Mutation in the two-component regulator BaeSR mediates cefiderocol resistance and enhances virulence in Acinetobacter baumannii

ABSTRACT Acinetobacter baumannii has become one of the most challenging pathogens in many countries with limited treatment options available. Cefiderocol, a novel siderophore-conjugated cephalosporin, shows potent in vitro activity against A. baumannii, including isolates resistant to carbapenems. To date, few reports on the mechanisms of cefiderocol resistance are available. In order to investigate potential mechanisms of cefiderocol resistance in A. baumannii, we performed in vitro evolution experiments at sub-lethal concentrations of the antibiotic. All four cefiderocol-resistant strains obtained harbored mutations in two-component system BaeS-BaeR. When we engineered the mutations of BaeS (D89V) and BaeR (S104N) into the genome of ATCC 17978, these mutations increased cefiderocol minimum inhibitory concentrations (MICs) by 8-fold to 16-fold. Transcriptome analyses showed that the expression of MacAB-TolC and MFS transporters was up-regulated in BaeSR mutants. Strains over-expressing MFS transporter and MacAB-TolC displayed higher MICs and higher median inhibition concentration (IC50) values, while MICs and IC50 decreased when efflux pump genes were knocked out. In a BaeR mutant with up-regulated csu operon, we observed a higher number of pili, enhanced surface motility, and increased biofilm formation compared to wild-type ATCC 17978. Using the Galleria mellonella infection model, we found that the BaeS mutant in which paa operon was up-regulated exhibited increased virulence. In conclusion, the mutations in BaeSR decreased cefiderocol susceptibility of A. baumannii through up-regulating efflux pumps gene expression. BaeS or BaeR also controls the expression of csu and paa, influencing biofilm formation, surface motility, and virulence in A. baumannii. IMPORTANCE The widespread prevalence of multi-drug-resistant A. baumannii (MDRAB) poses a significant therapeutic challenge. Cefiderocol is considered a promising antibiotic for the treatment of MDRAB infections. Therefore, it is necessary to study the potential resistance mechanisms of cefiderocol to delay the development of bacterial resistance. Here, we demonstrated that mutations in baeS and baeR reduced the susceptibility of A. baumannii to cefiderocol by up-regulating the expression of the MFS family efflux pump and MacAB-TolC efflux pump. We propose that BaeS mutants increase bacterial virulence by up-regulating the expression of the paa operon. This also reports the regulatory effect of BaeSR on csu operon for the first time. This study provides further insights into the role of BaeSR in developing cefiderocol resistance and virulence in A. baumannii.

were whole genome sequenced. DNA-Seq reads of every strain were assembled to less than 100 contigs, with more than 99% coverage (Table S1). Then the mutations were identified in reference to the parental strain ATCC 17978. Mutations in BaeS (G212D, A71T, D89V, D360N) or BaeR (S104N) appeared in all four cefiderocol-induced-resistant strains (Table 1).
We constructed the mutations BaeS (D89V) and BaeR (S104N) in the genome of the parental strain ATCC 17978 by employing genetic engineering. The obtained mutants allowed us to investigate the impact of single mutations on cefiderocol susceptibility. Antimicrobial susceptibility testing showed a 16-fold cefiderocol MIC increase in the strains ATCC 17978 BaeR S104N . The BaeS (D89V) mutation increased the MIC by eightfold, from 0.125 to 1 μg/mL.
To clarify the influence of mutations on BaeS or BaeR function, we knocked out gene baeS, baeR, or both baeS and baeR of wild-type ATCC 17978. Meanwhile, we introduced pYMAb2-BaeS D89V or pYMAb2-BaeR S104N into wild-type ATCC 17978, and introduced pYMAb2-BaeS WT or pYMAb2-BaeR WT into the mutants, respectively. Cefiderocol MICs reduced from 0.125 to 0.03 μg/mL for BaeSR knockout strains. However, cefiderocol MICs increased when mutated BaeS or BaeR were expressed in wild-type ATCC 17978. Cefiderocol MICs had no obvious changes when we expressed wild-type BaeS or BaeR in BaeSR mutants (Table S2). These results indicated that mutations in BaeS or BaeR were gain-of-function.
In order to explore the effect of mutations on the susceptibility of other antibiot ics, we tested the MICs of ceftazidime, cefepime, imipenem, meropenem, tigecycline, polymyxin, amikacin, ciprofloxacin, and erythromycin against cefiderocol-induced-resistant strains and BaeSR mutants. Compared with wild-type ATCC 17978, almost all antibiotics tested showed reduced susceptibility to induced-resistant strains XH1823 and XH1824. However, except tigecycline, there was no obvious susceptibility alterations for BaeSR mutants and BaeSR knockout strains. BaeS knockout caused a slight decrease in tigecycline MIC (Table S3). There may be unidentified genetic or epigenetic determinants of resistance in cefiderocol-induced-resistant strains XH1823 and XH1824, in addition to BaeSR mutation. These unidentified determinants lead to additional levels of resistance to cefiderocol and other antimicrobials, such as β-lactams and aminoglycosides.

