Technical replicates and multiple sequencing approaches reveal weak and inconsistent microbiome signals in tumour and blood samples

Technical Summary: Technical Replicates and Multiple Sequencing Approaches Reveal Weak and Inconsistent Microbiome Signals in Tumour and Blood Samples

Problem Statement
Recent literature has suggested the existence of unique microbial profiles within tumours, blood, and other low-biomass body sites, potentially serving as biomarkers for cancer diagnosis and treatment. However, significant controversy and inconsistency exist regarding these findings. The primary challenge lies in distinguishing true biological signals from technical artifacts in samples with low microbial biomass, where the ratio of microbial to host DNA is low. These samples are highly susceptible to contamination from laboratory reagents, sampling tools, and sequencing environments, leading to false positives and unreliable data. While high-biomass samples (e.g., stool) yield robust microbiome data, the reliability of current high-throughput sequencing methodologies for low-biomass samples (tumour, adjacent tissue, blood) remains unproven and debated.

Methodology
This observational prospective clinical study analyzed samples from 70 patients with non-small cell lung cancer (NSCLC) undergoing curative resection. The study employed a multi-faceted approach to characterize microbial signals across four sample types: lung tumour, adjacent normal lung tissue, blood, and saliva. To rigorously test the reliability of current methodologies, the researchers applied four distinct analytical techniques to the same biological samples where possible:

1. Full-length 16S rRNA gene sequencing: Utilizing PacBio Sequel II or Vega platforms with 27F/1492R primers to generate HiFi reads.
2. Nested short-read 16S amplicon sequencing: A two-step PCR approach targeting the V6-V8 regions, sequenced on Illumina MiSeq, designed to overcome low DNA input.
3. Metagenomic Sequencing (MGS): Shotgun sequencing on Illumina NextSeq2000, followed by rigorous host-read removal (human, PhiX, UniVec) and taxonomic classification using Kraken2 and Bracken, with verification via GeCoCheck.
4. Digital PCR (dPCR): Used for absolute quantification of 16S rRNA gene copies to measure bacterial biomass directly.

The study design included extensive controls: negative controls (surgical instrument swabs), positive controls (ZymoBIOMICS™ Microbial Community Standard and in-house human stool), and technical replicates (duplicate sequencing of the same extracted DNA) to assess reproducibility. Bioinformatic processing followed standard pipelines (QIIME2, DADA2, Deblur) with strict filtering for contaminants and low-abundance ASVs.

Key Results
The study yielded several critical findings regarding the limitations of current microbiome detection in low-biomass samples:

• Failure of Full-Length Sequencing: Full-length 16S sequencing failed to generate usable data for nearly all tumour, adjacent tissue, and blood samples after quality filtering. In contrast, saliva and positive controls retained high read counts and typical oral or stool microbiome profiles.
• Low Biomass and High Variability in Nested Sequencing: While nested PCR successfully generated sequence data for low-biomass samples, these samples contained very few taxa (median 3–5 ASVs) and exhibited substantial variability across technical replicates. The community composition of tumour, blood, and adjacent tissues was indistinguishable from negative controls.
• Lack of Reproducibility: Technical replicates of saliva and positive controls clustered tightly together, demonstrating high reproducibility. Conversely, replicates of tumour, blood, and adjacent tissue samples were highly disparate, with weighted UniFrac distances indicating significant dissimilarity in community composition between duplicates.
• Absence of a Core Microbiome: No consistent core microbiome was identified in tumour, blood, or adjacent tissues. The most abundant genera in these samples (e.g., Cutibacterium, Methylobacterium-Methylorubrum) were also dominant in negative controls, suggesting these signals likely originate from contamination rather than biological presence.
• Discrepancy Between 16S and Metagenomics: Metagenomic sequencing of low-biomass samples resulted in >99.8% of reads being classified as human or vector sequences. The few bacterial genera detected via MGS showed minimal overlap with those detected via nested 16S sequencing. In contrast, saliva samples showed significant overlap between the two methods.
• Absolute Quantification Confirms Low Biomass: dPCR confirmed that bacterial biomass in tumour, adjacent tissue, and blood samples was extremely low (mean 6–39 copies/µL), comparable to negative controls (7 copies/µL), whereas saliva and stool controls showed orders of magnitude higher biomass (mean ~29,000 copies/µL).

Significance and Claims
The paper concludes that current high-throughput sequencing methodologies struggle to detect microbial signatures in low-biomass samples that are distinct from the background noise inherent in collection and processing environments. The study claims that the lack of precision and reproducibility in these samples makes it difficult to identify robust microbial biomarkers for cancer using standard protocols.

The authors emphasize that apparent microbial signals in low-biomass tissues may reflect technical artifacts rather than true biological phenomena. Consequently, the study advocates for a more rigorous approach to low-biomass microbiome research, specifically recommending:

1. The use of technical replicates to assess the stochastic nature of PCR and sequencing in low-DNA samples.
2. The application of alternative sequencing strategies (e.g., comparing amplicon vs. metagenomic data).
3. The inclusion of absolute quantification (dPCR) to flag samples with insufficient biomass before sequencing.
4. The strict use of negative controls to distinguish biological signals from contamination.

The authors modestly state that while their findings challenge the reliability of current tumour and blood microbiome data, they provide essential insights into the specific challenges of assessing these sample types and highlight the necessity of multiple controls and methods to validate any potential cancer biomarkers.