Metatranscriptomic Analysis of Co-Occurring Phytoplankton Species in the Ross Sea Polynya

Primary production (PP) of the Ross Sea Polynya (RSP) contributes a significant proportion of the total PP of the Southern Ocean. As a result, the photosynthetic activities of phytoplankton communities in the RSP play important roles in the overall biochemical cycling of carbon and nutrients and structuring the marine food-web. Environmental change, especially in light and iron regimes, regulates variations in PP and affects the phytoplankton community dynamics. Since individual phytoplankton species have unique nutrient requirements, the limited resources impact these algal taxa in different ways. In this study, we used metatranscriptomic analysis to examine how the algal taxa Fragilariopsis, Thalassiosira, Pseudo-nitzschia, Micromonas, and Phaeocystis antarctica differ in their acclimation to their shared environment during the bloom season in the RSP. During the austral spring and summer of 2013-2014, the phytoplankton communities in the RSP were iron limited across all sampling sites and exposed to high-light conditions at some sampling sites. However, the acclimation of individual algal taxa to these conditions was different. Niche partitioning between the diatom group and the haptophyte Phaeocystis antarctica was detected. Phaeocystis dominated at the greater depths (80-100m) and showed relatively low abundances at the surface (average 8.9±5.8%). Pseudo-nitzschia showed optimal niche adaptation among near the surface with a largest population size (average relative abundance 32.0±18.0%) and increased genetic activity. The expression of key-genes of Pseudo-nitzschia showed potential acclimation to limited iron and cobalamin conditions. However, the expression levels of those genes were not only controlled by the related environmental parameters, but also by a set of physical and biochemical environmental factors. Our study is limited to examining the interaction between the phytoplankton communities and the environments. The potential effects of bacterial communities, viruses and predators should be included to further the understanding of individual algal species’ niche adaptation in the future. This study demonstrates that, although some environmental factors control the overall bulk PP, those factors have variable effects on the individual algal taxon. We also show the need to consider a combination of environmental parameters to accurately predict shifts in phytoplankton community composition within the context of long-term climate change.


