The probiotic strain Escherichia coli Nissle 1917 (EcN), a potential member of tumor-targeting bacteria, shows great promise for cancer treatment. By leveraging engineered EcN, we can design a bacteria-assisted, tumor-targeted therapy for the biosynthesis and targeted delivery of small-molecule anticancer agents. In this study, we aimed to use EcN as a base for synthesizing Romidepsin (FK228), an FDA-approved drug originally made by Chromobacterium violaceum No. 96. Through gene cluster reconstruction, promoter optimization, and genome modification, we created FK228-producing strains to boost anticancer efficacy. The engineered strain achieved a maximum in vitro yield of 1.5 mg/L. In 4T1 tumor-bearing BALB/c mouse xenograft models, six recombinant strains outperformed the wild-type EcN. Proteome showed that inflammatory response induced by EcN combined with intratumoral FK228 production improved treatment results. Also, targeted synthesis reduced FK228's cardiotoxicity and mortality. Engineered EcN enables drug biosynthesis and precise delivery, offering powerful anticancer activity.

Heterogeneous biosynthesis of FK228 in EcN

A previous study demonstrated that a 12-gene dep gene cluster encodes a hybrid modular NRPS-PKS pathway [18,22–24]. The original pathway comprises two PKS modules on the DepBC enzymes, which are deficient in a functional acyltransferase (AT) domain. Additionally, no evident AT-encoding gene is present within or near the gene cluster. It has been reported that a gene encoding a putative malonyl coenzyme A AT component of the fatty acid synthase complex was determined to be essential for FK228 biosynthesis in E. coli cells [24]. Furthermore, a gene (sfp) encoding a putative Sfp-type phosphopantetheinyltransferase was also found to be implicated in the heterogeneous expression of FK228 in E. coli cells.

To explore the FK228 biosynthesis in EcN, firstly, we constructed a plasmid based on a p15A vector and that contained a dep gene cluster from C. violaceum [18] under the control of a BAD promoter. However, no FK228 was detected from the fermentation of E. coli cells containing this plasmid because of the lack of a functional AT-encoding gene and a PPTase-encoding gene. As reported that a component of the fatty acid synthase takes responsibility of AT domain function and Sfp-type PPTase is necessary for FK228 synthesis in E. coli cells [24], the genes that encoded fabD1 and Sfp-type PPtase from C. violaceum were chosen to follow the regulation of constitutive promoter (Anderson promoters) (S1 Fig) to assist the FK228 biosynthesis in EcN. Considering that the synthesis of FK228 may influence the metabolism of the host cells, an inducible promoter (PBAD promoter, tetracycline inducible promoter, or PfnrS promoter) was applied to regulate the dep modules that were responsible for the most of the reactions involved in FK228 biosynthetic pathways. With codon optimization of all biosynthetic genes and the debugging of promoters, a new plasmid based on a p15A vector and that encoded dep modules A-J under the control of a BAD promoter, and that encoded fabD1 and Sfp-type PPtase, and other dep modules K-N under the control of promoter BBa-J23115 was reconstituted, which realized heterogeneous expression of FK228 in E. coli cells, especially in EcN (Figs 1 and 3A). As expected, the original FK228 biosynthetic genes from C. violaceum did not generate FK228 in EcN (S2 Fig). In addition, the yield of the E. coli BL21 that contains FK228 biosynthetic genes with codon optimization was higher than that without codon optimization (S3 Fig).

Reconstitution and generation of FK228-producing EcN strains

As we realized the biosynthesis of FK228 in EcN by use of plasmid 23 (Figs 2 and S4), we reconstructed assistant modules and adjusted promoters based on the plasmid 23 to generate more optimized biosynthetic pathways and corresponding FK228-producing EcN strains with expected high yield (Figs 2 and S4 and S1 Table). A series of optimized EcN strains including EcN-40, EcN (ΔaraC)-23, EcN-24, EcN-44, and EcN-25 were determined to heterogeneous express FK228 successfully (Fig 3) with higher yield under aerobic or anaerobic conditions (Fig 4A). The FK228 production of EcN-40 was more than that of others, up to 1.46 mg/L. Subsequently, the yield was 0.714 mg/L (EcN-44), 0.69 mg/L (EcN-24), 0.53 mg/L (EcN (ΔaraC)-23), 0.33 mg/L (EcN-23), and 0.001199 mg/L (EcN-25 cultivated under anaerobic condition) at the end of a culture with OD₆₀₀ of 3. What’s more, FK228 can be diffused from the EcN cells (Figs 4B and S5), which indicates that FK228 would diffuse to tumor microenvironment when administration of FK228-producing EcN strain. In subsequent investigations, we observed that the FK228-producing EcN strains reached the peak FK228 yield within 24 hours during in vitro fermentation. However, supplementing the inducer at 24 hours further increased FK228 production—this indicates that the peak yield within 24 hours is primarily due to inducer depletion, rather than the death of EcN (S5G Fig). Furthermore, at 6 hours post-induction, a considerable amount of FK228 was already present in the culture medium, reaching approximately half of the peak concentration in each group. Notably, the EcN strains had not yet entered the stationary growth phase at this time, with intact cellular morphology and no significant lysis observed (S5 Fig). These findings confirm that FK228 is released from EcN. Given that FK228 is a small-molecule compound, it is highly plausible that its release occurs via passive diffusion across the bacterial cell membrane. Furthermore, the growth curves of these strains under aerobic conditions with or without inducers have been determined and are presented in S6 Fig. It was observed that the engineered strains grew somewhat more slowly than EcN. Following induced expression, the growth rate of the EcN strains was further restricted, albeit not significantly. Only the OD₆₀₀ of EcN-23 decreased significantly at 4 hours post-induction, with a gradual recovery observed at 7 hours (S6 Fig). This indicates that the expression of FK228 imposes a relatively minor metabolic burden. We also captured photographs of both wild-type EcN and engineered EcN strains (S5H Fig). These images revealed that harboring the FK228 expression plasmid did not exert a significant impact on the morphology of EcN, with the engineered EcN retaining its rod-shaped structure.

