Abstract

   Parvimonas micra (P. micra), an opportunistic oral pathogen associated
   with multiple cancers, has limited research on its role in oral
   squamous cell carcinoma (OSCC). This study shows that P. micra is
   enriched in OSCC tissues and positively correlated with tumor
   metastasis and stages. P. micra infection promotes OSCC metastasis by
   inducing hypoxia/HIF-1α, glycolysis, and autophagy. Mechanistically, P.
   micra surface protein TmpC binds to CKAP4, a receptor overexpressed in
   OSCC, facilitating bacterial attachment and invasion. This interaction
   activates HIF-1α and autophagy via CKAP4-RanBP2 and CKAP4-NBR1
   pathways, driving metastasis. Targeting CKAP4 with masitinib or
   antibodies impairs P. micra attachment and abolishes P. micra-promoted
   OSCC metastasis in vitro and in vivo. Together, our findings identify
   P. micra as a pathogen that promotes OSCC metastasis and highlight that
   TmpC-CKAP4 interaction could be a potential therapeutic target for
   OSCC.

   Subject terms: Oral cancer, Bacterial pathogenesis, Oral cancer
     __________________________________________________________________

   Parvimonas micra (P. micra) is an oral opportunistic pathogen
   frequently present in oral lesions and infections. Here, authors report
   that P. micra promotes oral squamous cell carcinoma metastasis by
   regulation HIF-1α-induced glycolysis and autophagy following TmpC
   protein binding to CKAP4 receptor.

Introduction

   Oral squamous cell carcinoma (OSCC) is a prevalent type of head and
   neck malignancy, comprising 2% to 4% of all cancer diagnosed^[49]1. Due
   to the high rate of metastasis, the 5-year survival rate remains below
   50%, despite advancements in the diagnosis and treatment of
   OSCC^[50]2,[51]3. Of these, lymph node micrometastasis (LM) is
   considered the most significant prognostic risk factor for OSCC,
   serving as the primary reason for the low survival rates observed in
   OSCC patients^[52]4. Hence, a comprehensive mechanism study of LM can
   help to better understand the pathogenesis of OSCC and to identify
   potential therapeutic targets for OSCC patients.

   Recent studies have demonstrated a strong correlation between the
   microbiota present within tumor tissues and the metastasis of multiple
   types of human cancers^[53]5–[54]7. For example, microbes residing in
   breast tumor-tissue, including Streptococcus xylosus, Lactobacillus
   animalis, and Staphylococcus cuniculi, promote breast tumor
   metastasis^[55]8. Fusobacterium (F.) nucleatum can drive colorectal
   cancer (CRC) epithelial-mesenchymal transition (EMT) and metastasis by
   activating TLR4/Keap1/NRF2 signaling pathway^[56]9 or inducing the
   over-expression of LncRNA EVADR^[57]10. Likewise, a number of oral
   pathogenic bacteria are associated with the progression of OSCC^[58]11.
   Among them, Porphyromonas (P.) gingivalis promotes OSCC development by
   modulating tumor-immune microenvironment and activating PI3K-AKT
   pathway^[59]12,[60]13; F. nucleatum facilitates OSCC EMT and metastasis
   by activating autophagy and GLUT1-driven lactate
   production^[61]14,[62]15. Therefore, revealing how oral pathogen
   promotes OSCC metastasis is a crucial topic in current studies.

   Parvimonas micra (P. micra) is not merely an oral opportunistic
   pathogen that is frequently detected in various oral diseases,
   including endodontic abscesses, periodontitis lesions, and orofacial
   odontogenic infections^[63]16,[64]17. Furthermore, P. micra is
   implicated in the development of various tumor types. For example, P.
   micra promotes CRC development by inducing cell proliferation and
   altering Th17 immune response^[65]18,[66]19. P. micra may be involved
   in the progression of gastric cancer, with significant centrality in
   the gastric cancer ecological network^[67]20. However, as one of the
   commonly opportunistic pathogens in the oral cavity, whether and how P.
   micra plays a role in the progression of OSCC remains unclear.

   In this study, we examined the abundance of P. micra in OSCC tissues
   and delved into its function in facilitating OSCC metastasis. We
   utilized single-cell RNA sequencing (scRNA-seq) to evaluate the impact
   of P. micra on both OSCC and the tumor microenvironment (TME).
   Subsequently, we pinpointed the P. micra surface adhesin and cellular
   receptor responsible for mediating its tumor-promoting activity,
   thereby elucidating the underlying intracellular molecular mechanisms.
   Ultimately, we identified a small chemical compound to confirm the role
   of the cell receptor in OSCC LM.

Results

P. micra enriched in OSCC tissue and correlated with OSCC metastasis

   To examine the relationship between P. micra and OSCC, we analyzed the
   abundance of P. micra in OSCC patients using the microbiome data of
   OSCC tissues, paracancerous tissues and normal oral mucosa
   (PRJNA866676)^[68]21. The results showed that P. micra was
   significantly enriched in OSCC tissues compared with paracancerous
   tissues and normal oral mucosa (Fig. [69]1a). Its abundance in
   metastatic OSCC was significantly higher than that in non-metastatic
   OSCC (Fig. [70]1b). Compared with paracancerous tissues, the abundance
   of P. micra consistently increased in OSCC patients from stage 1 to
   stage 3 (Fig. [71]1c). Next, RNA sequencing data ([72]GSE227919)^[73]22
   derived from OSCC versus paracancerous tissues was used to validated
   the enrichment of P. micra in OSCC tissue. The microbial communities of
   OSCC and paracancerous tissues were significantly different
   (Fig. [74]1d) and P. micra was enriched in OSCC tissues (Fig. [75]1e).
   In addition, the potential cellular functions of P. micra was examined
   by pathway enrichment analysis, which found that multiple oncogenic
   pathways, such as autophagy of mitochondrion, canonical Wnt and cell
   growth, were positively correlated with the abundance of P. micra
   (Supplementary Fig. [76]1a). Gene set variation analysis (GSVA) also
   showed that several gene sets, including Nod-like receptor signaling
   and pathogenic infection activation pathways in OSCC, were positively
   correlated with the abundance of P. micra, while tight junction was
   negatively correlated with P. micra (Supplementary Fig. [77]1b).
   Moreover, the abundance of P. micra was positively correlated with the
   expression of the genes that involved in tumor progression, such as
   TWIST1^[78]23, GLUT1^[79]24 and LDHA^[80]25 (Supplementary
   Fig. [81]1c).

Fig. 1. P. micra enriched in OSCC tissue and correlated with OSCC metastasis.

   [82]Fig. 1
   [83]Open in a new tab

   a The abundance of P. micra in OSCC tissue (n = 37 samples),
   paracancerous tissue (PC, n = 37 samples) and anatomically matched
   contralateral normal oral mucosa (NOM, n = 38 independent samples) was
   plotted in microbiome data. PC vs NOM, P < 0.0001; OSCC vs NOM,
   P < 0.0001. b Relative abundance of P. micra in OSCC tissues with
   (n = 13 samples) or without (n = 23 samples) metastasis. P = 0.018. c
   Relative abundance of P. micra in different stage of OSCC (PC, n = 37
   samples; I: T1 N0 M0, n = 2 samples; II: T2 N0 M0, n = 13 samples; III:
   T1-2 N1 M0 and T3 N0-1 M0, n = 7 samples; IV: T1-3 N2 M0 and T4a N0-2
   M0, n = 13 samples. III vs PC, P = 0.0056; IV vs PC, P = 0.0007). d
   PCoA showing bacterial community variations between paracancerous
   tissue (PC) and OSCC tissues ([84]GSE227919 dataset). The lower and
   upper hinges of the box represent the 25th and 75th percentiles, and
   the whiskers extend to the minimum and maximum values within 1.5 times
   the interquartile range. The median value is depicted by the line
   within the box (a-c). e The abundance of P. micra in PC (n = 17
   samples) and OSCC tissue (n = 8 samples) of [85]GSE227919 datase. f and
   g Representative images (f) and quantification analysis (g) of P. micra
   distribution by FISH staining (Green: P. micra) in normal tongue
   samples (n = 9 samples), non-metastatic (n = 13 samples) or metastatic
   (n = 32 samples) OSCC tissues. h and i Representative images (h) and
   Spearman correlation analysis (i) of P. micra and E-cadherin in OSCC
   tissue (n = 30 samples). j and k Representative images (j) and Spearman
   correlation (k) of P. micra and N-cadherin in OSCC tissue (n = 30
   samples). Scale bar: 100 μm. Data were presented as the means ± SD.
   Two-sided Wilcoxon test (a–c, e), two-sided unpaired Student’s t test
   (g) and two-sided spearman test (i, k) were used to examine the
   statistical significance between groups.

   To visualize the enrichment of P. micra in OSCC tissues, fluorescence
   in situ hybridization (FISH) assay used a FITC-labeled P.
   micra-specific probe was conducted. Consistently, P. micra was enriched
   in the metastasized OSCC tissue compared with that in the OSCC lesion
   without metastasis (Fig. [86]1f, g). As tumor metastasis is facilitated
   by EMT, the abundance of P. micra and EMT markers was explored by FISH
   and immunofluorescence (IF) assay, respectively. The results showed
   that the abundance of P. micra was negatively correlated with the
   expression of E-cadherin while positively correlated with N-cadherin
   (Fig. [87]1h–k). Overall, these results suggest that P. micra may
   involve in the EMT and metastasis of OSCC.

P. micra promoted OSCC metastasis

   To determine the impact of P. micra on OSCC metastasis in vitro, two
   different OSCC cell lines, CAL27 and SCC15, were cultured with P. micra
   to examine cell proliferation and metastasis, respectively. The results
   showed P. micra slightly promoted OSCC cell proliferation at low MOI
   (Supplementary Fig. [88]2a), while dramatically accelerated cell
   migration and invasion in a dose-dependent manner (Fig. [89]2a, b and
   Supplementary Fig. [90]2b, c). The expression of EMT markers in CAL27
   and SCC15 cells with/without P. micra infection (MOI = 100) was
   analyzed, which showed that E-cadherin was markedly decreased by P.
   micra infection, while N-cadherin was dramatically upregulated
   (Fig. [91]2c, d and Supplementary Fig. [92]2d). These results
   demonstrate that P. micra potentiates OSCC metastasis in vitro.

Fig. 2. P. micra promoted OSCC metastasis in vitro and in vivo.

