Research ArticleHost-Microbe Interactions

Fusobacterium nucleatum host-cell binding and invasion induces IL-8 and CXCL1 secretion that drives colorectal cancer cell migration

See allHide authors and affiliations

Science Signaling  21 Jul 2020:
Vol. 13, Issue 641, eaba9157
DOI: 10.1126/scisignal.aba9157

Moved to metastasize by gut microbiota

Bacterial dysbiosis in the gut, particularly an increase in Fusobacterium nucleatum, is associated with colorectal cancer (CRC). Casasanta et al. found that CRC cell–resident F. nucleatum promotes proinflammatory cytokine secretion that stimulates tumor cell migration and invasion. Gene deletion revealed that F. nucleatum infection of HCT116 CRC cells required a membrane adhesin protein that did not mediate bacterial infection of immune cells. Infected CRC cells secreted the cytokines IL-8 and CXCL1 that promoted the invasive motility of infected and noninfected cells. The findings suggest that preventing F. nucleatum invasion of tumor cells may reduce bacterial-associated gut inflammation and metastatic progression in CRC patients.


Fusobacterium nucleatum is implicated in accelerating colorectal cancer (CRC) and is found within metastatic CRC cells in patient biopsies. Here, we found that bacterial invasion of CRC cells and cocultured immune cells induced a differential cytokine secretion that may contribute to CRC metastasis. We used a modified galactose kinase markerless gene deletion approach and found that F. nucleatum invaded cultured HCT116 CRC cells through the bacterial surface adhesin Fap2. In turn, Fap2-dependent invasion induced the secretion of the proinflammatory cytokines IL-8 and CXCL1, which are associated with CRC progression and promoted HCT116 cell migration. Conditioned medium from F. nucleatum–infected HCT116 cells caused naïve cells to migrate, which was blocked by depleting CXCL1 and IL-8 from the conditioned medium. Cytokine secretion from HCT116 cells and cellular migration were attenuated by inhibiting F. nucleatum host-cell binding and entry using galactose sugars, l-arginine, neutralizing membrane protein antibodies, or fap2 deletion. F. nucleatum also induces the mobilization of immune cells in the tumor microenvironment. However, in neutrophils and macrophages, the bacterial-induced secretion of cytokines was Fap2 independent. Thus, our findings show that F. nucleatum both directly and indirectly modulates immune and cancer cell signaling and migration. Because increased IL-8 and CXCL1 production in tumors is associated with increased metastatic potential and cell seeding, poor prognosis, and enhanced recruitment of tumor-associated macrophages and fibroblasts, we propose that inhibition of host-cell binding and invasion, potentially through vaccination or novel galactoside compounds, could be an effective strategy for reducing F. nucleatum–associated CRC metastasis.


The role of bacteria and viruses in the onset and progression of diverse cancers is well established (13). Recent studies on the oral, anaerobic, gram-negative bacterium Fusobacterium nucleatum in the acceleration of colorectal cancer (CRC) pathogenesis have revealed as many questions as they have answered for the mechanisms and proteins this bacterium uses to potentiate disease (49). An overarching question in the field is: How does F. nucleatum enter and reside in tumors after likely arriving via the bloodstream from its native oral cavity? In addition, a second theme of pathogenesis that remains understudied is: Are these bacteria capable of leaving the primary tumor on or within immune or cancerous cells to seed and accelerate metastatic cancer sites (10)? Answering these questions will be key in understanding both the host and bacterial mechanisms at play in microbe-accelerated cancers. Two studies reported that F. nucleatum directly induces cancer cell metastasis through nuclear factor-κB (NF-κB)–mediated increased expression of keratin 7 (KRT7) (11), as well as increased expression of caspase activation and recruitment domain 3 (CARD3), and down-regulation of E-cadherin (12). Our study here provides additional insights into the upstream bacterial binding as well as downstream chemokine/cytokine signaling that contribute to F. nucleatum–induced cellular migration.

Chemokines/cytokines play a crucial role in tumor initiation, progression, and metastasis (13). Initially discovered as chemotactic mediators of leukocytes, they are now known to be secreted by several cell types and can be expressed constitutively or induced by inflammatory stimuli, including bacterial infections, and function in a variety of roles including cell survival, proliferation, angiogenesis, and cell migration. In cancer, chemokines mainly function in regulating angiogenesis, activating tumor-specific immune responses, and directly stimulating the tumor through autocrine or paracrine mechanisms (13). In our study presented here, the results indicate that direct binding and invasion of host cancer and immune cells by F. nucleatum induce the secretion of the proinflammatory and prometastatic cytokines interleukin-8 (IL-8) and C-X-C motif chemokine ligand 1 (CXCL1) and that conditioned medium from F. nucleatum–infected HCT116 CRC cells causes non–Fusobacterium-exposed cells to chemotactically migrate as well.

The cytokines IL-8 (also known as CXCL8) and CXCL1 [also known as growth-regulated oncogene α (GROα)] play a pivotal role in CRC progression (14, 15). Multiple studies have characterized their role in influencing CRC invasiveness. Increased expression of CXCL1 (measured by immunohistochemistry in tissue micro‑arrays) is correlated with cancer progression and metastasis and ultimately poor prognosis in patients with CRC, thus indicating its potential as a biomarker for CRC (16). CXCL1 is an autocrine growth factor that binds to the receptor CXCR2 with high affinity. Many colorectal adenocarcinoma cell lines (LS174T, KM12L4, KM12C, and SW480) constitutively express CXCL1, and it is well known that high levels of this cytokine increase invasive potential (17). In addition, antibodies to CXCL1 or CXCR2 have been shown to inhibit colon cancer cell proliferation (18). It is relevant for these studies that nonmetastatic Caco2 and low-metastatic HT29 cell lines expressed lower levels of CXCL1 than the highly metastatic cancer cell line LS147T. In another study, Ogata et al. (19) showed that CXCL1 increases the number of invasive DLD-1 and LoVo cells and that these effects are quenched in the presence of anti-CXCL1 antibody . Furthermore, there is emerging evidence to indicate that CXCL1 participates in premetastatic niche formation in liver tissue, which, in turn, recruits CXCR2-positive myeloid-derived suppressor cells to support liver metastases of CRC (20).

IL-8 is a ubiquitously prevalent cytokine in CRC, where it has been characterized as the most potent chemoattractant and activator of neutrophils in both in vivo and in vitro studies. IL-8 expression at both the mRNA and protein levels is substantially up-regulated in pathological colorectal tissue liver metastases (21), and the abundance of both IL-8 and its receptor CXCR2 is increased in colon cancer (22). In addition, increased levels of IL-8 in the serum and the tumor microenvironment enhance the growth of human and mouse colon cancer cells in vivo and promote the dissemination of cancer cells to the lung and liver (23). IL-8 can bind to both CXCR1 and CXCR2, but it exerts different effects upon binding to either receptor (24); binding to CXCR1 induces neutrophil migration, whereas binding to CXCR2 modulates angiogenic activity (13). The angiogenic effect of IL-8 promotes tumor growth by providing access to oxygen and nutrients, as well as an opportunity to metastasize.

