Analytical separation of plasmid DNA isoforms using anion exchanging chromatographic monoliths with 6 µm channels

High‐performance liquid chromatography (HPLC)‐based analytical assays are used to effectively monitor purity and quantity of plasmid DNA (pDNA) throughout the purification process. However, the phenomenon of physical entrapment of open circular (OC) isoforms pDNA inside narrow channels of chromatographic support decreases its accuracy and precision and the effect increases with pDNA size. The purpose of the study was to develop a chromatographic method for accurate analytical separation between isoforms of <16 kbp pDNA using weak anion exchanging monolithic column with large (6 µm) convective channels. Purified samples of 4.7 and 15.4 kbp large pDNA with known isoform composition were prepared and their isoforms separated in ascending salt gradient. Both OC and supercoiled (SC) isoforms were baseline separated at a flow rate below 0.5 mL min−1 in a guanidinium chloride (GdnCl) gradient with a ≥95% OC pDNA elution recovery. However, these chromatographic conditions increased 2 times the peak width for linear (LIN) pDNA isoform compared to the results using monoliths with 1.4 µm channel size. If other chaotropic agents, such as urea or thiocyanate (SCN), were added to Gdn ions, the elution volume for LIN isoform decreased. Optimization of combined GdnCl/GdnSCN gradient for pDNA elution resulted in a simple and robust chromatographic method, where OC–LIN and LIN–SC pDNA (up to 15 kbp size) were separated with resolution above 1.0 and above 2.0, respectively. The accessibility and general acceptance of anion exchange chromatography for pDNA analytics give the newly developed method a great potential for in‐process control monitoring of pDNA production processes.


INTRODUCTION
Plasmid DNA (pDNA) is an enabling product, critical in production of messenger RNA (mRNA), adeno-associated viruses and other viral and non-viral therapeutic vectors [1][2][3].Increasing yield and purity in production of pDNA is a vital step in meeting such demand.pDNA isolated from Escherichia coli cells usually consists of different pDNA isoforms, such as open-circular pDNA (OC pDNA), supercoiled pDNA (SC pDNA), linear pDNA (LIN pDNA) and various multimeric forms [4].Purity and/or isoform composition requirements vary depending on the application of the pDNA.High purity of SC pDNA isoform is required in gene therapy platforms, whereas isoform composition is usually not defined for the preparation of IVT mRNA production, where linearization step follows the pDNA purification [5][6][7].In any case, pDNA must be purified from all other contaminants present in the E. coli lysate, such as cell debris, RNA, residual proteins and genomic DNA.A robust, accurate, selective and fast analytical method is needed to ensure unequivocal control of upstream and downstream pDNA production processes.
High-performance liquid chromatography (HPLC)based assays are powerful, high-resolution techniques for rapid qualitative and quantitative determination of different pDNA isoforms [8,9].pDNA isoform separation and quantification by anion exchange (AEX) chromatography is based on interaction between negatively charged phosphate groups in the DNA backbone and positively charged ligands on the matrix surface.Both strong quaternary amines and weak (such as diethylaminoethyl = DEAE) ion exchangers are used [10].Compact SC isoforms that have a higher charge density elute later than OC with a lower charge density [11].Base sequence and composition are known to affect the elution of pDNA from anion exchangers [12].By exploiting different physicochemical properties of the OC and SC pDNA, these two isoforms can be baseline separated, applying AEX chromatography in combination with shallow ascending salt gradients.By partial or complete replacement of NaCl with guanidinium (Gdn) salts during the elution, we can also gain insight into multimeric forms of pDNA due to improvement of chromatographic resolution [13].However, quantitative resolution of LIN pDNA can be an issue due to overlap with OC and SC isoform peaks [14].They have also shown physical retardation of LIN pDNA when it travels through the channels of chromatographic media.This was explained as slalom chromatography effect.This effect is related to LIN pDNA size, elasticity and other physicochemical factors related to hydrodynamics [15].Narrowing the convective pore size of the chromatographic support increases the LIN pDNA retention times relatively to OC or SC due to physical retardation of molecules by the channel/pore walls [14].
The efficiency of the HPLC analytical method is defined by the precise qualitative and quantitative determination of individual pDNA isoforms, especially in the final product, where combined amounts of OC, LIN and multimeric isoforms are very small (<5%), compared to the more abundant SC pDNA isoform (95% or more).Therefore, the recovery of pDNA components from the chromatographic media should be almost complete.As previously noted by Molloy et al., the use of chromatographic techniques often underdetermines the percentage of some pDNA isoforms, especially OC pDNA [4].This was proven by comparing the HPLC method with orthogonal techniques -agarose gel electrophoresis (AGE), chloroquine-AGE, multi-angle laser light-scattering and electron microscopy studies.Similar conclusions were obtained by Mahut et al. [16].They suggested a stronger binding of the OC pDNA to anion exchangers, in their case a silica column modified with quinine-carbamate ligands.The pDNA recovery was improved by using an ascending mixed gradient (pH from 7.0 to 7.6, salt from 0 to 0.6 M and isopropanol from 0% to 20%) to elute pDNA from the column.The improvement of pDNA recovery from the column was mainly induced with the negative charge increase of stationary phase due to silica surface deprotonation in this pH range.Relatively complex solution is against our aim, which is improvement of pDNA separation on standard anion exchangers through altering mobile phase composition without using organic modifiers.In previous works, we found the correlation between decreased recovery of OC pDNA isoform and reversible convective entrapment in the channels of chromatographic supports, based on convective mass transport [17].This effect was more pronounced with increasing pDNA size.This phenomenon could be reduced by decreasing the flow rate of the mobile phase or increasing the channel size of the chromatographic support.
In present work, the development of weak AEX analytical chromatographic monoliths with large (6 µm) convective channels enabled us to test their applicability for analytical separation of up to 16 kbp pDNA isoforms.We studied the OC pDNA elution recovery for two different pDNA sizes.The unexpected LIN isoform retention when passing through monolith channels with different channel diameter is also discussed.Understanding of the retention mechanisms for different pDNA isoforms on a monolithic column under investigation enabled an optimisation of a chromatographic method, which can be transferred into routine in-process pDNA monitoring after its validation.
Buffers were freshly prepared with European Pharmacopoeia grade purified water and analytical grade reagents.GdnCl and GdnSCN purity must be higher than 99.0%.All the buffer solutions were filtered using 0.22 µm PES filter (TPP, Trasadingen, Switzerland).

