Engineering Cycle

Previous studies have indicated that Staphylococcus xylosus isolated from primary tumor tissue possesses a certain level of invasion capability toward TNBC (Fu et al., 2022). Building on this observation, we employed the reference strain Staphylococcus xylosus ATCC 29971 as the chassis strain to verify whether this invasion phenotype could be reproduced in a standard genetic background and to quantitatively assess its invasion efficiency. Specifically, S. xylosus ATCC 29971 (TrojanHorseβ) was engineered to express sfGFP and co-cultured with TNBC cell lines to enable precise bacterial localization and quantification of invasion rates. Here, the invasion rate was defined as the percentage of host cells containing intracellular bacteria, as determined by flow cytometry. It is known that invasion efficiency correlates with the multiplicity of infection (MOI), although the precise relationship remains unclear. We hypothesize that this trend may approximate a bell-shaped distribution: at low MOIs, insufficient bacterial amount may result in reduced invasion efficiency, whereas at high MOIs, excessive metabolic burden could impair bacterial viability, thereby decreasing invasion efficiency. To investigate this, the bacteria were co-cultured with human MDA-MB-231 at various MOIs to determine the optimal invasion ratio.

Using the One-Step Cloning (OSC) method, the constitutive promoters P_SarA1 and P_cap, together with the sfgfp gene fragment, were inserted into appropriate sites of the plasmid pLI50 to construct two expression vectors: Reporter α and Reporter β (pLI50[P_SarA1-sfgfp] and pLI50[P_cap-sfgfp]). The plasmids were amplified in Escherichia coli JM110, extracted, and subsequently electroporated into Staphylococcus xylosus ATCC 29971 cells rendered competent by glycine treatment. Positive clones were selected on TSA plates containing 10 μg/ml chloramphenicol. Prior to co-culture, the bacteria were cultured for approximately 24 hours to reach the late stationary phase, during which toxin secretion induces specific phenotypic alterations and enhances their invasiveness. The culture were diluted to different MOIs and co-cultured with MDA-MB-231 cells in 12-well plates for at least 6 h. Following co-culture, gentamicin was added for 30 min to eliminate extracellular bacteria, after which the medium was removed and the cells were washed. The cells were then collected and examined by fluorescence microscopy to visualize GFP expression, while intracellular fluorescence was quantified by flow cytometry to assess the bacterial invasion rate.

In the pilot experiments, Staphylococcus xylosus reporter strains expressing sfGFP but lacking FnBPA (TrojanHorseβ[Reporter α]), were co-cultured with MDA-MB-231 cells at various multiplicities of infection (MOIs).

Pilot experiments A and B demonstrated that the Trojanβ[Reporter α] exhibited consistently low invasion rates, with invasion rates fluctuating between approximately 1% and 4%, and no clear MOI corresponding to a maximal invasion rate was observed. These findings suggest that our initial hypothesis of a bell-shaped distribution may not be valid. Although invasion rates of 1–4% are relatively high compared to typical intratumoral bacteria (Fu et al., 2022), they remain insufficient for constructing an effective TNBC intracellular delivery system, highlighting the need for further optimization.

To enhance the invasion efficiency of Trojanβ and develop a functional intracellular delivery platform, we retained sfgfp as a reporter gene and introduced heterologous expression of the canonical invasion protein FnBPA from Staphylococcus aureus USA300 to promote bacterial internalization. Since the relationship between the MOI and bacterial invasiveness remains undefined, we further assessed the invasion efficiency across a range of MOIs.

In this construct, the Staphylococcus aureus USA300 fnbA gene was positioned upstream of the sfgfp gene, with a ribosome binding site (RBS) interposed between them. A constitutive promoter (P_SarA1 or P_cap) was subsequently inserted immediately upstream of fnbA using the OSC method, generating two FnBPA expression constructs with distinct promoters: Sinon α and Sinon β (pLI50[P_SarA1-fnbA-RBS-sfgfp] and pLI50[P_cap-fnbA-RBS-sfgfp], respectively). The gene cassette from the former plasmid was uploaded to the iGEM Registry as a composite part (BBa_25LG0OCK), while the latter construct exhibited negligible fluorescence for unknown reasons. Following successful transformation, TrojanHorseβ[Sinon α] were co-cultured with MDA-MB-231 cells. The proportion of host cells harboring intracellular bacteria was determined by flow cytometry, and the intracellular localization of the bacteria was examined by fluorescence microscopy.

Pilot experiments C and D revealed that TrojanHorseβ[Sinon α] exhibited markedly enhanced invasion capability, with an invasion rate of up to 50.4%. The invasion rate displayed a bell-shaped dependence on MOI, reaching its maximum at an MOI of approximately 10. However, due to the limited dataset, this observation requires further validation. Notably, the overall invasion efficiency remained lower than anticipated, indicating that further optimization may be required.

In these experiments, invasion rates were significantly higher than those of the TrojanHorseβ[Reporter α] control group. However, due to limitations in experimental design and the limited number of replicates, the correlation between MOI and invasiveness could not be determined with precision. The optimal MOI can only be approximated to fall between 10 and 15, pending further validation.

