Utilizing CMP-Sialic Acid Analogs to Unravel Neisseria gonorrhoeae Lipooligosaccharide-Mediated Complement Resistance and Design Novel Therapeutics

Neisseria gonorrhoeae deploys a novel immune evasion strategy wherein the lacto-N-neotetraose (LNnT) structure of lipooligosaccharide (LOS) is capped by the bacterial sialyltransferase, using host cytidine-5’-monophosphate (CMP)-activated forms of the nine-carbon nonulosonate (NulO) sugar N-acetyl-neuraminic acid (Neu5Ac), a sialic acid (Sia) abundant in humans. This allows evasion of complement-mediated killing by recruiting factor H (FH), an inhibitor of the alternative complement pathway, and by limiting classical pathway activation (“serum-resistance”). We utilized CMP salts of six additional natural or synthetic NulOs, Neu5Gc, Neu5Gc8Me, Neu5Ac9Ac, Neu5Ac9Az, legionaminic acid (Leg5Ac7Ac) and pseudaminic acid (Pse5Ac7Ac), to define structural requirements of Sia-mediated serum-resistance. While all NulOs except Pse5Ac7Ac were incorporated into the LNnT-LOS, only Neu5Gc incorporation yielded high-level serum-resistance and FH binding that was comparable to Neu5Ac, whereas Neu5Ac9Az and Leg5Ac7Ac incorporation left bacteria fully serum-sensitive and did not enhance FH binding. Neu5Ac9Ac and Neu5Gc8Me rendered bacteria resistant only to low serum concentrations. While serum-resistance mediated by Neu5Ac was associated with classical pathway inhibition (decreased IgG binding and C4 deposition), Leg5Ac7Ac and Neu5Ac9Az incorporation did not inhibit the classical pathway. Remarkably, CMP-Neu5Ac9Az and CMP-Leg5Ac7Ac each prevented serum-resistance despite a 100-fold molar excess of CMP-Neu5Ac in growth media. The concomitant presence of Leg5Ac7Ac and Neu5Ac on LOS resulted in uninhibited classical pathway activation. Surprisingly, despite near-maximal FH binding in this instance, the alternative pathway was not regulated and factor Bb remained associated with bacteria. Intravaginal administration of CMP-Leg5Ac7Ac to BALB/c mice infected with gonorrhea (including a multidrug-resistant isolate) reduced clearance times and infection burden. Bacteria recovered from CMP-Leg5Ac7Ac-treated mice were sensitive to human complement ex vivo, simulating in vitro findings. These data reveal critical roles for the Sia exocyclic side-chain in gonococcal serum-resistance. Such CMP-NulO analogs may provide a novel therapeutic strategy against the global threat of multidrug-resistant gonorrhea.


Introduction
Sialic acids (Sias) are a family of 9-carbon sugars (nonoses, or nonulosonates (NulOs)) expressed in the tissues of every vertebrate and some "higher" invertebrates [1]. Sias serve a wide variety of biological roles, including modulation of several aspects of immune function [2]. For example, cell surface-associated Sia regulates the immune system by inhibiting complement activation. Sheep erythrocytes are resistant to lysis by the human alternative pathway because surface Sias increase the affinity of factor H (FH; inhibitor of the alternative complement pathway) for C3b, that is deposited on these or other host cells or microbial surfaces upon activation of complement [3]. Recent work has shown that FH C-terminal domains 19 and 20 bound simultaneously to C3b and glycosaminoglycans or Sias on host cells, which served to inhibit the alternative pathway [4]. This was further confirmed by structural data that suggest a trimolecular complex of the two C-terminal FH domains, Sia and C3b [5]. Neuraminidase treatment of sheep erythrocytes removes cell surface Sias and reduces the affinity of FH for C3b, which permits complement activation and promotes hemolysis. Typically, FH binds vertebrate cell surfaces via Sias to allow preferential protection of host cells (i.e., reduce complement-mediated damage).
Many microbes express Sias and other unique microbial NulOs (e.g., legionaminic (Leg) and pseudaminic (Pse) acid) on their surfaces that contribute to pathogenesis in numerous ways including subverting complement activation, promoting biofilm formation and facilitating colonization [6]. Some pathogens such as Neisseria gonorrhoeae (Ng), Haemophilus influenzae, Histophilus somni (Haemophilus somnus) and serogroup A N. meningitidis lack the ability to synthesize Sias, but scavenge these molecules (such as Neu5Ac or Neu5Gc, or the cytidinemonophospho (CMP)-activated form CMP-Neu5Ac) from the host. Other pathogens, for example, Escherichia coli K1, Streptococcus agalactiae, N. meningitidis, groups B, C, W and Y, Campylobacter jejuni and certain Leptospira, can synthesize their own nonulosonic acids such as Neu5Ac, Leg5Ac7Ac or Pse5Ac7Ac de novo. Sialylation of gonococcal lacto-N-neotetraose (LNnT; Galβ1-4GlcNAcβ1-3Gal-Glcβ1-4HepI) lipooligosaccharide (LOS) using CMP-Neu5Ac enhances resistance of Ng to complement-dependent killing by decreasing binding of IgG against select bacterial targets such as porin B (PorB) protein [7], which attenuates the classical pathway. LNnT LOS sialylation with Neu5Ac also enhances FH binding, which results in inhibition of the alternative pathway [8].
The purpose of this study was to use CMP-NulOs to define the structural basis of Neu5Acmediated complement inhibition by gonococci. CMP-NulO analogs that serve as substrates for gonococcal LOS sialyltransferase (Lst) and result in NulO modified LOS, may prevent Neu5Ac-mediated serum resistance. This could translate into a novel therapeutic approach to combat infections caused by Ng, a microorganism that has developed resistance to almost every conventional antibiotic.

