Escherichia coli DH5α (Stratagene, Santa Clara, CA) and BL21 (DE3) (Novagen, Darmstadt, Germany) were used as cloning and protein expression hosts, respectively. The genetic information and cultured conditions for E. coli, NTHi 3655 and the isogenic mutants are summarized in Supporting Table S1 . Construction of p5 ^CTD-knockout (NTHi 3655Δp5^CTD ) and p5-transcomplementation (NTHi 3655Δp5::p5) mutants is described in Supporting Figure S1 and Table S2 . NTHi mutants were verified for surface display of P5 by flow cytometry using rabbit anti-peptide P5^Loop3 polyclonal antibodies (pAbs) (Genscript, Piscataway, NJ) (Thofte et al., 2021). The human alveolar epithelial cell line A549 (ATCC CCL-185™) (American Type Culture Collection (ATCC), Manassas, VA) was maintained in F12 medium supplemented with 2 mM L-glutamine, and 10% fetal calf serum (FCS) (Gibco, Life Technologies, Carlsbad, CA).

Bacterial membrane fractions and OMVs were isolated from mid and late-log phase cultures, respectively as described (Su et al., 2019; Jalalvand et al., 2022). Briefly, membrane fractions were pelleted from the supernatant of sonicated bacterial lysate at 150,000×g at 4°C for 1 hour. Bacterial lysates or membrane fractions were separated on 12% SDS-PAGE followed by western blotting. Anti-peptide P5^Loop3 pAb and horse radish peroxidase (HRP)-conjugated donkey anti-rabbit pAb (Abcam, Cambridge, UK) were used to detect P5 on the blots (Thofte et al., 2021). The OM fractions (from three replicated preparations) were subjected to proteomic analysis (Proteomics Biomedicum, Karolinska Institutet, Sweden) ( Table S3 and Supporting Methodology ). The yield and particle size of purified OMVs were determined by nanoparticle tracking analysis (NTA) with a NanoSight NS300 (Malvern Panalytical, Malvern, UK).

DNA fragment encoding P5^CTD (M233 to K359) was amplified from plasmid P5³⁶⁵⁵-pET16b (Thofte et al., 2021) using specific primers ( Table S2 ) and cloned into the expression vector pET26(b)+ (Novagen). His-tagged recombinant protein was expressed in E. coli BL21 (DE3) with 1 mM IPTG induction and purified as described (Su et al., 2013).

The assay was performed as described (Fan et al., 2007; Park et al., 2012). Briefly, 3 µM of His-tagged recombinant protein was incubated with 100 µg of peptidoglycan from E. coli K12 (Invivogen, Toulouse, France) in 100 µl of ice-cold binding buffer (10 mM Tris-Cl, 10 mM MgCl₂, 50 mM NaCl, pH6.8) at 4°C for 16 hours. Mixtures were pelleted at 350,000×g at 4°C for 1 hour. The supernatants were collected as unbound protein; the pellets were washed twice with the ice-cold binding buffer. Finally, the pellets were dissolved in 2% SDS. The supernatants and pellets were analysed on 12% SDS-PAGE and western blotting. His-tagged proteins were detected with HRP-conjugated anti-His pAb (Abcam).

Bacterial adherence to human epithelial cells was performed as described (Su et al., 2016). Briefly, cells grown to 85% confluency were infected with mid-log phase bacteria (OD₆₀₀ = 0.5) to achieve a multiplicity of infection (MOI) of 100, at 37°C and 5% CO₂ in F12 media without FCS for 30 min. Cells were washed with PBS and detached with trypsin-EDTA (Sigma-Aldrich, St. Louis, MO). Cell solutions were plated on chocolate agar and incubated at 37°C and 5% CO₂ for 16 hours to quantify adherent bacteria in colony-forming units (CFU).

The assay was performed by flow cytometry as described (Su et al., 2019). Mid-log phase (OD₆₀₀ = 0.5) bacteria (5×10⁷ CFU) were incubated with indicated concentrations of human fibronectin (Sigma-Aldrich) in 100 µl of 1% BSA at 37°C and 5% CO₂ for 1 hour. Samples were washed and pelleted by centrifugation. Bacteria-bound fibronectin was detected with rabbit anti-human fibronectin pAb (Dako, Glostrup, Denmark) and FITC-conjugated swine anti-rabbit pAb (Dako). Samples were analysed in a BD FACSVerse™ flow cytometer (Becton-Dickinson, Franklin Lakes, NJ).

The assay was performed as described (Su et al., 2013). Suspension of mid-log phase (OD₆₀₀ = 0.5) bacteria (1.5×10³ CFU) in 150 µl of DVBS-BSA buffer was incubated with 5% normal human serum (NHS) for indicated time points at 37°C. Samples were plated on chocolate agar and incubated at 37°C and 5% CO₂ for 16 hours.

