Synthesis and Evaluation of 3‑Deoxy‑d-manno-oct-2-ulosonic Acid Derivatives to Perturb Escherichia coli Lipopolysaccharide Biosynthesis

Abstract

Lipopolysaccharides (LPS) play important roles in the Gram-negative bacterial cell envelope. LPS are located in the outer leaflet of the outer membrane and generally serve as the first defense layer against environmental stress. 3-Deoxy-d-manno-oct-2-ulosonic acid (Kdo) is a highly conserved monosaccharide that resides in the inner core region of LPS and that connects the lipid A moiety to the extending polysaccharide chain through the hydroxyl group on its C-5 position. Due to its central function in LPS, we hypothesized that metabolically incorporated Kdo derivatives modified on the C-5 position may impair LPS synthesis and therefore lead to a reduced outer membrane integrity. To test this, we successfully synthesized four Kdo derivatives, 5-epi-Kdo, 5-deoxy-Kdo, 5-epi-Kdo-8-N₃, and 5-deoxy-Kdo-8-N₃, and incubated Escherichia coli (E. coli) strains in the presence of these derivatives to investigate their influence on LPS production and labeling. Interestingly, while the 5-deoxy derivatives were not incorporated, 5-epi-Kdo-8-N₃ was successfully incorporated in cell envelope-associated LPS and increased the sensitivity of the bacteria to vancomycin, indicating that 5-epi-Kdo-8-N₃ incorporation in LPS interferes with outer membrane integrity.

Introduction

The outer membrane (OM) of Gram-negative bacteria is one of the most important features that protects bacteria against antimicrobial agents because it preserves the cellular integrity and acts as a potent permeability barrier. A major factor contributing to this strong barrier are the lipopolysaccharides (LPS), , which are found in virtually all Gram-negative bacterial membranes. LPS comprise approximately 75% of the outermost membrane leaflet in Escherichia coli (E. coli) , and are therefore a highly interesting target for antibacterial strategies.

LPS are glycolipids composed of a membrane-embedded core called lipid A, which is a disaccharide that carries multiple lipid chains and phosphate groups. Lipid A is extended by a large hydrophilic polysaccharide chain on the outer surface of the cell to constitute full-length LPS. Adjacent LPS structures are able to form a tight network through electrostatic interactions facilitated by divalent cations, such as Mg²⁺ and Ca²⁺, which stabilizes the outer membrane and decreases membrane permeability. ,, As a result, the tightly packed LPS layer forms a physical barrier surrounding the bacterial cell, which acts as a defense against external stress and greatly limits the diffusion of hydrophobic antimicrobial agents. ,,

The full length LPS structure consists of three domains (Figure A): (1) the lipid A region, embedded into the OM, (2) a core oligosaccharide region that can be further divided into the inner and outer core, and (3) the O-antigen, which is a polysaccharide chain extending out from the surface. LPS can exist as both smooth (high molecular weight) or rough (low molecular weight) forms, depending on the presence or absence of the O-antigen, as a result of adaptation to environmental stress stimuli, such as temperature changes, varying nutrient concentrations, or insults by the host-immune system. , In both smooth and rough LPS, the lipid A and the inner core glycan, containing 3-deoxy-d-manno-oct-2-ulosonic acid (Kdo) units, are conserved with only small variations in the oligosaccharide length and position of lipid chains. The E. coli K-12 derivative BW25113, a popular laboratory strain, produces a rough LPS structure with eight monosaccharide residues attached to the C-5 position on the first Kdo unit which is directly connected to the lipid A region (Figure B, C-5 attachment point highlighted in green). −

1.

(A) Schematic representation of rough- and smooth-LPS structures of Gram-negative bacteria. (B) Generalized structure of the lipid A-core region of (rough) LPS from enterobacterial species, including E. coli BW25113 (K-12 strain), with an approximate molecular weight of 3806 Da. The C-5 attachment point on the first Kdo unit (Kdo1) on the lipid A region is highlighted in green.

The central role of LPS in bacterial integrity and pathogenicity has inspired many strategies aimed at reducing LPS production and weakening its integrity. For example, chelating compounds like ethylene-diamino-triacetic acid (EDTA) or cationic peptides like polymyxin antibiotics , were used to target the tight LPS–cation network to increase OM permeability. The treatment of pathogenic E. coli O111:B4 cells with 5 mM EDTA resulted in the normally impermeable cell membrane becoming permeable to the antibiotic actinomycin. In another study, the well-known polymyxin-type antibiotic colistin was found to damage the OM of E. coli by electrostatically interacting with the anionic phosphate groups on lipid A. In turn, this resulted in an increased sensitivity to rifampicin, an OM-impermeable antibiotic. , Other strategies to target LPS pursued the inhibition of the enzymes responsible for the LPS biosynthesis and transport. For instance, the deacetylase enzyme LpxC, involved in lipid A synthesis, was successfully inhibited using N-aroyl-l-threonine hydroxamic acid, CHIR-090, resulting in cell death of E. coli and Pseudomonas aeruginosa. Another LpxC inhibitor, a phenyl diacetylene-based drug (LPC-009), exhibited inhibition of E. coli LpxC with K _i of 0.18 ± 0.03 nM and an increased potency against various Gram-negative bacteria (2–64-fold over CHIR-090). , The heptosyl transferase WaaC was inhibited using glycosylated (heptosylated) fullerenes with low micromolar inhibition levels (IC₅₀ 11–47 μM). Furthermore, the 2β-deoxy-Kdo derivative, which lacks the hydroxyl group on the C-2 position, was found to be a competitive potent inhibitor of E. coli KdsB, the cytidine monophosphate-Kdo synthase (K _i = 2.03 ± 0.08 μM). − More recently, a new class of macrocyclic peptides was reported to block LPS transport by binding to the LPS-bound transport complex and trap the newly synthesized LPS in the inner membrane.

Intriguingly, bacteria themselves can also be exploited to incorporate modified building blocks that alter the LPS structure. Through a process called metabolic glycan engineering (MGE), unnatural monosaccharides can be incorporated using the bacteria’s own metabolic enzymes. − For instance, uridine 5′-diphosphate-N-acetyl-glucosamine (UDP-GlcNAc) derivatives modified on the C-6 position were developed to inhibit the glycosyltransferase complex PgaCD in E. coli. By introduction of an unnatural GlcNAc salvage pathway through MGE, the polymerization of poly-N-acetyl-glucosamine (PNAG) on the cell membrane was successfully blocked. In Helicobacter pylori, the glycoprotein biosynthesis was metabolically inhibited using thioglycoside (S-glycoside) derivatives of rare bacterial carbohydrates, such as di-N-acetyl-d-bacillosamine and N-acetyl-d-fucosamine, which were discovered to have enhanced selectivity compared to their O-glycoside counterparts.

