Organometallic nucleosides induce non-classical leukemic cell death that is mitochondrial-ROS dependent and facilitated by TCL1-oncogene burden

We have been extensively studying butadiene Fe(CO)₃-complexes like Me-N69 and could show that their iron-fragment is crucial for the cell death inducing activity in BJAB lymphoma cells [21]. We found cytosine to be the nucleobase most effective for cell death induction as well as a non-polar protecting group at O5’ to be essential for this substance efficacy [21]. There was no significant difference in activity between furanoid compounds (partially hydrated furanes) like Me-N69 and its carbocyclic congeners [21]. Given these observations, we designed optimized MCNA. They were synthesized according to our established protocols [21‐23]. Based on chemical properties and extrapolated activity indices, 4 compounds were further selected for this study (Fig. 1). Huni132 is a ferrocene-based nucleoside analogue, which contains Fe(II). Huni218 is the corresponding ruthenium congener. Me-N69 is a butadiene Fe(CO)₃-complex, which contains Fe(O) and can donate up to three electrons upon oxidation. Huni132, Me-N69, and Huni218 resemble natural nucleosides with respect to the relative position of the cytosine unit and the 5’-side chain. Dia-Me-N69 is formed as a diastereomeric byproduct during the synthesis of Me-N69. It features a stereochemically inverse, hence ‘atypical’, configuration at C1 and was added as a control.

Flow cytometry with AnnexinV (early-apoptotic surface expression of the neoepitope phosphatidyl-serine) combined with 7AAD or Hoechst stain (uptake upon late-apoptotic or necrosis-associated membrane disintegration) assayed the ability of our MCNA to trigger death of primary tumor cells from CLL patients and of healthy-donor derived PBMC. Initial estimations of effective compound dosages were performed by LD50 titrations (Additional file 1 Fig. S1). Next, already after 24 h all 4 tested MCNA, with the highest efficacy noted for Huni132, induced pronounced cell death in 25 CLL samples whereas normal PBMC were significantly less sensitive towards MCNA treatment (Fig. 2a). Solvent (Dmso) induced cell death was negligible (mean/SEM: PBMC 3.7 %/0.9 and CLL 5.4 %/1.3). This overall selectivity by the MCNA was greater than observed for bendamustine at this time point and for fludarabine, compared to which the MCNA also proved to be more B-cell specific (Fig. 2a, Additional file 1 Fig. S2).

To evaluate the activity of the MCNA across different risk subsets of CLL (Additional file 1 Tables S1, S2), we compared post-exposure cell survival of cytogenetically high-risk cases [5, 6, 24] carrying 11q or 17p deletions (N = 9,) with 12 good-prognosis CLL (normal karyotype or isolated 13q14 deletion). The MCNA triggered cell death was at comparable rates, independent of cytogenetic disease subset (Fig. 2b). Responses were also mostly irrespective of IGHV mutational status or of previous therapy (Additional file 1 Fig. S3). Furthermore, the MCNA-induced biochemical response of PARP cleavage, a distal event in several routes of programmed cell death, was similar between samples of different karyotypes and from patients with distinct disease stages (Fig. 2c).

Next, we evaluated potential cooperative effects of our novel substances with bendamustine, a well-established clinically active drug with a good therapeutic index (low toxicity) in CLL. We preferred this well-described alkylator for this and some subsequent sets of experiments over the conventional adenosine nucleoside fludarabine based on bendamustine’s faster reaction kinetics (Additional file 1 Fig. S1). The fludarabine metabolism with synthesis of cellular F-ara-ATP peaking after ≈ 48–72 h allows reliable efficacy recordings for fludarabine starting only at 48 h. Therefore, and since we already expected the MCNA not to act as “classical” nucleosides [23], bendamustine appeared as the better comparison when the 24 h MCNA effects were studied.

