Journal of Chromatography B, 804 (2004) 327–335

Purification of pharmaceutical-grade plasmid DNA by anion-exchange

chromatography in an RNase-free process
Alex Eon-Duval∗ , Gemma Burke
BioPharmaceutical Development, GlaxoSmithKline Research and Development Ltd., South Eden Park Road, Beckenham, Kent BR3 3BS, UK

Received 17 October 2003; received in revised form 15 January 2004; accepted 19 January 2004

Abstract

Anion-exchange is the most popular chromatography technique in plasmid DNA purification. However, poor resolution of plasmid DNA
from RNA often results in the addition of bovine-derived ribonuclease (RNase) A to degrade RNA impurities which raises regulatory concerns
for the production of pharmaceutical-grade plasmid DNA. Low capacity for plasmid of most commercial media is another issue affecting the
suitability of anion-exchange chromatography for large-scale processing. This study reports the use of anion-exchange chromatography to
remove RNA in an RNase-free plasmid purification process. Resolution was achieved through careful selection of adsorbent and operating
conditions as well as RNA reduction steps before chromatography. Dynamic capacity for plasmid was significantly increased (to 3.0 mg/ml)
so that it is now possible to envisage the large-scale manufacturing of therapeutic-grade plasmid DNA in the absence of added RNase using
anion-exchange chromatography as a polishing step.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Purification; Plasmid DNA; RNase

1. Introduction size-exclusion [6–8], reversed-phase [9,10], hydrophobic

interaction [11–13], hydroxyapatite [14–16], silica [17–19]
In recent years, gene therapy and DNA vaccination [1,2] and triple-helix affinity [20,21]. Anion-exchange remains
have attracted a lot of interest as new ways of preventing the most popular chromatography technique as it offers
or treating disease through gene transfer. Non-viral tech- the advantages of rapid separation, no solvent requirement,
niques including the intra-muscular injection of naked plas- sanitisation with sodium hydroxide and a wide selection of
mid DNA or gene gun delivery of plasmid DNA coated onto industrial media. DNA is a polyanionic molecule due to the
gold particles into the epidermis appear very attractive as presence of phosphate groups on the nucleic acid backbone
they offer several advantages over viral vectors, especially and is therefore conveniently captured on a resin derivatised
low immunogenicity, better safety profile and easier manu- with positively charged functional groups. Protocols for
facture. There is therefore a need for large-scale processes both packed [22–24] and expanded bed [25–27] operation
to manufacture plasmid DNA of a high level of purity for have been developed.
use as therapeutic agent. At the same time, the requirements Current plasmid purification methods suffer from several
of regulatory agencies regarding purity, potency, safety and drawbacks that make them unsuitable for the manufacture
efficacy must be met [3–5]. of pharmaceutical-grade plasmid DNA. They often involve
Plasmid purification strategies usually involve at least the use of solvents (ethanol, isopropanol), toxic chemicals
one chromatography step either to capture the plas- (cesium chloride, ethidium bromide, phenol, chloroform)
mid or more often, for polishing. Techniques include: or animal-derived enzymes (ribonuclease A, lysozyme, pro-
teinase K) that are either not approved or not recommended
by regulatory agencies. Removal of key impurities such
∗ Corresponding author. Present address: Immunotherapeutics Depart-
as chromosomal DNA, RNA, proteins and endotoxins is
ment, GlaxoSmithKline Medicines Research Centre, Gunnels Wood Road,
Stevenage, Hertfordshire SG1 2NY, UK. Tel.: +44-1438-764-977;
often insufficient. Finally, many techniques were designed
fax: +44-1438-768-091. to produce small quantities of plasmid DNA for laboratory
E-mail address: [email protected] (A. Eon-Duval). use and are not suitable for the production of therapeutic

1570-0232/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.jchromb.2004.01.033
328 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335