The reconstitution and generation of FK228-producing EcN strains were as follows. Since tetracycline (Tet) inducible expression system [25] and the anaerobic-inducible promoter [8] have been reported to control the therapeutic genes expression in EcN, plasmid 24 and 25 were generated based on plasmid 23 by replacing the BAD promoter with Tet promoter and PfnrS promoter, respectively. Recombinant EcN strains were then generated using these plasmids. The FK228 production was analyzed by high-performance liquid chromatography−high resolution mass spectrometry (HPLC-HRMS). HPLC-HRMS spectra demonstrated that the recombinant EcN strains containing plasmid 24 (EcN-24) (Fig 3B) or plasmid 25 (EcN-25) (Fig 3C) could heterogeneous express FK228. The central peak of FK228 from the fermentation was observed compared with the FK228 standard sample. Both plasmid 23 and 24 are under the control of inducible promoter, the production of FK228 was increased with the increase of L-arabinose addition (Fig 4C) or anhydrotetracycline (AHT) addition (Fig 4D) in EcN.

Then we were concerned that the BAD operon within the EcN genome might interfere with the BAD promoter which governs the biosynthesis of FK228. The araC gene and the araCBAD gene in the genome were knocked out to generate EcN (ΔaraC) strain and EcN (ΔaraCBAD) strain as new hosts for FK228 biosynthesis. We found that EcN (ΔaraC) strain containing plasmid 23 could produce much more FK228 molecules than that of EcN (ΔaraCBAD) strain (S7 Fig). Further, we determined the yields of these strains. We found that the yield of EcN (ΔaraC) strain containing plasmid 23 was higher than that of EcN-23 (Fig 4A and 4B). EcN (ΔaraC)-23 could produce much more FK228 compared with EcN-23 when the same dosage of L-arabinose was used for startup (Fig 4C and 4D). The deletion of araC in the genome may disburden the excessive free araC domain binding to the operator to repress the downstream transcription. Thus, L-arabinose bonded to the araC domain to form complex to regulate FK228 artificial operon.

In addition, restriction modification systems (R-M) in bacterial cells are able to cut the exogenous DNA to protect bacteria themselves. It is rather challenging to transfer plasmids, particularly large ones, into EcN cells. Possibly, the restriction modification systems in EcN impede the survival of exogenous plasmids. Consequently, the relevant genes located in the EcN genome were successively knocked out to produce the recombinant strains depicted in S8A Fig. We found that the knockout of R-M-related genes can slightly increase the plasmid transformation. However, the results of bacterial colonization in tumor (S8B Fig) showed that the recombinant strains’ ability of colonizing in tumors declined compared with the original EcN. Thus, the original EcN is still the preferred host to accommodate FK228 pathways because of the best performance in colonization in tumor.

It is reported that the deduced product of depM is an aminotransferase. It is presumably responsible for catalyzing the removal of an amine group from a cysteinyl-S-peptidyl carrier protein (PCP) intermediate in a trans configuration, thereby forming 4-mercaptobutanyl-S-PCP in the initiation module of the FK228 biosynthetic pathway in C. violaceum [18]. The depN is regarded as a pseudogene; however, it contains regulatory elements upstream of depA [22]. The deficiency of depM or depN could decrease the production of FK228 in C. violaceum [22]. On the contrary, the deficiency of the putative genes depK or depL had no obvious influence on FK228 production in C. violaceum [22]. In our study, the depK-N genes were reconstructed behind sfp gene under the constitutive promoter as the assistant modules for FK228 biosynthesis in EcN. To investigate the effect of the depLNMK modules on FK228 biosynthesis in EcN, the gene depK-N was knocked out receptively based on plasmid 23 to generate plasmid 43 (ΔdepLNMK), 42 (ΔdepLNK), 41 (ΔdepLK), 40 (ΔdepK), and 39 (ΔdepL) and the corresponding recombinant EcN strains in order (S1 Table). As the results showed in Fig 4E, the deletion of depLNMK or depLNK could decrease the production of FK228 compared with EcN-23-4 strain. However, the deficiency of depLK could slightly increase the yield of FK228 compared with EcN-23-4 strain. Hence, the strain EcN-39 (ΔdepL based on plasmid 23) and EcN-40 (ΔdepK based on plasmid 23) were generated. The FK228 production displayed that the deficiency of depK could significantly increase the level of FK228 production in EcN. However, the high level FK228 production may slow down the growth of host to some extent (S6 Fig). The EcN-44 strain biosynthesized the highest level of FK228. Otherwise, the deletion of depL had no effect on the production of FK228. Since the deletion of depK plays a role in the level of FK228 production, plasmid 44 was constructed based on plasmid 24 by knockout of depK. The corresponding recombinant EcN strain containing plasmid 44 (EcN-44) becomes the one with the secondary level of FK228 production (Fig 4A).

