Chitin, Chitin Oligosaccharide, and Chitin Disaccharide Metabolism of Escherichia coli Revisited: Reassignment of the Roles of ChiA, ChbR, ChbF, and ChbG

Chitin is a homopolymer composed of β-1,4 linked N-acetylglucosamine (GlcNAc) found in the cell walls of fungi, exoskeletons of crustaceans and insects, as well as the shells of molluscs. It is the most abundant nitrogen-containing polysaccharide on earth, with an annual turnover rate of 10¹¹ tons, therefore playing a central role in the global nitrogen and carbon cycle [Tharanathan and Kittur, 2003]. Degradation of the recalcitrant chitin biopolymer is achieved by chitinolytic fungi as well as bacteria, e.g., gram-positive bacteria of the genus Bacillus or Streptomyces, and gram-negative bacteria of the genus Serratia, Aeromonas, or Vibrio, which play an important role in chitin recycling in soil and aquatic (marine) ecosystems [Blokesch, 2012; Hamid et al., 2013; Adams et al., 2019; Wucher et al., 2019]. These organisms secrete various forms of chitinases and N-acetylglucosaminidases that fragment polymeric chitin into shorter chitin oligosaccharides, N,N′,N′′-triacetylchitotriose (GlcNAc-GlcNAc-GlcNAc, triacetyl-chitotriose), N,N′-diace­tylchitobiose (GlcNAc-GlcNAc, diacetyl-chitobiose), and GlcNAc [Itoh and Kimoto, 2019]. Furthermore, they possess uptake and catabolic systems to grow on chitin and chitin oligosaccharides [Hamid et al., 2013; Itoh and Kimoto, 2019].

Escherichia coli K12 is unable to grow on chitin and chitin oligosaccharides, despite possessing a functional chitinase (ChiA/YheB) and an outer membrane porin that facilitates chitin oligosaccharide uptake (ChiP) [Francetic et al., 2000a; Soysa and Suginta, 2016]. Under normal laboratory growth conditions, the expression of ChiA and its secretion via the type 2 secretion system is silenced, requiring a mutation in the hns gene for activation [Francetic et al., 2000a, b]. As ChiA contains a native signal sequence, in an hns background, the chitinase is expressed and partially secreted to the periplasm via the Sec system and to the extracellular space via the type 2 secretome [Francetic et al., 2000a, b]. We became interested in the cryptic ChiA, since it had been described as a bifunctional enzyme with endo-chitinase and lysozyme activity, an assumption that, however, solely relied on the enzyme’s ability to cleave the artificial lysozyme substrate ethylene glycol-chitin [Francetic et al., 2000a]. We thus aimed to re-examine the function of ChiA and chitin metabolism in E. coli.

Growth of E. coli on diacetyl-chitobiose and triacetyl-chitotriose depends on the chitobiose operon chb­BCARFG. This operon was initially identified as the cryptic cel (cellobiose utilization) operon, which requires secondary mutations in order to metabolize β-glycosides such as cellobiose, arbutin, and salicin [Parker and Hall, 1990]. Later it was shown that the operon is not cryptic but induced by diacetyl-chitobiose, an end product of chitin degradation, and therefore renamed to chb [Keyhani and Roseman, 1997]. Encoded by the chbBCARFG operon are components of a phosphotransferase system (PTS) transporter (ChbBCA; EIIB, EIIC, and EIIA) involved in both, the transport of diacetyl-chitobiose and triacetyl-chitotriose and their simultaneous phosphorylation at the 6-hydroxyl group of the GlcNAc at the non-reducing end [Keyhani et al., 2000a, b, c]. Further encoded by this operon is a transcriptional repressor/activator (ChbR), repressing chb transcription by default and acting as an activator upon effector binding [Plumbridge and Pellegrini, 2004]. The chb operon is also subject to catabolite repression and is repressed by NagC, the transcriptional regulator of the GlcNAc catabolism, which is derepressed upon GlcNAc6P binding [Plumbridge and Pellegrini, 2004]. The enzymes ChbG and ChbF are involved in the intracellular processing of the phosphorylated diacetyl-chitobiose (diacetyl-chitobiose-P) and triacetyl-chitotriose (triacetyl-chitotriose-P). ChbG was identified as a monodeacetylase that is essential for growth on diacetyl-chitobiose and triacetyl-chitotriose but is dispensable for growth on cellobiose and chitobiose (chitosan dimer; GlcN-β-1,4-GlcN) [Verma and Mahadevan, 2012]. Verma and Mahadevan [2012] further showed that activation of the chb promoter by the regulatory protein ChbR is dependent on ChbG, suggesting that deacetylation of diacetyl-chitobiose-P and triacetyl-chitotriose-P is necessary for their function as co-activators of ChbR. The authors proposed that ChbG removes the N-acetyl group from the reducing end, yielding GlcNAc6P-GlcN and GlcNAc6P-GlcNAc-GlcN, which may act as the inducers of the chb operon upon ChbR binding [Verma and Mahadevan, 2012]. Subsequent cleavage of these saccharides would yield GlcNAc6P, which upon binding to NagC could relieve repression of the chb operon. We questioned these assumptions, because the ChbF protein, which is responsible for the hydrolysis of the ChbG products, monoacetyl-chitobiose-P and diacetyl-chitotriose-P, was initially characterized as cellobiose-6-phosphate (Glc6P-Glc) hydrolase (named CelF) [Parker and Hall, 1990; Thompson et al., 1999]. Since ChbF cleaves Glc6P-Glc and other 6-phospho-glucosides (e.g., arbutin-P and salicin-P) and as well as chitobiose-P (GlcN6P-GlcN), we suggest that ChbF likely cleaves GlcN6P-glycosides (GlcN6P-GlcNAc and GlcN6P-GlcNAc-GlcNAc), rather than GlcNAc6P-glycosides. ChbF is assigned to the family 4 of glycosidases (GH4). The proposed mechanism for this class of enzymes is very distinct from that of classical glycosidases, which operate via nucleophilic substitution-mechanisms. GH4 glycosidases instead depend on a divalent metal ion and NAD⁺ as cofactors and catalyze an oxidation/reduction of the hydroxyl group at C3 of the sugar at the non-reducing end of a disaccharide. Formation of an intermediate carbonyl at C3 renders the C2 proton more acidic and thereby facilitates the elimination of the glycosidic oxygen, that is, the cleavage of the glycosidic bond [Rajan et al., 2004; Yip et al., 2004, 2007; Yip and Withers, 2006a, b]. Since ChbF operates by an mechanism that involves reactions at C3 and C2, the N-acetyl group at C2 likely interferes with this mechanism [Thompson et al., 1999]. Here we revised the function of ChbF as well as of ChbG and unraveled the nature of the effector molecules of the transcriptional regulator ChbR. Furthermore, we characterized ChiA as a bona fide chitinase that cleaves chitin but not chitosan and peptidoglycan, thus disproving its previously suggested role as a bifunctional endochitinase/lysozyme.

