Microparasite manipulation of an insect: the influence of the egt gene on the interaction between a baculovirus and its lepidopteran host

Authors

  • J. S. CORY,

    Corresponding author
    1. Ecology and Biocontrol Group, NERC Centre for Ecology and Hydrology, Mansfield Road, Oxford OX1 3SR, UK
      †Author to whom correspondence should be addressed.
    Search for more papers by this author
  • E. E. CLARKE,

    1. Ecology and Biocontrol Group, NERC Centre for Ecology and Hydrology, Mansfield Road, Oxford OX1 3SR, UK
    2. Department of Biological Sciences, Imperial College of Science, Technology and Medicine, London SW7 2AZ, UK
    Search for more papers by this author
  • M. L. BROWN,

    1. Ecology and Biocontrol Group, NERC Centre for Ecology and Hydrology, Mansfield Road, Oxford OX1 3SR, UK
    Search for more papers by this author
  • R. S. HAILS,

    1. Ecology and Biocontrol Group, NERC Centre for Ecology and Hydrology, Mansfield Road, Oxford OX1 3SR, UK
    Search for more papers by this author
  • D. R. O'REILLY

    1. Department of Biological Sciences, Imperial College of Science, Technology and Medicine, London SW7 2AZ, UK
    Search for more papers by this author
    • §

      Present address: Syngenta, Jealotts Hill International Research Station, Bracknell RG42 6EY, UK.


  • Present address: Laboratory of Virology, Binnenhaven II, Wageningen University, 6709 PD, Wageningen, The Netherlands.

    E-mail: jennycory@tiscali.co.uk

†Author to whom correspondence should be addressed.

Summary

  • 1Parasites and pathogens manipulate their hosts in a variety of ways that are thought to enhance their fitness. However, it is rare to be able to link such phenotypic changes to specific genes.
  • 2Here the effect of a single pathogen gene is examined. The ecdysteroid UDP-glucosyltransferase (egt) gene of insect baculoviruses produces an enzyme that interferes with host moulting.
  • 3The effect of the egt gene was examined by comparing two baculoviruses that differed only in the expression of this gene. All three fitness traits examined – pathogenicity, infection duration and pathogen productivity – were affected by deletion of the gene.
  • 4Trichoplusia ni larvae in all five instars died earlier when infected with the egt-minus virus compared with those infected by the wild-type Autographa californica nucleopolyhedrovirus.
  • 5Unexpectedly, the egt-minus virus was more pathogenic to final instar larvae than the wild-type virus. Virus genotype and dose both influenced insect development.
  • 6Wild-type infected insects had a significantly higher yield of virus at death, cadaver weight and yield of virus per unit weight than those infected with the egt-minus virus.
  • 7The size of the virus challenge had a major influence on the outcome of the interaction. The consequences of these data for pathogen fitness are discussed.

Introduction

It is well documented that parasites can manipulate their hosts in a variety of ways that are thought to enhance their fitness. These effects can exert themselves through behavioural, physiological and morphological changes to the host whereby the parasite modifies endocrine, biochemical or immunological pathways (Moore 2002). The best-studied systems are those involving vertebrate macroparasites with complex life cycles, although there is a significant body of data on insects and their parasitoids (e.g. Edwards & Weaver 2001). In comparison, insects and their pathogens (microparasites) have been little studied (Cory & Myers 2003).

Behavioural modifications by parasites and pathogens have received the most attention as these changes are often quite marked and host behaviour can have a significant and direct influence on transmission to other susceptible hosts. There are several examples of insect pathogens that appear to manipulate the behaviour of their hosts to increase the likelihood of transmission (although transmission itself has rarely been estimated). For example, fungal infection is known to induce a wide range of insect hosts (flies, aphids, ants, grasshoppers) to adopt a posture with elevated wings when spores are due to be released (Hajek & St. Leger 1994). This is often accompanied by elevation-seeking behaviour. Similarly, lepidopteran larvae infected with baculoviruses often increase their activity and climb to the tips of the leaves and branches before dying, presumably to enhance dissemination to the foliage below (e.g. Vasconcelos et al. 1996).

