Tuning and functionalization of logic gates for time resolved programming of bacterial populations

Abstract

In order to increase our command over genetically engineered bacterial populations in bioprocessing and therapy, synthetic regulatory circuitry needs to enable the temporal programming of a number of consecutive functional tasks without external interventions. In this context, we have engineered a genetic circuit encoding an autonomous but chemically tunable timer in Escherichia coli, based on the concept of a transcription factor cascade mediated by the cytoplasmic dilution of repressors. As proof-of-concept, we used this circuit to impose a time-resolved two-staged synthetic pathway composed of a production-followed-by-lysis program, via a single input. Moreover, via a recombinase step, this synchronous timer was further engineered into an asynchronous timer in which the generational distance of differentiating daughter cells spawning off from a stem-cell like mother cell becomes a predictable driver and proxy for timer dynamics. Using this asynchronous timer circuit, a temporally defined population heterogeneity can be programmed in bacterial populations.

Graphical Abstract

Open in new tabDownload slide

Introduction

Industrial bioprocessing depends on metabolically engineered microorganisms in bioreactors to (over) produce compounds that they would natively not (or insufficiently) produce (1–3), and is becoming an increasingly important cornerstone in sustainable development and health (4). Indeed, the recent revolution in our capacity to understand and engineer metabolic pathways has resulted in microbial cell factories that can produce a wide variety of biochemical products, including commodities for which no alternative synthesis routes exist (5–8).

However, next to engineering and implementing the metabolic pathways themselves, outstanding challenges for unlocking the full potential of bioprocessing include finding novel ways to increase our overall command of the behaviour of microbial bioreactor populations. As such, synthetic genetic circuitry has already been designed that overlays metabolic pathways with dynamic feedback regulation, whereby the accumulation or depletion of key metabolic intermediates in the cell serves as a trigger to selectively pause up- or downstream branches of the pathway (9,10). Also, quorum sensing dependent circuits have meanwhile been elaborated to properly time and synchronize expression of metabolic pathways throughout bioreactor populations (11,12).

An intriguing class of novel circuitry in this context would be synthetic regulatory systems that aim to autonomously impose a defined timing or succession of gene expression events, as it would consequently allow to spread tasks of bioreactor populations in time. A number of interesting circuits and theoretical models implementing temporal and oscillatory aspects in gene regulatory circuits have already been pioneered (13–18), although usually not designed for the time-resolved programming of gene expression and often accompanied with restrictions on culturing or growth. Nevertheless, a brief succession in gene expression was recently accomplished by Pinto et al. (16), employing cascades of accumulating orthogonal sigma factors, thereby delaying transcriptional responses for up to two doubling times in Escherichia coli.

Despite this progress, our means to more elaborately program bioreactor populations and subpopulations with predefined consecutive tasks over several generations is still rather limited. In this study, we therefore devised a timer circuit that is based on cascaded repressors in which the delay is imposed by the gradual dilution of the most downstream repressor. The resulting delay can be chemically tuned and used to autonomously separate a production and subsequent lysis phase in an Escherichia coli population. Moreover, this synchronous timer was further engineered to become deterministically asynchronous by imposing recombinase-mediated asymmetric segregation of the DNA fragment encoding the delay-imposing repressor in the cascade.

Materials and methods

Strains, media and cultivation

For all experiments in this study, Escherichia coli K12 MG1655 (further referred to as MG1655) and derivatives thereof have been used. All plasmids constructed in this study were maintained in Escherichia coli DH10B sAJM.1504 (19). Various strains from the Marionette Sensor Collection, a gift from Christopher Voigt (Addgene Kit #1 000 000 137), were utilized in this study as polymerase chain reaction (PCR) templates and/or intermediate DNA assembly hosts (19).

