One pot synthesis of GDP‐mannose by a multi‐enzyme cascade for enzymatic assembly of lipid‐linked oligosaccharides

Abstract Glycosylation of proteins is a key function of the biosynthetic‐secretory pathway in the endoplasmic reticulum (ER) and Golgi apparatus. Glycosylated proteins play a crucial role in cell trafficking and signaling, cell‐cell adhesion, blood‐group antigenicity, and immune response. In addition, the glycosylation of proteins is an important parameter in the optimization of many glycoprotein‐based drugs such as monoclonal antibodies. In vitro glycoengineering of proteins requires glycosyltransferases as well as expensive nucleotide sugars. Here, we present a designed pathway consisting of five enzymes, glucokinase (Glk), phosphomannomutase (ManB), mannose‐1‐phosphate‐guanyltransferase (ManC), inorganic pyrophosphatase (PmPpA), and 1‐domain polyphosphate kinase 2 (1D‐Ppk2) expressed in E. coli for the cell‐free production and regeneration of GDP‐mannose from mannose and polyphosphate with catalytic amounts of GDP and ADP. It was shown that GDP‐mannose is produced at various conditions, that is pH 7–8, temperature 25–35°C and co‐factor concentrations of 5–20 mM MgCl2. The maximum reaction rate of GDP‐mannose achieved was 2.7 μM/min at 30°C and 10 mM MgCl2 producing 566 nmol GDP‐mannose after a reaction time of 240 min. With respect to the initial GDP concentration (0.8 mM) this is equivalent to a yield of 71%. Additionally, the cascade was coupled to purified, transmembrane‐deleted Alg1 (ALG1ΔTM), the first mannosyltransferase in the ER‐associated lipid‐linked oligosaccharide (LLO) assembly. Thereby, in a one‐pot reaction, phytanyl‐PP‐(GlcNAc)2‐Man1 was produced with efficient nucleotide sugar regeneration for the first time. Phytanyl‐PP‐(GlcNAc)2‐Man1 can serve as a substrate for the synthesis of LLO for the cell‐free in vitro glycosylation of proteins. A high‐performance anion exchange chromatography method with UV and conductivity detection (HPAEC‐UV/CD) assay was optimized and validated to determine the enzyme kinetics. The established kinetic model enabled the optimization of the GDP‐mannose regenerating cascade and can further be used to study coupling of the GDP‐mannose cascade with glycosyltransferases. Overall, the study envisages a first step towards the development of a platform for the cell‐free production of LLOs as precursors for in vitro glycoengineering of proteins.

transmembrane-deleted Alg1 (ALG1ΔTM), the first mannosyltransferase in the ERassociated lipid-linked oligosaccharide (LLO) assembly. Thereby, in a one-pot reaction, phytanyl-PP-(GlcNAc) 2 -Man 1 was produced with efficient nucleotide sugar regeneration for the first time. Phytanyl-PP-(GlcNAc) 2 -Man 1 can serve as a substrate for the synthesis of LLO for the cell-free in vitro glycosylation of proteins. A high-performance anion exchange chromatography method with UV and conductivity detection (HPAEC-UV/CD) assay was optimized and validated to determine the enzyme kinetics. The established kinetic model enabled the optimization of the GDP-mannose regenerating cascade and can further be used to study coupling of the GDP-mannose cascade with glycosyltransferases. Overall, the study envisages a first step towards the development of a platform for the cell-free production of LLOs as precursors for in vitro glycoengineering of proteins.

