Arming the living drugs: nanoengineered bacteria for precision tumor theranostics

Abstract

Bacteria-mediated tumor therapy exploits the innate ability of anaerobic bacteria to selectively colonize hypoxic tumor regions. However, clinical translation of this approach is hindered by three core challenges: insufficient monotherapy efficacy, the inability to non-invasively track bacterial biodistribution in vivo, and safety concerns. Nanotechnology offers solutions by enabling the construction of multifunctional nanoengineered bacteria. This review systematically summarizes six strategies for integrating nanomaterials with living bacteria—physical adsorption, covalent conjugation, bio-orthogonal chemistry, self-polymerization, biomineralization, and internalization—and critically analyzes how the interfacial properties of each strategy govern the resulting theranostic function. These nanoengineered systems address the key challenges by enabling multimodal imaging for real-time tracking, enhancing therapeutic efficacy through chemodynamic, photothermal, photodynamic, and immunotherapeutic modalities, and providing spatiotemporal control via external stimuli. We further examine the in vivo fate of these biohybrids, including immune clearance, tumor colonization, and the mechanisms for external and tumor-microenvironment-responsive activation. Safety considerations, including infection risk, cytokine release, nanomaterial immunogenicity, and regulatory requirements, are critically discussed. Finally, we outline future directions with an emphasis on adaptive coating design, hierarchical control and self-elimination strategies, and the manufacturing and standardization frameworks needed for clinical translation of this emerging class of anticancer theranostics.

Graphical Abstract

Nanoengineered Bacteria for Cancer Therapy：Strategies, Functions, and Clinical Translation.

Introduction

The hypoxic, immunosuppressed, and poorly vascularized tumor microenvironment (TME) severely impedes the penetration of chemotherapeutic drugs and nano-abiotic carriers. In contrast, various facultative or obligate anaerobic bacteria, such as Escherichia coli (E. coli) [1–4], Salmonella typhimurium [5, 6], and Bifidobacterium [7, 8], can preferentially colonize these hypoxic tumors through aerotaxis or chemotaxis pathways. Facultative anaerobes such as Salmonella, Escherichia, and Listeria target tumors through multiple mechanisms. They passively accumulate in tumors due to entrapment in the leaky, disordered tumor vasculature, actively migrate via chemotaxis toward tumor-derived compounds, and selectively proliferate within the nutrient-rich TME. Inflammatory cytokines like tumor necrosis factor (TNF) further enhance bacterial infiltration by increasing local blood flow. Chemotaxis is mediated by specific receptors, aspartate, ribose–galactose, and serine receptors, that direct Salmonella to viable, quiescent, and necrotic tumor regions, respectively. Auxotrophic strains such as Salmonella A1-R, which require leucine and arginine, achieve tumor specificity by thriving only in the amino-acid-rich TME while being unable to grow in normal tissues where these nutrients are scarce. Listeria employs an additional “Trojan horse” strategy by infecting myeloid-derived suppressor cells and other antigen-presenting cells, which transport the bacteria into the tumor. Once inside the immune-suppressive tumor microenvironment (TME), the bacteria evade clearance by residing within these host cells, whereas in normal tissues they would be rapidly eliminated. The deep penetration of bacteria into tumors is further aided by flagella-mediated active propulsion [9, 10]. Escherichia coli Nissle 1917 (EcN), known for its antimicrobial effects and gut microbiota regulation [11], exemplifies a facultative anaerobe used clinically (Mutaflor^®) against inflammatory bowel disease and gastrointestinal (GI) tumors [12–14].

Despite these advantages, the clinical advancement of bacteria-based tumor therapy is hindered by three challenges (Fig. 1). First, the efficacy of bacteria-targeted tumor therapy is compromised by dose-dependent toxicity and difficulties in precise dosage control, which limit the efficacy of bacterial monotherapy [15]. Second, a major translational hurdle is the inability to precisely localize and quantify bacterial populations in vivo [16, 17]. Currently, most trials rely on invasive tissue biopsied or clinical signs, which are invasive and lack real-time capability. Third, there are inherent safety concerns, including the risk of virulence reversion, especially with genetic modifications.

Fig. 1.

The main challenges of bacteria-targeted therapy

Progress in synthetic biology has partially addressed the safety and efficacy concerns, for example, by creating attenuated strains like Salmonella VNP20009 [18] or engineered bacteria to release CCL21 chemokine, cytokines, angiogenesis inhibitors, antigens, and antibodies therapeutic agents [19–29]. However, synthetic biology alone struggles with the challenge of real-time tracking and can introduce risks of genetic instability [30–33]. Non-invasive imaging techniques offer a superior alternative for tracking, allowing for repeated, comprehensive assessments and minimizing sampling error. In recent decades, various genetically encoded reporters, such as phytochrome-based proteins for photoacoustic imaging (PAI) [34], fluorescent proteins for fluorescence imaging (FI) [30], and magnetosome gene clusters for magnetic resonance imaging (MRI) [35], have enabled long-term tracking of engineered bacteria. However, these systems have limitations including laborious and inefficient genetic editing, unstable gene circuits that are prone to evolutionary failure and the potential for ecological escape, which raises biosafety concerns [31, 36–38].

Nanotechnology offers a powerful parallel strategy to overcome these hurdles. By integrating nanomaterials with exceptional catalytic, optical, thermal, acoustic, and magnetic properties, we can create “nanoengineered bacteria.” This approach enables three key advancements, Functionalization: nanomaterials equip bacteria with potent therapeutic modalities (e.g., photothermal, photodynamic) to overcome the limited efficacy of monotherapy. Visualization: imageable nanomaterials (e.g., NIR dyes, magnetic particles) allow for non-invasive, real-time tracking of bacterial distribution and density, addressing the “see” challenge. Controllability: nanomaterials can confer responsiveness to external stimuli (e.g., light, magnetic fields), enabling spatiotemporal control over bacterial activity and improving safety.

To date, several excellent thematic reviews have surveyed bacteria-based biohybrids from the perspectives of therapeutic applications, nanomaterial types, or specific bacterial chassis have been published [32–41]. However, a comprehensive systematically deconstructs the six fundamental conjugation strategies, critically compares their impact on bacterial physiology and interfacial stability, and then rationally links each strategy to its clinical theranostic functions is conspicuously absent.

The rapid diversification of conjugation tools in recent years, from bioorthogonal click chemistry to transporter-mediated internalization, has created an urgent need for such a unifying framework. Therefore, this review begins by systematically introducing a spectrum of fabrication methods for engineering bacteria (Fig. 2) from physical adsorption, covalent conjugation, bio-orthogonal chemistry, self-polymerization, biomineralization to internalization, analyzing their respective advantages, limitations, and impact on bacterial function. Subsequently, we provide a detailed insight into nanoengineered bacteria-mediated tumor visualization and therapeutics, mapping each application back to the core challenges they address. Finally, we conclude with a forward-looking perspective on potential improvements, aiming to establish guidelines for developing new multifunctional bacteria against cancer. This review fills that gap by providing a ‘from chemistry to function’ blueprint that guides the rational selection and design of nano-bacteria interfaces to overcome the core challenges of tumor theranostics.

Fig. 2.

Nanoengineered bacteria constructed by various methods (electrostatic interaction, covalent linking, bioorthogonal chemistry, etc.) to overcome the core challenges of bacterial therapy by enabling multifunctional tumor imaging and therapy

Nanoengineered bacteria construction, fate in vivo, and controllability

Strategies for nanoengineered bacteria

Multifunctional nanoengineered bacteria are constructed by coating bacterial surfaces with nanoparticle-based payloads. These systems utilize the tumor-targeting and tissue-penetrating capabilities of live bacteria to actively deliver theranostic agents deep into tumors. This section systematically reviews the construction strategies, emphasizing the nature of the interface and its implications for stability and function (summarized in Table 1).

Table 1.

Summary of construction strategies for multifunctional nanoengineered bacteria

Fabrication method	Nanomaterials/bacteria	Application	Ref
Electrostatic interaction	DOX@SP	FL-guided tumor therapy, drug delivery		[44]
	AIEgen@EcN	NIR-FL-guided PDT, PTT		[114]
	DSPE-PEG-AIE PSs@E.coli	NIR-FL-guided PDT		[46]
	PEI/DNA@VNP20009	Immunotherapy		[47]
	PAH/MIL-101@SO	Chemotherapy		[51]
	BPQDs@E.coli	PDT		[52]
	UCNP-PEG@E.coliMG1655	NIR-responsive protein therapy		[49]
Covalently linking	Nanobeads@VNP20009	Drug delivery		[56]
	Conjugated polymer@E.coli	PDT		[54]
	DOX@EcN	Drug delivery		[59]
	DOX@GDH-E.coli	ROS-mediated therapy		[60]
	NHS-N782, JQ-1@VNP20009	Immunotherapy, drug delivery		[62]
	Ce6@AMB-1	PDT, biotherapy		[61]
	INPs@YB-1	FL-guided PDT		[63]
	PEG-CuP-COF@∆St	CDT, SDT, Immunotherapy		[64]
	Fe3O4@Ec-pE	CDT		[66]
	NH2-AS1411@EcN	Immunotherapy		[67]
	LP@MTB	Drug delivery		[68, 69]
	PFH/PLGA@B.longum	Ultrasound irradiation therapy		[115]
	HCNs@YB-1	SDT, PDT		[73]
Bio-orthogonal chemistry	Fe-Bac-HlpA/CD47nb-BLPs	Immunotherapy		[75]
	Sgc8c-MMAE@VNP	Immunotherapy, drug delivery		[76]
	Antigen@ Salmonellae	Immunotherapy		[78]
	DOX@Lmat	Chemotherapy, immunotherapy		[79]
Self-polymerization	pDA@VNP20009	PTT, biotherapy		[80]
	Antigen-PDA@EcN	Immunotherapy		[82]
	PD1-S1-PDA@EcN	Immunotherapy		[83]
Biomineralization	Au@Bac	Toxin therapy		[86]
	Au@TAB	Cytolysin A, PTT		[87]
	MnS@SRB	Immunotherapy		[93]
	FeS@MR-1	CDT, immunotherapy		[91]
	CuS@VNP20009	immunotherapy		[92]
	FeS@SRB	CDT, drug delivery		[94]
	MOF@Bac	PTT		[102, 103]
Internalization	GP-Si-ICG@E.coliVNP	PTT, immunotherapy		[110]
	ICG-MD-NPs@E.coli DH5α	PTT		[111]
	GP-Si-BPs, GP-Si-Luc@Bac	BLI-guided PTT		[104]

Noncovalent physical adsorption

Physical adsorption, relying on electrostatic, hydrophobic, or affinity interactions, is the most straightforward method for functionalizing bacteria.

