Introduction to UF/DF Process Development
Ultrafiltration/diafiltration (UF/DF) is the final concentration and buffer exchange step in nearly every monoclonal antibody and recombinant protein downstream process. The UF/DF step concentrates the drug substance from 1-10 g/L post-polishing to 50-250 mg/mL in the final formulation buffer, while removing residual process impurities by buffer exchange. Getting this step right determines whether the drug substance meets its target concentration, viscosity specification, and aggregate limit.
Yet UF/DF process development is often treated as an afterthought, compressed into a few days of small-scale screening. The consequence is predictable: flux collapse during concentration, excessive process times, membrane fouling that limits reuse, and protein aggregation from high shear or high local concentrations at the membrane surface. A structured development workflow, moving from membrane screening through flux excursion analysis, concentration characterization, and diafiltration optimization, eliminates these failures before they reach manufacturing.
This guide walks through the complete UF/DF development workflow with quantitative targets at each stage. It covers membrane screening with 3-5 chemistries, flux excursion experiments to identify the optimal TMP, gel point determination by semi-log flux extrapolation, viscosity management strategies for high-concentration formulations, and scale-up from 50 cm2 cassettes to manufacturing-scale systems.
Membrane Screening: Chemistry, MWCO, and Format Selection
Membrane screening is the foundation of UF/DF development, and the choice of membrane chemistry, molecular weight cut-off (MWCO), and cassette format determines the operating envelope for every subsequent stage. Screen 3-5 membrane chemistries at two MWCO values using 50 cm2 flat-sheet cassettes with 200-500 mL of representative feed material.
The two dominant UF/DF membrane chemistries for biologics are polyethersulfone (PES) and regenerated cellulose (RC). PES membranes tolerate aggressive CIP with 0.5 N NaOH and deliver higher flux due to their asymmetric pore structure, but they adsorb more protein (1-5 g/m2) than RC membranes (<0.5 g/m2). RC membranes are preferred for high-value products where yield loss to adsorption matters, but they tolerate only 0.1-0.5 N NaOH for cleaning.
| Parameter | PES (Polyethersulfone) | RC (Regenerated Cellulose) | Modified PES |
|---|---|---|---|
| Protein adsorption (g/m2) | 1-5 | <0.5 | 0.5-2 |
| Typical NWP (LMH/psi) | 2-4 | 1.5-3 | 2.5-4.5 |
| NaOH CIP tolerance | 0.5 N (30 min) | 0.1-0.5 N (15 min) | 0.5 N (30 min) |
| Typical reuse cycles | 100-300 | 50-150 | 150-300 |
| Feed channel height (mm) | 0.5-1.0 | 0.5-1.0 | 0.5-1.0 |
| Relative cost (per m2) | 1.0x | 1.2-1.5x | 1.3-1.6x |
| Best application | Standard mAb, multi-cycle reuse | High-value, adsorption-sensitive | High-concentration, low fouling |
For MWCO selection, apply the 3-5x rule: the membrane MWCO should be 3-5 times smaller than the target protein molecular weight. For mAbs (~150 kDa), a 30 kDa MWCO gives a 5:1 ratio with >99.5% retention. A 50 kDa MWCO offers 15-30% higher flux but risks 1-3% product transmission during extended diafiltration (7-10 diavolumes). For smaller proteins (30-50 kDa), drop to 10 kDa MWCO.
Evaluate each membrane with three measurements: normalized water permeability (NWP) before and after a single-pass buffer flush, protein retention by UV absorbance of the permeate, and NWP recovery after a standard CIP cycle. Rank membranes by a composite score weighting flux (40%), retention (30%), and NWP recovery (30%). Advance the top two membranes to flux excursion testing.
Flux Excursion Analysis: Finding the Optimal TMP
The flux excursion experiment is the single most important characterization in UF/DF development. It maps the relationship between transmembrane pressure (TMP) and permeate flux at fixed crossflow rate and feed concentration, revealing the three operating regimes: pressure-dependent (flux rises linearly with TMP), transitional (the knee), and mass-transfer-limited (flux plateaus regardless of TMP increase). The optimal operating TMP sits at 70-80% of the knee pressure.
