Bacterial Outer Membrane Vesicles as Nano‐Scale Bioreactors: A Fatty Acid Conversion Case Study

Bacterial outer membrane vesicles (OMVs) are small unilamellar proteoliposomes, involved in various functions including cell‐to‐cell signalling and protein excretion. We have engineered the OMVs of Escherichia coli to nano‐scaled bioreactors for the biotransformation of fatty acids by targeting a fatty acid double bond hydratase of Stentrophomonas maltophilia (SmOhyA) and/or a photoactivated fatty acid decarboxylase from Chlorella variabilis NC64 A (CvFAP) into OMVs. Engineered OMVs containing both SmOhyA and CvFAP were able to catalyse the transformation of oleic acid ((Z)‐octadec‐9‐enoic acid) into 9‐hydroxyheptadecane via (R)‐10‐hydroxyoctadecanoic acid. The specific biotransformation rates of oleic acid reached 8.0×10−12 μmol/min per OMV.


Introduction
Lipid-based micellar and liposomal nanoparticles have been studied intensively as carriers for proteins, genes and other biomolecules. Liposomes, for example, have been used as model systems for plasma membranes, for drug delivery or imaging purposes. [1] Liposomes have also been explored as nano-scaled reaction vessels, [2] their small diameters between 50 and 250 nm allow for rapid diffusional mixing thereby enabling fast reaction kinetics. [3] Next to liposomes, bacterial outer membrane vesicles (OMVs), spherical, bilayered proteolipids with an average diameter of 20 to 200 nm are gaining increased interest. [4] OMVs can contain both, periplasmic and outer membrane proteins and therefore have been proposed to play a role in the delivery of toxins to cells and in the transfer of proteins between bacterial cells. [5] Bacterial OMVs have also been investigated as vaccine carriers, because they are structurally stable and may protect the encapsulated proteins and biomolecules from a variety of stresses. [6] In addition, bacterial OMVs have been reported as nano-scaled bioreactors. For example, an OMV loaded with an organophosphate hydrolase from Brevundimonas diminuta was used for the remediation of organophosphate pesticides. [7] In another example, cellulases encapsulated in OMVs enabled the 23-fold accelerated hydrolysis of cellulose as compared to the non-complexed enzymes. [8] Preparation of synthetic liposomes can be a tedious task including not only the heterologous expression of the biocatalysts and their purification but also the liposome preparation comprises various synthesis and purification steps. [9] In contrast, bacterial OMVs can readily be produced by a single synthesis and purification step. Furthermore, the composition of the vesicles can be engineered: [6a,10] protein engineering and synthetic biology tools enable to construct diverse protein-OMV complexes. [7][8]11] Overall, OMVs represent a versatile nano-scaled bioreactor platform and we became interested in evaluating OVMs for the conversion of fatty acids.
Fatty acids, one of the most abundant renewable resources in nature, serve as starting materials in the synthesis of biofuels and oleochemicals but also for the synthesis of signalling molecules and bioactive metabolites. [12] For instance, unsaturated fatty acids have been converted into hydroxy fatty acids by fatty acid double bond hydratases [12c,13] or into hydrocarbons through decarboxylation of the terminal carboxyl group by a photoactivated decarboxylase from Chlorella variabilis NC64A (CvFAP) (Scheme 1). [14] Moreover, functionalisation of fatty acids [15] sets the stage for the preparation of further products such as ω-amino fatty acids and α,ω-dicarboxylic acids. [15a,c, 16] The majority of the aforementioned reactions relies on wholecell biocatalysis. The major limitations here reside with the poor solubility and dispersibility of fatty acids in aqueous reaction media and with transport issues into the living cell. [17] In the present study, we have engineered E. coli OMVs to form nano-scale bioreactors for the biotransformation of unsaturated long chain fatty acids. The biocatalysts used were the fatty acid double bond hydratase of Stentrophomonas maltophilia (SmOhyA) [13a,b] and CvFAP. [14] Engineered OMVs, encapsulating the SmOhyA and/or CvFAP, were investigated for the biotransformation of oleic acid ((Z)-octadec-9-enoic acid, 1) into (R)-10-hydroxyoctadecanoic acid (2) and 9-hydroxyheptadecane (3) (Scheme 1). Also, the direct decarboxylation of oleic acid yielding (Z)-heptadec-8-ene (4) was investigated.

