Abbreviations

HNSCC

Head and Neck Squamous Cell Carcinoma

RA

relative abundance

OS

Overall Survival

CI

Confidence Interval

DSS

Disease‐Specific Survival

HR

Hazard Ratio

TCMA

The Cancer Microbiome Atlas

TCGA

The Cancer Genome Atlas

PFS

Progression‐Free Survival

HPV

Human Papillomavirus

OSCC

Oral Squamous Cell Carcinoma

MOI

Multiplicity of Infection

LDH

Lactate Dehydrogenase

ROC

Receiver Operating Characteristic

Fnuc

Fusobacterium nucleatum

Fper

Fusobacterium periodonticum

FnucHI

Group of patients where RA of Fnuc is above the cohort median

FnucLO

Group of patients where RA of Fnuc is below the cohort median

FperHI

Group of patients where RA of Fper is above the cohort median

FperLO

Group of patients where RA of Fper is below the cohort median

ATP

Adenosine triphosphate

16S rRNA

16S ribosomal RNA

Head and neck squamous cell carcinoma (HNSCC) is a devastating disease. Despite morbid treatment, 5‐year survival rates remain poor (28%‐67%) [1]. There is a significant knowledge gap regarding how the microbiota may impact HNSCC treatment efficacy [2]. We used microbiome data from two independent cohorts to test and validate the hypothesis that oral bacteria are associated with HNSCC prognosis and in vitro models to investigate mechanistic underpinnings. Methods are detailed inSupplementary Materials.

We first explored associations between the relative abundance (RA) of bacterial genera and overall survival (OS) time in 155 patients with mucosal HNSCC available in the Cancer Microbiome Atlas (TCMA, Supplementary TableS1, Supplementary Text). The distribution of bacterial genera is shown in Supplementary FigureS1. Linear stepwise and Cox regression modeling evaluated associations between these genera and OS/DSS. OnlyFusobacterium detectability was associated with both better OS (hazard ratio [HR] = 0.35, 95% confidence interval [CI] = 0.15‐0.83],P = 0.018, Supplementary FigureS2A) and better disease‐specific survival (DSS; 0.28 [0.15‐0.83],P = 0.031, Supplementary FigureS2B). Kaplan‐Meier survival analysis mirrored these results (Figure 1A‐B). Additionally,Fusobacterium was more abundant in tumors compared to normal tissue (Supplementary FigureS3A‐B), whereas a cognate Gram‐negative oral commensal anaerobe,Prevotella, was not (Supplementary FigureS3C‐D). Receiver operating characteristic (ROC) analysis identified aFusobacterium RA cutoff of 0.016 (specificity: 92.7%; sensitivity: 28.8%). Patients with RA above the threshold had better OS and DSS (Supplementary FigureS4).

Next, we questioned whether any particularFusobacterium species were associated with survival. Patients were stratified into groups with detectable and undetectable species (Supplementary FigureS5). In Cox regression, onlyFusobacterium nucleatum detectability was significantly associated with OS (HR: 0.43 [95% CI: 0.19‐0.97],p = 0.042; Supplementary FigureS6). Kaplan‐Meier modeling showed thatF. nucleatum detectability was associated with improved OS (P <0.001, Supplementary FigureS7A), with a trend for improved DSS (P = 0.096, Supplementary FigureS7B).

In multivariate Cox modeling with established predictors of survival (disease stage, smoking and Human Papilloma Virus [HPV] status), bothFusobacterium andF. nucleatum detectability were strongly associated with OS (P < 0.001 for both, Supplementary FiguresS8A/S9A) and DSS (P < 0.001 andP = 0.015 for each respectively, Supplementary FiguresS8B/S9B).

To test the validity of these results, we evaluated whether the abundance ofFusobacterium was also predictive of treatment efficacy in the separate MicroLearner cohort (n = 175; described inSupplementary Text and Supplementary TableS2) by dividing it into patient groups withFusobacterrium RA either below (FusoLO) or above (FusoHI) the cohort median, as the commensal nature ofFusobacterium in the oral cavity makes detectability nearly universal in saliva [3]. We used progression‐free survival (PFS) as an endpoint because, with a median follow‐up of 33.6 months (range 4‐57 months), very few deaths (n = 6, 3.4%) had occurred. FusoHI patients had a trend for better PFS (P = 0.054; Figure 1C). Only 10 events (15.6%) of progression were observed in patients with HPV‐positive oropharyngeal cancer, so we conducted a separate analysis including all patients except these (“HPVneg cohort”;n = 111, 29.7% event rate), where FusoHI patients had significantly better PFS (P = 0.011, Figure 1D). ROC analysis identified a salivaryFusobacterium RA cutoff of 2.760 in this cohort (specificity: 65.2%; sensitivity: 55.8%). Patients with RA above this threshold had better PFS (Supplementary FigureS10).F. nucleatum (Fnuc) andF. periodonticum (Fper) were the most abundant fusobacterial species. There was a non‐significant trend for better PFS in FnucHI (F. nucleatum RA > medianF. nucleatum RA) patients, but FperHI patients had significantly better PFS (P = 0.021; Supplementary FigureS11).