Substrate specificity of gonococcal lipooligosaccharide sialyltransferase
Humans, unlike other old-world primates, lack a functional CMP-Neu5Ac hydroxylase enzyme (CMAH) and cannot convert CMP-Neu5Ac to the C5 glycolyl (Gc) derivative, CMP-Neu5Gc. In addition to CMP-Neu5Gc (relevant for the mouse model of gonorrhea), we also included other analogs with additional structural variations at carbons 7, 8 and 9, the exocyclic moiety of NulOs. Fig 1 illustrates the CMP-NulOs selected and prepared for this study. In summary, the analogs synthesized include two C9 modified CMP-Neu5Ac analogs (CMP-Neu5Ac9Ac, CMP-Neu5Ac9Az), a C8 modified CMP-Neu5Gc analog (CMP-Neu5Gc8Me), and two C7/C9 modified CMP-Neu5Ac analogs (CMP-Leg5Ac7Ac, CMP-Pse5Ac7Ac) were synthesized. See Materials and Methods and also S1 and S2 Tables and S1 Fig for further details of synthesis/characterization. All of the CMP-NulO analogs except CMP-Pse5Ac7Ac exhibit the same absolute configuration as CMP-Neu5Ac, making them candidate substrates for gonococcal LOS sialyltransferase (Lst).
Gonococcal Lst adds a terminal Neu5Ac residue to LNnT LOS when CMP-Neu5Ac is supplied exogenously (gonococci are unable to synthesize CMP-Neu5Ac) [9]. To define the substrate specificity of gonococcal Lst, we used an isogenic LOS glycosyltransferase D (lgtD) deletion mutant of gonococcal strain F62, called Ng F62 ΔlgtD in subsequent in vitro experiments. LgtD is a phase-variable gene product that adds a terminal GalNAc residue to HepI LNnT [10]; 'capping' LNnT with GalNAc will prevent LOS sialylation. Thus, deleting lgtD permits more homogenous expression of LNnT and uniform sialylation. Ng F62 ΔlgtD was grown in media alone (unsialylated) or in media containing either CMP-Neu5Ac or one of the other six CMP-NulOs (listed in the Table 1 and Fig 1), each at a concentration of~30 μM (20 μg/ ml). Following incubation for 2 h at 37°C, bacterial LOS was examined by western blotting using monoclonal antibody (mAb) 3F11, which binds to the terminal lactosamine residue of LNnT; any extension beyond the terminal Gal (for example with a NulO) will abrogate mAb 3F11 binding. As shown in Fig 2A, growth in media containing CMP salts of Neu5Ac, Neu5Gc, Neu5Ac9Ac, Neu5Ac9Az, Neu5Gc8Me and Leg5Ac7Ac resulted in decreased binding of mAb 3F11. This suggests that these CMP-NulOs served as substrates for gonococcal Lst in the context of live bacteria and the respective NulOs are incorporated into LNnT. Only Pse5Ac7Ac was not incorporated into LNnT LOS. CMP-Pse5Ac7Ac, differs from the other CMP-NulOs stereochemically at C5, C7 and C8, and was not anticipated to be utilized by gonococcal Lst. Consistent with decreased mAb 3F11 binding and addition of a NulO residue, silver staining of LOS showed the appearance of a second, slower migrating band in the 6 inside lanes (Fig 2A). Whole cell ELISA with mAb 3F11 confirmed results of western blotting ( Fig  2B). Direct measurement of NulO incorporation into wild-type Ng F62 was shown for Neu5Gc using chicken polyclonal IgY Ab that specifically recognizes Neu5Gc (Fig 2C). This method directly demonstrates the presence of Neu5Gc on the bacterial surface. Finally, mass Structures of the CMP-nonulosonate (CMP-NulO) analogs used in this study. A General chemical structure, and B 2 C 5 chair chemical structure of CMP-nonulosonates used. All NulO sugars except for pseudaminic acid (Pse5Ac7Ac) have the same stereochemistry (D-glycero-D-galacto configuration), where Pse has an L-glycero-L-manno configuration, differing stereochemically at carbons 5, 7 and 8. For reference, the nine carbon atoms of the NulOs are numbered in A, where the NulO exocyclic moiety is highlighted in red. spectrometric analysis of LOS from bacteria grown in CMP-NulOs confirmed addition of the respective NulO onto LOS (S3 Table).

Serum resistance mediated by incorporation of NulOs
The addition of a terminal Neu5Ac residue to the LNnT LOS of Ng, as occurs in vivo or following the addition of CMP-Neu5Ac to growth media, results in resistance to complement-dependent killing [11]. We next determined the effects of incorporation of the five structural analogs of Neu5Ac on the ability of Ng F62 ΔlgtD ability to resist complement-dependent killing by normal human serum at concentrations of 10%, 6.7% or 3.3%. Bacteria were grown either in media alone, or media supplemented with 30 μM (~20 μg/ml) of each of the CMP-NulOs. As  Fig 4B). We next examined binding of Ng-specific IgG and IgM present in normal human serum (NHS), and deposition of complement components C3, C4 and factor B (FB) on Ng F62 ΔlgtD grown in the presence of the following representative CMP-NulOs: Neu5Ac represented a NulO that bound high levels of FH and conferred high level serum resistance; Neu5Gc8Me did not bind FH and conferred low level serum resistance; Neu5Ac9Ac bound modest amounts of FH and conferred low levels of serum resistance and Neu5Ac9Az and Leg5Ac7Ac neither bound FH nor conferred serum resistance. Ng F62 ΔlgtD with an unmodified (unsialylated) LNnT LOS was used as the control serum sensitive strain. Experiments were carried out using serum concentrations of 3.3% and 10%; the 3.3% serum concentration discriminated serum-killing of the Neu5A-c9Az phenotype from serum-resistance of the Neu5Gc8Me and Neu5Ac9Ac phenotypes.
Consistent with previous studies [7,8,12], modification of LOS with Neu5Ac decreased IgG binding and resulted in maximal inhibition of complement (lowest C3, C4 and FB deposition) at both serum concentrations tested, while high levels of complement activation products were deposited on Ng F62 ΔlgtD (unsialylated) (Fig 5). LOS substitution with Neu5Ac significantly decreased IgG and IgM binding compared to all other CMP-NulOs, which could at least in part, contribute to decreased C4 deposition. While LOS modification with Neu5Ac9Ac and Neu5Gc8Me (both resistant only to 3.3%, but not 10% serum) yielded low levels of C3, C4 and FB deposition in 3.3% serum, only a modest decrease in deposition of these components was noted in 10% serum when compared with Ng F62 ΔlgtD (unsialylated) that was not likely to be sufficient to subvert killing by complement. Neu5Gc8Me and Neu5Ac9Ac substitution also reduced IgG binding, but not to the extent seen with Neu5Ac. The two analogs that left Ng F62 ΔlgtD fully serum sensitive, Neu5Ac9Az and Leg5Ac7Ac, yielded the highest levels of Ig binding and complement deposition among the analogs tested. A summary of CMP-NulOs utilized by Lst, Ab and complement component binding, and serum resistance associated with each of the tested CMP-NulOs is found in Table 1.

