Screening and design of lysis unit
To obtain an efficient and quick-response cell lysis system, five lysis proteins, CoIM24, Lysep25, MS226, SRRz27, and X174E17, from E. coli and phages were selected and overexpressed in E. coli. The OD600 of cells expressing CoIM decreased from 0.61 to 0.25, which was 192%, 192%, 172%, and 52% lower than the corresponding values in cells expressing Lysep, MS2, SRRz, and X174E, respectively, after 1 h of induction. Similarly, the OD600 of cells expressing CoIM was 621%, 629%, 564%, and 93% lower than the corresponding values in cells expressing Lysep, MS2, SRRz, and X174E, respectively, after 2 h of induction. Furthermore, the cell viability decreased from 8.03 × 108 CFU/mL (control group without CoIM expression) to 2.73 × 105 CFU/mL after 2 h of induction (Supplementary Fig. 1). E. coli overexpressing CoIM exhibited the least cell growth among the five lysis proteins (Fig. 1c). In addition, it was found that only cells with the CoIM protein could provide a clear broth with no turbidity (Fig. 1a). The protein content of CoIM and X174E in the cell supernatant was 0.87 and 0.94 mg/mL, respectively, after 2 h of induction (Fig. 1b). Accordingly, 93.4% of the cells (control: 5.6%) were damaged, as determined by propidium iodide staining and flow cytometry (Fig. 1d). Cell damage was further demonstrated by scanning electron microscopy, as shown in Fig. 1e. The morphology of cells overexpressing CoIM was irregular or even completely broken. These results demonstrated that CoIM is an efficient and quick-response lysis protein that can release intracellular proteins from E. coli via cell lysis and that it is the most suitable element for the lysis unit.
a The effect of different lysis proteins on cell OD600 in 2 h. The clarity of LB medium represented E. coli has been lysed by lysis proteins. (P values = 0.6938725; 0.000165; 0.6108144; 0.1437364; 0.000358) b The effect of different lysis proteins on the release of contents in 2 h. The blue color of the reaction solution represented intracellular protein has been released into the supernatant. (P values = 0.0773454; 0.0002937; 0.0582133; 0.0015429; 0.0000676.) Strain JM109 harboring empty vector was used as the control group. All groups were grown at 37 °C and 200 r.p.m. in LB medium for 4 h. Statistical significance of values at 2 h was determined and was indicated as * for P < 0.05, ** for P < 0.01 and *** for P < 0.001, respectively. c Spot assays were used to test the effect of different lysis proteins on the viability of E. coli. The dilution of samples from left to right was 100, 101, 102, 103, 104, 105. Strain JM109 harboring empty vector was used as the control group. d PI staining was used to detect the effect of different lysis proteins on mortality rate. The percentage of fluorescence intensity represented the mortality ratio. The fluorescence intensity of living cells was less than 103, and that of dead cells was more than 103. Strain JM109 harboring empty vector was used as the control group. e Scanning electron microscope was used to detect changes on the surface of E. coli after expressing different lysis proteins. The white arrows indicated where the cells are broken. Three experiments were repeated independently with similar results. f Structure of wild-type CoIM and schematic of reconstructing CoIM*. Values are shown as mean ± s.d. from three (n = 3) biological replicates. Two-tailed t tests were used to determine statistical significance. Source data are provided as a Source Data file.
As shown in Fig. 1f, the wild-type CoIM consists of an N-terminal translocation domain, a central receptor-binding domain, and a C-terminal toxicity domain. To increase the lysis efficiency of CoIM, the N-terminal translocation domain, and the central receptor-binding domain were deleted, and then, the C-terminal toxicity domain was fused with different signal peptides and translocated into the periplasmic space to lyse E. coli24 (Fig. 1f, Supplementary Fig. 2a). Six signal peptides (pelB, NSs, ompA, Tat1, Tat2, and Tat3)28 from two secretion pathways (Sec-pathway and Tat-pathway) were introduced and tested. Among them, the C-terminal toxicity domain that fused with pelB to reconstruct CoIM* exhibited the most efficient cell lysis effect, with the OD600 decreasing to 0.12 (Supplementary Fig. 2b). The protein content of CoIM* (pelB group) reached 0.92 mg/mL after 2 h, which was only 2% and 3% higher than that of the NSs and ompA groups, respectively (Supplementary Fig. 2c). These results demonstrated that signal peptides from the Sec-pathway are more suitable for reconstructing CoIM, and pelB was selected for further testing because its translocation capacities were 1.86- and 1.15-fold higher than that of the other two signal peptides at 2 h, respectively (Supplementary Fig. 2h).
