Introduction

Fluctuations in marine environmental characteristics, such as nutrients, acidity, osmotic pressure, and temperature, can greatly affect unicellular organisms. Temperature plays a crucial role in the metabolism of phytoplankton. Some major primary producers live over a large temperature range (ranging from tropical + 30 °C to polar − 1.8 °C) and maintain major marine food webs (Toseland et al. 2013). Optimal growth temperature and temperature-driven nutrient stratification and mixing determine phytoplankton growth and biodiversity (Falkowski et al. 1998; Thomas et al. 2012), as well as phytoplankton productivity and distribution (Behrenfeld et al. 2006; Finkel et al. 2010). Because of the temperature sensitivity of phytoplankton growth and metabolism, temperature-dependent metabolic rates are widely used to predict the responses of phytoplankton to ocean warming (Marañón et al. 2014). Cells have programmed responses to stress, including rapid changes in protein phosphorylation and degradation processes, long-term transcriptional changes, and substantial changes in translation processes (Schade et al. 2004; Toseland et al. 2013).

A study on global climate change suggested a 1% decrease in global phytoplankton standing stocks per year was caused by rising sea surface temperatures in the twentieth century (Boyce et al. 2010). Diatoms are the phytoplankton group with the most drastic declines since the 1950s (Toseland et al. 2013). Marine species can thrive in cold ocean environments due to cold adaptation, acclimation, and plasticity (Parkinson et al. 1993; Thomas et al. 2002). According to biochemical studies, low temperature increases the abundance of some core metabolites in eukaryotic phytoplankton (Toseland et al. 2013). The adaptation of some cryophilic algae to low temperatures is reflected in the high intracellular concentration of certain enzymes involved in photosynthesis and carbon dioxide fixation, such as rubisco and ferredoxin (Hochachka and Somero 2002; Dolhi et al. 2013; Cvetkovska et al. 2018). Psychrophiles adjust their fatty acid compositions in response to low temperatures (Russell 2008). In cyanobacteria, low temperatures induce an increase in membrane viscosity and activate a two-component histidine kinase cascade that influences the expression profile of a specific fatty acid desaturase associated with enhanced membrane fluidity in response to low growth temperatures (Los and Murata 2002; Murata and Los 2006; Los et al. 2013). It has also attracted much attention to identify cold-regulated genes contributing to cold acclimation in green algae (Valledor et al. 2013; Li et al. 2020) and some model diatoms (Mock et al. 2017; Liang et al. 2019).

As a eurythermal and euryhaline diatom genus, Skeletonema is thriving in nutrient-rich coastal waters worldwide. Skeletonema cells have developed a series of strategies to adapt to declining temperatures: increasing enzyme concentrations, production of organic matter, and maintaining a stable photosynthesis rate (Jørgensen 1968; Toseland et al. 2013). Although these strategies reflect the interaction between dynamic environmental conditions and metabolic reactions accompanied by many aspects of physiology, biochemistry, and molecular genetics, the complex mechanisms of environmental–biological interactions have not been extensively explored. Therefore, identifying gene expression patterns and changes in physiological functions at low temperatures will provide new molecular evidence for understanding the cold tolerance of organisms.

We have sequenced Skeletonema marinoi (S. marinoi) transcriptomes from different growth phases, low temperature, and low silica concentrations and deposited the raw data to the NCBI's GEO (accession No. GSE77468). Here, we investigated the gene expression profiles in the S. marinoi exposed to low temperature compared with normal temperature. The high-throughput deep sequencing and digital gene expression (DGE) profiling were utilized to identify candidate genes and pathways associated with low temperature acclimation. Based on the GO and KEGG enrichment analysis, this study would facilitate the better understanding of gene differential expression pattern and metabolic pathways associated with cold tolerance in S. marinoi.