1) Electrostatic interaction

Due to negatively charged components like teichoic acid and lipopolysaccharide (LPS), the bacterial cell wall carries a negative charge [42, 43], which facilitates the direct adsorption of various cationic drugs, polymers, and nanoparticles (Fig. 3). For example, positively charged doxorubicin (DOX) can be loaded onto Spirulina platensis (SP) for lung metastases treatment (Fig. 4a) [44, 45]. Cation polymers like poly-ethylenimine (PEI) can act as a bridge to anchor anionic nanoparticles, such as AIE photosensitizer-loaded liposomes, on E.coli (Fig. 4b-c) [46, 47]. DSPE-PEG-NH₂ or NH₂-PEG-COOH modified nanoparticles to produce a positive surface charge, for example, super-small barium titanate [48] and lanthanide upconversion nanoparticles (UCNP) [49] (Fig. 4d) were conjugated onto the bacteria (EB) surface to construct nanoengineered bacteria by electrostatic coupling. Conversely, bacteria can be pre-cationized with poly(allylamine hydrochloride) (PAH) to then adsorb anionic nanoparticles like MIL-101 metal-organic framework (MOF) nanoparticles [50–52]. While simple and rapid, this method’s main limitation is the potential for payload desorption in vivo due to changes in ionic strength or pH, which can lead to premature drug release and expose bare bacteria to the immune system.

Fig. 3.

Fabrication pathways for nanoengineered bacteria via electrostatic interaction

Fig. 4.

Preparation of nanoengineered bacteria through electrostatic interaction force. (a) The negatively charged surface of SP containing with channels and pores can absorb positively charged DOX. Reproduced with permission [44]. Copyright 2020, Wiley-VCH. (b) The illustration of the TDNP nanoparticle formation process and the construction of TDNPP-E.coli nanoengineered bacteria. Reproduced with permission [46]. Copyright 2019, American Chemical Society. (c) Coating attenuated bacteria with DNA-loaded cationic polymers to construct nano-engineered bacteria. Reproduced with permission [47]. Copyright 2015, American Chemical Society. (d) The NH₂-PEG-COOH modified UCNP were anchored on EB surface to construct MG1655/HlyE-sfGFP@UCNP@PEG nanoengineered bacteria via electrostatic interaction, the nanohybrids release HlyE perforin destroy the membrane of cancer cells under NIR laser irradiation [49]. Copyright 2023, Elsevier LTD

2) Hydrophobic forces and hydrogen bonds

These forces can work synergistically with electrostatics. Amphiphilic conjugated polymers (CPs), for instance, can insert into the bacterial membrane via synergistic electrostatic and hydrophobic interactions forces (Fig. 5) [43, 53–55]. Similarly, the strong and specific biotin-streptavidin interaction provides a robust non-covalent linkages, used to attach poly (lactic-co-glycolic acid) (PLGA) nanoparticles to Salmonella VNP20009 [56].

Fig. 5.

The schematic illustrations of the two different tumor cells death strategies of CPs@E.coli and CPs@E.coli/toxin engineered bacteria. The chemical structures of CP and antibiotic polymyxin B (PLB). Reproduced with permission [54]. Copyright 2013, Wiley-VCH

Covalent linking

Covalent conjugation offers a more stable bond than physical adsorption, crucial for preventing payload detachment in the complex in vivo environment. The bacterial cell wall presents abundant functional groups like amino, carboxyl, phosphate, and sulfhydryl moieties [57, 58], enabling direct chemical linkage (Fig. 6).

Fig. 6.

Various functional groups on bacteria surface facilitate covalent bond formation for nanoengineering bacteria construction

1) Amide bonds

The most common approach involves forming amide bonds between carboxylated payloads and surface amines. For example, DOX has been linked to E. coli Nissle 1917 via an acid-sensitive linker, enabling tumor-targeted delivery with triggered release (Fig. 7a) [59, 60]. Similarly, the photothermal dyes (NHS–N782) [61] and protein inhibitors (JQ-1-COOH) have been co-conjugated to Salmonella VNP20009 for combined PTT and immunotherapy (Fig. 7b) [62]. This method can also be used to attach entire nanoparticles, such as PLGA-PEG-COOH-encapsulated indocyanine green (ICG) photosensitizer nanoparticles (INPs), to bacteria for enhanced photothermal therapy (Fig. 7c) [63]. Dual-amide-terminated polyethylene glycol polymers (NH₂-PEG2000-NH₂, MW = 200 Da) functionalized nanoparticles integrated with bacteria [64, 65]. For example, NH₂-PEG2000-NH₂ modified magnetic Fe₃O₄ nanoparticles (MNP) covalently with N-acetylmuramic acid residues abundantly present in the cell wall of Gram-negative bacteria, resulting in the formation of Ec-pE@MNP engineered bacteria by amination reaction (Fig. 7d). This system combines the innate tumor-targeting capability of bacteria with external magnetic guidance for enhanced localization. Within the tumor, a Fenton-like reaction occurs between the bacterially generated H₂O₂ and the MNPs, producing highly cytotoxic •OH radicals that effectively induce tumor cell death [66].

Fig. 7.

Preparation of nanoengineered bacteria by amine reaction. (a) Cis-aconitic anhydride modified DOX was covalently linked to the E. coli Nissle 1917 bacteria surface via amine action. Reproduced with permission [59]. Copyright 2017, Wiley-VCH. (b) The schematic illustration of the construction of N-V-J biohybrid system and its mechanism of tumor-targeting PTT. Reproduced with permission [62]. Copyright 2022, American Chemical Society. (c) The INPs are conjugated onto YB-1 through amide bond; Reproduced with permission [46]. Copyright 2019, Wiley-VCH. (d) The NH₂-PEG-NH₂ modified magnetic Fe₃O₄ nanoparticles (MNP) covalently linked with N-acetylmuramic acid of E. coli via amination reaction to construct Ec-pE@MNP, realizing bacteria-targeting tumor CDT. Reproduced with permission [66]. Copyright 2019, Wiley-VCH

In addition to equipping bacteria with novel theranostic capabilities, such as imaging/drugs/photosensitizers/photothermal agents, the conjugation of bacteria with targeting ligand represents a promising strategy to enhance their tumor-targeting accumulation. For example, Tan’s group developed a one-step amination method to conjugate aminated aptamers (NH₂-AS1411) to the carboxyl group of N-acetylmuramic acid existing in the cell wall of Gram-negative bacteria. The obtained aptamer-conjugated bacteria (ApCB) exhibited enhanced tumor localization capability. Besides, This aptamers conjugated approach not only enhanced antitumor efficacy but also highly activated immune responses within the tumor microenvironment [67].

A key advantage is the stable linkage, which minimizes systemic toxicity. However, care must be taken to ensure that the chemical conjugation does not block essential surface structures like flagella, which are vital for motility and tumor penetration, studies confirm that with optimized chemistry, bacterial viability and targeting ability can be preserved.

2) Carbodiimide chemistry

This is a specific type of amide bond formation used to conjugate carboxylated liposomes or drug-loaded PLGA nanoparticles to bacteria surfaces [68, 69]. It offers a controlled way to link pre-formed nanoparticles to bacteria. For example, carboxylated liposomes (DSPE-PEG-COOH, LP) were conjugated to Magnetococcus marinus MC-1 magnetotactic bacteria (MTB) through carbodiimide chemistry, forming MTB-LP nanoengineered bacteria. These MTB-LP nanoengineered bacteria can be directed to tumor hypoxic regions through their innate magnetotaxis and external magnetic guidance [69].

Bioorthogonal chemistry

Bioorthogonal reactions provide a highly specific and efficient method with minimal disturbance to the surrounding biological environment in vivo, offering numerous advantages for drug delivery [70]. Among these reaction, the click reaction between azide (N₃) and dibenzocyclooctyne (DBCO) is considered one of the most efficient bioorthogonal reactions in vivo and has been validated in animal model systems [71]. Leveraging this chemistry, a bacterial nanoengineering strategy usually employs a two-step process: first, -N₃ group is introduced onto the bacterial surface, typically via metabolic glycoengineering of the bacteria themselves. Subsequently, a complementary DBCO group pre-anchored on the functional materials enables a rapid, site-specific “click” reaction (Fig. 8). For example, N₃-modified hyaluronic acid (HA) hydrogel was shown to enhance the tumor-targeting capability of DBCO-engineered bacteria [72]. Sono-photosensitive engineered bacteria (YB1-HCNs) were constructed by conjugating DBCO-modified polymers to N₃-labeled bacteria (Fig. 9a) [73], and DBCO-modified photosensitive molecular (IN) was linked to D-ala-N₃ labeling VNP to construct VNP-PD-L1@IN through biorthogonal chemistry [74]. Furthermore, the method enables the construction of sophisticated “smart” systems. One representative system involves clicking DBCO-modified Fe₃O₄ magnetic nanoparticles onto N₃ moieties of bacteria surface (Fig. 9b). Upon exposure to an alternating magnetic field (AMF), these Fe₃O₄ NPs convert magnetic energy into heat, which acts as a signal to trigger the expression of a therapeutic protein CD47nb from a heat-inducible promoter within the bacteria [75]. Overall, this biorthogonal approach offers unparalleled specificity and mild reaction conditions, preserving bacteria viability. Its complexity primarily lies in the requirement for metabolic labeling or genetic engineering to introduce the initial chemical handle; nevertheless, the outcome is a highly stable and functional biohybrid [76–79].

Fig. 8.

Metabolic labeling bio-orthogonal groups onto the bacterial surface facilitate

Fig. 9.

Nanoengineered bacteria constructed by bioorthogonal reaction. (a) Preparation of YB1–HCNs nanoengineered bacteria realized bacteria-targeted sono–photodynamic therapy and released HCNs induce immunotherapy; Reproduced with permission [73]. Copyright 2025, American Chemical Society. (b) Scheme of site-specific coupling of DBCO-Fe₃O₄ onto the bacteria surface, and the TEM images of Fe₃O₄NPs@bacteria through site specific vs. non-site specific coupling method; Reproduced with permission [75]. Copyright 2023, The Springer Nature

Self-polymerization

The self-polymerization of dopamine into polydopamine (PDA) under mild, alkaline conditions provides a versatile coating method for bacterial nanoengineering. The PDA layer forms conformally on the bacteria surface, providing intrinsic photothermal properties [80]. More importantly, its rich catechol and amine groups serve as a reactive platform for secondary functionalization with PEG, drugs, or antibodies in a single step.

For example, tumor-specific antigens and checkpoint-blocking antibodies are simultaneously immobilized by PDA coupling agent for targeted immunotherapy [81]. In addition to serving as a coupling agent, PDA exerts a photothermal effect that promotes the repolarization of tumor-associated macrophages toward a pro-inflammatory phenotype. This secondary functionalization capability significantly expands the therapeutic efficiency of engineered bacteria, allowing the incorporation of diverse agents, such as PEG for prolonged circulation, chemotherapeutic drugs for localized release, and immune checkpoint inhibitors to boost antitumor immunity. Ultimately, this approach elevates bacterial carriers from simple monotherapies to combination treatment [82, 83].

