Introduction

Carnivorous plants use specialized traps to catch insects and absorb nutrients from their bodies. This unique feeding habit has been considered as an adaptation to nutrient-poor habitats because carnivorous plants can fill deficiency of nutrients in soil by digesting trapped insects (Thum 1988; Zamora et al. 1997; Thoren and Karlsson 1998). Probably due to this advantage, carnivory evolved at least six times independently in the angiosperms (Ellison and Gotelli 2001). Catching insects is, however, not always advantageous for carnivorous plants because insects may serve as both prey and pollinators (Ellison and Gotelli 2001). If the same insect species serves as both, catching pollinators may decrease pollination success. Most carnivorous plant species may avoid this Pollinator-Prey Overlap (PPO) by making traps functional only after a flowering season (temporal separation) or placing traps apart from flowers (spatial separation) (Jürgens et al. 2012). The degree of PPO has been studied in some sundew species (Drosera spp.) in which traps are arranged in a rosette close to the ground, whereas their flowers are located high at the top of erect stems. For D. pauciflora, Anderson (2010) demonstrated that no pollinators were caught in traps. Murza et al. (2006) showed that basal traps of another sundew D. anglica caught a minor pollinator species Thrips sp. (Murza et al. 2006). However, it comprised only 3% of all flower visitors and major pollinators (Diptera, 95% of the total) were not caught by the traps of D. anglica.

On the other hand, high PPO may be observed in highly selfing carnivorous species in which trapping pollinators may not reduce fitness under negligible inbreeding depression (Sciligo 2009; Jürgens et al. 2012). Sciligo (2009) tested this idea using autonomously selfing D. arcturi that showed no inbreeding depression, and showed that 96% of pollinator fly families were trapped. To test the expectation of high PPO in highly selfing species, further studies on other Drosera species are needed (Jürgens et al. 2012). Here, we examine this expectation in two selfing annuals Drosera makinoi Masamune and D. toyoakensis M. Watanabe in which the inbreeding coefficient is 0.497 and 0.260, respectively (Watanabe, unpublished), indicating that selfing rate is higher in D. makinoi. In D. makinoi and D. toyoakensis, flowers are close to trap leaves arising from flowering stems and pollinators may approach trap leaves more frequently.

We also examine whether the flowers of two Drosera species attract pollinator and prey insects. Because two species are moderately inbreeding, it may be unnecessary to attract many pollinators required for high outcrossing. On the other hand, it may be advantageous to attract more insects by flowers as prey (Zamora 1999; Salces-Castellano et al. 2016). To test how the flowers are attractive to pollinators and prey insects, we experimentally removed flowers of Drosera spp. In addition, we also experimentally removed flowers of co-flowering neighbor species because those could also increase the number of insects trapped by the two Drosera species.

Specific questions addressed in this paper are as follows. (1) How large is the degree of PPO in two selfing carnivorous species, D. makinoi and D. toyoakensis? (2) Does the experimental removal of the flowers of Drosera spp. decrease the numbers of insects trapped by Drosera spp.? (3) Does the experimental removal of the flowers of co-flowering neighbor plants decrease the number of insects trapped by Drosera spp.?

