Jack of all trades masters novel host plants: positive genetic correlations in specialist and generalist insect herbivores expanding their diets to novel hosts

Abstract

One explanation for the widespread host specialization of insect herbivores is the “Jack of all trades-master of none” principle, which states that genotypes with high performance on one host will perform poorly on other hosts. This principle predicts that cross-host correlation in performance of genotypes will be negative. In this study we experimentally explored cross-host correlations and performance among families in four species (two generalist and two specialist) of leaf beetles (Cephaloleia spp.) that are currently expanding their diets from native to exotic plants. All four species displayed similar responses in body size, developmental rates and mortality rates to experimentally controlled diets. When raised on novel hosts, body size of larvae, pupae and adults were reduced. Development times were longer and larval mortality was higher on novel hosts. Genotype × host plant interactions were not detected for most traits. All significant cross-host correlations were positive. These results indicate very different ecological and evolutionary dynamics than those predicted by the “Jack of all trades-master of none” principle.

Introduction

Plants and their associated phytophagous insects are an important component of the biological diversity (Mitter et al., 1991). One of the processes involved in the generation of this outstanding diversity is the co-adaptation between plants and their insect herbivores (Ehrlich & Raven, 1964; Janzen, 1980; Futuyma & Slatkin, 1983). A general pattern observed in plant-herbivore interactions is that the diets of most insect herbivores are specialized to one or a few host plants (Fox & Morrow, 1981; Jaenike, 1990; Thompson, 1995; Novotny & Basset, 2005). A broadly accepted explanation for this widespread resource specialization of insect herbivores is the “Jack of all trades-master of none” principle (McArthur, 1972; Futuyma & Moreno, 1988; Via, 1990; Futuyma et al., 1995). This principle proposes that there is a cost of adaptation to new hosts. Therefore a trade-off in performance between host plants is expected. Genotypes displaying high performance on a given host are expected to perform poorly on other hosts. Such a constraint of adaptation to multiple host plants implies that diet specialization would be selected over generalization. This principle is a central assumption of several models for the evolution of specialization (Barbosa, 1988; Futuyma & Moreno, 1988, Agrawal, 2007). A fundamental prediction of this principle is that the performance of herbivore genotypes will be negatively correlated across different hosts (Futuyma & Moreno, 1988; Ueno et al., 2003; Agosta & Klemens, 2009).

Negative genetic correlations in cross-host performance have been reported for some insect herbivores (Via, 1984; Mackenzie, 1996; Tilmon et al., 1998). However, most studies have reported positive or no correlation between a genotype’s performance on one host and its performance on a different host (Futuyma & Philippi, 1987; Ueno et al., 2003; Futuyma, 2008). An absence of correlation in cross-host performance means that performance of a genotype on a given host is not constrained by its performance on other host plants. A positive correlation means that genotypes that perform well on one host also perform well on other host plants. Positive correlations in cross-host performance represent very different ecological and evolutionary dynamics than those predicted by the “Jack of all trades-master of none” principle. With positive genetic correlations, high performance on one host plant is associated with high performance on other host plants, promoting generalization.

Although there has been considerable focus on the genetic architecture of specialist herbivores using different host plants, many insects are generalists, feeding on multiple host plants (Singer, 2001; García-Robledo et al., 2010). The observation that some insect species have been observed to use several environments (in this case multiple host plants), has led to several possible evolutionary scenarios. Evolution of composite generalized diets is a scenario in which there are multiple genotypes in a population, each displaying distinct performance among hosts. Evolution of plastic generalist genotypes is another scenario, one in which all the genotypes in a population can survive and reproduce successfully on several host plants (Powell, 1971; Powell & Wistrand, 1978; Fox & Morrow, 1981; Hawthorne, 1997; Agrawal, 2001, 2007). Relative performance of genotypes across host plants may differ between generalist and specialist herbivores. For example, composite generalization may translate into genotype × host plant interactions with negative correlations in performance across hosts, while plastic generalization would translate into similar performance among hosts with non-significant or positive correlations.

One critique of earlier studies that reported an absence of negative correlations is that these studies were conducted too late in the process of adaptation, at a time when the herbivores had already been interacting with “new” plants for several generations (Fry, 1993; Agosta & Klemens, 2008). After several generations, herbivores may have become adapted in which case trade-offs that would have been expected during the initial contacts of herbivores with new hosts may have already been reduced in the study populations. To address differential performance of genotypes during the initial phases of the process of adapting to new host plants, the critical study is one that is conducted on recently assembled plant-herbivore interactions. New plant-herbivore interactions between native insect herbivores and recently introduced exotic plants constitute a good study subject for such issues.

One of the oldest and most conservative plant-herbivore associations is the interaction between beetles of the neotropical genus Cephaloleia (Chrysomelidae: Cassidinae) and plants of the Zingiberales. Cephaloleia beetles and neotropical gingers evolved in the neotropics for the last 35–60 MY in isolation from paleotropical Zingiberales (Wilf et al., 2000; McKenna & Farrell, 2005; McKenna & Farrell, 2006; García-Robledo & Staines, 2008).

At La Selva Biological Station (Costa Rica, Central America), the tropical rain forest where we performed this research, several species of Cephaloleia are associated with native plants of the Zingiberales (Staines, 1996). During the last 10 – 20 years, at least five Paleotropical and one South American species of exotic Zingiberales have naturalized in areas surrounding our study site. At La Selva, the colonization and establishment of exotic Zingiberales inside La Selva is in an early stage (García-Robledo & Horvitz, 2011). In addition, exotic Zingiberales are removed by La Selva staff when detected inside the station. For these reasons, exotic Zingiberales inside La Selva are rare. At the time our study was conducted, at least seven Cephaloleia species had already been observed feeding on these exotic Zingiberales in the wild, generating 16 novel plant-herbivore interactions on the Station (Garcia-Robledo, 2010).

