Abstract

Understanding the evolutionary stability and possible context dependence of biological containment techniques is critical as engineered microbes are increasingly under consideration for applications beyond biomanufacturing. While synthetic auxotrophy previously prevented Escherichia coli from exhibiting detectable escape from batch cultures, its long-term effectiveness is unknown. Here, we report automated continuous evolution of a synthetic auxotroph while supplying a decreasing concentration of essential biphenylalanine (BipA). After 100 days of evolution, triplicate populations exhibit no observable escape and exhibit normal growth rates at 10-fold lower BipA concentration than the ancestral synthetic auxotroph. Allelic reconstruction reveals the contribution of three genes to increased fitness at low BipA concentrations. Based on its evolutionary stability, we introduce the progenitor strain directly to mammalian cell culture and observe containment of bacteria without detrimental effects on HEK293T cells. Overall, our findings reveal that synthetic auxotrophy is effective on time scales and in contexts that enable diverse applications.

New safeguards are needed for the deliberate release of engineered microbes into the environment, which has promise for applications in agriculture, environmental remediation, and medicine (1). Genetically encoded biocontainment strategies enable attenuation of engineered live bacteria for diverse biomedical applications (24), including as potential vaccines (510), diagnostics (11), and therapeutics (1215). Auxotrophy, which is the inability of an organism to synthesize a compound needed for its growth, is an existing strategy for containment. However, foundational studies of auxotrophic pathogens demonstrated proliferation in relevant biological fluids (16) and reversion to prototrophy upon serial passaging (17, 18). Modern genome engineering strategies can prevent auxotrophic reversion, and auxotrophy has been a key component of microbial therapies that have reached advanced clinical trials. However, the ability for auxotrophs to access required metabolites within many host microenvironments, and after leaving the host, remains unaddressed. Auxotrophy may not be effective in scenarios where engineered living bacteria encounter metabolites from dead host cells (19) or invade host cells (20). Growth of double auxotrophs is supported in vivo by neoplastic tissue (13). Auxotrophy may also be insufficient for tight control of cell proliferation in environments rich with microbial sources of cross-feeding (21), such as gut, oral, skin, and vaginal microbiomes. Given that most naturally occurring microorganisms are auxotrophs (22), it is also unlikely that auxotrophy will limit the spread of an engineered microbe once it leaves the body and enters the environment.

Synthetic auxotrophy may overcome these hurdles by requiring provision of a synthetic molecule for survival of the engineered bacteria. This strategy was first implemented successfully in Escherichia coli by engineering essential proteins to depend on incorporation of a nonstandard amino acid (nsAA) (23, 24). We previously engineered E. coli strains for dependence on the nsAA biphenylalanine (BipA) by computer-aided redesign of essential enzymes in conjunction with expression of orthogonal translation machinery for BipA incorporation (23). Among several synthetic auxotrophs originally constructed, one strain harbored three redesigned, nsAA-dependent genesadenylate kinase (adk.d6), tyrosyl-tRNA synthetase (tyrS.d8), and BipA-dependent aminoacyl-tRNA synthetasefor aminoacylation of BipA (BipARS.d6). This BipA-dependent strain, dubbed DEP, exhibited undetectable escape throughout 14 days of monitoring at an assay detection limit of 2.2 1012 escapees per colony-forming unit (CFU) (23). Although this strain demonstrates effective biocontainment in 1-liter batch experiments, its precise escape frequency and long-term stability remained unexplored.

Here, we perform the first study of evolutionary stability of a synthetic auxotroph with the aid of automated continuous evolution. Continuous evolution better emulates scenarios where biocontainment may be needed by fostering greater genetic variability within a population. We posited that decreasing BipA concentrations would add selective pressure for adaptation or for escape, either of which would be enlightening. Adaptive laboratory evolution of DEP may improve its fitness in relevant growth contexts, as previously demonstrated for its nonauxotrophic but recoded ancestor, C321.A (25). We report that DEP maintains its inability to grow in the absence of synthetic nutrient, even after three parallel 100-day chemostat trials. In addition, we find evidence of adaptation, with evolved DEP isolates requiring 10-fold lower BipA concentration to achieve optimal growth than ancestral DEP (0.5 M rather than 5 M). We resequence evolved populations and perform allelic reconstruction in ancestral DEP using multiplex automatable genome engineering (MAGE), identifying alleles that partially restore the adaptive phenotype. Last, we advance this technology toward host-microbe coculture applications, demonstrating direct mixed culture of DEP and mammalian cells without the need for physical barriers or complex fluidics.

