Long Serial Analysis of Gene Expression for Gene Discovery and Transcriptome Profiling in the Widespread Marine Coccolithophore Emiliania huxleyi

Abstract

The abundant and widespread coccolithophore Emiliania huxleyi plays an important role in mediating CO₂ exchange between the ocean and the atmosphere through its impact on marine photosynthesis and calcification. Here, we use long serial analysis of gene expression (SAGE) to identify E. huxleyi genes responsive to nitrogen (N) or phosphorus (P) starvation. Long SAGE is an elegant approach for examining quantitative and comprehensive gene expression patterns without a priori knowledge of gene sequences via the detection of 21-bp nucleotide sequence tags. E. huxleyi appears to have a robust transcriptional-level response to macronutrient deficiency, with 42 tags uniquely present or up-regulated twofold or greater in the N-starved library and 128 tags uniquely present or up-regulated twofold or greater in the P-starved library. The expression patterns of several tags were validated with reverse transcriptase PCR. Roughly 48% of these differentially expressed tags could be mapped to publicly available genomic or expressed sequence tag (EST) sequence data. For example, in the P-starved library a number of the tags mapped to genes with a role in P scavenging, including a putative phosphate-repressible permease and a putative polyphosphate synthetase. In short, the long SAGE analyses have (i) identified many new differentially regulated gene sequences, (ii) assigned regulation data to EST sequences with no database homology and unknown function, and (iii) highlighted previously uncharacterized aspects of E. huxleyi N and P physiology. To this end, our long SAGE libraries provide a new public resource for gene discovery and transcriptional analysis in this biogeochemically important marine organism.

Coccolithophores are an abundant and widespread phytoplankton functional group responsible for significant amounts of calcification in the ocean. This group is intensively studied for its roles in the marine carbon and sulfur cycles, the production of alkenones, and marine calcification. The coccolithophore Emiliania huxleyi is the most abundant species of this functional group in the modern ocean, and it blooms in both coastal and open ocean regions (24). E. huxleyi both fixes CO₂ through photosynthesis and generates CO₂ through the biomineralization of calcium carbonate (calcification). Photosynthesis and calcification are important components of the global carbon (C) cycle. Ultimately, both the presence of E. huxleyi blooms and the ratio of photosynthesis to calcification within the population mediate exchange between atmospheric and oceanic CO₂. As such, coccolithophores are being intensively studied for their role in the C cycle and their potential influence on global climate.

Nitrogen (N) and phosphorus (P) are two critical macronutrients for E. huxleyi growth, and their availability can impact when and where E. huxleyi blooms are able to occur (20). Further, N and P starvation can influence CO₂ exchange by changing rates of photosynthesis and calcification (24). For example, P starvation typically increases calcification rates relative to photosynthesis (25). In short, N and P availability in the field may influence bloom dynamics, calcification, and their concomitant impact on C cycling and on the ocean's ability to buffer changing CO₂ concentrations in the atmosphere.

To cope with low macronutrient availability in nature, marine phytoplankton have evolved inducible systems that enable them to efficiently scavenge dissolved inorganic N (DIN) and dissolved inorganic P (DIP), the concentrations of which are often growth limiting in marine systems. Phytoplankton also have the ability to utilize N and P from a diverse suite of dissolved organic N (DON) and P (DOP) compounds (1, 5). The concentrations of DON and DOP often exceed those of DIN and DIP in surface waters, so these organic compounds can be an important nutrient source in DIN- or DIP-depleted environments, such as the oligotrophic oceans. Understanding the complexity of phytoplankton nutrient scavenging systems and how they are expressed in response to depletion of N or P in the ocean is an ongoing area of study for biological oceanographers. Previous work with E. huxleyi cultures suggests that this coccolithophore has the ability to scavenge nitrogen from diverse sources. For example, it is able to grow on several DON substrates as a sole N source, including formamide, hypoxanthine, and urea (28). E. huxleyi is also able to scavenge P from diverse sources, expressing the enzyme alkaline phosphatase under low-DIP conditions allowing for the hydrolysis of certain DOP compounds (13, 32). In fact, E. huxleyi is known for being a good competitor relative to other algae in low-DIP systems and elevating phosphate uptake at growth-limiting DIP concentrations (32). Although some N and P starvation-inducible proteins have been identified for E. huxleyi (13, 29), our transcriptional understanding of E. huxleyi biology and particularly nutrient scavenging and nutrient starvation responses is limited.

While genomic research with marine cyanobacteria is rapidly advancing our understanding of their role in the sea (12, 27), there are few genome sequences (3), differential gene expression studies (2, 22, 37), and transcriptome analyses with eukaryotic marine algae. In the case of coccolithophores, fundamental gaps in our molecular-level understanding of calcification and even basic N and P scavenging mechanisms remain. Gene expression analyses are one way to work towards closing these gaps, providing a dynamic link between the E. huxleyi genome and its cellular functioning in the ocean.