Due to its central location in the LPS structure, Kdo has received much attention as a target for investigating LPS biosynthesis. An attractive method to study and visualize Kdo incorporation is using Kdo derivatives bearing a reactive handle in a MGE strategy. In previous publications, we and others demonstrated the successful uptake of the azide-containing Kdo derivative Kdo-8-N₃ via the salvage pathway and its metabolic incorporation into the extracellular LPS. ,−

Here we report our in-depth investigations to incorporate Kdo derivatives through MGE to impact LPS biosynthesis and weaken the bacterial LPS shield. By modifying the C-5 position on Kdo, where generally the core oligosaccharide is extended with the O-antigen, we hypothesized that LPS may not be fully synthesized. In addition, when successfully transferred to the outer membrane, the truncated LPS structures may result in impaired packing of the LPS conjugates, leading to a higher membrane permeability. We therefore developed straightforward synthetic procedures to synthesize 5-epi-Kdo (3-deoxy-d-altro-oct-2-ulosonic acid) and 5-deoxy-Kdo (3,5-dideoxy-d-arabino-octulopyranosylonic acid) and their respective 8-azido variants. The panel of Kdo derivatives was screened for their uptake and incorporation by nonpathogenic E. coli BW25113 and a selection of pathogenic clinical E. coli isolates. The 5-epi Kdo derivatives were successfully incorporated. However, no significant differences in LPS abundance and length distribution were detected, suggesting only minimal interference with LPS biosynthesis by these Kdo derivatives. Intriguingly, at low temperatures, incorporation of 5-epi derivatives seemed to destabilize the outer membrane of E. coli BW25113.

Results

Chemical Synthesis of Kdo Derivatives

To prevent extension of Kdo, we designed Kdo derivatives in which the C-5 hydroxyl was inverted (1, 5-epi-Kdo) or completely removed (2, 5-deoxy-Kdo) (Scheme ). The 8-azido-containing counterparts (compounds 3 and 4) were also produced to enable a straightforward visualization of their incorporation in the extracellular LPS.

1. Synthesis of 5-epi-Kdo (1), 5-Deoxy-Kdo (2), 5-epi-Kdo-8-N₃ (3), and 5-Deoxy-Kdo-8-N ₃ (4)­ .

^a Reagents and conditions: i. oxaloacetate, NaHCO₃, NaOH, H₂O, pH 13, rt; ii. NiCl₂, Amberlite-H⁺, pH 5, 50 °C (1: 33% (altro:allo 3:1), 2: 54% (arabino:ribo 3:1), 3: 28% (altro:allo 4:1), over two steps); iii. 0.18 N HCl in MeOH, 0 °C to rt; iv. I₂, imidazole, PPh₃, THF, 0 to 65 °C, overnight (S2: 77%, S6: 50%, over two steps); v. NaN₃, DMF, 80 °C, overnight; vi. Amberlite-H⁺, H₂O, pH 4, 70 °C (S4: 53%, S8: 32%, over two steps); vii. oxaloacetate, NaHCO₃, NaOH, H₂O, pH 14, rt; viii. NiCl₂, AcOH, pH 5, 50 °C (4: 36% over two steps).

The synthesis of compounds 1–4 was based on our previous synthetic strategy, which included the [3 + 5] aldol addition reaction between a d-pentose and oxaloacetate. For the 5-epi derivative (1, 5-epi-Kdo), the reaction was performed on d-ribose and for the 5-deoxy derivative (2, 5-deoxy-Kdo) on 2-deoxy-d-ribose. Both reactions were performed using the Cornforth methodology, and Kdo derivatives 1 and 2 were obtained in moderate yields. The aldol-addition step proceeded with good stereoselectivity for the desired C4-OH equatorial products with the altro and arabino configuration for 1 and 2, respectively. For the 5-deoxy-Kdo (2), a major byproduct was observed that was characterized to be the 4,8-anhydro-derivative presumably formed during the acidic decarboxylation step (see the Experimental Procedures for more details).

To synthesize the 8-azido Kdo derivatives 3 and 4, the azide groups were installed on the respective d-pentoses, similar to the synthesis of Kdo-8-N₃ (Scheme , see Supporting Information for more details). First, the methyl acetals of commercial d-ribose and 2-deoxy-d-ribose were prepared to lock the ring in the 5-membered conformation, followed by the iodination of the primary alcohols. Iodination was preferred over tosylation of the C-5 hydroxyl, as a low selectivity and overtosylation were observed even when 1 equiv of TsCl was used (data not shown). The primary iodide was then substituted by an azide, and subsequent acidic acetal hydrolysis yielded 5-azido-5-deoxy-d-ribose (compound S4) and 5-azido-2,5-dideoxy-d-ribose (compound S8), respectively. Finally, compounds S4 and S8 were subjected to the aldol reaction with oxaloacetate followed by decarboxylation, which gave the final compounds 5-epi-Kdo-8-N₃ (3) and 5-deoxy-Kdo-8-N₃ (4). For compound 3 an altro:allo-ratio of 4:1 was obtained, whereas for compound 4 only the arabino-configured product was characterized.

E. coli BW25113 Grows Well in the Presence of Synthetic Kdo Derivatives

Before investigation of the impact of the synthesized Kdo derivatives on bacterial LPS biosynthesis, the effect of compounds 1–4 on bacterial growth was assessed. E. coli BW25113 was selected as a benchmark E. coli strain because it has been shown to successfully incorporate Kdo-8-N₃ into its truncated LPS. ,,

E. coli BW25113 was inoculated into the nutrient-rich Lysogeny Broth (LB) or minimal M9 medium with or without supplementation of the Kdo derivatives 1–4. As shown in Figure S1, E. coli BW25113 grew to a similar extent in the presence or absence of the Kdo derivatives in LB, although the addition of 5-epi-Kdo resulted in slightly increased growth. In M9 medium, the growth was overall slower than in LB, and cultures reached a lower optical density at 600 nm (OD₆₀₀) at a stationary phase (OD₆₀₀ of 0.514).