These newly assembled plant-herbivore associations provided an opportunity to study the differential performance of herbivore genotypes on newly encountered, novel host plants. In the present study, we use a quantitative genetic approach to experimentally investigate the effects of native vs. novel hosts on body size, developmental rates and survival rates of insect herbivores, testing the “Jack of all trades-master of none” hypothesis. We explore whether diet breadth is associated with differences in performance by using two generalist and two specialist species of Cephaloleia. These beetles are currently expanding their diets and, collectively, now include four Paleotropical and one South American Zingiberales as hosts in addition to their native hosts (Table 1). The objectives of this research were (1) to determine the effects of using novel hosts vs. native hosts as a food source on body size, developmental rates and life expectancy, (2) to estimate genotype × environment interactions and correlations in performance across host plants and (3) to quantify the proportion of variance explained by genotype for body size and developmental rates across the life cycle and for life expectancy. Heritability (the proportion of phenotypic variance due to genetics) is generally expected to be higher for traits expressed early in development than for traits expressed later (Von Dassow & Munro 1999). In this study we defined genotype as family group, where each family was a mix of half- and full siblings, and environment was defined as host plant species. We discuss if genotypes in recently assembled plant-herbivore interactions are constrained in their use of novel hosts, displaying negative or positive correlations in cross-host performance.

Table 1.

Cephaloleia beetles and the native and novel host plants selected for quantitative genetics experiments.

Beetle species	* Native host plant	Novel host plants	Origin of novel hosts
Cephaloleia belti (Generalist)
	Heliconia latispatha (Heliconiaceae)	Heliconia psittacorum (Heliconiaceae)	Caribbean, northern South America
		Musa velutina (Musaceae)	India
Cephaloleia dilaticollis (Generalist)
	Renealmia alpinia (Zingiberaceae)	Alpinia purpurata (Zingiberaceae)	Pacific islands
		Hedychium coronarium (Zingiberaceae)	Eastern India
Cephaloleia dorsalis (Specialist)
	Costus malortieanus (Costaceae)	Cheilocostus speciosus (Costaceae)	Malay Peninsula of Southeast Asia
Cephaloleia placida (Specialist)
	Renealmia alpinia (Zingiberaceae)	Alpinia purpurata (Zingiberaceae)	Pacific islands
		Hedychium coronarium (Zingiberaceae)	Eastern India

^*The native hosts selected for preference and survival experiments are the native hosts where both adult and larvae are most frequently found in the field. See García-Robledo et al. 2010 for a detailed description in diet breadth and host use for each Cephaloleia species.

Methods

Study site and species

We conducted this research at La Selva Biological Station (hereafter La Selva) from August 2005 to March 2009. La Selva is a tropical rain forest site in Costa Rica, Central America (10o26’N, 83o59’W). For this study we selected four Cephaloleia beetle species with contrasting diet breadths (Table 1). At La Selva, Cephaloleia belti is the species with the broadest diet breadth, feeding on 15 species from three families of Zingiberales (García-Robledo et al., 2010). The beetle Cephaloleia dilaticollis is also a generalist, feeding on ten species from three families of Zingiberales (García-Robledo et al., 2010). We also selected two specialists. Cephaloleia dorsalis is a specialist on the family Costaceae. At La Selva C. dorsalis was found feeding on at four species of the genus Costus. Cephaloleia placida is a specialist on the family Zingiberaceae. At La Selva this species feeds on two species of the genus Renealmia (García-Robledo et al., 2010).

Adults of these Cephaloleia species feed on the leaf tissue of the young rolled leaves of their host plants. In contrast, larvae feed on the leaf tissue of expanded leaves (García-Robledo et al., 2010). These four beetle species are currently expanding their diets at La Selva by including naturalized exotic hosts from India, the Malay Peninsula, the Pacific Islands and South America into their diets (Table 1).

Larval development and survival on native and novel hosts

In insects colonizing novel hosts, the effects of feeding on a novel host may vary throughout larval development. For example, growth rates on novel hosts may be slower in younger than in older larvae (García-Robledo 2010). Mortality may also be higher at earlier than at advanced larval stages (García-Robledo & Horvitz 2011). Therefore it is of interest to study the relative performance of insect herbivores on native and novel diets at distinct times during development. To determine if generalist vs specialist Cephaloleia differ in the degree to which their larvae are affected by feeding on a novel host, we performed the following experiment.

We collected males and females of each species from the native host species on which it is most frequently encountered in the field at La Selva Biological Station (García-Robledo et al. 2010; Table 1). Mating couples were placed in separate 17 ×15 ×5 cm containers and fed ad libitum with young leaf tissue from their native host plants (Number of mating couples: N_C.belti = 38, N_{C.dilaticollis} = 32, N_C.dorsalis = 37, N _C.placida = 42). In each container we also included four 10 × 10 cm squares of tissue from fully expanded leaf from the native host as an oviposition substrate. Leaf tissue was replaced with fresh leaves every 48 h. Eggs were carefully removed from the leaf surface and placed in containers lined with moist filter paper. We recorded the mating couple from which each egg was obtained. After eclosion, larvae obtained from each mating couple were randomly assigned to one of the following diets: leaf tissue from the native host or leaf tissue from the novel host plant. Parenthood records were used in further quantitative genetics analyses (see next section).

Each larva was placed in an individual container lined with moist filter paper. Larvae were fed every 48 h with two 3.5 cm diameter disks of leaf tissue. Larvae were reared at a mean temperature of 27°C and a light regime of 12 h. light 12 h. darkness (sample sizes: Cephaloleia belti, Table 2; C. dilaticollis, Table 3; C. dorsalis, Table 4; C. placida, Table 5).

Table 2.

ANOVA results comparing body size and developmental rates of Cephaloleia belti (Generalist) reared on native and novel host plants.

Beetle species	Traits	Sample size			DF	F	P
		H. latispatha (Native)	H. psittacorum (Novel)	M. velutina (Novel)
Cephaloleia belti (Generalist)
	Length (mm)
	Newborn larvae	348	392	403	2	0.60	0.55
	Larvae instar 1	292	295	278	2	355.19	< 0.0001
	Larvae instar 2	254	272	252	2	47.13	< 0.0001
	Pupae	278	250	228	2	75.033	< 0.0001
	Adults	210	173	163	2	31.64	< 0.0001
	Weight (mg)
	Pupae	210	173	162	2	59.21	< 0.0001
	Adults	210	173	162	2	59.20	< 0.0001
	Development time (d)
	Time as a larva	278	250	228	2	115.17	< 0.0001
	Time as a pupa	210	173	163	2	1.99	0.1372

Table 3.

