Host growth form underlies enemy-free space for lichen-feeding moth larvae


Correspondence author. E-mail:


1. Natural enemies may direct the host use of herbivorous insects on those hosts that ensure highest survival, thus creating enemy-free space. Host structure may contribute to enemy-free space if the current host ensures better refuge from natural enemies than other potential hosts. So far, however, direct evidence of the role of host structure for enemy-free space is lacking.

2. This study looks at the effect of physical host structure on the previously demonstrated enemy-free space of a lichen-feeding moth, Cleorodes lichenaria by manipulating the structure of host lichens and the access of natural enemies to larvae in the field. It was predicted that if larvae receive enemy-free space on Ramalina lichens because of their shrubby appearance, larvae should survive better on shrubby than on flat lichens in the presence of natural enemies but not in the absence of natural enemies.

3. Larvae survived better on shrubby than flat lichens and when the access of natural enemies to larvae was prevented than in the presence of them. According to the prediction, larvae in the presence of natural enemies survived better on shrubby compared with flat thalli but not in the absence of natural enemies. Thus, shrubby host structure promotes survival of larvae and underlies the enemy-free space on Ramalina species in natural conditions.

4. Host structure as a mechanism for enemy-free space and the direct impact of host structure for the performance of C. lichenaria larvae are discussed. Other potential reasons, such as lichen secondary chemicals and host-induced colouration of larvae as a basis of enemy-free space, are also discussed.


According to the enemy-free space hypothesis, high selection pressure by natural enemies may lead to host use that ensures the best protection against natural enemies (Price et al. 1980; Jeffries & Lawton 1984; Berdegue et al. 1996). High selection pressure by natural enemies may restrict the diet breadth of herbivorous insects or lead to host shift to another host providing better protection against natural enemies, thus providing support for the enemy-free space hypothesis (Denno, Larsson & Olmstead 1990; Feder 1995; Stamp 2001; Oppenheim & Gould 2002; Mulatu, Applebaum & Coll 2004; Murphy 2004; Wiklund & Friberg 2008; Diamond & Kingsolver 2010). There are, however, surprisingly few attempts to unveil the mechanisms underlying enemy-free space. Enemy-free space is in general thought to be based either on sequestration of secondary chemicals for herbivores’ own defence or on specialization on a host that is less apparent to natural enemies (Denno, Larsson & Olmstead 1990; Feder 1995; Stamp 2001; Oppenheim & Gould 2002; Mulatu, Applebaum & Coll 2004; Murphy 2004; Wiklund & Friberg 2008; Diamond & Kingsolver 2010). The definition of enemy-free space (Jeffries & Lawton 1984) takes into account several other aspects, in principal all, niche characteristics, such as the structural refuge for herbivores on their hosts. So far, there is some comparative evidence of the impact of a host’s physical structure as an underlying cause for enemy-free space (Feder 1995; Oppenheim & Gould 2002). On those model systems where enemy-free space has been unequivocally demonstrated, empirical evidence of host structure as an underlying cause for enemy-free space is, however, currently lacking.

The aim of this study was to explore whether host physical structure contributes to the previously documented enemy-free space of a geometrid moth, Cleorodes lichenaria (Hufnagel) (Pöykkö 2011). Although this species is regarded a polyphagous lichen feeder (Mikkola, Jalas & Peltonen 1989; Robinson et al. 2000), in Åland Islands shrubby Ramalina lichens are preferred by ovipositing females and developing larvae (Pöykkö 2006). When natural enemies have access to larvae, survival is highest on Ramalina fraxinea (L.) Ach. and Ramalina farinacea (L.) Ach. on which larvae are less exposed to natural enemies than on other, flat lichens (Pöykkö 2011). In the absence of natural enemies, other potential host lichens ensure similar or even better performance of larvae indicating the presence of enemy-free space (Pöykkö 2011). Both Ramalina hosts are bush-like fruticose lichens attached to their substratum with a single holdfast. Ramalina fraxinea has broad and pendulous lobes up to few centimetres wide. Ramalina farinacea is a tufted lichen with abundant and numerous sparse lobes. Other potential hosts, like Xanthoria parietina (L.) Th. Fr. and Parmelia sulcata Taylor are foliose lichens attached tightly to their substratum with lower cortex and rhizomes. Accordingly, enemy-free space on Ramalina species may be based on the host growth form if shrubby lichens provide structural refuge for larvae. If enemy-free space of C. lichenaria is based on host growth form, larvae are expected to perform better on shrubby thalli compared with flat ones in the presence of natural enemies, whereas in the absence of natural enemies there should not be differences in the performance of larvae on different host forms.