To revisit the functions of ChbG and ChbF encoded by the chb operon, we first tested mutants deficient in either chbG or chbF for growth on diacetyl-chitobiose and chitobiose as the sole carbon source in M9 minimal medium, compared to the parental E. coli strain (wild type; wt). E. coli wt grew overnight to a final OD₆₀₀ of 4.2 in M9 medium supplemented with 2 mM diacetyl-chitobiose, while it failed to grow with chitobiose (data not shown). It has previously been described that E. coli wt grows with diacetyl-chitobiose but poorly with chitobiose as the only carbon source, presumably due to the lack of induction of the chb operon [Verma and Mahadevan, 2012]. As expected, the ΔchbG::kan and ΔchbF::kan strains were unable to grow on either diacetyl-chitobiose and chitobiose (data not shown). To identify accumulation products in ΔchbG::kan and ΔchbF::kan, we grew the mutants overnight in M9 minimal medium with a mixture of amino acids (casamino acids; CAA) that was supplemented with 2 mM of either diacetyl-chitobiose or triacetyl-chitotriose, isolated the cytosolic fractions, and analyzed them via high-performance liquid chromatography-mass spectrometry (HPLC-MS). In the cytosolic fraction of ΔchbF::kan cells, we detected large amounts of compounds with masses corresponding to monoacetyl-chitobiose-P (M+H)⁺ = 463.1324 m/z, which appears as a double peak due to separation of the α- and β-anomeric forms, and to monoacetyl-chitobiose (M+H)⁺ = 383.1660 m/z (Fig. 1a; a list of compounds with their corresponding monoisotopic masses can be found in Table 1). These compounds are likely the products of the activity of the monodeacetylase ChbG, which accumulate due to the lack of ChbF. The additional appearance of a dephosphorylated product indicates that a phosphatase partially cleaves the accumulating monoacetyl-chitobiose-P. To test whether the dephosphorylation event is growth phase-dependent, cytosolic fractions were isolated also from exponential phase-cells (OD₆₀₀ = 1). However, no difference in the relative amount of the phosphorylated and non-phosphorylated accumulation products was found in comparison to the stationary phase cells, indicating the expression of the enigmatic phosphatase is not growth phase-dependent (data not shown). Similarly, when the ΔchbF::kan strain was grown in CAA-M9 medium supplemented with 2 mM triacetyl-chitotriose, HPLC-MS analysis of cytosolic fractions revealed the presence of compounds with masses corresponding to diacetyl-chitotriose-P (M+H)⁺ = 666.2117 m/z (Fig. 2a) and diacetyl-chitotriose (M+H)⁺ = 586.2454 m/z, which both appear as double peaks due to separation of the α- and β-anomeric forms. These compounds are the products of the activity of the monodeacetylase ChbG, which apparently deacetylates the trisaccharide-P and possibly also the non-phosphorylated trisaccharide. The appearance of large amounts of diacetyl-chitotriose indicates that the enigmatic phosphatase preferentially cleaves diacetyl-chitotriose-P as compared to monoacetyl-chitobiose-P (cf. Fig. 1a).