We further hypothesize that bacterial expression of sfGFP was unstable and insufficient in intensity. In previous experiments, we observed substantial heterogeneity in overall sfGFP fluorescence intensity across independent experiments performed under comparable conditions, whereas the relative fluorescence levels within each experiment were consistent in pattern, although their absolute values differed significantly among experiments. This suggests that the sfGFP fluorescence was likely modulated by unknown environmental factors, resulting in experiment-specific fluctuations in overall signal strength. Moreover, all invasion-rate data were markedly lower than expected, which was likely attributable to the insufficient sfGFP signal causing false-negative events. Therefore, we plan to replace the sfGFP reporter system with a brighter and more stable fluorescent marker in future experiments to improve detection reliability.

To obtain stronger and more stable fluorescence signals for bacterial tracking, we employed the CellTrace™ Far Red Cell Proliferation Kit to label the bacteria. CellTrace™ Far Red is widely used in cell proliferation assays, as the dye can freely penetrate cell membrane and covalently bind to free amines within the cytoplasm, thereby retaining fluorescence through at least six cell divisions over several days to weeks. Moreover, CellTrace™ Far Red has an excitation maximum at 630 nm, which minimizes spectral overlap with the FITC channel and avoids the high background fluorescence typically observed at 488 nm. Collectively, these characteristics make CellTrace™ Far Red particularly suitable for fluorescence-based bacterial tracing and invasion analyses.

Considering that the invasion rate of TrojanHorseβ[Reporter α] may have been underestimated due to the weak and unstable fluorescence of sfGFP, we plan to apply CellTrace™ Far Red staining to this strain to re-evaluate their invasion efficiency toward multiple cell lines.

Given that expression of FnBPA may reduce the tumor-targeting specificity of S. xylosus and increase its off-target invasiveness in healthy tissues, we evaluated whether the engineered bacteria retain selectivity toward tumor cells by comparing their invasion rates between tumor and normal mammary epithelial cell lines.

An additional labeling step was introduced prior to co-culture. Bacteria were stained with CellTrace™ Far Red according to the manufacturer’s instructions, followed by three consecutive rounds of centrifugation and washing with PBS to remove residual dye. The labeled bacteria were then co-cultured with multiple cell lines, and invasion rates were quantified by flow cytometry.

To evaluate the labeling performance of the CellTrace™ Far Red dye, we compared its fluorescence intensity with that of sfGFP in the same bacterial population. Fluorescence imaging and quantitative analysis using ImageJ revealed that the CellTrace Far Red signal was significantly stronger than the sfGFP fluorescence under identical conditions, demonstrating the superior labeling efficiency of the dye.

We next labeled TrojanHorseβ[Sinon α] and TrojanHorseβ[Reporter α] with CellTrace™ Far Red and performed repeated (n = 3) co-culture experiments with cancer cell lines (4T1/MDA-MB-231) and corresponding normal mammary epithelial cell lines (HC11/HBL-100) to assess the invasion of the engineered bacteria by flow cytometry. Analysis showed that the negative control population exhibited a single, low-intensity peak, whereas the co-cultured samples displayed a distinct single peak that was completely shifted toward higher fluorescence, with signal intensity approximately three orders of magnitude greater than that of the negative control. When the gate was set to include 1% of the negative control, the maximum positivity reached 99.7%. These results indicate that CellTrace™ Far Red provides a strong and well-separated fluorescence signal with a high signal-to-noise ratio, enabling sensitive and reliable detection of labeled bacteria and substantially reducing the likelihood of false-negative events.

Notably, we observed that the laboratory reference strain Staphylococcus xylosus ATCC 29971 reporter strain (TrojanHorseβ[reporter α]) exhibited significant invasiveness towards MDA-MB-231 cells, with a maximum invasion rate of 54.2%. Parallel co-culture experiments with non-malignant HBL-100 cells revealed a statistically pronounced selectivity toward MDA-MB-231 at a higher MOI (25). These findings suggest the presence of an intrinsic, yet unidentified, host-cell invasion mechanism in Staphylococcus xylosus ATCC 29971.

As shown in Figure 9, flow cytometry analysis revealed distinct signal patterns between the APC-Cy7 and FITC channels in MDA-MB-231 cells co-cultured with the TrojanHorseβ[reporter α]. In the APC-Cy7 channel (640 nm), the experimental group displayed a clear bimodal distribution compared with the unimodal peak of the negative control, indicating the introduction of a specific APC-Cy7-positive subpopulation, excluding nonspecific or background signal. In contrast, the 488 nm (FITC) channel showed negligible differences between experimental and control groups, suggesting minimal sfGFP expression. These results confirm the presence of host-cell invasion events while explaining why preliminary assays relying solely on sfGFP fluorescence were insufficient to accurately assess bacterial internalization.

