The bacterial strains used in this study are listed in Table 1. These bacterial strains were grown as described previously (Ishikawa et al., 2012). Acinetobacter sp. Tol 5 and its mutant strains were grown at 28°C and E. coli strains were grown at 37°C in lysogeny broth (LB) medium unless otherwise noted. The following antibiotics were used at the following concentrations when necessary: ampicillin (500 μg/mL) and gentamicin (10 μg/mL) for Tol 5 mutant strains and ampicillin (100 μg/mL), gentamicin (10 μg/mL) and kanamycin (50 μg/mL) for E. coli. For the expression of ataA and its mutants, an overnight culture of bacterial cells harboring the expression vector was inoculated into LB medium supplemented with ampicillin, gentamicin, and 0.5% (w/v) arabinose to induce the expression of genes under the control of the P_BAD promoter, and the medium was incubated for 7 h at 115 rpm. Growth curves were obtained by measuring the optical density at 600 nm (OD₆₀₀) of the culture medium.

The plasmids and primers used in this study and synthetic DNA fragments prepared in this study are listed in Table 2, Supplementary Tables S1, S2, respectively. The schematic procedures for the construction of ataA mutants in pDONR221 plasmids are shown in Supplementary Figures S1–S7. The details of the procedure are described below.

To construct pIFD-∆Nhead, a DNA fragment was amplified by inverse PCR (iPCR) from pDONR221:Ntemp using primers ∆Nhead-f/∆Nhead-r and self-ligated, generating pDONR221:Ntemp_∆Nhead. A DNA fragment excised from pDONR221:ataA with EcoRV was inserted into the same site of pDONR221:Ntemp_∆Nhead, generating pDONR221:∆Nhead.

To construct pIFD-∆NS-A1, pTAKN-2:FragTrp3,4 was digested with BsaI, and the resulting fragment was inserted into the same site of pTAKN-2:FragTrp5,6, generating pTAKN-2:FragTrp3-6. A DNA fragment excised from pMD19:FragC with BglII/BamHI was inserted into the BamHI site of pMD19:FragB, generating pMD19:FragBC. Then, a DNA fragment was amplified by iPCR from pDONR221:FragA2_031 using primers ∆NS-A1-f1/∆NS-A1-r1 and self-ligated, generating pDONR221:FragA2_054. A DNA fragment excised from pTAKN-2:FragTrp3-6 with BsaI/BamHI was inserted into the same site of pDONR221:FragA2_054, generating pDONR221:FragA_054. A DNA fragment excised from pMD19:FragBC with BglII/BamHI was inserted into the BamHI site of pDONR221:pFragA_054, generating pDONR221:∆NS-A1.

To construct pIFD-ΔNS-A2, a DNA fragment was amplified by PCR from pDONR221:ataA using primers ∆NS-A2-f1/∆NS-A2-r1 and self-ligated, generating pDONR221:∆Nstalk-a. A DNA fragment was amplified by PCR from pMD19:FragB using primers ∆NS-A2-f2/∆NS-A2-r2 and assembled with the longer DNA fragment of pMD19:FragB digested with BglII/PstI using In-Fusion HD Cloning Kit (Takara Bio, Shiga Japan), generating pMD19:FragB_09. A DNA fragment excised from pMD19:FragC with BglII/BamHI was inserted into the BamHI site of pMD19:FragB_09, generating pMD19:FragBC_09. Then, a DNA fragment excised from pMD19:FragBC_09 with BsaI/BamHI was inserted into the same site of pDONR221:∆Nstalk-a, generating pDONR221:∆NS-A2.

