• Many EV therapeutics will never reach patients — not because the biology fails, but because the manufacturing economics miss the mark.
• RoosterBio builds downstream processing (DSP) platforms designed from day one for commercial scale, using scalable unit operations and proprietary yield-enhancing technology to eliminate the inefficiencies that make EV manufacturing cost-prohibitive.
• Combined with products like AgentV™-DSP, our recent process innovations advance a platform that achieves 40-70% particle recovery (compared to a typical industry baseline of 5-10%) and can drop an approximate production costs per dose from over $2,000 to under $250.

Downstream Processing (DSP) directly impacts the critical quality attributes (CQAs) of safety, identity, potency and stability of extracellular vesicle (EVs; aka “exosome”) therapeutic products. It matters for manufacturing costs as well. Thus, while in pursuit of optimized upstream processing [1, 2] RoosterBio has also sought valuable practical experience with DSP. This deep dive into problem-solving facilitates an end-to-end solution set for our partners and clients. What have we learned? To bring you up to speed, we recently prepared a well-attended webinar (The DSP Trade-Off: How Downstream Processing Decisions Impact EV Yield, CQAs, & Cost of Goods), presented by Jae Jung (Process Development Scientist) and Trey Picou, PhD (Product Manager). Watch it here. [3]

Until recently, [4, 5, 6] DSP hasn’t often gotten top billing in most EV therapy development conversations. Upstream cell and media productivity, cell source, or culture vessel format tended to dominate early discussions about EVs harvested from MSCs, hDFs (fibroblasts), dendritic cells, or HEKs. [7] However, as Jung and Picou explained, choices made in DSP can ripple through every dimension of a program’s commercial viability, from critical quality attributes to cost per dose.

First, Jung methodically unpacked what those choices are: how product type definition, unit operation selection, and hard-won process solutions determine whether a program can produce a well-characterized EV product. Next, Picou translated those choices into commercial consequences, showing that DSP yield is a multiplier on every upstream cost; the modeled difference between 10% and 40% particle recovery credibly contributes to the difference between a $2,000 dose and a $250 one. Interested in how to bring down costs and hassle? Read on!

How to Get Conditioned Media, Concentrated Secretome, or Purified EVs: It Starts with Definition

One person’s broth is another’s chowder. Similarly, the supernatant, comprised of spent cell culture medium, can be harvested and prepared in diverse ways after the cells release their product into it. [8, 9, 10] This EV-containing mix can be as complex as a stew or as refined as a consommé’. Regardless of what the end-product is, it’s important to define it carefully such that any sequence of downstream unit operations will be fit for purpose.  RoosterBio distinguishes three primary product categories: conditioned medium, concentrated secretome, and purified EVs (Figure 1). Each represents a progressively tighter definition of what counts as product versus impurity, and each requires a different combination of unit operations to manufacture consistently.

Figure 1. In EV downstream processing your components of interest are what define your EV product type. And your product type, in turn, drives every downstream processing decision you make. Conditioned Medium is the simplest output, the cell culture supernatant collected after cells have been expanding and secreting; it contains EVs, soluble proteins, cytokines, metabolites, and growth factors. Concentrated Secretome is a step up in processing. Here medium components and some soluble elements are removed, while the EV fraction is concentrated somewhat; some secreted proteins remain and this mix is equilibrated into a therapeutically relevant buffer. Purified EVs represent the highest level of downstream processing, with free proteins, lipid aggregates, and other impurities removed.

Conditioned medium retains the full secreted milieu and requires only normal flow filtration (NFF) for clarification. Concentrated secretome adds tangential flow filtration (TFF) to remove medium components and concentrate the EV and protein fraction, suited for therapies that derive potency from both EVs and secreted factors like cytokines. Purified EVs go further still, incorporating chromatography to selectively enrich the EV fraction and remove free proteins and lipid aggregates, the appropriate target for genetically modified or cargo-loaded EV products heading toward clinical or commercial manufacturing.

This tiered structure isn’t just conceptually useful. It has direct consequences for analytical assays and a quality monitoring plan. As Jung laid out, the assay panel that defines product identity shifts at each tier, and components that represent identity in conditioned medium become quantified impurities at the purified EV stage. “Do it right the first time” …says any teacher who ever lived. The same applies here. Avoid costly trouble by first defining your product meticulously.

“Sciencing” Through the Manufacturing Challenges

RoosterBio’s process development flow isn’t merely abstract. It is the product of empirical trial and error in the real world. When it hits a real manufacturing wall, the team digs in and “sciences” (verb) its way around it. This could mean developing a novel proprietary reagent or rethinking a core chromatography strategy from the ground up. Accordingly, Jung walked his presentation through two specific technical challenges that the RoosterBio team encountered during EV DSP development, and the solutions emerged.