However, the addition of the Kdo derivatives to the M9 medium resulted in faster growth and a higher OD₆₀₀. This could be explained by the potential capacity of E. coli cells to metabolize the exogenous Kdo derivatives by an endogenous Kdo aldolase (EC 4.1.2.23). ,

5-epi-Kdo-8-N₃ Is Incorporated into E. coli BW25113 LPS

Having established that the Kdo derivatives were well-tolerated by E. coli BW25113 in both LB and M9 media, the metabolic incorporation of azide-containing compounds 5-epi-Kdo-8-N₃ (3) and 5-deoxy-Kdo-8-N₃ (4) into cellular LPS was assessed. Freshly inoculated E. coli BW25113 cultures in LB or M9 medium were supplemented with 1 mM, 5 mM, or 10 mM of compounds 3 and 4, and 5 mM Kdo-8-N₃ was included as positive control. After incubation, bacteria were collected by centrifugation and treated with fluorescein-dibenzo-cyclooctyne (FAM-DBCO). After lysis, the samples were separated on a 16% Tris-Tricine SDS-PAGE gel; the labeled bands were visualized using fluorescence imaging; and the total LPS was visualized using a silver staining protocol (Figure ).

2.

Labeling of E. coli LPS with 5-epi-Kdo-8-N₃ and 5-deoxy-Kdo-8-N₃ (4). E. coli BW25113 grown in A–C: rich LB medium or D–F: M9 minimal medium, with 5 mM Kdo-8-N₃ (positive control), 1, 5, or 10 mM 5-deoxy-Kdo-8-N₃, 1, 5, or 10 mM 5-epi-Kdo-8-N₃ (3), or without any Kdo derivative (negative control). A, D: Fluorescence images taken in the green channel. B, E: Gels scanned after silver staining. C, F: Superimposed images of the fluorescence images (panels A or D) and the silver-stained gels (panels B or E). G: Relative fluorescence (%) of LPS labeling on E. coli BW25113 using a concentration range of 5-epi-Kdo-8-N₃ in LB (red) or M9 (blue) medium compared to using 5 mM Kdo-8-N₃ (set to 100%) based on SDS-PAGE. Lane 1: protein ladder; Lane 2: no additive; Lane 10: Kdo₂-lipid A standard.

Interestingly, addition of 5-epi-Kdo-8-N₃ (3) clearly resulted in successful labeling of LPS (Figure A and D, lanes 7–9), whereas 5-deoxy-Kdo-8-N₃ (4) did not (lanes 4–6). Comparison of the band intensities revealed that the relative fluorescence of labeled LPS with 5-epi-Kdo-8-N₃ (3) was higher when grown in LB compared to M9 medium (Figure G), which correlates with our earlier observations with Kdo-8-N₃. However, the labeling with 10 mM compound 3 was less efficient than with 5 mM Kdo-8-N₃.

We were intrigued to find that neither Kdo derivative seemed to have an effect on the size of the LPS produced. Figure shows clearly that the labeled LPS structure still consists of the standard core region (as depicted in Figure B), instead of the truncated structures more similar to Kdo₂-lipid A (lane 10; lipid A structure carrying two Kdo units).

Labeling of E. coli LPS Is Not a Result of 5-epi-Kdo-8-N₃ Epimerization

To understand why incorporation of the C-5-epi derivative did not impact the LPS size, we considered several explanations. One possible explanation is that 5-epi-Kdo-8-N₃ (3) is epimerized to the original C-5 axial Kdo, which then would allow for standard enzymatic extension again to attach the core region. For this to occur, we considered a direct enzyme-catalyzed epimerization on compound 3; however, a dedicated search of the E. coli genome and the literature did not reveal any potential Kdo C-5 epimerase. Alternatively, as depicted in Scheme , several enzymatic steps may together result in a net inversion of configuration at C-5: compound 3 could first be disconnected through a retro-aldol reaction (catalyzed by an aldolase, EC 4.1.2.23), after which the resulting 5-azido-d-ribose would undergo epimerization at C-2 to produce 5-azido-d-arabinose (catalyzed by Rib5P and Ara5P isomerases), which subsequently could be transformed into Kdo-8-N₃ by Kdo8P synthase, KdsA (EC 2.5.1.55). ,

2. Proposed Cellular Processing of 5-epi-Kdo-8-N₃ (3) to Form Kdo-8-N₃ .

^a Proposed enzymes in order: Kdo aldolase (EC 4.1.2.23), ribose-5-phosphate (Rib5P) isomerase RpiB (EC 5.3.1.6), arabinose-5-phosphate (Ara5P) isomerase KdsD (EC 5.3.1.13), and Kdo-8-phosphate (Kdo8P) synthase KdsA (EC 2.5.1.55).

To investigate the second part of this hypothesis, we envisaged that supplementing E. coli BW25113 with 5-azido-d-ribose (Rib-5-N₃) would allow the C-2 epimerization and subsequent Kdo-8-N₃ production and incorporation into the LPS. Labeling experiments were performed in LB or M9 media with 5 mM Rib-5-N₃, 5 mM Ara-5-N₃ (which may be converted to Kdo-8-N₃), and 5 mM 2-deoxy-Rib-5-N₃ (which may be converted to 5-deoxy-Kdo-8-N₃), including 5 mM Kdo-8-N₃ as a positive control.

No labeling was observed with Rib-5-N₃ (Figure A, lanes 4 and 7), suggesting that the route in Scheme , including the retro-aldol, epimerization, and aldol reaction, is not likely to occur, ruling out the possibility of 5-epi-Kdo-8-N₃ labeling through C-5 epimerization.

3.

Labeling of E. coli LPS with Ara-5-N₃ and Rib-5-N₃. E. coli BW25113 cells were grown in LB or M9 medium, with 5 mM Ara-5-N₃ or 5 mM Rib-5-N₃ or without any additional substrate (negative control). A: Fluorescence image taken in the green channel. B: Gels scanned after silver staining. C: Superimposed images of the fluorescence images (panel A) and silver stained gels (panel B). A–C: Lane 1: E. coli O55:B5 LPS standard; Lane 2 and 5: no additive. See Figure S2 for the complete gel images.

Intriguingly, Ara-5-N₃ supplementation actually resulted in LPS labeling, with higher intensity observed in LB (Figure A, lane 3) compared to M9 (lane 6), but the intensity was much lower than that achieved with Kdo-8-N₃ (Figure S2). As the E. coli BW25113 LPS structure does not include an arabinose residue (Figure B), we envisage that the fluorescent labeling arose from the metabolic conversion of Ara-5-N₃ to Kdo-8-N₃. Additionally, no fluorescent bands were detected in the samples of bacteria incubated with 2-deoxy-Rib-5-N₃ (Figure S2).

5-epi-Kdo-8-N₃ Labels the Cell Envelope

To verify that the fluorescent signal from LPS labeling with 5-epi-Kdo-8-N₃ (3) originated from LPS that is incorporated in the bacterial cell surface, the labeled cells were examined using fluorescence microscopy.

