Tunnel Engineering for Modulating The Substrate Preference in Decarbonylase P450BsβHI

An active site normally locates inside of enzymes, substrates should go through the tunnel to access the active site. Tunnel engineering is a powerful strategy for rening the catalytic properties of enzymes. Here, P450 Bsβ HI (Q85H/ V170I) derived from hydroxylase P450 Bsβ from Bacillus subtilis was chosen as study model, which is reported as a potential decarbonylase. However, this enzyme showed low decarboxylase activity towards long-chain fatty acids. Here, a tunnel engineering campaign was performed for modulating the substrate preference and improving the decarbonylase activity of P450 Bsβ HI. The nally obtained BsβHI-F79A variant had a 15.2-fold improved conversion for palmitic acid; BsβHI-F173V variant had a 3.9-fold improved conversion for pentadecanoic acid. The study demonstrates how the substrate preference can be modulated by tunnel engineering strategy.


Introduction
Enzymes are able to catalyze many speci c reactions and are widely used in practical application.
Previously, the enzyme functions were limited around their natural use; nowadays, the enzymes could be developed by engineering their activity and selectivity for meeting human demands (Bornscheuer et al., 2012;Damborsky and Brezovsky, 2014). Two common engineering strategies are the directed evolution based on Darwinian theory and rational design based on the structure-function relationship (Bornscheuer and Pohl, 2001). Rational design often focused on the substrate binding pocket which directly a uence the enzymatic process (Bornscheuer and Pohl, 2001). However, the experiences of directed evolution told us the residues outside the active site also in uence the enzyme properties (Kress et al., 2018).
Researching on those "non-hotspot" rather than typical active site may be bene cial for expanding the understanding of proteins.
The urgent problem now confronting us is how to obtain the important "non-hotspot" domains beyond the enzyme active site. Tunnel engineering may be one of the answers. Enzymes spanning all of the six classes are found the exist of the tunnels (Kingsley and Lill, 2015). It was reported that more than 64% enzymes annotated in Catalytic Site Atlas library have the buried active site with the tunnels connecting the enzyme binding pocket and the environment (Pravda et al., 2014). The tunnels could support the transport of solvent, product and solvent between the enzyme active site and bulk solvent, which play important role in enzymatic reaction (Kokkonen et (Kress et al., 2018). Here, we aim to develop the application of tunnel engineering for modulating the substrate preference. α-alkenes are multifunctional compounds that perform an important industrial value and an extraordinary economic importance due to their exible and active chemical performance. Particularly, long-chain αalkenes can be used in the synthesis of high-value biofuel, lubrication and surfactant (Lee et al., 2008). However, α-alkenes mostly come from non-renewable petroleum cracking (Dutta et al., 2014). Energyintensive process and harsh reaction conditions prompted researchers to focus on enzymatic synthetic strategy of α-alkenes (Schirmer et al., 2010). To date, three biotransformation strategies that convert fatty acids or its derivative to alkene were reported (Yi et al., 2014). A three-gene cluster from Micrococcus luteus represented a production of long-chain alkenes from a head-to-head condensation of fatty acids (Beller et al., 2010); a type I polyketide synthases from Synechococcus sp. was involved in a production of medium-chain α-alkenes via an elongation decarboxylation mechanism (Mendez-Perez et al., 2011); and some members from cytochrome P450 family are able to decarboxylate fatty acids and produce αalkenes directly (Grant et al., 2015;Hsieh and Makris, 2016). Among those pathways, the decarboxylic reaction catalyzed by P450 is the simplest strategy. Free fatty acids could be used as substrate directly, which represent potential of producing α-alkenes on a large-scale based in engineered cell factory.
in 1999. However, wild-type P450 Bsβ showed higher hydroxylation activity rather than decarboxylation activity (Matsunaga, et al., 1999). In contrast, P450 oleT presented prominent decarboxylation property, which is able to convert medium-or long-chain fatty acids to corresponding carboxylic acids (Rude et al., 2011). Interestingly, some P450 Bsβ variants also exhibited satisfactory decarboxylation ability. Xu et al.
reported a P450 Bsβ HI variant (Q85H/V170I) which displayed enhanced decarboxylation activity towards medium-or long-chain fatty acids (Xu et al., 2017). However, the yield of α-alkenes drops sharply as the length of the carbon chain of the substrate increases, which is the main limitation of the use of P450 Bsβ HI for long chain α-alkenes synthesis.
In the present work, we systematically analyzed the access tunnels in decarbonylase P450 Bsβ HI. In order to improve the substrate preference of P450 Bsβ HI to long chain fatty acids, two residues related to the access tunnels diameter were identi ed and mutated. In addition, the substrate selection mechanism controlled by tunnels of P450 Bsβ HI was brie y discussed.

