Identification of consensus head and neck cancer-associated microbiota signatures: a meta-analysis of 16S rRNA and The Cancer Microbiome Atlas datasets

35

Multiple reports have attempted to describe the tumour microbiota in head and 36 neck cancer. However, these have failed to produce a consistent microbiota signature which 37 may undermine understanding the importance of bacterial-mediated effects in head and neck 38 cancer. The aim of this study is to consolidate these datasets and identify a consensus 39 microbiota signature in head and neck cancer. 40

We analysed 11 published head and neck cancer 16S ribosomal RNA microbial 41 datasets collected from cancer, cancer-adjacent and non-cancer tissue to generate a consensus 42 microbiota signature. These signatures were then validated using The Cancer Microbiome 43 Atlas database. 44

We identified unique bacteria enrichment within tissue types and correlated it with 45 possible functional and clinical outcomes. 46

Our meta-analysis demonstrates a consensus microbiota signature for head and 47 neck cancer, highlighting its potential importance in this disease. 48 49

Tumour Microbiota, Head and Neck Cancer, 16s rRNA Sequencing, Meta-50 Analysis 51 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 3 52 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 4 1. Introduction 53 Recent studies have revealed that cancers previously thought to be sterile can contain unique 54 microbial communities. The extent of microbial infiltration varies across different cancer 55 types, with head and neck cancers (HNSC) containing one of the highest level of intratumoral 56 microbial infiltrates while glioblastomas having the least amount of microbes. 1-3 This 57 “intratumoral microbiota” can refer to bacterial infiltrates found in the extracellular matrix or 58 within the cellular components of the tumour such as cancer, immune and stromal cells. 2 It is 59 now widely appreciated that intratumoral bacteria can have direct and indirect effects on 60 tumours or the tumour microenvironment (TME). 4-6 The presence of specific intratumoral 61 bacteria has been reported to influence multiple features of tumour biology including 62 treatment efficacy, local immune composition and activity and promoting tumour 63 metastasis.7-10 64 65 Direct interaction between specific bacterial species with the tumour and the TME can induce 66 chemoresistance, promote tumour progression, enhance therapeutic responses and modulate 67 anti-tumour immunity through various mechanisms. 11-14 Bacteria can metabolise an active 68 drug into its inactive form or induce autophagy in cancer cells which can promote 69 chemoresistance.12-14 Moreover, specific bacterial species can mount or suppress anti-tumour 70 responses.15-17 Most notably, Fusobacterium nucleatum colocalises with cancer and immune 71 cells by binding to cell surface receptors such as Toll-like receptor 4 (TLR-4), T-cell 72 immunoreceptor with Ig and ITIM domains (TIGIT) and Carcinoembryonic Antigen-Related 73 Cell Adhesion Molecule 1 (CEACAM-1) receptors, or sugar groups (e.g. tumour expressed 74 Galactose-N-acetylgalactosamine), which may then promote chemoresistance and suppress 75 anti-tumour immunity13, 15, 18-22 . Alternatively, Bifidobacterium species enhance anti-tumour 76 immunity and efficacy of PD-1 immunotherapy responses.8, 23, 24 77 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 5 78 The release of bacterial metabolites such as short chain fatty acids (SCFA), amino acids, 79 vitamins and bile acids can indirectly affect the tumour and the TME. 25, 26 Butyrate, a SCFA 80 released by anaerobic bacteria through fermentation of carbohydrates, can decrease tumour 81 cell growth and invasion, while increasing CD8 + T cell-mediated anti-tumour responses.27-30. 82 However, butyrate has also been shown to have pro-tumorigenic effects by inducing 83 senescence-associated inflammatory phenotypes and inhibiting natural killer cell functions.31, 84 32 Bacteria-derived indole and its derivatives (i.e. indole-3-lactic acid) have been shown to 85 suppress anti-tumour immunity by activating immunosuppressive tumour-associated 86 macrophages in treatment-naïve pancreatic cancer, while improving chemotherapeutic and 87 immune-checkpoint inhibitor efficacy in pancreatic cancer and melanoma. 33-35 Together, 88 these studies demonstrate that the tumour microbiota can influence cancer clinical outcomes 89 in a context-dependent manner. 90 91 There are multiple reports describing the microbiota in HNSC. 36-89 Most of these studies 92 compared the microbiota diversity and bacterial relative abundance between cancer and 93 healthy samples using 16S ribosomal RNA (rRNA) sequencing 36-85, 88, 89 , while two studies 94 additionally correlated the impact of the microbiota with matched transcriptome analysis.86, 90 95 Samples studied include tissues, swabs, and oral fluids (saliva or oral rinse) from cancer and 96 healthy patients. Specifically for HNSC tissue microbiota analysis, samples included cancer, 97 cancer-adjacent (approximately > 5 mm away from the tumour), contralateral, and healthy 98 donor tissue samples. 36-55, 57-60, 85-89 Most bacteria identified in HNSC are oral commensal 99 bacteria from the genera Streptococcus, Rothia, Fusobacterium, Haemophilus and 100 Prevotella.36-38 However, changes in microbial composition have been identified when cancer 101 samples are compared to healthy controls. In general, there was an enrichment in 102 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 6 Fusobacterium within cancer tissue samples, that correlated with an inflammatory 103 phenotype.36, 37, 47 However, inconsistencies are observed for microbes such as Streptococcus, 104 Actinomyces and Prevotella warranting the need to identify a consensus microbiota signature 105 for HNSC.37, 38, 43, 54, 85 106 107 In this study, we systematically reviewed the literature and performed a meta-analysis to 108 consolidate the currently heterogenous HNSC-associated microbiota data. Selected 16s rRNA 109 sequencing datasets were analysed consistently to minimise variability between different 110 sample cohorts and adjusted for batch-effects.91 These consensus HNSC-associated microbial 111 signatures were then validated using whole genome sequencing (WGS) data from The Cancer 112 Microbiome Atlas (TCMA). 1 Finally, we correlated the presence of different microbiota 113 signatures with the HNSC tumour microenvironment and clinical outcomes. 114 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 7 2. Methods 115 This study was performed according to the Preferred Reporting Items for Systematic Reviews 116 and Meta-Analyses (PRISMA) Statement.92 117 2.1 Search and Study Selection 118 The following criteria were used to select datasets: 1) Tissue samples, 2) Presence of 119 metadata to distinguish sample types, 3) Illumina short-read amplicon sequencing of 16S 120 rRNA V3 to V5 primers (Figure 1). Database search was performed on 16 August 2022 and 121 datasets after this date were not included (Supplementary Table 1). The risk of biasness 122 assessment was conducted using RoB 2 (β v9) (Supplementary Table 1). 123 124 125 126 127 Figure 1: Study selection flow chart. 128 129 2.2 Download, pre-processing, and analysis of 16S rRNA datasets 130 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 8 Previously published raw sequences were retrieved from the National Center for 131 Biotechnology Information (NCBI) Sequence Read Archive (SRA) using pysradb.93 Samples 132 were divided into three main groups – cancer, cancer-adjacent and non-cancer tissues. Cancer 133 tissues are defined as tissues obtained directly from the tumour, while cancer-adjacent tissues 134 are cancer-free regions obtained > 5mm away from cancer tissues. Non-cancer tissues are 135 defined as tissues that were either obtained from healthy patients or contralateral tissues 136 obtained from cancer patients. FASTQ sequences files were obtained from SRA using 137 sratoolkit.94 These sequences were processed using QIIME2 DADA2 denoise-paired and 138 reads truncated using the same parameters (trim_left_f = 30, trim_left_r = 30, trunc_q = 15). 139 Sequences from different studies were merged before bacterial Operational Taxonomic Units 140 (OTU) classification using QIIME2 and SILVA reference database (version silva-138-99-nb-141 classifier).95 142 143 Raw microbial reads were filtered, central log-ratio (CLR) transformed and batch-adjusted 144 using Phyloseq and MixOmics as described previously. 96-98 Microbiome datasets are 145 inherently compositional, hence, CLR transformation addresses generates scale-invariant 146 values which allows datasets to remain unaffected by variations in library sizes among 147 samples.99 Briefly, low abundance of OTUs were filtered through proportional counts of all 148 samples (< 1%) and minimum counts per sample (< 10). Bacterial OTUs were agglomerated 149 at the genus level before transforming into CLR for their compositional nature.96, 98 The CLR-150 abundance was used for subsequent statistical and discriminant analysis. A total of 903 SRA 151 samples from 11 projects were downloaded (Table 1). 152 153 154 155 156 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 9 157 Table 1: Study accession and sample size post-filtering 158 159 160 161 162 163 164 165 166 167 2.3 Discriminant analysis of 16S rRNA dataset 168 To discriminate the microbial signature between sample types, we employed both 169 multivariate and univariate discriminant analysis. For β -diversity analysis, CLR-abundance of 170 all genera were ordinated using Euclidean distance and plotted on a principal component 171 analysis (PCA) using mixOmics R package. β -diversity for each sample were calculated as 172 distance to centroid for each tissue groups using betadisper (vegan v2.6-4). Group and 173 pairwise permutest (vegan v2.6-4, permutations = 9999) was performed to determine if 174 dispersions differed between sample types, while group and pairwise permutational 175 multivariate analysis of variance (PERMANOVA) was performed using adonis2 (vegan v2.6-176 4, method = “euclidean”, permutation = 9999) and pairwise.adonis2 (pairwiseAdonis, method 177 = “euclidean”, permutation = 9999) to determine statistical differences in β -diversity between 178 groups. Other statistical test such as Analysis of similarities (ANOSIM) (vegan v2.6-4, 179 distance = “euclidean”, permutation = 9999) and Fifty-fifty multivariate analysis of variance 180 (FFMANOVA) (nSim = 9999) were also applied as supplementary to distinguish between 181 sample types.100, 101 182 Accession number Sample size Primers Cancer Cancer- adjacent Non-cancer PRJNA412445 16 0 0 V4 -V5 PRJNA555458 0 0 4 V3 -V4 PRJNA596113 102 53 0 V3 -V4 PRJNA597251 19 20 0 V3 -V4 PRJNA666746 50 50 0 V3 -V4 PRJNA666891 7 0 10 V4 PRJNA685226 13 13 0 V3 -V4 PRJNA699728 37 0 201 V4 PRJNA803155 40 0 0 V4 -V5 PRJNA822685 75 79 0 V3 -V4 PRJNA866676 37 36 41 V3 -V4 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 10 Multivariate sparse partial linear discriminant analysis (sPLS-DA) was applied on batch-183 adjusted dataset to identify discriminating genera within each sample type.96 The Area Under 184 Curve (AUC) of the Receiver Operating Characteristics (ROC) curve was calculated using 185 mixOmics in Rstudio. 