Flux excursion is a stepwise TMP scan at constant crossflow. Starting from a low TMP (3-5 psi), increase pressure in 2-3 psi increments, holding each step for 5-10 minutes until flux stabilizes. Record permeate flux (LMH) at each TMP. Repeat the experiment at three feed concentrations spanning the process range: dilute (5 g/L, representing the starting pool), intermediate (25-50 g/L, the mid-concentration point), and concentrated (100-150 g/L, near the final target).
At low protein concentration, the flux-TMP curve shows a broad pressure-dependent region with a high plateau flux. As concentration increases, the plateau shifts downward and the knee moves to lower TMP values. This shift is the direct consequence of concentration polarization: the boundary layer of concentrated protein at the membrane surface generates osmotic back-pressure that opposes the applied TMP.
Setting TMP above the plateau delivers no additional flux but drives irreversible fouling by compacting the protein gel layer against the membrane surface. The penalty for over-pressurization is progressive: each cycle at excessive TMP reduces NWP by 2-5% more than operation at the optimal TMP, shortening membrane lifetime from 200+ to under 100 cycles.
For mAb UF/DF at typical feed concentrations (5-150 g/L), optimal TMP values typically fall in the 10-25 psi range for cassette-format membranes with 0.5 mm channel height. Hollow-fiber systems operate at lower TMP (5-15 psi) due to their open-channel geometry and lower shear rates.
Concentration Studies: Gel Point and Viscosity Limits
The concentration study determines how far you can push the protein concentration before flux collapses and viscosity exceeds pump limits. Concentrate the feed from its starting concentration (typically 1-10 g/L) to the highest achievable concentration in a batch ultrafiltration, recording permeate flux and retentate viscosity at regular intervals.
The stagnant film model describes the relationship between flux and concentration in the mass-transfer-limited regime:
J = k × ln(Cgel / Cb)
where J = permeate flux (LMH), k = mass transfer coefficient (LMH), Cgel = gel point concentration (g/L), Cb = bulk feed concentration (g/L)
Plot flux (y-axis) versus ln(Cb) (x-axis). In the mass-transfer-limited regime, the data fall on a straight line with slope equal to the mass transfer coefficient k and x-intercept at ln(Cgel). For typical mAbs in standard formulation buffers, Cgel ranges from 200 to 300 mg/mL. The practical maximum concentration target should be set 20-30% below Cgel to maintain workable flux and avoid the gel regime.
Viscosity is the operational ceiling that kills UF/DF processes at scale. Below 10 cP, standard peristaltic and diaphragm pumps perform well. Between 10-20 cP, feed flow rate drops and the retentate pressure (delta-P) across the cassette increases, compressing the actual TMP at the outlet end of the module. Above 20 cP, most pump systems cannot maintain the target crossflow rate, leading to flux collapse and stagnant zones in the retentate channel.
For mAbs requiring final concentrations above 150 mg/mL, viscosity mitigation is essential. Three strategies are available:
- Co-solute addition: Histidine (250-320 mM) or arginine (200-300 mM) reduces viscosity up to 10-fold at 250 mg/mL by disrupting electrostatic protein-protein interactions. This also increases Cgel by up to 100 mg/mL.
- pH adjustment: Formulating at pH values away from the mAb isoelectric point (pI) reduces self-association. Acidic formulations (pH 5.0-5.5) generally give lower viscosity than near-neutral pH for most IgG1s.
- Temperature control: Operating UF/DF at 15-25 °C rather than 2-8 °C reduces viscosity 2-3 fold, though this requires product stability data at elevated temperature during the 4-8 hour process time.
Diafiltration Optimization: Buffer Exchange Efficiency
Diafiltration removes residual process buffer and replaces it with the final formulation buffer. The efficiency of buffer exchange follows the exponential dilution equation: C/C0 = exp(-N × σ), where N is the number of diavolumes and σ is the sieving coefficient for the species being removed (σ = 1 for fully permeable small molecules). Five diavolumes remove 99.3% of a fully permeable solute; seven diavolumes remove 99.9%.