To construct pIFD-ΔNS-B, a DNA fragment was amplified by PCR from pDONR221:ataA using primers ∆NS-B-f1/∆NS-B-r1 and was self-ligated, generating pDONR221:∆Nstalk-b. A DNA fragment was amplified by PCR from pMD19:FragC using primers ∆NS-B-f2/∆NS-B-r2 and was assembled with the pTA2 vector (TArget Clone; TOYOBO, Osaka, Japan), generating pTA2:FragC_13. A DNA fragment excised from pTA2:FragC_13 with BsaI/BamHI was inserted into the same site of pMD19:FragA_013. The resulting plasmid was digested with BglII/BamHI and inserted into the BamHI site of pDONR221:∆Nstalk-b, generating pDONR: ΔNS-B.

To construct pIFD-∆NS-C∆Chead, a DNA fragment was amplified by iPCR from pDONR221:∆NS-A2 using primers ∆NS-C∆Chead-f/∆NS-C∆Chead-r and was self-ligated, generating pDONR221:∆NS-A2a. Subsequently, a DNA fragment excised from pTAKN-2:Trp11_047 & 048 with BglII/BamHI was inserted into the BamHI site of pMD19:FragB. The resulting plasmid was digested with BglII/BamHI and inserted into the BamHI site of pMD19:FragA. Then, the resulting plasmid was digested with BglII/BsaI and inserted into the BamHI/BsaI sites of pDONR221:∆NS-A2a, generating pDONR221:∆NS-C∆Chead.

To construct pIFD-∆Cstalk, a DNA fragment excised from pDONR221:ataA with KpnI/XbaI was inserted into the same site of the pTA2 vector, generating pTA2:Ctemp. A DNA fragment was amplified by iPCR from pTA2:Ctemp using primers ∆Cstalk-f/∆Cstalk-r and was self-ligated. The resulting plasmid was digested with BglII/XbaI and inserted into the same site of pDONR221:ataA, generating pDONR221:∆Cstalk.

To construct pIFD-mini-AtaA, a DNA fragment was amplified by iPCR from pDONR221:ataA using primers mini-AtaA-f/mini-AtaA-r and was self-ligated, generating pDONR221:mini-ataA.

Finally, all of these ataA mutants in pDONR221 plasmids were excised with EcoRI/XbaI and inserted into the same site of pARP3, generating plasmids for expression. Transformation of the ΔataA mutant strain Tol 5 4140 with these expression vectors was carried out by conjugal transfer from the E. coli S17-1 strain as previously described (Ishikawa et al., 2012).

The expression and production of mutant variants of AtaA were detected by SDS–PAGE followed by Coomassie brilliant blue (CBB) staining and immunoblotting as described previously (Ishikawa et al., 2012) with a slight modification. To confirm the production of AtaA and its mutants in E. coli, SDS-sample buffer (5% (v/v) 2-mercaptoethanol, 2% (w/v) SDS, 0.02% (w/v) bromophenol blue, 62.5 mM Tris-HCl, pH 6.8) supplemented with 8 M urea and anti-AtaA_59-325 rabbit antiserum (Aoki et al., 2020) was used.

Flow cytometry was performed as described previously (Ishikawa et al., 2012) with a slight modification. In brief, bacterial cells were resuspended in PBS containing 4% (w/v) paraformaldehyde and incubated at room temperature for 15 min. The samples were washed with PBS and treated with anti-AtaA_699-1014 (Ishikawa et al., 2012) or anti-AtaA_59-325 (Aoki et al., 2020) rabbit antiserum at a 1:10000 dilution in PBS containing 0.05% (v/v) Tween 20. After a 30-min incubation at room temperature, the samples were washed twice with NET buffer (150 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, 0.05% (v/v) Triton X-100, pH 7.6) and treated with Alexa Fluor 488-conjugated anti-rabbit antibody (Cell Signaling Technology, MA) at a 1:500 dilution in NET buffer for 30 min. The samples were washed twice with NET buffer and resuspended in deionized water, and the fluorescence was measured by FACS Canto II (Becton, Dickinson and Company, NJ, United States). Histograms were created using Flowjo software (Tomy Digital Biology, Tokyo, Japan).

