Central role of liver in anticancer and radioprotective activities of Toll-like receptor 5 agonist

Significance

Toll-like receptor 5 (TLR5) is an innate immunity receptor that specifically recognizes and triggers immune response to bacterial flagellins. In addition to resistance to Salmonella infection, TLR5 agonists protect mammals from radiation and have anticancer effects, including suppression of tumor metastases. Using mouse models, we defined the liver as a major target for TLR5 agonists. Administration of pharmacologically optimized flagellin derivative CBLB502 leads to rapid activation of prosurvival nuclear factor kappa B (NF-κB) and STAT3 pathways in the liver and rescues mice from lethal doses of hepatotoxic Fas-agonistic antibodies. Thus, TLR5 agonists can be considered for treatment and prevention of liver metastasis and hepatoprotective applications.

Abstract

Vertebrate Toll-like receptor 5 (TLR5) recognizes bacterial flagellin proteins and activates innate immune responses to motile bacteria. In addition, activation of TLR5 signaling can inhibit growth of TLR5-expressing tumors and protect normal tissues from radiation and ischemia-reperfusion injuries. To understand the mechanisms behind these phenomena at the organismal level, we assessed nuclear factor kappa B (NF-κB) activation (indicative of TLR5 signaling) in tissues and cells of mice treated with CBLB502, a pharmacologically optimized flagellin derivative. This identified the liver and gastrointestinal tract as primary CBLB502 target organs. In particular, liver hepatocytes were the main cell type directly and specifically responding to systemic administration of CBLB502 but not to that of the TLR4 agonist LPS. To assess CBLB502 impact on other pathways, we created multireporter mice with hepatocytes transduced in vivo with reporters for 46 inducible transcription factor families and found that along with NF-κB, CBLB502 strongly activated STAT3-, phenobarbital-responsive enhancer module (PREM), and activator protein 1 (AP-1–) -driven pathways. Livers of CBLB502-treated mice displayed induction of numerous immunomodulatory factors and massive recruitment of various types of immune cells. This led to inhibition of growth of liver metastases of multiple tumors regardless of their TLR5 status. The changed liver microenvironment was not, however, hepatotoxic, because CBLB502 induced resistance to Fas-mediated apoptosis in normal liver cells. Temporary occlusion of liver blood circulation prevented CBLB502 from protecting hematopoietic progenitors in lethally irradiated mice, indicating involvement of a factor secreted by responding liver cells. These results define the liver as the key mediator of TLR5-dependent effects in vivo and suggest clinical applications for TLR5 agonists as hepatoprotective and antimetastatic agents.

Toll-like receptors (TLRs) recognize and are activated by specific patterns in molecules that are produced by a broad range of microbial pathogens but are not present in host molecules. Activation of TLRs by these pathogen-associated molecular patterns leads to induction of infection-fighting innate immune responses (1). Various TLR agonists have been considered for multiple clinical applications, including cancer immunotherapy (2–4), and one, the TLR7 agonist imiquimod, is approved for topical treatment of basal cell carcinoma (5).

Although signaling pathways induced by different TLRs all result in mobilization of an innate immune response and involve activation of nuclear factor kappa B (NF-κB), the key regulator of immunity (6, 7), TLR5 is a particularly attractive candidate for therapeutic targeting for several reasons. First, bacterial flagellin, the natural ligand of TLR5, was found to have strong radioprotective effects in rodents and nonhuman primates (8). CBLB502 is a rationally designed derivative of Salmonella flagellin that lacks the highly immunogenic central globular domain and contains N- and C-terminal domains of the parental protein connected by a flexible linker. It is substantially less immunogenic than full-length flagellin but retains its TLR5-dependent NF-κB–inducing activity and radioprotective capability (8). CBLB502 (also called Entolimod) is currently under development as a medical radiation countermeasure capable of both reducing damage to radiosensitive hematopoietic (HP) and gastrointestinal (GI) tissues and improving their regeneration. Moreover, CBLB502 protected mice from dermatitis and mucositis associated with local fraction irradiation of head and neck area modeling radiation treatment of patients with head and neck cancer (9). Second, the TLR5 agonist CBLB502 was shown to be effective as a tissue protectant in mouse models of renal ischemia-reperfusion injury (10). Third, bacterial flagellin and flagellin-expressing bacteria (Salmonella) have shown antitumor effects in mouse models of colon and liver metastasis of pancreatic cancer (11, 12). Bacterial flagellin and its derivative, CBLB502, also demonstrated antitumor effects in several in vivo models (9, 13, 14). These effects required TLR5 expression by the tumor cells and were presumably mediated by innate immune cells recruited to the tumor following activation of TLR5 and subsequent production of immunomodulatory factors, such as cytokines. Fourth, TLR5 agonists are significantly less toxic than agonists of many other TLRs (e.g., the TLR4 agonist bacterial LPS). This is because “cytokine storm”-inducing cytokines, such as tumor necrosis factor (TNF) and interleukin one beta (IL1-β), which are major determinants of the toxicity associated with stimulation of other TLRs, are not included in the spectrum of cytokines induced following TLR5 activation (15–17).

