Design and Preparation of Artificial Heme-Copper Enzymes

Abstract

Natural and artificial metalloproteins play a critical role in biochemistry, with the first X-ray crystal structures ever solved belonging to heme proteins. Understanding the design principles for metal binding and its modulation in proteins is of great interest. Control of active site geometry is often the key requirement for catalysis and its mutagenesis helps probe a wide variety of biological and abiological reactions. In the case of small-molecule activation, introduction of new metal-binding sites to non-native heme scaffolds can unlock new chemistry. Due to their ability to carry out a diverse array of challenging reactions at ambient temperature, effective metalloenzyme design and isolation strategies are highly desirable. In this chapter, we will provide methods used in our lab for the design and experimental preparation of artificial metalloenzymes containing a heme-copper center to mimic and understand heme-copper oxidases. The methods can be applied to design other heterobinuclear centers containing heme, such as the heme-nonheme iron center in nitric oxidase reductases.

1. Brief Overview of Artificial Metalloprotein Design and Preparation

Making up almost half of the proteome, metalloproteins are essential to life, performing a wide range of biochemical functions that require metal ions and complex metallocofactors (Yu et al., 2022). These enzymes enable key processes such as respiration, nitrogen fixation, photosynthesis, and detoxification, often by facilitating electron transfer or activating small molecules. Because of their catalytic diversity, many biochemical and biophysical methods have been developed to understand the structure, function, and reaction mechanisms of these metalloproteins. There has been historic interest in the design of artificial metalloproteins (ArMs) from a bottom-up approach, replicating a target active site in a non-active scaffold for gain-of-activity characterization (Koebke et al., 2022; Zou et al., 2024; Chen et al., 2025). This complements the biochemical and biophysical studies of native enzymes and their variants. While a mutation in the native enzyme that abolishes the protein function can have multiple reasons, such as affecting overall protein stability, a mutation in the ArMs to gain function is a clearer confirmation of the role of the residue. Besides providing a complementary approach to understanding native enzymes, ArMs provide an alternative platform for engineering to improve substrate scope, enhancing performance under industrial conditions, and increasing enzyme tunability, which are pursued to improve their broader applications (Bell et al., 2024; Burgener et al., 2024; Yu and Ward, 2024). As a result, ArMs have emerged as a powerful class of catalysts that combine the functional diversity of metal cofactors with the programmability and selectivity of protein scaffolds (Kerns et al., 2022; Vornholt et al., 2024).

Early ArM designs, pioneered by Whitesides and Kaiser in the 1970s, focused on covalent or supramolecular incorporation of synthetic metal complexes into natural protein hosts (Yamamura and Kaiser, 1976; Wilson and Whitesides, 1978). More recent advances have leveraged tools like computational protein design, directed evolution, and noncanonical amino acid incorporation to expand ArM capabilities (Lu et al., 2009; Mirts et al., 2019; Morita and Ward, 2024; Vornholt et al., 2024). For instance, rational design approaches have taken advantage of high-resolution structural data to introduce or optimize metal-binding motifs into known protein scaffolds, enabling catalytic activity while preserving structural integrity (Zhou et al., 2024; Dai et al., 2025). Advances in structure prediction algorithms and energy function optimization have allowed researchers to pre-design scaffolds that bind metal ions with high specificity or support desired reactivities, such as carbene or nitrene transfer (Tinberg et al., 2013; Yang and Arnold, 2021). Computational models often require extensive experimental validation, especially when predicting second-sphere interactions critical for catalysis. Meanwhile, de novo protein design has opened new frontiers by enabling the creation of entirely new protein scaffolds from scratch, rather than modifying existing ones. Among state-of-the-art efforts, the Baker lab has led in de novo protein design, pushing the frontier of custom scaffolds tailored for specific catalytic tasks (Mills et al., 2013; Huang et al., 2016; Ahern et al., 2025). However, in contrast to the remarkable success in designing proteins that bind another protein or a small organic molecule, designing catalysts – especially metal-binding enzymes - are significantly more challenging due to the strict active site geometric control requirements. To meet this challenge, a wide array of ArM design strategies has evolved, each with distinct advantages and trade-offs.

2. General metalloprotein design and engineering considerations

To design ArMs, bioinformatics and homology modelling are common strategies. Identification of conserved sequences through sequence homology or other bioinformatic tools provides a list of the residues critical to the protein, which could otherwise be challenging to identify (Akcapinar and Sezerman, 2017). Additionally, it is also significant that the pool of homologs used to obtain mutational information is properly annotated and curated; this especially becomes important in enzymatic systems that are less well known. Finally, there are cases in which critical residues are co-evolved and require additional consideration to be properly identified (Kim et al., 2022).

Identification of critical residues using structural information, such as protein crystal structure, allows for a clear-cut identification of the probable binding and catalytic residues, especially if the literature on the protein’s mutagenesis is also considered. However, crystallography is not perfect. For instance, metal binding site parameters are prone to cryo-reduction (Yano et al., 2005). Over-reliance on one structure over the other homologs might result in misinterpretation of the residue roles or the strictness of the binding requirements within the enzyme. Finally, protein crystal structure usually differs from the solution structure due to the packing lattice. Dynamic backbone movements of proteins are not captured via a static low-energy snapshot; hence, crystal structures should not be the end-all point of reference for protein engineering (Fraser et al., 2011).

The selection of interactions to design should be determined by the experimental goal. In most cases, replication of the primary coordination sphere is a minimal requirement. The selection of the secondary coordination sphere is more challenging for a couple of reasons. For any given protein, there often does not exist a complete characterization of the role of SCS residues and their mode of action. Thus, the introduction of a homologous residue at a different placement might have a similar effect. Furthermore, the identity of the residue could drastically change while preserving functionality, i.e., including a tyrosine residue instead of a serine to serve a hydrogen bonding role. Finally, an SCS interaction could still fail even if the placement strongly mimics the native enzyme. Since the SCS interactions are non-covalent, they are much more sensitive to their local environment, and unintended interactions with other neighboring residues are not trivial to predict.