Transcriptome analysis of BaeS and BaeR mutants
In order to clarify the mechanism of the decrease in cefiderocol susceptibility caused by mutations in BaeSR, we compared the transcriptome of ATCC 17978 BaeS D89V and ATCC 17978 BaeR S104N with the wild-type ATCC 17978. The RNA-seq reads were aligned with the genome sequence of A. baumannii wild-type ATCC 17978 (GenBank accession no. NZ_CP018664.1). The analysis was based on a fold-change > 2 or < −2 (P < 0.05), where we observed a total of 77 differentially expressed genes in ATCC 17978 BaeS D89V , including 62 up-regulated genes and 15 down-regulated genes (Fig. 1A). In the case of ATCC 17978 BaeR S104N , we found 72 differentially expressed genes, of which 41 were up-regulated, and 31 were down-regulated (Fig. 1B). The Gene Ontology (GO) enrich ment analysis of differentially expressed genes showed no significantly enriched GO terms with P < 0.05 in the BaeR mutant (Fig. S1B). In contrast, the top 15 significantly enriched terms in the BaeS mutant are shown in Fig. S1A, including toxin metabolic and catabolic processes. As shown in Fig. 1C and Table 2, genes in paa operon were Research Article mSystems up-regulated in the BaeS mutant, which encode enzymes involved in the phenylacetate pathway and are associated with toxin metabolic processes. Among these differentially expressed genes, AUO97_RS10110, AUO97_RS10105, and AUO97_RS10100 coding for efflux pumps MacAB-TolC, and a major facilitator superfam ily (MFS) efflux pump coding gene (AUO97_RS00560) were up-regulated in ATCC 17978 BaeS D89V . Another MFS efflux pump locus on AUO97_RS10785 in ATCC 17978 BaeR S104N was 2.07-fold up-regulated compared to ATCC 17978. To our surprise, genes coding pilus protein expressed differentially in both ATCC 17978 BaeS D89V and ATCC 17978 BaeR S104N . csuA/B was down-regulated in ATCC 17978 BaeS D89V , while csuA/B, csuA, csuB, and csuC were up-regulated in ATCC 17978 BaeR S104N (Table 2). (B) compared to wild-type ATCC 17978. The orange dots represent the up-regulated genes with statistical significance (P < 0.05 and |log2(FC)| ≥1). The blue dots represent the down-regulated genes, while the gray ones represent genes with no statistical significance. Efflux pump coding genes, the csu operon, and the paa operon were marked in red. (C) The corresponding relationship between genes and significantly enriched GO terms in BaeS S104N mutant.

