Distinct virulence ranges for infection of mice by Bordetella pertussis revealed by engineering of the sensor-kinase BvgS

The whooping cough agent Bordetella pertussis coordinately regulates the expression of its virulence factors with the two-component system BvgAS. In laboratory conditions, specific chemical modulators are used to trigger phenotypic modulation of B. pertussis from its default virulent Bvg+ phase to avirulent Bvg- or intermediate Bvgi phases, in which no virulence factors or only a subset of them are produced, respectively. Whether phenotypic modulation occurs in the host remains unknown. In this work, recombinant B. pertussis strains harboring BvgS variants were tested in a mouse model of infection and analyzed using transcriptomic approaches. Recombinant BP-BvgΔ65, which is in the Bvgi phase by default and can be up-modulated to the Bvg+ phase in vitro, could colonize the mouse nose but was rapidly cleared from the lungs, while Bvg+-phase strains colonized both organs for up to four weeks. These results indicated that phenotypic modulation, which might have restored the full virulence capability of BP-BvgΔ65, does not occur in mice or is temporally or spatially restricted and has no effect in those conditions. Transcriptomic analyses of this and other recombinant Bvgi and Bvg+-phase strains revealed that two distinct ranges of virulence gene expression allow colonization of the mouse nose and lungs, respectively. We also showed that a recombinant strain expressing moderately lower levels of the virulence genes than its wild type parent was as efficient at colonizing both organs. Altogether, genetic modifications of BvgS generate a range of phenotypic phases, which are useful tools to decipher host-pathogen interactions.


Introduction
Bordetella pertussis is the agent of an acute respiratory disease, whooping cough. Despite the current global vaccination coverage of approximately 86% of the population, this bacterium a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 course of infection. In addition, we identified different ranges of virulence factor expression for colonization of, and persistence in, distinct sites of the mouse respiratory tract.

Strains and plasmids
The strains used to perform animal experiments, BPSM, BPSM SS1 (previously called BPSM BvgS-E113C+N177C ), BP-BvgS Δ65 and BP-BvgS Δ65-SS1 (previously called BvgS Δ65-VFT1-SS ) were described previously [12,20]. In all of them, expression of the bvgS variants from the natural chromosomal locus was achieved by allelic exchange as described in [12]. Similarly, to construct BP-BvgS T733M the mutation of interest was introduced by mutagenesis on a pUC19 derivative containing the appropriate region of the bvgS gene, followed by cassette exchange in pUC19mint, transfer of the EcoRI-HindIII fragment into pSORTP1 and homologous recombination into BvgS newΔAS as in [12].
The BvgS Δ65 R572L variant was expressed from a plasmid for activity measurement. It was constructed by mutagenesis on a pUC19 derivative containing the region of interest, followed by cassette exchange in pUC19mpla and then in pBBRmpla [12]. The recombinant B. pertussis strains were obtained by introducing the pBBRmpla variant by conjugation in B. pertussis BPSM newΔAS carrying the chromosomal ptx-lacZ transcriptional fusion [12]. bvgS Δ65 was cloned on a plasmid in a similar manner for comparison.

Animal experiments
After 36 h of growth on standard Bordet-Gengou (BG)-blood plates or plates containing 50 mM MgSO 4 to modulate the various strains, bacteria were resuspended in sterile PBS to 10 6 bacteria per 20 μL. After intraperitoneal anesthesia with a mixture of ketamine, atropine and valium, female 6-weeks-old JAX BALB/cByJ mice from Charles River were infected by intranasal inoculation with 10 6 bacteria. Groups of 5 animals per bacterial strain were sacrificed by cervical dislocation after 3 h and 3, 7, 14 or 28 days post-inoculation in the first experiment, and groups of 3 animals after 3 h and 7, 14 or 21 days in the second. Noses and lungs were collected and homogenized using an Ultra Turrax apparatus. Serial dilutions were performed in PBS and plated on BG agar plates to count the bacteria. All the experiments were carried out in accordance with the guidelines of the French Ministry of Research regarding animal experiments, and the protocols were approved by the Ethical Committees of the Region Nord Pas de Calais and the Ministry of Research (agreement number APAFIS#9107-201603311654342 V3).

RNA extractions
B. pertussis strains were grown on BG agar plates for 2 days at 37˚C and then cultured in modified Stainer Scholte (SS) liquid medium supplemented when indicated with 50 mM MgSO 4 at 37˚C under agitation. The bacterial cultures were stopped at mid exponential phase (OD 600 = 2) by adding 1 mL of a mixture of 5:95 phenol:ethanol (v:v) to 4 mL of bacterial suspensions. Bacteria were pelleted, and total RNA was extracted using TriReagent (Invitrogen) following the manufacturer's instructions. Genomic DNA was removed by DNase I treatment (Sigma Aldrich).
