Invasion-Potentiation Module Design

Component Selection and Rationale

The microbial landscape within triple-negative breast cancer (TNBC) cells indicates that Staphylococcus commensals possess inherent tumor-targeting potential. However, current evidence demonstrates that their native invasive efficiency toward host cells remains limited. For example, commensals isolated from tumor tissues exhibited infection rates of only ~5% when co-cultured with dissociated primary tumor cells (Fu et al., 2022), while in vitro co-culture with standard strains and MDA-MB-231 cells achieved a maximum infection rate of only ~30% [Kanghui Ruan, Personal communication, 2025]. Such inefficiency substantially constrains the utility of Staphylococcus as a therapeutic delivery platform.

It is well established that Staphylococcus and other Gram-positive bacteria rely predominantly on MSCRAMMs (microbial surface components recognizing adhesive matrix molecules) for host cell adhesion and invasion. We therefore hypothesize that the limited invasion observed in wild-type commensals reflects the suboptimal binding properties of their MSCRAMMs, including less favorable interaction modes, reduced affinity, and lower stability with host receptors.

To overcome this limitation, we systematically evaluated several MSCRAMMs with validated invasion-promoting activity. Among the various MSCRAMM-mediated invasion routes, the fibronectin-dependent pathway (FnBP–Fn–α5β1 integrin) has been most comprehensively characterized and is generally considered the predominant mechanism facilitating bacterial internalization. In contrast, non-fibronectin pathways (such as those mediated by Cna, IsdB, or members of the Clf/Sdr family) have been identified but function primarily in complementary or accessory roles, and are typically insufficient to support efficient internalization independently (Foster et al., 2014). Consequently, we focused on the FnBPs family. Within this family, GROMACS molecular dynamics simulations, MM/PBSA free energy calculations, and MEGA12 phylogenetic analysis collectively identified S. aureus fibronectin-binding protein A (FnBPA) as the most suitable candidate for heterologous expression.

Of note, S. aureus encodes two paralogs, FnBPA and FnBPB, which display highly similar structural architectures and partially overlapping functions (Gries et al., 2020). Clinical studies have shown that epidemic S. aureus strains lacking FnBPB retain full infectivity, indicating that FnBPA alone is sufficient to support pathogenic competence (Speziale and Pietrocola, 2020). Based on these findings, FnBPA was selected as the representative paralog for heterologous expression in our engineered strains.

Gene Cassette Organization

To enable quantitative tracking of bacterial invasion efficiency, we introduced the reporter gene sfgfp. Two expression strategies were considered: independent promoters for fnbA and sfgfp, or a single promoter driving both genes. We adopted the latter, constructing a bicistronic cassette where fnbA is placed upstream and terminates with a stop codon, and sfgfp is positioned immediately downstream with a high-TIR ribosome binding site (RBS) to initiate its translation. This configuration ensured efficient expression of sfgfp downstream of fnbA and produced coordinated expression levels of FnBPA and sfGFP, allowing sfGFP fluorescence to serve as a reliable proxy for FnBPA expression and thereby circumventing direct verification of its expression. In terms of expression regulation, reliance on the native promoter places FnBPA expression under the control of global regulators such as Sar and Agr, leading to robust expression during exponential growth but repression in stationary phase (Speziale and Pietrocola, 2020). Because bacteria in the tumor microenvironment are expected to encounter metabolic stress and persist primarily in non-exponential phases, such temporal regulation may therefore be suboptimal. To ensure stable expression, we employed two constitutive promoters: the strong promoter P_SarA1 and the intermediate-strength promoter P_cap (Rondthaler et al., 2024). These promoters were used across all plasmids to drive the expression of sfGFP and FnBPA, enabling not only consistent and predictable production of FnBPA but also direct comparison of invasion efficiencies under distinct promoter strengths, thereby elucidating the relationship between FnBPA expression level and invasion capacity.

Plasmid Construction

We constructed two plasmid series: the Sinon series carrying fnbA and the Reporter series lacking fnbA. The Sinon series comprises pLI50[P_SarA1-fnbA-RBS-sfgfp] and pLI50[P_cap-fnbA-RBS-sfgfp], hereafter referred to as Sinon α and Sinon β, which harbor both fnbA and sfgfp and were designed to assess the contribution of FnBPA to invasion efficiency. In parallel, the Reporter series comprised two plasmids: pLI50[P_SarA1-sfgfp], designated Reporter α, and pLI50[P_cap-sfgfp], designated Reporter β. Both carried only the sfgfp gene and were used to quantify invasion efficiency in the absence of FnBPA.

