A murine cellular model of necroinflammation displays RAGE‐dependent cytokine induction that connects to hepatoma cell injury

Abstract Unresolved inflammation maintained by release of danger‐associated molecular patterns, particularly high‐mobility group box‐1 (HMGB1), is crucial for hepatocellular carcinoma (HCC) pathogenesis. To further characterize interactions between leucocytes and necrotic cancerous tissue, a cellular model of necroinflammation was studied in which murine Raw 264.7 macrophages or primary splenocytes were exposed to necrotic lysates (N‐lys) of murine hepatoma cells or primary hepatocytes. In comparison to those derived from primary hepatocytes, N‐lys from hepatoma cells were highly active—inducing in macrophages efficient expression of inflammatory cytokines like C‐X‐C motif ligand‐2 , tumor necrosis factor‐α, interleukin (IL)‐6 and IL‐23‐p19. This activity associated with higher levels of HMGB1 in hepatoma cells and was curbed by pharmacological blockage of the receptor for advanced glycation end product (RAGE)/HMGB1 axis or the mitogen‐activated protein kinases ERK1/2 pathway. Analysis of murine splenocytes furthermore demonstrated that N‐lys did not comprise of functionally relevant amounts of TLR4 agonists. Finally, N‐lys derived from hepatoma cells supported inflammatory splenic Th17 and Th1 polarization as detected by IL‐17, IL‐22 or interferon‐γ production. Altogether, a straightforward applicable model was established which allows for biochemical characterization of immunoregulation by HCC necrosis in cell culture. Data presented indicate a remarkably inflammatory capacity of necrotic hepatoma cells that, at least partly, depends on the RAGE/HMGB1 axis and may shape immunological properties of the HCC microenvironment.


| INTRODUC TI ON
Necrosis and its related cell death entities necroptosis and pyroptosis initiate and/or perpetuate inflammation 1,2 thereby shaping the course of various diseases, among others hepatitides of diverse causes. 3,4 Specifically, hepatic ischaemia reperfusion injury 5 and acetaminophen-induced liver injury [5][6][7] as well as alcoholic (ASH) and non-alcoholic steatohepatitis (NASH) 4,8 are well-described examples of acute or chronic hepatic diseases in which necroinflammation is at the epicentre of pathophysiology. By fuelling chronic inflammation, necrotic cell death is moreover key to progression of NASH/ASH or viral hepatitides to hepatocellular carcinoma (HCC). [8][9][10] On a biochemical level, sterile inflammation induced by necrosis is mediated by release/misplacement of intracellular molecules normally invisible to receptors of innate immunity. According to the danger hypotheses, these danger-associated molecular patterns (DAMPs) initiate inflammatory signalling by activating key innate receptor systems. 4 Among DAMPs well characterized with regard to the inflamed liver are histones mainly targeting toll-like receptors (TLR)-2/4/9 (in case of TLR9 likely in combination with DNA), [11][12][13] RNA targeting discrete endosomal nucleic acids-sensing TLRs such as TLR3, 14 and high-mobility group box-1 (HMGB1) targeting mainly the receptor for advanced glycation end products (RAGE). 15,16 Resultant cellular responses not only maintain inflammatory hepatic diseases but also drive hepatocellular carcinogenesis 8,9 for which accumulating evidence indicates pivotal action of RAGE signalling.
Specifically, HMGB1/RAGE signalling has been implicated in sustaining an inflammatory tumour microenvironment that supports hepatic oval cell and carcinoma proliferation as well as tumour invasion and metastasis. [17][18][19][20] Herein, we established a feasible murine cell culture model that allows for investigation of immunoregulation by necrotic HCC cells under defined conditions. Data presented prove a highly inflammatory nature of necrotic HCC cells which translates into (at least) partly RAGE/HMGB1-dependent up-regulation of cytokine production. Among those are C-X-C motif ligand-2 (CXCL2), tumour necrosis factor (TNF)-α, interleukin (IL)-36α, IL-23-p19 and IL-17 which jointly have the capability to maintain an inflammatory pro-cancerous micromilieu in the diseased liver tissue.
Spleens were excised and transferred to 5 mL ice-cold RPMI 1640 medium. Tissue was destroyed over a nylon cell strainer (70 μm; BD Biosciences, Heidelberg, Germany). Cell suspensions were centrifuged at 500 g for 5 minutes at 4°C and resuspended in 2 mL 0.83% NH 4 Cl for 2 minutes at room temperature. Red blood cell lysis was stopped by adding 10 mL cold RPMI 1640 medium. Splenocytes were collected by centrifugation, washed once with RPMI and resuspended in RPMI 1640 supplemented with 10% heat-inactivated FCS and 100 U/mL penicillin, 100 μg/mL streptomycin. 3 × 10 6 cells were seeded on 24-well polystyrene plates (Greiner) in 0.5 mL culture medium.
Thereafter, remaining adherent cells on polystyrene plates were scraped off followed by centrifugation at 17000 g (4°C, 10 minutes). Insoluble pellets were discarded, whereas supernatants were aliquoted and denoted necrotic lysates (N-lys) when derived from Hepa1-6 cells or Pri-N-lys when derived from primary hepatocytes or N-lys-HepG2 when derived from HepG2 cells. For storage (at −80°C), 10% FCS was added to the lysates.
DNA content is regarded a most accurate correlate reflecting epithelial cell numbers. Accordingly, a PicoGreen-based highly sensitive and reliable method was applied for quantification of the dsDNA content in aforementioned cellular lysates. For that purpose, the Quant-iT PicoGreen assay kit was used according to the manufacturer's instructions (Thermo Fisher Scientific). Obtained dsDNA concentrations were used for sample calibration (in ng/mL) which enabled usage of cell lysates as stimulus for cultured cells.

