Molecular underpinnings and biogeochemical consequences of enhanced diatom growth in a warming Southern Ocean.

The Southern Ocean (SO) harbors some of the most intense phytoplankton blooms on Earth. Changes in temperature and iron availability are expected to alter the intensity of SO phytoplankton blooms, but little is known about how these changes will influence community composition and downstream biogeochemical processes. We performed light-saturated experimental manipulations on surface ocean microbial communities from McMurdo Sound in the Ross Sea to examine the effects of increased iron availability (+2 nM) and warming (+3 and +6 °C) on nutrient uptake, as well as the growth and transcriptional responses of two dominant diatoms, Fragilariopsis and Pseudo-nitzschia We found that community nutrient uptake and primary productivity were elevated under both warming conditions without iron addition (relative to ambient -0.5 °C). This effect was greater than additive under concurrent iron addition and warming. Pseudo-nitzschia became more abundant under warming without added iron (especially at 6 °C), while Fragilariopsis only became more abundant under warming in the iron-added treatments. We attribute the apparent advantage Pseudo-nitzschia shows under warming to up-regulation of iron-conserving photosynthetic processes, utilization of iron-economic nitrogen assimilation mechanisms, and increased iron uptake and storage. These data identify important molecular and physiological differences between dominant diatom groups and add to the growing body of evidence for Pseudo-nitzschia's increasingly important role in warming SO ecosystems. This study also suggests that temperature-driven shifts in SO phytoplankton assemblages may increase utilization of the vast pool of excess nutrients in iron-limited SO surface waters and thereby influence global nutrient distribution and carbon cycling. Copyright   2021 the Author(s). Published by PNAS.

The Southern Ocean (SO) occupies less than 10% of Earth’s ocean surface area but plays a major role in driving global climate and biogeochemical cycles. It connects the Pacific, Atlantic, and Indian ocean basins, supplies nutrients to lower latitudes, and absorbs a considerable portion of global heat and CO2 (1, 2). SO ecological processes are driven by seasonally productive phytoplankton assemblages that require sufficient light, macronutrients (e.g., nitrogen, phosphorus), micronutrients (e.g., iron, vitamin B12), and suitable temperatures to grow. These phytoplankton sequester atmospheric CO2 through the biological pump, sustain food webs, and influence the stoichiometry of nutrient supply to lower latitudes. Despite large regional variations, climate change models project overall warming in the SO, with current average sea surface temperatures (0 to 4 °C) predicted to increase by 1 to 2 °C in 2100 and ∼6 °C by 2300 (2–5). Additionally, small increases in iron supply (e.g., increased iron flux, by ∼0.01 nM and ∼0.02 nM day−1 by 2100 and 2300, respectively) are projected for the SO (2, 4) (), but large uncertainties remain in model projections of SO iron supply, which are complicated by the diversity of supply mechanisms and other physical/chemical factors (6, 7). Temperature and iron can be major drivers of SO phytoplankton growth (8, 9), yet we still have a limited understanding of how changes in these factors will influence the cellular mechanisms that govern phytoplankton dynamics and biogeochemistry in the SO.

Low iron availability limits phytoplankton growth and is the primary cause of the high-(macro)nutrient low-chlorophyll (HNLC) conditions in the SO (9, 10). Despite this, diatoms are highly successful in these environments and contribute substantially to primary productivity throughout the region. Low-iron–adapted diatoms utilize several strategies to survive chronic, episodic, or seasonal iron limitation. For example, they can reduce photosynthetic iron demand by substituting iron-containing ferredoxin with flavodoxin (11) or cytochrome c with plastocyanin (12). Increased light harvesting cross-section has also been observed in several iron-stressed diatoms, allowing for more light energy capture to facilitate photosynthesis (13, 14). Diatoms may also use iron-free rhodopsin to supplement photosynthetic energy capture (15), have transport mechanisms that can be up-regulated to maximize iron acquisition (16–18), and some species contain iron storage proteins (ferritin) to buffer the sporadic availability of this micronutrient (19).

