Sialic Acid Protects Nontypeable Haemophilus influenzae from Natural IgM and Promotes Survival in Murine Respiratory Tract

ABSTRACT

Nontypeable Haemophilus influenzae (NTHi), a common inhabitant of the human nasopharynx and upper airways, causes opportunistic respiratory tract infections that are frequently recurring and chronic. NTHi utilizes sialic acid from the host to evade antibacterial defenses and persist in mucosal tissues; however, the role of sialic acid scavenged by NTHi during infection is not fully understood. We previously showed that sialylation protects specific epitopes on NTHi lipooligosaccharide (LOS) targeted by bactericidal IgM in normal human serum. Here, we evaluated the importance of immune evasion mediated by LOS sialylation in the mouse respiratory tract using wild-type H. influenzae and an isogenic siaB mutant incapable of sialylating the LOS. Sialylation protected common NTHi glycan structures recognized by human and murine IgM and protected NTHi from complement-mediated killing directed by IgM against these structures. Protection from IgM binding by sialylated LOS correlated with decreased survival of the siaB mutant versus the wild type in the murine lung. Complement depletion with cobra venom factor increased survival of the siaB mutant in the nasopharynx but not in the lungs, suggesting differing roles of sialylation at these sites. Prior infection increased IgM against H. influenzae but not against sialic acid-protected epitopes, consistent with sialic acid-mediated immune evasion during infection. These results provide mechanistic insight into an NTHi evasive strategy against an immune defense conserved across host species, highlighting the potential of the mouse model for development of anti-infective strategies targeting LOS antigens of NTHi.

INTRODUCTION

Haemophilus influenzae, a gram-negative bacterium, persistently colonizes the upper respiratory tract of healthy individuals but can spread to infect other mucosal sites to cause acute and recurrent disease. Infections caused by both encapsulated and nonencapsulated H. influenzae, also termed nontypeable H. influenzae (NTHi), include otitis media, bronchitis, sinusitis, pneumonia, exacerbation of chronic obstructive pulmonary disease, and invasive disease (1, 2). Since the introduction of the vaccine for H. influenzae capsule type b (Hib), the incidence of invasive disease caused by non-type b and nontypeable strains, for which there are currently no vaccines, has increased. Additionally, other serious infections caused by NTHi remain prevalent, accentuating the need to understand the factors that contribute to NTHi pathogenesis in order to develop new strategies to combat infections (1, 3, 4).

Pathogens exploit host resources to aid in survival and immune evasion. To cause recurrent and persistent infections, H. influenzae must evade innate and adaptive immunity. Several factors can contribute to survival of NTHi, including the lipooligosaccharide (LOS) composition. Sialic acid, which NTHi cannot synthesize but must acquire from the host, is incorporated as a terminal extension on the LOS of most or all NTHi strains, and this LOS sialylation is known to contribute to resistance to antibody-mediated complement killing (5–7). Antibodies are vital in vertebrate immunity to invading pathogens. IgM antibodies are especially critical in early defense against microbes (8). IgM found in serum can possess a range of specificities for distinct surface antigens, giving them markedly different properties and functions during an immune response (9). For instance, antibodies against pathogens do not all exhibit the same degree of activity; i.e., they can be bactericidal or nonbactericidal (10). In previous studies with NTHi, we found that serum antibodies recognize various surface LOS structures, but the degree of bacterial killing in vitro may differ depending on the antibody specificity (7). Hence, it is important to identify bacterial antigens that elicit a protective, bactericidal antibody response that is effective in vivo.

Mice are commonly used to model aspects of disease caused by H. influenzae. Although mice are not naturally infected with H. influenzae, certain murine natural antibodies have been identified that recognize not only self antigens but also foreign antigens (11). For example, IgM specific for phosphorylcholine recognizes this structure on the surface of various microbes (12). Importantly, natural IgM, defined as antibody generated in the absence of exogenous antigen exposure, constitutes the majority of total circulating IgM in both humans and mice (13–15). However, the repertoire of natural IgM antibodies in mice has not been fully characterized, and despite our growing knowledge of the importance of IgM in immunity, this antibody isotype has not been well studied in the context of microbial infections in vivo.

We previously identified IgM antibodies in human serum which are inhibited from binding to specific LOS structures on the surface of various NTHi strains expressing sialylated LOS (Fig. 1) (7). In this report, we determined that mice possess bactericidal IgM specific to sialic acid-protected LOS structures similarly recognized by human IgM. We investigated the effect of prior immunization on these responses and demonstrated a role for sialylation in survival of H. influenzae in the mouse lung infection model. Our results indicate that LOS sialylation is important in protecting NTHi from host defenses, but this protective role may differ between respiratory mucosal sites, such as the lungs and the nasopharynx. Overall, our results support the utility of the mouse model to evaluate the role of LOS sialylation and provide insight into evasion of humoral immunity by H. influenzae during infection.