To understand the reasons behind the specific physiological response to and to facilitate rational exploration of potential clinical applications of TLR5 agonists, we sought to define the mechanism of action of CBLB502 in the context of the whole organism (using the mouse as a model system). Our focus was, foremost, on identification of the tissues and cells that are direct primary responders to CBLB502 and, subsequently, on analysis of the molecular effects of CBLB502 in primary responders (e.g., induction of signaling pathways, production of bioactive factors). These goals were met by (i) analyzing tissue specificity of TLR5 response by using NF-κB reporter mice that enable detection of organ sites responding to CBLB502, (ii) an in vivo adaptation of the recently developed FACTORIAL assay (18) to monitor the response of multiple transcription factors to CBLB502 simultaneously, and (iii) a surgical procedure for temporary occlusion of blood circulation through the liver to test the biological effects of factors secreted by liver cells in response to CBLB502. This work defined the liver as the major primary target organ of CBLB502, leading to activation of several prosurvival and immunoregulatory signaling pathways, as well as dramatic changes in the spectrum of secreted factors and immune cell content. CBLB502 treatment strongly suppressed growth of tumor cells in the liver regardless of their TLR5 status. These results, together with the finding that blood circulation through the liver is indispensable for CBLB502-mediated radioprotection of the HP system, revealed that the biological effects of TLR5 agonists are mediated, at least in part, by factors secreted by responsive liver hepatocytes. In addition to providing mechanistic insights regarding the radioprotective and anticancer activities of TLR5 agonists, this study defined hepatoprotection as a specific area of their prospective clinical application.

Results

Liver Is the Major Organ That Responds to the TLR5 Agonist CBLB502.

Stimulation of TLR5 is known to activate the major immunoregulatory transcription factor NF-κB, making NF-κB activation an accurate reflection of TLR5 agonist activity (1, 19). Therefore, to identify tissues that are primary responders to CBLB502 in vivo, we used noninvasive imaging to detect NF-κB–dependent luciferase expression in live BALB/c-Tg(IκBα-luc)Xen reporter mice following injection of PBS vehicle or CBLB502. Intense luciferase expression, indicative of strong NF-κB activation, was observed specifically in the livers of CBLB502-treated mice at 2 h postinjection (Fig. 1A).

Fig. 1.

NF-κB activation and immune cell mobilization in response to CBLB502 and LPS. (A) Bioluminescent imaging of NF-κB–dependent luciferase expression in BALB/c-Tg(IκBα-luc)Xen mice 2 h after s.c. injection of PBS or CBLB502 (0.2 mg/kg). (B) Luciferase activity in protein extracts of liver, small intestine (int.; ileum), large intestine (colon), kidneys, lungs, spleen, and heart tissue obtained from BALB/c-Tg(IκBα-luc)Xen mice 2 h after s.c. injection of PBS, CBLB502 (0.2 mg/kg), or LPS (1 mg/kg) (n = 3 per group). Bars represent average ± SD. (C) Immunohistochemical detection of NF-κB p65 (green) nuclear translocation in liver sections obtained from NIH Swiss mice 20, 40, or 60 min after s.c. injection of PBS [untreated (u/t)], CBLB502 (1 μg per mouse), or LPS (10 μg per mouse). Sections were costained with anti-cytokeratin 8 (ker8; red; epithelial cell marker) and DAPI (blue). Arrowheads indicate Kupffer and endothelial cells determined by morphological criteria. (Magnification: 40×.) (D) Immunohistochemistry with anti-p65 (green) and anti-cytokeratin 8 (red) in cultures of primary hepatocytes from NIH Swiss mice or humans treated with PBS (u/t), CBLB502 (100 ng/mL), or LPS (1 μg/mL) for 1 h. (Magnification: 40×.) (E) FACS analysis of NK-cell, neutrophil, and macrophage (Kupffer cell) populations in the liver and BM and T cells in the liver of mice treated with PBS or CBLB502 (1 μg, s.c.) for the indicated amounts of time. The absolute number of cells per organ (whole liver and BM from two hind tibias and two hind femurs) is indicated. Error bars represent mean ± SEM (n = 6 mice per group). An asterisk indicates that the difference from intact mice is significant (P < 0.05).

To quantitate NF-κB activity in different tissues, ex vivo luciferase activity assays were performed using tissue lysates from BALB/c-Tg(IκBα-luc)Xen reporter mice treated with PBS or CBLB502 for different amounts of time. For comparison, the TLR4 agonist LPS, which also activates NF-κB, was used at about 50% of its maximum tolerated dose. Again, the liver showed the strongest NF-κB response to CBLB502. Substantial reporter induction was also observed in the large intestines of CBLB502-treated mice (Fig. 1B). In contrast, LPS activated NF-κB specifically in the lungs, spleen, and kidney. In all responsive tissues, the kinetics of reporter activation were similar for both TLR agonists, peaking ∼2 h postinjection, declining by 6 h, and returning to near baseline by 24 h (data for representative organs are shown in Fig. S1).

To define specific CBLB502-responsive cell types within the liver, we used immunohistochemistry to detect nuclear translocation of the p65 subunit of NF-κB as an indicator of NF-κB activation. At 20 min after CBLB502 injection (the earliest tested time point), NF-κB translocation was observed in essentially all hepatocytes but not in other liver cell types, including Kupffer and endothelial cells (Fig. 1C). The response was opposite in LPS-treated mice: Early translocation of NF-κB was observed in Kupffer and endothelial cells but not in hepatocytes. Later (by 1 h postinjection), all liver cells in both TLR5 and TLR4 agonist-treated mice displayed nuclear NF-κB accumulation, presumably reflecting combined primary effects of TLR agonists on cells expressing their cognate receptors and secondary effects mediated by factors released by primary responding cells. NF-κB translocation was observed in pure cultures of primary hepatocytes from mice and humans (Fig. 1D) following treatment with CBLB502 but not LPS, regardless of the duration of treatment. These results indicate that liver hepatocytes of rodents and primates express functional TLR5 but not TLR4 and that hepatocytes are major specific primary targets of TLR5 agonists.

Activation of Immunostimulatory and Tissue Protective Pathways in Livers of CBLB502-Treated Mice.