BP-BvgS Δ65-SS1 and BP-BvgS T733M . For each RNA-seq sample, DNA-depleted total RNA was treated with the Ribo-Zero rRNA Removal Kit (Illumina) following the manufacturer's recommendations. The rRNA-depleted RNA was then used to make the Illumina library using the TruSeq RNA Library Preparation Kit, following sequencing on an Illumina NextSeq 500 benchtop sequencer on SR150 high output run mode. The RNA-seq data of each sample were analyzed using Rockhopper v2.0.3 with the default parameters to calculate the RPKM value for each CDS using the Bordetella pertussis Tohama I BX470248 genome annotation [21]. The RNA-seq data reported in this paper have been deposited in the Sequence Read Archive, www. ncbi.nlm.nih.gov/sra (submission SUB4097406; NCBI BioProj PRJNA474836; BioSample accessions: SAMN09374659, SAMN09374660, SAMN09374661, SAMN09374662, SAMN0 9374663, SAMN09374664, SAMN09374665, SAMN09374666).

Generation of cDNA and quantitative real-time polymerase chain reaction (qRT-PCR)
Fifteen micrograms of total RNA were treated with DNAse I, and then 500 ng of total RNA was reverse-transcribed using the Verso cDNA synthesis kit (Thermo Scientific). Polymerase chain reaction (PCR) was performed on 30 ng of cDNA using a LightCycler 480 SYBR Green I Master kit (Roche) and a Roche LightCycler 480 Instrument II. The efficiency for each primer pair was determined by serial dilutions. The experiments were performed three times for BPSM, MgSO 4 -treated BPSM, BP-BvgS Δ65 , MgSO 4 -treated BP-BvgS Δ65, BPSM SS1 , BPSM Δ65 SS1 , BPSM T733M and BP-BvgS Δ65-rev 79 , twice for BP-BvgS Δ65-rev 80 and from single biological sampling for the other BP-BvgS Δ65-rev strains. For each sample, at least three technical replicates were performed. The results were analyzed with the Light Cycler 480 software. The expression of the housekeeping gene bp3416 was used as reference to normalize the expression of the genes of interest.

Extraction of genomic DNA and sequence determination
The various segments of the bvgS gene were amplified from clarified lysates obtained by heating bacteria resuspended in H 2 O at 95˚C for 30 min, using the pairs of primers described in [22], and the amplicons were sequenced by the Sanger method. DNAseq experiments were performed on genomic DNA extracted using the Genomic tip 100/G kit (Qiagen). The sequencing libraries were prepared with the Nextera XT sample prep kit (Illumina) following the manufacturer's instruction and sequenced on HiSeq 2500 with 2x250 bp reads. The read mapping was performed with CLC Genomics Workbench V11, and variants were searched using the Basic Variant Detection module. β-galactosidase assay β-galactosidase assays were performed as described previously [12] with 3 different clones at different times, and the means and standard errors of the means were determined. TCEP was added to the cultures for 6 or 16 hours prior to harvesting the bacteria.

Statistical analyses
Statistical analyses were performed using an unpaired t test for β-galactosidase results, and using the one-way ANOVA followed by Bonferroni's multiple comparison post-test for the animal experiments and qRT-PCR results. For the RNA-seq data, the adjusted p-values were determined using the Benjamini-Hochberg procedure in Rockhopper.

Enhancement of animal colonization by reversion of the BvgS Δ65 regulation phenotype
The BP-BvgS Δ65 recombinant strain shows an inverted regulation phenotype relative to its wild type (wt) parent BPSM in laboratory conditions [20], as chemical modulation shifts BvgS Δ65 to a high-kinase mode of activity (Fig 1A and 1B). We characterized the full gene expression profile of BP-BvgS Δ65 in the 'default state' (ie, not modulated) by performing Illumina RNA sequencing (RNA-seq) experiments and compared it with that of BPSM modulated or not ( Fig  1C; S1 Table). The transcription profile of BPSM was in good agreement with previous reports [23][24][25], with high transcription levels of the vags and low transcription levels of the vrgs ( Fig  1C, green and red ovals; S1 Table). Except for a few sets of genes, the gene expression profile of BP-BvgS Δ65 was rather close to that of modulated BPSM. Thus, expression of late vags, e.g. coding for PTX or the type III secretion system, decreased to the same levels as in modulated BPSM ( Fig 1C, green oval; S1 Table). In contrast and congruent with the Bvg i phase, the expression levels of early vags such as fhaB, fimBCD or fhaC in BP-BvgS Δ65 , were similar to those in BPSM ( Fig 1A, blue oval), and bipA (bp1112), the hallmark gene of the Bvg i phase in B. pertussis, was expressed at higher level in BP-BvgS Δ65 than in BPSM grown in standard or modulated conditions ( Fig 1C, purple circle) [23]. Additionally, some other vags, like fhaS (bp2667), fmtB (bp2936), prn (bp1054), and other genes including lgmABC (bp0397, bp0398 and bp0399) and putA (bp2749) were expressed at higher levels in BP-BvgS Δ65 than in modulated BPSM (S1 Table). Most of the vrgs were expressed at levels similar to those in modulated BPSM, which was not previously reported for the Bvg i phase (Fig 1A, red ovals) [23]. Altogether, those results indicate that the default state of BP-BvgS Δ65 is the Bvg i phase.