Our initial goal was to investigate the role of outer membrane adhesins in F. nucleatum direct binding and invasion of cancer cells to determine whether this was critical for altered cell signaling. Multiple adhesins have been characterized in F. nucleatum binding and signaling, with important roles for the small multimeric adhesin FadA (6, 25) as well as the large, outer membrane autotransporter adhesin Fap2 (2527). Fap2 docks with host cells through Gal/GalNAc sugar residues, which are overexpressed on CRC cells, as well as through protein-protein interactions with the inhibitory receptor TIGIT (T cell immunoglobulin and ITIM domain) on natural killer cells (28). However, aside from these two adhesins, most outer membrane proteins of F. nucleatum have not been characterized in the bacterium’s interaction with cancer cell. Here, we expanded upon these analyses by developing a new, modified version of a galactose kinase markerless gene deletion system capable of creating strains with unlimited gene deletions. We implemented this system to functionally characterize the role of FadA, Fap2, and multiple uncharacterized type 5c trimeric autotransporter adhesins (CbpF, FvcB, FvcC, and FvcD) (29, 30). Our studies reveal that the invasion of HCT116 cells by F. nucleatum—which we found can be inhibited by small molecules, antibodies, or adhesin gene deletions—is critical to the induction of proinflammatory signaling that promotes CRC cell motility.


F. nucleatum 23726 outer membrane adhesins are critical for the binding and invasion of HCT116 CRC cells

Previous studies established that F. nucleatum is highly invasive and can undergo a non-obligate intracellular life stage within epithelial cells, endothelial cells, keratinocytes, and potentially immune cells (3134). We first confirmed the invasive potential of F. nucleatum subsp. nucleatum [American Type Culture Collection (ATCC) strain 23726] into HCT116 CRC cells using fluorescence microscopy (Fig. 1, A to C), and flow cytometry (Fig. 1, D to E), and antibiotic protection assays (Fig. 1F) by following surface-bound and intracellular F. nucleatum labeled with a fluorescent green lipid or membrane antisera to detect intracellular and surface-bound bacteria (Fig. 1, A to C). To then gain a deeper understanding of how F. nucleatum may contribute to the acceleration of CRC, we investigated the importance of direct interactions between F. nucleatum and cancer cells and verified bacterial cancer cell interactions, which are driven by bacterial outer membrane adhesins. Our goal was to confirm the roles of FadA and Fap2 in the bacterium’s binding to, and induction of signaling in, HCT116 cells as well as characterize multiple type 5c trimeric autotransporter adhesins that have a well-established role in the virulence potential of other gram-negative bacteria, such as Yersinia (35, 36). In previously reported bioinformatics, we identified five type 5c adhesins in the strain F. nucleatum 23726, with four of these (CbpF, FvcB, FvcC, and FvcD) containing all the classic domains that make up a complete adhesin (25). The fifth protein (FvcE) lacks a “head” domain that is predicted to coordinate adhesion (25) and was therefore not explored here. To characterize these outer membrane adhesins, we developed completely markerless, single-gene knockouts of fadA, fap2, cbpF, fvcB, fvcC, and fvcD in F. nucleatum 23726, as well as multiple gene deletions per strain (tables S1 to S3) using a new version of a galactose kinase (GalK) genetic system (fig. S1) previously used in F. nucleatum (37), as well as many classical studies in Clostridium (38).

Fig. 1 F. nucleatum binds to and invades HCT116 cells.

(A) Overview of experiments used to analyze binding and invasion of Fnn and adhesin mutants in HCT116 CRC cells. (B) Cocultures of HCT116 and Fnn (2 hours, MOI 50:1 Fnn:HCT116) stained for intracellular and extracellular Fnn. Fnn was labeled with the fluorescent green membrane–intercalating dye FM 1-43FX to detect intracellular and extracellular bacteria and labeled with a pan-Fusobacterium membrane antisera (Alexa Fluor 594 goat anti-rabbit antibody) for extracellular bacteria. Host-cell nuclei are stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). Arrows indicate intracellular bacteria. (C) Fnn bacterium that is half intracellular, half extracellular in HCT116 cells. (D) Comparison of HCT116 binding and invasion by wild-type (WT) F. nucleatum 23726 with F. nucleatum 23726 ∆galKT (Fnn) (n = 3 independent flow cytometry experiments). (E) Binding and invasion analysis of Fnn and adhesin gene deletion strains using flow cytometry (n = 3 independent flow cytometry experiments). (F) Invasion and survival of Fnn and Fnnfap2 in HCT116 cells (n = 3 independent antibiotic protection assays). CFU, colony-forming units. In (E) and (F), infection parameters were 50:1 MOI Fnn:HCT116 for 4 hours (tan) and 0.5:1 MOI Fnn:HCT116 for 2 hours (purple). nsP > 0.05, *P < 0.05, and ****P < 0.0001 by unpaired Student’s t test or two-way ANOVA for single or grouped analyses, respectively.

This optimized genetic system relies on a base strain lacking the galK and galT gene operon, thereby rendering it sensitive to growth in the presence of galactose (figs. S2 and S3). The F. nucleatumgalKT strain (hereby called Fnn) has no growth defects (fig. S3), as well as not being attenuated for binding or invasion with CRC cells (Fig. 1D). Target gene deletion is initiated by construction of a chloramphenicol/thiamphenicol-resistant plasmid containing a constitutively active galK gene, as well as fused regions of homology upstream and downstream of the target gene [750 base pairs (bp)]. Electroporation of this plasmid into F. nucleatumgalKT and selection on thiamphenicol results in single-crossover homologous recombinations onto the chromosome at the target gene. Next, colonies are then selected for double crossover gene deletions that remove the target gene and plasmid containing thiamphenicol resistance and galactose sensitivity by selecting on galactose-containing medium (fig. S4). We verified that all genes are accurate, complete gene deletions, and show no polar effects of gene deletion via reverse transcription polymerase chain reaction (RT-PCR) of upstream and downstream genes (fig. S5).

We next found that deletion of fap2 (Fnnfap2) significantly attenuated binding to HCT116 cells (Fig. 1F) but that FnnfadA did not result in such a decrease in HCT116 interactions. Although FadA has been shown to drive interactions with cancer cells, we note that these studies were done in the strain F. nucleatum 12230 (33, 39, 40). Upon bioinformatic analysis, we found that F. nucleatum 23726 contained four fadA ortholog genes (fadA2, fadA3a, fadA3b, and fadA3c), and F. nucleatum 12230 encoded for only one ortholog (25). This could mean that a single deletion of the fadA gene in F. nucleatum 23726 remains adhesive and that multiple fadA family proteins are contributing to binding and invasion. In addition, an Fnnfap2 fadA double mutant did not further reduce binding when compared to Fnnfap2 (Fig. 1E). Concurrently, we analyzed single mutants of the trimeric autotransporter adhesins cbpF, fvcB, fvcC, and fvcD, as well as a strain lacking all four genes (Fnncbpf fvcBCD). FvcB, FvcC, and FvcD contribute to invasion, albeit less potently than Fap2. Fnnfap2 cbpf fvcBCD quintuple mutants showed no significant decrease in their invasion of HCT116 cells when compared to the single Fnnfap2 gene deletion strain. We note here that the insignificant binding loss we observed for the FnncbpF mutant could be due to low levels of CEACAM1 expression on HCT116 cells (41), given that CEACAM1 was previously identified as the receptor for F. nucleatum CbpF (CEACAM binding protein of Fusobacterium) (29). The proportion of intracellular Fnn and Fnn ∆fap2 determined by antibiotic protection assays (Fig. 1F) correlates with the flow cytometry data, indicating that the trypsinization used to prepare infected HCT116 cells likely removes most of the surface-bound bacteria. Together, these data indicate that host-cell binding and invasion is largely driven by Fap2.

