Sometimes the most exciting moments in research happen when an experiment does not quite behave as expected. For me, the key observation came from what was supposed to be an enzyme inhibition experiment.
At the start of my PhD, I was working with enzymes engineered to catalyse the Morita-Baylis-Hillman reaction, using chromogenic substrates to probe their reaction mechanisms. We already had an absorbance assay based on suicide inhibition. In this system, the enzyme reacted with a substrate, eliminating a small leaving group and forming a covalently bound intermediate with a characteristic spectroscopic signal.
However, the assay had several limitations. The covalent enzyme-substrate species was a poor mimic of the true catalytic intermediate, which limited the value of the resulting observations. The species also resisted attempts at structural characterisation. In addition, the spectroscopic signal was proportional to the enzyme concentration rather than the amount of substrate turned over, and was therefore weak.
The plateau that was not quite flat
To improve the assay, I changed the substrate design. Instead of placing the chromophore on the reaction intermediate, I moved it to the leaving group. At first, the new compound behaved like the original suicide inhibitor: the signal increased quickly and then appeared to plateau. As expected, the enzyme reacted rapidly before becoming trapped in a covalent complex with the substrate.
Except the plateau was not completely flat. If I left the reaction running longer, the signal continued to creep upward – a small but reproducible increase. The supposedly inhibited enzyme was slowly hydrolysing the covalent intermediate, releasing the substrate and freeing itself to react again. This observation gave me an idea: if water could react with the intermediate, perhaps other molecules could too.
To test this idea, I added a small carbon nucleophile to the reaction. The result was immediate and dramatic. Instead of creeping upward, the absorbance signal shot up - the reaction was orders of magnitude faster. The signal continued to rise before plateauing at a much higher level, consistent with consumption of all the available substrate.
What was meant to be an inhibition assay had revealed a new catalytic activity. Rather than simply being trapped, the enzyme formed an activated species that could react with nucleophiles to generate a new bond. I had stumbled upon a previously undiscovered allylic transferase activity.
Inhibition assay absorbance traces, with and without added nucleophile.
From initial activity to biocatalytic platform
At this point, the project took off. To turn this reactivity into a useful biocatalytic platform, we used directed evolution: an iterative process of generating and testing enzyme mutants to optimise the target activity. After two years and eight rounds of evolution, we had developed a family of allylic transferases capable of synthesising a wide range of products with high conversion and enantioselectivity. The enzymes accepted at least 15 different nucleophiles, while the electrophile scope expanded to include a challenging, poorly electrophilic acrylate and a chiral electrophile.
One valuable feature of the evolved enzymes was their regiochemical control. If a substrate contains more than one reactive site, multiple regioisomeric products may be formed. Allylic transfer reactions are particularly prone to this issue. In small-molecule-catalysed systems, the intermediate can react again with another catalyst molecule, shuffling between states that lead to different regioisomeric products. The resulting mixtures can be difficult to separate and typically require protection and deprotection steps or more elaborate catalysts to avoid.
Our enzymes provided an elegant solution. Because the catalytic nucleophile and reactive intermediate are confined within a protein cavity, the intermediate cannot undergo the secondary catalyst additions that cause isomerisation. The reaction therefore produces a single regioisomer. This selectivity illustrates an appealing feature of biocatalysis: the enzyme not only accelerates the reaction, but also controls the environment around the intermediate in a way that is difficult to reproduce with small-molecule catalysts. What began as an anomalous assay trace had become a platform for highly regiocontrolled allylic transfer reactions. Little did I know there were more surprises in store.
Divergent reaction pathways for allylic transfer reactions produce different regioisomeric products. Confinement of the catalytic nucleophile and intermediate within a protein cavity leads to formation of a single product regioisomer (orange box).
Looking beyond the endpoint
Enantioselectivity, the preferential formation of one mirror-image product over another, is another sought-after form of selectivity. When the substrate nucleophile contained a chiral centre, our evolved enzymes often performed well, forming products with enantiomeric ratios of up to 98:2. When the chiral centre was instead on the electrophile, however, the results initially looked much worse, with ratios barely above 50:50. This was disappointing but not surprising: the enzymes had been evolved to control selectivity at chiral centres on the nucleophile, not the electrophile.
Then another odd observation appeared. When I varied the enzyme loading and reaction time, the apparent enantioselectivity changed. Reactions that looked poorly selective at one time point appeared much more selective at another. At the same time, the enantiomeric ratio of the racemic electrophile changed during the reaction. One substrate enantiomer was being consumed before the other.
To understand what was happening, we separated the electrophile enantiomers and fed them individually to the enzyme. The results were striking: both reactions formed product with a high enantiomeric ratio, but the preferred product enantiomer was reversed depending on which substrate enantiomer was used. The enzyme was transferring chiral information from substrate to product, even though the reaction passed through a planar, non-chiral intermediate.
This emergent stereospecificity had not been designed or screened for, but had arisen during evolution and became apparent only through investigating unexpected reaction behaviour. For me, this was one of the project’s most interesting lessons. Directed evolution can do more than improve activity towards a defined target: it can also confer unexpected behaviours on an enzyme. And there was one more surprise left to discover.
Reactions using a single substrate enantiomer generate products with the same stereocentre configuration. Stereospecificity is achieved via E/Z isomerism of the planar intermediate.
The leaving group that stayed behind
One powerful way to study an enzyme mechanism is to capture a crystallographic structure of the proposed reaction intermediate. Because our intermediate is covalently attached to the enzyme, we soaked protein crystals in a substrate solution before freezing, aiming to stop the reaction before hydrolysis. Reassuringly, the resulting structure showed the intermediate covalently bound to the catalytic nucleophile, supporting our proposed mechanism.
However, the structure also contained something unexpected. Beside the intermediate was an additional region of electron density that we initially could not explain. Eventually, we modelled in a molecule of the leaving group, para-nitrophenol (PNP), which fitted almost perfectly. The leaving group had not left: instead of immediately diffusing away, it remained bound in the active site. This surprising discovery raised an interesting question: did the bound PNP play a functional role in catalysis?
To find out, we replaced the PNP leaving group with a smaller acetate. The modified substrate reacted at a similar rate but with substantially lower stereoselectivity. Further analysis suggested that the bound PNP helps position the incoming nucleophile for stereoselective bond formation. The leaving group was not simply a side product, but had become part of the catalytic architecture of the evolved enzyme.
Crystal structure of the enzyme active site, showing electron density maps for the covalent intermediate and the bound PNP leaving group.
Lessons from the unexpected
Looking back, this project was shaped by a series of observations that could easily have been dismissed: a plateau that was not quite flat, reaction stereochemistry that changed with time, and a patch of unexplained electron density. Each clue led to a discovery more interesting than the experiment had been designed to find, ultimately shaping and enhancing the project.
For me, the work reinforced one of the most enjoyable aspects of enzyme engineering. Directed evolution is typically used to optimise enzymes for predefined goals, an application in which it is certainly powerful. But it can also introduce productive and unexpected catalytic behaviours that force us to investigate further, enriching our understanding of both the system and enzyme catalysis more broadly.
Sometimes the most important experiment is the one that refuses to give a perfectly flat line.