Overexpression of the Bam Complex Improves the Production of Chlamydia trachomatis MOMP in the E. coli Outer Membrane

A licensed Chlamydia trachomatis (Ct) vaccine is not yet available. Recombinant Chlamydia trachomatis major outer membrane protein (Ct-MOMP), the most abundant constituent of the chlamydial outer membrane complex, is considered the most attractive candidate for subunit-based vaccine formulations. Unfortunately, Ct-MOMP is difficult to express in its native structure in the E. coli outer membrane (OM). Here, by co-expression of the Bam complex, we improved the expression and localization of recombinant Ct-MOMP in the E. coli OM. Under these conditions, recombinant Ct-MOMP appeared to assemble into a β-barrel conformation and express domains at the cell surface indicative of correct folding. The data indicate that limited availability of the Bam complex can be a bottleneck for the production of heterologous OM vaccine antigens, information that is also relevant for strategies aimed at producing recombinant OMV-based vaccines.


Introduction
Chlamydia trachomatis is an obligate intracellular Gram-negative bacterium that is a major cause of common sexually transmitted infection (chlamydia) and blindness (trachoma) worldwide [1,2]. Chlamydia trachomatis has a complicated life cycle that begins with metabolically reduced elementary bodies (EBs) that are infectious and invade host cells. After endocytosis, the EBs transform into reticulate bodies (RBs) of low metabolic activity. The RBs then replicate and re-differentiate into the EBs that are released to infect neighboring cells [3,4].
Although treatments exist, chlamydial infections often start asymptomatic, which drives transmission in human populations. When left untreated, an infection can lead to chronic pelvic pain, infertility, and trachoma. The infection can increase the susceptibility to other sexually transmitted pathogens such as HIV, HPV, and Gonorrhea [5]. Therefore, an effective preventive vaccine is considered the best option to reduce chlamydial infection and disease [2,4,6]. A complicating factor in the development of a protective vaccine is the presence of different serotypes and their distinct surface-exposed antigens that correlate with specific immunity. Also, natural infection was shown to provide only transient immunity and the correlates of protection have not been fully elucidated [7]. In any case, both humoral and cellular immune responses appear to be required at the site of infection [8]. Early studies using whole-organism vaccines have been unsuccessful, showing immune responses that were serovar-specific, short-lived, and sometimes even causing delayed hypersensitivity upon re-infection [9]. Although a more recent wholecell vaccine showed promising results [10], vaccine efforts have shifted towards subunit formulations consisting of protective antigens.

Expression of Ct-MOMP in E. coli
To express natively folded Ct-MOMP in the OM of E. coli, a plasmid (pLemo-MOMP) was constructed in which the expression of the Ct-MOMP gene from serovar D strain UW3/Cx is under control of the L-rhamnose-inducible rhaBAD promoter. This promoter was chosen because it allows for a precise tuning of expression to avoid possible saturation of the membrane translocation channels and chaperones needed for the proper delivery of highly expressed secretory and membrane proteins [29][30][31]. The native signal peptide of Ct-MOMP was replaced by an E. coli signal peptide (OmpT) for an optimal engagement of the cognate Sec-translocon in the inner membrane [32]. Furthermore, 9 cysteine residues of Ct-MOMP were substituted by alanines to prevent potential aggregation resulting from improper intra-and inter-molecular disulfide bond formation by the periplasmic oxidoreductase DsbA (see Supplementary Document 1). The modified Ct-MOMP was expressed in E. coli BL21(DE3) upon induction with L-rhamnose (2-8 mM) for 2 h ( Figure  S1). Whole-cell lysates were analyzed by SDS-PAGE and Coomassie staining ( Figure S1A) or immunoblotting using Ct-MOMP-specific antiserum for detection ( Figure S1B). Tightly regulated expression of Ct-MOMP at its expected molecular weight (~41 kDa) was observed by immunoblotting, with increasing L-rhamnose concentrations leading to increasing levels of Ct-MOMP ( Figure S1B). However, even at the highest level of induction, Ct-MOMP could not be clearly distinguished by Coomassie staining ( Figure S1A, lane 4).