Methods for assessing gene expression on a genomic scale include DNA microarrays, massively parallel signature sequencing, differential display reverse transcriptase PCR (RT-PCR), subtraction hybridization, and serial analysis of gene expression (SAGE) (38). In long SAGE (33), a short, 21-bp sequence tag from the most poly(A) proximal NlaIII restriction site of an mRNA molecule is used to uniquely identify the source gene from within the genome. Short sequence tags are sampled from all NlaIII-positive transcripts in an mRNA sample and are linked together to form long concatenated molecules that are cloned and sequenced. Quantification of all tags provides a relative measure of gene expression (i.e., mRNA abundance). SAGE thus provides both the identities of expressed genes and the levels of their expression. Detection level and sensitivity in SAGE are a function of sampling depth, i.e., the more tags sampled, the more likely detection of rare transcripts and the stronger statistical resolution of differential abundance of transcripts among mRNA samples. Of the aforementioned approaches, SAGE is advantageous in that it requires no a priori knowledge of the genome, responsive genes, or gene sequences and directly samples transcript abundance. The material benefits of SAGE are its simplicity, low cost, and efficiency. Perhaps the most common approach to expression studies with eukaryotic marine algae involves expressed sequence tag (EST) sequencing (34, 39, 40). In contrast to EST sequencing, SAGE can provide a much deeper sampling of the transcriptome, is not subject to the difficulties involved in normalization, and is quantitative in nature.

Despite its clear value, SAGE is typically applied to model systems with available genome sequences (e.g., mouse, Arabidopsis thaliana) and its utility in nontraditional models has only recently been addressed (11). Here we used long SAGE to examine genes responsive to N or P starvation by sampling 21-bp sequence tags from three E. huxleyi libraries: replete, nitrogen starved (−N), and phosphorus starved (−P). Because P starvation resulted in increased calcification, we were also able to screen the −P library for genes potentially responsive to changes in calcification. To our knowledge, these are the first transcriptome analyses of N and P starvation for a coccolithophore.

MATERIALS AND METHODS

Cell culture.

Emiliania huxleyi CCMP 1516 was obtained from the Provasoli-Guillard Center for the Culture of Marine Phytoplankton, Bigelow Laboratories. Cultures were grown at 18°C on a 14 h:10 h light:dark cycle (140 μmol quanta m⁻² s⁻¹). Nitrogen- and phosphate-replete (35 μM NO₃⁻ and 1.5 μM PO₄³⁻), −N (10 μM NO₃⁻), and −P (0 μM PO₄³⁻) cells were grown in f/50 medium without Si (16). Locally collected seawater was filtered (pore size, 0.2 μm) and autoclaved. Filter-sterilized inorganic nutrients, trace metals, and vitamins (thiamine, biotin, and B₁₂) were added after autoclaving. The cells were grown in 8-liter batch cultures. The growth of cultures was monitored daily by cell number counted with a hemacytometer and by relative fluorescence with a Turner Designs AU fluorometer. Replete cells were harvested in mid-log phase, while −N and −P cells were harvested at the onset of stationary phase. At the onset of stationary phase, inorganic nitrogen (35 μM NO₃⁻) and inorganic phosphate (1.5 μM PO₄³⁻) were added to a 20-ml aliquot of −N cells and −P cells, respectively, and compared to a 20-ml aliquot of the cultures with no amendment. Nutrient samples were collected aseptically, filtered through a GF/F filter to remove cells, and collected into an acid-cleaned tube. All samples were stored frozen at −20°C until analysis and assayed throughout the experiment. Soluble reactive phosphate was assayed via the molybdate blue method (19), and nitrate was determined with a nutrient autoanalyzer by use of standard protocols. Calcification was assayed with a 10-ml aliquot of cells removed at the point of harvest from control, −N, and −P cultures as described elsewhere (26).

Total RNA extraction.

Approximately 2 × 10⁷ cells were harvested (8,000 × g for 10 min at 22°C), and the RNA was extracted using TRI reagent (Sigma, St. Louis, MO) according to the manufacturer's instructions. Briefly, cell pellets were resuspended and lysed in 0.75 ml TRI reagent. After complete lysis of the cells, 0.15 ml chloroform was added. Total RNA was precipitated with 0.35 ml of isopropanol and washed with 75% (vol/vol) ethanol. The RNA pellets were solubilized in 50 μl diethyl pyrocarbonate-treated water. RNA concentrations were obtained with a UV spectrophotometer. Integrity of the total RNA was assessed by 1% (wt/vol) agarose gel electrophoresis.

SAGE.