INTRODUCTION
As the level of CO 2 in the atmosphere has steadily increased and recently accelerated due to anthropogenic CO 2 productions (7-9 Pg C yr -1 ), global temperatures have risen with climate change Bazzaz, 1990;Long, 1991;Sarmiento et al., 1998;Takahashi et al., 2002). The Southern Ocean (SO, South of 50 °S) consists of only 10% of the global ocean surface, but it accounts for about 25 % of the oceanic uptake of atmospheric CO 2 (Takahashi et al., 2002(Takahashi et al., , 2009. Thus, in this era, the SO has received more and more attention because of its disproportionately high and constant total primary production (~2 Pg C yr -1 with interannual variability of ±4%) Le Quere et al., 2007;Moore & Abbott, 2000;Schlitzer, 2002). The SO consists of only 10% of the global ocean surface, but it accounts for about 25 % of the oceanic uptake of atmospheric CO 2 (Takahashi et al., 2002(Takahashi et al., , 2009). In addition, the high biological production in the SO plays a significant role in the biochemical cycles and the biological pump process by accommodating fecal pellet formation and aggregation of particulate organic carbon (Arrigo et al., 1999;Arrigo et al., 2000;Asper & Smith, 1999;Moloney et al., 1992).
Most of the total primary production of the SO is contributed by intense phytoplankton blooms of coastal polynyas in the Antarctic during the austral spring and summer (Arrigo & van Dijken, 2003;Arrigo et al., 2015;Sullivan et al., 1993). These polynyas are recurring areas of open water surrounded by pack ice and continental shelves (Smith & Barber, 2007). After long, dark winter periods in the Antarctic, the coastal polynyas provide oases to algal communities by providing open water areas where light can penetrate and delivering nutrient supplies from sea-ice melt water as well as the upwelling of deep water masses (Arrigo & McClain, 1994;Arrigo & van Dijken, 2003;Comiso et al., 1993). As a result, phytoplankton communities form intense blooms in these relatively fertile coastal polynyas, and their growth and biomass accumulation are much greater within polynyas than in other regions of the SO Becquevort & Smith, 2001;Gowing et al., 2001).
In the RSP specifically, the notable polynya eventually develops in December at the onset of the austral spring (Arrigo et al., 1999;Comiso et al., 1993;Smith & Gordon, 1997). The phytoplankton communities develop their blooms in early December supported by constantly improving environmental conditions (Arrigo & McClain, 1994;El-Sayed et al., 1983) and, triggered by the seeding with sea-ice algae, they follow on the heels of the ice-edge bloom Wilson et al., 1986). The blooms continue for several weeks to months depending on the environmental conditions and physiological properties of algal species in the community (Arrigo et al., 1998;DiTullio et al., 2000;Smith et al., 2000).
These blooms will decrease and disappear in early March when the surface water in the Ross Sea begins to refreeze.
The dynamic environmental conditions in the RSP cause spatial and temporal variations in phytoplankton growth rates and biomass in the RSP Smith et al., 1996). The vertical stability of the upper water column and the extremely low concentrations of dissolved iron (DFe) of ~ 0.1 nM in the surface layer have been considered as the major factors that control the local bulk primary production in the RSP Bertrand et al., 2011;Gerringa et al., 2015;Martin et al., 1990;Sedwick et al., 2000;Sverdrup, 1947;Zhu Z et al., 2016). In addition, grazing by zooplankton (Deibel & Daly, 2007;Lancelot et al., 1993), seeding by the sea ice algae (Asper & Smith, 1999;Smith & Nelson, 1985;, and the light limitation due to sea ice cover or continental shelf shading Perovich, 1990;Smith & Gordon, 1997) may also control the biological productivity of phytoplankton communities in the RSP.
Spatial and temporal variations in the composition of phytoplankton communities in the RSP have been reported. These communities are typically dominated by the haptophyte Phaeocystis antarctica during the spring and early summer when the mixed layer in deep, and several diatom species (Pseudonitzschia, Fragilariopsis, Thalassiosira, Chaetoceros) become more abundant in late summer and coincide with a shallower mixed layer depth (Delmont et al., (in prep); Smith et al., 2000;Tremblay & Smith., 2007). Some studies have presented that iron and light availability (Bertrand et al., 2011;Delmont et al., (in prep); Sedwick et al., 2000;Sverdrup, 1947), water stratification (Rozema et al., 2016) and temperature (Zhu Z et al., 2016) might be the major factors that control the relative abundance of individual algal species in the RSP during the bloom periods.
However, few studies conducted in the RSP investigate how several different species can co-occur in the same limited environmental conditions. Analysis of ecological and physiological status of individual algal species would further our understanding about algal species' relative contributions to bulk primary production in the RSP, in terms of co-existence and characterization of their niches.
It is also important for better predictions of changes in the phytoplankton community structure in response to long-term environmental change.
The main objectives of this research were 1) to examine how individual algal

Metatranscriptomic analyses
Water samples (4.0 -8.0 L depending on the biomass) for metatranscriptomic analysis were rapidly filtered onto 0.2 μm filters using a vacuum pump, and the total filtration process took 30 minutes or less for each sample. The six filters were flash frozen in liquid nitrogen and stored at -80 ˚C prior to RNA extraction.

RNA extraction, library preparation, and sequencing
Total RNA was extracted from each filter (total 6) using a modified