Immunofluorescence microscopy was performed as described previously (Ishikawa et al., 2016) with a slight modification. In brief, bacterial cells were placed onto a gelatin-coated glass plate, fixed with PBS containing 4% (w/v) paraformaldehyde solution for 15 min, and washed twice with PBS. Anti-AtaA_59-325 rabbit antiserum at a 1:10,000 dilution in PBS containing 0.05% (v/v) Tween 20, was placed onto the sample plate. After a 30-min incubation, the sample plate was washed with NET buffer, incubated for 30 min with Alexa Fluor 488-conjugated anti-rabbit antibody at a 1:500 dilution in NET buffer. The sample plate was washed twice with PBS and observed under a confocal laser-scanning microscope (FV-1000, Olympus Corporation, Tokyo, Japan).

Adhesion assays using microwell plates were performed as previously described (Ishikawa et al., 2012). In brief, 200 μl of a bacterial cell suspension in BS-N buffer (34.5 mM Na₂HPO₄, 14.7 mM KH₂PO₄, 15.5 mM K₂SO₄, pH 7.2) at an optical density at 660 nm of 0.5 was placed into a 96-well PS plate (353072; Becton, Dickinson and Company, NJ, United States) or a 96-well glass plate (FB-96; Nippon Sheet Glass Co., Ltd., Tokyo, Japan). After incubation for 2 h at 28°C without shaking, the cell suspensions were removed by a micropipette, and each well was rinsed three times with 200 μl of BS-N buffer. Cells adhering to the microwell surface were stained with 0.1% (w/v) crystal violet solution for 15 min. After three rinses with 200 μl of BS-N buffer, the stain was eluted from the cells with 200 μl of 70% (v/v) ethanol, and the absorbance at 590 nm of the elution was measured by a microplate reader (ARVO X3; PerkinElmer, MA, United States).

The immobilization of bacterial cells on polyurethane foam supports was performed as described previously (Yoshimoto et al., 2017), with slight modifications. The bacterial cells were suspended in 30 ml of BS-N buffer at an OD₆₀₀ of 1.0 in a 100-mL Erlenmeyer flask. Five pieces of polyurethane foam support with a specific surface area of 50 cm²/cm³ (CFH-40; Inoac Corporation, Nagoya, Japan) in the shape of a cube (1 cm³) were placed into the cell suspension and shaken at 115 rpm at 28°C for 60 min. After shaking, the OD₆₀₀ of the cell suspension was measured. The immobilization ratio of the cells was calculated using the following equation: I m m o b i l i z a t i o n   r a t i o % = 100 × O D 600   i n i t i a l − O D 600   a f t e r   s h a k i n g / O D 600   i n i t i a l (1)

For the detachment assay, the support with immobilized cells was placed into a 100-mL Erlenmeyer flask containing 25 mL of BS-N buffer supplemented with or without 1% (wt/vol) casein hydrolysate (casamino acids technical grade; Becton, Dickinson and company) and shaken at 115 rpm at 28°C. After 5 min of shaking, the OD₆₀₀ of the suspension containing detached cells was measured (OD₆₀₀-detached). The detachment ratio of the cells was calculated from the following equation: D e t a c h m e n t   r a t i o % = 100 × O D 600   d e t a c h e d / O D 600   i n i t i a l − O D 600   a f t e r   s h a k i n g (2)

For the re-immobilization assay, the cell suspension containing detached cells and 1% casein hydrolysate was centrifuged (5,000×g, 10 min, 25°C), and the precipitated cells were rinsed with BS-N buffer three times. The cells were resuspended in BS-N buffer at an OD₆₀₀ of 1.0 and shaken with 5 pieces of polyurethane foam support in a 100-mL Erlenmeyer flask at 115 rpm at 28°C. The OD₆₀₀ of the cell suspension was measured periodically, and the immobilization ratio was calculated from Equation 1.

