Fusobacterium nucleatum is associated with worse prognosis in Lauren’s diffuse type gastric cancer patients

Fusobacterium nucleatum (F. nucleatum) is frequently detected in primary colorectal cancer (CRC) and matching metastasis, and has been linked to a worse prognosis. We investigated the presence of F. nucleatum in gastric cancer (GC) and gastric preneoplastic conditions of the stomach, and its potential prognostic value in GC patients. Fusobacterium spp. and F. nucleatum were quantified in various specimens from gastrointestinal tract including paired CRC and GC tissues using probe-based qPCR. Fusobacterium spp. and F. nucleatum were more frequently found in tumorous tissue of CRC and GC compared to non-tumorous tissues. The frequency and bacterial load were higher in CRC compared to GC patients. F. nucleatum positivity showed no association to chronic gastritis or preneoplastic conditions such as intestinal metaplasia. F. nucleatum-positivity was associated with significantly worse overall survival in patients with Lauren’s diffuse type, but not with intestinal type GC. There was no association with gender, Helicobacter pylori-status, tumor stage or tumor localization. However, F. nucleatum was positively associated with patient’s age and a trend for a lower global long interspersed element-1 DNA methylation. In conclusion, our work provides novel evidence for clinical relevance of F. nucleatum in GC by showing an association between F. nucleatum positivity with worse prognosis of patients with Laurens’s diffuse type gastric cancer. Further studies are necessary to explore related mechanistic insights and potential therapeutic benefit of targeted antibiotic treatment in GC patients.

www.nature.com/scientificreports/ F. nucleatum has been correlated with expression of proinflammatory genes, lower CD3 + T-cell density and increased TNF-α gene expression in CRC [14][15][16] . The microbiome composition of the stomach is unique. Helicobacter pylori (H. pylori) is the predominant species and the key trigger for development of peptic ulcer disease and GC 17,18 . Despite years of research, the exact interaction of H. pylori with mucosa remains only partially understood. It is now clearly recognized that H. pylori is an infectious disease that causes chronic non-atrophic gastritis (CNAG) that can progress to preneoplastic conditions such as atrophic gastritis (AG), intestinal metaplasia (IM) and finally to dysplasia and cancer 19 . With new sequencing tools, it is increasingly appreciated that not H. pylori alone but rather the microbiome in whole complexity contributes to disease conditions. Several studies in detail reported about microbial alterations in stomach 20,21 . Fusobacterium spp. are frequently found in stomach mucosa [20][21][22][23] . According to few preliminary reports F. nucleatum have been found in tumorous GC tissues as well 24,25 , but there are still many unanswered questions. High-throughput techniques including 16 s RNA/DNA sequencing allow only a relative quantification of microbial community while polymerase chain reaction (PCR) based F. nucleatum analysis may provide an absolute quantification in relation to human cells. Next, whether F. nucleatum may be linked to preneoplastic conditions and contribute to carcinogenesis is still unknown. Most importantly, the clinical and prognostic relevance of F. nucleatum in GC has not been studied in detail.
In the present study, we performed in-depth characterization of Fusobacterium spp. and F. nucleatum in GC. To elaborate on its potential role in gastric carcinogenesis, we evaluated normal gastric mucosa (N), chronic gastritis samples with CNAG or with AG and IM, and correlated the positivity to clinicopathological characteristics and prognosis of GC patients.

Results
F. nucleatum in cRc . F. nucleatum have been previously evaluated in CRC tissues using PCR-based quantitative analysis. To confirm the analysis in our European cohort, we first validated the quantitative detection method and the reproducibility of Fusobacterium spp. and F. nucleatum analysis in a subset of samples from CRC patients. Based on our reproducibility results, the cycle threshold (Ct) values of ≤ 38 for both Fusobacterium spp. and F. nucleatum were classified as positive. In non-tumorous and tumorous CRC tissues we observed Fusobacterium spp. positivity in 69.23% (18/26) and 92.59% (25/27) (Fig. 1A, p = 0.0394), respectively. F. nucleatum positivity was present in 50% (13/26) N-CRC and 59.26% (16/27) T-CRC specimens (Fig. 1B). Overall, there was a significant correlation between F. nucleatum and Fusobacterium spp. (Fig. 1C, p < 0.0001). Analysis of the N-CRC and T-CRC samples (Fig. 1D,E) revealed only a trend for positive correlation for Fusobacterium spp. abundance in paired samples (p = 0.0817), while F. nucleatum load correlated significantly between N-CRC and T-CRC (p = 0.0112). Overall, we confirm that Fusobacterium spp. and F. nucleatum are more frequently detectable in T-CRC than in N-CRC and F. nucleatum load correlates significantly between tumorous and nontumorous tissues.
