Advances in Spatial Heterogeneity and Cellular Localization Specificity of Intratumoral Bacteria Authors: Minlan Ren, Jing Jia, Tingting Li, Yiwei Li, Nanxiang Shen, Yun Shi, Zhe Wang, Hao Fan, Guangli Sun, Guanying Xiong, Gang Li, Rui Peng Published: 2026 ABSTRACT ABSTRACTBacteria in tumors are increasingly recognized to be a part of the tumor microenvironment and their biological function depends on the location and relative abundance of a given microorganism or microbial community. Genetic sequencing data directly from tissue samples to detection of active metabolic signals from microorganisms, cell‐microbiome interactions, ultrastructural localization of microorganisms; these are all different kinds of inferences about the functional role of the microbiome, and the evidence for “microbial presence and function” grows in a stepwise fashion. Furthermore, most current microbial detection methods do not, alone, predict the viability, biological activity, or functional properties of microorganisms. In this review, we compare different existing microbial detection techniques and distinguish three spatial paradigms: organized communities in tissue, intracellular bacteria, and extracellular bacteria. We summarize how each paradigm influences signal transduction, metabolism, metastatic spread, treatment resistance, antitumor immunity, and immunotherapy responses. We focus on low‐biomass contamination, batch effects, negative controls, absolute quantification, and orthogonal validation. Finally, we classify results according to localization, biological role, clinical evidence, application, and applications, while distinguishing observations of human patients from mechanistic and therapeutic preclinical studies. Spatial microbiology provides useful information about tumor‐microbial interactions, but clinically actionable conclusions require standard analysis, reproducible localization, viability or activity, and functional disruption experiments to demonstrate microbial effects and independent clinical validation.Keywords: extracellular bacteria, immunotherapy, intracellular bacteria, intratumoral bacteria, spatial microbiology Abbreviations 16S ribosomal RNA Annexin A2 cyclic GMP‐AMP synthase—stimulator of interferon genes—interleukin‐17B colorectal cancer C‐X‐C motif chemokine ligand 16 digital PCR extracellular matrix Erb‐B2 receptor tyrosine kinase 2 fluorescence in situ hybridization histone deacetylase 3/8 human leukocyte antigen human papillomavirus immunohistochemistry interleukin‐17 interleukin‐8 mitogen‐activated protein kinase Multiplexed Error‐Robust Fluorescence In Situ Hybridization mismatch repai microsatellite instability microsatellite stable nuclear factor kappa‐light‐chain‐enhancer of activated B cells natural killer cell Penicillin‐Binding Protein 1B programmed death‐1 programmed death‐ligand 1 quantitative PCR v‐rel reticuloendotheliosis viral oncogene homolog B Ras homolog family member A Spatial Enhanced Resolution Omics‐sequencing T‐box transcription factor 21 transforming growth factor beta tumor‐infiltrating lymphocytes Toll‐like receptor 4 tumor mutational burden tumor microenvironment Transporter Membrane Protein C Introduction The tumor‐associated microbiota has become increasingly popular in recent years. As a microbial community colonizing tumor tissue, tumor‐associated microbiota, tumor cells, immune cells, stromal cells, non‐cellular components such as extracellular matrix and soluble factors constitute the tumor microenvironment (TME) and are involved in tumor initiation, immune remodeling, metastasis, and treatment response [1, 2]. Large scale pan‐cancer studies have shown that bacterial signals can indeed be detected in most human tumor tissues, but the composition and distribution of the bacteria vary depending on tumor type [3, 4]. This suggests that, while considering the presence of microorganisms in tumors, we should also consider the distribution of these communities and their possible actions. Different spatial and cell locations of microorganisms may lead to different patterns of interactions [5]. For instance, bacteria present in malignant cells may interfere with tumor initiation and progression by directly controlling signal transduction, the cytoskeleton, cell metabolism, or antigen presentation. Extracellular bacteria located between epithelial cells may directly alter intercellular junctions [6]. Also, bacteria clustered in hypoxic, necrotic, hypoperfused, or immune infiltration regions may regulate tumor growth by changing the local microenvironment [4]. These mechanisms are not interchangeable. As demonstrated in some studies, the intracellular and extracellular bacteria observed under a microscope in pan‐cancer tissues, as well as the spatial organization of the microbial microbiome described in certain cancer types, address different biological questions [3, 7]. Interpretation is further complicated by the low load of the microorganisms in many tumor specimens. The Race DNA from reagents, paraffin processing, lab surfaces, and cross‐contamination between samples can equal or exceed the true microbial signal [8, 9]. On the other hand, even a compelling spatial signal does not mean that an organism is viable or caused the host phenotype. Therefore, detection of microorganisms requires distinguishing four concepts: detection, spatial localization, viability or activity, and functional causality. We first construct a hierarchy for spatial and functional evidence, compare current detection platforms, define minimum control limits for studies with low biomass, and consider cancer type‐specific microbial ecosystems and distinguish intracellular, extracellular, and microcell spatial models. Finally, we summarize the clinical relevance of spatial localization using sequence‐based localization, biological consequences, and clinical relevance with immunotherapy, prognosis, and translational maturity. Hierarchy of Evidence, Detection Strategies, and Methodological Challenges for Intratumoral Microbial Spatial Localization A Five‐Tier Framework for Spatial Evidence of Intratumoral Microbes and Its Biological Implications To avoid comparing “detection of microorganisms” with “proof of function at a specific location”, spatial evidence for intratumoral microorganisms can be divided into five levels. The first level is the tissue level, where bacterial DNA, RNA, antigens, or metabolites can be detected in whole tumor or homogenate, the second level is the region level, where signals can be localized in hypoxic, necrotic, perivascular, marginal, or specific regions, the third level is the cellular neighborhood level, where microorganisms are associated with the local microenvironment composed of specific cells, the fourth level is the intracellular level, where microorganisms can directly control cell signaling, the fifth level is the subcellular level, localized to specific compartments such as the cytoplasm, phagosomes/lysosomes, and perinuclear region (Figure 1). The five levels are nested. Different levels represent different interpretations of biological conditions. The tissue level shows past exposure to microorganisms, the regional level shows the tumor selectively selects niches for certain microorganisms, the neighborhood level describes micro‐regional phenomena such as immune exclusion and metabolic gradients, the intracellular level indicates bacteria can interact directly with the host signaling network and escape immune clearance, the subcellular level also distinguishes sites of action. For example, lysosomal localization indicates degradation or antigen processing, perinuclear enrichment is spatial proximity and cannot be directly interpreted as intranuclear activity. The spatial distribution of bacteria is not synonymous with function nor is there a causal relationship between the two. Reviews and original studies should report both spatial and functional levels simultaneously, rather than summarize intratumoral microbiology with “present or absence” binary classification. A Five‐Tier Framework for Spatial Evidence of Intratumoral Microbes and Spatial Detection Technologies (Created with BioRender.com). Spatial Detection Technologies for Intratumoral Bacteria No single platform simultaneously provides taxonomic information, high spatial resolution, viability assessment, host status measurement, and causal evidence. 