As shown in Figure , peripheral fluorescence was observed, confirming that the LPS molecules on the outer cell membrane were labeled. The fluorescence intensity of cells labeled with compound 3 increased with increasing concentrations (Figure S3) but was overall lower than observed for cells labeled with 5 mM Kdo-8-N₃ (positive control). Furthermore, the labeling in M9 medium (Figure , D3) was less efficient than that in LB medium (Figure , A3), which was also corroborated by fluorescence gel imaging (Figure S3).

4.

5-epi-Kdo-8-N₃ labels the cell envelope of E. coli BW25113. Panels A, B, and C: cells grown in LB; panels D, E, and F: cells grown in M9 medium. Cells were visualized using a Nikon Ti-E microscope with fluorescence optics (A1–A3, D1–D3: green channel) or phase contrast (B1–B3, E1–E3). Colocalization of fluorescence and the bacterial cell envelope is visualized by merging the fluorescence signal and phase contrast images (C1–C3, F1–F3: merged channels).

Even though 5-epi-Kdo-8-N₃ (3) was not incorporated as efficiently as Kdo-8-N₃, this derivative was still effectively processed and is able to label the surface LPS to a certain extent.

5-epi-Kdo-8-N₃ Impacts Outer Membrane Integrity in E. coli BW25113

Having established that 5-epi-Kdo-8-N₃ (3) was successfully incorporated into the LPS of E. coli BW25113, yet without altering the size of LPS as determined by electrophoresis, we further investigated whether 5-epi Kdo derivatives impact the integrity of the outer membrane (OM), including Kdo-8-N₃ as a control. First, we explored the permeability of the E. coli BW25113 cell envelope when incubated with these derivatives using the LIVE/DEAD BacLight assay. , With this assay, viable cells with intact inner and outer membranes are stained green, while dead cells with compromised membranes are stained red. The microscopy images suggested that 5-epi-Kdo-8-N₃ incorporation into LPS did not significantly affect the bacterial viability and permeability of the cell envelope under these conditions (Table S1).

Subsequently, we conducted an antimicrobial susceptibility assay to specifically evaluate the OM integrity, using vancomycin which typically does not penetrate the OM of Gram-negative bacteria. The cells were incubated at 20 or 37 °C with varying concentrations of vancomycin (0–64 μg/mL) and 10 mM Kdo derivatives. The change in absorbance (ΔOD₆₀₀) after 24 h of incubation was used as a measure of antibiotic killing after membrane permeation by vancomycin (Figure ). At 37 °C, the OM permeability for vancomycin did not detectably change upon supplementation of the growth medium with any of the Kdo derivatives (Figure S4). However, at 20 °C, increased vancomycin sensitivity was observed for bacteria treated with 5-epi-Kdo (1) (Figure ). Interestingly, the cells incubated with 5-epi-Kdo-8-N₃ (3) at 20 °C did not grow at all, even without vancomycin supplementation. Given that the E. coli cells are known to have an altered OM integrity and sensitivity to vancomycin at low temperatures, the observed growth arrest and increased vancomycin sensitivity suggest a clear effect of the Kdo C-5 modifications on the LPS structure and outer membrane integrity.

5.

Incubation of E. coli BW25113 cells at 20 °C in the presence or absence of 10 mM of Kdo derivatives with varying concentrations of vancomycin (0–64 μg/mL). Change in OD₆₀₀ values after 24 h (n = 3) incubation with 10 mM Kdo-8-N3, 5-epi-Kdo (1), 5-epi-Kdo-8-N3 (3), or no Kdo derivative.

5-epi-Kdo-8-N₃ Labels LPS in Pathogenic E. coli

With the knowledge that 5-epi-Kdo-8-N₃ (3) is effectively incorporated into LPS from E. coli BW25113 while 5-deoxy-Kdo-8-N₃ (4) was not, we investigated whether these derivatives are processed by different E. coli serotypes and clinical isolates (Table S2). For this, six pathogenic E. coli strains were used, including two commercial strains (E. coli O55:B5, E. coli O119), an engineered strain (E. coli ET8), and three clinical E. coli isolates (SOR-A, SOR-B, SOR-E), which previously had shown varying levels of Kdo-8-N₃ incorporation.

Compound 3 was successfully incorporated into E. coli ET8, SOR-A, and SOR-B, but the incorporation efficiency varied greatly between the strains (Figure ). While the ET8 and SOR-A strains exhibited a higher extent of labeling in the LB medium (Figure A, lanes 6 and 7), the SOR-B strain displayed a more pronounced fluorescent band in the M9 medium (Figure B, lane 8). LPS labeling was predominantly observed in the lower molecular weight region (core-lipid A). However, higher molecular weight fluorescent bands were also faintly visible for the O55:B5 and SOR-E strains (Figure S5), which might correspond to labeled smooth LPS. The labeled LPS consistently contained a core region, demonstrating that the basic structure of the LPS was retained. Furthermore, 5-deoxy-Kdo-8-N₃ (4) did not lead to any LPS labeling among the pathogenic E. coli strains (Figure S6).

6.

E. coli cells were grown in A, C: rich LB medium or B, D: M9 minimal medium with 5 mM Kdo-8-N₃ (positive control) or 10 mM 5-epi-Kdo-8-N₃ (3), and total LPS was separated on 16% Tris-Tricine SDS-PAGE. A, B: Fluorescence images taken in the green channel. C, D: Gels scanned after silver staining. Lane 1: O55:B5 LPS standard; Lane 10: Kdo₂-lipid A standard.

Effect of Kdo Derivatives on the LPS Structures

As we previously observed that labeling with Kdo-8-N₃ impacted the size distribution and abundance of LPS structures in different media, we analyzed the overall LPS compositions of pathogenic E. coli strains upon growth with the Kdo derivatives in both LB and M9 medium (Figure ).

7.

Impact of Kdo derivatives 1–4 on LPS production in pathogenic E. coli strains. Total LPS was separated on 16% Tris-Tricine SDS-PAGE and visualized by silver staining. E. coli cells were grown in the rich LB medium (left panels) of M9 minimal medium (right panels) with or without 10 mM (A) 5-epi-Kdo-8-N₃ (3), (B) 5-deoxy-Kdo-8-N₃ (4), (C) 5-epi-Kdo (1), and (D) 5-deoxy-Kdo (2).

