Physiological and Transcriptomic Evidence for a Close Coupling between Chloroplast Ontogeny and Cell Cycle Progression in the Pennate Diatom Seminavis robusta

Despite the growing interest in diatom genomics, detailed time series of gene expression in relation to key cellular processes are still lacking. Here, we investigated the relationships between the cell cycle and chloroplast development in the pennate diatom Seminavis robusta . This diatom possesses two chloroplasts with a well-orchestrated developmental cycle, common to many pennate diatoms. By assessing the effects of induced cell cycle arrest with microscopy and ﬂow cytometry, we found that division and reorganization of the chloroplasts are initiated only after S-phase progression. Next, we quantiﬁed the expression of the S. robusta FtsZ homolog to address the division status of chloroplasts during synchronized growth and monitored microscopically their dynamics in relation to nuclear division and silicon deposition. We show that chloroplasts divide and relocate during the S/G2 phase, after which a girdle band is deposited to accommodate cell growth. Synchronized cultures of two genotypes were subsequently used for a cDNA-ampliﬁed fragment length polymorphism-based genome-wide transcript proﬁling, in which 917 reproducibly modulated transcripts were identiﬁed. We observed that genes involved in pigment biosynthesis and coding for light-harvesting proteins were up-regulated during G2/M phase and cell separation. Light and cell cycle progression were both found to affect fucoxanthin- chlorophyll a/c -binding protein expression and accumulation of fucoxanthin cell content. Because chloroplasts elongate at the stage of cytokinesis, cell cycle-modulated photosynthetic gene expression and synthesis of pigments in concert with cell division might balance chloroplast growth, which conﬁrms that chloroplast biogenesis in S. robusta is tightly regulated. Thalassiosira pseudonana v3.0) were added. In a BLASTN alignment of more 30 nucleotides with 80% coverage between the and the hit alignment achieved, hit sequence query for similarity searching with BLASTX. the homology (E-value cutoff, 1 3 10 2 3 lowest percentage identiﬁcation, 29) and identity of functional GO domains and InterPro domains, TDFs functionally annotated and classiﬁed into 11 functional groups. TDFs functional annotation classiﬁed as and InterPro annotations GO terms homologies identiﬁed National for annotations categories of speciﬁcity. sequences GenBank, complete sequences of their corresponding genes sequences. S10, acetone by means of sonication (tip sonicator; 40 W for 30 s). were ﬁltered over a 0.2- m m Alltech nylon syringe ﬁlter to remove particles and injected into a Agilent 1100 series HPLC system (ChemStation software) equipped with a Macherey-Nagel reverse-phase C18 column (Nucleodur C18 pyramid; 5 m m particle size). Pigments were analyzed according to the method of Wright and Jeffrey (1997) using a gradient of three solvents: 80% methanol-20% ammo-nium acetate, 90% acetonitrile, and ethyl acetate. Two detectors were con- nected to the HPLC system: an Agilent standard ﬂuorescence detector to measure ﬂuorescence of chlorophylls and their derivatives and an Agilent diode array detector to measure absorbance of each peak at 436 and 665 nm and absorbance spectra over a 400- to 700-nm range. Fucoxanthin was identiﬁed by comparison of retention times and absorbance spectra and quantiﬁed by calculating response factors using pure pigment standards (supplied by DHI). Sequence data in the GenBank/EMBL data libraries under accession numbers FJ388875 and GE343005 to GE343326.

Diatoms, an extraordinarily diverse group of heterokontophyte microalgae (Kooistra et al., 2003), domi-nate the primary production of many marine and freshwater ecosystems (Granum et al., 2005). Based on the shape and structure of their unique siliceous cell walls, two major architectural types are recognized: "centric" diatoms, a paraphyletic group with radially patterned valves, and "pennate" diatoms, a monophyletic group characterized by a feather-like valve structure. Whole-genome sequencing of Thalassiosira pseudonana (Armbrust et al., 2004) and Phaeodactylum tricornutum (Bowler et al., 2008), representatives of centric and pennate diatoms, respectively, has revealed that they combine plant-and animal-like characteristics and possess many genes that are completely unknown in other organisms. These genome studies enable the unraveling of the genetic basis of the unique properties underlying the ecological and evolutionary success of diatoms. Diatoms have long been known for their extremely high photosynthetic efficiency: compared with other photosynthetic eukaryotic unicells of comparable size, diatoms have the highest growth rates (Banse, 1982;Raven and Geider, 1988;Sarthou et al., 2005) and an unusually high photosynthetic flexibility that is essential for coping with large habitatintrinsic fluctuations in irradiance . Their high growth rates might be attributed to their high Rubisco carboxylase efficiency and the putative presence of an alternative C4-photosynthetic pathway, enabling temporary storage of carbon for use at times of low light irradiance (Riebesell, 2000;Wilhelm et al., 2006;Roberts et al., 2007). The photosynthetic flexibility of diatoms is related to their high capacity for energy dissipation through nonphotochemical quenching, which can reach a 5-fold higher level than that in plants (Ruban et al., 2004).
As in other groups of heterokontophyte algae, diatom chloroplasts originate from a secondary endosymbiosis, probably the engulfment of a red alga by a heterotrophic eukaryote (McFadden, 2001;Falkowski et al., 2004). As a result, diatom plastids are surrounded by four membranes instead of the usual two, typical for plants and green algae. Moreover, the outermost pair of membranes is connected with the nuclear envelope in a number of diatoms (Stoermer et al., 1965;Dawson, 1973;Hashimoto, 2005). These characteristics have implications for chloroplast protein targeting pathways (Apt et al., 2002;Kilian and Kroth, 2005) and possibly for its division mechanism and transmission to daughter cells (Hashimoto, 2005). While the chloroplast ultrastructure appears to be fairly uniform in diatoms, there is a wide variation in chloroplast morphology as well as in the number of chloroplasts and their arrangement within the cells. The chloroplasts of polyplastidic diatoms, which comprise most centric diatoms but also several genera of pennate diatoms, have a very simple morphology (for review, see Mann, 1996). In contrast, chloroplasts of mono-, di-, and tetraplastidic diatoms often have more elaborate shapes and can undergo complex changes in morphology and arrangement before and/or after cytokinesis. Chloroplast division involves constriction of the parent chloroplasts into two more or less equally sized daughter chloroplasts. Two types of chloroplast division are recognized among diatoms, referred to as autonomous and imposed division (Mann, 1996). In the autonomous type, believed to be the primitive condition, the chloroplast constricts into two without the obvious involvement of any other organelle. In imposed division, which appears to have evolved several times in different lineages of pennate diatoms, chloroplast division occurs synchronically with the formation of the cleavage furrow. In plants and some other eukaryotes, chloroplast division is orchestrated by a prokaryote-derived division machinery, whose regulation is not fully understood (Osteryoung and Nunnari, 2003;Margolin, 2005;Adams et al., 2008) but should somehow be coordinated with the regulation of the cell cycle machinery that paces the cell division rate on the surrounding environments and coordinates critical processes such as cell growth, karyokinesis, and cytokinesis (De Veylder et al., 2007).
Environmental factors mostly impinge on cell cycle control through cell cycle checkpoints at the G1/S or G2/M cell cycle phase transitions. Many universal core cell cycle genes that regulate these transitions have been found in the two sequenced diatom genomes (Armbrust et al., 2004;Bowler et al., 2008), but diatom-specific genes and gene regulations may contribute to the unique features that distinguish diatoms from other algae M.J.J. Huysman, A. De Martino, C. Martens, E. Rayko, J. Gillard, M. Heijde, B. Mathieu, A. Meichenin, A. Montsant, K. Vandepoele, Y. Van de Peer, L. De Veylder, D. Inzé, C. Bowler, and W. Vyverman, unpublished data). Chloroplast development is difficult to address in polyplastidic cells, such as centric diatoms and cells of higher plants (Pyke, 1999). However, the observed coordination of chloroplast development with cell cycle progression in many pennate diatoms creates an opportunity to address more easily chloroplast development in relation to other cellular processes, because if chloroplast ontogeny is regulated by the cell cycle, the controlled progression of chloroplasts can be studied in synchronized cultures.
Here, we combined physiological experiments, cytological observations, and a cDNA-amplified fragment length polymorphism (AFLP)-based genome-wide transcriptome analysis to identify cell cycle-dependent checkpoints in chloroplast development and gene expression during synchronized growth. As a model species, we used Seminavis robusta , a representative of the diverse and ecologically important group of pennate Naviculaceae. As in most genera of this family, the two chloroplasts in S. robusta divide in an autonomous manner and move from the girdle to the valves and back, once during each cell cycle (Chepurnov et al., 2002;Supplemental Fig. S1). S. robusta is particularly well suited for cytological studies because of its large size (up to 100 mm) and, being a benthic species adhering to surfaces (Thompson et al., 2008), the easy and nonintrusive observation of cellular behavior. Moreover, sexual reproduction can be controlled, resulting in the first diatom pedigree that currently comprises approximately 110 clones (Chepurnov et al., 2008). Therefore, this species offers interesting perspectives for functional genomics studies as well as for forward genetics to investigate key life history traits, including the dynamics of their photosynthetic apparatus and its interaction with other cellular processes and the environment.

