Flight performance of actively foraging honey bees is reduced by a common pathogen

Summary Sudden and severe declines in honey bee (Apis mellifera) colony health in the US and Europe have been attributed, in part, to emergent microbial pathogens, however, the mechanisms behind the impact are unclear. Using roundabout flight mills, we measured the flight distance and duration of actively foraging, healthy‐looking honey bees sampled from standard colonies, before quantifying the level of infection by Nosema ceranae and Deformed Wing Virus complex (DWV) for each bee. Neither the presence nor the quantity of N. ceranae were at low, natural levels of infection had any effect on flight distance or duration, but presence of DWV reduced flight distance by two thirds and duration by one half. Quantity of DWV was shown to have a significant, but weakly positive relation with flight distance and duration, however, the low amount of variation that was accounted for suggests further investigation by dose‐response assays is required. We conclude that widespread, naturally occurring levels of infection by DWV weaken the flight ability of honey bees and high levels of within‐colony prevalence are likely to reduce efficiency and increase the cost of resource acquisition. Predictions of implications of pathogens on colony health and function should take account of sublethal effects on flight performance.


Introduction
Decadal and on-going declines in the number of colonies of managed honey bees in the USA and Europe have been well documented and have been attributed to a number of stress factors (vanEngelsdorp and Meixner, 2010;Lee et al., 2015) that include pests and pathogens, pesticides and limited quality and availability of food resource (Klein et al., 2007;Neumann and Carreck, 2010;Potts et al., 2010;Becher et al., 2014;Goulson et al., 2015). These stressors interact with individual bees, resulting in lethal and sublethal effects that curtail longevity (Alaux et al., 2010;Aufauvre et al., 2012;Doublet et al., 2015;Retschnig et al., 2015) and alter fitness traits and behavioural and physiological performance, having implications for the entire colony (Becher et al., 2014;Rumkee et al., 2015). Pathogens affect behaviour directly through active manipulation evolved to facilitate transmission, although this is yet to be demonstrated in honey bees (see , and indirectly in response to an associated increase in the host's metabolic rate (Mayack and Naug, 2009;Naug and Gibbs, 2009;Mayack and Naug, 2015) or manipulating hormonal pathways .
Although living in social groups has fitness benefits (Wilson, 1975), one of the trade-offs is the increased risk of disease transmission because of close living quarters and high genetic relatedness (Schmid-Hempel, 1998;Tarpy, 2003). Honey bee colonies comprise thousands of individuals living in close contact and predictably, pests and pathogens are wide-spread and commonly occurring therein (Mouret et al., 2013;Manley et al., 2015; and have been implicated in honey bee colony losses in the U.S. and Europe (Higes et al., 2008;vanEngelsdorp and Meixner, 2010). Deformed wing virus (DWV), Varroa destructor virus-1 (VDV-1) and Nosema ceranae (Fries) are three of the most prevalent pathogens present in European honey bee colonies (Martin-Hernandez et al., 2007;Mouret et al., 2013;. The DWV complex (referred to henceforth as DWV) is a rapidly evolving and recombining group of closely related positive-sense, single-stranded RNA Iflaviruses, that includes VDV-1 (de Miranda and Genersch, 2010;Moore et al., 2011;Zioni et al., 2011;Martin et al., 2012;Mordecai et al., 2016). DWV is vectored by the parasitic mite, Varroa destructor Anderson & Truman, (Martin et al., 2012) and is transmitted both horizontally (faecalcannibal-oral) (Yue and Genersch, 2005;Mockel et al., 2011) and vertically (parent-offspring) (Chen et al., 2006;Yue et al., 2006;Yue et al., 2007;de Miranda and Fries, 2008;Yanez et al., 2012). Clinically relevant infections by DWV, defined as presence of DWV RNA in the brain (Mockel et al., 2011), do not always result in bees exhibiting a phenotype (deformed wings) (de Miranda and Genersch, 2010;Zioni et al., 2011). The microsporidian gut parasite, N. ceranae, historically a parasite of A. cerana, now includes A. mellifera as an alternative host (Higes et al., 2006) and causes no visible, external symptoms of infection. N. ceranae infects and reproduces inside epithelial cells of the midgut and is believed to be transmitted in the hive principally via the oral-oral pathway (Smith, 2012).
