Kingella kingae PilC1 and PilC2 are adhesive multifunctional proteins that promote bacterial adherence, twitching motility, DNA transformation, and pilus biogenesis

The gram-negative bacterium Kingella kingae is a leading cause of osteoarticular infections in young children and initiates infection by colonizing the oropharynx. Adherence to respiratory epithelial cells represents an initial step in the process of K. kingae colonization and is mediated in part by type IV pili. In previous work, we observed that elimination of the K. kingae PilC1 and PilC2 pilus-associated proteins resulted in non-piliated organisms that were non-adherent, suggesting that PilC1 and PilC2 have a role in pilus biogenesis. To further define the functions of PilC1 and PilC2, in this study we eliminated the PilT retraction ATPase in the ΔpilC1ΔpilC2 mutant, thereby blocking pilus retraction and restoring piliation. The resulting strain was non-adherent in assays with cultured epithelial cells, supporting the possibility that PilC1 and PilC2 have adhesive activity. Consistent with this conclusion, purified PilC1 and PilC2 were capable of saturable binding to epithelial cells. Additional analysis revealed that PilC1 but not PilC2 also mediated adherence to selected extracellular matrix proteins, underscoring the differential binding specificity of these adhesins. Examination of deletion constructs and purified PilC1 and PilC2 fragments localized adhesive activity to the N-terminal region of both PilC1 and PilC2. The deletion constructs also localized the twitching motility property to the N-terminal region of these proteins. In contrast, the deletion constructs established that the pilus biogenesis function of PilC1 and PilC2 resides in the C-terminal region of these proteins. Taken together, these results provide definitive evidence that PilC1 and PilC2 are adhesins and localize adhesive activity and twitching motility to the N-terminal domain and biogenesis to the C-terminal domain.


Introduction
Kingella kingae is a gram-negative bacterium that colonizes the upper respiratory tract in young children [1]. K. kingae colonization is a common occurrence, with approximately 70% of children being colonized by 48 months of age and approximately 10% of young children carrying this bacterium in the oropharynx at any given time [2,3]. While K. kingae is usually a commensal organism, on occasion it breaches the epithelial barrier, enters the bloodstream, and spreads hematogenously to sites of infection, causing invasive diseases such as septic arthritis and osteomyelitis [4]. Recent advances in molecular diagnostics have identified K. kingae as a leading cause of osteoarticular infections in children between 6 months and 4 years of age [5,6].
Adherence to epithelial cells is believed to be the first step in K. kingae colonization of the oropharynx and a prerequisite for invasive disease. K. kingae adherence to host cells is a twostep process. The initial step is mediated by type IV pili (T4P), which then retract, bringing the bacterium closer to epithelial cells [7,8]. Retraction displaces the polysaccharide capsule, allowing a surface expressed adhesin called the Kingella NhhA homolog (Knh) to mediate the second step required for full-level adherence [8,9]. Strains that lack retractile T4P exhibit an intermediate level of adherence, due to the lack of Knh-mediated adherence [8].
The K. kingae PilC1 and PilC2 pilus-associated proteins promote piliation and T4P adhesive activity, with expression of at least one of these proteins being required for piliation and adherence to epithelial cells in vitro [7]. These proteins have homologs in other bacterial species, including PilY1 in Pseudomonas aeruginosa, PilY1 in Legionella pneumophila, and PilC1 and PilC2 in Neisseria gonorrhoeae and Neisseria meningitidis (constituting the "PilC family"). These PilC family proteins have been shown to play a role in pilus assembly, pilus-mediated adherence, and pilus-mediated twitching motility [10][11][12][13][14]. Previous observations by Morand et al. showed that the N-terminal portion of the N. meningitidis PilC1 protein was required for PilC1-mediated bacterial adherence in engineered N. meningitidis strains [15]. Additionally, in studies of the N. gonorrhoeae PilC1 protein, Cheng and colleagues found that full-length recombinant PilC1 was able to block N. gonorrhoeae adherence to cultured epithelial cells, while the truncated PilC1 C-terminal domain had no blocking activity [16]. Because K. kingae PilC1 and PilC2 are required for surface piliation, it is unclear whether these proteins possess inherent adhesive activity.
The K. kingae PilC1 and PilC2 proteins share limited homology with each other, with only 7% identity and 16% similarity overall [7], contrasting with the other known PilC-containing systems, which possess only one protein as in P. aeruginosa and L. pneumophila or two highly homologous proteins as in N. gonorrhoeae and N. meningitidis. Interestingly, in previous work we found that while both K. kingae PilC1 and PilC2 were able to promote twitching motility, K. kingae exhibited significantly different levels of twitching motility depending on which PilC protein was present, with PilC1 promoting a hyper-motile phenotype in the absence of PilC2 and with PilC2 mediating slightly reduced twitching in the absence of PilC1 [17].
In the present study we examined the specific functions of the K. kingae PilC1 and PilC2 proteins. We demonstrate that PilC1 and PilC2 are adhesins and have distinct binding specificities. In addition, we establish that the N-terminal domains of these proteins harbor adhesive activity and are essential for twitching motility. In contrast, the pilus assembly function resides in the C-terminal domain of these proteins.

