Comparative Analyses of the Lipooligosaccharides from Nontypeable Haemophilus influenzae and Haemophilus haemolyticus Show Differences in Sialic Acid and Phosphorylcholine Modifications

Abstract

Haemophilus haemolyticus and nontypeable Haemophilus influenzae (NTHi) are closely related upper airway commensal bacteria that are difficult to distinguish phenotypically. NTHi causes upper and lower airway tract infections in individuals with compromised airways, while H. haemolyticus rarely causes such infections. The lipooligosaccharide (LOS) is an outer membrane component of both species and plays a role in NTHi pathogenesis. In this study, comparative analyses of the LOS structures and corresponding biosynthesis genes were performed. Mass spectrometric and immunochemical analyses showed that NTHi LOS contained terminal sialic acid more frequently and to a higher extent than H. haemolyticus LOS did. Genomic analyses of 10 strains demonstrated that H. haemolyticus lacked the sialyltransferase genes lic3A and lic3B (9/10) and siaA (10/10), but all strains contained the sialic acid uptake genes siaP and siaT (10/10). However, isothermal titration calorimetry analyses of SiaP from two H. haemolyticus strains showed a 3.4- to 7.3-fold lower affinity for sialic acid compared to that of NTHi SiaP. Additionally, mass spectrometric and immunochemical analyses showed that the LOS from H. haemolyticus contained phosphorylcholine (ChoP) less frequently than the LOS from NTHi strains. These differences observed in the levels of sialic acid and ChoP incorporation in the LOS structures from H. haemolyticus and NTHi may explain some of the differences in their propensities to cause disease.

INTRODUCTION

Nontypeable Haemophilus influenzae (NTHi) is a common commensal organism of the nasopharynx. NTHi can also cause disease in individuals with compromised upper airways (1–3). In adults, this organism is typically associated with causing pneumonia and bronchitis in individuals with chronic obstructive pulmonary disease (COPD) (4, 5), and in children, it is a causative agent of otitis media (6, 7). NTHi bacteria typically cause localized infections such as sinusitis or bronchitis, but they do have the ability to cause systemic infections such as bacteremia or meningitis (6). Haemophilus haemolyticus is also a commensal organism of the nasopharynx; it is generally considered a colonizer and is typically not associated with disease. The application of ribosomal analysis and specific gene differences has enabled resolution of these closely related organisms (8–12).

Lipooligosaccharides (LOS) are the major surface glycolipids expressed by both NTHi and H. haemolyticus. LOS has been shown to be a major factor in the pathogenesis of NTHi (13–17). LOS structures that terminate in N-acetyl-5-neuraminic acid (Neu5Ac), commonly referred to as sialic acid, have been shown to be important in resistance to complement killing (13, 15, 17) and in the ability of the organism to cause otitis media in the chinchilla model of infection (15, 16). Additionally, Weiser et al. have shown that phosphorylcholine (ChoP) can be expressed on NTHi LOS in a phase variable manner (18). The interaction between the platelet-activating factor (PAF) receptor and ChoP has been shown to be critical in airway cell invasion and human nasopharyngeal colonization by NTHi (14, 19). ChoP has also been found to be important in NTHi's interaction with the complement system (20). Studies by McCrea and coworkers showed that some strains of H. haemolyticus lacked a number of genes important in NTHi pathogenesis (21, 22). These genes include ones found in the lic1 operon, which is involved in the addition of ChoP to LOS. A complete lic1 operon was present in only 43% of the H. haemolyticus strains studied compared to 92% of the NTHi strains analyzed (22). In addition, genes involved in the development of LOS core structures (lic2A and lgtC) are found less frequently in H. haemolyticus strains compared to NTHi (21).

Structural analyses of the LOS from any H. haemolyticus strain have not been previously reported. The first goal of this study was to analyze the LOS structures from various NTHi and H. haemolyticus strains. The second goal was to perform a genetic comparison of the genes involved in the incorporation of sialic acid into the LOS of NTHi and H. haemolyticus. Last, we employed mass spectrometry and immunochemical techniques to examine the incorporation of ChoP and sialic acid into the LOS from H. haemolyticus and NTHi.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Forty-three NTHi strains and 14 H. haemolyticus strains used in this study are listed, and their source described, in Table 1. Twenty strains of NTHi were isolated from sputum samples from adults with COPD in an ongoing prospective observational study (5), 10 from children with otitis media, and 13 nasopharyngeal carrier isolates. The COPD study was approved by the institutional review board of the VA Western New York Healthcare System. Written informed consent was provided by study participants. The study was conducted according to the principles expressed in the Declaration of Helsinki. Patients with COPD were seen monthly and at the time of suspected exacerbation. A clinical evaluation using predetermined criteria was performed to assess whether the patient was experiencing an exacerbation or was clinically stable using previously described methods (5). The determination of whether the patient had stable disease or an exacerbation was made by one of two observers and was made before the results of sputum cultures were available. Expectorated sputum samples obtained during these visits were subjected to semiquantitative culture. An isolate was identified as H. influenzae on the basis of the following criteria: (i) colony morphology, (ii) satellite growth around a staphylococcus streak, (iii) growth requirement for hemin and NAD using Haemophilus ID Quad plates (Thermo Fisher), (iv) absence of porphyrin production (23), (v) absence of hemolysis on Quad plates, (vi) presence of the epitope on P6 recognized by the monoclonal antibody 7F3 in immunoblot assay, and (vii) the presence of the variable portion of iga by PCR (11, 24, 25) (Table 2). Some strains of H. haemolyticus showed hemolysis, but hemolysis alone was not a requirement to identify an isolate as H. haemolyticus. Five H. haemolyticus strains with published genome sequences (strains M19501, M19107, M21127, M21621, and M21639) were gifts of Leonard Meyer of the CDC (26). Three of the H. haemolyticus strains obtained from the CDC were isolated from patients with infections caused by the organism. The remaining H. haemolyticus strains were isolated from asymptomatic individuals.