96, 98 The AUC value served as a quantification of the discriminatory 186 potential between sample types. A higher AUC value, closer to 1, signified a test approaching 187 perfection in its ability to distinguish between the samples. Heatmap of all representative 188 bacteria in each sPLS-DA was presented with sample type clustered according to Euclidean 189 distance and Ward’s linkage. 190 191 Univariate Kruskal-Wallis test with Bonferroni multiple comparisons test was also performed 192 to determine microbial genera differences between sample types using microbiomeMarker in 193 RStudio v3.3.0, followed by a post-hoc Wilcoxon test (Mann-Whitney test) with Bonferroni-194 Dunn multiple comparison test to determine differences between groups (cancer– cancer-195 adjacent, cancer – non-cancer, non-cancer – cancer-adjacent). Additionally, Wilcoxon 196 matched-pairs signed rank test with Bonferroni-Dunn multiple comparison test was also 197 performed on paired cancer and cancer-adjacent samples. 198 199 2.4 Functional profiling analysis of 16S rRNA datasets in different sample types 200 To predict the microbial functions of genera detected from 16S rRNA sequencing between 201 each tissue sample type, Phylogenetic Investigation of Communities by Reconstruction of 202 Unobserved States 2 (PICRUST2) from QIIME2 was applied on raw 16S rRNA reads using 203 MetaCyc database. 102, 103 Functional abundance was processed and analysed similarly as 204 described for raw microbial reads. Univariate Kruskal-Wallis test and post-hoc Wilcoxon test 205 was performed as previously described to compare differences between groups. 206 207 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 11 2.5 Reanalysis of tissue microbiome data from TCMA 208 Decontaminated microbial read count derived from The Cancer Genome Atlas (TCGA) 209 HNSC whole genome sequences were obtained from TCMA repository.1 Data from a total of 210 177 cancer (TCGA annotation: primary tumour) and 22 cancer-adjacent (TCGA annotation: 211 solid tumour normal) tissues were obtained from TCMA repository (n = 22 paired cancer and 212 cancer-adjacent samples). Similar to 16S rRNA pre-processing, read counts were 213 agglomerated to the genus before CLR transformation as described in 2.2. As samples were 214 already pre-processed in the TCMA dataset, no further filtering or batch adjustment was 215 required. Microbiome statistical analysis were performed similarly as 16S sequencing 216 datasets. Metadata were obtained from cBioPortal for Cancer Genomics.104 217 218 2.6 Microbiome correlation analysis with tumour microenvironment and survival 219 analysis 220 The TME immune subtype and 29 functional gene expression signatures (FGES) scores were 221 previously described by Bagaev et al. (2021) using transcriptomics datasets from TCGA. 105 222 The 29 FGES represents the major functional components and immune, stromal, and other 223 cellular populations of the tumour. 105 Pearson’s correlation test was applied to determine the 224 correlation between FGES scores and selected bacteria genera. The four TME immune 225 subtypes were – Desert (D), Fibrotic (F), Immune-enriched (IE), Immune-enriched/Fibrotic 226 (IE/F) (Described in Supplementary Table 2). 105 Specifically, tissues with IE and IE/F 227 phenotype contains high T-cell infiltration, while D and F phenotypes have low T-cell 228 infiltration (Supplementary Table 2).105 Using a cut-off of high (top 35 th percentile) and low 229 (bottom 35th percentile) CLR-abundance, the proportion of each patient within the four TME 230 subtypes were determined, and survival analysis was performed. Since there were 153 231 TCGA-HNSC samples with both FGES/TME subtypes and microbiome datasets, these 232 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 12 samples were used for subsequent correlation and survival analysis. Chi-squared (χ 2) test was 233 performed in Prism9 to determine association between high/low bacterial genera CLR-234 abundance and proportion of patients within each tumour subtype. 235 236 237 238 2.7 Statistical analysis 239 For comparisons made between all unpaired tissue groups, Kruskal-Wallis test with 240 Bonferroni’s multiple comparison was used for comparisons made between all tissue groups 241 unless stated otherwise. Post hoc Wilcoxon matched pairs signed rank test with Bonferroni’s 242 multiple comparison was used to compare differences between unpaired tissue samples. For 243 all paired cancer and cancer-adjacent samples, Wilcoxon matched-pairs signed rank test was 244 performed. Univariate and multivariate Cox proportional hazard model was performed using 245 survminer in Rstudio v3.3.0. Statistical analysis was performed using RStudio v3.3.0 and 246 Prism9. 247 248 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 13 3 Results 249 3.1 Multivariate analysis identifies homogenous microbial abundance and functions between 250 cancer and cancer-adjacent samples, contrasting to non-cancer samples. 251 The 16S rRNA amplicon datasets were obtained for 903 head and neck tissue types ( 396 252 cancer, 251 cancer-adjacent, and 256 non-cancer) from 11 studies. 37, 38, 41-45, 87-89 Following 253 sample processing and aggregation of 16S data at the genus level, a total of 177 distinct 254 bacterial genera were identified. Differences in the microbiota and β -diversity between tissue 255 types were assessed using PCA and PERMANOVA test (Figure 2A-2B). The β -diversity 256 index was calculated for cancer (14.6 ± 5.7) and cancer-adjacent (15.0 ± 5.5) tissues, 257 revealing similar levels of β -diversity. In contrast, non-cancer tissues (8.61 ± 5.3) exhibited 258 lower β -diversity (PERMANOVA – Overall R2 = 0.006, p < 0.0001) (Figure 2B). Post-hoc 259 pairwise test identified significant differences in β -diversity between cancer and non-cancer 260 (R2 = 0.003, p = 0.002), cancer and cancer-adjacent (R 2 = 0.005, p < 0.001), and non-cancer 261 and cancer-adjacent samples (R 2 = 0.007, p < 0.001) (Figure 2B). These findings were 262 consistent with additional multivariate and univariate statistical analysis, ANOSIM (R = 263 0.027, p = 0.002) and FFMANOVA (p < 0.0001) (Supplementary Table 3). 264 265 Multivariate sparse partial least squares discriminant analysis (sPLS-DA) identified 116 266 representative bacterial genera in sPLS-DA component 1 and 2 which were discriminant 267 between tissue types (Figure 2C-E). The AUC values were computed for different sample 268 comparisons: cancer versus others (AUC = 0.74, p < 0.05), non-cancer versus others (AUC = 269 0.91, p < 0.05), and cancer-adjacent versus others (AUC = 0.84, p < 0.05). These results 270 demonstrate that sPLS-DA components 1 and 2 (Figure 2D) can effectively differentiate 271 between tissue types. Lastly, majority of cancer and cancer-adjacent samples clustered 272 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 14 together and were distinct from non-cancer samples, as determined by Euclidean distance 273 metric (Figure 2E). 274 275 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 15 Figure 2: Multivariate discriminant analysis (sPLS-DA and PERMANOVA) of tissue 276 16S rRNA microbiota to discriminant between cancer, cancer-adjacent and non-cancer 277 tissues. (A) Principal coordinates analysis (PCA) plot of tissue CLR-abundance microbiota 278 based on Euclidean distance. (B) Dispersion of β -diversity (top-right panel) for each sample 279 type, with error bar representing 95% confidence interval. PERMANOVA test was 280 performed with bacterial genera as variable for sample types. (C) sPLS-DA sample plot of 281 16S rRNA tissue microbiota. Ellipse displays 95% confidence interval for each sample group. 282 The batch-adjusted normalized abundance of tissue microbiota from 16S amplicon 283 sequencing was compared between cancer, cancer-adjacent and non-cancer tissue samples. 284 sPLS-DA identified 116 bacterial genera on component 1 and 2. (D) ROC curve and AUC 285 values determined from sPLS-DA analysis was used to access discriminatory potential of 286 sPLS-DA component 1 and 2. (E) Heatmap representing 86 bacterial genera after sPLS-DA 287 discriminant analysis. Each column and row represent a unique sample and bacterial genera 288 respectively, with OTUs clustered based on Euclidean distance and Ward linkage method. 289 290 3.2 Univariate analysis identifies differences in microbial abundance and functions 291 between sample types. 292 Next, unpaired univariate analysis was applied to determine the differences between tissue 293 types. Out of the 177 bacterial genera, 33 were identified as significantly different among 294 tissue types using Kruskal-Wallis test (Padjust < 0.05) (Supplementary Table 4). Notably, 18 of 295 these were also identified as representative bacterial genera in sPLS-DA discriminant 296 analysis (Supplementary Table 4). These 33 genera are denoted as bacterial genera of interest 297 (Supplementary Table 4). The top 20 differentially abundant genera, based on the effect size 298 (η 2), are presented in Figure 3. Post-hoc unpaired Wilcoxon test with Bonferroni-Dunn’s 299 multiple comparison test was performed on these genera to determine the mean differences in 300 the central log ratio transform (CLR) abundance between tissue types (Figure 3A, 301 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 16 Supplementary Table 5). Since most published studies compared cancer to non-cancer, or 302 cancer to cancer-adjacent tissues, we performed post-hoc test for these comparisons (Figure 303 3B). We identified 27 out of 33 genera as significantly different (P adjust (#) < 0.05) between 304 cancer and non-cancer tissues (Figure 3A-3B, Supplementary Table 5). Non-cancer tissues 305 contained more Fretibacterium (CLR-abundance diff. = 1.42, SE = 0.12), Stenotrophomonas 306 (CLR-abundance diff. = 0.80, SE = 0.12) and Tannerella (CLR-abundance diff. = 0.71, SE = 307 0.10), while cancer tissue had a greater CLR-abundance of Neisseria (CLR-abundance diff. = 308 2.32, SE = 0.15), Capnocytophaga (CLR-abundance diff. = 2.02, SE = 0.15), and 309 Streptococcus (CLR-abundance diff. = 1.98, SE = 0.19) (Figure 3A-3B). Capnocytophaga 310 abundance in cancer tissues was consistent to previous findings 46, 57, 85 , while contradicting 311 findings were identified for the abundance for Streptococcus38, 41, 52, 57, 85 and 312 Fusobacterium38, 41, 42, 52, 57, 58. 