Inducible complementation for antibiotic-free plasmid-based biomanufacturing in industrially relevant strains

Antibiotic-based plasmid maintenance is a core method in molecular biotechnology; however, reliance on antibiotics in biomanufacturing has significant drawbacks. Antibiotic use in novel industrial processes is being evaluated with increasing scrutiny due to rising levels of antimicrobial resistance (AMR). From an economic standpoint, antibiotic use is undesirable as supplementation on a manufacturing scale is prohibitively expensive and adds production costs, including purification steps to eliminate residual antibiotics and wastewater pre-treatment to prevent environmental contamination. Recently, we introduced a method for antibiotic-free plasmid selection and maintenance in Escherichia coli that retains the flexibility and ease of use enjoyed by antibiotic-based selection. This method uses inducible complementation of an essential gene. An essential gene in the production strain was replaced with an inducible copy so that in the absence of the inducer, a plasmid expressing the essential gene is required for growth. The ability to induce the genomic copy of the essential gene enables the engineered stain to be easily transformed with a complementing plasmid without requiring antibiotics, as non-supplemented media selects for plasmid uptake. Following the introduction of this method in an E. coli cloning strain, this study is designed to characterize and demonstrate the application of the method in various industrially relevant E. coli strains. Here, we modify a wide range of E. coli laboratory strains from the phylogenetic groups K-12, B and W as well as the probiotic strain Nissle1917, and demonstrate robust, antibiotic-free plasmid maintenance in all strain backgrounds in the absence of the inducer. We subsequently characterize growth rates and protein expression in the engineered strains, and find the modified strains performed equally well or better than their unmodified counterparts. Finally, we use this technology to scale up antibiotic-free protein expression to 10,000 L, demonstrating compatibility with large-scale production and downstream processing.

Rising antimicrobial resistance (AMR) is a pressing global health threat with wide-reaching medical, humanitarian and economic implications. A comprehensive review of AMR predicted that AMR could be responsible for up to 10 million annual deaths by 2050 without intervening measures and that the corresponding lack of effective antibiotics could undo decades of medical progress (O’Neill, 2014). Recently, the United Nations General Assembly (UNGA) held a High-Level Meeting on AMR, highlighting both the urgency and political will to address AMR. The meeting concluded with the release of a political declaration emphasizing the need for a concerted, global effort to reduce antibiotic use in all relevant sectors (UNGA, 2024).

In biotechnology, antibiotic use is mainly associated with plasmid DNA propagation. Plasmid DNA is a core component of industrial biomanufacturing, acting as the blueprint to produce pharmaceuticals, industrial enzymes and a diverse portfolio of small molecules for a variety of markets. The drive for industries to adopt more sustainable production methods, coupled with rapidly increasing biotechnological capabilities, positions industrial biotechnology to play a critical role in the future of manufacturing. The use of plasmid DNA represents a central technology, forming the basis of a rapidly growing market; it is therefore critical to optimize plasmid use.

Historically, the use of plasmid DNA has relied on antibiotics and antibiotic resistance genes (ARG) to enable plasmid selection and maintenance in cells. With rising levels of AMR, antibiotic use in novel industrial processes is being evaluated increasingly critically. For example, the European Medicines Agency (EMA) stated in a recent guideline, “If genetically modified cells are used in the product, any additional proteins expressed from the vector, e.g. antibiotic resistance factors or other selection markers should be analyzed and their presence in the product should be justified” (European Medicines Agency, 2024). Eliminating the use of antibiotics and ARG for plasmid selection in industrial manufacturing simplifies meeting regulatory requirements. Operating antibiotic-free also has significant financial benefits. Antibiotic supplementation on an industrial scale is expensive and leads to collateral costs in downstream processing and wastewater pre-treatment (Straub et al., 2020). Due to the prohibitive cost of antibiotic supplementation on a manufacturing scale, antibiotics are often omitted in the production phase, allowing the fermentation to run without selection pressure. However, several studies have shown that it is the induction of protein expression that creates the metabolic burden that leads to plasmid loss, highlighting the importance of plasmid selection pressure under production conditions (Cumming et al., 2022; Khananisho et al., 2023).