Chromatographic analysis
All chromatographic experiments were performed on a PATfix system, composed of quaternary pump, a multiwavelength UV-Vis detector (10 mm flow cell path length), thermostat oven and conductivity meter (Sartorius BIA Separations).pDNA separation was studied on CIMac pDNA analytical columns (0.32 mL bed volume) with 1.4 or 6 µm channel size (Sartorius BIA Separations).Before analysis, pDNA standards were diluted with equilibration buffer to concentration of 20 µg mL −1 , if not stated otherwise.Injection volume was 50 µL.Separation of pDNA isoforms was realized using NaCl, GdnCl, GdnSCN or combined GdnCl/GdnSCN LIN ascending gradient at 30 • C applying different flow rates of mobile phase (0.2, 0.5 or 1.0 mL min −1 ).Gradient slopes (g) in mM mL −1 were calculated according to the following equation: c F and c 0 represent final and initial salt concentrations in gradient in mmol L −1 , F v is volumetric flow rate in mL min −1 and t g is gradient time in min.pDNA isoforms were monitored using UV detection at 260 and 280 nm.Following gradients were used: Gradient 1: Equilibration buffer: 100 mM Tris, 300 mM NaCl and pH 8.0.Elution buffer: 100 mM Tris, 1000 mM NaCl and pH 8.0.CIMac pDNA column (1.4 µm, 6 µm), LIN gradient from 53% to 73% elution buffer in 25 column volumes (CVs) at different flow rates, corresponding to 17.5 mM salt increase per mL of mobile phase.
For instrument control and data processing, PATfix software (Sartorius BIA Separations) was used.The integration interval was from beginning of the OC isoform elution to the end of the SC isoform elution, resulting in total pDNA area (A pDNA gradient ) in the sample.A vertical rectangular line between the baseline and the minimum absorbance between two separating pDNA isoforms was used to divide pDNA areas into individual isoforms (A OC pDNA or A SC pDNA ).
Total pDNA elution recovery from CIMac pDNA column was determined using the following equation: A pDNA non-binding conditions is determined with a peak area of pDNA injected over the same column at the same flow rate in 100% elution mobile phase, where ionic interactions are completely suppressed.pDNA recovery, calculated in this way, gives an information about the amount of pDNA which is retained on the column due to interactions different from electrostatic interactions (unspecific adsorption or physical entrapment).
Elution recovery of individual pDNA isoforms (OC and SC) at flow rates of 0.5 and 1.0 mL min −1 was calculated using Equation ( 3), where F h represents faster volumetric flow rates (0.5 or 1 mL min −1 ) and F s is 0.2 mL min −1 .The calculation is based on determination of total pDNA elution recoveries at 0.2 mL min −1 , which were determined 100% (using Equation 2) regardless the pDNA size used in the study.