313 314 For cancer and cancer-adjacent tissue, 13 out of 33 bacterial genera were significantly 315 different (post-hoc unpaired Wilcoxon test P adjust (*) < 0.05) (Figure 3A and 3C, 316 Supplementary Table 5). Similar to many studies, Fusobacterium (CLR-abundance diff. = 317 1.11, SE = 0.20) displayed significantly higher CLR-abundance in cancer tissue than cancer-318 adjacent tissue, while Rothia (CLR-abundance diff. = 0.92, SE = 0.18), Stenotrophomonas 319 (CLR-abundance diff. = 1.33, SE = 0.15) and Serratia (CLR-abundance diff. = 0.70, SE = 320 0.12) had higher CLR-abundances in cancer-adjacent tissue than cancer tissue (Figure 3A). 36, 321 37, 43, 45, 52, 55, 58, 59 . Additionally, we found that Prevotella was elevated in cancer tissue as 322 compared to cancer-adjacent tissues.43, 45, 52, 55, 58 Unlike previous studies, we did not observe 323 any significant differences in Streptococcus abundance between cancer and cancer-adjacent 324 tissues.36, 37, 45, 51, 52, 55, 59 325 326 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 17 Lastly, 28 of the 33 top bacterial genera were significantly different (post-hoc unpaired 327 Wilcoxon test P adjust < 0.05) when comparing non-cancer to cancer-adjacent tissue samples 328 (Figure 3A, Supplementary Table 5). Genera Neisseria (CLR-abundance diff. = 2.83, SE = 329 0.19), Rothia (CLR-abundance diff. = 1.95, SE = 0.16) and Streptococcus (CLR-abundance 330 diff. = 1.95, SE = 0.16) were higher in CLR-abundance in cancer-adjacent, while 331 Fusobacterium (CLR-abundance diff. = 1.80, SE = 0.18) and Prevotella (CLR-abundance 332 diff. = 1.28, SE = 0.20 were greater in CLR-abundance in non-cancer tissue (Figure 3A). 333 334 To provide functional insights to microbial abundance between cancer tissues and other tissue 335 types, we applied Picrust2 to predict possible differences in MetaCyc pathway functional 336 CLR-abundance.102 After filtering low abundant functional pathways, we identified a total of 337 365 MetaCyc pathways. Using Kruskal-Wallis test, 162 MetaCyc pathways were identified 338 as significantly different among sample types (P adjust < 0.05) (Supplementary Table 5). Post-339 hoc analysis identified 129/162 and 7/162 pathways that were significantly different between 340 cancer – non-cancer, and cancer – cancer-adjacent tissues comparisons respectively 341 (Supplementary Table 6). 342 343 Cancer tissues, when compared to non-cancer tissues, were enriched in pathways involving 344 the synthesis of ubiquinol, L-methionine, inosine-5’-phosphate and cysteine and metabolic 345 pathways such as TCA cycle and pentose phosphate pathway, while non-cancer tissues were 346 enriched in the degradation of L-lysine, L-glutamine, N-Acetylglucosamine (GlcNac), N-347 acetylmannosamine (ManNac), and N-acetylneuraminate (Figure 3C). Cancer tissues were 348 more similar to cancer-adjacent tissues, albeit enrichment was identified in pathways 349 involving biosynthesis of ppGpp (guanosine pentaphosphate and tetraphosphate), cis-350 vaccenate, L-asparatate, L-asparagine, cob(II)yrinate a,c-diamide and CMP-legionaminate, 351 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 18 and enrichment in pathways involving degradation of pyruvate and L-lysine, when compared 352 to cancer-adjacent tissues (Figure 3C). 353 354 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 19 Figure 3. Comparison of bacterial CLR-abundance and functional prediction between 355 sample types. (A) Top 20 bacterial genera (based on effect size) in CLR-normalized 356 abundances between sample groups using Kruskal-Wallis test with Bonferroni’s multiple 357 comparison. 33 out of 177 genera were identified as significantly different (P adjust < 0.05) 358 using Kruskal-Wallis test. Post-hoc Wilcoxon test with Bonferroni-Dunn’s multiple 359 comparison was performed to identify group-wise differences between Cancer – Non-cancer 360 (#), Cancer – Cancer-adjacent (*), Non-cancer – Cancer-adjacent (^). Post-hoc unpaired 361 Wilcoxon test with Bonferroni-Dunn’s multiple comparison for (B) bacterial genera and (C) 362 functional CLR-abundance for Cancer – Non-cancer (Top panel), and Cancer – Cancer-363 adjacent (Bottom panel). 364 365 3.3 Paired cancer and cancer-adjacent tissues display similar bacterial abundance 366 differences using multiple sequencing techniques. 367 To understand microbial abundance differences between cancer tissue and cancer-adjacent 368 tissue within the same patients, we performed Wilcoxon matched-pairs signed rank test to 369 identify changes in microbial diversity and abundance within paired tissue samples in the 16S 370 rRNA datasets. Similar to unpaired data analysis, no significant differences in microbial β -371 diversity was identified between the patient’s paired cancer and cancer-adjacent tissues 372 (Supplementary Figure 2). 373 However, 76 bacterial genera were significantly different between paired tissue samples 374 (Figure 4A, Supplementary Table 7). Bacterial genera with the greatest differences in CLR-375 abundance were then identified by using a cut-off of > 0.4 and < -0.4 (Figure 3A). Using this 376 cut-off, we found that Fusobacterium, Prevotella, Alloprevotella, Catonella, Selenomonas 377 and Treponema were elevated in cancer tissue vs cancer-adjacent tissue, while 378 Stenotrophomonas, Rothia, Granulicatella, Serratia, Anoxybacillus, Actinomyces and 379 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 20 Bacteroides were greater in cancer-adjacent tissue compared to cancer tissue (Figure 4A-4B). 380 Similarly, nine of these bacteria were also found to be significantly different in unpaired 381 tissue analysis (Supplementary Table 4 and 7). Contrary to published studies on unpaired 382 samples, Streptococcus, an abundant oral commensal, was not significantly different in our 383 paired sample analysis.36, 37, 45, 51, 52, 55, 59 384 To validate this finding, we probed the publicly available TCMA dataset, a repository 385 containing microbiota reads derived from WGS of tissue samples. 1 Similar to the 16S rRNA 386 dataset, we observed that cancer tissues from TCMA displayed significantly (p < 0.05) higher 387 CLR-abundance for genera Fusobacterium, Selenomonas and Treponema, while Rothia and 388 Actinomyces were elevated (p < 0.05) in cancer-adjacent tissues (Figure 4D). In the TCMA 389 dataset, Anoxybacillus, Serratia, and Stenotrophomonas were not present due to pre-analysis 390 filtering, while no significant differences in CLR-abundance were observed for Prevotella, 391 Catonella, Alloprevotella, and Bacteroides (Figure 4C). Notably, similar trend in CLR-392 abundance between cancer and cancer-adjacent samples was still observed for Prevotella, 393 Catonella, and Alloprevotella in TCMA dataset. Overall, 16S rRNA and TCMA WGS 394 dataset showed similar trend for most bacteria genera, regardless of sequencing techniques. 395 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 21 396 Figure 4: Comparison of tissue microbiota in paired cancer and cancer-adjacent tissue 397 samples using different sequencing datasets. (A) Paired Wilcoxon matched-pairs signed 398 rank test on paired 16S rRNA sequencing cancer and cancer-adjacent tissue samples. 76 399 bacteria were significantly different in sample groups (p < 0.05) using paired Wilcoxon 400 matched-pairs signed rank test and 13 bacteria genera were identified as top bacteria with 401 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 22 differential CLR-abundance (Diff. CLR-abundance > 0.4 or < -0.4). Blue and red dot points 402 represent bacteria that were higher in abundance in cancer and cancer-adjacent tissues 403 respectively. CLR-abundance of paired cancer and cancer-adjacent samples from (B) 16s 404 rRNA sequencing and (C) TCMA WGS sequencing datasets. Wilcoxon matched pairs signed 405 rank test was performed for both 16s rRNA (n = 287) and TCMA (n = 22) datasets. *p < 406 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. 407 408 3.4 Tissue microbiota diversity correlates with cancer functional gene expression 409 signatures. 410 Since Fusobacterium, Selenomonas, Treponema, Actinomyces, and Rothia displayed 411 significant differences between paired cancer and cancer-adjacent tissues, we performed 412 correlation analyses to investigate the possible relationship between these genera and the 413 tumour transcriptional profile and patient clinical features found in matched TCGA patients 414 (n = 156). 105 Here, TCGA transcriptomic data were classified into 29 functional gene 415 expression signatures (FGES), which represent major functional components and 416 characteristics of cancer cell populations. 105 These 29 FGES can then be used to further 417 classify cancers into four major immune subtypes (Desert, Fibrotic, Immune-enriched/non-418 fibrotic, and Immune-enriched/fibrotic). 105 We correlated the CLR-abundance of 419 Fusobacterium, Selenomonas, Treponema, Actinomyces, and Rothia from TCGA-HNSC 420 patients with their respective FGES scores and immune subtype. 421 422 We first correlated CLR-abundance with the FGE signatures. The CLR-abundance of 423 Fusobacterium correlated (r > 0.3, p < 0.0001) with FGES related to angiogenesis, 424 neutrophils and granulocyte traffic (Figure 5A). Other FGES such as matrix remodelling, 425 protumour cytokines, MDSC traffic, M1 signature, antitumour cytokine, MHCI and EMT 426 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 23 signatures also positively correlated (p < 0.05) to CLR-abundance of Fusobacterium (Figure 427 5A). The CLR-abundance of Selenomonas showed a positive correlation (p < 0.05) to 428 angiogenesis, neutrophil signature, granulocyte traffic and antitumour cytokines signatures, 429 while negatively correlating (p < 0.05) to B cells (Figure 4A). Lastly, CLR-abundance of 430 Treponema displayed a negative correlation (p < 0.05) to endothelium, T reg traffic, T reg, 431 MHCII, Coactivation molecules, B cells, NK cells, Effector cells and T cells, while positively 432 correlating to (p < 0.05) neutrophils and granulocyte traffic (Figure 5A). 433 434 Next, we investigated how CLR-abundance correlated to tissue immune subtyping. Cancer 435 tissues classified as immune deserts (D) and fibrotic (F) which lack immune cell enrichment 436 correlated with higher Fusobacterium and Treponema CLR-abundance. On the other hand, 437 cancer tissues that are immune-enriched / non-fibrotic (IE) or immune-enriched / fibrotic 438 (IE/F) correlated with greater Rothia. No significant correlation in immune subtypes were 439 observed for Selenomonas and Actinomyces (Figure 5B). To identify the differences in 440 immune subtypes between high and low CLR-abundance of each bacterial genera, we further 441 segregated patients based on the upper and lower 35% CLR-abundance quartiles. As 442 expected, patients with IE and IE/F tumour subtypes showed significant association with low 443 CLR-abundance of Fusobacterium (chi-square test, p = 0.04). While not reaching statistical 444 significance, more patients with IE and IE/F tumour subtypes have low CLR-abundance of 445 Selenomonas (chi-square test, p = 0.33) and Treponema (chi-square test, p = 0.11), opposite 446 to high CLR-abundance for Rothia (Figure 5C). Conversely, patients with D and F subtypes 447 had higher CLR-abundance of Fusobacterium , Selenomonas or Treponema (Figure 5C). 