Tangential Flow Filtration (TFF): diafiltration

Tangential flow filtration (TFF) is well established as a method for concentrating biological samples, but concentration is only one part of what TFF systems can do. Diafiltration, the process of adding fresh buffer to a sample while simultaneously removing permeate, extends the technique into a far more powerful tool: one that can exchange buffers, remove small-molecule contaminants, and reduce the conductivity of a preparation to whatever level downstream applications require. For researchers working with extracellular vesicles (EVs), recombinant proteins, viral vectors, or any other biotherapy where formulation matters, understanding diafiltration in detail is practically useful from the first day it is applied.

This article covers the theoretical basis of diafiltration, the key parameters that determine its efficiency, the practical decisions researchers face when designing a diafiltration protocol, and how these principles apply specifically to EV and exosome processing.

What Diafiltration Actually Does

To understand diafiltration it helps to start with what TFF alone achieves. In a standard TFF concentration run, the sample is recirculated across a membrane under pressure. A fraction of the liquid, along with molecules smaller than the membrane's molecular weight cut-off (MWCO), passes through as permeate. The retained material, the retentate, becomes progressively more concentrated. At the end of the run, the volume is smaller but the composition of the retained species, relative to one another, is unchanged. Small molecules that do not pass the membrane are still present. Salts, buffer components, and process-related impurities that fall below the MWCO are removed along with the permeate water, but only in proportion to the volume reduction achieved.

Diafiltration changes this by replenishing the volume continuously or in steps. As permeate is removed, an equivalent volume of fresh diafiltration buffer is added to the retentate vessel. The retained macromolecules remain in place while each addition of fresh buffer progressively dilutes the small-molecule content of the retentate. After sufficient volume exchanges, the small-molecule composition of the retentate approaches that of the diafiltration buffer. Buffer exchange, contaminant removal, and conductivity adjustment all follow from this single principle.

The Mathematics: Diavolumes and Predicted Removal

The core metric in diafiltration is the diavolume (DV): one diavolume equals one retentate volume of fresh buffer added and an equivalent volume removed as permeate. The relationship between diavolumes applied and contaminant removal is well described by a simple exponential function, assuming ideal behaviour (complete mixing, 100% transmission of the contaminant, no binding to the membrane):

C / C₀ = e^−N

Where C is the contaminant concentration after diafiltration, C₀ is the starting concentration, and N is the number of diavolumes applied. This gives the following expected removal at each stage:

In practice, five to seven diavolumes is sufficient for most buffer exchange applications, delivering greater than 99% removal of freely transmitting species. Ten or more diavolumes is reserved for situations where residual contaminants must be reduced to trace levels, for example when removing dimethyl sulphoxide (DMSO) from cryopreserved EV preparations, or when clearing endotoxin from a formulation buffer.

The formula assumes ideal conditions. In real systems, incomplete mixing in the retentate vessel, partial sieving of the contaminant by the membrane, and concentration polarisation effects all reduce efficiency relative to the theoretical prediction. Understanding these deviations is part of developing a robust diafiltration protocol.

Continuous vs. Discontinuous Diafiltration

There are two practical implementations of diafiltration. In continuous (constant-volume) diafiltration, fresh buffer is added at the same rate that permeate is removed. The retentate volume stays constant throughout. This is the most common approach for buffer exchange applications because it maintains a consistent hydrodynamic environment at the membrane surface and keeps the concentration of the retained product relatively stable.

In discontinuous (sequential dilution) diafiltration, the sample is diluted with fresh buffer and then reconcentrated, and this cycle is repeated. This approach can be preferable when the product is sensitive to prolonged shear at the membrane surface, or when the available system does not easily support precise flow balancing. It requires greater operator attention but can be gentler on fragile particles.

For EV applications, continuous diafiltration is generally preferred. EVs are large relative to standard TFF membranes (typically 100 nm to 200 nm diameter versus membrane MWCOs in the 100 kDa to 750 kDa range), so they are retained efficiently. Continuous diafiltration minimises the number of times the EV preparation experiences concentration spikes, which can cause aggregation.