Select NulO derivatives inhibit Neu5Ac-mediated serum resistance
We next asked whether the two NulOs, Neu5Ac9Az and Leg5Ac7Ac, that were incorporated into gonococcal LOS but did not enhance serum resistance, could prevent the ability of CMP-Neu5Ac to enhance serum resistance of N. gonorrhoeae. These CMP-activated analogs were added at concentrations of 20, 2 or 0.2 μg/ml to growth media 15 min after the addition of CMP-Neu5Ac (20 μg/ml) to Ng F62 ΔlgtD (Fig 6A), and bacteria were allowed to grow for 2 h, followed by bactericidal assays using 10% NHS. As shown in Fig 6A, both CMP-Neu5Ac9Az and CMP-Leg5Ac7Ac prevented serum resistance mediated by CMP-Neu5Ac, even when the  latter was present in a 100-fold molar excess. Adding CMP-Leg5Ac7Ac and CMP-Neu5Ac to the growth media simultaneously, each a concentration of 20 μg/ml, also yielded similar results -the mean survival of Ng F62 ΔlgtD was only 7.6%. Similarly, CMP-Leg5Ac7Ac prevented CMP-Neu5Gc-mediated serum resistance ( Fig 6B). CMP-Neu5Ac9Az and CMP-Leg5Ac7Ac also prevented Neu5Ac-mediated serum resistance of a ceftriaxone-resistant (CRO-R) isolate, H041 (Fig 6C), however prevention was less than with Ng F62 ΔlgtD (minimal prevention seen with CMP-NulO concentrations of 0.2 μg/ml). In the unsialylated state, H041 is intrinsically more serum resistant than F62 ΔlgtD (H041 shows >100% survival in 3.3% NHS, while F62 shows <10% survival in 3.3% NHS).

Competing NulOs block Neu5Ac-mediated complement inhibition
To define the mechanism whereby Neu5Ac9Az and Leg5Ac7Ac blocked complement resistance by Neu5Ac, we studied Ab binding to, and complement component deposition on Ng F62 ΔlgtD that was incubated with CMP-Neu5Ac (20 μg/ml), followed 15 min later with either CMP-Neu5Ac9Az or CMP-Leg5Ac7Ac at concentrations of 20, 2 or 0.2 μg/ml. Controls included bacteria grown in the absence of any Sia, or bacteria grown in CMP-Neu5Ac alone.
As shown in Fig 7 (and above in Fig 5), incorporation of Neu5Ac into LOS decreased IgG binding and also diminished deposition of complement components C4 and C3. However, addition of CMP-Neu5Ac9Az or CMP-Leg5Ac7Ac at all concentrations tested resulted in lesser reduction in IgG binding in 10% NHS and also in the corresponding deposition of C4 and C3. Similar results were also seen when a lower concentration of serum, 3.3% NHS was used (S4 Fig). CMP-Neu5Ac9Az and CMP-Leg5Ac7Ac also resulted in lesser reduction in FB deposition (Figs 8A and S4); surprisingly these NulOs did not interfere with FH binding to Ng F62 ΔlgtD ( Fig 8B). These data seem paradoxical because under physiological conditions, FH functions to irreversibly dissociate factor Bb from C3bBb and also prevents factor B from binding C3; thus, FH and FB/FBb binding usually bear inverse proportions (reviewed in [13]). The relatively unimpeded activation of the alternative pathway, evidenced by high FB deposition seen in the presence of the competing CMP-NulOs that occurs in the face of FH binding, may suggest impaired FH function when Leg5Ac7Ac or Neu5Ac9Az are expressed on the bacterial surface concomitantly with Neu5Ac.
Ab-dependent classical pathway activation is essential for killing of N. gonorrhoeae [14,15]; disabling the classical pathway abrogates killing of otherwise fully serum sensitive gonococci [15]. To define the role of the alternative pathway in mediating killing by CMP-Leg5Ac7Ac, the function of FB was blocked with an anti-factor Bb mAb [16]. Blocking the alternative pathway enabled unsialylated Ng F62 ΔlgtD to survive 56% in the pooled NHS used in these assays ( Fig 8C). As expected growth of bacteria in CMP-Neu5Ac abrogated killing (CFU at t 30 was 115% of the CFUs at t 0 ) ( Fig 8C). Addition of CMP-Leg5Ac7Ac (20 μg/ml) alone to media, or 15 min after the addition of CMP-Neu5Ac (20 μg/ml), resulted in 76% and 112% survival, respectively, when FB function was blocked ( Fig 8C). By contrast (and consistent with the data in Fig 6A), the parallel control bactericidal reactions with all pathways intact (i.e. no anti-Bb) showed 0% survival when bacteria were grown in CMP-Leg5Ac7Ac alone, increasing only to 8% when this NulO was added to media 15 min after CMP-Neu5Ac. Thus, interfering with the function of LOS Neu5Ac by CMP-Leg5Ac7Ac on Ng F62 ΔlgtD appears to require a functioning alternative pathway.  that conversion efficiency for N. meningitidis and N. gonorrhoeae F62 α2,3-sialyltransferases with CMP-Leg5Ac7Ac as a donor is comparable to that with CMP-Neu5Ac [17]. Further, binding of mAb 3F11 (binds only to unsialylated LNnT; extensions beyond the terminal Gal of LNnT abrogates 3F11 binding) to Ng F62 ΔlgtD grown in concentrations ranging from 0.125 to 25 μg/ml of CMP-Neu5Ac or CMP-Leg5Ac7Ac were similar, providing additional evidence that Leg5Ac7Ac (when used alone) is not preferentially added to the Ng LNnT LOS (S5 Fig). Finally, the ratio of Leg5Ac7Ac:Neu5Ac incorporated into Ng F62 ΔlgtD LOS,after CMP-Le-g5Ac7Ac was added to growth media 15 min after CMP-Neu5Ac had been added, was approximately 1:2 (S6 Fig). Collectively, these data provide strong evidence that the effects of Leg5Ac7Ac in preventing the complement-inhibiting effects of Neu5Ac were not because of the preferential addition of Leg5Ac7Ac on to LOS.  (Ng) F62 ΔlgtD was incubated in media that contained CMP-Neu5Ac (20 μg/ml), followed by addition of either CMP-Neu5Ac9Az or CMP-Leg5Ac7Ac (20, 2 or 0.2 μg/ml). Bacteria were grown for 2 h and factor B (FB) binding (A) to bacteria following incubation with 10% normal human serum (NHS) was measured by whole cell ELISA as described in Fig 7. Binding to purified FH (B) was measured by flow cytometry using polyclonal goat anti-human FH. The Y-axis represents OD 405nm for FB binding and median fluorescence of FH binding. Mean (SD) of two independent observations is shown for each measurement. ****, P<0.0001 (ANOVA) compared to other reactions. C. Alternative pathway activation is required for maximal killing of Ng F62 ΔlgtD grown in CMP-Leg5Ac7Ac-containing media. Bacteria were grown in media alone, or media containing CMP-Neu5Ac (20 μg/ml) or CMP-Leg5Ac7Ac (20 μg/ml) or both CMP-NulOs together (each at 20 μg/ml) as indicated on the X-axis. Serum bactericidal assays were performed using normal human serum (all complement pathways intact; black bars) or NHS treated with an anti-factor Bb mAb that blocks FB function (AP inactivated, but classical and lectin pathways intact; grey bars). Y-axis, percent survival. The mean (range) of two independent experiments is shown.

