Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Decoding and Discrimination of Chemical Cues and Signals: Avoidance of Predation and Competition during Parental Care Behavior in Sympatric Poison Frogs

  • Lisa M. Schulte ,

    Current address: Amphibian Evolution Lab, Vrije Universiteit Brussel, Brussels, Belgium

    Affiliation Department of Biogeography, Trier University, Trier, Germany

  • Martin Krauss,

    Affiliation Department of Effect-Directed Analysis, Helmholtz Centre for Environmental Research, Leipzig, Germany

  • Stefan Lötters,

    Affiliation Department of Biogeography, Trier University, Trier, Germany

  • Tobias Schulze,

    Affiliation Department of Effect-Directed Analysis, Helmholtz Centre for Environmental Research, Leipzig, Germany

  • Werner Brack

    Affiliation Department of Effect-Directed Analysis, Helmholtz Centre for Environmental Research, Leipzig, Germany

Decoding and Discrimination of Chemical Cues and Signals: Avoidance of Predation and Competition during Parental Care Behavior in Sympatric Poison Frogs

  • Lisa M. Schulte, 
  • Martin Krauss, 
  • Stefan Lötters, 
  • Tobias Schulze, 
  • Werner Brack


The evolution of chemical communication and the discrimination between evolved functions (signals) and unintentional releases (cues) are among the most challenging issues in chemical ecology. The accurate classification of inter- or intraspecific chemical communication is often puzzling. Here we report on two different communication systems triggering the same parental care behavior in the poison frog Ranitomeya variabilis. This species deposits its tadpoles and egg clutches in phytotelmata and chemically recognizes and avoids sites with both predatory conspecific and non-predatory heterospecific tadpoles (of the species Hyloxalus azureiventris). Combining chemical analyses with in-situ bioassays, we identified the molecular formulas of the chemical compounds triggering this behavior. We found that both species produce distinct chemical compound combinations, suggesting two separate communication systems. Bringing these results into an ecological context, we classify the conspecific R. variabilis compounds as chemical cues, advantageous only to the receivers (the adult frogs), not the emitters (the tadpoles). The heterospecific compounds, however, are suggested to be chemical signals (or cues evolving into signals), being advantageous to the emitters (the heterospecific tadpoles) and likely also to the receivers (the adult frogs). Due to these assumed receiver benefits, the heterospecific compounds are possibly synomones which are advantageous to both emitter and receiver ‒ a very rare communication system between animal species, especially vertebrates.


Communication in the biological sense is defined by Wilson [1] as the action of one organism that alters the probability pattern of behavior in another. A major difficulty when defining communicative systems is to distinguish between evolved functions and incidental effects [2,3]. Chemical communication is generally regarded as the oldest and most widespread form of communication that occurs at all levels of biological organization [46]. Besides chemical signals, which are intentionally released by the sender, receivers also react towards unintentionally released chemicals, which are defined as chemical cues [7]. While chemical signaling is usually advantageous to the sender, releasing chemical cues is either neutral or damaging to it. However, if the reaction of the receiver towards chemical cues is beneficial to the sender, evolution into signaling is expected [8]. The evolution of chemical communication is therefore described as ‘chemical ritualization’ towards chemical signals [7]. Intraspecific signals are defined as pheromones [5], while signals between different species can be divided into synomones and allomones. While allomones are only beneficial to the sender (e.g. being used in predacious or parasitic relationships), synomones are beneficial to both, sender and receiver. That is, synomones are used in mutualistic relationships and are very rarely found between animal species (especially vertebrates) [9]. The proper categorical classification of a chemical substance that triggers a certain behavior in another individual is often puzzling and sometimes only a closer look at the chemical compounds involved brings clarification (for examples see [1012]).

In this study we combine behavioral in-situ bioassays and comparative chemical analyses in order to categorically classify intra- and interspecific communication systems in two species of poison frogs. Frogs that display advanced levels of parental care behavior (including the choice of exclusive breeding habitats used by various species) offer unique opportunities to investigate the influence of chemical communication on parental decisions. The Neotropical poison frog Ranitomeya variabilis [13] uses phytotelmata (such as small water bodies in plant axils [14]) for both egg clutch and tadpole deposition. It attaches its egg clutches just above the water surface and when the larvae hatch the male parent returns and transports them individually into different phytotelmata [15]. Because its tadpoles are cannibalistic, feeding on both conspecific tadpoles and eggs [16,17], parent R. variabilis need to locate unoccupied phytotelmata for their offspring. The recognition and avoidance of occupied water bodies is triggered by chemicals released by conspecific tadpoles [18]. An identical or even stronger avoidance reaction was shown towards chemicals released by tadpoles of another poison frog species of the same family, Hyloxalus azureiventris [19]. The tadpoles of this aposematic species occupy large ground phytotelmata, which sporadically are also used by R. variabilis for depositing tadpoles. We aimed to identify which chemical compounds trigger the avoidance of both con- and heterospecific tadpoles and to categorize the communication systems behind this behavior.

The avoidance of predatory conspecific tadpoles (i.e. the sender) increases the frogs’ (i.e. the receivers’) reproductive success, but is not advantageous to the sending tadpoles occupying the phytotelmata as they lose a possible food resource. In this case we might define the chemicals that trigger the avoidance behavior as chemical cues, released unintentionally by the tadpoles. However, this might not be the case for the chemicals released by H. azureiventris. These heterospecific tadpoles are not predatory and should therefore not pose a threat to (but may be in competition with) the offspring of R. variabilis. But the avoidance behavior of R. variabilis allows H. azureiventris tadpoles to evade threats of both predation and competition; so such behavior is presumably advantageous for them. Thus, the chemicals released by H. azureiventris tadpoles might be signals to the heterospecific adults of R. variabilis (or cues that are in the process of evolving into signals, see [7,8]). However, the phylogenetic relationship between these two poison frog species (even if not very closely related) raises the question of whether the chemicals released by both species might be the same, meaning that the avoidance of H. azureiventris might be a result of R. variabilis not being able to distinguish between con- and heterospecific tadpoles.

