Novel mechanism for the generation of human xeno-autoantibodies against the nonhuman sialic acid N-glycolylneuraminic acid

Sera from infants age 0–12 mo (cord, 3 mo, 6 mo, and 12 mo) were analyzed by ELISA for the presence of anti-Neu5Gc IgM and IgG antibodies against Neu5Gcα2–6Galβ1–4Glc (Gcα2–6Lac), an epitope against which most human adults possess high levels of IgM and IgG antibodies (Padler-Karavani et al., 2008). All sera were from infants who had been exclusively breastfed for the first 3 mo of life, and then switched to cow’s milk–based formula. Solid foods were also introduced starting at 3 mo of age and included both foods lacking Neu5Gc, such as fruits and vegetables, and Neu5Gc-rich foods, such as beef, pork, and lamb. Anti-Neu5Gc IgM antibodies were absent at birth (cord serum) and at 3 mo, appeared at 6 mo and achieved almost adult levels at 12 mo (Fig. 1 A). As expected because of transplacental transport of IgG, all cord sera contained anti-Neu5Gc IgG antibodies, at levels similar to maternal anti-Neu5Gc IgG. These anti-Neu5Gc IgG antibodies diminished at 3 mo, followed by increasing levels at 6 and 12 mo (Fig. 1 B). There was no difference between male and female anti-Neu5Gc IgM and IgG titers (unpublished data). The reactivity of 12 mo sera against Gcα2–6Lac was significantly reduced after truncation of the target epitope’s Neu5Gc side chain by mild periodate oxidation (Tangvoranuntakul et al., 2003), further demonstrating the specificity of the IgG antibodies for Neu5Gc-containing glycans (Fig. 1 C). The absence of anti-Neu5Gc IgM antibodies in cord sera suggests that anti-Neu5Gc antibodies are not germ-line encoded “natural antibodies” (Ochsenbein and Zinkernagel, 2000), but instead require a postnatal antigenic stimulus. And the early appearance and class switching of these antibodies indicate that humans are exposed to the Neu5Gc antigenic stimulus early in life. The nadir in anti-Neu5Gc IgG titer seen at 3 mo is also consistent with the half-life of maternally derived IgG (Morell et al., 1970), and suggests that these infants lack the production of endogenous anti-Neu5Gc IgG antibodies at 3 mo, when their diets were devoid of Neu5Gc. Interestingly, both infant IgM and IgG anti-Neu5Gc antibodies arise soon after the introduction of Neu5Gc in the diet in the form of cow’s milk formula and baby foods containing red meat.

The temporal correlation between the appearance of anti-Neu5Gc antibodies and the introduction of animal-derived foods suggested that dietary Neu5Gc might represent the antigenic stimulus. To study this issue experimentally, we used Cmah-null mice that have a human-like deficiency in Neu5Gc synthesis (Hedlund et al., 2007). However, despite consumption of ∼1 mg/kg/day of Neu5Gc in their regular chow (unpublished data; this represents approximately five times the amount present in the typical human Western diet; Tangvoranuntakul et al., 2003), Cmah-null mice did not spontaneously generate anti-Neu5Gc IgM or IgG antibodies after 6–9 mo (Fig. 2, A and B). In fact, as summarized in Table I, all attempts to induce anti-Neu5Gc antibodies in Cmah-null mice via dietary Neu5Gc exposure have been unsuccessful to date. These include: providing free Neu5Gc in the drinking water at 1 mg/ml for 12 wk; cross-fostering Cmah-null pups with WT dams whose milk contains Neu5Gc (simulating cow’s milk-based formula feeding in human infants); feeding with high-Neu5Gc content mucins; feeding with Neu5Gc-rich goat cheese; and oral gavage with cow’s milk (unpublished data; see Table I for details). These negative results are also interesting because Cmah-null mice actually have relatively hyper-reactive B cells, apparently caused by loss of Neu5Gc-containing Siglec ligands (Naito et al., 2007; Cariappa et al., 2009).