Given our clinical observations, we reasoned thatFusobacterium may contribute to HNSCC killing. We initially explored the effect ofF. nucleatum on oral SCC (OSCC) evaluated with an ATP‐based viability assay. TR146 cells were infected withF. nucleatum at multiplicity of infection (MOI) ranging from 0.5 to 5. With increasing MOI, a more significant reduction in OSCC cell viability was observed (Supplementary FigureS12A). Separate experiments using lactate dehydrogenase (LDH) activity and crystal violet assays validated these findings (Supplementary FigureS12B‐C). We also tested whether theF. nucleatum medium caused any OSCC death if added without any previous contact with bacteria and confirmed that it did not (Supplementary FigureS12D). A significant decrease in viability was observed from 24h post‐infection (Supplementary FigureS13).

To test whether the observed effects ofF. nucleatum on OSCC cytotoxicity were strain‐specific, cell‐line specific and not a general characteristic of oral commensal anaerobes, we co‐cultured multiple cell lines of OSCC (TR146, HN5 and HSC‐3) with either of twoF. nucleatum strains orPrevotella oralis (MOI = 100) and evaluated their effect on OSCC viability (Figure 1E), validated with a crystal violet assay (Supplementary FigureS14).P. oralis, likeF. nucleatum, is an oral commensal Gram‐negative anaerobe.P. oralis infection did not impact OSCC viability, while bothF. nucleatum strains caused a reduction in OSCC viability. We next questioned whether otherFusobacterium species caused OSCC killing. We tested the effect ofF. periodonticum on OSCC cultures at MOI 100 and found that it caused OSCC killing similarly toF. nucleatum (Figure 1F). At lower MOI (0.5‐5), OSCC killing was also overall similar between the two species and rose with MOI (Supplementary FigureS15). These results suggest that otherFusobacterium species which are phylogenetically close toF. nucleatum, but not all oral commensal Gram‐negative anaerobes can cause OSCC killing.

We next asked whether OSCC killing was mediated by a surface protein or by secreted compounds/metabolites (Figure 1G). Firstly, OSCC cells were infected withF. nucleatum, which was either alive or heat‐inactivated (inFnuc), and OSCC viability was assessed. We also tested whether the supernatant ofF. nucleatum culture was sufficient to cause OSCC death.F. nucleatum supernatant caused OSCC killing, whereas fresh medium did not. Co‐culture ofF. nucleatum washed in fresh broth significantly attenuated OSCC killing compared to growth broth, suggesting continued production of supernatant in co‐culture. inFnuc caused OSCC killing only when added to co‐culture with growth broth but not with fresh broth. Separately, we used transwell inserts to prevent direct contact ofF. nucleatum with OSCC while allowing for any secreted molecules to move freely between them (Supplementary FigureS16). Significant cell killing was observed in transwell replicates, more substantially whenF. nucleatum was in direct contact with OSCC, which may be attributable to higher local concentrations in direct contact co‐culture compared to transwell replicates. Taken together, these results indicate thatF. nucleatum mediates OSCC killing primarily via the bacterial secretome.

Although colorectal cancer studies indicate thatF. nucleatum contributes to oncoprogression and treatment resistance, these bacteria are not common constituents of the normal intestinal microbiota, whereas they are common components of the normal oral microbiota [4]. Previous studies often assume that a higher tumoral abundance ofFusobacterium, which we also detected, indicates its oncogenic role [5]. However, our findings suggest that its presence may enhance HNSCC treatment efficacy. Limitations of this study are discussed in theSupplementary Text.

In summary, our preliminary research suggests thatFusobacterium actively determines survival outcomes in HNSCC. Ongoing research will validate its role as a predictive biomarker in HNSCC and dissect the mechanism by which fusobacteria cause HNSCC killing.