RNA-Seq data analysis
Based on the Phred quality scores generated during the sequencing run, the raw reads were trimmed for the quality by Trimmomatic 0.36 (Bolger et al., 2014) in the paired end mode with parameters AVGQUAL 30 and CROP 134.
Reference-based assembly was performed using Bowtie2 (Langmead & Salzberg, 2012); quality filtered reads of each metatranscriptome were mapped against the The visualization of distributions of reads on the reference genomes was performed with Anvi'o (Eren et al., 2015). ORFs of each reference genomes were predicted with prodigal (Hyatt et al., 2010) and predicted ORFs were transformed to protein sequences and they were functionally annotated using COGs database (Tatusov et al., 2000) with Blastp e-value threshold 1e-5. The coverage of predicted genes in each reference genome were calculated by Filtration excluded the lowly expressed (low-coverage) genes with 'cpm' function (cpm < 0.5) in edgeR , and also the genes for which had mapped reads from less than 3 samples were excluded. The filtered data were normalized and tested for the differentially expressed genes using the "R" with edgeR package  and limma package (Ritchie et al., 2015).
'calcNormFactors' function for TMM normalization (Robinson & Oshlack, 2010) in edgeR was used to calculate the normalization factors relying on the library sizes and the composition bias; TMM normalization method reduces impacts of changes in species abundances on estimating gene expression by assumption that expression of a large proportion of genes in a certain organism un-change regardless the changes in environmental conditions. The calculated normalization factors were used during testing the differentially expressed genes through 'voom' function in limma (Law et al., 2014). 'Voom' function first adjusted the bias between the samples using the calculated normalization factors. This function also estimated the mean-variance relationship of the log-counts and calculated a precision weight for each sample. Processed data then was entered to the empirical Bayes analysis pipeline stored in limma package (Law et al., 2014;Phipson et al., 2016) to analyze the differential gene expression. Multidimensional plot (MDS) for investigating the differences in global gene expression patterns between the samples was produced with limma package on "R". The 100 most variable genes were also predicted. Based on the published papers, the genes that are related to the iron and light variations were extracted from the functional profiles of each reference genome mapping (key-genes). We used STAMP (Parks et al., 2014) software to show the differential expression levels of selected-genes (key-genes) across the samples for each reference species with heat maps and to estimate the correlation between the key-genes expression levels and the environmental parameters.

Pigment HPLC analysis was performed by Kate Lowry and Kate Lewis at
Stanford Woods Institute. The water samples for the HPLC analyses were filtered through Whatman glass-fiber filter (GF/F) with a nominal pore size of 0.7 μm (< 150 mm Hg). The pigment samples were immediately frozen in liquid nitrogen after filtration and stored at -80 ˚C until HPLC analysis. Phytoplankton pigments were extracted in 90 % acetone, and they filtered through 0.45 µm HPLC syringe cartridge into a 2 mL amber crimp-top vial. The filtered samples were injected into the HPLC system and the pigments were identified based on comparisons of their in-line diode array detector absorbance spectra. Standards were either purchased or fraction collected from algal monocultures. Calculating taxonomic composition of phytoplankton community from the HPLC data were performed with Chemtax version 1.95 (Wright, 2008).

Phytoplankton species characterization by FlowCAM
The FlowCAM was used to estimate the community composition at daily Samples were analyzed in triplicate and standard deviations were on average 2.8%.
Blanks were determined daily by loading a low iron seawater sample for 0, 5, 10 seconds. The blank values ranged from not detectable up to 14 pM. The average limit of detection, 16 pM was defined as 3 times the standard deviation of the mean blank and measured daily. Filtered and acidified (Seastar© baseline hydrochloric acid; pH 1.7) seawater was concentrated on a column containing aminodiacetid acid (IDA) which binds only transition metals and not the interfering salts. The column was rinsed with ultra-pure water, and eluted with diluted acid. After mixing with luminol, peroxide and ammonium, the oxidation of luminol with peroxide was catalyzed by iron and a blue light is produced and detected with a photon counter. The concentration of iron was calculated using a standard calibration line, where a known amount of iron was added to low iron containing seawater. Using this calibration line a number of counts per nM DFe is obtained.

Satellite remote sensing
Satellite remote sensing data was provided as a compressed kml format (kmz) via email during the cruise by an automated system at Stanford University under the supervision of Dr. Gert van Dijken.
The MODIS/Aqua ocean color satellite scenes were downloaded every 3 hours from NASA ftp-servers through two near real-time subscriptions using the NASA Ocean Biology Processing Group's data subscription service. The chlorophyll products were extracted from the downloaded files and mapped to a common projection. In addition, ice concentration data was downloaded from the National Snow & Ice Data Center (NSIDC) and re-projected in the same way.

