Tyrosine kinase inhibitor SU11274 increased tumorigenicity and enriched for melanoma-initiating cells by bioenergetic modulation

Small molecule inhibitor SU11274 was initially developed to specifically inhibit c-Met receptor signaling [1]. Receptor tyrosine kinase c-Met is a receptor for hepatocyte growth factor/scatter factor (HGF), pleiotropic cytokine controlling pro-migratory, anti-apoptotic and mitogenic signals [2]. c-Met activation evokes biological responses, globally referred to as ‘invasive growth’, thus being potential therapeutic target in metastatic cancer [3]. Promise of anti-c-Met drugs is based on their activity on multiple stages of cancer development, from initiation through progression to metastasis [4]. Moreover, the inappropriate c-Met signaling occurs in virtually all types of solid tumors [5]. In our previous study we have confirmed high expression of the c-Met receptor in a model cell line EGFP-A375/Rel3 derived as hypermetastatic and highly tumorigenic variant of human melanoma cell line A375 [6]. Our experiments in the above mentioned study have shown antiproliferative effect of SU11274 in vitro, but tumor supporting effect in vivo (see Fig. 6), when used as an augmentation to support antitumor effect of gene therapy-based approach [6]. It has been reported that the intraperitoneal administration of SU11274 achieved significant inhibitory effect on liver metastasis induced by the intrasplenic injection of human metastatic melanoma cells the HT168-M1 [7]. Moreover, the intratumor injection of SU11274 had high efficacy in vivo and this treatment reduced tumor volume by 7-fold as compared with control tumors induced by RU-P melanoma cells [8]. Based on the findings we wanted to explore the effect of SU11274 on various melanoma cell lines including our model of hypermetastatic variant, which was not examined so far.

Bulk tumor comprises subpopulations of non-tumorigenic and tumorigenic cells, which can reversibly transit among their states (tumorigenic stem cells vs. non-tumorigenic cells) [9]. Tumor-initiating cells (tumorigenic or so-called cancer stem cells) give rise to tumors in transplantation assays in vivo and they were associated with specific surface markers in human melanoma [10, 11]. Experiments suggested significant level of plasticity in melanoma cells and many markers were reversibly turned on and off in a manner that did not correlate with the ability to form a tumor [12‐14]. Growth of tumor cells in three-dimensional multicellular tumor spheroid cultures enables to maintain their tumorigenic potential [15‐17]; and therefore we suggested to use it to explore the effect of SU11274 treatment on tumor initiating potential of melanoma cells. It was previously shown that decreased tumor sphere formation by the inhibition of c-Met correlated with preventing metastatic disease and inhibiting stem cell function in pancreatic carcinoma [18].

There is not much known about the metabolic regulation of cancer stem cell function, but bioenergetic modulation was shown to counteract stem cell features and sensitized cells to kinase inhibitors [19‐22]. Bioenergetic modulators could be actually used in the antitumor treatment [21, 23, 24]; but recent evidence suggested specific metabolic behavior of melanoma cells [25, 26]. These reports prompted us to test potential of bioenergetic modulation to interfere with the tumor initiation in melanoma.

We evaluated the effect of SU11274 inhibitor in both adherent and spheroid melanoma cultures in vitro; and the effect on tumor growth and initiation in vivo. Our data show that SU11274-treated cells were enriched for melanoma-initiating cells; they had significantly increased tumorigenic potential. This effect could be counteracted by bioenergetic modulation with a glycolytic inhibitor dichloroacetate (DCA).

