Kingella kingae Expresses Four Structurally Distinct Polysaccharide Capsules That Differ in Their Correlation with Invasive Disease

Kingella kingae is an encapsulated gram-negative organism that is a common cause of osteoarticular infections in young children. In earlier work, we identified a glycosyltransferase gene called csaA that is necessary for synthesis of the [3)-β-GalpNAc-(1→5)-β-Kdop-(2→] polysaccharide capsule (type a) in K. kingae strain 269–492. In the current study, we analyzed a large collection of invasive and carrier isolates from Israel and found that csaA was present in only 47% of the isolates. Further examination of this collection using primers based on the sequence that flanks csaA revealed three additional gene clusters (designated the csb, csc, and csd loci), all encoding predicted glycosyltransferases. The csb locus contains the csbA, csbB, and csbC genes and is associated with a capsule that is a polymer of [6)-α-GlcpNAc-(1→5)-β-(8-OAc)Kdop-(2→] (type b). The csc locus contains the cscA, cscB, and cscC genes and is associated with a capsule that is a polymer of [3)-β-Ribf-(1→2)-β-Ribf-(1→2)-β-Ribf-(1→4)-β-Kdop-(2→] (type c). The csd locus contains the csdA, csdB, and csdC genes and is associated with a capsule that is a polymer of [P-(O→3)[β-Galp-(1→4)]-β-GlcpNAc-(1→3)-α-GlcpNAc-1-] (type d). Introduction of the csa, csb, csc, and csd loci into strain KK01Δcsa, a strain 269–492 derivative that lacks the native csaA gene, was sufficient to produce the type a capsule, type b capsule, type c capsule, and type d capsule, respectively, indicating that these loci are solely responsible for determining capsule type in K. kingae. Further analysis demonstrated that 96% of the invasive isolates express either the type a or type b capsule and that a disproportionate percentage of carrier isolates express the type c or type d capsule. These results establish that there are at least four structurally distinct K. kingae capsule types and suggest that capsule type plays an important role in promoting K. kingae invasive disease.


Introduction
Kingella kingae is being recognized increasingly as an important cause of bone and joint infections in young children, reflecting more sensitive cultivation techniques and increased availability of molecular-based diagnostic tools [1,2]. Among the key surface factors expressed by K. kingae is a polysaccharide capsule [3,4]. Capsules are recognized as important virulence factors in many gram-positive and gram-negative bacteria and have a variety of functions, including inhibiting complement deposition, reducing phagocytosis, and preventing desiccation [5][6][7]. Polysaccharide capsules conjugated to an immunogenic carrier protein also serve as effective vaccine antigens and have dramatically reduced morbidity and mortality caused by bacteria such as Streptococcus pneumoniae [8], Haemophilus influenzae type b [9], and Neisseria meningitidis [10].
In previous work, we described the structure of the capsule expressed by K. kingae strain 296-492 as a polymer of [3)-β-GalpNAc-(1!5)-β-Kdop-(2!] and identified the genes essential for capsule synthesis, assembly, and export [3,11]. In the course of this work, we established that the CsaA glycosyltransferase contains both a GalNAc-transferase domain and a Kdo-transferase domain and is sufficient for creating both the β-GalpNAc-(1!5)-β-Kdop linkage and the β-Kdop-(2!3)-β-GalpNAc linkage. In addition, the CsaA glycosyltransferase may catalyze addition of β-GalpNAc to the terminal β-Kdo residue of the poly-β-Kdo linker [3]. Bendaoud et al. recently reported that the structure of the capsule isolated from the surface of another K. kingae strain is a polymer of [6)-α-GlcpNAc-(1!5)-β-Kdop-(2!] [12], suggesting the existence of at least two different K. kingae capsule types. The presence of multiple capsule types is well documented in a variety of bacterial species, with examples including S. pneumoniae, N. meningitidis, H. influenzae, and Klebsiella pneumoniae. In some cases, specific capsule types are associated more commonly with carriage or more commonly with invasive disease. For example, there are at least 90 different S. pneumoniae capsule types, but 23 types account for more than 90% of invasive pneumococcal disease worldwide [13,14]. Similarly, in N. meningitidis 6 of the 13 characterized capsule types are responsible for 90% of invasive disease cases globally [15,16]. In K. pneumoniae, capsule types 1, 3, and 4 are associated with respiratory tract infection, and capsule types 9 and 10 are associated with urinary tract infection [17].
In this study, we set out to define the genetic and structural basis of capsule diversity in a large collection of K. kingae clinical isolates from Israel. In addition, we examined the relationship between specific capsule types and clinical presentations.