The critical decision is at what concentration to perform diafiltration. The optimal diafiltration concentration minimizes total process time (concentration + DF + final concentration) and is given by CDF,opt = Cgel/e, approximately 37% of the gel point. For a mAb with Cgel = 240 mg/mL, the optimal DF concentration is ~88 mg/mL.
| DF concentration (mg/mL) | Flux at DF (LMH) | DF buffer volume (L) | DF time (h) | Total process time (h) |
|---|---|---|---|---|
| 30 | 52 | 2,333 | 2.2 | 6.8 |
| 50 | 42 | 1,400 | 1.7 | 5.3 |
| 90 (optimal) | 30 | 778 | 1.3 | 4.5 |
| 120 | 22 | 583 | 1.3 | 4.9 |
| 150 | 15 | 467 | 1.6 | 5.8 |
Performing DF at too low a concentration wastes buffer (large retentate volume means each diavolume is large) and processing time. Performing DF at too high a concentration reduces flux dramatically, making each diavolume slow. The optimum at Cgel/e balances these two effects, yielding a broad minimum in total process time.
Monitor buffer exchange completion by measuring conductivity or UV absorbance of the permeate. For formulation buffers with conductivity significantly different from the process buffer, a conductivity target in the permeate within 5% of the formulation buffer conductivity confirms adequate exchange. For 7 diavolumes, the residual process buffer is 0.09% of its original concentration.
Membrane Fouling Control and CIP Strategy
Membrane fouling in UF/DF occurs through three mechanisms operating simultaneously: concentration polarization (reversible, cleared by crossflow), gel layer formation (partially reversible with CIP), and irreversible pore plugging from protein adsorption and aggregation. Effective fouling control manages all three through operating parameter selection and post-use cleaning.
The primary fouling control lever during operation is the TMP setpoint determined in the flux excursion. Operating in the pressure-dependent region (below the knee) maintains reversible fouling only. Above the knee, the gel layer compacts irreversibly and each process cycle deposits more protein into the membrane pores.
Post-use CIP restores membrane permeability for reuse. The standard CIP sequence for PES membranes processing mAb products follows five steps:
- Buffer flush: 10-15 L/m2 of WFI or 25 mM Tris at the process crossflow rate to displace product from the retentate volume
- NaOH recirculation: 0.5 N NaOH at 30-40 °C, recirculate for 30-60 minutes with periodic permeate-side back-pulse
- NaOH static soak: Close permeate valve, fill the system with 0.5 N NaOH, hold 30-60 minutes for diffusion-limited cleaning of pore interiors
- WFI rinse: Flush until permeate pH is within 0.5 units of WFI (typically 5-10 L/m2)
- NWP measurement: Measure NWP at standardized conditions (20 °C, 10 psi TMP). Compare to the membrane specification and the initial NWP. Alert limit: 80% of initial NWP. Replacement limit: 60% of initial NWP.
For stubborn fouling (NWP recovery <85% after standard CIP), a two-step CIP adds a 0.1 N NaOH + 150-300 ppm NaOCl step before the concentrated NaOH wash. The hypochlorite oxidizes adsorbed protein aggregates that resist alkaline hydrolysis alone. Track NWP trending over 200+ cycles to establish the reuse lifetime and replacement schedule.
UF/DF Scale-Up: From Bench to Manufacturing
UF/DF scale-up from 50 cm2 bench cassettes to manufacturing-scale systems (5-50 m2) follows three constant parameters: feed flux (L/m2/h, LMH), membrane loading (L of feed per m2 of membrane), and feed channel geometry (channel height and screen type). Maintaining these parameters preserves the mass transfer conditions and fouling behavior characterized during development.
The membrane area calculation starts from the minimum flux observed during the concentration step (Jmin, at the highest protein concentration) and the maximum allowable process time:
A = Vpermeate / (Jmin × tmax)
where A = membrane area (m2), Vpermeate = total permeate volume (L), Jmin = minimum flux (LMH), tmax = max process time (h)
The total permeate volume includes the UF concentration permeate (Vfeed - Vretentate) plus the DF permeate (NDV × Vretentate at DF) plus the final concentration permeate if applicable. Size the area with a 20-30% safety margin to account for membrane-to-membrane variability and fouling-related flux decline over the reuse lifetime.
