BALB/c mice (6–8 weeks old) were purchased from the Center of Medical Experimental Animals of Hubei Province (Wuhan, China). Mice were maintained in the accredited animal facility of Tongji Medical College, and used for studies approved by the Animal Care and Use Committee of Tongji Medical College. Murine H22, a human HepG2 hepatocarcinoma cell line, was purchased from China Center for Type Culture Collection (CCTCC, Wuhan, China) and cultured according to CCTCC guidelines.

Resveratrol (3,4′,5-trihydroxy-trans-stilbene) was purchased from Sigma-Aldrich (St. Louis, MO, USA). 6-Amino-4-(4-phenoxyphenylethylamino) quinazoline (QNZ) was purchased from Merck4Biosciences (Calbiochem, Darmstadt, Germany). Recombinant human HSP70 and HMGB1 were prepared from engineered bacteria carrying an expression plasmid kindly provided by Dr Richard Morimoto and Ji-Zhong Cheng (The University of Texas Medical Branch), purified and tested for endotoxin as previously described (16). The content of the endotoxin was <0.006 EU/ml for concentrations of recombinant HSP70 up to 50 μg/ml. Eukaryotic expression vectors psTLR2 and psTLR4 carrying cDNAs encoding the signal peptide and extracellular domain of murine TLR2 and TLR4 were constructed by the insertion of cDNA into plasmid pcDNA3.1 (Invitrogen, Carlsbad, CA, USA) in our laboratory. Antibodies were purchased from Cell Signaling Technology (Beverly, MA, USA) and Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA).

To downregulate HMGB1, RAGE or Beclin-1 in tumor cells, cells were transduced with HMGB1shRNA(h), RAGE-shRNA(h) or Beclin-1 shRNA(h) lentiviral particles or control shRNA lentiviral particles (Santa Cruz Biotechnology, Inc.), according to the manufacturer’s protocol.

For the treatment of tumors with DTC-Ms, HSP70 or HMGB1, BALB/c mice were inoculated with 1×10⁵ H22 cells by intramuscular injection into the right hind thigh. Following inoculation, the mice of treatment groups received intramuscular injection of DTC-Ms, HSP70 or HMGB1 (200 μg/mouse) at the inoculation site, starting on day 1 (d1) after tumor inoculation or d7 after tumor inoculation when the tumor was palpable, once every 2 days for 4 times. Mice from the control group received intramuscular injection of an equal volume of PBS. Mice were sacrificed and tumors were dissected and weighed at the indicated times.

For treatment of tumor with sTLR2 and sTLR4, BALB/c mice were inoculated with 1×10⁵ H22 cells by intramuscular injection into the right hind thigh. On d6, 8, 10, and 12 after inoculation, the mice of the treatment groups received an intramuscular injection (local naked DNA transfection) of 100 μg of psTLR2 or psTLR4 at the inoculation site. The control group mice received an equal volume of saline or an equal amount of pcDNA3.1 plasmid. In other experiments, mice received an intramuscular injection of psTLR2, psTLR4 or psTLR2/psTLR4 at the tumor inoculation site as described above, and also received intramuscular injection of HSP70 (200 μg/mouse) or saline at the inoculation site on d7, 9, 11 and 13 after tumor inoculation. Mice were sacrificed and tumors were dissected and weighed on d15 after inoculation.

For analysis of tumor metastasis after treatment with HSP70 or HMGB1, H22 cells or HMGB1shRNA-transfected H22 cells were injected into the liver of BALB/c mice (n=8/group). Mice were treated by intravenous injection of HSP70 or HMGB1 (200 μg/mouse) or saline once every 2 days 6 times after tumor inoculation. Mice were sacrificed and tumors were dissected and weighed on d16 after inoculation. The size of the main tumor nodule was measured and satellite tumor nodes were counted.