Temperature influences intracellular transport processes and enzymatic turnover rates and ultimately plays an important role in controlling phytoplankton growth and physiology (20). Temperature has also been shown to limit SO phytoplankton growth in the field and the laboratory. Warming incubations that simulate projected sea surface temperature changes by the year 2100 to 2300 (0 °C versus +4 °C) in iron-limited, diatom-dominated SO microbial communities resulted in increased nutrient drawdown and improved the growth of several diatom groups (21), with similar results reported in iron-limited laboratory diatom cultures grown under the same temperature treatments (22). Warming (up to a critical threshold) improves iron use efficiency (23–25), increases the turnover rate of iron-containing enzymes such as nitrate reductase (26, 27), and may thus cause increased utilization of SO nitrate even in the absence of iron addition (7, 28). Some Antarctic diatoms such as Fragilariopsis cylindrus also exhibit reduced cell size under elevated temperature, which can decrease cellular iron quotas, increase surface area to volume ratios, and facilitate nutrient uptake and growth (14). Other studies have shown a reduction in the upper thermal limit of Antarctic diatom growth (e.g., centric diatoms Chaetoceros neglectus, Chaetoceros flexuosus, Thalassiosira antarctica) under reduced iron availability (29, 30), and some diatoms like the pennate F. cylindrus still require iron supplementation (≥10 nM EDTA-buffered iron) to benefit from warming (3 °C and 6 °C compared to 1 °C), even when iron-conserving mechanisms are used (14). Additionally, concurrent increases in temperature and iron have been shown to have larger synergistic effects on growth and nutrient utilization compared to the individual effects of each factor (7, 21, 22). Here, we define “synergistic” as a nonlinear, multiplicative response, which exceeds the sum of the individual effects of iron and temperature (31).

Despite shared adaptations across various taxa, diatoms are an extremely diverse group of phytoplankton. Within the SO, major diatom groups have different morphological and physiological traits, with unique optimal growth temperatures and different tolerances to low iron (13, 29, 32–35). The molecular mechanisms that underpin these differences are still poorly understood and are overlooked by the majority of current marine ecosystem models, which consider all diatoms as one group (36). This oversimplification cannot capture how environmentally mediated changes in diatom community composition and structure (e.g., communities with different sinking rates) may affect important ecological processes, as well as nutrient and carbon biogeochemistry (37).

Here, we present an experimental manipulation study performed at the sea ice edge (IE) in the Ross Sea of the SO. Coastal areas of the SO, including the sea IE, are highly productive, and the Ross Sea alone contributes to ∼25% of all SO primary productivity (38), making this area of the SO ecologically and biogeochemically important. This region has also been shown in previous growing seasons [e.g., 2013, 2014 (39)] to be iron limited, suggesting that despite its coastal nature, phytoplankton here experience similar nutritional stressors found throughout the Ross Sea and broader SO. We investigated the growth, nutrient drawdown, and transcriptional responses of an SO microbial community to warming and changes in iron availability, mimicking changes that are projected to occur between the years 2100 and 2300 (2–5). We show that nutrient drawdown increases in response to temperature and iron, both independently and synergistically, and that two closely related diatom groups, Fragilariopsis spp. and Pseudo-nitzschia spp., respond differently to changes in temperature and iron. These two bloom-forming taxa (40) contribute substantially to diatom assemblages in the SO and strongly influence the ecology and biogeochemistry of the region (21, 41, 42). Pseudo-nitzschia cell numbers increased more than Fragilariopsis under warming alone, while the latter required iron addition to benefit from increased temperature. Our metatranscriptome data suggest that the growth of Pseudo-nitzschia under warming is likely facilitated by temperature-responsive light harvesting and iron management strategies that were not utilized by Fragilariopsis. This ability to grow and draw down nutrients in the face of warming, regardless of iron availability, could have major implications for the ecological and biogeochemical response of the HNLC SO to a changing climate.

Results and Discussion

We incubated SO surface water under constant growth-saturating light (65 to 85 µE m−2 sec−1) with and without 2 nM iron addition at three temperatures (−0.5 °C, 3 °C, 6 °C) to simulate increased iron availability and a range of SO surface water temperatures, spanning from current to projected conditions 100 to 300 y from now (2–4) (). We then examined the microbial community after 24 h (T1) and again after 5 (T5) and 7 d (T7). Unless otherwise stated, a temperature effect (i.e., effect of increased temperature) on the various measured parameters refers to an effect at both 3 °C and 6 °C (see for details).

Increasing temperature and iron separately caused a significant but small increase in community (>0.7 μm) primary productivity measured as bicarbonate uptake (Fig. 1 and , two-way ANOVA, temperature effect P = 1.1 × 10−6, Fe effect P = 2.4 × 10−6), as well as an increase in community nitrate uptake rates (Fig. 1 and , two-way ANOVA, temperature effect P = 0.001, Fe effect P = 7.9 × 10−5), nitrate consumption (Fig. 1 and , two-way ANOVA, temperature effect P = 2.9 × 10−11, Fe effect P = 1.2 × 10−11), and phosphorus consumption (Fig. 1 and , two-way ANOVA, temperature effect P = 2.3 × 10−6, Fe effect P = 1.0 × 10−6). Incubations under concurrent warming and iron addition showed a synergistic increase in primary productivity (Fig. 1, two-way ANOVA, temperature × Fe effect P = 4.7 × 10−5), nitrate uptake rates (Fig. 1, two-way ANOVA, temperature × Fe effect P = 0.007), nitrate consumption (Fig. 1 and , two-way ANOVA, temperature × Fe effect P = 6.0 × 10−9), and phosphorus consumption (Fig. 1 and , two-way ANOVA, temperature × Fe effect P = 7.5 × 10−5). This synergistic iron–temperature effect has been observed previously (21, 28) and suggests that future warming conditions can strongly influence SO productivity and nutrient drawdown, especially if combined with a concurrent increase in iron availability.