Strain, plasmid and chemicals
Escherichia. coli strain BL21(DE3) cells (used for gene expression) and TOP10 cells (used for molecular cloning) were purchased from TransGen Ltd. (China) The gene BsβHI-wt (NCBI Reference Sequence: WP_119898938.1) was synthesized by Inovogen Ltd. (China). Fast Mutagenesis kit was from Vazyme Biotech Co., Ltd (China). Plasmid extraction kits and gel extraction kits were obtained from Omega Bio-tek (USA). Capric acid, lauric acid, myristic acid, stearic acid, undecene and tridecene were purchased from BioRo Yee Ltd. (China). All chemicals used were of analytical grade.

Molecular modeling and simulation
The structure model of P450 Bsβ HI was constructed by PyMoL Molecular Graphics System (version 2.3.3) based using the crystal structure of P450 Bsβ from Bacillus subtilis (PDB code: 1IZO with the resolution in 2.10 Å) as template (Lee et al., 2003). Molecular dynamics simulation was carried out in YASARA (version 17.8.15) using the built-in MD macro "md_run.mrc" with the AMBER03 force eld (Hess et al., 2008). Structure were solvated into a 12 Å cube simulation cell of water molecules. The box was lled with 3857 water molecules. The simulations of the protein-water system was performed at 303 K, pH value of 8.0. Na + and Cl − were used to neutralize the systems. In all simulations, constant pressure periodic boundary conditions were used for 5 ns MD production. The simulation snapshots were capture every 100 ps from 2.5 ns to 5 ns (after RMSD stabilizes). with the analytical parameters as follow: probe radius of 1.4 Å, shell depth of 4 Å, shell radius of 3 Å, clustering threshold of 3.5, and the starting point of surrounding the residues of R242, P243, and heme. After the tunnel calculations, the in uence of the tunnel by related residues was extracted from the function of tunnel statistics and residue graph. Finally, amino acid residues with signi cant in uence on the tunnel bottleneck were selected for further mutagenesis experiments.

Relative folding free energies (ΔΔGfold) analysis
The ΔΔGfold values were calculated using FoldX employing the YASARA plugin (version 19.12.4). The structure model of P450 Bsβ HI and its variants were constructed by PyMoL Molecular Graphics System (version 2.3.3) based using the crystal structure of P450 Bsβ from Bacillus subtilis (PDB code: 1IZO with the resolution in 2.10 Å) as template. The initial structure of P450 Bsβ HI was constructed by the Automated Modeling Tool of Swiss Model Web Service (http://swissmodel.expasy.org/) using the crystal structure of P450Bsβ from Bacillus subtilis (PDB code: 1IZO with the resolution in 2.10 Å) as template.

