (reviewed in Bush & Clayton, 2018;Huffman, 2019).
"Anting" is a behaviour which involves the anointment of the feathers with ants and is performed habitually by over 200 bird species (Clayton et al., 2010). The formic acid that is released during anting is assumed to repel harmful parasites (Falótico et al., 2007;reviewed in Potter, 1970;Revis & Waller, 2004). Anointing with plant materials has been observed in only a few bird species and observations are rare or anecdotal (reviewed in Clayton & Wolfe, 1993). For instance, common grackles (Quiscalus quiscula) have been observed to preen with marigold flowers (genus Tagetes), which contain a natural insect repellent (Nero & Hatch, 1984), as well as with the pulp and rind of a lime fruit, which have insecticidal properties (Clayton & Vernon, 1993;Rodriguez & Wrangham, 1993). Additionally, the monarch flycatchers (Chasiempis sandwichensis) rub berries of Brazilian pepper (Schinus terebinthifolius), which has antibiotic properties, onto their feathers (VanderWerf, 2005).
The current study investigated anointing behaviour in Darwin's finches. In 2012, a green warbler finch (Certhidea olivacea) and later four other species of Darwin's finches were observed for the first time tearing off the leaves of the endemic tree Psidium galapageium and rubbing them onto their feathers (Cimadom et al., 2016). The authors observed two different methods of anointing: 1) the sponge method, in which the bird threads a piece of leaf through its feathers and 2) the lotion method, in which the bird chews the leaf first and applies the mashed leaf to its feathers. Cimadom et al., (2016) hypothesised that the birds use these leaves to repel ectoparasites that negatively impact the fitness of Darwin's finches, namely the blood-sucking fly Philornis downsi and mosquitoes. Several mosquito species are native to the Galápagos Islands but others have been introduced (e.g. Culex quinquefasciatus, Aedes aegypti; Sinclair, 2013) and transmit novel mosquito-borne pathogens such as avian poxvirus (Parker et al., 2011). The avian poxvirus causes lesions on toes, legs, and the tissue around the bill. Individuals that survive often have deformed or missing digits (Parker et al., 2011). The introduced parasitic fly P. downsi has an even stronger effect on the fitness of Darwin's finches than the poxvirus: approximately 55% of Darwin's finch nestlings die annually due to parasitism by this species (Fessl et al., 2018;Kleindorfer & Dudaniec, 2016). The fly's first larval stage is mainly found in the nestlings' nostrils. The second and third larval stages live in the bottom of the nest where they penetrate the skin of the nestlings and consume their blood (Fessl et al., 2006) and also attack incubating females (Cimadom et al., 2016;Knutie et al., 2013). Cimadom et al., (2016) demonstrated a repellent effect of extracts of P. galapageium on mosquitoes and adult P. downsi and a growth inhibiting effect on P. downsi larvae in the laboratory. The abundance of the above-mentioned parasites varies seasonally. P. downsi affects Darwin's finches mainly during the birds' breeding season (January -April), which is when their parasitic larval stage occurs in the finches' nests (Fessl et al., 2006). Mosquito abundance is also higher during the warm and rainy breeding season, as it is only then that favourable temperatures (> 20°C) and the availability of stagnant water pools stimulate mosquito breeding (Asigau et al., 2017;Khan et al., 2018). On the Galápagos islands, mosquito abundance increases with precipitation but decreases with altitude (Asigau & Parker, 2018;Bataille et al., 2010). Combined, these factors result in very low abundance of mosquitoes in the highlands of the Galápagos Islands during the cool non-breeding season (Asigau et al., 2017).
The seasonality of P.downsi and of mosquitoes allows for an indirect approach to test whether Darwin's finches use P. galapageium leaves to repel them. Here, we chose to concentrate on these parasites although Darwin's finches also suffer from other parasites, such as feather mites and feather lice (Palma & Peck, 2013;Villa et al., 2013).
If P. downsi and mosquitoes are targeted by this behaviour, the frequency of preening with leaves should also match the abundance of the targeted organisms.
Furthermore, it has been suggested that time of day and humidity have an effect on preening behaviour and could also influence preening with plants that release volatiles. Wet conditions of high humidity are known to facilitate the release of volatile organic compounds (VOC; Gouinguené & Turlings, 2002;Salerno et al., 2017;Vallat et al., 2005). In several bird species, the frequency of preening peaked in the early morning (Henson et al., 2007;Robbins, 1981) and increased with increasing humidity (Brown, 1974;Henson et al., 2007).
