Subject terms: Community ecology, Conservation biology, Evolutionary ecology

Butterfly collection and rearing

Butterflies used in this study were second-generation captives originating from the last extant population ofE. e. taylori in the South Puget Sound region, on an active artillery impact area at Joint Base Lewis-McChord, Washington, USA. At this site, butterflies oviposit and larvae feed primarily onPlantago, but occasionally onC. hispida as well. Butterflies were collected from the field in 2016 and allowed to oviposit in the lab; the eggs were then deployed for a field study³². In late winter 2017, we collected surviving larvae from the field study after they emerged from diapause. These larvae fed onPlantago during early instars in the field, and we continued to feed them this species until pupation. In the lab, larvae were reared in containers with other members of their sibling group. After eclosion, male and female butterflies were separated, and females (one per sibling group) were mated to non-sibling males using methods described by Barclay et al.³⁶. They were fed a 3:1 water:honey mixture daily. Butterflies used in this study were collected and handled in accordance with a USFWS permit.

Objective 1: oviposition choice trials

We tested oviposition preference of 29 matedE. e. taylori females (matrilines) in 2017. We created three-way choice trials by enclosing females with mesh screening on 2.5 L pots with all three host plant species growing in each pot. Mesh was tented with bamboo stakes and held to the pot with a rubber band. The host plants were grown from seed in greenhouses at the University of Washington Center for Urban Horticulture the previous year, then stored outdoors over winter and transplanted so all three species shared a pot (n = 42 pots) in spring 2017. The growing medium in each pot was covered in a layer of pea gravel. At the beginning of each choice trial, a mated female butterfly was placed on a cotton swab soaked in dilute honey solution and introduced to a screened-in pot while it was feeding, with the swab placed upright between the three plants. Pots were placed under full spectrum lighting in the lab, or when possible, in direct sunlight outdoors, until either one day had elapsed or we observed that the butterfly had laid eggs. At this point the butterfly was fed and introduced to a new enclosure using the same methods. After each choice trial, we inspected all leaves of all plants in the pot and removed leaves where eggs had been laid. Pots were re-cycled for trials with other individuals until ~ 20% of leaves had been removed from a plant. In all, 122 oviposition choice trials were made (mean = 4.2 choices per individual, range = 1–8). We quantified preference as the number of eggs laid. Checkerspots oviposit in clutches of varying size and preference can also be quantified in terms of number of egg clusters laid rather than individual eggs. We focus on number of eggs laid because we suspected the number of eggs per cluster would be confounded with plant species due to differences in architecture (Castilleja have smaller leaves which could lead butterflies to lay fewer eggs per cluster and more clusters overall). However, we also report the proportion of clusters laid on each species and results were qualitatively very similar (see “Results”).

Eggs were stored in 60 mL plastic containers lined with paper towel and with perforated lids until they neared hatching (indicated by darkening). At this time, clusters were separated into smaller groups with a fine-tipped paintbrush so they could be allocated to either neonate choice trials (Objective 2) or no-choice feeding trials (Objectives 3 and 4).

We analyzed oviposition preference at both the population and individual level. For the population-level assessment, we assigned each butterfly to the host species it laid the most eggs on, and used a chi-square test to evaluate whether the number of individuals preferring each species differed. For the individual-level assessment, we began by visually comparing the data for overall oviposition patterns among matrilines and then tested if preference among hosts changed over time using binomial generalized linear mixed models (GLMMs) with each oviposition trial as a replicate (n = 122). We included random effects for the matriline and for the pot used in the trial since individuals made multiple choices and pots were re-cycled. We tested each plant species separately. For each species, the response was the number of eggs laid on that species compared to the other species. We carried out these and all other statistical analyses in R version 3.6.2³⁷, and used thelme4³⁸ package for all mixed models.

Objective 2: neonate choice trials

Twenty-one of the 29 butterflies produced enough viable eggs to be used in neonate choice trials (3 sets of trials x minimum of 3 eggs per trial = 9 eggs required). For each matriline, we placed groups of 3–6 darkened eggs in choice trial arenas (3 × 5 cm plastic cells). We used groups of eggs rather than individual eggs because larvae are gregarious during early instars. We cut 6 mm diameter leaf discs from each plant species (field collected, see below) using an office hole puncher and arranged the discs around each cluster of eggs so that newly hatched larvae had equal access to all discs. Discs and eggs were placed on beds of moistened paper towel to keep them from desiccating. We rotated the spatial arrangement of leaf discs systematically to avoid spatial or directional effects.

We conducted three sets of choice trials (n = 21 for each). In one set of trials, neonates chose between leaves of all three species; in the other two they chose between leaves and bracts from the sameC. hispida orC. levisecta stem. After hatching, larvae fed for ~ 24 h or until we observed that they consumed a whole leaf disc, at which point we ended the trial. We photographed each leaf disc and assessed the area eaten using ImageJ³⁹.