448 Lastly, the proportion of patients in each immune subtype were similar in high and low CLR-449 abundance Actinomyces groups. Taken together, these show that Fusobacterium, 450 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 24 Selenomonas or Treponema are associated with poor T-cell infiltration compared to Rothia 451 which may have implications in selecting patients suitable for immunotherapy. 452 453 Figure 5: Correlation analysis of Fusobacterium, Selenomonas, Treponema, Rothia 454 and Actinomyces to the tumour transcriptional profiles. 455 (A) 29 functional gene expression (FGES) signature scores derived from Bagaev et al 456 (2021) were used to correlated with CLR-abundance of genera Fusobacterium, 457 Selenomonas, Treponema, Actinomyces, and Rothia, using Pearson’s correlation 458 method. Asterisk (*) represents significant correlation (p < 0.05), and red and blue 459 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 25 scales represents positive and negative correlation respectively. (B) The CLR-460 abundance of each bacterial genera within each tumour microenvironment immune 461 subtype (D – Desert, F – Fibrotic, IE – Immune-enriched/Non-fibrotic, IE/F – Immune-462 enriched/Fibrotic). Kruskal-Wallis test with uncorrected Dunn’s test was performed to 463 compare CLR-abundance in all immune groups. *p < 0.05, **p < 0.01. (C) The 464 proportion of patients in each tumour immune subtype with high and low CLR-465 abundance in each bacterial genera. High and low bacteria CLR-abundance groups 466 were determined by upper and lower 35% quartiles respectively. Chi-squared test was 467 performed to determine association between high/low bacterial genera CLR-abundance 468 and proportion of patients in each tumour subtype. 469 470 3.5 Evaluation of microbiota abundance with clinical features and survival 471 Univariate and multivariate Cox proportional hazard models were used to investigate the 472 association between the intratumoral microbiota and clinical features. Univariate Cox 473 proportional hazard model identified that current smokers (HR 2.235, 95% CI 1.146 – 4.359, 474 p = 0.018), HPV-negative (HR 2.273, 95% CI 1.158 – 4.459, p = 0.017), and low CLR-475 abundance of Fusobacterium (HR 0.8883, 95% CI 0.8183 – 0.9642, p = 0.005) were risk 476 factors for reduced overall survival (Table 2). Further multivariate Cox proportional hazard 477 models identified that HPV-negative (HR 2.853, 95% CI 1.1991 – 6.7882, p = 0.0178) and 478 low CLR-abundance of Fusobacterium (Continuous: HR 0.8482, 95% CI 0.7758– 0.9273, p 479 = 0.0003; Low: HR 2.579, 95% CI 1.3687 – 4.860, p = 0.0034) were independent hazards for 480 overall survival, but not current smokers (Table 2). 481 482 483 484 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 26 Table 2: Univariate and multivariable Cox proportional hazard models for overall survival 485 Univariate Multivariable n HR (95% CI) p-value HR (95% CI) p-value Age (years) < 65 ≥ 65 106 47 0.9832 (0.594 – 1.626) 0.947 Sex Female Male 41 112 0.868 (0.525 – 1.436) 0.581 Staging I II III IV 4 30 31 87 2.331 (0.302 – 17.97) 2.275 (0.2950 – 17.54) 3.275 (0.4492 – 23.87) 0.417 0.430 0.242 HPV status Positive Negative 37 107 2.273 (1.158 – 4.459) 0.017* 2.853 (1.1991 – 6.7882) 0.0178* Smoking Non-smoker Current Previous 37 43 71 2.235 (1.146 – 4.359) 1.488 (0.7804 – 2.838) 0.018* 0.227 1.3788 (0.5383 – 3.5317) 0.7821 (0.3130 – 1.9545) 0.50329 0.59894 Fusobacterium Continuous High Low 153 53 53 0.8883 (0.8183 - 0.9642) 2.0592 (1.17 – 3.625) 0.005** 0.0123* 0.8482 (0.7758– 0.9273) 2.579 (1.3687 – 4.860) 0.0003** 0.0034* Selenomonas Continuous High Low 53 52 0.9712 (0.8714 –1.082) 1.205 (0.7094 – 2.048) 0.597 0.49 Treponema Continuous High Low 153 53 53 0.9467 (0.8768 – 0.719) 1.432 (0.8092 – 2.535) 0.162 0.217 Rothia Continuous High Low 153 54 54 1.029 (0.8936 – 1.184) 0.6552 (0.3585 – 1.198) 0.694 0.17 Actinomyces Continuous High Low 153 53 52 0.9652 (0.8512 – 1.094) 1.006 (0.5679 – 1.783) 0.58 0.983 486 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 27 4 Discussion: 487 Several studies have investigated the microbial signature in HNSC using different sequencing 488 approaches and sample types, such as tissues, swabs, and oral fluids. However, these studies 489 have reported inconsistent findings regarding the presence of specific bacterial genera. 490 Consequently, a consensus microbial signature for head and neck tissues has yet to be 491 established. In this study, we aimed to address this gap by conducting a meta-analysis of 11 492 studies and presenting a consensus tissue microbiota signature for head and neck tissues. We 493 analyzed 16S rRNA sequencing datasets from 903 tissue samples, including 396 cancer 494 tissues, 251 cancer-adjacent tissues, and 256 non-cancer tissues. Our analysis revealed 495 significant differences in the abundance of 33 bacterial genera among the various tissue 496 types. Specifically, we observed that cancer tissues and cancer-adjacent tissues exhibited 497 greater similarity to each other compared to non-cancer tissues. These findings suggest 498 distinct microbial profiles in cancer and cancer-adjacent tissues compared to non-cancer 499 tissues. Non-cancer tissues exhibited the lowest differences in β -diversity and contained 500 elevated levels of bacterial genera such as Tannerella, Fretibacterium, Stenotrophomonas, 501 Fusobacterium, and Prevotella (Figure 6A). While cancer and cancer-adjacent tissues 502 displayed similar microbiota based on β -diversity indexes, further analysis using paired and 503 unpaired univariate methods enabled differentiation of these tissues at the genera level 504 (Figure 6B). Importantly, these abundance signatures were validated using additional data 505 from TCMA. Matching TCMA samples with transcriptomic data derived from TCGA) and 506 clinical features provided insights into the contributions of individual genera in HNSC. 507 Notably, we found that a high abundance of Fusobacterium was associated with better overall 508 survival in HNSC patients Overall, our study contributes to the establishment of a consensus 509 tissue microbiota signature for HNSC, shedding light on the distinct microbial profiles in 510 different tissue types and their potential implications for clinical outcomes. 511 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 512 Figure 6: Summary of bacteria genera within cancer, cancer-adjacent and non-cancer 513 tissue samples. (A) Elevated microbiota within non-cancer tissues compared to cancer and 514 cancer-adjacent tissues. (B) Elevated bacteria genera between cancer and cancer-adjacent 515 tissues. 516 517 Both multivariate and univariate discriminant analyses was able to differentiate different 518 tissue sample types based on microbial abundance. As previously reported, cancer and 519 cancer-adjacent tissues were more similar in microbial diversity when compared to non-520 cancer tissues. 36-38, 42, 43, 45, 49, 51, 52, 55, 57-59, 85 At the genus level, both paired and unpaired 521 abundance analysis of cancer and cancer-adjacent tissues showed consistent enrichment for 522 Fusobacterium and Rothia in cancer tissues. 36, 37, 45, 52, 55, 58, 59 In contrast, Prevotella was 523 enriched within cancer tissues compared to cancer-adjacent tissues, and no differences were 524 observed for Streptococcus.36-38, 41, 43, 45, 51, 52, 55, 57-59, 85 525 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 29 526 Previous studies have reported conflicting result where Fusobacterium was more in cancer 527 tissues as compared to non-cancer and cancer-adjacent tissues. 38, 42, 52, 57, 58 However, we 528 found that Fusobacterium was most abundant in non-cancer tissues. Fusobacterium is an 529 abundant commensal bacteria found largely in the oral cavity (buccal, hard palate, gingiva, 530 tonsils, tongue) and saliva of healthy individuals, suggesting a potential role within the 531 healthy oral microbiota. 106-108 In vitro experiments in HNSC cell lines showed that 532 Fusobacterium nucleatum infection promotes cancer cell invasion, proliferation, autophagy, 533 and PD-L1 expression. 109-113 It is unknown whether there are strain and species level 534 differences found in Fusobacterium isolated in cancer and non-cancer tissues to explain such 535 seemingly contradictory findings. Additionally, non-cancer tissue from cancer patients may 536 also have different tissue microbiota profiles from healthy donor tissues which is currently 537 unavailable for this study. Also, most of the experiments showing an oncogenic role for F. 538 nucleatum were carried out in vitro and thus did not consider a potential mitigating role of the 539 immune system. Moreover, the abundance of F. nucleatum both in absolute terms and 540 relative to other bacteria present in the tumour microbiota might influence the oncogenic 541 potential of F. nucleatum . Further experiments are required to evaluate the role of 542 Fusobacterium in HNSC. 543 544 We observed that Streptococcus, another highly abundant oral commensal genera 106-108, was 545 increased specifically in cancer and cancer-adjacent tissue when compared to non-cancer 546 tissues. However, there was no significant difference in Streptococcus abundance between 547 cancer and cancer-adjacent tissues. Within the oral cavity, certain pathogenic Streptococcus 548 species, like S. mutans, can contribute to periodontitis by acidifying the environment. 114 In 549 oral cancer, S. mutans has been shown to promote tumour proliferation and invasion, 550 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 30 potentially through upregulation of IL-6 in infected cells115. On the other hand, Streptococcus 551 species from the mitis ( S. oralis, S parasanguinis, S.mitis ) and sanguinis ( S. sanguinis, S. 552 gordonii) groups, can break down lactic acid or pyruvate into hydrogen peroxide, thereby 553 antagonising pathogenic species such as S. mutans.114 In oral cancer, S. mitis, S. salivarius, S. 554 anginosus were found to display anti-tumour effects, including reducing cancer cell viability 555 and promoting CD8 + cytotoxic T cell responses 116-120. These findings indicate that the 556 abundance of specific Streptococcus species may contribute to pathogenesis, disease severity, 557 or exert anti-tumour effects. It is important to note that these studies underscore the 558

