Down syndrome cell adhesion molecule 1: testing for a role in insect immunity, behaviour and reproduction

Down syndrome cell adhesion molecule 1 (Dscam1) has wide-reaching and vital neuronal functions although the role it plays in insect and crustacean immunity is less well understood. In this study, we combine different approaches to understand the roles that Dscam1 plays in fitness-related contexts in two model insect species. Contrary to our expectations, we found no short-term modulation of Dscam1 gene expression after haemocoelic or oral bacterial exposure in Tribolium castaneum, or after haemocoelic bacterial exposure in Drosophila melanogaster. Furthermore, RNAi-mediated Dscam1 knockdown and subsequent bacterial exposure did not reduce T. castaneum survival. However, Dscam1 knockdown in larvae resulted in adult locomotion defects, as well as dramatically reduced fecundity in males and females. We suggest that Dscam1 does not always play a straightforward role in immunity, but strongly influences behaviour and fecundity. This study takes a step towards understanding more about the role of this intriguing gene from different phenotypic perspectives.


Study insects and bacteria.
Tribolium castaneum (strain Cro1) were wild-collected in Croatia in 2010 (see 1) and adapted to laboratory conditions for at least 12 generations before the start of the experiments. Beetles were raised on organic wheat flour (Alnatura type 550) with 5 % brewer's yeast at 30 ºC, 70 % humidity, in a 12 hour light/dark cycle. Drosophila melanogaster (1_4WS) originated from inseminated females that were wild-collected in Münster, Germany in 2008 and allowed to adapt to laboratory conditions for at least 50 generations before the start of the experiments.
These species have been estimated to have last shared a recent common ancestor around 300-327 million years ago (2,3). For D. melanogaster we used whole body samples of eggs, first and third instar larvae, pupae, and adults were used. In addition we sampled fat body, haemocytes and brain from third instar larvae and fat body and brain from adults. We also sampled testes and ovaries from adults. To produce the samples, flies (4-5 days post pupal eclosion) were allowed to lay eggs for two hours on agar plates (1.5 % agar; 1 % vinegar) covered in a thin layer of yeast. The eggs were washed off the plates with PBS and three replicates of ten 12-14 hour old eggs were frozen. The flies were allowed to lay eggs for an additional 10 hours, after which time the eggs were washed off the plates with PBS and added to food vials at a constant density (methods after 4) and allowed to develop. Whole body samples of first instar larvae (35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45) hours after egg laying), third instar larvae (120-132 hours after egg laying), pupae (156-168 hours after egg laying), and adults (14 days after egg laying; i.e. around 4 day old adults) were used. The haemolymph from third instar larvae was obtained by carefully pulling animals apart on a glass slide without disrupting the gut. For fat body dissection third instar larvae were pulled apart by the mouth hooks and cuticle and the fat body was separated from other tissues. For adults, the fat body was obtained by removing the abdomen from the body; the gonads were discarded and the fat body tissue was collected into 50 µl of ice-chilled Drosophila Ringer's solution (182 mM KCl; 46 mM NaCl; 3 mM CaCl 2 ; 10 mM TrisHCl; (5)) and shock frozen. Ovary and testis sampling were performed in the same manner as for the fat body but with an additional centrifugation step. All samples were immediately centrifuged after dissection (4 ºC, 23 rcf, 5 min), the supernatant was discarded and the samples were snap frozen in liquid nitrogen and stored at -80 ºC until RNA extraction. All samples from the pupal and adult stages, other than testes and ovaries, contained 5 males and 5 females.
For RNA extraction a tissue specific volume of TRIzol reagent (Ambion, U.S.A) was added to the frozen homogenised tissue samples, incubated for 10 min at room temperature (RT) and vortexed two times during incubation. To ensure complete lysis of the fat body, samples were first incubated with 250 µl TRIzol with an additional shock freezing and sonification step ( The resulting cDNA was used for qPCR analyses using gene-specific primers (Table S1).