Materials and methods

A field experiment with a factorial setup, two levels of host growth forms (flat/shrubby) and natural enemies (presence/absence, referred as non-excluded and excluded treatments hereafter, respectively), was performed in two locations on the Åland Islands. Ramalina fraxinea thalli were collected from the study area in the middle of June 2009 and moistened with deionized water before placing them on tree trunks. One large thallus was split in four, and pieces were randomly distributed to one of the four treatments: flat excluded; flat non-excluded, shrubby excluded and shrubby non-excluded. These four experimental thalli were created from a single lichen thallus, so there should not be differences in the performance of larvae because of host quality between treatments. Two or three thalli were needed to create an individual experimental thallus of c. 10 cm in width. Experimental thalli, as well as exclosures were placed on trunks in the middle of June before the flight period of C. lichenaria to ensure that no naturally oviposited eggs occurred on the thalli. Flat thalli were formed by attaching lobes of R. fraxinea as tightly as possible on the tree trunks with a stapler (Fig. 1a). Shrubby thalli were achieved by attaching lobes of R. fraxinea to tree trunks only from the upper corner letting the rest of thallus to hang naturally (Fig. 1b). The four thalli (two shrubby and two flat) were attached near each other on the same tree within 1 m. All other macrolichens were scraped off from the experimental areas. Altogether 84 (21 of each treatment) thalli were formed on the trunks of 10 maples (Acer platanoides) and nine ashes (Fraxinus excelsior) at the height between 2 and 6 m to represent two most common tree species in the study area. At both locations, several tree individuals of both species were used as substrates for experimental thalli.

Figure 1.

 Host lichen forms and treatments used in the experiment: (a) flat thallus, lobes attached tightly to the tree with staples, (b) shrubby thallus, attached to the tree from the upper part allowing lobes to hang naturally, (c) excluded (upper) and non-excluded (lower) treatments. Larvae of Cleorodes lichenaria are shown with arrows.

Polytetrafluoroethylene (PTFE)-coated nylon glass cloth (hereafter teflon ring) was used to encircle thalli and to prevent larvae from escaping, and a cone of nylon haircloth excluded natural enemies from the larvae (Fig. 1c). This method has proven successful in preventing larval escape from experimental fences during the earlier study (Pöykkö 2011). Four-centimetre-wide teflon ring was sewed on the inner side of a haircloth cone (mesh size 0·5 mm) 2 cm above the bottom. A haircloth cone with a teflon ring was tightly attached on the tree trunk around each thallus with a stapler (Fig. 1c). This modified method from Pöykkö (2011) was applied, because direct attachment of a teflon ring on tree trunks may increase the humidity inside the ring and lead to moulding near the ring edge (personal observations). Thus, ‘detaching’ teflon rings from the trunk makes ventilation more effective and likely decreases excessive humidity inside the fences. The applied method may have also affected other factors such as shading or interspecific interactions to differ between excluded and non-excluded treatments. There is, however, no evidence that the applied exclusion treatment itself (even with smaller mesh size, 0·25 mm) would have any effect on the performance of larvae (Pöykkö 2011). Moreover, most likely competitors to larvae include different snail species, which were only occasionally observed inside the fences (personal observations).

To acquire larvae for the experiment, final instar larvae of C. lichenaria were collected from the study site in May 2009 and reared into adults under natural conditions. Adults were paired in small plastic containers (0·5 L) with small piece of R. fraxinea for females to oviposit on. After hatching, larvae were fed 4–5 days in laboratory on R. fraxinea before transferring them to experimental fences 10–13 July 2009. Larvae on the excluded treatments were placed on the experimental thalli through a cut in a nylon cloth above the teflon ring. For the non-excluded treatments, the upper part of a cone was cut above the teflon ring and bent aside (Fig. 1c). Two larvae were placed on each thallus simultaneously to ensure sufficient number of larvae in the experiment and to avoid intraspecific competition. The presence of larvae was checked 20–22 October 2009 and 3–6 May 2010 when the experiment was terminated. This procedure allowed to cover whole larval period and to specify whether the impact of natural enemies on the survival of larvae depends on larval developmental stage. Remaining larvae were reared into adulthood in natural conditions to identify potential emerging parasitoids. Missing larvae, along with the one parasitized larva, were considered dead, which is reasonable to assume, given that experimental fences efficiently prevent the escape of larvae (Pöykkö 2011).

