Sorafenib perpetuates cellular anticancer effector functions by modulating the crosstalk between macrophages and natural killer cells


  • Martin Franz Sprinzl,

    1. Institute of Virology, Technische Universität München/Helmholtz Zentrum München, München, Germany
    2. First Medical Department, Johannes Gutenberg Universität, Mainz, Germany
    Search for more papers by this author
  • Florian Reisinger,

    1. Institute of Virology, Technische Universität München/Helmholtz Zentrum München, München, Germany
    Search for more papers by this author
  • Andreas Puschnik,

    1. Institute of Virology, Technische Universität München/Helmholtz Zentrum München, München, Germany
    Search for more papers by this author
  • Marc Ringelhan,

    1. Institute of Virology, Technische Universität München/Helmholtz Zentrum München, München, Germany
    2. 2nd Medical Department, Klinikum rechts der Isar, Technische Universität München, Munich, Germany
    Search for more papers by this author
  • Kerstin Ackermann,

    1. Institute of Virology, Technische Universität München/Helmholtz Zentrum München, München, Germany
    2. Immune-Monitoring and Clinical Cooperation Group Antigen-specific Immunotherapy, Helmholtz Zentrum München, München, Germany
    Search for more papers by this author
  • Daniel Hartmann,

    1. Department of Surgery, Technische Universität, Klinikum rechts der Isar, München, Germany
    Search for more papers by this author
  • Matthias Schiemann,

    1. Institute for Medical Microbiology, Immunology and Hygiene, Technische Universität München
    2. Immune-Monitoring and Clinical Cooperation Group Antigen-specific Immunotherapy, Helmholtz Zentrum München, München, Germany
    Search for more papers by this author
  • Arndt Weinmann,

    1. First Medical Department, Johannes Gutenberg Universität, Mainz, Germany
    Search for more papers by this author
  • Peter Robert Galle,

    1. First Medical Department, Johannes Gutenberg Universität, Mainz, Germany
    Search for more papers by this author
  • Marcus Schuchmann,

    1. Immune-Monitoring and Clinical Cooperation Group Antigen-specific Immunotherapy, Helmholtz Zentrum München, München, Germany
    Search for more papers by this author
  • Helmut Friess,

    1. Department of Surgery, Technische Universität, Klinikum rechts der Isar, München, Germany
    Search for more papers by this author
  • Gerd Otto,

    1. Department of Hepato-biliary and Transplant Surgery, Johannes Gutenberg Universität, Mainz, Germany
    Search for more papers by this author
  • Mathias Heikenwalder,

    1. Institute of Virology, Technische Universität München/Helmholtz Zentrum München, München, Germany
    Search for more papers by this author
  • Ulrike Protzer

    Corresponding author
    1. Institute of Virology, Technische Universität München/Helmholtz Zentrum München, München, Germany
    2. Immune-Monitoring and Clinical Cooperation Group Antigen-specific Immunotherapy, Helmholtz Zentrum München, München, Germany
    • Institute of Virology, Technische Universität München / Helmholtz Zentrum München, Trogerstr. 30, D-81675 München, Germany
    Search for more papers by this author
    • fax: +49-89-41406823

  • Potential conflict of interest: Nothing to report.

  • Supported by the Helmholtz Alliance for Immunotherapy of Cancer (HA-202) and the DFG ( SFB TR 36). M.S. received a clinical leave stipend from the Helmholtz Alliance. M.H. was supported by ERC starting grant (Liver Cancer Mechanism).