Preparation and characterisation of the engineered OMVs
Hyper-vesiculating E. coli BL21(DE3) mutants were constructed by deletion of tolA or tolR (as parts of the Tol-Pal gram-negative bacteria [18] ). To target SmOhyA into the periplasm of the E. coli cells its gene sequence was fused with the signal sequence of pectate lyase B from Erwinia carotovora (PelBSS) [19] obtaining the plasmid pACYC-PelBSS-SmOhyA. Using this construct in wild type E. coli (BL21(DE3)) and its tolA and tolR deficient derivates expression of SmOhyA was achieved. The resulting OMVs were isolated from the cultivation broths via ultracentrifugation. As shown in Figure 1, the OMV concentrations in the ΔtolA and ΔtolR broths were 1.6 × 10 9 mL À 1 and 2.5 × 10 9 mL À 1 , respectively and thereby 13-and 21-fold higher than from wt E. coli.
The morphological properties of the engineered OMVs were investigated by transmission electron microscopy (TEM) and dynamic light scattering (DLS). Although the number of OMVs produced per recombinant cells was significantly different, the morphology of the OMVs appeared to be similar ( Figure S1). The OMVs were mostly spherical and ranged in diameter between 40 and 50 nm ( Figure S1B). This is in accordance with previous findings with OMVs isolated from E. coli DH5α. [20] The protein composition of the engineered OMVs was examined by SDS-PAGE and Western-Blot analysis ( Figure S2). Next to the typical OMV protein OmpF [11b,21] also the heterologous protein SmOhyA was found as a major protein in the OMVs ( Figure S2). Notably, the overall concentration of SmOhyA linearly correlated with the number of OMVs irrespective of which expression system was used ( Figure S3), indicating that the SmOhyA expression level did not vary significantly between the three strains.
Compared to the free enzyme, the OMV-confined SmOhyA exhibited a significantly higher apparent K M value indicating some diffusion limitation of the reagents over the OMV membrane.

Fatty acid biotransformation activity of the OMVs containing SmOhyA
To evaluate the fatty acid hydration activity of the SmOhyAencapsulating OMVs, the biotransformation of oleic acid (1) (Scheme 1) was used as model reaction.
The initial reaction rates were 0.6 mM h À 1 , 9.6 mM h À 1 and 10.8 mM h À 1 (10, 160 and 180 U L À 1 ) and thereby linearly correlated with the OMV concentration in the reaction mixture ( Figure 2d).
Similar reaction profiles were also observed with ricinoleic acid as starting material ( Figures S5 and S6).
SmOhyA activity was almost not detectable in the supernatant of the biotransformations indicating that surfactants (such as 0.1 % (v/v) Tween 80) did not significantly affect structural stability of the OMVs. It is interesting to note that the thermal stability of the OMV-confined SmOhyA was somewhat higher than of the purified enzyme. While the purified enzyme lost approx. 90 % of its activity already after 4 h incubation at 20°C, the OMV-confined enzyme exhibited more than 40 % of residual activity ( Figure S8A). Similar observations were made incubating the enzyme at 4°C ( Figure S8B).
Again, the initial rates linearly correlated with the concentration of the OMVs applied.
The two-step one-pot biotransformation (Scheme 1) was initiated by adding 3.5 mM oleic acid to the isolated OMVs (in 50 mM Tris buffer (pH 8) containing 0.1 % (v/v) Tween 80) under dark conditions. After 3 h, the second reaction was initiated by exposing the reaction setup to blue light for another 2 h (see the Experimental section for details). The first reaction proceeded to approx. 71 % conversion of oleic acid into (R)-10hydroxyoctadecanoic acid (Figure 4). After 3 h, illumination was initiated which resulted in the smooth decarboxylation of both, remaining oleic acid and the intermediate (R)-10-hydroxyoctadecanoic acid ( Figure S13).
Apparently, prolonging the dark, hydratation phase will increase the overall selectivity from oleic acid (1) to 9hydroxyheptadecane and diminish accumulation of the undesired (Z)-heptadec-8-ene (originating from the decarboxylation of non-converted oleic acid).
Finally, we compared the catalytic performance of the OMVconfined biocatalysts with the activity of the enzymes expressed in non-hyper vesiculating E. coli cells (Table 1). For this, we referred the product formation rate to the cell or the OMV as catalytic unit (micro/nano reactor). Interestingly, the specific catalytic activity of whole E. coli cell per catalytic unit (i. e., per cell or OMV, respectively) was only 5-6 times higher than that of OMVs. Considering the size difference between OMVs (d5 0 nm, V~0.5 × 10 À 3 μm 3 ) and E. coli (d~2 × 1 μm, V~0.6 μm 3 ) [23] the volumetric productivity within the catalytic unit was more than 100-fold higher in OMVs as compared to E. coli cells (Table 1). This indicates that the OMVs obtained in this study contained a highly concentrated enzyme mixture compared to the non-hypervesiculating E. coli cells.

