Ancestral sequence reconstruction - An underused approach to understand the evolution of gene function in plants?

Abstract

Whilst substantial research effort has been placed on understanding the interactions of plant proteins with their molecular partners, relatively few studies in plants - by contrast to work in other organisms - address how these interactions evolve. It is thought that ancestral proteins were more promiscuous than modern proteins and that specificity often evolved following gene duplication and subsequent functional refining. However, ancestral protein resurrection studies have found that some modern proteins have evolved de novo from ancestors lacking those functions. Intriguingly, the new interactions evolved as a consequence of just a few mutations and, as such, acquisition of new functions appears to be neither difficult nor rare, however, only a few of them are incorporated into biological processes before they are lost to subsequent mutations. Here, we detail the approach of ancestral sequence reconstruction (ASR), providing a primer to reconstruct the sequence of an ancestral gene. We will present case studies from a range of different eukaryotes before discussing the few instances where ancestral reconstructions have been used in plants. As ASR is used to dig into the remote evolutionary past, we will also present some alternative genetic approaches to investigate molecular evolution on shorter timescales. We argue that the study of plant secondary metabolism is particularly well suited for ancestral reconstruction studies. Indeed, its ancient evolutionary roots and highly diverse landscape provide an ideal context in which to address the focal issue around the emergence of evolutionary novelties and how this affects the chemical diversification of plant metabolism.

Keywords: APR, ancestral protein resurrection; ASR, ancestral sequence reconstruction; Ancestral sequence reconstruction; CDS, coding sequence; Evolution; GR, glucocorticoid receptor; GWAS, genome wide association study; Genomics; InDel, insertion/deletion; MCMC, Markov Chain Monte Carlo; ML, maximum likelihood; MP, maximum parsimony; MR, mineralcorticoid receptor; MSA, multiple sequence alignment; Metabolism; NJ, neighbor-joining; Phylogenetics; Plants; SFS, site frequency spectrum.

Graphical abstract

Fig. 1

Horizontal Vs vertical approach in the analysis of sequence data. The approach followed by ASR operates a shift from the classical, “horizontal” comparison of sequence data of extant species. Starting from a sequence multialignment and a phylogenetic tree (with branch lengths, here represented by t₁…t₈), the algorithms used by ASR infer the sequences in the ancestral nodes (blue dots). These ancestral sequences can be then aligned to the extant sequences (“vertical” comparison) to identify where and when the historical mutations occurred along the evolutionary trajectories. The ancestral coding sequences can be then expressed in heterologous systems for functional assays. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

The Fitch's algorithm of maximum parsimony (MP) to reconstruct ancestral states. To assign ancestral states the tree is traversed twice. The first time the algorithm proceeds from leaves to root, and assigns to each internal node a set of characters based on the intersection of descendant states (or the union of the intersection is empty). In the second step, the algorithm proceeds from the root to the leaves, and assigns to the internal nodes the state which is present both in the ancestral and in the descendant node. When different equally parsimonious reconstructions are possible, multiple solutions exist (see Suppl. Fig. 1).

Fig. 3

Example of a Maximum Likelihood (ML) algorithm for reconstruction of ancestral states. The figure represents a simple case of ancestral reconstruction using ML. We considered only a single site, with two possible character states (H or Q), across a phylogenetic tree with equal branch lengths. The algorithm first traverses the tree from the leaves to the root, and, for each internal node, computes the likelihood of all possible states taking also into account all possible states of the father node(s). In the second step, the algorithm traverses the tree from the root to the leaves assigning the ancestral states which maximise the likelihood. The figure represents the calculation for the subtree composed by the leaf nodes 4 and 5, the internal node 6 and father node 7 (blue rectangle).