Alternatively polarized macrophages (Mϕ) shape the microenvironment of hepatocellular carcinoma (HCC) and temper anticancer immune responses. We investigated if sorafenib alters the HCC microenvironment by restoring classical macrophage polarization and triggering tumor-directed natural killer (NK) cell responses. In vivo experiments were conducted with sorafenib (25 mg/kg)-treated C57BL/6 wildtype as well as hepatitis B virus (HBV) and lymphotoxin transgenic mice with and without HCC. Monocyte-derived Mϕ or tumor-associated macrophages (TAM) isolated from HCC tissue were treated with sorafenib (0.07-5.0 μg/mL) and cocultured with autologous NK cells. Mϕ and NK cell activation was analyzed by flow cytometry and killing assays, respectively. Cytokine and growth factor release was measured by enzyme-linked immunosorbent assay. Short-term administration of sorafenib triggered activation of hepatic NK cells in wildtype and tumor-bearing mice. In vitro, sorafenib sensitized Mϕ to lipopolysaccharide, reverted alternative Mϕ polarization and enhanced IL12 secretion (P = 0.0133). NK cells activated by sorafenib-treated Mϕ showed increased degranulation (15.3 ± 0.2% versus 32.0 ± 0.9%, P < 0.0001) and interferon-gamma (IFN-γ) secretion (2.1 ± 0.2% versus 8.0 ± 0.2%, P < 0.0001) upon target cell contact. Sorafenib-triggered NK cell activation was verified by coculture experiments using TAM. Sorafenib-treated Mϕ increased cytolytic NK cell function against K562, Raji, and HepG2 target cells in a dose-dependent manner. Neutralization of interleukin (IL)12 or IL18 as well as inhibition of the nuclear factor kappa B (NF-κB) pathway reversed NK cell activation in Mϕ/NK cocultures. Conclusion: Sorafenib triggers proinflammatory activity of TAM and subsequently induces antitumor NK cell responses in a cytokine- and NF-κB-dependent fashion. This observation is relevant for HCC therapy, as sorafenib is a compound in clinical use that reverts alternative polarization of TAM in HCC. (HEPATOLOGY 2013;57:2358–2368)

Tumor-associated macrophages (TAM) located in the hepatocellular carcinoma (HCC) environment increase HCC recurrence after resection and reduce patient survival.1, 2 TAM thereby fosters tumor cell proliferation and tumor spread.3 Natural killer (NK) cell numbers and activity, on the other hand, are associated with lower HCC stages and improved patient survival.4, 5 An outstanding feature of TAM is their cytokine-dependent inhibition of lymphocyte and NK cell functions.6 TAM also promote T-cell exhaustion7 and are associated with the intratumoral accumulation of regulatory cells contributing to immune tolerance.2 TAM themselves represent alternatively polarized macrophages (Mϕ), which are opposed by proinflammatory Mϕ populations.3 Mϕ plasticity therefore balances tumor protection and immunogenic tumor rejection. Hence, interference with Mϕ polarization leading to an anticancer immune response represents a potential approach for therapy.

Tyrosine kinase inhibitors are promising candidates for TAM-directed therapy, as Mϕ polarization is regulated by tyrosine kinases.3, 6 Sorafenib, a multi-tyrosine kinase inhibitor, has become a standard palliative treatment for HCC.8 Sorafenib blocks different tyrosine kinases, such as rat sarcoma (RAS), rat fibrosarcoma (RAF), and extracellular-regulated protein kinase (ERK), thereby inhibiting proliferation and survival of tumor cells. In combination with antiangiogenic effects, this eventually results in HCC regression.9 Previous reports also indicate that sorafenib subverts immune responses by mitigating mitogen-activated protein kinase (MAPK) and nuclear factor kappa B (NF-κB) signaling.10, 11 In addition, inhibition of MAPKp38 by sorafenib may affect Mϕ polarization and innate immune surveillance.12 Yet the impact of sorafenib on immune responses within the tumor milieu is unknown. Particularly the effect of sorafenib on the interaction between TAM and NK cells remains elusive.

In this work we studied sorafenib-triggered activation of polarized Mϕ, which show a TAM-like phenotype. Sorafenib-dependent Mϕ induction eventually affected NK cells, which displayed enhanced activity against tumor cells. This interaction with NK cells was confirmed for autologous TAM isolated from human HCC tissue. The observed sorafenib-triggered NK cell stimulation was dependent on NF-κB activation and cytokine induction in polarized Mϕ.


AFP, alpha-fetoprotein; C57BL/6wt, C57BL/6 wildtype; Cr, [51Cr]chromium; CSF-1, colony stimulating factor-1; DMSO, dimethylsulfoxide; EGTA, ethylene glycol tetraacetic acid; ELISA, enzyme linked immunosorbent assay; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; HBSS, Hank's balanced salt solution; IFN, interferon; IL, interleukin; JAK, Janus kinase; LPS, lipopolysaccharide; L-SIGN, liver/lymph node-specific ICAM-3-grabbing nonintegrin; LTα/β, lymphotoxin α/β; Mϕ, macrophage; MAPK, mitogen-activated protein kinase; NF-κB, nuclear factor “kappa-light-chain-enhancer” of activated B-cells; NK, natural killer; PMA, phorbol myristate acetate; RAF, rat fibrosarcoma; RAS, rat sarcoma; STAT, Signal Transducers and Activators of Transcription; tg, transgenic; TNF, tumor necrosis factor; UV, ultraviolet.

