Algal strains intended for biotechnological applications need to be produced as fast as possible in large-scale systems in order to ensure a sustainable process. Therefore microalgae displaying high growth rates are essential. In this context, water samples were supplemented with Algal growth medium and incubated for 7 days using growth conditions as described in Methods. This enrichment step facilitates the isolation procedure, allowing fast-growing strains to dominate, by competition, other microalgae of less interest under a set of desired growth conditions. If the enrichment step is omitted, fast-growing cells might be overlooked during the isolation procedure due to their low concentration. Thus, this stage is a key step in the selection of microalgae isolated by the present method.

FACS allows simultaneous measurements of individual cells. Light scatter angles and fluorescence intensity measured in different channels allows the distinction between different clusters of cells with heterogeneous characteristics. The information given by the different channels can be visualized in two-dimensional plots combining two of the variables at each time. Figure 1 shows the result of combining side scattering (SSC) with fluorescence in the FL3 channel. The combination of these two variables enabled us to separate cells with chlorophyll from unwanted cells and debris, mainly non-photosynthetic bacteria and sediments that are normally present in natural samples. This could be accomplished as flow cytometry allows the separation of cells by the inner cell complexity and endogenous fluorescence of pigments naturally occurring in microalgae. The first property can be estimated by the measurement of side angle light scatter (SSC) 22 , whereas fluorescence emitted by pigments (for example, chlorophyll, phycoerythrin and phycocyanin) can be detected in different channels upon excitation with a specific laser 19 23 . In this way, red fluorescence correlates with cellular chlorophyll content 18 19 24 , which was measured in the FL3 channel. At this stage, all cells present in the dot plot displaying more than 100 arbitrary units (AU) of autofluorescence, as detected in the latter channel, were gated; this first gate was named Chl.

The possibility of obtaining axenic cultures is one of the main advantages of the FACS approach, since it has been shown that the latter procedure is able to remove bacteria from microalgal cultures 25 . After this initial gating, several combinations of the signal obtained with the different channels were used. From all the combined plots the best separation between populations was obtained upon combining the FL2 and FL4 channels (Figure 2). During this acquisition, well resolved clusters of cells were clearly distinguished, which simplified the gating procedure. In this way, five gates (P1, P2, P3, P4 and P5) were drawn. The signal measured in the FL4 channel is associated with phycocyanin-derived autofluorescence 19 , while yellow-orange fluorescence detected in the FL2 channel is related with phycoerythrin-derived autofluorescence of cells 18 19 . In addition to measuring endogenous fluorescence, cells can also be stained with different dyes, providing a wide range of information. It has recently been shown that BODIPY is able to stain lipid bodies in photosynthetic unicellular organisms, allowing the establishment of lipid-rich microalgal cultures 26 , and also the development of selection programs to improve lipid production as demonstrated with the Nile red dye 14 15 . Although BODIPY fluorescence can be detected in the FL2 channel, the highest emission intensity of this fluorochrome is in the green fluorescence bandwidth range. Hence, cell complexity (SSC) was subsequently plotted against the fluorescence emission in the FL1 channel (centered at 530/30 nm) in addition to the FL2 channel. Results obtained using these settings suggested that each cluster was a different species (Figure 3). In this figure, each gated cluster was represented by a different color in order to distinguish clusters gated in Figure 2. Interestingly, the most discriminating factor for clusters P1, P2, P3 and P4 was the fluorescence associated with the FL1 and FL2 channels (Figure 3A, B) rather than inner cell complexity (SSC) or chlorophyll fluorescence (Figure 1). The combination of BODIPY fluorescence emission with cell complexity is shown in Figure 3A. The P4 cluster displayed the highest BODIPY fluorescence and was thus selected as a good candidate source for microalgal cells with potential use for biodiesel production. P1 and P3 gates are clustered together and cannot be resolved by this combination of signals. Figure 3B combines signals from SSC and FL2 channels in which P1 and P3 clusters are resolved, as the former cluster moved to the right with respect to the P3 cluster due to a higher fluorescence emission in the FL2 channel. That may be explained by higher phycoerythrin content present in the P1 cluster cells, since the fluorescence detected in the FL1 channel was very low. Fluorescence emission detected in the FL2 channel was in fact one of the best discriminant factors for these strains. Figure 4 presents a histogram relating event (cell) density with PE fluorescence. This graph showed that well defined cell populations with distinct PE fluorescence levels could be observed and that the P4 and P5 clusters displayed the highest cell density. The P5 cluster showed significantly lower complexity and fluorescence as compared with other clusters. This low degree of complexity and fluorescence emission indicated that the cells in the latter cluster were most probably cyanobacteria 18 27 .

