Glycosylation reactions in vitro separate the synthesis of oligosaccharides from cell metabolism. Purified enzymes and chemically synthesized sugar donors are used in a reaction vessel to produce homogeneous glycoforms. Compared to in-cell synthesis, advantages are greater control over the glycan structure and faster process development. Steps typically include expression of enzymes and preparation of activated nucleotide sugars, followed by either a one-pot reaction or a microfluidic synthesis with inline analysis and diafiltration purification. Drawbacks are expense of recombinant, often membrane tethered, glycosyltransferases and sugar donor handling.
In vitro glycosylation refers to glycosyltransferase-catalyzed sugar-chain assembly reactions performed in aqueous buffer in the absence of cells. Purified glycosyltransferases (or endoglycosidases/glycosynthases) transfer sugars from activated donors onto protein or peptide acceptors reproducing critical aspects of secretory pathway chemistry without cellular metabolism, competing biosynthetic pathways or variation due to membrane-trafficking.
Cell-free synthetic glycobiology systems for on-demand biosynthesis of designer glycomolecules.1,5
in vitro glycosylation can refer to any glycosylation reaction performed outside of a living cell. Reactions can be as simple as a single glycosyltransferase acting in a one-pot reaction or as complex as multiple enzymes acting in a defined order to recreate glycosylation steps occurring in the Golgi apparatus. They can also include chemoenzymatic steps, for example using chemical sugar donors like sugar oxazolines or fluorides with mutant glycosyltransferases (glycosynthases) to build unnatural sugar molecules or attach unnatural groups like azides, alkynes or photocleavable groups. Cell-free biosynthetic systems are created by adding glycosyltransferases expressed using cell-free protein synthesis to crude bacterial lysates and then mixing them together to create novel pathways. Using this technique, 37 hypothetical pathways were made and allowed for 23 unique glycans to be built in one publication. Microfluidic versions of this allow for protein synthesis, glycosylation and product capture to occur in separate chambers, recreating the idea of compartmentalization while allowing for dynamic control over each chamber. Protein glycosylation can be used for homogenizing therapeutic proteins, adding adjuvants to vaccine antigens or creating glycans arrays for binding assays. in vitro synthesis of lipid-linked oligosaccharide donors can be accomplished using yeast microsomes along with purified ALG1 to allow for enzymatic N-glycosylation reactions to occur without any cells. The advantage of these systems is that because the user is supplying all components used in the reaction they can manipulate the concentration of each nucleotide sugar to favor formation of one glycoform over another, add isotopic labeling to study enzyme mechanisms or link reactions together to high-throughput analyze glycosylation site preferences. With this ability, glycosylation can be turned into a predictable reaction instead of a stochastic event. However, proper characterization of enzymes and donors must be done to prevent inconsistencies.
The biggest difference between cell-based and cell-free synthesis is that, in cells, you're working with a pre-existing chassis defined by the relative expression levels of each enzyme in different subcellular locations. By contrast, synthesis in vitro is user-programmed with specific stoichiometry and reaction ordering. Competition between enzymes for shared donors and export pathways as well as cell-cycle dependent expression programs all contribute to micro- and macro-heterogeneity which persists through purification steps and complicates lot-to-lot release criteria. This 'background noise' is removed in cell-free systems by supplying recombinant enzymes at designed ratios and continuously feeding donors at a defined rate, achieving batch-to-batch cosine similarity >0.95 straight out of the box. Turnaround time differs as well. It typically takes 6-12 months of cell-line development and clone screening just to demonstrate stable glycoform output before process development can begin; in cell-free synthesis, one can iterate prototypes in the matter of days with synthetic gene circuits and commercial-grade lysates. There are challenges unique to cell-free synthesis that cells can 'get away' with, too. For example, many glycosyltransferases are membrane bound and need to be solubilized, stabilized, and purified into functionality. This often requires reconstitution into detergent micelles or nanodiscs, increasing costs and fouling your NMR. Lipid-linked oligosaccharide donors must also be pre-built using separate enzymatic assembly lines as they cannot be recycled from intracellular pools. Finally, there's a different set of regulatory considerations; cell-based synthesis benefits from established biologics frameworks, while in-vitro methods will need to be validated for donor impurities, enzyme clearance, and device leachables from microfluidics platforms. This leads to another major difference between the two approaches: cost. In cells, you're amortizing the cost of enzyme expression over total cell biomass. in vitro, you buy or over-express every enzyme individually, so the cost of your donors and the stability of your enzymes become key levers for cost savings.
