Response to ‘Purification of Mesenchymal Stromal Cell-Derived Small Extracellular Vesicles Using Ultrafiltration’

Dear Editor,

I read the recent article ‘Purification of mesenchymal stromal cell-derived small extracellular vesicles using ultrafiltration’ by Lei et al. (2025) with great interest, as they have thoroughly characterised an ultrafiltration method also used by me (Boysen et al. 2024). I would like to weigh in with some of my own experiences and findings on their conclusions. The extracellular vesicle (EV) characterisation techniques (TEM and NTA) used by the authors have size and concentration determination limitations. While TEM does a great job of identifying EVs in their cup-shaped appearance, it is not a great tool in size and concentration determination, as the amount of EVs present is few and they have been dried, thereby having lost their true size and morphology. Similarly, NTA has drawbacks as EVs have a low refractive index, the limit of detection is regarded to be around 50 nm (Dragovic et al. 2011), leading to a potentially skewed size profile and reduced particle concentration. The study of Lei et al. is therefore at risk of not showing a potential loss of sub-50 nm EVs using ultrafiltration and their specified molecular weight cut-off (MWCO) values. While they argue that the pore size of the membranes used is 6.13, 8.84 and 10.48 nm for 100, 300 and 500 kDa MWCO, respectively, this is only true from a theoretical perspective based on the hydrodynamic radius of molecules. The pore size of commercially available filters is not based on theoretical values but trial and error and the average pore size of the selected filters by Lei et al. are 10, 35 and 55 nm for 100, 300 and 500 kDa MWCO respectively as seen in Figure 1 (determined using cryo-EM, personal correspondence with Merck Millipore and as reported by Pall corporation (Pall Corporation 2022)). We have translated a copious amount of the knowledge from virology to EV research, as EVs and viruses have many biophysical similarities, and we could also learn from virology in this case. MS2, a icosahedral ∼27 nm bacteriophage, is not retained on a 300 kDa MWCO filter, and can be found in the permeate (Wick and McCubbin 1999). While fouling, particle deposition, and potential loss of EVs can be reduced by adding shear force to the membrane, this force will not be homogeneous, and loss will occur (Hwang et al. 2016; Zhao et al. 2023). Cryo-EM, in contrast to TEM and NTA, can show sub-50 nm particles in their native state (Yuana et al. 2013). As a true size distribution of an EV sample, I have made individual measurements of 130 EVs derived from human embryonic kidney cells (HEK293T) cultured in chemically defined media (HEK VIP NX, Satorius) and enriched using size exclusion chromatography (35 nm, Izon) on a 100 kDa MWCO spin filter (Thermo Fisher) as shown below.

Cryo-EM tends to exclude some of the larger EVs (Welsh et al. 2024) and I have also excluded a few of the largest asymmetrical EVs in my measurements, the size distribution of these HEK293T EVs is in line with what has previously been reported (Zabeo et al. 2017). This data shows that a large population of EVs are at risk of being lost at the larger MWCO filters used by Lei et al., which they cannot show with the methods they used in their manuscript. The stirred cell system utilised by Lei et al. and I is a powerful tool for scaling up from the small-scale and labour-intensive spin filters in an academic setting without investing in industrial or expensive large-scale solutions. While larger MWCO filters for this system seem attractive as they can reduce protein contaminants, this can come with the risk of losing small EVs that could carry valuable information as biomarkers or harbour biologically active payloads of key interest (Willis et al. 2017). I therefore highly encourage that more sensitive sizing tools are needed, like cryo-EM or atomic force microscopy (Zabeo et al. 2017; Ridolfi et al. 2020), before it can be concluded that the proposed methods of Lei et al. are good for small EV purification, as they have effectively not measured up to half of the small EVs in their purifications. Although these would not be biophysically identical, labelled sub-50 nm EV mimetics, liposomes or beads could also be used to validate whether the membranes can retain these.

I declare that I have no competing or conflicting interests.