material at large scale. A common limitation of commer- 996 Photodiode Array Detector from Waters (Watford,
cial chromatography media is the low capacity for plasmid UK). The tangential flow filtration Centramate rig was pur-
DNA due to the inability of large molecules to penetrate chased from PallFiltron (Portsmouth, UK) together with the
porous beads [22,28]. Omega 100 K Centramate 0.1 ft2 open channel membrane
Resolution of plasmid DNA from RNA, especially high and 500 ml ultrareservoir. The centrifuge used was a Sorvall
molecular weight molecules by anion-exchange chromatog- RC5C Plus from Kendro (Newtown, CT, USA). Sartobran
raphy is insufficient [24,26]. As a result, ribonuclease A P 0.45/0.2 ␮m sterile capsules were purchased from Sarto-
(RNase A) is commonly added to degrade RNA impuri- rius (Göttingen, Germany). The pump used for tangential
ties [22–25]. Apart from the obvious costs involved, there flow filtration was a WatsonMarlow (Falmouth, UK) 505S.
are concerns regarding RNase because it is purified from The 10 l bioreactor was purchased from New Brunswick
bovine pancreas and regulatory authorities recommend that Discovery. Slide-A-Lyzer 3–15 ml, 10,000 MWCO dialysis
bovine derived materials should be avoided in the production cassettes were bought from Pierce (Rockford, IL, USA).
of bio-therapeutics following the outbreak of new-variant All chemicals were bought from Merck (Poole, UK) and
Creutzfeld–Jakob disease in the UK [29]. were analytical grade.
In this paper, we investigate the suitability of anion-
exchange chromatography to purify plasmid DNA in the 2.2. Methods
absence of exogenous RNase. RNA reduction steps were
necessary prior to chromatography to allow complete 2.2.1. Bacterial cell culture
resolution from plasmid DNA. We evaluated a range of A 10 l bioreactor (5 l working volume) was inoculated
anion-exchangers for their ability to resolve plasmid DNA with 300 ml of an overnight shake-flask culture (230 rpm,
from RNA, high capacity and robustness. This allowed us 37 ◦ C). The medium was a complex growth medium
to identify one adsorbent with suitable characteristics for (in-house formulation) supplemented with 50 ␮g/ml
the large-scale production of pharmaceutical-grade plasmid kanamycin. The fermenter culture was grown at 37 ◦ C, pH
DNA. 7. Dissolved oxygen was set to 30% of the saturation value
and controlled by changing air flow and agitation speed. The
culture was harvested when a drop in oxygen consumption
2. Experimental was observed. Cells were centrifuged at 3200 × g for 20 min
on a MSE Mistral 6000 and the supernatant was discarded.
2.1. Materials
2.2.2. Lysis and plasmid purification
The E. coli host strain DH5␣ containing a 5.9 kb plas- Bacteria were lysed using a modification of the alkaline
mid encoding hepatitis B surface and core antigen was method described by Birnboim and Doly [30]. Bacteria cell
developed in-house. Anion-exchange chromatography me- paste (typically 100 g) was re-suspended in 500 ml of 25 mM
dia were purchased as follows: Q Sepharose Fast Flow, Tris, 10 mM EDTA, 55 mM dextrose, pH 8.0. Lysis was per-
DEAE Sepharose Fast Flow and SOURCE 30Q from formed by adding 834 ml of 0.96% (w/v) NaOH and 166 ml
Amersham Biosciences (Little Chalfont, UK), Q Ceramic of 6% (w/v) SDS for 30 min at 4 ◦ C. The lysate was neu-
HyperD F and DEAE Ceramic HyperD F from BioSepra tralised with 500 ml of 3 M potassium acetate for 30 min
(Cergy-Saint-Christophe, France), Fractogel EMD TMAE at 4 ◦ C. RNA levels were reduced by addition of 1 vol. of
(M), Fractogel EMD DMAE (M) and Fractogel EMD 5 M calcium chloride to 4 vol. of neutralised lysate. Pre-
DEAE (M) from Merck KGaA (Darmstadt, Germany), cipitated material including cell debris, chromosomal DNA,
Macro-Prep DEAE, Macro-Prep High Q and UNOSphere high molecular weight RNA and proteins was removed by
Q from BioRad (Hercules, CA, USA), POROS 50HQ, centrifugation at 8000 × g for 10 min followed by clarifica-
POROS 50PI and POROS 50D from Applied Biosystems tion through a 0.45/0.2 ␮m Sartobran P filter. The clarified
(Warrington, UK), Toyopearl DEAE 650 M and Toyopearl lysate was processed by tangential flow filtration (TFF) to
SuperQ M from TosoHaas (Stuttgart, Germany) and UFC remove remaining proteins and low molecular weight RNA,
DEAE and UFC PEI from UpFront Chromatography concentrate and buffer exchange. Clarified lysate was con-
(Copenhagen, Denmark). Analytical HPLC columns were centrated to 100 ml on a Pall Filtron Centramate rig fitted
obtained as follows: DNAPac PA-100, 4 mm × 250 mm with a 1 ft2 , 100 K Centramate membrane. Cross-flow rate
from Dionex (Sunnyvale, CA, USA) and TSKgel G-DNA was maintained at 1 l/(min ft2 ) using a WatsonMarlow 505S
PW, 7.8 mm × 300 mm and TSKgel G6000 PWXL, pump. Trans-membrane pressure was set at 35 kPa. Diafil-
7.8 mm × 300 mm from Phenomenex (Macclesfield, UK). tration with 50 vol. of 50 mM Tris, 0.54 M NaCl, pH 8.5 was
Preparative chromatography studies were conducted on an then carried out under the same conditions. The retentate
Äkta-FPLC system with UFC900 control unit, P920 pumps from TFF was used as the load material for anion-exchange
and Frac950 fraction collector from Amersham Biosciences chromatography after dialysis into a suitable buffer. Dialy-
(Little Chalfont, UK). Analytical HPLC equipment con- sis was performed against two changes of loading buffer in
sisted of a Waters 2690 Separations module and Waters 10,000 MWCO Slide-A-Lyzer cassettes.
 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335 329