F. nucleatum host-cell binding and invasion is critical for inducing the secretion of proinflammatory and prometastatic cytokines from cancer cells

After confirming the importance of outer membrane adhesins for invading HCT116 cells, we analyzed the ability of F. nucleatum to induce the secretion of cytokines from human cancer cells. Using cytokine arrays and enzyme-linked immunosorbent assays (ELISA) (Fig. 2, A to C), we found that F. nucleatum 23726 interactions with HCT116 cells induced the secretion of high concentrations (~1000 pg/ml after 4 hours) of the CXC family cytokines IL-8 and CXCL1 into the culture medium (Fig. 2D). To determine whether direct binding and invasion is necessary for induced cytokine secretion, as well as to rule out whether an unidentified secreted protein or molecule induces this phenomenon, we compared Fnn to Fnnfap2. Our results show that the ability of Fnnfap2 to induce cytokine secretion from HCT116 cells is significantly attenuated. This correlates well with reduced invasive potential and confirms that direct interaction is needed to alter host cells (Fig. 2D).

Fig. 2 F. nucleatum induces cytokine secretion from HCT116 cells.

(A) Schematic of experiments used to analyze Fnn-induced cytokine secretion from HCT116 CRC cells. (B and C) Representative broad cytokine array dot blots [of 36 proteins; (B)] analyzing effect of infection with Fnn and Fnn-adhesin deletion strains. Control spots (B) indicate successful Western blots and provide densitometry controls for quantitation, calculated as fold increase and shown as a heatmap in (C). (D) IL-8 and CXCL1 ELISA to quantitate cytokine secretion from HCT116 cells induced by Fnn and Fnnfap. n = 3 independent experiments. (E) IL-8 and CXCL1 ELISA to quantitate and compare cytokine secretion from HCT116 cells induced by F. nucleatum subsp. nucleatum 25586, F. nucleatum subsp. nucleatum 23726, and F. nucleatum subsp. animalis 7_1 (Fna). n = 3 independent experiments. In (D) and (E), infection parameters were 50:1 MOI Fusobacterium:HCT116 for 4 hours (tan) and 50:1 MOI Fusobacterium:HCT116 for 24 hours (purple). ****P < 0.0001 by unpaired Student’s t test or two-way ANOVA for single or grouped analyses, respectively.

To test whether this phenomenon was specific to strain F. nucleatum 23726, we compared the secretion of IL-8 and CXCL1 from HCT116 cells in response to F. nucleatum 25586 and 7_1, which cover subspecies nucleatum and animalis (Fna), respectively. We found that these strains also induce comparable, substantial levels of cytokine secretion (Fig. 2E). We note that both F. nucleatum 25586 and 7_1 contain the fap2 gene, potentially making their induction of cytokines also largely driven by Fap2-dependent binding to host cells. Last, we compared the 4-hour secretion levels to those at 24 hours in response to F. nucleatum 23726 and report an increase in CXCL1 and IL-8 to ~7000 and ~2500 pg/ml, respectively.

F. nucleatum–induced cytokine secretion in neutrophils and macrophages is not Fap2 dependent

Because bacterial-induced immune cell alterations have been implicated in a number of cancer processes (1, 2, 4), we next characterized bacterial-induced cytokine secretion from mouse immune cells using a mouse-specific cytokine array (Fig. 3A and fig. S7), similar to the process described for HCT116 cells. We first concurrently tested neutrophil cytokine secretion (Fig. 3B) and validated the bacterium’s direct binding to and invasion of neutrophils using fluorescence microscopy (Fig. 3C). F. nucleatum induces the secretion of CCL3, CXCL2, and tumor necrosis factor α (TNFα) from neutrophils, as measured by western blot (Fig. 3B) and ELISA (Fig. 3D). In mouse macrophages, F. nucleatum also induces robust CCL3, CXCL2, CCL5, and TNFα secretion (Fig. 3E), as well as additional cytokines when compared to either uninfected (no bacteria) or Escherichia coli–infected controls. CCL3 stimulates lymphocyte recruitment in early metastatic CRC (42), and CXCL2 promotes angiogenesis in the CRC microenvironment (43). Our data that show elongated extracellular neutrophil DNA (Fig. 3C) revealed that F. nucleatum could be inducing the formation of neutrophil extracellular traps (NETs), an alternative pathway for cytokine secretion (44). F. nucleatum and other oral bacteria were previously shown to degrade NETs, which is a potential escape mechanism from neutrophil-mediated death (45). In stark contrast to the Fap2-driven cytokine secretion seen in HCT116 cancer cells, Fnnfap2 did not cause a significant decrease in the amount of CCL3 and CXCL2 secreted from immune cells, as analyzed by cytokine arrays and ELISA (Fig. 3D and fig. S7). Therefore, we conclude that Fap2 selectively enables the bacterium’s binding and invasion of cancer cells, which is further supported by a lack of Gal/GalNAc sugar on immune cells. Last, because a major role of immune cells is to engulf and clear invading microorganisms, these cells have acquired the ability to recognize a broad range of bacterial ligands as endocytosis targets. Given that Fap2 does not drive immune cell interactions, our data indicate that the proteins or macromolecules on F. nucleatum and immune cells that coordinate these interactions remain unidentified.

Fig. 3 Cytokine secretion analysis from mouse neutrophils and macrophages.

(A) Overview of experiments used to analyze Fnn-induced cytokine secretion from immune cells. (B) Heatmap of cytokine array assessing secretion of the indicated cytokines from mouse neutrophils in the presence of E. coli, Fnn, or deletion control (columns as designated by colored dots, legend right). Infection parameters were 50:1 MOI Fnn:neutrophils for 4 hours. (C) Fluorescence microscopy of Fnn interacting with mouse neutrophils. DNA was detected using DAPI, and Fnn was labeled with the fluorescent red membrane–intercalating dye FM 4-64FX. Images are representative of n = 3 experiments. (D) ELISA confirming Fnn induction of CCL3 and CXCL2 and assessing the effect of deletion of fap2 on cytokine secretion in neutrophils. n = 3 independent experiments. (E) Cytokine array assessing the effect of Fnn infection on the secretion of several cytokines from mouse macrophages when compared to that of E. coli and control. Sample conditions designated by colored dots as defined in (B); relative abundance scale as indicated, right. In (B) to (E), infection parameters were 50:1 MOI Fusobacterium:neutrophils/macrophages for 4 hours. nsNot significant (P > 0.05) by unpaired Student’s t test.

