Improving the autotransporter‐based surface display of enzymes in Pseudomonas putida KT2440

Summary Pseudomonas putida can be used as a host for the autotransporter‐mediated surface display of enzymes (autodisplay), resulting in whole‐cell biocatalysts with recombinant functionalities on their cell envelope. The efficiency of autotransporter‐mediated secretion depends on the N‐terminal signal peptide as well as on the C‐terminal translocator domain of autotransporter fusion proteins. We set out to optimize autodisplay for P. putida as the host bacterium by comparing different signal peptides and translocator domains for the surface display of an esterase. The translocator domain did not have a considerable effect on the activity of the whole‐cell catalysts. In contrast, by using the signal peptide of the P. putida outer membrane protein OprF, the activity was more than 12‐fold enhanced to 638 mU ml−1 OD −1 compared with the signal peptide of V. cholerae CtxB (52 mU ml−1 OD −1). This positive effect was confirmed with a β‐glucosidase as a second example enzyme. Here, cells expressing the protein with N‐terminal OprF signal peptide showed more than fourfold higher β‐glucosidase activity (181 mU ml−1 OD −1) than with the CtxB signal peptide (42 mU ml−1 OD −1). SDS‐PAGE and flow cytometry analyses indicated that the increased activities correlated with an increased amount of recombinant protein in the outer membrane and a higher number of enzymes detectable on the cell surface.


Introduction
In the last years, Pseudomonas putida has been intensively studied with regard to its biotechnological application. This is due to some of the bacterium's beneficial properties, such as metabolic versatility, environmental stress resistance and low cultivation demands, making it particularly valuable for industrial biocatalysis (Poblete-Castro et al., 2012). The best characterized strain is P. putida KT2440 (Belda et al., 2016), which has been developed towards industrial applications and for which sufficient tools for heterologous gene expression and genetic modifications are available (Lieder et al., 2015;Loeschcke and Thies, 2015).
For whole-cell biocatalytic approaches, it is under certain conditions expedient not to express the desired enzyme intracellularly or to secrete, but instead to display it on the surface of the bacterium, i.e. to transport the enzyme to the outer membrane, rendering it accessible from the extracellular space while maintaining a stable linkage to the cell. This is specifically the method of choice when non-cell permeable substrates or products come into play or when enzymes require a membrane surrounding (Schuurmann et al., 2014). Surface display in Gram-negative bacteria can be realized by means of the monomeric autotransporter (type Va) secretion pathway (Henderson et al., 2004). The utilization of autotransporters for the display of recombinant enzymes (or other proteins) is termed autodisplay (Jose, 2006;Jose and Meyer, 2007;Jose et al., 2012).
In earlier studies, we established Pseudomonas putida as a host for autodisplay of cellulases and hemicellulases in the context of whole-cell catalytic biomass degradation (Tozakidis et al., 2016;Schulte et al., 2017). For this purpose, maximized autotransporter expression (MATE)-plasmids were used, which encode for fusion proteins consisting of the N-terminal CtxB signal peptide (SP) from Vibrio cholerae, followed by the passenger (that is the enzyme to be displayed) and the translocator domain of the EhaA autotransporter from E. coli. The translocator domain can be divided into b1-domain, a-helix (together termed linker) and b-barrel domain. Besides E. coli and P. putida, MATE plasmids were shown to also be compatible for autodisplay in the ethanologenic bacteria Zymomonas mobilis and Zymobacter palmae and our experience with several other hosts indicates universal applicability of MATE (Tozakidis et al., 2014;Sichwart et al., 2015). However, although the term autotransporter suggests an autonomous secretion, it is meanwhile known that several cellular entities participate in the folding of the protein into the outer membrane and that they include species-specific interactions (Robert et al., 2006). It was also found that expression and display levels of heterologous autotransporters correlate with the phylogenetic distance between host and donor species (Marin et al., 2010). A previous study has also identified the linker region as an important influence factor on autodisplay (Quehl et al., 2017). Besides this, it is obvious that the N-terminal SP also determines the amount of displayed protein as it mediates its translocation into the periplasm. It therefore seems promising to evaluate different homologues of these domains in order to optimize autodisplay.
Following this idea, we compared different translocator domains, namely EhaA from E. coli (Sichwart et al., 2015) EstP from P. putida (Asler et al., 2010) and a combination of EstP a-helix and EhaA b-barrel, in terms of their influence on the surface display of an esterase as an example enzyme. This influence, however, turned out to be negligible. Moreover, different signal peptides were tested, i.e. the V. cholerae CtxB SP (Maurer et al., 1997), the P. aeruginosa EstA SP (Wilhelm et al., 2007) as well as the native P. putida OprF SP (Kragelund et al., 1996). The last one had a strong positive effect on the surface display of an esterase as well as a bglucosidase with 12.4-and 4.3-fold increases in activity respectively.