Using effect-directed analysis [20] combining in-situ bioassays, fractionation and comparative chemical analyses, we tested the following hypothesis in order to identify and categorize this interspecific communication system:

(i) The chemical tadpole compounds avoided by R. variabilis are identical between R. variabilis and H. azureiventris. This would complicate the classification of the interspecific compounds produced by H. azureiventris tadpoles: (a) they could be a non-species-specific by-product and might therefore be defined as a chemical cue, not adapted for use in communication. Or (b) they might be produced in order to mimic the compounds of R. variabilis. In this way the H. azureiventris tadpoles would use chemical signals to mislead the adult R. variabilis in order to prevent sharing a pool with their predatory tadpoles. These signals could be defined as allomones, which benefit the H. azureiventris tadpoles (i.e. the sender) because they are not recognized as interspecific signals, but confused with intraspecific cues by the R. variabilis adults (i.e. the receiver).

Alternatively, we hypothesize that (ii) the biologically active compounds chemically differ between the two species. This would mean that the compounds released by H. azureiventris tadpoles can be defined as signals (or cues evolving into signals) that trigger a specific response in the heterospecific receiver (the R. variabilis adults). The chemicals involved here would be most likely synomones, since they benefit the sender (H. azureiventris) and are recognized but nevertheless avoided as heterospecific compounds by the receiver (R. variabilis). A possible benefit for the receiver might then be the avoidance of a competition situation for its own offspring.

Materials and Methods

An overview of the succession of the methodical steps is given in Fig 1. Chemicals and analytical instruments used in this study are given in the appendix.

Fig 1. Overview of methods.

Steps are numbered in accordance with the text (a—f). V and A stand for fractions of R. variabilis and H. azureiventris, respectively. Structural formulas shown in (f) are only examples.

Ethics statement

The research in Peru was done with permits and according to the guidelines of the Peruvian Ministry of Agriculture (DGFFS) in Lima (authorization No. 0200-2011-AG-DGFFS-DGFFS, 0176-2012-AG-DGFFS-DGFFS and 416-2013-AG-DGFFS-DGFFS; all permits allowed tissue collections of the species, but no tissues were sampled for the project). No animals were killed and all tadpoles were released after the experiment was finished. Study sites were situated at the western side of the Cainarachi River, Región San Martín, Peru: 6°25’53.4”S, 76°17’10.1”W, 6°25’50.3”S, 76°17’30.5”W, 6°25’44.8”S, 76°17’35.5”W.

Captive breeding groups in Germany were legally registered at the nature conservation authority of Trier (Germany). Since we only performed analyses of the water in which tadpoles were kept this was not classified as an animal experiment. Therefore, according to the German animal welfare law and the university representative, Prof. Dr. Thomas Schmitt, formal approval was not necessary.

(a) Search for a procedure to extract the chemical cues

To find a suitable solid-phase extraction (SPE) sorbent for concentrating the cues from water samples, preliminary experiments were conducted. The goal was to eliminate the active compounds from the tadpole water by sorption to the SPE material and later test these “cleaned” water samples for their activity in bioassays. Successful sorption implies that the active compounds are now in the SPE material and no longer in the water previously used by tadpoles. Accordingly, in following bioassays parental R. variabilis should not avoid this “cleaned” water for tadpole and egg clutch deposition. The tests of different SPE materials were conducted in-situ in a premontane late-stage secondary forest in the study site mentioned above.

(i) Activated carbon.

Between 23 March and 8 May 2011 27 tadpoles of each Ranitomeya variabilis and Hyloxalus azureiventris were collected from their natural habitat and raised in captivity in a wooden hut in the forest serving as a field laboratory during the trials. Each specimen was housed individually in 50 ml of rainwater in identical polypropylene cups. Every other day, we combined the water of the tadpoles (separated by species), refilled the cups with fresh rain water and fed the larvae ad libitum with flaked fish food (Tetra). After finishing the trials or when tadpoles reached Gosner developmental stage 41 [21], they were released back into the wild. The water obtained from the tadpoles was thereafter filtered through approximately 30 g of granular activated carbon (Merck Millipore, 1.5 mm diameter) in order to sorb the odor-producing substances [22].

(ii) Chromabond and DSC-18.

Two times twenty-five Ranitomeya variabilis tadpoles were collected and kept in the field laboratory from 25 March to 8 May 2011 together in one of two containers in 1250 ml of rainwater. Because tadpoles of this species are cannibalistic, each individual was enclosed in a small mesh bag. The mesh was large enough to allow feces to pass, but small enough to prevent physical contact between the tadpoles. A wide opening towards the water surface facilitated the feeding procedure and enabled the animals to take up oxygen. At the bottom of each container a passive sampler was placed (Sartorius membrane filter, 7 cm diameter, 1.2 μm pore size) containing either 1.7 g of Chromabond HR-X (Macherey-Nagel) or 4 g of Discovery DSC-18 (Supelco) to sorb the compounds. Differences in filter content weights were inevitable due to particle size of the sorbents. The water of the trials was collected and replaced every second day followed by feeding of the tadpoles (see (i)). The passive samplers were changed with every third water change. Trials were not conducted with Hyloxalus azureiventris because we could not find a sufficient number of tadpoles.

(iii) Bioassays in the forest.

We established three study sites on the western side of the Cainarachi River at an elevation between 540 and 580 m above sea level. In order to avoid pseudo-replication, i.e. repeated measures on the same specimens, we maintained a minimum distance of 30 m between the sites, which exceeds the known home range size of Ranitomeya variabilis [23]. We established artificial phytotelmata at three forest sites. Polypropylene plastic cups (200 ml volume, 10 cm height, 7 cm in diameter), wrapped in dark plastic membranes and with two-thirds of the opening covered, were fixed in pairs to trees at 0.5–1.5 m above ground. Each cup was filled with 25 ml water, one cup per pair with clean rainwater and the other cup with activated carbon, HR-X or C18 treated tadpole water as explained in (i) and (ii). For each treatment we spread 45 cup pairs throughout the three forest sites. They were checked every second day for newly deposited tadpoles and egg clutches of R. variabilis and both clean and treated water were changed if no depositions were found. Pairs that received a deposition were not used again but another pair was hung up at another tree instead. In order to minimize pseudo-replication, we kept a minimum of 4 m between cup pairs belonging to the same trial and scored only one randomly chosen deposition for any same-day deposition in neighboring cup pairs [18,24].

(b) Directed chemical extraction and primary fractionation of the chemical compounds

Based on the results of (a) we used Discovery DSC-18 (Supelco) for the extraction and collection of the active chemical compounds from the tadpole water. Although we had no data related to the extraction of chemical compounds produced by Hyloxalus azureiventris, we applied our methods in the laboratory to both Ranitomeya variabilis and H. azureiventris.