TABLE 1.

Strains used in this study

Strain	Source
NTHi strains
1P22H3	COPD colonization
5P11H1	COPD colonization
14P14H1	COPD colonization
33P18H1	COPD colonization
40P41H1	COPD colonization
67P23H1	COPD colonization
73P2H1	COPD colonization
101P3H1	COPD colonization
103P3H1	COPD colonization
2019	COPD colonization
6P8H1	COPD exacerbation
12P56H1	COPD exacerbation
18P16H1	COPD exacerbation
33P46H1	COPD exacerbation
67P38H1	COPD exacerbation
47P68H1	COPD exacerbation
84P5H1	COPD exacerbation
99P15H1	COPD exacerbation
102P16H1	COPD exacerbation
121P2H1	COPD exacerbation
1749H1	Otitis media
2536H1	Otitis media
56LH1	Otitis media
63RH1	Otitis media
956H1	Otitis media
992H1	Otitis media
3017H1	Otitis media
3113H1	Otitis media
3127H1	Otitis media
P81H1	Otitis media
B032.3H1	Nasopharyngeal carriers
C042.1H1	Nasopharyngeal carriers
C072.3H1	Nasopharyngeal carriers
C222.1H1	Nasopharyngeal carriers
D12.6H1	Nasopharyngeal carriers
D62.3H1	Nasopharyngeal carriers
D72.6H1	Nasopharyngeal carriers
D122.3H1	Nasopharyngeal carriers
D132.16H1	Nasopharyngeal carriers
L22.1H1	Nasopharyngeal carriers
IC-1	Nasopharyngeal carriers
IC-2	Nasopharyngeal carriers
IC-3	Nasopharyngeal carriers
H. haemolyticus strains
1P26H1	COPD colonization
3P5H1	COPD colonization
3P18H1	COPD colonization
11P18H1	COPD colonization
27P25H1	COPD colonization
70P28H1	COPD colonization
35P12H1	COPD colonization
73P10H1	COPD colonization
MK-1	COPD colonization
M19501	Colonization
M19107	Colonization
M21127	Invasive disease
M21639	Invasive disease
M21621	Invasive disease

TABLE 2.

Primers used in this study

Gene	Primer direction	Primer sequence	Reference
nanA	Forward	5′-TTATGACAAAAATTTCGCTTTCAGGTCT	This study
	Reverse	5′-ATGCGTGATTTAAAAGGTATTTTCAGTGC
licD	Forward	5′-GTGATGATATATTTGAAATG	This study
	Reverse	5′-TGGAAGCTTCATATAATCTCCATAA
siaP	Forward	5′-CACCATGATGAAATTGACAAAA	17
	Reverse	5′-TGGATTGATTGCTTCAATTTGTTT
iga conserved region	Forward	5′-TGAATAACGAGGGGCAATATAAC	25
	Reverse	5′-TCACCGCACTTAATCACTGAAT
iga variable region	Forward	5′-GTTCCACCACCTGCGCCTGCTAC	25
	Reverse	5′- GTTATATTGCCCCTCGTTATTCAT

NTHi bacteria were grown on supplemented brain heart infusion agar (BHI) (Difco Laboratories, Detroit, MI) supplemented with 10 μg/ml NAD and 10 μg/ml hemin. H. haemolyticus strains were grown on chocolate agar (Remel Company, Lenexa, KS). Both organisms were grown at 37°C in 5% CO₂.

Isolation and neuraminidase treatment of LOS.

Bacterial cultures were centrifuged, and cell pellets were washed with phosphate-buffered saline (PBS). Samples were digested with proteinase K, and LOS was then extracted from the cells using a modified hot phenol-water method as previously described (27). LOS was suspended in water at 1 mg/ml. Approximately 10 μg of LOS was treated with 5 milliunits of Vibrio cholerae neuraminidase (Roche Applied Science) in neuraminidase buffer (0.15 M NaCl, 4 mM CaCl₂ [pH 5.5]) at 37°C for 2 h.

Cloning, PCR, and mutagenesis.

