SEARCH

SEARCH BY CITATION

Keywords:

  • competition;
  • differential selection;
  • fitness;
  • host ecology;
  • mixed infection;
  • nucleopolyhedrovirus;
  • Panolis flammea;
  • pathogen;
  • resource partitioning;
  • virulence

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Mixed-genotype infections are common in many natural host–parasite interactions. Classical kin-selection models predict that single-genotype infections can exploit host resources prudently to maximize fitness, but that selection favours rapid exploitation when co-infecting genotypes share limited host resources. However, theory has outpaced evidence: we require empirical studies of pathogen genotypes that naturally co-infect hosts. Do genotypes actually compete within hosts? Can host ecology affect the outcome of co-infection? We posed both questions by comparing traits of infections in which two baculovirus genotypes were fed to hosts alongside inocula of the same or a different genotype. The host, Panolis flammea, is a herbivore of Pinus sylvestris and Pi. contorta. The pathogen, PfNPV (a nucleopolyhedrovirus), occurs naturally as mixtures of genotypes that differ, when isolated, in pathogenicity, speed of kill and yield. Single-genotype infection traits failed to predict the ‘winning’ genotypes in co-infections. Co-infections infected and caused lethal disease in more hosts, and produced high yields, relative to single-genotype infections. The need to share with nonkin did not cause fitness costs to either genotype. In fact, in hosts feeding on Pi. sylvestris, one genotype gained increased yields in mixed-genotype infections. These results are discussed in relation to theory surrounding adaptive responses to competition with nonkin for limited resources.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Advances in techniques to distinguish pathogen genotypes have revealed high frequencies of mixed-genotype infections in many host–pathogen interactions (Taylor et al., 1997; Fisher & Viney, 1998; Meijer & Leuchtmann, 1999; Paul et al., 1999; Hodgson et al., 2001; Read & Taylor, 2001). This observation emphasizes the need to consider within-host competition between parasite genotypes if we are to understand the epidemiology and coevolution of host–parasite interactions (Read & Taylor, 2001). Classical epidemiological models tend to consider within-host dynamics as a simple race between parasite genotypes to gain a disproportionate share of uniform host resources (Bremermann & Pickering, 1983; May & Nowak, 1995; Frank, 1996), analogous to the classic ecological ‘tragedy of the commons’ hypothesis (Hardin, 1968). Prudent exploitation of within-host resources, optimal for many single-genotype infections, should break down when parasites must share limited resources with nonkin. Where increased rates of exploitation cause greater harm to hosts, this argument forms the basis of the classic prediction of increased virulence of mixed-genotype infections (Frank, 1996), where virulence is defined as the impact of infection on host fitness (Read, 1994).

However, we lack a general, empirical understanding of the interactions between parasite genotypes within hosts, and the impact on parasite fitness of sharing limited host resources. This is particularly important now that models of evolutionary epidemiology have progressed to the point that different modes of competition and different scales of natural selection can radically alter predictions of optimal host resource exploitation and virulence (Brown et al., 2002; Griffin & West, 2002; West & Buckling, 2003). Previous empirical work suggests that the impact of co-infection on parasite performance is not easy to predict from phenotypic traits of single-genotype infections (Nakamura et al., 1992; Weeds et al., 2000; Cox, 2001; Thomas et al., 2003) and need not fit the classical predictions of increased virulence (Imhoof & Schmid-Hempel, 1998). In many host–parasite interactions, competition between co-infecting parasites will be mediated by the host immune system (Taylor et al., 1997; Davies et al., 2002). Evidence from nonparasite populations and communities demonstrates that competition for an apparently shared resource can be avoided by positive interactions or resource partitioning (Loreau, 2000; Hodgson et al., 2002a): host tissues and cell-types may differ sufficiently to allow resource partitioning between parasite genotypes and species, reducing or negating fitness costs in even the most resource-hungry infections. It is also possible that host ecology will influence the fitness of co-infecting parasites, just as it can affect the fitness of single-genotype infections (e.g. Hodgson et al., 2002b).

