Modulation of Haemophilus influenzae interaction with hydrophobic molecules by the VacJ/MlaA lipoprotein impacts strongly on its interplay with the airways

Airway infection by nontypeable Haemophilus influenzae (NTHi) associates to chronic obstructive pulmonary disease (COPD) exacerbation and asthma neutrophilic airway inflammation. Lipids are key inflammatory mediators in these disease conditions and consequently, NTHi may encounter free fatty acids during airway persistence. However, molecular information on the interplay NTHi-free fatty acids is limited, and we lack evidence on the importance of such interaction to infection. Maintenance of the outer membrane lipid asymmetry may play an essential role in NTHi barrier function and interaction with hydrophobic molecules. VacJ/MlaA-MlaBCDEF prevents phospholipid accumulation at the bacterial surface, being the only system involved in maintaining membrane asymmetry identified in NTHi. We assessed the relationship among the NTHi VacJ/MlaA outer membrane lipoprotein, bacterial and exogenous fatty acids, and respiratory infection. The vacJ/mlaA gene inactivation increased NTHi fatty acid and phospholipid global content and fatty acyl specific species, which in turn increased bacterial susceptibility to hydrophobic antimicrobials, decreased NTHi epithelial infection, and increased clearance during pulmonary infection in mice with both normal lung function and emphysema, maybe related to their shared lung fatty acid profiles. Altogether, we provide evidence for VacJ/MlaA as a key bacterial factor modulating NTHi survival at the human airway upon exposure to hydrophobic molecules.


Results
Generation and characterization of vacJ mutant strains in H. influenzae. Two-hundred and ninety eight available nontypeable H. influenzae whole genome sequences were downloaded and used to assess the presence of the Mla system through the TBLASTN tool by local BLAST. Gene homologs to those encoding the E. coli Mla system were shown to be present in all strains. Similar to E. coli, the vacJ gene and the mlaBCDFE operon occupy separate genomic locations (Fig. 1A). NTHi375 and RdKW20 strains were employed to generate vacJ mutants (vacJ gene accession numbers NF38_01540 and HI0718, respectively). The vacJ gene is predicted to encode a lipoprotein located at the bacterial outer membrane. VacJ NTHi375 and VacJ RdKW20 present a putative signal sequence (MKTKVILTALLSAIALTGC and MKTKTILTALLSAIALTGC, respectively) in amino acids 1-19, and a lipobox sequence (Leu-Thr-Gly-Cys) in amino acids [16][17][18][19]. At the protein level, VacJ NTHi375 and VacJ RdKW20 displayed 98.4% identity; VacJ NTHi presents homologs in other Gram-negative bacteria 20 .
NTHi mutants lacking the vacJ gene were selected on sBHI agar. These strains rendered normal size colonies (Fig. S1A). ΔvacJ mutants growth in sBHI was comparable to that of their respective isogenic wild type (WT) strains (Fig. 1B). Similar bacterial morphology was also observed for WT and mutant strains when assessed by transmission electron microscopy (TEM) (Fig. 1E). Deletion of the vacJ gene is known to impair outer membrane stability in several Gram negative bacteria including H. influenzae, leading to increased detergent sensitivity 17,[21][22][23] . In agreement, in the presence of sodium deoxycholate, both NTHi375 and RdKW20 vacJ mutants showed bacterial cell morphology changes, and a sodium deoxycholate dose dependent reduction of viability, which could be chromosomally complemented (Figs 1C-E and S1B).  influenzae growth in sBHI is not modified by vacJ gene mutation. Bacterial growth is shown for NTHi375 (left) and RdKW20 (right) WT and vacJ mutant strains. Growth in sBHI is shown as a mean of OD 636 at the indicated time points. Experiments were performed in triplicate, in three independent occasions (n = 9). (C-E) Sodium deoxycholate bactericidal effect on NTHi depends on VacJ. NTHi WT and ΔvacJ strains grown on chocolate agar were used to generate OD 600normalized bacterial suspensions for further incubation with sodium deoxycholate. After 20 min, deoxycholate dose dependent bactericidal effect was higher on the ΔvacJ mutant than on the WT strains, as measured by assessing bacterial growth on sBHI (C,D). In (C), bacterial counts are shown as log 10 c.f.u./ml (mean ± SE). For both NTHi375 and RdKW20, vacJ mutant viability was lower than WT strain at [deoxycholate] between 1 and 1.5 mg/ml (*p < 0.0005). Experiments were performed in duplicate and in three independent occasions (n = 6). Statistical comparison of the means was carried out using one-way ANOVA (Dunnett's multiple comparisons test). Moreover, it has been previously shown that mutation of the vacJ gene in H. influenzae strain R2866 increases the amount of ChoP on the LOS molecule 17 . NTHi375ΔvacJ and RdKW20ΔvacJ strains also showed significantly higher ChoP level than their respective WT strains (p < 0.005) when measured by flow cytometry with the murine monoclonal antibody TEPC-15, and this phenotype could be restored to WT levels in the ΔvacJ complemented strains (Fig. 1F). We asked if such ChoP increase could be related to the lic1A gene phase variation or to changes in lic1ABCD gene expression. The lic1A gene was sequenced in WT and vacJ mutant strains, showing the same number of repeats of the tetranucleotide (5′CAAT) n (Fig. S1C); similarly, lic1A gene expression was comparable between WT and vacJ mutant strains (Fig. 1G).
In summary, disruption of the vacJ gene did not modify NTHi morphology and growth under the conditions tested; differently, it affected NTHi outer membrane stability and increased bacterial surface decoration with ChoP, independently of the lic1A gene expression or phase variation.