To detect extracellular HMGB1 in the tumor milieu, BALB/c mice were inoculated with 1×10⁵ H22 cells by intramuscular injection into the right hind thigh. Tissues at the inoculation site or tumor peripheral tissues were surgically excised at the indicated time points after inoculation. Tissues adjacent to the tumor or inoculation site were used as controls. Interstitial molecules from the tissues were prepared by digesting the tissue with collagenase and removing debris by centrifugation, and were then used for detection of HMGB1 by western blotting. HMGB1 antibody was purchased from Cell Signaling Technology (Danvers, MA, USA).

Total RNAs were isolated from cells or muscle tissues of mice using TRIzol reagent (Invitrogen), according to the manufacturer’s instructions. RT-PCR was used to detect mRNAs of psTLR2 and psTLR4 as previously described (17). To detect the mRNAs of sTLR2 and sTLR4 expressed by expression vectors, the primer sequences were: sTLR2 sense, 5′-CCAAGCTGGCTAGCGTTTA-3′ and antisense, 5′-CAAATGTTCAAGACTGCCCA-3′; sTLR4 sense, 5′-GACCCAAGCTGGCTAGCGTTT-3′ and anti-sense, 5′-TTTGTCTCCACAGCCACCA-3′; β-actin sense, 5′-ATGGG TCAGAAGGACTCCTATG-3′ and antisense, 5′-ATCTCC TGCTCGAAGTCTAGAG-3′.

The quantitative real-time RT-PCR for MMP-9 and β-actin was performed as previously described (17). MMP-9 sense, 5′-CAGATGATGGGAGAGAAGCAG-3′ and antisense, 5′-GAAGGTGAAGGGAAAGTGAC-3′; β-actin sense, 5′-ATGGGTCAGAAGGACTCCTATG-3′ and antisense, 5′-ATCTCCTGCTCGAAGTCTAGAG-3′. The results are expressed as the expression level of the gene relative to that of the housekeeping gene β-actin.

Following incubation of cells in the presence or absence of HSP70 or HMGB1, tissue samples after in vivo transfection were lysed or homogenized for western blot analysis as previously described (17). Primary antibodies and horseradish peroxidase-conjugated secondary antibodies were purchased from Chemicon (Temecula, CA, USA), Santa Cruz Biotechnology, Inc., R&D Systems (Minneapolis, MN, USA) and Cell Signaling Technology, respectively.

MMP-9 assay of protein samples was performed as previously described (17). Briefly, proteins prepared from the tumor tissues or tumor cells of each group were separated by 7.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) containing 1% gelatin. Gels were incubated in MMP activation buffer containing 50 mM Tris (pH 8.0) and 10 mM CaCl₂ at 37°C overnight, and then stained with 1% Coomassie Brilliant Blue R-250 for 3 h and destained in 10% (v/v) methanol and 5% (v/v) acetic acid. The area and gray scale of bands were analyzed by the HPIAS-1000 analytical system. The relative activity of MMP-9 was calculated using the formula: (band gray scale - background gray scale) × band area.

Tumor cells were stained with PE-Cy7 anti-mouse TLR2 antibody, PE anti-mouse TLR4 antibody (eBioscience, San Diego, CA, USA), or isotype control antibody for flow cytometric analysis. Parameters were acquired on a FACSCalibur Flow Cytometer and analyzed with CellQuest Software (both from BD Biosciences).

H22 cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) and cultured for 3 days. Cell proliferation was analyzed by flow cytometry. Flow cytometric analysis acquired at least 1×10⁴ events on a BD LSR II Flow Cytometer. The proliferation index was calculated in the responder population gate using the ModFit LT for Win32 Software.

H22 cells were washed with phosphate-buffered saline (PBS), and resuspended in PBS to a final concentration of 5×10⁷/ml. After four-rounds of freeze-thaw cycles followed by vortexing for 30 sec, the cells were removed by centrifugation. The supernatant contained a mixture of DTC-Ms. The concentration of DTC-Ms was defined by the concentration of protein, determined using Coomassie Bradford reagent (Thermo Fisher Scientific, Inc., Rockford, IL, USA), according to the manufacturer’s instructions.

The results are expressed as the mean value ± standard deviation (SD) and are interpreted by ANOVA-repeated measures test. Differences are considered statistically significant when P<0.05.