Biomineralization

Biomineralization leverages the innate metabolic machinery of bacteria to synthesize nanomaterials in situ. Metal-reducing bacteria like Shewanella oneidensis MR-1 can enzymatically reduce metal ions (e.g., Pt³⁺, Pd²⁺, Au³⁺, Ag⁺, Fe³⁺) to form nanoparticles (Au) on their surface (Fig. 10a) under anaerobic conditions through respiratory-associated electron transfer processes [84–88]. This “green synthesis” approach is gentle and results in a tight integration of the material with the cell wall [89, 90]. In other hand, the bacterium is also biosynthesis of sulfides like FeS [91], CuS [92], and MnS [93], leveraging its respiratory electron transfer chain to reduce thiosulfate [91, 92, 94, 95]. The resulting FeS@MR-1 nanoengineered bacteria not only drives a continuous Fenton reaction by promoting Fe(III) to Fe(II) conversion, thereby enhancing CDT, but also reprograms the immunosuppressive tumor microenvironment through lactate consumption and functions as a natural immune adjuvant to enhance antitumor immune responses (Fig. 10b).

Fig. 10.

Preparation of nanoengineered bacteria by biomineralization reaction. (a) The schematic illustration of the formation of Bac@Au nanoengineered bacteria and its TTX releasing process for antitumor therapy. Reproduced with permission [86]. Copyright 2021, American Chemical Society. (b) The schematic illustration of the synthesis process of FeS@MR-1; TEM images of MR-1 and FeS@MR-1; The XPS curve of FeS@MR-1. Reproduced with permission [91]. Copyright 2024, American Chemical Society. (c) The construction of the PTB@ZIF-90/MB and its tumor-targeted PTT tumor therapy; Reproduced with permission [103]. Copyright 2020, Wiley. (d) The schematic illustration of the E.coli@ZIF-8/C&D nanoengineered bacteria. Reproduced with permission [102]. Copyright 2020, Wiley

Another approach is mineralize porous inorganic materials, MnO₂ [96, 97], metal-organic frameworks (MOFs) [98], iron oxides (Fe_xO_y) [97], and calcium phosphates (CaP) [99, 100]. around the bacterium, which can then be loaded with drugs for controlled release [86, 101]. For example, the porous zeolitic imidazolate framework-90 (ZIF-90) loading with the photosensitizer methylene blue (MB) has been mineralized onto live photothermal bacteria (PTB) to enhance the efficacy of PTT (Fig. 10c). Under light irradiation, ZIF-90/MB selectively releases MB into mitochondria, inducing mitochondrial dysfunction through singlet oxygen (¹O₂) generation [102]. In another study, Liu and colleagues employed a biomimetic mineralization strategy to coat E. coli (MG1655) with a ZIF-8 metal–organic framework, which was such as simultaneously loaded with DOX (D) and the photosensitizer chlorin e6 (Ce6) (Fig. 10d) [103].

Internalization

By functionalizing nanoparticles with maltodextrin (MD) or glucose polymers (GP), they can be recognized and actively transported into the bacterial cytoplasm via ATP-binding cassette (ABC) transporters. For example, a GP-conjugated ICG-loaded silicon nanoparticle system (GP-ICG-SiNPs) (Fig. 11a) was designed to exploit this pathway. The GP-ICG-SiNPs are efficiently internalized by facultatively anaerobic bacteria such as VNP20009 and Escherichia coli via the bacteria-specific ABC transporter, constructing the Trojan bacteria system [98, 99]. Similarly, MD-ICG-SiNPs were internalized into thermal bioswitch-engineered Escherichia coli (DH5α) via the ABC transporter; the same strategy has also been successfully applied to gold nanoparticles (AuNPs) and carbon dots (CDs) (Fig. 11b) [100]. This approach has been used to deliver imaging probes for highly sensitive bioluminescence imaging (Fig. 11c) [104, 105]. A key advantage of internalization over surface attachment is that it shields the nanomaterial cargo from extracellular enzymatic degradation and premature immune recognition, while preserving critical bacterial surface structures such as flagella, which are indispensable for motility and tumor targeting. This method achieves high payload loading and protects the cargo from the external environment until it reaches the tumor site.

Fig. 11.

Nanoengineered bacteria constructed by the bacteria-specific ABC transporter-mediated materials endocytosis. (a) The synthesis of GP-ICG-SiNPs nanoagents and its internalizing into bacterial cells by ABC transport; Reproduced with permission [110]. Copyright 2022, Springer Nature. (b) The schematic illustration of MD-mediated the internalized of ICG-MD-NPs nanoagents; Reproduced with permission [111]. Copyright 2022, Wiley. (c) Schematic illustration of ABC transporter-mediated the internalization of bioluminescent nanoprobes into bacteria, and the mechanism of bioluminescence imaging of in vivo of bacteria. Reproduced with permission [104]. Copyright 2023, Springer Nature

However, the post-internalization fate of the nanoparticles and their impact on bacterial physiology must be carefully assessed. Inorganic nanoparticles such as silicon-based particles, AuNPs, and CDs generally remain structurally intact within the cytoplasm, serving as stable functional materials [106–108]. For example, Glucose-polymer-modified fluorescent silicon nanoparticles (GP-SiNPs) are rapidly internalized into both Gram-negative and Gram-positive bacteria via the ABC transporter pathway, remaining structurally intact and retaining their fluorescence functionality within the bacterial cytoplasm [109]. Regarding bacterial viability, studies have shown that the internalized bacteria maintain normal proliferation, motility, and tumor-colonizing capability. Its major advantage is the protection of the cargo, but it requires the bacterium to possess the specific transporter, and the internalized material must not be toxic to the bacteria itself.

Except for the transporter-mediated pathways for material internalization, the intrinsic permeability of bacterial membranes offers an alternative route for functional molecule entry. For instance, cyanobacteria possess membranes that are naturally permeable various nutrients (e.g., glucose, uronic acids), dissolved gases (e.g., O₂, CO₂, NH₃), and hydrophilic biomolecules typically below 1000 Da. Moreover, the negatively charged cyanobacterial membrane facilitates the penetration of low polarity, positively charged molecules [112]. While photosynthetic microorganisms such as microalgae and cyanobacteria differ fundamentally from traditional tumor-colonizing bacteria in their oxygenic metabolism, they share a core set of advantages including amenability to surface modification, genetic programmability, and intrinsic optical/functional properties that position them within the same conceptual family of nanoengineered living platforms for tumor theranostics. For example, the cyanobacterium Synechococcus elongatus PCC 7942 was shown to internalize PEGylated photosensitizer chlorin e6 (Ce6), resulting in Ce6-loaded cyanobacteria (designated ceCyan) [113]. Under 660 nm laser irradiation, these engineered cyanobacteria continuously produce O₂ during the photosynthesis process, while the internalized Ce6 converts this O₂ into cytotoxic ¹O₂, enabling synergistic PDT.

A critical strategy comparison

The preceding sections have individually detailed the mechanisms and applications of six major nano-engineering strategies. However, the selection of an optimal strategy for clinical translation is not straightforward and requires a critical, cross-strategy comparison based on several interconnected criteria: stability, viability, immunogenicity, manufacturability, and safety (Table 2). A cross-strategy comparison reveals a fundamental trade-off: strategies that provide the most robust, high-stability coating often do so at the expense of bacterial metabolic activity. For example, covalent conjugation and self-polymerization form dense, chemical or mechanically stable shells, but the reaction conditions (e.g., alkaline pH for dopamine polymerization, reactive chemical handles) and the barrier to nutrient diffusion can significantly impair viability. Conversely, physical adsorption, while preserving near-native viability due to its mild nature, suffers from rapid payload desorption triggered by the complex ionic and proteinaceous environment in vivo, effectively nullifying the engineering effort.

Table 2.

Comparative analysis of nanoengineering strategies for living bacteria in tumor theranostics

Strategy	Stability	Bacteria viability	Immune clearance	Reproducibility	Scalability	Toxicity
Physical adsorption	Low; prone to desorption in vivo due to ionic/pH changes	High; mild conditions preserve viability	Low protection; bare bacteria exposed	High; simple mixing	High; rapid and simple	Low systemic toxicity, but premature payload release risks off-target effects
Covalent conjugation	High; stable covalent bonds	Moderate-High; dependent on reaction conditions	Moderate; exposed bacterial surfaces still immunogenic	Moderate; requires optimization of reaction stoichiometry	Moderate; multi-step chemistry adds complexity	Low-Moderate; risk of blocking motility-related structures
Bioorthogonal chemistry	High; site-specific and stable	High; mild, biocompatible reaction conditions	Moderate; depends on coating material	Moderate; requires metabolic engineering to introduce handles	Low-Moderate; two-step process with metabolic modification	Low; minimal interference with native biology
Self-polymerizatio	High; conformal and robust coating	Moderate; may restrictnutrient/waste exchange	Moderate-High; PDA provides a protective barrier	High; polymerization is self-driven andconsistent	High; one-step reaction in mild alkaline conditions	Low-Moderate; PDA is biocompatible, but thick coatings may impair division
Biomineralization	High; materials tightly integrated with cell wall	High; utilizes natural metabolic pathways	Moderate-High; mineral shell shields bacteria from immune recognition	Moderate; depends on bacterial metabolic activity and culture conditions	Low-Moderate; batch variability due tobiological factors	Low-Moderate; "green synthesis" minimizes chemical contaminants; heavy metal leaching must be assessed
Internalization	Very High; cargo protected within cytoplasm	High; preserves surface structures (flagella) and viability	High; internalized cargo shielded from extracellular environment and immune recognition	Moderate; dependent on transporterexpression levels	Low-Moderate; requires specific transporterand compatible cargo design	Low; avoids surface modification toxicity, but intracellular nanoparticle fate must be assessed

Strategies that best balance this trade-off are those that form stable interfaces under cytocompatible conditions. Bioorthogonal chemistry achieves high-stability, site-specific conjugation through metabolically inert “click” reactions that minimally disturb bacterial physiology. Similarly, biomineralization creates a semi-permeable, protective shell through a “green synthesis” process driven by the bacteria’s own metabolism. For therapeutic applications requiring sustained, active bacterial metabolism (e.g., continuous intratumoral enzyme or cytokine production), mild biomineralization and bioorthogonal labeling are therefore superior to highly cross-linked covalent coatings or dense polymer shells.

Physical adsorption strategies are most limited by uncontrolled immune clearance, as the naked bacterial surface remains fully exposed to opsonization and pre-existing antibodies. Covalent and self-polymerization strategies offer partial shielding but may introduce neo-antigens from chemical linkers or activate complement (particularly by positively charged surfaces). Bioorthogonal and internalization approaches provide the most effective stealth, with the former masking the surface with biocompatible polymers and the latter completely preserving the native, evolutionary-optimized bacterial surface. Beyond acute cytokine storm or sepsis, long-term toxicity from the nanocomponent itself is a largely unexplored risk. A major limitation for biomineralization and self-polymerization is the slow degradation of inorganic or synthetic polymer shells into fragments that may be nephrotoxic or bioaccumulate in the reticuloendothelial system. The long-term safety of introducing chemically cross-linked, non-degradable architectures onto a living, self-replicating vector is unknown and represents a critical knowledge gap. Internalization uniquely raises the concern of nanoparticle-induced genotoxicity or metabolic disruption within the bacterial host itself. In conclusion, this comparative analysis demonstrates that no single nanoengineering strategy is universally optimal. The choice of method must be dictated by the specific clinical context. For applications where strong, real-time imaging contrast is paramount and short-term action is sufficient, covalent conjugation of a probe is ideal. For long-term, self-sustaining metabolic therapy, the biocompatibility of bioorthogonal or biomineralization approaches is essential. Acknowledging these context-dependent trade-offs is a critical step toward rational, translation-focused design of next-generation living bacterial theranostics.

