Accuracy of nonmolecular identification of growth-hormone-transgenic coho salmon after simulated escape

Authors

  • L. F Sundström,

    1. DFO Centre for Aquaculture and Environmental Research, 4160 Marine Drive, West Vancouver, British Columbia V7V 1N6 Canada
    Search for more papers by this author
    • Present address: Uppsala University, Evolutionary Biology Centre, Department of Ecology and Genetics/Animal Ecology, Norbyvägen 18D, SE-75236 Uppsala, Sweden.

  • M. Lõhmus,

    1. DFO Centre for Aquaculture and Environmental Research, 4160 Marine Drive, West Vancouver, British Columbia V7V 1N6 Canada
    Search for more papers by this author
    • Present address: Karolinska Institute, Institute for Environmental Medicine, Nobels väg 13, SE-17177 Stockholm, Sweden.

  • R. H. Devlin

    Corresponding author
    1. DFO Centre for Aquaculture and Environmental Research, 4160 Marine Drive, West Vancouver, British Columbia V7V 1N6 Canada
    Search for more papers by this author

  • Corresponding Editor: M. E. Hellberg.

Abstract

Concerns with transgenic animals include the potential ecological risks associated with release or escape to the natural environment, and a critical requirement for assessment of ecological effects is the ability to distinguish transgenic animals from wild type. Here, we explore geometric morphometrics (GeoM) and human expertise to distinguish growth-hormone-transgenic coho salmon (Oncorhynchus kisutch) specimens from wild type. First, we simulated an escape of 3-month-old hatchery-reared wild-type and transgenic fish to an artificial stream, and recaptured them at the time of seaward migration at an age of 13 months. Second, we reared fish in the stream from first-feeding fry until an age of 13 months, thereby simulating fish arising from a successful spawn in the wild of an escaped hatchery-reared transgenic fish. All fish were then assessed from photographs by visual identification (VID) by local staff and by GeoM based on 13 morphological landmarks. A leave-one-out discriminant analysis of GeoM data had on average 86% (72–100% for individual groups) accuracy in assigning the correct genotypes, whereas the human experts were correct, on average, in only 49% of cases (range of 18–100% for individual fish groups). However, serious errors (i.e., classifying transgenic specimens as wild type) occurred for 7% (GeoM) and 67% (VID) of transgenic fish, and all of these incorrect assignments arose with fish reared in the stream from the first-feeding stage. The results show that we presently lack the skills of visually distinguishing transgenic coho salmon from wild type with a high level of accuracy, but that further development of GeoM methods could be useful in identifying second-generation fish from nature as a nonmolecular approach.

Ancillary