Colon Cancer-Associated Fusobacterium nucleatum May Originate From the Oral Cavity and Reach Colon Tumors via the Circulatory System

Fusobacterium nucleatum is a common oral bacterium that is enriched in colorectal adenomas and adenocarcinomas (CRC). In humans, high fusobacterial CRC abundance is associated with chemoresistance and poor prognosis. In animal models, fusobacteria accelerate CRC progression. Targeting F. nucleatum may reduce fusobacteria cancer progression and therefore determining the origin of CRC F. nucleatum and the route by which it reaches colon tumors is of biologic and therapeutic importance. Arbitrarily primed PCR performed previously on matched same-patients CRC and saliva F. nucleatum isolates, suggested that CRC F. nucleatum may originate from the oral cavity. However, the origin of CRC fusobacteria as well as the route of their arrival to the tumor have not been well-established. Herein, we performed and analyzed whole genome sequencing of paired, same-patient oral, and CRC F. nucleatum isolates and confirmed that CRC-fusobacteria originate from the oral microbial reservoir. Oral fusobacteria may translocate to CRC by descending via the digestive tract or using the hematogenous route during frequent transient bacteremia caused by chewing, daily hygiene activities, or dental procedures. Using the orthotropic CT26 mouse model we previously showed that IV injected F. nucleatum colonize CRC. Here, we compared CRC colonization by gavage vs. intravenous inoculated F. nucleatum in the MC38 and CT26 mouse orthotropic CRC models. Under the tested conditions, hematogenous fusobacteria were more successful in CRC colonization than gavaged ones. Our results therefore provide evidence that the hematogenous route may be the preferred way by which oral fusobacteria reach colon tumors.


INTRODUCTION
Fusobacterium nucleatum is a Gram-negative, non-spore forming oral anaerobe and one of the abundant species found in the oral cavity (Socransky et al., 1998;Nozawa et al., 2020). F. nucleatum is also among the pathobionts that outgrow during dysbiosis preceding periodontal disease (Socransky et al., 1998;Nozawa et al., 2020) to assist keystone species such as Porphyromonas gingivalis (Hajishengallis et al., 2011) to drive the breakdown of host-microbial homeostasis and induce periodontitis (Hajishengallis and Lamont, 2016;Lamont et al., 2018).
These observations that F. nucleatum colonizes CRC, promotes tumorigenesis and affects treatment outcome strongly support that strategies to selectively target bacteria (Zheng et al., 2019) and to hinder their pro-tumorigenic factors should be incorporated into colonic adenoma and CRC treatment. Revealing the origin of CRC fusobacteria and the route by which they reach the tumor, might facilitate development of approaches to reduce numbers of CRC fusobacteria. To date, the "biogeographic" origin of CRC fusobacteria remains under-explored. As F. nucleatum is a common oral microbe, and in step with current thinking (Han, 2015;Brennan and Garrett, 2018;Komiya et al., 2018) we hypothesized that F. nucleatum found in CRC originate from the oral cavity. Trafficking of oral F. nucleatum to CRC can occur by descending through the digestive tract. However, as transient bacteremia is frequent during chewing food, daily oral hygiene activities and during dental procedures (Wilson et al., 2008), hematogenous dissemination of oral fusobacteria to colon tumors is also likely.
Here we furnish evidence that CRC fusobacteria are likely oral in origin and that CRC colonization by blood-borne F. nucleatum is efficient.

Collection of Clinical Samples
The Hadassah Medical School IRB approved the use of human samples for this study. Informed consent was obtained from all patients. Patients received neomycin and flagyl orally 12 h before resection as well as metronidazole and cefuroxime IV 30 min prior to resection. For F. nucleatum isolation from the oral cavity, 1-2 ml of saliva was collected one day before resection or immediately following colonoscopy. Colonic adenocarcinoma samples (100-1,000 mm 3 ) were obtained from the pathology department in sterile tubes within 45 min of resection or colonoscopy. Fresh samples were immediately transferred to an anaerobic chamber (Bactron I-II Shellab, USA) with an atmosphere of 85% N 2 , 5% CO 2 , and 10% H 2 at 37 • C and streaked on Columbia agar plates (Oxoid, UK) supplemented with 5% defibrinated sheep blood (Novamed, Israel) and 0.15% crystal violet. Large resection samples were dissected with a scalpel in the anaerobic chamber before plating. Colonies consistent with F. nucleatum morphotypes were verified by Gram staining and microscopy followed by PCR as described below. Oral and CRC fusobacterial strains isolated from the same patient that appeared similar in morphology were kept for further examination.

