Results
The control goldfish group had 100% establishment success and a mean abundance of 17.0 ± 13.2. Of the 12 fish species, 8 were infected byG. kobayashii with 6.7-83.3% attachment success (Table 1). But only C. auratus gibelio became infected with 56.7% establishment success and 3.0 ± 4.4 mean abundance (Table 1), which were both significantly lower than those on the control group (P< 0.05). The silver crucian carp is defined as a low-competence host, while the other species with no establishment success are considered as incompetent hosts.
In the mixed-species culture, no G. kobayashii was detected on grass carp and swordtail (Fig 1). Compared with the goldfish in the control group (85.8 ± 25.1), the mean abundance of G. kobayashiion fins of goldfish was lower in the swordtail group (70.0 ± 22.2), the multi-species group (75.9 ± 30.8) and the silver crucian carp group (30.0 ± 16.5), but higher in the grass carp group (116.1 ± 33.2) (Table 2). There were significant differences in mean abundance on goldfish between the silver crucian carp group and all other groups (P< 0.05) (Fig 1). Mean abundance of G. kobayashii onC. auratus gibelio in the silver crucian carp group (7.1 ± 5.6, 66.0% prevalence) and the multi-species group (10.9 ± 3.8, 86.7% prevalence) were both significantly lower than that on C. auratusin the corresponding group (P < 0.05). Total population size of G. kobayashii was 857.8, 370.4, 1160.8, 698.8 and 791.4 in the five groups, respectively. At the host community level, the mean abundance was 18.5 (± 10.7) and 39.6 (± 15.1) in the silver crucian carp and the multi-species groups, respectively (Table 2). A significant difference in total mean abundance on all hosts was detected between the control group and the silver crucian carp group (P < 0.05) and the multi-species group (P < 0.05).
Observation of schooling patterns showed that population of goldfish mixed well with the silver crucian carp, but not with the grass carp and the swordtail (S Video). The average adjacency distance among goldfish significantly increased, while the average speed and cumulative contact time significantly decreased in all mixed groups when compared with those in control group (P < 0.05) (Fig 2). The trajectory heatmap showed that the goldfish tended to school along the tank wall both in the control and silver crucian carp groups. In contrast, goldfish were uniformly distributed in the whole tank in the grass carp group and in part of the tank in the swordtail group (Fig 3). A schematic model of the spatial distribution of the fishes under the different experimental conditions is presented in Fig 4, with the school of goldfish alone (Fig 4A), the goldfish school mixed with low-competence silver crucian carp (Fig 4B), and goldfish schooling separately from incompetent grass carp (Fig 4C).