Group Size, and Egg and Larval Survival in the Social Butterfly Melitaea Cinxia

Ann. Zool. Fennici 54: 213–223 ISSN 0003-455X (print), ISSN 1797-2450 (online) Helsinki 15 May 2017 © Finnish Zoological and Botanical Publishing Board 2017

Ilkka Hanski: The legacy of a multifaceted ecologist

Group size, and egg and larval survival in the social butterfly Melitaea cinxia

Mikko Kuussaari1,* & Michael C. Singer2

1) Finnish Environment Institute (SYKE), Natural Environment Centre, P.O. Box 140, FI-00251 Helsinki, Finland (*corresponding author’s email: [email protected]) 2) Biological Sciences, Plymouth University, Drake Circus, Plymouth PL4 8AA, U.K.

Received 1 Feb. 2017, final version received 10 Apr. 2017, accepted 10 Apr. 2017

Kuussaari, M. & Singer, M. C. 2017: Group size, and egg and larval survival in the social butterfly Melitaea cinxia. — Ann. Zool. Fennici 54: 213–223.

Gregarious, social caterpillars have stimulated research because group size may affect survival, growth rate, thermoregulation, and interactions with other species, yet group size is often variable both within and among populations. We used a combination of observations and experiments to study the importance of group size for egg and larval survival in the Glanville fritillary butterfly, Melitaea cinxia, which lives in groups from egg hatching until the last larval instar. Both experimental manipulation of egg clutches placed in the field and observations of naturally occurring groups showed that survival increased with increasing group size. This pattern was present independently during all four developmental stages studied: eggs, prediapause larvae, diapausing larvae and post-diapause larvae. However, it was significant only during two stages: pre-diapause (in one year only) and diapausing larvae (in all years). Large group size increased survival of entire larval groups as well as that of individual larvae within surviving groups. These results may explain why cluster size is large and why adults oviposit infrequently. Large cluster size, coupled with correlated survival of group members, in turn helps to explain the unstable local dynamics and short average persistence time of local M. cinxia populations.

Introduction satiation in turtles (Santos et al. 2016). Lepi- dopterans are no exception: even though some Aggregation of potentially independent indi- 90%–95% of species lay their eggs singly (Stamp viduals into social groups occurs in almost all 1980), egg clustering behaviour, followed by groups of animals (Hamilton 1971). Its adaptive gregarious behaviour of young larvae, occurs in functions are diverse: for example, collective a taxonomically diverse array of species, serves decision-making in slime moulds (Reid & Latty a diversity of functions and has evolved inde- 2016), efficiency of prey capture in lions (Caraco pendently on at least 15 occasions in butterflies & Wolf 1983), enhanced vigilance of vervet (Sillén-Tullberg 1988, Costa & Pierce 1997). In monkeys (Josephs et al. 2016), aposematic dis- most species, larval aggregation is restricted to play in Tropidothorax bugs (Gamberale & Sillen- the first few larval instars (Zalucki et al. 2002) Tullberg 1998) reduced risk of predator ambush but in some species the larvae remain in groups in seals (DeVos & O’Riain 2010), and predator almost until pupation (Fitzgerald 1993). 214 Kuussaari & Singer • ANN. ZOOL. FENNICI Vol. 54