ANOVA results comparing body size and developmental rates of Cephaloleia dilaticollis reared on native and novel host plants.

Beetle species	Traits	Sample size			DF	F	P
		R. alpinia (Native)	A. purpurata (Novel)	H. coronarium (Novel)
Cephaloleia dilaticollis (Generalist)
	Length (mm)
	Newborn larvae	566	556	215	2	2.9	0.05
	Larvae instar 1	253	255	61	2	42.03	< 0.0001
	Larvae instar 2	100	126	39	2	9.46	< 0,0001
	Pupae	142	129	27	2	14.94	< 0.0001
	Adults	74	75	10	2	16.14	< 0.0001
	Weight (mg)
	Pupae	142	129	27	2	51.03	< 0.0001
	Adults	74	75	10	2	25.24	< 0.0001
	Development time (d)
	Time as a larva	143	130	27	2	26.35	< 0.0001
	Time as a pupa	74	75	11	2	0.26	0.77

Table 4.

Welch two-sample T test results comparing the body size and developmental rates of Cephaloleia dorsalis reared on native and novel host plants.

Beetle species	Traits	Sample size		DF1	t	P
Cephaloleia dorsalis (Specialist)		C. malortieanus (Native)	Ch. speciosus (Novel)
	Length (mm)
	Newborn larvae	338	323	658.63	−1.24	0.2153
	Larvae instar 1	175	130	264.74	7.88	<0.0001
	Larvae instar 2	74	77	131.65	4.03	<0.0001
	Pupae	116	78	164.7	2.77	0.0062
	Adults	86	58	129.18	2.86	0.0049
	Weight (mg)
	Pupae	116	78	170.84	4.43	<0.0001
	Adults	85	58	117.08	4.07	<0.0001
	Development time (d)
	Time as a larva	115	80	153.88	−4.26	<0.0001
	Time as a pupa	84	58	137.49	0.8445	0.3999

¹Degrees of freedom associated with variance estimate are approximated using the Welch-Satterthwaite equation (Ruxton, 2006).

Table 5.

ANOVA and Welch’s t- test results comparing body size and developmental rates of Cephaloleia placida reared on native and novel host plants.

Beetle species	Traits	Sample size			DF	Statistic1	P
		R. alpinia (Native)	A. purpurata (Novel)	H. coronarium (Novel)
Cephaloleia placida (Specialist)
	Length (mm)
	Newborn larvae	272	220	240	2	F = 0.71	0.4893
	Larvae instar 1	109	49	82	2	F = 179.45	< 0.0001
	Larvae instar 2	39	39	63	2	F = 98.78	< 0.0001
	Pupae	91	26	9	2	F = 45.40	< 0.0001
	Adults	70	14	--	1	t = −2.54	0.021
	Weight (mg)
	Pupae	91	26	9	2	F = 36.89	< 0.0001
	Adults	70	14	--	1	t = −3.96	< 0.0001
	Development time (d)
	Time as a larva	91	26	9	2	F = 72.92	< 0.0001
	Time as a pupa	71	14	--	1	t = 0.01	0.922

¹F statistic provided for ANOVA tests exploring differences among three hosts. t statistics provided for Welch’s t-tests exploring

differences between two host plants.

To measure larval growth rates, we measured the length of each newborn larva from the tip of the head to the tip of the abdomen. This initial measure was performed to ensure that larvae assigned to different treatments displayed similar initial sizes. We performed two additional length measurements during larval development, one at the mean estimated time of transition on native hosts from first to second instar (i.e the end of the first instar) and the other at the time of transition from the second instar to the pupation on native hosts (i.e. the end of the second instar) (see estimated transition times for each species in Figure 1 and García-Robledo et al. 2010).

Figure 1.

Length (Mean ± SD) of larvae, pupae and adults of generalist and specialist Cephaloleia beetles reared on native and novel host plants. Larval lengths were measured after hatching, at instar 1, and the end of instar 2 development times. A. Cephaloleia belti. HL: Heliconia Latispatha, HP: Heliconia psittacorum, MV: Musa velutina. B. Cephaloleia dilaticollis. RA: Renealmia alpinia, AP: Alpinia purpurata, HC: Hedychium coronarium. C. Cephaloleia dorsalis. CM: Costus malortieanus, CS: Cheilocostus speciosus D. Cephaloleia placida. RA: Renealmia alpinia, AP: Alpinia purpurata, HC: Hedychium coronarium. Letters on the bars group similar categories (P < 0.05).

We measured larval lengths using a digital camera (Diagnostic Instruments Inc. Model 3.2.0) attached to a stereoscope (Leica MZ 12s). Lengths of the larvae were estimated on the digital images at an accuracy of 10⁻² mm, using the program Spot V.3.5.8 (Diagnostic Instruments Inc. Sterling Heights, MI). Data were log-transformed. Differences in length among larvae were tested with one-way ANOVA’s.

Pupal lengths and weights were measured on the day of pupation. Weight was measured using an analytic balance Scientech SA 40 with a precision of 10⁻⁴ g and log-transformed. Adult lengths and weights were measured on the day of adult emergence and log-transformed. For each individual we recorded the time from larval eclosion to pupation, and from pupation to adult emergence. Differences in pupal length, pupal weight, adult length, adult weight and development times of individuals on the native vs novel hosts were tested with one-way ANOVA’s, when there were more than two host plants and by Welch’s t-tests when there were only two hosts being compared. The Welch’s t-test is a modification of the t test for independent samples that does not assume equal population variances, where degrees of freedom are associated with the variances through the Welch-Satterthwaite equation (Welch, 1947).