To identify perturbations in signaling pathways in the mouse liver in response to TLR5 activation, we used a multiplexed reporter system (the FACTORIAL) enabling a simultaneous assessment of multiple transcription factor families within cells (18). As shown schematically in Fig. 2A, FACTORIAL multireporter mice were generated via in vivo transfection of mouse hepatocytes [using hydrodynamic stress (20)] with a “library” of 46 constructs, each containing a reporter transcription unit (RTU) controlled by a minimal promoter preceded by binding sites specific for a particular transcription factor. FACTORIAL multireporter mice were treated with CBLB502, LPS, or PBS for 1 h, and the 46 RTU-encoded transcripts were quantified in total liver RNA. This revealed dramatic activation of NF-κB (1,000-fold), STAT3 (200-fold), phenobarbital-responsive enhancer module (PBREM) (200-fold), and activator protein 1 (AP-1) (50-fold) signaling in livers from CBLB502-treated mice (Fig. 2B). In contrast, these pathways were either nonresponsive to LPS (for AP-1) or were at least 10-fold less responsive to LPS than to CBLB502. Rapid activation of NF-κB, STAT3, and AP-1 in livers of CBLB502-treated mice was confirmed by global gene expression profiling (using Illumina microarray hybridization) at 30 min and 2 h after CBLB502 injection (several examples are shown in Table S1). Consistent with the nature of activated transcription factors, this analysis showed strong changes (predominantly increases) in the abundance of mRNAs encoding multiple classes of bioactive factors, including NF-κB–responsive immunomodulators (cytokines, chemokines, and their receptors) and antiapoptotic and antimicrobial factors. mRNA induction for representative genes was confirmed using RT-PCR (Fig. 2C). Livers of CBLB502-treated mice also showed induction of IL-6, an NF-κB–regulated cytokine known to induce STAT3 signaling (21).

Fig. 2.

Activation of gene regulatory pathways in mouse liver in vivo by CBLB502 and LPS. (A) Schematic illustration of the in vivo FACTORIAL assay used to profile changes in the activities of transcription factors in the mouse liver following in vivo CBLB502 or LPS treatment. Two weeks after library transfection, mice were injected with CBLB502 (5 μg per mouse), LPS (10 μg per mouse), or PBS, and RTU activities in total liver RNA samples were isolated 1 h later and analyzed for the RTU activity as previously described (18). Description of transcription factors listed in the table can be found in www.attagene.com/cis-1-list.pdf. (B) The radial graph shows fold-induction values of mean activities of individual RTUs in CBLB502 or LPS treated mice vs. vehicle-treated mice (n = 8 mice per group). A fold induction value of 1× indicates that the pathway was not affected by the applied treatment. (C) mRNA levels for several apoptosis-related factors and cytokines implicated in TLR5 signaling (and GAPDH as a control) were analyzed by RT-PCR using total RNA from livers of untreated mice (“time 0”) or mice treated with CBLB502 (1 μg per mouse) for 30 min or 120 min; bcl2a1b, B cell leukemia/lymphoma 2 related protein A1b; bcl2a1d, B cell leukemia/lymphoma 2 related protein A1d; jun, jun proto-oncogene; cxcl1, chemokine (C-X-C motif) ligand 1; il6, interleukin 6; gapdh, glyceraldehyde-3-phosphate dehydrogenase.

These observations support our conclusion defining hepatocytes as primary responders to TLR5 agonists but not TLR4 agonists and provide a rational basis for observed biological activities of TLR5 agonists (e.g., tissue protection).

TLR5 Agonist Treatment Leads to Mobilization of Immune Cells to the Liver.

Because NF-κB is a major regulator of immune responses and induces numerous immunostimulatory factors, we predicted that NF-κB activation in the livers of CBLB502-treated mice would lead to recruitment of immune cells to the liver. This was verified by FACS analysis of total liver cells and bone marrow (BM), a known depot of immune cells, stained with specific Ab combinations (Fig. 1E). Five hours after CBLB502 treatment (the earliest tested time point), there was a 30% increase in the overall cellularity of the liver (Fig. S2A). This was due, at least in part, to rapid recruitment (by 5 h posttreatment or earlier) of natural killer (NK) cells and neutrophils to the liver (Fig. 1E). CBLB502-induced neutrophil accumulation was transient and completely resolved by 24 h. NK cells, however, remained elevated for at least 5 d (the duration of the experiment). Similar kinetics were observed for natural killer-T cells (Fig. S2B). The number of macrophages in the liver was only slightly increased by CBLB502 treatment. In addition to effectors of innate immunity, we observed recruitment of adaptive immune cells; total conventional αβ⁺ T cells, including CD4⁺/CD8⁺ T cells, were recruited to the liver within 5 h of CBLB502 injection and returned to baseline by day 5 (Fig. 1E and Fig. S2B). In the BM, NK-cell, neutrophil, and macrophage populations all displayed patterns of change following CBLB502 treatment that were opposite of those observed in the liver, thus suggesting that the BM is a likely source of at least some of the immune cells recruited to the liver.

CBLB502-Mediated Radioprotection of Hematopoietic Progenitor Cells Is Liver-Dependent.

The observed CBLB502 responsiveness of the liver suggested that it might play a role in the radioprotective activity of the compound by secreting CBLB502-induced bioactive factors. This possibility was supported by the failure of CBLB502 to protect HP cells against radiation-induced death in vitro (Fig. 3 A and B), despite its clear radioprotective effects on the HP system in vivo (8) (Fig. 3B). For example, treatment of primary mouse bone marrow cells (BMCs) with CBLB502 in vitro did not preserve granulocyte–macrophage colony formation potential (Fig. 3A). Moreover, although BM irradiated in the context of the whole animal showed CBLB502-dependent preservation of granulocyte–macrophage colony formation, BMCs that were isolated from CBLB502-injected mice and then irradiated in vitro did not (Fig. 3B).