(OC SC) pDNA recovery [%]
= 100 ×  h ×  OC or SC pDNA at f low rate 0.5 or 1.0 mL min −1  s ×  OC or SC pDNA at f low rate 0.2 mL min −1 (3) Resolution R between individual plasmid isoforms was calculated using the following equation: where 1.18 factor is calculated from Gaussian peak properties [19], t SC and t OC represent retention times of both isoforms, whereas w 0.5 SC and w 0.5 OC refer to the peak widths at half peak heights for both isoforms.

Agarose gel electrophoresis
Samples for AGE were diluted to a concentration of 3 µg mL −1 and dyed with TriTrack DNA Loading Dye (Thermo Fisher Scientific, Waltham, MA, USA).Samples were loaded on 1% agarose gel (Sigma Aldrich) in 1× TAE buffer (40 mM Tris, 20 mM acetic acid, 1 mM EDTA) and run at 100 V for 120 min (Mupid-One, Nippon Genetics).For visualization, agarose gel was dyed using SYBR Gold (Thermo Fisher Scientific).

Capillary gel electrophoresis
Capillary gel electrophoresis (CGE) analysis of pDNA was performed on a G7100A system (Agilent, Santa Clara, CA, USA) equipped with a laser induced fluorescence detector with excitation wavelength at 488 nm and a 515 nm band pass emission filter, Zetalif Laser 488UVI (Picometrics, Cambridge, MA, USA).Analysis was performed using the double-stranded DNA (dsDNA) 1000 kit (Sciex, Framingham, MA, USA) with modifications to manufacturer's instructions.The DNA capillary was cut to 42 cm with 30 cm separation length to detection window.The intercalating dye SYBR Gold (Thermo Scientific) was added to separation gel at the amount of 1 µL in 10 mL gel.Samples were diluted with 1× TBE buffer (Thermo Scientific) to a concentration of 2 µg mL −1 and injected with vacuum applied on the inlet capillary side at −5 mbar for 3 s.Separation was performed at 7.8 kV, 20 min.Each sample was analysed in three consecutive runs.Data analysis was performed with OpenLab CDS software (Agilent).

