Effect of plasmid DNA isoforms on preparative anion exchange chromatography

Increased need for plasmid DNA (pDNA) with sizes above 10 kbp (large pDNA) in gene therapy and vaccination brings the need for its large‐scale production with high purity. Chromatographic purification of large pDNA is often challenging due to low process yields and column clogging, especially using anion‐exchanging columns. The goal of our investigation was to evaluate the mass balance and pDNA isoform composition at column outlet for plasmids of different sizes in combination with weak anion exchange (AEX) monolith columns of varying channel size (2, 3 and 6 µm channel size). We have proven that open circular pDNA (OC pDNA) isoform is an important driver of reduced chromatographic performance in AEX chromatography. The main reason for the behaviour is the entrapment of OC pDNA in chromatographic supports with smaller channel sizes. Entrapment of individual isoforms was characterised for porous beads and convective monolithic columns. Convective entrapment of OC pDNA isoform was confirmed on both types of stationary phases. Porous beads in addition showed a reduced recovery of supercoiled pDNA (on an 11.6 kbp plasmid) caused by diffusional entrapment within the porous structure. Use of convective AEX monoliths or membranes with channel diameter >3.5 µm has been shown to increase yields and prevent irreversible pressure build‐up and column clogging during purification of plasmids at least up to 16 kbp in size.


K E Y W O R D S
dynamic binding capacity, entrapment, open circular isoform, pDNA, preparative chromatography

INTRODUCTION
Plasmid DNA (pDNA) is the central molecule of modern biotechnology, both in gene therapy and vaccination, and many clinical studies for DNA vaccines against the SARS-CoV-2 virus are currently underway [1].pDNA serves as the basis for the transcription of messenger RNA, which represents the active ingredient in the new generation of vaccines [2][3][4].It also serves as a cell line transfection vector used in the production of adeno-associated viruses for gene therapy [5].The preparation of large quantities of pDNA has been thoroughly studied and optimised, both upstream and downstream [6,7].Downstream processing represents a significant fraction of the production cost, often more than 50% of the total [8].The heart of downstream processing usually consists of one or more tangential flow filtrations (TFFs) and chromatographic steps [9,10].Among different chromatographic supports, monolithic columns have already been successfully used in the industrial scale purification process of pharmaceutical grade pDNA due to their flow rate-independent separation and high dynamic binding capacity (DBC) [11][12][13].The most standard choice for pDNA purification is anion exchange (AEX) chromatography, because DNA molecules are highly negatively charged, thus interacting with positively-charged AEX surface [14][15][16].
New development in the field increased the need for pDNA with sizes above 10 kbp (large pDNA), but their chromatographic purification is often challenging due to low process yields and column clogging [17].Krajnc et al. [15] reported a recovery decrease from 95% to 51% for pDNA of 4.7 and 61 kbp, respectively, when experiments were performed on CIM diethylaminoethyl (DEAE) monolithic supports.Purification of pCAMBIA-1303 (12.3 kbp) on a berenil-modified Q-Sepharose achieved a recovery of 45% [18].A low recovery together with a pressure built-up on the column was also measured using an arginine-modified monolithic support for the purification of pHV-1666/67 (8.7 kbp) [19].Molloy et al. confirmed the underdetermination of some pDNA isoforms, especially open circular pDNA (OC pDNA), by comparing the high-performance liquid chromatography (HPLC) method with agarose gel electrophoresis (AGE), chloroquine-AGE, multi-angle laser light-scattering and electron microscopy studies [20].Similar conclusions were obtained by Mahut et al. for the low recovery of ≥10 kbp OC pDNA from a silica bead column modified with quinine-carbamate ligands [21].Irreversible pressure build-up was sometimes observed (internal data, results not shown) during loading on CIMmultus DEAE with no apparent correlation with process parameters (sample preparation, such as filtration efficiency, pDNA size, pDNA concentration and presence of chromosomal DNA).Gabor et al. [22] described reversible entrapment [23] of pDNA in analytical chromatography using different AEX supports.Although they have not observed the substantial decrease of the recovered supercoiled pDNA (SC pDNA) and linear pDNA (LIN pDNA) isoforms of the pDNA with the increase of the plasmid size and with the increase of the flow rate, a pronounced decrease of the OC pDNA isoform recovery was observed.This was attributed to geometrical constraints of OC pDNA which adopts a more open structure in solution and is thus hindered from entering the stationary support.
The results of this study led us to the idea that the OC pDNA isoform could be responsible for the fouling of chromatographic columns during the preparative purification of pDNA.The main purpose of the present study was to test this hypothesis by evaluating the behaviour of pDNA of multiple different sizes in AEX monoliths with varying channel diameter, and to identify chromatographic support suitable for purification of pDNA between 10 and 16 kbp in size.Chromatographic performance was evaluated based on parameters, including DBC, plasmid recovery and elution volume.Hypothesis of pDNA isoform entrapment was also tested on other types of chromatographic media, such as porous beads and convective membranes.

