Rapid Process Development for High Yield Plasmid DNA Fed-batch Fermentation

November 1, 2009

BioPharm International

Volume 22, Issue 11

How to reduce plasmid-mediated metabolic burden for higher yields.


To commercialize DNA medicines, industrial plasmid DNA manufacturing processes that meet the quality, economy, and scale requirements projected for future products are needed. We have developed cell bank and fermentation process unit operation innovations that reduce plasmid-mediated metabolic burden, enabling improved upstream production of optimal plasmids to 2.6 g/L. Application of these processes also facilitated production of otherwise unstable direct repeat containing vectors in standard E. coli host strain DH5α, eliminating the need for specialized stabilizing strains. Fermentation yields with low yield plasmids also were improved up to three-fold by using a simple fermentation process development method requiring only one to two fermentations.

Plasmid DNA is increasingly finding its way into new, experimental non-viral gene therapeutics, including DNA vaccines,1 short hairpin RNA (shRNA) gene knockdown therapeutics, gene replacement vectors, and seed constructs for viral vector production.

Nature Technology Corporation

Host Escherichia coli strains for plasmid production require recA and endA mutations.2 Several studies on various plasmid host strains indicate that plasmid yield and quality are significantly affected by the choice of host strain. However, these studies also demonstrate that plasmid production in shake flasks is poorly predictive of plasmid production in fermentation, and that a strain's performance can be affected by the fermentation process.3 DH5α is a good host strain because it consistently produces high-quality plasmid DNA in both shake flask and fermentation culture. Evaluation of the individual contributions of the DH5α host strain, plasmid backbone, and production process to plasmid production has demonstrated high-yield plasmid fermentation is largely determined by process and vector-intrinsic factors, not strain-intrinsic factors.4


Constitutive, high plasmid copy number throughout the fermentation process is not necessary or desirable. Maintaining a high copy number during biomass accumulation creates an environment in which plasmid-free cells have a significant growth advantage.

Plasmid-mediated metabolic burden can inhibit biomass growth and may lead to stability or quality problems (e.g., deletions or dimers) with many plasmids. Thus, maintaining low cell stress or metabolic burden during biomass accumulation, and inducing high plasmid copy numbers for only the final portion of the process, leads to superior volumetric plasmid yields while preserving plasmid quality.

Gene therapy or DNA vaccine plasmids typically contain the pUC (temperature-sensitive) origin of replication. This temperature sensitivity is especially useful for inducing high-yield plasmid production in fermentation using a temperature-inducible, fed-batch process (Figure 1) with an optimized semi-defined medium.5 The initial setpoint temperature of 30 °C maintains the plasmid at a low copy number. This reduces metabolic burden and cell stress during bacterial growth such that the majority of the final biomass is formed under low stress conditions.

Figure 1. Illustration of the temperature-inducible fed-batch process. The initial temperature setpoint is 30 °C to keep the plasmid copy number low during growth. Feed medium containing concentrated glycerol is added according to an exponential feeding strategy to control cellular growth at about 0.12/h.

After sufficient biomass accumulation, the temperature is shifted to 42 °C and growth continues for up to approximately one doubling of cell mass. This process has resulted in volumetric plasmid yields up to 2,590 mg/L, and specific plasmid DNA yields up to 51 mg/g dry cell weight (DCW), or 5% of the total DCW.4 Figure 2 shows a time profile from a fermentation that reached a plasmid yield of 2,100 mg/L.

Figure 2. Time profile of a fermentation of DH5α containing a 6.5 kb DNA vaccine plasmid, temperature induced at 29 h


Critically, use of the low metabolic burden process was essential to produce otherwise unstable retroviral plasmids. For example, pVLTrap, a 6.7 kilobase (kb) retrotransposon plasmid vector containing two long terminal repeats (LTRs, known to cause stability problems), was produced successfully in DH5α at 785 mg/L using the aforementioned low metabolic burden fermentation process (Table 1). However, when grown under identical conditions but at 37 °C, the region between the LTRs was entirely deleted by the end of the fermentation (Figure 3), and the truncated plasmid yield only reached 214 mg/L (Table 1).

Figure 3. Agarose gel electrophoresis of samples from DH5α/pVLTrap fermentations from Table 1. Lane 1 is plasmid prepared directly from the seed stock, and lane 2 is plasmid from the 30–42 °C fermentation harvest (70 OD600). Lanes 3–8 are from the 37 °C throughout process, at 16 (14 OD600), 20, 24, 37, 38, and 40 (113 OD600 harvest sample) h post induction, respectively. Note the deletion product (arrow) is detected in all samples from the 37 °C fermentation.


For production of the various DNA medicines currently under investigation, it is essential that the fermentation process functions with a variety of different vector backbones and gene inserts. However, many plasmid vectors are unpredictably toxic or otherwise low yielding in standard fermentation processes. Numerous inserts also can confer low-yield plasmid production on an otherwise high-yielding plasmid backbone.