Membrane Selection for Diafiltration

The membrane is the most consequential single decision in a diafiltration protocol. Two properties matter most: MWCO and membrane chemistry.

MWCO determines what passes into the permeate. For EV diafiltration, the target is complete retention of vesicles with efficient removal of free proteins, nucleic acids, and small molecules. MWCOs of 100 kDa to 300 kDa are commonly used. A 100 kDa membrane retains proteins above approximately 100 kDa alongside the EVs; a 300 kDa membrane allows more protein through but also risks losing smaller vesicles or dense protein complexes. For most standard EV preparations, 300 kDa is a reasonable starting point. For preparations where protein co-purification is a specific concern, 100 kDa offers greater selectivity at the cost of slower flux.

Membrane chemistry affects non-specific binding and fouling behaviour. Polyethersulphone (PES) and modified PES membranes are most commonly used for EV applications because of their relatively low protein binding and reasonable flux characteristics. Regenerated cellulose membranes offer even lower binding but are mechanically less robust and more sensitive to certain cleaning agents. Hollow fibre geometries, in which the sample flows through the lumen of multiple parallel capillaries, are generally preferred over flat sheet cassettes for EV work because shear stress is more evenly distributed and there is no stagnant dead volume at channel edges.

Key Operating Parameters

Diafiltration performance is determined by three interconnected parameters: transmembrane pressure (TMP), cross-flow velocity (CFV), and the concentration of the retentate. Getting these right is the difference between a clean, reproducible process and one that fouls membranes, loses product, and delivers variable results.

TMP is the net pressure driving liquid through the membrane. Higher TMP increases flux up to a point, after which the membrane enters a pressure-independent regime where flux is limited by concentration polarisation rather than the driving pressure. Operating in this regime is wasteful of energy and accelerates fouling. For most EV diafiltration applications, TMP values between 0.5 and 1.5 bar represent the practical operating window, but the optimal value depends on the specific membrane and feedstock.

CFV is the velocity at which the sample flows tangentially across the membrane surface. High CFV sweeps away the concentration polarisation layer that forms as retained material accumulates near the membrane, maintaining flux and reducing fouling. For hollow fibre cartridges, CFV is controlled by the feed pump flow rate and cartridge internal diameter. Most hollow fibre EV diafiltration protocols operate at CFV values of 10 to 30 cm per second.

Product concentration in the retentate affects both flux and particle behaviour. As the retentate concentrates, viscosity increases and particle-particle interactions become more likely. For EVs, this can manifest as aggregation and irreversible membrane fouling. A practical guideline is to avoid concentrating EV preparations beyond approximately 10 to 50-fold during a single pass, and to perform diafiltration at a moderate concentration rather than at the maximum achievable concentration.

Diafiltration for Extracellular Vesicles: Specific Considerations

EVs present several specific challenges that do not arise in conventional protein diafiltration. First, the starting material is typically a conditioned cell culture medium or biological fluid such as plasma or urine, containing an enormous diversity of co-contaminants spanning six or more orders of magnitude in size. Lipoproteins, protein aggregates, and cell debris all behave differently under diafiltration conditions and some may foul the membrane more aggressively than the EVs themselves.

Second, EVs are heterogeneous in size, density, and surface chemistry. Smaller EVs in the 30 nm to 70 nm range approach the size exclusion limit of higher-MWCO membranes and may be lost in the permeate to a degree that is not always obvious from particle count measurements alone. Characterising the EV size distribution before and after diafiltration, for example by nanoparticle tracking analysis (NTA), is good practice when optimising a new protocol.

Third, the surface membrane of EVs contains lipid bilayer components, transmembrane proteins, and a corona of associated proteins that can interact with membrane surfaces. These interactions are generally reversible under physiological buffer conditions but become problematic at extreme pH, high ionic strength, or in the presence of organic solvents. Diafiltration buffers for EVs should generally be physiologically balanced: phosphate-buffered saline (PBS), HEPES-buffered saline, or equivalent formulations are standard choices.