   [93]Fig. 2
   [94]Open in a new tab

   a Wound-healing migration assay of OSCC cells co-cultured with P. micra
   in different MOI (n = 3 biological replicates per group). b Trans-well
   assays of OSCC cell pre-exposed with P. micra for 12 h in different
   MOI. The indicated migrated and invaded cells were quantified in five
   randomly selected fields. c Western blot assay of E-cadherin (E-cad)
   and N-cadherin (N-cad) in OSCC cells infected with P. micra for 24 h.
   The data are representative of at least two independent experiments
   with similar results. d qPCR assay of CDH1 and CDH2 (n = 3 biological
   replicates in each group) in OSCC cells stimulated with P. micra for
   24 h. e BALB/c nude mice were implanted with Luc-CAL27 or P. micra
   stimulated Luc-CAL27 cells on the left lateral edge of the tongue for 3
   weeks. f Images of tongue and quantification of tumor volume (n = 7
   mice per group). g Images of cervical lymph nodes (cLNs) and
   quantification of cLNs volume in vivo from 7 mice per group. h and i
   Representative images of immunofluorescence for CK5/6 (h) and
   quantification of metastatic area (i) in cLNs in vivo (n = 7 mice per
   group). Scale bar: 100 μm. j and k Western blot (j) and quantification
   (k) of E-cadherin and N-cadherin in CAL27 tumors in vivo (n = 3 mice
   per group). l Schematic diagram of single-cell RNA sequencing analysis
   of tumor cells from orthotopic CAL27 tumor model. m NNMF and UMAP of
   tumor cells. n Heatmap showed the differently expressed genes
   identified by NNMF. The corresponding gene signatures were numbered and
   selected genes indicated (red). o GSVA showing differences in hallmark
   pathways in tumor cells between P.micra-exposed tumors and control
   group. Data were presented as the means ± SD. One-way ANOVA with
   Turkey’s test (a, b, d, k) and two-sided unpaired Student’s t test (f,
   g, i) were used to examine the statistical significance between groups.
   The samples derive from the same experiment but different gels for
   E-cadherin, another for GAPDH and N-cadherin were processed in parallel
   (c, k).

   Next, a mouse oral cancer model was established by the implantation of
   Luc-CAL27 cells with/without P. micra infection on the left lateral
   edge of the tongue (Fig. [95]2e). The size tumor and cervical lymph
   nodes (cLNs) of the P. micra-treated group was significantly larger
   than that of the control group (Fig. [96]2f, g). CK5/6 staining showed
   P. micra increased the presence of OSCC cells in the cLNs (Fig.[97]2h,
   i). Furthermore, the expression of E-cadherin in P. micra-treated OSCC
   tissue was lower, while N-cadherin was significantly upregulated
   (Fig. [98]2j, k). These results indicate that P. micra enhances OSCC
   EMT and metastasis in vivo.

   To elucidate how P. micra promote OSCC metastasis, we performed
   scRNA-seq of the orthotopic CAL27 OSCC tumors with/without P. micra
   infection (Fig. [99]2l). A total of 49,765 single tumor cells were
   successfully characterized and classified into 6 separate clusters
   using non-negative matrix factorization (NNMF), and visualized by
   uniform manifold approximation and projection (UMAP, Fig. [100]2m).
   Clusters 2 and 3 exhibited a specific set of cell cycle-related genes
   indicative of the G1/S and G2/M phases, respectively (Fig. [101]2n).
   Cluster 4 was characterized with epithelial differentiation biomarkers,
   including KRT16, KRT6B, S100A7/8/9. Cluster 6 exhibited the highest
   activity in immune-related pathways, including TNF-α signaling,
   complement, and inflammatory response (Supplementary Fig. [102]3a).
   While Cluster 5 highly expressed genes associated with ECM (MMP1/10/13)
   and receptor of E-cadherin (ITGA2), which exhibited the higher activity
   in EMT related functions, including epithelial-mesenchymal transition,
   Wnt/β-catenin signaling, glycolysis and hypoxia (Supplementary
   Fig. [103]3a-d). Notably, P. micra increased the cell proportion of
   Cluster 5 compared with the control group (Supplementary Fig. [104]3e),
   suggesting P. micra may promoted OSCC LM via regulating EMT-related
   signaling pathways. Additionally, GSVA showed that the carcinogenic
   pathways, including Glycolysis, PI3K-AKT-mTOR, Hypoxia, and EMT, were
   significantly activated in all tumor cells from P. micra-treated group
   (Fig. [105]2o).

   Next, we clustered TME cells into six clusters by typical cell markers,
   which involved monocyte-macrophage, fibroblast, endothelial,
   neutrophil, dendritic and epithelial cells (Supplementary Fig. [106]4a,
   b). Odds ratios (OR) analysis showed that P. micra-treated group
   enriched in monocyte-macrophage, fibroblast, and endothelial cells in
   comparison to control group (Supplementary Fig. [107]4c). CellChat
   analysis was conducted to evaluate the interactions among six major TME
   components, which showed that P. micra markedly modified the overall
   interaction, especially enhanced the interaction between fibroblasts
   and other TME cells excluding epithelial cells (Supplementary
   Fig. [108]4d). For instance, P. micra promoted the crosstalk between
   fibroblasts and monocyte-macrophage through the Ccl7/8-Ccr1/2
   interaction (Supplementary Fig. [109]4e). These results suggest that
   fibroblast is an important cell type in P. micra-regulated OSCC TME.

   Five subtypes of cancer-associated fibroblasts (CAFs) were identified
   as well as normal fibroblasts (NFs) and smooth muscle cells (SMCs)
   (Supplementary Fig. [110]4f–g). P. micra-treated group was enriched in
   cancer-associated myofibroblasts (myCAFs) and antigen-presenting CAFs
   (apCAFs) in comparison to control group (Supplementary Fig. [111]4h).
   P. micra increased the expression of carcinogenic pathways, including
   angiogenesis, EMT and PI3K-AKT-mTOR in myCAFs (Supplementary
   Fig. [112]4i), and activated the immune-related pathways in apCAFs,
   including complement, inflammatory response and IL2-STAT5 signaling
   pathways (Supplementary Fig. [113]4j). These results suggest that P.
   micra may reshape CAFs in OSCC tissues.

   Tumor-associated macrophages (TAMs) are the most abundant immune cells
   in the OSCC TME (Supplementary Fig. [114]4a). The sub-population of
   monocyte-macrophage cells was characterized and identified 1 monocyte
   and 4 TAM subtypes (Supplementary Fig. [115]4k, l). P. micra increased
   the abundance of monocytes and activated the pathways of immune
   regulation, such as leukocyte transendothelial migration, chemokine
   signaling pathway, and cytokine-cytokine receptor interaction in
   monocytes (Supplementary Fig. [116]4m, n). These results indicate that
   P. micra may reshape the TAMs to promote OSCC metastasis.

P. micra promoted OSCC metastasis through HIF-1α-induced glycolysis and
autophagy

   To clarify the pathways involved in P. micra-promoted OSCC metastasis,
   we performed RNA-seq analysis on CAL27 with/without P. micra infection
   (MOI = 100). KEGG enrichment analysis revealed the upregulation of
   multiple oncogenic pathways, including HIF-1, NOD-like receptor,
   Glycolysis, and NF-κB in CAL27 cells stimulated with P. micra
   (Supplementary Fig. [117]5a). GSVA also showed that HIF-1 signaling
   pathway and Glycolysis/Gluconeogenesis were significantly activated by
   P. micra-infection (Fig. [118]3a).

Fig. 3. P. micra activated Hypoxia/HIF-1α, glycolysis and autophagy in OSCC
cells.

   [119]Fig. 3
   [120]Open in a new tab

   a The GSEA showing the enrichment of HIF-1 signaling pathway and
   Glycolysis/Gluconeogenesis pathway in P. micra stimulated group. The
   p-value is obtained through one-sided permutation test with 1000
   permutations. Adjusted p-value is the result of multiple validation
   correction on the original P-value. b Representative images of the
   Image-iT Green Hypoxia Reagent analysis. The data are representative of
   at least two independent experiments with similar results. c Western
   blot of HIF-1α, GLUT1, HK2 and PFKFB3 by P. micra infection with
   different MOI for 24 h. d Western blot of HIF-1α, GLUT1, HK2 and PFKFB3
   by P. micra infection (MOI = 100) and DMOG (1.0 mM) treatment for 24 h.
   e and f Glucose and lactate concentration in culture supernatant of
   CAL27 (e) and SCC15 (f) transfected with siHIF-1α for 24 h and followed
   by P. micra infection 24 h (n = 3 biological replicates in each group).
   g Western blot of HIF-1α, GLUT1, HK2 and PFKFB3 in OSCC cells
   transfected with siHIF-1α for 24 h and followed by P. micra infection
   24 h. h TEM showing autophagosomes in CAL27 and SCC15 following P.
   micra infection for 24 h (n = 5 biological replicates per group).
   Arrows indicated autophagosome. i Western blot of LC3B and P62 after P.
   micra infection 24 h. j Representative images and quantification of
   autophagosome and autolysosome in OSCC cells infected with mRFP-GFP-LC3
   adenovirus 24 h and stimulated with P. micra for another 24 h (n = 3
   biological replicates per group). White arrows indicated autophagosomes
   and red arrows indicated autolysosomes. scale bars, 100μm. k Western
   blot of LC3B and P62 in the presence of DMOG and P. micra for 24 h. l
   Western blot showing the impact of siHIF-1α on LC3B and P62. m Western
   blot showing the impact of 2-DG on LC3B and P62. Data were shown as
   mean ± SD. Blots are representative of at least two biological
   replicate experiments with similar results (c, d, g, i, k–m). The
   samples derive from the same experiment but different gels for GLUT1,
   another for HIF-1α and ACTB, another for HK2 and PFKFB3 were processed
   in parallel (c, d, g). One-way ANOVA with Turkey’s test (e, f, j) and
   two-sided unpaired Student’s t test (h) were used to examine the
   statistical significance between groups.

   P. micra-induced hypoxia was verified using the Image-iT Green Hypoxia
   Reagent (Fig. [121]3b, Supplementary Fig. [122]5b). Consistently, P.
   micra significantly upregulated the expression of HIF-1α in OSCC cells,
   while having no effect on HIF-2α (Fig. [123]3c, Supplementary
   Fig. [124]5c). Heat inactivated P. micra also increased the expression
   of HIF-1α (Supplementary Fig. [125]5d), indicating HIF-1α induced by P.
   micra is independent of intracellular P. micra. FISH and
   immunofluorescence (IF) analysis showed that the abundance of P. micra
   was positively correlated with the expression of HIF-1α in OSCC tissues
   (Supplementary Fig. [126]5e).

