Imaging of Escherichia coli K5 and glycosaminoglycan precursors via targeted metabolic labeling of capsular polysaccharides in bacteria

Abstract

The introduction of unnatural chemical moieties into glycosaminoglycans (GAGs) has enormous potential to facilitate studies of the mechanism and application of these critical, widespread molecules. Unnatural N-acetylhexosamine analogs were metabolically incorporated into the capsule polysaccharides of Escherichia coli and Bacillus subtilis via bacterial metabolism. Targeted metabolic labeled hyaluronan and the precursors of heparin and chondroitin sulfate were obtained. The azido-labeled polysaccharides (purified or in capsules) were reacted with dyes, via bioorthogonal chemistry, to enable detection and imaging. Site-specific introduction of fluorophores directly onto cell surfaces affords another choice for observing and quantifying bacteria in vivo and in vitro. Furthermore, azido-polysaccharides retain similar biological properties to their natural analogs, and reliable and predictable introduction of functionalities, such as fluorophores, onto azido-N-hexosamines in the disaccharide repeat units provides chemical tools for imaging and metabolic analysis of GAGs in vivo and in vitro.

Metabolic labeling of capsular polysaccharides in bacteria creates the chemical tools to observe and quantify glycosaminoglycans.

INTRODUCTION

Glycosaminoglycans (GAGs) are widely present in the extracellular matrix and on the surface of mammalian cells (1). These unbranched polysaccharides consist of repeating disaccharide units incorporating a uronic acid linked to a hexosamine, amino sugars created by adding an amine group to hexose. Hyaluronan (HA), heparin (HP), heparan sulfate (HS), and various chondroitin sulfates (CSs) are members of the GAG family that vary in the type of saccharide unit and the geometry of the glycosidic linkage they contain. These glycans are involved in various biological processes through interaction with extracellular signaling molecules (1, 2) and have become commonly used drugs (3–5). Development of chemical tools for imaging, activity-based glycan (or protein) profiling, and pharmacological perturbation of GAGs is urgently needed (6, 7).

Biomacromolecular labeling techniques have been developed in a range of systems through introduction of unnatural chemical moieties and unique functionalities onto desired groups via bioorthogonal chemistry (8, 9). Bioorthogonal chemistry has been rapidly applied in the field of bacterial glycobiology; glycoproteins, lipopolysaccharides, and peptidoglycan have been labeled and targeted (10). Numerous chemical ligation strategies, such as copper-free click chemistry, isocyanide-based click reactions, and quadricyclane ligation, have been developed for in vivo application without interfering with native biochemical processes. If bioorthogonal chemistry can be adapted for GAGs, it will be invaluable for assays of their activity and properties in vivo and in vitro (11). Recently, Leary et al. (12) successfully prepared azide-primed HS and CS using Chinese hamster ovary cells. However, selective labeling of GAGs with unnatural sugars using bacterial metabolism has not been reported.

Fortunately, some bacterial capsules (polysaccharide layers outside the cell envelope) have the same repeating disaccharide units as GAG precursors. These extracellular polysaccharide coatings are virulence factors for many pathogenic bacteria. For example, HA polysaccharide is produced by Gram-negative Pasteurella multocida type A (13) and Gram-positive groups A and C Streptococcus (14). Escherichia coli K5 (serovar O10:K5:H4) and P. multocida type D form capsules composed of unsulfated precursors of HP and HS, also called heparosan or K5 polysaccharide (15). Chondroitin, a CS precursor, was found in type F P. multocida, E. coli K4 (serovar O5:K4:H4) (16), and Avibacterium paragallinarum (17). Furthermore, the identification and biochemical characterization of the enzymes involved in the synthesis pathways of these bacterial polysaccharides has leapt forward (18). Therefore, hijacking capsular biosynthesis in recombinant bacteria may be a way to develop metabolic labeling methods for GAG precursors by introducing desired unnatural saccharide reporters.

The repeating disaccharide units of GAGs consist of an N-acetylglucosamine (GlcNAc) or an N-acetylgalactosamine (GalNAc) and a glucuronic acid (GlcA) residue. Meanwhile, N-azidoacetylglucosamine (GlcNAz), N-azidoacetylgalactosamine (GalNAz), and their tetra-acetyl derivatives (Ac₄GlcNAz and Ac₄GalNAz) are well-known polysaccharide-labeling agents, of which incorporation into cell-surface biomacromolecules has been demonstrated in yeast (19), plant (20), and mammalian cells (21). Ac₄GlcNAz and Ac₄GalNAz are hydrolyzed by nonspecific esterases upon entering cells (9, 21). Moreover, an artificial biosynthetic pathway of uridine-diphosphate–GlcNAc (UDP-GlcNAc) and UDP-GalNAc with broad tolerance to substrate modifications has been identified in vitro. C2′ derivatives of GlcNAc and GalNAc (GlcNAz and GalNAz) are tolerable substrates for this salvage pathway (22, 23). On the basis of these findings, we are interested in replacing N-acetylhexosamines with N-azidoacetylhexosamines in the repeating units of GAGs using metabolic engineering strategies in prokaryotic systems.

Here, unnatural saccharide reporters were metabolically incorporated into E. coli and Bacillus subtilis capsule polysaccharides, giving a way to obtain azido-bearing GAG precursors by fermentation. Then, site-specific introduction of functionalities (such as fluorophores) onto these azido groups was performed, affording the chemical tools to observe and quantify the bacteria and for imaging and metabolic analysis of the polysaccharides in bioassays with the aim of elucidating their biological activity and behaviors in animals and in vitro.

RESULTS

Targeted metabolic labeling of E. coli K5 capsule

The capsular polysaccharide of E. coli K5, usually named the K5 capsule or heparosan, consists of the repeat structure -4)GlcA-β(1,4)-GlcNAc-α(1- and is the same as that of the nonsulfated precursor of heparin. To determine whether the conserved GlcNAc residues within the cell-surface heparosan in E. coli K5 can be specifically targeted for metabolic replacement by synthetic sugar analogs, recombinant E. coli K5 strains were designed and constructed through four steps: (i) disruption of de novo biosynthesis of heparosan, (ii) introduction of inducibly expressed exogenous heparosan synthase [P. multocida heparosan synthase 2 (PmHS2)], (iii) disruption of de novo biosynthesis of UDP-GlcNAc, and (iv) exploitation of a constitutively expressed UDP-GlcNAc (UDP-GlcNAz) salvage pathway. The plasmids and recombinant strains used for this work are summarized in table S1.

In wild-type E. coli K5, the biosynthesis of the K5 capsule involves the concerted action of the enzymes KfiA and KfiC. The former was previously found to be a UDP-N-acetyl-d-glucosamine: heparosan α-1,4-N-acetyl-d-glucosaminyltransferase. However, KfiA has a narrow tolerance for monosaccharide donor substrates, and UDP-GlcNAz is not a suitable donor substrate for its α-1,4-GlcNAc activity (24). Fortunately, however, UDP-GlcNAz, the chemical reporter used in this work, is a tolerated substrate for the α1,4-GlcNAcT activity of PmHS2 (25). PmHS2 is a dual-function glycosyltransferase, responsible for the synthesis of heparosan as the capsular polysaccharide of P. multocida type D. Therefore, to incorporate GlcNAz into the K5 capsule of E. coli K5, KfiA was knocked out using the CRISPR-Cas9 system (fig. S1), followed by the introduction of PmHS2 under the control of an inducible promoter. Disaccharide analysis based on the degradation of polysaccharides by heparosan endo-β-eliminase HepIII (26) demonstrated that KfiAΔ strains lost K5 capsule-synthesis activity, but expression of PmHS2 could rescue this disability (fig. S2).

We hypothesized that genetic disruption of de novo UDP-GlcNAc biosynthesis in E. coli would force cells to scavenge environmental GlcNAc (or GlcNAz) through an artificial nucleotide-sugar salvage pathway to survive, which could enable the use of GlcNAz as a chemical reporter (Fig. 1A). The five-step de novo UDP-GlcNAc biosynthesis pathway is responsible for the conversion of fructose to UDP-GlcNAc in wild-type E. coli K5. Glutamine-6-phosphate fructose aminotransferase (GlmS), a highly conserved enzyme in this pathway, is essential for the viability of E. coli in normal growth conditions. However, glmSΔ strains can be rescued using the functional nagE–nagA pathway if extracellular GlcNAc is provided as a supplement to the culture medium (fig. S3). Then, because GlcNAz is known to be tolerated by biosynthesis machinery consisting of N-acetylhexosamine 1-kinase (NahK) (27, 28) from Bifidobacterium longum and Homo sapiens UDP–N-acetyl hexosamine pyrophosphorylase (AGX1) (29), we introduced these two exogenous enzymes into E. coli strain glmSΔ-KfiAΔ-K5. NahK and AGX1 were respectively placed under the control of the constitutively expressed promoter (OXB20) in vector pACYDuet. The final desired strain—E. coli K5 KfiAΔ-glmSΔ-NahK-AGX1-PmHS2-K5—was named E. coli K5ASSH.