of identifying microbiota at the genus level using short-read 16S rRNA 559 sequencing. To address these limitations, recent advances in sequencing technologies such as 560 long-read 16S rRNA amplicon sequencing (e.g., PacBio, Nanopore) or shotgun 561 metagenomics can be employed to reveal species- or strain-specific diversity within the 562 microbiota.121-123 Such advancements can provide a more comprehensive understanding of 563 the specific species and strains that play a role in oral cancer pathogenesis and anti-tumour 564 effects. 565 566 To compare the metabolic potential of different head and neck tissue types, a functional 567 prediction analysis was performed using PICRUSt2 on the 16S rRNA sequencing data. The 568 analysis revealed an enrichment of several amino acids and metabolites, including L-569 aspartate, L-asparagine, acetate, butanoate, and lactate, in cancer tissues compared to non-570 cancer and cancer-adjacent tissues. L-aspartate and L-asparagine, as substrates for nucleotide 571 biosynthesis and regulators of amino acid homeostasis and anabolic metabolism, have been 572 reported to promote tumour proliferation. 124-126 Butanoate, acetate, and lactate can serve as 573 energy sources for cells by converting into acetyl-CoA, which can then be utilized in the 574 tricarboxylic acid (TCA) cycle to produce ATP. 127-130 The role of butanoate in tumorigenesis 575 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 31 depends on the specific tumour and the TME, as it can exhibit tumour-promoting or 576 suppressive properties. 31, 131-133 Lactate, a well-studied metabolite produced by both cancer 577 cells and bacteria, can modulate the TME by inactivating natural killer cells, promoting 578 polarisation of M2-like tumour-associated macrophages, and stimulating the growth of T-579 regulatory cells.134 Collectively, these findings suggest that bacteria infiltrating HNSC tissues 580 possess functional capacities that may promote cancer progression. Further validation studies 581 are warranted to better understand the role of these metabolic pathways in HNSC and the 582 contribution of bacteria in shaping the TME. 583 584 We further explored the relationship between the abundance of the five cancer-associated 585 bacterial genera, Fusobacterium, Selenomonas, Treponema, Actinomyces, and Rothia, and the 586 TME phenotype and clinical outcomes. Fusobacterium was associated with a lack of T-cell 587 immune infiltration in HNSC, similar to colorectal and oesophageal cancers. 21, 135-137 588 Furthermore, Fusobacterium can chemoattract neutrophils via release of SCFA and can also 589 modulate neutrophils and endothelial cell functions in vitro.138-142 Interestingly, we observed 590 that patients with low levels of Fusobacterium within the tumour tissue had shorter overall 591 survival, consistent to previous reports in HNSC. 45, 60, 143 In contrast, opposite findings have 592 been reported for colorectal, gastric and oesophageal cancers, suggesting that Fusobacterium 593 may have a different role in HNSC. 144-147 We also found that Treponema correlated with an 594 lack of immune infiltration in HNSC. Although the effect of Treponema infiltration in HNSC 595 is still unknown, these bacteria have been associated with an upregulation of immune 596 suppressive cells and can suppress innate immune responses. 148-150 In our analysis, Rothia 597 was found to correlate with an immune-enriched TME. Limited information is available 598 regarding the role of Rothia in cancer; however, Rothia dentocariosa has been shown to 599 induce Toll-like receptor 2 (TLR-2) mediated TNF-alpha inflammatory response in human 600 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 32 embryonic kidney cells and THP-1 monocytes. 151 Selenomonas and Actinomyces did not 601 significantly correlate with TME subtypes in our analysis. However, Selenomonas sputigena 602 infected gingival epithelial cells can promote neutrophil and monocyte recruitment. 152 603 Actinomyces has been associated with young-onset colorectal cancers, showing a preferential 604 localisation with cancer-associated fibroblasts in the TME. 153 These findings underscore the 605 importance of validating and understanding the underlying mechanisms through which these 606 bacteria can modulate the tumour microenvironment in HNSC. 607 608 This study represents the first comprehensive comparison of 16S rRNA (V3-V5) microbial 609 sequencing across multiple studies to identify consensus HNSC-associated microbiota 610 signatures in cancer, cancer-adjacent and non-cancer tissues. To ensure consistency, a 611 uniform bioinformatics approach was employed. However, it is important to acknowledge the 612 inherent limitations of this study. Variations in sample collection, preparation, and 613 sequencing among different laboratories introduce batch effects that could contribute to the 614 inconsistencies observed across different reports. To mitigate these effects, we applied 615 PLSDA-batch adjustment to the pooled datasets. 91 Conventional short-read 16S rRNA 616 sequencing provides information only up to the genus level, which restricts the ability to 617 identify specific bacterial species or strains that may be relevant to disease outcomes. 121 618 Overcoming this limitation would require advanced sequencing technologies such as long-619 read 16S rRNA amplicon sequencing or shotgun metagenomics to reveal species- or strain-620 level diversity within the microbiota. Furthermore, the availability of complete clinical 621 metadata in published datasets reporting 16S rRNA sequencing is limited, restricting our 622 ability to make comprehensive clinical associations. Therefore, our clinical associations were 623 primarily based on TCMA/TCGA datasets. Despite these limitations, this study confirms 624 distinct differences in the microbiota composition among cancer, cancer-adjacent and non-625 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 33 cancer HNSC tissue samples. The strength of our study lies in the meta-analysis of a 626 substantial number of samples, totalling 903. Additionally, our analysis indicates that a high 627 load of Fusobacterium within HNSC tissues may be associated with a favourable survival 628 outcome. The correlation analysis of the microbiota with functional predictions, functional 629 gene enrichment signature, and immune subtyping of the tumour and TME provides novel 630 avenues for further exploration. 631 632 In conclusion, our study establishes a consensus microbial signature for head and neck 633 tissues, shedding light on the distinct microbial profiles present in head and neck cancer 634 (HNSC). These findings have the potential to serve as targets for future treatment approaches 635 in HNSC. Nevertheless, it is crucial to acknowledge the limitations identified in our study 636 and recognize the need for further research to address these limitations. Additional 637 investigations are required to gain a deeper understanding of the functional implications of 638 the identified microbiota differences in HNSC. By addressing these gaps, we can advance our 639 knowledge and pave the way for more effective therapeutic interventions in HNSC. 640 641 5. Conflict of Interest 642 Authors state no conflict of interest. 643 644 6. Author Contributions 645 Conceptualisation K.Y.,K.F.; methodology, investigation, and data analysis, K.Y.,R.L.,F.W., 646 E.S., G.B., and L.M.; resources A.P.,P.W. and S.V.; writing - original draft preparation, K.Y., 647 E.S., S.V., and K.F; writing-review and editing, K.Y., G.B., E.S., A.P., P.W., R.V., S.V., and 648 K.F.; supervision R.V.,A.P., S.V., and K.F.; funding acquisition, A.P.,P.W., and S.V.; All 649 authors have read and agreed to the published version of the manuscript. 650 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 34 7. Acknowledgments 651 This work is supported by an NHMRC investigator grant APP1196832 to P.W., a The 652 Garnett Passe and Rodney Williams Senior Fellowship to S.V., and The University of 653 Adelaide Postgraduate Research Scholarship to K.Y., R.L., F.W and L.M. Illustration in 654 Figure 6 was generated using Biorender. 655 656 8. Data Availability Statement 657 The data used to support the findings of this study are included within the article and within 658 supplementary material. 659 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 35

660 1. Dohlman AB, Arguijo Mendoza D, Ding S, et al. The cancer microbiome atlas: a pan-661 cancer comparative analysis to distinguish tissue-resident microbiota from contaminants. Cell 662 Host Microbe. Feb 10 2021;29(2):281-298.e5. doi:10.1016/j.chom.2020.12.001 663 2. Nejman D, Livyatan I, Fuks G, et al. The human tumor microbiome is composed of 664 tumor type-specific intracellular bacteria. Science. May 29 2020;368(6494):973-980. 665 doi:10.1126/science.aay9189 666 3. Xuan C, Shamonki JM, Chung A, et al. Microbial dysbiosis is associated with human 667 breast cancer. PLoS One. 2014;9(1):e83744. doi:10.1371/journal.pone.0083744 668 4. Jain T, Sharma P, Are AC, Vickers SM, Dudeja V. New Insights Into the Cancer-669 Microbiome-Immune Axis: Decrypting a Decade of Discoveries. Front Immunol. 670 2021;12:622064. doi:10.3389/fimmu.2021.622064 671 5. Xavier JB, Young VB, Skufca J, et al. The Cancer Microbiome: Distinguishing Direct 672 and Indirect Effects Requires a Systemic View. Trends Cancer. Mar 2020;6(3):192-204. 673 doi:10.1016/j.trecan.2020.01.004 674 6. Yang L, Li A, Wang Y, Zhang Y. Intratumoral microbiota: roles in cancer initiation, 675 development and therapeutic efficacy. Signal Transduct Target Ther. Jan 16 2023;8(1):35. 676 doi:10.1038/s41392-022-01304-4 677 7. Vétizou M, Pitt JM, Daillère R, et al. Anticancer immunotherapy by CTLA-4 678 blockade relies on the gut microbiota. Science. Nov 27 2015;350(6264):1079-84. 679 doi:10.1126/science.aad1329 680 8. Sivan A, Corrales L, Hubert N, et al. Commensal Bifidobacterium promotes 681 antitumor immunity and facilitates anti-PD-L1 efficacy. Science. Nov 27 682 2015;350(6264):1084-9. doi:10.1126/science.aac4255 683 9. Iida N, Dzutsev A, Stewart CA, et al. Commensal bacteria control cancer response to 684 therapy by modulating the tumor microenvironment. Science. Nov 22 2013;342(6161):967-685 70. doi:10.1126/science.1240527 686 10. Viaud S, Saccheri F, Mignot G, et al. The intestinal microbiota modulates the 687 anticancer immune effects of cyclophosphamide. Science. Nov 22 2013;342(6161):971-6. 688 doi:10.1126/science.1240537 689 11. Fu A, Yao B, Dong T, et al. Tumor-resident intracellular microbiota promotes 690 metastatic colonization in breast cancer. Cell. Apr 14 2022;185(8):1356-1372.e26. 691 doi:10.1016/j.cell.2022.02.027 692 12. Geller LT, Barzily-Rokni M, Danino T, et al. Potential role of intratumor bacteria in 693 mediating tumor resistance to the chemotherapeutic drug gemcitabine. Science. Sep 15 694 2017;357(6356):1156-1160. doi:10.1126/science.aah5043 695 13. Yu T, Guo F, Yu Y, et al. Fusobacterium nucleatum Promotes Chemoresistance to 696 Colorectal Cancer by Modulating Autophagy. Cell. Jul 27 2017;170(3):548-563.e16. 