Results
Four capsule synthesis loci are present in a diverse collection of K. kingae clinical isolates In initial experiments, we screened a collection of 417 Israeli invasive and carrier isolates for the presence of csaA, the capsule synthesis gene in our prototype K. kingae strain KK01. Using csaA-specific primers and PCR, we found that only 47 percent of all isolates contained the csaA gene. We hypothesized that other capsule types exist and that the region containing csaA represents the K. kingae capsule synthesis locus and differs in genetic content depending on the enzymatic machinery required to synthesize a specific capsule polysaccharide structure. To test this hypothesis, we designed a forward primer annealing to arg, the gene upstream of csaA in strain KK01, and a reverse primer annealing to hemB, the gene downstream of csaA in strain KK01, to amplify across the suspected capsule synthesis locus. As shown in Fig 1A, amplification across this locus in a group of representative isolates in the collection yielded four different amplicon sizes. Restriction mapping of these amplicons with NruI revealed similar banding patterns for strains with the same amplicon size (Fig 1B). Nucleotide sequencing of the amplicons from multiple strains with the same amplicon size revealed an absolute correlation between the amplicon size and the gene content, indicating that the four discrete amplicons represent four discrete loci (Fig 1C). After determining the predicted open reading frames (ORFs) in each amplicon, we searched for the presence of predicted domains or motifs using BLASTP and PHYRE2. The~3500 bp amplicon contained only the csaA gene (identical to our prototype strain KK01) and was named the csa locus. The~4000 bp amplicon contained a gene encoding a predicted GT-B type glycosyltransferase with homology to a GlcNAc transferase (designated csbA), a gene encoding a putative capsule synthesis enzyme with homology to a Kdo transferase (designated csbB), and a gene encoding a putative enzyme with homology to an acetyltransferase (designated csbC) and was named the csb locus. The~5000 bp amplicon contained two genes encoding putative enzymes with homology to halo-acid dehydrogenases (designated cscA and cscB) and a gene encoding a predicted glycosyltransferase (designated cscC) and was named the csc locus. The~5500 bp amplicon contained a gene encoding a predicted galactosyltransferase (csdA), a gene encoding a predicted GlcNAc transferase (csdB), and a gene encoding a predicted GT-A type glycosyltransferase (csdC) and was named the csd locus. Based on the sequence of the four unique loci, specific internal primer pairs were generated, producing locus-specific amplicons, as shown in Fig 1D-1G.
The four capsule synthesis loci are associated with four different polysaccharide capsule types To confirm that each of the four capsule synthesis loci is associated with a specific capsule type, we examined the glycosyl composition of purified polysaccharide capsule from representative strains that contain either the csa, csb, csc, or csd locus. In order to eliminate contamination with the galactan exopolysaccharide produced by K. kingae, we first deleted the pam locus from these strains [11,18]. As summarized in Table 1, strains KK01, PYKK98, and PYKK93 harbor the csa locus and produce a capsule containing GalNAc and Kdo, which we named capsule type a. Strains PYKK89, PYKK121, PYKK58, and PYKK59 harbor the csb locus and produce a capsule that contains GlcNAc and Kdo, which we named capsule type b. Strains PYKK60 and D7674 harbor the csc locus and produce a capsule that contains ribose and Kdo, which we named capsule type c. Finally, strains E3339, D7453, and BB270 contain the csd locus and produce a capsule that contains galactose and GlcNAc, which we named capsule type d. Considered together, these findings demonstrate complete agreement between the capsule synthesis locus and capsule glycosyl composition, indicating that genetic screening of the capsule synthesis locus is predictive of capsule type.

Structural analysis reveals distinct polysaccharide capsule structures
In previous work, we reported that the type a polysaccharide capsule is a polymer of [3)-β-GalpNAc-(1!5)-β-Kdop-(2!] [11]. To determine the chemical structure of the K. kingae type b, type c, and type d capsules, surface polysaccharide was purified from derivatives of Each locus shows the capsule synthesis genes unique to each capsule type (white) and the highly homologous flanking regions shared among all strains (gray). Solid arrows (size not to scale) denote the approximate location of the screening primers that anneal to the homologous flanking regions used to amplify across the flanking genes irrespective of internal sequence (as shown in Panel A). The dashed arrows above each locus denote the approximate location of the locus-specific screening primers that generate amplicons for (D) the csa locus (~2400 bp), (E) the csb locus (~2200 bp), (F) the csc locus (~2750 bp), and (G) the csd locus (~4100 bp strains PYKK58 (type b), PYKK60 (type c), and BB270 (type d) lacking the pam locus and was analyzed with a combination of linkage analysis and 1-D and 2-D NMR spectroscopy.
