Human Embryonic Kidney cells 293T (HEK 293T) ACS-4500, adapted to suspension, were purchased from ATCC (Virginia, United States). These were routinely sub-cultured to 0.6 × 10⁶ cells mL⁻¹ every 48 h when cell concentration reached 2–3 × 10⁶ cells mL⁻¹ using vented non-baffled shake flasks with BalanCD HEK293 medium (Irvine Scientific, California, United States) supplemented with 4 mM of GlutaMAX (Gibco, Fisher scientific, Hampton, United States) under a humidified atmosphere of 5% CO₂ in air at 37°C with controlled agitation (orbital diameter of 25 mm, 90 rpm). The AAVs feedstock generated for this study was obtained in a 20 L Biostat^® D-DCU (Sartorius, Göttingen, Germany) stirred-tank bioreactor (STB) equipped with two Rushton impellers and a ring-sparger for gas supply. The pO₂ was set to 40% of air saturation and was maintained by varying the agitation rate (70–200 rpm), the percentage of O₂ in the gas mixture (0%–100%), and the gas flow rate (0.01–0.04 vvm). The pH value was maintained by the automatic addition of either 1 M of NaHCO₃ or CO₂ within the gas mix. Cells were transfected with a DNA plasmid solution containing 1.5 µg of total plasmid DNA per 10⁶ cells. This mix included pDP 5 (reference: PF0435) and p-AAV-ssGFP (reference: PF1451), at a molar ratio of 1:1 (PlasmidFactory, Bielefeld, Germany), diluted in a specific volume of supplemented culture medium, corresponding to 5% of culture volume. Additionally, PEI MAX (PolySciences, Bergstrasse, Germany) transfection reagent was added with a 1:2 µg DNA/µg PEI ratio between total plasmid and reagent. This solution was incubated at room temperature for up to 15 min before addition.

Cells were lysed 72 h after transfection with 50 mM Tris-HCl, 1.0 (v/v) % Tween 20, and 2 mM of MgCl₂, at pH 8, followed by the addition of 50 Units per mL of Benzonase (catalog number: 1.01656.0001, Merck Millipore, Darmstadt, Germany). To prevent aggregation, salt-concentrated solutions of MgSO₄ and NaCl were supplemented to a final concentration of 37.5 mM and 200 mM, respectively. The cell lysate was harvested and clarified with two different filter trains. The first consisted of a depth filter capsule with 0.2 m² of the effective filtration area and retention range between 6 and 30 μm (Supracap™ 100, Pall Corporation, New York, United States) followed by a second filtration stage with a Gamma–Irradiatable MidiCap filter of 0.45 m² surface area and 0.2 μm of pore size (Sartopore^® 2, Sartorius Stedim, Göttingen, Germany) operated at flow rate of 600 mL min⁻¹.

The chromatography experiments were performed at room temperature of 20°C using an ÄKTA avant 25 (Cytiva, Uppsala, Sweden). A device of 0.4 mL of volume, consisting of a cellulose-based fiber matrix with geometry and flow properties similar to the commercially available HiTrap Fibro™ PrismA was provided by Cytiva (Uppsala, Sweden) (Hardick et al., 2013; Dods et al., 2015; Cytiva, 2020b). The fiber matrix is functionalized with an affinity ligand designed for purifying AAVs serotypes, including 1, 2, 3, and 5, similar to the commercially available affinity chromatography resins Capto AVB or Sepharose AVB (Cytiva, 2018; Cytiva, 2020a). The device was equilibrated with 15 column volumes (CV) of 20 mM Tris-HCl (Merck KGaA, Darmstadt, Germany), 500 mM NaCl (Merck KGaA, Darmstadt, Germany), and pH 8. After sample loading, the unit was washed with 30 CV of equilibrium buffer. The AAVs were eluted with 40 CV of 100 mM Glycine (Sigma-Aldrich, Missouri, United States), pH 2.5 and neutralized with 1 M Tris-HCl (Merck KGaA, Darmstadt, Germany), pH 9.