Fig. 1. Introduction of GlcNAc analogs into Escherichia coli K5 capsular polysaccharide.

(A) Design of reporter sugars for metabolic labeling of K5 capsule. De novo K5 polysaccharide biosynthesis was blocked by knocking out KfiA, which was followed by the introduction of an exogenous polysaccharide synthase (P. multocida heparosan synthase 2, PmHS2) under the control of an inducible promoter. A strategy for bypassing de novo UDP-GlcNAc biosynthesis was also used—glmS was knocked out, and then unnatural GlcNAc analogs (or GlcNAc) enter the salvage pathway and are converted into activated nucleotide sugars. To label K5 capsular polysaccharide, recombinant strains were grown in medium in which extracellular GlcNAc was provided as a supplement; then, well-grown K5 cells were cultured with GlcNAz and the expression of PmHS2 was simultaneously induced. KfiA, UDP-N-acetyl-d-glucosamine: heparosan α-1,4-N-acetyl-d-glucosaminyltransferase. KfiC, UDP–glucuronic acid: heparosan UDP-glucose dehydrogenase; GlmS, glutamine-fructose-6-phosphate transaminase; Pgm, phosphoglucomutase; GalU, UTP-glucose-1-phosphate uridylyltransferase; UgdA, UDP-glucose 6-dehydrogenase; GlmM, phosphoglucosamine mutase; GlmU, bifunctional GlcNAc-1-phosphate uridyltransferase/glucosamine-1-phosphate acetyltransferase; NahK, N-acetylhexosamine 1-kinase; AGX1, UDP-GlcNAc pyrophosphorylase 1. (B) Synthesis pathway of K5 capsular polysaccharide catalyzed by PmHS2 in recombinant E. coli strains. (C) GlcNAc analogs are incorporated into the disaccharide repeat units of K5 capsular polysaccharide located on the cell surface. Both GlcNAz and GlcNAl-supplemented E. coli K5 showed strong reactivity with fluorescein amidite (FAM)–alkyne (1) or rhodamine-N₃ chloride (2), neither of which was observed in the same cells grown in medium supplemented with GlcNAc. Furthermore, after HS-specific degradation by heparinase III (HepIII), the fluorescence almost disappeared. Structured illumination microscopy (SIM) images and fluorescence data were collected from three replicate experiments.

Notably, aiming to label the capsular polysaccharide of strain K5ASSH and minimize the ratio of GlcNAc to GlcNAz within the capsule, environmental GlcNAc was completely replaced with GlcNAz, and PmHS2 was simultaneously expressed by adding inducer (Fig. 1B). Briefly, recombinant E. coli K5 was grown in medium in which extracellular GlcNAc was provided as a supplement; then, well-grown cells were harvested, washed, and cultured further in medium containing GlcNAz and isopropyl β-d-1-thiogalactopyranoside (IPTG).

We conducted several experiments to assess the targeted metabolic labeling of the K5 capsule using the unnatural GlcNAc analogs in recombinant E. coli K5 strains. First, fluorescence microscopy assays were used to determine whether GlcNAz or N-pentynoacetylglucosamine (GlcNAl) was metabolized through the GlcNAc salvage pathway (Fig. 1C). In the case of azide-bearing GlcNAz-supplemented strain K5ASSH or alkyne-bearing GlcNAl-supplemented strain K5ASSH, we used an alkyne-functionalized fluorescein amidite (FAM) dye (i) or an azide-functionalized rhodamine (ii) to chemospecifically modify the azide or alkyne via Cu(I)-catalyzed azide-alkyne cycloaddition (30). GlcNAz- and GlcNAl-supplemented cells treated in this way displayed strong fluorescence on their surfaces, while GlcNAc-supplemented cells did not (Fig. 1C), suggesting that GlcNAc analog–supplemented K5ASSH cells were carrying azido or alkyne groups that reacted with the functionalized fluorescent dyes. The fluorescence intensity correlated with the amount of GlcNAz supplemented into the medium (fig. S4). Moreover, if supplemented with a GlcNAz/GlcNAc mixture, a decrease of the azido-dependent fluorescence of strain K5ASSH was observed, which depended on the GlcNAz/GlcNAc ratio (fig. S5).

To verify that azido and alkyne incorporation was limited to K5 capsular polysaccharide, we subjected the GlcNAz- and GlcNAl-supplemented K5ASSH cells to digestion with heparosan endo-β-eliminase HepIII (26), which specifically targets the α-1,4 glycosidic bonds between GlcNAc (or its analogs) and GlcA residues. We observed an overall downshift, almost disappearance, of the azide-dependent or alkyne-dependent fluorescence, consistent with degradation of the heparosan (Fig. 1C). To further verify the nonnatural sugar labeling of the capsule, K5 capsular polysaccharide was isolated from the GlcNAz-supplemented strain K5ASSH and incubated with HepIII; the resulting disaccharides were examined by high-performance liquid chromatography–mass spectrometry (HPLC-MS) (26) (fig. S6). The structures of the digestion products were identified as disaccharides with molecular mass 420.1150 ± 0.0043 Da, which is close to the calculated molecular mass of the Δ⁴-unsaturated disaccharide—4,5 unsaturated uronic acid-GlcNAz (ΔUA-GlcNAz) (420.1129 Da). This result supports the conclusion that an azide-bearing GlcNAc analog (i.e., GlcNAz) was successfully integrated into E. coli K5 capsular polysaccharide. In the authors’ opinion, some GlcNAc or GalNAc has the opportunity to be afforded through endogenous metabolic processes and subsequently incorporating into polysaccharides, although environmental N-acetylhexosamines were replaced with unnatural saccharide reporters (GlcNAz or GalNAz) when the polysaccharide synthases were expressed by adding inducer. However, the LC assays of disaccharides digested from modified polysaccharides indicated that the percentage of repeating units carrying natural N-acetylhexosamines was pretty low within the polysaccharides of GlcNAz- and GlcNAl-supplemented cells.

Imaging of E. coli K5

The incorporation of GlcNAz into the capsular polysaccharide enables labeling and visualization of E. coli K5, whether the bacteria are in a test tube or in the body of an animal. Addition via bioorthogonal reactions of functionalized fluorescent dyes, such as 4-ethynyl-N-ethyl-1,8-naphthalimide (31) or dibenzylcyclooctyne (DBCO) labeled with cyanine5 (11), enabled visualization of E. coli K5 cells (fig. S7). Here, we conducted two experiments as example applications of E. coli K5 imaging.

First, the progress of phagocytosis of strain K5 in contact with mouse macrophages was successfully observed (Fig. 2, A and B). A DBCO-functionalized cyanine5 dye with red fluorescence was used to visualize GlcNAz-supplemented E. coli strain within macrophages by fluorescence microscopy, followed by assay by flow cytometry (Fig. 2C). The beginning of the infection and bacterial clusters within infected macrophages were imaged. Notably, this observation was quantified using structured illumination microscopy (SIM) and flow cytometry to estimate the percentage of macrophages infected with K5 cells. Furthermore, the number of internalized bacteria, and the ability of these internalized E. coli K5 to replicate in macrophages, was determined.

Fig. 2. Click-labeling and fluorescence imaging of GlcNAz-treated E. coli K5 in macrophages and mice.

(A) Phagocytosis of fluorescently labeled GlcNAz-supplemented E. coli K5 by macrophages. (B) SIM of the progress of phagocytosis of strain K5 cells by macrophages. Scale bars, 10 μm. The K5 capsular polysaccharide was labeled with dibenzyl cyclooctyne-conjugated Cy5 dye (DBCO-Cy5; red fluorescence), followed by incubation with macrophages extracted from the abdominal cavity of mice and labeled with F4/80 monoclonal antibody (BM8)–fluorescein isothiocyanate (FITC) conjugate (green fluorescence). (C) Flow cytometry analysis of infected (red) and uninfected macrophages (blue). (D) Whole-body imaging of mice treated with GlcNAz-supplemented E. coli K5. The bacteria were alkyne-Cy7.5–labeled and gavaged into BALB/c mice. The intestinal tracts of mice were dissected after 2, 6, 8, 9, 12, and 24 hours, respectively. The bacterial distribution in the intestinal tract was observed by imaging. (E) Fluorescence changes in whole-body imaging of mice after intragastric administration of labeled GlcNAz-supplemented E. coli K5 at different concentrations [2 × 10⁸ or 4 × 10⁷ colony-forming units (CFU)]. Data were collected from three replicate experiments. Statistical analysis was performed by Wilcoxon matched-pair signed rank test, ***P = 0.0005. (F) Fluorescence imaging of mice treated with live (upper panel) and inactivated (lower panel) GlcNAz-supplemented E. coli K5 in vivo. Live and inactivated GlcNAz-supplemented E. coli K5 (2 × 10⁸ CFU) were Cy7.5-labeled and respectively gavaged into the intestinal tract of BALB/c mice.