SAGE libraries were constructed using ∼30 μg RNA isolated from extractions of replete, −P, and −N E. huxleyi cells following the I-SAGE Long kit protocol (Invitrogen, Carlsbad, CA). Recombinant pZEro1 clones produced by SAGE were purified using GeneMachines RevPrep Orbit (Genomic Solutions, Ann Arbor, MI) and were sequenced on an ABI 3730xl DNA sequencer (Applied Biosystems, Foster City, CA). Sequences collected were analyzed with software created at the Marine Biological Laboratory specifically for E. huxleyi SAGE analysis. The SAGE software extracts ditag sequences from the ABI 3730xl results according to the SAGE sequence grammar, parses out individual SAGE tags, excludes tags with sequence ambiguities, and reduces all SAGE tags to a look-up table of unique SAGE tag sequences and their observed frequencies among all of the E. huxleyi SAGE libraries. SAGE tags not found more than once in at least one SAGE library were excluded from analyses as putative sequencing errors, unless the tag sequence had a perfect match to available genomic and EST data. The unique tag sequences were mapped to all available E. huxleyi DNA sequences to determine the identities of expressed genes. For this, 11,880 genome and EST sequences were obtained from NCBI on 1 April 2005 and assembled with the computer program PHRAP (15). The 21-bp SAGE tags were mapped to assembled contigs and singletons based on exact sequence matches. Tags were then visualized with a generic model organism database using the Gbrowse interface (35) in the context of EST sequences and predicted open reading frames (ORFs). As SAGE tags are generated from the most 3′ NlaIII restriction site of transcripts, tags with significant differential expression among the SAGE libraries were manually assigned to predicted genes based on their positions relative to ESTs and predicted ORFs. These genes were annotated by BLAST and other homology detection methods. SAGE tags were scored for differential expression among the three libraries by using the R statistic of Stekel et al. (36), a log likelihood ratio statistic which scores tags by their deviation from the null hypothesis of equal frequencies given the tag sampling depth for each SAGE library. Higher scores represent a greater deviation from the null hypothesis, while scores close to zero represent near constitutive expression. To reduce the effects of sampling error in highlighting differential expression, only tags with an R value of 2 or greater were considered for RT-PCR and other post-SAGE analysis.

RT-PCR.

RNA was extracted from cultures as described above. Genomic DNA contamination was removed from the replete, −N, and −P RNA samples with a Turbo DNA-free kit (Ambion, Austin, TX) according to the directions provided. Equivalent concentrations of total RNA from each condition were reverse transcribed to single-stranded cDNA by using an iScript cDNA synthesis kit (Bio-Rad, Hercules, CA) and amplified using gene-specific primers (Table 1). All genes and conditions were tested for genomic contamination using no RT controls. PCR reaction conditions were as follows: 15-μl total volume of 1× PCR buffer, 0.25 mM deoxynucleoside triphosphates (Bio-Rad), 0.3 μM primers, 1 U Taq DNA polymerase (Bio-Rad) or Optimase high-fidelity polymerase (Transgenomic, Cambridge, MA), and 1 ng cDNA. PCR conditions were as follows: 1 cycle at 95°C for 5 min; 30 cycles at 95°C for 1 min, x°C for 1 min, and 72°C for 1 min; and 1 cycle at 72°C for 7 min. See Table 1 for annealing temperatures and amplicon sizes. PCR products were resolved on a 2% agarose gel, stained with ethidium bromide, and imaged with a ChemiImaging system (Alpha Innotech, San Leandro, CA).

TABLE 1.

PCR primers and amplification conditions for the genes used in the RT-PCR validation of the long SAGE analysis^a

Tag no./gene	Primer sequence	Annealing temp (°C)	Amplicon size (bp)
45	5′GCATCACGTTGACGTAGTCG	56.2	267
	3′GGAGTTCGCCTCACTGATTC
285	5′AGCTTGAAGGTCTTGGTGGA	63	115
	3′CGGGCTACTTCATCTTCGAG
12112	5′GATCATCGAGAGCACACCAA	59.5	286
	3′GATTCCGAAGATGACCCTCA
Type 1 actin (Q41205)	5′GATCTGGCACCACACCTTCT	55.5	116
	3′TGATCTGCGTCATCTTCTCG

^aPCR conditions are given in the text.

RESULTS

Culture physiology.

Any transcriptome analysis is highly dependent on the experimental culture conditions. Here we manipulated the dissolved inorganic N and P concentrations in batch cultures of E. huxleyi CCMP 1516 to induce N or P starvation and potentially capture the up-regulation of genes involved in N and P stress responses. Growth curves are presented for each condition (Fig. 1). At the onset of stationary phase, nitrate was drawn down in the −N culture and phosphate was drawn down in the −P culture (Fig. 1). The −N and −P cell cultures were split at the onset of stationary phase and re-fed with the appropriate nutrient to f/50 levels. Both of the re-fed cultures resumed growth relative to a no-addition control, confirming N and P starvation conditions (data not shown). Calcification was detectable under all conditions but enhanced nearly threefold in the −P culture relative to the −N and replete cultures (Table 2).

FIG. 1.

Growth curves of E. huxleyi under replete, −N, and −P conditions. Graphs are plotted as cell number (filled symbols) and nutrient concentration (open symbols) versus day in culture, noting the difference in scale of the y axes. Error bars denote standard errors of the means (n = 4). Filled arrowhead indicates point of harvest.

TABLE 2.

Summary of library characteristics, including calcification data and long SAGE analyses

Library condition	% Calcification	Total no. of tags sampled	Total no. of tag sequences	No. of tagsa
				Up-regulated twofold or greater	Down-regulated twofold or greater	Uniquely present	Uniquely absent
Replete	3.09	14,089	4,306			5	8
−N	3.57	15,055	4,219	38	31	4	37
−P	9.84	11,578	3,817	74	45	54	30

^aNumbers include only those tags with R values of 2 or greater.

SAGE analyses.