Site-speci c mutagenesis
The P450 Bsβ HI gene was inserted into plasmid pET22b(+) between the NdeI and HindIII restriction nuclease sites, with the His-tag encoding sequence at N-terminal. All of the variants were done by Fast Mutagenesis kit.
The variants with single mutation used P450 Bsβ HI as template, in which F79 and F173 site was dividually replaced with relatively small amino acids, including glycine, valine, alanine, serine, isoleucine, threonine, cysteine, leucine, and proline. Variants with double mutations used BsβHI-F173V variant as template.
2.6. Heterologous expression and puri cation P450 Bsβ HI and its variants were transferred to E. coli BL21(DE3) for expression. Recombinant E. coli BL21 (DE3) cells were grown in LB medium (5 g/L yeast extract, 10 g/L tryptone, and 5 g/L NaCl) supplemented with ampicillin (100 µg/ml) at 37 ℃ until the OD600 reached about 0.6. All of genes were induced by the addition of 0.1 mM of isopropyl-β-d-thiogalactopyranoside (IPTG) at 18℃ for 12 h. Then the cells were harvested and ultrasonic broken. Cell-free extractions were used for puri cation.
Enzymes were puri ed by His-tag a nity chromatography. Cell-free extractions was loaded onto a Ni-NTA column and equilibrated by Lysis buffer (50 mM of Tris-HCl, 10 mM of imidazole and 300 mM of NaCl, pH 7.8). Then the protein was sequential eluted by Wash buffer (50 mM of Tris-HCl, 30 mM of Iimidazole and 300 mM of NaCl, pH 7.8) and Elution buffer (50 mM of Tris-HCl, 300 mM of imidazole and 300 mM of NaCl, pH 7.8). Eluates were concentrated with an ultra ltration (Millipore, Germany). The concentrations of puri ed proteins were determined by BCA kit (Solarbio, China).

Enzymatic assay
The bio-catalytic system contained 10 µM P450 Bsβ HI enzyme (or its variants), 500 µM fatty acid substrate (from a 100 mM stock solution in ethanol) and 1 mM H 2 O 2 in a nal volume of 1 mL of 100 mM potassium phosphate buffer (pH = 8.0). The reaction was carried out at 30℃ for 2 h, then was quenched by additional 50 µL of 6 M HCl. The mixture was extracted by 800 µL hexane. Following extraction, the products were analyzed by gas chromatography-mass spectrometry (GC-MS-QP2020, Shimadzu, Japan) equipped with a Sh-Rxi-5Sil-Ms column (Shimadzu, Japan) using helium as carrier gas. The oven temperature was controlled initially at 50℃ for 2 min, then increased at the rate of 10℃ min − 1 to 280℃, and held for 10 min. The injecting temperature was 280℃. The peaks were identi ed by comparison of retention time and ion spectra with authentic references.