| Quantification of viable cells by WST-1 assay
RAW 264.7 macrophages or splenocytes were seeded onto 96-well polystyrene plates. Cells were stimulated as indicated or cultured as unstimulated control. After the indicated time-points, viable cells were quantified using WST-1 assay reagent (Roche Diagnostics) according to the manufacturer's instructions.

| Statistical analysis
Data are shown as means ± SD or as means ± SEM (as indicated) and presented as fold induction, pg/mL, ng/mL, or as per cent of N-lys alone. Statistical analysis was performed as indicated in the legends by one-way ANOVA with post hoc Bonferroni correction (GraphPad 5.0) or unpaired Student's t test.

| Immunoactivation and cytokine responses detected in RAW 264.7 macrophages under the influence of necrotic hepatoma lysates
In order to characterize immunoregulatory/stimulatory properties of cellular material derived from necrotic hepatoma tissue, murine RAW 264.7 macrophages were exposed to increasing amounts of necrotic cell lysates derived from murine hepatoma Hepa1-6 cells Immunostimulatory characteristics of necrotic hepatoma tissue were not confined to the murine system. Specifically, necrotic cell lysates derived from human hepatoma HepG2 cells were likewise able to potently activate human macrophage-like THP1 (mTHP1) cells as detected by analysis of IL-8 release ( Figure S1).

| Macrophage activation by necrotic hepatoma lysates is dependent on RAGE and ERK1/2 signalling
As already referred to, RAGE signalling is supposed to substantially impact inflammatory liver carcinogenesis. 16,23 In order to determine the relevance of RAGE concerning macrophage activation by N-lys, RAW 264.7 cells were exposed to RAGE antagonistic FPS-ZM1 (herein denoted Ri). As shown in Figure 2 In contrast, IL-1β-induced CXCL2 remained unaffected by Ri (E).
Since TLR9-activating cellular DNA may serve as biologically active DAMP 11 in this experimental setting, outcome of macrophage pre-incubation (30 minutes) with the TLR9 antagonist ODN2088 was evaluated. Whereas TLR9-activating B-type CpG-ODN1826induced CXCL2 was nullified in presence of ODN2088, this compound failed to significantly affect release of this chemokine under the influence N-lys ( Figure 2F).
Because potent activation of MAPK is a key feature of RAGE signal transduction, 24 the relevance of this pathway was investigated by its blockage using the MEK inhibitor U0126. In fact, activation by N-Lys strongly stimulated ERK1/2 activation ( Figure 2G) and pre-incubation (30 minutes) with U0126 actually suppressed associated CXCL2 secretion ( Figure 2H).