Incubation of E. coli O119 with azide-containing Kdo derivatives 3 and 4 resulted in reduced abundance of LPS with various molecular weights in both LB and M9 media (Figure A and B, lanes 6–7). A similar effect was observed with ET8 (lanes 8–9) and SOR-E (lanes 14–15) but only when incubated in M9 with 5-epi-Kdo-8-N₃ (3) (Figure A). On the contrary, the SOR-A strain incubated in LB with 5-deoxy-Kdo-8-N₃ (4) appeared to produce LPS in relatively higher abundance (Figure B, lanes 10–11).

Knowing that the azide may negatively affect the efficiency of metabolic processing of Kdo-8-N₃, we investigated whether the derivatives without the azides, 5-epi-Kdo (1) and 5-deoxy-Kdo (2), impacted the biosynthesis of LPS in the selected E. coli strains. Interestingly, 5-epi-Kdo supplementation resulted in an increase of low molecular weight LPS in some E. coli strains, such as O55:B5, ET8, SOR-A, SOR-B, and SOR-E (Figure C, lanes 4–5, 8–15). Additionally, new LPS bands with a molecular weight similar to that of Kdo₂-lipid A have appeared in nearly all pathogenic E. coli strains (Figure C, lanes 4, 6, 10, 12, and 14), suggesting that 5-epi-Kdo interfered with LPS biosynthesis, leading to shorter LPS production.

As expected from the lack of fluorescence labeling with its azide-containing counterpart, 5-deoxy-Kdo had almost no effect on LPS synthesis in pathogenic E. coli (Figure D), with the exception of E. coli O119 strain, which showed a slight change in the distribution of LPS (Figure D, lanes 6–7), similar to the effect of compound 3 (Figure B). The effects observed with 5-deoxy-Kdo derivatives 2 and 4 are much less pronounced, and together these results highlight the importance of the C-5 hydroxyl for metabolic processing by the E. coli LPS biosynthesis enzymes.

5-epi Derivatives Are Processed by KdsB

Our results indicate that 5-epi derivatives are incorporated in the LPS, whereas 5-deoxy derivatives are not. As the Kdo derivatives need to be processed by the biosynthetic machinery of the cell in order to be incorporated effectively, it is likely that at least one enzyme of the LPS pathway discriminates among these derivatives. Upon intake, the first metabolic enzyme is KdsB (EC 2.7.7.38) that catalyzes the reaction between Kdo and cytidine triphosphate (CTP) to synthesize CMP-Kdo. − To assess the activity of KdsB on the internalized Kdo derivatives, we overexpressed and purified the enzyme and performed an in vitro assay that determines conversion by monitoring released inorganic pyrophosphate (PPi) using the colorimetric eikonogen reagent (EK) assay.

In agreement with earlier reports, Kdo and Kdo-8-N₃ at 1 mM substrate concentration were shown to be substrates for KdsB (100 nM), with the first being more efficiently processed (Figure A). ,, However, under these reaction conditions, no conversion was observed for Kdo derivatives 1–4. When the concentration of compounds 1–4 was 10-fold increased and reaction time was extended to 24 h, KdsB activity was observed with both 5-epi compounds 1 and 3 (Figure B). In contrast, KdsB did not reveal any activity toward the 5-deoxy compounds 2 and 4, suggesting that the lack of a C-5 hydroxyl group negatively impacts enzyme catalysis.

8.

KdsB activity assays on Kdo derivatives: 100 nM KdsB enzyme was incubated with 3 mM CTP and (A) 1 mM Kdo derivative (Kdo, Kdo-8-N₃, compounds 1–4) or (B) 10 mM of C-5 modified Kdo derivative (compounds 1–4). Error bars indicate standard deviations (n = 3).

Molecular Origin of the Impaired KdsB Activity on 5-epi-Kdo and 5-deoxy-Kdo

The clear differences between KdsB catalysis of the 5-epi and 5-deoxy Kdo derivatives inspired us to investigate the underlying molecular cause. Using the crystal structure of E. coli K-12 KdsB in complex with the inhibitor 2-deoxy-Kdo and CTP (PDB: 3K8D), we generated models of the complex of KdsB with Kdo, as well as with derivatives 1 and 2 in the active site. After preparing the structure for computational analysis (see Methods), we manually added the missing hydroxyl group on C-2, added hydrogen atoms on all positions, and predicted the hydrogen bonding (H-bond) network using YASARA.

For Kdo, the resulting complex at pH 10 revealed tight coordination by an intricate network of hydrogen bonds within the active site of KdsB (Figure A). The reactive hydroxyl oxygen on C-2 is situated near the Mg²⁺-coordinated alpha phosphate of the CTP cosubstrate, while the OH on C-5 is stabilized by hydrogen bonds to glutamine (Q98) and tyrosine (Y185) (Figure B, green arrows). We included in our models a second Mg²⁺ ion, previously postulated to be present in the active site (Figure B, orange arrow), which coordinates to the residues D253 and D100 and possibly the C-1 oxygen. , When we repeated the prediction at a lower pH of 7.5, we observed protonation of Y184, causing the optimal network to be established when Q98’s amide group is flipped (Figure S7). As both conformations are plausible, it is possible that the Q98 amide rotation (and thus binding) is in fact pH-dependent, which could explain the difference in reported K _M values for KdsB with Kdo at neutral and basic pH. ,

9.

Active site binding of Kdo derivatives in KdsB with CTP. (A) Schematic representation of Kdo interactions in the active site of KdsB at pH 10. (B) Docking of 2-deoxy-Kdo with added OH at C-2, hydrogens, and a predicted H-bond network. Seven active-site residues (sticks with gray carbons) bind tightly to Kdo (ball and sticks with blue carbons), and the C-2 OH is positioned near the Mg²⁺ (yellow sphere) coordinated to the CTP cosubstrate (ball and sticks with cyan carbons). The C-5 OH forms H-bonds to Q98 and Y185. (C) Modeling of 5-epi-Kdo (1) in the active site leads to the loss of the H-bond between the C-5 oxygen with the side chain of Q98 and two unsatisfied donor/acceptor groups. (D) Modeling of 5-deoxy-Kdo (2) in the active site shows a disruption of two H-bonds.

We next modeled the 5-epi-Kdo (1) (Figure C) by manually altering the configuration of the C-5 OH and repredicting all hydrogen bonds. We found that the optimal network is essentially unchanged, except that the C-5 OH can now only H-bond via its hydrogen to Y185, whereas the H-bond to Q98 is lost, leaving two unsatisfied hydrogen bonds at Q98 and the Kdo derivative. The resulting impaired binding energetics can thus explain the reduced activity of KdsB with this compound. Finally, our third model shows that in the case of 5-deoxy-Kdo (2), in addition to the loss of interaction with Q98, the H-bond to Y185 is also lost, and both residues’ side chains now feature energetically unfavorable unsatisfied hydrogen bonds (Figure D). This model suggests a further diminishing of the binding affinity of KdsB, and it explains why the enzymatic reaction with this compound is hampered. These findings explain the lower affinity of KdsB toward the 5-epi-Kdo (1), necessitating the higher substrate concentrations in the reactions and the inactivity toward 5-deoxy-Kdo (2).