Most studies on host manipulation by macro- and microparasites have focused on phenotypic alterations; few studies have been able to identify a mechanism or genetic basis for the observed changes nor to study genetic effects by gene deletion. In insect baculoviruses a gene has been identified, the ecdysteroid UDP-glucosyltransferase (egt) gene, which alters host development and thereby influences the rate at which the baculovirus kills its host (O’Reilly & Miller 1991). This in turn influences the productivity of the virus (Slavicek, Popham & Riegel 1999; Wilson et al. 2000). Baculoviruses produce an infective stage that can survive outside the host's body, the occlusion body (OB). They only infect larval stages and infection frequently results in the death of the host. In Lepidoptera, most tissues of the larval body are converted to virus; in late instar noctuid larvae this can result in the production of over 109 OBs per larva (e.g. Hodgson et al. 2001). However, the vast majority of the OBs are not released until the insect dies, so transmission is delayed until this point.

The egt gene was first identified in the NPV of the Alfalfa Looper, Autographa californica (AcMNPV) (O’Reilly & Miller 1989). The gene was present in a region of the genome that was frequently deleted when AcMNPV was passaged in cell culture (Kumar & Miller 1987), indicating that egt is not needed for in vitro replication. However, the egt gene is retained during larval to larval passage. Additionally more recent studies have shown that the egt gene is found in all but one lepidopteran baculovirus for which the complete sequence is known (Herniou et al. 2001). This implies that the egt gene plays a key role at the organismal level. The enzyme produced by this gene, EGT, conjugates ecdysteroids with sugars (UDP-glucose or UDP-galactose) (O’Reilly & Miller 1989; Kelly et al. 1995; Clarke et al. 1996). Ecdysteroids (via interaction with juvenile hormone) play a critical role in the control of metamorphosis in insects, in particular, the initiation of moulting, the inhibition of larval-specific genes and the transcription of pupal-specific genes. EGT can conjugate all of the main hormones involved in ecdysteroid metabolism; 3-dehydroecdysone, ecdysone, and the most active form of the hormone in vivo, 20-hydroxyecdysone. However, the prohormone ecdysone appears to be its preferred substrate (Evans & O’Reilly 1998).

In baculoviruses it is possible to delete specific genes so it is feasible to examine the influence of the egt gene directly. Earlier studies indicated that viral expression of EGT interfered with the larval moulting process. In the absence of the egt gene, insects infected with the virus followed the development of untreated control insects, whereas those treated with wild-type AcMNPV did not undergo larval–larval or larval–pupal moults, resulting in earlier death and reduced virus yield (O’Reilly & Miller 1991). This was the first direct evidence that baculoviruses altered the development of their insect hosts in ways that were likely to enhance their fitness and also indicated the genetic basis for these changes. Studies have since shown that the deletion of egt gene function in several baculovirus species results in a more rapid speed of kill but this does not occur in all situations (Cory et al. 2001). The influence of the egt gene on virus productivity has received little attention, but in the two species examined insects infected with an egt-minus virus usually produced fewer virus OBs (Slavicek et al. 1999; Wilson et al. 2000). These results also indicated that factors such as age (instar) of the insect, time of infection within the instar and virus dose could all be important in determining whether and to what extent EGT expression impacts on host development and virus fitness.

We have been carrying out a detailed investigation of the range of phenotypic effects produced by the baculovirus egt gene using the Cabbage Looper, Trichoplusia ni (Hübner), and AcMNPV as a model system. Initial experiments demonstrated that deletion of the AcMNPV egt gene reduced speed of kill in two larval instars (second and fourth) of T. ni and also reduced virus yield in the fourth instar larvae (Wilson et al. 2000). In terms of virus transmission in the field, effects on final instar larva are likely to be critical as these produce the greatest yield and the potential reservoir of overwintering inoculum. Additionally, final instar larvae undergo endocrinological and behavioural changes in readiness for pupation, which might have an adverse effect on virus replication or virus transmission. Here we examine three issues. Firstly, we address whether the egt gene exerts an effect on all larval instars by measuring virus speed of kill. Secondly, we carry out a detailed examination of final instar T. ni larvae to ascertain whether the egt gene influences pathogenicity, infection duration and virus production and the possible mechanisms behind these changes. Finally, we extend earlier studies by investigating the effect of virus dose on changes mediated by EGT expression.