Second, K5ASSH carrying azido K5 capsular polysaccharide was used to monitor the distribution and colonization of cells in the mouse intestinal tract (Fig. 2D). GlcNAz-supplemented cells were labeled using alkyne-Cy7.5 dye and then gastrically gavaged into BALB/c mice. In mice fed labeled K5ASSH cells, the areas around the stomach displayed strong fluorescence (Fig. 2D), demonstrating that the E. coli cells were successfully imaged in the animal body. As expected, fluorescence remained longer in mice that were fed 2 × 10⁸ K5 colony-forming units (CFU) than those that were fed 4 × 10⁷ CFU (Fig. 2E). Furthermore, the mean level of fluorescence remaining in disembodied intestinal tracts and live animals showed a similar trend, and a change in the colonization location was visualized (Fig. 2E and fig. S8). When living and ethanol-inactivated alkyne-Cy7.5–labeled K5 cells (2 × 10⁸ CFU/mL) were respectively gavaged into mice, live bacteria remained in the intestinal tract longer, and intestinal colonization was observed (Fig. 2F).

Production of azido-labeled GAG precursors in recombinant Bacillus subtilis

Encouraged by our success in targeted azido labeling of heparosan in E. coli, we tried to produce clickable HA and chondroitin polysaccharides via similar strategies. This time, B. subtilis 168 was used as the host strain because of its clear glycan background and its “generally recognized as safe” status (32). HA is a nonsulfated GAG composed of repeating -4)GlcA-β(1,3)-GlcNAc-β(1- disaccharide units. Chondroitin, consisting of the repeat structure -4)GlcA-β(1,3)-GalNAc-β(1-, is the nonsulfated precursor of CS.

P. multocida HA synthase (PmHAS) (33) and chondroitin synthase (KfoC) (16) from E. coli O5:K4:H4 were respectively introduced into B. subtilis strain glmSΔ-NahK-AGX1-168 (table S1) as the glycan polymerases because of their tolerance of UDP-GlcNAz as a substrate (figs. S9 and S10). These polymerases were expressed under the control of the inducible promoter PxylA (table S1). It is worth noting that KfoA, the gene encoding UDP-glucose-4-epimerase in E. coli O5:K4:H4, was previously coexpressed with KfoC to produce UDP-GalNAc (or UDP-GalNAz) in vivo (34). KfoA can convert both natural substrates and azido substrates.

The biosynthesis of HA and chondroitin in GlcNAc-supplemented B. subtilis strains containing the functional nucleotide-sugar salvage pathway and desired polysaccharide polymerases, BS168SSHAS to HA and BS168SSAC to chondroitin, respectively, was confirmed by HPLC-MS analysis of their disaccharides digested by chondroitin AC exolyase (35) (figs. S11 and S12). Meanwhile, these strains produced azido-labeled polysaccharides if switched to a GlcNAz supplement in the culture medium, which was proved by a series of assays including fluorescence microscopy, flow cytometry, disaccharide analysis, and HPLC-MS (figs. S13 and S14). Moreover, metabolically labeled polysaccharides were purified, and the yield of azido-chondroitin (~26 mg/liter) and azido-HA (~75 mg/liter) produced by recombinant B. subtilis was determined by HPLC-MS. The yield of azido-heparosan in E. coli K5 was approximately 25 mg/liter.

Azido-labeled GAG precursors and E. coli K5 retain similar protein-binding specificities to most lectins

A key criterion for development of chemical tools is the ability of the materials to mirror natural molecules in native conditions. Therefore, we further characterized the azido-labeled GAG precursors and E. coli K5 cells. Interactions between glycans and glycan-binding proteins (GBPs) are essential for a wide range of biological processes in all kingdoms of life. Lectins are one of the major classes of GBP and are widely used to study the roles of complex carbohydrates in vitro and in vivo (36). Therefore, the specificities of labeled and natural GAG precursors toward lectins are an apt comparison to assess the impact of the azido modifications.

A microarray including 70 lectins was used in these experiments (fig. S15) (https://www.raybiotech.com/lectin-array-70/). Here, HA, a linear unsulfated GAG, was used as a representative azido-labeled polysaccharide. As shown in Fig. 3A, we used static (one-step procedure) and dynamic competitive (two-step procedure) assays simultaneously to determine the influence of azido modification on binding of the polysaccharide to the lectins. In the static experiment, before the addition of azido-HA polysaccharide to the chip, an excess of natural competitor HA (referred to as the cold probe) was added to the sample mix. The slides were then labeled with DBCO-conjugated Cy5 fluorescent dye (DBCO-Cy5), followed by determination of the fluorescence intensity (referred to as data 1 and shown in table S3) for each lectin on the chip (Fig. 3A and fig. S16). In the dynamic competitive assays, pure azido-HA was put on the lectin chip in the first step; then, this lectin chip was incubated with buffer containing natural HA (1000-fold excess over the azido-HA in step 1) as a competitor, followed by labeling of the polysaccharide bound to the lectins with fluorescent dye. These fluorescent data for each lectin were named data 2 (fig. S16 and table S3). Normalized data for the static and dynamic competitive assays were then compared to determine any change in the interaction between azido- and natural HA toward each specific lectin (Fig. 3B and fig. S16).

Fig. 3. Lectin binding profiles of azido GAGs and E. coli K5 in a microarray.

(A and B) A 70-lectin microarray was used to determine the impact of azido modifications on the ability of hyaluronan (HA) to bind specific lectins. (A) Schematic representation of simultaneous static (one-step procedure) and dynamic competitive assay (two-step procedure) to determine the effects of azide modification. In the static experiment, azido-HA polysaccharide and excess of natural competitor (HA, 1000-fold excess over azido-HA) were mixed and added to the chip (top). In the dynamic competitive assays, azido-HA was placed on the lectin chip, followed by competition with natural HA (1000-fold excess over azido-HA) (bottom). The chips for the static and dynamic competitive assays were labeled with fluorescent DBCO-Cy5 dye and washed, followed by determination of the fluorescence intensities at an excitation wavelength of 635 nm and an emission wavelength of 662 nm (table S3). (B) Normalized data for the static and dynamic competitive experiments for each lectin. (C) Grouping of the effects of azide modification on glycan-lectin specificities. Group I: The ratio of dynamic to static fluorescence was in the range of 0.50 to 2.00, or the fluorescence in both processes was negative, indicating that there was no binding of either glycan. Group II: ratio ≥ 2.00, P < 0.05; group III: ratio ≤ 0.50, P < 0.05. (D) Schematic representation of the determination of specificity differences toward lectins of azido-modified GAGs. Pure azido-HA (left) and its mixture with FITC-labeled heparin (right) were respectively applied on parallel lectin chips; then, the chips were incubated with DBCO-Cy5 and simultaneously scanned at an excitation wavelength of 635 nm and an emission wavelength of 662 nm or at an excitation wavelength of 532 nm and an emission wavelength of 540 nm. (E) The relative change in HA binding in the presence of heparin.

The effects of azide modification on the glycan–lectin specificities can be divided into three groups:

1) Group I: Similar fluorescence intensities were observed for HA and its azido-modified analog (i.e., the ratio of data 2 to data 1 was in the range of 0.5 to 2.0, P < 0.05). This group included 54 lectins (54 of 70, 80% of those on the chip). Note that lectins that displayed negative values for data 2 and data 1, demonstrating that neither natural- nor azido-HA polysaccharide was recognized by the lectin, were classified in this group.

2) Group II: The azido-polysaccharide bound the lectin more strongly than HA; hence, the fluorescence in the dynamic process was higher than that in the static process (ratio of data 2 to data 1 ≥ 2.0, P < 0.05). Six lectins were in this group. It appears that glucosamine residues carrying azido groups increased the affinity of the polysaccharide for these lectins.

3) Group III: The binding affinity of azido-modified HA to these lectins was lower than that of HA; 10 lectins on the chip were in this group.

Then, the binding specificities to lectins of the precursors of sulfated GAGs, and of E. coli K5-harboring heparosan polysaccharide with azido modification, were respectively determined in similar microarray-based interaction studies (Fig. 3C, figs. S17 to S19, and tables S4 to S6). These polysaccharides (or K5 cells) and their azido-labeled analogs bound with similar specificities to a wide range of lectins in vitro—the numbers of lectins unaffected by the azido modification of the polysaccharides were 42 (60%), 50 (~70%), and 53 (~75%) for chondroitin, heparosan, and E. coli K5 cells, respectively (Fig. 3C).