Fig. 1.

(A) Absolute bicarbonate and nitrate uptake rates at T0 (prior to the incubations) and T7 for each temperature and iron condition. Uptake rates were measured in duplicate from a single incubation bottle per treatment. (B) Dissolved nitrate and phosphate concentrations in triplicate incubation bottles at T0 and T7. Phosphate concentration is scaled (16×) for better visual comparison with nitrate. (C) Initial (T0) cell counts of eukaryotic taxa identified by light microscopy, n = 3, taxa are arranged from low to high abundance. The large variation in Phaeocystis cell counts at T0 likely resulted from random variability in colony abundance and size in the relatively small-volume samples used for microscopic cell counts. Cell counts for other taxonomic groups under the different temperature and iron treatments are shown in . (D) Measurements of Fragilariopsis and Pseudo-nitzschia cell counts at T0 and T7 in triplicate incubation bottles for each temperature and iron condition. (E) Pie charts representing percent contribution of taxa to the total reads mapped to ORFs and to the total 18S rRNA Amplicon Sequence Variant (ASV) abundance at T0. “Other” represents all other taxonomic groups, including prokaryotes and viruses. (F) Percent contribution of Fragilariopsis and Pseudo-nitzschia to the total ASV abundance at T0 and T5 for each temperature and iron condition. Each point represents triplicate means; error bars represent ±1 SD and fall within the bounds of the symbol when not visible.

Microscopy observations showed that Phaeocystis, dinoflagellates, ciliates, Fragilariopsis, and Pseudo-nitzschia were among the most abundant eukaryotic taxa at the beginning of the incubation experiment (the in situ community) and after temperature and iron incubations (Fig. 1 and ). We focused our analyses here on the responses of the two most abundant diatom groups in our samples, Fragilariopsis (identified by light microscopy and 18S ribosomal RNA sequencing as mostly Fragilariopsis kerguelensis and Fragilariopsis cylindrus) and Pseudo-nitzschia (identified by light microscopy, 18S ribosomal RNA sequencing, and a protein Basic Local Alignment Search Tool (BLAST-p) analysis of Pseudo-nitzschia transcriptomes as mostly Pseudo-nitzschia subcurvata) (reference ). Pseudo-nitzschia but not Fragilariopsis cell counts increased significantly under warming without added iron (Fig. 1, two-way ANOVA, temperature effect P = 0.003 and P = 0.5, respectively) and showed the largest increase in abundance throughout the incubations, compared to other plankton groups (). Additionally, concurrent warming and iron supplementation caused an increase in Fragilariopsis (Fig. 1, two-way ANOVA, temperature × Fe effect P = 0.001) and a much larger synergistic increase in Pseudo-nitzschia cell counts (Fig. 1, two-way ANOVA, temperature × Fe effect P = 1.13 × 10−6). This highlights the ability of Pseudo-nitzschia to increase its growth under warmer conditions, regardless of iron availability.

We used metatranscriptomics to comprehensively capture gene expression patterns and examine the molecular processes underpinning the distinct growth responses of Fragilariopsis and Pseudo-nitzschia to shifts in iron availability and temperature (Fig. 2). Metatranscriptomics allows for the interrogation of cellular pathways utilized across various taxa and can be used to infer metabolic states under different environmental conditions (39, 43, 44). Here, we identified transcripts belonging to viruses, prokaryotes, and eukaryotes (); with dinoflagellates, ciliates, Pseudo-nitzschia, and Fragilariopsis contributing 37% of the total reads mapped to open reading frames (ORFs) (Fig. 1). The contribution of Phaeocystis to the metatranscriptome was lower than anticipated based on microscopy, possibly due to difficulty in harvesting cells/colonies without rupturing them (45). Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthology (KO) term enrichment analysis on ORFs belonging to Fragilariopsis and Pseudo-nitzschia showed that pathways corresponding to photosynthesis, nitrogen metabolism, and genetic information processing/translation were highly differentially expressed in both taxa following temperature and iron increase (). We explored these pathways in more detail by grouping individual ORFs into clusters of similar sequences using Markov clustering (MCL) (46) (). Several clusters contained differential expression patterns that were similar in Fragilariopsis and Pseudo-nitzschia under iron addition (63 clusters) and temperature increase (212 clusters). However, a larger number of clusters were uniquely differentially expressed in either Fragilariopsis or Pseudo-nitzschia under these conditions (Fig. 3), suggesting that Fragilariopsis and Pseudo-nitzschia utilize different molecular strategies to respond to changes in temperature and/or iron.