K E Y W O R D S
cell-free synthesis, enzymatic catalysis, kinetic modeling, in vitro N-glycoengineering, nucleotide sugar regeneration 1 | INTRODUCTION N-linked protein glycosylation is a co-translational modification in eukaryotes that affects protein folding directly or indirectly (Culyba et al., 2011;Hanson et al., 2009;Helenius & Aebi, 2004;Shental-Bechor & Levy, 2008). N-linked glycans play a role in protein stability, solubility and cell trafficking as well as cell signaling (Taylor & Drickamer, 2011). Therefore, the glycosylation of proteins is also an important parameter in the optimization of animal cell culture-derived drugs including monoclonal antibodies, growth factors, and hormones (Dekkers et al., 2016;Hossler, Khattak, & Li, 2009;Lalonde & Durocher, 2017;Sha, Agarabi, Brorson, Lee, & Yoon, 2016;Spearman, Rodriguez, Huzel, Sunley, & Butler, 2007). In addition, over the past years efforts have been made to modify the N-glycosylation machinery in yeast and E. coli for the production of therapeutic proteins at lowcosts with tailored glycosylation in vivo (Srichaisupakit, Ohashi, Misaki, & Fujiyama, 2015;Valderrama-Rincon et al., 2012;Wildt & Gerngross, 2005). An alternative approach is the in vitro glycoengineering of proteins by modifying the glycostructure via enzymatic reactions with purified glycosyltransferases and nucleotide sugars (Thomann et al., 2015). Case studies have shown very promising results in terms of increasing the level of galactosylation and sialylation on IgG (Chung et al., 2006;Raju, Briggs, Chamow, Winkler, & Jones, 2001;Thomann et al., 2015). To satisfy the high demand of nucleotide sugars UDPgalactose and CMP-sialic acid for in vitro glycoengineering, Raju et al. (2001) have designed an in vitro enzymatic nucleotide sugar regeneration cascade for these two co-substrates and demonstrated galactosylation and sialylation of tumor necrosis factor receptor IgGs.
In order to in vitro N-glycosylate proteins by enzymatic reactions, lipid-linked oligosaccharides (LLO) are needed as substrates (Ramírez, Boilevin, Biswas, et al., 2017;Ramírez, Boilevin, Lin, et al., 2017). The ER associated biosynthesis of the LLO is a highly conserved process in eukaryotic cells. The core glycan (GlcNAc) 2 -Man 9 -Glc 3 is assembled on a membrane-localized dolichyl-pyrophosphate by a cascade of 12 glycosyltransferases and is then transferred to a nascent polypeptide chain by an oligosaccharyltransferase (OST). GDP-mannose (GDPman), UDP-GlcNAc and UDP-Glc serve, directly and indirectly, as mannose, N-acetylglucosamine and glucose donors for the attachment of sugars to the LLO (Aebi, 2013;Helenius & Aebi, 2004). So far, to the best of our knowledge, there is no cell-free platform or process for the preparative synthesis of ER-LLOs with efficient nucleotide sugar regeneration. Challenges are, in particular, the expression and the purification of ER membrane-associated glycosyltransferases, and the provision of key enzymatic reactions with expensive sugar nucleotides, namely GDP-man, UDP-GlcNAc, and UDP-Glc. GDP-man is enzymatically produced from mannose-1-phosphate and GTP by mannose-1phosphate guanyltransferase (ManC). In nature there are two pathways for the production of mannose-1-phosphate starting either from glucose or mannose in the salvage pathway (Kuettel et al., 2012;Pfeiffer, Bulfon, Weber, & Nidetzky, 2016). Several studies have been published on exploiting and modifying these pathways for the cell-free production and isolation of GDP-man (Jia et al., 2011;Pfeiffer et al., 2016;Wang, Shen, Wang, Ichikawa, & Wong, 1993). Honghong et al.
have designed an enzyme cascade based on the salvage pathway.
Using a raw extract of E. coli containing recombinant glucokinase (Glk), phosphomannomutase (ManB), and mannose-1-phosphate-guanyltransferase (ManC) the GDP-man was produced from mannose, ATP, and GTP (Jia et al., 2011). To avoid product purification after one-pot multi-enzyme cascade synthesis of nucleotide sugars, the in vitro coupling of an enzyme cascade regenerating nucleotide sugars to the glycosyltransferase reactions is advantageous. For example, Chung et al. (2006)  designed an in vitro enzyme cascade with pyruvate kinase, inorganic pyrophosphatase, and mannose-1-phosphate-guanyltransferase to regenerate GDP-man from mannose-1-phosphate. The cascade was coupled to α-1,2-mannosyltransferase to attach mannose on Omannosylglycopeptides.
Here we present a systematic, model-supported development of a cell-free synthetic enzyme cascade consisting of five enzymes to synthesize and continuously regenerate GDP-man from mannose and polyphosphate with catalytic amounts of GDP and ADP. The cascade was optimized for effective GDP-man production and tested at pH values 7-8, temperatures 25-35°C, and co-factor concentrations 0-20 mM MgCl 2 to characterize the synthesis reactions at various conditions. In addition, the cascade was in vitro coupled to a transmembrane-deleted β-1,4mannosyltransferase (Alg1ΔTM) in a one-pot reaction to produce phytanyl-PP-(GlcNAc) 2 -Man 1 . To identify inhibition and bottlenecks in the multi-enzyme cascade reaction a kinetic model was established using the MATLAB® systems biology toolbox.