697 d oi:10.1016/j.cell.2017.07.008 698 14. Spanogiannopoulos P, Kyaw TS, Guthrie BGH, et al. Host and gut bacteria share 699 metabolic pathways for anti-cancer drug metabolism. Nat Microbiol. Oct 2022;7(10):1605-700 1620. doi:10.1038/s41564-022-01226-5 701 15. Gur C, Ibrahim Y, Isaacson B, et al. Binding of the Fap2 protein of Fusobacterium 702 nucleatum to human inhibitory receptor TIGIT protects tumors from immune cell attack. 703 Immunity. Feb 17 2015;42(2):344-355. doi:10.1016/j.immuni.2015.01.010 704 16. Kalaora S, Nagler A, Nejman D, et al. Identification of bacteria-derived HLA-bound 705 peptides in melanoma. Nature. Apr 2021;592(7852):138-143. doi:10.1038/s41586-021-706 03368-8 707 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 36 17. Naghavian R, Faigle W, Oldrati P, et al. Microbial peptides activate tumour-708 infiltrating lymphocytes in glioblastoma. Nature. May 2023;617(7962):807-817. 709 doi:10.1038/s41586-023-06081-w 710 18. Abed J, Emgård JE, Zamir G, et al. Fap2 Mediates Fusobacterium nucleatum 711 Colorectal Adenocarcinoma Enrichment by Binding to Tumor-Expressed Gal-GalNAc. Cell 712 Host Microbe. Aug 10 2016;20(2):215-25. doi:10.1016/j.chom.2016.07.006 713 19. Parhi L, Alon-Maimon T, Sol A, et al. Breast cancer colonization by Fusobacterium 714 nucleatum accelerates tumor growth and metastatic progression. Nat Commun. Jun 26 715 2020;11(1):3259. doi:10.1038/s41467-020-16967-2 716 20. Zhang S, Yang Y, Weng W, et al. Fusobacterium nucleatum promotes 717 chemoresistance to 5-fluorouracil by upregulation of BIRC3 expression in colorectal cancer. 718 J Exp Clin Cancer Res. Jan 10 2019;38(1):14. doi:10.1186/s13046-018-0985-y 719 21. Wu J, Li Q, Fu X. Fusobacterium nucleatum Contributes to the Carcinogenesis of 720 Colorectal Cancer by Inducing Inflammation and Suppressing Host Immunity. Transl Oncol. 721 Jun 2019;12(6):846-851. doi:10.1016/j.tranon.2019.03.003 722 22. Gur C, Maalouf N, Shhadeh A, et al. Fusobacterium nucleatum supresses anti-tumor 723 immunity by activating CEACAM1. Oncoimmunology. 2019;8(6):e1581531. 724 doi:10.1080/2162402x.2019.1581531 725 23. Yoon Y, Kim G, Jeon BN, Fang S, Park H. Bifidobacterium Strain-Specific Enhances 726 the Efficacy of Cancer Therapeutics in Tumor-Bearing Mice. Cancers (Basel). Feb 25 727 2021;13(5)doi:10.3390/cancers13050957 728 24. Asadollahi P, Ghanavati R, Rohani M, Razavi S, Esghaei M, Talebi M. Anti-cancer 729 effects of Bifidobacterium species in colon cancer cells and a mouse model of carcinogenesis. 730 PLoS One. 2020;15(5):e0232930. doi:10.1371/journal.pone.0232930 731 25. Rossi T, Vergara D, Fanini F, Maffia M, Bravaccini S, Pirini F. Microbiota-Derived 732 Metabolites in Tumor Progression and Metastasis. Int J Mol Sci. Aug 12 733 2020;21(16)doi:10.3390/ijms21165786 734 26. Krautkramer KA, Fan J, Bäckhed F. Gut microbial metabolites as multi-kingdom 735 intermediates. Nat Rev Microbiol. Feb 2021;19(2):77-94. doi:10.1038/s41579-020-0438-4 736 27. Bachem A, Makhlouf C, Binger KJ, et al. Microbiota-Derived Short-Chain Fatty 737 Acids Promote the Memory Potential of Antigen-Activated CD8(+) T Cells. Immunity. Aug 738 20 2019;51(2):285-297.e5. doi:10.1016/j.immuni.2019.06.002 739 28. Wang W, Fang D, Zhang H, et al. Sodium Butyrate Selectively Kills Cancer Cells and 740 Inhibits Migration in Colorectal Cancer by Targeting Thioredoxin-1. Onco Targets Ther. 741 2020;13:4691-4704. doi:10.2147/ott.S235575 742 29. Liang Y, Rao Z, Du D, Wang Y, Fang T. Butyrate prevents the migration and 743 invasion, and aerobic glycolysis in gastric cancer via inhibiting Wnt/β -catenin/c-Myc 744 signaling. Drug Dev Res. May 2023;84(3):532-541. doi:10.1002/ddr.22043 745 30. He Y, Fu L, Li Y, et al. Gut microbial metabolites facilitate anticancer therapy 746 efficacy by modulating cytotoxic CD8(+) T cell immunity. Cell Metab. May 4 747 2021;33(5):988-1000.e7. doi:10.1016/j.cmet.2021.03.002 748 31. Okumura S, Konishi Y, Narukawa M, et al. Gut bacteria identified in colorectal 749 cancer patients promote tumourigenesis via butyrate secretion. Nat Commun. Sep 28 750 2021;12(1):5674. doi:10.1038/s41467-021-25965-x 751 32. Zaiatz-Bittencourt V, Jones F, Tosetto M, et al. Butyrate limits human natural killer 752 cell effector function. Sci Rep. Feb 15 2023;13(1):2715. doi:10.1038/s41598-023-29731-5 753 33. Bender MJ, McPherson AC, Phelps CM, et al. Dietary tryptophan metabolite released 754 by intratumoral Lactobacillus reuteri facilitates immune checkpoint inhibitor treatment. Cell. 755 Apr 27 2023;186(9):1846-1862.e26. doi:10.1016/j.cell.2023.03.011 756 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 37 34. Tintelnot J, Xu Y, Lesker TR, et al. Microbiota-derived 3-IAA influences 757 chemotherapy efficacy in pancreatic cancer. Nature. Mar 2023;615(7950):168-174. 758 doi:10.1038/s41586-023-05728-y 759 35. Hezaveh K, Shinde RS, Klötgen A, et al. Tryptophan-derived microbial metabolites 760 activate the aryl hydrocarbon receptor in tumor-associated macrophages to suppress anti-761 tumor immunity. Immunity. Feb 8 2022;55(2):324-340.e8. doi:10.1016/j.immuni.2022.01.006 762 36. Chang C, Geng F, Shi X, et al. The prevalence rate of periodontal pathogens and its 763 association with oral squamous cell carcinoma. Appl Microbiol Biotechnol. Feb 764 2019;103(3):1393-1404. doi:10.1007/s00253-018-9475-6 765 37. Zhou J, Wang L, Yuan R, et al. Signatures of Mucosal Microbiome in Oral Squamous 766 Cell Carcinoma Identified Using a Random Forest Model. Cancer Manag Res. 767 2020;12:5353-5363. doi:10.2147/cmar.S251021 768 38. Torralba MG, Aleti G, Li W, et al. Oral Microbial Species and Virulence Factors 769 Associated with Oral Squamous Cell Carcinoma. Microb Ecol. Nov 2021;82(4):1030-1046. 770 doi:10.1007/s00248-020-01596-5 771 39. Henrich B, Rumming M, Sczyrba A, et al. Mycoplasma salivarium as a dominant 772 coloniser of Fanconi anaemia associated oral carcinoma. PLoS One. 2014;9(3):e92297. 773 doi:10.1371/journal.pone.0092297 774 40. Chan JYK, Ng CWK, Lan L, et al. Restoration of the Oral Microbiota After Surgery 775 for Head and Neck Squamous Cell Carcinoma Is Associated With Patient Outcomes. Front 776 Oncol. 2021;11:737843. doi:10.3389/fonc.2021.737843 777 41. De Martin A, Lütge M, Stanossek Y, et al. Distinct microbial communities colonize 778 tonsillar squamous cell carcinoma. Oncoimmunology. 2021;10(1):1945202. 779 doi:10.1080/2162402x.2021.1945202 780 42. Zakrzewski M, Gannon OM, Panizza BJ, Saunders NA, Antonsson A. Human 781 papillomavirus infection and tumor microenvironment are associated with the microbiota in 782 patients with oropharyngeal cancers-pilot study. Head Neck. Nov 2021;43(11):3324-3330. 783 doi:10.1002/hed.26821 784 43. Sarkar P, Malik S, Laha S, et al. Dysbiosis of Oral Microbiota During Oral Squamous 785 Cell Carcinoma Development. Front Oncol. 2021;11:614448. doi:10.3389/fonc.2021.614448 786 44. Zhou X, Hao Y, Peng X, et al. The Clinical Potential of Oral Microbiota as a 787 Screening Tool for Oral Squamous Cell Carcinomas. Front Cell Infect Microbiol. 788 2021;11:728933. doi:10.3389/fcimb.2021.728933 789 45. Chen Z, Wong PY, Ng CWK, et al. The Intersection between Oral Microbiota, Host 790 Gene Methylation and Patient Outcomes in Head and Neck Squamous Cell Carcinoma. 791 Cancers (Basel). Nov 18 2020;12(11)doi:10.3390/cancers12113425 792 46. Perera M, Al-Hebshi NN, Perera I, et al. Inflammatory Bacteriome and Oral 793 Squamous Cell Carcinoma. J Dent Res. Jun 2018;97(6):725-732. 794 d oi:10.1177/0022034518767118 795 47. Al-Hebshi NN, Nasher AT, Maryoud MY, et al. Inflammatory bacteriome featuring 796 Fusobacterium nucleatum and Pseudomonas aeruginosa identified in association with oral 797 squamous cell carcinoma. Sci Rep. May 12 2017;7(1):1834. doi:10.1038/s41598-017-02079-798 3 799 48. Schmidt BL, Kuczynski J, Bhattacharya A, et al. Changes in abundance of oral 800 microbiota associated with oral cancer. PLoS One. 2014;9(6):e98741. 801 doi:10.1371/journal.pone.0098741 802 49. Wang H, Funchain P, Bebek G, et al. Microbiomic differences in tumor and paired-803 normal tissue in head and neck squamous cell carcinomas. Genome Med. Feb 7 2017;9(1):14. 804 doi:10.1186/s13073-017-0405-5 805 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 38 50. Guerrero-Preston R, Godoy-Vitorino F, Jedlicka A, et al. 16S rRNA amplicon 806 sequencing identifies microbiota associated with oral cancer, human papilloma virus 807 infection and surgical treatment. Oncotarget. Aug 9 2016;7(32):51320-51334. 808 doi:10.18632/oncotarget.9710 809 51. Gong H, Shi Y, Zhou X, et al. Microbiota in the Throat and Risk Factors for 810 Laryngeal Carcinoma. Appl Environ Microbiol. Dec 2014;80(23):7356-63. 811 doi:10.1128/aem.02329-14 812 52. Gong HL, Shi Y, Zhou L, et al. The Composition of Microbiome in Larynx and the 813 Throat Biodiversity between Laryngeal Squamous Cell Carcinoma Patients and Control 814 Population. PLoS One. 2013;8(6):e66476. doi:10.1371/journal.pone.0066476 815 53. Pushalkar S, Ji X, Li Y, et al. Comparison of oral microbiota in tumor and non-tumor 816 tissues of patients with oral squamous cell carcinoma. BMC Microbiol. Jul 20 2012;12:144. 817 doi:10.1186/1471-2180-12-144 818 54. Burcher KM, Burcher JT, Inscore L, Bloomer CH, Furdui CM, Porosnicu M. A 819 Review of the Role of Oral Microbiome in the Development, Detection, and Management of 820 Head and Neck Squamous Cell Cancers. Cancers (Basel). Aug 25 821 2022;14(17)doi:10.3390/cancers14174116 822 55. Yang K, Wang Y, Zhang S, et al. Oral Microbiota Analysis of Tissue Pairs and Saliva 823 Samples From Patients With Oral Squamous Cell Carcinoma - A Pilot Study. Front 824 Microbiol. 2021;12:719601. doi:10.3389/fmicb.2021.719601 825 56. Gopinath D, Kunnath Menon R, Chun Wie C, et al. Salivary bacterial shifts in oral 826 leukoplakia resemble the dysbiotic oral cancer bacteriome. J Oral Microbiol. Dec 9 827 2020;13(1):1857998. doi:10.1080/20002297.2020.1857998 828 57. Gong H, Shi Y, Xiao X, et al. Alterations of microbiota structure in the larynx 829 relevant to laryngeal carcinoma. Sci Rep. Jul 14 2017;7(1):5507. doi:10.1038/s41598-017-830 05576-7 831 58. Dong Z, Zhang C, Zhao Q, et al. Alterations of bacterial communities of vocal cord 832 mucous membrane increases the risk for glottic laryngeal squamous cell carcinoma. J 833 Cancer. 2021;12(13):4049-4063. doi:10.7150/jca.54221 834 59. Shin JM, Luo T, Kamarajan P, Fenno JC, Rickard AH, Kapila YL. Microbial 835 Communities Associated with Primary and Metastatic Head and Neck Squamous Cell 836 Carcinoma - A High Fusobacterial and Low Streptococcal Signature. Sci Rep. Aug 30 837 2017;7(1):9934. doi:10.1038/s41598-017-09786-x 838 60. Chan JYK, Cheung MK, Lan L, et al. Characterization of oral microbiota in HPV and 839 non-HPV head and neck squamous cell carcinoma and its association with patient outcomes. 840 Oral Oncol. Dec 2022;135:106245. doi:10.1016/j.oraloncology.2022.106245 841 61. Choi YS, Kim Y, Yoon HJ, et al. The presence of bacteria within tissue provides 842 insights into the pathogenesis of oral lichen planus. Sci Rep. Jul 7 2016;6:29186. 