Where genes contained more than one exon, primers were designed such that the forward or reverse primer spanned an intron. Amplification efficiencies (E) of the primer pairs were determined with five dilutions (undiluted, 1:10, 1:100, 1:1,000, 1:10,000) of template cDNA, where E = 10 -1/slope . The qPCR was performed in a 384-well plate format, with a total reaction volume of 10 µl in each well. From each cDNA sample two technical replicate qPCR reactions were performed using the Kapa SYBR® Fast qPCR Mastermix according to the manufacturer's instructions. The reaction was run on a LightCycler480 (Roche) using the following protocol: 95 °C for 5 min, followed by 40 cycles of annealing and amplification at 60 °C for one min and denaturation at 95ºC for 15 sec. As a final step the products were heated up to 95 °C with continuous fluorescence measurements to obtain the melting curves and subsequently cooled to 40 °C. The crossing point (Cp) values (see raw data) were calculated using the Fit Point method using a Noise Band threshold of 10, with the LightCycler®480 software and the average Cp values from each of two technical replicates were used for analyses. Dscam1 expression for each sample was calculated relative to the geometric mean of two reference genes (ribosomal protein 49 (Rp49) and ribosomal protein 13a (RpL13a)), used in the formula: E (reference) -E (target) = ∆ct. The genes Rpl13a and Rp49 were chosen as reference because they have been found to have relatively stable expression across treatments, life history stages and tissues (e.g., 6, 7-9); however it is important to recognise that it is difficult to find universally suitable reference genes (9). Although across tissues and life history stages, the Ct values of the two reference genes were tightly correlated with each another (D. melanogaster: r 2 = 0.991, df = 34, p < 0.0001; T. castaneum: r 2 = 0.986, df = 37, p < 0.0001), it can be expected that reference genes do not necessarily show identical expression across different tissues and life history stages therefore we discuss the results in terms of expression of Dscam1 relative to the mean expression of the two reference genes.
A similar experimental set up was used for T. castaneum: whole body samples of eggs, ten and fifteen day-old larvae, three day-old pupae and seven day-old adults were used. In addition we sampled fat body, haemocytes and brain from fifteen day-old larvae and from seven day old adults, and the testes and ovaries from adults. In detail, to produce the samples, two to three week old adult beetles were allowed to lay eggs for 24 hours. Three replicates of ten eggs were collected and frozen, and the remaining eggs were further cultivated as described above. After ten days, larvae were individualised in 96-well microtitre plates containing ad libitum flour and 5 % yeast and incubated until sampling. Haemolymph was taken by pricking the larvae and adults with a fine needle (2 -10 µm Ø) between the head and thorax. Clear haemolymph was collected with 1 µl glass capillaries (Hirschmann, Germany) and immediately transferred into 50 µl ice-chilled Schneider's insect medium (Sigma) containing 10 % fetal calf serum (FCS). For fat body dissections, larvae were cooled on ice and transferred to a microscope slide with a drop of PBS. Larvae were cut transversally at the first and last segment. The gut was pulled out and the fat body was then squeezed out with forceps and collected in 50 µl ice-chilled PBS. For adult fat body dissection, equal numbers of males and females were cooled and decapitated. The gut and gonads were removed and the fat body was collected in 50 µl ice-chilled PBS. Testes were sampled by cooling the beetle and cutting the abdomen by cutting beneath the last pair of legs on the ventral side. After transfer into a drop of PBS the testes were squeezed out of the abdomen and collected in 50 µl icechilled PBS. To dissect ovaries, females were cooled and cut on the ventral side between second and third pair of legs. The abdomen was then cut laterally and transferred into a drop of PBS. Ovaries were push out of the abdomen and transferred into 50 µl ice-chilled PBS. The qPCR was performed in a 96-well plate format, with a total reaction volume of 15 µl in each well. All samples were further treated as described for D. melanogaster.
Dscam1 expression in larval D. melanogaster and T. castaneum upon infection.