Recently we described novel hypermetastatic human melanoma cell line EGFP-A375/Rel3 (designated Rel3 in the following text) [6]. It was derived from parental EGFP-A375 [28] by three rounds of consecutive in vivo passaging as the third relapse which could regrow after the experimental therapy with prodrug-converting mesenchymal stromal cells. Rel3 cell line is highly tumorigenic and produces massive lung colonization upon intravenous injection indicative of its aggressiveness. We decided to examine the antitumor potential of SU11274 in malignant melanoma cell lines M14, M4Beu [27], A375 [28] and Rel3 cells (derived hypermetastatic variant of A375) [6]. We confirmed high level of the c-Met receptor on cell surface by flow cytometry in these cells. The c-Met expression was detected on the 46 % of M14 cells, 97.7 % of M4Beu cells, 98.0 % of A375 cells and 95.2 % of Rel3 cells (Fig. 1a). SU11274 can inhibit HGF-stimulated phosphorylation of c-Met on Tyr1234/1235 [31]. C-Met is not phosphorylated on these Tyr residues in A375 [32]. However, our data have shown that SU11274 increased phosphorylation of c-Met on Tyr1349, which was phosphorylated in both A375 and Rel3 cells (Fig. 1b). In addition to SU11274 as a selective c-Met inhibitor with the IC₅₀ of 10 nM in a cell-free assay, Crizotinib (PF-02341066) as another ATP-competitive small-molecule inhibitor of the catalytic activity of c-Met the IC₅₀ of 11 nM and in cell-free assay was used [33, 34]. Examined cell lines exhibited very similar IC₅₀ for these two different c-Met inhibitors SU11274 ranging 4–5 μM (Fig. 1c) and crizotinib ranging 1.25–3 μM (Fig. 1d) in standard adherent cultures. SU11274 treatment caused alteration in cellular morphology from narrow spindle shape to flatter rounded morphology and less scattered colonies (Fig. 2a). SU11274-treated cells were subsequently injected as xenografts and they were more tumorigenic in vivo (Fig. 2b). When we injected 5 × 10⁵ cells, the median tumor volume was significantly higher in the SU11274-pretreated group - 721.2 mm³ versus 395.8 mm³ in control group by day 16). When 2 × 10⁵ cells were injected, tumor take rate was also higher in the SU11274-pretreated group with 4 out of 4 injected tumors growing in contrast to the untreated group, where only 2 out of 4 inoculates grew with a longer dormancy period. Tumor growth kinetic is shown in (Fig. 2b).

In order to examine the effect of SU11274 on melanoma-initiating cells we switched cells to non-adherent melanosphere cultures [35, 36]. M4Beu cells were not able to proliferate under these conditions and did not form melanospheres (Fig. 3a). M14, A375 and Rel3 cells propagated and expanded in spheroid conditions (at least for more than 10 consecutive passages). Direct comparison of adherent and non-adherent cultures unraveled increased sensitivity to SU11274 in melanospheres (Fig. 3b–d). SU11274 also significantly inhibited cell proliferation. There were 6.5 × 10⁵ Rel3 cells in control versus 4.2 × 10⁵ Rel3 cells after SU11274 treatment, which is a 35 % inhibition of the proliferation. (Fig. 2e). The effect of SU11274 on tumor initiation frequencies was evaluated by extreme limiting dilution assay (ELDA) [30]. We injected gradually decreasing number of the cells after adherent and melanosphere culture and determined a proportion of growing tumors. For adherent parental A375 cells, frequency of tumor initiating cells was one in 8.5 × 10⁵. Frequency of tumor initiating cells in Rel3 was one in 2.4 × 10⁵, which was 3.5 higher corresponding to increased tumorigenicity. More importantly, melanosphere cultures further increased the frequency of tumor initiating cells to 1 out of 3.3 × 10⁴ spheroid Rel3 cells (7-fold increase in comparison to adherent culture). Tumor-initiating cell frequencies in SU11274-treated cells was determined to be 1 out of 3289 cells (p value 1 × 10⁻⁵), which was a 10-fold enrichment for tumor-initiating cells by SU11274. Same effect was achieved in M14 cells, where stem cell frequencies determined in vivo were 1 out of 3.8 × 10⁴ spheroid M14 cells in contrast to 1 out of 1.0 × 10³ SU11274-treated spheroid M14 cells. This represents a 4-fold enrichment in tumor initiating cell frequency (Table 1).