The polyurethane support on which the bacterial cells were immobilized was rinsed in BS-N buffer. The support was sliced into 1 mm-thick and immersed in 4% paraformaldehyde phosphate buffer solution (FUJIFILM Wako, Osaka, Japan) for 15 min. After a rinse in BS-N buffer, the immobilized cells on the support were stained with 60 µM propidium iodide (Invitrogen, Waltham) dissolved in 50 mM HEPES buffer (pH 7.4) for 15 min. The support was rinsed in BS-N buffer and observed in BS-N buffer under a digital microscope (KEYENCE, Osaka, Japan) at ×100 magnification.

For reactions in cell suspension, cells of E. coli JCM-Tfu0937 (bgl)-Blc derivatives expressing β-glucosidase (BGL) constitutively were suspended in 10 ml of BS-N buffer in a 15-mL centrifuge tube to an OD₆₀₀ of 1.0. The reaction was initiated by adding 4-nitrophenyl β-D-glucopyranoside (FUJIFILM Wako) at a final concentration of 1 mM, and the absorbance at 405 nm was measured after shaking at 37°C for 30 min. The concentration of p-nitrophenol produced by BGL was calculated from the increase in absorbance at 405 nm.

For reactions with immobilized cells, the polyurethane support on which cells of E. coli JCM (bgl, mini-ataA) were immobilized was rinsed in BS-N buffer and dewatered on paper towel for 2 min. One piece of the support with the immobilized cells were incubated in 5 ml of BS-N buffer containing 0.2 mM 4-nitrophenyl β-D-glucopyranoside at 37°C with shaking and the absorbance of at 405 nm was measured periodically. After 50-min reaction, the support was collected, washed with fresh BS-N buffer, transferred to a fresh reaction solution, and subjected to a new cycle of the reaction.

To evaluate the contribution of each domain of AtaA to the adhesiveness, we constructed six in-frame deletion (IFD) mutant genes for plasmid-based expression in ΔataA strain of Acinetobacter sp. Tol 5, as shown in Figure 1A. When constructing domain-deletion mutant genes of large and complex proteins, great care and precise design are necessary to prevent undesirable structural distortion and steric conflicts in the resulting mutant proteins. Recently, we confirmed that preserving the periodicity of hydrophobic residues in coiled coils at the deletion region is effective for displaying the very large homotrimeric fiber structure and maintaining the proper folding state of AtaA on the cell surface after a domain is deleted (Aoki et al., 2020). Therefore, each IFD mutant protein was carefully and precisely designed and constructed so that coiled-coil periodicity was maintained after deletion, following established procedures for TAA domain prediction (Szczesny and Lupas, 2008; Bassler et al., 2015). The IFD mutant genes were expressed in the Tol 5 ΔataA strain Tol 5 4140, and successful recombinant protein production was confirmed by SDS–PAGE followed by CBB staining. All of the IFD mutant proteins were detected as well as full-length AtaA (FL-AtaA) (Figure 1B, dotted line). A band corresponding to a higher molecular weight was also detected in IFD-ΔNS-CΔChead (Figure 1B, arrowhead). The band was considered to be an undenatured trimeric state of the mutant protein because many TAAs are known to have a tolerance to heat denaturation (Grosskinsky et al., 2007; Riess et al., 2007) that is not easily resolved by excess heating. Subsequently, we examined the cell surface display of the IFD mutant proteins by flow cytometry. Tol 5 4140 cells expressing the IFD mutant genes except IFD-ΔNS-A2 were immunostained with anti-AtaA_699-1014 antiserum and Alexa 488-conjugated anti-rabbit antibody. For IFD-ΔNS-A2, which lacks the antigenic region relevant for the anti-AtaA_699-1014 antiserum, anti-AtaA_59-325 antiserum was used instead. All of the cells expressing the IFD mutant genes exhibited a high fluorescence intensity similar to that of the cells expressing FL-AtaA (Figure 1C). These results suggest that all of the IFD mutant proteins were properly produced, secreted, and displayed on the cell surface.