Elimination of the K. kingae PilT retraction ATPase restores surface piliation but not adherence in a ΔpilC1ΔpilC2 mutant
In previous work, we observed that insertional inactivation of pilC1 and pilC2 in K. kingae resulted in an extreme piliation defect and a non-adherent phenotype [7], suggesting that PilC1 and PilC2 are critical for pilus assembly. To determine whether elimination of PilC1 and PilC2 results in a loss of adhesive activity simply because the resulting organisms are nonpiliated, we insertionally inactivated the K. kingae pilT gene in the KK03ΔpilC1ΔpilC2 double knockout strain (Table 1), using strains KK03 (wild type), KK03ΔpilA1 (lacking the major pilin subunit and thus lacking surface fibers), KK03ΔpilT (lacking PilT-mediated retraction), KK03ΔpilC1ΔpilT (lacking PilC1 and PilT-mediated retraction), and KK03ΔpilC2ΔpilT (lacking PilC2 and PilT-mediated retraction) as controls. As expected, the resulting mutant regained piliation to greater than 90% of the level of wild type strain KK03 as assessed by densitometry analysis of the PilA1 band in pilus preparations normalized to the GAPDH loading control (Fig 1A), consistent with previous work indicating that the pilT gene encodes an inner membrane ATPase that mediates pilus retraction in K. kingae and N. gonorrhoeae [8,18]. As shown in Fig 1B, while strains KK03ΔpilC1ΔpilT and KK03ΔpilC2ΔpilT were adherent in assays with Chang cells of HeLa origin, the KK03ΔpilC1ΔpilC2ΔpilT mutant was non-adherent ( Fig 1B), supporting the possibility that PilC1 and PilC2 have inherent adhesive activity.

K. kingae PilC1 and PilC2 proteins are adhesins with distinct protein structures
To further explore the possibility that PilC1 and PilC2 are adhesins, we purified recombinant N-terminal 6X histidine-tagged PilC1 and PilC2 (Fig 2A) and examined binding to epithelial cell monolayers. As shown in Fig 2B and 2C, PilC1 exhibited saturable binding with a K d of 74.4 nM, while PilC2 displayed saturable binding with a K d of 165.6 nM. Examination of the purified proteins by circular dichroism (CD) spectroscopy revealed that PilC1 and PilC2 have distinct CD spectra when comparing mean residue ellipticity (Fig 2D). Based on the CD spectra, PilC1 is predicted to contain 9.8% helices, 8.3% antiparallel sheets, 21.6% parallel sheets, and 10.9% turns (Fig 2E), while PilC2 is predicted to contain 7.0% helices, 11.6% turns, 42.5% antiparallel sheets, and no parallel sheets (Fig 2E). Taken together, these data demonstrate that the K. kingae PilC1 and PilC2 proteins are adhesins and have different binding affinities and distinct global protein structures.
The K. kingae PilC1 and PilC2 C-terminal domains promote production of pilus fibers that lack adhesive activity Previous work on the P. aeruginosa PilY1 protein established that the PilY1 C-terminal domain has a β-propeller fold and plays a critical role in P. aeruginosa type IV pilus assembly [26]. To localize the adhesive regions of PilC1 and PilC2, we generated deletion constructs lacking the N-terminal domains but containing intact C-terminal β-propeller folds (KK03ΔpilC2-ErmPilC1 Cterm and KK03ΔpilC1-KanPilC2 Cterm ) based on modeling with Phyre2 software [27] (S1 Fig). We hypothesized that the strains expressing only the C-terminal region of PilC1 or PilC2, in the absence of the other PilC, would still produce type IV pili but that these fibers would not be able to mediate adherence due to lack of the adhesive N-terminal region. Strains KK03ΔpilC2-ErmPilC1 and KK03ΔpilC1-KanPilC2, which contain an antibiotic marker upstream of the wild-type pilC1 gene or downstream of the pilC2 gene, respectively, were used as controls when evaluating the phenotypes of strains KK03ΔpilC2-ErmPilC1 Cterm and KK03ΔpilC1-KanPilC2 Cterm . As shown in Fig 4A, piliation was reduced in strain KK03ΔpilC2-ErmPilC1 compared to the parent strain KK03ΔpilC2, indicating that the erythromycin resistance marker upstream of pilC1 has a negative impact on pili levels. However, comparison of strains KK03ΔpilC2-ErmPilC1 and KK03ΔpilC2-ErmPilC1 Cterm revealed similar piliation levels, indicating that the PilC1 Cterm fragment is able to promote piliation at levels equivalent to full-length PilC1. Similarly, comparison of strain KK03ΔpilC1-KanPilC2 with strain KK03ΔpilC1-KanPilC2 Cterm revealed equivalent levels of piliation, indicating that (B, C) Adherence was detected by ELISA using polyclonal antiserum CHP-GP7 for PilC1 (B) or GP103 for PilC2 (C) and a secondary antiguinea pig antibody conjugated to horseradish peroxidase (HRP). Non-linear regressions were fit using the GraphPad Prism one site specific binding model fit to total data from 3 independent runs, and the K d was calculated based on the regression. Error bars represent standard error of the mean, n = 3. (D, E) Circular dichroism was carried out on purified PilC1 and PilC2 (D) and the resultant spectra were used to predict protein secondary structures (E). The CD spectra are representative graphs of three independent analyses carried out on different batches of purified protein. https://doi.org/10.1371/journal.ppat.1010440.g002 the PilC2 C-terminal domain is also able to promote similar piliation levels as full-length PilC2. Despite the presence of pili, these strains were non-adherent in assays with cultured Chang cells of HeLa origin (Fig 4B and 4C), suggesting that the N-terminal domains of PilC1 and PilC2 are critical for adhesive activity. These results indicate that the C-terminal domains of PilC1 and PilC2 are sufficient for pilus assembly and suggest that the N-terminal domains of PilC1 and PilC2 harbor adhesive activity.

