Hepatitis C virus (HCV) is a major health problem worldwide. This virus infects approximately 3% of the global population. HCV causes chronic infection in most infected people and is one of the main reasons for liver transplantation [1]. To date, there are no preventive or therapeutic vaccines against this virus. The best antiviral treatment available, which is uses a combination of pegylated interferon and ribavirin, causes numerous adverse effects and is effective for only 50% of cases [2]. Therefore, the development of new treatments and prophylactic interventions for HCV is a priority.

DNA immunization can potentially induce both humoral and cellular immune responses thus, comprises an attractive approach for the development of an effective vaccine against HCV. Viral capsid proteins are able to interact with nucleic acids and naturally protect viral genetic material from degradation [3]. Recently, we demonstrated the safety of CIGB-230. This novel vaccine candidate is based on the mixture of pIDKE2, a plasmid for DNA immunization expressing HCV structural antigens, with a recombinant HCV core protein. Vaccination with CIGB-230 in HCV chronically-infected individuals was safe, well-tolerated, and did not impair the ability of the subjects to respond to non-HCV antigens [4].

A downstream process for pIDKE2 plasmid purification was developed previously. This process is based on alkaline lysis and tangential flow filtration, combined with size exclusion and perfusion chromatography using a porous particle matrix [5]. Although this process fulfills all regulatory requirements and renders pharmaceutical-grade plasmid DNA (pDNA) with 95% purity [5], its yields and recoveries are low. In an effort to improve these metrics, the downstream process was modeled to identify factors with the highest influences on production cost and annual amount of plasmid produced [6]. The model revealed that the reverse phase chromatography step was the most critical point because both the capacity of the resin under process conditions and the flow rate used for loading the sample were low. Due to the effect of salt on matrix topology, high concentrations of ammonium sulfate can shield the electrostatic repulsion between the outer phosphate groups of pDNA, allowing the interaction of the hydrophobic inner with the matrix and contributing to the differential retention of all pDNA species [7]. Therefore, the process capacity would likely be raised by increasing the concentration of ammonium sulfate. However, because the shape and structure of the pDNA also change with the salt concentration in the running buffer, the maximum useful concentration is 2 mol/L [6, 7].

Technological advances in fixed-bed liquid chromatography, based on favorable hydrodynamic properties in comparison with conventional supports, have introduced membrane and monolith technologies as alternatives to porous particles [8]. Membranes and monoliths differ fundamentally from porous particle media in having a mass transport mechanism that is established by convection instead of mainly by molecular diffusion. For membrane and monolith technologies, values for capacity and resolution do not depend on flow rates, even if velocities are 10-20 times higher than those commonly used in work with conventional porous particles. High flow rates offer numerous practical benefits, such as increased manufacturing productivity and accelerated process development and validation. Short process times may also be beneficial for biomolecules that are labile under the conditions used to conduct the purification step.

Membranes are very thin beds and usually consist of more than one layer, stacked one onto another. Solute transport to their binding sites takes place predominantly via convection, thereby reducing both process time and recovery liquid volume. The binding efficiency is generally independent of the feed flow rate over a wide range, therefore, very high flow rates may be used. Furthermore, membranes provide a reduced pressure drop along the chromatographic unit, allowing increased flow rates and consequently higher productivity [9]. Membrane chromatography is particularly suitable for pDNA. This biomolecule rarely enters the pores present in conventional bed chromatography media and only binds on the externally available surface area of such media. Therefore, for large pDNA molecules, the surface area available for binding is significantly greater for membranes than for conventional bed chromatography media. Another major advantage of membrane adsorbers is the relative ease for scale-up in comparison with packed beds. Notably, different separation chemistries are utilized in membrane chromatography [9]. Montesinos et al. [10] reported that, for ion-exchange membrane columns, the dynamic capacity for pDNA was found to be highly dependent on flow rates and concentrations.

A typical monolith is a continuous bed consisting of a single piece of a highly porous solid material [6]. Like membranes, the most important feature of this support is that the entirety of the mobile phase is forced to flow through the large interconnected pores of the monolith. Consequently, mass transport is enhanced by convection, dramatically reducing the long diffusion time required by conventional particle-package chromatographic columns [6, 11]. The "large" channels of about 2 μm in diameter enable the accessibility of all active surfaces for binding large biomolecules such as pDNA, resulting in high capacities. Moreover, high porosity (more than 50%) leads to a low pressure drop. Sousa et al. [12] reported that the analytical results and transfection studies performed with a pDNA sample purified using a carbonyldiimidazole monolith confirmed the suitability of that pDNA to be used in pharmaceutical applications. Additionally, other experiments confirmed the possibility of using single hydrophobic interaction chromatography (HIC) to separate supercoiled (SC) pDNA from open circular (OC) pDNA and other host contaminants [13].

