**4. Discussion**

Glycosyltransferases are assumed to be very specific enzymes and misincorporation of NDP-sugars is not common in normal circumstances. Nevertheless, the incorporation of non-natural analogues of N-acetamido sugars carrying various N-groups is a well-known approach for carbohydrate labelling or drug design [24,25]. To some extent, KfoC has been found to be able to take up UDP-GlcNAc as a donor, but no further elongation could be observed after that [7]. In this study, we showed that in vivo, elongation after the incorporation of GlcNAc was still possible, leading to a co-polymer composed of hyaluronic acid and chondroitin. To our knowledge, the ability of KfoC to use UDP-Gal as a donor has not been investigated, nor the presence of galactose among chondroitin backbone. The crystal structure of KfoC has been elucidated, but the binding of UDP-Gal in the binding pocket has not been investigated [26].

β α Here we showed that in vivo, in the absence of UDP-GalNAc, galactose was incorporated by KfoC leading to the synthesis of a chondroitin-like polymer. A similar phenomenon has already been observed with the heparosan synthase PmHS2. In vitro experiments show that PmHS2 incorporates glucose in absence of UDP-GlcNAc, resulting in a new heparosan-like polymer [-4-GlcAβ1-4-Glcα1-]n named hepbiuronan [26]. Following the same rationale, we named the new chondroitin-like polymer: chondbiuronan. Interestingly, the PmHS enzyme used for in vitro hepbiuronic synthesis showed a 10-fold decrease in maximum velocity. Here, comparison of the chondbiuronan and chondroitin synthesis suggests a similar phenomenon occurring with KfoC.

Another work dealing with human α-1,3-GalNAc-T (GTA) involved in blood group reported that the enzyme could bind UDP-Gal but could not efficiently stabilize the complex in a catalytically active conformation [27]. Such a control based on the enzyme mobility does not seem to be very efficient in KfoC. This lack of selectivity may be due to the fact that in the context of recombinant strains, engineered metabolic pathways are different from what they are in wild type strains from which recombinant enzymes are issued, and this can result in enhancing aberrant enzymatic behaviour. In our work, only *kfoA* and *kfoC* genes belonging to the chondroitin operon of K4 have been expressed. It cannot be excluded that in normal circumstances, protein interactions occur, which play a role in enzymatic stability and specificity. Moreover, recombinant *kfoC* gene was cloned in high copy plasmid which supposedly differs from *E. coli* K4 wild type expression. Eventually, we pointed out the fact that galactose could be incorporated in the chondroitin backbone when overexpressing KfoC. The NMR study we have conducted should be of great help to investigate the presence of galactose in bacterial chondroitin produced in other biological systems.

We showed that chondbiuronan could be digested by chondroitin AC lyase and by hyaluronidase. Previously, it has been observed that hepbiuronan could be digested by the bacterial heparin lyase III [26]. These results suggest that the specificity for acting on the hexose-substituted GAGs is relaxed in both classes of bacterial enzymes which are widely used in the industry for quality control of GAGs.