Prevalence and diversity of disease pathogens in honey bee colonies are probably greater than previously thought (Tentcheva et al., 2004;Siede et al., 2008;Furst et al., 2014;, since infection is often inapparent (Zioni et al., 2011;Mouret et al., 2013) or below the level of detection (Martin et al., 2013a). It is unsurprising, therefore, that research into the sublethal effects of commonly occurring pathogens on honey bee behaviour is limited, yet of increasing interest. Other than understanding transmission (Bowen-Walker et al., 1999;Yanez et al., 2012;Manley et al., 2015), and influence on gene expression (Steinmann et al., 2015), physiology (Yang and Cox-Foster, 2007) and learning (Iqbal and Mueller, 2007) of DWV, research into sublethal behavioural effects of pathogens has largely been limited to N. ceranae. This gut parasite has been shown to modify many aspects of honey bee behaviour, including increased maturation (Dussaubat et al., 2013;Goblirsch et al., 2013), impaired learning (Mallon et al., 2003;Kralj and Fuchs, 2010), enhanced energetic stress (Mayack and Naug, 2009;2010;Mayack and Naug, 2015) and changes to flight and homing behaviour (Alaux et al., 2014;Naug, 2014;Wolf et al., 2014;Perry et al., 2015). N. ceranae has previously been shown to increase the number of foraging trips and flight duration, reduce the time spent in the hive (Alaux et al., 2014;Naug, 2014;Retschnig et al., 2015) and reduce foraging efficiency (Naug, 2014), but since individuals were not tracked once they had left the hive, the proportion of the time spent flying or resting was unknown. In contrast, exploring whether DWV affects flight behaviour in bees that do not exhibit the visual symptoms of deformed wings typical of high levels of infection (de Miranda and Genersch, 2010), yet may already be suffering altered physiological, neurological or immunological function remains to be done.
Flight performance of an individual can determine its potential resource-gathering capability and in social insects, efficient resource acquisition can have profound effects at the colony level (Becher et al., 2014). The farther the distance and longer the duration an individual is able to travel allows more of the landscape to be exploited for resources. Understanding flight performance of foraging honey bees challenged by pathogens is therefore of key importance not only for effective colony management, but also for protecting pollination service provision (Potts et al., 2010) by managed and wild species. Measuring flight performance of individual bees is notoriously difficult though; bees are small, fast flyers covering vast foraging areas. Tracking individuals using harmonic radar (Riley et al., 1996) is currently the only technology available to record bee flight routes in the field, however, it is not possible to simultaneously track two or more individuals to estimate the spatial and temporal limits of bee flight under similar environmental conditions. In contrast, assessing flight performance using tethered individuals on flight mills provides an elegant opportunity to explore individual endurance limits allowing maximal control of environmental factors other than pathogen load (Brodschneider et al., 2009).
Here, we sought to test the null hypothesis that natural levels of infection by N. ceranae and the DWV complex (comprising DWV and VDV-1) in forager honey bees have no effect on flight performance. Thus, the aims of this work were to (i) quantify the natural levels of infection by N. ceranae and DWV 1 VDV-1 in actively foraging, apparently healthy honey bees and, (ii) elucidate the sublethal effects of these commonly occurring pathogens on flight distance and duration.

Results and discussion
Of 127 bees that were analysed, 73 tested positive for one of the two pathogens that were screened, 20 tested positive for both pathogens and 34 tested negative for neither pathogen. DWV was more prevalent (83 bees) than N. ceranae (30 bees) and the level of co-infection was lower (20 bees) than for single infection (N. ceranae 10 bees; DWV 63 bees). Of the bees that tested positive for DWV and N. ceranae, mean loads were 3.6 3 10 10 6 SD 1.8 3 10 11 copies head 21 and 1.7 3 10 4 6 SD 2.1 3 10 4 mid-gut 21 , respectively. The levels of infection we recorded are comparable to those reported elsewhere for standard, managed apiaries (Gauthier et al., 2007;Steinmann et al., 2015).