Mutations were made by insertion of kanamycin resistance cassettes into nanA for NTHi strain IC-1 and H. haemolyticus strain 18P16H1 and licD for NTHi strain IC-1 using methods previously described (17). Primers utilized in PCR are described in Table 2. All mutations were verified by sequence analysis of the mutated gene.

Whole-genome sequencing and annotation.

The DNA preparations of NTHi strain 2019 and H. haemolyticus strains 1P26H1, 3P5H1, 11P18H1, and 27P25H1 for Roche 454 sequencing were made using an Epicenter MasterPure purification kit (Madison, WI). Genome sequence data were generated using the GS-FLX titanium system (Roche, Branford, CT) at the University of Iowa DNA Facility using the manufacturer's recommended protocols. Briefly, 500 ng of genomic DNA was sheared via nebulization to give 400- to 800-bp fragments. Sequencing libraries containing barcoded adaptors were generated using the GS-FLX titanium rapid library preparation kit (Roche) in combination with the GS-FLX titanium rapid MID adaptor kit (Roche). The GS-FLX titanium LV emPCR kit (Lib-L) (Roche) was used for emulsion-based clonal amplification of each DNA library. The DNA sequencing run was performed using the GS-FLX titanium XLR70 sequencing kit (Roche) in combination with the GS-FLX titanium PicoTiterPlate kit (70 × 75; Roche). The sequences were assembled de novo using Newbler v2.3 (454 runAssembly software). The assembled sequences were queried for specific genes using Clustal Omega analysis. Annotation of the genomes was performed by the National Center for Biotechnology Information (NCBI).

Whole-cell ELISA.

Whole-cell enzyme-linked immunosorbent assays (ELISAs) were performed using the monoclonal antibody (MAb) 12D9 using a modified version of the method of Abdillahi and Poolman (28). Antibody 12D9 was a gift from Richard Lynch of the University of Iowa and binds to ChoP (29). Studies were also performed using a lectin from Maackia amurensis (MAL II), which binds Neu5Ac (Vector Laboratories, Burlingame, CA). Assays were repeated six times. All values presented were normalized to the value for the positive control (NTHi strain 2019) in each experiment to control for variation of absorbance readings. An ELISA value was considered positive if it fell outside 2 standard deviations of the mean of the antigen negative control. The nonparametric Mann-Whitney U test was used to test the null hypothesis that the distribution of absorbance relative to NTHi 2019 is the same for both H. haemolyticus and NTHi. The two-sample two-sided test was applied for each antibody or lectin at the 5% nominal significance level. Bonferroni's adjustment for multiple-hypothesis tests changes the nominal significance level from 0.05 to 0.0125. When ties were present, a normal approximation was applied to obtain the P values (30).

Silver-stained SDS-PAGE and Western blot analysis of LOS.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) of the LOS was performed by the method of Lesse et al. (31). The gel was subsequently silver stained by the method of Tsai and Frasch (32). MAb 3F11 was raised against Neisseria gonorrhoeae cells (28, 33) and was subsequently shown to react with N-acetyllacto-N-neotetraose (34). Western blotting was performed using MAb 3F11, using the method of Towbin et al. (35). A peroxidase-labeled goat anti-mouse IgM secondary antibody (Jackson ImmunoResearch Laboratories) and the Super Signal West Pico chemiluminescent substrate (Thermo Scientific) were used to detect the antibody. The previously characterized LOS from N. gonorrhoeae strain PID2 was used as a molecular mass marker (36).

Isothermal titration calorimetry.

Isothermal titration calorimetry (ITC) measurements were performed using a VP-ITC microcalorimeter (Microcal, Northhampton, MA) on purified SiaP from each strain as previously described (37).

MALDI-MS analyses of LOS.

To generate LOS more amendable to mass spectrometric analyses, O-deacylated LOS (O-LOS) samples were prepared by treating ∼50 μg of LOS with 50 μl of anhydrous hydrazine followed by acetone precipitation as previously described (38). The oligosaccharide (OS) was liberated from the lipid, and the phosphate groups were removed from the OS by treatment with 48% aqueous hydrofluoric acid (HF) for 16 to 24 h at 4°C. HF was removed and neutralized by drying the sample with nitrogen gas under vacuum and over sodium hydroxide, as previously described (39). Neuraminidase treatment of O-LOS was performed using type IV Clostridium perfringens neuraminidase immobilized on agarose beads (Sigma). Briefly, approximately 25 μg of O-LOS was reconstituted in 10 mM ammonium acetate buffer (pH 6) containing the beads, and samples were incubated at 30°C for 20 h with shaking. Beads were pelleted, and the sample was removed. All O-LOS samples were desalted by drop dialysis using 0.025-μm-pore-size nitrocellulose membranes (Millipore, Bedford, MA) and were subsequently lyophilized. Samples were reconstituted in 5 to 20 μl of high-performance liquid chromatography (HPLC)-grade H₂O, and 1 μl of the reconstituted sample was loaded onto the target, allowed to dry, and then overlaid with 1 μl of matrix (50 mg/ml 2,5-dihydroxybenzoic acid [DHB] [Laser Biolabs, Sophia-Antipolis, France] in 70% acetonitrile). Samples were subsequently analyzed using matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) on an LTQ linear ion trap mass spectrometer coupled to a vMALDI ion source (MALDI-LIT) (Thermo Fisher, Waltham, MA). The vMALDI source uses a nitrogen laser which operates at 337.1-nm wavelength, 3-ns pulse duration, and 60 Hz repetition rate. Data were collected in the negative-ion or positive-ion mode using the automated gain control (AGC) and the automatic spectrum filter (ASF) settings. Tandem mass spectrometry (MSⁿ) data were collected using a precursor ion selection window of 2 to 3 m/z and normalized collision energy of 35 to 40%.