The fate of nanoengineered bacteria in vivo

Host immune clearance

For bacteria to effectively reach and colonize tumors following systemic administration, they must first survive the formidable clearance mechanisms of the host bloodstream. The innate immune system rapidly eliminates circulating bacteria through a cascade of events: complement activation via the alternative and lectin pathways, deposition of opsonins such as C3b and immunoglobulins, recognition by Fc and complement receptors on hepatic Kupffer cells and splenic macrophages, and subsequent phagocytic destruction [116]. This clearance can reduce the viable bacterial burden in circulation by several orders of magnitude within minutes, severely compromising tumor-targeting efficiency and necessitating higher, potentially toxic, doses. Overcoming this barrier is therefore a prerequisite for clinical translation, and the nanoengineering strategies described in Sect. "Nanoengineered bacteria construction, fate in vivo, and controllability" offer distinct, tunable approaches to attenuate these host defenses.

Steric shielding via polymer conjugation

Covalent attachment of hydrophilic polymers, most notably polyethylene glycol (PEG), creates a hydrated steric barrier around the bacterial surface that physically hinders the approach of large opsonins and prevents receptor engagement on phagocytes [117]. As described in Sect. "Strategies for nanoengineered bacteria", the NH₂-PEG2000-NH₂-modified Fe₃O₄ nanoparticles conjugated to E. coli (Ec-pE@MNP) not only enable chemodynamic therapy but also impart a “stealth” effect that reduces macrophage uptake [66]. Similarly, PLGA-PEG-COOH nanoparticles anchored to Salmonella YB1 (INPs-YB1, Sect. "Strategies for nanoengineered bacteria") prolong bacterial circulation by minimizing protein adsorption [63]. However, a critical concern is the increasing prevalence of pre-existing anti-PEG antibodies in the human population, which can trigger accelerated blood clearance upon repeated administration; strategies using alternative polymers such as polyzwitterions or polyoxazolines are being explored to mitigate this risk.

Biomimetic camouflage through surface coating

Conformal coatings that physically mask the antigenic determinants on the bacterial cell wall provide an alternative route to immune evasion. The self-polymerized polydopamine (PDA) layer on Salmonella VNP20009 (pDA-VNP, Sect. "Strategies for nanoengineered bacteria") forms a robust, biocompatible shell that shields surface lipopolysaccharide and flagella from pattern-recognition receptors and complement components, resulting in reduced opsonization and prolonged blood residence [80]. Likewise, the biomineralized metal–organic framework ZIF-8 shell on E. coli (E.coli@ZIF-8/C&D, Sect. "Strategies for nanoengineered bacteria") acts as a physical barrier that can delay immune recognition until the carrier reaches the tumor microenvironment where the pH-responsive shell degrades [102]. However, the potential immunogenicity of the coating material itself and its degradation products must be carefully characterized.

Preservation of native surface via internalization

The internalization strategy (Sect. "Strategies for nanoengineered bacteria") tackles the clearance problem from a fundamentally different angle. By actively transporting nanoparticle cargos into the bacterial cytoplasm via ABC transporters, the external surface of the bacterium remains in its native, evolutionary-optimized state, complete with functional flagella and chemotactic machinery [104]. This completely avoids introducing exogenous chemical moieties or surface coatings that could serve as neoantigens or activate the complement cascade. While this approach does not actively suppress clearance of the carrier bacterium itself, it ensures that the therapeutic cargo is shielded from extracellular degradation and that the bacterial targeting capacity is preserved until the tumor is reached.

Charge and chemical modulation

Complement activation and phagocyte recognition are highly sensitive to surface charge and chemistry. Positively charged coatings tend to adsorb anionic opsonins and activate complement more readily, whereas neutral or slightly zwitterionic surfaces exhibit lower clearance rates. The choice of coupling chemistry, whether electrostatically adsorbed cationic polymers or covalently bound neutral linkers, therefore directly influences the degree of opsonization and the bactericidal activity of serum components. Many of the covalent conjugation and biomineralization approaches described above achieve the dual benefit of functional loading and immune evasion through precisely tuned surface neutrality.

It should be emphasized that complete evasion of immune clearance is neither possible nor necessarily desirable; an efficient oncolytic bacterial platform must strike a delicate balance between evading premature destruction in the blood and retaining sufficient immunostimulatory activity to prime anti-tumor immunity upon tumor colonization. Coatings that fully shield Pathogen-Associated Molecular Patterns (PAMPs) may inadvertently abolish the adjuvant effects that make bacteria such potent immunotherapeutic agents (Sect. "Bacteria as intrinsic immunotherapeutic agents"). Future designs will therefore require dynamic, stimuli-responsive coatings that degrade specifically within the tumor microenvironment to unmask the bacterial surface and unleash its full immunostimulatory potential. Incorporating these principles into the rational design of nanoengineered bacteria, as systematically reviewed here and informed by critical pharmacokinetic analyses, is essential for translating bacterial therapeutics from promising preclinical observations to effective clinical realities [117].

Controllability of nanoengineered bacteria: external stimuli and environmental triggers

The clinical translation of live bacterial therapeutics is fundamentally contingent on the ability to control their behavior in vivo—where they go, when they act, and when they are eliminated. Uncontrolled bacterial proliferation or systemic distribution can lead to severe sepsis-like symptoms, even with highly attenuated strains. Nanoengineering offers a unique method for achieving spatiotemporal precision control high intratumoral activity with minimal systemic toxicity. These controllability mechanisms fall into two complementary categories: external physical stimuli that can be remotely applied, and tumor-intrinsic biochemical that trigger localized activation.

Magnetic guidance and targeting

One of the most direct and clinically relevant forms of control is magnetic actuation. By decorating bacteria with superparamagnetic Fe₃O₄ nanoparticles (as described in Sect. "Strategies for nanoengineered bacteria"), their systemic distribution can be manipulated using external magnetic fields. For example, the magnetotactic bacterium Magnetococcus marinus MC-1 conjugated with drug-loaded liposomes (MTB-LP) can be steered toward hypoxic tumor regions by combining its innate magnetotaxis with externally applied magnetic fields [68]. Similarly, non-magnetotactic bacteria such as E. coli coated with Fe₃O₄ NPs (Ec-pE@MNP) can be magnetically guided to tumors, where they perform chemodynamic therapy [65]. Magnetic control thus addresses the twin challenges of poor tumor accumulation and off-target colonization, providing a physical, instantly responsive steering mechanism.

Light-triggered activation and therapy

Light offers an exceptionally versatile trigger for spatial and temporal control. Near-infrared (NIR) light can penetrate tissue to a few centimeters and, when absorbed by photothermal agents on the bacterial surface, generates localized heat that serves multiple controllable functions. First, photothermal heating directly ablates tumor cells (PTT, Sect. "ROS-based therapies: integrating chemodynamic and photodynamic approaches"). Second, it can induce structural changes in nano-coatings, triggering the release of drugs or antigens. For instance, polydopamine coatings degrade under NIR-induced hyperthermia to release co-immobilized checkpoint inhibitors [79]. Third, and most elegantly, photothermal energy can be used to activate heat-inducible gene circuits within engineered bacteria. The site-specific conjugation of DBCO-Fe₃O₄ nanoparticles onto azide-labeled bacteria (Sect. "Strategies for nanoengineered bacteria") enabled alternating magnetic field–induced hyperthermia that served as a physical signal to drive the expression of a therapeutic protein (CD47nb) from a thermosensitive promoter [74]. This represents a paradigm of hybrid control where an externally applied physical stimulus is transduced by the nanomaterial into a genetic output.

Tumor microenvironment responsive

While external stimuli provide on-demand control, The acidic pH, hypoxia, and elevated enzyme levels of tumor-intrinsic cues enable autonomous, localized activation [118]. Biomineralized shells such as ZIF-8 (Sect. "Strategies for nanoengineered bacteria") remain stable in the neutral pH of blood but rapidly degrade in the acidic tumor microenvironment, releasing their cargo [100, 101]. Similarly, the FeS coating on Shewanella oneidensis MR-1 dissolves under acidic conditions to release Fe²⁺ ions and H₂S, simultaneously activating CDT and gas therapy [90]. These systems minimize systemic exposure by passively confining therapeutic action to the tumor site, thus enhancing safety without requiring external intervention.

Drawing on the principles of material-assisted biological control discussed by Liu et al. [119, 120] It becomes evident that the ideal nanoengineered bacterial platform must integrate multiple layers of control: physical steering for tumor homing, environment-responsive activation for tumor-specific action, and an inducible clearance mechanism for post-treatment safety, as discussed in Sect. "Bacteria as intrinsic immunotherapeutic agents". Such hierarchical controllability will transform bacteria from uncontrollable biological agents into truly programmable, physician-guided therapeutic robots.

Nanoengineered bacteria-mediated tumor imaging

Precise quantification of bacterial colonization within tumors is a prerequisite for optimizing dosing regimens, validating targeting efficiency, and ensuring safety. While genetically encoded reporters (e.g., luciferase, fluorescent proteins, magnetosome gene clusters, phytochrome-based proteins) have enabled longitudinal tracking in preclinical models, they are constrained by limited signal intensity, shallow tissue penetration, and potential genetic instability [34, 35, 121–124] Nanoengineering provides an alternative strategy by equipping bacteria with synthetic contrast agents that are compatible with a wide range of deep-tissue clinical imaging modalities. This section critically reviews the imaging techniques employed for tracking nanoengineered bacteria, organized by their depth of penetration, sensitivity, and clinical maturity, rather than by modality alone.

Optical and optoacoustic imaging approaches

Optical imaging techniques, including near-infrared fluorescence imaging (NIR-FI), bioluminescence imaging (BLI), and photoacoustic imaging (PAI), offer the advantages of real-time acquisition, high sensitivity, and the absence of ionizing radiation. Their shared limitation, however, is the rapid attenuation of light by biological tissues, which fundamentally restricts their standalone utility for whole-body, deep-organ tracking in patients.

Near-infrared fluorescence imaging (NIR-FI)

NIR light (700–1700 nm) provides improved tissue penetration and lower autofluorescence compared to visible light, enabling surface and subsurface visualization [125, 126]. Conjugating NIR fluorophores to tumor-homing bacteria enables real-time optical tracking of biodistribution. For instance, the heptamethine cyanine dye NHS–N782 was covalently linked to Salmonella VNP20009, creating an N–V-J theranostic system that permitted simultaneous fluorescence imaging and photothermal therapy [127]. Aggregation-induced emission (AIE) luminogens (AIEgens) exhibit intense fluorescence with tunable emission wavelengths spanning the entire visible range and extending into the near-infrared (NIR) region upon aggregation [128], owing to the restriction of intramolecular motion. This unique property enables them to serve as “light-up” probes that turn on fluorescence in response to external stimuli or microenvironmental changes, making them highly suitable for biosensing, imaging, and even integrated diagnosis and therapy [129–132]. Anchoring an AIEgen (INX-2) onto E. coli Nissle 1917 enabled high-contrast multimodal imaging of tumor tissues while simultaneously facilitating photodynamic/photothermal/immunotherapy (Fig. 12) [114]. Furthermore, certain photosynthetic bacteria, such as Rhodopseudomonas palustris, intrinsically emit NIR-II fluorescence, acting as self-reporting, living contrast agents without the need for exogenous labeling [133]. Despite these advances, NIR-FI for bacterial tracking faces two major hurdles to clinical translation. First, even NIR-II photons are exponentially attenuated, limiting reliable detection to superficial tumors or intraoperative settings. Second, no NIR dye has yet received regulatory approval specifically for whole-body bacterial imaging, and the long-term toxicity of these agents in replicating carriers requires thorough evaluation.