Table 1 Comparative architecture of in vitro versus cell-based glycosylation
| Feature | in vitro (cell-free) | Cell-based (mammalian) | Strategic implication |
| Heterogeneity source | Donor/enzyme batch | Golgi dynamics | Lot consistency |
| Development timeline | Days to weeks | Months to years | Speed to clinic |
| Donor logistics | Synthetic/exogenous | Metabolic synthesis | Cost concentration |
| Membrane protein handling | Purification burden | Native insertion | Technical hurdle |
| Non-natural sugar scope | Permissive | Restricted | Functionalisation |
Steps generally include independently preparing the protein acceptor substrate, glycosyltransferases enzymes, and nucleotide-linked sugar donors, followed by either stepwise or simultaneous incubation reactions in aqueous buffer solution(s), inline analytical assessment of reaction progress, and purification by diafiltration. Ultimately, what could take months of cell-line engineering can be accomplished in days using a series of benchtop enzymatic reactions that recapitulate the enzymatic logic of the secretory pathway in a defined and modular way outside of the cell.
First, the protein of interest is purified from expression hosts. Native glycans are trimmed down to a consistent GlcNAc stub through treatment with Endo-S or Endo-M. This step is typically validated by liquid chromatography-mass spectrometry to ensure >98% conversion. This must be quantitative since any remaining complex or high mannose glycans would introduce unwanted variability and defeat the purpose of complete control in vitro. It is then buffer exchanged into a metal and low conductivity buffer using tangential flow filtration to ensure removal of glycerol/detergents/nucleotide sugars which will inhibit/downregulate glycosyltransferases. The protein is confirmed using SDS-PAGE to ensure molecular weight decrease and LC-MS to ensure desired stub population is >98%. Next, the protein acceptor is filtered sterilized and cooled until ready for use.
Choice of enzyme(s) depends on the desired glycan structure. For N-glycan engineering, endoglycosynthase mutants are often used since these have had their hydrolytic activity removed, but retained transglycosylation capability. These enzymes allow the orthogonal addition of pre-activated oxazoline donors onto the GlcNAc stub in one step. Addition of individual sugars for elongation is achieved using recombinant β1,4-galactosyltransferase, α2,3/α2,6-sialyltransferase and α1,3/α1,4-fucosyltransferase sequentially. Donor substrates (commonly UDP-Gal, CMP-Neu5Ac, GDP-Fuc) must be prepared endotoxin-free and metal-analyzed (trace copper or iron will catalyze donor β elimination stopping reaction). Addition of modified sugars (eg click enabled azido-sugars, sulphated donors) must be custom synthesized with proof of enzyme acceptance. Ratio of enzyme to acceptor is fixed at process validation stage. Insufficient enzyme will result in incomplete conversion, however too much enzyme can lead to secondary hydrolysis (if using wild-type enzymes) or non-specific proteolysis.
The goal is to maintain enzyme activity while also maintaining donor stability and protein stability. Therefore the buffer pH is typically chosen such that it falls within the optimal pH range of most glycosyltransferases and such that donor hydrolysis is minimized. Reaction temperatures are kept at room temperature or below to match physiological conditions. Higher temperatures will increase reaction rates but also the rate of UDP-sugar β elimination so a kinetic model of the reaction will be utilized to determine the ideal timeframe/temperature range. Necessary metal ions are added at catalytic concentrations. All metal ions must also be Chelex-treated to remove trace metals that will promote side reactions. The reaction volume should initially be minimized to avoid unnecessary use of donor, but can be linearly scaled up once reaction kinetics have been established and validated. Reaction quenching can occur inline with addition of EDTA/acetonitrile. Monitoring of reaction conversion by HPLC or Raman can help determine the rate of donor depletion and can be used in a feedback loop to control the rate of enzyme addition to maintain desired reaction velocity.