The first example involved filter fouling. As upstream process input volume increased, NFF (normal flow filtration) recovery dropped sharply due to fouling of the filter membrane by cell debris and extracellular matrix components, which were co-retaining EVs. After systematic root cause analysis, the team devised a proprietary processing reagent, AgentV-DSP, that ameliorated the fouling mechanism directly (Figure 2). The results were substantial NFF cumulative yield improved from 56% to 89%, and final purified EV vial yield rose from 4% to 52%, a greater than tenfold improvement in overall recovery.

Figure 2. AgentV-DSP treatment significantly improves particle yields through filtration steps, shown above in a sequence of EV concentration and purification unit operations. Additional yield and improvement in the product stream quality cumulates in over 10-fold improvement in total process yield. 10-fold improvement allows for 10-fold reduction in batch size to meet fixed demand. [11]

The second challenge involved high-protein collection media. When producer cells are expanded and EVs collected in nutrient-rich media containing abundant protein, the particle-to-protein ratio in the harvest is skewed enough to make standard flow-through chromatography economically unworkable. Although this remains a key purity metric, column loading scales with protein burden, and resin cost adds up to a major fraction of DSP COGS. RoosterBio’s response was to develop a bind-and-elute chromatography method (Figure 3). Rather than trapping impurities and collecting EVs in the flow-through, the column is conditioned to bind EVs selectively while proteins pass through. The result was improved purity, competitive particle recovery, and a column volume requirement that no longer scales adversely with protein load.

Figure 3. To remove protein impurities and reduce column matrix costs, RoosterBio devised a bind-elute chromatography operation for EV purification. Bind and elute chromatography typically binds multiple components and resolves them by optimizing the elution buffer conditions, as shown in a representative elution profile (above left) with a tight peak of EV elution in fractions after ~75mL of eluate.  The text summary highlights key development steps taken to build the final recipe (above right).

In short, Rooster’s DSP team surmounted a yield challenge and a scalability challenge. Its process development expertise and platform methods are now the foundation to routinely apply toward new client programs and accelerate the booming EV therapy programs appearing worldwide. With that, Jung then turned the webinar over to Trey Picou.

Between DSP Yield and Patients Healed… Drive COGS Toward Commercial Appeal

Picou framed a business case around three requirements: CQA maintenance, manufacturing scalability, and cost of goods reduction. His central point is that none of these can be optimized in isolation. For example, a process that achieves high purity but destroys yield will demand more upstream runs. And a process that works at bench scale but relies on ultracentrifugation cannot be linearly scaled to commercial volumes without fundamental redesign. And both of those problems ultimately express themselves as cost per dose.

Picou next pointed to three interdependent levers for pulling cost per dose toward commercial viability: (1) improving DSP yield, (2) adopting a platform process approach, and (3) reducing the number of process steps. A platform process approach reduces development time and enables process transfer across products and indications. Reduction in process steps means fewer unit operations and thus fewer consumables, less labor, shorter batch cycles, and fewer failure points. However, DSP yield improvement is by far the most powerful. Typical legacy processes recover only 5 to 10 percent of produced EVs, meaning 90 cents of every dollar spent on upstream manufacturing is discarded before a single dose is filled. Moving from that baseline to 50 percent (or more) recovery doesn’t just improve DSP efficiency in isolation. It reduces upstream bioreactor runs, media consumption, labor, and facility time by a proportional factor, cascading savings across the entire manufacturing chain.

The numbers illustrate the stakes. A published 3D fed-batch culture system operating at 10% DSP recovery requires eight 50-liter bioreactors to produce 2,500 doses at 10¹¹ EVs per dose, at a cost exceeding $2,000 per dose from upstream and DSP inputs alone. The RoosterBio platform, running at higher cell densities and achieving a conservative estimate of 40% DSP recovery through AgentV-DSP treatment, produces the same lot output from a single 50-liter bioreactor in 15 days or less, at under $250 per dose. That is not a marginal improvement. It is the difference between a program that reaches patients and one that stalls on manufacturing economics.

 

Figure 4. Interdependent, strategic DSP decisions to (1) Improve DSP Yield, (2) Employ a Platform Process Approach, and (3) Reduce Process Steps can have a dramatic effect on the cost and upstream process volume to produce a dose of EVs. Above, we have compared a competitor EV process (left) with RoosterBio’s manufacturing innovations (right) and shown that RoosterBio’s methods can help economize extracellular vesicles (aka “exosomes”) into viable future medicines.