Disruption of the vacJ gene modifies NTHi fatty acid content. The content of NTHi surface phospholipids has been shown to increase upon vacJ gene mutation 17 . Such modification may relate to variations in bacterial fatty acid composition, in terms of fatty acyl residues type and/or amount. Based on this notion, bacterial total fatty acid methyl ester composition was analysed by gas chromatography and mass spectrometry (GC-MS), showing increased content of total fatty acids in vacJ mutant compared to WT strains (for NTHi375 and RdKW20, p < 0.005) (Fig. S2A). The same type of fatty acyl residues, consisting of straight saturated C14:0 (myristic and 3-hydroxymyristic), C16:0 (palmitic), C18:0 (stearic) and unsaturated C16:1 (palmitoleic) acid was observed in all tested strains. ΔvacJ mutants displayed increased content of palmitic (for NTHi375 and RdKW20, p < 0.0005) and palmitoleic acid (for NTHi375 and RdKW20, p < 0.0005), compared to their WT isogenic strains. The content of palmitoleic shown by vacJ complemented strains was similar to that of the WT strains; this was not the case for palmitic acid content (for NTHi375 and RdKW20, p < 0.0005). No significant differences were observed for myristic, 3-hydroxymyristic (C14:0-3-OH) and stearic acid levels between WT and vacJ mutant strains ( Fig. 2A).
The H. influenzae lipid A disaccharide backbone is known to be composed of two 2-amino-2-deoxyglucose residues linked by a β-(1-6) glycosidic linkage and phosphorylated at positions 1 and 4′. The C2/C2′ and C3/C3′ positions are substituted by amide-linked and ester-linked myristic acid, respectively; fatty acid chains on C2′ and C3′ are further esterified by 14:0, and fatty acid chains on C2 and C3 are 3-hydroxymyristic acid 24 . We next assessed WT and vacJ mutant strains lipid A species by MALDI-TOF mass spectrometry. All strains showed a major lipid A species corresponding to the hexa-acylated form (m/z 1,824) 24,25 . The ion peak corresponding to m/z 1,744, likely to represent the monophosphorylated hexa-acylated lipid A, showed lower intensity in vacJ mutant than in WT strains; no differences among strains were observed for the ion peak with m/z 1,388, which may represent a tetra-acyl lipid A 24,25 (Fig. 2B).
The observed increased fatty acid content led us to assess bacterial surface hydrophobicity. Although we did not observe striking differences between isogenic WT and ΔvacJ strains, a slight but consistent tendency to increased bacterial surface hydrophobicity in the absence of the vacJ gene could be seen (Fig. S2C).
Together, our results show increased amount of total fatty acids and identify specific fatty acyl residues raised in NTHi vacJ mutant strains, i.e. palmitic and palmitoleic acid. Such fatty acid increase may relate to changes in bacterial phospholipid composition, without having a gross influence on lipid A acylation patterns, altogether supporting the notion that NTHi vacJ mutants may have increased surface hydrophobicity compared to their isogenic parental strains.
Inactivation of the vacJ gene lowers NTHi resistance to hydrophobic antibiotics. The observations above prompted us to speculate that changes in bacterial surface hydrophobicity may modify antibiotic susceptibility. We tested bacterial susceptibility to 12 hydrophilic antibiotics, showing no differences between WT and mutant strains (Table S1). In contrast, NTHi375 and RdKW20 vacJ mutants were more susceptible to the hydrophobic antibiotics erythromycin, rifampicin and azithromycin (Table 1). In this case, phenotypic restoration could not be performed because the complemented strains are resistant to erythromycin (see Methods section). Together, bacterial changes associated to vacJ mutation were found to contribute increasing NTHi susceptibility to hydrophobic antibiotics.
The vacJ gene mutation increases H. influenzae susceptibility to ionic antimicrobials. Changes in resistance to hydrophobic antibiotics may also be extended to other antimicrobials with hydrophobic regions. Polymyxins consist of a cyclic heptapeptide ring with a tripeptide side chain covalently bound to a fatty acid via (n = 12). Statistical comparison of the means with one-way ANOVA (Dunnett's multiple comparisons test) was carried out. (G) Comparable expression of the lic1A gene between WT and vacJ mutant strains, exponentially grown in sBHI. Experiments were performed in duplicate and in three times (n = 6). Statistical comparison of the means using two-tail t-test was performed.
SCIENTIfIC REPORTS | (2018) 8:6872 | DOI:10.1038/s41598-018-25232-y an acyl group. The polycationic peptide ring interacts with the lipid A, and the fatty acid portion interacts with the hydrophobic region of the outer membrane, driving loss of membrane integrity and bacterial cell death 26,27 . Following this notion, we next tested NTHi WT and mutant strains resistance to the antimicrobial peptide polymyxin E (PxE, also known as colistin). We observed increased PxE susceptibility for both ΔvacJ strains (MIC PxE of 0.032 mg/l), compared to that of the WT strains (for NTHi375, MIC PxE of 0.094 mg/l; for RdKW20, MIC PxE of 0.064 mg/l) ( Table 1).