We tested whether neonate preference, expressed as the proportion of total leaf area consumed, varied among host species using a GLMM with a binomial distribution and matriline as a random effect. This model had a singular fit so we refit it as a GLM with matriline included as the first fixed effect. Finally, we tested whether neonates ate more tissue from lower leaves or from bracts. These tests were conducted separately for eachCastilleja species, using paired t-tests.

Objective 3: larval growth

Twenty matrilines produced enough viable eggs to be used in no-choice feeding trials (5 treatments x minimum of 4 larvae per treatment = 20 larvae required for a matriline to appear in all treatments). We assigned freshly-hatched sibling neonates from each matriline to five feeding treatments. Larvae were fed ad libitum diets ofP. lanceolata leaves, C. hispida leaves,C. hispida bracts,C. levisecta leaves, orC. levisecta bracts. We cutCastilleja stems in half so members of the same matriline ate the sameCastilleja stems but were restricted to either leaf or bract. Three of the 20 matrilines produced enough larvae to be included in only three of the five treatments, so for these we omitted the treatments withCastilleja leaves because in the field we generally observe larvae eating bracts. Numbers of individual larvae per replicate varied from 4 to 10 among matrilines depending on availability, but group sizes were equal among treatments within a matriline. Larvae in each treatment were raised in either 3 × 5 cm rectangular containers or 60 ml cups until second instar (container treatment was consistent within matrilines). They were transferred to 700 ml rectangular containers at the second instar and remained in these containers until they reached diapause, which usually occurs at the end of the fourth instar. Plant materials were replaced daily, and containers were cleaned every 1–2 days. Plants were field-collected from Glacial Heritage Preserve (46.87° N, 123.04°) weekly, with cut stems placed in water vials to prevent desiccation. Cut plants were stored under moist paper towels in a plastic cooler within a walk-in refrigerator (4 °C). Larvae were weighed to the nearest 0.1 mg at the third instar and again after reaching diapause.

Effects of diet on larval mass at third instar and again at diapause were tested using LMMs with matriline as a random effect. Significant effects of diet were followed by pairwise Tukey contrasts with theemmeans⁴⁰ package.

Objective 4: iridoid glycoside sequestration by larvae

We used gas chromatography to measure iridoid glycoside sequestration at diapause^41,42. We used 1–2 larvae per treatment per matriline for these measurements. Caterpillars were frozen, then ground whole and extracted in 95% methanol for 24 h. The solid material was filtered out and methanol evaporated. After adding the internal standard, phenyl-β-D-glycopyranoside (PBG) at 0.500 mg/mL, each sample was partitioned with ether (3 times) to remove hydrophobic compounds. The ether layer was removed and the water layer (which contains the iridoid glycosides) was evaporated. The residue was suspended in 1.0 mL methanol, and a 100 µL aliquot removed for analysis. The methanol was evaporated and the remaining residue derivatized using Tri-Syl-Z (Thermo-Fisher Chemical Company) in pyridine before injection into an Agilent 7890A gas chromatograph equipped with a DB-1 column (30 m, 0.320 mm, 0.25 µm particle size) and using flame ionization detection. Amounts of four individual IG compounds, aucubin, catalpol, macfadienoside and methyl shanziside, were quantified using ChemStation B-03-01 software. Six samples were excluded due to labelling or processing issues.

Iridoid glycoside uptake was expressed both as the concentration and the total amount of iridoid glycosides in diapausing larvae. Each measure of uptake was tested using a LMM with matriline as a random effect. Significant effects of diet were followed by pairwise Tukey contrasts. Since caterpillars can contain multiple iridoid glycosides, we also tested if the composition of these compounds differed among diets using permutational multivariate ANOVA (PERMANOVA) in the R packagevegan⁴³. Values were relativized by the total IG amount in each larva so they expressed each compound as a proportion of the total. We used a Euclidean distance measure with 10,000 permutations.

Objective 5: correlation of maternal preference, neonate preference, and larval growth

We tested if maternal preference correlated to that of their neonate offspring, and if larval mass gain on a given species could be predicted by maternal or sibling preference for that species (i.e., preference-performance relationships). In each case, we ran a separate linear model for each host plant species, first with maternal preference as a predictor and neonate preference of that matriline as a response, then with either maternal preference or neonate preference as a predictor and mean mass gain of that matriline as a response. We used mass data from larvae that ate bracts ofCastilleja spp., rather than leaves, because this is what larvae generally feed on in the field.

Objective 1: oviposition choice trials

Population-level oviposition preference differed among the three host species (χ²_[2] = 11.7, p < 0.01). Of the 5417 eggs laid, 44% were placed onPlantago, 31% were onC. hispida, and 24% onC. levisecta. However, individual preference was quite variable. Some butterflies oviposited mostly onP. lanceolata, others mostly onC. hispida, and some used both or all three species (Fig. 2). Thirteen individuals allocated more than half their eggs toPlantago; eight allocated more than half toC. hispida, and none allocated more than half toC. levisecta. The remaining eight individuals apportioned their eggs more evenly among the three species. Eggs were laid in 312 clusters; 40% of these were onPlantago, 35% onC. hispida, and 25% onC. levisecta (this calculation omits rare instances when eggs were laid singly). Of the 29 butterflies, 16 laid their first cluster of eggs onPlantago, 10 onC. hispida, and 3 onC. levisecta (if they laid eggs on multiple species in the first trial, we considered the species that received more eggs to have been chosen).