Correcting for sequencing error, we have sampled 11,000 to 15,000 long SAGE tags from each library and detected 3,817 to 4,306 unique tags per library (Table 2). With the use of log likelihood statistics (R ≥ 2) to reduce the effects of sampling error, we identified a total of 38 unique tags that were up-regulated twofold or greater in the −N library and 74 unique tags that were up-regulated twofold or greater in the −P library (Table 2). In addition, a total of 4 tags were uniquely present in the −N library, and 54 tags were uniquely present in the −P library (Table 2). Approximately 48% of all differentially expressed tags could be mapped to publicly available sequence data. Expression data for tags that map to genes with a putative function are presented in Table 3. The remaining tags either map to E. huxleyi sequence data with no database homology, match more than one EST or genome sequence (unresolved tags), or do not have a sequence match (orphan tags) and apparently represent nascent E. huxleyi sequences (Table 4) (also see the supplemental material). The five most highly expressed (i.e., greatest change [n-fold]) tags encompass these tag categories (Table 4). The complete data set including all of the differentially expressed genes is available in Table S1 in the supplemental material. Additional information on all the gene tags, including those that are not differentially expressed, is available on our website (http://gmod.mbl.edu/emiliania_huxleyi).

TABLE 3.

Successfully annotated differentially expressed (R ≥ 2) SAGE tags

Tag IDa	Tag sequence	No. of copiesb in:			R value	Change (n-fold)c or result for:		Annotation
		−N library	Replete library	−P library		−N library	−P library
1064	CATGGGATAATGAAATAGGAC	21	41	137	32.03	−2.09	4.07	Probable 18S rRNA transcript
24856	CATGGCGTTAACGGGTAACGG	28	50	137	26.78	−1.91	3.33	Probable 18S rRNA transcript
11533	CATGATTTGTCGACCGAAGAT	0	2	40	19.30	Uniquely absent	24.34	Putative fructose-1,6-bisphosphate aldolase
11736	CATGCTAATTTTTAAAAGAAA	0	1	38	19.22	Uniquely absent	46.24	Mitochondrial-encoded large-subunit rRNA
24531	CATGCAAGTCGAACGAGAGTT	0	6	37	15.45	Uniquely absent	7.50	Plastid-encoded small-subunit rRNA
289	CATGCCGACTAGGGATTGGAG	20	29	75	12.49	−1.55	3.15	Probable 18S rRNA transcript
11766	CATGGTTGTCGTCAGTTCGTG	0	0	21	11.48	NA	Uniquely present	Mitochondrial-encoded large-subunit rRNA
25306	CATGGCCGTTCTTAGTTGGTG	40	71	104	10.68	−1.90	1.78	Probable 18S rRNA transcript
2053	CATGTTTTCCGTGCCGGTCTG	9	7	39	9.28	1.20	6.78	Putative chloroplast ferredoxin
25385	CATGTTGGCTAGTGTGATCTT	8	14	43	9.09	−1.87	3.74	Putative light-harvesting complex protein
285	CATGCGCCGATTCGTGCGCAC	12	47	41	6.92	−4.19	−0.94	Putative fucoxanthin chlorophyll a/c binding protein
24557	CATGGATAGTATAGAGAGTCC	1	2	16	5.68	−2.14	9.74	Putative polyphosphate synthetase
39	CATGATGGCCAAGTAGGCTCC	34	40	5	5.58	−1.26	−6.57	Putative calmodulin
30	CATGGCTGCGATGTACGACCC	35	16	3	5.32	2.05	−4.38	Putative polyphosphoinositide-binding protein
196	CATGGGAGCGCAGGACGGCGC	3	5	20	4.96	−1.78	4.87	Putative fucoxanthin chlorophyll a/c binding protein
12215	CATGTGGAGTGTCCTCCTTTT	0	0	9	4.92	NA	Uniquely present	Putative enolase (2-phosphoglycerate dehydratase)
12088	CATGAATTGGGTTTAAAACGA	0	0	9	4.92	NA	Uniquely present	Mitochondrial-encoded large-subunit rRNA
25078	CATGTAGAGGTATTCTACACC	11	20	36	4.81	−1.94	2.19	Putative fucoxanthin chlorophyl a/c protein
25102	CATGCCCCTTATGCCCTGGGC	1	4	15	4.51	−4.27	4.56	Plastid-encoded small-subunit rRNA
3320	CATGTGTTGTTGTTTTGGATA	3	14	21	4.44	−4.99	1.83	Putative fucoxanthin chlorophyll a/c binding protein
