Main

To test the predictions for one daughter lineage, the authors first investigated patterns of DNA sequence evolution3. It was shown that after the duplication of the progenitor DFR gene, the lineage leading to the common ancestors of DFR-A and DFR-C exhibited a non-synonymous-to-synonymous rate ratio significantly exceeding 1, indicative of adaptive evolution by positive selection. However, subsequent biochemical assays showed that these enzymes have “essentially no activity” on any ancestral substrate3 and therefore do not demonstrate improvement of ancestral function, violating the above criterion (ii).

The other daughter gene from the duplication, DFR-B, showed no evidence of positive selection from sequence analysis3, calling into question whether adaptive evolution has occurred. To test potential improvement of ancestral enzyme function after duplication, the authors made recombinant proteins from a heterologous system and measured catalytic activities on five substrates. Although the same volume of cell culture extract was used in all assays, the recombinant proteins were neither quantified nor equalized in concentration. It is known that the protein yield from heterologous systems can vary by several orders of magnitude, depending on a wide variety of factors7. In fact, an earlier study in a heterologous system nearly identical to the one used by Des Marais and Rausher reported very different yields for different DFR enzymes8. Thus, it is crucial to quantify the concentration of the obtained recombinant protein for any quantitative measure of activity, as has been done in previous studies of duplicate gene evolution9,10. Without enzyme concentration information, the apparently higher activity of DFR-B depicted in their Fig. 2 could be due to a larger amount of enzyme rather than a truly higher enzyme activity as the authors interpreted3.

One may argue that the ratios of activities against different substrates can be compared among enzymes to control for differences in protein concentration. For example, one could infer from their Fig. 2 (ref. 3) that the relative activity on dihydrokaempferol (DHK) to dihydroquercetin (DHQ) is enhanced in DFR-B, compared to that of the progenitor enzyme. However, such relative activity information does not indicate whether the enhancement is due to increased activity on DHK, reduced activity on DHQ, or both. It could also be posited that the activity of DFR-B on DHQ is unlikely to be reduced compared to the progenitor enzyme if DHQ is the main precursor of most anthocyanins produced by post-duplication species. However, because Michaelis–Menten kinetic parameters were uncharacterized for these enzymes, even if the activity of DFR-B is lower than that of the progenitor, it remains possible that it is high enough for normal physiology, particularly if DHQ concentrations are high. Furthermore, one cannot exclude the possibility that the activity on DHQ is retained in DFR-A and/or DFR-C but reduced in DFR-B. Even if different enzymes are compared using these ratios, it is clear that DFR-B is no better with naringenin and eriodyctiol (the less typical but proposed to be improved-upon substrates) compared to DHK, DHQ and dihydromyricetin (DHM) (the typical substrates of DFR) than the progenitor enzyme from Evolvulus (Fig. 1). Thus, there does not seem to be any unambiguous evidence for improvement of any ancestral function in DFR-B. Coupled with the failure to detect positive selection in DFR-B, these results go against both criteria (i) and (ii). Because there is no evidence of improvement of ancestral function in any of DFR-A, -B and -C, criterion (iii) obviously cannot be established.

Figure 1: Results of enzyme activities against different substrates.
figure 1

Shown are enzyme activities against naringenin (grey bars) and eriodyctiol (white bars), relative to those against DHK, DHQ and DHM. A greater number indicates a relative preference for narigenin (or eriodyctiol) compared to DHQ, DHK and DHM. Error bars denote one standard error. For a given enzyme, we calculate its activities against naringenin (or eriodyctiol) relative to DHK, DHQ and DHM, and show the average of the three relative activities and its standard error. The evolutionary relationships among the four enzymes are shown by the phylogeny. The relative activities are calculated from the data presented in Fig. 2 of ref. 3, graciously supplied by the authors.

PowerPoint slide

Full size image

In conclusion, none of the three key features of the EAC model have been clearly demonstrated in the evolution of DFRs. Although EAC may be a valid model for describing gene family evolution, rigorous tests of the predictions associated with it are needed to describe its general importance.