Nursery pollinators, and the plants they use as hosts for offspring development, function as exemplary models of coevolutionary mutualism. The two pre-eminent examples – fig wasps and yucca moths – show little variation in the interaction: the primary pollinator is an obligate mutualist. By contrast, nursery pollination of certain Caryophyllaceae, including Silene spp., by two nocturnal moth genera, Hadena and Perizoma, ranges from antagonistic to potentially mutualistic, offering an opportunity to test hypotheses about the factors that promote or discourage the evolution of mutualism. Here, we review nursery pollination and host–plant interactions in over 30 caryophyllaceous plants, based on published studies and a survey of researchers investigating pollination, seed predation, and moth morphology and behavior. We detected little direct evidence of mutualism in these moth–plant interactions, but found traits and patterns in both that are nonetheless consistent with the evolution of mutualism and merit further attention.
Plant–pollinator relationships have figured prominently in our understanding of mutualism and floral trait evolution (Thomson, 2003; Fenster et al., 2004). Moreover, the pattern of selection generated in the plant–pollinator interaction depends on a diverse web of organisms, which also exhibit spatio-temporal variation and inherent conflicts of interest (Herre et al., 1999; Fenster & Dudash, 2001; Rutter & Rausher, 2004). The tightly evolved mutualisms of Yucca-Tegeticula moths and Ficus-Agaonid wasps are considered paradigms for the importance of such interactions in driving trait evolution and speciation (Thompson & Cunningham, 2002; Datwyler & Weiblen, 2004). In these obligate mutualisms, pollinating species use developing seeds as resources for progeny (i.e. are nursery pollinators). Although advantageous as models with readily quantifiable fitness trade-offs, these partnerships are exceptional, as most nursery pollination systems are less specific, facultatively mutualistic, or antagonistic (Thompson & Pellmyr, 1992; Dufay & Anstett, 2003).
To quantify the processes underlying the evolution of nursery pollination, and contribute to a broader understanding of mutualisms, we need model systems that exhibit variation in their interactions. Well-studied mutualisms are often tightly coevolved, making it difficult to reconstruct the ecological conditions leading to one-to-one mutualisms. However, two nocturnal moth genera (Hadena, Noctuidae; Perizoma, Geometridae) interact with Silene and several allied genera (Caryophyllaceae) in diverse ways, which suggest that these systems are capable of shifting between antagonism and mutualism (Collin et al., 2002; Dufay & Anstett, 2003; Westerbergh, 2004). In this system, the moths are simultaneously effective seed predators and pollinators, with wide variation in the abundances of interacting species, in the costs exacted by larval feeding and in the ecological contexts influencing selection (Pettersson, 1991a; Westerbergh, 2004). The Caryophyllaceae is a particularly rich and well-studied system for examining the evolution of mutualisms, yet important questions remain unanswered. For example, empirical data and modeling of the Silene–Perizoma interaction in Finland provide compelling evidence for potential mutualism (Westerbergh & Westerbergh, 2001; Westerberg, 2004), whereas S. vulgaris–Hadena interactions appear to be largely antagonistic (i.e. destructive to plants with little apparent benefit) (Pettersson, 1991a). In addition, very few families provide suitable candidates for the study of nursery pollination: among 13 families recently reviewed, the interactions in the Caryophyllaceae (i.e. Silene–Hadena) emerge as particularly variable, with relationships to pollinator-seed predators that imply a relatively unspecialized, ineffectively regulated ‘primitive state’ (Dufay & Anstett, 2003). Finally, over 600 species of Silene occur worldwide, in complexes including nursery pollinators, parasitoids and pathogens (Elzinga et al., 2005; Biere & Honders, 2006; Giles et al., 2006). Here, we investigate whether moth seed predators are effective pollinators, evaluating traits in both the moths and their host plants that affect the outcome, whether mutualistic or antagonistic.
This review integrates a scattered literature on plant-pollinator/seed predator interactions in Silene and related Caryophyllaceae with unpublished data from a meta-survey to evaluate the evolution of their close association with nocturnal moths. Using over 50 case studies for geographically diverse species, we ask:
(1) Is there evidence for specialization of Hadena and Perizoma moths as either parasites or mutualists of caryophyllaceous hosts?
(2) Do moths act as selective agents on floral traits associated with both pollination and seed predation?
(3) Are specific behavioral, chemical, or morphological characteristics evident in Hadena or Perizoma that facilitate pollination?
Incorporating empirical data and theory spanning ≈ 30 yr (Brantjes, 1976a,b,c) to date, we seek to stimulate studies of Caryophyllaceae–moth interactions that quantify the sign of the interaction and the ecological conditions that either inhibit or promote the evolution of mutualism.