   Next, glycolysis was evaluated by comparing the expression of
   glycolytic proteins, and the levels of glucose uptake and lactate
   output in OSCC cells with/without P. micra infection (MOI = 100). The
   results showed that P. micra simultaneously elevated the expression of
   GLUT1, HK2 and PFKFB3 (Fig. [127]3c, d and Supplementary Fig. [128]6a,
   b), and significantly promoted glucose uptake and lactate output
   (Fig. [129]3e, f). These results indicate that P. micra activates
   glycolysis in OSCC cells.

   As a transcription factor, HIF-1α could promote the expression of
   glycolytic proteins, such as HK2, PFKFB3, and GLUT1 to active
   glycolysis^[130]26. We tested whether P. micra-induced glycolysis was
   regulated by HIF-1α. Firstly, we compared the expression of HIF-1α and
   glycolytic proteins in OSCC cells with dimethyloxalylglycine (DMOG, a
   HIF stabilizer)^[131]27, P. micra, or DMOG + P. micra. The expression
   of HIF-1α induced by P. micra at an MOI of 100 was comparable to that
   elicited by a 0.5 mM concentration of DMOG (Supplementary Fig.
   [132]7a). DMOG alone elevated the expression levels of HK2, GLUT1, and
   PFKFB3, and in combination with P. micra, the expression of these
   proteins displayed an additive upregulation effect (Fig. [133]3d and
   Supplementary Fig. [134]6b). The small interfering RNAs targeting
   HIF-1α (siHIF-1α) or inhibiting HIF-1α by LW6^[135]28 markedly reduced
   the expression of glycolytic proteins induced by P. micra, including
   HK2, GLUT1, and PFKFB3 (Fig. [136]3g and Supplementary Figs. [137]6c,
   [138]7b). This was accompanied by a reduction in glucose uptake and
   lactate production (Fig. [139]3e, f and Supplementary Fig. [140]7c).
   These results suggest that P. micra-promoted glycolysis is regulated by
   HIF-1α in OSCC cells.

   Besides, RNA-seq analysis on CAL27 cells also showed that a group of
   autophagy genes were upregulated by P. micra infection (Supplementary
   Fig. [141]8). Transmission Electron Microscope (TEM) showed a
   significant increase of autophagosomes in P. micra-exposed CAL27 and
   SCC15 cells compared to non-exposed cells (Fig. [142]3h). Western blot
   also showed an increased expression of the autophagy marker LC3B II by
   P. micra infection, whereas P62 was decreased (Fig. [143]3i and
   Supplementary Fig. [144]9a). CAL27 and SCC15 cells were transfected
   with mRFP-GFP-LC3 adenoviruses to evaluate the autophagic flux.
   Confocal fluorescence microscope showed that P. micra significantly
   enhanced the autophagic flux indicated by the increased number of
   autophagosome and autolysosome (Fig. [145]3j). These results suggest
   that P. micra can activate the autophagy of OSCC cells.

   Hypoxia/HIF-1α can induce autophagy^[146]29 and glycolysis can also
   facilitate autophagy in response to glucose
   deprivation^[147]30,[148]31. Therefore, whether P. micra-activated
   autophagy could be regulated by HIF-1α and/or glycolysis was
   investigated. DMOG increased the protein level of LC3BII while
   decreased the expression of P62 (Fig. [149]3k and Supplementary
   Fig. [150]9b). siHIF-1α and LW6 as well as 2-DG (glycolysis inhibitor)
   dramatically inhibited the P. micra-induced autophagic flux and rescued
   the expression of LC3B II and P62 in both CAL27 and SCC15 cells
   (Fig. [151]3l, m and Supplementary Figs. [152]9c–d, [153]10a–c). These
   results imply that P. micra-induced autophagy in OSCC cells is
   regulated by HIF-1α and glycolysis.

   Next, we investigated whether P. micra induced OSCC metastasis was
   mediated by HIF-1α, glycolysis and autophagy. siHIF-1α, LW6 or 2-DG
   abolished the effect of P. micra on OSCC cell migration and invasion,
   respectively (Supplementary Fig. [154]11a–d). The expression of
   N-cadherin and E-cadherin induced by P. micra were also rescued by LW6
   and 2-DG, respectively (Supplementary Fig. [155]11e). The autophagy
   inhibitor, CHQ, abolished the effect of P. micra on migration and
   invasion in both CAL27 and SCC15 cells (Supplementary Fig. [156]11f,
   g). Also, CHQ rescued the expression of E-cadherin and N-cadherin
   induced by P. micra (Supplementary Fig. [157]11h). The above results
   indicate that P. micra promotes the OSCC metastasis by activating
   HIF-1α, glycolysis and autophagy.

   To further verify the role of HIF-1α in P. micra-promoted OSCC
   metastasis in vivo, CAL27 cells, and CAL27 cells transfected by
   siHIF-1α with/without P. micra infection were implanted on the tongue
   of mice model, respectively. The results showed that siHIF-1α + P.
   micra group displayed smaller size of tumors and cLNs as well as lower
   tumor cell metastatic area in cLNs compared to the P. micra group
   (Supplementary Fig. [158]12). These results suggest that HIF-1α is an
   important mediator of P. micra-promoted OSCC metastasis.

TmpC was an effector of P. micra promoted OSCC metastasis

   As the surface and intra-cellular of CAL27 and SCC15 cells contained
   amounts of live P. micra (Supplementary Fig. [159]13a), suggest P.
   micra may exert its pathogenicity by adhering and invading OSCC cells.
   Scanning electron microscopy (SEM) and TEM confirmed the attachment and
   invasion of P. micra to OSCC cells (Fig. [160]4a, b). CFSE-labeled P.
   micra was also observed on the surface and plasma of CAL27 and SCC15
   cells by confocal microscopy (Supplementary Fig. [161]13b). These
   results suggest that P. micra may exert its pathogenic effects to OSCC
   cells by the surface proteins.

Fig. 4. P. micra interacted with OSCC cells via its surface protein TmpC.

   [162]Fig. 4
   [163]Open in a new tab

   a Representative SEM images of CAL27 and SCC15 cells after co-culture
   with P. micra (MOI = 100). The higher magnification images showed
   attachment between bacteria and the cells. b Representative TEM images
   of CAL27 and SCC15 cells after co-culture with P. micra (MOI = 100. Red
   arrows indicated intracellular P. micra and white arrows indicated
   cell-surface P. micra. c Wound-healing migration assays of OSCC cell
   stimulated with TmpC (0.05 and 0.1 mg/mL) for 24 h (n = 3 biological
   replicates per group). Scale bar: 100 μm. d Trans-well assays of OSCC
   cells pre-stimulated with TmpC (0.05 and 0.1 mg/mL) for 24 h (n = 5
   biological replicates per group). Scale bar: 100 μm. e Representative
   images of the Image-iT Green Hypoxia Reagent assay of OSCC cell
   stimulated with TmpC for 24 h. f The schematic representation of the
   insertional inactivation of TmpC in P. micra. DNA fragment containing
   ermB gene was constructed and transformed into P. micra wild type
   (PmWT) for homologous recombination. g PCR of insertion mutants of
   PmWT. Bacteria chromosomal DNA was used as template for the PCR and
   different primers indicated in (f) were used to validate ermB
   insertion. h PmΔtmpC showed decreased adhesion and invasion ability
   compared with PmWT (n = 3 biological replicates per group). i BALB/c
   nude mice were implanted with PmWT or PmΔtmpC treated CAL27 cells on
   the left lateral edge of the tongue for 3 weeks. j and k Images of
   tongue (j) and quantification of tumor volume (k, n = 8 mice per
   group). l and m Representative images of cLNs (l) and quantification of
   volume quantification of cLNs volume (m) from 8 mice per group. n and o
   Representative images of immunofluorescence for CK5/6 (n) and
   quantification of metastatic area (o, n = 8 mice per group) in cLNs.
   Scale bar: 500 μm for 2X and 100 μm for 10X. Images are representative
   of two independent experiments (a, b, e, g). Data were shown as
   mean ± SD. One-way ANOVA with Turkey’s test (c, d, k, m, o) and
   two-sided unpaired Student’s t test (h) were used to examine the
   statistical significance between groups.

   To elucidate the mechanism responsible for P. micra attachment and
   invasion to OSCC cells, we focused on P. micra surface protein TmpC,
   which was an important adhesins of P. micra. TmpC was obtained and used
   to simulate OSCC cells, which showed that TmpC significantly promoted
   the migration and invasion of OSCC cells (Fig. [164]4c, d and
   Supplementary Fig. [165]14a, d), and induced cell hypoxia
   (Fig. [166]4e, Supplementary Fig. [167]14c) in both CAL27 and SCC15
   cells. The expression of HIF-1α, HK2 and LC3B II was upregulated by
   TmpC (Supplementary Fig. [168]14d, e). TmpC significantly increased the
   mRNA level of glycolytic proteins, such as HK2, LDHA, PFKFB4
   (Supplementary Fig. [169]14f), albeit to a lesser extent compared to P.
   micra. Furthermore, HIF-1α served as a pivotal mediator in TmpC-induced
   glucose uptake and lactate production in OSCC cells (Supplementary
   Fig. [170]14g). These results indicate that TmpC is an important
   virulence effector of P. micra promoted OSCC metastasis.

   To validate the essential role of TmpC, we utilized insertional
   inactivation (Fig. [171]4f) to generate a TmpC mutant strain (PmΔtmpC,
   Fig. [172]4g). Compared with wild-type P. micra (PmWT), PmΔtmpC
   demonstrated a significant reduction in the adhesion and invasion
   abilities to OSCC cells in vitro (Fig. [173]4h) and was unable to
   elicit LM of OSCC in vivo (Fig. [174]4i–o). Together, these results
   suggest that TmpC is the necessary effector of P. micra to promote OSCC
   metastasis in vivo.

TmpC interacted with CKAP4 to promote OSCC metastasis

   The receptor of TmpC on OSCC cells was analyzed by His-tagged TmpC and
   HLPC-MS from both CAL27 and SCC15 cells (Fig. [175]5a). CKAP4, a type
   II transmembrane protein, was identified as one of the TmpC receptors
   by overlapping all candidate proteins detected in both cell
   lines (Supplementary Table [176]1). Immunoprecipitation analysis showed
   CKAP4 could be pulled down by His-TmpC in both CAL27 and SCC15 cells
   (Fig. [177]5b). Confocal fluorescence microscopy showed that CKAP4 was
   co-localized with CFSE-labeled P. micra (Supplementary Fig. [178]15a).
   These results indicate that CKAP4 is a receptor of TmpC on OSCC cells.