Inhibition of bacterial invasion blocks cytokine secretion

Because Fap2 is a Gal/GalNAc sugar-binding lectin, and this sugar is overrepresented on the surface of CRC cells, we next tested a panel of galactose- and non–galactose-containing sugars for their ability to inhibit bacterial binding to HCT116 cells and consequent IL-8 and CXCL1 secretion. We found that galactose, GalNAc, and lactose (galactose disaccharide) potently inhibited the bacterium’s binding of and invasion into HCT116 cells, whereas control sugars—including glucose and maltose—did not (Fig. 4A). In addition, we tested l-arginine as an inhibitor because the F. nucleatum surface adhesin RadD, which is predominantly thought to coordinate interspecies bacterial interactions in the oral biofilm, is inhibited by l-arginine (4648). We found that l-arginine inhibited the invasion of F. nucleatum into HCT116 cells but not to the extent achieved by galactose sugars (Fig. 4A). All of the tested sugar molecules and l-arginine led to reduced secretion of IL-8 and CXCL1, with the most potent being galactose-containing sugars (Fig. 4B) as also seen in the binding and invasion assay. We believe that the addition of 10 mM glucose and maltose results in less potent and nonspecific inhibition of cytokine secretion or could be binding to an unidentified F. nucleatum lectin that may affect the host intracellular response after infection. Next, we developed a pan-Fusobacterium membrane antisera (FMAS; DJSVT_MAS1) with the goal of blocking HCT116 cell invasion, cytokine secretion, and migration. The pan-Fusobacterium membrane antisera significantly decreased cancer cell binding (Fig. 4C) and cytokine secretion (Fig. 4D). However, Fnnfap2 was still more highly attenuated for invasion and cytokine secretion when compared to the effect of the pan-Fusobacterium membrane antisera, indicating that the antisera were unable to inhibit all Fap2-HCT116 docking at the concentrations used.

Fig. 4 Inhibition of Fnn binding to and invasion of HCT116 cells using small molecules and membrane antibodies.

(A) Fnn invasion is significantly inhibited by galactose-containing sugars and l-arginine (10 mM). n = 3 independent flow cytometry experiments. (B) Effect of sugars and l-arginine (10 mM) on the secretion of IL-8 and CXCL1 from HCT116 cells. n = 3 independent experiments. (C) Fnn invasion in the presence of Fusobacterium membrane antisera (FMAS) versus control rabbit antisera (RAS). n = 3 independent flow cytometry experiments. (D) Effect of FMAS and Fnn ∆fap2 on secretion of IL-8 and CXCL1 from HCT116 cells. n = 3 independent experiments. nsP > 0.05, *P < 0.05, ***P < 0.001, and ****P < 0.0001 by unpaired Student’s t test or two-way ANOVA for single or grouped analyses, respectively.

In addition, we tested the autophagy and endosome maturation inhibitor chloroquine for its ability to inhibit bacterial invasion and consequent cytokine signaling (fig. S8). This experiment was performed because the two previous studies reporting that F. nucleatum induced metastasis showed that chloroquine inhibited CRC cell migration (11, 12); thus, we set out to test whether this might be due to reduced cytokine levels that we observed. However, we found that chloroquine does not inhibit bacterial invasion when applied at 10 μM and only slightly decreases invasion when applied at 50 μM (fig. S8A). Curiously, we saw the reverse effect of chloroquine (50 μM) on cytokine signaling, with a slight increase in secreted IL-8 and CXCL1 induced by Fnnfap2 (fig. S8B). We elaborate on this finding below in our discussion of the results, but certainly, the mechanism of chloroquine’s effects—and, seemingly separately, the intracellular mechanism(s) of cytokine induction observed here—in modulating F. nucleatum–associated CRC cell motility and metastasis requires further investigation.

F. nucleatum–conditioned supernatants from HCT116 infections are enriched in IL-8 and CXCL1 and cause cancer cell migration

We first tested whether F. nucleatum–conditioned culture medium from HCT116 infections could cause noninfected cells to migrate. We show through transwell assays (Fig. 5, A to C) that F. nucleatum–conditioned HCT116 culture medium significantly induces non–Fusobacterium-exposed cancer cells to migrate starting at 8 hours, with increasing significance through 16-hour incubations with conditioned medium (Fig. 5D). Not only did F. nucleatum subsp. nucleatum–conditioned media induce cellular migration but also F. nucleatum subsp. animalis (Fna) induced significant migration in HCT116 cells (Fig. 5E). We subsequently found that conditioned medium from cells infected with Fnnfap2, which led to lower levels of IL-8 and CXCL1 secretion, resulted in significantly reduced cell migration compared to that induced by Fnn-infected cell-conditioned medium (Fig. 5E). We were able to reduce cellular migration by adding the pan-Fusobacterium membrane antisera at 1:250 dilution to the Fnn-conditioned medium, but this concentration did not result in a significant decrease in migration. This could be due to insufficient antibody concentrations added to the assay, or because the antisera were developed with 11 strains of Fusobacterium, its specific inhibitory effects on F. nucleatum 23726 have been distributed across non-Fnn proteins.

Fig. 5 HCT116 migration is driven by Fnn-induced secretion of CXCL1 and IL-8.

(A and B) Experimental setup (A) and representative three-dimensional confocal imaging (B) of HCT116 transwell migration experiments. Blue, DAPI-stained DNA; red, CellTracker Red; white arrows, cross-barrier migrated cells. (C) Representative images of migrated HCT116 cells after a 16-hour exposure to the indicated conditioned medium. n = 3 independent experiments. (D) HCT116 cellular migration when cultured in Fnn-conditioned medium containing high CXCL1 and IL-8 concentrations. n = 3 independent experiments. (E) Effect of deletion of fap2 (Fnnfap2) and Fusobacterium membrane antisera (FMAS) on bacterial-induced HCT116 cellular migration. n = 3 independent experiments. (F) Effect of the addition of purified IL-8 and CXCL1 to culture medium within the lower transwell chamber on HCT116 cell migration. n = 5 independent experiments. (G) Analysis of CXCL1 and IL-8 levels after concentration, with and without antibody depletion of cytokines and before loading into the bottom chamber of the transwell apparatus. n = 3 independent experiments. (H) HCT116 cellular migration upon depleting CXCL1 and IL-8 from medium versus controls. n = 3 independent experiments. nsP > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by unpaired Student’s t test or two-way ANOVA for single or grouped analyses, respectively.

Last, we found that adding purified cytokines IL-8 and CXCL1 stimulated HCT116 cellular migration (Fig. 5F), whereas bead-conjugated antibody-mediated depletion of IL-8 and CXCL1 (Fig. 5G) from F. nucleatum–conditioned culture medium significantly reduced the number of migrated cells (Fig. 5H). Because the depletion of IL-8 and CXCL1 did not completely abolish Fnn-induced cell migration, we hypothesize that F. nucleatum modulates multiple additional host proteins and pathways involved in cellular migration and metastasis.