Chemicals
Buffers were freshly prepared with deionised water and analytical grade reagents.Buffer solutions were filtered through a 0.2 µm PES filter (Thermo Fisher Scientific, Nalgene Rapid-Flow, USA

Plasmid isolation
pAAV2/8 (7.3 kbp), pKLAC1 (9.1 kbp), pGMAAV (11.6 kbp) and pAdDeltaF6 (15.4 kbp) were purified using HIP2 Plasmid Process pack (Sartorius BIA Separations, Ajdovščina, Slovenia) by optimising the manufacturer's protocol for each plasmid [12].E. coli cell pellets were thawed in heated bath.Afterwards, cell pellet was resuspended in 50 mM TRIS, 10 mM EDTA, pH 8.0 by mixing on a laboratory mixer at 500 rpm for 60 min.The volume of resuspension buffer (and lysis and neutralisation buffers in later steps) equals 10, 20, 50 and 20 mL of buffer per 1 g of pKLAC1, pAdDeltaF6, pAAV2/8 and pGMAAV cell paste, respectively.We increased the buffer consumption when this was still feasible at bench scale, as this improved pDNA extraction from cells significantly.Afterwards, different procedures of alkaline lysis were followed, each tailored to selected plasmid.Optimal conditions were determined in preliminary experiments (not part of this manuscript) and are listed in Table 1.
Lysis buffer (same volume as resuspension) was added to cell resuspension and left to incubate for predetermined time-period (Table 1).The solution was gently stirred throughout the lysis and followed steps by glass stirring rod or by the circular motion of the container.Neutralisation solution (3 M CH 3 COOK pH 5.5) was added in the same volume as lysis buffer and enough of 5 M CaCl 2 solution to reach a final concentration of 0.5 or 0.75 M CaCl 2 .
Lysate solution was afterwards filtered using depth and membrane filter.Filtered lysate was TFF purified with the intention of concentrating and exchanging buffer to 50 mM TRIS, 10 mM EDTA, pH 7.2.TFF was followed by AEX chromatography using CIMmultus DEAE columns (8-or 80-mL CV) to remove host cell DNA and RNA.For pAAV2/8, a purification column with a 2 µm channel diameter was used, whereas a 6 µm channel diameter column was used for plasmids pKLAC1, pAdDeltaF6 and pGMAAV.Unless stated otherwise, final pDNA samples were buffer exchanged to 50 mM TRIS, 10 mM EDTA, 0.2 M NaCl, pH 7.2.

2.4
pDNA samples enriched with OC or SC pDNA isoforms pAdDeltaF6 (15.4 kbp) samples enriched with OC and SC pDNA isoforms were produced with CIMmultus C4

Monolith characterisation
The specific surface area of the monoliths (S BET ) was measured via nitrogen adsorption by TriStar II 3020 (Micromeritics Instrument Corporation, Norcross, GA, USA).Nitrogen of 99.999% purity was used.Before analysis, monolith samples were dried in nitrogen flow at 70 • C for 1 h.Pore size distribution was measured by a Pascal 440 (ThermoQuest Italia, Rodano, Italy) mercury porosimeter within the range of 15-10 000 nm. Approximately 0.1 g of dried monolith sample was used for the measurements.
The differential pressure on the convective interaction media (CIM) monolithic columns was recorded in 50 mM TRIS, 10 mM EDTA, 0.2 M NaCl pH 7.2 buffer prior to and after pDNA sample loading by a differential manometer at different flow rates.