For cultivation either Lysogeny Broth (LB; 10 g/l tryptone, 5 g/l yeast extract, 5 g/l NaCl; with 1.5% agar added for agar plates), or AB medium [2 g/l (NH₄)₂SO₄, 7.5 g/l Na₂HPO₄× 2H₂O, 3 g/l KH₂PO₄, 3 g/l NaCl, supplemented with 0.1 mM CaCl₂, 1 mM MgCl₂, 0.003 mM FeCl₃, 10 mg/l thiamine, 25 mg/l uracil, 0.2% casamino acids and 0.4% glucose] was used. Where needed, the medium was supplemented with the following antibiotics or inducers at the indicated final concentrations: ampicillin (100 μg/ml), chloramphenicol (30 μg/ml), kanamycin (50 μg/ml), tetracycline (10 μg/ml), gentamicin (20 μg/ml), cuminic acid (Cuma, 100 μM), vanillic acid (Van, ≤ 100 μM) and anhydrotetracycline (aTc, ≤ 200 nM).

All strains were aerobically grown on an orbital shaker at 37°C with the exception for strains harboring the pKD46 (20) helper plasmid or pCP20 [encoding FLP recombinase (21)], which were grown aerobically at 30°C. For stationary-phase cultures, cells were grown for 16–18 h, while for exponential-phase cultures, stationary phase cells were diluted 1/1000 in fresh medium and grown for an additional 3–4 h. When cultures needed to be kept in continuous exponential phase, ca. 2/3 of the culture volume was replaced with fresh pre-warmed medium every hour. In the case of pulsed inductions, after 1 h of growth in the presence of inducer, cells were washed twice with pre-warmed, inducer free medium.

Plasmid and strain construction

All plasmids created in this study were constructed via Gibson assembly using homology overhangs of 20–40 bp. Gibson assembly master mix was made by 5-fold diluting 5 × ISO buffer (0.5 M Tris–HCl, pH 7.5, 50 mM MgCl₂, 1 mM gATP, 1 mM gTTP, 1 mM gGTP, 1 mM gCTP, 50 mM DTT, 5 mM NAD, and 25% PEG-8000) into MilliQ and supplementing with 5.3 U/ml T5-exonuclease (NEB), 33.3 U/ml Phusion DNA polymerase (NEB), and 5.3 U/μl Taq DNA ligase (ABclonal). In preparation for the assembly, DNA fragments were first PCR amplified using primers with respective homology overhangs at their 5′end. If needed, DPN1 digestion was performed in Tango buffer, using < 1 μl DPN1 restriction enzyme (10 U/μl, Thermo Scientific) and incubating for 1 h at 37°C, followed by heat inactivation (80°C for 20 min). The purified DNA fragments were mixed in a ca. 3:1 ratio (insert:vector), and 5 μl of this mix was added to 15 μl aliquots of Gibson assembly master mix thawed on ice. The resulting samples were incubated in a heat block (VWR) pre-heated to 50°C for 1 h, dialysed and used for electroporation.

For chromosomal replacements and insertions, Lambda Red recombineering was used (20). For the construction of MG1655 AT, the lac regulon was first replaced with a tetA–sacB cassette (22) in an MG1655 strain harboring the pKD46 plasmid (20). Next, using the tetA–sacB counter-selection protocol of Li et al. (22), the lac::tetA–sacB region was substituted with a P_Tet*-vanR^AM construct [based on parts obtained from Meyer et al. (19)]. Subsequently, this construct was expanded by transcriptionally fusing an mCerulean3-frt-Cm^R-frt cassette downstream of vanR^AM. Next, the frt-Cm^R-frt fragment was replaced by a cymR^AM-P_CymRC-T7RNAP-frt-Km^R-frt cassette [constructed based on pAJM.657 (19)], after which the region comprising P_CymRC-T7RNAP-frt-Km^R-frt was replaced by P_CymRC-tetR-mKate2-P_VanCC-sYFP2-frt-Cm^R-frt [based on parts obtained from Meyer et al. (19)], yielding MG1655 AT.

For the construction of MG1655 AAT, tetR of MG1655 AT was first replaced with a bxb1-Gm^R cassette flanked with Bxb1 irreversible recombination sites [based on parts obtained from Bonnet et al. (23)]. Next, its P_Tet*-vanR^AM–mCerulean3-frt-Cm^R-frt fragment was replaced by an frt-Km^R-frt cassette, after which this cassette was flipped with pCP20 (21). Finally, the Gm^R marker was replaced by the previously deleted P_Tet*-vanR^AM– mCerulean3-frt-Cm^R-frt fragment, resulting in MG1655 AAT.