Materials and Methods

Animal Experiments.

C57BL/6 wildtype mice (C57BL/6wt), hepatitis B virus replicating HBV1.3.32 (HBV-tg),13 or albumin-promoter-controlled lymphotoxin-α/β transgenic mice (LTα/β-tg)14 were maintained under pathogen-free conditions. C57BL/6wt mice were used for experiments at the age of 6 months, LTα/β-tg mice at 14-28, and HBV-tg mice at 21-25 months. Animal experiments were performed in accordance with the German legislation governing animal studies and the Principles of Laboratory Animal Care guidelines (National Institutes of Health, NIH).

Cell Culture Experiments.

NK cells (CD3/CD56+) were sorted from blood leukocytes with a MoFlow (Beckman Coulter, Krefeld, Germany) or were purified untouched using magnetic beads (Miltenyi Biotech, Bergisch-Gladbach, Germany). Circulating CD14+ monocytes were enriched by positive magnetic isolation (Miltenyi) and were cultured in the presence of 10 ng/mL colony stimulating factor-1 (CSF-1) (Peprotech, Hamburg, Germany) for 1 week to generate polarized macrophages (Mϕ). Also, 3 × 104 NK cells, Mϕ, and TAM per well were cultured in flat-bottom 96-well plates with RPMI-1640 medium (Gibco, Carlsbad, CA), supplemented with fetal calf serum (10%), L-glutamine (1%), penicillin (1%), and streptomycin (1%) (all Sigma-Aldrich, St. Louis, MO). K562, Raji, and HepG2 cells were maintained under equal conditions. Then 3 × 104 NK cells and TAM per well were used for coculture at the day of isolation.

Human Tissue Preparation.

Following patient informed consent and local Ethics Committee approval, TAM were obtained from histological confirmed HCC tissue. Tumor was dissected in Hank's balanced salt solution / ethylene glycol tetraacetic acid (HBSS/EGTA) (2 mM) / heparin (500 units/mL) and transferred to RPMI-1640 (Gibco) containing calcium-chloride (2 mM) and collagenase-IV (5 U/mL) (Worthington, Lakewood, NJ). Tissue suspension was incubated (37°C, 30 minutes), ground through a 100-μm strainer (BD Biosciences, Franklin Lakes, NJ), and eventually centrifuged (50g, 5 minutes). The supernatant was transferred onto a two-step density gradient (Optiprep; Sigma-Aldrich) and centrifuged (800g, 20 minutes). Cells from the gradient interface were transferred onto flat-bottom 96-well plates (5 × 105 cells/well) for 30 minutes at 37°C and nonadhesive cells were removed. Expression was analyzed in TAM seeded on 24-well plates (1 × 106 cells/well). Murine liver associated lymphocytes were isolated as described.15

Culture Treatments.

Sorafenib (sc-220125; Santa Cruz Biotech., Santa Cruz, CA) or dimethyl sulfoxide (DMSO) carrier (mock) was added to cell cultures (0.07-2.5 μg/mL [0.02-10 μM]) for 3 hours. Treatment was followed by a medium exchange and NK cell coculture for 24 hours. Lipopolysaccharide (LPS) (1 ng/mL), Celastrol, TPCA-1 (all Sigma-Aldrich), or Q-VD-OPh (R&D Systems, Minneapolis, MN) were applied as indicated. Blocking experiments were performed with IL12 (BD Biosciences), IL15 (R&D Systems) and IL18 (MBL, Woburn, MA) specific antibodies (5 μg/mL) for 24 hours or with IgG1 control antibodies (Abcam, Cambridge, UK).

Flow Cytometry.