We used tadpoles from captive breeding colonies at Trier University and kept them from 12 November 2011 to 24 January 2012 at the Department of Effect-Directed Analysis, UFZ Leipzig, where fractionation and chemical analysis were conducted. During this time we had between six and nine R. variabilis tadpoles and 11 to 23 H. azureiventris tadpoles. They were kept individually and their water was changed twice a day: six hours (+/-1) a day they were kept in approximately 50 ml of tap water with abundant food supply (see (a)) and 18 hours (+/-1) a day in 10 ml without food. The water without food, containing excretions produced by the tadpoles only, was collected over two days, separated by species, and vacuum filtered for further analysis (Whatman GF/F glass microfiber filters, pore size 0.7 μm). Thereupon a solid phase extraction was conducted, using Discovery DSC-18 (Supelco) as sorbent. The sorbent was preconditioned using dichloromethane (DCM), methanol (MeOH) and bidistilled water before the water sample was passed through the cartridge at a flow rate of about 10.0 ml/min. Once the entire sample had passed through, cartridges were dried under nitrogen and eluted two times with 6 ml of MeOH and DCM each. Elutriates were collected and reduced in volume to dryness in a rotary evaporator. The dried compounds were collected at 4°C and while 214 R. variabilis and 652 H. azureiventris samples were kept in the dry form as total samples (Vtotal and Atotal) for later use in the bioassays, 238 R. variabilis and 406 H. azureiventris samples were reconstituted in MeOH (10 μL per tadpole sample) for fractionation.

As a control sample to check for background contamination that derived from the fish food, a water sample containing only fish food but no tadpoles was processed by SPE as described above. The LC fractionation was carried out using a Lichrospher 100 RP-18 column (Merck, 250 x 4 mm, 5 μm particle size). A gradient elution was carried out using bidistilled water:methanol 95:5 (eluent A) and water:methanol 5:95 (eluent B) at a flow rate of 1.0 ml/min. The gradient program was 0–4 min 100% A, 4–22 min 0–100% B, 22–25.5 min 100% B, and re-equilibration to 100% A for 5 min. Fractions of samples A and V were collected every minute. Based on LC-HRMS analyses of the sample the one-minute fractions were combined as follows: Fraction A1-12/V1-12 10–18 min, A2-12/V2-12 18–21 min, A3-12/V3-12 21–25 min, A4-12/V4-12 25–30 min, A5-12/V5-12 30–40 min.

(c) Bioassays to test which fraction(s) contain the avoided chemical compound(s)

Between 10 February and 18 April 2012 we conducted bioassays in the field to find out which fraction(s) contained the compounds(s) avoided by the frogs. We set up 20 pairs of artificial phytotelmata for each trial as described in (a) and conducted water changes every other day with either clean rainwater or with water containing either the total samples (Vtotal or Atotal) or one of the fractions (V1-12 - V5-12 or A1-12 - A5-12) produced in the lab. To mix the dried samples into water, we acted in accordance with the concentrations used in trials with fresh tadpole water before (compare (a) and [18]). We first dissolved our dried samples in 15 μl rainwater per tadpole sample and deep-froze them in measured portions. Those portions were then mixed with rainwater in the field, mixing 15 μl each (i.e. one tadpole that used water in the lab for one day) in 25 ml water. Due to limited chemical compounds, we had to use a smaller amount of water in our cups than before, but the concentration stayed the same. In trials with samples from R. variabilis we filled 10 ml in each cup and in trials with samples from H. azureiventris we used 15 ml per cup. Despite this difference, cups were still frequently used for both tadpole and egg clutch depositions and there was no disparity between cups of the same experiments which might have influenced the results [24].

(d) Second fractionation of the chemical compounds from the tadpole water

During the bioassays after the first fractionation we found several of the tested fractions to be avoided by Ranitomeya variabilis. Because each of them was composed of various compounds, a finer fractionation was conducted subsequently. From 18 May to 19 June 2012 and 23 August 2012 to 4 January 2013 water samples of tadpoles from Trier University were collected daily as described in (b). The number of tadpoles varied between five and 23 in Ranitomeya variabilis and two and 15 in Hyloxalus azureiventris. Chemical compounds were extracted at the Soil Science Department, Trier University, in the way described in (b). Of R. variabilis 1089 tadpole samples were thereafter not further processed and kept in dried form for a second test of the total sample (Vtotal) in the field. The remaining samples (1197 of R. variabilis and 978 of H. azureiventris) were further fractionated at UFZ Leipzig. We thereby focused on the fractions that were avoided in the previous bioassays (see results) or where we found similarities between fractions of the two species.

Fractionation of the R. variabilis samples was conducted in the same manner as described in (b), but the fractions collected every minute were pooled differently and those with no effects in 2012 were not used. Fractions were combined and named as follows: Fraction V1a-13 13–16 min, V1b-13 16–20 min, V2c-13 20–23 min. The numbers in the new names refer to the according fractions of the previous year, while the letters give them a new counting system. Unlike the R. variabilis samples, the H. azureiventris samples were at first fractionated and pooled together as in 2012. For a finer fractionation with different selectivity we fractionated fraction A2-12 (avoided in 2012, see results) using a Polaris 5 Amide C18 column (Varian, 150 x 4.6 mm, 5 μm particle size). A gradient elution was carried out using bidistilled water:methanol 50:50 (eluent A) and water:methanol 5:95 (eluent B) at a flow rate of 1.0 ml/min. The gradient program was 0–10 min 100% A, 10–15 min 0–100% B, 15–16 min 100% B, and re-equilibration to 100% A for 3.5 min. Fractions of A2-12 were collected every minute and combined as follows: Fraction A2a-13 0–7 min, A2b-13 7–15 min, A2c-13 15–20 min.

(e) Bioassays to test which fraction(s) contain the avoided chemical compound(s)

Bioassays to test which fractions/compounds were avoided by the frogs were conducted in the field, in the same Peruvian sites as mentioned above, with artificial phytotelmata again (compare (a) and (c)). Between 5 February 2013 and 24 May 2013 we offered the fractions and mixtures of fractions to the frogs. The mixtures of the fractions of H. azureiventris were tested for the case that the compounds were not active in isolated form. We tested the mix of A2a-13 and A2b-13 (hereafter called A2ab-13), A2c-13, the mix of A2a-13 and A2c-13 (hereafter called A2ac-13), and A2b-13. Thus, with the combination of single and mixed fractions, we were able to determine if a single cue was avoided by the frogs as well. Dried samples were diluted in the same manner as described in (c) with one difference: due to the non- or only nearly significant results for R. variabilis in 2012 we doubled the concentration of both the total sample as well as the fractions of this species used in 2013. The amount of water used in the cups in the forest was 10 ml each.