The K. kingae PilC1 and PilC2 N-terminal domains bind to epithelial cell monolayers
To obtain definitive evidence that the adhesive activity of PilC1 and PilC2 resides in the N-terminal domains of these proteins, we purified recombinant truncated 6X histidine-tagged PilC1 and PilC2 proteins corresponding to the N-terminal domains, lacking the predicted βpropeller region (Figs 5A and S1). As shown in Fig 5B,

The PilC1 and PilC2 C-terminal domains are not sufficient for twitching motility but support transformation
To begin to localize other PilC1-and PilC2-mediated functions, we examined twitching motility in strains KK03ΔpilC2-ErmPilC1 Cterm and KK03ΔpilC1-KanPilC2 Cterm . As shown in Fig  6A, compared to strain KK03ΔpilC2-ErmPilC1, strain KK03ΔpilC2-ErmPilC1 Cterm was devoid of twitching motility, similar to the non-piliated ΔpilA1 mutant. Compared to strain KK03ΔpilC1-KanPilC2, strain KK03ΔpilC1-KanPilC2 Cterm was also devoid of twitching motility. These results suggest that the C-terminal domains of PilC1 and PilC2 are not sufficient for promoting twitching motility, localizing this activity to either the N-terminal domains or to regions in both the N-terminal and the C-terminal domains. Twitching motility requires sequential pilus extension, adhesion to a substrate, and retraction. To determine if the defect in twitching motility in KK03ΔpilC2-ErmPilC1 Cterm and KK03ΔpilC1-KanPilC2 Cterm was due to a defect in pilus retraction, we examined natural transformation in these strains, recognizing that natural transformation requires pilus retraction in other systems [28]. The WT strain exhibited transformation efficiencies ranging from 0.3% to 1.5%, normalized to 1.0 in Fig 6B. As expected, the non-piliated ΔpilA1 mutant and the , and KK03ΔpilC1-KanPilC2 Cterm were boiled and separated using SDS-PAGE. KK03ΔpilF lacks the PilF assembly ATPase and does not assemble PilA1 into pilus fibers. For pili preps, the PilA1 pilin monomer band was stained with Coomassie blue. For whole cell lysates, the PilA1 pilin monomer band was detected by Western blot analysis using polyclonal antiserum GP65 [52] to PilA1. GAPDH was detected by Western blot analysis to control for total protein using polyclonal antiserum CHP-GP22. The values above the gel and blot images are densitometry measurements of the PilA1 band from the sheared pilus preparations (pili prep) gel that were normalized to the α-GAPDH band intensities and are expressed as a percentage of the band intensity of the wild type strain KK03. The densitometry values are listed ± standard error of the mean from three biological replicates, and representative gel and blot images are shown. (B, C) Strains KK03, KK03ΔpilA1, KK03ErmΔpilC2-PilC1, KK03ΔpilC2-ErmPilC1 Cterm , KK03ΔpilC1-KanPilC2, and KK03ΔpilC1-KanPilC2 Cterm were added to monolayers of Chang epithelial cell of HeLa origin and evaluated for adherence. Percent adherence was calculated based on the ratio of recovered bacteria to the inoculum. Error bars represent standard error of the mean, n = 3. Cyan color denotes strains expressing either fulllength PilC1 or the C-terminal region of PilC1 only. Magenta color denotes strains expressing either full-length PilC2 or the C-terminal region of PilC2 only. � indicates significance of P < 0.05 as determined by a one-way ANOVA using Bonferroni correction for multiple comparisons.
https://doi.org/10.1371/journal.ppat.1010440.g004 retraction-deficient ΔpilT mutant were completely non-transformable. Strain KK03ΔpilC1Δ-pilC2 was capable of low but measurable levels of transformation, while strains KK03ΔpilC2-ErmPilC1 and KK03ΔpilC1-KanPilC2 displayed wild type levels of transformation, suggesting that the presence of either PilC1 or PilC2 is sufficient for wild type transformation efficiency. While strain KK03ΔpilC1-KanPilC2 Cterm displayed wild type transformation efficiency, strain KK03ΔpilC2-ErmPilC1 Cterm exhibited a statistically significant decrease in transformation efficiency when compared to KK03ΔpilC2-ErmPilC1. This lower level of transformation was above the level observed for strain KK03ΔpilC1ΔpilC2, indicating that the PilC1 C-terminal domain promotes transformation and that full transformation efficiency requires additional sequence N-terminal to the C-terminal domain fragment. Taken together, these data indicate that the PilC1 and PilC2 C-terminal domains are sufficient for retraction, and thus the lack of twitching motility in KK03ΔpilC2-ErmPilC1 Cterm and KK03ΔpilC1-KanPilC2 Cterm is not a consequence of absent pilus retraction.