Fig. 3.

CBLB502-mediated protection of BM granulocyte/macrophage (GM) progenitors from radiation damage. BM GM colony-forming units (GM-CFUs) were quantified as described in Materials and Methods. Bars in graphs represent the average number of colonies in triplicate cultures ± SD. (A) GM-CFUs in BMCs isolated and treated in vitro with CBLB502 (100 ng/mL) or PBS vehicle, followed by 10-Gy irradiation 30 min later. (B) GM-CFUs in BMCs isolated immediately after 10-Gy total body irradiation (TBI) from the mice that were injected with either CBLB502 (1 μg per mouse) or PBS 30 min before irradiation. As indicated by an asterisk, BMCs in one regimen were isolated from mice 30 min after CBLB502 injections and then irradiated in vitro with 10 Gy. (C) GM-CFUs in BMCs isolated from mice treated with the indicated combinations of CBLB502 (1 μg per mouse) or PBS and 10-Gy TBI under conditions of LER (occlusion) or sham-treatment (surgery without occlusion).

To test whether the liver response to CBLB502 is indeed important for its radioprotective activity, we assessed CBLB502-mediated protection of HP progenitor cells in irradiated mice in which blood circulation through the liver was blocked during the 30-min period of CBLB502 treatment and then restored. The effectiveness of this liver exclusion/reperfusion (LER) procedure was confirmed by quantifying NF-κB–dependent luciferase levels in liver tissue extracts from BALB/c-Tg(IκBα-luc)Xen reporter mice 2 h after application of LER or sham-LER clamps and injection of CBLB502 or PBS (Fig. S3). NIH Swiss mice were then used to assess CBLB502-mediated HP radioprotection under conditions of LER and sham-LER. BMCs were obtained from CBLB502- or PBS-injected mice immediately after irradiation and analyzed in granulocyte–macrophage colony formation assays. Under conditions of sham-LER, CBLB502 treatment resulted in significantly greater postirradiation colony formation than PBS treatment (Fig. 3C). However, under conditions of liver exclusion, there was no difference between CBLB502 and PBS treatment; drastic radiation-induced depletion of HP progenitors was observed in both cases (Fig. 3C). Injection of CBLB502 under liver exclusion did not alter the colony-forming potential of BMCs in nonirradiated mice, indicating that neither the drug nor the applied procedure had an adverse impact on the BM. The results obtained using LER strongly support our hypothesis that CBLB502-stimulated liver cells produce blood-borne factor(s) essential for protection of HP progenitor cells from radiation damage.

TLR5 Agonist Treatment Suppresses Liver Metastasis.

The liver is a frequent site for metastatic growth of multiple tumor types, and it is the primary site of colon cancer metastasis (22). Our observation of CBLB502-induced recruitment of large numbers of immune cells to the liver suggested that the changed liver microenvironment might affect the growth of liver metastases. We tested this in several syngeneic mouse models of liver metastasis, including colon carcinoma CT26, lymphoma A20, and breast carcinoma 4T1.

CT26 and A20 tumor cells were introduced into mice either by s.c. injection to mimic primary tumor formation or by intrasplenic injection (followed by rapid spleen removal), which leads to formation of multiple liver nodules mimicking liver metastases of the tumor. Formation of such experimental liver metastases was confirmed by bioluminescent imaging of live mice injected with luciferase-expressing CT26 cells and treated with PBS (Fig. 4A). Neither CT26 nor A20 tumor cells express TLR5 mRNA or respond to CBLB502 with NF-κB activation (Fig. S3). Consistent with this result and with previous results showing that antitumor effects of CBLB502 required expression of functional TLR5 by the tumor, we found that CBLB502 treatment did not alter growth of s.c. CT26-derived metastases (Fig. S4A). However, growth of CT26 metastases in the liver was suppressed by CBLB502 treatment on days 5 and 6 (Fig. 4 A and B, Top) or on days 1, 3, and 5 (Fig. S4B). Both regimens of CBLB502 treatment significantly delayed tumor growth in livers and increased the proportion of animals that remained tumor-free for at least 50 d. CBLB502 was even more effective in suppressing hepatic growth of A20 tumors. Mice treated with 1 μg of CBLB502 on days 5 and 6 or on days 5–9 or with 5 μg of CBLB502 on days 5–9 after intrasplenic A20 cell inoculation were >95% tumor-free on day 60, whereas only 40% of mice treated with vehicle were tumor-free (Fig. 4B, Bottom). These results illustrate a liver-specific antitumor effect of TLR5 agonists that does not require TLR5 expression by the tumor cells.

Fig. 4.