Cell and Chloroplast Division Are Both Arrested in Response to Darkness and Cell Cycle Inhibitor Treatments
Using light deprivation and two cell cycle inhibitors, we investigated whether a relationship exists between the cell cycle checkpoints and the chloroplast division cycle. For light deprivation, exponentially growing cultures were transferred to complete darkness for 24 h and compared with light-grown control cultures by flow cytometry (Fig. 1A). The tetraploid peak (G2+M phase) present in the light-grown cultures was absent in the 24-h dark-treated cultures, indicating that in the latter the majority of cells were arrested in the G1 phase of the cell cycle. Microscopically, the lightgrown cultures consisted of both dividing and nondividing cells (Fig. 1B). In contrast, the dark-arrested cultures only contained nondividing cells (Fig. 1C). These cells have their two chloroplasts undivided and located at the girdle sides of the cell (Fig. 1, D-G). In most cases, the ventral chloroplast had four subcentral lobes that, across the valves, reached the dorsal side of the cell (Fig. 1, F and G). Occasionally, some postcytokinetic doublet cells could also be found in the darkarrested cultures. These doublet cells are at a stage after cytokinesis but before cell separation, during which the cell is devoted to frustule formation (Coombs et al., 1967a). Microscopically, the doublet cells in dark-arrested cultures contained the G1 type of girdle-located undivided chloroplasts.
Next, we studied chloroplast behavior upon treatment with chemical cell cycle inhibitors. First, hydroxyurea (HU) was applied to exponentially growing cultures, thereby inhibiting S-phase progression by depleting the cell of deoxynucleoside triphosphates (Young and Hodas, 1964). A complete S-phase arrest of cell division by HU was confirmed by flow cytometry after 72 h of treatment (Supplemental Fig. S2), while the cultures remained healthy, because cell growth restarted after the inhibitor had been washed away (data not shown). The cells in these S-phase-arrested cultures all contained girdle-located undivided chloroplasts as in dark-arrested G1-phase cells, but the subcentral lobes were not observed ( Fig. 2A). From these observations, we conclude that chloroplast division and movements do not take place before the S phase. However, these results should be interpreted with caution: because HU depletes the cell's deoxynucleoside triphosphate pool, chloroplast DNA replication might be affected by the drug and, thus, inhibition of plastokinesis might be a direct effect of the chloroplast's inability to replicate its DNA. To validate these findings, aphidicolin, another cell cycle S-phase inhibitor that blocks the nucleus-specific DNA polymerase a (Ikegami et al., 1978), was applied on dark-arrested cultures that were reilluminated immediately after addition of the drug. Microscopic analysis after 24 h revealed that treated cells did not divide and contained undivided chloroplasts located at the girdle sides (Fig. 2, B-D), while control cultures divided normally. In contrast with HU-arrested cells, ventral chloroplast lobes were observed during aphidicolin arrest (Fig. 2D).

Cytological Changes during Cell Cycle Progression in Synchronized Cultures
Based on the uniform cell cycle arrest in darkarrested cultures, we established a synchronization procedure to study cell cycle-modulated processes. To this end, exponentially growing cultures (12 h:12 h light:dark) were transferred into darkness for a period of 24 h and reilluminated for 12 h. Two monoclonal strains, designated F 1 -8B and F 1 -9A, were synchronized in this manner and were used for the transcriptome analysis described below. Reillumination resulted in the synchronous reactivation of cell cycle progression, starting from the G1 phase of the darkarrested cells. Synchrony was evaluated by estimating the amount of dividing cells at hourly intervals, complemented by observations of chloroplast dynamics (Fig. 3). Dividing cells were operationally defined to include only doublet cells, characterized by the presence of a cleavage furrow, which is visible as an area of high contrast separating the chloroplast pairs (see Supplemental Fig. S3 for an illustration of the counting procedure). Immediately upon reillumination, all cells were in the G1 phase, containing undivided chloroplasts as previously shown. After 4 h, few cells with divided chloroplasts appeared in the cultures, and after 6 h, some cells were dividing. From 6 h onward, the cultures contained S-, G2-, and M-phase cells, in which the proportion of dividing cells progressively increased until 50% at the 9-h sampling Figure 1. Response of S. robusta to dark treatments. A, Flow cytometric DNA content histograms of an exponentially growing culture before transfer to darkness (top), after 24 h of dark incubation (middle), and after 24 h of light (bottom). B to D and F, Light microscopic photographs of the 24-h light-treated culture (strain F 1 -8B; B), the 24-h dark-treated culture (strain F 1 -8B; C), a 24-h dark-arrested cell in girdle view with focus on the ventral chloroplast (strain F 2 -31B; D), and a 24-h darkarrested cell in valve view (strain F 2 -31B; F). The chloroplasts appear brownish, and the cell wall is visible due to its high contrast. E and G, Confocal fluorescence microscopic photographs of cells from darkarrested cultures (strain F 2 -31B) in which the chloroplastidic signal is red (autofluorescence), the nuclear signal is green (SYBR Safe), and the cell wall signal is blue (PDMPO; Lysosensor). Chloroplasts of darkarrested cultures are undivided and pressed against the girdle. The ventral chloroplast has four subcentral lobes that extend across the valve toward the dorsal side of the cell. Bars = 10 mm. point. At 9 h, the amount of dividing cells decreased as a consequence of daughter cell separation and initiation of the second cell division cycle.
As shown previously by Chepurnov et al. (2002), chloroplast division occurs by central constriction (  Fig. 4E), the difference in location between undivided and divided chloroplasts could easily be seen. During this "reorganized chloroplast configuration" stage, the nucleus, positioned in the center of the cytoplasm, was still undivided (Fig. 4E). Taking into consideration that the chloroplasts divide after S-phase initiation, this shows that plastokinesis and subsequent chloroplast relocation occur during S/G2 phase and are fully established before M phase.
Before signs of mitosis appeared, but after chloroplast rearrangement, the first PDMPO signal was detected as an elliptically shaped thin band (Fig. 4F). This signal probably represents a girdle band (Supplemental Fig. S4) that is added to the girdle as a result of cell growth. Chloroplast reorganization was followed by karyokinesis (Fig. 4G) and cytokinesis (Fig. 4,. Early on during cytokinesis, PDMPO fluorescence was initially detected inside two "vacuolelike" compartments located against the division plane running from pole to pole (Fig. 4H), while a PDMPOstained cell wall was not yet visible. Since the fluorescence emission and intensity of PDMPO typically shift when the intracompartmental silica concentration increases above 3.2 mM (Shimizu et al., 2001;Hazelaar et al., 2005), the observed fluorescence of PDMPO in these compartments could be paralleled with silicon influx in preparation for silica deposition during the subsequent formation of the frustule. The deposition of cell wall material was observed as a straight line that ran between the two daughter nuclei and located  . Evaluation of the synchrony of cell and chloroplast division in S. robusta synchronized strains F 1 -8B and F 1 -9A. The proportion of dividing cells is presented as a function of time, during reillumination of the 24-h dark-arrested cultures. Data are means 6 SE of six microscopic fields (0.09 mm 2 ). The predominant morphological stages in each cell cycle phase are represented from light to dark gray in this order: undivided chloroplasts, dividing chloroplasts, dividing cells, and separating cells. G1 corresponds to the phase in which cells with undivided girdle-located chloroplasts are present. The asterisk denotes the first observed cells with divided chloroplasts and also approximates the moment of S-phase initiation, according to the transcriptional induction of the MCM5 prereplication factor (Supplemental Fig. S8). G2/M corresponds to the phase in which dividing cells were increasingly represented with a maximum in the amount of dividing cells at 9 h. M/G1 corresponds to the stage of cell separation.
at the site of the previously observed cleavage furrow (Fig. 4I). Only after this initial deposition of silica, while both cells were still attached to each other, the chloroplasts moved back from the valves to the ventral and dorsal girdle sides (Fig. 4J). During this last relocation event, each chloroplast elongated until it completely covered the girdle area. Cell division was finalized when daughter cells separated (Fig. 4K), creating two new cells containing one unstained valve, inherited from its mother, and one newly built valve stained with PDMPO (Fig. 4L).

