A bioorthogonal chemistry approach to detect the K1 polysialic acid capsule in Escherichia coli

Most Escherichia coli strains associated with neonatal meningitis express the K1 capsule, a sialic acid polysaccharide that is directly related to their pathogenicity. Metabolic oligosaccharide engineering (MOE) has mostly been developed in eukaryotes, but has also been successfully applied to the study of several oligosaccharides or polysaccharides constitutive of the bacterial cell wall. However, bacterial capsules are seldom targeted despite their important role as virulence factors, and the K1 polysialic acid (PSA) antigen that shields bacteria from the immune system still remains untackled. Herein, we report a fluorescence microplate assay that allows the fast and facile detection of K1 capsules with an approach that combines MOE and bioorthogonal chemistry. We exploit the incorporation of synthetic analogues of N-acetylmannosamine or N-acetylneuraminic acid, metabolic precursors of PSA, and copper-catalysed azide–alkyne cycloaddition (CuAAC) as the click chemistry reaction to specifically label the modified K1 antigen with a fluorophore. The method was optimized, validated by capsule purification and fluorescence microscopy, and applied to the detection of whole encapsulated bacteria in a miniaturized assay. We observe that analogues of ManNAc are readily incorporated into the capsule while those of Neu5Ac are less efficiently metabolized, which provides useful information regarding the capsule biosynthetic pathways and the promiscuity of the enzymes involved. Moreover, this microplate assay is transferable to screening approaches and may provide a platform to identify novel capsule-targeted antibiotics that would circumvent resistance issues.