In the present study, a Sartobind phenyl membrane column (hereinafter referred to as Sartobind) and a CIM^® C4-HLD monolithic column (hereinafter referred to as CIM^® C4-HLD), which both support hydrophobic-interaction chromatography, were evaluated to purify pIDKE2 plasmid in 2 mol/L ammonium sulfate. The pIDKE2 capacities under various flow conditions and the selectivity between SC and OC isoforms were determined. A mixed-level factorial experimental design was used to find the best process conditions (optimal conditions) for the CIM^® C4-HLD monolithic column in terms of both the pDNA purity in the elution fraction and the pDNA binding capacity. The optimal conditions were fixed on a CIM^® C4-HLD 800-mL monolithic column in a real production process, because the clinical trials for this vaccine candidate will require large amounts of plasmid production.

1 Experimental

1.1 Materials and reagents

Chemicals were purchased from Merck (Darmstadt, Germany). CIM^® C4-HLD (1 mL and 800 mL) and CIMac^TM pDNA-0.3 analytical columns were obtained from BIA Separations (Ajdovscina, Slovenia). Sartobind were obtained from Sartorius Stedim (Gottingen, Germany). The plasmid used in this study was the pIDKE2 (5.5 kbp) obtained by performing the Limonta et al. [6] method until the reverse phase chromatography step of the downstream purification process, achieving 85% purity of the product.

Hydrophobic phenyl groups were covalently attached to the base matrix (hydrophilic regenerated stabilized cellulose) for Sartobind. A high ligand-butyl was the hydrophobic ligand for CIM^® C4-HLD.

1.2 Breakthrough curve determinations

Initially, a flow distribution analysis was performed for all columns used in the experiments to check the leakage or possible bypass of the columns. This consisted of pulse response experiments with 0.5 mol/L phosphate buffer injected into the water mobile phase at 1 mL/min and 3 mL/min with the absorbance measured at 200 nm. The dynamic binding capacity measurements were performed using pure plasmid dissolved in the loading buffer (LB), pH (7.0±0.2), 25 mmol/L 2-amino-2-(hydroxymethyl)propane-1, 3-diol hydrochloride (Tris-HCl), 10 mmol/L ethylene diamine tetraacetic acid (EDTA) and 2 mol/L (NH₄)₂SO₄. The plasmid solution was loaded onto the Sartobind or the CIM^® C4-HLD columns until 60% (v/v) of the breakthrough and fractions of flow-through were collected every two column volumes (CVs). The CVs were 3 mL for the Sartobind column and 1 mL for the CIM^® C4-HLD column.

Each column was then washed with loading buffer, and the plasmid was subsequently eluted with the elution buffer (EB), pH (7.0±0.2), 25 mmol/L Tris-HCl and 10 mmol/L EDTA. All flow-through, washing and elution fractions were collected and analyzed by high performance liquid chromatography (HPLC). Finally, the cleaning in place process (CIP) of both matrices was performed for 1 h with 1 mol/L NaOH.

1.3 pIDKE2 purity determination

All pDNA analytics were performed on an Agilent HP 1200 Series chromatograph (Santa Clara, CA, USA), coupled with a multi-wavelength detector. The chromatographic separations were run on a CIMac^TM pDNA-0.3 analytical column (BIA Separations). The chromatographic separations were executed with 0.2 mol/L Tris-HCl (pH 8.0) as the binding buffer and 0.2 mol/L Tris-HCl containing 1 mol/L NaCl (pH 8.0). After the sample injection, the column was washed for 1 min with binding buffer, and a step gradient to 0.6 mol/L NaCl in 0.2 mol/L Tris-HCl (pH 8.0) was then applied for the next minute, followed by a 10-min shallow gradient from 0.6 mol/L NaCl to 0.7 mol/L NaCl in 0.2 mol/L Tris-HCl (pH 8.0) buffer. Finally, two step gradients were applied for column equilibration before the next chromatographic run: the first was to 0.2 mol/L Tris-HCl containing 1 mol/L NaCl (pH 8.0) over 1 min and the second was to 0.2 mol/L Tris-HCl (pH 8.0) over 3 min. The flow rate was 1.0 mL/min, and the injection volume was 100 μL. The corresponding chromatogram was monitored at 260 nm and 280 nm [14, 15].