Cell Cycle-Modulated Expression of the S. robusta Chloroplast Division Protein FtsZ
The microscopically derived indications of cell cycle phase-dependent chloroplast division were validated by transcript quantification of the bacterial cell division homolog FtsZ in synchronized S. robusta cultures. Therefore, a FtsZ gene fragment was amplified with a set of nested degenerate primers, cloned, and sequenced, enabling the design of a specific primer pair suited for real-time (RT) quantitative (Q) PCR (Supplemental Fig. S5). Three constitutively expressed genes were used for data normalization (see "Materials and Methods"), and RT-Q-PCR of the S. robusta FtsZ ortholog was done on two replicated synchronizations of strain F 1 -8B (Fig. 5). Both showed that FtsZ expression is modulated during synchronization. Its expression 5 to 8 h after reillumination increased 2.5to 4-fold when compared with its average expression between 0 to 4 h. In both replicates, FtsZ was maximally expressed at 7 h after darkness, being 2 h before most cells were dividing.

cDNA-AFLP Expression Profiling during Cell Cycle Progression
In parallel with the above-described sampling (Fig.  3) for cytological observations, samples were taken at 1-h intervals to conduct a cDNA-AFLP-based genomewide transcript profiling to identify modulated gene expression profiles correlated with the cytological changes during the cell cycle. The complete data set, including annotations and expression data, is available online (Supplemental Table S1).
The cDNA-AFLP analysis resulted in 2,908 transcriptderived fragments (TDFs) that were scored quantitatively. A considerable amount of expression variation between the two strains originated from genotypedependent DNA polymorphisms in the transcripts, resulting in the absence/presence of cDNA-AFLP fragments . Cytological events during synchronized growth of S. robusta observed by confocal fluorescence microscopy. The chloroplastidic signal is red (autofluorescence), the nuclear signal is green (SYBR Safe), and the silica tracer signal is blue (PDMPO; Lysosensor). In photographs A, C, D, and H, the transmission light channel was permitted in order to visualize the outline of the cell. B to D, Cells shown in valve view. A, E to J, and L, Cells shown in girdle view. A, Cell with focus on a dividing dorsal chloroplast, displaying the central constriction (arrow). B, Cell with the dorsal chloroplast divided. C, Cell with both chloroplasts divided. D and E, Cell with fully rotated, valve-located chloroplasts and containing an undivided nucleus. F, Cell with valve-located chloroplasts and showing a blue band in the middle of the cell, extending from pole to pole. This band probably represents a girdle band (Supplemental Fig. S4). G, Cell just after karyokinesis. Each daughter nucleus is situated on a different side of the future division plane. H, Cell during cytokinesis. The division of the protoplast is apparent by the cleavage furrow, visible as a line of higher contrast running from pole to pole and in between both daughter nuclei. The PDMPO signal is confined to two "vacuole-like" compartments; as such, it appears that deposition of silica at the site of the cleavage furrow has not yet begun. I, Dividing cell during frustule formation. Deposition of silica into the frustule is visible as a line running from pole to pole and situated between the chloroplast pairs and daughter nuclei. J, Dividing cell with chloroplasts moving back from the valves toward the girdle. K, Separating daughter cells in slightly tilt valve view with focus on the dorsal chloroplasts, which are completely covering the girdle area. From the ventral chloroplasts, only the subcentral lobes are visible. L, A newly divided cell with the newly formed valve stained with PDPMO. The other valve is inherited from the mother cell. Bar = 10 mm. (see Supplemental Fig. S6 for an example of a cDNA-AFLP electropherogram; Vuylsteke et al., 2006). These expression profiles are of no value for expression analysis and were removed from the data set; only expression profiles that were highly reproducible (Pearson's correlation, P , 0.05) across the two strains were kept for further analysis. In this manner, 955 TDFs were selected and used for adaptive quality-based clustering (De Smet et al., 2002). With highstringency settings (see "Materials and Methods"), 917 TDFs were grouped in nine expression clusters (designated C1-C9; Supplemental Fig. S7). Figure 6 shows the expression patterns of all these TDFs during the cell cycle of both genotypes by hierarchical clustering (Eisen et al., 1998). TDFs in clusters C1 and C2 had their maximal expression during darkness (0 h), and the level of expression decreased drastically within the first few hours after darkness. Expression of TDFs in clusters C3 and C4 was induced after illumination, peaking at 2 and 3 h, respectively, after darkness. Average peak expression in cluster C5 occurred at 4 to 6 h after darkness, corresponding with the initiation of chloroplast division and the appearance of the first dividing cells. This cluster contained all identified TDFs with nucleotide-binding regions ( Table I). Two of these were the universal S-phase-specific genes, histone H3 and MCM5. Histone H3 (Sr025; Kapros et al., 1992;Menges et al., 2002) was maximally expressed at 7 h (Supplemental Fig. S8). The universal DNA replication licensing factor MCM5 (Sr027) had a discrete expression at 4 h. This latter gene is specifically involved in the initiation of DNA synthesis and unwinding of the DNA at the replication forks and has been shown to be temporally expressed at the G1/S transition (Tsuruga et al., 1997;Takisawa et al., 2000). Therefore, the expression of the MCM5 ortholog represents a good marker for S-phase initiation and was used to demarcate the G1 and S phases during synchronization. While genes in cluster C6 were discretely expressed at 7 h, genes in cluster C7 had a broader expression range. They were induced at 6 h, in correspondence with the first occurrences of dividing cells, until the 12th h at the moment of cell separation. Genes in this cluster were expressed simultaneously with the enrichment of G2-phase cells and dividing cells. They consisted of an early and late expression subgroup, indicated as C7-Early (C7-E), containing genes induced at 6 h, and C7-Late (C7-L), containing genes induced at 8 h. Clusters C8 and C9 comprised a rather small number of expression profiles, peaking at three (3, 7, and 10-12 h) and two (0 and 12 h) nonconsecutive time points, respectively.
In total, 378 TDFs (41%) with reproducible expression profiles across the two genotypes were selected for sequencing. Since we were interested primarily in the progression of the cell cycle, proportionally more TDFs were selected from clusters C5 and C7, in which expression was up-regulated after reillumination (Supplemental Table S2). Good-quality sequences were found in 322 TDFs (85%), of which 100 had significant (E-value , 10 23 ) similarity (Table I) with genes in the in-house-constructed database (see "Materials and Methods"). Expression patterns of these annotated TDFs are presented in Supplemental Figure  S9. Seventy-two TDFs showed homology with a gene with a putative allocated function, 18 were homologous with sequences without an allocated function (hypothetical proteins), and 10 TDFs matched with diatom orphan genes, for which no homologous counterparts exist (Table I). These latter genes were all confined to the late phases of the cell cycle, indicating that diatom-specific genes might be functionally associated with mitosis and cell separation. Based on the gene homology and the mapping of Gene Ontology (GO) labels and InterPro domains (see "Materials and Methods"; Supplemental Tables S3 and S4), the TDFs with annotated functions were classified into 11 distinct functional groups (Table I).

