Highly variable individual donor cell fates characterize robust horizontal gene transfer of an integrative and conjugative element

Significance

DNA conjugation is a fascinating process that bestows the prokaryotic world with the ability to exchange hundreds of genes in a single event. What has long been puzzling is the low rate of horizontal gene transfer at the population level. Here, we develop a fluorescent tool to study the dynamic fate of a chromosomally located conjugative DNA. We find that donor cells becoming transfer proficient undergo highly variable cell fates, likely as a result of the conjugative element wreaking havoc in the cell. Modeling of the partnership between the cell and integrative and conjugative element (ICE) suggests that ICE fitness is optimal at low activation rates, explaining why gene transfer in bacterial populations occurs at low frequencies despite its ecological importance for adaptation.

Abstract

Horizontal gene transfer is an important evolutionary mechanism for bacterial adaptation. However, given the typical low transfer frequencies in a bacterial population, little is known about the fate and interplay of donor cells and the mobilized DNA during transfer. Here we study transfer of an integrative and conjugative element (ICE) among individual live bacterial cells. ICEs are widely distributed mobile DNA elements that are different than plasmids because they reside silent in the host chromosome and are maintained through vertical descent. Occasionally, ICEs become active, excise, and transmit their DNA to a new recipient, where it is reintegrated. We develop a fluorescent tool to differentiate excision, transfer, and reintegration of a model ICE named ICEclc (for carrying the clc genes for chlorocatechol metabolism) among single Pseudomonas cells by using time-lapse microscopy. We find that ICEclc activation is initiated in stationary phase cells, but excision and transfer predominantly occur only when such cells have been presented with new nutrients. Donors with activated ICE develop a number of different states, characterized by reduced cell division rates or growth arrest, persistence, or lysis, concomitant with ICE excision, and likely, ICE loss or replication. The donor cell state transitions can be described by using a stochastic model, which predicts that ICE fitness is optimal at low initiation rates in stationary phase. Despite highly variable donor cell fates, ICE transfer is remarkably robust overall, with 75% success after excision. Our results help to better understand ICE behavior and shed a new light on bacterial cellular differentiation during horizontal gene transfer.

Bacterial genomes represent dynamic constellations of core and accessory genomic regions (1–3). The latter are dominated by mobile elements; DNA that can be transferred horizontally from a donor to a recipient bacterium, even of a different species. Integrative and conjugative elements (ICEs) are frequently detected mobile genome inhabitants. They can contribute to host adaptation by dispersing dozens to hundreds of genes in a single transfer event (4–6). It has been estimated that conjugative systems of ICEs are more abundant among bacteria than those of plasmids (7), yet we know far less about ICE behavior because they are difficult to follow and isolate. ICEs have attracted interest because they can transmit antibiotic resistance genes (8) and suspected virulence factors (9–11), but also genes for toxic compound degradation (12, 13) and heavy metal resistance (14). Several evolutionary distinct families of ICEs have been described, which have a mechanistically similar lifestyle (4, 6, 15). In contrast to conjugative plasmids, ICEs invade cells through conjugation but subsequently insert at one or more specific sites in the genome, from where they need to excise for a next round of transfer (4–6). Similar to a prophage, the ICE is coreplicated with the bacterial chromosome in its integrated form, ensuring stable vertical transmission. Conjugation frequencies of ICEs vary widely from approximately 1 × 10⁻² to 1 × 10⁻⁷ per donor cell, which is considered to be the outcome of an infrequent bistable switch triggering activation of the ICE (16, 17). Environmental cues eliciting the “SOS response” (18) or growth on specific carbon substrates (19) can enhance ICE transfer rates, likely by influencing the frequency of the bistable switch.

Horizontal transmission of ICEs starts by their excision from the host chromosome through site-specific recombination and subsequent processing of the excised circularized DNA for conjugation, similar to conjugative plasmids (20, 21) (Fig. 1A). In the receiving cell, the ICE must ensure effective integration to avoid being lost upon cell division because of its limited self-replication capacity (22, 23) (Fig. 1B). We showed for the model ICEclc (named for carrying the clc genes for chlorocatechol metabolism) in the bacterium Pseudomonas knackmussii B13 (13, 24) that conjugative transfer necessitates development of the host bacterial cell into a “transfer competence” state (25). Development of transfer competence is orchestrated by ICEclc and results in a bistable differentiation of transfer competent (tc) cells and non-tc cells (in which ICEclc remains silent; Fig. 1C) (25). Initiation of transfer competence occurs in stationary phase, when the bacteria in culture run out of nutrients and stop dividing, but affects only 3–5% of the cells in a clonal population harboring WT ICEclc (17, 19, 26). This proportion can increase to 50% or more in strains with a deletion of the ICEclc gene mfsR (major facilitator superfamily regulator) (27). Initiation is assumed to trigger a cascade of events (17, 27), leading to the synthesis of the ICEclc conjugation proteins (28, 29) and, finally, to activation of the P_int promoter driving intB13 integrase expression and causing ICEclc excision (Fig. 1A) (17, 27, 30). Intriguingly, tc cells do not only specifically commit to ICE transfer, but their further cell division becomes compromised by factors encoded on ICEclc (25). This limited cell division leads to the formation of small groups of tc donor cells, which are thought to benefit the overall transfer success of ICEclc by increasing the chance to contact recipients (31).

Fig. 1.

Life cycle of the ICEclc element in Pseudomonas. (A) ICEclc is normally integrated in the host’s chromosome at the 3′ end of a gene for tRNA^Gly (white-green box). Excision occurs through site-specific recombination between two 18-bp sequences (purple boxes) (39). Note how the stronger P_circ promoter faces outward from ICEclc in the integrated form but is placed upstream of the weaker P_int promoter in front of the intB13 integrase gene in the excised form (30). The developed ICEclc reporter has an egfp gene located in the same transcriptional unit as intB13. (B) Upon transfer to a recipient cell, ICEclc integrates through site-specific recombination in the reverse process of A. Shown here is the constructed recipient with the conditional trap, consisting of a promoterless mcherry gene downstream of the integration site. Upon integration, mcherry will be strongly transcribed from the P_circ promoter. (C) ICEclc activation is a bistable process arising in tc cells. Tc cells are characterized by activation of the P_int promoter, which can be visualized through EGFP fluorescence. Cells with excised ICEclc are expected to have stronger EGFP expression as a result of the P_circ promoter. Incoming ICEclc-intB13-egfp in the recipient may lead to temporary green fluorescence for as long as the ICE is in its excised form. When it has been integrated into the trap, the cell will appear orange because of mCherry expression (plus the remaining green). (D) Transfer competence is initiated in stationary phase (stat) cultures but only in a small proportion of cells. Some individuals transferred from flask to nutrient surface for time-lapse microscopy (t = 0) will be tc starters; others will be non-tc cells. Upon renewed cell growth (expo), non-tc cells divide normally to microcolonies, and, in stationary phase (stat), a proportion of cells again starts the tc development (tc formers). tc starter cells divide abnormally and cease growth earlier than non-tc cells. (E) Starting condition of a tc starter cell (tcs) and non-tc cells of P. putida ICEclc-intB13-egfp,ΔmfsR 1 h after inoculation on nutrient surface (EXPO), and the same area with tc formers (tcf) in stationary-phase (STAT) microcolonies after 28 h.

With few exceptions (25, 32), most of the current knowledge on ICE behavior derives from population-level genetic studies. Very little is known about the cellular events during transfer, which, given the low frequencies of ICE conjugation, are challenging to study. Of particular interest is what ICE transfer provokes in a donor cell and how or whether ICE can maximize successful transfer. Our starting hypothesis based on previous results (25) was that tc donor cells, when they have been initiated, would efficiently deliver the ICEclc at a high rate of successful transfer. Single-cell studies have been previously used to follow plasmid conjugation in which case donor cell fates have not been particularly addressed (33–35), but because ICEs behave very differently, those results are not immediately adaptable. Hence, to better study and understand donor cell fates, we develop a fluorescence reporter that distinguishes the four key steps of ICE transmission: activation, excision, actual transfer, and integration into new recipients. The reporter is calibrated on donor populations alone and in real-time transfer experiments between donors and recipients, and compared with quantitative PCR (qPCR) measurements of ICEclc excision. In contrast to our expectations, we find that donor cells display highly variable phenotypes characterized by dynamic ICEclc behavior, which can be explained by stochastic modeling of different cell state transitions. Despite the variability in cell fates, horizontal transmission is successful in 75% of donor cells in which ICEclc excises. Further work will be needed to understand the molecular events underlying the emergence of the different cell fates.