We demonstrated that CellTrace Far Red dye markedly reduces false negatives in flow cytometry, making it a robust and stable alternative to sfGFP as a high-intensity reporter signal. This substitution enables more accurate quantification of bacterial invasion efficiency and improves the reliability of infection assays. Furthermore, in situations where spectral overlap occurs or when microscopes lack a 640 nm excitation source, other dyes from the CellTrace series, as well as CFSE and other cell proliferation dyes, can serve as effective substitutes for bacterial tracing. These findings collectively highlight the versatility and robustness of cell proliferation dyes as reliable labeling tools in bacterial invasion studies.

Building upon these methodological advances, we conducted pairwise co-cultures involving two bacterial strains (TrojanHorseβ[Sinon α] and TrojanHorseβ[Reporter α]) and four mammary cell lines (human and murine normal and TNBC cell lines) to further characterize the invasive capacity of our engineered bacteria. This systematic approach allowed us to quantitatively evaluate both invasion efficiency and cell-type selectivity. Notably, the introduction of FnBPA substantially enhanced bacterial invasion toward TNBC cells, achieving invasion rates up to 99.7%, while also conferring pronounced cell-line selectivity under specific MOIs. These results demonstrate that the FnBPA-expressing bacteria possess a remarkable capacity to recognize and preferentially invade TNBC cells.

While not previously reported to exhibit any cell-invasive phenotype, the laboratory reference strain S. xylosus ATCC 29971 was unexpectedly found to invade the human TNBC cell line MDA-MB-231, achieving a maximum infection rate of 52.4%. The invasion rate of this strain increased in a dose-dependent manner with the MOI, indicating an active and quantifiable invasion process. Moreover, parallel co-culture assays with the non-malignant HBL-100 cell line revealed that, under high MOI conditions (MOI = 25), S. xylosus ATCC 29971 displayed significant selectivity toward TNBC cells over normal epithelial cells. Collectively, these findings demonstrate that S. xylosus ATCC 29971 possesses intrinsic host cell invasive potential and a natural preference for malignant cells, despite its conventional classification as a nonpathogenic laboratory strain.

Taken together, our findings not only identify S. xylosus ATCC 29971 as a previously unrecognized invasive bacterium, but also suggest its potential utility as a foundation for developing a commensal-based TNBC-targeted intracellular drug delivery platform. Moving forward, we plan to retain and further characterize this strain to elucidate the molecular determinants underlying its selective invasion and assess its suitability for therapeutic engineering applications.

The fluorescence intensity of the labeled bacteria is sufficiently strong to effectively prevent false negatives at the signal detection level. Nevertheless, other potential sources of bias throughout the experimental workflow must be carefully controlled to ensure data reliability.

Specifically, insufficient bactericidal efficacy during the gentamicin protection assay may allow extracellular bacteria to survive and adhere to host cell surfaces, thereby generating false-positive results. Moreover, engineered bacteria that are internalized by host cells but fail to escape from lysosomes would also produce misleading positive signals, as only bacteria that successfully reach the cytoplasm can evade degradation, establish intracellular colonization, and subsequently release therapeutic molecules.

To rigorously evaluate the robustness and accuracy of our bacterial platform, we designed two complementary experiments aimed at independently validating these two critical steps: (i) verifying the killing efficiency of gentamicin under co-culture conditions, and (ii) assessing the ability of intracellular bacteria to escape from endosomal compartments into the cytoplasm.

To validate these two aspects, we established two experimental modules:

1. Gentamicin bactericidal efficiency assay:
This assay was designed to confirm that the concentration and treatment duration of gentamicin used in the co-culture system are sufficient to eliminate extracellular bacteria. Bacterial suspensions were incubated with gentamicin in DMEM supplemented with 10% FBS, and aliquots were collected at 5-min intervals. Following centrifugation and resuspension, serial dilutions were plated for CFU enumeration to quantify bacterial survival over time. This validation step ensures that the antibiotic treatment completely removes extracellular contaminants, thereby preventing false-positive signals in subsequent infection experiments.

2. Intracellular localization and endosomal escape assay:
To examine whether the engineered strain TrojanHorseβ[Sinon α] can escape from endosomal compartments after internalization, multichannel confocal fluorescence imaging was performed. Bacterial cytoplasm was labeled with CFSE (green), endosomes and lysosomes were stained with LysoTracker Red (red), and host nuclei were counterstained with Hoechst 44334 (blue). Infected host cells were co-cultured under hypoxic conditions (5% O₂) for approximately 10 hours prior to imaging. Distinct spatial distributions of bacteria were observed, allowing evaluation of endosomal escape efficiency and intracellular localization fidelity.

To evaluate the reliability of the two critical experimental steps, we performed validation assays for both gentamicin bactericidal efficiency and intracellular endosomal escape capability.

1. Gentamicin bacterialcidal efficiency assay:

The red-marked area indicates the untreated control group (without gentamicin). Labels 5/10/15/20/25/30 represent the treatment durations (minutes). No bacterial colonies were observed in any gentamicin-treated samples, indicating that gentamicin exhibited strong bactericidal activity under the tested conditions.

