Process development for cross‐flow diafiltration‐based VLP disassembly: A novel high‐throughput screening approach

Virus‐like particles (VLPs) are particulate structures, which are applied as vaccines or delivery vehicles. VLPs assemble from subunits, named capsomeres, composed of recombinantly expressed viral structural proteins. During downstream processing, in vivo‐assembled VLPs are typically dis‐ and reassembled to remove encapsulated impurities and to improve particle morphology. Disassembly is achieved in a high‐pH solution and by the addition of a denaturant or reducing agent. The optimal disassembly conditions depend on the VLP amino acid sequence and structure, thus requiring material‐consuming disassembly experiments. To this end, we developed a low‐volume and high‐resolution disassembly screening that provides time‐resolved insight into the VLP disassembly progress. In this study, two variants of C‐terminally truncated hepatitis B core antigen were investigated showing different disassembly behaviors. For both VLPs, the best capsomere yield was achieved at moderately high urea concentration and pH. Nonetheless, their disassembly behaviors differed particularly with respect to disassembly rate and aggregation. Based on the high‐throughput screening results, a diafiltration‐based disassembly process step was developed. Compared with mixing‐based disassembly, it resulted in higher yields of up to 0.84 and allowed for integrated purification. This process step was embedded in a filtration‐based process sequence of disassembly, capsomere separation, and reassembly, considerably reducing high‐molecular‐weight species.