There are various antibiotic-free plasmid maintenance technologies available that eliminate the need for antibiotic use. These methods impose selection pressure for plasmid presence using strategies such as post-segregational killing (PSK) or complementation. PSK systems achieve antibiotic-free plasmid maintenance through expression of a toxin-antitoxin pair. For example, the STABY ® technology uses plasmid-based expression of an antitoxin (ccdA) to neutralize a genomically encoded toxin (ccdB) (Reschner et al., 2013). Similarly, plasmid maintenance can be achieved through co-expression of a bacteriocin and the corresponding immunity gene on a plasmid; in this case, the diffusible bacteriocin from neighboring cells eliminates all plasmid-free cells in the culture (Gabant, 2013). In a related strategy, the RNA-OUT system uses a plasmid-encoded antisense regulator to inhibit expression of a genomically encoded sacB gene, which metabolizes sucrose to create a toxic metabolite (Luke et al., 2009). A drawback of these strategies, however, is the reliance on negative selectable markers (i.e. expression of a detrimental gene) because the selection pressure can be diffused by any mutation inactivating expression of the marker. Examination of loss-of-function (LOF) mutations in sacB revealed an estimated LOF mutational rate in vivo of approximately 0.001% per kb (Hersch et al., 2024). Since LOF mutations arise with a relatively high frequency, negative selectable markers are poorly suited for processes with extended growth periods. In contrast, complementation strategies, which use plasmid-based expression of an essential gene in a deficient host, rely on positive selection and are intrinsically more robust. Numerous complementation strategies have been demonstrated for antibiotic-free plasmid maintenance, using genes such as valS (Skogman and Nilsson, 1984), dapD (Degryse, 1991), glyA (Vidal et al., 2008), proAB (Fiedler and Skerra, 2001), and leuABCD (Xiao et al., 2016). Indeed, due to the high stability, complementation was recently evaluated and shown to be a useful tool to improve plasmid stability in continuous processing fermentation (Outram et al., 2025).

While well-established, plasmid maintenance through complementation has not received widespread adoption in E. coli, likely due to cumbersome handling. Since the engineered strain is not viable without a complementing plasmid, repetitive genome engineering or the use of a helper plasmid is required to generate new strain-plasmid combinations, unless the plasmid-free strain can be rescued by permissive conditions. Many auxotrophic strains can be rescued through media supplementation; however, this makes the process susceptible to cross-feeding (Shi and Wang, 2015; Outram et al., 2025). Alternatively, some complementation strategies have been devised using temperature-sensitive mutants to allow plasmid-free growth at a permissive temperature (Skogman and Nilsson, 1984; Pagni et al., 2022). However, this in turn restricts the manufacturing temperature, which is a critical factor for process optimization. In an interesting example, Cranenburgh and colleagues developed a complementation system where the engineered strain could be rescued using the inducer isopropyl-β-D-1-thiogalactopyranoside (IPTG) (Cranenburgh et al., 2001). Here, an essential gene was placed under the control of the IPTG-inducible lactose promoter (P lac) to enable antibiotic-free maintenance of any plasmid containing the lac operator (lacO); the lacO sites on the multicopy plasmid bind the lac repressor (lacI), relieving repression of the genomic P lac, which allows expression of the regulated essential gene. While this strategy addresses the challenge of cumbersome handing, the design is incompatible with P lac expression systems and requires balancing of lacI expression and lacO copies in new strain-plasmid combinations.

We recently introduced an antibiotic-free method for plasmid selection and maintenance in E. coli that retains the flexibility and ease of use characteristic of antibiotic-based selection systems (Brechun et al., 2024). Here, the promoter of an essential gene in the production strain is replaced with an inducible promoter, and the same essential gene is expressed constitutively on a plasmid. Without the inducer, the plasmid is therefore essential for growth. The ability to induce the genomic copy of the essential gene allows the engineered strain to be propagated without a plasmid, enabling new strain-plasmid combinations to be generated using a standard transformation without needing antibiotic selection. Inducible expression was achieved using the tightly controlled arabinose promoter (P Ara). The essential gene chosen is infA, encoding transcription initiation factor 1 (IF1). IF1 is small (219 bp), essential under all culture conditions, and cannot be taken up from the media, ensuring that cross-feeding does not occur (Cummings and Hershey, 1994). As such, infA has attracted significant attention for complementation, and its suitability for plasmid maintenance is well-established (Isaksson et al., 2002; Kang et al., 2018; Pagni et al., 2022; Sathesh-Prabu et al., 2022; Park et al., 2024; Outram et al., 2025; Yiakoumetti et al., 2025). By controlling the expression of the endogenous infA with a tight inducible promoter, we improve the handling of this strategy.