Flow cytometry was performed with a Canto-II reader (BD Biosciences) and FlowJo software (Tree Star, Ashland, OR). Cells were permeabilized in Cytofix/Cytoperm reagent (BD Biosciences) and live-dead discriminated by EMA or propidium-iodide. Fluorochrome-conjugated antihuman antibodies: CD3-FITC, CD16-PerCP, CD69-FITC (all eBioscience, San Diego, CA), CD56-APC (R&D Systems), interferon-gamma (IFN-γ)-FITC (BD Biosciences), CD107a-PE-Cy5.5 (Biolegend, San Diego, CA), or antimurine antibodies: CD3-PacB, IFN-γ-PE, CD69-FITC, CD107a-APC, NK1.1-PerCp/Cy5.5, NK1.1-APC (all eBioscience) as well as appropriate isotype controls (BD Biosciences) were subsequently applied. NK cells designated for CD107a degranulation assays and IFN-γ stains were stimulated with K562 (effector to target ratio [E:T] = 1) or ionomycin/phorbol myristate acetate (PMA) (500/5 ng/mL) for 5 hours; Brefeldin-A (5 μg/mL) was added after 1 hour of stimulation (all Sigma-Aldrich).

Quantitative PCR.

Quantitative real-time polymerase chain reaction (RT-qPCR) was performed as described.15 Primers specific for CD14, CD68, albumin, α-fetoprotein (AFP), liver/lymph node-specific ICAM-3-grabbing non-integrin (L-SIGN), interleukin-6 (IL6), IL10, interleukin-12 p40 (IL12), IL18, tumor necrosis factor-α (TNF-α), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were used (Supporting Information). Messenger RNA (mRNA) expression was normalized to GAPDH.

Enzyme-Linked Immunosorbent Assay (ELISA).

Cytokines were quantified by ELISA for IL6, IL10, TNF-α (all BD Biosciences), IFN-γ, IL12 (p40/70) (all Biolegend), and IL18 (Bender Med Systems, Burlingame, CA).

Transwell and Migration Assay.

Mϕ (12-well plate, 1 × 105 cells/well) were sorafenib- (1.2 μg/mL) or mock-treated for 3 hours, followed by a medium exchange. Transwells (0.4 μm pores; Corning, Corning, NY) carrying 1 × 105 NK cells were subsequently placed on top of the cultured Mϕ for 24 hours, either with or without addition of LPS (1 ng/mL). Mϕ/NK cocultures served as control.

Migration assays were modified by 5 μm pore transwells (Corning) carrying 1 × 105 [51Cr]chromium (Cr)-labeled NK cells (see below). Transmigration was quantified by autoradiography within the destination compartment after 5 hours. NK cell migration in the presence of IL15 (10 ng/mL) (Peprotech) served as reference.

Killing Assay.

K562, Raji (2 × 106 cells), and HepG2 (5 × 105 cells/well) were Cr-labeled for 1.5 hours with 250 μCi/mL or 50 μCi/mL, respectively. NK cells were added for 5 hours at defined E:T ratios. Maximal and minimal lysis referred to Triton X-100-treated (0.1%) (Sigma-Aldrich) or nontreated targets, respectively. Culture supernatant (30 μL) was transferred to a γ-counter (TopCount; Packard, Meriden, CT) and specific cell lysis was calculated (lysis(%) = [(lysisx-lysismin)/(lysismax − lysismin)] × 100).

Western Blot.

Cells were lysated in buffer (Tris-HCL [10 mM], NaCl [100 mM], EDTA [5 mM], Triton X-100 [5%]) containing protease inhibitor (Roche), sodium-fluoride (50 mM), and sodium-o-vadanate (1 mM) (Sigma-Aldrich). Lysates were subjected to 10%-15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (Bio-Rad, München, Germany) and blotted on nitrocellulose membranes (Bio-Rad). Stains were performed with p100/p52, phospho-RelA, cleaved caspase-3, and β-actin (all Cell Signaling, Beverly, MA) specific antibodies. Staining was visualized with horseradish peroxidase (HRP)-conjugated antibodies (Cell Signaling) on film (Thermo Scientific, Waltham, MA).

Statistical Analysis.

Bars represent mean values with standard deviation and boxplots indicate median, quartiles, and range. P-values are based on Student's t test at a local significance level of 95%.


Sorafenib Activates Hepatic NK Cells in Mice.

First, C57Bl/6wt mice were screened for immune activation following administration of sorafenib. Hepatic NK cells (CD3/NK1.1+) from sorafenib-treated mice showed a higher CD69 expression compared to those from mock-treated mice (Fig. 1A). Splenic NK cells, in contrast, displayed a constitutively lower CD69 expression in comparison to hepatic NK cells (P < 0.0001) and did not respond to sorafenib. Serum transaminase activity was not significantly increased, excluding relevant sorafenib toxicity (Fig. 1A). Analysis of hepatic NK cells further showed increased cellular degranulation and IFN-γ secretion after sorafenib treatment (Fig. 1B,C). HBV-tg mice and one LTα/β-tg mouse with histologically confirmed HCC (Supporting Fig. S1A,B) were used to analyze activation of NK cells in a cancerogenic environment. Sorafenib triggered NK cell activation in HBV-tg mice (Fig. 1D), and in the HCC-bearing LTα/β-tg mouse, but not in younger LTα/β-tg mice without established HCC (Fig. S1C). Based on the latter experiments we concluded that sorafenib can activate hepatic NK cells under physiological and cancerogenic conditions.

