Dialysis is a classic laboratory technique that relies on selective diffusion of molecules across a semi-permeable membrane to separate molecules based on size. Dialysis is used for a wide variety of applications: desalting, buffer exchange, removal of labeling reagents, drug binding studies, cell growth and feeding, virus purification, and blood treatment. Typically, a sample and a buffer solution (called the dialysate) are placed on opposite sides of a dialysis membrane which contains pores of a manufactured size-range. Sample molecules that are larger than the pores are retained on the sample side of the membrane, but small molecules pass through the membrane, reducing the concentration of those molecules in the sample (Figure 1). Alternatively, desired components in the external buffer solution can be slowly brought into the sample.

Figure 1. How dialysis membranes work. A dialysis membrane is a semi-permeable film (usually a sheet of regenerated cellulose) containing various sized pores. Molecules larger than the pores cannot pass through the membrane but small molecules can do so freely. In this manner, dialysis may be used to perform purification or buffer exchange for samples containing macromolecules.

The separation characteristic determined by the pore size-range of a dialysis membrane is most often referred to as the molecular weight-cutoff (MWCO) of the membrane. Traditionally, a membrane’s MWCO refers to the smallest average molecular mass of a standard molecule that will not effectively diffuse across the membrane. Typically, the smallest size globular macromolecule (in Daltons) that is retained by greater than 90% upon extended dialysis (overnight) defines the nominal MWCO. Thus, a dialysis membrane with a 10K MWCO will generally retain proteins having a molecular mass of at least 10kDa.

It is important to note that the MWCO of a membrane is not a sharply defined value. The diffusion of molecules near the MWCO will be slower compared to molecules significantly smaller than the MWCO. And dialysis membranes, which are composed of regenerated cellulose, contain a broad range of pore sizes; it is practically impossible to achieve 100% retention of even very large molecules. To ensure adequate time for removal of contaminants in a dialysis application, it is essential to understanding these properties of dialysis membranes and the influence of other factors, such as surface area-to-sample-volume ratio (SA:V).

In this article, we characterize the separation properties of dialysis membranes having nominal MWCO ratings of 2K, 3.5K, 7K, 10K and 20K. We also compare dialysis rates and other specifications among various sizes of Thermo Scientific Slide-A-Lyzer Devices and Thermo Scientific SnakeSkin Dialysis Tubing, which are designed to process samples from 0.1mL to 250mL.

Results and discussion

To characterize and define the MWCO and retention properties of our 2K, 3.5K, 7K, 10K, and 20K dialysis membranes, we examined a number of molecules to determine the percent retained in the sample after overnight dialysis in 3mL-capacity Slide-A-Lyzer Dialysis Cassettes (Figure 2, Panels A to E).

Figure 2A. Retention with 2K MWCO dialysis membrane:


Figure 2B. Retention with 3.5K MWCO dialysis membrane:


Figure 2C. Retention with 7K MWCO dialysis membrane:


Figure 2D. Retention with 10K MWCO dialysis membrane:


Figure 2E. Retention with 20K MWCO dialysis membrane:

Figure 2. Determination of the MWCO for a series of dialysis membranes. Panels chart the percent retention for solutions of various test molecules (see graphs) after overnight (17 hours) dialysis at 4°C in 3mL-capacity Slide-A-Lyzer Dialysis Cassettes having dialysis membranes with the indicated 2K to 20K MWCO ratings. Samples were prepared at a starting concentration of 0.5 to 1mg/mL in either PBS (pH 7) or 0.2M carbonate bicarbonate buffer (pH 9.4). Retention was measured using either the Thermo Scientific Pierce BCA Protein Assay (Part No. 23225) or absorption at 360nm (for vitamin B12). In each panel, the division between molecules considered to be smaller than the MWCO and those considered to be slightly larger than the MWCO is demarcated by a change from gray to colored bars. Bar colors correspond to the product cassette colors.

With each membrane (panel), retention levels increase for molecules of increasing mass (size) until a plateau is reach at approximately 90% retention. Further increases in sample molecular size are accompanied by only slight (or no) increases in retention (see Panels A and B). A large proportion (even a majority) of tested peptides and biomolecules that are smaller than the nominal MWCO of the membrane are retained as well. This demonstrates that dialysis is not an effective method for separating molecules of similar size. Instead, dialysis is generally most suitable for exchange of buffering salts, inorganic chemicals, and other media components that are very much smaller (e.g., two or three orders of magnitude smaller) than the MWCO-rating of the membrane (see Figures 3 and 4 below).

However, it is also important to note that MWCO ratings are based on globular molecules (e.g., proteins). More linear molecules, such as DNA or RNA, which may have a small diameter in two of three dimensions, may be able to pass through the pores more freely despite having molecular weights that exceed the stated MWCO (data not shown). To ensure proper retention of DNA or RNA samples, researchers typically select a dialysis membrane whose MWCO is one-third to one-half the molecular weight of the nucleic acid of interest.