Freshly isolated primary leukemic cells from 5 CLL patients were treated for 24 h with 5 μM of each MCNA, alone or in combination with 25 μM bendamustine followed by AnnexinV/7AAD flow cytometry and immunoblot (PARP cleavage) analysis. There were at least additive effects across all combinations, with the most significant cooperation observed for bendamustine with Huni132 (P = 0.006) showing even a synergistic relationship (Huni132: 79.8 ± 9, Bendamustine 70.8 ± 9, Bendamustine + Huni132 46.4 ± 7.5; all Mean ± SEM; Fig. 2d). These effects were further corroborated by an increased cleavage of PARP seen for those drug combinations (Fig. 2d, bottom).

Fludarabine resistance in CLL remains a clinical problem. We therefore analyzed samples from 9 CLL patients with clinically fludarabine-refractory disease for responses to the MCNA; as 4 cases to all 4 MCNA (see cytogenetic high-risk subsets, above and Fig. 2b) or as all 9 cases to only Huni132 and Me-N69 (Fig. 2e). Importantly, the MCNA and in part also bendamustine were still efficient in the context of clinical and in vitro fludarabine resistance (Fig. 2e).

The high apoptotic threshold of CLL B-lymphocytes is significantly determined by milieu-derived signals. The direct contact between leukemic cells and stromal components, i.e. BMSC, was shown to protect from spontaneous and drug-induced apoptosis [25, 26]. We therefore explored the influence of the MCNA on CLL cell viability in a stromal cell coculture system as this represents a higher fidelity model of niche-mediated resistance than suspension cultures. For this, CLL cells were kept either in suspension or as cocultures with the human BMSC line NK-tert. After 24 h of drug exposure, cell death rates and levels of cleaved PARP were recorded. In line with our previous findings [26], we observed upregulation of protein levels of the pro-survival oncogene TCL1 in CLL cells upon NK-tert co-culture (Fig. 2f). The BMSC coculture protected CLL cells from spontaneous and in part from bendamustine-induced cPARP-induction/cell death. In contrast, the MCNA consistently evoked marked cell death also in the presence of BMSC-mediated pro-survival stimuli; in the case of Huni132 nearly to the same extent as in the suspension cultures. For cPARP densitometry see Fig. 2f; corresponding AnnexinV/7AAD double-negative fractions of surviving cells (not shown) were (suspension vs coculture; percentage-means ± SEM): Huni132: 23.3 ± 3.2 vs 34.6 ± 10.2, Huni218: 56.8 ± 8.2 vs 70.9 ± 7.6, Me-N69: 38.8 ± 7 vs 62 ± 7.2, Dia-Me-N69: 66.5 ± 8.2 vs 86 ± 4.9, and bendamustine: 89.3 ± 4.5 vs 94.5 ± 0.9. As the feeder cells themselves remained unaffected by the treatment with the drugs (Additional file 1 Fig. S4), the differential induction of leukemia cell apoptosis was not due to loss of BMSC protection, rather CLL-cell specific. Furthermore, only incubation of CLL cells in supra-physiological 100 % (instead of 10 %) serum of either fetal calf or human origin confered some degree of protection from Huni132 mediated cell death (Additional file 1 Fig.S5).

Taken together, we demonstrated thus far that our novel MCNA efficiently and specifically trigger cell death in CLL, irrespective of the disease risk group or the presence of protective cell-cell mediated micro-environmental signals.

Currently, at least 3 mayor pathways of programmed cell death are distinguished by characteristic cell fates, namely apoptosis, autophagy, and necroptosis [27‐29]. Upon MCNA exposure, we identified morphologic changes that were rather characteristic for apoptosis, namely pyknotic nuclei and cytoplasmic condensations (not shown). They paralleled the pronounced AnnexinV induction (above). However, to discriminate in more detail among the pathways involved in the MCNA-mediated CLL cell death, we employed Z-VAD and necrostatin-1 as specific inhibitors of caspase-mediated apoptosis and caspase-independent necroptosis, respectively. In addition, we tracked the autophagy-specific proteolytic conversion of the marker protein LC3B-I to autophagosome associated LC3B-II. We also analyzed the cleavage of PARP, which is mainly, but not exclusively [30], triggered by the activation of the caspase signaling cascade.