Methods

Cell culture and sample collection

S. marinoi was isolated from Jiaozhou Bay of Qingdao on the coast of northern China and morphologically and genetically verified as Skeletonema marinoi JZB1 (Wang et al. 2017). Subsequently, cultures from a single cell were established and maintained in f/2 medium at 20 ± 1 °C, 80–100 μmol photons m−2 s−1 of photosynthetically active radiation under a 14/10 h light/dark photocycle. Axenic cell culture conditions and concentration monitoring were the same as described by Jing et al. (2019). To quantitatively compare and analyze the transcriptional changes at low temperature (LT), S. marinoi was cultured at 12 °C (the lowest temperature at which the diatom can still grow normally according to a low-temperature tolerance gradient culture pre-experiment) and harvested during the exponential growth phase after 6 days by centrifugation (4000 g for 10 min), covered with RNAlater solution (Invitrogen, CA, USA) and stored at − 80 °C for RNA extraction. A culture grown at 20 °C for 6 days (exponential phase) was used as a control (EP). The growth curves are shown in Supplementary Material 1. Each culture group was performed in triplicate.

RNA sequencing, de novo assembly, and function annotation

Total RNA extraction, mRNA purification, cDNA library construction, de novo transcriptome assembly using Illumina Hiseq 2000 sequencing data with Trinity Assembler and transcriptional analysis were performed by Novogene Co. Ltd. (Beijing, China).

To obtain high-quality clean data for downstream analyses, adapter, poly-N, and low-quality reads were removed from raw data. Transcriptome assembly was performed based on Trinity with default parameters (Grabherr et al. 2011). Sequence information of unigenes obtained by blastx alignment (e-value < 0.00001) was compared with the public protein databases NR (NCBI non-redundant protein sequences), Swiss-Prot, and KEGG (Kyoto Encyclopedia of Genes and Genomes database, http://www.genome.jp/kegg/). Unsuccessful aligned unigenes were then annotated with ESTScan to predict their coding regions to determine sequence information. To identify differentially expressed transcripts, each of the original sequence files were aligned to the de novo assembled transcriptome using Bowtie and then RSEM was used to estimate expression values based on the alignment. The gene expression level was estimated by RNA-Seq by Expectation Maximization (RSEM) for LT and EP samples: rsem− 1.2.0 used Bowtie to align each clean read file to the reference transcripts and estimate the expected number of counts for each transcript (Li et al. 2011). In the RNA-seq technique, RPKM is Reads Per Kilobase of transcript per Million reads mapped, it also considers the impact of sequencing depth and gene length on fragments count, so is currently the most commonly used method to estimate gene expression level (Trapnell et al. 2010). The unigene expression level was measured by comparing the mapping number of clean reads to annotated unigenes by RPKM normalization and adjusted by a scaling factor for the normalization is calculated as a weighted trimmed mean of the log ratios between the two treatment samples (Mortazavi et al. 2008; Robinson et al. 2010).

Analysis of differentially expressed genes

Differential expression of RNA-seq data was analyzed with the DESeq R package (1.10.1), and the Benjamini and Hochberg method was used to adjust the resulting p value to control the false discovery rate (Anders et al. 2010). As a result, adjusted p values < 0.05 and log2(fold change) ≥ 1 were set as the threshold for significantly differential expression.

Differential expressed genes (DEG) gene ontology (GO) enrichment analysis was performed with the GOSeq R package using Wallenius’ non-central hypergeometric distribution to adjust for gene length bias (Young et al. 2010). The GO database was used to analyze the functional annotation of DEGs, including homologous protein annotations. DEGs were blasted for functional gene ontology annotation based on sequence homology using the Blast2Go software. To obtain the predicted PPI (protein–protein interaction in the STRING database: http://string-db.org/), sequences of DEGs were queried against the database (blastx) into the genome of a related species and visualized in Cytoscape (Shannon et al. 2003). KEGG statistical enrichment analysis of DEGs was tested by KEGG Orthology-Based Annotation System (KOBAS) software (v 2.0) for understanding the advanced functions and utilities of biological systems such as cells, organisms, and ecosystems (Mao et al. 2005).

Results

Overview of transcriptional analysis

To understand the low-temperature gene expression profile of S. marinoi, six total RNA samples were sequenced by Illumina. The overall quality assessment of the data showed that there were more than 10 million clean reads, with an average length of 50 bp per sample. More than 92% of the clean reads per sample could be mapped back to the assembled transcriptome of S. marinoi. More detail information about the transcriptome quality and functional annotation are in Supplementary Material 2.

Gene differential expression analysis was then performed to identify transcriptomic differences between low temperature (LT) and normal temperature (EP) groups. A total of 573 DEGs of 20,319 genes (2.8% of the detected genes) were detected in all samples, including 414 differentially upregulated (differences were between 2.61 and 12.33 times) and 159 differentially downregulated genes (differences were between 2.83 and 10.12 times, Fig. 1 and Supplementary Table S1).