Oviposition preference among the three plant species changed over time (Table1, Fig. 3). Younger butterflies favoredP. lanceolata, but preference for this species declined with time. Preference forC. hispida also declined with time, though not as steeply, while preference forC. levisecta increased.

Objective 2: neonate choice trials

We found no evidence of population-level neonate preference among the three host species (deviance = − 0.70, p = 0.70), although there was a non-significant trend for larvae to feed less onC. hispida than the other species. The average percent of total leaf area eaten by larval groups was 26%C. hispida, 37%C. levisecta, and 37%P. lanceolata. However, sibling groups from the various matrilines made diverse choices. Two groups fed only onC. hispida, two only onC. levisecta, and two only onPlantago; the remaining groups fed on two or all three species (Fig. 4). Thus, the population-level maternal oviposition preference towardPlantago and away fromC. levisecta was not expressed by neonates.

Neonates strongly preferred bract over leaf tissue of bothCastilleja species. When choosing among tissue types ofC. hispida, neonates on average fed 78% on bract tissue and 22% on leaf (t_[20] =  − 4.15, p < 0.01). When choosing among tissue types ofC. levisecta, they fed 86% on bracts and 14% on leaves (t_[20] =  − 7.83, p < 0.01) (Fig. 4).

Objective 3: larval growth

Mass gain differed strongly among feeding treatments both during third instar (F = 81.99, p < 0.01) and upon entering diapause (F = 54.09, p < 0.01). During third instar, larvae feeding onC. levisecta bracts were largest, followed by those feeding onC. hispida bracts and then those feeding onPlantago, while larvae feeding on leaves of eitherCastilleja species were smaller (Fig. 5A). At diapause, larvae that fed onPlantago or bracts of eitherCastilleja species were of similar mass and significantly larger than those that ate onlyCastilleja leaves (Fig. 5B).

Objective 4: iridoid glycoside uptake

The total amounts of iridoid glycosides sequestered by larvae depended on their food plant, regardless of whether IGs were considered by concentration (F_[4,71] = 46.38, p < 0.01) or by mass (F_[4,71] = 32.76, p < 0.01). Uptake was highest in larvae that atePlantago, intermediate in those that ateC. hispida leaves, and lowest in larvae fed other diets (Fig. 5C,D). We detected aucubin in every larva, along with up to three other iridoid glycosides depending on diet (Table2, Fig. 6, PERMANOVA pseudo-F_[4,93] = 88.95, p < 0.01). Larvae that fed onP. lanceolata always contained catalpol. Larvae that fed onC. hispida sometimes contained catalpol but were more likely to contain macfadienoside and/or methyl shanziside. Larvae that fed onC. levisecta usually contained methyl shanziside, sometimes contained macfadienoside, and did not contain catalpol.

There were strong differences in iridoid glycoside uptake among larvae that ate different tissues withinCastilleja. Most larvae that fed onC. hispida bracts contained only aucubin; a minority of these sequestered small amounts of catalpol and/or methyl shanziside, but none contained detectable amounts of macfadienoside. In contrast, those that ate leaves of the same species contained both aucubin and methyl shanziside, usually contained macfadienoside, and sometimes contained catalpol. Larvae that ateC. levisecta bracts always contained aucubin but usually also contained a small amount of methyl shanziside, occasionally contained macfadienoside, and never contained catalpol. Finally, those that ateC. levisecta leaves contained almost exclusively aucubin and methyl shanziside, with macfadienoside detected in only one individual and catalpol in none (Table2, Fig. 6).

Objective 5: correlation of maternal and neonate preference

Neonate preference for a given host did not correspond to the oviposition preference of their parent for that host (Table3). Similarly, larval growth on each host was not predicted by either their mother’s or their siblings’ preference for that host (Table4).

Causes and consequences of adopting a novel host

The ease with which checkerspots adoptPlantago and the number of taxa that have done so independently is remarkable^13–15. A recent meta-analysis found that butterflies and moths generally perform poorly on novel exotic hosts compared to ancestral ones⁴⁴, although there are certainly exceptions to this trend⁴⁵. Checkerspots seem to illustrate ecological fitting, as they are broadly pre-adapted toPlantago^11,46. Naïve larvae are often immediately able to develop on it, although in at least some cases adults initially reject it as an oviposition plant and preference for it must evolve¹³. A number of additional factors complement this close ecological fit and help explain why checkerspots repeatedly adoptPlantago and often perform well on it.Plantago can be very abundant in disturbed herbaceous habitats. In the case of Taylor’s checkerspot it is not only very easy to find but is also more phenologically available than native hosts because it senesces later³². Finally, the iridoid glycosides thatPlantago produces, aucubin and catalpol, are already present in bothCastilleja species investigated here, indicating that shifting fromCastilleja toPlantago does not require metabolic adjustments as might be needed if a herbivore encountered a novel host containing new IG compounds. These two compounds are also common across other checkerspot hosts such asChelone glabra⁴⁷, the ancestral host ofE. phaeton in Eastern North America. Many populations of this butterfly have also adoptedPlantago¹⁴.

