The small molecule curcumin analog FLLL32 induces apoptosis in melanoma cells via STAT3 inhibition and retains the cellular response to cytokines with anti-tumor activity

A375, HT144 and Hs294T human melanoma, and the K562 leukemia cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, VA) and 1106 MEL, 1259 MEL, MEL-39 and F01 human melanoma cell lines were provided by Dr. Soldano Ferrone (University of Pittsburgh, Pittsburgh, PA) and cultured as described [17]. Melanoma cell lines were authenticated via karyotype analysis in the Molecular Cytogenetics Core Laboratory of The Ohio State University. The radial growth phase WM 1552c and vertical growth phase WM 793b human melanoma cell lines were provided by Dr. M. Herlyn (Wistar Institute, Philadelphia, PA) and cultured as described [18]. Primary cultures from patients with recurrent cutaneous melanomas were cultured as previously described [17]. Tetramethylrhodamine ethyl ester perchlorate (TMRE) was purchased from Invitrogen (Carlsbad, CA). The pan-caspase inhibitor (Z VAD-FMK), control (Z-FA-FMK) and recombinant human IFN-γ were purchased from R & D Systems, Inc. (Minneapolis, MN). Recombinant human interleukin-6 (IL 6) was purchased from Peprotech, Inc. (Rocky Hill, NJ). Recombinant human IL-2 (specific activity = 10⁷ U/mg) was purchased from Hoffmann-La Roche Pharmaceuticals (Nutley, NJ). The JSI-124 and Stattic inhibitors were purchased from Calbiochem (Gibbstown, NJ). WP1066 was synthesized in the laboratory of Dr. P-K Li [19]. FLLL32 and curcumin were synthesized, purified and evaluated for purity as previously described [16, 20, 21].

Peripheral blood mononuclear cells (PBMCs) were isolated from source leukocytes of healthy donors (American Red Cross, Columbus, OH) via density gradient centrifugation using Ficoll-Paque (Amersham, Pharmacia Biotech, Bjorkgatan, Sweeden) as described [22]. NK cells were enriched from source leukocytes by negative selection with Rosette Sep reagents (Stem Cell Technologies, Inc., Vancouver, British Columbia, Canada).

Lysates were prepared from melanoma cell lines or PBMCs and assayed for protein expression by immunoblot analysis as previously described with antibodies (Ab) to STAT1 (BD Biosciences), Survivin (Novus Biologicals, Littleton, CO), pSTAT1, STAT3, pSTAT3, pSTAT5, STAT5, pJAK2, JAK2, PARP, Cyclin D1, Caspase-3, Caspase-8, Caspase-9, phosphorylated and total Akt (pAkt), Src (pSrc), p38 MAPK (p-p38 MAPK), ERK (pERK) (Cell Signaling Technology, Danvers, MA), or β-actin (Sigma) [23]. Following incubation with the appropriate horseradish-peroxidase-conjugated secondary Ab, immune complexes were detected using the SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific).

Phosphatidyl serine exposure was assessed in tumor cells by flow cytometry using APC-Annexin V and propidium iodide (PI; BD Pharmingen, San Diego, CA) as described [23]. Analyses were performed utilizing at least 10,000 events.

STAT3 DNA binding was measured with the Pierce LightShift Chemiluminescent EMSA kit used according to manufacturer's instructions (Thermo Fisher Scientific Inc, Rockford, IL). Nuclear protein was collected using the NucBuster™ Protein Extraction kit (EMD Chemicals Inc, Gibbstown, NJ). Binding reactions using equal amounts of nuclear protein were incubated for 20 minutes at room temperature with DNA probes. A biotinylated STAT3 binding sequence in the human survivin promoter (sense 5'-GAGACTCAGTTTCAAATAAATAAATAAAC-3') was purchased from Operon Biotechnologies (Huntsville, AL). Reactions with biotinylated target DNA only and nuclear protein with biotinylated target DNA and excess unlabelled target DNA to compete for binding were included. STAT3 specificity was confirmed by incubation with 6μg of anti-STAT3 Ab (Santa Cruz Biotechnology Inc, Santa Cruz, CA) to interfere with the protein-DNA complex. Following electrophoresis, DNA was transferred to a nylon membrane, cross-linked and detected by chemiluminescence.

The mitochondrial membrane potential (ΔΨ_m) was assayed using 150 nM TMRE in regular medium at 37ºC for 15 minutes and by subsequent flow cytometric analysis as described [24].

Real-time PCR was used to evaluate the expression of the IFN-γ stimulated gene (IRF1) as described [25, 26] with pre-designed primer/probe sets (Assays On Demand; Applied Biosystems, Foster City, CA) and 2X TaqMan Universal PCR Master Mix (Applied Biosystems) per manufacturer's recommendations. Primer/probe sets for 18s rRNA (Applied Biosystems) were used to normalize expression values (housekeeping gene). Data were acquired and analyzed using the ABI Prism 7900HT Sequence Detection System (Applied Biosystems).