Discussion
In this study we examined the role of the K. kingae PilC1 and PilC2 proteins in K. kingae adherence. Using purified full-length and truncated PilC1 and PilC2, we established that PilC1 and PilC2 are adhesins, with the adhesive activity located in the N-terminal domains of these proteins. Using deletion constructs, we found that PilC1-mediated and PilC2-mediated twitching motility requires the N-terminal domains. In contrast, our results indicate that the C-terminal domains of PilC1 and PilC2 are sufficient for promoting pilus assembly and natural transformation, although the PilC1 C-terminal domain is not sufficient for wild type transformation efficiency. Adherence was detected by ELISA using polyclonal antiserum CHP-R1 for PilC1 Nterm (B) or GP103 for PilC2 Nterm (C) and a secondary anti-rat or anti-guinea pig antibody conjugated to HRP. Non-linear regressions were fit using the GraphPad Prism one site specific binding model fit to total data from 3 independent runs, and the K d was calculated based on the regression. Error bars represent standard error of the mean, n = 3. (D, E) Circular dichroism was carried out on purified PilC1 Nterm and PilC2 Nterm (D) and the resultant spectra were used to predict protein secondary structures (E). The CD spectra are representative graphs of three independent analyses carried out on different batches of purified protein. Earlier studies suggested that the K. kingae PilC1 and PilC2 proteins might be the adhesive components of the pilus, as expression of either the PilC1 or PilC2 protein is required for bacterial adherence to epithelial cells [7]. In this work, we found that deletion of the PilT retraction ATPase restored production of K. kingae surface fibers in the absence of a functional PilC protein but did not restore adherence. Furthermore, we observed that purified PilC1 and PilC2 were capable of binding to epithelial cell monolayers to saturating levels, providing strong evidence that these proteins are adhesins that interact with epithelial cells in a receptormediated process. Interestingly, the predicted K d measurements of the K. kingae PilC1 and PilC2 N-terminal domains mimicked results with the full-length proteins, supporting the conclusion that full adhesive activity resides in the N-terminal domains [16].
Adhesins expressed by host-adapted bacteria commonly bind to extracellular matrix (ECM) components. Heiniger et al. showed that P. aeruginosa preferentially binds to the exposed basolateral cell surface of injured polarized epithelial cells in a PilY1-dependent manner [13]. As ECM proteins are located at the basolateral side of epithelial cells in the basement membrane and connective tissue, they are well positioned to serve as targets for bacterial adherence when there is damage to the epithelium. In this report, we demonstrate that K. kingae PilC1, but not PilC2, mediates bacterial adherence to ECM proteins. During the K. kingae pathogenic process, damage to the epithelial surface mediated by the K. kingae RTX toxin [21] or viral infection [29][30][31][32] may promote bacterial access to the basement membrane, where PilC1 may bind to ECM proteins, potentially facilitating invasive disease.
Our studies examining adherence to ECM proteins establish that the K. kingae PilC1 and PilC2 proteins have distinct binding specificities, consistent with the limited homology between PilC1 and PilC2 and the different global structures as assessed by CD spectroscopy. Interestingly, the N. meningitidis PilC1 and PilC2 proteins also have distinct binding specificities, despite the fact that these proteins are highly homologous. In particular, N. meningitidis PilC1 has been demonstrated to promote meningococcal adherence to HEC-1-B, ME180, and HUVEC cell lines, while PilC2 promotes adherence only to ME180 cells [33,34].
It is notable that K. kingae PilC1-mediated adherence to ECM proteins was increased in the absence of PilC2, similar to our earlier observation that PilC1-mediated twitching motility was increased in the absence of PilC2 [17]. We have established that PilC1 and PilC2 production is not affected by the presence or the absence of the other protein (S2 Fig), suggesting that changes in expression cannot account for the differences in the phenotypes observed. One possibility is that PilC2 obscures PilC1 along the pilus and interferes with PilC1 interaction with its receptor. Another possibility is that PilC1 localization along the pilus is affected when PilC2 is eliminated. A third possibility is that PilC1 and PilC2 interact with each other and modify the function of each other.
The PilC family of proteins share a predicted C-terminal β-propeller fold domain. This fold was first identified when the C-terminal region of the P. aeruginosa PilY1 protein was crystallized [26], revealing a calcium-binding motif that is present in the β-propeller region of multiple PilC proteins [16,17,26,35]. Orans et al. showed that prevention of calcium-binding resulted in a defect in surface piliation and twitching motility, while mimicking the charge of a bound calcium ion produced an abundance of non-functional surface pili [26], suggesting a role for this domain in pilus biogenesis and retraction. Porsch et al. demonstrated that the K. kingae PilC1 and PilC2 calcium-binding motifs promoted twitching motility [17]. In this study, we observed that the C-terminal domain of K. kingae PilC1 and PilC2 was sufficient for promoting surface piliation, consistent with a role for the C-terminal domain in supporting pilus dynamics.
Studies in gram-negative bacteria have shown that T4P components and PilC proteins promote natural competence for transformation [28,[36][37][38]. Because the expression of at least one PilC protein is required for surface piliation, ascertaining whether the proteins promote DNA uptake by promoting piliation or through another mechanism has been a challenge. KK03ΔpilC2-ErmPilC1 Cterm displays reduced transformation efficiency compared to strain KK03ΔpilC2-ErmPilC1 despite expressing similar levels of pili (Fig 4A), suggesting that factors beyond piliation level influence transformation efficiency. Other possible mechanisms of PilC protein involvement include direct binding of DNA or influencing other known competence proteins such as the pilus-associated ComP, which binds to DNA [39,40], or the periplasmic ComE or ComA, which function to promote DNA uptake and translocation [41][42][43][44][45][46][47]. As PilC1 and PilC2 localize to surface pili (S2 Fig), it is feasible that they may interact with ComE or ComA in the periplasm to modulate DNA translocation, or ComP along the pilus fiber to modulate DNA binding. The transformation defect observed in KK03ΔpilC2-ErmPilC1 Cterm but not KK03ΔpilC1-KanPilC2 Cterm raises the possibility that PilC1 and PilC2 have different functions during transformation. The PilC1 N-terminal region may function in combination with the C-terminal region to promote transformation, while the PilC2 C-terminal region may function alone to promote transformation using a distinct mechanism.
Twitching motility is believed to be carried out by sequential T4P extension, adhesion to a substrate, and PilT-mediated retraction of the adhering pilus fibers, which exert a pulling force allowing the bacteria to move along a surface [28,[48][49][50][51]. In PilC-containing bacteria, at least one functional PilC protein is required for a twitching phenotype [10,17,26]; however, it remains to be determined if the PilC proteins promote twitching motility by promoting adherence, by regulating the pilus extension/retraction dynamic, or both. Strains KK03ΔpilC2-Erm-PilC1 Cterm and KK03ΔpilC1-KanPilC2 Cterm produce retractile pilus fibers, as evidenced by their natural transformation. Despite being retractile, the present fibers are unable to promote