Although successful adoption of novel host plants can result in loss of adaptation to ancestral ones^4,12, that was not the case here. The population we studied was almost entirely restricted to usingPlantago, yet our results show that it has retained a breadth of oviposition and neonate feeding preferences. These findings contrast with outcomes for anotherE. editha population (subspeciesmonoensis) located about 900 km south of the one we studied, in Nevada, USA^12,13. This population historically fed onCollinsia parviflora but switched to feeding entirely onPlantago over the course of three decades and lost the preference for its traditional host. The lost adaptation was this population’s downfall; it went extinct whenPlantago became unavailable due to changes in agricultural management, despiteCollinsia remaining available for oviposition¹². There could be several plausible reasons why our study population did not lose adaptation to ancestral hosts. First, we do not know when our study population incorporatedPlantago into its diet. This population could be derived partly or wholly from ancestors that adoptedPlantago more than a century ago³³, or the switch could be very recent (there could also be a history of repeated switches). If the switch occurred recently, loss of adaptation toCastilleja could still be in progress. Second, when the now-extinct Nevada population switched fromCollinsia toPlantago it did so in the face of an evolutionary tradeoff in which butterflies experienced higher fecundity but slower development onPlantago, making for a poorer phenological fit withCollinsia¹². We have not found evidence that Taylor’s checkerspot faces this sort of trade-off, so addingPlantago to its diet may not require losing adaptations toCastilleja. There is, however, some evidence thatPlantago exerts selective pressure for larger clutch sizes, as larger sibling groups are more likely to survive to later instars onPlantago but not onCastilleja³².

Variation in oviposition preference

Oviposition preference varied considerably among individual butterflies. While some strongly preferred eitherPlantago orC. hispida, many spread their effort more evenly among all three species (i.e., the 8 individuals in the center section of Fig. 2). Oviposition preference also varied within individual butterflies over time, as the proportion of eggs laid onPlantago andC. hispida decreased and the proportion laid onC. levisecta increased. There are multiple potential explanations for this. First, ovipositingE. editha become less selective over time⁴⁸. However, if this was the only mechanism we would expect preference among the three species to become equal over time, and it did not. Second, butterflies may have innately avoidedC. levisecta but learned and gradually accepted it over time—although individuals of otherE. editha subspecies do not appear to learn regarding oviposition⁴⁹. Finally, the suitability and attractiveness of plants may have shifted over the course of the study, as we moved plants daily between full-spectrum artificial light and sunlight during the two-week oviposition period. IGs, nitrogen, and water content ofPlantago can change over short time intervals⁵⁰, and relative attractiveness could also have changed over time as eggs were laid and leaves were removed, possibly inducing chemical responses in the plants. In light of these possibilities, we interpret oviposition patterns early in the study as being more indicative of preference in general. Field studies would be required to track oviposition over time and assess whether oviposition preference also varies in that context.

Our findings regarding oviposition contrast with previous work in which the same population appeared to prefer eitherCastilleja species overPlantago³⁴. This difference could be attributable to variation in host plant characteristics, to the smaller number of lineages used in the previous study (5), or to differences in methodology, as the earlier study used abdomen-curling behavior as a signal of host plant acceptance⁴⁸ rather than quantifying oviposition per se.

Preference and performance

We did not find evidence for links between adult oviposition preference and neonate feeding choices, and neither of these forms of preference predicted larval performance. We would have interpreted a significant correlation between preference for, and performance on, a given host species as evidence that subsets of the population had evolved different host plant specializations, but this possibility was not supported. However, the ordering of oviposition preference (Plantago > C. hispida > C. levisecta)did match the frequency of interaction Taylor’s checkerspot from this region probably had with these species over the last several decades²⁹, and their survival rates on them in the field³². In the field, bothCastilleja species senesce beforePlantago and early-instar larvae die in especially high numbers onC. levisecta as it begins to senesce in May and June³². In the present study we only fed non-senescent plants to larvae; the fact that they developed similarly on non-senescentC. levisecta as on other host species is more evidence that early senescence is the driving factor makingC. levisecta less suitable in the field.

Neonate choices were variable, but neonates did not express strong population-level preference among host species, and when they selected a certain species this did not correlate to better performance on that species by their siblings. However, they expressed strong preference between tissue types inCastilleja, choosing bracts over leaves. This choice appears adaptive since eating bracts resulted in higher mass gain (Fig. 5A,B). Neonates must choose which tissues to eat within a plant but are rarely required to choose among plant species (they typically remain on the natal plant until instar 2 or 3); this aligns with our finding that they strongly preferred the higher-quality tissue type withinCastilleja but did not have strong preferences among species.