To efficiently control CoIM*, a protease-trigger mechanism was introduced: the TEVp protease cleavage site (tev site) was fused to the N-terminal of pelB with an F degron (Supplementary Fig. 2d). Following the expression of the protease TEVp, the N-terminal of the F degron was exposed, leading to the degradation and inactivation of the reconstructed CoIM*. As shown in Supplementary Fig. 2e, f, when CoIM* was expressed, the OD600 decreased from 0.63 to 0.27, and the protein content in the supernatant increased to 0.93 mg/mL. When CoIM* and TEVp were expressed at the same time, the OD600 increased to 1.31, and the protein content was maintained at 0.76 mg/mL. These results indicated that although small amounts of CoIM* were not degraded and partly caused cell lysis due to the high efficiency of the pelB translocation system, the accumulation of TEVp in cells could block CoIM* from entering the periplasmic space and lysing E. coli.
Construction and evaluation of the programmed switch
E. coli expressing CoIM* with stationary phase promoters could not grow (Supplementary Fig. 3). Given their lethal effect on E. coli, proteases were used to build a protease-based regulatory switch (programmed switch) to delay the expression of CoIM*. This switch consisted of an action arm and a repression arm. As shown in Fig. 2a, the repression arm expressed the protease TEVp, which is regulated by a more stringent stationary phase promoter (Pfic) (Supplementary Fig. 4c). The action arm expressed the protease TVMVp, which was regulated by a weaker growth phase promoter (PrpsM) (Fig. 2a, Supplementary Fig. 4f). To construct this programmed switch, these two proteases were modified by fusing their N terminal with an F degron and another protease cleavage site that could be specifically recognized and degraded by the corresponding proteases (Fig. 2a). In the protease-based regulatory switch, protease TEVp could be activated only when TVMVp was completely degraded and vice versa.
a Designing and characterizing the programmed switch. PrpsM and PrpsT are growth phase promoters; Pfic is a stationary phase promoter; RBS is a ribosome binding site; TEVp is tobacco etch virus protease; TVMVp is tobacco vein mottling virus protease; mKate2 is a red fluorescent protein. b Designing of the programmed switch, and characterizing programmed switch by expression unit. c The fluorescence abundance curve of the repression arm was used to characterize the modification effect of different regulation strategies on the programmed switch. (P values = 0.016130; 0.000202; 0.000202; 0.002192; 0.002192; 0.000089; 0.000132; 0.000013; 0.000001; 0.000001; 0.000059; 0.000002.) The increase of fluorescence intensity represented the accumulation of mKate2, and the decrease of fluorescence intensity represented the degradation of mKate2 and indicated repression arm began to take effect (programmed switch was activated). After 24 h, 0.1 mL of 100 mg/mL yeast extract was added into medium every 4 h. All groups were grown at 37 °C and 200 r.p.m. in LB medium for 44 h. The OD600 and fluorescence intensity were measured every 1 h. Statistical significance of switch was determined and was indicated as * for P < 0.05, ** for P < 0.01 and *** for P < 0.001, respectively. d, e Programmed switch inhibited enzyme activity of β-galactosidase. The absorbance at 420 nm was measured after cells were sampled each 6 h and lysed by ultrasound. The left d was the absorbance curve, and the right e was the reaction solution. In strain fic (the control group), TEVp was driven by Pfic. The enzyme activities were measured every 2 h. Values are shown as mean ± s.d. from three (n = 3) biological replicates. Two-tailed t tests were used to determine statistical significance. Source data are provided as a Source Data file.