2.2.3. Analytical chromatography 2.2.5. Agarose gel electrophoresis

Plasmid concentration was determined by anion-exchange A 0.8% agarose gel containing 0.5 ␮g/ml ethidium bro-
HPLC (IE-HPLC) on a Dionex DNAPac PA-100 column. mide was run in a horizontal gel electrophoresis unit
Injection of 10 ␮l of sample was followed by a linear (Mini-Sub DNA cell, BioRad). The running buffer was TAE
gradient of 0.7–0.9 M KCl in 20 mM Tris, pH 7.5 for (40 mM Tris, 20 mM acetic acid, 1 mM EDTA, pH 8.0).
9.5 min. The column was re-equilibrated with 0.7 M KCl Electrophoresis was carried out at 100 V for 1 h on an Amer-
in 20 mM Tris, pH 7.5 for 4.5 min. The flow rate was sham Pharmacia Biotech power supply unit ECPS3000/
1 ml/min. Absorbance was monitored at 260 nm. Plasmid 150.
standards of concentrations ranging from 5 to 275 ␮g/ml
were injected on the column and the surface area under
600
the peak of absorbance at 260 nm was plotted against pH 5.5 1000
the plasmid concentration. Samples of unknown plasmid 400
800
concentration were injected and from the resulting peak 200 600

area, concentration was determined from the standard 0

600
400

curve. pH 6.5 1000

400
800
RNA levels were measured as a percentage of RNA 200 600
Sodium Chloride conc (mM)

to plasmid by size-exclusion HPLC (SEC-HPLC). Af- 0 400

Abs 254nm (mAU)

ter the tangential flow filtration step samples were free 600
pH 7.5 1000
of contaminant chromosomal DNA and proteins so there 400
800
was no interference with the assay [31]. The only sub- 200 600

stances absorbing at 260 nm present at this stage were 0 400

600
plasmid and RNA. Injection of 100 ␮l of sample was 400
pH 8.5 1000
800
followed by isocratic elution in 0.1 M Tris, 0.3 M NaCl, 200 600
1 mM EDTA, pH 7.5 for 70 min at 0.5 ml/min on TSKgel 0 400
600
G-DNA PW followed by TSKgel G6000 PWXL in se- pH 9.5 1000
400
ries. Absorbance was monitored at 260 nm. The surface 800
200 600
area under the first peak (plasmid) and the second peak
0 400
(RNA) was integrated and the percentage of each peak
calculated. 40 45 50 55 60 65 70 75 80 85 90
(a) Time (min)
2.2.4. Preparative anion-exchange chromatography
HR5/5 Amersham Biosciences columns were packed 600
pH 5.5 1000
with 1 ml of anion-exchange adsorbent according to 400
750
manufacturer’s instructions. All columns were equilibrated 200 500
with 5 ml of loading buffer (typically 50 mM sodium phos- 0
250
600 0
phate, pH 7.0) at a linear flow rate of 150 cm/h. Dialysed pH 6.5 1000
400
retentate from TFF was injected on each column (typically 750
200 500
0.5 ml except for breakthrough experiments). The columns
Sodium Chloride conc (mM)

250
0
Abs 254nm (mAU)

were washed with 5 ml of loading buffer. 600 0

pH 7.5 1000
For the screening experiments, a 40 ml gradient of 0 to 400
750
1 M NaCl in loading buffer was applied to resolve RNA and 200 500
plasmid. For the step elution experiments, RNA was eluted 0
250
600 0
in one 5 ml step of loading buffer containing sodium chloride pH 8.5 1000
400
at various concentrations, followed by plasmid elution in a 750
200 500
10 ml gradient to 1 M NaCl. 250
0
For the breakthrough experiments, the TFF retentate was 1000 0
750 pH 9.5 1000
dialysed into loading buffer containing sodium chloride at 750
500
the concentration determined in the step experiment to pre- 250
500
250
vent binding of RNA to the column without affecting plas- 0
0
mid recovery. The column was washed with 5 ml of dialy- 10 20 30 40 50 60 70 80 90
sis buffer followed by plasmid elution in a gradient to 1 M (b) Time (min)
NaCl.
All columns were then regenerated with 4 ml of 1 M NaCl Fig. 1. Elution profile of plasmid and RNA on a 1 ml Fractogel EMD
TMAE (a) and Fractogel EMD DEAE (b) column using a linear NaCl
in loading buffer, sanitised with 5 ml of 0.5 M NaOH fol-
gradient at five different buffer pH values: 5.5, 6.5, 7.5, 8.5 and 9.5. The
lowed by 5 ml re-equilibration in loading buffer. Absorbance first peak represents the elution of RNA while plasmid elutes in a second
was monitored at 254 nm. peak.
330 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335

