SEARCH

SEARCH BY CITATION

Keywords:

  • biological control of weeds;
  • food webs;
  • fundamental host range;
  • host range expansion;
  • host specificity testing;
  • non-target effects;
  • risk assessment

Introduction

  1. Top of page
  2. Introduction
  3. Host specificity testing and risk to non-target plant species
  4. Avoiding undesirable indirect non-target effects
  5. Final remarks
  6. Acknowledgements
  7. References
  8. Biosketch

Many ecologists profess a negative opinion of biocontrol, whilst practitioners argue that it offers a cost-effective solution for many invasive weed problems. Practitioners are under pressure to implement effective weed biocontrol more quickly, cheaply and safely. In our practitioner’s perspective, we focus on two key areas, host range testing and indirect non-target effects, where advances in ecological research could progress these stakeholder-driven aims and minimize potential negative outcomes of biocontrol that concern ecologists and practitioners.

Host specificity testing and risk to non-target plant species

  1. Top of page
  2. Introduction
  3. Host specificity testing and risk to non-target plant species
  4. Avoiding undesirable indirect non-target effects
  5. Final remarks
  6. Acknowledgements
  7. References
  8. Biosketch

Linking Fundamental and Realised Host Ranges

Host specificity testing is used to discard weed biological control agents that might damage native or valued exotic plants. ‘No-choice’ starvation testing is almost infallible at identifying plants that cannot be hosts and those which have the potential to support development of a biocontrol agent – which defines the fundamental host range. However, such tests can produce ‘false positives’ and reject agents that, under real host selection situations in the field, cause no direct non-target effects. For example, Chrysolina spp. beetles are the basis for New Zealand’s most successful weed biocontrol programme but would probably not be released nowadays because retrospective laboratory host range testing demonstrates that indigenous Hypericum spp are within the agents’ fundamental host ranges. Nevertheless, recent field studies have not shown significant impacts on these indigenous congenerics (Groenteman, Fowler & Sullivan 2011). This example is far from unique: many successful weed biocontrol programmes in the past relied on agent species that would not pass modern regulatory scrutiny despite there being no evidence of harmful post-release non-target effects (H. Hinz, personal communication). For the future then, unless we improve our risk assessment, we will reject potentially successful and safe agents based on overly conservative host range testing. In reality, determining the realised host range (i.e. the plant species that will support agent populations in the field) has proven difficult. Previously, if laboratory tests showed that an agent could develop from egg to adult on a test plant, it was believed that choice tests would better predict the risk of attack on that plant species in the field. This approach has proven unreliable. The ideal testing option is open-field specificity tests conducted in the native range, but these cannot always be performed (e.g. quarantine restrictions may prohibit importing key test plants into the native range of a weed).

We need a framework for assessing whether a fundamental host is likely to become a significant realised host without having to rely on native-range open-field specificity testing. Choice tests have failed to assess the realised host range of agents such as seed feeders, where asynchrony between an agent’s activity and host plant availability can result in ‘no-choice’ situations in the field (e.g. Paynter et al. 2008). Are choice tests reliable for agents that attack non-ephemeral host plant structures (e.g. stems or roots) where phenological asynchrony is unlikely, or can non-target attack result from ‘no-choice’ situations arising when agents disperse away from their normal hosts? Could investigating the influence of deprivation on no-choice testing results predict whether a non-target host is likely to be permanently colonised or only subject to spillover attack (Withers 1997)? More, well-studied examples are needed. In some cases, it may be relatively simple to predict when non-target attack is likely to be minor spillover: Taylor et al. (2007) found that the native leguminous shrub Neptunia major (Benth.) Windler supported the development of the continuously brooded biocontrol agent Neurostrota gunniella (Busck) that targeted the invasive weed Mimosa pigra L. However, N. major died back during the dry season. Therefore, N. gunniella populations could not persist on this native host, and only those individuals growing close to M. pigra were regularly attacked.