Improvement of Ct-MOMP Levels upon Co-Overexpression of the E. coli Bam Complex
We then considered the possibility that the observed low expression of Ct-MOMP in E. coli was due to inefficient insertion into the OM provoking recognition and degradation by the periplasmic quality control systems [33]. Potentially, this may originate from a mismatch between the heterologous Ct-MOMP and the E. coli Bam complex. Recently, we have shown that co-overexpression of the E. coli Bam complex improves the secretion of difficult-to-secrete recombinant autotransporter chimeras that include an OM-inserted β-barrel domain and are sensitive to degradation when accumulating in the periplasm [28]. To test whether this approach could also relieve a bottleneck in the production of Ct-MOMP, we co-transformed cells carrying pLemo-MOMP with a compatible plasmid, pJH114, that encodes all subunits of the E. coli Bam complex under control of the IPTG-inducible trc promoter [34]. To ensure the availability of additional Bam complexes for Ct-MOMP biogenesis, production of the Bam subunits was induced 1 h prior to induction of Ct-MOMP expression (with 8 mM L-rhamnose). Cell samples were subsequently taken after 2 h and analyzed by SDS-PAGE and Coomassie staining ( Figure 1A) or immunoblotting using antisera against either Ct-MOMP or the Bam complex ( Figure 1B,C, respectively). As expected, gradually increasing levels of Bam subunits accumulated in the cells ( Figure 1C) in an IPTG-concentration dependent manner. BamA could be readily detected upon Coomassie staining, whereas detection of other Bam subunits (BamBCDE) appeared obscured by endogenous proteins ( Figure 1A). Importantly, improved expression of Ct-MOMP was observed by immunoblotting ( Figure 1B) that resulted from increasing the Bam complex levels ( Figure 1C). This suggests that co-overexpression of the Bam complex has a favorable impact on the biogenesis of Ct-MOMP in E. coli. A regime involving induction of Bam expression with 100 µM IPTG 1 h prior to the addition of 8 mM L-rhamnose yielded the highest levels of Ct-MOMP and was selected for our further studies on Ct-MOMP biogenesis ( Figures 1B and S3).
To exclude that the observed effects were strain-specific for E. coli BL21(DE3), we tested the influence of Bam overexpression on Ct-MOMP biogenesis in different E. coli and Salmonella Typhimurium backgrounds. E. coli DH5α was chosen as representative of the E. coli K12 lineage, while BL21(DE3)-derivative ClearColi(DE3) [35] and Salmonella Typhimurium SL3261-∆msbB [36] were selected because these strains express detoxified versions of lipopolysaccharide, which is relevant for bacteria-based vaccine development. Induction of the Bam complex from pJH114 with IPTG resulted in a Coomassie-detectable increase in BamA levels ( Figure S2, lanes 2, 4, 6). Importantly, Ct-MOMP levels were also found to increase in these cells ( Figure S2A, lane 2, 4, 6) compared to cells not induced with IPTG ( Figure S2B, lane 1, 3, 5), which suggests that the favorable impact of Bam overexpression on Ct-MOMP biogenesis is not specific and can be achieved in relevant vaccine strains. To exclude that the observed effects were strain-specific for E. coli BL21(DE3), we tested the influence of Bam overexpression on Ct-MOMP biogenesis in different E. coli and Salmonella Typhimurium backgrounds. E. coli DH5α was chosen as representative of the E. coli K12 lineage, while BL21(DE3)-derivative ClearColi(DE3) [35] and Salmonella Typhimurium SL3261-∆msbB [36] were selected because these strains express detoxified versions of lipopolysaccharide, which is relevant for bacteria-based vaccine development. Induction of the Bam complex from pJH114 with IPTG resulted in a Coomassie-detectable increase in BamA levels ( Figure S2, lanes 2, 4, 6). Importantly, Ct-MOMP levels were also found to increase in these cells ( Figure S2A, lane 2, 4, 6) compared to cells not induced with IPTG ( Figure S2B, lane 1, 3, 5), which suggests that the favorable impact of Bam overexpression on Ct-MOMP biogenesis is not specific and can be achieved in relevant vaccine strains.

The Complete Bam Complex Is Needed for Optimal Ct-MOMP Expression
Within the Bam complex, the BamA subunit has a central role in folding and integrating OMPs into the OM lipid bilayer [37], whereas BamBCDE subunits have regulatory functions [38]. To investigate which of these roles are most relevant for improving Ct-MOMP biogenesis, we monitored cellular Ct-MOMP levels upon overexpression of either the BamA or BamBCDE subunits and compared these to cells overexpressing Bam units in three independent experiments. As a negative control to the Bam complex, the unrelated β-barrel OMP PhoE was expressed together with Ct-MOMP. IPTG-induced expression of BamA, BamBCDE, or Bam subunits, as well as PhoE was confirmed by analysing whole-cell samples by SDS-PAGE and Coomassie staining ( Figure S4). Notably, the levels of BamA appeared somewhat lower when expressed by itself (e.g., Figure S4, cf lanes 2 and 4), which may relate to the stabilizing effect of BamD on BamA [38]. Semi-quantitative immunoblotting on whole-cell samples ( Figure S5) was performed to compare Ct-MOMP expression levels in the different co-expression conditions (Figure 2A). Levels of Ct-MOMP were normalized against those of RpoB (DNA-directed RNA polymerase subunit β), a cytosolic protein that is expected to be insensitive to variations in Bam expression