Fig. 1 
figure 1

Volcano plot of differential expressed genes between low temperature (12℃) exposure and control group (20℃, EP). The red and green dots represent significantly upregulated and downregulated genes upon low temperature exposure

Full size image

GO enrichment analysis of DEGs

In the 573 DEGs, 408 upregulated genes were functionally classified into 54 subcategories, belonging to at least one term of GO biological processes (BP), cellular components (CC), and molecular functions (MF). Figure 2 shows the top 36 GO enrichment terms of DEGs, which include 20 terms in the BP category, 12 in the CC category, and 4 in the MF category. In the BP category, genes related to metabolic, cellular metabolic, organic substance metabolic, and primary metabolic processes were the four most significant upregulated metabolic items, followed by three synthetic items: biosynthetic processes, organic substance synthetic processes, and cellular synthetic processes (Fig. 2). These results suggested extensive metabolic and biosynthetic activities in S. marinoi at low temperatures. In the CC category, genes related to cell, cell part, and intracellular represented the four major subcategories of upregulated expression, followed by macromolecular complex, cytoplasm, and cytoplasmic part items (Fig. 2). In contrast, besides structure constituents of ribosome and structure molecule activities, most of the genes in the MF category were upregulated slightly (Fig. 2). No downregulated DEGs were significantly enriched for GO categories, because corrected overrepresented p value of each GO item was over 0.05, indicating no significant GO enrichment results.

Fig. 2
figure 2

GO enrichment analysis of the differential expressed genes of S. marinoi grown in low temperature or normal temperature. 408 upregulated out of the 573 DEGs were assigned to 10,826 GO annotations, which were divided into 3 categories: biological processes, cellular components, and molecular functions

Full size image

To illustrate the DEGs clustered into three main categories, GO functional enrichment analysis was conducted. A GO item over-presented with a corrected p value of < 0.05 was defined as a significantly enriched GO item. In the biological process category, many DEGs were related to translation functions, mainly reflected in molecular processes, such as protein metabolism and cellular macromolecule biosynthesis (Fig. 3). Other important and significantly enriched genes were involved in ribosome biogenesis processes belonging to cellular component processes (Fig. 3). In accordance with the results above, ribosomes belonging to the cellular component category were overrepresented because of the enrichment of ribonucleoprotein and cytoplasmic parts mainly (Fig. 4). These results indicated that low-temperature stress had some influence on translation-related processes. Only structural constituents were overrepresented in the molecular function category (Fig. 5).

Fig. 3
figure 3

Significantly enriched GO-item biological process categories in DEGs between LT and EP. Each node shows GO ID, GO description, GO over-presented corrected p value, DEG item, background item in turn. The program defaults to setting the top ten GO terms in terms of significance as squares and the other GO terms as circles. The arrows represent inclusion relationships, where all genes of this node are equally annotated into their parent nodes. The colored nodes represent the significantly over-presented GO terms, darker the color of node, the higher it was enriched

Full size image
Fig. 4
figure 4

Significantly enriched GO-item cellular component categories in DEGs between LT and EP. Each node shows GO ID, GO description, GO over-presented corrected p value, DEG item, background item in turn. The program defaults to setting the top ten GO terms in terms of significance as squares and the other GO terms as circles. The arrows represent inclusion relationships, where all genes of this node are equally annotated into their parent nodes. The colored nodes represent the significantly over-presented GO terms, darker the color of node, the higher it was enriched

Full size image
Fig. 5
figure 5

Significantly enriched GO-item molecular function categories in DEGs between LT and EP. Each node shows GO ID, GO description, GO over-presented corrected p value, DEG item, background item in turn. The program defaults to setting the top ten GO terms in terms of significance as squares and the other GO terms as circles. The arrows represent inclusion relationships, where all genes of this node are equally annotated into their parent nodes. The colored nodes represent the significantly over-presented GO terms, darker the color of node, the higher it was enriched

Full size image

Metabolic pathways related to low-temperature tolerance

The KEGG database enrichment method was used to analyze the main metabolic or signaling pathways to understand the low-temperature tolerance mechanisms of S. marinoi. The 20 most enriched pathways were identified in the low-temperature group and 5 significant pathways were ribosome, porphyrin and chlorophyll metabolism, fatty acid metabolism and biosynthesis, biotin metabolism (Supplementary Table S2).