Introduction
Although Escherichia coli is an important part of the commensal microbiota colonizing the digestive tract of humans and animals, certain strains of this species have developed virulence attributes and cause serious diseases. Among these pathogenic strains, the meningitis/sepsis-associated E. coli (MNEC) pathotype is an important cause of neonatal infections. Mostly transmitted from mother to infant during birth, they are known to migrate to the vascular system after mucosal colonization, then penetrate the blood-brain barrier. This generally leads to meningitis or to septicaemia (blood-poisoning), both lifethreatening conditions with high mortality and morbidity rates. 1 80% of MNEC strains express the K1 capsule, a mucous layer of poly-a-2,8-sialic acid (PSA) that surrounds the bacterium thereby shielding its immunogenic proteins from detection by the host. 2,3 Mimicking the human glycan structure found in the neural cell adhesion molecule (NCAM), PSA is not recognized as an external threat and allows the encapsulated bacteria to evade the immune system, and the severity of these infections is directly related to the amount of K1 antigen found at the bacterial surface. [4][5][6][7] This capsular polysaccharide is a linear homopolymeric chain constituted of N-acetylneuraminic acid (Neu5Ac) monomers belonging to the sialic acid (Sia) family of carbohydrates. 8 Its de novo biosynthesis requires the condensation of N-acetylmannosamine (ManNAc) and phosphoenolpyruvate to form Neu5Ac, which is then converted to the sugar nucleotide donor cytidine 5 0 -monophospho N-acetylneuraminic acid (CMP-Neu5Ac) and assembled into PSA in the cytosol prior to translocation to the cell surface 9 (Fig. 1). E. coli serotype K1, as well as other bacteria that produce polysialic acid capsules such as Neisseria meningitidis, Pasteurella haemolytica A2 or Moraxella nonliquefaciens, can also scavenge free Neu5Ac from the host. 10 Monosaccharide analogues modified with an additional chemical moiety can be used as molecular tools to engineer sialylated glycoconjugates in metabolic oligosaccharide engineering (MOE) approaches. Reutter and co-workers first pioneered MOE, with synthetic ManNAc analogues bearing an elongated N-acyl side chain that were successfully metabolized into cell surface sialoglycoproteins. 11 Modern labelling methodologies combine MOE and click chemistry or bioorthogonal chemistry, an ensemble of reactions originally pioneered by the groups of Sharpless, Meldal and Bertozzi, who were recently awarded the Nobel prize for this achievement, [12][13][14] and further developed by them and other groups in the past twenty years. A sugar reporter capable of entering the metabolic pathway under scrutiny and equipped with a reactive handle is first introduced in a living organism of interest, followed by the bioorthogonal ligation of an appropriately functionalized probe, most often a fluorescent dye for optical bioimaging. 15 These methods rely on biocompatible click reactions that involve p 4 s + p 2 s pericyclic processes, namely the coppercatalysed azide-alkyne (3+2) cycloaddition, the strainpromoted azide-alkyne (3+2) cycloaddition and the inverse electronic demand Diels-Alder reaction (CuAAC, SPAAC and IEDDA, respectively). Indeed, such mechanisms are mostly absent from living systems, thus greatly minimizing interference with biological processes. 16 Although many bioorthogonal methods have been used in prokaryotic cells to track proteins with non-canonical amino acids, 17 efforts to apply MOE to detect glycans in prokaryotes remain rather meager by comparison, as these strategies were developed in human models first and foremost. Some glycoengineering methods were reported in various species to label bacterial glycans, which were extensively reviewed by Banahene et al. 18 They aim at detecting components of bacterial cell walls, such as peptidoglycans with N-acetylmuramic acid derivatives 19,20 or lipopolysaccharides with, for example, azidemodified analogues of N-acetylglucosamine, 21 3-deoxy-Dmanno-octulosonic acid 22 or legionaminic acid precursors. 23 However, very few bacterial species have been specifically labelled on their capsular polysaccharides by bioorthogonal chemistry. We could only identify two reports, in which capsules of commensal bacteria of the Bacteroides genus have been effectively tagged by MOE with modified N-acetylgalactosamine reporters. 24,25 The authors visualized the capsular component of prelabelled bacteria in live mice intestines, opening an avenue to better understand host-pathogen interactions and the evolution of related diseases.
To the best of our knowledge, capsular polysaccharides constituted of sialic acids have thus never been investigated by MOE in combination with click chemistry, despite the importance of these structures as virulence factors. Herein, we report a bioorthogonal labelling method to detect the polysialic acid capsule using alkyne-and azide-modified ManNAc and Neu5Ac reporters in E. coli K1 strains. Although this reporter toolbox is commonly used in mammalian systems to label N-glycoproteins, 26,27 it is the first time that these reporters are used successfully in E. coli. We aimed to take advantage of the fact that, unlike non-pathogenic strains, K1 strains are capable of synthesizing Neu5Ac de novo, and developed a test that allows facile and specific detection of the PSA capsule. The method was miniaturized into a microplate assay for transferability to screening approaches, which could serve as an interesting platform to decipher K1 pathways and identify external factors that influence PSA biosynthesis. Such assays could help in the discovery of new capsule-targeted classes of antibiotic drugs that decrease the virulence of a pathogenic bacterium while circumventing known resistance mechanisms.