A typical EV diafiltration workflow, combined with upstream concentration, follows this sequence:

Downstream Coupling: Why Diafiltration Format Matters for SEC and Other Polishing Steps

In many EV and biotherapy workflows, TFF diafiltration is not the final purification step. It is a preparation step for a downstream polishing technique such as size exclusion chromatography (SEC), density gradient ultracentrifugation, or ion exchange chromatography. In each case, the condition in which the sample arrives from diafiltration has a direct bearing on the efficiency of the downstream step.

For SEC, which is increasingly used to separate EVs from co-purified proteins and lipoprotein particles, the diafiltration buffer must be compatible with the SEC running buffer. Conductivity mismatches between sample and column buffer cause peak broadening and can reduce resolution. Using the SEC running buffer as the diafiltration buffer is a practical way to ensure compatibility and avoid an additional manual buffer adjustment step.

For ion exchange, sample conductivity matters critically. Ion exchange resins bind charged species only when the ionic strength is low enough that electrostatic interactions are not screened. TFF diafiltration is an efficient way to reduce conductivity to the sub-5 mS/cm range that most ion exchange protocols require, without the dilution penalty of dialysis or the sample loss associated with desalting spin columns.

Common Diafiltration Problems and How to Approach Them

Flux decline during diafiltration is the most commonly encountered problem. If flux drops progressively during a diafiltration run rather than remaining roughly constant, the most likely cause is membrane fouling by accumulated retentate material. Prevention is preferable to remedy: ensuring adequate pre-clarification, avoiding excessive retentate concentration before diafiltration begins, and maintaining sufficient CFV to sweep the membrane surface all reduce fouling risk. Fouled membranes can sometimes be partially recovered with an alkaline clean-in-place (CIP) cycle using 0.1 to 0.5 M NaOH, but this is best treated as a recovery option rather than a planned step.

Lower than expected contaminant removal after the predicted number of diavolumes is usually caused by incomplete mixing in the retentate vessel. If the diafiltration buffer added at one point in the vessel is not well mixed before reaching the membrane, local dilution is lower than assumed and the exponential removal formula overpredicts performance. Using magnetic or paddle mixing, and ensuring that buffer addition and retentate return are not at the same location in the vessel, resolves most mixing problems.

EV aggregation during diafiltration typically appears as a shift in particle size distribution toward larger sizes and an increase in polydispersity on NTA. It usually indicates either that the diafiltration buffer has a suboptimal ionic strength (too low leads to charge-driven aggregation; too high screens stabilising surface charges), that the retentate concentration is too high, or that there is a significant pH mismatch between the sample and the diafiltration buffer. Running a short buffer compatibility check on a small sample aliquot before committing a full preparation to diafiltration is good practice when working with a new EV source.

Scaling Diafiltration

One of the practical advantages of TFF diafiltration is that scale-up is relatively predictable. The key scaling parameter is membrane area, which scales linearly with the volume to be processed at a given target flux and processing time. Hollow fibre cartridge formats are particularly amenable to scale-up because adding cartridge area does not fundamentally change the fluid mechanics at the membrane surface, unlike scale-up of flat sheet cassette systems where channelling and edge effects can become more significant at larger format.

Retentate vessel geometry matters more at scale. At small scale, a 50 ml conical tube with a magnetic stir bar mixes adequately. At larger scale, the mixing regime needs deliberate engineering: turbulent mixing, baffled vessel design, or a dedicated recirculation loop to ensure homogeneity. Protocols that work well in research-scale vessels (100 ml to 500 ml) sometimes perform worse than expected at pilot scale (5 L to 50 L) for this reason rather than because of any change in the membrane performance.