Glycosylation in vitro removes carbohydrate attachment from the complicated and variable biochemical systems found inside cells. This allows bioengineers to define precise levels of sialylation, branching structure and core-fucosylation without genetic manipulation or large scale clone sorting. The benefit is precision: all enzymes, sugars and reaction conditions are provided at known concentrations allowing for a predictable reaction that outputs one glycoform rather than a mixture of microspecies.
In contrast to protein glycosylation in cells, where Golgi trafficking rates, import rates of nucleotide-sugars and differential expression of competing enzymes result in a heterogeneous mixture of glycoforms, glycoproteins are trimmed in vitro in an open reactor system where all of these parameters can be tightly controlled. One chooses which endoglycosidase to use (Endo-S trimming an Fc protein or Endo-M trimming everywhere else) and trims down the native glycoprotein uniformly to a GlcNAc stub that conversion can be confirmed with LC-MS. Each rebuilding enzyme chosen is recombinant and has fully defined donor specificities: β1,4-galactosyltransferase will add only β1,4 Gal, α2,3-sialyltransferase will add only α2,3-linked sialic acid, and mutant enzymes can accept unnatural donors with azide or click functionalization. Traces metals are eliminated from the reaction cocktail due to the aqueous copper-free conditions preventing trace-metal catalyzed deglycosylation via donor β-elimination, allowing exact stopping of the reaction when the desired antenna size is reached. Enzyme feeds can be monitored with real-time Raman spectroscopy by following UDP-sugar depletion allowing for a feedback loop to be put in place that prevents over-building. Macro-heterogeneity caused by partial sequon occupancy is avoided as seen in cell culture by using a purified protein acceptor and fully exposing the GlcNAc stub quantitatively before beginning the rebuild. The finished product then acts as a single species during PK modelling, biosimilar evaluation, and regulatory filings allowing glycan structure to move from being a liability during development to an engineered attribute.
Table 2 Control levers in vitro versus cellular systems
| Control dimension | in vitro handle | Cellular constraint | Functional dividend |
| Enzyme stoichiometry | User-defined mixing | Expression-level competition | Predictable kinetics |
| Donor availability | Synthetic feed rate | Metabolic flux limitation | Custom motifs |
| Reaction timing | Temporal staging | Golgi transit time | Sequential fidelity |
| Spatial architecture | Microfluidic zones | Membrane topology | Compartment mimicry |
| Non-natural sugars | Engineered donors | Rejection/ toxicity | Functional tags |
Cell culture faces this challenge of reproducibility forever: passage-dependent enzyme drift, dissolved-oxygen crashes and medium tweaks alter glycoform ratios from one batch to the next. By bypassing these variables, in vitro glycosylation uses one lot of each enzyme, qualified against a reference substrate, provided in single-use aliquots that prevent freeze/thaw degradation. Buffers are treated to remove ppb-level metals. Donors are metal analyzed by ICP-MS before release. Oxidative side reactions that would hydrolyze UDP-sugars are avoided. The reaction takes place in microfluidic segment-flow reactors, which ensure laminar mixing and consistent temperature: every catalyst molecule feels the same pH, same donor concentration. Scale-up is direct: after confirming conversion kinetics at 5 mL, simply scale to 5 L applying the same stoichiometry, mixing physics and quench timing. No more comparability exercises to validate scale-up like those that plague cell-culture campaigns. Glycohomogeneity is confirmed by LC-MS overlaying the reference standard within ±5 % relative abundance of any peak; site-specific glycopeptide mapping then confirms occupancy and linkage purity. Since the entire process is defined in the electronic batch record as critical process parameters, every lot made going forward—even at commercial scale—retraces the same glycoform roadmap. The product released has identical specifications, linear expiry dating and minimal regulatory burden. For biosimilars, where the glycoform ensemble must overlay the originator within predefined windows, or maternal vaccines where minor glyco-drift necessitates time-consuming bridging studies, this reproducibility advantage is decisive.