Importantly, with toxic plasmids, maintaining a low metabolic burden should begin as early as the transformation process. Propagation of cultures at 30 °C rather than 37 °C during seed stock manufacture can dramatically improve downstream fermentation yields when producing plasmids containing toxic inserts,4 presumably by reducing the copy number to limit plasmid-mediated metabolic burden during this critical step.

Use of the low metabolic burden process with 30 °C manufactured seed stocks further improved yield. A two-fold yield improvement was observed with the low metabolic burden at 30–42 °C inducible fermentation process (compared to fermentation at 37 °C throughout) with low-yielding short-hairpin RNA (shRNA) plasmids containing multiple short palindromic repeats (Table 1).

Table 1. Comparison of plasmid quality and yield from 37 °C throughout and 30–42 °C inducible fermentations

Modest yield improvement was observed with the 30–42°C inducible fermentation process compared to 37 °C throughout with a vector containing a 43 basepair (bp) palindromic repeat. Specific yields of this vector were reduced three-fold compared to a control vector without the repeat (Table 1). Reduced yields are caused by an inherent difficulty of DNA polymerases in replicating self-associating hairpin structures.

Palindromes may be deleted by intramolecular or intermolecular recombination events. The observed sequence stability of the palindromes (in both processes; Table 1) establishes that shRNA plasmids are inherently stable in DH5α fermentation cells.

Collectively, these results demonstrate plasmid production processes that use 37 °C continuously do not perform well with many unstable or suboptimal plasmids because of detrimental metabolic burden from plasmid maintenance at constitutively high levels. Using the low metabolic burden cell bank and fermentation process facilitated production of unstable plasmids such as inverted or direct repeat containing vectors in standard strains such as DH5α, eliminating the need to use specialized stabilizing host strains.


Some plasmids are low yielding in the 30–42 °C inducible fermentation process because of cell lysis after plasmid induction. We hypothesize that these plasmids are toxic to the production cell at moderate copy number, such that lysis occurs at a threshold copy number reached shortly after plasmid induction. Consistent with this, fermentation yields (and cell bank viability) from low yielding plasmid-containing cell lines were much higher when seed stocks were created at 30 °C, rather than at 37 °C.4

We have determined that fermentation yield with such plasmids can be improved through simple process modifications to: 1) Increase biomass before induction by performing temperature induction at a higher OD600; or 2) Induction for a shorter duration to limit cell lysis; or 3) Induction at a lower temperature (37 °C) to reduce copy number amplification.


The following case studies using seed stocks manufactured at 30 °C in strain DH5α illustrate rapid process improvement using combinations of these modifications. In all cases, culture lysis and poor plasmid yields after the first pilot fermentation using standard induction conditions were indicative that the plasmid was toxic to the host at moverate copy number. Volumetric yield (mg/L plasmid) improvements ranged from 60–170% (Table 2: one process development run) to 194% (Table 3: three-fold yield increase with two process development runs).

Table 2. Improved plasmid yield after fermentation process modification

Case Study 1

In this optimization with a proprietary 9-kb plasmid, fermentation run 1 induction was at 42 °C at 50 OD600 for 10.5 h. After the observed extensive lysis at harvest, the second fermentation was induced at 55 OD600 for 10 h (shorter induction at higher OD600). This resulted in improved harvest biomass and specific yield, resulting in a 60% improvement in volumetric yield (Table 2).

Case Study 2

In this optimization with a proprietary 12-kb plasmid, induction at 42 °C was at 50 OD600 for 10.5 h in fermentation run 1. After the observed extensive lysis and very poor yield at harvest, the second fermentation was induced at 37 °C at 60 OD600 for 9 h (shorter, reduced temperature induction at higher OD600). This resulted in improved harvest biomass and specific yield, resulting in a 170% improvement in volumetric yield (Table 2).

Case Study 3

In this optimization with a proprietary 3.5-kb eukaryotic GFP expression plasmid, fermentation run 1 induction was at 42 °C at 50 OD600 for 9 h. After the observed extensive lysis and poor yield at harvest, the second fermentation was induced at 55 OD600 for 10 h (induction at higher OD600). This resulted in improved harvest biomass and specific yield, resulting in a 72% improvement in volumetric yield (Table 2).

Case Study 4

This was a proprietary 6-kb DNA vaccine plasmid containing a toxic influenza serotype H1 hemagglutinin gene. In the first fermentation (run 1), culture lysis was visible after 4 h induction at 42 °C (starting OD600 was 60). Only modest plasmid specific yield improvement (1.2 to 3.2 mg/L/OD600) was observed by 30–42 °C induction resulting in a final volumetric yield of 260 mg/L (Table 3).

Table 3. Improved plasmid yield after fermentation process modification

For the first process development run (run 2), a strategy to more modestly induce copy number (using 37 °C rather than 42 °C induction) was attempted. The 37 °C induction was performed at 53 OD600 for 6 h. However, specific yield during the 37 °C stage increased from 1.0 to only 2.3 mg/L/OD600. Therefore the 37 °C stage was followed by a shift to 42 °C at 83 OD600 for 4 h. Specific yields further increased to 4.6 mg/L/OD600 (44% improvement compared to run 1) after the 42 °C stage. This improved overall volumetric yield compared to run 1 by 68% to 436 mg/L (Table 3).

