Prognostic value of intratumoral Fusobacterium nucleatum and association with immune-related gene expression in oral squamous cell carcinoma patients

Changes in the oral microbiome, particularly Fusobacterium nucleatum, are associated with oral squamous cell carcinoma (OSCC). F. nucleatum has been reported to modulate local immunity in cancers. We aimed to assess the association between intratumoral F. nucleatum and clinico-pathological features, relapse, and overall survival (OS) in two independent cohorts of patients with OSCC, and to explore the interplay with immune-related genes. We retrospectively analyzed tissue samples from a first cohort of 122 patients with head and neck squamous cell carcinoma, including 61 OSCC (cohort #1), and a second cohort of 90 additional OSCC (cohort #2). We then performed a sensitivity analysis on the merged cohort of OSCC patients (N = 151). F. nucleatum 16S rRNA gene sequences were quantified using real-time quantitative PCR. The presence of gram-negative bacteria and macrophages was confirmed by LPS and CD163 immunostainings, respectively. F. nucleatum positivity was associated with older age, less alcohol and combined alcohol plus tobacco consumption, and less frequent lymph node invasion. There was a trend for a lower recurrence rate in F. nucleatum-positive cases, with less metastatic relapses compared to F. nucleatum-negative tumors, and significantly longer OS, relapse-free and metastasis-free survival. F. nucleatum status was independently associated with OS in multivariate analysis. Immune-related gene and immunohistochemistry analyses showed that gram-negative bacteria load inversely correlated with M2 macrophages. F. nucleatum-associated OSCC has a specific immune microenvironment, is more frequent in older, non-drinking patients, and associated with a favorable prognosis.