Previously, we introduced the strategy of conditional complementation for antibiotic-free plasmid maintenance and provided detailed characterization of the method in the cloning strain E. coli HS996 (Brechun et al., 2024). Here, we engineer infA complementation strains in additional E. coli strain backgrounds and provide detailed characterization of this method in all strains, demonstrating the effectiveness of this strategy outside of the E. coli cloning strain. Testing this mechanism in detail in different strain backgrounds is important as other selection methods have been noted to have host-dependent efficiencies (Cranenburgh et al., 2001; Liao et al., 2022). We include in our analysis several B and K-12 strains, as well as the more phylogenetically diverse strains, W and Nissle1917. We examine the activity of several T7-expression strains, including both the prominent B strain BL21(DE3) and a derivative of the K-12 T7-expression strain HMS174(DE3) (Mairhofer et al., 2014; Hausjell et al., 2018). E. coli W is the only laboratory strain able to metabolize sucrose, a preferred carbon source for industrial bioprocesses, and is therefore an important industrial fermentation host (Archer et al., 2011). E. coli Nissle1917 is a probiotic strain with GRAS (Generally Recognized As Safe) status that has been used to treat a variety of gastrointestinal disorders (Lynch et al., 2022). While we find that the engineered strains show expected strain-specific differences in growth rate, protein production levels and their propensity to lose a plasmid, the infA -engineered strains invariably performed equally well or better than their unmodified counterparts. Finally, we demonstrate successful scale-up with antibiotic-free protein production on an industrial scale. Therefore, the aim of this study was achieved, demonstrating not only the successful use of this technology in a wide range of industrially applicable E. coli strains, but also the application of this technology for antibiotic-free manufacturing of an industrial enzyme.

Bacterial strains and growth conditions

All cloning was performed in E. coli HS996. Strain engineering was performed in numerous strains, as outlined in Table 1 (see Section 3.1) and as described in detail in Supplementary Table S1. Reference strains and infA -modified strains with a complementing plasmid were grown in LB. All infA -modified strains without a complementing plasmid were grown in LB supplemented with 0.25% arabinose (Ara). When indicated, the following antibiotics were used: Chloramphenicol (Cm, 30 µg/ml for high-copy plasmids and 15 µg/ml for genomic expression), Zeocin (Zeo, 50 µg/ml), Kanamycin (Kan, 15 µg/ml), Tetracycline (Tet, 3 µg/ml) and Ampicillin (Amp, 100 µg/ml). Handling of bacterial cultures in the absence of antibiotic selection was performed in an MSC-Advantage™ Class II Biological Safety Cabinet (Thermo Fisher Scientific) and growth was accompanied with an uninoculated medium control to monitor for contamination.

Genome engineering

Genome engineering was performed using λ-Red recombineering with the helper plasmid pSC101-BAD-gbaA (CGB002), as described previously (Wang et al., 2006). Genome integration was performed by electroporating 200 – 1,000 ng of linear DNA with 50–200 bp of homology. Electroporation was performed using electrocompetent cells and an Eppendorf Electroporator2510 set to 1350 V. DNA fragments were prepared using standard PCR conditions, digested with DpnI and purified using a GeneJET PCR Purification kit (Thermos Fisher Scientific) with elution in water. All modifications were confirmed using Sanger sequencing (Microsynth AG). Recombinase-mediated marker deletion was performed using a low-copy, temperature-sensitive helper plasmid with temperature-inducible expression of Flp recombinase (CGB086) or Cre recombinase (CGB011), as described previously (Brechun et al., 2024). Marker deletion was confirmed using PCR and sequencing.

Plasmid assembly

All plasmids used and developed in this study are shown in Table 1. Plasmid CGB310 was assembled by using λ-Red recombineering to replace the original KanR sequence in CGB145 with a cassette containing the endogenous infA expression cassette (including the promoter and terminator) and a CmR expression cassette flanked by NotI recognition sites. The resulting plasmid had only two NotI sites flanking the Cm marker, allowing the marker to be removed by digesting with NotI and re-ligating the plasmid. The ligation was transformed into HS996- infA and plated on LB, using the lack of Ara for plasmid selection. Individual colonies were patched on LB and LB-Cm to identify clones that had lost CmR after re-ligation. Clones showing Cm sensitivity were further analyzed with restriction digestions and sequencing. Plasmid CGB323 was assembled from CGB310 by using λ-Red recombineering to insert a fragment containing the low dimerizing Emerald (mEmerald) coding sequence driven by the T7 promoter and a CmR cassette downstream of the infA coding sequence. To increase the plasmid copy number, the rop and bom regulatory sequences were removed by replacing them with a ZeoR cassette flanked by XmaJI sites. The Zeo marker was subsequently removed by digesting with XmaJI and re-ligating the plasmid. Zeo sensitive clones were further analyzed with restriction digestions and sequencing. Plasmid CGB333 was assembled from CGB310 using λ-Red recombineering to insert a fragment containing a Kan marker cassette and mStrawberry driven by a constitutive promoter.