16S ribosomal RNA (16S rRNA) sequencing and shotgun metagenomics provide general information on community composition and structure, but disrupt tissue architecture. Quantitative PCR (qPCR) and digital PCR (dPCR) can detect absolute or relative amounts of a specific DNA sample with high sensitivity, but do not provide any spatial localization information and are called blind assays [10]. Culture is the strongest direct evidence of viability, but is insensitive to fastidious, dormant, intracellular, or previously uncharacterized organisms and may selectively detect contamination introduced in tissue processing [11]. Microscopy preserves a wide range of environmental or taxonomic groups. Fluorescence in situ hybridization (FISH) can identify the RNA of a particular bacteria, immunohistochemistry (IHC) can label antigens on surfaces like lipopolysaccharides and stain markers on host cells to reveal which immune cells surround the bacteria. Confocal microscopy can also reconstruct 3D structures; electron microscopy can reveal ultrastructures of bacteria hidden inside organelles, but these methods are limited by the fact that they usually detect only a few known bacteria and cannot perform full, objective screening; antibodies may also cross‐react, which requires very strict control validation. High‐multiplex imaging platforms expand on these capabilities. Imaging Mass Cytometry and Multiplexed Ion Beam Imaging quantify host and microbial antigens at the cellular to subcellular level (nearly) but depend on the localization panel [12, 13]. Multiplexed Error‐Robust Fluorescence In Situ Hybridization (MERFISH), CosMx spatial molecular imaging, and Xenium target multiplexed transcript localization at single‐cell or subcellular level, and Spatial Enhanced Resolution Omics‐sequencing (Stereo‐seq) captures large tissues at high density array [14]. These platforms were developed for host transcripts, and specially designed probes are needed to detect bacterial transcripts; bacterial RNA is short and lacks a poly‐A tail, and interference from tissue autofluorescence must be addressed. Thus, high spatial resolution does not guarantee high specificity of the microbes. Based on the five levels listed above, detection of the microbes and host interactions depends on different technical approaches at each level: tissue level sequencing or culture can provide information but not location specificity; regional level in situ signals must be matched to specific histopathological structures; neighborhood level multiplex fluorescence labeling and spatial point‐pattern statistical methods should be used to quantitatively confirm the enrichment of targets around specific cell populations; intracellular level 3D imaging should check membrane boundaries and avoid adhesion or invasion artifacts; subcellular level observation of contact and physical relationships between bacteria and organelles to avoid overlap artifacts. (Figure 1) It is generally hard to directly measure the viability of bacteria at lower levels, and once microbes enter cells, functional validation is usually required. In order to understand host–microbe interactions, intervention experiments must combine in situ observations with functional perturbations. These five levels, along with viability and interaction evaluations, form the chain of evidence from localization to association and from phenotype to mechanism. Methodological Challenges and Quality Control Validation Strategies for Low‐Biomass Samples Tumor specimens harbor far less microbial biomass than the gut, creating a bottleneck for intratumoral microbiome detection. Bacterial DNA can enter at every step, from extraction kits, PCR reagents, water, paraffin, microtome surfaces, and the surgical environment to library preparation. And at low biomass, contaminants dominate and generate batch‐specific structures [15]. Consensus guidelines mandate contamination control from collection to reporting [16]. Rigorous studies should include sampling, extraction, paraffin/slide, and PCR/library blanks plus positive controls such as mock communities or dilution series, randomized across batches and blinded where feasible. Absolute load must be measured by calibrated qPCR, digital PCR, or microscopy rather than relative abundance alone, with host DNA, tissue mass, section area, and limits of detection reported. Computational filtering is ancillary and cannot replace controls. Prevalence‐based tools flag negative‐enriched sequences, yet contamination often correlates negatively with input DNA; mock dilutions show optimal filtering depends on biomass, controls, and batch, so universal blacklists or fixed thresholds fail. Because true low‐abundance taxa may be miscalled contaminants and reagent contaminants may seem tumor‐relevant, orthogonal validation is indispensable: confirm taxa by FISH, RNAscope, qPCR, or independent‐primer dPCR in a second cohort; assess intracellular localization via 3D boundary analysis with ultrastructural or functional support; substantiate viability through culture, membrane integrity, replication, or metabolic labeling; and base functional claims on causal designs such as microbial addition or removal, genetic perturbation, or metabolite rescue. Extremely low biomass with high contamination has already revised several early reports. Reports must disclose pre‐analytical details, all control results, batch allocation, absolute biomass, decontamination rules, image thresholds, and evidence level. Reproducibility across cohorts, labs, or platforms is integral to validation, not optional. [15, 16]. Spatial Localization and Functional Implications of Intratumoral Microbiota Cancer‐Specific Characteristics and Intratumoral Spatial Heterogeneity The interpretation of intratumoral microbiota must also address two equally important but non substitute scales simultaneously: ecological differences among cancer types and spatial heterogeneity among individual tumors. Pan‐cancer studies have shown that tumor types differ in terms of absolute bacterial load, community composition, dominant taxonomic groups, cell types preferences and detectable proportions, even on the same platform, breast cancer, colorectal cancer, oral/head and neck tumors, lung cancer and pancreatic cancer can differ markedly [3, 4]. The difference between cancer types determines background probability and interpretable range of signal, while intratumoral spatial heterogeneity determines which regions, cells neighborhoods or compartments in the same