preclinical and clinical trials, and recently, a vaccine against the circumsporozoite protein of the malaria pathogen has been locally approved (European Medicines Agency, 2015). Recent efforts to develop VLPbased severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) vaccines underline the flexibility and simplicity of chimeric VLPs (Ghorbani et al., 2020;Yang et al., 2021). However, the versatile platform of chimeric VLPs not only comes with promises, but also with fundamental challenges, such as the ability to form stable capsids (Borisova et al., 1999;Böttcher et al., 2006;Nassal et al., 2005;Pumpens & Grens, 2001). Challenges related to the production process include the limited solubility of candidate molecules (Jegerlehner, 2002;Karpenko et al., 2000;Vormittag et al., 2020a) and the dependence of process parameters on molecular properties, which are both influenced by the amino acid sequence of the inserted foreign antigenic epitope (Rüdt et al., 2019). A typical production process is shown in Figure 1a. Process steps that are influenced by the epitope insertion are, for example, precipitation, where a varying amount of ammonium sulfate is required for VLP precipitation (Hillebrandt et al., 2020;Vormittag et al., 2020b), and VLP reassembly, which has been proposed to depend on VLP zeta potential (Rüdt et al., 2019). VLPs are disassembled (dissociated into capsomeres) and reassembled (capsomeres triggered to form capsids) to improve structural homogeneity, stability, and immunogenicity (Zhao, Allen, et al., 2012;Zhao, Modis, et al., 2012). Disassembly is realized by a high pH and low ionic strength, often adding denaturants or reducing agents (Bin Mohamed Suffian et al., 2017;Mach et al., 2006;McCarthy et al., 1998;Singh & Zlotnick, 2003;Strods et al., 2015;Zhang et al., 2021), while reassembly is achieved at neutral pH and high ionic strength (Mach et al., 2006;Zlotnick et al., 1996). As disassembly releases capsid-internal impurities and typically leads to incomplete disassembly or aggregate formation, a capsomere separation step is added between dis-and F I G U R E 1 (a) Process of intracellularly produced, in vivo-assembled capsids, highlighting unique virus-like particle (VLP) process steps. Compared with other biotechnological products, upstream processing and primary purification are followed by a sequence of disassembly, capsomere separation, and reassembly. This sequence allows for the improvement of particle homogeneity and removal of encapsulated impurities. (b) Concept of a filtration-based VLP purification cascade. Capsids are disassembled by cross-flow diafiltration (DF) into a disassembly buffer while the capsomeres are retained by the membrane. Encapsulated impurities are released during disassembly and washed out if smaller than the membrane molecular weight cut-off. Non-disassembled capsids and potential aggregates are separated by a dead-end ultrafiltration step. The capsids are then reassembled in a second cross-flow DF step into a DF buffer that favors the assembled state HILLEBRANDT ET AL. | 3927 reassembly (Zlotnick et al., 1996). As a denaturant, urea has been investigated at several concentrations for hepatitis B core antigen (HBcAg) VLP disassembly into HBcAg dimers (capsomeres) (Singh & Zlotnick, 2003;Zhang et al., 2021), further referred to as dimers. These publications show that an increasing urea concentration leads to a more complete and rapid disassembly, while urea concentrations of ≥4 M resulted in protein denaturation, which, in turn, can lead to aggregation or the inability to reassemble. It, therefore, seems reasonable to use urea concentrations for disassembly which are high enough to maximize the dimer yield but do not result in protein denaturation. For chimeric VLPs, it is conceivable that the inserted epitope influences the optimal disassembly solution conditions. Unpublished results on chimeric HBcAg disassembly by our laboratory confirm these assumptions.
VLP disassembly is typically achieved by the addition of disassembly agents (Mach et al., 2006;McCarthy et al., 1998;Singh & Zlotnick, 2003;Zlotnick et al., 1996), mixing VLP product with a disassembly buffer (Bin Mohamed Suffian et al., 2017;Lee & Tan, 2008;Liew et al., 2012;Zhang et al., 2021), or by dialysis (Holmes et al., 2015;Strods et al., 2015). While mixing is fast, it leads to dilution. Dialysis does not significantly change the original concentration and has the capability to remove some of the (encapsulated) impurities through the dialysis membrane (Mach et al., 2006) but is a slow process (Phillips & Signs, 2004). In recent publications, we have demonstrated the utility of transferring VLP process steps, namely capture and VLP reassembly, to a cross-flow filtration (CFF) unit (Hillebrandt et al., 2020;Rüdt et al., 2019). Figure 1a shows that these two process steps frame the disassembly step, which is one of the reasons why it was presumed to be useful to transfer this process step to cross-flow diafiltration (DF). As CFF process development is time-and material-consuming, it can be accelerated if the optimal DF buffer composition is known before.
In this study, we developed a time-resolved high-throughput disassembly screening for (chimeric) VLP candidates. This approach aims to reduce the experimental effort to identify optimal disassembly conditions for different VLPs and for CFF process development. As candidate VLP material is scarce at an early stage of process development, it is highly desirable to develop a screening method which requires a small amount of VLPs. To this end, we developed a low-volume, fast, and accurate screening method that allows for the assessment of VLP dimer yield and kinetic data based on a high-performance liquid chromatography (HPLC) system. Disassembly of a C-terminally truncated HBcAg and a chimeric C-terminally truncated HBcAg VLP with a polyhistidine tag was investigated (referred to as Cp149 and VLP A, respectively). Optimal disassembly conditions were selected for DF process development. This integrated DF process step allows for disassembly and depletion of impurities simultaneously. Additionally, we show that the developed disassembly process step can be integrated into a filtration-based sequence of disassembly, dimer separation, and reassembly ( Figure 1b). The results presented in this study underline the influence of inserted peptides on the optimal conditions for disassembly and demonstrate the usefulness of the developed high-throughput screening method and its transferability to a filtration-based process.