In contrast, deliberate immunization with an artificial immunogen rich in Neu5Gc, such as chimpanzee RBCs, and complete Freund’s adjuvant, did elicit anti-Neu5Gc IgM and IgG antibodies in Cmah-null, but not WT, mice (Hedlund et al., 2008). Thus, despite the lack of spontaneously generated anti-Neu5Gc antibodies, Cmah-null mice are capable of generating a humoral immune response against Neu5Gc upon active immunization (Fig. 2, A and B). Recently, another group also reported that a different strain of Cmah-null mice do not spontaneously express anti-Neu5Gc antibodies, and require artificial immunization to generate a humoral immune response (Tahara et al., 2010).

Some other postnatally appearing anticarbohydrate antibodies in humans, such as anti–α-Gal and antiblood group antibodies, have been shown to be elicited in response to resident gut bacteria expressing these epitopes (Springer and Horton, 1969; Galili et al., 1988; Posekany et al., 2002). Interestingly, we found that the appearance of anti-Neu5Gc antibodies in human infant sera coincides with the appearance of both anti–α-Gal (Fig. 1 D) and anti-blood group antibodies (not depicted), suggesting that anti-Neu5Gc antibodies may also be elicited by colonizing bacteria during infancy. In this regard, many human microbial pathogens/commensals express Sias on their surface glycans, creating a resemblance to “self” glycoproteins and glycolipids found in mammals (Vimr and Lichtensteiger, 2002; Vimr et al., 2004; Lewis et al., 2009). This molecular mimicry of host glycans can help pathogens evade, dampen, and/or inhibit host immune defenses (Estabrook et al., 1997; Ram et al., 1998; Carlin et al., 2009). Several bacterial species, including Escherichia coli K1, Campylobacter jejuni, most meningococcal strains, and Streptococcus agalactiae are able to synthesize Sia de novo (Vimr and Lichtensteiger, 2002). However, the parent Sia structure is always Neu5Ac, and no microbe has ever been shown to synthesize Neu5Gc. Indeed, the biosynthesis of the N-glycolyl group of Neu5Gc appears to be a singular event in evolution, being confined to the Deuterostome lineage of animals (Schauer, 1982).

On the other hand, certain sialylated bacteria including Neisseria gonorrhoeae, Haemophilus influenzae, Haemophilus ducreyi, and Corynebacterium diphtheriae lack the biosynthetic machinery necessary for Sia synthesis and instead scavenge host-derived Sias (Mandrell et al., 1990; Mattos-Guaraldi et al., 1998; Schilling et al., 2001; Bouchet et al., 2003). Non-typeable H. influenzae (NTHi) colonizes most humans and can transition from a commensal to a pathogenic state in diseases such as infantile otitis media. NTHi can efficiently scavenge minute amounts of environmental free Sias via a specific transporter (Allen et al., 2005; Severi et al., 2005), and then use an endogenous CMP-Sia synthetase (Hood et al., 1999) and sialyltransferases (Jones et al., 2002) to decorate its LOS. These sialylated LOS molecules are critical for NTHi human serum resistance (Hood et al., 1999; Allen et al., 2005) and virulence in a chinchilla otitis media model (Bouchet et al., 2003; Jurcisek et al., 2005).

Of the aforementioned bacteria, only H. influenzae has the potential to access dietary Neu5Gc while living as part of the human flora. We hypothesized that NTHi might scavenge exogenous dietary Neu5Gc and express it as an immunogenic epitope, providing a mechanism for the generation of anti-Neu5Gc antibodies in human infants. To pursue this hypothesis, we grew the NTHi strain 2019 (Campagnari et al., 1987) in a Sia-free defined medium and analyzed the bacteria by flow cytometry using a Neu5Gc-specific chicken IgY antibody (Diaz et al., 2009). When 2019 was grown in Sia-free defined media, the anti-Neu5Gc antibody background staining was similar to the control IgY, confirming the inability of NTHi to synthesize endogenous Neu5Gc (Fig. 3 A). However, when exogenous free Neu5Gc (0.1 µM–1 mM) was added to the defined media, there was a dose-dependent increase in anti-Neu5Gc binding, again, compared with no binding by the control IgY antibody (Fig. 3 A). Similar results were found (unpublished data) using two other NTHi strains, Int1 (Nizet et al., 1996) and DH1 (Houliston et al., 2007), as well as one encapsulated type b strain, Eagan (Anderson et al., 1972). Thus, uptake and surface expression Neu5Gc is a common feature among different strains of H. influenzae.