Effect of CBLB502 treatment in experimental and spontaneous metastatic tumor models. (A) Bioluminescent imaging of CT26 tumors in livers of BALB/c mice injected (s.c.) with PBS (untreated) or CBLB502 (0.04 mg/kg) on days 5 and 6 after intrasplenic tumor cell inoculation. Images shown are from day 14. (B) Proportion of mice without tumor growth in the liver was determined by bioluminescent imaging. (Top) Mice were injected as in A. (Middle) Following intrasplenic CT26 inoculation, mice were left untreated, injected with CBLB502, or treated with anti-asialo GM1 Ab in combination with CBLB502 given on days 1, 3, and 5 (P value indicates the difference from intact and Ab-injected mice). (Bottom) A20 lymphoma cells were delivered to BALB/c mice by intrasplenic injection. The combined results of three regimens of CBLB502 treatment [0.05 mg/kg of CBLB502 on days 5 and 6 (n = 8) or on days 5–9 (n = 8) or 0.2 mg/kg of CBLB502 on days 5–9 (n = 9)] are shown. (C) H&E-stained liver sections obtained from mice with established CT26 liver metastases were treated with vehicle (A and B) or CBLB502 (C and D). The areas boxed in A and C are shown in B and D. (Magnification: A and C, 10×; B and D, 40×.) Arrows indicate CT26 metastases (white), mitotic tumor cells (red), neutrophils and lymphocytes at the boundary between metastases and liver parenchyma (green), and extravasation of inflammatory cells from hepatic vessels into liver tissue (blue). (D) Quantification of spontaneous liver metastasis of 4T1 tumors in mice given two rounds of PBS or CBLB502 treatment. Clonogenic assays were performed by plating total liver cells in 6-thioguanine–containing medium. The mean number of 4T1 colonies per mouse as determined from triplicate wells is shown. P values were calculated for comparison with PBS-only treatment (paired t test). (E) Mouse survival (%) was monitored during 75 d postsurgery. The difference between CBLB502-treated groups (n = 5 each) and PBS-injected (n = 4) mice was significant (log-rank test, P < 0.05).

CBLB502 also suppressed spontaneous liver metastasis of mouse breast adenocarcinoma 4T1 cells, which do express functional TLR5 (Fig. S3). In this model, metastasis was evaluated in a clinically relevant setup following surgical removal of primary tumors. The 4T1 cells were inoculated into mammary fat pads of syngeneic BALB/c mice, and the mice were given three daily treatments of PBS or CBLB502 when primary tumors were about 2 mm in diameter. Primary tumors were resected when they reached 400 mm³ in volume, and the mice were treated again with PBS or CBLB502 on days 1, 3, and 5 postsurgery. Twenty-one days after surgery, clonogenic assays were performed to quantify liver metastases as previously described (23). As shown in Fig. 4D, mice treated with CBLB502 after primary tumor resection (with either PBS or CBLB502 treatment before resection) had significantly fewer colony-forming 4T1 cells in their livers compared with mice treated with PBS both before and after resection (P < 0.05). Two out of four mice treated with CBLB502 before resection and with PBS after resection also had significantly fewer 4T1 cells in their livers. Consistently, there was a significant increase in the proportion of animals that survived following surgical removal of primary tumors in CBLB502-treated groups (Fig. 4E), presumably reflecting eradication of tumor cells infiltrating distal organs of tumor-bearing mice.

Metastasis Suppression by CBLB502 Is Mediated by Mobilization of NK Cells to the Liver.

To determine whether immune cell mobilization plays a role in the observed suppression of liver metastasis in CBLB502-treated mice, we evaluated the morphology of mouse livers containing CT26 metastases. Luciferase-expressing CT26 cells were delivered to mice by intrasplenic injection, and 20 d later, when tumor growth in livers was detected by luminescent imaging, mice were given two daily s.c injections of PBS or CBLB502 (1 μg per mouse). At 5 h after the second injection, H&E-stained liver sections showed multiple metastases in livers of vehicle-treated animals (Fig. 4C, white arrows) characterized by frequent mitoses within tumor cells (Fig. 4C, red arrows) and no visible inflammatory infiltrates. CBLB502-treated mice had liver metastases of similar size and distribution (Fig. 4C, white arrows) as vehicle-treated animals but also showed marked accumulation of immune cells in the liver. Neutrophils and lymphocytes were seen at the rim of growing metastases bordering the liver parenchyma (Fig. 4C, green arrows). In some areas, recruitment of inflammatory cells from hepatic vessels into the liver was evident (Fig. 4C, blue arrows).

The importance of NK cells in particular for CBLB502-mediated suppression of liver metastases was assessed in mice depleted of NK cells using antiasialo GM1 Abs (24), which specifically depleted NK cells without affecting neutrophils or macrophages (Fig. S5). Mice were treated with anti-asialo GM1 Ab 1 d before and immediately before CBLB502 injection on days 1 and 5; CBLB502 injections (1 μg per mouse) were given on days 1, 3, and 5 after intrasplenic injection of CT26 cells. Imaging (IVIS Imaging System, 100 series; Xenogen Corp.) was used to monitor development of CT26 metastases in the liver and showed that the previously observed antitumor effect of CBLB502 (Fig. 4B, Top) was completely abrogated by depletion of NK cells (Fig. 4B, Middle). This result is consistent with previous work demonstrating the importance of innate immune cells for TLR5-mediated antitumor effects (14).

CBLB502 Treatment Protects Mice from Fas-Mediated Hepatotoxicity.

Cytotoxicity of activated immune cells, including those mobilized to the liver by CBLB502 (e.g., NK cells, neutrophils), is often mediated by their expression of Fas ligand (25, 26). This, together with the high sensitivity of liver cells to Fas-mediated apoptosis (27, 28), raises the possibility that CBLB502-induced inflammation in the liver might be hepatotoxic. However, on the other hand, TLR5 agonists are strong activators of NF-κB, which has powerful antiapoptotic effects, including protection against Fas-mediated apoptosis (29). To test the effect of CBLB502 on liver sensitivity to Fas-mediated apoptosis, we treated mice with anti-Fas agonistic Ab (anti-Fas Ab) at doses capable of inducing lethal hepatotoxicity associated with signs of apoptosis, liver tissue necrosis, and hemorrhage (30, 31). Injection of NIH Swiss mice with anti-Fas Ab resulted in 100% mortality within 1–2 d (Fig. 5A). In contrast, 100% of mice survived when they were given CBLB502 30 min or 2 h before anti-Fas Ab. CBLB502 had a lesser but still beneficial effect on survival when injected 10 min or 6 h before anti-Fas Ab. The protective effect of CBLB502 on normal liver tissue in this model was confirmed using H&E staining, TUNEL staining, and erythrocyte autofluorescence to detect necrosis, apoptotic cells, and hemorrhagic lesions, respectively (Fig. 5C). In addition, Fas-induced elevation of serum alanine aminotransferase levels [an indicator of liver damage (32)] was significantly reduced by CBLB502 pretreatment (Fig. 5B).

