Rivulus, Poey, 1860

RIVULUS View in CoL EMBRYONIC DEVELOPMENT

Through observation of embryos over the time course of development, we were able to confirm that several named and currently unnamed Rivulus species exhibit the dispersed cell phase, diapause II and delayed hatching/diapause III. These results are presented in Table 2 and described briefly below.

No embryos from any species exhibited diapause I. However, a brief dispersed cell phase separating epiboly from the formation of the embryonic axis (Supporting Information, Figs S4–S View Figure 4 6 View Figure 6 ) was observed in all species of the Owiyeye clade of Laimosemion that were examined, but not the outgroup taxa, Rivulus (Laimosemion) xiphidius (Supporting Information, Fig. S3 View Figure 3 ), Rivulus (Laimosemion) gransabanae , Rivulus (Anablepsoides) hartii (Supporting Information, Figs S1 View Figure 1 , S 2 View Figure 2 ) and K. marmoratus . The dispersed cell period was short, lasting between 1 and 7 days. In other annual species, harsh environmental conditions, such as lack of oxygen, chemical factors produced by adult fish and cold temperatures (either alone or in some combination), tend to induce diapause I during the dispersed cell phase, whereas in benign conditions (such as those of a laboratory with regular embryo collection and incubation in fresh medium) it is a less frequent and lengthy occurrence ( Levels & Denuce, 1988; Podrabsky et al., 2010b; Furness et al., 2015a). Although diapause I was not observed during the dispersed cell phase in any of the Rivulus species studied here, it is possible that diapause I might occur given different harsher incubation conditions.

Diapause II was clearly observed in the R. (L.) tecminae complex (sp. Ventuari, tecminae and sp. Atabapo) and Rivulus (L.) sp. Maroa ( Figs 3 View Figure 3 , 4 View Figure 4 ), but not in R. (L.) tomasi , R. (L.) rectocaudatus , R. (L.) xiphidius , R. (L.) gransabanae , R. (A.) hartii and K. marmoratus . Pairwise tests of the proportion of embryos that exhibited diapause II were performed; all pairwise comparisons among species with and without diapause II were highly significant (P <0.0001). Embryos that halted development at the diapause II stage did so when they reached 38–42 somite pairs, the same stage at which other annual killifish from South America and Africa halt development ( Wourms, 1972c). Of the species that exhibited diapause II, only R. sp. Atabapo exhibited variation among embryos in whether diapause II was entered (30 of 37 embryos exhibited diapause II and seven of 37 skipped diapause II). These alternative developmental trajectories have been termed the ‘diapause’ and ‘direct-developing’ or ‘escape’ pathways ( Podrabsky et al., 2010a; Furness et al., 2015b). Furness et al. (2015b) demonstrated that R. (L.) sp. Atabapo embryos following the direct-developing pathway exhibit significantly more robust head morphologies relative to embryos at an identical stage of development that entered diapause II ( Fig. 5 View Figure 5 ; supplementary material of Furness et al., 2015b). This divergence in morphology began at ~20 somites, well before the 38–42 somite stage at which development was arrested in diapause II embryos.

All species examined had embryos capable of surviving on peat moss without hatching for> 10 days after development was complete. This is probably the case because such conditions are generally unsuitable for hatching and therefore required for survival. This phenomenon could therefore realistically be called delayed hatching. The degree to which hatching is postponed when it could presumably occur at any time (i.e. water incubation) might therefore be a more realistic measure of the potential for diapause III. Among species, there was significant variation in the proportion of water-incubated embryos that delayed hatching for> 10 days (general linear model analysis with binomial distribution: χ 2 = 59.6, P <0.0001) and in the duration of this delay (generalized least-squares analysis: F = 7.3, P <0.0001; Fig. 3 View Figure 3 ). However, there was no apparent phylogenetic trend to this variation given that the longest delays were observed in R. (L.) sp. Atabapo and the outgroup taxa K. marmoratus .

PHYLOGENETIC ANALYSES

The combined alignment included a total of 3462 bp, of which 1923 bp were constant, 293 bp were variable but parsimony uninformative, and 1246 bp were parsimony informative. Figure 6 View Figure 6 shows the maximum likelihood phylogeny and Supporting Information, Figure S8 the maximum parsimony phylogeny (5048 steps). The maximum likelihood and maximum parsimony trees were largely congruent (identical topologies with regard to the ingroup taxa sequenced in the present study) but with differences in the placement of Rivulus (Laimosemion) kirovskyi and some outgroup Laimosemion taxa derived from GenBank. The sub-genus Laimosemion was recovered as strongly monophyletic. Within Laimosemion , there was a basal split between a clade that has been named Owiyeye ( Costa, 2006) that includes the species of interest with regard to the evolution of diapause, and the rest of Laimosemion .

ANCESTRAL STATE RECONSTRUCTIONS

Figure 7 View Figure 7 shows parsimony ancestral state reconstructions for diapause II. The parsimony reconstruction suggests that diapause II evolved in the common ancestor of R. (L.) sp. Ventuari, R. (L.) tecminae , R. (L.) sp. Atabapo and R. (L.) sp. Maroa. Maximum likelihood reconstructions also support this general evolutionary scenario, but with considerably less certainty (Supporting Information, Table S5).