PLOS PATHOGENS
Characterization of the Kingella kingae PilC1 and PilC2 adhesins bacterial adherence to epithelial cells. For this reason, the defect observed in twitching motility is likely due to a defect in adherence, rather than a defect in pilus retraction.
Our data demonstrate that K. kingae PilC1 and PilC2 differentially promote K. kingae adherence using the N-terminal domains to directly interact with epithelial cells and the C-terminal domains to promote surface piliation. Future studies should elucidate the specific adhesive motifs in PilC1 and PilC2 and should address whether PilC1 and PilC2 interact with each other to modulate type IV pilus-mediated phenotypes.

Generation of K. kingae mutants
K. kingae gene disruptions and mutations were generated as described previously [7,17,52]. Plasmid-based disruption and truncation constructs were generated in E. coli, linearized, and introduced into K. kingae using natural transformation of linearized plasmid DNA followed by selection for mutants on chocolate agar plates with the appropriate antibiotic. Mutations were confirmed by genomic DNA preparation of putative mutant strains followed by PCR amplification and evaluation by Sanger sequencing.
To generate the unmarked KK03ΔpilC2 mutant, the plasmid pUC19/ΔpilC2 [17] lacking a resistance cassette was transformed into the kanamycin-marked strain KK03ΔpilC2 using the spot transformation technique [53], and patch plating was used to screen for the loss of kanamycin resistance. To generate strain KK03ΔpilC1ΔpilC2, linearized pUC19/ΔpilC1 [7] was introduced into KK03ΔpilC2 and transformants were recovered by selection on chocolate agar with 2 μg/ml tetracycline. To generate strain KK03ΔpilC2ΔpilT, linearized pUC19/ΔpilT [17] DNA was introduced into strain KK03ΔpilC2 and transformants were recovered by selection on chocolate agar with 1 μg/ml erythromycin. To generate strain KK03ΔpilC1ΔpilT, linearized pUC19/ΔpilT [17] DNA was introduced into strain KK03ΔpilC1 [17], as described above. To generate strain KK03ΔpilC1ΔpilC2ΔpilT, linearized pUC19/ΔpilT [17] DNA was introduced into strain KK03ΔpilC1ΔpilC2 as described above. For pUC19/ΔpilA1:tet, the aphA3 kanamycin resistance cassette in pUC19/ΔpilA1:kan [7] was excised as a MluI-fragment and replaced with a tetM tetracycline resistance cassette. For strain KK03ΔpilA1, linearized plasmid pUC19/ ΔpilA1:tet DNA was introduced into KK03 and transformants were recovered by selection on chocolate agar with 2 μg/ml tetracycline.
To generate strain KK03ΔpilC2-ErmPilC1 Cterm , site-directed mutagenesis using primers C1ΔNterm_mut_sense and C1ΔNterm_mut_anti (see Table 2 for the primer sequences used in this study) and the QuikChange II XL Site-directed mutagenesis kit (Agilent, Santa Clara, CA) was used to delete sequence encoding amino acids 32-438, using plasmid pUC19/Erm-pilC1 [17] as the template. The C-terminal domains were designated as the region of the proteins encompassing the predicted β-propeller fold based on modeling with Phyre2 software [27], and the N-terminal regions were defined as the polypeptide excluding the signal sequence and the predicted β-propeller fold. The resulting construct, pUC19/ErmpilC1ΔNterm, which contains an erythromycin resistance marker upstream of the pilC1 promoter region and the PilC1 signal sequence fused in-frame to the C-terminal domain, was linearized and transformed into strain KK03ΔpilC2, generating strain KK03ΔpilC2-ErmPilC1 Cterm . To generate strain KK03ΔpilC1-KanPilC2 Cterm , Gibson assembly was employed using DNA fragments amplified with Q5 Hi-Fidelity Master Mix with primer pairs pilC2up_F and pilC2up_R (Table 2) (using KK03 genomic DNA as template) and pilC2downKan_F and pilC2down-Kan_R (using KK03ΔpilC1-KanPilC2 [17] genomic DNA as the template) and EcoRI-digested pUC19. The resulting plasmid, pUC19/pilC2ΔNtermKan, which contains sequence encoding the PilC2 signal sequence fused in-frame to sequence encoding the C-terminal domain and a kanamycin-resistance marker downstream of the pilC2 ORF, was linearized and transformed into strain KK03ΔpilC1, generating strain KK03ΔpilC1-KanPilC2 Cterm . A schematic representation of the pilC1 and pilC2 loci in the panel of mutants used in this study is shown in S3 Fig. For the transformation efficiency experiments, the plasmid pTrc99a/knh:tetM [8] was modified by excising the tetM cassette as a ClaI fragment and inserting the aphA3 kanamycin resistance cassette or the ermC erythromycin resistance cassette, generating pTrc99a/knh:aphA3 and pTrc99a/knh:ermC, respectively.

