Toward Automated Enzymatic Glycan Synthesis in a Compartmented Flow Microreactor System

Article abstract of DOI:10.1002/adsc.201900709

Immobilized microfluidic enzyme reactors (IMER) are of particular interest for automation of enzyme cascade reactions. Within an IMER, substrates are converted by paralleled immobilized enzyme modules and intermediate products are transported for further conversion by subsequent enzyme modules. By optimizing substrate conversion in the spatially separated enzyme modules purification of intermediate products is not necessary, thus shortening process time and increasing space-time yields. The IMER enables the development of efficient enzyme cascades by combining compatible enzymatic reactions in different arrangements under optimal conditions and the possibility of a cost-benefit analysis prior to scale-up. These features are of special interest for automation of enzymatic glycan synthesis. We here demonstrate a compartmented flow microreactor system using six magnetic enzyme beads (MEBs) for the synthesis of the non-sulfated human natural killer cell-1 (HNK-1) glycan epitope. MEBs are assembled to build compartmented enzyme modules, consisting of enzyme cascades for the synthesis of uridine 5′- diphospho-α- d-galactose (UDP-Gal) and uridine 5′-diphospho-α-d-glucuronic acid (UDP-GlcA), the donor substrates for the Leloir glycosyltransferases β4-galactosyltransferase and β3-glucuronosyltransferase, respectively. Glycan synthesis was realized in an automated microreactor system by a cascade of individual enzyme module compartments each performing under optimal conditions. The products were analyzed inline by an MS-system connected to the microreactor. The high synthesis yield of 96% for the non-sulfated HNK-1 glycan epitope indicates the excellent performance of the automated enzyme module cascade. Furthermore, combinations of other MEBs for nucleotide sugars synthesis with MEBs of glycosyltransferases have the potential for a fully automated and programmed glycan synthesis in a compartmented flow microreactor system. (Figure presented.).

Full text of DOI:10.1002/adsc.201900709

Title: Toward Automated Enzymatic Glycan Synthesis in a
Compartmented Flow Microreactor System
Authors: Raphael Heinzler, Thomas Fischöder, Lothar Elling, and
Matthias Franzreb
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Adv. Synth. Catal. 10.1002/adsc.201900709
Link to VoR: http://dx.doi.org/10.1002/adsc.201900709

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
VERY IMPORTANT PUBLICATION
FULL PAPER
DOI: 10.1002/adsc.201((will be filled in by the editorial staff))
Toward Automated Enzymatic Glycan Synthesis in a
Compartmented Flow Microreactor System
Raphael Heinzler a,c, Thomas Fischöder b,c, Lothar Elling b*, Matthias Franzreb a*
a
Institute of Functional Interfaces, Karlsruhe Institute of Technology, Karlsruhe, Germany
Fax: (++49)-721-608-2-3478, phone :(++49)-721-608-2-3595 e-mail: Matthias.franzreb@kit.edu
b
Laboratory for Biomaterials, Institute for Biotechnology and Helmholtz Institute for Biomedical Engineering, RWTH
Aachen University, Aachen, Germany
Fax: (++49)-241-802-2387, phone: (++49)-241-802-8350; e-mail: l.elling@biotec.rwth-aachen.de
c
These authors contributed equally to this work
Received: ((will be filled in by the editorial staff))
Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/adsc.201######.((Please
delete if not appropriate))
Abstract. Immobilized microfluidic enzyme reactors (IMER) uridine 5’-diphospho-α-D-glucuronic acid (UDP-GlcA),
are of particular interest for automation of enzyme cascade the donor substrates for the Leloir glycosyltransferases β4-
reactions. Within an IMER, substrates are converted by galactosyltransferase and β3-glucuronosyltransferase,
paralleled immobilized enzyme modules and intermediate respectively. Glycan synthesis was realized in an
products are transported for further conversion by subsequent automated microreactor system by a cascade of individual
enzyme modules. By optimizing substrate conversion in the enzyme module compartments each performing under
spatially separated enzyme modules purification of optimal conditions. The products were analyzed inline by
intermediate products is not necessary, thus shortening an MS-system connected to the microreactor. The high
process time and increasing space-time yields. The IMER synthesis yield of 96% for the non-sulfated HNK-1 glycan
enables the development of efficient enzyme cascades by epitope indicates the excellent performance of the
combining compatible enzymatic reactions in different automated enzyme module cascade. Furthermore,
arrangements under optimal conditions and the possibility of combinations of other MEBs for nucleotide sugars
a cost-benefit analysis prior to scale-up. These features are of synthesis with MEBs of glycosyltransferases have the
special interest for automation of enzymatic glycan synthesis. potential for a fully automated and programmed glycan
We here demonstrate a compartmented flow microreactor synthesis in a compartmented flow microreactor system.
system using six magnetic enzyme beads (MEBs) for the
synthesis of the non-sulfated human natural killer cell-1
(HNK-1) glycan epitope. MEBs are assembled to build
compartmented enzyme modules, consisting of enzyme
cascades for the synthesis of uridine 5’- diphospho-α-D-
galactose (UDP-Gal) and
Keywords: Glycoconjugate; Microreactors; Biocatalysis;
Immobilization; Magnetic beads
protected and deprotected.[2] This is even more severe
in case of an automated chemical glycan synthesis, as
each step must be optimized for high product yields.