Fig. 12.

Schematic depiction of the EcN@INX-2 engineered bionic robot realized tumor-targeting multimodal imaging and its corresponding immunotherapy mechanisms. Reproduced with permission [114]. Copyright 2025, Springer Nature

Bioluminescence imaging (BLI)

BLI offers an inherently background-free signal because it relies on enzymatic light production (luciferase–luciferin reactions) without external excitation. A sophisticated “Trojan horse” internalization strategy, in which glucose-polymer-modified nanoparticles carrying luciferase and its substrate D-luciferin are transported into bacteria via ABC transporters, has achieved highly sensitive detection of deep-seated infections in mice through a BRET-FRET cascade that shifts emission into the NIR range (Fig. 13) [104, 134]. However, the clinical applicability of BLI for bacterial tracking is negligible. The physical principle governing photon attenuation is absolute: the millimeter-to-centimeter penetration depth achievable in rodents does not scale to human tissues. Even the most red-shifted bioluminescent systems cannot provide whole-body, quantitative bacterial maps in adult patients. BLI therefore remains a powerful preclinical tool but is not considered further as a clinically translatable imaging modality for bacterial therapy.

Fig. 13.

Bioluminescence imaging of S. aureus-induced nephritis in mice with different treatments as indicated. (a) ~ 1.0 × 10⁸ CFU S. aureus-induced nephritis in mice, and (b) ~ 1.0 × 10⁶ CFU S. aureus-induced nephritis in mice, five different probes were injected. (c) BLI of health mice, glycerin nephritis-bearing mice and S. aureus nephritis-bearing mice by using the Trojan probes. (d) BLI of health mice, glycerin nephritis-bearing mice and S. aureus nephritis-bearing mice by intraperitoneal administration of luminol (282 mM, 200 µL). Reproduced with permission [104]. Copyright 2023, Springer Nature

Photoacoustic imaging (PAI)

PAI combines optical excitation with ultrasonic detection, achieving spatial resolution at centimeter-scale depths that is unattainable by pure optical methods. A variety of synthetic contrast agents have been used, including coronene-based nanoparticles [135], gold nanorods [136–142], carbon-based nanomaterials, and NIR dyes [143], iron-based photosensitizers [144–146], biopolymer-melanin complexes [147], and inorganic semiconductor photosensitizers [148–150]. For example, conjugating ICG and polymerized dopamine to melanin-producing bacteria resulted in a Mel-pDA-ICG construct with a four-fold enhancement in PA signal (Fig. 14a-b) [151]. In addition to various nanoparticles carrying exogenous PA signal modifying bacteria realized bacteria-targeting tumor PAI. Interestingly, an attenuated Salmonella strain (ΔppGpp) was found to induce tumor-specific thrombosis, causing local hemorrhage and hemoglobin accumulation that provided strong endogenous NIR absorption, enabling PAI-based tumor visualization without any exogenous contrast agent (Fig. 14c) [152]. PAI occupies a middle ground between optical and deep-tissue imaging. It is radiation-free, relatively portable, and achieves depths of 3–5 cm, which is sufficient for many solid tumors (e.g., breast, head and neck). Its current limitation for bacterial tracking is a lack of clinical maturity: PAI instruments are not yet standardized across medical centers, and no PA contrast agent has been approved for oncological bacterial imaging. As these technologies mature, PAI may become valuable for intraoperative or superficially accessible tumor settings.

Fig. 14.

Strategy for nanoengineered bacteria-mediated photoacoustic imaging (PAI). (a) The ICG and pDA nanoparticular were polymerized onto bacteria surface; Reproduced with permission [151]. Copyright 2023, American Chemical Society. (b) The organs bacteria content detection after Mel-pDA-ICG and Mel-pDA-ICG+laser treatment; Fluorescence efficiency of tumors at 0/6/24/48 h different injection time with Mel-pDA-ICG or ICG (n = 5); Reproduced with permission [151]. Copyright 2021, American Chemical Society. (c) The schematic illustration of the tumor thrombosis formation after bacteria injection and hemoglobin accumulation-derived NIR absorption enhancing PAI tumor visualization and PAI-mediate photothermal therapy. Reproduced with permission [152]. Copyright 2025, Science

Clinical deep-tissue imaging modalities

For whole-body, quantitative, and repeatable tracking of bacterial therapeutics in human patients, only modalities already integrated into clinical oncology workflows are truly realistic. Positron emission tomography (PET) and magnetic resonance imaging (MRI) meet this criterion, each with distinct strengths and limitations.

Positron emission tomography (PET)

PET offers unlimited depth penetration, picomolar sensitivity, and absolute quantification of radiotracer accumulation. These features are indispensable for establishing the pharmacokinetic profiles and dose–response relationships required in Phase I clinical trials. Bacteria can be tracked using radiolabeled substrates that are selectively accumulated by metabolically active cells. For instance, [¹⁸F]-FDG and the bacteria-specific probe [¹⁸F]-FDS [153] have been successfully employed to track E. coli Nissle 1917 in a 4T1 breast cancer model, with signal corroborated by optical imaging (Fig. 15) [154]. PET has the unique advantage that the imaging infrastructure (PET/CT scanners) and many radiotracer production pipelines already exist in major hospitals. The primary limitations are the ionizing radiation dose, which restricts the frequency of repeated scans, and the high operational costs. Additionally, short-lived isotopes (e.g., ¹⁸F, t₁/₂ = 110 min) require on-site or nearby cyclotron facilities, complicating logistics.

Fig. 15.

¹⁸F-FDG and ¹⁸F-FDS realized tumor-targeting PET. Reproduced with permission [153]. Copyright 2020, Wiely

Magnetic resonance imaging (MRI)

MRI provides excellent soft-tissue contrast and anatomical detail without ionizing radiation, making it ideal for repeated, long-term monitoring [155]. Some bacteria naturally produce MR contrast agents, Magnetospirillum magneticum AMB-1 biosynthesizes intracellular magnetite crystals that serve as inherent T₂-weighted contrast agents [35, 156], and when cultured in iron-rich medium, they can generate particles that induce T₁-weighted positive contrast (Fig. 16a-b) [157, 158]. Alternatively, bacteria can be artificially labeled with superparamagnetic iron oxide nanoparticles (Fe₃O₄ NPs). Spirulina platensis (SP) microalgae coated with superparamagnetic Fe₃O₄ NPs obtained MR visibility (Fig. 16c), as demonstrated by T1-weighted signal changes with the different time 0 h/6 h/24 h/72 h in porcine skin gelatin phantoms and in vivo tracking within the rat gastrointestinal tract under magnetic guidance (Fig. 16d) [159]. MRI’s main limitation for bacterial tracking is its relatively low sensitivity, requiring micromolar concentrations of contrast material, which may be challenging to achieve when only a small fraction of the injected bacterial dose colonizes the tumor. Furthermore, MRI cannot easily distinguish viable from dead bacteria, as the contrast persists regardless of metabolic activity.

Fig. 16.

Engineered bacteria realized bacteria-targeted MRI. (a) Representative TEM images of AMB-1 cultured with different concentration of Fe-containing medium. Scale bars, 100 nm; Reproduced with permission [158]. Copyright 2017, Science. (b) T1-weighted MRI and corresponding tumor signal intensities after injection with AMB-1 (right tumor) and AMB-1 growth medium (left); Reproduced with permission [158]. Copyright 2017, Science. (c) The schematic illustration of the Fe₃O₄ NPs coating onto S. platensis surface. The FESEM and FL imaging of synthesized MSP-24 h, MCR-24 h, and MTS-24 h samples. Reproduced with permission [156]. Copyright 2021, Wiley. (d) Representative T2-weighted cross-sectional MRI of subcutaneously injected MSP swarms at 4 mg/mL or 8 mg/mL. Reproduced with permission [156]. Copyright 2021, Wiley

Comparative clinical potential and multimodal integration

When weighing these modalities against the stringent requirements of initial clinical trials, PET currently stands out as the most clinically realistic imaging modality for whole-body tracking of bacterial biodistribution. It uniquely provides the combination of unlimited depth, picomolar sensitivity, and absolute quantification that is essential for dose-finding studies. MRI serves as a highly complementary modality, providing radiation-free anatomical co-registration and enabling repeated imaging sessions over the course of treatment. Optical and optoacoustic techniques, while not suitable for whole-body screening, may play important ancillary roles: NIR fluorescence for real-time intraoperative guidance during tumor resection, and PAI for high-resolution, radiation-free monitoring of accessible lesions. A rational clinical imaging protocol for bacterial theranostics would therefore integrate multiple modalities: baseline and post-treatment PET/CT to quantify tumor colonization and biodistribution, serial MRI to monitor tumor morphology and surrounding tissue response without cumulative radiation, and intraoperative NIR fluorescence to guide surgical margins. As next-generation NIR-II probes and standardized PAI systems advance through regulatory pathways, these optical modalities may gradually transition from preclinical tools to clinical adjuncts. The choice of imaging strategy must be guided by the specific clinical question—whether it is pharmacokinetic profiling, therapeutic response assessment, or surgical planning—and by the available hospital infrastructure.

Nanoengineered bacteria-mediated combination biotherapy

Bacteria as intrinsic immunotherapeutic agents

Long before the advent of nanotechnology, live bacteria were recognized for their inherent capacity to stimulate potent anti-tumor immunity. The pioneering work of William Coley in the 1890s demonstrated that intratumoral injection of heat-inactivated Streptococcus pyogenes and Serratia marcescens could induce durable tumor regression in sarcoma patients, establishing the first systematic bacterial immunotherapy [160]. The molecular basis of this phenomenon is now understood to rely on the recognition of highly conserved microbial structures, collectively termed pathogen-associated molecular patterns (PAMPs), by the host innate immune system [161–163]. Lipopolysaccharide (LPS) on the surface of Gram-negative bacteria engages Toll-like receptor 4 (TLR4), flagellin activates TLR5, and unmethylated CpG motifs in bacterial DNA stimulate TLR9. These interactions trigger the NF-κB and type I interferon signaling cascades, resulting in a local burst of pro-inflammatory cytokines such as tumor necrosis factor (TNF-α), interleukin-1β (IL-1β), and interleukin-12 (IL-12). This cytokine milieu potently recruits and activates neutrophils, natural killer (NK) cells, and macrophages, while simultaneously promoting dendritic cell (DC) maturation and antigen cross-presentation. Collectively, these events transform an immunologically “cold” tumor microenvironment into a “hot” one, breaking pre-existing immune tolerance and priming a systemic, T-cell-mediated adaptive response against tumor-associated antigens.