Surface display of esterase with different translocator domains
To investigate the influence of different translocator domains on autodisplay of enzymes in P. putida, plasmids were constructed that encode for autotransporter fusion proteins consisting (from N-to C-terminus) of the V. cholerae CtxB SP, the esterase EstA from B. gladioli as an example passenger (Schultheiss et al., 2002) with a 6xHis epitope at its N-terminus and three different translocator domains (Fig. 1A): (i) the translocator domain of the EhaA autotransporter from E. coli as described previously (Sichwart et al., 2015) and utilized so far for surface display in P. putida (Tozakidis et al., 2016;Schulte et al., 2017). This protein (and the encoding plasmid) was named MATE-EstA according to the previous terminology. (ii) A translocator in which the a-helix and b-barrel of EhaA were replaced by the corresponding domains from the EstP autotransporter from P. putida KT2440 (Asler et al., 2010). To define these domains, the tertiary structure of EstP was modelled via RaptorX (Kallberg et al., 2012), and the a-helix and b-barrel relevant for autotransporter secretion could be identified ( Figure 1B). The resulting protein was termed MATE2-EstA. (iii) A translocator that combines the a-helix of EhaA and the b-barrel domain of EstP (termed MATE3-EstA). The b1-domain derived from E. coli EhaA remained constant in all constructs.
The strains as obtained after transformation, P. putida pMATE-EstA, pMATE2-EstA and pMATE3-EstA, were cultivated and the encoded fusion proteins were expressed under the control of the araBAD promotor. Strains without induction of protein expression were used as controls. To determine the expression of surface displayed esterase, the cultures were harvested and washed, and the cells were incubated with anti-6xHis antibodies. After washing to get rid of unbound antibodies, cells were treated with secondary DyLight 633-coupled antibodies, washed again and subjected to flow cytometer analysis. As the antibodies are too large to penetrate the membrane, they can only interact with their epitope in case it is accessible from the outside. Fluorescence intensity of the cells should therefore be proportional to the number of enzymes displayed on their surface. As seen in Fig. 1C-E, all strains expressing one of the autotransporter fusion proteins showed enhanced fluorescence compared to the control, with similar mean fluorescence (mf) values (MATE-EstA: 1591; MATE2-EstA: 1307; MATE3-EstA: 1217), indicating that they contained correctly folded autotransporters and displayed EstA on their surface.
To evaluate the influence of the different translocator domains on the activity of P. putida whole-cell catalysts, the strains were cultivated, protein expression was induced, and cell suspensions with an optical density at 578 nm (OD) of 0.2 were incubated with 4-nitrophenyl acetate at a temperature of 30°C to photometrically monitor the formation of the reaction product 4-nitrophenolate ( Fig. 1F and G). Surface display of the esterase by means of MATE2-EstA resulted in 44% higher activity (101 mU ml À1 OD À1 ) compared with MATE-EstA (70 mU ml À1 OD À1 ). The fusion of EhaA a-helix and EstP b-barrel (MATE3-EstA) had no significant positive effect on the activity (79 mU ml À1 OD À1 ).
Considering these findings, it is either conceivable that the compatibility of the EhaA autotransporter with P. putida periplasmic chaperones is such high that the secretion cannot be optimized significantly at this point or that the recognition of the translocator domain is not the limiting step of the secretion mechanism. In the latter case, the transport of the autotransporter fusion protein into the periplasm and consequently the recognition of the N-terminal signal peptide should represent the bottleneck of surface display. We therefore shifted our focus on substituting the signal peptide in order to optimize autodisplay for P. putida.