The assay was performed as described (Mychack et al., 2019). Bacteria grown to mid-log phase (OD₆₀₀ = 0.5) were pelleted, washed, and resuspended in PBS to OD₆₀ = 1.0. Bacterial suspensions were 10-fold serially diluted in PBS to generate 10⁴-10⁹ CFU/ml, and 2 µl from each dilution was spotted on chocolate agar supplemented with 50 mM or 100 mM NaCl. Plates were incubated at 37°C and 5% CO₂ for 16 hours.

Broth microdilution (BMD) was performed in three independently repeated-experiments according to ISO 20776-1 on custom freeze-dried Sensititre plates (Thermo Scientific, Basingstoke, UK) using the EUCAST recommended broth for fastidious organisms (Mueller Hinton Fastidious (MH-F) broth) to estimate minimal inhibitory concentrations (MICs) of antibiotics. MIC endpoints after 16-20 hours of incubation at 35 ± 1°C were visually evaluated according to ISO and EUCAST recommendations.

We used GraphPad Prism^® 9.0 (GraphPad Software, La Jolla, CA) for statistical analyses. One- or two-way ANOVA tests were used as indicated. Differences were considered statistically significant at P ≤ 0.05.

Based on bioinformatic analysis, P5 of NTHi 3655 belongs to the OmpA family of proteins (InterPro entry number: IPR002368) ( Figures 1A , S2 ). An OmpA-like domain (IPR006665) was found at the C-terminal domain (CTD) of P5 (P5^CTD) that was located between residue M233 and K359. The P5^CTD shared 20.6-56.2% of sequence similarity (34.6-67.2% identity) with some peptidoglycan-associated lipoproteins (Pal) and the C-terminus of other OmpA proteins that are known to have an OmpA-like domain, and experimentally bind peptidoglycan ( Figure 1B ). While the protein sequence homology of CTDs among OmpA family proteins is generally poor, two key residues (i.e., D283 and R298) that were crucial for peptidoglycan binding of OmpA in Acinetobacter baumannii and E. coli were also found on the P5^CTD (Park et al., 2012; Samsudin et al., 2016).

The OmpA-like domain has a peptidoglycan-binding property via non-covalent interactions (Hancock et al., 1994; Park et al., 2012; Samsudin et al., 2016). We therefore hypothesize that P5 also might function in an OmpA-like domain-dependent peptidoglycan binding. We observed that purified His-tagged P5^CTD formed a complex with peptidoglycan, which was almost comparable to His-tagged P6 (positive control) ( Figures 1C, D ). P6 is a Pal with an OmpA-like domain and is known to bind peptidoglycan in NTHi (Karalus and Murphy, 1999; Berenson et al., 2005). P6 does not have a transmembrane beta-barrel domain as seen in OmpA. The negative control, His-tagged P4 of NTHi 3655 (Kemmer et al., 2001; Su et al., 2016) did not bind peptidoglycan. Taken together, our data demonstrates that the P5^CTD has a peptidoglycan-binding activity as predicted by in silico analysis thus may play a role in the OM integrity of NTHi.

The OmpA-like domain serves as an anchor between the OM layer and the peptidoglycan. The interaction is crucial for OM-cell wall stability, and the assembly of membrane and cell wall proteins (Hancock et al., 1994; Knowles et al., 2009; Park et al., 2012; Calmettes et al., 2015; Ortiz-Suarez et al., 2016; Samsudin et al., 2016; Graham et al., 2021; Mamou et al., 2022). Since we showed that P5^CTD of NTHi 3655 has peptidoglycan-binding activity, we hypothesized a possible role for P5 in the distribution of NTHi periplasmic and membrane protein (membrane proteome). To this end, we generated NTHi 3655Δp5^CTD that expressed CTD-deleted P5, and a P5-transcomplemented mutant (NTHi 3655Δp5::p5). In addition, NTHi 3655Δp5 that has a full deletion of P5 was included as a control for comparison to the deletion of P5^CTD (Thofte et al., 2021).

Comparative SDS-PAGE of bacterial whole cell lysates and membrane fractions revealed some differences in the intensity of a few protein bands among the NTHi 3655 wild-type and mutants ( Figures 2A, B ). NTHi 3655Δp5^CTD and Δp5 exhibited a roughly similar pattern of protein profiling; while a simpler similarity of protein profiling pattern was observed between the wild-type and Δp5::p5. In western blotting, wild-type P5 and CTD-truncated P5 were detected by anti-P5^Loop3 pAb in NTHi 3655 and Δp5^CTD , respectively. This indicated that the deletion of CTD did not hamper protein expression of the remaining extracellular-loops and the transmembrane parts of P5 in NTHi 3655Δp5^CTD . Any corresponding signal was not detected in NTHi 3655Δp5 that was used as a negative control. Interestingly, we observed reduced detection of P5 on the surface of Δp5^CTD compared to the wild-type by flow cytometry analyses ( Figure 2C ). The signal was, however, higher than the P5-knockout mutant, suggesting partial or reduced exposure of P5 extracellular structures on the surface of Δp5^CTD . Unexpectedly, Δp5::p5 also exhibited partial detection of P5 by flow cytometry compared to the wild-type. It is currently unclear what causes the partial exposure of P5 extracellular structures in Δp5::p5; it is not known if there are any spontaneous genetic mutations that could affect periplasmic post-translational modifications essential for OM insertion of P5 (Chen and Henning, 1996; Sklar et al., 2007). Since the transcomplemented strain could not fully restore the surface expression of P5, this strain was excluded from downstream experiments.

