Introduction The search for sustainable sources of liquid transportation fuels has led to renewed interest in microalgae as potential feedstocks and rising research activity focused on the basic biology of algae. Microalgae can accumulate large quantities of oils (triacylglycerols) and carbohydrates, particularly when nutrient-deprived [1], [2]. Recent estimates taking into account different locations predict that microalgal photosynthesis can produce between 40,000 and 50,000 L ha−1 year−1, which is 5-to-6 times the yield observed for oil palm [3]. To realize this potential, it will be necessary to understand photosynthetic growth and metabolism of specific model algae. Even though genomic information and basic molecular tools are available for a range of organisms such as the diatoms Phaeodactylum tricornutum [4], [5], the brown algae Ectocarpus siliculosus [6] or the tiny chlorophyte Ostreococcus tauri [7], the mechanistic study of microalgal gene functions is currently lagging behind models such as Arabidopsis. Of all algae, Chlamydomonas reinhardtii is currently the most thoroughly studied based on the number of entries in the Public Library of Medicine (http://www.ncbi.nlm.nih.gov/pubmed/). Despite its proven versatility, Chlamydomonas is still somewhat limited with regard to available tools for its molecular analysis. For example, efficient targeted inactivation of genes by gene disruption technology is not available and loss-of-function mutants can be difficult to obtain by RNA interference and related techniques. The recent achievement of homologous gene replacement in Nannochloropsis oceanica [8] opens up potential opportunities to develop this alga into an alternate model organism representing marine, oleaginous microalgae. Nannochloropsis is classified under the class Eustigmatophyceae of the Heterokontophyta [9], a diverse algal group that includes brown algae and diatoms. The plastid of this alga is surrounded by four membranes derived from a secondary endosymbiotic event [10]. Strains from this genus have been investigated for their lipid composition and lipid accumulation, e.g. [11]–[14]. In addition, the biomass production by strains of Nannochloropsis grown under different conditions has been increasingly studied in recent years, e.g. [15]–[19]. Given the potential of this alga as an industrial feedstock and the progress made in developing homologous gene replacement, several research groups have set out to sequence the genome of different Nannochloropsis strains and draft genomes of Nannochloropsis oceanica [20] and Nannochloropsis gaditana [21] have recently become available. Here we focus on the publicly available strain Nannochloropsis oceanica CCMP1779, which we chose based on its growth in culture, its sensitivity to antibiotics, and ease of integrating transformation markers into its nuclear genome. We sequenced its genomic DNA and two sets of cDNAs obtained from two different growth conditions to aid in the annotation of genes. Its genome has been tentatively compared to that of N. gaditana. In addition a team of scientists has begun to manually annotate and examine the gene repertoire for specific pathways and processes to better understand the biology of this alga. Results/Discussion Strain selection—antibiotic sensitivity, growth and introduction of selectable markers Out of 20 axenic Nannochloropsis strains obtained from the Provasoli-Guillard National Center for Marine Algae and Microbiota (NCMA, formerly CCMP), strains of the N. salina (CCMP369), N. gaditana (CCMP1775 and 536) and N. granulata (CCMP529), as well as two not further specified strains (CCMP1779 and CCMP531) were selected based on uniformly dispersed, robust growth in enriched artificial sea water (16 g/L marine salt content) in batch culture as well as on agar-solidified medium. Both unspecified Nannochloropsis sp. strains cluster with strains of the N. oceanica species in a rooted tree [22] based on 26 published 18S rRNA nucleotide sequences (Figure 1) using Eustigmatos vischeri (Eustigmatophyceae) as an out-group [23]. For this reason, these strains are hereafter referred to as N. oceanica. Because of poor growth under the conditions we have used, N. oculata and the fresh water species N. limnetica were not further analyzed. 