Linkage analysis of the type b capsule gave 1,5,6-tri-O-acetyl-2-deoxy-2-methylacetamido-3,4-di-O-methyl-1-2 H-glucitol, derived from 6-linked GlcpNAc, and 1,2,5,6-tetra-O-acetyl-3-deoxy-4,7,8-tri-O-methyl-1,1,2-tri-2 H-octitol, derived from 5-linked Kdo (S1A and S2A Figs). Absolute configuration analysis gave D-GlcNAc. Characteristic peaks in the 1-D proton spectrum (Fig 2A) included one major anomeric signal at 5.08 ppm, two signals corresponding to the H-3 protons of Kdo, one N-acetyl peak from GlcNAc, and one O-acetyl of unknown origin. Tracing the connectivities of GlcNAc from H-1 and of Kdo from H-3 in the COSY and TOCSY spectra together with the carbon chemical shifts derived from the HSQC spectrum led to the complete assignment of the chemical shifts belonging to each residue (Table 2 and Fig  3A). Due to the high molecular weight of the sample, the peaks in the spectra were broad and not suitable to measure proton-proton coupling constants for the determination of the anomeric configurations of GlcNAc and Kdo. However, the proton and carbon chemical shifts of the GlcNAc residue agreed with the α-configuration. Comparison of the chemical shifts of the Kdo residue with literature values [19] showed that Kdo was in the β-configuration. The downfield displacement of carbon chemical shifts GlcNAc-C6 and Kdo-C5 indicated the linkage positions as 6-linked GlcNAc and 5-linked Kdo. The downfield displacement of the proton chemical shifts of Kdo-H8 together with the intensity (3H) and chemical shifts of the O-acetyl signal (2.13/23.1 ppm) indicated acetylation on O-8 of Kdo. Taken together, these results indicated that the polymer is composed of a disaccharide repeating unit with the structure [6)-α-D- Linkage analysis of the type c capsule gave 1,3,4-tri-O-acetyl-2,5-di-O-methyl-1-2 H-ribitol, derived from 3-linked ribofuranose, 1,2,4-tri-O-acetyl-3,5-di-O-methyl-1-2 H-ribitol, derived from 2-linked ribofuranose, and 1,2,4,6-tetra-O-acetyl-3-deoxy-5,7,8-tri-O-methyl-1,1,2-tri-2 H-octitol, derived from 4-linked Kdo (S1B and S2B Figs). Absolute configuration analysis gave D-ribose. Characteristic peaks in the 1-D proton NMR spectrum ( Fig 2B) included three anomeric signals at 5.34, 5.29, and 5.18 ppm, two pairs of signals corresponding to the H-3 protons of Kdo, an acetyl methyl signal, and several resonances in the carbohydrate ring region. The presence of two sets of Kdo signals of unequal intensity (ratio 2:3) together with the presence of an acetyl signal with an area three times that of the larger Kdo-H3 peak suggested that 60% of the Kdo residues in the polysaccharide were O-acetylated. To reduce the heterogeneity of the sample, we performed de-O-acetylation. The 1-D proton NMR spectrum of the de-O-acetylated material (Fig 2C) was simplified compared to the native polysaccharide and displayed only a single set of Kdo H-3 peaks. Tracing the connectivities of the three anomeric signals from H-1 and of Kdo from H-3 in the COSY and TOCSY spectra together with the carbon chemical shifts derived from the HSQC spectrum led to a complete chemical shift assignment and revealed the presence of two 2-linked and one 3-linked ribofuranose residues as well as one 4-linked Kdo residue (Table 3 and Fig 3B). The proton and carbon chemical shifts of the ribose residues agreed with β-anomeric configuration [20,21]. Comparison of the chemical shifts of the Kdo residue with literature values [19] showed that Kdo also had the β-  Fig 2D) included two α-anomeric signals (ratio 3.4:1), a cluster of several β-anomeric signals, and a group of N-acetyl peaks. The β-anomeric cluster was resolved into four distinct resonances in the HSQC spectrum. Tracing the connectivities from these anomeric signals in the COSY and TOCSY spectra together with the carbon chemical shifts obtained from the HSQC spectrum allowed a complete assignment of a total of six different residues that were grouped into three pairs of residue types (Table 4 and Fig 3C2). The chemical shifts of two of the six residues identified them as 3-linked α-GlcNAc (A and A'), another two of the six as 3,4-linked β-GlcNAc (B and B'), and the final two of the six as terminal β-Gal (C and C'). This information suggested the presence of two similar trisaccharide repeating units in the polysaccharide. Inter-residue linkages were assigned from HMBC and NOESY correlations. The HMBC ( Fig 3C2) and NOE correlations (S4 Fig) between H1 of B/B' and H3 of A/A' and between H1 of C/C' and H4 of B/B' confirmed the presence of two slightly different 3)-[β-Gal-(1!4)]-β-GlcNAc-(1!3)-α-GlcNAc-(1! trisaccharides, but there were no HMBC or NOE correlations to H1 of A or H3 of B that would link these trisaccharides together. The unusual downfield shifts of A-H1, B-H3, and B'-H3 suggested substitution by an electronegative group, such as acetate, sulfate, or phosphate. The fact that no significant amount of  3C1). Taken together, these results strongly indicated that the polymer is composed of a trisaccharide repeating unit with the structure [P-   The csa, csb, csc, and csd capsule synthesis loci are necessary and sufficient for polysaccharide capsule synthesis To confirm that the csa, csb, csc, and csd loci are essential for production of capsule, we deleted each of these loci and then examined the resulting strains for surface material that stains with Alcian blue. As shown in Fig 5, targeted deletion of the csa, csb, csc, or csd locus resulted in loss of surface extractable capsule from strains KK01, PYKK58, PYKK60, and BB270, respectively. Chromosomal complementation of each of these regions at the native locus resulted in restoration of encapsulation. These results demonstrate that the capsule synthesis loci are essential for production of capsule in representative type a, type b, type c, and type d K. kingae strains.