tumor actually carry higher local exposure. Comparing only the average abundance across tumors ignores local microenvironments. Mapping hot spots in one tumor while disregarding cancer types baselines may confuse common surface colonization with tumor colonization. The greater detection or capacity of microorganisms of some types of cancer may be from “opportunities of entry” and “local ecological permissiveness”. Tumors in the mouth, stomach, and other mucosal surfaces are near high‐biomass microbial reservoirs, barrier loss, ulceration, biofilm formation, and contact with luminal contents increase the chances for entry, Tumors in organs with nominally low biomass may harbor microorganisms by hematogenous or lymphatic dissemination, retrograde spread through ducts, barrier displacement or carried by immune cells and malignant cells [17]. Once inside the tumor, hypoxia, necrosis, low pH, release of nutrients, abnormal vascular permeability, dense extracellular matrix, local immunosuppression and radiation, chemotherapy or antibiotic exposure can explain differences in cancer types and even patients of the same type, but current evidence is mostly correlational, especially low‐biomass samples from pancreas and liver, where contamination by reagents, paraffin and processing procedures must first be excluded. Differences between cancer types and intratumoral heterogeneity should be viewed as two nested coordinates. Cancer specific origins, barrier states and metabolic environments provide the “ecological baseline”, hypoxic zones, necrotic cavities, perivascular regions, invasion fronts and areas of immune cell aggregation create local selection on top of this baseline [18]. We must compare the absolute load and composition between cancer types and map regional, neighborhood, intracellular and subcellular profiles of each tumor. Only linking “which types of tumors are more susceptible to colonization” and “where specific in that tumor colonization occurs” can distinguish universal exposure, cancer type selection, local functional microhabitats. The Influence of Tumor Histological Microstructures on the Spatial Distribution of Bacteria Within Tumors The localization of tumor‐associated bacteria depends on the tissue structure and its own physiological properties. Hypoxic and necrotic areas with low oxygen tension, accumulation of debris and inadequate perfusion often are preferred sites of tumor‐associated bacteria. Blood vessels and perivascular regions provide nutrients and circulation pathways for tumor‐associated bacteria. Fibrotic or immunocompetent areas form physical and immunological barriers [18]. For example, in oral and colorectal tumors, Fusobacterium nucleatum forms microcolonies within CD45+ immune zones and interepithelial spaces, which are often characterized by low numbers of T cells and inflammatory or oncogenic transcription programs. However, this work demonstrates spatial associations at the neighborhood level rather than definitive intracellular colonization [5]. Many studies on breast cancers show clusters of bacteria are concentrated in hypoxic and necrotic regions and separated from activated T cells, natural killer cells, and other tumors [19]. Other studies have shown that multiple bacteria in breast cancer are most commonly found in necrotic cavities of malignant cells, perivascular spaces, and perinuclear regions suggesting that the pathology and adaptation to the local environment can coexist [20]. Experimental bacterial therapies also tend to have region‐specific behavior; for example, Clostridium novyi‐NT spores germinate preferentially in hypoxic tumor tissue to exert their antitumor function [21] whereas genetically modified Salmonella colonize vascular, proliferative or quiescent regions depending on their chemotactic and growth properties [22]. Even though these models support hypoxia‐targeted strategies, the distribution of engineered bacteria cannot be equated to natural colonization patterns of endogenous intratumoral bacteria. Intratumoral bacteria have highly conserved spatial organization patterns based on local microenvironment characteristics. This pattern was repeatedly tested in different tumor types and experiments. It can thus be concluded that spatial localization is an important prerequisite for microbiota function; spatial distribution of bacteria can influence physical contact frequency and interaction distance between host cells, activating pro‐inflammatory or oncogenic signaling mechanisms by controlling local immunity and metabolism [18]. Using only microbiota abundance and neglecting spatial heterogeneity might underestimate or even misunderstand the biological role of microbiota (Table 1). TABLE 1Representative studies on intratumoral bacteria: Localization patterns, biological consequences, and clinical relevance.Cancer typeBacteria strainsBacterial localizationBiological consequenceClinical relevanceReferencesEsophageal squamous cell carcinoma Lactobacillus, Leptotrichia and others Cytoplasm of tumor and immune cellsPromotes immunosuppressive tumor microenvironmentRelated to prognosis[23]Oral squamous cell carcinoma and colorectal cancerRepresented by Fusobacterium nucleatum and Treponema Immunosuppressive and Low proliferation AreaAffects tumor cell migrationRelated to cancer progression[5]Oral squamous cell carcinoma and colorectal cancerMultiple bacteria ExtracellularDisruption of cell communication and cell cycle regulationChemotherapy resistance and tumor recurrence[24]Esophageal cancer, colorectal cancerMultiple bacteria Myeloid cells (especially monocytes‐macrophages)Upregulates antigen presentation genes in tumor cells downregulates in myeloid cells—[25]Breast cancerMultiple bacteria Immune‐excluded hypoxic necrotic niches——[19]Breast cancer Fusobacterium nucleatum Necrotic cavities, vascular regions, and perineural malignant‐cell regionsModulates the expression of tumor cell protein and MAPK‐associated tumor progressionImpacts tumor progression and providing novel therapeutic targets[20]Breast cancer Staphylococcus, Lactobacillus, Streptococcus and Enterococcus Tumor cellsModulates Toll‐like receptor signaling pathwayPotential metastasis target[26]Breast cancer Pseudomonas aeruginosa Cytoplasmic regions distal to the nucleus in tumor cellsQuorum‐sensing molecule disrupts TGF‐β/ERBB2 signaling and reduces transformationImmunotherapy resistance[27]Breast cancerSame bacterial strains in different statesIntracellular and ExtracellularIntracellular engagement drives suppressive neutrophil; extracellular exposure promotes antitumor responsesPossible localization‐guided therapy[7]Lung cancerTotal bacterial abundanceTumor cellsPositively correlates with oncogenic signaling pathwaysHypothesis‐generating biomarker[28]Gastric cancer Streptococcus anginosus Tumor cellsActivates TMPC‐ANXA2‐mediated oncogenic pathwaysPromote the progression of gastric cancer[29]Gastric cancer Streptococcus anginosus Tumor cellsMetabolizes arginine to suppress anti‐tumor CD8+ T cellsPromote the progression of gastric cancer[6]Gastric cancer Fusobacterium nucleatum Tumor cellsActivate the IL‐17/NF‐κB/RelB signaling axisPromotes gastric cancer progression and can serve as a new biomarker for predicting response to immunotherapy[30]Intrahepatic cholangiocarcinoma Paraburkholderia fungorum Tumor tissue: Tumor cells; Adjacent tissue: T cells, Parenchyma; Plasma: GranulocytesInhibits tumor proliferation and migrationAntitumor[31]Hepatocellular carcinoma Klebsiella pneumoniae Tumor cellsPBP1B on bacteria binds to TLR4 in HCC cellsPromote cancer progression[32]Hepatocellular carcinomaMultiple bacteria Tumor cells, CD8+ T cells, MacrophagesModulate immune functionPredicting liver cancer prognosis[33]Pancreatic cancerMultiple bacteria All cells, especially tumor cellsEnhances host cell motility, ECM interactions, and immune signalingRelated to tumors, immune response, and patient survival[34]Pancreatic cancerTotal bacterial abundancePredominantly immune cells (especially CD8+ T cells); Minority in tumor cellsHigher bacterial abundance in pancreatic cancer associated with immune responseSpatially restricted immune microbiomes have different effects on tumor biology[35]Colorectal cancer Fusobacterium nucleatum Tumor cellsPromotes antitumor exosome secretionPromote tumor metastasis[36]Colorectal cancer Fusobacterium nucleatum CD8+ T cellsModulates PD‐1 expression on CD8+ TILs via butyrate‐HDAC3/8‐TBX21 axisOptimize the effectiveness of immunotherapy[37]Colorectal cancer and liver metastases Fusobacterium nucleatum Tumor cellsCo‐migrates with tumor cells to maintain malignancyPromotes cancer progression and spread, linked to poor prognosis[38]Melanoma Fusobacterium nucleatum, Odontolyticus and Caprae Tumor cellsIntracellular bacterial peptides trigger immune reactivityAs a potential target for immunotherapy[39]Pan‐cancerMultiple bacteria Cancer‐related specificityRemodeling the immune microenvironmentAffecting immunotherapy and predicting its effectiveness[4]Taking gliomas and soft tissue sarcomas as examples Clostridium novyi‐NT Hypoxic regionsInduce an antitumor responseAntitumor[21]Non‐specific tumorLarge coloniesNear blood vesselsInduce an antitumor responseProvide a theoretical foundation for developing precisely controllable bacteria‐mediated tumor‐targeted therapy[22]Small coloniesPredominantly localized away from blood vesselsPancreatic, breast, and ovarian cancersFungiTumor cells—Cancer type specificity and location specificity[40]Melanoma and lung cancerMacrophages Representative studies on intratumoral bacteria: Localization patterns, biological consequences, and clinical relevance. Tissue‐Specific Localization and Subcellular Compartmentalization of Intratumoral Bacteria Intratumoral bacteria are not only found in tumor tissue but also spread in different cell types. A 2020 article published in Science shows that most of the intratumoral bacteria reside in tumor cells and immune cells, as part of the tumor microenvironment and have various effects on tumor formation, progression, metastasis and treatment [3]. Cell type specificity varies significantly among different types of cancer. Aikun Fu and colleagues observed that the number of intracellular bacteria in breast cancer cells was higher than normal breast cells. By controlling the Toll‐like receptor signaling pathway and remodeling the cytoskeleton of the cell, intracellular bacteria such as Staphylococcus, Lactobacillus, Streptococcus and Enterococcus help tumor cells fight blood flow shear stress and spread tumors, but not the primary tumor growth [26]. Pseudomonas aeruginosa localizes to the perinuclear cytoplasm of breast tumor cells, and its secreted N‐acyl homoserine lactone blocks the transforming growth factor beta (TGF‐β)–Erb‐B2 receptor tyrosine kinase 2 (ERBB2) pathway to impair trastuzumab efficacy, thereby promoting immune evasion and resistance [27]. Fusobacterium nucleatum concentrates in perinuclear regions of malignant breast cells, altering expression of RNAs/proteins linked to proliferation and metastasis while activating the mitogen‐activated protein kinase (MAPK) pathway [20]. In lung cancer spatial macro transcriptomics shows that the burden of bacterial cells in malignant cells positively correlates to many oncogenic transcriptional programs and that bacterial signaling is concentrated in tumor cells rather than immune cells [28]. Li Yuan et al. showed Streptococcus anginosus invading gastric cancer cell lines, promoting proliferation, migration, and invasion by inhibiting anti‐tumor CD8+ T cells and metabolizing arginine [29]. Streptococcus anginosus also activates oncogenic pathways via Transporter Membrane Protein C‐ Annexin A2 (TMPC‐ANXA2) interaction [6]. Fusobacterium nucleatum invades gastric cancer cells, activating interleukin‐17/nuclear factor kappa‐light‐chain‐enhancer of activated B cells/v‐rel reticuloendotheliosis viral oncogene homolog B (IL‐17/NF‐κB/RelB) signaling, recruiting Tumor‐Associated Neutrophils, polarizing them toward a pro‐tumor phenotype, and upregulating programmed death‐ligand 1 (PD‐L1) to enhance immune evasion [30]. Dysbiosis of Klebsiella pneumoniae disrupts the gut barrier and exacerbates hepatocellular carcinoma progression via Penicillin‐Binding Protein 1B‐ Toll‐Like Receptor 4 (PBP1B‐TLR4) signaling, suggesting specific localization links to tumor promoting or fighting mechanisms [32]. Bassel Ghaddar et al. identified microbes within pancreatic cancer cells, enhancing host cell motility, extracellular matrix (ECM) interaction, and immune signaling, and they prove that microbes are related to tumors, immune response, and patient survival [34]. Susan Bullman et al. found Fusobacterium nucleatum not only promotes colorectal cancer initiation but also co‐migrates with metastatic cells, maintaining malignancy [38]. In colorectal cancer, the bacterium colonizes malignant cells and induces host cells to produce exosomes rich in miR‐1246/92b‐3p/27a‐3p and C‐X‐C motif chemokine ligand 16/Ras homolog family member A/interleukin‐8 (CXCL16/RhoA/IL‐8) which accelerate metastases through multiple targets [36]. In melanoma, Fusobacterium nucleatum, Odontolyticus and Caprae in the tumor cells can be recognized by the body as antigens, thus extending the notion of cellular tropism from “where bacteria dwell” to “to whom bacteria are presented” [39]. Intratumoral bacteria are also present in immune cells, and spatiotropism in spatial tropism is closely related to immunoregulatory functions. In oral squamous cell carcinoma, the abundance of bacteria like Lactobacillus and Leptotrichia in tumor cells and immune cells is strongly associated with the infiltration levels of natural killer cells (NK cell), PD‐L1+ endothelial cells and tumor associated macrophages, suggesting that this type of intracellular bacteria might be widely involved in shaping an immunosuppressive environment and could become a marker for disease prognosis [23]. Specifically, in esophageal and colorectal cancers, bacteria in myeloid cells especially monocytes‐macrophages have been shown to bidirectionally regulate the expression of antigen‐presenting genes in the tumor cells, and in the myeloid cells [25]. Microscopic organisms in hepatocellular carcinoma can be present in tumor cells and immune cells. It is reported that microbial signaling in CD8+ T cells and high microbial loads in macrophage‐rich areas are both associated with patients' postoperative survival [33]. In pancreatic cancer, because the tumor is low‐mass, lipopolysaccharide is concentrated in the stroma or inhibitory compartments rather than in the malignant cells. The results differ between cohorts, so it is prudent to draw functional conclusions, suggesting that the spatial localization of bacteria in low‐mass tumors may shift toward the stromal compartments [35]. Xueliang Wang et al. detected the Fusobacterium nucleatum metabolite butyrate within T cells. This specific localization may facilitate the butyrate histone deacetylase 3/8‐ T‐box transcription factor 21 (HDAC3/8‐TBX21) axis, reducing programmed death‐1 (PD‐1) expression on CD8+ tumor‐infiltrating lymphocytes (TILs) and enhancing anti‐PD‐1 efficacy in microsatellite stable colorectal cancer (MSS CRC) [37]. Cell type dependent bidirectional gene regulation illustrates the role of spatial tropism in immunoregulatory output. In addition, intratumoral fungi also exhibit cancer‐specific cell tropism: In pancreatic, breast, and ovarian cancer, fungi are mostly found in cancer cells, in melanoma and lung cancer, fungi are mostly found in macrophages, and the fungal composition differs greatly in different anatomical sites, suggesting that while the basic pattern of spatial localization may transcend microbial types, combinations of microbial types vary depending on cancer type [40]. Once bacteria enter certain cell types, subtle differences in their subcellular localization also influence their functional output. Perinuclear localization could provide structure to the physical proximity between bacterial products to nuclear transcription machinery, thus facilitating signal regulation. In intrahepatic cholangiocarcinoma, tumor specific subcellular localization is striking: bacteria are typically found in tumor tissue in lysosomes of malignant cells, but in peritumoral tissue, they are predominantly located in T cells, and a representative species of bacteria, Paraburkholderia fungorum, are antitumor [31]. Cellular tropism and subcellular localization differ significantly between tumor and peritumoral tissue in the same organ; this suggests that localization itself may be a better predictor of functional orientation than bacteria. (Table 1). Extracellular Bacterial Communities Within Tumors Although intracellular bacteria in tumors have long been considered as driving tumor regulation, recent studies have demonstrated that extracellular bacteria also play functions. Based on their discovery of stable intracellular bacteria in breast cancer cells, Cai Shang's team found bacteria outside tumor cells can induce an antitumor function subset of neutrophils, activate immune responses and suppress the recurrence [7]. In colorectal cancer and oral squamous cell carcinoma regions rich in extracellular bacteria especially Fusobacterium nucleatum are significantly reduced cell density, weak proliferation and lower transcriptional activity suggesting low proliferation regions. In addition, these bacteria can arrest tumor cell cycle and increase cancer cell resistance to 5‐fluorouracil [24]. The disruptive effects of this “extracellular microbiome” on tumors challenge previous notions and provide new insights for future oncological diagnostic and therapeutics. Intratumoral microbes are not randomly distributed, but have cell‐specific localization and cell type tropism. Different cancer types do not harbor microbes randomly but instead screen for specific communities by their microenvironment. The same bacteria can drive either a pro‐ or anti‐tumor effect depending on cell type or spatial localization, and the mode of contact between bacteria and host cells can determine direction and intensity of downstream signaling pathway activation. Based on this, it is reasonable to conclude that spatial orientation is not just a distribution characteristic, but also a prerequisite for host cell fate regulation. Localization between bacteria and cells directly determines the molecular interface to host factors that controls immune recognition, signaling and drug response. Subcellular localization can be either perinuclear enrichment, intracellular transport, or retention for degradation. Assuming abundance, diversity and differences, spatial localization should also be taken into account in order to understand its biological function. (Table 1). Differences in Biological Functions Among Different Spatial Structural Patterns The spatial localization of bacteria in tumors can be classified into three structures: organized communities in space, intracellular bacteria, and extracellular communities, each representing different host interactions patterns and biological consequences. Organized communities in space cannot be limited to single bacterial‐cell pair. In this scenario, bacteria may be free extracellular state, adherent state, or in only few cells. However, the main point is remodeling surrounding tissue microenvironments (i.e., immune evasion, amplification of inflammatory signals, metabolic competition or local proliferation inhibition). Intracellular bacteria are the most direct host–microbe interaction mode. Bacteria colonization in malignant cells opens the signal transduction networks and cytoskeleton of the host cell directly to bacteria, and provides the bacteria with a refuge from extracellular immune clearance, allowing them to be metastasized and spread. Fusobacterium nucleatum is consistently detected in paired primary and metastatic colorectal cancer samples, suggesting that it may spread along tumor cells. Infected colorectal cancer cells can release metastasis‐promoting exosomes, which is a direct mechanistic explanation of how local intracellular interactions affect distant target cells [36]. Extracellular bacteria colonize a different kind of cells. Mechanistic experiments show that extracellular bacteria colonize cells through a species like Fusobacterium nucleatum can disrupt cell–cell contact, arrest G0‐G1 phase arrest, quit transcription and increase resistance to 5‐fluorouracil [24]. The same bacteria can induce opposite immune responses depending on the type of contact. In a breast cancer model, intracellular bacteria reprogram neutrophils into immunosuppressive cells by activating the cyclic GMP‐AMP synthase—stimulator of interferon genes‐ interleukin‐17B (cGAS‐STING‐IL‐17B) pathway in tumor cells, while exposure to extracellular bacteria promotes antitumor neutrophil activity and invasion leads to poor clinical results [7]. This suggests that location is a function determinant, not a descriptive label. The same unit of taxons can result in very different biological consequences in different spatial contexts. Future studies should not consider average abundance in tumor cells as a biological description but report both the physical state of bacteria and their spatial environment. From Microbial Localization to Biological and Clinical Translations Mechanisms by Which Microbial Localization Influences Function and Treatment Resistance Bacteria can influence key cell processes such as receptor signal transduction, DNA damage repair, cytoskeletal remodeling, autophagy, and metabolic availability [41, 42, 43]. Breast cancer studies link intracellular bacteria to metastatic survival and trastuzumab resistance [27]. Streptococcus pharyngitis in gastric cancer directly links epithelium invasion to metabolic reprogramming and oncogenic signaling [6, 29]. Fusobacterium nucleatum in colorectal cancer is persistently present in primary and metastatic lesions, and release of metastasis‐promoting exosomes suggests tumor‐associated bacteria may support tumor spreading [36]. Treatment resistance can develop through at least four spatially distinct mechanisms: direct bacterial metabolism of drugs, activation of bacterial signals in malignant cells, cell quiescence initiated by extracellular colonization, and remodeling of the immune microenvironment in the bacterial niche. These four mechanisms should not be confused because they correspond to different intervention strategies: intracellular bacteria may require antimicrobial drugs able to penetrate the cell membrane or host targeted therapies; extracellular bacteria may be sensitive to locally delivered antibiotics, bacteria, biofilm disruptors, or agents that induce cell cycle reversal [44]. Microecological niches dominated by immunosuppression may require immunomodulation rather than mere microbial elimination. Intratumoral Bacterial Localization and the Immune Microenvironment The tumor immune microenvironment is composed of T cells, B cells, NK cells, myeloid cells, and various stromal components. Functional orientation is determined jointly by cell type composition, activation status, spatial arrangement, and gradients of soluble mediators. Intratumoral bacteria can shape this microenvironment by pattern recognition receptors, nucleic acids sensing, metabolites, virality factors, and antigen presentation, but immune effects are not determined by species alone but depend on load, activity, host cell type, and spatial location [45]. One microorganism may cause inflammation and antigen presentation or may suppress myeloid suppression, T‐cell exhaustion, or immune rejection. Colorectal cancer and oral squamous cell carcinoma are enriched with myeloid cells, decreased T‐cell numbers, and immune suppression [5]. Head and neck squamous cell carcinoma contains a higher amount of bacteria with neutrophils, T cells, and exhaustion [46]. It is not proved that bacteria are the only culprit in the immune suppression: hypoxia, necrosis, and hypoperfusion can affect the distribution of bacteria and immune cells simultaneously. The host cell also determines which type of host cell initiates the immune signal. Bacteria can activate intracellular signaling pathways such as chemokines, PD‐L1, antigen processing, and cell death programs in malignant cells. In breast cancer models, invading tumor cells induce immunosuppressive neutrophils, while extracellular exposure induces the opposite effect [7]. Bacteria signaling in myeloid cells can be active infection or phagocytic clearance; the latter continuously reprograms immune function, while the latter mimics antigen processing or an inflammatory response [25]. Melanoma cells can load bacterial‐derived peptides onto human leukocyte antigen (HLA) molecules for recognition by T cells, suggesting that bacterial antigens enter the tumor antigen processing pathway [39]. Cytoplasmic localization may activate nucleic acid recognition pathways such as cGAS‐STING, while endosomal or lysosomal localization triggers TLR activation, degradation, and antigen processing; however, lysosomal colocalization does not equal live bacterial colonization, and perinuclear signals may also be caused by transport or overlap projections. Extracellular localization activates surface receptors, complement and neutrophils by microcolonies, biofilm‐like structures and soluble products alter epithelial junctions, perfusion and metabolites to form microenvironments of immune activation or suppression without cells invade. Notably, the same bacteria can respond differently intracellularly and extracellularly, suggesting that immune effects cannot be predicted by a single total tissue abundance [7]. The intracellular, extracellular, and subcellular organelle localizations of bacteria are major parameters for the pathology of the immune microenvironment, thus providing a theoretical basis for precise immune interventions based on localization. (Figure 2). Impact of intratumoral bacteria on the tumor immune microenvironment and immunotherapy. Intratumoral bacteria preferentially localize in hypoxic, necrotic, hypoperfused, and immune infiltration areas within tumors. As components of the immune microenvironment, they influence immunotherapy efficacy, manifesting as immunoresistance or immunosensitization (Created with BioRender.com). Intratumoral Bacterial Localization and Immunotherapy Intratumoral bacteria may alter immune checkpoint blockade by altering antigen presentation, interferon signaling, myeloid suppression, T‐cell infiltration and metabolic competition [43], but current evidence suggests that their direction depends on tumor type and location [3, 5]. The strongest human evidence is head and neck squamous cell carcinoma; Silver et al. found that higher total tumor bacteria burden was associated with poorer immune checkpoint blockade response in clinical trials and independent validation cohorts, and mechanistic models show that increasing the intratumoral bacteria burden can increase neutrophil enrichment, T‐cell exhaustion and anti‐PD‐L1 tolerance [46]. This paper supports the possibility of “total load” as treatment predictor in some cancer types. However, the main indicator is tumor load, and it has not yet been shown that a single bacteria or subcellular location can predict treatment efficacy for all cancer types. In microsatellite stable colorectal cancer, higher levels of intracellular Fusobacterium nucleatum correlate with better anti‐PD‐1 responses, and mechanistic studies suggest that its metabolites may improve CD8+ T cells [37]. In a single‐center retrospective study of patients with high PD‐L1 expression in lung cancer, there was also a general association between tumor microbiome composition and immunotherapy response, but because of small sample size, external validation is still needed [47]. These inconsistent results are not simply contradictory but may reflect differences in baseline cancer characteristics, dominant bacteria species, bacterial load, intracellular/extracellular state, and microenvironment of immune cells. In fact, it is not appropriate to generalize the presence of bacteria as universally sensitizing or universally resistant. Possible mechanisms by which intratumoral bacteria promote resistance to immunotherapy include at least: formation of an immune barrier dominated by neutrophils and myeloid suppressor cells in hypoxic or necrotizing areas, activation of oncogenic signaling, chemokines or immune checkpoint programs in malignant cells, alteration of antigen presentation and inflammatory polarization in myeloid cells, and decreased ability of effector T cells to approach tumors or kill tumors through extracellular colonies, metabolites, and cell quiescence. On the other hand, T‐cell function enhanced by HLA‐presented bacteria, extracellular immune stimulation, or specific molecules, or tumor suppressive signaling pathways, may increase sensitivity to immunotherapy. Localization information is therefore critical for distinguishing sensitizing from resistance mechanisms. At this stage, intratumoral bacteria cannot yet be a universal, independent and clinically valid biomarker for immunotherapy effectiveness. Total bacterial burden in head and neck squamous cell carcinoma has independent predictors [46], but prospective multicenter studies, predefined thresholds, absolute burden measurements, and standard tissue sampling are still needed, while confounding factors such as antibiotic use, oral or intestinal origin, necrosis proportion, sample preservation, tumor purity and number of treatment lines must also be adjusted, and more importantly, it should be shown that spatial indicators add predictive value over existing markers such as PD‐L1, microsatellite