As in vitro modulation shifts BP-BvgS Δ65 to the Bvg + phase, we investigated whether animal experiments might reveal specific conditions present in the mammalian respiratory tract that induce phenotypic modulation, thus causing a shift of BP-BvgS Δ65 to the colonization-proficient, Bvg + phase. If this occurred, the 'up-modulated' recombinant bacteria should be able to colonize the animal lungs, while they would be cleared quickly if they remained in the Bvg i phase, as described earlier using other Bvg i -phase strains [26,27]. In contrast, we did not expect them to be rapidly cleared from the mouse noses, as B. pertussis locked in the Bvg i phase was reported to survive in the mouse upper respiratory tract in a way similar to virulent bacteria [26,27]. Mouse colonization experiments were thus performed with BP-BvgS Δ65 cultured in standard conditions (i.e., 37˚C without modulator) prior to inoculation, or at 37˚C in the presence of 50 mM MgSO 4 to set the bacteria in the Bvg + phase prior to colonization. We reasoned that if the bacteria encountered modulating conditions at a late stage of mouse colonization rather than early on, pre-modulation of BP-BvgS Δ65 might facilitate initial survival in the animals, before induction of virulence factor expression resulting from in vivo modulation. If no modulation occurred in vivo, we expected pre-modulated BP-BvgS Δ65 to progressively lose virulence factor expression and therefore to be cleared like its non-modulated counterpart. The wt control strain, BPSM, was cultured in the same two conditions before mouse inoculation. The mice were infected intranasally with 10 6 bacteria. After 3 h, approximately 10 6 and 10 5 bacteria were found in the lungs and in the noses of mice infected with the BPSM control or with BP-BvgS Δ65 (Fig 1D and 1E). The colonization profiles by BPSM were as typically reported. At day 3, the bacterial loads increased in both organs, the bacteria were found in similar numbers at day 7, and then their numbers markedly decreased at days 14 and 28. Bacterial multiplication and clearance from both organs appeared to be slightly delayed for BPSM chemically down-modulated prior to inoculation, possibly because virulence factor production was initiated only when the bacteria encountered the in vivo environment of the animals' respiratory tract and shifted to the Bvg + phase.
For BP-BvgS Δ65 , in contrast, no bacterial multiplication was observed at day 3 in either organ. On the contrary, marked decreases of the bacterial loads were seen irrespective of prior up-modulation of the bacteria. Thus, in vitro modulation before inoculation failed to restore a wt-like colonization profile. The fast decreases of the BP-BvgS Δ65 populations in both organs do not support the hypothesis that up-modulating signals were present in the respiratory tract that might have shifted BvgS Δ65 to a high-kinase mode of activity.
At later time points, however, BP-BvgS Δ65 appeared to survive at similar levels to BPSM. BP-BvgS Δ65 is non-hemolytic on blood agar plates in non-modulated culture conditions, due to the lack of or the very low-level expression of the hemolysin/adenylate cyclase gene cyaA, a late vag whose transcription requires high concentrations of phosphorylated BvgA [8]. Intriguingly, however, all colonies recovered at day 3 from the lungs of mice infected with non-modulated BP-BvgS Δ65 , and 40% of those recovered from the lungs of mice infected with BP-BvgS Δ65 modulated prior to infection were hemolytic (Fig 2A). At day 7, only hemolytic colonies were recovered from the lungs of the two groups of mice. Hemolytic colonies were also recovered from the noses, in low proportions at day 3 that appeared to increase at day 7 ( Fig 2B). Nevertheless, non-hemolytic clones were detected in the noses up to day 28. The hemolytic phenotype was maintained after re-streaking those clones on Bordet Gengou agarblood plates with or without modulators, suggesting that the bacteria were locked in the Bvg + phase.