To perform continuous evolution of E. coli, we constructed custom chemostats for parallelized and automated culturing (Fig. 1A). Our design and construction were based on the eVOLVER system (26), an open-source, do-it-yourself automated culturing platform (figs. S1 to S4). By decreasing BipA concentration over time in our chemostats, we provide an initial mild selection for escape and steadily increase its stringency. This design is analogous to a morbidostat, where a lethal drug is introduced dynamically at sublethal concentrations to study microbial drug resistance (27), but with synthetic auxotrophy providing selective pressure. Our working algorithm for automated adjustment of BipA concentration as a function of turbidity is shown in Fig. 1B, and a representative image of our hardware is shown in Fig. 1C (see also fig. S5).

(A) Illustration of a smart sleeve connected to separate nonpermissive media and biphenylalanine (BipA; structure shown in blue) feed lines for automated adjustment of BipA concentration based on growth rate. Pumps and optics are integrated with Arduino controller hardware and Python software based on the eVOLVER do-it-yourself automated culturing framework. (B) Working algorithm for maintenance of cultures in continuous evolution mode. Criteria for lowering the BipA concentration are based on the difference in time elapsing between OD peaks (tpeak OD). Smaller time elapsed between OD peaks is indicative of higher growth rates, triggering decrease in BipA concentration when below a threshold value. (C) Representative configuration of hardware for parallelized evolution in triplicate, with three empty sleeves shown. Photo credit: Michael Napolitano, Harvard Medical School.

Our long-term culturing experiments featured two phases. The first phase included one chemostat (N = 1) that was inoculated with DEP for an 11-day incubation, with an initial concentration of BipA of 100 M and automated adjustment based on growth rate (Fig. 2A). Because we observed no colony formation when the outgrowth from this population was plated on nonpermissive media, we then began a second phase in replicate. We used our population grown for 11 days to inoculate three chemostats in parallel (N = 3) where BipA supply decreased automatically over the following 90 days from 100 M to nearly 100 nM. One controller provided identical BipA concentrations to all three vials at any given time. To determine whether the decrease in BipA supply was due to escape from dependence on BipA, we periodically performed escape assays. We continued to observe no escape, including when we seeded liter-scale cultures and plated the associated outgrowth on nonpermissive media. Evolved isolates were obtained after this procedure (fig. S6), and th
eir growth was characterized across BipA concentrations (Fig. 2B and fig. S7). At 0.5 to 1 M BipA, we observed growth of all evolved isolates and no growth of the ancestral DEP strain.

(A) Timeline for continuous evolution, with detection limits for escape frequency assays shown in parentheses. (B) Doubling times of progenitor and evolved synthetic auxotrophs as a function of BipA concentration, normalized to the doubling time of DEP at 100 M BipA. Error bars represent the SD across technical triplicates within the same experiment.

To identify the causal alleles contributing to decreased BipA requirement of all three evolved isolates, we performed whole-genome sequencing and mutational analysis. We expected that mutations in auxotrophic markers or orthogonal translation machinery associated with aminoacylation of BipA would be observed. However, no variants were detected in the plasmid-expressed orthogonal translation machinery (aminoacyl-tRNA synthetase and tRNA) reference sequence. Instead, in all three evolved isolates, variants were observed in three nonessential genes, all of which are implicated in molecular transport: acrB, emrD, and trkH (Fig. 3A). AcrB and EmrD are biochemically and structurally well-characterized multidrug efflux proteins (28), and TrkH is a potassium ion transporter (29). These exact mutations have no precedent in the literature to our knowledge. Because they are missense mutations or in-frame deletions, it is unclear whether they cause loss of function or altered function (table S1). Because permissive media contain four artificial targets of efflux (BipA, l-arabinose, chloramphenicol, and SDS), mutations that confer a selective advantage during continuous evolution could disable BipA/l-arabinose efflux, improve chloramphenicol/SDS efflux, or affect transport of these or other species more indirectly. Given the strong selective pressure enforced by decreasing BipA concentration, we hypothesize that mutations observed are more likely to affect BipA transport. We also observed mutations in all evolved populations to the 23S ribosomal RNA (rRNA) gene rrlA (table S2). 23S rRNA mutations have been found to enhance tolerance for D-amino acids (30) and -amino acids (31). However, 23S rRNA mutations could also be related to increased tolerance of chloramphenicol (32).

