Diatomaceous Earth Filtration Scales Up Viral Vector Harvest for Gene Therapy

Gene therapy has moved decisively from promise to practice, with recombinant adeno-associated viral (rAAV) vectors now serving as the dominant delivery vehicles for treatments that replace, silence, or add genes in a single intervention. Yet the industrial machinery behind these medicines has struggled to keep pace with demand. A new study published in Applied Microbiology and Biotechnology by researchers at Roche Diagnostics GmbH in Penzberg, Germany, reports a practical solution to one of the most stubborn bottlenecks in vector manufacturing: the harvest step, where freshly produced viral particles must be separated from the debris of the host cells that made them. The team demonstrates that alluvial filtration, a technique borrowed from older industrial filtration practice, can clarify rAAV-containing cell lysate at scales up to 500 liters while keeping the entire process enclosed within single-use equipment.

The significance of the harvest step is easy to underestimate. In a typical manufacturing train, human embryonic kidney 293 (HEK293) cells are cultivated in bioreactors and transfected to produce viral vectors, after which a chemical lysis step releases the particles from the cells. What remains is a murky soup of viral vectors, host cell proteins, host cell DNA, and fragmented cellular material. Before chromatography-based purification can begin, this crude lysate must be clarified, and the efficiency of that clarification determines how much product survives and how smoothly downstream operations run. At laboratory scales of up to roughly 10 liters, centrifugation handles the job well, but single-use centrifuge technologies cannot deliver the required centrifugal forces at production scale, forcing manufacturers to rely on filtration instead.

Depth filtration has long been the industry standard for this task. These filters contain a porous multilayer matrix of fibrous material and filter aids such as diatomaceous earth, and they are valued for their high retention capacity and their ability to remove not only cell debris but also meaningful quantities of host cell proteins and DNA. However, the Roche team identified a critical weakness when they systematically screened ten depth filters or filter combinations using lysate from rAAV8-producing cultures in 1 to 3 liter test volumes. Capacity values ranged widely, from 20 to 300 liters per square meter of filter area, with turbidity reductions generally exceeding 90 percent for the better performers. The synthetic Millistak+ HC Pro D0SP filter emerged as the clear favorite, achieving approximately 300 liters per square meter, followed by Clarisolve 20MS at roughly 240 and Millistak+ HC Pro C0SP at 120 liters per square meter.

The problem lay not in performance but in geometry. The largest filter area available per fully closed single-use capsule of the best-performing depth filter was limited to 0.027 square meters. Scaling calculations for a 500 liter harvest revealed that closed processing would require roughly 50 such lab-scale capsules, an impractical proposition, while the cassette-based alternative that provides adequate area requires a multi-use filter holder whose disassembly cannot be performed under closed conditions. This matters because rAAV vector variants classified as risk group 2 fall under German biosafety regulations that prescribe a fully closed system whenever no separate space exists for the disassembly and cleaning of equipment outside the production area. A fully closed system, as defined in the study, fully isolates cells and viral vectors from the environment throughout processing.

Alluvial filtration, also known as body feed filtration, offered a way around this constraint. Rather than relying on a fixed filter matrix alone, the technique adds diatomaceous earth directly to the feed, where it forms a porous, incompressible cake on the filter sheet that acts as the working filtration medium. This cake resists blockage, supports high flow rates of around 300 liters per square meter per hour or more, and enables efficient impurity removal. The Roche team, building on earlier small-scale work conducted at merely 50 milliliters by other groups, optimized the method across combinations of filter sheets, diatomaceous earth grades, and dosing ratios. The winning configuration paired the PURAFIX CH 31 filter sheet with Celpure C 300 diatomaceous earth at a ratio of 2 grams of filter aid per 3 grams of biological wet mass, the latter determined gravimetrically from centrifuged samples of each feed.