For ArM design, the selection of the candidate scaffold is a compromise between protein active site similarity, mutability, and experimental feasibility. For example, while myoglobins and hemoglobins have similar active sites, hemoglobin is present as a tetramer in the solution, making experimental assessment of the mutagenesis effects significantly more difficult than myoglobins. In addition, the candidate scaffolds must be expressible in the selected host. For E. coli expression, this rather trivial fact needs to be reconciled with two realities. If the engineering process is done via screening of a large protein database, it is not uncommon that such in-expressible proteins pass through the original filtering process. Second, if the protein has not been expressed in E. coli yet, it doesn’t mean it cannot be expressed in it and a plethora of options can be pursued to achieve this (Pouresmaeil and Azizi-Dargahlou, 2023).

If candidate scaffolds and/or residue choices are limited, or if the geometrical parameterization of the interactions is not feasible, rational design will be the most practical engineering approach. Active sites can be overlayed and critical residues identified for mutagenesis. However, to assess a database of candidate scaffolds especially when the interactions are already known, a computational design pipeline is more feasible. If the scope of the project permits, the stability of the protein should also be probed via MD simulations.

3. Heterologous protein expression with the pET vector system

Metalloprotein preparation throughout the last decades has advanced exponentially. One of the earliest and most historically significant milestones in protein experimentation was the isolation of myoglobin by John Kendrew and his colleagues in the 1950s (Kendrew et al., 1954, 1958). Myoglobin, was among the first metalloproteins to be purified in sufficient quantity and homogeneity for structural analysis. The purification strategy that yielded the crystal structure relied on extraction of the protein from tissue samples through steps such as blending and multiple salting out precipitations (Michael Rothgeb and Gurd, 1978; Wittenberg and Wittenberg, 1981) (Fig. 1). This was a groundbreaking moment that established the importance of metal cofactors in protein structure and function. Despite all the efforts, there were still challenges in ensuring that protein fold, the redox state of the cofactor, and protein activity remained intact. The development of heterologous expression systems has alleviated many of these constraints. By expressing metalloenzyme genes in tractable organisms such as E. coli, researchers can now produce large quantities of recombinant protein under defined conditions (Fig. 1). This allows for co-expression with metallochaperones, supplementation with specific metals, and incorporation of isotopically labeled cofactors (Waldron and Robinson, 2009; Rosano and Ceccarelli, 2014; Cai et al., 2019). Heterologous expression enables the study of previously inaccessible or low abundance metalloenzymes, greatly expanding the toolkit for understanding and engineering these vital biological catalysts.

Fig. 1:

Schematic for common construct designs.

Initially created in the 1980s by Sturdier and coworkers at Brookhaven National Laboratory, pET is the most popular vector system for heterologous E. coli protein production (Fig. 2) (Rosenberg et al., 1987) Expression of the gene of interest is controlled by the lambda phage T7 promoter and terminator, originally governing capsid protein production and specific to T7 polymerase. Lab E. coli strains are lysogenized to host lambda T7 RNA polymerase genes, which are regulated by lacUV5 promoter and repressed by lacI (Dubendorf and Studier, 1991). In the presence of IPTG, repression by lacI is turned off, enabling polymerase production and subsequent transcription. Mimicking the viral mechanism, the T7 polymerase system drives cells to divert most of their metabolic resources towards protein production.

Fig. 2:

Crystal structural of the engineered myoglobin variant containing a copper B site, CuB-Mb (PDB ID 4FWY) on the left and the overlay with the best overlap shown in three different angles. His64 already matches one of the histidines in the oxidase. Phe43 and Leu29 have adequate positioning to be mutated into the ligating histidine.

Widely adopted by commercial platforms, there are now more than 100 pET-based vectors in the market. Primary factors to consider during construct design include promoter strength, cloning site selection, antibiotic choice, plasmid copy number, and host/vector combination (Francis and Page, 2010; Samuelson, 2011; Rosano and Ceccarelli, 2014). In many cases, fine-tuning cells’ metabolic burden for maximal protein expression is a protein-specific optimization problem, with multiple factors like solubility, mRNA/protein stability, growth, and induction conditions necessitating a trial-and-error approach. For constructs producing toxic protein, glucose supplementation and the use of the T7 lysozyme-encoding PlySs strain can improve plasmid stability and decrease basal transcription, creating higher yields (Studier, 1991). Maintaining low plasmid copy number can also aid in toxic protein production (Rong Fu Wang and Kushner, 1991). During construct design, it is strongly recommended to codon-optimize the target protein based on the expression host’s codon usage (Gustafsson et al., 2004). When this is not feasible, rare codon compatible E. coli strains such as Codon Plus or Rosetta DE3 can be used. The inclusion of an affinity tag can significantly streamline purification. The most common tags are the hexa-histidine, strep, and maltose binding protein tags, with the latter also aiding the expression of difficult-to-fold proteins (Kimple et al., 2013). Introduction of the metal-binding histidine tag requires an additional removal step, as it can interfere with either binding the protein metallocofactor or creating side reactions in relevant downstream assays. Common removal strategies include the usage or sequence specific proteases such as the Tobacco Etch Virus cysteine protease (TEV-P), factor XA, or SUMO protease for the SUMO tag, to name a few (Butt et al., 2005; Waugh, 2011; Kimple et al., 2013). New non-enzymatic cleavage strategies are also constantly evolving (Dang et al., 2019). When expressing multiple proteins simultaneously, the use of pET Duet constructs (Fig. 3), multiple plasmids of different replicon-selection marker combinations, or multiple T7 promoter/terminator pairs can be helpful.

Fig. 3:

Constraint file used for the Rosetta Matcher run. For the first histidine ligation, as the constraint has to be classical and hence include the virtual atoms of the copper parameter file, the angle A and torsion angles A and AB values are arbitrary.

With many metalloenzyme systems, we have noticed that tuning down the expression rate via lowering the temperature to sub-37 degrees during growth and induction is beneficial. It is well known that overexpression of the target protein can induce severe metabolic strain to its heterologous host (Snoeck et al., 2024). Additionally, it is not uncommon for a heterologous expression machinery that the transcription rate of the protein outpaces the cell’s protein folding mechanism and results in the formation of in-vivo protein aggregates called inclusion bodies (Fahnert et al., 2004). It has been shown that inclusion body formation can be minimized by carefully optimizing expression media formulation or temperature control (Studier, 2005; Chou, 2007). Alternatively, strategies are also employed to resolubilize the target protein from the inclusion bodies (Fahnert et al., 2004).

