Moonlighting Enzymes

Ten years ago, when I was explaining the molecular basis of enzyme specificity to a group of students, I mentioned that the enzyme, aconitase exhibted two independent functions. It led me to write a short Blog post on what was at the time a relatively new concept in enzymology called “moonlighting”. That is, the unexpected alternative functions or activities of proteins whose function was previously thought to be unique. It’s a little like finding that a 10-year-old copy of the Guinness Book of Records can also be used to keep your office door open. Or as our CEO Rob suggested, using a screw driver to uncork a wine bottle may do the job, but not as well as a cork-screw.

The word moonlighting (in this context) is defined as holding a second job, especially at night (originally). The unexpected concept of moonlighting activities has now become much more mainstream  in the intervening ten years. It also has a significant resonance with those interested in the evolution of enzyme catalysis. 

A well-studied example of a moonlighting enzyme is the cytosolic form of aconitase. This enzyme is “famous” for catalysing the isomerisation of citrate to isocitrate, via aconitate (an enzymatic reaction as old as the discovery of the Krebs Cycle itself and described in detail by WA Johnson in 1939). However, roll on another 50 years a couple of independent labs (see here for example)  established that iron regulatory protein 1 (IRP1, for short) that interacts with mRNA under conditions of low iron concentrations in the cell, is one and the same protein. The reduction in iron levels leads to a collapse of an iron sulphur centre in the enzyme, opening the way for it to become a regulator of a group of mRNAs associated with iron uptake, homeostasis and utilisation.

The discovery that a single amino acid sequence can specify two quite independent biochemical functions challenges much of our thinking about enzymes over the last century. In addition, when a new gene sequence is reported, does it get categorized as (in the above example) as aconitase or IRP1? Prior to 1990, such a sequence would have automatically been annotated as encoding an aconitase function. Moonlighting introduces functional redundancy into genomics and proteomics, and immediately points to the economies inherent in molecular evolution. This brings to mind the discussions amongst the RNA Tie Club members (pioneers of the Molecular Biology Revolution formed in the aftermath of the publication of the Double Helix paper in Nature), trying to address the nature of the Genetic Code and its likely intrinsic, informational redundancy. (You can read about this in the excellent book by Horace Freeland Judson.)

The enzyme dihydrolipoamide dehydrogenase services the catalytic activity of several multienzyme complexes in core metabolism. Which is interesting in itself: one gene encodes one protein that serves several masters! There are some mutations in this gene that turn the metabolic enzyme into a protein degrading enzyme, or protease. A totally different type of biochemical function. You can read about the latest examples at a superb resource site managed by Constance Jeffrey’s lab at Illinois, Chicago.

I find this phenomenon fascinating, since it suggests to me that the discovery of the function of a protein is determined by technology available at the time or perhaps the targeted search for a particular activity: in other words the classification  of protein function confirms hypotheses, but doesn’t throw up new ones: a little like site directed mutagenesis. So an enzyme gets branded with a function, making it difficult to subsequently to think of it as having a different role in the cell. It also begs the question which function was selected first during evolution! This makes genome annotation even more challenging, on the one hand, but more complex and intriguing on the other.

The two faces of the same enzyme: aconitase becomes IRP1. PDB file 2B3Y kindly made available at the RCSB website.

Let us now consider whether the concept of “moonlighting” has any implications for the outcome of a directed evolution experiments? I shall begin with the discussion of an early success of knowledge-based protein engineering.

In a typical scenario, the primary aim of a directed evolution experiment might be to alter the substrate specificity of an enzyme, such as that achieved by changing for example a lactate dehydrogenase to a malate dehydrogenase. In a landmark publication in 1988 the protein engineering group at the University of Bristol, used site directed mutagenesis to generate a highly efficient malate dehydrogenase (MDH). In this work the authors used the Bacillus stearothermophilus lactate dehydrogenase (LDH) gene as a ground state (recently re-named as Geobacillus), which had around 50% sequence similarity to the organism’s wild type MDH. With a knowledge of the structure they were able to predict that 3 amino acid changes at the active site were sufficient to switch specificity. The new engineered LDH variant exhibited a two-fold increase in catalytic efficiency over the native MDH, with a near 10 000-fold discrimination between the two substrates. Very impressive, and probably as important a step in enzyme engineering as the switch in repressor specificity achieved by the Ptashne laboratory four years earlier.

These early successes in the application of structural knowledge for the rational engineering of protein function have not however proved easy to translate to other systems; which goes some way to explain the motivation behind directed evolution. However, it does illustrate for me that directed evolution is an important step on the way to developing the rules for more robust predictive methods and algorithms: a journey we are undertaking at Entropix.

Back to moonlighting. The example of LDH was chosen to illustrate the way in which successful rational design strategies can successfully re-purpose an enzyme with respect to substrate specificity. However, it is not clear how far such improvements in catalytic efficiency can be taken by strategic substitution for enzymes acting in solution. In vivo, if mutation delivers a competitive edge to an organism, sufficient for that variant to repopulate a pool of species, it will generally prevail. If the new variant brings added versatility, then its competitive impact may be even more profound. (This is at the heart of Darwin’s reference to adaptation as a key driver in evolution by natural selection.) A gene that has acquired multiple functionality through evolutionary change, represents a much greater challenge for rational design. This is where the concept of moonlighting and directed evolution converge.

Starting from an LDH ground-state, if a client requires an MDH that binds an unrelated ligand, perhaps one that on binding, inhibits catalysis i.e. an allosteric inhibitor, or toggle switch, then rational design becomes much more of a challenge. The rational design of a moonlighting enzyme, comparable to that discussed above, at the time of writing, is much more likely to emerge from a programme of directed evolution than by rational design. On the other hand, by systematically documenting the pathways to success (and importantly failure) in directed evolution, predictive tools will continually improve. The close interplay of experimental directed evolution, bioinformatics and computational chemistry is steadily catching up with Nature. At Entropix we are building our capability around this exciting vision.