First, to characterize the programmed switch behaviors and switch time, a repression reporter unit was constructed, resulting in strain PS1008 (Supplementary Table 8). The repression reporter unit consisted of mKate2, which was modified by fusing its N-terminal with an F degron and a TEVp cleavage site (tev site) that could be specifically recognized and degraded by TEVp (Fig. 2a). The expression of modified mKate2 was under the control of PrpsT (a stronger growth phase promoter for producing more mKate2), and modified mKate2 could be degraded by the accumulation of TEVp, but the degradation was removed by cleaving TEVp via TVMVp. As shown in Fig. 2c, the switch time of strain PS1008 was 10 h, indicating that it was delayed by 1 h compared with that of the control group (TEVp was driven by the stationary phase promoter Pfic). These results indicated that TVMVp was rapidly degraded after being cleaved by TEVp. Therefore, three different strategies (RBS regulation, initiation codon regulation, and degradation tag regulation) (Supplementary Fig. 5a–c) were selected to fine-tune TEVp abundance and to regulate the switch time. To verify the effect of these strategies on TEVp abundance, green fluorescent protein (GFP) was fused to TEVp. As shown in Supplementary Fig. 6, the fluorescence intensity gradually decreased, indicating that the abundance of TEVp also decreased. Therefore, as demonstrated in Fig. 2c and Supplementary Fig. 7, when we replaced RBS8 with RBS1 (PS1008–PS1001), the switch time increased from 10 to 11 h. Based on PS1001, when changing the initiation codon from ATG to GGA (PS1011–PS1061), the switch time increased from 11 to 28 h. Based on these two strategies, the degradation tags (DSA, AAV, and LAA) (namely PS1361‒PS1061) were added at the C-terminal of TEVp, causing the switch time to increase to 32 h. In a computational model of the programmed switch (Supplementary Fig. 19), the increase in TEVp abundance (At) caused the decreased switch time (T).
Then, to further characterize the behaviors of the action arm and the relative switch time (when the fluorescence density increased), an action reporter unit was constructed, consisting of mKate2, which was modified by fusing its N-terminal with an F degron and a TVMVp protease cleavage site (tvmv site) that could be specifically recognized and degraded by TVMVp (Supplementary Fig. 8a). The expression of modified mKate2 was under the control of Pfic (a stationary phase promoter), and modified mKate2 could be degraded by the accumulation of TVMVp, but the degradation was removed by cleaving TVMVp via TEVp. The stationary phase promoter Pfic was selected to regulate mKate2 expression, resulting in strain PS2008. As shown in Supplementary Fig. 8b, c, when this strain entered the stationary phase, mKate2 fluorescence began to increase after 10 h. Moreover, according to the switch time, four programmed switches were constructed (PS2031, PS2041, PS2161, and PS2361) by replacing the repression reporter unit with the action reporter unit (promoter PrpsT was replaced by promoter Pfic, and the tev site was replaced by the tvmv site). The fluorescence intensity began to increase at 20, 27, 28, and 32 h for PS2031, PS2041, PS2161, and PS2361, respectively. These results demonstrated that the programmed switch could also be fine-tuned to obtain a different switch time by fine-tuning the expression of TEVp. Furthermore, to test the universality of the programmed switch, the reporter protein mKate2 was replaced with β-galactosidase, resulting in five strains (PS3008–PS3361). As shown in Fig. 2d, e, both the PS3008 group and the control group began to show enzymatic activity after 12 h (1.7 U/mL), which continued to increase (up to 13.0 U/mL), while the PS3031, PS3041, PS3161, and PS3361 groups did not show enzymatic activity until 24, 30, 30, and 36 h, respectively, suggesting that the results were similar to those of the fluorescence measurements. These results demonstrated that programmed switches were portable and could be used with different genetic components and expression systems. Furthermore, the enzymatic activity of strain PS3361 was the last to appear, indicating that strain PS3361 was the most suitable programmed switch to express CoIM* so that E. coli could grow normally at the growth stage.