In the field, eggs are typically laid on leaves toward the bottom of the stem, and larvae move to the plant apex by the second instar to feed on bract and flower tissues (although they still eat some lower leaves as well; Haan pers. obs.). It is counterintuitive that adults lay eggs on a tissue type that larvae avoid and that is nutritionally inferior, but they could be choosing oviposition sites to hide eggs from enemies, based on within-plant differences in secondary chemistry²² or based on abiotic conditions required for egg development.

Larval growth varied among diets and over time.Castilleja bracts gave an early advantage to larvae, but by diapause, larvae that fed onPlantago had caught up and were similar in size (Fig. 5A,B). This may have occurred becauseCastilleja bracts are soft and contain fewer trichomes (Haan pers. obs.), making them easy for early instars to eat quickly. We also noticed that larvae in the first and second instars ate entireCastilleja bracts but skeletonized the thickerPlantago leaves. However, during the third and fourth instars larvae were able to eat entirePlantago leaves rather than skeletonizing them and quickly made up for lost growth.

Iridoid glycoside uptake

In general, overall IG concentrations in larvae were similar to those in previous studies of related taxa at similar ontogenetic stages^41,51,52. Concentrations were highest when larvae atePlantago orC. hispida leaves and lower if they ateC. hispida bracts or either tissue type ofC. levisecta. This is consistent with data from individuals that fed on these species in a field setting: those that ateC. levisecta sequestered substantially less than those that ate the other species⁵³. Since sequestered IGs are thought to deter natural enemies^54,55, the reduced ability to sequester fromC. levisecta could increase predation risk. Our results also indicate that intra-specific and intra-individual variation withinC. hispida is important with respect to larval performance: larvae that fed only onC. hispida bracts grew larger but had lower IG concentrations than those that fed only onC. hispida leaves. In a field setting, larvae would be mobile and able to forage across both tissue types, perhaps optimizing larval performance and sequestration.

Larvae carried distinct chemical fingerprints depending on their diet. If they atePlantago, they contained aucubin and catalpol (the two IGs found inPlantago), while if they ateCastilleja they contained up to two additional compounds, macfadienoside and methyl shanziside (found in bothCastilleja species). There were striking differences between larvae that ate differentC. hispida tissues; they sequestered large amounts of macfadienoside when they ate leaves, but none if they ate bracts (Fig. 6). IGs can be spatially segregated amongCastilleja tissues; for example, withinCastilleja integra, macfadienoside is abundant throughout the plant but especially in leaves, methyl shanziside is much more common in leaves than in bracts or flowers, and catalpol dominates flowers³⁵.

In previous work we found that greenhouse-grownC. levisecta produced largely aucubin, andE. editha larvae that ate it contained both aucubin and catalpol⁵²; this contrasts with our findings here. These differences in plant chemistry are likely because plants in the previous study were grown in the greenhouse, whereas plants used in the present study were field collected. Greenhouse grown plants are often lower in secondary compounds than those grown outside or collected from the wild⁵⁶. It is also not known if the four compounds derived fromCastilleja have different costs or benefits to specialist herbivores. Previous work with another IG specialist, the buckeye,Junonia coenia (Nymphalidae), showed that aucubin and catalpol have positive synergistic effects on survival, growth, and sequestration rates but negative effects on immune response^57,58. The specific effects of macfadienoside and methyl shanziside on herbivores and their enemies have not been investigated. Given that amounts of these compounds in larvae differ sharply depending on host species and on tissue type eaten withinCastilleja spp., they could have important implications for herbivore immune function and/or interactions with higher trophic levels.

Implications for conservation efforts

Taylor’s checkerspot has been reduced to a very small number of populations that mostly rely onPlantago, and our data show that the population we studied prefers it and performs well on it. However, we also found that a preference for, and ability to develop on, other hosts continues to persist. While we did not directly assess genetic diversity, it seems that this relict population is well-suited for reintroductions to sites where host plant availability differs from the source site.

Our results cast further doubt on the suitability ofC. levisecta as a primary host plant for Taylor’s checkerspot in Washington lowlands. Although it was almost certainly an ancestral host, it is rare and may not have been used as frequently, at least during the last several decades²⁹. Adults will oviposit on it but tend to favor other species and, although larvae do not innately avoid it, they have lower survival rates on it in the field³² and sequester relatively lower concentrations of IGs from it. However, sinceC. levisecta does not appear to be especially attractive to ovipositing butterflies and larvae can successfully develop on it if needed, it could be acceptable as a secondary host plant and does not appear to pose undue risk to recovery efforts. Furthermore, given thatE. editha host plant affiliations evolve rapidly, populations that are reintroduced whereC. levisecta is the main available host might readily adapt to it.

