CABI Bioscience, UK Centre (Ascot), Silwood Park, Ascot, Berks SL5 7TA, UK

Biological control of weeds using fungal pathogens is a relatively new field in plant pathology, and therefore risk assessment can be said to be still on a learning curve. Two broad approaches are distinguished: classical biological control for exotic or alien weeds, involving the release of co-evolved pathogens from the centre of origin of the target weed; and inundative control typically for use against endemic weeds, involving the mass production and application of indigenous pathogens as formulated products (mycoherbicides).

Classical biological control
This is based on the same principles that have been established and refined for insect biocontrol agents over the past 100 years, in which co-evolved, natural enemies are selected from the purported native range of the weed, initially employing field observations and taxonomic affinities in order to assess their safety (specificity). However, since all introduced or alien organisms present a potential hazard, the next phase of risk assessment, host specificity screening, is the most critical. This centres on a macroevolutionary approach using a centrifugal, phytogenetic testing method [1], whereby potential hosts are inoculated under predetermined optimum conditions for infection and assessed macroscopically. In addition, detailed studies of host-pathogen interactions are studied at the microscopic level using a clear staining technique. The resistance mechanisms in operation are discussed in relation to the rusts Maravalia cryptostegiae and Puccinia abrupta, and at least five reaction types are identified. These data are presented to the relevant quarantine authorities for risk evaluation, and as additional evidence to support the introduction of the potential agent. The ambiguous (equivocal) results often obtained from these greenhouse-based screens are analysed: typically these involve an artificial extension of the pathogen's host range, and it is concluded that extreme care should be exercised in their interpretation in order to avoid rejection of potentially useful agents. Simulation of more natural infection conditions, using a wind tunnel, and comparison with well documented crop-pathogen associations, can be employed to clarify such equivocal results and to demonstrate the inherent safety (specificity and stability) of the classical biological control approach. However, despite a high success rate and a perfect safety record, this weed management strategy is still viewed with unfounded scepticism, tinged with pathophobia.

Inundative control
This approach usually involves the selection of endemic, necrotrophic pathogens and thus strict host specificity is not a necessary prerequisite, since they occur naturally within the area of potential use and, like chemical herbicides, are site-directed. Nevertheless, detailed data on host range are still required for the risk assessment in order to predict possible conflicts of interest and to determine in which cropping systems they can be applied. The first registered product, DeVine, based on Phytophthora paimivora, was known to have a wide host range, including annual crop species, but was assessed to be safe if its use was confined to citrus orchards. Similarly, although Collego was considered initially to be specific to its target weed, Aeschynomene virginica, the pathogen involved, Colletotrichum gloeosporioides f.sp. aeschynomene, was later shown to attack other leguminous genera in expanded host-range tests. Again, the risk assessment determined that the product was safe since it was site-directed to rice cropping systems, where susceptible legumes are never cultivated. These and other examples of mycoherbicide risk assessments are discussed in relation to epidemiology. Since the current trend is to move towards relatively non-specific pathogens, including virulent crop pathogens, a case study of the silverleaf fungus, Chondrostereum purpureum, for control of black cherry in the Netherlands [2], is analysed.

1. Wapshere AJ, 1974. Annals of Applied Biology 77, 201-211.
2. De Jong MD, Scheepens PC, Zadoks JC, 1990. Plant Disease 74, 189-194.