Firstly, we chose 11 common redox partners and built a redox partner library. Based on molecular docking and the calculation of binding free energy, we initially screened out four redox partners.

Next, we analyzed the interaction mechanism between P450 and redox partners from multiple perspectives.

In addition, we predicted 32 key residues on the petF interface and mutated them to enhance the interaction between P450 and petF.

Finally, we constructed a novel sfGFP sensor to test 23 mutants and successfully obtained a mutant (D68P) that the catalytic conversion rate can be improved from 85.6% to 89.2%.

P450 enzymes can catalyze thousands of reactions and various substrates with complex chemical structures, which have been described as versatile catalysts. There are several steps involved in the P450-catalysed reaction, including the reduction of molecular oxygen via an electron transfer system and the addition of oxygen atoms to the substrate (RH), which requires NADPH.

The P450 enzymes were classified according to the different kinds of redox partner. Class I and Class II are the most common types, which were studied in this project.
The first class of redox partner for P450 enzymes (Fig. 1), commonly found in bacterial and eukaryotic mitochondria, contains a ferredoxin (Fdx) and an iron-sulfur cluster (Fe2S2) with a FAD-containing ferredoxin reductase (FdR).

Fig. 1 Class I

The second class of redox partner (Class II) for P450 enzyme is the FAD/FMN-containing cytochrome P450 enzyme reductase (CPR), commonly found in eukaryotes and many cases as a membrane protein (Fig. 2).

Fig. 2 Class II

The previous work mainly focused on the screening of suitable redox partners for P450 enzymes, but the mechanism of between P450 enzymes and redox partners is still unknown. In addition, the fast and accurate screen of the most suitable redox partner for P450 enzymes is another big challenge for researchers.

The importance of redox partners for the catalytic efficiency and functional regulation of bacterial P450 enzymes, redox partner screening, and redox partner modification have received increasing attention in recent years.^[1]

Deshuai Lou used Maestro, based on the amount of binding free energy change (DDG), to perform virtual saturation mutagenesis, and screened seven mutants, but experimentally verified that the activities of all seven mutants decreased. ^[2] Tanja Sagadin found that mutants with a shorter distance between the heme iron in P450 enzyme and the Fdx iron-sulfur cluster had higher catalytic activity.^[3] Shuaiqi Meng found that the introduction of aromatic amino acids on the redox partner can improve the catalytic activity. ^[4]

Adoption of rational indicators to explain the interaction mechanism between P450 and redox partners.

Establishment of hotspot residue screening model.

Selection of rational scoring function for virtual saturation mutagenesis.

Table 1 Building a redox companion library

The catalysis of P450 enzymes involves electron transfer, which must be transferred by redox partners, and the combinatorial expression of P450 enzymes with different redox partners can achieve the reconstruction and enhancement of reaction activity. To screen out suitable redox partners, we selected more than ten common redox partners as libraries, simulated the spatial structure of the complex formed by the interaction between the P450 enzyme and redox partners based on GRAMM molecular docking, which improves docking accuracy compared to other docking platforms. Through using the Lennard-Jones method for calculating intermolecular energies during docking instead of the traditional lattice method calculation, and screening the optimal docking pose based on the distance between the iron-sulfur clusters and the heme iron, we calculated the binding free energy of each complex structure. ^[5]

Fig. 3 The binding free energy with MM/GBSA

Based on the 11 complex structures after docking, their binding free energies were calculated using Prime MM/GBSA, respectively, and the top four redox partners (BM3-R, petF, SelFdx1499, CamB) were screened according to the binding free energies (DG_bind, the smaller the better).

DG_bind=G_com- (G_rec+G_lig)

Where DG_bind is binding free energy, G_com,G_rec,G_lig are Gibbs functions for complex, receptor, and ligand respectively.

Fig. 4 Free energy calculation on P450 and redox partners

After validation of the sensor and fermentation experiments, among the four pairs of redox partners that have been screened, it was found that the highest intensity of sensor fluorescence was observed when petH/petF was used, while the catalytic efficiency also showed the best performance. This suggests that petF interacts most strongly with Olep.