Taylor’s checkerspot host affiliations are not static; the patterns we documented in this study are likely to change in the future both for this population and for the reintroduced ones that are derived from it. While there may be uncertainty about relative suitability of host resources, the diversity in preference and performance we documented here is probably an asset. Management for Taylor’s checkerspot should aim to maintain many populations using diverse host resources. Population-level differences in host specialization, if they occur, will help buffer against future changes and make recovery efforts more likely to succeed over the long term.

• 1.Strauss SY, Lau JA, Carroll SP. Evolutionary responses of natives to introduced species: what do introductions tell us about natural communities? Evolutionary responses of natives to introduced species. Ecol. Lett. 2006;9:357–374. doi: 10.1111/j.1461-0248.2005.00874.x. [DOI] [PubMed] [Google Scholar]
• 2.Smith DC. Heritable divergence of Rhagoletis pomonella host races by seasonal asynchrony. Nature. 1988;336:66–67. doi: 10.1038/336066a0. [DOI] [Google Scholar]
• 3.Filchak KE, Roethele JB, Feder JL. Natural selection and sympatric divergence in the apple maggot Rhagoletis pomonella. Nature. 2000;407:739–742. doi: 10.1038/35037578. [DOI] [PubMed] [Google Scholar]
• 4.Carroll, S. P., Dingle, H., Famula, T. R. & Fox, C. W. Genetic architecture of adaptive differentiation in evolving host races of the soapberry bug,Jadera haematoloma. inMicroevolution Rate, Pattern, Process (eds. Hendry, A. P. & Kinnison, M. T.) vol. 8 257–272 (Springer Netherlands, 2001). [PubMed]
• 5.Nice CC, Fordyce JA, Shapiro AM, Ffrench-Constant R. Lack of evidence for reproductive isolation among ecologically specialised lycaenid butterflies. Ecol. Entomol. 2002;27:702–712. doi: 10.1046/j.1365-2311.2002.00458.x. [DOI] [Google Scholar]
• 6.Graves SD, Shapiro AM. Exotics as host plants of the California butterfly fauna. 2003;110:413–433. [Google Scholar]
• 7.Thomas JA, Simcox DJ, Hovestadt T. Evidence based conservation of butterflies. J. Insect Conserv. 2011;15:241–258. doi: 10.1007/s10841-010-9341-z. [DOI] [Google Scholar]
• 8.Battin J. When good animals love bad habitats: Ecological traps and the conservation of animal populations. Conserv. Biol. 2004;18:1482–1491. doi: 10.1111/j.1523-1739.2004.00417.x. [DOI] [Google Scholar]
• 9.Casagrande, R.A. & Dacey, J. E. Monarch butterfly oviposition on swallow-worts (Vincetoxicum spp.).Environ. Entomol.36, 631–636 (2007). [DOI] [PubMed]
• 10.Davis SL, Cipollini D. Do mothers always know best? Oviposition mistakes and resulting larval failure of Pieris virginiensis on Alliaria petiolata, a novel, toxic host. Biol. Invasions. 2014;16:1941–1950. doi: 10.1007/s10530-013-0637-2. [DOI] [Google Scholar]
• 11.Janzen DH. On ecological fitting. Oikos. 1985;45:308. doi: 10.2307/3565565. [DOI] [Google Scholar]
• 12.Singer MC, Parmesan C. Lethal trap created by adaptive evolutionary response to an exotic resource. Nature. 2018;557:238–241. doi: 10.1038/s41586-018-0074-6. [DOI] [PubMed] [Google Scholar]
• 13.Thomas CD, et al. Incorporation of a European weed into the diet of a North American herbivore. Evolution. 1987;41:892–901. doi: 10.1111/j.1558-5646.1987.tb05862.x. [DOI] [PubMed] [Google Scholar]
• 14.Bowers MD, Stamp NE, Collinge SK. Early stage of host range expansion by a specialist herbivore Euphydryas phaeton. Ecology. 1992;73:526–536. doi: 10.2307/1940758. [DOI] [Google Scholar]
• 15.Severns PM, Breed GA. Behavioral consequences of exotic host plant adoption and the differing roles of male harassment on female movement in two checkerspot butterflies. Behav. Ecol. Sociobiol. 2014;68:805–814. doi: 10.1007/s00265-014-1693-z. [DOI] [Google Scholar]
• 16.United States Fish and Wildlife Service. Endangered and threatened wildlife and plants; proposed designation of critical habitat for the bay checkerspot butterfly (Euphydryas editha bayensis); proposed rule. (2000).
• 17.United States Fish and Wildlife Service. Endangered and threatened wildlife and plants; designation of critical habitat for the Quino checkerspot butterfly (Euphydryas editha quino). (2002).