843 doi:10.1038/srep29186 844 6 2. Baek K, Lee J, Lee A, et al. Characterization of intratissue bacterial communities and 845 isolation of Escherichia coli from oral lichen planus lesions. Sci Rep. Feb 26 846 2020;10(1):3495. doi:10.1038/s41598-020-60449-w 847 63. Bao K, Li X, Poveda L, et al. Proteome and Microbiome Mapping of Human Gingival 848 Tissue in Health and Disease. Front Cell Infect Microbiol. 2020;10:588155. 849 doi:10.3389/fcimb.2020.588155 850 64. Sawant S, Dugad J, Parikh D, Srinivasan S, Singh H. Identification & correlation of 851 bacterial diversity in oral cancer and long-term tobacco chewers- A case-control pilot study. J 852 Med Microbiol. Sep 2021;70(9)doi:10.1099/jmm.0.001417 853 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 39 65. Frank DN, Qiu Y, Cao Y, et al. A dysbiotic microbiome promotes head and neck 854 squamous cell carcinoma. Oncogene. Feb 2022;41(9):1269-1280. doi:10.1038/s41388-021-855 02137-1 856 66. Sharma AK, DeBusk WT, Stepanov I, Gomez A, Khariwala SS. Oral Microbiome 857 Profiling in Smokers with and without Head and Neck Cancer Reveals Variations Between 858 Health and Disease. Cancer Prev Res (Phila). May 2020;13(5):463-474. doi:10.1158/1940-859 6207.Capr-19-0459 860 67. Lau HC, Hsueh CY, Gong H, et al. Oropharynx microbiota transitions in 861 hypopharyngeal carcinoma treatment of induced chemotherapy followed by surgery. BMC 862 Microbiol. Nov 9 2021;21(1):310. doi:10.1186/s12866-021-02362-4 863 68. Hsueh CY, Gong H, Cong N, et al. Throat Microbial Community Structure and 864 Functional Changes in Postsurgery Laryngeal Carcinoma Patients. Appl Environ Microbiol. 865 Nov 24 2020;86(24)doi:10.1128/aem.01849-20 866 69. Panda M, Rai AK, Rahman T, et al. Alterations of salivary microbial community 867 associated with oropharyngeal and hypopharyngeal squamous cell carcinoma patients. Arch 868 Microbiol. May 2020;202(4):785-805. doi:10.1007/s00203-019-01790-1 869 70. Vesty A, Gear K, Biswas K, Radcliff FJ, Taylor MW, Douglas RG. Microbial and 870 inflammatory-based salivary biomarkers of head and neck squamous cell carcinoma. Clin 871 Exp Dent Res. Dec 2018;4(6):255-262. doi:10.1002/cre2.139 872 71. Lee WH, Chen HM, Yang SF, et al. Bacterial alterations in salivary microbiota and 873 their association in oral cancer. Sci Rep. Nov 28 2017;7(1):16540. doi:10.1038/s41598-017-874 16418-x 875 72. Amer A, Galvin S, Healy CM, Moran GP. The Microbiome of Potentially Malignant 876 Oral Leukoplakia Exhibits Enrichment for Fusobacterium, Leptotrichia, Campylobacter, and 877 Rothia Species. Front Microbiol. 2017;8:2391. doi:10.3389/fmicb.2017.02391 878 73. Chen MY, Chen JW, Wu LW, et al. Carcinogenesis of Male Oral Submucous Fibrosis 879 Alters Salivary Microbiomes. J Dent Res. Apr 2021;100(4):397-405. 880 doi:10.1177/0022034520968750 881 74. Debelius JW, Huang T, Cai Y, et al. Subspecies Niche Specialization in the Oral 882 Microbiome Is Associated with Nasopharyngeal Carcinoma Risk. mSystems. Jul 7 883 2020;5(4)doi:10.1128/mSystems.00065-20 884 75. Kumpitsch C, Moissl-Eichinger C, Pock J, Thurnher D, Wolf A. Preliminary insights 885 into the impact of primary radiochemotherapy on the salivary microbiome in head and neck 886 squamous cell carcinoma. Sci Rep. Oct 6 2020;10(1):16582. doi:10.1038/s41598-020-73515-887 0 888 76. Wolf A, Moissl-Eichinger C, Perras A, Koskinen K, Tomazic PV, Thurnher D. The 889 salivary microbiome as an indicator of carcinogenesis in patients with oropharyngeal 890 squamous cell carcinoma: A pilot study. Sci Rep. Jul 19 2017;7(1):5867. doi:10.1038/s41598-891 01 7-06361-2 892 77. Zhu XX, Yang XJ, Chao YL, et al. The Potential Effect of Oral Microbiota in the 893 Prediction of Mucositis During Radiotherapy for Nasopharyngeal Carcinoma. EBioMedicine. 894 Apr 2017;18:23-31. doi:10.1016/j.ebiom.2017.02.002 895 78. Furquim CP, Soares GM, Ribeiro LL, et al. The Salivary Microbiome and Oral 896 Cancer Risk: a Pilot Study in Fanconi Anemia. J Dent Res. Mar 2017;96(3):292-299. 897 doi:10.1177/0022034516678169 898 79. Zhang J, Liu H, Liang X, et al. Investigation of salivary function and oral microbiota 899 of radiation caries-free people with nasopharyngeal carcinoma. PLoS One. 900 2015;10(4):e0123137. doi:10.1371/journal.pone.0123137 901 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 40 80. Hu YJ, Wang Q, Jiang YT, et al. Characterization of oral bacterial diversity of 902 irradiated patients by high-throughput sequencing. Int J Oral Sci. Mar 2013;5(1):21-5. 903 doi:10.1038/ijos.2013.15 904 81. Pushalkar S, Mane SP, Ji X, et al. Microbial diversity in saliva of oral squamous cell 905 carcinoma. FEMS Immunol Med Microbiol. Apr 2011;61(3):269-77. doi:10.1111/j.1574-906 695X.2010.00773.x 907 82. Hu YJ, Shao ZY, Wang Q, et al. Exploring the dynamic core microbiome of plaque 908 microbiota during head-and-neck radiotherapy using pyrosequencing. PLoS One. 909 2013;8(2):e56343. doi:10.1371/journal.pone.0056343 910 83. Minarovits J. Anaerobic bacterial communities associated with oral carcinoma: 911 Intratumoral, surface-biofilm and salivary microbiota. Anaerobe. Apr 2021;68:102300. 912 doi:10.1016/j.anaerobe.2020.102300 913 84. Orlandi E, Iacovelli NA, Tombolini V, et al. Potential role of microbiome in 914 oncogenesis, outcome prediction and therapeutic targeting for head and neck cancer. Oral 915 Oncol. Dec 2019;99:104453. doi:10.1016/j.oraloncology.2019.104453 916 85. Gopinath D, Menon RK, Wie CC, et al. Differences in the bacteriome of swab, saliva, 917 and tissue biopsies in oral cancer. Sci Rep. Jan 13 2021;11(1):1181. doi:10.1038/s41598-020-918 80859-0 919 86. Jain V, Baraniya D, El-Hadedy DE, et al. Integrative Metatranscriptomic Analysis 920 Reveals Disease-specific Microbiome–host Interactions in Oral Squamous Cell Carcinoma. 921 Cancer Res Commun. 2023;3(5):807-820. doi:10.1158/2767-9764.CRC-22-0349 922 87. Wang X, Zhao Z, Tang N, et al. Microbial Community Analysis of Saliva and 923 Biopsies in Patients With Oral Lichen Planus. Front Microbiol. 2020;11:629. 924 doi:10.3389/fmicb.2020.00629 925 88. Zhang Z, Feng Q, Li M, et al. Age-Related Cancer-Associated Microbiota Potentially 926 Promotes Oral Squamous Cell Cancer Tumorigenesis by Distinct Mechanisms. Front 927 Microbiol. 2022;13:852566. doi:10.3389/fmicb.2022.852566 928 89. Nie F, Wang L, Huang Y, et al. Characteristics of Microbial Distribution in Different 929 Oral Niches of Oral Squamous Cell Carcinoma. Front Cell Infect Microbiol. 930 2022;12:905653. doi:10.3389/fcimb.2022.905653 931 90. Qiao H, Li H, Wen X, Tan X, Yang C, Liu N. Multi-Omics Integration Reveals the 932 Crucial Role of Fusobacterium in the Inflammatory Immune Microenvironment in Head and 933 Neck Squamous Cell Carcinoma. Microbiol Spectr. Aug 31 2022;10(4):e0106822. 934 doi:10.1128/spectrum.01068-22 935 91. Wang Y, KA LC. PLSDA-batch: a multivariate framework to correct for batch effects 936 in microbiome data. Brief Bioinform. Mar 19 2023;24(2)doi:10.1093/bib/bbac622 937 92. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated 938 guideline for reporting systematic reviews. Rev Esp Cardiol (Engl Ed). Sep 2021;74(9):790-939 7 99. Declaración PRISMA 2020: una guía actualizada para la publicación de revisiones 940 sistemáticas. doi:10.1016/j.rec.2021.07.010 941 93. Choudhary S. pysradb: A Python package to query next-generation sequencing 942 metadata and data from NCBI Sequence Read Archive. F1000Res. 2019;8:532. 943 doi:10.12688/f1000research.18676.1 944 94. Leinonen R, Sugawara H, Shumway M. The sequence read archive. Nucleic Acids 945 Res. Jan 2011;39(Database issue):D19-21. doi:10.1093/nar/gkq1019 946 95. Bolyen E, Rideout JR, Dillon MR, et al. Reproducible, interactive, scalable and 947 extensible microbiome data science using QIIME 2. Nat Biotechnol. Aug 2019;37(8):852-948 857. doi:10.1038/s41587-019-0209-9 949 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 41 96. Rohart F, Gautier B, Singh A, KA LC. mixOmics: An R package for 'omics feature 950 selection and multiple data integration. PLoS Comput Biol. Nov 2017;13(11):e1005752. 951 doi:10.1371/journal.pcbi.1005752 952 97. McMurdie PJ, Holmes S. phyloseq: an R package for reproducible interactive analysis 953 and graphics of microbiome census data. PLoS One. 2013;8(4):e61217. 954 doi:10.1371/journal.pone.0061217 955 98. Moentadj R, Wang Y, Bowerman K, et al. Streptococcus species enriched in the oral 956 cavity of patients with RA are a source of peptidoglycan-polysaccharide polymers that can 957 induce arthritis in mice. Ann Rheum Dis. May 2021;80(5):573-581. 958 doi:10.1136/annrheumdis-2020-219009 959 99. Gloor GB, Macklaim JM, Pawlowsky-Glahn V, Egozcue JJ. Microbiome Datasets 960 Are Compositional: And This Is Not Optional. Front Microbiol. 2017;8:2224. 961 doi:10.3389/fmicb.2017.02224 962 100. Khomich M, Måge I, Rud I, Berget I. Analysing microbiome intervention design 963 studies: Comparison of alternative multivariate statistical methods. PLoS One. 964 2021;16(11):e0259973. doi:10.1371/journal.pone.0259973 965 101. Cao Y, Dong Q, Wang D, Zhang P, Liu Y, Niu C. microbiomeMarker: an 966 R/Bioconductor package for microbiome marker identification and visualization. 967 Bioinformatics. Aug 10 2022;38(16):4027-4029. doi:10.1093/bioinformatics/btac438 968 102. Douglas GM, Maffei VJ, Zaneveld JR, et al. PICRUSt2 for prediction of metagenome 969 functions. Nat Biotechnol. Jun 2020;38(6):685-688. doi:10.1038/s41587-020-0548-6 970 103. Caspi R, Billington R, Keseler IM, et al. The MetaCyc database of metabolic 971 pathways and enzymes - a 2019 update. Nucleic Acids Res. Jan 8 2020;48(D1):D445-d453. 972 doi:10.1093/nar/gkz862 973 104. Gao J, Aksoy BA, Dogrusoz U, et al. Integrative analysis of complex cancer 974 genomics and clinical profiles using the cBioPortal. Sci Signal. Apr 2 2013;6(269):pl1. 975 doi:10.1126/scisignal.2004088 976 105. Bagaev A, Kotlov N, Nomie K, et al. Conserved pan-cancer microenvironment 977 subtypes predict response to immunotherapy. Cancer Cell. Jun 14 2021;39(6):845-865.e7. 978 doi:10.1016/j.ccell.2021.04.014 979 106. Eren AM, Borisy GG, Huse SM, Mark Welch JL. Oligotyping analysis of the human 980 oral microbiome. Proc Natl Acad Sci U S A. Jul 15 2014;111(28):E2875-84. 981 doi:10.1073/pnas.1409644111 982 107. Mark Welch JL, Rossetti BJ, Rieken CW, Dewhirst FE, Borisy GG. Biogeography of 983 a human oral microbiome at the micron scale. Proc Natl Acad Sci U S A. Feb 9 984 2016;113(6):E791-800. doi:10.1073/pnas.1522149113 985 108. Wilbert SA, Mark Welch JL, Borisy GG. Spatial Ecology of the Human Tongue 986 Dorsum Microbiome. Cell Rep. Mar 24 2020;30(12):4003-4015.e3. 987 doi:10.1016/j.celrep.2020.02.097 988 109. Shao W, Fujiwara N, Mouri Y, et al. Conversion from epithelial to partial-EMT 989 phenotype by Fusobacterium nucleatum infection promotes invasion of oral cancer cells. Sci 990 Rep. Jul 22 2021;11(1):14943. doi:10.1038/s41598-021-94384-1 991 110. Chen G, Gao C, Jiang S, et al. Fusobacterium nucleatum outer membrane vesicles 992 activate autophagy to promote oral cancer metastasis. J Adv Res. Apr 13 993 2023;doi:10.1016/j.jare.2023.04.002 994 111. Geng F, Zhang Y, Lu Z, Zhang S, Pan Y. Fusobacterium nucleatum Caused DNA 995 Damage and Promoted Cell Proliferation by the Ku70/p53 Pathway in Oral Cancer Cells. 