Although BODIPY, phycoerythrin and phycocyanin fluorescence were used in the strain isolation for this specific environmental sample, different combinations of channels can be chosen to produce a similar two-dimensional plot as shown in Figure 2. In this particular case, FL2 and FL4 produced the best clustering for the chosen gating method. Though the other channels gave relevant information, they were not used in the procedure for separation of the different naturally occurring cell populations since they formed overlapped clusters. The use of fluorescence emission due to BODIPY staining in the gating procedure should be done with care since the concentration of lipids in the cells varies with the culture growth stage. Cells normally present higher lipid concentrations during stationary phase as compared with cells growing exponentially. As in the present work environmental samples underwent a pre-enrichment step to isolate fast-growing cells able to withstand competition from other microalgae, it is possible that lipid-rich strains growing actively might have been overlooked.

To assess cell morphology, the gated clusters were then sorted directly to microscope slides and observed in a Zeiss Axio Imager Z2 fluorescence microscope. Cells from the P1 cluster did not survive the cell sorting procedure, since upon microscopic observation only disrupted cells were found. The same problem was described by Reckermann 19 with Fibrocapsa japonica that disrupted soon after sorting. Microscopic observations suggested that the established gates were able to isolate monoalgal cultures since all observed fields from the same cluster showed the same strain. As expected, P5 was an exception and microscopic observation showed different cyanobacterial strains and debris.

Microscopic observation of cells from the P4 cluster, which presented the highest levels of fluorescence in the FL1 channel, confirmed the presence of several lipid bodies. As can be observed in Figure 5, several dots of BODIPY fluorescence were detected (Figure 5B, D), confirming that the dye was internalized into the cells, as it can be seen in the merged differential interference contrast (DIC) system plus BODIPY image (Figure 5A, C). This observation strongly supports our FACS results, suggesting that the combination of the gates used was efficient in the isolation and identification of lipid-rich microalgae that can be later tested for biodiesel production (Figure 5).

Growth was visible in solid medium in approximately 70% of the wells after 2 weeks of incubation (Figure 6), whereas 20% did not show any sign of bacterial growth (Figure 6A). Absence or presence of bacteria was later confirmed by means of PCR amplification of bacterial 16S ribosomal DNA (Figure 7) using primers as described by Weisburg et al. 28 . In liquid medium, however, after 2 weeks only 45% of wells presented algal growth and almost 95% of these displayed visible bacterial growth. However, in this medium species that did not grow on solid medium were successfully isolated. Cells from the P3 cluster did not grow on solid growth medium, although with some difficulty these microalgae were able to proliferate in liquid medium. In fact, some strains and even entire algal taxa (for example, dinoflagellates) hardly grow on agar growth medium 12 . However, isolation in solid medium showed obvious advantages, such as (1) faster biomass growth at the early stages of the scale-up process; (2) easy application and maintenance of cultures, since in liquid medium microalgae need to be kept afloat to promote growth efficiently; and (3) bacterial growth is detected more easily.

The efficiency of the sorting procedure obtained in this work (70%) was considerably higher than that reported by Sensen et al. 25 for the removal of bacteria from unialgal cultures (20% to 30%). Global recovery, however, was lower than reported by Doan et al. 20 in novel isolates from an environmental water sample (82% to 100%). Still, sorting efficiency is highly dependent on the original sample, including the starting group of species and their abundance, thus demonstrating the importance of the initial enrichment step. For instance, Sinigalliano et al. 21 compared the isolation efficiency of dinoflagellates between FACS and manual picking using a micropipette and the former approach appeared to be less efficient than the latter (0.5% vs 2%, respectively). This could be due to the frailness of algae belonging to this taxonomic group. The enrichment step can also lead to a selective enrichment of some species. In our work, Algal growth medium, which contains low levels of silica, was used 29 . Therefore, domination of mixed cultures by diatoms and other silicate-requiring algae was not favored during the enrichment process.

To ensure algal growth and estimate the best starting number of cells for an expedited recovery, different numbers of events (cell-containing droplets) were tested. For each well 1, 2, 10 or 20 events were sorted. Interestingly, the starting number of cells per well was not a constraint in the scale-up procedure on solid medium. Regardless of the number or events sorted per well, visual growth was obtained after a 2-week incubation period. Single cell sorting in solid medium (one event) yielded a single axenic colony in most wells (Figure 6A), which was easily transferred onto a Petri dish with solid medium. Wells with > 1 event (2, 10, 20 events) gave rise to a large number of colonies; however, all of them presented visible bacterial growth (Figure 6B-D).