Up-regulation of efflux pumps genes is associated with increased cefiderocol MICs
As we observed high expression levels of efflux pump related genes, we deter mined the impact of these genes on cefiderocol susceptibility. Thus, we carried out gene knockout and over-expression experiments based on ATCC 17978. We confirmed the over-expression of macAB-tolC, AUO97_RS00560, and AUO97_RS10785 through RT-qPCR (Fig. S2). The BaeS mutant, in which we knocked out either macB or gene AUO97_RS00560, or both macB and gene AUO97_RS00560 showed slight but distinguishable decreases in cefiderocol MICs. However, cefiderocol MIC did not significantly decrease when we knocked out these genes from the genome of wild-type ATCC 17978, which indicated that the up-regulation of the efflux pump genes caused by the mutation in BaeS resulted in the reduction in cefiderocol susceptibility (Table S2). The cefiderocol MICs of ATCC 17978 strains with overexpression efflux pumps for pYMAb2-p ompA -macAB-tolC, pYMAb2-p ompA -MFS00560, or pYMAb2-p ompA -macAB/tolC-MFS00560 increased onefold to twofold (Table 3; Table  S2). Considering that cefiderocol MIC may not be sensitive enough to detect slight changes in antibiotic susceptibility, we measured the IC 50 values for cefiderocol in these strains. Here, we observed obvious changes, especially in the macB knockout strain, which showed increased susceptibility to cefiderocol, while the macAB-tolC over-expressing strain exhibited dramatically increased resistance (Table 3 and Fig. 2).

The efflux pump inhibitor CCCP restores the susceptibility of A. baumannii to cefiderocol
To demonstrate that it is indeed the action of the efflux pumps that confer cefiderocol resistance, we used carbonyl cyanide m-chlorophenylhydrazine (CCCP) as an inhibitor. In the presence of 25 μM CCCP, MICs to cefiderocol were significantly reduced from ≥16 to 1 µg/mL for the strains XH1823 and XH1824. The MIC values of ATCC 17978 BaeS D89V and ATCC 17978 BaeR S104N were reduced to 0.25 μg/mL with a fourfold and eightfold reduction when CCCP was added (Table 3). Moreover, cefiderocol MICs of strains in which MFS efflux pumps were over-expressed also decreased when CCCP was added. Therefore, up-regulation of efflux pumps, driven by the proton motive force, appears to be a key mechanism for reduced susceptibility to the antibiotic in A. baumannii with BaeSR mutations. However, the MIC of strains with MacAB-TolC over-expression showed no obvious change in the presence of CCCP, which might be explained by the fact that CCCP could not inhibit the activity of MacAB-TolC.

A. baumannii ATCC 17978 with a D89V mutation in BaeS exhibits increased virulence
We previously found that the strain ATCC 17978 BaeS D89V up-regulated the gene expression involved in toxin catabolic or metabolic processes. Thus, we used the Galleria mellonella infection model to evaluate the effect of BaeSR mutations on the virulence of A. baumannii. As shown in Fig. 3, ATCC 17978 BaeR S104N showed similar virulence as ATCC 17978 (P = 0.3379). In contrast, the strain with the BaeS (D89V) mutation appeared to be significantly more virulent, as less than half of the larvae survived 72 h post-infection (P = 0.0087). However, all tested strains exhibited less virulence than the highly virulent AB5075 strain we used as a control.

Mutations in BaeS and BaeR impact motility and biofilm formation via the expression of csu genes
Genes coding for the pilus protein expressed differently in the BaeS mutant and the BaeR mutant. Using cryo-electron tomography (cryo-ET) analysis, we could observe that pili  were rarely present in the case of the ATCC 17978 strain. Similarly, few pili were seen on the surface of the BaeS mutant. In contrast to these two strains, pili were abundant in the BaeR mutant ( Fig. 4; Videos 1 to 3). When assessing the biofilm formation of mutants, we observed that ATCC 17978 BaeR S104N had a greater ability to form biofilms on plastic surfaces than the parent strain ( Fig. 5A), while ATCC 17978 BaeS D89V did not show significant differences in comparison with the parent strain. We also tested bacterial adhesion to human bronchial epithelial (HBE) cells. No significant differences were detected between the BaeSR mutants and ATCC 17978 ( Fig. S3; Fig. 5B). Next, we investigated the swarming motility of the strains as we observed an impact on the expression of csu genes, with ATCC 17978 BaeR S104N leading to higher and ATCC 17978