Host Range Expansion

Evidence of past host shifts can be seen in the phylogenies of close relatives of biocontrol agents. With some rusts, past host range expansions have even led to the fungus exploiting taxonomically unrelated alternative hosts, often as part of complex life cycles. Do past evolutionary events indicate a higher risk of host range expansions/shifts? Could host range testing help to assess this risk by revealing a ‘ghost of past evolutionary adaptation’ to other, sometimes taxonomically distant, host plant taxa?

The host specificity of biocontrol agents can vary across their native ranges. Does this suggest a greater risk of post-release host range expansion even if only monophagous populations were used to source agents? Would modern host range testing show that the fundamental host range reflected the wider geographic host range rather than local monophagy? To our knowledge, these questions have not been tested.

What else could we test to examine the evolutionary potential for host range expansion? If a very low proportion of larvae survive to adulthood on a non-target plant, compared to the target weed, it seems likely that a population of the candidate agent would not be a threat to the non-target plant in the field. However, could post-release selection result in a higher proportion of larvae that are capable of maturing on non-target hosts? Should we host-range test the offspring of individuals that survived on non-target hosts to determine the potential for the evolution of improved performance on non-target hosts? How many generations should be considered sufficient?

Another issue is whether oligophagous weed biocontrol agents are more likely to expand their host ranges after release compared with tightly evolutionarily constrained monophagous species. Where target weeds have no valued close taxonomic relatives in New Zealand, we have used oligophagous biocontrol agents rather than strictly monophagous species. Indeed, if the target weed has close alien relatives that are current or potential weeds, we have deliberately utilised non-target damage. The release of these ‘multi-targeting’ biocontrol agents is intended to attack both the primary target weed and a range of closely related alien plant species that might currently be minor weeds or perhaps not even have naturalised yet. In some cases, this might include ornamental plants, particularly if they are not widely used or could be readily protected from released biocontrol agents. This ‘multi-targeting’ is proactive and economically prudent, but is it safe?

We end this section with the one example we know where post-release host range expansion is intended. Multiple blackberry rust strains were released in Australia, so that adaptation to different subspecific host taxa will occur (Morin et al. 2006). The risk assessment assumption here is that adaptation will remain constrained within the Rubus fruticosus L. aggregate and not expand onto indigenous Rubus spp. Time will tell whether such a strategy is wise.

Avoiding undesirable indirect non-target effects

  1. Top of page
  2. Introduction
  3. Host specificity testing and risk to non-target plant species
  4. Avoiding undesirable indirect non-target effects
  5. Final remarks
  6. Acknowledgements
  7. References
  8. Biosketch

Recently, the debate over the safety of weed biocontrol agents has moved from direct effects on non-target plants to indirect non-target effects, for example, via interactions in food webs. A high profile example has been the increase in abundance of deer mice Peromyscus maniculatus (Wagner) in rangeland in the USA from feeding on the introduced knapweed gall fly Agapeta zoegana (L.) (Pearson & Callaway 2008). Higher deer mice populations can then cause negative effects on native plants by destroying higher proportions of their seed. Furthermore, as deer mice are the main natural reservoir of hantavirus, the levels of this virus were shown to increase, with potential impacts on humans in which the virus causes a rare but serious disease (Pearson & Callaway 2006). In an entomological example, an introduced tephritid seed fly Mesoclanis polana (Munro) appeared to harm a native insect community by causing local increases in shared native natural enemies (Carvalheiro et al. 2008). More such examples are needed to give a better basis for generalisations. For example, we do not know whether the deer mice–gall fly interaction is unusual, although the potential link to human health seems likely to be.

From here on, we concentrate on food webs, but other non-trophic interactions (e.g. direct competition, pollination or other mutualisms) could contribute to indirect non-target effects. First, we look at practical and theoretical food web science and ask how this might be relevant to predicting indirect non-target effects from weed biocontrol agents. Then, we move onto issues of scale of effects and uncertainty in the context of risk assessment.

Food Web Research and Weed Biocontrol

In a review of insect biocontrol, Hawkins et al. (1999) concluded that biological control successes were most often characterised by artificially simplified food webs, typical of productive sector systems (where biocontrol was more successful per se), in contrast to ‘natural control’ that resulted from multiple links in complex food webs. Recently, biocontrol of weeds has moved from a focus on productive sector weeds to weeds in the natural environment. Does this suggest that biological control of weeds in more complex natural environments is likely to be even less successful than it has been against productive sector weeds, because agents are subjected to ‘multiple links in more complex food webs’ (Hawkins et al. 1999)? Would we expect biocontrol agents released against environmental weeds to cause more indirect non-target effects than those released against economic weeds?