The depR encodes an LysR-type transcriptional activator, which acts as a pathway regulatory gene governing FK228 biosynthesis in C. violaceum [22,26]. The deletion of depR abolished FK228 production in C. violaceum [22]. However, the deficiency of depR had no obvious effect on FK228 production in EcN in our study. The reason may be that a certain homolog of the LysR family of transcriptional regulator in EcN takes the place of depR in FK228 biosynthesis.

As mentioned, fabD1 module takes the place of a functional AT domain involved in FK228 synthesis in E. coli. We found that the original fabD1 had a negative effect on growth of E. coli. The optimized-code fabD1 gene was better. Meanwhile, a gene circuit which sfp was followed by fabD1 was generated (plasmid 62). However, the yield of FK228 of EcN-62 had no obvious improvement compared with that of the former one (Fig 4E). Considering that Sfp-type PPTase plays a role in the FK228 biosynthesis in E. coli, a series of Sfp-type PPTase including sfp from C. violaceum, MtaA from Myxobacterium Stigmatella aurantiaca, sfp from Bacillus subtilis, Pcps from Pseudomonas aeruginosa PAO1, SpiPcps from Pseudomonas sp. Q71576 was chosen to explore the effect on FK228 yield in EcN. The corresponding plasmids and strains were generated (S1 Table). However, the levels of FK228 production indicated that the optimized-code sfp gene from the original C. violaceum seemed the best for FK228 biosynthesis in EcN (Fig 4E).

The assistant modules including fabD1, sfp, and depLNMK have been reconstructed for higher FK228 production. We further modified the constitutive promoter that controls the assistant modules based on the depK deletion FK228 biosynthetic pathway. A series of Anderson promoters with more strength took the place of the BBa-J23115, which controls the transcription of sfp-PPTase, fabD1, and depLNM (S1 Table). Although these plasmids have been generated successfully, they failed to work in EcN. Thus, these plasmids were transferred into E. coli BL21 for FK228 fermentation. However, these plasmids with more strength promoters led to less FK228 compared with the original plasmid 23 in E. coli BL21 (S9 Fig). Maybe the strong promoter leading to high transcription of assistant modules disturbs the balance of original growth and metabolism, which resulted in less final products.

The maintenance of the plasmids in bacterial cells without antibiotics selection is a challenge in real-world applications. Antibiotics are not appropriate for clinical applications due to multiple reasons. One significant concern is the horizontal gene transfer of resistance genes, which can lead to the spread of antibiotic resistance among different organisms. Additionally, the use of antibiotics may disrupt the native microbiota, the natural community of microorganisms residing in the body, which can have various consequences on health and physiological balance [27].Except for antibiotics selection, several mechanisms for stability of plasmid persistence in more complex environments have been developed. The toxin-antitoxin systems have been used for ensuring plasmid maintenance for a certain period of time [28,29]. An the toxin–antitoxin system Axe-Txe from Enterococcus faecium has been utilized to stabilize plasmids in the absence of antibiotics in vitro and in vivo [27]. In our study, the Axe-Txe system and their homologues yefM-yoeB from E. coli were used for the stability of FK228 plasmids in EcN. However, the plasmids carrying yefM-yoeB were failed to be constructed. The plasmids carrying axe-txe gene or codon-optimized axe-txe gene and the corresponding recombinant strains were constructed successfully. However, the yield of these strains (EcN-34-3, EcN-35-1) showed that the addition of TA system led to a massive reduction of FK228 in EcN. Further investigation should be taken for explanation. Next, we investigated the plasmids maintenance in mouse model (S10 Fig). The results displayed that the codon-optimized AT plasmids led to higher plasmids maintenance compared with the original AT module group. However, the plasmid maintenance rate of the plasmids carrying the AT modules was similar to the one without AT (EcN-23). Considering that the ones with AT modules had low FK228 production in vitro, these strains were not tested in anticancer activities in mouse models.

The transcriptomic differences between engineered EcN strains and EcN before and after induction with L-arabinose

We further test the transcriptional profiles of the engineered FK228-producing strains under the control of L-arabinose to investigate the genetic transcription of FK228 biosynthetic gene cluster in EcN. The transcriptional profile data in Fig 5 revealed that after 24 hours of L-ara induction, a significant amount of dep gene cluster mRNA remained in EcN-40, while the residual amounts in EcN-23 and EcN ΔaraC-23 were relatively lower. It is hypothesized here that, depK might influence the degradation of dep gene cluster mRNA or inhibit the utilization of L-ara and gene transcription as a zinc finger protein. Consequently, the yield of FK228 is upregulated after the knockout of depK. Meanwhile, the data also indicated that the transcriptional level was markedly attenuated at depD and depE, suggesting that we could further enhance the synthesis of FK228 by adding an additional promoter at these sites. Furthermore, we observed that, in contrast to the decreasing trend of transcriptional levels that occur with the extension of the gene cluster, the transcriptional levels of the depI and depJ genes exhibited an upward trend instead. So, we guess that depI and depJ possess additional transcriptional pathways. This is also in line with the speculation by V. Y. and others [22].

