Antibiotic-free selection in E. coli: new considerations for optimal design and improved production
© Peubez et al; licensee BioMed Central Ltd. 2010
Received: 23 February 2010
Accepted: 7 September 2010
Published: 7 September 2010
The increasing regulatory requirements to which biological agents are subjected will have a great impact in the field of industrial protein expression and production. There is an expectation that in a near future, there may be "zero tolerance" towards antibiotic-based selection and production systems. Besides the antibiotic itself, the antibiotic resistance gene is an important consideration. The complete absence of antibiotic-resistance gene being the only way to ensure that there is no propagation in the environment or transfer of resistance to pathogenic strains.
In a first step, we have designed a series of vectors, containing a stabilization element allowing a complete elimination of antibiotics during fermentation. Vectors were further improved in order to include alternative selection means such as the well known poison/antidote stabilization system. Eventually we propose an elegant positive pressure of selection ensuring the elimination of the antibiotic-resistance gene through homologous recombination. In addition, we have shown that the presence of an antibiotic resistance gene can indirectly reduce the amount of expressed protein, since even in absence of selection pressure the gene would be transcribed and account for an additional stress for the host during the fermentation process.
We propose a general strategy combining plasmid stabilization and antibiotic-free selection. The proposed host/vector system, completely devoid of antibiotic resistance gene at the end of construction, has the additional advantage of improving recombinant protein expression and/or plasmid recovery.
Antibiotics are commonly used during bacterial fermentation, and the vast majority of expression vectors contain antibiotic resistance genes as selection markers. The degree of acceptance of these markers is both antibiotic and context dependent: Kanamycin and, to a lesser extent, Tetracycline, are still acceptable to the health authorities. In contrast the use of β-lactams, often associated with allergic responses, is strictly prohibited. Requirement can also be variable according to the nature of the therapeutic product, the presence of an antibiotic resistance gene, tolerated on a vector expressing a recombinant biopharmaceutical, will be totally undesirable on a gene therapy plasmid.
Different strategies are currently used and are described below.
Post-segregational killing is a mechanism by which plasmids are stably maintained by expressing a gene product that would be toxic to cells becoming plasmid-free upon division.
The most representative example of post-segregational killing of plasmid-free cells is the Hok/Sok system described for the first time in 1986. The translation of the Hok (host killing) messenger, encoding a toxin lethal to the bacteria, is completely blocked by the anti-messenger Sok (suppression of killing). In the absence of plasmid, Sok, which is less stable than Hok, is lost first, allowing the translation of the Hok mRNA and expression of the toxin lethal to the cell .
Other examples of the mechanism known as post-segregational killing have been reported such as plasmids carrying type II restriction modification (RM) gene complexes inhibiting the propagation of a cell population following chromosome breakage .
Essential gene complementation
• The example of dapD
The most common way to achieve selection in the absence of antibiotics is via complementation of an essential gene making use of an expression vector in a strain with a defect or inhibited expression of the same essential gene. The dapD gene, which has a role in the Lysine biosynthetic pathway as well as cell wall assembly, has been selected as a preferred candidate by several authors, knowing that mutations in the dap pathway are lethal . The limitations, in that case, are the intrinsic difficulty in construction of a dapD - mutant strain and dependence towards defined culture media composition.
• dapD and repressor titration
A more elaborated strategy still based on dapD was described: the so-called "operator repressor titration for antibiotic-free plasmid maintenance", proposed by Cobra Therapeutics, is a model in which the plasmid loss induces the downregulation of the essential dapD gene, and thus bacterial death .
• Translation initiation factor infA
As alternatives to dapD, other essential genes have been targeted for the same purpose, particularly infA coding for a translation initiation factor . The authors show that the system is tightly regulated and that no cross-feeding effect is observed since initiation factors released into the media from lyzed cells are not absorbed by plasmid-free cells.
• Amino-acid auxotrophy complementation
A proline-auxotrophic K12 strain obtained through chromosomal pro BA gene deletion is described  for the expression of antibody Fab fragment. In that particular case the plasmid-mediated complementation is used as a second selection mechanism to completely abolish plasmid loss during fermentation.