Fig. 2.

Schematic representations of Fragilariopsis and Pseudo-nitzschia cells showing cellular processes, with each process comprised of several transcript clusters (MCL clusters). Photosynthesis: Lhcf = light harvesting complexes-f, OEC = oxygen evolving complex, Cyt b6f complex = cytochrome b6f complex, PCYN = plastocyanin, Cyt c6 = cytochrome c6. Nitrogen Assimilation: NRT = nitrate transporter, NR = nitrate reductase, NiT = nitrite transporter, NiR = nitrite reductase, AMT = ammonium transporter, GOGAT = glutamine oxoglutarate aminotransferase cycle. Iron Regulation: ISIP = iron starvation induced protein. Each row in a heatmap represents one MCL, each column represents a temperature and iron treatment at T5 with each block representing three biological replicate means with individual replicates shown in . Heatmaps were constructed using taxon-normalized RPKM values. Empty heatmap placeholders represent clusters found in Fragilariopsis but not Pseudo-nitzschia. Arrows represent energy/electron flow in photosynthetic light reactions and steps involved in nitrogen assimilation using nitrate or ammonium.

Fig. 3.

(A) Venn diagram showing the number of transcript clusters that are significantly up- (black) or down- (blue) regulated with iron addition or temperature increase in Fragilariopsis and Pseudo-nitzschia. The intersection represents clusters up- or down-regulated in both Fragilariopsis and Pseudo-nitzschia. The number outside of the circles represents the number of clusters from both taxa that were not differentially expressed. In both taxa, warming caused more clusters to be differentially expressed than iron addition. (B) Differential expression (DE) of various clusters under increased temperature. DE was calculated using the quasi-likelihood test (glmQLFTest) in EdgeR, and fold change was calculated for high-temperature (6 °C) versus low-temperature (−0.5 °C) treatments at T5. (C) DE of various clusters after iron addition. DE was calculated using the quasi-likelihood test (glmQLFTest) in EdgeR, and fold change was calculated for iron-added versus no iron-added treatments at T5. In both B and C, positive and negative Log2 fold change values represent up- and down-regulation, respectively. Filled circles are clusters with statistically significant DE (adjusted P value < 0.05). Point size represents total normalized transcript abundance under all iron and temperature treatments.

Photosynthesis.

Iron addition resulted in up-regulation of iron-containing cytochrome b6f complex transcripts and down-regulation of flavodoxin transcripts in both Fragilariopsis and Pseudo-nitzschia (Figs. 2 and 3). This indicates that both diatoms in our study were undergoing some degree of iron stress at the time of sampling; ambient dissolved iron concentrations at the time of sampling were growth limiting at 0.47 nM (47), and iron-containing photosynthetic processes including the electron transport chain can account for up to 40% of cellular iron quotas (48) and are typically down-regulated under low iron (49, 50). Here, we did not detect the expression of iron-dependent electron acceptor ferredoxin (PetF) in Pseudo-nitzschia, regardless of iron status. Antarctic Pseudo-nitzschia species may have constitutively reduced dependence on PetF (51, 52) to minimize photosynthetic iron demand.

Light-Harvesting Complexes.

Most members of the light-harvesting complex (LHC) superfamily primarily absorb and direct light energy to photosynthetic reaction centers where charge separation occurs, but some of them (Lhcx clade) are also involved in stress responses and photoprotection (53, 54). Expression patterns of transcripts belonging to the major Lhcf group and photosystem I (PSI)-associated Lhcr groups were markedly different between the two taxa under the different temperature and iron treatments (Fig. 4). In Fragilariopsis, there was a general pattern of decreased expression with warming and marked up-regulation with iron addition. In contrast, Pseudo-nitzschia showed a pattern of increasing expression with elevated temperatures. Up-regulation of Lhcf and Lhcr groups contributes to increased light harvesting efficiency and provides more energy for downstream temperature-dependent catalytic processes. This also alleviates photosynthetic iron demand by reducing the number of iron-containing photosynthetic components required to process light energy (55). Increased light harvesting cross-section has been measured under warming (3 °C and 6 °C compared to 1 °C) in iron-stressed F. cylindrus in the laboratory (14). However, these environmental data indicate that Pseudo-nitzschia may be even better equipped to use this mechanism to support growth under low iron in the SO, provided that the LHC up-regulation observed here reflects an increase in light harvesting cross-section, which remains speculative at this time. Several Lhcf and Lhcr clusters were also up-regulated in both Fragilariopsis and Pseudo-nitzschia after iron addition. Given that iron addition is expected to decrease light harvesting cross-section (13, 14, 56, 57), it might be predicted that iron addition could decrease the demand for Lhcf and Lhcr expression. However, this impact appears to be overridden by increased demand for LHC expression resulting from iron-induced increases in photosynthetic unit (PSU) abundance. Additional measurements of light absorption cross-section and photosynthetic efficiency could provide further insight into how light harvesting responds to changes in LHC expression under different iron and temperature conditions.