CMP-Leg5Ac7Ac attenuates gonococcal infection in mice
Given the urgent need for novel therapies against multidrug-resistant gonorrhea that has spread globally, we next tested the ability of CMP-Leg5Ac7Ac to attenuate gonococcal infection in mice. CMP-Leg5Ac7Ac was chosen because it most effectively blocked Neu5Ac-mediated complement inhibition (Figs 5 and 7 and 8) and because Leg5Ac7Ac-α2,3-lactose-FCHASE substrates are more resistant to microbial and human sialidases (that Ng may encounter in the genital tract) than Neu5Ac-α2,3-lactose-FCHASE substrates [18]. Two groups of 17β-estradiol-treated BALB/c mice (10 mice per group) were infected with wild-type strain F62. We used wild-type Ng for the experiments in vivo because phase variation of LOS (i.e., 'GalNAc capping' via LgtD) could constitute a possible mechanism of 'drug resistance'. One group was treated with 10 μg CMP-Leg5Ac7Ac intravaginally daily; the control group was untreated. A third group of mice (n = 10) was infected with F62 Δlst (unable to sialylate LOS) and served as a control. As shown in Fig 9A, mice treated with CMP-Leg5Ac7Ac cleared infection faster than untreated mice (median clearance time was 6 days in treated mice compared to 10 days for untreated mice; P<0.0001, Kaplan-Meier survival curve; clearance times were compared between groups using a log-rank test). Mixed model analysis indicated significant differences in colonization trends of F62 wild-type over time between the two groups comparing CMP-Leg5Ac7Ac versus untreated mice (P = 0.0005) (Fig 9B). A significant difference in the Mean Areas Under the Curve (mean AUCs) (log 10 CFU versus time) between the treated and untreated groups challenged with F62 WT was also observed (Fig 9C). To address the possibility that non-specific stimulation of the immune system by CMP-Leg5Ac7Ac may have accounted for its activity, we examined the effects of CMP-Leg5Ac7Ac treatment on the burden of infection caused by an Lst deletion mutant of F62 (F62 Δlst) that is unable to add NulO to its LOS. As shown in S7 Fig, CMP-Leg5Ac7Ac treatment did not affect the course of infection (measured as time to clearance or AUC) of F62 Δlst.
To confirm that CMP-Leg5Ac7Ac treatment reduced serum resistance of Ng in vivo, we incubated bacteria recovered directly from mice (without sub-passage on media) with normal human serum and measured their survival after 30 min. As shown in Fig 9D, the wild-type bacteria from untreated mice remained resistant to complement over all 5 days; bacteria recovered from the CMP-Leg5Ac7Ac treatment group became progressively more serum sensitive over time. As expected, the Δlst mutant was uniformly serum sensitive on all days. This experiment was limited to 5 days, because the numbers of bacteria recovered from the treatment and Δlst groups thereafter were too low to perform the assay.
Similar to findings with F62, CMP-Leg5Ac7Ac significantly attenuated the duration and burden of infection caused by CRO-R isolate H041 in mice (Fig 10). Of note, the H041 isolate is naturally more serum resistant than F62, resulting in minimal clearance at day 10 of salinetreated mice (Fig 10A and 10B).

