Biomass from ten diatom isolates was extracted sequentially, and the extracts were tested for ability to induce hepatocyte and leukemia cell death. We found strong induction of leukemia cell death by seven of the aqueous extracts. Only three aqueous extracts induced hepatocyte cell death (Table 1). While several of the methanol extracts showed moderate leukemia cell death induction, none of the organic extracts showed any death induction (Table 1).

A subclone of the IPC leukemia cells expressing the prosurvival oncogenic protein Bcl-2 is resistant towards several apoptotic stimuli [10,11], including current anti-leukemic chemotherapy drugs like the anthracycline. Although the IPC-Bcl-2 cells were completely resistant to some extracts (from isolates ND 50, 51, 53, 71 and 79) they were only partly resistant to extracts from isolates ND52, 55, 56, 58, 73 (Table 1). This is of interest since an important challenge in cancer therapy is to overcome treatment resistance due to cancer cell overexpression of survival proteins like Bcl-2. A well-known group of microbial toxins are protein phosphatase targeting substances like microcystin, nodularin, calyculin A and okadaic acid [12]. These are also able to overcome the cell death protection by Bcl-2 [13], but we found no such activity in any of the extracts (not shown) using a highly sensitive assay [14].

During the screening we discovered that the extracts produced distinct forms of cell death. The cell death morphology was therefore studied in more detail (Figures 1A–C, Figure 2B). When tested at a concentration inducing about 50% overall cell death, the aqueous extract from ND73 induced chromatin margination without alteration of the nuclear shape or fragmentation of the nucleus of IPC-81 cells (Figure 2B). In contrast, the other aqueous extracts produced massive nuclear fragmentation (Figures 1B and C). At higher concentrations the extracts caused complete nuclear disintegration (not shown). An in depth morphological analysis by transmission electron microscopy was undertaken of IPC-81wt cells treated with the ND58 aqueous extract (Figures 1 D–G). IPC-81wt cells treated with ND58 aqueous extract for 8 h showed two distinct morphologies. A typical apoptotic cell (Figure 1E) had disintegrated into apoptotic bodies, which could contain nuclear fragments with hypercondensed chromatin, mitochondria, and other organelles. In some cells, we noted the formation of vesicles resembling autophagosomes (Figure 1F), probably representing an early stage towards excessive autophagy where the cell interior was dominated by autophagic vesicles containing partially digested organelles (Figure 1G). The nucleus appeared to be disintegrated, and the nuclear envelope and chromatin could not be identified.

We next wanted to find out if different toxins caused the cell death morphologies induced by ND58 (Figure 1C) and ND73 (Figure 2B). We found that reversed phase chromatography of an anion exchange extract from ND73 gave a toxic fraction at around 5 minutes (Figure 2), which reproduced the distinct nuclear morphology seen in cells treated with the crude extract (Figure 2B). In contrast, the toxicity of ND58 eluted later under identical chromatographic conditions (Figure 3A and data not shown). We have not been able to purify enough material from ND73 to continue the analyses of the toxic fraction, but noted that the UV-spectrum of the active fraction of ND73 (inset in Figure 2A) was different from that of ND58 (inset in Figure 3A). Moreover, the anion exchange extract of ND73 produced a peak at 8 min (Figure 2A) with similar UV-spectrum (not shown) as that of the active peak of ND58 (inset in Figure 3A). We conclude that the diatoms contained water-soluble compounds able to induce various types of cell death in leukaemia cells, including classic apoptotic and autophagic death, both via pathways blocked by Bcl-2 and pathways overriding Bcl-2.

Some of the anti-leukemic activity of the ND58 extract was irreversibly bound to the QMA-resin during solid phase separation, and less than half of the initial activity was recovered in the flow-through and wash fraction. This apoptosis-inducing fraction was loaded onto semipreparative reversed phase HPLC, and nearly all the cytotoxic activity was found to be associated with one peak (Figure 3A). When analysed by Q-tof mass spectrometry (Figure 3B and C) the peak was found to contain multimers of m/z 268 (Figure 3B), which when fragmented all produced a daughter-ion with m/z of 136 and a corresponding neutral loss of 132 Da (not shown), like adenosine (Figure 3C). When exposed to adenosine deaminase the activity was lost and a compound appeared with the retention time (Figure 3D and E) and mass of inosine (Figure 3F and G). We conclude that part of the antileukemic activity in the ND58 extract could be ascribed to adenosine. This was surprising in view of the high dilution of the extracts used for screening, and led us to determine the adenosine content in fresh samples of the various diatom strains. We found that while isolate ND58 contained 0.17 μg adenosine per mg dry weight, four isolates contained even more (like ND73: 0.31 μg/mg dry weight), while two isolates (ND53 and ND 71) had very little adenosine (Table 2).