Plasmid maintenance assays

To monitor plasmid presence, each strain was electroporated with the plasmid CGB285. Wild-type strains were plated on LB-Cm, and infA -modified strains were plated on LB. From each transformation, a single clone was inoculated in LB without antibiotics and grown at 37°C with agitation for 24 h. For the spot plates, the optical density at 600 nm (OD 600) was determined for each culture by measuring absorbance using an Eppendorf BioPhotometer. Each culture was adjusted to an OD 600 of 2.0, a 10-fold dilution series was generated in LB, and 2.5 µl of dilutions 10– 2 to 10– 7 were spotted on LB, LB-Ara and LB-Cm plates. The plates were incubated overnight at 37°C. The spot plates were performed 3 times per strain. To additionally analyze single colonies, the 24 h LB cultures were streaked on LB plates and incubated overnight at 37°C, extending the growth period in non-supplemented LB media to 48 h (i.e. 24 h in liquid culture, and 24 h on solid media). Then 18–24 isolated clones were patched on LB and LB-Cm to test for plasmid presence. Colonies were patched prior to development of pigment from the model plasmid. For the extended plasmid maintenance trials, the experiment was performed in the same manner with the exception that the cultures were grown at 37°C for a total of 96 h with passaging every 24 h prior to streaking the cultures. Cultures were passaged by inoculating fresh LB medium without antibiotics with an inoculating loop. The agar plates were stored at 4°C to allow for pigment development prior to imaging. Plate images were modified by adjusting global brightness and contrast.

To monitor the presence of the plasmid CGB323 after protein expression, the expression cultures were analyzed using the spot plates and colony patching described above, with the exception that the expression cultures were used directly to create the 10-fold dilution series for the spot plates and to create the LB-streak plates.

Strain genotyping

Cultures were genotyped using standard PCR conditions to amplify the infA locus. The following primers were used: Primer1 GAGCTTGTCGGCGGTATGTAC, Primer2 CGGACACCTCACCAAATTG. The wild-type infA locus yielded a 1059 bp fragment. The modified infA locus containing P Ara yielded a 2212 bp fragment.

Plasmid isolation and analysis

To recover plasmid from the 96 h cultures, a single colony was inoculated into 10 ml of LB and grown overnight at 37°C with 250 rpm. Plasmid was recovered using a GeneJET Plasmid Miniprep kit (Thermo Fisher Scientific) and subsequently digested using PstI. To isolate plasmid from the Nissle1917 culture, a larger culture volume (150 ml) was used to perform a Maxiprep from a frozen pellet using the PureLink, HiPure Plasmid Maxiprep Kit (Invitrogen), due to difficulties obtaining plasmid from Nissle1917 cultures.

Microtiter plate analysis

Bacterial growth rates and protein expression levels were monitored in the multimode microplate reader Varioskan™ ALF (Thermo Fisher Scientific) and analyzed using SkanIt software (Thermo Fisher Scientific). Strains were transformed with the appropriate plasmid using non-supplemented LB for selection in infA -modified strains and antibiotics for selection in unmodified strains. Precultures were grown overnight in 1.4 ml of LB (with antibiotics for the unmodified strains) at 37°C in an Eppendorf Thermomixer with agitation. The overnight cultures were diluted 1:100 in LB (with antibiotics for the unmodified strains) in flat-bottom, clear microtiter plates with a lid. For the cultures with the constitutively expressed model protein, an additional preculture was created (50-fold dilution in 1.4 mL LB, 37°C with agitation for 6 h) prior to inoculation into the microtiter plates to avoid high starting fluorescence levels from the stationary phase cultures. For this analysis, only the overnight cultures of the wild-type strains were supplemented with antibiotics. The growth curves and constitutive protein expression experiments were performed in 96-well plates with a total volume of 200 µl. Expression of the IPTG-inducible model protein was performed in 6-well plates with a total volume of 2 ml. Plates were incubated at 37°C with non-continuous orbital shaking (5 s on, 10 s off) for 16–28 h. For inducible protein expression, the incubation temperature was modified; the cultures were incubated for 3 h at 37°C, induced with 1 mM IPTG and then incubated at 30°C for a further 24 h. OD 600 was monitored with measurements every 10–20 minutes. Fluorescence measurements for cultures containing CGB333 were performed using the Resorufin family filter pair (excitation 555 nm, and emission 615 nm), with filter bandwidths of 15 nm and 45 nm full width at half maximum (FWHM) for excitation and emission respectively. Fluorescence measurements for cultures with CGB323 were performed using the Fluorescein family filter pair (excitation 485 nm, and emission 525 nm), with filter bandwidths of 15 nm FWHM for both excitation and emission. Specific growth rates were calculated using linear regression from the blank-subtracted exponential growth phase.

Commercial-scale fermentation

The fermentations were performed using rich, defined media and fed-batch fermentation parameters similar to the conditions described previously (Stanbury et al., 2017; Ferreira et al., 2018). The exact conditions of this commercial fermentation are not disclosable. Conditions for the infA -modified strain and the unmodified parental strain were identical, with the exception that all the cultures of the unmodified strain were supplemented with Kan (for the 3 L and 750 L scale fermentations). The target protein produced in the fermentation is an enzyme with application in food.