To measure granzyme B (GrB) and IFN-γ secretion, ELISPOT experiments were conducted using MultiScreen 96-well plates (Millipore, Bedford, MA) and biotinylated monoclonal anti-human GrB or IFN-γ detecting Ab (Mabtech) as described [27]. Freshly isolated NK cells (effectors) were incubated overnight in IL-2-containing media (1 nM) with either 5μM FLLL32 or DMSO. Effector cells were then co-incubated in triplicate with K562 cells as targets at an effector:target ratio of 10:1 for four hours. Targets and effectors cultured alone were used as controls. Spots were visualized and counted using the ImmunoSpot Imaging Analyzer (Cellular Technology Ltd, Cleveland, OH).

The 4-parameter logistic or Hill model [28] was the assumed dose-response relationship for FLLL32 concentration and proportion of apoptotic cells. Nonlinear least squares regression was used to estimate the parameters. ELISPOT data were compared between groups using a two-sample t-test. All analyses were performed in Statistical Analysis System (version 9.2; SAS Institute). P-values were considered significant at the 0.05 level and all tests were two-sided.

The pro-apoptotic effects of FLLL32 (Figure 1A) were examined by flow cytometry following Annexin V/PI staining of a panel of metastatic human melanoma cell lines with basal STAT3 phosphorylation (A375, Hs294T, FO1, HT144, MEL-39) and the pSTAT3 negative 1106 MEL and 1259 MEL cell lines [17]. Dose-response studies revealed consistent induction of apoptosis in pSTAT3-positive metastatic human melanoma cell lines following a 48 hour treatment with FLLL32 as compared to DMSO (vehicle) treated cells (Figure 1B). The pSTAT3-positive A375 cell line was particularly sensitive to the pro-apoptotic effects of FLLL32 (IC₅₀ = 1.3μM at 48 hours). Similar data were obtained in multiple pSTAT3 positive human melanoma cell lines (IC₅₀ range = 1.9 -- 2.8 at 48 hours). The pSTAT3 negative 1106 MEL and 1259 MEL cell lines were poorly sensitive to FLLL32 (Figure 1B). FLLL32 was more potent than curcumin at inducing apoptosis (Figure 1C). Consistent with prior studies from our group, a 10-fold greater concentration of curcumin was required to achieve the same degree of apoptosis at the 48 hour time point [15]. FLLL32-induced apoptosis was also confirmed in pSTAT3⁺ human melanoma cell lines derived from other disease phenotypes, including the WM 1552c radial growth phase and WM 793b vertical growth phase lines following treatment with FLLL32 (data not shown).

FLLL32 inhibited STAT3 phosphorylation at Tyr⁷⁰⁵ but not at Tyr⁷²⁷ in multiple human melanoma cell lines after a 24 hour treatment (Figure 1D). Prior studies indicated FLLL32 could inhibit Jak2 kinase activity in an in vitro cell-free assay [16]. However, we did not observe an appreciable alteration in Jak2 phosphorylation even at a concentration of 8 μM, suggesting that this compound likely acted directly against the STAT3 protein (Figure 1D). Time course studies also revealed that fulminant cell death occurred after 24 hours of continuous culture, yet exposure to FLLL32 at 2 - 4 μM for only 4 hours was sufficient to reduce pSTAT3 and induce cell death (Additional File 1: Figure S1A-B). FLLL32 did not appear to inhibit the phosphorylation of other key signaling pathways that are constitutively active in malignant cells (e.g. Src, Akt) at doses capable of inhibiting STAT3 phosphorylation after 24 hours. Consistent with reciprocal activation of the p38 MAPK and STAT3 pathways [29], FLLL32 treatment led to increased levels of total p38 MAPK protein once pSTAT3 decreased (Figure 1D). Importantly, FLLL32 was capable of reducing pSTAT3 levels, cyclin D1 expression and inducing apoptosis in primary human melanoma cell cultures derived from recurrent cutaneous melanoma tumors [17] (Figure 1E). Finally, treatment of basal pSTAT3-positive human melanoma cell lines with FLLL32 for 24 hours led to reduced STAT3 DNA binding as determined by gel shift assays and expression of the STAT3-regulated genes, cyclin D1 and survivin as measured by immunoblot (Figure 2A-B).