Introduction
Carbohydrate molecules linked to other compounds Moreover, lengthy and time-consuming procedures
such as proteins and lipids are called glycoconjugates are also required for the removal of by-products, such
and serve various functions, including cell-to-cell and as
stereoisomers,
regioisomers,
unreacted
cell-to-matrix communication as well as cross-linking intermediates. Elegant strategies have been developed
between proteins.[1] For chemical stereoselective and to address these challenges for automated chemical
regioselective glycosylations, multi-step syntheses are glycan synthesis leading up to hexasaccharides and
necessary because the numerous hydroxyl groups even a 50mer of a homopolysaccharide.[3] However,
present on individual sugars must be selectively yields are still low or moderate. In contrast, enzyme-
1
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
assisted glycan synthesis is an attractive alternative to glycosyltransferases is now available for enzymatic
chemical synthesis as it has the advantage of synthesis of complex glycans.[4] Automated
achieving regio- and stereoselective glycosylations in enzymatic synthesis is, therefore, an emerging and
a single step. Although strict enzyme substrate still developing technology. With glycosyltransferases
specificity appears as a disadvantage, a whole enzyme in solution, automated glycan using solid-phase,
toolbox for glycosylation reactions and synthesis of polymer-bound[5] or tagged
nucleotide
sugars
as
precursors
of
Scheme 1. Enzyme cascade using magnetic enzyme beads (MEB). A: Enzyme modules (dotted line) with magnetic
enzyme beads (MEB) for the synthesis of the donor substrates UDP-Gal (3) and UDP-GlcA (5) were combined for
glycosylation reactions by the glycosyltransferases β4GalT to yield LacNAc (7) and β3GlcAT for the synthesis of HNK-1
epitope (8). B: Flowchart of the reaction cascade in the compartmented flow microreactor system (CFMS). Box: reaction
2
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
compartment with magnetic enzyme beads (MEB); circle: product compartment for transport to the next reaction
compartment. Each compartment has a volume of 150 µL; EM reactions were carried out at 37 °C.
substrates for capture and release of products[6] have
The herein used enzyme modules (EM) have been
been developed, each strategy with its own pros and optimized and combined in our previous studies
cons as discussed in recent reviews.[7] A common employing soluble enzymes (Scheme 1A).
disadvantage is that glycosyltransferase reactions are Galactokinase (GalK) was combined with UDP-sugar
not optimized for high product yields in short reaction pyrophosphorylase (USP) in an enzyme cascade
time, and the precious biocatalysts are not recovered reaction for the synthesis of UDP-galactose (UDP-Gal,
and reused. Relevant numbers for space-time-yield 3) from D-(+)-galactose (Gal,1) via α-D-galactose-1-
(STY, g product L-1 day-1) and product-specific total phosphate (Gal-1-P, 2) in the EM-UDP-Gal[12].In the
turnover number TTN (g product/g catalyst) are EM-UDP-GlcA,
therefore low for automated enzymatic synthesis. (UGDH) generates Uridine 5 ′ ‐ diphosphate ‐
Only few immobilized Leloir-type glucuronic acid (UDP-GlcA, 5) from UDP‐glucose
UDP-glucose-dehydrogenase
a
glycosyltransferases have been applied in glycan (UDP-Glc, 4), while NADH oxidase (NOX)
synthesis.[8] Key to high enzyme activity is the regenerates the co-substrate NAD+[13]. In the EM-
oriented immobilization on surfaces or polymers GalT, β1,4-galactosyltransferase (GalT) transfers
mediated by terminal peptides or proteins. Leloir-type galactose from UDP-Gal onto a tert-butyloxycarbonyl
glycosyltransferases and enzymes for nucleotide protected N-acetylglucosamine (GlcNAc-linker-tBoc,
sugar synthesis have been immobilized via their 6) to synthesize N-Acetyl-D-lactosamine (LacNAc-
polyhistidine tags.[9] Immobilized His-tagged linker-tBoc, 7).[14] The UDP-GlcA, produced in the
enzymes
EM-UDP-GlcA, was used for glucuronosyltransferase
can show activities close to those in solution due to (GlcAT) to transfer the GlcA onto LacNAc-linker-
the highly specific orientation provided by the His6 tBoc in the EM-GlcAT, resulting in the product non-
linkage.[10] Magnetic particles are well suited as sulfated human natural killer cell-1 (HNK-1) epitope
carriers because of their easy separation by magnetic (8).[15]
fields, thus resulting in an enzyme-free product
without time-consuming and expensive purification glycan synthesis by enzyme cascade reactions,
steps. conducted in the CFMS, using six different enzymes
In this work, we demonstrate automated enzymatic
In addition, magnetic carriers simplify the immobilized on magnetic particles (Scheme 1A).
implementation of automated processes and protocols, Nucleotide sugars are synthesized in parallel enzyme
thus allowing the automation of complex multistep modules (EM) and subsequently delivered to the
enzymatic cascades. A sophisticated example is reaction of two glycosyltransferases to synthesize the
demonstrated in this work by handling enzymes final product 8 with a yield of 96% (Scheme 1B).
immobilized on magnetic particles in an automated Product formation of individual glycosyltransferase
compartmented flow microreactor system (CFMS). steps was confirmed by an inline MS-system. To the
The principle of the CFMS was described before[11] best of our knowledge, such a high number of
illustrating the system well suitable for automated magnetic enzyme beads (MEB) has not yet been
optimization of bioprocesses applying single employed for biocatalytic cascades before; it is the
immobilized enzymes, in this case, immobilized β1,4- first example for automated glycan synthesis in a
galactosyltransferase. The reaction progress can be compartmented
monitored online with spectroscopic methods and the
temperature can be controlled precisely.
microfluidic
microreactor.
Table 1. Loading yields of all immobilized enzymes investigated. The first column shows the theoretical loading in g
immobilized enzyme per L settled beads. The enzyme concentration used for immobilization was 0.15 g enzyme per L
reaction solution in all cases.
Theoretical
enzyme
Loading yields of His6-tagged enzymes [%]
loading (g/L)
UGDH NOX
GalK
USP
GalT
GlcAT
2.5
5
64
56
67
63
n.d
70
74
93
83
n.d
n.d
99
95
94
n.d
n.d
n.d.