Materials and methods

insects

The Cabbage Looper, Trichoplusia ni, larvae used in all experiments originated from a culture maintained at the Centre for Ecology and Hydrology, Oxford. Insects were maintained at 25 °C ± 1 °C on a 16 : 8 h light/dark cycle at 30–40% RH. Adults were kept in insect breeding cages in an approximately equal sex ratio and fed on 10% v/v solution of honey. Eggs were laid on strips of filter paper which were collected three times a week and surface-sterilized in 10% formalin for 30 min. Larvae were reared communally on artificial diet (Hunter, Crook & Entwistle 1984) in plastic sandwich boxes (27·5 cm × 15·5 cm × 9 cm deep). To standardize the size and age of insects within an instar for bioassays, we collected larvae that had moulted over a specified period (12 h) and reared them through together. Numbers were reduced for fourth and fifth instar larvae to avoid cannibalism.

viruses

The wild-type AcMNPV strain used was the L1 clone. We compared this virus with a derivative of L1, vEGTDEL, with a deletion of 1094 base-pairs within the egt gene (O’Reilly et al. 1991). Both viruses were amplified in third or fourth instar T. ni by inoculating the insects with 104 OBs and then rearing them individually until death. Infected larval cadavers were macerated in sterile distilled water and filtered through a double layer of muslin, keeping the macerate on ice to reduce melanization. After low speed centrifugation at 400 g to remove larval debris, the virus was pelleted and then purified using a discontinuous (50%/60% w/w) sucrose gradient at 90 000 g for 2 h. The virus band that formed at the interface was removed and washed and pelleted twice. The concentration of OBs was estimated using a haemocytometer (Improved Neubauer 0·01 mm depth, BS 748, Weber, Teddington) and then the virus was stored at −20 °C until required.

bioassays

Experiment 1. The influence of larval stage on the action of the egt gene

We collected larvae 24 h prior to setting up the bioassay and starved them overnight. First, second and third instar T. ni were infected by droplet feeding; fourth and fifth instar larvae were individually fed a droplet of virus suspension. For droplet feeding the virus suspension is coloured using blue food dye; drops of virus suspension are placed on a hydrophobic surface (Parafilm) and the insects are left to imbibe the virus for 30 min. Larvae that have ingested the virus are easily identified by the blue coloration in their gut. We used these methods of treatment because they ensured that the virus dose was ingested over a very short period, allowing time to death to be recorded with precision. Larvae in the first three instars were fed the same concentration of virus (2 × 106 OBs/ml). As larger larvae (later instars) will ingest a greater volume, this means that the virus dose ingested increased with increasing instar. Fourth and fifth instars were fed with 1 µl of virus suspension containing 2000 OBs. Up to 50 insects were treated per assay and each bioassay was replicated three times.

Insects were reared individually at 24 °C ± 1 °C in 30-ml plastic pots each containing approximately 4·5 ml of diet. The larvae were checked after 24 h; any deaths at this stage were assumed to be due to handling and were removed. We checked the insects for viral mortality approximately every 8 h until no insects had died for at least 24 h. Nucleopolyhedrovirus-induced mortality is usually obvious with the insects becoming pale and then lysing to release large quantities of white OBs. Larvae that showed no clear symptoms of infection were checked for virus presence using Giemsa staining and light microscopy.

Experiment 2. Detailed analysis of fifth instar larvae: the influence of the egt gene on virus-induced mortality

We compared the pathogenicity of the two viruses in fifth instar T. ni. Insects were infected using small plugs of diet with a range of doses between 100 OBs and 40 000 OBs per insect, in two blocks. Twenty-five larvae were used per treatment. The virus doses were applied in a 1-µl droplet to the diet plug and the larvae were left to feed for 24 h. Twenty-five larvae per replicate were dosed with sterile water as untreated controls. Only those larvae that had eaten the whole diet plug were utilized in the bioassay. These larvae were transferred to individual polypots containing diet and monitored every day until death or pupation. We also recorded the impact of virus dose on development in terms of whether the insect attempted to pupate or not.