25010	CATGGCTGTCGTCAGCTCGTG	1	8	16	4.41	−8.55	2.43	Plastid-encoded small-subunit rRNA
12892	CATGCAACATTTGTAAAAATG	0	0	8	4.38	NA	Uniquely present	Mitochondrial-encoded large-subunit rRNA
25026	CATGCTGTGTGAAGGGGCACA	4	10	22	4.28	−2.67	2.68	Putative phosphoglycerate kinase
9469	CATGCGAGCCTTCTCGTAGGC	0	3	11	4.24	Uniquely absent	4.46	Phosphate-repressible phosphate permease
9481	CATGTGTTTTTAATAAACAGT	0	2	10	4.04	Uniquely absent	6.08	Mitochondrial-encoded large-subunit rRNA
25244	CATGCAGGAGTTCCCGACTCA	12	16	33	4.02	−1.42	2.51	Probable 18S rRNA transcript
8393	CATGCCGATATGTTGTCTGCC	1	0	9	3.94	Uniquely absent in replete	Uniquely absent in replete	Putative 30S ribosomal protein S1
24482	CATGGGAGCTGGTCATACCCA	0	3	10	3.80	Uniquely absent	4.06	Plastid-encoded small-subunit rRNA
1484	CATGGGCTTCGTCTCGGAGGC	4	12	20	3.55	−3.21	2.03	Putative fucoxanthin chlorophyll a/c binding protein
5002	CATGTATGTATAAATTACTCG	1	0	8	3.44	Uniquely absent in replete	Uniquely absent in replete	Putative sulfate adenylyltransferase (ATP-sulfurylase)
4313	CATGTCGTGTGTTTGTGTCCT	2	0	8	3.07	Uniquely absent in replete	Uniquely absent in replete	Putative chlorophyll a/b binding protein
37	CATGTCGTGGCAGGCCTTTGT	45	24	12	2.93	1.75	−1.64	Putative calcium-dependent protein kinase
191	CATGGGTGCCGTGCGCCGGGG	17	40	15	2.89	−2.51	−2.19	Putative ribosomal protein S15
2203	CATGGCCTTCGTCTCCGAGTC	6	11	0	2.87	−1.96	Uniquely absent	Putative fucoxanthin chlorophyll a/c binding protein
271	CATGTCGCCCTGCCCAGCGCC	11	8	0	2.82	1.29	Uniquely absent	Similarity to a number of hypothetical proteins
10410	CATGATTCTTGATCGTGGCGC	0	0	5	2.73	NA	Uniquely present	Putative photosystem II stability/assembly factor HCF136
11952	CATGATCAACGAGGTTGACGC	0	0	5	2.73	NA	Uniquely present	Putative calmodulin
255	CATGGGGTTTACATATGCACT	14	4	1	2.59	3.28	−3.29	Putative P450 protein
639	CATGCTCAACGAGGACGAGCT	7	20	4	2.54	−3.05	−4.11	ADP-ribosylation factor
1714	CATGGAGTGCGTTTAGCGAGA	3	12	1	2.49	−4.27	−9.86	Putative acetyl-coenzyme A:acetoacetyl-coenzyme A transferase
18	CATGATACCCCGTTTGTGCGA	43	19	36	2.43	2.12	2.31	Putative nitrate transporter
5573	CATGCGATTGTGCGTCAGTGA	1	6	10	2.42	−6.41	2.03	Putative RAB11A (member RAS oncogene family)
470	CATGTGTGCTCCCGCATACGC	6	9	0	2.35	−1.60	Uniquely absent	Putative Myb-related domain
5434	CATGTAGACGCGTCTGTACAG	0	5	0	2.30		Uniquely present in replete	Putative glutaredoxin-related protein
2558	CATGTACCGATCGCCGCTACG	6	7	17	2.30	−1.25	2.96	Putative light-harvesting chlorophyll a/b protein of photosystem I
90	CATGGATAGAAGAGGCGCTGC	38	20	11	2.27	1.78	−1.49	Possible sterol desaturase-related protein
1848	CATGCAAGAGCGACTGCCTGC	0	4	6	2.20	Uniquely absent	1.83	Probable 18S rRNA transcript
12112	CATGAGGAGAGCGACACTTTG	0	0	4	2.19	NA	Uniquely present	Putative inorganic pyrophosphatase
12181	CATGTCCATCCAGGGCAAGTC	0	3	6	2.18	Uniquely absent	2.43	Putative photosystem I subunit X
1388	CATGACCGGGCCGAGAGGGAT	5	0	0	2.16	Uniquely present	NA	Putative GDP dissociation inhibitor
4062	CATGCAGCGCAGCGCGCGTCG	1	4	9	2.15	−4.27	2.74	Putative acyl-coenzyme A-desaturase with similarity to 4-methyl-5(B-hydroxyethyl)-thiazole
10890	CATGCAGGGTGTCGTGCAGTC	0	1	5	2.02	Uniquely absent	6.08	Monophosphate biosynthesis protein
17129	CATGCTCGTCTGCCCGACCGA	0	1	5	2.02	Uniquely absent	6.08	Putative U6 snRNA-associated Sm-like protein LSm7

^aThe tag identification (ID) can be used to view the data on a particular tag at http://gmod.mbl.edu/emiliania_huxleyi.

^bRaw frequencies in each SAGE library are presented for each tag.

^cChange (n-fold) includes correction for sampling depth, with negative symbols used for down-regulation. NA, tag was not detected.

TABLE 4.