| Materials, buffers, and VLPs
All chemicals were purchased from Merck KGaA, unless otherwise stated. Solutions and buffers were prepared with ultrapure water (PURE-LAB Ultra; ELGA LabWater). All buffers were pH-adjusted with 32% HCl or 4 M NaOH using a SenTix62 pH electrode (WTW) at a HI 3220 pH meter (Hanna Instruments). Solutions were filtered before use and analysis through 0.2 µm cellulose acetate (VWR) or Millex-GV 0.22 µm polyvinylidene fluoride filters (Merck Millipore), often with glass fiber pre-filtration (Minisart GF; Sartorius Stedim Biotech GmbH). The plasmid for Cp149, a C-terminally truncated HBcAg protein (amino acids 1 to149 (Zlotnick et al., 1996)), was generously provided by Prof. Adam Zlotnick (Indiana University). BioNTech Protein Therapeutics GmbH generously provided the chimeric HBcAg VLP A plasmid. VLP A was C-terminally truncated, contained an inserted epitope, and incorporated a C-terminal polyhistidine tag, similar as in (Schumacher et al., 2018). The Cp149 and VLP A protein dimers had a molecular weight of approximately 34 and 40 kDa, respectively. The 280 nm extinction coefficients were derived from the web-tool ProtParam (Gasteiger et al., 2005) and were 1.764 L·g −1 ·cm −1 for Cp149 and 1.558 L·g −1 ·cm −1 for VLP A. Concentrations were calculated using Beer's law and the 280 nm absorbance peak area derived from size-exclusion chromatography (SEC) HPLC. The HBcAg concentration of the VLP feedstock was determined analogously using the total 280 nm absorbance and a NanoDrop 2000c spectrometer (Thermo Fisher Scientific). HBcAg was expressed in Escherichia coli, liberated by lysis, precipitated and re-dissolved applying a centrifugation protocol similarly as described in a recent article (Hillebrandt et al., 2020). Additionally, the re-dissolved and sterile-filtered VLP solution was purified by DF and multimodal SEC as described in the Supporting Information Section S1.

| Disassembly buffer compositions
The disassembly time series (DisA-TS) is a two-step procedure and described in detail in Section 2.3. Briefly, a batch disassembly reaction is followed by SEC-HPLC analysis under the same liquid phase conditions as in the batch reaction. This batch reaction is initiated by mixing equal volumes of HBcAg VLP solution and disassembly buffer to reach the target disassembly conditions after mixing. For all conditions, the common target concentrations were 50 mM Tris and 1 g·L −1 HBcAg. Urea concentrations (c urea ) and pH were screened in the ranges from 0 to 4 M and pH 7.2 to 9.0, respectively. The selection of the condition ranges was based on unpublished pre-experiments and other publications investigating HBcAg VLPs (Schumacher et al., 2018;Singh & Zlotnick, 2003). Each target condition required a distinct composition of the added disassembly buffer which was composed of Tris, urea, and titrant. Therefore, the required titrant concentration (c titrant ) of each disassembly buffer was determined to eventually reach the target pH after mixing with VLP solution. The procedure of this disassembly buffer composition determination is explained in Figure 2a. It was assumed that the protein buffer capacity of HBcAg at a concentration of 1 g·L −1 is negligible. Under this assumption, 50 mM Tris at pH 7.2 was used to mimic the VLP solution and thereby minimize the VLP (product) consumption. The experiments were performed in duplicates and at a 200 mL scale to minimize pipetting errors. The results were exemplarily confirmed by mixing VLP solution and disassembly buffer at a 200 µL scale and measuring the resulting pH using an Orion PerpHecT ROSS combination microelectrode (Thermo Fisher Scientific).