Fig. 5.

Protection against Fas-mediated hepatotoxicity by CBLB502. (A) Survival of NIH Swiss mice on day 20 after i.p. injection of 4 μg of anti-Fas Ab alone (PBS, n = 5) or in combination with CBLB502 (1 μg per mouse) injected s.c. 10 min (n = 5), 30 min (n = 16), 2 h (n = 10), or 6 h (n = 10) before anti-Fas injection. (B) Mice were treated with CBLB502 given 30 min before anti-Fas Ab. Serum alanine aminotransferase (ALT) levels were determined at 6 h after anti-Fas injection (Top), caspase-8 (Middle) and caspase-3/7 (Bottom) activities in liver tissue protein lysates were determined at 5 h. Bars represent average ± SD (n = 3–4 mice per group). For all three assays, the difference between mice given anti-Fas alone vs. anti-Fas + CBLB502 was statistically significant (P < 0.05). (C) Evaluation of histological sections of liver tissue prepared 5 h after injection of NIH Swiss mice with vehicle or with anti-Fas Ab alone or in combination with CBLB502. Tissue morphology was assessed by H&E staining; apoptotic cells were detected by TUNEL staining; and hemorrhage was visualized by erythrocyte autofluorescence (red), mouse IgG control staining (Cy5-conjugated anti-mouse IgG Ab, purple), and DAPI staining of nuclei (blue). (Magnification: 20×.) (D) Western blot analysis of full-length and cleaved Bid in liver tissue protein lysates from NIH Swiss mice left untreated (“vehicle”) or injected with anti-Fas Ab, CBLB502, or both. Lysates were prepared 2 h after anti-Fas injection.

Fas-dependent apoptosis involves activation of caspase-8, which leads to activation of effector caspase-3 and caspase-7 either directly (in type I cells) or indirectly (in type II cells) via a chain of events involving BH3 interacting domain death agonist (Bid) activation, mitochondrial cytochrome c release, and caspase-9 activation (33). CBLB502 pretreatment only slightly reduced caspase-8 activation in livers of anti-Fas Ab–treated mice (Fig. 5B) and only partially suppressed cleavage of the endogenous caspase-8 target, Bid (Fig. 5D). In contrast, activation of caspase-3 and caspase-7 was nearly eliminated (Fig. 5B; also observed in BALB/c and C57BL/6 mice, Fig. S6). These results suggest that the liver cells affected by anti-Fas Ab are type II cells and that prevention of apoptosis by CBLB502 occurs predominantly downstream of caspase-8 and Bid, presumably at the level of cytochrome c release, as predictable by the observed induction of the B-cell lymphoma 2 family proteins (Fig. 2C) and their upstream regulator, STAT3 (Fig. 2B). These proteins are known to block production of reactive oxygen species (ROS) and suppress the mitochondrial apoptotic pathway (34). In addition, activation of STAT3 signaling in hepatocytes was previously shown to have an antiapoptotic effect (35).

Fas-mediated apoptosis has also been implicated in the pathogenicity of Salmonella infection in mice, which is predominantly localized in the liver (36–38). Consistent with this, treatment with CBLB502 (but not with LPS) improved survival of mice infected with a lethal dose of Salmonella typhimurium (Fig. S7). This protective effect was likely due to a combination of suppression of Fas-mediated apoptosis in the liver by CBLB502 (as described above) and its ability to induce production of antimicrobial factors (detected by microarray-based gene expression analysis; some examples are shown in Table S2; to be reported in detail in a separate paper).

Discussion

Stimulation of TLR5 by its natural ligand, flagellin, or by the flagellin derivative CBLB502 causes NF-κB activation on the cellular level and multiple biological effects, including tissue protective and antitumor effects with strong clinical potential, on the organismal level. Fundamental to understanding the mechanism(s) of action of TLR5 agonists is identification of tissues that are primary responders to these agents. This report demonstrates that the strongest NF-κB activation in response to TLR5 agonist CBLB502 occurs in the liver and GI tract. Although these observations possibly reflect the tissue specificity of TLR5 expression, this assumption cannot be directly tested due to lack of a reliable immunohistochemical assay for TLR5. Hepatocytes were shown to be responsible for the primary CBLB502 response in the liver, and cells of the lamina propria [presumably dendritic cells as reported by Uematsu et al. (39)] were identified as likely primary responders in intestinal tissue. Expression and function of TLR5 in the liver make biological sense, given that the liver is the primary site of Salmonella residence during infections (36). Our finding that TLR5 and TLR4 are expressed in two different nonoverlapping liver compartments, hepatocytes and Kupffer cells, respectively, further differentiates TLR5 from TLR4, the latter of which is predominantly expressed in cells of the immune system (1). Operation of these two receptors in different tissue microenvironments/epigenetic backgrounds provides a plausible explanation, along with differences in TLR5- and TLR4-activated signal transduction cascades, for the distinct biological outcomes of TLR5 agonist (flagellin) and TLR4 agonist (LPS) exposure. This includes induction of different cytokines (1, 4), which translates into cardinal differences in toxicity. Although LPS strongly induces the highly toxic cytokines TNF and IL-1β, both of which are capable of triggering a harmful cytokine storm (40–42), these cytokines are not induced by flagellin or CBLB502, which predominantly up-regulate granulocyte colony-stimulating factor (G-CSF), IL-6, IL-8, and IL-10 (8, 16, 17).