Surface display of esterase with different signal peptides
Plasmids were constructed that encode for autotransporter fusion proteins with different signal peptides: (i) the routinely used CtxB SP from V. cholerae (Jose and Meyer, 2007), i.e. the protein MATE-EstA; (ii) the SP of the outer membrane protein OprF from P. putida (Kragelund et al., 1996), resulting in OprF SP -MATE-EstA; OprF is a dominant outer membrane protein of P. putida KT2440 that plays a similar role as OmpA in E. coli (Wang, 2002); and (iii) the SP of EstA from Pseudomonas aeruginosa, resulting in the protein EstA SP -MATE-EstA. In P. aeruginosa, this SP is involved in the surface display of a virulence factor (Wilhelm et al., 2007). The following C-terminal parts of the fusion proteins remained as described before, including the EhaA translocator. The corresponding domain structures are shown in Fig. 2A, the amino acid sequences of the signal peptides are depicted in Fig. S2.
After transforming P. putida with these plasmids, cultivating the resulting strains and inducing protein expression, outer membrane proteins were isolated and separated via SDS-PAGE (Fig. 2B). In all samples except the control (P. putida with empty vector), a protein band at an apparent molecular weight of approximately 100 kDa was visible. This band was assigned to the recombinant autotransporter fusion proteins with calculated molecular weights between 91 and 93 kDa after release of the SP in all of the three samples. This protein band was most pronounced in samples of P. putida pOprF SP -MATE-EstA, whereas the faintest band appeared in samples of P. putida pMATE-EstA, giving a first hint on the different translocation efficiencies of the signal peptides. Alongside the CtxB SP, both the OprF SP and the EstA SP were therefore functional in terms of translocating the protein into the periplasm.
Next, the strains were analysed via flow cytometry as described above (Fig. 2C-E). All three of them exhibited enhanced fluorescence intensities compared with the controls, indicating the surface localization of EstA. Consistently with the observed protein band intensities in the SDS-PAGE, P. putida pOprF SP -MATE-EstA showed the highest mf of 2064. Contrary to the faint band in the SDS-PAGE, P. putida pMATE-EstA had a higher mf (1591) than P. putida pEstA SP -MATE-EstA (1035). At this point, it is unclear whether the secretion process itself, the subsequent labelling with antibodies or both factors are responsible for the observed difference between protein amount in the outer membrane and detectable EstA on the cell surface.
Whole-cell esterase activity tests of each strain were performed as described above and revealed that replacing the CtxB SP had much stronger influence on the activity than replacing the EhaA translocator ( Fig. 2F and G). Cells expressing the fusion protein with the OprF SP possessed a 12.4-fold higher esterase activity (638 mU ml OD À1 ) than cells with the CtxB SP (52 mU ml OD À1 ). The activity of cells with the EstA SP was between these values with 152 mU ml À1 OD À1 .
To assess whether these effects originated from a codon usage bias, the codon adaption index (CAI) of the CtxB-and the OprF-sequences were calculated using the online tool JCat (Grote et al., 2005). The CAI value of the CtxB sequence was significantly higher (0.54) than that of the OprF sequence (0.35), indicating that the higher activities of P. putida pOprF SP -MATE-EstA cells were not an effect of codon bias. Additionally, RBS calculator (Salis, 2011) was used to predict the translation rates of the described constructs, which can be influenced by secondary structures of the gene transcripts. The translation rate for pMATE-EstA was more than twice as high as for pOprF SP -MATE-EstA (16 542 vs. 7628). Hence, this parameter could also not account for the observed differences in whole-cell activities. It can thus be assumed that the signal peptides have different secretion efficiencies in P. putida.
It was observed that all strains expressing the autotransporter fusion proteins exhibited an increased sensitivity towards antibiotics, which could be an effect of the increased number of b-barrel proteins within the outer membrane (Fig. S4). Also, the viability of the strains decreased significantly by a factor of 1000 in terms of colony-forming units. This could reflect the burden of protein expression or an impairment of the cell's outer membranes by the b-barrel proteins (Fig. S5).