According to the same principle an E. coli M15-derivated glycine-auxotrophic strain has been constructed and shown to produce comparable amounts of recombinant protein as a conventional system .
• The murA gene and RNA/RNA interaction
The principle of down regulating an essential gene upon plasmid loss has been exploited in a very original way for the design of new vectors for gene therapy [8, 9]. The expression of the essential gene murA encoding an enzyme essential for the biosynthesis of cell wall is under control of the Tet repressor, TetR expression is inhibited by an RNA-RNA antisense interaction with RNAI derived from plasmid origin of replication ColE1. The major advantage of this system is that no additional sequence is required on the plasmid.
• RNA-OUT based antibiotic-free selection system
In a recent paper , a RNA based selectable marker, not restricted to ColE1 containing vectors is described. Briefly, a counter-selectable marker (sacB) levansucrase from Bacillus subtilis, under control of the RNA-IN promoter is integrated into the bacterial chromosome induces cell death in presence of sucrose.
Plasmid maintenance is ensured by the presence of the plasmid-borne regulator RNA-OUT anti-messenger acting as a down regulator of the expression of levansucrase.
Plasmid selection using an endogenous essential gene marker
A potential drawback of essential gene complementation is the requirement of engineering mutant host strains and using specifically defined media. The fabl-triclosan model system recently proposed is based on the over-expression of a host essential gene in presence of a chemical inhibitor of its product .
• Mutual dependence: the pCOR system
Other systems such as pCOR, based on the complementation of an amber mutation, and conditional origin of replication have also been established. The original feature of the model is that an additional degree of refinement was introduced, since the dependence created between the host and the vector has become bilateral. Nevertheless, the requirement for a minimal medium for culture means these systems are more likely to be used for DNA production rather than recombinant protein over expression .
Gene ccdB (the poison), inserted into the bacterial genome of the E. coli strain BL21λ (DE3) and encoding a stable protein (100 aa), binding gyrase, an essential element for cell division. Upon binding gyrase, the ccdB gene product impairs DNA replication and induces cell death.
Gene ccdA (the antidote), plasmid-born, encodes an instable protein (90 aa) under control of the mob promoter, acting as a natural inhibitor of ccdB.
It has been shown that after 20 generations on a non selective medium 100% of the bacteria still contain the plasmid. Two hours after induction, the plasmid is still present into all bacteria, which is not the case with a standard pET/BL21λDE3 system.
The various complementation-based expression systems or more generally antibiotic-free systems have the common drawback of being strain dependant, since genetic knockout or modification of an essential gene is not easily transferable from one strain to another and has to be done independently for each strain. Strain modification by chromosomal mutagenesis is still a tedious approach, notwithstanding the existence of technologies making use of counter-selectable vectors or PCR amplified fragments [14–16].
A second aspect to be considered during the expression of a recombinant biopharmaceutical is the plasmid stabilisation that can be achieved by insertion of a genetic element such as the cer locus, which allows stable inheritance of ColE1 and related plasmids by preventing the runaway accumulation of multimers known as "dimer catastrophe" [17, 18]. Multimer resolution is achieved through action of the XerCD site-specific recombinase at the cer site. Cloning of the cer locus into various expression vectors has been extensively documented and the proof of principle largely established in high-cell density cultures .
Here, we report a new strategy to overcome the need for antibiotics in the production process, including increasing plasmid stability and maintenance during fermentation as well as the use of alternative, "antibiotic-free" selection markers.
Results and Discussion
First step: Plasmid stabilisation
Plasmid stabilisation and increased maintenance during a fermentation process can be considered as a first step towards antibiotic starvation. Genetic elements allowing plasmid maintenance during cell division and limiting the probability of plasmid loss over generations should be considered.
In a first series of experiments the stabilisation effect of the cer locus was demonstrated during the expression of different recombinant antigens from Helicobacter pylori, at laboratory scale in shake flasks.
Two vectors, based on pET28 backbone, expressing the AlpA or Urease antigens from Helicobacter pylori have been modified in order to include a DNA fragment containing the cer locus in both orientations as compared to the expression cassette. From a technical point of view, the construct was done by ligation between two PCR fragments containing respectively the cer fragment and the entire pET28 plasmid devoid of the F1 origin. The assembly was made possible by the complementarity of two newly created unique restriction site brought by the PCR primers used during the amplification steps.