Discussion
Over 40 years ago, Ward and colleagues showed that gonococci recovered directly from male urethral secretions fully resisted killing by human complement [19]. However, sub-passage of gonococci on routine culture media abrogated the property of serum resistance and thus was termed 'unstable' serum resistance. Almost two decades later, seminal work performed by Smith and co-workers [11,20,21] identified CMP-Neu5Ac as the molecule scavenged by Ng in vivo that was responsible for conferring unstable serum resistance. The only molecule modified by Neu5Ac on Ng is LOS [9]. Sialylation of LOS is mediated by Lst, an enzyme located on the Ng outer membrane [22]. The two gonococcal HepI LOS extensions that can be modified by Neu5Ac are lacto-N-neotetraose (LNnT) (Neu5Acα2-3Galβ1-4GlcNAcβ1-3Galβ1-4Glcβ1-4HepI) and the P K LOS (Neu5Acα2-6Galα1-4Galβ1-4Glcβ1-4HepI) [23]. Here we show that substrate specificity of gonococcal Lst in the context of intact bacteria extends to five additional CMP-Sia analogs listed in Table 1.
Several previous studies have elucidated mechanisms of gonococcal LNnT LOS Neu5Acmediated serum resistance and point to regulation of both the classical and alternative pathways of complement [7,8,24,25]. The presence of LOS Neu5Ac on gonococci decreases binding of select mAbs directed against gonococcal PorB [7]. Another report suggested that LOS Neu5Ac did not affect the quantity of anti-LOS IgM binding to bacteria, but perhaps masked epitopes usually recognized by pentameric IgM molecules that resulted in limited engagement of C1q and subsequent activation of the classical pathway [25]. The addition of Neu5Ac via an α2-3 linkage to the terminal Gal of LNnT also enhances binding of human FH, the principal inhibitor of the alternative pathway of complement. Enhancement of FH binding to sialylated gonococci also requires the presence of gonococcal PorB specifically; specifically replacement of gonococcal PorB with meningococcal PorB3 does not augment FH binding [26]. Neu5Ac that enhances FH binding and results in high-level serum resistance of gonococci is also specific for the LNnT LOS structure. Previously we reported that addition of α2-6-linked Neu5Ac to the terminal Gal of the P K -like (or L1) LOS did not enhance FH binding and conferred resistance only to lower concentrations of serum relative to serum resistance seen with Neu5Ac α2-3-linked to LNnT LOS [23]. Binding of FH hastens dissociation of alternative pathway C3/C5 convertases (C3b,Bb and C3bC3b,Bb, respectively) [13] and likely contributes to the barely detectable levels of FB on the Neu5Ac and Neu5Gc-bearing strains even in 10% serum.
In contrast to the high-level serum resistance seen with LOS bearing Neu5Ac and Neu5Gc, NulOs with further modifications at C7, C8 and C9 were equally well incorporated and yet conferred only low level serum resistance (Neu5Ac9Ac or Neu5Gc8Me) or no serum resistance (Leg5Ac7Ac and Neu5Ac9Az). These data also suggest that the negative charge of Sias alone is not sufficient for FH binding or complement evasion-Neu5Gc8Me, Leg5Ac7Ac and Neu5A-c9Az are all negatively charged, yet do not enhance FH binding or classical pathway inhibition. In contrast, N-glycolyl substitution at the 5-carbon position of Neu5Ac (to Neu5Gc) facilitates FH binding and increases serum resistance. The importance of the C9 substitutions reported here are in accordance with several previous studies as discussed below. Removal of the 9-carbon was associated with loss of 90% of the complement inhibition by Sia [27]. The extent of 9-Oacetylation of Sias on erythrocytes varies widely across mouse strains and correlates directly with the susceptibility of erythrocytes to lysis by the alternative pathway [28]. Similarly, increased expression of 9-O-acetylated sialoglycans on the red cells of individuals with visceral leishmaniasis is associated with greater alternative pathway activation [29]. Fearon showed that removal of C8 and C9 substitutions of Sia by NaIO 4 treatment rendered sheep erythrocytes susceptible to lysis by the alternative pathway [3]. Using ligand-based nuclear magnetic resonance (NMR), Blaum et al recently showed that the O8 and O9 hydroxyl groups of the glycerol chain of Neu5Ac that was α2-3-linked to lactose formed hydrogen bonds with the amide and carbonyl groups, respectively, of the W1198 residue in FH domain 20 [5]. In addition the O9 also formed a hydrogen bond with the W1183 backbone [5]. Effective and unimpeded alternative pathway inhibition when the 5Ac group is replaced with 5Gc is also consistent with a prior study [27]. Previously we reported that the addition of Neu5Ac to gonococcal LOS diminished classical pathway activation (less C4 fragment deposition) [12]. In the present study, using relatively low numbers of bacteria in Ab binding and complement deposition assays to more closely simulate conditions in serum bactericidal assays, we observed a substantial reduction in binding of IgG to sialylated (Neu5Ac and Neu5Gc) bacteria. Reduced IgG binding could account, at least in part, for classical pathway inhibition mediated by Neu5Ac. Downstream events, including C4, C3 and FBb deposition (the latter a measure of alternative pathway recruitment), were effectively regulated only in low serum concentrations when LOS expressed Neu5Gc8Me and Neu5Ac9Ac, but were not likely to reach a threshold that permitted bacterial survival in 10% complement. Of note, the intermediate phenotype exhibited by CMP-Neu5Ac9Ac treated bacteria, compared to CMP-Neu5Ac9Az treatment, is likely due to the instability of the C9 O-acetyl group where some of the Neu5Ac9Ac is converted to Neu5Ac, as this instability was observed during preparation and storage of CMP-Neu5Ac9Ac.
Sialic acids (Sias) are present on all vertebrate cell types and play diverse roles in immune function [2]. Humans (and birds and reptiles) lack a functional CMP-Neu5Ac hydroxylase enzyme (CMAH) and therefore cannot convert Neu5Ac to Neu5Gc [30]. In contrast to humans, most other mammals studied (with the exception of ferrets [31] and new world monkeys [32]) including mice, possess a functional CMAH and therefore express both Neu5Ac and Neu5Gc glycoconjugates. Thus, the LOS of Ng recovered from mice during experimental studies of gonococcal colonization may be modified with either Neu5Ac and/or Neu5Gc. The ability of Neu5Gc on Ng LOS to also inhibit complement validates the use of the mouse model to study complement interactions with gonococci and evaluate vaccines and therapeutics that may rely on complement activation for their efficacy. The observation that Neu5Gc also facilitates binding of human FH and results in high-level serum resistance that simulates the findings with Neu5Ac would validate the use of human FH transgenic mice to elucidate the role of complement in pathogenesis of sialylated gonococci.
Binding of purified FH (independent of C3 fragment deposition as occurs in context of normal serum) to sialylated gonococci is human specific, with weak binding observed to chimpanzee FH [33]. This raises the question of the utility of wild-type mice, whose endogenous FH is not known to bind directly to sialylated gonococci, for studies related to sialic acid. Distinct from the mechanism that involves direct FH binding described above, gonococcal LOS Neu5Ac also regulates non-human (e.g., rhesus macaque) C3 deposition on gonococci, albeit to a lesser extent than seen in species (chimpanzee and human) whose FH directly bind (i.e., in a C3 independent manner) to sialylated bacteria [33]. Although not formally demonstrated on sialylated gonococci, we have shown that Neu5Ac substitution of meningococcal LNnT LOS limits non-human (in this instance, rat) C3 deposition [34]. This mechanism represents a cooperative interaction of the C-terminal FH domains with C3 fragments that bind to bacteria and Neu5Ac-LNnT [34] that is concordant with the prior model proposed by Kajander and colleagues [4]. A key role for sialic acid in bacterial survival in vivo in wild-type mice has been shown in Fig 10 (lower 'virulence' of the Lst mutant), and replicates prior findings [35,36].
We and others previously have proposed a role for bacterial Neu5Ac (reviewed in [6]), including Leg5Ac7Ac in immune evasion [37,38], possibly through increased FH binding. However, in the context of Ng, the current data (Figs 4B and S3) refute this notion for Leg5A-c7Ac. Paradoxically, incorporation of Leg5Ac7Ac by Ng LOS appears to confer heightened susceptibility of Ng to the immune system.
Ng has developed widespread resistance to almost every conventional antibiotic currently in clinical use [39]. High-level resistance to ceftriaxone was first reported in 2011 [40,41] and has ushered in an era of potentially untreatable gonorrhea. There is an urgent need for novel therapeutics and vaccines against this disease [42,43]. LOS sialylation is an important component of gonococcal pathogenesis; isogenic mutants that cannot sialylate their LOS are at a disadvantage in vivo compared to their wild-type counterparts [44]. Manipulation of gonococcal LOS 'sialylation' could represent a novel preventive or treatment strategy. We acknowledge that further work is needed to elucidate the mechanistic details of the 'dominant negative' effect of Leg5A-c7Ac/ Neu5Ac9Az over Neu5Ac substitution of gonococcal LOS, however given the dire need for alternative therapeutics against drug-resistant gonorrhea, we decided to focus first on the practical utility of these findings. Here we have shown that select derivatives of Sia, such as Neu5Ac9Az and Leg5Ac7Ac block Neu5Ac-mediated serum resistance (Fig 6). In vitro, the effects of these CMP-Sia derivatives are also evident when added to bacteria shortly after adding CMP-Neu5Ac. In vivo, however, bacteria actively replicate for several generations and are likely to incorporate the competing Sia analog over time. Of note, the effects of CMP-Neu5Ac were negated even when CMP-Neu5Ac9Az or CMP-Leg5Ac7Ac were provided at~100-fold lower concentrations. These competing CMP-NulOs diminished inhibition of the classical pathway (determined by measuring both increased IgG binding and C4 deposition) and serum resistance elicited by CMP-Neu5Ac. Surprisingly, FBb continued to bind to Ng in an uninhibited manner (Fig 8A) in the face of high FH binding (Fig 8B). This raises the possibility that FH is rendered 'non-functional' (i.e., does not display decay accelerating activity and liberate FBb from C3/C5 convertases) when Leg5Ac7Ac-or Neu5Ac9Az-substituted LOS and Neu5Ac-subsituted LOS are concomitantly expressed on the bacterial surface. The reasons for the 'dominant-suppressive' effect of the competing NulOs over complement inhibition by Neu5Ac remain to be elucidated. Based on the phylogeny of NulOs [45] it is likely that Ng Lst may have evolved from an ancestral transferase that originally used CMP-bacterial NulOs such as CMP-Leg5Ac7Ac as substrates.
In addition to its ability to counteract Neu5Ac-mediated complement inhibition effectively, CMP-Leg5Ac7Ac was chosen to study therapeutic efficacy in vivo because of the sialidase resistance exhibited by Leg5Ac7Ac, as evidenced by the resistance of Leg5Ac7Ac-α(2,3)-lactose-FCHASE to mammalian and microbial sialidases [18]. Sialidases are present in the female genital tract [46,47,48,49,50]; resistance of Leg5Ac7Ac-substituted LNnT on Ng to sialidases would be a desired attribute for a therapeutic CMP-NulO. Indeed, mice treated with CMP-Leg5A-c7Ac cleared wild-type Ng F62 more rapidly (Fig 9A) and harbored lower bacterial burdens over the course of infection (Fig 9B and 9C) analogous to clearance and colonization levels using a Lst knockout mutant of strain F62. Furthermore, as daily treatment of mice with CMP-Leg5Ac7Ac took hold, surviving bacteria isolated from mice during the course of infection became progressively more sensitive to killing by complement (Fig 9D), consistent with organisms having incorporated Leg5Ac7Ac (Figs 2 and 3). Under the influence of CMP-Le-g5Ac7Ac, conversion of the infecting strain into a phenotype unable to sialylate only with Neu5Ac (and/or Neu5Gc in the mouse [51]) resulted in an attenuation of bacterial burden similar to that seen with the Lst mutant, which manifests incomplete virulence in the mouse model of gonococcal infection [36]. In contrast to strain F62, gonococcal strain H041 is fully resistant to ceftriaxone (CRO-R), resistant to 3.3% serum, and more resistant to the bactericidal properties conferred by NuIOs Neu5Ac9Az and Leg5Ac7Ac than strain F62 ΔlgtD (Fig 6A and 6C). Nevertheless, mice treated with CMP-Leg5Ac7Ac also cleared strain H041 more rapidly and harbored lower bacterial burdens than saline-treated infected animals (Fig 10).
In conclusion, these findings have shed new light on the molecular basis of gonococcal LOS sialylation and how this translates to complement regulation on the bacterial surface. CMP-Sia analogs that interfere with serum resistance of gonococci mediated by CMP-Neu5Ac may predict their efficacy as topical anti-gonococcal therapeutics in vivo. The identification of such Sia analogs that can interfere with a key virulence mechanism of Ng paves the way to develop novel therapeutic strategies against rapidly spreading multi-drug resistant gonococcal infections.
Purified CMP-nonulosonates were dissolved in >99% D 2 O and structural analysis performed by mass spectroscopy (MS) using either a Varian Inova 500 MHz ( 1 H) spectrometer with a Varian Z-gradient 3 mm probe or a Varian 600 MHz ( 1 H) spectrometer with a Varian 5 mm Z-gradient probe. All spectra were referenced to an internal acetone standard (δ H 2.225 ppm and δ C 31.07 ppm). Results that are shown in S1 Table and S1 Fig verify the production and purity of each CMP-nonulosonate compound made for this study; CMP-LegAc7Ac and CMP-Pse5Ac7Ac, which were confirmed based on NMR data presented in [38] and [37].
CMP-NulOs were also characterized using capillary electrophoresis (CE)-MS analysis. Separation of ions was achieved by CE (Prince CE system [Prince Technologies, Netherlands]) in a 90 cm long bare fused-silica capillary (365 μm OD x 50 μm ID). The 30 mM morpholine aqueous running CE buffer (adjusted to pH9 with formic acid) was coupled with the capillary sheath fluid (isopropanol: methanol [2:1]) at their interface prior to mass spectrometry (API3000 mass spectrometer [Applied Biosystems/Sciex, Concord, ON, Canada]). S2 Table indicates measured m/z ions of each CMP-nonulosonate compound using CE-MS. Measurements correspond to calculated masses.