| MATERIALS AND METHODS
For a comprehensive list of chemicals used including vendors and purity grades, see the supplementary information.

| Cultivation of enzyme variants in
Transformants were grown in 1 L shaking flasks with baffles in a volume of 500 ml of LB medium supplemented with 50 µg/ml Kanamycin. The cultures were grown at 37°C (His 6 -Glk, ManB-His 6 -ManC, His 6 -Alg1ΔTM, PmPpA-His 6 ) and 24°C (His 6 -1D-Ppk2), respectively, and shaken at 80 rpm. The induction of the LacZ promotor was forced by addition of IPTG with a final concentration of 1 mM to the culture at an OD 600 of 0.5-0.6. Expression time was terminated after 4 hr. Biomass was separated from the medium by centrifugation at 6,000 × g for 10 min. Successful expression of the respective protein was analyzed by SDS-PAGE following standard operating procedures (Laemmli, 1970). The wet biomass was stored at −20°C.

| Purification of enzymes by immobilized metal affinity chromatography
For purification, typically 30 ml of equilibration buffer were added to 3 g of frozen biomass. The equilibration buffer consisted of 50 mM Tris/HCl (pH 7.5), 500 mM NaCl, 10 mM imidazole, and 10 mM MgCl 2 .
In case of His 6 -1D-Ppk2 purification, 5 % glycerol (v/v) was added to stabilize the enzyme (Bradbury & Jakoby, 1972) according to (Zhang et al., 2001). For purification of His 6 -Alg1ΔTM, the concentration of imidazole in the equilibration buffer was 30 mM; in addition the buffer contained 0.25% (w/v) of Triton X-100. Following thawing at 4°C under stirring, cells were disrupted by four passages through a high pressure homogenizer (Emulsiflex C5, Avestin Inc., Ottawa, Canada) at 1,000 bar with intermediate cooling on ice. After centrifugation (45 min, 20,000 × g), the supernatant was applied to an equilibrated Immobilized Metal Affinity Chromatography (IMAC) column (10 ml CV) containing Ni 2+ Sepharose™ High Performance chromatography material from Amersham Biosciences (Uppsala, Sweden). Unbound proteins were washed out using equilibration buffer. Immobilized protein was eluted in 1 ml fractions using elution buffer containing 50 mM Tris/HCl (pH 7.5), 500 mM NaCl, 500 mM imidazole, and 10 mM MgCl 2 with changes in equilibration buffer composition also applied to the elution buffer. To remove excess imidazole, the eluted pool (typically 10 ml) was dialyzed two times against 3 L of reaction buffer containing 20 mM Tris/HCl (pH 7.5), 50 mM NaCl, and 10 mM MgCl 2 (again with respect to the above mentioned changes). Finally, the enzyme solutions were concentrated by Centrifugal Filter Units Amicon® Ultra-15 with a 50 kDa cut-off from Merck Millipore (Darmstadt, Germany). No enzyme loss was observed during the ultrafiltration. The enzymes were stored in 50% glycerol at −20°C. The protein concentration was determined by Bradford assay using BSA as standard (Bradford, 1976