Evaluation of anion exchangers by injecting analytical amounts of pDNA
First evaluation of CIMmultus DEAE 1 mL columns was performed with the injection of analytical quantities (15 µg) of pDNA on 1 mL columns.Samples were loaded in 50 mM TRIS 10 mM EDTA 0.5 M NaCl pH 7.2 buffer and eluted with gradient 35%-75% (3.3 min, 6 mL/min, if not stated otherwise) of buffer 50 mM TRIS 10 mM EDTA 1 M NaCl pH 7.2.Load and elution fractions were collected and analysed using CIMac pDNA 0.3 mL analytical column (6 µm channels) to determine isoform composition.
Péclet number (P e ) provides information about the mass transport of the plasmid through the monolith channels and suggests the type of entrapment inside the pores or channels of the stationary phase.When the P e is higher than 1, convection overcomes diffusion and at least partial entrapment of the pDNA is expected.P e was calculated according to which was derived from Ref. [22] considering the requirements of radial column flow distribution.Φ represents the volumetric flow rate in mL/min, D is the OC pDNA isoform diffusivity and was calculated according to Robertson et al. and Prazeres [25,26]; d p is the average pore diameter, obtained from Hg porosimeter measurements (2020, 3080 and 5630 nm for three different monoliths).ε max is the total bed porosity (0.6 for the monoliths used), OD is the outer diameter of radial CIMmultus 1 mL column, being 18.6 mm, h is the column bed height (4.2 mm).d c is a theoretical OC pDNA diameter if the plasmid molecule shape would be a regular circle (µm), and d is a theoretical distance the pDNA would have to diffuse to escape from a constriction (it is a constant of 17 nm, for a detailed explanation check Ref. [22]).

2.7.1
Determination of dynamic binding capacity DBC measurements for BSA and pDNA samples were performed.The procedure described below was used for pDNA.The only difference when determining it for BSA was the buffer composition and loading BSA concentration.
Columns (CIMmultus DEAE 1 mL) were washed with 10 CVs of 1 M NaOH and left in contact with solution for 30 min.Water flush (10 CVs) and equilibration with 50 mM TRIS, 10 mM EDTA, 0.2 M NaCl pH 7.2 (loading buffer) (10 CV), then 50 mM TRIS, 10 mM EDTA, 1 M NaCl pH 7.2 (elution buffer) (10 CV flush and 30 min contact time), and finally loading buffer (10 CV) followed.Purified plasmid samples in loading buffer (concentrations: pAAV2/8 65 µg/mL, pKLAC1 35 µg/mL, pAdDeltaF6 24 µg/mL) were loaded on the column until saturation was visible as an increase in absorbance signal due to plasmid breakthrough.Loading was performed at 6 mL/min flow rate (unless stated otherwise), and elution was performed in two different flow rates for each run, first at 6 mL/min, then at 0.5 mL/min (unless stated otherwise).The UV signal was monitored at 260 nm.After elution, a wash step was performed with 5 CV of 0.1 M NaOH solution.NaOH fraction was immediately neutralised with 5 mL 0.5 M phosphate pH 6.5.All pDNA fractions (loading sample, flow-through fraction, elution at 6 mL/min, elution at 0.5 mL/min and 0.1 M NaOH wash) were collected and analysed by UV spectroscopy for an overall concentration of nucleic acids or HPLC analysis for OC pDNA and SC pDNA isoform ratio determinations (see the following sections for details).
The loading buffer for BSA capacity was 20 mM TRIS, and pH 7.4 and 1 M NaCl were added to the elution buffer.BSA loading concentration was 1 mg/mL and was loaded until column saturation.The UV signal was monitored at 280 nm.The fractions at the column outlet were not additionally analysed.
DBC was calculated according to where t 10% is the time when the breakthrough signal during loading at the end of the column reaches 10% of the maximum absorbance, V d is the dead volume of the system (mL), γ pDNA or BSA is the concentration of pDNA or BSA in the loading buffer (mg/mL) and V col is the CV (mL).Pressure drop on column was followed with a differential manometer throughout the whole chromatographic run.

Comparison between monoliths, porous beads and membranes
The experiments were performed with pGMAAV (11.6 kbp) at 24 µg/mL concentration in a loading buffer.OC pDNA percentage in the loading sample was 22%.Each column was loaded with 500 µg of pGMAAV.Experimental set-up was same as described in previous section; flow rate used was 1.3 mL/min for loading and 0.4 mL/min for elution and NaOH wash.SOURCE 15Q, POROS 50 D, Fractogel EMD DEAE, Sartobind Q Nano, CIMmultus DEAE 2 µm and CIMmultus DEAE 6 µm were evaluated, all of them with 1 mL column bed volume.The chromatographic fractions were analysed for pDNA isoform ratio and pDNA concentration.