Modern bacterial therapy builds on this foundational principle by employing well-characterized, attenuated strains engineered for safety while retaining immunostimulatory capacity [164]. Salmonella typhimurium VNP20009, auxotrophic mutants of Listeria monocytogenes, and the probiotic Escherichia coli Nissle 1917 have all been shown to preferentially colonize solid tumors and elicit robust innate and adaptive immune responses in preclinical models [5, 6, 12–14]. For instance, VNP20009 colonization in murine melanoma and breast cancer models leads to a significant influx of CD8⁺ T cells and a concomitant reduction in immunosuppressive regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs) within the TME. Listeria-based platforms further exploit the bacterium’s unique intracellular lifecycle to deliver tumor-associated antigens directly into the cytoplasm of antigen-presenting cells, generating potent tumor-specific cytotoxic T-lymphocyte responses that have advanced to phase II/III clinical trials for cervical and pancreatic cancers. E. coli Nissle 1917, already approved for clinical use in gastrointestinal disorders (Mutaflor^®), has demonstrated intrinsic oncolytic and immunomodulatory properties that make it an attractive chassis for tumor immunotherapy [11, 12].

Despite these immunological advantages, bacteria as standalone immunotherapeutic agents face two major constraints: dose-dependent toxicity and the profound immunosuppressive networks operative within the tumor. Increasing the bacterial dose amplifies the anti-tumor immune response but concurrently elevates the risk of systemic inflammation, cytokine release syndrome, and sepsis. Furthermore, the TME actively counteracts bacterial-induced immune activation through multiple mechanisms, including upregulation of programmed death-ligand 1 (PD-L1), expression of indoleamine 2,3-dioxygenase (IDO), and recruitment of regulatory cell populations. Consequently, while live bacteria constitute a powerful self-adjuvanted therapeutic platform, their intrinsic immunotherapy alone is rarely curative against large, established tumors.

This fundamental limitation has motivated the integration of nanotechnology with living bacterial vectors. The overarching goal of nanoengineering is not to replace but to augment the bacteria’s innate immunostimulatory functions: by equipping them with additional therapeutic modalities that directly kill tumor cells, thereby releasing a fresh wave of tumor antigens, and by co-delivering agents that specifically counteract the immunosuppressive barriers that restrain the bacteria-triggered immune response [165]. the following subsections detail how nanoengineering enables this synergistic enhancement through ROS-based killing, thermal ablation, and rationally combined immunomodulation, each building on the intrinsic immunotherapeutic backbone described here.

ROS-based therapies: integrating chemodynamic and photodynamic approaches

Both chemodynamic therapy (CDT) and photodynamic therapy (PDT) ultimately exert their anticancer effects through the generation of cytotoxic reactive oxygen species (ROS), primarily hydroxyl radicals (•OH) and singlet oxygen (¹O₂), respectively. CDT is triggered by Fenton/Fenton-like chemistry, while PDT is triggered by photon absorption. Despite this difference, they share common challenges, including dependence on the availability of tumor substrates (H₂O₂, O₂) and the need for catalytic centers or photosensitizers (PSs) to be precisely localized within the tumor [166]. Nanoengineered bacteria are uniquely suited to address these challenges by combining the bacteria’s metabolic machinery with synthetic nanocatalysts or photosensitizers.

Chemodynamic therapy (CDT)

CDT exploits Fenton or Fenton-like reactions to convert endogenous H₂O₂ into •OH within tumor sites for inducing apoptotic cell death [167–170]. As an in situ therapeutic modality, CDT offers several distinct advantages, including inherent tumor specificity, without requiring external stimulation or energy input, and minimal side effects. However, the efficacy of CDT relies heavily on the concentration of endogenous H₂O₂ in TME, and numerous studies have shown that the steady-state concentration of H₂O₂ in tumors is typically in range of 50–100 µM, which is far from sufficient to sustain the continuous generation of therapeutic levels of ·OH via Fenton or Fenton-like reactions driven by iron- or copper-based catalysts. In addition, the amount of superoxide anions (.O₂⁻) produced by bacterial metabolism is extremely low, and bacteria themselves do not carry highly active Fenton catalytic metal centers. Therefore, bacterial metabolism alone cannot maintain therapeutically relevant ROS levels in vivo. Consequently, wild-type bacteria are engineered to bypass the insufficient supply of endogenous H₂O₂, enabling the sustained production of millimolar levels of H₂O₂, while nanocatalysts are constructed on the bacterial surface to continuously generate therapeutic ·OH. For example, magnetic Fe₃O₄ nanoparticles (MNPs) were chemically anchored to engineered E. coli (Ec-pE) overexpressing the respiratory enzyme respiratory chain enzyme II (NDH-2), creating the hybrid system Ec-pE@MNP [66]. During bacterial respiration, NDH-2 transfers electrons from nicotinamide adenine dinucleotide (NADH) to oxygen, enabling continuous endogenous production of H₂O₂ [171]. Following tumor accumulation, the MNPs catalyzes the bacteria generated H₂O₂ into highly cytotoxic •OH radicals without requiring external H₂O₂ supply. Similarly, biomineralized FeS nanoparticles on Shewanella oneidensis MR-1 (FeS@MR-1) surface release Fe(II) ions and H₂S gas under the acidic tumor microenvironment [91]. The H₂S inhibits catalase activity, leading to intracellular H₂O₂ accumulation, while Fe(II) drives the Fenton reaction, and the bacteria continuously regenerate Fe(II) from Fe(III), creating a self-amplifying CDT loop.

Photodynamic therapy (PDT)

PDT utilizes PSs to generate ROS under light irradiation, effectively eliminating cancer cells. However, conventional PDT faces several limitations, including the poor penetration of PSs, tumor hypoxia and limitation of light penetration depth. PDT requires photons to excite the PSs to a high-energy state, and then energy transfer is used to generate ¹O₂. This requires a high photon flux to ensure an adequate yield of ¹O₂. Therefore, in cases where the light penetration is severely attenuated, the reaction rate of PDT tends to drop rapidly to the ineffective threshold. In addition, Most PDT relies on the Type II mechanism, in which the excited-state PSs transfers energy to ground-state oxygen molecules (³O₂), generating highly reactive ¹O₂. This process irreversibly consumes O₂ molecules in TME. Therefore, PDT is somewhat effective against superficial, O₂-rich tumor regions, but is often ineffective against deep, hypoxic areas that determine recurrence and invasion. To address hypoxia, Nanoengineered bacteria can serve as in situ oxygenators. Photosynthetic cyanobacteria, such as Synechococcus elongatus PCC 7942, produce O₂ through photosynthesis under light irradiation. By conjugating the photosensitizer Ce6 to these cyanobacteria (Cyan@Ce6), the bacterially generated O₂ is immediately converted into cytotoxic ¹O₂ by the PS, significantly enhancing PDT efficacy in hypoxic tumors [113].

Researchers have integrated PSs with tumor-targeting engineered bacteria, enabling spatiotemporally controlled ROS generation directly within the TEM [61, 113, 172, 173]. For example, photosensitizer-encapsulated nanoparticles (HSA/ICG) were constructed via intermolecular disulfide crosslinking and covalently conjugated to the bacterial surface through amide bonds, forming a biomimetic hybrid S/HSA/ICG [174]. Beyond organic photosensitizers, inorganic nano-photosensitizers and AIE PSs (Fig. 17a) have been increasingly employed to functionalize bacteria for enhanced photodynamic therapy [46].

Fig. 17.

PSs integrates tumor-targeted engineered bacteria to realize PDT. (a) AIE PSs nanoparticles-E.coli biohybrids system used for bacteria broken and photodynamic killing; Reproduced with permission [46]. Copyright 2019, American Chemical Society. (b) PEG-BPNSs conjugated cyanobacteria by amination reaction, the constructed Cyan@BPCNs realized tumor-targeted PDT; Reproduced with permission [179]. Copyright 2021, Wiley. (c) DSPE-PEG-NH₂ modified bacteria further conjugating with BPQDs via amination reaction, the BPQDs@E.coli-NH₂ nanoengineered bacteria realized tumor targeting and efficient photodynamic therapy. Reproduced with permission [52]. Copyright 2021, American Chemical Society

Black phosphorus nanosheets (BPNSs) and black phosphorus quantum dots (BPQDs) inorganic two-dimensional nanomaterials have been demonstrated that broad-spectrum light absorption and exceptional photothermal conversion efficiency, along with a remarkable ability to generate ¹O₂ [175–178]. Shi’s group utilized BPNSs to modify photosynthetic cyanobacteria via amidation, constructing a hybrid system Cyan@BPNSs (Fig. 17b). The cyanobacteria can continuously produce oxygen through photosynthesis process under 660 nm laser irradiation, while the BPNSs convert this oxygen into cytotoxic ¹O₂, inducing significant oxidative stress and potent tumor cell death [179]. Similarly, Li’s group electrostatically adsorbed BPQDs onto catalase-expressing engineered E. coli, creating an E. coli/BPQDs (EB) composite. Upon light exposure, the BPQDs not only generate ROS but also disrupt bacterial membranes, leading to the release of catalase. The exposed enzyme further decomposes endogenous H₂O₂ to produce O₂, which in turn amplifies the efficacy of BPQD-mediated PDT (Fig. 17c). This self-reinforcing cycle resulted in effective tumor cell elimination both in vitro and in vivo [52]. To further mitigate hypoxia and reduce dependence on prolonged external irradiation, Shi’s group developed a blue persistent luminescence material (CaAl₂O₄: Eu, Nd) as an inplanted light source. Functioning as an “optical battery,” this material activates both cyanobacteria and the photosensitizer verteporfin without requiring continuous external light. This innovative system enables sustained O₂ production and ¹O₂ generation, effectively alleviating tumor hypoxia and improving the effect of PDT [137].

The convergence of CDT and PDT under a unified ROS-based framework reveals a recurring design principle: the bacteria’s metabolic activity supplies the limiting substrate (H₂O₂ or O₂), while the nanomaterial provides the catalytic or photosensitizing function to convert that substrate into lethal ROS. This synergy is a hallmark of effective nanoengineered bacterial therapeutics and can be further extended by combining ROS generation with thermal or immune-stimulating effects.

