The organic crude extracts from the 11 blue colonies analyzed showed high inhibitory activity against breast SKBR3, colorectal H-116, lung A-549, and pancreas PSN-1 cancer cell lines and displayed essentially no antiproliferative activity against the glioblastoma T98G cancer cell line (Table 1). Previous studies proved that the ascidian showed cytotoxicity against HL-60 and P338 leukemic cells, and the MCF7 breast cancer cell line [25, 27, 35, 36]. Morover, when the blue ascidians were kept in aquarium for up to 75 days, the antiproliferative activities were conserved with high levels, (IC₅₀ <5 μg/ml in A-549, H-116, PSN-1 and SKBR3 cancer lines).

In good agreement with toxicity data from the larvae reported by Tarjuelo et al. [31], we found that larva crude extracts did not display antiproliferative activity against human lung carcinoma A549 or colon adenocarcinoma H116. The lack of antitumor activity in larva extracts could be due to the absence of cytotoxic compounds in the tunic tissue of the larvae. Adults and larvae of a given species may use distinct chemical and physical defence strategies [31, 37–41] and unpalatable larvae do not always come from unpalatable adults [39]. Larval defence could thus be associated with the reproductive strategy of the species rather than the chemical and physical defence mechanisms [37, 38, 40, 42].

On the other hand, our results showed a significative difference between antiproliferative activities between green and blue ascidian colonies. The cytotoxic activity detected in the green colonies was very low, IC₅₀ ≈ 25 μg/mL, and only the blue pigmented colonies showed high values. Within the population of C. dellechiajei in the sampling area, the blue pigmented colonies were much more abundant (around 3 fold) than the green colonies. It could thus be possible that the high cytotoxicty of the blue colonies provide them with an adaptative advantage compared to green colonies.

Although we have not analyzed the chemical nature of the active compounds, it has been widely reported that the main compounds extracted from C. dellechiajei are pyridoacridine alkaloids. These compounds are ascididemins, 11-hydroxyascididemin, cystodytins A-I, shermilamine B, kuanoniamine D, and sebastianines A and B [20, 22, 25, 28, 32, 35, 43–45], some of which showed antitumor properties [25–29]. However, only ascididemin, 11-hydroxyascididemin and sebastianines A and B have been detected in the blue colonies [20, 21, 29]. Therefore, the cytotoxic activities we have observed could be due to these products. In other hand, Rottmayr et al. [20] and Turon et al. [21] reported that the pyridoacridine compounds were located in the pigmented cells of the tunic tissue in purple color colony, and, in addition, López-Legentil et al. [46] showed seasonal variations in the production of these compounds, with minimum values in summer, attributable to sexual exhaustion and seasonally varying biotic interaction or abiotic parameters. Therefore, according to the data from the location of these compounds inside the ascidian tissues [20, 21], our results provided from the different parts of the ascidian analyzed, zooid, tunic and larva, suggest that the cytotoxic compounds could be also located inside the pigmented cells of the tunic tissue.

Antibiotic treatment did not produce any apparent detrimental effect on the ascidian growth, as no morphology and/or pigmentation differences were observed between treated and untreated colonies. In vitro antitumor activities were very similar in treated and control colonies in triplicate samples. For both control and treated colonies, the antitumor activity remained constant during the experiment (Figure 1), showing a range of IC₅₀ values from 4 to 6 μg/mL, except for the control sample after 2 weeks, which showed a higher value, IC₅₀ ≈ 8 μg/mL against A-549 cell line, since one specimen of the three replicates samples at that time displayed a low antiproliferative activity. Although, as shown in Figure 1, treated samples during two weeks showed the IC₅₀ value lower than the control samples.

The data provided by DGGE analysis showed that bacterial communities associated with sample controls (sample C0, before starting the experiment, and samples C1 and C4, cultured during one and four weeks respectively) were similar (above 95% similarity, Figure 2), indicating that the community underwent few changes when the tunicate was maintained out of its natural environment.

After 4 weeks of antibiotic treatment (T4), the bacterial community in the tunicate had changed remarkably, and displayed 75% similarity (Figure 2b) to the rest of the samples, which were all above 95% similar to each other. T4 DGGE pattern displayed changes in the intensity of some bands present in the rest of the samples (2, 3, 4 and 5 in Figure 2a) as well as some new bands (bands 1, 6 and 7).