Tethered flight mills have been successfully used to measure the relative flight performance of different taxa under controlled, standardised conditions (Riley et al., 1997;Blackmer et al., 2004;Spiewok and Schmolz, 2006;Brodschneider et al., 2009;Taylor et al., 2010; Flight performance is reduced by a common pathogen 729  Dorhout et al., 2011;Jones et al., 2015), however it is important to recognise the limitations of this experimental technique. Flight mills restrict the physical and biophysical dynamics of natural flight, where reduced drag and a lack of need to produce uplift have been shown to result in lower levels of expended energy than are readily available (Riley et al., 1997) which could result in enhanced measures of flight performance than are possible when insects are in free flight. Equally, a lack of stimuli, such as olfactory cues from sources of forage, to initiate and sustain flight behaviour, could result in reduced measures of flight performance. Despite these differences to natural free-flight conditions, tethered flight mills remain an important instrument for measuring the relative flight performance of worker honey bees, given the assumption that handling, tethering and restriction of natural cues affect the behaviour and performance of the test bees equally.
Contrary to our expectation and the findings of previous work (e.g. Alaux et al., 2014;Naug, 2014;Wolf et al., 2014), we found no effect of either presence of N. ceranae on flight distance (F 1, 121.5 5 0.71, P 5 0.400) or duration (F 1, 121.9 5 1.39, P 5 0.240) or the amount of N. ceranae on flight parameters (Table 1). Honey bees exclusively use carbohydrates to power flight activity (Sacktor, 1970;Rothe and Nachtigall, 1989) that accounts for 30% of the total energy expenditure of a forager bee (Harrison and Fewell, 2002). In this study, bees were flown to exhaustion before being fed a known and finite amount of energy in the sucrose meal (c.f. Gmeinbauer and Crailsheim, 1993;Hanauer-Thieser and Nachtigall, 1995;Brodschneider et al., 2009) that fuels the subsequent test flight. The assumption is, therefore, that bees flown to exhaustion have no remaining energy reserves available to them (Gmeinbauer and Crailsheim, 1993). As an obligate gut parasite without mitochondria, Nosema species have been shown to cause energetic stress by reducing the amount of energy available to an infected bee. Trehalose, which is synthesised in invertebrate haemolymph from dietary sucrose and used for the rapid release of energy used in flight (Thompson, 2003), is decreased in bees naturally infected with Nosema and is thought to lead to significant decreases in flying ability (Mayack and Naug, 2010). In response to Nosema-induced energetic stress, infected honey bees consume more energy-rich food (Mayack and Naug, 2009;Martin-Hernandez et al., 2011) and reduce food-sharing with nest-mates (Naug and Gibbs, 2009). Indeed, Nosema-induced energetic limitations have been suggested as an underlying mechanism behind the increased likelihood of failure of foragers to return to the hive (Wolf et al., 2014), increased periods of time spent on foraging trips (Kralj and Fuchs, 2010;Alaux et al., 2014;Naug, 2014), and increased number of foraging trips (Dussaubat et al., 2013). However, the different experimental approaches of these studies may explain the apparent conflict in our results that naturally occurring, low levels of infection by Nosema have no effect on flight performance. Firstly, the studies did not directly measure flight duration and distance of individuals, rather they measured time spent outside the colony and were unable to distinguish between bee movement (flight) and resting. Secondly, the studies did not administer known quantities of energy prior to measuring flight activity, and so they were unable to determine the effects of Nosema-induced energetic stress on honey bee flight. Thirdly, and perhaps most significantly, in some of the previous work, bees were inoculated with Nosema spores (Kralj and Fuchs, 2010;Dussaubat et al., 2013;Alaux et al., 2014;Naug, 2014;Wolf et al., 2014) that resulted in spore loads in the whole abdomen (Alaux et al., 2014) and mid-gut (Dussaubat et al., 2013;Wolf et al., 2014) orders of magnitude greater than their controls, while the natural level of Nosema infection we report here was lower than the control groups (Dussaubat et al., 2013;Alaux et al., 2014). Finally, whilst we recognise that we tested for just two pathogens in our experiment, screening exclusively for Nosema in some previous work excluded other pathogens that may have influenced flight behaviour (Mayack and Naug, 2009;Naug and Gibbs, 2009;Kralj and Fuchs, 2010;Martin-Hernandez et al., 2011;Dussaubat et al., 2013;Alaux et al., 2014).