When performed in vitro, glycosylation allows great control and precision, however issues still remain which can negatively impact development timelines. The process is highly dependent on a three-part combination including enzyme, donor substrate, and acceptor substrate. The purity and stability requirements for each are stringent; if any one suffers, the conversion efficiency decreases, glycoform ratios shift and heterogeneity that was thought to be abolished reappears in the form of micro-heterogeneity that resembles cell-culture noise.
The most pressing limitation is enzyme supply. Of the hundreds of glycosyltransferases encoded by nature, only a fraction have been cloned, expressed and characterised under preparative conditions; unusual linkages (sulfated Lewis-x, C-glycosides or phospho-glycans) necessitate dedicated directed-evolution efforts that take months or bespoke synthesis of unnatural donors that may not even be recognized by wild-type enzymes. And even if such an enzyme can be found, its specific activity may change erratically from expression lot-to-lot: a single oxidised methionine or N-terminal protease clip can skew the kinetic equilibrium towards hydrolysis over trans-glycosylation, turning the intended product into sugar and deglycosylated protein. Similarly, poor substrate tolerance restricts workflow: a β1,4-galactosyltransferase will not recognise β1,3-linked galactose or GlcNAc unless mutating the active site, requiring developers to procure, express and qualify another catalyst whenever a linkage is changed. Donor sugars (UDP-Gal, CMP-Neu5Ac) are costly at multi-kilogram scales and susceptible to β-elimination by aqueous buffers; trace metals (Cu2+, Fe3+) accelerate this decomposition. To top it all off, the protein acceptor itself can be problematic: any leftover high-mannose or hyper-sialylated species that escape purify will act as an alternative substrate, altering stoichiometry and generating unforeseen side-products that necessitate re-purification. As a result, every new enzyme expression lot and sugar donor batch must be qualified with a reference substrate before use in GMP manufacturing, making enzymatic glycosylation a supply-chain issue rather than one of simple reagent addition.
Expenses associated with scale-up kill most in vitro programs quietly. While bench-top reactions in academic labs can easily be performed on the milligram scale, gram-to-kilogram scale reactions require grams of UDP-sugar donor compounds which cost more than the protein product itself. Enzyme costs drive expenses too: recombinant glycosyltransferases need to be expressed in mammalian or insect cells for correct folding and post-translational modifications, but are typically produced at low yields and require purification over multiple chromatography columns driving cost-of-goods higher still. Recycling of enzymes through immobilization on magnetic beads or membrane-bound cartridges is possible, but creates another validation headache: removal of any leached catalyst to undetectable levels must be demonstrated, and loss of activity through multiple turnover cycles must be modeled in order to predict when the resin needs to be regenerated/replaced. Reaction volume can also cause problems at scale: large stirred tanks have mixing dead-zones that allow local UDP-donor depletion which causes the enzyme to fall-back on hydrolysis, creating a shoulder of under-glycosylated product not seen at smaller scales. To avoid this, developers need to incorporate static mixers or segmented-flow micro-reactors to ensure mixing times stay in the sub-millisecond range at high-flow rates, creating capital expense only justified for platform programs or highest-value targets. Downstream, the analytical burden increases with batch size: every production batch needs LC-MS glycoform profiling, site-specific glycopeptide analysis, and enzyme-activity validation, creating a quality control cost-center that must be amortized over the lifetime of the product. Taken together, these issues mean in vitro glycosylation platforms are primarily reserved for high-margin drug products (oncology antibodies, rare disease enzymes) but are cost-prohibitive for vaccines or other high-volume therapeutic diagnostics unless the process is defined very early-on and leveraged across multiple products.
Enzymatic glycosylation in vitro is also increasingly being used to engineer glycoproteins for therapeutic, diagnostic, and research purposes. This approach allows researchers and developers to attach defined glycans to purified proteins. This detachment of glycosylation from the host cell can allow for homogeneous glycosylation on therapeutic proteins where glycan pattern can impact function. This allows for consistency where host cell expression may vary, as well as the ability to sidestep challenges with expression hosts and customize glycan structures for better pharmacokinetics or to create unique conjugates. In vitro glycosylation has many applications including therapeutic antibodies, enzyme therapeutics, and vaccine antigens.