Nucleotide sequence accession numbers.

The following NCBI accession numbers were assigned: H. haemolyticus strain 3P5H1 genome accession number PRJNA282678, H. haemolyticus strain 1P26H1 genome accession number PRJNA282677, H. haemolyticus strain 27P25H1 genome accession number PRJNA282658, H. haemolyticus strain 11P18H1 genome accession number PRJNA282951 and NTHi strain 2019 genome accession number PRJNA283325.

RESULTS

Whole-cell ELISA studies.

H. haemolyticus and NTHi are closely related genetically yet have considerable differences in their tendency toward pathogenicity. To study the potential role of the LOS in these clinical differences, we compared the expression of LOS structures known to be associated with pathogenicity using MAb 12D9, which reacts with ChoP, and a Neu5Ac binding lectin MAL II. The results of these ELISA studies comparing the differences between H. haemolyticus and NTHi are shown in Fig. 1. The MAL II lectin detects the presence of Neu5Ac regardless of the acceptor. Mean values of the H. haemolyticus and NTHi strains tested with the lectin MAL II were 0.05 and 0.40, respectively (Fig. 1A). MAL II bound H. haemolyticus significantly less frequently than NTHi did, with a P value of P < 0.0002. Our ELISA data showed that H. haemolyticus ChoP expression is significantly less than that seen among NTHi strains (P < 0.0443 [Fig. 1B]) but was not considered significant when Bonferroni's adjustment for multiple-hypothesis tests was applied.

FIG 1.

ELISA studies with the lectin, MAL II (A) and MAb 12D9 (B). The binding of the MALII lectin was significantly reduced in H. haemolyticus strains compared to the NTHi strains. MAb 12D9 bound significantly greater to the NTHi strains than to the H. haemolyticus strains, but application of Bonferrroni's adjustment suggested that the differences might not be significant. Each symbol represents the value for an individual. The mean for a group of individuals is indicated by a horizontal line. Data shown represent the mean values from six experiments.

Genomic analyses of sialic acid incorporation genes.

Genomic analysis for the four sialyltransferase genes known to be involved in sialylation of LOS in NTHi was performed on the 10 sequenced H. haemolyticus strains. These analyses showed that siaA was absent in all of the H. haemolyticus strains examined (Table 3). Nine of the 10 strains examined did not contain homologues for lic3A or lic3B. M21639, an invasive strain, contained these genes, but they appeared truncated. The predicted protein sequences for Lic3A and Lic3B from strain M21639 had minimal homology to only the first 146 and 96 amino acids, respectively, to these proteins from NTHi strain 2019. The sialyltransferase gene lsgB was found in 9 of the 10 strains examined, with the colonizing strain M19107 being the exception. The predicted protein sequences of LsgB from the various H. haemolyticus strains showed strong homology to LsgB from NTHi 2019 over the first 222/304 amino acids. The genes involved in sialic acid uptake (siaP and siaT) and sialic acid degradation were intact in all of the H. haemolyticus strains examined (data not shown).

TABLE 3.

Genomic analyses of genes involved in Neu5Ac transfer in sequenced H. haemolyticus genomes^a

Strain	Presence of the following geneb:
	lsgB	lic3A or lic3B	siaA
HK386	Y	N	N
M19501	Y	N	N
M19107	N	N	N
M21127	Y	N	N
M21639	Y	Y	N
M21621	Y	N	N
1P26H1c	Y	N	N
3P5H1c	Y	N	N
11P18H1c	Y	N	N
27P25H1c	Y	N	N
2019c,d	Y	Y	Y

^aNeu5Ac, N-acetyl-5-neuraminic acid.

^bThe presence of the gene is indicated as follows: Y (for yes) or N (for no). Positive results are shown in boldface type for emphasis.

^cThe genomes of these strains were sequenced and annotated in this study.

^dThis strain is the prototypical NTHi strain, strain 2019.

Isothermal titration calorimetry.