Oviposition in clusters has been suggested stages of the larval development: prediapause, as an adaptation of adult female butterflies to winter-diapause and postdiapause larvae (van conditions in which their reproduction would be Nouhuys et al. 2003, Tack et al. 2015). Typical time-limited rather than egg-limited if they laid to the study system is that in any one year a con- eggs singly (Courtney 1984, Parker & Courtney siderable proportion of local populations consists 1984). However, most of the adaptive explana- of only one group of diapausing larvae (usually tions put forward to account for the evolution of the offspring of one female butterfly) and also egg clustering address the consequences of gre- inbreeding is common (Saccheri et al. 1998). gariousness for success of the larvae rather than The aim of this study was to investigate consequences of clutch size for their mothers’ the importance of group size for egg and larval realized fecundities. Several field and labora- survival in field populations of M. cinxia and tory experiments have demonstrated increased to assess its role in population dynamics. We larval growth rate and survival with increasing conducted a field experiment in which group size group size (Lawrence 1990, Clark & Faeth 1997, was experimentally manipulated — with group Denno & Benrey 1997, Costa & Ross 2003, size varying from very small (10 eggs) to very Reader & Hochuli 2003, Fordyce 2006). These large (500 eggs) — in order to be able to detect effects reflect benefits accruing to larvae from potential negative effects of unusually small and group living that have been shown to include: unusually large group sizes. Additional data on group size and larval survival was gathered by 1. feeding facilitation during the first larval monitoring a large number of naturally occurring instar (Fitzgerald 1993, Clark & Faeth 1997, larval groups. In all cases we followed larval Zalucki et al. 2002, Reader & Hochuli 2003), survival by counting larvae and measuring their 2. improved thermoregulation through group locations in successive stages of their develop- basking (Porter 1982, Stamp & Bowers 1990, ment in order to see how the effects of group size Bryant et al. 2000) and larval webs (Casey vary during larval growth. 1993), 3. amplified strength of warning signals (Bowers 1981, 1993, Stamp 1980, Fitzgerald Material and methods 1993, Riipi et al. 2001, Reader & Hochuli 2003), and Study organism 4. active defences like head-jerking behaviour (Stamp 1984), regurgitation (Stamp 1984, Melitaea cinxia is a widely distributed butterfly Peterson et al. 1987) and construction of species in the temperate regions of Europe and protective webs (Fitzgerald 1993) and leaf Asia (Tolman 1997). In Finland, the distribution shelters (Damman 1987). of M. cinxia is restricted to the Åland Islands in SW Finland, where the butterfly occurs in small The Glanville fritillary Melitaea cinxia (Nym- local populations on several hundred dry mead- phalidae), is a butterfly whose larvae remain ows containing the larval host plants Plantago in groups until close to the time of pupation lanceolata (Plantaginaceae) and Veronica spi- (Thomas & Simcox 1982, Kuussaari 1998). This cata (Plantaginaceae) (Hanski 1999, Nieminen species has been the subject of detailed long- et al. 2004, Ojanen et al. 2013, Hanski et al. term studies of metapopulation dynamics in the 2017). In Finland, the butterfly has one genera- Åland Islands (Hanski 2011, Ojanen et al. 2013). tion per year. Investigations have shown that individual popu- Adult butterflies fly in June and lay eggs in lations are typically unstable with frequent local batches underneath the leaves of host plants. extinctions and colonizations observed annually The larvae are gregarious and spin a communal (Hanski 1999, Ehrlich & Hanski 2004, Hanski “feeding” web on the host plant immediately & Meyke 2005, Tack et al. 2015, Hanski et al. after hatching in July. On entering diapause, 2017). Mortality of larval groups is often high, usually in August, the larvae change their colour but it varies much among years and in different from light-brown to jet black and form a com- ANN. ZOOL. FENNICI Vol. 54 • Group size, and egg and larval survival in Melitaea cinxia 215 pact group inside a “winter nest” made of dense Group size of larvae in feeding webs was silk very different in appearance and texture estimated directly by counting the larvae. This from the feeding web. Larvae become active was not done when counts were made at the again early in the spring, in late March or early winter nest stage, because opening up the nest April. They are still black and spend much time would have been destructive. Winter nest size basking in the sun in conspicuous aggregations. (area covered by the winter nest in cm2) correlated They remain in groups typically until the begin- well with group size counted earlier in the autumn ning of the last larval instar and pupate within (Pearson’s correlation for the log-transformed var- the vegetation close to the ground in early May. iables: r = 0.91, n = 24, p < 0.001). The strength of this correlation justified the use of the winter nest size as an estimator of group size. By this Egg batch size means, group size in winter nests was estimated on a scale from 1 to 4 based on the size of the nest. We gathered data on egg batch sizes by catch- We marked the locations of egg batches and larval ing 151 butterflies in the field and allowing groups in the field with sticks and drew a map of them to oviposit on potted host plants in outdoor the locations of the groups in each population in cages with one female per cage. Half (48%) of order to be able to monitor their development. these females laid at least three egg clusters in During spring we recorded the number of captivity. The sizes of the egg batches obtained larvae in each larval group. No information from the caged females were counted under a was collected from groups that had potentially microscope. In addition, the sizes of 68 naturally become mixed with their neighbours. occurring egg batches were estimated in the field without breaking the clusters of eggs. Field experiment