2. Intracellular localization and endosomal escape assay:
Bacterial cytoplasm was labeled with CFSE (green) to allow long-term tracking, while late endosomes and lysosomes were stained with LysoTracker Red (red). Nuclei were marked by Hoechst 44334 (blue), and cellular boundaries were outlined based on bright-field images due to instrument limitations in DID detection. Following co-culture of infected host cells under hypoxic conditions (5% O₂) for approximately 10 hours, confocal fluorescence microscopy revealed distinct cellular distributions of the labeled bacteria. Representative fluorescence micrographs are presented in Figure 11, where nuclei (blue), endosomes/lysosomes (red), and CFSE-labeled bacteria (green) are shown, with white contours delineating individual cell boundaries. The left and right panels correspond to the same microscopic field, displaying different channel combinations: the left panel shows the merged image of all fluorescence channels, whereas the right panel excludes the 488 nm channel to enhance visualization of bacterial spatial distribution.

Fluorescence microscopy revealed that TrojanHorseβ[Sinon α] bacteria were predominantly distributed within the host cytoplasm rather than confined to endosomal or lysosomal vesicles. Only a small subset of fluorescent signals showed co-localization with LysoTracker staining, indicating limited retention within endosomal compartments. This spatial pattern suggests that most intracellular bacteria successfully escaped endosomal confinement and accessed the host cytosol. Collectively, these results demonstrate that TrojanHorseβ[Sinon α] enables efficient endosomal escape, facilitating stable cytoplasmic localization following internalization.

Together, the results from these validation experiments confirm the robustness of our workflow and the reliability of intracellular signal interpretation. The gentamicin protection assay effectively eliminated extracellular bacteria, ensuring that the observed fluorescence signals predominantly originate from internalized cells. Meanwhile, TrojanHorseβ[Sinon α] demonstrated efficient endosomal escape, supporting its potential as a chassis strain for intracellular drug delivery applications.

However, it remains technically challenging to completely exclude the possibility that residual dead bacteria adhered to the cell surface might contribute to background fluorescence. To further address this limitation, future experiments could incorporate an extracellular fluorescence quencher to selectively suppress surface-associated signals prior to imaging or flow cytometric analysis. Implementing this step would provide an additional layer of validation, ensuring that all detected fluorescence originates exclusively from bacteria residing within host cells.

Based on previous studies, most researchers have employed temperature-sensitive plasmids (e.g., pBT2) for homologous recombination in Staphylococcus xylosus, allowing integration of the target gene into the chromosome for expression (Brückner, 1997). However, since this method is time-consuming and technically demanding, we selected the shuttle vector pLI50, which can replicate in both Escherichia coli and Staphylococcus aureus. Staphylococcus possesses a DNA methylation–based restriction–modification (R–M) system that protects against the invasion of foreign DNA. Since the strength of the R–M system in Staphylococcus xylosus ATCC 29971 cannot be determined, we planned to initially disregard the R–M system in our preliminary experiments and introduce the plasmid into Staphylococcus xylosus to test whether it could be directly expressed in the host.

To facilitate plasmid transformation, the plasmid pLI50 was initially amplified in Escherichia coli DH5α and subsequently extracted. The purified plasmid was then electroporated into electrocompetent Staphylococcus xylosus ATCC 29971 cells that had been pretreated with sucrose. Positive transformants were selected on Tryptic Soy Agar (TSA) plates containing 10 μg/ml chloramphenicol.

No colonies grew on the plates, indicating that plasmid introduction was unsuccessful. Consistent results were obtained across multiple independent attempts.

The experimental results demonstrate that bypassing the R-M system is an essential step for achieving successful transformation. Therefore, further plasmid editing or modification of the Escherichia coli host used for plasmid amplification is required. Meanwhile, the preparation protocol for competent cells of Staphylococcus xylosus ATCC 29971 may still be suboptimal and requires further optimization.

Literature review revealed that Staphylococcus xylosus ATCC 700404 commonly employs E. coli TG1 for plasmid amplification (Brückner, 1997). Given the genomic similarity between Staphylococcus xylosus ATCC 700404 and Staphylococcus xylosus ATCC 29971, and the limited accessibility of Staphylococcus xylosus ATCC 700404 domestically, we propose that Escherichia coli TG1 could also be applied to Staphylococcus xylosus ATCC 29971 for this purpose. In addition, we identified an alternative protocol for preparing competent cells using glycerol solution washes, which we plan to evaluate.

To facilitate plasmid transformation, the plasmid pLI50 was initially amplified in Escherichia coli TG1 and subsequently extracted. The purified plasmid was then electroporated into electrocompetent Staphylococcus xylosus ATCC 29971 cells that had been pretreated with glycerol. Positive transformants were selected on Tryptic Soy Agar (TSA) plates containing 10 μg/ml chloramphenicol.

No colonies grew on the plates, indicating that plasmid introduction was unsuccessful. Consistent results were obtained across multiple independent attempts.