We found that the MCNA lead to PARP cleavage in CLL cells without preceding caspase-3 activation (Fig. 3a). Neither Z-VAD nor necrostatin-1, were able to prevent the cleavage of PARP by the MCNA (Fig. 3a,b, Additional file 1 Fig. S6). However, the PARP-inhibitor olaparib mitigated MCNA-mediated CLL cell death (Additional file 1 Fig. S7). This suggested that the MCNA-mediated cell death is to a large degree PARP dependent, but not promoted through classical apoptosis or necroptosis. Although the presence of Z-VAD could not fully prevent the generation of cleaved PARP by bendamustine exposure, this pan-caspase inhibitor suppressed the otherwise robust induction of caspase-3 cleavage products by bendamustine. Furthermore, while bendamustine triggered strong P53 Ser15-phosphorylation, none of the MCNA evoked such p53 phospho-(p)-activation (Fig. 3c). PARP cleavage and higher cellular death rates by MCNA treatment were not associated with autophagic LC3B conversion (Fig. 3c), which however was the case for the positive control of rapamycin (Additional file 1 Fig. S8).

Together these data suggest that the programmed cell death mediated by the MCNA can be considered as a PARP mediated, caspase-3 independent, hence non-classical, way of apoptosis that does not involve P53. This distinguishes these substances strikingly from conventional agents like bendamustine, which acts at least to a relevant part caspase-3 dependent and via P53.

Caspase-independent cell death is important for organismal protection upon failure of the caspase-mediated pathways, and can be triggered in response to cytotoxic agents [12, 23, 31]. Damage of organelles including mitochondria and increases of cellular ROS levels are frequently involved in such caspase-independent cell death [12]. Based on their chemical properties (ionic metals) we predicted our MCNA to induce ROS in a Fenton’s reaction (iron-catalyzed hydrogen peroxide synthesis). Therefore, we analyzed cellular levels of ROS in MCNA-treated CLL (n = 3) cells in relation to induced apoptosis to test for this possible caspase-independent mode of MCNA action. Recordings of cellular retention of the fluorescent (oxidized) CM-H2DCFDA dye measured ROS levels at different time points after drug exposure (2 h, 4 h, 6 h, and 12/24 h). Increased ROS levels were revealed already at 4–6 h after MCNA exposure (Huni132, Fig. 4a), while bendamustine-treated cells showed unchanged ROS levels. The earliest AnnexinV expression, however, was detectable after 12/24 h of drug treatment demonstrating that the ROS increase precedes the induction of apoptosis. According to these data we speculated that increasing ROS levels are causally linked to the induction of early apoptosis by the MCNA, which stands in contrast to the p53-mediated activation of a ROS-independent DNA-damage stress response and caspase-mediated apoptosis by bendamustine.

To highlight the functional and specific role of ROS in MCNA mediated killing of leukemic cells, we investigated cell death (3 CLL cases) in the presence of the cell-permeable ROS scavengers Tiron (chelator, scavenger of superoxide ions and free electrons [32]), NAC (multimodal; antioxidant through GSH synthesis [33]), and Deferasirox (oral iron-chelator in clinical use). These substances all mitigated the redox-induced apoptosis by the MCNA as we observed a subtotal reduction of ROS and a reduced degree of MCNA-triggered cell death in their presence (Tiron, Fig. 4b). From that, we conclude that the mode of anti-leukemic action of the MCNA is at least to a high degree, but possibly not entirely, ROS dependent. In agreement with our data on a rather selective cell death induction by MCNA treatment in CLL tumor cells over healthy-donor PBMC (Fig. 2a), we were also not able to detect MCNA-induced ROS increases preceding the low degree of apoptosis in these normal cells (Fig. 4b). This could provide a rationale for a low cytotoxicity of MCNA in healthy tissues.