We used ITC to compare the binding affinities of SiaP from NTHi strain 2019 with two H. haemolyticus strains, one isolated from the nasopharynx of a healthy individual (strain M19501) and one from a patient with invasive disease (strain M2711217). The dissociation constant for SiaP binding to sialic acid for the NTHi strain 2019 was 34.7 nM (Fig. 2A). Under the same conditions as NTHi strain 2019, the dissociation constants for H. haemolyticus strains M19501 and M21127 were 119 nM and 253 nM, respectively (Fig. 2B and C). BLAST analysis and alignment, using Clustal Omega, of the sequences from 7 NTHi strains showed that SiaP is consistent across the entire 310 amino acids (see Fig. S1 in the supplemental material). The SiaP sequences from 10 H. haemolyticus strains were determined to be ∼93% identical to NTHi 2019; however, the changes which do occur are localized primarily in the region from amino acids 100 to 149. This region of the SiaP sequences from both NTHi and H. haemolyticus strains can be seen in Fig. S1. While the specific amino acids in this region are not involved in Neu5Ac binding, conformational changes caused by these amino acid substitutions may account in part for the lower affinities of SiaP from H. haemolyticus strains M19501 and M21127 for Neu5Ac compared to SiaP from NTHi 2019.

FIG 2.

ITC studies performed on affinity-purified SiaP. Panel A demonstrates a dissociation constant (K_d) of 35 nM for SiaP isolated from NTHi strain 2019. Panels B and C show K_ds of 119 and 472 nM for purified SiaP isolated from H. haemolyticus strains M21127 and M19501, respectively. All studies were performed on two separate occasions, and similar results were obtained.

SDS-PAGE and Western blot analyses of LOS.

LOS from a subset of four H. haemolyticus strains and five NTHi strains were analyzed by silver-stained SDS-PAGE (Fig. 3A and B). Neuraminidase treatment, which removes sialic acid from LOS, was used to determine whether sialylated LOS was expressed in the various strains. LOS was examined before and after neuraminidase treatment to determine whether any changes in the banding pattern occurred. The LOS from the H. haemolyticus colonizing strain, 1P26H1, and from two of the NTHi strains, IC-1 and 18P16H1, showed altered banding patterns after neuraminidase treatment (Fig. 3A and B), indicating that at least one of the LOS glycoforms expressed by these strains contains sialic acid. NTHi strain 2019 is known to express sialylated LOS structures, and as expected, neuraminidase treatment affected the banding pattern of the 2019 LOS (Fig. 3B).

FIG 3.

Silver-stained SDS-PAGE and Western blot analyses of the LOS from a subset of NTHi and H. haemolyticus strains. The LOS from four H. haemolyticus strains (strains 1P26H1, 3P5H1, 11P18H1, and 27P25H1) and six NTHi strains (strains IC-1, 2019, 5P11H1, 6P8H1, 14P14H1, and 18P16H1) were analyzed by silver-stained SDS-PAGE (A and B) or Western blot analyses with MAb 3F11 (C and D). Lanes labeled with an asterisk represent LOS samples treated with neuraminidase. LOS from the previously characterized Neisseria gonorrhoeae strain PID2 was included in the first and last lanes of each gel and blot as a molecular mass marker.

Western blot analyses using MAb 3F11 were also performed on the subset of H. haemolyticus and NTHi strains. Figure 3C shows that the H. haemolyticus colonizing strain 3P5H1 and the NTHi strain IC-1 reacted with the antibody before and after neuraminidase treatment, so the presence of sialic acid had no impact on the ability of the antibody to bind the LOS. In contrast, the 3F11 antibody reacted with NTHi 2019 LOS only after neuraminidase treatment (Fig. 3D). This reaction has been previously observed for strain 2019 and is due to the presence of sialic acid on the terminal N-acetyllactosamine structure in strain 2019 (38). None of the remaining LOS samples reacted with MAb 3F11 (Fig. 3C and D).

MALDI-MS analyses of O-LOS.

LOS from a subgroup of NTHi and H. haemolyticus strains were analyzed by MALDI-MS and, for the most part, were consistent with previously reported NTHi LOS structures (Fig. 4) containing a conserved triheptose core linked to the lipid A moiety through a single phosphorylated 2-keto-3-deoxyoctonate (KDO) (40–44). In NTHi, all three core heptoses can be further substituted with oligosaccharide chains, free phosphate (P), phosphoethanolamine (PEA), ChoP, acetate groups, and glycine (13, 41–48). The addition of these various moieties can create highly heterogeneous LOS structures both between strains and within a strain. The MALDI-MS spectra of O-LOS from the five representative NTHi strains examined in this study showed these variable structural features (Fig. 4A and B and see Fig. S2 to Fig. S4 in the supplemental material). For example, all five strains expressed LOS with at least some ChoP-containing glycoforms (Table 4). The addition of ChoP has been previously observed in a number of NTHi strains and can occur at any of the three core heptose (Hep) units (Fig. 4) (41, 44–48). These findings are in contrast to the most prevalent glycoforms expressed by NTHi strain 2019 (see Fig. S5 and Table S1 in the supplemental material) (38, 42). O-LOS from NTHi 2019 typically does not express high levels of ChoP-containing glycoforms (Fig. S5 and Table S1) (38, 42). Additionally, the five NTHi strains, examined in this study, also showed a propensity toward expressing glycoforms with an additional heptose. The addition of a fourth heptose in the branch region has been previously observed and seems to typically occur on the HepI unit, the first core heptose linked to KDO (Fig. 4) (43, 44). We have not observed the addition of an extra Hep in the branch structure of NTHi 2019 LOS (Fig. S5 and Table S1) (38, 42).