To estimate larval survival on native and novel host plants, we monitored each larva every 48 h until death or pupation. Differences in larval mortality between native and novel hosts were determined by Cox proportional hazard survival analyses. Larvae that pupated were counted as right-censored observations. Cox proportional hazard examines the effects of covariates or continuous independent variables on the risk of death at different ages. It is particularly useful for making contrasts based on covariate values without being particularly concerned with the shape of the survival function for the group designated as the base line pattern. Thus the risk of death is modeled for individuals based on the values of their covariates, relative to a baseline pattern. This type of analysis is a regression model similar to life table survival analysis such as Kaplan-Meier tests. Cox proportional hazard models can include interactions among multiple covariates, allowing exploration of the genotype × environment interaction effects on survival. The model is specified as:

$$h_i(t)=h_0(t)exp({\beta }_1{X}_{i1}+{\beta }_2{X}_{i2}+\dots +{\beta }_k{X}_{ik})$$

Where h_o(t) is the baseline or reference hazard function that is changed by the values. The covariates X and regression coefficients β’s. The values of the covariates X_ik’s varies among individuals (i) essentially this type of analysis considers the relative likelihood of death at a given time among individuals based on the values of their covariates, in our case genetic family and host plant (Fox, 2001).

Quantitative genetics analyses

During the development and survival experiments we recorded the female from which each larva was obtained. The offspring of each female (here after a genetic family) represents a mixture of half and full siblings because the females were collected in the wild, potentially having already mated with one or more males before being incorporated into our experiment. The eggs from one female constitute a family group that are at least half sibs. The members of these family groups share both additive and non-additive genetic variance such as dominance, epistatic and maternal effects. Therefore the analyses reported in this study are based on broad-sense heritabilities (H²).

A change in diet may have different effects on performance at early or advanced developmental stages. The response of genotypes to the novel conditions may vary across development. For this reason, in the following quantitative genetics analyses we report genotype responses to native and novel diets at different points during development.

For the following analyses we grouped the data obtained for all individuals during the experiments, testing for differences in development and survival on native and novel hosts by their respective genetic families. In cases where genetic families had no individuals surviving on one of the host species, genetic families were removed from the analyses (see sample size in Figure 5).

Figure 5.

Summary of pairwise correlations for developmental traits and larval survival of Cephaloleia families reared in original and novel host plants. For each comparison, abbreviations indicate a host plants (Host plant abbreviations as in Figure 1). Sample size (number of families) is given in parentheses for each correlation. A. Cephaloleia belti. B. Cephaloleia dilaticollis. C. Cephaloleia dorsalis. D. Cephaloleia placida. Asterisks indicate levels of significance (based on Pearson’s product moment correlation coefficients) *P < 0.10, **P < 0.05, ***P < 0.01.

Family × host interactions of development and survival on native and novel hosts

To determine if generalist and specialist Cephaloleia species have genetic variation in body size and developmental rates, we estimated the proportion of variance due to family and compared the mean performance of genetic families for the following traits on native and novel hosts: a. length of larvae at instar 1, b. larval length at instar 2, c. pupa length, d. adult length, e. pupa weight, f. adult weight, g. development time from larval eclosion to pupation and h. development time from pupation to adult emergence. Larvae, pupae and adult lengths were log-transformed. For each trait, differences among families were estimated with an ANOVA model.

Family × host interactions were estimated for each trait. Analyses were performed used linear mixed-effects models, employing restricted maximum likelihood (REML) (R-Development-Core-Team, 2009). Only larvae, pupae and adult lengths were log-transformed. The diet assigned to each larva (native vs novel hosts) was included as a fixed factor and genetic family as a random factor.

To determine if generalist and specialist Cephaloleia species reared on native and novel hosts display genetic variation in survival, we performed the following analyses. Differences in survival among families were tested for each beetle species with mixed Cox proportional hazard models (Package Survival, R-Development-Core-Team, 2009). The model included larval diet as a fixed factor and genetic family as a random factor (see sample size in Table 7). We determined family × host plant interactions, by comparing the survival models with or without the family × host plant interaction term. Comparisons between models were performed using likelihood-ratio tests.

Table 7.

Cox proportional hazard analyses for differences in larval survival in Cephaloleia beetles reared on native and novel host plants. N = the number of families.

Beetle species	Log-likelihood	χ2	N	P
Cephaloleia belti (Generalist)
Host plant	−2501.1	35.21	36	<0.0001
Family	−2492.0	18.11	36	<0.0001
* Family × Host	−2492.04; −2492.29	0.49	36	0.78
Cephaloleia dilaticollis (Generalist)
Host plant	−6719.5	14.04	30	0.0008
Family	−6685.3	68.34	30	<0.0001
* Family × Host	−6685.3; −6691.6	12.49	30	0.0019
Cephaloleia dorsalis (Specialist)
Host plant	−2728.6	5.44	35	0.0196
Family	−2711.6	33.99	35	<0.0001
* Family × Host	−2711.6; −2712.4	1.53	35	0.2158
Cephaloleia placida (Specialist)
Host plant	−3404.4	17.352	38	0.0002
Family	−3361	86.83	38	<0.0001
* Family × Host	−3361.0; −3361.9	1.87	38	0.3924

^*Log-likelihoods for model comparisons between the model without Family × Host

interaction and the model including the interaction term.

Genetic correlations of body size, development rates and survival on native and novel hosts

To determine if performances of genetic families on native and novel hosts are negatively correlated, as predicted by the “Jack of all trades-master of none” principle, we performed the following analysis. For each genetic family we estimated the mean performance on native and novel hosts for the following traits: a. larval length at instar 1, b. larval length at instar 2, c. pupa length, d. adult length, e. pupa weight, f. adult weight, g. development time from larval eclosion to pupation, h. development time from pupation to adult emergence and i. larval survival (see sample size in Figure 5). We estimated correlations in the mean relative performance among host plants using Pearson’s product moment correlations. To meet the assumptions of normality and homocedasticity required for these analyses, we performed the following data transformations: In all Cephaloleia species, larval, pupal and adult lengths were log-transformed. Survival proportions were arcsin transformed.