We proposed that a major underlying factor for the tumor-specific activity of the MCNA is the disturbed redox equilibrium present in CLL cells [8, 10, 34]. In the search for causative factors of elevated ROS in CLL [8], we postulated that besides an increased mitochondrial mass [34], oncogenic stress represents a leading force towards this part of the aberrant redox signature of CLL. Therefore, we decided to investigate this potential vulnerability aspect further by following the hypothesis that one of the central oncogenes in CLL, TCL1, might represent a causal link. This is reasoned by TCL1’s activating interaction with AKT [19, 35], a central growth-promoting signaling node and an established inducer of oxygen free radicals [36].

Apart from their function in programmed cell death, intracellular ROS are known to regulate a number of tumor-associated signaling cascades particularly PI3K/AKT and NFkB, the 2 major effector pathways of the lymphoid oncogene TCL1 [35, 37, 38]. To explore the potentially tight relation of tumor cell inherent ROS to TCL1, first the number of TCL1+ cells within the CD5+/CD19+ CLL tumor cell population was correlated with the overall leukemic ROS levels by flow cytometry. In PB samples from 20 CLL, we observed a significant association (rho 0.66, P < 0.002) of ROS levels with the percentage of TCL1+ cell within the malignant clone (Fig. 5a left). These results were confirmed in systems of modulated TCL1. Eμ-TCL1-tg mice develop noticeable CD5+ B-cell proliferations at 6–10 months resulting in overt CLL after 12–15 months [18]. Fittingly, the ROS levels in splenocytes from 7-months old pre-leukemic, (Fig. 5a right-top) and leukemic (not shown) TCL1-tg mice were significantly higher than in splenocytes from age-matched C3HxC57BL/6 wild-type controls. Similarly, introduction of TCL1 in the CLL-like cell line JVM3 led to strongly elevated ROS levels as compared to the parental cells (Fig. 5a right-bottom). The findings from both experimental systems argue against the possibility that the elevated ROS levels are predominantly a general cancer phenomenon unrelated to TCL1 (TCL1 transfection of an already fully transformed cell line; analysis of pre-leukemic mice) or to be seen in the context of organismal/lymphocyte ageing (age-matched animals). More likely, the results suggest that the elevated TCL1 expression, characteristic for CLL [19, 20], stands in a causal relation to a dysbalanced redox state resulting in aberrantly high ROS levels.

We next analyzed whether TCL1 overexpression, linked through the elevated ROS levels translates into altered sensitivity towards our MCNA. Indeed, upon MCNA treatment the amount of cleaved PARP was noticeably higher in the TCL1-carrying JVM3 CLL-like B-cells than in the parental line, which was confirmed by the number of AnnexinV expressing cells (not shown). In contrast, bendamustine sensitivity was slightly reduced by stable TCL1 transfection (Fig. 5b, top), a phenomenon we observe more strikingly for fludarabine in vitro and in the clinical setting [19]. Similarly, TCL1-driven murine leukemias (Eμ-TCL1-tg splenocytes) showed a good response to our novel agents (Fig. 5b bottom-left). Furthermore, MCNA treatment elicited much stronger cytotoxic responses in TCL1-positive CLL as compared to TCL1-negative cases (Fig. 5b bottom–right). Together, these data demonstrate that high TCL1 oncogene expression directly associates with elevated cellular ROS levels and by that mediates an increased vulnerability towards MCNA.