Agarose gel electrophoresis
For agarose gel analysis we loaded 20 µL of sample in each well.Prior to analysis, the samples were diluted with double deionised water to achieve the loading of 55 ng of nucleic acids per well.AGE was conducted with 1% agarose gels at 100 V for 60 min or 100 V for 120 min and stained with Sybr Gold from Thermo Fisher Scientific (Waltham, MA, USA) [27].

Analytical evaluation of pDNA isoform composition
pDNA isoform composition was performed using CIMac pDNA 0.3 mL analytical column (6 µm channels).Convective channels of 6 µm enabled the accurate quantification of pDNA isoforms of plasmids up to 16 kbp sizes [28].Loading buffer was 100 mM TRIS, 0.3 M guanidinium hydrochloride 0.3 M NaCl pH 8.0, whereas the elution buffers consisted of 100 mM TRIS, 0.3 M guanidinium hydrochloride, 0.7 M NaCl pH 8.0.Isoform separation was achieved in 6 min linear gradient from 12% to 32% of elution mobile phase at a flow rate of 0.5 mL/min.Example of chromatogram is shown in Figure S1.The quantifica-tion of the plasmid was performed using a calibration curve, performed with in-house prepared 15 kbp large pDNA standard.Analyte concentrations between 5 and 100 µg/mL were used to construct the calibration curve.Total chromatographic pDNA areas (OC + SC pDNA together) were integrated and correlated with UV spectrophotometric measurements of the same samples.One unit of OD 260 nm in a 10 mm cuvette was assumed to correspond to 50 µg/mL of a double-stranded DNA.
Isoform ratio in each collected fraction was calculated from HPLC data according to the following equation: where w iso pDNA is mass fraction for specific isoform, whereas A iso pDNA represents the determined chromatographic areas for the same isoform.A OC pDNA , A OC pDNA and A OC pDNA are chromatographic areas for the sum of all pDNA isoforms present in the sample.Mass of a specific isoform (m iso pDNA ) in the collected fractions was calculated by multiplying the mass fraction of each isoform with the sample volume in mL (V) and with the concentration of overall plasmid (γ pDNA ) in the same collected fraction: Elution recovery or NaOH wash recovery (R) was calculated by dividing the mass of pDNA (overall pDNA or individual isoform separately) in specific fraction by the total mass of loaded pDNA according to the following equation: The mass balance for the whole chromatographic run was calculated by subtracting the amounts of pDNA (overall pDNA or each isoform separately) in each fraction collected at the column outlet (m iso pDNA outlet ) from the amount of pDNA loaded on the column (m iso pDNA load ), see Equation (6).The difference represents the pDNA, irreversibly entrapped on the column (m iso pDNA not recovered ): iso pDNA load = ∑  iso pDNA outlet +  iso pDNA not recovered (6)

Separation of pDNA in linear region of adsorption isotherm
The first goal of our research was to confirm the OC pDNA entrapment, as it was proven by Gabor et al. [22], but this time using preparative columns.For this reason, monolithic CIMmultus DEAE 1 mL columns with three different channel sizes (2, 3 and 6 µm, respectively) were provided by Sartorius BIA Separations and characterised (Table 2).Permeability increase was proportional to channel size and inversely proportional to surface area (see Figure S2).DBC for a standard protein biomolecule, such as BSA, was proportional to surface area and inversely proportional to channel size and according to the well-known literature data [29,30].
The pDNA separation was performed with plasmids of three different sizes -pAAV2/8 (7.3 kbp), pKLAC1 (9.1 kbp) and pAdDeltaF6 (15.4 kbp).Small amount of the plasmid (15 µg) was injected into the 1 mL monolith columns and separated in an ascending NaCl gradient.Resolution between eluted SC and OC pDNA isoforms in elution was not sufficient to calculate the percentage of OC pDNA isoform.Therefore, the elution fractions were collected from all nine performed experiments, and isoform ratio (indicator of OC pDNA entrapment) was determined by CIMac pDNA chromatographic analysis, as described in Section 2. Each experiment was performed in duplicates.Example of analytical chromatogram is shown in Figure S1.
The first experiments on 1 mL columns were conducted at a flow rate of 2 mL/min (2 CV/min), and no difference in the pDNA isoforms ratio between elution and load samples was shown.The calculated P e for 15.4 kbp pDNA at this flow rate was 1.7 (according to Equation 1), which is a boundary condition for reversible entrapment.To validate the results from [22] and confirm entrapment of OC isoform we increased the flow rate to 6 mL/min.Mass transport at this flow rate should be predominantly convection-controlled according to a calculated P e of 5. Indeed, the OC pDNA percentage decreased from 31% in load to 14% in the elution of pAdDeltaF6 (15.4 kbp) on the column with 2 µm channel size (Figure 1).The elution composition did not change on 3 and 6 µm columns, and in addition, we did not observe this phenomenon for smaller plasmids -pAAV2/8 (7.3 kbp) and pKLAC1 (9.1 kbp).