FIG 1.

LOS diagram of NTHi 375 and serum IgM targets. The diagram is based on structural information described elsewhere (5) and previously determined targets of IgM in human serum (7). Glc, glucose; Gal, galactose; Neu5Ac, N-acetylneuraminic acid; GalNAc, N-acetylgalactosamine; PC, phosphorylcholine; Hep, heptose; Kdo, 2-keto-3-deoxyoctulosonic acid. Competition between Lic3A and LgtC is indicated by dashed line. Lic3B is a bifunctional α-2,3- and α-2,8-sialyltransferase present in some NTHi strains (indicated by dashed brackets), including 375.

RESULTS

Mice have natural IgM antibodies specific for NTHi LOS epitopes that are protected by sialic acid.

Previously, we identified targets of natural anti-Gal (galactose) IgM antibodies in humans (7). These antibodies recognize Gal-containing epitopes within the terminal βGal (Glcβ1-4Gal) and terminal αGal (Glcβ1-4Galα1-4Gal) structures created by sequential galactose additions to the LOS by Lic2A and LgtC (Fig. 1). Because synthesis of these structures requires galactosyltransferases Lic2A and LgtC, and because with no Gal residues the epitopes are absent, we deduce that at least the terminal Gal of each structure is a required component of the epitopes. However, addition of a terminal sialic acid (Neu5Ac) to the LOS (Glcβ1-4Galα2-3Neu5Ac) can inhibit IgM from binding each of these targets on the surface of NTHi (7). To determine if mice produce IgM specifically recognizing sialic acid-protected LOS epitopes on NTHi similar to those recognized by human serum, heat-inactivated C57BL/6 mouse serum was preabsorbed with 375 wild-type (WT) and Δlic2A strains grown on chemically defined MIc medium containing no sialic acid (see Materials and Methods). The truncated LOS of the Δlic2A mutant lacks the sialic acid-protected IgM epitopes containing βGal and αGal (7). Thus, antibodies against the terminal βGal and αGal structures, designated t-βGal and t-αGal, remain in Δlic2A preabsorbed serum (Δlic2A^SA− serum^abs) (Fig. 2A). In contrast, the 375 WT strain grown without sialic acid displays the galactose targets and consequently depletes serum of IgM specific for epitopes dependent on Lic2A and LgtC. We used this absorption strategy to evaluate murine serum IgM binding to 375 WT grown in either the absence (WT^SA−) or presence (WT^SA+) of sialic acid (Fig. 2B). Our results show that there is significantly more IgM in Δlic2A^SA− serum^abs that binds to 375 WT^SA− than WT^SA+. In contrast, in WT preabsorbed serum (WT^SA− serum^abs), IgM binding to WT^SA− decreases to levels nearly equal to that of WT^SA+. These results indicate that mouse serum contains IgM specific for sialic acid-protected LOS epitopes on NTHi strain 375.

FIG 2.

Sialylation protects NTHi LOS epitopes from bactericidal IgM in mouse serum. Heat-inactivated C57BL/6 mouse serum was preabsorbed with NTHi 375 wild type (WT) and an isogenic Δlic2A mutant grown on MIc^SA−. (A) Diagram depicting the differences in LOS structures on HepIII between these two strains and the IgM antibodies expected to remain in preabsorbed sera. The indicated galactose residues (βGal and αGal) are within the terminal βGal and αGal structures (t-βGal and t-αGal) which contain IgM epitopes. Also shown are corresponding antibodies to both epitopes predicted to remain in Δlic2A^SA− preabsorbed serum. Neither of these IgM species is predicted to remain in WT^SA− preabsorbed serum. (B) 375 WT grown on MIc^SA± was incubated with each preabsorbed serum (serum^abs) and IgM binding was measured by flow cytometry. Binding is depicted as a percentage of the maximum, calculated by normalizing the median fluorescence intensity (MFI) values of triplicate samples to the highest MFI. Statistical significance was evaluated by two-way ANOVA with Bonferroni’s multiple-comparison test. (C) 375 WT and the lgtC′ mutant were grown on MIc^SA± and incubated with each preabsorbed serum (Serumabs). Percent survival is the ratio of the number of CFU recovered from serum^abs-treated samples at 30 min to the number of input CFU at time zero. Means for triplicate samples are shown. The dashed line indicates the lower limit of detection at approximately 0.07%. Statistical significance of log-transformed survival ratios was evaluated by one-way ANOVA with Bonferroni’s multiple-comparison test (****, P < 0.0001; ***, P < 0.001; **, P < 0.01; ns, not statistically significant).

