Inclusion body expression in E. coli remains the dominant production route for non-glycosylated proteins, Fab fragments, and industrial enzymes. The bottleneck is not expression. It is refolding. Unoptimized refolding protocols typically recover 15-30% of solubilized protein as native, active product. The rest is lost to aggregation, misfolding, and incomplete disulfide formation. Yet many labs still optimize refolding buffers one factor at a time, missing the interactions between pH, redox conditions, and additives that account for 30-50% of yield variance.
This article shows you how to use design of experiments (DOE) to systematically optimize refolding conditions. We compare dilution, dialysis, and on-column refolding methods head-to-head with quantitative performance data, present a decision tree for method selection, and walk through a complete worked example: an 8-factor definitive screening design that pushed a Fab fragment from 28% to 72% soluble yield in 17 experiments.
Why Refolding Needs DOE, Not OFAT
Refolding buffer optimization is a textbook DOE problem because the factors interact strongly with each other. Changing pH from 8.0 to 9.0 has a different effect on yield depending on whether arginine is present at 0 M or 0.6 M. A one-factor-at-a-time (OFAT) approach misses these interactions entirely and typically finds a local optimum 20-40% below the true best conditions.
Published DOE studies on refolding consistently show that:
- Two-factor interactions contribute 30-50% of explained variance in soluble yield
- The pH-by-arginine interaction is the single largest interaction effect in most screens
- The OFAT-optimized buffer underperforms the DOE-optimized buffer by 1.5-2.5x in head-to-head tests
- A screening DOE followed by response surface optimization requires 30-35 total experiments, versus 100+ for systematic OFAT
The refolding buffer has 6-10 adjustable factors (pH, protein concentration, arginine, redox pair ratio, total thiol, salt, osmolyte, detergent, temperature, denaturant residual). Even a modest DOE screen captures the critical main effects and interactions in under 20 runs.
Dilution vs Dialysis vs On-Column: Head-to-Head Comparison
The three principal refolding methods differ fundamentally in how they remove denaturant and at what protein concentration the refolding transition occurs. No single method is universally best. The right choice depends on the protein's properties, the required scale, and whether purification can be integrated.
| Parameter | Dilution | Dialysis | On-Column (IMAC/IEX) |
|---|---|---|---|
| Mechanism | Rapid denaturant removal by volume increase | Gradual denaturant removal across membrane | Linear denaturant gradient while bound to resin |
| Typical soluble yield | 25-45% | 30-50% | 35-60% |
| Protein conc. during refolding | 0.01-0.5 mg/mL | 0.5-5 mg/mL | 1-10 mg/mL (on resin) |
| Buffer volume (L per g protein) | 50-100 | 20-40 | 2-5 |
| Refolding time | 2-24 h | 24-72 h | 2-6 h |
| Scale limit (practical) | Up to 10,000 L tank | Up to 50 L (membrane area) | Up to 5 L resin bed (column hardware) |
| Aggregate content | 15-30% | 10-25% | 5-15% |
| Purification integration | None (separate capture step) | None | Refolding + capture in one step |
| Key advantage | Simple, scalable | Controlled rate of denaturant removal | High conc., low volume, purity gain |
| Key limitation | Enormous buffer volumes | Slow, limited scale | Requires affinity tag or charge handle |
Dilution refolding is the simplest: solubilized inclusion bodies are diluted 10-100x into refolding buffer, dropping the denaturant concentration below the threshold for protein folding. The critical limitation is that the protein concentration must stay below 0.1-0.5 mg/mL to avoid aggregation during the folding transition. At manufacturing scale, this means very large refolding tanks.
Dialysis refolding removes denaturant gradually across a dialysis membrane or by tangential-flow diafiltration. The controlled rate of denaturant removal allows the protein to traverse the folding landscape slowly, which helps for proteins with complex folding pathways or multiple disulfide bonds. However, dialysis is slow (24-72 h for complete buffer exchange) and scales poorly.