Fig. 5. Strategies to construct sfGFP sensor to screen redox partners. (A) The scheme of constructing sfGFP sensor. (B) The self-assembly of Olep and Fdx based on the three-dimensional structure of sfGFP (PDB: 5BT0). (C) Screening proper redox partners for OleP from different sources. The G1 strain that contains the empty pRSFDuet-1 plasmid was used as a control. The fluorescent intensities were calculated and the color of cells and fluorescent images were presented for G2-G5 strains that express different redox partners-sfGFP-1-10 and sfGFP-11-Olep, respectively. (D) The conversion rates were calculated for R2-R5 strains that express different redox partners and Olep, respectively. The R1 strain that contains the empty pRSFDuet-1 plasmid was used as a control. The blue-filled triangle represents the fluorescent intensity/OD600. The red hollow triangle represents the conversion rate (%). Values and triangles represent the means and standard deviations of biological triplicates.

Catalysis by P450 enzymes involves electron transfer, and their electron transfer pathways have a great influence on their catalytic efficiency.

While the electron transfer process is inherently complex, it is primarily governed by the spatial separation between the electron donor and electron acceptor. In the context of an approximate non-adiabatic process, such as enzymatic reactions, it is often feasible to simplify the electron transfer rate with a distance-dependent equation.

k_ET = A(0)e^-βR

Herein, A(0) represents a constant, R denotes the distance between the electron donor and electron acceptor, and β serves as a parameter to gauge the capacity of the intervening matrix in facilitating electronic coupling or superexchange.

So we measured the spatial distance between the iron-sulfur cluster in the chaperonin and the heme in Olep (Pymol), and we have labeled the measurements in the docking model above. For the Olep-petF complex, the distance between the iron-sulfur cluster and the iron atom of the heme is 17.7 Å (Fig. 7a), which is shorter than the corresponding distances observed in the other complexes, suggesting that the shorter iron-sulfur cluster and heme iron atom distance may increase the efficiency of electron transfer, which in turn affects the catalytic activity.

The interfacial areas of the interactions of these complexes were calculated using PDBePISA, with the largest interfacial area of the interactions being formed between petF and P450, which amounts to more than 3600 Ų, which implies a larger range of electrostatic interactions between them. The larger interaction area helps to achieve more efficient electron transfer. (Fig. 7b)

Fig. 7 P450-Fdx docking models. (a) The distance between the iron-sulfur cluster and the iron atom of the heme. In the docking models, the green is Olep, the other colors present are redox partners, the heme is represented by sticks, and the iron-sulfur clusters are represented by balls (b) Interface area in the P450 enzyme binding with different redox partners. The green area is the interface area on Olep.

At the same time, we took into account the long-range interactions that play an important role in protein binding, mainly manifested as electrostatic interactions between P450 and redox partners, and we analyzed the electrostatic surfaces (pymol) of these several proteins and computed the total charge at the interface of each protein interaction. In Fig. 8, positively and negatively charged surfaces are colored blue and red, respectively. We can visualize from the figure that petF has more negatively charged surfaces compared to other redox partners, with an interfacial charge of -8, which implies stronger electrostatic interactions. This gives petF a greater chance to form an ideal protein-interaction interface with P450, resulting in more efficient electron transfer.

Fig. 8 Electrostatic surface and interface charge

Interfacial Interaction Force Statistics

To obtain more details of the interface, we predicted the hydrogen bonds and salt bridges formed at the interaction interface (PDBePISA), as can be seen from the number counts on the left side, petF and P450 formed the most number of hydrogen bonds and salt bridges at the interface with a total of 27 forces, which suggests that petF is a more stable complex with P450 than the others, whereas on the right side of this figure, we locate exactly which residues are involved in the formation of the forces at the interface formed by P450 and petF, and these strong forces prove the strong interaction between the P450 enzyme and petF.(Table 2)

Table 2 Interfacial interaction force statistics

i. Concept of Hotspot Residues

The key amino acid residues at the interface of protein complexes are called hotspot residues, which play a crucial role in the formation, stability and specific recognition of the complexes, and it is necessary to focus on these amino acid residues when performing semirational modification.

ii. Properties of Hotspot Residues (Biological Principles)

At the interaction interfaces

Thirty-two interaction interface residues on petF were screened using the Pymol InterfaceResidues plugin, and since the hotspot residues were all at the interaction interface, the screening model was based on these 32 residues.