Data were analysed with R 2.9.1 (R Development Core Team). Generalized linear mixed model (GLMM) with the number of dead and alive larvae within a fence as a response variable and lichen growth form and the exclusion treatment as fixed factors (including also interaction) was used to explore the survival of larvae before and after overwintering. GLMM was fitted with the function ‘glmer’ with binomial error distribution and logit link function using Laplace approximation (Bates, Maechler & Dai 2008). Fitting a random factor to the models was started with tree nested within locality and proceeded with only tree as a random factor. Finally, a GLM model without any random effects (with ‘glm’ function) was also fitted to the data. Estimation of model goodness-of-fit of GLMMs with different random structures (including also a GLM model without a random factor) were performed by visual evaluation of residual plots, with Akaike’s information criterion and by comparing residual deviances to chi-square distribution with residual degrees of freedom.

As the impact of host structure on the survival of larvae was parallel between non-excluded and excluded treatments (see ‘Results’ section), it is necessary to perform statistical tests among non-excluded and excluded treatments to reveal whether the impact of natural enemies on the survival of larvae differs between these two treatments. Thus, to test the specific prediction (i.e. that survival of larvae on shrubby thalli is higher than on flat thalli only among exposed treatments), similar GLMMs with growth form as an explanatory variable and tree individual as a random factor (model goodness-of-fit was best with this random structure, see ‘Results’ section) were used to test the survival of larvae both among non-excluded and excluded treatments.


Before overwintering, larval survival was higher on shrubby lichens than on flat ones in both excluded and non-excluded treatments. In addition, larvae in the excluded treatment had a higher survival rate than the ones in non-excluded treatment. (Fig. 2a, Table 1). The trend in the survival of larvae across treatments after overwintering was similar to that before overwintering (Fig. 2b), although statistically insignificant (Table 1). Comparisons of models with different random structures revealed that tree as a random factor resulted in best model goodness-of-fit (Table 2). Moreover, inclusion and structure of random effects had relatively little impact on goodness-of-fit of models, indicating that the impact of localities and individual tree on the survival of larvae was negligible (Table 2). The tests carried out within non-excluded and excluded treatments revealed that survival of larvae was higher on shrubby than flat thalli in the presence of natural enemies but not in the absence of them (Fig. 2, Table 1). Only one larva on a flat non-excluded thallus was parasitized by a braconid wasp (Microgastrinae spp.).

Figure 2.

 Survival (mean + 1 SE) of Cleorodes lichenaria larvae on flat and shrubby lichens in the absence (excluded treatment) and presence (non-excluded treatment) of natural enemies during (a) the autumn and (b) the whole experimental period.

Table 1.   Results of generalized linear mixed models for survival of Cleorodes lichenaria larvae
Fixed effectsEstimate ± SEzP-value
  1. GF, growth form (flat/shrubby); EXT, exclusion treatment (non-excluded/excluded).

Before overwintering
 Intercept2·39 ± 0·584·11 <0·0001
 GF−1·43 ± 0·65−2·190·0282
 EXT−1·30 ± 0·66−1·980·0480
 GF × EXT0·61 ± 0·810·750·4554
  Among non-excluded treatment
   Intercept−2·67 ± 0·66−4·03<0·0001
   GF1·58 ± 0·69−2·300·0216
  Among excluded treatment
   Intercept−1·07 ± 0·37−2·860·0042
   GF0·78 ± 0·47−1·650·1000
After overwintering
 Intercept4·12 ± 1·143·590·0003
 GF−1·22 ± 1·28−0·950·3421
 EXT−1·57 ± 1·25−1·260·2085
 GF × EXT0·49 ± 1·470·330·7403
Table 2.   Comparisons of GLM model (without random factor) and GLMM models with different structures of random factors. Fixed factors were same in all comparisons including growth form, exclusion treatment and interaction term
Random factorsAICaResidual d.f.Residual devianceP-valueb
  1. GLMM, generalized linear mixed model.

