The gram-negative bacterium Escherichia coli possesses various protein secretion systems (SS), including types I-VI (T1SS-T6SS) and type VIII (T8SS). Among these, the T5SS system-especially T5SSa, an autotransporter (AT) pathway-stands out as the simplest polyprotein related to protein secretion, comprising three functional domains—an inner membrane transport domain, a passenger domain on the surface, and a β-domain that serves as a translocation pore for the passenger domain (1).

Antigen 43 (Ag43), which belongs to the T5SSa system, is expressed on the surface of E. coli outer membrane (OM) in an embedded state and functions in cell aggregation and biofilm formation (2). Ag43 is a high-copy expression system (> 50,000 copies per cell) (3) and, has been used in several applications for the display of various recombinant proteins (4-6). Ag43 was initially synthesized as a 1,039 amino acid (aa)-long polypeptide consisting of a signal peptide (SP), passenger domain, and translocation unit (TU) (6). TUs consist of a concentric α-helical linker domain (α-HLD) and a 12 β-subunit, which form a stable pore-like barrel structure on the OM, leading to an exterior orientation of the passenger domain in the transmembrane cylindrical pore (7). While TUs share structural characteristics, their ability to display heterologous proteins on the surface varies markedly. Understanding Ag43β-barrel (Ag43β) structures would provide an insight into the utility of the Ag43 AT system for producing engineered biomaterials.

Although the domains and common architectures of Ag43β have been studied, the Ag43β expression module has not been extensively characterized, prompting us to systematically investigate this aspect. We used the Molecular Operating Environment (MOE) for protein structure modeling to show that the α-HLD of the Ag43β TU effectively forms stable hydrogen bonds with neighboring β-strands. While the structure of the full-length Ag43β from E. coli has not been reported to date, MOE allows three-dimensional molecular visualization of Ag43β through sequence-to-structure prediction and intermolecular docking characterization (8, 9).

Bacterial outer membrane vesicles (OMVs) are lipid-bilayer membrane nanovesicles formed by the bulging and pinching-off of a portion of the OM. They contain biologically active components typically found in the gram-negative OM and offer several advantages as nanocarriers (10). To date, nanoscale OMVs have been exploited as attractive immune boosters, in drug delivery, for intra/inter-species communication, etc. (11-13). OMVs can be systematically engineered to express heterogeneous proteins, offering the possibility of acquiring OMVs with exterior or interior display of chimeric proteins.

The yield of naturally released OMVs (nOMVs) is often too low for therapeutic or industrial applications. Strategies to increase OMV yields have been developed based on gene disruption using the Tol-Pal system (14), mutation of lipopolysaccharide (LPS) (15), and physical or chemical treatments (16). In this study, we isolated and purified OMVs using a lysozyme-based detergent (LD-OMVs).

To systematically analyze whether Ag43β variants can be used as a platform for the simultaneous display of chimeric proteins, we generated fluorescent chimeras using genetic fusion of the Ag43β variants with enhanced green fluorescent protein (eGFP). Next, LD-OMVs were compared with nOMVs to assess whether these LD-OMVs would have higher yield than that of nOMVs for the engineering of bacteria carrying Ag43β variants to produce the desired target.

Chimeric Ag43β surface display model construction

E. coli Ag43β is a well-known bacterial AT that functions via self-recognition of the passenger domains. Ag43β AT polypeptides are 1,039 aa long and contain the following: a signal peptide (SP; aa 1-52), an N-terminal unstructured domain of unknown function (aa 53-138), a functional passenger domain (aa 139-600) with a proteolytic cleavage site at residue 552, an autochaperone (AC) domain (aa 601-700), an α-HLD (aa 701-740), and a C-terminal β-barrel (aa 741-1,039) (Fig. 1A). The β-barrel assembly machinery catalyzes the essential β-barrel insertion into the OM. The C-terminal β-barrel pore is necessary for the translocation of the functional passenger domain to the cell exterior (17). The AC domain is required for proper extracellular folding and secretion in numerous ATs. However, some studies have demonstrated that stable surface display of heterologous globular proteins can be accomplished in the absence of both the native passenger and AC domains (6, 17). The inconsistencies in the literature regarding the AC domain prompted us to further investigate the relevance of this structural element in Ag43β.

To explore the influence of crucial components of the Ag43β surface display system, we designed a series of truncated variants of Ag43β. Supplementary Fig. 1 shows the domain organization of Ag43β and the structures of the following generated plasmids: Ag43β138_eGFP, Ag43β552_eGFP, Ag43β700_eGFP, Ag43β740_eGFP, and Ag43β926_eGFP. Ag43β138_eGFP encodes the native Ag43 SP, Ag43 amino acids 138 through 1,039, and eGFP. In order to investigate the minimum surface display module, the downstream regions of the passenger domain (aa 552-600), native Ag43 AC region (aa 601-700), and α-HLD in the TU (aa 701-740) were deleted. Finally, to assess the potential for maintaining the β-barrel anchor, the N-terminal β-barrel compartments (aa 741-926) were deleted. All transformants were successfully expressed and observed on the basis of their fluorescence in cells (data not shown). However, the expression of Ag43β_eGFP was insufficient to provide information on the extracellular surface display.