On-column refolding immobilizes the denatured protein on a chromatography resin (IMAC for His-tagged proteins, IEX for charged proteins) and applies a decreasing denaturant gradient. The protein refolds while physically separated on the resin surface, suppressing intermolecular aggregation. This method combines refolding with a purification step and uses 10-50x less buffer than dilution.
Bar chart: On-column refolding achieves 48% mean soluble yield vs 35% for dilution and 40% for dialysis, with only 10% aggregates vs 22% for dilution.
How to Choose the Right Refolding Method
The best refolding method depends on four protein characteristics: disulfide bond count, affinity tag availability, aggregation propensity, and production scale. The decision tree below routes you to the method most likely to give the highest yield for your specific case.
Decision tree diagram. His-tagged proteins route to on-column IMAC refolding. Non-tagged proteins at large scale route to pulse or batch dilution. Small-scale proteins with disulfide bonds route to dialysis; without disulfides, to dilution.
The decision tree prioritizes on-column refolding for His-tagged proteins because the method integrates refolding with capture chromatography, reducing process steps and buffer consumption by 10-50x. If the protein lacks a suitable affinity tag, scale and disulfide complexity determine the choice between dilution and dialysis.
The 8 Factors That Control Refolding Yield
Refolding buffer optimization involves 6-10 adjustable parameters. The following 8 factors consistently appear as significant across published refolding DOE studies and are the recommended starting set for a screening design.
| Factor | Low Level | High Level | Rationale |
|---|---|---|---|
| pH | 7.5 | 9.5 | Solubility and charge state of folding intermediates; alkaline pH favors disulfide formation |
| Protein concentration (mg/mL) | 0.05 | 0.5 | Higher concentration increases throughput but raises aggregation risk |
| L-arginine (M) | 0 | 0.8 | Suppresses aggregation by binding to hydrophobic patches on folding intermediates |
| GSH:GSSG ratio | 1:1 | 10:1 | Controls disulfide shuffling rate; ratio determines redox potential |
| Total thiol concentration (mM) | 1 | 5 | Total GSH+GSSG sets the thiol exchange capacity available for disulfide formation |
| NaCl (M) | 0 | 0.5 | Ionic strength screens electrostatic interactions; too high can salt-out intermediates |
| Glycerol (% v/v) | 0 | 15 | Osmolyte that stabilizes compact folding intermediates via preferential exclusion |
| Temperature (°C) | 4 | 20 | Lower temperature slows aggregation more than folding, improving yield for aggregation-prone proteins |
L-arginine is the single most impactful additive in refolding buffer optimization. At 0.4-0.8 M, arginine suppresses aggregation by interacting with exposed hydrophobic surfaces on partially folded intermediates, increasing their apparent solubility without destabilizing the native state. This distinguishes it from chaotropes like urea, which improve solubility at the cost of native-state stability.
The GSH:GSSG redox pair is essential for any protein with disulfide bonds. The ratio sets the effective redox potential of the refolding buffer, and the total thiol concentration determines the rate of disulfide exchange. A ratio of 5:1 (GSH:GSSG) with 2-3 mM total thiol is a reasonable starting point for most Fab fragments and disulfide-containing enzymes.
For proteins without disulfide bonds, the redox factors can be dropped, reducing the design to 5-6 factors. A definitive screening design with 6 factors needs only 13 runs.
Setting Up a Definitive Screening Design for Refolding
A definitive screening design (DSD) is the optimal choice for refolding buffer optimization because it estimates main effects cleanly, detects curvature (nonlinear responses), and captures some two-factor interactions in 2k+1 runs. For 8 factors, that is 17 experiments, compared to 256 for a full factorial and 12 for a Plackett-Burman design (which confounds two-factor interactions with main effects).
The DSD workflow for refolding follows three stages:
- Screen (17 runs) — Run the DSD across 8 factors. Measure soluble yield by SEC-HPLC and optionally bioactivity. Identify the 3-4 statistically significant factors from the Pareto chart.