Bacterial Strains and Growth Conditions
F. nucleatum strains used in this study are listed in Table 1. Fusobacterial strains were grown in Wilkins Chalgren broth (Oxoid, UK) or on Columbia agar plates (Oxoid, UK) supplemented with 5% defibrinated sheep blood (Novamed, Israel) in an anaerobic chamber described above.

Murine CRC Model
All experiments were performed in accordance with the guidelines of our institutions' animal welfare committee. Mouse orthotopic rectal colon tumors were performed as follow. Male 6-7 weeks old C57BL/6 or Balb/C mice were anesthetized with inhaled isoflurane and injected via the rectum (using a 26gauge syringe) into the mucosa and submucosa with luciferase expressing colon carcinoma cells MC38-luc or CT26-luc cells, respectively (1 × 10 6 cells in 100 µl PBS). Real-time imaging using a CCD camera was used to validate tumor development. Luciferin 5 mg/ml was injected intraperitoneally 15 min before image capture. A caliper was used to assess the volume of the tumors and inoculations were conducted when tumor volumes

Quantification of Bacteria Using Plating and qPCR
Murine samples were homogenized using sterile 2 ml tubes with 2.8 mm stainless steel beads in a Fastprep (MP Biomedicals, USA) for 60 s with speed of 6 m/s. Homogenates were serially diluted and plated on Columbia agar plates supplemented with 0.15% (final concentration) crystal violet and 5% (final concentration) defibrinated sheep blood. Plates were incubated in an anaerobic chamber at 37 • C for 6 days, and colonies were enumerated. DNA extraction from fresh tissues was performed using the DNeasy Blood & Tissue Kit (Qiagen, Germany), samples were incubated at 56 • C overnight. A custom TaqMan primer and probe set was used to amplify F. nucleatum DNA. The cycle threshold (Ct) values for F. nucleatum was normalized to the amount of murine gDNA in each reaction by using a primer and probe set for the reference gene (Gapdh). The fold difference (2 − Ct ) in F. nucleatum abundance in tumor vs. normal tissue was calculated as described before (Castellarin et al., 2012). Each reaction contained 1 ng of DNA and was assayed in triplicate in 20 µL reactions containing 2× qPCRBIO Lo-ROX Probe Mix as appropriate for individual qPCR StepOne Real-Time PCR system. Reaction conditions were as follows: 2 min at 50 • C, 10 min at 95 • C, and 40 cycles of 15 s at 95 • C and 1 min at 60 • C. The primers and probe sequences for each assay were as follows: Immunofluorescence MC38 and control HT-29 CRC cells were cultured in 48 well plates overnight. The cells were fixated for 15 min using 4% paraformaldehyde, blocked with PBS supplemented with 10% BSA, 10% FBS, and 0.5% Triton X-100 for 2 h at room temperature and stained overnight at 4 • C with FITC-labeled PNA (50 µg/ml in PBS, Sigma-Aldrich). The cells were then incubated with Hoechst 33258 (Sigma-Aldrich) diluted 1:5,000 for 15 min at room temperature washed three times with PBS followed by mounting using Fluoromount aqueous mounting medium (Sigma-Aldrich). Pictures were obtained using the Nikon Eclipse Ti microscope.

Hemagglutination Assay
Fusobacterial strains were grown overnight, washed twice in PBS, and brought to an OD600 of 1 (∼10 9 CFU ml −1 ). Sheep erythrocytes (RBCs) were washed twice in PBS and brought to a concentration of 2% (vol/vol). 50 µl of fusobacterial cells was mixed with 50 µl of sheep erythrocytes (2% in PBS) in round-bottom 96-well plates (Nunc, Denmark) and incubated at room temperature for 2 h. For inhibition assays, washed bacteria were pre-incubated with 25 mM GalNAc (Sigma-Aldrich) for 30 min prior to incubation with erythrocytes (Coppenhagen-Glazer et al., 2015).