| Parameter | Bench (50 cm2) | Pilot (0.5 m2) | Manufacturing (20 m2) |
|---|---|---|---|
| Membrane area | 50 cm2 | 0.5 m2 | 20 m2 |
| Feed volume | 200-500 mL | 20-50 L | 2,000 L |
| Feed flux (constant) | 300 LMH | 300 LMH | 300 LMH |
| TMP (constant) | 15 psi | 15 psi | 15 psi* |
| Channel height | 0.5 mm (C-screen) | 0.5 mm (C-screen) | 0.5 mm (C-screen) |
| Path length | 17 cm | 17 cm | 17-115 cm** |
| Cassettes per holder | 1 | 5-10 | 15-42 |
| Retentate hold-up (%) | 30-50% | 10-20% | 3-8% |
*At manufacturing scale, longer path lengths increase the pressure drop across the cassette (delta-P), which creates a TMP gradient from inlet to outlet. The inlet TMP may exceed the optimal value even when the average TMP is on target. For path lengths above 30 cm, measure inlet and outlet pressures independently and verify that the inlet TMP stays below the flux excursion knee pressure.
**Longer path lengths (e.g., Pellicon 3 at 115 cm vs. Pellicon 2 at 20 cm) reduce the number of cassettes per holder but increase delta-P. At high protein concentrations where viscosity rises, the delta-P penalty grows further. Compensate by reducing feed flux 10-20% to keep inlet TMP within the optimal range.
What Is the Optimal TMP for a UF/DF Process?
The optimal TMP for a UF/DF process is the transmembrane pressure that delivers 90-95% of the maximum achievable (plateau) flux while operating entirely within the pressure-dependent or early transition regime. In practice, this is 70-80% of the TMP at which the flux-TMP curve transitions from linear to flat. For most mAb UF/DF processes on 30 kDa PES membranes, the optimal TMP falls in the range of 10-20 psi at the starting concentration and decreases to 8-15 psi as the protein concentration increases during the batch.
There are two philosophies for managing TMP during UF/DF concentration: constant-TMP and constant-flux (permeate control). In constant-TMP operation, the TMP is fixed at the value optimized for the mid-point concentration, and flux drops naturally as the protein concentrates. This is simpler to implement and is the standard approach for most mAb processes. In constant-flux operation (permeate control), TMP is increased automatically to maintain a target permeate flow rate. This maximizes throughput early in the concentration but risks over-pressurization as the system approaches the plateau at high concentration. Constant-flux operation requires real-time TMP monitoring with an upper limit interlock to prevent irreversible fouling.
For high-concentration formulations above 150 mg/mL, a hybrid approach works best: constant TMP during the initial concentration (5 to 50 mg/mL), then switch to constant flux for the intermediate phase (50-100 mg/mL), and finally reduce both flux and TMP targets for the final concentration above 100 mg/mL where viscosity and fouling accelerate.
Worked Example: 2,000 L mAb UF/DF Process Design
Worked Example: Sizing a mAb UF/DF Step
Given: Post-polish pool: 2,000 L at 5 g/L mAb (10 kg total). Target: 150 mg/mL in 20 mM histidine, 150 mM NaCl, pH 6.0. Diafiltration: 7 diavolumes. Membrane: 30 kDa PES, C-screen. From flux excursion: Jmin at 150 mg/mL = 15 LMH. Cgel = 240 mg/mL.
Step 1: Optimal DF concentration
CDF = Cgel/e = 240/2.718 = 88 mg/mL
Round to 90 mg/mL for operational simplicity.
Step 2: Volumes at each stage
- UF1 retentate volume: 10,000 g / 90 g/L = 111 L
- UF1 permeate volume: 2,000 - 111 = 1,889 L
- DF permeate volume: 7 × 111 = 778 L
- UF2 retentate volume: 10,000 g / 150 g/L = 66.7 L
- UF2 permeate volume: 111 - 66.7 = 44.3 L
- Total permeate: 1,889 + 778 + 44.3 = 2,711 L
Step 3: Membrane area sizing
A = Vtotal permeate / (Jmin × tmax)
For a 6-hour maximum process time and Jmin of 15 LMH:
A = 2,711 / (15 × 6) = 30.1 m2
With 25% safety factor: A = 30.1 × 1.25 = 37.6 m2. Use 38 m2 (e.g., 76 × 0.5 m2 Pellicon 3 cassettes or equivalent).
Step 4: Process time estimate
- UF1 (5 → 90 mg/mL): avg flux ~35 LMH, time = 1,889/(35 × 38) = 1.4 h
- DF (7 DV at 90 mg/mL): flux ~30 LMH, time = 778/(30 × 38) = 0.7 h
- UF2 (90 → 150 mg/mL): avg flux ~18 LMH, time = 44.3/(18 × 38) = 0.06 h
- Total: ~2.2 h process time (well within 6 h limit)
Step 5: Buffer preparation
DF buffer needed: 778 L × 1.1 (10% overage) = 856 L
Filtration Calculator
Size TFF membranes, calculate diafiltration volumes, and estimate flux for UF/DF process development.