Generation of recombinant PilC1 and PilC2 constructs
Sequences corresponding to the ORFs of pilC1 and pilC2 lacking sequence encoding the predicted signal sequences were amplified from genomic DNA from either KK03 or KK03 producing 6X histidine-tagged PilC1 using Q5 High-Fidelity Master Mix (New England Biolabs, Ipswich, MA) with the primers GibCytoC1F and GibCytoC1R for PilC1 and GibCytoC2F and GibCytoC2R for PilC2. The signal sequence cleavage sites were predicted using SignalP [54,55]. The 6X histidine tag was placed at the N-terminus of the protein immediately following the predicted signal peptide cleavage site. The recombinant PilC1 full-length protein encompassed amino acids Asp 35 -Phe 1359 , and the recombinant PilC2 full-length protein encompassed amino acids Asn 37 -Phe 1502 . Gibson assembly was used to introduce sequence encoding 6X histidine-tagged PilC1 and PilC2 into the vector pET22b, generating pET22b-PilC1 and pET22b-PilC2, respectively. For PilC2 an N-terminal 6X histidine tag was added using primers GibCytoC2F and GibCytoC2R. The assembly reaction was carried out using the Gibson Assembly Master Mix (New England Biolabs)

PLOS PATHOGENS
Characterization of the Kingella kingae PilC1 and PilC2 adhesins according to the manufacturer's instructions, and the reaction was transformed into electrocompetent E. coli DH5α followed by the selection for transformants on LB agar with 100 μg/ml ampicillin. The plasmids were purified, and the constructs were confirmed using Sanger sequencing. The N-terminal regions of PilC1 and PilC2 were amplified from pET22b-PilC1 and pET22b-PilC2 using Q5 High-fidelity Master Mix with primers CNtermF and C1NtermR for PilC1 and CNtermF and C2NtermR for PilC2. The C-terminal domains were designated as the region of the proteins encompassing the predicted β-propeller fold based on modeling with Phyre2 software [27], and the N-terminal regions were defined as the polypeptide excluding the signal sequence and the predicted β-propeller fold. The amplified PilC1 N-terminal region encompassed amino acids Asp 35 -Gln 672 , while the PilC2 N-terminal region encompassed amino acids Asn 37 -Ser 1022 . Gibson assembly was used to introduce sequence encoding the 6X histidine-tagged PilC1 N-terminal region and the 6X histidine-tagged PilC2 N-terminal region into pET22b, generating pET22b-PilC1 Nterm and pET22b-PilC2 Nterm , which were purified and confirmed using Sanger sequencing. The confirmed constructs were introduced into electrocompetent E. coli BL21 followed by selection for transformants on LB agar with 100 μg/ml ampicillin.

PilC1 and PilC2 protein purification
Overnight cultures of E. coli BL21 with either pET22b-PilC1, pET22b-PilC2, pET22b-PilC1 Nterm , or pET22b-PilC2 Nterm were back-diluted 1:200 in LB broth with 100 μg/ml ampicillin. After reaching an OD 600 of 0.4, gene expression was induced for 3 hours at 30˚C using 0.04 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Subsequently, the cultures were centrifuged at 6700 x g for 20 minutes, and the supernatant was discarded. The cell pellets were resuspended in 50 mM tris-HCl, 5 mM EDTA, and 10 mM NaCl at pH 8.0 at a ratio of 3 ml buffer to 1 gram of cells. AEBSF was added to a final concentration of 1 mM. For each 1 gram of cells, 0.8 mg of lysozyme was added. The solution was mixed and placed in a 37˚C water bath until viscous. The solution was sonicated 3 x at 25% amplitude for 30 seconds using a QSonica Q500 sonicator to shear DNA. The solution was centrifuged at 39,191 x g for 30 minutes, and the supernatant was discarded. The resulting pellets were resuspended in 20 mM Na 2 HPO 4 , 20 mM NaCl, 5 mM EDTA, and 25% sucrose at pH 7.2 at a ratio of 3 ml buffer to 1 gram of material. AEBSF was again added to a final concentration of 1 mM. Triton X-100 was added to a final concentration of 1%. The solution was mixed and centrifuged at 48,384 x g for 20 minutes, and the supernatant was discarded. The pellets were solubilized using 50 mM tris-HCl, 40 mM imidazole, 8 M urea, and 1mM β-mercaptoethanol at pH 8.0 (solubilization buffer) for several hours at 37˚C, adding more buffer until no more pellet would solubilize. The solubilized pellet was centrifuged at 48,384 x g for 20 minutes. The remaining supernatant was filtered using vacuum filtration through a 0.22 μM membrane, and the 6X histidine-tagged PilC1, PilC2, PilC1 Nterm , or PilC2 Nterm was purified from the supernatant using affinity chromatography over an Ni-NTA agarose column. The 2 ml column bed volume was equilibrated with the solubilization buffer, and the supernatants were added to the column and incubated at room temperature on a rotator for 3 hours to facilitate protein-binding. The column was washed wth 20 ml of solubilization buffer. Proteins were eluted with 50 mM tris-HCl, 500 mM imidazole, 8 M urea, and 1 mM β-mercaptoethanol at pH 8.0. The samples were concentrated over a 100,000 Da molecular weight cutoff filter for PilC1 and PilC2 or a 50,000 Da molecular weight cutoff filter for PilC1 Nterm and PilC2 Nterm . The samples were then dialyzed into 1 L 50 mM tris-HCl at pH 8.5 at 4˚C for 24 hours. The buffer was removed and replaced with 1 L fresh buffer, and samples were dialyzed at 4˚C for an additional 24 hours. The protein samples were stored at 4˚C in 50 mM Tris-HCl at pH 8.5.

