1. Application to release Lathronympha strigana & Chrysolina abchasica into NZ for biological control of Tutsan Hypericum androsaemum

2. Topics 1.Need for biocontrol – already covered by Craig 2.Why two agent species? 3.Risk of non-target attack on native Hypericum spp. I won’t discuss ecosystem effects (e.g. indirect competition mediated by parasitoids) as we feel they have been adequately covered in the application/response to submissions

3. Why two agent species? Increased chance of success: Weed biocontrol agent establishment rate =~84%, so releasing 2 spp. increases chance of getting 1 agent sp. established from 84% to ~97% Main reason: complementary feeding = potentially more impact –Georgian populations of Lathronympha strigana larvae bore into fruits/eat seeds; –Chrysolina abchasica defoliates plants

4. Seed-feeding agents rarely destroy enough seeds to reduce densities of existing weed infestations BUT simulation modelling 1 & field studies 2 indicate that reducing seed production can reduce the rate that weeds spread into novel (previously uninvaded) habitats or reinvade following control (e.g. by herbicides) L. strigana probably won’t control tutsan on its own, but can potentially slow tutsan invasion, making infestations easier to contain & control using conventional means & may be particularly useful if outlying plants escape attack by Chrysolina abchasica (complementary impact) Complementary feeding 1 Coutts, SR et al. 2011. Biological invasions, 13, 1649-1661 2 Norambuena, H., & Piper, G. L. (2000). Biological Control, 17(3), 267-271

5. Complementary feeding 1 Hayes, L. et al., 2013. Biocontrol of weeds: Achievements to date and future outlook. Ecosystem services in NZ–conditions & trends. Manaaki Whenua Press, Lincoln, NZ, 375-385. Good reason to believe Chrysolina abchasica has excellent potential to control existing infestations: Closely-related Chrysolina hyperici (introduced in 1943) successfully controlled St John’s wort Hypericum perforatum in NZ NZ’s most successful biocontrol programme 1 & an almost identical weed/agent combination

6. Potential for non-target attack

7. Host-range testing Species closely-related to the target weed are at biggest risk of non-target attack 1 Modern host-range testing, therefore, follows a “centrifugal” phylogenetic approach (i.e. testing the most closely-related spp., then more distantly related plant spp. until the host-range is circumscribed 2 ) Track record of host-range testing is good 1 Pemberton, RW, 2000. Oecologia 125, 489-494 2 Briese, DT & Walker A 2002. Biological control, 25, 273-287.

8. Track record host-range testing No significant non-target attack on native plants or on economically important exotic spp. in NZ 1,2 Of 512 weed biocontrol agent spp. released worldwide only 4 (0.8%) have serious non-target impacts: all were on thistles or cacti & within the same genus as the target host plant & with predictable outcomes that resulted from an earlier era of lower standards of biosafety than prevail today 3. Host-range testing works! Testing has been improved to reduce potential risks still further 1,2 1 Paynter et al. 2004 New Zealand Plant Protection 57 102-107 2 Fowler et al. 2012. J. Appl. Ecol. 49 307-310 3 Suckling, D. M., & Sforza, R. F. H. 2014. PloS one, 9, e84847.

9. Predicting non-target attack Problems associated with agents that are capable of developing on a potential non-target plant, where they perform relatively poorly on that plant Such plants are termed “physiological” hosts, because they support development, but they may or may not be field hosts in natural conditions A Gung-ho approach to releasing such agents risks non-target attack, but not releasing such agents risks rejecting a safe & effective agent

10. No-choice starvation tests: NZ native Senecio wairauensis is a (poor) physiological host 1 No evidence of non-target attack in the field since 1983 release 2 Rejecting this agent would have needlessly prevented a highly successful programme Example: Longitarsus jacobaeae in NZ 1 Syrett 1985. NZ J. Zool. 12: 335-340 2 Paynter et al. 2004. NZ Plant Prot. 57:102-107

11. Predicting non-target attack on “physiological” hosts Ideally, field tests would be performed in agent native range, but it is often impossible to grow NZ native plants in other countries A recently developed scoring system predicts non-target attack based on relative performance of agent on test plant versus target plant 1 e.g. if an ovipositing ♀ lays 20% of the normal no. eggs on a test plant & subsequent larval survival is 30% of that on the host plant then the relative performance on that test plant (expressed in proportions) is 0.2 Χ 0.3 = 0.06 1 Paynter, Q., et al. 2015. Biological Control 80, 133-142.

12. Relative performance All attack(including spill-over)Relative risk score (combining oviposition & starvation test results) from past NZ biocontrol programmes shows a clear-cut threshold: all plants with a combined risk score > 0.33 were attacked; no plants attacked when the risk score was < 0.33 Paynter, Q., et al. 2015. Biological Control 80, 133-142.

