全球变化知识资源中心

Chemostat cultures.

For a full description of chemostat culturing methods see ref. 12. Briefly, axenic T. pseudonana cells in four biological replicates (duplicate chemostats × 2 experimental runs) were acclimated to nitrate limitation at 70% (1.5 day⁻¹) of maximum growth rate for more than ten days (>15 generations) under a continuous light level of 80 μmol photons m⁻² s⁻¹. Cell biomass was maintained at ~2 × 10⁵ cells ml⁻¹ by 10 μM nitrate, carbonate chemistry stabilized to 300, 475 or 800 μatm CO₂, verified by calculating³¹ f_CO2 from pH and dissolved inorganic carbon (DIC) measurements. After steady-state acclimation, ~1 × 10⁸ cells were harvested on 0.2 μm polycarbonate filters by gentle vacuum filtration and then flash frozen. Transition samples and carbonate chemistry were collected daily from chemostat cultures as CO₂ levels were increased from ~300–800 μatm at a rate ≤ 0.2 μatm min⁻¹ over four consecutive days (six generations) after pre-acclimation to 300 μatm CO₂ and nitrate limitation (Fig. 1a, b). During transition, ~1.5 × 10⁷ cells were harvested daily on 0.2 μm polycarbonate filters by gentle vacuum filtration and then flash frozen.

Nutrient-replete cultures.

Triplicate cultures of axenic T. pseudonana were grown in f/2 medium under continuous illumination (300 μmol photons m⁻² s⁻¹) at 20 °C with constant aeration. Exponentially growing cultures were harvested with a cell density of ~4.2 × 10⁵ cells ml⁻¹ at 240 ± 19 μatm or 857 ± 54 μatm CO₂, verified by DIC and pH measurements (Supplementary Table 3). To inhibit the hydrolysis of cAMP, 1.0 mM 3-isobutyl-1-methylxanthine (IBMX) was added to the cultures, as described in previous work with the diatom Phaeodactylum tricornutum³². Before and after a 100 min exposure to 1.0 mM IBMX, ~4 × 10⁸ cells were harvested by gentle vacuum filtration on a 0.8 μm filter and then flash frozen for RT-qPCR analysis.

SOLiD libraries.

A total of 28 barcoded transcriptome libraries were prepared from chemostat culture samples. RNA was extracted from filtered cells using the ToTALLY RNA kit (Life Technologies). Messenger RNA was selectively amplified using MessageAmp II aRNA Amplification kit (Life Technologies) and used to prepare SOLiD barcoded libraries (SOLiD Total RNA-seq kit, Life Technologies). Libraries were sequenced on a SOLiD 5500XL sequencer in two runs: one containing steady-state barcoded libraries and one containing transition barcoded libraries.

In silico read processing.

Reads were quality controlled (using a cutoff of p = 0.99 and minimum length = 30), trimmed, and aligned to T. pseudonana gene models using the Burrows–Wheeler Alignment tool and the SEAStAR tool (https://github.com/armbrustlab/SEAStAR). The aligned reads were counted for each gene model (Joint Genome Institute: Thaps3 extended models) using the SEAStAR tool. RNA sequences and analysis products were deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO series accession number http://www.nature.com/nclimate/journal/v5/n8/full/GSE67971.

Gene set enrichment analysis.

Read counts from transition and steady-state transcriptomes were first normalized by the trimmed mean of M-values method (TMM: R package, edgeR; ref. 33), then each gene was normalized to its own average expression level at low CO₂ (<312 μatm for transition samples and <350 μatm for steady-state samples) and log₂ transformed. The normalized gene expression data were tested for correlation to CO₂ using the Spearman rank correlation test (cor. test, R stats). Genes with significant (p < 0.05) correlation to CO₂ in transition and steady-state transcriptomes were tested for significant gene set enrichment with the hypergeometric test (phyper, R stats) in categories defined by Kyoto Encyclopedia of Genes and Genomes (KEGG 58.1, 1 June 2011), Gene Ontology (GO 2009).

Gene clustering analysis.

A data set of 98 microarray and RNA-seq samples representing ten experiments^{15, 16, 17, 18, 19, 20} was used to identify co-expressed gene clusters. All RNA-seq data were TMM normalized, and log₂ ratios versus triplicate control samples were combined with microarray log₂ ratios. To estimate the co-expression of transcripts, Pearson pairwise correlation distances were computed across all samples, following normalization such that all within-sample standard deviations were equal to one. Hierarchical clustering of these distances using Ward’s method (fastcluster)³⁴ identified a hierarchy of co-expressed genes, from which 400 co-expressed groups were selected using arbitrary cut height. Quality controls required a cluster to have at least 15 genes expressed with normalized mean square residual³⁵ <0.6 across the CO₂ experiments and to be significantly correlated with CO₂ in transition or steady-state experiments (FDR < 0.0001) to be considered for further analysis. Multiscale bootstrap resampling was performed to estimate the significance and reproducibility of sub-clusters among this hierarchy. Candidate cis-regulatory regions were identified using MEME (ref. 36) from 0 to 800 base pairs upstream of gene start sites, and TOMTOM (ref. 37) was used to assess similarity of candidate cis-regulatory motifs to previously characterized cis-regulatory elements.

Homology and alignments.

Searches for homologous sequences were performed by a hidden Markov model search tool (HMMsearch; ref. 38) across the gene models. Multiple alignments were performed with MAFFT (ref. 39) and visualized in Jalview⁴⁰.

RT-qPCR.

To quantify gene expression we used primers (Supplementary Table 4) to amplify genes for δ-CA3 (ref. 41) and a putative transporter (pID 262258) from poly(A) selected RNA (MicroPoly(A)Purist kit, Life Technologies) on a StepOnePlus instrument with the Power SYBR Green RNA-to-CT 1-Step kit (Life Technologies). The gene copies were quantified in technical triplicate by a standard curve generated from the gene amplicon in a linearized 2.1 TOPO vector (Life Technologies) and normalized to mRNA in each sample. A two-way ANOVA (R stats, aov) was performed on δ-CA3 gene expression with a Tukey HSD post hoc test (R, stats, TukeyHSD) to determine significant differences between treatments and groups respectively. A Student’s t-test (R, stats, t-test) was performed on gene expression of the putative transporter at high CO₂ only, because the low-CO₂ gene expression was below the quantification limit of the standard curve.