(f) Instrumental analysis and LC-HRMS data analysis

To determine the chemical cues by LC-HRMS, we used the Lichrospher 100 RP-18 column (250 x 4 mm, 5 μm particle size) and gradient described above. The LC was connected to an ESI source and an LTQ Orbitrap XL mass spectrometer. Samples were analyzed in positive and negative full scan mode at a nominal resolving power of 100,000 with data-dependent acquisition of MS/MS spectra at a resolving power of 15,000 triggered by the most and second-most intense peaks in the chromatograms.

The compound identification procedure followed the nontarget screening approach described by Hug et al. [25]. Briefly, fractions were finally analyzed by LC-HRMS using a Kinetex Core-Shell C18 column (100 mm × 3.0 mm; 2.6 μm; Phenomenex) and a gradient elution with water (A) and methanol (B) both containing 0.1% formic acid at a flow rate of 0.2 ml/min. Deconvolution of the full scan chromatograms and peak detection was carried out using the software MZmine 2.9. From the obtained peak lists, those peaks occurring in blank and fish food control samples were removed. For the peaks remaining in active fractions, molecular formulas were calculated based on accurate masses and isotope patterns, and the Chemspider and KEGG databases were searched for all candidate structures of the particular molecular formula. To exclude unlikely candidate structures, we employed MS2 fragmentation prediction and retention time prediction as detailed in Hug et al. [25].

(g) Statistical analysis

For each treatment tested in (a), (c) and (e) we pooled egg clutch and tadpole depositions because chemical cues of both species are avoided for both deposition types [19]. We compared the deposition frequencies in each water type (clean or treated) using a G-test [26], which is suggested to be better for limited observations [27,28]. Since each dataset had less than 200 observations, we adjusted the G-test following Williams [29]. As a null hypothesis, we assumed that the frequency of deposition events was random (0.5 in either pool).

Because the switch from rainy to dry season is known to have an effect on the deposition decisions of parental Ranitomeya variabilis [30], we conducted a changepoint analysis in R (R Development Core Team [31]) with the rainfall measurements taken during each field season, using the package `changepoint´ [32,33]. While there was a significant seasonal change in 2011 (compare [30]), no change could be measured in 2012 and 2013 (bootstrap p > 0.05). All data were taken during the rainy season. We therefore did not have to split our data before analysis.


(a) Sorbent to extract the chemical cues

Three different sorbents were tested to extract potential cues or signals from water used by tadpoles of the frogs Ranitomeya variabilis and Hyloxalus azureiventris. The extraction efficiency was characterized by in-situ avoidance testing. The water used by R. variabilis or H. azureiventris was still avoided by parental R. variabilis for tadpole and egg clutch deposition after extraction with activated carbon (24 in clean and 9 in extracted R. variabilis water, G = 6.97, p < 0.01; 31 in clean and 3 in filtered H. azureiventris water, G = 26.45, p < 0.001; Fig 2). Applying passive sampling with Chromabond HR-X could not sufficiently bind the avoided chemical cues either. Parental R. variabilis still preferred clean over treated water significantly (24 in clean and 4 in passively filtered R. variabilis water, G = 14.51, p < 0.001). However, water used by R. variabilis tadpoles extracted with a Discovery DSC-18 passive sampler was not avoided anymore (18 in clean and 13 in passively filtered R. variabilis-water, G = 0.08, p = 0.37; Fig 2). An influence of the beginning dry season on this result as shown by Schulte and Lötters [30] can be ruled out here. Experiments were stopped a few weeks after the seasonal change (compare changepoint in [30]) and treatments that involved Discovery DSC-18 were not only distributed uniformly after, but also before the changepoint (before changepoint: 4 depositions in clean and 4 in treated water, after changepoint: 14 in clean and 9 in treated water).

Fig 2. Results of the preliminary trials to find a sorbent to extract the active tadpole compounds.

Ratio (in percent) of offspring depositions in clean water (grey narrow bars) and water used by tadpoles and treated with sorbents afterwards (black narrow bars). The expected distribution (50:50) is shown in lighter shades of grey in the background. V and A refer to R. variabilis and H. azureiventris, respectively. When frogs showed a significant preference for the clean water we assumed that the treated water still contained tadpole compounds, i.e. the sorbents did not filter them sufficiently out of the water to end the avoidance behavior. Only after the treatment with DSC-18 did frogs not show avoidance of tadpole-treated water. * p < 0.05, ** p < 0.01, *** p < 0.001.

(b) Ranitomeya variabilis fractions

The separation of DSC-18 extracts on C18 provided several fractions which were tested for avoidance in-situ. The results of these bioassays conducted in 2012 and 2013, testing the frogs’ reactions towards the different chemical fractions obtained from Ranitomeya variabilis tadpole-water, are shown in Table 1 and Fig 3. None of the fractions tested in 2012 were avoided significantly, but frogs showed a strong tendency towards the clean water over fraction 1 (V1-12). When splitting V1-12 into smaller fractions in 2013, we found a significant avoidance of fraction b (V1b-13). One of the compounds found in V1b-13 (see Table 2) was detected in the water of Hyloxalus azureiventris as well.

Fig 3. Bioassay results after offering the fractionated compounds of the tadpoles to the frogs.

The ratio of offspring depositions in clean water (grey narrow bars) and in water treated with chemically processed (A) R. variabilis and (B) H. azureiventris substances (black narrow bars) is shown in percent. V and A stand for R. variabilis and H. azureiventris, respectively. The distribution of the total samples (Vtotal and Atotal) is shown with lighter grey shades in the background. Connecting lines leading from results from 2012 to 2013 show which fractions were further processed in 2013 and contain identical compounds. * p < 0.05, ** p < 0.01, *** p < 0.001.

Table 1. Pooled egg and tadpole depositions placed by parental Ranitomeya variabilis in cups with clean or treated water.

Table 2. Compounds found in fractions avoided in the bioassays by Ranitomeya variabilis.