| Chromatography
Reaction substrates and products were separated and quantified by high performance anion-exchange chromatography (HPAEC). A BioLCType DX320 system from Dionex (Sunnyvale) with UV (wave length: 260 nm) and conductivity detection was used. Chromatographic separation was performed at a system flow of 0.35 ml/ min by two analytical columns, AS11 (250 × 2 mm) operated in-series.
The eluent gradient (5-100 mM KOH) for chromatographic separation was based on previous studies with some modifications (Ritter, Genzel, & Reichl, 2006). The gradient is shown in Figures

| Reaction conditions
Reaction volumes for kinetic measurements were 1 ml. All reactions were carried out with a co-factor concentration of 10 mM MgCl 2 at 30°C in 50 mM Tris/HCl buffer of pH 7.5 and incubated at 30 rpm in a thermomixer from Eppendorf AG (Hamburg, Germany) unless stated otherwise. Sample aliquots of 100 μl were quenched in 400-900 μl of MilliQ water, preheated in a closed Eppendorf tube to 90°C, followed by another 10 min of heating at 90°C. To ensure enzyme inactivation, the quenching protocol was tested on all enzymes.
All reactions were tested for reversibility, inhibition, and long-term stability when stored as well as enzyme inactivation during assaying by Selwyn's Test (Selwyn, 1965). Overall six data sets-Glk (16 reactions

| Data-fitting and simulations
The systems biology toolbox SBTOOLBOX2 for MATLAB® (Version R2013b) from The MathWorks (Natick) was used for data fitting and simulations (Schmidt, 2007;Schmidt & Jirstrand, 2006). Typically, the adjusted Nelder-Mead Simplex and the Particle swarm pattern search algorithm were employed interchangeably to find function minima (Olsson & Nelson, 1975;Press, Teukolsky, Vetterling, & Flannery, 1996;Vaz & Vicente, 2007). For data fitting one fit per data set was performed.
His 6 -Glk was produced as mentioned in the section 2 (Lunin et al., 2004;Meyer et al., 1997). From 2 g bio wet mass, 164 mg highly pure enzymes were isolated. The yield was 27-fold higher compared to the literature (Lunin et al., 2004). Aqueous enzyme stock solution contained 8.2 mg/ml His 6 -Glk.
PmPpA-His 6 was produced according to the literature (Lau et al., 2010). The enzyme precipitated during the dialysis and was therefore used in non-dialyzed form. From 2 g bio wet mass, 53 mg highly pure enzyme was isolated. PmPpA-His 6 concentration in the stock solution was 2.03 mg/ml.
ManC and ManB-His 6 were co-expressed according to the literature using one plasmid, separated by a 50 bp linker region, each nucleotide sequence having its own ribosomal binding site (Koizumi et al., 2000;Lee, Han, Park, & Seo, 2009). So far, both enzymes were applied as crude enzymes or as raw extracts (Koizumi et al., 2000;Lee et al., 2009). In order to prevent side-reactions from other host proteins, a purification protocol was developed based on IMAC. Untagged ManC was co-eluted with the adsorbed ManB-His 6 .
This result can be either explained by an intermolecular interaction of both enzymes or by the 14 histidines present in the amino acid sequence of ManC (see supplement). Eight of these histidine residues are located in the C-terminal part in close vicinity to each other. The enzymes were not separated before use and were free of contaminating host cell proteins from E. coli. From 3 g bio wet mass, 74 mg highly gradient. Suppression of hydroxide ions was conducted by the electronically regenerated suppressor ERS500 from Dionex. The large peak at 5 min was caused by Clions. Man1P, man6P, phosphate (P), and diphosphate (DP) concentrations were measured by conductivity detection. The assay was validated pure enzymes were isolated. Enzyme stock solution contained 1.71 mg/ml ManB-His6/ManC. His 6 -1D-Ppk2 was produced according to the literature (Nocek et al., 2008;Zhang et al., 2001). During expression of His 6 -1D-Ppk2 at 37°C the formation of inclusion bodies was detected. Lowering the temperature to 24°C after induction increased the yield of the soluble enzyme substantially. After purification, His 6 -1D-Ppk2 was soluble in concentrations of about 0.5 mg/ml in presence of 50% glycerol. Above this concentration the enzyme precipitated. From 3 g bio wet mass, 25 mg highly pure enzyme was isolated. His 6 -1D-Ppk2 stock solution contained 0.15 mg/ml of the enzyme.
His 6 -Alg1ΔTM was produced according to the literature using 0.25 % Triton X-100 as a stabilizer (Revers, Bill, Wilson, Watt, & Flitsch, 1999). Without the addition of Triton X-100 to the binding buffer His 6 -Alg1ΔTM was not bound to the IMAC stationary phase.
The addition of 5 mM DTT to the dialysis buffer solved the problem of inactivation/precipitation. From 3 g bio wet mass, 5 mg highly pure enzyme was isolated and the stock solution contained 0.55 mg/ml His 6 -Alg1ΔTM after purification and addition of glycerol.