Proof-of-concept Experiment

To verify and quantify the role of FnBPA in promoting bacterial invasion, we established a series of bacteria–cell co-culture assays, followed by downstream analytical procedures. Bacteria were first cultured for approximately 24 h to reach the late stationary phase, a condition associated with toxin secretion–induced phenotypic changes and maximal invasive potential [Kanghui Ruan, Personal communication, 2025]. In parallel, host cells were pre-conditioned at 5% O₂ for at least 6 h to ensure activation of hypoxia-responsive pathways prior to co-culture. For the co-culture experiments, bacterial suspensions of varying concentrations were added to adherent cells in multi-well plates and maintained at 5% O₂ for 6–24 h, with detectable invasion already observed as early as 4 h. Upon completion, extracellular bacteria were removed by using the gentamicin protection assay, as gentamicin is membrane-impermeable and selectively eliminates extracellular but not intracellular bacteria. The host cells were then harvested and fixed with paraformaldehyde (PFA). Finally, the prepared cell suspensions were analyzed by flow cytometry or processed for microscopy, enabling comprehensive assessment of invasion through quantification of bacteria-positive cells and evaluation of intracellular bacterial distribution.

Progress

The invasive capability of TrojanHorseβ[Sinon α] (FnBPA-expressing S. xylosus) was comprehensively evaluated through co-culture with both human and murine mammary cell lines across a range of MOIs. After six hours of co-culture under hypoxic conditions (5% O₂), the bacterium achieved a full-scale infection rate exceeding 99% in MDA-MB-231 cells, displaying significant cell-line selectivity at lower to moderate MOIs (5–15). In the murine system, TrojanHorseβ[Sinon α] exhibited robust invasive potential in 4T1 cells, maintaining infection rates above 92% even at the lowest tested MOI (20) and reaching up to 99% at higher doses. Statistical analyses revealed consistent and significant tissue selectivity relative to non-tumorigenic HC11 cells across all MOIs, confirming a stable and reproducible cell-line selectivity.

It is noteworthy that even among the selected chassis bacteria, distinct species and strains exhibit preferential infectivity toward different cell lines. Staphylococcus epidermidis shows greater efficiency in infecting human-derived cell lines (Kanghui Ruan, personal communication, August 18, 2025), consistent with its predominance in human triple-negative breast cancer (TNBC) cells (Dr. Shang Cai, [webinar], July 30, 2025). In parallel, Staphylococcus xylosus preferentially infects murine-derived cell lines (Kanghui Ruan, personal communication, August 18, 2025), reflecting its enrichment in murine TNBC cells (Dr. Shang Cai, [webinar], July 30, 2025). Based on these preferences, we primarily matched bacterial species with the corresponding host cell type in our co-culture experiments. At the same time, we also tested cross-infection with non-corresponding species–cell line pairs, aiming to assess the potential generalizability of infectivity across host backgrounds. Moreover, strains isolated from primary tumor tissues generally display stronger intrinsic invasive capacity and are presumed to possess enhanced tumor-targeting specificity, possibly shaped by selective pressures and adaptive differentiation within the tumor microenvironment. Although our experiments were designed to approximate such physiological conditions, restrictions on the use of primary tissues limited our ability to obtain strains with these advantageous traits.

Hypoxia-Responsive Module (HRM) Design

Component Selection and Rationale

Given the inherent risk of non-specific cell invasion mediated by FnBPA, the incorporation of a tumor microenvironment (TME)–responsive sensing system is essential to ensure the activation of safety modules and to safeguard the overall biosafety of the engineered construct. Because Staphylococcus species lack well-established genetic tools and no TME-responsive elements are readily available, we sought oxygen-regulated regulatory modules encoded in staphylococcal genomes. We identified a hypoxia-responsive module composed of the endogenous nitrogen metabolism operon nreABC and its downstream promoter P_narT. The nreABC operon is highly conserved among staphylococci and encodes a three-component regulatory system, in which P_narT functions as an integrative promoter responsive to both oxygen (via NreB/NreC) and nitrate (via NreA).

Gene Cassette Organization

As nreABC is a conserved module encoded in the staphylococcal genome, we initially considered employing only P_narT to confer hypoxia responsiveness on the plasmid. However, in the context of high-copy backbones (pUC57 and pLI50), the presence of multiple promoter copies would likely sequester the limited pool of regulatory proteins, thereby causing a titration effect and attenuating the intended response. To prevent this issue and ensure robust regulation, we incorporated the entire nreABC–P_narT module into the plasmid construct.