Besides, considering that the biosynthesis of colibactin, as a native secondary metabolite in EcN, may disturb the heterogeneous expression of FK228 in EcN, the genetic transcription of colibactin gene cluster located in EcN genome were analyzed too. Colibactin, a genotoxic metabolite produced by the majority of intestinal bacteria, is capable of inducing damage to cellular DNA. Simultaneously, it can also trigger prophage induction and potentially affect human diseases [30]. EcN can also express colibactin, but due to its bacterial adhesin FimH mutation, it does not exhibit pathogenicity because of colibactin [31]. However, E. L. and others have recently discovered that colibactin can induce DNA damage in bacteria hundreds of micrometers away [32]. Therefore, we cannot completely disregard the impact of colibactin. According to transcriptional profiles, we have discovered that the expression of colibactin in EcN strains carrying the FK228 expression plasmid, especially after inducible expression, is suppressed. Moreover, the higher the expression level of FK228, the more pronounced the inhibitory effect. We hypothesize that this might be attributed to the competitive action of FK228 (Fig 5D–5G). The genes HtpG, DnaK, UvrY, MarA, AcrA, AcrB, and Fur promote or directly participate in the synthesis of colibactin and are downregulated after the induction of FK228 expression. The CrsA gene inhibits the synthesis of colibactin and is significantly upregulated after the induction of FK228 expression (Fig 5H–5K) [33]. This further demonstrates that the expression of the FK228 gene cluster suppresses the synthesis of colibactin. This can reduce the impact of colibactin on the entire process.

All in all, the deletion of depK from the FK228 biosynthetic pathway make great contribution to the heterogeneous expression of FK228 in EcN. The knockout of araC from the EcN genome cooperating with BAD promoter also plays a role in high production of FK228. Considering that tumor microenvironment is anaerobic, the anaerobic-inducible promoter has potential to behave better to control FK228 biosynthesis in tumor microenvironment. Thus, the optimized strain with a high yield of FK228 including EcN-40, EcN (ΔaraC)-23, EcN-23, EcN-44, and EcN-24, and the EcN-25 strain under the control of low oxygen were chosen for further antitumor investigation.

The activity of FK228-producing EcN strains

Fermentation broth of FK228-producing EcN strain under the control of L-arabinose was tested with mouse breast cancer cells 4T1 and mouse mammary epithelial cells HC11, and their cell viabilities are shown in S11 Fig The cell viability revealed that the fermentation of broth from the FK228-producing EcN strain (approximately 0.16 μM FK228) had a significantly inhibition on 4T1 cells compared with that of fermentation of broth from the original EcN strain, and had an even more inhibition than that of positive drugs FK228 (2.5 μM). Meanwhile, the fermentation of the FK228-producing EcN strain had a similar inhibition on HC11 compared with positive drugs FK228. The fermentation of the original EcN also had a slight inhibition on HC11 cell viability. The IC₅₀ values of FK228 on cells were measured and shown in S12 Fig. This result indicated that the FK228-producing EcN had a better inhibition on cell viability of cancer cells. However, the in vivo tumor microenvironment is more complex. Tests in xenograft mouse models are necessary to see if these engineered strains could have anticancer activities.

In vivo anticancer activity of FK228-producing EcN strains

The FK228-producing EcN strains EcN-23, EcN-40, EcN (ΔaraC)-23, EcN-24 and EcN-44 were chosen to be investigated for in vivo anticancer activities using a xenograft mouse model due to their high levels of FK228 production. EcN-25 was also selected because continued hypoxia in tumor microenvironment may promote this strain to produce more FK228 for a long time. The FK228 detection results of the FK228-producing EcN strains under the control of L-arabinose are shown in Fig 6. The FK228 signal was detected in the tumor tissues of the FK228-positive group subsequent to either intratumoral injection or intravenous injection. Then, the mice bearing 4T1 tumors were intravenously injected with a suspension of EcN-23, EcN-40, or EcN (ΔaraC)-23. The inducer L-arabinose was intraperitoneally injected 48 hours after the administration of these strains. The results showed that FK228 was detected from the tumor tissues 6 hours and 24 hours post administration of inducer. And the highest FK228 was detected from the EcN-40 group, even more than that of FK228-positive group that mice were intraperitoneally injected with FK228 solution. The level of FK228 detection from the EcN-23 group was in the next place. The lowest one was EcN (ΔaraC)-23 group. Even if the yield of FK228 from the fermentation of EcN (ΔaraC)-23 was more than that of EcN-23. The results in Fig 6 revealed that the FK228 can be detected in the tumor tissues dissociated from the EcN-24 and EcN-44 groups 6 and 24 hours post administration of inducer anhydrotetracyclin. The FK228 level of EcN-44 was also the highest group 6 hours post administration. The FK228 was also detected from the EcN-25 group 1–4 days post intravenous administration of engineered bacteria. The mice of EcN-25 group were not to be administrated by inducers because the PfnrS promoter senses the oxygen of environment to regulate the downstream pathway. These results indicated that all engineered EcN strains can produce FK228 in tumor environment in situ under the control of a small molecule-inducer. We realized tumor-targeting bacteria EcN heterogeneous express and deliver FK228 in tumor environment after strains colonized in tumors. Additionally, we detected serum FK228 levels at various time points after intraperitoneal injection of either inducers or FK228 itself to investigate its metabolism in the blood of each treatment group. Measurements were taken within 24 hours post-injection. Results showed that FK228 persisted in the blood for 24 hours after injection, with a peak concentration of 0.015 μg/mL at 6 hours. In contrast, residual FK228 in the blood of the bacterial treatment group remained extremely low throughout the 24-hour period. Instead, FK228 tended to accumulate in tumors in these groups, further highlight the advantages of tumor colonization by these strains, which enables FK228 to be produced and accumulated in tumor tissues (Fig 6D).