Genomic Sequencing and Bioinformatics Analysis
Libraries of extracted gDNA were created using the Nextera R XT DNA kit (Illumina, San Diego, USA) according to manufacturer's instructions and sequenced on the Illumina MiSeq platform (Illumina, San Diego, USA) Individually tagged libraries were sequenced in a common flow cell as 2 × 150 base pairedend reads using the Illumina MiSeq platform (Illumina, San Diego, USA).

Comparative Analyses
Illumina reads for all F. nucleatum genomes sequenced with Illumina technology and deposited at the Sequence Read Archive were downloaded, for a total 18 additional F. nucleatum genomes ( Table 1). Our comparative dataset included 10 F. nucleatum subsp. animalis strains: 7_1, 4_8, 11_3_2, 3_1_33, OT420, CTI1, CTI3, CTI5, SRR5945983, and SRR2093830; three F. nucleatum subsp. vincentii strains: 4_1_13, CTI7, and CC53; three F. nucleatum subsp. nucleatum strains: CTI2, DRR015879, and DRR015964; and 2 F. nucleatum subsp. polymorphum strains: CTI6 and 13_3C. For each strain, including T2 and O2, reads were aligned to the F. nucleatum subsp. animalis 4_8 reference genome (AEIB00000000) using BWA mem version 0.7.12 (Li and Durbin, 2009). Pilon v.1.12 (Walker et al., 2014) was run in variant discovery mode using default settings to generate a list of single nucleotide polymorphisms (SNPs) for each data set. For phylogenetic reconstruction, a nucleotide alignment was built from all positions containing at least one passing SNP; ambiguous calls in other strains at these positions were scored as Ns. A phylogeny was then estimated using RAxML v7.3.0 (Stamatakis, 2006) using a GTR+Gamma substitution model with 1,000 bootstrap replicates, and midpoint-rooted. The RAxML phylogeny was based on 84,537 positions. To calculate the number of SNPs that differed between two strains, we considered only positions in the Pilon output that had highquality, passing variant calls, as well as good mapping quality (MQ > 10) in both strains.
For each genome, we obtained draft assemblies using the SPAdes assembler (Bankevich et al., 2012). Using these assemblies, we calculated genome-wide Average Nucleotide Identity (ANI) (Goris et al., 2007) using the online slidingwindow implementation available at the Kostas lab (http:// enve-omics.ce.gatech.edu/ani/index) with a default window size of 1000bp.

Statistical Analysis
GraphPad Prism software (Version 6.0) and IBM SPSS Statistics for Windows (Version 25, 2017, Armonk, NY: IBM Corp) were used for statistical presentation and analysis as described in the figure legends.