Agents released against environmental weeds are likely to be in closer proximity to native biota and may well have more food web links with this biota. If success rates are also reduced, then there may be more agent species that fail to suppress the target weed but remain reasonably abundant (Pearson & Callaway 2005). However, we are unsure whether a greater number of food web links will necessarily reduce the effectiveness of weed biocontrol agents. For example, insect herbivores do not necessarily show reduced population densities when attacked by more species of natural enemies, provided intra-guild predators are also present (Denno & Finke 2006). Moreover, interspecific plant competition can enhance the impact of biocontrol (Sheppard 1996), and environmental weeds appear to be easier targets for biocontrol than agricultural weeds (Thomas & Reid 2007). Could it be that lower herbivore impacts against environmental weeds are sufficient to achieve weed suppression because these weeds are normally subjected to greater interspecific competition than weeds invading regularly disturbed productive sector systems?

We also note that complex food webs have now been linked to community stability in contrast to the previous argument that complexity caused instability (Polis 1998). If so, then the perturbation caused by introducing a biocontrol agent into a complex food web (to blunt an environmental weed invasion) might be expected to create many minor ripples in the web as opposed to fewer major indirect effects. Overall, this probably concurs with the view of a weed biocontrol practitioner (Harris 1988) that in successful biocontrol ‘the weed is starting to be cycled through various food chains, a sign of ecological health’.

A related issue is whether the strength of a food web perturbation would generally decline with the increasing number of links away from the direct interaction between an introduced biocontrol agent and its host plant. We suspect this would usually be true unless a trophic cascade was set off by impacting a keystone component of a web. A key question then is whether ecological science can help us identify when biocontrol agents might impact on such keystone components of a food web and adjust the risk assessment accordingly.

A vital issue is how to avoid introducing agents that become common but still fail to suppress their target weed. Historically, most insect or pathogen species established for biological control of exotic weeds have failed to impact on their targets. Much effort has been invested in modelling to improve agent climate match and to target the vulnerable life stages of a plant, with arguably limited benefit on biocontrol success. Biotic interference reducing the impact of a biocontrol agent because of itself being parasitised or preyed upon, or interacting unfavourably with epiphytic or endophytic microbiota, has been neglected as a factor that influences biocontrol success. Recent studies of these complex biotic interactions indicate that the likelihood of parasitism (e.g. Paynter et al. 2010) and even the structure of a food web (Veldtman et al. 2011) can be predicted prior to an agent’s release. More work is required to determine whether the likelihood of predation or disease can similarly be predicted and, if so, what the population level impacts are likely to be. There certainly needs to be an attempt to use food webs as a predictive tool for indirect non-target effects in weed biocontrol (Memmott 2009) and then test these predictions once agents are established: both areas where ecologists could assist the practical discipline of weed biocontrol.

The Scale of Indirect Non-target Effects Caused by Biocontrol Agents

The magnitude, and the spatial/temporal extent, of indirect non-target effects is critical for risk assessment in biocontrol. From the above-mentioned discussion of food webs, it is clear that some indirect non-target effects are expected from introducing weed biocontrol, and these will range from positive on some indigenous biota to negative on others. The timing and duration of indirect non-target effects need to be considered in any risk assessment. Do they impact on the valued non-target species throughout the season? Are they likely to decline over time, for example, if the biocontrol programmes become successful (because of either the agent concerned or the addition of further agent species)? With the knapweed gall fly example, it appears that spotted knapweed Centaurea stoebe L. may finally be coming under successful biological control in some areas (e.g. Story et al. 2008), which should prevent the unwanted increases in deer mice populations (Van Driesche, Hoddle & Center 2008).