Chromatographic characterisation of columns at preparative pDNA loading
To address the effect of OC pDNA entrapment on the industrial purification of pDNA, we performed preparative experiments using AEX monoliths.The initial testing included the determination of DBCs for all combinations of pDNA-monolith column types.Plasmid samples were loaded on the column in 0.2 M NaCl, because it is the salt concentration at which the plasmid capacity using DEAE monoliths is close to maximal value for the plasmid sizes below 10 kbp size [15].The loading was carried out at 6 mL/min (6 CV/min) until the column was saturated with pDNA (observed as an increase of UV signal), or the loading was stopped earlier due to the increase in pressure on the column.An example of the elution from preparative chromatographic run is shown in Figure S3.Calculated pDNA capacities, pressure issues and experimental notes from the chromatographic runs for all combinations of pDNA and channel sizes are gathered in Table 3. DBCs around 7 mg pDNA per mL of column were consistent with expectations and literature data [11] for the smallest, 7.3 kbp large pAAV2/8 molecule.The DBCs of columns with flow channel diameters of 3 and 6 µm were expectedly lower in accordance with the reduction of the surface area (Table 2) of the monoliths with larger channels [29,31].
The DBC for pKLAC1 (9.1 kbp) on 6 µm monoliths was 2.3 mg/mL.We were not able to complete the capacity measurement on 2 and 3 µm monoliths at 6 mL/min with pKLAC1 due to pressure increase over the column specification.However, we were able to finish the pDNA loading on 3 µm monoliths at a decreased flow rate (0.5 mL/min) to achieve 10% pDNA breakthrough, followed by elution at reduced flow rate.Column with 2 µm channels was irreversibly blocked before the column capacity for pDNA was reached.
In the case of pAdDeltaF6 (15.4 kbp), loading on 2 µm channel column was stopped due to increasing back pressure even before the full capacity was reached and elution was possible only at 0.1 mL/min.The results were slightly better with 3 µm diameter channels, where the  full column capacity was reached, and we were able to perform an elution at 3 mL/min.Only the column with a channel diameter of 6 µm allowed the usage of 6 mL/min through the whole chromatographic run.