Next, we evaluated a long-term serial propagation of cells in the non-adherent conditions with or without SU11274. Rel3 cells could be long-term propagated, although the cumulative cell numbers differed significantly due to the antiproliferative action of the inhibitor (Fig. 4a). Cells from melanospheres were viable; they adhered and proliferated after switching to adherent conditions. Cell morphology after spheroid culture remained similar to morphology of adherent cultures in the presence or absence of SU11274 shifted from irregular spiked shape to flatter cobblestone morphology (Fig. 4b). Obvious discrepancy between minor decrease in the viability and severe decrease in the cell numbers mediated by SU11274 was further examined by BrdU incorporation assay. DNA synthesis and cell cycle progression was substantially more inhibited in comparison to the decrease of ATP level measured by relative viability assay (Fig. 4c). Relative ATP-content per 100,000 cells was significantly higher in cells propagated in SU11274 (Fig. 4d). Further analysis confirmed no significant difference in the glucose uptake, but higher lactate release from the SU11274-treated cells, indicative of their higher dependence on (or a metabolic switch to) aerobic glycolysis (Fig. 4e and f). No effect on ATP levels/cells and tumorigenicity was be observed with crizotinib (data not shown).

Next, we examined alterations induced by SU11274 on pluripotent stem cell proteins and phosphokinase proteome profile. Melanoma cells express many proteins associated with pluripotency, but SU11274-treated spheroids have increased levels in comparison to the untreated ones (Fig. 5a). Higher level of these transcription factors correlates with increased capability of the treated cells to induce tumor growth. Phosphokinase proteome array demonstrated that SU11274 activated p53 (Fig. 5b, lower panel b), which correlates with inhibition of the cell proliferation shown in Fig. 4a and c. RSK1/2/3 phosphorylation was increased after SU11274 exposure. We detected phosphorylation of following target kinases in Rel3 cells: ERK1/2, p-RAS40, Akt 1/2/3, p38 alpha, AMPK alpha1, CREB, GSK-3 alpha/beta, WNK-1 and HSP60 in both treated and untreated cells.

Based on the previous data suggesting involvement of bioenergetic modulation in SU11274-treated Rel3 cells, we decided to test the ability of bioenergetic modulators to suppress increased tumorigenicity of SU11274-treated cells. Spheroid non-adherent cells were hypersensitive to bioenergetic modulators dichloroacetate (DCA, inhibitor of pyruvate dehydrogenase kinase [23]) and 3-bromopyruvate (3BrPA, a hexokinase and GAPDH inhibitor [24]) in comparison to the adherent cells (Fig. 6a and b). Switch to melanosphere cultures leads to higher cellular dependence on the aerobic glycolysis, which correlates with increased tumor initiating properties of melanosphere cells. Contrastingly, chemotherapeutic alkylating agent dacarbazine, which was clinically approved for the treatment of malignant melanoma [37], was significantly less toxic to spheroid cells in comparison to adherent cells. This demonstrated inherent drug resistance in cultures enriched for melanoma-initiating cells (Fig. 6c). Finally, tumor take rate of the cells treated with SU11274 combined with selected compounds was examined by previously published approach [38]. We expected that in vitro pretreatment targeting tumor-initiating cells would eradicate these from culture and lessen the tumorigenicity. Therefore, SU11274-stimulated cells were co-treated with 3BrPA, DCA or dacarbazine at the IC₅₀ for 24 h. Dacarbazine was previously shown to spare the tumor initiating cells; thus, it was used in this assay as a control. Cells were cultured for next 48 h to allow the cell death to occur and viable cells to recover. Combination treatment with DCA did not significantly change SU1174-mediated increase of the ATP content per cell. Compounds 3BrPA or dacarbazine further increased relative ATP level (Fig. 6d). However, DCA decreased tumorigenicity of SU11274-treated cells in vivo, which was not the case for SU11274 combination with 3BrPA or dacarbazine (Fig. 6f). Median tumor volume in the SU11274-treated group was 115.5 mm³ in contrast to 14.5 mm³ in the DCA pulsed SU11274-treated cells. Moreover, three out of the eight animals did not develop any tumor in contrast to the eight out of the eight in the SU11274 group. DCA treatment alone did not significantly change tumor take rates or median tumor volumes. Our data show that antimelanoma chemotherapeutic drug dacarbazine similarly to bioenergetics modulator 3BrPA does not affect tumor initiating cell subpopulation in Rel3 cells. Combination of the DCA with SU11274 also did not completely eradicate tumor initiation capabilities. We attribute this outcome to the fact that melanoma cells derived from A375 cells harbor mutated B-Raf (V600E), thus tumor initiation capabilities without targeted inhibition of hyperactivated oncogenic pathway were retained. SU11274 compound does not interfere with this signaling axis.