Meanwhile, the anticancer activities of these strains were investigated in 4T1 tumor-bearing mouse model. The original EcN was taken as a control. The FK228 group was a positive control. The group without treatment was a blank control. The results are displayed in Figs 7–11. Fig 7 shows the antitumor activities of engineered strains under the control of L-arabinose. The strains EcN-40, EcN-23, and EcN (ΔaraC)-23 showed significant inhibition on tumor growth compared with the blank group. Compared with the original EcN, the engineered strains showed much more inhibition of tumor growth. Moreover, the treatment of EcN-23 resulted in a similar inhibition of tumor growth compared with FK228 group. So did EcN-40 group. Figs 8 and 9 displayed the antitumor activities of engineered strains under the control of anhydrotetracycline. The treatment of EcN-24 (Fig 8) and EcN-44 (Fig 9) resulted in the inhibition of tumor growth, which was similar to positive FK228 group. EcN-25 strains including clone 8, clone 11 and clone 16 also showed significant anticancer activities compared with blank group (Fig 10). All EcN-25 strains performed more inhibition on tumor growth beyond the original EcN group. In addition, some of mice in the FK228 treatment group died approaching endpoint. The administration of engineered EcN strains resulted in less mortality of mice in experiments. The reason may be that the tumor-targeting EcN producing and delivering FK228 in tumor microenvironment could reduce the side effects of FK228. The tumor colonization of engineered EcN ensured the concentration and accumulation of FK228 in tumor microenvironment, which reduced the side effects of FK228 compared with direct intravenous injection of FK228. At the endpoint, the bacterial colonization in the tumors has been determined and shown in S13 Fig. Most engineered strain groups have CFU counts in tumors that are similar to or lower than those in the EcN group. Only EcN-23 has colonized tumors to a greater extent compared to EcN. Besides, there is a possibility that the mice will die within the FK228 treatment group throughout the approximately three-week FK228 treatment course (S14 Fig). In contrast, the mice in the engineered EcN groups will survive during the roughly three-week treatment period. The results suggested that FK228 has side effects in cancer treatment.

In addition, we selected EcN-23, EcN-40, EcN-24, EcN-44, and EcN-25, which exhibited relatively good performance, and re-conducted animal experiments to evaluate their anti-tumor activity. In this experiment, the injection concentration of FK228 was lowered, and the injection frequency was reduced. Fortunately, no mouse deaths occurred in the FK228 group during the treatment period. However, as anticipated, the therapeutic efficacy of the FK228-positive group decreased, reaching a level comparable to that of the EcN group. The EcN-44, EcN-23, and EcN-24 groups showed relatively superior tumor growth inhibition effects compared with the EcN group (Fig 11). At the end of the experiment, the mice’s tumors were dissected and photographed (S15 and S16 Figs). In general, these FK228-producing strains demonstrate potential for anti-tumor activity.

The proteomic differences of tumors after treatment with different treatment groups

We further investigate the proteome of tumor tissues that are treated with FK228-producing EcN strains. In the screening of significantly differential proteins, with the criteria of fold change (FC) > 1.5 times (up-regulated more than 1.5 times or down-regulated less than 0.67 times) and P value < 0.05 (T test or other), the number of up-regulated and down-regulated proteins between the comparison groups was obtained (Fig 12A and 12B).

Perform Gene Ontology (GO) terms on differentially expressed genes to classify their functions (Fig 12C). The results of the GO analysis show that engineered EcN strains can initiate and enhance immune responses through antigen presentation, complement activation, response to interleukin, and other means, thereby exerting a killing effect on tumor cells. Additionally, they can induce apoptosis and ferroptosis in cells via oxidative stress [34,35]. Simultaneously, it can also inhibit angiogenesis and tumor development. However, the inflammatory reaction caused during this process may also affect cancer treatment and development. We can also observe that the direct use of FK228 can down-regulate centromere-associated proteins to inhibit cell proliferation and promote cell apoptosis by regulating ubiquitination modification, especially the expression of cullin family proteins [36]. Nevertheless, the direct use of FK228 does have an impact on cardiomyocytes and myocardial tissue [37], which does not occur in the EcN treatment group (S17 Fig). The GO analysis of engineered EcN strains in comparison with the EcN group and the FK228 group reveals that: in contrast to EcN, engineered EcN strains possess not only stronger immunostimulatory and oxidative stress-stimulating effects but also the capabilities of modulating ubiquitination modification and promoting histone acetyltransferase activity. Compared with FK228, engineered EcN strains can exert additional immunostimulatory and oxidative stress-stimulating effects and promote the function of p53, thereby achieving a better cancer treatment outcome (S17 Fig). Therefore, engineered EcN strains not only combine the tumor treatment capabilities of EcN and FK228 but also surpass them in many aspects, while reducing the side effects such as cardiotoxicity produced by FK228 during the treatment process.