3. Results late, agarose, polystyrene divinylbenzene, acrylamide) but

always sanitisable with sodium hydroxide.
3.1. Anion-exchange chromatography: effect of pH Only four anion-exchangers gave baseline resolution of
RNA from plasmid: Fractogel DEAE (Fig. 2a), Q Sepharose
To determine the effect of pH on the performance of FF (Fig. 2b), POROS 50HQ (Fig. 2c) and Q Ceramic Hy-
a selected strong (Fractogel TMAE) and weak (Fractogel perD F (Fig. 2d).
DEAE) anion-exchanger, both columns were run in a univer-
sal buffer of 20 mM Tris, 20 mM MES, 20 mM glycine at pH 3.3. Anion-exchange chromatography: step elution of RNA
5.5, 6.5, 7.5, 8.5 and 9.5. Results are shown in Fig. 1a (Frac-
togel TMAE) and Fig. 1b (Fractogel DEAE). As expected, A step elution was devised to remove RNA adsorbed to
pH had a very limited effect on plasmid retention on the the anion-exchanger followed by plasmid elution in a gradi-
strong anion-exchanger within the range investigated. Reso- ent. The conditions required for complete RNA elution in the
lution of RNA (first peak) from plasmid (second peak) was step were determined by varying the salt concentration in the
insufficient. The weak anion-exchanger (Fractogel DEAE) step. Under optimised conditions, only RNA eluted in the
gave baseline separation of RNA from plasmid in the range step while plasmid essentially free of RNA was recovered in
of pH between 5.5 and 7.5. As pH was increased above 7.5, the gradient. The levels of plasmid and RNA in the step and
not only was plasmid retention significantly reduced but res- gradient fractions were measured by size-exclusion HPLC
olution was also lost. and expressed as a percentage of each species present. Re-
Although results obtained from an experiment with just sults in Table 2 show that all four selected anion-exchangers
one weak and one strong anion-exchanger cannot be directly gave complete RNA removal and recovery of almost pure
generalised to all other adsorbents, a phosphate buffer at plasmid under optimised conditions.
pH 7 was selected for all subsequent experiments so that
separation of RNA from plasmid was maximised. 3.4. Anion-exchange chromatography: breakthrough
experiments
3.2. Anion-exchange chromatography: screening of
adsorbents Sequential elution of RNA and plasmid has now been
achieved by selecting the anion-exchanger together with a
Table 1 shows the characteristics of 18 anion-exchange suitable pH and salt concentration for RNA elution. How-
chromatography media screened, all of which are suitable ever, the fact that RNA binds to the anion-exchanger means
for large-scale manufacturing. Pore sizes of 50–100 ␮m that fewer binding sites are available for plasmid capture and
were preferred for better comparability and potential as a result dynamic capacity for plasmid is reduced. By se-
scale-up. A broad range of weak (DEAE, DEAP, DMAE, lecting loading conditions that prevent binding of RNA, plas-
PEI non-Q) and strong (Q, TMAE, PEI Q) functional mid capture can be maximised and dynamic capacity signif-
groups was included. Matrices were diverse (methacry- icantly increased. This was achieved by optimising loading

Table 1
Characteristics of the commercial anion-exchange chromatography media evaluated
Support Functional group Matrix Particle size (␮m) Pore size (Å) Ligand density (␮eq./ml)

Q Sepharose FF Q Agarose 90 1900 180–250

DEAE Sepharose FF DEAE Agarose 90 1900 110–160
SOURCE 30Q Q PSDBa 30 20–1000 NA
Q Ceramic HyperD F Q Hydrogel/ceramic 50 3000 >200
DEAE Ceramic HyperD F DEAE Hydrogel/ceramic 50 3000 >180
Fractogel EMD TMAE TMAE Methacrylate 40–90 800 NA
Fractogel EMD DMAE DMAE Methacrylate 40–90 800 NA
Fractogel EMD DEAE DEAE Methacrylate 40–90 800 NA
Macro-Prep DEAE DEAE Methacrylate 50 1000 175
Macro-Prep High Q Q Methacrylate 50 1000 400
UNO Sphere Q Q Acrylamide 120 10000 NA
POROS 50HQ PEI Q PSDBa 50 <8000 NA
POROS 50PI PEI non-Q PSDBa 50 <8000 NA
POROS 50D DEAP PSDBa 50 <8000 NA
Toyopearl DEAE 650 DEAE Methacrylate 40–90 1000 80–120
Toyopearl SuperQ 650 Q Methacrylate 40–90 1000 200–300
UFC DEAE DEAE Agarose 100–300 NA 50–100
UFC PEI PEI Agarose 100–300 NA 180–200
NA: not available.
a Polystyrene divinylbenzene.
 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335 331

350 1200 100 1200

Sodium Chloride conc (mM)

Sodium Chloride conc (mM)

300 1000 1000

80
Abs 254nm (mAU)

Abs 254nm (mAU)

250 800
800
60 Plasmid
200
600 600
Plasmid 40
150
RNA 400
400
100 RNA
20
200 200
50
0 0
0 0
0 20 40 60 80 100 0 20 40 60 80 100
(a) Time (min) (b) Time (min)

500 1200 400 1200

450 350
Sodium Chloride conc (mM)
Sodium Chloride conc (mM)

1000 1000
400
300
Abs 254nm (mAU)
Abs 254nm (mAU)

350 800 800

250
300
600 200 600
250
200 Plasmid 400 150 400
150 Plasmid
100
100 200 RNA 200
RNA 50
50
0 0
0 0
0 20 40 60 80 100 0 20 40 60 80 100
(c) Time (min) (d) Time (min)

Fig. 2. Elution profile of plasmid and RNA on a 1ml Fractogel EMD DEAE (a), Q Sepharose FF (b), POROS 50HQ (c) and Q Ceramic HyperD F (d)
column using a linear NaCl gradient in 50 mM sodium phosphate buffer at pH 7.0.