The readout of the stability testing experiment was a simple counting, after plating on selective medium, of bacterial colonies still containing the plasmid at different times of culture in absence of antibiotic selection.
Example of Helicobacter pylori AlpA protein produced in erlen flask in absence of Kanamycin
Plasmid without cer
With cer orientation1
With cer orientation2
2 h(IPTG added)
Example of Helicobacter pylor i Urease produced in erlen flask in absence of kanamycin
Plasmid without cer
With cer orientation1
With cer orientation2
The real value of the 24 hours end points is matter of discussion since one can argue that an expression driven by the T7 promoter is likely to cause a certain level of metabolic breakdown of the cellular system and make questionable a cultivation experiment for more than 25 hours post induction. Even if protein and/or induction condition dependent, this is perfectly true but nevertheless important to document plasmid stability at end and after normal end of fermentation to document regulatory files.
We have demonstrated, or at least re-demonstrated that an appropriate stabilisation of the expression vector is a first step towards the suppression of antibiotic selection during culture.
Second step: evaluation of the poison/antidote system in fermentation conditions
the host/vector couple is commercially available from Delphigenetics,
the engineered strain is a BL21λ(DE3) derivative, making comparison with our standard process easier.
In a pharmaceutical context, the presence of minute amount of ccdB as contaminant would be acceptable, since no equivalent of gyrase is present in human cells. The experiment was designed in order to show whether this antibiotic-free selection would be applicable in fermentation at 1 liter and larger scale.
The model system was based on the pM1816 vector used for the production of recombinant Pseudomonas exotoxin (rEPA), a pM1800 derivative vector containing the kanamicyn resistance gene and the cer fragment. The ccdA encoding gene extracted from the pStaby cloning vector obtained from Delphigenetics was cloned into the pM1816 expression vector generating the pSP1 construct. In order to eliminate the Kanamycin resistance gene a digestion with ClaI and AscI restriction enzymes followed by self-religation has been done.
Evaluation of antibiotic-free system in fermenters (production of rEPA protein)
Induction at early stage
Induction at advanced stage of growth
System based on kana resistance
Antibiotic free system
System based on kana resistance
Antibiotic free system
Cell dry weigh (g/L)
Plasmid retention (%)
Product yield (mg/L)
a standard induction at high cell density (A600 = 25),
an early induction (A600 = 1), not representative of normal fermentation conditions, to increase the number of cell divisions between induction and end of fermentation and thus artificially creating extended high stress conditions. In that case the higher stress would be due to high metabolic charge to ensure multiple cell divisions in parallel to protein production. It is likely to say that in such condition the probability to lose plasmid would be greater than using induction at high cell density. The 3 following parameters have been monitored at the end of fermentation:
percentage of cells still containing a plasmid,
amount of recombinant protein expressed.
Scale-up of antibiotic-free system
Third step: possible contribution of the cer locus
Summary of plasmids used
Derived from plasmid
Recombinant protein expressed
Kanamycin resistance gene
antibiotic-free selection - cer locus gives superior production level (16,5%) compared to standard antibiotic selection - cer (14,5%).
antibiotic-free selection + cer gives also superior production (17,9%) compared to the same construct devoid of cer locus (16,5%).
The following hierarchy can be established:
antibio-free + cer > antibio-free - cer > standard - cer
Even if limited to a single antigen, the experiment shows a benefit, in terms of protein yield, by combining the antibiotic-free selection and the stabilizing element.
Final step: elimination of the antibiotic resistance gene
Antibiotic selection and therefore, the presence of an antibiotic resistance gene on the vector until the last step of the expression are required to ensure the absence of external contaminant cells that would not be counter-selected by the poison/antidote system.
Even if antibiotic selection pressure is not used during the fermentation process, removal of this antibiotic resistance marker is of major importance to prevent horizontal transfer in the environment. This is particularly true for vectors to be used in gene therapy or DNA vaccination protocols. For above mentioned reasons (see second step) it is also important to eliminate this sequence to improve both vector stability and recombinant protein expression level.