Thermal ablation therapies

Photothermal therapy (PTT) employs photothermal agents (PTAs) to convert light energy into heat, causing localized tumor ablation. However, its clinical translation has been hampered by the poor tumor-targeting ability of conventional PTAs and the intrinsic heat resistance and heterogeneous heat distribution of tumor tissues [180, 181]. It is well known that PTAs nanoparticles accumulate in tumor sites through the enhanced permeability and retention (EPR) effect of solid tumors. However, intratumoral heterogeneity causes PTAs to distribute mainly in the perivascular regions with abundant blood supply, making it difficult to penetrate in the hypoxic tumor core. This leads to heterogeneous heat distribution. Moreover, differences in thermal conductivity among various tissue components impede uniform heat conduction. In addition, heterogeneous heating can cause localized hyperthermia, which induces the expression of heat shock proteins and triggers thermal resistance at the cellular level. Therefore, the development of bacteria-based photothermal platforms offers a promising strategy to overcome these limitations by leveraging the natural bacteria-targeting ability toward deep/hypoxic of tumor, this active infiltration and colonization result in more uniform heating of the tumor and enhancing thermal efficacy within the tumor microenvironment [182]. Bacteria, with their ability to actively penetrate and colonize hypoxic tumor cores, offer a solution by delivering PTAs uniformly throughout the tumor mass, thereby achieving more homogeneous heating and reducing thermal resistance [89, 90, 103]. A common strategy involves decorating tumor-targeting bacteria with synthetic PTAs such as noble metal nanoparticles, indocyanine green (ICG)-loaded nanoparticles [63, 183–188], or polydopamine (PDA) [80] (Fig. 18a), or melanin [151] perylene diimides (PDIs) organic dyes [189–191]. For instance, Shewanella oneidensis MR-1 can biologically reduce Pd²⁺ to form palladium nanoparticles (Pd@MR-1), combining bacterial tumor tropism with strong NIR photothermal conversion for deep tumor ablation. PDA-coated Salmonella VNP20009 (pDA-VNP) provides intrinsic photothermal properties while the coating itself serves as a reactive platform for secondary functionalization. ICG-loaded PLGA-PEG nanoparticles conjugated to bacteria (INPs-YB1) via amide bonds enable NIR fluorescence-guided PTT.

Fig. 18.

PTAs integrate tumor-targeted engineered bacteria to realize PTT. (a) The pDA-VNP nanoengineered bacteria system realized bacteria-targeted PTT and PTT-mediated produced cell lysis further enhance the biotherapy. Reproduced with permission [80]. Copyright 2018, American Chemical Society. (b) Schematic illustrates that the photosynthetic bacteria used as PTAs for hypoxia-targeted tumor ablation. Reproduced with permission [196]. Copyright 2021, American Chemical Society

Beyond synthetic modifications, certain bacteria exhibit inherent photothermal capabilities, photosynthetic bacteria like Rhodobacter johrii possess natural chlorophylls that absorb NIR light, functioning as biocompatible, self-contained PTAs (Fig. 18b) [192–196]. Moreover, the discovery that attenuated Salmonella ΔppGpp can induce tumor-specific thrombosis and hemorrhage, thereby increasing local NIR absorption without any exogenous agent, opens a path toward “agent-free” photothermal ablation guided by photoacoustic imaging [152]. Thermal ablation, whether achieved by synthetic or biological PTAs, not only destroys tumor cells directly but also generates a reservoir of tumor antigens and danger signals. This natural progression from local hyperthermia to systemic immune activation forms the conceptual bridge to the next therapeutic category.

Immune-modulating therapies: converging on immunogenic cell death

A central, cross-cutting theme in bacteria-based combination therapy is that both ROS generation and thermal ablation can induce immunogenic cell death (ICD) [60, 129, 197, 198]. ICD is characterized by the release of damage-associated molecular patterns (DAMPs, e.g., CRT, HMGB1) and tumor-associated antigens (TAAs). This “danger signal” potently activates dendritic cells, which then prime a systemic, adaptive T-cell response against the cancer, turning the treated tumor into an in situ vaccine [199–201]. This synergistic effect enables combined photothermal agents (PTAs) and immunotherapeutic components strategies to effectively anti-tumor therapy for primary tumors and distant metastases [202–205]. Nanoengineered bacteria are exceptionally well-positioned to exploit this ICD–immune axis because they inherently deliver immunostimulatory bacterial components (PAMPs such as LPS and flagellin) alongside the ICD-inducing therapy, creating a potent adjuvant-rich microenvironment. For example, Salmonella VNP20009 engineered to biosynthesize copper sulfide nanoparticles (CuS@VNP20009) generates photothermal ablation under NIR irradiation, triggering CRT exposure and HMGB-1 release that enhance cytotoxic T lymphocyte infiltration [92]. Similarly, the AIEgen-based bionic robot EcN@INX-2 leverages synergistic PDT/PTT to induce robust ICD, activating CD4⁺ and CD8⁺ T cells and culminating in a powerful antitumor immune attack [114, 206, 207]. More sophisticated platforms combine multiple immune-activating components on a single bacterial chassis. E. coli engineered to simultaneously produce melanin (for PTT) and express an anti-PD-1 checkpoint inhibitor generates both potent photothermal ablation and sustained checkpoint blockade, creating a self-sustaining cycle of immune activation [208–211].

Importantly, the immune consequences of bacterial therapy are not limited to photothermal or photodynamic triggers. ROS generated via CDT, as well as metabolic interventions (discussed below), can also contribute to ICD. The key insight from a translational perspective is that the therapeutic endpoint, a durable, systemic antitumor immune response, can be reached through multiple mechanistic entry points, all of which are potentiated by the bacteria’s intrinsic immunoadjuvanticity. Future designs should therefore prioritize the rational integration of ICD induction with checkpoint modulation, rather than optimizing a single physical modality in isolation.

Metabolic intervention and other emerging strategies

Beyond ROS and hyperthermia, nanoengineering can potentiate bacteria for metabolic and gas-based therapies. For example, photocatalytic carbon nitride nanoparticles assembled on E. coli (CCN@E. coli) transfer photoelectrons to bacterial nitrate-reducing enzymes under light, boosting the conversion of non-toxic NO₃⁻ to toxic nitric oxide (NO) for controlled metabolite therapy [212, 213]. Another promising area is metabolic intervention. Shewanella coated with MnO₂ (Bac@MnO₂) can oxidize and deplete lactate, a key oncometabolite that promotes immune evasion and drug resistance, leading to tumor regression without conventional drugs [214–220]. The MnO₂ coating acts as an electron acceptor, facilitating lactate oxidation and catabolism while simultaneously decomposing endogenous H₂O₂ to generate oxygen [221]. This system effectively reduces both extracellular and intracellular lactate levels, leading to tumor regression without the need for conventional drugs. Similarly, sulfate-reducing bacteria (SRB) biomineralizing MnS (MnS@SRB) simultaneously depletes lactate and activates the cGAS-STING pathway for potent metabolic-immunotherapy (Fig. 19) [93].

Fig. 19.

Schematic illustration of tumor targeting of MnS@SRB and its tumor inhibition mechanism. Reproduced with permission [93]. Copyright 2025, Wiley

Integrated theranostic platforms based on nanoengineered bacteria

While Sect. "Optical and optoacoustic imaging approaches"–"Clinical deep-tissue imaging modalities" and Sect. "Bacteria as intrinsic immunotherapeutic agents"–"Metabolic intervention and other emerging strategies" separately detail the imaging and therapeutic functions, a transformative advantage of nanoengineered bacteria is their ability to integrate both capabilities within a single living platform. Such “all-in-one” theranostic systems enable real-time, image-guided intervention, providing critical feedback on bacterial localization, dosing, and treatment efficacy. Several representative examples illustrate this integrated design. Salmonella VNP20009 covalently conjugated with the near-infrared dye NHS–N782 (the N–V–J system) permits simultaneous fluorescence imaging and photothermal therapy (PTT), allowing visualization of bacterial biodistribution immediately prior to photothermal ablation (Sect. "A critical strategy comparison" [62]),. The AIEgen-based bionic robot EcN@INX-2, constructed by anchoring aggregation-induced emission luminogens onto E. coli Nissle 1917, achieves multimodal imaging (near-infrared fluorescence, photoacoustic, and photothermal) while generating both reactive oxygen species and heat for combined photodynamic/photothermal/immunotherapy (Sect. "Strategies for nanoengineered bacteria" [114]),. Similarly, melanin-producing bacteria coated with polydopamine and indocyanine green (Mel-pDA-ICG) exhibit a four-fold enhancement in photoacoustic signal, enabling high-contrast PA imaging to guide subsequent photothermal ablation (Sect. "Optical and optoacoustic imaging approaches" [151]),. ICG-loaded PLGA nanoparticles conjugated to Salmonella YB1 (INPs-YB1) provide NIR fluorescence guidance for tumor-targeted PTT (Sect. "A critical strategy comparison" [63]),. Beyond optical and optoacoustic modalities, some systems integrate clinical deep-tissue imaging with therapy. Magnetospirillum magneticum AMB-1 naturally synthesizes magnetite crystals that serve as both T₂/T₁ MRI contrast agents and mediators for magnetic hyperthermia, while also enabling magnetically guided drug delivery (e.g., MTB-LP carrying liposomal drugs, Sect. "Nanoengineered bacteria construction, fate in vivo, and controllability" [69, 158]),. Engineered E. coli coated with superparamagnetic Fe₃O₄ nanoparticles (Ec-pE@MNP) not only catalyze sustained chemodynamic therapy via the Fenton reaction but, owing to the magnetic core, also hold strong potential for MRI-based tracking (Sect. "Nanoengineered bacteria construction, fate in vivo, and controllability [66]),.These examples highlight a common design principle: the nanocoating simultaneously serves as the diagnostic contrast agent and the therapeutic effector.

Clinical translation: safety, regulatory, and clinical translation

The ultimate clinical value of nanoengineered bacteria hinges on risk profile. While the preceding sections have demonstrated how nanotechnology amplifies therapeutic efficacy and imaging capability, this section confronts the safety dimensions that from systemic infection and hyperinflammation to examine the regulatory landscape that any candidate must navigate, which have largely determined the translational fate of bacterial therapies.

The safety landscape

Systemic infection and sepsis risk

The fundamental hazard of administering live bacteria is the potential for systemic infection, bacteremia, and sepsis, particularly in immunocompromised cancer patients. Even highly attenuated strains carry residual virulence, and the hypoxic tumor core, while promoting colonization, can also serve as a reservoir for bacterial leakage into the bloodstream. The Phase I clinical trial of Salmonella typhimurium VNP20009 in patients with metastatic melanoma revealed a narrow therapeutic window: while tumor colonization was confirmed, dose-limiting toxicities including bacteremia, hyperbilirubinemia, and hypotension were observed at doses above 3 × 10⁸ CFU/m² [222]. These adverse events underscore the challenge of balancing bacterial load for efficacy against systemic tolerance. The key lesson is that attenuating mutations alone is insufficient to guarantee safety; additional external control mechanisms, such as those enabled by nanotechnology (Sect. "Nanoengineered bacteria construction, fate in vivo, and controllability") are essential. Nanocoatings that confine bacterial immune recognition until the tumor is reached, and magnetic or light-triggered activation localized to the tumor, represent practical strategies to reduce systemic exposure. Future studies should integrate physiological-based pharmacokinetic (PBPK) models with non-invasive imaging surrogates to enable adaptive dosing.