Conversely, in the ΔchbG::kan cytosolic fractions, the expected accumulation products, diacetyl-chitobiose-P and triacetyl-chitotriose-P could not be detected (data not shown). One could speculate that these saccharides do not accumulate because they are metabolized by an unknown, ChbG-independent pathway. If diacetyl-chitobiose is metabolized via a different pathway and thus is used as a source of energy and carbon, growth on this saccharide would be expected. The ΔchbG::kan mutant, however, failed to grow with diacetyl-chitobiose. It has previously been reported that ChbG is crucial for the generation of monoacetyl-chitobiose, which supposedly acts as the inducer of the chb operon [Verma and Mahadevan, 2012]. Thus, the more plausible explanation for the absence of an accumulation product is that no induction occurred and, consequently, the ChbBCA-transporter was not expressed and diacetyl-chitobiose was not taken up.

To further characterize the accumulation products of the ΔchbF::kan strain, the gene encoding the 6-phospho-β-glycosidase ChbF (CelF) was cloned and recombinantly expressed. Recombinant ChbF completely degraded the phosphorylated accumulation product monoacetyl-chitobiose-P and yielded products with masses corresponding to GlcN-P (M+H)⁺ m/z = 260.0530 and GlcNAc (M+H)⁺ m/z = 222.0972, but was not able to cleave the non-phosphorylated accumulation product (Fig. 1b). It has previously been shown that diacetyl-chitobiose is phosphorylated at the non-reducing end during its transport into the cell by the ChbBCA-PTS system [Keyhani et al., 2000a]. On this basis, the cleavage of the accumulation product of ΔchbF::kan into GlcN-P and GlcNAc by ChbF suggests that ChbG deacetylates diacetyl-chitobiose-P at the non-reducing end, yielding GlcN6P-GlcNAc. To unequivocally exclude that the dephosphorylated accumulation product of ΔchbF::kan is GlcNAc-GlcN, we treated it with Bacillus subtilis exo-N-acetylglucosaminidase NagZ. This enzyme plays a role in the recycling of peptidoglycan and specifically cleaves off GlcNAc residues from the non-reducing end [Litzinger et al., 2010a, b]. While GlcNAc is released from diacetyl-chitobiose with NagZ (data not shown), no product is observed during digestion of the accumulation products of ΔchbF::kan, indicating the absence of GlcNAc entities at the non-reducing terminus (Fig. 1c). Similarly, it was shown that ChbF is able to cleave diacetyl-chitotriose-P, which accumulates in E. coli ΔchbF::kan cells grown in the presence of triacetyl-chitotriose, yielding diacetyl-chitobiose and GlcN-P (Fig. 2b). This indicates that triacetyl-chitotriose is also phosphorylated and deacetylated at the non-reducing end by the ChbBCA-PTS and ChbG, respectively, yielding GlcN6P-GlcNAc-GlcNAc, from which GlcN6P at the non-reducing end is cleaved-off by ChbF.

A method to identify the sugar at the reducing end beyond doubt is the reduction with sodium borohydride followed by enzymatic cleavage and/or fragmentation by tandem MS (MS/MS). In these conditions, the reduction of the accumulation products of ΔchbF::kan was complete and the two reduced products could be detected (Fig. 3a). When ChbF was added, the peak of the phosphorylated, reduced substrate became smaller and a distinct peak emerged, with a mass that corresponds to reduced GlcNAc (M+H)⁺ m/z = 224.1129 (Fig. 3b). A mass for GlcN-P, which would also be expected, was not detected. GlcN-P does not ionize very well in MS and its ionization could have been worsened in the samples that were reduced with sodium borohydride due to high salt concentrations. These conditions are also suboptimal for ChbF and, accordingly, the substrate was not completely digested by ChbF. Nevertheless, the detection of reduced GlcNAc ([M+H]⁺ m/z = 224.1129) in significant abundance (83% relative to the reduced monoacetyl-chitobiose-P, which was set to 100%) indicates that there had been GlcNAc at the reducing end of the ChbF substrate (Fig. 4). Both compounds could also be detected in anhydrous forms with lower intensities. The mass difference between reduced monoacetyl-chitobiose-P and reduced GlcNAc corresponds exactly to uncharged GlcN-P -H₂O (M) = 241.0351 Da. Furthermore, the mass of GlcN-P -H₂O itself could be detected, although in very low abundance of <0.5%. This indicates once more that GlcNAc ionizes readily in MS positive ion mode compared to GlcN-P.