The prediction of breakthrough curves and adsorber capacity is frequently performed recurring to Thomas’s model, which assumes negligible external and internal diffusion resistances, Langmuir kinetics of adsorption-desorption and second-order reversible reaction kinetics (Atar et al., 2011; Futalan et al., 2011; López-Cervantes et al., 2018; Patel, 2019).

The linearized form of the Thomas model is as follow: ln C 0 C t − 1 = k T h q 0 w Q − k T h C 0 t (1)

Where C₀ and C_t are the column inlet and outlet AAVs concentration at time t, k_Th is Thomas rate constant, q₀ is the saturation capacity of the device, w is the mass of the adsorbent and Q is the flow rate. Eq. 1 was modified to take into consideration the volume of the adsorber, replacing w with the packed nanofiber volume (0.4 mL). To keep Eq. 1 dimensionally consistent, q₀ should therefore be expressed in terms of adsorbed AAVs per unit volume of packed nanofiber adsorber.

Permeability of the nanofiber adsorber can be obtained by the linear fitting of the superficial velocity versus ∆ P/L (Herigstad et al., 2015; Lemma et al., 2021), according to Darcy’s Law: v = k μ ∆ P L (2)

Where v is the superficial velocity, ∆ P is the pressure drop along the adsorber height (L), µ is the viscosity of the fluid, and k is the flow permeability of the porous medium.

Pressure drop in chromatography media limits the superficial velocity and consequently flow rate, affecting process productivity. The nanofiber adsorber permeability, using water and running buffer, were determined under a range of superficial velocities between 57 and 170 cm h⁻¹ (corresponding to flow rates of 5.0–15.0 mL min⁻¹). Additionally, pressure drop across the chromatographic support was also evaluated using clarified AAV5 material and compared against water and equilibration buffer. The linear relationship between pressure drop and flow rate, demonstrated in Supplementary Figure S1, indicates, not only, a laminar flow regime, but as well, the stability of the porous nanofiber matrix by not contracting at higher flow rates for water and equilibration buffer (Mihelič et al., 2005). Under these circumstances, the Darcy law is valid, and the permeability can be obtained through a linear fitting of the superficial velocity as a function of the pressure drop along the unit length. As the equilibrium buffer presents a low solid concentration, it was assumed that its viscosity is equal to the water, 10^–3 Pa s (Ho and Zydney, 2000). Figure 1A shows the linear relation between ΔP/bed height (ΔP/L) versus the superficial velocity; the slope was used to calculate the flow permeability for water, 2.65 × 10⁻¹⁵ m² and for buffer, 2.55 × 10⁻¹⁵ m². The values obtained are in the same order of magnitude as the values obtained by Trilisky et al. for CIM DEAE disks (6.30 × 10⁻¹⁵ m²) (Trilisky et al., 2009), Herigstadt et al. for CIM protein A disks (5.74 × 10⁻¹⁵ m²) (Herigstad et al., 2015) and Lemma et al. for PBT membranes (2-3×10⁻¹⁵ m²) (Lemma et al., 2021).

The chromatographic purification of biological materials has to deal with colloids, such as lipids, cell debris, and carbohydrates, as well as soluble components, such as proteins and nucleic acids. Although cells and cell debris are usually removed during clarification, the remaining components can adversely affect adsorption performance, especially in cases of prolonged exposure, as in the case of a capture step. These components can occupy the interparticle pore space by competing for the binding groups or by blocking the pores of the chromatography medium. Consequently, this can be manifested as reduced resolution, reduced adsorption capacity, or increased pressure drop, all of which affect the efficiency of the purification step (Siu et al., 2006). The impact of the loading volume on pressure drop during the capture step was evaluated using AAV5 clarified material. The pressure drop was monitored at different superficial velocities (57–198 cm h⁻¹) and is reported in Figure 1B as averages over 100 mL loading intervals. These were used to calculate the variation of the group k µ⁻¹ of the Darcy equation over the loading of more than 2100 CV. As seen in Figure 1C, k µ⁻¹ decreases over the continued loading. As the viscosity of the AAVs feedstock solution remains constant over the loading volume, this decrease can therefore be explained by a reduction in the permeability of the nanofiber adsorbents as a consequence of such high loading volumes. Considering that the amount of loaded material is approximately twice the highest value calculated for the dynamic binding capacity of the unit, as demonstrated in the next section, the observed variation in permeability value is lower than 40%.