Results

Development of a Tool to Visualize ICEclc Excision and Transfer from Single Donor Cells.

To better understand the dynamic fate of ICEclc during transfer, we developed a tool to follow experimentally single donor cell fates in absence or presence of recipients. The tool consists of an ICEclc reporter variant in which egfp is placed directly downstream of the intB13 integrase gene in the same bicistronic transcription unit (Fig. 1A, Table S1, and Fig. S1). Initiation of transcription of the integrase gene is a key feature of donor cells in which ICE becomes active (25) and would be visible as single cells producing EGFP above background levels (Fig. S2) (36). However, upon excision and ICEclc-DNA circularization, a stronger constitutively transcribed promoter (P_circ) is placed upstream of intB13 (30), which we expected would be reflected in higher EGFP production (Fig. 1A). Pseudomonas putida carrying a single integrated copy of such ICEclc-intB13-egfp (strain 4611) expressed EGFP in 5.9 ± 0.4% of individual cells after 72 h in stationary phase (Fig. S3A). This is consistent with previously measured proportions of initiated tc cells in stationary-phase cultures of P. putida ICEclc and P. knackmussii B13, estimated from EGFP expression of single-copy P_int-egfp insertions located outside ICEclc (17, 37). In P. putida carrying the same ICEclc-intB13-egfp bicistronic fusion but with an additional deletion of the mfsR global negative regulator [strain 4612 (27)], 45 ± 5% of individual cells expressed EGFP after 72 h in stationary phase (Fig. S3A). The ICEclc variant with the mfsR deletion is useful for studying single-cell transfer, as it causes a higher transfer frequency and increases the likelihood of observing such events in live-cell time-lapse microscopy. Strains carrying an additional randomly inserted mcherry reporter construct controlled by the P_inR promoter (strain 4673), which is also exclusively transcribed in tc cells (17, 26), coexpressed EGFP and mCherry in the same cell (Fig. S3B), confirming that those cells initiate the ICEclc transfer competence process.

Table S1.

Strains used in this study

No.	Strain	Characteristic	Source
E. coli
3044	DH5αλpir	sup E44, ΔlacU169 (ΦlacZΔM15), recA1, endA1, hsdR17, thi-1, gyrA96, relA1, λpir phage lysogen	V. de Lorenzo
P. putida
1291	UWC1	Plasmid-free derivative of P. putida KT2440, RifR	Ref. 49
2737	UWC1-clc5	Derivative of strain 1291 with one ICEclc copy integrated into tRNAgly-5	Ref. 40
4322	UWC1-clc5 ΔmfsR	Derivative of strain 2737 with an internal deletion in mfsR	Ref. 37
4611	UWC1-clc5 ICEclc-intB13-gfp	Derivative of strain 2737 with a promoterless KmR gene inserted 27 bp downstream of intB13	This study
4612	UWC1-clc5 ICEclc-intB13-gfp, ∆mfsR	Derivative of strain 4322 with a promoterless egfp gene inserted 27 bp downstream of intB13	This study
4669–4672	UWC1-clc5 ICEclc-intB13-gfp, Tn5 PinR-echerry	Derivatives of strain 4611 with random chromosomal insertion of miniTn5 PinR-echerry. Four independent clones, KmR	This study
4673–4676	UWC1-clc5 ICEclc-intB13-gfp, ∆mfsR, Tn5 PinR-echerry	Derivatives of strain 4612 with random chromosomal insertion of miniTn5 PinR-echerry. Four independent clones, KmR	This study
4754	UWC1-clc5 ICEclc-intB13-gfp, ∆attL	Derivative of 4611 with 18 bp attL deletion	This study
4755	UWC1-clc5 ICEclc-intB13-gfp, ∆mfsR, ∆attL	Derivative of 4612 with 18 bp attL deletion	This study
2744	UWC1GC	Derivative of strain 1291 with single copy miniTn7 Ptac-echerry, GmR	Ref. 29
2756	UWC1 miniTn5 JimX-gfp	Derivative of UWC1 with miniTn5 insertion of the jimX-gfp fragment, KmR	Ref. 40
5001–5003	UWC1 miniTn5 JimX-echerry	Derivatives of UWC1 with miniTn5 insertion of the jimX-echerry fragment, KmR	This study

Fig. S1.

Detail of the egfp transcription fusion to the intB13 integrase gene of ICEclc. (A) Localization of intB13 nearby the attR end of ICEclc. Approximate positioning of a promoterless egfp downstream and in between intB13 and orf2848. GenBank accession number of ICEclc: AJ617740.2. (B) Sequence detail of the end of the intB13 reading frame (red), the 68-bp spacer, and the start of the egfp gene. RBS, ribosome binding site.

Fig. S2.

Subpopulation determination of transfer competent cells within the population of P. putida cells carrying ICEclc-intB13-egfp,∆mfsR. Cells ranked according to their average EGFP fluorescence as a function of the expected normal distribution (theoretical quantiles). The program calculates the normal distribution (green line) on the basis of the fraction of cells with the lowest EGFP values plus the 95% CI (blue dotted lines). Cells with values above the upper 95% confidence boundary are considered as belonging to a second subpopulation with higher EGFP expression. The slope of the line through this second population would again be normally distributed. Calculation was performed according to Reinhard and van der Meer (31).

Fig. S3.

Transfer competence initiation in P. putida carrying ICEclc-intB13-egfp. (A) Proportion of tc cells in suspended cultures in stationary phase after 24, 48, and 72 h for ICEclc-intB13-egfp (WT) and ICEclc-intB13-egfp without the mfsR global regulator (27). Detection of tc cell subpopulations using the quantile-quantile normalization procedure (Fig. S2) (31). Bars are averages from biological triplicates. Error bars represent calculated SDs. (B) Colocalization of EGFP and mCherry fluorescence produced from a randomly inserted P_inR promoter, which is characteristic of the tc state (25).

To distinguish potentially different states of ICEclc in donor cells (i.e., initiation of intB13 expression in the integrated state and ICEclc excision), we followed EGFP production over time by live-cell time-lapse microscopy. Individual cells of P. putida carrying WT ICEclc-intB13-egfp or ICEclc-intB13-egfp with the ∆mfsR mutation were placed on solid agarose surfaces with nutrients in closed microscope chambers (38) and imaged during exponential growth and subsequent stationary phase. Importantly, because cells inoculated on the surface come from liquid precultures that were grown to stationary phase, some of them already initiated the ICEclc transfer competence process (hereafter named tc starters; Fig. 1D). Individual cells visible on the nutrient surface at the start of the experiment can therefore be a tc starter or a non-tc cell, which can be distinguished on the basis of their EGFP fluorescence level (Fig. 1 D and E, EXPO, and Fig. S2). Dividing non-tc cells reach new stationary phase on the surface after 9–10 h, during which expression of EGFP from the intB13-egfp fusion increases again in some cells, indicating initiation of the transfer competence process (Fig. 1 D and E, STAT). In the following, we will name such cells tc forming cells (or tc formers) to distinguish them from tc starters and their daughters (Fig. 1D). tc starters divide not at all (Fig. 1E) or only a few times, even in the presence of new nutrients (as discussed later).

EGFP production rates in individual cells of P. putida ICEclc-intB13-egfp,∆mfsR (strain 4673), measured from the slope of EGFP fluorescence increase over time for at least 150 min in individual cells (Fig. 2A, Inset), were statistically significantly higher in tc starter cells and their daughters than in tc formers or non-tc cells (P < 0.0001, multiple-comparison one-way ANOVA, Tukey test; Fig. 2B). Also, EGFP fluorescence maxima observed during the lifetime of individual cells were statistically significantly higher in tc starters and daughters than in tc formers (P < 0.0001; Fig. S4). Visibly, EGFP production rates were also more variable in tc starter cells than in tc formers (Fig. 2A and Fig. S5A). No significant differences in EGFP production rates could be detected in tc starter cells of P. putida with the ∆mfsR mutation compared with the WT ICEclc-intB13-egfp fusion (strain 4611; P = 0.9898; Fig. 2B). This indicates that, whereas the mfsR deletion strongly increases the proportion of tc cells in a population, it does not change the level of ICEclc activation in individual cells compared with WT ICEclc. The difference in EGFP production rates between non-tc, tc starter, and tc former cells was thus in agreement with our hypothesis that the reporter could detect different ICEclc states in donor cells.