Results And Discussion
The result and discussion part is divided into three parts. In the rst part, the P450 Bsβ HI access tunnels were analyzed and two amino acids were identi ed as bene cial key residues. In the second part, the variants based on the two key residues were characterized with lauric acid substrate then 5 best variants were selected. In the third part, the substrate preference of selected P450 Bsβ HI variants was investigated.
3.1. P450 Bsβ HI access tunnel analysis and hot spots identi cation Access tunnels is responsible for ligand transportation between active site and solvent environment in enzymes with buried active site (Kokkonen et al., 2019). Using MOLEonline serve (L. Pravda et al., 2018), two access tunnels were found in P450 Bsβ HI (Fig. 1A, 1B and Table 1). The tunnels showed typical cytochrome features. The F and G helix de ne the most common tunnels among cytochrome P450s (Cojocaru et al., 2007). The peripheral exible F-G loop can control the tunnel topology then in uence substrate recognition. Representative tunnel in P450 Bsβ HI is Tunnel 1, which goes through the A helix, B' helix, B-B' loop B'-C loop, and F-G loop. Tunnel 1 consists of a large number of non-polar residues, which facilitates the access of hydrophobic substrates (Table S1). Tunnel 2 threads through B-B' loop and B'-C loop, which is also common among P450s. Compared with Tunnel 1, Tunnel 2 is shows higher polarity and contains a hydrophilic area near the surface of P450 Bsβ HI ( Fig. 1C and Table 1). In addition to participating in substrate transport, this tunnel may be also involved the controlling water diffusion. In order to nd potentially important residues, we set four criterions: 1.) Located in the tunnel bottleneck area; 2) being high in uential among the entire tunnel; 3) Located in the loop area; 4) being not completely conserved in its homologous cytochromes. Caver Analyst 2.0 software was chosen for analyzing the residues in uence, which provide an opportunity to explore the portion of the tunnel in uenced by a particular amino acid (Jurcik et al., 2018). We analyzed the tunnel properties and bottleneck residues in uence over 5 ns (Fig. 2). The results indicate that residues of F173 and V74 have high impact on Tunnel 1, and H85 and F79 are the main bottleneck residues in access Tunnel 2. Among high impact residues, H85 is the core residues for P450 Bsβ HI decarboxylation activity (Xu et al., 2017), V74 is located in the middle of B' helix, while the two phenylalanine, F79 and F173, are non-conservative ( Figure S1) and located in the B'-C loop and F-G loop, respectively. Given that, F79 and F173 residues were chosen for further analysis to optimize the enzymatic property while avoid ruining tunnel architecture.
Theoretically, the transport ability of substrates between enzyme active site and solvent environment could be adjusted by gates located at the access tunnels (Gora et al., 2013). Aromatic amino acids in access tunnels bottleneck are often participated in the control of the putative gates (Pavlova et al., 2009). As show in Fig. 2, the huge aromatic side chains of F79 and F173 were involved in the tunnel bottleneck regions. Here, the two phenylalanine residues F79 and F173 are presumed to be the "gatekeeper", i.e.
control the access of substrate from solvent environment to P450 Bsß HI active site. In addition, the F79 and F173 are also deemed to stabilize the fatty acid substrates via hydrophobic interactions (Lee et al., 2003). Based on the critical location of the two residues, it is assumed the large phenyl side chain of phenylalanine maintains substrate stability, but hinders the entry of long-chain fatty acid substrates. A more exible substrate entrance may bene t to increase enzyme activity towards long-chain fatty acids substrate. To test the hypothesis, the F79 and F173 residues were replaced by relatively small amino acid including cysteine, isoleucine, leucine, alanine, glycine valine, serine, threonine and proline, in mutagenesis experiments, respectively. The variants stability was evaluated by analysis of the relative free energy of folding (ΔΔG fold ) (Cui et al., 2020) and no unstable substitution was found ( Figure S2).

Decarbonylation activity of P450 Bsβ HI variants
As a hydrogen peroxide-dependent decarboxylase, P450 Bsβ HI relies on H 2 O 2 as an electrons provider and oxygen supplier (Xu et al., 2017). To select promising variants of P450 Bsβ HI, we examined the enzyme activity with the substrate of lauric acid using H 2 O 2 as the sole co-factor. Decarboxylation e ciency was characterized by α-undecene yield. As shown in Fig. 3, 9 of 18 single-site variants showed improved yield of α-undecene. Particularly, BsβHI-F79A, BsbHI-F79S, BsβHI-F79T, BsβHI-F79V, and BsβHI-F173V exhibited more than 1.5 fold increase of α-undecene yield compared to P450 Bsβ HI. The optimal variant BsβHI-F79T and BsβHI-F79V showed the 3.1-fold improvement of α-undecene yield. Then we focused on the alterations of identi ed tunnels in bene cial variants using MOLEonline serve (Table S3). The trends of decreased tunnel length and increased bottleneck radius were observed in the putative substrate tunnels (Tunnel 1) among variants, which is consistent with our previous assumptions. F79A variant showed the shortest substrate tunnel and the widest bottleneck radius, which were 2.8 Å decrease and 0.4 Å improvement, respectively. Interestingly, for both F79 or F173, we observed that the best decarboxylation performances were achieved by the valine substitution. The obvious feature of those two variants is the wider Tunnel 2. It is speculated that Tunnel 2 also undertakes the tunnel transfer task although this tunnel prefer polar water. Compared with Tunnel 1, Tunnel 2 has a longer length and a narrower bottleneck, which indicates that the Tunnel 2 may only serve as an auxiliary transportation for substrate. In short, the results indicated that the replacement of phenylalanine residue in positions 79 and 173 to relatively small amino acids (less bulky) had signi cant effect on the tunnels property and enzyme activity of P450 Bsβ HI.
To further investigate the in uence of substitution combination on the catalytic performance of P450 Bsβ HI, we recombined the bene cial substitutions. Therefore, a series of double-site variants were constructed, which were BsβHI-F79A/F173V, BsβHI-F79S/F173V, BsβHI-F79T/F173V, and BsβHI-F79V/F173V. However, all of the double-site variants show very low α-undecene production. One possibility is that the fatty acid substrates locate in the binding pocket lost the su cient stability in the double site variants. The original F79 and F173 residues stabilize the substrate by hydrophobic interaction, so that substrate is structurally complementary to the P450 Bsβ HI active site. Although the lauric acid substrate is more accessible to the binding pocket after double-site variants, the lack of strong hydrophobic interaction with F79 and F173 residues resulted in exible and instable binding of substrate around heme. Hence, the single-site variants of BsβHI-F79A, BsbHI-F79S, BsβHI-F79T, BsβHI-F79V, and BsβHI-F173V were chosen for further test.