Disruption of the vacJ gene in H. influenzae increases bacterial susceptibility to free fatty acids. Free fatty acids are considered as natural detergents; in fact, a bactericidal effect on H. influenzae has been described for arachidonic acid (C20:4) 10 . The results above led us to speculate that lacking the vacJ gene may further increase bacterial susceptibility to fatty acids. Based on this notion, we assessed the result of . Bacteria grown on chocolate agar were used to extract fatty acids, by following saponification, methylation, extraction and washing steps. vacJ mutants showed increased amounts of C16:0 and C16:1 compared to the WT strains (for NTHi375 and RdKW20, *p < 0.0005). ΔvacJ complemented strains showed amounts of C16:1 comparable to those of the WT strains; increased amounts of C16:0 were maintained in the ΔvacJ complemented strains (for NTHi375 and RdKW20, *p < 0.0005). Quantifications were performed in duplicate and in three independent occasions (n = 6). (B) Determination of NTHi lipid A species by MALDI-TOF for NTHi375 (left) and RdKW20 (right) WT and vacJ mutant strains. (C) TLC band corresponding to PE and PG phospholipids were scraped to analyse their fatty acid composition. Assays were performed in duplicate and in three independent occasions (n = 6) (means ± SE are shown). PG C16:0 content was higher in the vacJ than in the WT strains (for NTHi375, *p < 0.05; for RdKW20, **p < 0.005). PG C18:0 content was higher in the vacJ than in the WT strains (for RdKW20, *p < 0.05). PE C16:0 content was also higher in the vacJ than in the WT strains (for NTHi375, *p < 0.05; for RdKW20, ***p < 0.0001). Statistical comparisons of the means were carried out with two-way ANOVA (Sidak's multiple comparisons test).
SCIENTIfIC REPORTS | (2018) 8:6872 | DOI:10.1038/s41598-018-25232-y NTHi interaction with exogenous fatty acids. To do so, bacterial growth in a defined medium free of fatty acids (MM-FFA), where a carbon source can be added ad hoc, was first optimized. Medium composition and NTHi bacterial growth when using glucose 20 mM as carbon source are shown in Table S2 and Fig. 3A, respectively. NTHi is unlikely to degrade exogenous long-chain fatty acids due to the absence of β-oxidation, therefore excluding the use of such molecules as sole carbon and energy source by this pathogen. As expected, bacterial growth in MM-FFA could not be recorded when using either arachidonic acid or vehicle solution as carbon source; vacJ mutants showed similar behaviour to their respective WT strains (Fig. 3A). To further determine NTHi viability in the presence of free fatty acids, NTHi WT and vacJ mutant strains were inoculated in MM-FFA in the presence of arachidonic (C20:4) or oleic (C18:1) acid. Such fatty acid selection was based on their presence in the human lung [7][8][9] . NTHi375 and RdKW20 viability decreased in the presence of both fatty acids in a dose dependent manner. Arachidonic acid susceptibility was higher for RdKW20 than for NTHi375, leading us to assess non-identical concentrations of this fatty acid for both strain backgrounds. As expected, mutant strains lacking the vacJ gene displayed higher susceptibility to both arachidonic and oleic acid than their respective WT strains (NTHi375, arachidonic acid 25 µM, p < 0.0005; NTHi375, oleic acid 0.8 to 1.6 mM, p < 0.0005; 2 mM, p < 0.05; RdKW20, arachidonic acid 3 to 12.5 µM, p < 0.0005; 25 µM, p < 0.05; RdKW20, oleic acid 1.2 to 2 mM, p < 0.005). (Fig. 3B,C). Complete or partial restoration was obtained for the ΔvacJ complemented strains, for NTHi375 and both free fatty acids, and for RdKW20 and oleic acid (Fig. 3B, left panel; Fig. 3C).
Together, these results showed that NTHi does not grow using free fatty acids as sole carbon source; in turn, free fatty acids have a killing effect, further increased by inactivation of the vacJ gene.

VacJ contributes to NTHi interplay with human airway epithelial cells. Given the determinant role
of NTHi interplay with the human airway epithelium in the progression of infection 2,28 , we speculate that VacJ mediated changes on the bacterial surface may alter such interplay. We next asked whether the lack of VacJ alters this host cell-pathogen interaction by infecting NCI-H292 human bronchial epithelial cells with NTHi WT and vacJ mutant strains 29 . NCI-H292 cell invasion of ΔvacJ mutants was lower than that shown by the WT strains (for NTHi375 and RdKW20, p < 0.0005) (Fig. 4A). ΔvacJ lower epithelial infection rate was also observed during invasion of A549 human type II pneumocytes (for NTHi375, p < 0.05; for RdKW20, p < 0.0005) 30,31 (Fig. 4B, left and middle panels). Phenotypic restoration could not be achieved with the ΔvacJ complemented strains (Fig. 4). Differently, the vacJ gene mutation did not modify epithelial inflammatory response in terms of IL-8 secretion by A549 cells upon infection (Fig. 4B, right panel). In summary, VacJ is likely to contribute significantly to H. influenzae entry within airway epithelia.

Inactivation of the vacJ gene attenuates H. influenzae virulence in vivo.
A genome-wide screen for H. influenzae genes required in the lung previously revealed vacJ involvement in pathogenesis 18 . Following this observation, we sought to quantify the impact of vacJ disruption in a NTHi mouse respiratory infection model system previously used for NTHi375 29,31-34 . Mice were infected with NTHi375 WT and vacJ mutant strains, and bacterial loads quantified in lung and bronchoalveolar lavage fluid (BALF) samples at 24 and 48 h post-infection (hpi). At both post-infection time points, NTHi375ΔvacJ lung and BALF bacterial numbers were lower than those recovered for the WT strain (at 24 hpi, lung and BALF samples, p < 0.005 and p < 0.0001 respectively; at 48 hpi, p < 0.0001) (Fig. 5A).