Next, the adhesiveness of Acinetobacter cells expressing the IFD mutant proteins was evaluated by an adhesion assay using microwell plates. The cell suspensions of the Tol 5 4140 cells expressing the IFD mutant genes were placed into hydrophobic polystyrene (PS) and hydrophilic glass microwell plates. After an incubation for 2 h, the cell suspensions containing non-adhering cells were removed, and the adhered cells that remained in the microwell plates were quantified by crystal violet staining. The cells expressing the Nhead-deleted mutant (IFD-ΔNhead; Δ60-313) exhibited a significantly lower adhesiveness to both PS and glass surfaces than that of the cells expressing FL-AtaA (Figure 2), although a similar amount of AtaA molecules was displayed on the cell surface as described above (Figure 1C). In contrast, cells displaying other IFD mutant proteins maintained the same degree of high adhesiveness as that of FL-AtaA. Thus, Nhead was revealed to be the essential domain for the adhesive function of AtaA.

To extend the applicability of efficient heterologous expression of ataA to bacteria other than Acinetobacter species, we designed a miniaturized AtaA (mini-AtaA; Δ364-3218 aa). The mini-AtaA consists of a signal peptide (1-59 aa) for secretion through the inner membrane into the periplasm, Nhead (60-325 aa) for adhesive function, FGG1 (326-363 aa) for seamless connection to the following domains, Cstalk (3219-3561 aa) as the C-terminal region of the PSD to ensure effective secretion, and TM (3562-3630 aa) to display the PSD on the surface of OM (Figure 3A). Note that PSD C-terminal regions are often involved in the secretion of the whole PSD through the TM in autotransporter families (Rojas-Lopez et al., 2017). The number of amino acids in mini-AtaA (775 aa) was reduced to 21% of that in FL-AtaA (3630 aa), and its molecular weight was only 76 kDa as a monomer. E. coli BL21 (DE3) cells were transformed with an expression plasmid for mini-AtaA or FL-AtaA, and their protein production was analyzed by SDS–PAGE, followed by immunoblotting using anti-AtaA_59-325 antiserum. From the E. coli cells expressing the FL-ataA gene (E. coli (FL-ataA)), a faint band corresponding to the theoretical molecular weight of FL-AtaA was detected, but several bands corresponding to lower molecular weights were also detected, suggesting that most of the FL-AtaA was degraded (Figure 3B). From the E. coli cells expressing the mini-ataA gene (E. coli (mini-ataA)), a single band corresponding to the theoretical molecular weight of mini-AtaA was clearly detected. Subsequently, we examined their cell surface display by flow cytometry and immunofluorescence microscopy using anti-AtaA_59-325 antiserum. The flow cytometry results clearly showed a much larger amount of Nhead on the cell surface of E. coli (mini-ataA) than on that of E. coli (FL-ataA) (Figure 3C). Immunofluorescence microscopy also showed that a higher amount of mini-AtaA was displayed on the E. coli cell surface compared to the weak expression signal of FL-AtaA (Figure 3D).

We tried to immobilize the E. coli cells expressing mini-ataA onto a polyurethane foam, which is often used in bioprocesses as a support (Al-Amshawee et al., 2020). For comparison, E. coli wild type (E. coli WT), E. coli (FL-ataA), and Acinetobacter sp. Tol 5 were also subjected to the cell immobilization experiment using polyurethane foam support (see materials and methods section). After shaking for 60 min in the presence of polyurethane foam support, E. coli WT and E. coli (FL-ataA) cell suspensions remained cloudy, indicating that many of the cells did not adhere to the polyurethane foam support (Figure 4A). In contrast, the cell suspension of E. coli (mini-ataA) became clear, as did Tol 5, suggesting that most of the cells adhered to the support. Microscopy revealed polyurethane fibers covered with adherent E. coli (mini-ataA) cells (Figure 4B). Then, the immobilization ratio was quantified based on the decrease in the OD₆₀₀ of the cell suspension after shaking. While the immobilization ratios of E. coli WT and E. coli (FL-ataA) were only 8% and 40%, respectively, that of E. coli (mini-ataA) was 85% (Figure 4C). This immobilization ratio was slightly lower than the 97% of Acinetobacter sp. Tol 5 but significantly improved compared to that of E. coli (FL-ataA). Therefore, we achieved efficient immobilization of E. coli cells by the expression of a single recombinant gene, mini-ataA.