Figure 1.

NK cell induction by sorafenib in mice. Wildtype (C57Bl/6wt) and HCC-bearing HBV-tg mice were analyzed 24-48 hours after a single application of sorafenib (25 mg/kg, intraperitoneal) or mock (DMSO) as indicated. (A) CD69 activation on splenic/hepatic NK cells and corresponding alanine aminotransferase (ALT) levels in treated C57Bl/6wt mice. (B,C) IFN-γ secretion and CD69/CD107a surface expression of hepatic NK cells following PMA/ionomycin stimulation ex vivo. Panels refer to pooled NK cells derived from C57Bl/6wt mice after previous treatment as indicated (n = 3). (D) CD69 activation or IFN-γ secretion of hepatic NK cells and corresponding ALT levels in treated HBV-tg mice (n = 4). IFN-γ secretion was assayed after ex vivo stimulation with PMA/ionomycin.

Sorafenib Triggers Cytokine Production in Macrophages.

Following prior mouse experiments, we recapitulated sorafenib-triggered immune activation in human polarized Mϕ cultures, which resemble characteristics of TAM.16 Mϕ cultures upon stimulation were monitored for the influence of sorafenib on inducible cytokine profiles. Compared to untreated controls, sorafenib (1.2 μg/mL) primed an induction of IL6 (7.5-fold), IL18 (3.5-fold), IL12 p40 (2.3-fold), and TNF-α (2.3-fold) transcription in cultured Mϕ after LPS stimulation. In contrast, a relevant IL10 induction (1.1-fold) was not observed. Corresponding cytokine secretion culminated in a 1.7-fold, 2.9-fold, and 3.2-fold increase of IL6, TNF-α, and IL12, respectively (Fig. 2). IL10 secretion was slightly reduced by sorafenib (Fig. 2), whereas IL18 was not detectable. Hence, we surmised that sorafenib triggers proinflammatory cytokines in polarized Mϕ.

Figure 2.

Sorafenib triggers proinflammatory cytokine secretion in macrophages. Mϕ cultures were treated with sorafenib (0.07-2.5 μg/mL). Medium was exchanged after 3 hours and LPS (1 ng/mL) was added. Cytokine secretion into culture supernatants was determined after 24 hours by ELISA (n = 3).

Sorafenib-Treated Macrophages Trigger NK Cell Activation.

Induction of cytokines by sorafenib prompted us to analyze NK cells in the presence of cultured Mϕ, as IL12 and also IL18 are NK cell activators.17 Therefore, Mϕ were cocultured with autologous NK cells of characteristic phenotype and morphology (Fig. 3A,B). Sorafenib triggered CD69 activation on CD56dim NK cells in a dose-dependent manner during coculture with LPS-stimulated Mϕ. In contrast, NK cells in the absence of Mϕ showed no CD69 activation upon sorafenib treatment (Fig. 3C). NK cell degranulation leads to IFNg release to orchestrate tumor-directed immunity.18 We were able to confirm both events in sorafenib-triggered NK cells during target cell contact (Fig. 3D). Moreover, Mϕ/NK cocultures secreted more IFN-γ into the culture supernatant upon treatment with sorafenib and/or LPS (Fig. 3E). Finally, NK cell mobility towards sorafenib pretreated Mϕ was increased (Fig. 3F), which confirmed the profound functional NK cell activation.

Figure 3.