The hemolytic bacteria appear to have undergone a reversion from the Bvg i phase to the Bvg + phase in most of the animals. Possibly because of this reversion, we observed a moderate increase of the bacterial loads of BP-BvgS Δ65 in the lungs between days 3 and 7, for both the bacteria chemically modulated and those not modulated prior to inoculation (Fig 1D and 1E). Nevertheless, the bacterial loads remained 1000-fold lower than those of BPSM even at their peak at day 7. The numbers of bacteria in the lungs of those two groups of mice leveled off at day 14 and decreased thereafter. In the noses, where the proportions of hemolytic colonies were initially lower, the bacterial counts remained steady between days 3 and 7 and started to decrease from day 14. Thus, infection of mice with BP-BvgS Δ65 resulted in the appearance of hemolytic variants whose proportions in the bacterial populations recovered from both organs increased over time. This suggested that it could be due to selective pressure in mice. As only 0.1% of the initial bacterial loads were recovered from the lungs after 3 days, the remaining The VFT domains and Histidine Kinase (HK) domains are connected to one another with the two-helix linkers 1 and 2 and the intervening PAS domain in BPSM, or with a single two-helix linker X in BPSM-BvgS Δ65 . The receiver and Hpt domains of BvgS were omitted for clarity. B. Activities of the BvgS variants as measured using the ptx-lacZ reporter system. The strains were grown in standard conditions (0) or with 50 mM MgSO 4 (Mg 50). The measurements were performed at least three times, and the means and standard errors of the mean are given. Statistical analyses were performed, and significant p values are indicated ( ���� , p<0.0001). C. Illumina RNA sequencing was performed for BPSM, BPSM grown in the presence of 50 mM MgSO 4 (BPSM Mg), and BP-BvgS Δ65 . The trancriptomes of BPSM grown in standard conditions and in 50 mM MgSO 4 were used as the reference Bvg + and Bvgtranscriptomes, respectively. The data are plotted as the ratios of gene expression in BP-BvgS Δ65 relative to BPSM (y-axis) versus the ratios of gene expression in modulated BPSM (denoted BPSM Mg) relative to BPSM (x-axis). Thus, genes found on a straight line starting from the origin with a slope of 1 are regulated similarly in BP-BvgS Δ65 and in modulated BPSM, while genes found above this line are expressed at higher levels in the former than in the latter. Two distinct groups of vags are circled in green and blue, with the blue oval encompassing some early vags. Most vrgs are circled in red, and bipA is shown in purple. The complete datasets of these experiments are presented in S1 Table.  Virulence phenotypes of Bordetella pertussis with BvgS variants 99.9% bacteria, which were presumably non-revertants, most likely died in the first three days. The potential selective pressure for virulent bacteria-using the hemolytic phenotype as a proxy for virulence-appears to be less intense in the nose than in the lungs, as more than 40% of the BP-BvgS Δ65 bacteria recovered from the noses were non-hemolytic at day 14 ( Fig 2B).

Colonization profiles by distinct Bvg i -phase bacteria
Since hemolytic revertants were obtained from several animals, the most likely explanation is that they spontaneously arose at low frequency in our master stock of BP-BvgS Δ65 and were selected for by the host environment. However, when BP-BvgS Δ65 was streaked on blood-agar plates and incubated for 5 days at 37˚C, no hemolytic colony was detected, indicating that hemolytic bacteria, if present, would represent a very small proportion of the stock. Furthermore, deep Illumina whole-genome sequencing of BP-BvgS Δ65 did not reveal any mutations.
We nevertheless performed a new round of single colony isolation starting from our initial BP-BvgS Δ65 stock, which yielded BP-BvgS Δ65new , and we performed another mouse colonization experiment with the newly isolated clone. In this second experiment, in addition to BPSM and BP-BvgS Δ65new we included two other strains that express BvgS variants with different levels of enzymatic activities, BPSM SS1 [12] and BP-BvgS Δ65 SS1 [20] (Fig 3A and 3B). In BPSM SS1 , two selected residues at the lips of the lobes of the VFT1 domain of BvgS were replaced with Cys residues, which results in the formation of an inter-lobe disulfide (S-S) bond that  artificially closes that domain when the bacteria are growing in liquid culture [12]. This forced closure abolished BvgS kinase activity in both standard and modulated conditions using the ptx-lacZ reporter [12]. However, using the fhaB-lacZ reporter in standard growth conditions, we observed in this study that BvgS SS1 expresses the fhaB gene, an early vag whose transcription requires only small concentrations of phosphorylated BvgA [7,8] (Fig 3B). The addition of a reducing agent, TCEP, to the bacterial cultures increased the β-gal activity with the same reporter. Thus, in non-modulating conditions, BvgS SS1 appears to be in a 'low-kinase' (Bvg i ) mode of activity, rather than in the bona fide phosphatase mode. Conversely, introduction of the same S-S bond in BvgS Δ65 , yielding BvgS Δ65-SS1 , shifted it to a high-kinase mode of activity, as shown with the ptx-lacZ reporter in standard conditions [20] (Fig 3B). Upon addition of TCEP to the culture, the β-gal activity decreased significantly but not fully, suggesting that S-S bond reduction was most likely incomplete in those conditions.