Studies have illuminated a strong correlation between the presence of F. nucleatum within colorectal tumors and increased tumor microenvironment inflammation, enhanced oncogenic signaling, and poor patient prognosis (4952). To address the role of F. nucleatum in tumor microenvironment signaling, we set out to characterize potential signaling pathways that drive cellular migration and immune cell–driven cancer acceleration. Our data show that direct F. nucleatum cellular interactions with cancer and immune cells, largely coordinated in CRC cells (minimally the HCT116 line used here) by the outer membrane adhesin Fap2, drive host cells to define a locally prometastatic, inflammatory microenvironment. The ability of this bacterium to cause selective induction of the proinflammatory and prometastatic cytokines IL-8 and CXCL1 from CRC cells complements previous studies in which these gross phenotypes of oncogenic acceleration from human patients and mouse models are consistently reported (10, 53, 54). In addition, our findings add to a growing understanding that F. nucleatum directly induces metastasis through cytokine release, increased NF-κB expression, and subsequent expression of KRT7 (11), increased CARD3, and down-regulation of E-cadherin (12). These studies have now shown active roles for F. nucleatum in mouse models of metastasis, which complement the seminal study by Bullman et al. (10) first reporting viable F. nucleatum within human CRC cell metastases in the liver (Fig. 6).

Fig. 6 A model of F. nucleatum–induced metastasis through cytokine signaling.

Model of our findings, that Fap2-dependent F. nucleatum invasion into CRC cells induces IL-8 and CXCL1 secretion, and both cytokines have been characterized as key players in cancer metastasis and subsequent downstream cell seeding. HCT116-derived IL-8 and CXCL1 can participate in autocrine signaling back to cancer cells as a metastatic signal, as well as paracrine signaling to recruit neighboring immune cells, which further secrete their own cytokine signatures (CXCL2, CCL3, and TNFα) that alter the tumor microenvironment through metastatic, inflammatory, and immune cell programming. Our data show that Fap2 drives CRC cell docking, which is a key step in initiating the prometastatic cytokine cascade but is not necessary for immune cell signaling.

CXCL1 and IL-8 both play roles in immune cell recruitment, particularly neutrophil recruitment and programming. Through paracrine signaling, F. nucleatum interactions with CRC cells could be releasing factors that create not only a metastatic environment but also one that provides protumor inflammation. There is the potential for recruited neutrophils to be activated to tumor-associated neutrophils (TANs) that could further accelerate tumor progression. In addition, if F. nucleatum or the tumor microenvironment induces NET formation, these structures have been shown to increase metastasis and sequester circulating tumor cells to promote reseeding (5557). Last, it was shown that high intratumor loads of F. nucleatum resulted in lower CD3+ T cell density in CRC, providing another layer to the complexity of F. nucleatum immune cell regulation (58). To address these potential scenarios, a priority should be placed on investigating the potential for this bacterium to induce the formation of additional pro-oncogenic cell types including cancer-associated fibroblasts (59), tumor-associated macrophages (60), and TANs (55).

Bacteria such as Streptococcus gallolyticus are known to induce the secretion of IL-8 in colon tumor cells (61). It has also been shown that increased expression of IL-8 leads to a significant resistance to the cytotoxic effects of oxaliplatin, a platinum-based chemotherapeutic drug (62). As F. nucleatum was previously shown to induce chemoresistance through the induction of autophagy and subsequent inhibition of apoptosis (51), our report of the induction of IL-8 and CXCL1 by this bacterium adds another dimension to this story.

Our data definitively show the importance of F. nucleatum binding and invasion of human cancer cells in producing prometastatic cytokines. We and others have proposed that the development of small-molecule compounds or proteins that are able to block F. nucleatum docking to host cells could be effective strategies to reduce cancer and immune cell signaling that accelerates metastasis from the colon. A starting point could be the selective inhibition of Fap2, which we have now shown drives binding and cytokine secretion. Fap2 was previously characterized as a lectin and shown to bind to Gal/GalNac sugars that are in abundance on the surface of most cancer cells, especially colorectal (63). The potential for targeting of bacterial lectins for disease treatment has been validated in urinary tract infections by using mannoside sugar derivative compounds that block the uropathogenic E. coli lectin FimH from binding to host receptors, thereby resulting in depletion of these pathogenic bacteria from the gut (64).

Two studies have reported that chloroquine blocks metastasis by lowering F. nucleatum–induced NF-κB–driven expression of KRT7, a protein linked to increased cancer cell motility and invasive potential (65, 66), and overexpression of CARD3, a prometastatic kinase previously characterized in breast cancer (67). We were therefore interested in whether the IL-8– and CXCL1-regulated cell migration we have observed might be mitigated by chloroquine administration. Our results show that chloroquine treatment actually increased the secretion of IL-8 and CXCL1, suggesting that multiple pathways may be at play in regulating F. nucleatum–driven metastasis (fig. S8). This outcome is also the opposite of the effect chloroquine has on reducing the amount of IL-8 secreted upon Campylobacter jejuni infections (68). Together, these observations could mean that the IL-8 and CXCL1 secretion we observed is not caused by a Toll-like receptor (TLR) maturation–dependent mechanism, which chloroquine regulates (68, 69), but is still NF-κB dependent. In addition, inhibiting autophagy could lead to increased intracellular survival of F. nucleatum. This should be characterized further before these inhibitors are considered for use as more than chemical tools to understand host biological pathways.

Last, because the consensus belief is that F. nucleatum leaves the oral cavity and traverses the human body through the blood and potentially lymph, it could be advantageous to use a vaccine-based strategy whereby antibodies block and clear this bacterium before leaving the bloodstream, thereby preventing downstream tumor interactions. Our data using a pan-Fusobacterium membrane antisera show effective blocking of F. nucleatum entry into cancer cells, but this strategy needs refinement to increase strain-specific potency. Hence, interfering with F. nucleatum interactions in the human body could be a targeted approach that provides an alternative to using nonspecific antibiotics. Last, we hypothesize that future studies characterizing how F. nucleatum intricately influences multiple cell signaling pathways during cancer progression and metastasis will lead to targeted approaches for controlling additional oncomicrobes in cancer.


Culturing F. nucleatum

F. nucleatum subsp. nucleatum ATCC 23726, F. nucleatum subsp. nucleatum ATCC 25586, and F. nucleatum subsp. animalis 7_1 (Fna) were cultured on solid agar plates made with Columbia broth (Gibco) substituted with hemin (5 μg/ml) and menadione (0.5 μg/ml) (CBHK) under anaerobic conditions (90% N2, 5% H2, 5% CO2) at 37°C. Liquid cultures started from single colonies were grown in CBHK medium under the same conditions. For infections, overnight cultures were back-diluted 1:1000 and grown to mid-exponential phase [~0.5 OD600 (optical density at 600 nm)] before subsequent experiments unless otherwise noted.

Culturing HCT116 CRC cells

HCT116 (ATCC CCL-247) cells were purchased from ATCC and grown on tissue culture–treated plates and flasks in McCoy’s 5A medium supplemented with 10% fetal bovine serum (FBS), penicillin, and streptomycin. Cells were grown to no more than passage 15 at 37°C with 5% CO2. For maintenance between infections, cells were passaged by gentle trypsinization and reseeding. For infection experiments, cells were grown to >90% confluency in 6- or 24-well plates. Unless otherwise indicated, all experiments were conducted with HCT116 medium supplemented with 10% FBS.