LOS-specific IgM in mice exhibits bactericidal activity against NTHi.

We further characterized these LOS-specific IgM antibodies found in mouse serum by investigating their bactericidal potential, demonstrated by the ability to activate complement via the classical pathway and ultimately lyse the bacteria. Here, we used 375 WT and a mutant containing a disruption in lgtC (lgtC′) to narrow in on the murine antibody response to the terminal αGal and βGal structures, respectively. In our previous studies, we concluded that the N-acetylgalactosamine (GalNAc) added by LgtD is likely not part of a major IgM epitope because a lgtD deletion mutant retains a significant amount of IgM binding and protection by sialic acid. More importantly, IgM binding to epitopes within the terminal αGal and βGal structures is not blocked by the LgtD-dependent GalNAc present on nonsialylated 375 WT. Both the exposed terminal αGal structure and the underlying terminal βGal structure, also exposed on the surface of the lgtC′ mutant, make these strains much more susceptible to bactericidal activity in human serum in the absence of sialylation (7).

To differentiate between bactericidal and nonbactericidal murine antibodies targeting NTHi, we supplemented preabsorbed mouse sera (WT^SA− serum^abs or Δlic2A^SA− serum^abs) with active human serum complement (see Materials and Methods) and performed bactericidal assays on WT and the lgtC′ mutant. When grown with sialic acid, all strains exhibited nearly uniform survival regardless of the preabsorbed sera they had been incubated with (Fig. 2C). Sialic acid protection was comparable between WT^SA+ and lgtC′ ^SA+ in both preabsorbed sera, indicating that sialylation efficiently protects the LOS IgM targets present in each strain. 375 WT^SA− was significantly more serum sensitive than 375 WT^SA+ in both preabsorbed sera; however, WT^SA− serum^abs kills less efficiently because of the relatively small amounts of antibodies present against the terminal galactose structures compared to those in Δlic2A^SA− serum^abs. Removal of mouse IgM directed against Lic2A- and LgtC-dependent epitopes nearly tripled the survival of WT^SA− (2.8-fold). In the absence of sialic acid, the lgtC′ mutant is extremely serum sensitive, as we previously determined (7), despite its lack of expression of the LgtC-dependent galactose target. The underlying galactose displayed by the nonsialylated lgtC′ mutant is bound by murine IgM that remains in Δlic2A^SA− serum^abs but is removed from WT^SA− serum^abs. Consequently, survival of lgtC′ ^SA− decreases in Δlic2A^SA− serum^abs. Together, these results confirm that sialic acid effectively inhibits bactericidal IgM in mouse serum targeting terminal galactose LOS structures on the surface of NTHi.

Sialic acid protects against shared epitope specificities of human and mouse IgM.

To compare murine and human antibodies to NTHi, we measured IgM in preabsorbed sera from both species by ELISA. Overall, human and murine antibodies in preabsorbed sera exhibited a similar pattern of reactivity, with increased binding to 375 WT and lgtC′ grown on MIc^SA− versus MIc^SA+ (Fig. 3A and B). Differential antibody binding to strains in Δlic2A^SA− serum^abs was diminished in WT^SA− serum^abs. We also measured binding to lic2A phase variants in the lgtC′ mutant background. These variants provide a direct comparison to further confirm that the lic2A gene is required to generate the epitope targeted by IgM (Fig. 3C and D). The lic2A gene is active in the lgtC′ lic2A^ON variant, but the gene has a nonsense mutation and is inactive in the lgtC′ lic2A^OFF variant, so that it behaves like a Δlic2A mutant. Without the galactose added by Lic2A, both mouse and human IgM bound similarly and exhibited no sialic acid-mediated inhibition. Interestingly, a degree of differential binding to sialylated and nonsialylated NTHi is retained in mouse WT^SA− serum^abs, unlike human serum, which could be attributed to a higher abundance of antibodies against the sialic acid-protected epitopes in mouse serum. It is difficult to speculate on the quantitative difference in IgM between species, but we expect the mouse antibody repertoire to be different overall. However, most notably, these results indicate that IgM antibodies recognized the same structures between species.

FIG 3.