The use of mice with normal lung function may limit modelling NTHi infection in the context of an emphysematous lung. When tested for NTHi strain 2019, murine lung treatment with elastase resulted in damage consistent with COPD/emphysema and impaired clearance of NTHi 35 . We next speculate that vacJ mutant attenuation may be favoured, among others, by lung fatty acids. Lung emphysema, where fatty acids seem to mediate airway inflammation 6,7 , could in turn amplify the attenuation effect of vacJ disruption. To elicit pulmonary damage consistent with emphysema, CD1 mice were treated with porcine pancreatic elastase delivered via nonsurgical intratracheal instillation 36 . Emphysema-related features in elastase-treated mice were confirmed by pulmonary function tests and micro-computed tomography (micro-CT) (Fig. S3). Namely, elastase-treated lungs displayed higher physiological compliance and lower resistance and elastance than vehicle-treated mice. These results suggest an increment in the lung capacity to expand and a reduction in the resistance to the airflow, as a result of the parenchymal degradation induced by the effect of elastase activity. Moreover, micro-CT image scans revealed lower X-ray density and swollen lungs in elastase-treated mice, which again correlates with a decrease in lung stiffness and collapsed lung parenchymal structures. All these findings are consistent with those previously reported 36 .  . In all strains, glucose allowed exponential growth. Ethanol and arachidonic acid did not render bacterial growth, similar to only MM-FFA conditions. Growth curves were performed in quadruplicate, three times (n = 12) (B) Arachidonic and (C) oleic acid have a bactericidal effect on NTHi. WT, ΔvacJ and ΔvacJ complemented strains grown on chocolate agar were used to generate normalized bacterial suspensions in MM-FFA, for further incubation with arachidonic or oleic acid. Results are expressed as percentage of bacterial survival (means ± SE), referred to that in the presence of vehicle solution. A dose dependent bactericidal effect was observed for both WT strains and fatty acids. Both vacJ mutants were more susceptible to the two fatty acids tested than their respective parental strains (NTHi375, arachidonic acid 25 µM, ***p < 0.0005; NTHi375, oleic acid 0.8 to 1.6 mM, ***p < 0.0005; 2 mM, *p < 0.05; RdKW20, arachidonic acid 3 to 12.5 µM, ***p < 0.0005; 25 µM, *p < 0.05; RdKW20, oleic acid 1.2 to 2 mM, **p < 0.005). Complete or partial restoration was observed for NTHi375 and both free fatty acids, and for RdKW20 and oleic acid; increased arachidonic acid killing was maintained for the RdKW20ΔvacJ complemented strain (arachidonic acid 3 to 12.5 µM, ***p < 0.0005). Survival assays were performed in duplicate and at least three times (n ≥ 6). Statistical comparisons of the means were performed with two-way ANOVA (Sidak's multiple comparisons test).
Overall, these results support that VacJ is a bacterial factor involved in NTHi pulmonary infection. Under the conditions tested, elastase-treated mice showed a subtle impairment of NTHi375 pulmonary clearance; induced emphysema did not modify the dynamics of NTHi375ΔvacJ clearance. The panel of saturated and unsaturated fatty acids quantified in murine lung samples was comparable for normal function and emphysematous lungs, which may contribute to the observed vacJ mutant attenuation in both mouse infection models. NCI-H292 bronchial epithelial cells and (B) A549 type II pneumocytes were used to quantify invasion by WT, ΔvacJ and ΔvacJ complemented strains. Inactivation of the vacJ gene triggered significantly lower entry into NCI-H292 (for NTHi375 and RdKW20, ***p < 0.0005) and A549 (for NTHi375, p < 0.05; for RdKW20, ***p < 0.0005) cells than that shown by the WT strains. Same observation was made for ΔvacJ complemented strains (NCI-H292 cells: for NTHi375 and RdKW20, ***p < 0.0005; A549 cells: for NTHi375, **p < 0.005; for RdKW20, ***p < 0.0005). IL-8 release was comparable in cells infected by NTHi375 WT and vacJ mutant strains. Cell invasion assays were carried out in triplicate, at least three times (n ≥ 9). IL-8 levels were quantified twice in two independent assays (n = 4). Means ± SD are represented and statistical comparisons of the means were performed using two-tail t-test.

Discussion
Lipids are important components of the host immunity by organizing membrane signalling complexes and releasing lipid-derived mediators. Changes in the lipid content of the airway epithelium play important roles in cystic fibrosis, COPD and asthma, and several lipid molecules are key molecular mediators of disease in these chronic respiratory conditions 6 . In fact, COPD is linked to dysregulation of many metabolic pathways including lipid biosynthesis, and reconstruction of a COPD sputum lipid signalling network indicates that arachidonic acid may be a critical and early signal distributer upregulated in this disease 37 . Likewise, pulmonary injury is a trademark of acute respiratory distress syndrome associated to oleic acid 8,9 . Chronic respiratory patients are often persistently infected at their lower airways by NTHi, which may encounter free fatty acids at this host niche. Free fatty acid antibacterial properties have been observed for several pathogens including group A streptococci, Neisseria gonorrhoeae, mycobacteria, Staphylococcus aureus and H. influenzae 10,[38][39][40][41][42] , but mechanisms counteracting their detergent effect have not been detailed. Outer membrane lipid asymmetry plays an essential role in Gram negative bacteria barrier function. Its maintenance contributes preserving surface hydrophilicity, further counteracting the antibacterial effect of hydrophobic antimicrobials such as fatty acids and numerous conventional antibiotics 14 . Three systems are known to prevent damage resulting from surface exposed phospholipids in Gram negative bacteria, including the phospholipase PldA, the palmitoyltransferase PagP and the VacJ-MlaBCDEF phospholipid transport system 15 . In this study, we elucidate the involvement of VacJ in NTHi respiratory infection, as part of the only system known to maintain the outer membrane lipid asymmetry found to be present in available genomes of this bacterial pathogen, likely to be part of its core genome. Based on the absence of gene homologs, our observations are likely to be unrelated to a concerted action of VacJ NTHi with PldA. Previous evidence indicates that VacJ is important for NTHi resistance to serum mediated killing via the classical pathway of complement activation 17 , which may relate to vacJ gene implication in lung pathogenesis 18 . Here, we link the vacJ gene deficiency with increased specific fatty acyl residues in bacterial global fatty acid and phospholipid composition, which may contribute to raise bacterial surface hydrophobicity therefore jeopardizing the outcome of NTHi interaction with hydrophobic and lipophilic molecules. Of note, similar outcomes were observed upon NTHi vacJ mutant interaction not only with hydrophobic antibiotics and synthetic antimicrobial peptides, but also with free fatty acids shown to be present in the host lungs, further strengthening the importance of VacJ to maintain NTHi membrane stability during its persistent interplay with the host at the human airways.