Chromatographic characterization of monoliths using different pDNA samples
Poor chromatographic recovery of OC pDNA isoform recovery is well known, especially when large pDNA molecules were separated by use of ion-exchange columns [4,16,17].This was confirmed for pFix15 on CIMac pDNA column with 1.4 µm channels, bearing DEAE functional group (a weak AEX ligand), where practically no OC pDNA eluted regardless of flow rate (0.2, 0.5 and 1 mL min −1 ), see Figure S1.Chromatographic monoliths with 6 µm convective channels were therefore evaluated to overcome this limitation.Calculation of Peclet numbers (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 number >1, convection overcomes diffusion and at least partial entrapment of the pDNA is expected.The calculated P e for 15 kbp OC pDNA using CIMac monolith with 6 µm channels were 0.64 and 1.24 at 0.5 and 1.0 mL min −1 , respectively [17].This value suggested some OC entrapment at 1.0 mL min −1 but should become almost negligible at ≤0.5 mL min −1 due to diffusion-controlled pDNA escape from the restriction pores.For more information on P e calculation, see the Calculation S2.Theoretical considerations were evaluated thoroughly by chromatography using small (4.7 kbp) and large (15.4 kbp) pDNA standards.Samples were first analysed by orthogonal analytical methods (AGE and CGE) to obtain independent information on their composition (Figure 1, Table 1 and Figure S3).Analysis of pFix5 standard by CGE resulted in baseline separated peaks corresponding to main plasmid isoforms, with SC isoform eluting first, followed by dimer/multimer, LIN and OC isoform.Relative percentage of isoforms, calculated from corrected time areas are gathered in Table 1.With CGE analysis we did not detect other nucleic acid-related impurities, such as RNA and genomic DNA.In Figure 2a chromatographic profile for pFix15 plasmid using ascending NaCl gradient in a buffer with constant pH (Gradient 1) is shown.These represent standard separation conditions [9,20] applied on CIMac pDNA columns.The improved elution of OC pDNA using a column with 6 µm channels was evident (in contrast to 1.4 µm channels); however, the resolution of 1.4 between the isoforms was not satisfactorily (Table 2).The LIN isoform (3% according to CGE analysis), which should elute between OC and SC peaks, could not be confirmed due to low resolution.In addition, if Gradient 1 was applied, only 73% recovery of pFix15 plasmid was achieved (Table 2).A partial or complete replacement of NaCl with GdnCl has been shown to improve the resolution on columns with 1.4 µm channels when working with smaller plasmids [13].Therefore Gradients 2 and 3 were tested on the column with 6 µm channels.Combination of two elution salts, NaCl and GdnCl (Gradient 2), resulted in an improved resolution of 2.4 between the OC and SC isoform (Table 2, Figure 2b).It can be assumed that GdnCl silences OC isoform interaction with the stationary phase more efficiently than SC-stationary phase interaction.GdnCl is a strong chaotropic salt as well as donor-acceptor of hydrogen.It weakens hydrogen bonding in the same way that salts generally weaken electrostatic interaction [13].Hydrogen bonding between plasmid and DEAE ligand was interrupted and less ionic strength was necessary for elution of pFix15 isoforms.Using Gradient 2 and a flow rate of 1 mL min −1 , pDNA recovery increased to 88% (Table 2).
Although the recovery increased, the result still indicated the partial entrapment of some pDNA inside the monolith channels.
In Gradient 3, NaCl was completely replaced with GdnCl (Figure 2c).In this experiment, 300 mM concentration of GdnCl was applied at loading conditions and elution was achieved with increasing GdnCl concentration.OC isoform was eluted at a conductivity of 61.1 mS cm −1 and SC isoform at 62.6 mS cm −1 with resolution of 3.3 (Table 2).This is approximately 2 mS cm −1 lower than in NaCl gradient, proving the weakening of hydrogen-bonding between plasmid and DEAE ligand and increasing selectivity for separation of different plasmid isoforms.In comparison with combination of GdnCl and NaCl, elution with GdnCl (Gradient 3) was not as efficient in terms of pFix15 recovery from the column (77%).Overall, Gradients 2 or 3 enabled a baseline separation of OC and SC pDNA isoforms and these chromatographic conditions were used in further experiments.

Optimization OC pDNA recovery
Figure 3 shows fractionation of pFix15 plasmid on 6 µm CIMac pDNA column using Gradient 2 at different flow rates (0.2, 0.5 and 1 mL min −1 ).SC isoform eluted in a sharp peak at around 17 CV.Different flow rates did not have a significant influence on SC pDNA isoform elution efficiency.SC pDNA recoveries at flow rates 1.0, 0.5 and 0.2 mL min −1 according to Equation (3) were 98%, 100% and 100%, respectively (Table 3).The peak representing OC isoform of pFix15 eluted at approximately 13 CV.There was a significant increase of OC recovery from 75% at 1 mL min −1 to 85% at 0.5 mL min −1 to 100% at 0.2 mL min −1 (Table 3).This observation confirms the hypothesis of reversible entrapment of pDNA, predicted by P e calculations in Section 3.1.At a flow rate of 1.0 mL min −1 in Gradient 2, a resolution of 2.4 can be obtained between pFix15 OC and SC isoforms (Table 3).Decreasing the flow rate to 0.5 mL min −1 improved the OC peak shape (sharper peak) and consequently also the resolution to 2.6.A further decrease in flow rate (to 0.2 mL min −1 ) resulted in a significant increase in resolution for the SC-OC pair to 3.6.The reduced flow rate resulted in a sharper OC elution peak and consequently higher resolution.The conditions described thus far can be implemented for effective monitoring of SC concentration and its relative percentage in the sample.Gradient 2 at reduced flow rate provides a high-resolution separation of SC from OC isoform.
In these conditions, however, LIN pDNA appeared as a shoulder after the OC elution peak.Integration of the LIN pDNA isoform was not possible, and further improvement was needed for efficient characterisation of all isoforms of large plasmids.