Human and mouse serum share similar antibody specificities for LOS epitopes on NTHi 375. IgM binding to 375 wild type (WT) versus lgtC′ (A and B), and WT versus lgtC′ lic2A^OFF, a lic2A phase-off variant in the lgtC′ mutant background (C and D), all grown on MIc^SA±, were analyzed by ELISA (absorbance read at 620 nm) after incubation with human (A and C) and mouse (B and D) serum preabsorbed with 375 WT and the Δlic2A mutant grown on MIc^SA− (serum^abs for all at a final dilution of 1:25). lic2A is phase-on for the WT and lgtC′ strains and phase-off in the lgtC′ lic2A^OFF strain. IgM binding is shown as a percentage of the maximum normalized absorbance values, with the means and standard errors of the means (SEM) of triplicate samples graphed. Statistical significance was evaluated by one-way ANOVA with Bonferroni’s multiple-comparison test (****, P < 0.0001; **, P < 0.01; *, P < 0.05; ns, not statistically significant).

Sialic acid enhances H. influenzae survival in the murine lung independent of IgM induced by prior infection with NTHi.

Throughout life and in disease states, humans are sequentially colonized by different NTHi strains, infections that are in part facilitated by immune evasion through antigenic diversity of NTHi outer membrane proteins (16). The LOS, in contrast, contains conserved epitopes that may provide targets of a protective response. Because mice are not naturally infected with H. influenzae, and to more closely reflect the immune response in humans, we employed a two-part inoculation strategy to evaluate how prior infection affects subsequent bacterial colonization. During infection with wild-type bacteria, H. influenzae may express heterogenous sialylated and nonsialylated LOS structures, and the nonsialylated structures could elicit IgM antibodies that enhance subsequent clearance of NTHi with defects in LOS sialylation. Therefore, we anticipated that prior infection might enhance the survival advantage that wild-type bacteria have in evading IgM directed at LOS structures compared to the siaB′ mutant. For prior infection we used a heterologous strain, NT127lic2A^C_on, containing a “phase-locked” lic2A gene to prevent phase variation of the sialic acid-protected structure (see Materials and Methods). NT127 differs to an extent in LOS gene composition and structure from NTHi 375. Specifically, NT127 contains the lex2A gene, which is involved in extending another LOS chain from the first heptose which may contain epitopes resembling those added by lic2A and lgtC (17). Because of these structural differences between strains, we confirmed that similar protection by sialylation occurs in NT127 strains displaying alternative LOS chains. To determine if the lex2A-dependent chain affects sialic acid-mediated protection of epitopes added by lic2A and lgtC on the neighboring heptose, we measured IgM binding to NT127lic2A^C_on variants containing either active (phase-on) or inactive (phase-off) lex2A genes. We found that the two variants displayed similar IgM binding, indicating that lex2A does not influence sialic acid-mediated protection (Fig. S1A). Serum absorption experiments were used to confirm that the sialic acid-protected epitope was still present in the NT127lic2A^C_on lex2A phase-off mutant (Fig. S1B), and this strain was used to preclude any unanticipated effects of lex2A on the display of sialic acid-protected epitopes during infection. This strain was inoculated into mice 14 days prior to challenge with WT and siaB′ strains to allow development of IgM responses and concurrent clearance of the primary infection.

To verify IgM induction, on day 11 after the primary infection (“immunization”) with NT127lic2A^C_on we measured the amount of IgM bound to the surface of strains that would be used on day 14 for the secondary infection, WT and siaB′ mutants in the Rd background, a serotype d-derived strain that is heterologous to NT127. Significantly more IgM was detected on the surface of both strains incubated with serum from immune mice than naive mice (Fig. 4), indicating that cross-reactive IgM had been induced. LOS sialylation in Rd WT notably reduced overall binding of IgM relative to siaB′ mutants in naive as well as immune serum. However, the difference in binding to the siaB′ mutant relative to WT was not markedly enhanced after immunization. Therefore, immunization induced a murine antibody response to antigens distinct from the LOS epitopes on NTHi that are protected by sialic acid.

FIG 4.

Immunization induces cross-reactive IgM to H. influenzae. Blood was collected from naive (n = 3) and immune (n = 3) mice on day 11 after immunization. Serum was obtained and heat inactivated before incubating with wild-type (WT) H. influenzae Rd and the siaB′ mutant grown on sBHI to measure IgM binding by flow cytometry. MFI values were log transformed [Y = log(y)] and normalized to get the percentage of the maximum, and means and SEM are shown. Statistical significance was evaluated by one-way ANOVA with Bonferroni’s multiple-comparison test (****, P < 0.0001; ***, P < 0.001; **, P < 0.01).