Modulated Expression of Genes Implicated in Cell Metabolism
The majority of the GO-labeled TDFs were active in cellular and metabolic processes (Supplemental Fig. S10). TDFs involved in protein biosynthesis were dominant in S-phase cluster C5. Two genes were found to be involved in amino acid synthesis, namely N-acetyl-g-glutamyl-phosphate reductase (Sr029) and the bifunctional ATP sulfurylase-adenosine 5#-phosphosulfate kinase (Sr033). Various components of the translation machinery dominated this cluster: three TDFs encoded ribosomal proteins (Sr030, Sr031, and Sr032) and one encoded an aspartyl-tRNA synthetase (Sr034). A translation initiation factor 3 (Sr050) was induced shortly after these genes, in cluster C7-E. In addition, nine posttranslational modification proteins were identified (Sr001, Sr002, Sr009, Sr010, Sr028, Sr047, Sr048, Sr049, and Sr073). Four of these genes were expressed in cluster C7-E but two of them, a ClpX homolog (Sr001; Weart et al., 2005) and a ubiquitinconjugating enzyme (Sr002), were highly expressed in the G1-phase cluster C2 when cells were released from the dark and were rapidly down-regulated after the first hour of light. A series of hypothetical proteins with undefined function displayed an identical regulation.
Several genes of the chloroplast-localized fatty acid biosynthetic pathways (Ohlrogge and Browse, 1995) were modulated during different phases of the cell cycle. b-Hydroxyacyl-ACP-dehydratase (Sr023) was identified in cluster C5, while two fatty acid desaturases (Sr044 and Sr045) were expressed some time later in cluster C7-E. Transcription of glycerol-3-phosphate dehydrogenase (Sr091), also referred to as dihydroxyacetone phosphate reductase (Gee et al., 1988), was activated during cell separation in cluster C9, starting 10 h after reillumination.
To drive the anabolic pathways of protein and fatty acid synthesis, the cell depends on an acetyl-CoA pool, which can be produced by the aerobic oxidation of carbohydrates (Fernie et al., 2004). In this respect, the glycolytic enzyme enolase (Sr017) was expressed in cluster C5 together with the mitochondrion-localized enzyme isocitrate dehydrogenase (Sr015) and a mitochondrial phosphate-carrier protein (Sr016), both needed during active citric acid cycling.
Two TDFs (Sr051 and Sr095), a subunit of the vacuolar (V)-type H + -ATPase and a V-type H + -pyrophosphatase, are known to play a role in vacuolar transport. More particularly, they are responsible for the acidification or maintenance of the acidity of organelles (Maeshima, 2001). Sr051 was activated during G2/M phase in cluster C7-E, while Sr095 was initially expressed slightly during darkness and again abundantly during cell separation (cluster C9).

Modulated Expression of Photosynthesis-Related Genes and Genes Involved in Chloroplast Movement
Fifteen TDFs were assigned to the functional category of photosynthesis (Table I). Using a x 2 correlation test (Supplemental Table S5), this functional category was found to be dominantly expressed (P , 0.001) in cluster C7-L: one TDF was found in C7-E and two in C9. Nine copies of the chromophyte-specific fucoxanthin-chlorophyll a/c-binding proteins (FCPs; Green and Durnford, 1996) were identified, and they were complemented with four TDFs involved in the biosynthesis of their bound photosynthetic pigments: porphobilinogen synthase (Sr063), protoporphyrinogen IX oxidase (Sr068), and protoporphyrin IX magnesium chelatase subunit H (Sr046). They all encoded enzymes from the chlorophyll a/c biosynthetic pathway (von Wettstein et al., 1995). In addition, a z-carotene desaturase (Sr062) was identified, being an intermediate enzyme in the biosynthetic pathway of b-carotene and other carotenoids, such as fucoxanthin (Wilhelm et al., 2006). Several Calvin cycle enzymes were also expressed in C7-L and C9: phosphoribulokinase (Sr070), transketolase (Sr093), and glyceraldehyde-3-phosphate dehydrogenase (Sr089). However, because glyceraldehyde-3-phosphate dehydrogenase is also an enzyme involved in glycolysis, it was assigned to the functional category of carbohydrate metabolism. Four genes operating in reactive oxygen metabolism were found in cluster C7-L: one TDF (Sr074) was highly similar to manganese superoxide dismutase (SOD; Wolfe-Simon et al., 2006), and three TDFs (Sr075, Sr076, and Sr077) were highly similar to the peroxisomal membrane protein MPV17/PMP22, which is known to up-regulate the activity of SODs in animal cells (Iida et al., 2003). These genes probably relate to photosynthetic reactive oxygen species generation, which is an inevitable part of oxygenic photosynthesis (Horton et al., 1996;Apel and Hirt, 2004). . Hierarchical clustered expression profiles of 917 TDFs reproducibly modulated during the cell cycle of synchronized cultures of strain F 1 -8B and F 1 -9A. Clustering was performed using TMEV software (Saeed et al., 2003) and the hierarchical clustering algorithm (Eisen et al., 1998). Each row represents a tag with the relative transcript accumulation patterns shown over 12 consecutive time points (columns) after reillumination of dark-arrested cultures. Yellow and blue color intensities reflect up-and down-regulation of gene expression relative to a range of +3 to -3, respectively; gray represents missing data. Cluster names (in accordance with adaptive quality-based clus-tering) and cell cycle phases are indicated at right and left, respectively. The arrow corresponds with the expression profile of the DNA replication licensing factor MCM5.  b-Tubulin (Sr019), which plays a fundamental role in cytoskeleton-based cellular movements, was highly induced 4 h and maximally expressed 6 h after darkness. This induction (clustered within C5) slightly preceded chloroplast reorganization. Another TDF with similarity to a WD40-repeat protein (Sr018), putatively involved in cytoskeleton assembly, was analogously expressed; both genes were down-regulated 6 h after darkness but b-tubulin was again slightly induced at 8 h, in parallel with cell division (Supplemental Fig. S11). The involvement of microtubulebased cytoskeleton dynamics during chloroplast movement was expected and further validated by treatment of synchronized cultures with the microtubule inhibitor nocodazole. Nine hours after treatment, chloroplast division and reorganization were impaired when compared with untreated control cultures. Only after 26 h did chloroplast division and reorganization proceed, arresting the cells at G2+M phase (data not shown).

Cell Cycle versus Light Regulation of Photosynthesis and Pigment Biosynthesis
The microscopic observations (Fig. 4) showed that chloroplasts elongated after cell division initiation when they relocated to the girdle sides of the cell. To investigate whether this elongation is paralleled with chloroplast growth, we determined fucoxanthin pigment turnover and validated the FCP expression with RT-Q-PCR in synchronized cultures. In addition, these data were compared with those from synchronized cultures to which the cell cycle inhibitor aphidicolin had been added shortly before reillumination (Fig. 7). This was done to assess the dependence of pigment turnover and FCP expression on cell cycle progression versus light regulation.
Upon reillumination, fucoxanthin content per cell increased under both conditions, but the trend was more pronounced in the dividing cultures than in the S-phase-arrested cultures. This fucoxanthin accumulation in the dividing cultures was obvious after 5 h of light and decreased after 10 h, corresponding with cell separation (Fig. 3). In the S-phase-arrested cultures, the accumulation curve was more gradual, without a clear peak and with a smaller maximum. Although the difference between both profiles was obvious, a longitudinal ANOVA test was unable to prove its significance, which could be due to the small number of replicates.
In parallel, in response to reillumination in both cultures, FCP expression was activated and subsequently down-regulated after 5 h in the arrested cultures, whereas in cell cycle-progressing cultures, its expression was maintained throughout the 13-h time course (Fig. 7). These findings suggest that chloroplast growth occurs in advance of chloroplast elongation at cytokinesis, while chloroplast lightharvesting complexes are regulated by both the cell cycle and light.