78
n.d.
85
67
10
15
30
58
73
78
n.da
56
76
73
n.d.
n.d.
3
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
50
64
80
87
94
84
86
a not determined.
theoretical loading would be unusual since the
number of free binding sites decreases and more steric
hindrance occur. For the reaction cascade in the
CFMS, the 50 g/L MEB were used.
Results and Discussion
Enzyme loading yields of the immobilization step
Determination of the
optimum reaction
Table 1 displays an overview of the loading yields of
the PureCube Ni-IDA particles for all tested enzyme
amounts, expressed as theoretical loadings in g
immobilized enzyme per L settled beads. For UGDH
the loading yields varied between 58% and 67% and
for NOX between 56% and 80%. The loading yields
of GalK, GalT, and GlcAT were around the same
range of 73%-93%. Immobilizing USP led to loading
yields of nearly 100%. Analyzing correlations
between loading yield and theoretical loading, UGDH,
USP, and GlcAT show a decreasing trend of the yield
with increasing theoretical loading, the yield at 50 g/L
being an exception. No correlation of loading yield
and theoretical loading was found for immobilized
NOX, GalK and GalT, if the 50 g/L tests are excluded.
An increase of the loading yield with increasing
parameters of the magnetic enzyme beads MEB
To find the parameters giving the highest specific
activity U/mg immobilized enzyme, for each
immobilized enzyme, the effects of the reaction
parameters were analyzed in a partially automated
procedure in the CFMS in a ‘one-factor-at-a-time’
approach. The resulting optimal parameters of our
immobilized enzymes and the optimal parameters of
the respective free enzymes from literature are listed
in Table 2. As can be seen, the values of the optimal
reaction parameters for the immobilized enzymes
largely correspond with those reported in the
literature.[13,15-17]
Table 2. Experimentally determined optimal reaction conditions and kinetic data, compared to data from the literature in
brackets.
UGDH
NOX
GalK
USP
β4GalT
β3GlcAT
pH-optimum
9.5 (9.7)
6 (6-6.5)
7.5 (7.5)
7.5 (7.5)
7.5 (7.5)
6.5 (6.5)
Temperature
-optimum
40°C
(30°C)
25°C
(30-40°C)
40°C
(40°C)
45°C
(45°C)
40°C
(30°C)
50°C
(45°C)
Loading
2.5 g/L
2.5 g/L
5 g/L
50 g/L
50 g/L
15 g/L
(theoretical)
vmax
UDP-Glc:
0.12 U/mg
NADH+H+:
16.8 U/mg
Gal:
Gal-1-P:
GlcNAc-
linker-tBoc:
0.68 U/mg
(1.13 U/mg)
LacNAc-
linker-tBoc:
1.1 U/mg
1.9 U/mg
(5 U/mg)
10.6 U/mg
(13.1 U/mg)
(0.81 U/mg) (116 U/mg)
(0.48 U/mg)
NAD+:
0.11 U/mg
(0.92 U/mg)
ATP:
1.13 U/mg
(6.5 U/mg)
UTP:
11.7 U/mg
UDP-Gal:
1.1 U/mg
UDP-GlcA:
1.2 U/mg
(0.45 U/mg)
Km
UDP-Glc:
NADH+H+:
Gal:
Gal-1-P:
GlcNAc-
linker-tBoc:
0.72 mM
LacNAc-
linker-tBoc:
0.99 mM
0.09 mM
0.37 mM
1.54 mM
1.1 mM
(0.03 mM)
(0.024 mM)
(0.24 mM)
(0.83 mM)
(3.07 mM)
(0.24 mM)
NAD+:
0.16 mM
(0.44 mM)
ATP:
0.7 mM
(0.31 mM)
UTP:
1.6 mM
UDP-Gal:
5.4 mM
UDP-GlcA:
0.59 mM
(0.19 mM)
Ki
-
-
Gal:
UTP:
GlcNAc-
linker-tBoc:
3.24 mM
-
1.8 mM
1.87 mM
(33.1 mM)
(2.66 mM)
ATP:
3.15 mM
(5.8 mM)
4
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
[13,15]
[16,18]
[13]
[12]
[14,17]
[15]
Reference
The highest specific activity of immobilized UGDH,
NOX, and GalK was achieved with the lowest
loadings (see Figure S1, supporting information). By
increasing the enzyme load, the specific activity
decreased. This effect often occurs with immobilized
enzymes and is explained by the fact that immobilized
enzymes that are in close proximity to each other
influence each other by steric hindrance. This impairs
the binding of the substrates with the active sites and
thus the activity [19]. In contrast, immobilized USP,
GalT, and GlcAT displayed an increase of specific
activity by rising enzyme loading. In this case, steric
hindrance seems not to occur, which could be
explained by a superior arrangement of these
immobilized enzymes. The optimal values for pH and
temperature did not differ much from published data.
The occurrence of the maximum specific activity at
higher temperatures indicates a stabilization of the
MEB against heat denaturation. In comparison with
the literature, the Km value differed from around 2.5
times smaller to nearly 15 times higher and the vmax
value from about 8 times smaller to more than two
times higher (see Figure S2-S4, supporting
Recycling of the MEB
The results of the recycling of immobilized UGDH
are shown as the percentage residual activity in Figure
1A). After the second cycle, 81% residual activity
could be measured. Subsequently, it gradually
decreased by 9% to 16%. During the fourth cycle, the
residual activity of 55% was measured and in the
seventh cycle about 35 %. Wahl et al. achieved a
residual activity of 50% after 3 h reaction time with
free UGDH [13]. This corresponds to about six cycles,
where the immobilized UGDH showed about 37% of
the original specific activity. In the case of MEB
loaded with NOX, the specific activity decreased by
nearly 85% after the first recycling. Published data of
free NOX already shows the low stability of NOX,
with a half-life time of 10 min and 10–20% residual
activity after 30 min.[13] The results of the repeated
assays with immobilized GalK show a fairly constant
decrease of the residual activity until the fifth cycle.