To further confirm Neu5Gc incorporation by NTHi, 2019 was grown in 1 mM Neu5Gc, treated with sialidase or heat-inactivated sialidase, and then probed in a whole-cell ELISA with the chicken anti-Neu5Gc antibody. Sialidase treatment decreased anti-Neu5Gc staining when compared with treatment with heat-inactivated sialidase, demonstrating release of Neu5Gc from NTHi surface glycans by active sialidase (Fig. 3 B). Finally, no Neu5Gc expression was seen in NTHi strain 2019siaT (Allen et al., 2005), an isogenic mutant that lacks a critical Sia transporter (see Fig. 6 A, top left). Collectively, our data indicate that dietary Neu5Gc may be efficiently taken up and then expressed on cell surface molecules of NTHi.

Most humans are colonized by NTHi (Turk, 1984). We next asked whether there was a temporal correlation between the timing of initial NTHi colonization or infection and the appearance of anti-Neu5Gc antibodies. As neither nasopharyngeal nor middle ear cultures were available from the infants in our cohort, we used a whole-cell ELISA to screen for antibodies against NTHi as an indicator of colonization or infection. For these studies, NTHi was grown in Sia-free defined media to ensure absence of Neu5Gc on the LOS of the bacteria. In all infants analyzed, we could detect IgM antibodies against NTHi that increased significantly between birth, 3, 6, and 12 mo (Fig. 4). Because it is unlikely that all infants in our cohort have had NHTi otitis media by 3 mo of age, it is likely that these increased titers of antibody against whole-cell NTHi are a response to NTHi colonization. These results are consistent with those of others who have shown that most infants acquire NTHi within the first year of life (Vives et al., 1997), and that infants generate antibodies against outer membrane proteins on NTHi during the course of asymptomatic colonization (Faden, 2001). Thus, infants in our cohort generated an adaptive immune response against colonizing NTHi at about the same time that anti-Neu5Gc antibodies appear.

Given the temporal relationship between the appearance of anti-Neu5Gc antibodies and colonization with NTHi, we next asked whether Neu5Gc-expressing NTHi could elicit an anti-Neu5Gc antibody response in Neu5Gc-deficient mice, which did not spontaneously express anti-Neu5Gc antibodies (Table I). To this end, Cmah-null mice were injected intraperitoneally with heat-killed NTHi, which had been grown in Sia-free media with Neu5Gc (generating Neu5Gc expressing NTHi, Gc-NTHi) or without Neu5Gc (generating Sia-free NTHi, Sia-free NTHi). To assess the inherent immunogenicity of the Gc-NTHi, no adjuvant was used to enhance antibody responses. Sia-free NTHi represents an ideal negative control because, except for the absence of Neu5Gc, it is identical to Gc-NTHi. Indeed, only Cmah-null mice injected with Gc-NTHi generated anti-Neu5Gc IgM antibodies and underwent class switching to generate IgG antibodies after 2–3 injections (Fig. 5, A and B). In contrast, Cmah-null mice injected with Sia-free NTHi did not generate anti-Neu5Gc antibodies (Fig. 5, A and B). Multiple attempts to elicit anti-Neu5Gc antibodies in Cmah-null mice via intranasal administration of Neu5Gc-expressing NTHi have been unsuccessful (unpublished data). This is not surprising, given that mice show very rapid mucosal clearance of this human-specific microorganism and limited nasopharyngeal colonization (Zola et al., 2009). Furthermore, human infants are prone to major upper respiratory infections with NTHi, and the immune response we are studying would likely be enhanced by such inflammation in infants. The anti-Neu5Gc IgG antibodies generated in Cmah-null mice were also tested for reactivity against different Neu5Gc containing antigens and found to be of similar titer and specificity to those we have found in human infants (Fig. 5 C) and adults (Padler-Karavani et al., 2008).