In our study, we tested the hypothesis that preening with leaves by Darwin's finches serves to repel the invasive parasitic fly P. downsi and/or native and introduced mosquitoes. We predicted that if preening with P. galapageium serves to protect incubating females from P. downsi, the behaviour should be more frequently observed in females and more frequently during their breeding season. If preening serves to protect against mosquitos, we expected the behaviour to be evenly distributed across both sexes and to increase with the high abundance of mosquitoes in the breeding season. To test these hypotheses, we gathered quantitative data on leaf-preening behaviour of Darwin's finches during the breeding and non-breeding season. We measured whether the occurrence of leaf-preening is influenced by season, time of day or wetness of leaves and whether it differs between the sexes.

| Study area
This study was conducted in Los Gemelos (0°37'34" S, 90°23'10" W), at an elevation of around 600 m, in the humid "Scalesia" forest on the Island of Santa Cruz, Galapágos from January 19- The climate on the Galápagos Islands is highly seasonal. During the breeding season, from January to April, mean air temperatures in the humid highlands range between 16 and 28°C and the skies are usually clear with occasional heavy rain showers. During the nonbreeding season from June to December, the temperature is lower (ranging between 13 and 20°C) and while there is hardly any precipitation in the lowlands, the highlands are continuously wet due to a consistent, dense mist (Causton et al., 2019;Jackson, 1993).

| Behavioural observations
We selected 30 points for behavioural observations. Each point contained at least three medium-sized P. galapageium trees (>3 m high) and points were 100 m apart from each other ( Figure S1). The distance between observation points was chosen to minimise possible overlaps of bird territories between points and was measured via GPS. At each point, we recorded the following habitat parameters: canopy height as well as the total number of and height of P. galapageium within a radius of 15 m (hereafter "point radius"). The point radius was measured with a laser range finder. Each point was visited five times per season between 6:00 and 11:00, which is the period of highest bird activity. On average, eight points were visited per day and routes between points were chosen so that each point was observations were made by one person (TS). At the beginning of each visit, the observer noted the number and species identity of the passerines present within the point radius. Then, for a duration of 30 min, preening events were recorded, along with the specification of whether preening was conducted with or without P. galapageium. The observer did not record whether birds that preened with leaves used the sponge or lotion method because this would have required focal observation of preening individuals and would have increased the probability of missing preening individuals. The following parameters were recorded for each preening individual: species identity, life stage (adult vs juvenile) and sex (male vs female/ immature male combined; see below for explanation of latter categorization). Adult males were identified using plumage characteristics, beak colour and song activity: small ground finch (Geospiza fuliginosa), medium ground finch (Geospiza fortis) and large ground finch males (Geospiza magnirostris) show a streaky plumage from crown to chin and their overall plumage darkens with every moult until they are completely black (Kleindorfer et al., 2019). Small tree finch (Camarhynchus parvulus) females and immature males have a light brown colouration. Mature males have black feathers on the head, starting at the beak and later forming a dark hood that extends down to throat and breast (Kleindorfer et al., 2019). Green warbler finch females and immature males have a grey-greenish colouration while males can be distinguished by their orange throat (Kleindorfer et al., 2019). In Darwin's finches, only males sing, thus any bird that sang was identified as male. However, yearling males of all Darwin's finches that do not sing cannot be distinguished from females and were therefore categorised as being in the female-immature male group (other). Juveniles were identified by pink beak colouration and begging behaviour. We distinguished individuals that were present at the same time and new arrivals based on species, sex and plumage colouration. Observations of individuals that looked the same and were not observed simultaneously were excluded from the data set.
At the start of each observation, leaves of P. galapageium at the observation point were scored as wet when the leaf surface exhibited visible dew or water droplets. Otherwise, they were scored as dry. The initial time was noted as well as the weather conditions, which were categorised as rain versus no rain and sun versus fog.
Temperature and humidity were recorded with two DS1923 hygrochron temperature/humidity data loggers (iButton ® ).

| Statistical analyses
For the statistical analyses, juveniles were excluded because they were not observed preening with leaves during the breeding season.