Conceptual background for nursery pollination
Insects that rear offspring on the seeds of flowers they pollinate constitute ‘nursery’ pollination systems (sensu Dufay & Anstett, 2003) and potentially impose large fitness costs on their plant hosts. Inherent conflict over seeds as larval resources vs future progeny may stimulate trait evolution (Herre et al., 1999), with efficient active pollination stabilizing obligate mutualisms by ensuring food for offspring while reducing overexploitation. Of at least 13 nursery pollinator systems (Dufay & Anstett, 2003), the most strongly mutualistic (Yucca-Tegeticula moths and Ficus-Agaonid wasps) show active pollination (i.e. trait adaptations in a coevolutionary context facilitating pollination) (Weiblen, 2002; Pellmyr, 2003). Here, mutualism is clear as removal of either partner leads to reproductive failure of both. In nonobligate systems, however, evolutionary conflicts are poorly resolved – the mutual benefit of the pollinating seed predator is less obvious because effective pollination by copollinators may shift the direction of the interaction. For example, in S. dioica, a positive benefit by Perizoma moths occurs only if copollinators service ≤ 60% of flowers (Westerbergh, 2004). Otherwise, the ratio of seeds gained through pollination vs lost to predation may result in commensalism, at best.
Both the ecological context of pollination, and the frequency and relative effectiveness of nursery pollinators and copollinators, influence the net positive effect on a host. In senita cacti, where active pollination by the nursery pollinating moth mutualist is clear (Fleming & Holland, 1998), the role of bee copollinators in fruit-set is environmentally dependent, evident only in years of abundant precipitation and cool temperatures (Holland & Fleming, 2002). Similar uncertainty accompanies interactions between Silene and Hadena or Perizoma because the prevalence of important copollinators varies with site and year, potentially swamping any mutualism, and shifting the interaction towards parasitism (Pettersson, 1991b; Westerbergh, 2004).
Silene and allied genera, an emerging model system
Moth pollinators that act as seed predators occur in multiple caryophyllaceous genera (e.g. Dianthus sylvestris, Collin et al., 2002; S. latifolia; Wright & Meagher, 2003). Silene–Hadena interactions are of particular interest as a potentially less specialized, basal form of nursery pollination, lacking regulation of seed damage by moths (Dufay & Anstett, 2003). Diverse ‘pollination syndromes’ (e.g. bird, fly, moth, bee) and breeding systems (e.g. dioecy, gynomonecy, trioecy) also allow exploration of the contexts associated with these potentially mutualistic to antagonistic interactions. Hadena moths are similarly diverse, with over 145 species in palearctic and nearctic regions (Troubridge & Crabo, 2002). Varying flight patterns and oviposition choices by these moths also affect hybrid formation between co-occurring silenes (Goulson & Jerrim, 1997). Nursery pollinators are both prey for parasitoids (Elzinga et al., 2005) and vectors for spores of an anther-smut, Microbotryum violaceum, which reduces floral attractiveness and influences seed set (Shykoff & Bucheli, 1995; Carlsson-Graner et al., 1998).
Natural history of the Silene–moth interaction
Accurate knowledge of the direction of interactions between seed predating moths and Caryophyllaceae hosts requires data on pollinator frequency and effectiveness, and on predation pressure from larval offspring, ideally from multiple species, years and populations. At least 20 insect genera develop on species of Dianthus, Silene, Lychnis and Viscaria (Seppänen, 1970; Robinson et al., 2005), but most are not pollinators. For example, the anthomyid fly, Delia flavifrons, consumes ovules of S. vulgaris, yet has no effect on pollen deposition or seed set (Pettersson, 1992). Only Hadena and Perizoma are widely recognized as nursery pollinators of Silene and as specialist, or rarely, obligate seed predators (Brantjes, 1976a; Westerbergh, 2004). Hadena moths gather nectar, pollinate and oviposit on caryophyllaceous hosts (Fig. 1; Brantjes, 1976a) whereas Perizoma pollinates Silene but obtains nectar elsewhere (e.g. Veronica; Westerbergh, 2004).
Seed-eating behaviors are best known for H. bicuris and H. confusa on S. latifolia and S. vulgaris, respectively. Early instar larvae consume ovules in young capsules at the site of oviposition, and subsequently enter secondary capsules through the top or via holes chewed by these larvae (Brantjes, 1976a). The rewards to the predispersal seed predator, Hadena, include the ovules eaten by larvae, nectar consumed by adults (Brantjes, 1976a; Pettersson, 1991a), and protection within the calyx from parasitoids (Biere & Honders, 2006). The percentage of flowers and capsules damaged ranges from 0 to 100% (Wolfe, 2002); and larval cannibalism occurs in H. bicruris (Brantjes, 1976b) potentially limiting damage to plant progeny.