F. nucleatum in Gc. Next, we investigated F. nucleatum in tumour tissues of gastric cancer (T-GC) and its adjacent mucosa (N-GC). Based on our validation and reproducibility results, the Ct-value of ≤ 38 cycles were defined as positivity also in gastric mucosa. Fusobacterium spp. was detectable in 65.38% (51/78) of N-GC and 77.78% (63/81) of T-GC samples ( Fig. 2A). F. nucleatum was positive in 23.08% (18/78) of N-GC and 28.75% (23/80) of T-GC samples (Fig. 2B). In similar fashion as in CRC, we observed a statistically significant correlation between abundance of Fusobacterium spp. and F. nucleatum (p < 0.0001) in mucosa of GC patients (Fig. 2C). Furthermore, the abundance of Fusobacterium spp. and F. nucleatum correlated significantly between N-GC and T-GC (each p < 0.0001, Fig. 2D,E). Differences in F. nucleatum abundance between cRc and Gc. Following normalization to prostaglandin transporter (PGT), we observed a significant correlation between Fusobacterium spp. and F. nucleatum in CRC (Fig. 3A, p < 0,0001) and in GC (Fig. 3B, p < 0,0001). In comparison to non-normalized values presented in Figs. 1 and 2, the normalized abundance of Fusobacterium spp. and F. nucleatum was similar between N-CRC and T-CRC and between N-GC and T-GC, respectively ( Supplementary Fig. S1). Next, we evaluated the differences in bacterial load of Fusobacterium spp. and F. nucleatum between CRC and GC. Despite the anatomical distance to oral cavity, abundance of Fusobacterium spp. in N-CRC and T-CRC was significantly higher than in N-GC and T-GC, respectively (Fig. 3C,D). In addition, F. nucleatum was higher in N-CRC and T-CRC compared to N-GC and T-GC, respectively (Fig. 3E,F).

F. nucleatum in preneoplastic conditions in comparison to Gc.
To explore the potential involvement of Fusobacterium spp. and F. nucleatum we compared samples from patients with normal mucosa (N), CNAG, AG/IM, N-GC and T-GC. The analysis of Ct-values, revealed relatively similar pattern of Fusobacterium spp. and F. nucleatum abundance in normal and chronic gastritis with or without preneoplastic conditions in comparison to GC, suggesting that F. nucleatum may be probably involved in rather late stages of classical Correa's cascade of gastric carcinogenesis (Fig. 3G,H). F. nucleatum was present in 16.7% (3/18) of N, 17.65% (3/17) CNAG mucosa and 0% (0/9) in AG/IM mucosa, which was not significantly different to N-GC and T-GC. Since none of the AG/IM mucosa samples were positive for F. nucleatum, we did not perform any correlation to OLGA/OLGIM. To evaluate potential association between F. nucleatum and H. pylori in non-neoplastic mucosa, we compared Fusobacterium spp. and F. nucleatum levels between subjects with and without active H. pylori infection irrespective of gastritis type or severity in total cohort of non-neoplastic mucosa. As shown in the Supplementary Fig. S2, we observed no difference in Fusobacterium spp. between H. pylori positive and negative gastric mucosa, while a slightly lower level of F. nucleatum was found in H. pylori positive mucosa (p = 0.046).  Fig. 4A) and higher F. nucleatum (p = 0.0031, Fig. 4B) abundance in T-GC. Furthermore, patients with F. nucleatum positive T-GC were overall older than patients with negative T-GCs. Based on the median age with cut-off of 68 years, older groups with GC had significantly higher F. nucleatum load compared to younger patients (Fig. 4C).