Azido HA and E. coli K5 facilitate the determination of relative specificity differences toward lectins among GAGs

GAGs bind hundreds of proteins, including cytokines, enzymes, and membrane receptors, with great specificity. This network of GAG-binding proteins is involved in most biological progresses in mammals. Therefore, it would be valuable to determine whether amide-labeled HA and GAG precursors can be applied to demystify these GAG-protein interactions and determine the different target protein specificities among these polysaccharides. As noted above, azido-HA retained similar specificity to that of unlabeled HA toward 80% of lectins in an on-chip assay. Thus, azido-HA and fluorescein isothiocyanate (FITC) conjugate–labeled heparin were used as model GAGs in competitive binding assays. Although HA and HP contain the same type of saccharide unit, they vary in the geometry of the glycosidic linkage. Briefly, FITC-labeled heparin and its mixture with azido-HA were respectively applied to parallel lectin chips (Fig. 3D). The slides capturing the polysaccharide mixture were incubated with DBCO-Cy5 and simultaneously scanned for fluorescence emission at 532 and 635 nm; meanwhile, fluorescence values for azido-HA were collected at 635 nm (Fig. 3D and fig. S20). The data are summarized in table S7. Then, the fluorescence intensity at 635 nm from the two parallel lectin chips was compared, coupled with analysis of the data at 532 nm, to determine any relative change in HA-binding levels in the presence of HP.

For a few proteins on the chip, such as human Malectin, Galectin 2, and Galectin 9, the binding of HA and heparin was competitive. For such proteins, the extent of the change in fluorescence for each lectin could be measured relatively accurately, providing a quantification of the binding preference of these lectins for heparin and HA (Fig. 3E). Meanwhile, for lectins that were not ligands for either heparin or HA, such as the soluble lectin from Burkholderia cenocepacia (BC2LCN), the strength of the fluorescence at 635 nm on the two parallel chips differed insignificantly (both close to 0). For a few proteins on the chip, the presence of heparin increased their affinity toward azido-HA; it appears that heparin becomes a “helper” or “coreceptor” for HA binding to these proteins, such as disconldin II from Dictyostelium discoideum and human galectin 3C-S.

Recombinant E. coli K5 with azido-modified capsule polysaccharide, and purified polysaccharides from such strains, are ideal models and tools to demystify the impact of capsule polysaccharides on bacterial adhesion to, and invasion of, host cells. The lectin-binding selectivity of independent azido-capsule polysaccharide (heparosan) and of whole E. coli K5 cells harboring azido heparosan was determined and compared (fig. S21 and tables S5 and S6). The 70 lectins on the chip were used to identify which displayed affinity for both bacteria and purified polysaccharide, or the ones for which the excessive polysaccharides alone cannot compete bacterial adhesion as “cold probes.” Some lectins on the chip bound to capsular polysaccharides, but the fluorescence intensity became weak when they were incubated with bacterial cells, probably because the size of whole cells (~2 μm) means that, spatially, only a few cells can bind to the receptor proteins on the chip (37). It is also possible that cells are more likely than capsular polysaccharides to be eluted during chip experiments in vitro. In the authors’ opinion, these unfavorable in vitro effects would not occur in experiments in which mammalian cells were used as the host, or in vivo in animals, because the binding of membrane proteins carried by mammalian cells to bacterial cells (or capsular polysaccharides) would not suffer the artifactual interference from the chip or the in vitro washing processes.

Azido-HA: Chemical biology tools for assessing HA activities and properties in the body

A method for the preliminary quantitative analysis of HA in vivo was established by taking advantage of the fact that azido groups can be selectively labeled and thus become visible in animal tissues. This approach enables imaging and metabolic analysis of HA in animal tissues, whether these azido-labeled polysaccharides are used as HA-based biomedical materials or as imitators of endogenous HA.

First, in animal tissues, after bioorthogonal labeling with alkynyl dyes through intraarticular injection, the amount of azide-HA showed a linear relationship with the fluorescence intensity observed (fig. S22). Briefly, various amounts of azido-HA, ranging from 0.1 to 1.2 mg, were injected into the left knee joints of 12 mice simultaneously, while the right knee joints were treated with the corresponding amount of native HA. One hour later, DBCO-Cy5 was injected into the knuckle articular cavity of the mice to chemically bind with the azide groups of the polysaccharides. As expected, all left joints (treated with azido-HA) were positive for Cy5 after 48 hours; neither the right knee joints nor other tissues showed obvious fluorescence intensity. Any DBCO-Cy5 that had not reacted with azide-labeled HA had been excreted by this time point. It is worth mentioning that such polysaccharide derivatives displayed the ability to capture compounds containing bioorthogonal groups that can bind with the azido groups, which suggests further applications of azido-polysaccharides in biomedical engineering.

Our next goal was to determine whether azido-HA could be used as an imitator of endogenous HA to verify the persistence time of HA in vivo. Therefore, animal experiments were designed and carried out as shown in Fig. 4A and fig. S23. Twenty-two mice were simultaneously injected with 0.8 mg of azido-HA into the knee joints. Every day, one mouse was randomly selected, and DBCO-Cy5 was injected into the knuckle articular cavity of that mouse; fluorescence determination was performed 2 days later. Twenty-four days after the selection of the first mouse, the residual azido-HA concentration in each of the mice was quantified on the basis of a standard curve of azide-HA fluorescence (Fig. 4, B and C, and fig. S24).

Fig. 4. HA labeling enables targeting and characterization of HA activities and properties in the animal body.

Targeted imaging and characterization of HA activities using azido-HA in mice. (A) Schematic representation of azido-HA visualization. (B) Azido-HA (0.8 mg) was injected into mouse knee joints, and fluorescence determination was performed on the third day after DBCO-Cy5 was injected into the knuckle articular cavity. Representative fluorescence imaging of BALB/c mice, knee bones, and cartilage tissue section. (C) Each day, for 22 days, one mouse injected in the knee joint with azido-HA on day 1 of the experiment was randomly selected and injected with DBCO-Cy5; 2 days later, that mouse was fluorescently imaged. In total, 22 mice were used in this experiment, and the concentration of azido-HA in the mouse knee joint over time (24 days) was determined. (D) Concentration of azido-HA over time in the mouse model of OA. (E) HA hydrogel displacement in mice observed based on targeted imaging.

Meanwhile, preplaced left flank azido-HA was generated by ectopic injection of fluorescein, which circulated in the blood and bound to the azido-HA (fig. S25). Here, fluorescence determination was performed 24 and 48 hours after DBCO-Cy5 was intravenously injected.

Next, the effectiveness of our method was further verified in HA metabolism experiments in animal models of osteoarthritis (OA) (38). HA degradation in OA and healthy animals was analyzed and compared using azido-HA as imitator of endogenous HA in knee joints. The finding that arthritic knees metabolized HA significantly faster than healthy knees confirmed that azido-HA not only is an eligible imaging tool for HA research but also displays potential for application in diagnosis and evaluation of the effect of drug treatments of OA (Fig. 4D and fig. S26).

Last, a polysaccharide hydrogel was prepared using azido-HA and injected subcutaneously into mice. The molecular weight of the polysaccharide was approximately 12 kDa (fig. S27). Then, in vivo degradation and displacement of this HA-hydrogel were assessed (fig. S28). The in vivo retention time and elimination half-life were 90 and 21 days, respectively. It is worth noting that in vivo displacement of hydrogels was successfully observed on the basis of our imaging approach (Fig. 4E and fig. S28C). On the third day, obvious fluorescence of HA could be seen on the mouse flank. On the 68th day, 90° displacement of HA could be observed when the body position of the mouse was flipped. Our results demonstrate that azido-HA will be an effective tool for studying the subcutaneous displacement of biomedical materials containing HA, e.g., by material fluidity or body movement.

DISCUSSION

High-efficiency, site-specific replacement of N-acetylhexosamines (GlcNAc for HA and heparosan and GalNAc for chondroitin) with unnatural analogs in the disaccharide repeat unit of capsular polysaccharides allows for reliable and predictable introduction of functionalities on the surface of bacteria with little risk of perturbing cell behavior. Our results demonstrate that the azido analogs GlcNAz and GalNAz can be metabolically incorporated into the capsule polysaccharides of E. coli and B. subtilis. These species are model Gram-negative and Gram-positive bacteria, respectively. We envision azido analogs as a convenient means for reliable introduction of fluorophores directly onto the cell surfaces of bacteria, providing options for observing and quantifying bacteria, for example, in assays of the intestinal tract in vivo, and for experimenting on the bacteria-derived polysaccharide-mediated communication patterns that are critical in bacteria-host cross-talk.

Furthermore, azido-bearing HA and the precursors of heparin and CS were successfully obtained through fermentation and purification using these recombinant strains. A method for the preliminary quantitative analysis of HA in vivo was established by taking advantage of the fact that azido groups can be selectively labeled with fluorescent dyes and thus become visible in animal tissues. This approach enabled imaging and metabolic analysis of HA in mouse tissues. In addition to imaging, these azido-polysaccharides, their enzymatically sulfated forms (GAGs), and azido-capsule–labeled bacteria will be effective chemical biology tools to facilitate discovery of specific biological processes in vitro and even in vivo. For example, they can serve as activity-based probes, not only for affinity analysis between proteins and a single polysaccharide but also in applications probing the networks of GAG-protein interactions.