While the use of pET has vastly simplified many purifications, challenges remain, particularly for oxygen-sensitive metalloenzyme production. Many metalloenzymes have internal cavities to support cofactor binding, which can make them unstable when expressed rapidly in a cofactor-free environment. Lack of chaperones and native assembly machinery can cause the target protein to be insoluble or get incorporated with incorrect metal, necessitating more downstream processing. Metal supplementation during growth can be helpful, but large quantity of metal can cause toxicity. Use of accessory operons (Jansing et al., 2023), specialized host (Fan et al., 2020), and cell-free protein synthesis (Fan et al., 2020; Jansing et al., 2023; Koo, 2023) are several alternative strategies for fully active protein production, each with their pros and cons.

For the later sections, we will focus on several standard E. coli expressions relying on in-vitro cofactor incorporation, which is the most common strategy since many engineered proteins express in apo form.

4. Computational Design and Validation of ArMs containing heme-copper center to mimic HCO

4.1. Rational Design

The design of the metalloenzymes has significantly evolved through the last decade, especially with the advent of generative AI design methodologies. Here we discuss the original approaches that resulted in our published metalloenzymes (Sigman et al., 1999, 2000; Miner et al., 2012).

Procedure

1. Heme copper oxidase RCSB crystal structure with PDB ID of 1V54 and nitric oxide reductase PDB ID of 3O0R were selected as modelling candidates.

2. Myoglobin RCSB crystal structure of 1MBN and Cytochrome C Peroxidase Model of 2CYP is selected for modelling.

3. Engineering candidate features for HCO is selected: it consists of the ligating three histidine residues for the PCS and additional interactions such as the His-Tyr crosslink

4. Proteins are loaded through the ChimeraX (Pettersen et al., 2021) and their heme rings are superimposed. As seen in Figure 2, while overlaying can be done with the alignment of different heme-ring nitrogen atoms, there is usually a singular, intuitive ring overlay that minimizes the number of required mutations.

5. Comparable residue positions are selected in the CcP and Mb scaffolds and mutagenesis is done in silico and rotamers are selected to replicate the ligation environment for the respective copper and Fe binding sites and used for downstream computational assessment.

4.2. Computational Design using Rosetta Software Suite

While this method has not been originally used for the design of the above proteins, it has become increasingly relevant for screening a large library of candidates. Hence, it will be showcased here.

1. Download the PDB ID’s for monomeric heme proteins in the RCSB database that are expressed in E. coli. These would serve as the candidate scaffolds.

2. Construct the Rosetta Matcher constraint files using the measured binding geometries for the HCO and NOR systems using either the model structures in 4.1.1

3. Prepare the input proteins for the Rosetta Matcher run:

   1. Renumber PDB sequences so that the first resolved residue is always numbered as residue 1
   2. Identify the distal pocket residue lists of the heme proteins. A potential pseudo code is as follows:

      1. Identify the heme iron atom
      2. Identify the proximal ligation residue
      3. To construct the pos file, identify the distal pocket as residues within the vicinity (less than 6 angstrom) of the heme iron atom and not on the same side as the proximal residue, this can be done for example by calculating the dot product sign of the heme iron-ligating atom vector and heme iron and the Ca of the potential distal residue.
         Note: this will not work for cytochromes. In this context, they are also not reliable candidates since many of them contain heme c cofactors. These usually needs a maturation enzyme for incorporation and removal of their axial residues, which usually results in drastically less stability.
   3. Clean the PDB files by removing all the non-protein molecules but keep the heme cofactor.
4. Benchmark the specified constraint file against the native heme-copper oxidase, input preparation should be done the same way as 4.2.3.

   1. If the constraint file does not produce any matches, the binding geometry ranges need to be expanded, or the actual geometries should be readjusted. Readjustment only becomes necessary if the ligand residue in the parent protein is in a non-common rotamer; otherwise, expanding the scan range values almost always results in a positive match.
   2. Constraint geometries should have high tolerance values up to 0.3 Angstroms for bond distances and up to 25 degrees for the angles. Higher constraint tolerance or looser constraints usually result in many false positives and increase the effort needed to filter the results. Lower constraint tolerance reduces the Matcher run time but also results in more false negatives since Matcher does not perform structural optimization. As a rule of thumb, no more than ten million rotamers should be examined per constraint.

5. Run Rosetta Matcher on the candidate scaffolds with the benchmarked constraints

6. Run a Rosetta structure relaxation script on the Rosetta Matcher output. The results of this simulation are not necessarily chemically accurate but provide much better input structures for further energy minimization compared to Rosetta Matcher outputs, as seen in Fig. 5. MD simulations and can also be used to evaluate additional computational mutations. The script is an adjustment of the script used in (Mills et al., 2013; Heinisch et al., 2015)

   1. Generate a rotamer library of the heme ring if needed or use the native pdb heme ring as the parameter file.
   2. Parameterize the heme-ligation interactions in a separate constraint file.
   3. Run the Rosetta script code to stabilize the matched structure.

Fig. 5:

Overlay of CuB-Mb (PDB ID 4FWY, pale green) with the Rosetta Matcher output on WT myoglobin in cyan and the relaxed structure via Rosetta Scripts in magenta. Structural relaxation usually optimizes the interaction geometry and balances it with small backbone shifts.

Notes:

1. You can alternatively download the PDBs of all XRD structures of the enzymes from the RCSB database and measure the desired interaction through a python scripting pipeline. This will provide a better geometrical representation of the active site but could also introduce artifacts from poorly resolved crystal structures and given the lack of chemical accuracy of Rosetta simulations, the potential accuracy gain in this approach is negligible especially if the chosen modelling enzyme has been selected carefully as a faithful representation.

2. For histidine ligations, using the N_His atom type is recommended if there is no preference between the epsilon and delta ligating atoms. In this case Rosetta treats the constraint block as a variable constraint with the atoms one, two, and three of the histidine residue either as NE2,CE1,ND1 or as ND1,CE1,NE2.

3. The default copper cation parameter file in Rosetta orders it in the center of a tetrahedron with four virtual atoms (V1–4) on each corner. Hence the bond and dihedral angles that involve these virtual atoms are arbitrary as seen in Fig. 3. As the first constraint block needs to have all six degrees of freedom specified, an arbitrary number without any scan range should be used for those specific constraints. In the constraint blocks of the second and third histidine ligation, these three degrees of freedom should be removed and a secondary constraint interactions, specified by the “SECONDARY_MATCH: DOWNSTREAM,” should be used.