FACTORIAL assay (18) combined with global gene expression profiling identified multiple signaling pathways and downstream genes activated in hepatocytes of CBLB502-treated mice that are consistent with observed biological effects of TLR5 stimulation. In particular, NF-κB–regulated immunomodulators, such as cytokines, are likely responsible for rapid and massive infiltration of the liver with different types of immune cells following CBLB502 treatment. In this way, CBLB502 has a negative effect on hepatic growth of tumor cells, regardless of their TLR5 expression status (Fig. 6). In contrast, in TLR5-negative environments outside of the liver, the tumor itself must express TLR5 to mobilize an antitumor immune response following TLR5 agonist treatment (9, 14). The hypothetical scheme of CBLB502 tumor suppressive activity in the liver is shown in Fig. 6.

Fig. 6.

Schematic description of a plausible mechanism of the liver metastasis suppression effect of CBLB502.

The fact that the liver is a principal site for tumor metastasis (22) and for residence of immature NK cells (43) adds to the significance of our finding that NK cells are critical for the activity of CBLB502 against CT26 liver metastasis. NK cells are known to participate in surveillance and elimination of tumor cells through their cytolytic activity and production of cytokines that affect other components of the innate and adaptive immune systems (44). Cytolytic activity of NK cells against tumors cells is typically not antigen-dependent but may involve alternative receptor-mediated recognition of “stress-induced” ligands commonly expressed on tumor cells. The importance of neutrophils for the antitumor effects of TLR5 agonists noticed by others (13, 14), together with our finding of neutrophils' mobilization to the liver, suggests that cross-talk between neutrophils and NK cells may be involved. This notion is supported by previous work showing the involvement of neutrophils in NK-cell function (45).

In addition to regulating immunomodulatory factors, NF-κB controls expression of numerous other factors, including antiapoptotic proteins, scavengers of ROS, and growth factors (46). Such factors may play roles in CBLB502’s protection of (i) the HP and GI systems against radiation (8), (ii) the kidney against ischemia-reperfusion injury (10), and (iii) the liver against Fas-mediated apoptosis (this report). In these cases, additional “prosurvival” TLR5 agonist responses may include those controlled by the STAT3 (47, 48) and AP-1 pathways (49), which were previously shown to be responsible for liver resistance to Fas-mediated apoptosis (35), as well as by the PBREM, which regulates genes involved in the detoxifying metabolic capacity of the liver (50). The positive effect of TLR5 stimulation on liver resistance to severe stresses is illustrated by rescue of mice pretreated with CBLB502 from otherwise lethal doses of Fas agonistic Ab. This result explains the lack of liver damage in CBLB502-treated animals regardless of massive mobilization to this organ of inflammatory cells, the cytotoxic effect of which is largely mediated by their expression of Fas ligand (25, 26).

Our studies testing the radioprotective effect of CBLB502 under conditions of temporary exclusion of the liver from blood circulation indicate that the liver is a critical source of soluble factors acting directly or indirectly (i.e., via modulation of cellular trafficking) to protect the HP system against radiation damage (Fig. 3). The nature of these factors remains to be determined. However, likely candidates include the cytokines G-CSF and IL-6, which have been found elevated in the blood of CBLB502-treated mice and monkeys (8) and are known stimulators of HP stem cells (G-CSF) and the thrombopoietic lineage of hematopoiesis (IL-6) (51–53).

In addition to providing results consistent with other data and predictable hypotheses, our in vivo FACTORIAL assay pointed to previously unknown properties of CBLB502. For example, although not activated to the same level as the top four pathways (NF-κB, AP-1, STAT-3, and PBREM), the antioxidant response element [nuclear factor (erythroid-derived 2)-like 2 (NRF2)] RTU showed a strong (20-fold) and rapid response to CBLB502 but not LPS (Fig. 2B). Based on the known downstream targets of this pathway, this may indicate early induction of antioxidant enzymes (e.g., superoxide dismutase, GST) in hepatocytes of CBLB502-treated animals (54). Secreted antioxidants produced in this way could act as systemic ROS scavengers, and thus contribute to both the HP radioprotective activity and liver tissue protective activity of CBLB502. This possibility, as well as the significance of similar TLR5 agonist-specific induction of octamer-binding transcription factor (Oct) pathway (Fig. 2B), will be explored in future studies.

In summary, the key finding of this study, that hepatocytes are a major and specific primary target of TLR5 agonists, provides insight into the mechanism(s) of action of these agents, which have strong potential as therapeutics for radiation protection and reduction of cancer treatment side effects (8, 9); prevention of ischemia-reperfusion injury (10); immunotherapy of TLR5-positive tumors (9, 13, 14); and, as revealed by this report, suppression of liver metastases of different types of cancer regardless of their TLR5 status and protection against hepatotoxic insults.

Materials and Methods

Mice.

NIH Swiss female mice (National Cancer Institute) and BALB/c and C57BL/6 female mice (Jackson Laboratory) aged 10–14 wk were used. Balb/C-Tg(IκBα-luc)Xen reporter mice were originally purchased from Xenogen Corp. and bred in our domestic colony. All animal experiments followed protocols approved by the Roswell Park Cancer Institute Institutional Animal Care and Use Committee.

Reagents.

TLR5 agonistic agent CBLB502 was obtained from Cleveland BioLabs, Inc. (8, 55), and LPS from Escherichia coli 055:B5 was purchased from Sigma. Purified hamster anti-mouse Fas Ab, clone Jo2, was purchased from BD Biosciences. Purified anti-asialo GM1 Ab was purchased from Wako Chemicals.