Experiment 3. Detailed analysis of fifth instar larvae: the influence of the egt gene on speed of kill, yield and cadaver weight

Fifth instar larvae were selected from mass culture as described above and used within 24 h of moulting. We challenged each larva with either a high (LD99) dose, 20 000 OBs, or a low dose, 250 OBs, of either the AcMNPV wild-type or egt-minus virus. Thirty larvae were inoculated for the high dose and between 100 and 125 for the low dose to ensure sufficient cadavers for analysis. Thirty larvae were also dosed with water as untreated controls. We inoculated the insects using diet plugs small enough to be eaten in 6 h to obtain precise estimates of time to death. Only those insects which had eaten the complete diet plug were transferred to individual polypots for monitoring. The insects were then monitored every 6 h until death or pupation. Larvae were weighed at death and then frozen until later yield analysis. The bioassay was repeated twice. We carried out virus yield analysis on a random subsample of between 38 and 52 larvae per treatment. Larvae were individually homogenized in 1 ml of sterile water; the homogenate was sonicated for 2 min and virus polyhedra counted using an Improved Neubauer haemocytometer with two counts per sample.

statistical analysis

The data were analysed using a generalized linear modelling program (GLIM version 3·77, 1985, Royal Statistical Society). Initially, all explanatory variables and their interactions were fitted to the data and the contribution of each term was tested for significance. Non-significant terms were removed leaving the minimal adequate model. Percentage mortality and proportion of insects attempting pupation were modelled using binomial errors, using the scale parameter to adjust deviances if required. Virus yield and weight data were transformed using natural logs before analysis. Normal model checking procedures were employed.

Results

experiment 1. the influence of larval stage on the action of the egt gene

Deletion of the egt gene resulted in a more rapid death for larvae in all instars, compared with insects infected with wild-type AcMNPV. However, the magnitude of this difference varied with instar (genotype × instar: F4,1044 = 9·3, P < 0·001) ranging from a reduction of 15% in the first instar to only 2% in third instar T. ni (Fig. 1).

Figure 1.

Mean time (± standard errors) taken for different instar larvae of T. ni to die after infection with AcMNPV wild-type (L1) virus (shaded bars) or egt minus AcMNPV (unshaded bars). (Sample sizes: Instar 1, wild type (wt) n = 129, vEGTDEL = 106; Instar 2; wt = 133, vEGTDEL = 96; instar 3, wt = 90, vEGTDEL = 48, instar 4, wt = 119, vEGTDEL = 123; instar 5, wt = 97, vEGTDEL = 113).

experiment 2. detailed analysis of fifth instar larvae: the influence of the egt gene on virus-induced mortality

The egt-minus virus was more pathogenic to fifth instar T. ni than its parent wild-type AcMNPV (genotype: χ2 = 5·44, df = 1, P < 0·05). The LD50 dose for the wild-type AcMNPV was 737 OBs (95% confidence intervals, 553 and 953, logit mortality = −6·992 + 1·059 ln virus dose) and for vEGTDEL, 355 OBs (95% confidence intervals, 254 and 466; logit mortality = −6·2184 + 1·059 ln virus dose). There was no difference between blocks (χ2 = 2·94, df = 1, P > 0·05).

Both virus genotype and dose affected whether the final instar larvae attempted to pupate, with virus dose affecting insects treated with the two viruses differently (genotype × dose interaction: χ2 = 13·15, df = 1, P < 0·001) (Fig. 2). There was no difference between the two bioassays for the wild-type virus (χ2 = 0·067, df = 1, P > 0·05) and the number of larvae attempting to pupate decreased markedly with increasing dose. However, the response varied significantly between the two egt-minus assays (genotype × block interaction: χ2 = 163·3, df = 1, P < 0·001). In one assay, vEGTDEL-treated insects responded like the untreated controls with almost all larvae attempting to pupate regardless of virus dose. However, in the second assay, fewer insects attempted to pupate and the number failing to do so increased with increasing virus dose, although to a lesser degree than for insects treated with the wild-type virus (Fig. 2). All untreated control insects successfully formed pupae.

Figure 2.