4. Depending on your Rosetta version, you might need to specifically include the “-match::enumerate_ligand_rotamers false flag” for the run even though the copper cation does not have any rotamers.

5. Running Rosetta Matcher for the Manganese CcP needs additional considerations as the matcher does not allow the matching of two non-protein residues. The more straightforward solution is to make a singular parameter file of the heme and manganese cation together and do the matching to this “ligand”.

6. The Rosetta Scripts relaxation can also be done via running the optimization steps in a Montecarlo mover or with enabling additional mutagenesis. In our case, it has been more informative to repeatedly run the output of the optimization as the input of the next relaxation step using the same script in order to obtain a trajectory of structure optimization.

7. It is strongly recommended that a rotamer library of the heme parameter is generated for the Rosetta optimization step. However, this can be avoided if the optimization step is used as purely a pre-relaxation for MD or QM/MM runs.

8. Depending on the residue database of your Rosetta installation, it might be likely that the heme residue name of HEM is already used for another small molecule, in this case the heme name can be simply renamed in the parameter file, and the name should also be used in the pdb input file.

5. Metalloenzyme molecular dynamics

While classical dynamics is routinely used in evaluating protein stability, simulations for metallo-protein remain a bit more challenging to setup due to requirements for force field construction. Several metal-ligand bond treatments exist, the most popular of which include non-bonded model (Stote and Karplus, 1995), cationic dummy atom model (Pang et al., 2000), or bonded model (Hoops et al., 1991). For a non-compromising treatment, building customized force field parameters using software such as AMBER MCPB is recommended (Li and Merz, 2016). Regardless of the methods used, the nature of the metal-ligand interaction is difficult to handle qualitatively at the classical molecular dynamic level, even for inert metal like zinc (Hitzenberger and Hofer, 2016). Assessing stability of a metal binding site in-silico is not simple since bonds need to be predefined during setup but cannot be broken during simulation.

Due to the inherent complexity of parameterization, here we will describe a molecular dynamic protocol with approximations for the metal site parameters. Bond length fixing and harmonic constraints are common strategies for metalloenzyme treatment (Sinnecker and Neese, 2006; Waller et al., 2007; Sproviero et al., 2008), with protocols having MD relaxation followed by a QM/MM optimization. The rationale is that the initial equilibration would remove crystal packing strains from protein backbone while the fixed active site ensures the metal center would not collapse or deviate significantly from crystal geometry. This active site would receive more accurate treatment during a follow-up QM/MM procedure and benefit from having its starting point resemble experimental evidence.

One approach to estimate a metal binding site favorability is from an active site preorganization perspective. A simple exercise is to observe the rotamer profile of the ligation residues without metal sampled during molecular dynamics. Our working assumption is that an active site that supports metal binding would not involve a significant backbone shift to accommodate the ligating rotamer. Here, we will evaluate the copper ligating site of CuB Mb using this approach.

5.1. Methods:

5.1.1. Installation

1. Conda environment for ASH and OpenMM (https://github.com/RagnarB83/ASH)

2. A molecular graphics software. Here, we will use PyMOL

The provided python script is adapted from ASH’s Molecular dynamics tutorial and offers a simple template for simulation setup. A github repository containing the codes for this demonstration is https://github.com/Lisa-Phan/HCO_Rosetta_MD_MIE_2025

5.1.2. Setup

For system preparation, the first step is to select an appropriate starting structure and clean up the PDB file. Here, we start with the copper-bound version of CuB Mb, 4FWY.

1. The first step would be to select all the necessary components of the system and save it as a separate PDB. This includes the protein chain, heme, and single Cu at the coordination site. This will apply PyMOL’s atom ordering rule, making atom numbering more consistent downstream.

2. With the new PDB, we first protonate the heme through command h_add. Delete excess hydrogens on the propionate group and rename the heme hydrogen based on CHARMM’s format. (Fig. 6) This is needed for the program to recognize the ligand. This can be done through PyMOL’s rename command, or through direct edit of the raw PDB

3. Determine the protonation state of coordination residues. Here, all relevant histidines that coordinate the CuB center and heme are protonated at the epsilon position, giving them HID residue labels. Fill out the dictionary RES_VARIANTS

4. Create a special_residue.xml for copper. Refer to ASH tutorial and OpenMM’s README for more specifications.

5. Modify input file paths and adjust parameters for water box in template. Run the first step of the script to generate a solvated system.

   ‘conda activate ash_env; python md.py solvate’

   Visually check finalsystem.pdb for protonation state and potential unwanted water molecules in the active site.

6. Fill out the BOND_DICTIONARY and FREEZE_ATOM list. Keys are there to make atoms easier to track and are non-sensitive. Note that atom indexing is based on the absolute order of the atom in the file counting from 0, which does not necessarily match the PDB atom column. Here, we are constraining the histidine ligation and CuB center. Frozen atoms are treated as massless and freezing specified heme atoms in this case prevent significant distortion. Sometimes, incorrect indexing creates ill-conditioned constraint matrix, causing simulations to fail. This is a relatively common error that can be detected by looking at bond distance associated with given indexing, which are printed in the output.

7. Run the second step of the setup, ‘python md.py run’. This would carry out a simple equilibration protocol: Quick minimization, Warm up, NPT box equilibration, and NVT canonical ensemble, with all bonds between hydrogen-heavy atom restrained.

Fig. 6:

CHARMM naming scheme for heme.

Once simulation has finished, the trajectory can be viewed with PyMOL. First, open a representative PDB, then load corresponding .dcd file with command load_traj. Observation of the backbone around CuB active site suggests that the apo structure largely supports the rotamer needed for copper binding, with only HIS 29 making a small rotation away from the site (Fig. 7). The copper coordination, however, is not perfect. Compared to the Apo structure of CuB Mb PDB: 4FWX, we see that HIS 43 takes on a different rotamer profile from the one required for coordination. This, in conjunction with constraining bonds of the CuB site, could explain why RMSD of the Cu-bound form is higher than that of the Cu-free from (Fig. 8). This tells us that some strain is present at the CuB center in CuB Mb.

Fig. 7:

One snapshot of the molecular dynamics simulation (green) alongside copper bound (blue PDB: 4FWY) and copper free (orange PDB: 4FWX) CuB Mb.