| DisA-TS and disassembly on column (DisA-OC)
The DisA-TS is a hybrid disassembly screening in nature. It consists of a batch disassembly reaction followed by SEC-HPLC analyses.
During the latter, the disassembly reaction continues until detection, similar to the on-column disassembly reaction described below. The time series was started at time t 0 by adding 250 µL disassembly buffer to 250 µL VLP solution in a 2.2 mL deep-well plate (VWR) in intervals of 7.5 min between each well. The mixtures were incubated at 23°C and repeatedly analyzed by SEC-HPLC over a period of 24 h at times t i (Figure 2b). Analytical SEC was performed with 20 µL injections on an AdvanceBio SEC 300 Å, 4.6 × 150 mm, 2.7 μm column (Agilent Technologies) at a Dionex Ultimate 3000 RS UHPLC system with a diode array detector controlled by Chromeleon version 6.8 SR15 (Thermo Fisher Scientific). The mobile phase was adapted to the sample's target disassembly condition, and a flow rate of 0.35 mL·min −1 was applied. Samples at pH 9.0 were analyzed at pH 8.0 due to the limited pH compatibility of the column. The efficiency of a disassembly condition was described by the (total) dimer yield is the total dimer concentration at the time t after start of the disassembly at t 0 . Furthermore, c t ( ) VLP 0 and c t ( ) dimer 0 are the mean initial concentrations of HBcAg VLPs and dimers, respectively, which are present in the VLP feedstock prior to the disassembly reaction. Mean concentrations from 28 samples of the VLP solution at 1 g·L −1 HBcAg were determined by SEC-HPLC as described above but with 50 mM Tris at pH 7.2 as mobile phase. During the 24 h DisA-TS, evaporation was observed. To estimate an average volumetric evaporation rate, the total HBcAg concentration of a VLP solution was analyzed twice with an interval of 20 h. The mean evaporation rate was then converted into evaporation correction where c t ( ) dimer,tot is the measured concentration during the DisA-TS.
A derivation of f t ( ) v and Equation (2) can be found in Appendix A.
To investigate the disassembly reaction with a shorter time interval and without prior mixing or batch disassembly, the abovementioned SEC-HPLC method was used as an additional screening tool. This approach is in the following referred to as DisA-OC and shown in Figure 2c. To this end, a VLP solution with 1 g·L −1 HBcAg (50 mM Tris, pH 7.2) was analyzed by SEC-HPLC applying the same flow rate and mobile phases at target disassembly conditions as described above. For our SEC-HPLC setup, we determined the ob- The disassembly rate k 0 was calculated by In the following, we define k 0,DisA TS as the rate obtained by DisA-TS r e t and k 0,DisA OC as the rate obtained by DisA-OC at r e t , where t 1 is the time of the first sampling after approximately 30 min.

| Filtration-based disassembly, dimer separation, and reassembly
The sequential process of disassembly, dimer separation, and reassembly was realized in three steps as shown in Figure 1b. Step (I) was DF-based disassembly into a disassembly buffer using a 10 kDa molecular weight cut-off (MWCO), 88 cm 2 Ultracel Pellicon 3 membrane (Merck Millipore) followed by an 18 h overnight hold at 5°C and subsequently by filtration through a 0.2 µm pore size cellulose acetate syringe filter (VWR). Note, that urea concentration and pH of the DF disassembly buffer were at target disassembly conditions. Step (II) consisted of dimer separation by dead-end ultrafiltration using Vivaspin Turbo 15 RC centrifugal filters with 100 kDa MWCO regenerated cellulose membranes (Sartorius Stedim Biotech GmbH). The disassembly solution was split into six centrifugal filters, which were operated at a relative centrifugal force of 1000 for 15 min, and the product was collected in the filtrate/ permeate. The remaining retentate was equally reprocessed with a new filter.
Step (III) was DF-based reassembly into 50 mM Tris buffer at pH 7.2 with 650 mM NaCl using a 10 kDa MWCO, 200 cm 2 Sartocon Slice 200 (Sartorius Stedim Biotech GmbH) where the product was collected from the retentate. Both DF process steps were realized on a KrosFlo Research KRIIi CFF system (Spectrum Labs) at a constant volume of 30 mL, a feed flow rate of 30 mL·min −1 , and permeate flow rate control at 2 mL·min −1 as implemented previously (Hillebrandt et al., 2020). The corresponding permeate flux setpoints were 13.6 and 6.0 L·m −2 ·h −1 for disassembly and reassembly, respectively. The whole process was performed at room temperature The dimer concentrations after disassembly/hold (c dimer,DisA ) and after dimer separation (c dimer,Sep ) represent the bulk (feed) and permeate concentrations, respectively.