General Environmental Conditions at each Sampling Site
The surface area of the Ross Sea Polynya (RSP) increases in size throughout the austral spring-summer sampling period (Fig. 2) Fig. 3).
The concentrations of nitrite/nitrate and phosphate were similar across the sampling sties, with average concentrations of 20.0 ± 0.75 µM and 1.39 ± 0.19 µM, respectively (Table 1). Based on the Redfield ratio (N/P~16) (Redfield, 1958), only station 4 showed potentially P-limited environmental conditions, whereas the remainder of the stations had ratios indicative of N-limited conditions (Table 1).
In general, the concentration of dissolved iron (DFe) was extremely low at the surface throughout the RSP during the cruise (<0.10 nM) (Gerringa et al., 2015). The average DFe concentration at sampling sites 1, 2, 4, 5, and 6 was 0.05 ± 0.01 nM. The DFe concentration at station 3 was much higher than the others, with a concentration of 0.2 ± 0.003 nM (Table 1). Gerringa et al. (2015) interpreted the higher DFe concentration at station 3 as an indication of iron input at the surface originating from dust deposition or melting sea ice. These processes were also suggested as the explanation of lower salinity at station 3 ( Table 1).

Evaluation of Metatranscriptomic Libraries
To examine the genetic response of individual algal species (Fragilariopsis, Thalassiosira, Pseudo-nitzschia, Macromonas, and Phaeocystis) to their shared environmental conditions in the RSP, we sequenced metatranscriptomes for each sampling site. The total yield of extracted RNA from each sample varied from 128.5 to 745.9 ng µL -1 . From this extracted total RNA, messenger RNA (mRNA) was enriched for each sample, yielding 0.066 ± 0.014 ng µL -1 to 1.621 ± 0.026 ng µL -1 ( Table 2). The percentage of mRNA among the total RNA was less than 0.1% in samples 1 and 2, and more than 0.1% in the remainder of the samples ( Table 2).
The number of quality controlled reads varied from 326,364 to 31,307,197 and had an average GC content of 48%; the metatranscriptomic libraries for stations 1 and 2 have 326,364 reads and 427,800 reads respectively, while metatranscriptomes for the other stations yielded more than 23,000,000 reads (Table 4).

Selection of the Reference Genomes
In order to select genomes for reference-based mapping, we compared several phytoplankton genomes and transcriptomes from a variety of original sampling locations and culture conditions to the overall alignments with our transcriptomes. The candidate algal taxa for reference genome selection were determined from inspection of phytoplankton communities in the RSP during our cruise period by FlowCAM, HPLC pigment, and 18S diatom community analyses (Filliger, unpublished) (Fig. 4b, 4c and Fig. 5b) and from publications on phytoplankton communities in the RSP during the bloom period (Delmont et al., (in prep); Smith et al., 2000;Tremblay & Smith., 2007). As a result, we included Fragilariopsis sp., Thalassiosira sp., Pseudo-nitzschia sp., Micromonas sp. and Phaeocystis antarctica reference genomes and transcriptomes for downstream metatranscriptome analysis (Table 3).
The overall alignments from our metatranscriptomes between the three different genomes and transcriptomes of Fragilariopsis sp. were similar across the samples (Table 4). Overall alignments to three different Thalassiosira sp. genomes and transcriptome were also similar across the samples, although the culture conditions of these reference Thalassiosira sp. were different. Thalassiosira-antarctica-CCMP982 (MMETSP) and Thalassiosira oceanica (NCBI, (Lommer et al., 2012)) were cultured under iron-limited conditions, whereas Thalassiosira pseudonana CCMP1335 (NCBI, (Armbrust, 2004)), known as a coastal species, was originally isolated from Moriches Bay and continuously grew in culture before DNA extraction.
The antarctica), only one source was available from the previous study in our group (Delmont et al., (in prep)). antarctica) as reference genomes and transcriptomes in this study.

Reference Genome Mapping
Of the total reads, 26.7 % (sample 1), 47.0 % (sample 2), 41.5 % (sample 3), 45.5 % (sample 4), 17.1 % (sample 5), and 22.7 % (sample 6) did not map to any of the reference genomes/transcriptomes (Table 5) Table 5). The relatively high alignment of reads to the P. nitzschia antarctica genome is in agreement with the algal community metatranscriptomic analysis of the on-deck incubation experiments completed during the same cruise (Delmont et al., (in prep)). Overall alignment percentages to the two other selected diatom genomes varied from 1.6% to 19.6% of the total reads from individual libraries.
Less than or equal to 0.4 % of reads from individual libraries mapped to P.