Cytokine storm and hyperinflammation

The very mechanism that makes bacteria potent immunostimulants, PAMPs-driven innate activation also poses the risk of a cytokine release syndrome (CRS) or sepsis-like hyperinflammation. Lipopolysaccharide, flagellin, and bacterial DNA are among the most powerful triggers of TNF-α, IL-6, and IL-1β secretion [223]. In animal models, high-dose intravenous administration of Gram-negative bacteria can induce a rapid, fatal cytokine storm. Nanoengineering offers a potential solution by making PAMPs during systemic circulation (as discussed in Sect. "The fate of nanoengineered bacteria in vivo") and allowing them to be re-exposed selectively in the tumor after coating degradation. However, this strategy must be calibrated precisely: full and permanent shielding of PAMPs may abolish the adjuvanticity required for robust anti-tumor immunity (Sect. "Bacteria as intrinsic immunotherapeutic agents"), while excessive exposure causes systemic toxicity. The ideal nano-coating is therefore a dynamic one, and this remains a significant design challenge.

Uncontrolled proliferation and ecological containment

A self-replicating therapeutic agent introduces unique challenges not encountered with conventional drugs or nanoparticles. Once administered, bacteria can proliferate exponentially if not constrained, leading to unpredictable pharmacokinetics and potential ecological escape. Genetic kill switches, such as auxotrophy or inducible suicide genes, have been developed but are susceptible to mutational inactivation. Nanotechnology can add an orthogonal layer of control: photothermally or magnetically triggered hyperthermia can induce bacterial death independently of genetic circuits (Sect. "Nanoengineered bacteria-mediated combination biotherapy"). Moreover, biomineralized shells that degrade in the TME can be engineered to release bactericidal agents’ post-treatment. Designing a “self-eliminating” nanoengineered bacterium, one that performs its function and then autonomously degrades, remains a central goal. The risk of horizontal gene transfer to the microbiome must also be considered, especially with genetically modified strains. To facilitate regulatory acceptance, we propose a minimal panel of characterization criteria (surface charge, conjugation density, viability/motility metrics) following quality-by-design principles.

Immunogenicity and toxicity of nanomaterials

The nanocomponents themselves are not immunologically inert. Polyethylene glycol (PEG), widely used for stealth coatings, can induce anti-PEG IgM and IgG antibodies after administration, leading to accelerated blood clearance and, in severe cases, complement activation-related pseudoallergy (CARPA) [224]. Cationic polymers used for electrostatic adsorption (Sect. "Strategies for nanoengineered bacteria") are known to disrupt cell membranes and activate complement [225]. The long-term fate of iron oxide, gold, or silica inorganic nanoparticles remains poorly characterized in the context of replicating bacterial carriers. They may accumulate in the liver, spleen, and kidneys, with unknown chronic toxicity [226, 227]. Biodegradable coatings (e.g., polydopamine, MOFs) are preferable, but their degradation products must be profiled for safety. A key recommendation is that future studies systematically evaluate the immunotoxicology of the nano-biohybrid, not just its acute therapeutic efficacy.

Biodistribution, clearance, and long-term persistence

Complete understanding of the in vivo behavior of nanoengineered bacteria is lacking. After tumor colonization and therapeutic action, what happens to the bacterial corpse and its associated nanoparticles? Preclinical imaging data (Sect. "Nanoengineered bacteria-mediated tumor imaging") often shows signal persisting in liver and spleen for weeks, suggesting that a significant fraction of the dose is cleared by the reticuloendothelial system. Whether this leads to chronic inflammation, granuloma formation, or organ dysfunction requires long-term toxicology studies. The biodegradability of the nano-coating is critical: rapidly degradable shells may be eliminated, whereas inert inorganic materials could bioaccumulate. Standard pharmacokinetic models designed for static drugs are inadequate for self-replicating vectors; new modeling frameworks that account for bacterial proliferation and immune clearance is urgently needed.

Regulatory considerations

From a regulatory perspective, nanoengineered bacteria sit at the intersection of multiple categories: live biotherapeutic products (LBPs), nanomedicines, and, if genetically modified, gene therapy products. In the United States, the FDA’s Center for Biologics Evaluation and Research (CBER) provides guidance on LBPs, emphasizing the need for well-characterized master cell banks, environmental containment measures, and rigorous monitoring for genetic stability [228]. A nanoengineered bacterium would likely be classified as a combination product, requiring coordinated review by multiple centers and demanding extensive characterization of both the biological and nanomaterial components. Early and iterative engagement with FDA CBER is strongly advised, and a consensus white paper defining first-in-human evidence expectations would benefit the field.

Clinical experience with bacterial cancer therapy, though limited, provides instructive precedents. The VNP20009 Phase I trial, while not successful due to toxicity and insufficient tumor colonization, established a framework for dose escalation, antibiotic susceptibility testing, and patient monitoring that remains relevant [222]. The more successful clinical development of Listeria monocytogenes-based vaccines (e.g., ADXS11-001 for cervical cancer) has demonstrated that repeated intravenous dosing of an attenuated intracellular bacterium can be safe when coupled with antibiotic “safety switches” and rigorous cytokine monitoring [229]. These trials highlight that translation requires not just efficacy data but a comprehensive pharmacovigilance plan that includes: (i) predefined stopping criteria based on inflammatory markers, (ii) pre-established antibiotic susceptibility and de-escalation protocols, and (iii) long-term follow-up for autoimmune sequelae and second malignancies.

In summary, the safety challenges of nanoengineered bacteria are multifaceted and interdependent. Solving one (e.g., systemic infection) without worsening another (e.g., immunogenicity of the coating) requires an integrated design philosophy. The path to clinical translation will demand not only more sophisticated nano-biohybrid engineering but also new regulatory frameworks that recognize the unique, living nature of these therapeutics.

Conclusions and future perspectives

Bacteria-mediated tumor therapy has been advanced by the integration of synthetic biology and nanotechnology. This review has systematically deconstructed the six fundamental strategies for constructing nanoengineered bacteria, such as physical adsorption, covalent conjugation, bioorthogonal chemistry, self-polymerization, biomineralization, and internalization, and critically mapped each strategy’s interfacial properties to its ultimate theranostic function. Overall, we provide a “chemistry-to-function” framework that directly addresses the three translational bottlenecks of conventional bacterial therapy: (i) insufficient monotherapy efficacy, overcome by functionalizing bacteria with catalytic, photothermal, photodynamic, and immunomodulatory nanomaterials; (ii) the inability to track bacterial biodistribution in real time, resolved by arming bacteria with contrast agents for multimodal imaging; and (iii) safety concerns, mitigated through externally controllable nano-activation and the potential for reduced bacterial dosing. Despite this progress, incremental optimization of individual components is no longer sufficient. The next generation of nanoengineered bacteria must be conceived as integrated, adaptive, and translation-ready living therapeutic systems. We identify three key priorities that will define this transition.

From static coatings to adaptive, dynamic bio-interfaces

The current generation of nano-bio interfaces is largely static. Although capable of providing stealth or catalytic function, these coatings lack the environmental responsiveness required to navigate the complex spatiotemporal dynamics of the tumor microenvironment (TME). As highlighted in our cross-strategy comparison (Sect. "A critical strategy comparison"), every conjugation approach involves a trade-off between stability and bacterial viability; a translational solution must resolve this challenge through dynamic adaptability. Future coatings should be engineered with TME-triggered programmable logic: (a) pH- or enzyme-labile cross-linkers that selectively degrade within the tumor to unmask bacterial immunoadjuvants (e.g., LPS, flagellin) and initiate localized immunotherapy; (b) hypoxia-responsive linkers that release chemotherapeutic payloads exclusively in the poorly vascularized tumor core where bacteria preferentially colonize; and (c) stimuli-responsive polymers that expose therapeutic ligands upon binding to tumor-specific metabolites. The integration of dynamic covalent chemistry (e.g., boronic ester, imine, and disulfide bonds) with bacterial surface engineering will be central to this effort. Moreover, the coating itself could serve as a functional sensor: for instance, fluorescently quenched probes that activate only upon detachment would provide built-in feedback on coating integrity and payload release kinetics.

Hierarchical, clinically compatible control and self-elimination

The clinical translation of living bacterial therapeutics depends on the ability to control their lifecycle, from systemic administration and tumor colonization to therapeutic action and definitive elimination. There is no single control mechanism that can achieve this, hierarchical, multi-layered control architecture is required. External physical stimuli (magnetic fields, focused ultrasound, NIR light) should govern macroscopic biodistribution and tumor localization, as demonstrated with magnetically guided bacteria and photothermal activation. Tumor-intrinsic factors (pH, hypoxia, elevated ATP) should then trigger the therapeutic activation, ensuring that payload release and immunostimulatory unmasking occur only at the target site. Finally, a third layer of orthogonal control must enable post-treatment elimination. This could be realized through nanomaterial-mediated induction: for example, a second, longer-wavelength NIR pulse that specifically triggers photothermal ablation of the bacterial carrier without damaging surrounding tissue, or magnetic hyperthermia-activated suicide circuits. The development of “self-clearing” nano-coatings, such as enzymatically degradable biomineralized shells that disassemble into renally excretable fragments and concurrently release a bactericidal payload would combine this elimination step with the clearance of the nanomaterial itself, addressing the long-term biodistribution and bioaccumulation concerns discussed in Sect. "The safety landscape".

Manufacturability, standardization, and regulatory convergence

Finally, clinical development of nanoengineered bacteria will require a transition from customized laboratory fabrication to robust, scalable, and quality-controlled manufacturing. Drawing on the critical strategy comparison (Table 2) and the quality-by-design framework introduced in Sect. "Regulatory considerations", we propose that the field converge on a set of minimal characterization criteria: (a) nanoparticle conjugation density and homogeneity; (b) post-engineering bacterial viability, motility, and growth kinetics; (c) coating stability in whole human serum; and (d) the residual PAMPs exposure. These should be established as consensus critical quality attributes (CQAs). Self-polymerization and biomineralization strategies, which proceed under cytocompatibility conditions with minimal batch-to-batch variation, may offer inherent manufacturability advantages over multi-step covalent or bioorthogonal protocols. In parallel, the field must proactively engage with regulatory agencies to co-develop evaluation frameworks that account for the unique, self-replicating nature of these combination products. Longitudinal safety studies specifically examine chronic nanomaterial toxicity, immunogenicity of the hybrid surface, and ecological containment of genetically modified strains are not optional but requisite.

In conclusion, nanoengineered bacteria represent a promising direction in oncology, moving from passive drug delivery systems to tumor-targeting, programmable living therapeutics. Achieving their clinical potential will require a coordinated interdisciplinary effort that bridges materials science, synthetic biology, and clinical oncology. A central principle is to engineer not just the bacterium or the nanoparticle, but the interface between them and, more importantly, the interface between the hybrid system and the host. It is at these interfaces that further advances are most likely to occur, supporting the translation of nanoengineered bacteria from a preclinical concept into a new class of clinically validated, image-guided, and tightly controlled anticancer agents.

Author contributions

Youyi Yu: Funding acquisition, Investigation, Writing–original draft, Methodology, Writing–review&editing. Weitao Yang: Resources, Visualization. Bingbo Zhang: Conceptualization, Supervision, Project administration.

Funding

This work was supported by the National Natural Science Foundation of China (82302363, 82272055, 82272137), the National Key Research and Development Project (2022YFB3804500).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.