Monitoring of naturally-occurring eggs To study the effects of a full range of potential and larval groups group sizes on egg and larval survival in a natural environment, we conducted an egg-transplant We collected information on a total of 1049 natu- experiment in which group size was experimen- rally occurring larval groups by surveying dry tally manipulated. We used six egg-batch sizes: meadows occupied by M. cinxia in two larval 10, 50, 100, 200, 300 and 500 eggs. Each size generations in 1993–1994 and 1994–1995. A class was replicated 10 times, except size class total of 588 and 461 prediapause larval groups 500, which was replicated 11 times. The total of were examined during the autumns of 1993 and 61 egg batches was divided among five meadow 1994. During the springs in 1994 and 1995, 178 patches occupied by M. cinxia so that each and 368 groups were re-examined after their meadow received one or two full sets of experi- winter diapause. In the summer of 1995, we mental egg batch sizes. searched for naturally occurring egg batches In order to obtain eggs for the experiment in four M. cinxia populations and checked the we caught female butterflies in the field, allowed survival of the hatched larval groups (n = 68) them to lay eggs in cages and transplanted the later in August. During these surveys we always eggs into the same habitat patch from which each recorded information on every larval group and butterfly was originally collected. After counting egg batch that we encountered. the eggs in the laboratory into equally sized clus- During the autumn surveys we recorded the ters, we placed them in tiny nylon mesh baskets following variables for each larval group: local and transferred them to the meadows used for population, day of observation, estimated egg the experiments. Some of the experimental egg batch size (only in summer 1995), larval group batches, especially those in the largest size class, size, larval web type (feeding web or winter included eggs from different females. nest) and larval winter nest size (area covered In the field, we selected plants for the experi- by the winter nest in cm2; only in autumn 1994). ment similar in shape, size and microhabitat to 216 Kuussaari & Singer • ANN. ZOOL. FENNICI Vol. 54

25 100 A B 50 C

20 s 80 40

60 15 30

40 10 20 Number of egg batches 20 Number of larval groups 5 10 Number of larval group

0 0 0 9 0 0 0 1–9 1–9

1–50 > 40 0 10–19 20–29 30–39 40–49 50–59 60–69 70–79 80–89 90–99 30–39 40–49 50–59 60–69 70–79 80–89 90–99 10–19 20–2 51–100 100–109 100–109 110–119 110–119 101–15 151–200 201–25 251–30 301–350 351–400 Egg batch size Larval group size in autumn 1993 Larval group size in spring 1994 Fig. 1. Group size distribution in (A) eggs, (B) prediapause larvae during autumn (n = 80), and (C) postdiapause, penultimate instar larvae during spring (n = 137). In A the grey bars show the distribution of egg batch sizes (n = 304, counted in the laboratory) of wild-caught butterflies allowed to oviposit in outdoor cages, whereas the white bars show the distribution of estimated cluster sizes for naturally occurring egg clusters (n = 68). the plants that M. cinxia females use for ovi- Statistical analyses position, but excluded especially small as well as isolated plant individuals in order to avoid We analysed the effect of group size on egg starvation of larvae due to food limitation during and larval survival separately for four phases the prediapause development. We hung the mesh of development: egg stage, prediapause stage in baskets gently from a host leaf located close to autumn, diapause stage in winter, and postdia- the ground in similar places to those that the pause stage in spring. Survival of larval groups females prefer for oviposition. The leaf covered (0 or 1) and larval survival within the surviving the basket completely and the group of eggs was groups (0–1) were analysed separately using located at the bottom of the basket within 5 mm logistic and linear regressions, respectively. below the host leaf. This method was expected Linear regression could be used because larval to minimize egg predation, while the larvae survival was normally distributed (Shapiro- hatched in a natural position, immediately below Wilk’s test). The effects of laying order on egg a host leaf. An advantage of this method was cluster size was tested using one-way ANOVA that the unhatched eggs remained in the basket as egg cluster size was approximately normally and could be subsequently counted in the labora- distributed (estimated visually, see Fig. 1A). tory. In the beginning of the experiment, all egg clusters were located on plants with several other Plantago individuals within the normal move- Results ment distance of prediapause larval groups of M. cinxia. Group size We assessed egg survival by collecting the egg baskets after larvae had hatched and count- Wild-caught females allowed to oviposit in out- ing the remaining unhatched eggs. Larval sur- door cages in 1996 laid 304 egg batches. The vival was estimated by counting the surviving mean egg-batch size was 174 eggs (commonly larvae in the field twice in the late summer and varied from 50 to > 250 eggs; Fig. 1 and Table 1). autumn and three times in the following spring. The maximum egg batch size recorded in the In contrast to the naturally occurring larval laboratory was 453 eggs while the size of the groups, in the transplant experiments it was pos- smallest egg batch was 10 eggs only. The size dis- sible to measure larval survival accurately also at tribution of egg batches that were observed in the the prediapause stage in the autumn. field in 1995 matched the distribution observed in laboratory-counted egg batches (Fig. 1). ANN. ZOOL. FENNICI Vol. 54 • Group size, and egg and larval survival in Melitaea cinxia 217