Generation of polyclonal antisera
To generate a guinea pig antiserum to PilC1, the sequence encoding PilC1 minus the predicted signal peptide was amplified from KK03 genomic DNA with primers pBADpilC1_F and pBADpilC1_R and was cloned into KpnI-digested pBAD18 using Gibson assembly. The resulting plasmid pBAD-pilC1 was transformed into electrocompetent DH5α, and gene expression was induced with 0.2% arabinose. The bacterial pellet was harvested and sonicated. After centrifugation, the insoluble fraction containing recombinant PilC1 was separated on a 7.5% SDS-PAGE gel and stained with Coomassie blue, and the PilC1 band was excised and sent to Cocalico Biologicals to immunize a guinea pig (CHP-GP7) according to their standard antiserum generation protocol. To generate a rat antiserum to PilC1, 6XHisPilC1 was used to immunize a rat (CHP-R1) at Cocalico Biologicals according to their standard antiserum generation protocol. To generate an antiserum to PilC2, a PilC2 fragment corresponding to amino acids 868-1502, generated as previously described [17], was used to immunize a guinea pig (GP103) at Cocalico Biologicals according to their standard antiserum generation protocol. To generate a rabbit polyclonal antiserum to the PilC1 N-terminal region, recombinant PilC1 Nterm was used to immunize a rabbit (Rab128) at Cocalico Biologicals according to their standard antiserum generation protocol.
To generate an antiserum to the K. kingae glyceraldehyde phosphate dehydrogenase (GAPDH) protein, the full gapdh gene sequence (minus the start codon) was amplified using Q5 High-Fidelity Master Mix with primers gapdh_F and gapdh_R using KK03 genomic DNA as template. The resulting amplicon was digested with EcoRI and BamHI and ligated into EcoRI/BamHI-digested pHAT10, generating plasmid pHAT10-gapdh, encoding N-terminal histidine affinity tag (HAT)-tagged GAPDH. The plasmid was transformed into E. coli BL21 (DE3), and expression of the recombinant gapdh gene was induced with 0.4 mM IPTG for 3 hours at 30˚C. The resulting bacterial pellet was suspended in binding/wash buffer (20 mM sodium phosphate pH 7.4, 500 mM NaCl, 40 mM imidazole), sonicated, clarified via centrifugation at 20,000 x g for 20 min, and applied to a pre-equilibrated 5 ml HisTrap column (Cytiva, Marlborough, MA), using an AKTA Protein Purifier 10. After washing with 10 column volumes of binding/wash buffer, the bound protein was eluted with 20 mM sodium phosphate pH 7.4, 500 mM NaCl, 500 mM imidazole. The fractions containing the fusion protein were pooled, buffer exchanged and concentrated in 20 mM sodium phosphate pH 7.4, 500 mM NaCl using a 10,000 Da molecular weight cutoff filter, and then frozen at -80˚C. The purified fusion protein was then sent to Cocalico Biologicals for injection into a guinea pig (CHP-GP22) using their standard antiserum generation protocol. Reactivity of all antisera was assessed using Western blot analysis.

Cell-based protein-binding assays
Chang epithelial cells of HeLa origin were seeded into 96 well tissue culture-treated plates at a density of 3.24 x 10 4 cells per well, and the plates were incubated overnight for~18 hours at 37˚C, 5% CO 2 . The cells were then fixed with 2% glutaraldehyde in sodium phosphate buffer and washed 3x with Tris-buffered saline (TBS). The monolayers were blocked using 2% dry milk powder in phosphate-buffered saline (PBS) for 1.5 hours at 37˚C . Protein dilutions ranging from 1.85 nM to 205.55 nM were prepared in 50 mM Tris-HCl, pH 8.5. The blocking buffer was removed, and the diluted proteins were added to either tissue culture treated plates only or tissue culture treated plates coated with epithelial cell monolayers. Protein-loaded plates were incubated at 37˚C for 3 hours before being washed 4 x with PBS. Polyclonal antisera diluted 1:500 in 2% dry milk powder/PBS was added to the plates, which were incubated at 37˚C for 45 minutes and then washed 4 x with PBS. A secondary antibody conjugated to horseradish peroxidase diluted 1:2000 in 2% dry milk powder/PBS was added to the plates, and plates were incubated at 37˚C for 45 minutes before being washed 4 x with PBS. Peroxidase substrate (3, 3', 5, 5'-Tetramethylbenzidine) was added to the plates, and the color change was measured at 655 nm after 11 minutes of development using a multimode plate reader. Non-specific protein adherence to the plate and antibody adherence to epithelial cells were subtracted from the total protein adherence detected to epithelial cell monolayers to generate specific binding data.

Circular dichroism
PilC1, PilC2, PilC1 Nterm , and PilC2 Nterm proteins were diluted to a final concentration of 50 μg/ml in 16 mM tris-HCl, pH 8.5 and analyzed using a Jasco J-810 spectropolarimeter, using a 0.1 cm cuvette at a wavelength range of 250-190 nm. The spectropolarimeter was blanked using 16 mM Tris-HCl, pH 8.5 prior to taking measurements. The parameters were set to 1 nm data pitch, standard sensitivity, 1 s DIT, 1 nm bandwidth, immediate start mode, 20 nm/min scan speed, and 6 total accumulations. Spectra were converted to mean residue ellipticity to account for differences in the size of the proteins, and protein secondary structures were predicted using BeStSel [56,57].