(c) Hyloxalus azureiventris fractions

In 2012 parental Ranitomeya variabilis avoided both the total sample of the chemically processed Hyloxalus azureiventris tadpole-water (Atotal) as well as fractions 2 (A2-12) and 4 (A4-12) (Table 2, Fig 3). However, the avoidance of fraction 4 was much weaker compared to fraction 2 (compared with a Fisher’s exact test [34]: p = 0.049), as well as compared to the usually very strong general avoidance of Hyloxalus azureiventris tadpoles (see [19]; Fisher’s exact test: p = 0.009). Fraction 4 was therefore tested again without further fractionation in the following year and its avoidance could not be confirmed. Fraction 2 however was avoided very strongly in 2012 and was therefore further fractionated in 2013 (Table 2, Fig 3). The mix of the fractions A2a-13 and A2c-13 (= A2ac-13) was avoided significantly. Other fractions or mixes of fractions did not show significant results. One of the compounds found in A2ac-13 (compound Azuri-2, see below) was also detected in the water of R. variabilis (compound Vari-2). Nonetheless, this compound was not active independently (but in combination with Azuri-1 or Vari-1, respectively), because it could also be found in A2ab-13 (compare Table 1), which was not avoided by the frogs.

(d) Identification of chemical cues

In the final active fractions of 2013 a small number of compounds could be detected which were present in the active fractions of 2012 as well (Table 2). Fraction V1b-13 from 2013 had two compounds in common with fraction V1-12 from 2012. In LC-HRMS these compounds were detected in positive ion mode as ions with m/z 134.0594 (compound Vari-1) and 171.1488 (compound Vari-2), respectively. The latter was also found in fraction A2a-13 of fraction-mix A2ac-13 (accordingly compound Vari/Azuri-2). Since this compound did not exhibit activity in fraction mix A2ab-13, we further analyzed A2c-13 (the second fraction in A2ac-13). Compound Azuri-1 found in this fraction showed an ion at m/z 228.1962, but it could not be found in A2-12.

Based on the accurate mass and isotope patterns and the underlying assumption that protonation in ESI+ occurred, we determined molecular formulas for these compounds (Table 2) and searched the Chemspider database for corresponding structures. For compound Vari-1 (C8H7NO) 186 candidate structures were found, for compound Vari/Azuri-2 (C9H18N2O) 1653 and for compound Azuri-1 (C13H25NO2) 582.

Based on the approach by Hug et al. [25], we reduced the number of candidate structures by comparing predicted and measured MS2 spectra using the software MetFrag [35]. Candidate structures with a score < 0.8 were excluded. Subsequently, measured and predicted LC retention (as expressed by the chromatographic hydrophobicity index, CHI) was compared based on the method of Ulrich et al. [36] and all candidates removed which were not within ±6 units of CHI.

This candidate reduction procedure resulted in 11 candidates for compound Vari-1 (S1 Table), of which two were considered unlikely due to an anticipated lack of stability. The remaining candidate chemicals include hydroxylated benzonitriles, substituted 1H-pyrrole and pyridine carbaldehydes, and pyridinones. For compound Vari/Azuri-2 14 candidate structures were left (S2 Table), most of them glycinamides. For compound Azuri-1 the candidate reduction procedure did not provide any meaningful MS/MS spectra at different collision energies. Thus, no reduction of candidate structures based on MS/MS data was possible.


Our comparative chemical analysis of the fractions that triggered avoidance behavior in adult Ranitomeya variabilis revealed that one compound (C9H18N2O) was produced by tadpoles of both R. variabilis and Hyloxalus azureiventris. This finding would be concordant with our hypothesis (i) that the avoidance behavior of R. variabilis of both con- and heterospecific tadpoles is based on a reaction to the same chemical compound. The shared compound is most likely a glycinamide, known to contain salty tastants which are shown to act as strong stimuli of the tongue epithelium and the gustatory nerve of bullfrogs (Lithobates catesbeianus [3739]). This finding is well in agreement with our observations and suggests that glycinamides, being recognized by the adults via taste, play a role in intra- and interspecific communication in poison frogs. It might be a compound that tadpoles of all species have in common and is possibly used by adult R. variabilis to identify if a phytotelm is generally occupied or not. However, for H. azureiventris the fraction containing this compound could only be shown to be active (i.e. triggering avoidance in the frogs) in combination with one other fraction, but not in another combination. Hence, we suggest that the shared compound is only biologically significant to the frogs in combination with a second compound that gives the frogs more species-specific information about the tadpole species. This leads us to reject hypothesis (i) and support the alternative hypothesis (ii) stating that the active compounds produced by inter- and intraspecific tadpoles are different, as at least one of the avoided compounds is unique in each species (C8H7NO in R. variabilis and C13H25NO2 in H. azureiventris).

Several of the possible structural formulas for the compound unique to R. variabilis tadpoles are pyridine alkaloids, but none of them has ever been found in anurans. Furthermore, R. variabilis tadpoles lack alkaloids (pers. comm. Ralph Saporito). With regard to the definition of intraspecific chemical communication between adult R. variabilis (the receiver) and their tadpoles (the sender), we specified the chemical compounds as chemical cues, because the avoidance is advantageous to the frogs (i.e. their offspring), but not to the tadpoles [18]. However, this definition might be context dependent. While conspecific tadpoles are strongly avoided during the rainy season, with the change to the dry season they are not avoided, but preferred instead. This can be interpreted as a kind of offspring-provision: older tadpoles are fed with younger ones in order to help them survive the dry season [30]. However, since in our study we only focused on the rainy season, we cannot tell if this feeding behavior is triggered by the same chemical compounds as the avoidance behavior, possibly initiating the evolution into chemical signaling (as suggested by Steiger et al. [8]), or if it is triggered by another compound that might be defined as a chemical signal (i.e. pheromone). A switch of preference between rainy and dry season towards H. azureiventris tadpoles could not been shown [19].

Regarding the communication between adult R. variabilis and H. azureiventris tadpoles, the results that at least one of the actively avoided chemical compounds is species specific brings us a step further towards the definition of the interspecifically operating chemicals. We suggest that these species-specific chemical compounds are either chemical signals or chemical cues which are being co-opted for chemical signaling. The reason for this is the obvious benefit the tadpoles of H. azureiventris receive when R. variabilis avoid offspring deposition in the same pools. Not only are the predatory tadpoles of R. variabilis dangerous to the omnivorous, non-predatory H. azureiventris tadpoles, but also competition in the relatively small water bodies might be a relevant factor [40]. We do not know how the compounds are excreted (e.g. with feces or through the epidermis) or if they are intentionally released by the H. azureiventris tadpoles. If they are intentionally released, this would be a case of chemical signaling. If accidently released by-products are avoided, this would suggest they are chemical cues. However, since the avoidance of these cues by the receiver (R. variabilis) turns out to be beneficial to the sender (H. azureiventris), this might be an example of selection for signaling, as described by Steiger et al. [7] and Wyatt [8].