Buffer Calculator
Calculate formulation buffer recipes for your diafiltration step, including pH adjustment and ionic strength.
Frequently Asked Questions
What is a flux excursion in UF/DF?
A flux excursion (also called a TMP scan or flux-TMP experiment) is a systematic characterization where transmembrane pressure is increased stepwise while measuring permeate flux at a fixed feed concentration and crossflow rate. The resulting curve reveals the pressure-dependent region, the transition zone (knee), and the mass-transfer-limited plateau. The optimal operating TMP is set at 70-80% of the plateau onset pressure.
How do you determine the gel point concentration for a protein in UF/DF?
Plot permeate flux on the y-axis versus the natural log of retentate protein concentration on the x-axis during a batch concentration experiment. In the mass-transfer-limited regime, the data form a straight line. Extrapolate to the x-intercept where flux equals zero. That x-intercept gives ln(Cgel). For typical mAbs, Cgel ranges from 200 to 300 mg/mL depending on buffer composition, pH, and mAb self-association properties.
What membrane MWCO should I use for mAb UF/DF?
For mAbs (~150 kDa), use a 30 kDa MWCO membrane, giving a 5:1 size ratio with >99.5% retention. A 50 kDa MWCO offers higher flux but risks 1-3% mAb transmission during extended diafiltration. For proteins under 50 kDa, use 10 kDa MWCO. The 3-5x rule (MWCO should be 3-5 times smaller than the target protein MW) provides near-complete retention while avoiding excessive flux limitation from oversized pores.
How much membrane area do I need for a UF/DF process?
Calculate from A = Vpermeate / (Jmin × tmax), where V is total permeate (concentration + diafiltration + final concentration), Jmin is the minimum flux at peak concentration (typically 10-20 LMH for mAbs at 150-200 mg/mL), and tmax is the maximum process time (typically 4-8 hours). Add a 20-30% safety factor. For a 2,000 L mAb batch at 5 g/L concentrated to 150 mg/mL with 7 DV, expect 15-40 m2.
Why does flux drop during UF/DF concentration and how can I prevent it?
Flux drops because increasing protein at the membrane surface raises osmotic pressure and viscosity. The stagnant film model (J = k × ln(Cgel/Cb)) shows flux decreases logarithmically with bulk concentration. Mitigation: increase crossflow rate (raises k), diafilter at the optimal concentration (Cgel/e, ~37% of gel point), add co-solutes (histidine 250-320 mM raises Cgel by up to 100 mg/mL and reduces viscosity up to 10-fold), or switch to a wider-channel cassette to reduce delta-P at high viscosity.
Related Tools
- Filtration Calculator — Size TFF membranes, calculate diavolumes, and estimate process times for UF/DF operations
- Buffer Calculator — Prepare formulation buffers for diafiltration with precise pH and ionic strength targeting
- Chromatography Calculator — Size the upstream polish steps that feed into your UF/DF process
References
- van Reis R. & Zydney A. (2007). Bioprocess membrane technology. Journal of Membrane Science, 297(1-2), 16-50. doi:10.1016/j.memsci.2007.02.045
- Baek Y. et al. (2017). Ultrafiltration behavior of monoclonal antibodies and Fc-fusion proteins: Effects of physical properties. Biotechnology and Bioengineering, 114(10), 2057-2065. doi:10.1002/bit.26326
- Hung J.J. et al. (2016). High concentration tangential flow ultrafiltration of stable monoclonal antibody solutions with low viscosities. Journal of Membrane Science, 508, 113-126. doi:10.1016/j.memsci.2016.02.031
- Cunha F. et al. (2024). Intramodule pressure profiles and protein accumulation during tangential flow filtration. Biotechnology Progress, 40(1), e3389. doi:10.1002/btpr.3389
- Yu H. et al. (2025). Process optimization mitigated the retention loss of an Fc-fusion protein during ultrafiltration/diafiltration. Biotechnology Progress, 41(2), e70021. doi:10.1002/btpr.70021