DEGs were mapped to the KEGG standard pathways to determine their biological pathways. There were 101 KEGG pathways covering 573 DEGs, and 5 pathways were significantly enriched (adjusted p value < 0.05). The largest cluster was ribosome with 69 members, followed by chlorophyll and porphyrin metabolism with 13 members, and fatty acid metabolism with 11 members (Fig. 6). Moreover, all the genes related to these metabolic pathways were upregulated, such as RP-L genes encoding large-subunit ribosomal proteins and RP-S genes encoding small-subunit ribosomal proteins for ribosome biogenesis and proliferation; fabB encoding 3-oxoacyl-synthase I related to fatty acids biosynthesis, metabolism pathways, or biotin metabolism (Table 1). Gene expression patterns in porphyrin and chlorophyll metabolism also were significantly upregulated in S. marinoi at low-temperature stress, e.g., hem genes encoding porphyrin enzymes required for heme biosynthesis and chl genes encoding components of chlorophyll.

Fig. 6
figure 6

KEGG classification of 106 DEGs between LT and EP. The x-axis represents number of differentially expressed genes; the y-axis represents metabolic pathways. 106 out of the 573 DEGs were assigned to 101 KEGG pathways. The top five most abundant KEGG pathways are shown

Full size image
Table 1 The significantly enriched KEGG pathways and the identified genes between low temperature and normal temperature conditions
Full size table

Discussion

Translation processes responding to cold stress

A strong transcriptional response was associated with the functional category of protein synthesis at low-temperature culture conditions. Previous biochemical studies on the psychrophilic diatom Fragilariopsis cylindrus and the globally distributed temperate Thalassiosira pseudonana showed that the eukaryotic ribosomal protein S14 was more abundant at low temperatures (Toseland et al. 2013). In S. marinoi, ribosomal proteins S14 transcripts were fivefold upregulated at 12 °C (Fig. 3), showing a similar temperature-dependent transcription. Compared with normal temperatures, many genes involved in ribosomal biogenesis and assembly exhibited higher transcriptional levels at 12 °C (Fig. 3–5). This response was most pronounced in KEGG enrichment, where the mRNA levels of 38 genes encoding large-subunit ribosomal proteins and 29 genes encoding small-subunit ribosomal proteins were elevated at 12 °C (Table 1). Upregulated expression of ribosomal protein-coding genes at low temperatures would be consistent with translation compensation for potential cold-induced problems. It was in agreement with the report in Chlamydomonas reinhardtii that 14 transcripts of GO-enriched proteins involved in “translation” and “ribosome” processes of protein synthesis showed significant expression changes in cold exposure (Li et al. 2020). The fact that expression of genes related to protein synthesis is elevated in diatoms adapted to low temperatures may be a beneficial response to the reduced rate of biosynthesis at low temperatures (Liang et al. 2019).

Low temperature significantly affects the stability of the RNA structure and, thus, the initiation of translation (Farewell et al. 1998). Studies in cold-adapted diatom species suggest that high baseline expression of pathways such as aminoacyl-tRNA biosynthesis, ribosomal proteins, RNA transporters, and translational factors is to better compensate for lower translational efficiencies at lower temperatures (Liang et al. 2019). In the current study, we find that low growth temperature influences the expression of genes associated with translation and protein synthesis too. Among the genes upregulated at 12 °C, eight genes (EIF4E, EEF2, EIF4A, TUF, TSF, TEF, ETF1, and fusA) encode translation initiation factors, elongation factors, or termination factors, and seven genes (LARS, GARS, AARS, HARS, WARS, YARS, and CARS) are involved in aminoacyl-tRNA synthetase (Log2 (fold change) > 1, Supplementary Table S2). Besides, three genes (typA, K06942, and ABCF3) related to GTP (guanosine triphosphate) binding proteins were uprelated due to low temperature.