The Complete Bam Complex Is Needed for Optimal Ct-MOMP Expression
Within the Bam complex, the BamA subunit has a central role in folding and integrating OMPs into the OM lipid bilayer [37], whereas BamBCDE subunits have regulatory functions [38]. To investigate which of these roles are most relevant for improving Ct-MOMP biogenesis, we monitored cellular Ct-MOMP levels upon overexpression of either the BamA or BamBCDE subunits and compared these to cells overexpressing Bam units in three independent experiments. As a negative control to the Bam complex, the unrelated β-barrel OMP PhoE was expressed together with Ct-MOMP. IPTG-induced expression of BamA, BamBCDE, or Bam subunits, as well as PhoE was confirmed by analysing whole-cell samples by SDS-PAGE and Coomassie staining ( Figure S4). Notably, the levels of BamA appeared somewhat lower when expressed by itself (e.g., Figure S4, cf lanes 2 and 4), which may relate to the stabilizing effect of BamD on BamA [38]. Semi-quantitative immunoblotting on whole-cell samples ( Figure S5) was performed to compare Ct-MOMP expression levels in the different co-expression conditions (Figure 2A). Levels of Ct-MOMP were normalized against those of RpoB (DNA-directed RNA polymerase subunit β), a cytosolic protein that is expected to be insensitive to variations in Bam expression ( Figure S5, see Supplementary Document 2). In line with the visually observed differences, a considerable~4-fold improvement of Ct-MOMP expression was found upon co-expression of the full Bam complex compared to cells only induced for Ct-MOMP (Figure 2A, compare Bam+ to PhoE−). Co-expression of only BamA also resulted in significantly higher levels of Ct-MOMP but to a lesser extent compared to the effect of full Bam complex co-expression. The difference could be due to the lower expression of the orphan BamA subunit as discussed above. Co-expression of BamBCDE had a more moderate effect, despite the efficient expression of these subunits in the absence of BamA (Figure 2A, compare BamBCDE+ to Bam+). In any case, the combined data indicate that both the accessory subunits and BamA contribute to Ct-MOMP expression, but the complete Bam complex yields the most pronounced effect.
higher levels of Ct-MOMP but to a lesser extent compared to the effect of full Bam complex co-expression. The difference could be due to the lower expression of the orphan BamA subunit as discussed above. Co-expression of BamBCDE had a more moderate effect, despite the efficient expression of these subunits in the absence of BamA (Figure 2A, compare BamBCDE+ to Bam+). In any case, the combined data indicate that both the accessory subunits and BamA contribute to Ct-MOMP expression, but the complete Bam complex yields the most pronounced effect. Overexpression of periplasmic proteins and OMPs may induce cell envelope stress [39], and may have resulted in the moderately decreased growth observed for cultures induced for Bam and MOMP ( Figure S3). To test whether Bam overexpression induces cell envelope stress, we monitored its effect on the production of the chaperone/protease DegP, which is upregulated under several envelope stress conditions [33]. Indeed, a clear IPTG-dependent increase of DegP expression was observed in pJH114-containing cells ( Figure S3). This prompted us to rule out that enhanced Ct-MOMP production is a consequence of secondary stress-related effects rather than of Bam overexpression itself. Overexpression of the unrelated OM protein PhoE also induced upregulation of DegP, similar to BamA and the complete Bam complex ( Figure 2B, S4, S6), but did not promote expression of Ct-MOMP (Figure 2A, S4, S5). In fact, Ct-MOMP expression was even lower than in cells not induced for PhoE expression (Figure 2A, compare PhoE+ to PhoE-), which could be caused by DegP-mediated degradation of Ct-MOMP intermediates accumulating in the periplasm ( Figure 2B, compare PhoE+ to PhoE-). Taken together, the data support a scenario in which co-overexpression Bam directly improves the expression of Ct-MOMP.  Figure S5, Figure S6 was shown. Ct-MOMP (A) and DegP (B) expression levels in cells co-expressing BamA, BamBCDE, and PhoE were calculated relative to those of cells co-expressing the Bam complex (Bam+). Quantification of Ct-MOMP and DegP levels was normalized for quantification of RpoB levels in the same sample. Measurements were based on three independent experiments. Statistical differences were calculated using Student's t-test for unpaired means. "*": p < 0.05; "**": p < 0.01.

Localization of Ct-MOMP in the OM
Overexpression of periplasmic proteins and OMPs may induce cell envelope stress [39], and may have resulted in the moderately decreased growth observed for cultures induced for Bam and MOMP ( Figure S3). To test whether Bam overexpression induces cell envelope stress, we monitored its effect on the production of the chaperone/protease DegP, which is upregulated under several envelope stress conditions [33]. Indeed, a clear IPTG-dependent increase of DegP expression was observed in pJH114-containing cells ( Figure S3). This prompted us to rule out that enhanced Ct-MOMP production is a consequence of secondary stress-related effects rather than of Bam overexpression itself. Overexpression of the unrelated OM protein PhoE also induced upregulation of DegP, similar to BamA and the complete Bam complex ( Figures 2B, S4 and S6), but did not promote expression of Ct-MOMP (Figures 2A, S4 and S5). In fact, Ct-MOMP expression was even lower than in cells not induced for PhoE expression (Figure 2A, compare PhoE+ to PhoE−), which could be caused by DegP-mediated degradation of Ct-MOMP intermediates accumulating in the periplasm ( Figure 2B, compare PhoE+ to PhoE−). Taken together, the data support a scenario in which co-overexpression Bam directly improves the expression of Ct-MOMP.