Fig. 2.

ICEclc excision and transfer from tc starter cells. (A) EGFP fluorescence development in individual tc starters and daughters (tcs; gray), tc formers (tcf; blue) and non-tc cells (red; Right) of P. putida ICEclc-intB13-egfp,ΔmfsR,P_inR-mcherry (strain 4673). (Inset) How the slope of EGFP fluorescence increases during at least 150 min (dashed line) is operationally used to define tc starters with ICEclc excision (>1.65 EGFP U⋅min⁻¹; example yellow trace) or without (example magenta trace). Traces in tc formers are aligned to the point of last cell division (t = 0). Further individual tc starter traces are shown in Fig. S5. (B) Box-plot distributions of EGFP production rates among non-tc cells, tc former, and tc starter cells of P. putida ICEclc-intB13-egfp (WT, strain 4611), P. putida ICEclcintB13-egfp,ΔmfsR,P_inR-mcherry (strain 4673), and P. putida ICEclc-intB13-egfp,ΔmfsR,ΔattL (strain 4755) in absence of recipient. TRA⁺ and TRA⁻, tc starters of P. putida ICEclc-intB13-egfp,ΔmfsR (strain 4612) in the presence of recipient, with confirmed ICEclc transfer or without transfer, respectively. Green asterisk indicates tc formers in presence of recipient with confirmed ICEclc transfer. Dashed line indicates the threshold above which cells are considered to have an excised ICEclc. Letters above box plots indicate statistically significant similarity among groups by multiple-comparison one-way ANOVA, Tukey test. N, number of cells analyzed. Whiskers represent the minimal and maximal values. (C) ICEclc transfer from tcs donors (d) of P. putida ICEclc-intB13-egfp,ΔmfsR (strain 4612) to P. putida P_tac-mcherry, resulting in two transconjugants (t) receiving ICEclc. Overlay image combined from individual images recording EGFP (green) and mCherry (red). Schematic shows expected color transitions. (D) Single tc donor double transconjugant event. (E) Single tc donor triple transconjugant event. (F) Single-donor ICEclc transfer and integration into P. putida with the conditional attR-mcherry trap (r). PhC, phase-contrast. (G) Stimulation of ICEclc excision in tc starters by addition of 0.1 mM 3-CBA on silica-gel patches (+3-CBA) or without (−). P = 0.0015, two-tailed t test, equal variance.

Fig. S4.

EGFP fluorescence maxima in cells of P. putida carrying ICEclc derivatives. Graph shows box-plot distributions of EGFP production rates among non-tc cells, tc former (tcf), and tc starter (tcs) cells of P. putida ICEclc-intB13-egfp (WT; strain 4611), P. putida ICEclc-intB13-egfp,∆mfsR,P_inR-mcherry (strain 4673), and P. putida ICEclc-intB13-egfp,∆mfsR,∆attL (strain 4755) in absence of recipient. TRA⁺ and TRA⁻, tc starters of P. putida ICEclc-intB13-egfp,∆mfsR (strain 4612) in presence of recipient, with confirmed ICEclc transfer or without transfer, respectively. Letters above box plots indicate statistically significant similarity groups in multiple-comparison one-way ANOVA, Tukey test. N, number of cells analyzed. Whiskers represent the minimal and maximal values.

Fig. S5.

EGFP fluorescence traces over time of tc starter cells in absence or presence of recipients. (A) EGFP fluorescence traces over time of only tc starter cells (colored trace) and their daughters (gray traces) in microcolonies, deduced from dividing cells on live time-lapse microscopy. Note the large variability in EGFP expression (absolute levels and rates of increase or decrease). (B) EGFP fluorescence in tc starter cell donors (green traces) and recipients (blue, from incoming ICEclc-intB13-egfp). Note the double and triple transfers (multiple blue traces). Asterisk denotes a case in which EGFP fluorescence continues to increase in the donor cell even though it has already started in the recipient. This suggests ICEclc replication in the donor. Note that fluorescence scales (y axis) may be different for individual panels.

ICEclc Excision Occurs in Donor Cells Only After Renewed Nutrient Addition.

To verify whether higher EGFP expression in tc starters was indeed a consequence of ICEclc excision, we compared with a P. putida strain with ICEclc in which the recombination sequence attL (Fig. 1A) was deleted (strain 4799). A ∆attL mutant cannot recombine between attR and attL and, therefore, ICEclc excision and transfer are strongly impaired (25) (Table 1). As a consequence, IntB13 and EGFP cannot be expressed from P_circ but only from P_int (Fig. 1A). EGFP production rates and fluorescence maxima from tc starters of the ∆attL mutant were indeed significantly lower than those of P. putida tc starters with intact attL (P < 0.0001, multiple-comparison one-way ANOVA, Tukey test; Fig. 2B and Fig. S4).

Table 1.

ICEclc transfer between P. putida donor and recipients

Donor*	Transfer rate†	Dividing tcs						Nondividing tcs						tcf		Total no. traco
		No. at start	With contact to recipient				No. tcm without contact	No. at start	With contact to recipient				Without contact	No. of contacts‡	No. traco
			No. of contacts‡	No. tcm (cells) with transfer	No. tcm (cells) without transfer	No. traco			No.of contacts‡	No. with transfer	No. w/o transfer	No. traco
P. putida (4612)	12.8 ± 4.6 × 10−2	99	294	28 (93)	19 (73)	43	52	75	34	10	34	14	31	316	3	60
P. putida (4755)	2 ± 2 × 10−4	44	39	0	11 (30)	0	33	41	51	0	14	0	27	28	0	0

Data pooled from independent triplicate live-cell microscopy transfer experiments, each following three to five randomly located areas of observation on the surface. tcf, transfer competent forming cells; tcm, microcolonies formed from tcs; tcs, transfer competent starters; traco, transconjugants (i.e., ICEclc in recipient). Number of cases with one transconjugant from one donor, 52 (87%); two transconjugants from a single donor, 2–4 (7%); and three transconjugants from a single donor, 1 (5%).

^*P. putida (4612) is ICEclc-intB13-egfp,∆mfsR; P. putida (4755) is ICEclc-intB13-egfp,∆mfsR,∆attL. Recipient is P. putida UWC1 (2744).

^†Number of colonies on plates with 3-CBA as sole carbon source plus gentamicin divided by the number of colonies on plates with 3-CBA only.

^‡Multiple contacts per donor allowed.

To further confirm that ICEclc excision is more prevalent among tc starters than among tc formers, we measured by qPCR the ratio of covalently closed excised ICEclc DNA compared with the ICEclc-located clcB gene in stationary-phase cultures of P. putida ICEclc-intB13-egfp,∆mfsR and in freshly diluted cultures after 4 h growth with 0.1 mM 3-chlorobenzoate [3-CBA; a selective carbon substrate for growth of cells carrying ICEclc (39)]. If the hypothesis from live-cell microscopy observations is correct, tc cells in stationary-phase cultures should have proportionally low ratios of excised ICEclc DNA (i.e., tc formers), but addition of new nutrients would result in stimulation of excision of ICEclc DNA (i.e., tc starters). Indeed, the measured ratio of amplified ICEclc circular junction compared with amplified clcB fragment was, on average, 1.49 times higher (SD = 0.35; triplicate independent cultures) in exponential than in stationary-phase cultures. This difference seems small but is in agreement with a predicted calculated ratio when considering the difference in growth rate between tc starters and non-tc cells and the average measured number of generations before tc starters stop dividing (Table S2 and Fig. S6). As expected, we were unable to detect by PCR any circular form of ICEclc in stationary phase or in exponential cultures of P. putida carrying ICEclc with the ∆attL mutation.

Table S2.