| RESULTS AND DISCUSSION
DisA-Ts: The long-term development of the disassembly reaction was monitored using an initial batch disassembly reaction followed by 12 influences the pK a of weak acids/bases, such as Tris, and ultimately the solution pH (Beynon & Easterby, 1996). Therefore, the pH does not change linearly with the volume shares upon mixing of VLP solution and disassembly buffer. This nonlinear behavior is not expected for nondissociating species, such as urea in this screening. Next to the ionic strength, the urea concentration influences the pH of aqueous solutions (Bull et al., 1964). Instead of correcting the pH of the mixture by titration, the exact amount of titrant was determined beforehand in a separate experiment ( Figure 2a). The disassembly buffer for the DisA-TS was then prepared according to the results of this experiment. Reaction analysis was carried out by an SEC-HPLC method, where the mobile phase composition was the same as the examined disassembly reaction conditions. Compared with using a standard analysis mobile phase, the disassembly reaction is less influenced using the mobile phase at disassembly conditions. The selected 300 Å pore size SEC-HPLC column led to separation of VLPs, dimers (capsomeres), and lower-molecular- Here, VLPs in neutral buffer were directly injected into the mobile phase at disassembly conditions, resulting in an on-column disassembly reaction. Figure 3 shows a comparison of the two experimentally determined disassembly rates. While Cp149, this condition already resulted in the presence of HBcAg dimers, indicated by an initial dimer yield of 0.07, as similarly observed previously (Singh & Zlotnick, 2003 Figure S1), the decline in dimer yield was accompanied by a con-

| Synergistic effects
The combination of increasing pH and c urea generally led to higher dimer yields as compared with pH or c urea increase alone. Figure 4c shows Cp149  Generally, it has to be noted that dimer yields below 1 are commonly observed for disassembly of in vivoassembled VLPs, for example, 0.58 to 0.89 in a recent publication . This behavior is expected for in vivo-assembled VLPs, where dis-and reassembly aim to remove inactive protein .
A comprehensive overview of the impact of reaction conditions on maximum dimer yield and k 0,DisA OC is given in Figure 5. While k 0,DisA OC was similar for Cp149 and VLP A at pH 7.2, the maximum dimer yield was significantly higher for Cp149. At higher pH and c urea , dimer yields were comparable between Cp149 and VLP A, while k 0,DisA OC was generally higher for VLP A. This illustrates that not only the dimer yield but also k 0,DisA OC is influenced by the molecular structure of the VLP, in this case, the insertion of a foreign epitope and addition of a C-terminal polyhistidine tag for VLP A. In essence, the screening experiments showed that higher pH and c urea lead to higher dimer yields, which is, however, limited by aggregation, especially for VLP A and at pH 9. The highest dimer yields after 24 h were 0.71 for Cp149 and 0.69 for VLP A, achieved at pH 8.5,

| Filtration-based dis-and reassembly
At a 30 mL scale, DF-based disassembly of Cp149 and VLP A was performed, followed by an 18 h overnight hold, a dimer separation step by dead-end filtration, and DF-based VLP reassembly ( Figure 1b). In this study, disassembly is achieved by buffer exchange of 6 diafiltration volumes (DV) into the DF disassembly buffer, which HILLEBRANDT ET AL. during the hold step and was implemented in all presented processes. Another potential measure to prevent aggregation is the supplementation of the disassembly buffer with additives such as NaCl (Singh & Zlotnick, 2003), glycerol (Schumacher et al., 2018), or surfactants (Shi et al., 2005). A screening for additives and their optimal concentration could easily be performed using the developed DisA-TS method, which is, however, out of the scope of this study. As observed in the DisA-TS, the highest dimer yield was 0.71. The remainder of the protein is regarded as inactive protein and is therefore removed in a separation step (Zlotnick et al., 1996. Here, dead-end filtration with a 0.2 µm syringe filter and a 100 kDa MWCO membrane aimed for removal of undesired species with higher molecular weight than dimers. The MWCO of 100 kDa was selected as it successfully retained VLPs during the preceding purification (Supporting Information Section S1). The permeation of dimers through the centrifugal filter membrane was confirmed by SEC-HPLC in a preliminary test (data not shown). For the subsequent VLP reassembly, DF has proven to be a valuable tool (Liew et al., 2012;Rüdt et al., 2019) and was therefore applied for 3 DV in this study. were adapted to the current theoretical buffer composition in the retentate of the CFF unit (Section 2.4) aiming for a minimal bias of the analysis procedure with regard to the measured dimer yield. The same theoretical 280 nm extinction coefficient was used for all mobile phases in this study. It has to be noted that the absorption of proteins at 280 nm increases with increasing urea concentration (Pace et al., 1995) and thereby leads to a relative overestimation of the protein concentration for samples with higher urea F I G U R E 5 Effect of screening conditions on k , 0 DisA OC and total dimer yield of Cp149 and VLP A. The center of each bubble determines the screening conditions, urea concentration, and pH. The color intensity and the area of a bubble represent k , 0 DisA OC and the maximum of the total dimer yield, respectively. Note that the DisA-OC approach was not performed for conditions at pH 9 due to the pH limit of the SEC-HPLC column. Bubbles of conditions which did not allow for k , 0 DisA OC determination are transparent and marked with a black "X." DisA-OC, disassembly on column; HPLC, high-performance liquid chromatography; -k , 0 DisA OC , disassembly rate from disassembly on column; SEC, sizeexclusion chromatography concentration, which was considered negligible for this study.