Sorafenib triggers NK cell activation in the presence of macrophages. Flow cytometric phenotype (A) of NK cells and phase contrast microscopy (B) of cocultured Mϕ (black arrow) or NK cells (white arrow). Mϕ cultures were treated with sorafenib for 3 hours followed by addition of autologous NK cells and LPS (1 ng/mL) for 24 hours as indicated. Separate NK cell and Mϕ cultures served as control. NK cells were assayed by flow cytometry for CD69 activation in CD56dim/CD56high subpopulations (C) directly after coculture (n = 4-6) or for CD107a/IFN-γ expression (D) after K562 (E:T = 1.5) contact (n = 3). (E) IFN-γ secretion into Mϕ/NK supernatants as determined by ELISA (n = 3). (F) NK cell transmigration towards sorafenib (1.2 μg/mL) and/or LPS (1 ng/mL) treated Mϕ cultures. IL15-stimulated (10 ng/mL) NK cell migration served as reference (n = 4).

Sorafenib-Treated Macrophages Enhance NK Cell Killing.

NK cells were passaged from NK/Mϕ cocultures onto target cells to assess their killing capacity. Sorafenib was carefully removed before NK cell transfer to prevent sorafenib exposure of target cells. Mϕ coculture reduced NK cells killing of K562 targets compared to NK cells without previous Mϕ contact (8.0 ± 1.3% versus 19.7 ± 1.6%, P = .0015 [mean ± SD, n = 4]) (Figs. 4A, S2). Sorafenib pretreatment restored NK cell killing and enhanced K562 cell lysis in doses between 0.6 and 2.5 μg/mL. The latter experiment was repeated with MHC-I-positive Raji and HepG2 targets, which are resistant to resting NK cells. In this setting, sorafenib more than doubled NK cell killing during LPS stimulation (Fig. 4A). Finally, killing assays with increasing E:T ratios conclusively proved NK cell-dependent killing of different targets (Fig. 4B).

Figure 4.

Sorafenib enhances NK cell killing. Mϕ cultures were treated with sorafenib for 3 hours followed by addition of autologous NK cells and LPS (1 ng/mL) for 24 hours as indicated. Separate Mϕ cultures served as control. (A) Killing of K562 (E:T = 1.5), Raji (E:T = 1.5), and HepG2 (E:T = 0.2) targets by cocultured NK cells in response to sorafenib (n = 4-6). (B) Correlation of target killing and E:T ratios (K562, Raji) within the sorafenib (1.2 μg/mL; 3 hours) and/or LPS (1 ng/mL; 24 hours) treatment groups (n = 4). Killing was assayed by Cr-release.

Sorafenib-Triggered Cytokines Activate NK Cells.

Cytokine induction led us to propose a link between sorafenib-triggered cytokine secretion in Mϕ cultures and NK cell induction. This hypothesis was supported by a transwell experiment, showing sorafenib-triggered NK cell activation without direct Mϕ contact (Fig. 5A). Nevertheless, direct Mϕ/NK interaction provided a stronger NK cell activation, indicating additional involvement of Mϕ surface molecules. Eventually, blocking experiments confirmed IL12 or IL18 as sorafenib-triggered NK cell stimulus (Fig. 5B), whereas IL15 neutralization and isotype antibodies did not affect NK cell activity. IL12 and IL18 acted synergistically on NK cells, as reduction in killing efficacy was more pronounced if both cytokines were blocked simultaneously (IL12 versus IL12/IL18; K562: P = 0.0012; Raji: P = 0.0001) (Fig. 5B). In conclusion, NK cell activation was cytokine-dependent and was partially enhanced by direct contact between Mϕ and NK cells.

Figure 5.

Sorafenib-triggered NK cell activation is cytokine-mediated. Mϕ cultures were treated with sorafenib (1.2 μg/mL, 3 hours) followed by addition of autologous NK cells and LPS (1 ng/mL) for 24 hours as indicated. (A) Mϕ/NK cell contact was prevented by transwells, which allowed cytokine passage. NK cells were assayed for CD69 activation by flow cytometry (n = 5) and for IFN-γ secretion into culture supernatants by ELISA (n = 6). (B) Cytokine blocking antibodies (5 μg/mL) were added to Mϕ/NK cocultures. Isotype antibodies served as control. NK cells were assayed for CD69 activation by flow cytometry (n = 4) and killing activity (K562, Raji targets; E:T = 1.5) by Cr-release (n = 4). IFN-γ secretion into culture supernatants was quantified by ELISA (n = 4).

Sorafenib Triggers NF-κB Activation and Sensitizes Macrophages for Apoptotic Cells.

