Lilly's Exploring A Universal AAV Purification Method

A conversation with Juan-Carlos Rosario, Eli Lilly and Company

Contorting AAV purification methods to match serotypes is one of gene therapy development’s great headaches. It’s expensive and time-consuming, with costs ballooning during scale-up. Today, affinity chromatography remains the dominant purification method. However, as capsid engineering accelerates, developers are running up against its limitations. Serotype-specific resins don't scale gracefully.

Juan-Carlos Rosario, a senior principal scientist of purification and virology development at Eli Lilly and Company, was part of a team to develop what they call a universal purification method, one that’s serotype-independent and skips the affinity resins.

In a paper published earlier this year in the Journal of Chromatography A, the group showed a viral genome recovery of nearly 52% with impurity removal of 99.9% for DNA and 99.8% for protein. Their method uses steric exclusion chromatography (SXC) for the capture step combined with anion exchange chromatography (AEX) to separate full/empty capsids.

SXC captures AAV particles based on size rather than capsid surface properties, while AEX separates full of empty capsids by charge.

The research presents a potentially powerful new way to think about AAV purification, and we had questions for the authors. Rosario agreed to help us out, and here’s what he told us. His answers have been lightly edited for clarity.Please briefly summarize the work and your part in it

Rosario: This research was the fruit of a collaboration between Lilly and Sartorius Germany, initiated in 2022 following my review of a paper on applying steric exclusion chromatography to lentiviral vectors. After a series of productive discussions, the joint venture was established. My principal responsibility was to rigorously assess the SXC prototype devices (MA-15 and MA-100) against established commercial capture technologies for AAV vectors. Additionally, I laid the groundwork for optimizing PEG molecular weight (2K, 4K, 6K) and percentage parameters, through this internal data, arising from our proprietary pipeline, which was outside the scope of the published article.

The collaborative team conducted a design of experiments, testing variables such as PEG molecular weight, concentration, and flow rate to refine the process featured in the publication. My contributions to the article focused on all work performed with AAV9-eGFP material derived from the SF-9 insect cell system using both the MA-15 and scale-up to MA-100, plus subsequent polishing using the PRIMA T system.

What were the key shortcomings of resin-based affinity chromatography that motivated this study? Was it mainly about serotype inflexibility, or were there other pain points?

Rosario: While serotype resin specificity was a significant concern, our motivation was chiefly driven by the high-priced costs associated with affinity resins, which are not only expensive but also in most cases, serotype-specific. An additional complication arises from the harsh elution conditions required by many affinity resins—often at sub-physiological pH—which can compromise the integrity of the AAV capsids. Furthermore, high salt concentrations in the eluate necessitate substantial dilution, sometimes as much as 40-fold, to reduce conductivity (typically below 3 mS/cm) before the material is suitable for downstream polishing steps. This multi-layered complexity underscored the imperative for a universal, cost-effective approach.

A universal approach doesn't obviate variation. Were there any surprises in how AAV9 behaved compared to affinity workflows?

Rosario: Any universal methodology must contend with inherent serotype variability. However, SXC offers the advantage of permitting similar buffer formulations across multiple serotypes, thereby enhancing operational flexibility. With AAV9, we observed predictable variations, yet SXC consistently supported scalable and robust performance, circumventing many of the idiosyncrasies that challenge affinity resin-based workflows.

AEX can struggle to distinguish full from empty capsids at scale. Did you see this with AAV9? If so, how did you mitigate it?

Rosario: Traditionally, ultracentrifugation was viewed as the gold standard for full/empty capsid separation, though its scalability is inherently constrained. Contemporary chromatographic techniques—ranging from mixed-mode to weak and strong AEX resins, including monolithic supports—now achieve separations on par with ultracentrifugation, yet are more amenable to scale-up. In our hands, these “different ice cream flavors,” as I like to call them, have consistently rendered AEX a preferred strategy, particularly as we transcend the limitations of traditional ultracentrifugation. With AAV9, appropriately chosen chromatographic conditions allowed us to effectively resolve full and empty populations without significant compromise.

The report notes polyethylene glycol residues. Is buffer exchange or ultrafiltration required to address that?

Rosario: Indeed, residual PEG is a consideration inherent to SXC-based workflows. Our standard protocol incorporates tangential flow filtration, combining diafiltration and ultrafiltration, to effectuate buffer exchange and achieve the desired final concentration. This ensures minimal PEG carryover and preserves product quality.

From your perspective, is this workflow ready for GMP use? If not, what's holding it back?

Rosario: The workflow is nearing readiness for GMP manufacturing; however, it is still an R&D project and not commercialized. The prototype is available for testing upon request. Its robustness at the largest operational scale — namely, the 2-L capsule system — must still be validated. This critical step is scheduled for evaluation in the coming months and will inform the readiness for potential full GMP deployment.

About The Expert:

Juan-Carlos Rosario is a senior principal scientist, purification and virology department, at Eli Lilly and Company where he develops optimized downstream purification processes for mAbs, peptides, and AAVs. Previously, he worked in roles at Merck; Janssen, Johnson & Johnson Innovative Medicines; and Amgen. He is finishing his Ph.D. at IU Indianapolis. His thesis focuses on developing gene therapies (AAV) to rescue cognitive and motor skills phenotypes in Ts65Dn mouse Down syndrome models. Contact him on LinkedIn