Table 2
Sequential elution of RNA in a step and plasmid in a gradient by anion-exchange chromatography
Anion-exchange column Step NaCl concentration (M) SEC-HPLC

RNA content in step fraction (%) Plasmid content in gradient fraction (%)

Q Sepharose FF 0.67 100.0 94.5

Q Ceramic HyperD F 0.67 100.0 92.8
Fractogel DEAE 0.70 100.0 99.5
POROS 50HQ 0.81 100.0 97.7
0.5 ml of TFF retentate containing plasmid and RNA was loaded on a 1 ml anion-exchange column. RNA was recovered by step elution with NaCl in
50 mM sodium phosphate buffer at pH 7.0. Plasmid was then eluted with a NaCl gradient to 1 M in 50 mM sodium phosphate at pH 7.0. Step and
gradient elution fractions were analysed by size-exclusion HPLC (SEC-HPLC) and the percentage of plasmid and RNA determined in each fraction.

salt concentration in a breakthrough experiment where RNA

and plasmid levels were measured in the column flowthrough Table 3
during loading by size-exclusion HPLC. Dynamic capacity Dynamic capacity for plasmid of several anion-exchangers when loading
in high salt
was determined as the amount of plasmid loaded when plas-
mid starts to appear in the column breakthrough. Table 3 Anion-exchange column NaCl concentration Dynamic capacity
shows that loading salt concentrations determined in the pre- in the load (M) (mg plasmid/ml gel)
vious section were too high and salt concentration in the load Q Sepharose FF 0.61 0.72
Q Ceramic HyperD F 0.63 >5.3
had to be reduced to prevent early plasmid breakthrough.
Fractogel DEAE 0.69 2.45
Only three of the four selected anion-exchangers had suffi- POROS 50HQ 0.72 2.12
cient capacity: Q Ceramic HyperD F, Fractogel DEAE and
Each 1 ml anion-exchange column was loaded with a mixture of plasmid
POROS 50HQ. Q Ceramic HyperD F had a dynamic capac-
and RNA in high NaCl in 50 mM sodium phosphate buffer at pH 7.0.
ity of >5.3 mg/ml when loading in 50 mM phosphate, 0.63 M Dynamic capacity was determined as the amount of plasmid loaded when
NaCl, pH 7 at 150 cm/h. Fractogel DEAE had a dynamic plasmid begins to appear in the column breakthrough as detected by
capacity of 2.4 mg/ml under the same conditions except for size-exclusion HPLC (SEC-HPLC).
332 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335

Table 4
Effect of NaCl concentration in the load on RNA removal and plasmid recovery when loading anion-exchange columns to capacity
Anion-exchange column NaCl concentration SEC-HPLC IE-HPLC
in the load (M)
RNA content in Plasmid content in Plasmid
breakthrough (%) eluate (%) recovery (%)
Q Ceramic HyperD F 0.63 100.0 100.0 72.0
Fractogel DEAE 0.69 100.0 100.0 97.6
0.61 100.0 100.0 100.0
0.59 100.0 99.2 100.0

POROS 50HQ 0.73 99.7 100.0 95.0

0.72 100.0 100.0 96.0
0.71 100.0 99.7 95.0
Each 1 ml anion-exchange column was loaded to capacity in high NaCl in 50 mM sodium phosphate buffer at pH 7.0. The breakthrough fraction was
recovered and analysed by size-exclusion HPLC (SEC-HPLC) to determine the RNA content. The columns were then washed in high NaCl containing
buffer and plasmid eluted with a linear gradient of NaCl to 1 M in 50 mM sodium phosphate at pH 7.0. The eluate was analysed by size-exclusion HPLC
to determine the plasmid content and by anion-exchange HPLC (IE-HPLC) to measure plasmid recovery.