Correlation with global and gene specific methylation changes. It has been recently suggested that F. nucleatum may be associated with distinct molecular alterations in cancer. We evaluated possible correlation between F. nucleatum and global DNA hypomethylation using surrogate long interspersed element-1 (LINE-1) methylation and miR-137 promoter methylation, which are frequently deregulated in GC and CRC. Overall, correlation analysis between LINE-1 and F. nucleatum revealed a non-significant trend for lower LINE-1 methylation in subjects with higher F. nucleatum load (p = 0.156, Fig. 4D). LINE-1 methylation in the F. nucleatumpositive group was slightly lower as in F. nucleatum-negative group although the difference did not reach statistical significance (60.1 ± 9.6 vs. 63.4 ± 7.4, p = 0.09) (Fig. 4E). For comparison, gene specific DNA methylation analysis of miR-137 and F. nucleatum revealed no difference between the groups (data not shown).  5D). Comparison of clinicopathological characteristics of F. nucleatum positive and negative diffuse type of GC revealed only differences in age, but no other major differences, suggesting stage-independent effect of F. nucleatum positivity on the prognosis in diffuse type GC (Table 2).

Discussion
Increasing evidence suggests that F. nucleatum may be involved in tumour development and associated with worse prognosis in CRC and other cancers. However, only limited data is available on the role of F. nucleatum in GC and gastric preneoplastic conditions. Using a well-characterized cohort of GC patients, we showed that Fusobacterium spp. and F. nucleatum may be frequently found not only in N-and T-CRC, but also in N-and T-GC although less frequently and at lower abundance. F. nucleatum was furthermore detected in normal mucosa and chronic gastritis. Interestingly, F. nucleatum was found in N-CRC and T-CRC in higher abundance despite the anatomical distance compared to N-GC and T-GC, respectively. Overall survival analysis revealed a significantly worse prognosis of patients with F. nucleatum-positive T-GC only in Lauren's diffuse type GC, but not in intestinal type GC. www.nature.com/scientificreports/ Amounting evidence has been collected to confirm the presence of F. nucleatum in CRC. Our data are in the frame of existing reports showing the F. nucleatum positivity in up to 60% of CRC specimens 8,11,26 . Only few preliminary reports have been dealing with this topic in GC and no data to prognostic relevance of F. nucleatum in GC has been studied yet. Yamamura et al. studied 20 samples from various GI cancers and detected F. nucleatum   25 . In our cohort, we observed F. nucleatum positivity in GC patients in up to 28.75%. Surprisingly, the absolute abundance of F. nucleatum in T-GC was not different to N-GC, which is different to CRC studies. We next performed the comparison of F. nucleatum absolute load between tumorous and non-tumorous colon and gastric mucosa. Both Fusobacterium spp. and F. nucleatum were at higher abundance in N-CRC compared to N-GC, as well as in T-CRC compared to T-GC. It is remarkable as the anatomical distance and proximity to an oral cavity would probably rather suggest higher abundance of F. nucleatum in the stomach as in the colon. Two reports have recently published results elaborating on the potential mechanism of F. nucleatum transfer to the tumours. Abed et al. 27 have recently shown that host polysaccharide Gal-GalNAc, which is overexpressed www.nature.com/scientificreports/ in CRC, recognizes fusobacterial Fap2, which may trigger binding of F. nucleatum to the tumours. In another report, the authors confirmed an increased Gal-GalNAc levels in various tumours including GC 28 . Although the level of Gal-GalNAc was high in both CRC and GC tissues, the level in non-tumorous CRC samples was much lower as in non-tumorous GC which may explain the load differences. Overall, our results may support the hypothesis of potential hematogenous route of F. nucleatum spreading. Some time ago, several initial studies have reported the capability of F. nucleatum to form biofilms. For instance, Zilm et al. reported that F. nucleatum may form biofilms and optimize its adhesion characteristics 29 . This property of F. nucleatum was dependent on the host environment in response to alkaline pH 30 . In CRC using the 3-dimensional tumour spheroid model, Kasper et al. observed development of biofilm-like structure in the tumour spheroid microenvironment by F. nucleatum 31 . The pathogenicity of F. nucleatum in the stomach may however be different as its low pH creates a unique microenvironment and microbial interplay. Low abundance of F. nucleatum in stomach in comparison to colon allows us to speculate on protective properties of acidic milieu preventing F. nucleatum dissemination. From another side, we observe no clear pattern for an increased abundance of F. nucleatum in AG/IM tissues where higher pH due to mucosa atrophy is expected. Considering the increasing interest in biofilm formation in the colon, further studies will be also necessary to address this point in the stomach microenvironment.