Chemical reporters
Per-O-acetylated monosaccharides are commonly used for MOE in mammalian models. However, per-O-acetylation can cause issues as it may generate non-specific signal and false positives. 28,29 It may also be conducive to acidification of the intracellular milieu, metabolic perturbation due to partially acetylated forms, or cytotoxicity. 16 In addition, it has been suggested that low levels of non-specific esterase activity prevent the efficient release of the free sugar form thus thwarting the use of such reporters in some prokaryotic species including E. coli. 23 Conversely, this species is capable of active transport mediated ManNAc uptake via the ManXYZ transporter and can also scavenge exogenous sialic acid from the host via the sialic acid transporter NanT. We therefore decided to use unprotected reporters and synthesized N-(2-azidoacetyl)-D-mannosamine (ManNAz), N-4-pentynoyl-D-mannosamine (ManNAl), N-(2-azidoacetyl)-neuraminic acid (SiaNAz) and N-4-pentynoylneuraminic acid (SiaNAl) (see experimental section). CuAAC was chosen as the bioorthogonal reaction, owing to its fast kinetics and ease of use. In addition, the azide and terminal alkyne tags are easily interchanged thus allowing comparison of the chemical handle's impact.
CuAAC whole cell labelling of K1 expressing E. coli EV36 We first assessed whether these reporters could be taken up and metabolically incorporated into living E. coli EV36, a strain containing all fourteen genes of the kps PSA biosynthetic cluster that produces the K1 capsule. [30][31][32] EV36 cultures were incubated overnight in lysogeny broth (LB) medium complemented with the modified monosaccharides, prior to CuAAC labelling with an appropriate clickable tetramethylrhodamine dye bearing an alkyne or azide reactive moiety (TAMRA-N 3 for alkynyl reporters and TAMRA-Alk for azido reporters). The same fluorochrome was used in both instances in order to allow intensitybased comparisons. The reactive function is not p-conjugated to the rhodamine core owing to the presence of an oligoPEG spacer arm, and both dyes present the same photophysical characteristics (F = 0.1; e = 92.000 M À1 cm À1 , superimposable spectra). The assay was implemented using a fluorescence microplate reader to enable facile testing of multiple conditions. The bioorthogonal ligation was first optimized to obtain a robust labelling and satisfying signal-to-noise ratio in EV36 whole cells. Various parameters were screened, with TAMRA dye concentrations ranging from 62.5 nM to 25 mM in EV36 cultures incubated with reporters at a fixed 1 mM concentration, giving the best signal to noise ratio at 0.25 mM of fluorophore for 45 minutes ( Fig. 2A), and then with chemical reporter concentrations ranging from 100 mM to 2 mM at a fixed 0.25 mM TAMRA concentration, resulting in an optimal chemical reporter concentration found at 600 mM (Fig. 2B).
In contrast to what has been reported in activity-based protein profiling studies in mammalian cell lysates using dye concentrations 200 to 400 fold higher, 33,34 we observed a higher level of background using TAMRA-N 3 at 0.5 mM. This could be attributed to non-covalent interactions of azide dipoles at the bacterial cell surface at higher concentrations, or to the fact that alkyne-dyes exert less non-specificity in the nanomolar range.
Next, we sought to perform statistically robust tests to determine the significance of the difference in labelling after the incorporation of the chemical reporters at 600 mM and 1 mM. In order to allow transferability of the protocol to prospective screening approaches, we developed this assay in 96-well plates, using an intensity-based readout with a fluorescence microplate reader. For all four compounds, we observed significant labelling at 600 mM but not at 1 mM, except for ManNAz being the only analogue where the labelling was significant in both cases (Fig. 2C). This suggested that ManNAl, SiaNAl and SiaNAz might exert a negative effect on the physiological state or viability of bacteria at higher concentrations.
Experiments in which the bacteria were labelled with SiaNAz and SiaNAl generally exhibited a lower fluorescence intensity when compared to ManNAz and ManNAl, which came as a surprise. Indeed, these microbial cells are generally thought to be able to scavenge, process and install Sia derivatives but to metabolize ManNAc very slowly, 35,36 although this is highly species dependent. ManNAz led to the most efficient labelling. The noticeable difference in intensity between ManNAz and ManNAl tagging indicates that the azide handle may be more compatible with the metabolic network overall, or that the alkyne moiety might induce toxicity during the metabolic incorporation step. Distinct uptake and/or metabolisms should account for differences in labelling efficiency between ManNAc or Sia derivatives bearing the same chemical tag (e.g., ManNAz and SiaNAz). 36