In summary, our results indicate that ChbG deacetylates its substrate diacetyl-chitobiose-P at the non-reducing end. This finding fits well into the previous reports that E. coli is able to utilize the β-glucoside sugars cellobiose, arbutin and salicin by expressing the chb (formerly designated cel) operon [Parker and Hall, 1990]. In each of these sugars, the non-reducing end consists of glucose, while the residue at the reducing end is variable. These β-glucosides, after phosphorylation, were shown to be substrates for ChbF (previously named CelF) [Thompson et al., 1999]. Our data also fit well into the proposed enzymatic mechanism of GH4 members, which posits that the decisive reaction mechanism takes place at the carbohydrate at the non-reducing end. Indeed, oxidation/reduction of the hydroxyl group at C3 of the sugar at non-reducing end of a disaccharide to a carbonyl renders the C2 proton more acidic and thereby facilitates the elimination of the glycosidic oxygen, that is, the cleavage of the glycosidic bond [Rajan et al., 2004; Yip et al., 2004, 2007]. In the second step of the proposed mechanism, a proton gets released from the non-reducing end sugar’s C2 hydrogen to a catalytic base. This process is favored by a positive dipole on the C2, which is more pronounced with an amino instead of a N-acetyl functional group. The binding of Mn²⁺ to C2 and C3 functional groups would also be enhanced with the more negative dipole on the amino group. A well-studied member of GH4 is the 6-phospho-α-glucosidase GlvA from B. subtilis [Yip et al., 2007]. The crystal structure of GlvA bound to glucose 6-phosphate (Glc6P) revealed that at C2 there is relatively little space in the binding pocket and steric hindrance could occur with larger side chains. An amino acid sequence alignment among ChbF, GlvA, and other described GH4 enzymes revealed high similarities especially at the site next to the C2 functional group, where a cysteine residue is conserved across all queried enzymes. To further clarify the difference between amino and N-acetyl groups at C2 for ChbF we modeled its structure using the GlvA-Glc6P cocrystal structure (pdb code: 1U8X) and exchanging the C2 hydroxyl group of the ligand with an amino or N-acetyl group. Two amino acids are in proximity of the functional group, the conserved cysteine and methionine in GlvA, the latter replaced with the isoleucine of ChbF. This modeling suggests that an amino group fits well between the residues (Fig. 5a), whereas the N-acetyl group on the other hand does not, as the oxygen collides with the isoleucine and the methyl group collides with the sulfur of cysteine (Fig. 5b).

To test in vivo the enzymatic activity of ChbF on diacetyl-chitobiose-P, E. coli ΔchbG::kan was transformed with a plasmid that allows IPTG-inducible expression of the ChbBCA-PTS. Since under these conditions, due to the missing ChbR coactivator, the natural induction of the chb operon does not occur, we expected the accumulation of diacetyl-chitobiose-P, when grown overnight in M9 CAA minimal medium supplemented with diacetyl-chitobiose. Accordingly, diacetyl-chitobiose-P (M+H)⁺ m/z = 505.1429 accumulated in the cytosolic fraction (Fig. 6a). Intriguingly, ChbF is not able to digest this compound (Fig. 6b), which confirms the inability of ChbF to take substrates with an N-acetyl group at the non-reducing end. Notably, as seen throughout in the accumulation studies, also here phosphorylated and non-phosphorylated saccharide represented the major accumulation product. This further suggests the action of a phosphatase that partially dephosphorylates the accumulation product GlcNAc6P-GlcNAc.

E. coli ChiA was previously characterized as an endo-chitinase, since it cleaved the artificial chitinase substrate 4-methylumbelliferyl (4-MU)-(GlcNAc)₃ but not 4-MU-GlcNAc [Francetic et al., 2000a]. Furthermore, in the latter report, ChiA was described as a bifunctional chitinase/lysozyme, on the basis of its ability to cleave ethylene-glycol chitin (EGC), which is a common lysozyme substrate [Wang and Chang, 1997]. We aimed to re-evaluate the enzyme’s specificity using natural substrates, chitin (-oligomers), chitosan (-oligomers), and peptidoglycan, and thus expressed and purified ChiA as a polyhistidine-tagged recombinant protein. Hydrolysis of 4-MU-(GlcNAc)₃ (Sigma-Aldrich) was confirmed with the recombinant protein (data not shown). A preferred substrate of chitinases is colloidal chitin, which was prepared by the acetylation of low molecular weight chitosan (<3 kDa in size; obtained from Carbosynth). ChiA was able to cleave colloidal chitin yielding diacetyl-chitotriose and triacetyl-chitotriose (Fig. 7b), whereas an endo-chitinase from Streptomyces griseus, in contrast, almost exclusively yielded diacetyl-chitobiose (Fig. 7c). That ChiA releases dimers and trimers in approximately equal amounts could be due to the fact that the enzyme features five chitin-binding type-3 domains which mediate binding to the glycan chains. To understand whether ChiA requires acetylated sugars, we incubated it with chitosan oligomers up to 3 kDa in size. The oligomer mix contained large amounts of chitotriose and chitotetraose and smaller amounts of chitobiose (Fig. 8a). Incubation with ChiA did not result in the release of small chitosan-oligosaccharides, but released significant amounts of diacetyl-chitotriose (Fig. 8b). As the diacetyl-chitotriose product of ChiA could not be further cleaved by the exo-N-acetylglucosaminidase NagZ from B. subtilis, we suggest the diacetyl-chitotriose product of ChiA is composed of GlcN-(GlcNAc)₂. These data indicate that ChiA is able to cleave particular bonds within polymeric, acetylated chitosan releasing products that have GlcNAc at the reducing end. As a positive control, the chitosanase Csn from B. subtilis was cloned, expressed, and purified and the digestion was performed according to the conditions described [Rivas et al., 2000]. After incubation of the chitosan oligomers with Csn an increased amount of chitobiose could be measured, while chitotetraose was completely digested (Fig. 8c).