(A) List of alleles identified through next-generation sequencing. Sequencing results originally obtained during the project identified this EmrD allele as a 33-bp deletion, which was then reconstructed in the experiment shown in (B). However, resequencing performed at the end of the project identified the allele as a 39-bp deletion and was confirmed by Sanger sequencing. A repetitive GGCGCG nucleotide sequence corresponding to G323-A324 and G336-A337 creates ambiguity about the precise positional numbering of the deletion. However, the three possible 13amino acid deletions (323335, 324336, and 325337) result in the same final protein sequence. (B) Effect of reconstructed allele in DEP progenitor on doubling time as a function of BipA concentration, normalized to the doubling time of DEP at 100 M BipA. Error bars represent the SD across technical triplicates within the same experiment.

To learn how identified transporter alleles may contribute to increased growth rates at low BipA concentration, we performed allelic reconstruction in the progenitor DEP strain using MAGE (33). Among four mutants that we generated in DEP, we observed growth of all mutants at 2 M BipA, a condition in which progenitor DEP could not grow (Fig. 3B and fig. S8). Furthermore, only emrD mutants exhibited near-normal growth at 1 M BipA. To investigate possible differential sensitivity of strains that contain reconstructed alleles to other media components of interest (SDS, l-arabinose, tris buffer, and chloramphenicol), we varied the concentration of these components and measured doubling times (fig. S9). We observed no significant deviation in doubling time from DEP in any of these cases. These results collectively suggest that observed transporter alleles are linked to BipA utilization.

The unobservable escape of DEP even after 100 days of evolution encouraged us to explore the possibility of an improved in vitro model for host-microbe interactions. In vitro models allow direct visualization and measurement of cells and effectors during processes such as pathogenesis (34). They are more relevant than animal studies for several human cell-specific interactions due to biological differences across animal types (35, 36). A nonpathogenic E. coli strain engineered to express heterologous proteins could be particularly useful for studying or identifying virulence factors and disease progression. However, an obstacle associated with coculture of microbial and mammalian cells is microbial takeover of the population. Approaches used to address this are bacteriostatic antibiotics (37), semipermeable Transwell membranes (3840), microcarrier beads (41), microfluidic cell trapping (42), peristaltic microfluidic flow (43, 44), and microfluidic perfusion (45). However, the use of a well-characterized synthetic auxotroph capable of limited persistence could offer a superior alternative for spatiotemporal control of microbial growth, especially for studying longer duration phenomena such as chronic infection or wound healing. Our study demonstrates how temporal control can be achieved by removal of BipA; we anticipate that spatial control could be achieved by patterning BipA onto a variety of solid surfaces with limited diffusion, such as a skin patch.

We investigated mammalian cell culture health, growth, and morphology after simple transient exposure to a hypermutator variant of DEP that we engineered by inactivating mutS during allelic reconstruction (DEP*). The use of DEP* rather than DEP is yet another form of a stress test to increase opportunity for escape under coculture conditions. We directly cocultured adherent human cell line human embryonic kidney (HEK) 293T with either no bacteria, nonauxotrophic E. coli DH5, or DEP* overnight (24 hours). HEK293T cells were cultured in selection media that allow only growth of desired but not contaminant strains while selecting for bacterial plasmid maintenance. After coculture, we washed cells and replenished cells with media varying in inclusion of BipA and/or an antibiotic cocktail (penicillin/streptomycin/amphotericin B). We continued incubation and imaged cells at days 2, 4, and 7 after initial coincubation. HEK293T cells contain a copy of mCherry integrated into the AAVS1 locus, and they appear red. DH5 and DEP* were transformed with Clover green fluorescent protein before coculture and appear green.

Compared to the control culture where bacteria were not added (Fig. 4A), HEK293T cells cocultured with DH5 display visible bacterial lawns with no attached human cells in the absence of the antibiotic cocktail at all days of observation (Fig. 4B). In the presence of antibiotic, cocultures containing DH5 sharply transition from bacterial overgrowth to apparent bacterial elimination (Fig. 4C). In contrast, cells cocultured with DEP* in the absence of BipA exhibited similar morphology to the control at all days of observation and no detectable bacteria by fluorescence microscopy on day 7, without the need for antibiotics to achieve bacterial clearance (Fig. 4D). Thus, DEP* addition was not detrimental to HEK293T cells in the absence of BipA, and DEP* remains biocontained and cannot survive because of cross-feeding. Clearance of bacterial cells from human cells appears to occur faster for DEP* when not provided BipA (Fig. 4D) than for DH5 when provided with the antibiotic cocktail (Fig. 4C).