Localization of Ct-MOMP in the OM
To examine whether recombinant Ct-MOMP was properly delivered in the OM, we subjected E. coli cultures expressing Ct-MOMP to subcellular fractionation. Initially, we tested whether the protein is present in crude cell envelopes, comprising both the bacterial inner and outer membranes ( Figure 3). Cells expressing Ct-MOMP either in the absence or presence of induced Bam complex were lysed and subjected to differential centrifugation followed by analysis of the fractions by SDS-PAGE and Coomassie staining ( Figure 3A). The major OMP OmpF/C and OmpA were almost exclusively detected in the cell membranes ( Figure 3A; compare lanes 9-10 to lanes 7-8), validating the fractionation procedure. Likewise, although BamD and BamE were difficult to detect by Coomassie staining, BamA, -B, and -C subunits localized to the cell envelopes ( Figure 3A, lanes 9-10), indicating a proper localization of the Bam components even when overexpressed, consistent with the earlier studies [40,41]. Ct-MOMP localization was monitored by immunoblotting ( Figure 3B). The protein did not appear to be aggregation-prone as it was predominantly recovered in the low-speed supernatant, irrespective of co-expression of the Bam complex ( Figure 3B, lanes 3-4). Upon subsequent high-speed centrifugation, Ct-MOMP was almost exclusively detected in the pellet fraction indicative of membrane association ( Figure 3B, lanes 9-10). ent, the increased levels of Ct-MOMP in cells co-overexpressing Bam also largely co-fractionated with BamA with a slight peak in fractions 10-12 ( Figure 4B). Importantly, it was hardly detected in the fractions containing the inner membrane marker YidC ( Figure 4B, fractions 2-4). Virtually no Ct-MOMP was found at the bottom of the gradient ( Figure 4B, fraction 17), confirming that the recombinant protein did not form large aggregated structures. In conclusion, the recombinant Ct-MOMP predominantly localized in the OM regardless of Bam co-overexpression. Next, we wished to verify the localization of Ct-MOMP in the OM. Cell envelopes isolated from cells co-expressing Ct-MOMP with (Bam+) or without (Bam−) the Bam complex were subjected to isopycnic sucrose gradient centrifugation to separate the inner and outer membranes [42] (Figure 4). Analysis of the fractions by SDS-PAGE and Coomassie staining showed that OM marker proteins OmpF/C and OmpA peaked in fractions 10-12 irrespective of co-expression of the Bam complex ( Figure 4A). Immunoblotting of the fractions further demonstrated that the integral inner membrane protein YidC was mostly present in fractions 2-4 ( Figure 4B). As expected, endogenous BamA peaked in fractions 10-12 together with OmpF/C and OmpA (Figures 4A and S7). Induced Ct-MOMP from cells not induced for Bam overexpression also peaked in fractions 10-12, suggesting localization with BamA in the OM ( Figure 4A). Though distributed more broadly over the gradient, the increased levels of Ct-MOMP in cells co-overexpressing Bam also largely co-fractionated with BamA with a slight peak in fractions 10-12 ( Figure 4B). Importantly, it was hardly detected in the fractions containing the inner membrane marker YidC ( Figure 4B, fractions 2-4). Virtually no Ct-MOMP was found at the bottom of the gradient ( Figure 4B, fraction 17), confirming that the recombinant protein did not form large aggregated structures. In conclusion, the recombinant Ct-MOMP predominantly localized in the OM regardless of Bam co-overexpression.
To analyze the protein content of OM fractions isolated from cells overexpressing or not overexpressing the Bam complex (Bam+ and Bam-, respectively) in more detail, the membranes from fractions 10-12 of the respective sucrose gradients were pooled, collected, and reanalyzed by SDS-PAGE and silver staining ( Figure 5A). While similar levels of OmpF/C and OmpA were detected in both samples, elevated levels of BamA, -B and -C were clearly detected in the Bam+ samples. Interestingly, in the~41 kDa size range, novel protein bands emerged in the Bam+ sample ( Figure 5A, lane 2). The identity of the lower band (*) remains unclear, but immunoblotting confirmed that the upper band represented Ct-MOMP ( Figure 5B, lane 2). Remarkably, immunoblotting detection using Ct-MOMP antiserum revealed a second band at~60-65 kDa. This band may represent an SDS-resistant multimeric form of Ct-MOMP, which also has been observed for recombinant Ct-MOMP produced in nanodiscs [43]. Moreover, a protein product with very similar electrophoretic behaviour was described to represent a compact trimeric form of native Ct-MOMP when purified from EBs [16]. In any case, the data demonstrate the production of a substantial, silver stain-detectable amount of Ct-MOMP in the E. coli OM, but only upon co-expression of the Bam complex.
of OmpF/C and OmpA were detected in both samples, elevated levels of BamA, -B and -C were clearly detected in the Bam+ samples. Interestingly, in the ~41 kDa size range, novel protein bands emerged in the Bam+ sample ( Figure 5A, lane 2). The identity of the lower band (*) remains unclear, but immunoblotting confirmed that the upper band represented Ct-MOMP ( Figure 5B, lane 2). Remarkably, immunoblotting detection using Ct-MOMP antiserum revealed a second band at ~60-65 kDa. This band may represent an SDS-resistant multimeric form of Ct-MOMP, which also has been observed for recombinant Ct-MOMP produced in nanodiscs [43]. Moreover, a protein product with very similar electrophoretic behaviour was described to represent a compact trimeric form of native Ct-MOMP when purified from EBs [16]. In any case, the data demonstrate the production of a substantial, silver stain-detectable amount of Ct-MOMP in the E. coli OM, but only upon co-expression of the Bam complex.   Figure 4) were pooled and subjected to ultracentrifugation to collect the OMs. After washing, the membranes were analyzed by SDS-PAGE and silver staining (A) or immunoblotting using Ct-MOMP antiserum (B). The same samples were used for analyzing the heatmodifiability of Ct-MOMP. Membranes were solubilized in the presence of 0.1% (w/v) SDS to obtain semi-native samples or 2% (w/v) with or without 4 M urea to obtain denatured samples. Samples were kept at room temperature (RT) or heated (98 o C) prior to analysis by semi-native SDS-PAGE followed by Coomassie staining (C) and immunoblotting (D). A silver-stained band at ~41 kDa representing Ct-MOMP that could be identified through immunoblotting was indicated (Ct-MOMP). A similar-sized silver-stained product that was not detected by Ct-MOMP antiserum (*) and the ~60-65 kDa band potentially representing a higher-order form of Ct-MOMP was also annotated (Ct-MOMP+). In panel C, bands representing folded BamA (fBamA) and denatured unfolded BamA (uBamA) were indicated.