The functionality of eGFP fused with the Ag43β variants was assessed using flow cytometry (Fig. 1B). Recombinant bacterial cells bearing the Ag43β variants Ag43β700_eGFP and Ag43β740_eGFP were grown and induced overnight to trigger eGFP expression and surface display. Cells harboring Ag43β700_eGFP had a significant double-positive eGFP/Alexa Flour^TM 647 area (Q1) of 70.19%, whereas that of cells harboring Ag43β740_eGFP was 13.26%.

To assess whether eGFP was displayed on the exterior or interior, we carried out extracellular protease digestion (Supplementary Fig. 2). Targets of interest are accessible to extracellular trypsin if displayed on the exterior of the cell but remain resistant when trapped beneath the cellular membrane; periplasmic or cytoplasmic eGFP is not degraded unless the bacterial cells are lysed by treatment with proteases and detergents. Fig. 1C shows SDS-PAGE and western blot results for eGFP from trypsin-treated (+) and non-treated (−) cells. The most significant decrease in band intensity, compared to that in the non-treated control was observed for cells expressing Ag43β700_eGFP. Therefore, Ag43β700 can effectively present chimeric proteins on the exterior of the OM. In the case of Ag43β740_eGFP, eGFP remained largely undigested even at high trypsin concentrations (up to 1 μg/ml), consistent with flow cytometry results. Ag43β926_eGFP exhibited greater resistance to trypsin digestion, as shown in Supplementary Fig. 2, which suggested that the eGFP was trapped in the bacterial periplasm. Additionally, Ag43β926_eGFP showed low expression levels. Therefore, we hypothesized that Ag43β926_eGFP probably did not form a complete β-barrel structure. Owing to the absence of the α-HLD, the eGFP may not be inserted into the membrane and could potentially float freely in the periplasm. These observations led us to investigate eGFP translocation across the β-barrel pore (Fig. 1D). We found that the embedded α-HLD was necessary for the insertion of chimeras into the β-barrel pore, indicating its crucial structural role. Furthermore, in the α-HLD deletion variant Ag43β740, β-barrel strands led to the retention of the chimeric proteins in the periplasm. This affected both structural plasticity and periplasmic degradability.

α-HLD is an essential cue for achieving surface display and Ag43β machinery

The C-terminal TU consists of a β-barrel and an α-HLD. While these structures enable Ag43β to mediate extracellular display through transmembrane pore formation, they offer limited insights into the exterior orientation of Ag43-mediated surface display. The Ag43β700_eGFP construct included a native TU, including the α-HLD, whereas Ag43β740_eGFP did not. Initial studies using cells harboring Ag43β700_eGFP indicated that functional surface display of eGFP could be achieved in this context. Therefore, we next systematically investigated the contribution of α-HLD in the TU.

The three-dimensional structure of the TU predicts a central α-HLD (aa 701-740) winding through an antiparallel 12-strand β-barrel. The TU also shows a fully zipped conformation between its C-terminal strand (β12) and strand β1 (Fig. 2A, strands colored in green). The α-HLD, as indicated by its topological features and amino acid sequence, comprises two interfacing parts—a pore-facing plug (V727-E740) and a pore-releasing tip (R701-V726). Interestingly, the pore-releasing tip is positioned at the top edge of the β-barrel domain, extending in an elongated conformation toward the external environment. In the pore-facing bridge region, residues G729, R731, H733, Q734, and E740 are located near the periplasmic end of the barrel strands, engaging in hydrogen-bond interactions with the internal residues R777, E779, N830, E872, Q890, Q998, and R1004 in strands β2, β4, β6, β7, and β10 (Fig. 2B, C, and Supplementary Fig. 3). These findings reveal that the conformational assembly necessitates the burial of the pore-facing plug into the internal cavity. Therefore, it is most plausible that α-HLD serves as the hinge linker for recruiting the protein of interest into the proximity of the primed β-strand pore for spontaneous integration. Accordingly, α-HLD may be essential for the stable conformation of the β-barrel structure.

Enhanced purification yield of OMVs and extracellular targeting of Ag43β-eGFP on the OMVs

Gram-negative bacteria naturally release OMVs with typical OM compositions, including LPS, phospholipids, and OM proteins. The potential therapeutic applications of engineered OMVs have been described (12, 18). Furthermore, OMVs have been utilized as carriers for exogenous enzymes in bioremediation and industrial applications (19). In the present study, we adopted the lysozyme-based detergent isolation technique (Fig. 3A) and compared it to the common isolation method involving nOMVs (Table 1). Both LD-OMVs and nOMVs showed no differences in terms of OMV particle size. However, the LD-OMVs displayed significantly higher particle concentrations than those isolated using nOMVs.

