Forty-one benthic Baltic Sea cyanobacterial strains of the genera Anabaena, Nostoc, Calothrix and Cyanothece (Table S1 and Figure S1), previously screened for death-inducing activity against primary hepatocytes [17], were extracted sequentially with water, aqueous 70% methanol, and 1:1 methanol:dichloromethane (A, B and C extracts, respectively). They were first screened for cell death-inducing activity against the AML cell line IPC-81wt [22], which originates from the BNML rat model, known to be a reliable predictor of the outcome of anticancer treatment [23]. Nineteen A extracts, 13 B extracts, and 6 C extracts contained cell death inducers. Ten strains (M4, 8, 9, 11, 12, 13, 23, 26, 33, 37) contained cell death inducing activity in more than one extract, like strain M26 that contained activity in the A and C extract, but not in the intermediate B extract (Figure 1). Several of the extracts with anti-AML activity, most notably the A extracts from the Anabaena sp. M4, M27, M30, M44 and M45 (Figure 1), were previously found to lack activity against normal primary hepatocytes (see [17]). We therefore conclude that marine benthic cyanobacteria are not only a rich source of cell death inducing activity, but also contain apoptogens selective for AML cells.

Thrombocyte function is critical in leukemia, which is characterized by platelet deficiency due both to thrombocytopenia and thromboembolic complications [8,24,25]. Moreover, treatment of leukemia with cytotoxic drugs may change platelet function in vivo (reviewed in [26]). We therefore studied if any of the cyanobacterial extracts contained blood platelet activating activity that could limit usefulness as anti-AML agents. Interestingly, we found that a high proportion of the C extracts contained blood platelet activators (Figure 2). Of more importance for the present study is that the antileukemic A extracts of M27, M30, M44 and M45, unlike the anticancer drug daunorubicin [26], lacked platelet-modulating activity (Figure 2A).

Although the IPC-81 AML cells used in the primary screen respond in a clinically relevant way to present day anti-leukemic drugs both in vivo and in vitro [23,27], it is important to know if the IPC-81 apoptogens induce death in other relevant AML cell lines. In this respect, the aqueous extracts from M27, M30 and M44 were overall more efficient than those from M45 and M4 (not shown). M27, M30 and M44 induced apoptosis rapidly in human HL-60 AML cells (Figure 3), even compared to a higher-than-therapeutic concentration of daunorubicin in these cells ([27], data not shown). The high potency and rapid action of M27, M30 and M44 towards the leukemia cells are advantageous for in vivo therapy, since there will be a shorter burst of the drug, thus allowing less time for the formation of potentially generally harmful metabolites of the drug. For these reasons, the aqueous extracts from M27, M30 and M44 were selected for further studies.

Cell death induced by M27, M30 or M44 had the morphological hallmarks of apoptosis, such as chromatin condensation, nuclear fragmentation, surface budding and shedding of apoptotic bodies (Figure 4A–D). The apoptotic phenotype was as striking as the one observed in cells incubated with the benchmark anti-AML drug daunorubicin (Figure 4E).

Cells overexpressing Bcl-2 are resistant towards stimuli leading to mitochondrial instability with release of pro-apoptotic proteins [28]. IPC-81 AML cells with enforced expression of Bcl-2 [29] were resistant towards apoptosis induced by either daunorubicin or by the apoptogens in M27, M30 or M44 (not shown). This indicates that like daunorubicin, the cyanobacterial apoptogens signaled through the mitochondrial cell death pathway. Significant differences were discovered, however, between the AML apoptogens with respect to the involvement of caspase dependent cell death pathways. The apoptosis induced by activation of the cAMP-dependent protein kinase by a cAMP analogue was completely abrogated by the caspase inhibitor zVADfmk, whereas apoptosis induced by M30 or M44 appeared to be caspase independent and apoptosis induced by daunorubicin or M27 was partially inhibited by the caspase inhibitor (Figure 5A).

Drug resistance is the major cause of therapeutic failure in AML [9,10]. We have found that LEDGF/p75 is consistently upregulated in AML blasts from patients with chemoresistant AML [11]. LEDGF/p75 is a survival factor reported to inhibit a caspase-independent lysosomal cell death pathway [13]. As previously observed [11], we found, that IPC-81 cells with enforced expression of LEDGF/p75 were less sensitive towards apoptosis induced by either cAMP analog or daunorubicin. In contrast, such cells had intact sensitivity towards M44, and even showed enhanced apoptosis, compared to wild-type cells, at low M44 concentrations (Figure 5B). M27 and M30 induced a response intermediate between M44 and daunorubicin. They resembled M44 by tending to enhance apoptosis in the LEDGF/p75 overexpressing cells at low concentrations, but were like daunorubicin less efficient against these cells at high concentrations (Figure 5B). We conclude that M44 and to a lesser extent M27 and M30 had better ability than daunorubicin to circumvent the chemoresistance provided by LEDGF/p75.

Since M44, and to a lesser extent M27 and M30, differed from daunorubicin in cell death mechanism and in efficiency against LEDGF/p75 overexpressing cells, they could be useful adjuncts to daunorubicin. In fact, all three synergized strongly with daunorubicin to induce apoptosis whether the AML cells had enforced expression of LEGDF/p75 or not (Figure 6A and B). This suggests that the cyanobacterial apoptogens have potential to enhance the efficiency of daunorubicin, also in cells expressing the chemoresistance-associated LEDGF/p75.