Regulation of Chloroplast Dynamics with Respect to Cell Cycle Progression and Frustule Silicification
In diatoms, chloroplast division and development have not yet been considered in relation to cell cycle regulation. For example, darkness was shown to arrest the cell cycle in several diatom species Brzezinski et al., 1990), but microscopic observations of the effect on chloroplasts are, to our knowledge, completely lacking. A putative reason is that mostly polyplastidic centric diatoms have been studied in which the small chloroplasts are stochastically partitioned upon cell division and, hence, have not such an obvious developmental cycle. In addition, the fact that flow cytometry is traditionally preferred to investigate cell cycle effects might have contributed to the general neglect of morphology and cytology.
As shown here for the pennate diatom S. robusta, chloroplast division and development are intimately linked with cell cycle progression. First, conditions that activate the G1-to-S cell cycle checkpoint (i.e. light deprivation, HU, or aphidicolin) arrest the development of chloroplasts at the plastokinesis stage without inducing aberrant cell morphologies. Second, both divided chloroplasts rotate from the girdle to the valves before karyokinesis is initiated. And, at last, the chloroplast division homolog FtsZ is expressed in concert with chloroplast division in a cell cycle phasedependent manner, during the S/G2 phases. We conclude that the G1-to-S-checkpoint controls chloroplast division and relocation and enables their synchronous division in S. robusta cultures. Previous observations of chloroplast dynamics during the cell cycle in S. robusta had already shown that chloroplasts and the nucleus divide in a coordinated manner (Chepurnov et al., 2002). Now, we observe that both also depend on each other and that common mechanisms to regulate their division should be involved. Similarly, in the red alga Cyanidioschyzon merolae, chloroplast and mitochondrion divisions are regulated at distinct cell cycle checkpoints (Nishida et al., 2005). However, in contrast with aphidicolin-arrested S. robusta cells, in C. merolae the mitochondrial division is restrained while the chloroplast divides multiple times (Itoh et al., 1996;Nishida et al., 2005).
Since its initial discovery in Arabidopsis (Arabidopsis thaliana; Osteryoung and Vierling, 1995), homologs of the bacterial cell division protein FtsZ are held responsible for organelle division in many other eukaryotes (Beech and Gilson, 2000;Takahara et al., 2000;Kiefel et al., 2004) that occurs by FtsZ assembling into the bacterial cytokinetic apparatus, called the Z-ring (Osteryoung and Nunnari, 2003;Margolin, 2005). Accordingly, FtsZ expression is modulated during synchronized growth of tobacco (Nicotiana tabacum BY2; El-Shami et al., 2002) and C. merolae (Takahara et al., 2000;Nishida et al., 2005), in a circadian rhythm in Chlamydomonas reinhardtii (Hu et al., 2008), and in light/dark-synchronized cultures of S. robusta as well. Its expression is up-regulated after S-phase progression and prior to cell division, in agreement with microscopic observations. This shows that the chloroplast developmental cycle can be monitored unmistakably through synchronization of cell division in S. robusta.
Control of chloroplast division is governed by conserved cell cycle regulators of bacterial origin (Adams et al., 2008). One of these, the chaperone ClpX, identified here, was down-regulated upon reillumination when cells reentered the cell cycle. ClpX has been found to inhibit Z-ring formation in Bacillus subtilis (Weart et al., 2005). It is tempting, therefore, to suggest that the activation of ClpX in a dark-arrested cell ensures the inhibition of plastokinesis in S. robusta. Figure 7. Comparison of fucoxanthin pigment turnover and FCP expression between light/dark-synchronized cultures and light/dark-synchronized cultures to which the cell cycle inhibitor aphidicolin was added shortly before reillumination. A difference in fucoxanthin pigment (mg mg 21 dry weight) accumulation is apparent after 5 h of reillumination. In the S-phase-arrested cultures, the buildup is less pronounced. In parallel, FCP expression is up-regulated between 1 and 5 h after reillumination in both dividing and arrested cultures, but its expression is down-regulated after the initial increase in arrested cultures while it is maintained in the dividing cultures. [See online article for color version of this figure.] Another resemblance can be found with the regulation of plastokinesis at the G1/S transition in Arabidopsis (Raynaud et al., 2005), in which chloroplast division was regulated by the activity of the prereplication factor CDT1. In S. robusta, chloroplast division was paralleled by activation of the S-phase-specific gene MCM5, which is also a universal member of the prereplication complex. Furthermore, our observations also suggest that "retrograde" mechanisms (Koussevitzky et al., 2007), which involve signaling from the chloroplast to the nucleus, might be present as another way to coordinate the chloroplast and nuclear division cycle. Retrograde signaling was, for example, reported in response to membrane tension as a possible way to regulate shape, size, and division of chloroplasts in Arabidopsis (Haswell and Meyerowitz, 2006). Analogous feedback mechanisms might exist within diatoms, for example, to control the G2/M transition in response to the chloroplast developmental status, ensuring high-quality chloroplast segregation of daughter chloroplasts to the newly divided cells. In any case, control mechanisms may not be as strict in centric polyplastidic diatoms because, similar to what is known for land plants, a stochastic partitioning of chloroplasts upon cell division will hardly result in cells without any chloroplast (Coleman and Nerozzi, 1999).
The molecular mechanism of chloroplast movement in diatoms is still unclear. In polyplastidic diatoms, the red alga C. merolae, and plants, chloroplast movements are known to be actin dependent (de Francisco and Roth, 1977;Nishida et al., 2005;Krzeszowiec et al., 2007). The transcriptional induction of b-tubulin in S. robusta suggests the involvement of microtubules during chloroplast rearrangement but does not exclude the possibility that actin may also be involved.
The intriguing question remains why pennate diatoms display this sometimes complicated cycle of repositioning the chloroplast during cell cycle progression. One possibility is that the position of the chloroplasts at the valves might create a disadvantage for cell movement needed for adequate reaction to environmental factors, such as light and nutrients (Cohn and Disparti, 1992). Because cells glide across the surface by mucilage secretion from a valvar slit (called the raphe; Hoagland et al., 1993;Wetherbee et al., 1998;Chiovitti et al., 2006), the underlying actinmyosin system that is held responsible for the raphe's functionality could be impaired by a valvar location of the chloroplasts. In cultures of S. robusta, cells with a reorganized chloroplast configuration have indeed never been observed moving. This reasoning was first introduced in Mereschkowsky's "law" on diatom chloroplasts, which states that chloroplasts of diatoms have a tendency to leave the raphe as much as possible uncovered (Mereschkowsky, 1904). This is achieved in numerous diatoms (Brebissonia spp., Cymbella spp., Gomphonema spp., Dickieia ulvacea, Didymosphenia spp., and Lyrella spp.) by plastid invaginations beneath the raphe (Mann, 1996), whereas in other diatoms, such as S. robusta, it is achieved by positioning of the chloroplasts at the valves only during a short time of the cell cycle. As observed here, this positioning at the valves is necessary only to free the girdle area for the addition of girdle bands, cytokinesis, and frustule formation. The identified girdle bands were indeed observed late during the cell cycle, in cells with a reorganized chloroplast configuration. As such, the cyclic movement of chloroplasts could create a large enough time frame for a cell to position itself optimally within the environment, maximally favoring cell growth, before cell division is committed. It is an interesting speculation that because centric diatoms are planktonic, lack a raphe, and are thus unable to move, they did not evolve such a well-orchestrated chloroplast development cycle. Also, variations in the chloroplast cycle in diatoms might explain the differences in the timing of girdle band deposition among species (Coombs et al., 1967b;Chiappino and Volcani, 1977;Krö ger and Wetherbee, 2000;Hildebrand et al., 2007). For example, T. pseudonana, which is polyplastidic, was shown recently to synthesize its girdle bands already during G1 phase (Hildebrand et al., 2007).
Girdle bands accommodate cell growth and, like frustules, are formed by polymerization of silica in a silica deposition vesicle (SDV) that is exocytosed (Pickett-Heaps et al., 1990;Krö ger and Wetherbee, 2000;Zurzolo and Bowler, 2001). Not much is known about the origin of the SDV, but it probably originates from other internal vesicle compartments, such as the endoplasmic reticulum, Golgi, and lysosome (Lee and Li, 1992). In this regard, the TDF Sr051, identified as a subunit of the V-type H + -ATPase, is of particular interest because such energy-consuming proton transport proteins have a known functionality in the acidification of organelles (Maeshima, 2001). It was maximally expressed 1 h before the initiation of cell separation (at 8 h after darkness), when the two "vacuolelike" compartments also were visible by PDMPO accumulation. Because diatom silicification occurs inside an acidic SDV (Mayama and Kuriyama, 2002), the V-type H + -ATPase could be related to girdle or valve formation at the stage of SDV acidification. A P-type ATPase was recently also hypothesized to be involved in SDV formation, during expansion of its plasma membrane in T. pseudonana (Frigeri et al., 2006). Possibly, both identified V-type and P-type ATPases are required in parallel during valve formation in diatoms.