a loading salt concentration of 0.69 M. POROS 50HQ had anion-exchangers gave complete separation of RNA in the
the lowest dynamic capacity at 2.1 mg/ml under the same breakthrough from plasmid in the eluate. Fig. 3a shows the
conditions except that the salt concentration in the load was size-exclusion HPLC analysis of the loading material. The
0.72 M. first elution peak between 25 and 30 min is plasmid with 72%
of the total peak area while RNA elutes as a second peak
3.5. Anion-exchange chromatography: loading to capacity between 40 and 45 min and represents 28% of the total peak
area. Fig. 3b shows the eluate material from anion-exchange
The robustness of the three anion-exchangers selected for chromatography on Fractogel DEAE: only the plasmid peak
their high capacity was tested by determining their load- is present. Pure plasmid free of contaminant RNA was recov-
ing range of operation at capacity. The highest loading salt ered at high yield (>95% for Fractogel DEAE and POROS
concentration that resulted in complete recovery of plasmid 50HQ) except for Q Ceramic HyperD F where plasmid re-
was determined. Similarly, the lowest loading salt concen- covery was low (72.0%). Subsequent sanitisation cycles on
tration that prevented binding of RNA to the column was Q Ceramic HyperD F revealed that some of the plasmid was
also determined. Table 4 shows that Fractogel DEAE had strongly bound to the column and difficult to elute with a
a wider range of operation (0.61–0.69 M NaCl in the load) salt gradient.
compared with POROS 50HQ (0.72–0.73 M NaCl) and Q Three repeat runs on Fractogel DEAE loading plasmid
Ceramic HyperD F. In fact, with Q Ceramic HyperD F con- at 3 mg/ml in 50 mM phosphate, 0.63 M NaCl, pH7.0 at
sistency of column performance was a problem and loading 150 cm/h showed good reproducibility in plasmid recovery
conditions could not be reproducibly established. Loading and RNA clearance (Table 5).
the Fractogel DEAE column at the lower end of the operat- Three repeats runs on POROS 50HQ loading plasmid at
ing range (0.63 M NaCl) also resulted in improved dynamic 1.9 mg/ml in 50 mM phosphate, 0.72 M NaCl, pH 7.0 at
capacity (3 mg/ml). Within their range of operation, all three 150 cm/h also showed good reproducibility (Table 5).

Table 5
RNA clearance and plasmid recovery after three repeat runs on anion-exchange chromatography columns under optimised conditions
Anion-exchange column NaCl concentration IE-HPLC SEC-HPLC
in the load (M)
Plasmid recovery (%) RNA content in breakthrough (%) Plasmid content in eluate (%)

Fractogel DEAE 0.63 94.0 98.9 100.0

94.0 99.4 100.0
94.0 99.3 100.0
POROS 50HQ 0.72 95.0 99.6 100.0
95.0 99.3 100.0
90.0 100.0 100.0
Each 1 ml anion-exchange column was loaded to capacity in high NaCl in 50 mM sodium phosphate buffer at pH 7.0. The breakthrough fraction was
recovered and analysed by size-exclusion HPLC (SEC-HPLC) to determine the RNA content. The columns were then washed in high NaCl containing
buffer and plasmid eluted with a linear gradient of NaCl to 1 M in 50 mM sodium phosphate at pH 7.0. The eluate was analysed by size-exclusion HPLC
to determine the plasmid content and by anion-exchange HPLC (IE-HPLC) to measure plasmid recovery. This process was repeated a further two times
with sanitisation in 0.5 M NaOH between each run.
 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335 333

0.14 decreased to 28:72 wt.%. RNA removal steps prior to chro-

0.12 matography were necessary not only to reduce the RNA bur-
den on the column but also to allow better control of loading
0.10
buffer conditions which is not possible when loading lysate
Abs 260nm (AU)

0.08 directly.
0.06 Plasmid Careful selection of the anion-exchange chromatogra-
0.04 phy adsorbent is essential. But before starting a screening
0.02
RNA experiment, it was important to have the right conditions
for optimal resolution of plasmid from RNA. Since it
0.00
would have been too time-consuming to scout each adsor-
-0.02 bent for the effect of pH, only one weak and one strong
0 10 20 30 40 50 60 70
anion-exchanger were investigated (Fractogel DEAE and
(a) Time (min)
TMAE, respectively). As expected, pH had no significant
effect on plasmid retention on the strong anion-exchanger
0.6 but was an important factor in the selectivity of the weak
0.5 anion-exchanger. Only at pH 7.5 or less was baseline reso-
lution of RNA from plasmid achieved. As pH was increased
0.4
Abs 260nm (AU)