To investigate the lysozyme activity of ChiA, peptidoglycan was purified from E. coli and B. subtilis. Unlike the positive controls, Atl (Glc), the glucosaminidase domain of the bifunctional N-acetylmuramoyl-L-alanine amidase/endo-β-N-acetylglucosaminidase Atl from Staphylococcus aureus [Oshida et al., 1995], and mutanolysin, the endo-N-acetymuramidase from Streptomyces sp. [Rau et al., 2001], ChiA did not release products from the peptidoglycans of either source. From E. coli peptidoglycan, Atl (Glc) released MurNAc-GlcNAc in smaller and MurNAc-GlcNAc-tetrapeptide ([MurNAc-GlcNAc]-4P) in larger quantities (Fig. 9c). From B. subtilis peptidoglycan, Atl released disaccharides with and without peptides (online suppl. Fig. S1C; for all online suppl. material, see www.karger.com/doi/10.1159/000515178). To test if ChiA possibly only cleaves denuded peptidoglycan, that is, poly-GlcNAc-MurNAc chains devoid of peptides, the N-acetylmuramoyl-L-alanine amidase CwlC from B. subtilis [Shida et al., 2000] was used to remove the peptide stems from the peptidoglycan as reported elsewhere [Müller et al., 2021]. CwlC quantitatively released tripeptides (3P) and tetrapeptides (4P) as well as crosslinked 4P-4P and 3P-4P peptides (Fig. 9d) from E. coli peptidoglycan. From B. subtilis peptidoglycan it released 3P*, 3P*-4P*, and 3P-4P*, in which the amino acid meso-diaminopimelic acid is amidated (assigned with *) (online suppl. Fig. S1E). However, ChiA was not able to release any products from denuded peptidoglycan chains (Fig. 9e; online suppl. Fig. S1F). Atl (Glc), on the other hand, releases significant amounts of MurNAc-GlcNAc as the only product in both cases. The fact that no disaccharide with peptides was released using Atl (Glc) indicates that CwlC has completely removed the peptides (Fig. 9f; online suppl. Fig. S1G). Mutanolysin, in contrast, cleaved native peptidoglycan, but was unable to degrade denuded peptidoglycan (online suppl. Fig. S1H). In summary, the results show that ChiA can digest chitin but neither chitosan nor peptidoglycan polymers irrespectively of whether they are native or denuded.

With the sole role of ChiA as chitinase confirmed in vitro, the growth of E. coli wt and ΔchiA::kan in M9 glucose minimal medium was compared in the presence of diacetyl-chitobiose or penta-acetyl-chitopentaose. As expected both strains were able to grow on glucose and diacetyl-chitobiose exhibiting a biphasic growth pattern due to carbon catabolite repression (Fig. 10b). With penta-acetyl-chitopentaose no biphasic growth was observed, as only glucose was utilized (Fig. 10c). However, the wt strain was expected to utilize penta-acetyl-chitopentaose if ChiP and ChiA were active. This was not the case, because chiA expression is usually silenced by the nucleotide-structuring protein H-NS [Francetic et al., 2000a]. In contrast, ChiP expression is controlled by the small RNA ChiX but, expression of the chb transcript relives chiP inhibition by functioning as trap-RNA [Figueroa-Bossi et al., 2009; Rasmussen et al., 2009]. Thus, expression of ChiA in the periplasm should allow growth on chitooligomers (e.g., penta-acetyl-chitopentaose), which are transported through ChiP into the periplasm. Cleavage by ChiA releases diacetyl-chitobiose and triacetyl-chitotriose, which can be taken up by the ChbCBA-PTS and metabolized by ChbG and ChbF. To test this, ChiA was artificially expressed in E. coli BL21 (DE3) from plasmid pET22b-chiA in an IPTG-inducible manner. Compared to E. coli K12 wt during growth in M9 CAA (to avoid carbon catabolite repression) minimal medium, the expression of ChiA clearly restored the availability to utilize penta-acetyl-chitopentaose (Fig. 10d). To exclude that E. coli BL21 (DE3) expressed endogenous ChiA we monitored its growth on penta-acetyl-chitopentaose and observed the same growth pattern as that of E. coli K12 wt (data not shown). This clearly shows that E. coli has the potential to utilize chitin oligosaccharides in a ChiA-dependent manner. However, since E. coli ChiA expression and secretion is silenced, the bacterium either relies on scavenging chitinase end products from other organisms or, if there is enough evolutionary pressure, selection against H-NS repression of ChiA and the type 2 secretome might occur.

