Abstract
Nearly 60 years ago, Eck and Dayhoff (Science 152:363-366, 1966) aligned amino acids from first and second halves of a ferredoxin sequence, revealing a symmetric CX(2)CX(2)CX(3)CX(18)CX(2)C(2)C(3)C spacing signature in which X(n) denoted intervening residues. This symmetry, along with other cyclic patterns, suggested that a tandem duplication shaped ferredoxin evolution and that the ancestral sequence may have drawn from a reduced amino acid repertoire. Here, I revisit the duplication model using the deep learning-based AlphaFold2 ab initio pipeline, benchmarked against the I-TASSER threading tool. Predicted ancestral structures were obtained with high confidence, with some aligning to the two halves of a reference ferredoxin (PDB entry 1CIF) at acceptable RMSD and TM-score values. A chronology of loops and structural domains further identified which duplicate was ancestral, reinforcing the antiquity of the fold. Loops and domains also dissected the evolution of the [4Fe-4S] ferredoxin superfamily. The resulting structural models provided strong support for the tandem duplication hypothesis and the idea that modular units underpinned early molecular evolution. However, they also challenged the notion that the duplication event arose from a reduced amino acid alphabet. This work revisits Eck and Dayhoff's seminal insights and commemorates Dayhoff's pioneering contributions on the centenary of her birth.