Antagonistic to potentially mutualistic relationships involving both moth genera appear to be context- and species-dependent. H. compta and H. bicruris are frequent, effective pollinators of D. sylvestris and S. latifolia, respectively (Collin et al., 2002; Bopp & Gottsberger, 2004). In Swedish populations of S. vulgaris, however, H. bicruris is one of 26 visiting species and a relatively ineffective pollinator, suggesting that seed gain from pollination dwarfs its loss from predation (Pettersson, 1991a,b). Recent studies of Silene and Hadena highlight the cost of larval seed predation as a selective force favoring traits that limit exposure to the moth (e.g. flowering phenology) (Wright & Meagher, 2003) and plant mating system characters (Collin et al., 2002).
Web survey analysis and key variables
To document the role and behavior of moths on specific caryophyllaceous plants, we invited researchers studying Silene and related genera to complete a web survey detailing the names of interacting taxa plus qualitative and quantitative data on: (1) pollination and seed predation, (2) floral traits and breeding systems, and (3) habitat, population and community characteristics. We limited the case study to two large lineages in the subfamily Silenoideae with nursery pollination: Dianthus–Saponaria, and the Lychnis–Silene–Viscaria complex, now circumscribed as distinct genera based on molecular data (Oxelman et al., 2001).
Whether moths enhance plant fitness via seed production is dependent on the pollinator community. Thus, we asked investigators to identify major pollinators (i.e. those with the greatest overall positive effect on plant fitness) based on pollinator frequency, pollen load on insects and, when available, effectiveness in pollen transfer to stigmas (as in Inouye et al., 1994; Dafni et al., 2005). We gathered data on pollen or resource limitation, as it might affect the sign of the moth–plant interaction. We further explored whether a functional group of pollinators (sensu Fenster et al., 2004) might exert similar selective pressures on floral design. We quantified associations between nocturnal moth pollination and floral traits predicted for this guild (e.g. related to time of anther dehiscence, flower color and fragrance). We included scent as a response variable in the metasurvey, as it may be an important distinguishing feature of nocturnally pollinated Caryophyllaceae (Jürgens et al., 2002a, 2003). We also noted structures or behaviors that might signify active pollination, and whether plants exhibited traits that facilitate oviposition or minimize seed loss to larval predation. As moths must oviposit in flowers with female function for larvae to gain access to ovules, variation in breeding systems might partly reflect mechanisms to minimize seed loss.
We assessed moths as seed predators, using investigator-provided data for fruit damaged by larvae, and analyzed any potential relationship to ovule number, which could reflect selection to provide a reward to moths while simultaneously minimizing seed predation. We conducted a separate survey of host–plant interactions, using major websites recommended by noctuid moth experts, and cross-checked all sites against published data, our metasurvey and taxonomic synonyms. We also sought to compare data on seed predation and pollinator effectiveness across populations, variable habitats and community structures. However, most survey taxa inhabit similar open environments, and data on community composition were variable in quality, preventing full analysis.
We analyzed 36 taxa (nine Dianthus, 23 Silene and four others), augmenting survey data from published sources to supply missing variables before statistical analysis (SAS Institute, 2004). For multiple entries in widely studied taxa (e.g. S. latifolia), we used modal or mean values for categorical and continuous data, respectively, restricting analysis to native populations for characters influenced by residency (e.g. flowering time, pollinators). We used Fisher's exact test to analyze associations between categorical variables (some cell-expected values were ≤ 5) and t-tests to determine whether categorical variables were significant sources of variation in continuous variables. Correlation analysis allowed the examination of linear relationships between continuous variables. To detect multivariate patterns in scent chemistry, we ran nonmetric multidimensional scaling (NMDS) ordination (pc-ord 4.0) using Sorensen distance and slow, thorough, autopilot settings.
Caryophyllaceae–moth interactions as potential specialized mutualisms
The first goal was to evaluate evidence for specialization in the interaction of Silene and related genera with Hadena or Perizoma. If these are emerging as specialized mutualisms, we hypothesized that investigators would classify the moths as primary pollinators.
Two insect groups comprise over 75% of the animal taxa designated as major pollinators (Table 1): diurnal and nocturnal moths (48.1%; e.g. Noctuidae and Sphingidae); and bees (29.6%; Apidae, especially Bombus). By far, lepidopterans pollinated the majority (55.5%) of the 27 taxa with a known major pollinator or suite of pollinators, with diurnal moths and butterflies poorly represented (14.8%), relative to nocturnal moths (40.7%). Of plants with lepidopterans as primary visitors, 73.3% are nocturnally pollinated, either by noctuids (usually Hadena) or sphingids (e.g. hawkmoth Deilephila porcellus). Moreover, Hadena is a major or common pollinator for most moth-pollinated plants (67%, Table 1).