CRC Fusobacteria Originate From Oral Cavity
To investigate the possible oral origin of CRC-colonized F. nucleatum, saliva was obtained from seven patients prior to CRC surgical resection. A matching colon adenocarcinoma sample was obtained from each patient tumor immediately following resection. Fusobacteria could be isolated from all saliva samples, and fusobacterial gDNA was detected by PCR in all tested CRC samples. However, live fusobacteria were only cultured from one colon adenocarcinoma sample ( Figure 1A). This CRC isolate, F. nucleatum T2, and the paired oral isolate, F. nucleatum O2, were biobanked for sequencing and experimental analyses.
Attempts to culture fusobacteria from CRC samples from an additional 23 fusobacteria PCR+ patients were unsuccessful. We hypothesized that the antibiotics that patients received preoperatively resulted in the low post-operative isolation rate of live fusobacteria. In order to isolate CRC fusobacteria from patients that would not be pre-treated with antibiotics, we identified two patients with high clinical suspicion for CRC who were scheduled for a colonoscopy-based biopsy. CRC tumor samples were collected from these two patients undergoing colonoscopy, who as a matter of practice do not receive pre-procedural antibiotics. Many fusobacterial colonies were obtained from the colonoscopy biopsy samples and from the paired saliva samples. F. nucleatum isolates Oral003 and JAoral001 were isolated from these patients' saliva and CRC003 and JAcrc001 were isolated from their CRC samples, respectively ( Figure 1A).
Genomic DNA samples from the paired oral and adenocarcinoma fusobacteria were extracted, sequenced and compared to each other and to the fusobacterial genomes previously sequenced using Illumina technology and deposited in the Sequence Read Archive (SRA) database. Two of CRC isolates (T2 and CRC003) belonged to F. nucleatum subsp. animalis and the third (JAcrc001) to F. nucleatum subsp. polymorphum. Both F. nucleatum subspecies animalis and polymorphum are commonly associated with the oral cavity (Henne et al., 2018). Whole-genome Average Nucleotide Identity (ANI) (Goris et al., 2007) comparisons of the genomes of each isolated oral/CRC pair revealed very high homology with percentage identity for each pair as follows: O2/T2 100.00 ± 0.03%, Oral003/CRC003 99.98 ± 0.38%, and JAoral001/JACRC001 100.00 ± 0.04% (Figure 1B). The paired-patient isolates O2/T2, Oral003/CRC003, and JAoral001/JAcrc001 shared much greater genome sequence identity with one another than they did with other previously sequenced Fusobacterium genomes ( Figure 1C). Of the reference genome positions where high-quality read mapping was obtained for both T2 and O2 (82% of the F. nucleatum subsp. 4_8 reference genome or 1,842,979 positions), there were no single nucleotide polymorphisms (SNPs) separating T2 and O2. Of the reference genome positions where high-quality read mapping was obtained for both Oral003 and CRC003 (82% of the F. nucleatum subsp. 4_8 reference genome, or 1,857,684 positions), Oral003 and CRC003 were separated by only 183 SNPs. Of the reference genome positions where high-quality read mapping was obtained for both JAoral001 and JAcrc001 (61% of the F. nucleatum subsp. 4_8 reference genome, or 1,371,204 positions), JAoral001 and JAcrc001 were separated by only 2 SNPs. This comparison was performed over a smaller fraction of the reference genome because JAoral001 and JAcrc001 are members of subspecies polymorphum, not animalis. In comparison, the next most closely homologous fusobacterial genomes, 7-1 and CTI-1, isolated from inflamed biopsy tissue from a Crohn's patient and a CRC tumor, respectively, were separated by 9,553 SNPs ( Figure 1C). Thus, our data reveal an extremely close evolutionary relationship between T2 and O2, CRC003 and Oral003, and between JAcrc001 and JAoral001.
These data indicate that the strains from each of the paired samples most likely shared a recent common ancestor within each patient and support the hypothesis that fusobacteria from the oral cavity may seed and become enriched in colorectal cancers. Our results support a previous report that used random primed PCR and found very close homology between F. nucleatum isolated from CRC and oral cavity of patients (Komiya et al., 2018).