Fig. 8:

Rolling RMSD sample of NVT simulation run.

6. Expression, purification, and cofactor incorporation of Cytochrome C Peroxidase mutants

6.1. Culture growth and expression

6.1.1. Equipment

VWR mini incubated shaker incubator

Quincy lab, Inc. 10–140 Bench Top Incubator

New Brunswick Excella E-25 large incubator

Sorvall Lynx 6000 centrifuge

Fiberlite F9–6 × 1000 LEX Fixed Angle Rotor

Thermo Scientific^™ 1000 mL Fiberlite Bottles

−80 Freezer

IMPORTANT: the same equipment is used for the culture growth and the harvesting of the other proteins explained in this chapter.

6.1.2. Buffers and chemicals

Tryptone, Yeast Extract, Sodium Chloride, deionized water, SOC outgrowth medium, IPTG (Gold Bio), Ampicillin (Gold Bio), Kanamycin (Gold Bio), agar cell stocks/colony plates of CcP mutant

6.1.3. Procedure

Media is a variant of the 2xYT consisting of 16 g/L tryptone, 10 g/L yeast extract, 5 g/L NaCl in DI water or water filtered via Milli-Q IQ water purification system from Millipore Sigma. The media is autoclaved before usage.

1. 5 mL media cultures of the protein with 50 μg/mL of the antibiotic (in our case kanamycin) is made from its colony or frozen cells stocks and grown between 10–14 hours at 37C with 200 RPM shaking rate.

2. Large growth was done via either 6-liter flask containing 2 liters or a 4.5-liter flask containing 1.5 liters of media. All the 5 mL of cell culture is then added to per media flask, antibiotic stock solution is added to the concentration of 50 mg/L and shaken for at least four hours at 37 °C and 200 RPM or until its optical density with respect to ungrown media reaches the range of 0.4–1.0.

3. Afterwards, the shaker is cooled down to 18 C and then aqueous IPTG stock solution is added to a final concentration of 300 μM. The flasks are shaken for 18–24 hours.

4. Cells were harvested from media via centrifugation at 13800 rcf for 10 minutes at 4 C. The supernatant solution was discarded, and the pellet was flash frozen with liquid nitrogen and stored in a −20 °C freezer until further use.

6.1.4. Notes

1. Transformation of the plasmid is done via the NEB protocol to the BL21DE3 cell line.

2. Colony plates are functional for up to a month however it is recommended to use them less than a week after plating them. Flash-frozen glycerol stocks are recommended for long term storage, obtained from a 5mL overnight growth culture, and mixed of 900uL of cell media with 600uL of 60% glycerol and then flash frozen and stored in a −80C freezer.

3. Harvested cell pellets are stable for months at −20C.

4. The typical yield is about 7mL of pellet per 1L of media

6.2. Protein purification

6.2.1. Equipment

Misonix S-4000 ultrasonic sonicator

Fiberlite^™ F21–8×50y fixed angle rotor

Sorvall LYNX 6000 centrifuge

Millex^® 33 mm, 0.45-micron PES syringe filter

AKTA Go FPLC system

Cytiva C16–20 column with adapter, C26–100 column with adapter

Amicon^® Stirred Cell with 10kDa MWCO

IMPORTANT: the same equipment is used for the purification of the other proteins explained in this chapter.

6.2.2. Buffer and Reagents

Ethanol, K₂HPO₄, KH₂PO₄, imidazole, Beta-mercaptoethanol (BME), 1,4-dithiothreitol (DTT), Tris, NaCl, KCl, EDTA, phenylmethylsulfonyl fluoride (PMSF), egg white lysozyme (Gold Bio), Triton X-100, Pierce^™ Universal Nuclease, Cytiva Ni Sepharose High Performance nickel-charged IMAC resin 25mL, Cytiva Sephacryl^™ S-200 High Resolution (HR) resin, Ethylenediaminetetraacetic (EDTA).

1. Buffer A: 50mM K₂HPO4/KH₂PO4 pH 7.7, 250mL KCl, 5mM BME, 10mM Imidazole

2. Buffer B: 50mM K₂HPO4/KH₂PO4 pH 7.7, 250mL KCl, 5mM BME, 500mM Imidazole

3. TEV-P Buffer: 50mM Tris-HCl pH 8, 1mM EDTA, 1mM DTT

4. SEC Buffer: 50mM Tris-HCl pH 8, 1mM EDTA, 150mM NaCl, 5mM BME

6.2.3. Procedure

6.2.3.1. Lysis and Ni-IMAC purification

1. For each 1mL of pellet the following solutions are added: 9mL K-A Buffer, 10uL Triton X-100, 10mg egg white lysozyme, 100uL of 100mM PMSF in ethanol and to the final solution, regardless of volume, 10uL nuclease is added. Pellet is resuspended and solution is stirred in the cold room for 1 hour.

2. Solution is sonicated for 5 minutes total in 2 second pulses with 8 seconds of rest in between, at amplitude 70.

3. Sonicated solution is then centrifuged in the at 38000rcf for 30 minutes in 4C. Supernatant is filtered via the syringe filter.

4. Protein is purified from the clarified cell lysate via a Ni-IMAC chromatography using the FPLC system. Protein is purified using a combination of buffer A and B.

   1. Column is equilibrated with buffer A.
   2. Clarified cell lysate is loaded into the column and afterwards is washed using 5–10 column volumes (CV) of 15% buffer B.
   3. Protein elution is done by 50% buffer B. Elution is done for three column volumes. Then a wash with only buffer A buffer for 1 column volume is done followed by a last isocratic wash with 100% elution buffer for two column volumes. Protein elution is monitored from the 280 absorption peaks of the chromatogram and is confirmed via either a protein gel or through mass spectrometry.

6.2.3.2. Tag removal and final purification

1. 1.The protein peaks in the 50% and 100% elution buffer wash steps are pooled and buffer exchanged via the Amicon into TEV-P buffer.

2. 2.Protein is then incubated with a purified TEV-protease either purchased or purified using the protocol of (Tropea et al., 2009) at a A280 ratio of between 1:20–100 overnight with gentle stirring at 4C.

3. 3.Afterwards, solution is pelleted, and supernatant is buffer exchanged again into buffer A.

4. 4.Tag-cleaved protein is separated from the remainder TEV-protease and un-cleaved protein using the following elution profile: solution is loaded into the HP Ni-IMAC column equilibrated with buffer A and the flowthrough is collected in addition to a 5CV 5% buffer B eluent.

