There's a 2001 paper by Ritz,
et al., that describes a precisely known mutation in a strain of
E. coli that "created" an enzyme that breaks disulfide bonds from a previous enzyme that reduced peroxides. The change in activity was from adding a single amino acid into the peroxiredoxin enzyme - Coragyps's guess is that the change made the active site of the enzyme large enough to accomodate a sulfur-sulfur bond rather than its original oxygen-oxygen.
Abstract:
Pathways for the reduction of protein disulfide bonds are found in all organisms and are required for the reductive recycling of certain enzymes including the essential protein ribonucleotide reductase. An Escherichia coli strain that lacks both thioredoxin reductase and glutathione reductase grows extremely poorly. Here, we show that a mutation occurring at high frequencies in the gene ahpC, encoding a peroxiredoxin, restores normal growth to this strain. This mutation is the result of a reversible expansion of a triplet nucleotide repeat sequence, leading to the addition of one amino acid that converts the AhpC protein from a peroxidase to a disulfide reductase. The ready mutational interconversion between the two activities could provide an evolutionary advantage to E. coli.
Science 5 October 2001:
Vol. 294 no. 5540 pp. 158-160
DOI: 10.1126/science.1063143
It should be publically available if you want to read the whole thing -
Science | AAAS