Bacteria were added to HEK293T cell cultures and coincubated for 24 hours before washing and replenishing media. HEK293T cells express mCherry, whereas bacterial cells express Clover green protein marker. Images were taken at days 2, 4, and 7 after coincubation. (A) Untreated HEK293T cells. (B) HEK293T with commercial E. coli DH5 in the absence of antibiotic cocktail. (C) HEK293T with DH5 in presence of antibioti
c cocktail. (D) HEK293T and DEP* (mismatch repair inactivated to create hypermutator phenotype) in the absence of BipA. (E) HEK293T cells and DEP* in the presence of BipA. (F) HEK293T and DEP* in the absence of BipA until day 2 [identical at this point to condition in (D)], and then 100 M BipA was added to this condition daily until day 7.

To learn how the synthetic auxotroph behaves when supplied its essential nutrient in these coculture settings, we tested DEP* cocultures with continual resupply of 100 M BipA. Here, DEP* proliferates and in turn decreases proliferation and viability of HEK293T cells (Fig. 4E). A bacterial lawn begins to form on day 2, and at later times, human cell debris is overtaken by DEP*. This demonstrates that DEP* is fully capable of taking over the coculture if supplied with BipA. Replicates for these experiments can be found in figs. S10 to S12.

Given that DEP* grows in cocultures when BipA is provided, we sought to understand whether it could be rescued by readdition of BipA after multiple days of withholding. The possible time scale of reemergence influences applications where the duration of bacterial activity would need to be prolonged and/or repeated via limited BipA introduction while remaining contained. We find that coculturing DEP* with HEK293T cells for 2 days in the absence of BipA followed by the addition of BipA at day 2 does not rescue the DEP* growth (Fig. 4F and fig. S13). Human cells still grow and look morphologically similar to untreated cells, and bacteria are not visible. To look at analogous questions for nonauxotrophic E. coli, we removed antibiotics after 2 days of coculturing and do not observe bacterial rescue (fig. S13). We also investigated whether bacterial clearance could be delayed by the addition of antibiotic after some growth of DH5. DH5 cells grown in the absence of the antibiotic cocktail for 2 days before addition of the cocktail and maintenance to day 7 result in bacterial lawns (fig. S13, A and D). This demonstrates that antibiotic cocktails ordinarily used in mammalian cell culture maintenance can become ineffective beyond a certain amount of nonauxotrophic bacterial growth, whereas synthetic auxotrophy is subject to fewer and different constraints.

To further investigate the persistence of progenitor DEP and its evolved descendants, we performed BipA readdition studies in Lennox lysogeny broth (LB-Lennox) monoculture. Within 7 hours of BipA removal, DEP cell populations that are harvested from midexponential or stationary phases can be reactivated upon delayed BipA addition with unperturbed growth kinetics after a highly tunable lag phase (fig. S14). Further studies are ongoing to investigate the amount of time after which BipA reintroduction can recover growth of synthetic auxotrophs under different contexts.

We have shown that synthetic auxotrophy can exhibit long-term stability and function in unique contexts, enabling reliable control of microbial proliferation. Recent work has also shown that the escape rate and fitness of multiple synthetic auxotrophs can be improved by increasing the specificity of nsAA incorporation machinery (46). Collectively, these engineering and characterization efforts advance synthetic auxotrophy as a powerful safeguard for basic and applied research when using engineered microbes.

Cultures for general culturing, growth rate assays, biocontainment escape assays, MAGE, and fluorescent protein assays were prepared in LB-Lennox medium [bacto tryptone (10 g/liter), sodium chloride (5 g/liter), and yeast extract (5 g/liter)] supplemented with chloramphenicol (15 g/ml), 0.2% (w/v) l-arabinose, 20 mM tris-HCl buffer, 0.005% SDS, and variable concentration of L-4,4-biphenylalanine (BipA). Unless otherwise indicated, all cultures were grown in 96-well deep plates in 300 l of culture volumes at 34C and 400 rpm. The above media are permissive for growth of the synthetic auxotroph. Nonpermissive media are identically formulated as permissive media except for BipA, which is not included.