To assess the dynamic response, the shape and time of breakthrough curves were obtained from frontal experiments performed with AAV5 clarified samples at a feed concentration of 5.45 × 10¹¹ TP mL⁻¹. The collected samples during the unit loading were analyzed using ELISA. The breakthrough curves generated with experimental data were fitted simultaneously (Figure 2) using the Thomas model. This model is applied for adsorption processes where external and internal diffusion limitations are negligible (Futalan et al., 2011), and can be suitable for convective materials with some degree of approximation. In this sense, the saturation capacity and Thomas kinetic constant were assumed as independent of the residence time. The model fitting in the range of residence times scoped allowed to determine the saturation capacity of the nanofibers (q₀) of 8.29 × 10¹⁴ TP mL⁻¹. Additionally, the modelling of the breakthrough curves also allows the estimation of the unit’s dynamic binding capacity (DBC), which is used as a measure of the performance of the chromatographic material. The estimated DBC for a percentage of 10% of the breakthrough (DBC_10%) at different residence times is reported in Figure 3, where for longer residence times the estimated DBC_10% approaches q₀. This is corroborated in Figure 2, where the breakthrough curves move to earlier elution times as the residence time is decreased. The same effect in the breakthrough time would be observed if the inlet concentration is increased for a given flow rate, assuming that the nanofibers explored are dominated by convective flow, with presumably negligible resistance to mass transfer (Sridhar, 1996; van Beijeren et al., 2012).

The load volume effect in the AAVs elution recovery was tested using a set of chromatographic runs where the load volume was varied (100–850 mL or 250–2125 CV) for a fixed residence time (2.4 s). The results are reported in Figure 4. As observed, the elution recovery varies from 85% to approximately 40% with the load volume increase. This is in line with the data of Figure 3, where for a constant residence time, the continuous increase in the load volume leads to exceeding the DBC_10% at that point. As a result, this surplus is no longer adsorbed to the nanofiber and the final recovery decreases.

In the second stage, to analyze the effect of residence time in the elution recovery, this parameter was varied from 4.8 to 1.2 s using a fixed loading volume of 250 mL (625 CV) of feedstock (Figure 4). In this case, the AAVs recovery varies between 77% and 20%, following Figure 3, for a fixed volume, as the residence time decreases, the dynamic binding capacity of the nanofibers decreases, reducing virus adsorption and consequently recovery. Additionally, the content in total protein and ds-DNA of elution samples was analyzed (Table 1), revealing a removal of around 90% for total protein and between 80–90% for ds-DNA. As expected, results show that overloading the nanofibers or decreasing residence time is reflected in flowthrough losses as adsorptive capacity is exhausted. This can be minimized with applications such as continuous chromatography where media saturation is desired and loading zones composed of several interconnected adsorptive beds are used (Araújo et al., 2007; Silva et al., 2015; Mendes et al., 2021).

The reproducibility of the purification was evaluated at a residence time of 4.8 s, using new nanofiber adsorbents for each run (total of 3). The RT was selected based on Figure 3 where for RT below 4.8 s DBC would be greatly affected. Conversely, higher RT would not significantly improve DBC, as DBC tends to a constant value. All the procedures were automated, and the sample volume injected corresponds approximately to the calculated value of DBC_10%. The chromatograms are depicted in Supplementary Figure S2.