On day 14 after the initial immunization, naive and immune mice were challenged with either the Rd WT or an isogenic siaB′ mutant strain mixed at a 1:1 ratio with the RdLacZ competitor strain. Mice were euthanized at 20 h postinfection, and lungs were collected to quantitate bacterial burden. Lung homogenates were plated to enumerate CFU of each experimental (LacZ ⁻) strain, Rd WT and siaB′, in comparison to the reference competitor (LacZ⁺) strain, RdLacZ. These ratios are reported as competitive indices for either strain in each immunization condition (naive or immune) (Fig. 5A). The siaB′ mutants were significantly attenuated relative to WT in both naive and immunized mice, and the extent of attenuation of the siaB′ mutants was not increased by prior infection with NTHi. To further analyze the role of sialylation in vivo, we examined survival of the siaB′ mutant in mice depleted of complement versus in untreated mice. Cobra venom factor (CVF), a structural and functional analog of complement component C3, continuously activates the complement system and ultimately depletes serum complement activity. Bacterial burden in competitive infection with the RdsiaB′ mutant versus RdLacZ was evaluated in CVF-treated versus phosphate-buffered saline (PBS)-treated mice (Fig. 5B). In the mouse lungs, survival of the siaB′ mutant was similar with and without CVF treatment. In contrast, the competitive index of the siaB′ mutant was significantly lower in the nasopharynx of mice treated with PBS than mice treated with CVF, indicating that the ability of the siaB′ mutant to colonize the nasopharynx is significantly reduced in the presence of active complement. Consistent with the in vivo role of siaB in protection against killing by antibody-directed complement in the murine model, the siaB′ mutant is more sensitive than WT to human complement in the presence of heat-inactivated mouse serum as an antibody source but not with Ig-depleted human serum alone (Fig. 6).

FIG 5.

Sialic acid enhances in vivo survival of H. influenzae. (A) Mice received intranasal inoculations of Rd wild type (WT) or siaB′ mutants, each mixed with equal numbers of CFU of a SiaB⁺ competitor strain, RdLacZ (LacZ⁺), on day 14 postimmunization. Lung CFU were enumerated 20 h postinfection, and the ratio of the indicated experimental (LacZ⁻) strains to the reference strain (LacZ⁺) was calculated. The ratios, or competitive indices, are shown. Lines indicate geometric means. Statistical significance was evaluated by two-way ANOVA with Bonferroni’s multiple-comparison test (****, P < 0.0001). (B) Mice were treated with cobra venom factor (CVF) or PBS and inoculated intranasally with an equal mixture of RdsiaB′ and RdLacZ. Lungs and nasal septa were collected 22 h postinfection and plated for CFU determination. Shown are competitive index ratios at each site, calculated as for panel A. Lines indicate geometric means. Statistical significance was evaluated via unpaired, one-tailed t test (*, P < 0.05; ns, not statistically significant).

FIG 6.

Deletion of siaB increases sensitivity of H. influenzae to mouse serum antibody-directed killing by complement. Survival of Rd wild type (WT) and a siaB mutant grown on MIc^SA following incubation with or without 10% heat-inactivated mouse serum (MS^Δi) as a source of antibody and 3% IgG/IgM antibody-depleted human complement active serum (HC) as a source of complement at 37°C for 30 min. Heat-inactivated HC (HC^Δi) was used as a control for complement-mediated killing. Percent survival is the ratio of the number of CFU recovered at 30 min relative to the input CFU at 0 min of incubation. Limit of detection was 0.2%. Log-transformed survival ratios were evaluated by one-way ANOVA with Bonferroni’s multiple-comparison test (***, P < 0.001; **, P < 0.01; ns, not statistically significant). The means for duplicate samples are shown.

These results indicate that LOS sialylation mediated by siaB is critical to the survival of NTHi in the lungs and nasopharynx. The results in CVF-treated mice suggest that complement is more important in nasopharyngeal colonization than in the lungs and that a host immune mechanism other than complement is likely responsible for the siaB-mediated differences in the lungs, such as responses of Fc-receptor-bearing lymphocytes to IgM-opsonized bacteria. Surprisingly, while prior infection with NTHi did elicit increased IgM against the bacteria, it did not lead to increased levels of IgM specific to sialic acid-protected epitopes, nor did it increase clearance of the nonsialylated siaB′ mutant relative to WT. It appears that during infection with wild-type NTHi, sialylation effectively prevented a protective IgM response to specific LOS structures, while nonprotective IgM responses against other structures increased. These results suggest that LOS sialylation not only blocks binding by bactericidal natural IgM but also prevents induction of protective IgM specificities during infection. This evasion mechanism may contribute to reinfection by different NTHi strains despite expression of common underlying LOS structures.