Table 1. Primary mode of pollination, breeding system, and floral traits in Caryophyllaceae where Hadena is present (Y) and known to be a a major pollinator (MP) or a common pollinator (CP)a
Time of pollination
d, dioecy; g, gynodioecy or gynomonecy; h, hermaphroditic for diurnal (di) or nocturnal (noc) pollination in flowers with funnelform (f), narrow (n), or ovoid (o) floral tubes.
In some populations of S. dioica, the moth Perizoma affinitata (Geometridae) is probably the main effective pollinator (see Westerbergh 2004 and the text for details).
Hadena clearly emerges as an important pollinator, but clear evidence of specialization is critical to establishing whether Hadena–Caryophyllaceae interactions represent a pathway to obligate mutualism, as exemplified in Yucca and Ficus. Originating > 80 Ma (Datwyler & Weiblen, 2004), the highly evolved fig–wasp interactions provide stringent baseline criteria for an exclusive mutualism: (1) a moth species lays eggs on a single caryophyllaceous taxon; (2) a caryophyllaceous taxon has one moth taxon responsible for ovule and seed predation; and (3) this single moth species is a plant's sole or major pollinator. Specialization in nursery pollination can be inferred from various criteria, including a close, potentially 1 : 1, association of host and pollinator-seed predator and concordant geographical ranges or activity periods of the two partners.
We uncovered little evidence for the tight association manifested in figs or yuccas (Fig. 2a,b), but several types of data demonstrate the evolutionary potential for such interactions. First, our survey detected 14 Hadena species whose larvae use the flowers and seed capsules of 26 caryophyllaceous plants as larval hosts (Appendix 1). Second, Hadena caterpillars feed almost exclusively on a small subset of genera in the Caryophyllaceae (i.e. Dianthus, Silene–Lychnis–Viscaria, and rarely others). Only H. caesia appears to also rear progeny on plants outside the carnation family (i.e. Fragaria, Primula; Seppänen, 1970). Third, host-plant records of the Natural History Museum of London (Robinson et al., 2005), and our survey data, show that 16 of 26 plant taxa probably host only a single species of Hadena (Fig. 2a; mean 2.0 ± 0.38 Hadena species per plant). In a more limited data set of five species of Silene–Lychnis with Perizoma associations, three host a single Perizoma species.
A plant species might host diverse moth taxa, but specificity must also be viewed from the perspective of the moth's ability to use one or many hosts. Perizoma may have a broad range of hosts, but our survey revealed only four Perizoma species with Caryophyllaceae hosts, whereas many noncaryophyllaceous plants are known to host Perizoma moths, implying low overall specialization in Perizoma–Caryophyllaceae lineages. Similarly, London's Natural History Museum lists 16 Perizoma species on 29 angiosperm host genera, and 15 species of Hadena on 14 genera: for Hadena, 95% of the records with names are caryophyllaceous plants, compared with 23% for Perizoma (Robinson et al., 2005. In Northern Europe, P. affinitata has an obligate dependence on S. dioica for rearing its young, and both moth and plant have the same distribution, flowering and flight periods in Finland (Westerbergh, 2004). However, three other Hadena moths and P. flavofasciata also occur on this Silene (Appendix 1), and bumblebees predominate in most S. dioica populations (Table 1), potentially swamping any mutualistic effect, except in isolated, serpentine populations where plants appear to be dependent on their nursery pollinator (Westerbergh, 2004).
Of 14 Hadena species using carnation family hosts, seven occur on one or two plant hosts (Fig. 2b). Moreover, in three instances involving two hosts, the plants used by the moth are close relatives (e.g. H. ectypa on S. virginica and S. stellata); in two cases, they have been treated historically within a single species (i.e. H. filograna on S. vulgaris and S. uniflora var. petraea; H. variolata on varieties douglasii and rupinae of S. douglasii). Three Hadena–Silene interactions show a 1 : 1 correspondence between the moth seed predator and its plant host (i.e. H. circumvadis–S. spaldingii, H. irregularis–S. otites and H. sancta–S. stockenii; Appendix 1). Most have diurnal pollination, so the moth seed-predators are probably not mutualistic, and they are not major pollinators, with the caveat of limited night observations (Lesica & Heidel, 1996; Talavera et al., 1996).