Results

Larval body size and developmental rates on native and novel hosts

Larvae pupae and adult length

For all four species of Cephaloleia, larvae, pupae and adults were larger on the native than on the novel host plants. For the generalist beetle Cephaloleia belti, larvae, pupae and adults were larger on the native host H. latispatha than on the novel hosts (Figure 1A, Table 2). Larvae in first instar, were larger on the novel host M. velutina than on the novel host H. psittacorum. However, individuals in second larval instar, pupa and adults were larger when reared on the novel host H. psittacorum than when reared on the novel host M. velutina. (Figure 1A, Table 2).

Larvae, pupae and adults of the generalist beetle C. dilaticollis reared on the native host R. alpinia and the novel host A. purpurata were larger than individuals reared on the novel host H. coronarium (Figure 1B, Table 3).

Larvae, pupae and adults of the specialist C. dorsalis reared on the native host C. malortieanus were larger than individuals reared on the novel host Ch. speciosus (Figure 1C, Table 4). In the specialist beetle C. placida, larvae and pupae were larger in larvae reared on the native host than in larvae reared on the novel hosts. Larval and pupal mortality was very high for C. placida larvae reared on the novel host H. coronarium; no adults emerged from pupae reared in this novel host (Figure 1D, Table 5). Adults were larger when reared on the native host R. alpinia than on the novel host A. purpurata (Table 5).

Pupal and adult weight

In general, pupae and adults were heavier when reared on native hosts. Pupae and adults of the generalist C. belti were heavier when reared on the native host H. latispatha than individuals reared on the novel hosts H. psittacorum and M. velutina (Figure 2A, Table 2). Pupae and adults reared on the novel host H. psittacorum displayed the lowest weights (Figure 2A, Table 2).

Figure 2.

Weight (Mean ± SD) of pupae and adults of generalist and specialist Cephaloleia beetles reared on native and novel host plants. A. Cephaloleia belti. B. Cephaloleia dilaticollis. C. Cephaloleia dorsalis. D. Cephaloleia placida. Letters on the bars group similar categories (P < 0.05). Host plant abbreviations as in Figure 1.

Pupae and adults of the generalist C. dilaticollis were heavier when reared on the native host R. alpinia than individuals reared on the novel hosts A. purpurata and H. coronarium (Figure 2B, Table 3). Pupae and adults reared on the novel host H. coronarium displayed the lowest weights (Figure 2B, Table 3).

Pupae and adults of the specialist C. dorsalis were heavier when reared on the native host C. malortieanus than individuals reared on the novel host Ch. speciosus (Figure 2C, Table 4). Pupae of the specialist C. placida reared on the native host R. alpinia were heavier than individuals reared on the novel hosts A. purpurata and H. coronarium. Pupal mortality was very high for larvae and pupae of C. placida on the novel host H. coronarium; no adults emerged from pupae reared in this novel host (Figure 2D, Table 5). Adults were heavier when reared on the native host than on the novel host A. purpurata.

Development time

In general, development from egg eclosion to pupation was faster in individuals reared on the native than in individuals reared on the novel host plants (Figure 3). However, development time from pupation to adult emergence was not affected by larval diet (Figure 3, Tables 2 to 5).

Figure 3.

Larval and pupal development times (Mean ± SD) of generalist and specialist Cephaloleia beetles reared in original and novel host plants. A. Cephaloleia belti. B. Cephaloleia dilaticollis. C. Cephaloleia dorsalis. D. Cephaloleia placida. Letters on the bars group similar categories (P < 0.05). Host plant abbreviations as in Figure 1.

Larval survival on native and novel hosts

For all four species of Cephaloleia, larval survival was usually higher on native than on novel host plants (Figure 4, Table 7). Larval survival of the generalist C. belti was higher on the native host H. latispatha, and equivalent on the novel hosts H. psittacorum and M. velutina (Figure 4A, Table 7). The generalist C. dilaticollis displayed equivalent survival on the native host R. alpinia and the novel host A. purpurata (Figure 4B, Table 7). Mortality of C. dilaticollis larvae in H. coronarium was high (Figure 4B, Table 7).

Figure 4.

Larval survival of generalist and specialist Cephaloleia beetles reared in original and novel host plants. A. Cephaloleia belti. B. Cephaloleia dilaticollis. C. Cephaloleia dorsalis. D. Cephaloleia placida. Differences in survival among host plants, P < 0.05. Differences between host plants: Kaplan-Meier Survival Analysis with Log-rank Significance Test, P<0.05. Host plant abbreviations as in Figure 1.

Larval survival of the specialist C. dorsalis was higher on the native host C. malortieanus than on the novel host Ch. speciosus (Figure 4C, Table 7). Larval survival of the specialist C. placida was higher on the novel host R. alpinia than on the novel hosts A. purpurata and H. coronarium (Figure 4D, Table 7). Mortality of larvae reared on the novel host H. coronarium was high (Figure 4D, Table 7).

Quantitative genetic analyses

Family × host interactions of development and survival on native and novel hosts

In general, we found that families differed significantly for most of the body size parameters and developmental rates that we measured in the laboratory. However, for most of the traits, we did not find family × host plant interactions (Table 6). Families of the generalist C. belti displayed differences in mean larval, pupal and adult size and weight as well as in development times (Table 6). We only found differences in family × host interactions for the larval length measured during the first instar (Table 6, see estimates of H² for each trait in Appendix 1).

Table 6.

ANOVA table for the comparison among genetic families and genotype × environment interactions for body size and developmental rates of Cephaloleia on native and novel host plants.