The mechanism by which FLLL32 induces apoptosis was subsequently investigated in the A375 melanoma cell line. Immunoblot analysis demonstrated a concentration-dependent increase in the processing of both initiator (caspase-8 and caspase-9) and effector caspases (caspase-3) following a 24 hour treatment with FLLL32 (Figure 3A). Treatment of with FLLL32 also resulted in a concentration-dependent loss of mitochondrial membrane potential as measured by flow cytometry (Figure 3B). These data support the involvement of the mitochondrial amplification loop in promoting cell death in response to this treatment. Apoptosis was caspase-dependent, as culture with a pan-caspase inhibitor (Z-VAD-FMK) inhibited melanoma cell death as compared to culture with the Z-FA-FMK control compound (Figure 3C and Additional File 1: Figure S2). These data were confirmed at the 48 hour time point by flow cytometry following annexin V/PI staining, and by reduced PARP cleavage by immunoblot analysis (Figure 3C-D). Interestingly, reduced levels of pSTAT3 and cyclin D1 occurred following treatment of A375 cells with FLLL32 in the presence of the pan-caspase inhibitor (Figure 3D). These data are consistent with a mechanism that places reduced pSTAT3 and its cellular targets upstream of the caspase cascade and subsequent apoptosis.

Since many cytokines act via homologous STAT proteins (e.g. STAT1), it was imperative to test whether FLLL32 had deleterious effects on the action of cytokines that might promote an anti-tumor response. Of concern were the effects of FLLL32 on signal transduction in response to IFN γ, a cytokine that mediates its cellular effects via phosphorylation of STAT1, and a resulting STAT1-STAT1 homodimer [25]. To test these interactions in a biologic system, we investigated the effects of FLLL32 or curcumin pre-treatment on IFN-γ-induced signaling and gene expression. Pre treatment of pSTAT3 positive A375 and Hs294T cells with FLLL32 or curcumin led to reduced pSTAT3 versus DMSO-treated cells. However, in contrast to curcumin, FLLL32 did not adversely affect IFN-γ-induced pSTAT1 (Figure 4A and Additional File 1: Figure S3). A unique advantage of FLLL32 versus other STAT3 pathway inhibitors was its apparent specificity. Despite a similar degree of cytotoxicity and the ability to reduce basal pSTAT3 in human melanoma cells (Additional File 1: Figure S4), the WP1066, JSI-124, and Stattic compounds also inhibited IFN-γ-induced STAT1 phosphorylation (Figure 4B). Pre-treatment with FLLL32 also enhanced transcription of the pro-apoptotic interferon-regulatory factor-1 (IRF1) gene in response to IFN-γ stimulation as determined by Real Time PCR (Figure 4C). This IFN-γ responsive gene has been shown to be transcribed via STAT1-STAT1 homodimers binding to a gamma-activated sequence (GAS) element [25]. Consistent with our prior studies [15], IFN-γ stimulated IRF1 transcription was reduced in all cells pre-treated with curcumin (Figure 4C). The induction of IRF1 was not enhanced in the pSTAT3 negative 1106 MEL cell line (Figure 4C), suggesting that cross-reactivity of FLLL32 with STAT1 was negligible, and that IFN-γ driven gene transcription can be augmented via STAT3 inhibition. These data indicated that IFN-γ-induced signal transduction and gene expression were not reduced by FLLL32 and that its inhibitory actions were specific for STAT3 and not other homologous STAT proteins that function as tumor suppressors (e.g. STAT1) [30‐33].

STAT3 function in immune cells can promote tolerance to developing or established tumors. We therefore evaluated whether FLLL32 would affect the responsiveness of PBMCs to stimulation with clinically relevant cytokines that mediate tumor progression (IL 6), immunosurveillance (IFN-γ) or T and NK cell survival (IL 2) [34‐36]. Pre treatment with increasing doses of FLLL32 reduced basal pSTAT3 in PBMCs from healthy donors and led to reduced IL-6-induced pSTAT3 in PBMCs (Figure 5A). FLLL32 pre-treatment also did not adversely affect the level of IFN-γ-induced pSTAT1 or IRF1 gene expression in PBMC (Figure 5B-C). The level of IL 2-induced pSTAT5 also was not altered by FLLL32 pre-treatment (Figure 5D). The FLLL32 compound did not decrease viability of PBMCs after a 24 hour treatment as compared to treatment with DMSO alone as determined by Annexin V/PI staining or PARP cleavage (Figure 6A-B). Similarly, NK cell viability from healthy donors cultured with IL-2 (1 nM) was not reduced following a 24 hour treatment with FLLL32 as compared to treatment with DMSO (Figure 6C). In addition, the production of granzyme b and IFN-γ by NK cells from normal donors when cultured with the K562 target cell line was not adversely affected in the presence of FLLL32 (Figure 6D). The mean difference (FLLL32 - DMSO) for granzyme b was 41.0 spots/well (95% CI: -79.0 to 161.0) and 65 spots/well for IFN-γ (95% CI: -146.0 to 277.9).