DISCUSSION
As a global regulator, the main function of the TCS BaeSR is to up-regulate efflux pump expression in response to specific stress. BaeSR has been reported in Salmonella Typhimurium and E. coli to modulate the expression of outer membrane proteins or specific efflux pumps, thus affecting antibiotic susceptibility (15)(16)(17). In A. baumannii, BaeSR has been associated with tigecycline resistance via regulating AdeIJK and AdeABC pumps (14,18). In this study, ATCC 17978 showed higher cefiderocol MICs when valine substituted aspartic acid at 89 of BaeS or asparagine substituted serine at 104 of BaeR. The 104th amino acid is located in the cheY-homologous receiver domain of BaeR, and the amino acid substitution at this position may affect the reception of upstream signals. Transcriptome analysis showed that the BaeS mutation caused the over-expression of macAB-tolC and an MFS efflux pump coding gene AUO97_RS00560. Another MFS efflux pump, AUO97_RS10785, was up-regulated in strains with the BaeR mutation. Although previous reports describe the role of BaeSR as regulating the expression of resistance-nodulation-division (RND) efflux family (AdeABC and AdeIJK) or the ATP-bind ing cassette (ABC) family (MacAB-TolC), we demonstrated for the first time that the BaeSR participated in the regulation of the MFS transporter expression. MFS is by far the largest and most diverse of secondary transporters superfamily known (19), and its members are ubiquitous across all domains of life (20). MFS transporters in the important pathogen A. baumannii have not been studied in detail, with CraA, AmvA, AbaF, AbaQ, and TetA being associated with susceptibility of chloramphenicol, erythromycin, fosfomycin, quinolones, and tetracyclines, respectively (20-23).