13. Host-range testing L. strigana & C. abchasica Phylogenetic approach predicts biggest risk of undesirable non-target feeding is to 4 native NZ Hypericum spp., - H. rubicundulum; H. minutiflorum; H. pusillum & H. involutum (formerly H. gramineum). Three of these were tested - H. minutiflorum was not tested (hard to obtain & notoriously difficult to grow). According to Peter Heenan (Landcare Research, pers. comm.), H. rubicundulum & H. minutiflorum are genetically almost identical 1, so we would expect almost identical test results. These 2 spp. are ecologically similar too, so the risk profile for these spp. is almost identical – essentially H. rubicundulum can be considered a surrogate for H. minutiflorum. 1 Heenan, PB 2008 Three newly recognised species of Hypericum (Clusiaceae) from NZ, NZ J. Botany, 46:4, 547-558

14. A B C D E H. androsaemum H. calycinum NZ natives Phylogeny of genus Hypericum: Meseguer et al. / Molecular Phylogenetics & Evolution 67 (2013) 379–403 NZ native Hypericum spp. belong to clade B; H. androsaemum belongs to clade C. Other commonly naturalised Hypericum spp. in NZ included in testing belong to clades D & E. Clade B diverged from Clades C-E ~34 MYA Clades C, D & E diverged ~27 MYA, Closest relatives to H. androsaemum in NZ in order of relatedness are exotic H. perforatum & H. calycinum, then NZ natives 34 MYA 27 MYA H. perforatum

15. Summary of host-range testing Lathronympha strigana: –Oviposition & larval development restricted to Genus Hypericum. –Both oviposition & no-choice development/larval starvation tests indicate little risk to native Hypericum spp. (no or v few eggs laid; no larval development). Chrysolina abchasica: –Oviposition & larval development restricted to Genus Hypericum. –No oviposition or larval development on native H. involutum (not a host). –Significantly lower oviposition & larval survival on native H. pusillum & H. rubicundulum compared to H. androsaemum (tutsan) & low survival of resulting adults indicate these spp. are very poor hosts.

16. Host-range testing Chrysolina abchasica All attack(including spill-over) Relative risk score for C. abchasica ranged from 0.01-0.06 for H. pusillum & H. rubicundulum – well below threshold for non- target attack - i.e. predicts that C. abchasica populations will not colonise & persist on native NZ Hypericum spp. & even spill- over (incidental feeding on non-target plants that does not persist or only occurs in the presence of the target) is unlikely

17. Risk to native Hypericum spp. Although the scoring system predicts that spill-over attack is unlikely, we nevertheless examined this risk in more detail: 2 scenarios how spill-over might occur: 1.Larvae dispersing from defoliated tutsan plants might feed on native Hypericum plants (C. abchasica larvae are not very mobile so this could only occur in v close proximity to the source plant) 2.High adult densities may result in oviposition on native Hypericum plants growing near tutsan, with subsequent larval feeding. The risk of damage would fall off rapidly with distance, so is also likely to be relatively localised (few beetles likely to disperse > a few hundred m).

18. Risk of spill-over attack Significant spill-over can only occur if: 1.The host plant & the non-target species co-exist regionally 2.The non-target plant could attract & arrest the dispersal of adult C. abchasica 3.Non-target damage to plants affects their survival status 4.There is spill-over risk from host plants spp. other than tutsan (e.g. other invasive Hypericum spp.)

19. 1. Do host plant & non-target spp. co-exist regionally Tutsan is a common weed in higher rainfall areas, esp N & W NZ, but is less suited to, or absent from drier & cooler eastern or upland regions H. pusillum has closest range overlap with tutsan, but tolerates a wider range of habitats & is commonly found where tutsan is absent. H. rubicundulum occurs in dry areas of the S Island (& 1 site in Hawkes Bay). Range overlap with H. androsaemum minimal. H. minutiflorum almost exclusively grows on pumice soils in the Taupo basin & central volcanic plateau where tutsan is rare. Range overlap with H. androsaemum also minimal

20. 2. Could non-target plants attract dispersing adult C. abchasica? During oviposition tests, adult beetle feeding was recorded as well as the no. eggs laid Adults did not feed/only trivial feeding on H. involutum & H. rubicundulum indicating that adults encountering these spp. in the field would likely re-disperse rather than settle & feed. Hypericum minutiflorum was not tested, but a similar response is probable.

21. 3. Damage to non-target plants affected survival status of the non-target species Photographs were taken following the completion of larval starvation/development tests Larvae caused heavy damage on H. androsaemum (top) Barely discernible damage by 11 larvae on H. rubicundulum (middle) & ditto for 12 larvae on H. pusillum (bottom)

22. 4. Spill-over risk from host plants species other than tutsan The most closely-related plant to tutsan in NZ is hybrid H. x inodorum (clade C; parent plants: tutsan & H. hircinum). Given it’s derivation, it is likely to be a host. Not included in testing as it is rarely naturalised (Mainly Auckland & Northland: little overlap with native Hypericum spp.) Hypericum canariense is another naturalised sp. belonging to clade C. It is restricted to a roadside near Gisborne (almost no overlap with native Hypericum spp.) St John’s wort (H. perforatum) is a fundamental host for C. abchasica, but performance scores (0.17-0.21) predict that it will not be field host in NZ – backed up by field surveys in Georgia, where tutsan & St John’s wort co-occur that indicate St John’s wort is not a host there.

23. Relative performance Finally, it is noteworthy that retrospective testing performed on Chrysolina hyperici & C. quadrigemina, which have trivial impacts on H. pusillum & H. involutum in NZ 1, indicates much higher equivalent combined risk scores than for C. abchasica – further evidence that C. abchasica is likely to be safe Groenteman, R., et al. 2011. St. John's wort beetles would not have been introduced to NZ now: A retrospective host range test of NZ's most successful weed biocontrol agents. Biological Control 57, 50-58.