Fig. 5. P. micra surface TmpC bound to CKAP4 on OSCC cells.

   [179]Fig. 5
   [180]Open in a new tab

   a OSCC cell membrane protein was incubated with His-TmpC together with
   His magnetic beads for His-pull-down assay and mass spectrometry
   analysis. b Immunoprecipitation showing His-TmpC directly bound to
   CKAP4 in OSCC cells. c CKAP4Ab suppressed the attachment and invasion
   of P. micra in OSCC cells (n = 3 biological replicates in each group).
   d and e Representative images (d) and quantitation (e) of cell
   migration and invasion of OSCC cells preincubated with CKAP4Ab and
   treated by PmWT and Pm∆tmpC. The indicated migrated and invaded cells
   were quantified in five randomly selected fields. Scale bars: 100 μm. f
   Western blot of E-cadherin (E-cad), N-cadherin (N-cad), HIF-1α, GLUT1,
   HK2, PFKFB3, P62 and LC3B in OSCC cells preincubated with CKAP4Ab and
   treated by PmWT and Pm∆tmpC. The samples derive from the same
   experiment but different gels for N-cadherin and GLUT1, another for
   HIF-1α and ACTB, another for HK2 and PFKFB3, another for E-cadherin,
   P62 and LC3B were processed in parallel. g Plausible binding mode of
   CKAP4 (green) with TmpC (pink) from molecular docking study. h
   Schematic illustration of the full-length CKAP4 protein and the
   truncated forms of CKAP4 tagged with GST. Red segments indicated
   predicted binding sites. i and j SPR binding sensorgrams and affinity
   fit curves for the interaction of TmpC with CKAP4 in vitro. k Western
   blot of proteins expression in shCKAP4 OSCC cells transfected with
   full-length CKAP4 (CKAP4^WT) or CKAP4 mutant lacking the TmpC-binding
   site (CKAP4^Δ991-1350) and stimulated by P. micra. The samples derive
   from the same experiment but different gels for N-cadherin and GLUT1,
   another for HIF-1α and ACTB, another for HK2 and PFKFB3, another for
   E-cadherin, P62 and LC3B, another for Flag were processed in parallel.
   l and m Representative images (l) and quantitation (m) of cell
   migration and invasion of OSCC cells transfected with CKAP4^WT and
   CKAP4^Δ991-1350 and stimulated by P. micra (n = 5 biological replicates
   in each group). Scale bars: 100 μm. Data were shown as mean ± SD. Blots
   are representative of at least two biological replicate experiments
   with similar results (f, k). One-way ANOVA with Turkey’s test (e, m)
   and two-sided unpaired Student’s t test (c) was used to examine the
   statistical significance between groups.

   Next, we investigated the role of CKAP4 in P. micra/TmpC promoted OSCC
   metastasis. Knockdown of CKAP4 by siRNAs (siCKAP4) or blockade of CKAP4
   with neutralizing antibodies (CKAP4Ab) markedly inhibited the
   attachment and invasion of P. micra to OSCC cells (Fig. [181]5c and
   Supplementary Fig. [182]15b). siCKAP4 attenuated P. micra/TmpC-promoted
   migration and invasion as well as glucose uptake, lactate output and
   autophagic flux in OSCC cells (Supplementary Figs. [183]16 and
   [184]17). Preincubation of OSCC cells with CKAP4Ab eliminated the
   stimulatory effects of PmWT on the migration and invasion of OSCC
   cells, whereas no effect was determined on PmΔtmpC (Fig. [185]5d–e). In
   vivo experiment showed siCKAP4+P. micra group displayed smaller size of
   tumors and cLNs as well as lower tumor cell metastatic area in cLNs
   compared to the P. micra group (Supplementary Fig. [186]12). Moreover,
   CKAP4Ab and siCKAP4 inhibited PmWT or TmpC-promoted expression of EMT
   markers (E-cadherin and N-cadherin), HIF-1α, glycolytic proteins
   (GLUT1, HK2 and PFKFB3) and autophagy proteins (P62 and LC3 II) in OSCC
   cells, whereas it had no effect on the PmΔtmpC (Fig. [187]5f and
   Supplementary Fig. [188]18a–c). PmΔtmpC could not affect OSCC cells
   migration and invasion as well as the expression of EMT markers,
   HIF-1α, glycolytic proteins and autophagy proteins (Fig. [189]5d–f and
   Supplementary Fig. [190]18a). These results suggest that CKAP4 is
   essential for P. micra/TmpC promoted OSCC metastasis.

   To reveal the domain of CKAP4 bound to TmpC, GST-tagged full length and
   truncated forms of CKAP4 were constructed according to the predicted
   domain architecture bound with TmpC (Fig. [191]5g, h and Supplementary
   Fig. [192]19) and subjected to Surface Plasmon Resonance (SPR) assays.
   The results revealed that aa 331-475 of CKAP4 containing two sites
   binds with TmpC (Fig. [193]5i, j). Furthermore, the OSCC cells with
   stable silencing of CKAP4 (shCKAP4) were constructed (Supplementary
   Fig. [194]18d) and then was transfected with full-length CKAP4
   (CKAP4^WT) or CKAP4 mutant lacking the TmpC-binding site
   (CKAP4^Δ991-1350) for rescue experiments. The expression of E-cadherin,
   N-cadherin, HIF-1α, HK2, PFKFB3, and LC3B II were effectively rescued
   by CKAP4^WT in shCKAP4 OSCC cells, but not by CKAP4^Δ991-1350
   (Fig. [195]5k and Supplementary Fig. [196]20). CKAP4^WT also enhanced
   the migration and invasion capability of shCKAP4 cells stimulated by P.
   micra or TmpC, while not CKAP4^Δ991-1350 (Fig. [197]5l, m).

TmpC-CKAP4 stabilized HIF-1α by RanBP2 and induced autophagy by NBR1

   The gene expression of HIF-1α was unchanged in OSCC cells with P.
   micra/TmpC stimulation (Supplementary Fig. [198]21a), while the protein
   level of HIF-1α was markedly increased (Fig. [199]3), suggesting that
   P. micra/TmpC may promote the translation or stability of HIF-1α. We
   further deciphered the proteins involved in TmpC-CKAP4-mediated HIF-1α
   translation or stability. RanBP2, a SUMO (small ubiquitin-like
   modifier) E3 ligase^[200]32, was identified from the overlap of the
   candidate proteins interacted with CKAP4 by HPLC-MS in CAL27 cells with
   P. micra/TmpC stimulation (Fig. [201]6a, Supplementary Table [202]2).
   Co-immunoprecipitation (Co-IP) revealed a direct interaction between
   RanBP2 and CKAP4 (Fig. [203]6b). RanBP2 may facilitate the SUMOylation
   process of HIF-1α, thereby enhancing the stability of the HIF-1α
   protein^[204]33. Co-IP assay showed HIF-1α could bind with SUMO1 in
   both CAL27 and SCC15 cells under the stimulation of P. micra/TmpC
   (Fig. [205]6c). To assess whether SUMOylation primarily accounts for
   HIF-1α protein stability, the SUMOylation enzymatic cascade was
   inhibited using TAK981, a specific inhibitor^[206]34. The results
   showed that TAK981 significantly decreased the abundance of HIF-1α and
   HK2 in CAL27/SCC15 cells under the stimulation of P. micra/TmpC
   (Fig. [207]6d and Supplementary Fig. [208]21b). Moreover, siRanBP2
   suppressed TmpC-induced stability of HIF-1α (Fig. [209]6e and
   Supplementary Fig. [210]21c). siRanBP2 or TAK981 attenuated P.
   micra-induced up-regulation of LC3BII and down-regulation of P62
   (Supplementary Fig. [211]21d). These results suggest that
   RanBP2-dependent SUMOylation is involved in the stability of HIF-1α.

Fig. 6. TmpC-CKAP4 stabilized HIF-1α by binding RanBP2 and induced autophagy
by binding NBR1.

   [212]Fig. 6
   [213]Open in a new tab

   a Immunoprecipitation of CKAP4 followed by HPLC-MS identified
   CKAP4-bound proteins. b Co-IP and immunoblotting detection showed the
   interaction between RanBP2 and CKAP4. c Co-IP and immunoblotting
   detection showed the interaction between HIF-1α and SUMO1. d TAK981
   decreased the expression of HIF-1α and HK2. The samples derive from the
   same experiment but different gels for HIF-1α and ACTB, another for HK2
   were processed in parallel. e siRanBP2 decreased the expression of
   HIF-1α. f Co-IP and immunoblotting detection showed the interaction
   between NBR1 and CKAP4. g Co-IP and immunoblotting detection showed the
   binding of P62 with CKAP4 and NBR1. h siNBR1 increased P62 and
   decreased LC3B II. i siNBR1 decreased the P62 bodies (n = 5 cells per
   group). The samples derive from the same experiment but different gels
   for NBR1 and ACTB, another for P62 and LC3B were processed in parallel.
   j–l Representative immunofluorescence images (j) and quantification (k,
   l, n = 15 cells per group) showing siCKAP4 inhibited the
   co-localization of NBR1 with P62 bodies. The ratio quantified by
   comparing the number of NBR1-positive P62 bodies to all P62 bodies per
   cell. Scale bar: 10 μm. m A schematic diagram illustrating the
   mechanisms of P. micra TmpC promoted OSCC metastasis, created in
   figdraw.com. Data were shown as mean ± SD. Blots are representative of
   at least two biological replicate experiments with similar results
   (b–h). One-way ANOVA with Turkey’s test was used to examine the
   statistical significance between groups.

   Meanwhile, HPLC-MS data also showed the autophagy proteins NBR1 and
   SQSTM1(P62) were interacted with CKAP4 in CAL27 cells under the
   stimulation of P. micra/TmpC (Fig. [214]6a). The interaction of NBR1
   and CKAP4 was confirmed by Co-IP in CAL27 and SCC15 cells with P.
   micra/TmpC stimulation, respectively (Fig. [215]6f). Different from
   NBR1, P62 could bind with CKAP4 in CAL27 or SCC15 cells with/without P.
   micra/TmpC stimulation (Fig. [216]6g). The knockdown of NBR1 by siRNA
   (siNBR1) inhibited TmpC-induced autophagy, as evidenced by the
   restoration of P62 and LC3B II expressing levels (Fig. [217]6h and
   Supplementary Fig. [218]21e). These results indicate that NBR1 is a
   main mediator involved in TmpC-CKAP4 induced autophagy.