These bands were reamplified and sequenced, and they were found to be related to the bacterial 16S rRNA gene sequences shown in Table 2. The uppermost gel bands belong to 16S rRNA sequences of chloroplast of photosynthetic epibionts of C. dellechiajei, like diatoms and rhodophyte algae [33]. The rest of bands corresponded to 16S rRNA sequences of bacteria. Most of the DGGE bands were present in all samples without significative variation, representing bacteria that were either resistant or inaccesible to gentamicine and bacitracine at the assayed doses. However, the treatment favoured the growth of new bacteria related to the alpha-proteobacterium Stappia sp. and beta-proteobacterium Janthinobacterium agaricidamnosus (DGGE bands bands 1, 6 and 7 in Figure 2a) and other alpha-proteobacteria that were previuolsy present in the ascidian tissues (bands 2 and 3). Finally, only bacteria related to Mesorhizobium sp. (bands 4 and 5 in Figure 2a) were directly sensitive to the antibiotic treatment. Since treated colonies had stable and high antitumor activity, those bacteria related to Mesorhizobium sp. (bands 4 and 5), Stappia sp. (bands 6 and 7) and Janthinobacterium agaricidamnosus (band 1) were most likely not involved in the synthesis of cytotoxic compounds.

On the other hand, as expected, the antibiotic treatment did not affect to archaeal community associated to C. dellechiajei. As shown in Figure 3, treated and control samples displayed the same DGGE pattern.

A total of ninety-four isolates were obtained, and twenty-seven of them were randomly selected and analyzed by 16S rRNA gene PCR amplification and sequencing as described above. All the isolates were Gram negative. A predominant (85% of the selected colonies) colonial morphotype (number 1, see Table 3) was observed.

These bacteria formed irregular highly mucous colonies with white-yellow pigmentation, and were easily distinguishable from the other colonies. However, inside this morphotype, five different ARDRA patterns were found. 16S rRNA gene sequences analysis revealed that the isolates Dell4, Dell5, Dell6, Dell7, and Dell8 (representatives of each ARDRA pattern) were related to alpha-proteobacteria associated with invertebrate marine organisms (Figure 4a), like strain NW001 associated with the sponge Rhopaloeides odorabile [47], common throughout the Great Barrier Reef, strain SB89 isolated from the Mediterranean sponge Aplisina aerophoba [48], strains MBIC3368 and JE064 isolated from sponges [49, 50] and strain Z143-1 isolated from a Philippine tunicate [51].

Strains Dell 4, Dell 5 and Dell 8 displayed >98.2% partial rRNA gene sequence similarity to their closest cultured relatives, and Dell 6 and Dell 7 showed 96.1 and 91% respectively. The second morphotype was represented by Dell 9, which formed irregular colonies without pigmentation and was related to Erythrobacter litoralis (93.5% complete 16S rRNA sequence similarity), an aerobic anoxygenic phototroph, previously cultured from open surface waters [52]. The rest of the three strains isolated, Dell 1, Dell 2 and Dell 3, were related to gamma-proteobacteria (Figure 4b). Strain Dell 1, which showed pleomorphic forms in pure culture as well as cocci and rods, was related to the gamma-proteobacterium strain C1, isolated from C. dellechiajei [53, 54]. Strain Dell 2 was clustered with Halomonas sp. group (97.6% similarity with the closest specie, H. salina). Dell 2 colonies were circular, convex and sligthly opaque with smooth edges. Strain Dell 3, which formed circular, smooth, raised and cream-colored colonies, was related to Alteromonas macleodi (98.9% similarity). 16S rRNA gene clone sequences related to strains Dell1 (97.3% similarity) and Dell9 (99% similarity) have been found in the bacterial community associated with tunic tissues of C. dellechiajei using a culture independent approach [33].

All the ninety-four isolates were screened for antitumor compounds. None of the strains showed in vitro antitumor activity against breast SKBR3, colorectal H-116, pancreas PSN-1, glioblastoma T98G or lung A-549 cancer cell lines. Interestingly, phylogenetic analysis (Figure 4) showed that the isolates Dell 4, 5, 6, and 8 were closely related to the strains alpha-proteobacteria SB89 and Z143-1 associated with sponges and tunicates that produce secondary metabolites with antimicrobial activity against Staphylococcus aureus and Gram negative and positive reference strains [51]. Although antimicrobial activities were not analysed, since this was out of the scope of this study, others authors described cytotoxic compounds with antimicrobial properties extracted from C. dellechiajei [36, 55]. According to the close relationship between isolates Dell 4, 5, 6 and 8 with alpha-proteobacteria that produce antimicrobial products, these antibacterial activities detected in the ascidian extracts could be due to secondary metabolites produced by the associated bacteria.