Ample evidence of generalization also exists among Caryophyllaceae–Hadena interactions. V. vulgaris, S. nutans, and S. vulgaris each host five to nine different species of Hadena, as well as the larvae of one or two Perizoma species (Appendix 1). All five Silene species known as hosts for Perizoma also support at least one Hadena species, highlighting again the absence of tight 1 : 1 associations between moth and caryophyllaceous species. These groups might also reflect a diffuse coevolution (Strauss et al., 2005), with the unrelated moth species acting as a single functional group of nursery pollinators (sensu Fenster et al., 2004). Not only do the relative roles of moths as pollinators–seed predators merit more study, but quantifying the ancestral and derived character states would allow us to determine the direction of evolution of the moth–Caryophyllaceae interactions (Pellmyr, 2003).
Despite the 1 : 1 correspondence among several interactions, in only four survey species did one Hadena species act as a seed predator and major pollinator, indicating the potential for mutualism to evolve. In the three nocturnally pollinated species (H. compta–D. sylvestris, H. ectypa–S. stellata and H. variolata–S. douglasii), moths consumed or damaged ≈ 18–51% of capsules. Because we lack estimates of the total seed production attributable to Hadena, the interactions could be mutualistic or antagonistic. Clearly, future studies need to quantify the Hadena contribution to population-level seed set in these species relative to seed loss through larval predation. In S. vulgaris, Pettersson (1991a,b) quantified similar variables for a guild of Hadena species delivering ≈ 7% of the pollen on stigmas, but consuming ≈ 5–68% of capsules, indicating that this interaction is probably antagonistic. Presently, H. ectypa may account for up to ≈ 45% of nocturnal visitors in populations of S. stellata and ≈ 35% seed loss (R. J. Reynolds et al., unpublished). Although other moth copollinators exist, S. stellata could have a mutualistic relationship with H. ectypa if it contributes more to seed production than loss. These preliminary data emphasize the need to quantify the abundances and effects of copollinator and nursery moths across populations and years, with respect to pollination and larval herbivory.
Trait evolution in plants and moths
Floral evolution in response to moths as pollinators
Another survey goal was to identify traits that might reflect the evolution of a specific moth–plant interaction, whether mutualistic or not. In an evolving nursery mutualism, we expect floral traits associated with pollination (e.g. in attraction or efficient pollen transfer) to reflect a response to Hadena or Perizoma.
Our analysis suggests that indeed some floral traits, commonly associated with attracting nocturnal lepidopterans, may have evolved in response to moths as selective agents. Fisher's exact test demonstrates a significant association between nocturnal pollination and white flower color (P < 0.004), the presence of scent (P = 0.004), and crepuscular or nocturnal anther dehiscence (P < 0.001). Only three nocturnally visited plants had nonwhite corollas, and all but D. sylvestris are pollinated by hawkmoths (Table 1). We detected a weak association between nocturnal pollination and plants with fringed petals (P = 0.11).
Scent is a key attractant for nocturnally visited Caryophyllaceae: its emission coincides with crepuscular opening of S. latifolia flowers (Dötterl et al., 2005) and it both initiates seeking behavior in Hadena and guides its landing at close range (Brantjes, 1976a,c). For 13 plant species, our survey allowed us to associate the role of Hadena or other visitors as main pollinators with specific scent profiles (Jürgens et al., 2002a, 2003; Jürgens, 2004). For these plants, NMDS generated a three-dimensional solution with clear separation of nocturnal and diurnal patterns (Fig. 3), for which axes 1 and 2 explain 20% and 69% of the variation. Diurnally pollinated species that serve as host plants for Perizoma and/or Hadena emit primarily fatty acid derivatives (mean 42.3%) and secondarily benzenoids (30%). In contrast, benzenoids (mean 51.9%) and isoprenoids (32.7%) are the main floral volatiles for nocturnally pollinated species. Only the four species with Hadena as a common/major pollinator emitted lilac aldehydes or methyl benzoate as dominant compounds (this survey; Jürgens et al., 2002a, 2003). In S. latifolia–H. bicruris, dose-dependent tests of electrophysiologically active lilac aldehydes also demonstrate their role in attracting Hadena (Dötterl et al., 2006). However, these compounds also attract other nocturnal visitors (Jürgens et al., 2002a; 2003), obscuring the relationship between these special compounds and Hadena.
The combined survey data on floral traits, including scent, provide strong evidence of a nocturnal pollination syndrome for many Silene species. However, we know little about the exact role of Hadena and Perizoma moths as selective agents for the evolution of these floral traits. We detected no clearly documented case where a moth species is the sole pollinator of Silene or related genera, and these nursery systems have noctuid, sphingid and bumblebee copollinators that may also influence floral characters (Table 1). We need more complete data on Hadena and Perizoma species as pollinators and selective agents, including visitation rates, pollination effectiveness, effect on seed-set and, ultimately, the selection intensities and direction exerted by these moths.
