Molecule of the Month: Artificial Metalloenzymes

Enzymes are the cell's chemists, catalyzing the numerous reactions that are essential for life. At least a third of all known enzymes rely on a metal ion in their active site to help with catalysis. These proteins are known as metalloenzymes. Metal ions, typically iron, zinc, copper, or manganese, enable functions that amino acids sidechains alone could not support, such as shuttling electrons, binding to oxygen, and coordinating substrates and intermediates in the proper orientation for reactions to occur. Hemoglobin, carbonic anhydrase, superoxide dismutase, cytochrome p450, alcohol dehydrogenase, and the respiratory supercomplex are all examples of metalloenzymes previously featured as the Molecule of the Month.

Natural metalloenzymes can perform difficult chemical tasks that, if done within a chemistry lab, would often require high heat and pressure, harsh reagents, and result in the production of toxic waste. In contrast, metalloenzymes and other biocatalysts work in water, at ambient temperature and pressure, and with a high level of selectivity. The range of reactions that natural metalloenzymes carry out, however, is limited; nature only evolved enzymes for reactions that life happened to need, and many chemical reactions that would be useful for industrial purposes appear to have no natural catalyst. Artificial metalloenzyme engineers aim to close this gap, bringing together the broad reactivity of a transition metal with the selectivity, water-compatibility, and potential for "evolvability" of a protein scaffold.

A major way that researchers have been able to engineer molecules with new or improved function is through directed evolution, an approach that involves repeated rounds of mutation and selection. The laboratory of Frances Arnold, who was recognized with the 2018 Nobel Prize in Chemistry for her pioneering work in directed evolution, was interested in finding enzymes that could catalyze carbon-silicon bonds -- a reaction that had not been observed to occur in nature. Arnold hypothesized that heme proteins might be able to catalyze carbon-silicon bond formation, and identified cytochrome c from Rhodothermus marinus as a promising candidate. The main job of this cytochrome c, a small, extremely stable protein, is to shuttle electrons using its iron-containing heme.

After rounds of mutagenesis and selection, these researchers successfully converted R. marinus cytochrome c from an electron carrier into an enzyme that forges carbon–silicon bonds. Remarkably, functional conversion only required changing three amino acids. One of the mutations removes an amino acid that normally caps the heme iron, opening a pocket wherein a reactive iron carbene can form and then insert into the silicon–hydrogen bond of a silane. The evolved enzyme (shown on the right, pdb_00006cun) catalyzes carbon-silicon bond formation in benchmark substrates with near-perfect control of the product's handedness, and works inside living bacteria.

Another strategy developed to engineer metalloenzymes involves anchoring an organometallic catalyst within a protein using "molecular velcro." This strategy, pioneered by the laboratory of Thomas Ward and others, utilizes the extraordinarily tight interaction between the protein streptavidin and the small molecule biotin. If a synthetic metal catalyst is fused to a biotin "anchor," it slots neatly and semi-permanently into the biotin binding pocket. The protein wraps around the catalyst. Residues lining the pocket can be mutated to optimally steer the reaction.

This platform has been used to build metalloenzymes capable of catalyzing reactions that no known natural enzyme performs, including asymmetric hydrogenation, transfer hydrogenation, C–H activation, and olefin metathesis. An artificial metathase, which includes a ruthenium metathesis catalyst held inside streptavidin, is shown on the left (pdb_00005f2b). The metathase was expressed in the periplasm of E. coli cells (where it could avoid naturally occurring inhibitors present in the cytosol) and underwent rounds of in vivo directed evolution. The evolved enzyme compared favorably with commercial (and much dirtier) catalysts and could catalyze reactions with a variety of metathesis substrates.

The newest approach towards engineering artificial metalloenzymes utilizes generative artificial intelligence models to design a protein that can hold the atoms of a substrate at the precise orientation and location to catalyze a reaction. This method was used in 2025 by a team led by David Baker (who received a 50% share of the 2024 Nobel Prize in Chemistry for his work on computational protein design) to create a zinc metallohydrolase, an enzyme that uses a zinc ion to activate a water molecule and cleave a chemical bond (the same reaction that natural carbonic anhydrases and carboxypeptidases catalyze). The researchers specified only key elements of the active site: a zinc ion coordinated by three histidines and a neighboring glutamate that deprotonates a zinc-bound water molecule, activating it for attack on the ester. The AI, in this case, a deep-learning model called RFdiffusion2, then designed a protein built around the active site that was unlike any other protein found in nature. Remarkably, the best designs coming from this approach were active straight out of the computer, without subsequent laboratory-based evolution needed. These designs reached catalytic efficiencies comparable to those of natural enzymes. The illustration on the right shows ZETA_2 (pdb_00009pyl), one of the most active designs.