The restriction-modification system profile of Escherichia coli TG1 may not be compatible with that of Staphylococcus xylosus ATCC 29971. Further exploration of alternative solutions is still required.

Database revealed that the restriction–modification system of Staphylococcus xylosus ATCC 29971 specifically recognizes and cleaves methylated DNA sequences, including the dam and dcm methylation sites commonly present in Escherichia coli K-12 strains such as DH5α and TG1. To circumvent interference from this system, the methylation-deficient Escherichia coli JM110 (dam⁻/dcm⁻ double mutant) was employed for plasmid amplification in this study.

To facilitate plasmid transformation, the plasmid pLI50 was initially amplified in Escherichia coli JM110 and subsequently extracted. The purified plasmid was then electroporated into electrocompetent Staphylococcus xylosus ATCC 29971 cells that had been pretreated with glycerol. Positive transformants were selected on Tryptic Soy Agar (TSA) plates containing 10 μg/ml chloramphenicol.

No colonies grew on the plates, indicating that plasmid introduction was unsuccessful. Consistent results were obtained across multiple independent attempts.

We speculate that the restriction–modification system of Staphylococcus xylosus ATCC 29971 has been circumvented; however, plasmid transformation was still unsuccessful. A likely explanation is the poor quality of electrocompetent cells prepared using the glycerol-wash method. Therefore, further optimization of the competent cell preparation protocol is required.

Standard washing protocols employing glycerol, sucrose, or a glycerol–sucrose mixture were ineffective for Staphylococcus xylosus. Therefore, we employed a washing procedure using 10% glycine, as described by Dr. Shang Cai (Dong et al., 2022). All steps were performed under strict ice-bath conditions, and plasmid DNA pretreated with E. coli JM110 was introduced at relatively high concentrations.

Plasmid pLI50 was first propagated in Escherichia coli JM110 and subsequently purified using a standard plasmid extraction protocol. The plasmid DNA was then introduced into Staphylococcus xylosus ATCC 29971 electrocompetent cells, which had been prepared by washing with 10% glycine under ice-bath conditions. Transformants were selected on tryptic soy agar (TSA) plates supplemented with chloramphenicol at a final concentration of 10 µg/ml, and colonies exhibiting resistance were further verified.

After two days of incubation, several large colonies were observed on both plates. Colony PCR and microscopic analysis confirmed the presence of the plasmid and the identity of the isolates as Gram-positive cocci. Initial colony PCR following mechanical disruption produced positive results in two groups and negative results in two groups; however, repeated PCR after further disruption of the latter yielded positive results.

The gel electrophoresis results shown in the figure were obtained from bacterial samples lysed using a physical method. Lanes 1 and 3 displayed positive bands, whereas lanes 2 and 4 were negative. Samples 2 and 4 were then re-sampled from the same colonies and subjected to NaOH-based chemical lysis, which yielded positive results. The expected target band size was 979 bp.

Based on the recognition mechanism of the restriction–modification (R-M) system in Staphylococcus xylosus, we inferred that the previous failure of plasmid transformation was due to methylation of the target plasmid by Escherichia coli DH5α and TG1, which rendered the plasmid susceptible to cleavage by the Staphylococcus xylosus ATCC 29971 R-M system. After switching to an alternative amplification host, positive results were obtained, demonstrating that the pLI50 plasmid amplified in Escherichia coli JM110 can replicate and be stably maintained in Staphylococcus xylosus. Thus, we successfully circumvented the R-M system of Staphylococcus xylosus ATCC 29971 and achieved efficient plasmid introduction.

Nevertheless, this method remains suboptimal, as the transformation efficiency of competent cells prepared using the glycine-wash approach was extremely low, yielding only a small number of transformant colonies in each attempt. Extending the post-electroporation recovery period to 3 hours modestly increased the colony yield, but this strategy still requires further validation and optimization.

Restriction–modification (R–M) systems are widely distributed among Staphylococcus species, frequently impairing plasmid transformation and thereby complicating genetic manipulation. However, the activity and strength of the R–M system in Staphylococcus epidermidis ATCC 14990 (TrojanHorseα) have not yet been characterized. To preliminarily evaluate the feasibility of heterologous gene expression in this strain, the potential impact of R–M systems was not considered in the initial experiments. Instead, an expression plasmid was directly introduced into the host to determine whether it could replicate autonomously and support gene expression. The shuttle vector pLI50 was employed for this purpose, as it possesses a medium copy number in Escherichia coli and a high copy number in Staphylococcus species (Lee et al., 1991), making it an appropriate choice for assessing plasmid replication and expression capability in S. epidermidis.

To facilitate plasmid transformation, the plasmid pLI50 was initially amplified in E. coli DH5α and subsequently extracted. The purified plasmid was then electroporated into electrocompetent Staphylococcus epidermidis ATCC 14990 cells that had been pretreated with sucrose. Positive transformants were selected on Tryptic Soy Agar (TSA) plates containing 10 μg/mL chloramphenicol.

No colonies grew on the plates, indicating that plasmid introduction was unsuccessful. Consistent results were obtained across multiple independent attempts.