Construction and evaluation of a programmed lysis system
A PLS was constructed by integrating a lysis unit (consisting of the reconstructed CoIM*) with a programmed switch (based on PS3361) (Fig. 3a). To demonstrate that the PLS could be effective in different strains of E. coli, it was introduced into seven different E. coli strains and the results are illustrated in Fig. 3c and Supplementary Fig. 9. The mortality ratios of the E. coli strains were <10.0% in the first 30 h (determined by propidium iodide staining), indicating that E. coli could grow normally. Four hours later, the mortality ratios increased to a maximum of 94.1% (JM109) and a minimum of 59.0% (BL21) and was over 86.7% at 38 h. Meanwhile, the protein content in the supernatant of all E. coli strains was higher than 1.62 mg/mL at 38 h, with a value of only 0.72 mg/mL for the control strain, which did not carry the PLS (Fig. 3d). The mortality ratios of the control groups (strains harboring empty vectors) were maintained at <10.0%. These results indicated that the PLS could efficiently lyse E. coli and release cellular inclusions into the broth after 30 h. In addition, the differences in cell growth caused strain-to-strain differences in mortality rate. Firstly, the PLS was based on a stationary phase promoter, which meant that the lysis time of the PLS was related to cell growth. Secondly, the cell growth of strains BL21 and ATCC8739 was slower than that of other strains (Supplementary Fig. 10a). Thus, these two strains were lysed at a slower rate than the other strains. This phenomenon was due to the genotypic differences in different E. coli subtypes, causing differences in growth and gene expression.
a Schematic diagram of constructing programmed lysis system. b Schematic diagram of engineering strains and populations. c Mortality ratios comparison of different strains with PLS (30, 34, and 38 h). Mortality ratios of programmed lysis system (PLS) in different E. coli were measured by PI staining. The fluorescence intensity of living cells was less than 103, and that of dead cells was more than 103. For each sample, at least 20,000 counts were recorded using a 0.5 mL/s flow rate. All data were exported in FCS3 format and processed using Flow Jo software (FlowJo-V10). d Protein content comparison of different strains with/without PLS system. (P values = 0.000608; 0.001569; 0.001427; 0.001045; 0.001948; 0.001042; 0.001351.) Statistical significance was indicated as * P < 0.05, ** for P < 0.01 and *** for P < 0.001, respectively. e The regulation of seeding ratios and programmed lysis system on population. The green fluorescence intensity represented the accumulation of GFP in strain SG and SG5361. The red fluorescence intensity represented the accumulation of mKate2 in SM. f Fluorescence curve of POPS1 and POPS2. It was measured every 2 h and the seeding ratio was 10:1. After 24 h, 0.1 mL of 100 mg/mL yeast extract was added into medium every 4 h. Samples were taken at 22 and 52 h, respectively. Images with green and red fluorescence were also taken by a fluorescence microscope and their colony-forming units (CFUs) were counted to calculate their ratios in the population (the green percentage represented the percentage of green E. coli). Values are shown as mean ± s.d. from three (n = 3) biological replicates. Two-tailed t tests were used to determine statistical significance. Source data are provided as a Source Data file.