Model estimation of abundance of ICEclc excised form in early exponential-phase cultures

Time, min	No. of tc cells*	Generations tc cells†	No. of non-tc cells	Generations non-tc cells	Proportion tc/non-tc	No. of excised ICE‡	No. of total clcB	Ratio exc ICE/clcB	Enrichment§
0.00	50.00	—	50.00	0.00	0.50	3.00	100.00	0.03	1.00
30.00	59.38	0.25	63.68	0.35	0.48	32.69	123.05	0.27	8.85
60.00	70.51	0.50	81.09	0.70	0.47	38.25	151.60	0.25	8.41
90.00	83.73	0.74	103.28	1.05	0.45	44.86	187.01	0.24	8.00
120.00	99.43	0.99	131.53	1.40	0.43	52.71	230.95	0.23	7.61
150.00	118.07	1.24	167.50	1.74	0.41	62.04	285.58	0.22	7.24
180.00	140.21	1.49	213.32	2.09	0.40	73.11	353.53	0.21	6.89
210.00	166.50	1.74	271.67	2.44	0.38	86.25	438.17	0.20	6.56
240.00	197.72	1.98	345.98	2.79	0.36	101.86	543.70	0.19	6.24
270.00	234.80	2.23	440.62	3.14	0.35	120.40	675.42	0.18	5.94
300.00	278.82	2.48	561.14	3.49	0.33	142.41	839.97	0.17	5.65
330.00	200.00	—	714.64	3.84	0.22	103.00	914.64	0.11	3.75
360.00	200.00	—	910.11	4.19	0.18	103.00	1110.11	0.09	3.09
390.00	150.00	—	1159.06	4.53	0.11	78.00	1309.06	0.06	1.99
420.00	150.00	—	1476.10	4.88	0.09	78.00	1626.10	0.05	1.60
450.00	100.00	—	1879.87	5.23	0.05	53.00	1979.87	0.03	0.89

^*n(t) = 50*2^(t − t₀/t_gen)). Cell division time (t_gen): tc starters, 121 min; non-tc cells, 86 min.

^†Cell division arrest of tc starter cells after 2.5 generations.

^‡Assuming excision in 50% of tc starters once dividing.

^§Enrichment calculated as (ratio exc ICE/clcB)_t/(ratio exc ICE/clcB)_t0, to be conform to qPCR results.

Fig. S6.

Probability distribution of the different model parameters, derived from experimental data. (A and B) Distribution of cell divisions for tc and non-tc cells, respectively, based on experimental data. (C) The time at which a cell divides sampled from the Gaussian distribution, again depending on their cell type. The dots and bars show mean and SD of each distribution. (D–F) Probability for lysis times of different cell types at different phases. (G) Time of excision of tc cells since their birth. (H) Time of differentiation of non-tc cells into tc cells since the beginning of the stationary phase. (I) Probability for each of the aforementioned processes for a given cell. For example, with a probability of 0.53, a tc will be destined to lyse during growth phase. The time at which this lysis occurs is determined by randomly sampling from the distribution of the green data in D with some added noise. Each plot shows experimental data values in green against the randomly generated values for the simulation in orange. Simulation values were chosen by randomly sampling from the experimental data, with added Gaussian noise (mean ± SD, 0 ± 0.5 for A and B and 0 ± 60 for D–I). These values are assigned at the beginning of the simulated growth or stationary phase.

Finally, we followed ICEclc transfer from individual P. putida donors mixed with ICEclc-free recipient cells, expecting that transfer would occur only from donors in which the ICE excises. Transfer can be followed at the single-cell level, because the egfp gene is inserted downstream of intB13 on ICEclc, and, upon its entry into a new recipient (temporary), egfp expression may occur (as illustrated in Fig. 1C). As recipients, we used derivatives of P. putida UWC1 constitutively expressing mcherry (29) to distinguish them from donor cells or equipped with a conditionally fluorescent “trap” (lighting up after ICEclc insertion) (40). Recipient cells producing EGFP were observed exclusively when they had been in direct contact with tc starters or their daughters (57 transfers among 328 observed donor–recipient contacts with P. putida ICEclc-intB13-egfp,∆mfsR in triplicate independent experiments; Table 1), but never with non-tc cells (n = 202 observed contacts with recipient cells). Three recipients with increased EGFP fluorescence occurred next to tc formers (n = 316 contacts; Table 1). In contrast, no EGFP-producing recipient cells were detected in matings with P. putida donors carrying the ICEclc ∆attL mutant (n = 90 tc starter contacts; Table 1). The EGFP production rates of tc starters with confirmed ICEclc transfer (Fig. 2B, TRA⁺) were statistically indistinguishable from those of tc starters in donor populations alone, but higher than of tc starters without observed transfer (P = 5.6 × 10⁻⁵, multiple-comparison one-way ANOVA, Tukey test; Fig. 2B, TRA⁻). Also, the rate of EGFP production of the three tc formers with confirmed ICEclc transfer fell within the distribution of TRA⁺-tc starters (Fig. 2B, green asterisk). This confirmed that tc starters with the highest EGFP expression rates are the ones in which ICEclc excises.

Collectively, these results indicate that, although ICEclc transfer competence is initiated in cells in stationary phase (i.e., tc formers), the element predominantly excises when such cells receive fresh nutrients again (i.e., tc starters on the agarose surface or in diluted liquid batch cultures with new 3-CBA). To corroborate this more specifically, we repeated experiments with P. putida ICEclc-intB13-egfp,∆mfsR cells inoculated on silica-gel surfaces in presence or absence of 0.1 mM 3-CBA. Silica gel, in contrast to agarose, contains no further traces of carbon substrate. Although cells divided more slowly on silica gel than on agarose surface, EGFP production rates among tc starters were significantly higher on silica gel with 3-CBA than without, suggesting that tc starter cells more frequently excise ICEclc in the presence of new nutrients (Fig. 2G).

ICEclc Transfer Competent Donors Display Variable Cell Fates.

Given the high variability of EGFP production rates of tc starter cells of P. putida carrying WT ICEclc-intB13-egfp or its variant lacking ∆mfsR (Fig. 2 A and B and Fig. S5A), we wondered whether all tc starters actually follow the same cell fate. From EGFP production rates in tc starters with confirmed ICE transfer (Fig. 2B, TRA⁺), we operationally defined a rate of 1.65 EGFP U⋅min⁻¹ during at least 150 min (on agarose surface, a rate higher than the 25th percentile of the tc starter population with confirmed ICEclc transfer; Fig. 2B, dotted line) to be considered as an individual cell with ICEclc excision. As the average EGFP production rate of tc starters in direct contact with recipient cells but for which no ICEclc transfer could be detected (n = 27) was statistically significantly lower than that from confirmed ICEclc transferring tc starters, this suggested that these cells did not excise the ICE. Averaged over all experimentally observed tc starters, the threshold of 1.65 EGFP U⋅min⁻¹ indicated an overall probability of 0.65 to excise ICEclc, but a decreasing likelihood over time to excise for a given tc starter cell (Fig. S6 G and H).

Time-lapse transfer experiments also showed further tc starter variability. Although, in most cases, one transconjugant was observed from contact to a tc starter with excised ICEclc, in two to four cases, two transconjugant cells arose simultaneously from a single tc starter (Fig. 2D), and, in one case, three transconjugant cells arose (Table 1 and Fig. 2E). In some donor cells, EGFP continued to accumulate even after EGFP fluorescence in the transconjugants started to appear (Fig. S5B, asterisks). This suggests that, at least sometimes, ICEclc is replicated in the donor cell during transfer and that intB13-egfp continues to be transcribed. Many donor cells showed, at some point, an abrupt decrease in EGFP levels (Fig. S5), which, for 50% of the cases, corresponded to a loss of cell integrity in phase-contrast images and possible lysis (Fig. S7). In other cases, EGFP levels of donor cells decreased more smoothly (Fig. 2A and Fig. S5) and cells did not visibly lose integrity, suggesting that EGFP expression is turned off as a result of ICEclc reintegrating into the hosts’ genome.

Fig. S7.

ICEclc excision and cell lysis in tc starter cells. tc starter cells (tcs) and non-tc cells of P. putida ICEclc-intB13-egfp,∆mfsR at different times after inoculation on an agarose patch. Note appearance of tc forming cells at new stationary phase (30 h; tcf). Lysis judged from disappearance of contrast in phase-contrast (PhC) images.