Only data from four species (the green warbler finch, small ground finch, medium ground finch and small tree finch) were analysed.
We fitted a generalised linear mixed regression model to the binary preening data (with leaves versus without leaves) to test for effects of season (non-breeding versus breeding season), sex (male versus other), wetness of leaves (dry versus wet), and time of day.
Season, sex, and wetness of leaves were included as binary categorical predictors and time was treated as a numerical predictor (time of day at preening event, ranging from 6:00 to 11:00). To account for the fact that the observed preening events varied substantially between observation points, we included point ID, ranging from one to 30, as a random factor (random intercept) in our model. The statistical significance of coefficients was assessed using likelihood ratio tests.
We had no clear a priori hypotheses concerning which species might anoint with P. galapageium more or less than the others nor concerning how rain, sun or temperature would affect preening behaviour. We did not conduct hypothesis tests for these variables and did not include them in the regression model, but we explored their effects by visualising their associations with preening.

| RE SULTS
We observed 229 preening events in total for the four species included in the analysis (the green warbler finch, small ground finch, medium ground finch and small tree finch). Of these observations, Both sexes preened with leaves at a similar frequency (effect of sex was not statistically significant in the model, p =.16; Figure 2).
There were species differences in the preening patterns between the seasons (Figure 1a). For example, medium ground finches preened predominantly in the non-breeding season and almost exclusively with leaves, whereas warbler finches preened mainly without leaves in the breeding season.
The behaviour of preening with leaves occurred when wet leaves were present in the non-breeding season ( Figure 3) and both preening with leaves and occurrence of wet leaves showed the same decreasing daily trend (Figure 4). However, the effect of leaf wetness on preening with versus without leaves was strongly correlated with season (r = 0.66) and therefore was itself not statistically significant in the model (p =.10). Humid conditions were ever-present in the non-breeding season, so the vegetation was frequently wet (correlation between rain and presence of wet leaves was r = 0.77, correlation between season and rain was r = 0.70).

| D ISCUSS I ON
Contrary to our prediction, leaf preening occurred more frequently in the non-breeding season, when mosquito activity is low and TA B L E 1 Observations of preening with and without leaves for the 8 observed species. Species in the four bottom rows were excluded from the analysis because they were never or only once observed to preen with leaves on Darwin's finches, but these numbers were too low to enable investigation of seasonal patterns. Given the small numbers found on Darwin's finches, it does not seem likely that the blood-feeding mites are the main target of leaf-preening behaviour.
In addition to mites, three genera of lice have also been found ter (Burtt & Ichida, 1999). The authors suggest that the higher incidence may be related to higher temperature and humidity but also to increased exposure to ultraviolet (UV) light, as it kills the vegetative cells of bacteria and their spores (Madigan et al., 1997). Humidity was always high at our study site, but temperature and the number of hours of sunshine were low during the non-breeding season.
This combination of environmental factors could potentially lead to seasonality in feather-degrading bacteria. The influence of climate conditions on feather-degrading fungi is less clear. A study in house sparrows (Passer domesticus L.) showed no overall seasonal pattern (Hubálek, 1976).
Analysis of the ethanolic extract and the essential oil of P. galapageium and its respective fractions (Cimadom et al., 2016)  Pandey & Shweta, 2011). Combined, these findings raise the possibility that preening with P. galapageium reduces feather-damaging microbes. However, this needs to be tested in future studies with an experimental approach in which the incidence and seasonality of feather microbes is sampled. The antimicrobial properties of P.
galapageium could be tested in vitro on the samples taken from the birds' feathers. P. galapageium also contains eucalyptol (Cimadom et al., 2016, Martina et al. submitted), which has skin soothing properties (Gilles et al., 2010). If parasites cause itchy skin irritations, a soothing effect of P. galapageium could elicit the behaviour of preening with leaves.