In head-to-head small-scale comparisons using 1 to 3 liter feeds, the optimized alluvial method matched the best depth filter in capacity at approximately 300 liters per square meter while surpassing it on nearly every other measure. Alluvial filtration achieved complete removal of host cell DNA, a 75 percent reduction in host cell protein, 97 percent turbidity reduction, and product recoveries of 95 percent for viral genomes and 97 percent for capsids. The absolute turbidity after filtration was slightly higher than with the leading depth filter, 63 versus 40 nephelometric turbidity units, but remained comfortably within the range suitable for subsequent sterile filtration. The method also operated at a recommended flux of 480 liters per square meter per hour, more than three times the depth filter recommendation, translating into substantially shorter processing times.

The real test came at scale. The researchers processed lysate from bioreactors at 10, 100, and 500 liters, spanning process development, initial supplies, and pilot-scale study material. The runs covered three rAAV variants, rAAV2, rAAV8, and rAAV9, providing a stress test of robustness across different vector serotypes and feedstock compositions. Average filter capacities came in at 223, 256, and 229 liters per square meter for the three scales respectively, a remarkably consistent band. Product recovery and turbidity reduction stayed high at every scale with no decline at 500 liters, and standard deviations remained acceptable. Impurity reduction for host cell DNA and protein showed larger variability, a deliberate trade-off in favor of capacity, which the authors judged acceptable since robust lysate processing, not maximal impurity clearance, is the critical objective at this initial stage.

Pressure profiles from a representative 500 liter run told a story of a process operating well within its limits. Pressure before the alluvial filter rose to 1.15 bar at the start of filtration and then held nearly constant, while downstream pressure stabilized at 0.75 bar and flow declined only marginally from 12 to 10 liters per minute, a drop attributed to the centrifugal pump rather than filter clogging. The entire 405 liter filtration finished in approximately 30 minutes, matching the calculated projection, and pressure never approached the 1.5 bar safety limit. The filter cake volume also stayed below the module maximum, indicating that the system’s full capacity was never reached and that the process could tolerate even higher-turbidity feedstocks. A sterile filtration stage downstream of the alluvial filter performed without blocking or product loss, confirming the filtrate was fit for purification.

Safety engineering rounded out the design. Because filtration trains are generally classified as high-risk unit operations prone to equipment failure and pressure buildup, the pilot-scale setup incorporated single-use pumps, sterile couplings, and pressure-rated tubing, along with continuous monitoring by pressure and flow sensors. If system pressure reached the 1.5 bar maximum, flow was automatically reduced to prevent exceedance. Before any viral vector-containing fluid entered the train, all connections were checked, filters were vented during a water pre-rinse, and the housing and connections were validated by a water pressure test. Disassembly was performed by tube welding and sealing, allowing the used filter to be removed and autoclaved for viral inactivation without ever opening the system.

The study’s authors conclude that alluvial filtration meets both the performance and biosafety requirements that the best available depth filter could not satisfy at pilot scale, offering up to 2.71 square meters of filter area in a single closed module where the leading depth filter capsule provides only 0.027 square meters. The method does demand roughly an hour of additional preparation to measure biological wet mass and suspend the filter aid, but this is more than compensated by filtration times cut more than threefold. As demand for rAAV-based gene therapies continues to climb, the ability to harvest viral vectors safely, quickly, and at the 500 liter scale within a fully closed single-use train may prove a meaningful step toward industrial-scale production of these transformative medicines.

Subject of Research: Alluvial filtration for scalable, closed-system harvest clarification of recombinant adeno-associated viral vectors in gene therapy manufacturing

Article Title: Alluvial filtration as a scalable and safe single-use technology for the harvest of viral vectors

Article References: Wiesner, J., Bauer, A., & Eisenkraetzer, D. (2026). Alluvial filtration as a scalable and safe single-use technology for the harvest of viral vectors. Applied Microbiology and Biotechnology, 110(1), Article 293. https://doi.org/10.1007/s00253-026-14035-1

Image Credits: AI Generated

DOI: 10.1007/s00253-026-14035-1

Keywords: alluvial filtration, rAAV vectors, gene therapy, viral vector manufacturing, depth filtration, diatomaceous earth, harvest clarification, single-use technology, bioprocessing, HEK293 cells, pilot-scale production, biosafety