Trait	Factor	Herbivore species
		Cephaloleia belti (Generalist)			Cephaloleia dilaticollis (Generalist)			Cephaloleia dorsalis (Specialist)			Cephaloleia placida (Specialist)
		DF (num/den)	F	P	DF (num/den)	F	P	DF (num/den)	F	P	DF (num/den)	F	P
Length (mm)
Instar 1
	Host	2/756	370.95	<0.0001	2/422	32.64	<0.0001	1/232	62.68	<0.0001	1/110	189.7	< 0.0001
	Family	34	2.17	0.0002	22	1.94	0.0069	31	2.29	<0.0001	20	2.38	0.0019
	Family × Host	68/756	1.40	0.0218	44/422	0.82	0.7852	31/232	1.08	0.3551	60/110	0.73	0.9054
Instar 2
	Host	2/669	46.49	<0.0001	2/155	8.19	<0.0001	1/89	18.37	<0.0001
	Family	34	48.94	0.0003	17	3.43	<0.0001	24	2.02	0.0094	8	2.69	0.0197
	Family × Host	68/669	1.22	0.1188	34/155	0.95	0.5558	24/89	1.49	0.0928	24/28	1.53	0.1405
Pupae
	Host	2/647	81.27	<0.0001	1/218	1.02	0.3139	1/127	7.00	0.0092
	Family	34	81.84	<0.0001	23	1.77	0.0193	24	2.12	0.004	--*	--*	--*
	Family × Host	68/647	0.98	0.535	23/218	0.59	0.9333	24/127	1.39	0.1262	--*	--*	--*
Adults
	Host	2/438	32.71	<0.0001	1/102	7.52	0.0072	1/83	6.16	0.0151
	Family	30	1.89	0.0035	21	0.94	0.5443	18	3.33	<0.0001	--*	--*	--*
	Family × Host	60/438	8.84	0.7976	21/102	0.91	0.5785	18/83	1.62	0.073	--*	--*	--*
Weight (mg)
Pupae
	Host	2/647	79.53	<0.0001	1/217	20.24	<0.0001	1/127	19.97	<0.0001	--*	--*	--*
	Family	34	3.26	<0.0001	22	1.94	0.0089	24	1.91	0.0118	--*	--*	--*
	Family × Host	68/647	1.13	0.2302	22/217	0.72	0.8148	24/127	1.14	0.375	--*	--*	--*
Adults
	Host	2/437	63.50	<0.0001	1/102	18.00	<0.0001	1/82	17.54	<0.0001
	Family	30	2.61	<0.0001	21	1.08	0.3776	18	3.47	<0.0001	--*	--*	--*
	Family × Host	60/437	0.86	0.7601	21/102	0.98	0.4957	18/82	1.55	0.0934	--*	--*	--*
Development time (d)
Egg to pupa
	Host	2/647	129.83	<0.0001	1/220	17.85	<0.0001	1/107	18.32	<0.0001
	Family	34	3.90	<0.0001	23	1.45	0.0885	23	1.55	0.0662	--*	--*	--*
	Family × Host	68/647	1.17	0.1731	23/220	0.75	0.7919	46/107	0.66	0.9438	--*	--*	--*
Pupa to adult
	Host	2/438	0.98	0.3765	1/102	0.61	0.4356	1/66	0.08	0.7730
	Family	30	1.92	0.0029	21	0.43	0.986	17	0.91	0.5695	--*	--*	--*
	Family × Host	60/438	1.28	0.0881	21/102	0.53	0.9504	34/66	0.27	0.9999	--*	--*	--*

^*Trials not performed because larval mortality was high, precluding obtaining the required adults for longevity trials.

In the generalist C. dilaticollis, pupal and adult mortality were high on the novel host H. coronarium (Figure 4B). For this reason, we only explored genetic variation and family × host plant interactions for families reared on the native host R. alpinia and the novel host A. purpurata for the following traits: pupal and adult lengths, pupal and adult weights and development times (Table 6). Families of the generalist C. dilaticollis displayed differences in mean larval and pupal length, pupal weight and development time from egg eclosion to pupation (Table 6). The length and weight of adult C. dilaticollis and the development time from pupation to adult emergence were similar among families. Family × host interactions were not significant (Table 6, see estimates of H² for each trait in Appendix 1).

Families of the specialist C. dorsalis displayed different larval and adult lengths and weights. Families of C. dorsalis displayed equivalent development times from larval eclosion to pupation and from pupation to adult emergence (Table 6). Family × host interactions were not significant (Table 6, see estimates of H² for each trait in Appendix 1).

Families of the specialist C. placida displayed differences in larval length (Table 6). Larval and pupal mortality were very high on the novel host plants (Figure 4D). Therefore, differences among families in size, weight and development times were not measured for pupae and adults of C. placida.

For the four species of Cephaloleia beetles, larval survival was different among genetic families (Table 7). We found family × host interactions in larval survival only for the generalist C. dilaticollis. (Table 7, see estimates of H² for each trait in Appendix 1).

Genetic correlations of development and survival on native and novel hosts

In general, we did not detect negative correlations for the developmental traits or survival in families reared on native or novel hosts. In the four Cephaloleia species, all correlations were positive or not significant (Figure 5).

In the generalist C. belti, nine correlations were positive, and 18 not significant (Figure 5A). In the generalist C. dilaticollis, pupal and adult mortality were high on the novel host H. coronarium (Figure 4B). For this reason in some of the traits we only analyzed the correlations in performance between the native host R. alpinia and the novel host A. purpurata (Figure 5B). In C. dilaticollis, two correlations were positive and 13 were not significant (Figure 5B).

In the specialist C. dorsalis we found three positive and six non-significant correlations (Figure 5C). In the specialist C. placida, larval mortality was high in both novel hosts (Figure 4D). For this reason we only analyzed the correlations in performance for the traits directly associated with larval development and survival (Figure 5D). In this species we found three positive and six non- significant correlations.

Discussion

Our results suggest no evident physiological advantage for generalist Cephaloleia species over specialists in colonizing novel hosts. Instead, we recorded similar responses for all beetle species. Development time was extended and survival was reduced on novel hosts. As a result, mean phenotypes on native and novel hosts were different. Adults were smaller on novel than on native hosts. The high mortalities of larvae on novel hosts show that novel hosts are challenging. This is not a surprising result, as the fitness of insect herbivores is usually reduced during early colonization of new environments (Futuyma & Moreno, 1988; Scheirs et al., 2000; Forister et al., 2009).