Plasmid Construction

The HRM was used to construct an sfgfp-based reporter plasmid and two safety mechanisms: a controlled drug release system and a suicide switch.

1. A reporter plasmid carrying the nreABC–P_narT promoter linked to sfgfp, Reporter ε (pLI50[nreABC–P_narT–sfgfp]Sep) and Reporter ζ (pLI50[nreABC–P_narT–sfgfp]Sxy) were constructed for quantitative characterization of the Hypoxia-Responsive Module. The corresponding Staphylococcus xylosus ATCC 29971 strain constructed: TrojanHorseβ[Reporter ζ] (S. xy[nreABC–P_narT–sfgfp]).

2. For the drug-controlled release system, the Hypoxia-Responsive Module was coupled with a therapeutic payload (represented in our project by apoptin), ensuring that off-target bacteria are unable to release cytotoxic agents outside the tumor microenvironment. Plasmids constructed: Odysseus α (pLI50[nreABC–P_narT–apoptin]Sep) and Odysseus β (pLI50[nreABC–P_narT–apoptin]Sxy). The corresponding Staphylococcus xylosus ATCC 29971 strain constructed: TrojanHorseβ[Odysseus β] (S. xy[nreABC–P_narT–apoptin]).

3. In parallel, we designed a suicide switch by integrating the MLSB-resistance gene ermC downstream of the Hypoxia-Responsive Module, thereby confining its expression to the hypoxic tumor microenvironment. Upon bacterial dissemination into normoxic tissues, transcription from P_narT ceases, resulting in the loss of ermC-mediated resistance. Under continuous systemic antibiotic administration, off-target bacteria are thus rendered vulnerable and can be efficiently eliminated, preventing endotoxin-associated damage to healthy tissues. As ermC confers resistance to MLSB antibiotics, we selected azithromycin as the selective agent due to its superior pharmacokinetics and clinical safety, ensuring a robust and durable selective pressure while preserving patient quality of life. Plasmids constructed: SuicideSwitch α (pUC57[nreABC–P_narT–ermC]Sep) and SuicideSwitch β (pUC57[nreABC–P_narT–ermC]Sxy). The corresponding Staphylococcus xylosus ATCC 29971 strain constructed: TrojanHorseβ[SuicideSwitch β] (S. xy[nreABC–P_narT–ermC]).

Proof-of-concept Experiment

For the characterization of the hypoxia-responsive module, overnight bacterial cultures were diluted, plated, and incubated under either normoxic (21% O₂) or hypoxic conditions (5% O₂). Fluorescence intensity was assessed the following day using fluorescence microscopy.

For validation of the suicide switch, diluted bacterial suspensions were plated onto erythromycin-containing agar and incubated under normoxic (21% O₂) or hypoxic (5% O₂) conditions. Colony formation was examined the next day to evaluate switch activity.

For assessment of the anticancer effect, bacteria were added to cell culture systems and co-cultured with mammalian cells. Cells were subsequently stained to distinguish between live and dead populations, and further analyzed by flow cytometry to determine therapeutic efficacy.

Progress

We employed an Olympus spinning disk confocal fluorescence microscope to examine the fluorescence of TrojanHorseβ[Reporter ζ] cultured under normoxic (21% O₂) and hypoxic (5% O₂) conditions. The results showed that the mean fluorescence intensity under 5% O₂ was significantly higher than that under 21% O₂, indicating that the hypoxia-inducible module was effectively activated in low-oxygen environments and exhibited strong signal responsiveness. These findings confirm the sensitivity and reliability of the module in responding to hypoxic stimuli, establishing a solid foundation for the subsequent development of hypoxia-triggered suicide switches and targeted anticancer delivery systems.

To evaluate whether nitrate supplementation could enhance bacterial metabolism under anoxic conditions, bringing it closer to the normoxic state, growth curves were measured under three conditions: normoxia without potassium nitrate, anoxia with potassium nitrate, and anoxia without potassium nitrate. The results showed that at 23 hours, the anoxic group supplemented with potassium nitrate exhibited approximately a 100% increase in growth compared with the anoxic group without nitrate. These findings suggest that potassium nitrate supplementation partially restored metabolic activity under oxygen-limited conditions, rendering the metabolic phenotype more similar to that observed under normoxia and potentially improving the colonization or invasiveness performance of the engineered bacteria.

