Nine species of diatoms were isolated from the Western Antarctic Peninsula along the PalmerLTER sampling grid in 2013 and 2014. Isolations were performed using an Olympus CKX41 inverted microscope by single cell isolation with a micropipette (Anderson 2005). Diatom species were identified by morphological characterization and 18S rRNA gene (rDNA) sequencing. DNA was extracted with the DNeasy Plant Mini Kit according to the manufacturer’s protocols (Qiagen). Amplification of the nuclear 18S rDNA region was achieved with standard PCR protocols using eukaryotic-specific, universal 18S forward and reverse primers. Primer sequences were obtained from Medlin et al. (1982). The length of the region amplified is approximately 1800 base pairs (bp). Pseudo-nitzschia species are often difficult to identify by their 18S rDNA sequence, therefore, additional support of the taxonomic identification of P. subcurvata was provided through sequencing of the 18S-ITS1-5.8S regions. Amplification of this region was performed with the 18SF-euk and 5.8SR_euk primers of Hubbard et al. (2008). PCR products were purified using either QIAquick PCR Purification Kit (Qiagen) or ExoSAP-IT (Affymetrix) and sequenced by Sanger DNA sequencing (Genewiz). Sequences were edited using Geneious Pro software (http://www.geneious.com, Kearse et al., 2012) and BLASTn sequence homology searches were performed against the NCBI nucleotide non-redundant (nr) database to determine species with a cutoff identity of 98%.

Diatom phylogenetic analysis was performed with Geneious Pro and included 71 additional diatom 18S rDNA sequences from publically available genomes and transcriptomes, including those in the MMETSP database. Diatom sequences were trimmed to the same length and aligned with MUSCLE (Edgar 2004). A phylogenetic tree was created in Mega with the Maximum-likelihood method of tree reconstruction, the Jukes-Cantor genetic distance model (Jukes and Cantor 1969), and 100 bootstrap replicates.

Illumina TruSeq adapters and poly-A tails were trimmed from raw reads using the Fastx_toolkit clipper function. Fastq_quality_filter was used to remove poor quality sequences, such that remaining sequences had a minimum quality score of 20 with a minimum of 80% of bases within a read meeting this quality score requirement. Any remaining raw sequences less than 50 base pairs in length were also removed. Merged files were assembled de novo using Trinity (Grabherr et al. 2011). The resulting assembly was filtered to remove contigs less than 200 bp in length. Trinity-assembled contigs which exhibited sequence overlap were grouped into isogroups which were then used for sequence homology searches (BLASTx E-value ≤ 10-4) against the Kyoto Encyclopedia of Genes and Genomes (KEGG) databases (Kanehisa 2006).

BUSCO (Benchmarking Universal Single-Copy Orthologs) was used to assess the completeness of genomes and transcriptomes based on sets of single copy orthologous groups derived from OrthoDB that are highly conserved within multiple lineages (Felipe et al. 2015). Completed, duplicated and fragmented orthologs were determined by meeting an ‘expected score’ and having aligned sequences within two standard deviations of the BUSCO gene’s length. A second metric of completeness was performed by evaluating conserved pathways, such as the ribosome and spliceosome, using the single-directional best-hit method in the KEGG Automatic Annotation Server (KAAS) (Moriya et al. 2007). Finally contiguity, was calculated at the 0.75 level as according to Martin and Wang (2011) with custom scripts.

For each transcriptome, unassembled sequence reads were aligned to the final Trinity assembly using Bowtie 2 (Langmead 2012). Mapped reads were normalized by the Reads per Kilobase per Million reads method (RPKM) (Mortazavi et al. 2008).

Gene biogeographical distributions - 20 genes of interest were selected in the study to investigate the molecular basis of iron and light limitation in polar diatoms. Reference sequences for each of these genes were obtained from the F. cylindrus and P. tricornutum JGI genome portals and T. pseudonana and T. oceanica NCBI and GenBank repositories. Reference sequences were identified in the transcriptomes by translated nucleotide homology searches (tBLASTn) with an e-value cutoff of <10-5. A reciprocal tBLASTn homology search was performed for each transcriptome against the KEGG GENES database, using the single-directional best-hit method in the KAAS online tool to ensure consistent gene annotations (Moriya et al. 2007).

Subsequently, reference sequences were identified in the MMETSP protein database by BLASTp (e-value <10-5) homology searches among the diatom transcriptomes. The transcriptomes and their associated latitude and longitude were obtained from iMicrobe Data Commons (Project Code CAM_P_0001000) and the National Center for Marine Algae and Microbiota (NCMA). Custom Matlab scripts allowed global biogeographical distribution of key genes of interest to be mapped.