For both generalists and specialists, we detected substantial differences in performance among genetic families when reared on native and novel hosts. Differences of performance among families on original hosts suggest that genotypes within the population have intrinsic differences in their ability to use the native host plants. The fact that Cephaloleia beetles are capable to some degree of recognizing, feeding and growing on novel hosts is evidence of diet plasticity. However, an absence of genotype × host plant interactions show that the plastic responses of genotypes to novel hosts are proportional to their performance on native hosts. This pattern is opposite to the trade-off predictions of the “Jack of all trades-master of none” principle, where it is expected that genotypes with high performance in one host plant will perform poorly in other hosts (Futuyma & Moreno, 1988; Jaenike, 1990). Our results suggest for the novel plant-herbivore interactions described in this study a scenario of pre-adapted genotypes, where a high performance in the original environment is a precondition for success in novel environments.

Our results exploring cross-host correlations in performance also support a scenario of pre-adapted genotypes. Similar to other studies, we found that all significant correlations in performance were positive (Rausher, 1984; Agosta, & Klemens, 2009). We did not find evidence for negative correlations in cross-host performance, that is main prediction of the “Jack of all trades-master of none” principle (Falconer & Mackay, 1996).

Some authors suggest that the absence of negative cross-host correlations in performance is an artifact of the inadequacy of genetic correlations to detect trade-offs (Fry, 1993, 1996). The three main arguments against genetic correlations as a method to detect trade-offs in performance are: 1. Trade-offs in host use may be quickly ameliorated by selection. 2. Genetic trade-offs may be obscured by some environmental effects such as maternal effects and 3. Statistical methods such as genetic correlations have limitations in detecting negative trade-offs if genetic covariances are non-negative (Rausher, 1988; Fry, 1993, 1996).

Plant-herbivore interactions included in this study are recent. Therefore, most likely natural selection did not have the opportunity to ameliorate trade-offs in host use. Although it is true that our experiments include environmental effects, natural selection acts on phenotypes. Even if genetic correlations are masked by non-additive genetic variance components or non-negative genetic covariances, negative correlations in cross-host performance should be of sufficient magnitude for natural selection to generate trade-offs in performance.

An alternative experimental design to detect trade-offs in performance is selection experiments (Fry, 1993). However, evidence from selection experiments is also not conclusive. Although in some cases selection experiments support trade-offs in performance, there are also several cases that support scenarios of pre-adaptated genotypes or no trade-offs at all (Fry, 1990; Agrawal, 2000; Yano et al., 2001).

It is noteworthy that although we tested for genotype × environment interactions and genetic correlations in performance for four herbivore species reared on multiple native and novel host plants, we found absolutely no evidence for trade-offs in host use. There is no doubt that trade-offs play an important role shaping evolutionary processes (Roff, 2002). However, in the case of recently assembled plant-herbivore interactions such as those described in this study, genotypes apparently are not constrained by negative genetic correlations (Rausher, 1984).

A potential reason why genotypes of Cephaloleia beetles don’t seem to be constrained is because all diet expansions described in this study are phylogenetically conservative. As reported for the majority of insect herbivores acquiring new host plants, Cephaloleia beetles expanded their diets to exotic plants that are taxonomically and maybe chemically similar to their original hosts (Janzen, 1968; Schultz, 1988). Insect herbivores usually possess some degree of behavioral and physiological plasticity to cope with the chemical variation among host plants (Chew, 1981; Rausher, 1984;Thomas et. al., 1987). This variation originates from selection pressures that promote the use of multiple plant genotypes, plant parts or even different host plant species (Ueno et al. 2003). It is possible that the chemical and nutritional characteristics of exotic Zingiberales are within the plastic range of Cephaloleia herbivores. Therefore herbivores can recognize and consume leaf tissue from these novel hosts.

Plasticity in the use of different resources is known to play a central role in the origin of ecological novelty (West-Eberhard, 1989, 2003). For example, insect herbivore phenotypic plasticity will facilitate the initial colonization of a novel host plant (Chew, 1981; Berenbaum & Zangerl, 1991; Graves & Shapiro, 2003). However, if the novel conditions are challenging, and mortality on the novel host is high, the failure or success of colonization will depend on genetic variation available for adaptation to the novel conditions (Rausher 1984). Colonization of novel environments is also possible if insect herbivores display limited genetic variation. For example, insect herbivores will be able to colonize novel hosts without substantial evolutionary change if the vital rates in the novel environment lead to positive population growth rate (Janzen, 1985).

In this study we found that all four herbivores showed some degree of plasticity in the use of new resources. However, broad-sense heritability estimates were very low (see H² estimates in Appendix 1). One question that rises from our results is if these initial conditions will preclude or facilitate the initial colonization of novel hosts by Cephaloleia beetles.

An experimental approach to answer to this question is to estimate the instantaneous population growth rates for each Cephaloleia species on both native and novel host plants. Negative instantaneous growth rates suggest that the initial colonization of a host plant would lead to extinction, and evolutionary change is required within the population before colonizing the novel host (Carey, 2001). Positive instantaneous growth rates suggest that the genetic variation and/or phenotypic plasticity already available within the population would lead to a successful colonization of a novel host, and natural selection can act on the genotypes that already colonized the novel environment (Carey, 2001).

In a previous demographic study, we estimated instantaneous growth rates for all Cephaloleia species included in this study on both native and novel hosts (García-Robledo & Horvitz, 2011). For all herbivore species, instantaneous population growth rates were lower on novel than on the native host plants, showing that fitness of both generalist and specialist herbivores are reduced during early colonization of host plants (García-Robledo & Horvitz, 2011). In this previous demographic study, we recorded negative instantaneous population growth rates for some of the novel plant-herbivore interactions such as the generalist C. dilaticollis and the specialist C. placida colonizing the novel host Hedychium coronarium. Instantaneous population growth rates were also negative for Cephaloleia placida colonizing the exotic host Alpinia purpurata. However, instantaneous population growth rate were positive in several novel plant-herbivore interactions. The generalist C. belti and the specialist C. dorsalis displayed positive instantaneous population growth rates on native and exotic host plants. Instantaneous population growth rate was also positive for generalist C. dilaticollis beetles colonizing the novel host Alpinia purpurata (García-Robledo & Horvitz, 2011). This suggests that although in some cases the characteristics of novel hosts will prevent herbivores diet expansions, it is also common for herbivores to be pre-adapted and colonize novel hosts without the requirement of substantial evolutionary change (Chew, 1981; Rausher, 1984; Berenbaum & Zangerl, 1991).