Only 10 of the bees that were flown subsequently screened positive exclusively for N. ceranae, leading, retrospectively, to low statistical power (32%) for detecting differences in flight performance of Nosema-infected and uninfected bees of the magnitude reported by Naug (2014) for inoculated bees. However, 99% confidence intervals for the observed ratios of flight durations for the two groups, based on either all bees flown or only on bees uninfected with DWV (see Supporting Information Appendix S1), did not contain values as extreme as the halving of flight duration for uninfected relative to Nosema-infected bees that was reported by Naug (2014). Thus, we were unable to reject the null hypotheses of no statistical differences in flight performance between groups of bees uninfected and infected by low levels of N. ceranae, but we do not preclude or dismiss previously reported effects at higher levels of infection Bees that were infected by DWV, and yet presented no obvious morphological symptoms of infection, flew shorter distances (F 1, 121.4 5 10.17, P 5 0.002) and durations (F 1, 117.1 5 9.08, P 5 0.003) than bees uninfected with DWV (Figs. 1A and B DWV is transmitted to honey bees horizontally and vertically within the hive (Chen et al., 2006;Yue et al., 2006;Yue et al., 2007;de Miranda and Fries, 2008;Yanez et al., 2012) and is vectored by V. destructor, parasitising adults, larvae and pupae (Yang and Cox-Foster, 2007;Gisder et al., 2009;Mockel et al., 2011). V. destructor, and by association, DWV, have been implicated in disrupting immunological responses and behaviour in asymptomatic honey bees. It has previously been shown that genes for protein repair and the labelling of protein for degradation were up-regulated in pupae that were parasitized by V. destructor, while genes involved in wing development processes were down-regulated (Navajas et al., 2008) suggesting disruption of larval and adult development. Furthermore, parasitism by the varroa mite of young adult worker bees with normal wings inhibited protein metabolism, energy production and expression of immune genes (Yang and Cox-Foster, 2005;Alaux et al., 2011) and reduced longevity (Yang and Cox-Foster, 2007). Significantly, studies have successfully linked varroa mite parasitism with the direct effects of infection by DWV, providing evidence for DWVinduced reduced immune gene expression (Steinmann et al., 2015), impaired associative olfactory learning and memory formation (Iqbal and Mueller, 2007). Thus, it is clear that there are diverse effects of sublethal infection by DWV on honey bees. Our data suggest, for the first time, that DWV may affect another important behavioural function, in reducing flight performance. The mechanisms behind these reductions in flight duration and distance are, as yet unclear, but it is possible that the disruption in expression of genes associated with protein metabolism, energy production and internal wing development may reduce the physical fitness characters of the forager bees. Another explanation for reduced flight performance in bees infected with DWV may be related to pathogen-induced accelerated behavioural development. Enhanced behavioural maturation from in-hive to forager bees has been observed in bees infected with N. ceranae (Dussaubat et al., 2013;Goblirsch et al., 2013; and reduced flight performance has been recorded for forager bees from colonies exposed to high levels of the DWV-vector, V. destructor (Blanken et al., 2015). Blanken et al. (2015) found that increased body mass, a character associated with precocious foragers (Vance et al., 2009)  honey bees and those infected with DWV or N. ceranae suggest that any provocation of precocious foraging occurs because of self-removal from the colony, as a form of social immunity (Meunier, 2015). We are unable to confirm an association between body mass, and thence DWV-induced precocious foraging in our experiment, however our finding that there was no direct effect of DWV load on wing size (F 1, 17 5 0.005, P 5 0.945; N 5 19), infers this may be a plausible explanation that warrants further investigation.
Predictive models including only DWV indicate a weakly positive relationship between amount of DWV and flight distance (regression coefficient 5 0.060, SE 5 0.0199, P 5 0.003) and duration (regression coefficient 5 0.039, SE 5 0.0132, P 5 0.004) ( Table 1). Whilst these relationships were significant, the models only accounted for 8.6% and 7.1% of the variation in distance and duration, respectively ( Fig. 2A and B). This perplexing result requires further investigation, not least because so little of the variation was accounted for by DWV in the models, but also because the weak relationship over the range of levels of infection predicts such small increases in distance and time travelled (regression coefficients of 0.06 and 0.04), which cannot be considered to be biologically significant. Conducting doseresponse assays of DWV and N. ceranae on flight behaviour and genomic response are prime areas of future research.
Co-infection by the pathogens in this experiment occurred in 15.7% of the bees we tested and in agreement with Martin et al. (2013b), we found no association between the presence and absence of N. ceranae and DWV (Pearson v 2 5 0.03, df 5 1, P 5 0.863) and nor were there interactions between the two pathogens and either distance (F 1, 121.6 5 1.35, P 5 0.248) or duration (F 1, 122.1 5 1.08, P 5 0.301). It is unlikely, then, that there were confounding effects of these pathogens on flight behaviour in this experiment.