As observed previously (25), approximately 40% (SD = 20% in triplicate experiments; Table 1) of tc starter cells did not divide at all (Fig. S5A), whereas the others divided irregularly but, on average, only 2.2 times (SD = 1.1, n = 18; Fig. S6A). Frequent lysis occurred among dividing tc starters, probably as a collateral consequence of the ICEclc activation (25) (Fig. S6D and Fig. S7). The average cell division rate of tc starter cells was also slower and more heterogenous than that of non-tc cells [121 ± 55 min (n = 115 cells) vs. 96 ± 9 min (n = 228 cells), respectively; Fig. S6C).

Using average EGFP expression values validated here earlier as proxies for ICEclc behavior (e.g., EGFP production rates greater than 1.65 U⋅min⁻¹ means ICEclc excision; Fig. 2A), we interpreted the fates of ICEclc in microcolonies formed from individual tc starters mixed with recipients (Fig. 3 A and B and Fig. S8). In the example of the microcolony shown in Fig. 3A, ICEclc excised after the first cell division in one of the daughters (d1), but remained integrated in the other (d2). Subsequent division of the cell with excised ICEclc (d1) led to one daughter cell (d1.1) with rapidly increasing EGFP fluorescence levels and confirmed ICE transfer, perhaps even to two recipients (t1 and t2; Fig. 3A). The other daughter cell (d1.2) showed a rapid decrease in EGFP fluorescence (Fig. 3B), and lost contrast in phase-contrast images (Fig. 3A), which would be in agreement with that cell losing the ICE and possibly lysing. Daughter cell d2 divided once more, but ICEclc excised only after cell division in one of the daughters (d2.2), given the increase in EGFP fluorescence in that cell (Fig. 3B, green asterisk). Daughter d2.2 later transferred ICEclc successfully to a single recipient (t3; Fig. 3A). Daughter cell d2.1 displayed a gradual decline of EGFP fluorescence, suggesting arrest of P_int expression and degradation of EGFP (Fig. 3B). Analysis of a wider set of microcolonies from tc starters provided further evidence for replication and segregation of ICEclc in daughter cells (Fig. S8A), and cell division even after ICEclc transfer might take place (Fig. S8A). Missegregation of ICE in daughter cells also seemed to occur with occasionally ICE-free cells appearing (Fig. 3B and Fig. S8; note the rapid decrease in EGFP fluorescence in cells e13 and e31 in Fig. S8A). The dynamic fate of ICEclc and tc starters was similar in microcolonies of P. putida carrying WT ICEclc-intB13-egfp, indicating that the variable cell fates were not a result of the mfsR deletion (Fig. S8C).

Fig. 3.

Different cell fates in an ICEclc transferring tc starter cell microcolony. (A) Selected time-lapse microscopy images of cell fates in a tc starter microcolony of P. putida ICEclc-intB13-egfp,ΔmfsR transferring ICEclc to P. putida Ptac-mcherry. PhC, phase-contrast. Arrows point to acquisition of ICEclc by the recipient (transconjugants). Colored outlines on the left represent cell origin (green, tc starter; red, recipient; yellow, transconjugants; gray, other tc starter cell). d1, d1.1, etc., represent tc starter and daughters in B and C. EGFP panels all scaled to maximum intensity of 840, and mCherry panels are autoscaled for intensity. (B) EGFP fluorescence traces of individual cells in the tc starter microcolony. Stars indicate ICEclc excision deduced from the slope of EGFP fluorescence production rate (>1.65 U⋅min⁻¹). Note the rapid decline in EGFP fluorescence in cell d1.2 after 8 h, which occurs after ICEclc excision in the sister cell (d1.1), and thus might represent a case of missegregation and ICE loss. Further traces are shown in Fig. S5B. (C) Deduced genealogy of cell fates and ICEclc transfer. Fig. S8 provides more examples.

Fig. S8.

tc starter cell variability in growing microcolonies in presence of recipient. (A) Divergent cell fates in a growing microcolony of P. putida ICEclc-intB13-egfp,∆mfsR mixed with P. putida P_tac-mcherry–labeled recipient. a, EGFP fluorescence traces of individual cells (colors corresponding to inset on top with colony genealogy). b, Corresponding live-cell microscopy images, with cells appropriately indicated. Arrowheads point to transconjugants. Note the high slopes of EGFP production rates of cells e1, e14, e24, and e25, above the defined 25% quantile of ICEclc transferring cells, indicating ICEclc excision. ICEclc transfer was confirmed for cells e1, e23, and e24. Further note how EGFP production continues to increase rapidly in sister cells e14 and e24, suggesting ICEclc replication and correct segregation among daughter cells. Excision in cell e1 is followed by a continued increase of EGFP fluorescence in daughter e25 but a more gradual decline in e1, which might indicate ICEclc missegregation and reintegration in e1. Finally, note the sudden “drop” in EGFP fluorescence in e13 and e31, and their loss of integrity in PhC at 13 h, suggesting loss of the ICE and cell lysis. (B) Further examples of deduced cell fates in microcolonies (a–f) of P. putida ICEclc-intB13-egfp,∆mfsR mixed with P. putida P_tac-mcherry–labeled recipient. Notes cases of ICEclc transfer, followed by further cell division for microcolonies a and f. (C) Deduced cell fates in microcolonies (a–d) of P. putida with WT ICEclc-intB13-egfp mixed with P. putida P_tac-mcherry–labeled recipient. Note the similar cell fates to microcolonies of P. putida ICEclc-intB13-egfp,∆mfsR.

Robust ICEclc Transfer Despite Variable tc Cell Fates.

Despite this highly variable individual cell behavior, ICEclc transfer was quite robust. Among 174 tc starters, 91 contacted recipient cells (alone or one of their daughters after division), resulting in 57 transconjugants (Table 1). In total, 43 transconjugants among 57 (75%) originated from tc starters that had formed a small microcolony (3.0 ± 1.4 cells), whereas 14 (25%) transfer events came from single, nondividing tc starter cells (Table 1). Excision did not lead to ICEclc transfer in all cases, despite visible contact to recipients. Among 27 measured tc starters with contact to recipient cells but without detected transfer (Fig. 2B, TRA⁻), eight displayed EGFP production rates greater than the 25th percentile defined for tc starters with confirmed transfer, suggesting that these cells did excise ICEclc but somehow were unable to successfully deliver ICEclc to the recipient. The success of transfer upon donor–recipient contact in cells with excised ICEclc can thus be estimated as 75% (i.e., 28 TRA⁺ events among 28 + 8 cells with excision). As previously suggested (31), the formation of small tc donor microcolonies is indeed advantageous for ICEclc transfer, given that the 60% of tc starters that could divide contacted almost 10 times more recipients (294 contacts) than the nondividing tc starters (34 contacts), and provided 75% of all transfers (Table 1).

A Model for tc Cell Fate Transitions.

To better understand the occurrence of the different cell fates accompanying ICEclc transfer, we developed a model that describes transitions between a limited number of donor cell states, which we defined and parameterized based on experimental observations (Fig. 4A and Fig. S6). The first transformation starts when cells with integrated ICEclc (i.e., non-tc cells) initiate ICEclc transfer competence (i.e., tc cells) during stationary phase when nutrients and carbon substrate become limiting for growth (Fig. 4B, stationary phase). This transformation rate, t_non-tc, has a low probability (0.04 for WT ICEclc; Fig. S6I) and occurs, on average, after 4–5 h into stationary phase (Fig. S6H), leading to a proportion of approximately 4% tc cells over time (Fig. 4C, stationary phase). When tc cells are again provided with nutrients (Fig. 4B, growth phase), their fates become more variable. Non-tc cells divide with a relatively coherent birth rate, b_non-tc, whereas tc cells divide more irregularly (b_tc; Fig. S6 A–C). With a probability of 0.65, tc cells excise the ICE (e_tc; Fig. 4A). They can again divide in excised state or lose the ICE and lyse (Fig. S6 D–I). Simulations of population development by stochastic sampling of experimentally derived parameters suggest that this leads to a temporary increase of the total number of tc starters and ICEclc excision, and also to a small increase in the population of cells without ICE (Fig. 4B, growth phase). Proportionally speaking, however, the number of tc starters rapidly decreases as a result of more consistent cell division of non-tc cells (Fig. 4C, growth phase), declining from 4% to an estimated 0.014% after 10 generations (mostly non-tc doublings). Non-tc and tc starter cell numbers counted on time-lapse imaging of P. putida ICEclc-intB13-egfp WT followed the simulated predictions quite nicely (Fig. 4D), although tc starters seem to persist slightly better than predicted (parameterization was performed on the basis of the ∆mfsR mutant data). Predictions from simulations using different non-tc to tc cell transformation rates (i.e., t_non-tc) suggest that ICEclc fitness markedly decreases if t_non-tc increases from 4% (WT) to 20% (Fig. 4E). Fitness is counted here as the copy number of ICEclc in the donor population plus 0.75× the number of cells in which ICEclc excises (assuming that these would successfully transfer the ICE to a recipient). Similarly, ICEclc fitness would decrease if tc cells form already during growth phase instead of stationary phase (even at 4% WT transformation rate; Fig. 4E). These predictions would imply that the ICEclc system was selected for low tc cell activation in stationary phase, given the highest fitness.