In additional experiments, we examined the ability of the type a, b, c, and d loci to complement a deletion of the csa locus in prototype strain KK01 and produce the corresponding capsule. In performing these studies, we engineered a deletion of csaA with no effect on the flanking arg and hemB genes, producing a strain called KK01Δcsa. Subsequently, we generated a plasmid called pSwap, which contains the arg and hemB genes, a kanamycin resistance marker, and a partial pUC19 multiple cloning site (MCS) (Fig 6A). Using this plasmid, we inserted each of the four capsule synthesis loci into the MCS, generating pSwapcsa, pSwapcsb, pSwapcsc, and pSwapcsd. Each of these plasmids was linearized and transformed into strain KK01Δcsa, producing strains KK01Swapcsa, KK01Swapcsb, KK01Swapcsc, and KK01Swapcsd. As shown in Fig 6B, each of these strains produced a capsule as assessed by Alcian blue staining of surface extracts. To confirm that the capsule in each of these strains corresponded to the specific capsule synthesis locus, surface polysaccharide was extracted and examined initially by Alcian blue staining. As expected, the Alcian blue staining profile of the capsule extracted from the csa, csb, csc, or csd swap strains was similar to the profile of the parental capsule locus source strain, suggesting that the capsules produced in a KK01Δcsa background strain retain their native migration pattern (Fig 6B). 1-D Proton NMR analysis demonstrated that strain KK01Swapcsa produced the type a capsule, strain KK01Swapcsb produced the type b capsule, strain KK01Swapcsc produced the type c capsule, and strain KK01Swapcsd produced the type d capsule (Table 5). These results demonstrate that the csa, csb, csc, and csd loci encode the  The type a and type b capsules are enriched in invasive isolates of K. kingae With our knowledge of the type a, type b, type c, and type d capsule loci in hand, we used a PCR approach to examine a large collection of K. kingae clinical isolates for capsule type. A total of 417 Israeli strains isolated between 1990 and 2014 were investigated. The collection contains 239 strains isolated from healthy pharyngeal carriers and 178 strains recovered from patients with a variety of invasive infections, including skeletal system infections, bacteremia, and endocarditis, allowing characterization of the capsule types elaborated by the full range of K. kingae isolates in the country. Overall, 413 of the 417 (99.0%) strains were genotyped by pulsed-field gel electrophoresis (PFGE) [22] and found to belong to 60 distinct clones, including 16 clones that were represented in the collection by !7 strains and that collectively accounted for 345 (83.5%) of all typed strains. One of the four capsule synthesis loci was identified in all of the strains, except for strain KK183 belonging to PFGE clonal group Tnc, which was isolated from the synovial fluid of a child with septic arthritis. This strain was shown to be nonencapsulated and did not generate capsule locus flanking or capsule locus specific PCR amplicons (S7 Fig). Therefore, this strain was not included in the analysis of the association between capsule type and invasiveness or clonal distribution. A second nonencapsulated strain (KK56, PFGE clone S) was isolated from a child with arthritis and was found to have a csaA gene with a 512 bp internal deletion in the ORF, which is predicted to introduce a frameshift mutation leading to a truncated CsaA protein (S7 Fig). This strain was included among organisms with capsule type a for the purposes of the data analysis. While capsule type a was common among both carrier and invasive isolates, the distribution of capsule types b, c, and d in the invasive versus carrier groups showed statistically significant    (Fig 7B and S2 Table). Statistical analysis using the χ 2 test demonstrated a significant association of capsule type and PFGE clones (P<0.001). Of note, all 16 common PFGE clones showed a clear predominance of a single capsule type: capsule type a in PFGE clones A, B, C, H, J, M, P, and S; capsule type b in PFGE clones K, N, and V; capsule type c in PFGE clones D and R; and capsule type d in PFGE clones F, G, and U (Fig 7C and S3 Table).