Assuming the compounds released by H. azureiventris and recognized by R. variabilis are an interspecific signal, their ecological context remains unclear. Due to its advantage for the sender, it can be either classified as an allomone (advantageous only to the sender) or a synomone (advantageous to both the sender and the receiver). The fact that the H. azureiventris tadpoles do not seem to imitate the compounds of the R. variabilis tadpoles (at least one compound is different) is evidence against a classification as allomones, which are usually replicates of intraspecific pheromones (see below), and rather for the classification as synomones. This classification is further supported by the fact that R. variabilis is able to distinguish the heterospecific compounds as such and nevertheless avoid them even more intensely than the compounds of its conspecific tadpoles [19].

However, to confirm this classification, the question to be discussed is whether the avoidance of the signal emitted by larval H. azureiventris is advantageous to R. variabilis (i.e. its offspring) or not. Since Ranitomeya tadpoles are able to feed on H. azureiventris tadpoles (observation J. L. Brown) the avoidance of this potential prey does not seem to be beneficial to R. variabilis. This would suggest that the signals are not synomones, but allomones, i.e. beneficial only to the sender. The use of allomones is primarily known from predator-prey systems, where predators lure prey animals with imitations of prey specific pheromones (e.g. [41,42]). This would be the case (in the opposite direction) if H. azureiventris imitates the cues of the heterospecific tadpoles in order to be avoided by adult R. variabilis, that is a chemical-based version of Batesian mimicry (which turned out not to be the case). Another possibility are so-called defensive allomones which are toxins used by prey to repel predators (for review see [43]). However, whether those repellents that indicate the toxicity of a prey animal are indeed defined as allomones [9] or rather as synomones [44] is controversial (both definitions were used in the literature to describe avoidance of signals that indicate inedible prey; as allomones: [43,45,46]; as synomones: [47,48]).

Adult poison frogs are a classic example of signaling toxicity. While they use visual signals (aposematic coloration) to warn predators such as birds and mammals [4952], there is evidence that predators such as spiders, ants and possibly snakes receive chemical signals from the frogs as an indicator of toxicity [5355]. Due to the fact that toxins (skin alkaloids) are assimilated through food uptake by adult dendrobatid frogs (for review see [56]), it is thought that toxins do not exist in their tadpoles. However, recent research has shown that Oophaga pumilio tadpoles can sequester alkaloids from nutritive eggs provided by their parents [57]. But because H. azureiventris does not egg-feed its offspring, we do not believe that toxins (i.e. repellent allomones/synomones) are involved in the predator-prey relationship between larval R. variabilis and H. azureiventris. Furthermore, even the toxicity of adult H. azureiventris is controversial [58].

Because H. azureiventris tadpoles grow bigger than R. variabilis tadpoles and R. variabilis tadpoles are very small when deposited (Gosner stage 25 [21]), we assume that the avoidance by parental R. variabilis is associated with competition. Individual small- or medium-sized R. variabilis larvae are apparently not capable of preying on larger H. azureiventris tadpoles (N = 11 per species, paired for seven days in 100 ml water contained in artificial pools between 20 May and 2 June 2006; J. L. Brown, unpublished data). We therefore suggest that the chemical signal released by larval H. azureiventris and avoided by parental R. variabilis might be a synomone and that the avoidance is advantageous for both species in terms of preventing competition (and for young H. azureiventris, preventing predation). However, the exact benefits and costs for both species are still to be determined.

There are many examples for synomones between plants and animals, for example between flowers and pollinators (e.g. [59,60]) or between parasite-hosting plants and natural enemies of those parasites (e.g. [61,62]). Synomones between different animal species are less common (aside from predator-warning toxins, depending on the definition, see above). The best known examples are synomones released by sea anemones that initialize their mutualistic relationship with clown fish [63,64]. Chemical communication based on synomones between two species of vertebrates (aside from predator-warning toxins), lacks further examples, to the best of our knowledge. However, by combining ecological studies with chemical analyses, such a communication system might turn out to be more common in mutual competition situations between other vertebrates than is currently known.

Supporting Information

S1 File. Chemicals and analytical instruments.



S1 Table. Chemspider candidate structures for unknown compound Vari-1 (C8H7NO) remaining after candidate selection procedure.



S2 Table. Chemspider candidate structures for unknown compound Vari/Azuri-2 (C9H18N2O) remaining after candidate selection procedure.




We especially thank Jason L. Brown for the contribution of his records about the interactions between Ranitomeya variabilis and Hyloxalus azureiventris and Daniela Rößler for help with captive frogs and language editing our manuscript. We further thank the technicians in Leipzig (particularly Angela Sperreuter) for help in the lab and Evan Twomey for language edits. The Soil Science Department of Trier University, especially Sören Thiele-Bruhn, allowed us to use their lab. Ralph Saporito helped us with the identification of possible alkaloids. The Asociación de Productores de Ranas Venenosas, Progreso, in particularly Rainer Schulte, allowed us to use their field station. Research permits were obtained from the Peruvian Ministry of Agriculture (DGFFS) in Lima (authorization No. 0200-2011-AG-DGFFS-DGFFS, 0176-2012-AG-DGFFS-DGFFS and 416-2013-AG-DGFFS-DGFFS) and were conducted in the framework of a research cooperation between Trier University and Centro de Ornitología y Biodiversidad (CORBIDI; special thanks to Pablo J. Venegas for support). Captive breeding groups were declared at the nature conservation authority of Trier. A free academic license of Instant JChem for the processing of chemical structures was provided by ChemAxon Ltd. (

Author Contributions

Conceived and designed the experiments: LMS WB MK TS SL. Performed the experiments: LMS MK TS. Analyzed the data: LMS MK TS. Wrote the paper: LMS SL MK.