Bacterial strains and growth conditions
A mutant of Ng strain F62 [57] that lacked expression of lipooligosaccharide glycosyltransferase D (lgtD), called F62 ΔlgtD [58], was provided by Dr. Daniel C. Stein (University of Maryland). LgtD adds a GalNAc residue to the terminal Gal of the HepI lacto-N-neotetraose species [10]. Therefore, any extension of the HepI of F62 ΔlgtD is limited to the addition of a NulO residue that is transferred from the CMP-NulO added to growth media.
Bacteria (F62 ΔlgtD) grown overnight on chocolate agar plates were suspended in gonococcal liquid media supplemented with IsoVitaleX [59] that contained specified concentrations of the CMP-NulO. Bacteria were then incubated at 37°C for the period specified in each experiment.
A spontaneous streptomycin-resistant mutant of Ng F62 [60], kindly provided by Dr. Ann E. Jerse, was used in mouse infection studies. An isogenic lst deletion mutant, derived from this strain, was constructed using plasmid pUC18-lst-Kan [23].
Strain H041 (sequence type (ST) 7363; NG-MAST ST 4220) was isolated from a female commercial sex-worker in Kyoto, Japan [40]. This isolate is highly resistant to ceftriaxone (MIC 2-4 μg/ml) and to several other antibiotics [40]. H041 was rendered resistant to streptomycin by transformation with rpsL derived from streptomycin resistant Ng strain FA1090. Both wildtype H041 and streptomycin resistant H041 expressed predominantly the sialylatable HepI lacto-N-neotetraose LOS, confirmed by silver stain and western blot with mAb 3F11.