At the last cycle, the residual activity of around 50%
remains about the same. For the immobilized USP,
information). The GalK, USP and GalT MEB the graph (see Figure 1B) shows a similar course,
displayed a substrate inhibition as stated in the
literature for the free enzymes.[12,13,17] Immobilization
via complex binding can cause a change in Km values,
for example, if the binding of the active site is
impaired by the Ni-IDA functionalization of the
magnetic particles. Also, if the immobilization causes
the orientation of the catalytic center which is not
optimal for the substrate to bind, the Km value can
change. In summary, the usefulness of the CFMS for
semi-automated parameter screening and optimization
of enzymatic reactions could be demonstrated with a
multitude of different immobilized enzymes and
tested parameters.
with a slightly higher residual activity of around 55%
after six assays. The specific activity of immobilized
β4GalT decreased by almost 40% after the first assay
but reached an approximately constant residual
activity of around 50% during the further cycles up to
seven repetitions. Finally, the course of the residual
activity of immobilized β3GlcAT did not change
significantly for the first three cycles. Subsequently,
the residual activity decreased to 61% at the sixth
cycle with an outlier in the fifth cycle. In summary,
with the exception of NOX, all other MEB showed a
residual activity of around 40 - 60% after six or seven
cycles. This offers the possibility for multiple reuses
of the MEB also in complex enzymatic cascades,
where the MEB could be easily recovered
Figure 1. Reuse of immobilized (A) UGDH, NOX, GalK and (B) USP, GalT, GlcAT in multiple cycles.
5
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
investigated (see Figure S5A/B, supporting
and replaced individually, while the use of free
enzymes would lead to a mixture after the first run of information). With 2 mM of the starting substrates 1
the cascade with limited use for further runs.
and 4, EM-UDP-GlcA and EM-UDP-Gal reached
conversions of 97.7% and 99.2%, respectively. The
EM-GalT performed with 99.7% conversion of the
available substrate 6. Finally, after additional 60 min,
the EM-GlcAT converted 98% substrate 7 from EM-
GalT and 99.3% substrate 5 from EM-UDP-GlcA,
resulting in a total yield of 96.3% for 8 with respect to
initial substrate concentrations (Figure 2A). The
respective HPLC analysis is displayed in Figure S6 in
the supporting information.
Automated enzymatic glycan synthesis
With optimal reaction conditions for MEB, the
synthesis of the non-sulfated HNK-1 epitope in the
reactor system was investigated. Different mixing
ratios of MEB and substrate concentrations were
Figure 2. Yields of the optimized reaction cascade with MEB in the reactor system as (A) experimental, (B) simulated
data. Applied enzyme modules: (I) EM-UDP-GlcA, (II) EM- UDP-Gal, (III) EM-GalT and (VI) EM-GlcAT. Intermediates
and products: (-◌-) UDP-GlcA (5), (-■-) UDP-Gal (3), (-▼-) LacNAc-linker-tBoc (7) and (-△-) non-sulfated HNK-1
epitope (8). For the substrates, 2 mM of UDP-Glc, Gal, ATP, UTP, GlcNAc-linker-tBoc, 4 mM NAD+, 8 mM MgCl2 and
6 mM MnCl2 were used (see Table 3). The initial concentrations of the immobilized enzymes were 1 µg/µL for UGDH
and NOX and 0.5 µg/µL for GalK, USP, GalT and GlcAT.
Product formation in the MEB CFMS system of 100%. In the EM-UDP-Gal GalK had initially
(Scheme 1B) was analyzed by ESI-Q-ToF MS. The converted almost 91% of 1 and USP 83% of 2, which
mass spectrum of the reaction solution before adding led to the total yield of 75.5% for UDP-Gal 3. The
β4GalT (Figure 3A) shows the substrate GlcNAc- free β4GalT also converted almost 83% of 3.
linker-tBoc 6 with a molecular mass (m/z) of 422. The
In comparison, the use of immobilized enzymes in
molecular mass m/z 444 Da derives from the sodium the reactor system could synthesize almost 40% more
adduct replacing a proton by a sodium of the buffer non-sulfated HNK-1 epitope. The advantage of
solution. Transfer of galactose (180 Da) by β4GalT compartmentation is that optimal reaction conditions
onto 6 results in a water molecule and the 162 Da for a defined enzyme module are applied. In addition,
bigger LacNAc-linker-tBoc 7 (Figure 3B). In the EM- the time- and cost-intensive removal of soluble
GlcAT, the 194 Da GlcA from previously produced enzymes from intermediate and final product volumes
UDP-GlcA is transferred resulting in the 760 Da non- is not necessary with the MEB.
sulfated HNK-1 epitope 8 (Figure 3C).
For comparison, the same reaction sequence was
performed with soluble enzymes with the same
reaction solution composition (see Table 3). In
contrast to the application of MEB, the soluble
enzymes could not be separated from the reaction
solution after use and remained in the solution during
the following steps of the cascade. After 210 min, a
total yield of 57.5% for 8 was achieved (Figure S5C,
supporting information). The EM-UDP-Gal had the
lowest yield of 75.5%, the EM UDP GlcA the highest
Theoretical simulation of automated enzymatic
glycan synthesis
In addition to the experimental work, the reaction
cascade with MEB was also simulated based on the
kinetic data obtained with single MEB systems (see
Figure 2B). The simulated EM-UDP-GlcA and EM-
UDP-Gal both achieved a conversion of 100%,
whereby in the experiment the EM-UDP-GlcA had
97.7% and EM-UDP-Gal 99.2% conversion. Based
6
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
on these substrate amounts, EM-GalT was simulated accessible substrate, which led to a total yield of
to convert 99.2%. In the experiment, EM-GalT 99.2%. The experimental EM-GlcAT converted 98%
surpassed the simulation and converted 99.7% of 6. of 7 and 99.3% of 5, resulting in the total yield of
Further, simulated EM-GlcAT converted 100% of the 96.3%
for
8.
