We performed two experiments to assess how nectar influences the longevity and fecundity of adult female Boisduval’s blue butterflies (Icaricia [= Plebejus] icarioides) and how diets affect the population growth rate. Sugar and amino acids are the two major nutritional components of nectar (Corbet 2003; McDade and Weeks 2004; Arnold 2016) and are present in varying ratios in flowers visited by Boisduval’s blues (unpublished data). In our first experiment, we modified available nectar sugar; in our second experiment, we varied sugar and amino acids.
Study species and sites
Boisduval’s blues butterflies are a complex of ~ 25 recognized subspecies (Pelham 2008; Warren et al. 2017). The complex includes the federally threatened Fender's blue (I. i. fenderi) and endangered Mission blue (I. i. missionensis), the extinct pheres blue (I. i. pheres), the Washington state candidate for endangered species listing Puget blue (I. i. blackmorei)(WDFW 2022), and the Alberta/Saskatchewan imperiled pembina blue (I. i. pembina) (Alberta Environment and Parks 2017; Government of Saskatchewan 2021). Boisduval’s blues are non-migratory and univoltine, lay eggs singly, are nectar generalists, and overwinter as second instar larvae in the soil at the base of a host lupine (Lupinus spp.)(Schultz and Dlugosch 1999; James and Nunnallee 2011; Schultz et al. 2012). The species occurs in meadows and prairies, where lupine and nectar are present during their flight period. We use Puget blues in 2019 and pembina blues in 2020 due to COVID-19 shutdowns during the Puget blue flight period in 2020.
Puget blue butterflies are a candidate for listing as a Washington State endangered species that inhabits the Puget Sound area, Olympic Peninsula of Washington, and southern British Columbia, Canada. They use Lupinus albicaulis as a host at the collection site, Joint Base Lewis-McChord (46.92, -122.73; Rainier, Washington, U.S.A.), which is owned by the U.S. Department of Defense and managed jointly by the U.S. Army and U.S. Air Force. Pembina blues are common throughout the Cascade Mountains in Oregon and Washington and their range is the largest of all subspecies, present throughout most of the interior of Canada and the United States. We collected pembina blues within the Gifford-Pinchot National Forest (45.78, -122.17; Yacolt, Washington, U.S.A.) and Mt. Hood National Forest and Wilderness Area (45.34, -121.67; Government Camp, Oregon, U.S.A.). They typically use Lupinus latifolius (varying ssp.) as their host plant at these sites.
Experimental Procedures
In the first experiment, testing sugar levels, we worked with Puget blue in 2019, and for the second experiment, testing sugar and amino acids, we worked with pembina blue in 2020. All butterflies in the experiment were collected as newly eclosed adult females indicated by minimal wing wear from wild populations. Butterflies were chilled after capture in the field and then placed in individual housing before transport to the Washington State University Vancouver greenhouse. Butterflies were randomly assigned to a treatment and provided with nectar or water. Females that died within 48 hours of capture are excluded from the trial (n = 2) because death is more likely due to conditions the butterfly was experiencing before capture. We also excluded infertile females or females damaged during captivity (n = 2). Butterflies were housed in the Washington State University Vancouver greenhouse with a lupine stem (see ESM 2 for husbandry). Fresh sponges with the diet were placed daily, and eggs were removed from the lupine daily.
To assess response to experimental treatments, we collected the following data: daily number of eggs laid, the butterfly’s weight every three days, longevity and unlaid eggs, and weight at death. Longevity was measured as the number of days from the date of collection to the date of death. We measured unlaid eggs by dissecting the abdomen of females within 48 hours of their death, following O’Brien et al. (2004). Eggs are classified either as developed eggs (> 0.5 mm in diameter) or partially absorbed eggs (< 0.5 mm in diameter; partially absorbed eggs). A small percentage of partially absorbed eggs (> 2%) were likely eggs that never developed; these eggs were small (~ 0.1 mm), clear, and hard to detect. Nearly all eggs classified as partially absorbed eggs were green or cloudy white indicating development had occurred. If Boisduval’s blues are pro-ovigenic we would expect that a majority of eggs are developed at the time of eclosion (Hill and Pierce 1989; O’Brien et al. 2004; Miller 2005; Jervis et al. 2005), and undersized colored eggs at death are more likely to be those that are being reabsorbed than never developed.
