Extracellular Vesicle Isolation: Ultracentrifugation, SEC and TFF

By Benn Bluestein-Veyra, M.Sc. Organic Chemistry · Updated 20 October 2026

Extracellular vesicle isolation methods separate vesicles from cells, debris, soluble protein and lipoproteins by exploiting one physical property at a time: sedimentation rate, buoyant density, size, solubility or a surface protein.^[1] Each method trades recovery against purity and scale, and published comparisons measure that trade directly.^[2][3]

This reference covers differential ultracentrifugation, density gradients, size-exclusion chromatography, tangential flow filtration, polymer precipitation and immunoaffinity capture. It then summarizes what side-by-side studies report and how the field records isolation choices through MISEV2023 and EV-TRACK.

Extracellular vesicle isolation methods exploit distinct physical properties

Vesicles share their size and density ranges with other particles in conditioned media and body fluids, so separation from non-vesicular particles is a central task of every study.^[1] Théry and colleagues framed purification this way in their 2006 protocol, which pairs isolation from culture supernatants and fluids with tests of purity and homogeneity.^[4]

Sedimentation rate drives differential ultracentrifugation, buoyant density drives gradients, and size drives chromatography and filtration (Fig. 1).^[4][5] Polymer precipitation acts on solubility, and immunoaffinity capture selects vesicles that carry a chosen surface protein.^[6][7]

The 2016 worldwide survey led by Gardiner found conditioned cell culture media to be the most widely used starting material, at 83 percent of respondents. Sample volume, sample type and the intended downstream analysis all shaped which isolation method a laboratory chose.^[8]

Differential ultracentrifugation separates vesicles by sedimentation rate

Differential ultracentrifugation applies increasing centrifugal force in sequence. Low-speed steps remove cells and debris, an intermediate step pellets larger vesicles, and a final high-speed step pellets small vesicles, which are then washed and pelleted again.^[4]

It remains the most widely used method. Ultracentrifugation was used by 81 percent of respondents in the 2016 ISEV survey,^[8] and the 2019 follow-up survey found ultracentrifugation and density gradients still the most common separation methods.^[9]

Rotor type, run time and force shape the pellet

Cvjetkovic and colleagues compared swinging bucket and fixed angle rotors across run times. Pelleting of small vesicles depended on rotor type and run time as well as g-force; longer runs raised RNA and protein yield, and very long runs added excess protein to the pellet.^[10]

Pelleting also packs particles together. Linares and colleagues showed by electron microscopy and flow cytometry that high-speed centrifugation formed aggregates of vesicles with mixed phenotypes,^[11] and Lobb and colleagues reported that repeated ultracentrifugation steps lowered vesicle yield.^[3]

In plasma, the high-speed pellet also collects protein aggregates. Böing and colleagues described this co-isolation as the motivation for a single-step chromatography method that separates vesicles from protein and lipoproteins.^[5]

Density gradients resolve vesicles by buoyant density

Density gradient ultracentrifugation places a sample on top of, or beneath, a continuous or stepped gradient of iodixanol or sucrose. During centrifugation each particle moves to the layer that matches its own density, which separates membrane vesicles from denser protein complexes.^[1]

Iodixanol is a dimeric, non-ionic and iso-osmolar iodinated compound, developed as a radiographic contrast medium and widely used as a gradient medium (Fig. 2).^[12] Its two tri-iodinated benzene rings give it the formula C35H44I6N6O15.^[12]

Kowal and colleagues first separated dendritic cell vesicles by sedimentation speed and then floated them upward into iodixanol gradients. Quantitative proteomics showed that classical markers such as MHC, flotillin and HSP70 were present in all vesicle populations.^[7]

Van Deun and colleagues compared an iodixanol density gradient with ultracentrifugation and two commercial precipitation reagents. The gradient gave the most CD63-positive vesicles and the strongest enrichment of exosomal marker proteins, and it separated vesicles from extracellular Argonaute-2 complexes.^[2]