For the second process development run (run 3) the fermentation was induced for 9 h at 42 °C (starting OD600 was 50). Despite culture lysis during the last several hours of induction resulting in overall lower biomass yields, three-fold improved overall plasmid yields were obtained (Table 3: 194% improvement in volumetric yield, and 253% improvement in specific yield compared to run 1). This demonstrates that, although a small subpopulation of cells is sensitive to plasmid accumulation and lysis early, the bulk of the cells are high producers with this plasmid. Optimization of the process to favor accumulation of the high producing cells led to dramatic overall process yield improvement.

Thus, consistent with the hypothesis that these plasmids are toxic to the production cell at moderate copy number, simple process modifications that increased biomass and delayed high copy plasmid induction dramatically improved yield.

Expression from cryptic promoters in plasmid sequences could drive copy number-dependent expression of toxic peptides (from the sense or antisense strand) in E. coli. Alternatively, RNA sequences or structures may be involved. Surprisingly, two classes of toxic plasmids were identified that are either 1) universally toxic or 2) conditionally toxic to a subset of cells (e.g., influenza serotype H1 hemagglutinin plasmid, case study 4). Interestingly, an influenza serotype H3 hemagglutinin plasmid that was toxic at moderate copy number in freshly transformed cells, but not fermentation cells, has also been identified.4 Although universal toxicity may be attributed to the production of cryptic proteins, peptides, or RNA structures that disrupt core metabolic functions or membrane integrity, the cellular targets for conditional plasmid toxicity are unknown.


The temperature-inducible fermentation process has been successfully scaled-up to 300 L and used for GMP production of DNA vaccine plasmids. Fermentation cells are compatible with a variety of vectors,6–8 lysis,9 and downstream purification strategies.10–11 This low metabolic burden process is ideal as a generic plasmid DNA production platform to produce previously unstable and toxic plasmid DNA, as well as optimized plasmids at high yields in standard host strain DH5α, thus eliminating the need to use specialized stabilizing host strains. Contract manufacturers can use the base process with simple defined process optimization steps if necessary, to produce a wide range of customer-developed plasmids.


We thank Sheryl Anderson, Sarah Langtry, Justin Vincent, and Angela Schukar for their tireless efforts cleaning, batching, and operating fermenters. This paper described work supported by NIH grant R44GM072141.

James A. Williams is the vice president of research and development, Clague P. Hodgson is the president, and Aaron E. Carnes is the director of process development, all at Nature Technology Corporation, Lincoln, NE, 402.472.6530, jim@natx.com


1. Kutzler MA, Weiner DB. DNA vaccines: ready for prime time? Nat Rev Genet. 2008;9:776–88.

2. Carnes AE. Fermentation design for the manufacture of plasmid DNA. BioProcess Int. 2005;3(9):36–42.

3. Williams JA, Carnes A, Hodgson CP. Plasmid DNA vaccine vector design: impact on efficacy, safety and upstream production. Biotechnol Adv. 2009;27:353–70.

4. Williams JA, Luke J, Langtry S, Anderson S, Hodgson CP, Carnes AE. Generic plasmid DNA production platform incorporating low metabolic burden seed-stock and fed-batch fermentation processes. Biotechnol Bioeng. 2009;103:1129–43.

5. Carnes AE, Hodgson CP, Williams JA. Inducible Escherichia coli fermentation for increased plasmid DNA production. Biotechnol Appl Biochem. 2006;45:155–66.

6. Williams JA, Luke J, Johnson L, Hodgson C. pDNAVACCultra vector family: high throughput intracellular targeting DNA vaccine plasmids. Vaccine. 2006;24:4671–6.

7. Luke J, Carnes AE, Hodgson CP, Williams JA. Improved antibiotic-free DNA vaccine vectors utilizing a novel RNA based plasmid selection system. Vaccine. 2009. Epub.

8. Williams JA. Vectors and methods for genetic immunization. World Patent Application WO2008153733; 2008.

9. Carnes AE, Hodgson CP, Luke J, Vincent J, Williams JA. Plasmid DNA production combining antibiotic-free selection, inducible high yield fermentation, and novel autolytic purification. Biotechnol Bioeng. 2009;104:505–15.

10. Carnes AE, Williams JA. Plasmid DNA manufacturing technology. Recent Patents Biotechnol. 2007;1:151–66.

11. Hoare M, Levy MS, Bracewell DG, Doig SD, Kong S, Titchener-Hooker N, Ward JM, Dunnill P. Bioprocess engineering issues that would be faced in producing a DNA vaccine at up to 100 m3 fermentation scale for an influenza pandemic. Biotechnol Prog. 2005;21:1577–92.