In conclusion, the bees tested here were representative of colonies with natural levels of N. ceranae and DWV infection, where inapparent, but clinically relevant infection by DWV was shown to reduce the distance and duration of flights in forager bees. If the reduced flight abilities we recorded on the tethered flight mills operate in the field under natural, free-flight conditions, it is likely that sublethal effects of DWV infection are more widespread than previously thought. Indeed, recorded Dainat et al., 2012) and predicted (Kielmanowicz et al., 2015) over-winter colony losses have been attributed to natural levels of DWV infection. Possible consequences of reduced flight endurance per unit of energy include less efficient and more costly resource acquisition for the individual and for the colony, particularly in landscapes where forage resources are spatio-temporally sparsely and patchily distributed, and enhanced risk of premature death because of increased exposure to predators and physiological fatigue. Predicting and scaling the implications of our findings under artificial conditions to colonies in natural field conditions where DWV is persistently and highly prevalent will require further studies to understand behavioural responses to pathogen-mediated compromises in flight performance, including for example, whether there are trade-offs between resource utilised by the bee and that contributed to the colony.

Honey bees
Returning, actively foraging honey bees with no visible signs of disease (dysentery or malformed wings) and carrying corbicular pollen were randomly selected on the morning of the flight test from five, conventionally managed apiaries, comprising 14 colonies, within 10 km of Rothamsted Research, UK (51848 0 28.83 00 N, 000822 0 31.58 00 W) in July, August and September 2012 and August and September 2013. The bees were placed into hoarding cages (Williams et al., 2013) with free access to 1M sucrose syrup and water and then placed in an incubator set at 328C to allow preparation of the bees for flight performance testing on the flight mill.

Flight performance
We recorded flight distance and duration as two measures of flight performance using roundabout flight mills, similar to those used in many studies to characterise insect flight ability (Bradley and Altizer, 2005;Brodschneider et al., 2009;Dorhout et al., 2011;Sappington and Burks, 2014). A set of five flight mills, that consisted of a lightweight arm suspended at the centre by two magnets forming an almost resistance-free axis (see Chapman et al., 2015) and surrounded by equally sized and spaced monochrome vertical stripes surround each flight mill to provide the illusion of movement (Hrassnigg and Crailsheim, 1999), were located in a controlled environment room set at 248C with constant overhead lighting. Test bees had an Opalith disc attached to the thorax (Human et al., 2013), before being allowed to rest in the hoarding cage for 45 minutes before being tested on the mill. Immediately prior to connecting to the flight mill, an attachment (15 mm x 1 mm) was glued (Evo-Stik impact multipurpose adhesive) onto the Opalith disc. The bee, holding a small ball of paper between her legs as a stimulus for spontaneous flight (Brodschneider et al., 2009;Sappington and Burks, 2014), and complete with attachment, was then connected to the flight mill arm and a counter weight of similar mass was attached to the other end of the arm. The flight mill allows the bee to fly in a circular trajectory, with a circumference of 1m, and the embedded microcontroller board records the total distance and duration flown by the bee at 5 s intervals to the nearest 20 cm.
To control the amount of energy available for flight, each bee was allowed to fly to exhaustion in order to deplete the sugar reserves in the honey stomach prior to being fed a known amount of energy (Gmeinbauer and Crailsheim, 1993;Brodschneider et al., 2009). The exhaustion flight was completed when, despite being stimulated, the bee did not recommence flying for more than 30 s. The bee was removed from the mill, fed 10 ml of 1M sucrose solution using a pipette before being reattached to the mill for flight performance testing until the bee again ceased flying because of lack of energy. The bee was removed from the flight mill, placed in an Eppendorf tube and stored at 2808C. Test flights were terminated when, despite stimulation following a pause in flight, bees did not resume flight.

Disease analysis
Nosema spore counts midgut 21 were determined microscopically using a Neubauer improved 5x5 haemocytometer following the methods of Human et al. (2013). The digestive tract was removed from the bee, and the midgut was isolated and homogenised in 500 ml of distilled water using a micropestle. Nosema spores were counted in four haemocytometer chambers and the total number of spores in a volume of 0.28 mm 2 was counted per chamber. To confirm species identification (N. ceranae or N. apis), spores were identified using species-specific PCR (Fries et al., 2013;Wolf et al., 2014).