Plasmid and sequence images

Plasmid maps and annotated DNA sequences were generated using SnapGene ® software (Dotmatics).

Strain development

Genome engineering for inducible expression of an essential gene

Each modified strain was engineered using λ-Red recombineering and recombinase-mediated deletion. Four infA -complementation strains were introduced previously [BL21(DE3) -infA, T7E2 -infA, HS996 -infA, W3110 -infA (Brechun et al., 2024)]; the remaining strains used in this analysis were developed in a similar manner. The final genomic sequence after modification is shown in Figure 1.

Strain optimization through inactivation of the DE3 phage

The K-12 T7 expression strain HMS174(DE3) was additionally modified to inactivate the DE3 prophage. Both BL21(DE3) and HMS174(DE3) express T7 RNA polymerase from a lysogenized DE3 phage, allowing protein expression to be driven using the strong T7 promoter (Studier and Moffatt, 1986). The lysogenized DE3 phage has been reported to spontaneously undergo int -independent prophage excision, leading to initiation of the lytic phage lifecycle (Shuman, 1989). This presents a costly risk for industrial scale biomanufacturing. To mediate this risk, we previously modified BL21(DE3) by removing the majority of the DE3 prophage to leave behind the lac repressor (lacI) and T7 RNA polymerase gene, creating the strain T7E2 (Noll et al., 2013). Likewise, we modified HMS174(DE3) to eliminate the risk associated with the DE3 prophage; 38 kb of the DE3 phage was deleted, leaving behind the T7 RNA polymerase and the lacI repressor. We refer to this strain as HMS174(T7). The strain modification steps are shown in the SI (Supplementary Figure S1).

The modified strains are summarized in Table 2. A detailed list of the strains with the genotypes or reference sequences can be found in the SI (Supplementary Table S1).

Plasmid maintenance

Investigation of short-term antibiotic-free plasmid maintenance

To examine plasmid maintenance, a model plasmid was used containing the wild-type infA sequence under the control of its endogenous promoter, a Cm R marker, a high-copy origin of replication (ColE1) and a constitutively expressed chromoprotein as a model gene of interest. The chromoprotein consists of the fluorescent proteins E2-Crimson and mStrawberry connected with a short flexible linker (Shaner et al., 2004; Strack et al., 2009; Liljeruhm et al., 2018). This plasmid is termed CGB285 (see Table 1 for a list of plasmids). A plasmid map is shown in the SI (Supplementary Figure S2). The plasmid was transformed into the infA engineered strains using a standard electroporation protocol, with the exception that prior to transformation, the plasmid-free cells were grown with Ara supplementation and after transformation, cells were plated on non-supplemented LB for plasmid selection. The wild-type strains were transformed with the same plasmid using Cm selection. A single colony of each strain was inoculated in non-supplemented LB media and incubated at 37°C with agitation for 24 h. Next, the liquid cultures were examined for plasmid maintenance using two complementary methods. First, serial dilutions were spotted on LB, LB-Ara and LB-Cm plates (Figure 2; Supplementary Figure S3). In parallel, the 24 h cultures were streaked on LB plates to isolate single colonies, and the resulting colonies were patched on LB and LB-Cm plates (Figure 3).

Plating cultures on different media allowed direct examination of the 24 h culture composition. The unmodified strains showed equal growth on LB and LB-Ara plates, and in most cases, decreased growth on the Cm plates, indicating loss of the plasmid during the 24 h incubation in LB. The colony pigmentation provides an indication of plasmid presence; the white colonies arising under non-selective conditions show the model plasmid with the constitutively expressed chromoprotein had been lost. In contrast, the infA -modified strains showed equal growth and uniform pigmentation (excluding the LB-Ara plates as discussed below) in all three conditions, indicating that the plasmid was maintained due to the selection pressure imposed by the lack of Ara in the 24 h culture. The LB-Ara plates provide non-selective conditions for the infA -modified strains. As a result, plasmid-loss can occur in these cells during colony formation, leading to variations in pigmentation. Plating an aliquot from the 24 h culture on LB-Ara tests whether the corresponding liquid culture contained infA -modified cells that had lost plasmid and ceased growth but remained viable. None of the strains showed an increase in colony numbers on LB-Ara plates, indicating that if cells had lost the plasmid in the LB culture, such cells were not able to proliferate enough to create an appreciable population. Representative results from 3 strain pairs are shown in Figure 2, while representative results from the remaining strains are shown in the SI (Supplementary Figure S3).