Quantitative bacterial adherence assays
Quantitative adherence assays were performed as described previously [7,8,17,52]. For determining adherence levels, Chang epithelial cells of HeLa origin were seeded into 24-well tissue culture treated plates and incubated overnight at 37˚C, 5% CO 2 . The cells were then fixed with 2% glutaraldehyde in 0.2 M sodium phosphate buffer pH 7.4 and washed 3 x with TBS. For determining adherence to extracellular matrix (ECM), BioCoat plates coated with either collagen I, collagen IV, fibronectin, or laminin were purchased from Corning (Corning, NY). The ECM plates were equilibrated at room temperature for one hour prior to inoculation. The bacteria were cultured on chocolate agar plates at 37˚C, 5% CO 2 for 18-20 hours, swabbed from the plate, and resuspended in brain heart infusion media (BHI) to an OD 600 of 0.8. A volume of 10 μl of resuspended bacteria was added to 300 μl of prewarmed 37˚C MEM in 24-well tissue culture treated plates coated with either epithelial cell monolayers or ECM. The plates were incubated at 37˚C, 5% CO 2 for 25 minutes and were then washed 4 x with PBS to remove unbound bacteria. A volume of 100 μl 0.05% trypsin-EDTA was added to the plates, followed by incubation at 37˚C, 5% CO 2 for 20 minutes to facilitate bacterial recovery. The recovered bacteria were diluted and plated onto chocolate agar, and percent adherence was calculated based on the ratio of recovered bacteria to the inoculum.

Pilus preparations
Pilus preparations were performed using a large-scale method modified from the small-scale method described previously [17]. Bacteria were grown at 37˚C, 5% CO 2 for 20 hours on chocolate agar plates, swabbed from the plate, and resuspended in 12 ml PBS to an OD 600 of 0.8. Samples were vortexed at full speed for 1 min and centrifuged at 4,000 x g for 30 min to pellet bacteria. A total of 10 ml of the bacteria-free supernatant was subjected to 20% ammonium sulfate precipitation on ice for 2 hours. Precipitated pili were collected via centrifugation at 20,000 x g for 20 min and resuspended in 1x SDS-PAGE loading buffer. Pilus preparations were separated on 15% SDS-PAGE gels, stained with Coomassie blue, and imaged with a Syngene G:Box system.

Densitometry analysis
To quantitate piliation levels, the PilA1 major pilin subunit band densities of Coomassie bluestained pilus preparations and the Western blot GAPDH band densities of whole cell lysates (from the same bacterial pellet from which the pili were sheared for the pilus prepartions) were measured using ImageJ software [58,59]. The PilA1 band densities were divided by their matched GAPDH loading control band densities for normalization. The normalized values were then divided by the wild type KK03 normalized value to set wild type to 100% and compare all of the other strains to the wild type level. This analysis was completed for three independent biological replicates, and the average relative normalized PilA1 levels are presented ± standard error of the mean.

Transformation efficiency assays
Strains were suspended to an OD 600 of 0.8 in BHI broth, and 250 μl of the bacterial suspension was added to wells of a 24-well plate. A total of 1.0 μg of plasmid DNA containing either a kanamycin or erythromycin resistant cassette in place of the knh gene, using a plasmid backbone previously used to generate a knh deletion via allelic exchange [8], was added to the bacterial suspension. The transformation mixture was left at room temperature for 30 minutes, followed by the addition of 250 μl of BHI/20% lysed horse blood. The mixture was incubated at 37˚C, 5% CO 2 for 2.5 hours to allow for bacterial recovery before plating on chocolate agar containing 50 μg/ml kanamycin or 1 μg/ml erythromycin. The plates were incubated at 37˚C, 5% CO 2 overnight, and transformant colony forming units (cfu) were then enumerated. Transformation efficiency was calculated based on the ratio of kanamycin-resistant or erythromycin-resistant transformed cfu relative to the inoculum cfu.

Twitching motility assays
Twitching motility assays were performed as described previously [17]. The strains were suspended to an OD 600 of 0.8 in BHI. A 1 μl volume of the bacterial suspension was stab inoculated into the center of a chocolate agar motility plate (chocolate agar with 1% agar) to the plate-agar interface using a pipette tip. Plates were incubated at 37˚C, 5% CO 2 for three days. After three days the chocolate agar was removed, and the zone of bacterial spread at the plateagar interface was stained with 0.1% crystal violet. The diameter of the crystal violet-stained bacterial spread was measured in millimeters.

PLOS PATHOGENS
Characterization of the Kingella kingae PilC1 and PilC2 adhesins using polyclonal antiserum GP103 to PilC2, and GAPDH was detected by Western blot analysis using polyclonal antiserum GP22 to GAPDH. The PilA1 pilin monomer band was stained with Coomassie blue. (TIF) S3 Fig. Cartoon representation of the pilC1 and pilC2 loci and pilC1 and pilC2 mutants used in this study. Strain KK03 (WT) produces PilC1 and PilC2, which are encoded by the pilC1 and pilC2 genes. Strain KK03ΔpilC1ΔpilC2 contains a tetracycline resistance cassette in place of pilC1 and an unmarked deletion of pilC2 and does not produce PilC1 or PilC2. Strain KK03ΔpilC2 contains an unmarked deletion of pilC2 and produces full-length PilC1. Strains KK03ΔpilC2-ErmPilC1 and KK03ΔpilC2-ErmPilC1 Cterm contain an unmarked deletion of pilC2 and an erythromycin resistance cassette upstream of pilC1 and produce full-length PilC1 and the C-terminal domain of PilC1, respectively. Strain KK03ΔpilC1 contains a tetracycline resistance cassette in place of pilC1 and produces full-length PilC2. Strains KK03ΔpilC1-KanPilC2 and KK03ΔpilC1-KanPilC2 Cterm contain a tetracycline resistance cassette in place of pilC1 and a kanamycin resistance cassette downstream of pilC2 and produce full-length PilC2 and the C-terminal domain of PilC2, respectively. SS represents the predicted signal sequence. Cyan color denotes strains producing either full-length PilC1 or the C-terminal region of PilC1; magenta color denotes strains producing either full-length PilC2 or the C-terminal region of PilC2. The dashed lines indicate a deletion. Tet R , Kan R , and Erm R indicate tetracycline, kanamycin, and erythromycin resistance cassettes, respectively. (TIF)