10.1371/journal.pgen.1003064.g001 Figure 1 Rooted neighbor joining tree of 18s rRNA sequences of different Nannochloropsis species using Eustigmatos vischeri as an outgroup. Labels refer to strain identification numbers from the respective culture collections, if applicable the synonym is given as 2nd name. CCMP, Provasoli Guillard Culture Collection for Marine Phytoplankton, USA; CCAP, Culture Collection of Algae and Protozoa, UK; MBIC, Marine Biotechnology Institute Culture Collection, Japan, AS3-9 from [177]. The use of antibiotics is essential for eliminating contaminants from cultures and genes conferring resistance to antibiotics are frequently used as markers for the introduction and genomic insertion of foreign DNA. Therefore, we tested the Nannochloropsis strains for their sensitivity to a range of antibiotics. Cells were plated at high density on agar-solidified medium containing the antibiotics at high density to determine the appropriate dosage (Table S1). Zeocin (5 µg/mL), and Hygromycin B (25 µg/mL) were chosen for use in subsequent selection marker studies, because of the consistent inhibition of growth at low concentrations by these antibiotics. Sensitivity to Paromomycin and Hygromycin B varied among the Nannochloropsis strains; Paromomycin had promise as a selective agent for the two N. oceanica strains (CCMP1779 and CCMP531), which were also the most sensitive to Hygromycin B. Of those four antibiotics, plasmids with genes that confer resistance to Zeocin, Hygromycin B, or Paromomycin are readily available and commonly used for transformation of Chlamydomonas as reviewed in [24]. Sensitivity to antibiotics is often determined by its rate of entry into the respective cells, which may be determined by the cell membrane and its transporters and the physical barrier provided by the cell wall. Differences in cell wall composition or thickness allowing more efficient cell entry of antibiotics are possible explanations for increased sensitivity in N. oceanica strains. Since efficient uptake of antibiotics or other supplemented molecules (such as metabolic substrates, inhibitors or nucleic acids) is a desirable trait for a laboratory model organism, we focused on N. oceanica. All Nannochloropsis strains were resistant to low concentrations of Rifampicin (10 µg/mL), Benomyl (5 µg/mL), Nystatin (5 µg/mL), and higher concentrations of Spectinomycin (100 µg/ml), Ampicillin (200 µg/ml), and Chloramphenicol (100 µg/mL). Hence these antibiotics can be useful for selecting against bacterial and other possible contaminants in Nannochloropsis cultures. Basic growth characteristics of N. oceanica CMP1779 were determined. The growth curves were fitted to a sigmoidal curve and the averaged exponential growth rate k, maximum cell density amax and time of half maximum cell density xc were determined (Table S2). Under photoautotrophic conditions in enriched sea water the exponential growth rate of the population, k, reached an average of 0.66±0.17 d−1 and cultures grew to a cell density of approximately 6×107 cells mL−1 (amax). The addition of vitamins did not enhance growth in liquid culture, whereas the addition of an external carbon source drastically increased final cell densities in stationary phase, reaching up to 8.7×107 or 1.5×108 cells mL−1 when the medium was supplemented with 30 mM glucose or fructose, respectively. The intrinsic growth rate did not increase, indicating a positive effect of sugars on cell division only during the later log phase and/or early stationary phase when self-shading limited growth in the photoautotrophic culture. Introduction of foreign DNA and stable integration into the genome are crucial for many reverse-genetics approaches. Recently, efficient protocols using an electroporation approach have been published for N. oceanica sp. and N. gaditana [8], [21]. We tested the strain CCMP1779 for nuclear transformation using an endogenous promoter region of a structural lipid droplet surface protein [25] driving the aphVII gene that confers resistance to Hygromycin B. Transformation was performed by electroporation without prior enzymatic treatments [26], and selection on 50 µg/mL Hygromycin B resulted in a transformation rate of 1.25×10−06±0.6×10−06 per µg plasmid DNA (Table S3). This equals a more than 10-fold increase in transformation events compared to plasmid pHyg3 [27] that was engineered for C. reinhardtii. The insertion of the transgene into the genome was confirmed for selected clones of both constructs by Southern hybridization (Figure S1). Genome sequencing strategy, assembly, and annotation The N. oceanica