Plasmid
above 7.5, not only was plasmid retention affected but res-
0.3 olution was also lost. As a result, phosphate buffer at pH
0.2
7.0 was used to conduct all subsequent experiments.
A wide range of anion-exchangers (18 in total) was tested
0.1 under the same conditions of elution with a linear salt gra-
0.0 dient. Only four gave baseline separation of RNA from
plasmid: Q Sepharose FF, Q Ceramic HyperD F, Fractogel
-0.1
0 10 20 30 40 50 60 70
DEAE and POROS 50HQ. We found no obvious explanation
for the high selectivity of these adsorbents under the same
(b) Time (min)
operating conditions: the heterogeneity of functional groups
Fig. 3. Size-exclusion HPLC analysis of loading material (a) and eluate (DEA and Q) and matrix (agarose, ceramic, methacrylate
(b) from anion-exchange chromatography on Fractogel EMD DEAE under and polystyrene divinylbenzene) cannot account for their
optimised conditions.
performance.
The screening experiments show that both plasmid and
4. Discussion RNA bind to the chromatography support but that RNA has
a lower affinity. Both plasmid and RNA are polyanionic
Anion-exchange chromatography as a tool for the purifi- molecules and interact with the positively charged functional
cation of pharmaceutical-grade plasmid DNA suffers from groups on the chromatography support through the nega-
several drawbacks. Loading crude lysate containing large tively charged phosphate residues on the nucleic acid back-
amounts of impurities such as chromosomal DNA, RNA, bone. With nucleic acids, overall charge is a function of size
proteins and endotoxins directly on an anion-exchanger is and RNA molecules are therefore expected to be displaced
not recommended and primary purification involving pre- ahead of the plasmid. However, high molecular weight RNA
cipitation or filtration steps is essential. Resolution from molecules have been found to co-elute with plasmid [24,26].
impurities with a similar chemical composition and struc- It is thought that the supercoiled conformation of plasmids
ture (chromosomal DNA, RNA) or charge (endotoxin) is reduces the overall charge density and may be responsible
limited. Low capacity for plasmid of most commercial for early elution. RNA reduction steps before the chromatog-
chromatography media is also a concern for large-scale raphy step are therefore essential (Fig. 4).
manufacture. Bovine-derived RNase is commonly added to Anion-exchange chromatography may be considered
degrade RNA but is not recommended by regulatory agen- for large-scale processing only if capacity for plasmid
cies. The absence of added RNase in the purification process is sufficient. The typical anion-exchanger capacity of
results in very high RNA levels: about 25 times the amount 40–200 ␮g/ml quoted by Prazeres et al. [28] is well below
of plasmid DNA by weight in the clarified lysate. We have that for globular proteins (up to >100 mg/ml) and is a chal-
developed a plasmid purification protocol incorporating an lenge for scale-up. We have found that dynamic capacity
anion-exchange polishing step that addresses these issues. for plasmid can be increased significantly by using a suit-
First, RNA levels were reduced by precipitation with cal- able salt concentration in the load that prevents binding
cium chloride to remove high molecular weight molecules of RNA to the adsorbent, therefore maximising plasmid
[31] followed by tangential flow filtration to clear low molec- capture. These loading conditions were optimised for each
ular weight RNA [32]. As a result the ratio of RNA to plas- of the four selected anion-exchangers and breakthrough
mid DNA in the load of the chromatography column was experiments conducted to determine capacity. Q Sepharose
334 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335

from the column. Significant peak tailing during gradient

elution and removal of material absorbing at 254 nm during
sanitisation with sodium hydroxide suggests diffusion lim-
itations. In terms of robustness, Fractogel DEAE showed a
wider range of operation: selectivity at high capacity was
achieved within a wider range of loading salt concentration
than with POROS 50HQ or Q Ceramic HyperD F. Plasmid
recovery and reproducibility over three repeat runs was also
excellent.
In conclusion, one anion-exchanger (Fractogel DEAE)
used under optimal conditions met the criteria of pu-
rity, high capacity, high recovery, robustness and repro-
ducibility required for the large-scale manufacture of
pharmaceutical-grade plasmid DNA. Anion-exchange chro-
matography was the polishing step in a downstream process
with no added RNase. Primary purification was achieved
by precipitation and TFF to reduce RNA levels before
Fig. 4. Analysis of RNA removal by electrophoresis on a 0.8% agarose chromatography.
gel: molecular weight markers (lanes 1 and 6); clarified lysate (lane 2);
lysate post calcium chloride precipitation (lane 3); TFF retentate (lane
4), and Fractogel DEAE eluate (lane 5). Plasmid isoforms include open
circular (OC), supercoiled (SC), and linear (L). Acknowledgements

