The Glanville fritillary (M. cinxia L.) is widely distributed from Western Europe to Asia and North Africa. The species' habitat consists of dry meadows and grasslands with one or more of the larval host plant species, typically Plantago lanceolata and Veronica spicata in northern Europe. For a detailed description of the ecology and life history of the Glanville fritillary see Nieminen et al. (2004), and for demographic and eco-evolutionary dynamics see Hanski (1999, 2011).
We collected study material from four large regional populations in the Baltic Sea region, including the well-studied metapopulation in the Åland Islands (ÅL) in Finland, the Swedish east-coast population in Uppland (UP), and the populations on the large Swedish island of Öland (ÖL) and the large Estonian island of Saaremaa (SA) (Fig. 1A). The effective size of the ÅL metapopulation is roughly Ne ≈ 10,000, based on long-term census data (Hanski 2012b). Based on their areas, the other regions have comparable large regional populations. ÅL and UP have highly FL, with most habitat patches smaller than one hectare (in ÅL the average size is 0.17 ha; Nieminen et al. 2004). The bedrock in these regions is made up of volcanic rocks formed more than 1700 million years ago. (Katrantsiotis 2013). Following land up-lift since the last glacial period, the landscape is a small-scale mosaic of granite outcrops and intervening areas with sedimentary deposits supporting forests and presently small cultivated areas. In ÅL, some loss of seminatural meadows may have occurred in the past 100 years (Ojanen et al. 2013), but in any case, the landscapes in ÅL and UP have always been highly fragmented. In contrast, the habitat in ÖL and SA occurs in large continuous calcareous grasslands (alvars), often extending across several hundred hectares. In SA, the area of alvar grasslands has declined, with average area of individual (contiguous) alvars decreasing from 3.6 km2 in the 1930s to 20 ha in 2000 (Helm et al. 2006), which nonetheless is still two orders of magnitude greater than the average meadow size in ÅL. There is no detailed mapping of habitat patches at large spatial scale in UP and ÖL, but field observations indicate that landscape structure in these regions is very similar to that in ÅL and SA, respectively, and hence, the contrast between fragmented and CL is very clear.
Figure 1. (A) Map of the Baltic Sea region with the populations of the Glanville fritillary included in the present study: Åland Islands (ÅL), Uppland coastal region (UP), Öland (ÖL), and Saaremaa (SA). (B) Average monthly temperatures between 1992 and 2001 in ÅL (Blue), UP (Red), and SA (Green). The dashed lines give the average monthly temperatures recorded in ÅL.
Download figure to PowerPoint
According to monthly average temperature records for 1992–2001, ÅL, ÖL and SA have similar climatic conditions, with average temperatures approaching 20°C during summer and dropping below zero in winter (Fig. 1B). Note that the average temperatures for ÖL and SA are well within the range of yearly variation in ÅL during the 10-year period (Fig. 1B). Nonetheless, some slight differences in climatic conditions may be present due to the slightly more northern latitude of ÅL and UP (~60°N) than SA (~58°N) and ÖL (~57°N). The larval host plants P. lanceolata and V. spicata are present in all study areas. All four regional populations of the Glanville fritillary used in this study belong to the eastern clade (Wahlberg and Saccheri 2007) based on genome-wide SNP data (P. Somervuo, J. Kvist, S. Ikonen, P. Auvinen, L. Paulin, P. Koskinen, L. Holm, M. Taipale, A. Duplouy, A. Ruokolainen, S. Saarnio, J. Sirén, J. Kohonen, J. Corander, M. Frilander, V. Ahola, I. Hanski, unpubl. data) with some admixture with the central European clade in ÖL (based on COI haplotypes; Genbank #KC465909-KC465917).
The Glanville fritillary lays clutches of 150–200 eggs, and larvae remain gregarious during most of their development (Nieminen et al. 2004). Material for the present experiments was collected from the field as prediapause larvae in two occasions, in September 2006 (G0-2006) and in September 2009 (G0-2009) (“G0” referring to generation zero; Table 1). The larval families that were sampled were widely distributed across each region, roughly within area of 10 by 10 km, and hence, it is fair to assume that the larval families were unrelated to each other and representative of the respective regional populations. The sampling included 34 families in 2006 (9 ÅL, 12 SA and 13 UP) and 200 families in 2009 (50 in each population). The G0-2006 prediapause individuals were reared in family groups, while the G0-2009 material was reared in four mixed groups of 50 unrelated individuals, each originating from a different larval family. We intended to use only one such mixed group for the cage experiment (below), but as some mortality occurred during the larval period, we supplemented the material with some extra individuals from the other groups. We later estimated the relatedness of the experimental individuals in the G0-2009 material based on their genetic similarity (SNPs, data not shown). We found potential sibling relationships between eight pairs of individuals from ÅL, six pairs and one triplet from ÖL, 15 pairs and one triplet from SA and ten pairs from UP, leaving a total of 27, 27, 42, and 32 independent families, respectively.