Table 3 Interface residues on petF

The contribution to the binding free energy is large compared to other interfacial residues

The contribution of hotspot residues to the binding free energy is higher than that of other interfacial residues, and alanine scanning using FoldX to calculate the change in binding free energy can reflect the contribution of this residue to the binding free energy to some extent.

DDG = DG_M - DG_WT

Where DG_WT binds binding free energy of the wild type, DG_M binds binding free energy of the mutant.Where DG_WT binds binding free energy of the wild type, DG_M binds binding free energy of the mutant.

Fig. 9 Top 15 residues in Alascan

Located at the hub of the force network, interacting with multiple neighboring residues to form multiple interaction forces

Hot spot residues have more forces at the interaction interface. Using RING, Cytoscape performed a residue interaction network analysis of Olep-petF and calculated the degree (i.e., the number of interfacial interaction forces) of the petF residue nodes in the network.

G = (R_i, I_j)

Where, G is the Residue Interaction Network.R_i is the set consisting of Residue Nodes, I_j is the set consisting of Interaction Forces.

D_i = degree(R_i)

Where, D_i is the set of corresponding degrees for each residue node.

Fig. 10 Residue interaction network. Red nodes are residues at the Olep interface, and cyan nodes are residues at the petF interface.

More conserved compared to other interfacial residues during evolution

Hotspot residues are more evolutionarily conserved, and the evolutionary conservation scores for each residue on petF were calculated using the Consurf platform, based on MSA.

M_ij = [((∆p_ij)/σ_p)² + ((∆v_ij)/σ_v)²]^(1/2)

Where, M_ijis the estimation of the spatial distance between the i,j-th residue. ∆p_ij ∆v_ij are the difference of the i,j-th residue caused by polarity and volume respectively. σ_p,σ_v are the standard error of polarity and volume respectively.

P_K^' = (P_k-〈P_k〉)/k

Where, P_K^' is the evolutionary score after normalization, k is the number of residues, 〈P_k〉 is the mean value of P_k.

Fig. 11 Conserved staining of protein residues

iii. The modeling process was designed based on the above principle and hotspot residues (7) were obtained

Based on the above principles, combined with the TOPSIS method, a commonly used evaluation method for mathematical modeling, the model flow was designed as follows. Screening for hotspot residues based on TOPSIS scores.

Fig. 12 Flowchart of the hotspot residue prediction model

The entropy weighting method was utilized to calculate the weights, and the results show that Degree has a weight of 33.43%, Consurf has a weight of 39.02%, and Ala has a weight of 27.55%, where the maximum value of the indicator's weights is Consurf (39.02%) and the minimum value is Ala (27.55%).

Table 4 Results for Predicting Hotspot Residue Models Top 7 residues

Virtual mutagenesis based on the seven hotspot residues on petF screened by the hotspot residue screening model described above and screening using the combined free energy change (DDG) as a scoring function yielded 23 positive mutants.

Fig. 13 Positive results for virtual saturation mutations

We verified the results of virtual mutations based on the sfGFP sensor. The results of fluorescence intensity showed that 4 of the 23 designed mutants showed enhanced fluorescence intensity compared to the wild type. (Fig. 14)

Fig. 14 The results of fluorescence intensity

Further, we selected nine mutants with the highest fluorescence intensity for catalytic experiments to explore the influence of petF mutants on the catalytic conversion rate. Finally, we successfully obtained a mutant that increased the catalytic conversion rate from 85.6% to 89.2% (22.D68P).(Fig. 15)

Fig. 15 The results of catalytic experiments on mutants

We designed a rapid screening strategy for redox partners based on molecular docking and binding free energy calculations. The experimental results demonstrate the effectiveness of the strategy. Further, we analyzed the mechanism of P450 enzyme interaction with redox partners, which can strongly explain our experimental results. In addition, we built a prediction model of hotspot residues and predicted the hotspot residues located on the petF interface, which gave us a deeper understanding of their interaction mechanism. We carried out virtual saturation mutation on the hotspot residues and verified them by experiments. Finally, we successfully obtained a mutant that increased the catalytic conversion rate from 85.6% to 89.2%(D68P).