Heat-Modifiability of Ct-MOMP
In general, the folding state of a β-barrel membrane protein can be assessed using semi-native SDS-PAGE [44,45]. A β-barrel OMP usually remains folded even in the presence of low concentrations of SDS, provided that the samples are not heated. This results in an altered, usually faster, migration of the folded protein in the gel when compared to a heated, fully denatured sample. For native Ct-MOMP such a heat-modifiability assay previously resulted in a more complicated pattern. Bands at ~41 kDa represented stable and denatured forms of Ct-MOMP monomer while a ~60-65 kDa band was likely to represent stable trimers with also less-stable multimers running higher position in the gel [16]. To test the heat-modifiability of recombinant Ct-MOMP expressed in E. coli  Figure 4) were pooled and subjected to ultracentrifugation to collect the OMs. After washing, the membranes were analyzed by SDS-PAGE and silver staining (A) or immunoblotting using Ct-MOMP antiserum (B). The same samples were used for analyzing the heat-modifiability of Ct-MOMP. Membranes were solubilized in the presence of 0.1% (w/v) SDS to obtain semi-native samples or 2% (w/v) with or without 4 M urea to obtain denatured samples. Samples were kept at room temperature (RT) or heated (98 • C) prior to analysis by semi-native SDS-PAGE followed by Coomassie staining (C) and immunoblotting (D). A silver-stained band at~41 kDa representing Ct-MOMP that could be identified through immunoblotting was indicated (Ct-MOMP). A similar-sized silver-stained product that was not detected by Ct-MOMP antiserum (*) and the~60-65 kDa band potentially representing a higher-order form of Ct-MOMP was also annotated (Ct-MOMP+). In panel C, bands representing folded BamA (fBamA) and denatured unfolded BamA (uBamA) were indicated.

Heat-Modifiability of Ct-MOMP
In general, the folding state of a β-barrel membrane protein can be assessed using seminative SDS-PAGE [44,45]. A β-barrel OMP usually remains folded even in the presence of low concentrations of SDS, provided that the samples are not heated. This results in an altered, usually faster, migration of the folded protein in the gel when compared to a heated, fully denatured sample. For native Ct-MOMP such a heat-modifiability assay previously resulted in a more complicated pattern. Bands at~41 kDa represented stable and denatured forms of Ct-MOMP monomer while a~60-65 kDa band was likely to represent stable trimers with also less-stable multimers running higher position in the gel [16]. To test the heat-modifiability of recombinant Ct-MOMP expressed in E. coli BL21(DE3), we incubated the pooled OM fractions from the sucrose gradients in different loading buffers (0.1% SDS or 2% SDS). All samples were kept at RT or heated (98 • C) prior to loading on semi-native SDS-PAGE. The gels were either stained with Coomassie or subjected to immunoblotting to detect Ct-MOMP ( Figure 5C,D). The Coomassie-stained gels showed that BamA in the OM fractions was clearly heat-modifiable, ( Figure 5C, lanes 1-4), with folded forms of BamA (fBamA) disappearing upon heat treatment in favour of unfolded forms (uBamA), similar to the previous reports [46,47]. In all samples, regardless of whether the Bam complex was overexpressed or not, a major Ct-MOMP band running at~41 kDa did not shift upon heating of the 2% SDS samples ( Figure 5D, lanes 1-4). Additionally, in the heated OM sample of Bam+ cells, the band of~60-65 kDa appeared ( Figure 5D, lane 3), possibly representing the stable trimer revealed previously for native Ct-MOMP [16]. In the non-heated OM sample with 0.1% SDS, this band was much less prominent, perhaps because it resolved into higher-order complexes detected as a smear high up in the gel ( Figure 5D, lane 4), consistent with the previous study [43].
In an attempt to fully denature the~60-65 kDa band, we incubated the OM fractions in loading buffer containing both 2% SDS and 4 M of urea, then heated these samples ( Figure 5D, lanes 5-6). As expected, the addition of urea resulted in the disappearance of the Ct-MOMP+ band. Strikingly, heating in 4 M urea also changed the migration of the monomeric Ct-MOMP bands to a slightly higher position ( Figure 5D, compare lanes 5-6 to lanes 1-4), which could reflect further denaturation of the monomeric form also observed for native Ct-MOMP [16].