The sensitivity to our MCNA that TCL1 oncogene overexpression conferred via elevated ROS was intriguing as it stands in contrast to a general TCL1-associated resistance towards classical apoptosis inducing agents [19]. Therefore, we asked for the specific underlying mechanisms of this vulnerability. Thus far, our results showed that the apoptosis induced by MCNA does not rely on active caspase-3 (Fig. 3a). We next investigated the release of mitochondrial proteins in response to apoptotic stimuli with particular focus on the flavoprotein apoptosis-inducing factor (AIF), a known mediator of caspase-independent apoptosis [31], Smac, the suppressor of the inhibitors of apoptosis proteins (IAP) [39], and cytochrome c, an electron transport protein released from mitochondria during the early stages of apoptosis [40]. JVM3 CLL-like cells and their stable TCL1 transfectants were treated with Huni132 or Huni218, and the protein levels of AIF, Smac, and cytochrome c were analyzed in cytosolic and mitochondrial fractions by immunoblots. There were no obvious uniform changes in AIF, Smac, or cytochrome-c levels at baseline (untreated cells). However, most consistently, exposure to the MCNA resulted in mitochondrial depletion of all three proteins in the TCL1-transfected JVM3-cells (Fig. 6a). A simultaneous cytosolic accumulation of these released molecules was not captured at the parallel time points, which points to different kinetics of their degradation and established nuclear trans-localization. Overall, these data implicate that the presence of TCL1 promotes the apoptotic effects of the MCNA at the mitochondrial level and involves mediators of caspase-independent pathways.

We next monitored bio-energetic rates of the two major sources of cellular ATP generation, glycolysis and mitochondrial respiration, in the context of substance challenge and in relationship to the impact of TCL1. Interestingly, basal mitochondrial respiration (OCR) was not increased in TCL1-transfected cells (Fig. 6b). However, basal aerobic glycolysis as indicated by lactate production (ECAR), was significantly lower in the JVM3-TCL1 cells (Fig. 6b). Furthermore, treatment with the complex-V inhibitor oligomycin reduced the OCR on average to 41 ± 7 % (of baseline level) in the TCL1-transfected JVM3 cells vs 84 ± 4 % in the TCL1-negative JVM3 control (Fig. 6c). Since the mitochondrial complex-V (F0F1-ATPase) couples the electron chain to ATP synthesis, this result of a higher oligomycin-sensitive fraction indicates that TCL1 abundance is associated with a significantly higher fraction of oxygen consumption being coupled to ATP-synthesis (higher ATP turn-over). The maximal respiratory capacity as determined by addition of the uncoupling agent FCCP was not impacted in TCL1’s presence. The antimycin/rotenone insensitive OCR fraction was lower in the TCL1 condition indicating a lower contribution of the non-mitochondrial component of OCR to overall respiration under TCL1 impact.

In conjunction, these data implicate a TCL1-mediated shift towards a higher relative contribution of the mitochondrial metabolism over glycolysis to overall ATP generation and over non-mitochondrial sources of oxygen consumption to global respiration. It could partially provide an explanation for the increased ROS levels mediated by TCL1 (Fig. 5). At the same time a stronger dependence on mitochondrial energetic flux seems to render TCL1+ cells more susceptible towards agents interfering with mitochondrial homeostasis such as the tested MCNA.

Under physiologic conditions, production of ROS is a byproduct of normal ATP-dependent mitochondrial respiration [34]. Given the selective efficacy of our MCNA in CLL acting through ROS induction and based on the modulatory impact of the CLL oncogene TCL1 on mitochondrial metabolism, we hypothesized that the MCNA induce cell death in primary CLL cells by significantly affecting their mitochondrial function. We analyzed the respiration and membrane potential of mitochondria in CLL samples after incubation with MCNA. We found that mitochondrial respiration (OCR) is significantly reduced upon MCNA exposure compared to bendamustine-, fludarabine-treated or untreated CLL samples (Fig. 7a). In addition, we assessed the mitochondrial membrane potential (∆ΨM) as an established marker for mitochondrial fitness. The energy produced during mitochondrial respiration is maintained as a high ∆ΨM whereas declining mitochondrial function or their destruction is associated with a ∆ΨM loss. As anticipated, application of the MCNA, but neither bendamustine nor fludarabine, resulted in a rapid mitochondrial depolarization in primary CLL cells suggesting that mechanistically our MCNA promote ROS production in tumor cells by affecting mitochondrial respiration and membrane polarization (Fig. 7b, Additional file 1 Fig. S9).