To address whether there is a change in Dscam1 expression after bacterial challenge we performed experiments using two different infection routes (e.g., see 10, 11). The first two experiments examined gene expression after haemocoelic bacterial exposure of both D. melanogaster and T. castaneum. In the third we took advantage of the fact that B. thuringiensis can also be orally administered and cause pathogenesis to T. castaneum (1) and therefore examined Dscam1 expression after oral exposure. Infections and dissections were performed blind with respect to treatment as far as possible and treatment orders were randomised within each replicate block.
As a positive control that the immune system had been activated, we additionally tested the expression of three immune genes for each host species. We monitored the expression of Imd and two species-specific antimicrobial peptides (AMPs), Attacin2 (Att2) and Coleoptericin1 (Col1) for T. castaneum (12)(13)(14)(15) and Diptericin (Dpt) and Drosomycin (Drs) for D.
melanogaster (15)(16)(17)(18)(19). In brief, Imd plays a signal transduction role in the Imd signalling pathway, this pathway together with the Toll pathway, is responsible for the expression of AMPs after recognition of a bacterial infection (16). Imd has been found to be upregulated in T. castaneum after infection with some microorganisms (12). In D. melanogaster the Imd pathway is activated by Gram-negative bacteria (or Gram-positive bacteria with DAP-type peptidoglycan) and leads to the expression of, e.g., Dpt; the Toll pathway is activated by Gram-positive bacteria and fungi and leads to the expression of e.g., Drs (19,20). Both bacteria species we tested, B. thuringiensis and E. coli, have previously been found to result in increased expression of at least one of our chosen AMPs (e.g., 16,17). It has been suggested that T. castaneum has more promiscuous activation of AMPs than D. melanogaster (14), and the AMPs that we chose to test have been shown to be upregulated after infection by bacteria, including E. coli (12) Bacillus subtilis (14), and another Gram-positive bacteria,
The experiment consisted of six replicates, each replicate being performed on a different day.
Adult beetles were therefore allowed to lay eggs for 24 hour periods over a course of six days.
10 day old T. castaneum larvae were individualised in 96-well microtitre plates containing flour and 5 % yeast and incubated as described above. Fifteen day old larvae were pricked dorsolaterally between the second and third segment with a fine needle (2-10 µm Ø) that had previously been dipped into a bacterial suspension or PBS (treatment control, hereafter termed TC), or they were left untreated (naïve control, hereafter termed Naïve). The bacteria concentrations used were: B. thuringiensis: 1 x 10 10 mL -1 ; E. coli: 1 x 10 10 mL -1 ; P. fluorescens: 2 x 10 7 mL -1 all suspended in PBS. For this and the following Dscam1 expression after infection experiments we generally aimed to use bacterial concentrations that resulted in low mortality (typically between 0 and 20 % over three days following exposure). We did not want to create strong selection for a small sub-group of survivors from which we estimated gene expression, yet we wanted to elicit an immune response. The bacterial concentration information was gained from a combination of preliminary experiments and data that we have previously published. In detail, the B. thuringiensis concentration for pricking (1 x 10 10 cells / mL -1 ) was the same as in (11). Similarly as for D. melanogaster, the E. coli strain we used is non-pathogenic for T. castaneum, therefore we used the same concentration (1 x 10 10 cells / mL -1 ) as we used for D. melanogaster. The P. fluorescens concentration was based upon preliminary experiments showing that 2 x 10 7 cells / mL -1 resulted in approximately 20 % mortality at 3 days post exposure. After pricking, the larvae were put into fresh 96-well microtiter plates containing flour and 5 % yeast. The fat body and haemocytes were sampled six and 18 hours after treatment as described before. We produced six replicates per combination of treatment / tissue / time point, whereby each replicate consisted of tissue from ten pooled animals (6 replicates x 5 treatments x 2 tissues x 2 time points x 10 animals = 1,200 larvae). Survival at six and 18 hours was examined at the time of sampling and we also produced an extra 14 animals per replicate / treatment / tissue / time point to estimate survival at 6, 18 and 24 hours (n = 1,680). Average mortality 6 hrs after pricking was less than 1 % for all five treatment groups; at both 18 and 24 hrs after pricking mortality in the B. thuringiensis pricked larvae was ~ 10 %, and for all other treatment groups was it was ~ 3 % or less.