Table 1. Sample sizes in the different study materials.
|Landscape||Fragmented (FL)||Continuous (CL)|
|Population||Åland (ÅL)||Uppland (UP)||Öland (ÖL)||Saaremaa (SA)|
|Sex||♂||♀||♂||♀||♂||♀||♂|| ♀ |
|G0-2006 (larval stage)||16||15||18||29||–||–||22||15|
|G0-2009 (larval and adult stages)||21||14||26||18||22||17||27||31|
|Flight metabolic rate||12||9||10||8||4||4||13||11|
Larval and pupal development
The field-collected larvae over-wintered in the laboratory in incubators in the constant temperature of 3°C. Following the end of diapause in April, the larvae were reared indoors under common garden conditions: 12:12 L/D and 25/15°C day/night temperature, respectively. Larvae were fed on excess of cut leaves of P. lanceolata. G0-2006 postdiapause larvae were first reared in small family groups and individually from the seventh (final) larval instar onward in small containers (4 cm diameter, 7 cm height). G0-2009 postdiapause larvae were reared individually from the fifth instar (end of diapause) onward. Laboratory-reared offspring of the G0-2009 butterflies, denoted as G1-2009 (the first laboratory generation), were reared in small family groups under the same conditions as the parental generation.
The G0-2006 larvae were weighed individually in the beginning of the seventh instar. For the G0-2009 material, we recorded the dates of the fifth, sixth, and seventh instar molts as well as the weight of each caterpillar at the beginning of each instar. We calculated the time of postdiapause development as the number of days between the fifth instar molt and pupation. Similarly, the development time of each postdiapause instar is the number of days between two consecutive molts. We recorded the pupal weight and pupal development time. The relative larval growth rate (RGR) was calculated as the difference between the pupal weight and the fifth instar larval weight, divided by the fifth instar weight and by the development time.
Outdoor cage experiment
Butterflies were marked individually by writing a number on the underside of the hindwing. Marking took place on the day following eclosion, thus ensuring that all butterflies had fully extended and dry wings prior to handling. Due to a problem with the timing of larval development, the experiment on the adults of the G0-2006 material largely failed; hence, we only analyzed larval traits for this material. Table 1 summarizes the sample size for the different experiments.
G0-2009 butterflies were released 24 h after their eclosion into a large outdoor cage in June 2010 (32 × 26 × 3 m, Fig. 2). The cage covered a dry meadow rich in flowering plants and closely resembling the natural habitat of the Glanville fritillary (Hanski et al. 2006; Saastamoinen 2008). The central part of the cage had 200 potted host plants, 100 individuals of P. lanceola and V. spicata each, individually labeled and placed within a regular grid within an area of 3.9 × 9.5 m (Fig. 2). These plants were provided for females to oviposit on, while all other host plants had been removed from the cage. After the 10th sunny day of the experiment, each remaining female butterfly was placed singly into a small cage of 10 (radius) × 20 cm, topping a potted host plant, to allow the female to continue ovipositing. Remaining males were similarly collected and kept in small groups (<15 individuals) in cages of 40 (radius) × 50 cm. The butterflies were kept in the small cages for the rest of their life to allow the measurement of remaining ovipositions and longevity, while minimizing the loss of dead butterflies that are difficult to recover from the outdoor cage. The dead butterflies were sampled for DNA for further phenotype–genotype association studies (not reported here). Four assistants worked continuously in the cage during the butterflies' active daily period, from 8 am to 6 pm, to collect data.
Figure 2. Schematic representation of the large outdoor population cage, with the transect surveyed five times a day (dashed line), the potted host plants in the central part of the cage (gray areas), and the grid division of the cage. (A) An inside view of the cage with an assistant monitoring the 200 potted host plants (© A. Jussila). (B) Close-up of the Glanville fritillary butterfly (Melitaea cinxia) (© T. Delahaye).
Download figure to PowerPoint
The cage was surveyed for marked butterflies five times a day (Fig. 2) every second hour, starting at 9, 11, 13, 15, and 17 h, weather permitting. We recorded the position in the cage and the behavior of each butterfly encountered as follow:
- Basking: butterfly steady on the ground, on a plant or on the cage wall, wings open.
- Resting: butterfly steady on the ground, on a plant or on the cage wall, wings closed.
- No movement: butterfly resting or basking.
- Flying: butterfly flying.
- Mating: butterfly paired with a sexual partner.
- Laying: female ovipositing on a host plant.
Mating pairs were additionally searched continuously throughout the cage during the experiment. For each mating pair encountered, the identity of both the male and the female, their position in the cage as well as the time of the day and the ambient air temperature were recorded. The male mating success of butterflies originating from the different populations was compared with a null hypothesis based on random mating given the total numbers of females and males from each population present in the cage in each day (empirical data compared with 10,000 randomizations).
For each butterfly, we regressed the total number of grid cells in the cage (Fig. 2) in which the butterfly was observed against the respective total number of observations during the transect surveys. As the age may influence butterflies' mobility, we measured mobility over two periods, first over the first 3 days of life, and second over the butterflies' entire life span. The residual from the regression line was used as a measure of mobility for each butterfly (Hanski et al. 2006). The distance covered by a butterfly was calculated as the cumulative distance between all observations. We regressed the distance covered by the butterfly against its life span and used the residual from the regression line as a measure of life-time independent movement distance in the cage.
