Chromatography is a key part in the AAV manufacturing process. The egocentric chromatographer might view the clarification and concentration steps, post-bioreactor harvest, as merely preparation for chromatographic separations! However, clarification removes cells and cell debris, lowering the turbidity to make the feed stream compatible with chromatography. This prepares the way for chromatographic separation, but as AAV concentrations out of the bioreactor can be relatively modest compared to resin capacity, it is common practice to also concentrate the feed stream via tangential flow filtration (TFF). This reduces the load volume onto the first chromatography step, often by 60-fold or more, subsequently reducing the loading time.
Routinely, two chromatographic steps are employed. A capture step to remove the bulk of the contaminants including host cell proteins (HCP) and DNA, followed by an anion exchange step to further reduce those contaminants, HCP and DNA, along with a reduction in the number of empty capsids that do not contain the therapeutic DNA payload.
Affinity chromatography
Before the advent of affinity ligands optimized for AAV, purification was performed over two or three steps, using a range of chromatography methods including ion exchange, hydrophobic interaction (HIC), and size exclusion (SEC).
The introduction of affinity resin for AAV, such as Capto™ AVB from Cytiva, has simplified process development. Often the affinity step results in greater than 99% pure AAV in a single step with very minimal process development. Optimization can be performed around the loading step. Higher loadings may result in higher yield, due to relatively low losses attributed to non-specific binding. Additionally, higher loading may result in higher purity, perhaps by blocking binding sites for non-specific interactions.
Elution is another step that can be optimized. In my experience, I have found that pH 3 has not been sufficient to fully elute AAV from affinity resins, so lower pH values could be worth experimenting with. However, the main parameter to optimize for AAV affinity is regeneration. To date, affinity resins have depended upon camelid derived single chain antibodies. These are not yet sodium hydroxide stable, so regeneration of the resins can be challenging, even for as few as 5-10 cycles. In fact, for process development, it is probably more effective to use the 1 mL or 5 mL columns a single time.
Anion exchange chromatography
Following affinity chromatography, a second chromatography step is required, for several reasons including the removal of specific contaminants such HCP, DNA, and perhaps adventitious viruses. Additionally, this step is the only one in the whole downstream purification process that can enrich for capsids containing the gene of interest, the so-called “full” capsids. Anion exchange chromatography (AEX) has proven to be effective in meeting all of these challenges. The biggest challenge, however, is removal of empty capsids that do not contain the DNA payload. This is difficult because the empty and full capsids are so similar.
They do have different densities, which can be exploited by ultracentrifugation, but as discussed in a previous blog, this is not a scalable manufacturing method. The encapsulated DNA does, however, confer an additional net negative charge to full capsids, and it is this that can be exploited for separation via anion exchange chromatography. This is fortuitous, as AEX, by nature of its positive charge, can effectively remove DNA, HCP, and viruses which are negatively charged. The separation and removal of empty capsids by AEX is still very challenging. It requires significant optimization, and this may be impacted not just by serotype, but also by the encapsulated gene of interest. Linear elution gradients have been shown to separate empty and full capsids, but experience shows it is a struggle to get linear gradients to readily give enough resolution. Small conductivity steps of around 1 mS/cm can be a handy process development tool that can enable easy assessment of the separation. This technique has brought success with Mustang™ Q membrane chromatography in making ~5-fold enrichment of full capsids for serotypes 5, 8, and 9. The separation here was elution-based and as such, conductivity increases and the amount of membrane or column volumes of the steps could be optimized along with pH, elution buffer, elution salt species, and excipients such as MgCl2 and poloxamers, to improve purification.
This type of success clearly provides a framework for further optimization and development. Perhaps, when moving from process development into a GMP manufacturing setting, it will be desirable to elute in two or three discrete steps, and the multiple optimization steps described can guide us in that endeavor. It is clear from the small steps that the discrete capsid species elute based upon differences in their conductivity. This information is not so clearly available with linear gradient experiments.