7
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
Figure 3. MALDI-TOF mass spectra of the substrates and products of each step of the cascade reaction of immobilized
GalT and GlcAT. In A: the substrate GlcNAc-linker-tBoc (6) is marked, calculated mass (m/z): 422.2, in (B) the
intermediate LacNAc-linker-tBoc (7): calculated mass (m/z): 584.2; (C) the product non-sulfated HNK-1 epitope (8):
calculated
mass
(m/z):
760.3.
All
samples
were
diluted
1:10000.
beads have a binding capacity of up to 70 mg (IDA) His-
tagged protein/mL of settled beads.
The total turnover number (TTN) of the experiment is
therefore 0.37 g product * g-1 of all immobilized
enzymes and for the simulation 0.38 g product * g-1.
Thus, the simulation almost matches the experiment.
Based on the experiment in the CMFS and the results
of the recycling experiments, the reactor system
would be able to produce in total 40 mg non-sulfated
HNK-1 epitope in 18 h, which would be 6 cycles. It
was assumed that a 3 mL compartment with 5 mM of
substrates would be used and that six cycles with the
same enzymes would be run. This would result in a
space-time yield (STY) of HNK-1 epitope of 17.63 g
* L-1 * day-1. Considering the simulation, it can be
possible to reach a STY of 18.16 g * L-1 * day-1.
Enzymes and chemicals
The enzymes studied in this work are fusion constructs
with an N-terminal polyhistidine (His6)-tag. The enzyme
production was done as described in our previous studies
[15]
by Engels et al.
for UDP-glucose-dehydrogenase
(UGDH, EC 1.1.1.22), NADH oxidase (NOX, EC 1.6.99.-)
and β1,3glucuronyltransferase (β3GlcAT, EC 2.4.1.17, and
by Wahl et al.[13] for galactokinase (GalK, EC 2.7.1.6) and
[12]
for UDP-sugar pyrophosphorylase (USP, EC 2.7.7.64)
,
and by Fischöder et al.[14] for β1,4-galactosyltransferase
(β4GalT, EC 2.4.1.38). The linker‐modified substrate N-
acetylglucosamine with a tert-butyloxycarbonyl protected
amino group (GlcNAc-linker-tBoc) was kindly provided by
Prof. Vladimír Křen (Institute of Microbiology, Czech
Academy of Sciences).[21] The inhibitory by-products PPi
and UDP were removed by inorganic pyrophosphatase
(PPiase) and alkaline phosphatase (FastAP), respectively,
purchased from Thermo Fisher Scientific (Rockford, USA).
Standard compounds comprising Uridine 5’-diphosphate-α-
D-glucose (UDP‐Glc), nicotinamide adenine dinucleotide
(NAD+), nicotinamide adenine dinucleotide hydride
(NADH), D-(+)-Galactose (Gal), adenosine 5’-triphosphate
ATP, uridine 5’-triphosphate (UTP), α-D-Galactose-1-
phosphate (G-1-P), Uridine 5’-diphosphate-α-D-galactose
(UDP-Gal) and Uridine 5’-diphosphate-α-D-glucuronic
acid (UDP-GlcA) were purchased from Sigma Aldrich
(Deisenhofen, Germany) or VWR International GmbH
(Bruchsal, Germany). All chemicals were analytical grade
and used without further purification. The water used for
all experiments was deionized and purified using a Milli-Q
Ultrapure system from Merck Millipore KGaA (Darmstadt,
Germany).
Conclusion
We here demonstrate for the first time glycan
synthesis in an automated compartmented flow
microreactor by a reaction cascade of six different
magnetic enzyme beads. Compared to other published
enzyme cascades in microreactors utilizing enzymes
immobilized on magnetic beads,[20] the number of
applied enzymes and the degree of automation is
clearly increased, indicating the flexibility of the
applied immobilization method and of the developed
compartmented reactor device. Despite the high
number of reaction steps, a yield of 96% non-sulfated
HNK-1 epitope could be achieved. Thus, compared to
an approach with soluble enzymes, the yield was
almost 40% higher. This again demonstrates the great
advantage
of
compartmentalization
in
the
microreactor system, which can provide optimal
reaction conditions for each reaction step and prevent
inhibition. The microreactor system can product-
specific to an MS-system for inline reaction control
and product analysis. Due to the modularity and
multiplexing of the microreactor system, the
productivity could be increased by adding more tubes
for parallel reactions or even more complex cascades
consisting of more enzymes could be utilized. With
further immobilized enzyme modules in hand,
automated synthesis of more complex glycans can be
realized in a compartmented flow microreactor.
Immobilization of the enzymes
Purified His6-tagged enzymes were immobilized onto the
IDA-Ni functionalized surfaces of the magnetic particles.