Materials and methods

Study species and sites

Drosera makinoi Masamune and D. toyoakensis M. Watanabe, previously included in D. indica (Watanabe et al. 2013; Kagawa 2015), are both annual plants with a single stem growing up to 10–20 cm. Linear and alternate trap leaves 5–7 cm long are scattered along the stem and trap leaves at the upper position are close to flowers (Fig. 1a). Traps have glandular hairs called tentacles and glands secrete sticky liquid and digestive enzyme (Sun et al. 2014). The two species are different in flower color: white in D. makinoi and pink in D. toyoakensis. Both Drosera species grow in open bogs of a few restricted localities and have been listed as threatened plants (Japan Wildlife 2015). Drosera makinoi has a scattered distribution in Japan and the distribution of D. toyoakensis is now restricted to Aichi Prefecture, Japan. Drosera toyoakensis germinates early in May and has three to four flowers at the top of the stem from July to September. Flowers open around 8 am and close by 1 pm. A flower blooms only once, and it does not open again after closing even if closing is due to strong rain or wind. About 500 seeds per flower mature in the middle of September, and the plants begin to wither in the middle of October. Drosera makinoi has similar phenology but a longer flowering period from July to October. We made observations and field experiments of D. makinoi in a Sphagnum bog at Watarase (Tochigi City, Tochigi Pref., Japan). The bog is about 200 m2 and neighboring to a population of common reed Phragmites australis (Cav.) Trin. ex Steud. (Poaceae). Studies of D. toyoakensis were carried out in a Sphagnum bog at Toyoake (N 35°4′26″ E 137°1′7″, Toyoake City, Aichi Pref., Japan). The bog of about 200 m2 is surrounded by a protection fence and is neighboring to paddy fields.

Fig. 1
figure 1

Drosera makinoi and two co-flowering neighbors at the study sites of Drosera spp. a Drosera mainoi. b Eriocaulon decemflorum, a co-flowering species growing with D. makinoi at Watarase site. c Lysimachia fortunei, a co-flowering species growing with D. toyoakensis at Toyoake site. Color figure online

Full size image

Flower visitor observations

To determine whether pollinator species were trapped by D. makinoi and D. toyoakensis, we observed flower visitors from August to September for three years (2012–2014). For D. makinoi, we observed pollinators from 8 am to 2 pm, and for D. toyoakensis, from 7 am to 2 pm while careful not to damage the bog. We observed each flower visitor until it visited five flowers of Drosera or it flew away before visiting five flowers, and recorded the taxonomic identity of insect species and whether it touched stamens and/or pistils. Some flower visitors of D. makinoi and D. toyoakensis visited Eriocaulon decemflorum Maxim. (Eriocaulaceae) and Lysimachia fortunei Maxim. (Primulaceae), respectively (Fig. 1b, c). The plant species was recorded if the flower visitor moved from Drosera species to another species (co-flowering neighbor species) and vice versa. A pollinator was defined as an individual that (1) touched a stamen and/or a pistil and (2) carried pollen grains. Flower visitors were caught with a butterfly net and fixed with 95% ethanol. Later in the laboratory, the entire body of each insect visitor was stained with 45% aceto-carmine solution and observed under a stereomicroscope to check whether it carried pollen grains. Based on results from the flower-visitor survey and the following prey survey, we calculated an index of pollinator-prey species overlap J as the number of trapped pollinator species divided by the total number of pollinator species; J = 0 when no pollinator species was trapped, while J = 1 when all pollinator species were trapped.

We computed all statistical analyses using R 3.1 (R Core Team 2010). To examine whether there is a difference in the number of flower visits between Drosera and co-flowering neighbor plant species, we used a generalized linear mixed model (GLMM) with the plant species, pollinator species, and the interaction term as explanatory variables, and survey date as a random factor. At first, we verified the significance of the interaction term by comparing the model with and without the interaction term. In Watarase, we adopted the model without the interaction term because the effect of the interaction was not significant. In Toyoake, the interaction term was significant because some pollinators visited Drosera flowers more frequently than flowers of neighbor plant species, and others visited flowers of neighbor plant species more frequently than Drosera flowers. Thus, we only used the data of pollinator species that visited flowers of both Drosera and neighbor plant species at least once in our survey. We used a likelihood-ratio test (type II) to determine any significant effect of explanatory variables. These GLMM analyses were performed with a log link and a Poisson distribution (Crawley 2005) using the R package “lme4” (Douglas et al. 2015). To describe the degree of pollinator sharing between Drosera and co-flowering neighbor plant species, we calculated the niche overlap index (NOI) (Pianka 1973); the index ranges from 0 (no overlap) to 1 (complete overlap) using the R package “EcoSimR” (Gotelli et al. 2015).