Antibodies
Anti-FH mAb (Quidel, catalog no. A254 (mAb 90X)) or goat anti-human FH were used in flow cytometry assays to detect human FH binding to bacteria. Goat polyclonal antibodies against C3, C4 and factor B (FB) were obtained from Complement Technology, Inc (Tyler, TX). Anti-factor Bb (Quidel) added to serum at 100 μg/ml was used to block factor B function [16], thereby disabling the alternative pathway in serum bactericidal assays. Alkaline phosphatase conjugated anti-human IgG and IgM were purchased from Sigma (St. Louis, MO). mAb 3F11 (mouse IgM, kindly provided by Dr. Michael A. Apicella, University of Iowa) binds to the unsialylated HepI lacto-N-neotetraose structure; sialylation of LOS results in decreased binding of mAb 3F11 [61]. MAb 2-8C-4-1 [62] recognizes Neisserial H.8 lipoprotein and was used in whole cell ELISA assays to measure capture of bacteria on microtiter wells. FITC conjugated anti-mouse IgG and anti-goat IgG, and alkaline phosphatase conjugated anti-mouse IgM, antimouse IgG and anti-goat IgG were all obtained from Sigma. Neu5Gc incorporation into LNnT LOS was detected using a Neu5Gc-specific chicken polyclonal IgY Ab (1:2,000) [63] followed by FITC conjugated donkey anti-chicken IgY secondary Ab (1:200; Jackson ImmunoResearch).

SDS-PAGE and Western blotting for LOS analysis
Gonococcal lysates treated with protease K (100 μg/ml) and NuPAGE LDS Sample Buffer (4X) (Invitrogen) were separated on NuPAGE 12% Bis-Tris (Invitrogen) gels using Novex MES running buffer (Invitrogen) followed by transfer to an Immobilon PVDF membrane (Millipore) by western blotting. Membranes were blocked with PBS/1% milk and probed with tissue culture supernatants that contained mAb 3F11. mAb 3F11-reactive LOS bands were disclosed with anti-mouse IgM conjugated to alkaline phosphatase followed by the addition of BCIP/ NBT-Purple Liquid Substrate (Sigma). Silver staining for LOS was performed following electrophoresis on 16.5% Criterion Tris-Tricine gels (Bio-Rad) using the Bio-Rad Silver Stain kit.

Mass spectrometry of gonococcal LOS
Ng F62 ΔlgtD was grown in media containing the CMP salts of Neu5Ac, Neu5Gc, Neu5Ac9Ac, Neu5Gc8Me, Neu5Ac9Az and Leg5Ac7Ac and LOS was extracted, de-O-acylated and analyzed by MS as described previously [64].

Flow cytometry for factor H binding
Factor H (FH) binding to bacteria was performed using flow cytometry as described previously [16]. Briefly, bacteria (Ng F62 ΔlgtD) were harvested from chocolate agar plates and grown in liquid media that contained the specified concentration of the CMP-NulO as described above. Bacteria were then washed with Hanks Balanced Salt Solution (HBSS) containing 1mM Ca 2+ and 1 mM Mg 2+ (HBSS ++ ) and incubated with FH purified from human plasma (Complement Technology, Inc.; concentration specified for each experiment). Bound FH was detected using either anti-FH mAb (Quidel, catalog no. A254 (mAb 90X)) or affinity-isolated polyclonal goat anti-human FH, followed by FITC conjugated anti-mouse IgG or anti-goat IgG, respectively (Sigma); both Abs had similar performance characteristics. All reaction mixtures were carried out in HBSS ++ /1% BSA in a final volume of 50 μl. Flow cytometry was performed using a FACSCalibur instrument (Becton Dickinson) and data were analyzed using FlowJo (version 7.2.5; Tree Star, Inc.).
Whole cell ELISA for complement component deposition C3 and C4 fragment deposition on, and FB binding to bacteria were measured by whole cell ELISA as described previously [12,16]. Briefly,~10 6 organisms in HBSS ++ were incubated with NHS (at concentrations of 3.3% or 10%) in a reaction volume of 100 μl for 10 min at 37°C. This time point was chosen based on the kinetics of complement deposition on gonococci in previously published data [12]. Reactions were stopped after 10 min by washing three times with ice-cold HBSS containing 5 mM phenylmethylsulfonyl fluoride at 4°C. Organisms were resuspended in 200 μl of the same buffer, and 50 μl of each sample applied per well of a 96-well U-bottomed polystyrene microtiter plate (Dynatech Laboratories, Inc., Chantilly, VA) and incubated for 3 h at 37°C. Plates were washed with PBS containing 0.05% Tween 20. Primary antibodies (polyclonal goat anti-human C3, C4 and FB) were diluted in PBS, and secondary antibodies diluted in PBS-0.05% Tween 20 prior to use. To ensure similar capture of bacteria incubated under different conditions, we measured the amount of gonococcal H.8 antigen [65] expression using mAb 2-8C-4-1 [62], followed by anti-mouse IgG-alkaline phosphatase conjugate.