For the immobilization, 5 µL particle slurry were used to
bind 12.5/25/50/75/150/250 µg of the enzymes resulting in
a theoretical loading of 2.5/5/10/15/30/50 g enzyme per L
particle slurry. The immobilization reaction was performed
at room temperature for 30 min in 1.5 mL Eppendorf tubes
(Wesseling-Berzdorf, Germany) and done in duplicates,
using 250 µL binding buffer. The binding buffer consisted
of 20 mM Na2PO4/200 mM NaCl at pH 6.8. After the
immobilization step, the samples were washed with (0.5
mL) 1 M NaCl buffer and twice with storage/binding
buffer 1 (0.5 mL) 100 Tris/25 mM KCl buffer pH 7.5 or
storage/reaction buffer 2 (0.5 mL) 50 mM Tris buffer pH
8.7 to desorb non-bound enzymes. The enzyme particles
were diluted to 10 µL particle slurry per mL storage buffer
and stored at 4°C. To calculate the amount of bound
enzyme, the protein content in the incubation- and
Experimental Section
Magnetic microcarriers
PureCube Ni-IDA MagBeads were purchased from Cube
Biotech GmbH (Monheim am Rhein, Germany) and used
as enzyme carriers without further modification. The
ferrimagnetic magnetite beads are coated with 6% cross-
linked agarose, functionalized with Ni-IDA and have a
diameter of 25-30 µm. According to the manufacturer, the
8
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
washing-supernatants was determined using the µm, 4.6
x
150 mm). The concentration of the
bicinchoninic acid (BCA) Protein Assay Reagent Kit from substrates/cosubstrates and the products/coproducts were
Thermo Scientific Pierce (Rockford, USA) according to calculated by the ratio of each peak area compared to the
manufacturer’s instructions.
sum of the two peak areas and the corresponding
calibration curve of standards. From the resultant
concentrations, the activity was calculated by the increase
of product over time. One enzyme unit
Activity assay of immobilized enzymes in the
CFMS
correlates with one μmol product per minute. The mass-
specific activity U/mg was calculated for one enzyme unit
per mg immobilized enzyme.
To characterize the activity of the MEB in the microreactor
system with regard to pH, temperature, and loading, MEB
were pumped into the reaction tubing of the system. The
MEB were separated from the storage buffer (as described
before) [22], resuspended in 100 µL of a preheated reaction Recycling of the MEB
buffer compartment and pumped to the temperature control
module (TCM). Details of the IR-unit for preheating and
the TCM are discussed by Heinzler et al.,[11] describing the
reactor system. Compound composition for the activity
assays of the analyzed enzymes are listed in Table S1
(supporting information). For each assay, 15 µg of MEB
with a loading of 15 mg immobilized enzyme per mL
settled beads were used in a 100 µL compartment. To avoid
possible effects of enzyme deactivation, fresh MEB of the
same batch were used in each assay, except for the assays
on particle recycling. To test different temperatures, the
activity of 15 mg/mL MEB was investigated at the
respective optimal pH value of the free enzymes according
to literature (UGDH/NOX/GalK,[13] USP,[12] GalT,[14]
GlcAT[15]). The influence of different pH values was
investigated at the respective optimal temperature of the
free enzymes. Assays with various enzyme loadings were
conducted at the optimal pH values and temperatures. For
the kinetic studies, 15 mg/mL MEB were used at the
optimal conditions with variable substrate concentrations.
The kinetic constants were obtained by fitting a Michaelis-
Menten equation v=vmax*[S]/(Km +[S]) or a Michaelis-
Menten equation including additional substrate inhibition
v= vmax/(1+ Km/[S]+[S]/Ki) to the experimental data, using
the software Sigma Plot 11 (Systat Software GmbH,
Erkrath, Germany). If two substrates were used, the
substrate, which was not varied in each case, was added in
excess. Thus, limiting effects could be avoided and the
corresponding kinetic parameters could be determined for
each substrate. Assay conditions for kinetic studies are
listed in Table S2 (supporting information). To determine
the specific activity of the enzymes UGDH and NOX, the
concentrations of the coproduct NADH in the samples were
determined by measuring the absorption of the samples at
340 nm. For GalK, USP, GalT and GlcAT 30 μL of a
sample of the sample were injected in an HPLC system
(Agilent 1100 series, Waldbronn, Germany) and measured
at 254 nm. For analytics of GalK and USP, a normal phase
column (TSKgel Amide-80, Tosoh Bioscience, Germany, 5
µm, 2.0 x 250 mm) was used and for analytics of GalT and
GlcAT a reversed-phase column (SunFire, Waters, USA, 5
To determine the recyclability of the different MEB,
several activity assays were carried out in succession in
which the MEB were reused. Between the reactions, the
MEB were separated and washed three times with 200 µL
reaction buffer. Thereafter, they were resuspended in a new
reaction solution and transported to the TCM for the next
cycle. To have enough volume for taking samples, a 300 µl
compartment was used with the same compound and MEB
concentration as stated before. The specific activity of the
first assay of the respective enzyme was set as 100%
residual activity.
Reaction cascade in the CFMS
The reaction cascade (see Scheme 1B) was conducted in
two separate tubes of the temperature control module
(TCM) of the CFMS (Scheme S1A and B, supporting
information) at 37°C in 150 µL compartments. A
compartment is an aqueous reaction solution, separated by
ethyl acetate (EtOAc) and created by alternately pumping
EtOAc and reaction solution within a tubing (see Scheme
S1C, supporting information). The corresponding
components of the reaction solutions of the compartments
are listed in Table 3. The concentrations of the immobilized
enzymes were 1 µg/µL for UGDH and NOX and 0.5 µg/µL
for GalK, USP, GalT and GlcAT. Synthesis of UDP-GlcA
was conducted for 150 min parallel to a synthesis of
LacNAc-linker-tBoc. The reaction compartment EM-UDP-
Gal contained MEB and 1 µL of 0.1 U/µL PPiase to
generate UDP-Gal within 90 minutes reaction time. After
separation of MEB, the product compartment was
transported to the MEB of the EM-GalT. The solution
already contained the substrate GlcNAc-linker-tBoc and 1
U/µL FastAP was added to hydrolyze the byproduct UDP.
GalT synthesized LacNAc-linker-tBoc within 60 minutes.