In conclusion, our results support the idea that the genotypes of insect herbivores expanding their diets to novel hosts are not constrained by trade offs in performance, leading to diet specialization. On the contrary, better performance of genotypes on native hosts is associated with better performance on novel hosts and diet generalization. This study shows that the assemblage of novel plant-herbivore interactions is plastic. Diet expansions to novel hosts such as those described for Cephaloleia beetles, in addition to facilitate the colonization of novel hosts, may also play a fundamental role on the recolonization of historical hosts, the colonization of new geographic localities and the survival of native insect herbivores on exotic host plants.

APPENDIX 1

To estimate broad-sense heritabilities (H²), we calculated for each trait the variance components from linear mixed models that included genetic family as a random factor (Lynch & Walsh, 1998). Variance components were estimated for families on native and novel hosts, using the function lmer, package lme4 in Program R (R-Development-Core-Team, 2009). Estimates of H² for survival on native and novel hosts were calculated using only the uncensored data from the Cox proportional hazard analyses.

Broad-sense heritability estimates were obtained from unknown mixes of half and full siblings. Therefore H² must be a value between estimates assuming that larvae within each family are half-siblings (Equation 1) or full-siblings (Equation 2).

$${H^2}_{\text{half}\text{siblings}}=4\times \frac{{{\sigma }^2}_{\text{Family}}}{{{\sigma }^2}_{\text{Total}}}$$ (Equation 1)

$${H^2}_{\text{full}\text{siblings}}=2\times \frac{{{\sigma }^2}_{\text{Family}}}{{{\sigma }^2}_{\text{Total}}}$$ (Equation 2)

In this appendix, we report the observational variance components, i.e. the fraction $\frac{{{\sigma }^2}_{\text{Family}}}{{{\sigma }^2}_{\text{Total}}}$ (Lynch & Walsh, 1998). Estimates of causal variance assuming half sib or full sib relationships can be calculated from values in appendix 1 by multiplying the observational variance component by four or two respectively (Lynch & Walsh, 1998).

Appendix 1A.

Estimates of H² for larvae of Cephaloleia belti reared on the native host Heliconia latispatha (HL), or the novel hosts Heliconia psittacorum (HP) and Musa velutina (MV).

Traits	Host plants
	HL (Native)	HP (Novel)	MV (Novel)
Length (mm)
Instar 1	~ 0	1.62 × 10 −1	6.35 × 10 −2
Instar 2	5.32 × 10 −2	1.11 × 10 −1	6.59 × 10 −2
Pupa	8.30 × 10 −2	1.14 × 10 −1	1.18 × 10 −1
Adult	6.84 × 10 −11	2.85 × 10 −11	~ 0
Weight (g)
Pupa	1.05 × 10 −1	1.33 × 10 −1	6.88 × 10 −2
Adult	1.13 × 10 −1	1.49 × 10 −9	9.62 × 10 −10
Development time (d)
Larva-pupa	1.51 × 10 −1	1.45 × 10 −1	8.56 × 10 −2
Pupa-adult	~ 0	2.56 × 10 −10	6.85 × 10 −2
Survival (d)
Time to death	~ 0	9.49 × 10 −2	1.58 × 10 −10

Appendix 1B.

Estimates of H² for larvae of Cephaloleia dilaticollis reared on the native host Renealmia alpinia (RA), or the novel hosts Alpinia purpurata (AP) and Hedychium coronarium(HC).

Traits	Host plants
	RA (Native)	AP (Novel)	HC (Novel)
Length (mm)
Instar 1	2.53 × 10 −10	3.21 × 10 −2	~ 0
Instar 2	1.49 × 10 −1	2.55 × 10 −1	1.16 × 10 −1
Pupa	5.41 × 10 −11	6.47 × 10 −12	--*
Adult	~ 0	~ 0	--*
Weight (g)
Pupa	8.72 × 10 −2	9.54 × 10 −10	--*
Adult	5.53 × 10 −11	~ 0	--*
Development time (d)
Larva-pupa	7.60 × 10 −2	~ 0	--*
Pupa-adult	~ 0	6.90 × 10 −12	--*
Survival (d)
Time to death	~ 0	6.41 × 10 −2	~ 0

^*Trials not performed because larval mortality was high, precluding obtaining the required adults for longevity trials.

Appendix 1C.

Estimates of H² for larvae of Cephaloleia dorsalis reared on the native host Costus malorteanus(CM), or the novel hosts Cheilocostus speciosus (CS).

Traits	Host plants
	CM (Native)	CS (Novel)
Length (mm)
Instar 1	8.56 × 10−2	1.58 × 10−1
Instar 2	1.23 × 10−11	1.26 × 10−1
Pupa	2.23 × 10−11	3.57 × 10−1
Adult	2.27 × 10−1	3.91 × 10−1
Weight (g)
Pupa	6.68 × 10−2	1.93 × 10−1
Adult	2.96 × 10−1	3.56 × 10−1
Development time (d)
Larva-pupa	1.10 × 10−1	1.26 × 10−10
Pupa-adult	~ 0	5.88 × 10−11
Survival (d)
Time to death	8.16 × 10−2	~ 0

Appendix 1D.

Estimates of H² for larvae of Cephaloleia placida reared on the native host Renealmia alpinia (RA), or the novel hosts Alpinia purpurata (AP) and Hedychium coronarium (HC).

Traits	Host plants
	RA (Native)	AP (Novel)	HC (Novel)
Length (mm)
Instar 1	2.51 × 10−1	~ 0	1.35 × 10−1
Instar 2	1.81 × 10−1	6.25 × 10−1	1.60 × 10−1
Pupa	--*	--*	--*
Adult	--*	--*	--*
Weight (g)
Pupa	--*	--*	--*
Adult	--*	--*	--*
Development time (d)
Larva-pupa	--*	--*	--*
Pupa-adult	--*	--*	--*
Survival (d)
Time to death	2.57 × 10−10	1.65 × 10−9	4.43 × 10−10

^*Trials not performed because larval mortality was high, precluding obtaining the required adults for longevity trials