   The mechanism by which NBR1 regulates TmpC-CKAP4 induced autophagy was
   further examined. As a P62-interacting protein, NBR1 enhanced the
   formation of P62 bodies^[219]35. Confocal fluorescence microscopy
   revealed that TmpC stimulation led to a dramatically increase in the
   quantity of P62 bodies in both SCC15 and CAL27 cells, whereas siNBR1
   resulted in a decrease in the number of P62 bodies (Fig. [220]6i and
   Supplementary Fig. [221]21f). Moreover, siCKAP4 increased the protein
   levels of P62 and NBR1 in both SCC15 and CAL27 cells under the
   stimulation of TmpC (Supplementary Figs. [222]18b and [223]21g),
   indicating that CKAP4 contributed to the TmpC-induced degradation of
   P62 and NBR1. Similarly, siCKAP4 inhibited TmpC-induced P62 bodies and
   the co-localization of P62 and NBR1 (Fig. [224]6j–l). These results
   suggest that, with the stimulation of TmpC, CKAP4 promotes NBR1 binding
   to P62, facilitating the formation of P62 bodies, thereby induce OSCC
   cell autophagy (Fig. [225]6m).

Blockade of CKAP4 attenuated P. micra/TmpC-promoted OSCC metastasis

   To further validate the role of CKAP4 in P. micra/TmpC-promoted OSCC
   metastasis, the small chemicals targeting CKAP4 were selected through
   molecular docking-based virtual screening and identified 9454 compounds
   with promising scores (Supplementary Table [226]3). Considering various
   factors, four compounds including imidurea, masitinib, fostamatinib,
   and quizartinib, were chosen for the following experiments
   (Fig. [227]7a). Concentrations of these chemicals that had no effect on
   OSCC cell viability were selected (Supplementary Fig. [228]22a),
   subsequently enabling the performance of bacterial attachment/invasion
   assays. Three chemicals (imidurea, masitinib, and fostamatinib),
   particularly masitinib, effectively suppressed the attachment and
   invasion of P. micra. However, their efficacy was negated in the
   presence of siCKAP4 (Fig. [229]7b-c). Masitinib effectively inhibited
   the motility of OSCC cells promoted by P. micra through a scratching
   test (Fig. [230]7d and Supplementary Fig. [231]22b) and influenced the
   expression of EMT markers (Supplementary Fig. [232]22c).

Fig. 7. Blockade of CKAP4 attenuated P. micra/TmpC-promoted OSCC metastasis.

   [233]Fig. 7
   [234]Open in a new tab

   a Plausible binding mode of CKAP4 with imidurea, masitinib,
   fostamatinib, and quizartinib. b and c The impact of imidurea,
   masitinib, fostamatinib, and quizartinib on P. micra attachment (b) and
   invasion (c) to the OSCC cells (n = 3 biological replicates per group).
   d Wound-healing migration assays showed the impact of masitinib on P.
   micra-promoted OSCC motility. The average rate quantified in five
   randomly selected fields. Scale bar: 100 μm. e Schematic diagram
   showing OSCC orthotopic xenograft therapy model. f and g Quantification
   of tumor volume (f) and cLNs volume (g) from 7 mice per group. h
   Representative luciferase images of tongue and cLNs. i and (j)
   Quantification of bioluminescence in tumor (i) and cLNs (j) from 5 mice
   per group. k Representative immunofluorescence images of CK5/6 in cLNs.
   Scale bar: 100 μm. l Quantification of metastatic area in cLNs (n = 7
   mice per group). m Western blot of E-cadherin, N-cadherin, HIF-1α, HK2,
   PFKFB3 and P62 in OSCC orthotopic xenograft model (mixed 3 mice per
   group, blots are representative of at least two biological replicate
   experiments with similar results. The samples derive from the same
   experiment but different gels for N-cadherin and GLUT1, another for HK2
   and ACTB, another for HIF-1α and PFKFB3, another for E-cadherin and P62
   were processed in parallel). n and o Representative images (n) and
   Spearman correlation (o) of P. micra abundance and CKAP4 expression in
   OSCC tissue. Scale bar: 100 μm. p The expression of CKAP4 in TCGA-HNSCC
   (Normal, n = 30 independent samples; OSCC, n = 262 independent samples)
   and GEO [235]GSE42743 dataset (Normal, n = 29 independent samples;
   OSCC, n = 74 independent samples). The lower and upper hinges of the
   box represent the 25th and 75th percentiles, and the whiskers extend to
   the minimum and maximum values within 1.5 times the interquartile
   range. The median value is depicted by the line within the box. q
   Kaplan-Meier survival analysis and the log-rank test of overall
   survival rate in GEO [236]GSE42743 and [237]GSE41613 dataset. Data were
   shown as mean ± SD. One-way ANOVA with Turkey’s test (b, c, d, f, g, i,
   j, l), two-sided spearman test (o) and two-sided Wilcox test (p) were
   used to examine the statistical significance between groups.

   Next, we established an OSCC orthotopic xenograft therapy model to
   investigate the therapeutic effect of masitinib targeting CKAP4 in
   vivo, with CKAP4Ab as a positive control (Fig. [238]7e). Consequently,
   P. micra increased tumor and lymph node size, bioluminescence in tumors
   and cLNs, and promoted OSCC LM, while CKAP4Ab or masitinib treatment
   dramatically inhibited the P. micra-promoted OSCC metastasis
   (Fig. [239]7f–l). The combination therapy of masitinib with CKAP4Ab
   (Pm/CKAP4Ab/ masitinib group) failed to yield superior therapeutic
   outcomes in comparison to either CKAP4Ab or masitinib monotherapy
   (Fig. [240]7f–l). Additionally, CKAP4Ab or masitinib abolished P.
   micra-induced expression of EMT markers (E-cadherin and N-cadherin),
   HIF-1α, glycolytic proteins (GLUT1, HK2 and PFKFB3) and autophagy
   protein (P62) in mice OSCC tissues (Fig. [241]7m and Supplementary
   Fig. [242]23). These results offer additional proof that CKAP4 serves
   as the primary mediator through which P. micra/TmpC exerts its
   pathogenic influence on OSCC.

   We further explored the relationship between P. micra and CKAP4 in OSCC
   patients by FISH and IF assay. Spearman correlation showed that the
   abundances of P. micra was positively correlated with the expression of
   CKAP4 in OSCC tissues (Fig. [243]7n, o). RNA-seq data from the
   TCGA-HNSCC cohort and GEO database ([244]GSE42743)^[245]36 concordantly
   showed that CKAP4 was markedly higher in OSCC tissues compared to
   adjacent normal tissues (Fig. [246]7p). The survival rate analysis of
   OSCC patients from the [247]GSE42743 and [248]GSE41613 dataset^[249]37
   showed that the overall survival (OS) rate was significantly lower for
   CKAP4-overexpressing OSCC patients (Fig. [250]7q). Moreover, CKAP4 was
   positively associated with the expression of HIF-1α and N-cadherin
   while negatively associated with E-cadherin in OSCC patients
   (Supplementary Fig. [251]24). These results suggest that CKAP4 serves
   as a promising target for OSCC therapy.

Discussion

   OSCC exhibited a significant association with the presence of various
   oral bacteria, such as P. gingivalis, F. nucleatum, and Streptococcus
   specie^[252]38. However, previous study rarely focused on the role of
   P. micra in oral cancer. This study demonstrated that P. micra is an
   important pathogen of OSCC.

   A study showed P. micra can translocate from the subgingival cavity to
   the gut and promote colorectal tumorigenesis^[253]39. This process was
   facilitated by the induction of colonocyte proliferation through the
   activation of the Ras/ERK/c-Fos signaling pathway^[254]18. In our
   study, the abundance of P. micra was higher in OSCC tissues from
   patients with LM than in those from patients without LM. In vivo and in
   vitro study showed P. micra slightly promoted OSCC proliferation while
   remarkably enhanced OSCC metastatic potential and facilitated LM.
   Therefore, we focused on how P. micra facilitated OSCC metastasis.

   The intra tumor tissue provided feasible conditions for the invasion of
   oral pathogenic bacteria, which in turn promoted tumor progress through
   various ways^[255]40. For example, a study on CRC showed P. micra
   induce Th17 immune cells infiltration and expression of Th17
   cells-secreted cytokines in the colon, thereby promoting the
   establishment of a pro-inflammatory microenvironment to favor
   colorectal tumorigenesis^[256]19. The scRNA-seq analysis in this study
   also revealed that P. micra alters the OSCC TME to facilitate
   metastasis by elevating the proportions of CAFs, particularly myCAFs
   and apCAFs, as well as monocytes, while also activating immune
   regulatory pathways in these cells and OSCC cells.

   Oral pathogenic bacteria can colonize multiple cancer types through
   specific adhesins. We elucidated TmpC was an essential adhesin and
   virulence effector of P. micra on OSCC cells to facilitate OSCC
   metastasis. Consistently, studies showed that TmpC is essential for S.
   pneumoniae-related lung infection^[257]41 and for gastric
   tumorigenesis-promoting effect of S. anginosus^[258]42. However, the
   extent of HIF-1α-dependent gene induction by TmpC alone was weaker
   compared to that observed with wild-type P. micra. This observation
   hinted that while TmpC was indeed a prerequisite for this process,
   additional factors or mechanisms associated with P. micra may be
   required to fully activate HIF-1α signaling. These factors could
   include other bacterial components, secreted metabolites, or
   synergistic interactions with host pathways. Next, we proved that TmpC
   binds to CKAP4 in OSCC cells to mediate the attachment and invasion of
   P. micra to OSCC and promote OSCC metastasis. Importantly, our
   observation of a positive correlation between CKAP4 expression levels
   and P. micra abundance suggested a potential host-microbe interaction
   that may contribute to the selective enrichment of P. micra in OSCC.
   CKAP4 was initially identified as a protein resident within the
   endoplasmic reticulum (ER)^[259]43 which also expressed on the cell
   membrane of various cells and was an important membrane
   receptor^[260]44. Notably, CKAP4 functioned as a receptor for Dickkopf1
   (DKK1), and the DKK1/CKAP4 signaling cascade facilitated the
   proliferation of both normal and cancer cells via the PI3K/AKT
   signaling pathway^[261]44. A recent study has revealed that CKAP4 acted
   as a crucial membrane receptor for the SARS-CoV-2 spike
   protein^[262]45. Notably, we demonstrated that the binding site of TmpC
   to CKAP4 exhibited a high similarity to the active domains of
   SARS-CoV-2 spike protein (Supplementary Fig. [263]25). This suggests
   that CKAP4 possesses considerable potential in mediating pathogen
   infection within human body.