To determine whether the PLS could regulate the structure of the engineered populations, three different strains, namely SG5361 (carrying the PLS and constitutively expressing GFP), SG (only constitutively expressing GFP), and SM (only constitutively expressing mKate2), were constructed (Fig. 3b). Then, two populations, POPS1 (consisting of SG5361 and SM) and POPS2 (consisting of SG and SM), were reconstructed and cultured for 60 h to detect their structure by fluorescence. When strains SG5361 and SG were seeded at different ratios (100:1, 10:1, 1:1, 0.1:1, 0:1, and 1:0) with respect to strain SM, POPS1, and POPS2 showed different intensities of red and green fluorescence at 60 h (Fig. 3e). When the seeding ratio in populations POPS2 (SG:SM) and POPS1 (SG5361:SM) was 10:1, strain SG accounted for 84.4% of POPS2, whereas strain SG5361 accounted for only 10.6% (Supplementary Fig. 11) after 60 h. As shown in Fig. 3f, when the seeding ratio was 10:1, the red and green fluorescence of POPS2 remained stable after 22 h. However, the green fluorescence decreased from 3804.22 to 481.69, and the red fluorescence increased from 88.59 to 582.48 between 32 and 44 h in POPS1, and then became stable after 44 h. These results indicated that when the seeding ratio was 10:1, the PLS could change the structure of the populations by lysing the original dominant strains, and this structure could remain stable.
Enhancement of poly(lactate-co-3-hydroxybutyrate) release by a PLS
Poly(lactate-co-3-hydroxybutyrate) (PLH) is a macromolecular polymer that cannot be transported outside the cell after intracellular synthesis29,30. Therefore, an ideal solution would enable the release of PLH into the fermentation broth through cell lysis after fermentation is complete. The PLH production process could be divided into two stages (Fig. 4a): (1) PLH synthesis stage in which the precursor’s lactate and acetyl coenzyme A are produced using the original pathway, and four key enzymes, propionyl-CoA transferase (encoded by pct), β-ketothiolase (encoded by phaA), acetoacetyl-CoA reductase (encoded by phaB), and polyhydroxyalkanoate synthase (encoded by phaC), are overexpressed to produce PLH; (2) PLH release stage in which cells are lysed by the PLS. For this, the PLS was introduced into strain B0032, a PLH-producing strain reconstructed in a previous study29, resulting in the engineered strain B0033 and the population PLH-5 consisting of B0033. As shown in Fig. 4b, 0.45 g/L intracellular PLH was produced by PLH-5 in 56 h, and about 71.21% of the total PLH (0.32 g/L) was released into the fermentation broth. These values were −2% and 82% higher than the corresponding values of PLH-1, respectively. To ascertain the applicability of the PLS to bench-top bioreactors, the performance of PLH-5 (with PLS) in a 5 L fermenter was investigated. We found that 2.26 g/L free PLH was released into the supernatant, which was 335% higher than that of PLH-1 (without PLS) (Supplementary Fig. 12e). The OD600 of PLH-5 reached 40.02 at 48 h, which was only 4% lower than that of PLH-1 (Fig. 4f and Supplementary Fig. 12d). In addition, the yield and productivity of PLH-5 improved by 3.68- and 3.83-fold, respectively, compared with those of PLH-1. Thus, the PLS was an effective tool for releasing intracellular products. After changing the programmed switch, we found that the earlier the PLS was activated, the less PLH was produced (Supplementary Fig. 12b, c). Thus, the optimization of lysis time was necessary.
a The process of producing PLH using programmed lysis system. Stage I was producing PLH and stage II was a programmed lysis system that took effect, lysed E. coli, and released PLH. b PLH production of engineered populations PLH-1 and PLH-5 in shake flasks. PLH-1 consists of B0032, and PLH-5 consists of B0033 which harbors PLS based on PS3361. PLH released into the medium was defined as ‘free PLH’. (P values = 0.655746; 0.000873.) c Fluorescence microscopy was used to detect PLH production in cells stained with Nile red. The PLH granule was stained by Nile red. d Morphological of PLH-1 and PLH-5. Three experiments were repeated independently with similar results. e PI staining to detect cell mortality in fermentation broth at different times. The percentage of fluorescence intensity represented the mortality ratio. The fluorescence intensity of living cells was less than 103, and that of dead cells was more than 103. The fluorescence intensity of living cells was less than 103, and that of dead cells was more than 103. For each sample, at least 20,000 counts were recorded using a 0.5 mL/s flow rate. All data were exported in FCS3 format and processed using Flow Jo software (FlowJo-V10). f pH-stat fed-batch cultures of PLH-5 in a 5-L bioreactor. Values are shown as mean ± s.d. from three biological replicates. Values are shown as mean ± s.d. from three (n = 3) biological replicates. Two-tailed t-tests were used to determine statistical significance. Statistical significance was indicated as * P < 0.05, ** for P < 0.01 and *** for P < 0.001, respectively. Source data are provided as a Source Data file.