Fatty acid composition has been previously determined for H. influenzae RdKW20 purified outer membrane vesicles and outer membranes, showing abundance of myristic, palmitic, palmitoleic and stearic acid 43,44 . To our knowledge, this is the first report on NTHi total fatty acid composition, further determined in two genetically different strains. Our findings support previous observations on major fatty acyl species, and also highlight the presence of hydroxymyristic acid. This fatty acid is present in the lipid A molecule 24 , but absent in total, outer membrane vesicle and outer membrane phospholipids of NTHi strains ( Fig. 2 and 43 ), therefore narrowing its location at the lipid A. Likewise, phospholipid content and species composition in the PE fraction has been previously determined in NTHi purified outer membrane vesicles and outer membranes 43,44 . Here, we report NTHi phospholipid total content and fatty acyl composition, and show palmitic and stearic acid as shared species in both PG and PE. vacJ deletion has been shown to result in the asymmetric expansion of the outer leaflet, supporting the outward budding of the outer membrane to finally form outer membrane vesicles enriched in phospholipids with decreased palmitic and increased myristic acid amounts 43 . Our results showed that vacJ mutation increases total fatty acid content, and vacJ deficient strains showed higher amount of fatty acyl residues, including palmitic acid.
Besides VacJ involvement in NTHi membrane integrity and serum resistance 17 , it also contributes maintaining E. coli, Actinobacillus pleuropneumoniae, Haemophilus parasuis and Shigella flexneri membrane integrity 15,20,21,23 , A. pleuropneumoniae and H. parasuis serum resistance and biofilm formation 20,23 , and H. parasuis virulence in vivo 20 . Moreover, VacJ participates in H. parasuis epithelial adhesion and invasion, and in S. flexneri cell to cell spread 20,21 . In this study, we show that VacJ plays a role in NTHi airway epithelial infection, although a direct role for VacJ as a bacterial ligand could not be established. Next to chromosome complementation, the vacJ NTHi375 gene HA-tagged at the 3′ end, together with its predicted promoter region, was cloned and expressed in E. coli as a heterologous host; despite VacJ NTHi375 -HA expression, confirmed by western blot using a monoclonal mouse anti-HA antibody, infection of A549 cells was similar to that shown by control E. coli strain harbouring the empty plasmid (Fig. S4). We cannot exclude that defective cell infection could be an indirect effect of vacJ disruption, causing outer membrane alterations not only in phospholipids and fatty acids but also in bacterial ligands, which may altogether hinder complementation. The reason for this unsuccessful complementation is currently unknown. Operon prediction by using the MicrobesOnline tool 45 showed that the vacJ gene is unlikely to be part of an operon in H. influenzae, therefore excluding possible polar effects. In fact, phenotypic restoration was successful when testing bacterial deoxycholate survival and ChoP levels, and partial when testing fatty acid composition and fatty acid survival, further supporting the complementation approach employed in this study.
LOS decoration with ChoP is known to relate to NTHi resistance to antimicrobial peptides 46 . Unexpectedly, vacJ gene deficiency was shown to increase NTHi ChoP levels ( Fig. 1F and 17 ) but also antimicrobial susceptibility, therefore excluding a VacJ-mediated relationship between bacterial ChoP decoration and resistance to PxE. This observation could relate to the lower intensity of the ion peak representing the monophosphorylated hexa-acylated lipid A observed in the vacJ mutant strains (Fig. 2B). Likewise, the vacJ gene is known to play a role in resistance to oxidative stress in Campylobacter jejuni 47 . Oxidative stress occurs in the small airways, lung parenchyma and alveolar regions in COPD; in asthma, the larger airways are its major site of action. Different to C. jejuni, the diameter of NTHi bacterial growth inhibition around H 2 O 2 soaked discs was comparable between NTHi ΔvacJ and WT strains (Fig. S1D). Lastly, this work shows VacJ requirement for NTHi pulmonary infection in mice, supporting previous observations 18 , and NTHi375ΔvacJ showed a comparable level of attenuation in mice with either normal lung function or lung emphysema. The panel of fatty acids identified and quantified in SCIENTIfIC REPORTS | (2018) 8:6872 | DOI:10.1038/s41598-018-25232-y murine lung samples, next to other soluble and cellular host factors such as antimicrobial peptides or alveolar macrophages, may contribute to the observed vacJ mutant decreased numbers, compared to those by the WT strain. Similar fatty acid composition for both normal function and emphysematous murine lung samples may contribute explaining the comparable vacJ mutant loads in the two tested in vivo models. We acknowledge that the emphysema model used in this study generated lung lesions (Fig. S3), but did not modify lung fatty acid content, which may relate to potential limitations of the procedure used for emphysema induction over a 17 days time window. Further work will be needed to assess the effect of longer elastase treatment on the murine lung pathophysiology, also considering fatty acid content.