   Mechanically, the combination of CKAP4 and TmpC activates the
   intracellular signaling cascades in OSCC cells. Specifically, we found
   that TmpC-CKAP4 could promote SUMOylation of HIF-1α and stabilize
   HIF-1α protein^[264]33 by SUMO E3 ligase RanBP2^[265]32. The
   accumulated HIF-1α in turn activated cellular glycolysis, which
   facilitated OSCC EMT and metastasis. However, the mechanism by which
   TmpC alone induced hypoxia was currently unknown and warrants further
   study. Furthermore, CKAP4 interacted with the autophagy protein
   NBR1^[266]46, which promoted the formation of P62 bodies and P.
   micra-induced autophagy. Meanwhile, the autophagy activated by P.
   micra/TmpC was enhanced by HIF-1α and glycolysis (Fig. [267]6m).

   The small molecule chemical of Masitinib, targeting CKAP4, was employed
   to verify whether the pathogenic effects of P. micra on OSCC were
   mainly mediated by CKAP4. Similar to the CKAP4Ab, masitinib abrogated
   the attachment, invasion, and tumor-promoting effects of P. micra on
   OSCC cells in vitro and inhibited OSCC LM in vivo. In line with our
   results, masitinib has been investigated in the treatment of pancreatic
   cancer^[268]47. Besides, anti-CKAP4 antibody also has shown antitumor
   activity in pancreatic cancer^[269]48, esophageal cancer^[270]49, and
   lung cancers^[271]50. Collectively, these results exhibited the
   potential application of masitinib as a therapeutic strategy for OSCC
   LM, especially with the present of P. micra.

   Clinically, CKAP4 was higher in OSCC patients and associated with poor
   overall survival as well as positively correlated with HIF-1α and EMT
   in OSCC patients. Our scRNA-seq data analysis revealed that CKAP4 was
   expressed across all subpopulations of OSCC cells, with particularly
   high expression levels observed in clusters 1, 2, and 3. Of the OSCC
   TME, CKAP4 exhibited high level of expression in NFs, mCAFs, and
   monocytes-macrophages in contrast to other subpopulations. Also, CKAP4
   has been characterized as pro-tumor molecule that plays a regulatory
   role in the progression of multiple cancer types, such as lung
   cancer^[272]51, esophageal squamous cell carcinoma^[273]52 and
   hepatocellular carcinoma^[274]53, and regulated the signaling pathways
   including PI3K/AKT and MAPK1/3. This suggests that CKAP4 may play an
   important role in the progression and prognosis of OSCC.

   Collectively, this study showed P. micra was an important intratumor
   bacterium promoting OSCC LM. P. micra surface protein, TmpC, binds to
   CKAP4 on OSCC cells, thereby triggering P. micra attachment, invasion,
   and the subsequent activated HIF-1α and autophagy signaling by
   CKAP4-RanBP2 and CKAP4-NBR1 interaction, respectively (Fig. [275]6m).
   The blockade of CKAP4 with masitinib or CKAP4ab has potential
   therapeutic effects on OSCC LM.

   Our study has the following limitations. While we have clarified the
   interaction between P. micra and OSCC cells, future research utilizing
   appropriate models should delve into the effect of P. micra on
   inflammatory factors and the intercellular communication they
   facilitate among OSCC cells and immune cells. Furthermore, additional
   studies are essential to conclusively determine whether the impacts of
   P. micra on CAFs and TAMs are CKAP4-dependent. Moreover, the mechanisms
   underlying the enrichment of P. micra in OSCC remain elusive and
   necessitate further exploration.

Methods

Ethics

   This research complies with all relevant ethical regulations approved
   by the Ethics Committee of the School and Hospital of Stomatology,
   Cheeloo College of Medicine, Shandong University.

Cell lines

   Human OSCC cell lines CAL27 and SCC15 were maintained in DMEM and
   DMEM/F-12 (Shanghai BasalMedia Technologies), respectively,
   supplemented with 1% penicillin-streptomycin (Gibco, Washington, USA)
   and 10% fetal bovine serum (FBS), henceforth referred to as DMEM
   complete media. Pro. Zhiquan Huang (Sun Yat-Sen University) gifted
   CAL-27 luciferase-labeled cells (Luc-CAL27) cultured at the same
   conditions as CAL27 cells.

Bacteria

   P. micra ATCC33270 was purchased from the Chinese Academy of Sciences.
   P. micra was maintained in SchaEdler anaerobic broth (BD Biosciences,
   New Jersey, USA) in an anaerobic tube (80% N2, 10% CO[2] and 10% H[2])
   and incubated for 3-4 days at 37 °C.

Mouse models

   BALB/c nude mice (male, 5-6 weeks) were purchased from the Vital River
   Laboratory Animal Technology (Beijing, China) and bred at Shandong
   University Model Animal Center. The stable supply and immunodeficiency
   characteristics of male homozygous nude mice make them the preferred
   choice for experiments. The mice were randomly assigned to different
   groups for the experiments, and no animals were excluded from the
   analyzes. After a week of adjustable feeding, 5 × 10^6 CAL-27
   luciferase-labeled cells with/without siRNA mixed with P. micra
   (MOI = 1) or PBS in 50 μL of Matrigel and DMEM were implanted into the
   left tongue edge of BALB/c nude mice. After 3 weeks, cervical LNs and
   tongue tumors were gathered for imaging, immunofluorescence,
   histopathology and size measurement to measure the proliferation of
   tumor cells, LM, and EMT. Tumor and LNs volume (mm^3) = Width
   (mm)^2 × length (mm)/2^[276]54. For drug treatments in vivo, the tongue
   mucosa of each mouse was injected with 5 × 10^6 Luc-CAL27 cells. After
   one week, the mice were randomly assigned to seven groups. Then, they
   received the following treatments: (1) IgG (50 μg, i.p., twice per
   week); (2) CKAP4Ab (Santa, 50 μg, i.p., twice per week); (3) Masitinib
   (30 mg/kg, i.p., daily); (4) P. micra (5 × 10^5, i.t., weekly); (5) P.
   micra + CKAP4Ab; (6) P. micra+Masitinib and (7) P.
   micra + CKAP4Ab+Masitinib. After 2 weeks, cervical LNs and tongue
   tumors were gathered for imaging, immunofluorescence, histopathology
   and size measurement to measure the proliferation of tumor cells, LM,
   and EMT.

   According to the guidelines (GB/T 35892-2018), the ethics committee
   specified that the maximal tumor burden is no more than 10% of the body
   weight of animals and the average diameter is less than 20 mm. During
   the experiment, the tumor sizes of the mice complied with the
   regulations. All animal protocols were reviewed and approved by the
   Ethics Committee of the School and Hospital of Stomatology, Cheeloo
   College of Medicine, Shandong University (approval no. 20230105). This
   study adhered to the guidelines outlined in the National Institutes of
   Health Guide for the Care and Use of Laboratory Animals.

Tissue microarray and fluorescence in situ hybridization (FISH)

   Human OSCC tissue microarray (HN054Oc01) was purchased from Zhongke
   Guanghua Intelligent Biotechnology Co., Ltd. (Xi’an, China,
   [277]https://www.bioaitech.com/). Use of clinical samples was approved
   by the Ethical Committee and Institutional Review Board of the School
   and Hospital of Stomatology, Cheeloo College of Medicine, Shandong
   University in compliance with ethical standards and patient
   confidentiality (approval no. 20241339). Informed consent was obtained
   from all the patients regarding the use of tissue samples. The
   participants ranged in age from 25 to 81 years, of whom 91 % were male.
   The basic characteristics of this microarray were provided in
   Supplementary Table [278]4. Due to the limited samples available, sex
   and gender were not considered in the study design.

   A FITC-conjugated P. micra probe, possessing the sequence
   5’-CTGAGCGTCAGTAAAAGTCC-3’, was used for detecting the colonization of
   P. micra in tissue microarray. After rehydration and deparaffinization,
   specimens were treated with Proteinase K and 0.2 N HCl for 10 minutes
   each. After incubation in a blocking buffer for 2 h at 55 °C, the probe
   (diluted 1:50 in 35% preheated hybridization buffer for 3 minutes at
   88 °C) was added and hybridized overnight at 42 °C in a humid, dark
   chamber. The specimens were then washed in a Tris-HCl buffer (20 mM, pH
   7.2) containing 40 mM NaCl and mounted using a DAPI-antifade solution.
   The resulting images were captured using Pannoramic 250FLASH. The
   positive fluorescent signal intensity and area proportion was
   calculated utilizing ImageJ v1.54 software.

Migration and invasion assay

   Trans-well chambers (Corning, 3422) were applied to assess cell
   migration and invasion. OSCC cells treated with P. micra (Indicated MOI
   or MOI = 100) or stimulated with TmpC (Indicated concentration or
   0.1 mg/mL) for 24 h in 6-well plate. In some experiments, the cells
   were transfected with siRNA 24 h prior to infection. For migration
   assay, 5 × 10^4 cells were transplanted along with P. micra
   simultaneously from a 6-well plate into the upper chamber in 200 μl of
   serum-free medium, while the lower chamber was full of 600 µL medium
   supplemented with 15% FBS. For invasion assay, the upper chamber was
   pre-coated with 25 µL of Matrigel (BD, 3422356234). After 24-h
   incubation, the invaded or migrated cells were fixed and stained.
   Random images were taken under a light microscope, and three
   independent replicates were analyzed for consistency.

Wound healing assay

   The back of each 6-well plate was marked for a uniform position before
   the experiment. OSCC cells (1 × 10^6 per well) were grown until
   confluence and damaged with a pipette tip. Cells from each well were
   then co-incubated with P. micra (Indicated MOI or MOI = 100) or
   stimulated with TmpC (Indicated concentration or 0.1 mg/mL)
   simultaneously. Cell images were taken after 24 h. The wound size of
   each well was calculated as the distance between the edges.

Western blots

   Mice fresh OSCC tumor tissues or cultured OSCC cells underwent lysis in
   RIPA buffer containing phosphatase and protease inhibitors. Proteins of
   equal mass were loaded onto an SDS-PAGE gel (EpiZyme, China) and
   transferred to a PVDF membrane (Millipore, Billerica, CA, USA) for
   blotting with specific antibodies. Among, CKAP4, E-cadherin,
   N-cadherin, GLUT1, HK2, et al. were purchased from ProteinTech (Wuhan,
   China), HIF-1α, Flag Tag, Goat anti-Rabbit IgG, and Goat anti-Rabbit
   IgG were purchased from Zen Bio (Chengdu, China), RanBP-2 was purchased
   from Santa (USA), LC3B was purchased from Cell Signaling Technology
   (USA). All antibodies are detailed in Supplementary Table [279]5.
   Protein bands were visualized using the Amersham Imager 680 (GE,
   America). Samples derive from the same experiment and that blots were
   processed in parallel. Images have been cropped for presentation. All
   uncropped images for all blots are provided as source data.