Fig. 4.

Heatmaps of MCL clusters of LHCs Lhcf, Lhcr, Lhcx, and Lhcz (reference ) in Pseudo-nitzschia and Fragilariopsis measured after 24 h (T1) and 5 d (T5) of incubation under the various iron and temperature treatments. Heatmaps were constructed using taxon-normalized RPKM values. I.E represents IE samples; T0 represents in situ samples processed immediately before any incubations. Each block is one biological replicate measurement. Black and white up/down pointing triangles represent transcripts that were significantly (glmQLFTest-EdgeR P < 0.05) up- or down-regulated due to warming or iron addition (respectively) at T5. Stars represent transcripts that were significantly (glmQLFTest-EdgeR P < 0.05) up-regulated due to an interactive iron–temperature effect at T5.

Each transcript cluster belonging to the Lhcx clade was down-regulated after 24 h of incubation at all temperatures and iron conditions in both Fragilariopsis and Pseudo-nitzschia (Fig. 4). Lhcxs are highly sensitive to environmental change, including fluctuations in light and nutrient levels (53, 54, 58). Given these expression patterns and previous observations of Lhcx regulation, the Lhcx transcripts detected here may be responding to changes in the light levels that cells would have undergone during sample collection—brief exposure to unattenuated daylight during sampling, followed by lower light intensities in the incubations. Notably, expression of these Lhcx clades showed only minimal responses to iron addition.

Plastocyanin.

Plastocyanin is a copper-containing protein that acts as an electron shuttle in the electron transport chain and plays an important role in reducing iron requirements in phytoplankton by substituting for cytochrome c6 (12, 47, 59). Here, iron addition had no significant effect on plastocyanin transcript expression, while warming caused its up-regulation in Pseudo-nitzschia and down-regulation in Fragilariopsis (Fig. 3). Plastocyanin expression has been observed to be insensitive to iron availability in low-iron–adapted diatoms (43). However, our data suggest that Pseudo-nitzschia plastocyanin transcripts are phylogenetically distinct from those of Fragilariopsis and may play a temperature-responsive role to support growth under elevated temperatures (). Plastocyanin transcript abundance in Pseudo-nitzschia was higher at 3 °C and 6 °C compared to −0.5 °C at T1, showing a rapid response to warming (). This, in combination with the strong and immediate response of LHCs may give Pseudo-nitzschia a photosynthetic advantage over Fragilariopsis across a wide range of likely warming scenarios in the SO.

Nitrogen Metabolism.