Sugar Nutrition Experiment
Our experiment included three treatments to investigate how sugar affected butterfly longevity and fecundity. We made one batch of sucrose solution at 300 mg per mL, which is thought to be an ideal viscosity for proboscis feeding (following results of Kim et al. 2011), and froze it in aliquots and used it as needed; we also froze aliquots of water and used those as required. Common composite flowers in the habitats where Boisduval’s blue butterflies reside are commonly ~ 65 mg sucrose/flower (R. Bonoan personal communication). We used three nectar treatments: ad libitum, restricted, and water. The females in the ad libitum (A) treatment were fed twice daily, 2mL of the sucrose solution (1200 mg sucrose/day, 18x composite flower), in the morning and the afternoon. The females in the restricted nectar treatment (R) received 1 mL of nectar (300 mg sucrose/day, 4.6x composite flower) in the afternoon for 1 hour, with water available throughout the rest of the day. The water treatment (W) received only water on their sponges. Each treatment had 10 individually housed females, except one ad libitum female had a cystic mass in the abdomen that prevented dissection and was excluded (n = 9).
Sugar and Amino Acid Experiment
This experiment was conducted in 2020 with pembina blues. After our first experiment, we saw that the sugar levels appear to exceed the daily requirement for Puget blues and reduced the amount of sucrose provided. We used four treatments, water (W), lupine (L), flower (F), and flower plus lupine (F + L) with eight females per treatment. The water treatment (W) received fresh sponges daily with water. Nectar treatments were simulated after two nectar species, sickle-keeled lupine, lupine (L) and composite flowers, flower (F), each contained 65 mg of sucrose per day, but varied the quantity of amino acids. In our final treatment, flower plus lupine (F + L), butterflies had access to both nectar treatments, where each treatment (F or L) was given daily on separate sponges (total availability is 130 mg of sucrose and 23 mg of amino acid per day). Amino acid levels were based on field sampling of histidine, as an indicator of amino acids (R. Bonoan personal communication), with the lupine treatment having higher amino acids, and flower treatment is lower (17 and 6 mg/day respectively of histidine, and the quantity of other amino acids are dependent upon the blend used: see ESM 2 Table 1). The amino acid blend included all nine essential amino acids plus cystine and tyrosine. Each treatment, including water, had the addition of sea salt (2 mM of sodium), as an estimate for nectar salt (McDade and Weeks 2004). We prepared the two nectars and the water in one batch and froze the solution into aliquots to be used as needed. Each treatment had 8 individually housed females.
Table 1
Mean and 95% bootstrap prediction intervals by treatment across all butterflies (Icaricia icarioides) in both the Sugar and the Sugar and Amino Acids experiments. Total fecundity is eggs laid from collection day to death, and longevity is the number of days lived from collection day to death and unlaid eggs are those inside an individual at time of death determined via dissection. Superscript letters refer to the significant differences between pairs via estimated marginal means with Tukey’s post-hoc correction.
Experiment | Treatment | Toal Fecundity (95% CI) | Longevity (95% CI) | Unlaid Eggs (95% CI) |
---|
Sugar (n = 10 females per treatment) | Ad Libitum | 228 (127, 342)a | 11 (8, 14)a | 46 (31, 62)a |
Restricted | 280 (164, 396)b | 12 (8, 16)b | 46 (29, 62)b |
Water Only | 68 (43, 95)ab | 5 (4, 6)ab | 89 (68, 111)ab |
Sugar and Amino Acids (n = 8 females per treatment) | Flower & Lupine | 175 (100, 274)a | 20 (14, 25)a | 26 (5, 51)a |
Flower | 101 (53, 164) | 13 (7, 18)b | 43 (16, 80) |
Lupine | 136 (42, 252) | 13 (10, 17)c | 19 (10, 28)b |
Water Only | 34 (7, 70)a | 4 (3, 4)abc | 89 (70, 103)ab |
Analysis
We performed the same types of analyses for the two experiments. We used R for statistical analyses (R Core Team 2024), and we used ANOVA from the car library created generalized linear and generalized linear mixed models, visually assessing if models met the appropriate model assumptions. We performed Wald χ2 tests to detect evaluate treatment effects using ANOVA from the car package (Fox and Weisberg 2019). We performed pairwise comparisons of treatments with the Tukey’s contrast of the estimated marginal means using emmeans (Lenth 2024). We created 95% bootstrapped prediction intervals using glm.predict (Schlegel 2024) and bootpredictlme4 for mixed models (Duursma 2023). Data was manipulated and figures were made using tidyverse (Wickham et al. 2019) and ggpubr (Kassambara 2023).