Discussion

LPS is an integral component of the OM of Gram-negative bacteria, with great potential as a target to disrupt the structural integrity of the membrane. We hypothesized that C-5 modified Kdo derivatives would result in a truncated LPS structure and therefore increased membrane permeability, rendering Gram-negative bacteria more susceptible to antibiotics. Interestingly, although the LPS structure of the investigated E. coli strains was not significantly truncated as anticipated, the 5-epi-Kdo-8-N₃ (3) derivative was successfully incorporated into the LPS in several strains, and 5-epi derivatives (1 and 3) influenced the OM integrity of E. coli BW25113 at low temperatures.

Supplementation of the growth medium of E. coli BW25113 with 5-epi-Kdo-8-N₃ (3) led to LPS labeling without disrupting the synthesis of the core region. We established that this was not the result of epimerization of the C-5 hydroxyl group within the cells via the breakdown of compound 3 into its metabolite, Rib-5-N₃. Serendipitously, we discovered that Ara-5-N₃ supplementation resulted in effective LPS labeling, likely by being used to synthesize Kdo-8-N₃ in the cells prior to incorporation into LPS. This implies, against all expectations, that the Kdo8P synthase KdsA, which normally requires arabinose-5-phosphate for enzyme activity, , is able to accommodate an azide group on the C-5 position of arabinose. Future in vitro studies with isolated KdsA will further verify the hypothesis of substrate promiscuity.

Incorporation of 5-epi-Kdo into the LPS did not result in the expected truncation and lower molecular weights. To explain this finding, we considered that the incorporated 5-epi Kdo unit may be extended on its equatorial C-5 hydroxyl or possibly even on another position than the C-5. However, since reported examples of structurally characterized LPS all describe the core extension to appear on the axial C-5 position, ,,− we deemed this option highly unlikely. Finally, we concluded that it is most likely that 5-epi-Kdo-8-N₃ is incorporated as the second Kdo unit (Kdo2, Figure B). The E. coli BW25113 LPS structure contains two Kdo units, and given the fact that constitutive production of natural Kdo is expected to be unaffected by the derivatives, it is likely that natural Kdo, with the axial C-5 hydroxyl, is incorporated as the Kdo1 unit. A possible explanation for this is that the bifunctional E. coli KdtA transferase enzyme, which is responsible for the addition of both Kdo units, may not accommodate the altered C-5 configuration for the first Kdo transfer onto lipid IV_a, which is the tetraacylated precursor of lipid A, while the second Kdo addition is more promiscuous. Although the crystal structure of E. coli KdtA has not been reported, a study comparing the bifunctional E. coli KdtA and the monofunctional Haemophilus influenzae KdtA enzyme highlights possible active site residues involved in the binding of the Kdo residues. In vitro biochemical reactions with KdtA and CMP-5-epi-Kdo will reveal their substrate specificity, and extensive NMR studies on isolated LPS fractions may provide the proof of incorporation of 5-epi-Kdo as the Kdo2 unit. Studies with other Gram-negative bacteria, such as H. influenzae, which possesses a monofunctional transferase KdtA, may also elucidate the placement of the derivatives in the LPS depending on whether these derivatives are processed or not. After KdtA-catalyzed Kdo transfer, the subsequent acyltransferases LpxL (EC 2.3.1.241) and LpxM (EC 2.3.1.243) and the flippase MsbA are needed to transport the modified LPS (sub)­structure onto the cell membrane. If the 5-epi derivatives are incorporated on the lipid IV_a as the first Kdo unit, it is possible that the acyltransferases may be unable to continue the addition of acyl chains necessary to form lipid A. These acyl chains are crucial for recognition and processing by the MsbA flippase in order to be transported on the outer cell membrane. − Enzyme activity assays with MsbA and Lpx enzymes using substrates consisting of C-5 modified Kdo derivatives can shed light on the mechanism of incorporation. Taken together, although the exact mechanism of incorporation is currently unclear, our studies reveal that incorporation of the Kdo derivatives likely happens at the Kdo2 position, which offers the opportunity to attach (sterically challenged) functional cargo to the LPS, for instance, to study the fate of LPS, downstream effects, and binding partners.

When comparing LPS labeling between pathogenic E. coli strains, we observed that E. coli isolates SOR-A and ET8 showed LPS labeling with 5-epi-Kdo-8-N₃ (3) in higher intensity when incubated in LB medium, while SOR-B showed higher intensity when incubated in M9 medium. Furthermore, differences in LPS composition and amounts were seen when comparing the effect of Kdo derivative supplementation across the pathogenic E. coli strains. As these results may originate from differences in the activity of the metabolic enzymes, we sequenced and compared the SOR-A and SOR-B genomes from these clinical isolate strains. We found that the two strains were genetically nearly identical, differing by only one gene target, the zinc-binding GTPase YeiR. Furthermore, we compared the genes for KdsB (CMP-Kdo synthase, EC 2.7.7.38), KdtA (CMP-Kdo transferase, EC 2.4.99.12), and MsbA (lipid A-core flippase, EC 7.5.2.6), between E. coli BW25113 (accession number CP009273.1), SOR-A, and SOR-B, but no major differences in the respective gene sequences (kdsB, kdtA, msbA) were observed (Figures S8, S9, and S10). Interestingly, the sequences of E. coli SOR-A and SOR-B revealed a gene annotated as kpsU, which encodes for a CMP-Kdo synthase homologue that is absent from E. coli BW25113. Although this KpsU enzyme catalyzes the same reaction as E. coli KdsB (EC 2.7.7.38), this enzyme is predicted to be involved in capsular polysaccharide synthesis and shares only 45% sequence identity (BLAST) with KdsB (Figure S11). Additional assays using KdsB knockout mutants of E. coli strains might provide further valuable insights into whether KpsU activity contributes to the synthesis of the activated Kdo derivatives in these strains. Given that the variations in LPS production and labeling among the investigated strains cannot be explained by the gene alignments, we hypothesize that these differences may be attributed to transcriptional regulation and phase variations between these strains.