Floral evolution in response to moths as seed predators
If present, a mutualism of Silene and sister groups with seed-eating moths should reflect the evolution of traits in response to both pollination and seed loss. So, is there evidence that seed predation also influences floral traits? First, our survey detected significantly higher rates of fruit predation with nocturnal pollination (30.6 ± 6.1; P = 0.01, t = 2.79) and nonhermaphroditic breeding systems (28.1 ± 7.0; P = 0.04, t = 2.21) compared with diurnal pollination (10.3 ± 3.5) and hermaphrodites (11.2 ± 3.5). These results implicate moths as potential selective agents for traits minimizing predation, in conjunction with effective pollination. Because diurnally pollinated plants experience significantly less predation than nocturnal species, diurnal pollination could represent a mechanism that avoids seed predation or, alternatively, an ‘escape route’ from nocturnal pollination. A complication is that diurnal pollination can be linked to environments outside the moth's range (e.g. S. caroliniana is an early spring ephemeral; S. acaulis occurs at high latitudes and elevations). Second, if selection favors the evolution of a mutualism, we expect floral traits that minimize seed predation after pollination and egg-laying by moths. In dioecious S. latifolia, the rapid, 24 h decline of scent emission in pollinated vs unpollinated flowers is consistent with a response by plants to reduce seed predation over the life span of female flowers (Dötterl et al., 2005) and to limit costs in nonhermaphrodite systems. The high predation in nonhermaphrodites in our survey also suggests that moths can specialize on female flowers. Thus, another promising avenue for study is that avoidance of seed loss may drive the evolution of breeding systems towards hermaphroditism.
We predicted initially that pollen limitation or pollinator scarcity might create opportunities for the evolution of mutualism if plants attain more pollination and fruit-set in the presence of seed predators than is possible in their absence. In nocturnally pollinated systems, the survey shows no inherent predation cost when these moths are major pollinators [i.e. investigator-estimated damage to fruits is similar regardless of whether Hadena is a major pollinator (35%) or not (30%)]. Thus, selection to increase the role of these moths as pollinators vs seed predators is possible. Pure seed predation and a minor role for moths as pollinators may be antecedents to evolution of a larger pollination role by these moths, but testing this hypothesis requires a more resolved phylogeny. Of species surveyed to date, nearly all are pollen-limited, implying that moth pollination might be important to overall plant reproductive success (Dudash & Fenster, 1997; Brown & Kephart, 1999; Alatalo & Molau, 2001).
If oviposition delivers pollen to stigmas, floral morphology may have evolved to facilitate and regulate egg-laying behavior by Hadena within accessible flowers. The survey provides some support for this: among lepidopteran-pollinated plants, those with Hadena as a main or common pollinator are almost exclusively ovoid or funnel form (P = 0.02, Fisher's Exact Test), including 75% of nocturnally visited plants (Table 1). However, we do not as yet know whether oviposition is associated with high pollen transfer, or if broad calyces (Fig. 1c) enhance egg-laying or larval survival (e.g. larvae may gain shelter from parasitoids, sensu Biere & Honders, 2006, if they can develop within a few large flowers). Morphological differences in stigma height and curvature may be regulatory in S. dioica: short-styled stigmas, presented at same level as the corolla tube opening, create difficult access for Perizoma females, thwarting oviposition despite successful pollination (Westerbergh, 2004). This trait variability in Hadena- or Perizoma-visited flowers might ensure that some flowers escape oviposition, thus reducing reproductive failure and stabilizing the mutualism.
This survey also revealed a weakly positive association between ovule number per flower and larval herbivory (P = 0.06, F = 4.6; correlation = 0.56, r2 = 0.32), which is consistent with trait changes expected in response to high seed predation. High ovule number might either compensate for seed predation or act as a reward to seed predators. Data from additional species will be critical for understanding this pattern, as the relationship garnered from our survey is dominated by a high value in S. latifolia (> 500 ovules per flower). The ovule number is also large in S. dioica (> 250 ovules per flower), the sole host for the obligate nursery pollinator P. affinitata (Carlsson-Graner et al., 1998; Jürgens et al., 2002b; Westerbergh, 2004). In contrast, of closely related S. caroliniana, S. stellata and S. virginica, S. stellata has the highest seed loss to Hadena, yet fewer ovules dispersed into greater per-plant flower production, a potential strategy to reduce overall seed predation by moths in this pollen-limited plant (R. J. Reynolds et al., unpublished).