In a similar way, UDP-GlcA was
Table 3. Composition of the reaction mixtures utilized in the optimized reactor cascade experiments in the CFMS and in
the approach with soluble enzymes. The volume of the EM-UDP-GlcA, EM-UDP-Gal and the reaction solution for the
soluble enzymes was 150 µL respectively. All experiments were conducted at 37°C.
Experiments
(all numbers in mM)
9
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
Reaction
cascade
EM-UDP-
100
-
2
-
4
-
-
-
-
-
-
GlcA
in CFMS
EM-UDP-
Gal/
100/
25
-
2
2
2
8
6
6
2
EM-GalT
100/
25
Approach with enzymes in
solution
-
2
4
2
2
2
8
2
produced in a parallel reaction compartment within 150 the reaction solutions after 90 min, before starting the EM-
min reaction time. After MEB separation, both product GalT, after 150 min, before starting the EM-GlcAT, and at
compartments were combined and added to the MEB of the the end of the reaction cascade after 210 min were
EM-GlcAT. The synthesis was conducted for 60 minutes. measured (Scheme 1B)
The product concentrations were determined by HPLC.
For analytics of EM-UDP-GlcA, a reversed-phase
column (TSKgel ODS-100V, Tosoh Bioscience, Germany,
4.6 x 150 mm) was used. The mobile phase consisting of
Acknowledgments
30% (v/v) of 50 mM ammonium acetate (pH 4.5) in
acetonitrile was used for 30 min at a flow rate of 0.3
mL*min-1, increased after 10 min to 40% (v/v) in 5 min
and after 5 min decreased in 2 min to 30% (v/v). At 260 nm,
30 µL of the sample were analyzed. The EM-UDP-Gal was
analyzed with the normal phase column TSKgel Amide-80
using 20 mM tert-Butylamine as mobile phase at a flow
rate of 0.9 mL*min-1. In 35 min 10% (v/v) methanol was
added. A sample of 30 µL was analyzed at 260 nm.
Analytics of the EM-GalT and EM-GlcAT were done with
the reversed-phase column SunFire with 24% (v/v)
acetonitrile dissolved in MilliQ water with 0.1 % formic
acid as eluent at a flow rate of 0.3 mL*min-1. For 30 min,
30 µL of sample mas measured at 254 nm.
In addition to the experiment in the CFMS, the reaction
cascade was simulated with the application SimBiology
(MathWorks, USA). The same reaction periods as in the
experiment (see Scheme 1B) were used. The previously
obtained kinetic data (see Table 2) were used as parameters
for the simulation. For substrate concentrations, the same
amounts were utilized as in the experiment (see Table 3).
For comparison, an experiment was conducted with the
same amount of soluble enzymes in the same parallel
reaction sequence using the same composition of the
reaction solutions in the reactor system. The soluble
enzymes were sequentially added to the reaction solutions
for each EM. GalT was added to the reaction solution of
EM-UDP-Gal after 90 min and subsequently incubated for
a further 60 min. In parallel, the EM-UDP-GlcA was
conducted for 150 min. The solutions of both reactions
were combined and GlcAT was added. The reaction was
stopped after 60 min, by adding 50 µL acetonitrile (ACN).
The authors thank Prof. Dr. Vladimír Křen and co-workers
(Academy of Science of the Czech Republic) for providing
GlcNAc-linker-tBoc. The authors gratefully acknowledge
financial support by the Federal Ministry for Education and
Research (BMBF) through the project “The Golgi Glycan
Factory 2.0” (AZ: 031A557D and AZ: 031A557A) as part of the
BMBF program Biotechnology 2020+—Basic Technologies and
the support by Deutsche Forschungsgemeinschaft and Open
Access Publishing Fund of Karlsruhe Institute of Technology.
References
[1]
a) R. K. Yu, Y. Suzuki, M. Yanagisawa, FEBS
Lett. 2010, 584, 1694-1699; b) K. Ohtsubo, J. D.
Marth, Cell 2006, 126, 855-867.
[2]
a) C.-C. Wang, J.-C. Lee, S.-Y. Luo, S. S.
Kulkarni, Y.-W. Huang, C.-C. Lee, K.-L. Chang,
S.-C. Hung, Nature 2007, 446, 896; b) X. Zhu, R.
R. Schmidt, Angew. Chem., Int. Ed. 2009, 48,
1900-1934.
[3]
[4]
a) C.-H. Hsu, S.-C. Hung, C.-Y. Wu, C.-H. Wong,
Angew. Chem., Int. Ed. 2011, 50, 11872-11923; b)
M. Guberman, P. H. Seeberger, J. Am. Chem. Soc.
2019, 141, 5581-5592; c) M. Panza, S. G. Pistorio,
K. J. Stine, A. V. Demchenko, Chem. Rev. 2018,
118, 8105-8150.
a) K. W. Moremen, A. Ramiah, M. Stuart, J. Steel,
L. Meng, F. Forouhar, H. A. Moniz, G. Gahlay, Z.
Gao, D. Chapla, S. Wang, J.-Y. Yang, P. K.
Prabhakar, R. Johnson, M. d. Rosa, C. Geisler, A.
V. Nairn, J. Seetharaman, S.-C. Wu, L. Tong, H. J.
Gilbert, J. LaBaer, D. L. Jarvis, Nat. Chem. Biol.
2017, 14, 156; b) C. A. G. M. Weijers, M. C. R.
Franssen, G. M. Visser, Biotechnol. Adv. 2008, 26,
436-456; c) I. Brockhausen, Front. Immunol. 2014,
Inline connection between the reactor and a mass
spectrometer
To provide inline analytics, the reactor system was
connected to an ESI-Q-ToF MS-System (QSTAR Pulsar I,
SCIEX, USA) (see Scheme S2, supporting information).