| At-line monitoring of DF-based disassembly
Compared with DisA-TS, Figure 6 shows a 19%-increased initial

| Process data of the filtration-based process sequence
In addition to at-line analysis for dimer quantification during disassembly, all process steps were analyzed by off-line SEC-HPLC using a 1000 Å pore size column. This column allowed for a better separation and quantification of differently sized species as compared with the column used in the disassembly screening. Besides VLPs and dimers, a peak with HMWS larger than VLPs was detected.
As already shown in a previous study (Hillebrandt et al., 2020), these HMWS are expected to be forms of HBcAg, such as partially reversible aggregates of VLPs or dimers (Newman et al., 2009;Schumacher et al., 2018), as mainly dimers are detected after disassembly. The recovered mass of each species after each process step is listed in Table 1. Besides the aforementioned species, a shoulder of the dimer peak and LMWS were detected in the 280 nm chromatogram. According to their ultraviolet light spectra, the dimer shoulder is a protein species, which could constitute aggregated or partially unfolded forms (Samandoulgou et al., 2015) of the HBcAg dimers. According to their ultraviolet light spectra, the LMWS are a mixture of nucleic acids, buffer species, and/or proteins (data not shown). The content (by peak area) of LMWS after reassembly was 1.8% for Cp149 and 6.4% for VLP A while the dimer shoulder content was 4.0% and 0%, respectively. As no clear trend was observed, both species were not further investigated.
During the 5°C overnight hold, the Cp149 dimer yield further increased from 0.73 ( Figure 6) to 0.84 (Table 1) showing no HMWS or VLPs. A potential reason for the higher dimer yield is the temperature-related pH increase. The strong temperature dependence of the Tris pK a (Beynon & Easterby, 1996) resulted in a measured increase of~0.3 pH units for the used DF disassembly buffer. Furthermore, decreasing the temperature might increase the extent of disassembly as the opposite reaction, that is, VLP assembly, is favored at higher temperature . Due to the cooling costs and the long downtime, the overnight hold seems not profitable at a larger scale and an immediate continuation of the process is suggested in this case. Note, that a potential yield loss due to the 0.2 µm filtration is included in the aforementioned yields and was not separately investigated. For the dimer separation step, the apparent retention coefficient of dimers R dimer was determined under F I G U R E 6 Total dimer yield determined by at-line SEC-HPLC during DF-based disassembly. The samples were taken at every DV, and the analysis was completed after a median duration of 13 min. Dashed lines were added to guide the eye. DF, diafiltration; DV, diavolume; HPLC, high-performance liquid chromatography; SEC, size-exclusion chromatography T A B L E 1 Summary of process data for the filtration-based downstream process Note: The product species differ between process steps. HMWS, VLPs, and dimers are regarded as product species of the feed, while dimers are regarded as product species of disassembly and hold and dimer separation. For reassembly, VLPs are regarded as product species. The recovered product species mass of each step is shown in bold.
Step yield, concentration, and A260/A280 refer to the product species of each step.