More case studies of indirect non-target impacts of introduced insects and pathogens as weed biocontrol agents are probably needed before valuable generalisations emerge. Whether microbial or insect focussed, we urge that future case studies take a holistic approach to risk assessment, considering spatial and temporal scales as well as the straightforward magnitude of negative (or positive) effects. Overall of course, risk assessment needs to consider the impact of the status quo with the invasive weed. What is needed is a more holistic view – more of an environmental balance sheet. For example, Pearson & Callaway (2008) acknowledge that the indirect effects of the introduced knapweed gall fly on native forb seedlings (from higher populations of deer mice) were dwarfed by the direct negative effects of herbicide used to control spotted knapweed C. stoebe.

In another example, a highly host plant–specific weed biocontrol agent, recently introduced into Australia to control bitou bush Chrysanthemoides monilifera (L.) Nordlindh, is associated with declines of local insect communities (Carvalheiro et al. 2008). As the agent shares natural enemies (predators and parasitoids) with seed herbivore species from native plants, this study strongly implicated locally significant apparent competition causing negative effects on indigenous seed feeding insects. However, for risk assessment, we need to know whether this was localised within a few metres of invading bitou bush or more widespread throughout the natural range of the impacted indigenous insects? Work on direct non-target attack has shown that spillover effects can be quite localised (e.g. Taylor et al. 2007) – is the same true for food web effects?

If enough ecological studies have been carried out, then we should be informed about the current/potential impact of the weed and know that doing nothing is an option that will result in strong negative effects. Alternatives to biocontrol always exist of course, but in our experience, investment in weed biocontrol is seldom chosen unless these alternatives are expensive or involve unacceptable non-target damage.

Final remarks

  1. Top of page
  2. Introduction
  3. Host specificity testing and risk to non-target plant species
  4. Avoiding undesirable indirect non-target effects
  5. Final remarks
  6. Acknowledgements
  7. References
  8. Biosketch

We hope that the above-mentioned discussion demonstrates many of the areas where the practical discipline of weed biocontrol could benefit from appropriate ecological science. There are relatively few biocontrol practitioners in the world, and often political/logistical/funding constraints to funding ecological science within operational weed biocontrol programmes. However, as mentioned almost ad nauseum in past reviews, biological control systems offer remarkable experimental systems for ecologists.

On the positive side, biocontrol practitioners can forge links between the academic end of ecology and on-the-ground stakeholders, and there are areas where the melange of pure and applied science has worked well, for example, release strategies and establishment success. We want to see more such mixing – ecological science should not be the icing on the applied weed biocontrol cake; we prefer a layered Black Forest gateau approach and believe there is an abundance of fascinating ecological research to be undertaken alongside weed biocontrol programmes.

Acknowledgements

  1. Top of page
  2. Introduction
  3. Host specificity testing and risk to non-target plant species
  4. Avoiding undesirable indirect non-target effects
  5. Final remarks
  6. Acknowledgements
  7. References
  8. Biosketch

Lynley Hayes, Matt McGlone, John Hoffmann, Phil Hulme and an anonymous reviewer improved the manuscript. Funding was provided by FRST contract C09X0905.