In order to better assess the importance of adenosine and related metabolites for diatom toxicity, adenosine deaminase was added to the aqueous extracts and they were then rescreened for antileukemic activity (data not shown). The adenosine deaminase was active in the extract, and abolished the adenosine peak and produced an inosine peak in HPLC-chromatograms of the extracts (not shown). Deaminase treatment only decreased significantly the cytotoxic activity in extracts ND53 and ND79. This was puzzling since ND53 had very little adenosine (Table 2), and suggested that these diatoms had other adenosine deaminase-sensitive toxins. A number of such toxins have been reported in microorganisms or have been produced synthetically, and we have shown previously that some of them can induce cell death [15,16].

The LC₅₀ value of adenosine is 7 μM in IPC-81wt cells. The highest adenosine level found in our diatom strains was 0.31 μg/mg biomass DW, corresponding to 4.5 μM at the highest concentrations used in the cell assay. This is enough to produce only 20% cell death (not shown). We conclude that although diatoms can contain high levels of adenosine and other toxic adenosine deaminase sensitive metabolites, most of them owe their antileukemic activity to other types of molecules.

We next tested if the aqueous diatom extracts could modulate blood platelet activation by the thrombin agonist TRAP (thrombin receptor agonist peptide). Extract from strain ND53 activated the blood platelets. The activation was insensitive to adenosine deaminase treatment (Table 2), suggesting it was due to another compound than the one responsible for IPC cell death induction, which was deaminase sensitive (see the preceding paragraph).

The other extracts produced strong to moderate inhibition of platelet activation (Table 2). Adenosine deaminase treatment abolished the inhibitory activity only for extracts ND73 and ND79 (Table 2). Since adenosine deaminase failed to abolish the antileukemic activity of extract ND73 the platelet inhibitory and antileukemic activity in this extract presumably resided in different molecules.

The fact that diatoms previously have been only sparsely associated with production of bioactive compounds made us determine the identity of our richly bioactivity producing isolates. We employed a genetic approach by sequencing parts of the large subunit (LSU, 28S rDNA) from the isolates, and sequences were aligned together with sequences from other pennate diatom species. Since relatively few sequences from pennate diatoms are available at GenBank, we also studied the diatom morphology in detail (a thorough description of the methodology and morphology and taxonomy of the strains is given in the supplementary data). The identification of the isolates is presented in Table 2.

Diatoms were isolated by dilution series of samples from microbial biofilms on rocks, sediments and marine plants in the intertidal zone along fjords on the western part of Norway (see table 1). All cultures were grown in Conway medium prepared from filtrated seawater diluted with distilled water to 80% [18]. Diatom cultures were cultured at 20 °C in 250 mL glass cylinders with conical bottom (inner diameter 3.5 cm) and humified air mixed with CO₂ (1% CO₂ final concentration) was filtered through bacterial filters (0.45 μm) and bubbled through the cultures. The cultures were provided with 116 μmol photons m⁻² sec⁻¹ continuous light from fluorescent tubes (Phillips). Cultures were harvested in early stationary growth phase after one week by centrifugation at 900 × g for 10 min, washed twice with water and the pellets were frozen and lyophilized. The lyophilized samples were stored at −20 °C. Water was always Milli-Q quality from Millipore systems.

Between 3 and 10 mg dry diatom biomass was measured and transferred to Eppendorf tubes and added 0.4 mL MQ water. The cells were added 0.5 mL 0.45 mm glass beads and mixed vigorously by a vortex blender at seven one-min pulses and cooled on ice between the vortex pulses. The extract was separated from the glass beads, and the glass beads were washed twice with 0.6 mL MQ water. The complete destruction of the cells was confirmed by microscopy. The combined extracts and washes were pooled and extracted on ice for 1 h before centrifugation at 1600 × g for 10 min. The supernantant was the water extract. The pellet was resuspended in 0.4 mL of ice-cold 70% aqueous methanol, left on ice for 1 h, and centrifuged and washed with 70% methanol to produce the 70% methanol extract. The pellet was extracted further with 0.4 mL ice-cold 1:1 methanol-dichloromethane for 12 h in the dark. The mixture was then centrifuged, the pellet washed twice, and the supernatants were combined to give the organic extract. The extracts were dried in a vacuumed centrifuge and resuspended in either 100 μL of MQ water (aqueous extract) or 100 μL of 25/75% dimethylsulfoxide/water (v/v). The concentrations given in the tables and figures correspond to amount dry diatom biomass extracted per mL cell or blood platelet suspension (mg DW mL⁻¹).