  2. aAIC-values are for GLMM models including random factor (with glmer function) and for a GLM model without random factor (with glm function).

  3. bModel goodness-of-fit.

  4. cVariances in the survival of larvae because of localities and tree individuals were 0·0 and 0·51, respectively, indicating that locality had no impact on the survival of larvae, whereas survival of larvae slightly differed between tree individuals.

Before overwintering
After overwintering


Natural enemy–mediated and direct impact of host structure on larval survival

Both lichen growth form and exclusion treatment affected larval survival. According to the prediction, C. lichenaria larvae survived better on shrubby lichens compared with flat ones only in non-exclusion treatment, i.e. in the presence of natural enemies. Thus, host lichen growth form significantly contributes to the enemy-free space of C. lichenaria on fruticose Ramalina species in field conditions (Pöykkö 2011). The specificity of natural enemies on the mortality of C. lichenaria larvae was not explored during this study, except one parasitized larva, which is also the only parasitoid species reared from C. lichenaria larvae in the study population (H. Pöykkö, unpublished data). There exist several other natural enemies like avian predators, e.g. blue tit (Cyanistes caeruleus) and pied flycatcher (Ficedula hypoleuca), as well as predaceous invertebrates such as ants, beetles and bugs roaming on the tree trunks likely contributing to low survival of larvae. For example, a pentatomid bug (Picromerus bidens) found also on the study trees is fatal to lepidopteran larvae (Mahdian, Tirry & De Clercq 2007). Fruticose host structure may itself create shelter for larvae against both avian and invertebrate natural enemies, especially during the early instars when larvae can hide within the lobes of a lichen thallus. Moreover, attachment of host lichens by a single holdfast to the substratum may decrease their availability for generalist invertebrate predators roaming on tree trunks, thus creating enemy-free space through herbivores spatial dispersion pattern (Stamp 2001). This is also supported by the fact that young larvae are often found on the lobe tips of hanging lichens (personal observations).

Results of this study suggest that host structure–mediated escape from natural enemies may be mainly caused by reduced predation pressure from generalist natural enemies, in line with previous results from the same model species (Pöykkö 2011). Parasitism rate by specialist parasitoids appeared to be very low as only one larva of 84 in non-excluded treatments were found to be parasitized. This observation, however, likely underestimates parasitism rate as many parasitized larvae may have been attacked by predators or died otherwise during the experiment. Interestingly, other studies of host structure–mediated natural enemy escape have found that host structure reduces parasitism rate by specialized natural enemies rather than predation pressure from generalist enemies (Oppenheim & Gould 2002; Feder 1995). Both of those studies used larvae with unexposed feeding habits and their parasitoids as a model system. In addition, Oppenheim & Gould (2002) found that the use of enemy-free space is coupled with behavioural adaptations of herbivores decreasing their availability for parasitoids. Thus, the relative importance of generalist and specialist natural enemies on the host specialization and host range in general are likely species and context specific.

The impact of natural enemies on the survival of C. lichenaria larvae was significant only during early stages before overwintering, whereas after overwintering, no differences were detected in the survival of larvae between different host lichen forms. This is in agreement with many other studies on lepidopteran larvae that young instars are most vulnerable to natural enemies, especially to generalist predators (reviewed in Zalucki, Clarke & Malcolm 2002). Thus, the results hint that predaceous invertebrates are behind the low survival of larvae on flat lichens in the field (Pöykkö 2011). Larvae of C. lichenaria spend their whole larval period, lasting from the beginning of August at the latest until May of the next spring, on the tree trunks and are accordingly vulnerable to predation for a relatively long period. Predation pressure may vary between seasons being likely higher in autumn compared with winter and early spring. Predation by invertebrates is probably also higher in autumn (August–October) than in spring (April–May) and resident avian predators may be most important natural enemies during winter. Accordingly, high predation pressure by invertebrate predators during early larval stages might have selected for the existence of enemy-free space and preference of females and neonate larvae towards Ramalina species (Pöykkö 2006, 2011).