Molecular cloning of a galactose-selectable gene deletion system

We have developed a new iteration of a galactose-selectable gene deletion system in F. nucleatum ATCC 23726 (Fnn) capable of deleting an unlimited number of genes in a single strain. We showcase the power of our genetic system by developing strains containing multiple gene deletions from ~300 bp to 12 kb, as well as chromosomally complementing tagged versions of these virulence factors for detection and purification. The plasmids used for these genetic mutations are all new, developed in house (table S2). This system is different than a previously reported system in that it relies on the deletion of both the galactose kinase (galK) and galactose-1-phosphate uridylyltransferase (galT) in the Leloir pathway (fig. S1), ultimately creating a system that allows for the selection of target gene deletions on solid medium containing galactose and not the more expensive 2-deoxy-d-galactose. The first step in creating the chloramphenicol- and thiamphenicol-selectable plasmid pDJSVT1 was to PCR-amplify the catP gene from a pJIR750 backbone (Clostridium shuttle vector) followed by using overlap extension PCR (OLE-PCR) to fuse a pMB1 E. coli origin of replication from pUC19 (fig. S2A) (see table S1 for primers). This linear product, which contains a multiple cloning site with GC-rich DNA restriction sites (XhoI, NotI, KpnI, and MluI) that are optimized for cloning in DNA from AT-rich bacteria such as Fusobacterium (~75% AT), was cut with NotI and ligated to circularize the plasmid. This final high-copy pDJSVT1 construct is small (~1800 bp), which likely enhances transformation efficiency.

pDJSVT1 is the base vector to create multiple additional vectors to complete the genetic system. To delete the galK and galT genes in F. nucleatum 23726 to make the base strain DJSVT02 (fig. S3A), 1000 bp directly upstream and downstream of the galKT gene cluster was amplified separately and then fused using OLE-PCR. This 2-kb PCR product was then digested with KpnI/MluI and ligated into pDJSVT1 digested with the same enzymes. This final vector is pDJSVT13. This vector was electroporated (1 to 3 μg of DNA, 2.5 kV, 50-μF capacitance, 360-ohm resistance) into competent F. nucleatum 23726 and selection on chloramphenicol (single chromosomal crossover), followed by selection on solid medium containing 1% 2-deoxy-d-galactose to select for either galKT gene deletions, as the absence of the galT gene makes 2-deoxy-d-galactose nontoxic to F. nucleatum (fig. S1).

The next plasmid is for targeted gene deletion in the F. nucleatum 23726 ∆galKT (Fnn: strain DJSVT02) background. This vector, pDJSVT7 (fig. S2B), contains a FLAG::galK gene under the control of a Fusobacterium necrophorum promoter. Transformation allows for initial chromosomal integration and selection with thiamphenicol, followed by selection for double crossover gene deletions on solid medium containing 3% galactose. For this study, this vector was used to create gene deletions in fap2, fadA, cbpF, fvcB, fvcC, and fvcD in F. nucleatum 23726 ∆galKT. As shown in fig. S4, 750 bp directly upstream and downstream of a target gene for intrabacteria homologous recombination was amplified by PCR, making complementary fragments fused by OLE-PCR. This product (fig. S4A) is ligated into pDJSVT7 using KpnI/MluI restriction sites. This vector is then electroporated (2.5 kV, 50-μF capacitance, 360-ohm resistance) into competent F. nucleatum 23726 ∆galKT and selected on chloramphenicol (single chromosomal crossover), followed by selection on solid medium containing 3% galactose, which produces either complete gene deletions or wild-type bacteria revertants. Gene deletions are verified by PCR and sequencing, and we show that this system has been accurate down to the single base level (fig. S5). For creating multiple gene deletions in a single strain (table S3), additional vectors were created and mutants were made in sequential fashion, as gene deletion leaves no trace of vectors, including excision of the chloramphenicol and galactose selection genes.

The third vector in the suite, pDJSVT11 (fig. S2C; not used in this study), is to create single-copy chromosomal complementations at a static chromosomal location within the arsB gene, which is only necessary during times of high arsenic exposure and therefore does not change the phenotype of Fnn under any conditions tested (Fig. 1E and fig. S3H).

RNA extraction and RT-PCR

F. nucleatum cultures were grown to stationary phase and pelleted by high-speed centrifugation (12,000g for 2 min at room temperature). TRIzol Extraction Isolation of total RNA was performed following the manufacturer’s instructions (Invitrogen). Briefly, cell pellets were resuspended in 1 ml of TRIzol reagent (Invitrogen), and 0.2 ml of chloroform was added. Solution was centrifuged for 15 min at 12,000g at 4°C. The RNA-containing aqueous phase was collected, and the RNA precipitated after 500 μl of isopropanol had been added. The RNA pellet was then washed with 75% ethanol and centrifuged at 10,000g for 5 min at 4°C. After drying at room temperature for 10 min, the RNA pellet was resuspended in 30 μl of sterile ribonuclease-free water and solubilized by incubating in water bath at 55°C for 10 min. Total RNA was quantified using the Qubit RNA HS Assay Kit (Thermo Fisher Scientific ).

Before RT-PCR, RNA samples were subjected to deoxyribonuclease (DNase) treatment. Briefly, 500 ng of total RNA was incubated with DNase I (Invitrogen) for 2 hours at 37°C. After treatment, DNase I was inactivated using EDTA and heating mixture for 5 min at 65°C. RT-PCR was performed using the TaKaRa PrimeScript One Step RT-PCR Kit according to the manufacturer’s instructions. The PCR conditions consisted of reverse transcription for 30 min at 50°C, initial denaturation for 2 min at 94°C, followed by 30 cycles (30 s at 94°C, 30 s at 50° to 62°C, and 30 s at 68°C) and elongation at 68°C for 1 min. The expected bands around 250 bp were confirmed on a 1.5% agarose gel. Specific primers to detect knockout of gene and validate intact genes upstream and downstream of gene of interest (table S1) were used to amplify from RNA extracts.

Development of a pan-Fusobacterium outer membrane antisera

To test whether broadly neutralizing polyclonal antisera against Fusobacterium outer membrane proteins could be effective at blocking cellular binding and entry, rabbits were injected (New England Peptide) with purified total membrane proteins from 11 strains of Fusobacterium that span seven species (F. nucleatum subsp. nucleatum 23726, F. nucleatum subsp. animalis 7_1, F. nucleatum subsp. polymorphum 10953, F. nucleatum subsp. vincentii 49256, F. periodonticum 2_1_31, F. varium 27725, F. ulcerans 49185, F. mortiferum 9817, F. gonidiaformans 25563, F. necrophorum subsp. necrophorum 25286, and F. necrophorum subsp. funduliforme 1_1_36S). Twenty milliliters of each strain was grown as previously described to stationary phase before pooling. Pooled Fusobacterium culture (220 ml) was centrifuged, and the pellet was resuspended in 40 ml of phosphate-buffered saline (PBS; pH 7.4) with two Roche EDTA-free protease inhibitor tablets, frozen at −80°C, and passed through an Emulsiflex C3 at >15,000 psi to lyse cells. Lysate was centrifuged at 18,000g for 25 min to pellet debris and unlysed cells. The supernatant was carefully transferred to ultracentrifuge tubes and centrifuged in a 50-Ti rotor at 38,000 rpm (~144 kG average) for 1 hour and 40 min. The supernatant was discarded, the membrane protein pellet was gently washed with PBS, and 0.5 ml of PBS with 1% n-octyl-b-d-glucopyranoside (BOG) was added to each of the two tubes and gently resuspended by stirring overnight at 4°C. The protein concentration was determined by bicinchoninic acid (BCA) assay before shipment to New England Peptide for inoculation into rabbits. The resulting membrane antiserum is herein named DJSVT_MAS1 but for clarity is described in figure legends as FMAS for Fusobacterium membrane antisera.