5. Collected eluents are concentrated to a volume of less than 5mL.

6. Final purification is done via size exclusion chromatography (SEC) through a Cytiva S-200 HR resin in a C26–100 column with a volume of ~400mL equilibrated to the SEC buffer. Purified proteins were confirmed by either ESI-MS and SDS-PAGE and either concentrated below 2.5mL and transferred to the glovebox or used for aerobic cofactor incubation.

6.2.4. Notes

1. Not more than 30mL of pellet at a time is purified.

2. If the lysate supernatant after centrifuging is still cloudy, it should be centrifuged again under the same conditions.

3. The added steps in 5.3.3.1.4.3 ensure to our experience that no CcP is left inside the column. In principle a 100% buffer B wash should achieve the same outcome.

4. TEV-P reaction will result in the TEV-P enzyme to precipitate, this is expected.

5. TEV-P can be reapplied to the reaction system if the cleavage yield is not high.

6. The TEV-P reaction tolerates sodium chloride concentrations up to 200mM NaCl and a pH range of 7–9. The presence of EDTA and DTT, however, is mandatory.

7. The protein solution transfer to the glovebox will introduce a small amount of air to the box, however it will be negligible since it is transferred in a 1.5mL tube and the buffer will be exchanged in the next steps.

8. Purified CcP is not amenable to freeze thaw cycles for storage and is not stable for more than a week.

6.3. Heme incubation

6.3.1. Equipment:

AKTA Go FPLC system (Cytiva)

C26–100 column with adapter (Cytiva)

HiTrap^™ Q Sepharose Fast Flow anion exchange column (Cytiva)

Disposable PD 10 Desalting Columns (Cytiva)

UV-Vis spectrometer

1cm quartz cuvette

Amicon^® Ultra Centrifugal Filter, 10kDa MWCO

Amicon^® stirred cell with 10kDa MWCO membrane

6.3.2: Buffers and chemicals

Hemin chloride

1. PD10 buffer A: 50mM Tris pH 8 150mM NaCl

2. QFF buffer A: 50mM Tris pH 8

3. QFF buffer B: 50mM Tris pH 8 500mM NaCl

4. SEC buffer: 50mM Tris pH 8 150mM NaCl

5. Sodium hydroxide 1M

6.3.3. Procedure

1. Apo protein is diluted to 100uM in PD10 Buffer A and is stirred at 4C to cool down.

2. Hemin chloride is dissolved first in a minimal amount of 1M NaOH and then diluted to 5mM of PD10 buffer A. Additions of 0.25 eq hemin solution are done slowly to the protein while stirring, and the additions are monitored by UV-Vis and repeated every 20 minutes.

3. After adding a max of 3 eq of hemin, the solution is left to stir overnight at 4C.

4. Holo protein solution is purified with QFF column through a gradient of 0–500mM NaCl, the protein usually eluted at 100–200mM NaCl depending on the variant

5. Column eluent fractions are assessed with the UV-Vis and those with the highest Soret peak to 280nm absorption are pooled, concentrated and purified through the SEC column via the 6.2.3.2 protocol.

6.3.4. Notes

1. An Rz value of 2 or higher is considered satisfactory for heme incubation.

2. The heme additions can conclude before 3eq if the latest addition has not correspondingly increased the Rz or if the Soret peak starts to significantly broaden, which is a sign of free heme absorption.

3. To increase the column longevity, one can buffer exchange the holo protein solution with a PD10 column first to prevent heme precipitation into the subsequent columns.

7. Expression, purification, and cofactor incorporation of Myoglobin mutants

7.1. Culture growth and expression:

Protein is compatible with all common rich media, with 2XYT being the default recommended choice. One round of purification typically use of 4–8L media total, giving 10 – 14g cell mass.

1. Start small culture by picking single colony/scrape some glycerol cell stock and inoculate 5mL LB media with ampicillin(100mg/L). Put into 37°C incubator shaker for 6–8 hours.

2. Add 1mL small growth culture into 2L of large growth media in 6L flat bottom flask. Shake at 200 RPM, 37°C overnight. (~12–16hrs)

3. To induce, add IPTG(100mg/L) to overnight culture. Let shake at 200 RPM, 37°C for 4hrs.

4. Harvest by pelleting at 13800 rcf, 10 min. Flash freeze pellet in liquid N2 and store in −80 for further use.

Overnight low temperature growth at 18 degrees sometimes increases soluble expression. Although Mb is generally purified without an affinity tag, the addition of polyHis tags can help during rapid smaller scale trial purifications of soluble variants. In this case, it is recommended to remove all imidazole before heme incorporation.

7.2. Purification

7.2.1. Buffers and chemicals

1. Lysis buffer (12mM Tris, 1mM EDTA, 1mM DTT), 10ml buffer for 1g cell mass, 100ml per batch. Immediately before use, add 1ul Nuclease and 1ml 100mM PMSF.

2. Cold MiliQ water

3. Beaker with 750ml of cold MiliQ water

4. Dialysis buckets. (2 buckets per batch, each consisting of 11L H2O + 1g monobasic potassium + 12g dibasic potassium. Pre-chill at 4 degrees for at least 3 hours before use.)

5. Guanidinium hydrochloride (GnHCl) in a glass bottle. Use 33.3 g for each batch.

6. Hemin solution: add 14mg of hemin chloride into small falcon tube, add 10ml H2O and dissolve heme by adding a few drops of 1M NaOH

7. Ion exchange resin, either DEAE Sepharose, QFF, or CM

8. S-200 Sephracyl resin for SEC Column

9. Spectrum^™ Spectra/Por^™ 3 RC Dialysis Membrane Tubing 3500 Dalton MWCO

7.2.2.2. Procedure

A. Inclusion body purification protocol:

Most engineered Mb express as inclusion bodies, with the following protocol being a generalizable approach for solubilizing globular heme protein. One typical batch uses 8L or more of expression media, yielding 10 −14 g of cell mass

1. All purifications are done with protein kept on ice or at low temperature. First, thaw and resuspend pellet in a metal cup and let stir until homogeneous for 15–30 minutes. Addition of lysozyme can help, especially if culture is grown at low temperature overnight

2. Cells are lysed by sonication under temperature control at amplitude = 70, 12min process time with 6s/12s on/off pulse interval to prevent overheating. Lysates are then centrifuged at 20000 rcf for 30 minutes.