To understand the functional role of F. nucleatum in GC, we next analysed F. nucleatum in non-/preneoplastic gastric mucosa under consideration if H. pylori status and performed survival analysis. Aviles-Jimenez et al. have recently linked certain alterations in stomach microbiota composition to Correa's cascade stages from CNAG to IM to intestinal type gastric cancer 32 . In our specific quantitative analysis, we did not observe any difference in F. nucleatum in preneoplastic conditions as well as no clear signal was found for H. pylori status. Since the sample size was sufficient only for pilot analysis, further studies will be needed to take a closer look at the F. nucleatum abundance in preneoplastic conditions with its variables and influencing factors specifically.
F. nucleatum has been repeatedly associated with worse prognosis in patients with oesophageal cancer 9 , pancreatic cancer 10 and colorectal cancer 26,33 , but the data on GC are not available, yet. Although the overall survival analysis revealed only a non-significant trend toward a worse prognosis, we further performed subgroup www.nature.com/scientificreports/ analysis based on the Lauren's classification, which is one of the most simple and valuable classifications of GC that partially mirrors the molecular GC classification and is frequently underappreciated in scientific work related to GC 34 . While no pattern was observed for intestinal type, we observed significantly worse overall survival in diffuse type GC patients with F. nucleatum positive tumours. It has been reported that F. nucleatum may promote carcinogenesis in CRC via FadA adhesin, which binds to E-cadherin, activated β-catenin signalling and accordingly various inflammatory and oncogenic properties of the cells 35 . Since diffuse type of GC is strongly associated with E-cadherin deregulation one may speculate for potential molecular mimicry of F. nucleatum to diffuse type of GC and probably specific prognostic relevance. In one of the pivotal reports, F. nucleatum was associated with CIMP positivity, hMLH1 methylation, MSI and CHD7/8 positivity 11 . We analysed correlation between F. nucleatum and LINE-1 as a global methylation marker and miR-137 methylation 36 . F. nucleatum positive GC tumours showed a trend to lower LINE-1 methylation with overall positive correlation, while no association was found for miR-137. Although this may suggest that indeed, F. nucleatum positivity could be associated with certain epigenetic alterations such as global DNA hypomethylation, from another side, the lower LINE-1 DNA methylation could also be related to the aging as F. nucleatum positivity correlated strongly also to older age.
Despite intriguing results, we would like to underline that this is one of the first analyses and multiple remaining questions need to be addressed in future work. First, the study aimed to evaluate specifically the translational role of F. nucleatum in GC, therefore the data acquired may allow only a partial view on the microbial changes. Microbiome-sequencing may provide in-depth view on microbial alterations in GC. Second, our work provides only some preliminary molecular analysis on correlation with LINE-1 methylation. Additional in vitro and in vivo studies should provide mechanistic insights and explanation. Third, in particular from the clinical point of view, the data to F. nucleatum may have substantial clinical consequences. It has been recently reported that antibiotic treatment of tumours harbouring F. nucleatum led to reduced tumour growth in mice 12 . Therefore, use of antibiotics (for example metronidazole) could be a possible therapeutic consequence in patients with diffuse www.nature.com/scientificreports/ type GC with F. nucleatum positivity. Furthermore, the impact of Fusobacterium on the treatment response especially in the era of immunotherapy may be quite intriguing. Recently, it has been reported that prudent diets rich in whole grains and dietary fibres were associated with lower risk of F. nucleatum positive CRC while diets that may promote intestinal inflammation were associated with increased risk of F. nucleatum positive tumours 37,38 . Diet has been shown to provide a great source of various microRNAs including xenomiRNAs 39 , therefore, taking into account an association between diet and F. nucleatum positivity one may speculate on the role of exogenous microRNA or even various drugs. Further studies will be needed to address the impact of