Oral Fusobacteria Reach CRC via the Blood Circulation
While our data suggest that CRC fusobacteria may originate from the oral fusobacterial community, the route of their oral to CRC transmission remained to be resolved. Our prior studies indicated that fusobacteria can reach the colon tumors by descending from the oral cavity (Kostic et al., 2013) or through the blood (Abed et al., 2016), perhaps as can occur during frequent gingival injury and bleeding (Ashare et al., 2009). Such hematogenous transfer of oral fusobacteria to the placenta was demonstrated previously and may explain the high occurrence of this organism in preterm births (Han et al., 2004).
To resolve the preferred oral-CRC route, we used the C57BL/6 mouse MC38 colon cancer model. As the intrarectal MC38 CRC model has not been previously used in studies of fusobacteria, we first confirmed that fusobacteria can bind to MC38 cells. N-Acetyl-D-galactosamine (Gal-GalNAc) is displayed at high levels on CRC (Yang and Shamsuddin, 1996) and is correlated with CRC attachment and colonization by fusobacteria (Abed et al., 2016). First, we tested the Gal-GalNAc levels displayed by MC38 cells using FITC-labeled peanut agglutinin (PNA), a Gal-GalNAc [Gal-β(1 → 3)GalNAc] specific lectin and confirmed that MC38 cells display high levels of Gal-GalNAc. We employed human CRC HT29 cells, which display low Gal-GalNAc levels (Abed et al., 2016), as a negative control (Figure 2A). Next, we confirmed the activity of Fap2, the fusobacterial lectin that mediates attachment to CRC-displayed Gal-GalNAc (Abed et al., 2016), in the oral and tumor isolates using standard hemagglutination ( Figure 2B). Given these encouraging findings with respect to MC38 cells and the fusobacterial isolates, we proceeded with experiments wherein we implanted MC38 cells stably transfected with the luciferase (luc) gene under the mucosa of the rectum of C57BL/6 wild-type mice. Tumor volume was monitored both by real-time imaging of luciferase expression and by direct measurement of rectal tumors. At day 9 posttumor implantation, mice were randomized to the oral or the intravascular (IV) inoculation groups and inoculated with two of the new oral F. nucleatum isolates O2 or Oral003 according to the experimental scheme presented in Figure 2C. Oral inoculation was performed at day 9 via gavage of ∼10 10 bacteria and repeated two additional times (on days 12 and 15). Intravascular inoculation was performed at day 15, by a single intravenous injection of 5 × 10 6 -1 × 10 7 fusobacteria via the tail vein. Tumors and adjacent non-cancerous colon samples were harvested 24 h after the last oral or single intravenous inoculation, and fusobacteria were enumerated by plating and by qPCR.
CRC colonization with IV inoculation was far more efficient than with gavage of the oral fusobacteria. Fusobacteria could be cultured from the CRC tumors of all 12 mice IV inoculated with O2 (mean 4.9 × 10 3 CFU/gr tissue) and from 9 of the 10 mice inoculated IV with Oral003 (mean 5.7 × 10 3 CFU/gr tissue). In contrast, fusobacteria could not be cultured from any mice orally inoculated with O2 (n = 10 mice) or with Oral003 (n = 10 mice) ( Figure 2D). Fusobacterial gDNA was detected and quantified in the CRC tumors of all above 12 mice IV inoculated with O2 and in 9 out of the 10 inoculated with Oral003. In contrast, fusobacterial DNA was below the limit of detection in the CRC tumors of all 10 mice inoculated by oral gavage with O2 and in 9 out of the 10 mice inoculated with Oral003 ( Figure 2E). These results were repeated using a different F. nucleatum sub