To globally understand how P5 is involved in regulating membrane protein organization, we primarily focused on analysing the proteomic composition of the membrane from Δp5 and compared with the wild-type strain. The bacterial membrane preparations used in this study contained mainly OM and periplasmic proteins. Based on the ranking of relative abundance, virulence factors such as major adhesins (i.e., High molecular weight adhesin (HMW) and Haemophilus adhesin protein (Hap)) and IgA protease were less represented while P2, PD, PF and PE were more abundant in the P5-knockout mutant, with respect to the wild-type strain ( Table S3 ) (Duell et al., 2016). We also noticed that periplasmic chaperones (SurA, Omp26 and DegP) and Bam complex subunit proteins (BamA, C, D and E) that are essential for proper folding and assembly of OM proteins, respectively, were less abundant in the mutant (Knowles et al., 2009; Calmettes et al., 2015; Mamou et al., 2022).

In conclusion, our results suggest that P5 has an influence on the proteomic composition of OM and periplasm in NTHi. P5 may help to maintain optimal interactions between the OM layer and the periplasmic peptidoglycan network that is crucial for cell envelope integrity and protein assembly.

The aberrant membrane proteome composition in NTHi upon the deletion of P5 is likely to affect some cellular functions that are crucial for NTHi pathogenesis (Erwin and Smith, 2007; Duell et al., 2016). Previous studies have shown that P5 deletion in different NTHi isolates results in defective adherence to host cells and increased serum sensitivity compared to the P5-expressing NTHi (Hill et al., 2001; Rosadini et al., 2014; Euba et al., 2015; Wong et al., 2016; Thofte et al., 2021).

Considering that NTHi 3655Δp5^CTD has a reduced exposure of P5 on the surface while showing almost similar membrane profiling on SDS-PAGE as the NTHi 3655Δp5, we wanted to investigate whether the P5^CTD truncation could also affect bacterial pathogenicity in a similar fashion as when P5 was fully deleted.

We therefore tested the bacteria by in vitro functional assays related to adherence to host epithelial cells and evasion of host immune system since these two mechanisms are crucial for NTHi persistence to establish infection in the respiratory tract (Duell et al., 2016). The adherence of NTHi 3655Δp5^CTD to the lower airway A549 epithelial cells and extracellular matrix protein (fibronectin) was almost comparable to NTHi 3655Δp5, which were significantly (P ≤ 0.001) reduced compared to the NTHi 3655 wild-type ( Figures 3A, B ). When exposed to 5% NHS for 10 min, deletion of P5^CTD rendered the Δp5^CTD mutant to be more serum sensitive compared to the wild-type ( Figure 3C ). However, the Δp5^CTD mutant survived better (53% of survival) than the NTHi 3655Δp5 which was almost fully killed (1.9% survival) when exposed to NHS. The data confirm that deletion of P5^CTD could impair NTHi pathogenicity similarly as the full deletion of P5. We concluded that P5^CTD has an important impact on bacterial pathogenesis.

Gram-negative bacteria with genetic knockouts of OmpA-like domain superfamily proteins (InterPro: IPR036737) are unable to establish bonds between the OM and peptidoglycan, leading to increased release of OMVs (Avila-Calderon et al., 2021). To evaluate whether the deletion of either CTD or full length P5 could affect the vesiculation of NTHi, naturally released OMVs were purified from the late-log phase bacterial culture and normalized by cell density. Both NTHi 3655Δp5^CTD and Δp5 produced extensive amounts of OMVs as compared to the NTHi 3655 ( Figure 3D ). Interestingly, NTHi 3655Δp5 was the most hypervesiculating strain, producing almost 2-fold more OMVs than the Δp5^CTD mutant ( Figure 3D ).

The increased OMV production suggested an OM instability among the Δp5^CTD and Δp5 mutants. To test this hypothesis, we performed an osmotic shock assay by growing bacteria on chocolate agar supplemented with different NaCl concentrations. When cultured on chocolate agar without additional NaCl or supplemented with 50 mM NaCl, similar to the wild-type, NTHi 3655Δp5^CTD and Δp5 mutants did not show any growth defect ( Figure 3E ). However, at a higher NaCl concentration (100 mM), NTHi 3655Δp5^CTD and Δp5 mutants were equally sensitive to the high salt hence showed defective growth, compared to the wild-type. In line with these results, the Δp5^CTD and Δp5 mutants equally displayed a 2- and 4-fold susceptibility to β-lactam antibiotics ampicillin (MIC=0.12 µg/ml) and imipenem (0.12 µg/ml), respectively compared to NTHi 3655 (MIC=ampicillin 0.25 µg/ml; imipenem 0.5 µg/ml). Collectively, our data indicate that the CTD of P5 is important in maintaining OM stability and integrity, hence bacterial vesiculation and beta-lactam antibiotics resistance.