| Single-enzyme kinetics
The kinetics of enzyme-catalyzed in vitro reactions are influenced by multiple parameters such as enzyme concentration, purity, and buffer components. Published reaction conditions such as temperature, pH value, co-factors, co-factor concentration, and type of reaction buffers for different enzymes differ significantly.
Therefore, in the present study, the enzyme kinetics of each reaction was determined experimentally for the reaction conditions stated in the section 2.
The kinetic parameters for each enzyme were determined by fitting the generated experimental data sets to the equations detailed in Table 1 All experimental kinetic data is available on request. In the following, the results for the single enzymes are discussed.

| His 6 -Glk
Glk catalyzes the reaction of glucose and ATP to glucose-6-phosphate and ADP. Previous studies of His 6 -Glk from E. coli primarily focused on the conversion of glucose. Instead of glucose, the enzyme also accepts mannose as a substrate. However, experimental data on the enzyme kinetics are very scarce. The only report states that the maximum rate of mannose conversion is "reduced" compared to glucose. For mannose conversion kinetic data are missing (Arora & Pedersen, 1995;Lunin et al., 2004;Meyer et al., 1997;Miller & Raines, 2004).
In our experiments, it was found that the His 6 -Glk-catalyzed conversion of ATP and mannose to man6P and ADP is irreversible.
Reactions with His 6 -Glk concentrations of 1 mg/ml (25.6 µM) were not inhibited by mannose (up to 10 mM) ( Figure 4A). There was no loss of enzyme activity over 11 months when stored at −20°C. No man6P was produced when His 6 -Glk (1 mg/ml) was incubated with 2 mM of GTP.
Reactions with His 6 -Glk concentrations of 0.5-2 mg/ml (ATP: 2 mM, mannose: 4 mM) over a time course of up to 180 min revealed no enzyme inactivation according to Selwyn's plot (data not shown). The experimental data were fitted to mass action kinetics and a good fit was achieved.

| PmPpA-His 6
PmPpA from P. multocida catalyzes the hydrolysis of pyrophosphate and has not been characterized and kinetically investigated yet. Own pre-tests at 37°C measured by UV spectroscopy with pyrophosphate concentrations between 0.015-0.225 mM confirmed the activity of the PmPpA-His 6 (see supplementary information). At the employed concentrations of PmPpA-His 6 (0.445 μg/ml; equivalent to 21.2 nM) the reaction rate to mass ratio is several orders of magnitude higher compared to the other enzymes of the cascade. In the model, it was  therefore assumed that pyrophosphate is consumed instantaneously.
For more details on the kinetics see the supplementary information.