Investigation of substrate preference in P450 Bsβ HI variants
Four additional fatty acids (capric acid, myristic acid, pentadecanoic acid, and palmitic acid) were investigated to probe the substrate pro le of P450 Bsβ HI variants and explain whether the decarboxylation activity of P450 Bsβ HI variants are improved towards other medium-and long-chain fatty acids. In addition, given that the important industrial application of styrene, phenylpropionic acid was also added for testing. As shown in Fig. 4, the decarboxylation activities of BsβHI-F79T, BsβHI-F79V, and BsβHI-F173V are higher than P450 Bsβ HI with the capric acid as substrate. While toward long-chain fatty acids substrate, as a general trend, all candidates of P450 Bsβ HI variants exhibit higher catalytic e ciency than P450 Bsβ HI. Intriguingly, optimal decarboxylation e ciency towards different substrates is exist in different variants. BsβHI-F173V possesses the highest activity towards myristic acid and pentadecanoic acid, while BsβHI-F79A is the best to palmitic acid. Both the two variants possessed broader access tunnels than the tunnels in P450 Bsβ HI (Table S3). We reasoned that this observation may stem from the compatibility between the substrate and access tunnels. Alanine has the smallest side chain, so it provided the F79A variant the widest substrate tunnel, which is able to accommodate relatively larger fatty acid substrate, like palmitic acid. Meanwhile, hydrophobic interaction provided by nonpolar alanine or valine also reduces substrate exibility to fatty acids substrate, leading to higher stability of enzymesubstrate complex during the reaction cycle. In addition, only trace styrene was detected when used the phenylpropionic acid as substrate, probably due to the strong preference of P450 Bsβ HI towards fatty acid substrates (Xu et al., 2017).

Conclusions
Overall, an access tunnel engineering was carried out to understand the substrate preference as well as for improving the decarboxylation activity and of P450 Bsβ HI. The structure indicates that two residues (F79 and F173) locate in the bottleneck of tunnels. A series of variants of P450 Bsβ HI were generated and investigated. Signi cantly improved decarboxylation activity was observed in BsβHI-F79A and BsβHI-F173V variants towards long-chain fatty acids. The results reveal that the appropriate reduction of the amino acid size at the gate of tunnels improves the enzymatic activity towards larger substrates, like long-chain fatty acids. Furthermore, our study shows that identifying and engineering key residues lining the access tunnels may be a valuable and e ciency strategy for improving the performance of enzymes with buried active sites.

Declarations
Ethics approval and consent to participate Not applicable

Consent for publication
Not applicable Availability of data and materials All data generated or analyzed during this study are included in this article and the supplementary information le.

Competing interests
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper.   Evolution of the activity of P450BsβHI variants towards the decarboxylation of lauric acid. The αundecene products were qualitatively determined by contrasting with standard sample, and quanti ed by external standard method. Results shown are mean ± SD of duplicated experiments.

Figure 4
Decarboxylation performance of P450BsβHI and its variants towards various organic acid. Capric acid, myristic acid, pentadecanoic acid, palmitic acid, and phenylpropionic acid were used as substrate. All of the α-alkene products were qualitatively determined by contrasting with standard sample, and quanti ed by external standard method. Results shown are mean ± SD of duplicated experiments.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.