Annotated sequences of the resulting engineered circuits in MG1655 AT, MG1655 AAT and pTApp have been deposited as GenBank files (cfr. Data Availability statement).

Lycopene production and quantification

Overnight cultures of MG1655 AT and its derivatives were diluted 1:500 into Erlenmeyer flasks containing 25–100 ml AB medium, and grown at 37°C on orbital shakers for 4 h. Cultures were subsequently normalized to an OD₆₀₀ of 0.2 using pre-warmed AB medium, and 50 ml of culture supplemented with appropriate inducers were further incubated in Erlenmeyer flasks at 37°C. All strains harboring the pTApp plasmid were continuously grown in the presence of 50 μg/ml kanamycin.

For lycopene extraction, 1 ml culture samples were centrifuged for 10 min at 4°C and 3 400 × g. The cells were then washed once with 0.085% NaCl and centrifuged again under the same conditions. Next, all supernatant was carefully removed followed by resuspension of the pellet in pure acetone (99.5%). After overnight storage at 4°C, heat extraction was performed for 20 min in a 55°C heat block shaking at 400 rpm, after which absorption was measured at 450 nm using a microplate reader (Multiscan FC, Thermo Scientific).

Microscopy

Time-lapse fluorescence microscopy experiments were performed by placing cells between AB agarose pads (AB medium supplemented with 1.5% agarose) and a cover glass using Gene Frames (Life Technologies), followed by incubation at 37°C using an Okolab cage incubator (Okolab, Ottaviano, Italy).

Cells were imaged using a Ti-Eclipse inverted microscope (Nikon, Champigny-sur-Marne, France), equipped with a 60 × Plan Apo λ oil objective, a TI-CT-E motorized condenser and a Nikon DS-Qi2 camera. Cyan fluorescent protein (CFP), yellow fluorescent protein (YFP), and red fluorescent protein (RFP) were imaged with triple dichroic (475/540/600 nm) emission filter, and a SpecraX LED illuminator (Lumencor, Beaverton, USA) as a light source.

Image analysis

Fluorescence microscopy images were analyzed using the DeLTA 2.0 image analysis pipeline (24) for cell segmentation and fluorescence feature extraction, after which the output was further processed using Python. As a first step, background subtraction was separately applied for each respective condition. Next, in order to accommodate for crowding effects in time-lapse fluorescence experiments, relative fold changes of fluorescence intensities were calculated by dividing the average cellular fluorescence value of an induced condition by the otherwise equally treated basal fluorescence level of the non-induced control. Therefore, values are given as multiples of the non-induced control basal level at the same time point. The number of objects identified by cell segmentation in each frame of the respective time-lapse image series was used to extrapolate the average doubling time of cells.

Statistical analysis

GraphPad Prism was used to determine the statistical significance. When appropriate, two-tailed, paired Student’s t-tests and one-way Analysis of variance with Tukey’s tests were applied. Significance intervals are given as such: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Best fit values were determined according to the Hill function with uncertainty represented as 95% prediction bands.