Two peaks representing pDNA, corresponding to the OC and SC isoforms, were obtained during the shallow gradient, and the ratios of peak areas between the SC and OC were determined.

1.4 Experimental design

To optimize the pIDKE2 purification process, a screening, mixed-level factorial, 3² experimental design was created focusing on two factors: ammonium sulfate concentration (c(AS), mol/L) and pDNA mass concentration (C(pDNA), mg/mL). Experiments were performed using a CIM^® C4-HLD column (1 mL). The tested levels were selected as: 1.5 mol/L and 2 mol/L as the low and high levels for c(AS), respectively; 0.1 mg/mL and 0.3 mg/mL as the low and high levels for C(pDNA), respectively. The loading solution and elution fraction were analyzed for each experiment. Fractions were evaluated on a UV-VIS spectrophotometer (SmartSpec 3000, Biorad) to determine the mass concentration of pDNA. Elutions were analyzed on a pDNA analytical column to determine the ratio between SC and OC isoforms; this ratio was obtained via peak area comparison. The purity (% of SC), capacity (mg/mL) and recovery (%) were calculated for each experiment, and each experiment was performed with two replicates. The purity, recovery, and capacity were calculated as:

(1)

(2)

(3)

All pDNA analytics were performed on an Agilent HP 1200 Series chromatograph, coupled with a multi-wavelength detector.

2 Results and discussion

2.1 Breakthrough curves

Firstly, columns were checked for leaking or bypass by performing pulse response experiments to avoid using damaged columns for the experiments. Based on the resulting retention times and pulse shapes, no leakage or bypass appeared in any of the columns used.

Dynamic binding capacity experiments were then performed on a 1-mL CIM^® C4-HLD column; pDNA in LB was loaded onto the column, and fractions were collected every two CVs. As it can be seen from Fig. 1, a double breakthrough was obtained (marked with an arrow). The first breakthrough corresponds to the OC isoform, and the second breakthrough corresponds to the SC pDNA isoform. Their difference in hydrophobicity can explain this behavior [16]. Ammonium sulfate promotes the adsorption of hydrophilic species, and it contributes to the differential retention of OC and SC due to the effect that salt has on the matrix.

Fig. 1

Fig. 1 Breakthrough curves of CIM® C4-HLD and Sartobind at 2 CV/min The arrow shows the double breakthrough curves. The CVs were 3 mL for the Sartobind column and 1 mL for the CIM® C4-HLD column.

Chromatograms of OC and SC isoforms on CIM^® C4-HLD and Sartobind columns at 2 CV/min were shown in Fig. 2 for the initial sample for both matrices (Fig. 2a), flow-through fractions (Fig. 2b and Fig. 2c) and the elution fractions (Fig. 2d). An analysis of the flow-through fractions from the CIM^® C4-HLD column showed that only OC pDNA was present in the first seven fractions (fraction seven for Fig. 2b) and SC pDNA appeared only in the last two fractions (nine and ten) (fraction nine for Fig. 2c). In contrast, OC pDNA and SC pNDA were both presented in all flow-through fractions for the Sartobind column.

Fig. 2

Fig. 2 Chromatograms of OC and SC isoforms on CIM® C4-HLD and Sartobind columns at 2 CV/min

Interestingly, the concentrations of SC in the flow-through fractions on the CIM^® C4-HLD column increased over the fractions, meaning that the SC isoform acts as a displacer of the OC isoform (Fig. 2b and 2c). This phenomenon of sample displacement for pDNA on monolith columns has been previously described in the literature [16]. In contrast, on Sartobind under the same conditions, both OC and SC pDNA were present from the beginning in the flow-through fractions, indicating that no OC pDNA displacement occurred in this column.

Additionally, breakthrough curves on the Sartobind column were standard (no double breakthrough occurred), indicating that no displacement took place under our chosen experimental conditions. This result suggests that no additional purification of the SC isoform occurred on this column. Furthermore, the ratio between the SC and OC isoforms in the eluted fraction was only slightly improved on the Sartobind column (Table 1).