The staining of PLH-1 and PLH-5 (48 h) with Nile red revealed some significant intercellular red spots, indicating that PLH was accumulated (Fig. 4c). At 0–48 h, the OD600 of PLH-1 and PLH-5 were the same, but after 48 h, the OD600 of PLH-5 gradually decreased to 2.82 (Supplementary Fig. 12a), indicating that PLH-5 was lysed by the PLS. This result was confirmed by scanning electron microscopy, as shown in Fig. 4d. At 24 h, the morphology of PLH-1 and PLH-5 remained elliptical and intact, and only a few cells were damaged in PLH-5. However, at 56 h, most of the cells were damaged in PLH-5. Similarly, the mortality ratio of PLH-1 and PLH-5 was 27.6% and 92.8%, respectively, as determined by propidium iodide staining at 56 h (Fig. 4e). These results demonstrated that the stages of PLH-5 were switched from PLH synthesis to PLH release through the introduction of the PLS and that intracellular PLH could be successfully released without affecting the total PLH titer.
Enhancement of butyrate production by a PLS
Using fatty acids as a substrate to synthesize butyrate by engineered E. coli causes a huge metabolic burden because the butyrate synthesis pathway contains 11 enzymes31,32. An ideal solution to reduce the metabolic burden would be to divide the metabolic pathway into two strains and achieve microbial cooperation. The first strain (strain I) would contain the substrate utilization pathway that utilizes fatty acids to produce acetate, and the second strain (strain II) would contain the product synthesis pathway that utilizes acetate to synthesis butyrate. Therefore, to efficiently produce butyrate from fatty acids, three stages of fermentation were hypothesized (Fig. 5a): Stage I: strain I acts as the dominant strain to utilize fatty acids to produce acetate; Stage II: strain I can be lysed by the PLS, and strain II starts to grow; and Stage III: strain II becomes the new dominant strain to utilize acetate to synthesize butyrate and achieve temporal cooperation with stage I. To validate this hypothesis, four engineered strains were constructed: BUT001, which contained the complete fatty acid degradation (substrate utilization) pathway and butyrate synthesis (product synthesis) pathway; BUT002, which contained only the fatty acid degradation pathway; BUT003 (strain I), which contained the fatty acid degradation pathway and the PLS; and BUT004 (strain II), which contained only the butyrate synthesis pathway (Fig. 5a). Based on these strains, four engineered populations were constructed, namely POP6 (consisting of strain BUT001), POP7 (consisting of strains BUT002 and BUT004), POP8 (consisting of strains BUT003 and BUT004), and POP9 (consisting of strain BUT004).