- Optimize (15 runs) — Run a Box-Behnken or central composite design on the 3-4 significant factors. Build a response surface model and find the predicted optimum.
- Confirm (3-5 runs) — Run confirmation experiments at the predicted optimum. Verify that the model predicts within its confidence interval.
The response variable should be % soluble monomer by SEC-HPLC (monomeric peak area / total peak area). This is more reliable than total soluble protein by BCA or Bradford, which includes soluble aggregates. If bioactivity is the ultimate criterion, measure both SEC monomer yield and specific activity, and use the desirability function to optimize both responses simultaneously.
Design Your Refolding Screen
Set up a definitive screening design for 6-8 refolding buffer factors. Enter your factor ranges, generate the run table, and analyze results with built-in Pareto charts. DOE Pro adds QbD reporting and D-optimal designs for irregular factor spaces.
Worked Example: Fab Fragment Refolding DSD
This worked example demonstrates the full DOE workflow for optimizing the refolding of a 25 kDa Fab fragment with 2 disulfide bonds, expressed in E. coli BL21(DE3) as inclusion bodies at 3 g/L wet cell weight.
Worked Example: Fab Refolding DSD
Protein: Anti-TNF-alpha Fab fragment, 25 kDa, 2 intramolecular disulfide bonds (VH-CH1 interface + light chain CL-CL'), His6-tagged at C-terminus.
Baseline: Standard refolding in 50 mM Tris pH 8.0, 0.1 mg/mL protein, 1 mM GSH, 0.2 mM GSSG, 4 °C. Soluble yield by SEC-HPLC: 28 ± 3%.
Design: Definitive Screening Design, 8 factors, 17 runs (including 4 center points). Factors and levels per Table 2. Response: % soluble monomer by SEC-HPLC (280 nm, Superdex 200 Increase 10/300 GL).
Analysis: Forward stepwise regression with Bonferroni-corrected significance threshold (alpha = 0.05/8 = 0.00625). Results in Table 3 below.
| Factor | Coefficient | Standard Error | p-value | Significant? |
|---|---|---|---|---|
| L-arginine (M) | +18.2 | 2.1 | < 0.001 | Yes |
| pH | +12.7 | 2.1 | < 0.001 | Yes |
| GSH:GSSG ratio | +9.4 | 2.1 | 0.002 | Yes |
| Temperature (°C) | -6.8 | 2.1 | 0.011 | Marginal |
| Protein conc. (mg/mL) | -5.3 | 2.1 | 0.032 | No (Bonferroni) |
| Total thiol (mM) | +3.1 | 2.1 | 0.18 | No |
| NaCl (M) | -1.9 | 2.1 | 0.39 | No |
| Glycerol (% v/v) | +1.2 | 2.1 | 0.58 | No |
Pareto chart. L-arginine has the largest positive effect (+18.2), followed by pH (+12.7) and GSH:GSSG ratio (+9.4). Temperature has a negative effect (-6.8). NaCl and glycerol are not significant.
Optimized Conditions
Based on the DSD screening results, the optimized refolding buffer for this Fab fragment is:
- Buffer: 50 mM Tris-HCl, pH 8.5
- L-arginine: 0.55 M
- Redox: 5:1 GSH:GSSG, 2 mM total thiol (1.67 mM GSH + 0.33 mM GSSG)
- Protein concentration: 0.1 mg/mL
- NaCl: 0.15 M
- Glycerol: 5% v/v
- Temperature: 10 °C
- Refolding time: 16 h (overnight)
Soluble yield (SEC): 72 ± 4% (n=3 confirmation runs)
Baseline yield: 28 ± 3%
Improvement: 2.6-fold
The three significant factors, L-arginine, pH, and GSH:GSSG ratio, account for 89% of the model's explained variance. This is consistent with published refolding DOE studies, where arginine and pH are almost always the top two factors. The negative temperature coefficient means that lower temperatures favor yield for this Fab, likely because aggregation kinetics are more temperature-sensitive than folding kinetics for this specific protein.