表 1 (Table 1)

Table 1 Analytical results of samples at two different flow rates Column Flow rate/ (CV/min) Fraction Peak area of OC Ratio of SC to OC loading solution 243.8 6.5 CIM® C4-HLD 1 fraction 9 308.0 fraction 10 298.4 elution 13.0 CIM® C4-HLD 2 fraction 9 256.7 fraction 10 270.2 elution 13.7 Sartobind 1 fraction 8 175.6 fraction 9 214.8 elution 7.8 Sartobind 2 fraction 8 209.7 fraction 9 213.6 elution 7.6

Table 1 Analytical results of samples at two different flow rates

It can also be observed from Fig. 1 that the pressure started to increase simultaneously with plasmid loading for both columns, because the pores were filled gradually. Later, when the surface was already covered by pDNA and the pores were already occupied by adsorbed pDNA, the additional pDNA needed to pass through the column through a reduced pore size, resulting in a larger pressure drop. Moreover, the pore size for the Sartobind is higher than 3 μm, but, for CIM^® C4-HLD, it is only 2 μm. The difference in pore size influences the pressure drop. Higher pressure drop values could be observed in CIM^® C4-HLD than in Sartobind at similar column volumes.

The dynamic binding capacities for both the monolith and membrane columns were comparable, but the displacement phenomena made it difficult to determine the exact capacity for pDNA sample loading on the monolithic column. Additionally, the purity ratios between the SC and OC in the elution fractions were improved on both monolith or membrane column, but the ratio was more than doubled for the CIM^® C4-HLD column (Table 1). This drastic increase is a consequence of the displacement phenomena that occurred only in the monolithic column under our selected conditions.

Table 1 shows the integrated OC peak area of the initial and flow-through samples on the CIM^® C4-HLD column and Sartobind column at two different flow rates.

Breakthrough curves for the CIM^® C4-HLD column and Sartobind column at two different flow rates are shown in Fig. 3. The dynamic binding capacity slightly diminished as the flow rate increased for both columns, likely due to the convection mass transport that is predominant in these matrices.

Fig. 3

Fig. 3 Breakthrough curves of the (a) CIM® C4-HLD column and (b) Sartobind column at two different flow rates

The plasmid quantities adsorbed onto the surface and eluted from the column were evaluated from fraction analyses to calculate the recovery values. Mean results for the CIM^® C4-HLD and Sartobind columns are shown in Table 2.

表 2 (Table 2)

Table 2 Loading of pDNA on a CIM® C4-HLD column and Sartobind column at three different flow rates Column Flow rate/ (CV/min) Mass concentrations (SDs, n=2)/(mg/mL) Recovery/ % pDNA bound pDNA eluted CIM® C4-HLD 0.5 2.36 (0.01) 2.32 (0.13) 98.27 1.0 2.10 (0.26) 2.18 (0.11) 103.81 2.0 2.03 (0.05) 1.95 (0.08) 95.88 Sartobind 0.5 1.99 (0.02) 1.67 (0.01) 83.92 membrane 1.0 1.67 (0.16) 1.72 (0.18) 102.99 2.0 1.61 (0.01) 1.81 (0.02) 112.42

Table 2 Loading of pDNA on a CIM® C4-HLD column and Sartobind column at three different flow rates

Considering the possible experimental errors (those emerging from pipetting inconsistencies, fraction volume measuring, etc.) that are multiplied over the course of the experiment, the recovery values were very high (above 95%) for both systems, except at the flow rate of 0.5 CV/min for the membrane column.

2.2 Design of experiments to model the purification process with a monolith

Due to the recovery and especially final purity (SC/OC ratio) results obtained from the pDNA loading experiments on Sartobind and CIM technologies, the CIM^® C4-HLD column was selected for use in additional examinations. To further optimize the loading conditions, a design of experiments (DoE) evaluating the pDNA concentration and ammonium sulfate concentration in the loading solution was performed via a set of screening mixed-level factorial 3² tests.

The initial samples prepared for the optimization experiments were first analyzed by the HPLC method, resulting in the purities between 80.87%-85.24% and SC/OC ratios from 4.23 to 5.78 (Table 3). The main sample impurity was the OC isoform.