  1. 1. Wilson EO. Chemical communication within animal species. In: Sondheimer E, Simeone JB, editors. Chemical ecology. New York: Academic Press; 1970. pp. 133–155.
  2. 2. Otte D. Effects and functions in the evolution of signaling systems. Annu Rev Ecol Syst. 1974; 5: 385–417.
  3. 3. Weiss I, Rössler T, Hofferberth J, Brummer M, Ruther J, Stökl J. A nonspecific defensive compound evolves into a competition avoidance cue and a female sex pheromone. Nature communications. 2013; 4: 1–11.
  4. 4. Agosta WC. Chemical communication: the language of pheromones. New York: Scientific American Library; 1992.
  5. 5. Wyatt TD. Pheromones and animal behaviour: communciation by smell and taste. Cambridge: Cambridge University Press; 2003.
  6. 6. Bradbury JW, Vehrencamp SL. Principles of animal communication. Massachusetts: Sunderland; 2011.
  7. 7. Steiger S, Schmitt T, Schaefer HM. The origin and dynamic evolution of chemical information transfer. Proc R Soc B Biol Sci. 2011; 278: 970–979. doi: 10.1098/rspb.2010.2285. pmid:21177681
  8. 8. Wyatt TD. Pheromones and signature mixtures: defining species-wide signals and variable cues for identity in both invertebrates and vertebrates. J Comp Physiol A. 2010; 196: 685–700. doi: 10.1007/s00359-010-0564-y. pmid:20680632
  9. 9. Nordlund DA, Lewis WJ. Terminology of chemical releasing stimuli in intraspecific and interspecific interactions. J Chem Ecol. 1976; 2: 211–220.
  10. 10. Fraker ME, Hu F, Cuddapah V, McCollum SA, Relyea RA, Hempel J, et al. Characterization of an alarm pheromone secreted by amphibian tadpoles that induces behavioral inhibition and suppression of the neuroendocrine stress axis. Horm Behav. 2009; 55: 520–529. doi: 10.1016/j.yhbeh.2009.01.007. pmid:19470366
  11. 11. Ferland-Raymond B, March RE, Metcalfe CD, Murray DL. Prey detection of aquatic predators: Assessing the identity of chemical cues eliciting prey behavioral plasticity. Biochem Syst Ecol. 2010; 38: 169–177.
  12. 12. Saleh N, Scott AG, Bryning GP, Chittka L. Distinguishing signals and cues: bumblebees use general footprints to generate adaptive behaviour at flowers and nest. Arthropod-Plant Interactions. 2007; 1: 119–127.
  13. 13. Zimmermann H, Zimmermann E. Etho-Taxonomie und zoogeographische Artengruppenbildung bei Pfeilgiftfröschen (Anura: Dendrobatidae). Salamandra. 1988; 24: 125–160.
  14. 14. Varga L. Ein interessanter Biotop der Biocönose von Wasserorganismen. Biol Zentralbl. 1928; 48: 143–162.
  15. 15. Brown JL, Twomey E, Morales V, Summers K. Phytotelm size in relation to parental care and mating strategies in two species of Peruvian poison frogs. Behaviour. 2008; 145: 1139–1165.
  16. 16. Summers K. The effects of cannibalism on Amazonian poison frog egg and tadpole deposition and survivorship in Heliconia axil pools. Oecologia. 1999; 119: 557–564.
  17. 17. Schulte LM. Feeding or avoiding? Facultative egg-feeding in a Peruvian poison frog (Ranitomeya variabilis). Ethol Ecol Evol. 2014; 26: 58–68.
  18. 18. Schulte LM, Yeager J, Schulte R, Veith M, Werner P, Beck LA, et al. The smell of success: choice of larval rearing sites by means of chemical cues in a Peruvian poison frog. Anim Behav. 2011; 81: 1147–1154
  19. 19. Schulte LM, Lötters S. A danger foreseen is a danger avoided: how chemical cues of different tadpoles influence parental decisions of a Neotropical poison frog. Anim Cogn. 2014; 17: 267–275. doi: 10.1007/s10071-013-0659-2. pmid:23852187
  20. 20. Brack W. Effect-directed analysis: a promising tool for the identification of organic toxicants in complex mixtures? Anal Bioanal Chem. 2003; 377: 397–407. pmid:12904950
  21. 21. Gosner KL. A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica. 1960; 16: 183–190.
  22. 22. Greenberg B. Animal litter formulation, patent number 4638763. United States Patent. 1987.
  23. 23. Brown JL, Morales V, Summers K. Home range size and location in relation to reproductive resources in poison frogs (Dendrobatidae): a Monte Carlo approach using GIS data. Anim Behav. 2009; 77: 547–554.
  24. 24. Brown JL, Morales V, Summers K. Divergence in parental care, habitat selection and larval life history between two species of Peruvian poison frogs: an experimental analysis. J Evol Biol. 2008; 21: 1534–1543. doi: 10.1111/j.1420-9101.2008.01609.x. pmid:18811668
  25. 25. Hug C, Ulrich N, Schulze T, Brack W, Krauss M. Identification of novel micropollutants in wastewater by a combination of suspect and nontarget screening. Environ Pollut. 2014; 184: 25–32. doi: 10.1016/j.envpol.2013.07.048. pmid:24012788
  26. 26. Woolf B. The log likelihood ratio test (the G-test); methods and tables for tests of heterogeneity in contingency tables. Ann Hum Genet. 1957; 21: 397–409. pmid:13435648
  27. 27. Gotelli NJ, Ellison AM. A primer of ecological statistics. Sunderland, Mass: Sinauer; 2004.
  28. 28. Sachs L. Angewandte Statistik: Planung und Auswertung—Methoden und Modelle. Berlin-Heidelberg-New York: Springer; 1974.
  29. 29. Williams DA. Improved likelihood ratio tests for complete contingency tables. Biometrika. 1976; 63: 33–37.
  30. 30. Schulte LM, Lötters S. The power of the seasons: rainfall triggers parental care in poison frogs. Evol Ecol. 2013; 27: 711–723.
  31. 31. R Development Core Team. R: A language and environment for statistical computing. Vienna. ISBN 3-900051-07-0, URL R Foundation for Statistical Computing. 2011.
  32. 32. Killick R, Eckley IA. Changepoint: An R package for changepoint analysis. wwwlancsacuk/~killick/Pub/KillickEckley2011pdf. 2011.
  33. 33. Killick R, Eckley IA. An R package for changepoint analysis. (R package). 2011.
  34. 34. Fisher RA. On the interpretation of X² from contingency tables, and the calculation of P. J Roy Stat Soc Ser A (Stat Soc). 1922; 85: 87–94.