NF-κB regulates Mϕ activation and promotes cytokine expression. We therefore analyzed sorafenib-triggered NF-κB activation in Mϕ cultures (Fig. 6A). Sorafenib activated the canonical and noncanonical NF-κB pathway in polarized Mϕ cultures in a dose- and LPS-dependent fashion, as shown by p100/p52 processing and RelA phosphorylation (Fig. 6A). Celastrol, an inhibitor of both NF-κB pathways, and TPCA-1, specifically subverting the canonical NF-κB pathway (Fig. 6A), were employed for NF-κB blocking experiments. Both compounds coadministered with sorafenib reduced NK cell killing (Fig. 6B) as well as NK cell degranulation (Fig. 6C).

Figure 6.

Sorafenib-triggered NK cell activation involves NF-κB activation and is promoted by apoptotic cells. Mϕ cultures were treated with sorafenib for 3 hours followed by LPS (1 ng/mL; 3 hours) as indicated. (A) NF-κB pathway activation in Mϕ cultures was confirmed by NF-κB2 (p100/p52) processing and RelA phosphorylation using western blot (10% SDS-PAGE). Celastrol (2 μM) and TPCA-1 (5 μM) abolished NF-κB induction, when added to Mϕ cultures 3 hours prior to sorafenib treatment. Actin served as loading control. NK cells were analyzed after Mϕ/NK coculture following sorafenib (1.2 μg/mL) and/or LPS (1 ng/mL) treatment. Effect of Celastrol (2 μM) and TPCA-1 (5 μM), added 3 hours before and during sorafenib treatment, on NK cell (B) killing (n = 6) and (C) degranulation (CD107a) (n = 4) was detected by Cr-release (K562; E:T = 1.5) and flow cytometry, respectively. (D-F) Following sorafenib treatment (1.2 μg/mL; 3 hours) UV-treated (40 mJ/m2) or nontreated HepG2 cells (5 × 105 cells/well) were added to Mϕ/NK cocultures for 24 hours. (D) NK cell degranulation (CD107a) and (E) IFN-γ secretion were analyzed by flow cytometry (n = 3). (F) Western blot (10% SDS-PAGE) of cleaved caspase-3 and actin loading control in nontreated or in UV-treated (40 mJ/m2) HepG2 after 24 hours.

We next investigated if sorafenib sensitizes polarized Mϕ to apoptotic cells, as this reflects the constellation during cytotoxic HCC treatment in vivo. In fact, sorafenib-treated Mϕ provided a stronger stimulus on NK cells in the presence of ultraviolet (UV)-irradiated apoptotic HepG2 cells. Control experiments showed that this was not the case after addition of untreated HepG2 cells and that caspase-3 cleavage distinguished UV-irradiated from untreated HepG2 (Fig. 6D-F). On the other hand, sorafenib did not induce apoptosis in Mϕ (Fig. S3A) and NK cell activation was not abolished by a caspase inhibitor during Mϕ/NK coculture experiments (Fig. S3B,C), indicating that apoptotic Mϕ did not contribute substantially to NK cell activation in our model.

Sorafenib-Activated NK Cells Through Macrophages Isolated From HCC.

Complex TAM polarization is not completely resembled by in vitro models. We therefore isolated macrophages from freshly resected HCC tissue. Primary human TAM displayed a bipolar morphology in contrast to spherical monocytes derived from peripheral blood (Fig. 7A). CD68 and CD163 mRNA expression confirmed TAM identity, whereas AFP, albumin, and L-SIGN transcripts indicating tumor cells, hepatocytes, and endothelial cells were barely detectable (Fig. 7A).

Figure 7.

NK cells are activated by sorafenib-treated HCC associated macrophages. (A) Morphology of TAM isolated from HCC in comparison to monocytes. Cell marker expression in TAM compared to HCC as determined by RT-qPCR (technical triplicates). (B) Cytokine expression determined by RT-qPCR after sorafenib treatment (0.6-1.2 μg/mL; 3 hours) and LPS stimulation (1 ng/mL; 24 hours) of TAM isolated from different HCC (n = 3-4). TAM from one donor were sorafenib treated (1.2 μg/mL; 3 hours) followed by addition of autologous bead-enriched NK cells and LPS (1 ng/mL) for 24 hours as indicated. Separate NK and TAM cultures served as control. NK cell IFN-γ response (C) against K562 (E:T = 1.5) was assayed by flow cytometry (n = 3-4). (D) NK cell degranulation (CD107a) and (E) target killing was determined by flow cytometry or Cr-release (K562, E:T = 1.5), respectively (n = 4).