Hepatocytes were isolated from male Wistar rats (80 to150 g) by in vitro collagenase perfusion [19,20]. They were resuspended (8.0 × 10⁵ cells mL⁻¹) in pre-gassed (5% CO₂/95% O₂) 10 mmol L⁻¹ Hepes buffer (pH 7.4) with 5 mmol L⁻¹ lactate, 5 mmol L⁻¹ pyruvate, 120 mmol L⁻¹ NaCl, 5.3 mmol L⁻¹ KCl, 0.01 mmol L⁻¹ KH₂PO₄, 1.2 mmol L⁻¹ MgSO₄, and 1.0 mM CaCl₂. For incubation with solvent (water or 25% aqueous DMSO) or extract, the cells were incubated (0.12 mL) in 48-well tissue culture dishes at 37 °C in 5% CO₂ atmosphere with rotatory shaking (120 cycles min⁻¹) for 60 min before adding fixative (2% formaldehyde in PBS, pH 7.4 with 0.5 mg mL⁻¹ of the DNA-stain Hoechst 33342).

The IPC-81 rat leukemia cells [21] were cultured at 37 °C in a 5% CO₂ atmosphere in DMEM medium supplemented with 10% horse serum. The experiments were performed as described above for the hepatocytes, except that the incubation was for 18 h.

The surface morphology was judged by differential interference microscopy and the degree of chromatin condensation by fluorescence microscopy, and the cells scored for apoptosis and necrosis as described previously [11,22]. The major criteria to judge if a cell was apoptotic were cell shrinkage, budding of the cell surface to produce protrusions that eventually detached to form apoptotic bodies, and chromatin hypercondensation. Apoptotic cells excluded Trypan blue. Necrotic cells underwent swelling and lost their ability to exclude Trypan blue.

For transmission electron microscopy the cells were fixed for 15 min in 1.5% glutaraldehyde in ice-cold Na-cacodylate buffer (pH 7.4), post-fixed for 60 min in 1% osmium tetraoxide and dehydrated by ethanol. After embedding the cells in Agar100 resin, ultra-thin sections were cut and stained in uranyl acetate and lead citrate. Sections were examined in a Philips EM 300/II electron microscope.

Human venous donor blood (144 mL) was collected into 16 mL of anticoagulant (71 mmol L⁻¹ citric acid, 85 mmol L⁻¹ Na₃-citrate, 122.1 mmol L⁻¹ glucose). Concentrated platelet-rich plasma was prepared as described previously [23] and the platelets were further separated from plasma proteins by size-exclusion on a column packed with Sepharose CL-2B gel matrix (Pharmacia Biotec, Sweden). The platelets were eluted with Ca²⁺-free Tyrode’s buffer enriched with 0.1% BSA and 5.6 mmol L⁻¹ glucose (pH 7.3), counted (Coulter Counter, Beckman Coulter UK Ltd, Bedfordshire, UK) and adjusted to 3.5 × 10⁸ platelets mL⁻¹.

For measurement of platelet inhibition or activation, 2 μL of aqueous extract were added either 10 μL PBS or 10 μL adenosine deaminase (Sigma, CA, USA) in PBS (0.9 U mL⁻¹) and incubated for 15 min before addition to 40 μL mixture of platelets and PE-conjugated anti-CD62 (BD biosciences, San Jose, CA, USA). After another 10-min incubation, 20 μmol L⁻¹ thrombin receptor agonist peptide (TRAP, SFLLRN, Biotechnology centre of Oslo, Norway) was added and the platelets were left to activate for 15 min before fixation in 0.2% paraformalaldehyde in PBS. The ratio of PE-positive platelets (expressing P-selectin on the surface) was measured by a FACSCalibur (BD Biosciences, Franklin Lakes, NJ, USA) flow cytometer as previously described [24].

Six grams of lyophilized dried diatom material from ND58 (P. tricornutum) was mixed with 350 mL of water at 4 °C, homogenized, 100 mL ice-cold water added and sonicated for 5 min in a sonication water bath, left for 2 h at 4 °C, and centrifuged at 14 000 × g for 10 min. The supernatant was subjected to anion exchange solid phase extraction (QMA 10 g, 35 cc Waters, Sep Pak®, Waters Corporation, Milford, MA, USA), conditioned with 100 mL of methanol and equilibrated with 200 mL of water (flow rate: 5–10 mL min⁻¹). Each cartridge was loaded with sample corresponding to 1.5 g diatom dry mass, and washed with 75 mL water before elution with 100% methanol. The combined flow-through and water wash, and the methanol-fraction were collected separately and evaporated.