Scale-Up Considerations for Each Refolding Method
Refolding scale-up is the step where method choice has the largest practical impact. The buffer volume, mixing requirements, and process time all scale differently for dilution, dialysis, and on-column refolding.
Dilution refolding at scale
For batch dilution at manufacturing scale, the volume requirement can be calculated directly. If 1 L of solubilized IBs contains 10 g protein at 10 mg/mL, and the target refolding concentration is 0.1 mg/mL, the dilution factor is 100x. The refolding tank must hold 100 L of buffer per liter of solubilized IBs. For a 50 L fermentation (yielding roughly 5 L of solubilized IBs), batch dilution requires a 500 L refolding vessel.
Pulse (fed-batch) dilution reduces this by adding the denatured protein in small aliquots (typically 5-10% of the refolding volume) every 30-60 minutes over 4-8 hours. Each pulse refolds before the next addition, keeping the instantaneous concentration of unfolded protein low while accumulating refolded product in the tank. This reduces peak tank volume by 40-60% compared to batch dilution and is the standard approach for dilution refolding above 10 L scale.
On-column refolding at scale
On-column refolding scales by increasing column bed volume. A 5 cm diameter column with 20 cm bed height (400 mL) loaded at 20 mg protein per mL resin handles 8 g of protein per run. At manufacturing scale, the same process uses standard preparative IMAC or IEX columns. The critical scale-up parameter is the linear flow rate during the denaturant gradient, which must allow enough residence time for the protein to refold (typically 30-60 cm/h for a 2-4 h gradient).
Dialysis refolding at scale
Dialysis scales the least well. Membrane dialysis cassettes are limited to about 50 mL per cassette, and tangential-flow diafiltration with a 10 kDa MWCO membrane can handle larger volumes but requires 24-48 h for complete buffer exchange. For proteins that specifically require the controlled denaturant removal rate that dialysis provides, membrane-assisted refolding using hollow-fiber TFF modules at constant volume diafiltration is the best scale-up path.
Generate a Starting Refolding Buffer
Enter your protein's molecular weight, disulfide bond count, and expression system. The Refolding Generator creates a starting buffer recipe based on published protocols, giving you a baseline to improve with DOE.
How to Set Up a Refolding Screen in the DOE Generator
The BioProcess Tools DOE Generator handles the complete workflow for designing a refolding buffer screen. Here is the step-by-step setup for the 8-factor DSD used in the worked example.
- Select design type: Choose Definitive Screening Design from the design menu. For 8 continuous factors, the tool generates a 17-run design with center points.
- Enter factors: Add each factor from Table 2 with its name, low level, and high level. Use the continuous factor type for all 8 factors. The GSH:GSSG ratio can be entered as a single continuous variable (1 to 10, representing the ratio of GSH to GSSG).
- Add responses: Enter "Soluble monomer yield (%)" as the primary response. Optionally add "Specific activity (U/mg)" as a secondary response if you plan to use desirability optimization.
- Generate and randomize: The tool produces the 17-run table with randomized run order. Export to CSV or follow the on-screen layout.
- Enter results: After running the experiments, enter the measured yield for each run. The built-in analysis produces the Pareto chart, coefficient table, and predicted optimum.
For the optimization phase, switch to a Box-Behnken design with the 3 significant factors identified in the screening phase. The DOE Generator supports this transition with a "Follow-up design" option. DOE Pro adds D-optimal designs for irregular factor spaces and generates a QbD-compliant report with design space boundaries.
Optimize Upstream Expression First
Refolding yield depends on inclusion body quality. Use the E. coli Expression Optimizer to find the best IPTG concentration, induction temperature, and growth conditions before starting your refolding screen.
Frequently Asked Questions
What is the best refolding method for proteins with multiple disulfide bonds?
For proteins with 2 or more disulfide bonds, on-column refolding on IMAC resin or dialysis with controlled redox gradients give the highest native yields. Both methods allow gradual removal of denaturant while maintaining the GSH:GSSG ratio needed for correct disulfide shuffling. Dilution refolding can also work but requires careful redox control and very low protein concentrations (below 0.1 mg/mL) to avoid intermolecular disulfide scrambling.