Experimental results indicate that the R–M system of Staphylococcus epidermidis ATCC 14990 exhibits strong activity. Therefore, circumventing this system is a prerequisite for achieving successful transformation. In addition, the current protocol for preparing competent S. epidermidis ATCC 14990 cells may be suboptimal and requires further optimization.

Literature review revealed that Staphylococcus xylosus ATCC 700404 commonly employs Escherichia coli TG1 as a host strain for plasmid amplification (Brückner, 1997). In contrast, reports on genetic manipulation of Staphylococcus epidermidis are limited, and no plasmid amplification strains have yet been identified for direct reference. Therefore, we considered applying this strategy to S. epidermidis ATCC 14990. In addition, we identified a protocol for preparing competent cells through glycerol-wash treatment and plan to evaluate its applicability.

To facilitate plasmid transformation, the plasmid pLI50 was initially amplified in Escherichia coli TG1 and subsequently extracted. The purified plasmid was then electroporated into electrocompetent Staphylococcus epidermidis ATCC 14990 cells that had been pretreated with glycerol. Positive transformants were selected on Tryptic Soy Agar (TSA) plates containing 10 μg/ml chloramphenicol.

No colonies grew on the plates, indicating that plasmid introduction was unsuccessful. Consistent results were obtained across multiple independent attempts.

The R–M system profile of Escherichia coli TG1 may be incompatible with that of Staphylococcus epidermidis ATCC 14990, and further investigation into alternative strategies is still required.

We investigated the R–M system profile of Staphylococcus epidermidis ATCC 14990. Literature reports and databases indicate that this strain possesses an intrinsic DNA methylation gene cluster composed of hsdM-hsdS (with partial overlap) and hsdR, which together form a type I R–M system (Lee et al., 2019). Since the recognition sequence of the specificity subunit HsdS remains undefined and no relevant information is available in public databases, direct optimization of the plasmid’s methylation recognition sites was not feasible. To circumvent cleavage by HsdR, we planned to perform hsdM-hsdS Plasmid Artificial Modification (PAM) in an amplification host prior to their introduction into S. epidermidis ATCC 14990.

Given the compatibility of Escherichia coli and its wide use as a plasmid amplification host, we explored strategies to endow it with type I modification capability through lambda red recombination, ensuring plasmids stably maintained across generations. We initially evaluated E. coli JM110, a dam⁻ dcm⁻ strain that lacks the common type IV modification system. However, literature review indicated that JM110 is rarely used for plasmid construction, likely because deletion of dam and dcm increases genomic mutation rates. In addition, JM110 and related strains are often engineered to inactivate recA or other recombination genes to suppress homologous recombination, which could further complicate genomic modification. Eventually, we selected E. coli DH5α (dam⁺ dcm⁺) as the host for modification and planned to replace its native dcm locus with the hsdM–hsdS gene cluster from S. epidermidis ATCC 14990 via λ-Red homologous recombination. The dcm locus was targeted for replacement because the recognition sequence of HsdS is unknown; potential overlap between Dcm and HsdS recognition sites could interfere with methylation, leading to PAM failure. Furthermore, to minimize the risk of plasmid degradation by endogenous type II restriction endonucleases, the native dam methylation system in E. coli DH5α was preserved to maintain protective adenine methylation at recognition sites.

We planned to perform λ-Red homologous recombination in Escherichia coli DH5α using the pKD46 plasmid, replacing the native dcm locus with the hsdM-hsdS fragment directly derived from the Staphylococcus epidermidis ATCC 14990 genome, with a kanR cassette inserted upstream as a selectable marker. This resulted in the construction of the strain DH5α[Δdcm-yedJ-J23100-RBS-kanR-T2-J23104-RBS-hsdM-hsdS-vsr] (DH5α[hsdM-hsdS]) (Yu et al., 2018).

The strain exhibited both ampicillin resistance (conferred by pKD46) and kanamycin resistance (from genomic integration of kanR). Since the downstream plasmid of interest also carried an ampR marker, pKD46 was eliminated by overnight cultivation at 40 °C without ampicillin to avoid interference during subsequent selection.

Following successful removal of pKD46, the target plasmid pLI50[Reporter α] (pLI50[P_SarA1-sfgfp]) was introduced into the engineered strain, and the modified plasmid was subsequently extracted. This plasmid was then electroporated into glycerol-prepared competent cells of S. epidermidis ATCC 14990, and positive colonies were obtained by selection on TSA plates supplemented with 10 μg/mL chloramphenicol.

No colonies grew on the plates, indicating that plasmid introduction was unsuccessful. Consistent results were obtained across multiple independent attempts.

We reasonably speculate that the restriction–modification system of Staphylococcus epidermidis ATCC 14990 has been circumvented; however, plasmid transformation was still unsuccessful. A likely explanation is the low quality of electrocompetent cells prepared using the glycerol-wash method. Therefore, further optimization of the competent cell preparation protocol is necessary.