In the second animal experiment, two Bvg + -phase strains, BPSM and BP-BvgS Δ65-SS1 , and two Bvg i -phase strains, BP-BvgS Δ65new and BPSM SS1 , were thus used to inoculate mice. The bacteria were counted in their lungs and noses over the time course of infection. The colonization profiles of both organs by BP-BvgS Δ65-SS1 were similar to those of BPSM (Fig 3C and 3D). This indicates that formation of the S-S bond in the VFT1 domains of BvgS Δ65 SS1 enabled the bacteria to express the virulence factors necessary for colonization and survival in both organs. In contrast, the two Bvg i -phase strains, BP-BvgS Δ65new and BPSM SS1 , were rapidly cleared from the lungs but colonized the noses of mice.
In contrast to the first animal experiment, no hemolytic colonies of BP-BvgS Δ65new were obtained from either organ at any time point. This supported the hypothesis that the initial BP-BvgS Δ65 bacterial stock used for the first experiment contained a very small proportion of revertants that were not detected on blood-agar plates prior to inoculation of the mice. In the lungs, BP-BvgS Δ65new was rapidly eliminated and was not detectable after day 7. In the noses, this strain did not multiply, and the bacterial load markedly decreased at day 7, followed by a milder decrease thereafter.
In the lungs, BPSM SS1 did not multiply and was rapidly cleared, though not quite as quickly as BP-BvgS Δ65new . In contrast, BPSM SS1 persisted in the noses in a manner similar to BPSM, or even better at day 21. Thus, whereas the recombinant strains that express BvgS variants displaying intermediate modes of activity were rapidly cleared from the lungs, they survived much longer in the noses. Intriguingly, BPSM SS1 persisted in the noses at higher bacterial counts than BP-BvgS Δ65new . The observation that the nose colonization profiles of two strains in the Bvg i phase were not identical indicates that they might display slightly different gene expression patterns.
Expression range of the various recombinant strains. We performed RNA-seq experiments on BPSM SS1 and BP-BvgS Δ65-SS1 to characterize their gene expression profiles (Fig 4A  and 4C). An additional strain, BPSM T733M , which represents a prototypical Bvg i -phase-locked strain, was included in the experiments (Fig 4B). The T 733 M mutation, which localizes to the DHp domain of BvgS, spontaneously occurred in Bordetella bronchiseptica and led to the initial description of the Bvg i phase [9]. This mutation was introduced in the BPSM chromosome by allelic exchange to serve as the Bvg i reference strain for RNA-seq.
The gene expression profiles of BPSM SS1 and BPSM T733M were broadly similar to each other and rather close to that of BP-BvgS Δ65 (compare Fig 4A and 4B with Fig 1C; S1 Table). Subtle differences were nevertheless found among the Bvg i -phase strains, notably regarding the levels of expression of specific vags (Fig 1C and Fig 4A-4C, blue ovals; S1 Table). Furthermore, bipA (bp1112), a specific marker of the Bvg i phase [23], was overexpressed in both BP-BvgS Δ65 and BPSM T733M (purple circles), but not to the same extent in BPSM SS1 . Altogether, BP-BvgS Δ65, BPSM T733M and BPSM SS1 present transcriptomic patterns characteristic of the Bvg i phase, but within a certain range of gene expression levels. The reason why BPSM SS1 was more efficient at colonizing the mice noses than BP-BvgS Δ65 is difficult to ascribe to specific differences between their transcriptomes.
BP-BvgS Δ65-SS1 expressed several vags, including ptx, vag8 (bp2315), tcfA (bp1201), brkA (bp3494) and bfrD (bp856), at significantly lower levels than BPSM, but higher than the three Bvg i -phase strains (S1 Table). The expression of bipA in BP-BvgS Δ65-SS1 was higher than in BPSM and in the same range as in the Bvg i -phase strains. BP-BvgS Δ65-SS1 also expressed the genes of the chemotaxis and flagellar operons at higher levels than modulated BPSM (Fig 4C,  grey circle). BP-BvgS Δ65-SS1 is thus closer to the Bvg + phase than the other variants but with lower expression levels of several vags than BPSM, although it colonized mice as efficiently as BPSM. Genetic engineering of BvgS can thus generate a range of intermediate phenotypes.
We also performed quantitative real-time polymerase chain reaction (qRT-PCR) experiments to quantify the transcripts of selected genes in the various strains (Fig 5). We included late vags (cya, ptx-S1, tcfA), an early vag (bp1881, i.e. fimB), the intermediate phase marker bipA, and the vrg bp2782. We tested the effect of modulation on gene expression in BP-BvgS Δ65 .