The life span of each individual was defined as the time between eclosion and natural death. If the death was not recorded and the butterfly's body was lost, the date of the last observation alive in the cage was used as the date of death. This method gives an accurate result, as we conducted five surveys per day and the probability of detecting a butterfly that was alive in each survey was ~0.5.
The 200 potted host plants were constantly monitored between 8 am and 6 pm to record all ovipositions. For each oviposition, we recorded the identity of the female, the plant species, the time of the day, and the temperature at the start of the oviposition. The positions of individual plant pots in the grid of plants were randomized in each evening. After the oviposition, the leaf with the egg clutch was detached, placed on a petri dish, and brought to the laboratory for rearing in an incubator. After 3 days, eggs were counted to determine the clutch size. Lifetime-corrected number of eggs for each female was calculated as the total number of eggs laid by the female divided by its life span. The oviposition rate was defined as the total number of ovipositions (egg clutches) divided by the life span. The preference of a female to oviposit on P. lanceolata was calculated as the fraction of clutches laid on this host plant species.
The G1-2009 butterflies, the laboratory-reared offspring of the G0-2009 butterflies, were used for three laboratory experiments described below:
The G1-2009 females from each population were mated in small cages with a male from the same population but from a different family, thus avoiding outbreeding and inbreeding. Mated females were placed individually in small cages with a potted host plant (V. spicata) on which they oviposited. Each egg clutch was collected singly onto a petri dish. Clutches were weighed, and the eggs were counted 3 days after oviposition. The average egg weight was calculated by dividing the weight of the clutch by the number of eggs.
Ten random G1-2009 butterflies (five males and five females) from each family were frozen on the day after eclosion. After dissection, abdomen and thorax (head discarded) were dried for 24 h at 60°C and weighed. Wings were scanned and the right forewing areas were measured as described by Mattila et al. (2012). If the right forewing was damaged, the left forewing was used instead. Thorax-wing load was calculated by dividing the thorax dry weight by the wing area (mg/mm2), and total wing load by dividing thorax plus abdomen dry weights by the large wing area.
The flight metabolic rate (FMR) of four G1-2009 adults (two males and two females) from each family was measured as described by Niitepõld et al. (2009). In brief, a butterfly was placed into a hermetic chamber through which CO2-free air was pumped and was forced to fly continuously for 10 min. The CO2 concentration inside the chamber was recorded for more than 12 min, including the 10 min of forced flight and two mins of inactivity prior to the forced flight. Four measures were calculated: (1) The highest level of CO2 produced during the experiment (CO2 peak; mL/h); (2) the total volume of CO2 (mL) produced during the experiment; (3) the total volume of CO2 (mL) produced during the last 5 min of the experiment (reflecting flight endurance); and (4) the average resting metabolic rate (RMR; mL/h), calculated as the rate of CO2 emitted during the 60 sec prior to the start of the forced flight period. Temperature in the metabolic chamber was regulated at 30°C (Niitepõld et al. 2009).
Many of the above traits are correlated with each other. We nonetheless present results on a large number of traits (Appendix S1) as they may be of value for comparisons with other species and studies. To aid interpretation of results on many correlated traits, we have also performed multivariate analyses of several sets of traits (Appendix S2).
Statistical analyses were carried out using R64 (R Development Core Team 2009). All data were tested for normality and transformed prior analysis when appropriate. We used log transformation to normalize data on the age at first oviposition, the age at first mating, and the first oviposition event, and ArcSin transformation on proportions and probabilities. We analyzed larval, pupal, and adult traits (G0-2006 and G0-2009 samples) using ANOVAs with sex, population, and landscape type (FL or CL) as fixed factors, with population (ÅL, ÖL, SA, and UP) nested within landscape. While analyzing the larval developmental time and seventh instar weight, the year (2006 or 2009) was included as another fixed factor. We tested for the effects of pupal weight on lifetime-corrected number of eggs laid, of the clutch rank on clutch size, and of the clutch size and clutch rank on the length of ovipositing the clutch and on the period between two consecutive ovipositions. Longevity was analyzed with a general linear model with inverse link function. Traits related to egg weight, metabolic rates, wing size, and body parts (G1-2009) were tested using a mixed model, with family nested in population as a random factor and landscape type and sex as fixed factors. When appropriate, we calculated pairwise comparisons using the post hoc Tukey HSD test and corrected for multiple testing by applying the Bonferroni's adjustment (α = 0.05).
We used R64 to calculate four principal component analyses (PCA). In the first analysis, we included larval, pupal, and adult traits (including flight behavior, mobility and longevity) shared by males and females. The second PCA included only the larval traits for both sexes, while the third and the fourth PCAs included only female traits, either all traits or adult traits only, respectively. We analyzed the principal components with ANOVAs, with landscape type and population nested within landscape type as explanatory factors, and sex in the PCA1 and PCA2.