CCMP1779 genome was sequenced with 454 and Illumina technology. Both types of reads were used to generate a hybrid assembly with 3,731 contigs, an assembly size of 28.7 Mb and an N50 of 24,152 bp (see Materials and Methods; Figure 2, NCBI/SRA SRP013753). The coverage of the hybrid assembly was calculated to be ∼116-fold (30-fold for 454, and 86-fold for Illumina data). In addition to genomic sequences, we conducted RNA-sequencing (RNA-seq) and generated a de novo assembly of 65,321 transcripts. Using these transcripts, we assessed the parameter choice for genome assembly (see Materials and Methods). RNA-seq reads were also mapped to the final genome assembly and assembled into 35,756 transcripts to facilitate structural annotation. 10.1371/journal.pgen.1003064.g002 Figure 2 Hybrid assembly strategy using Illumina and 454 reads. N50: the length N for which 50% of all bases in the sequences are in a sequence of length L 1 contigs. We also used de novo transcript assemblies (see next section) to assess genome assembly quality. The genomic sequence data are deposited in NCBI SRA (SRP013753). Transcript assembly and differential expression analysis De novo transcript assemblies were generated from 55 bp directional single-end Illumina reads of N-replete and N-depleted conditions (NCBI/GEO GSE36959) using Oases (http://www.ebi.ac.uk/~zerbino/oases/). First, Oases was run for k-mer lengths of 23, 25, 27, 29, 31, 33, 35, and 37, and the results were compiled. To identify a set of high confidence transcripts from the de novo assemblies, proteins from six sequenced heterokont genomes, including Ectocarpus siliculosus [6], Pythium ultimum [161], Phytophthora sojae [162], Phytophthora ramorum [162], Thalassiosira Pseudonana [64], and Phaeodactylum tricornutum [4], were aligned to the de novo transcripts and only those with significant matches to known proteins were kept. These transcripts with cross-genome matches were mapped back to the Illumina genome assemblies to evaluate genome assembly quality. In addition to de novo transcript assembly, we generated a genome-based transcript assembly. Transcriptomic reads from N-replete and N-depleted conditions were separately mapped to the hybrid genome assembly using Tophat [160] (parameters: -I 10 –I 3000 –library-type fr-unstranded –g 1). The mapped reads were assembled into transcripts using Cufflinks [163] (-I 3000 –library-type fr-secondstrand) and a set of transcripts was generated for each condition. Genome annotation The MAKER genome annotation pipeline [28] was used to annotate the genome. The first run of MAKER was performed using the est2genome option in the absence of a trained gene predictor. Transcripts from both N-replete and N-deprived growth conditions were provided to MAKER along with protein sequences from the above mentioned six sequenced heterokonts. Gene models obtained from the first run were used to train ab initio gene prediction programs SNAP [164] and Augustus [165]. With the trained models, MAKER was rerun. The gene models from the rerun were used for training SNAP and Augustus again. The second round training models were provided to run MAKER for the third time to generate the final annotations. The protein sequences were searched for Pfam domain Hidden Markov Models using HMMER3 [166] with trusted cutoffs. CEGMA was run on the genome assembly using default settings [30]. A total of 11,973 genes (12,012 protein models considering alternative splice forms) were recovered with an average AED score of 0.555. During the course of the study, a new version of MAKER was released. Thus we conducted a second annotation run with the most recent MAKER version, a more recent repeat library, and a larger protein evidence dataset. Given that the AED distributions were highly similar between these two annotation datasets (Figure S12A, S12B) only annotation results from the first set of analysis were used throughout. InterProScan [167] was used to identify Pfam protein domains within the predicted protein sets from Nannochloropsis oceanica CCMP1779 and six other heterokonts. Protein families were identified by grouping proteins with identical protein domains, and the number of proteins from each species that were classified into each protein family was tallied. Figure S2 shows the percentages of proteins that have at least one InterPro domain, and those that have none, of each species. Functional annotation and determination of