Turon et al. [21] and Rottmayr et al. [20] showed that in C. dellechiajei cytotoxic pyridoacridine alkaloids involved in chemical defense were stored in pigmented cells of the tunic, with maximum abundance in upper tunic zones. In addition, Turon et al. [21], detected signals corresponding to the pyridoacridine alkaloids shermilamine B and kuanoniamine D and their deacetylated from the compartments of the ascidian tissues, mainly pigmented cells, but no signals from the associated bacteria, by X-ray spectra. Therefore, these authors pointed out that the host produces these cytotoxic compounds, and not the associated bacteria. In order to clarify this point, we compared the microbial communities associated to the blue and green colonies by DGGE analysis in order to study the differences in the composition of the communities and correlate these data with antiproliferative results against tumor cells obtained from the two colour morphs extracts. As shown in figure 5, the DGGE patterns between the blue and green colonies was very similar. However, the green colonies did not show antiproliferative activity against tumor cells (see Table 1). Since the microbial communities associated to green and blue colonies are very similar and only the blue colonies show cytotoxicity, it does not seem probably that the associated bacteria are the source of the cytotoxic compounds nor they have an essential role in the production. These data, together with the lack of the activity in the isolated bacteria extracts and the results obtained from the antibiotic treatments of the ascidian, and data from Rottmayr et al. [20] and Turon et al. [21] do not indicate a microbial origin for the compounds.

Specimens of the two varieties, blue and green, of the ascidian C. dellechiajei were collected by scuba diving to depths of 2–5 m in the Mediterranean sea (Cape Palos, Murcia, Spain). Samples for in vitro antitumor activity assays were collected in May 2002 and March 2003. Colonies for antibiotic treatment were harvested in October 2003. Colonies and seawater were transferred directly to a container with an autonomous aeration pump, and immediately brougth back to the laboratory. Samples were processed immediately for antiproliferative assays or cultured in aquarium for antibiotic treatment. Larvae were obtained from ripe colonies under aseptic conditions, rinsed three times in sterile seawater and processed for antiproliferative assays.

Twenty four adult colonies were collected and maintained in a seawater aquarium for one month. Half of the colonies were treated every day with gentamicine sulfate (Sigma, 100 mg) and bacitracine (Sigma, 100 mg) per litre of seawater, and the other half were used as controls without antibiotic treatment. Antibiotics active against Bacteria that did not cause detrimental effects against eukaryotic (i.e. ascidian) cells were chosen, according to their action spectra [60]. Antibiotic concentration was adjusted according to Davidson et al. [18]. The aquarium water was changed every day. At the end of each week, triplicates of treated and control colonies were collected from the aquarium and used for antiproliferative activity measurements and DGGE analysis. Activity was measured for each of the three replicates while DGGE was carried out with only one of them. Prior to the experiment of the antibiotic treatment, five adult colonies were maintained in a seawater aquarium for seventy-five days, then collected and screened for antiproliferative activity in order to study the effect of the artificial maintenance of the colonies on the production of cytotoxic compounds.

Colonies from the antibiotic treatment experiment were dissected into tunic and zooids under aseptic conditions and rinsed three times in sterile seawater. Prior to DNA extraction, tunic tissues (1.3 g) were frozen and homogenized in liquid N₂ with a sterile mortar. Pulverised tissue was suspended in 25 ml TE buffer (10 mM Tris-HCl and 1 mM EDTA, pH 8) with 0.5% SDS. DNA was purified by phenol-chloroform-isoamyl alcohol 25:24:1 (all Sigma) and ethanol precipitated essentially as described by Sambrook et al. [61]. The DNA pellet was air-dried at room temperature, resuspended in ultrapure sterile water (200 μL) and stored at −80 ºC. PCR amplification of partial 16s rRNA gene for Bacteria was performed using primers 341-GC and 907R [34], and 344-GC and 907R primers for Archaea [62]. The PCR conditions for Bacteria were as follows: 5 min at 94 ºC, 1 min at 65ºC, 3 min at 72ºC, 10 cycles of 1 min at 94ºC, 1 min at 64ºC, 3 min at 72ºC, 20 cycles of 1 min at 94ºC, 1 min at 55ºC, 3 min at 72ºC, with a final extension step of 72ºC for 10 min in a PTC-100 (MJ Instruments). The PCR conditions for Archaea were as follows: 94°C for 5 min (initial denaturation), and 30 cycles of 94°C for 30 s, 56 °C for 45 s, 72 °C for 2 min. The length of extension step was increased to 10 min in the last cycle.