As predicted, the behaviour of preening with leaves co-occurred with the presence of wet leaves in the non-breeding season. During the breeding season, almost all of the few observations of preening with leaves occurred before 9:00. Both humidity and moistness of leaves decreased with time of day, which could explain why preening with leaves occurred more frequently during the early hours of the day. Wet conditions facilitate the release of volatile organic compounds (VOCs) by plants (Gouinguené & Turlings, 2002;Salerno et al., 2017;Vallat et al., 2005) and thus more VOCs may be present in the early morning hours when humidity is high, making leaf preening at this time more effective. Whether time of day, rain, wetness of leaves or other factors, such as the activity of herbivorous insects (Hare, 2011), increase the release of VOCs remains to be shown. This could give greater insight into the possible function of this behaviour, but also its annual pattern. For example, if any of the above-mentioned factors triggers volatile emission, the effect on the fitness of the bird could be episodic and impact different ectoparasites at different times of the year but also lead to variation between years. On the Galápagos Islands, climatic conditions vary strongly between years (Jackson, 1993). Thus, sampling of volatile emission and preening behaviour throughout the year and over several years would be necessary for a comprehensive picture. An alternative explanation for the peak of leaf-preening behaviour in the morning is that the preening pattern follows the activity of the parasites. For instance, Amblyceran chewing lice but also the larvae of P. downsi have a diurnal feeding rhythm. (O'Connor et al., 2010;Stenkewitz et al. ,2017).
A higher concentration of volatiles could also be the mechanism by which preening with leaves is triggered. The ability to detect VOCs through olfaction (Amo et al., 2011;Krause & Caspers, 2012;Nelson Slater & Hauber, 2017) or taste (Berkhoudt, 1992;Niknafs & Roura, 2018)  fixed components but also components which are developed in ontogeny through individual learning (Tebbich et al., 2001). A similar scenario can be envisaged for preening with leaves, but this needs to be tested experimentally. Alternatively, this behaviour could have been discovered by one or several individuals of one species and then passed on between species by way of a cross-species transfer of information (Avarguès- Weber et al., 2013;Krebs, 1973).
Additionally, our results show that juveniles apparently only start preening with leaves in the non-breeding season, which may indicate that this behaviour needs time to mature or that Darwin´s finches need time to acquire this behaviour through social learning, as has been shown in other bird species (Slagsvold & Wiebe, 2011;Zentall, 2004). An alternative explanation is that juveniles suffer less from parasites and therefore show less leafpreening behaviour.
In conclusion, currently we can only speculate about the function of preening with leaves of P. galapageium, but our study is the first to quantify topical application of a plant species and to report this as a predictable, habitual behaviour that varies seasonally in a closely related species group. In the non-breeding season, the frequency of this behaviour was surprisingly high (56% of all preening events), which suggests that this behaviour has adaptive value. Anointing feathers with secondary products (ants, millipedes, beetles, caterpillars, plant materials and pesticides) is taxonomically widespread, but rare, which makes it hard to evaluate the significance of this behaviour to birds' fitness (reviewed in Bush & Clayton, 2018;reviewed in Potter, 1970). For instance, although anting has been observed in over 200 species, evidence that it reduces parasite load is scarce (reviewed in Bush & Clayton, 2018). This is probably attributable to methodological difficulties associated with measuring the effect of anointing and other forms of self-medication (Bush & Clayton, 2018;de Roode et al., 2013). Clayton and Wolfe (1993) established three criteria for defining self-medication behaviour: (I) the medicinal substance must be deliberately contacted by the meditator; (2) the substance must be detrimental to one or more parasites when contacted and (3) the detrimental effect on parasites leads to an increase in host fitness.
Preening with P. galapageium leaves already satisfies two criteria of self-medication, as the leaves are actively applied by the birds and Cimadom et al., (2016) have demonstrated repellent qualities against potential parasites. However, we were unable to identify which parasites the birds are targeting and cannot provide evidence for parasitereducing or fitness-enhancing effects. While we are not currently able to identify the function of anointing in Darwin's finches, this behaviour is another example of the incredible behavioural diversity of this species group and is one of several examples in which animal behaviour has led humans to the discovery of the pharmaceutical properties of an endemic plant species (Huffman, 2002). An international research group is currently investigating short and long-term measures to mitigate the effects of P. downsi and experiments with Nerolidol, one of the main compounds of P. galapageium, are yielding the first promising results showing that it is an effective repellent of adult flies.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

E TH I C A L S TATEM ENT
Permission to conduct this study was granted by the Galápagos National Park Directorate (Project PC-35-19: Control of the Invasive Parasite, P. downsi and its Impact on Biodiversity). The study was observational only and did not impact the study subjects. All applicable institutional and/or national guidelines for the care and use of animals were followed.