Surface display of b-glucosidase with different signal peptides
In order to confirm the beneficial effect of the OprF SP on the surface expression with autodisplay, as found with esterase, we investigated a second passenger protein, the b-glucosidase BglA from Caldicellulosiruptor saccharolyticus (Hong et al., 2009). To this end, the plasmids pMATE-BglA and pOprF SP -MATE-BglA were constructed, encoding for the autotransporter fusion proteins as described before, with either CtxB or OprF SP at the N-terminus, followed by BglA as the passenger domain and the EhaA translocator. As before, the proteins included a 6xHis epitope for immunological detection. After cultivation of the strains and induction of protein expression, outer membrane proteins were isolated and separated by SDS-PAGE (Fig. 3A). A protein band at an apparent molecular weight between 100 and 130 kDa was detected in samples of P. putida pOprF SP -MATE-BglA and, to a much lesser extent, in samples of P. putida pMATE-BglA. It could be assigned to the recombinant fusion proteins with a calculated theoretical molecular mass between 111 and 112 kDa after release of the SP during transport across the inner membrane. With both signal peptides, surface display could be demonstrated in flow cytometry experiments  Fig. 3B and C). In addition, b-glucosidase activity of these cells using 4-nitrophenyl-b-D-glucopyranoside as substrate was 4.3-fold higher (181 mU ml À1 OD À1 ) than the activity of cells with CtxB SP (42 mU ml À1 OD À1 ) ( Fig. 3D and E).
Hence, replacing the CtxB SP by the OprF SP had a similar impact on autodisplay of the b-glucosidase as on autodisplay of the esterase, although the effect seems to be biased by the passenger domain.

Conclusion
These findings support the hypothesis that the transport of fusion proteins across the inner membrane of the bacterium could represent a bottleneck when using MATE for autodisplay in P. putida (Tozakidis et al., 2016;Schulte et al., 2017). Transport efficiency appears severely affected by the SP, but not by the translocator domains used. This study helps optimizing P. putida whole-cell biocatalysts with surface displayed enzymes and identified the OprF signal peptide as well suited for autodisplay in this host.

Construction of plasmids
All constructed plasmids are based on pMATE-EstA (Tozakidis et al., 2014). For the construction of pMATE2-EstA, the sequence encoding for the a-helix and b-barrel of EstP (Asler et al., 2010) was amplified from the chromosomal DNA of P. putida KT2440. The sequence was inserted into pMATE-EstA instead of the EhaA a-helix and b-barrel via In-Fusion â Cloning as described by the manufacturer. For the construction of pMATE3-EstA, the sequence encoding for the b-barrel of EstP was amplified and inserted into pMATE-EstA instead of the EhaA b-barrel. For the construction of pOprF SP -MATE-EstA, the DNA sequence of the signal peptide (SP) of OprF (Kragelund et al., 1996) was amplified from the chromosomal DNA of P. putida KT2440. The sequence of the original CtxB SP of MATE-EstA was substituted by the OprF SP via In-Fusion â Cloning. For the construction of pEstA SP -MATE-EstA, the CtxB SP sequence was substituted by the SP of P. aeruginosa EstA (Wilhelm et al., 2007). The DNA sequence of the EstA SP was synthesized commercially and inserted into MATE-EstA via In-Fusion â Cloning. For the construction of pMATE-BglA and pOprF SP -MATE-BglA, the DNA sequence of C. saccharolyticus BglA (NCBI GenBank: P10482) was inserted into the plasmids pMATE-EstA and pOprF SP -MATE-EstA via restriction/ligation cloning with XhoI and KpnI. To enable the translation of full-length MATE-BglA and OprF SP -MATE-BglA, the stop codon of C. saccharolyticus BglA had to be removed beforehand. All cloning primers are listed in Table S1. The sequences of all constructed plasmids were verified by Sanger sequencing (Microsynth Seqlab, G€ ottingen, Germany).