The three chromatographic experiments show a similar UV absorption profile. The eluates were analysed by ELISA to determine AAVs recovery. The average value obtained, 71.3% is following the previous observations of the impact of residence time on loading volume (Figure 4). The sample loading of each run varied △P from 0.05 MPa (initial) to 0.08 MPa (end). These values are in accordance with what is observed in Figure 1C for the same flow rate. The quantification of flowthrough pools of each run was close to the lower detection limit of ELISA method used and highly prone to quantification errors. Considering the elution recoveries obtained and the possible low concentration of AAVs in flowthrough pools it is plausible that some of the AAVs adsorbed are only removed from the nanofibers during harsher elution conditions or cleaning in place.

The eluates were further characterized by SDS-PAGE and western blot to assess purity and identity. Additionally, transmission electron microscopy (TEM), HPLC-SEC and DLS were also performed to assess capsid morphology, the presence of aggregates, particle size and size distribution. The results are presented in Figures 5, 6; Table 2. Finally, the reduction in total protein and ds-DNA was analysed during the chromatographic step (Table 1).

The main AAVs capsids proteins, VP1, VP2, and VP3 are present in SDS-PAGE (Figure 5A), for each eluate and its identity is confirmed by western blot (Figure 5B), with the expected proportion of 1:1:10 (VP1:VP2:VP3). No visible impurities are seen in the SDS-PAGE. TEM images (Figure 5C) show that the size, shape, and capsid integrity of AAVs particles are kept between runs. The analysis of size exclusion chromatography (Figure 6A; Table 1) reveals two peaks. The first one (smaller) with a relative area of 2% at 9.0 min is indicative of the presence of aggregates, although this is not visible in TEM analysis. The second peak, with a retention time of 12.0 min and a relative area of 98%, corresponds to the AAVs purified. The observable difference in peak height of the HPLC-SEC chromatograms (Figure 6A) is explained by the different volumes of buffer added to dilute and neutralize the elution pools. This dilution is also noticeable in the TEM analysis (Figure 5). Regarding the characterization by particle size distribution (Figures 6B–D), the eluted AAVs have an average diameter of 26 nm. The differences observed in the peak shape, between intensity and number distributions, are due to the presence of large particles (aggregates) above 1 μm, that cause a peak extension, confirming the SEC results. Regarding the removal of total protein and ds-DNA (Table 1), fibro affinity units allowed a substantial reduction of 78–79% of total protein and 80–90% of ds-DNA.

The nanofiber adsorbents evaluated for AAV5 affinity capture showed that it is possible to operate at flow rates as high as 15 mL min⁻¹ (RT = 1.6 s), with pressure drops lower than 0.4 MPa. On one hand, these values are in accordance with those reported for affinity purification of monoclonal antibodies with similar cellulose-based nanofiber adsorbents (Cytiva, 2020b). On the other hand, the nanofiber adsorbents for AAVs allow 10 times higher operational flow rates in comparison to packed beds of resin with 2.5 times its volume (Cytiva, 2018).

The nanofiber adsorbents reveal a good dynamic binding capacity and recovery yields even at low residence times in comparison to packed beds. Mendes et al., using periodic counter-current chromatography, reported an AAVs recovery of 82.8% using Capto AVB™ of 0.2 mL, with 10-fold higher residence times in comparison to the used nanofibers (Mendes et al., 2022). With respect to the experimental binding capacity, the highest value (6.08 × 10¹⁴ TP mL⁻¹) was obtained for a residence time of 4.8 s. Although capacity is in the same order of magnitude as what is reported by some manufacturers there is a substantial improvement in diminished residence times Cytiva, 2018; Cytiva, 2020a; Fisher Scientific, 2017; Repligen Corporation, 2022). To the present moment there are no reports of convective materials functionalized with affinity ligands for AAVs purification. However, a great effort has been made in the development of new ligands for different AAVs serotypes (Zhao et al., 2019; Chevrel et al., 2022). Thus, the cellulose-based nanofibers adsorbents reported here, arise as the first convective adsorber for AAVs affinity purification, enabling a meaningful removal of impurities even processing a large volume of AAVs-clarified material with a single chromatography step. To enable the application of the nanofiber advantages to the larger-scale of AAVs production, the scale-up of the device of 0.4 mL is under development using the same principles of HiTrap Fibro™ PrismA (Cytiva, 2020b).