We tested the effects of diet on weight using a Gaussian generalized linear mixed model with the individual butterfly as a random effect on intercept (weight ~ treatment *day in trial) using the lme4 library (Bates et al. 2015). We modeled daily fecundity as an effect of treatment and day on each diet (daily eggs laid ~ treatment * day in trial) with individual butterfly as a random effect on intercept using a negative binomial generalized linear mixed model with a log link. We used quasi-Poisson generalized linear models to evaluate longevity (longevity ~ treatment), total fecundity (total fecundity ~ treatment), and unlaid eggs (unlaid eggs ~ treatment) after finding overdispersion in Poisson models. We evaluated all model distributions by comparing residuals and Q-Q plots, evaluating dispersion parameters and selected the best fitting distribution using AIC or qAIC from MASS (Venables and Ripley 2002).
Population model
We used experimental data to estimate parameters in a population model and to estimate relative population growth rates. We estimate two parameters from our data, butterfly longevity, L, (Eq. 1) and fecundity, F (Eq. 2).
\(L \sim QPois\left({\mu }_{x}\right)\) (Eq. 1)
The average longevity of females, L, is estimated assuming Poisson distribution adjusted for overdispersion (Quasi-Poisson) using the mean of the given treatment, x.
\(F=\sum _{D=1}^{L}{f}_{D} \sim Negbinom\left({\mu }_{xD}\right)\) (Eq. 2)
We use daily fecundity, fD, across the lifespan, L, summed to estimate the average total fecundity per female, F (Eq. 2). fD, is the fecundity on day, D, from a negative binomial distribution using the mean fecundity on that day, D, for the given treatment, x, for each day from 1 until the longevity, L, of the butterfly. The daily fecundity, fD, is summed to estimate total fecundity, F, and then we estimate population growth rate, λ (Eq. 3), where we assume a 50:50 sex ratio, and therefore the number of eggs per capita is ½ F.
\({\lambda }=\frac{1}{2}\times F \times \text{s}\) (Eq. 3)
Estimates of immature survival, s, set at 2.7%, based on the mean of estimates from prior studies of Boisduval’s blue survival from egg in summer to post-diapause survival the following spring (n = 46 site-years) (Schultz and Crone 1998; Warchola et al. 2017; Schultz and Ferguson 2020). Post-diapause to eclosion survival is not easily measured in situ (pupae are in the soil or cryptic), individual larvae are free to move out of survey areas, and some proportion of individuals observed are near pupation (variable by site-year); therefore, 2.7% survival, on average, represents the minimum survival of eggs to mid-way to pupation after the breaking of diapause. In the laboratory, Puget blue egg to post-diapause survival was 37% and post-diapause to adult survival was 20% across two rearing environments, resulting in 7.5% of eggs surviving to become adult butterflies, which is thought to be only a moderate improvement in survival over what occurs in nature (Schultz et al. 2009). Therefore, given the limits of the existing data, we assume that post-diapause to eclosion survival is 100%. Additionally, we assume that immature survival is not affected by nectar treatments, because nectar quality has shown no effect (Woods et al. 2010; Niitepõld 2019) or has improved immature survival (e.g. Jensen et al. 1974; Song et al. 2007; Marchioro and Foerster 2012), and therefore our model conservatively represents the effect of nectar on the population growth rates.
The population growth rate was calculated 10,000 times per diet, with randomly selected longevity (L) and fecundity (fD) for each calculation to sum for total fecundity (F); therefore, variation in λ represents variation from individual daily fecundity and longevity within the experiment. When comparing resulting population growth rates to determine if treatments provide sufficient resources across treatments, we consider λ = 1 to be the minimum indication of stable population, and λ = 1.55 as the threshold for long-term stable populations given previous research on the species documenting high stochasticity and possibility of density dependence (Schultz and Hammond 2003). We evaluated the sensitivity of the population model (Eq. 4) to the survival parameter by solving the equation for fecundity, Fi, that results from a given immature survival, si, when λ = 1.55.
\({F}_{i}=1.55 ÷({s}_{i}\times \frac{1}{2})\) (Eq. 4)