• 18.United States Fish and Wildlife Service. ESA Proposed Listing Taylor’s Checkerspot.Fed. Regist.77, (2012).
• 19.Ehrlich, P. R. & Hanski, I.On the wings of checkerspots: a model system for population biology. Oxford University Press (2004).
• 20.Singer MC, Ng D, Thomas CD. Heritability of oviposition preference and its relationship to offspring performance within a single insect population. Evolution. 1988;42:977–985. doi: 10.1111/j.1558-5646.1988.tb02516.x. [DOI] [PubMed] [Google Scholar]
• 21.Singer MC, McBride CS. Multitrait, host-associated divergence among sets of butterfly populations: implications for reproductive isolation and ecological speciation. Evol. Int. J. Org. Evol. 2009;64:921–933. doi: 10.1111/j.1558-5646.2009.00866.x. [DOI] [PubMed] [Google Scholar]
• 22.Peñuelas J, Sardans J, Stefanescu C, Parella T, Filella I. Lonicera implexa leaves bearing naturally laid eggs of the specialist herbivore Euphydryas aurinia have dramatically greater concentrations of iridoid glycosides than other leaves. J. Chem. Ecol. 2006;32:1925–1933. doi: 10.1007/s10886-006-9118-8. [DOI] [PubMed] [Google Scholar]
• 23.Nieminen, M., Suomi, J., Nouhuys, S. V., Sauri, P. & Riekkola, M.-L. Effect of iridoid glycoside content on oviposition host plant choice and parasitism in a specialist herbivore.J. Chem. Ecol. 22 (2003). [DOI] [PubMed]
• 24.Bowers MD. Unpalatability as a defense strategy of Euphydryas phaeton (Lepidoptera: Nymphalidae) Evolution. 1980;34:586–600. doi: 10.1111/j.1558-5646.1980.tb04846.x. [DOI] [PubMed] [Google Scholar]
• 25.Bowers MD. Unpalatability as a defense strategy of western checkerspot butterflies (Euphydryas Scudder, Nymphalidae) Evolution. 1981;35:367–375. doi: 10.1111/j.1558-5646.1981.tb04895.x. [DOI] [PubMed] [Google Scholar]
• 26.Dobler S, Petschenka G, Pankoke H. Coping with toxic plant compounds–the insect’s perspective on iridoid glycosides and cardenolides. Phytochemistry. 2011;72:1593–1604. doi: 10.1016/j.phytochem.2011.04.015. [DOI] [PubMed] [Google Scholar]
• 27.Bowers MD, Stamp NE. Effects of plant age, genotype and herbivory on Plantago performance and chemistry. Ecology. 1993;74:1778–1791. doi: 10.2307/1939936. [DOI] [Google Scholar]
• 28.Dyer, L. A. & Deane Bowers, M. The importance of sequestered iridoid glycosides as a defense against an ant predator.J. Chem. Ecol.22, 1527–1539 (1996). [DOI] [PubMed]
• 29.Dunwiddie PW, et al. Intertwined fates: Opportunities and challenges in the linked recovery of two rare species. Nat. Areas J. 2016;36:207–215. doi: 10.3375/043.036.0214. [DOI] [Google Scholar]
• 30.Stinson, D. Washington State Status Report for the Mazama Pocket Gopher, Streaked Horned Lark, and Taylor’s Checkerspot. Washington Department of Fish and Wildlife (2005).
• 31.Cavers, P. B., Bassett, I. J. & Crompton, C. W. The biology of Canadian weeds 47.Plantago lanceolata L.Can. J. Plant Sci.60, 1269–1282 (1980).
• 32.Haan NL, Bakker JD, Dunwiddie PW, Linders MJ. Instar-specific effects of host plants on survival of endangered butterfly larvae. Ecol. Entomol. 2018;43:742–753. doi: 10.1111/een.12656. [DOI] [Google Scholar]
• 33.Danby WH. Food plant of Melitaea taylori Edw. Can. Entomol. 1890;22:121–122. doi: 10.4039/Ent22121-6. [DOI] [Google Scholar]
• 34.Buckingham DA, Linders M, Landa C, Mullen L, LeRoy C. Oviposition preference of endangered Taylor’s checkerspot butterflies (Euphydryas editha taylori) using native and non-native hosts. Northwest Sci. 2016;90:491–497. doi: 10.3955/046.090.0409. [DOI] [Google Scholar]
• 35.Mead EW, Stermitz FR. Content of iridoid glycosides in different parts of Castilleja. Phytochemistry. 1993;32:1155–1158. doi: 10.1016/S0031-9422(00)95082-9. [DOI] [Google Scholar]
• 36.Barclay E, Arnold M, Anderson MJ, Shepherdson D. Husbandry manual: Taylor’s checkerspot (Euphydryas editha taylori) Portland OR: Oregon Zoo; 2009. [Google Scholar]
• 37.R Core Team.R: a language and environment for statistical computing. R Foundation for Statistical Computing (2020).
• 38.Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models usinglme4.J. Stat. Softw.67, (2015).
• 39.Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods. 2012;9:671–675. doi: 10.1038/nmeth.2089. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 40.Lenth, R. V. Least-Squares Means: The R package lsmeans.J. Stat. Softw.69, (2016).