3. Discard supernatants and resuspend pellets in cold MiliQ water.

4. Centrifuge again at 20000 rcf for 30 minutes to wash the pellet, then discard supernatants.

5. Resuspend each 50 ml pellet in 2 ml of cold water, and transfer solution into a small bottle containing GnHCl.

6. Add H2O to increase total volume to 100ml. Let GnHCl slurry unfold protein for 1 hour by shaking slowly under cold temperature.

7. Add heme solution into the beaker with 750 mL MiliQ water using a transfer pipet while stirring.

8. After protein has finished unfolding, slowly add GnHCl slurry dropwise into the stirring beaker, allowing protein to refold in the presence of heme while local GnHCl concentration is diluted by the solution. Flask should turn red or brown, with slower addition leading to better refolding yield.

9. Let stir slowly for 1 hour after all GnHCL addition. Centrifuge solution at 13800 rcf for 30 minutes to pellet out aggregates.

10. Transfer supernatant into 10kDA dialysis membrane. Let dialyze in the potassium phosphate (KPi) bucket for 3 hours.

11. Transfer tubing into the second bucket and dialyze overnight.

12. The following day, centrifuge dialyzed solution at 13800 rcf 20 minute to remove precipitated protein.

13. Concentrate solution into 3–8ml with 10kDA Amicon and run protein through a small gravity DEAE Sepharose column to remove excess heme.

14. Briefly filter protein using 22 uM PES membrane. Protein can then be purified using size exclusion chromatography (50mM potassium phosphate KPi pH 7, Cytiva S-200 resin C16 column 50cm).

15. If the ratio between the Soret and the 280 peak is not at least 3.5 or above, a follow-up purification with anion exchange column can be helpful

16. Pool all colored Mb fractions from SEC run, concentrate to less than 0.5 CV, and load onto QFF resin equilibrated in 50mM KPi pH 8. Flow sample through using this same buffer.

7.2.3.3. Note:

Most Mb variants do not bind anion exchange resin at the above conditions. Thus, the final QFF column enables additional removal of free heme and further separate samples by their Rz ratio. A good purification with homogeneous and complete heme incorporation would contain some proteins with Rz value above 4. Anion exchange fractions that fit these criteria can be further used in assay and crystallographic screening.

B. Soluble holo or apo Mb protocol:

All steps up to post-sonication are similar to inclusion body protocol. Oftentimes, inclusion body – soluble is not a binary outcome, with growth and IPTG concentration during induction affecting how much of the protein is expressed as soluble. If protein expresses as soluble apo, perform hemin chloride addition using solution f) at the initial lysis step prior to sonication.

7.2.3.1. Procedure

• 2. Adjust the supernatant pH to <6.1 b y adding 0.3M HCl dropwise. To minimize protein loss, adjust the pH with the supernatant on ice with fast stirring. Stirring should be at least at a speed of 4–5 on a standard stir plate.

• 3. Use milliQ water to dilute the solution to a conductivity less than 2 mS.

• 4. Check pH again after adjusting conductivity to ensure pH is <6.2. Adjust pH/conductivity until pH is less than equal to 6.2 and conductivity less than 2mS. If there is visible white precipitate or if solution is slightly cloudy, centrifuge supernatant.

• 5. To prepare CM beads for batch binding, equilibrate beads with 10mM KPi pH6

• 6. Before mixing all of the supernatant with the CM beads, mix a small amount of the protein with the CM beads in an Eppendorf tube (100uL CM bead slurry + 1mL protein supernatant) and invert the mixture. Do a quick spin to pellet the CM beads to the bottom and make sure the beads are a pink/red color signifying the myoglobin is binding to the beads.

• 7. Add pH/conductivity-adjusted protein to CM beads. Shake in the cold for at least 2 hours. If needed, overnight shaking can also be done

• a. For overnight shaking, add PMSF (few speckles dissolved in ethanol) as a protease inhibitor

• b. To determine if batch binding is complete, allow the beads to settle and visually inspect them: the beads should be dark red and more intensely colored than the solution.

• 8. Decant the beads onto a coarse frit filter attached to a vacuum filtration flask and draw vacuum to pass the flow through off the beads. Some Mb will flow through

• 9. Wash the beads with 10mM KPi pH6 until the filtrate is colorless.

• 10. Elute the protein with 100mM KPi pH7 + 100mM NaCl. Beads should return to pale white once protein is fully eluted.

• a. Add ~50–100mL of elution buffer and stir the beads to create a slurry. Let the beads sit in the elution buffer for 5–10 minutes before pulling vacuum.

Concentrate the protein by Amicon to 10mL or less and filter it through a 0.45 uM syringe filter. Load the protein onto SEC equilibrated in KPi7and collect colored fractions.

8. Copper titration protocol for both Cu_BCcP and Cu_BMb

8.1. Equipment:

Glove box anaerobic Chamber (VTI) containing a bead bath setup

Disposable PD 10 Desalting Columns (Cytiva)

Schlenk line manifold setup

90-degree vacuum adapter with glass stopcock

250mL round bottom flask

UV-Vis spectrometer

1cm quartz cuvette

Amicon^® Ultra Centrifugal Filter, 10kDa MWCO

Amicon^® stirred cell with 10kDa MWCO membrane

8.2. Buffers and chemicals

1. Freshly made aqueous 10mM CuSO₄

2. trakis(acetonitrile)Cu(I)hexafluorophosphate

3. Acetonitrile

4. Sodium Hydrosulfite

5. Copper inoculation buffer 1: 50mM Tris pH 8

6. Copper inoculation buffer 2: 50 mM BIS-Tris pH 7.3

7. Desalting buffer: 20mM MOPS pH 7

8. SEC buffer: 50mM Tris pH 8 150mM NaCl

8.3. Aerobic titration with Cu²⁺

1. Add 1–1.5 equivalent of this copper solution into the protein solution. The ideal range for protein concentration is 100uM - 300 uM. Protein should be in Tris pH 8.

   1. Let protein solution sit with stir bar on ice, stir slowly.
   2. Pipette 0.25 equivalent of the CuSO₄ solution. Add the volume slowly over the span of a few minutes, hold the pipette tip above the stir bar.
   3. Wait ~15 minute before next addition.