In this study we refined the understanding of the catabolic pathways of diacetyl-chitobiose (GlcNAc-GlcNAc) and triacetyl-chitotriose (GlcNAc-GlcNAc-GlcNAc) in E. coli, as summarized in Figure 11. Catabolism of the saccharides involves phosphorylation via the ChbBCA PTS transporter, as shown previously [Keyhani et al., 2000a, b, c], and subsequent deacetylation via the monodeacetylase ChbG, which specifically removes the N-acetyl group from non-reducing end terminal GlcNAc6P entities, as shown in this study. Since activation of the chb promoter by the transcriptional regulator ChbR depends on the ChbG activity [Verma and Mahadevan, 2012], the correction of the function of ChbG indirectly identified GlcN6P-GlcNAc and GlcN6P-GlcNAc-GlcNAc as the inducers of the chb operon, which upon binding to ChbR activate chbBCARFG transcription. These saccharides were found to accumulate in a ΔchbF strain, thus constituting the natural substrates of ChbF. Intriguingly, ChbF failed to cleave GlcNAc6P-GlcNAc, because the N-acetyl group of GlcNAc6P-glycosides sterically interferes with amino acids in the active site pocket and thus prevents binding to ChbF, as shown by structural modeling. Possibly, the unusual redox/elimination-type mechanism of ChbF, belonging to family 4 of Cazy glycosidases (URL: http://www.cazy.org/GH4.html) [Henrissat, 1991; Davies and Henrissat, 1995; Yip and Withers, 2012], excludes GlcNAc6P-glycosides as suitable substrates. Accumulation of non-phosphorylated saccharides in the chbF mutant indicates that an unidentified phosphatase is active intracellularly that cleaves the phosphorylated 6P-glycosides. We assume that the dephosphorylation occurs due to the accumulation of high concentrations of phosphorylated saccharides and, thus, likely constitutes a detoxification mechanism counteracting sugar phosphate stress. The identity of the phosphatase is so far unclear; however, it could be identical to the phosphatase YigL, which has been reported to relieve glucose-phosphate stress in E. coli [Kuznetsova et al., 2006; Sun and Vanderpool, 2013].

The reason for this particular catabolic pathway in E. coli is unclear, but may be linked to the function of the chitinase ChiA of E. coli. We showed that the chitinase ChiA specifically cleaves chitin as well as acetylated chitosan, yielding diacetyl-chitobiose/triacetyl-chitotriose and GlcN-GlcNAc-GlcNAc, respectively, but does not cleave intact or peptide-free, denuded peptidoglycan as well as chitosan. Chitin in nature barely occurs as a fully acetylated polymer. Since ChiA is able to cleave GlcNAc-beta-1,4-GlcN bonds, it thus generates saccharides that upon uptake and phosphorylation by the ChbBCA PTS directly act as inducers of the chb operon. A still open question is why ChiA expression is silenced in E. coli. It is appealing to assume that E. coli laboratory strains have adapted to the use of LB- or similar media made up of yeast extract, which contains diacetyl-chitobiose and triacetyl-chitotriose but lacks chitooligomers; thus, ChiA is dispensable under these condition. It could be interesting to screen for undomesticated E. coli strains that lack a silent ChiA system and hence retain their ability to grow on polymeric chitin.

The bacterial strains used in this study are listed in online supplementary Table S1. B. subtilis 168 wild-type and E. coli K12 were used for the isolation of peptidoglycan. Bacteria were cultured at 37°C in lysogeny broth (LB broth Lennox, Carl Roth) with continuous shaking at 140 rpm. Overnight cultures (∼16 h) were used to inoculate fresh LB medium. Cells were harvested by centrifugation (3,000 × g, 20 min, 4°C). E. coli BL21 (DE3) cells (New England Biolabs) used to heterologously express recombinant enzymes were grown in LB medium supplemented with 50 µg/mL kanamycin or 100 µg/mL ampicillin depending on the expression plasmid until OD₆₀₀ 0.7 was reached, followed by induction with 1 mM IPTG and further propagation for 3 h. Cells were harvested by centrifugation (3,000 × g, 20 min, 4°C) and used for the purification of recombinant proteins. For the extraction of accumulation products E. coli ΔchbG::kan, ΔchbF::kan, and E. coli ΔchbG::kan pQE32-chbBCA were grown at 37°C o/n in 5 mL M9 minimal medium [Harwood and Cutting, 1990] with 6 g/L casamino acids (CAA) instead of glucose and 2 mM diacetyl-chitobiose. The growth curves were conducted in a Tecan plate reader at 37°C in 24-well plates filled with 400 µL M9 minimal medium [Harwood and Cutting, 1990] either with 0.1% glucose or 0.1% CAA and 0.1% of test substrate (diacetyl-chitobiose, triacetyl-chitotriose or penta-acetyl-chitopentaose; obtained from Carbosynth).