表 3 (Table 3)

Table 3 SC/OC ratios and purities of initial samples Condition c(AS)/ (mol/L) C(pDNA)/ (mg/mL) SC/OC ratio Purity of SC/% 1 2.00 0.10 4.23 80.87 2 2.00 0.30 4.71 82.49 3 1.75 0.10 4.36 81.35 4 1.75 0.20 4.94 83.16 5 1.75 0.30 4.56 82.02 6 1.50 0.10 5.26 84.01 7 1.50 0.30 5.78 85.24

Table 3 SC/OC ratios and purities of initial samples

Data from the elution streams are shown in Table 4. The SC/OC ratios, purities of SC (%), capacities of the monolith for pDNA and recoveries were studied as dependent variables.

表 4 (Table 4)

Table 4 Response of the SC/OC ratios, purities, capacities and recoveries obtained for each run defined by the DoE Run c(AS)/ (mol/L) C(pDNA)/ (mg/mL) SC/OC ratio Purity of SC/% Capacity/ (mg/mL) Recovery/ % 1 2.00 0.10 12.82 92.76 1.99 100.0 2 2.00 0.30 16.14 94.17 1.95 86.0 3 1.75 0.10 93.98 98.95 1.65 99.5 4 1.75 0.20 64.31 98.47 1.83 85.9 5 1.75 0.30 38.58 97.47 1.19 90.7 6 1.50 0.10 146.88 99.32 0.90 100.0 7 1.50 0.30 96.32 98.97 0.92 90.3 8 2.00 0.10 25.24 96.19 1.97 93.3 9 2.00 0.30 14.10 93.38 1.99 80.9 10 1.75 0.10 94.98 98.96 1.86 85.6 11 1.75 0.20 67.72 98.54 1.85 91.1 12 1.75 0.30 43.13 97.73 1.11 100.0 13 1.50 0.10 174.24 99.43 0.90 100.0 14 1.50 0.30 131.16 99.24 0.89 89.3

Table 4 Response of the SC/OC ratios, purities, capacities and recoveries obtained for each run defined by the DoE

A higher SC/OC ratio and higher level of purity of SC (%) were obtained in the elution stream for all conditions, as compared with those of the initial samples (Tables 3 and 4). Purities greater than 92% (92.76%-99.43%) were achieved. The data were analyzed using Software Statgraphics Centurion v15.0 to determine which independent variables affect the dependent variables.

A Pareto's diagram of the data shows that the main factor affecting the SC/OC ratio is the ammonium sulfate concentration, although the pDNA concentration of the loading sample and the product of both variables also have a statistically significant impact at a 95% confidence level (Fig. 4a). The fitted model explains 96.7% of the variability in the SC/OC rate and has a mean absolute error of 7.71 (average value of the residuals). This result suggests that a higher proportion of SC pDNA can be obtained by diminishing the ammonium sulfate concentration.

Fig. 4

Fig. 4 Standardized Pareto's diagrams of (a) SC/OC ratio, (b) purity and (c) capacity A: c(AS)/(mol/L); B: C(pDNA)/(mg/mL); AB: A×B; AA: A2.

Similar results were obtained from a purity analysis (Fig. 4b). A Pareto's diagram shows that the ammonium sulfate concentration has a significant influence on the purity, for a 95% confidence level. The fitted model explains 91.4% of the variability with a mean absolute error of 0.46% (average value of the residuals). As above, a low level of ammonium sulfate increases the elution stream purity, and this finding agrees with the SC to OC pDNA proportion in the elution stream.

A Pareto's diagram for capacity (Fig. 4c) shows that the ammonium sulfate concentration also influences capacity, at a 95% confidence level. We adjusted a model with a coefficient of determination (R²) of 77.7% and a mean absolute error of 0.17 mg/mL. The capacity increased with increasing ammonium sulfate concentrations.

Recoveries were higher than 80%, but there was not a statistically significant relationship among the ammonium sulfate concentration, pDNA concentration and recovery at a 95% confidence level.

To integrate the effect of the ammonium sulfate concentration and pDNA concentration on response variables, a multiple response optimization was performed, considering the mathematical function named desirability as the optimization function. Statgraphics Centurion XVI uses this criterion to determine experimental factor configurations that cover the desired characteristics for several simultaneous responses. A desirability function is calculated from the response models fitted for each variable. Multiple optimizations were performed to maximize SC/OC, capacity and recovery (Fig. 5). As a high SC/OC rate is determinant for this vaccine, the highest impact factor on the global desirability function was selected for this variable. The optimum conditions that reported the highest desirability and maximized results were 1.56 mol/L for c(AS) and 0.1 mg/mL for C(pDNA). A SC/OC ratio of 144.1 (equivalent to a purity of 99.3%), capacity of 1.1 mg/mL and a recovery of 97.8% could be reached under these conditions.