We compared the pathogenicity, time to kill, and yield of single- and mixed-genotype nucleopolyhedrovirus (NPV; Baculoviridae) infections in larval Lepidoptera feeding on two different food plant species. Overt NPV infections must kill the host to transmit, therefore pathogenicity, usually defined as the capacity to cause disease, is quantified here as the number of hosts killed by NPV after challenge from NPV inoculum. It is therefore a combination of infectivity, and the probability of overt disease causing host death. Several features of baculovirus infections make them useful candidates for empirical work on within-host competition. First, genotypic diversity is high in baculovirus populations (Lee & Miller, 1978; Knell & Summers, 1981; Hodgson et al., 2001). Secondly, overt NPV infections (such as those studied here) use most of the host's tissues during an infection: co-infecting NPV genotypes must share this limited resource, therefore basic competition seems certain to occur. Thirdly, there is a well-known correlation between the time to host death and yield of overt NPV infections, as a result of growth of hosts during infection (Hodgson et al., 2001; Clarke et al., 2004). This favours prudence for single-genotype infections (maximize yield by delaying host death). Fourthly, baculoviruses have great utility as biopesticides, lending applied significance to our understanding of their fitness as mixed-genotype inocula (Muñoz & Caballero, 2000; Bull et al., 2003). Fifthly, larval hosts are usually simple to rear in the laboratory, and the combination of high virus yields and established molecular analysis techniques facilitate the study of relative yields of co-infecting genotypes.

We studied the Panolis flammeaPfNPV – pine food plant system. Panolis flammea D&S is a native herbivore of Pinus sylvestris L. forests in Europe, and has become an economically important pest of Pi. contorta Douglas ex Loudon plantations in Scotland. Molecular analysis of a single larva infected with a wild isolate of PfNPV demonstrated that at least 24 NPV genotypes shared one host individual. Further work revealed phenotypic differences between individual genotypes (Hodgson et al., 2001), and differential fitness of two genotypes when hosts fed on different food plant species (Hodgson et al., 2002b). Further details of this host–pathogen system, and a discussion of the maintenance of genotypic variation in PfNPV populations, are published elsewhere (Hodgson et al., 2003). Two isolated genotypes, Pf4 and Pf6, have similar pathogenicities in single-genotype infections of hosts fed on Pi. contorta (Hodgson et al., 2001). We used single- and mixed-genotype infections of these two NPV genotypes to ask three critical questions:

  • 1
    Do NPV genotypes compete in mixed infections?
  • 2
    Does the host's food plant species influence the outcome of within-host competition?
  • 3
    Can the ‘winning’ genotype be predicted from single-genotype infection traits?

The first question has an apparently simple answer when unlimited hosts are available for infection: given that most host tissue is converted to virus transmission bodies, competition can only be avoided if co-infected hosts grow to at least double the size of singly infected hosts. More interesting, however, is the outcome of competition given the assumption that all or most hosts are infected, as was the case during the epizootic from which the PfNPV genotypes were originally isolated. Our experimental design therefore framed the questions in the following way: what is the effect on the fitness of a virus inoculum of sharing host individuals with an equal-sized inoculum of the same, or a different, virus genotype?

Null hypothesis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Assuming no antagonistic or synergistic interactions between co-infecting genotypes, our null hypothesis was that pathogenicity should not differ between single- and mixed-genotype infections given that total inoculum size was kept constant, and that other co-infection traits (time to kill, virus yield) should be intermediate between the traits of the single-genotype infections, irrespective of host ecology.

Alternative hypotheses

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Despite the simplicity of the experiment, several alternative outcomes were possible: different increases or decreases in any or all of the measured infection traits for each host ecology. However, three possibilities seemed likely a priori. Our first a priori alternative hypothesis, based on classical ‘tragedy of the commons’ theory, predicted a facultative race for resources in mixed infections, resulting in faster kill and lower yields. Note that in nonobligate-killing parasites, this would also be predicted to increase the virulence of mixed infections (Frank, 1996). We make no such prediction here because overt NPV infections must kill the host to transmit: virulence is divorced from rates of resource exploitation. Secondly, decreased exploitation rates caused by interference between virus genotypes should result in delayed host death and increased yield from co-infections. Thirdly, positive interactions (i.e. facilitation) or resource partitioning between co-infecting genotypes should result in increased total virus yields due to more complete conversion of host resources into transmission bodies, and no change in time to kill.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Panolis flammea larvae were raised to fourth instar on Pi. contorta saplings at the Centre for Ecology and Hydrology, Banchory, Aberdeenshire, Scotland. A total of 475 larvae were randomly allocated to three virus treatments (Pf4, Pf6 or both genotypes) and two food plant treatments (Pi. sylvestris or Pi. contorta). Thirty more larvae were reared as controls. Individual P. flammea larvae were inoculated with monoclonal NPV infections (genotypes hereafter called Pf4 and Pf6), or mixed infections (1 : 1 ratio of Pf4 and Pf6). NPV transmission stages are called occlusion bodies (OBs). Larvae were dosed individually in chambers of a 96-well microtitre plate by feeding them the appropriate species of pine needle coated with either 1000 OBs of Pf4 or Pf6, or a mixture of 500 OBs of each genotype. The design therefore compares the fitness of 500 OBs of either genotype when fed to hosts alongside 500 OBs of the same or a different genotype. Control doses contained no OBs. Larvae were transferred to individual rearing chambers when the needle dose had been eaten, then reared at 18 °C, 16 : 8 L : D photoperiod. Larvae were monitored every 12 h until death or pupation, and fed ad libitum with decontaminated shoots of the appropriate pine species.