DISCUSSION

Murine models are commonly used to study NTHi lung infections and have also been used to investigate whether sialic acid plays a role in reduced bacterial clearance. However, the mechanism for enhanced survival of H. influenzae mediated by sialic acid in the lung has not been established. Swords et al. found that a siaB-deficient NTHi mutant was attenuated for colonization and persistence in a rat pulmonary model and suggested that this attenuation was at least partially due to the role sialic acid plays in promoting biofilm formation, based on reduced biofilm formation of the siaB mutant in vitro (18). Complement is important in defense against H. influenzae infections, and extensive studies, including our own, have demonstrated the ability of sialic acid to inhibit various aspects of the complement system. Specifically, sialic acid can inhibit activation of the classical pathway and binding of IgM to NTHi (6, 7). Moreover, we have established that sialic acid mediates this protection by inhibiting bactericidal IgM from binding to specific LOS epitopes on NTHi while also conferring protection from C3 deposition by the alternative pathway via mechanisms that remain to be determined (7). Other studies support an in vivo role for sialic acid in blocking complement, and our finding of IgM blocking in human and murine serum is consistent with that mechanism (20, 45). We have demonstrated that complement is likely involved in the mechanism of protection utilized by sialylated H. influenzae in vivo depending on the site of colonization. Although complement is known to be present in the lung as well as the nasopharynx, the role that sialic acid plays in protecting NTHi from host immune defenses in these tissues appears to differ. In the nasopharynx, the SiaB-mediated ability to sialylate the LOS was required for H. influenzae to evade complement-mediated clearance; however, in the lungs, the role of SiaB was complement independent, and it may aid in evasion of other effector mechanisms, such as recognition of IgM-opsonized bacteria by Fc receptor-bearing lymphocytes at this site (9).

Intriguingly, we identified IgM antibodies in naive mice specific for H. influenzae LOS epitopes added by Lic2A and LgtC, even though mice are not naturally infected with this bacterial species. Mice share these serum antibody specificities, which exhibit bactericidal activity against NTHi in vitro, with humans. Our data indicate that at least for several NTHi strains, most of the sialic acid-protected IgM epitopes are on the lic2A-dependent chains, but we cannot rule out lesser contributions from adjacent LOS chains with different sugar moieties that also act as acceptor sites for Neu5Ac and exhibit sialic acid-mediated protection from IgM binding. Sialylation may also potentially inhibit antibodies from binding to epitopes on adjacent LOS chains via steric hindrance. A well-developed repertoire of anti-carbohydrate/glycan antibodies is needed to initiate successful, immediate immune responses against invading pathogens, so it is feasible that more than a few of these antibody specificities are conserved among host species (13, 21). Although several different pathogens, including bacteria, viruses, and fungi (8, 22–25), can be efficiently targeted by these antibodies, the issue remains that NTHi and other bacteria can utilize host sialic acid to escape recognition by effectors of the immune response.

We investigated whether prior infection with NTHi would enhance the role of sialic acid, potentially by better reflecting prior exposure that occurs in humans. We evaluated a short-term immunization of 14 days to favor responses by IgM capable of recognizing conserved LOS structures of antigenically distinct strains. Our strategy used heterologous immunizing and infecting strains to better replicate human infections because individuals are typically colonized by different strains over time (26–29). We theorized that prior infection may increase the IgM response to sialic acid-protected epitopes. Alternatively, NTHi may successfully evade responses induced against the epitopes that are protected in the presence of sialic acid. Ultimately, we observed greater IgM responses to NTHi in immunized mice; however, the response to sialic acid-protected epitopes was not increased relative to the overall response. Moreover, this immunization did not result in increased clearance of the sialic acid-deficient mutant from the lungs. Thus, immunization did induce IgM production, but sialic acid present in vivo appears to have efficiently protected the NT127 immunizing strain so that the antibody response was generated against other structures. In addition, the sialic acid-protected epitopes may be primarily recognized by innate-like B-1 lymphocytes that typically do not expand in response to antigen, though an increase in IgM production is still expected when these cells encounter antigenic targets (30, 31). Since the LOS-specific mouse IgM that we identified appears to be present naturally (that is, prior to H. influenzae exposure), it is likely already present at levels sufficient to kill bacteria that are not protected by sialic acid.

Natural IgM has been implicated in protection against various bacterial, viral, and fungal pathogens (21, 32). Unfortunately for the host, NTHi has acquired mechanisms of protection from this IgM, mechanisms which we show to include LOS sialylation. We observed an antibody response to exposed NTHi epitopes induced by prior infection (Fig. 4); however, this induced response was actually not protective against infection, as it did not enhance clearance (Fig. 5). This could be considered an example of a pathogenic survival strategy in which humoral immunity is diverted to decoy epitopes and away from the targets of protective antibodies, a mechanism that is employed by various bacterial and viral pathogens (33–36). Several studies on the outer membrane proteins (OMPs) of Haemophilus influenzae, including OMP P1, P2, and P5, have found that these surface structures contain several immunodominant epitopes that elicit strong antibody responses but that not all of the antibodies produced are protective against NTHi (37–39). Thus, an epitope on the bacterial cell surface that is a major antibody target, or immunodominant, may polarize the immune response by acting as a decoy inducing nonprotective antibodies, thereby protecting a less immunodominant but more functionally important epitope, which ultimately impairs the ability of the host immune system to successfully clear the pathogen (40). Host-microbe interactions are important in driving the evolution of evasive mechanisms that aid in bacterial survival and pathogenesis (41–43). For pathogens like NTHi to be efficient, long-term colonizers, they must have an arsenal of defenses against the host and this decoy technique acts to further protect LOS epitopes from immune recognition.