Construction of appropriate fluidics and chambers followed the eVOLVER framework (26) (figs. S1 and S2). The following components were included: (i) fluidics and chambers (reactor vial, inlet and outlet lines, filters, pumps, stirrers, and inlet and outlet reservoirs); (ii) light source and detector (LED and photodiode); (iii) controller hardware (circuit and microprocessors); and (iv) controller software (Arduino for controlling tasks, Raspberry Pi for computing tasks, and Python code for programming tasks) (full build of materials included in table S3). Briefly, our apparatus consisted of a custom smart sleeve (fig. S3), with the following modifications: Each vial was constructed without temperature control and was supplied by two media pumps (one for permissive media and another for nonpermissive media) and connected to one waste pump. All pumps were RP-Q1 from Takasago Fluidics, each driven off a standard N power MOSFET (metal oxide semiconductor field-effect transistor) with an Arduino controlling the gate. Like the eVOLVER system, we installed a stirring fan underneath each sleeve that consisted of magnets attached to a computer fan. By including a small stir bar within each reactor vial, we enabled efficient mixing of 1-ml working volumes. To enable automated measurement of turbidity [optical density (OD)], we used a 605-nm LED (LO Q976-PS-25) and an OPT101P-J photodiode detector. We mounted the LED and detector on custom printed circuit boards mounted to the vial sleeve to enable easier construction and better control of ambient light leakage into the light path (fig. S4). To monitor turbidity within each vial and to control pump arrays in response, we constructed printed circuit board designs in Gerber format as is standard for circuit fabrication. We attached an Arduino Mega microcontroller with an analog-digital converter and directed it using a PyMata script (47).

Chemostats were operated by automated maintenance of culture OD within a specified parameter range within exponential growth phase (20 to 80% of dynamic range) depending on linearity of photodiode measurements. Constant fixed dilutions of permissive media were used to decrease OD until desired equilibrium of cell growth and dilution rates. This resulted in a sawtooth curve (27), where time between peaks is recorded as a proxy for growth rate. Our program gradually decreased the ratio of permissive to nonpermissive media as step functions, with a specified number of dilution cycles allowed to elapse before the next decrease to provide time for acclimation. Time between OD peaks lengthened as strain fitness decreased. Once a threshold difference between ancestral peak-to-peak time and current peak-to-peak time was passed, the ratio of permissive to nonpermissive media remained fixed. This allowed cells to evolve until peak-to-peak time returns to ancestral values, which initiated the next phase of decrease in BipA concentration. To assess the quality of our continuous evolution process, we paused chemostat trials on a weekly basis for strain storage, strain evaluation, chemostat cleaning, and investigation of contamination.

Growth assays were performed by plate reader with blanking as previously described (25). Overnight cultures were supplemented with different BipA concentrations depending on the strain. The DEP progenitor strain was grown in permissive media containing 100 M BipA, and evolved DEP strains DEP.e3, DEP.e4, and DEP.e5 were grown in permissive media containing 1 M BipA. Saturated overnight cultures were washed twice in LB and resuspended in LB. Resuspended cultures were diluted 100-fold into three 150-l volumes of permissive media. BipA concentrations used in this assay were 0, 0.001, 0.01, 0.1, 0.5, 1, 10, and 100 M. Cultures were incubated in a flat-bottom 96-well plate (34C, 300 rpm). Kinetic growth (OD600) was monitored in a Biotek Eon H1 microplate spectrophotometer reader at 5-min intervals for 48 hours. The doubling times across technical replicates were calculated as previously indicated. We refer to these as technical replica
tes because although triplicate overnight cultures were used to seed triplicate experiment cultures, the overnight cultures were most often seeded from one glycerol stock.

Escape assays were performed as previously described with minor adjustments to decrease the lower detection limit for final evolved populations (23, 46). Strains were grown in permissive media and harvested in late exponential phase. Cells were washed twice with LB and resuspended in LB. Viable CFU were calculated from the mean and SEM of three technical replicates of 10-fold serial dilutions on permissive media. Twelve technical replicates were plated on noble agar combined with nonpermissive media in 500-cm2 BioAssay Dishes (Thermo Fisher Scientific 240835) and monitored daily for 4 days. If synthetic auxotrophs exhibited escape frequencies above the detection limit (lawns) on nonpermissive media, escape frequencies were calculated from additional platings at lower density. The SEM across technical replicates of the cumulative escape frequency was calculated as previously indicated.