RNA was extracted, cDNA synthesised and qPCRs run as described in before. For each sample, Dscam1 (Table S1) expression was examined as well as the expression of three immune-related gene: Attacin2 (Att2), Coleoptericin1 (Col1) and Imd (Table S1). Cp values were calculated using the Fit Point method with a Noise Band threshold of 3.8844. In some cases for the infection experiments, the melting curves fulfilled our quality criteria (i.e., were above detection limit and had a single clear peak at the expected melting temperature), but the standard deviation (SD) of the Cp value across the two technical replicates was ≥ 0.8. If after repeating the qPCR the SD was still ≥ 0.8, or if there was no mRNA signal at all (e.g., T. castanuem, Attacin2, 18 hour haemocyte sample) we removed the sample (gene/tissue/time point/treatment combination). For D. melanogaster this was the case for 38 out of 576 samples, and for T. castaneum this was the case for 36 out of 720 samples. These removals explain why in Figure 1 there are sometimes less than 6 biological replicates. Of the remaining samples, in some cases the melting curve fell below the detection limit of the LightCycler®480 software. In these cases we re-did the qPCR. If after the re-run the melting curve was still below the detection limit we used one technical qPCR replicate for that biological replicate where the melting curve of the remaining technical replicate met our quality criteria. This was the case for 19 of the remaining 538 samples for D. melanogaster and 16 of the remaining 684 samples for T. castaneum. We analysed the data using REST© 2009 (relative expression software tool; 21) and compared the gene expression of two groups at a time: in each case the naïve control group was tested against the treatment control or one of the bacteria-exposed groups. REST calculates the relative fold expression differences by using the expression of reference genes (Rpl13a and Rp49) to normalise the expression levels of the target genes (Dscam1, Att2, Col1 and Imd), whilst taking the reaction efficiency (E) of the PCR into account; it is based on the following formula (21,22): A pair wise fixed reallocation randomisation test© is performed to examine whether there are significant differences between the two groups. We allowed 2000 random reallocations of the observed Cp values to the two groups being tested; REST© notes the expression ratio change for each reallocation, and the proportion of these effects gives the p-value assuming a 2-sided test. In our figures we present the mean and standard errors as calculated according to the REST© software, i.e., the results of the 2000 random reallocations.
Dscam1 expression after haemocoelic bacterial exposure in larval D. melanogaster.
The experiment consisted of six replicates, each replicate being performed on a different day.
four to seven days post-eclosion, flies were allowed to lay eggs on agar plates, as described before. Eggs were collected in the morning and evening for six days, and adults were replaced every two days. Prior to bacteria challenge, late second instar D. melanogaster larvae were thuringiensis is more pathogenic than E. coli, and preliminary experiments showed that 1 x 10 7 cells / mL -1 resulted in 20-30 % mortality 3 days post-infection, and 1 x 10 6 cells / mL -1 resulted in low mortality, therefore we chose the concentration of 7.5 x 10 6 cells / mL -1 .
Bromophenol blue was used to check the injection success, since it is not toxic when fed to larvae (e.g., 23) and it did not induce mortality of larvae after injection and did not negatively affect bacteria survival (personal observation). We produced six replicates per combination of treatment / tissue / time point, whereby each replicate consisted of tissue from ten pooled animals (6 replicates x 4 treatments x 2 tissues x 2 time points x 10 animals = 960 larvae).
After injection the larvae were placed individually into 0.2 mL PCR reaction tubes containing fly food, recipe as described above. Survival at 6 and 18 hours was examined at the time of sampling and we also produced an extra 7 or 8 animals per replicate / treatment / tissue / time point to estimate survival at six, 18 and 24 hours (n = 725). Average mortality 6 and 18 hrs after pricking was less than 2 % for all five treatment groups; at 24 hrs after injection mortality in the B. thuringiensis exposed larvae was less than 4 %, and for all other treatment groups it was it was 0 %. The methods for RNA extraction, cDNA synthesis and qPCR were the same as for T. castaneum except that the immune genes examined were: Diptericin (Dpt), Drosomycin (Drs) and Imd (Table S1). Relative fold expression calculation and statistical analysis was done as described for T. castaneum.