Differentially (R ≥ 2) expressed tags with the greatest change (n-fold) in the −N and −P libraries

Tag IDa	Tag sequence	No. of copiesb in:			R value	Change (n-fold)c or result for:		Annotationd
		−N library	Replete library	−P library		−N library	−P library
11736	CATGCTAATTTTTAAAAGAAA	0	1	38	19.22249452	Uniquely absent	46.24131974	Mitochondrial-encoded large-subunit rRNA
1855	CATGATTGTTAAGAAAGCGCA	1	1	32	14.48724537	−1.068564128	38.94005873	Unresolved SAGE tag
11533	CATGATTTGTCGACCGAAGAT	0	2	40	19.30477518	Uniquely absent	24.33753671	Putative fructose-1,6-bisphosphate aldolase
25408	CATGAACTACTTCTTGAATTC	0	1	19	9.127231112	Uniquely absent	23.12065987	Orphan SAGE tag
25238	CATGGCCAGTGTGAAGCTTAG	1	2	35	15.10682258	−2.137128256	21.29534462	Orphan SAGE tag
1083	CATGAGCTACCTTTACGATGA	11	1	1	2.731229859	10.29418798	0.821775854	Orphan SAGE tag
294	CATGGTTAGGCGAGTGGCAGT	11	1	2	2.31564119	10.29418798	2.433753671	Orphan SAGE tag
827	CATGCTCGGTGTGGGGTGGCG	9	1	0	2.934929692	8.422517436	Uniquely absent	Orphan SAGE tag
1041	CATGGCGCAACCCGGGGGCGC	9	1	0	2.934929692	8.422517436	Uniquely absent	Orphan SAGE tag
128	CATGGTGCTCCTCCCGCTGTC	8	1	0	2.551511149	7.486682165	Uniquely absent	EST has no meaningful BLASTX hit

^aThe Tag identification (ID) can be used to view the data on a particular tag at http://gmod.mbl.edu/emiliania_huxleyi.

^bRaw frequencies in each SAGE library are presented for each tag.

^cChange (n-fold) includes correction for sampling depth, with negative symbols used for down-regulation. The top five tags are presented for each library.

^dUnresolved SAGE tags are tags matching more than one EST or genome sequence. Orphan SAGE tags are tags not having a sequence match in all available EST and genome sequences.

Type 1 actin and three tags (285, 45, and 12112) that mapped to gene sequence data were examined with qualitative RT-PCR to validate the expression patterns observed with the long SAGE analysis (Fig. 2). Under the conditions used in this study, type 1 actin appears to be constitutively expressed (Fig. 2). Tag 285 maps to a putative fucoxanthin chlorophyll a/c binding protein (fcp) of the E. huxleyi light-harvesting complex, and it is down-regulated over fourfold in the −N library (Table 3). This gene also appears to be down-regulated by −N conditions when assayed with RT-PCR (Fig. 2A), as the band intensity is lower in the −N sample than in the replete sample. Additional RT-PCR analysis was performed with tag 45. Tag 45 maps to an EST encoding for the C-terminal portion of a hypothetical protein with no database homology (the complete tag sequence can be found in Table S1 in the supplemental material). This tag is up-regulated over fourfold in the −N library, and this expression pattern was confirmed with qualitative RT-PCR, where the band intensity of the tag 45 amplicon was substantially reduced in the replete sample relative to the band intensity observed for the −N sample (Fig. 2B). Tag 12112 maps to an EST with homology to an inorganic pyrophosphatase (Table 3). This tag is uniquely present in the −P library, with a frequency of 4 (Table 3). Even at this low copy number the expression pattern observed with long SAGE was also observed with RT-PCR, as no band was detected for the replete sample (Fig. 2C). In no case was amplification observed for the no-RT controls (data not shown).

FIG. 2.

(A) Differential expression of long SAGE tag 285. Frequency of tag 285 in the SAGE libraries generated from replete (control) and −N E. huxleyi (top). RT-PCR amplification of the fucoxanthin chlorophyll a/c binding protein gene that maps to tag 285 (115 bp) and type 1 actin (116 bp) from control and −N conditions (bottom). (B) Differential expression of long SAGE tag 45. Frequency of tag 45 in the SAGE libraries generated from control and −N E. huxleyi (top). RT-PCR amplification of the EST sequence which maps to tag 45 (267 bp) and type 1 actin (116 bp) from control and −N E. huxleyi (bottom). (C) Differential expression of long SAGE tag 12112. Frequency of tag 12112 in the SAGE libraries generated from control and −P E. huxleyi (top). RT-PCR amplification of a 286-bp fragment of the putative inorganic pyrophosphatase gene that maps to tag 12112 and a 116-bp fragment of type 1 actin from control and −P samples (bottom). All PCRs were performed with equal template concentrations.

DISCUSSION

Genomic research with marine phytoplankton is rapidly advancing our understanding of how they function and interact with their environment. By comparing three long SAGE libraries (replete, −N, and −P), we have identified hundreds of differentially regulated tags (Table 2). To our knowledge, this is the first time long SAGE has been employed for gene expression profiling over multiple conditions with a photosynthetic marine alga, although a modified approach has been used for gene discovery with a dinoflagellate (11).

Validation.