Conclusions

Our study reveals that the incorporation of C-5 modified Kdo derivatives into the LPS of E. coli strains did not result in the expected truncation of the LPS structure. However, the successful incorporation of the 5-epi-Kdo-8-N₃ derivative into the LPS has uncovered a significant feature in LPS biosynthesis. The observed LPS labeling and initial observations on KdsB activity suggest that the 5-epi Kdo derivatives can be utilized by the LPS biosynthetic pathway, potentially affecting the outer membrane and susceptibility to antibiotics, especially at relatively low temperatures. Although the underlying mechanism and reasons for the specific incorporation pattern and growth effects remain yet to be determined, our findings provide insights into the importance of the C-5 hydroxyl group’s presence and conformation for biological activity, and they also reveal the structural impact of the Kdo derivatives on E. coli LPS biogenesis. Further studies into individual enzyme activities involved in LPS biosynthesis and into the LPS biosynthetic pathways of other Gram-negative bacteria can shed further light on the mechanisms underlying the present observations.

Experimental Section

General Procedure for the Chemical Synthesis of Kdo Derivatives 1–4

To a solution of NaHCO₃ (0.05–0.1 equiv) in distilled water (0.15–0.80 M) was added 10 M aqueous NaOH at 0 °C until pH 13–14. Oxaloacetic acid (1–1.5 equiv) in distilled water (1.2–2.5 M) and 10 M aq. NaOH (24–31% (v%)) was added simultaneously dropwise, while keeping the pH around 12–14 (as monitored by pH meter). The respective pentose (d-ribose, 2-deoxy-d-ribose, 5-azido-d-ribose, or 5-azido-2-deoxy-d-ribose, 1–1.3 equiv) was dissolved in distilled water (1.6–3.9 M) and added dropwise into the reaction mixture. During the addition the pH was adjusted to 12–13 with 10 M aq. NaOH. The reaction was left to stir at ambient temperature for 2 h. Subsequently, NiCl₂ (0.02–0.1 equiv) was added, and the reaction mixture was heated to 50 °C. Amberlite IR120 (H⁺) resin or acetic acid was added in portions over a period of 1 h until pH was stable at 4–5 for 15 min. The reaction mixture was filtered, and the filtrate was applied to an anion exchange column (freshly prepared HCO₃ ^– or HCO₂ ^– resin). Product was eluted with a gradient of aqueous NaHCO₃ (0.05–0.2 M) or HCO₂H (0.5–2 M) solution. Combined fractions were neutralized using Amberlite IR120 (H⁺), filtered, and then concentrated in vacuo to give Kdo derivatives 1–4. See Supporting Information for further details regarding the individual syntheses.

Bacterial Strains and Growth Conditions

Bacterial strains used in this project are E. coli BW25113 (E. coli K-12 derivative, Leibniz Institute; DSM 27469), E. coli O55:B5 (E. coli O55:K59­(B5), Leibniz Institute; DSM 4779), E. coli O119 (ATCC: The Global Bioresource Center), E. coli ET8 (E. coli ET12567 derivative obtained from Zhu et al.), and clinical isolates of E. coli obtained from the strains collection at the Department of Medical Microbiology and Infection Prevention at the University Medical Center Groningen (UMCG). Clinical E. coli isolates listed here as SOR-A, SOR-B, and SOR-E were obtained from independent blood cultures (see Table S1 for details). All strains were preserved at −80 °C in a lysogeny broth (LB) containing 25% glycerol.

The strains were grown either in LB (containing 10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl, pH 7.5, adjusted using 6 M HCl and 6 M NaOH) or in M9 minimal medium (M9, containing 0.3% KH₂PO₄, 0.6% Na₂HPO₄, 0.5% NaCl, 0.1% NH₄Cl, 2 mM MgSO₄, 0.1 mM CaCl₂, supplemented with 0.2% maltose, sterilized through a filter) as indicated below. Overnight cultures were typically prepared by inoculation of LB medium and incubation at 37 °C at 220 rpm.

Copper-Free Click Reaction

Overnight cultures were inoculated into 1 mL of fresh LB or M9 medium containing 1 or 5 mM of desired Kdo derivative (5-deoxy-Kdo, 5-deoxy-Kdo-8-N₃, 5-epi-Kdo, 5-epi-Kdo-8-N₃) or an equivalent amount of sterile Milli-Q water to an OD₆₀₀ of 0.05. The liquid cultures were incubated at 37 °C at 220 rpm for 16 h, then pelleted at 7000 rpm for 2 min, and washed three times using fresh medium. The bacterial cell pellets corresponding to an OD₆₀₀ of 0.3 were resuspended in fresh M9 medium and treated with the appropriate fluorophore to a final concentration of 0.25 mM in 20 μL. The reaction was performed at 37 °C for 1 h in the dark. Subsequently, the bacteria were pelleted at 7000 rpm for 2 min and washed three times with fresh M9 medium. The resulting bacterial cell pellets were prepared either by SDS-PAGE or fluorescence microscopy.

SDS-PAGE Analysis and LPS Visualization

Sample Preparation

Cell pellets corresponding to an OD₆₀₀ of 0.3 in 1 mL were resuspended in 50 μL of Tricine SDS sample buffer (4% β-mercaptoethanol, 200 mM Tris-HCl, 2% SDS, 0.04% Coomassie Blue in a 40% glycerol solution in Milli-Q), and boiled at 100 °C for 10 min. After the samples were cooled to ambient temperature, Streptomyces griseus protease (Type XIV, ≥ 3.5 units/mg solid, Sigma-Aldrich) was added to the samples to a final concentration of 3.3 mg/mL in a total volume of 60 μL, and the samples were incubated overnight at 55 °C. Next day, the samples were vortexed.

16% Tris-Tricine PAGE for LPS Separation

The loading of samples was adjusted in a way where each well contained a portion of the sample that equals an OD₆₀₀ of 0.025 (for gels consisting of 10 wells) or an OD₆₀₀ of 0.012 (for gels consisting of 15 wells). The gels were run approximately at 20 mA for ∼5 h (corresponding to 30–80 V). The in-gel fluorescence was analyzed on Typhoon FLA 9500 (GE Healthcare) in the green channel green: Ex = 488 nm, Em = 524 nm.

Gel Staining Protocols and Imaging for Total LPS

The gels were stained using a silver staining protocol with slight modifications to the previously described method: The gels were incubated overnight in 50 mL of fixing solution (40% ethanol, 5% AcOH in water) with gentle shaking. Next day, the gels were incubated in oxidizing solution (2.1 g of sodium periodate, 120 mL of ethanol, 15 mL of AcOH in 165 mL of water) for 10 min with gentle shaking and then washed with fresh water for 15 min. The washing step was repeated 2 times. The gels were then incubated in the freshly prepared silver staining solution (151.6 mg NaOH, 1.32 g silver nitrate, 2.9 mL of 25% NH₄OH solution in 200 mL water) for 10 min with gentle shaking and subsequently washed with water 3 times in 15 min intervals. The LPS bands were developed using the developing solution (20 mg of citric acid, 200 μL of 37% formaldehyde solution in 200 mL of water) for 3 to 5 min, and the development was stopped by incubating the gels in 7% AcOH solution. The images of stained gels were taken on a GelDoc EZ Imager (Bio-Rad).