proton-pump-inhibitors and antibiotics on positivity and variation of F. nucleatum in stomach and CRC. In summary, the results of our work strongly support the potential involvement of F. nucleatum in gastric carcinogenesis. F. nucleatum is frequently found in normal, preneoplastic and neoplastic mucosa although substantially lower than in colon. Even though there were no specific clinicopathological differences related to F. nucleatum positive gastric cancer patients, F. nucleatum positivity was associated with significantly worse overall survival in diffuse Lauren's type GC patients. Further studies are needed to evaluate possible therapeutic implications and molecular alterations responsible for this phenotype. Samples collection. The collection and characterization of biological material was partly described in our previous studies 36,40 . Briefly, specimens from GC and CRC were prospectively collected during surgical interventions. Samples from controls (N) and patients with various stages of chronic gastritis were obtained during endoscopy. The samples were immediately snap-frozen in liquid nitrogen and placed in − 80 °C freezer. The updated Sydney classification was applied for histological characterization of gastritis 41 . The Lauren's classification was used for histological assessment of GC tumours. H. pylori status was analysed either by H. pylori ELISA IgG test (Virion\Serion GmbH, Germany) for GC patients or using multistep approach via serology, microbiology and histology as previously reported 42,43 . We obtained 81-paired samples from patients with GC including tumour tissues (T-GC) and their corresponding adjacent non-tumorous gastric mucosa (N-GC). Histopathological assessment of GC tissues was performed by an experienced pathologist at the tertiary centre form Lithuania. For preliminary analysis we included samples from 18 patients with histologically confirmed normal gastric mucosa (N), 17 patients with CNAG and 9 patients with AG/IM. In addition, we included samples from 27 patients with colorectal cancer (T-CRC) and their corresponding adjacent non-tumorous colon mucosa (N-CRC). An overview for sample collection and methods are presented in Supplementary Table S1 and the clinical and demographic data in Table 1.

Materials and methods
DNA isolation and quantitative real-time PCR. DNA was extracted from frozen tissue samples, pretreated with QIAzol Lysis reagent (Qiagen, Valencia, CA) and chloroform based on manufacturer's recommendations as described previously 36,40 . Probe-based quantitative real-time PCR was performed using Bio-Rad CFX96 real-time PCR cycler (BioRad, CA). Following probe-based primer were used: Fusobacterium spp. 44 ; F. nucleatum 9 ; prostaglandin transporter (PGT), also known as solute carrier organic anion transporter family, member 2A1 (SLCO2A1), as endogenous control for normalization as previously described 8 . Primer and probe sequences are provided in Supplementary Table S2. Ct-values for Fusobacterium spp. and F. nucleatum were set to 40 if PCR analyses revealed a negative result. Normalization was performed using 2^deltaCt-method. The values of the samples with undetectable Fusobacterium spp. and F. nucleatum were set to the lowest measurable normalized values.

Methylation analysis.
Purified genomic DNA from tissue samples was used for global long interspersed nucleotide element-1 (LINE-1) and miR-137 promoter methylation analyses. The procedure was in detail described in our previous reports 36,40 . Briefly, we applied Cells-to-CpG Bisulfite Conversion Kit (Life Technologies, Carlsbad, CA) for bisulphite modification, thereafter the standard PCR with biotin-labelled primers and eventually the pyrosequencing on PyroMark Q96 ID (Qiagen) using PyroMark Gold Q96 reagents (Qiagen). The mean methylation level of analysed CpG motifs was used for quantitative methylation analysis. Statistical analysis. Statistical evaluation was conducted with GraphPad Prism 7.0 (San Diego, CA), statistical software. We applied χ 2 -test for qualitative analysis and for quantitative analysis we used either Wilcoxon test for paired samples or Mann-Whitney U test for unpaired samples. For comparison of more than two groups we used the Kruskal-Wallis test. Spearman's test was applied for correlation analysis. Survival analyses were performed with the Mantel-Cox test. Two-sided p-values of < 0.05 were considered as statistically significant.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.