For the overexpression and purification of the enzymes used in this work, E. coli K12 ChbF (JW1723) and ChiA (JW3300), B. subtilis 168 Csn (BSU26890), NagZ (BSU01660), and CwlC (BSU17410), and S. aureus USA300 Atl (Glc) (ABD22514.1), the corresponding genes were amplified by PCR using the respective genomic DNA and primers listed in online supplementary Table S1. Prior to its insertion in pET28a, the NcoI site in cwl was deleted. The PCR products were purified (GeneJET purification kit and GeneRuler, 1-kb marker, Thermo Fisher Scientific). PCR product and vector (specified in primer name in online suppl. Table S1) were digested with restriction enzymes (specified in primer name in online suppl. Table S1; New England Biolabs) and ligated with T4 DNA ligase (Thermo Fisher Scientific). Chemically competent E. coli DH5α cells were transformed using the constructed plasmids and selected for growth on the respective antibiotic. From positive clones plasmids were purified and checked by sequencing.

The positive plasmids were used to transform E. coli BL21 (DE3) cells, which were grown as described above and lysed in a French pressure cell. Recombinant His-tagged enzymes were purified by Ni²⁺-affinity chromatography using a 1-mL HisTrap column (GE Healthcare) followed by size exclusion chromatography on a HiLoad 16/60 Superdex 200 pg column (GE Healthcare) and purity was checked with a 12% SDS-PAGE. From a 1-L culture 3.6 mg ChbF, 1.75 mg CwlC, and 12.5 mg Csn were obtained. The enzymes were stored at −20°C in 20 mM Na₂HPO₄ buffer (500 mM NaCl, pH 7.6). Gibson Assembly was used for the construction of a PTS transporter mutant in E. coli ΔchbG::kan. The corresponding gene sections for chbBCA were amplified by PCR from E. coli K12 genomic DNA using the appropriate primers (online suppl. Table S1). The purified PCR product was cloned via Gibson Assembly into pQE32, which was linearized via digestion with BseRI and HindIII. Chemically competent E. coli ΔchbG::kan cells were transformed with pQE32-chbBCA. Positive clones selected by antibiotics were confirmed by sequencing.

For the preparation of B. subtilis 168 cell walls 2 L of culture (exponential growth phase, OD₆₀₀ = 1.0) were harvested and resuspended in 30 mL piperazine-acetate buffer (50 mM, pH 6) with 12 U proteinase K and boiled for 1 h. The cytosolic fractions were removed by centrifugation (3,000 × g, 15 min, 4°C). The pellet was resuspended in 6 mL buffer (10 mM Tris, 10 mM NaCl, 320 mM imidazole, adjusted to pH 7.0 with HCl) and 600 µg α-amylase, 250 U RNase A, 120 U DNase I, and 50 mM MgSO₄ were added. The sample was incubated at 37°C for 2 h while shaking, 12 U Proteinase K was added, and the incubation continued for 1 h. 4% SDS solution was added 1:1 and the mixture was boiled for 1 h. SDS was removed by repeated ultracentrifugation steps (20 times at 140,000 × g, 30 min, 40°C) and suspension in H₂O_bidistilled as well as dialysis against H₂O_bidistilled. The SDS content was controlled with the methylene blue assay described earlier [Hayashi, 1975]. The cell wall preparation was dried in a vacuum concentrator. Wall teichoic acids were removed by incubation with 1 M HCl for 4 h at 37°C.

For the preparation of E. coli K12 cell walls 2 L of culture (exponential growth phase, OD₆₀₀ = 1.0) were harvested and resuspended in 90 mL ice-cold H₂O_bidistilled (to prevent peptidoglycan autolysis activity). Boiling SDS solution (8%) was added dropwise until a final volume of 100 mL and boiled for 1 h further while stirring constantly. The SDS was removed as described for B. subtilis cell wall. After resuspending the pellet again in 20 mL H₂O_bidistilled, 300 µL ɑ-amylase stock solution (12 mg α-amylase + 0.5 mL H₂O_bidistilled + 0.7 mL 10 mM Tris-HCl pH 7) were added and incubated at 37°C under constant shaking for 2 h. Pronase solution (100 mg pronase + 10 mL 10 mM Tris-HCl buffer pH 7) was pre-incubated at 60°C for 2 h. 500 µL pronase solution was added to the sample and incubation was continued for 2 h further. The sample was boiled for 1 h while stirring followed by one ultracentrifugation step. The cell wall preparation was dried in a vacuum concentrator.