Host structure itself had a significant impact on the survival of larvae in both excluded and non-excluded treatments, especially during the early stage of larval development (Fig. 1). Thus, physical appearance of host lichen had a direct effect on the survival of larvae independent of natural enemies. Larvae may, for example, find shelter from direct sunshine or strong rain under the lobes of Ramalina lichens. On flat thalli direct sunshine, especially during July and August may increase the surface temperature of lichen thallus increasing also temperature of larvae and evaporation through larval cuticle, which may be detrimental for tiny larvae. Additionally, small scale topographical variation at the surface of flat thallus may decrease air flow near thallus surface preventing cooling of microclimate near the surface of flat lichen thallus. In turn, shrubby appearance of host thallus may allow stronger air flow through host lichen creating milder environment for larvae. Strong rain may either wash larvae away from flat thalli or tiny larvae may drown in water drops. In laboratory experiments, first and second instar larvae are very prone to drown in small water drops (personal observations).

Other potential factors contributing to enemy-free space

Potential host lichens for C. lichenaria larvae differ also in several other characteristics besides host structure, e.g. in secondary chemistry. Xanthoria parietina contains relatively high amount of parietin, c. 3% dry mass (Pöykkö, Hyvärinen & Bačkor 2005). In P. sulcata atranorin (including chloroatranorin) is a major cortical secondary compound (c. 0·5% dry mass). Both R. fraxinea and R. farinacea contain small amounts of usnic acid (0·05% and 0·5% dry mass, respectively) as a cortical compound (Pöykköet al. 2010). Usnic acid is known of its antiherbivore and antimicrobial activity (Cocchietto et al. 2002). Hence, one potential cause for enemy-free space might be that larvae sequester lichen secondary chemicals for their own defence, as several other insects feeding on higher plants (Nishida 2002). Although some lichen-feeding Lepidoptera indeed do sequester lichen secondary chemicals (Hesbacher et al. 1995; Karunaratne et al. 2002), C. lichenaria seems to have adapted to metabolize rather than sequester these chemicals (Pöykköet al. 2010). Thus, it is unlikely that enemy-free space of C. lichenaria on Ramalina species is based on sequestered lichen secondary chemicals.

Enemy-free space for C. lichenaria might also be explained by best background matching, as potential host lichens differ also in colouration. Ramalina species are greenish lichens, P. sulcata is blue-grey and X. parietina is yellow-orange. Anaptychia ciliaris L. Krber ex Massal., fifth potential host lichen used in field studies (Pöykkö 2011), is grey when dry, but turns to dark green after wetting. Colouration and patterns of lepidopteran larvae may vary according to host plants larvae feed on (Greene 1989; Fink 1995; Noor, Parnell & Grant 2008; Canfield, Chang & Pierce 2009). When considering larvae reared on X. parietina, it is not likely that larvae might receive as bright yellow-orange colouration as their host lichens. This is because yellow-orange colouration of X. parietina depends on its major cortical secondary compound, parietin and C. lichenaria larvae metabolize most of it and excrete the rest with faeces (Pöykköet al. 2010). Thus, because of colour difference between larva and host lichen, larvae reared at least on X. parietina are probably more detectable to predators and parasites compared with larvae reared on Ramalina species.


In conclusion, shrubby host structure itself increased the survival of larvae and simultaneously made larvae more difficult to be detected by natural enemies on shrubby host lichens. Further experiments are needed to explore the specific causes of mortality to confirm the results and to reveal interspecific interactions behind the narrowed host range of C. lichenaria. Besides natural enemies alone, also host as a physical environment need to be taken into account when evaluating the roles of different factors driving the evolution of host use in insects. Finally, enemy-free space of larvae on Ramalina species in natural conditions might best be explained by combined effects of host structure and background matching. In general, the role of secondary chemistry as a basis for enemy-free space should not be regarded self-evident.


This study was funded by Finnish Academy (project no. 120964). Comments and statistical advice from Sami Kivelä, Panu Välimäki and Juha Tuomi are highly appreciated. I am thankful to several anonymous referees of their invaluable comments on the manuscript, to the staff of Zoological Museum of Oulu University for preparing haircloth fences and to the staff of Husö Biological Station for their help during field work. Identification of a braconid species by Dr Gergely Várkonyi is highly appreciated.