Tumor Microbiome Sequencing from Host-Rich and Archived Tissue

Multiple studies have reported low-abundance bacterial and fungal signals in human tumor tissues, including evidence for intracellular bacteria and spatially localized fungi in selected tumor types. These observations have opened new research questions about tumor-associated microbial ecology, host–microbe interactions, and potential relationships with tumor biology. However, tumor microbiome sequencing is also one of the most technically challenging areas of low-biomass microbiome research.

Microbial DNA can represent only a very small fraction of the total nucleic acid recovered from a tumor specimen. Host genomic DNA may dominate the library, archived formalin-fixed paraffin-embedded (FFPE) tissue can contain highly fragmented and chemically modified DNA, and microbial contaminants introduced during tissue procurement, sectioning, extraction, library preparation, or bioinformatic classification can approach or exceed the authentic biological signal. For this reason, sequence detection alone should not be interpreted as proof that a microorganism is tumor-resident, intracellular, viable, or mechanistically involved in cancer biology.

This guide provides translational researchers, pathology biobanks, biotechnology teams, and CRO project managers with a framework for designing defensible tumor microbiome studies from fresh-frozen, limited, and archived tissue. It focuses on sample provenance, contamination-aware controls, evidence grading, sequencing method selection, and the specific role of reduced-representation 2bRAD-M when the primary objective is species-level taxonomy from host-rich or degraded material.

Key Takeaways

• Tumor Microbial Signals Require Unusually Strong Quality Control: Low microbial biomass, abundant host DNA, reagent background, and potential read misclassification can all produce apparently convincing microbial signals.
• Tissue Context Matters: Microbial signal strength and biological plausibility can differ substantially among tumor types, anatomical sites, mucosal versus deep tissues, and individual specimens.
• FFPE Is a Feasibility Question, Not an Automatic Exclusion: Archived tumor blocks can support microbiome research, but DNA fragmentation, chemical damage, block history, tissue area, host background, and microbial biomass should be evaluated before method selection.
• Adjacent Tissue Is a Comparator, Not a Negative Control: Normal-adjacent tissue can support within-patient comparison but should not automatically be considered sterile, cancer-free baseline tissue, or a substitute for technical negative controls.
• Sequence Detection Does Not Establish Intracellular Residency: Spatial or orthogonal validation becomes particularly important when the biological claim depends on tissue localization or intracellular microorganisms.
• Sequencing Methods Provide Different Evidence: 16S sequencing, shotgun metagenomics, reduced-representation 2bRAD-M, and spatial assays answer different questions and should not be treated as interchangeable.
• 2bRAD-M Is Primarily a Species-Level Taxonomic Strategy: Published benchmarks support low-input, degraded-DNA, and high-host-background feasibility, but the method does not replace unrestricted gene profiling, strain genomics, MAG reconstruction, or spatial validation.

The Core Technical Challenges of Tumor Microbiome Profiling

Why Is Tumor Tissue Difficult to Analyze?

Nejman et al. reported tumor-type-associated bacterial signals across several solid cancers and used complementary microscopy and staining approaches to support bacterial localization within tumor and immune cells (Nejman et al., 2020). Narunsky-Haziza et al. subsequently described cancer-type-associated fungal ecologies and bacteriome–mycobiome relationships across multiple cancer types (Narunsky-Haziza et al., 2022).

These studies provide important evidence that microbial signals can be detected and spatially localized in selected tumor contexts. At the same time, tumor microbiome analysis remains highly sensitive to methodological artifacts. Four challenges should be considered before sequencing begins:

• Low microbial biomass: Tumor-associated microbial abundance can be very low and varies substantially across tumor types and specimens.
• High host-DNA background: Human genomic DNA can dominate extracted DNA, reducing the effective microbial depth of untargeted sequencing.
• Preservation-related DNA damage: FFPE fixation can produce fragmentation, crosslinking, base damage, and reduced library complexity, with severity varying among blocks.
• Contamination and classification artifacts: Reagent DNA, sectioning equipment, paraffin processing, laboratory background, human-read misclassification, and contaminated microbial references can all generate false microbial assignments.