How many experiments do I need for a refolding buffer DOE screen?
A definitive screening design (DSD) with 8 factors requires only 17 runs (2k+1, including center points), compared to 256 runs for a full factorial at 2 levels. If you have fewer factors (5-6), a Plackett-Burman design needs 12 runs. For the optimization phase after screening, a Box-Behnken design on 3 significant factors adds 15 runs. The total two-stage DOE workflow typically requires 30-35 experiments to go from an unknown buffer to an optimized formulation.
Why does L-arginine improve protein refolding yield?
L-arginine at 0.4-0.8 M suppresses protein aggregation during refolding by binding to exposed hydrophobic patches on partially folded intermediates, increasing their solubility without stabilizing misfolded states. Unlike other aggregation suppressors such as urea or GdnHCl, arginine does not significantly destabilize the native state at these concentrations. It acts as a chemical chaperone that widens the kinetic window for correct folding by slowing the competing aggregation pathway.
Can I refold inclusion bodies without a DOE screen?
Yes, but yields are typically 15-30% without optimization versus 50-80% with a systematic DOE approach. A one-factor-at-a-time (OFAT) screen misses interactions between buffer components, which account for 30-50% of the variance in refolding yield. At minimum, use a published starting recipe or the Refolding Generator to get a baseline buffer, then run a screening DOE to identify which factors matter most for your specific protein.
What protein concentration should I use for dilution refolding?
Keep the final protein concentration below 0.1-0.5 mg/mL for dilution refolding to minimize aggregation. The critical parameter is the concentration of unfolded protein in the refolding buffer, not the starting denatured stock. For aggregation-prone proteins, stay below 0.05 mg/mL. Pulse (fed-batch) dilution adds denatured protein in small boluses over 4-8 hours, allowing each aliquot to refold before the next addition and reducing peak volume by 40-60%.
How do I set up a refolding screen in the BioProcess Tools DOE Generator?
Open the DOE Generator, select Definitive Screening Design, and enter 6-8 continuous factors with their low and high levels (for example, pH 7.5-9.5, L-arginine 0-0.8 M, GSH:GSSG ratio 1:1 to 10:1). Set your response as soluble yield measured by SEC-HPLC or activity assay. The tool generates the run table with randomized factor combinations. After entering your experimental results, use the built-in analysis to identify the Pareto chart of significant effects and the predicted optimum.
Related Tools
- DOE Generator — Design and analyze experiments for bioprocess optimization, including refolding buffer screens
- Refolding Generator — Generate optimized refolding buffer recipes from protein properties and expression system
- E. coli Expression Optimizer — Optimize IPTG concentration, induction temperature, and growth conditions for inclusion body expression
References
- Yamaguchi H, Miyazaki M. Refolding techniques for recovering biologically active recombinant proteins from inclusion bodies. Biomolecules. 2014;4(1):235-251. doi:10.3390/biom4010235
- Singh A, Upadhyay V, Upadhyay AK, Singh SM, Panda AK. Protein recovery from inclusion bodies of Escherichia coli using mild solubilization process. Microb Cell Fact. 2015;14:41. doi:10.1186/s12934-015-0222-8
- Vincentelli R, Canaan S, Campanacci V, et al. High-throughput automated refolding screening of inclusion bodies. Protein Sci. 2004;13(10):2782-2792. doi:10.1110/ps.04806004
- Mannall GJ, Titchener-Hooker NJ, Dalby PA. Factors affecting protein refolding yields in a fed-batch and batch-refolding system. Biotechnol Bioeng. 2007;97(6):1523-1534. doi:10.1002/bit.21377
- Schmoeger E, Berger E, Trefilov A, Jungbauer A, Hahn R. Matrix-assisted refolding of autoprotease fusion proteins on an ion exchange column. J Chromatogr A. 2009;1216(50):8460-8469. doi:10.1016/j.chroma.2009.10.012