Dscam1 expression after oral infection in larval T. castaneum.
The experiment consisted of six replicates, with two replicates being performed per day.
Adult beetles were therefore allowed to lay eggs for 24 hour periods over a course of three days. Seventeen day old T. castaneum larvae were exposed to B. thuringiensis sporecontaining diet or to a control diet. The B. thuringiensis spore-containing diet was prepared as described in (1) and using the same concentration. In brief, the spore concentration was adjusted to 1 x 10 9 mL -1 with PBS, and 0.15 g of flour with 5 % yeast was added per mL of spores. Forty microlitres of liquid diet per well was pipetted in a 96 well plate; the plates were covered with breathable sealing foil for culture plates (Kisker Biotech) and placed into individual plastic boxes (Curver, New Grand Chef, 2.6 L) with holes in the lids, which were plugged with foam stoppers to allow for air circulation. To dry the diet the boxes were placed at 50 °C for approx. 17 hours. The diet for the control larvae was prepared in the same way, except flour with yeast was mixed with only PBS. Larvae were exposed to spore-containing or naïve diet for three hours and subsequently transferred to 96-well plates containing the diet without spores, at which point the survival was also monitored. Gut samples were taken six and 18 hours after the initial exposure. To do this, larvae were ice anesthetised, the first and the last segments were removed using a scalpel, and a drop of PBS was added to the sample.
The gut was carefully pulled out with a pair of forceps and the fat body was removed. The guts were washed in a droplet of clean PBS, and ten guts were pooled in 1.5 mL centrifuge tubes containing 50 µL of ice-chilled PBS and the tube was frozen in liquid nitrogen. We produced six replicates per combination of treatment and time point, whereby each replicate consisted of ten pooled animals (6 replicates x 2 treatments x 2 time points x 10 animals = 240 larvae). In addition to monitoring survival at the point of transfer from the sporecontaining to the spore-free diet, survival at six, 18 and 24 hours was monitored from additional larvae (n = minimum 480). No larvae had died by 3 or 6 hours after exposure, but at 18 hours 5.1 % B. thuringiensis-exposed and 0.3 % control larvae had died; additional mortality was noted at 24 hours (total ~ 8 %) in the B. thuringiensis group, but there was no further mortality in the control group. Before RNA extraction, the guts were homogenized over liquid nitrogen with a pestle. Further RNA isolation, cDNA synthesis and qPCR was done as described in before, with the addition of immune gene expression (Att2, Col1, and Imd). Cp values were calculated using Fit Point method with a Noise Band threshold of 6. RNAi is a powerful and well established molecular technique in this species (see 13,24). To perform knockdown of Dscam1 in T. castaneum larvae we followed the protocol of Posnien et al. (25). A non-alternatively spliced region within the Dscam1 gene, exon 15, was used for RNAi ( Figure S1, Table S1; See end of this file for the annotated T. castaneum Dscam1 gene), which did not overlap with the Dscam1 qPCR primer pair. A gene-specific primer pair µl H 2 O) for one hour at -20 ºC. After centrifugation (10,000 rcf, 30 min, 4 ºC) and washing the pellet with 70 % EtOH, the pellet was air-dried for 20 to 30 min at RT. The dsRNA was resupended in PBS and an annealing step was done where the resuspended dsRNA was incubated at 95 ºC for two min. Immediately after incubation the dsRNA was incubated in pre-boiled water until the water temperature reached 70 ºC, after which followed another incubation at 95 ºC in a thermo block. After five min the thermoblock was removed from the heating source and cooled to room temperature. The dsRNA concentration was measured using NanoDrop (NanoPhotometer TM Pearl, Implen) and stored at -80 °C until further use.