Immunofluorescence sample preparation using pan-Fusobacterium membrane antisera

Confluent HCT116 cells on slides were infected with FM 1-43FX–labeled (green fluorescent lipid) F. nucleatum 23726 in wells containing McCoy’s 5A medium with 10% FBS and no antibiotics at a multiplicity of infection (MOI) of 50:1 for 4 hours at 37°C with 5% CO2. After infection, medium was removed and thoroughly washed with PBS with gentle orbital shaking. Infected cells were then fixed with PBS/3.2% paraformaldehyde for 20 min at room temperature, followed by washing with PBS. Cells were blocked with Sea Block (Abcam) for 2 hours at 37°C, followed by the addition of 1:100 dilution of DJSVT_MAS1 overnight at 4°C. Slides were washed in PBS and incubated for 1 hour with Alexa Fluor 594 goat anti-rabbit antibody diluted in Sea Block. After washing with PBS, cells were permeabilized with PBS/1% Triton X-100 for 20 min. After permeabilization, cells were washed in PBS and labeled with 4′,6-diamidino-2-phenylindole (DAPI) for 30 min. After three final PBS washes, coverslips were mounted with 80% glycerol/0.1 M tris (pH 8.5) and sealed onto glass slides. Fluorescence microscopy was performed on a Zeiss LSM 800 confocal microscope.

Flow cytometry

Mid-log phase F. nucleatum were incubated with FM 1-43FX Lipophilic Styryl Dye (5 μg/ml ) to stain outer membranes and allow for green fluorescence detection. After washing in PBS, bacteria were resuspended in PBS at 100× concentration (MOI 50:1) before adding to cultures of HCT116 in medium containing 10% FBS. Infections lasted 4 hours before removing the culture medium (in many cases used for ELISAs as described), and cells were washed twice with PBS, followed by cell recovery using 0.05% trypsin. After neutralizing trypsin with medium containing 10% FBS, cells were pelleted and resuspended in PBS containing 20 mM EDTA. Cells were loaded onto a Guava easyCyte 5 flow cytometer (Luminex), and 10,000 cells were collected using an initial single-cell gate to measure the median green fluorescence induced by intracellular F. nucleatum labeled with FM 1-43FX. After data acquisition, FlowJo 10 software was used to further refine gating to single cells as well as determine median fluorescence of all samples. FlowJo analysis was then transferred to GraphPad Prism for statistical analysis and figure generation. We show intracellular F. nucleatum 23726 using imaging flow cytometry on an Amnis ImageStream X Mk II (Fig. 1D). For this experiment, the same protocol for bacterial and HCT116 growth, FM 1-43FX labeling, and infection times was used.

Invasion and survival antibiotic protection assays

HCT116 monolayers were washed once with PBS, and the cells were incubated in medium containing 10% FBS and no antibiotics. HCT116 cells were then infected at an MOI of 0.5:1 with exponential-phase F. nucleatum for 2 hours at 37°C with 5% CO2. After infecting cells for 2 hours, the cell culture medium was aspirated off and cells were washed two times with antibiotic-containing cell culture medium to remove any unbound bacteria. Cells were then incubated in medium containing penicillin and streptomycin (readily kills F. nucleatum 23726) for 1 hour to kill any remaining extracellular bacteria. At the end of the incubation period, cells were washed twice with PBS. Epithelial cells were then incubated with warm sterile water to lyse HCT116 cells. Lysates were plated on CBHK plates and incubated at 37°C under anaerobic conditions for 48 hours followed by colony counting.

Inhibition of F. nucleatum binding to HCT116 cells and subsequent signaling using chemical compounds and antibodies

F. nucleatum invasion and induced host-cell signaling inhibition by chemical compounds and neutralizing antibodies were analyzed using the standard infection conditions described above. Just before adding bacteria, with no preincubation time, 10 mM d-galactose, GalNAc, lactose, d-glucose, maltose, or l-arginine was added to the culture medium. For DJSVT_MAS1 Fusobacterium membrane antisera, serum was added from between 1:25 and 1:250 dilutions per infection. Chloroquine was added to HCT116 cells at 10 μM for 4 hours before infection and remained in the culture medium during infections. For all compounds, after a 4-hour infection at MOI 50:1, binding and invasion was analyzed using flow cytometry.

Isolation of mouse neutrophils and macrophages

Mouse bone marrow neutrophils were isolated from wild-type C57BL/6 mice over a 62.5% Percoll gradient with centrifugation at 1100g for 30 min. Purity was >90% as determined by flow cytometry analysis with Ly6G+CD11b+ staining. Neutrophils were used immediately after isolation. For mouse bone marrow–derived macrophages, bone marrow cells were cultured for 5 days in RPMI 1640 medium supplemented with 10% FBS, 2 mM l-glutamine, 1% penicillin/streptomycin, and macrophage colony-stimulating factor (M-CSF) (10 ng/ml). Fresh medium was replaced every other day. Wild-type C57BL/6 mice were bred and maintained in the animal facility at Virginia Tech in accordance with the Institutional Animal Care and Use Committee–approved protocol.

Human and mouse inflammatory cytokine arrays

Exponential-phase F. nucleatum 23726 or F. nucleatum 23726 ∆fap2 were used to infect HCT116 cell monolayers (100% confluency, 1.2 × 106 cells per well for six-well plate) in antibiotic-free cell culture medium for 4 hours at an MOI of 50 bacteria per human cell (MOI, 50:1). For neutrophils and macrophages, 1 × 106 cells in suspension were used. After 4 hours, all media (1.5 ml) were collected and sterile-filtered through 0.2-μm filters (MilliporeSigma) to remove bacterial and human cells from the sample. The cytokine array membranes [R&D Systems: Proteome Profiler Human Cytokine Array (ARY005B), Proteome Profiler Mouse Cytokine Array Kit, Panel A (ARY006)] were then blocked for 1 hour in tris-buffered saline–3% bovine serum albumin (BSA) at room temperature on a rocking platform. While the membrane was blocking, 1.5 ml of each sample of collected infection medium was incubated with 15 μl of the human cytokine array detection antibody cocktail at room temperature for 1 hour. After 1 hour, the blocking buffer was poured off and the sample/antibody mixture was incubated with the array membrane at 4°C overnight. After incubation with the sample/antibody mixture, the array membrane was washed three times with PBS. A streptavidin–horseradish peroxidase (HRP) conjugate was then added to the membrane and incubated for 30 min on a rocking platform. The array membrane was then washed three times with wash buffer to remove any unbound streptavidin-HRP. After the membrane was sufficiently washed, it was incubated with a chemiluminescent substrate solution and results were analyzed using the G:Box gel imaging platform.