KO term enrichment analysis showed that transcripts encoding nitrogen metabolism were enriched in Pseudo-nitzschia and depleted in Fragilariopsis after temperature increase (). Nitrogen is required for protein synthesis and would be necessary to support the observed increase in Pseudo-nitzschia cell counts under warming, and notably, dissolved nitrogen drawdown was enhanced due to warming alone and elevated further upon iron addition (Fig. 5 ). However, nitrogen acquisition and assimilation from nitrate requires iron (60), and the observed Pseudo-nitzschia growth under elevated temperature would necessitate iron-economic nitrogen metabolism. Our data show that ammonium transporter transcripts were up-regulated in Pseudo-nitzschia but not in Fragilariopsis under warming (Figs. 3 and 5 ), indicating that Pseudo-nitzschia is better equipped to utilize ammonium as an additional and/or alternative nitrogen source in response to elevated nitrogen demand, which is likely a result of increased growth in response to warming. Elevated ammonium transporter expression would provide Pseudo-nitzschia with an additional nitrogen source at a lower iron cost for use in amino acid synthesis and growth (60–62). In fact, community ammonium drawdown at 6 °C was greater than at lower temperatures, regardless of iron status (Fig. 5), consistent with the notion that ammonium use at high temperature and low iron could have contributed to the success of Pseudo-nitzschia. The ability to uptake and utilize ammonium rapidly, relative to nitrate, also allows diatoms to grow faster when ammonium is available (63, 64). However, the high Pseudo-nitzschia growth we observed cannot be supported by ammonium alone; total community ammonium drawdown comprised only 25.2 ± 2.4% of dissolved inorganic nitrogen (nitrate + ammonium) drawdown in the 6 °C no added iron treatment and 7.9 ± 1.0% of dissolved inorganic nitrogen drawdown in the added iron 6 °C treatment (Fig. 5 ). Nitrate drawdown was responsible for the majority of dissolved nitrogen uptake regardless of iron status, though significantly (two-way ANOVA, P = 9.6 × 10−11) more so under elevated iron. This suggests that Pseudo-nitzschia may have utilized additional parsimonious mechanisms such as internal protein recycling to further reduce nitrogen requirements and thereby reduce iron demand (65). While nitrate drawdown was elevated at high temperatures both with and without added iron (Fig. 5), neither Pseudo-nitzschia nor Fragilariopsis changed their expression of nitrate transporters in response to elevated temperature. Both diatom groups up-regulated nitrate transporters in response to iron addition regardless of temperature (Figs. 3 and 5 ). Similarly, nitrate reductase transcript expression was up-regulated in response to iron addition but not elevated temperature (Figs. 3 and 5 ). However, the observed elevation in nitrate drawdown and nitrate uptake rates with warming (Figs. 1 and 5), in the absence of a change in gene expression, is consistent with the well-characterized temperature dependence of nitrate reductase activity (26). This suggests that the enhanced nitrate assimilation under elevated temperature was indeed accomplished without the increase in iron demand required to produce additional copies of active nitrate reductase.

Fig. 5.

(A and B) Total nitrate and ammonium drawdown from beginning to end of the experiment (difference in concentration between T7 and T0). (C–H) Taxon-normalized RPKM expression values for clusters belonging to ammonium transporters (AMT), nitrate transporters (NT), and nitrate reductase (NR) in both Pseudo-nitzschia and Fragilariopsis prior to incubation (T0) and at T5 with and without added iron at −0.5, 3, and 6 °C.

Genetic Information Processing/Translation.

Protein synthesis is an energetically costly process that is required for metabolism and growth. Increased kinetic energy due to warming enhances translation efficiency, reduces the total RNA per cell, and reduces the number of ribosomes required to maintain protein synthesis rates (66). Consistent with this response, we observed that several transcript clusters corresponding to ribosomes were down-regulated in both Fragilariopsis and Pseudo-nitzschia under warming (Figs. 2 and 3). Warming also caused up-regulation of transcripts encoding ribosomal recycling proteins only in Pseudo-nitzschia (Fig. 2), which can reduce nitrogen demand and the energetic costs associated with ribosomal synthesis. While the reduction in phosphorus-rich ribosome abundance is thought to increase cellular N:P ratios, leading to increased nitrogen demand in warmer ocean regions (66), our bulk nutrient drawdown ratios () show that N:P drawdown is not significantly influenced by warming (two-way ANOVA, P = 0.07). This suggests that the increasing dominance of diatoms, with their considerably lower N:P ratios than other SO plankton groups such as Phaeocystis (22, 67), is likely an overall driver of N:P ratios and can counterbalance the expected reduction in cellular phosphate demand at high temperature from reduced ribosome content. It is also possible that temperature changes alone, without additional interactive effects due to changes in other variables such as CO2 and light, might not cause a strong increase in N:P ratios (68).

Micronutrient Acquisition, Trafficking, and Storage.

Increased iron uptake, storage, and intracellular transport are strategies to improve growth under low iron availability. In the laboratory, Zhu et al. (22) report higher iron uptake rates in P. subcurvata than in F. cylindrus under warming. Our data show down-regulation of iron stress-induced proteins (ISIP1, ISIP2A, ISIP3) in both Pseudo-nitzschia and Fragilariopsis after iron addition, and ISIP2A up-regulation only in Pseudo-nitzschia after warming (Figs. 2 and 3). ISIPs are involved in iron acquisition and storage systems in diatoms, and specifically, ISIP2A contributes to nonreductive iron uptake (16–18, 50). No previous studies have inspected ISIP2A response to temperature, but it is likely that the apparent temperature response of ISIP2A contributes to the ability of Pseudo-nitzschia to acquire and utilize iron to support growth under warming. Additionally, transcripts for ferritin, an iron storage protein (19), were significantly up-regulated only in Pseudo-nitzschia after iron addition (Fig. 3). Several studies have reported ferritin up-regulation in Pseudo-nitzschia under iron-replete conditions (44, 69), and Fragilariopsis can also utilize ferritin to store iron (19). However, consistent with low Fragilariopsis ferritin expression under iron-replete conditions in the field (70), our study shows that Fragilariopsis ferritin transcripts were not significantly up-regulated after iron addition (Fig. 3). These trends suggest that in addition to reducing iron demand, Pseudo-nitzschia was better able to increase iron uptake and utilized stored ferritin-bound iron to support growth under warming. This suggests that Pseudo-nitzschia may be particularly well suited to environments with episodic iron availability.