This study provides valuable insight into sialic acid-mediated immune evasion in NTHi, and the anti-H. influenzae activity observed with murine IgM could help us better understand the mechanisms of antibody-mediated protection against these infections in humans. Identification of common antibody specificities against sialic acid-protected epitopes will allow us to evaluate NTHi’s interactions with these responses in mice in vivo to model this aspect of immune evasion and to increase our understanding of this model’s utility for the development of potential therapeutic antibodies.

MATERIALS AND METHODS

Strains and culture conditions.

H. influenzae strain RdAW (referred to here as Rd) (44) and NTHi clinical isolates NT127 (19) and NTHi 375 (45) were grown at 35°C ± 1.5°C on agar plates containing brain heart infusion supplemented with 10 μg/ml NAD and 10 μg/ml hemin (sBHI) or a chemically defined medium (MIc) (46). MIc agar contained either no sialic acid or sialic acid (Neu5Ac) at 25 μg/ml (MIc^SA±). Kanamycin (Km), tetracycline (Tet), 3,4-cyclohexenoesculetin-β-d-galactopyranoside (S-Gal), and d-xylose were added to sBHI at 20 μg/ml, 3 μg/ml, 300 μg/liter, and 1 mM, respectively. S-Gal is a chromogenic substrate for β-galactosidase for detection of bacterial colonies with the Lac⁺ phenotype; d-xylose is used for induction of the lacZ gene, encoding β-galactosidase under the control of the xylA promoter at the xyl locus (47). Deletions of siaB in NT127 and lic2A in NTHi 375 and Rd were created as previously described (48), and we thank Derek Hood for the Hi375siaB′ strain (45). To make the RdsiaB′ strains, we PCR amplified the entire siaB gene disrupted by a kanamycin resistance cassette from Hi375siaB′. DNA was transformed into naturally competent H. influenzae strain Rd as previously described (49), and transformants were selected on sBHI agar containing Km. The NTHi 375 lgtC′ mutant contains a lgtC gene disrupted by a kanamycin resistance cassette (50). Several LOS genes in H. influenzae strains are subject to phase variation, the random and reversible switching between the active, in-frame (phase-on) and inactive, frame-shifted (phase-off) state that is mediated by gain or loss of tandem DNA repeats in the coding region of each gene by slipped-strand base mispairing. Relevant genes, including lic2A, lgtC, and lex2A, were analyzed for their phase-on/off status via PCR with gene-specific primers encompassing the tandem repeat regions followed by Sanger sequencing (Eurofin Genomics). The NT127lic2A^C on strain was generated by transformation of NT127 with plasmid pLic2Aon (48), delivering a lic2A gene that is phase-on locked (no phase variable tandem repeats within the coding region) into the xyl locus of NT127. The phase-on/off status of lic2A, lgtC, and lex2A in two independent isolates of NT127lic2A^C_on was confirmed via sequencing. Isolates differed only in the phase status of lex2A and were indistinguishable phenotypically with regard to IgM binding and serum resistance (see Results; also, see Fig. S1 in the supplemental material). The isolate with lex2A phase off was chosen for the murine immunization experiment.

Absorption of sera.

Absorption of heat-inactivated (35 min at 56°C) pooled C57BL/6 mouse serum (Bioreclamation Inc.) was performed as previously described (7) using NTHi 375 wild-type and Δlic2A strains grown on MIc agar containing no sialic acid (MIc^SA−). Preabsorbed serum (serum^abs) is referred to as either Δlic2A^SA− or WT^SA− serum^abs.

Flow cytometry.

NTHi strains were grown on MIc^SA±, unless noted otherwise in the figure legends, and diluted in Hanks balanced saline solution with calcium and magnesium chloride (HBSS⁺⁺) to an optical density at 600 nm (OD₆₀₀) of ∼0.3. Mouse serum heat inactivated for 35 min at 56°C was added to bacteria at a final concentration of 20% and incubated for 30 min at 37°C. Fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgM (Southern Biotech) was used to measure binding of serum IgM. Data were acquired on an ACEA Biosciences NovoCyte flow cytometer and analyzed using ACEA NovoExpress software. All experiments included an antibody-only background control for each strain. Median fluorescence intensity (MFI) values were normalized to the maximum MFI so that all IgM binding is depicted as a percentage of the maximum.