• 41.Bowers MD, Stamp NE. Effect of hostplant genotype and predators on iridoid glycoside content of pupae of a specialist insect herbivore, Junonia coenia (Nymphalidae) Biochem. Syst. 1997;25:571–580. doi: 10.1016/S0305-1978(97)00058-6. [DOI] [Google Scholar]
• 42.Bowers MD. Hostplant suitability and defensive chemistry of the Catalpa sphinx Ceratomia catalpae. J. Chem. Ecol. 2003;29:2359–2367. doi: 10.1023/A:1026234716785. [DOI] [PubMed] [Google Scholar]
• 43.Oksanen J, et al. Package ‘vegan’. Community Ecol. Package Version. 2013;2:1–295. [Google Scholar]
• 44.Yoon S, Read Q. Consequences of exotic host use: Impacts on Lepidoptera and a test of the ecological trap hypothesis. Oecologia. 2016;181:985–996. doi: 10.1007/s00442-016-3560-2. [DOI] [PubMed] [Google Scholar]
• 45.Cogni R. Resistance to plant invasion? A native specialist herbivore shows preference for and higher fitness on an introduced host. Biotropica. 2010;42:188–193. doi: 10.1111/j.1744-7429.2009.00570.x. [DOI] [Google Scholar]
• 46.Agosta SJ, Klemens JA. Ecological fitting by phenotypically flexible genotypes: implications for species associations, community assembly and evolution. Ecol. Lett. 2008;11:1123–1134. doi: 10.1111/j.1461-0248.2008.01237.x. [DOI] [PubMed] [Google Scholar]
• 47.Bowers MD, Boockvar K, Collinge SK. Iridoid glycosides of Chelone glabra (Scrophulariaceae) and their sequestration by larvae of a Sawfly, Tenthredo grandis (Tenthredinidae) J. Chem. Ecol. 1993;19:815–815. doi: 10.1007/BF00985011. [DOI] [PubMed] [Google Scholar]
• 48.Singer MC. Quantification of host preference by manipulation of oviposition behavior in the butterfly Euphydryas editha. Oecologia. 1982;52:224–229. doi: 10.1007/BF00363841. [DOI] [PubMed] [Google Scholar]
• 49.Parmesan C, Singer MC, Harris IAN. Absence of adaptive learning from the oviposition foraging behaviour of a checkerspot butterfly. Anim. Behav. 1995;50:161–175. doi: 10.1006/anbe.1995.0229. [DOI] [Google Scholar]
• 50.Quintero C, Lampert EC, Bowers MD. Time is of the essence: direct and indirect effects of plant ontogenetic trajectories on higher trophic levels. Ecology. 2014;95:2589–2602. doi: 10.1890/13-2249.1. [DOI] [Google Scholar]
• 51.Gardner DR, Stermitz FR. Host plant utilization and iridoid glycoside sequestration by Euphdryas anicia (Lepidoptera: Nymphalidae) J. Chem. Ecol. 1988;14:2147–2168. doi: 10.1007/BF01014022. [DOI] [PubMed] [Google Scholar]
• 52.Haan NL, Bakker JD, Bowers MD. Hemiparasites can transmit indirect effects from their host plants to herbivores. Ecology. 2018;99:399–410. doi: 10.1002/ecy.2087. [DOI] [PubMed] [Google Scholar]
• 53.Haan NL. Ecological interactions between Euphydryas editha larvae and their host plants. Seattle: University of Washington; 2017. [Google Scholar]
• 54.Bowers, M. D. Aposematic caterpillars: life-styles of the warningly colored and unpalatable, inCaterpillars: ecological and evolutionary constraints on foraging (eds. Stamp, N.S., and Casey, T.M.). Chapman & Hall (1993).
• 55.Theodoratus DH, Bowers MD. Effects of sequestered iridoid glycosides on prey choice of the prairie wolf spider Lycosa carolinensis. J. Chem. Ecol. 1999;25:283–295. doi: 10.1023/A:1020894729188. [DOI] [Google Scholar]
• 56.Cirak C, et al. Phenological changes in the chemical content of wild and greenhouse-grown Hypericum pruinatum: hypericins, hyperforins and phenolic acids. Res Rev J Bot. 2015;4:37–47. [Google Scholar]
• 57.Richards LA, et al. Synergistic effects of iridoid glycosides on the survival, development and immune response of a specialist caterpillar, Junonia coenia (Nymphalidae) J. Chem. Ecol. 2012;38:1276–1284. doi: 10.1007/s10886-012-0190-y. [DOI] [PubMed] [Google Scholar]
• 58.Smilanich AM, Dyer LA, Chambers JQ, Bowers MD. Immunological cost of chemical defence and the evolution of herbivore diet breadth. Ecol. Lett. 2009;12:612–621. doi: 10.1111/j.1461-0248.2009.01309.x. [DOI] [PubMed] [Google Scholar]
• 59.Hamilton, N.E. & Ferry, M. ggtern: Ternary diagrams using ggplot2.J. Stat. Softw., Code Snippets,87, 1–17 (2018).

Data Availability Statement

Data from this study were submitted to Dryad.https://doi.org/10.5061/dryad.612jm642h.