R-M System Bypass

Restriction–Modification (R–M) systems represent a fundamental bacterial defense against foreign DNA, consisting of paired restriction endonucleases and DNA methyltransferases. Restriction enzymes cleave DNA at unmethylated recognition sites, while methyltransferases protect the host genome by site-specific methylation. This methylation-based discrimination enables efficient elimination of exogenous DNA and constitutes a major barrier to genetic transformation. As a result, targeted inactivation or evasion of host R–M systems (Types I–IV) is indispensable for successful bacterial engineering. In this study, we characterize Staphylococcus epidermidis ATCC 14990, Staphylococcus xylosus ATCC 29971, and Escherichia coli Nissle 1917 as chassis organisms. We delineate their strain-specific restriction barriers and propose strategies to overcome these obstacles, thereby facilitating efficient genetic manipulation and advancing their utility in synthetic biology.

Staphylococcus epidermidis ATCC 14990

In the standard laboratory strain Staphylococcus epidermidis ATCC 14990, we identified one functional Type I R–M system, one defective Type I R–M system, and a putative Type II R–M system. This information was derived from the REBASE Genome database, NCBI genomic data, BLAST analyses, and supporting literature.

1. Functional Type I R–M system:

   In Staphylococcus epidermidis ATCC 14990, we identified a functional Type I R–M system with the gene cluster arranged as hsdM–hsdS–hsdR, encoding the methyltransferase, sequence recognition, and restriction subunits, respectively. The HsdS and HsdM subunits form the minimal functional complex (S–2M), which subsequently recruits HsdR to form the complete restriction complex capable of cleaving unmethylated DNA (Loenen et al., 2014). To circumvent this barrier, we employed a strategy of heterologous expression of the hsdM–hsdS operon in Escherichia coli DH5α (dam⁺ dcm⁺), thereby allowing the HsdMS complex to pre-methylate host-specific recognition sites on plasmids prior to transformation. However, since the recognition sequence of the HsdMS complex is unknown, potential overlap with the E. coli DH5α dcm recognition motif could reduce methylation efficiency (Hermann and Jeltsch, 2003). To address this issue, we applied λ-Red homologous recombination to replace the chromosomal dcm locus with an hsdM–hsdS–KanR cassette, ensuring consistent and specific modification of plasmids.

2. Defective Type I R–M system:

   Based on the REBASE database and multiple literature reports, Type I R–M systems generally exhibit a conserved gene organization in which hsdM and hsdS are located adjacently, while hsdR is positioned at one end of the operon (Lee et al., 2019). This arrangement likely reflects functional requirements: the HsdS and HsdM subunits must form a complex as the minimal functional unit, and their proximity facilitates co-transcription and proper assembly (Loenen et al., 2014). In contrast, aberrant expression of HsdR could result in insufficient restriction or non-specific cleavage, necessitating tighter regulatory control through its separation at the operon boundary. Within the genome of Staphylococcus epidermidis ATCC 14990, we identified a gene cluster organized as hsdR–hsdS. Downstream of hsdS, however, a disrupted open reading frame was annotated as a pseudogene with labels such as “frameshifted,” “incomplete,” and “product = IS5/IS1182 family transposase.” We infer that the ancestral hsdM gene was interrupted by insertion of an IS element, resulting in a frameshift mutation and loss of hsdM function, while the residual sequence was automatically annotated as a transposase. Since the minimal functional complex of Type I systems requires the HsdM subunit, we conclude that this Type I R–M system in Staphylococcus epidermidis ATCC 14990 is non-functional and therefore does not pose a barrier to genetic transformation.

3. Type II RM system (putative):

   The functionality of an R–M system requires the restriction endonuclease (REase) and its cognate methyltransferase (MTase) to recognize identical DNA sequences. Because of this strict coupling, evolutionary divergence requires at least one component to alter its recognition sequence. We hypothesize that REases are highly conserved, as any alteration of their recognition specificity without a corresponding methyltransferase would be lethal to the host. By contrast, MTases can exist independently to provide epigenetic modification, often with redundancy across multiple enzymes, rendering their recognition sequences more susceptible to evolutionary change. Based on this reasoning, we propose that Staphylococcus epidermidis ATCC 14990 harbors a putative Type II R–M system, consisting of a Sau3AI-like REase and Dcm-like MTase. The former has been automatically annotated by NCBI as mutH. However, as the MMR (mismatch repair) pathway in staphylococci lacks mutH, and given that Sau3AI shares high similarity with MutH and recognizes the same GATC motif, we infer that this annotation is likely erroneous. The latter gene exhibits sequence similarity to dcm in BLAST analysis. We suggest that if its recognition specificity has co-evolved to align with that of the Sau3AI-like REase, the two proteins may together form a functional Type II R–M system with restriction activity. In summary, the functional status of this system remains unresolved; if active, it may discriminate the methylation state (m⁵C) of GATC sites. To circumvent potential restriction, we employed Escherichia coli DH5α (dam⁺), ensuring that all plasmid GATC sites are modified with m⁶A, thereby blocking Sau3AI restriction activity and preventing degradation of exogenous plasmids (Hermann and Jeltsch, 2003).