References

  1. Top of page
  2. Introduction
  3. Host specificity testing and risk to non-target plant species
  4. Avoiding undesirable indirect non-target effects
  5. Final remarks
  6. Acknowledgements
  7. References
  8. Biosketch
  • Carvalheiro, L.G., Buckley, Y.M., Ventim, R., Fowler, S.V. & Memmott, J. (2008) Apparent competition can compromise the safety of highly specific biocontrol agents. Ecology Letters, 11, 690700.
  • Denno, R.F. & Finke, D.L. (2006) Multiple predator interactions and food-web connectance: implications for biological control. Trophic and Guild Interactions in Biological Control (eds J. Brodeur & G. Boivin), pp. 4570. Springer, Dordrecht, The Netherlands.
  • Groenteman, R., Fowler, S.V. & Sullivan, J.J. (2011) St. John’s wort beetles would not have been introduced to New Zealand now: a retrospective host range test of New Zealand’s most successful weed biocontrol agents. Biological Control, 57, 5058.
  • Harris, P. (1988) Environmental impact of weed-control insects. BioScience, 38, 542548.
  • Hawkins, B.A., Mills, N.J., Jervis, M.A. & Price, P.W. (1999) Is the biological control of insects a natural phenomenon? Oikos, 86, 493506.
  • Memmott, J. (2009) Food webs: a ladder for picking strawberries or a practical tool for practical problems? Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 364, 16931699.
  • Morin, L., Aveyard, R., Batchelor, K.L., Evans, K.J., Hartley, D. & Jourdan, M. (2006) Additional strains of Phragmidium violaceum released for the biological control of blackberry. 15th Australian Weeds Conference, Papers and Proceedings, Adelaide, South Australia, 24–28 September 2006: Managing weeds in a changing climate, 565568.
  • Paynter, Q., Gourlay, A.H., Oboyski, P.T., Fowler, S.V., Hill, R.L., Withers, T.M., Parish, H. & Hona, S. (2008) Why did specificity testing fail to predict the field host-range of the gorse pod moth in New Zealand? Biological Control, 46, 453462.
  • Paynter, Q., Fowler, S.V., Gourlay, A.H., Groenteman, R., Peterson, P., Smith, L. & Winks, C.J. (2010) Predicting parasitoid accumulation on biological control agents of weeds. Journal of Applied Ecology, 47, 575582.
  • Pearson, D.E. & Callaway, R.M. (2005) Indirect nontarget effects of host-specific biological control agents: implications for biological control. Biological Control, 35, 288298.
  • Pearson, D.E. & Callaway, R.M. (2006) Biological control agents elevate hantavirus by subsidizing deer mouse populations. Ecology Letters, 9, 443450.
  • Pearson, D.E. & Callaway, R.M. (2008) Weed biocontrol insects reduce native-plant recruitment through second-order apparent competition. Ecological Applications, 18, 14891500.
  • Polis, G.A. (1998) Ecology: stability is woven by complex webs. Nature, 395, 744745.
  • Sheppard, A.W. (1996) The interaction between natural enemies and interspecific plant competition in the control of invasive pasture weeds. Proceedings of the IX International Symposium on Biological Control of Weeds (eds V.C. Moran & J.H. Hoffman), pp. 4753. University of Cape Town, Rondebosch, South Africa.
  • Story, J.M., Smith, L., Corn, J.G. & White, L.J. (2008) Influence of seed head-attacking biological control agents on spotted knapweed reproductive potential in western Montana over a 30-year period. Environmental Entomology, 37, 510519.
  • Taylor, D.B.J., Heard, T.A., Paynter, Q. & Spafford, H. (2007) Nontarget effects of a weed biological control agent on a native plant in Northern Australia. Biological Control, 42, 2533.
  • Thomas, M.B. & Reid, A.M. (2007) Are exotic natural enemies an effective way of controlling invasive plants? Trends in Ecology & Evolution, 22, 447453 (supplementary file).
  • Van Driesche, R., Hoddle, M. & Center, T.D. (2008) Control of Pests and Weeds by Natural Enemies: An Introduction to Biological Control. Blackwell Publishing, Malden, MA.
  • Veldtman, R., Lado, T.F., Botes, A., Procheş, Ş., Timm, A.E., Geertsema, H. & Chown, S.L. (2011) Creating novel food webs on introduced Australian acacias: indirect effects of galling biological control agents. Diversity and Distributions, 17, 958967.
  • Withers, T.M. (1997) Changes in plant attack over time in no-choice tests: an indicator of specificity. New Zealand Plant Protection, 50, 214217.

Biosketch

  1. Top of page
  2. Introduction
  3. Host specificity testing and risk to non-target plant species
  4. Avoiding undesirable indirect non-target effects
  5. Final remarks
  6. Acknowledgements
  7. References
  8. Biosketch

Simon Fowler and Quentin Paynter are entomologists with over 20-year experience in weed biocontrol in the UK, Australia, USA and New Zealand (NZ). Sarah Dodd has 19-year experience in plant pathology and biological control. Ronny Groenteman obtained her PhD in NZ in 2008, studying multi-targeting biocontrol agents. The authors have experience with classical biological control using >50 agent species on >20 weeds affecting both the productive sector and the natural environment in temperate and tropical regions.