Analysis of H. influenzae genome sequences revealed the absence of homologs to E. coli PldA, therefore excluding a relationship between vacJ disruption and phospholipase-mediated removal of phospholipids at the bacterial surface. Similarly, a homolog to E. coli PagP is absent in NTHi, excluding NTHi lipid A hepta-acylation, as shown by lipid A determination for WT and vacJ mutant strains. Moreover, vacJ disruption did not modify lipid A hexa-acylation, excluding a functional relationship between VacJ and the acyltransferase HtrB. Analysis of H. influenzae genome sequences revealed a homolog to the recently characterized E. coli pqiB gene, encoding a multi Mammalian Cell Entry (MCE) domain-containing protein displaying a syringe-like architecture periplasmic bridge, which may be involved in maintaining cell envelope homeostasis 48,49 , and will be subject of future study. H. influenzae genome sequences also revealed homologs to the E. coli fadL and fadD genes, annotated as an exogenous long-chain fatty acid transporter and an acyl-CoA synthase/fatty acid-CoA ligase 11,19 , but they lack the fadE, fadB, fadH and fadA genes, encoding four enzymes involved in fatty acid degradation via β-oxidation 50 , and genes encoding enzymes of the so called β-oxidation complex II 51 . Conversely, NTHi contains the type II fatty acid synthesis (FASII) system, only lacking the fabF gene 50,52,53 , whose products are substrates for acyltransferases catalysing the initial steps in the biosynthesis of phospholipids and lipid A. Those acyltransferases encoding genes (plsB and plsC for phospholipids; lpxA, lpxD and msbB for lipid A) are also present in NTHi genomes 11,19 . Based on this gene distribution (for a summary, see Table S3), NTHi may transport long-chain fatty acids to be converted to acyl-CoAs, which could contribute to phospholipid synthesis together with FASII-derived acyl-ACPs, but is unlikely to use such fatty acids as carbon and energy source. In fact, when using arachidonic acid as a carbon source, NTHi did not grow in a chemically defined medium but, in turn, its survival was reduced in a VacJ-dependent manner. Same observations were made for oleic acid, supporting the notion that free fatty acid bactericidal effect may be a selective pressure for NTHi to develop counteracting adaptive strategies at the human airways during chronic infection. Interestingly, Helicobacter pylori is also susceptible to free fatty acids, lacks β-oxidation, and actively exchanges genetic material by natural competence and homologous recombination [54][55][56][57] . Despite occupying separate niches at the human body, both pathogens could develop common adaptive strategies contributing to colonization linked to chronic disease conditions. Identification and characterization of such selective pressures and patho-adaptive counteracting mechanisms will be subject of future studies.
Prediction of VacJ NTHi characteristics and lipobox motif. Subcellular location of the VacJ lipoprotein was predicted using the Cell-Ploc package (http://chou.med.harvard.edu/bioinf/Cell-PLoc/) 62 . VacJ characteristics were predicted with the PROTEAN program and proteomics tools from the ExPASy website. VacJ lipobox sequence was predicted using the DOLOP program 63 .
Bacterial growth. Strains were grown on chocolate agar for 16 h. Bacterial suspensions collected in PBS were normalized to OD 600 = 1, diluted to OD 600 = 0.05 in sBHI, and 200 µl aliquots were transferred to individual wells in 96-well microtiter plates (Falcon). Plates were incubated with agitation at 37 °C for 12 h in a Multiskan instrument (Thermo Scientific), and OD 636 was monitored every 15 min. Each growth curve was corrected to its blank values (sBHI). Alternatively, a bacterial suspension recovered with 1 ml MM-FFA from a freshly grown chocolate agar plate was adjusted to OD 600 = 1. Arachidonic acid stock solution (100 mM) was prepared and diluted to the requiredworking concentrations in ethanol. Then, 160 μl of (i) MM-FFA, (ii) MM-FFA with arachidonic acid 25 μM, (iii) MM-FFA with fatty acid vehicle solution, i.e. ethanol volumes identical to those used for arachidonic acid 25 μM, (iv) MM-FFA with glucose 20 mM, were transferred to individual wells in 96-well microtiter plates (Sarstedt). Next, 40 μl of the previously prepared bacterial suspensions were added to each well. Plates were incubated in a SpectraMAX 340 microplate reader at 37 °C, and OD 600 was recorded every 30 min for 12 h.
Lipid A purification and analysis. Lipid A was extracted and processed by using an ammonium hydroxide/ isobutyric acid method and subjected to negative-ion matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry 25 . Freshly grown bacteria on chocolate agar were scraped from the plates, resuspended in 400 µl isobutyric acid-1 M ammonium hydroxide (5:3 [v/v]) and incubated in a screw-cap test tube at 100 °C for 2 h, with occasional vortexing. Samples were cooled in ice water and centrifuged (2,000 × g for 15 min). The supernatant was transferred to a new tube, diluted with an equal volume of water, and lyophilized. The sample was then washed twice with 400 µl methanol and centrifuged (2,000 × g for 15 min). The insoluble lipid A was solubilized in 100 to 200 µl chloroform-methanol-water (3:1.5:0.25 [v/v/v]). Analyses were performed on a Bruker autoflex ® speed TOF/TOF mass spectrometer (Bruker Daltonics Inc.) in negative reflective mode with delayed extraction. The ion-accelerating voltage was set at 20 kV. To analyse the samples, 1 to 3 µl of lipid A suspension (1 mg/ml) were desalted with a few grains of ion-exchange resin (Dowex 50W-X8; H) in a 1.5 ml microcentrifuge tube. A 1 µl aliquot of the suspension (50 to 100 µl) was deposited on the target and covered with the same amount of dihydroxybenzoic acid matrix (Sigma-Aldrich) dissolved in 0.1 M citric acid. Different ratios between the samples and dihydroxybenzoic acid were used when necessary. Alternatively, lipid A was mixed with 5-chloro-2-mercapto-benzothiazole (Sigma-Aldrich) at 20 mg/ml in chloroform-methanol (1:1 [v/v]) at a ratio of 1:5. Each spectrum was an average of 300 shots. A peptide calibration standard (Bruker Daltonics) was used to calibrate the MALDI-TOF. Further calibration for lipid A analysis was performed externally using lipid A extracted from E. coli strain MG1655. Interpretation of the negative-ion spectra was based on earlier studies showing that ions with masses higher than 1,000 gave signals proportional to the corresponding lipid A species present in the preparation 64 .