The effect of TLR2/4 signaling on the growth of hepatocarcinoma cells was measured after inoculating 1×10⁵ H22 cells into the hind thigh muscle and treating mice with DTC-Ms containing TLR2/4 ligands (18). Treatment with HSP70 and HMGB1 was used as control. Tumor growth increased significantly after two weeks of treatment with DTC-Ms. Notably, the effect of HSP70 on H22 cells was very similar to DTC-Ms stimulation (Fig. 1A). Moreover, at the same time interval, a rise in HMGB1 was detected in DTC-Ms and HSP70 treated tissues at the site of tumor cell inoculation, and increased in local tissues with the development of the tumor (Fig. 1B). These results suggested that HSP70 is a critical molecule(s) in DTC-Ms that can upregulate the expression of HMGB1 and may promote tumor development. As triggering of TLR4 and TLR2 in tumor cells promotes tumor cell proliferation (19) and both TLR2 and TLR4 were expressed on H22 cells (Fig. 1C) we verified whether TLR2 and TLR4 mediate the increased expression of HMGB1 and tumor promoting effects of HSP70. We expressed sTLR2 and sTLR4 in the tissue at the site of the palpable tumor by intramuscular transfection of sTLR2 and sTLR4 expression vectors (Fig. 1D). Expression of sTLR2 and sTLR4 suppressed tumor growth (Fig. 1E). The promoting effect of HSP70 on tumor growth was reduced by sTLR2 or sTLR4 expression (Fig. 1E), but the production of HMGB1 was not significantly influenced by blockade of TLR2 and TLR4 (Fig. 1F). Thus, HSP70-inducible expression of HMGB1 in the tumor microenvironment, independent of both TLR2 and TLR4, may promote tumor development.

Based on the above results, we next investigated whether HSP70 could increase H22 cell capability for malignant invasion by inducible expression of HMGB1 in the tumor microenvironment. In mice receiving 1×10⁵ H22 cells in the liver, injection of HSP70 increased the expression of HMGB1. Production of active MMP-9 also increased significantly in the HSP70 treatment groups (Fig. 2A), suggesting it is a major contributor to tumor invasion (20). Both increased tumor growth and increased satellite tumor nodes were observed in the HSP70 treatment groups (Fig. 2B), whereas treatment of mice with HMGB1shRNA transfection suppressed tumor growth and the formation of satellite tumor nodes (Fig. 2B). However, in the presence of HMGB1, increased satellite tumor nodes in the liver and increased production of active MMP-9 was observed (Fig. 2A and B), suggesting the increased capability for malignant invasion of the tumor was not induced by HSP70 but was due to HSP70-inducible expression of HMGB1. We used specific antibodies to neutralize HMGB1 activity that may be present in the HSP70 preparation to confirm the direct effect of HMGB1 on H22 cells cultured for 5 or 24 h in the presence of HSP70 or HMGB1. The production of active MMP-9 was increased in the presence of HMGB1 for 5 or 24 h but was unaffected by incubation in the presence of HSP70 for 5 h (Fig. 2C). In addition, anti-HMGB1 antibody treatment significantly reduced the increased production of active MMP-9 in the presence of HSP70 incubated for 24 h (Fig. 2C).

TLR2 and TLR4 can activate NF-κB that mediates the effects of HSP70 (20). The proliferation of H22 cells was promoted by administration of HSP70 (Fig. 3A). Moreover, stimulation of H22 cells with HSP70 induced a biphasic temporal pattern of I-κB activation, an inhibitory protein of NF-κB, and increased the quantity of NF-κB in the cell nucleus, with a rapid increase in phosphorylation within 30 min, followed by a second rise at ~6 h (Fig. 3B). Preincubation of H22 cells with resveratrol, a TLR4 and TLR2 signaling inhibitor (21), inhibited the HSP70-induced first phase of NF-κB phosphorylation but not the second phase (Fig. 3B), and the proliferation of H22 cells was significantly suppressed (Fig. 3A). QNZ, an inhibitor of NF-κB, abrogated the effect of HSP70 (Fig. 3A) suggesting that both TLR2 and TLR4 mediated the first phase of NF-κB activation and were involved in promoting the effect of HSP70 on tumor cell proliferation. Next, we looked at whether the second phase of NF-κB phosphorylation may last if HSP70 was removed. To test this, H22 cells were treated with HSP70 for 4 h and then specific HSP70 neutralizing antibodies were administered. I-κB activation increased at ~6 h, although not as high as that when in the continuous presence of HSP70 (Fig. 3B). Collectively, the results suggest that HSP70 stimulation in tumor cells may result in a previously unrecognized activation of NF-κB but not dependent of TLR4 and TLR2 signal pathways.