The Correlation between Read Alignments and Relative Abundances
We normalized the overall alignments to each reference genome to 100% across individual libraries (not including un-mapped reads) by calculating FlowCAM and HPLC pigment analyses ( Fig. 4b and 4c). However, there was no clear agreement between the read alignments (TPM) to each diatom species shown here and the relative abundance of individual diatom species determined by 18S diatom community data (Filliger, unpublished) (Fig. 5). The relative TPM values to each diatom species changed regardless of the changes in their relative abundances within the diatom community. The TPM value of P. nitzschia antarctica increased whereas the TPM value of F. cylindrus and T. oceanica decreased over the sampling period (Fig. 5a). The changes in the relative abundances of haptophytes and chlorophytes in the phytoplankton communities across the samples were also not correlated with TPM values of P. antarctica and Micromonas (Fig. 4).

Global Gene Expression Patterns
Reads that mapped to P. nitzschia antarctica and F. cylindrus were evenly distributed over their reference genomes whereas reads mapping to other species were concentrated on partial areas of the whole genome (Fig. 6). Since the number of mapped reads affiliated with each species were not evenly distributed between the samples (Fig. 6, box plots), we converted the coverage values of predicted genes of each reference genome to cpm (counts-per-milion), which normalized the different sequencing depths, to compare the differences between the samples. The differences in global gene expression patterns of each algal species across the sampling sites were examined with multidimensional scaling (MDS) plots; each sample site has different environmental and geographical conditions (Table 1; Fig.   1 and Fig. 2). Since stations 1 and 2 have lower quantity and quality metatranscriptome reads, which might cause a significant bias in the MDS plots (e.g. Appendix 1), we used only the metatranscriptomes of stations 3, 4, 5 and 6 for further analyses (Fig. 7). The overall gene expression patterns of P. nitzschia antarctica and P. antarctica were relatively similar at stations 5 and 6 and different at stations 3 and 4 ( Fig. 7a and 7c). On the other hand, the global gene expression patterns of T. oceanica and Micromonas were similar for stations 3 and 4 ( Fig. 7d and 7e). F. cylindrus showed different gene expression patterns across all samples (Fig. 7b).