In parallel, each 24 h culture was streaked on LB plates, and the resulting isolated colonies were patched on LB and LB-Cm plates to examine plasmid maintenance in single cells. The isolated colonies were patched prior to development of the chromoprotein pigment. A representative image from one culture pair is shown in the SI (Supplementary Figure S4), and a summary of colonies containing plasmid-DNA from all cultures is shown in Figure 3. Streaking the cultures on LB extends the incubation period of the LB growth trial to 2 days. We note that LB provides selective conditions for plasmid maintenance for the infA -modified strains. Isolating colonies with non-selective conditions (i.e. LB-Ara) is problematic because plasmid loss can occur during colony formation (see phenotypic variation in Figure 2; Supplementary Figure S3), and the supplemented media does not represent relevant fermentation conditions.

The results of the two complementary methods of plasmid analysis are in good agreement. Both methods showed that the wild-type strains lost the plasmid to varying degrees, while the infA -modified cultures maintained the plasmid. The variation in plasmid-loss observed between replicate wild-type cultures is to be expected, as spontaneous events providing a growth advantage (e.g. plasmid loss or a mutation inactivating or decreasing chromoprotein expression) would lead to enrichment of these populations. The results also suggest that different strains have a different propensity to lose the plasmid, likely reflecting differences in genes involved in plasmid segregation and replication control; however, our results do not provide a characterization of plasmid-loss of the wild-type strains in non-selective conditions. The rapid loss of the model plasmid in the wild-type strains aligns with the previous observation that overexpression of chromoproteins impose a high metabolic burden, giving cells that have lost the plasmid a growth advantage (Liljeruhm et al., 2018). Despite the high metabolic burden posed by the model plasmid, all infA -modified strains showed tight plasmid maintenance regardless of the strain background in both methods of analysis.

Extended antibiotic-free plasmid maintenance

Next, an extended plasmid maintenance trial was performed. Previously, infA complementation in an infA deletion strain was shown to enable stable plasmid maintenance and high productivity in a continuous fermentation for over 1000 h, showing that infA complementation offers extremely robust selection (Yiakoumetti et al., 2025); here, we examine the stability of the system when using inducible complementation. For this, we focus on the infA -modified versions of four common industrial strains, BL21(DE3), Nissle1917, W3110 and W, as well as the Δ ara strain, HS996. It has been suggested that strains unable to metabolize Ara may show leaky expression with P Ara because the inducer persists in the cells (Guzman et al., 1995); HS996- infA therefore acts as a model for other Δ ara strains. For comparison, the unmodified HS996 and Nissle1917 strains were selected as controls because these strains showed a higher propensity to maintain the plasmid in the previous experiments. To mimic longer fermentations (e.g. for secondary metabolites), the cultures were incubated for 96 h with passaging every 24 h, and then the cultures were streaked on LB to isolate individual colonies. As before, colonies were patched on LB and LB-Cm to examine plasmid maintenance. All colonies from the infA -modified strains contained plasmid; despite the extended incubation, no mutants were recovered that could proliferate in LB media without plasmid (Figure 4). Importantly, the complete plasmid maintenance in the strain HS996- infA indicates that any potential leaky expression from P Ara caused by the Δ ara genotype is insufficient to compromise the selection pressure imposed by this system, confirming that this strategy is also compatible with Δ ara strains. The unmodified HS996 cultures had completely lost the plasmid after the extended incubation, while the unmodified Nissle1917 cultures showed partial plasmid loss. To confirm that the plasmid did not integrate into the genome, a single clone from each infA -modified strain after the incubation period was analyzed using primers to amplify the genomic infA locus. In addition, plasmid was isolated from each strain post incubation and analyzed with restriction enzymes. The colony PCR indicated the infA genomic locus had not changed (Supplementary Figure S5), and the restriction digest analysis showed the plasmid was present with the expected banding pattern (Supplementary Figure S6). The positive results from the extended trial indicate that this inducible complementation strategy is robust against mutations. Possible bypass mechanisms allowing infA -modified strains to grow in the absence of an infA -plasmid or Ara induction would include mutation of the genomic P Ara leading to uninduced expression, or genomic integration of an infA plasmid. The P Ara is well-known for its tight expression control and is commonly used when precise control is required (Guzman et al., 1995). Expression from P Ara is controlled by the regulator AraC, which blocks transcription in the absence of Ara and activates transcription when bound by Ara. While mutations in AraC have been found that make P Ara constitutively active (Dirla et al., 2009), these specific function-tuning mutations are less likely to arise compared to mutations silencing expression. During generation of the infA -modified strains, a copy of araC was integrated adjacent to the P Ara (see Figure 1), giving cells with an endogenous araC gene a second, ‘back-up’ copy to ensure repression in the uninduced state. Genomic integration of the infA plasmid was not detected in our analysis; however, if such recombination proves to be problematic, this could be eliminated by using alternative codons in one of the infA gene copies to reduce homology between the genomic and the plasmid copies. Alternatively, infA homologues from other organisms could be used (either in the genome or on the plasmid), as it has been shown that complementation is also possible with infA homologues (Kapralou et al., 2008).