For the preparation of colloidal chitin chitosan (<3 kDa; obtained from Carbosynth) was acetylated with acetic anhydride. 20 mL 1 M sodium bicarbonate (freshly prepared) were added to 500 mg chitosan and mixed by swirling until a brown-colored foam developed. 0.5 mL of >99% acetic anhydride were added dropwise while mixing slowly, followed by incubation for 20 min at room temperature. The solution was precipitated by centrifugation (1,500 × g) and the pellet was resuspended again in 20 mL 1 M sodium bicarbonate with 0.5 mL acetic anhydride and further incubated for 20 min. The reaction was boiled for 10 min and washed with phosphate buffered saline until pH 7 was attained. The colloidal chitin was stored at 4°C.

For the digestion of E. coli and B. subtilis peptidoglycan, colloidal chitin, and chitosan oligomers, 0.2 mg substrate were used. The following buffers were used for the enzymes: 0.1 M potassium phosphate pH 8 for Atl and CwlC; 0.1 M sodium phosphate pH 7.2 for ChiA, chitinase from S. griseus, NagZ and ChbF; 0.1 M potassium phosphate pH 6 for mutanolysin; 0.05 M sodium acetate pH 5.7 for Csn. In each case 5 µg of enzyme were used for one preparation and all reactions took place at 37°C. Peptidoglycan preparations were incubated for 1 h, colloidal chitin, chitosan, and cytosolic fractions for 2 h. The digestion volume was 50 µL total, except for the pre-digestion of peptidoglycan by CwlC, which had a volume of 25 µL. After CwlC pre-treatment the pH was adjusted to the corresponding pH value of the subsequent enzyme with 0.1 M potassium dihydrogen phosphate and the volume was adjusted to 50 µL. After each digestion the enzymes were heat-inactivated at 95°C for 10 min.

A 5-mL culture was harvested by centrifugation at 16,000 × g for 10 min and was washed twice in 1 mL Tris-HCl (pH 7.6, 10 mM) buffer. The cells were resuspended in 200 µL buffer and incubated for 10 min at 95°C followed by a further centrifugation step. The supernatant was removed by pipetting and mixed with 800 µL acetone. After a further centrifugation step the supernatant was removed and dried under vacuum at 37°C. Samples were dissolved in 30 µL H₂O_bidistilled for LC-MS analysis [Gisin et al., 2013].

Samples were reduced with sodium borohydride as described earlier [Schaub and Dillard, 2017; Kluj et al., 2018]. To prepare the reducing solution, 500 µL of 0.5 M borate buffer (pH 9) was added to 5 mg sodium borohydride. Samples dissolved in 30 µL H₂O_bidistilled were added to 30 µL reducing solution and incubated for 20 min at room temperature. By the addition of 10 µL 8.5% phosphoric acid the pH was adjusted to 3–4 before the samples were dried in a vacuum concentrator at 37°C and dissolved in 30 µL H₂O_bidistilled for LC-MS analysis.

Sample analysis was conducted using an electrospray ionization-time of flight (ESI-TOF) mass spectrometer (MicrOTOF II; Bruker Daltonics), operated in positive ion-mode that was connected to an UltiMate 3000 high-performance liquid chromatography (HPLC) system (Dionex). For HPLC-MS analysis 5 µL of the sample supernatant were injected into a Gemini C18 column (150 × 4.6 mm, 5 µm, 110 Å, Phenomenex). A 45-min program at a flow rate of 0.2 mL/min was used to separate the compounds as previously described [Gisin et al., 2013]. An Orbitrap Elite mass spectrometer (Thermo Scientific) was used to record higher energy collisional dissociation (HCD) MS/MS spectra.

We acknowledge Marina Borisova, Tim Teufel, and Amanda Duckworth for providing plasmids and proteins as well as Johannes Madlung and Boris Macek for valuable expertise with MS/MS measurements.

Ethics approval was not required.

The authors have no conflicts of interest to declare.

C.M. acknowledges funding by the ministry of science, research and art of the state Baden-Württemberg (program glycobiology and glycobiotechnology) and by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), grants SFB766, Project-ID 398967434 – TRR 261 and Project-ID 174858087 – GRK1708. This work was further supported by the DFG-funded Cluster of Excellence EXC 2124 Controlling Microbes to Fight Infections.

A.W. conducted the experiments. S.F. provided the plasmid pQE32-ChbBCA. C.M. formulated the original problem and provided guidance throughout the study. C.M. and A.W. designed the experiments and developed the methodology. C.M., S.F., and A.W. wrote the manuscript.