Sorafenib treatment triggered a stronger IL12 and IL18 mRNA expression in isolated TAM under LPS stimulation compared to untreated controls (Fig. 7B). Homologous TAM/NK cocultures derived from the same donor were used to confirm an interaction between both cell types. Upon coculture with sorafenib-treated TAM, NK cells showed increased IFN-γ expression, degranulation, and killing capacity (Fig. 7C-E). Additional experiments with heterologous TAM/NK cocultures derived from different donors indicated comparable sorafenib-triggered NK cell activation (Fig. S4A-C). In these experiments, resting TAM as well as LPS-stimulated TAM reduced the cytolytic NK cell function by 35% (P = 0.0098) and 27% (P = 0.0074), respectively.

Taken together, we confirmed that freshly isolated HCC-associated macrophages display immune inhibitory functions ex vivo, which in response to sorafenib were reversed, eventually leading to NK cell activation.


Our study shows that sorafenib sensitized TAM for exogenous immune stimuli, which eventually accelerated cytotoxicity of cocultured NK cells against tumor cells. This sorafenib effect was based on an NF-κB-dependent switch from inhibitory towards immune stimulatory TAM responses. We therefore propose that sorafenib—besides its direct effect on tumor cells—has an immune stimulatory function.

We observed that Mϕ or TAM are pivotal for sorafenib-triggered NK cell stimulation. Blocking experiments confirmed cytokine signaling between both cell types, albeit cell contact remained relevant for NK cell activation by sorafenib. Nonsecreted cytokines exposed on the Mϕ surface might explain this partially contact-dependent NK cell activation.19, 20

Cytokine expression in Mϕ and TAM is to a large part NF-κB-dependent.21 This also applied to sorafenib-stimulated Mϕ cultures, in which sorafenib triggered NF-κB activation despite its known NF-κB inhibitory properties10, 22 under certain conditions. This may include the polarization of TAM, as inhibition of NF-κB in TAM has been reported to induce antitumoral immune responses.23 Sorafenib targets, such as the CSF-receptor-1,24 MAPKp38,12 and JAK/STAT,25 which are pivotal for Mϕ maturation, support this finding.26-28 Sorafenib-mediated cytokine induction in polarized Mϕ might therefore rely on complex signaling networks. In line with this hypothesis, cytokine induction peaked at a certain sorafenib dose range (0.6-1.2 μg/mL), indicating a shift between immune-activating versus blocking effects. The activating dose range fits with sorafenib concentrations in the extravascular compartment29 and therefore seems appropriate to study intratumoral effects. NF-κB inhibition by sorafenib as observed in other studies was achieved using higher concentrations10, 22 resembling sorafenib plasma levels.30

Sorafenib-triggered NK cell activation was dependent on LPS as an exogenous Mϕ stimulating factor. LPS is abundant in the portal system31 and contributes to hepatocarcinogenesis.32In vivo relevance of LPS for sorafenib-triggered NK cell activation was demonstrated by a selected activation of NK cells in the murine liver, in contrast to unresponsive splenic NK cells, which receive no LPS stimulation. The apoptotic cell encounter, which maintains an antiinflammatory tumor environment by way of STAT-3 activation in TAM,33 also led to NK cell activation after sorafenib treatment in cell culture. Sorafenib may therefore interrupt the immunosuppressive feedback loop driven by apoptosis in HCC.

Sorafenib eventually induced essential tumor-directed NK cell killing. Given that sorafenib increases tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)-mediated apoptosis by MCL-1 suppression34 one may speculate that sorafenib could also sensitize tumor cells for NK cell activity.18 Finally, sorafenib triggered IFN-γ secretion of NK cells,35 which prevents Mϕ polarization by mitigating CSF-136 or IL437 signal transduction. IFN-γ secretion may therefore enhance pattern recognition by Mϕ38 or could ameliorate their antiinflammatory IL1 receptor antagonist and IL10 expression.39, 40

Taken together, sorafenib primes proinflammatory responses of macrophages located within the HCC microenvironment and perpetuates cytotoxic NK cell activity. This provides an additional mechanism of how tyrosine kinase inhibitors could elicit anticancer effects and may provide new insights for immune stimulatory treatments.


We thank Ruth Hillermann, Lynette Henkel, and Daniel Kull for excellent technical assistance, Melissa Schlitter and Norbert Hüser for tissue preparation. We are grateful to Frank Chisari for providing mouse strain HBV1.3.32.