The qRT-PCR data showed that the strains expressing BvgS Δ65 , BvgS T733M and BvgS SS1 have similar patterns of expression that represent the Bvg i phase. This analysis also confirmed that BP-BvgS Δ65 treated with MgSO 4 and BP-BvgS Δ65 SS1 are closer to the Bvg + phase than the three Bvg i -phase strains, although they expressed some vags at lower levels than BPSM. Thus, the BvgS variants analyzed in this work display a range of gene expression patterns between the fully Bvg + and Bvgphases, which most likely correlate with their respective behaviors in animal colonization.

Identification of a mutation in hemolytic colonies of BP-BvgS Δ65
Finally, we characterized hemolytic colonies of BP-BvgS Δ65 that appeared in the first animal experiment to test the hypothesis that their bvgS gene harbored a mutation that locks it in the kinase mode. Such 'constitutive' mutants were previously reported to spontaneously occur in wt BvgS, with the substitutions mapping in particular in the PAS domain and the linker 1 [28,29]. We thus PCR-amplified and sequenced the corresponding region of bvgSΔ65 from a number of hemolytic clones obtained at various time points from the two sites of the mice's  respiratory tracts. A mutation replacing the CGC codon of the Arg572 residue with a Leu CTC codon was identified in a majority of the clones recovered from the lungs, and in some of those recovered from the noses (Fig 6A). Arg572 is localized in linker X of BvgS Δ65 .
To confirm the effect of this mutation on the activity of BvgS Δ65 , the R 572 L substitution was introduced in that variant by site-directed mutagenesis. The recombinant BvgS Δ65-R572Lexpressing strain was hemolytic on blood agar plates irrespective of the addition of 50 mM MgSO 4 as the modulating agent. It also showed a high level of β-galactosidase (β-gal) activity using the ptx-lacZ reporter fusion that did not respond to modulation (Fig 6B), unlike BP-BvgS Δ65 . This confirms that the R 572 L mutation is sufficient to lock the BvgS Δ65 variant in a kinase state unresponsive to modulation.
Intriguingly, other hemolytic variants selected for sequencing did not harbor the R 572 L or any other substitution in the linker X region of BvgS Δ65 , and whole-genome sequencing on seven such clones did not identify mutations to account for their hemolytic phenotype. We performed qRT-PCR analyses on a hemolytic variant harboring the R 572 L substitution in BvgS Δ65 (BP-BvgS Δ65 -rev79) and on several other revertants devoid of that substitution. The transcription patterns of the revertant strains showed rather similar trends (Fig 6C). They expressed cyaA at levels similar to, or moderately lower (in BP-BvgS Δ65 -rev80) than BPSM, in agreement with their hemolytic phenotypes on blood-agar plates. Similarly, the expression levels of ptx or fimB were similar to or slightly below those of BPSM, whereas those of another vag, tcfA, were lower. The revertants transcribed the Bvg i -phase gene bipA at levels similar to or higher than BPSM. The expression levels of the selected vrg, bp2782, were significantly lower in several of them, but not in BP-BvgS Δ65 -rev79 harboring the R 572 L mutation, than in BPSM.

Discussion
We genetically engineered a chimera, BvgS Δ65 , in which the linker X of a BvgS homolog replaces the region encompassing the linker 1, the PAS domain and the linker 2 of BvgS, between the transmembrane and the DHp domains [20]. In standard growth conditions, recombinant BP-BvgS Δ65 is in the Bvg i phase, and modulation causes an increase of BvgS Δ65 kinase activity, contrary to its effect on wt BvgS. We thus used the inverted regulation properties of this chimera as a tool to investigate host-pathogen interactions in an animal model of infection. Additional strains harboring other engineered versions of BvgS were also included in mouse colonization experiments. Two distinct groups of strains were thus revealed. BP-BvgS Δ65 and BPSM SS1 , which are in the Bvg i phase, were able to colonize and to survive in the mice noses but were rapidly cleared from the lungs, while BP-BvgS Δ65 SS1 could colonize and survive in those two compartments, similar to the Bvg + -phase control strain BPSM. There thus appears to be distinct virulence thresholds for the colonization of the mouse nose, which both Bvg i -and Bvg + -phase bacteria can colonize, and the mouse lungs, which only Bvg + -phase bacteria can. Our transcriptomic analyses showed that those Bvg i -and Bvg + -phase strains populate distinct ranges of phenotypes, based on gene expression levels in the various strains. Such recombinant strains are useful tools to finely dissect the interactions of B. pertussis with its host. A. Numbers of hemolytic colonies selected for sequencing that present the R 572 L mutation. The variants were isolated from the lungs (left panel) and noses (right panel) of the mice. The cause of the hemolytic phenotype of the clones that do not carry the above mutation is not identified (ni). B. The ptx-lacZ reporter system was used to determine the activities of the BvgS Δ65 R572L compared to BvgS Δ65 in standard conditions (0) or after growth in the