HA has been recommended to treat knee OA and dry eye syndrome (39, 40), and it is also used in ophthalmic viscoelastic devices in surgical procedures (41). In addition, natural polysaccharides and a variety of their derivatives have been developed as good candidates for biomedical engineering and drug delivery systems (42). Azido-HA retains similar physicochemical, biochemical, and biological properties to HA, such as its rheological properties and biocompatibility. Our results confirmed that the targeted labeling of polysaccharides, affording the visualization of HA, is a chemical biological tool that can be developed to target and characterize HA activities in the animal body. The azido groups, metabolically incorporated and finely distributed within the saccharide chains, easily undergo covalent attachment to desired molecules through biorthogonal reactions, meaning that such polysaccharides can replace chemically modified derivatives of HA in biomedical engineering.

Although we tested only azido analogs in this study, we anticipate that a similar strategy will allow assimilation of additional nonnatural sugars into GAGs and their precursors with the help of substrate engineering of desired enzymes involved in nucleotide sugar and glycan biosynthesis. We further believe that targeted metabolic labeling of GAGs with various unnatural sugars will afford numerous applications that were not previously possible.

MATERIALS AND METHODS

Materials

GlcNAz, GlcNAl, Ac₄GlcNAz, and Ac₄GalNAz were purchased from Jinan Samuel Pharmaceutical Co. Ltd. (Jinan, SD, China). GlcNAc and GalNAc were purchased from Shanghai Yuanye Biotechnology Co. Ltd. (Shanghai, China). FAM alkyne (5-isomer), DBCO-Cy5, and cyanine 7.5 alkyne were purchased from Lumiprobe Ltd. (MD, USA). 4-Ethynyl-N-ethyl-1,8-naphthalimide was purchased from R&D Systems Inc. (Minneapolis, MN, USA). Chloramphenicol, ampicillin sodium, neomycin sulfate, kanamycin sulfate, IPTG, rhodamine-N₃ chloride, and d-(+)-xylose were purchased from MedChemExpress LLC (NJ, USA). Centrifugal filters (Amicon Ultra-0.5 Millipore) and acetonitrile (hypergrade for HPLC-MS) were purchased from Merck KGaA (Darmstadt, Germany). M9 minimal salts, cupric sulfate pentahydrate, sodium l-ascorbate, tris-(hydroxymethyl) aminomethane, 4% paraformaldehyde fix solution, chloral hydrate, lysozyme, and poly-l-lysine were purchased from Sangon Biotech (Shanghai). Ammonium acetate for mass spectroscopy (eluent additive for HPLC-MS, ≥99.0%) was purchased from Aladdin (Shanghai, China). Phosphate-buffered saline (PBS) and Dulbecco’s modified Eagle’s medium (DMEM) were obtained from Invitrogen (Carlsbad, CA, USA). Penicillin-streptomycin and inactivated fetal bovine serum (FBS) were obtained from Gibco Life Technologies (Grand Island, NY, USA). A RayBiotech Lectin Array 70 was purchased from RayBiotech Inc. (Peachtree, the Netherlands). ProLong Glass Antifade Mountant, F4/80 Monoclonal Antibody (Thermo Fisher Scientific, catalog no. 62-4801-80, RRID: AB_2723151 BM8, eBioscience), and FITC isothiocyanate conjugate heparin were purchased from Thermo Fisher Scientific Corporation (MA, USA).

Instrumentation

Scanning microscopy images were taken on a structured illumination microscope (Danaher Corporation, Washington, DC, USA). Flow cytometry analyses of cells were conducted using a ZE5 Cell Analyzer (Bio-Rad, Hercules, CA, USA). HPLC-MS analyses were performed on a liquid chromatograph in conjunction with a quadrupole time-of-flight tandem high-resolution mass spectrometer Electrospray-ionisation quadrupole time-of-flight mass spectrometry (ESI-Q-TOF) (Bruker) and an analytical Hydrophilic interaction liquid chromatography (HILIC) column (Luna HILIC 200 Å, 150 × 2.0 mm, 3 μm; Phenomenex, Torrance, CA, USA). Fluorescent images of animals were obtained using an in vivo imaging system (IVIS) Spectrum in vivo imaging system (PerkinElmer, Waltham, MA, USA).

Bacterial strain construction

E. coli

The plasmids and E. coli strains used in this study are listed in table S1. The primers used are listed in table S2. E. coli strain K5A was constructed using the CRISPR-Cas9 system by eliminating gene KfiA from E. coli O10:K5 (L):H4 (American Type Culture Collection, 23506, E. coli K5). The fragment glmS_F-Kn-glmS_R was amplified by polymerase chain reaction (PCR) using plasmid pKD4 as the template, Pfu DNA polymerase, and primers glmS_F (E. coli)/glmS_R (E. coli). The glmS_F-Kn-glmS_R fragment contained homologous flanking sequences of glmS and a kanamycin-resistance gene. Competent E. coli K5A cells were electrotransformed with pKD46 and cultured in liquid LB medium containing ampicillin. The glmS_F-Kn-glmS_R fragment was transferred into E. coli K5A/pKD46 by electrotransformation. E. coli K5AS was cultured in liquid LB medium containing kanamycin and GlcNAc (final concentration, 100 μg/ml). pCP20 was transferred into E. coli K5AS-Kn by electrotransformation to knock out the kanamycin resistance gene.

PCR amplification of a Hind III–POXB20–Nde I fragment (primers H_POXB20_1/N_POXB20_2) with 5′–Hind III and 3′–Nde I restriction sites, and an Apa I–POXB20–Bam HI fragment (primers A_POXB20_1/B_POXB20_2) with 5′–Apa I and 3′–Bam HI restriction sites, used pET28a-POXB20-GFP as the template. Constitutive promoter POXB20 fragments with flanking restriction sites were obtained, respectively. The target DNAs were purified from agarose gels. Hind III–POXB20–Nde I and pACYDuet-Ptrc-AGX1-Ptrc-NahK were digested with restriction endo-eliminases Hind III and Nde I, and then the plasmid skeleton and promoter POXB20 were ligated using T4 ligase to obtain pACYDuet-Ptrc-AGX1-POXB20-NahK. Apa I–POXB20–Bam HI and pACYDuet-Ptrc-AGX1-POXB20-NahK were digested with restriction endo-eliminases Apa I and Bam HI, and then the plasmid skeleton and promoter POXB20 were ligated using T4 ligase to obtain pACYDuet-POXB20-AGX1-POXB20-NahK.

The gene encoding P. multocida heparosan synthase 2 (PmHS2) was amplified by PCR from pET28a-INP-PmHS2, with 5′–Bam HI and 3′–Hind III restriction sites (primers B_PmHS2_1/H_PmHS2_2). Subsequently, the Bam HI–PmHS2–Hind III PCR product was linked to restriction-digested vector pET28a-PT7-elmA to generate pET28a-PT7-PmHS2. PCR amplification of promoter Ptrc (primers B_Ptrc_1/N_Ptrc_2) with 5′–Bgl II and 3′–Nco I restriction sites was from pACYDuet-Ptrc-AGX1-Ptrc-NahK. pET28a-Ptrc-PmHS2 was produced by linking the PCR product containing promoter Ptrc to restriction-digested plasmid pET28a-PT7-PmHS2.

All inserted genes were verified using sequencing primers: For verification of homologous recombinant knockout of glmS, primers glmS_F (E. coli)/glmS_R (E. coli) were used. For verification of kanamycin-resistance gene deletion, primers glmS_outF/glmS_outR were used. For pACYDuet-POXB20-AGX1-POXB20-NahK, primers H_POXB20_1/N_POXB20_2 and A_POXB20_1/B_POXB20_2 were used. For pET28a-Ptrc-PmHS2, primers B_PmHS2_1/H_PmHS2_2 and B_Ptrc_1/N_Ptrc_2 were used.

pACYDuet-POXB20-AGX1-POXB20-NahK and pET28a-Ptrc-PmHS2 were successively transferred into E. coli K5AS by electrotransformation. Thus, an E. coli K5 strain containing the plasmids, with both KfiA and glmS knocked out, was obtained; this was named strain K5ASSH (i.e., KfiAΔ-glmSΔ-NahK-AGX1-PmHS2 K5).

B. subtilis

The plasmids involved in the construction of recombinant Bacillus subtilis strains are listed in table S1. pUC57-neo was digested with Bam HI to obtain a neomycin-resistance gene between homologous flanking fragments of the gene glmS. After recovery and purification, the fragment was transferred into B. subtilis 168 to achieve gene glmS knockout from the genome, generating strain BS168S (i.e., glmSΔ-168). pUC57-Pveg-AGX1-Pveg-NahK was cut with Acc65 I and Sma I to obtain the homologous arm fragments of the neomycin resistance gene Pveg-AGX1-Pveg-NahK with kanamycin resistance gene. The Pveg-AGX1-Pveg-NahK gene fragment with homologous flanking sequences of the neomycin-resistance gene was integrated into the genome of BS168S (i.e., glmSΔ-168) by homologous recombination, and positive transformants were obtained as strain BS168SS (i.e., glmSΔ-NahK-AGX1-168).