NTHi that are colonizing the infant’s oropharynx and even upper airways can have direct access to dietary Neu5Gc, since reflux of ingested liquids into the infant upper respiratory tract is commonly observed (Cober and Johnson, 2005). To determine if NTHi can take up and express Neu5Gc from baby food, NTHi was grown in Sia-free media with or without various commercially available semi-solid baby foods and analyzed by flow cytometry for cell-surface Neu5Gc. As shown in Fig. 6 A, WT NTHi strain 2019 (WT) was found to express Neu5Gc when grown in the presence of baby foods consisting of red meat (beef, pork, and lamb) but not in poultry (chicken and turkey; unpublished data), vegetables or fruits. In contrast, there was no Neu5Gc expression seen in the siaT mutant of 2019, confirming that uptake through the Sia transporter is required for expression of dietary Neu5G by NTHi. Furthermore, an even greater shift in anti-Neu5Gc staining (an approximately fourfold increase in MFI compared with WT) was seen with a NTHi sialic acid lyase-deficient mutant (2019nanA; unpublished data), which develops a hypersialylated phenotype in the presence of exogenous Sia (Allen et al., 2005). The finding that Neu5Gc expression by NTHi was restricted to uptake from baby foods containing red meat is not surprising, as red meat is known to contain high levels of Neu5Gc, whereas poultry contains only Neu5Ac and plants contain no Sia (Tangvoranuntakul et al., 2003). NTHi do not produce a sialidase (Vimr and Lichtensteiger, 2002), and therefore require free (nonglycosidically bound) Sia for LOS sialylation. Indeed, when Sia levels were quantified in the baby food, considerable levels of free Neu5Gc (∼2 µg/g; Table II) were detected in baby food containing red meat, but not in those containing poultry or plants (Table II). Of course, many other oral and nasopharyngeal commensal bacteria express sialidases (Corfield, 1992), which could further increase the local concentrations of free Neu5Gc for use by NTHi in vivo.

To further corroborate our hypothesis, we asked if anti-Neu5Gc antibodies that were affinity purified from normal human serum (Padler-Karavani et al., 2008) could interact with Neu5Gc-expressing NTHi in a whole-cell ELISA. Indeed, human anti-Neu5Gc antibodies bound specifically to Neu5Gc-expressing NTHi (Gc-NTHi) and not to nonsialylated NTHi (Sia-Free NTHi; Fig. 6 B).

In this study, we propose a model for how NTHi and dietary Neu5Gc cooperate to generate anti-Neu5Gc antibodies in humans. Collectively, our data indicate a mechanism by which humans may generate anti-Neu5Gc antibodies in early life, by simultaneous exposure to Neu5Gc-containing foods and the incorporation and surface expression of the nonhuman Sia by colonizing NTHi. As a Gram-negative bacterium that expresses pathogen-associated molecular patterns such as LPS, this Neu5Gc-expressing organism would simultaneously provide both immunogen and adjuvant. It is also interesting that this carbohydrate antigen elicits class switching to generate sometimes high levels of IgG antibodies. In this regard, direct engagement of LPS with TLR4 on B cells of mice is known to promote proliferation, class switching, and immunoglobulin secretion (Peng, 2005). Furthermore, LPS-induced secretion of BAFF and APRIL by dendritic cells and monocytes could contribute to T cell–independent induction of class switch recombination (Fagarasan and Honjo, 2000; Litinskiy et al., 2002), especially in the setting of inflammation (Ueda et al., 2007) that is commonly associated with NTHi-mediated otitis media. Future studies of mice and humans with various genetically defined immunodeficiencies should help to define the cellular and molecular pathways required for the generation of anti-Neu5Gc antibodies, and address the question of T cell dependence versus independence, and the role of Toll-like receptors. Identifying the specific B cell populations that produce anti-Neu5Gc antibodies and determining if these B cells undergo somatic hypermutation will also help to characterize the anti-Neu5Gc response in humans and mice. An additional contributing factor may be the relative overreactivity of human B cells to stimulation (Soto et al., 2010).

We have shown here that NTHi is capable of scavenging Neu5Gc from the diet and expressing it as an immunogenic epitope, apparently contributing to the generation of anti-Neu5Gc antibodies in humans. Thus, our normal flora can act as “antigen-presenting cells” of bacterial rather than host origin, eliciting humoral immune responses that could contribute to inflammatory or autoimmune pathologies. Although NTHi is the first known commensal shown to express scavenged dietary Neu5Gc, it is possible that other as yet unknown commensal/pathogenic bacteria can do the same. In this regard, the wide variation in adult human anti-Neu5Gc titers and specificity between adult individuals (Padler-Karavani et al., 2008) may reflect multiple routes of xeno-autoimmunization.