Results and discussion

Construction and characterization of an autonomous timer circuit

Since the gradual dilution of repressors has already been shown to impose an intrinsic (and often unwanted) delay in gene expression (25), we coined a synthetic genetic circuit based on a regulatory cascade of three repressors in Escherichia coli (Figure 1A). More specifically, we chose to employ the high dynamic range P_Tet*/TetR, P_CymRC/CymR^AM and P_VanCC/VanR^AM promoter/repressor pairs described by Meyer et al. (19) that can also be chemically derepressed (via anhydrotetracycline or aTc, cuminic acid or Cuma, and vanillic acid or Van, respectively). Using these pairs, the transcriptional units of the cascade were arranged such that in the ground state of the circuit (Figure 1A, Stage 0), the constitutively expressed CymR^AM repressor silences the P_CymRC promoter, which is responsible for the expression of tetR and mKate2 (referred to as rfp). Therefore, in the absence of the TetR repressor, vanR^AM and mCerulean3 (referred to as cfp) are actively expressed from the P_Tet* promoter. Consequently, in Stage 0, the VanR^AM silenced promoter (P_VanCC) will not express sYFP2 (referred to as yfp). The timer is activated by inducing P_CymRC with Cuma (Figure 1A, Stage 1), resulting in the production of RFP and TetR. The latter represses P_Tet*, thereby halting vanR^AM expression, and in turn causing the eventual de-repression of P_VanCC reported by YFP fluorescence (Figure 1A, Stage 2). Since the initial abundance of the gradually diluting VanR^AM determines the actual delay of P_VanCC derepression, we employed a translation-maximized vanR^AM ribosome binding site (RBS). Moreover, since leakiness of TetR expression could potentially compromise VanR^AM expression levels, tetR was fitted with an RBS of moderate strength and an SsrA-tag (expediting degradation of the resulting TetR protein). This autonomous timer circuit was chromosomally inserted as a single copy in the Escherichia coli MG1655 lac operon, yielding MG1655 AT. As such, MG1655 AT cells should only exhibit CFP fluorescence in the non-induced ground state and, upon induction with Cuma, first become red fluorescent, lose their CFP fluorescence and eventually become yellow fluorescent.

When microscopically assessed on agarose pads (Figure 1B), MG1655 AT showed no signs of leakiness and presented a delay time of ca. 3 h, as determined by the temporal separation between start of rfp expression (Stage 1) and yfp expression (progression to Stage 2; Figure 1B). Likewise, besides the immediate increase of RFP fluorescence, CFP fluorescence started to decline within the first hour of induction (Figure 1B). Therefore, P_Tet* was readily silenced by induction of TetR expression, indicating that the majority of the delay was imposed by the cytoplasmic dilution of VanR^AM. At the same time, during the first 3 h of timer induction barely any YFP signal was observed, while at ca. 3 h it readily increased. Based on calculations of the average generation time of cells (being ca. 38 min), this 3 h delay between Stage 1 and Stage 2 corresponds to around 4–5 divisions being required to sufficiently dilute out the VanR^AM repressor.

In order to more precisely assess the temporal dynamics of the individual circuit steps of MG1655 AT, we analysed the time-lapse fluorescence images for average singe cell fluorescence (Figure 1C). Considering that the maturation time (t₅₀) of our RFP (mKate2) is ca. 34 min (26), a significant fluorescence increase within the first 45 min suggests that Stage 1 is initiated readily upon induction. Stage 2 activity is being reported by the extremely fast maturing YFP [SYFP2, t₅₀ = 4 min; (26)] and can therefore be estimated with a negligible offset by direct fluorescence readout. At ca. 3 h of induction, the YFP fluorescence starts exceeding the basal level of the non-induced control by 10-fold, while at ca. 4 h the half maximum expression is reached compared to direct Van-induced Stage 2 derepression (which directly derepresses P_VanCC; cfr. Supplementary Figure S1). At the same time, supplementing pads with both Cuma and aTc prevented the (delayed) increase in YFP fluorescence as P_Tet*-VanR^AM expression would not be blocked by the Cuma induced P_CymRC-TetR (Supplementary Figure S1), proving that the delayed YFP increase in Cuma triggered MG1655 AT indeed stems from TetR silenced VanR^AM expression and the subsequent cytoplasmic dilution of VanR^AM. Therefore, the circuit cloned into MG1655 AT reliably imposes time resolved gene expression.