The load and wash fraction was further fractionated by semi-preparative RP-HPLC (Kromasil 100-5 C₁₈ 250 × 10 mm column fitted with a guard column Kromasil 100-5 C₁₈ 10 × 10 mm, both from AKZO Nobel, EKA Chemicals, Bohus, Sweden), with water (A) and acetonitrile (ACN, Rathburn, UK; B) as mobile phases, the gradient was: 0 min: 96% A and 4% B, 10 min: 84% A and 16% B, 24 min: 0% A and 100% B. The column was then washed with 100% B for 3 min. The flow rate was 4.5 mL min⁻¹ and the monitoring wavelength was 238 nm. Fractions were collected every 30 s, evaporated in a vacuumed centrifuge, redissolved, and tested for cytotoxicity.

The cytotoxic fraction from the preparative column was evaporated and resuspended in 40 μL of water and injected onto an analytical C18 HPLC column (Kromasil 100-5 C₁₈ 250 × 4.6 mm) fitted with a guard column (10 × 4.6 mm). Isocratic mobile phase mode (98 water, 2% ACN) enabled full resolution of the component peaks. The active peak was collected manually and further analysed by Q-tof MS and MSMS.

The spectrometer, Micromass Q-TOF Ultima Global, interface was operated in positive ion mode, capillary voltage: 3.00 kV, cone: 70 V, source temperature: 79, desolvation temperature: 100 °C, gas flow (L h⁻¹) (a) cone: 50, (b) desolvation: 300. The purified diatom compounds and the adenosine and inosine standards were directly infused into the spectrometer at 1.0 μL min⁻¹. Collision energies between 15 and 60 V were used to obtain fragments of the nucleosides.

Aqueous extracts were prepared as described for the crude extracts, and subjected to anion exchange solid phase extraction (QMA Sep-Pak® Light, Waters). The cartridges were conditioned with 2.0 mL methanol and equilibrated with 2.0 mL water. Volumes of extract corresponding to not more than 5 mg dry weight were loaded onto the columns followed by a wash with 2.0 mL MQ water and a final elution with 1.0 mL 20% methanol.

Ninety μL of the QMA-Load+Wash fractions were injected into the analytical HPLC (instrumentation as described above). The gradient was from 95:5% water-ACN to 85:15% water-ACN in 9 min, followed by 100% ACN to wash the column. The flow rate was 0.8 mL/min. Adenosine standards (20 μL) from 0.0012 mmol L⁻¹ to 0.08 mmol L⁻¹ were chromatographed under identical conditions to create a standard curve. The adenosine peak was identified by retention time and UV-spectrum, and the area under the peak was used to determine the adenosine concentration.

For the stereo-specific enzymatic degradation of adenosine four units of adenosine deaminase were added to adenosine (1 mM) or the bioactive 11.9-min peak (Figure 3) to a final volume of 0.1 mL, and the mixtures allowed to react for 30 s before stopping the reaction by adding 50 μL of ACN. The product was added to either adenosine or Apo-58-11, and the mixtures were chromatographed.

Partial sequences from genomic DNA of LSU (28S) from the diatom isolates have been deposited in GenBank (at http://www.ncbi.nlm.nih.gov/) with the following accession numbers: FJ214161(ND51); FJ214162(ND52); FJ214163(ND53); FJ214164(ND55); FJ214165(ND56); FJ214166(ND58); FJ214167(ND 71); FJ214168(ND73); FJ214169(ND79).

Large batch cultures were grown in Conway medium [18] prepared from filtrated seawater diluted with distilled water to 80% at 20 °C. Humified air mixed with pure CO₂ to obtain 1% CO₂ was filtered (0.45 μm) and bubbled through the cultures, and a magnetic stirrer at the bottom prevented sedimentation of the cells. The 10 L culture vessels were polycarbonate cylinders (height 110 cm, inner diameter 11 cm). Cultures were provided with 250 μmol photons m⁻² sec⁻¹ continuous light from fluorescent tubes (Phillips). Scalar irradiance was measured at the front of the cultures with a spherical sensor (Biosperical Instruments Inc. QSL-100). Cultures were harvested in early stationary growth phase after one week by continuous centrifugation at 1,000 rpm, washed with MQ water (MilliQ quality, Millipore system, Millipore Billerica, MA), frozen and lyophilized. Approximately 6 g wet weight was typically obtained from 10 L cultures.