Ct-MOMP Is Exposed at the Surface of E. coli
When fully integrated into the OM, Ct-MOMP is expected to expose several loops to the extracellular milieu [48]. To assess surface exposure of heterologous Ct-MOMP in E. coli BL21(DE3) co-overexpressing Ct-MOMP and the Bam complex, cells were subjected to immunofluorescence microscopy using Ct-MOMP-specific antiserum ( Figure 6A). A clear circumferential, punctuate labeling of intact cells was observed, indicating Ct-MOMP indeed exposed domains at the cell surface. Interestingly, similar to Ct-MOMP, the Bam complex has been described to localize in bright distinct foci in immunofluorescence studies [49,50]. Permeabilization of the cells with Triton X-100 and lysozyme prior to the antiserum incubation yielded a more homogenous circumferential labeling, which could indicate improved access of the antiserum to also intracellular parts of Ct-MOMP. As a further control for cell integrity, we performed labeling of lipoprotein BamB, which is located on the periplasmic side of the OM and thus not accessible to antisera when cells are intact. Indeed, BamB could not be detected with BamB antiserum, unless cells were first permeabilized ( Figure 6B). In conclusion, we demonstrated that Ct-MOMP inserted into the E. coli OM exposes domains at the cell surface.
to the antiserum incubation yielded a more homogenous circumferential labeling, which could indicate improved access of the antiserum to also intracellular parts of Ct-MOMP. As a further control for cell integrity, we performed labeling of lipoprotein BamB, which is located on the periplasmic side of the OM and thus not accessible to antisera when cells are intact. Indeed, BamB could not be detected with BamB antiserum, unless cells were first permeabilized ( Figure 6B). In conclusion, we demonstrated that Ct-MOMP inserted into the E. coli OM exposes domains at the cell surface. Figure 6. Ct-MOMP was exposed at the surface of E. coli. Cells co-expressing Ct-MOMP and Bam complex were fixed and subjected to indirect immunofluorescence microscopy using rabbit polyclonal antisera against Ct-MOMP (A) or BamB (B) and Alexa Fluor 488 labeling/conjugation of antiserum. The intact cells were permeabilized with Triton-X100 and lysozyme prior to antisera incubations when indicated. Corresponding phase-contrast microscopy fields were also shown. The scale bar was 10 μm. Images were processed using ImageJ software (http://rsb.info.nih.gov/ij/).

Discussion
The abundance and immunogenicity of Ct-MOMP make it the most attractive Chlamydia vaccine target despite its moderate conservation and despite the challenges to express and purify sufficient amounts of properly folded Ct-MOMP from native or heterologous hosts. Here, we show that the expression and localization of recombinant Ct-MOMP in the E. coli OM can be substantially improved by co-overexpression of the Bam complex, suggesting that Bam-mediated assembly of Ct-MOMP is a bottleneck in its biogenesis. Figure 6. Ct-MOMP was exposed at the surface of E. coli. Cells co-expressing Ct-MOMP and Bam complex were fixed and subjected to indirect immunofluorescence microscopy using rabbit polyclonal antisera against Ct-MOMP (A) or BamB (B) and Alexa Fluor 488 labeling/conjugation of antiserum. The intact cells were permeabilized with Triton-X100 and lysozyme prior to antisera incubations when indicated. Corresponding phase-contrast microscopy fields were also shown. The scale bar was 10 µm. Images were processed using ImageJ software (http://rsb.info.nih.gov/ij/).