Bacteria screening from transcriptome data

   The transcriptome data were downloaded from the GEO database under the
   accession number [280]GSE227919. The clean sequencing data in fastq
   format were aligned to the reference genome GRCh38 via bowtie2 software
   (version 2-2.5.1). Quantitative analysis of the mapped data was
   performed utilizing the featurecounts method (subread 2.0.6) to obtain
   the mRNA expression. The unmapped data were handled with kraken2
   software to align the Kraken microbes standardDB. Bracken software was
   then used to organize the output of kraken2. Ultimately, we got both
   mRNA expression and bacteria abundance in each sample.

Real-time quantitative polymerase chain reaction (RT-qPCR)

   Cellular RNA was extracted using TRIzol reagent (CWBIO, Beijing, China)
   and converted to cDNA via the HiFiScript cDNA Synthesis kit (Vazyme).
   Then, qRT-PCR was conducted with UltraSYBR Mixture (CWBIO) to quantify
   relative mRNA levels, normalized against β-actin using the 2^(-ΔΔct)
   method. The primer sequences used in this process are detailed in
   Supplementary Table [281]5.

Single-cell RNA sequencing (scRNA-seq) and data analysis

   Fresh OSCC tissues were retrieved from tongue orthotopic xenograft
   mice, and three repeats were prepared for each group. Samples were then
   sent to Singleron Company for subsequent scRNA-seq, and sequencing data
   was obtained through the pipeline of Singleron Company. Single cells
   were captured using the SCOPE-chip microfluidic technology, and
   millions of barcoding beads carrying unique cell barcodes were added to
   the chip micropore, ensuring that only one barcoding bead fell into
   each micropore. After cell lysis, barcoding beads with unique cell
   labels (Barcode) and molecular labels (UMI) capture mRNA by binding to
   the poly (A) tail on mRNA and then label the mRNA and cells. Barcoding
   beads were collected and mRNA was reversely transcribed into cDNA
   before amplification. A sequencing library suitable for the Illumina
   sequencing platform was constructed by steps such as fragmentation and
   splicing cDNA. The scRNA-seq data were first mapped with the human
   genome (GRCh38), and the unmapped sequences were mapped with the mouse
   genome (mm10).

   The R package Seurat v5.0.1 was applied to analyze these data. In the
   quality control phase, any cell complying with one of the criteria
   below was removed: having more than 20% transcripts of mitochondrial
   genes, expressing fewer than 200 genes, or having fewer than 500 unique
   molecular identifiers (UMIs). The doublets were predicted and removed
   using the R package DoubletFinder v2.0.3. Data were integrated using R
   package harmony v1.2.0. The non-negative Matrix Factorization (NMF)
   method was used for dimensionality reduction to obtain an NMFfit
   object, and then the RunUMAP function was performed to cluster and the
   results were displayed using the DimPlot function. The function of
   “FindAllMarkers” in the Seurat package was used to analyze the changes
   in gene expression between clusters, and the highly expressed genes
   within the group were shown in heatmaps or volcano plots. Further
   functional enrichment analysis was performed on the findings of
   differential gene expression analysis using the clusterProfiler v4.10.0
   package, and the significant enrichment pathways of each cell type were
   displayed using the dotplot. Moreover, the irGSEA package was selected
   to calculate enrichment scores under Hallmark gene sets using the
   AUCell method, and visualize the enrichment results of clusters using
   heatmap. In addition, GSVA demonstrated differences between the
   experimental group and the control group based on the hallmark and KEGG
   pathways. To characterize the tissue distribution of meta clusters
   (cell type), odds ratio (OR) values were calculated to evaluate the
   tissue preference of single cells. Intercellular communication network
   analysis was performed using CellChat v1.6.1 and differences in cell
   communication between the experimental group and the control group were
   compared.

RNA sequencing and analysis

   Total RNA was extracted from CAL27 cells stimulated by P. micra
   (MOI = 100) or PBS, and three repeats were prepared for each group.
   Then, samples were sent to Novogene Tech for subsequent sequencing.
   Quality control was performed on raw data using fastp (0.23.4). The
   clean data obtained through quality control were aligned to the
   reference genome GRCh38 via bowtie2 (2-2.5.1). Quantitative analysis
   was performed using feature Counts (subread 2.0.6) to obtain the
   expression profile for each sample. Differential analysis was performed
   using the limma (version 3.58.1) method. The differentially expressed
   genes were screened by p value < 0.01 and foldchange > 1.5, and plotted
   in the volcano and heatmap. Enrichment of GO and KEGG pathways was
   implemented using filtered differentially expressed genes, and
   significant results were shown in bubble plots with p < 0.05. On the
   other hand, GSEA analysis was performed using all gene sets to compare
   the pathway changes between the experimental group and the control
   group. The significant result (p < 0.05) was shown using GseaVis
   package (version 0.0.5).

Immunofluorescence (IF)

   OSCC cells were fixed with 4% paraformaldehyde for 20 min after they
   were seeded on chamber slides. The slides were incubated with 0.2%
   TritonX-100 (Beyotime, China) for 15 min and blocked with 3% BSA for
   30 min. After three times washing by PBS, cells were incubated with
   primary antibodies in a wet chamber overnight at 4°C. The next day, the
   slides were washed by PBST three times and incubated with secondary
   antibodies for 1 h at room temperature. These cells were then stained
   with DAPI.

   For detection of P. micra and CKAP4 colocalization, P. micra was
   labeled with CFSE (20 μM) for 0.5 h before co-culture with OSCC cells
   for another 2 h. Cells were fixed with 4% paraformaldehyde for 15 min
   and incubated with the anti-CKAP4 antibody (1: 100 dilution) at 4 °C
   overnight. After washing with PBS, secondary antibody incubation was
   performed at room temperature for 1 h using Alexa Fluor 594-conjugated
   goat anti-rabbit IgG (1:500 dilution). Images were acquired using a
   confocal microscope.

Autophagic flux detection

   CAL27 and SCC15 cells were infected with mRFP-GFP-LC3 adenoviruses
   following the manufacturer’s protocol. After adenoviruses infection
   24 h, the cells were stimulated with P. micra for another 24 h or
   transfected with siRNA 24 h prior to infection and then fixed with 4%
   paraformaldehyde for 30 min. Nuclei were re-stained with DAPI for
   5 min. All procedures were performed in the dark at RT. Finally, a
   confocal laser scanning microscope was used to image the cells in
   confocal dishes. The autophagic flux is enhanced when there is an
   increase in both yellow and red dots within cells. Conversely, the
   autophagic flux is deemed to be obstructed, if only the yellow dot
   signal rises without a change in the red dot signal, or if both yellow
   and red dots decrease.

In vitro gene silencing

   OSCC cells were seeded in 24 or 6-well plates and incubated in a
   culture medium with 10% FBS overnight. Thereafter, the medium was
   replaced with a fresh medium and siRNAs were transfected into cells via
   Lipofectamine 3000 transfection reagents (Invitrogen, Carlsbad,
   California, USA) in opti-MEM (Genom, China). The following sequences of
   siRNAs are listed in Supplementary Table [282]5.

Lactate production assay and glucose uptake assay

   The L-Lactate Assay Kit (Elabscience, China) was used to accurately
   measure lactate production. Similarly, the Glucose Uptake Colorimetric
   Assay Kit (Elabscience, China) facilitated the assessment of glucose
   uptake, following the recommended protocols. In the lactate assay, the
   cells were seeded in 6-well culture plates and incubated for 24 h at
   37 °C. After a starvation period of 2 h, the supernatants of each group
   were gathered and underwent deproteinization. The subsequent
   measurement of lactate production was conducted using a microplate
   reader at 450 nm.

   Regarding the glucose uptake assay, the cells underwent specific
   treatments to assess their glucose uptake capabilities. The levels of
   glucose uptake were quantitatively determined by measuring the optical
   density (OD) in a microplate reader at 412 nm. All experiments were
   conducted in triplicate to ensure reproducibility and reliability of
   the results.

Rescue experiments

   The pLenti-U6-CKAP4-shRNA-CMV-copGFP-P2A-puro vector was designed and
   constructed by WZ Biosciences Inc. (Shandong, China). OSCC cells were
   transfected with vectors using Lipo 3000. The empty vector was used as
   a control. 72 h post-transfection, puromycin (10 μg/mL) was added to
   the cells in a 6-well plate. Fresh medium containing puromycin was
   replaced every 2 days for a continuous selection period of 14 days,
   until the proportion of fluorescent cells observed under a microscope
   reached 100%. Western blot was used to verify the knockout of CKAP4 in
   CAL27 and SCC15 cells. The wild-type CKAP4 and CKAP4 mutant lacking the
   TmpC-binding site (Δ991-1350) were synthesized and cloned into pENTER
   vectors by WZ Biosciences Inc. These vectors were transfected into
   CKAP4 knockdown cells used Lipo 3000. After transfection 48 h, cells
   were stimulated with P. micra for 24 h, Then, cells were used for
   western blot, migration and invasion assay. The expression of CKAP ^WT
   and CKAP4 ^Δ991-1350 were verified by Flag expression.

Construction of P. micra ΔtmpC

   The gene knockout of tmpC is based on the natural competence of P.
   micra. Targeted mutagenesis against ATCC33270 was designed using
   cloning-independent methodologies.^[283]55,[284]56 Initially, a gene
   knockout construct was built, utilizing ermB as the resistance
   selection marker, which was synthesized by BGI. The ermB gene was
   flanked by approximately 1 kb of upstream and downstream homologous
   sequences of tmpC. During the knockout process, approximately 1 ug DNA
   fragment was first added dropwise onto a BHI plate and allowed to be
   fully absorbed under anaerobic conditions. Subsequently, bacterial
   suspension in the logarithmic growth phase (OD = 0.4) was overlaid onto
   the fully absorbed DNA fragment and incubated for 1–2 days. The
   bacteria were then harvested and plated onto BHI plates containing
   erythromycin for anaerobic cultivation for 5–7 days. Transformants were
   screened using primers 1 F/2 R.