For broader cancer–microbiome biological context, see our review on sequencing technologies and the relationship between the human microbiome and cancer. This page focuses instead on whether a specific tumor-tissue project can generate technically defensible microbial evidence.

Tumor Microbiome Evidence Is Tissue- and Method-Dependent

Intratumoral microbiome research should not assume that every solid tumor contains the same type or amount of microbial signal. Tumors arising in anatomically exposed or mucosal tissues may have different microbial backgrounds from tumors in deeper organs. Surgical route, tissue necrosis, prior instrumentation, antibiotics, anatomical communication with colonized surfaces, and local tissue physiology can all influence what is detected.

The field has also experienced important methodological debate. Gihawi et al. reanalyzed previously published cancer microbiome sequencing data and argued that human-read misclassification, database problems, and data-processing artifacts could generate false microbial signals in some public cancer datasets (Gihawi et al., 2023). Sepich-Poore et al. subsequently reanalyzed cancer microbiome signals across a broad range of computational choices and reported that multiple signals remained robust under alternative methodological conditions (Sepich-Poore et al., 2024).

These differing conclusions reinforce a practical point rather than requiring a project team to assume one side of the debate: low-biomass cancer sequencing should incorporate stringent host filtering, curated reference databases, process controls, transparent pipeline versioning, and orthogonal confirmation when strong biological claims are being made.

Sample Matrix Feasibility: Fresh-Frozen, FFPE, and Limited Biopsies

No preservation format is universally optimal for every tumor microbiome endpoint. The important variables are recoverable microbial signal, host background, DNA integrity, available tissue quantity, and the genomic information required by the study.

FFPE should not automatically be excluded from microbiome studies, but neither should a published low-input benchmark be converted into a universal FFPE submission requirement. Block age, fixation history, tissue cellularity, tumor area, microbial biomass, sectioning workflow, extraction efficiency, and host background all influence feasibility. For a dedicated discussion of degraded and archived material, see our FFPE and degraded-DNA microbiome method-selection guide.

Pathology and Sample Provenance Matter

A tumor microbiome profile cannot be interpreted independently of pathology metadata. Two specimens assigned the same cancer diagnosis may differ substantially in tumor cellularity, necrosis, stromal content, anatomical location, surgical exposure, mucosal involvement, and preservation history.

Important metadata can include:

• Anatomical site and tumor type: Record the exact tissue origin rather than relying only on a broad cancer label.
• Mucosal or externally exposed status: Tumors directly connected to colonized surfaces may require different contamination and biological interpretation from deep internal tissues.
• Tumor cellularity and necrosis: These factors can affect host-DNA content, DNA quality, and spatial interpretation.
• Biopsy versus resection: Limited specimens may constrain replicate extraction, orthogonal validation, and confirmatory assays.
• Preservation history: FFPE fixation duration, block age, storage, and previous sectioning can influence DNA recovery.
• Pre-analytical exposure: Surgical handling, antibiotics, bowel preparation, instrumentation, or other procedures may affect microbial signals.
• Section selection: Pathology-guided macrodissection or microdissection can help define which tissue compartment enters the sequencing workflow.

These variables should be captured prospectively where possible and included in statistical interpretation rather than treated as incidental metadata.

Contamination Control Across Surgical, Pathology, and Sequencing Workflows

Because authentic microbial signal may be very low, controls should follow the points at which contamination can enter the actual study workflow. There is no universal requirement for a fixed number or type of blanks in every tumor project. For general low-biomass control principles, see our low-biomass microbiome study design guide.

Collection and Procurement Controls

If contamination can be introduced during biopsy, surgery, irrigation, transport, or specimen handling, controls should be designed to represent those specific processes. Their purpose is to identify procedural background rather than to reproduce an arbitrary list of operating-room blanks.