The pentatricopeptide repeat (PPR) family is potentially involved in many important metabolic processes, including thylakoid, ribosome, integral to membrane, cytoplasm, and photosystem I processes. Most importantly, as an RNA-binding protein in mitochondria and chloroplasts, PPR is involved in complex post-transcriptional processes, such as RNA processing and translation (Lurin et al. 2004). In S. marinoi, four upregulated DEGs were annotated in Swiss-prot as PPR (Gene ID: comp16273_c0, comp 12502_c0, comp 242154_c0, comp 18063_c0; Supplementary Table S2). Here, as the eukaryotic translation initiation factor 3 complex, the transcriptional response of PPR to cold stress indicated that low temperature could affect translation processing by regulating PPR gene expression. Translation initiation ribosomes may be related to the rate of protein synthesis at low temperatures (Toseland et al. 2013).

In addition, a certain proportion of genes with upregulated expression of low-temperature encoding proteins are involved in transcriptional regulation. For example, secA and GST are nucleotide-binding proteins involved in the transcriptional regulation of special substrates ATP. The PET domain is a transcription termination factor nusG. VCP, EGD1, and another three genes are transcriptional regulators of RNA polII (Supplementary Table S2).

There were also three temperature-sensitive proteins regulating transcription. One gene encoding the cold-shock DNA-binding domain was upregulated 5.77-fold in the LT cultures (Supplementary Table S2). In contrast, the heat shock factor protein transcript was downregulated 3.74-fold. However, another gene encoding the heat stress transcription factor (HSF) was upregulated significantly. Cold-shock proteins can be easily induced by cold shock and overcome decreased DNA replication, transcription, and translation efficiency, thus maintaining the stabilization of the secondary structures in RNA and DNA (Phadtare et al. 1999). In plants, HSFs are closely related to heat shock protein (HSP) stress-dependent and developmental expression (Schoffl et al. 1998a, b). Plants seem to contain a large number of different HSFs mediating the expression and protection of target genes, such as genes encoding HSP. These genes are important for stress tolerance but are not the only protective component induced by heat stress (Panchukm et al. 2002). Stress proteins combat protein toxicity by preventing protein denaturation and holding it in a folding or assembling state to facilitate repair. This protein complex is involved in key physiological processes and protects cells from stress-induced damage (Bierkens et al. 1998). A study on Saccharomyces cerevisiae’s response to different stresses indicated that yeast cells produce a large number of HSPs regulated by the transcription factor hsf1p under rapid heat shock (Estruch 2000).

Low temperature as a positive transcriptional regulation of fatty acid metabolism

In addition to transcription factors, many significant DEGs were involved in fatty acid metabolism or biosynthesis. The temperature has an important influence on the fatty acid composition of lipids in some organisms (Palma et al. 2008; Sumner et al. 1969). In S. marinoi, ten genes involving fatty acid metabolism were upregulated at LT conditions, and seven overlapped between fatty acid metabolism and biosynthesis (Table 1). As annotated in KEGG, fabB, fabD, fabF, and fabH were involved in acyl-carrier-protein (ACP) synthase or transferase. ACP plays a key role in fatty acid biosynthesis. It participates in the fatty acid synthesis cycle by transferring acyl groups. fabG and fabI were encoding ACP reductase, which is involved in fatty acid degradation metabolism. ACAC, EHHADH, and PHS1 encode the enzymes involved in acyl-CoA synthesis. Acyl-CoA metabolism plays a key role in temperature adaptation in diatoms (Liang et al. 2019). Acyl-CoA is a temporary compound formed when CoA attaches to the end of long-chain fatty acids in living cells and participates in fatty acid metabolism. Acyl-CoA synthetase participates in the fatty acid beta-oxidation pathway by acting on medium-chain fatty acids in peroxisomes (Hettema et al. 1996). After β-oxidation, the compound is oxidized to form one or more acetyl-CoA molecules, which enter the citric acid cycle and eventually form several ATP molecules. Related reports pointed out that at lower temperatures, higher rates of fatty acid biosynthesis help consume excess cytosolic acetyl-CoA due to the reduced nuclear acetyl-CoA requirements when growth rates are slow (Weinert et al. 2014).