Research Article mSystems
In this study, we found that efflux pump inhibitor CCCP restored the susceptibility of BaeSR mutants to cefiderocol. The main mechanism of CCCP's action was via disrupting the proton motive force of membranes. Thus, CCCP inhibited efflux pumps such as RND and MFS family, which were driven by the proton motive force (21). The compound inhibits the activity of MFS transporters, but not that of MacAB-TolC, which belongs to the ATP-driven ABC transporters. We showed that cefiderocol MICs of BaeSR mutants decreased significantly when CCCP was added, which provided clear evidence that MFS transporters played a role in the cefiderocol removal as a consequence of BaeSR mutations. The addition of CCCP decreased the cefiderocol MIC of macB knockout strain compared to ATCC17978 BaeS D89V , which indicated that MacAB-TolC and the MFS transporter potentially act together to allow bacterial survival in the presence of cefiderocol.
Although the transcriptome analysis and the efflux pumps inhibition assay confirmed that the up-regulation of certain efflux pumps played an important role in cefiderocol resistance, we did not observe the same effect when performing single-gene knockout and over-expression experiments. When we knocked out AUO97_RS00560 or macB, we only observed a slight reduction in MIC values. A possible reason is that other efflux pumps may be up-regulated when one efflux system is absent or inactive. Bacteria express numerous efflux transporters with functional redundancy and partially overlapping substrates. The expression of other alternative transporters may increased when one of them are disrupted (24). Another possibility should be mentioned: the fact that MFS transporters are energized inner membrane transporters, the simple over-expression might not be sufficient to remove cefiderocol from the periplasm to the surrounding media across the outer membrane, which might require a carefully balanced over-expression network containing MFS transporters, the periplasmic adaptor proteins and outer membrane factor channel-like proteins (25). In addition, even when we over-expressed both MFS transporter and MacAB-TolC at the same time, the MIC of cefiderocol was not as high as that of the BaeS mutant, which suggested that maybe factors other than the two pumps we did not find out also contributed to the decreased cefiderocol susceptibility.
We also found paa operon up-regulated in the BaeS mutant. paa operon encodes enzymes responsible for phenylacetic acid (PAA) degradation, which is linked to bacterial virulence; it was previously shown that knocking out gene paaA in ATCC 17978 results in attenuated virulence in a zebrafish infection model (26). A previous study showed that the virulence of ATCC 17978 was attenuated in ΔpaaB mutant in a murine cathe ter-associated urinary tract infection model (27). Consistent with this, we observed an increase in A. baumannii ATCC 17978 virulence in the BaeS mutant with up-regulated paa in the G. mellonella infection model. The study mentioned above also indicated that Csu expression nearly halved when the paa operon was over-expressed, and could be restored by adding an exogenous PAA (27). Similarly, we observed paa up-regulation, accompanied by csu down-regulation in the BaeS mutant strain. Therefore, we speculate that this process may be regulated by BaeS (Fig. 6).
The csu cluster encodes Csu pili, which are formed from four different subunits, namely CsuA/B, CsuA, CsuB, and CsuE, and are assembled via a novel archaic chaperoneusher pathway (28). Csu pili are required for biofilm formation on abiotic surfaces (29), but are not necessary for bacterial adhesion to epithelial cells (30). In our study, ATCC 17978 BaeR S104N showed higher expression of the csu operon than the reference strain. This resulted in increased biofilm formation on plastic surfaces and enhanced motility on agar plates. However, the biofilm formation of ATCC 17978 BaeS D89V with decreased expression of csuA/B was not significantly different from that of the reference strain. This may be because CsuA, CsuB, and CsuE subunits are still assembled to some form of short pili or tip fibrillum and mediate biofilm formation on abiotic surfaces, as described in a previous study (31). No difference in epithelial cell attachment between the reference strain and the BaeSR mutants was observed. BfmR, the cytoplasmic response regulator of TCS BfmSR, is a transcriptional activator that regulates the expression of csu operon (32). Another study found that A1S_2811, a two-component regulator in ATCC 17978, might influence surface motility and biofilm formation by regulating the csu operon (33). This suggests that the expression of csu is under the control of multiple transcriptional regulators. To our knowledge, this is the first time that BaeSR has been reported to regulate csu expression, which required further investigation into whether this function of BaeSR occurs via a cross-talk with other regulators.
In conclusion, mutations in BaeS (D89V) and BaeR (S104N) decrease the susceptibility to cefiderocol in A. baumannii via the up-regulation of MFS transporters and the efflux pump MacAB-TolC. In addition, the BaeS mutation increases virulence by up-regulating the paa operon. The TCS BaeSR plays a key role in regulating the expression of Csu pili, which affects biofilm formation and swarming motility. This study provides further insights into the role of BaeSR in cefiderocol resistance and virulence in A. baumannii.

Bacterial isolates, culture conditions, and antimicrobial susceptibility testing
All isolates were cultured in Mueller-Hinton broth (Oxoid, UK), or on Mueller-Hinton agar plates at 37°C overnight. Minimum inhibitory concentrations (MICs) of cefiderocol were determined by broth microdilution method in cation-adjusted Mueller-Hinton broth or iron-depleted cation-adjusted Mueller-Hinton broth (ID-CAMHB) according to CLSI guidelines (34).
Half maximal inhibitory concentration (IC 50 ) of cefiderocol was detected on a 96-well cell culture plate through microdilution broth method. Cefiderocol was serial diluted with 100 µL ID-CAMHB to reach a final concentration of 0.0039 µg/mL (from 4 µg/mL. Then ~5 × 10 5 CFU/mL bacterial cells were added to each well. Bacteria cells were inoculated in the medium containing no cefiderocol as control. OD 600 values of each well were measured using a Multiskan Go microplate reader (Thermo Scientific, USA) after incubation for 18 h. Three independent experiments were performed for every isolate.