FIG 4.

Comparison of O-LOS glycoforms from representative NTHi and H. haemolyticus strains. Negative-ion MALDI-MS spectra of O-LOS from NTHi strains IC-1 (A) and18P16H1 (B) and H. haemolyticus strains M19107 (C) and M21621 (D) are shown. Major glycoforms for each strain are labeled, and their proposed OS branch structures are shown in the table included in the figure. All glycoforms are built on the conserved core structure shown in the figure; the OS branch structures are added to the core Hep sugars. The subscript number in each glycoform represents how many PEA are present in the O-LOS structure. The observed and calculated monoisotopic masses [M-H] are listed in Table S1 in the supplemental material. Less abundant unlabeled peaks typically represent observed major glycoforms with salt adducts, plus or minus H₂O, or minus phosphate.

TABLE 4.

Analysis of LOS glycoforms

Strain	Presence of an LOS glycoform with the followinga:
	ChoP	Neu5Ac
NTHi strains
6P8H1	Y	N
14P14H1	Y	N
5P11H1	Y	N
18P16H1	Y	Y
IC-1	Y	Y
2019	N	Y
H. haemolyticus strains
11P18H1	Y	N
3P5H1	N	N
27P25H1	N	N
1P26H1	Y	Y
M21639	Y	N
M19107	Y	N
M21621	N	N
M19501	N	N
M21127	N	N

^aChoP, phosphorylcholine; Neu5Ac, N-acetyl-5-neuraminic acid. The presence of a glycoform with ChoP or NeuAc is indicated as follows: Y (for yes) or N (for no). Positive results are shown in boldface type for emphasis.

In addition to NTHi strain 2019, two NTHi strains analyzed by MALDI-MS were shown to be sialylated (Table 4). NTHi strain 18P16H1 and its corresponding nanA mutant were examined (Fig. 4B; see Fig. S3A and S3C in the supplemental material). The parent strain, 18P16H1, predominantly expressed an LOS with a branch consisting of Hex₄, HexNAc₁, PEA₂ and ChoP at m/z 3090.18 (Fig. 4B; Fig. S3A and Table S1). Sialylated glycoforms were also observed in strain 18P16H1 at m/z 3016.09 and 3307.36, corresponding to structures consisting of Hex₃, PEA₂ with one or two NeuAc, respectively (Fig. 4B; Fig. S3A and Table S1). In the 18P16H1 nanA mutant, sialic acid cannot be catabolized; therefore, more sialic acid is available for incorporation into the LOS. We observed an increase in the level of the sialylated glycoforms in the 18P16H1 nanA O-LOS (Fig. S3C) similar to what has been previously observed in a strain 2019 nanA mutant (38). Neuraminidase treatment of the O-LOS removed the sialic acid from the sialylated glycoforms (Fig. S3B and S3D).

Low abundant sialylated glycoforms were also observed in NTHi strain IC-1 at m/z 2892.82 and 3015.82, corresponding to an LOS with a branch structure containing Hex₃ and NeuAc, with either one or two PEA, respectively (Fig. 4A and see Fig. S4A and Table S1 in the supplemental material). Neuraminidase treatment confirmed that these glycoforms were sialylated (Fig. S4B). Similar to the 18P16H1 nanA mutant, an IC-1 nanA mutant also showed increased levels of sialylation (Fig. S4C). The presence of ChoP was observed in at least some of the glycoforms from all of the NTHi strains examined by MALDI-MS in this study. ChoP is transferred from choline diphosphonucleoside to the LOS by the LicD transferase. A licD mutant from strain IC-1 was used to confirm the presence of ChoP in its LOS (Fig. S4D). MALDI-MS of the IC-1 licD O-LOS showed a structure lacking ChoP, consisting of Hex₃ and one or two PEA with and without NeuAc (Fig. S4D).

The predominant glycoforms in the strain 6P8H1 O-LOS were observed at m/z 3291.27 and 3414.45 and correspond to branch structures consisting of Hex₆, HexNAc, and ChoP with one or two PEA substitutions, respectively (see Fig. S2A and Table S1 in the supplemental material). Two predominant glycoforms were observed in the O-LOS from NTHi strain 14P14H1 (Fig. S2B). The first major glycoform was observed at m/z 2847.91, corresponding to a structure consisting of Hex₃, PEA₃, and ChoP (Fig. S2B and Table S1). The second major glycoform observed in 14P14H1 O-LOS was observed at m/z 3037.00 and corresponds to a branch structure consisting of Hex₄, PEA₃, and Hep (Fig. S2B and Table S1). The major glycoform present in O-LOS from strain 5P11H1 was observed at m/z 3282.36, and it corresponds to a structure with a composition of Hex₄, HexNAc, PEA, Hep, and ChoP (Fig. S2C and Table S1). Strain 5P11H1 is the only strain examined in this study to express an LOS containing both ChoP and the additional Hep in its branch structure.