Patients and methods
Patients and samples. We retrieved samples from patients with HNSCC who underwent upfront surgery at the Curie Institute between 1990 and 2006. We selected patients with complete clinical, histological, and biological data and long-term follow-up.
We used samples from two independent cohorts: the cohort #1 consisted of 122 patients with HNSCC from various primary sites: oral cavity (OSCC; n = 61), oropharynx (n = 22), hypopharynx (n = 17), and larynx (n = 22). The cohort #2 consisted of 90 OSCC patients. Finally, we grouped the OSCC patients from the first cohort with the second cohort for analysis (merged cohort). The flow chart of patients is displayed in Fig. 1.
The study was conducted in accordance with the ethics principles of the Declaration of Helsinki and GDPR regulations. In accordance with the French regulations, informed consent was obtained from all subjects. This study was approved by the French Ethical Committee (Agreement number D-750602, France) and the ethics committee of the Institut Curie (Agreement number C75-05-18).
Genomic DNA and total RNA extractions. Tumor samples were frozen in liquid nitrogen in a cryotube immediately after surgery and stored at − 80 °C, under temperature control.
A tumor fragment of 5-40 mg and 10-50 mg was used for DNA and RNA extraction, respectively. Tumoral cellularity was evaluated on a 4-5 µm cryosection and samples with less than 50% of tumoral cells were excluded from the study. Nucleic acid extraction was performed on macrodissected tumoral zones, according to the following protocols. Total genomic DNA was extracted from with phenol-chloroform after proteinase K digestion, followed by the precipitation of nucleic acids in ethanol. Total RNA was extracted with miRNeasy Mini kit Qiagen following supplier's recommendations. The quality of RNA was verified by migration on agarose gel. Nucleic acids were quantified using Nanodrop spectrophotometer ND-1000 (ThermoScientific, Wilmington, DE, USA).
In order to rule out external contaminations for F. nucleatum analysis, we included negative controls (buffers/ reagents without tumor samples) and the samples were manipulated under a hood and with mask and gloves.
Immune-related gene expression analysis by real-time quantitative reverse transcription PCR. The immune-related genes were selected in collaboration with immunologists from our group, consistent with our previous publications 21,22 .
Immune-related gene expression levels were normalized on the basis of TBP contents (Genbank accession number NM_003194) 23 . We chose TBP, which encodes the TATA box binding protein, as endogenous control because the prevalence of its transcripts is moderate, and because there are no known TBP retropseudogenes (retropseudogenes lead to co-amplification of contaminating genomic DNA and thus interfere with RT-PCR, www.nature.com/scientificreports/ despite the use of primers in separate exons) 23 . We had previously selected and used the same gene as endogenous control in different cancer types including HNSCC 21,24 . Results, expressed as N-fold differences in target gene expression and termed "Ntarget, " were determined as Ntarget = 2 ΔCtsample , where the ΔCt value of the sample is determined by subtracting the average Ct value of the target gene from the average Ct value of the control gene.
Primers for immune-related genes were designed with the assistance of Oligo 6.0 computer program (National Biosciences, Plymouth, MN). We searched the dbEST and nr databases to confirm the total gene specificity of the nucleotide sequences chosen as primers and the absence of single nucleotide polymorphisms 25 . In particular, the primer pairs were selected to be unique relative to the sequences of closely related family member genes or of the corresponding retropseudogenes 25 . Agarose gel electrophoresis was used to verify the specificity of PCR amplicons. The nucleotide sequences of the oligonucleotide hybridization primers and the average Ct value are shown in Suppl. Table 1. Sixty-four genes mainly involved in the immune process were selected, in particular 19 checkpoint T cell and tumor cell genes, 16 chemokine genes and 18 immune cell population genes. For each primer pair we performed no-template control, no-RT control (RT negative) and RT control with genomic DNA assays, which produced negligible signals that were usually greater than 40 in Ct value, suggesting that primerdimer formation and genomic DNA contamination effects were negligible.
All of the real-time quantitative PCR reactions were performed using the Power SYBR Green master mix (ThermoFisher Scientific, Waltham, MA) using an ABI Prism 7900HT Sequence Detection System (Applied Biosystems, ThermoFisher Scientific, Waltham, MA). The thermal cycling conditions were an initial denaturation step at 95 °C for 10 min, followed by 50 cycles at 95 °C for 15 s and 65 °C for 1 min, and then the melt curve steps. The amplifications specificity was confirmed by melting curve analysis.
Statistical methods. Relationships between F. nucleatum and clinical, biological, and pathological parameters were assessed by using the Chi-square, Chi-square with Yates correction or Fisher tests, as appropriate. Spearman rank correlation non-parametric test was used to determine relationships between F. nucleatum and immune-related genes levels. Bonferroni correction was applied to adjust for multiple tests. Differences were considered significant at confidence levels greater than 95% (p < 0.05).
Survival endpoints were defined according to the DATECAN consensus 27 . OS was determined from the time of initial diagnosis to the time of death, regardless of the cause of death. Relapse-free survival (RFS) was determined from the time of initial diagnosis to the time of relapse (locoregional and/or metastatic) or death, whichever occurred first, regardless of the cause of death. Metastasis-free survival (MFS) was determined from the time of initial diagnosis to the time of metastatic relapse or death, whichever occurred first, regardless of the cause of death. In the absence of event, patients were censored at the date of last follow-up. Survival distributions were estimated by the Kaplan-Meier method, and the significance of differences between survival rates were ascertained with the log-rank test. The multivariate Cox proportional hazards regression model was used to assess the prognostic significance of F. nucleatum and clinical markers on OS; parameters with p values < 0.05 in univariate analysis were entered into the final multivariable Cox regression model, after considering redundancy between variables. The results are presented as hazard ratios (HR) and 95% confidence intervals (95%CIs). www.nature.com/scientificreports/

Results
Association of Fusobacterium nucleatum with clinico-pathological features and overall survival (OS) in the cohort #1. The first cohort (cohort #1) comprised 122 patients with untreated HNSCC from various primary sites, with a majority of tumors arising from the oral cavity (OSCC, n = 61, 50%). Patient characteristics are listed in Suppl. Among the 122 HNSCC samples tested, 35 (28.7%) were scored F. nucleatum-negative and 87 (71.3%) were scored F. nucleatum-positive. Among the 87 F. nucleatum-positive tumors, major differences of Nfn values (determined as described in "Patients and Methods") were observed, ranging from 1.0 to 1613 (Suppl. Fig. 1).
In this overall HNSCC population, there was a non-significant trend toward longer OS in patients with F. nucleatum-positive tumors (HR: 0.64, p = 0.092) ( Fig. 2A). We analyzed the association of F. nucleatum status with OS according to tumor primary site (Fig. 2B,C and Suppl. Fig. 2). We observed that this trend was mainly driven by the oral cavity subgroup (HR: 0.51, p = 0.089) (Fig. 2B), while there was no significant association with OS in patients with HNSCC tumors from other primary sites ( Fig. 2C and Suppl. Fig. 2). Therefore, we decided to focus on the subset of patients with oral cavity tumors for the next steps. In this subgroup of 61 OSCC patients from cohort #1, F. nucleatum positivity (78.7%) was associated with lower pN stage (p = 0.043) and a non-significant trend for more non-drinkers (p = 0.092) compared to F. nucleatum-negative tumors (Suppl. Table 4). Among the clinical, biological and pathological parameters listed above, TP53 mutation (HR: 2.76, p = 0.0093) and UICC stage (HR: 2.79, p = 0.043 for stage II and HR: 2.70, p = 0.049 for stage IV) were significantly associated with OS ( Table 1).