| HMGB1 serves inflammatory functions in necrotic hepatoma lysates
HMGB1 is a key DAMP and agonistic RAGE ligand that potently initiates RAGE-dependent inflammatory signalling. 16,23 In order to determine the functional relevance of HMGB1 in the context of pro-inflammatory N-lys properties, lysates were preincubated with RAGE antagonist glycyrrhizin (Gly). 25 Viability of RAW 264.7 cells was not affected by Gly at chosen concentrations. 26 In fact, pharmacological inhibition of HMGB1 by this compound significantly reduced N-Lys-mediated CXCL2 release by 37.6% at 500 μmol/L ( Figure 3A). Recombinant box-A domain F I G U R E 1 RAW 264.7 macrophages were kept as unstimulated control or stimulated with the indicated concentration of N-lys for 6 h (A, B, D, E, G, L) or 16 h (C, F, H-K). mRNA expression of indicated genes (A, D, G-L) was determined by real-time PCR, normalized to that of glycerinaldehyde-3-phosphatedehydrogenase and is shown as mean fold induction compared to unstimulated control ± SD (A, D, G-K: n = 4-5; L: n = 3). *P < 0.05, **P < 0.01, ***P < 0.001 compared to unstimulated control; raw data were analysed by one-way ANOVA with post hoc Bonferroni correction. Release of CXCL2 (B, C) or TNFα (E, F) was determined by ELISA. Data are shown as means ± SD (B, C: n = 5; E, F: n = 4-7). *P < 0.05, **P < 0.01, ***P < 0.001 compared to unstimulated control; statistical analysis, one-way ANOVA with post hoc Bonferroni correction of HMGB1 was used as a further specific HMGB1 antagonist. 15 Notably, pre-incubation with recombinant box-A domain resulted in inhibitory effects very similar to those of Gly ( Figure   S2). Accordingly, HMGB1 was readily detectable by immunoblot analysis in N-lys obtained from Hepa1-6 or Hep5.1D cells.
Immunoblot analysis likewise revealed that, compared to N-lys derived from aforementioned hepatoma cells, those derived from primary murine hepatocytes (Pri-N-lys) displayed considerably less HMGB1 protein expression ( Figure 3B). Specifically, hepatocyte isolates from three individual C57Bl/6J mice were analysed (H1-H3). Of note, longer exposure times revealed welldetectable HMGB1 expression also in the lysate denoted H1 F I G U R E 2 A-E, H: RAW 264.7 macrophages were kept as unstimulated control or stimulated with N-lys (as indicated 40 or 13 ng/mL) or IL-1β (20 ng/mL) for 6 h. Where indicated, cells were pre-incubated with Ri or U0126 (at indicated concentrations) for 30 min. All cultures were adjusted to a final concentration of 0.03% dimethyl sulfoxide (DMSO) (vehicle for Ri) or 0.1% DMSO (vehicle for U0126). Secreted CXCL2 (A, E, H) or TNFα (B) was determined by ELISA. Data are shown as means ± SD (A: n = 3-9, B: n = 9, E: n = 6, H: n = 3). *P < 0.05, **P < 0.01, ***P < 0.001 compared to unstimulated control; # P < 0.05, ## P < 0.01, ### P < 0.001; statistical analysis, one-way ANOVA with post hoc Bonferroni correction. IL-36α (C) and Irg1 (D) mRNA was determined by real-time PCR, normalized to that of GAPDH, and is shown as mean fold induction compared to unstimulated control ± SD (C: n = 7-8, D: n = 4-5). ***P < 0.001 compared to unstimulated control, ## P < 0.01, ### P < 0.001; raw data were analysed by one-way ANOVA with post hoc Bonferroni correction. F, RAW 264.7 macrophages were kept as unstimulated control or stimulated N-lys (40 ng/mL) or with agonistic B-type ODN1826 (2 µmol/L) for 16 h. Where indicated, cells were pre-incubated with antagonist ODN2088 (10 µmol/L) for 30 min. Secreted CXCL2 was determined by ELISA. Data are shown as means ± SD (n = 3). ***P < 0.001 compared to unstimulated control; ### P < 0.001; statistical analysis, one-way ANOVA with post hoc Bonferroni correction. G, RAW 264.7 macrophages were kept as unstimulated control or were stimulated with 67 ng/mL N-lys. After 10 min, cellular content of p-p44/p-p42 and total p44/p42 was determined by immunoblot analysis. One representative of three independently performed experiments is shown (data not shown). Those observations precisely confirm a previous report demonstrating that Hepa1-6 cells in fact express far higher levels of HMGB1 protein as compared to primary hepatocytes. 17 Remarkably, Hepa1-6 hepatoma cells were also found to express far higher levels of HMGB1 as compared to RAW264.7 macrophages ( Figure S3).
In order to test whether different levels of HMGB1 in cellular lysates may affect their immunostimulatory properties, RAW 264.7 macrophages were exposed to N-lys (from Hepa1-6 cells) or to each of the lysates derived from aforementioned primary hepatocytes (Pri-N-lys H1 to H3 reflecting n = 3 in Figure 3C-E). In concurrence with the view of HMGB1 contributing to inflammatory functions of N-lys, experiments revealed that, as compared to Pri-N-lys, N-lys was considerably more potent in stimulating expression of CXCL2 ( Figure 3C,D) or TNFα ( Figure 3E). Notably, albeit on a much lower level, Pri-N-lys induced significant inflammatory CXCL2 and TNFα production (Figure 3, see insets).