Synthetic glycosylation systems constructed from the bottom-up.2,5
One of the most developed uses of in vitro glycosylation technologies has been in the remodeling of Fc glycans of therapeutic antibodies in order to control their effector functions. For example, IgG antibodies contain one N-linked glycan on the Fc domain which can strongly modulate ADCC and CDC activity. Antibodies can be deglycosylated using glycosidases or fucosyltransferase KO variants in vitro, and then enzymatically rebuilt afucosylated to induce up to 10–50-fold improvements in ADCC activity. Removing core fucose from the Fc domain has now become standard practice for oncology antibodies that require strong effector function. Engineered adjustments to sialic acid content can also be done selectively in vitro: expression of α2,6-sialyltransferases adds an "anti-inflammatory cap" that improves serum half-life by abrogating FcγR binding, whereas expression of α2,3-sialylation has been shown to potently enhance anti-viral activity. Glycoengineering technologies can also circumvent challenges in developing biosimilar antibodies by obviating the need for comparatory bridging studies to match the glycoform profile of the originator. Finally, while therapeutic antibodies are generally glycosylated at only one site, IgM antibodies can have up to five N-linked glycosylation sites. in vitro glycosylation allows for selective remodeling away from antigen-binding sites to permit chemical functionalization for purposes like imaging or drug conjugation. Following chemoenzymatic assembly, click-modified glycans can serve as attachment points for site-specific drug conjugation with defined drug-to-antibody ratios (DAbs), solving issues with heterogeneity caused by traditional lysine or cysteine conjugation.
In addition to antibodies, other therapeutic enzymes and diagnostic proteins are also being designed using in vitro glycosylation. Enzymes used for the treatment of lysosomal storage diseases are often glycoengineered to display terminal mannose residues. Target cells will take up these glycoengineered enzymes through the mannose-6-phosphate receptor allowing for receptor mediated endocytosis. This increases therapeutic potency allowing for less frequent dosing. Vaccine antigens are being designed with hyper-sialylation linkages or defined glycan arrays that increase their immunogenicity and stability profile allowing for single dose administration. Glycoproteins used for diagnostic assays can be functionalized with clickable glycans allowing for precise attachment of imaging reporters. These molecules can then be used as cell-specific MRI contrast agents or to fluorescently track cancer cells in vivo. Lectin arrays and glycan-binding assays use uniformly glycosylated proteins as controls. Fully new glycoprotein constructs can also be made using chemoenzymatic glycosylation which uses chemical synthesis of unnatural donors and enzymatic ligation to build bispecific antibodies with altered, asymmetric glyco-signatures that simultaneously allow for high ADCC potency with prolonged serum half-life. When analytical techniques improve to allow higher throughput and sensitivity glycan profiling can be used to select clones during development or begin production with glycoengineered hosts that exclusively produce antibodies with the desired glycosylation. Increased speed and sensitivity will also allow for continual monitoring of fermentations.
When developing glycoproteins, if cell-based expression systems cannot produce glycoforms that consistently meet the product specifications due to limitations with metabolic capacity, enzyme specificity, or otherwise necessary unnatural modifications, in vitro glycosylation becomes the optimal solution. Common cases where in vitro glycosylation becomes necessary involve many trials of protein expression in cell culture resulting in a mixture of glycoforms that fall outside of predetermined acceptance criteria. Alternatively, the goal glyco-signature may need greater control than what is afforded by trafficking through the Golgi. An example of this could include consistently producing a fully afucosylated glycoprotein to increase ADCC or to hyper-sialylate a glycoprotein for anti-inflammatory response.