Fig. 5

Fig. 5 Surface responses for multiple response optimization

The association of recovery and purity output variables assures the maintenance of the conditions required by the regulatory agencies regarding DNA vaccines. Three 800-mL batches were produced using the suggested conditions to supply the pDNA active ingredient for use in a future clinical trial. The certificate of analysis of the final purified pDNA indicated that each batch had a pIDKE2 purity of more than 95%. The content of genomic DNA was lower than 5 μg per dose, RNA was not detectable by agarose gel electrophoresis, the endotoxin content was below 5.0 endotoxin unit (EU) per body weight (kg), and the protein content was 1.4 μg per dose, which is lower than the established limit (data not shown). These results assure the conditions required by the regulatory agencies regarding DNA vaccines. Furthermore, the results agree with those of previous and our DoE, so the process could be scaled-up at this interval.

3 Conclusions

Here, we performed a comparison between CIM^® C4-HLD and Sartobind columns for use in the pDNA polishing step. The dynamic binding capacities, SC/OC ratios and recoveries at two different flow rates were determined. A minor decrease in the dynamic binding capacity was observed with the doubled flow rate for both the CIM^® C4-HLD and Sartobind columns, demonstrating that convection was the predominant mass transport in these columns. During the capacity measurements on the CIM^® C4-HLD column, a double breakthrough was observed; this points to the previously described self-displacement phenomena of pDNA isoforms, which enabled a very high purity (the SC/OC ratio was doubled) of the eluted fraction in comparison with the elution from the Sartobind column where no self-displacement was detected.

The process was scaled up to an 800-mL CIM^® C4-HLD monolithic column, according to DoE results, to produce enough pharmaceutical-grade plasmid for use in clinical trials. Importantly, the optimized process still fulfills the conditions required by the regulatory agencies regarding DNA vaccines.

References

[1]	Thitinan S, McConville J T. Int J Pharm, 2009, 369(1/2): 121.
[2]	Syed E, Rahbin N, Weiland O, et al. Scand J Gastroenterol, 2008, 43(11): 1378. doi: 10.1080/00365520802245395
[3]	Lian S Z, Shan Q, Tao Y Z, et al. World J Gastroenterol, 2000, 6(2): 239.
[4]	Castellanos M, Cinza Z, Dorta Z, et al. J Gene Med, 2010, 12(1): 107. doi: 10.1002/jgm.v12:1
[5]	Rey I, Pupo M, Márquez G, et al. Biopharm Int, 2008, 21(9): 38.
[6]	Limonta M, Lendero N, Vidic U, et al. Biochem Eng J, 2013, 80(11): 14.
[7]	Limonta M, Zumalacárregui L, Soler D. Chinese Journal of Chromatography, 2012, 30(5): 522.
[8]	Rajamanickam V, Herwig C, Spadiut O. Chromatography, 2015, 2(2).
[9]	Kuczewski M, Fraud N, Faber R, et al. Biotechnol Bioeng, 2010, 105(2): 296. doi: 10.1002/bit.v105:2
[10]	Montesinos R, Vega J, Ortega J, et al. Biotechnol Prog, 2007, 23(4): 881. doi: 10.1021/bp070054d
[11]	Urthaler J, Shlegl R, Podgornik A, et al. J Chromatogr A, 2005, 1065: 93. doi: 10.1016/j.chroma.2004.12.007
[12]	Sousa A, Tomaz C T, Sousa F, et al. J Chromatogr A, 2011, 1218(46): 8333. doi: 10.1016/j.chroma.2011.09.033
[13]	Bo H, Wang J, Chen Q, et al. Pharm Biol, 2013, 51(1): 42. doi: 10.3109/13880209.2012.703678
[14]	Cernigoj U, Vidic U, Barut M, et al. J Chromatogr A, 2013, 1281: 87. doi: 10.1016/j.chroma.2013.01.058
[15]	BIA Separations. pDNA Downstream Processing Using CIM Monoliths. [2016-05-25]. http://www.hp-ne.com/wp-content/uploads/2013/05/BIA_Sep_Brochure_pDNA_purification.pdf
[16]	Cernigoj U, Martinuc U, Cardoso S, et al. J Chromatogr A, 2015, 1414: 103. doi: 10.1016/j.chroma.2015.08.035