Jeppesen and colleagues combined high-resolution density gradient fractionation with direct immunoaffinity capture. They assigned Argonaute 1 to 4, glycolytic enzymes and cytoskeletal proteins to non-vesicular material and identified annexin A1 as a marker of microvesicles shed from the plasma membrane.^[13]

Density separation is especially informative for blood. In an iodixanol gradient of rat plasma, Onódi and colleagues found vesicle markers at 1.128 to 1.174 grams per milliliter, a fraction that kept only 1.4 percent of the LDL and chylomicron marker and 3.0 percent of the HDL marker.^[14]

Size-exclusion chromatography separates vesicles from soluble protein by size

Size-exclusion chromatography passes a sample through a bed of porous beads. Proteins and small molecules enter the pores and elute late, while vesicles larger than the pores travel around the beads and elute in the early fractions.^[5]

Böing and colleagues ran platelet-free supernatant of platelet concentrates on a sepharose CL-2B column. The early fractions held 61 percent of the collected platelet-derived vesicles with under 5 percent of the HDL and under 1 percent of the protein, an 8-fold and 70-fold enrichment over each.^[5]

Vesicle recovery in that study was 43 percent of the starting material.^[5] Lobb and colleagues later compared four isolation techniques and found size exclusion comparable to density gradient purification.^[3]

Gámez-Valero and colleagues reported that SEC removed most soluble plasma protein and was the only method in their comparison that allowed detection of CD9, CD63 and CD81. The authors suggested that precipitating agents may interfere with vesicle structure or composition.^[15]

Lipoproteins are the main co-isolates in blood. Karimi and colleagues found vesicle markers in the same SEC fractions as lipoprotein particles, and adding a density cushion to SEC reduced lipoprotein particles 100-fold and allowed 1,187 proteins to be identified in plasma vesicles.^[16]

The equipment is simple: a column of porous resin, run by gravity or a low-pressure pump, with fractions collected in sequence below it (Fig. 3).^[5]

Tangential flow filtration and chromatography scale isolation to large volumes

Tangential flow filtration sweeps the feed along the surface of a membrane. Vesicles larger than the pores stay in the recirculating retentate and become concentrated, while water, salts and small proteins pass through as permeate.^[17]

Busatto and colleagues compared tangential flow filtration with ultracentrifugation for conditioned cell culture media. TFF concentrated vesicles of comparable physicochemical characteristics with higher yield, fewer free macromolecules and aggregates, and better batch-to-batch consistency in half the processing time.^[17]

Haraszti and colleagues grew umbilical cord mesenchymal stem cells on microcarriers in three-dimensional culture. That culture raised exosome yield 20-fold over flat culture isolated by ultracentrifugation, and tangential flow filtration raised it a further 7-fold.^[18]

Ultrafiltration offers a similar route at bench scale. Nordin and colleagues found that ultrafiltration followed by liquid chromatography gave a significantly higher vesicle yield than ultracentrifugation, with unchanged vesicle protein composition and preserved biophysical properties.^[19]

Charge offers another handle. Heath and colleagues used the net negative charge of vesicles to isolate them by anion exchange chromatography, recovering 2.4 × 10¹¹ vesicles from 1 liter of culture supernatant, with less protein and debris than in tangential flow preparations.^[20]

Precipitation and immunoaffinity capture select by solubility and by surface protein

Polymer precipitation favors yield

Polymers such as polyethylene glycol lower the solubility of vesicles so they pellet at low speed. Rider and colleagues developed this approach as ExtraPEG, which enriched vesicles from large volumes of media using low-speed centrifugation and one small-volume ultracentrifugation step.^[6]

Comparative studies place precipitation at the high-yield, lower-purity end. Lobb and colleagues found precipitation gave the least pure preparations from both culture media and plasma,^[3] and Van Deun and colleagues reported higher purity from density gradients than from commercial precipitation reagents.^[2]

Gámez-Valero and colleagues found that PEG precipitation retained the abundant soluble plasma proteins that SEC removed. They advised matching the isolation method to the intended use of the vesicles.^[15]