Fluorescence Imaging Microscopy

The bacterial cell pellets corresponding to an OD₆₀₀ of 0.3 in 1 mL were resuspended in 50 μL PBS buffer, and 4–6 μL of each sample was mounted on a 1% agarose pad on a microscopy slide. The samples were imaged using a Nikon Ti-E microscope (Nikon Instruments, Tokyo, Japan) equipped with a Hamamatsu Flash 4.0 camera. Images were acquired with Nikon Instruments Elements 4.10 software and processed using ImageJ 1.52p.

Antibiotic Susceptibility Assay

Antibiotic susceptibility assays were performed on a 96-well plate, using 100 μL as the final volume per well. Overnight cultures were inoculated into 100 μL of fresh LB containing 10 mM of the desired Kdo derivative or an equivalent amount of sterile Milli-Q water and a fixed concentration of vancomycin (0, 4, 16, 32, or 64 μg/mL) to an OD₆₀₀ of 0.003. The prepared plate was then incubated at 20 or 37 °C with continuous shaking at 200 rpm for 24 h. OD₆₀₀ measurements were done in a microplate reader (BioTek^R, Synergy H1) and recorded at 0 and 24 h. The growth analyses were performed using BioTek^R Gen5 Software. Each experiment was repeated in triplicate, and the average of these measurements was used to depict the graphs on Microsoft Excel.

In Vitro Enzymatic Assays with KdsB (CMP-Kdo Synthase)

The reactions were performed in safe-lock microcentrifuge tubes and KdsB activity was assessed in a 96-well plate using the reported eikonogen assay (EK). , The reaction buffer consisted of 100 mM glycine, 5 mM NaOH, and 5 mM MgCl₂, and was adjusted to pH 10. First, 70 μL of Kdo/modified Kdo derivatives and 140 μL of CTP in reaction buffer were added to the corresponding tubes and mixed. Then, 70 μL of KdsB in reaction buffer was added into each tube. Each reaction mixture contained either 1 mM or 10 mM of Kdo/modified Kdo derivative, 3 mM of CTP and 100 nM of KdsB in a total volume of 280 μL. The reaction mixtures were incubated at ambient temperature, and at each time point (0 min, 10 min, 1 h, 2 h, 5 h, 24 h) an aliquot of 40 μL was quenched by mixing with an equal volume (40 μL) of 100% ethanol in a separate tube. Then, 40 μL of each quenched reaction mixture was transferred into a well on a 96-well plate. The enzymatic activity was evaluated by the release of pyrophosphate (PP_i) during the reaction upon coupling of CTP with the sugar substrate using the EK assay reagents). To each well was added 50 μL of ammonium molybdate solution (13.3 μL H₂SO₄, 2.5 g ammonium molybdate tetrahydrate in 100 mL distilled H₂O), and then 70 μL of 2:5 mixture of EK solution (0.25 g sodium sulfite, 14.65 g sodium metabisulfite and 0.25 g 1-amino-2-naphtol-4-sulfonic acid (eikonogen, or EK) in 100 mL hot distilled H₂O) and aqueous β-ME (3.5 mL β-mercaptoethanol in 100 mL distilled H₂O). After thorough mixing of the components in the wells, the plate was incubated at 37 °C for 10 min to allow the color to develop. Then the absorbance was measured at 595 nm on a microplate reader (BioTek^R, Synergy H1). Each reaction was performed in triplicates. The EK assay controls consisted of dilutions of an aqueous Na₄P₂O₇ solution.

Molecular Modeling of Kdo Derivatives and KdsB

The holo enzyme (CTP and 2-deoxy-Kdo bound-) crystal structure of KdsB (PDB: 3K8D) was downloaded from the PDB server. The KdsB protein part was prepared for docking by fixing common crystallography issues and adding hydrogens using YASARA (YASARA commands Clean and OptHydAll). Next, a crystal water (HOH 254 in chain A) was exchanged with a magnesium ion and the missing OH group at the C-2 position of Kdo, hydrogens on the two substrates (CTP and Kdo), and the crystallographic water molecules were added. After the pH was set to 10, OptHydAll was used again to predict the new optimal hydrogen bonding network, which was carefully scrutinized for biophysical logic. An equivalent approach was utilized for the 5-epi-Kdo and 5-deoxy-Kdo derivatives. For each derivative, the optimal network was recalculated, as were alternative networks resulting from the deprotonation of C-2 oxygen of Kdo, and/or the protonation at KdsB enzyme residue histidine 181, and/or the swapping of the crystal water hydrogen-bonding to the aspartic acid 235 for a magnesium ion, to rule out that such a change in the model would disagree with the hypothesis on the interactions at the C-5 position on Kdo presented in the main text.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.5c00338.

• Additional figures and tables, chemical synthesis, materials and methods, and NMR and HRMS spectra (PDF)

Z.S.Z., M.T.C.W., and D.-J.S. conceived and designed the study. Z.S.Z. performed the chemical synthesis, labeling experiments, imaging, and enzymatic assays. I.M.A.B. contributed to the chemical synthesis. W.S.W. performed the expression, isolation, and purification of the KdsB enzyme. M.J.L.J.F. conceived and performed the molecular modeling. J.M.v.D. contributed the clinical isolates. Z.S.Z. and M.T.C.W. wrote the manuscript. All authors approved the final version. CRediT: Zeynep Su Ziylan conceptualization, investigation, methodology, writing - original draft, writing - review & editing; Imke M.A. Bartels investigation, methodology, writing - review & editing; Wahyu S. Widodo investigation, methodology, writing - review & editing; Jan Maarten van Dijl methodology, supervision, writing - review & editing; Maximilian J. L. J. Fürst investigation, methodology, writing - original draft, writing - review & editing; Dirk-Jan Scheffers investigation, methodology, supervision, writing - review & editing; Marthe T. C. Walvoort conceptualization, funding acquisition, supervision, writing - original draft, writing - review & editing.

This work was financially supported by the Faculty of Science and Engineering, University of Groningen. M.J.L.J.F. gratefully acknowledges funding from the COST-Action COZYME (CA21162) from the European Union.

The authors declare no competing financial interest.