Serum bactericidal assays.

Bactericidal assays using absorbed or unabsorbed sera were performed as described previously (6). Briefly, bacteria grown overnight on MIc^SA± were harvested and resuspended in HBSS⁺⁺–0.1% bovine serum albumin (BSA). Bacteria were diluted to ∼2,000 CFU and incubated with the indicated preabsorbed or unabsorbed C57BL/6 mouse serum (10% final concentration) and IgG/IgM antibody-depleted human complement (HC) serum (3% final concentration) (Pel-Freez Biologicals) as a source of active complement. Reaction mixtures were plated at timer zero and after 30 min incubation at 37°C to determine percent survival. Heat-inactivated HC was used as a control to establish that killing depends on the presence of active complement.

ELISAs.

Enzyme-linked immunosorbent assays (ELISAs) were used to measure mouse and human serum IgM against NTHi strains in preabsorbed sera. Bacteria grown on sBHI or MIc^SA± were diluted in HBSS⁺⁺, and wells were coated with ∼10⁷ bacteria/100 μl and incubated at 4°C overnight. Unbound antigen was removed by washing wells before blocking with 200 μl of Tris-buffered saline (TBS) with 1% BSA at 4°C for 1h. Plates were washed 3 times with TBS–0.025% Tween 20, incubated with 50 μl of heat-inactivated serum samples as indicated in the figures (diluted 1:25) at room temperature for 1 h, and then washed again 5 times before addition of alkaline phosphatase (AP)-conjugated goat anti-mouse IgM (1:2,000) (Southern Biotech) or horseradish peroxidase (HRP)-conjugated goat anti-human IgM (1:2,000) (Southern Biotech) for 30 min at room temperature. After 5 washes to remove excess detection antibody, 100 μl of BluePhos microwell phosphatase substrate system (KPL) or SureBlue Reserve 3,3',5,5'-tetramethylbenzidine (TMB) microwell substrate for HRP (KPL) was added to each well. Plates were read at a wavelength of 620 nm. Wells with secondary antibody alone were included as controls. Absorbance measurements were normalized to the maximum reading, with IgM binding depicted as a percentage of the maximum.

Immunization and murine lung infection.

H. influenzae organisms grown to mid-log phase (OD₆₀₀, 0.3 to 0.5) in 5 ml sBHI were pelleted, resuspended at appropriate concentrations in HBSS⁺⁺, and inoculated (40 μl) into the nares of 6- to 15-week-old C57BL/6 mice (Charles River Laboratories, Wilmington, MA) anesthetized with ketamine (50 mg/kg) and xylazine (5 mg/kg) by intraperitoneal (i.p.) injection. “Immunization” of mice was by prior infection, as described above, which was cleared within 72 h postinoculation. Mice were immunized with 10⁸ CFU of NTHi strain NT127lic2A^C_on or HBSS⁺⁺ buffer alone (mock) on day 0. Clearance of this strain was verified in a representative group of mice on day 11. On day 14, 10⁸ CFU of H. influenzae wild-type Rd and RdsiaB′ were inoculated with an equal number of CFU of an H. influenzae LacZ-expressing competitor strain, RdLacZ (48), into the nares of mice. Blood was collected from the heart to obtain serum. Lungs were harvested and homogenized in 2 ml of BHI at 20 h after bacterial inoculation and plated onto sBHI agar with S-Gal and d-xylose and grown for CFU determination. Ratios of CFU of the experimental strain (white colonies, LacZ⁻) to the competitor strain (black colonies, LacZ⁺) are reported as the competitive index. For complement depletion, mice were treated i.p. with 20 μg of cobra venom factor (CVF) from Complement Technology Inc. (Tyler, TX) in 100 μl of PBS or PBS alone (control) for 24 h followed by intranasal inoculation with 1 × 10⁸ CFU of both RdsiaB′ and RdLacZ mixed at a ratio of 1:1 in 40 μl. Lungs and nasal septa were collected 22 h postinoculation, homogenized, and plated for enumeration of CFU. Competitive index ratios were calculated as described above to compare survival of the siaB′ mutant in CVF-treated and PBS-treated mice. Experiments were conducted with approval and in accordance with guidelines of the Institutional Animal Care and Use Committee at the University of Mississippi Medical Center.

Statistical analyses.

Statistical significance was determined by one-way or two-way analysis of variance (ANOVA) with Bonferroni’s multiple-comparison test using GraphPad Prism software (GraphPad Software, San Diego, CA) as indicated in the figure legends.