Fig. 4.

Stochastic modeling of tc cell and ICEclc fates. (A) Schematic cell fate transitions during growth and stationary phase, following experimental observations. Process parameters stochastically sampled from experimentally derived probability distributions (Fig. S6). (B) Mean predictions (solid lines) plus SD (lighter shading) of subpopulation development from 20 independent simulations in stationary and growth phase during 1,000 min each under starting conditions as indicated. (C) Same data as in B but displayed as the proportion of tc cells in the population over time. (D) Predicted and observed population development of tc and non-tc cells in P. putida WT ICEclc-intB13-egfp (strain 4611) during growth phase, at the indicated observed starting cell numbers. (E) ICEclc fitness (as ICE copy number in donor populations plus 0.75× the number of excised tc cells, assuming these can successfully transfer to recipient) after one or five cycles of alternating stationary-growth phases, and as a function of rate and timing of tc transformation (t_non-tc). Study cases (i to vii) are shown on the right. Letters indicate statistically significantly different groups (Kruskal–Wallis test, df = 38, P < 0.001). Data points indicate the mean abundance in 20 independent simulations, with whiskers indicating calculated SDs.

Discussion

Horizontal gene transfer is a fascinating process in the prokaryotic world, through which large pieces of DNA can be actively transmitted between cells. Until now, there has been only limited insight into and appreciation for the actual cell fates during transfer of mobile DNA (32, 34), and how these fates determine the ecological and evolutionary success of the types of mobile DNA becoming dominant in bacterial populations. ICEs are still remarkably poorly understood considering their wide distribution in bacterial genomes and importance among conjugative systems (7). The system we study here, ICEclc, is an example of an abundant ICE family in a wide variety of γ- and β-proteobacteria (14, 17), among which are several P. aeruginosa isolates (10, 11, 41). Our results enable us to draw a number of conclusions about the basic steps in the behavior of the ICEclc family. First, the process of ICEclc transfer competence development is initiated in a small proportion of stationary-phase cells from non-tc cells (i.e., tc formers). This differentiation to tc formers can be observed in time-lapse microscopy (Figs. 1E and 2A). When tc formers are presented with new carbon substrate, they develop different cell fates (as deduced here from cells we named tc starters). Although we did not experimentally show here that tc formers (appearing in stationary phase) are the same cells as tc starters (at the start of new growth phase), because they appear at different stages in the experimental setup (e.g., Figs. 1E and 2A), modeling suggests that this is the most likely explanation (Fig. 4C). In contrast, it is extremely unlikely that tc starters are senescent cells or persisters that maintain silently in the population and grow out when conditions have become more favorable [like in the case of antibiotic persisters (42, 43)]. We do acknowledge that tc starters can persist over long durations in growth phase (Figs. 1E and 4D), even though they frequently lyse and stop dividing or do not divide at all (Fig. S6 A and D and Fig. S7), and are very rapidly overgrown by dividing non-tc cells (Fig. 4D). However, as previous experiments have shown, elongation and division of such persisting tc cells cannot be induced a second time (that is, after passage through stationary phase as in Fig. 1 D and E) by renewed addition of carbon source (25). The molecular details of this cell arrest are currently unknown.

Contrary to our expectations, in most instances, ICEclc does not excise or transfer from tc formers in stationary phase, but this predominantly (Fig. 2 C–F) occurs when these cells receive fresh nutrients (i.e., tc starters; Fig. 1D). Our data strongly suggest that availability of carbon or some other nutrient is needed as a trigger for excision and subsequent transfer. Because at least 40% of tc starters can excise and transfer ICEclc without further cell division (Table 1), we conclude that it is not so much a renewed division cycle after stationary phase that acts as a trigger, but the actual presence of nutrients, as demonstrated by the qPCR data and silica-gel experiments (Fig. 2G). One could imagine that nutrients may be needed to provide sufficient energy to set in motion the complex DNA transfer machinery. As tc donor cells have already initiated the transfer competence cascade in stationary phase, it is a priori unclear how, mechanistically, nutrients can provide the trigger for ICE excision and transfer. Either somehow the tc cells in stationary phase only transcribe but do not translate the ICE mRNAs (which seems unlikely because EGFP is produced in such cells from the combined intB13-egfp mRNA), or all necessary transfer proteins and the integrase are produced in stationary-phase tc cells but some other unique factor only becomes available upon new nutrient stimulation that aids the integrase to excise the ICE DNA and unleashes transfer. These and other questions should be studied in future experiments.

As deduced from our intB13-egfp reporter tool, ICEclc behavior is highly variable among individual P. putida tc starters, and different cell (and ICE) fates can be inferred (Fig. 4A). Excision is not automatic after initiation of ICEclc transfer competence, but occurs at a probability of 0.65, whereas, in other cells, the ICE remains integrated. In some tc starters, the ICE replicates and correctly segregates copies among daughter cells (concluded from consistently high EGFP production rates in both daughter cells), whereas others seem to lose the ICE upon division (concluded from rapid decrease in EGFP production rate). An estimated 75% of cells with excised ICEclc successfully deliver the ICE upon cell contact to a new ICE-free recipient. Inferred ICEclc replication and correct segregation in some tc starters would be in agreement with accumulating evidence on temporary replication of other ICE after excision using qPCR methods (22, 44). It is tempting to suggest that the variance in EGFP production rates in tc starters with confirmed ICEclc transfer (Fig. 2B, TRA⁺) is a result of different ICE copy numbers in cells, but this cannot be firmly concluded from our reporter tool. Different excised ICEclc copy numbers would be a reasonable assumption to explain the small proportion of tc starters that transfer ICEclc twice or even three times almost simultaneously (Fig. 2 D and E and Fig. S5). The mechanisms for multiple transfer from single donor cells is ill understood, but can contribute to the overall transfer success. It may be the result of a TraI-dependent ICEclc replication after excision analogous to what has been described for ICESXT from Vibrio cholerae (22) and ICEBs1 from Bacillus subtilis (44) and/or a consequence of its double origin of transfer (29), which might enable successive rounds of ICEclc DNA processing for conjugative transfer.

Our results demonstrate clearly that tc starter cells, when the ICEclc excision and transfer process has started, undergo highly variable fates, exemplified by limited or arrested cell division and cell lysis. Previous data indicated that this is, at least in part, the result of ICEclc-encoded factors (25). Further observations using cell staining for reactive oxidative species suggested that tc starters are more frequently damaged than non-tc cells (25), which would be in agreement with the idea that such cells are in poor shape. One could thus imagine that the various cell fates of tc starters are not just spontaneous stochastic events but a result of how the ICE tries to manipulate and transform the host cell into a successful transfer machine, with dire consequences for that cell. It is too early to conclude whether this behavior is general for horizontal gene transfer or specific for ICEs from the ICEclc family, as few other conjugative systems have been studied at the single-cell level. Studies on F and R751 plasmid transfer from Escherichia coli (33, 34) and ICEBs1 transfer in B. subtilis (32) did not particularly describe phenotypic variability among donor cells.

In conclusion, our results provide a unique, fascinating, and unprecedented view on dynamic cell fates during horizontal gene transfer. The robust transfer of ICEclc despite highly variable behavior at the level of individual donors can help to explain the success of distribution of ICEclc-like elements in a wide variety of γ- and β-proteobacteria (14, 17), but our results and modeling also help to understand why ICE fitness benefits from the element keeping a low transfer frequency.