Comparison of specific growth rates in engineered and unmodified strain pairs

We next examined strain fitness by monitoring the growth rate of the modified infA -inducible strains and the wild-type counterparts. To avoid bias from non-uniform protein expression, new model plasmids were used that differ only in their selection cassettes. Both plasmids contained a ColE1 origin of replication with the rop regulatory sequence, leading to low copy number plasmid maintenance, and either the wild-type infA sequence with its endogenous promoter (plasmid CGB310) or a Kan R cassette (plasmid CGB145). Plasmid maps are shown in the SI (Supplementary Figures S7, S8). For each strain, the specific growth rate was compared between the wild-type and infA -modified strain (Figure 5). No statistical differences between any of the infA -modified and wild-type strain pairs were observed (2-tailed Welch’s t -test, p = 0.05). This indicates that redirection of infA transcription to the plasmid does not impair the strains. Overall, the strain-dependent differences in growth rates are consistent with previous reports. As expected, the W strains showed faster growth rates compared to the K-12 and B strains (Sabri et al., 2013). BL21(DE3) and T7E2 showed similar growth rates, consistent with the fact that T7E2 is a close derivative of BL21(DE3) (Noll et al., 2013).

Comparison of protein expression levels in engineered and unmodified strain pairs

Antibiotic-free protein expression using a constitutively expressed reporter

Next, we examined strain productivity by monitoring the production of a fluorescent protein. We used the plasmid CGB333 to monitor constitutive expression of the reporter gene mStrawberry in all strain backgrounds. For each strain, precultures were inoculated in a 96-well plate, and the fluorescence and optical density were monitored. The fluorescence results were normalized to account for variation in growth. Analysis of the mean fluorescence after 18 h revealed comparable fluorescence signals in all strain pairs (Figure 6). As expected, the B strains (BL21(DE3) and T7E2) show similar fluorescence levels, reflecting their common genetic background, and the K-12 strains showed slightly higher fluorescence, consistent with higher plasmid copy numbers in K-12 strains. The mean fluorescence was compared for each strain pair using a 2-tailed Welch’s t -test. There is moderate, but not strong, evidence that there is a difference observed in three strain pairs (alpha = 0.05 for W3110, HMS174(T7) and W); none of the strain pairs have a statistically significant difference when alpha is set to 0.01. The observed difference could be caused by the small sample size (n = 4) and relatively high biological variability due to subtle differences in the culture conditions. Overall, the mean fluorescence values in all strain pairs are highly similar.

Antibiotic-free protein expression using T7-inducible expression

Next, we examined inducible expression of a fluorescent protein using the T7-expression strain backgrounds, BL21(DE3), T7E2 and HMS174(T7). For this analysis, a plasmid was assembled containing mEmerald under the control of the T7 promoter, as well as a Cm marker and the infA coding sequence to enable antibiotic and infA -based selection respectively. A map of the plasmid, termed CGB323, is shown in the SI (Supplementary Figure S10). The wild-type and the infA -modified T7-expression strains were transformed with CGB323 using Cm or a lack of Ara for selection respectively. Expression cultures were inoculated in 6-well plates by diluting precultures 1:100 in fresh LB. The cultures were grown for 3 h at 37°C, induced using 1 mM IPTG and then incubated further for 24 h at 30°C while monitoring OD 600 and fluorescence. All wild-type cultures were supplemented with 30 µg/ml Cm to ensure plasmid selection pressure. The expressions were performed for each strain three times on separate days in order to account for variations in the precultures that could lead to differences at the point of induction, as such changes have been shown to significantly affect protein production levels (Mühlmann et al., 2017). Fluorescence curves are shown in Figure 7A. In agreement with the literature, the B-strains showed very similar fluorescence levels and produced significantly higher fluorescence than the K-12 strain pair. Process optimization would be expected to improve the productivity of the HMS174(T7) strains (Hausjell et al., 2018). Lower fluorescence values were observed in all wild-type cultures compared to the antibiotic-free expression cultures, despite supplementation of the wild-type cultures with Cm.

