Temperature during pupal development affects hoverfly developmental time, adult lifespan and wing length

Hoverflies (Diptera, Syrphidae) are cosmopolitan, generalist flower visitors and among the most important pollinators after bees and bumblebees. The dronefly Eristalis tenax can be found in temperate and continental climates across the globe, often synanthropically. Eristalis tenax pupae of different generations and different climate zones are thus exposed to vastly different temperatures. In many insects, the ambient temperature during the pupal stage affects development, adult size, and survival; however, the effect of developmental temperature on these traits in hoverflies is comparatively poorly understood. We here reared E. tenax pupae at different temperatures, from 10°C to 25°C, and quantified the effect on adult hoverflies. We found that pupal rearing at 17°C appeared to be optimal, with high eclosion rates, longer wings, and increased adult longevity. Rearing temperatures above or below this optimum led to decreased eclosion rates, wing size, and adult survival. Similar thermal dependence has been observed in other insects. We found that rearing temperature had no significant effect on locomotor activity, coloration or weight, despite evidence of strong sexual dimorphism for each of these traits. Our findings are important as hoverflies are key pollinators, and understanding the effects of developmental temperature could potentially be useful for horticulture.

Methods
Animals

We collected egg batches from female Eristalis tenax hoverflies wild caught under permit in Wittunga Botanic Garden (Adelaide, South Australia) and reared the larvae in fresh rabbit dung (Nicholas et al. 2018) at room temperature (22.8 ± 1.0°C). Each batch contained around 200 eggs (Nicholas et al. 2018). Third instar larvae were moved to small containers for rearing in different temperature conditions until the adult hoverflies emerged from the pupae. All pupae were kept in small, transparent containers fitted with a mesh under a 16 hour light: 8 hour dark cycle, with light provided by Arlec LED lights (UC0168, 350 lumens, warm white, Arlec Electrical Services, Australia) controlled by a timer. We exposed the pupae to 5 different temperatures. The coldest was achieved using a fridge (Hisense, HR6AF243, Hisense Australia) set to 10°C (10.1 ± 0.47°C). The 12°C (12.1 ± 0.54°C) and 17°C (17.4 ± 0.37°C) pupae were kept in wine coolers (Kogan 8 Bottle Thermoelectric Wine Cooler, Kogan Australia Pty Ltd). The 23°C pupae (22.8 ± 1.0°C) developed at room temperature in insect rearing cages. The 25°C (25.4 ± 1.76°C) pupae were housed on a plant propagation heating mat controlled by a THD digital controller (Aldoheat Horticultural Products, Australia). In one case, 10 pupae reared at 25°C were moved to room temperature for the last 2 days of their pupation, but treated as part of the 25°C cohort.

Upon eclosion, the adult flies were moved to insect rearing cages (BugDorms, Australian Entomological Supplies) with a 24.5 cm or 32.5 cm side for maintenance at room temperature under laboratory lights (Nicholas et al. 2018). When counting the eclosion rate, we treated the hoverflies that died before completely crawling out of the pupae as unsuccessful.

Adult hoverflies were labelled with unique color marks (Semco acrylic paint) on the thoracic dorsum to enable tracking of individuals. We recorded the lifespan of adult hoverflies from the day of eclosion to their death. Hoverflies that died from unnatural causes (e. g. by freezing for experimental reasons) appear as censored datapoints, and 10 individuals were omitted from the survival analysis due to lacking data about their sex.

Morphometrics

For the hoverflies reared at 12°C, 17°C and 25°C, we assigned a group of individual hoverflies (30-40 per temperature, balanced sample from each sex and batch) that were kept in a separate cage and regularly weighed at an interval of 5-15 days. The first measurement took place within 1-2 weeks of emergence. Each hoverfly was weighed by placing it in a small tissue culture dish (35 x 10 mm, Sarstedt AG & Co. KG, Germany), using a Sauter AR 1014 electronic balance (Sauter GmbH, Germany). These measurements were done un-blinded.