Cmah^−/− mice (Hedlund et al., 2007) were backcrossed onto a congenic C57BL/6 background. WT C57BL/6 mice were purchased from Harlan Laboratories. Mice were fed standard chow (PicoLab Rodent Diet 20; LabDiet) and water ad libitum and maintained on a 12-h light/dark cycle. All animal work was performed in accordance with The Association for Assessment and Accreditation of Laboratory Animal Care and under a protocol approved by the Institutional Animal Care and Use Committee of the University of California, San Diego, La Jolla, CA.

Collection of maternal and infant blood samples for immunological studies was approved by the University of La Frontera, Temuco, Chile, Institutional Review Board and approved by the Regional Ethical Committee of the Chilean National Health Service for the Araucania Region in which the samples were collected. Written informed consent for study of infants was obtained from both parents. Serum was collected from 15 Chilean infants at birth (from cord blood), 3, 6, and 12 mo of age. Serum from 9 of the 15 mothers in the study was obtained during the third trimester of pregnancy, with consent.

NTHi strains 2019 (Campagnari et al., 1987) and 2019siaT (Allen et al., 2005) were a generous gift from Michael Apicella, Department of Microbiology, University of Iowa. Sialic acid–free bacterial stocks were prepared by passaging 2019 several times in sialic acid–free media: RPMI 1640 media (Sigma-Aldrich) supplemented with 1 µg/ml protoporphyrin IX (Sigma-Aldrich), 1 µg/ml l-histidine (Sigma-Aldrich), 10 µg/ml β-nicotinamide adenine dinucleotide (Sigma-Aldrich), 0.1 mg/ml hypoxanthine (Sigma-Aldrich), 0.1 mg/ml uracil (Sigma-Aldrich), and 0.8 mM sodium pyruvate (Invitrogen; Greiner et al., 2004; Allen et al., 2005). The absence of sialic acid was confirmed by HPLC and mass spectrometry. Commercial baby foods from Gerber Product Company and Beech-Nut Nutrition Company were purchased at a local grocery store.

NTHi strain 2019 or 2019siaT grown to mid-log (OD₆₀₀ ∼0.3−0.4) in sialic acid–free media, was grown for 2 h with or without various amounts of Neu5Gc (Inalco), washed twice with PBS, and then incubated for 1 h at RT with chicken anti-Neu5Gc IgY (1:2,000; Diaz et al., 2009), control chicken IgY (1:2,000; Jackson ImmunoResearch Laboratories), or PBS alone. After washing once with PBS, bacteria were resuspended in FITC donkey anti–chicken antibody (1:200; Jackson ImmunoResearch Laboratories) for 1 h at room temperature and analyzed by flow cytometry (FACSCalibur; BD). For uptake of Neu5Gc from baby food, each food was diluted with an equal volume of sialic acid–free media and the particulate matter pelleted by centrifugation. After filter sterilization, the filtrate was added to an equal volume of mid-log bacteria in sialic acid–free media, underwent incubated shaking (250 RPM) for 2 h at 37°C, washed twice with PBS, and stained for Neu5Gc by flow cytometry, as described above. Sialidase treatment after Neu5Gc loading of NTHi was performed by resuspending 500 µl (OD₆₀₀ 0.4) in 150 µl PBS, pH 6.0 + 9 mM CaCl₂ with 10 mU active or heat-inactivated (10 min at 100°C) Vibrio cholerae sialidase (Sigma-Aldrich) for 3 h at 37°C. Neu5Gc was detected in a whole-cell ELISA by resuspending the bacteria in Milli-Q water, adding to a 96-well plate (Corning), and evaporating overnight. Wells were washed with Tris-buffered saline, pH 7.4, + 0.1% Tween (TBST), and then incubated with chicken anti-Neu5Gc IgY (1:1,000), isotype control chicken IgY (1:1,000), or TBST alone for 1 h at room temperature, 100 µl/well. Wells were then washed three times with 150 µl TBST, incubated with 100 µl/well alkaline phosphatase (AP) donkey anti–chicken IgY (1:5,000; Jackson ImmunoResearch Laboratories) for 1 h at room temperature, washed as before, and then developed with p-nitrophenyl phosphate, with product measured at 405-nm wavelength on a SpectraMax 250 (MDS Analytical Technologies).