The expression frame of the gene encoding PmHS2 (in a PxylA-PmHS2-T fragment) was amplified by PCR from pUC57-PxylA-PmHS2 with 5′–Acc65 I and 3′–Sma I restriction sites (primers PmHS2_1F/PmHS2_1R). The PCR product was then linked to pHT43 to produce pHT43-PxylA-PmHS2. The expression frames of genes KfoA and KfoC were amplified by PCR from pUC57-PxylA-KfoA-PxylA-KfoC with 5′–Acc65 I and 3′–Sma I restriction sites (primers KfoC_F/KfoC_R, and KfoA_F/KfoA_R, respectively). The PCR product was then linked to pHT43 to produce pHT43-PxylA-KfoA-PxylA-KfoC. The expression frame of gene PmHAS (in a PxylA-PmHAS-T fragment) was amplified by PCR from pUC57-PxylA-PmHAS with 5′–Acc65 I and 3′–Sma I restriction sites (primers PmHAS_1F/PmHAS_1R). The PCR product was then linked to pHT43 to produce pHT43-PxylA-PmHAS.

All inserted genes were identified using sequencing primers: To verify homologous recombinant knockout of glmS, primers glmS_F (B. subtilis)/glmS_R (B. subtilis) and neo_F/neo_R were used. To verify neomycin-resistance gene knockout by homologous recombination, primers NahK_2F/NahK_2R and tuaD_F/tuaD_R were used. For pHT43-PxylA-PmHS2, primers PmHS2_F/PmHS2_R were used. For pHT43-PxylA-KfoA-PxylA-KfoC, primers KfoC_F/KfoC_R and KfoA_F/KfoA_R were used. For pHT43-PxylA-PmHAS, primers PmHAS_F/PmHAS_R were used.

pHT43-PxylA-PmHS2 was transferred into the receptive state of BS168SS (i.e., glmSΔ-NahK-AGX1-168) by electrotransformation to generate B. subtilis strain BS168SSHS (i.e., glmSΔ-NahK-AGX1-PmHS2-168). pHT43-PxylA-KfoA-PxylA-KfoC was transferred into the receptive state of BS168SS (i.e., glmSΔ-NahK-AGX1-168) to obtain strain BS168SSAC (i.e., glmSΔ-NahK-AGX1-KfoA-KfoC-168). pHT43-PxylA-PmHAS was transferred into the receptive state of BS168SS (i.e., glmSΔ-NahK-AGX1-168) to obtain strain BS168SSHAS (i.e., glmSΔ-NahK-AGX1-PmHAS-168). Strains can be made available via a material transfer agreement (MTA) by contacting J.-Z.S.

Animals

Male BABL/c mice aged 4 to 6 weeks old and male BABL/c mice aged 10 to 12 weeks old were purchased from Vitonlihua (Beijing, China). Mice were kept in specific pathogen–free conditions, and each group was kept in a separate cage and allowed to eat and drink freely. Artificial light was provided in a 12-hour light/12-hour dark cycle. All animal experiments were conducted according to guidelines approved by the Animal Care and Use Committee of Shandong University Cheeloo Medical College. For all the animal studies, mice were randomly allocated to groups. The investigator was aware of the group allocation during the animal studies, as demanded by the experimental design.

In vitro and in vivo azido or alkyne labeling of capsular polysaccharides

E. coli strain K5ASSH was activated at 225 rpm and 37°C for 12 hours in liquid LB medium containing chloramphenicol (34 μg/ml), kanamycin (50 μg/ml), and GlcNAc (100 μg/ml). After consumption of GlcNAc for 1 hour, GlcNAz (500 μg/ml) or GlcNAl (500 μg/ml) was added to the culture, which was continued at 37°C and 225 rpm until the OD₆₀₀ (optical density at 600 nm) was 0.6. Meanwhile, IPTG was added to the medium (0.2 mM). Culture was continued for 10 hours at 22°C. Then, the bacteria were collected (5000g, 5 min) and washed three times with 600 μl of PBS. In the control group, GlcNAc was added without unnatural monosaccharides; the other conditions were the same.

B. subtilis strains BS168SSHS2, BS168SSAC, and BS168SSHAS were respectively activated overnight in liquid LB medium containing GlcNAc (100 μg/ml), chloramphenicol (5 μg/ml), and kanamycin (50 μg/ml) at 225 rpm and 37°C. On the second day, the bacteria were centrifuged (4427g, 5 min), washed three times with PBS, and transferred to M9 medium. After consuming the original GlcNAc for 1 hour, GlcNAz (final concentration, 500 μg/ml) was added. The culture was continued at 37°C and 225 rpm until the OD₆₀₀ reached 0.45. Then, xylose (20 mg/ml) was added, and the culture was continued for a further 6 hours at 37°C and 225 rpm. Then, the bacteria were collected (4427g, 5 min) and washed three times with PBS. In the control group, GlcNAc was added without azide monosaccharide; the other conditions were the same. Fluorescence imaging and fluorescence-activated cell sorting (FACS) of E. coli K5ASSH cells were remodeled with different concentrations (w/v) of GlcNAz (100, 200, 300, 400, and 500 μg/ml), and then the azido-containing strains were labeled by FAM alkyne (5-isomer). Fluorescence imaging and FACS of E. coli K5ASSH cells were remodeled with different ratios of GlcNAz and GlcNAc, and different proportions of azide monosaccharides and nonazide monosaccharides were added into the culture medium: 1:1 (0.45 M:0.45 M), 1:2 (0.3 M:0.6 M), 1:3 (0.225 M:0.675 M), 1:4 (0.18 M:0.72 M), and 1:8 (0.1 M:0.8 M), after which the azido-containing strains were labeled by FAM alkyne (5-isomer). Flow cytometry and fluorescence microscopy were used to find the lowest limit of azide monosaccharide addition.

The azido-containing strains were labeled by bioorthogonal reactions. For Cu(I)-catalyzed bioorthogonal chemistry, the collected bacterial cells were suspended in 200 μl of PBS (2 × 10⁸ CFU); 10 mM FAM alkyne (5-isomer), 32 mM copper sulfate pentahydrate, and 0.32 M sodium ascorbate were added sequentially to the bacterial suspension. After incubation at 37°C for 1 hour, the cells were washed three times with PBS (1 ml).

For biorthogonal chemistry without copper catalysis, bacterial cells were suspended in 200 μl of PBS (2 × 10⁸ CFU). DBCO-Cy5 (10 mM) was added to the bacterial suspension, which was incubated at 37°C for 1 hour. Then, the cells were washed three times with PBS (1 ml).

For Cu(I)-catalyzed click-activated bioorthogonal chemical reaction, bacterial cells were suspended in 200 μl of PBS (2 × 10⁸ CFU); 10 mM fluorescein 4-ethynyl-N-ethyl-1,8-naphthalimide, 32 mM copper sulfate pentahydrate, and 0.32 M sodium ascorbate were sequentially added to the bacterial suspension and incubated at 37°C for 1 hour.

Bacteria imaging

Bacterial fluorescence was observed under a structural illumination microscope as follows: A cover glass (Zeiss; 18 mm × 18 mm, 0.17 ± 0.005 mm) was treated with poly-l-lysine (0.1 mg/ml) at room temperature for 3 hours. The cover glass was washed three times with sterile water and air-dried at room temperature. Labeled bacteria (20 μl, PBS, 2 × 10⁸ CFU) were placed in the center of the cover slide and incubated at room temperature for 30 min for the cells to adhere. The bacteria were washed three times with PBS, fixed with 500 μl of 4% paraformaldehyde, and placed at 4°C for 30 min. The cover glass with fixed bacteria was cleaned three times with PBS, and 5 μl of ProLong Glass Antifade Mountant agent was placed on the slide glass to cover the whole glass. FAM alkyne (5-isomer), DBCO-Cy5, and 4-ethynyl-N-ethyl-1,8-naphthalimide were stimulated by laser excitation at 488, 568, and 405 nm, respectively. The exposure time of the camera was set to 100.0 ms with a Z value of 5812.5 μm. Using SIM analysis software, processing and filtering settings remain unchanged, and image intensity is preserved. Reconstruction was performed using OMX SI Reconstruction and maximum intensity projection was carried out using Quick Projection. The analysis software generates two-dimensional images.

For flow cytometry, bacteria labeled with FAM alkyne (5-isomer), DBCO-Cy5, or 4-ethynyl-N-ethyl-1,8-naphthalimide were washed and resuspended in 100 μl of PBS (1 × 10⁸ CFU). Flow cytometry was performed on a ZE5 Cell Analyzer, and each sample was backwashed with deionized water before and after. The samples were swirled for 10 s before each run. The cells of each sample were analyzed, and the fluorescence intensity of the R1 gated sample was measured. FAM alkyne (5-isomer), DBCO-Cy5, and 4-ethynyl-N-ethyl-1,8-naphthalimide were stimulated by laser excitation at 488, 568, and 405 nm, respectively. Histograms of fluorescence intensity (height) were generated and overlaid.