At an intraspecific level, some tantalizing evidence exists for plant response to limit seed predation by moths. Capsule wall thickness is significantly greater in native European populations of S. latifolia receiving higher predation by Hadena than in their introduced North American counterparts (Wolfe, 2002; Blair & Wolfe, 2004), but confirmation that capsule thickness actually impairs damage by Hadena larvae is needed. For example, during primary attack, the soft wall of a young ovary seems unlikely to impede initial larval penetration. In a secondary attack, mobile larvae typically gain access to ovules or seeds by chewing into the top of developing fruits (Elzinga et al., 2005).
Evolution of moth behavior and morphology
If selection for mutualism favors the provision of benefits by the moth, adaptations by nursery pollinators should include specific behavioral, biochemical, or morphological characteristics that not only enhance their performance as seed predators, but facilitate effective pollination. Yucca moths use maxillary tentacles to collect and compact pollen, storing it in a cavity under the head (Pellmyr, 2003). In Hadena, no morphological structures seem to be specifically adapted for pollination. The calyx tube in D. sylvestris, and the proboscis in H. compta, are both ≈ 23 mm, and Dianthus pollen is most abundant on the proboscis and labial palps of H. compta and H. caesia (Erhardt, 1988, 1990), but few specimens have been studied.
The survey reveals much diversity and some selectivity in Hadena behavior, which has been studied in detail in European S. latifolia (Brantjes, 1976a,b,c). H. bicruris selectively oviposits in its flowers over co-occurring Dianthus, S. dioica, S. nutans, S. vulgaris and Saponaria (Erhardt, 1988; Goulson & Jerrim, 1997; Bopp & Gottsberger, 2004). As in cases known among the caryophyllaceous plants we surveyed, typically Hadena imbibes nectar before successful oviposition (71%), and flowers receive only one egg (all survey taxa; Brantjes, 1976b). During nectar feeding on S. latifolia, the initial floral contact with the proboscis precedes contact with the head as moths repeatedly and more vigorously pump flowers for nectar (Brantjes, 1976a). During oviposition, the legs, abdomen and ovipositor also contact floral parts (Brantjes, 1976a). Similarly, in close congeners S. uniflora var. petraea and S. vulgaris, Hadena brushes stamens and pistils, then bends its body while ovipositing, vigorously inserting its abdomen (M. W. Pettersson, unpublished; H. Prentice, unpublished). Hadena also contacts anthers and stigmas on North American S. douglasii and S. stellata, gathering nectar first, then ovipositing in a subset of these flowers (S. Kephart, unpublished; R. J. Reynolds et al., unpublished). Nectar collection typically precedes oviposition on a different flower (71% of survey cases), but can also occur simultaneously with it in Silene and Dianthus; either behavior can result in pollination (this survey; Brantjes, 1976b), but definitive tests of the relative effectiveness of these behaviors in securing pollination and fruit set are sorely needed.
The behavior of geometrid P. affinitata on S. dioica reveals both commonalities and distinct differences from Hadena (Westerbergh, 2004). Hadena and Perizoma typically lay one egg per flower, have cannibalistic larvae that leave seeds in some capsules, and show stereotypical oviposition behaviors (Brantjes, 1976a,b; this survey). Perizoma alone exhibits a dense brush of hair on the ovipositor that retains pollen, exclusive development of larvae within one fruit and the absence of nectar feeding because the proboscis is too short to reach S. dioica nectar (Westerbergh, 2004). Perizoma females differ in behavior and visit length during probes of male and female flowers of dioecious S. dioica (one per plant in 94% of visits); both the abdomen and ovipositor enter the floral tube making contact with anthers and/or stigmas (Westerbergh, 2004). In both Perizoma and Hadena interactions with Caryophyllaceae, however, we need fuller exploration of moth behavior in relation to pollination and seed loss, within and among populations.
Conclusions and future directions
Although our survey of over 30 plant taxa spans only a fraction of the diversity in Silene and allied genera, the review identifies promising avenues of future research. Moth seed predators can be major pollinators, and evidence exists that floral traits have probably evolved in response to the selection pressures they exert. Nocturnal moth pollination is associated with floral traits classically assigned to moth pollination, including white color, fragrance and nocturnal anthesis (Faegri & van der Pijl, 1979); all are consistent with a response to selection exerted by Hadena and Perizoma, especially when they serve as major pollinators. In some nocturnally pollinated Silene, the petals also close (i.e. roll towards the center) during the day, either blocking (S. latifolia) or reducing (S. douglasii, S. nutans) access to nectar; for two of these species, Hadena moths are important pollinators (Table 1). Flowers of D. deltoides, a diurnal species found in open habitats, act in reverse, closing at night (Jennersten, 1988b), suggesting that the phenomenon is not simply a mechanism of water conservation.