Conclusion
Overall, the present study demonstrates the synthetic potential of OMVs as nano-scaled reactors for biocatalytic and multi-step biocatalytic reactions. Especially if the biocatalysts are specifically targeted to the OMVs, e. g., using the signal sequence PelBSS, OMVs containing highly concentrated biocatalysts can be obtained, which especially for multi-step cascade transformations is very promising.

Experimental Section
A complete description of the experimental and analytical conditions can be found in the supporting information.

Microbial strains and cultivation conditions
Recombinant E. coli was cultivated at 37°C in lysogeny broth medium (5 g/L yeast extract, 10 g/L tryptone, and 10 g/L NaCl) supplemented with appropriate antibiotics for plasmid DNA preparation (Table S2). Terrific broth (TB) medium (24 g/L yeast extract, 12 g/L tryptone, and 4 mL/L glycerol) with the appropriate antibiotics was used for the seed and main cultivation at 37°C. Recombinant SmOhyA expression was induced by adding 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) to the culture broth at the exponential growth phase (optical density at 600 nm: 0.5) followed by a 16 h incubation at 20°C. Coexpression of SmOhyA and CvFAP was induced by 0.03 mM IPTG, which allowed expression of the both enzymes to a soluble form in E. coli (see Fig.  S12 for details).

Construction of hyper-vesiculating E. coli mutants and recombinant plasmids
Hyper-vesiculating E. coli mutants (i. e., tolA and tolR gene knockout mutants) were constructed by the method described by Datsenko and Wanner. [24] The pET22b-PelBSS-CvFAP was constructed via amplification of the gene from pET28a-CvFAP [14f] and insertion into pET22b vector including the pelB signal sequence (Merck Millipore, Darmstadt, Germany) by In-Fusion cloning protocol (Takara Bio. Inc, CA, USA). The pACYC-PelBSS-CvFAP was constructed by using the pET22b-PelBSS-CvFAP. The plasmids, strains, primers, and protein sequences used in this study are summarised in Tables S1, S2, and S3.

Isolation of outer membrane vesicles (OMVs)
The culture broth of E. coli cells (150 mL) was centrifuged twice at 10,000 rpm for 10 min at 4°C, and filtered through a 0.2 μm membrane to remove intact bacteria and undesired large cellular materials and concentrated by Amicon® Stirred Cells (Millipore, Bedford, MA, USA), as previously reported. [7] The OMVs were then pelleted at 29,000 rpm (150,000 g) in a Sorvall WX Ultra 90 centrifuge using an AH-629 rotor (Thermo Fisher Scientific, Waltham, MA, USA) for 3 h at 4°C. The culture media was decanted and the OMV pellet was suspended in 1 mL of Tris buffer (50 mM, pH 8.0).

Biotransformation of fatty acids by OMVs and whole-cell biocatalysts
The hydration of oleic acid and ricinoleic acid by the isolated OMVs was conducted in the 50 mM Tris buffer (pH 8.0) at 35°C and 400 rpm (reaction volume: 2 mL). The hydration of fatty acids was initiated by adding 5 mM oleic acid or ricinoleic acid with 0.1 % (v/ v) Tween 80 into the reaction medium containing the OMVs. The fatty acids were added from stock solution, which had been made by sonication with Tween80. The biotransformations were terminated by mixing with two volumes of ethyl acetate containing 5 g/ L palmitic acid as internal standard.
The hydration of fatty acids by whole-cell biocatalysts was carried out based on our previous studies. [2,8] Briefly, after cultivation of the recombinant E. coli cells in TB medium (cell density: 3.6 g dry cells/ L), the cells were harvested and washed with 50 mM Tris buffer (pH 8.0). The resulting recombinant cells were resuspended in Tris buffer (pH 8.0). The biotransformations were initiated by adding 10 mM fatty acids with 0.1 % (v/v) Tween 80 into the reaction medium at 35°C and 200 rpm (reaction volume: 10 mL).  = 0.18 to 6.7 × 10 10 OMV mL À 1 , Tris buffer pH 8.0 (50 mM, with 10 % (v/v) DMSO), blue light (intensity = 13.7 mEL À 1 s À 1 ), total volume 2 mL.
The decarboxylation of whole-cell biocatalyst was carried out based on previous studies. [14f] Briefly, after cultivation of the recombinant E. coli cells in the TB medium (cell density: 3.6 g dry cells/L), the cells were harvested and washed with the 50 mM Tris buffer (pH 8.0). The biotransformation was initiated by 5 mM oleic acid and 0.1 % (v/v) Tween 80 into the reaction medium under blue light (λ = 450 nm). The reaction mixture was stirred at 37°C and 400 rpm. , blue light (intensity = 13.7 mEL À 1 s À 1 ), total volume 2 mL. The biotransformations were terminated by mixing with two volumes of ethyl acetate containing 5 g/L palmitic acid.