Purification of HepIII and AsChnAC

HepIII was expressed in E. coli BL21(DE3) harboring a pET28a-Ptrc–HepIII vector, and AsChnAC was expressed in E. coli BL21(DE3) harboring a pET28a-Ptrc–AsChnAC vector. HepIII and AsChnAC were purified using a Ni-NTA agarose column. Purification of HepIII (43) and AsChnAC (35) was performed as previously reported. The enzyme activity was measured and calculated based on the increase in absorbance at 232 nm when GAG was depolymerized to unsaturated uronic acid by HepIII and AsChnAC.

Macrophage infiltration and immunostaining

Extraction of mouse peritoneal macrophages was performed as previously reported (44). Next, we studied the interaction between the macrophages and DBCO-Cy5–labeled E. coli K5ASSH cells. The medium was siphoned from the six-well plate, the cover glass was washed twice with PBS and transferred to a new six-well plate, and 1 ml of DMEM high-sugar medium containing 10% inactivated FBS was added without antibiotics. The plate was incubated at 37°C in a 5% CO₂ incubator for 1 hour. DBCO-Cy5–labeled E. coli K5ASSH cells were added (100 μl, 2 × 10⁸ CFU) into the macrophage culture medium, and the plate was incubated at 37°C under 5% CO₂ for 1 hour. Then, 100 μl of 10% FBS was added to the medium, and the plate was incubated at 4°C in the dark for 20 min and subsequently eluted twice with 1% FBS (500 μl). F4/80 monoclonal antibody (BM8)–FITC (0.5 μg; eBioscience) was added, and the plate was incubated at 4°C in the dark for 1 hour. Then, the cells were washed three times with PBS (500 μl). Paraformaldehyde (4%, 500 μl) was added to the plate for fixation for 20 min and then siphoned away, and the cells were washed three times with PBS (500 μl). After that, 5 μl of ProLong Glass Antifade Mountant agent was added to cover the slide. SIM photography used dual-color channel 488- and 568-nm laser excitation. Macrophages were labeled with FITC (green) and E. coli K5ASSH with DBCO-Cy5 (red), which enabled us to directly observe the interaction between the immune cells and the bacteria.

Flow cytometry analysis was also performed on the interaction between F4/80 monoclonal antibody (BM8)–FITC–labeled macrophages and DBCO-Cy5–labeled E. coli K5ASSH. DBCO-Cy5 labeled E. coli K5ASSH (PBS suspension, 100 μl, 2 × 10⁸ CFU) was added to culture plates with macrophages. The plate was incubated at 37°C under 5% CO₂ for 1 hour. Then, the culture medium was sucked out, and the plate was washed twice with PBS. Trypsin 0.25%–EDTA buffer (200 μl) was added for digestion for 10 min. Then, 5% FBS (500 μl) terminated the digestion. The digested cells were recovered into a 15-ml centrifuge tube. After centrifugation at 5000g for 3 min, the cells were retained. They were washed with 3 ml of PBS and centrifuged at 5000g for 3 min. Next, the cells were resuspended in 100 μl of PBS, 10 μl of FBS (10%) was added, and the cells were incubated on ice for 20 min in the dark and then washed twice with 1% FBS. F4/80 monoclonal antibody (BM8)–FITC (10 μl) was added, and the cells were incubated at 4°C in the dark for 1 hour and then washed twice with 1% FBS. The cells were resuspended in 1% FBS (200 μl) for flow cytometry analysis. In flow cytometry, macrophages that consumed DBCO-Cy5–labeled E. coli K5ASSH showed red fluorescence at an excitation wavelength of 646 nm and an emission wavelength of 662 nm.

Oral administration of labeled bacteria to mice

The GlcNAz-supplemented strains were labeled by bioorthogonal reactions

For Cu(I)-catalyzed bioorthogonal chemistry, the collected bacterial cells were suspended in 200 μl of PBS; 10 mM alkyne-Cy7.5, 32 mM copper sulfate pentahydrate, and 0.32 M sodium ascorbate were added sequentially to the bacterial suspension. After incubation at 37°C for 1 hour, the cells were washed three times with PBS (1 ml).

Imaging of labeled E. coli in gastrointestinal tract

The bacteria (200 μl of PBS, 2 × 10⁸ CFU) were alkyne-Cy7.5–labeled and gavaged into BALB/c mice (4 to 6 weeks old). The intestinal tracts of mice were dissected after 2, 6, 8, 9, 12, and 24 hours, respectively (n = 3). The bacterial distribution in the intestinal tract was observed by imaging. Fluorescence (excitation filter, 788 nm; emission filter, 808 nm) was also observed in whole-body imaging at 2, 6, 8, 9, 12, and 24 hours after intragastric administration. Furthermore, two groups of BALB/c mice (4 to 6 weeks old) were respectively treated with 4 × 10⁷ and 2 × 10⁸ CFU-labeled E. coli, and then the fluorescence was observed at 4, 6, 8, 10, 12, 14, 24, 26, 28, 30, 32, and 34 hours.

Imaging of mice treated with live or inactivated labeled E. coli in vivo

The experimental group was given living, labeled bacteria (200 μl of PBS, 2 × 10⁸ CFU). In the negative control group, the labeled bacteria were suspended in 75% (v/v) ethanol, shaken at room temperature for 30 min, washed three times with PBS, and then resuspended in 200 μl of PBS (200 μl of PBS, 2 × 10⁸ CFU). Mice were divided into two groups. One group of mice was given the alkyne-Cy7.5-labeled E. coli K5ASSH (200 μl of PBS, 2 × 10⁸ CFU), and the other group was given the same dose of ethanol-treated bacteria (200 μl of PBS, 2 × 10⁸ CFU). Fluorescence was also observed at 3, 8, 11, 24, 36, and 48 hours after intragastric administration.

Mouse imaging

Mice were treated with isoflurane air anesthesia for 10 min. The parameters were as follows: fluorescence imaging; exposure time, 1 s; binning, medium; f-stop, 2; electron multiplying gain, off. The excitation and emission filters were based on the excitation and emission wavelengths of different fluoresceins (Cyanine5 DBCO: Ex, 646 nm; Em, 662 nm; Cyanine7.5 alkyne: Ex, 788 nm; Em, 808 nm). Photography parameters were as follows: binning, medium; f-stop, 16; overlay. Fluorescence intensity at selected regions of interest was quantified using the IVIS Spectrum imaging software.

Enzymatic hydrolysis of azide or alkyne-GAGs and disaccharide HPLC-MS analysis

E. coli K5ASSH was labeled with FAM alkyne (5-isomer). For Cu(I)-catalyzed bioorthogonal chemistry, the collected bacterial cells were suspended in 200 μl of PBS (2 × 10⁸ CFU). Then, 10 mM FAM alkyne (5-isomer), 32 mM copper sulfate pentahydrate, and 0.32 M sodium ascorbate were added sequentially to the bacterial suspension. After incubation at 37°C for 1 hour, the cells were washed three times with PBS (1 ml). The bacteria were suspended in enzymatic hydrolysis system buffer (10 mM CaCl₂ and 50 mM tris-HCl, pH 7; 100 μl), and 100 μl of heparin endo-eliminase HepIII (0.2 mg/ml) was added. The incubation conditions for the negative control group were the same, but no HepIII was added. After incubation at 37°C for 24 hours, fluorescence images were taken, and flow cytometry analysis was performed.

Lysozyme was used to treat engineered bacteria with GAG modified by an azide group. Enzymolysis for 3 days was required. Cells were maintained at 225 rpm and 37°C, and fresh lysozyme (1 mg/ml) was added every 24 hours. After the reaction was terminated, the supernatant was taken, and salt was removed by 3-kDa ultrafiltration, yielding GAG polysaccharides produced by bacteria. Then, disaccharides were obtained by enzymolysis with the corresponding GAG endo-eliminase (26, 35). The heparosan was suspended in 200 μl of tristeamed water (0.1 mg/ml) containing 10 mM CaCl₂, 50 mM tris-HCl (pH 7.0), and heparin endo-eliminase HepIII (0.2 mg/ml). The HA and chondroitin were suspended in 200 μl of tristeamed water (0.1 mg/ml) containing 50 mM tris-HCl (pH 7.0) and endo-eliminase AsChnAC (0.2 mg/ml). Protein was removed by 30-kDa ultrafiltration, and the disaccharides were analyzed by HPLC and high-resolution MS. The LC method was as follows: mobile phase A, 5 mM ammonium acetate aqueous solution; mobile phase B, 5 mM ammonium acetate, 98% acetonitrile solution. The liquid phase gradient was 0 to 5 min, 95% B; 5 to 15 min, 95 to 77% B; 15 to 20 min, 77 to 40% B; 20 to 24 min, 40% B; 24 to 26 min, 40 to 95% B; 26 to 30 min, 95% B. The flow rate was 0.15 ml/min. The absorbance of the elution peak was measured at 232 nm. MS parameters were as follows: spray voltage, 4.5 kV; spray gas flow rate, 20 arb; capillary transfer tube temperature, 200°C; scanning range, 50 to 800 m/z (mass-to-charge ratio). All data were processed and analyzed on Bruker Data Analysis software. The quality errors were all <20 ppm.