Mutualisms may evolve from an ancestral state where only one of the species in the interaction benefits (Dale et al., 2001) but currently, of the moth–plant associations surveyed, no strict one-to-one interaction involving both pollination and seed predation exists throughout the geographical range of a given species. To assess the generality of this finding, however, we need more field observations documenting the taxonomic identity of pollinators and seed predators of caryophyllaceous plants. We have observed cases of parasitism by moths without benefit to plants, and cases suggesting mutualism. To discriminate between antagonistic relationships and the presence of nursery mutualisms, we encourage field studies comparing the relative cost-benefits of these pollination systems among related plant species. Studies should incorporate detailed measurements of both reductions in seed production as a result of predation and gains in plant fitness arising from moth pollination, particularly in pollen-limited populations. Spatio-temporal components would help us to evaluate whether the evolving interactions form a ‘coevolutionary mosaic’, a perspective that has been fruitful thus far for nursery pollination systems (Thompson & Cunningham, 2002).
In the Silene–Hadena systems, the stereotyped behaviors and retention of the hairs on the ovipositor of P. affinitata are potential precursors to effective active pollination. Study of additional species will reveal the commonality of such characteristics within Perizoma, as the presence of active pollination could shift the interaction towards mutualism. To determine if evolved mutualisms characterize these interactions, however, more detailed morphological measurements are needed, along with the pollen distribution on moths, amounts of pollen transferred to stigmas during nectar feeding and oviposition, and the extent to which these behaviors augment seed-set. Presently, for species that have been studied intensively, the evidence that Hadena moths are more effective than other nocturnal moths in pollen transfer is equivocal (Pettersson, 1991b; Collin et al., 2002; R. J. Reynolds et al., unpublished). Identifying the pollinator's sex is important, because nonovipositing male pollinators may only contribute to positive components of the interaction while gathering nectar, unlike female moths that feed and oviposit, becoming potential parasites. Focal species in which seed-predating moths are the major pollinators deserve special attention.
While relatively few questions are definitively answered by our survey, this review demonstrates how richly variable this system is for future study of species interactions. For lesser-known nursery pollination systems, we must define the ecological conditions that might ultimately foster the evolution of an exclusive mutualism without copollinators. Just as importantly, embedding the results of pollination and predation studies within a well-supported phylogenies of both moth and plant species will permit major advances in our understanding of the direction and frequency of evolutionary change for traits underlying nursery pollination, and of the factors shaping the form and timing of transitional stages in the evolution of mutualisms.
We thank investigators from nine countries and 33 institutions who provided invaluable data, photographs, or sources on plants or insects as part of our survey: J. Alatalo, H. Alexander, J. Antonovics, E. Barthelmess, A. Biere, M. Buide, C. Collin, L. Delph, R. Dolan, S. Dötterl, A. Erhardt, M. Glaettli, B. Giles, P. Goldstein, D. Goulson, P. Hammond, D. Inouye, O. Jennersten, A. Jürgens, P. Lesica, D. Marr, D. McCauley, T. McCabe, T. Meagher, E. Menges, B. Oxelman, M. Pettersson, M. Pogue, H. Prentice, J. Rawlins, C. Richards, J. Shykoff, L. Wolfe, J. Wright. Also, J. Kephart designed the web interface; B. Lindh contributed multivariate analyses; and J. Apple, J. Butler, B. Giles and A. Jürgens added insightful comments on the paper. The work is supported by: NSF-R0A 0108285 (S. Kephart, C. Fenster, M. Dudash), NSF 0427922 (C. Fenster, M. Dudash), NSF 0315972 (C. Fenster, M. Rutter), and Earthwatch Institute (S. Kephart). R. Murtha (Willamette University) and H. Wilbur and E. Nagy (Mountain Lake Biological Station) gave valuable logistical and technical assistance.
Table 2. Plant host–moth relationships in 26 species of Silene and related Caryophyllaceae based on our survey and two comprehensive websites (1 and 2)
Moth seed predator(s)
Numbered moth taxa in Hadena and Perizoma (n= 17 species in combined genera) are shown in parentheses. Bold type indicates moth species known to visit only a single species of Silene in its native geographical area, to our best knowledge.
Sometimes treated as Sideridis rivularis.
Sometimes treated as Conisania luteago; also uses Spergularia rupicola as a host, although not included in this survey. Similarly, Stellaria media is listed by Seppänen (1970) as a host for Perizoma taeniata.