Effect of
To produce the experimental animals, two to three week old adult beetles were allowed to lay eggs for 24 hours. Eleven day old larvae were injected with dsRNA (D-ex15 RNAi : n = 256; TC RNAi : n = 256; N RNAi : n = 256) as described in (25). Briefly, the concentration of dsRNA was 2.7 µg/µl for D-ex15 RNAi and 2.5 µg/µl for TC RNAi . To inject the dsRNA we used glass capillaries that had been pulled with a micropipette puller (as mentioned before). Capillaries were filled with 10 µl dsRNA solution and approximately 300 nl of each dsRNA (ca. 0.8 -0.9 µg per larvae as recommended by Posien et al. (25)) was dorsolaterally injected between the first and second larval abdominal segment using a FemtoJet microinjector. Injected and naïve beetles were individualised into 96-well plates with flour and 5 % yeast. To test the efficiency of the knockdown in the haemocytes and the whole body, we sampled haemocytes from 2 x 10 animals and 2 x 10 whole body larvae from each injection group (D-ex15 RNAi and TC RNAi ) four days after dsRNA injection. Dscam1 expression was examined for all samples by using qPCR as described before, using Rp49 and RpL13a as reference genes and TC animals as a control group.
The B. thuringiensis concentration for pricking was 1 x 10 10 (as in 11) and 3 x 10 10 cells / mL -1 to increase mortality and potentially any differences between the knockdown treatments.
The E. coli concentration was 1 x 10 11 cells / mL -1 as used in (27 N RNAi : n = 192). The concentration of dsRNA was 2.6 µg/µl for D-ex15 RNAi and 2.9 µg/µl for TC RNAi . Injections and preparation of the dsRNA were performed as described before. The efficiency of the knockdown was tested by dissecting three pools of five guts, as described before, and by pooling 3 x 10 whole body larvae from the D-ex15 RNAi and TC RNAi injected groups four days after dsRNA injection. Dscam1 expression was examined for all samples by using qPCR as described before.
Four days after the dsRNA injections we checked larval survival from a subset of injected larvae (mortality: N RNAi : 6 %; TC RNAi : 17 %; D-ex15 RNAi : 16 %), and from each injection group 48 size-selected (34) larvae were randomly assigned to a B. thuringiensis sporecontaining diet and 48 to a naïve non-spore containing diet. The oral infections were performed as described before with the exception that the spore concentration for this experiment was higher 5 x 10 9 mL -1 , so that we would induce a higher mortality. Larvae were exposed to spore-containing and naïve discs such that 16 larvae from each treatment were assigned to one infection or control 96-well plate. This was replicated three times resulting in a sample size of 48 larvae per treatment group. Survival was monitored for 4 consecutive days. Larval survival after exposure to spores was analysed as described before. None of the larvae from the N RNAi , TC RNAi or D-ex15 RNAi groups exposed to only the spore-free diet died; because there was no event, there was no contribution to the likelihood, and a Cox mixed effect model could not be fitted. Therefore we denoted one larva as dead in each of these groups allowing us to fit the model. The fixed effect was knockdown treatment and plate was included as a random factor. However, the latter was removed from the model as it was not statistically significant and the final model with only the fixed effect was tested using Cox proportional hazards.
Life history effects of Dscam1 knockdown in T. castaneum.
To examine whether there is a fitness effect, measured via fecundity, and an adult behavioural phenotype after Dscam1 knockdown we performed two simultaneous experiments. We then performed a third experiment to further investigate aspects of fecundity. Unfertilised T.
castaneum females will lay eggs (35), therefore we could examine whether knockdown affects egg production even when the mating was unsuccessful. T. castaneum is highly polygamous -males can mate with up to seven different virgin females in 15 minutes (36).
Males reach sexual maturity approximately two days after imaginal eclosion and females after four days (35); furthermore, copulations are brief, Edvardsson and Arnqvist (37) found that on average they last for around 100 seconds. Therefore they make an ideal model with which to examine questions relating to copulation and mating success.
Experimental animals were produced as described before. For these experiments we produced a second Dscam1 knockdown treatment using exon 12 (D-ex12 RNAi ; Table S1; Figure S1), the dsRNA was produced as described before. We injected 11 day old larvae with dsRNA (D-

Behavioural tests.