Nearly half of the long SAGE tags could be mapped to gene sequence data from an E. huxleyi strain. Although this mapping is most straightforward with an available genome sequence, we show herein that it is tractable to map these sequence tags to existing sequence information in the absence of a genome. It is worth noting that the data mapped to multiple strains, suggesting that analysis of these SAGE data are not be restricted to strain 1516. Using RT-PCR, we examined the relative expression patterns of three tags (285, 45, and 12112), which mapped to publicly available sequence data and were either up- or down-regulated in one of the libraries. In all cases, RT-PCR generated an expression pattern similar to that observed with the long SAGE analysis. Although these tags were consistently differentially regulated, type 1 actin was constitutively expressed across the conditions tested in this study. Particularly striking are the results for tag 12112, which amplified from the −P sample but not the replete sample and is uniquely present (four copies) in the −P library. In this case, long SAGE results could be confirmed even for a low copy number tag with a small R value. Overall, these validation results demonstrate the value of SAGE libraries as a tool for transcriptome profiling, where subtle changes in expression, not just presence/absence, can be examined across multiple conditions. With the increasing number of whole-genome sequences or EST collections for diatoms, dinoflagellates, and other groups of eukaryotic marine phytoplankton, long SAGE is a promising new approach for transcriptome profiling with these different taxa. However, long SAGE is equally promising for phytoplankton lacking available EST or genome information, as high-throughput methods exist to identify full transcripts from SAGE tag sequences (6, 7). In some cases, SAGE tag sequences have been used to prime 5′ rapid amplification of cDNA ends for isolation of nearly complete transcripts (A. G. McArthur, unpublished data).

Nitrogen responses.

Little is known about transcriptional responses to N starvation in E. huxleyi. However, many algae become chlorotic, with a decrease in light-harvesting pigments, under N starvation, and the −N library was characterized by the down-regulation of genes mapping to multiple fucoxanthin chlorophyll a/c binding proteins (fcp) of the E. huxleyi light-harvesting complex. RT-PCR validation confirmed the regulation of one fcp (tag 285) (Fig. 2). The expression of this gene is also down-regulated by −N conditions in Pleurochrysis carterae (10), suggesting this may be a common transcriptional response to N starvation. Of the 38 up-regulated tags (e.g., 255, a putative 21-hydroxylase-like P450 protein) in the −N library, only 6 could be assigned a putative function (Table 3). Many of the genes up-regulated in the −N library were either orphan tags or mapped to sequence with no database homology (e.g., Table 4). This observation emphasizes how SAGE can be an important tool for the identification of new genes as well as for the regulation of genes with no putative function. For example, tags 1083, 294, 827, and 1041 (Table 4) are all orphan tags which do not map to available sequence data but are strongly up-regulated eightfold or more in the −N library. Tag 128 maps to an EST sequence without a significant BLASTX hit and thus has an unknown function (Table 4). Although the function of the gene is unclear, the specific N regulation observed with our long SAGE analysis (Table 4) suggests an involvement with N scavenging or N homeostasis. Here long SAGE has identified several new genes or gene targets for more-directed study of N scavenging in this organism.

Phosphorus responses.

E. huxleyi blooms in many oligotrophic, low-DIP areas, such as the North Atlantic (17), and has a strong physiological response to P starvation in culture, inducing several phosphate-regulated surface-associated proteins, as well as increasing alkaline phosphatase activity and phosphate uptake rate (13, 32). As was the case with the −N library, we have identified many orphan tags and tags which map to sequences with no database homology that are up-regulated in the −P library. These tags warrant further study, yet the overall results from the −P library suggest a robust transcriptional response to P starvation, with the up-regulation of several genes likely to be involved in P scavenging or P metabolism. One previously observed response to P deficiency in eukaryotic algae is the up-regulation of genes encoding phosphate-repressible permeases, allowing cells to transport phosphate with high affinity (8). In our analysis, tag 9469 maps to a putative E. huxleyi phosphate-repressible permease (9) that is up-regulated over fourfold in the −P library. Although this putative phosphate-repressible permease was previously identified (9), this is the first data confirming its regulation by phosphate supply. E. huxleyi can increase phosphate uptake under P-deficient conditions (32), and this gene is likely involved with this process.

Tag 12112 is uniquely present in the −P library (Table 3), and its expression was confirmed with RT-PCR. This tag has been successfully mapped to a complete ORF with homology to soluble inorganic pyrophosphatases (sPPase) (data not shown). Transcriptional activation of sPPase genes has been observed for P-starved cyanobacteria (14), and genes for sPPase have been identified for a number of protists (30), but the presence and P regulation of an sPPase gene has not previously been demonstrated with a coccolithophore.

An additional tag (24557) mapped to a gene with a putative role in polyphosphate metabolism. Polyphosphate is ubiquitously found in microorganisms and is either demonstrated to be or thought to be involved in diverse biological functions, including acting as a reservoir for inorganic phosphate (P_i) (18). In Saccharomyces cerevisiae, genes for polyphosphate synthesis and polyphosphate hydrolysis are both up-regulated in response to low extracellular phosphate as part of the PHO pathway involved in the acquisition of phosphate (23). In E. huxleyi, tag 24557, a putative polyphosphate synthetase, was up-regulated over ninefold in the −P library (Table 3). As highlighted above, we observed the up-regulation of a putative sPPase (tag 12112) which releases P_i from pyrophosphate, but we were unable to map any tags to a putative exopolyphosphatase, a processive enzyme that releases the terminal orthophosphate group from linear polyphosphates and would be required for the degradation of polyphosphate. It is important to note that a transcript encoding an exopolyphosphatase could be present in the −P library but not up-regulated sufficiently to be detected in this analysis. The seemingly paradoxical increase in the cell's ability to convert P_i into polyphosphate in response to P starvation might represent a strategy for accumulating and preserving P_i. Studies with yeast suggest that polyphosphate accumulation is required, presumably as a sink, to sustain a high rate of P_i uptake (23), and Escherichia coli has been shown to transiently accumulate polyphosphate in P-deficient medium (31). In E. huxleyi, where P deficiency results in the up-regulation of a phosphate permease (tag 9469) and increased rates of P uptake (32), the production of polyphosphate may be an important short-term storage pool for P_i. Ultimately the ability to both store and mobilize polyphosphate might help explain why E. huxleyi is such a good competitor for P in the oceans, where a given cell may transiently experience fluctuations in DIP concentration.