Isoform composition of collected fractions from preparative chromatography
The elution of pDNA from preparative runs of the previous section was, unless stated otherwise, performed at two different flow rates -6 mL/min, followed by 0.5 mL/min.Afterwards, the column was washed with 0.1 M NaOH.
Figure 2A shows the isoform ratio in collected chromatographic fractions from different 1 mL DEAE monolith supports for pAAV2/8 (7.3 kbp), expressed as OC pDNA isoform percentage according to Equation (3).Ratio between OC and SC pDNAs in loading sample and in the main elution at 6 mL/min (elution 1) was similar (w OC pDNA ∼ 13%), whereas it increased to 20%-25% of OC pDNA in elution at reduced flow rate (elution 2) regardless of monolith channel size.However, the mass of the pAAV2/8 in elution 2, calculated using Equation ( 4), corresponds to less than 10% of the pDNA in elution 1 regardless of the monolith channel size.The NaOH wash fraction from the 2 µm column contained approximately 1% of the total loaded pAAV2/8 (see Figure S4).On larger channels (3 and 6 µm), no pAAV2/8 was detected in the NaOH wash, see Figure 3A,B.The data show the suitability of DEAE monolith supports for purifying pDNA of sizes up to 7 kbp, but an increase of OC pDNA in reduced flow rate elution already indicates some entrapment.
For plasmid pAdDeltaF6 (15.4 kbp), we only obtained useful data for 3 and 6 µm channels, because the 2 µm column was blocked during loading.On the 3 µm channels, cca twofold decrease of OC pDNA percentage was observed in the elution (from 26% in load to 14%-15% in elutions 1 and 2).The 0.1 M NaOH wash resulted in an OC pDNA ratio comparable to the load.The depleted OC pDNA percentage of 15.4 kbp plasmid in all elution and NaOH wash fractions from 3 µm column suggested its irreversible entrapment.On the contrary, the 6 µm column showed an increase of the OC pDNA per cent in elution 2 compared to the load (from 30% in the load to 52.4% in elution 2).The 0.1 M NaOH wash also showed an increased OC pDNA ratio (73%).The results indicate the higher efficiency of OC pDNA isoform washing from 6 µm column compared to 3 µm.To prove the hypothesis, a mass balance for OC and SC pDNA isoforms was calculated for the whole chromatographic experiment according to Equation (6). Figure 3 shows the'mass balance for all three plasmids on 6 and 3 µm columns.
The results show a decrease in recovery with increasing plasmid size on 3 µm channels.The effect was more pronounced for OC pDNA isoform with recoveries of 94%, 58% and 40% for 7.3, 9.1 and 15.4 kbp plasmids, respectively (Figure 3A).Elution 2 (lower flow rate) and NaOH wash did not improve the recovery significantly for 9.1 and 15.4 kbp pDNAs regardless of the isoform, confirming their irreversible entrapment.
Using larger 6 µm channels, the effect of plasmid size was much less pronounced with recoveries above 85% for all plasmids and isoforms (Figure 3B).A significant portion of the plasmids was present in elution 2 (up to 20%), and in the NaOH fraction (up to 10%), more predominantly visible for the OC pDNA isoform.
These findings suggest that a low flow rate elution strategy can be successfully implemented to improve recoveries of large plasmids only where entrapment is reversible (i.e. on 6 µm channels, but not on 3 µm channels).Furthermore, suitable column cleaning should enable the re-use of the column.

Loading pDNA samples enriched with OC or SC pDNA isoforms
Based on the results obtained in binding capacity experiments, loading OC pDNA-enriched sample of pAdDeltaF6 (15.4 kbp) onto a 2 µm column should increase the pressure drop and clog the column much faster than SC pDNA enriched sample.In this experiment, SC-enriched (99% SC) and OC-enriched (55% OC) samples were loaded on a smaller chromatographic column with 2 µm channels (CIMmic DEAE 0.1 mL disc).A higher flow rate (10 CV/min) was used to exaggerate the pDNA entrapment (high P e numbers, convection-controlled regime).OC-and SC pDNA-enriched samples were loaded onto chromatography units up to full capacity and eluted at two subsequent flow rates, 10 and 1 CV/min (Figure 4).The pressure at the start of the loading was 0.25 bar.At the end of the loading, the pressure with the SC-enriched sample was 0.45 bar, whereas the pressure with the OC-enriched sample was 3 bar.During elution of the SC-enriched sample, 98% of the plasmid was recovered, and pressure on the column returned to initial 0.25 bar in mobile phase A. Elution of the OC-enriched sample only recovered 2% of the plasmid (flow rate 10 and 1 CV/min combined), and exposure of the column to 1 M NaOH resulted in instant clogging.We suspect that the sudden release of pDNA from the surface into the lumen of the channels upon exposure to NaOH tangled the plasmid in a Gordian knot, causing pressure increase and column blockage.
Column blockage and pressure issues are therefore directly related to the presence of OC pDNA isoform, observed when large plasmids are purified on small channel monoliths.

OC pDNA recovery problems on other types of stationary phases
Different commercially available porous beads as well as membrane-type supports were evaluated for pDNA recovery to prove the general nature of the observed phenomena.Experiments were performed with purified pGMAAV (11.6 kbp), serving as model for large pDNA.Sartobind Q (3-5 µm channel size) was evaluated as another member of convective stationary phases.We have selected beads with different particle sizes (SOURCE 15Q with 15 µm, POROS 50 D with 50 µm, Fractogel EMD DEAE with 40-90 µm particle sizes), because they are often used by other researchers for pDNA capture [16,32].The effective interstitial pore diameter for packed beds is between 15% and 35% of the bead diameter [33].The minimal interstitial pore diameters are therefore 2.25, 7.5 and cca 13.5 µm for chromatographic supports packed with 15, 50 and 90 µm beads, respectively, compared with the monolith channel sizes evaluated in the study.CVs were in all cases 1 mL, and the pDNA was loaded at the same flow rate, 1.3 mL/min, which corresponds to 150 cm/h for porous beads.Samples containing 0.5 mg pDNA were loaded on each 1 mL chromatographic column, and no column was overloaded.Constant amount of loaded pDNA enabled direct comparison between the columns for the isoform ratio and elution recoveries in all collected elution and NaOH wash fractions.The results of all tests are gathered in Table 4, which shows elution recoveries as well as NaOH wash recov-  eries for total pDNA and separately for each individual isoform.