Twenty five days post-knockdown we used ten males and ten females from each injection group for the behavioural assays (n total = 80). On the same day, for each of the four treatment groups, we froze three pools of two female and three pools of two male beetles for later qPCR to check the knockdown (methods as described before), except for D-ex12 RNAi , where we froze two pools of females and four pools of males. The assays were performed between 09:30 and 17:00 (daylight hours for the beetle) at room temperature and under light conditions. The assays were carried out in ten blocks, where each block contained one male and one female of each of the treatment groups, i.e. eight beetles, processed in a random order with respect to treatment and sex within each block. We noticed that at the time when beetles were removed from their glass vials for the first behavioural assay that some beetles were positioned ventral side down and others were dorsal side down in the flour, we therefore noted for each beetle whether this was the case or not. When given the opportunity, T.
castaneum has an innate response to climb. We therefore adapted an assay from Michalczyk et al. (39) to test the speed at which beetles climb vertically. The beetles were removed from their glass vials and placed on their backs in individual glass Petri dishes. The beetle was passed blind with respect to treatment to a second experimenter who offered the beetle a white strip of paper (2.5 mm wide x 100 mm long, with a pencil mark at 30 mm), such that the bottom edge of the paper was in contact with the beetle's tarsi. When the beetle gripped onto the paper, the paper was lifted up from the Petri dish by approximately 10 cm. We  we counted the number of larvae that had hatched. This is the first study to use this wild type stock population, Cro1, in mating assays, and as such we note that the hatching rate of eggs (~ 80 %) estimated from control (TC RNAi x TC RNAi ) pairings is comparable to wild type T.
castaneum hatching rates that have been found in other populations (35,43). Egg cannibalism is known to occur in adult T. castaneum (e.g., 35, 44) therefore we cannot exclude the possibility that it might have occurred during the 12-day adult pairing period and affected some of our egg counts. We removed the eggs from the adult pair every three days (the same frequency was used for fecundity analyses by, e.g., 42), meaning that the resulting cohorts of developing larvae were synchronised, which is important because older larvae have been found to have higher egg cannibalism rates than younger larvae (45).
Female survival was analysed as described above, using Cox proportional hazards. The model was fitted with day of death as the response variable. The fixed effect was the combination of the female and male knockdown treatments giving a seven-level factor). There were no female deaths in the TC RNAi x D-ex12 RNAi group, therefore for the same rationale as given laid eggs, we therefore only tested whether the TC females differed significantly in the number of eggs that they laid depending upon who they had been paired with. The data were normally distributed and had equal variances so we performed an ANOVA using JMP. Apart from two D-ex12 RNAi knockdown females paired with TC males, larvae only hatched from the pairings between TC RNAi treated females and males, we therefore did not statistically analyse this data set.
Mating behaviour and physiology.
In the fecundity experiment above the females and males were placed together continuously and no observations were made of mating behaviour. In this experiment we therefore aimed to examine whether knockdown females and males mate and also whether female knockdown beetles show evidence of reduced ovaries. Experimental animals were produced as described before. We   (37), we left the pair together. In these cases where there were multiple attempts by the male to mate the female, we observed them until there was a 10 minute period with no interaction and then separated them. With the exception of one TC RNAi x TC RNAi pairing that had two mating attempts, multiple attempts to mate always involved a knockdown female paired with a TC RNAi or N RNAi male. The pairs were given a maximum of one hour in which to mate, after this time they were separated and females were placed in individual plastic vials containing 4 g of pre-sieved flour plus 5 % yeast and kept in controlled conditions as described above. The flour was sieved after three days and the number of eggs was counted.
The eggs were kept for ten days and after this time the numbers of larvae were counted. The ovaries of twenty females were dissected on the same day that the eggs were counted; they were then photographed with a Canon EOS 5D Mark II under a dissecting microscope. We present the data from this experiment as descriptive because the responses measured were mostly binary and because of the relatively low sample sizes.