Association of Fusobacterium nucleatum with clinico-pathological features and OS in the cohort #2.
To further explore the association of F. nucleatum with OS in patients with OSCC, we used an independent cohort composed of 90 patients with HNSCC tumors exclusively from this primary site (cohort #2). Patient characteristics and association with F. nucleatum status as displayed in Suppl.  (Table 1). Similarly to the cohort #1, we observed a non-significant trend toward longer OS in the F. nucleatum-positive group (HR: 0.54, p = 0.066) (Fig. 2D).    www.nature.com/scientificreports/

Sensitivity analysis for the association of Fusobacterium nucleatum with OS in the merged cohort.
In order to increase the study power, we performed a sensitivity analysis after grouping the two cohorts of patients with OSCC tumors (merged cohort, n = 151). Patients characteristics are described in Table 2 (Fig. 2E).

Multivariate analysis of OS predictors and stratification according to pT and pN stage. Using
a Cox proportional hazards model, we also assessed the prognostic value for OS of parameters that were significant in univariate analysis, i.e. pT, pN, UICC stage, TP53 mutation (Table 1) and F. nucleatum status (Fig. 2E). As pT/pN and UICC were redundant variables, the model was built with UICC stage, TP53 and F. nucleatum. The prognostic significance of these three parameters was maintained in Cox multivariate regression analysis (Table 3).
In addition, we assessed the prognostic value of F. nucleatum on OS after stratification on pT and pN as main prognostic factors in OSCC. Interestingly, the combinations of pT1/T2 and F. nucleatum positivity (n = 49) and pN0 and F. nucleatum positivity (n = 72) identify a subgroup of patients with a very favorable survival (Fig. 4).

Relationship between Fusobacterium nucleatum load and mRNA expression of immune-related genes.
To explore the underlying mechanisms of the favorable prognosis of F. nucleatum-positive tumors, as several studies recently suggested that F. nucleatum modulates the local immunity of various cancers (particularly, macrophages and Tregs) 18-20 , we tested possible association between F. nucleatum load and expression of various immune-related genes in 115 evaluable F. nucleatum-positive samples from the merged cohort (Suppl. Table 6). We observed a significant negative association between F. nucleatum load and markers of B lymphocytes (CD20, p = 0.027), T helper lymphocytes (CD4, p = 0.013), M2 macrophages (CD163, p = 0.020), and fibroblasts (PDGFRß, p = 0.0067). Toll-like receptor (TLR) 4 (p = 0.020) and OX40 ligand (TNFSF4) (p = 0.0067) expressions were significantly decreased in tumors with high F. nucleatum load. Notably, TNFSF9 receptor (TNFRSF9) expression was decreased (p = 0.0067) while the expression of its ligand (TNFSF9) increased with F. nucleatum load, along with the pro-inflammatory cytokine IL-1ß (p = 0.020). Correlations between F. nucleatum load and TNFSF4, TNFSF9, IL1B, and CD163 expression levels are displayed in Supplementary Fig. 3. There was no association with cell proliferation and APOBEC genes. www.nature.com/scientificreports/ Validation by LPS and CD163 immunostainings. In order to validate the interplay between bacteria and immune cells, particularly macrophages, we performed immunostainings to detect bacteria using an antilipopolysaccharide (LPS) antibody and an anti-CD163 labelling M2 macrophages. We show that, consistent with our PCR results, tumors with high (or low) CD163 and F. nucleatum RNA levels display high (or low) CD163 and LPS expressions, respectively, with an inverse correlation between CD163 and LPS expression levels ( Fig. 5 and Suppl. Table 7). LPS staining was located mainly in the cytoplasm of tumor and macrophage cells and more rarely in the form of extracellular bacterial vesicles (Suppl. Fig. 4).