Discussion
The abundance and immunogenicity of Ct-MOMP make it the most attractive Chlamydia vaccine target despite its moderate conservation and despite the challenges to express and purify sufficient amounts of properly folded Ct-MOMP from native or heterologous hosts. Here, we show that the expression and localization of recombinant Ct-MOMP in the E. coli OM can be substantially improved by co-overexpression of the Bam complex, suggesting that Bam-mediated assembly of Ct-MOMP is a bottleneck in its biogenesis.
Expression of recombinant Ct-MOMP in E. coli has shown to be problematic [32,51], which may be due to the intricate species-specific conformation of native Ct-MOMP. Its three-dimensional structure has not been solved to date, but sequence-based predictions suggested it folds into a 16-stranded β-barrel [15,52]. These models placed the immunogenic VDs of Ct-MOMP at the cell surface. We have used the recent AlphaFold protein structure prediction method [53] to model the native Ct-MOMP. Interestingly, and in contrast to the previous predictions, it yielded a structural model of high confidence that includes only 10 very long antiparallel β-strands, while still placing all VDs at the cell surface ( Figure S8A). Possibly, the extended β-barrel structure is related to the relatively long fatty acid chains of Chlamydia trachomatis lipid A, when compared to other Gram-negative bacteria [54]. This creates a relatively thick hydrophobic core in the OM, which may impact the nature of transmembrane anchoring and explain the longer β-strands. If correct, this is also expected to complicate the Bam-mediated assembly of heterologous Ct-MOMP as it is thought to involve an intermediate "hybrid-barrel" folding state [55]. The formation of a poorly matched interface of the nascent Ct-MOMP β-strands and the short β-strands that flank the lateral gate of the "insertase" BamA in E. coli may potentially slow down Ct-MOMP membrane integration. This, in turn, could induce cell-envelope quality control mechanisms, as suggested by the increased levels of DegP detected, triggering degradation of nascent Ct-MOMP. Overproduction of the Bam complex may overcome titration of available Bam complexes handling recombinant Ct-MOMP resulting in improved assembly rates. A similar beneficial effect of the Bam complex overproduction was recently reported for the surface display of difficult-to-secrete chimeric autotransporter constructs that also obstruct the Bam complex [28].
An additional problem in recombinant expression may be posed by the 9 cysteines in Ct-MOMP, an unusually high number for an OMP. In Chlamydia trachomatis they form intermolecular disulfide bonds that contribute to a supportive connected network of OMPs in the cell envelopes know as the COMC [56]. However, in the absence of cognate partner proteins and in the presence of a non-cognate oxidoreductase system, non-native disulfide bonding of these cysteines may lead to misfolding and consequent degradation in the E. coli periplasm. The substitution of the cysteines for alanines has been shown to improve Ct-MOMP production to some extent [32], but our results, obtained with a cysteineless variant, indicate that the OM integration is still a bottleneck for this variant. Of note, an AlphaFold model for this variant closely resembles that of non-mutated Ct-MOMP ( Figure S8B). If the model is correct, the relatively long antiparallel β-strands may also explain the harsh conditions that appear required for full denaturation of the integrated Ct-MOMP.
It is difficult to compare the extent of Ct-MOMP expression, localization, folding, and surface display presented here with other attempts to improve recombinant Ct-MOMP expression as different assays for detection have been applied. In general, expression of recombinant Ct-MOMP has benefited from the use of an E. coli signal peptide [32], tailored growth conditions, optimal tuning of protein expression by codon harmonization [57] and the use of well-regulatable promoters such as the L-rhamnose-inducible rhaBAD promoter applied in this study. Conceivably, the gradual expression of Ct-MOMP leads to an optimal use of the Sec-translocon, which is known to be a potential bottleneck for membrane protein production [58]. Alternative reported approaches to increase production of Ct-MOMP antigen include cell-free expression in nanolipoprotein particles [43] and grafting VDs of Ct-MOMP in another β-barrel OMP, PorB of Neisseria lactamica [17,18]. Parallel efforts are directed towards the high-level expression of non-native Ct-MOMP (lacking a signal peptide) that can be relatively easily produced in the cytosol of E. coli. For example, the engineered version of Ct-MOMP (CTH522) that comprises VDs of different serovars may lack conformational B-cell epitopes but still generates neutralizing antibodies and cellular immune responses [59,60].

Conclusions
In conclusion, co-overexpression of the Bam complex is a promising strategy to increase the expression of challenging heterologous OMPs in the E. coli OM. This may prove useful for the purification of native OMP antigens for vaccine development or for structural and functional analysis. Furthermore, the inclusion of substantial amounts of native antigens in the OM may enable the generation of recombinant live or OMV-based vaccines. From that perspective, it is encouraging that Ct-MOMP expression in an attenuated Salmonella Typhimurium strain suitable for vaccine production can be achieved using this strategy. Future efforts will be directed towards the development of OMV-based vaccines that include Ct-MOMP and preferably also other Chlamydia trachomatis OMPs.