Pathogenicity, time to death, yields and relative yields of transmission stages from cadavers were measured. Viral deaths were confirmed by Giemsa staining of macerated cadaver tissue. Speed of kill was measured as the number of days between transfer to rearing chamber and larval death. Virus yields were measured by macerating each cadaver in 1 mL distilled water and counting OBs in two subsamples using a haemocytometer and light microscope.

To determine relative yields in mixed-genotype infections, DNA was extracted separately from each of 37 co-infected cadavers using phenol-chloroform extraction from macerated cadavers (Hunter-Fujita et al., 1998), then analysed by gel electrophoresis after digestion with HindIII endonuclease. Fragmented DNA from each larva was run on a single gel lane, then band intensities were quantified using Phoretix® software (Non-Linear Dynamics, Newcastle upon Tyne, UK). Background signal was removed using a rolling disc algorithm with radius of 10 units. The proportion of Pf4 DNA in each cadaver was calculated as the intensity of a restriction fragment unique to Pf4 relative to the intensity of a similar-sized fragment shared by both genotypes. The unique Pf4 band (22 kbp) was chosen to be as large as possible (for detection purposes), and as close as possible on the gel to the shared band (10.1 kbp). The fading of band intensity with decreasing fragment length was accounted for by multiplying the raw proportion intensity of Pf4 by the relative intensity of band 3 to band 1 in a pure Pf4 DNA sample run on each gel (Hodgson et al., 2003).

Relative yield data from mixed-genotype infections were used to calculate the yield of OBs from an initial 500 OB inoculum: this was calculated as (total yield × 0.5) in single-genotype infections, and (total yield × relative representation) for each genotype in mixed-genotype infections. Note that relative representation and total yield information was only available for 37 mixed-genotype infection hosts.

Data were analysed using generalized linear models (GLM). GLM promotes the use of data transformations and alternative error structures without loss of statistical power. Pathogenicity was analysed using a binary error structure and logit link function. Times to death and virus yields were log-transformed and analysed using normal errors. Relative yields of Pf4 from co-infected cadavers, because bounded by 0 and 1, were arcsine-transformed then analysed using normal errors. Virus and food plant treatments were factors, and time to death was used in yield models as a linear and a quadratic covariate. Model simplification transferred all sums of squares from nonsignificant interactions and main effects into the error term of each model, therefore degree of freedom for significant model terms differ from those of nonsignificant terms in the results section. All transformations and error structures satisfied model-checking procedures (Crawley, 1993).

Pathogenicity

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Mixed-genotype doses of PfNPV were significantly more likely to kill hosts (binary analysis of successful virus-producing deaths: inline image = 10.3, P < 0.01) than either single-genotype dose, which shared similar pathogenicities (inline image = 0.7, ns; Fig. 1). Overall, pathogenicity was high at the chosen dose of 1000 OBs: the proportions of hosts killed by overt virus infection were 90 (Pf4), 92 (Pf6) and 98% (Pf4 and 6). None of the control larvae died of virus infection, confirming a lack of contamination during the experiment. Neither food plant (inline image =1.1, ns) nor the interaction between food plant and virus treatment (inline image = 0.9, ns) influenced pathogenicity.

image

Figure 1. Logit-transformed virus-caused mortality of doses of fourth instar Panolis flammea larvae challenged by 1000 occlusion bodies (OBs) of Pf4 or Pf6, or a 500 : 500 mixture of Pf4 and Pf6.