Separation of LIN pDNA isoform on chromatographic monoliths
LIN pDNA has become an enabling molecule for production of mRNA [21].Linearization is performed using restriction endonucleases and the cleaved product is further purified and quantified.Its robust and fast analysis is therefore one of the requirements of modern vaccine production processes.
Chromatographic separation of pFix5 plasmid, spiked with LIN pFix5 (see isoform composition in Table 1), on CIMac pDNA column with narrow channels (1.4 µm), performed by use of Gradient 3 at 1 mL min −1 is presented in Figure S4.LIN isoform eluted between OC and SC pDNA.The resolution between OC and LIN isoform of pFix5 was 1.1 and despite low resolution, quantification of both isoforms was feasible.The resolution between LIN and SC isoform in these conditions was 2.9.A width at half of peak height (w 0.5 ) parameter was introduced, because it served as an indirect measure of chromatographic efficiency.w 0.5 for OC, LIN and SC isoforms of pFix5 were 0.16, 0.23 and 0.18 min in Gradient 3 at 1 mL min −1 using 1.4 µm CIMac pDNA column, respectively (Table 4).pFix5 containing LIN isoform was also analysed on CIMac pDNA 6 µm column using Gradient 3 at 1 mL min −1 and a typical chromatogram is presented in Figure S5.The larger channels resulted in a decrease in resolution between the OC-LIN pair to 0.9, and between LIN/SC pair to 1.0 (Table 4).The resolution decrease was mostly due to the broadening of the chromatographic peaks.The w 0.5 parameters on the CIMac pDNA 6 µm were 0.22, 0.47 and 0.28 min for OC, LIN and SC isoform, respectively.This represents a 1.4fold increase of w 0.5 for SC and OC isoforms compared to the CIMac pDNA 1.4 µm.The increase was expected due to the bulk phase dispersion, which correlates with bead particle size increase.Increasing the channel diam-  eter in chromatographic monoliths can be mechanistically understood in a similar way as increasing particle size in chromatography using resins.Unexpectedly large, almost twofold increase of w 0.5 parameter when comparing results from 1.4 and 6 µm channel monoliths, was observed for an LIN isoform of pFix5 standard.
A similar observation was confirmed with pFix15 standard spiked with its LIN isoform (see the isoform composition in Table 1).Analysis using the 6 µm column with Gradient 3 at 1 mL min −1 (Figure 4a) showed a partial coelution of LIN and OC isoforms of pFix15.The LIN pDNA elution peak was broad with a w 0.5 of 0.48 min (Table 4).LIN pFix15 standard was injected separately on 6 µm CIMac pDNA column using Gradient 3 to exclude that peak broadening is due to simultaneous influence between LIN and other pDNA isoform (Figure 4b).As expected, similar chromatographic parameters for LIN pDNA were obtained regardless the sample used.