Proportion of final instar T. ni attempting to pupate after inoculation with virus. The two blocks of insects inoculated with wild-type AcMNPV (L1) have been combined: data shown as ▴; model fit, logit proportion attempting pupation = 5·298 − 0·7296 * ln dose (dashed line). First block of insects inoculated with vEGTDEL: data shown as •; model fit, logit proportion attempting pupation = 2·874 + 0·0546 * ln dose (heavy line). Second block of insects inoculated with vEGTDEL block 2: data shown as ○; model fit, logit proportion attempting pupation = 3·118 − 0·3106 * ln dose (light line).

experiment 3. detailed analysis of fifth instar larvae: the influence of the egt gene on speed of kill, yield and cadaver weight

Speed of kill

Fifth instar T. ni larvae were killed more rapidly by the egt-minus virus (genotype: F1,181 = 157, P < 0·001) and this altered with dose, with larvae treated with the lower dose suffering a longer duration of infection (dose: F1,181 = 24·34, P < 0·001) (Fig. 3). The effect of virus genotype and dose was additive (genotype × dose: F1,181 = 1·3844, P = 0·2409). Duration of infection was reduced by approximately 15% regardless of virus dose. The average mortality for a dose of 250 OBs was approximately 51% and for 20 000 OBs, 97%.

Figure 3.

Mean times to death (± standard errors) for fifth instar T. ni treated with a low (250 OBs) and a high (20 000 OBs) dose of wild-type AcMNPV and vEGTDEL. Sample sizes: high dose, n = 50 wt, n = 45 vEGTDEL, low dose, n = 38 wt, n = 52 vEGTDEL.

Virus productivity

Larvae killed by the egt-minus virus produced a significantly lower yield of OBs than those killed by the wild-type virus (genotype: F1,180 = 16·28, P < 0·001). Additionally, more virus was produced per insect at the higher dose than the lower (dose: F1,180 = 20·98, P < 0·001), despite the fact that these larvae died more rapidly (Fig. 4). Infection duration did not affect the amount of virus OBs produced (duration: F1,180 = 0·566, P = 0·4528). Infection with the egt-minus virus reduced the yield of OBs by approximately 17% compared with insects infected with the wild-type virus.

Figure 4.

Mean yield (back-transformed means ± standard errors) of OBs produced per fifth instar T. ni infected with either a low (250 OBs) or a high (20 000 OBs) dose of wild-type AcMNPV or vEGTDEL. Sample sizes as for Fig. 3.

Differences in virus yield could result from differences in insect growth or differential growth of virus genotypes within the larva. Differential growth can be examined by estimating how much insect tissue has been converted into virus OBs. Insects infected with wild-type AcMNPV had a higher yield per unit weight than those killed with the egt-minus virus (genotype: F1,180 = 5·063, P < 0·05) (Fig. 5). The virus dose ingested also significantly influenced the outcome with yield per unit weight being greater at the higher dose for both virus types (dose: F1,180 = 18·2, P < 0·001). OB yield per unit weight was not influenced by the length of the infection period (duration: F1,180 = 0·3081, P = 0·5795). Larval weight differences could indicate whether the different virus genotypes influence insect growth. Larvae killed with wild-type AcMNPV were significantly heavier than those killed by the egt-minus virus (genotype: F1,180 = 12·9, P < 0·001) at any given time to death. The two doses produced different effects. In insects infected with the high dose, time to death increased in a positive manner with cadaver weight whereas at the low dose this relationship was negative (duration × dose: F1,180 = 16·82, P < 0·001) (Fig. 6).

Figure 5.

Mean yield OBs per mg cadaver weight (back-transformed means ± standard errors) produced in fifth instar T. ni infected with either a low (250 OBs) or a high (20 000 OBs) dose of wild-type AcMNPV or vEGTDEL. Sample sizes as for Fig. 3.

Figure 6.

Relationship between mean weight (mg) of virus-killed T. ni cadavers and time to death for fifth instar larvae inoculated with low (250 OBs) and high (20 000 OBs) virus doses. Insects inoculated with wild-type AcMNPV, high-dose data shown as ▴; low-dose data shown as ▵. Model fit: ln weight cadaver killed by wild-type AcMNPV = 6·4985687 − 0·1307274 * ln dose + 0·00107863 * time to death * ln dose − 0·0076088 * time to death. Insects inoculated with egt-minus virus, high-dose data shown as •; low-dose data shown as ○. Model fit: ln weight cadaver killed by vEGTDEL = 6·41470593 − 0·1307274 * ln dose + 0·00107863 * time to death * ln dose − 0·0076088 * time to death. Samples sizes as for Fig. 3.