NTHi strain 2019 was grown to mid-log in sialic acid–free media with or without 1 mM Neu5Gc (Inalco), heat-killed, and injected (200 µl of OD₆₀₀ 0.4) intraperitoneally into Cmah-null and WT (C57BL/6) mice (age 5−8 wk, female). All mice were injected a total of three times at 2-wk intervals. No adjuvant was used with any of the bacterial injections. Erythrocyte immunizations were performed as described previously (Hedlund et al., 2008). In brief, Cmah-null mice (age 6−9 wk, male and female) were injected intraperitoneally with 200 µg chimpanzee (Neu5Gc-rich; Yerkes National Primate Research Center, Emory University, Emory, GA) or human (Neu5Gc-free) erythrocyte membrane ghosts in 100 µl PBS with equal volume Freund’s complete adjuvant (Difco), and boosted twice (2 and 8 wk later) with the same amount of immunogen in Freund’s incomplete adjuvant (Difco). Serum for antibody analysis was collected 7 d after the second boost.

Mouse sera were analyzed for anti-Neu5Gc antibodies against Neu5Gcα-PAA (Glycotech) by ELISA, as previously described (Hedlund et al., 2008). In brief, 96-well plates (Costar 9018; Corning) were coated overnight at 4°C with 250 ng/well Neu5Gcα-PAA in 50 mM sodium carbonate-bicarbonate buffer, pH 9.5. Wells were emptied of coating solution and blocked with 200 µl/well TBS + 0.1% Tween, pH 7.4 (TBST) for 2 h at room temperature. Sera were added to the wells for 2 h at room temperature in triplicate, diluted 1:200 in 100 µl TBST. After washing three times with 150 µl TBST, wells were incubated with 100 µl alkaline phosphatase-conjugated goat anti–mouse IgM (Calbiochem) or IgG (Jackson ImmunoResearch Laboratories), diluted 1:5,000 in TBST for 1 h at RT. Wells were washed again, as described, developed with p-nitrophenyl phosphate, and absorbance was measured at a 405-nm wavelength on a SpectraMax 250 (MDS Analytical Technologies). Alternatively, mouse anti-Neu5Gc antibodies against Neu5Gc α2–6 or α2–3 linked to Galβ1–4Glcβ-HSA, or Galβ1–4GlcNAcβ-HSA (synthesized as described previously; Yu et al., 2007) were detected by ELISA, as described above, using pooled sera from mice injected with NTHi strain 2019 grown in 1 mM Neu5Gc, as described above. Antibody levels were quantified after subtracting the readings obtained from coating with HSA alone.

Human anti-Neu5Gc antibodies were detected against Neu5Gcα2–6 or α2–3 linked to Galβ1–4Glcβ-HSA (Yu et al., 2007) and Galα1–3Gal-Polyacrylamide (α-Gal-PAA; Glycotech) by ELISA, as previously described (Padler-Karavani et al., 2008). In brief, 96-well microtiter plates (Costar 9018; Corning) were coated overnight at 4°C, in triplicate, with saturating concentrations of Neu5Gc containing glycoconjugates α2–6 or α2–3 linked to Galβ1–4Glcβ-HSA, HSA (Sigma-Aldrich) alone or with saturating concentrations of α-Gal-PAA or PAA alone, (GlycoTech) in 50 mM sodium carbonate-bicarbonate buffer, pH 9.5. To standardize anti-glycan Ig signals, each plate was also coated with serial dilutions of human IgG and IgM, (Jackson ImmunoResearch Laboratories). Wells were blocked for 2 h at RT with 1% ovalbumin (Grade V; Sigma-Aldrich; free of Neu5Gc) in PBS, followed by incubation with serum samples diluted 1:100 in the same blocking solution for 2 h at RT. The plates were washed three times with PBS containing 0.1% Tween (PBST) and subsequently incubated for 1 h at RT with HRP-conjugated goat anti–human IgM (Kirkegaard and Perry Laboratories), 1:4,000 diluted in PBS, or goat anti–human IgG (Bio-Rad Laboratories) 1:7000–1:6000 diluted in PBS. After washing three times with PBST, wells were developed with an O-phenylenediamine in a citrate-PO₄ buffer, pH 5.5, and absorbance was measured at 490 nm on a SpectraMax 250 (MDS Analytical Technologies). Anti-glycan Ig values were defined by subtracting readings obtained from HSA or PAA alone from those obtained from the Neu5Gc-glycoconjugate-HSA or α-Gal PAA antigens, respectively, and were quantified into nanograms/microliter using the standard dilution curves of the corresponding purified human IgG or IgM.