FFPE Sectioning and Paraffin Controls

For archived tissue, sectioning equipment, paraffin processing, exposed block surfaces, and prior handling may contribute microbial DNA. Depending on the workflow, projects can include paraffin-only controls, sectioning blanks, or other instrument-associated controls. Surface trimming may be appropriate in some protocols, but a universal number of sections should not be prescribed across all blocks.

Extraction and Library Controls

Extraction negatives and appropriate library or no-template controls help characterize reagent and processing background. Positive or mock-community controls can provide additional information about library success and taxonomic recovery.

Statistical tools such as decontam can help identify candidate contaminant features using patterns such as higher prevalence in negative controls or concentration-dependent frequency relationships (Davis et al., 2018). Such tools should not be interpreted as automatic subtraction algorithms. Candidate contaminants should be evaluated together with controls, batch structure, taxonomic plausibility, prevalence, read characteristics, and reference quality.

Tumor Tissue, Adjacent Tissue, and Healthy Controls Are Not Interchangeable

Paired normal-adjacent tissue can be highly valuable because it provides a within-patient anatomical comparator. However, it is not a technical negative control and should not automatically be treated as equivalent to healthy tissue from an individual without cancer.

Tumor-adjacent tissue may share local exposures, inflammatory conditions, surgical handling, anatomical microbiota, or molecular alterations with the tumor. Depending on the research question, a study may distinguish among:

• Technical negative controls: Used to characterize contamination introduced by the workflow.
• Paired adjacent tissue: Used to compare tumor with nearby histologically non-tumor tissue from the same individual.
• Distant tissue: Provides a more anatomically separated within-patient comparison when scientifically and ethically available.
• Healthy reference tissue: Can provide a different biological baseline but may be difficult to obtain and may differ in age, exposure, surgery, or other covariates.

The anatomical definition of "adjacent" tissue should be established with the pathology team and reported explicitly. A single universal distance from the tumor margin is not appropriate across different organs and cancer types.

The Tumor Microbiome Evidence Ladder

Different claims require different levels of supporting evidence. A sequence read assigned to a bacterial species does not provide the same evidence as microscopy showing organisms within tumor cells.

1. Level 1: Candidate Sequence Detection — Microbial reads or markers are detected after quality filtering and taxonomic classification. At this stage, contamination and misclassification remain possible explanations.
2. Level 2: Contamination-Aware Detection — Signals are evaluated against procurement controls, sectioning controls, extraction negatives, library controls, host-read filtering, and curated microbial references.
3. Level 3: Quantitative Confirmation — Independent qPCR, ddPCR, or another validated assay confirms that selected microbial DNA signals are reproducibly measurable above technical background.
4. Level 4: Spatial or Orthogonal Confirmation — Methods such as taxon-appropriate FISH, bacterial LPS/LTA staining, fungal stains or probes, culture where feasible, or another independent assay support localization within the tissue.
5. Level 5: Biological Replication and Association — Findings are reproduced in independent samples or cohorts and evaluated against pathology, phenotype, outcome, or mechanistic experiments appropriate to the research hypothesis.

This hierarchy is especially important when a study moves from saying that microbial DNA is "tumor-associated" to claiming that microorganisms are "tumor-resident," "intracellular," or biologically active.

Tumor-Associated Does Not Automatically Mean Tumor-Resident or Intracellular

Sequencing Platform Selection for Tumor and Archived Tissue

Short-Region 16S rRNA Sequencing

16S amplicon sequencing remains a sensitive option for bacterial community screening because targeted amplification reduces the sequencing overhead associated with human genomic DNA. Species-level classification, however, varies according to the selected marker region, taxon, primer design, read length, reference database, and classifier. Non-target amplification can also occur in some assay designs.

16S is therefore useful when broad bacterial community profiling is sufficient, but it does not directly profile fungi, unrestricted microbial genes, or strain-level genomic variation. For general marker-based analysis, see our microbial diversity analysis platform.

Whole-Metagenome Shotgun Sequencing

Shotgun metagenomics can provide direct microbial gene information, broader species profiling, strain-associated variants, and genome-resolved data when sufficient effective microbial coverage is achieved. It should therefore be considered when the study requires information beyond taxonomy.