Each 50 μL reaction sample contained 1.5 mM MgCl₂, 10 mM Tris-HCl, 50 mM KCl, 200 μM of each dNTP, 1 U Taq I DNA polymerase (Invitrogen), 0.25 μM of each primer and 100 ng of DNA. The PCR products were quantified by comparison with a Low DNA Mass Ladder (Invitrogen) using agarose gel electrophoresis. DGGE analysis was carried out using a DCODE system™ (Universal mutation detection system, Bio-Rad) with a denaturing gradient of 45–60% for Bacteria and 40–55% for Archaea (100% was defined as 7 M urea and 40% deionized formamide) in a 0.75 mm-thick 6% polyacrylamide gel. Approximately 600–800 ng of PCR product were applied to each lane in the gel. Electrophoresis was performed at 5 V/cm at 60ºC for 16 h in 1X TAE buffer (40 mM Tris, 20 mM sodium acetate, 1 mM EDTA, pH 7.4). DGGE gels were stained with SYBR Gold (Molecular Probes), 100 μL/L in 0.5X TAE buffer, for 15 min, rinsed with 1X TAE buffer for 15 min, visualized and photographed with a Typhoon 9410 image analysis system (Amersham Biosciences). Selected DGGE bands were excised with a sterile blade and incubated overnight at 4ºC in ultrapure sterile water (20 μL). The eluent was used as template DNA for reamplification with the primers and conditions described above. PCR products were sequenced using primer 907R in a genetic analyzer ABI PRISM 310 (Applied Biosystem).

A similarity dendrogram of the DGGE patterns from the antibiotic treatment experiment was computed using SPSS® 12.0 software (SPPS, Inc., Chicago, IL) with the Jaccard index (shared characters/total characters).

All culture media contained an ascidian organic extract (AOE) prepared as follows: a sample of the the colony (2 g) was homogenized using a mortar and boiled in filtered sterile seawater (35 mL) for 15 min. The homogenized colony was centrifuged at low-speed (2,000 rpm at room temperature, Labofuge 400R, Heraeus) for 1 min and the resulting supernatant constituted the AOE. Three different solid (2% agar) media were prepared with 12% v/v of AOE in filtered seawater collected at the sampling area. Medium 1 did not contain additional components, medium 2 was supplemented with 0.001% (p/v) yeast extract, and medium 3, with 0.001% (p/v) yeast extract and 1% (p/v) sucrose. The pH of the media was adjusted to 8 and they were sterilized at 121 ºC for 15 min. The plates were inoculated by either spreading a small piece of tunic (1–2 mm³) or homogenized tunic (100 μL, 2 g of tunic in 1 mL of filtered and sterile seawater). Cultures were incubated at room temperature under aerobic conditions for at least two weeks, and representatives of each bacterial colonial morphotype were serially streak-plated until pure cultures were obtained. Bacterial colonies were then suspended in TE buffer (400 μL, 10 mM Tris-HCl and 1 mM EDTA, pH 8) with 0.5 % SDS. DNA extraction and purification were carried out as described previously. PCR amplification of bacterial 16S rRNA genes was performed using primers 27f and 1492r [63]. Cycling conditions were as follows: 3 min at 94ºC, 30 cycles of 15 sec at 94ºC, 30 sec at 55ºC, 2 min at 72ºC and a final extension of 10 min at 72ºC. PCR products of bacterial 16S rRNA were analyzed using ARDRA [64] with the enzymes HinfI and MboI (Invitrogen) according to the manufacture's protocol. The digestion products were analyzed in 2% agarose gel (LE, FMC Bioproducts) in 0.5X Tris-boric acid-EDTA buffer. 16S rRNA genes from each restriction pattern were sequenced using primers B1055, 16S5r, 338f and 785f [65].