Flow cytometry analyses
For flow cytometry analyses, the cells were cultivated in LB medium to an OD of 0.5 at 30°C and 200 rpm. Protein expression was induced by supplementation of 0.2% of L-(+)-Arabinose for 2 h. The cells were then harvested and diluted with PBS to an OD of 0.4. After three rounds of centrifugation and washing with particle-free PBS (12 000 g, 1 min), the cells were suspended in 100 ll primary antibody solution (THE TM His tag Antibody, mouse, 1:100 in PBS; GenScript, Piscataway, NJ, USA) and incubated for 45 min at room temperature (RT). The cells were again centrifuged and washed three times with particle-free PBS (12 000 g, 2 min) and suspended in 100 ll of a secondary antiserum in PBS (antimouse IgG (H+L) Secondary Antibody, DyLight 633coupled, goat, 1:100 in PBS; Thermo Fisher Scientific, Waltham, MA, USA) and incubated for 60 min at RT in the dark. The cells were centrifuged and washed three times with particle-free PBS (12 000 g, 2 9 2 min, 2 9 1 min) and each pellet was suspended in 500 ll particle-free PBS. The fluorescence of the cells was analysed with a FACSAria III (Becton Dickinson, Heidelberg, Germany). 50000 events of every sample were analysed. The DyLight 633 fluorescence was excited at 633 nm and analysed at 660/20 nm (band-pass filter, 537 V).

Photometric activity assays
For esterase activity determination, the cells were sedimented, washed with 50 mM sodium phosphate buffer pH 7 and suspended in the same buffer to an OD of 1. 40 ll of this suspension was mixed with 160 ll of a 4nitrophenyl acetate solution (4 mM in 50 mM sodium phosphate buffer) in a microtiter plate and incubated at 30°C under shaking. At given time points, the absorption at 405 nm was measured with a microtiter plate reader. For the b-glucosidase activity assay, the cells were washed and suspended in 100 mM sodium citrate buffer pH 6 to an OD of 1, mixed in a 1:1 ratio with a 4-nitrophenyl-b-D-glucopyranoside solution (5 mM in 100 mM sodium citrate buffer) and incubated at 55°C under shaking. At given time points, the reaction samples were centrifuged, 50 ll of the supernatant was mixed with 50 ll of a 2 M Na 2 CO 3 solution and absorption was detected at 405 nm. In both assays, the concentration of the generated 4-nitrophenol was calculated by means of 4-nitrophenol calibration curves.

Outer membrane protein isolation and separation
For the isolation of membrane proteins, adapted protocols of Lee et al. (2005) and Park et al. (2015) were used. 100 ml of cell suspensions were cultivated to an OD of 0.5 and induced with 0.2% of L-(+)-Arabinose. Protein expression was induced for 2 h, and the cells were harvested (5000 g, 5 min, 4°C). The cell pellets were suspended in 10 ml of Tris-HCl buffer (10 mM, pH 7.5) and centrifuged (5000 g, 5 min, 4°C). The pellets were suspended in 5 ml Tris-HCl buffer (10 mM, pH 7.5) with 2 mM PMSF, 10 lg ml À1 aprotinin and 4 lg ml À1 pepstatin. Crude extracts of P. putida cells were prepared by three cycles of sonication (40 s, 25% of maximum output). Each crude extract was supplemented with 0.05 mg ml À1 DNase I and incubated for 20 min at 37°C. 2 ml of each crude extract was centrifuged (8000 g, 7 min, 4°C) to remove partially disrupted cells. The supernatants were centrifuged (20 000 g, 30 min, 4°C) to isolate membrane proteins and lipid layers. The pellets were suspended in Tris-HCl buffer (10 mM, pH 7.5) and centrifuged again (20 000 g, 30 min, 4°C). This step was repeated. The supernatant was then removed, and the membrane proteins suspended in Millipore water. The suspensions were supplemented with SDS sample buffer (containing 30 mg ml À1 dithiothreitol), incubated for 30 min at 95°C and separated via sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) with PageRuler Unstained Protein Ladder as size standard. The outer membrane protein OprF with a molecular weight of 35 kDa was used as a loading control (Hancock et al., 1990). For the analysis, 12.5% acrylamide gels were used and afterwards stained in ProBlueSafe Stain (GiottoBiotech, Sesto Fiorentino, Italy) and de-stained with water.