Flower removal experiments

To determine the effect of flowers of Drosera and neighbor plants (E. decemflorum for D. makinoi and L. fortunei for D. toyoakensis) on the number of insects trapped by D. makinoi or D. toyoakensis, we removed flowers of D. makinoi, D. toyoakensis and neighbor plants during the flowering periods at study sites of D. makinoi and D. toyoakensis: September 5 and 6, 2014 for D. makinoi and September 2 and 3, 2014 for D. toyoakensis. For each species, we set the following four classes of 40 cm × 40 cm quadrats at an interval of 40 cm or longer: (1) a treatment where flowers of both Drosera species and neighbor species were left, (2) a treatment where flowers of Drosera species were left but flowers of neighbor plants were removed, (3) a treatment where flowers of neighbor plants were left but flowers of Drosera species were removed, and (4) a treatment where flowers of both Drosera species and neighbor plants were removed. For each treatment, we set one quadrat containing 20 individuals of each Drosera species and co-flowering neighbor species. When the quadrat had more than 20 individuals, we cut flowers of excessive individuals to equalize the sample size; here, we regarded the sample size as 20 by considering a plant as a unit of replication. On each observation day (from 6 am to 7 am), we removed all flower buds of Drosera and/or neighbor plants that were just before flowering and we removed all dead insects on the surface of traps. For both Drosera species, we collected prey insects on traps from noon to 1 pm to identify species and recorded the number of prey individuals per day per plant in each quadrat. We then measured the body length of a prey insect, distance between each prey insect and the nearest flower, the presence/absence of flowers, and the height (from the ground) of traps where the prey was trapped.

To determine whether the flowers of Drosera species and neighbor species affected the number of prey individuals trapped, we used a generalized linear model (GLM) with the number of insects trapped per day per plant, the number of pollinators trapped per day per plant or the number of non-pollinators trapped per day per plant as the response variable. At first, the presence/absence of flowers of Drosera, the presence/absence of flowers of neighbor species and the interaction of these two variables were considered as the explanatory variables. Because the interaction effect on the response variable was not significant, we used a model without the interaction term to test effects of the two variables. Because the observed number of prey insects showed over-dispersion, we used a negative-binomial error distribution and a log link (Zuur et al. 2009) with a function “glm.nb” in the R package “MASS” (Venables and Ripley 2002).

Results

Pollinator-prey species overlap

In D. makinoi, five pollinator species of Diptera, Hymenoptera and Lepidoptera were recorded (Table 1a, Fig. 2a) and all five pollinator species except Mathias oberthueri (Lepidoptera) were trapped as prey (Table 1a). In D. toyoakensis, six pollinator species of Diptera and Lepidoptera were recorded (Table 1b, Fig. 2b) among which only a syrphid fly Sphaerophoria menthastri (Diptera) was trapped as prey (Table 1b). The index of pollinator-prey species overlap J was 0.83 ± 0.21 (mean ± SD, n = 5) in D. makinoi and 0.071 ± 0.18 (mean ± SD, n = 6) in D. toyoakensis; it was positive in every month in D. makinoi, but positive only in September 2014 in D. toyoakensis.

Table 1 Pollinator and prey species of Drosera and co-flowering neighbor species
Full size table
Fig. 2
figure 2

Monthly variations of pollinator and prey species composition. Green, blue and red bars respectively indicate the number of prey species that were not regarded as pollinators, the numbers of species regarded as both prey and pollinators, and the number of pollinator species not regarded as prey. a Drosera makinoi, b D. toyoakensis

Full size image

A syrphid fly (S. menthastri, Fig. 3) was the most frequent visitor to flowers of D. makinoi (22.6% of all flower visits; mean ± SD = 1.3 ± 1.8) and D. toyoakensis (67.3% of all flower visits; mean ± SD = 3.3 ± 7.3). Sphaerophoria menthastri carried 4.6 ± 3.2 (mean ± SD, n = 5) pollen grains of D. makinoi and 14 ± 8.2 (mean ± SD, n = 5) pollen grains of D. toyoakensis.