In liquid medium, after a 2-week incubation period algal growth was only visible in the wells with 10 or 20 events. Most of the obtained cultures were not unialgal and showed abundant bacterial contamination as verified by microscopic observation (data not shown). Algal growth in wells with one or two events was only visible after 4 weeks. A few of these cultures were, however, axenic, though at a very low frequency (5%). Therefore, in liquid medium, increased culture growth can be achieved by sorting more events, which results in a faster scale-up, but the use of several replicates is recommended when sorting to liquid medium to isolate axenic cultures as this improves the scale-up procedure efficiency. Overall, the scale-up procedure was considerably faster in solid medium (4 weeks) than the required period for achieving the same growth and culture volume using liquid medium (7 weeks).

The fact that BODIPY cell staining did not affect cell recovery in both solid and liquid medium is essential for present and future work in the detection and isolation of lipid hyper-producing algae. Although it has already been reported that culturing cells after Nile red staining is possible 14 15 , BODIPY presents several advantages when compared with Nile red: (1) BODIPY is able to stain a wide range of algae groups without the need of using high concentrations of dimethylsulfoxide (DMSO) or acetone to carry the dye in, which can be key to wider and faster cell recovery and scale-up; and (2) BODIPY preferentially traces lipid bodies instead of other cytoplasmic compartments 26 .

Although light and electron microscopy have been traditionally used to identify microalgae, small subunit ribosomal RNA gene (16S or 18S rDNA) sequences are considered as an ideal tool in the identification of microorganisms, including microalgae 30 31 32 . Thus, after the scale-up procedure, isolated strains were identified by microscopy and amplification with specific primers by PCR of the 18S rDNA sequences. All cyanobacteria were kindly identified by the microbiology company AquaExam by means of microscopy. Isolates belong to three different phyla (Table 1). Strains P4 and P2 are chlorophytes, Tetraselmis sp. and Nannochlorum sp., respectively. The P3 gate corresponded to an unclassified chrysophyte. From the P5 gate one cyanobacterial strain of the genus Synechococcus was successfully isolated. The monoalgal status of the cultures was ensured by repeated subculturing from isolated colonies followed by 18S rDNA sequencing, which yielded sequences 100% identical when comparing different colonies of a given strain.

Upon its optimization, this method, coupling fluorescence activated cell sorting with BODIPY staining, was efficiently applied to several water samples collected in different marine habitats, namely salt marsh inlets (Ria Formosa, Almargem) and former salt ponds near the estuary of the Guadiana river used nowadays for fish farming (Atlantik Fish), which resulted in the isolation of 56 strains. From these, 12 fast-growing strains were selected, isolated and successfully scaled up to higher volumes (Table 2). The majority of strains isolated by this method belong to the phylum Chlorophyta (approximately 70%), Nannochloris and Tetraselmis being the most abundant genera. During this work, only 10% of the isolated strains were diatoms due to the unfavorable conditions of the initial enrichment step of the water samples. Accordingly, only two diatom strains (phylum Heterokontophyta) were successfully scaled up: Nitzschia sp. and Cyclotella sp. Cyanobacteria was the second most abundant phylum isolated by this method, representing 20% of the total isolates and included Pseudanabaena sp., Leptolyngbya sp. and Synechococcus sp. In order to isolate more cyanobacterial strains a different gating method will be needed.

In order to confirm the assessment of lipid-rich microalgae by BODIPY staining (Figure 3A, B), the growth curve and total lipid contents of the Nannochlorum sp. P2 and Tetraselmis sp. P4 strains were determined (Figure 8). Both strains presented high growth rate, reaching the exponential growth stage in approximately 10 days and cell concentrations of about 2.5 × 10⁸ (P2) and 4.0 × 10⁶ cells/ml (P4) in the stationary phase. The disparity in number of cells at the stationary phase between the two strains is most probably due to low cell size of Nannochlorum sp. P2 (approximately 2 μm) when compared to Tetraselmis sp. (approximately 9 μm). Total lipid concentration was significantly higher (P < 0.01) in Tetraselmis sp. P4 (25.4% of dry weight) as compared to Nannochlorum sp. P2 (18.1% of the dry weight). These results confirm that cells derived from the P4 cluster accumulate higher amounts of lipids as anticipated by the fluorescence measurements obtained during FACS.

Water sampling was performed in aquaculture ponds at the facilities of Atlantik Fish SA on the south-eastern coast of Portugal. In every sampling spot the water was collected and stored in 1 l bottles and kept at room temperature. At a later stage, the water samples were transferred to 80 ml test tubes containing liquid Algal growth medium, which was prepared with sterile seawater supplemented with an Algal stock solution concentrated 1,000 × 29 . Cultures were kept in an incubator for 7 days at 21°C, with a 12:12 h dark/light photoperiod, at a photon flux density of 80 μmol/m²/s. Growth on solid medium was carried out in either Petri dishes or 96-well plates containing Algal growth medium solidified with 1.5% agar.