EVlution TFF: An Integrated Platform for EV Diafiltration

Implementing TFF diafiltration reproducibly at laboratory scale requires a system that can control transmembrane pressure and cross-flow velocity accurately, monitor flux in real time, and accommodate the hollow fibre cartridges and small retentate volumes typical of EV research workflows. The EVlution TFF system from Cell Guidance Systems is designed specifically for this context.

The EVlution TFF platform integrates peristaltic pump control, pressure monitoring, and a compact instrument footprint suited to a standard cell culture hood or biosafety cabinet. It supports the hollow fibre cartridge formats most commonly used for EV purification and is compatible with the concentration and diafiltration workflows described in this article. A dedicated EV processing service is also available for groups that want processed material without in-house instrumentation.

For groups characterising EVs post-purification, ExoLISA assays provide quantitative surface marker detection on EV preparations, and NTA size profiling is available as a service to verify particle size distribution and concentration after TFF processing.

Diafiltration in Context: Where It Sits in the Purification Landscape

It is worth being clear about what diafiltration can and cannot do, because it is sometimes positioned as a standalone purification method when it is more accurately a conditioning step. Diafiltration removes small molecules efficiently and can significantly reduce the protein content of an EV preparation if a membrane MWCO is chosen that allows free proteins to pass. It does not, however, separate EVs from co-purified material of similar size, such as lipoproteins in the 100 nm to 200 nm range. For applications where lipoprotein removal is required, a downstream SEC or density-based step remains necessary.

Similarly, diafiltration does not remove endotoxin reliably unless the membrane MWCO is high enough that endotoxin micelles (which aggregate to form structures of 100 kDa or larger) are allowed to pass. Specific endotoxin reduction requires either affinity-based endotoxin removal columns or dedicated diafiltration membranes designed for endotoxin clearance.

Understanding these boundaries is what allows researchers to use diafiltration confidently: it is an exceptionally efficient tool for what it does, and integrating it into a workflow that uses its strengths appropriately is the most reliable path to reproducible, well-characterised EV preparations.

Summary

Diafiltration transforms TFF from a concentration technique into a complete buffer management platform. The underlying mathematics are straightforward: five to seven diavolumes remove more than 99% of freely transmitting contaminants, and the exponential relationship between diavolumes and removal is a practical guide for protocol design. For EV applications, continuous diafiltration against a physiologically balanced target buffer, at moderate retentate concentration and adequate cross-flow velocity, delivers reproducible buffer exchange with acceptable particle recovery. The technique scales predictably with membrane area, integrates readily with downstream polishing steps, and is considerably more time-efficient than dialysis for research-scale volumes.

The practical challenges, principally membrane fouling, incomplete mixing, and EV aggregation, are all manageable with good protocol design and appropriate equipment. For groups establishing or refining EV purification workflows, diafiltration is not an optional add-on: it is the step that determines whether the final preparation is in the right buffer, at the right concentration, and free of the process-related impurities that would compromise downstream experiments or clinical applications.

Further Reading

• Witwer KW, et al. Standardization of sample collection, isolation and analysis methods in extracellular vesicle research. J Extracell Vesicles. 2013;2:20360.
• Nordin JZ, et al. Ultrafiltration with size-exclusion liquid chromatography for high yield isolation of extracellular vesicles preserving intact biophysical and functional properties. Nanomedicine. 2015;11(4):879-883.
• Heinemann ML, Vykoukal J. Sequential Centrifugal Ultrafiltration of Plasma Enables Measurement of Lipoprotein and LDL Particle Size Subfractions. J Lipid Res. 2019.
• van der Pol E, et al. Particle size distribution of exosomes and microvesicles determined by transmission electron microscopy, flow cytometry, nanoparticle tracking analysis, and resistive pulse sensing. J Thromb Haemost. 2014;12(7):1182-1192.
• Bioprocess International. Diafiltration: A Practical Guide. BPI Supplement. 2012.

Learn more about EVlution TFF and EV purification tools from Cell Guidance Systems: