Nasopharyngeal carcinoma, or NPC, is a cancer of the upper airway mucosa that is strongly linked to Epstein-Barr virus, or EBV. The nasopharynx is constantly bathed in commensal and environmental microorganisms, which makes it a notoriously difficult place to separate meaningful tumor-associated bacterial signals from background noise. Previous work has reported altered oral and nasopharyngeal microbiota in NPC patients, and some studies have linked intratumoral bacterial burden with clinical outcomes. What remained unclear was whether bacteria detected within tumor tissue are compositionally distinct from organisms present on the lesion surface — a question most earlier studies never addressed directly because they sampled only one compartment or the other.

The research team, led by Nan Shi, Zongjian Huang, and Wei Xia, enrolled 38 newly diagnosed NPC patients and 26 people with benign nasopharyngeal lymphoid hyperplasia, an inflammatory but noncancerous condition that served as the control. During endoscopic biopsy, each excised specimen was rinsed, swabbed with a sterile nylon-flocked swab to capture its surface microbiome, rinsed three more times, and then flash-frozen for tissue analysis. An air-exposed swab left in the operating room for thirty minutes served as an environmental control. This paired design allowed the investigators to compare surface and tissue communities within the same lesion from the same patient, minimizing the interindividual variability that plagues most microbiome studies.

Before any sequencing, the team first established that bacteria were genuinely present in the tumor tissue. Fluorescence in situ hybridization with a pan-bacterial 16S rRNA probe revealed discrete punctate bacterial signals within NPC biopsies, signals that were rare in the benign control tissue. Quantitative PCR on an independent cohort of 24 NPC and 18 NLH samples detected a median of roughly 7,760 bacterial 16S copies per nanogram of DNA in NPC tissue, about eight times higher than the 907 copies found in NLH tissue, and both values exceeded the non-template and environmental background controls. Cultures of tissue homogenates and enzymatically dissociated single-cell suspensions yielded bacterial growth, while PBS washes and environmental controls remained sterile — orthogonal evidence that viable bacteria inhabit NPC tissue.

Perhaps the most provocative culture finding was that growth persisted after treatment with ampicillin plus gentamicin, two antibiotics that struggle to penetrate host cells, but was suppressed by doxycycline, which penetrates cells far more readily. The authors interpret this pattern as compatible with a protected or intracellular bacterial fraction, echoing the landmark 2020 Science study by Nejman and colleagues that described tumor type-specific intracellular bacteria across many cancer types. They are careful, however, to note that the aerobic culture assay did not identify the recovered colonies taxonomically, so culture positivity supports viability but does not independently validate the specific taxa detected by sequencing.

For the microbial profiling itself, the team deployed a five-region 16S rRNA gene sequencing strategy, simultaneously targeting the V2, V3, V5, V6, and V8 hypervariable regions to squeeze more taxonomic resolution out of low-biomass samples than conventional single-region approaches allow. Sequencing was performed on an Illumina NovaSeq 6000, processed through QIIME 2, and subjected to a stringent contaminant-removal pipeline: any taxon appearing in at least half of the negative-control libraries was discarded as a probable laboratory or sampling artifact. Samples with fewer than 1,000 total reads were excluded entirely. These safeguards matter enormously, because low-biomass microbiome research is famously vulnerable to reagent and environmental contamination that can masquerade as biological signal.

The results were striking. Surface swabs showed higher species richness and diversity than the paired tissue samples, as expected for a mucosal surface exposed to the environment, but swab communities overlapped almost completely between NPC and NLH patients — an analysis-of-similarity statistic of negative 0.038 with a p-value of 0.88, meaning essentially no separation whatsoever. Tissue communities, by contrast, separated cleanly between the two groups, with ANOSIM R values above 0.5 and p-values below 0.01. Even after adjusting for age and sex, disease status remained significantly associated with tissue community composition (PERMANOVA p < 0.001) but not with swab composition. NPC tissue also retained higher Shannon diversity than NLH tissue after covariate adjustment. An exploratory random-forest classifier underscored the gap: tissue-derived features achieved an area under the ROC curve of 0.927, while swab-derived features managed only 0.408 — worse than a coin flip.