2. After finishing titration, let protein sample stir for at least 30 minutes

3. Switch the protein into a non-ligating buffer, like 50mM MOPS pH8, using an Amicon protein concentrator or spin Centricons.

4. Run protein through a PD10 column to remove excess copper

5. An example of the UV-vis spectra is shown in Fig. 9.

Fig. 9:

UV-vis spectra of high-spin ferric WTMb (solid line) and Cu_BMb (dashed line). Spectra were recorded with the samples in 100 mM KH₂PO₄ pH 7 at 20°C. The figure is adopted from Sigman 2000.

8.3.1. Notes:

1. Spectra of Cu²⁺ bound and non-bound Mb protein tends to be very similar.

2. For most variants, adding Cu²⁺ can cause partial precipitation, especially in alkaline conditions.

8.4. Anaerobic titration with Cu⁺

All manipulations from this stage, unless explicitly stated, are done within an anerobic chamber and use oxygen free buffers.

8.4.1. Anaerobic solution preparation:

1. Buffers are degassed through three cycles of freeze-pump thawing.

2. Alternatively, high concentration of stock solutions can be prepared, if possible, degassed through freeze-pump-thawing and milliQ water can be degassed by overnight sparging with N₂.

For low volume salts, the measured powder is transferred inside the box and the dissolved in the appropriate solvent.

8.4.2. Procedure:

1. Transfer protein into anaerobic chamber and reduce with dithionite by adding a freshly made 1M stock to a final concentration of 1mM.

2. Buffer exchange sample into 50 mM BIS-Tris pH 7.3 using PD10 column to make sample fully anaerobic and remove the dithionite.

3. Concentrate protein down to 100uM – 300uM, with absorbance monitored via Soret band. Let the protein cool down in a cold temperature metal bath with a stir bar on low speed.

4. Make a 10mM solution of Cu(I) by dissolving trakis(acetonitrile)Cu(I)hexafluorophosphate in acetonitrile.

5. Titrate until reaching 1–1.2 equivalent:

   1. Pipette 0.25 equivalence of copper solution.
   2. Add volume above stir bar. Add slowly over the span of one minute. This is because the combination of metal and organic solvent can precipitate protein when added with quick succession.
   3. Wait 15 minutes to perform next addition.

6. Remove excess copper by concentrating sample and running it through a PD10.

7. Protein can be transferred outside the glove box after copper addition

8.4.2.3. Note:

The amount of copper needed to reach saturation for each protein varies. Unlike Cu(II), Cu(I) titration produce a strong UV signal change (Bhagi-Damodaran et al., 2017), providing a convenient method for monitoring. For Cu_BMb, due to the extra metal binding outside of the tri-histidine center, up to 3 equivalences of Cu(I) can be used to reach Cu_B saturation.

9. Characterization of the Cu_BCcP and Cu_BMb

The structural integrity of the Cu_BCcP and Cu_BMb can be assessed by several methods. First, the UV-vis spectra of these mutants can be compared with those of wild-type proteins (see Fig. 9), as a well-folded protein with heme in the designed position should display a sharp Soret band around 400 nm. In addition, other spectroscopic methods, such as electronic paramagnetic resonance (EPR), magnetic circular dichroism (MCD), resonance Raman (RR), and x-ray absorption spectroscopy (XAS) can be used to characterize their electronic structures in comparison to those of either WTCcP/WTMb or HCOs. The most direct evidence for the structural integrity of these designed proteins is 3D structures of these proteins using either x-ray crystallography or NMR.

Fig. 4:

Rosetta Matcher run command with relevant flags. These specific flags will group different rotamers of the same residue match positions as a cloud pdb file to save file space.

Table 1.

Notable metalloproteins and their activating cofactors. Cofactor element colors are as follows: S: Yellow, Ca: Green, Mn: Red, Fe: Orange, Ni: Green, Cu: Brown, Mo: Cyan.

Protein and protein family	Cofactor(s) [Formula]	Function	Notable Host/ Expression Organism	Expressed in E. Coli?	Reference
Myoglobin	Heme B [C34H32O4N4Fe]	Oxygen transport	Horse heart	Yes	(Springer and Sligar, 1987)
Cytochrome P450	Heme B [C34H32O4N4Fe]	Oxidation, Hydroxylation, Monooxygenation, Biosynthesis and Xenobiotic degradation	Rabbit liver	Yes	(Omura and Sato, 1962)
Heme Copper Oxidase	Heme O [C49H58O5N4Fe], Heme B [C34H32O4N4Fe], CuA, CuB centers	Oxygen reduction, Nitric oxide reduction	Thermus thermophilus, bovine heart	No	(Person et al., 1953; Tiefenbrunn et al., 2011; Shinzawa-Itoh et al., 2021)
Dissimilatory Sulfite Reductase	Siroheme [C42H44FeN4O16], Fe4S4 cluster	Sulfite reduction	Desulfolvibrio desulfuricans E. coli	Yes	(Peck, 1959; Cepeda et al., 2018)
Nitrogenase	FeMoco [Fe7 MoS9C], P-cluster [Fe8S7]	Nitrogen reduction	Azotobacter vinelandii, Clostridim pasteranium, Klebsiella oxytoca	Yes	(Bulen and LeComte, 1966; Solomon et al., 2024; Warmack et al., 2024)
Carbon Monoxide Dehydrogenase	NiFe4S4 cluster	CO oxidation	Rhodopseudomonas, Rhodopspiridum rubrum	No	(Uffen, 1976)
Soluble Methane Monooxygenase	Fe-Fe core	C-H activation	Methylosinus trichosporium OB3	No	(Tonge et al., 1977)
Particulate methane monooxygenase	CuB, CuC, CuD centers	C-H activation	Methylococus capsulatus	No	(Ribbons and Michalover, 1970)
Azurin	T1 Cu center	Electron transport	Pseudomonas aeruginosa	Yes	(Sutherland and Wilkinson, 1963)
Laccase, blue multicopper oxidase	T1 Cu center, T2/T3 trinuclear Cu center	Biopolymer degradation, oxidation	Coprinus cinereus	Yes	(YA Pamphlet Collection (Library of Congress) DLC, 1901)
Photosystem II	Oxygen evolution complex [Mn4O5Ca]	Water splitting	Rhodopseudomonas veridis	No	(Kok and Datko, 1965)