differential expression Blast2GO [32] (http://blast2go.com/b2ghome) was used for functional annotation of predicted protein models with the default settings for the mapping and annotation step. The initial BLAST [159] search was performed with an e-value cut-off of 10−5 and a maximum of 20 blast hits. This results in Gene Ontology (GO) annotations of 5,980 N. oceanica genes (in 4,012 GOs) and 3,008 N. gaditana genes (in 3,205 GOs). Fisher's exact test was used to assess if either the number of conserved or species-specific genes are over-represented in any GO category. Cuffdiff from Cufflinks package [163] was used to analyze the differential gene expression under N-replete and N-deprived growth conditions. Fisher's exact tests were performed to determine the enrichment of each GO category in up- and down-regulated gene clusters and at the 1% significance level based on p-values. Comparison of Nannochloropsis genomes OrthoMCL [168] was used to identify Orthologous Groups (OGs) of genes in N. gaditana, N. oceanica CCMP1779, and E. siliculosus (run parameters: percentMatchCutoff = 50, evalueExponentCutoff = −5). BLAST [169] was used to identify significant matches of lineage-specific genes across species. A significant match was defined as identity ≥47.04% (5 percentile in the identity distribution of one-to-one orthologs between N. gaditana and N. oceanica), Expect value≤10−5, alignment length ≥30 amino acids, and ≥50% of the protein sequence covered in the alignment. The orthologous group assignments as well as lists of species-specific genes are detailed in Table S5. Database tools To allow easy access to the CCMP1779 genome data, we released a public version of the genome browser along with a basic BLAST tool to search nucleotide and protein databases, accessible at www.bmb.msu.edu/nannochloropsis.html. The genome browser contains EST data aligned to the latest genome assembly as well as alternative gene models in addition to the final models retrieved from the MAKER gene annotation pipeline described above. Collection and identification of repetitive sequences Repetitive sequences were first collected with RECON (version 1.06, [170], http://www.repeatmasker.org/), with a cutoff of 5 copies. This resulted in a total of 175 repetitive sequences. Two sequences matching non-transposase proteins were considered to represent gene families and were excluded. Thereafter, repetitive sequences with more than 10 copies were manually curated to verify their identity, individuality and 5′/3′ boundaries. This was achieved by pair wise comparison of sequence contigs containing the relevant repeats using the “gap” program available from the GCG package (version 11.0, Accelrys Inc., San Diego, CA). A boundary was defined as the position to which sequence homology is conserved between the aligned sequences, and sequences flanking the boundary of the putative element were compared with that of a known transposable element (TE). Furthermore, the sequences immediately flanking the element boundaries were examined for the possible presence of target site duplication, which is created during transposition. Each transposon family has unique terminal sequences and target site duplication, which can aid in the identification of a specific transposon [171]. For some large transposable elements, fragmented sequences identified by RECON were joined to derive a compete sequence. To recover transposable elements that are less than 5 copies, the assembled sequence was masked using the repeat library generated by RECON. Thereafter, the masked sequence was used to search against known transposons at the protein level (BLASTX E ’ indicates the highest concentration of the respective antibiotic tested and no detectable impact on cell growth observed. All of the Nannochloropsis strains listed here were found to be resistant to the following antibiotics with the respective concentrations in µg/mL given in parenthesis: Rifampicin (10), Benomyl (5), Nystatin (5), Spectinomycin (100), Ampicillin (200), Chloramphenicol (100). (DOCX) Click here for additional data file. Table S2 Growth parameters of N. oceanica CCMP1779 in f/2 medium using different supplements. V = f/2 Vitamine mix, Gl = Glucose, Fr = Fructose, curves have been determined in triplicates based on cell density and fitted to a sigmoidal logistic function type 1 individually using OriginPro software (y = a/1+exp(−k*(x*xc))). Parameters a (Amplitude, here: max. cell