The team then integrated the microbial data with matched bulk RNA sequencing of the same tissues to ask whether bacterial abundance covaries with host and viral biology. As expected, EBV transcript levels were higher in NPC, and the tumors showed elevated activity of proliferation and cell-cycle programs including E2F, MYC, G2M checkpoint, and mTORC1 signaling, alongside a remodeled immune landscape inferred by CIBERSORTx deconvolution. Within the NPC group, several tissue-enriched bacterial taxa — including species-level assignments for Comamonas, Streptococcus, Prevotella intermedia, Corynebacterium striatum, and Peptostreptococcus stomatis — showed associations with higher EBV transcript levels, proliferative pathway activity, and lower inferred CD8 T-cell and plasmacytoid dendritic-cell signatures. Neisseria perflava stood out for its extensive connectivity, correlating with MYC targets, the G2M checkpoint, DNA-repair programs, EBV transcripts such as A73 and RPMS1, and inferred CD8 T-cell abundance.

The authors are appropriately measured about what these correlations mean. Because the data derive from 16S profiling and bulk RNA sequencing, the directionality of the relationships cannot be resolved: bacteria might promote viral activity or immune remodeling, EBV-driven transformation might create a tissue environment permissive to particular bacteria, or the microbial signatures might simply mark a pre-existing tumor niche shaped by hypoxia, altered barrier function, and dampened immune surveillance. Co-occurrence networks built from the compositional 16S data suggested that the NPC tissue microbial community is more clustered yet less densely connected and apparently less robust than that of benign tissue — a hypothesis-generating observation, the authors stress, rather than proof of ecological instability.

Several limitations temper the conclusions. The cohort was modest and drawn from a single institution, the NLH controls are inflammatory rather than histologically normal mucosa, and lifestyle variables such as smoking, diet, oral health, and recent antibiotic use were unavailable, leaving residual confounding possible. The biopsy-surface swabs were taken from excised, rinsed specimens rather than the living mucosa, so they cannot speak to non-invasive screening strategies. Despite the rinsing protocol, residual surface-adherent bacteria may still contribute to the tissue profiles. Still, the core message stands: nasopharyngeal carcinoma harbors a tissue-resident microbiome that is compositionally distinct from both the surface of the lesion and from benign inflammatory tissue, and its members covary with EBV activity and immune composition. The authors call for larger cohorts, spatial localization techniques such as FISH-based co-localization, isolate-level characterization by whole-genome sequencing, and epithelial–EBV–bacteria co-culture models to determine whether these bacteria play a functional role in one of the world’s most geographically concentrated cancers — or merely serve as fingerprints of an EBV-conditioned tumor microenvironment.

Subject of Research: Tissue-associated bacterial microbiome signatures and their association with EBV transcription and immune composition in nasopharyngeal carcinoma

Article Title: Paired biopsy-surface and tissue microbiome profiling reveals tissue-associated bacterial signatures in nasopharyngeal carcinoma

Article References: Shi, N., Huang, Z., Wang, X., Xie, Y., Tang, A., & Xia, W. (2026). Paired biopsy-surface and tissue microbiome profiling reveals tissue-associated bacterial signatures in nasopharyngeal carcinoma. Frontiers in Cellular and Infection Microbiology, 16, Article 1892233. https://doi.org/10.3389/fcimb.2026.1892233

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1892233

Keywords: nasopharyngeal carcinoma, tumor microbiome, Epstein-Barr virus, 16S rRNA sequencing, intratumoral bacteria, biopsy-surface swabs, immune microenvironment, RNA-seq, Neisseria perflava, bacterial culture, Guangxi, cancer biomarkers

The review builds on the concept of the microbiota-gut-brain axis, a bidirectional communication network linking the gastrointestinal tract with the central nervous system through neural, endocrine, immune, and metabolic pathways. Gut-derived signals can reach the brain via vagal afferents, circulating immune mediators, microbial metabolites, and gut hormones, while autonomic and hypothalamic-pituitary-adrenal signaling can in turn modify intestinal motility, barrier function, and microbial composition. Microbial products such as short-chain fatty acids, tryptophan derivatives, and lipopolysaccharide are known to influence blood-brain barrier homeostasis, microglial activity, immune-cell differentiation, and epigenetic regulation. Much of this mechanistic evidence, however, comes from germ-free animals and non-cancerous neurological models, and its direct relevance to human glioblastoma remains incompletely established.