Materials and Methods

Bacterial strains used in this study are listed in Table S1. E. coli strains used for plasmid cloning were routinely grown at 37 °C on Luria–Bertani (LB) medium. P. putida strains were cultured at 30 °C on LB or 21C minimal medium (MM) (45) complemented with 5 mM 3-CBA or 10 mM sodium succinate. If necessary, antibiotics were added at the following concentrations: ampicillin 100 µg⋅mL⁻¹, kanamycin 25 µg⋅mL⁻¹, and gentamicin 20 µg⋅mL⁻¹.

Targeted chromosomal mutations in P. putida (insertion and deletions) were created by the I-SceI counterselection double recombination system as described elsewhere (26, 46). A 752-bp promoterless egfp gene was amplified from a derivative of plasmid pJAMA23 (47) and inserted 27 bp downstream of the stop codon of intB13 (Fig. S1) in P. putida carrying WT ICEclc (strain 2737) or ICEclc-∆mfsR [strain 4322 (27)], with the ICE inserted in the gene for tRNA-gly5 (40). The resulting strains (P. putida 4611 and 4612, respectively) were further used to create an 18-bp deletion in attL of ICEclc by using the same double-recombination techniques (producing P. putida strains 4754 and 4755, respectively). A single copy of a P_inR-echerry reporter gene cassette was randomly inserted in P. putida ICEclc-intB13-gfp (strains 4669–4672) and P. putida ICEclc-intB13-gfp,∆mfsR (strains 4673–4676) using mini-Tn5 delivery, as previously described (17). P. putida UWC1 strains containing an attR-mcherry conditional trap for ICEclc insertion (Fig. 1B) were constructed by cloning the attR fragment [formerly named ∆P_int-fragment (40)] in front of a promoterless mcherry on plasmid pJAMA39 (17). The NotI attR-mcherry fragment was then recovered and cloned into pCK218 (48), and the resulting plasmid was electroporated into P. putida UWC1 to deliver the mini-Tn5 cassette, as previously described (40), to produce strains 5001–5003.

Transfer frequencies of ICEclc from P. putida donors to P. putida recipients on plates were determined following experimental conditions described previously (27) and are further specified in SI Materials and Methods.

Time-lapse microscopy was performed as described previously with minor changes (37). Briefly, all derivatives of UWC1 carrying ICEclc were pregrown overnight in LB medium supplemented with antibiotics to select for maintenance of their respective traits. An aliquot of 100 µL of this preculture was then transferred in 20 mL MM with 5 mM 3-CBA (without further antibiotics) and allowed to grow for 96 h, after which 1 mL of cells were 2- to 10-fold diluted in MM before being spotted on agarose patches containing 0.1 mM 3-CBA. Recipient cells were grown for 24 h in 20 mL MM with 10 mM sodium succinate before being mixed in a 2:1 (vol/vol) ratio with the 96-h-old donor cell culture. Cells were not further diluted but were washed once in MM before being spotted onto the agarose patches. Agarose patches were prepared as previously described (27). Silica gel-based patches were prepared by mixing 0.5 mL of twofold concentrated MM with 0.42 mL of a silica-gel solution [composed of 10 g of silicate gel (no. 60738; no. 112926–00-8; Fluka) and 7 g of KOH dissolved in 100 mL Milli-Q water] with or without 3-CBA (at 0.1 mM final concentration). Immediately after addition of 83 µL phosphoric acid solution (fivefold dilution pure orthophosphoric acid), 130 μL was pipetted on a coverslip to create 1-mm-thick patches as described previously (27). Ten arbitrarily defined regions with cells were followed automatically for 50 h with image acquisition every 30 min. Image series were deconvoluted, allowing consecutive automated cell tracking using MetaMorph (series 7.5, MDS; Analytical Technologies) and the mean fluorescence intensities of single cells at all time points were extracted by using an in-house–written Matlab script. EGFP production rates in single cells were calculated by linear regression of periods of increasing EGFP fluorescence values, taking at least six consecutive time measurements into consideration (150 min). The 25th percentile value of EGFP production rates of confirmed ICEclc transferring tc starters (1.65 U⋅min⁻¹; n = 28) was used as a cutoff, above which a tc cell was considered having excised ICEclc. Cells with negative slopes of EGFP fluorescence over time, or for which fewer than six time points were available, and/or with R² < 0.80, were discarded for statistical analysis. Because the fluorescence intensities of non-tc cells remained constant and low, their slopes were calculated from two random time points by linear regression. P. putida P_tac-mcherry recipients were considered as having acquired ICEclc-intB13-egfp if EGFP values augmented for at least six consecutive measurement points (150 min) and were higher than the average EGFP of recipients not in contact to tc starters. Live-cell microscopy transfer data were pooled from triplicate independently started agarose patch growth experiments, either with donors alone or with donor-recipient mixtures, using tc starters present on three to five randomly selected image locations for each replicate.

For qPCR, three independent cultures of P. putida ICEclc-intB13-egfp,∆mfsR (strain 4612) and P. putida ICEclc-intB13-egfp,∆attL,∆mfsR (strain 4755) were grown for 96 h in 20 mL MM with 5 mM 3-CBA. At this time point, 1 mL of each culture was collected for DNA extraction (stationary phase), whereas another 1 mL was transferred into 20 mL MM containing 0.1 mM 3-CBA. Cells were allowed to grow for 4 h and then collected by centrifuging the 20-mL cultures (early exponential phase). DNA was extracted using the Wizard Genomic DNA purification kit (Promega) and used to quantify by real-time PCR the relative abundances of the clcB gene of ICEclc and of attP (formed exclusively when ICEclc excises) (40). The primer set for clcB amplified a 177-bp fragment (5′-GGTTCAGAGAGCGTGCCTTC + 5′-GCGCTGAAACCATCAAGGTC), whereas the attP primer set (5′-CTTCGCTGGCCACCTCGG + 5′-GTGGCGCTCGCTGGAATGA) amplified a 166-bp fragment. Primer set calibration curves were prepared by using 10-fold serial dilutions of DNA obtained from P. putida ICEclc-intB13-egfp,∆mfsR grown for 96 h in MM with 5 mM 3-CBA. Real-time qPCR was performed on an ABI Prism Step One Plus Real-Time PCR system (Applied Biosystems) with 5 μL of SYBR Mix Select (Life technologies), 3.2 μL of water, 0.4 μL of each primer (5 μM), and 1 μL of DNA. Reactions were held for 2 min at 50 °C and 2 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 72 °C. Products were verified by melting curves. qPCR efficiencies for both primer sets were above 1.83 and with R² > 0.996. Samples were diluted such that clcB and attP threshold cycle-amplification values fell within both calibration curves. The abundance of the attP fragment was calculated relative to clcB and taking into account the qPCR efficiencies of both primer sets.

A stochastic model was built that predicts the fates of cells and ICEclc during a population growth phase (sufficient nutrients available for cell division) and during stationary phase (no net population growth). Four cell types were assumed (Fig. 4A): (i) cells that have initiated the tc pathway but have not excised the ICE yet (tc), (ii) tc but with excised ICE (excised tc), (iii) tc cells that have lost the ICE (non-ICE), and (iv) cells in which ICEclc remains silent (non-tc). During growth, the different cell types can divide or lyse. In addition, tc cells can excise their ICE and become excised tc cells, and excised tc cells can lose their ICE and become non-ICE cells, as illustrated schematically in Fig. 4A. Because of the need to carry ICEclc for metabolizing the substrate 3-CBA, non-ICE cells are not allowed to divide. Following experimental observations, none of the cells are allowed to divide during stationary phase, but non-tc cells can differentiate to tc cells, which can lyse (Fig. 4A), whereas the lysis rate for non-tc cells is set to zero. Transition rates are determined by stochastic sampling from probability distributions derived from experimental data (Fig. S6). Starting populations for simulations were composed based on experimental data. For example, populations enter stationary phase without any tc cells, and their appearance essentially follows the differentiation rate from non-tc cells (Fig. 4B). Growth phase starts with an initial population of 1,000 cells with 4% tc cells as for the WT ICEclc situation. Both phases were allowed 1,000 time steps of simulation. We further simulated the effects of various tc differentiation rates and the timing of tc differentiation (stationary phase only or during growth and stationary phases) on ICE fitness after single or multiple growth/stationary phase cycles. Multiple cycles were simulated by proportionally diluting the final abundance of all cell types at the end of the stationary phase to a starting size of 1,000 cells. A new stationary phase started with the number of non-tc cells from the end of the previous growth phase and no tc cells, matching experimental observations. ICE fitness was scored as the copy number in the donor population plus 0.75 times the number of cells with excised ICE (as these would be able to successfully transmit the ICE to a recipient) at the end of each cycle. The simulations were coded in Matlab R2014b. Statistical significance was inferred on averages from 20 independent simulations by using the Kruskal–Wallis test.