density in cell/ml), xc (time of ½a in d) and k (coefficient, intrinsic growth rate d−1) are arithmetic means with standard deviation. (DOCX) Click here for additional data file. Table S3 Number of resistant colonies achieved by electroporation of N. oceanica CCMP1779 cells in the presence of linearized pHyg3, pSelect100 plasmids per µg linearized plasmid DNA and transformation rates. Arithmetic means are given from three (pSelect100) or four (pHyg3 and no plasmid control) independent experiments with standard deviation. All transformation reactions contained denatured salmon sperm DNA in 10-fold excess compared to plasmid DNA. (DOCX) Click here for additional data file. Table S4 Enriched GO categories in up- and down-regulated genes during N-deprived versus N-replete conditions based on RNAseq data. (DOCX) Click here for additional data file. Table S5 Comparison of N. gaditana and N. oceanica CCMP1779 protein sets. (XLS) Click here for additional data file. Table S6 Enriched GO categories in conserved OGs and N. oceanica CCMP1779-specific and N. gaditana-specific genes. (DOCX) Click here for additional data file. Table S7 Putative genes identified to be involved in photosynthetic electron transport in CCMP1779. In cases where no gene model was structurally annotated, genome coordinates are given. (DOCX) Click here for additional data file. Table S8 Genes predicted to encode for Violaxanthin-Chlorophyll binding proteins (VCP) in CCMP1779 genome and there designation in the phylogenetic tree (Fig. 6). (DOCX) Click here for additional data file. Table S9 Putative genes identified to be involved in xanthophyll synthesis. (DOCX) Click here for additional data file. Table S10 Genes putatively involved in central carbon metabolism and possible carbon concentrating mechanism. (DOCX) Click here for additional data file. Table S11 Functional annotation of putative genes involved in H2 metabolism and oxidative phosphorylation identified in the CCMP1779 genome. (DOCX) Click here for additional data file. Table S12 Fatty acid composition of the major glycerolipids of Nannochloropsis CCMP1779. Averages are presented (n = 3) with standard deviation in parenthesis. (DOCX) Click here for additional data file. Table S13 Functional annotation of putative genes involved in fatty acid and glycerolipid biosynthesis. (DOCX) Click here for additional data file. Table S14 Genes predicted to encode enzymes putatively involved in fatty acid mobilization and degradation. (DOCX) Click here for additional data file. Table S15 Genes predicted to encode enzymes putatively involved in cell wall metabolism. (DOCX) Click here for additional data file. Table S16 Predicted genes in the biosynthetic pathways of Asp-derived, aromatic and branched-chain amino acids and in nitrogen assimilation in CCMP1779. (DOCX) Click here for additional data file. Table S17 Presence of fused genes in essential amino acid biosynthesis in representative bacteria, cyanobacteria, green algae, diatoms, Nannochloropsis and higher plants. (DOCX) Click here for additional data file. Table S18 Putative Nannochloropsis genes involved in sulfate assimilation and metabolism. (DOCX) Click here for additional data file. Table S19 Putative chloroplast protein import related genes identified in the CCMP1779 genomic sequence. (DOCX) Click here for additional data file. Table S20 Summary of testing the HECTAR heterokont protein localization prediction tool. Detailed information on the tested sequences and results is available in Table S25. (DOCX) Click here for additional data file. Table S21 Predicted genes involved in organelle division. (DOCX) Click here for additional data file. Table S22 Genes predicted to be involved in light signaling. (DOCX) Click here for additional data file. Table S23 Putative transcription factors and transcriptional regulators. (DOCX) Click here for additional data file. Table S24 Protein domain search results for 6 different heterokonts. (XLSX) Click here for additional data file. Table S25 Predicted subcellular localization of proteins. (XLSX) Click here for additional data file. Text S1 Supplemental results and discussion. Additional annotation is provided for genes predicted to be involved in ROS scavenging systems, oxidative phosphorylation, amino acid biosynthesis, degradation of branched chain amino acids, sulfate uptake and metabolism, and histones and histone variants. (DOC) Click here for additional data file.