| His 6 -1D-Ppk2
His 6 -1D-Ppk2 from P. aeruginosa catalyzes the synthesis of both ATP and GTP from ADP and GDP, respectively, as one substrate with inorganic polyphosphate as the second substrate. The enzyme is dependent on Mg 2+ ions and was characterized and kinetically investigated first by . Later it was discriminated from the two-domain Ppk2 (2D-Ppk2) which catalyzes the phosphorylation of AMP and GMP (Nocek et al., 2008).
In our experiments it was found that the phosphorylation of ADP and GDP, respectively, by His 6 -1D-Ppk2 are equilibrium reactions (see Figure 5). This is in disagreement with Nocek et al. (2008)  concentrations showed that their effect on the reaction rate is complex FIGURE 4 (a) Irreversible man6P synthesis by His 6 -Glk (1 mg/ml) from ATP (1.7 mM) using different mannose concentrations. The lines represent the model (see Table 1). (b) Reverse reaction of the ManB-His 6 /ManC (0.171 mg/ml) complex: GDP-mannose and pyrophosphate are converted to GTP and man6P catalyzed by ManB and ManC. Subsequently, the synthesis of the GDP-man is largely dependent on the conversion of pyrophosphate to phosphate. The fits depicted are fits to the entire data set of His 6 -Glk and ManB-His 6 /ManC, respectively, catalyzed reactions  Table 1). In additional experiments, it was observed that at lower concentrations of PolyP 14 (1-3 mM) the measured reaction rates were much lower than predicted by the kinetic equations. With 0.1 mM PolyP 14 , the substrate affinity (K eq ) was higher for GDP than for ADP. This is in accordance with previous studies Nocek et al., 2008;Zhang et al., 2002). As 4 mM PolyP 14 was found to be the optimal concentration, no attempt was made to establish a more complex rate equation to describe the inhibition of PolyP 14 at initial concentrations above or below 4 mM.
No loss of enzyme activity was observed over 13 months when storing His 6 -1D-Ppk2 stock solutions at −20°C (data not shown).

| ManB-His 6 /ManC
ManB catalyzes the conversion of man6P into man1P. ManC catalyzes the reaction of man1P and GTP into GDP-man and pyrophosphate.
manB and manC are neighboring genes in the polycistronic gene cluster for colanic acid biosynthesis in Enterobacteriaceae and both enzymes are usually used as a complex and produced in a bicistronic expression system (Koizumi et al., 2000;Lee et al., 2009)

. ManB and ManC from
Salmonella enterica have been expressed in E. coli. Using purified enzymes, kinetic models for the reactions have been proposed and  (Elling, Ritter, & Verseck, 1996;Fey, Elling, & Kragl, 1997). Because of the competitive inhibition of recombinant GDP-man pyrophosphorylase of S. enterica for GTP by GDP-man and the uncompetitive inhibition for man1P by GDP-man (Yang et al., 2005), we investigated the production of recombinant ManB and  (Naught & Tipton, 2001;Orvisky et al., 2003).
The equilibrium conversion of man1P to man6P catalyzed by the ManB-His 6 /ManC complex was studied independently from the GDP-man synthesis. It was found that ManB-His 6 does not require exogeneous glucose-1,6-bisphosphate but the reaction was faster with glucose-1,6-biphosphate. Moreover, glucose-6-phosphate peaks emerged when glucose-1,6-biphosphate was used. It is likely that there are multiple reaction equilibria between mannosephosphates and glucose-phosphates as proposed in the literature for ManB from Galdieria sulphuria (Oesterhelt, Schnarrenberger, & Gross, 1997). In the presence of glucose-1,6-bisphosphate Reverse reactions were carried out for 1.7 mM GDP-man and pyrophosphate, respectively, and for two different enzyme concentration (0.171 and 0.342 mg/ml). It was found that after rapid consumption of about 90% of the substrates, GDP-man was again produced with a constant rate of 2.1 ± 0.4 μmol/min after 40 min due to cleavage of pyrophosphates (see Figure 4b).
The ManB-His 6 /ManC enzyme complex did not lose activity within 7 months when stored at −20°C.