NTHi lipid extraction and phospholipid separation by TLC. Bacterial total lipids were extracted using the Bligh and Dyer Method 65 . Briefly, 10 ml of bacterial suspensions normalized in distilled water (OD 600 = 1) were prepared by collecting biomass from NTHi strains grown on chocolate agar. Bacterial suspensions were pelleted, resuspended in 0.2 ml distilled water and 0.75 ml chloroform:methanol (1:2 [v/v]), incubated for 5 min with-and for 15 min without shaking, followed by centrifugation (15,000 g, 20 min). Supernatants were transferred to a new tube with 0.5 ml chloroform:water (1:1 [v/v]), agitated for 5-15 min, and centrifuged (15,000 g, 20 min) to separate organic and aqueous phases. Next, 0.4 ml of the organic phase (bottom) were transferred to a clean tube. All steps were performed at room temperature (RT). Chloroform was evaporated from samples (SpeedVac) and dried extracts were stored at −20 °C prior TLC. Dried lipids were dissolved in 60 μl chloroform and spotted in duplicate onto a TLC plate (silica gel 60, Merck). TLC plates were placed in a solvent saturated tank (chloroform:methanol:acetic acid, 13:5:2 [v/v/v]) for phospholipid separation. After separation, dried TLC plates were placed in an iodine chamber for band visualization, scraped or, alternatively, developed by charring with sulfuric acid 15% in ethanol at 180 °C.
Determination fatty acid composition. Fatty acid composition and quantity were determined by GC-MS of the corresponding fatty acid methyl esters (FAMEs). NTHi FAMES were prepared as it follows: triplicates of chocolate agar grown bacteria were normalized in 1 ml PBS to OD 600 = 3 and pelleted to obtain ∼40 mg bacterial biomass. Both these samples and scraped phospholipids from TLC separation (see section above), were treated with NaOH 15% (1 ml) in methanol:water (1:1, v/v) for 30 min at 100 °C. Saponified material was methylated with methanol (2 ml) in 50% (v/v) HCl 6 N for 10 min at 80 °C. In this step, 20 μl of the internal standard C17:0 (1 mg/ml) were added. NTHi FAMEs were extracted with 1.25 ml hexane:methyl tertbutyl ether (1:1, v/v), and added to a new tube (900 μl) containing 4 ml NaOH 3% in distilled water as washing step. ∼300 μl of remaining the organic phase (top) were carefully transferred to a GC vial and capped (Technical note #101, MIDI). For murine lung FAMEs preparation, lung homogenates (see below) were stored at −80 °C. Five hundred μl of each homogenate were mixed with 750 μl HCl-methanol (3 N), 150 μl toluene and 20 μl of the GC-MS internal standard (C17:0, 1 mg/ml), and incubated at 90 °C for 1 h. Next, 500 µl NaCl 0.9% and 500 µl hexane were added, mixed vigorously, and centrifuged at 2,000 r.p.m for 5 min at RT. Supernatants (~300 µl) were transferred to a new tube, dried overnight in a chemical cabinet at RT, and the pellet obtained was resuspended in 100 µl hexane for further GC-MS analysis.
GC-MS analyses of bacterial and murine lung FAMEs were performed using a 7890A GC device coupled to a 5975C Inert XL MSD mass selective detector (Agilent Technologies, Santa Clara, USA). A volume of 1 μl was injected on an Agilent J&W DB-WAX column (diameter, 0.25 mm; film thickness, 0.25 μm; length, 30 m) with a 1 ml/min helium flow. Injection parameters were as it follows: split mode injection (1:10) and injector temperature 250 °C. The temperature gradient was 1 min at 50 °C; 25 °C/min up to 200 °C; 3 °C/min up to 230 °C and then 18 min at 230 °C, and the solvent delay was 5 min. The source was set to 230 °C and 70 eV, scanning at 20 scans/ min, from 40 to 500 m/z. FAMEs were identified by comparing their mass spectra with those of the NIST library and by comparison of the retention times with a FAME standard mix (Sigma-Aldrich).
Bacterial hydrophobicity. Duplicates of chocolate agar grown bacteria were normalized in 2.5 ml PBS to OD 600 = 0.5 (A 1 ) in capped tubes. Next, 0.5 ml xylene (PanReac AppliChem) were added per tube, the mixture was incubated at 44 °C for 10 min, vigorously vortexed for 1 min, incubated under the same conditions for 1 h, and the aqueous phase OD 600 (A 2 ) was measured. Results are expressed as percentage of bacterial surface hydrophobicity ([(A 1 − A 2 )/A 1 ] × 100). Strong hydrophobic and hydrophilic bacteria get values >50% and <20%, respectively 66 .