The limitation in tumor tissue is that host reads can consume a large fraction of the sequencing output. The amount of sequencing required depends on the actual host fraction, microbial biomass, community complexity, and desired endpoint rather than on one universal read-count threshold.

Short-read shotgun sequencing can also be applied to FFPE-derived DNA; it does not universally require high-molecular-weight intact DNA. However, fixation-associated fragmentation and chemical damage can reduce library complexity, usable microbial coverage, and confidence in low-abundance observations. For broader genome-based analysis, explore our metagenomic shotgun sequencing platform and microecology and cancer research solutions.

2bRAD-M Reduced Metagenomics

2bRAD-M uses type IIB restriction enzymes to generate short genomic tags and analyzes a reduced representation of microbial genomes. Sun et al. developed the method for species-resolved microbiome profiling and demonstrated bacterial, archaeal, and fungal profiling using species-specific reference tags (Sun et al., 2022).

Published benchmark experiments included total DNA inputs down to 1 pg, severely fragmented DNA, and separate high-host-background tests. These benchmarks support evaluation of limited or degraded material but should not be interpreted as a guarantee that every FNA, thin FFPE section, or highly host-dominated tumor sample will generate equivalent results.

Jiang et al. later evaluated 2bRAD-M in host-rich samples without prior host depletion, including mock communities containing more than 90% human DNA, saliva samples, and oral cancer specimens (Jiang et al., 2025). In the saliva comparison, 2bRAD-M captured host-specific and temporal community patterns using approximately 5%–10% of the WMS sequencing effort under that study design. Published host-rich benchmarking also suggests that differences in type IIB restriction-site representation between microbial and human genomes contribute to reduced host-sequencing overhead.

These findings support 2bRAD-M analysis for microbiome research as a method worth evaluating when the primary goal is species-level microbial taxonomy from host-rich, limited, or degraded tissue. The oral-cancer evidence should not be generalized into a universal performance claim for every solid tumor or FFPE cohort. Projects requiring unrestricted gene content, strain genomics, mobile elements, MAGs, or mechanistic localization still require additional approaches. For difficult sample planning, see our low-biomass and host-rich microbiome solutions.

What Each Method Cannot Establish

One of the most important design principles in tumor microbiome research is matching the strength of the biological claim to the type of evidence produced.

• 16S sequencing can identify candidate bacterial community patterns but cannot establish fungal ecology, unrestricted microbial gene content, intracellular localization, or viability.
• Shotgun metagenomics provides substantially richer genomic information but sequence detection still does not prove that a microorganism is physically located inside tumor cells.
• 2bRAD-M can provide species-level bacterial, fungal, and archaeal reference-tag profiles but does not provide complete microbial genomes or direct evidence of spatial localization.
• FISH and targeted staining can strengthen spatial evidence but typically interrogate selected organisms or broad microbial groups rather than providing an unbiased complete community profile.
• Statistical association with tumor phenotype can identify candidate research relationships but does not by itself demonstrate mechanism or causality.

A Practical Tumor Microbiome Study Design Workflow

Step 1: Define the Biological Claim

Decide whether the project requires detection of tumor-associated microbial DNA, species-level comparison, direct gene profiling, intracellular localization, viable organisms, or mechanistic evidence. These are different endpoints.

Step 2: Characterize Tissue Provenance

Document tumor type, anatomical site, specimen format, tumor cellularity, necrosis, exposure to colonized surfaces, fixation history, sectioning history, and pre-analytical treatment.

Step 3: Build Controls Around the Real Workflow

Use procurement, sectioning, extraction, and library controls that reflect the actual opportunities for contamination. Adjacent tissue should be analyzed as a biological comparator, not used in place of technical negative controls.

Step 4: Match the Sequencing Strategy to the Required Information

Use marker-based sequencing for selected bacterial community questions, reduced-representation profiling when species-level taxonomy is the main endpoint in difficult samples, and shotgun metagenomics when direct genomic information is required and sufficient effective microbial depth can be obtained.