HILLEBRANDT ET AL.
| 3935 process conditions. Based on the results in Table 1, R dimer was 0.10 for Cp149 and 0.18 for VLP A. The higher retention of VLP A dimers probably results from the higher level of HMWS. The retained HMWS can build up a fouling layer that has been shown to influence the overall selectivity of a fouled membrane in the case of albumins (Meireles et al., 1991). Another possible explanation could be in- showed a greater tendency to aggregate. Next to the disassembly process, aggregation challenges have also been reported for reassembly (Ding et al., 2010;Rüdt et al., 2019). For the interested reader, a detailed interpretation of the VLP A product loss can be found in the Supporting Information Section S3.
The disassembly experiments showed an initial increase of the transmembrane pressure (Supporting Information Figure S2). As this pressure increase is concomitant with the degree of buffer exchange and comparably constant toward the end, it can most probably be attributed to the viscosity increase (van Reis & Zydney, 2010) due to the increasing urea concentration (Kawahara & Tanford, 1966). Reassembly experiments showed slightly decreasing transmembrane pressures over time (Supporting Information Figure S2). Overall, DFbased dis-and reassembly resulted in low mean transmembrane pressures of 0.15 and 0.10 bar, respectively, with mean absolute deviations of 0.01 bar for each run. For both membranes, cleaning according to the manufacturer's instructions recovered the water permeabilities compared with the ones before the experiment (note that new membranes were conditioned during pre-experiments to avoid yield or permeability loss due to adsorptive effects). To this end, membrane fouling had no noticeable effect on the filtration performance and was not irreversible. Nevertheless, the permeate flux and membrane loading (amount of retained solutes per membrane area) were comparably low (Liew et al., 2012;Rosenberg et al., 2009;Rüdt et al., 2019;van Reis et al., 1997) and not optimized in this study. Hence, membrane fouling and product quality should be carefully investigated when these parameters are increased for economic reasons.
The 260 to 280 nm absorbance ratio (A260/A280) is an indicator for the nucleic acid content in a protein solution (Layne, 1957;Porterfield & Zlotnick, 2010) but is also influenced by the solution conditions, such as the urea concentration (Donovan, 1969;Pace et al., 1995). Therefore, only samples analyzed under the same solution conditions can be compared, which are feed and reassembly as well as disassembly/hold and dimer separation. For Cp149, the A260/A280 was equal for these pairs (Table 1). It was 0.61 for the feed and after VLP reassembly, where 0.60 was previously regarded as pure non-truncated HBcAg monomer in water, based on theoretical considerations (Porterfield & Zlotnick, 2010). The A260/A280 of dimers after disassembly was 0.55 for Cp149 and 0.56 for VLP A and remained constant after dimer separation. The Cp149 A260/ A280 is comparable to 0.57 obtained by affinity chromatography at a urea concentration of 4 M . The A260/A280 of the VLP A feed was 0.67 and was decreased to 0.63 after reassembly, indicating the removal of nucleic acids. It is important to mention that the feed A260/A280 in Table 1 is calculated based on the peak areas of HMWS, VLPs, and dimers while it was 0.64 for the VLP peak, suggesting that the depleted nucleic acids were mainly associated with (or bound to) the HMWS. This was also observed in a recent study with murine polyomavirus VLPs (Gerstweiler et al., 2021). Both VLPs used in this study lack the C-terminal protaminelike region of the wild-type HBcAg, which reduces packaging of nucleic acids (Crowther et al., 1994;Zlotnick et al., 1997). Considering VLPs with a higher nucleic acid burden, the developed process sequence could demonstrate even better separation capacities. For further improvement of the purification performance, strongly bound nucleic acids could be digested by a nuclease after disassembly  and nucleotides washed out as described previously (Hillebrandt et al., 2020). Another reason for dis-and reassembly lies in the improvement of particle structure and homogeneity (Mach et al., 2006;Zhao, Allen, et al., 2012). This could be shown, especially for VLP A, by the reduction of the HMWS content, suggesting improved VLP homogeneity.
In summary, the filtration-based process sequence has proven efficient in the realization of dis-and reassembly, depleting impurities, and decreasing the HMWS content. An observation during DF-based reassembly was the presence of a small fraction of unassembled protein at the end of the process. A polishing step by flow-through multimodal SEC (Hillebrandt et al., 2020)

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request. The amino acid sequence of VLP A cannot be made available because it is confidential industry data.    Due to volume reduction, the measured dimer concentration c dimer,tot is increased due to mass conservation. The corrected concentration is