The volcano plot displayed the abnormally expressed genes in the two datasets. We selected the top 10 proteins with the most significant up-regulation and down-regulation in each group compared with the Blank group for the study (Fig 12C and S2 Table). We found that the tumor suppressor genes Ptrh1, Rpl36a-ps1, Rps25, as well as immunoglobulin-related genes including Ighg3, Igkv9-120, Igkc, Ighm, Ighg2b, and Iglv2, showed a significant increase in expression in all EcN groups or engineered EcN strains treatment groups [38–41], while there was no significant change in the FK228 group. The oncogenes Fn1, Fbn1 Gcsh, Ptma, Mecp2, Fxn, Ctsh, Got2, Csnk1g2, Lrp6, Prss23, Mylk, and Adamtsl4 exhibited a significant decrease in expression in all EcN groups or Engineered EcN strains treatment groups [42–54], with no obvious change in the FK228 group. This indicates that EcN or engineered EcN strains, rather than FK228, exert an anti-cancer effect by regulating these proteins.

The genes Ppp3cb, Phf11, Sepsecs, and Uck1 are among the most significantly up-regulated tumor suppressor genes in the FK228 group [55–58]. The genes Gpr84, Faah, Mettl2, Taf12, Cd163, and Cyth1 are the most significantly down-regulated oncogenes in the FK228 group [59–64]. Among them, Ppp3cb contributes to Herceptin resistance, and the combination of Cd163 and GPR84 antagonists with anti-PD-1 antibodies has enhanced the anti-tumor response. These existing research results suggest to us that these engineered EcN strains may achieve a more excellent therapeutic effect in combination with related drugs [55,63,65].

The significantly differentially expressed proteins in the direct comparison between the engineered EcN strains group and both the EcN group and the FK228 group showed that compared with the EcN treatment group, the oncogene and drug resistance-related genes Anxa3, Kif14, and Ecd in the engineered EcN strains group were downregulated [66–68]. In comparison with the FK228 treatment group, the tumor suppressor genes Siglec12, Stfa2, and immunoglobulin-related genes in the engineered EcN strains group were upregulated [69,70]. Meanwhile, compared with the other two groups, the oncogene and drug resistance-related genes B2m, Pf4, and Rps25 in the engineered EcN strains group all exhibited downregulated expression [41,71,72], while the tumor suppressor genes Csnk1g2, Hfe, Col4a1, Ptma, Fbn1, Taf1a, Ahsa2, Adap2, Zfp871, and Dhx34 all showed upregulated expression (Fig 12C and S2 Table) [43,45,50,73–79]. This indicates that the engineered EcN strains may inhibit the development of cancer by more effectively regulating these genes. These existing research results also provide some inspiration for our subsequent combination drug therapy.

Strains, culture conditions, and plasmids

S1 and S3 Tables detail the bacterial strains, plasmids, and primers employed in this study, with gene synthesis and codon optimization performed by Sangon Biotech. The Escherichia coli (E. coli) strains were cultivated in Luria-Bertani (LB) medium at 37°C within a thermostatic oscillator operating at 950 rpm. The liquid LB medium comprised 5 g/L yeast extract, 10 g/L pancreatin, and 0.1% or 1% sodium chloride. Solid LB medium was formulated by supplementing liquid LB medium with 1.2% agar. The sodium chloride was sourced from Genview, antibiotics, and the remaining components were provided by Thermo Fisher Invitrogen.

Construction of plasmid and recombinant strains

The tool plasmid pSC101-BAD-ccdA-Rha-Gbaa-tet was introduced into EcN through electroporation. Subsequently, the gene OxyR and genes associated with the arabinose operon were substituted with spectinomycin and chloramphenicol resistance genes flanked by lox66 and lox71 sites via recombination. After screening for the correct single clones using colony PCR, the plasmid pSC101-BAD-ccdA-Rha-Gbaa-tet was removed by culturing at 37°C. Similarly, the plasmid RK2-BAD-Cre-SacB was transformed into the modified EcN. Upon induction of Cre enzyme expression, recombination was triggered by the recognition of the two lox sites, leading to the deletion of the resistance selection marker between them. The single clones successfully removing the selection marker were identified by colony PCR. Subsequently, the obtained correct strain was serially subcultured three times in a sucrose-containing medium, streaked in three sectors on LB plates, and ultimately, the target gene-knocked-out ECN without plasmid was obtained by double-plate streaking.

The original core genes and the codon-optimized core genes required for FK228 biosynthesis were partially synthesized by Genewiz, Sangon Biotech, and Beijing Genomics institution. The primers for FK228 synthesis and the synthesized oligonucleotides are listed in S3 and S4 Tables, respectively. We employed the engineered E. coli strain GB05dir-gyrA462 to conduct multipiece DNA assembly, thereby constructing plasmids 01-p15A-cm-FK228CODON-F12-3-neoccdB, 02-BR322-cmccdB-FK228codon-F47, and 03-BR322-cmccdB-FK228codon-F811. Subsequently, we digested them with the SnaI enzyme and assembled them into the plasmid p15A-cm-pBAD-FK228CODON using the ExoCET technology. After digesting the plasmid p15A-cm-pBAD-FK228CODON with ScaI and AatII, the digested products were combined with the PCR product kam-oriT through linear plus linear homologous recombination (LLHR) to form the plasmid 23-p15A-km-pBAD-FK228. Other plasmids were constructed based on plasmid 23 by linear plus circular homologous recombination (LCHR), LLHR, and RedEx methods. amp-ccdB, cm-ccdB, and spect-ccdB served as a selection-counterselection marker in this process. The accurate single colonies were identified through restriction enzyme digestion verification and then subjected to sequencing. EcN and engineered EcN can transform with constructed plasmid DNA by means of conjugation with the Escherichia coli donor strain WM3064.