SI Materials and Methods

Strains and Culture Conditions.

Escherichia coli strains used for plasmid cloning were routinely grown at 37 °C on LB medium. Pseudomonas putida strains were cultured at 30 °C on LB or 21C MM (45) complemented with 5 mM 3-CBA or 10 mM sodium succinate. If necessary, antibiotics were added at the following concentrations: ampicillin 100 µg mL⁻¹, kanamycin 25 µg mL⁻¹, and gentamicin 20 µg mL⁻¹.

Strain Constructions.

All strains used in this study are listed in Table S1. Targeted chromosomal mutations in P. putida (insertion and deletions) were created by the I-SceI counterselection double recombination system as described elsewhere (26, 46). A 752-bp promoterless egfp gene was amplified from a derivative of plasmid pJAMA23 (47) and inserted 27 bp downstream of the stop codon of intB13 (Fig. S1) in P. putida carrying WT ICEclc (strain 2737) or ICEclc-∆mfsR [strain 4322 (27)] inserted in the gene for tRNA-gly5 (40). The resulting strains (P. putida 4611 and 4612, respectively) were further used to create an 18-bp deletion in attL of ICEclc by using the same double recombination techniques (producing P. putida strains 4754 and 4755, respectively). A single copy of a P_inR-echerry reporter gene cassette was randomly inserted in P. putida ICEclc-intB13-gfp (strains 4669–4672) and P. putida ICEclc-intB13-gfp,∆mfsR (strains 4673–4676) using mini-Tn5 delivery as previously described (17). P. putida UWC1 strains containing an attR-mcherry conditional trap for ICEclc insertion were constructed by cloning the attR fragment (formerly named ∆P_int-fragment) (40) in front of a promoterless mcherry on plasmid pJAMA39 (17). The NotI attR-mcherry fragment was then recovered and cloned into pCK218 (48), and the resulting plasmid was electroporated into P. putida UWC1 to deliver the mini-Tn5 cassette, as previously described (40), to produce strains 5001–5003.

ICEclc Transfer on Plates.

Transfer frequencies of ICEclc from P. putida donors to P. putida recipients on plates were determined following experimental conditions described previously (27). As recipient strains, we used P. putida UWC1 P_tac-mcherry (strain 2744) or UWC1 mini-Tn5-attR-mcherry (strains 5001–5004). Briefly, donor strains were grown for 48 h in 20 mL MM + 5 mM 3-CBA and mixed at 1:2 (vol/vol) with recipient strains that were grown in 20 mL MM + 10 mM sodium succinate for 24 h. Cell mixtures (1.2–1.5 mL) were concentrated by centrifugation at 13,000 × g for 2 min, resuspended into 50 µL MM solution, spotted on MM plates containing 0.5 mM 3-CBA, and placed at 30 °C for 48 h. After this period, cell spots were resuspended in 1 mL of a sterile MM solution, which was serially diluted and plated on MM + 5 mM 3-CBA to count the number of donors and MM + 5 mM 3-CBA kanamycin to count transconjugants. The transfer frequency was determined from the ratio of colony-forming units of transconjugants and of the donor.

They were first tested in plate mating assays by using P. putida with one ICEclc copy integrated in the tRNA-Gly gene number 5 as a donor strain. All three recipients received ICEclc at the similar frequency (2.6–2.8%). Moreover, as many as 17.7 ± 5.7% of the colonies were bright red under the blue light, a proportion close to the 20% expected (four natural insertion sites plus the conditional trap, see ref. 40). One of these recipients was used in time-lapse microscopy transfer experiments using P. putida ICEclc-intB13-egfp,∆mfsR as donor strain. Here, recipient cells are expected to turn green only when ICEclc as been transferred, and then red if insertion occurred within the chromosomal conditional trap. Indeed, we were able to detect cells producing EGFP and subsequently mCherry.

Time-Lapse Microscopy.

Time-lapse microscopy was performed as described previously with minor changes (37). Briefly, all derivatives of UWC1 carrying ICEclc were pregrown overnight in LB medium supplemented with antibiotics to select for maintenance of their respective traits. An aliquot of 100 µL of this preculture was then transferred in 20 mL MM + 5 mM 3-CBA (without further antibiotics) and allowed to grow for 96 h, after which 1 mL of cells were 2- to 10-fold diluted in MM before being spotted on agarose patches containing 0.1 mM 3-CBA. For time-lapse transfer experiments, recipient cells were grown for 24 h in 20 mL MM + 10 mM sodium succinate before being mixed in a 2:1 (vol/vol) ratio with the 96-h-old donor cell culture. Cells were not further diluted and were directly spotted onto the agarose patches. Agarose patches were prepared as previously described (27). Ten arbitrarily defined regions with cells were followed automatically for 50 h with image acquisition every 30 min. Image series were deconvoluted, allowing consecutive automated cell tracking using MetaMorph (series 7.5, MDS; Analytical Technologies), and the mean fluorescence intensities of single cells at all time points were extracted by using an in-house written Matlab script. EGFP production rates in single cells were calculated by linear regression of periods of increasing EGFP fluorescence values over time, taking at least six consecutive time measurements into consideration (150 min). The 25th percentile value of EGFP production rates of confirmed ICEclc transferring tc starters (1.65 U⋅min⁻¹; n = 28) was used as a cutoff above which a tc cell would be considered having excised ICEclc. Cells with negative slopes of EGFP fluorescence over time, or for which fewer than six time points were available, and/or with R² < 0.80, were discarded for further analysis. Because the fluorescence intensities of non-tc cells remained constant and low, their slopes were calculated from two random time points by linear regression. P. putida P_tac-mcherry recipients were considered as having acquired ICEclc-intB13-egfp if EGFP values augmented for at least six consecutive measurement points (150 min) and were higher than the average EGFP of recipients not in contact to tc starters. Live-cell microscopy transfer data were pooled from triplicate independently started agarose patch growth experiments, either with donors alone or with donor–recipient mixtures, using tc starters present on five randomly selected image locations for each replicate.

qPCR.

Three independent cultures of P. putida ICEclc-intB13-egfp,∆mfsR (strain 4612) and P. putida ICEclc-intB13-egfp,∆attL,∆mfsR (strain 4755) were grown for 96 h in 20 mL MM + 5 mM 3-CBA. At this time point, 1 mL of each culture was collected for DNA extraction (stationary phase), whereas another 1 mL was transferred into 20 mL MM containing 0.1 mM 3-CBA. Cells were allowed to grow for 4 h and then collected by centrifuging the 20 mL cultures (early exponential phase). DNA was extracted by using the Wizard Genomic DNA purification kit (Promega) and used to quantify by real-time PCR the relative abundances of the clcB gene of ICEclc and of attP (formed exclusively when ICEclc excises) (40). The primer set for clcB amplified a 177-bp fragment (5′-GGTTCAGAGAGCGTGCCTTC + 5′-GCGCTGAAACCATCAAGGTC), whereas the attP primer set (5′-CTTCGCTGGCCACCTCGG + 5′-GTGGCGCTCGCTGGAATGA) amplified a 166-bp fragment. Primer set calibration curves were prepared by using 10-fold serial dilutions of DNA obtained from P. putida ICEclc-intB13-egfp,∆mfsR grown for 96 h in 5 mM 3-CBA. Real-time qPCR was performed on an ABI Prism Step One Plus Real-Time PCR system (Applied Biosystems) with 5 μL of SYBR Mix Select (Life Technologies), 3.2 μL of water, 0.4 μL of each primer (5 μM), and 1 μL of DNA. Reactions were held for 2 min at 50 °C and 2 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 72 °C. Products were verified by melting curves. qPCR efficiencies for both primer sets were greater than 1.83 with R² > 0.996. Samples were diluted such that clcB and attP threshold cycle-amplification values fell within both calibration curves. The abundance of the attP fragment was calculated relative to clcB and taking into account the qPCR efficiencies of both primer sets.