In addition to iron, cobalamin has also been shown to be a critical micronutrient driver of phytoplankton growth and community composition in the SO (39). Fragilariopsis and Pseudo-nitzschia’s cobalamin acquisition and stress response machinery also showed markedly different responses to iron and warming (Fig. 3 and ), suggesting that these two diatoms may employ different strategies to cope with low cobalamin availability ().

Synergistic Responses and Gene Expression.

The observed synergistic iron–temperature effects on growth and nutrient drawdown were not accompanied by a synergistic effect on gene expression (). For example, we did not observe an interactive iron–temperature effect on the expression of genes involved in nitrogen uptake, yet nitrate and ammonium drawdown increased synergistically due to iron and temperature increase (Fig. 1). It is unlikely that synergistic nutrient drawdown was influenced by synergistic gene expression in taxa other than Fragilariopsis and Pseudo-nitzschia. Except for the prokaryotic Alteromonadales group, we observed relatively little or no interactive iron–temperature effect on gene expression in all the taxa we assessed (). This suggests that improved enzymatic efficiency under warming can accelerate metabolic functions, even without a proportional increase in transcript abundance. It is worth noting that, without extensive genetic manipulation and biochemical investigations using monocultures in controlled conditions, observational studies like this lack the ability to disentangle causal relationships between gene expression patterns and physiology/growth. Nevertheless, observations of gene expression trends under different environmental conditions provide insight into the cellular mechanisms that facilitate growth and those which underpin differences between ecologically important groups.

Ecological and Biogeochemical Relevance.

Most recent model projections show warming trends across the SO with sea surface temperatures increasing by 1 to 2 °C by 2100 and ∼6 °C by 2300 (2, 4, 71). Fragilariopsis grows in a wide range of environments throughout the SO but is most successful in high-salinity, high-iron sea ice while Pseudo-nitzschia is most successful in warmer low-salinity, low-iron meltwaters (33, 72, 73). The growth and gene expression patterns we observed in both diatoms may then be driven by niche-specific adaptations due to competition for resources. Our results are consistent with a 2013 incubation study conducted at the IE in the Ross Sea where Pseudo-nitzschia dominated the community after warming (28), suggesting that Antarctic Pseudo-nitzschia are better equipped to dominate in a warmer (4 °C compared to 0 °C) ocean and even more so if iron availability increases. A shift into a Pseudo-nitzschia dominated community raises concerns for domoic acid (DA) production and its toxic effects in mammals and other trophic levels (74, 75). A search for DA biosynthesis transcripts (76) did not yield strong evidence for expression of the DA biosynthesis gene cluster in this experiment (). However, other studies have measured DA concentrations as high as 220 ng L−1 in the SO (77, 78), but the temporal, spatial, and environmental factors that elicit DA production remain unclear. Further work is required to investigate whether resident Pseudo-nitzschia can produce the toxin under conditions not examined in our study or if DA producing Pseudo-nitzschia are likely to migrate into a warmer SO.

A change in SO phytoplankton community structure could have important biogeochemical ramifications on local and global scales. Changes in cell size, frustule thickness, and metabolic functions under changing temperature and iron conditions could influence ecosystem and biogeochemical processes by altering sinking rates, grazing pressures, and cellular nutrient quotas (37, 42, 68). Warming-induced increases in the expression of ISIP2A, as well as iron-conserving plastocyanin, LHCs, and nitrogen assimilation processes suggest the involvement of a suite of mechanisms that support growth under warming. These cellular changes resulted in increased nutrient drawdown in our multiday experiment, even in the face of iron limitation. Such warming-enhanced SO nutrient drawdown could have profound consequences for northward nutrient supply, especially if cells are able to utilize these parsimonious processes over weeks or months and if the trends we observe in this coastal experiment hold true across broad swaths of the SO. As more nutrients such as nitrogen and phosphorous are consumed by phytoplankton and exported out of the surface ocean around Antarctica, less nutrients will be available to fuel productivity at lower latitudes (2). Here, we have identified molecular mechanisms that appear to enable diatoms, particularly Pseudo-nitzschia, to accomplish enhanced growth and nutrient drawdown with rising temperatures under low-iron conditions. These mechanisms, up-regulation of iron-conserving photosynthetic processes, utilization of iron-economic nitrogen assimilation mechanisms, and increased iron uptake and storage, underpin what may be a pattern of increasing nutrient utilization in the SO and decreasing availability of nutrients for lower latitudes.