Applications fall into three broad categories: Therapeutic antibodies are the flagship use case, where remodeling Fc glycans allows tuning of antibody effector functions. Protein-engineering platforms can reliably install afucosylated biantennary glycans that enhance ADCC, whereas achieving consistent afucosylation in CHO cells requires cell engineering using metabolic engineering or gene knock out/knock in approaches that create clonal instability. Vaccine antigens are another prime example, where installing defined glycans onto recombinant viral proteins can improve immunogenicity/stability/cross-reactivity to the point that you can deploy single dose vaccinations instead of variable doses from cell-culture harvests. Finally, tools like click-enabled glycans installed by mutant transferases can enable antibody–drug conjugates (ADCs) with site-specific conjugation that achieves defined drug-antibody ratios (DAR) instead of the heterogeneous mixtures when coupling drugs to lysines or cysteines. Other examples include biosimilar production where matching the glycoform profile of the originator molecule within sub-% can be accomplished with in vitro rebuilding rather than heterogeneous glycoforms from cell culture that would require bridging studies. Enzyme replacement therapies for lysosomal storage disorders can be engineered to expose terminal mannose residues for improved receptor-mediated uptake and less frequent dosing. Glycoengineered research tools can include everything from glycan arrays to lectin-binding assays where uniformly glycosylated proteins are needed as controls. In every case, the glycan heterogeneity seen with native protein expression is not something you simply test at release, but rather functions as a critical quality attribute that affects the clinical performance of the drug.
Batch-release failures due to glycoform drift may suggest when a change of tactics is needed (if you find yourself continually failing specs with under-sialylated batches that clear too quickly or over-fucosylated lots yielding insufficient ADCC activity you may want to consider changing your approach). Metabolic limitations are also a good indication (UDP-Gal enrichment or CMP-Neu5Ac supplementation strategies have plateaued or you've maximized your feeding strategy but sucrose simply cannot feed enough activated precursor for the extension of tetra-antennary glycans). Targets requiring non-natural modifications are obviously a dead giveaway (needs sulfated LeX, C-linkages or your protein needs to be clickable with azido sugars?). Failures to translate micro-sparger aeration patterns from 20L to 2000L causing oxidative stress and shedding of high mannose forms due to up-regulated mannosidases can also be a sign that it's time to look at your process from a glycoengineering perspective. Glycoform-drift during cell culture development and scale up ( Again, if your continually failing due to glycoform drift) or if regulatory agencies start requiring single defined glycoforms over a mixture of glycoforms and your cell culture comparability studies start requiring fancy statistical analyses in order to prove that your process changes did not alter the product you are making.
Table 3 Indicators of Cell-Based System Insufficiency and In Vitro Solutions
| Insufficiency Indicator | Manifestation in Cell Culture | in vitro Solution |
| Glycoform drift | Batch-to-batch failure | Locked enzyme set |
| Metabolic bottleneck | Diminishing UDP-sugar returns | Synthetic donor feed |
| Exotic modification | Synthetic machinery absent | Custom donor addition |
| Scale-up stress | Oxidative high-mannose | Controlled reactor |
| Regulatory demand | Ensemble not accepted | Single-peak product |
in vitro glycosylation provides a powerful alternative to cell-based systems by enabling precise control over enzymes, substrates, and reaction conditions. When designing or optimizing cell-free glycosylation workflows, specialized glycosylation and analytical services help maximize reproducibility, efficiency, and structural accuracy.
in vitro and enzymatic glycosylation services support cell-free workflows by delivering controlled glycan attachment and remodeling using purified components. These services enable stepwise glycosylation, site-specific modification, and fine control over glycan structure without the variability introduced by cellular processing. Such approaches are particularly valuable for Fc glycan remodeling, structure-function studies, and projects where precise glycosylation outcomes are required but difficult to achieve in living systems.
Robust analytical validation is essential for evaluating the outcomes of in vitro glycosylation workflows. Glycan analysis and profiling services provide detailed characterization of glycan composition, structure, and relative abundance, enabling accurate assessment of glycosylation efficiency and homogeneity. By integrating glycan profiling with in vitro glycosylation workflows, these services help identify incomplete reactions, detect residual heterogeneity, and support data-driven optimization of cell-free glycosylation processes.
References