Enzyme-linked immunosorbent assay

HCT116 cells were seeded to confluence in 24-well plates (2 × 105 cells per well at 100% confluence), and F. nucleatum was added to 500 μl of these wells at an MOI of 50:1. The plates were then incubated at 37°C, 5% CO2 for 4 hours. The medium was then collected from the wells, sterile-filtered using a 0.2-μm filter (MilliporeSigma), and diluted to concentrations within the range of the R&D Systems DuoKit ELISA to analyze human IL-8 and CXCL1 concentrations. For ELISAs detecting mouse CCL3 and CXCL2, 1 × 106 fresh neutrophils in suspension were used as described for the HCT116 protocol.

Transwell HCT116 migration assays

Transwell HCT116 migration assays were performed with Corning 8-μm transwell inserts in 24-well plates. HCT116 cells were first cultured in McCoy’s 5A (ATCC) to 90% confluence, followed by collection with trypsin. Cells were then stained using CellTracker Red (Thermo Fisher Scientific) and resuspended at a concentration of 2 × 106 cells/ml in medium supplemented with 1% FBS. One hundred microliters (2 × 105 cells) of the cell suspension was added to the top chamber of the 8-μm transwell insert precoated with 100 μl of Matrigel (Corning; 250 μg/ml). The lower chamber contained 600 μl of medium with 1% FBS in addition to adding the following: (i) chemokines [purified recombinant human IL-8 (Thermo Fisher Scientific) and CXCL1 (Sigma-Aldrich), individually or together at a concentration of 100 ng/ml] and (ii) conditioned and concentrated media obtained from 4-hour F. nucleatum infections of HCT116 cells. To prepare concentrated medium for each sample, three T-75 flasks with confluent HCT116 cells were used. Before infection, the complete medium in each flask was replaced with 10 ml of serum-free and penicillin/streptomycin-free medium. The bacteria resuspended in serum-free medium were added at 50:1 MOI, and the flasks were incubated in a hypoxic chamber (1% O2) for 4 hours. The medium containing the cell secretions was collected and pooled from three flasks, spun down at 3000g for 5 min, and passed through a 0.22-μm filter. The samples were then concentrated from 30 to 1.5 ml using a 3000 molecular weight cutoff Amplico concentrator (MilliporeSigma) at 4°C. The resulting sample was used for ELISA and transwell migration assays. The transwells were incubated at 37°C, 5% CO2 for 6 to 16 hours, after which the cells on the top were removed using a cotton-tipped applicator, and migrated cells at the bottom of the membrane were stained with DAPI and imaged on a Zeiss LSM 800 confocal microscope using a 10× objective. ImageJ was used to count the number of cells in five representative images per transwell in biological triplicate (70).

Depletion of IL-8 and CXCL1 from conditioned medium

Conditioned and concentrated media were obtained as described. Human IL-8/CXCL8 biotinylated antibody (R&D Systems, BAF208) and Human/Primate CXCL1/GROα/KC/CINC-1 biotinylated antibody (R&D Systems, BAF275) were added to the medium to a final concentration of 40 ng/ml and incubated at room temperature with gentle shaking for 30 min. Magnetic streptavidin particles (150 μl; Sigma-Aldrich, 11641778001) were first washed twice with PBS and spun down at 1500g and added to the solution. The mixture was incubated at room temperature with gentle shaking for another 30 min. The samples were then spun down at 1500g, and the supernatant was collected containing the conditioned medium with reduced cytokines. An ELISA quantified and confirmed the depletion. This medium was subsequently used in transwell migration assays as described.

Statistical analysis

All statistical analyses were performed in GraphPad Prism version 8.2.1. For single analysis, an unpaired Student’s t test was used. For grouped analyses, two-way analysis of variance (ANOVA) was used. In each case, the following P values correspond to star symbols in figures: nsP > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. To obtain statistics, all studies were performed as three independent biological experiments. For all experiments in which statistical analysis was applied, an N of three independent experiments was used (details in figure legends).


Fig. S1. Using the Leloir pathway for targeted gene deletions in bacteria.

Fig. S2. Development of vector for markerless gene deletion in F. nucleatum.

Fig. S3. Deletion of the galKT gene operon in F. nucleatum 23726.

Fig. S4. Target gene deletion in F. nucleatum 23726 ∆galKT (Fnn).

Fig. S5. Validation of markerless fap2 and fadA gene deletions in F. nucleatum 23726 ∆galKT (Fnn).

Fig. S6. Complementation of gene deletions at the static arsB gene site in F. nucleatum 23726.

Fig. S7. Human and mouse cytokine arrays to detect F. nucleatum–induced immune cell signaling.

Fig. S8. Chloroquine does not inhibit F. nucleatum cellular invasion and cytokine secretion.

Table S1. Primers used in this study.

Table S2. Plasmids used in this study.

Table S3. Bacterial strains used in this study.

References (7175)


Acknowledgments: We thank the following individuals for help and guidance with these studies: S. Melville (Virginia Tech) and C. Caswell (Virginia Tech) for critical insight into bacterial genetics, J. Lemkul (Virginia Tech) for critical manuscript insights, and M. Makris (Virginia-Maryland School of Veterinary Medicine) for imaging flow cytometry. Select figures were made with a paid subscription of We thank the Statistical Applications and Innovations Group (SAIG) at Virginia Tech for critical guidance during statistical analysis. Funding: This research was supported by the NIH through an NCI R21 Award (grant no. 1R21CA238630-01A1 to D.J.S. and S.S.V.), an NIAID R01 Award (grant no. 5R01AI136386-03 to L.L.), and an NSF Career Award (grant no. CBET-1652112 to S.S.V.); the Fralin Life Sciences Institute at Virginia Tech (to D.J.S.); the Institute for Critical Technology and Applied Science at Virginia Tech (to D.J.S., S.S.V., and L.L.); and the USDA National Institute of Food and Agriculture (to D.J.S.). Author contributions: M.A.C., C.C.Y., and B.U. curated and analyzed the data, designed/optimized the methodology, and reviewed and edited the manuscript. B.E.S., A.U., H.P., A.J.D., and Y.W. curated the data and reviewed and edited the manuscript. L.L. and S.S.V. helped conceptualize, supervise, and acquire funding for the study, performed data analysis, and reviewed and edited the manuscript. Y.Z curated the data and reviewed and edited the manuscript. D.J.S. contributed all the above. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Materials are available upon reasonable request with a material transfer agreement with Virginia Tech or through the Addgene plasmid repository. All data needed to evaluate conclusions are presented in the paper or Supplementary Materials, and raw data can be accessed on the Open Science Framework repository at

Stay Connected to Science Signaling

Navigate This Article