After the hoverflies died, we took high-resolution photographs of them using an Olympus SZX10 camera equipped with DF PLAPO 1X-4 JAPAN lens (130 hoverflies, done blinded), or Olympus E-M10 Mark II camera equipped with Olympus ED 12-50 mm f/3,5-6,3 EZ lens (57 hoverflies, unblinded). 226 hoverflies were photographed while still alive using the Olympus E-M10 Mark II camera, and to prevent their movement during photographing, we covered them with a Petri dish lid pressed against a soft cellulose square. Camera type and status during photographing (alive/dead) were used as predictors for statistical models to filter out potential systematic variability. Wings were cut from living or dead specimens using scissors and were stretched with a Petri dish for photographing. We used a ruler to calibrate the photographs, and then extracted wing length, measured between the point where the transversal h-vein joins the upper edge of the wing and the point where the R4+5 vein joins the tip of the wing (as in Ottenheim & Volmer 1999). We measured the length of the thorax, defined as the distance between the scutellum-metathorax border and the head-prothorax border, along the center of the metanotum. All measurements were conducted using ImageJ software (Schneider, Rasband & Eliceiri 2012).

From the photographs, we defined the coloration of tergite 2 and tergite 3, which can be yellow, orange, brown, or black (Francuski et al. 2011). We used a color scale with 6 steps (pictograms, Fig. 5), with 3 lighter grades and 3 darker grades (as defined in Heal 1979). For this classification the size of the colored patches is more important than shade (Heal 1979). However, note that this is a simplified scale as other authors have used up to 22 different Eristalis tenax color morphs (Francuski et al. 2011).

Activity

To quantify locomotor activity we used the Locomotor Activity Monitoring system (LAM25, TriKinetics Inc, Waltham, MA, USA) with 25 mm diameter × 125 mm long Pyrex glass tubes (PGT25 × 125, TriKinetics Inc) positioned horizontally. Hoverflies were individually placed in each tube, where the ends were sealed with a cotton ball with water, some pollen, and honey (Thyselius & Nordström 2016). Additional water was added to the cotton balls twice daily (before 9:00 and after 16:00) by a Pasteur pipette to prevent drying out. Each time the hoverfly crossed the center of the tube, it would break an infrared beam, and this would be counted as an activity measurement. The hoverflies were kept in the Locomotor Activity Monitoring system (LAMS) for approximately 48 hours, starting in the morning, and the average activity between 10:00 and 16:00 on the second day of recording was used for quantification. We quantified the mean activity at 12°C, 17°C, or RT by placing the LAMS in wine cooler fridges or at room temperature, under a 16 hour light: 8 hour dark cycle.

Sometimes a hoverfly would remain stationary in the middle of the tube for extended periods of time, continuously breaking the beam. To avoid giving erroneously high activity measurements, all measurements indicating more than 10 crossings per minute were replaced with a 1, and all non-null measurements immediately following it were replaced with a 0 (as in Thyselius & Nordström 2016).

Statistical analysis

Throughout the paper, N refers to the number of flies. In the text, all data given as mean ± stdev unless otherwise mentioned. Where the figures show boxplots, these indicate median and interquartile ranges, and the whiskers extend up to 1.5x of the interquartile range. Any data beyond this distance are considered as outliers, shown with individual points.

All statistical analyses were performed using R 4.2.1 software. Where data were normally distributed, we conducted ANOVA analysis combined with Tukey’s HSD. In data with non-normal residuals, we used appropriate transformation. In non-normally distributed data, we used generalized linear model (GLM) of either quasi-poisson (over-dispersed count data) or quasi-binomial (proportional data) family. For ordinal response variable (coloration), we used cumulative link model (CLM). When dealing with repeated measurements (weight gain data), we checked the data using partial autocorrelation function (PACF) and found only weak temporal autocorrelation. Subsequently, we compared results and AIC of the weight gain model with and without autocorrelation structure, and as these two reported similar results, we implemented the model without it. For survival analysis, we used Cox proportional hazards model with time-splitting (tt) correction.