Genomic DNA was obtained from evolved populations and ancestral clone using the Wizard Genomic DNA purification kit (Promega). Sequencing libraries were prepared as described in Baym et al. (48). Sequencing was performed using a NextSeq instrument, producing 75base pair (bp), paired-end reads. Resulting data were aligned to the E. coli C321.delA nonauxotrophic but recoded reference sequence (GenBank no. CP006698.1) and the sequence of the plasmid encoding nsAA incorporation machinery. The Millstone software suite was used to identify variants, provide measures of sequencing confidence, and predict their likelihood of altering gene function (49). Genomic variants of low confidence, low sequence coverage, or presence in the ancestral strain were discarded, prioritizing variants observed in three nonessential genes that encode membrane proteins: acrB, emrD, and trkH.

Subsequent genomic sequencing was performed on genomic DNA extracted from the evolved populations and ancestral clone using the DNeasy Blood and Tissue Kit (Qiagen). Genomic DNA was then sent to the Microbial Genome Sequencing Center (MiGS) in Pittsburgh, PA. Variants were identified through the variant calling service from MiGS.

MAGE (33) was used to inactivate the endogenous mutS gene in the DEP strain. Overnight cultures were diluted 100-fold into 3 ml of LB containing chloramphenicol, BipA, l-arabinose, and tris-HCl buffer and grown at 34C until midlog. The genome-integrated lambda Red cassette in this C321.A-derived strain was induced in a shaking water bath (42C, 300 rpm, 15 min), followed by cooling the culture tube on ice for at least 2 min. The cells were made electrocompetent at 4C by pelleting 1 ml of culture (8000 rcf, 30 s) and washing thrice with 1 ml of ice-cold 10% glycerol. Electrocompetent pellets were resuspended in 50 l of dH2O containing the desired DNA; for MAGE oligonucleotides, 5 M of each oligonucleotide was used. Allele-specific colony polymerase chain reaction (PCR) was used to identify desired colonies resulting from MAGE as previously described (50). Oligonucleotides used for MAGE and for allele-specific colony PCR are included in table S4.

This assay was performed using a similar protocol as described in the Measurement of doubling times section. The cultures for DEP and its single mutants were grown overnight in 100 M BipA. Then, cultures were diluted 100 in the media specified. Those conditions include standard media conditions and single component changes: 0% SDS, 0.01% SDS, 0.02% (w/v) arabinose, 0 mM tris-HCl, and chloramphenicol (30 g/ml). The cultures were grown in triplicate for each condition and in a SpectraMax i3 plate reader, shaking at 34C for 24 hours. The OD600 was measured about every 5 min. The doubling times were then calculated as previously described.

HEK293T cells containing one copy of mCherry marker (red) integrated into the AAVS1 locus were grown at 40 to 50% confluency in DMEM (Dulbeccos modified Eagles medium) high-glucose medium (Thermo Fisher Scientific, catalog no. 11965175) with 10% inactivated fetal bovine serum (FBS; Thermo Fisher Scientific, catalog no. 10082147), 100 MEM NEAA (nonessential amino acids; Thermo Fisher Scientific, catalog no. 11140050), and 100 diluted anti-anti cocktail [antibiotic-antimycotic: penicillin (10,000 U/ml), streptomycin (10,000 g/ml), and Gibco amphotericin B (25 g/ml); Thermo Fisher Scientific, catalog no. 15240112). Commercially acquired E. coli DH5 bacteria were used as control to the E. coli DEP mutS or DEP* strain. A plasmid containing Clover (green marker) containing a UAA stop codon compatible with the biocontained strain DEP, and under the selection marker ampicillin was transformed into both DH5 and DEP* strains to visualize them with the mammalian cells (red). BipA-dependent auxotroph DEP* bacteria were grown to an OD of 0.6 in LB medium supplemented with 1% l-arabinose, 100 M BipA, carbenicillin (100 g/ml), and chloramphenicol (25 g/ml) and then washed three times with 1 phosphate-buffered saline (PBS). DEP* culture conditions with l-arabinose, carbenicillin, and chloramphenicol supplements did slightly affect HEK293T early cell growth compared to untreated cells, although insufficient to affect conclusions drawn from these experiments. DH5 strain was grown to an OD of 0.6 with carbenicillin (100 g/ml). The pellet of 10-ml bacterial cell culture was resuspended in mammalian cell medium as described above without any antibiotics and anti-anti, and split equally among all conditions and their replicates. Auxotroph bacteria are added to HEK293T cells plated in pretreated 12-well plates in 2 ml of mammalian cell medium. The coculture is incubated overnight before the medium that contains the bacterial cells is removed. HEK293T cells were washed three times with 1x PBS (Thermo Fisher Scientific, catalog no. 10010023) and replenished with fresh media as conditions indicate. Media were replaced and added fresh to all conditions daily for 7 days. Imaging of cells was done with the inverted microscope Nikon Eclipse TS100 at days 2, 4, and 7 after initial coculture at 200 magnification.