The annotation of the long SAGE data yielded some unexpected results. Notably, there was up-regulation of a tag (5002) which maps to a putative sulfate adenylyltransferase (ATP-sulfurylase) in the −P library. In plants, this enzyme mediates the first step in the metabolism of sulfate (21), and the up-regulation of this gene would be a potential mechanism for E. huxleyi cells to accumulate S during P deficiency. E. huxleyi is known to produce high levels of dimethylsulfoniopropionate (DMSP), and in diatoms there can be an accumulation of intracellular DMSP in nutrient-limited batch cultures (4). This accumulation of DMSP may be fueled by increased sulfate metabolism. We are in the initial stages of understanding E. huxleyi S metabolism, and additional expression studies of the putative ATP-sulfurylase and the extent to which its expression is linked to DMSP accumulation are certainly warranted. Also, we identified SAGE tags from rRNA in our libraries. This has been observed with other organisms from the genera Giardia and Trypanosoma (McArthur, unpublished), as rRNA contains stretches of poly(A). In most cases, these tags are not differentially expressed, but results from the −P library suggest that mitochondrial and plastid rRNA transcripts are more abundant in the −P condition (Table 3). There was some up-regulation of nuclear rRNA genes in the −P library as well, which is sometimes observed for Trypanosoma spp. at different cell densities (McArthur, unpublished). At this juncture, the mechanisms driving increased abundance of mitochondrial rRNA and plastid rRNA are unclear; however, this may be related to increased calcification in the −P condition. Additional targeted genome expression studies in this organism over a greater range of conditions would address the consistency of this observation.

Calcification.

Calcification was enhanced in the −P condition, and nearly twofold more genes were up-regulated in the −P library than in the −N library. In addition, a total of 54 tags were detected only in the −P library. The high number of unique and up-regulated tags in the −P library suggests that we may be observing the up-regulation of genes involved in calcification or calcium homeostasis (e.g., tag 11952 [calmodulin]) in addition to those involved in P scavenging. Substantial efforts have been made to identify the genes involved in calcification, and the most recent approaches have compared E. huxleyi EST libraries (40) or used suppressive subtractive hybridization with calcifying and noncalcifying cells (22). Of the five tags most highly expressed in our long SAGE −P library (Table 4), none clearly map to the most prevalent ESTs in the calcifying library analyzed by Nguyen et al. (22). Moreover, of the tags up-regulated or unique to the −P library, over 60 are orphan tags, which do not map to any of the genes identified to date in calcifying cells (22, 40). Considering the sensitivity of the transcriptome to culture conditions and the differences between EST, suppressive subtractive hybridization, and long SAGE analyses, these inconsistencies are not surprising; rather, these data underscore the value of using multiple genomic approaches over a range of conditions to identify the genes involved in uncharacterized metabolic pathways such as calcification. Further analysis of the tags up-regulated in our −P library will be helpful for future studies of transcriptional responses associated with increased calcification, and these data emphasize the potential of long SAGE analyses for building our understanding of the genes involved in calcification as well as other poorly understood pathways, such as alkenone biosynthesis.

Conclusions.

Genomic tools such as SAGE hold great promise for phytoplankton ecology and biological oceanography, where the molecular genetics of key marine organisms and the biogeochemical processes they mediate are still poorly understood. With our initial SAGE analyses, we have identified many new E. huxleyi sequences, assigned regulation data to sequences of unknown function, and highlighted previously uncharacterized aspects of E. huxleyi P scavenging, including polyphosphate synthesis. As discussed in the introduction, both N and P bioavailability can have a substantial influence on the timing, magnitude, and C cycling of E. huxleyi blooms. Our data demonstrate the nutrient regulation of a number of potential targets for RT-PCR assays in field populations of E. huxleyi, where the expression patterns of key genes relative to actin could be used to identify the N or P status of a population. As sequence data from the E. huxleyi genome project are released, an increasing number of E. huxleyi long SAGE tags will be resolved to genes in our public database (http://gmod.mbl.edu/emiliania_huxleyi), thus providing a dynamic tool for future gene discovery, transcriptome profiling, and genome annotation efforts. As scientists work toward an understanding of how E. huxleyi influences the marine C cycle and global climate, further analysis of the genes examined herein will help identify the molecular underpinnings that drive the cellular functioning of this organism in the sea.