Expression Patterns of Key Genes
Since the environmental conditions across the sampling sites were not significantly different, we expected that there would be no remarkable shifts in gene expression. Transcripts for the 100 most variable predicted genes from all reference species were examined to determine if these genes were closely linked to environmental factors. These genes did not change expression levels according to the environmental conditions or they were unannotated. This implies that the environmental conditions between the sampling sites were too similar to elicit a significant difference in transcript abundance of the genes related to the environmental parameters. In a next step, we examined how the phytoplankton communities responded to their environmental conditions, particularly to iron and light regimes and cobalamin availability between the samples. Based on the literature, we determined a set of key-genes that are differentially expressed under varying iron and light conditions and concentrations of cobalamin in these phytoplankton taxa (Fig. 8). Transcript abundances for the chlorophyte Micromonas were too low to determine difference in expression levels of these key genes. The overall coverage of key-genes before the normalization was highest for P. nitzschia antarctica (Fig. 8).
To determine the differential expression patterns of key-genes between samples, we applied TMM normalization (Robinson & Oshlack, 2010) and limma empirical Bayes analysis pipeline (Law et al., 2014). The key-gene expression patterns were not significantly different across the sampling sties, but they were different for the individual species (Fig. 9). Overall, key-gene expression patterns of T. oceanica and P. antarctica were similar. The few key-genes that were differentially expressed were related to vitamin B 12 (cobalamin) biosynthesis and to iron stress response (Flavodoxin) in the functional profiles of T. oceanica and P.
antarctica. Overall cobalamin biosynthesis genes showed a low transcript abundance, but the Flavodoxin transcript abundances was distinctly higher for stations 5 and 6 ( Fig. 8), suggesting that iron limiting conditions may have prevailed for these taxa at the two locations. Transcript abundances for F. cylindrus and P. nitzschia antarctica indicated a greater number of differentially expressed key-genes than was found in T. oceanica and P. antarctica, but expression levels for each gene were distinctly different for these two species ( Fig.   8 and Fig. 9). Genes involved in low light acclimation (Plastocyanin, Photosystem II stabilization factor) were most prominently expressed in F. cylindrus, whereas transcript abundances for iron stress related genes (Flavodoxin, Ferredoxin, Rieske protein etc.) were much higher in P.nitzschia antarctica.
Flavodoxin was highly expressed in all species except Micromonas.
Plastocyanin was expressed only in the pennate diatoms T. oceanica and F.
cylindrus. Gene expression for multicopper oxidase, a predicted ferric reductase, a putative heme iron utilization, and a heat shock protein was specific to P. nitzschia antarctica (Fig. 9). Genes for cobalamin biosynthesis protein were highly expressed in all species except Micromonas. The methionine synthase I (cobalamin-dependent) gene was shown in the functional profiles of P. antarctica, F. cylindrus, and P. nitzschia antarctica, and transcripts for methionine synthase II (cobalamin-independent) were detected only in F. cylindrus (Fig. 9).
Since the RSP metatranscriptomes were highly dominated by transcripts of P. nitzschia antarctica (~50% of total reads) ( Fig. 4a and Fig. 6c), we further analyzed the key-genes expression levels of P. nitzschia antarctica between the sampling sites. Figure 10 shows the same heat map as Fig. 9 (P. nitzschia antarctica) in descending order of mean gene expression levels in a log2 scale across the samples. We determined that the key-genes expression patterns were more similar between stations 5 and 6 ( Fig. 10), a pattern that was also observed for the global gene expression patterns of P. nitzschia antarctica (Fig. 7c).
Altogether, genes that encode components of the photosynthetic apparatus and that are involved in light acclimation responses were amongst the most highly expressed genes, whereas genes implicated in Fe-dependent redox processes were much less expressed (Fig. 10). These findings suggest that Pseudo-nitzschia was acclimated to low light rather than to nutrient (specifically iron) limitation.
We further compared transcript abundances of the key-genes for their correlation with environmental parameters, leading to sometimes surprising observations. The genes of deoxyribodipyrimidine photolyase, ferredoxin-NADP reductase, and NADPH-dependent glutamate synthase beta chain correlated with variations in temperature and salinity, whereas the Mg-chelatase subunit ChlD gene had an inverse correlation with these variables (Table 6). Transcript abundances for the genes encoding 6-phosphofructokinase, cobalamin biosynthesis protein CobN, and Mg-chelatase were positively correlated with DFe concentrations, but where inversely correlated with chl-a concentrations and with the relative abundance of Pseudo-nitzschia in phytoplankton communities ( Table  6). The genes related to carbon fixation (Fructose-bisphosphate aldolase class 1 and Photosystem II stability/assembly factor) were inversely correlated with light intensity (Table 6).