Optimization of LIN pDNA separation on monoliths with 6 µm channels
It can be anticipated that the LIN pDNA is reversibly retarded in the channels of the monolith due to its shape.It can escape only with diffusion according to a mechanism similar to OC pDNA reversible entrapment.The hypothesis was excluded by performing the separation at reduced flow rates and applying the same salt gradient slope, but no improvement of LIN pDNA was achieved (results not shown).Then, a LIN pDNA was injected on 6 µm pDNA column at 1 mL min −1 in 100% elution buffer, that is in non-binding conditions (Figure S6a).LIN pFix15 eluted as a sharp peak with (w 0.5 0.09 min) without any tailing and any retardation relative to column void volume.Next, the LIN pDNA was injected in binding conditions (low salt concentration) followed by a step to 100% elution buffer (Figure S6b).LIN pFix15 was again eluted as a sharp peak with w 0.5 of 0.08 min and no retardation.Both experiments excluded a slalom chromatography mechanism, which is typical for large, linearized dsDNA inside very narrow chromatographic channels [14].Because the similar peak broadening was not observed on the 1.4 µm pDNA column, open volume of large (6 µm) monolith channels in combination with electrostatic interactions between the stationary phase and pDNA may be the reason.Doublestranded LIN DNA fragments up to ca. 100 bp in length can be treated as a straight rod with a certain degree of elasticity.The longer the molecule becomes, the more it forms a random coil in free solution.[22][23][24].LIN pDNA in kbp size range could therefore adopt random coil conformation in free solution and it is anticipated that similar is possible within wide 6 µm channel of chromatographic columns.Each formed pDNA conformation has slightly different charge density, therefore adsorbing to positively charged chromatographic surface with different binding constant.Consequently, different ionic strengths of the mobile phase are needed for desorption of all LIN pDNA population, resulting in a broad chromatographic peak during the elution using Gradient 3 conditions.When the same LIN pDNA solution flows through the narrow channels of 1.4 µm column, however, the random coil conformation of large DNA molecules may be stretched by laminar shear forces [15], which are then undergoing interactions with the stationary phase [14].Ionic charge of phosphate groups on plasmid is evenly distributed throughout the whole pDNA structure (length), resulting in constant and homogeneous interaction with stationary phase.This is chromatographically reflected as LIN pDNA elution in a sharp peak.The described hypothesis led us to the idea of using additional chaotropic agents in the pDNA solution to prevent the random coil conformations of LIN pDNA.6 M urea can cause denaturation of secondary and tertiary structures of nucleic acids [25] and may homogenize the structure of linearized plasmid and improve the elution of LIN pFix15 on 6 µm pDNA column.The w 0.5 indeed decreased from 0.48 to 0.33 min with addition of urea in GdnCl buffers (Gradient 3), see Figure S6c.However, the improvement was not sufficient for the present analytical application.Additional optimization was performed by inclusion of stronger chaotropes in the chromatographic buffers.