| Immunostimulatory properties of necrotic hepatoma lysates as detected in murine splenocytes
In order to investigate characteristics of N-lys against the background of a naturally occurring murine leucocyte composition, the cellular model of freshly isolated splenocytes was applied. In fact, N-lys efficiently stimulated induction of CXCL2 by splenocytes which was observed on mRNA and protein level ( Figure 4A,B). By investigating splenocytes derived from TLR4-deficient mice, a potential LPS contamination of N-lys was excluded ( Figure 4B). As expected, in these same experiments LPS-induced CXCL2 release was nullified in TLR4 −/− mice ( Figure 4C). A variety of further indicators of immunoactivation were up-regulated by N-lys in a manner similar to CXCL2. Those included TNFα ( Figure 4D,E), IL-23-p19 ( Figure 4F), IL-6 ( Figure 4G), IL-10 ( Figure 4H) and IRG1 ( Figure 4I).

Finally, inflammatory polarization of splenic T cells for Th1-and
Th17-like activation was enhanced under the influence of N-lys, a F I G U R E 3 A, RAW 264.7 macrophages were kept as unstimulated control or were stimulated with 4 ng/mL N-lys for 16 h. Where indicated, N-lys was pre-treated with glycyrrhizin (Gly; 500 or 100 µmol/L, final concentrations on cells) for 30 min. All cultures were adjusted to a final concentration of 0.4% DMSO (vehicle for Gly). Secreted CXCL2 was determined by ELISA. Data are shown as means ± SD (n = 5). ***P < 0.001 compared to unstimulated control; # P < 0.05; statistical analysis, one-way ANOVA with post hoc Bonferroni correction. B, Content of HMGB-1 in N-Lys or Pri-N-lys (equalling 20 ng of DNA) was determined by immunoblot analysis. C-E, RAW 264.7 macrophages were kept as unstimulated control or stimulated with 40 ng/mL N-lys or indicated concentrations of Pri-N-lys (40 or 80 ng/mL) for 24 h. CXCL2 (C) and TNFα (E) mRNA was determined by real-time PCR, normalized to that of GAPDH, and is shown as mean fold induction compared to unstimulated control ± SD (n = 3; ***P < 0.001 compared to unstimulated control; ## P < 0.01, ### P < 0.001 compared to N-lys; raw data were analysed by one-way ANOVA with post hoc Bonferroni correction. Insets: mRNA expression of CXCL2 (C) or TNFα (E) by Pri-N-lys (Pri, 80 ng/mL) stimulated or unstimulated RAW 264.7 macrophages is shown at a larger y-axis scale. *P < 0.05 compared to unstimulated control; raw data were analysed by unpaired t test. D: Secreted CXCL2 was determined by ELISA. Data are shown as means ± SD (n = 3). **P < 0.01 compared to unstimulated control; ## P < 0.01 compared to N-lys; raw data were analysed by one-way ANOVA with post hoc Bonferroni correction. Inset: Secreted CXCL2 by Pri-N-lys (Pri, 80 ng/mL) stimulated or unstimulated RAW 264.7 macrophages is shown at a larger y-axis scale. *P < 0.05 compared to unstimulated control; raw data were analysed by unpaired Student's t test phenomenon well detectable by amplified production of respective signature cytokines, namely IFNγ ( Figure 4J), IL-22 ( Figure 4K) and IL-17 ( Figure 4L).