In vitro evolution experiments for the selection of cefiderocol-resistant strains
A. baumannii ATCC 17978 served as a parental culture to perform serial passage experiments as previously described (35). Briefly, a single clone of ATCC 17978 was inoculated in 2 mL MH broth with an initial cefiderocol concentration of 0.5-fold MIC (0.06 µg/mL) and transferred into 2 mL fresh media containing a double concentration of cefiderocol at a dilution of 1:100 every 24 h until no further bacterial growth was observed. All the cultures in serial passage experiments were stored at −80°C. One single colony was randomly selected from the final populations of four parallel, serial passage experiments and stored as XH1799, XH1800, XH1823, and XH1824 (Table 4).

Whole genome sequencing and analysis
The genomic DNA of XH1799, XH1800, XH1823, and XH1824 were extracted using QIAamp DNA Mini Kit (Qiagen, USA) according to the manufacturer. Zhejiang Tianke (Hangzhou, China) was entrusted with the whole genome sequencing (WGS) via Illumina Hiseq. With reference to the original genome of ATCC 17978 (GenBank accession no. NZ_CP018664), mapping and putative mutations detection were carried out using breseq v0.33.0 (36). Putative mutations were confirmed by PCR and Sanger sequencing (primers are listed in Table S4).

Reconstruction of mutations
Reconstruction experiments were performed as described previously (37,38). Fragment of baeS (D89V), and baeR (S104N) was amplified from corresponding induced-resist ant strains. Purified PCR products were inserted into the pMo130-Hyg r vector diges ted with BamHI and XbaI using 2× Hieff Clone Enzyme Premix (Yeason, China). The recombinant plasmid was elecro-transformed into ATCC 17978 and selected on plates containing 100 µg/mL hygromycin. Then, ATCC 17978 harboring inserted pMo130-Hyg rtarget fragment construct was cultured overnight in Luria-Bertani broth containing 20% sucrose to select isolates of which target gene achieved allelic replacement by a second cross-over ( Table 4). The reconstructed mutations were confirmed through PCR and Sanger sequencing.

Transcriptome analysis
To further clarify the regulatory effect of BaeS and BaeR mutations on gene expression, RNA-sequencing of ATCC 17978, ATCC 17978 BaeS D89V , and ATCC 17978 BaeR S104N was performed. Briefly, the overnight bacterial solution was diluted in 100 mL MH broth at a 1: 100 and grown to log phase. The culture was centrifuged at 5,000 rpm for 10 min. After grinding with liquid nitrogen, total RNA was extracted using TRIZOL Reagent (Invitrogen, USA). Zhejiang Tianke (Hangzhou, China) was entrusted with bacteria mRNA sequence library construction and sequencing. The RNA-seq reads were mapped to the reference genome of A. baumannii ATCC 17978 using Rockhopper v2.0.3 (39). The output data were analyzed by edgeR (40). Differentially expressed genes were selected with significance cutoffs of BH adjusted P < 0.05 and |fold-change| ≥2. GO annotation was performed by eggNOG (http:// eggNOG-mapper.embl.de), and enrichment analysis was performed using ClusterProfiler R package (41).

Gene knockout and over-expression
Knockout and over-expression experiments were performed to further verify the role of the high expression of efflux pumps in cefiderocol resistance. Gene knockout was performed according to the previous study (37). Briefly, amplify up-stream and downstream fragments of target genes and clone purified PCR products into pMo130-Hyg r vector digested with BamHI and XbaI using 2× Hieff Clone MultiS Enzyme Premix (Yeason, China). The recombinant plasmid was electro-transformed into wild-type ATCC harboring inserted pMo130-Hyg r -target fragment (Up/Down) construct was cultured in Luria-Bertani broth containing 20% sucrose to induce the second cross-over and allelic replacement. Select clones which were white when sprayed with 0.45 M pyrocathechol and were susceptible to 100 µg/mL hygromycin and confirm that the target gene was deleted via PCR and Sanger sequencing (primers were listed in Table S4). Target genes together with ompA promoter were cloned into pYMAb2-Hyg r plasmid digested with BamHI and SalІ using 2× Hieff Clone MultiS Enzyme Premix (Yeason, China), and then electro-transformed into ATCC 17978 to over-express target genes in host bacteria (Table 4). qRT-PCR was used to verify the over-expression of the target genes. The expression of the baeS, baeR, AUO97_RS00560, AUO97_RS10785, and macB were assessed using specific primers (Table S4), employing real-time PCR using TB Green Premix Ex Taq (Takara, Japan) with standard procedure for two-step PCR amplification according to the instruction in the LightCycler 480 System (Roche Diagnostics). Three independent experiments were performed for every isolate.