The presence of DWV-complex RNA in the brains of honey bees with apparently normal wings indicates clinically relevant, overt infection by DWV Mockel et al., 2011), so we quantified absolute copy numbers of positive-strand DWV and VDV-1 in the heads of the test bees. Each head was homogenised in 600 ml of lysis buffer (RLT Buffer, Qiagen, Manchester, UK) with 1% b-mercaptoethanol (Qiagen, Manchester, UK). RNA was extracted from this supernatant using the RNeasy Mini kit affinity column purification (Qiagen, Manchester, UK) in a QIAcube robot (Qiagen, Manchester, UK) and quantified using an Epoch microplate spectrophotometer (BioTek, Swindon, UK). Total cDNA was synthesized cDNA from 800 ng RNA using M-MLV reverse transcriptase (Promega). For absolute quantification, duplicate qRT-PCR was performed for each sample using SYBRgreen Sensimix (Bioline, Luckenwalde, Germany) in the following program: 5 min at 958C, followed by 40 cycles of 10 s at 958C, 30 s at 578C, and 30 s at 728C (read) and data were normalised using the honey bee reference gene RP49 (Lourenco et al., 2008). An RNA-free HPLC-water and a virus-positive sample cDNA were run as negative and positive control respectively in each reaction run. Following PCR, DNA was denatured for 1 min at 958C and cooled to 558C for 1 min. A melting profile was generated from 558C to 958C (0-58C s 21 increments). Absolute quantification of DWV and VDV-1 was calculated using duplicate DNA standard curves of purified flanking PCR products with efficiencies between 90% and 100% and correlation coefficients (R 2 ) from 0.990 to 0.999. To account for Flight performance is reduced by a common pathogen 733 potential variation in sample quality, an upper cycle threshold (Ct) of 35 was set for RP49, above which samples were not included in quantitative analysis (Blanchard et al., 2007;de Miranda et al., 2013). The primers used were: DWV: DWV-F2, DWV-R2a; VDV: VDV-F2, VDV-R2a (McMahon et al., 2015).

Wing size
Acute DWV infection frequently results in, amongst other symptoms, malformed wings that render bees incapable of flight. In order to estimate effects of pathogen on wing size and subsequent flight performance, wings were removed from a subsample of the bees stored for pathogen analyses. Both forewings from 19 bees were mounted between two microscope slides and scanned (4800 dpi) before measuring the combined lengths of three wing venation characters as surrogate measures of wing size (D2, D3, D7, after Dedej and Nazzi, 1994;Jaffe and Moritz, 2010). Each character on both wings was measured three times, to account for measurement error, giving a mean total length of the characters per bee.

Statistical analysis
The flights of 476 bees were assessed on the mills. Since we sought to test the effect of pathogens on the range of flight abilities of apparently healthy bees recorded on the flight mill, nonflyers (N 5 31) were excluded from the analysis. Laboratory constraints necessitated the creation of a subset of individuals from the remaining 445 bees for subsequent disease analysis. Therefore, bees classified as strong or reluctant flyers (test flight duration greater or less than 10 min, respectively), were matched, as closely as possible, by test date and colony. This process created a pooled subset of 127 bees for which the effects of presence and amount of pathogen load on flight performance (Supporting Information Table S1) were analysed.
The disease variables were skewed and so were transformed to logarithms (base 10) after adding an offset of 0.01 to allow for the absence of pathogens. The measures of flight performance (distance and duration) were also logged (base 10) to achieve homogeneity of variances and normality.
We first tested whether disease status (based on a 2 3 2 factorial treatment structure representing presence and absence of each of the two pathogens) affects flight distance or duration using a linear mixed model fitted using restricted maximum-likelihood (REML), with flight mill included as a random effect.
We then tested, for diseased bees only, the relationship between measures of flight performance and quantitative pathogen load, using an intra-block regression approach (Welham et al., 2015). Initial analyses with calendar month included in the random model suggested negligible temporal effects hence the random model was subsequently simplified to include flight mill effects only. Block effects (flight mill) were fitted before pathogen terms, the latter fitted first individually, then together. Statistically nonsignificant fixed effects were dropped and the resulting parsimonious models were used to obtain predictions. Finally, we analysed the effect of DWV load on wing length (transformed to logarithms (base 10) after adding an offset of 0.01) using simple linear regression. All analyses were done using GenStat 17 (VSNI, 2014).