The Neu5Gc specificity of serum antibodies from 12-mo-old infants was confirmed by mild periodate oxidation followed by borohydride reduction of the Neu5Gcα2–6Galβ1–4Glcβ-HSA glycoconjugate, which selectively truncates the side chain of sialic acids while maintaining their negatively charged carboxyl group (Van Lenten and Ashwell, 1971). Neu5Gcα2–6Galβ1–4Glcβ-HSA glycans were coated onto ELISA plates as described above, except 384-well microtiter plates (MaxiSorp; Nunc) were used to conserve reagents. After overnight coating of Neu5Gcα2–6Galβ1–4Glcβ-HSA or HSA alone, wells were washed twice with 60 µl PBS, pH 6.5. Wells were periodate treated with 80 µl fresh 2 mM sodium metaperiodate (NaIO₄; Thermo Fisher Scientific) in PBS, pH 6.5, at 4°C in the dark for 20 min, and then quenched with 20 µl 100 mM sodium borohydride (NaBH₄; EMD Chemicals; 20 mM final concentration) in PBS, pH 6.5, for 10 min at room temperature in the dark. Mock treatment was performed in parallel by premixing NaIO₄ with NaBH₄, which renders the NaIO₄ inactive. Borates formed during the borohydride treatment were removed by washing wells three times with 60 µl of 50 mM NaOAc/100 mM NaCl, pH 5.5. Wells were then washed twice with 60 µl PBS, pH 7.4. Blocking and detection of human serum IgG antibody binding was performed as described above.

Detection of human infant antibodies binding to NTHi was performed by whole-cell ELISA as described above, except the 96-well plates were coated with saturating amounts of Sia-free NTHi strain 2019, as described above.

Anti-Neu5Gc antibodies were affinity purified from the serum of an adult human (S34) and biotinylated as previously described (Padler-Karavani et al., 2008). Detection of purified human anti-Neu5Gc antibodies binding to Neu5Gc on NTHi was performed by whole-cell ELISA, as described above, using NTHi strain 2019 grown in Sia-free media with or without 1 mM Neu5Gc (Inalco).

For analysis of nonglycosidically bound (free) Neu5Gc in baby food, 50 mg of each baby food was resuspended in 200 µl Milli-Q water (for analysis of total Neu5Gc, samples were heated to 80°C in 2 M acetic acid for 3 h to release sialic acids) and centrifuged at 10,000 g for 10 min at 4°C. The supernatant was then filtered through a 10,000 molecular weight cutoff filter, and sialic acids in the filtrate were derivatized with 1,2-diamino-4,5-methylenedioxybenzene (DMB; Sigma-Aldrich) as described previously (Manzi et al., 1990) and analyzed by reverse-phase HPLC using a C18 column (Phenomenex) at a flow rate of 0.9 ml/min, using 88% water, 5% acetonitrile, and 7% methanol. The excitation and emission were at 373 and 448 nm, respectively. The DMB-derivatized sialic acids were identified and quantified by comparing elution times and peak areas to known standards.

Statistical analysis was performed using Prism v5.0a (GraphPad Software).

We thank Dr. Michael Apicella for generously providing the NHTi 2019 strains, Maria Hedlund and Pam Tangvoranuntakul for early Neu5Gc feeding studies in mice, and Maria J. Martin for guidance during the initial stages of this work. Chimpanzee blood samples were provided by the Yerkes National Primate Research Center.

This work was supported by National Institutes of Health grants R01GM32373 and U01 CA128442 (to A. Varki), ISEF (to V. Padler-Karavani), and R01GM076360 (to X. Chen).

A. Varki is a co-founder of, and shareholder in Sialix, Inc. (formerly Gc-Free, Inc.), a startup biotech company focused on solving problems arising from the pathological consequences of Neu5Gc and Neu5Gc-antibody interactions in humans. D. Ghaderi is currently an employee of Sialix, Inc. The authors have no additional competing financial interests.