Download figure to PowerPoint

Time to death

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

The time elapsed between infection and host death of virus infections was differentially determined by the food plant species being fed on by the host (Fig. 2). This was described statistically by a significant interaction between virus treatment and food plant species (F2,387 = 11.19, P < 0.001). Graphical interpretation confirmed (Fig. 2) that the interaction was caused by differential effects of food plant on time to kill single-genotype infections. Mean times to kill mixed-genotype infections were intermediate between the single-genotype infection mean values (Fig. 2), and did not differ between food plant species (t144 = 0.732, ns). Main effects of virus treatment and food plant species are made irrelevant by their statistical interaction.

image

Figure 2. Time to death of fourth instar Panolis flammea larvae killed by pure and mixed infections of Pf4 and Pf6 after feeding on Pinus sylvestris (solid bars) and Pi. contorta (grey bars), showing mean values of all treatments ±1 SE.

Download figure to PowerPoint

Virus yields

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Diminishing returns of virus yield with increasing time to death were suggested by a significant quadratic relationship between these two measurements (Fig. 3: linear covariate F1,388 = 17.06, P < 0.001; quadratic F1,388 = 14.16, P < 0.001). However, rather than saturating, yields from late deaths were smaller than peak yields, suggesting (speculatively) that they were caused by late infections or came from hosts that showed some resistance to infection growth. Irrespective of time to death, virus yields differed between virus treatments (F2,388 = 6.90, P < 0.01). Mixed-genotype infections gained highest yields but not significantly higher than Pf6 infections (F1,388 = 1.62, ns). Introduction to the statistical model of more complicated interaction terms showed that regression slopes did not differ between virus treatments (quadratic × virus-treatment interaction, F2,386 = 2.17, ns; linear × virus-treatment interaction, F2,386 = 0.10, ns). Food plant species had no impact on the quadratic fit (F1,387 = 0.643, ns), but the linear slope between yield and time to kill was significantly higher for hosts feeding on Pi. sylvestris than on Pi. contorta (F1,387 = 3.886, P = 0.05).

image

Figure 3. Mean yields (±1 SE) of pure and mixed infections of Pf4 and Pf6 in Panolis flammea larvae. Pf4 infections diamonds and solid lines, Pf6 squares and dashed lines, mixed infections triangles and dotted lines. Fitted lines represent minimal adequate models fitted by ancova.

Download figure to PowerPoint

RFLP analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Pf4 was absent from the yield of only one of the 37 mixed infection cadavers analysed: its relative yield averaged 44.6% [95% confidence interval (CI): 35.0–54.5]. Relative yield of Pf4 was not affected by time to death (F1,35 = 0.14, ns), food plant species (F1,35 = 1.82, ns), or by their interaction (F1,34 = 3.11, P = 0.09).

Relative yield analyses

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

We compared the yield attributable to 500 OBs of inoculum between single- and mixed-genotype infections of each genotype, using two separate statistical models. For Pf4, half the yield from each pure Pf4 infection was compared with the yield of mixed-genotype infections multiplied by the proportional representation of Pf4, using only those mixed infection cadavers where both pieces of information were measured. The same logic was used for the Pf6 comparison. Where Pf4 was not detected by restriction fragment length polymorphism (RFLP), model structures were improved by giving the apparently ‘absent’ genotype a proportional representation of 0.1%, below any RFLP detection threshold.

For Pf4, sharing hosts with OBs of the same genotype or of Pf6 had no effect on yield (single vs. mixed infection: F1,167 = 0.42, ns), irrespective of food plant species (food × virus-treatment interaction, F1,166 = 2.16, ns; food plant main effect, F1,167 = 0.37, ns; Fig. 4a). For Pf6, however, yield per 500 OBs was explained by an interaction between virus treatment and food plant species (F1,169 = 10.37, P < 0.01). When hosts fed on Pi. contorta, yield per 500 OBs did not differ between single- and mixed-genotype infections, but in hosts fed on Pi. sylvestris, Pf6 gained significantly higher yields in mixed- than in single-genotype infections (Fig. 4b). Note the potential link between increased relative yields of Pf6, short-time to kill of pure Pf6 infections and the significantly steeper increase in yield per time, all in hosts fed on Pi. sylvestris.

image

Figure 4. Yield attributable to 500 occlusion bodies (OBs) of (a) Pf4 or (b) Pf6, forced to share hosts with 500 OBs of the same (‘single’) or the alternative (‘mixed’) genotype. Hosts fed on either Pinus sylvestris (solid bars) or Pi. contorta (grey bars).