Lectin microarray analysis of polysaccharides and bacteria

The lectin microarray was manufactured as previously described (45, 46) and was provided by Wayen Biotechnologies (Shanghai, China). For the dynamic competitive assay, 150 μl of azide-modified polysaccharide (PBS, 0.1 mg/ml) or 150 μl of azide-modified E. coli K5ASSH (~10⁵ cells, PBS) was used for sample loading. Before sample incubation, the sample was centrifuged at 6000g for 5 min to remove any particles or precipitates. PBS-diluted sample (150 μl) was added to the corresponding well, sealed with tape, and incubated at 4°C overnight. The sample was removed completely from the well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). PBS was completely removed after each wash. DBCO-Cy5 [5 mM, dimethyl sulfoxide (DMSO)] was briefly centrifuged at 6000g for 5 min. Eighty microliters of diluted DBCO-Cy5 was added to each well and incubated at room temperature in the dark for 1 hour. Then, DBCO-Cy5 was removed completely from each well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). After washing, labeling, and washing again, images were scanned at an excitation wavelength of 635 nm and an emission wavelength of 662 nm. After scanning, the frame was installed again, and 150 μl of common GAG backbone sample stock solution (100 mg/ml) or wild-type E. coli K5 (~10⁸ cells) was added respectively into the corresponding wells and incubated at 4°C for 10 hours. The sample was removed completely from the well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). Subsequently, the chip was scanned again at an excitation wavelength of 635 nm and an emission wavelength of 662 nm.

For static assay, PBS was added to the diluted sample (150 μl) and mixed with 75 μl of azide-modified polysaccharide (0.2 mg/ml, PBS) with 75 μl of normal GAG backbone sample (200 mg/ml, PBS), or the PBS-diluted sample (150 μl) mixed with 75 μl of azide-modified E. coli K5ASSH (~2 × 10⁵ cells) was mixed with 75 μl of wild-type E. coli K5 (~2 × 10⁸ cells) to the corresponding well, with tape pasted on the well at 4°C overnight. The sample was removed completely from the well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). PBS was completely removed after each cleaning. DBCO-Cy5 (5 mM, DMSO) was briefly centrifuged at 6000g for 5 min. Eighty microliters of diluted DBCO-Cy5 was added to each well and incubated at room temperature in the dark for 1 hour. Then, DBCO-Cy5 was completely removed from each well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). After washing, labeling, and washing again, images were scanned at an excitation wavelength of 635 nm and an emission wavelength of 662 nm. Then, 150 μl of FITC-labeled heparin (FITC-HP) (0.01 mg/ml, PBS) was added to the corresponding well, sealed with tape, and incubated at 4°C overnight. The sample was removed completely from the well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). PBS was completely removed after each cleaning and then scanned at an excitation wavelength of 532 nm and an emission wavelength of 540 nm. Next, an equal amount of azido-HA (75 μl, 0.02 mg/ml, PBS) and FITC-HP (75 μl, 0.02 mg/ml, PBS) was incubated in another well on the chip. The sample was removed completely from the well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). PBS was removed completely after each cleaning. DBCO-Cy5 (5 mM, DMSO) was briefly centrifuged at 6000g for 5 min. Eighty microliters of diluted DBCO-Cy5 was added to each well and incubated at room temperature in the dark for 1 hour. Then, DBCO-Cy5 was removed completely from each well, and 150 μl of PBS was added and used to gently wash the sample five times at room temperature (5 min each wash). After washing, labeling, and washing again, images were scanned at an excitation wavelength of 635 nm and an emission wavelength of 662 nm or at an excitation wavelength of 532 nm and an emission wavelength of 540 nm. Previous experiments only performed the incubation of azido-HA.

In vivo labeling of azido-HA in knee joints

Mouse knee joints were injected with azido-HA. Twenty-two male BALB/c mice (aged 12 to 14 weeks) were randomly selected. Each mouse was injected in the cavity of the left knee joint with azido-HA (0.8 mg) dissolved in 10 μl of PBS, and each mouse was injected in the cavity of the right knee joint with ordinary HA (0.8 mg) dissolved in 10 μl of PBS. The first mouse was injected with 10 μl of DBCO-Cy5 (5 mM, DMSO) fluorescein in both knee joints on the first day, and then body imaging was performed on the third day. Similarly, the 22nd mouse was injected with 10 μl of DBCO-Cy5 (5 mM, DMSO) fluorescein in both knee joints on the 22nd day, and then in vivo imaging was performed on the 24th day. In this way, the dynamic changes of knee joint fluorescence in the 22 mice indicated the dynamic changes of the amount of azido-HA acid in the mice over 24 days. Because of the inability of ordinary HA to bind DBCO-Cy5 in vivo, there was little or no fluorescence from the right knees of the mice. Fluorescent images of animals were obtained using the IVIS Spectrum in vivo imaging system (PerkinElmer, Waltham, MA, USA), and the signal was quantified as total radiant efficiency. Parameters for the in vivo imaging were as follows: exposure time, 1 s; binning, medium; f-stop, 2; EM gain, off; excitation filter, 646 nm; emission filter, 662 nm. Photograph parameters were as follows: binning, medium; f-stop, 16; overlay.

The molecular weight was determined by referring to a previously reported method (47). To establish the relationship between fluorescence intensity and the amount of azido-HA, we used fluorescence intensity and different concentrations of azido-HA (0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, and 1.2 mg) to generate a standard curve. The curve was established by injecting different concentrations of azido-HA (10 μl, PBS) labeled with DBCO-Cy5 (5 mM, DMSO) into the knee joint of BALB/c mice, and then body imaging was performed 48 hours later. In this way, we converted the fluorescence intensity into the amount of azido-HA remaining in the mice over the 24-day experimental period, and the azido-HA metabolism of the knee joint could be assessed.

An OA model was established by operation induction in ten 10-week-old BABL/c mice. Under anesthesia, the anterior cruciate ligament was transected in combination with destabilization of the medial meniscus generated from the transection of the medial meniscotibial ligament in the left knee joint to establish the OA model, resulting in medial meniscus instability. At the same time, the anterior cruciate ligament was cut, and the OA model was established, with the right knee joint as the control. One week after surgery, 10 μl of HA (0.5 mg, PBS) modified with DBCO-Cy5 (5 mM, DMSO) was injected into the OA (left) and healthy (right) knee joints of the OA model mice. Whole-body imaging of the mice was performed after 1, 5, 10, 15, 20, and 35 days to study the metabolism of HA in the joints. On day 35, the knee joint was taken out and made into a coronal section. The joint of the mouse OA model was stained with saffron and solid green to locate the cartilage (48).

In vivo labeling of azido-HA in flanks

To study the biodistribution of the azido-HA biopolymer in mice, three male BALB/c mice were treated with a 100-μl right flank subcutaneous injection of azido-HA biopolymer (1 mg/ml). The azido-HA biopolymer used was modified with 5 mM DBCO-Cy5 (DMSO) to allow for in vivo detection. Biopolymer residency was assessed over 100 days using the in vivo imaging system on days 1, 2, 3, 4, 7, 10, and 14 after injection and then twice or thrice weekly until the last day. Fluorescence imaging was conducted using 646-nm excitation and 662-nm emission filters to visualize Cy5, and the signal was quantified as total radiant efficiency of the biopolymer injection site.

Last, three male BALB/c mice were treated with a 100-μl left flank subcutaneous injection of azido-HA biopolymer (1 mg/ml, PBS) and a 100-μl right flank subcutaneous injection of HA biopolymer (1 mg/ml, PBS). DBCO-Cy5 (10 μl, 5 mM, DMSO) is injected into the tail vein. Fluorescence determination was performed 24 and 48 hours after DBCO-Cy5 was intravenously injected.

Statistical analysis

Statistical analysis was performed by Wilcoxon matched-pair signed rank test and one-way analysis of variance with post hoc Fisher’s least significant difference test using GraphPad Prism 9 software. The results were deemed significant at 0.01 < *P ≤ 0.05, highly significant at 0.001 < **P ≤ 0.01, and extremely significant at ***P ≤ 0.001. Sample size was empirically set at n = 3 to 6 for in vitro cell experiments and n = 3 or 4 for in vivo biodistribution and imaging studies.

Supplementary Materials

This PDF file includes:

Figs. S1 to S28

Tables S1 to S7

View/request a protocol for this paper from Bio-protocol.