Examination of the growth curves showed a distinct difference between the infA -based cultures and the wild-type cultures. Initially, all cultures showed logarithmic growth that slowed shortly after induction (at 3 h), consistent with reallocation of resources from growth to expression. Then, while the cultures with infA -based plasmid maintenance showed slow but steady growth, the wild-type cultures showed a second rapid growth phase (Figure 7B). We hypothesized that the increase in growth rate in the wild-type cultures was due to plasmid loss despite Cm supplementation. We analyzed one culture from each strain after expression using the plating assays described previously. The results showed that the wild-type cultures contained plasmid-free cells; the wild-type cultures showed reduced growth when spotted on LB-Cm (Figure 7C) and relatively few isolated wild-type colonies retained Cm R (Figure 7D). Streaking the cultures on LB media to obtain isolated colonies likely leads to an overestimation of plasmid loss in the original LB-Cm expression culture; however, without induction, plasmid-loss during growth on the LB plates should be reduced (Khananisho et al., 2023). The infA -modified B-strain cultures showed similar growth on LB, LB-Ara and LB-Cm, indicating high plasmid maintenance. The HMS174(T7)- infA culture showed similar growth on LB and LB-Cm, but increased growth on LB-Ara. This indicates that the expression culture contained growth-arrested plasmid-free cells, suggesting some plasmid loss. However, the inability of these cells to grow on LB means that the plasmid-free cells cannot outcompete the productive, plasmid-containing cells in the LB culture. This plasmid loss is consistent with the observations of Khananisho and colleagues, who reported elevated plasmid loss despite antibiotic selection when using T7-based expression (Khananisho et al., 2023). They attributed this to the extremely high metabolic burden caused by T7-expression systems and subsiding antibiotic concentrations over the course of the fermentation. Likewise, we observed significant plasmid loss using T7-based expression, and we reason that the active Cm concentration in the wild-type cultures subsided due to acetylation from the Cm R cells, allowing plasmid-free cells to proliferate. In contrast, when using infA -based complementation for plasmid maintenance, the selective pressure remained constant because the selection mechanism does not require media supplementation; as a result, infA -based plasmid selection led to improve plasmid maintenance and productivity.

Scale-up of antibiotic-free protein expression to manufacturing scale

Next, the BL21(DE3)- inf A strain was evaluated for scaled-up production of an undisclosed enzyme with application in food. First, expression was compared in a 3-L fermenter using BL21(DE3)- inf A in an antibiotic-free expression and the parental BL21(DE3) strain with antibiotic selection (i.e. Kan added in the basal media). The resulting product was purified via homogenization, centrifugation, and a combination of ultrafiltration and diafiltration steps followed by bactofiltration. The standard downstream processing (DSP) was not affected by the use of the modified strain. In fact, the product yield and specific activity (Table 3) obtained after DSP were comparable in both fermentations.

Next, the antibiotic-free fermentation was scaled up to 750 L. Triplicate results showed high reproducibility with comparable product yield and specific activity, the latter being slightly higher than the one achieved with the strain carrying antibiotic resistance (Table 3). For production, the process was then successfully scaled up to 10,000 L delivering a commercial product with similar specific activity (Table 3).

Characterization of the modified strains has demonstrated that inducible infA complementation offers robust plasmid maintenance in a variety of E. coli strain backgrounds without compromising strain fitness or strain productivity. No proliferation of plasmid-free cells was detected in any of the infA -modified strains in selective growth conditions (i.e. the absence of Ara). Indeed, reliable selection pressure was observed in all test scenarios, even with model proteins that led to frequent plasmid loss in the wild-type strains. Plasmid-loss was observed in the wild-type strains even in cultures with antibiotic pressure, showing the risk associated with subsiding antibiotic concentrations. Although antibiotic-based plasmid maintenance is accepted as the current state-of-the-art, the method is in many ways poorly suited for industrial fermentation. The high cost of antibiotic supplementation on a production scale leads many manufacturers to forgo the use of antibiotics in the final production phase. The high variability in plasmid loss in the wild-type strains here demonstrates the unpredictability of expression in the absence of selection pressure. While the ability to tightly control fermentation parameters (e.g. nutrient availability, pH, agitation rate, temperature) can decrease the propensity for plasmid loss, the complete absence of selective pressure leaves a significant risk, a surprising flaw in an otherwise meticulously controlled procedure. In contrast, this inducible complementation strategy does not require supplementation, and the selection pressure remains constant over the course of the fermentation. Furthermore, the complete lack of antibiotic resistance genes (in both the strain and the plasmid) offers a significant advantage in terms of regulatory compliance. It is also noteworthy that no additional regulatory concerns would be expected to arise with this technology, as the modified strains do not contain any foreign DNA; the integrated sequence (araC-P Ara) is of E. coli origin. The successful manufacturing-scale fermentation demonstrated here shows the compatibility of this method with large-scale fermentations and DSP. Overall, the current study shows that using infA as a selectable marker through conditional complementation results in flexible and reliable plasmid maintenance without any detectable reduction in strain fitness or productivity, making this a very promising technology for improved plasmid selection. Through eliminating the need for antibiotic-based plasmid selection, this technology offers a greener and more cost-efficient solution for biomanufacturing.