Download figure to PowerPoint

Do NPV genotypes compete for limited host resources?

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Unexpectedly, the answer to this question was, at least for the two genotypes studied, ‘no’. If NPV infections were simple ‘infect then grow until resources are exhausted then transmit’ processes, the fastest replicating genotype should dominate the yield from a mixed-genotype infection, and determine the time to death of the host, and by correlation the yield of OBs. This was not the case for PfNPV. Clearly there were insufficient host resources available to allow each NPV genotype to replicate independently in mixed-genotype infections: yields of each genotype were reduced relative to single-genotype infection yields. In other words, host tissue did not provide a nonoverlapping resource base for each virus genotype. However, as a comparison of the fitness of virus that must share hosts with virus of the same or different genotypes, our experiment demonstrated no cost to sharing hosts with another genotype. Indeed, under some conditions fitness was increased (see Section below). Increased yields of mixed infections were not achieved by increasing the length of time taken to kill the host, suggesting that facultative prudence was not the cause of our results. Instead, we believe that some level of facilitation or resource partitioning between NPV genotypes promoted the lack of cost to, or increase in, yield. The exact mechanism causing facilitation can only be determined by sacrificial sampling of hosts during the infection cycle. This is important future work. Considering also the increased pathogenicity of (i.e. successful transmission from) mixed-genotype infections, it is clear that sharing hosts need not be costly to the PfNPV genotypes tested, and will under some circumstances be beneficial.

Does the host's food plant species influence the outcome of within-host competition?

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Previous work on this host–pathogen interaction has demonstrated differential selection of PfNPV genotypes mediated by the species of food plant fed on by hosts during infection.

Results from single-genotype infections provided further support for differential selection of NPV genotypes in hosts feeding on different species of plant (see Hodgson et al., 2002b). Pf4 killed hosts quicker on Pi. contorta than on Pi. sylvestris: Pf6 results were vice versa. Time to kill in mixed infections was similar for both food plant species, and intermediate between single-genotype values.

Critically, food plant species mediated the outcome of within-host interactions between genotypes in terms of relative yields. In hosts feeding on the native host plant, Pi. sylvestris, Pf6 gained increased yields in mixed-genotype infections of hosts feeding on the sharing hosts compared with other OBs of the same genotype. This result was independent of time to death, and therefore represents an unexpected but potentially important function for host food plant in determining the outcome of co-infection. We speculate that increased growth rates of host larvae on Pi. sylvestris, found in several feeding trials (Vanbergen et al., 2003), caused the faster increase in yield per time on this food plant. This allowed Pf6, normally a fast killer in single infections of hosts eating Pi. sylvestris, to gain extra yield when achieving longer time to kill in co-infections with Pf4. It is also intriguing that this facilitation of growth occurred when hosts fed on the native food plant, but not on the exotic alternative.

Can the ‘winning’ genotype be predicted from single-genotype infection traits?

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

Neither time to kill nor yield could be predicted for mixed-genotype infections based on single-genotype infections and the null hypothesis of no antagonistic or synergistic interactions. Pathogenicity was significantly greater in mixed infections than predicted by either single infection trait. The roughly equal relative representation of both genotypes in mixed-genotype yields was expected to result in mixed-genotype yields lying somewhere between the yields caused by pure Pf4 or Pf6 infections. Instead, mixed-genotype yields were at least as great as for the highest yielding single genotype, Pf6, and were significantly higher than predicted based on weighted relative yield analysis. Only time to kill matched the null hypothesis that mixed-genotype infection traits should be intermediate between those of the respective single genotypes.

Our results at first appeared to confirm the predictions of conventional adaptive virulence theory: that mixed-genotype infections will impose greater impacts on host fitness (i.e. pathogenicity) via increased rate of host resource use (i.e. large yield per unit time) (Bremermann & Pickering, 1983; Frank, 1996). However, in obligate-killing infections, host mortality is essential for transmission and therefore should be maximized, even in single-genotype infections. This is a classic example where the commonly assumed trade-off between exploitation rate and virulence breaks down (Ebert & Bull, 2003). The only discernable fitness benefit for obligate-killing pathogens with low pathogenicity in single-genotype infections is for cheating genotypes, dysfunctional when alone but performing well in mixed infections by parasitising the replication and transmission machinery of other pathogen genotypes. This behaviour is predicted to result in overall lower yield and transmission from mixed-genotype infections (Muñoz & Caballero, 2000; Bull et al., 2003; West & Buckling, 2003), and so does not fit our results either. Adaptive virulence theory also predicts that co-infecting parasite genotypes should race to gain disproportionate shares of host resources. This should result in faster speed of kill because of earlier exhaustion of host resources and, because of the trade-off between speed of kill and yield, lower yields from mixed-genotype infections. The opposite occurred: high yields, relative to those of single-genotype infections, were gained at no cost to speed of kill.