16S rRNA gene sequences were initially compared with the reference sequences at NCBI (http://www.ncbi.nlm.nih.gov) using BLASTn [66]. The correlation between primary sequence and secondary structure was analyzed using the ARB sequence editor. Complete and partial sequences obtained in this study were added to an alignment of over 50.000 primary aligned gene structures available at http://db-central.arb-home.de/. Sequences not included in the ARB database were obtained from the GenBank database. Phylogenetic analysis was performed by using the three algorithms implemented in the ARB package: maximum-likelihood, neighbour joining using Jukes-Cantor correction and maximum parsimony. Phylogenetic trees calculated by neighbour-joining were evaluated after 1.000 bootstrap resampling of the data. In addition, the TREE-PUZZLE program of the ARB package was used to reconstruct phylogenetic trees by maximum-likelihood with quartet puzzling support values for each internal branch [67, 68].

For in vitro antitumor assays, samples of whole colonies (2 g), tunic tissue (1.8 g), zooids (100 mg) and larvae (50 mg) were homogenyzed separately in methanol-acetone (10 mL, 1:1 v/v) with a mortar and centrifuged at 4,500 rpm (Labofuge 400R, HERAEUS) for 1 min. The resulting supernatants were dried under vacuum with a rotary evaporator (Laborata 4000, Heidolph, France) at room temperature and resuspended in ultrapure sterile water at different final concentrations (50, 25, 10, 5 and 2.5 μg/mL).

Microbial cultures were grown in liquid media designed to favour the synthesis of the secondary metabolites (C. Acebal, personal communication). Media 1, 2 and 3 were supplemented with pharmamedia® (1.5% p/v), soya flour (0.8% p/v), corn step (0.8% p/v), yeast extract (0.2 % p/v), MgSO₄ 7H₂O (5% p/v), and CaCO₃ (0.3% p/v). pH was adjusted at 7.1. All media were incubated at 30ºC, 220 rpm for 4 days. Once grown, each culture (5 mL) was lyophilized, resuspended and homogenized in methanol-acetone (10 mL, 1:1 v/v) and centrifuged at 4,500 rpm (Labofuge 400R, HERAEUS) for 5 min. The resulting supernant was dried and resuspended at different final concentrations (50, 25, 10 and 5 μg/mL).

All the tumor cell lines were obtained from the American Type Culture Collection (ATCC). Human lung carcinoma A549, colon adenocarcinoma H116, breast carcinoma SKBR3 and pancreatic adenocarcinoma PSN1 were cultured in RPMI medium [69] containing 2 mM glutamine, 50 IU/ml penicillin, 50 μg/ml streptomycin. Media were supplemented with 5% fetal bovine serum (FBS) for A549 and H116 carcinomes, and 10% FBS for SKBR3 and PSN1 carcinomes. Caucasian glioblastoma T98G was maintained in RPMI 1640 medium [69] containing 2mM glutamine, 50 IU/mL penicillin, 50 μg/mL streptomycin, 1 mM pyruvate supplemented with 0.1 mM nonessential amino acids, and 10% FBS.

The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT, Sigma Chemical Co., St. Louis, MO) dye reduction assay in 96-well microplates was used, essentially as previously described [70]. Tumor cells (4x10³ A-549 cells or 6x10³ H-116, 6x10³ PSN1, 6x10³ SKBR3 and 6x10³ T98G cells in a total volume of 200 μL of complete medium) were incubated in each well with different amounts of the ascidian and bacterial extracts (see above). After 2 days of incubation (37° C, 5% CO₂ in a humid atmosphere), MTT (50 μL, 5 mg/mL in PBS) were added to each well and the plate was incubated for a further 2 h (37ºC). The resulting formazan was dissolved in DMSO (100 μL) and the absorbance read at 490 nm. All determinations were carried out in triplicate. IC₅₀ was calculated as the concentration of drug yielding a 50% cell survival rate. The cytotoxicity assays were performed in two stages: first the extracts were tested for activity against human lung carcinoma A549 and colon adenocarcinoma H116. Only active extracts were further analyzed for activity against breast carcinoma SKBR3, pancreatic adenocarcinoma PSN1 and caucasian glioblastoma T98G.