Gene therapy's bottleneck is a chromatography adsorbent, not a clinical pipeline

A grant awarded in May by NIIMBL to two North Carolina State University engineers is aimed at the least glamorous step in gene therapy manufacturing: the chromatographic capture of adeno-associated virus vectors ^[8]. Stefano Menegatti, a professor in NC State's Department of Chemical and Biomolecular Engineering, told GEN that limited availability of viral vectors has become a major bottleneck for the sector, and that the shortage is a downstream materials problem ^[1][2].

The specific defect he names is selectivity. Current purification materials, usually resin-based affinity adsorbents, cannot distinguish full AAV capsids carrying genetic payload from empty ones that carry nothing ^[4]. AAVs are the leading delivery vehicle for these therapies ^[3], so a capture step that is blind to payload is a capture step that spends capacity on product that cannot dose a patient. By Menegatti's own figures, raw material runs roughly 20 to 30 percent full capsids ^[9], which means 70 to 80 percent of what the column sees is packaging with no gene inside ^[15].

Two further constraints are economic rather than analytical. According to Menegatti, these technologies operate at slow flow rates, require harsh chemical conditions that can damage the product, and wear out quickly, all of which raise manufacturing time and cost, and ultimately price ^[5][6]. Flow rate and cycle life are not clinical variables. They are properties of a material, and they set how many doses a given suite of columns can turn out per year.

The claimed remedy is a membrane-format adsorbent. Menegatti says AvXcel affinity adsorbents, developed by ChromaGenix, accelerate purification and enrich full capsids at the capture step, with preliminary data showing the full fraction rising from roughly 20 to 30 percent in raw material to 34 to 48 percent in the affinity eluate, which he describes as far outperforming the industry benchmark ^[7][9][10]. On matched endpoints that is an enrichment factor of about 1.6 to 1.7 ^[16]. It is also, on the same numbers, an eluate in which every full capsid is still accompanied by between roughly 1.1 and 1.9 empty ones ^[17]. He also says the membranes tolerate up to 50 caustic cleaning cycles, in line with industry needs ^[11], and characterises the combination of speed, selectivity and durability as "genuinely unprecedented," with the goal of compressing months-long process development into weeks ^[12].

The second half of the grant addresses the tuning problem. Michael Daniele, Menegatti's collaborator, says the standard approach requires dozens to hundreds of experiments per new AAV serotype and transgene combination, repeated essentially from scratch each time ^[13]. Their platform, Beacon, uses a Gaussian Process Bayesian algorithm to pick the next most informative experiment, which Daniele says cuts experimental burden by 30 to 50 percent against conventional design of experiments while optimising yield, full-capsid enrichment, impurity clearance and productivity together ^[14]. It warm-starts from North Carolina's VVIRAL database of thousands of AAV purification experiments and reports through SHAP analysis rather than as a black box ^[18].

Note what the source material does and does not contain: every performance figure here is a quote from the two investigators, and the material includes no published dataset and no number for the "industry benchmark" being outperformed ^[19]. AvXcel is a ChromaGenix product ^[7].

What to watch: whether the 34 to 48 percent enrichment holds across serotypes and at manufacturing scale rather than in preliminary runs, whether the 50-cycle caustic figure survives real cleaning-in-place schedules, and whether Beacon's 30 to 50 percent saving is ever measured against a documented DOE baseline.