The Many Nuanced Evolutionary Consequences of Duplicated Genes

Abstract

Gene duplication is seen as a major source of structural and functional divergence in genome evolution. Under the conventional models of sub or neofunctionalization, functional changes arise in one of the duplicates after duplication. However, we suggest here that the presence of a duplicated gene can result in functional changes to its interacting partners. We explore this hypothesis by in silico evolution of a heterodimer when one member of the interacting pair is duplicated. We examine how a range of selection pressures and protein structures leads to differential patterns of evolutionary divergence. We find that a surprising number of distinct evolutionary trajectories can be observed even in a simple three member system. Further, we observe that selection to correct dosage imbalance can affect the evolution of the initial function in several unexpected ways. For example, if a duplicate is under selective pressure to avoid binding its original binding partner, this can lead to changes in the binding interface of a nonduplicated interacting partner to exclude the duplicate. Hence, independent of the fate of the duplicate, its presence can impact how the original function operates. Additionally, we introduce a conceptual framework to describe how interacting partners cope with dosage imbalance after duplication. Contextualizing our results within this framework reveals that the evolutionary path taken by a duplicate's interacting partners is highly stochastic in nature. Consequently, the fate of duplicate genes may not only be controlled by their own ability to accumulate mutations but also by how interacting partners cope with them.

Fig. 1.

Selection schemes. We simulate evolution under seven different selection schemes, five of which include the duplication of protein B, denoted as B′. All simulations assume selection for the stability of A, B, and B′ if applicable. Two simulations assume that a duplication event does not occur. Selection scheme 1 (no duplication, ND) describes a scenario without duplication where the stability of the A–B interface is included in the fitness function. Selection scheme 2 (ND no bind) describes a scenario without duplication where the stability of the A–B interface is not included in the fitness function. To examine how the duplication of a subunit affects evolutionary dynamics, we consider five additional selection schemes. Selection scheme 3 (bind both) describes a scenario where both duplicates (B and B′) need to bind A, and the stability of both the A–B and the A–B′ interface is included in the fitness function. This type of selection pressure could occur in a situation where increased dosage of B is beneficial. Selection scheme 4 (bind max) describes a competition scenario where the stability of only one interface, that of the maximum stability of binding for either the A–B or the A–B′ interface, is included in the fitness function. Selection scheme 5 (bind B) describes the process of B′ nonfunctionalization, and the stability of only the A–B interface is included in the fitness function. Selection scheme 6 (bind B and not B′) describes diversifying selection, and the stability of the A–B interface is included in the fitness function whereas the stability of the A–B′ interface is used as a fitness penalty. This sort of selection scenario mimics that of dosage imbalance, where an excessive amount of unbound B′ is harmful. Selection scheme 7 (no bind) describes a control duplication experiment where the stability of neither the A–B nor the A–B′ interface is included in the fitness function.

Fig. 2.

Stability of the A–B interface versus time, for all seven selection schemes of figure 1. $\Delta G$ values are averaged over replicate simulations. Under selection schemes that do not reward binding, the interface stability tends to decay rapidly (more positive $\Delta G$ indicates less stable binding). In contrast, selection schemes that do reward binding tend to maintain the interface stability throughout. One exception to this pattern is the bind B and not B′ case, which first shows rapid destabilization of the A–B interface, followed by subsequent regaining of binding stability.

Fig. 3.

Significantly differing sites after adaptation to different selection schemes. The highlighted sites differ significantly in their amino acid composition between the bind both and the bind B and not B′ simulations. These sites include site 20 in protein A and sites 9, 44, 46, 47, and 68 in protein B′. No sites were found to significantly differ in their amino acid composition in protein B.

Fig. 4.

Distribution of stickiness for sites with differing amino acid compositions, shown at every 100th generation. Grayed-out distributions indicate time points at which sites do not display significant differences. (A) Site 20 in protein A. Whereas the stickiness of site 20 in protein A in the bind B and not B′ scenario is initially reduced, some stickiness is regained and by the end of the simulation the distributions of residue stickiness are similar for the two selection schemes. This observation suggests that once dosage imbalance is escaped, the A–B interface is refined and reoptimized. (B) Site 9 in protein B′. This site displays a slight shift toward more sticky residues under selection to avoid binding B′, though a larger portion of this distribution still resembles the bind both scenarios. However, distributions differ significantly only from generation 400 to generation 1,100, and this transient behavior may be related to the restabilization of the B′ structure. (C) Site 44 in protein B′. For this site, the differing amino acid composition between the two selection schemes does not appear to be reflected in the distribution of amino acid stickiness. (D) Site 46 in protein B′. The dynamics at this site suggest that selection to avoid binding B′ initially shifts the distribution towards stickier residues. Interestingly, this shift also occurs under the bind both selection scheme, just later in time. It appears that selection to avoid binding B′ results in an accelerated exploration of sequence space. (E) Site 47 in protein B′. (F) Site 68 in protein B′. Sites 47 and 68 increase in stickiness under selection to avoid binding B′. However, this effect sets in only around generation 500, indicating that these changes are related to the restabilization of B′.

Fig. 5.

All possible ways in which the proteins in a three-member network could adapt after duplication. Grayed-out structures indicate no adaptation and structures in full color indicate diversifying change. In options 1 and 2, only one of the duplicated interacting partners acquires diversifying changes. Option 3 describes a situation where only the nonduplicated interacting partner accumulates diversifying changes. Option 4 describes the case where both of the duplicated genes acquire diversifying changes. In options 5 and 6, the nonduplicated partner and one of the duplicate partners accumulate diversifying changes. Finally, in option 7 all proteins and in option 8 none of the proteins acquire diversifying changes. Black circles denote options that we observe in this study.

Fig. 6.

Comparison of functionalization pathways under dosage imbalance. Each bar represents an experiment initialized with a different starting structure. (A) Fraction of functionalization pathways of the nonduplicated interacting partner. (B) Fraction of functionalization pathways of the duplicated interacting partner that is under selection to maintain binding. Data are not shown for the deleterious duplicate because it defunctionalizes in all replicates.