Discussion
In this work, we assessed the association between intratumoral F. nucleatum and clinico-pathological features and OS, RFS and MFS in two independent cohorts of patients with OSCC, and explored the interplay between F. nucleatum and well-known immune-related genes. Overall, we showed that F. nucleatum identified a subgroup of OSCC, more frequent in older, non-drinking patients, and associated with less frequent lymph node invasion and distant relapse, and favorable OS (independent predictor), RFS and MFS outcomes in the merged cohort. Other independent prognostic indicators included UICC stage and TP53 mutational status, as previously reported 6 .  www.nature.com/scientificreports/    www.nature.com/scientificreports/ Interestingly, the association of low pT or pN stage with F. nucleatum positivity allowed the identification of a patient subgroup with remarkably good prognosis. The prevalence of F. nucleatum positivity in our study was higher than previously described (82% vs. 16% in a recently published meta-analysis), which may be explained by the high sensitivity of our real-time quantitative PCR assay and the selection of OSCC tumors only (in which F. nucleatum are enriched) 28 .
The positive correlation between F. nucleatum and survival was unexpected as this bacteria is usually associated with poor prognosis in cancers, particularly in colorectal cancer [29][30][31][32] . Based on several studies suggesting that F. nucleatum modulates the local immunity of cancers [18][19][20] , we assessed the association between F. nucleatum loads and expression of various immune-related genes to explore the underlying mechanisms of the favorable prognosis of F. nucleatum positive tumors. We showed that tumors with high F. nucleatum loads displayed low RNA levels of M2 macrophages (CD163), CD4 lymphocytes, fibroblasts (PDGFRß), TLR4, OX40 ligand (TNFSF4), and TNFRSF9, and high levels of TNFSF9 and IL-1ß. These results are consistent with our previously published work in which we identified high OX40 ligand and high PDGFRß as factors associated with poor survival 21 . Immunohistochemistry analyses confirmed that gram-negative (LPS-positive) bacteria (including F. nucleatum) load inversely correlated with CD163-positive cells.
Data regarding the effects of F. nucleatum on inflammation are conflicting. In most studies, F. nucleatum infection was shown to expand myeloid-derived immune cells and Tregs and promote M2 polarization of macrophages, inducing pro-tumoral inflammation, which inhibit T-cell proliferation and induce T-cell apoptosis 33,34 . It also directly inhibits cytotoxic T-cell activities via proteins such as Fap2 35,36 . TLRs are a family of receptors involved in the detection of microbial agents to induce activation of inflammatory and antimicrobial innate immune responses 37 . TLR4 is a receptor expressed at the surface of macrophages and tumor cells and has been involved in these pro-inflammatory/immunosuppressive activities of F. nucleatum 38,39 . As an apparent paradox, in our study we observed that high F. nucleatum loads were associated with low levels of TLR4 and M2 macrophages. Interestingly, in a mice model of intestinal inflammation, TLR2/TLR4 knock-out induced increased colonization of F. nucleatum and production of pro-inflammatory cytokines including IL-1ß (as observed in our study) 18 .
On another hand, another work suggested that F. nucleatum enhanced the TNFSF9/IL-1ß signaling inducing M1 macrophage polarization 19 . This is consistent with the positive correlation that we observed between F. nucleatum loads and expression levels of TNFSF9 and IL-1ß cytokines. The lack of effect on M1 polarization could be explained by the concomitant decrease in TNFSF9 receptor.
It is noteworthy that inflammation and M2 infiltrates were found to be associated with poor prognosis in HNSCC [40][41][42] .
Taken together, these data suggest that in OSCC F. nucleatum may be associated with "permissive" tumor microenvironment, insensitive to pro-inflammatory signals, with low TLR4 signaling and low recruitment of M2, resulting in favorable clinical outcomes. Of note, defects in TLR functions have been associated with ageing, which may partially account for the higher proportion of older patients in the F. nucleatum positive group 43,44 .
This work provides a new insight into the prognostic role of intratumoral F. nucleatum in OSCC patients and opens new avenues regarding the biological interplay between this bacteria and OSCC tumor immune microenvironment. Yet our study has some limitations. First, the small sample size did not allow reaching statistical significance in each individual cohort and merging the two cohorts was necessary to obtain sufficient statistical power for OS. In addition, the immune-related gene analysis was based on selected genes, which are not fully specific of each immune cell subtypes. Overall, these results would require further validation in larger prospective cohorts from randomized clinical trials and using more comprehensive methods such as RNA sequencing. Moreover, saliva samples and normal oral tissue were not available for analysis in our study; it could be of interest to assess the correlation between intratumoral F. nucleatum expression and saliva/normal tissue levels in further studies.
In conclusion, we highlight a unique association between F. nucleatum and OSCC patient survival and tumor immune microenvironment. This can give a rationale for further exploration of the role of F. nucleatum in OSCC carcinogenesis and response to treatment, particularly immune therapy.