Tumor Cells.

Murine mammary carcinoma 4T1, B-cell lymphoma A20, and colon undifferentiated carcinoma CT26 cells (American Type Culture Collection) were transduced with constitutive luciferase expression constructs and were cultured in RPMI plus 10% (vol/vol) FBS with standard supplements and antibiotics.

Primary Hepatocytes.

Primary human hepatocytes were purchased from BD Biosciences. Primary mouse hepatocytes were isolated from the liver of NIH-Swiss or BALB/c-Tg(IκBα-luc)Xen mice by collagenase digestion as described previously (56). Hepatocytes were cultured in William’s modified E medium supplemented with 10% FBS, 1% penicillin/streptomycin, 2 mM glutamine, 50 ng/mL EGF, 10 mM nicotinamide, 10⁻⁷ M dexamethasone, and 1× insulin-transferrin-selenite (57).

Imaging of Luciferase Expression in Live Mice.

BALB/c-Tg(IκBα-luc)Xen mice treated with CBLB502 or PBS were injected i.p. with 15 mg/mL d-luciferin (Promega) in PBS (0.1-mL volume), anesthetized using 1–2% isoflurane, and imaged (within 5 min of d-luciferin injection) using an IVIS Imaging System and Living Image software (Caliper LifeSciences). The same method was used to detect luciferase-expressing CT26 and A20 tumor cells in BALB/c mice.

Quantification of NF-κB–Driven Luciferase Expression in Protein Extracts.

Tissue samples of liver, lungs, kidney, spleen, heart, and intestine from BALB/c-Tg(IκBα-luc)Xen mice treated with PBS, CBLB502, or LPS were homogenized in protein lysis buffer (Promega) containing proteinase inhibitor mixture (Calbiochem). Luciferase activity was measured in 20 μL of lysate after adding 30 μL of luciferin reagent (Bright-Glo Luciferase Assay System; Promega) and normalized to the extract protein concentration measured using a BioRad DC protein assay kit II. Luciferase fold induction was calculated as the ratio between the average normalized luciferase activity in organs of TLR ligand-treated mice and PBS-injected mice. To assess NF-κB activation in tumor cell lines, cells were transiently transfected with the p5XIP10 κB reporter construct containing five tandem copies of the NF-κB binding site from the interferon-inducible protein 10 promoter upstream of the luciferase gene using Lipofectamine Plus (Invitrogen Life Technologies). Luciferase activity was measured 5 h after the cells were treated with CBLB502, LPS, or TNF using the Bright-Glo Luciferase Assay System.

Immunohistochemical Staining of p65.

Paraffin-embedded sections and primary mouse and human hepatocytes were stained with a rabbit polyclonal Ab against NF-κB p65 (catalog no. 7970; Abcam) and a rat monoclonal Ab against cytokeratin 8 (Troma-1; Developmental Studies Hybridoma Bank, University of Iowa), followed by secondary fluorochrome-conjugated Abs [p65 (green) and cytokeratin-8 (red)] and DAPI (blue) DNA stain.

FACS Analysis of Immune Cell Populations.

Single cell suspensions were obtained from mouse livers by perfusion with PBS followed by digestion with 0.05% collagenase type IV (Sigma–Aldrich) as described (58). Single cell suspensions of BMCs were obtained by flushing hind leg femurs and tibias with PBS. Following lysis of RBCs, cells were stained with a mixture of mAb against CD45 (clone 30-F11), CD11b (clone M1/70), GR-1 (clone RB6-8C5), F4/80 (clone BM8), CD63 (clone FA-11), CD3ε (clone 145-2C11), bTCR (clone H57-597), and CD49b (clone DX5) (all from BioLegend); fixed in 2% (vol/vol) formalin; and analyzed using a BD Biosciences Fortessa instrument and WinList 7.0 software (Verity Software House). Neutrophils were defined as CD45⁺, CD11b⁺, GR-1^hi (mean fluorescence intensity >10³); macrophages were defined as CD45⁺CD11b⁺F4/80⁺; NK cells were defined as CD45⁺CD3ε⁻CD49b⁺; and T cells were defined as CD45⁺CD3ε⁺αβTcR⁺.

FACTORIAL Analysis of Signaling Pathway Activity.

A FACTORIAL library consisting of 46 different plasmids each containing a single unique pathway-specific reporter transcription unit (RTU) (Attagene, Inc., www.attagene.com) was transfected into livers of NIH Swiss mice by a “hydrodynamic” transfection trough the tail vein that predominantly targets the plasmids to liver hepatocytes (20). Mice were allowed to recover after transfection for 2 wk and then were injected (s.c.) with CBLB502 (5 μg per mouse), LPS (10 μg per mouse), or PBS (eight mice per group). RTU activities were evaluated in total liver RNA samples collected 1 h later using multiplex RT-PCR detection with RTU-specific primers (18). Changes in transcriptional factor activities were calculated by normalizing the mean activities of RTUs in the livers of mice treated with CBLB502 or LPS to those in PBS-treated mice.

LER Procedure.

LER was achieved by using a nontraumatic clamp to occlude the hepatoduodenal ligament containing the hepatic artery and portal vein (59). CBLB502 or PBS was injected; 30 min later, clamps were removed to restore liver blood circulation, abdomens were closed, and mice were immediately subjected to total body irradiation using a ¹³⁷Cs Mark I-30 irradiator (J. L. Shepherd and Associates) with a dose rate of 2.2 Gy/min. Sham-LER was performed identically, with the exception that the hepatoduodenal ligament was occluded.