Staphylococcus xylosus ATCC 29971

The reference strain Staphylococcus xylosus ATCC 29971 is predicted to encode a functional Type IV R–M system. Distinct from Type I–III systems, Type IV R–M systems comprise solely restriction endonucleases that cleave DNA carrying foreign modifications. In the NCBI database, the putative gene associated with this system is annotated as a “DUF3427 domain-containing protein,” indicating the presence of a conserved yet functionally uncharacterized DUF3427 domain that has been identified across multiple taxa. Notably, the Type IV enzyme SauUSI in Staphylococcus aureus carries a DUF2427 domain at its C-terminus, which has been proposed to function as a target recognition domain (TRD) (Xu et al., 2011). DUF2427 is also frequently found within bacterial defense islands, often fused to the PD-(D/E)XK nuclease domain (Lutz et al., 2019). Such fusion proteins exhibit negligible endonuclease activity in vitro yet confer pronounced anti-phage resistance when expressed in Escherichia coli, suggesting an atypical role in recognizing modified DNA (Xu et al., 2011). In consideration of these findings, plasmids were propagated in E. coli JM110 (dam⁻ dcm⁻) to avoid host-derived DNA methylation and thereby minimize the risk of degradation by the putative Type IV R–M system present in S. xylosus ATCC 29971.

E. coli Nissle 1917 (EcN)

REBASE annotates eight R–M systems in Escherichia coli Nissle 1917 (EcN). Among these, one Type I system (recognition motif: RTACNNNNGTG) and one Type II system (recognition motif: GGTCTC) are the most likely to impose restriction barriers. Given that EcN is extensively employed as a probiotic strain and numerous reports describe successful cloning and transformation without explicit R–M evasion strategies, it is likely that these systems exhibit attenuated or incomplete functionality under standard laboratory conditions. To mitigate potential restriction, plasmids were initially propagated in E. coli JM110 (dam⁻ dcm⁻), thereby avoiding host-derived methylation. Should transformation efficiency in EcN remain suboptimal, further strategies will be considered, including site-directed mutagenesis to remove recognition motifs or in vitro methylation pretreatments to precondition plasmids against EcN-specific restriction.

Progress

1. Staphylococcus epidermidis ATCC 14990

   Using the λRed recombination system, we successfully engineered an Escherichia coli DH5α derivative in which the chromosomal dcm gene was replaced with a synthetic fragment (yedJ–J23100–RBS–kanR–T2–J23104–RBS–hsdM–hsdS–vsr). The recombinant strain was verified, confirming accurate genomic integration. The constructed plasmids were intended to be successfully expressed in Staphylococcus epidermidis ATCC 14990. However, despite employing various competent-cell preparation methods, transformation of the strain remained unsuccessful. Literature evidence later indicated that Staphylococcus epidermidis ATCC 14990 is poorly suited for electroporation-based transformation, providing a plausible explanation for these challenges and guiding our subsequent methodological adjustments.

2. Staphylococcus xylosus ATCC 29971

   Escherichia coli JM110 (dam⁻ dcm⁻) was selected as the plasmid propagation host to eliminate methylation patterns that might activate the Type IV restriction system. The shuttle vector pLI50 was amplified and purified from JM110 and subsequently transformed into glycine-treated S. xylosus ATCC 29971 cells. Colonies harboring the correct plasmid were successfully obtained, indicating that this approach effectively circumvented the host R–M system. The number of transformants remained very low, which is consistent with the generally poor transformation efficiency of S. xylosus ATCC 29971.

3. Escherichia coli Nissle 1917 (EcN)

   Plasmids propagated in E. coli JM110 were successfully expressed in E. coli Nissle 1917 (EcN), indicating that the methylation system of EcN does not interfere with gene expression.

Reference