Cell Cycle Synchronization of S. robusta
In many studies using phytoplankton cultures (for review, see Pirson and Lorenzen, 1966;Tamiya, 1966;Krupinska and Humbeck, 1994), the endogenous cell cycle control mechanisms that respond to naturally phased light/dark conditions are exploited to study cell division in synchronous cultures (Otero and Goto, 2005). Diatoms are known to be capable of sustaining long periods of darkness and retain the ability to start growing rapidly upon reillumination (Furusato et al., 2004). The synchronization protocol established here for S. robusta relies on this apparently fast release from the dark-arrested G1 phase. As a result of the identified common regulatory mechanisms in chloroplast division and cell cycle progression, this procedure was found successful for studying both processes simultaneously.
The uniformly dark-induced G1-phase arrest in S. robusta appears unusual when compared with other diatoms. The centric diatom Thalassiosira weissflogii accumulates, besides G1-phase cells, 60% G2+Mphase cells Vaulot et al., 1986;Brzezinski et al., 1990), and other centric diatoms (Chaetoceros muellerii, T. pseudonana, Chaetoceros simplex, and Minutocellus polymorhus) plus one pennate diatom (Cylindrotheca fusiformis) all accumulate G1-phase cells together with a smaller fraction (approximately 10%) of G2+M-phase cells (Brzezinski et al., 1990). The pennate diatom P. tricornutum is the only diatom reported without a dark-sensitive G2+M phase, and its facultative silicon requirement has been put forward as the main reason for this observation (Brzezinski et al., 1990). However, in diatoms, the G2+ M-phase fraction in flow cytometric signatures includes, besides G2-and M-phase cells, also postcytokinetic doublet cells, which have completed cell wall formation but have not yet separated. Therefore, the "G2+M-phase" fraction in the dark-arrest experiment by Brzezinski et al. (1990) might have been postcytokinetic doublet cells. In S. robusta, the occasionally detected G2+M-phase cells at 24 h of dark arrest were microscopically observed to be all postcytokinetic doublet cells containing girdle-located chloroplasts, typical for G1-phase cells. Apparently, only the process of cell separation and not cytokinesis is hampered in these cells by the absence of light. Therefore, we suggest that it might be useful to address cell separation in diatoms analogously as in yeast cells, where it is defined as a postmitotic phase and known to be molecularly distinct from the septum formation process during M phase (Yeong, 2005;Sipiczki, 2007).
Based on the expression of the prereplication factor MCM5 and the occurrence of dividing chloroplasts, the duration of the G1 phase in S. robusta can be estimated to last for 4 h (33% of the total division time) during synchronization under the applied culture conditions. Since the last 3 h were dominated by separating cells, 5 h (41% of the total division time) are left to fulfill the S, G2, and M phases. Our results suggest that cell division during synchronization occurred faster (,12 h) than during normal exponential growth, even when compared with the highest recorded maximum division rate for S. robusta, being 16.6 h per division (approximately 1.5 divisions per day under continuous light at 200 mE). As suggested Vaulot et al., 1986), part of the G1 phase might already be accomplished during dark arrest. Taking into account this comment, our microscopically estimated cell cycle stage durations of S. robusta correspond roughly with the flow cytometric estimations from P. tricornutum, reported to spend half its cell division time in G1 phase (Brzezinski et al., 1990).

Gene Expression in Relation to Chloroplast Functioning and Photosynthesis
The overall results from the cDNA-AFLP experiment confirm the hypothesis that several biological processes are under temporal transcriptional control during the cell cycle of diatoms, as shown previously in similar studies of higher plants and animals (Cho et al., 2001;Menges et al., 2002;Breyne et al., 2002). One of the most distinct patterns in gene modulation concerns those genes encoding chloroplast proteins, associated with the functioning of chloroplasts during photosynthesis, including light-harvesting proteins, biosynthetic pigment enzymes, and genes coping with oxidative damage. Nearly all of these genes were expressed during G2/M-phase progression and clustered into the expression cluster C7-L. Most notable was the expression of FCP, a chromophyte-specific type of light-harvesting protein (Bhaya and Grossman, 1993;Rhiel, 1998, 2000;Guglielmi et al., 2005), in concert with cell division.
In S. robusta, FCP expression was found to be light activated in the first place, confirming previous results in centric (Leblanc et al., 1999;Oeltjen et al., 2002Oeltjen et al., , 2004 and pennate  diatoms. But after 5 h, its expression became cell cycle dependent, because it was maintained in dividing cultures while being repressed in S-phase-arrested cultures. A possible molecular mechanism to explain this difference relates to the retrograde inhibition of light-harvesting proteins known in C. reinhardtii and Arabidopsis. In these organisms, light-induced expression of light-harvesting proteins is repressed by accumulation of the chlorophyll precursor Mg-protoporphirin IX (Johanningmeier and Howell, 1984;Nott et al., 2006). Its accumulation, leakage from the chloroplast, and nuclear repression of photosynthetic genes are thought to occur during reduced chloroplast function to coordinate pigment biosynthesis with the expression of light-harvesting proteins (Nott et al., 2006). In an analogous way, if chloroplast development is impaired in S. robusta, a retrograde signal might repress FCP. The fact that fucoxanthin does not decrease together with FCP could be a result of the low turnover rates of diatom pigments, ranging from days to weeks (Riper et al., 1979;Goericke and Welschmeyer, 1992), compared with the low stability of mRNA.
A recent study by Ragni and Ribera d'Alcalà (2007) on the circadian variations in pigment content in P. tricornutum showed that pigment synthesis follows the somatic growth preceding cell division. Furthermore, they observed that a pronounced circadian pattern of pigment synthesis was lost when cultures were continuously illuminated and became asynchronous. This study substantiates our finding that fucoxanthin content increases in preparation of cell division, until it is distributed into daughter cells.
Taking together our microscopic analysis, cDNA-AFLP results, and pigment accumulation patterns, we suggest that cell cycle modulation of photosynthetic light harvesting occurs in concert with chloroplast growth to prepare for its elongation at cytokinesis. Because chloroplasts in S. robusta elongate at the stage of cytokinesis, regulated synthesis of more functional light-harvesting complexes would be expected to provide a balanced growth of chloroplasts at this stage, thereby optimizing the cell's photosynthetic capacity during the subsequent G1 phase. Together with light acclimation and circadian regulation, cell cycle-modulated biosynthesis of the large amount of pigments that are contained within diatoms (Chan, 1978(Chan, , 1980Falkowski et al., 1985;Tang, 1996) could as such coordinate cellular investments and improve cell fitness.