Fig. 3
figure 3

The most frequent pollinator species: Sphaerophoria menthastri. The hoverfly S. menthastri was a pollinator species common to Drosera makinoi and D. toyoakensis. a S. menthastri landing on a D. toyoakensis flower. b S. menthastri landing on a D. makinoi flower. c S. menthastri trapped by D. makinoi (also trapped in D. toyoakensis, not shown). Color figure online

Full size image

Four pollinator species visited E. decemflorum, a co-flowering neighbor species of D. makinoi. All four species were overlapped with pollinator species of D. makinoi and three of four species were overlapped with prey species of D. makinoi (Table 1a). Drosera makinoi and E. decemflorum showed a large overlap of their pollinator species (NOI = 0.761). There was no significant difference between the number of flower visits in D. makinoi and E. decemflorum (Fig. 4a GLMM P = 0.730). Seven pollinator species visited L. fortunei, a co-flowering neighbor species of D. toyoakensis. Four out of seven species were overlapped with pollinator species of D. toyoakensis and one of seven species was overlapped with prey species of D. toyoakensis (Table 1b). Drosera toyoakensis and L. fortunei showed a large overlap of their pollinators (NOI = 0.834). The effect of the interaction of plant species and pollinator species on the number of flower visits was significant (GLMM P < 0.001), and some pollinator species visited D. toyoakensis more frequently (e.g. Eurema hecabe, 5.5 ± 5.5 times on D. toyoakensis, 2.5 ± 0.50 times on L. fortunei) while S. menthastri visited L. fortunei more frequently (4.5 ± 7.6 times on D. toyoakensis, 9.7 ± 8.5 times on L. fortunei). The number of flower visits by pollinators was significantly higher in L. fortunei than in D. toyoakensis when we used the data of pollinator species that visited D. toyoakensis and L. fortunei at least once in our survey (Fig. 4b, GLMM Estimate ± SE = 1.8 ± 0.82, P < 0.001).

Fig. 4
figure 4

Number of pollinator visits for Drosera and co-flowering neighbor species. a There was no significant difference in pollinator visits between D. makinoi and a co-flowering neighbor species Eriocaulon decemflorum at the Watarase site (GLMM P = 0.73). b Pollinators visited a co-flowering neighbor species Lysimachia fortunei more frequently than D. toyoakensis at the Toyoake site (GLMM P < 0.001)

Full size image

Flower removal experiments

Flowers of D. makinoi had no significant effect on the number of insects trapped, whereas flowers of a co-flowering neighbor species E. decemflorum had a significant positive effect on the number of both pollinator and non-pollinator species trapped (Table 2a, Fig. 5a–c). Flowers of D. toyoakensis had no significant effect on the number of insects trapped, whereas flowers of a co-flowering neighbor species L. fortunei had a significant positive effect on the number of pollinator species trapped (Table 2b, Fig. 5d–e). Flowers of co-flowering neighbor species L. fortunei had no significant effect on the number of non-pollinator species trapped (Table 2b, Fig. 5f).

Table 2 Results of generalized linear model analyses on the effect of flowers of Drosera and co-flowering neighbor plants on (a) Watarase site (Drosera makinoi and Eriocaulon decemflorum) and (b) Toyoake site (D. toyoakensis and Lysimachia fortunei)
Full size table
Fig. 5
figure 5

Kernel density (black line) and observed value (gray circles) of trapped prey individuals on four treatments with the combination of presence/absence of Drosera and presence/absence of co-flowering neighbor plants flowers. D&C both Drosera and co-flowering neighbor plants, D only Drosera, C only co-flowering neighbor plants, N neither Drosera nor co-flowering plants existed. Flowers of neighbor plants positively affected the number of a all prey species and b pollinator species trapped by Drosera makinoi and d all prey species and e pollinator species trapped by D. toyoakensis. Flowers of neighbor plants positively affected the number of c non-pollinator species trapped by D. makinoi, while did not affect significantly the number of f non-pollinator species trapped by D. toyoakensis