The standard glycerol-wash method proved ineffective for Staphylococcus epidermidis ATCC 14990. Therefore, we opted to test several alternative competent cell preparation protocols.

The preparation of competent cells was tested under various conditions, including cultivation in BHI or TSB, and different washing procedures: water-only wash, sequential water and glycerol wash, and glycine wash.

Electroporation parameters were varied with respect to: whether the Staphylococcus epidermidis cells were subjected to a 1-minute heat treatment at 55 °C to inactivate the R–M system (under heat treatment, electroporation was performed after recovery at room temperature; without heat treatment, electroporation was conducted on ice), and the choice of recovery medium, including BHI, TSB, or BHI supplemented with 500 mM sucrose (Fey, 2014; Costa et al., 2017; Dong et al., 2022).

Among the various conditions tested, only the glycine-treated group successfully yielded single colonies. Microscopic examination confirmed that the isolates were Gram-positive cocci; however, colony PCR results were negative, indicating that plasmid introduction was unsuccessful. Repeated experiments consistently produced the same outcome.

Plasmid introduction was unsuccessful. Although single colonies occasionally appeared on the plates, this may have been due to antibiotic inactivation or spontaneous genetic mutations. A literature review indicated that S. epidermidis is inherently unsuitable for electroporation, which may explain our repeated failures (Yu et al., 2018). Furthermore, previous attempts to employ phage-mediated plasmid transfer were also unsuccessful. Therefore, additional methods need to be investigated in future studies.

We planned to use protein engineering to identify the most suitable FnBP for introduction into engineered bacteria.

Molecular dynamics (MD) simulations were employed to study the binding process of FnBP and Fn, and MM/PBSA was applied to quantify binding affinities, with simulation parameters closely aligned with experimental conditions.

Relevant literature was reviewed to collect all commonly studied FnBPs.

Based on FnBP–Fn complex structures predicted by AlphaFold-Multimer, we conducted MD simulations using GROMACS with the following parameters:

• CHARMM36 force field
• TIP3P water model,0.15 M NaCl
• 310 K
• cubic box with a 1.0 nm boundary, explicit solvent
• total simulation length of 1000 ns

Systematic analyses were performed on the MD trajectories. MM/PBSA binding free energy calculations were carried out during RMSD-stable intervals to quantify binding strength.

All successfully simulated FnBP datasets were organized and compared. From a modeling perspective, we concluded that FnBPs from Staphylococcus aureus and Streptococcus pyogenes exhibit stronger binding to Fn.

Consultating with experimental group, FnBPs were decided to be further screened based on modeling results to maximize experimental success rates.

To further enhance the experimental success rate of introducing exogenous proteins into engineered bacteria, we decided to incorporate phylogenetic analysis into our modeling. By considering both evolutionary relationships and molecular-level interaction results, we aimed to select FnBPs that are not only effective but also closely related to the host strain, thereby improving feasibility and stability.

16S rRNA genes from different bacterial species were aligned, and phylogenetic trees were constructed using MEGA12.

The resulting tree topology was validated against published literature and NCBI Taxonomy to ensure consistency with known phylogenetic classifications.

The phylogenetic tree showed that Staphylococcus epidermidis (one of our engineered strains) clusters closely with S. aureus, forming a tight branch. Staphylococcus xylosus (another engineered strain) diverged earlier but remains within the same evolutionary clade, while S. pyogenes belongs to a more distant genus.

Although MD simulations of S. aureus FnBP revealed two distinct binding conformations, one of them showed the highest binding affinity among all FnBPs, stabilizing for 0 ~800 ns. This suggests that S. aureus FnBP is highly likely to adopt this favorable binding conformation in reality.

Taking both molecular dynamics and phylogenetic proximity into account, we selected S. aureus FnBP as the optimal exogenous protein for engineered bacteria.

We aimed to construct a mathematical model to describe bacterial diffusion, convection and growth within tumor cells, thereby capturing both temporal proliferation and spatial dynamics.

Starting from the mass conservation equations:

$$\frac{\partial b}{\partial t} + \nabla \cdot \mathbf{J}_{b} = R_{b}$$ $$\frac{\partial c}{\partial t} + \nabla \cdot \mathbf{J}_{c} = R_{c}$$

Flux was decomposed into diffusion and convection terms.
Diffusion flux (Fick’s law):

$$\mathbf{J}_{\mathrm{diff},b} = -D_{b}\nabla b$$ $$\mathbf{J}_{\mathrm{diff},c} = -D_{c}\nabla c$$

Convection flux:

$$\mathbf{J}_{\mathrm{adv},b} = \vec{u} b$$ $$\mathbf{J}_{\mathrm{adv},c} = \vec{u} c$$

Source terms (exponential growth):

$$R_{b} = \mu b$$ $$R_{c} = -\frac{1}{Y}\,\mu b$$

The model becomes:

$$\frac{\partial b}{\partial t}+\nabla\cdot\!\bigl(-D_{b}\nabla b+\vec{u}b\bigr)=\mu b$$ $$\frac{\partial c}{\partial t}+\nabla\cdot\!\bigl(-D_{c}\nabla c+\vec{u}c\bigr)=-\frac{1}{Y}\,\mu b$$

COMSOL simulations showed reasonable spatial diffusion and convection patterns. However, bacterial concentrations at fixed points increased explosively without noticeable substrate limitation, suggesting that bacterial growth far outpaced transport, which does not match biological reality.