Antimicrobial susceptibility testing. Susceptibility to ampicillin, amoxicillin-clavulanic acid, cefuroxime, cefepime, cefotaxime, ceftriaxone, imipenem, meropenem, chloramphenicol, tetracycline, ciprofloxacin, cotrimoxazole and azithromycin was determined by microdilution according to the criteria of the Clinical Laboratory Standards Institute (CLSI) 67 . Susceptibility to the hydrophobic antibiotics erythromycin, azithromycin and rifampicin was also determined by E-test (Biomérieux) and/or disc diffusion (Becton Dickinson). For PxE, bacterial susceptibility was determined by E-test (Biomérieux). Detergent susceptibility testing. PBS normalized bacterial suspensions (OD 600 = 1) were prepared by using NTHi strains freshly grown on chocolate agar. Then, 100 µl of normalized suspensions were transferred to individual wells in 96-well microtiter plates (Sarstedt), to be incubated with 100 µl sodium deoxycholate (Alfa-Aesar) for 20 min at RT in static conditions. A deoxycholate stock solution (43.25 mg/ml) was prepared in dH 2 O and diluted to the required working concentrations in PBS (0.5; 0.75; 1; 1.25; 1.5 mg/ml). Bacterial suspensions were next used to (i) measure OD 405 ; (ii) ten-fold dilution in PBS and plating on sBHI agar for c.f.u. counting (represented as log 10 c.f.u./ml); (iii) five µl of serial culture dilutions (10 −1 ) from each well were spotted on sBHI agar for visualization.
TEM. PBS normalized NTHi suspensions (OD 600 = 1) were prepared by using NTHi strains freshly grown on chocolate agar. Then, 100 µl of such suspensions were incubated with none or with 100 µl sodium deoxycholate 0.5 mg/ml for 20 min at RT in static conditions. Bacteria were next applied to Formvar-coated grids, air dried, negatively stained with 1% phosphotungstic acid in distilled water for 10 s, and examined with a JEM-1011 transmission electron microscope (JEOL) operating at 80 kV and equipped with an Orius SC1000 charge-coupled device (CCD) camera (Gatan).
Free fatty acids susceptibility testing. MM-FFA normalized bacterial suspensions (OD 600 = 0.1) were prepared by using NTHi strains grown on chocolate agar. Oleic and arachidonic acid stock solutions (100 mM) were prepared and diluted to the required working concentrations in ethanol. For arachidonic acid, tested concentrations ranged from 3 to 75 μM; for oleic acid, tested concentrations ranged from 0.8 to 2 mM. MM-FFA with fatty acid (160 µl) was transferred to individual wells in 96-well microtiter plates (Sarstedt); 40 µl of the previously prepared bacterial suspensions were added to each well, and incubated for 20 h at 37 °C with 5% CO 2 in static conditions. Vehicle solution, consisting of an ethanol volume equivalent to that used for the highest fatty acid concentration tested, and only MM-FFA controls were performed in parallel. After incubation, bacteria were serially diluted in PBS and plated on sBHI agar. Cell culture and bacterial infection. A549 human alveolar basal epithelial cells (ATCC CCL-185) were maintained and seeded as described 68 . NCI-H292 mucoepidermoid pulmonary human carcinoma epithelial cells (ATCC CRL-1848) were maintained and seeded as described 29 . Invasion assays were performed and processed as described 29,30,32,33,68  f.u. was placed at the entrance of the nostrils until complete inhalation by each mouse, previously anesthetized (ketamine-xylazine, 3:1). When indicated, mice were euthanized and lungs aseptically removed. The left lung was individually weighed in sterile bags (Stomacher80, Seward Medical) and homogenized 1:10 (w/v) in PBS. Each homogenate was serially 10-fold diluted in PBS and plated in triplicate on sBHI agar to determine the number of viable bacteria. Results are shown as log 10 c.f.u./lung. When required, remaining homogenate material was stored at −80 °C to determine lung fatty acid composition (see section above). In parallel, BALF samples were obtained in animals with normal lung function by perfusion and collection of 0.7 ml of PBS, with the help of a sterile 20 G (1.1-mm diameter) Vialon intravenous catheter (Becton-Dickinson) inserted into the trachea. Each recovered BALF fraction was serially 10-fold diluted and plated on sBHI agar. Results are shown as log 10 c.f.u./ml BALF.

Animal preparation and pulmonary function tests (PFT). Pulmonary function tests were performed
in all animals before micro-CT acquisition. To this end, animals were first anesthetized with an intraperitoneal injection of 90 mg/kg ketamine (Imalgene ® , Merial, France) and 10 mg/kg xylazine (Rompun ® , Bayer AG, Germany). Anesthetized animals were intratracheally cannulated and connected to a Flexivent rodent ventilator (Scireq, Montreal, Canada) set at a rate of 200 breaths/min and a tidal volume of 10 ml/kg. Animals were kept breathing isoflurane at 2% concentration until completely relaxed. Lung resistance (R, measured in cmH 2 O.s/ mL), compliance (C, measured in mL/cmH 2 O) and elastance (E, measured in cmH 2 O.s/mL) parameters were measured by using a single-frequency-forced oscillation, fitting the measured data to a single compartment model of the lung. All measurements were repeated at least five times.
Breath-hold gated micro-CT imaging and image analysis. Lung 3D tomographic images were acquired using X-ray micro-CT (Micro-CAT II, Siemens PreClinical Solutions, Knoxville, Tennessee) with the following parameters: 80 kVp X-ray source voltage, 500 μA current and 450 ms exposure time per projection. Seven hundred micro-CT projections were acquired during 650 ms iso-pressure breath holds at 12 cm H 2 O. Normal breathing was induced during two complete respiratory cycles between breath holds. A total lung capacity perturbation was performed every 20 breath holds to prevent atelectasis. The tomographic three-dimensional images obtained had a total of 640 slices with isotropic 46 μm voxel size and a resolution of 1024 × 1024 pixels per slice. Micro-CT images were automatically reconstructed by using the Cobra software (Exxim Computing