DISCUSSION
The "paradox of the plankton" postulate (Hutchison, 1961) states that many phytoplankton species coexist since they compete for the same resources in constantly changing environments. However, the phytoplankton community is sometimes dominated by one taxon, which usually happens during the summer blooms when the water layers are stratified and the environmental conditions do not change as quickly as in other seasons. In the case of the phytoplankton communities in the RSP, they only flourish throughout the austral spring and summer when dissolved iron is supplied by dust deposition and sea ice melting, and light reaches below the surface. Several algal taxa that often co-occur (Arrigo et al., 1999;Smith et al., 2000), participated in the intense RSP phytoplankton bloom studied in this thesis. The seasonal variations in iron, light and water stratification have been considered as major limited resources controlling the bulk primary production (Bertrand et al., 2011;Delmont et al., (in prep); Sedwick et al., 2000). Closer to the peak of the summer season, the environmental conditions of the RSP became more uniform; water masses were more stratified because of ice melt water (less salty) and increasing sea surface temperature due to stronger solar radiation; the dissolved iron were limited throughout the RSP (Boyd, 2002;Smith et al., 2014). As a result, reduced variations in environmental factors near the surface would causes increased competition between the phytoplankton species for limited resources. However, how those limited environmental resources interact with the physiological and ecological properties of individual algal species in the phytoplankton community is less well understood. Additionally, different algal taxa have different C:N:P ratio as well as different iron requirements, resulting in differences in how these taxa responds to environmental stressors (Alexander et al., 2015) and affecting the contribution that each species makes to primary production (Arrigo et al., 1999;Arrigo et al., 2000;Smith & Asper, 2001). These responses are mediated via the expression of genes that drive acclimation to changing nutrient and light regimes.
Thus, environmental conditions likely control the relative abundance of individual algal species and their contribution to primary productivity by adjusting their growth requirements via the expression of stress response genes (Arrigo et al., 1999;Delmont et al., (in prep).; Tagliabue & Arrigo, 2005).
During the blooms in the RSP, the phytoplankton community often consists of several diatom species and the haptophyte Phaeocystis antarctica (Delmont et al., (in prep); Sweeney et al., 2000). P. antarctica is the first species to form a bloom in spring followed by an extensive diatom blooms in summer (Smith et al., 2000;Tremblay & Smith., 2007). Based on culture studies, P. antarctica grows better under low iron and low irradiance conditions than diatom species, and after iron input or light level change, the growth rate of P. antarctica change more significantly than diatom species (Coale et al., 2003;Sedwick et al., 2000). P.
antarctica also showed higher efficiency in converting carbon dioxide to organic carbon via photosynthesis Sweeney et al., 2000;Tagliabue & Arrigo, 2005). These studies suggested a potential role for iron and light regimes in were also acclimated to high-light conditions. We found that chl-a concentrations and chlorophyll fluorescence were inversely correlated with Photosynthetically Active Radiation (PAR) (Fig. 3). Chl-a concentrations and fluorescence are often used indicators of photosynthetic biomass (Cullen, 1982;Falkowski & Kiefer, 1985;Steele, 1962). Therefore, lower chl-a concentrations and fluorescence values at higher Photosynthetically Active Radiation (PAR) would indicate that some populations in the phytoplankton communities at the surface had acclimated by reducing their pigment content to avoid photoinhibition (van Hilst & Smith, 2002). This was also in agreement with the analysis of the This is because phytoplankton communities in shallow mixed layers or more stable water are often dominated by diatoms whereas deeper mixed layer or relatively unstable water columns are often dominated by P. antarctica in the RSP (Arrigo et al., 1998;Goffart et al., 2000;Smith & Asper, 2001). An alternative interpretation could be that bacterial community composition changed with depth. Many algal taxa are auxotrophic for cobalamin (vitamin B12 dependent) and often this compound is provided by bacteria (Bertrand et al., 2015;Delmont et al., 2014;Shields & Smith, 2008). The fact that cobalamin biosynthesis genes were abundantly transcribed for the taxa that mapped to reference genomes, is an indication that cobalamin was a limiting factor for the phytoplankton community at To examine how individual algal species responded to their shared environment, we extracted the key-genes associated with major limited environmental factors, such as light and iron regimes that control primary production in the RSP (Sedwick et al., 2000). As expected, since there were no significant changes in the environmental conditions between the sampling sites, no notable shifts in the overall key-genes expression patterns of each species were detected (Fig. 9). Instead, the expression patterns for these key-genes were characteristic for the individual species. Flavodoxin, one of the key-genes expressed under the iron-deficient condition as a replacement for ferredoxin under iron limiting conditions (McKay et al., 1997;Palenik, 2015), was commonly expressed in the functional profiles of all species, except Micromonas. As we mentioned above, this indicates, with the once exception, all species had been exposed to iron-limitation and their acclimation response is shown from high transcript abundance of the flavodoxin gene. The differences in transcript abundance between flavodoxin and ferredoxin genes were more pronounced in P.