Discussion
In this study we examined a large collection of K. kingae clinical isolates and established that there are four different K. kingae capsule types. In addition, we identified the underlying capsule synthesis genes for each capsule type. Using a combination of mass spectroscopy and NMR, we also determined the structure of two previously uncharacterized capsule types (type c and type d), complementing previous work on the structure of the type a capsule and the type b capsule [11,18]. Finally, we used a genetic screen to determine the capsule type of invasive disease isolates and healthy carrier isolates and discovered that capsule type a and type b account for 96% of all invasive disease isolates and that capsule type c and type d are disproportionately present among healthy carrier isolates.
At the outset of our study, we hypothesized that our large collection of K. kingae clinical isolates would contain multiple polysaccharide capsule types. Bacterial polysaccharide capsules are traditionally typed using one of two methods: 1) genetically, based on the presence of specific capsule synthesis genes in the capsule locus, or 2) immunologically, based on agglutination reactions using capsule-specific sera. In this study we used a PCR-based genetic screening method, similar to methods that assess the capsule polysaccharide synthesis region for capsule typing of K. pneumoniae [23][24][25], Pasteurella multocida [26], and N. meningitidis [27]. Using this approach, we established that there are four different capsule types in K. kingae, with each strain containing only one of four distinct capsule synthesis loci. 1-D Proton NMR analysis of purified capsule from isogenic capsule synthesis locus swap strains confirmed that capsule type is determined by the gene content of the capsule synthesis locus ( Table 5).
It is interesting to speculate regarding the potential for interstrain capsule type switching in K. kingae. In N. meningitidis, capsule switching has been shown to result from recombination of the polysialyltransferase gene (siaD) or the capsule biosynthesis operon [33], with evidence for capsule switching between strains implicated in carriage and strains associated with invasive disease [34]. Pneumococcal isolates can also undergo capsule switching, with the serotype of a clone changing due to alteration in the capsule biosynthesis locus via mutations or through genetic recombination [35][36][37]. In support of the possibility of intraspecies capsule switching in K. kingae, several K. kingae clonal groups are associated with multiple capsule types.
Asymptomatic carriage provides an ideal environment for interspecies exchange of genetic material among bacteria that occupy the same niche [38]. The human nasopharynx has been shown to harbor diverse bacteria, including N. meningitidis, H. influenzae, and S. pneumoniae as well as nonpathogenic Neisseria spp. and Moraxella spp. Evidence supporting horizontal gene transfer between phylogenetically distant species is seen in the meningococcal genome, which harbors three independent domains of Haemophilus-like DNA. Uptake and integration of DNA in the upper respiratory tract is a probable mechanism to explain the capsule diversity observed in K. kingae in this study. Actinobacillus spp., Moraxella spp., and Kingella spp. are all found in normal human flora of the upper respiratory tract, providing the opportunity for horizontal gene transfer from other genera as the genesis of the four K. kingae capsule synthesis loci. M. nonliquefaciens has been shown to be present in the respiratory tract of young children [39,40]. A. pleuropneumoniae is primarily a swine pathogen, but other Actinobacillus spp. can be found in humans [41].
All of the K. kingae strains in our collection gave a PCR product for the capsule export and assembly genes ctrABCD, lipA, and lipB, suggesting that all of these strains contain the machinery necessary to display a capsule polymer on their surface [3]. However, out of 417 isolates, two strains demonstrated atypical PCR capsule typing results. First, strain PYKK56 yielded a csa locus PCR product, but the product was smaller than expected (S7C Fig). Sequencing of the csa locus in this strain revealed a 512 bp internal deletion in the csaA gene, resulting in frameshift that is predicted to lead to a truncated CsaA protein (S7G Fig). Alcian blue staining of surface extracts from this strain revealed no capsule, suggesting that the csaA mutation resulted in abrogation of capsule expression (S7A Fig). Second, strain PYKK183 yielded no capsule locus flanking product and no capsule locus-specific PCR product (S7B- S7G Fig). Alcian blue staining of surface extracted material revealed that this strain is not encapsulated, suggesting that this strain lacks capsule synthesis genes, rather than possessing a unique capsule synthesis locus (S7A Fig). Amit et al. determined that K. kingae PFGE clonal groups B, H, K, N, and P account for 72.9% of all invasive isolates and that PFGE clonal groups A, C, D, F, G, J, R, S, and U are rare among invasive disease isolates [42]. Interestingly, only capsule types a and b are represented in the B, H, K, N, and P clonal group isolates. Overall, the type a and type b capsules account for 96% of invasive isolates but less than 70% of the carrier isolates. The type c capsule is most prevalent in the D and R clonal groups, and the type d capsule is most prevalent in the F, U, and G clonal groups, all of which fall into the subset of rare-disease PFGE clonal groups. The significant difference in the capsule type distribution between strains carried by asymptomatic children and those isolated from patients with invasive infections may suggest that the type c and type d capsules provide incomplete protection to K. kingae organisms, enabling them to colonize the oropharyngeal epithelium but not allowing their survival in the bloodstream, the skeletal system, or the endocardium. This phenomenon of specific capsule type association with invasive disease is reminiscent of encapsulated H. influenzae, an upper respiratory tract colonizer that elaborates six distinct polysaccharide capsules, with strains elaborating the type b capsule accounting for almost all cases of disease prior to the introduction of the conjugate vaccines [43]. However, the association between capsule type and virulence may not be causal. Close examination of the data in Fig 7C and [42], elaborate polysaccharide capsule types a or b, indicating that determinants other than capsule type likely play an important role in the potential of the organism to cause invasive disease. This possibility is indirectly supported by the fact that two nonencapsulated strains, KK183 and KK56, were able to cause septic arthritis in otherwise healthy children.
