Over the past 30 years, protocols for the selection of test plants used to determine the host range of candidate biological control agents have remained largely unchanged. Using the case of the root-feeding flea beetle, Longitarsus sp., a candidate agent for biological control of Heliotropium amplexicaule in Australia, this paper describes a "modernized" protocol, based more strongly on phylogeny, and refined by ecological and biogeographic similarities. Taxonomic nomenclature is de-emphasized in favour of strict phylogenetic relationships and the use of so-called "safeguard species" is abandoned. This is the first time that a biological control agent has been tested for host-specificity and application made for release, based solely on the new protocol, and the changes were acceptable to the regulatory organisations in Australia. The testing showed that adult feeding extended to plant species with up to five degrees of phylogenctic separation from H. amplexicaule, indicating that there would be a moderate risk that more distantly related plants suffer some feeding damage by adult Longitarsus sp. when they co-occur with infestations of the target weed that have large flea-beetle populations. Such damage would not have severe consequences for survival and reproduction of the plants. However, Longitarsus sp. was able to complete its life-cycle on plants related to the target weed by two degrees of phylogenetic separation or less, leaving indigenous Heliotropium and Tournefortia species at some risk of colonisation. While these species had different life-histories and/or only slightly overlapped with the actual and potential range of the target weed, a minority of reviewers were concerned that insufficient information was available on the dispersal abilities of Longitarsus sp. to dismiss this risk. Release was therefore not approved. While disappointing for the biological control project, the outcome was not unexpected, as the assessment was based on factors that modified the effects of host range alone. The new protocols highlighted some problems in the review process concerning an overreliance on taxonomic nomenclature as opposed to actual genetic relationships and an inadequate understanding of the nature of risk. However, they also directed attention to knowledge gaps in biogeography and agent biology that might refine the assessed risk. As such, this process can be considered an improvement in methodology. Moreover, rejection of the application to release Longitarsus sp. demonstrated that fears expressed by some researchers/regulators that the removal of "safeguard" species would somehow weaken the data and allow unsafe agents to be released were unwarranted. Crown copyright (C) 2008 Published by Elsevier Inc. All rights reserved.