antarctica, T. oceanica, and F. cylindrus than in P. nitzschia antarctica. This could be interpreted as P. nitzschia antarctica experiencing a lesser degree of iron stress under the same iron-deficient condition. A few additional genes, which encode proteins for adjusting under low-iron condition (Multicopper oxidase (Shaked et al., 2005), Predicted ferric reductase (Morrissey & Bowler, 2012), Putative heme iron utilization protein (Hogle et al., 2014;Hopkinson et al., 2008)), were only expressed in P. nitzschia antarctica. Since a few of studies observed the roles of cobalamin in the physiology of phytoplankton communities (Bertrand et al., 2007(Bertrand et al., , 2015, we also analyzed the genes related cobalamin availability. Abundantly transcribed cobalamin biosynthesis protein gene in our metatranscriptomes infers that the phytoplankton communities were stressed by cobalamin-limited condition. However, we identified again that P. nitzschia antarctica adjust under this condition better than other taxon, which could be confirmed with higher expression levels of methionine synthase I (cobalamin-dependent) (Ellis et al., 2017). That is, P. nitzschia antarctica might succeed to obtain cobalamin in the water by competing with other algal species and used it for enhancing the activity of this enzyme (Banerjee & Matthews, 1990). Thus, niche adaptation in P. nitzschia antarctica includes a strong ability to respond to iron-deficient condition and cobalamin-limited condition, which may be a leading cause for P. nitzschia antarctica dominance in the RSP phytoplankton communities. The better adjustment of P. nitzschia antarctica than other taxon was also detected regarding the light variation with distinctly higher photosystem II stability/assembly factor gene expression. This also infers more photosynthetic activities of P. nitzschia antarctica in the phytoplankton community with better acclimation to limited environmental conditions.
We further analyzed the correlation between the key-genes expression levels of P. nitzschia antartica and the environmental variations (Table. 6) to examine if a specific environmental factor controlled transcript abundance for a collection of genes that are involved in acclimation response to that factor. The key-genes in the functional profile of P. nitzschia antartica changed their overall expression patterns gradually over the sampling period; the key-gene expression levels and patterns are higher and similar in sample 5 and 6 than others (Fig. 10).
The expression levels of most key-genes do not linearly correlated with the variations of major limiting factors, but might be related with general environmental parameters such as temperature, salinity or fluorescence (Table. 6).
The key-genes expression patterns of P. nitzschia antarctica showed the potential acclimations to limited cobalamin and iron availability with abundantly transcribed genes related to these parameters (Fig. 10). Since we did not measure cobalamin concentrations, the correlation between the cobalamin availability and the expression level of related genes could not be concluded. However, we were able to identify that the extremely low iron concentration throughout the RSP caused the higher expression of iron-related genes. We also assumed, based on the linear correlation between the cobalamin biosynthesis protein gene expression level and the iron concentrations (Table. 6), the possibility of a combination of iron and cobalamin regulating the key-genes expression patterns (Bertrand et al., 2007(Bertrand et al., , 2011(Bertrand et al., , 2015. We also detected the highest expression of the flavodoxin gene at station 3 although the highest dissolved iron concentration was observed at that station. This may be because the dissolved iron at the surface, supplied from sea ice melting and/or degrading of organic matters by bacterial organisms (Gerringa et al., 2015;Mack et al., 2017;Sedwick et al., 2000), had not been taken up by P.
nitzschia antarctica when the sample was taken. Thus, the functional profiles of P.
nitzschia antarctica showed stress from low iron conditions by expressing higher levels of flavodoxin genes, despite high iron supplies at the surface.
The genes related to carbon fixation (fructose-bisphosphate aldolase class 1 and photosystem II stability/assembly factor) inversely correlated with Photosynthetically Active Radiation (PAR) variation (Table 6). This might indicate that increased Photosynthetically Active Radiation (PAR) caused photoinhibition (van Hilst & Smith, 2002) on the photosynthetic capacity of P. nitzschia antarctica.
In sum, we see individual species showed different acclimations to their shared environment.

CONCLUSIONS
This study demonstrates that critical environmental parameters, such as iron and light affect bulk primary production and act as stressors to the phytoplankton communities. Individual algal taxa in the phytoplankton communities showed different acclimations to their shared environmental condition, which likely determined relative abundances of each algal taxon. The phytoplankton communities during the bloom in the RSP in 2013-2014 were dominated by P. nitzschia sp. and acclimation of P. nitzschia sp. to iron and cobalamin co-limitation as well as to light variation favored this taxon.
Thalassiosira sp. and Phaeocystis antarctica were stressed more severely by low iron condition than by the cobalamin limitation. Compared to these two taxa, Fragilariopsis sp. and P. nitzschia sp. displayed better adjustments to iron and cobalamin co-limitation. Acclimation to low light condition was mostly detected in Fragilariopsis sp. whereas acclimation to low iron conditions was observed to a higher degree in P. nitzschia sp. than Fragilariopsis sp. We also detected nichepartitioning between diatom species and the haptophyte Phaeocystis antarctica along depth gradients. Overall, our findings indicate that species specific traits underpin differences in acclimation responses in coexisting phytoplankton taxa.