Elimination of the antibiotic resistance marker can be very simply obtained by restriction enzyme digestion using unique sites flanking the sequence to be removed followed by self-ligation of the vector. The only issue being that the discrimination between correctly deleted and parental vectors is not directly possible and would involve time consuming analytical techniques such as, restriction mapping following a small scale DNA preparation or alternatively a PCR-based approach.
Recombinant protein production and/or Plasmid DNA preparation
The techniques for midi or maxi plasmid DNA extraction suffer from a certain batch to batch variability (at least at lab scale),
A real head to head comparison is difficult, since the comparison has to be done with non-identical bacterial strains (CYS21 vs DH10B in our case).
Nevertheless, we would say that, in addition to a potential superiority in production, the real added value of our system to plasmid DNA preparation is the one-step selective elimination of the antibiotic resistance gene that is mandatory for therapeutic applications.
Antibiotic-free selection is a general and ultimate goal that can be attained by the implementation of various and combined approaches. An increasing knowledge of bacterial physiology will give access to more and more essential genes or pathways that would be an unlimited source of inspiration for the design of novel selection means. The major driver for the definition of antibiotic-free systems is a desire to anticipate and fulfil future recommendations from health authorities to overcome safety concerns but, interestingly it can give access to unexpected properties such as a marked increase in recombinant protein production or plasmid recovery. The complete elimination of any antibiotic resistance gene is, for different reasons, of critical importance for both recombinant protein production and DNA immunisation vectors. The proposed homologous recombination process has been demonstrated to be fully functional and might be associated to different antibiotic-free expression systems or transferred to other microorganisms.
Vector containing the cer fragment: pM1800 and its derivatives
Table 5 summarizes the properties of mentioned plasmids.
The pET28c vector (Novagen) was amplified by PCR using the « expand long template PCR system » kit from Boehringer Mannheim in order to suppress the f1 origin.
The 5' primer carried the site AscI and the 3' primer, AscI and PacI. After amplification, the PCR product was purified, AscI (New England Biolabs) digested and self-ligated with T4 DNA ligase, (New England Biolabs).
A 300 bp fragment containing the sequence of cer fragment was obtained by PCR amplification using the pXL 2979 plasmid (gift of Rhône Poulenc) as matrix. After purification, the PCR product was digested by AscI and PacI (New England Biolabs), and was cloned into pET28c devoid of f1 origin previously obtained digested by these same enzymes. This new plasmid was termed pM1800.
The same strategy has been used to obtain the same plasmid with a cer fragment in the opposite orientation.
To express the recombinant exoprotein A from Pseudomonas aeruginosa (rEPA), the plasmid pVC45 D (an Ampicillin resistant plasmid provided by Temple University, USA) containing the recombinant exoprotein A (rEPA) gene cloned in fusion with the OmpA signal sequence between HindIII and EcoRI sites has been used. To avoid a PCR step, this plasmid was digested by XbaI (New England Biolabs) and EcoRI (New England Biolabs) to isolate the fragment containing the RBS sequence and the OmpA-rEPA fusion (2024 bp). After purification, the fragment was cloned into plasmid pM1800 digested by the same enzymes to create plasmid pM1816 (7195 bp).
Vector containing the ccdA fragment: pSP301, pSP1, pSP2 and their derivatives
Then, pSP301 vector was digested by PacI (New England Biolabs) and AscI (New England Biolabs) to eliminate the cer fragment. The cohesive ends were digested by the Mung Bean exonuclease (New England Biolabs), and the plasmid was self- ligated to create the new vector pSP2.
Homologous recombination process to obtain a functional ccdA
To use the ccdA/ccdB selection system, the Kanamycin resistance gene was eliminated through homologous recombination. The pSP301 and derived plasmids were digested by ClaI (New England Biolabs). The CYS21 strain (DH10B derived with ccdB in the chromosome) (Delphigenetics) was transformed with the linear plasmid, and plated on LB agar. Only the clones containing a correctly recombined plasmid encoding a functional ccdA gene could grow. The presence of ccdA was checked by enzymatic digestion and sequencing.