The tumor microenvironment of glioblastoma provides a plausible target for such systemic signals. The tumor is dominated by myeloid cells, particularly tumor-associated macrophages that arise from both resident microglia and infiltrating monocytes. Single-cell studies show these populations occupy distinct spatial niches and transcriptional states, rendering the conventional M1/M2 classification overly simplistic. Glioma cells recruit and shape myeloid cells through mediators including colony-stimulating factor 1, CCL2, transforming growth factor-beta, and interleukin-10, while regulatory T cells and exhausted T cells expressing PD-1, TIM-3, LAG-3, and CTLA-4 further entrench immune suppression. Immune dysfunction also extends beyond the tumor itself: treatment-naive patients show sequestration of naive T cells in the bone marrow, impairing T-cell egress and helping explain why checkpoint blockade alone has failed in this disease.

Against this backdrop, experimental evidence from mouse models has begun to suggest that gut microbes can indeed modulate glioma-associated immunity. In one of the earliest glioma-specific studies, chronic antibiotic exposure in mice bearing intracranial GL261 tumors altered intestinal bacterial composition, reduced cytotoxic natural-killer-cell subsets, modified microglial inflammatory proteins, and was accompanied by accelerated tumor growth. A subsequent orthotopic model showed that antibiotic-induced dysbiosis reduced short-chain fatty acid concentrations in blood and tumor tissue and sped tumor progression, while oral supplementation with these metabolites increased glycolytic activity in tumor-associated macrophages and improved outcomes. More strikingly, fecal microbiota transfer experiments demonstrated that mice receiving microbiota from patients with high-grade glioma developed faster-growing tumors than mice receiving microbiota from healthy donors, with bacterial genera such as Eisenbergiella and Merdimonas enriched in the recipients.

Metabolite signaling sits at the heart of these mechanisms, but the review stresses a critical caveat: a metabolite relevant to glioblastoma is not necessarily microbiota-derived. Short-chain fatty acids such as butyrate act through G-protein-coupled receptors and epigenetic mechanisms, inhibiting histone deacetylases and influencing regulatory T-cell differentiation. Sodium butyrate has reduced proliferation and promoted apoptosis in human glioma cell lines, and in one mouse study it increased PD-L1 expression and enhanced the efficacy of anti-PD-1 therapy. Yet the millimolar concentrations used in vitro do not demonstrate that microbiota-derived butyrate reaches human glioblastoma tissue at comparable levels. Similarly, the kynurenine pathway engages the aryl hydrocarbon receptor, but in glioblastoma the strongest evidence points to kynurenine produced by tumor cells themselves rather than by gut microbes. A recent study also showed that tumor-cell-derived spermidine suppressed CD8-positive T-cell activity and that blocking its synthesis prolonged survival in mice, again without any established microbial contribution.

Perhaps the most provocative findings concern microbial material detected within human brain tumors themselves. A prospective multi-institutional study of 243 samples from 221 patients combined RNA-based fluorescence in situ hybridization, immunohistochemistry, spatial imaging, sequencing, and culturomics, detecting bacterial 16S rRNA and lipopolysaccharide signals in subsets of glioma and metastatic samples, including within tumor, immune, and stromal cells. Signals assigned to genera including Fusobacterium, Prevotella, and Veillonella partially overlapped with matched oral or gut communities. Separately, an immunopeptidomic analysis of tumors from nineteen patients identified 344 distinct bacteria-associated peptides bound to HLA class II molecules, and some of these peptides were recognized by tumor-infiltrating T cells, including a CD4-positive T-cell clone that cross-recognized both a mutated glioblastoma antigen and several microbial peptides.