Lipid bodies were stained with a solvatochromic fluorochrome, BODIPY 505/515 (Life Technologies Europe BV, Porto, Portugal), as described by Cooper et al. 26 . Cells were stained with a 1 mM aqueous solution of BODIPY dissolved in DMSO (0.2%) to attain a final concentration of 1 μM. Upon addition of the fluorochrome, tubes were vortexed for 1 minute and incubated at room temperature in darkness for 10 minutes.

The flow cytometer used in our studies was a Becton Dickinson FACS Aria II (BD Biosciences, Erembodegem, Belgium). Fluorescence readings were performed by excitation with a blue and red laser (488 and 633 nm, respectively). The emission signal was measured in three channels upon excitation with the blue laser: FL1 channel centered at 530/30 nm; FL2 centered at 585/42 nm; and FL3 centered at 695/40 nm. A fourth channel, FL4, registered the emission at 660/20 nm after excitation with the red laser.

Samples were acquired with the software FACSDiva version 6.1.3 (BD Biosciences, Erembodegem, Belgium). After the acquisition of samples, images were treated with the analysis software, Infinicyt 1.5.0 (Cytognos S.L., Santa Marta de Tormes, Spain).

The settings and compensations of all channels and lasers were the same for all sorting procedures. The flow cytometry sheath fluid used in all experiments was sterile filtered seawater. Filters (PALL) used had a pore size of 0.2 μm. Sorting was performed at 2,000 events/s flow rate using 'single cell' sort precision mode, with a 100 μm nozzle.

Cells were sorted directly into wells of 96-well plates containing 250 μl of either liquid or solid (agar) Algal growth medium. In order to assess the best number of cells needed to achieve visible culture growth in a feasible time, sorting conditions were set in order to direct 1, 2, 10 or 20 droplets into each well. After the cell sorting procedure, cells were incubated for 2 weeks in the same incubator and growth conditions as described above.

Microscopic images were acquired in a Zeiss AXIOMAGER Z2 microscope, with a coollSNApHQ2 camera and AxioVision software version 4.8 (Carl Zeiss MicroImaging GmbH, Gõttingen, Germany), using the 100 × lens. For the fluorescence images, we used Zeiss 38 He filter set (Carl Zeiss MicroImaging GmbH, Gõttingen, Germany) for fluorescein isothiocyanate (FITC) and the transmitted light images were acquired using differential interference contrast. Z stacks were acquired and the resulting image was a maximum intensity projection of the stacks. For the plate cultures the images where acquired in a Zeiss SteREO Lumar.V12 stereoscope, equipped with an Axiocam MRC, using AxioVision software release 4.8 (Carl Zeiss MicroImaging GmbH, Gõttingen, Germany). Images were treated using Image J software (Research Service Branch, NIH, Bethesda, MD).

The scale-up procedure was carried out by streaking a single colony from a well that did not show any signs of bacterial growth onto a Petri dish containing solid Algal growth medium. After 1 week axenic plates were scrapped to 100 ml Erlenmeyer flasks with liquid Algal growth medium. A week later the 100 ml culture was transferred to 1 l reactors with aeration. In liquid medium, the scale-up was performed by transferring the 250 μl of the 96-well plates into 5 ml test tubes containing 1 ml of liquid Algal growth medium under minor aeration. Cells were allowed to grow for 1 week and were then transferred successively to 20 ml test tubes, 100 ml test tubes and finally 1 l reactors with aeration.

Lipid extraction was performed according to a modified protocol by Bligh and Dyer 33 . Briefly, the obtained algal biomass was homogenized at room temperature with an IKA Ultra-Turrax disperser (IKA-Werke GmbH, Staufen, Germany), in a mixture of chloroform, methanol and water (2:2:1). The mixture was centrifuged to allow phase separation, and a known volume of the organic phase was pipetted into a new preweighed tube. The extract was then evaporated until dryness in a warm bath (50°C) and weighed carefully to estimate lipid contents.

The resulting unialgal cultures were identified by optical microscopy and 18S rDNA sequencing. DNA extraction was performed with the EZNA DNA plant extraction kit (Omega Bio-Tek, Norcross, GA) according to the manufacturer's procedure. The obtained DNA was amplified by PCR with specific primers (Table 3) and sequenced at an in-house DNA sequencing facility equipped with an Applied Biosystems 3130XL DNA sequencer (Life Technologies BV, Porto, Portugal).