996 DNA Cell Biol. Jan 2020;39(1):144-151. doi:10.1089/dna.2019.5064 997 112. Zhang S, Li C, Liu J, et al. Fusobacterium nucleatum promotes epithelial-998 mesenchymal transiton through regulation of the lncRNA MIR4435-2HG/miR-296-999 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 42 5p/Akt2/SNAI1 signaling pathway. Febs j. Sep 2020;287(18):4032-4047. 1000 doi:10.1111/febs.15233 1001 113. Binder Gallimidi A, Fischman S, Revach B, et al. Periodontal pathogens 1002 Porphyromonas gingivalis and Fusobacterium nucleatum promote tumor progression in an 1003 oral-specific chemical carcinogenesis model. Oncotarget. Sep 8 2015;6(26):22613-23. 1004 doi:10.18632/oncotarget.4209 1005 114. Baty JJ, Stoner SN, Scoffield JA. Oral Commensal Streptococci: Gatekeepers of the 1006 Oral Cavity. J Bacteriol. Nov 15 2022;204(11):e0025722. doi:10.1128/jb.00257-22 1007 115. Tsai MS, Chen YY, Chen WC, Chen MF. Streptococcus mutans promotes tumor 1008 progression in oral squamous cell carcinoma. J Cancer. 2022;13(12):3358-3367. 1009 doi:10.7150/jca.73310 1010 116. Baraniya D, Jain V, Lucarelli R, et al. Screening of Health-Associated Oral Bacteria 1011 for Anticancer Properties in vitro. Front Cell Infect Microbiol. 2020;10:575656. 1012 doi:10.3389/fcimb.2020.575656 1013 117. Xu Y, Jia Y, Chen L, Gao J, Yang D. Effect of Streptococcus anginosus on biological 1014 response of tongue squamous cell carcinoma cells. BMC Oral Health. Mar 20 1015 2021;21(1):141. doi:10.1186/s12903-021-01505-3 1016 118. Baraniya D, Chitrala KN, Al-Hebshi NN. Global transcriptional response of oral 1017 squamous cell carcinoma cell lines to health-associated oral bacteria - an in vitro study. J 1018 Oral Microbiol. 2022;14(1):2073866. doi:10.1080/20002297.2022.2073866 1019 119. Wang J, Sun F, Lin X, Li Z, Mao X, Jiang C. Cytotoxic T cell responses to 1020 Streptococcus are associated with improved prognosis of oral squamous cell carcinoma. Exp 1021 Cell Res. Jan 1 2018;362(1):203-208. doi:10.1016/j.yexcr.2017.11.018 1022 120. Wang J, Yang L, Mao X, Li Z, Lin X, Jiang C. Streptococcus salivarius-mediated 1023 CD8(+) T cell stimulation required antigen presentation by macrophages in oral squamous 1024 cell carcinoma. Exp Cell Res. May 15 2018;366(2):121-126. doi:10.1016/j.yexcr.2018.03.007 1025 121. Curry KD, Wang Q, Nute MG, et al. Emu: species-level microbial community 1026 profiling of full-length 16S rRNA Oxford Nanopore sequencing data. Nat Methods. Jul 1027 2022;19(7):845-853. doi:10.1038/s41592-022-01520-4 1028 122. Johnson JS, Spakowicz DJ, Hong BY, et al. Evaluation of 16S rRNA gene sequencing 1029 for species and strain-level microbiome analysis. Nat Commun. Nov 6 2019;10(1):5029. 1030 doi:10.1038/s41467-019-13036-1 1031 123. Gehrig JL, Portik DM, Driscoll MD, et al. Finding the right fit: evaluation of short-1032 read and long-read sequencing approaches to maximize the utility of clinical microbiome 1033 data. Microb Genom. Mar 2022;8(3)doi:10.1099/mgen.0.000794 1034 124. Garcia-Bermudez J, Baudrier L, La K, et al. Aspartate is a limiting metabolite for 1035 cancer cell proliferation under hypoxia and in tumours. Nat Cell Biol. Jul 2018;20(7):775-1036 78 1. doi:10.1038/s41556-018-0118-z 1037 125. Krall AS, Xu S, Graeber TG, Braas D, Christofk HR. Asparagine promotes cancer 1038 cell proliferation through use as an amino acid exchange factor. Nat Commun. Apr 29 1039 2016;7:11457. doi:10.1038/ncomms11457 1040 126. Halbrook CJ, Thurston G, Boyer S, et al. Differential integrated stress response and 1041 asparagine production drive symbiosis and therapy resistance of pancreatic adenocarcinoma 1042 cells. Nat Cancer. Nov 2022;3(11):1386-1403. doi:10.1038/s43018-022-00463-1 1043 127. Donohoe DR, Garge N, Zhang X, et al. The microbiome and butyrate regulate energy 1044 metabolism and autophagy in the mammalian colon. Cell Metab. May 4 2011;13(5):517-26. 1045 doi:10.1016/j.cmet.2011.02.018 1046 128. Mashimo T, Pichumani K, Vemireddy V, et al. Acetate is a bioenergetic substrate for 1047 human glioblastoma and brain metastases. Cell. Dec 18 2014;159(7):1603-14. 1048 doi:10.1016/j.cell.2014.11.025 1049 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 43 129. Lyssiotis CA, Cantley LC. Acetate fuels the cancer engine. Cell. Dec 18 1050 2014;159(7):1492-4. doi:10.1016/j.cell.2014.12.009 1051 130. Comerford SA, Huang Z, Du X, et al. Acetate dependence of tumors. Cell. Dec 18 1052 2014;159(7):1591-602. doi:10.1016/j.cell.2014.11.020 1053 131. Donohoe DR, Collins LB, Wali A, Bigler R, Sun W, Bultman SJ. The Warburg effect 1054 dictates the mechanism of butyrate-mediated histone acetylation and cell proliferation. Mol 1055 Cell. Nov 30 2012;48(4):612-26. doi:10.1016/j.molcel.2012.08.033 1056 132. Koh A, De Vadder F, Kovatcheva-Datchary P, Bäckhed F. From Dietary Fiber to 1057 Host Physiology: Short-Chain Fatty Acids as Key Bacterial Metabolites. Cell. Jun 2 1058 2016;165(6):1332-1345. doi:10.1016/j.cell.2016.05.041 1059 133. van der Hee B, Wells JM. Microbial Regulation of Host Physiology by Short-chain 1060 Fatty Acids. Trends Microbiol. Aug 2021;29(8):700-712. doi:10.1016/j.tim.2021.02.001 1061 134. Li Z, Wang Q, Huang X, et al. Lactate in the tumor microenvironment: A rising star 1062 for targeted tumor therapy. Front Nutr. 2023;10:1113739. doi:10.3389/fnut.2023.1113739 1063 135. Mima K, Sukawa Y, Nishihara R, et al. Fusobacterium nucleatum and T Cells in 1064 Colorectal Carcinoma. JAMA Oncol. Aug 2015;1(5):653-61. 1065 doi:10.1001/jamaoncol.2015.1377 1066 136. Kim HS, Kim CG, Kim WK, et al. Fusobacterium nucleatum induces a tumor 1067 microenvironment with diminished adaptive immunity against colorectal cancers. Front Cell 1068 Infect Microbiol. 2023;13:1101291. doi:10.3389/fcimb.2023.1101291 1069 137. Kosumi K, Baba Y, Yamamura K, et al. Intratumour Fusobacterium nucleatum and 1070 immune response to oesophageal cancer. Br J Cancer. Apr 2023;128(6):1155-1165. 1071 doi:10.1038/s41416-022-02112-x 1072 138. Dahlstrand Rudin A, Khamzeh A, Venkatakrishnan V, Basic A, Christenson K, 1073 Bylund J. Short chain fatty acids released by Fusobacterium nucleatum are neutrophil 1074 chemoattractants acting via free fatty acid receptor 2 (FFAR2). Cell Microbiol. Aug 1075 2021;23(8):e13348. doi:10.1111/cmi.13348 1076 139. Mendes RT, Nguyen D, Stephens D, et al. Endothelial Cell Response to 1077 Fusobacterium nucleatum. Infect Immun. Jul 2016;84(7):2141-2148. doi:10.1128/iai.01305-1078 15 1079 140. Wang Q, Zhao L, Xu C, Zhou J, Wu Y. Fusobacterium nucleatum stimulates 1080 monocyte adhesion to and transmigration through endothelial cells. Arch Oral Biol. Apr 1081 2019;100:86-92. doi:10.1016/j.archoralbio.2019.02.013 1082 141. Wright HJ, Chapple IL, Matthews JB, Cooper PR. Fusobacterium nucleatum 1083 regulation of neutrophil transcription. J Periodontal Res. Feb 2011;46(1):1-12. 1084 doi:10.1111/j.1600-0765.2010.01299.x 1085 142. Zhou T, Meng X, Wang D, Fu W, Li X. Neutrophil Transcriptional Deregulation by 1086 t he Periodontal Pathogen Fusobacterium nucleatum in Gastric Cancer: A Bioinformatic 1087 Study. Dis Markers. 2022;2022:9584507. doi:10.1155/2022/9584507 1088 143. Neuzillet C, Marchais M, Vacher S, et al. Prognostic value of intratumoral 1089 Fusobacterium nucleatum and association with immune-related gene expression in oral 1090 squamous cell carcinoma patients. Sci Rep. Apr 12 2021;11(1):7870. doi:10.1038/s41598-1091 021-86816-9 1092 144. Mima K, Nishihara R, Qian ZR, et al. Fusobacterium nucleatum in colorectal 1093 carcinoma tissue and patient prognosis. Gut. Dec 2016;65(12):1973-1980. 1094 doi:10.1136/gutjnl-2015-310101 1095 145. Lehr K, Nikitina D, Vilchez-Vargas R, et al. Microbial composition of tumorous and 1096 adjacent gastric tissue is associated with prognosis of gastric cancer. Sci Rep. Mar 21 1097 2023;13(1):4640. doi:10.1038/s41598-023-31740-3 1098 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 44 146. Hsieh YY, Kuo WL, Hsu WT, Tung SY, Li C. Fusobacterium Nucleatum-Induced 1099 Tumor Mutation Burden Predicts Poor Survival of Gastric Cancer Patients. Cancers (Basel). 1100 Dec 30 2022;15(1)doi:10.3390/cancers15010269 1101 147. Zhang N, Liu Y, Yang H, et al. Clinical Significance of Fusobacterium nucleatum 1102 Infection and Regulatory T Cell Enrichment in Esophageal Squamous Cell Carcinoma. 1103 Pathol Oncol Res. 2021;27:1609846. doi:10.3389/pore.2021.1609846 1104 148. Jo AR, Baek KJ, Shin JE, Choi Y. Mechanisms of IL-8 suppression by Treponema 1105 denticola in gingival epithelial cells. Immunol Cell Biol. Feb 2014;92(2):139-47. 1106 doi:10.1038/icb.2013.80 1107 149. Babolin C, Amedei A, Ozolins D, Zilevica A, D'Elios MM, de Bernard M. TpF1 from 1108 Treponema pallidum activates inflammasome and promotes the development of regulatory T 1109 cells. J Immunol. Aug 1 2011;187(3):1377-84. doi:10.4049/jimmunol.1100615 1110 150. Hashimoto M, Asai Y, Ogawa T. Treponemal phospholipids inhibit innate immune 1111 responses induced by pathogen-associated molecular patterns. J Biol Chem. Nov 7 1112 2003;278(45):44205-13. doi:10.1074/jbc.M306735200 1113 151. Kataoka H, Taniguchi M, Fukamachi H, Arimoto T, Morisaki H, Kuwata H. Rothia 1114 dentocariosa induces TNF-alpha production in a TLR2-dependent manner. Pathog Dis. Jun 1115 2014;71(1):65-8. doi:10.1111/2049-632x.12115 1116 152. Hawkes CG, Hinson AN, Vashishta A, et al. Selenomonas sputigena Interactions with 1117 Gingival Epithelial Cells That Promote Inflammation. Infect Immun. Feb 16 1118 2023;91(2):e0031922. doi:10.1128/iai.00319-22 1119 153. Xu Z, Lv Z, Chen F, et al. Dysbiosis of human tumor microbiome and aberrant 1120 residence of Actinomyces in tumor-associated fibroblasts in young-onset colorectal cancer. 1121 Front Immunol. 2022;13:1008975. doi:10.3389/fimmu.2022.1008975 1122 1123 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint Supplementary Figures 1124 1125 1126 1127 1128 Figure S1: CLR-normalized abundances for remaining 13 bacteria between sample groups using unpaired 1129 Kruskal-Wallis test with Bonferroni’s multiple comparison. Post-hoc Wilcoxon test with Bonferroni-Dunn’s 1130 multiple comparison was performed to identify group-wise differences between Cancer – Non-cancer (#), 1131 Cancer – Cancer-adjacent (*), Non-cancer – Cancer-adjacent (^). 1132 1133 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint 1134 Figure S2: Beta-diversity between paired cancer and cancer-adjacent tissue samples. Paired Wilcoxon test was 1135 performed on Euclidean distance between each samples. No significant differences between cancer and cancer-1136 adjacent tissue samples. 1137 1138 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 27, 2023. ; https://doi.org/10.1101/2023.07.25.23293137doi: medRxiv preprint