The O-LOS from nine H. haemolyticus strains were also examined by MALDI-MS. Our data were consistent with a conserved core structure for H. haemolyticus LOS that was the same as that observed in NTHi, i.e., a triheptose linked to a phosphorylated KDO (Fig. 4). Overall, there seemed to be less complexity in the number of glycoforms expressed by each H. haemolyticus strain compared to the number of various glycoforms expressed by the individual NTHi strains that were examined (Fig. 4 and see Fig. S2 to S7 in the supplemental material). Four of the nine (44%) H. haemolyticus strains expressed glycoforms containing ChoP (Table 4 and Fig. 4; Fig. S6 and S7), and only one of the nine strains (11%) expressed a sialylated glycoform (Table 4; Fig. S7A).

The glycoforms consisting of Hex₂ and ChoP with two to four PEAs were observed as the predominant structures in H. haemolyticus strains M19107, 11P18H1, M21639 (an invasive strain), and 1P26H1 at m/z 2685.82, 2808.82, 2562.73, and 2562.73, respectively (Fig. 4C; see Fig. S6C and S6D, Fig. S7A, and Table S1 in the supplemental material). Similar glycoforms, lacking ChoP, were observed in the O-LOS from H. haemolyticus colonizing strain M19501 (Fig. S6B). Invasive strains M21621 and M21127 also expressed LOS structures that lacked ChoP. The predominant glycoforms present in their O-LOS were observed at m/z 2598.91 and 2721.73 for M21621 and at m/z 2721.73 and 2844.91 for M21127, corresponding to LOS with branch structures consisting of Hex₄ with one to three PEA substitutions (Fig. 4D; Fig. S6A and Table S1). Less abundant and previously unidentified glycoforms were also observed in the M21127 O-LOS at m/z 2194.64 and 2414.73 (Fig. S6A). The masses of these glycoforms are consistent with an LOS having a core structure which lacks the phosphate group and instead consists of Hep₃ with one or two KDOs. Minor peaks, which most likely correspond to glycoforms expressing this alternate core structure, were also observed in the O-LOS from strain 3P5H1 (Fig. S7C). These glycoforms were observed at m/z 2414.73, 2575.82, 2738.00, and 2861.09 and correspond to the proposed alternative core structure with the addition of zero to three hexoses, respectively (Fig. S7C).

The O-LOS from H. haemolyticus colonizing strains 3P5H1 and 27P25H1 were found to predominantly express high-molecular-mass glycoforms (see Fig. S7C and S7D and Table S1 in the supplemental material). The most abundant glycoforms observed in the O-LOS from strain 27P25H1 were found at m/z 2884.18 and 3007.27, and they correspond to a branch structure consisting of Hex₂ and HexNAc₃ with one or two PEA substitutions, respectively (Fig. S7D and Table S1). The most abundant glycoforms observed in strain 3P5H1 were observed at m/z 2964.09 and 3087.18, and they correspond to a branch structure consisting of Hex₅ and HexNAc with one or two PEA substitutions, respectively (Fig. S7C and Table S1). Confirmation of the proper composition for strain 27P25H1 was performed using multistage MALDI-MS analyses of an HF-treated oligosaccharide (OS). The HF treatment removes any phosphate groups from the LOS and releases the OS from the O-deacylated lipid A. The dephosphorylated OS was examined in positive-ion mode and was observed at m/z 1752.1 in its sodiated form (Fig. S8A). Figure S8A shows that MS² fragmentation of this precursor ion generated a large fragment ion at m/z 1549.2, corresponding to the loss of one HexNAc (−203 Da). MS³ fragmentation of this fragment ion generated a large fragment ion at m/z 1346.1, corresponding to the loss of another HexNAc (−203 Da) (Fig. S8B). Further fragmentation (MS⁴) of the ion observed at m/z 1346.0 generated a predominant fragment ion at m/z 1143.1, corresponding to the loss of a third HexNAc (−203 Da) from the OS (Fig. S8C). MS⁵ fragmentation of the ion observed at m/z 1143.0 showed the loss of one or two Hex (−162 Da each) seen at m/z 981.1 and 819.0, respectively (Fig. S8D). The peak observed at m/z 819.0 corresponds to a sodiated structure consisting of Hep₃ with one KDO. Overall, these fragmentation analyses confirmed the composition of the proposed Hex₂ and HexNAc₃ branch for the 27P25H1 LOS.

Of the nine H. haemolyticus strains analyzed, only strain 1P26H1 (a colonizing strain) expressed any sialylated LOS glycoforms (Table 4 and see Fig. S7A in the supplemental material). The sialylated glycoforms observed in strain 1P26H1 were very minor LOS species observed at m/z 2874.82 and 3201.82 (Fig. S7A). The overall composition of these structures is currently unknown, but the presence of sialic acid on the LOS structures was confirmed by neuraminidase treatment of the 1P26H1 O-LOS (Fig. S7B). After treatment with neuraminidase, the peaks previously observed at m/z 2874.82 and 3201.82 were no longer detected; however, their nonsialylated counterparts were still detected at m/z 2583.73 and 2910.73, respectively (Fig. S7B).