instability/mismatch repair (MSI/MMR), tumor mutational burden (TMB), and immune infiltration and distinguish prognosis and treatment benefit. Future translation candidates will not be relative abundance of one single bacterium, but composite biomarkers consisting of absolute load, activity, regional/cellular localization, and host immune status. (Figure 2). Association Between Intratumoral Bacteria and Clinical Outcomes Intratumoral bacteria are associated with diagnosis, prognosis, and treatment response in patients with different types of cancers. The association between Helicobacter pylori and gastric cancer [48], human papillomavirus (HPV) and cervical cancer [49] has been confirmed in many studies; pathogen testing for both has become part of risk management and early diagnosis systems for both cancers. A study of 312 head and neck cancer patients showed that high tumor bacterial load and low microbial diversity are independent adverse prognostic factors for head and neck cancer (especially HPV‐negative subtype) [50]. Combined indicators can predict better than other individual clinical and pathological factors alone. In addition, many clinical cohort studies have shown that high tumor bacterial load is associated with resistance to immunotherapy [46]. A lung cancer study with 507 surgically treated patients showed that high tumor bacterial load is associated with overall survival and recurrence‐free survival based on average abundance obtained by tissue homogenate sequencing; 16S FISH only validated histological spatial data, and localization status has not yet been taken into account for prognostic models [51]. Colorectal cancer is associated with poorer survival and resistance to chemotherapy [24]. In pancreatic cancer, intratumoral microbial composition can distinguish long‐term survivors from short‐term survivors [52]. Bladder cancer and microsatellite‐stable colorectal cancer bacteria correlate with PD‐1 blockade efficacy and predictive power [37, 53]. The repeated patterns of association across these types suggest that intratumoral information may provide a new dimension for risk stratification than standard prognostic assessment based on staging and molecular subtyping and a new perspective for explaining differences in immunotherapy effectiveness. However, current evidence is based mostly on average abundances from tissue homogenate sequencing, and hence it is hard to establish causality. Further, spatial localization is not yet included in the prognostic or predictive model, and no intratumoral bacterial biomarkers or therapies have been included in clinical decisions. Specifically, spatially targeted engineered bacterial therapies are changing intratumoral bacteria from prognostic markers to direct intervention targets, using tumor‐seeking bacteria to deliver in situ drug and offering a new therapeutic approach for traditionally refractory lesions [54]. Future work will focus on standardizing detection methods and testing the incremental predictive value of spatial localization information and the clinical feasibility of engineered bacterial therapies in prospective cohorts to drive this field from discovery to clinical translation. Therapeutic Translation of Spatial Localization of Intratumoral Bacteria: Strategic Foundations, Hierarchy of Evidence, and Barriers to Clinical Development Spatial Localization‐Guided Therapeutic Strategies Treatment based on spatial information can be divided into two ways: using exogenous engineered bacteria to deliver payloads directly to tumor regions via chemotaxis or directly kill tumor cells, or using elimination, blockage, or immunological exploits based on the locations of endogenous intratumoral bacteria [43]. The first approach relies on the natural hypoxic tropism of exogenous engineered bacteria. Anaerobic and facultative anaerobic strains generally preferentially germinate or develop in low perfusion, hypoxia, and necrosis, where they release local cytokines, toxins, enzymes, nanomaterials, or immunostimulants (the area that drugs and effector lymphocytes struggle to reach) [55]. However, “regional targeting” is simply showing the spread of bacteria in the stroma or necrosis, not “automatically” allowing delivery into tumor cells. Extracellular and intracellular states correspond to different intervention strategies: Extracellular bacteria are better suited for local antimicrobial agents, bacteriophages, biofilm disruptors, cell wall‐targeting antibodies/nanoparticles, or engineered bacterial vectors activated in hypoxic environments. Cytokines, enzymes, metabolites, drugs and immunostimulants release can act on neighboring cells from the extracellular space [42, 56]. Babarian bacteria actually found in malignant or immune cells may escape extracellular antibodies and some antibiotics, need cell‐penetrating drugs, modulating host cell autophagy or lysosomal pathways, or immune clearance by the antigen presentation by infected cells. The clinical safety of engineered bacteria designed for active invasion or direct killing of tumor cells is unknown yet. The second approach directly uses endogenous intratumoral bacteria as targets or sources of antigens. Preclinical studies show that targeted delivery of anticancer drugs using liposomes or nanoparticles as well as multivalent antimicrobial vaccines can eliminate tumor bacterial populations locally, or activate antitumor immunity using microorganisms as antigen sources for therapeutic effects against cancer [57, 58, 59]. However, broad‐spectrum elimination can remove bacterial populations beneficial to immunotherapy and alter gut microbiome or normal tissue; spatially targeted interventions are necessary. Progress in clinical translation of these two approaches provides early evidence for spatial selection but each faces its own challenges. After intravenous administration of the attenuated Salmonella strain VNP20009, bacteria colonized tumors only in some patients, and no objective response was obtained [60]. Intratumoral injection of Clostridium novyi‐NT spores can germinate in hypoxic tumors and cause local oncolysis, but this is associated with infection and inflammation and not injectable [61]. Phase Ib study combining this approach and pembrolizumab also assessed safety and preliminary activity but is still in early development. The injection of the engineered Escherichia coli Nissle strain SYNB1891 in a single injection can induce STING‐related signaling, but the results have not been proven as monotherapy or in combination with atezolizumab [62]. The trials support hypoxic/necrotic regions and local injection sites, but they do not show stable intratumoral delivery in patients. For the endogenous microbial strategy, translational research also needs to determine the absolute burden, activity, distribution, intracellular or extracellular status via paired biopsies before treatment, and track the clearance, redistribution and changes in the host immune response after treatment. Both approaches depend on “spatial information” as the basis of treatment decisions. However, the exogenous approach is spatially precise for “delivery and release” and the endogenous approach is spatially selective for “targeting and clearance”. Future directions lie in the integration of spatial omics with engineered bacterial design to upgrade from passive colonization to active response. (Figure 3). Specific localization of engineered bacteria within the