| D ISCUSS I ON
By using an in vitro cell culture model, we demonstrate an augmented inflammatory potential of murine necrotic hepatoma cells that, at least partly, relies on the RAGE/HMGB1 axis. Albeit in principle capable of mediating inflammatory macrophage activation, N-lys derived from primary hepatocytes displayed far modest biological activity in this context. Those functional differences associated with strikingly divergent HMGB1 protein expression detectable in hepatoma cells and primary hepatocytes, respectively ( Figure 3 and reference 17). Notably, HMGB1 which connects to inflammatory carcinogenesis [19][20][21][22] is increased in cancerous HCC tissues as well as in patients' sera and serves as prognostic marker indicating F I G U R E 4 Splenocytes of male C57BL/6J wt mice were kept as unstimulated control or were stimulated with indicated concentrations of N-lys for 6 h (A, F) or 16 h (D, E, G, H, I). mRNA (A, D, F, H, I) for indicated genes was determined by real-time PCR, normalized to that of GAPDH, and is shown as mean fold induction compared to unstimulated control ± SEM (A: n = 4-5; D: n = 5; F: n = 6; H: n = 5; I: n = 7). *P < 0.05, **P < 0.01 compared to unstimulated control; raw data were analysed by one-way ANOVA with post hoc Bonferroni correction (A, D, H) or unpaired Student's t test (F, I). Secreted TNFα (E) or IL-6 (G) was determined by ELISA. Data are shown as means ± SEM (E: n = 5; G: n = 3-5). ***P < 0.001 compared to unstimulated control; statistical analysis, one-way ANOVA with post hoc Bonferroni correction. B and C, Splenocytes of male C57BL/6J wt mice or male TLR4 deficient mice were kept as unstimulated control or were stimulated with N-lys (B, 67 ng/mL) or LPS (C, 1 µg/mL) for 16 h. Secreted CXCL2 was determined by ELISA. Data are shown as means ± SEM (n = 3). ***P < 0.001 compared to unstimulated control of the same genotype, ### P < 0.001; statistical analysis, one-way ANOVA with post hoc Bonferroni correction. J-L, Splenocytes of male C57BL/6J wt mice were kept as unstimulated control or were cultivated under Th1 (J) or Th17 conditions (K, L) with or without N-lys (67 ng/mL). After 4 d, secreted IFNγ (J), IL-22 (K) and IL-17 (L) were determined by ELISA. Data are shown as means ± SEM (J, L: n = 5, K: n = 6). ***P < 0.001 compared to unstimulated control, # P < 0.05, ### P < 0.001; statistical analysis, oneway ANOVA with post hoc Bonferroni correction progression of clinical HCC. 17,[27][28][29][30][31][32][33] Notably, HMGB1 antagonizing glycyrrhizin and recombinant box-A domain only partly abolished inflammatory action of N-lys which concurs with the view that RAGE is a multi-ligand receptor 23 likely activated by further, to be identified, DAMPs present in N-lys. In that context, it will be interesting in coming studies to characterize in detail differential properties of hepatic necroinflammation in cancerous vs non-cancerous settings.
This should include identification of crucial DAMPs and their potential to regulate associated immunoactivation.
A focus of the current study was to characterize in murine RAW 264.7 macrophages and freshly isolated splenocytes cytokine responses initiated by necroinflammation under the influence of N-lys. In that context, we report on a remarkably potent gene induction of CXCL2 which is regarded a murine functional homologue of IL-8 and was previously identified as crucial for neutrophil-driven experimental HCC. 34 Interestingly, CXCL2 likewise correlates with progression in HCC patients. 35   IFNγ displays complex immunoregulatory characteristics that include induction of potentially tumour-supporting IL-18 binding protein, 52,53 its tumour-suppressive properties apparently override in the context HCC. 54 In contrast, Th17-derived IL-17 which is supported by tumour-associated macrophages clearly promotes progression in experimental and clinical HCC. [55][56][57] Similarly, IL-22 is a well-defined pathogenic factor in HCC that likewise indicates patients' prognosis. [58][59][60] Altogether, the cell culture model used herein validates a strong inflammatory potential of necrotic hepatoma cells that, at least partly, depends on the RAGE/HMGB1 axis and mediates production of key cytokines known to determine progression of HCC. Further biochemical analysis of necrotic normal hepatocytes vs necrotic cancerous hepatoma cells may sheet further light on the role that necroinflammation plays in HCC pathogenesis.

ACK N OWLED G EM ENTS
This study was supported by departmental funding (pharmazentrum frankfurt, General Pharmacology and Toxicology) to HM and by a grant from the DFG Graduate School GRK2336 to HM/LL and HM/ SG.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.