Materials and Methods Summary

Seawater was collected at a 3-m depth at the IE from McMurdo Sound, Antarctica on January 15, 2015 (165°24.7985’E, 77°37.1370’S) using previously described trace-metal-clean techniques (39). Triplicate bottles of each treatment (temperature: −0.5 °C ± 0.2 °C, 3 ± 0.5 °C, and 6 ± 0.5 °C; iron: not added [+0 nM], added [+2 nM]) were incubated indoors at a constant irradiance of 65 to 85 uE m−2 sec−1. This incubation irradiance is generally growth saturating for Antarctic phytoplankton (57, 79) and falls within observational bounds of mean mixed layer light levels (80, 81). Metatranscriptome samples were taken at the sea IE, in the laboratory immediately following bottle incubation setup (T0), on January 16 (T1), and again on January 20 (T5). Cell counts were measured on T0 and January 22 (T7). Dissolved nutrients (nitrate, phosphate, silicate) were measured on T0, T1, T3, T5, and T7, and ammonium was measured on T0 and T7. Primary productivity (bicarbonate uptake) and nitrate uptake rates were measured on T0, T1, T3, and T7 ().

For metatranscriptomics, the microbial community was harvested on 0.2-μm Sterivex filters, total RNA was extracted using TRIzol reagent (Thermo Fisher Scientific), and ribosomal RNA was removed with Ribo-Zero Magnetic kits (Illumina). The resulting messenger RNA (mRNA) was purified and subjected to amplification and complementary DNA (cDNA) synthesis, using the Ovation RNA-Seq System V2 (TECAN). One microgram of the resulting cDNA pool was fragmented to a mean length of 200 base pairs, and libraries were prepared using Truseq kit (Illumina) and subjected to paired-end sequencing via Illumina HiSeq. Illumina paired reads were filtered to eliminate primer sequences and quality trimmed to Phred Q33, and rRNA was identified and removed using riboPicker (82). Transcript contigs were assembled de novo using CLC Assembly Cell (https://www.qiagenbioinformatics.com/), and ORFs were predicted using FragGeneScan (83). ORFs were annotated for putative function using hidden Markov models and BLAST-p against PhyloDB (39) and filtered to eliminate those with low mapping coverage (<50 reads total over all samples) and proteins with no BLAST hits and no known domains. ORFs were assigned to taxonomic groups of interest () based on best Lineage Probability Index (LPI) taxonomy (39, 84). Reads per kilobase mapped (RPKM) expression values for each ORF were calculated and taxon normalized using a normalization factor representing the summed taxonomic group contribution to total nuclear-assigned reads per sample. ORFs were clustered into orthologs and protein families using MCL (46). Group normalized cluster (average/total) RPKM expression values were calculated by pooling the taxon-normalized expression values for each group within a cluster. Cluster annotations were aggregated by annotation type (KEGG, KO, euKaryotic Orthologous Groups [KOG], KOG class, Pfam, TIGRfam, Enzyme Commission [EC], Gene Ontology [GO]), and a single annotation was chosen to represent each cluster based on the lowest Fisher's exact test P value (fisher.test in R).

Differential gene expression at T5 (foldchange magnitude and adjusted P value) was calculated using empirical Bayes quasi-likelihood F-tests (glmQLFTest in edgeR) on taxon-normalized expression values to account for the change in abundance under the different iron and temperature treatments. A P value cut off of 0.05 was used for statistical significance. Iron effect was determined by comparing all −Fe treatments at T5 to all +Fe treatments at T5. Temperature effect was determined by comparing all −0.5 °C to 3 °C treatments at T5 and −0.5 °C to 6 °C at T5. A gene was considered up-regulated with temperature if it was up-regulated in both −0.5 °C versus 3 °C and −0.5 °C versus 6 °C tests.

Stable isotope tracer techniques, using 15N-labeled nitrate and 13C-labeled bicarbonate (Cambridge Isotope Laboratories), were used to determine uptake rates, similar to methods previously described in Ross Sea nutrient utilization studies (28). Tracer amounts of labeled substrates were added to bottles containing the initial IE community (T0) and to subsampled volume from treatments at T1, T3, and T7. After incubations (∼6 h), the microbial communities (>0.7 μm) were collected on glass fiber filters (Whatman GF/F; combusted at 450 °C for 2 h) and stored frozen. Isotopic enrichments of 15N and 13C were measured on a Europa 20/20 isotope ratio mass spectrometer, and absolute uptake rates (μM h−1) were calculated according to ref. 85. Complete methods and uptake rate measurements from all time points can be found in .