Reporter accumulation in the capsule
We pursued by investigating whether the observed labelling originates from reporters incorporated into the K1 capsule. To show this, E. coli EV36 was grown in LB medium complemented with ManNAz or ManNAc, then the K1 capsular polysaccharides were extracted and purified according to a previously described procedure. 37 A polyacrylamide gel electrophoresis of the extracted K1 capsule was performed, followed by alcian blue staining (a dye that specifically reveals acidic polysaccharides 38 ) to confirm the presence and purity of PSA (Fig. 3A). To evaluate the incorporation of the chemical reporters in the isolated capsule, the extracted acidic polysaccharides were then labelled in solution by CuAAC and fluorescence intensity was measured as described hereabove. Specific labelling was observed with both ManNAc derivatives ManNAl and ManNAz, but not with the Neu5Ac analogues SiaNAl and SiaNAz ( Fig. 3B). Therefore, this indicates that chemical reporters ManNAz and ManNAl enter the bacterial cell (possibly through the ManXYZ-encoded transporter) and are converted to SiaNAz and SiaNAl, respectively, in the bacterial cytoplasmic compartment. 39 The derived tagged Sia is subsequently activated as CMP-Sia, polymerized and exported at the cell surface by various enzymes and transporters. 9 E. coli has been described to convert ManNAc to ManNAc-6-P during uptake to redirect the product on the degradation pathway. 39,40 Our results suggest that E. coli can also incorporate ManNAc into the K1 capsule biosynthesis via an unidentified pathway able to either transport ManNAc via the manXYZ phosphotransferase system, producing intracellular ManNAc-6-P which is then converted to ManNAc by an unknown phosphatase, or to transport ManNAc without ManNAc-6-P conversion. The former hypothesis carries a much stronger weight as the presence of this enzyme was previously suggested in

Paper RSC Chemical Biology
Neisseria meningitidis, which also expresses K1 capsules, though no proper characterization has yet been reported. 41 Furthermore, Neu5Ac is transported by the NanT sialate permease to the interior of the bacterial cells and is known to be directly degraded. 30,35,40,42,43 Our results are in good agreement with this observation. Neu5Ac is indeed transported inside the cell as demonstrated by the whole cell fluorescence observed (Fig. 2), but the monosaccharide is not directed to the PSA biosynthesis as suggested by the lack of signal detected on the purified capsule (Fig. 3B).
To check whether constitutive biomolecules other than the capsule were labelled by our strategy, we carried out the same experiments on whole E. coli BL21 cells, a B strain that does not express capsules. All four reporters were tested, and no labelling was observed on this strain (Fig. 2D), thereby confirming the specific incorporation of the reporters into the K1 PSA capsule in the EV36 bacterium.

Effect of unnatural monosaccharides on the K1 capsule biosynthesis
To ensure that the capsule was correctly expressed in the different labelling conditions, we performed immunofluorescence labelling of whole E. coli EV36 cells with an anti-K1 antibody followed by observation in fluorescence microscopy. A neat peripheral labelling pattern was observed, confirming the presence and integrity of capsular PSA. The experiments were carried out after growth in the presence of a ManNAc or Neu5Ac analogue (ManNAz and SiaNAl, respectively), or after the thorough rinsing and shaking conditions required during the CuAAC labelling protocol (''Shaken'' condition) (Fig. 4). These results demonstrate a robust expression of the K1 capsule, without side effects of the unnatural monosaccharides on the K1 pathways.
Impact of unnatural monosaccharides on the growth and physiological state of the bacteria.
To further investigate the results obtained in Fig. 2C, E. coli EV36 growth and viability were evaluated for ManNAz, ManNAl, SiaNAz, or SiaNAl at 1 mM and 0.6 mM in comparison to ManNAc and LB alone. To evaluate the impact of the chemical reporters on the bacterial growth we measured the optical density at 600 nm (OD 600 ) of LB suspensions at regular intervals over the course of several hours. No difference was observed between the various conditions, illustrating that the reporters do not impair cell growth (Fig. 5A). To evaluate the impact on long-term cell viability, serial dilutions of the suspensions were carried out after growth that were plated for viable counts. All chemical reporters except ManNAz induced rather significant toxicity on the long term when compared to ManNAc. Interestingly, when ManNAc was added to the medium the cell viability was also significantly lower by comparison to LB medium alone (Fig. 5B). ManNAz and ManNAc equally affect bacterial long-term viability to a mild extent. On the contrary, ManNAl drastically reduces the viability at both concentrations, revealing an increased toxicity linked to the appended pentynoyl moiety. This toxicity leads to decreased levels of specific PSA labelling (Fig. 2), and might enhance non-specific binding. Dead bacteria have been recently described to form nanotubes by membrane extrusion or vesicles. These structures may impact background levels. 44 Such an effect due to the nature of the bioorthogonal handle has been evidenced before in human Jurkat cells, where azidofucose derivatives proved to be more efficiently incorporated but more toxic than their alkynyl counterparts. 45 In the case of K1-expressing E. coli, it is the alkyne tag that decreases the viability of EV36 cells, which might be explained by the fact that a lower level of incorporation in capsular PSA leads to increased catabolism and accumulation of alkyne by-products interfering with other metabolic networks in the cell. Nevertheless, ManNAl is efficiently incorporated into the capsule (Fig. 3B).
It could thus be hypothesized that the pentynoyl sidechain is less compatible with the E. coli NeuB-encoded sialic acid synthase or NeuA-encoded CMP-sialic acid synthase than its 2-azido counterpart in ManNAz. This result calls attention to the use of unnatural monosaccharides to study bacterial physiology. To avoid any misinterpretation, viability assays should always be performed when testing MOE strategies on a new bacterial strain.