In search for an agent to prevent both growth and metastatic dissemination of melanoma cells we hypothesized that the c-Met receptor was a suitable target [3]. However, experimental evidence suggested that c-Met receptor plays a dual role in oncogenesis. (i) In the mutated, amplified or otherwise genetically altered form, c-Met generates and maintains transformed phenotype, and drives clonal evolution; (ii) in the wild-type form, c-Met contributes to maintain - in the cancer stem cell - the phenotype ‘inherent’ in the stem/progenitor cell of origin [4]. Extensive redundancy of the receptor-tyrosine kinase signaling in cancer cells and receptor cross-talk suggested that there might be inherent or acquired resistance mediated by other signaling cascades compensating for inhibitory effect of the particular small-molecule inhibitor [39]. We detected high surface expression of c-Met receptor in tested melanoma cell lines (Fig. 1a). These cells did not produce detectable HGF into cell culture medium thus excluding c-Met autocrine stimulation (data not shown). We focused predominantly on potential role of the c-Met inhibitor SU11274 in highly metastatic aggressive variant Rel3 [6], as it was suggested as efficient atimelanoma agent [7, 8]. Antiproliferative activity of small molecule inhibitor SU11274 in vitro, unexpectedly, was in contrast to its protumorigenic effect on the Rel3 cells in vivo [6]. More importantly, SU11274 significantly increased frequency of tumor initiating cells. We hypothesize that there might be several reasons for the protumorigenic outcome including individual response of given model cell line, different route of administration or experimental setup. Although intraperitoneal administration of SU11274 decreased metastatic burden in liver of intrasplenically injected mice in orthotopic model [7], the same route of administration supported tumor growth of subcutaneously xenografted melanoma cells Rel3 in heterotopic model (Fig. 2b). Etnyre et al. [8] achieved antitumor effect by direct intratumor injection of compound in model melanoma. Taken together these data stress extreme plasticity of melanoma cells and context-dependent nature between protumorigenic and antitumorigenic action of small molecule inhibitor.

Hierarchical organization of melanoma remains a matter of a debate. It was shown that melanoma cells possess considerable plasticity and represent a tumor type with shallow if any hierarchy [16]. Our data support this notion as we detected high expression of pluripotent proteins in unaffected Rel3 cells. Switching melanoma cells to spheroid non-adherent culture conditions further enriched for melanoma-initiating cells as confirmed by the in vivo assay. We were able to propagate the melanospheres long-term both with or without SU11274 (more than 10 passages corresponding to more than 10 weeks) thus demonstrating the presence of self-renewing cells in vitro and no detrimental effect of the SU11274 on them. We have examined surface marker expression with a particular focus on putative melanoma cancer stem cell markers [10]. We could not find any significant alteration (up- or down-regulation) in any of these markers tested such as c-Kit, CD271, CD133, ABCB5, ABCB1, ALDH1 in the SU11274-treated cells versus the untreated ones (data not shown). Non-adherent culture conditions did not alter melanoma differentiation marker CD146 (M-CAM). It also did not change expression of VEGFR2, VE-cadherin CD144 or angiogenic marker CD31. Based on the data we excluded that increased tumor initiation could be due to a vasculogenic mimicry [40].