We would like to thank Cora Henwood for her comments

FF had a low capacity (<1 mg/ml) compared with POROS and Hella Bosteels for the fermentation.
50HQ (2.1 mg/ml), Fractogel DEAE (3 mg/ml) and espe-
cially Q Ceramic HyperD F (>5.3 mg/ml). Most commer-
cial anion-exchangers were designed for the purification References
of globular proteins and have a porous bead structure with
[1] A. Mountain, Trends Biotechnol. 18 (2000) 119.
pore sizes of up to 2000 Å. Supercoiled plasmids are several
[2] H.L. Robinson, Int. J. Mol. Med. 4 (1999) 549.
thousands of angstroms in length and are therefore essen- [3] M. Schleef, Biotechnology, vol. 5a, Wiley, Weinheim, 1999, p. 443.
tially excluded from the pores [22–28]. Binding is achieved [4] US Food and Drug Administration, Points to consider on plasmid
mostly through interaction with functional groups on the DNA vaccines for preventive infectious disease indications, US Food
surface of the beads resulting in low capacity. This is the and Drug Administration, Rockville, MD, USA, 1996.
[5] US Food and Drug Administration, Guidance for industry: guidance
case for the Q Sepharose FF adsorbent with a pore size of
for human somatic cell therapy and gene therapy, US Food and Drug
about 1900 Å. POROS 50HQ exhibits large through-pores Administration, Rockville, MD, USA, 1998.
(6000–8000 Å) which may allow some penetration of plas- [6] N.A. Horn, J.A. Meek, G. Budahazi, M. Marquet, Hum. Gene Ther.
mids [25]. Fractogel DEAE has a “tentacular” structure 6 (1995) 565.
where functional groups are located at the end of long arms [7] J.K. McClung, R.A. Gonzales, Anal. Biochem. 177 (1989) 378.
[8] G.J. Raymond, P.K. Bryant, A. Nelson, J.D. Johnson, Anal. Biochem.
grafted to the bead surface that circumvent the steric hin-
173 (1988) 125.
drance caused by large plasmid molecules [25]. Q Ceramic [9] R.B. Sparks, J.H. Elder, Anal. Biochem. 135 (1983) 345.
HyperD F adsorbent is made of a polymeric cross-linked hy- [10] A.P. Green, G.M. Prior, N.M. Helveston, B.E. Taittinger, X. Lu, J.A.
drogel homogeneously distributed within a porous ceramic Thompson, BioPharm 10 (1997) 52.
material. It is unlikely that plasmid molecules penetrate the [11] M.M. Diogo, J.A. Queiroz, G.A. Monteiro, D.M.F. Prazeres, Anal.
Biochem. 275 (1999) 122.
hydrogel contained within the ceramic shell. One explana- [12] M.M. Diogo, J.A. Queiroz, G.A. Monteiro, S.A.M. Martins, G.N.M.
tion for high capacity is that plasmids not only bind to the Ferreira, D.M.F. Prazeres, Biotechnol. Bioeng. 68 (2000) 576.
hydrogel coating the surface of the particle but also to the [13] M.M. Diogo, S. Ribeiro, J.A. Queiroz, G.A. Monteiro, P. Perrin, N.
hydrogel filling the pores although surface ruggedness has Tordo, D.M.F. Prazeres, Biotechnol. Lett. 22 (2000) 1397.
also been mentioned as a possible source of high capacity [14] A. Colman, M.J. Byers, S.B. Primrose, A. Lyons, Eur. J. Biochem.
91 (1978) 303.
[33]. [15] T.R. Johnson, J. Ilan, Anal. Biochem. 132 (1983) 20.
When the three anion-exchangers selected for their high [16] R. Giovannini, R. Freitag, Biotechnol. Bioeng. 77 (2002) 445.
capacity and selectivity were loaded to capacity under opti- [17] M.J. Carter, I.D. Milton, Nucleic Acids Res. 21 (1993) 1044.
mal conditions of high loading salt concentration, plasmid [18] R. Boom, C.J.A. Sol, M.M.M. Salimans, C.L. Jansen, P.M.E.
free of contaminating RNA (as determined by size-exclusion Wertheim-Van Dillen, J. Van der Noordaa, J. Clin. Microbiol. 28
(1990) 495.
HPLC) was eluted from the column while RNA was cleared [19] R. Lakshmi, V. Baskar, U. Ranga, Anal. Biochem. 271 (1999) 109.
in the column breakthrough. Plasmid recovery was a prob- [20] P. Wils, V. Escriou, A. Warnery, F. Lacroix, D. Lagneaux, M. Ollivier,
lem with Q Ceramic HyperD F: only 72% could be eluted J. Crouzet, J.-F. Mayaux, D. Scherman, Gene Ther. 4 (1997) 323.
 A. Eon-Duval, G. Burke / J. Chromatogr. B 804 (2004) 327–335 335

[21] T. Schluep, C.L. Cooney, Nucleic Acids Res. 26 (1998) 4524. [28] D.M.F. Prazeres, G.N.M. Ferreira, G.A. Monteiro, C.L. Cooney,
[22] D.M.F. Prazeres, T. Schluep, C. Cooney, J. Chromatogr. A 806 (1998) J.M.S. Cabral, Trends Biotechnol. 17 (1999) 169.
31. [29] G.D. Cooke, R.M. Cranenburgh, J.A.J. Hanak, P. Dunnill, D.R.
[23] R.N. Hines, K.C. O’Connor, G. Vella, W. Warren, BioTechniques 12 Thatcher, J.M. Ward, J. Biotechnol. 85 (2001) 297.
(1992) 430. [30] H.C. Birnboim, J. Doly, Nucleic Acids Res. 7 (1979) 1513.
[24] G. Chandra, P. Patel, T.A. Kost, J.G. Gray, Anal. Biochem. 203 [31] A. Eon-Duval, K. Gumbs, C. Ellett, Biotechnol. Bioeng. 83 (2003)
(1992) 169. 544.
[25] I. Theodossiou, M. Sondergaard, O.R.T. Thomas, Bioseparation 10 [32] A. Eon-Duval, R.H. MacDuff, C.A. Fisher, M.J. Harris, C. Brook,
(2001) 31. Anal. Biochem. 316 (2003) 66.
[26] G.N.M. Ferreira, J.M.S. Cabral, D.M.F. Prazeres, Bioseparation 9 [33] J. Coffman, E. Boschetti, Bioseparation and Bioprocessing, vol. 1,
(2000) 1. Wiley, Weinheim, 1998, p. 157.
[27] E. Thwaites, S.C. Burton, A. Lyddiatt, J. Chromatogr. A 943 (2002)
77.