species, F. nucleatum ATCC 23726 subsp. nucleatum and a different mouse model, the CT26-luc colon cancer model in the BALB/cJ mouse ( Figure 3A). Once again, the hematogenous route (a single injection of 5 × 10 7 -1 × 10 8 fusobacteria) appeared much more efficient than the digestive tract for fusobacteria trafficking to CRC orthotopic tumors. Relative fusobacterial gDNA abundance revealed that gavage-inoculated F. nucleatum ATCC 23726 was 1.13 × 10 5 fold less frequent in the colon cancer as compared to intravascular injected ones, as quantified by qPCR (mean 2 − Ct of 1.356 × 10 −7 vs. 0.01538, respectively) ( Figure 3B). Increasing the dose of the IV injected bacteria 10-fold from 5 × 10 6 -1 × 10 7 used in the MC38 model to 5 × 10 7 -1 × 10 8 used in the CT26 model, resulted in higher levels of fusobacteria in the CRC. However, these differences might also result from the different CRC models and fusobacterial strains used. Intravenously-inoculated fusobacteria (B) Hemagglutination demonstrating Fap2-dependent Gal-GalNAc binding by matching oral and CRC isolates O2/T2 with Oral003/CRC003 and JAoral001/JAcrc001 in the absence (left) and in the presence (right) of 25 mM GalNAc. The F. nucleatum ATCC 23726 Fap2 inactivated mutant K50 was used for negative control. Non-hemagglutinated erythrocytes settle in the bottom of the round bottom well. (C) in vivo experimental scheme of the orthotopic rectal MC38-luc mouse CRC model. At day 9 post-tumor implantation mice were randomized to an oral or intravenous inoculation group. Oral inoculations were performed on day 9, 12, and 15. A single intravenous inoculation was performed on day 15. (D,E) CRC colonization by hematogenously or orally administered fusobacteria. (D) Abundance (CFU/gr tissue) and relative fusobacterial gDNA abundance (2 − Ct ) (E) in tumor (T) samples and in adjacent normal (N) colon samples from MC38 transplanted mice inoculated once with 5 × 10 6 -1 × 10 7 intravenously (IV) or three times by oral gavage (Gavage) with 10 ∼10 F. nucleatum O2 or Oral003. ****p < 0.0001, Bonferroni-corrected two-tailed Mann-Whitney test. ***p < 0.001 Bonferroni-corrected two-tailed Mann-Whitney test for gavage vs. IV, Bonferroni-corrected one-tailed Wilcoxon test for tumor vs. normal. **p < 0.01 Bonferroni-corrected one-tailed Wilcoxon test. n.s., not statistically significant. Each symbol represents data from individual mice. Error bars show Mean ± SEM.
were found in mouse CT26 CRC tumors 2 h post-delivery and levels remained stable at 6 h post-infection. This was followed by fusobacterial proliferation in the tumor that continued 24 and 72 h post-infection (Figures 3C,D).
The magnitude of bacteremia resulting from a dental procedure and from routine daily activities is significantly lower (<10 4 CFU/ml) (Wilson et al., 2008) than that tested in our experiments. We therefore tested whether fusobacteria inoculated in doses mimicking these physiological ones (5 × 10 3 -1 × 10 7 ) would be found in CRC tumors. We employed the orthotropic MC38 CRC model and F. nucleatum JAoral001 in these experiments. CRC tumor-associated fusobacteria could be detected in mice inoculated across the more physiologic dose range. Increasing the dose resulted in increases in the proportion of mice with fusobacteria detected in the tumors. Fusobacteria were found in the tumors of 45% of mice inoculated with 5 × 10 3 -1 × 10 4 , 60% inoculated with 5 × 10 4 -1 × 10 5 , and 100% inoculated 5 × 10 6 -1 × 10 7 (Figures 4A,B). These results may explain the heterogeneity observed in fusobacterial occurrence and levels in human CRC.