For this purpose, a syringe pump, an additional high-
pressure multiport valve (Cetoni, Korbußen, Germany) and
a mixing reservoir (1.5 mL Eppendorf tube, Wesseling-
Berzdorf, Germany) were added to the reactor system. The
syringe pump delivers samples to the mixing reservoir via
the multiport valve, where they are diluted with Milli-Q
water before being pumped into the MS system. Samples of
5; d) P. Bojarová, V. Křen, Trends
Biotechnol.
2009, 27, 199-209; e) L. Engels, L. Elling, Engels,
in: Handbook of Carbohydrate-Modifying
Biocatalysts (Eds.: P. Grunwald, ed.), Stanford
Publishing, Stanford, 2016, pp 297-320.
[5]
a) P. Sears, C.-H. Wong, Science 2001, 291, 2344-
2350; b) R. L. Halcomb, H. Huang, C.-H. Wong, J.
Am. Chem. Soc. 1994, 116, 11315-11322; c) O.
10
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
Blixt, T. Norberg, J. Carbohydr. Chem. 1997, 16, [21]
B. Sauerzapfe, K. Křenek, J. Schmiedel, W. W.
Wakarchuk, H. Pelantová, V. Křen, L. Elling,
Glycoconjugate J. 2009, 26, 141-159.
J. Hübner, R. Heinzler, C. Arlt, S. Hohmann, W.
Brenner, M. Franzreb, React. Chem. Eng. 2017, 2,
349.
143-154; d) T. Matsushita, I. Nagashima, M.
Fumoto, T. Ohta, K. Yamada, H. Shimizu, H.
Hinou, K. Naruchi, T. Ito, H. Kondo, S.-I. [22]
Nishimura, J. Am. Chem. Soc. 2010, 132, 16651-
16656; e) X. Huang, K. L. Witte, D. E.
Bergbreiter, C. H. Wong, Adv. Synth. Catal. 2001,
343, 675-681; f) J. Zhang, C. Chen, M. R. Gadi, C.
Gibbons, Y. Guo, X. Cao, G. Edmunds, S. Wang,
D. Liu, J. Yu, L. Wen, P. G. Wang, Angew. Chem.
2018, 130, 16880-16884.
[6]
a) C. Cai, D. M. Dickinson, L. Li, S. Masuko, M.
Suflita, V. Schultz, S. D. Nelson, U. Bhaskar, J.
Liu, R. J. Linhardt, Org. Lett. 2014, 16, 2240-
2243; b) J. Hwang, H. Yu, H. Malekan, G.
Sugiarto, Y. Li, J. Qu, V. Nguyen, D. Wu, X.
Chen, Chem. Commun. 2014, 50, 3159-3162; c) T.
Li, L. Liu, N. Wei, J.-Y. Yang, D. G. Chapla, K.
W. Moremen, G.-J. Boons, Nat. Chem. 2019, 11,
229-236.
[7]
[8]
a) L. Wen, G. Edmunds, C. Gibbons, J. Zhang, M.
R. Gadi, H. Zhu, J. Fang, X. Liu, Y. Kong, P. G.
Wang, Chem. Rev. 2018, 118, 8151-8187; b) N. L.
B. Pohl, Nat. Chem. 2019, 11, 201-203.
a) T. Ito, R. Sadamoto, K. Naruchi, H. Togame, H.
Takemoto, H. Kondo, S.-I. Nishimura, Biochem.
2010, 49, 2604-2614; b) K. Naruchi, S.-I.
Nishimura, Angew. Chem. 2011, 123, 1364-1367;
c) N. Nagahori, K. Niikura, R. Sadamoto, M.
Taniguchi, A. Yamagishi, K. Monde, S.-I.
Nishimura, Adv. Synth. Catal. 2003, 345, 729-734.
J. Nahalka, Z. Liu, X. Chen, P. G. Wang, Chem. -
Eur. J. 2003, 9, 372-377.
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
T. Cha, A. Guo, X.-Y. Zhu, PROTEOMICS 2005,
5, 416-419.
R. Heinzler, J. Hübner, T. Fischöder, L. Elling, M.
Franzreb, Front. Bioeng. Biotechnol. 2018, 6, 189.
C. Wahl, M. Spiertz, L. Elling, J. Biotechnol.
2017, 258, 51-55.
C. Wahl, D. Hirtz, L. Elling, Biotechnol. J. 2016,
11, 1298-1308.
T. Fischöder, D. Laaf, C. Dey, L. Elling,
Molecules 2017, 22, 1320.
L. Engels, M. Henze, W. Hummel, L. Elling, Adv.
Synth. Catal. 2015, 357, 1751-1762.
W. Hummel, B. Riebel, Biotechnol. Lett. 2003, 25,
51-54.
C. Rech, R. R. Rosencrantz, K. Křenek, H.
Pelantová, P. Bojarová, C. E. Römer, F. G.
Hanisch, V. Křen, L. Elling, Adv. Synth. Catal.
2011, 353, 2492-2500.
[18]
[19]
[20]
Z. Findrik, A. Vrsalović Presecki, D. Vasić-Racki,
J. Biosci. Bioeng. 2007, 104, 275-280.
Z. Guo, S. Bai, Y. Sun, Enzyme Microb. Technol.
2003, 32, 776-782.
T. Peschke, M. Skoupi, T. Burgahn, S. Gallus, I.
Ahmed, K. S. Rabe, C. M. Niemeyer, ACS Catal.
2017, 7, 7866-7872.
11
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900709
Advanced Synthesis & Catalysis
FULL PAPER
Toward Automated Enzymatic Glycan Synthesis in
a Compartmented Flow Microreactor System
Adv. Synth. Catal. Year, Volume, Page – Page
Raphael Heinzler, Thomas Fischöder, Lothar
Elling*, Matthias Franzreb*
12
This article is protected by copyright. All rights reserved.