Step 5: Reserve Material for Orthogonal Validation

If tissue localization, intracellular residence, or viability is central to the hypothesis, preserve sufficient sections or independent material for microscopy, targeted molecular assays, culture, or another relevant validation strategy.

Step 6: Separate Discovery from Strong Biological Conclusions

Low-abundance candidate taxa identified during exploratory sequencing should be evaluated against controls, quantitative confirmation, tissue localization, replication, and biological context before being described as tumor-resident or mechanistically important.

Summary: Tumor Microbiome Studies Require Evidence-Aware Design

Tumor microbiome research combines several of the hardest problems in sequencing-based microbial ecology: low biomass, high host background, variable tissue provenance, archival DNA damage, contamination sensitivity, and potentially strong biological claims.

• Do not assume every tumor contains the same microbial ecology: Evidence can vary substantially among tissues and tumor types.
• Treat FFPE as a sample-specific feasibility problem: Do not rely on universal fragment-size, section-count, or DNA-input thresholds.
• Separate technical negatives from biological comparators: Adjacent tissue, healthy tissue, and extraction blanks answer different questions.
• Use contamination-aware bioinformatics: Host-read filtering, reference curation, negative controls, and pipeline versioning are essential in low-biomass datasets.
• Match claims to evidence: Sequence detection, tumor enrichment, tissue residence, intracellular localization, viability, and mechanism represent progressively stronger conclusions.
• Use 2bRAD-M for the question it is designed to answer: species-level reference-tag profiling from challenging material, rather than complete functional or strain-resolved tumor metagenomics.

The strongest tumor microbiome project is therefore not the one that detects the greatest number of microbial taxa. It is the one in which tissue provenance, controls, sequencing architecture, bioinformatic filtering, and orthogonal validation support the specific biological conclusion being made.

Frequently Asked Questions (FAQ)

References

1. Nejman D, Livyatan I, Fuks G, et al. The human tumor microbiome is composed of tumor type-specific intracellular bacteria. Science. 2020;368(6494):973–980. [DOI: 10.1126/science.aay9189]
2. Narunsky-Haziza L, Canale FP, Cao Y, et al. Pan-cancer analyses reveal cancer-type-specific fungal ecologies and bacteriome interactions. Cell. 2022;185(20):3789–3806.e17. [DOI: 10.1016/j.cell.2022.09.005]
3. Sun Z, Huang S, Zhu P, et al. Species-resolved sequencing of low-biomass or degraded microbiomes using 2bRAD-M. Genome Biology. 2022;23:36. [DOI: 10.1186/s13059-021-02576-9]
4. Walker SP, Tangney M, Claesson MJ. Sequence-Based Characterization of Intratumoral Bacteria—A Guide to Best Practice. Frontiers in Oncology. 2020;10:179. [DOI: 10.3389/fonc.2020.00179]
5. Jiang Y, Liu J, Zhang Y, et al. High-resolution microbiome analysis of host-rich samples using 2bRAD-M without host depletion. npj Biofilms and Microbiomes. 2025;11:223. [DOI: 10.1038/s41522-025-00851-2]
6. Gihawi A, Ge Y, Lu J, et al. Major data analysis errors invalidate cancer microbiome findings. mBio. 2023;14:e01607-23. [DOI: 10.1128/mbio.01607-23]
7. Sepich-Poore GD, McDonald D, Kopylova E, et al. Robustness of cancer microbiome signals over a broad range of methodological variation. Oncogene. 2024;43:1127–1148. [DOI: 10.1038/s41388-024-02974-w]
8. Davis NM, Proctor DM, Holmes SP, Relman DA, Callahan BJ. Simple statistical identification and removal of contaminant sequences in marker-gene and metagenomics data. Microbiome. 2018;6:226. [DOI: 10.1186/s40168-018-0605-2]

For Research Use Only (RUO). Not for use in diagnostic procedures.