  35. 35. Wolf S, Schmidt S, Müller-Hannemann M, Neumann S. In silico fragmentation for computer assisted identification of metabolite mass spectra. BMC Bioinformatics. 2010; 11: 148. doi: 10.1186/1471-2105-11-148. pmid:20307295
  36. 36. Ulrich N, Schüürmann G, Brack W. Linear solvation energy relationships as classifiers in non-target analysis—a capillary liquid chromatography approach. J Chromatogr. 2011; 1218: 8192–8196. doi: 10.1016/j.chroma.2011.09.031. pmid:21968343
  37. 37. Soeda H, Sakudo F, Kajiya H, Okabe K. Epithelial responses of the frog (Rana catesbeiana) tongue to glycyl-amides and their relationship to gustatory nerve responses. Comp Biochem Physiol, A: Physiol. 1994; 107: 85–93.
  38. 38. Soeda H, Sakudo F. Epithelial responses to glycinamide and glycylglycinamide in the frog (Rana catesbeiana) tongue. Comp Biochem Physiol, A: Physiol. 1991; 100: 315–321.
  39. 39. Kiyonori Y, Yukari K, Kenzo K. Taste receptor mechanism of salts in frog and rat. Comp Biochem Physiol, A: Physiol. 1986; 85: 501–507.
  40. 40. Summers K, McKeon CS. The evolutionary ecology of phytotelmata use in Neotropical poison frogs. Misc Publ Mus Zool Univ Mich. 2004; 193: 55–73.
  41. 41. Haynes KF, Gemeno C, Yeargan KV, Millar J, Johnson KM. Aggressive chemical mimicry of moth pheromones by a bolas spider: how does this specialist predator attract more than one species of prey? Chemoecology. 2002; 12: 99–105.
  42. 42. Mizuno T, Yamaguchi S, Yamamoto I, Yamaoka R, Akino T. "Double-trick" Visual and chemical mimicry by the juvenile orchid mantis Hymenopus coronatus used in predation of the oriental honeybee Apis cerana. Zool Sci. 2014; 31: 795–801. doi: 10.2108/zs140126. pmid:25483791
  43. 43. Whitman DW, Blum MS, Alsop DW. Allomones: chemicals for defense. In: Evans DL, Schmidt JO, editors. Insect Defenses: adaptive mechanisms and strategies of prey and predators. Albany: State University of New York; 1990. pp. 229–251.
  44. 44. Dicke M, Sabelis MW. Infochemical terminology: based on cost-benefit analysis rather than origin of compounds? Funct Ecol. 1988; 2: 131–139.
  45. 45. Ruther J, Meiners T, Steidle JLM. Rich in phenomena-lacking in terms. A classification of kairomones. Chemoecology. 2002; 12: 161–167.
  46. 46. Schlee D. Ökologische Biochemie. Berlin: Spektrum Akademischer Verlag; 1999.
  47. 47. Augner M. Should a plant always signal its defence against herbivores? Oikos. 1994; 70: 322–332.
  48. 48. Sbarbati A, Osculati F. Allelochemical communication in vertebrates: kairomones, allomones and synomones. Cells Tissues Organs. 2006; 183: 206–219. pmid:17159346
  49. 49. Amézquita A, Castro L, Arias M, González M, Esquivel C. Field but not lab paradigms support generalisation by predators of aposematic polymorphic prey: the Oophaga histrionica complex. Evol Ecol. 2013; 27: 769–782.
  50. 50. Richards-Zawacki CL, Yeager J, Bart HPS. No evidence for differential survival or predation between sympatric color morphs of an aposematic poison frog. Evol Ecol. 2013; 27: 783–795. doi: 10.1521/pedi_2013_27_115. pmid:23795756
  51. 51. Saporito RA, Zuercher R, Roberts M, Gerow KG, Donnelly MA. Experimental evidence for aposematism in the dendrobatid poison frog Oophaga pumilio. Copeia. 2007; 2007: 1006–1011. pmid:18002130
  52. 52. Maan ME, Cummings ME. Poison frog colors are honest signals of toxicity, particularly for bird predators. Am Nat. 2012; 179: E1–E14. doi: 10.1086/663197. pmid:22173468
  53. 53. Szelistowski WA. Unpalatability of the poison arrow frog Dendrobates pumilio to the ctenid spider Cupiennius coccineus. Biotropica. 1985; 17: 345–346.
  54. 54. Fritz GN, Rand AS, Depamphilis C. The aposematically colored frog, Dendrobates pumilio, is distasteful to the large, predatory ant, Paraponera clavata. Biotropica. 1981; 13: 158–159.
  55. 55. Brodie ED, Tumbarello MS. The antipredator functions of Dendrobates auratus (Amphibia, Anura, Dendrobatidae) skin secretion in regard to a snake predator (Thamnophis). J Herpetol. 1978; 12: 264–265.
  56. 56. Saporito RA, Donnelly MA, Spande TF, Garraffo HM. A review of chemical ecology in poison frogs. Chemoecology. 2012; 22: 159–168.
  57. 57. Stynoski JL, Torres-Mendoza Y, Sasa-Marin M, Saporito RA. Evidence of maternal provisioning of alkaloid-based chemical defenses in the strawberry poison frog Oophaga pumilio. Ecology. 2014; 95: 587–593. pmid:24804437
  58. 58. Saporito RA, Spande TF, Garraffo HM, Donnelly MA. Arthropod alkaloids in poison frogs: a review of the'dietary hypothesis'. Heterocycles. 2009; 79: 277–297
  59. 59. Tan KH, Nishida R. Mutual reproductive benefits between a wild orchid, Bulbophyllum patens, and Bactrocera fruit flies via a floral synomone. J Chem Ecol. 2000; 26: 533–546.
  60. 60. Harrewijn P, Minks AK, Mollema C. Evolution of plant volatile production in insect-plant relationships. Chemoecology. 1994; 5: 55–73.
  61. 61. Boevé J-L, Lengwiler U, Tollsten L, Dorn S, Turlings TCJ. Volatiles emitted by apple fruitlets infested by larvae of the European apple sawfly. Phytochemistry. 1996; 42: 373–381.
  62. 62. Sabelis MW, De Jong MCM. Should all plants recruit bodyguards? Conditions for a polymorphic ESS of synomone production in plants. Oikos. 1988; 53: 247–252.
  63. 63. Murata M, Miyagawa-Kohshima K, Nakanishi K, Naya Y. Characterization of compounds that induce symbiosis between sea anemone and anemone fish. Science. 1986; 234: 585–586. pmid:17835564
  64. 64. Konno K, Qin G-W, Nakanishi K, Murata M, Naya Y. Synthesis of amphikuemin and analogs: a synomone that mediates partner-recognition between anemonefish and sea anemone. Heterocycles. 1990; 30: 247–251.