DISCUSSION

Distinguishing between NTHi and H. haemolyticus has challenged investigators for many years. Recent studies employing ribosomal 16S sequencing and whole-genome analysis have provided a number of means by which this differentiation can be made (8–12). Genotypic analysis of both species has demonstrated differences in the frequency of genes in the lic1 operon and in transferases responsible for core LOS structures (21, 22). Since the LOS of NTHi has been shown to be important in colonization (19) and pathogenesis (15), a detailed phenotypic and genotypic comparison of H. haemolyticus LOS to NTHi LOS was undertaken and described here.

ChoP expression on NTHi LOS has been shown to be important in initiating uptake by binding to the platelet-activating factor (PAF) receptor (49). This binding initiates a cell signaling cascade and allows the bacteria to invade the host cell (49). Additionally, ChoP-expressing bacteria have been shown to exhibit increased resistance to bactericidal antibody killing and antimicrobial peptides (20, 50). ChoP has also been shown to play a role in biofilm maturation (51, 52). Paradoxically, ChoP also increases binding of the C-reactive protein (CRP) and activation of the classical complement pathway (53); therefore, bacteria expressing ChoP need mechanisms for controlling its surface expression. ChoP is synthesized and transferred to LOS by genes in the lic1 operon. Using dot blot hybridization, McCrea and coworkers showed that less than half (42.6%) of 109 H. haemolyticus strains had a completely intact lic1 operon compared to more than 90% of 88 NTHi strains (22). Similarly, mass spectrometric analyses of LOS demonstrated that less than half (44%) of the H. haemolyticus strains examined in the present study expressed LOS structures with ChoP. LOS from all five of the NTHi strains examined in this study contained glycoforms with ChoP. ELISAs with the ChoP-specific MAb 12D9 further confirmed these results. These data demonstrate that NTHi has a greater propensity than H. haemolyticus to express ChoP on its LOS. These differences seem to arise at least in part due to the presence or absence of the lic1 operon.

Sialylation of NTHi LOS has been shown to be important in bacterial resistance to human serum (13, 15). LOS sialylation has also been shown to be important for biofilm formation and persistence in vitro and in vivo (16, 54, 55). Therefore, expression of sialylated LOS glycoforms clearly gives organisms a pathogenic advantage. In the present study, the ELISA results with the lectin MAL II showed that the LOS from 40% of the NTHi strains (17/42 strains) examined expressed sialic acid, while only 11% or 1/9 strains of H. haemolyticus (1P26H1) showed evidence of its expression. Similarly, mass spectrometric analyses showed that only one H. haemolyticus strain, 1P26H1, had a low-abundance sialylated LOS glycoform. These differences in expression of sialic acid-containing LOS structures between NTHi and H. haemolyticus strains could be due to several factors, such as the lack of sialyltransferases that add Neu5Ac to the Galα1-3Gal structure (Lic3A and Lic3B) and/or the paucity of N-acetyllactosamine acceptors for the LsgB and SiaA sialyltransferases. In addition, while all of the H. haemolyticus strains studied had copies of siaP, the Neu5Ac periplasmic binding protein, and siaT, the Neu5Ac transporter, ITC studies of SiaP from two H. haemolyticus strains demonstrated that they had affinities for Neu5Ac 3.4- to 7.3-fold lower than NTHi strain 2019 SiaP, suggesting an uptake defect in H. haemolyticus strains. Based on this data, it appears that H. haemolyticus can utilize Neu5Ac as a carbon and nitrogen source but is for the most part unable to incorporate it into its LOS.

NTHi and H. haemolyticus both reside in the nasopharynges of humans and are generally considered commensals. However, NTHi has the ability to cause both localized and systemic disease in children and in adults with respiratory problems. H. haemolyticus has typically been considered a colonizer with very limited potential to cause human infection. In a study by Murphy et al., they were unable to culture H. haemolyticus from normally sterile sites in the body while they were able to culture NTHi from these sites (11), further confirming H. haemolyticus as a commensal organism. Recently, a select group of H. haemolyticus strains were found to be able to cause systemic disease, but these patients were likely immunocompromised, primarily due to other illnesses or surgery (56). These studies strongly suggest that H. haemolyticus is typically a commensal organism and is only able to cause disease under constrained conditions such as an immunocompromised individual. Several recent studies further indicate that H. haemolyticus is rarely pathogenic. A retrospective analysis of 161 invasive NTHi strains at the CDC in 2012 demonstrated that 7 were reclassified as H. haemolyticus (56), and a study of 514 NTHi otitis media strains indicated that only 3 were reclassified as H. haemolyticus (57, 58). Our data presented in this study demonstrate that H. haemolyticus LOS structures are generally unsialylated and express ChoP on their LOS at lower levels than NTHi strains. These differences in the incorporation of Neu5Ac and ChoP into the LOS of NTHi and H. haemolyticus may be a factor contributing to their differing propensities for causing disease.