Instead our results are suggestive of some level of facilitation or resource partitioning between co-infecting NPV genotypes during both the infection and the growth phase. Facilitation may be an adaptive response to the high frequency of natural mixed infections, avoiding competition between NPV genotypes within hosts, and sometimes gaining benefits from infection diversity. Alternatively, facilitation may be a purely ecological consequence of phenotypic differences between NPV genotypes.

Two further mechanisms may explain our results. First, is the possibility of a trade-off between quantity and quality of transmission stages from each infection. Measuring only individual aspects of infections risks ignorance of life history trade-off that will be critical determinants of parasite fitness (Frank, 1996; Day, 2003). A comprehensive comparison of parasite fitness in single- and mixed-genotype infections should account for between-host competition (Cory et al., 1994; Goulson et al., 1995). Mixed-genotype yields may have been composed of OBs containing few virus genomes or with less resistance to degradation outside the host. Transmission experiments are costly and time-consuming but our results warrant a comparison of the transmission success of mixed- and single-genotype OBs. A second alternative explanation follows the immune response hypothesis of Taylor et al. (1997) and Davies et al. (2002). It is possible that P. flammea larvae vary in their resistance to specific PfNPV genotypes. If so then challenge by mixed inocula, with each component genotype being pathogenic towards a subset of hosts, could promote the observed increase in pathogenicity and, potentially, yield. Specificity of insect larval immune responses remains poorly understood and deserve further study.

An important caveat to our results is that they represent only one interaction between two genotypes, isolated from a diverse mixture of genotypes originally sampled from a single host cadaver. We therefore avoid the claim that all NPV co-infections are neutral or commensal. Instead, we publish this specific interaction example as a demonstration that ecological interactions between parasite genotypes within hosts can produce co-infection results that do not fit with adaptive hypotheses. It remains possible that other pairs or groups of co-infecting genotypes may show neutral, commensal, mutualistic, parasitic or competitive interactions, or facultative changes in prudence strategies. It seems clear that adaptive hypotheses should be tested using experimental evolution (e.g. the selection of NPV genotypes under conditions of single- and mixed-genotype infections over several infection cycles), and that ecological hypotheses deserve further study via manipulations of co-infection diversity. It is also possible that variation in the ingested dose of OBs, known to be important to single-genotype pathogenicity, may influence the outcome of co-infection.

Our results have several implications for both biological pest control and fundamental epidemiology. Mass production of baculovirus pesticide sprays (in cell cultures or in mass host rearing) either use or may favour single genotypes. High pathogenicity and yield of mixed-genotype infections, and the differential performance of NPV genotypes in different ecological conditions, should instead favour the retention of genotypic variation in biopesticide culturing. The selection pressures assumed by most models of competition for limited resources (Hardin, 1968), and by most classical models of virulence evolution, are that competing species or strains have completely overlapping resource needs. Thus, competition should cause a reduction in diversity over time as poor competitors are excluded from resource patches. However, evidence for compensation for competition such as presented here may help explain the naturally high diversity of baculovirus populations, and the maintenance of diversity in other host–pathogen systems.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References

The authors thank Penny Luntz, Rachel King, Mo Docherty and Ben Woodcock for practical assistance. Thanks also to Bernadette Green, Allan Watt, Sue Hartley, Angus Buckling and Judith Myers for contributions. Our research was funded by NERC EDGE thematic grant number GST/02/1838.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Null hypothesis
  5. Alternative hypotheses
  6. Materials and methods
  7. Results
  8. Pathogenicity
  9. Time to death
  10. Virus yields
  11. RFLP analysis
  12. Relative yield analyses
  13. Discussion
  14. Do NPV genotypes compete for limited host resources?
  15. Does the host's food plant species influence the outcome of within-host competition?
  16. Can the ‘winning’ genotype be predicted from single-genotype infection traits?
  17. Acknowledgments
  18. References