Fermentation and high-performance liquid chromatography-mass spectrometry analysis of the metabolites derived from Engineered EcN

Engineered EcN strains were inoculated from a plate into 50 mL LB in 250 mL flasks and incubated with shaking at 200 rpm at 30°C for about 3 hours. When it grew to the logarithmic growth phase, the corresponding inducer was added. XAD16 macroporous adsorption resin was added at 72 hours of cultivation, and extraction was carried out with methanol at the end of a culture with OD₆₀₀ of 3 (96 hours). With the exception of the EcN strain harboring the fnrs promoter, which was subjected to anaerobic fermentation throughout the entire process, all other strains were cultured under aerobic fermentation conditions. The extract was filtered through a 0.22 μm filter and then subjected to HPLC–MS analysis.

High-resolution mass spectrometry analysis was performed on high-Resolution Q-TOF mass spectrometry (impactHD). The mobile phases consisted of H₂O with 0.1% (v/v) formic acid (solvent A) and acetonitrile (ACN) with 0.1% (v/v) formic acid (solvent B). The ion source was ESI and the flow rate was 0.3 ml/min. The gradient solvent conditions were as follows: maintain at 5% B for 3 min, increase from 5% B to 95% within 15 min, maintain at 95% B for 4 min, and then maintain 5% B for 3 min. The quantification of compounds was carried out on the UPLC-Triple Quadrupole Mass Spectrometer System (Triple Quad 5500+ QTRAP+X-5). The gradient solvent conditions were: maintain at 5% B for 1 min, increase from 5% B to 98% within 4 min, maintain at 98% B for 2 min, and then maintain 5% B for 3 min. The remaining conditions were the same as those for high-resolution mass spectrometry analysis.

Cell culture and mouse model

The cells we used were mouse mammary epithelial cells HC11 (purchased from DINGGUO CHANGSHENG) and mouse breast cancer cells 4T1 (Lab stock). The reagents used included Gibco’s RPMI 1640 medium, Pen-Strep antibiotic, Trypsin-EDTA (0.25%), Fetal Bovine Serum (FBS), and PBS from Procell. The RPMI 1640 used was supplemented with 1% Pen-Strep antibiotic and 10% FBS. All cell lines were cultured and maintained in a humidified incubator at 37°C with 5% CO₂.

Cell viability assay: Firstly, cells were trypsinized to obtain cell suspensions, with 1 × 10⁴ cells being carefully seeded into each well of a 96-well plate. The plate was then incubated overnight to allow cells to adhere firmly. Meanwhile, the fermentation broth was filtered under sterile conditions and subsequently introduced into the 96-well plate containing the adhered cells. FK228 served as the positive control throughout the experiment. Following a 24-hour incubation period, the culture medium was gently aspirated, and the cells were rinsed 2–3 times using PBS. Next, a CCK-8 solution was added. After an incubation of 3–4 hours, the absorbance values were measured to calculate the cell survival rate.

All animal studies were conducted in accordance with the guidelines of the Ethics Committee and the IACUC protocols of the School of Life Sciences, Shandong University. All animal experiments were reviewed and approved by the IACUC of Shandong University (Approval No.: SYDWLL-2022-023). The experimental animals employed in this study were 8-week-old female BALB/c mice procured from Shandong Jinan Pengyue Animal Co.. Immediately upon receipt, the mice were housed in a controlled environment with a precisely regulated 12-hour light/dark cycle and furnished with sterile bedding, feed, and water for a period of one week to allow them to acclimate and reach a stable physiological state. Subsequently, a mouse allograft tumor model was established by subcutaneously injecting 5 × 10⁵ 4T1 murine breast cancer cells into the right forelimb of each mouse. When the tumors had grown to a volume ranging from 300 to 600 mm³, 100 μL of EcN strains culture (with an OD₆₀₀ = 0.2, approximately 1.2 × 10⁷ CFU) (S16 Fig) containing was administered to the mice via tail vein injection. Forty-eight hours after the injection of the engineered bacteria, the 100 μL inducer, PBS, or 0.3 mg/mL FK228 was intraperitoneally injected, and then the injection was repeated every 48 hours [88,89]. Tumor volume = length × width²/2. Upon completion of the experiment, the mice were anesthetized with isoflurane prior to euthanasia via cervical dislocation.

Antitumor activity and FK228 detection in tumor

The day of injecting the bacterial solution was designated as day zero. The body weight of the mice as well as the length and width of the tumors were measured every 48 hours. On the last day of the experiment, the tumors were removed, washed with PBS, dried with sterile paper to remove surface moisture, and then weighed. After homogenizing the tumors, they were diluted and spread for counting colony-forming units (CFU). For the detection of FK228 in the tumors, the tumors were removed 6 hours and 24 hours after injecting the inducer (for the intratumoral injection and intraperitoneal injection FK228 groups, the tumors were removed 1 hour and 2 hours after injection). After homogenizing the tumors, extraction was carried out. The extract was used for FK228 quantification by the UPLC-Triple Quadrupole Mass Spectrometer System. When the tumor volume reached 1,500 cubic millimeters, the mice were considered dead and euthanized. Proteomic sequencing assays were subsequently performed on the tumor tissues of the animals. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [90] partner repository with the dataset identifier PXD073783.

S1 Fig. The necessity of the genes fabD1 and sfp for the heterologous expression of FK228 in Escherichia coli.