MOE microscopy
Given that ManNAz enables specific labelling of the capsule with no significant toxicity nor impact on growth, we chose to use this analogue for MOE followed by CuAAC labelling of the capsule for observation by fluorescence microscopy. After ManNAz incorporation, cells were treated with a CuAAC buffer whose composition was adapted from a previously reported procedure. 22 A bright, heterogeneous and specific labelling was observed after ManNAz incorporation compared to ManNAc as negative control (Fig. 6). All cells appear to incorporate the reporter. In most cells, the labelling is of rather low intensity (although significant when compared to the control) but on some others, a bright peripheric fluorescent signal was visible ( Fig. 6B and C), typical of capsule localisation. Additionally, a number of cells exert a labelling pattern at the poles of the bacterium (Fig. 6D), and most cells that do appear to be undergoing or to have just undergone cell division. This could be interpreted as a physiological state in which the older labelled PSA capsule has been relocated at the polar regions while new unlabelled capsule is being produced at the septum region during binary fission. Like most capsular polysaccharides in Gram-negative bacteria, PSA assembly and export through the periplasm is controlled by a multiprotein complex. The K1 antigen belongs to group 2, which identifies strains for which this export involves an ABC-transporter. 46 Our results might indicate that this translocation to the outer membrane is spatially regulated at the equatorial region during cell division, similarly to what has been proposed for Gram-positive Streptococcus pneumoniae, whose capsule export is mediated by the Wzx-Wzy system. 47 Indeed, Henriques et al. suggested that this spatial control ensures the coordination of capsule and cell wall biogenesis at the division site, leaving the bacteria unexposed to the host's immune system throughout. The heterogeneous labelling points out the incorporation of ManNAz with a potential dependence on the physiological state of the bacteria during their growth. Similarly, the bacteria need to adjust their metabolism to ManNAc and ManNAz import and successfully export the polysialic acid at the cell surface, which may also result in heterogeneity. Finer localisation may be obtained with super-resolution microscopy, to refine the visualization of capsule export zones at the poles of the cells. 48 In order to further confirm the specificity of our method for the K1 capsule and ascertain its applicability to various pathogenic bacteria, we then tested three other wild-type E. coli strains (Fig. 7). In addition to E. coli BL21, 49 a strain devoid of capsule mentioned hereabove as negative control (Fig. 2D), we also evaluated the ability of E. coli Nissle 1917 bacteria to metabolize ManNAz. The Nissle 1917 strain expresses another capsular acidic polysaccharide, namely the K5 antigen heparosan constituted of a disaccharide repeating unit that contains glucuronic acid and N-acetylglucosamine. 8,50 As expected, this strain was unable to incorporate ManNAz, further confirming the specificity of this assay for the K1 PSA capsule. Moreover, the two E. coli K1 pathogenic strains K-235 and U5/41, which produce significantly higher levels of PSA than the EV36 model, exhibited strong fluorescence (Fig. 7). This indicates an increased level of incorporation of SiaNAz units into PSA derived from the metabolization of ManNAz when compared to experiments on EV36. The heterogeneity was lesser in the K-235 and U5/41 wild-types and the pattern observed was consistent, with bacteria marked at the periphery typical of