Following egg hatching in July, larval group Natural survival of larvae varied greatly size declined gradually until the time of pupa- among groups and years (Table 1). The results tion in the following spring when the larvae from monitoring of survival of naturally occur- dispersed individually (Fig. 1). There was some ring groups were in broad agreement with the variation in average larval group size between experimental results in that all observed pat- years. For example, in spring 1994 the average terns comprised positive effects of group size on postdiapause group size was two times greater survival of both groups and individuals within than in the previous spring (Table 1, Figs. 1B groups (Table 2). The effect on overwintering and C). The long tail in the group size distribu- survival of groups (Table 2 and Fig. 3A) was tion in the spring (see Fig. 1C) indicates high particularly clear with very high significance variation in within-group survival during larval in both years of study, aided by large sample development. sizes. This result was almost identical in the field observations and in the experiment (Fig. 2D). In contrast, other positive effects of group Group size and survival size in naturally occurring groups were less clear, being significant only in single study years for the In the egg transplant field experiment, in the following: within-group survival of prediapause 61 groups the number of surviving individuals and diapausing larvae, and group-level survival declined gradually at various stages of prepu- after diapause (Table 2). No negative effects of pal development (Fig. 2A). In this experiment, even the largest group sizes on survival were survival of entire groups was significantly posi- detected, either in the experiment or in the obser- tively affected by group size in two out of four vations on naturally occurring groups. developmental stages (Table 2 and Fig. 2C–D): in prediapause larval groups during autumn and in diapausing groups during winter. A similar Discussion pattern at the egg stage failed to reach significance (Fig. 2B), and group-level survival of The results showed that there is much variation post-diapause larvae was 100% (Table 2). Again in egg and larval group size in natural popula- in the experiment, within-group survival of indi- tions of M. cinxia, that survival increases with vidual larvae tended to increase with increasing group size throughout larval development and group size in all life stages studied, but this pat- that larval survival is highly variable both within tern reached significance only for diapausing and among years. Below we discuss each of insects (Table 2 and Fig. 2E). these three issues separately.

Table 1. Variation in mean group size and survival of eggs and larvae of Melitaea cinxia in natural populations during 1993–1995. Number of egg batches and larval groups studied indicated by n.

Developmental Years Group size Surviving groups Within-group survival stage mean SD n fraction n mean SD n

Eggs 1995a 159.0 67.5 068 – – – – – 1996b 174.0 75.6 304 – – – – – Prediapause larvae 1993 035.5 25.5 121 – – – – – 1994 – – – 0.84 050 0.30 0.13 42 1995 035.6 54.3 026 0.62 050 0.18 0.23 25 Diapause 1993–1994 – – – 0.77 170 0.52 0.33 61 1994–1995 – – – 0.97 368 0.62 0.22 44 Postdiapause larvae 1994 022.3 21.6 105 0.94 064 0.70 0.30 60 1995 042.4 33.1 232 1.00 037 0.89 0.17 37 a Field estimated sizes of naturally-occurring egg batches. b Laboratory counted egg batch sizes from field caught females allowed to lay eggs on potted plants in cages. 218 Kuussaari & Singer • ANN. ZOOL. FENNICI Vol. 54

500 1.0 A BC1.0

400 0.8 0.8 e 300 0.6 0.6

200 0.4 0.4 to diapaus

100 0.2 0.2 Fraction of groups surviving Number of surviving individuals 0 Fraction of hatching egg batches 0 0 e e 10 50 100 200 300 500

