Phenotypic stasis with genetic divergence

Abstract

Whether or not genetic divergence in the short-term of tens to hundreds of generations is compatible with phenotypic stasis remains a relatively unexplored problem. We evolved predominantly outcrossing, genetically diverse populations of the nematode Caenorhabditis elegans under a constant and homogeneous environment for 240 generations and followed individual locomotion behavior. Although founders of lab populations show highly diverse locomotion behavior, during lab evolution, the component traits of locomotion behavior - defined as the transition rates in activity and direction - did not show divergence from the ancestral population. In contrast, transition rates' genetic (co)variance structure showed a marked divergence from the ancestral state and differentiation among replicate populations during the final 100 generations and after most adaptation had been achieved. We observe that genetic differentiation is a transient pattern during the loss of genetic variance along phenotypic dimensions under drift during the last 100 generations of lab evolution. These results suggest that short-term stasis of locomotion behavior is maintained because of stabilizing selection, while the genetic structuring of component traits is contingent upon drift history.

Figure 1 –

A. Experimental design. One hybrid population (A0) was created from the intercross of 16 inbred founders. Six replicate populations were then domesticated to a defined lab environment and after 140 generations one of these (A6140) was the ancestor to six other replicate populations maintained for an extra 100 generations under similar conditions (CA). Inbred lines were derived by selfing hermaphrodites (colored circles) from A6140 and three replicate CA populations at generation 50 and 100 (blue and red). Horizontal lines indicate outbred population samples that were phenotyped. B. Modelling locomotion behavior from component traits, defined by the transition rates between moving forward (F), moving backward (B) or being stationary (S). We consider the 6 independent non-self rates, shown in colored arrows. C. Evolution of locomotion behavior. Stationary frequency in the founders (pink dots) and outbred populations during lab evolution. Colored overlays indicate three stages of lab evolution: hybridization, domestication and focal. Ticks are sampled time points, while colored points during the focal stage indicate populations from which inbred lines were derived. Point mean estimates are shown for 3–6 replicate populations at other generations, with 95% confidence intervals for each one of them. The evolution of the component traits of locomotion behavior in hermaphrodites and males, the transition rates between movement state and direction, can be found in Figures S4 and S5.

Figure 2 –

G-matrix evolution and divergence during the focal stage. A. A6140 G-matrix. Shown are the 15 genetic covariances between transitions rates (top) and six the genetic variances of transition rates (bottom), as bars and dots the 95% and 83% credible intervals (black and red) and mean of the posterior distribution, respectively. “S”, “F”, “B” stand for still, forward and backward movement states, with letter ordering indicating the direction of movement. G-matrices of the CA populations can be found in Supplementary Figure S7. B. Total amount of genetic variance computed as the sum of the G diagonal elements (trace). All observed posterior means differ from the null 95% of posterior means (orange). C. Genetic variance along the phenotypic dimension encompassing most genetic variation in the ancestral A6140 population ($g_{max}$, red mean, 83% and 95% CI). A6140 $g_{max}$ explains 64% of the total genetic variance Table S4). D. The angle ($\mathrm{\Theta }$, see Methods) between the A6140 $g_{max}$ and the first three eigenvectors of the evolved G-matrices ($g_{max},g_2$ and $g_3$ of the CA[1–3]100 populations). There is no alignment between the evolved populations’ first three eigenvectors with the ancestral $g_{max}$. Dots show the mean estimate with bars the 83% and 95% credible interval of the posterior G-matrix distribution. The null expectation was obtained by computing the angle between pairs of random vectors sampled from a uniform distribution.

Figure 3 –

Genetic divergence. A. Spectral decomposition of variation among A6140 and CA[1–3]100 G-matrices. The variance ${\mathit{\alpha }}_{\mathbf{i}}$ associated with the $i$th eigentensor ${\mathbf{E}}_{\mathbf{i}}$ is compared to a null permutation model where variation among matrices is due to sampling (see Methods). Although several eigentensors are different from zero (black bars, 95% credible interval) only the first one, ${\mathbf{E}}_1$, do not overlap the null (red and orange bars, 83% credible intervals). B. The coordinates of the G-matrices in the space of the first eigentensor ${\mathbf{E}}_1$ for each population tested. Absolute values of the coordinates in the first eigentensor represents its contribution to the difference between matrices. C. Contribution of specific transition rate combinations to coordinated changes among G-matrices. The amount of genetic variance in the direction of the greatest variation among all Gs for the first eigenvector of ${\mathbf{E}}_1\left(e_{11}\right)$, for each population. Eigentensor decomposition of the CA[1–3]50 G-matrices, testing for differentiation at generation 50, can be found in Figure S9.

Figure 4 –

Selection surface of locomotion behavior. Canonical analysis of the $\gamma$-matrix shows positive phenotypic dimensions ($y_1-y_3$) of transition rate combinations under disruptive selection (as measured by the eigenvalue $\lambda$), and negative dimension ($y_4-y_6$) under stabilizing selection. Stars show the mode of the posterior empirical distribution (see Methods). These estimates are to be compared to the posterior distribution of null modes (dots and colored bars, the mean and 83% and 95% credible intervals). The $\gamma$-matrix before canonical rotation can be found in Figure S12.

Figure 5 –

G-matrix evolution in the selection surface. Loss of genetic variance along axes $y_2-y_5$, which contain most of the genetic variance in the evolved populations and are under very weak or no selection, is compatible with expectations from genetic drift under the assumption of infinitesimal trait inheritance (dashed lines, for $N_e={10}^3$). We modelled drift as a per generation loss of genetic variance of. $\frac{V_A}{2N_e}$ Along $y_1$ and $y_6$, genetic variance was much reduced relative to the founders of experimental evolution (green). The genetic variance of each canonical axis $y_i$ was obtained by rotation of the original G-matrices, with 95% (grey) and 83% (colored) credible intervals from sampling 400 matrices in the posterior distributions for each G-matrix. Dots show the median estimates. See also Figure Figure S13.