Functionalizing the autonomous timer circuit

Since our circuit performed as a robust timer, allowing for an autonomous and synchronous delay between two distinct stages, we proceeded with a functional proof-of-concept by imposing an MG1655 AT population with a time-resolved production-followed-by-lysis program via a single input (Figure 2A). In fact, bioprocessing often requires physical or chemical disruption of cells after the production phase to better recover metabolites or recombinant proteins, making production and lysis two meaningful consecutive actions for bioreactor populations (27). More specifically, a low copy number timer application plasmid (i.e. pTApp) was constructed encoding (i) a P_CymRC-crtEBI circuit, conferring a lycopene production module, that via its P_CymRC promoter is coupled to Stage 1 activation and (ii) a P_VanCC-M4Lys circuit, conferring a lysis module, that via its P_VanCC promoter is coupled to Stage 2 activation (Figure 2B). While the crtEBI operon stems from Erwinia herbicola and enables lycopene production in Escherichia coli from its natural farnesyl pyrophosphate intermediate (28), the M4Lys (or ORF_38) gene stems from the BSPM4 phage and enables lysis independently from the presence of holin or the Sec pathway (29).

As intended, when a batch culture of MG1655 AT equipped with pTApp was Cuma induced, it first started to accumulate lycopene for ca. 3 h (Stage 1), after which lysis and reduced viable cell count (Stage 2) autonomously set in (pTApp + Cuma in Figure 2C–E). In fact, almost complete lysis occurred at 4 h post Cuma triggering, suggesting that time delayed expression of plasmid-based functionalization genes follow the same temporal profile as inferred from the TLFM-assayed fluorescent reporters (Figure 1), despite the (ca. 10-fold) copy number difference of the pTApp plasmid [∼9 copies/cell; (30)] to the chromosomal YFP reporter. Therefore, the additional P_VanCC copies do not significantly titer VanR^AM away, and the autonomous timer progression can be used as a temporal queue for pTApp performance. After lysis, the amount of lycopene that could be recovered from cell mass (at t = 5 h; Figure 2D) was greatly reduced, indicating its release into the culture supernatant.

As a control, MG1655 AT cells were also locked in Stage 1 after Cuma induction by simultaneously providing aTc (pTApp + Cuma + aTc in Figure 2C–E). This indeed allowed lycopene production to continue (and lycopene levels to increase) past 3 h without being followed by lysis (as P_Tet* repression and subsequent P_VanCC derepression were abolished). Moreover, when such Stage 1-fixed MG1655 AT cells were forced to proceed to Stage 2 by providing Van after 3 h (pTApp + Cuma + aTc + Van in Figure 2C–E), they displayed virtually similar lysis and viability dynamics as the autonomously operating pTApp + Cuma population. These same dynamics could also be visually observed as indicated by representative pictures of the cultures after 5 h of incubation in flasks (Figure 2F). While all Cuma induced cultures had a characteristic pink colour, only the conditions allowing to progress to Stage 2 revealed massive cell lysis.

Since the sequentially expressed functionalization pathways are located on a small low copy number plasmid harbouring insulated low leakiness promoters from the Marionette Sensor Collection (19,30), the pTApp system is modular and orthogonal to the chromosomal timer. It therefore also allows for genes encoding other sets of consecutive tasks to be placed on pTApp without the risk of significantly altering timer performance. Serial expansion of the timer circuit would in turn enable a larger number of tasks to be autonomously executed in sequence, thereby enabling in vivo cascade catalysis approaches. In fact, such cascades require a strictly sequential enzyme expression to circumvent incompatible enzymatic reactions and reach the desired final product (31–33). Moreover, genetic circuitry for autonomously timed consecutive tasks could also provide new venues for microbial therapy, where the lack of proper access to therapeutic microbial populations within the body prevents the time resolved addition of inducers to impose consecutive tasks.

Chemical tuning of the autonomous timer circuit

To further establish the versatility of the autonomous timer circuit, we examined how easily the delay between Stage 1 and Stage 2 could be tuned. In fact, varying the background concentrations of aTc or Van should affect the time it takes to repress P_Tet* (directly correlated with aTc concentrations) or to derepress P_VanCC (inversely correlated with Van concentrations), respectively (Figure 3A).