Cell Culture and Image Acquisition
The Seminavis robusta strains F 1 -8B and F 1 -9A were selected from a first generation of siblings obtained by crossing the wild-type clones 75 and 80 (Chepurnov et al., 2002). F 1 -8B and F 1 -9A belong to opposite mating types (Chepurnov et al., 2008) and had at the time of study an average apical cell length of 22.7 61.6 and 18.9 61.2 mm, respectively. Currently, they are maintained cryopreserved in the culture collection of the Laboratory of Protistology and Aquatic Ecology (http://www.pae.ugent.be/collection. htm). Cell cultures were grown in F/2 medium (Guillard, 1975) made with filtered (GF/C grade microfiber filter; Whatman) autoclaved seawater collected from the North Sea. F/2 nutrients were sterilized through 0.2-mm filters and added to the filtered autoclaved seawater. Na 2 SiO 3 was added at a concentration of 30 mg L 21 medium. Cultures were cultivated at 18°C with a 12:12-h light:dark period and approximately 85 mmol photons m 22 s 21 from cool-white fluorescent lights. Stock cultures were reinoculated weekly by transferring small aliquots of cell suspension into fresh medium. Experimental cultures were prepared from stock cultures by inoculating an aliquot of cells from cell suspensions created by scraping the cells from the surface (cell scrapers; Sarstedt) or by detaching the cells by pipetting. Observations and cell culture photography were done with a Zeiss Axiovert 40 light microscope and a digital camera (Powershot G3; Canon). Fluorescence microscopy of cells was performed using Zeiss Axioplan 200, Axiocam MRm for image capturing, and Zeiss filter set 14. Confocal fluorescence images were taken with a laserscanning confocal microscope (Zeiss Confocal LSM 510) equipped with software package LSM510 version 3.2 (Zeiss) and equipped with a 633 water-corrected objective (numerical aperture 1.2). Chloroplast fluorescence was visualized with helium-neon laser illumination at 543 nm. Nuclear DNA was stained with SYBR Safe (Molecular Probes) that was visualized with argon laser illumination at 488 nm and a 500-to 530-nm band emission filter. Cell walls were stained with the Lysosensor Yellow/Blue DND-160 probe (PDMPO; Molecular Probes) as described by Leblanc and Hutchins (2005) and visualized by illumination at 351 nm. Emission fluorescence was captured in the line-scanning mode, and for transmission light images, differential interference contrast optics were used.

Cell Cycle Progression Assays and Flow Cytometry
The effect of darkness on cell morphology was tested on exponentially growing cultures of strain F 1 -8B (Chepurnov et al., 2008) that were inoculated at 1,000 cells mL 21 in tissue culture flasks (Cellstar; 175-cm 2 growth surface with filter screw cap; Greiner Bio-One) with 200 mL of growth medium. After 2 d of growth, the old medium was decanted from the flask-attached cells and the flask was refilled with 200 mL of fresh medium before transferring the cultures to complete darkness to ensure that no nutrient limitation occurred during dark incubation. HU (Fluka) was tested at a final concentration of 6.5 mM, added to 100 mL of exponentially growing cultures of strain F 1 -31B. After addition, the cultures were incubated in continuous light during 72 h and compared with blank cultures to which an equal volume of dimethyl sulfoxide was added. The effect of aphidicolin (Sigma-Aldrich; 0.5 mg mL 21 ) was tested in triplicate on F 1 -8B cultures containing approximately 5,000 darkarrested cells in 10 mL of medium and compared after 24 h of light with dimethyl sulfoxide-treated control cultures.
DNA content was measured on intact fixed cells. Therefore, at least 10 mL of a culture was centrifuged for 10 min at 1,500g, and the cells were fixed in 10 mL of ice-cold methanol in the dark at 4°C . Fixed cells were rinsed two to three times with 4 mL of TE (10 mM Tris and 1 mM EDTA, pH 8.0) by centrifugation for 10 min at 1,500g and resuspended in 1 mL of TE. Samples were incubated for 40 min at 37°C with 10 mL of 30.5 mg mL 21 RNase A (R4642; Sigma-Aldrich) and then placed on ice in the dark. To each sample, 1 mL of 4,6-diamidino-2-phenylindole from a stock of 1 mg mL 21 was added, stained for at least 10 min on ice, and filtered over a 50-mm nylon mesh (CellTrics; Partec). The stained cell suspensions were analyzed with a CyFlow flow cytometer and FloMax software (Partec).

Synchronization of S. robusta Cells and Sampling of Material
For the synchronization, each S. robusta strain was grown in two tissue culture flasks. Because of the difference in cell size, F 1 -8B and F 1 -9A culture flasks were inoculated with 7.5 3 10 5 and 1 3 10 6 cells, respectively, into 200 mL of medium. Cell density estimates of stock cultures were obtained by microscopically counting cells in a 100-mL aliquot of a suspended culture on a 96-well TC plate. Cultures inoculated for synchronization were grown for 2 d in a 12:12-h light:dark regime. At the end of day 2, the dark period was extended for another 12 h, arresting the cell culture at the G1 phase. A first sample was taken just before the light was switched on, followed by 12 samples taken every hour after reillumination. Just before sampling of each culture, six photographs were taken with the Axiovert 40 microscope and a connected digital camera (Canon Powershot G3). Images were used to count different cell types using the open-source software ImageJ (http://rsb.info. nih.gov/ij/index.html) and the cell counter plug-in. M-phase cells were easily identified and distinguished from interphase cells by the presence of a newly built cell wall, situated between the two valve-located chloroplasts. As soon as daughter cells were separating, they were counted as nondividing cells. In preparation for cell culture harvesting, the cells were concentrated by reducing the medium of the first flask to less than 50 mL by aspiration with a water pump and attached Pasteur pipette. After suspension through scraping (Sarstedt), the cells were added to the second bottle, from which the complete medium was aspirated. The collection of suspended cells from the two bottles was transferred to a 50-mL Falcon tube and centrifuged for 6 min at 1,500g. The supernatant was poured off, and the tube containing the pellet was frozen in liquid nitrogen and stored at 280°C until RNA preparation.

cDNA-AFLP-Based Transcript Profiling
Total RNA was extracted from each cell sample using the RNeasy Plant Mini Kit (Qiagen). Cell lysis was achieved by mechanical disruption in 600 mL of RNeasy Lysis buffer (Qiagen) by highest speed agitation with glass/ zirconium beads (0.1 mm diameter; Biospec) on a bead mill (Retsch). All other steps for RNA extraction were done according to the manufacturer's instructions. RNA concentration and purity were assessed by spectrophotometry (Nanodrop ND-1000 spectrophotometer) and denaturing agarose gel electro- phoresis. First-and second-strand cDNA synthesis and cDNA-AFLP analysis, with BstYI and MseI as restriction enzymes, were carried out according to Vuylsteke et al. (2007), starting from 2 mg of total RNA. The final selection of the cDNA-AFLP amplification products was done with three selective nucleotides, resulting in 128 primer combinations. Expression profiles across the 13 time points for each genotype were quantified with AFLP-QuantarPro software (Keygene). The raw expression values within each genotype were normalized per gene by subtracting the average expression value of each gene from each data point in the time series and dividing it by the SD. Reproducibility of expression profiles between strains F 1 -8B and F 1 -9A was estimated for every gene by calculating a Pearson correlation between normalized expression data. Expression profiles showing a positive and significant (P , 0.05) correlation across the two strains were submitted to adaptive qualitybased clustering (De Smet et al., 2002). The minimal number of genes in a cluster was set to 10, and the minimal probability of genes belonging to a cluster was set to 0.95. Hierarchical clustering (Eisen et al., 1998) was performed with TMEV software (Saeed et al., 2003).

Identification of Differentially Expressed Genes
TDFs were purified from the gel, followed by amplification and subsequent sequencing as described by Vuylsteke et al. (2007). The sequence quality was checked through inspection of the electrophoretic peaks. With the goodquality sequences, a similarity search was done with BLASTN and BLASTX sequence alignments (Altschul et al., 1997) against nucleotide and protein sequences in the publicly available GenBank databases, to which the latest versions of the more recently released genomes of Phytophthora soja (U.S. Department of Energy Joint Genome Institute [JGI], v1.1), Phytophthora ramorum (JGI, v1.1), Phaeodactylum tricornutum (JGI, v2.0), and Thalassiosira pseudonana (JGI, v3.0) were added. In case a BLASTN alignment of more than 30 nucleotides with 80% coverage between the query and the hit alignment was achieved, the hit sequence was used as a query for similarity searching with BLASTX. Based on the homology (E-value cutoff, 1 3 10 23 ; lowest percentage identification, 29) and the identity of functional GO domains and InterPro domains, the TDFs were functionally annotated and classified into 11 functional groups. TDFs without functional annotation were classified as unknown. GO and InterPro annotations were performed with the online version of the Blast2GO v1.7.2 program (www.Blast2GO.de; Conesa et al., 2005). The program extracts the GO terms associated with homologies identified with the National Center for Biotechnology Information's QBLAST and returns a list of GO annotations represented as hierarchical categories of increasing specificity. Because the diatom genome sequences are not included yet in GenBank, we used the complete sequences of their corresponding diatom genes as input sequences. For the construction of Supplemental Figure  S10, the data were analyzed at level 2 and GO classes were set to represent at least five different TDFs.