Immunoaffinity capture selects a marker-defined subset

Immunoaffinity capture binds vesicles through an antibody to a surface protein, most often one of the tetraspanins CD9, CD63 and CD81 (Fig. 4). Kowal and colleagues found exosomal and non-exosomal subpopulations within small vesicles and proposed immuno-isolation with CD63, CD81 or CD9 to separate them.^[7]

CD63 is a tetraspanin glycoprotein encoded by a gene at human chromosome 12q13.2.^[21] Jeppesen and colleagues used direct immunoaffinity capture to define the protein, RNA and DNA content of classical exosomes, separately from the non-vesicular material collected alongside them.^[13]

Recovery by capture reflects the share of vesicles that carry the target, which varies between vesicle subtypes.^[7] Immunoaffinity results therefore describe a defined subpopulation of the vesicles in a sample, which complements the bulk view given by density and size separation.^[13]

Fig. 4 | Tetraspanin CD9, Tetraspanin CD63 (predicted) and Tetraspanin CD81 with cholesterol.

Combined workflows and published comparisons map recovery against purity

Most laboratories combine methods. In the 2016 survey, 59 percent of respondents used more than one,^[8] and the 2019 survey found rising use of size-exclusion chromatography, with tangential flow and microfluidic techniques each used by more than 10 percent of respondents.^[9]

Benedikter and colleagues concentrated bronchial epithelial cell media by ultrafiltration and then ran size-exclusion chromatography. The UF-SEC vesicles tended toward higher yield and a higher vesicle-to-protein purity than ultracentrifuged vesicles, and the same run returned the vesicle-free secreted protein for comparison.^[22]

For plasma, Onódi and colleagues followed an iodixanol gradient with bind-elute chromatography, which brought albumin and lipoprotein below the detection limit with a lower overall yield.^[14] Brennan and colleagues compared common techniques alone and in combination on human serum and found that the procedure shaped yield, lipoprotein content and protein content.^[23]

Asymmetric flow field-flow fractionation resolves subpopulations by size

Asymmetric flow field-flow fractionation separates particles in a thin channel against a perpendicular cross flow, with no packed stationary phase. Zhang and colleagues used it to resolve large exosome vesicles of 90 to 120 nm, small exosome vesicles of 60 to 80 nm, and non-membranous nanoparticles of about 35 nm that they named exomeres.^[24]

Across these comparisons a consistent pattern emerges. Precipitation and tangential flow favor recovery and scale, and SEC and combined workflows sit in the middle. Density gradients favor purity, and immunoaffinity capture isolates a marker-defined subset (Fig. 5).^[2][3][22]

Purity ratios and EV-TRACK make isolation choices reportable

Webber and Clayton proposed a simple purity estimate: the ratio of vesicle counts from nanoparticle tracking to protein concentration from a colorimetric assay. They suggested it as a routine quality step that would let laboratories compare purity across studies.^[25]

MISEV2023 summarizes the advantages and limitations of current approaches to producing, separating and characterizing vesicles from cell culture, body fluids and solid tissues.^[1] It compiles feedback from ISEV task forces and more than 1,000 researchers.^[1]

The EV-TRACK Consortium described its knowledgebase in 2017 as a crowdsourced resource that centralizes EV biology and methodology.^[26] Its checklist for authors and editors makes the methods of EV experiments transparent, and the database supports queries and meta-analysis of published experiments.^[27]

In the 2019 survey, most researchers said separation and characterization deserve more attention.^[9] Reporting the isolation method with a purity ratio and marker data is how the field now compares preparations across laboratories.^[1][25]

Where to go from here

The protein marker reference explains the MISEV categories that isolation studies measure, and the purity ratios reference covers particle-to-protein and related metrics in depth. The single-vesicle analysis reference describes how individual vesicles are counted and phenotyped after isolation.

The quality references on NTA and MRPS particle counting and on how exosomes are tested show how these measurements appear in supplier documentation. Verified practices can request lot documentation, and our team is glad to discuss the methods described on this page.