These discoveries come with heavy methodological caveats. Brain tumors contain extremely low microbial biomass, making sequencing studies vulnerable to reagent contamination, environmental DNA, and computational misclassification. The 2024 retraction of a prominent TCGA-based cancer microbial-signature study underscored the danger. Standard culture failed to recover a readily cultivable bacterial community from the prospective brain-tumor samples, the signals were sparse and heterogeneous, and no association with clinical outcome was established. The review therefore recommends reserving the terms microbiota and microbiome for settings where a diverse, biologically active community is supported by adequate evidence, and instead using phrases like microbial signals or bacterial elements until viability, replication, and function are demonstrated. Rigorous negative controls, orthogonal spatial validation, and viability assessment are deemed essential before any claim of stable intratumoral colonization can stand.

On the translational front, the review finds everything still at the earliest stage. No gut- or tumor-derived microbial biomarker has been independently validated for glioblastoma diagnosis, prognosis, or treatment selection, and current prospective studies such as THERABIOME-GBM remain observational. Fecal microbiota transplantation has shown clinical proof of principle in melanoma, where it can promote responses to checkpoint inhibitors, but no therapeutic trial exists in glioblastoma, and safety concerns are serious: FMT has transmitted multidrug-resistant organisms to immunocompromised recipients, including a fatal infection. Probiotic combinations of Bifidobacterium lactis and Lactobacillus plantarum reduced tumor volume in mice, and even dietary manipulations such as a short-term high-glucose drink or aspartame exposure altered tumor biology in animal models, but the authors explicitly caution that none of these findings justifies recommending any microbiome-directed intervention to patients.

Looking ahead, the review identifies several promising but speculative directions. One is the intersection between microbiota-associated inflammation and the circadian-glymphatic-melatonergic network, where experimental sleep-restriction data suggest melatonin can preserve glymphatic function, and a proposed but unproven pathway links aryl hydrocarbon receptor activity to melatonin metabolism and N-acetylserotonin signaling through the TrkB receptor. Another is bacteriophage-based modulation, which in colorectal cancer models has successfully targeted Fusobacterium nucleatum to augment chemotherapy, though no viable bacterial driver has been validated in glioblastoma and delivery across the blood-brain barrier remains unsolved. The authors conclude that progress will depend on source attribution, orthogonal validation, perturbation-and-rescue experiments, longitudinal human sampling, and replication across centers. The key distinction, they argue, is between a microbiota-associated signal and a clinically established mechanism, and only evidence that a defined microbial feature is reproducible, mechanistically linked to tumor or immune phenotypes, and safely modifiable will allow microbiota research to contribute meaningfully to precision neuro-oncology.

Subject of Research: The role of the gut microbiome and microbial metabolites in glioblastoma tumor-immune interactions and translational oncology

Article Title: The Microbiome in Glioblastoma: Mechanisms, Tumor–Immune Interactions, and Translational Perspectives

Article References: AlRamadneh, T. N., Jyothi‐S, R., Priyadarshini‐Nayak, P., Nanda, A., Al‐Hasnaawei, S., Bhatt, A., Singh‐Chauhan, A., Singla, S., & Mishra, M. K. (2026). The Microbiome in Glioblastoma: Mechanisms, Tumor–Immune Interactions, and Translational Perspectives. MicrobiologyOpen, 15(5), Article e70431. https://doi.org/10.1002/mbo3.70431

Image Credits: AI Generated

DOI: 10.1002/mbo3.70431

Keywords: glioblastoma, gut microbiome, microbiota-gut-brain axis, tumor microenvironment, short-chain fatty acids, kynurenine, aryl hydrocarbon receptor, tumor-associated macrophages, immune checkpoint inhibitors, fecal microbiota transplantation, intratumoral bacteria, HLA peptides