Measurements on the basis of cryo-TEM images revealed a range of Ag43β variant LD-OMVs, with diameters approximately 240 nm, consistent with the OMV particle diameter measured using the qNano Gold system (Fig. 3B). Extracellular trypsin treatment of LD-Ag43β700_eGFP-OMVs and LD-Ag43β740_eGFP-OMVs produced different band patterns as observed after bacterial trypsin treatment (Fig. 3C).

Furthermore, a comparison of the two OMV purification methods revealed the presence of internal components. A recent study demonstrated the presence of common cytoplasmic proteins in culture-derived OMVs (10). Moreover, the possibility of gene transfer between bacteria and host cells via OMVs has been reported (20, 21). This suggests that OMV-associated genetic materials may play an important role in bacterial pathogenesis. Analysis of OMV-derived 16s rDNA showed high levels of bacterial DNA amplification in nOMVs, which was not observed in LD-OMVs (Fig. 3D).

This study demonstrated that the bacterial Ag43β AT platform facilitates efficient presentation of heterologous chimeric proteins on the OM and OMV surfaces. Our first strategy, using truncated Ag43β forms fused with eGFP, allowed straightforward characterization of the key components of the β-barrel scaffold, based on the basic β-barrel topology, using MOE prediction. Based on MOE evaluation of the Ag43β truncation variants, we speculated that the α-HLD facilitated the translocation of eGFP to the cell exterior. In this scenario, Ag43β700-eGFP correctly recruits eGFP to the surface, leading to the interaction of the α-HLD with several β-strands (β2, β4, β6, β7, and β10, see Fig. 2B). In contrast, Ag43β740-eGFP showed intracellular trapping. In addition, we found an abrupt alteration in Ag43β926-eGFP due to the lack of support from periplasmic plugs (N-terminal strands β1 to β8), perturbing the formation of the membrane spanning pore.

The second approach we used to understand the Ag43β AT was to use OMVs containing the OM-associated bacterial compartment. OMVs produced by gram-negative bacteria have been widely investigated for application as biomimetic nanocarriers because they are engineerable. We developed two Ag43β-derived modules (Ag43β700-eGFP and Ag43β740-eGFP) to target chimera to different OMV compartments, including the lumen and the outer surface. Early studies on this topic have shown that enzymatic activity can be preserved through OMV-luminal enzyme localization (19, 22). Furthermore, antigens presented on the surfaces of OMVs can be phagocytosed via the pathogen-associated pattern recognition system to stimulate a strong and specific immune response (23, 24). However, important drawbacks of these applications are their low OMV yield and the toxicity of LPS. Current approaches involve using a genetically modified vesicle-overproducing host and detergent extraction, which reduces the content of reactogenic LPS (14-16, 25). However, detergent extraction leads to the removal of chimeric components (26) and promotes vesicle aggregation by altering the OMV structure (27). Recent efforts to improve OMV extraction have focused on novel techniques that allow chimeric functionality retention. Chemical modification of Ag43β700-OMVs using lysozyme induces the shedding of OM sheets, releasing the cytosolic components. LD-Ag43β-OMVs had a morphology and diameter similar to those of nOMVs but were produced in higher yields. In general, nOMVs typically exhibit a low extraction yield (100 ng/10⁸ colony forming units), which renders clinical application challenging (28). Increasing the yield of vesicles appears to be sufficiently feasible for clinical use as vaccines, drug delivery vehicles, and adjuvants (29, 30). For example, Park et al. developed an effective vaccine platform that can prevent Pseudomonas aeruginosa (P. aeruginosa) and SARS-CoV-2 infection by increasing the amount of OMV using a detergent-based OMV extraction method (31). Therefore, lysozyme digestion may be an improved method for isolating OMVs, with a faster isolation process and higher yield.

In conclusion, this study provides insights into the autonomous functionality of Ag43β in facilitating the expression of chimeric proteins, irrespective of the exterior/interior expression machinery. Notably, our findings highlight the pivotal role played by the interaction between α-HLD and periplasmic plug β-strands in enhancing the efficiency of this process. Moreover, our study demonstrates that LD-OMVs can be obtained in a significantly shorter timeframe and a greater yield than those for nOMVs. This not only streamlines the production process but also provides new avenues for leveraging OMVs in various applications. Furthermore, the remarkable biological activity displayed by the OMVs emphasizes their potential for pioneering surface display strategies, such as antigen target delivery. Overall, our results offer valuable insights into enhancing OMV-based platforms for a wide range of biochemical and molecular biological applications.

Materials and methods are available in the Supplementary File.

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (No. 2020R1A6A1A06046235), the Korea Institute of Planning and Evaluation for Technology in Food, Agriculture and Forestry (IPET) through the Crop Viruses and Pests Response Industry Technology Development Program, funded by the Ministry of Agriculture, Food and Rural Affairs (MAFRA; No. 321108-04), and the Commercializations Promotion Agency for R&D Outcomes (COMPA) grant funded by the Korea government (MSIT; No. 1711173792).

The authors have no conflicting interests.