Full size image

Discussion

The two Drosera species trapped their pollinators, four of five pollinator species in D. makinoi and one of six species in D. toyoakensis, including syrphids that are known as effective pollinators of many flowering plants (Willmer 2011; Woodcock et al. 2014). This finding supported our prediction that selfing species may show large PPO. Because both D. makinoi and D. toyoakensis do not have mechanisms such as temporal, spatial, and chemical separation of traps from flowers (Jürgens et al. 2012; El-Sayed et al. 2016), flying pollinators are likely to be trapped by accident. A similar case is reported in a selfing species D. arcturi, which has flowers close to traps and catches major pollinator species (Sciligo 2009).

Whereas the two Drosera species trapped their pollinators, the experimental removal of flowers of neither D. makinoi nor D. toyoakensis significantly changed the number of insects trapped. This may be because autonomous selfing species like D. makinoi and D. toyoakensis allocate small amounts of resources to pollinator attraction (Barrett 2002). On the other hand, the experimental removal of flowers of co-flowering neighbor plants significantly reduced the number of pollinator insects trapped. This finding shows that the number of flowers of not Drosera but neighbor plants maintains the density of pollinator species, although the sample size is small and further tests using more quadrats are desirable. This positive relationship is similar to the relationship called pollination facilitation in pollination biology: the presentation of flowers by one species attracts pollinators to the other neighboring plant species (Feldman et al. 2004; Ghazoul 2006). Pollination facilitation can occur when two or more co-flowering plants form large and collective floral displays (Morales and Traveset 2009; Ye et al. 2014) that increase pollinator density in a patch. Facilitative pollination especially benefits species occurring at low densities (Schemske 1981; Morales and Traveset 2009) and rewardless species growing near rewarding species (Johnson et al. 2003). In our study, the existence of flowers of co-flowering plants benefits Drosera plants by increasing the density of prey insects in a patch. Therefore, Drosera plants may experience pollination facilitation in the presence of co-flowering neighbors. On the other hand, Drosera plants may impose a cost on co-flowering neighbors by decreasing the density of pollinators and pollination efficiency. Therefore, co-flowering neighbors may experience exploitation rather than facilitation.

Drosera makinoi trapped more pollinator species than D. toyoakensis. This difference may be explained by the difference of flower color: D. makinoi has white flowers, which are similar to neighboring E. decemflorum, whereas D. toyoakensis has pink flowers, which are not similar to neighboring L. fortunei. Corresponding to this difference, the frequency of flower visits was not significantly different between D. makinoi and E. decemflorum in all pollinator species, but was significantly different between D. toyoakensis and L. fortunei for some pollinator species. Under these circumstances, pollinator species visiting neighboring plant flowers are expected to be trapped more frequently in D. makinoi, because pollinators of D. makinoi probably share a similar search image to white flowers of Drosera and L. fortunei. Alternatively, it may be explained by the difference of selfing level between the two species: the inbreeding coefficient was higher in D. makinoi (F = 0.497) than in D. toyoakensis (F = 0.260) (Watanabe, unpublished). Pollinator dependence is expected to be lower in the more highly selfing species D. makinoi in which trapping pollinator species is considered to be more advantageous. This result may support the prediction that species with a higher selfing rate are likely to trap pollinators more frequently. For a more rigorous test of this prediction, however, further studies on the degree of PPO in plants with a wide range of selfing rates are needed.

In conclusion, by studying selfing species of Drosera, we showed evidence of trapping major pollinator species in carnivorous plants that were attracted by flowers of neighbor plants. This finding suggests that Drosera plants are exploiting pollinator resources of co-flowering neighbor species. We suggest that selfing carnivorous species provide a unique opportunity to test various ideas on pollinator-prey relationships in carnivorous plants. More species of selfing carnivorous plants are known in Drosera and other carnivorous plants and further studies on those species would deepen our understanding on PPO as well as possible exploitative relationship between carnivorous species and co-flowering neighbor species.