The underlying principle is straightforward: DSP yield is a multiplier on every upstream cost. At 10% recovery, 90 cents of every dollar spent growing cells and producing EVs is discarded. Through the power of integrated platform approaches, technologies like AgentV-DSP can simultaneously improve yield, maintain CQAs, and operate within scalable, GMP-ready systems. We believe these will help trailblaze the path to make EV therapeutics commercially viable and accessible to patients.

Designing for Scale from Day One

Together, Jae and Trey were emphatic on this point: scalability must be a design criterion from the earliest stages of process development, not a retrofit applied after the science is settled.

NFF, TFF, and chromatography all have bench-scale implementations that faithfully represent commercial conditions. Starting there means the process you develop is the process you scale. Moreover, the data you generate along the way is interpretable in a commercial context rather than having to be rebuilt from scratch.

Questions from the post-webinar Q&A reflect exactly where development teams are wrestling with these decisions in practice. For example, the audience asked:

• When should sterile filtration be introduced given its known impact on EV size distribution?
• How do you develop at 100mL bench scale while keeping an eye on clinical volumes?
• Can AgentV-DSP extend beyond the RoosterBio media platform?

These are completely addressable questions, and RoosterBio will be following up on them soon in a forthcoming blog. Practically speaking, such questions will get cheaper to answer as earlier scalable technologies are built into the process.

If your team is working through EV DSP decisions and wants to discuss how RoosterBio’s solution set could support your program, reach out to start a conversation with our process development team.

References

1. Lenzini, Stephen. Extracellular Vesicle/Exosome Upstream Process Development: Maximizing Productivity to Accelerate Clinical Adoption. 2022; Available from: https://www.roosterbio.com/blog/extracellular-vesicle-exosome-upstream-process-development-maximizing-productivity-to-accelerate-clinical-adoption/.
2. RoosterBio. On the Exosome/Extracellular Vesicle Frontier, Choose Your Own Adventure. 2023; Available from: https://www.roosterbio.com/blog/on-the-exosome-extracellular-vesicle-frontier-choose-your-own-adventure/.
3. Jae Jung, Trey Picou. The DSP Trade-Off: How Downstream Processing Decisions Impact EV Yield, CQAs, & Cost of Goods. 2026; Available from: https://share.hsforms.com/1aSahmdx5T1Kcne8QvZsKZQ3564o.
4. Staubach, S., et al., Scaled preparation of extracellular vesicles from conditioned media. Adv Drug Deliv Rev, 2021. 177: p. 113940. 10.1016/j.addr.2021.113940
5. Veerman, R. E., et al., Molecular evaluation of five different isolation methods for extracellular vesicles reveals different clinical applicability and subcellular origin. J Extracell Vesicles, 2021. 10(9): p. e12128. 10.1002/jev2.12128
6. Jonathan Carson, Stephen Lenzini, Jae Jung. Countdown to Zero: Overcoming Downstream Processing’s Top Five Challenges for Viral Vectors & Exosomes. 2024; Available from: https://www.roosterbio.com/blog/countdown-to-zero-overcoming-downstream-processings-top-five-challenges-for-viral-vectors-exosomes/.
7. Katrina Adlerz, Josephine Lembong, Tim Olson, Jon Rowley, Taby Ahsan. Xeno-Free Manufacturing of MSC-EVs in Scalable Bioreactor Culture. 2019; Available from: https://www.roosterbio.com/wp-content/uploads/2019/10/RoosterBio_Poster_ISEV2019-_-APR-2019.pdf.
8. Lamparski, H. G., et al., Production and characterization of clinical grade exosomes derived from dendritic cells. J Immunol Methods, 2002. 270(2): p. 211-26. 10.1016/s0022-1759(02)00330-7
9. Kim, S., et al., Dual-mode action of scalable, high-quality engineered stem cell-derived SIRPalpha-extracellular vesicles for treating acute liver failure. Nat Commun, 2025. 16(1): p. 1903. 10.1038/s41467-025-57133-w
10. Lee, Y. X. F., et al., Considerations and Implications in the Purification of Extracellular Vesicles – A Cautionary Tale. Front Neurosci, 2019. 13: p. 1067. 10.3389/fnins.2019.01067
11. Stephen Lenzini, Jae Jung, Madeline Cramer, Elie Zakhem, Jon Rowley. Scalable GMP-compatible Process Solution for MSC-EV Purification with 10X Yield Improvements. 2025; Available from: https://www.roosterbio.com/resource/scalable-gmp-compatible-process-solution-for-msc-ev-purification-with-10x-yield-improvements/.