Discussion

We have examined the influence of a single virus gene, the egt gene, on three key parameters likely to impact on the fitness of an insect virus (Cory, Hails & Sait 1997). All three of the parameters investigated, pathogenicity, infection duration and virus productivity, were significantly altered by EGT expression in final instar T. ni larvae. The pathogenicity of the egt-minus virus was higher than that of its parent wild-type virus. Deleting the egt gene resulted in larvae of all instars being killed more rapidly, although the proportionate reduction varied with instar. In final instar larvae, virus yield was also reduced in insects infected with an egt-minus virus and this was accompanied by a reduction in cadaver weight and OB yield per unit weight. The study also demonstrated the major impact that virus dose can have on the outcome of the interaction. The likelihood of EGT-related developmental changes decreased as virus dose was reduced. Additionally, infections resulting from larger virus doses produced greater yields of virus progeny overall and per unit cadaver weight than insects infected with lower doses, despite the fact that larvae at the higher doses died earlier.

For baculoviruses, which require host death before transmission can occur, the level of host mortality (pathogenicity) is obviously crucial and likely to have a major impact on virus fitness. All previous studies have shown that the egt gene has no impact on baculovirus pathogenicity (O’Reilly & Miller 1991; Popham, Li & Miller 1997; Treacy, All & Ghidui 1997; Bianchi et al. 2000; Chen et al. 2000). The exception is a recent study by Pinedo et al. (2003) in which pathogenicity was also increased by egt deletion in Anticarsia gemmatalis NPV, however, in this case the egt gene was replaced by the Escherichia coli LacZ gene, which might itself also affect pathogenicity. The increase in pathogenicity in the current study was unexpected; if a prediction was to be made about whether the expression of EGT altered virus pathogenicity, it would be that it increased it. Although the function of the egt gene can be deleted in cell culture, it is retained when the virus is multiply passaged from one larva to the next. This implies that the gene has a function at the within-host level. There is indirect evidence that it could counteract the potentially negative effects of ecdysteroids on virus replication. When a recombinant egt-minus AcMNPV was constructed to express prothoraciotropic hormone (PTTH) (which stimulates the production of ecdysone) from Bombyx mori, pathogenicity was reduced by two orders of magnitude (O’Reilly, Hails & Kelly 1995). Additionally, an injection of 20-hydroxyecdysone into NPV-infected Heliothis larvae caused a delay in the onset of NPV-induced pathology as well as reducing virus mortality (Keeley & Vinson 1975). Why an increase in pathogenicity in the absence of the egt gene was observed in the present study cannot at present be explained.

Other studies that have addressed the impact of the baculovirus egt gene have tended to focus on single larval instars or single parameters. However, even among these studies, the increase in baculovirus speed of action is not universal (Cory et al. 2001). This is the first study to show that the egt gene alters speed of kill across all larval instars of one host species. Even within this host–virus system the results have not been consistent as the reduction in time to death for fifth instar T. ni ranged from 4% to 15% in the two experiments. It is difficult to compare across studies because of the differing virus doses and techniques used, as well as the incorporation of marker genes in some constructs. However, the reductions in other systems range from 0 to 33% (Cory et al. 2001). Differences between species and viruses may be genuine; alternatively, as is perhaps indicated by our results, they may be a result of a high sensitivity to within instar changes in hormone levels, making the timing of infection crucial.

Virus yield has only been measured in a few studies as interest in the egt gene has tended to focus on its capacity for producing recombinant insecticides with an enhanced speed of action. There is frequently a trade-off between the duration of infection and the amount of virus produced (e.g. Burden et al. 2000; Hernández-Crespo et al. 2001, Hodgson et al. 2001). Thus it would be expected that insects killed more rapidly by the egt-minus virus would produce fewer OBs. However, despite the fact that we measured a reduction in virus yield in fifth instar T. ni larvae infected with the egt-minus virus, there was no trade-off with speed of kill at the within genotype level. This may be in part due to the increased variability of the data at the lower dose. Reductions in virus yield have been reported in other studies but not in every case (Cory et al. 2001), although this may in part be due to the accuracy of counting virus in small samples (insects).