Serum bactericidal assay
Serum bactericidal assays were performed as described previously [59]. Bacteria harvested from an overnight culture on chocolate agar plates and~10 5 CFU of Ng were grown in liquid media containing the specified concentration of CMP-Sia as specified for each experiment. Bacteria were diluted in Morse A and~2000 CFU of Ng F62 ΔlgtD were incubated with NHS (concentration specified for each experiment). The final reaction volumes were maintained at 150 μl. Aliquots of 25 μl of reaction mixtures were plated onto chocolate agar in duplicate at the beginning of the assay (t 0 ) and again after incubation at 37°C for 30 min (t 30 ). Survival was calculated as the number of viable colonies at t 30 relative to t 0 .
To measure the sensitivity of bacteria recovered directly from mouse secretions to human complement, an equal aliquot of secretions from each mouse within each group were pooled (pools were made to ensure adequate number of bacteria for the assay over the first 5 days). Secretions from each pool were serially diluted, each dilution was incubated with 10% NHS and survival at t 30 relative to t 0 was carried out as described above. Only those dilutions that yielded~100-300 CFU at t 0 were considered in order to ensure similar bacteria-to-complement ratio across groups.

Mouse vaginal colonization model
Female BALB/c mice 5-6 weeks of age (Jackson Laboratories) in the diestrus phase of the estrous cycle were started on treatment (that day) with 0.5 mg of water soluble 17β-estradiol (Sigma) in 200 μl of water given subcutaneously on each of three days; −2, 0 and +2 days (before, the day of and after inoculation) to prolong the estrus phase of the cycle and promote susceptibility to Ng infection. Antibiotics (vancomycin, colistin, neomycin, trimethoprim and streptomycin) ineffective against Ng were also used to reduce competitive microflora [66]. Mice (n = 20) were then infected with 3.4 x 10 5 CFU of strain F62. One group of mice (n = 10) was treated with 10 μg CMP-Leg5Ac7Ac (1 mg/ml in sterile H 2 O) daily intravaginally while the remaining 10 mice (controls) were given saline (vehicle control). A third group (n = 10) was infected with 2.5 x 10 5 CFU of Ng F62 Δlst (unable to sialylate) and received no treatment. In a separate control experiment, Ng F62 Δlst infection was compared with Ng F62Δlst infection 'treated' with CMP-Leg5Ac7Ac.

Statistics
Experiments that compared clearance of N. gonorrhoeae in independent groups of mice estimated and tested three characteristics of the data [67]: Time to clearance, longitudinal trends in mean log 10 CFU and the cumulative CFU as area under the curve (AUC). Statistical analyses were performed using mice that initially yielded bacterial colonies on Days 1 and/or 2. Median time to clearance was estimated using Kaplan-Meier survival curves; the times to clearance were compared between groups using a log-rank test. Mean log 10 CFU trends over time were compared between groups using a linear mixed model with mouse as the random effect using both a random intercept and a random slope. A quadratic function in time was determined to provide the best fit; random slopes were also quadratic in time. A likelihood ratio test was used to compare nested models (with and without the interaction term of group and time) to test whether the trend differed over time between the two groups. The mean AUC (log 10 CFU vs. time) was computed for each mouse to estimate the bacterial burden over time (cumulative infection); the means under the curves were compared between groups using the nonparametric rank sum test because distributions were skewed or kurtotic.

Ethics statement
Collection of human sera and its use were approved by the University of Massachusetts Medical School Institutional Review Board (IRB). Informed, written consent was obtained from all serum donors (Docket # H00005614). Use of animals in this study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Massachusetts Medical School (Docket # A-1930).
Online supplemental material  Supporting Information S1 Fig. The 1 H spectrum and 1 H-13 C HSQC correlation spectrum of different CMP-nonulosonic acids used in this study. Spectra for CMP-Pse5Ac7Ac and CMP-Leg5Ac7Ac have been described previously [37,38]. Spectra were recorded on a Varian Inova Unity 500 MHz spectrometer with standard Varian pulse sequences in D 2 O at 25°C, with 16 scans for the 1 (Ng) strain F62 lgtD grown in media containing no CMP-NulO, or in media containing CMP-Neu5Ac alone (20 μg/ml), CMP-Leg5Ac7Ac alone (20 μg/ml), or in media where CMP-Leg5Ac7Ac was added 15 min after CMP-Neu5Ac (both CMP-NulOs at 20 μg/ ml). LOS was extracted on a small-scale from a 12 ml culture volume using a modification of the phenol-chloroform method [68]. A. Estimation of the relative amounts of LOS in the samples. Equal volumes of each preparation taken just prior to the final lyophilization step were loaded on a 4-12% Bis-Tris gel (Invitrogen/Life Technologies) with MES running buffer and LOS was revealed by silver staining. The relative intensities of the LOS bands were estimated using ImageJ software (NIH) and are indicated below each lane. B. Estimation of the relative amounts of Neu5Ac and Leg5Ac incorporated by Ng F62 lgtD lacto-N-neotetraose LOS. Lyophilized LOS extracts samples were dissolved in H 2 O. Acid hydrolysis of the NulOs was performed with 0.1 M hydrochloric acid at 80°C for 1 h to release them from the underlying LOS backbone Samples were cooled to room temperature, neutralized with NaOH and then derivatized with 1,2-diamino-4,5-methylene-dioxybenzene (DMB) and analyzed by high performance liquid chromatography (HPLC), as described below. The DMB derivatization reagent was made by mixing 14 mM DMB (Sigma), 18 mM sodium hydrosulfite, 0.75 M 2-mercaptoethanol, and 1.6 M acetic acid, followed by incubation at 50°C for 2.5 h [69,70]. DMB-derivatized samples were analyzed on a LaChrom Elite HPLC System (Hitachi) using a Phenomenex Gemini 5μ C18 250 × 4.6-mm HPLC column at room temperature. Fluorescence was detected at 448 nm using excitation at 373 nm. To separate NulOs, an isocratic solvent composition of 88% water, 7% methanol and 5% acetonitrile was used at a flow rate of 0.9 mL/ min; the data collection time was expanded to 90 min. A Neu5Ac standard curve is shown on the left. Relative amounts of each of the NulOs (normalized to the respective LOS band densities) are indicated on the graph on the right as the normalized peak area.  Table. NMR chemical shifts for CMP-nonulosonates prepared in this study. All spectra were referenced to an internal acetone standard (δ H 2.225 ppm and δ C 31.07 ppm). (DOC) S2 Table. CE-MS data of purified CMP-nonulosonate sugars prepared in this study. The detected ions indicated here were not present in negative control samples. (DOCX) S3 Table. MS analysis of LOS of Ng F62 lgtD grown in CMP-NulOs. (DOCX)