As shown in Figure 3B, emergence of YFP (as a proxy of Stage 2 activation) can indeed be expedited in the presence of Van and delayed in the presence of aTc. In fact, while 1.6 or 3 μM of Van fails to significantly derepress P_VanCC when VanR^AM is constitutively expressed and thus replenished (cfr. Supplementary Figure S1), these concentrations do manage to rapidly attenuate VanR^AM activity during the cytoplasmic dilution of this repressor. Overall, the delay (measured as the time after Cuma triggering at which the average cellular YFP expression surpasses the basal level of the non-induced controls 10-fold) of the timer circuit could be tuned between 1.9 h (in the presence of 3 μM Van) and 4.4 h (in the presence of 60 nM aTc; Figure 3C). Since the doubling time of MG1655 AT remained stable at ca. 38 min throughout all tested conditions, it can also be stated that Stage 2 activation can be chemically tuned to occur anywhere within a range of 3–7 cell divisions.

Construction and validation of an asynchronous autonomous timer circuit

In order to further our abilities to impose a differential but still deterministic timing within bacterial populations, the autonomous timer circuit was adapted so that (i) tetR became replaced with a likewise SsrA-tagged bxb1 [encoding the Bxb1 irreversible recombinase with an expedited degradation rate; (23)] and (ii) the P_Tet*-vanR^AM-cfp fragment became flanked with Bxb1 recognition sites (attB/attP). The resulting circuit now allows for (pulsed) Cuma-triggering of Bxb1 activity that physically excises the P_Tet*-vanR^AM-cfp fragment (‘vanR^AM-fragment’) from the chromosome and turns it into a circular non-replicating plasmid that inevitably segregates asymmetrically upon cell division (Figure 4A). Upon each division, cells containing the excised P_Tet*-vanR^AM-cfp plasmid (referred to as vanR^AM carrier cells) will give rise to a daughter cell lacking the P_Tet*-vanR^AM-cfp circuit. Since these latter cells thus lack the genetic information to replenish VanR^AM, they will upon further proliferation experience the progressive dilution of cytoplasmically inherited VanR^AM and eventual de-repression of P_VanCC (reported by YFP; Figure 4B). As such, the generational distance of daughter cells from respective vanR^AM carrier cells (i.e. the time since spawning off from a cell expressing VanR^AM) now is the sole queue for delayed Stage 2 activation (Figure 4B). The resulting Escherichia coli MG1655 strain harbouring this asynchronous autonomous timer circuit is further referred to as MG1655 AAT.

When MG1655 AAT cells subsequently became triggered by a short Cuma pulse, they indeed developed into heterogeneous microcolonies that include cells retaining their bright CFP fluorescence over many generations because they asymmetrically inherit the vanR^AM-fragment (underscoring the stability of this episomal fragment), as well as their daughter cells that cytoplasmically dilute their CFP and VanR^AM content with every generation and in time become YFP fluorescent (Figure 4C and Supplementary Figure S2). Mapping the CFP/YFP content of cells therefore allows us to infer the following subpopulations: (i) vanR^AM carrier cells (Stage 1; displaying high CFP and low YFP), (ii) VanR^AM diluting cells (Stage 1.5, in between Stage 1 and 2; displaying moderate CFP and low YFP) and (iii) VanR^AM depleted cells (Stage 2; displaying low CFP and high YFP). Please note that the fraction displaying high CFP and low YFP also contains a number of (Stage 0) cells (ca. 5–10%) that serendipitously failed to Bxb1-excise the vanR^AM-fragment during pulsing, and (passively) remain in the population without partaking in the dynamics. But as shown recently, this fraction could further be suppressed by counter selecting against non-excised cells (34).

Based on this fluorescence gating, a more quantitative analysis of a Cuma-pulsed exponentially growing MG1655 AAT population could be performed (Figure 5). Since, upon Bxb1-mediated excision, the number of vanR^AM carrier cells remain constant within a population, the VanR^AM diluting daughter cells are starting to take over the population. In agreement with the synchronous timer (Figure 1C), there is a 2 h delay between the first cells experiencing VanR^AM dilution reported by decreased CFP fluorescence (as a proxy for entering Stage 1) on the one hand, and subsequently exceeding the basal YFP level of the non-induced control by 10-fold (as a proxy for entering Stage 2) on the other (Figure 5B).