presence of 2 mM chloronicotinate (CN 2). The measurements were performed at least three times, and the means and standard errors of the mean are given. Statistically different values relative to BPSM are indicated by ���� , p<0.0001; ns, p>0.05. C. Quantitative RT-PCR analysis of selected genes in various recombinant strains. The values represent the Log2 values of the mean expression of each gene normalized to bp3416 using the 2ΔCt method. The error bars represent the Ct standard deviations. BP-BvgS Δ65 -rev79 (obtained from a mouse nose at day 14) is a hemolytic clone with BvgS Δ65 carrying the R 572 L substitution. BP-BvgS Δ65 -rev26 (from nose; day 3), -rev47 (from lung; day 3), -rev54 (from nose; day 7) and -rev80 (from nose; day 14) are other hemolytic revertants without that mutation. The results represent biological triplicates for BP-BvgS Δ65-rev79 , biological duplicates for BP-BvgS Δ65-rev80, and single biological sampling for the other strains. In all cases, the measurements were performed in triplicates. Statistical analyses were performed using gene expression in BPSM as a control and indicated as follows: ��� , p<0.001; �� , p<0.01; � , p<0.05. https://doi.org/10.1371/journal.pone.0204861.g006

Virulence phenotypes of Bordetella pertussis with BvgS variants
Our initial goal in this study was to determine whether modulation might occur in the host. Closing of the inter-lobe cavity upon ligand binding triggers signaling in other VFT domainbased systems [30,31]. Several observations have suggested that the VFT1 domains of BvgS might function similarly. Thus, they are open in the available crystal structure, their putative solute-binding cavity is conserved among B. pertussis isolates, and their closing by inter-lobe S-S bond formation down-regulates the kinase activity of wt BvgS [12]. Therefore, we took advantage of the inverted regulation phenotype of BP-BvgS Δ65 to test the hypothesis that VFT1 might bind modulating ligands in the course of infection by B. pertussis. As S-S-bond-induced closing of the VFT1 domains of BvgS Δ65 restored kinase activity and a wt-like colonization profile by BP-BvgS Δ65 SS1 in a mouse model of infection (this work), ligand-induced closing of the VFT1 domains of BvgS Δ65 in the respiratory tract of mice might similarly have enhanced colonization by BP-BvgS Δ65 . Our animal experiments, however, provided no indication that BP-BvgS Δ65 could colonize mouse lungs in the absence of reversion, arguing that modulation of BvgS activity did not occur in that organ. One caveat to our conclusion is that temporally or spatially restricted in vivo modulation might be missed in animal experiments. Of note, highlevel expression of several vrgs in mouse infection were recently reported [32]. Using a fluorescent reporter under the control of the ptx promoter in BP-BvgS Δ65 , we also found no evidence that modulation might occur inside macrophages or dendritic cells (our unpublished data), in line with recent proteomic analyses of B. pertussis that showed increased production of some virulence proteins inside human macrophages [33]. Nevertheless, as both Bvg i -and Bvg +phase strains can colonize the nose, we cannot exclude the possibility of up-modulation of BvgS Δ65 in the upper respiratory tract of mice. It is also possible that modulating signals are present in specific environments or conditions encountered by the bacteria in the human respiratory tract, but not in that of the mouse.
In the first animal experiment, colonies with hemolytic phenotypes progressively outnumbered non-hemolytic colonies. Transcriptomic analyses of those variants showed increased expression levels of several virulence factors relative to parental BP-BvgS Δ65 . The rapid increase of the proportions of those more virulent revertants and their ability to outcompete non-hemolytic clones suggest that they were selected for by innate immune responses in the mice, and that this selective pressure is very strong in the lungs. In the nose, in contrast, nonhemolytic bacteria in the Bvg i phase could survive for up to four weeks, similarly to those in the Bvg + phase. Nevertheless, the occurrence of both BP-BvgS Δ65 and hemolytic revertants in the noses of mice from day 3 de facto resulted in a mixed infection, and the initially low proportions of revertants in the noses increased relative to their Bvg i -phase parent over time. This shows that they may have a selective advantage in the upper respiratory tract as well, in agreement with previous reports that the prototypical Bvg i -phase-locked strain harboring the T 733 M substitution, was less competitive in both compartments [26]. Nonetheless, as selective pressure is less intense in the nose, it is conceivable that attenuated mutants could persist in that compartment in the absence of more virulent competitors. The report that avirulent B. pertussis clones harboring an IS481 insertion in bvgAS were detected in the nasopharynx of experimentally infected monkeys suggests that even avirulent bacteria might be able to persist for some time in the upper respiratory tract once infection has been established [34].