A major limitation for drugs and drug combinations is unwanted side effects [10,16]. Since the major side effect of anthracyclines in cancer therapy is cardiotoxicity [8,16], we studied whether any of the cyanobacterial activities had intrinsic cardiomyoblast toxicity or enhanced the toxicity of daunorubicin in such cells. The intrinsic ability of M27, M30 or M44 to induce cardiomyoblast death was far lower than for the AML cells, confirming their AML selectivity (Figure 6 and data not shown). Furthermore, at concentrations effective against AML cells, M27, M30 and M44 failed to enhance the daunorubicin toxicity towards cardiomyoblasts. On the contrary, M44 significantly reduced the daunorubicin toxicity (Figure 6C).

Dichloromethane, dimethylsulfoxide (DMSO), Hoechst 33342, penicillin/streptomycin, 8-chlorophenylthio-cAMP and daunorubicin were from Sigma-Aldrich (St. Louis, USA). Methanol and formaldehyde were from VWR (West Chester, USA). Dulbecco’s Modified Eagle Medium (DMEM), Fetal Calf Serum (FCS) and Horse Serum (HS) were supplied by EuroClone® Life Sciences Division (Milan, Italy). RPMI 1640 Medium (Gibco®invitrogen cell culture) was from Invitrogen AS (Carlsbad, USA). PE-conjugated anti-human CD62P antibody (CAT. #348107) was from Becton Dickinson Biosciences (San Jose, USA). Thrombin receptor agonist peptide (TRAP, SFLLRN) was from the Biotechnology centre of Oslo (Oslo, Norway). Z-val-ala-DL-asp-fluoromethylketone (zVAD-fmk) was from Alexis Biochemicals (San Diego, USA).

The cyanobacterial strains used in this study [17] were collected from sediment, sand, surface of stones, rocks, water plants, macroalgae, mussels or molluscs from the Baltic Sea at Porkkala (Gulf of Finland). Strains had previously been identified [17] based on morphology [36] as Anabaena, Nostoc, Calothrix or Cyanothece (see supplementary material: Figure S1 and Table S1) and mass cultivated as previously described [17]. Cyanobacterial cells were cultured for 20–60 days before they were harvested by centrifugation and freeze dried.

Extraction of cyanobacterial biomass was performed as described by Herfindal et al. [17]. In brief, freeze-dried cyanobacterial biomass was first extracted in water, followed by extraction in 70% aqueous methanol and finally 1:1 (v/v) methanol:dichloromethane. The extracts were then dried in a centrifugal vacuum concentrator (Savant SpeedVac®AES1010) and resuspended in either 0.2 mL water (aqueous; A extract), 0.05 mL water (70% aqueous methanol; B extract) or 0.05 mL DMSO (methanol:dichloromethane; C extract).

The rat promyeloid leukemia cell lines IPC-81 wild-type cells [22], and IPC-81 with enforced expression of LEDGF/p75 [11] or Bcl-2 [29], were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 50 U/mL of penicillin and 0.05 mg/mL streptomycin and 10% (v/v) heat inactivated horse serum. The rat cardiomyoblastic cell line H9c2 (ATCC: CRL-1446) was cultured in DMEM with antibiotics as described above but enriched with 10% (v/v) heat inactivated fetal calf serum. When the cells reached 80% confluence, they were detached by mild trypsin treatment (0.33 mg/mL trypsin for 5 min at 37 °C), washed and reseeded in fresh medium to 25% confluence. The human myeloid leukemia cell line HL-60 (ATCC: CCL-240) was cultured in RPMI medium with antibiotics as described above and enriched with 10% (v/v) heat inactivated fetal calf serum. All cells were cultured at 37 °C in a humid atmosphere of 5% CO₂.

For assays of cell death, the suspension cells were centrifuged at 160 × g for 4 min, resuspended in fresh medium and 1.5 × 10⁴ cells were seeded in 0.1 mL in a 96-well tissue culture plate. The cardiomyoblasts were detached, centrifuged, and 2000 cells were seeded in each well of a 96-well tissue culture plates 24 hours before start of the cell death assays. The experiments were stopped by addition of 0.1 mL PBS containing 4% formaldehyde (pH 7.4) and 0.01 mg/mL of the DNA specific fluorescent dye Hoechst 33342. Apoptosis was scored by differential interference contrast and fluorescence microscopy (Axiovert 35M, Zeiss) as previously described [37]. Untreated cells or cells added vehicle exhibited a viability of more than 97%.

Isolation of human blood platelets was from fresh whole blood, as previously described [38,39]. In brief, platelet rich plasma obtained by centrifugation of whole blood was gel filtered and platelets eluted into a calcium-free Tyrode’s buffer containing glucose and bovine serum albumin at pH 7.3.

The gel-filtered human blood platelets (3.5 × 10⁷ per sample) were incubated under non-stirring conditions for 30 min with cyanobacterial extract or Thrombin Receptor Agonist Peptide SFLLRN (TRAP), PBS (5 μL) and anti-CD62PE (0.24 μg/mL) in a total volume of 25 μL. The platelets were then fixed 1:16 with 0.5% (v/v) paraformaldehyde in PBS and analysis of P-selectin (CD62) expression was carried out using a FACSort flow cytometer as described by Selheim et al. [40]. Ten thousand platelets, identified by size (light scatter gating), were analyzed for CD62PE fluorescence per sample.