These results demonstrate that even in a well-mixed environment the pulse-triggered asynchronous timer mediates the creation of temporally defined, and genetically distinct subpopulations over multiple hours. Although in this proof-of-concept the fully differentiated cells inevitably take over the population through growth, this aspect could be quenched or alleviated in case Stage 2 would be engineered to impose a more terminal differentiation such as the cell lysis imposed by the pTApp in the previous section or by selecting for the chloramphenicol resistance encoded on the vanR^AM fragment. In fact, terminal differentiation for bioprocessing has recently been demonstrated to be highly beneficial (35,36), as it prevents take-over of non-productive escape mutants when product formation is burdensome. In this context, our asynchronous timer would enable to continuously spawn off non-proliferating producer cells (e.g. having P_VanCC^ON driving production of a lethally toxic compound) from non-producing VanR^AM carrier cells. The latter would thereby continuously replenish the pool of producers without themselves experiencing any negative selection pressure from the production pathway. Furthermore, the defined creation of phenotypically diverse subpopulations will eventually enable to impose more complex division-of-labor strategies. Indeed, incompatible metabolic tasks could be allocated to different subgroups within a clonal bacterial population in order to functionally mimic the relevant multi-step microbial processes currently seen in microbial consortia (32,37,38).

Conclusion

Embedding functional genetic pathways and cargo within intricate regulatory schemes is becoming increasingly important in synthetic biology, as it allows to increase our command over the behaviour of engineered microbial populations. While there are many ways to control successions in gene expression with serial chemical or physical inducers, such inducers are often expensive (39) and/or not always able to properly reach or accumulate at the targeted population when it operates in complex structured environments [such as soil or the human body; (40,41)]. In this report, we demonstrate that the intrinsic temporal dynamics of repressor logic gates can be readily engineered and exploited for the autonomous relay of time-resolved instructions to bacterial populations. The elaborated timer-tools allow to synchronously spread tasks over time, and even to asynchronously divide tasks between clonal siblings in a predictable fashion.

In fact, ways to program population heterogeneity in a deterministic and thus predictable manner are scarce. Irreversible recombinases have often been used to invert the orientation of functional genetic elements (17,42–44), or -more recently- to terminate bacterial chromosome replication by excising the origin of replication (36). However, the apparent stability and inherent asymmetric segregation of a functionalized (i.e. providing the VanR^AM repressor) Bxb1-excised fragment proves to yield a deterministic motor for cellular differentiation.

In summary, the timer tools designed and characterized in this study provide new handles in programming complex autonomous behaviours in engineered microbial populations that can be applied in bioprocessing, bioremediation and microbial therapy.

Data availability

The data underlying this article are available in Zenodo, at https://dx.doi.org/10.5281/zenodo.11258520. DNA sequences of constructs described in this study have been deposited as GenBank files. The circuit in MG1655 AT and MG1655 AAT as well as the plasmid pTApp can be found in the respectively named files: ‘at-circuit.gb’, ‘aat-circuit.gb ’ and ‘pTApp.gb’. The strains presented in this article will be shared on reasonable request to the corresponding author.

Supplementary data

Supplementary Data are available at NAR Online.

Acknowledgements

The authors would like to thank Kristel Bernaerts and Steffen Waldherr for fruitful discussions.

Conceptualization, L.E.B. and A.A. Methodology, L.E.B. and S.K. Investigation, L.E.B., K.B. and L.W. Formal analysis, L.E.B., K.B. and A.A. Writing—original draft preparation, L.E.B. and A.A. Writing—review & editing, L.E.B. and A.A. Supervision, A.A. Funding acquisition, A.A.

Funding

Fonds Wetenschappelijk Onderzoek [11H0323N to L.E.B., 11Q4T24N to K.B., G0C5322N]; KU Leuven Research Fund [C14/20/087]. Funding for open access charge: Fonds Wetenschappelijk Onderzoek.

Conflict of interest statement. None declared.

References