Column
An important observation was the high overall recovery (91%) of the pDNA on CIMmultus DEAE 2 µm compared to the results from Figure 3, but this goes well with the decreased flow rate used in present experiment (1.3 vs. 6 mL/min in Figure 3) and smaller pDNA (11.6 vs 15.4 kbp in Figure 3).This suggests efficient capture of 10-15 kbp pDNA using 2 µm monolith if the flow rate is kept around 1 CV/min, which is usually practised on industrial-scale columns.The total plasmid recovery from 6 µm channels (98%) was higher than on 2 µm channels, with no difference in the recovery of OC and SC isoforms.
The most pronounced effect of OC pDNA reversible entrapment was observed with the smallest porous bead size (SOURCE 15Q), where OC pDNA isoform percentage in elution drops from 22% (determined in load) to 17%.On the contrary, isoform pDNA composition in elution fraction from Fractogel EMD was not changed compared to loading sample.This is a result of wide convective channels formed between the Fractogel EMD beads having the largest diameter of all beads tested in our study.Similar behaviour was expected for 50 µm POROS 50 D particles, but very low overall pDNA recovery (15% in NaCl gradient) cannot be rationalised within the bounds of the experiments performed.Overall, 54% of pDNA was eluted only with NaOH.This is not due to reversible OC pDNA entrapment, but most probably due to the nature of the porous structure inside the bead.POROS 50 D consists of perfusion-based beads with flow-through pores less than 800 nm in diameter [32,34].These large pores enable the penetration of large pDNA molecules inside the particle, increasing the overall accessible surface for binding.But the molecules are kept entrapped after the application of an elution buffer.On the other side, Fractogel EMD with average diffusion pores within the bead of 80 nm, as well as SOURCE with pore size between 2 and 100 nm [32], do not allow substantial entering of the pDNA into the particle and the majority of adsorption processes take place on the external bead surface [34].However, a substantial elution of both pDNA isoforms in NaOH wash fraction (11% for Fractogel and 25% for SOURCE) suggested partial pDNA entrapment inside these particles as well.
As expected, this was not the issue on all tested convection-based phases, including Sartobind Q membranes.Sartobind Q has shown similar results as CIMmultus DEAE 6 µm.Almost 100% OC and SC pDNA recoveries in the elution step, confirming our hypothesis of a good balance between convection-based stationary phase and large channel size for capturing large plasmids.The results lead to two important conclusions -(a) at least 4 µm convective channels/pores are needed for high OC pDNA recovery of cca 15 kbp sized pDNA; (b) even the porous beads with the largest pores are not suitable for the chromatographic capture of such large plasmids.

CONCLUDING REMARKS
Monolith chromatographic columns with 2, 3 and 6 µm channel sizes were thoroughly characterised for pDNA capture.We were able to confirm that OC pDNA isoform was the driver of reduced chromatographic performance when anion exchanging columns are used.Entrapment of OC pDNA was confirmed by: (1) analytical injection of 15.4 kbp pDNA onto a column with a channel diameter of 2 µm; (2) decreased recovery of OC pDNA in preparative experiments of pAdDeltaF6 (15.4 kbp) and pKLAC1 (9.1 kbp) on columns with channel diameters of 2 and 3 µm; (3) by irreversible clogging and absence of pDNA elution from the column, when loading OC pDNA enriched sample of 15 kbp pDNA onto a 2 µm chromatographic column.The reversible entrapment of large OC pDNA was proven on porous beads with particle size below 30 µm.However, the more acute problem of porous beads for the capture of >10 kbp pDNA is overall low recovery of pDNA.Advantages of convection-based monolithic chromatographic supports for large pDNA were additionally proven by highest overall and individual isoform recoveries observed.
The results suggest the application of anion exchanging convective monoliths or membranes with >4 µm channels size for plasmids at least up to 16 kbp in size.Plasmids with sizes below 10 kbp could also be responsible for some pressure issues, if the OC pDNA content is above 30%.In addition, attention to the thorough regeneration of the column after the preparative cycle is needed.The back pressure of the regenerated column should return to the value before the chromatographic run; otherwise, this is an indication of partial column clogging.Our main research efforts are now focused on increasing pDNA binding capacity and recovery while avoiding column clogging.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors have declared no conflict of interest.

D ATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

R E F E R E N C E S
1. Dong Y, Dai T, Wei Y, Zhang L, Zheng M, Zhou F. A systematic review of SARS-CoV-2 vaccine candidates.Signal Transduct Target Ther.2020;5(1):237.2. Karikó K, Muramatsu H, Ludwig J, Weissman D. Generating the optimal mRNA for therapy: HPLC purification eliminates immune activation and improves translation of

TA B L E 2
Basic characteristics of CIMmultus diethylaminoethyl (DEAE) columns with three different channel sizes.Note: Geometry, volume and porosity are the same for all three column types.Abbreviations: BSA, bovine serum albumin; DBC, dynamic binding capacity.F I G U R E 1 Decreasing the open circular plasmid DNA (OC pDNA) percentage from loading to elution fractions on CIMmultus diethylaminoethyl (DEAE) 1 mL columns with different channel sizes at 6 column volume (CV)/min at small injections (15 µg pDNA per mL of support) of three pDNA samples.The relative standard deviation for calculated OC pDNA recovery was 10%.

TA B L E 3
Chromatographic evaluation of columns through dynamic binding capacity (DBC) measurements for three different plasmid DNA (pDNA) molecules.Starting flow rate of loading was always 6 mL/min (6 CV/min).

F I G U R E 3
Plasmid DNA (pDNA) mass balance calculations, separately for open circular (OC) and supercoiled (SC) pDNA isoforms, for preparative chromatographic experiments using CIMmultus diethylaminoethyl (DEAE) 1 mL columns: (A) 3 µm channels; (B) 6 µm channels.Loaded pDNA mass was normalised to 100%, elution 1 (at higher flow rate), elution 2 (at reduced flow rate) and NaOH wash combined represent total collected pDNA at the column outlet during chromatography run.The relative standard deviation for calculated pDNA amount in each fraction was 5%.

F I G U R E 4
Comparison of elution profiles for supercoiled plasmid DNA (SC pDNA) sample (black) and open circular pDNA (OC pDNA)-enriched sample (hatched line), both pAdDeltaF6 (15.4 kbp).Pressure drop of the elution was not monitored due to anticipated unpredictable pressure changes that could damage the manometer.SC pDNA sample exhibited expected dual elution profile corresponding to two different flow rates used.The pDNA elution was negligible with OC pDNA-enriched sample, confirming the irreversible pDNA entrapment.TA B L E 4 Evaluation of open circular plasmid DNA (OC pDNA) reversible entrapment and overall as well as OC and supercoiled pDNA (SC pDNA) recoveries for porous beads and convection-based stationary phases.
Ana Benčina Kosmač, Tadej Sever, Sara Rotar, Rok Miklavčič, Ana Pavšič, Ana Ferjančič, Maja Kobal, Tjaša Plesničar, Nejc Pavlin, Mojca Bavčar, Klemen Božič and Rok Sekirnik from Sartorius BIA Separations d.o.o. are acknowledged for laboratory support and discussions.We thank COBIK d.o.o., Ajdovščina, Slovenia and Generi Biotech (Hradec Kralove, Czech Republic) for providing us for E. Coli pellets containing plasmid DNA used in the study.This work was partially financially supported by the Slovenian Research Agency (Grant Agreement No. J2-1719).The content provided in this manuscript reflects the author's views only.

TA B L E 1
Experimental conditions for plasmid DNA (pDNA) alkaline lyses and final composition of pDNA samples.
NaOH wash -overall pDNA percentage eluted in NaOH wash determined from UV measurements; W OC pDNA -OC pDNA percentage in elution fraction, determined from CIMac pDNA analytics.OC pDNA percentage in load was 22%; R OC pDNA and R SC pDNA -Elution recovery (without NaOH wash) for individual pDNA isoform separately, calculated from CIMac pDNA analytics.
a C -convective channel size; P -pore size; R -elution recovery from UV measurements using NanoDrop (without accounting pDNA in NaOH wash fraction), calculated for overall pDNA; R