Guanidinium thiocyanate as chaotropic additive for suppressing non-specific interactions
Gdn cation is a denaturant of biological molecules, usually added in solution as a chloride salt.Substitution of chloride anion with a stronger chaotrope, such as thiocyanate, increases its denaturing efficiency [26].Subjecting pFix5 and pFix15 to an ascending GdnSCN gradient successfully eliminated peak broadening (data not shown).However, the LIN pDNA retention time shifted to earlier elution and overlapped with the OC pDNA.It was anticipated that the effect of GdnSCN abolished the structural differences (conformation, geometry, charge-size ratio etc.) between these two isoforms to an extent that they interacted with the stationary phase equally strong.Subsequent attempts focused on combining GdnCl and GdnSCN into mixed gradients and an optimised method at 1 mL min −1 is summarised with Gradient 4 (Figure S7).The pDNA binding was achieved in presence of only GdnCl, whereas the elution was performed in ascending mixed GdnCl and GdnSCN gradient.Most importantly, the w 0.5 for LIN pFix15 decreased from 0.48 min (Gradient 3 with GdnCl) to 0.28 min (Gradient with mixed GdnCl and GdnSCN), whereas the separation between LIN and OC pDNA improved from unresolved peaks to a resolution of 1.2.The final method optimization included a flow rate decrease to 0.5 mL min −1 to improve OC elution recovery and gradient slope decrease from 17.5 to 10.5 mM mL −1 to increase the resolution between isoforms (Gradient 5).Chromatograms of different samples in Gradient 5 are shown in Figure 5. GdnSCN disrupted unspecific interaction between pDNA and DEAE ligand more efficiently resulting in elution at 10 mS cm −1 lower conductivity in comparison with Gradient 3 method (Figure 4a).LIN isoform elution was shifted to higher conductivity compared to OC and simultaneously LIN pDNA w 0.5 decrease improved the separation with SC isoform.This was valid for both pDNA sizes and enabled quantification of all three isoforms in such complex samples.However, even the optimized method still did not separate pDNA dimers/multimers, which were probably co-eluted with SC pDNA.
Table 5 summarizes the analyses of chromatograms from Figure 5.The isoform composition for both pFix5 samples matches perfectly with CGE results (Table 1).Nonbinding recoveries for both pFix5 standards confirmed ≥97% elution from CIMac pDNA 6 µm column (Table S8).Even for the spiked sample with cca 50% LIN isoform the separation efficiency was high enough for accurate peak integration.The interpretation of data for pFix15 and comparison with CGE results was more complex.OC % from AEX analysis was underestimated for AEX (18% vs. 25% determined with CGE), whereas LIN % was overestimated (7% vs. 3%).Similar discrepancy of cca 5% was determined for pFix15 sample spiked with LIN as well.It seems that OC isoform of large plasmid tails into LIN elution, blurring the exact quantification of both isoforms.The SC percentage for pFix15 was overestimated between 2% and 5% with AEX method.Multimer/dimer contribution accounted for 1%-2%, whereas the rest was probably due to less than 100% elution recovery of OC isoform at applied flow rate of 0.5 mL min −1 .This was indeed confirmed with 91% elution recoveries at 0.5 mL min −1 for both pFix15 standards used, see Table S1.Elution recoveries of pFix15 standards were further improved above 96% by decreasing the flow rate to 0.2 mL min −1 (Table S1), but the chromatographic separations were not performed to undoubtedly confirm the discrepancies between CGE and AEX results.
Overall, we have shown that the large-channelled monolith column in combination with strong chaotropic salts as pDNA displacers can be used for separation of 15 kbp large pDNA isoforms.We have confirmed applicability of this method on 5 kbp pDNA size as well.Additional improvement of the separation between OC and LIN pDNA is needed to qualify and validate the method for analytical use.However, it should be noted that different pDNA samples may show different behaviour than pFix15 and the conditions shown may provide sufficient resolution.Additionally, typical process samples during pDNA manufacturing contain lower amounts of LIN isoform (except during pDNA linearisation for mRNA production).

CONCLUDING REMARKS
OC pDNA isoform of 15 large pDNA did not elute from AEX chromatographic monolith (DEAE) with 1.4 µm channel size due to entrapment inside the channels.We have demonstrated that increasing the channel size to 6 µm enabled ≥90% OC pDNA elution recovery, where an ascending gradient of guanidinium chloride was used in combination with ≤0.5 mL min −1 flow rate.Separation between OC and SC pDNA was achieved with a resolution ≥3.However, the presence of higher percentage of LIN pDNA isoform in the sample resulted in co-elution or overlap between the isoforms.The elution volume of LIN pDNA in GdnCl gradient was approximately two times larger than for OC and SC isoforms, indicating inhomogeneous interaction between LIN molecules and stationary phase.The different behaviour of LIN pDNA in 1.4 and 6 µm channels is explained by the structure adopted by the molecule in each case.The larger 6 µm channel enables the molecule to adopt a more three-dimensional structure which affects its charge density, resulting in inhomogeneous charge density, and consequently a wider elution peak.Addition of stronger chaotropic salts in the elution buffer, for example GdnSCN, decreased the elution volume of LIN pDNA isoform by suppressing the folding of LIN pDNA molecules into different 3D structures.Optimization of combined GdnCl/GdnSCN gradient for pDNA elution resulted in a simple and robust chromatographic method, where OC-LIN and LIN-SC pDNA (up to 15 kbp size) were separated with resolution above 1.0 and above 2.0, respectively.The pDNA isoform composition, calculated from AEX results, was compared with CGE analysis, gold standard for pDNA analytics.Both methods correlated for the 5 kbp pDNA and up to 5% underestimation of 15 kbp OC and 5% overestimation for 15 kbp LIN were found for new AEX method.This goes along with slightly lower elution efficiency of pFix15 at higher flow rate of the mobile phase.However, the simplicity, accessibility and general acceptance of AEX chromatography for pDNA analytics give the newly developed method a great potential in process monitoring.Additional improvement of the separation between OC and LIN pDNA is needed to qualify and validate the method for analytical use.