a Schematic of the butyrate biosynthetic from fatty acid pathway and process in the fermentation of engineered population POP8. Each gene encodes the following: FadL, long-chain fatty acid outer membrane porin; FadD, fatty acyl-CoA synthetase; FadK, short chain acyl-CoA synthetase; FadE, acyl-CoA dehydrogenase; FadB, α component of the fatty acid oxidation complex; FadJ, α component of the anaerobic fatty acid oxidation complex; FadA, β component of the fatty acid oxidation complex; atoB, acetoacetyl-CoA thiolase; hbd, 3-hydroxybutyryl-CoA dehydrogenase; crt, 3-hydroxybutyryl-CoA dehydratase; ter, trans-enoyl-CoA reductase; tesB, acyl-CoA thioesterase II. b The effect of programmed lysis system on producing butyrate. POP6 consists of strain BUT001; POP7 consists of strain BUT002 and strain BUT004; POP8 consists of strain BUT003 and strain BUT004; POP9 consists of strain BUT004. Plus (+) indicates that the corresponding gene or system was introduced. Minus (−) indicates that the corresponding gene was knocked out. Wave (~) indicates that the corresponding gene remained the same. FadR, GntR family transcriptional regulator, a negative regulator for fad regulon and positive regulator of fabA; PTA, phosphate acetyltransferase; ACK phosphate acetyltransferase; ACS, medium-chain acyl-CoA synthetase. (P values = 0.000425; 0.000168; 0.000057; 0.000185; 0.002112;0.000254.) c pH-stat fed-batch culture of POP8 in a 5-L bioreactor. Stages I, II, III were multiple stages during fermentation. (P value = 0.000002). d PI staining was used to detect mortality ratios in fermentation broth at different times. The percentage of fluorescence intensity represented the mortality ratio. The fluorescence intensity of living cells was less than 103, and that of dead cells was more than 103. For each sample, at least 20,000 counts were recorded using a 0.5 mL/s flow rate. Values are shown as mean ± s.d. from three (n = 3) biological replicates. Two-tailed t tests were used to determine statistical significance. Statistical significance was indicated as *P < 0.05, ** for P < 0.01 and *** for P < 0.001, respectively. Source data are provided as a Source Data file.
The fermentation results are illustrated in Fig. 5b. The butyrate titer of POP8 was 18.66 g/L in the shake flask culture, being 56%, 22%, and 74% higher than the corresponding values in POP6 (8.15 g/L), POP7 (14.64 g/L), and POP9 (4.78 g/L), respectively. The residual fatty acid content in POP8 was 1.57 g/L, being 543%, 39%, and 1089% lower than that of POP6 (10.09 g/L), POP7 (2.18 g/L), and POP9 (18.66 g/L), respectively. Moreover, there was only 0.71 g/L acetate in POP8, which was 104% less than that in POP7 (1.45 g/L). These results demonstrated that POP8 could utilize more fatty acids to produce more butyrate, with fewer intermediate metabolites (acetate) remaining. To identify the applicability of the PLS to bench-top bioreactors, the performance of these four populations in 5 L fermenters was investigated. We found that 41.61 g/L butyrates was produced by POP8 (with PLS) (Fig. 5c and Supplementary Fig. 13d–f), which was 115% higher than that of POP6 (without PLS) (19.31 g/L). At 48 h, the OD600 of POP8 reached 40.82, which was 0.05-fold higher than that of POP6. In addition, the yield and productivity of POP8 increased by 48% and 115% compared with those of POP6, respectively. Thus, temporal cooperation with the PLS was a suitable strategy for improving butyrate production. When E. coli cells were lysed at 12 h, the titer of butyrate was reduced by 66%. This result was attributed to the incomplete fatty acid degradation following premature lysis of BUT003 (Supplementary Fig. 13b, c).
To verify the temporal cooperation that occurred during fermentation, we measured the OD600 and cell mortality at different times. As shown in Supplementary Fig. 13a, at 12 h, the OD600 of POP6 was only 7.15, being 0.93, 0.97, and 1.00 less than that of POP7 (8.08), POP8 (8.12), and POP9 (8.15), respectively. This indicated that excessive enzyme expression induced a heavier burden and slowed the cell growth rate. After 48 h, the OD600 of POP6, POP7, and POP9 remained above 8.18, but that of POP8 decreased at 60 h (6.01) and then recovered to above 8.25, indicating that cell lysis occurred from 48 to 60 h and that the growth resumed after 60 h. As shown in Fig. 5d, <9.8% of the cells were damaged in all populations (POP6, POP7, POP8, and POP9) before 36 h. Twenty-four hours later, cell mortality increased to 80.1% in POP8, but that of POP6, POP7, and POP9 remained at less than 21.1%. This result indicated that the PLS successfully lysed strain BUT003 in POP8. At 96 h, the cell mortality of POP8 decreased to 28.4%, indicating that BUT004 grew and became the dominant strain. In summary, the PLS serves as an advantageous tool for temporal cooperation by changing the phenotypic structure of populations to improve the performance of microbial cell factories.