Molecules whose sizes (masses) are near to the same order of magnitude as the MWCO have variously restricted dialysis rates, depending on their shape and solubility characteristics. By contrast, relatively very small molecules (especially highly soluble ones) usually have very similar rates of diffusion because they can pass through a membrane’s pores freely and unconstrained.

To demonstrate the influence of MWCO on the dialysis rate of small molecules, we dialyzed 200mL of 1M NaCl versus water using Thermo Scientific Slide-A-Lyzer Dialysis Flasks possessing 2K, 3.5K, 10K and 20K dialysis membranes (Figure 3). Dialysis rates for the 3.5K, 10K, and 20K membranes were very similar, each resulting in complete salt removal in less than 10 hours. The sodium and chloride ions of salt have molecular weights (23 and 35g/mol, respectively) that are orders of magnitude less than the MWCOs of these membranes, which have similar thicknesses and pore-densities. By contrast, the dialysis rate for the 2K membrane is significantly slower because it has much smaller pores and a much thicker (50µm vs. ~25µm) membrane compared to the others.

Figure 3. Membrane MWCO and time-course of dialysis. Rates of removal of 1M NaCl from 200mL samples in 2K, 3.5K,10K, and 20K MWCO Thermo Scientific Slide-A-Lyzer Dialysis Flasks at room temperature. At the indicated times (triangles), the dialysis buffer (4L) was changed and the percentage of NaCl removal was determined by measuring the conductivity of the sample. Greater than 95% of NaCl was removed in 8 to 18 hours (41 hours for the 2K condition). The average thickness of the 2K, 3.5K, 10K, and 20K membranes is 50, 23, 30, 25 um, respectively.

Although the size and number of the pores in a dialysis membrane, along with its thickness, have a major effect on determining the rate (or probability) at which molecules of different sizes diffuse through the pores into the external buffer, the rate of dialysis is also directly proportional to the surface area of the membrane in relationship to the volume of the sample. The more a sample can be spread over a membrane surface, the faster dialysis will occur as molecules will more frequently interact with the membrane during diffusion. High-performance dialysis products, such as Slide-A-Lyzer Dialysis Cassettes, Flasks and MINI Devices, are designed with a surface area-to-volume ratio optimized for both speed and ease of handling for a variety of different volumes of sample.

To examine the influence of the surface area-to-volume ratio, we dialyzed 1M NaCl samples versus water in four different sizes of dialysis devices having the same (3.5K) MWCO membrane (Table 1, Figure 4).

Table 1. Devices and parameters of dialysis rate experiment with sodium chloride. The surface area-to-volume ratio (SA:V) is calculated based on the shape of each chamber with the tested volume of sample.
Thermo Scientific Product Device Capacity Tested Volume SA:V
Slide-A-Lyzer MINI Device
(Part No. 88403)
2mL 2mL 1.25cm2/mL
SnakeSkin Dialysis Tubing
(Part No. 88244)
35mm dia. 70mL 1.32cm2/mL
Slide-A-Lyzer G2 Cassette
(Part No. 87726)
70mL 70mL 1.32cm2/mL
Slide-A-Lyzer Dialysis Flask
(Part No. 87761)
250mL 200mL 0.65cm2/mL
Figure 4. Influence of surface area to volume ratio on dialysis rate. Graph displays rates of removal of 1M NaCl from 2mL, 70mL, 70mL, and 200mL samples dialyzed in four respective sizes of Thermo Scientific Dialysis Devices (see Table 1), each equipped with 3.5K MWCO membrane. Dialysis was conducted at room temperature against very large volumes (e.g., 4L) of water (dialysate). At the indicated times (triangles), the dialysis buffer was changed and the percentage of NaCl removal was determined by measuring the conductivity of the sample.

The three sample with similar SA:V values (~1.3cm2/mL) exhibited similar dialysis rates (~95% removal of salt in 4 to 6 hours). The larger Slide-A-Lyzer Dialysis Flask, with only half the SA:V of the other devices, required about twice as long (10 hours) to achieve the same 95% removal of salt. This demonstrates that, all else being equal, dialysis rate is directly proportional to ratio of surface area to sample volume. In addition, the differently shaped dialysis devices interact with the dialysate in subtly different ways that affect efficiency of sample diffusion.


These simple experiments demonstrate several key characteristics of dialysis as it relates to the common life science laboratory applications of sample desalting and buffer exchange. They help to clarify the meaning of molecular weight cutoff (MWCO) values – that these are nominal classifications for membranes rather than discrete and precise boundaries. Dialysis is not an effective method to separate molecules of similar size.

For buffer exchange and desalting, the rate of dialysis is directly proportional to the membrane surface area-to-volume ratio (SA:V). Therefore, it is important to select a dialysis device that maximizes SA:V for the intended sample while still providing convenient and trouble-free sample addition and recovery.

It is important to note that every molecule is different; the concentration, interactions, and hydrophobicity of molecules can influence their ability to diffuse through a dialysis membrane. The temperature, volume, agitation rate and frequency of exchange of the external buffer are also important factors. Therefore, some amount of empirical testing is usually necessary to optimize a dialysis protocol for a specific sample and application.