SU11274 was previously shown to alter expression profile of the treated cells attributed to its off-target action. Thirty-nine genes belonging to the apoptosis/necrosis, inflammation, oxidative/metabolic stress, heat shock, proliferation/carcinogenesis and growth arrest/senescence pathways were altered at least 2-fold (by increasing or decreasing them) by SU11274 in ovarian cancer cells [41]. These data show its broader action and capability to induce multiple target genes involved in oxidative and metabolic stress [41], so this compound cannot be considered as a c-Met specific inhibitory agent. Crizotinib represents more targeted agent in comparison to SU11274 and it did not alter cellular ATP content in treated cells. Based on the correlation to SU11274-mediated increase in tumor initiation in vivo we concluded that the off-target action of SU11274 is responsible for its protumorigenic action. Furthermore, it favors our hypothesis that melanoma-initiating capability is linked to the metabolic state of cells. The experiments investigating these effects in other tumor types might bring further insight how altering bioenergetic state might potentially support the tumor initiation. SU11274 upregulated almost 2-fold several stem cell markers (Oct3/4, Nanog, AFP and Gata4) in treated cells (Fig. 5a). It also increased activity of RSK1/2/3 kinase based on the phosphotyrosine array analysis. Martin et al. [25] also reported increased ERK activity resulting in RSK1 activation correlating with protumorigenic action in metformin treated melanoma cells A375. However, our data did not confirm any VEGF expression upon SU11274 treatment in contrast to their conclusions that VEGFA upregulation led to protumorigenic action of metformin [25]. It suggested that SU11274 compound exerts its protumorigenic action in the absence of increased VEGF secretion and prompted us to examine bioenergetic regulation.

As recently reviewed, increasing evidence suggests that many types of stem cells rely on anaerobic glycolysis and their stem cell function is regulated by bioenergetic signaling [22, 42]. Similar mechanisms might be operating in cancer stem cells, in fact some studies have already suggested critical role of the metabolic de-regulation for stemness [43]. These findings open novel therapeutic intervention points in cancer. Liu et al. suggested that glycolytic inhibitor 3-BrOP could be combined with standard chemotherapy to target both side population and bulk tumor mass. It was sufficient to treat cells for 24 h with 3-BrOP to achieve antitumor effect in contrast to a platinum-derived agent, which did not affect tumor growth whilst sparing the side population [38]. Higher glycolytic rate seems to be a general characteristic of melanoma cells. Oncogene BRAF, which is constitutively activated also in our melanoma model Rel3 was implicated to be directly involved in reprogramming of cellular metabolism. Dichloroacetate (DCA) as a pyruvate dehydrogenase kinase inhibitor, exerted antimelanoma effect and potentiated its response to specific BRAF inhibition by vemurafenib [21]. Chemosensitivity was not significantly altered in the SU11274-treated cells in vitro and in vivo. We were not able to find any combination of the SU11274 molecule with another chemotherapeutic drug to achieve synthetic lethality (data not shown).

Over the last years, several strategies to target melanoma stem cells were suggested [44, 45]. There were attempts to target a self-renewal pathway of melanoma stem cells thus disabling their ability to replicate as rewieved in [11]. Bioenergetic modulation seems to emerge as novel strategy to target melanoma cancer stem cells. In our work, we present data from the experiments in vivo, which support this hypothesis. Protumorigenic action of small molecule SU11274 could be counteracted by bioenergetic modulator dichloroacetate. On the other hand, glycolytic inhibitor 3-bromopyruvate did not prove suitable, thus showing the specificity in the signaling cascade induced by SU11274 in melanoma cells. More importantly, we observed no antitumorigenic action when dacarbazine was used. Our study further underlines the importance of drug testing in non-adherent spheroid cultures as these might better reflect the efficiency against tumor initiating cells [46].

Our work highlights a role of bioenergetic modulation in melanoma initiation. It shows that antiproliferative effect in vitro can actually lead to increased tumorigenicity in vivo. In summary, melanosphere cultures were substantially enriched for melanoma-initiating cells in vivo in the absence of any alteration in cancer stem cell markers. Small molecule SU11274 originally intended as a specific c-Met inhibitor significantly reduces melanosphere proliferation, but increases intracellular ATP content, which correlates with an increased tumorigenicity. Tumorigenicity could be reduced in SU11274-treated cells by bioenergetic modulator DCA indicating that glycolytic inhibition could counteract SU11274 mediated effect on melanoma initiation.