Considering the effectiveness of many polysaccharide-conjugate vaccines in reducing childhood morbidity and mortality, it is interesting to speculate that a K. kingae capsule polysaccharide-conjugate vaccine may be an effective strategy to prevent K. kingae disease, pending additional analysis of the global burden of K. kingae disease. While more studies are needed, the discovery that the capsule repertoire of a diverse collection of K. kingae carrier and invasive disease isolates is represented by only 4 capsule types, with two capsule types accounting for >95% of invasive disease, is an important first step in establishing the feasibility of a vaccine for the prevention of K. kingae disease.

Bacterial strains and growth conditions
The strains representative of each capsule type that were used for the fundamental studies in the work are listed in Table 6. The complete list of clinical isolates that were examined for capsule type are shown in S4 Table. K. kingae strain 269-492 was isolated from the joint fluid of a child with septic arthritis at St. Louis Children's Hospital, St. Louis, MO. K. kingae strain KK01 is a stable natural variant of strain 269-492 that grows as a non-spreading, non-corroding colony type and was used as the primary strain in this study [44]. K. kingae and E. coli strains were grown and stored as previously described [4,11].
Clinical isolate strain collection K. kingae isolates were selected from a large assortment of Israeli strains that have been gathered at the Soroka University Medical Center since the early 1990's. The collection contains isolates from patients with a variety of invasive infections and from healthy pharyngeal carriers in the course of epidemiological studies on K. kingae carriage and transmission.
A total of >200 K. kingae invasive strains and >600 K. kingae carrier isolates have been typed by pulsed field gel electrophoresis (PFGE) [42], and a sample of the predominant PFGE clones has been further characterized by MLST and rtxA gene sequencing [46]. Based on genotyping results, strains were selected to meet the following study goals while maintaining a manageable number for analysis: strains isolated over more than two decades, clones that collectively cause the vast majority of invasive infections in Israel [47], clones that are primarily associated with asymptomatic pharyngeal colonization [48], strains isolated from patients with a variety of clinical syndromes (bacteremia, skeletal system infection, or endocarditis) [42], and strains associated with clusters of disease in daycare center facilities [49]. Because the different genotyping schemes of K. kingae exhibit remarkable congruency [46], it was assumed that studying strains belonging to rare PFGE clones would increase the chances to detect novel capsule types. Thus, the strain collection was enriched with a large number of uncommon invasive as well as colonizing isolates.

Ethics Statement
The Israeli isolates used in this study are part of a preexisting anonymized collection and as such did not require IRB approval for use.

Molecular methods and strain manipulation
Targeted gene disruptions and complementation constructs in K. kingae were generated as previously described [4,45]. Briefly, plasmid-based gene disruption constructs were created in E. coli, linearized, and introduced into K. kingae using natural transformation. Transformants were recovered by selectively plating on chocolate agar plates with the appropriate antibiotic. Gene disruptions and complementation constructs were confirmed by PCR.
The primers used in this study are listed in Table 7. To delete the capsule synthesis locus, we generated the plasmid pSwapEmpty. Briefly, fragments of homologous recombination targeting sequence corresponding to~1 kb upstream of csaA and~1 kb downstream of csaA were PCR amplified from strain KK01 genomic DNA using primers pSwapFor5'/pSwapRev5' and pSwapFor3'/pSwapRev3' , respectively, and were ligated into pUC19. A kanamycin resistance cassette was then ligated into the pUC19 KpnI site, which is located between the cloned upstream and downstream homologous recombination targeting sequences, to generate pSwa-pEmpty. The plasmid was linearized with NdeI and transformed into strain KK01.