Egg batch size 1–20 > 300 21–50 End of 51–100 of eggs 101–200 201–300 Hatching diapaus Molting to Beginning sixth insta r Number of hatching eggs of diapaus

1.0 e D 1.0 E l 0.8 0.8 e

0.6 0.6

0.4 0.4 during diapaus

0.2 Within-group surviva 0.2

Group survival during diapaus 0 0 1–20 21–50 51–100 > 100 0 100 200 300 400 Prediapause larval group size Prediapause larval group size Fig. 2. Egg and larval survival in the group size field experiment. (A) Numbers of surviving individuals in the experimental groups at the various stages of development. The relationship between group size and the fraction of surviving groups (B) at the egg stage, (C) in prediapause larval groups and (D) in diapausing larval groups. (E) Within-group survival during diapause in relation to group size in those groups that survived over the winter. In C–E the relationship between group size and survival was statistically significant: (C) logistic regression: coefficient = 0.041, p = 0.030, 92% of cases classified correctly, n = 37; (D) logistic regression: coefficient = 0.102, p = 0.048, 93% of cases classified correctly, n = 29; (E) linear regression: coefficient = 0.0008, r 2 = 0.13, p = 0.045, n = 25.

Variation in group size balance between these opposing selective forces on clutch size is likely to shift with female age if Average cluster size was 174 eggs, which gen- older females mature eggs more slowly (Boggs erated a fairly high probability of producing & Nieminen 2004), as is suggested by the last instar larvae. However, small clusters with observed decline in size of successive clutches low probability of survival also occurred both in females that laid several times. Based on the in the field and in captivity. Why did adults lay observations of 65 females that laid at least three clusters smaller than those that would maximize egg clusters, the effect of laying order on egg- survival of their offspring? There are several cluster size was significant (one-way ANOVA: very different reasons why this may occur. First, F2,192 = 9.15, p = 0.0002). The first clusters laid there is likely to be parent–offspring conflict. A in captivity (mean size = 201 eggs) were larger female can increase offspring survival by delay- than the second clusters (mean size = 175), and ing oviposition until an extremely large clutch the second clusters were larger than the third can be laid. But once enough eggs have been ones (mean size = 155). matured to generate a clutch with relatively high A second reason for the observed clutches survival probability, further delay poses a risk to be so small as to jeopardize larval survival is of the mother dying before those eggs are laid. that oviposition may be interrupted at any stage, Therefore, such delay may reduce maternal fit- for example by attack from other invertebrates ness even if it increases offspring survival. The (Wahlberg 1995), reduction in solar insolation ANN. ZOOL. FENNICI Vol. 54 • Group size, and egg and larval survival in Melitaea cinxia 219

due to cloud cover (authors’ pers. obs.) and pre- p sumably also due to harassment by conspecific 0.080 0.343 males and grazing livestock. Egg predation by insect predators, e.g. lady beetles and lacewing (+) (+) 1995 larvae (van Nouhuys & Hanski 2004), may also effect direction be a reason for unusually small egg batches

being observed in the field in cases when the n entire cluster is not eaten. 33 19

Especially large larval groups in the field

result from the fusion of groups when more than p 0.003 0.107 one egg batch has been laid on the same host 0.0001 0.428 plant, a phenomenon that is not uncommon in M. cinxia. A survey of naturally-occurring egg + batches on 4300 randomly selected Plantago + (+) (+) effect 1994–1995 lanceolata plants in five M. cinxia populations direction all survived demonstrated that — even in relatively low all survived density populations of M. cinxia — some host Naturally occurring groups 26 13 17 19 17 n plants receive more than one egg cluster, and 392 that multiple egg batches on one plant occur more often than predicted by chance (Kuussaari et al. p

2004). Laboratory experiments produced see captions to Figs. 2 and 3. 0.147 0.878 0.778 0.009 0.035 evidence for two mechanisms explaining the < 0.0001 pattern observed in the field. Firstly, some host + + + (+) (+) plant individuals tend to be especially attractive (+) effect 1993–1994 to egg-laying females (Singer & Lee 2000), and direction secondly, ovipositing butterflies are attracted to 8 n 13 plants bearing conspecific eggs (Singer et al. 61 64 60 169 2017). The lack of observed negative effects on larval survival of even the largest group sizes in