| Simulation and experiment
Multi-enzyme experiments were performed to demonstrate GDPman production by the cascade (see Figures 6 and 7). The model composed of the kinetics measured for the single-enzyme reactions is compared against the experimental data (see Figure 6). While the fit between simulated and experimental data is reasonable for ADP and ATP, the model underestimates GDP-man concentrations and overestimates GDP and GTP concentrations. One potential expla-  3.3.2 | Effect of pH value, temperature, and cofactors on the one-pot multi-enzyme cascade reaction To check the effect of pH value, temperature, and co-factors on the GDP-man synthesis, a number of one-pot multi-enzyme cascade reactions were carried out (see Figure 7). It was found that there is no significant difference in metabolite concentration profiles and, in particular, GDP-man concentration over time at pH 7.0 (HEPES/ NaOH), pH 7.5, and pH 8.0 (both Tris/HCl) (see Figure 7b). Increasing the temperature to 35°C did not change the rate of GDP-man production within 240 min while at 25°C only about 200 μM (35% less) of GDP-mannose was produced after 240 min (see Figure 7c).
Moreover, it was found that ADP and GDP were not converted to ATP and GTP when the cascade reaction was initiated without addition of the co-factor MgCl 2 which indicates that His 6 -1D-Ppk2 requires MgCl 2 for activity (data not shown). At 5 mM MgCl 2 around 45% less GDP-man was formed after 240 min compared to the reaction with 10 mM MgCl 2 (see Figure 7d). Overall, the maximum amount of GDPman produced after 240 min was 566 μM (30°C, 10 mM MgCl 2 ) with a constant reaction rate of 2.7 μM/min after a lag phase of 60 min. The initial concentrations were 0.8 mM GDP and 6 mM mannose and, thus, the yield with respect to GDP was 71%. No GDP-man was detected by HPAEC-UV/CD which means that the GDP-man to Man1 yield should be 100%. However, only about 80 μM phytanyl-PP-(GlcNAc) 2 -Man 1 was produced after 240 min which is equivalent to a yield of 11% with respect to phytanyl-PP-(GlcNAc) 2 . Thus, phytanyl-PP-(GlcNAc) 2 -Man 1 production in the multi-enzyme cascade was much slower than predicted by the kinetic simulation (see Figure 9). Comparison between the simulation and the experimental data show that the steady-state equilibrium of the His 6 -1D-Ppk2-catalyzed conversions of ADP to ATP and GDP to GTP, respectively, is lower with respect to the nucleoside triphosphates than predicted. Parameter estimations reveal that the kinetic model fits the experimental data when constants k 2 and k 3 are adjusted to 0.097 and 0.041 L/(min g) values.

|
This finding indicates that nucleoside diphosphate conversion is inhibited by phytanyl-PP-(GlcNAc) 2 , Alg1ΔTM or Triton X-100, which is present in the Alg1ΔTM stock solution. Triton X-100 is necessary to solubilize membrane-bound enzymes and cannot be omitted.

| CONCLUSIONS
A cell-free cascade of five enzymes expressed in and purified from E. coli BL21-Gold (DE3) was designed to produce and regenerate GDP-mannose from mannose and polyphosphate with catalytic amounts of GDP and ADP. The maximum reaction rate of GDP-mannose was 2.7 μM/min at 30°C and 10 mM MgCl 2 producing 566 nmol GDP-mannose from 800 nmol GDP and 6 μmol mannose after 240 min. Furthermore, it was demonstrated that the cascade can be in vitro coupled with a purified glycosyltransferase to donate mannose for the LLO assembly. A kinetic model based on single-enzyme reactions was established to investigate inhibition and enhancement in the multi-enzyme cascade. The model can be used to study and optimize coupling of the GDP-mannose cascade with one or more glycosyltransferases. Overall, the study envisages a first step toward the development of a platform for the cell-free production of LLOs as precursors for in vitro glycoengineering of proteins.