tumor microenvironment affects cancer therapy. (A) Exogenous engineered bacteria localize to hypoxic, necrotic, hypoperfused, immune infiltration niches, secreting cytokines, enzymes, metabolites, and delivering drugs, immune agonists to trigger immune‐mediated or direct tumor killing. However, intracellular bacteria raise safety concerns due to poor post‐treatment clearance. (B) Endogenous engineered bacteria activate immunity via antigen presentation, antibacterial vaccines or target anticancer drug delivery for tumor cell killing (Created with BioRender.com). Barriers to Clinical Development The clinical translation of bacterial therapies is challenging. Engineered bacteria have proved feasibility in the preclinical stage but only a few have actually been clinically validated. Preclinical efficacy cannot replace the validation of colonization capacity, safety, and benefits required in the patient. Even when the mechanism of action and efficacy of engineered bacteria have been validated in animal models, delivery efficiency remains the main bottleneck on the path to clinical translation. The oral route is limited by gastric acid, bile, competition from gut microbiota and epithelium; intravenous injection involves complement activation, phagocytic clearance, and vascular filtration; direct intratumoral injection is often difficult in deep, multifocal, or metastatic tumors [63, 64, 65, 66]. Even if the bacteria reach the tumor site, dense extracellular matrix, high interstitial pressure, oxygen heterogeneity, and immune clearance can still lead to uneven distribution in the tumor [67]. Pharmacokinetic studies in humans have shown that Escherichia coli Nissle and its derivatives have limited persistence in vivo and are heavily dependent on the microenvironment, so reaching the tumor is only the first step, and stable colonization is equally critical [68]. Colonization stability and genetic controllability are also the second issue [69, 70]. Under selective pressures of the tumor environment, engineered bacteria can lose therapeutic elements, experience spontaneous mutations, horizontal gene transfers, and even phenotypic drift. Protective design should include nutritional modification, kill switch, replication restriction, antibiotic rescue, monitoring detachment and dissemination, genetic stability testing, and the production process must also be controlled to ensure activity, potency, purity, and batch consistency; regulatory oversight must adapt to live evolutionarily capable products. Lack of stability not only weakens therapeutic effectiveness but also increases safety risks: sepsis, persistent infection, toxin release, chronic immune activation, horizontal transfer of drug resistance genes, and unintended intestinal disruption. These risks are especially high in neutropenia, mucosal damage, obstruction of the biliary tract, or immunosuppression patients [67, 71] In view of this, long‐term follow‐up and clear‐cut procedures are needed. To address these challenges, spatial pharmacology should be integrated with microbiology monitoring in the design. At the clinical trial level, baseline intratumoral bacterial load, target intention is validated, bacterial viability and load expression are measured, systemic distribution and clearance dynamics are tracked, and the correlation between these variables and pharmacodynamic and clinical endpoints is established. The main benefit of such a design is that it can objectively assess the reproducibility of spatial targeting strategies in patient populations rather than only extrapolating from particular animal models, which is an important test for the field to move from proof of concept to clinical application. Conclusions and Future Directions The work on intratumoral bacteria has evolved from high‐throughput detection to multilevel analysis consisting of regional spatial mapping, cell neighborhoods, host‐cell interactions, and subcellular localization. The main conclusion of this evolution is that while taxonomy is the source of their functional potential, spatial localization is also crucial to their biological function. Specifically, intracellular bacteria, extracellular communities, and organized communities correspond to different action strategies; even if taxonomic units are identical, differences in location can lead to significant differences in functional outputs. Also, evidence level must match conclusions: detection of sequencing signal is not active colonization, spatial colocalization differs from sub‐level confirmation of intracellular localization, statistical associations must be distinguished from causal effects, and preclinical therapeutic effects must be differentiated from clinical benefits. For low biomass studies, comprehensive controls, absolute quantification, transparent decontamination, and orthogonal validation should also be included. Future research should prioritize longitudinal sampling, three‐dimensional and multimodal imaging, pairing detection of load and vitality, organoid or tissue models of preserved spatial structure, microbial genetic perturbation systems, and independent clinical validation. Incorporating bacteria classification and spatial status into a multidimensional framework comprising host genome, immune system, metabolism, and therapeutic exposure could yield clinically relevant biomarkers or interventions. Until then, spatial microbiology should be used to generate and test mechanistic hypotheses rather than to assume that every signal is a confirmed, causal, or therapeutically available intratumoral biological entity. Author Contributions Minlan Ren: conceptualization, investigation, writing – original draft, writing – review and editing. Jing Jia: conceptualization, investigation, writing – original draft, writing – review and editing. Tingting Li: investigation, validation. Yiwei Li: investigation, validation. Nanxiang Shen: investigation, validation. Yun Shi: investigation, validation. Zhe Wang: conceptualization, writing – original draft. Hao Fan: conceptualization, writing – original draft. Guangli Sun: conceptualization, writing – original draft. Guanying Xiong: supervision, writing – original draft. Gang Li: writing – original draft, writing – review and editing, supervision. Rui Peng: writing – original draft, writing – review and editing, supervision, conceptualization. Funding This work was supported by National Natural Science Foundation of China, 82203226, 82303602, 82404002. Yishan Research Project of Jiangsu Cancer Hospital, YSZD202404. Research Project of Jiangsu Cancer Hospital, XHMS202401. China Postdoctoral Science Foundation, 2022M721410. Jiangsu Provincial Association for Science and Technology Youth Talent Lifting Project, JSTJ‐2023‐WJ013. Jiangsu Province “333 Project” High‐Level Talent Training Program, 2024‐3‐2249. Ethics Statement The authors have nothing to report. Conflicts of Interest The authors declare no conflicts of interest. Acknowledgments The authors have nothing to report. Contributor Information Guanying Xiong, Email: xiong_gy@126.com. Gang Li, Email: ligang@njmu.edu.cn. Rui Peng, Email: pengrui@njmu.edu.cn. Data Availability Statement Data sharing not applicable to this article as no datasets were generated or analysed during the current study. References Associated Data Data Availability Statement Data sharing not applicable to this article as no datasets were generated or analysed during the current study.