In our previous study (Ohara et al., 2019), immobilized Tol 5 cells on a polyurethane support were easily detached from the support by rinsing the cells with a solution containing casein hydrolysates, then the cells were re-immobilized onto the support without impairing their adhesiveness by removing the casein hydrolysates. Here, we examined whether a similar process is applicable to E. coli (mini-ataA). Five pieces of polyurethane foam support on which E. coli (mini-ataA) cells were immobilized were placed in 100-mL Erlenmeyer flasks containing 25 mL BS-N buffer supplemented with or without 1% casein hydrolysates. After 5 min of shaking, only 6% of E. coli (mini-ataA) cells detached from the support in BS-N buffer alone, but 78% of the cells detached in the BS-N buffer supplemented with 1% (w/v) casein hydrolysate (Figure 5A). Subsequently, the detached cells were collected by centrifugation, washed with BS-N buffer, resuspended in BS-N buffer at an OD₆₀₀ of 1.0, and tested again for immobilization on the polyurethane foam support. The detached cells were immobilized as quickly as fresh cells, and the immobilization ratio reached 80% in 30 min (Figure 5B). These results imply that mini-AtaA retains adhesive properties similar to those of full-length AtaA, including a sensitivity to inhibitors of Nhead-surface interactions.

We intended to enzymatic reaction by E. coli cells immobilized with mini-AtaA on polyurethane foam support. For this purpose, E. coli JCM-Tfu0937 (bgl)-Blc, the β-glucosidase gene-introduced E. coli strain JCM20137 (Tanaka et al., 2014), were transformed with the mini-ataA gene, generating E. coli JCM (bgl, mini-ataA). First, we examined the effect of mini-ataA expression on the growth of the E. coli cells and cellular BGL activity reflecting the production level of the enzyme. E. coli JCM (bgl, mini-ataA) showed a similar growth curve with the control cells harboring an empty vector (Figure 6A); the maximum specific growth rates (μ _max) of JCM (bgl, mini-ataA) and the control cells were 1.9 h⁻¹ and 2.0 h⁻¹, respectively. The BGL activity of the cell suspensions was measured using 4-nitrophenyl β-D-glucopyranoside as a substrate. As a result, E. coli JCM (bgl, mini-ataA) showed approximately 80% BGL activity of the control cells (Figure 6B).

Next, JCM (bgl, mini-ataA) were immobilized on the polyurethane foam support, resulting in the similar immobilization ratio (>80%) (Figure 6C) as that of BL21 (DE3) expressing mini-AtaA (Figure 4C), although the control cells were hardly immobilized (Figure 6C). These immobilized cells were subjected to the repetitive enzymatic reaction to assess their reuse stability. One piece of the support with immobilized cells was incubated in BS-N buffer containing the substrate. After 50-min incubation at 37°C with shaking, the color of the reaction mixture was changed into yellow due to p-nitrophenol produced by BGL (Figure 6D). The absorbance at 405 nm was periodically measured during the incubation to obtain the p-nitrophenol concentration. Then, the support was picked up, washed with fresh BS-N buffer, transferred to a new reaction solution, and thereafter, another cycle of the reaction was repeated. The support with cells immobilized by expressing mini-AtaA maintained high BGL activity at least five cycles over a 5-h period (Figure 6E). These results demonstrate that the immobilized E. coli cells by mini-AtaA can be used for repetitive enzymatic reactions as a whole-cell catalyst.