DISCUSSION
Our results, admittedly in a limited subset of patients, support the concept that fusobacteria, which are abundant in CRC and play a role in prognosis and in treatment outcome, may originate from the oral cavity. In a pre-clinical model, the circulatory system appears to be the most efficient route by which fusobacteria reach colon tumors. In humans, this most likely occurs during transient physiologic bacteremias originating from the mouth.
There are many diseases associated with oral bacteria disseminating to a distant site as a result of transient bacteremia and the paragon is infective endocarditis (IE). Oral viridans group streptococci are implicated as causal organisms in 35-45% of cases (Dayer et al., 2015). Administration of antibiotic prophylaxis before invasive dental procedures for prevention of IE has been practiced for over 50 years and remains the global standard of care for patients at risk (Dayer et al., 2015). Transient bacteremia is common with manipulation of the teeth and periodontal tissues. However, there is a wide variation in the reported frequencies of bacteremia in patients resulting from dental procedures: tooth extraction (10-100%), periodontal surgery (36-88%), scaling and root planing (8-80%), and teeth cleaning (up to 40%) [For review see Wilson et al. (2008)].

FIGURE 4 | Efficiency of CRC colonization by hematogenous fusobacteria is dose-dependent. (A)
Relative fusobacterial gDNA abundance (2 − Ct ), and proportion CRC colonization (B) in tumor samples from MC38 transplanted mice intravenously inoculated daily three times with 5 × 10 3 -1 × 10 4 , 5 × 10 4 -1 × 10 5 or with 5 × 10 6 -1 × 10 7 F. nucleatum JAoral001. Each symbol represents one mouse. Bar show median. *p < 0.05, the proportion of mice with fusobacterial-colonized tumors increases significantly with the amount of bacteria injected, as determined by an exact one-sided Permutation Test implemented by StatXact* (p = 0.0442 with the scores set equal to the amount of bacteria, p = 0.04633 with the scores equal to the log).
Transient bacteremias also occurs during frequent daily activities such as tooth brushing and flossing (20-68%) and chewing food (7-51%) (Wilson et al., 2008). Thus, the frequency of bacteremia from routine daily activities is far greater than that from dental treatment.
In our experimental data shown in Figure 4B, lowering the fusobacterial inoculation dose did not abolish CRC colonization but rather reduced its efficiency (measured as frequency of CRC-fusobacteria colonized mice) in a dose-dependent manner. A large body of evidence suggests that the abundance of oral bacteria in the blood during transient bacteremia is approximately 10 4 CFU/ml. Of course, not all oral bacteria are fusobacteria and the average blood volume of an adult 70 kg male is ∼5 L.
Resistance to killing by phagocytes in the blood might be a mechanism to advance hematogenous dissemination by bacteria. F. nucleatum can modulate neutrophil function to increase tissue damage (Katsuragi et al., 2003) and deplete neutrophils at the site of oral infection (Jewett et al., 2000;Wu et al., 2018a). However, a report of microbicidal assays indicated that <98% of F. nucleatum organisms were killed by PMNs within 60 min (Mangan et al., 1989). Complicating the matter further, different strains of F. nucleatum impact neutrophil responses differently (Kurgan et al., 2017). Therefore, it remains to be elucidated whether F. nucleatum is specifically adapted for survival while in transit in host blood to avoid killing by neutrophils.
We do not know if fusobacteria within the bloodstream travel to the tumor as free bacteria or bound to erythrocytes which fusobacteria are well-known to agglutinate. As fusobacteria are non-motile, it is plausible that the altered vasculature of tumors with enhanced blood vessel permeability may assist fusobacteria in reaching tumors. Unlike other bacteria that may arrive in the vicinity of a tumor in a similar manner, fusobacteria are equipped with the Fap2 lectin that binds Gal-GalNAc, which is specifically and highly displayed on many colon tumors (Abed et al., 2016), enabling fusobacteria to bind, persist, and colonize. Tumor-induced local hypoxia and immune effectors might be additional factors that provide fusobacteria with a competitive advantage in the CRC tumor milieu.
Our observation that the intravenous route for CRC colonization by oral fusobacteria is more efficient than a gastrointestinal route in orthotropic CRC mouse models cannot rule out that oral fusobacteria constantly swallowed in humans may colonize CRC through the digestive tract. However, hematogenous spread of oral fusobacteria to CRC is biologically more plausible, as travel through the bloodstream circumvents the toxicity of low gastric pH as well as bile acids encountered when descending through the gastrointestinal tract. Furthermore, bloodstream travel affords fusobacteria an escape from competition with the endogenous colonic microbiota. Also, genomic studies have not uncovered type III secretion systems as part of the F. nucleatum pan-genome that would facilitate epithelial cell invasion (McGuire et al., 2014). Tumor arrival via the circulatory system might also explain fusobacterial absence from biofilms in the colonic mucus layer of adenomas of familial adenomatous polyposis patients and some sporadic CRC patients (Dejea et al., 2018).
F. nucleatum is currently classified into five subspecies: F. nucleatum subsp. nucleatum, F. nucleatum subsp. polymorphum, F. nucleatum subsp. fusiforme, F. nucleatum subsp. vincentii, and F. nucleatum subsp. animalis. On average, human adults carry three F. nucleatum sub species in their oral cavity (Henne et al., 2018). Two out of the three CRC F. nucleatum isolated by us (T2 and CRC003), belong to F. nucleatum subsp. animalis. Our results support previous reports that F. nucleatum subsp. animalis is the most prevalent F. nucleatum subspecies in human colorectal cancers (Ye et al., 2017). Interestingly, F. nucleatum subsp. animalis is not the most prevalent F. nucleatum subspecies in the oral cavity (Henne et al., 2018) suggesting that F. nucleatum subsp. animalis may be better adapted for CRC colonization. A recent study suggested that fusobacteria travel intracellularly with the primary tumor cells to distant sites, as part of metastatic tissue colonization (Bullman et al., 2017). We previously showed that Gal-GalNAc is over displayed in CRC metastases (Abed et al., 2016) that may facilitate fusobacterial attachment via Fap2. Our results here suggest that fusobacteria can also reach metastases from the primary tumor or from the oral cavity via the circulatory system.
In conclusion, it is our hope that understanding the mechanisms by which fusobacteria colonize colorectal cancers and accelerate their development and progression may provide a means to more effectively detect, prevent and treat CRC.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://www.ncbi.nlm.nih. gov/bioproject/?term=PRJNA362799, PRJNA362799.

ETHICS STATEMENT
The animal study was reviewed and approved by the Hebrew University of Jerusalem Ethics Committee. The studies involving human participants were reviewed and approved by Hadassah Medical School IRB. The patients/participants provided their written informed consent to participate in this study.