Cla I is a unique site in the kanamycin resistance gene, and EcoR V is unique in the vector. When the vector contains the Kanamycin resistance gene two bands are present for the ClaI-EcoRV digestion. If the recombination event behaves as expected and eliminates the kanamycin resistance gene, only one band is likely to be present for the ClaI-EcoRV digestion (data not shown).
Expression at lab scale
To express the recombinant proteins the SE1 bacteria derived from BL21λ(DE3) (with ccdB in the chromosome) were transformed with pSP1, pSP3, pSP4, pSP5 and pSP6, and plated on LB agar. Similarly, the BL21λ(DE3) strain (Invitrogen) was transformed with the PM1816 or pM-Hp3.1 and plated on LB kanamycin (25 μg/ml). An overnight culture for each construction was diluted 1/100 in 50 ml of fresh LB with or without Kanamycin. All cultures were grown at 37°C to an A600 = 0.4-0.6. The cultures were divided in 2 × 25 ml and the expression of the recombinant exoprotein A was induced using 1 mM Isopropyl-β-D-thiogalactopyranoside (IPTG) in one of 25 ml culture. The induced and non-induced bacteria were grown at 37°C up to 25 H post induction.
Evaluation in fermenters
The comparison at bench scale of the new system with the existing system based on the kanamycin resistance involves studies of plasmid retention, product yield and robustness.
The cultures at lab scale were conducted with the Biostat Q BBi Sartorius system which includes 4 individual fermentors of 500 ml as working volume.
The culture at pilot scale was conducted in a Applikon fermentor of 30 liters as working volume. This is the current scale for clinical batches for a phase 1 study.
The medium used is a "in house" propietary complex medium (Yeast extract as source of nitrogen and dextrose as source of carbon) without the use of antibiotics.
Cultures were regularly sampled to monitor the in-process parameters such as A600, Cell Dry Weight (CDW) (g/L), numeration (CFU/ml), product yield and plasmid retention (%)
Each sample of cultures was treated by osmotic shock to release the periplasmic content. Following this, each extract was titrated by ELISA and loaded on SDS-PAGE. A densitometry quantification has been done using GeneTools software to compare the protein of interest versus total proteins.
Plasmid stability monitoring
Samples of cultures were diluted and then, plated on solid media for the numeration
For the system based on the kanamycin resistance, 100 individual colonies were picked on LB agar supplemented with kanamycin (25 μg/mL).
For the antibiotic free system, 96 colonies were picked to inoculate individual liquid cultures in microplates. After 24 h of incubation, the plate was centrifuged and then placed into a robotic pDNA extraction system (QIAgen bioroBot® System). Finally, the samples were loaded on 1% agarose gel to detect the presence or absence of plasmid.
The readout of the stability is provided by a simple numeration of colonies still containing the plasmid as a percentage of the total analyzed colonies.
We are grateful to Dr Cedric Szpirer and Dr Philippe Gabant for helpful discussions and scientific input on the poison antidote system.
Very special thanks to Dr. Catherine Moste, head of the research department, for enthusiastic support to the project and fruitful discussions.