Reagents, Enzymes, and Sera
Plasmid DNA Miniprep and MACHEREY-NAGEL NucleoSpin Gel, PCR Clean-up gel extraction kits were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Phusion High Fidelity DNA polymerase and DNA restriction enzymes were bought from New England Biolabs (Ipswich, MA, USA). In-Fusion HD Plus Kit for seamless DNA cloning was obtained from Takara Bio (Mountain View, CA, USA). Coomassie Blue G250 was from BioRad (Hercules, CA, USA). Silver staining kit was obtained from Thermo Fisher Scientific (Waltham, MA, USA). Lumi-Light substrate for immunoblotting was purchased from Roche (Mannheim, Germany). All other reagents and chemicals were supplied by Sigma-Aldrich (Saint Louis, MO, USA).
Rabbit antiserum against Ct-MOMP was kindly provided by Frank Follmann (Statens Serum Institut, Copenhagen, Denmark). Rabbit antiserum against DegP was a gift of Jon Beckwith, Harvard Medical School, USA. The YidC polyclonal antiserum was raised in rabbits against a peptide that consisted of the 17 C-terminal amino acids of YidC by Agrisera (Umea, Sweden) [63]. Mouse antiserum against E. coli RpoB was bought from BioLegend (San Diego, CA, USA). Rabbit antisera against Bam, BamA and BamB were provided by Tanneke den Blaauwen (University of Amsterdam, Amsterdam, The Netherlands). Alexa488 goat anti-rabbit IgG was obtained from Life Technologies (Willow Creek Rd, Eugene). Anti-mouse and anti-rabbit peroxidase IgG antisera generated in goats were ordered from Rockland (Limerick, PA, USA).
The pJH114 plasmid [34] encoding the ABCDE subunits of the Bam complex under control of the IPTG-inducible trc promoter was kindly provided by Harris Bernstein (Bethesda, USA). To generate versions of pJH114 expressing either the BamA or BamBCDE subunits, BamA and BamBCDE encoding DNA sequences were amplified by PCR using pJH114 as the template and primers 3-4 and primers 5-6, respectively (see Supplementary Document 1- Table S1). PCR fragments were ligated into NdeI/XbaI-digested pJH114 using In-Fusion cloning so as to yield pTrc-BamA and pTrc-BamBCDE. A DNA sequence encoding E. coli K12 PhoE was cloned under lacUV5 promoter control into vector pEH1 [64], yielding plasmid pEH1-PhoE. All constructs were confirmed by sequencing (Macrogen, Amsterdam, The Netherlands).

General Protein Production and Analysis
E. coli BL21(DE3) cells harboring pLemo-MOMP were cultured until an OD 600 of 0.4-0.5, at which time expression of Ct-MOMP was induced by adding 0, 2, 4, 8 mM L-rhamnose and growth was continued for 2 h. For co-expression, cells harboring pLemo-MOMP and any of pJH114, pTrc-BamA, pTrc-BamBCDE or pEH1-PhoE, were induced with 100 µM IPTG at an OD 600 of 0.4-0.5 for production of the Bam complex, Bam subunits (BamA or BamBCDE) or PhoE (negative control), respectively. Cultures were grown for 1 h, after which expression of Ct-MOMP was induced by adding L-rhamnose to a final concentration of 8 mM, and further grown for 2 h. Then, cells and culture media were separated by centrifugation (10,000× g, 10 min, 4 • C). Cell pellets were resuspended in an ice-cold lysis buffer (5 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM EDTA). Next, the cells were disrupted by tip sonication (Branson Sonifier 250), yielding whole-cell lysates. Whole-cell lysates were analysed by SDS-PAGE and Coomassie Blue G250/silver staining or immunoblotting. Densitometric scanning on Coomassie-stained or silver-stained gels was conducted using a Molecular Imager GS800 Calibrated Densitometer from Life Science (Bulletin, USA). Immunoblots were imaged using an AI600 Imager from GE Healthcare (Canton, MA, USA).

Subcellular Fractionation
The cell pellets were resuspended in a lysis buffer (5 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM EDTA) plus the complete protease inhibitor from Roche (Mannheim, Germany) to a working concentration (a tablet per 50 mL extraction solution). Cell suspensions were kept in ice for 15 min. Cells were broken by two passages through the One-Shot disruptor from Constant System Ltd. (Daventry, UK) at 1.9 kbar. Removal of cell debris was performed by centrifugation (10,000× g, 10 min, 4 • C). Soluble protein and membrane fractions were separated by ultracentrifugation (293,000× g, 60 min, 4 • C). The whole-cell lysates, low-speed supernatants/pellets, and high-speed supernatants/pellets were kept at −20 • C for further analysis.

Heat-Modifiability Assay
Heat-modifiability assay was performed essentially as previously described [65]. Shortly, membrane samples were dissolved in SDS-PAGE loading buffer (125 mM Tris-HCl, pH 6.8, 20% glycerol, 0.02% bromophenol blue) containing various amounts of SDS and urea. To obtain semi-native conditions SDS was included to a final concentration of 0.1% (w/v), whereas 2% SDS (w/v), with or without 4 M urea, was included for denatured samples. Samples were subsequently kept at room temperature or incubated at 98 • C for 10 min prior to loading on semi-native polyacrylamide gels. These gels were prepared as for SDS-PAGE, but with no SDS added to the running and stacking gels. The gels were run in a normal buffer (25 mM Tris-HCl, 250 mM glycine, 0.1% SDS) at 15 mA whilst cooling the gel-cassette to prevent heating and stained with Coomassie Blue G250 or immunoblotting for further analysis.

Statistical Analysis
Statistical analysis of Ct-MOMP expression was done using the GraphPad Prism software ver. 8.0. p < 0.05 "*" or p < 0.01 "**" was taken as significant. The error bars represent the standard error of mean (SEM).