Virus productivity could be altered in several ways; reduced yield could simply be the result of a shorter host life span, it could be a product of changes in the growth rate of the insect host or it could be the result of an alteration in viral replication rate. At the high dose, cadaver weight increased as time to death increased. However, insects infected with the egt-minus virus were lighter than those infected with the wild-type virus at any time point. This implies that larvae infected with the wild-type virus had enhanced growth at some point. We found a similar result previously with second and fourth instar T. ni (Wilson et al. 2000). The increase in weight was also accompanied by a higher yield of OBs per unit weight in insects infected with the wild-type virus, suggesting a replication advantage. However, in the previous study on second and fourth instars there was no difference (Wilson et al. 2000), so it is possible that EGT expression exerts a different effect in final instar larvae.

An important outcome of the study was the demonstration that the virus dose ingested had a significant impact on all aspects of the virus–host interaction. More rapid speed of kill at higher doses is commonly seen in baculovirus bioassays (e.g. Burden et al. 2000; Hernández-Crespo et al. 2001), however, this was also associated with a greater yield at the higher dose, even though the insects died more rapidly. In fact, the difference in yield between doses was similar to that between the two virus genotypes. As insects infected with a higher dose of virus also had a higher yield per unit weight, the resulting increase in yield might be tied to the fact that higher doses are more likely to result in multiple foci of infection, and the more rapid utilization of host tissues. Curiously, while cadaver weight increased as might be expected with time to death at the higher dose, it decreased at the lower dose. This might relate to developmental differences, such as the number of insects that achieved moult, but further studies are needed to investigate this. Virus dose had a major impact on developmental changes in final instar larvae with higher doses leading to an increased likelihood of developmental arrest. This could imply that EGT needs to reach a critical level before it has a significant impact on the host. The difference between the two assays with the egt-minus virus is intriguing and might result from small, but biologically important, differences in the point within the instar that the insects were inoculated. Studies using final instar Heliothis virescens and wild-type AcMNPV have shown that the likelihood of developmental arrest decreases later in the instar and this is accompanied by a reduction in virus yield (O’Reilly, Hails & Kelly 1998). Our results indicate that the egt gene is unlikely to produce a significant difference in virus yield under all conditions and that virus dose, timing of infection and the degree of developmental arrest could all be crucial factors.

How does the egt gene enhance baculovirus fitness? The experiments on T. ni and AcMNPV clearly show that expression of EGT increases both the time taken for the insects to die in all instars and viral yield in later instars at least. The question remains as to whether the observed increase in yield, which is assumed to lead to a higher transmission rate, is sufficient and outweighs the perceived disadvantage of delaying a further round of transmission. Using a simplified experimental set up (branches in jugs), Dwyer et al. (2002) demonstrated that the transmission rate of an egt-minus virus from the Gypsy Moth, Lymantria dispar, is less than that of the wild-type virus. By introducing these data into a mathematical model of Gypsy Moth–NPV dynamics, Dwyer et al. (2002) also showed that this enhanced transmission gave the wild-type virus a significant fitness advantage over the egt-minus virus. This analysis has been extended to examining the potential competition between genetically modified baculoviruses with enhanced speed of kill in general (Dushoff & Dwyer 2001). They concluded that faster-acting viruses were unlikely to become dominant unless the reduction in yield was small. However, their model also indicated that it might take decades for faster-acting viruses to become extinct and that the outcome was significantly influenced by the overwintering capacity of the viruses, implying that egt-plus and minus viruses could coexist. Environmental persistence of the virus could thus be a crucial factor in understanding the conditions in which an egt-minus virus could dominate or coexist with the wild type. It is feasible that EGT expression could influence OB decay rate indirectly by modifying host behaviour. Final instar larvae usually undergo behavioural changes prior to pupation; these can include leaving the plant to pupate in the soil and spinning a cocoon inside which to pupate. Neither of these behaviours is likely to promote virus transmission in the short term but could decrease OB decay rates by depositing the virus in protected microhabitats. More detailed manipulative studies and longer-term field-based experiments are needed to fully understand the impact of the egt gene and the conditions under which it would confer a selective advantage.

Acknowledgements

We would like to thank Tim Carty for providing the insects and the diet. We would also like to thank Dave Hodgson for his thoughtful and constructive comments on the manuscript. This work was funded by NERC grant GR3/8967 to DOR and JSC.

Ancillary