Figure 1
shows AGE analysis of different pFix5 and pFix15 samples, used in this study.pFix5 standard (column b) shows two main bands.The lower bands represent SC isoform of pFix5 plasmid (4.7 kbp, SC ladder), the F I G U R E 1 Agarose gel electrophoresis analysis of pFix5 and pFix15 plasmid: (A) supercoiled (SC) DNA ladder; (B) pFix5; (C) linear (LIN) pFix5; (D) pFix5 spiked with LIN pFix5; (E) pFix15; (F) LIN pFix15; (G) pFix15 spiked with LIN pFix15 and (H) GeneRuler ladder.F I G U R E 2 Elution profile of pFix15 on 6 µm CIMac plasmid DNA (pDNA) column at flow rate of 1 mL min −1 using different gradients: (A) Gradient 1 (elution with NaCl); (B) Gradient 2 (elution with mixed NaCl and GdnCl) and (C) Gradient 3 (elution with GdnCl).upper OC isoform of pFix5 plasmid.LIN form of pFix5 plasmid is shown as one band, see column c. pFix15 standards are shown in column e-g.AGE showed no visual contamination with RNA.

F I G U R E 3 Note:
Separation of the pFix15 isoforms at different flow rates (0.2, 0.5 and 1.0 mL min −1 ) on CIMac 6 µm plasmid DNA (pDNA) column.Separation was achieved using Gradient 2 as described in Section 2.2.TA B L E 3 Calculated plasmid DNA (pDNA) recoveries and resolution between isoforms of pFix15 at different flow rates using 6 µm CIMac pDNA column in Gradient 2. Total pDNA elution recovery was calculated from Equation (2), whereas OC and SC pDNA recoveries were obtained using Equation (3).Abbreviations: oc, open circular; SC, supercoiled.a Calculated according to Equation (3).b Calculated according to Equation (2).

TA B L E 4
Resolution and width at half of peak height for pFix5 obtained using CIMac plasmid DNA (pDNA) 1.4 µm and CIMac pDNA 6 µm column, and pFix15 obtained using CIMac pDNA 6 µm column, all in Gradient 3 at flow rate of 1 mL min −1 .

F
I G U R E Chromatograms representing the plasmid DNA (pDNA) isoform separation using an optimized guanidinium thiocyanate (GdnSCN) and guanidinium hydrochloride (GdnCl) gradient (Gradient 5) at a flow rate of 0.5 mL min −1 on a CIMac pDNA 6 µm column: (A) pFix15 standard; (B) pFix15 standard spiked with linear (LIN) pFix15.TA B L E 5 Resolution and relative percentages of plasmid DNA (pDNA) isoforms, determined from anion exchange (AEX) analysis using CIMac pDNA 6 µm column and Gradient 5 at 0.5 mL min −1 , for pDNA samples used in the study. pDNA Tomas Kostelec, Ana Ferjančič Budihna, Ana Benčina Kosmač, Mateja Pečnik and Jana Vidič from Sartorius BIA Separations d.o.o. are acknowledged for laboratory sup-port, discussions and other activities.We thank COBIK d.o.o., Ajdovščina, Slovenia, for providing us for E. coli pellets containing plasmid DNA used in the study.The content provided in this manuscript reflects the author's views only.Open access funding provided by CTK.C O N F L I C T O F I N T E R E S T S TAT E M E N TThe 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 TThe data that support the findings of this study are available from the corresponding author upon reasonable request.O R C I DUrh Černigoj https://orcid.org/0000-0001-9397-0759RE F E R E N C E S Relative percentages of plasmid DNA (pDNA) isoforms, determined from capillary gel electrophoresis (CGE) analysis and 260 nm/280 nm absorbance ratio in pDNA samples, used in the study.

sample pDNA size (kbp) OC % LIN % SC % R (OC-SC) a R (OC-LIN) a R (LIN-SC) a
Abbreviations: LIN, linear; oc, open circular; SC, supercoiled.a It was not possible to calculate representative OC-LIN or LIN-SC resolutions in pDNA samples without spiked LIN isoform.