- Gerdes K, Rasmussen PB, Molin S: Unique type of plasmid maintenance function: postsegregational killing of plasmid-free cells. Proc Natl Acad Sci USA. 1986, 83: 3116-20. 10.1073/pnas.83.10.3116.View Article
- Handa N, Kobayashi I: Post-segregational killing by restriction modification gene complexes: observations of individual cell deaths. Biochimie. 1999, 81 (8): 931-938. 10.1016/S0300-9084(99)00201-1.View Article
- Degryse E: Stability of a host-vector system based on complementation of an essential gene in Escherichia coli. J Biotechnol. 1991, 18: 29-39. 10.1016/0168-1656(91)90233-L.View Article
- Cranenburgh RM, Hanak JA, Williams SG, Sherratt DJ: Escherichia coli strains that allow antibiotic-free plasmid selection and maintenance by repressor titration. Nucleic Acids Res. 2001, 29: E26- 10.1093/nar/29.5.e26.View Article
- Hagg P, de Pohl JW, Abdulkarim F, Isaksson LA: A host/plasmid system that is not dependent on antibiotics and antibiotic resistance genes for stable plasmid maintenance in Escherichia coli. J Biotechnol. 2004, 111: 17-30. 10.1016/j.jbiotec.2004.03.010.View Article
- Fiedler M, Skerra A: ProBA complementation of an auxotrophic E. coli strain improves plasmid stability and expression yield during fermenter production of a recombinant antibody fragment. Gene. 2001, 274: 111-8. 10.1016/S0378-1119(01)00629-1.View Article
- Vidal L, Pinsach J, Striedner G, Caminal G, Ferrer P: Development of an antibiotic-free plasmid selection system based on glycine auxotrophy for recombinant protein overproduction in Escherichia coli. J Biotechnol. 2008, 134: 127-36. 10.1016/j.jbiotec.2008.01.011.View Article
- Pfaffenzeller I, Mairhofer J, Striedner G, Bayer K, Grabherr R: Using ColE1-derived RNA I for suppression of a bacterially encoded gene: implication for a novel plasmid addiction system. Biotechnol J. 2006, 1: 675-81. 10.1002/biot.200600017.View Article
- Mairhofer J, Pfaffenzeller I, Merz D, Grabherr R: A novel antibiotic free plasmid selection system: advances in safe and efficient DNA therapy. Biotechnol J. 2008, 3: 83-9. 10.1002/biot.200700141.View Article
- Luke J, Carnes AE, Hodgson CP, Williams JA: Improved antibiotic-free DNA vaccine vectors utilizing a novel RNA based plasmid selection system. Vaccine. 2009, 27: 6454-6459. 10.1016/j.vaccine.2009.06.017.View Article
- Goh S, Good L: Plasmid selection in Escherichia coli using an endogenous essential gene marker. BMC Biotechnol. 2008, 11: 61-10.1186/1472-6750-8-61. 10.1186/1472-6750-8-61.View Article
- Soubrier F, Cameron B, Manse B, Somarriba S, Dubertret C, Jaslin G, Jung G, Caer CL, Dang D, Mouvault JM, Scherman D, Mayaux JF, Crouzet J: pCOR: a new design of plasmid vectors for nonviral gene therapy. Gene Ther. 1999, 6: 1482-8. 10.1038/sj.gt.3300968.View Article
- Szpirer CY, Milinkovitch MC: Separate-component-stabilization system for protein and DNA production without the use of antibiotics. Biotechniques. 2005, 38: 775-81. 10.2144/05385RR02.View Article
- Palmeros B, Wild J, Szybalski W, Le Borgne S, Hernández-Chávez G, Gosset G, Valle F, Bolivar F: A family of removable cassettes designed to obtain antibiotic-resistance-free genomic modifications of Escherichia coli and other bacteria. Gene. 2000, 247: 255-64. 10.1016/S0378-1119(00)00075-5.View Article
- Datsenko KA, Wanner BL: One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc Natl Acad Sci USA. 2000, 97: 6640-5. 10.1073/pnas.120163297.View Article
- Mizoguchi H, Tanaka-Masuda K, Mori H: A simple method for multiple modification of the Escherichia coli K-12 chromosome. Biosci Biotechnol Biochem. 2007, 71: 2905-11. 10.1271/bbb.70274.View Article
- Summers DK, Sherratt DJ: Multimerization of high copy number plasmids causes instability: CoIE1 encodes a determinant essential for plasmid monomerization and stability. Cell. 1984, 36: 1097-103. 10.1016/0092-8674(84)90060-6.View Article
- Balding C, Blaby I, Summers DA: Mutational analysis of the ColE1-encoded cell cycle regulator Rcd confirms its role in plasmid stability. Plasmid. 2006, 56: 68-73. 10.1016/j.plasmid.2005.12.001.View Article
- Stirling CJ, Stewart G, Sherratt DJ: Multicopy plasmid stability in Escherichia coli requires host-encoded functions that lead to plasmid site-specific recombination. Mol Gen Genet. 1988, 214: 80-4. 10.1007/BF00340183.View Article
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.