And the Anderson promoters that can be used to construct the plasmids that encoded the fabD1 and sfp-type PPtase from Chromobacterium violaceum for assisting the FK228 biosynthesis in EcN.

https://doi.org/10.1371/journal.pbio.3003657.s001

(TIFF)

S2 Fig. The MS analysis of FK228 detection from the fermentation of engineered EcN clones.

The FK228 had not been detected from the fermentation of the EcN that contained the original FK228 biosynthetic genes from Chromobacterium violaceum (clones 15, 18, and 20). The FK228 had been detected from the fermentation of the EcN that contains the codon optimized FK228 gene cluster (clone 1,426).

https://doi.org/10.1371/journal.pbio.3003657.s002

(TIFF)

S3 Fig. The MS analysis of FK228 detection from the fermentation of engineered Escherichia coli BL21 clones.

(A) The FK228 had been detected from the fermentation of the E. coli BL21 that contains codon optimized FK228 gene cluster (clone 1–4 and 10) and the E. coli BL21 that contain the original FK228 gene clusters from Chromobacterium violaceum (clone 15 and 18). (B) The FK228 yield of the E. coli BL21 that contains FK228 biosynthetic genes with codon optimization was higher than that without codon optimization (n = 3). The assay was performed in triplicate and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s003

(TIFF)

S4 Fig. Restriction analysis of constructed plasmids involved in anticancer therapy.

(A) XmnⅠ restriction analysis of plasmid 23, 24, and 25. (B) BamHⅠ restriction analysis of plasmid 40 and 44. Correct clones are indicated with red check mark.

https://doi.org/10.1371/journal.pbio.3003657.s004

(TIFF)

S5 Fig. Quantification of FK228 production and the CFU counting of fermentation culture.

(A–C) Data from Batch 1 experiments (24 h–96 h): (A) Quantification of FK228 production in supernatant of fermentation and precipitation of engineered EcN strains (n = 3). (B) The proportion of FK228 that diffused into the culture medium to the total FK228 produced (n = 3). (C) Analysis of viable cell count during fermentation of engineered EcN strains (n = 3). (D–F) Data from Batch 2 experiments (0 h–24 h): Quantification of FK228 production in the fermentation supernatant and pellet of engineered EcN strains, as well as the OD600 and CFU of the bacteria within 24 hours (n = 3). (G) FK228 production of EcN-23 at 24 h, 48 h, and 48 h with an additional inducer supplementation at 24 h (n = 3). (H) Scanning electron microscopy (SEM) images of EcN strains at different time points during fermentation. The magnification of all SEM images is 20,000× . The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s005

(TIFF)

S8 Fig. The tumor colonization of restriction-modification enzymes knockout strains.

(A) Genes related to restriction-modification enzymes and the corresponding EcN strains with gene knockouts. (B) The colonization status of EcN strains with gene knockouts in tumors Data are mean values of results from duplicate experiments, with error bars indicating standard deviation. The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s008

(TIFF)

S10 Fig. The plasmid maintenance of AT series strains.

The mice bearing 4T1 tumors were intravenously injected with the suspension of AT series strains EcN-34, EcN-35, EcN-DR004-34, EcN-DR004-35, and CK strain EcN-23 respectively. The blank group was intravenous administration of PBS. The mice in EcN-23 were intragastric administration of kanamycin solution daily. The mice in the AT series strains group were not intragastric administration of antibiotics (n = 4). Data are mean values of results from duplicate experiments, with error bars indicating standard deviation. The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s010

(TIFF)

S11 Fig. Cell viability.

Cell viability of 4T1 (A) cells and HC11 (B) cells incubated with the fermentation supernatant of engineered EcN stain (approximately 0.16 μM FK228), EcN wild type stain or FK228 standard solution (2.5 μM). Data are mean values of results from duplicate experiments, with error bars indicating standard deviation. The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s011

(TIFF)

S12 Fig. The proliferative effects of FK228 on mouse breast cancer cells and mouse mammary epithelial cells.

The CCK-8 kit was employed to determine the cell inhibitory effect on 4T1 (A) cells and HC11 (B) cells subsequent to their incubation with FK228. Data are mean values of results from duplicate experiments, with error bars indicating standard deviation. The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s012

(TIFF)

S13 Fig. The colonization of engineered EcN strains and EcN in tumors.

(A) The colonization of engineered EcN strains and EcN in tumors 15 days after injection. (B) The colonization of EcN-25 in tumors 20 days after injection (n = 3). Data are mean values of results from duplicate experiments, with error bars indicating standard deviation. The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s013

(TIFF)

S14 Fig. Survival curves of mice bearing 4T1 tumors treated with engineered EcN strains, or EcN, FK228, or PBS.

(A) supporting information for Fig 7. (B) Supporting information for Fig 8. (C) Supporting information for Fig 9. (D) Supporting information for Fig 10 (n = 4). Data are mean values of results from duplicate experiments, with error bars indicating standard deviation. The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s014

(TIFF)

S16 Fig. The colony-forming units (CFU) of EcN strains to be used for tail vein injection.

The EcN strains to be used for tail vein injection were serially diluted, and the diluted samples were spread on LB agar plates. The colonies were counted to calculate the CFU (n = 3). Data are mean values of results from duplicate experiments, with error bars indicating standard deviation. The assay was performed in triplicate (biological replicates) and is represented as the mean ± standard error of the mean (SEM). The underlying data can be found in https://figshare.com/s/340f560874499f66444e.

https://doi.org/10.1371/journal.pbio.3003657.s016

(TIFF)