To create the complementation/capsule swap constructs, the capsule synthesis loci were PCR amplified as follows: for the csa locus, using genomic DNA from strain KK01 and primers csaswapFor/csaswapRev; for the csb locus, using genomic DNA from strain PYKK58 and primers csbswapFor/csbswapRev; for the csc locus, using genomic DNA from strain PYKK060 and primers cscswapFor/cscswapRev; and for the csd locus, using genomic DNA from strain BB270 and primers csdswapFor/csdswapRev. The csa and csb locus amplicons were cloned into pSwa-pEmpty using standard restriction cloning, generating pSwapcsa and pSwapcsb, respectively. The csc and csd amplicons were cloned into pSwapEmpty using the Gibson Assembly Cloning kit (New England Biolabs, Ipswich, MA), generating pSwapcsc and pSwapcsd, respectively.
For the capsule swap studies, we transformed each swap construct (pSwapcsa, pSwapcsb, pSwapcsc, or pSwapcsd) harboring a kanamycin cassette into the nonencapsulated isogenic strain KK01Δcsa (Erm R ) and screened for loss of Erm R and gain of Kan R . For capsule synthesis locus complementation, the capsule swap plasmids pSwapcsa, pSwapcsb, pSwapcsc, and pSwapcsd were transformed into KK01Δcsa, PYKK58Δcsb, PYKK60Δcsc, and BB270Δcsd, respectively, using the unmarked transformation protocol described below.
To generate unmarked gene disruptions and complements, we used the following procedure without antibiotic selection. First, K. kingae was grown overnight on chocolate agar, resuspended in Brain Heart Infusion (BHI) broth containing 50 mM MgCl 2 to an OD 600 of 0.7, and diluted 1:25 in BHI broth. The initial dilution was then serially diluted 1:4 a total of 9 times, and 5 μl of each dilution was transferred to a microfuge tube containing 5 μl of linearized transforming plasmid at a concentration of 50 ng/μl. The 10 μl total mixture was then plated on chocolate plates and allowed to dry in ambient air conditions (approximately 5 minutes) before placement into the CO 2 incubator at 37°C. Single colonies were screened by PCR for recombination at the locus of interest after two rounds of single colony purification of the potential transformants.

Polysaccharide capsule extraction, purification, and analysis
In preparation for extraction and purification of capsule, the pam locus involved in synthesis of the galactan exopolysaccharide was deleted from the relevant strains [11]. Extraction, purification, and visualization of migration patterns on 7.5% SDS-PAGE gels using Alcian blue staining of capsule material were performed as previously described [4,11].
De-O-acetylation. The sample solutions (~1.5 mg/0.5 ml) were adjusted to pH 11.0 by addition of 2 M ammonium hydroxide solution and incubated overnight at room temperature. The resulting solutions were dialyzed through a 3.5 kDa regenerated cellulose membrane to remove salts. The retentates were then deuterium-exchanged for NMR analysis.
NMR spectroscopy. The retentates after xylanase digestion (total amount) were deuterium exchanged 2 times by lyophilization in D 2 O. The dry residues were re-dissolved in 300 μl D 2 O (99.96%, Cambridge Isotopes) and placed in 3-mm OD NMR tubes. 1-D Proton and 2-D COSY, TOCSY, HSQC, HMBC and NOESY spectra were acquired at 55°C on an Agilent Inova 600 MHz spectrometer equipped with a cryoprobe using standard Agilent pulse sequences. Chemical shifts were measured relative to internal acetone peak (δ H /δ C = 2.218/33.0 ppm).
Approximately 1 mg quantities of the samples were used for linkage analysis. The samples were suspended in 200 μl of dimethyl sulfoxide and left to stir for 1 d. Permethylation was effected by two rounds of treatment with sodium hydroxide (15 min) and methyl iodide (45 min). Following sample workup, the permethylated material was hydrolyzed using 2 M TFA (2 h in sealed tube at 121°C), reduced with NaBD 4 , and acetylated using acetic anhydride/TFA. The resulting PMAAs were analyzed on an Agilent 7890A GC interfaced to a 5975C MSD (mass selective detector, electron impact ionization mode). Separation was performed on a 30 m Supelco SP-2331 bonded phase fused silica capillary column.
Linkage analysis of the Kdo residues. To prepare the partially methylated alditol acetate (PMAA) derivatives of Kdo, the permethylated polysaccharide samples were subjected to sequential steps: reduction of the Kdo carboxymethyl groups with lithium triethylborodeuteride (Superdeuteride, Aldrich, St. Louis, MO) in THF (200 μl, 2 h at room temperature); mild hydrolysis (0.1 M trifluoroacetic acid, 100°C, 30 min) to cleave the Kdo-ketosidic linkages; reduction of Kdo residues at C-2 carbonyl group (using NaBD 4 in water/ethanol 1:1 v/v); normal hydrolysis (2 M trifluoroacetic acid, 100°C, 30 min) to cleave the sugar linkages; reduction of the newly formed aldehydo sugars (using NaBD 4 in 50 mM NH 4 OH); and acetylation of the resulting partially methylated alditols to yield the PMAA derivatives. Acetylation was performed in acetonitrile-pyridine-acetic anhydride containing 4-N,N 0 -dimethylaminopyridine as a catalyst for 4 h at room temperature as described [51].