Motility assay, biofilm formation on abiotic surface, and HBE adhesion
Swarming motility assays were performed as described previously (42). Firstly, Luria-Ber tani media containing 0.35% agar was poured into 9 cm Petri dishes. And then, a colony was inoculated in the center of the above-described medium and incubated at 37°C. The diameter of growth was determined after 14 h of incubation.
The biofilm formation assay was performed according to the previous description with minor modifications (43). Briefly, one colony was inoculated in MH broth overnight at 37°C, and then transferred into a fresh broth at a dilution of 1:100 in a 96-well cell culture plate and incubated overnight. Each culture was added to at least three wells. Adherent cells were washed with phosphate buffered saline (PBS), stained with 0.1% crystal violet at 4°C for 30 min, and washed with PBS three times. Use ethanol and shake for 30 min at 200 rpm to release the dye. Absorbance at 595 nm was measured on a Multiskan Go microplate reader. Three independent experiments were performed.
HBE cell line (#CRL2741; ATCC) grew in 5% CO 2 at 37°C in Dulbecco's Modification of Eagle's Medium (DMEM, Corning, USA) with 10% fetal bovine serum as well as 1% penicillin/streptomycin. Monolayers were washed once with PBS lightly, and then log-phase bacteria were added to cells at a MOI of 100:1 and incubated for 1 h at 37°C. Infected monolayers were washed three times with PBS and lysed with 0.2% Triton X-100. Ten microliters of gradient diluted lysates were dropped onto MH agar plates. Count the number of clones to assess the attachment of BaeS and BaeR mutants to HBE cells.

Cryo-electron tomography and three-dimensional visualization
ATCC 17978 and BaeSR mutants were cultured in Luria-Bertani broth overnight and were centrifuged at 5000 rpm for 10 min. The obtained pellets were suspended in PBS, and the cell concentration was adjusted to ~OD 600 = 0.8. 5 µL of the cell suspension was placed on freshly glow-discharged (for 40 s) holey carbon grids (Quantifoil Cu R2/1,200 mesh) after adding 15 nm gold marker solution. The grids were blotted with Whatman filter paper and then rapidly frozen in liquid ethane, using a homemade plunger apparatus. The grids were imaged using a 200-kV electron microscope (Talos F200C, Thermo Fisher Scientific) equipped with a field emission gun and a fast-imaging Ceta camera (Thermo Fisher Scientific). Serial EM was used to collect all tilt series (44). Images were acquired at 22,000× magnification with an effective pixel size of 4.81 Å at the specimen level. The defocus was set as close to −5 µM. A total dose of ~90 e − /Å 2 (the dose was estimated based on the screen current) is distributed among 35 tilt images covering angles from −51° to 51° with a tilt step of 3°. All recorded images were then stacked and aligned by IMOD (45) and tomograms (binned four times; 19.24 Å pixel size) were reconstructed by simultaneous iterative reconstruction technique (SIRT) reconstruction using Tomo3D (46). Segmentations of representative tomographic reconstructions from ATCC 17978, ATCC 17978 BaeS D89V , and ATCC 17978 BaeR S104N cells were manually constructed using IMOD (45).

G. mellonella infection model
The virulence of BaeSR mutants was evaluated using the G. mellonella infection model. Log-phase bacterial cultures were resuspended in PBS to 10 7 CFU/mL. Thirty larvae of G. mellonella in each group were inoculated with 10 µL of bacterial suspension and incubated at 37°C. The death of G. mellonella was defined as no movement. Viability was assessed every 12 h for a total of 72 h.

ADDITIONAL FILES
The following material is available online.