Paper
RSC Chemical Biology capsule staining and other bacteria showing localization at the poles. In contrast, the heterogenous incorporation observed in the EV36 construct might be attributed to a less efficient metabolic channeling process, induced by the expression of exogenous genes from K1 strains in a K-12 background (e.g., the manXYZ operon). The presence of PSA in the K-235 and U5/41 strains and its absence in the Nissle 1917 strain were further confirmed by immunofluorescence using a fluorescent anti-K1 antibody (Fig. S1, ESI †). Thus, our method can be used to detect K1 antigen production in pathogenic strains, which could help decipher the dynamics of capsule expression and the factors that regulate it in future studies. Furthermore, it provides a platform for screening new types of antibiotics targeting K1 capsule metabolism.

Microplate fluorescence
Overnight cultures were adjusted to OD 600 1 by diluting the suspensions with PBS. 200 ml of these suspensions were split in 2 ml microtubes (minimum of 3 per condition) then centrifuged (2 min, 10 000 G), and pellets were resuspended in 200 ml CuAAC buffer and agitated for 45 min, 600 rpm at room temperature in the dark. These suspensions were rinsed 3 times with 1 ml PBS (2 min, 10 000 rpm) before being resuspended in 200 ml PBS. CuAAC whole cell suspensions were split in a dark opaque 96-well plate (100 ml per well). Fluorescence (l em 535 AE 20 nm/l exc 585 AE 30 nm) was measured on a CLARIOstar Plus microplate reader.

Fluorescence microscopy
Agar pads were made by pouring 10 ml of hot LB agar on a microscopy slide and quickly covering it with a glass coverslip. Whole-cell suspensions were deposited on the hardened agar pad by gently raising the coverslip and placing it back down. These slides were then observed on a Leica AF6000 LX inverted video microscope with differential interference contrast (DIC). For immunofluorescence staining of the capsule, overnight cultures were adjusted to 1 OD 600nm by diluting the suspensions with PBS and treated with anti-K1 rabbit antibody (ENZ-ABS559-0100 Enzo Life Sciences) (1/100, 45 min), rinsed with 200 ml PBS and treated with anti-rabbit AF488 antibody (1/250, 30 min), rinsed with 200 ml PBS and finally mounted on agar pads and observed as described above. For CuAAC fluorescence labelling, cells were treated as described for the microplate fluorescence with the fluorophore concentration adjusted to 100 mM. After the reaction and rinsing steps, cells were resuspended in 200 ml PBS and 10 ml of these suspensions were mounted on agar pad before observation on a fluorescence videomicroscope.

Viability assay and growth curves
For viability assay, OD 600 of overnight cultures was harmonized to 1 by diluting the suspensions with sterile PBS. 200 ml of these suspensions were transferred in a 96-well plate. Serial dilutions RSC Chemical Biology Paper ranging from 10 À1 to 10 À10 were performed. 20 ml of these suspensions were streaked on LB agar Petri dishes and grown overnight at 37 1C. Petri dishes showing between 5 to 250 colonies were selected, colonies counted, and the number of CFU in the original suspension was deducted from the dilution. For growth curves, from an isolated colony, 2 ml of LB liquid culture were grown in a 15 ml conical tube for 2 hours at 37 1C. 10 ml of this suspension was used to inoculate 10 ml LB supplemented with the desired chemical reporter or monosaccharide in a 50 ml conical tube. OD 600 of these suspensions was then recorded at regular intervals.