We found it necessary to strengthen the coupling between bacterial growth and substrate concentration, as well as to account for environmental carrying capacity.

We refined the bacteria term \(R_{b}\) by incorporating Monod nutrient-dependent kinetics and Logistic density limitation.

The substrate term \(R_{c}\) was also modified to couple bacterial growth with substrate utilization.

Optimized bacterial growth term:

Monod kinetics (substrate-dependent growth):

$$\mu(c) = \mu_{max}\cdot \frac{c}{K_{s}+c}$$

Logistic growth (density limitation):

$$b\left(1-\frac{b}{K_{B}}\right)$$

Optimized substrate consumption:

$$R_{c}=-\frac{1}{Y}\,\mu(c)\,b = -\frac{1}{Y}\,\mu_{max}\frac{c}{K_{s}+c}\,b$$

Final model:

Bacterial equation:

$$\frac{\partial b}{\partial t} -\nabla\cdot(D_{b}\nabla b)+\vec{u}\cdot\nabla b =\mu_{max}\cdot\frac{c}{K_{m}+c}\cdot\left(1-\frac{b}{K_{B}}\right)\cdot b$$

Substrate equation:

$$\frac{\partial c}{\partial t} -\nabla\cdot(D_{c}\nabla c)+\vec{u}\cdot\nabla c =-\frac{1}{Y}\,\mu_{max}\frac{c}{K_{m}+c}\,b$$

Simulation results showed that bacterial populations first increased then down, finally stabilized, which matches biological expectations.

COMSOL results confirmed that incorporating Monod kinetics and Logistic growth dynamics was essential. The resulting diffusion–convection–growth model provides a more accurate description of bacterial dynamics inside tumor cells.

At the beginning, we only had a broad vision: education should not be limited to one group, but should cover all age groups. Children, youth, middle-aged, and elderly people at different stages of life have different ways of understanding and different life backgrounds, so targeted educational approaches are needed.

We envisioned linking the class through a “problem chain,” pushing knowledge points forward in a continuous question-and-answer sequence rather than simply instilling information. We hoped that students would gradually enter the world of synthetic biology through the process of thinking – answering – thinking again.

We designed a problem chain in the PPT: from “What is a microorganism?” to “Can microorganisms be edited?” and then to “Can bacteria help treat cancer?” At the same time, we set up Q&A sessions, image-matching games, and a point mechanism to ensure an active classroom atmosphere.

In the classroom of Wuxi Xinwu District No.1 Experimental School, students actively participated in interactions, and the classroom atmosphere was lively. However, Principal Fang Yun reminded us that whether a class is interesting depends not only on interaction but also on whether the logic and analogies are clear. We realized that “too much interaction” might disrupt the rhythm.

We realized that problem-driven teaching is feasible in middle school, but at the high school level adjustments are needed, requiring stronger logical chains. Therefore, middle school classes became our first step in exploring the balance between logic and fun.

We tried to combine “classical Chinese text + science,” using classical literature to inspire children’s thinking about synthetic biology. For example, we selected the story of Yan Shi’s Mechanical Man from Liezi·Tang Wen, which resonates with the themes of artificial life and ethical boundaries.

Under the guidance of teacher Hairong Zhang, we prepared classical Chinese text slides, annotations, and translations, and tried to use this ancient text to introduce the issue of “synthetic life” and ethics.

The results were not ideal: most children and parents could not understand the relationship between the text and our project, and the classroom fell silent. Teacher Zhang reminded us: “A lecture without context is like a castle in the air.”

We realized that no matter how advanced the content is, without proper context and background introduction, it cannot be understood. The reading club deeply impressed on us the importance of background and storytelling in education, and it pushed us to improve later sessions.

After the attempts with middle school and the reading club, we planned to enter the high school stage, designing courses with clearer logic and more direct scientific content. We envisioned combining hands-on experiments + critical thinking to let students truly experience the integrity of scientific research.

Under the guidance of teacher Yongtao Zhu, we participated in the summer school course at Suzhou High School SIP Campus Westmark International School. The class was taught by teacher Zhu, and we served as teaching assistants, guiding students in sampling, inoculation, staining, and microscopic observation.

Students actively engaged in experiments and continuously raised questions, showing strong curiosity. After class, teacher Zhu reminded us: “A good class not only lets students work with their hands but also makes them think with their minds.” Practice is the entry point for emotions, but theory is where rationality lands.

We understood that theory and practice must run as two wheels in parallel: experiments can ignite interest, but true understanding requires a logical framework. The summer school helped us establish the “hands-on + minds-on” model, which later provided methodological support for elementary school and community teaching.