There was also a FAD2 encoding a microsomal omega-6 fatty acid desaturase. In many organisms, the degree of unsaturation of fatty acids varies with the growth temperature. In S. marinoi, the desaturation of fatty acids was significantly correlated with the growth temperature, increasing the synthesis of unsaturated fatty acids with decreasing growth temperature. Similar phenomena have also been reported in bacteria and cyanobacteria, yeasts, and fungi (Nozawa 2011). Chlorella minutissima can produce a large amount of EPA at low temperatures, suggesting that EPA may be a temperature-sensitive enzyme as an important enzyme for fatty acid desaturation and chain elongation (Seto et al. 1984). The biological phenomenon that low temperature promotes an increase in unsaturation can be explained by increasing the content of unsaturated fatty acids to maintain the fluidity of membrane lipids (Marr 1962). Apart from the mechanisms discussed above, the bioB gene encoding biotin synthase showed an upregulated expression (Table 1). As a necessary component of cell growth, fatty acid production, and lipid and amino acid metabolism, biotin changing at the transcriptional level also proved that fatty acid metabolism was responding to low temperatures in S. marinoi. In summary, the high expression level of these genes indicated that low temperatures could promote cell growth by stimulating cells to produce more energy and intermediate metabolites, which have a positive regulatory effect on fatty acid metabolism.

Photosynthesis acclimated transcriptionally to low temperature

As an important environmental variable, low temperature can disturb the balance between energy input and consumption and cause photosynthetic changes (Maxwell 1994). Acclimation and adaptation are two important strategies of cold tolerance in photosynthetic organisms (Huner et al. 2020). Literatures point out that low temperature have some effects on organisms: increased in maximum photosynthesis rate, increased resistance to low-temperature-induced photoinhibition, decreased sensitivity to temperature changes, decreased optimal temperatures for photosynthesis, and decreased tolerance to high temperature, so the photosynthetic response of algae exposed to abnormal growth temperature may be a short-term effect of temperature on photosynthetic metabolism (Davison 1991). The current study indicated that some significant differential expression genes were involved in porphyrin and chlorophyll metabolism. As shown in Table 1, six differential expression genes (hemB, hemC, hemE, hemF, hemL, and hemY) encoded enzymes correlated with porphyrin precursors, and porphyrin binding to magnesium is the active component of chlorophyll. We also found three upregulated genes (chlD, chlH, and chlI) encoding magnesium chelatase subunits and two genes encoding enzymes correlated with chlorophyll content. These positive regulations at the transcriptional level suggest low temperatures can induce photosynthetic pigment gene expression in S. marinoi, promoting photosynthesis. This agrees with the results of a prior study that S. costatum has higher photosynthetic rates at low growth temperatures (Mortain-Bertrand et al. 1988). This is also true for many higher plants, and chlorophyll content is inversely related to growth temperature (Huner et al. 1993). In the green alga Chlorella vulgaris, the structure and function of the chlorophyll photosynthetic apparatus change at low temperatures (5 °C, Maxwell et al. 1994). Laminaria saccharina (Phaeophyta) and Phaeodactylum tricornutum (Bacillariophyta) growing at low temperatures for a long time had higher maximum CO2 fixation rates (Li et al. 1982; Davison 1987). To cope with the dark end of the light spectrum, cold-adapted algae have evolved mechanisms to significantly reduce their light compensation point (LCP). A low LCP can be achieved by efficient energy conversion and carbon fixation under low light and low temperatures. To maintain the accumulation of inorganic carbon, the algal cells showed high expression of carbon fixation genes under stress conditions. Previous research also showed that carbon fixation genes were upregulated in low-temperature S. marinoi (Liu et al. 2020).

In summary, S. marinoi had a photosynthetic response to low temperatures similar to other species, suggesting a photosynthetic temperature domestication mechanism, that low environmental temperatures can modulate the structural and functional properties of the photosynthetic apparatus at multiple levels.

Conclusion

In this study, we analyzed the expression profiles of S. marinoi during cold exposure by RNA-seq. These identified DEGs from S. marinoi have diverse functions, such as ribosome, fatty acid biosynthesis, porphyrin and chlorophyll metabolism, fatty acid metabolism, and biotin metabolism. The regulation of genes related to translation processes, fatty acid metabolism, and photosynthesis provide new molecular-level insight into cold stress responses in eukaryotic marine phytoplankton. The regulation mechanism may allow S. marinoi to adapt to environments with changing temperatures. The primary pathways highlighted by the present work and the genes directly responsive to low temperature provide a clear regulation network for further molecular, biochemical, and physiological analyses in diatoms and phytoplankton in general. Characterizing these constraints will allow us to make improved forecasts of species survival and may prove critical for understanding the fate of thermo-sensitive diatom communities facing climate change.