Recombinant protein production and purification

   The full TmpC gene of P. micra and the CKAP4 gene of Homo sapiens were
   synthesized by Sangon Biotech (Shanghai, China). After rigorous
   verification, the prokaryotic expression vector was successfully
   transformed into E. coli BL21(DE3) strains. These E. coli cells were
   cultivated in LB medium until they reached an OD600 value of 0.6,
   indicating optimal growth conditions. Subsequently, isopropyl
   β-d-1-thiogalactopyranoside (IPTG) at 0.5 mmol·L-1, sourced from
   Aladdin in Shanghai, China, was used to induce the cultures for 2.5 h.

   A His-tag Protein Purification Kit (Byotime, Shanghai, China) was
   employed for TmpC purification and a GST-tag Protein Purification Kit
   was applied for CKAP4 purification. Furthermore, Amicon® Ultra-15
   Centrifugal Filters from Millipore were used to efficiently desalt,
   diafilter, and concentrate the purified protein, ensuring its purity
   and concentration for further experiments.

Agarose gel electrophoresis

   To fully dissolve the agarose, combine 0.8 grams of agarose with 80
   milliliters of 1 × TAE buffer solution and heat the mixture in a
   microwave. Before adding 8 microliters of Solargel Red, vigorously stir
   the mixture. Once the mixture reaches approximately 50 °C, pour it into
   the tank and allow it to cool fully. Ensure the agarose gel is entirely
   submerged in the TAE buffer by positioning it in the electrophoresis
   tank. Load the sample holes with DNA samples, and then adjust the
   voltage to 100 volts for a duration of 30 minutes. Capture images using
   the UV imaging system.

Blockade of CKAP4 with neutralizing antibodies

   For CKAP4 blockade in vitro, OSCC cells were pre-incubated with
   anti-CKAP4 antibodies (Santa, catalog number sc-393544; dilution ratio,
   1:100) or isotype control (Santa, catalog number sc-2025; dilution
   ratio, 1:50) for 2 h before the addition of P. micra or P. micraΔtmpC.

His pull-down assay

   The Pierce pull-down polyhis protein: protein interaction kit (Thermo
   Fisher Scientific, MA, USA) was used for His pull-down analysis. After
   the isolation of pull-down proteins, they underwent proteolytic
   digestion to yield peptides, which were subsequently identified by
   Liquid Chromatography-Mass Spectrometry (LC-MS), also provided by
   Thermo Fisher Scientific. This approach can accurately determine the
   interacting proteins.

Immunoprecipitation (IP) assay

   To investigate the interaction between TmpC and CKAP4, OSCC cells were
   first pre-incubated with 0.1 mg/ml of TmpC for 120 min. Subsequently,
   the cells were incubated at 4 °C overnight with either an IgG from
   Santa Cruz Biotechnology (Texas, USA) or anti-His antibody from
   Proteintech. Then, protein A/G agarose beads from Santa Cruz
   Biotechnology were employed to isolate and capture the resulting
   protein complexes. After thorough washing, the antibody-bound complexes
   were eluted and Western blot analysis was performed as previously
   described, allowing for a detailed examination of the TmpC-CKAP4
   interaction.

Co-immunoprecipitation (Co-IP) assay

   Cells stimulated with P. micra/TmpC were lysed at 4 °C for 30 min with
   a complete protease inhibitor cocktail and phosphatase inhibitor as
   well as 1% NP-40 lysis buffer (Beyotime, P0013F). Centrifugation was
   conducted to discard lysates. The supernatant at 100 μl was used as an
   input and the left was incubated with anti-CKAP4 (Proteintech,
   16686-1-AP), anti-NBR1(Santa, sc-130380), anti-RanBP2 (Santa, sc-74518)
   or ant-P62 (Proteintech, 18420-1-AP). The mouse or rabbit IgG was added
   to the supernatant and incubated overnight at 4 °C. Protein A/G-
   magnetic beads (Santa, sc-2003) were incubated at 4 °C for 2 h and
   boiled for 10 min with loading buffer, and the complex was washed with
   1% NP-40 lysis buffer three times.

Surface plasmon resonance (SPR) analysis

   SPR analysis was conducted using a Biacore T200 (GE Healthcare, MA,
   USA). Amine coupling chemistry was used to immobilize about 200 RU of
   His-tagged recombinant TmpC on a sensor chip CM5. Unreacted moieties
   were blocked with ethanolamine. Recombinantly purified CKAP4 went
   through a sensor chip at various concentrations. None coated flow cell
   was used as a nonspecific binding control. Data was analyzed using
   Biacore T200 software.

Multiple sequences alignment analysis

   The amino acid sequence of spike protein of SARS-CoV-2 ([285]P0DTC2)
   was downloaded from the NCBI website
   ([286]https://www.ncbi.nlm.nih.gov/). Stereochemistry structure of
   SARS-CoV-2 spike protein was searched in the PDB database and obtained
   with entry ID 7ek6 ([287]https://www.rcsb.org/). Multiple sequences
   alignment analysis was performed by the mafft software^[288]57 using
   the default parameters and visualized by the ESPript^[289]58.

Docking-based interaction

   The amino acid sequences of protein TmpC and CKAP4 were obtained from
   the UniProt website. Sequences of two proteins were then uploaded to
   the alphafold server (AlphaFold 3, [290]https://alphafoldserver.com/)
   to do the protein-protein docking analysis. After the docking forecast,
   molecular conformation with lowest energy was selected and Pymol
   software was utilized to further analyze and visualize the.cif files of
   alphafold server.

Docking-based virtual screening

   From the DrugBank database (Release Version 5.1.12), we downloaded the
   structures of 9468 drug molecules in sdf format. Utilizing the
   Openbabel software, we segmented the dataset and then converted the
   molecules in batches into the mol2 format. The amino acid sequence of
   the protein CKAP4 was downloaded from UniProt. Then, the sequence was
   uploaded to the alphafold server ([291]https://alphafoldserver.com/) to
   obtain an optimal structure of CKAP4. To improve the reliability of the
   predicted structure, a sequence fragment with very low pIDDT score was
   removed by the Pymol software. The LeDock software^[292]59 was then
   used to conduct a batch virtual screening to dock small molecules (mol2
   format) and CKAP4. Ultimately, results with low energy scores
   (kcal/mol) were selected for further experimental verification.

Bacterial attachment and invasion assay

   OSCC cells were cultured with P. micra (MOI = 100) in a 6-well plate
   for 2 h in antibiotic-free DMEM or DMEM/F-12 medium supplemented with
   10% FBS under anaerobic conditions. For the invasion assay and
   bacterial attachment, the cells were thoroughly washed with warm PBS
   three times, followed by lysis with 1% TritonX-100 for 5 min. The
   resulting cell lysate was then gathered and plated on Schaedler agar
   for an enumeration of attached bacterial counts through serial
   dilution. For the invasion assay, OSCC cells were treated with
   ampicillin and gentamicin (100 μg/mL) for another 2 h before
   enumerating bacterial colony-forming units (CFU). For flow cytometry
   analysis, P. micra was treated by CFSE (20 μM) for 0.5 h before
   co-culture with OSCC cells for another 2 h. Antibiotic-treated or
   non-treated cells were collected and subjected to flow cytometric
   analysis.

Scanning electron microscopy (SEM)

   OSCC cells were seeded in sterile coverslips and co-cultured with P.
   micra (MOI = 100) for 2 h. Subsequently, the cells were then washed
   thoroughly three times with freshly prepared Sorensen’s phosphate
   buffer (0.1 M, pH 7.2), followed by fixation overnight using 2.5%
   glutaraldehyde fixative. The fixed coverslips were then washed three
   more times with Sorensen’s phosphate buffer and further fixed using 1%
   osmium tetroxide for a duration of 2 hours. The specimens were then
   rinsed with distilled water, dehydrated, and coated with gold
   palladium. Finally, the prepared specimens were measured under a Cold
   Field Scanning Electron Microscope (Hitachi SU8010).

Transmission electron microscopy (TEM)

   OSCC cells were co-cultured with P. micra (MOI = 100) for 2 h. The
   cells were initially pre-fixed for 2 h in 0.1 M sodium cacodylate
   buffer (SCB) at pH 7.2 using a 2.5% (v/v) glutaraldehyde solution.
   Subsequently, the cells were thoroughly washed three times with 0.2 M
   SCB and post-fixed at room temperature for 1 h in 1% (w/v) osmium
   tetroxide dissolved in 0.2 M SCB. Afterwards, the cells underwent a
   dehydration process through a series of graded ethanol concentrations,
   culminating in 100% acetone. Once dehydrated, the cells were
   infiltrated with a combination of acetone-Epon 812 resin mixture,
   eventually transitioning to 100% Epon 812 resin. Ultra-thin serial
   sections were carefully gathered on a copper formvar-coated slot grid,
   and these sections were stained with 2% (w/v) uranyl acetate, leading
   citrate to enhance contrast. Finally, the sections were visualized and
   examined under an electron microscope.

Luciferase imaging

   Mice were anaesthetized and administered with D-luciferin at the dose
   of 150 mg/kg body weight. After 10 min, tumor and lymph tissues were
   extracted and Luciferase imaging (IVIS Spectrum) was performed.

   Key resources were list in Supplementary Table [293]5.

Statistics and reproducibility

   Each experiment was performed at least two biological replicates.
   Statistical analysis was performed using GraphPad Prism 9.5 (GraphPad
   Software, Inc.). two-sided unpaired Student’s t test was performed to
   compare the means between two groups. Multiple group comparison was
   conducted by o One-way analysis of variance (ANOVA) with Turkey’s test.
   Two-sided Wilcoxon test was used to examine the abundance of P. micra
   in different groups. The survival curves were plotted using the
   Kaplan-Meier method and compared with the log-rank test. Details of
   statistical analysis were mentioned in each Figure legends. The
   probability values of < 0.05 were considered statistically significant.
   Exact p-value was provided.

Reporting summary

   Further information on research design is available in the [294]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [295]Supplementary Information^ (18.3MB, pdf)
   [296]41467_2025_57530_MOESM2_ESM.pdf^ (73KB, pdf)

   Description of Additional Supplementary Information
   [297]Supplementary Data 1^ (22.9KB, xlsx)
   [298]Supplementary Data 2^ (233.9KB, xlsx)
   [299]Supplementary Data 3^ (5.6MB, xlsx)
   [300]Supplementary Data 4^ (23.6KB, xlsx)
   [301]Supplementary Data 5^ (18.2KB, xlsx)
   [302]Reporting Summary^ (1.6MB, pdf)
   [303]Transparent Peer Review file^ (6.4MB, pdf)

Source data

   [304]Source Data^ (59.2MB, xlsx)

Acknowledgements