HSP70 regulates diverse signaling pathways involving NF-κB in different types of cells (22). However, the mechanism(s) of the second phase of NF-κB activation by HSP70 is unknown. Given that HSP70 can induce HMGB1 release via JNK activation and Beclin-1 expression and that HMGB1 can trigger NF-κB activation (8,23), we hypothesized that the second phase of NF-κB activation in H22 cells is caused by Beclin-1-derived HMGB1 production induced by HSP70. To test this hypothesis, we examined the temporal production of HMGB1 by HSP70. In H22 cells, HSP70 induced rapid JNK phosphorylation, significant upregulation of Beclin-1 and HMGB1 at 4 h (Fig. 4A and B), which preceded the second phase of NF-κB activation. Then, we examined the effect of exogenously delivered HMGB1 on NF-κB signaling. Stimulation of H22 cells with HMGB1 induced rapid NF-κB phosphorylation (Fig. 4D). These results indicated that H22 cells activated NF-κB rapidly but the second phase of NF-κB activation was not induced by HMGB1. The second phase of NF-κB activation induced by HSP70 in H22 cells could be due to HMGB1 generation via JNK mediating Beclin-1. However, when cancer cells were pretreated with a selective JNK inhibitor (SP600125) HSP70-induced JNK, phosphorylation was inhibited (Fig. 4B), which prevented Beclin-1 and HMGB1 production in H22 cells (Fig. 4C). HMGB1 production by HSP70 was also inhibited by Beclin-1 shRNA (Fig. 4C). Moreover, the JNK inhibitor prevented the second, but not the first, phase of HSP70-induced activation of NF-κB (Fig. 4D). From these results, we conclude that HSP70 induces the second phase of NF-κB activation via JNK-mediated Beclin-1-dependent HMGB1 production in H22 cells.

HMGB1 mediates the response to inflammation-based carcinogenesis by binding with high affinity to several receptors including the RAGE and TLRs 2 and 4, thereby augmenting tumor growth and invasion. To determine whether RAGE and/or TLR2/4 mediate HMGB1-induced increased production of active MMP-9, a target-specific shRNA against these receptors was transfected into tumor cells. Knockdown of RAGE in cancer cells diminished HMGB1-induced increased production of MMP-9 (Fig. 5A). By contrast, there was no effect on HMGB1-induced increased production of MMP-9 when resveratrol was used to inhibit TLR2/4 (Fig. 5A). This suggested RAGE is required for HMGB1 promotion of tumor invasion by increased production of active MMP-9. To analyze the pathway(s) involved in the effect of RAGE-mediated NF-κB signaling, we stimulated H22 cells with HMGB1 in the presence of QNZ (NF-κB inhibitor) and resveratrol (TRIF inhibitor). The effect of RAGE-mediated NF-κB signaling was completely abolished by QNZ, but not resveratrol (Fig. 5B), suggesting that NF-κB was involved in RAGE signaling for MMP-9 expression in H22 cells. A human hepatocarcinoma cell line, HepG2, was used to evaluate the effect of RAGE signaling. The human cancer cell line showed a similar pattern of increased MMP-9 expression in response to HMGB1 stimulation (Fig. 5C). Taken together, these results suggested that RAGE signaling activated NF-κB to regulate the expression of MMP-9 in hepatocarcinoma H22 cells.