D and L determination. The samples were freeze-dried and hydrolyzed in 2M TFA (500 μl) at 120°C for 1.5 h. The hydrolysates were dissolved in 200 μl S-(+)-2-butanol (Fluka, St. Louis, MO), 15 μl acetyl chloride (Aldrich) was added, and nitrogen gas was bubbled through the solutions for 30 seconds. The mixtures were capped tightly and incubated at 80°C for 16 h. After incubation, the mixtures were dried under a stream of dry nitrogen and then redried with absolute methanol. A 250 μl volume of TMS reagent (Tri-Sil, Thermo Scientific Pierce, Waltham, MA) was added to the dry sample, and derivatization was carried out at 80°C for 20 min.
The same procedure was applied to authentic monosaccharide standards. Two sets of standards were derivatized, and either S-(+)-2-butanol or R-(-)-2-butanol was added separately to these standards. The derivatized samples and standards were analyzed on Agilent 5975C GC interfaced with 7890A MS detector.
Size exclusion chromatography (SEC). The samples were dissolved in deionized water (2 ml), and 20 μl volumes of the resulting solutions were injected for separation on a Superdex-75 SEC column using an Agilent 1200 HPLC system. The eluent was 50 mM ammonium acetate, pH 5, the flow rate was 1 ml/min, and signal was monitored with an ELS detector. Peaks were integrated using Chemstation software.
Nanospray-mass spectrometry(NSI-MS). NSI-MS analyses in full mass and MS n mode were performed using an Orbitrap Fusion mass spectrometer (Thermo Fisher, Waltham, MA) equipped with a nanospray ion source. Intact capsule from type d strain BB270 was dissolved in 1 mM NaOH in 50% methanol and then infused directly into the instrument at a constant flow rate of 1 μl/ min. A full FTMS spectrum was collected at 30,000 resolution with 20 microscans. The capillary temperature was set at 210°C, and MS analysis was performed in the negative ion mode.
Genetic screen of capsule synthesis loci K. kingae sequence outside of the capsule synthesis locus from strain 269-492 was used to design flanking primers hemBFor and argRev in the hemB (delta aminolevulinic-acid dehydratase) and arg (arginine-succinate synthase) genes flanking csaA (Table 7). PCR amplicons were sequenced, and the resulting sequence was the basis for design of interior primers specific for each of the four capsule synthesis loci. To screen for the presence of each locus, we used both universal flanking primers that amplified all capsule loci and locus-specific primers that annealed to the interior portion of each locus. The presence of either csaA, csbABC, cscABC, or csdABC (the csa, csb, csc, or csd capsule synthesis locus) was determined by PCR amplification using interior primers (see Table 7) and confirmed by determining the size and restriction map of the flanking primer amplicon.

Restriction analysis
To obtain restriction digest patterns, PCR products were amplified using the primers hemBFor and argRev listed in Table 7 in a total reaction volume of 25 μl. 2.5 μl of 10x digestion buffer and 1 μl of NruI enzyme (New England Biolabs) were mixed with PCR products and incubated overnight at 37°C. The digests were resolved on a 1.2% agarose gel and visualized for banding pattern.

Statistical analysis
The statistical significance of the differences in the distribution of the different capsule types among carrier vs. invasive strains, among bacteremia vs. skeletal system infections, and among the PFGE clones was determined by the χ 2 test using the Statistical Package for the Social Sciences (SPSS) version 21 software. The link between the different capsule types and invasiveness was further explored with a logistic regression model in which the capsule type with the lowest percentage of associated invasive strains was employed as a reference and the odds ratio (OR) for invasiveness, p-value, and 95% confidence intervals (CI) for the other capsule types were calculated. A p-value <0.05 was considered significant for all comparisons. , and PYKK183 (lane 6) were subjected to Alcian blue staining of surface extracts (A) and PCR of the capsule synthesis locus using the flanking primers (B), csa-specific primers (C), csb-specific primers (D), csc-specific primers (E), and csd-specific primers (F). PYKK56 and PYKK183 lack Alcian blue-stainable material in surface extracts. For PYKK56, the capsule synthesis locus flanking PCR amplicon is smaller than all of the four control amplicons, and the csa-specific PCR amplicon is smaller than the control csaspecific PCR amplicon. Sequencing of the capsule synthesis locus flanking PCR amplicon from PYKK56 (Panel B, Lane 5) revealed a csaA gene with a 512 bp deletion in the open reading frame (G), which also introduces a frameshift mutation leading to a predicted truncated CsaA protein of 316 amino acids versus the wild type 811 amino acid protein. The capsule locus flanking and capsule locus specific PCRs failed to generate amplicons for PYKK183. (TIF) S1