The transcription profile of BP-BvgS Δ65-SS1 shows that it is not a fully virulent Bvg + -phase strain, but an additional intermediate between the bona fide Bvg i and Bvg + phenotypes. Nevertheless, expression of vags in that strain appears to be sufficient to establish lung infection, indicating that the virulence phenotype of BP-BvgS Δ65-SS1 is beyond the threshold at which mouse lung colonization can occur. Whether such a moderately virulent strain might be successful in natural human infections is unknown, and it might be interesting to address this question in a baboon model of infection [1]. Its phenotype might be advantageous in the face of the immune pressure, notably linked to vaccination. The loss of specific virulence factors or lower levels of virulence gene expression have indeed been reported to occur in currently circulating B. pertussis strains [35][36][37][38][39]. On the contrary, however, immune pressure appears to currently select for strains that express greater levels of pertussis toxin [40]. Thus, distinct virulence factors make different contributions to the success of infection. In this respect, it is worth noting that a live attenuated vaccine candidate, BPZE1, which lacks some bona fide virulence factors, efficiently colonizes the respiratory tract of animals and humans without causing disease [41,42].
B. pertussis has conserved its capability to shift between the Bvg + phase and the Bvgor Bvg i phases in the laboratory. However, the relevance of the latter phases in the host and the conditions that trigger phenotypic modulation in its current lifestyle remain unclear. For B. pertussis, the shift to the Bvgphase might be a remnant of an ancestor able to survive outside mammalian hosts, as the Bvgphase of Bordetella bronchiseptica allows its growth and dissemination in amoeba as an environmental reservoir [43]. The occurrence and the role of the Bvg i phase in the infection remain unclear as well. The presence of antibodies directed against Bvg iphase-specific antigens in the sera of convalescent children was reported [44] but contradicted by subsequent work [26]. An in vivo shift to the Bvg i phase in the nasopharynx might be favored by the somewhat lower temperature there than in the rest of the respiratory tract, but a recent study revealed that virulence factor expression in B. pertussis persists at suboptimal temperatures [45]. Several studies including this one have shown that Bvg i -phase locked B. pertussis strains are proficient for infection of the mouse nose. However, Bvg + -phase locked B. bronchiseptica remains proficient for infection and transmission in a pig model of disease, casting doubt on the idea that the Bvg i phase contributes to transmission [27]. It is possible that the bacteria transiently shift to the Bvg i phase before or after transmission, but the Bvg i phase passes undetected by the immune system. It was proposed that virulence genes expressed by B. pertussis could be classified in four classes, and that vrgs were expressed mainly in absence of phosphorylated BvgA, that is, in the Bvgphase [4]. Nevertheless, we showed in this study that expression of several vrgs was at similarly high levels in the Bvg i -phase strains as in the modulated, Bvgphase wt strain. In contrast, vrgs were not overexpressed in the Bvg i -phase-locked strain described earlier [23]. Whether these differences stem from the different genetic backgrounds of the strains used in the two studies remains to be determined. It is also possible that the vrgs form distinct subgroups that are regulated in slightly different ways, as shown for vags [24].
The majority of hemolytic revertants of BPSM-BvgS Δ65 isolated from the mice had a specific mutation in linker X that results in the substitution of an Arg residue with a Leu. In the subset of BvgS homologs devoid of a PAS domain, the linker X harbors two antagonistic coiled-coil registers, and regulation of activity implies interconversion between the two marginally stable coiled coils defined by those registers [20]. In BvgS Δ65 , Arg572 appears to be involved in repulsive interactions between the two helices in the kinase register of the coiled coil, but not in the phosphatase register [20]. Therefore, the removal of unfavorable interactions by the Arg to Leu replacement is likely to stabilize the two-helix coiled coil in the kinase register, which might lock BvgS Δ65 in the kinase state of activity.
Finally, some of the hemolytic revertants from BPSM-BvgS Δ65 isolated from the noses of mice at early time points, did not harbor the R 572 L point mutation to account for their hemolytic phenotype. While whole-genome sequencing did not enable us to identify any other specific mutation, restoration of a hemolytic phenotype may have arisen from genomic reshuffling of insertion sequences or epigenetic modifications. The observation that the hemolytic phenotype appears to be maintained upon sub-culturing suggests that the BvgS Δ65 variant might generate bistability, in particular in the mouse nose. Of note, bistable phenotypes have been reported in the lower respiratory tract of mice infected with specific mutants of the broad-host range pathogen Bordetella bronchiseptica [46]. Supporting information S1 Table. Raw data of the RNA seq experiments. The RPKM for each open reading frame are provided for the 6 conditions. The corrected p-values and the fold changes (FC) are also provided. Fold changes > 2 or >-2 (in log2) are considered to be significant. (XLSX)