Capsule extraction and quantification
Capsular extracts were purified by following a method described previously. 37 Overnight grown cultures were suspended in 600 ml lysis buffer (100 mM SDS, 50 mM Tris, 0.128 mM NaCl). 600 ml phenol/chloroform/isoamyl alcohol (25 : 24 : 1) were added and agitated 15 min at 65 1C. After centrifugation (16 000 g for 15 min at 4 1C), the upper phase was transferred and completed with 600 ml of ice-cold absolute ethanol. The tubes were stored overnight at À20 1C. Tubes were then centrifuged (10 min, 13 000 g) and the white precipitate was rinsed with ethanol then dried under N 2 flow. 20 ml DNAse was added (45 min, 37 1C, 300 rpm) followed by 20 ml Proteinase K (1 h, 56 1C, 300 rpm). 560 ml H 2 O and 600 ml phenol/chloroform/isoamyl alcohol (25 : 24:1) were added. Tubes were centrifuged (15 min, 13 000 rpm) and the upper phase transferred in a new microtube. 200 ml H 2 O, 50 ml sodium acetate 3 M and 1 m absolute ethanol were added and the tubes were stored at À20 1C overnight. The tubes were centrifuged (15 min, 13 000 rpm) and the white precipitate was rinsed with 1 ml ice cold absolute ethanol and dried under N 2 flow. Capsular extracts were finally resuspended in 50 ml of H 2 O. The capsule extracted was labelled and quantified on a microplate reader as follow: 10 ml of the capsular extract solutions were added to 190 ml CuAAC buffer in a microtube and agitated 45 min, 600 G at RT. After the reaction 50 ml sodium acetate 3 M and 1 ml ice cold absolute ethanol were added and tubes were stored overnight at À20 1C.
Tubes were then centrifuged (15 min, 13 000 G and the white pellet was rinsed with ice cold absolute ethanol, dried under N 2 flow, resuspended in 20 ml H 2 O and transferred to a dark opaque 96-well plate for fluorescence readout on a microplate reader. The purity of the capsular extract was performed by polyacrylamide gel electrophoresis, for this 10 ml of the same capsular preparation were loaded on a 15% TBE-PAGE (Tris-Boric-EDTA polyacrylamide gel electrophoresis) as described previously. 51 The capsular polysaccharides were stained with 5% alcian blue in water for 15 min followed by three washing steps with acetic acid 5% in a solution of 50% methanol until the gel background get unstained.

Conclusions
We developed a method for the specific bioorthogonal labelling of K1 capsules in E. coli after the metabolic incorporation of ManNAc analogues equipped with alkyne and azide chemical handles. These sialylation reporters did not exert any significant effect on bacterial growth. ManNAz was determined as the better choice among the tested reporters, as ManNAl showed inherent long-term cytotoxicity. While both ManNAc and Neu5Ac derivatives were readily incorporated in growing E. coli EV36, resulting in fluorescent labelling of whole bacteria, only the signal obtained from ManNAc chemical reporters could be attributed to their incorporation into the polysialic acid capsule. This allowed us to refine our understanding of the capsule metabolic pathways. The method was miniaturized as a microplate assay amenable to screening approaches. With the ability to track the K1 capsule biosynthesis, this platform might be a useful tool for future studies aiming at impacting capsule expression, which is of great interest in the context of increasing pathogen resistance.

Author contributions
V. R. and Y. R. performed the experiments. V. R., Y. R. and C. L. analysed and interpreted the data. C. L., Y. R., and V. R. wrote the manuscript. V. R. and C. L. prepared figures. C. L. and C. B. conceptualised the project. C. B. and C. L. acquired funding and supervised the work. C. L., V. R. and Y. R. revised the paper.

Conflicts of interest
There are no conflicts to declare.