Conditions:

Control: HEK293T grown in regular 10% FBS media with anti-anti and NEAA as described above.

DH5: HEK293T cells cocultured with this strain in mammalian cell media supplemented with carbenicillin (100 g/ml) to maintain plasmid during growth and absence of anti-anti.

DH5; anti-anti (antibiotic cocktail): HEK293T cells cocultured with this strain in mammalian cell media supplemented with carbenicillin (100 g/ml) to maintain plasmid during growth and presence of anti-anti cocktail.

DH5; anti-anti after day 2: HEK293T cells cocultured with this strain in mammalian cell media supplemented with carbenicillin (100 g/ml) to maintain plasmid during growth and absence of anti-anti cocktail. At 48 hours, anti-anti added and maintained to day 7.

DH5; anti-anti; no anti-anti after day 2: HEK293T cells cocultured with this strain in mammalian cell media supplemented with carbenicillin (100 g/ml) to maintain plasmid during growth and presence of anti-anti until day 2. After day 2, no anti-anti added and maintained to day 7.

DEP*: HEK293T cells cocultured with the biocontained strain in media supplemented with l-arabinose, chloramphenicol (25 g/ml), and carbenicillin (100 g/ml) to maintain bacteria and green marker. No bipA or anti-anti was added.

DEP*; bipA: HEK293T cells cocultured with the biocontained strain in media supplemented with l-arabinose, chloramphenicol (25 g/ml), and carbenicillin (100 g/ml) to maintain bacteria and green marker. One hundred micromolar bipA and no anti-anti added.

DEP*; bipA after day 2: HEK293T cells cocultured with the biocontained strain in media supplemented with l-arabinose, chloramphenicol (25 g/ml), and carbenicillin (100 g/ml) to maintain bacteria and green marker. No bipA or anti-anti added. At 48 hours, bipA at 100 M concentration added and maintained to day 7.

DEP*; anti-anti: HEK293T cells cocultured with the biocontaine
d strain in media supplemented with anti-anti, l-arabinose, chloramphenicol (25 g/ml), and carbenicillin (100 g/ml) to maintain bacteria and green marker. No bipA added.

DEP*; bipA; anti-anti: HEK293T cells cocultured with the biocontained strain in media supplemented with anti-anti, l-arabinose, chloramphenicol (25 g/ml), and carbenicillin (100 g/ml) to maintain bacteria and green marker. One hundred micromolar bipA added.

Persistence was evaluated by two kinds of assays: plate reader and colony count. For the plate reader case, DEP, DEP.e3, DEP.e4, and DEP.e5 cultures were grown overnight in permissible media conditions with 100 M BipA. For cells harvested at midexponential phase, the cultures were diluted 100 and grown to that state. Both stationary-phase and midexponential-phase cultures were then washed twice with LB media and resuspended in the original volume of nonpermissible media containing all specified media components except BipA. The resuspended cultures were then diluted 100 into nonpermissible media in triplicate for each time point to be tested. The specified concentration of BipA was then added back to those cultures at the specified time points. Typically, the BipA readdition occurred at 10 or 5 M concentrations and at hourly or daily intervals. The cultures were then incubated with shaking in SpectraMax i3 plate readers in a flat, clear-bottom 96-well plate with breathable and optically transparent seal for an upward of 84 hours at 34C. Approximately every 5 min, the OD600 was measured to determine cell growth kinetics.

More:
Synthetic auxotrophy remains stable after continuous evolution and in coculture with mammalian cells - Science Advances