Cloning of Putative FtsZ cDNA from S. robusta
Based on conserved amino acid regions of known FtsZ genes in P. tricornutum (JGI, v2.0 protein identifiers 14995 and 14426) and T. pseudonana (JGI, v3.0 protein identifiers 35728, 269655, and 15398), 12 oligonucleotide primers were designed with CODEHOP (CDH; Rose et al., 1998) and by manual (m) design (Supplemental Fig. S5). These primers were tested in pairwise combinations, and three (m_FP1, m_RP1b, and CDH_FP4) were successfully used to amplify FtsZ cDNA. First, PCR with m_FP1 and m_RP1b primers corresponding to amino acid sequences W(A/S)(I/L/V)NTDAQA and (I/V)NVDFAD, respectively, was carried out using hot-start PCR (AmpliTaq Gold DNA polymerase; Applied Biosystems) in five steps: 9 min at 95°C; four cycles of 1 min at 94°C, 1 min at 35°C, and 1 min at 72°C; four cycles of 1 min at 94°C, 1 min at 45°C, and 1 min at 72°C; 30 cycles of 1 min at 94°C, 1 min at 50°C, and 1 min at 72°C; and one cycle of 10 min at 72°C. A 20-folddiluted, pooled cDNA sample constituting all cell cycle phases during synchronization was used as template, and 20 pmol of each primer was used in the 25-mL reaction mixture. The PCR products were analyzed on a 1.5% agarose gel, and a band of the correct size (approximately 500 bp) was isolated from the gel and extracted with Nucleospin Extract II (Macherey-Nagel). This extract served as a template in the subsequent PCR with nested primer CDH_FP4, corresponding to amino acid sequence VGIVTKPF, and primer m_RP1b under the following conditions: 9 min at 95°C; 35 cycles of 45 s at 94°C, 1 min at 52°C, and 1 min at 72°C; followed by one cycle of 10 min at 72°C. The resulting DNA fragment of correct size (approximately 300 bp) was ligated into a pCR 4 TOPO vector (Invitrogen) and sequenced. The DNA sequence of the insert was 92% identical to that of the P. tricornutum FtsZ protein 14995. Two specific primers (Sr FtsZ-FP and Sr FtsZ-RP) were designed with Roche ProbeFinder on the cloned sequence for RT-Q-PCR.

RT-Q-PCR Assay
RNA was isolated and quantified as before. Isolated RNA was treated with DNaseI (GE-Healthcare) to remove all contaminating genomic DNA. An aliquot of 0.5 mg of total RNA from each sample was used for cDNA synthesis. The reverse transcription was carried out in a total volume of 20 mL with oligo(dT) primers and the SuperScript II kit (Invitrogen) according to the manufacturer's instructions. A 0.5-mL aliquot of a 5-fold dilution of the cDNA (2.5 ng) served as template for each RT-Q-PCR. Primers for FCP quantification and normalization were designed using the Beacon Designer 7.0 (Premier Biosoft International; Supplemental Fig. S5) together with a stringent set of primer design criteria, including predicted melting temperatures of 58.0°C 6 2.0°C, primer lengths of 17 to 22 nucleotides, and amplicon lengths of 75 to 150 bp. Primer pairs were tested by RT-Q-PCR on a pooled cDNA sample under the same conditions as described below. Primer reliability was confirmed by the appearance of a single peak in the melting curve analysis performed by the PCR machine after completion of the amplification reaction.
Eight constitutively and moderately expressed cDNA-AFLP TDFs were selected as candidate genes to serve for internal reference during RT-Q-PCR. These TDFs were excised from the cDNA-AFLP gels, reamplified, and sequenced, and primer pairs were designed as before. The expression stability (M) of the putative normalization genes was analyzed with the geNORM program: genes with the lowest M value are the most stably expressed and were selected as normalization genes (Vandesompele et al., 2002). TDFs sr13af_M281.1 and sr07ae_M536.3 were identified by geNORM to be the two most stably expressed genes (M = 0.426), followed by sr14ah_M385.6 (M = 0.540). The value for the pair-wise variation between two sequential normalization factors (geNORM factor V 2/3 ) was 0.146 during synchronization, which is smaller than the cutoff value of 0.15 proposed by Vandesompele et al. (2002) for reliable data normalization. Therefore, all three genes were used for data normalization in the subsequent RT-Q-PCR analyses.
RT-Q-PCR was performed on the Lightcycler 480 (Roche) platform. Each sample was assayed in triplicate under the following conditions: 2.5 ng of template cDNA, 2.5 mL of the Lightcycler 480 SYBR Green I Master Mix (Roche Applied Science), and 2 mL of primers at concentration of 0.5 mM. The cycling conditions comprised 10 min of preincubation at 95°C and 45 cycles of 10 s at 95°C, 15 s at 58°C, and 15 s at 72°C. Amplicon dissociation curves (i.e. melting curves) were recorded by heating at 95°C for 5 s and at 65°C for 1 min. Samples were cooled at 40°C for 10 s. The relative comparison DDC t method (Pfaffl, 2001) was used to evaluate expression levels of the selected genes relative to the expression of the normalization genes in the same sample. Data were analyzed and normalized with qBase (Hellemans et al., 2007) for expression profile generation.

HPLC Pigment Analysis
S. robusta pigments were sampled from light/dark-synchronized cultures (see above) by filtering 25 mL of suspended culture over a preweighed 25-mm Whatman glass fiber filter. Filters were wrapped in aluminum foil, frozen in liquid nitrogen, and stored at 280°C. Before analysis, the filters were lyophilized for 8 h and weighed (0.1-mg accuracy) to calculate the dry weight of each sample. Pigments were extracted from the filters in 90% acetone by means of sonication (tip sonicator; 40 W for 30 s). Extracts were filtered over a 0.2-mm Alltech nylon syringe filter to remove particles and injected into a Agilent 1100 series HPLC system (ChemStation software) equipped with a Macherey-Nagel reverse-phase C18 column (Nucleodur C18 pyramid; 5 mm particle size). Pigments were analyzed according to the method of Wright and Jeffrey (1997) using a gradient of three solvents: 80% methanol-20% ammonium acetate, 90% acetonitrile, and ethyl acetate. Two detectors were connected to the HPLC system: an Agilent standard fluorescence detector to measure fluorescence of chlorophylls and their derivatives and an Agilent diode array detector to measure absorbance of each peak at 436 and 665 nm and absorbance spectra over a 400-to 700-nm range. Fucoxanthin was identified by comparison of retention times and absorbance spectra and quantified by calculating response factors using pure pigment standards (supplied by DHI).
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers FJ388875 and GE343005 to GE343326.

Supplemental Data
The following material is available in the online version of this article.
Supplemental Figure S1. Schematic presentation of the chloroplastidic events during the cell cycle of S. robusta.
Supplemental Figure S2. Flow cytometric DNA content histograms for assessing the S-phase cell cycle arrest by HU.
Supplemental Figure S3. Estimation method for percentage of cells in different cell cycle phases based on photographs of the synchronized culture.
Supplemental Figure S4. Laser-scanning confocal photographs of a S. robusta cell to illustrate the addition of one girdle band.
Supplemental Figure S5. Identification of a S. robusta FtsZ ortholog and design of specific RT-Q-PCR primers for FtsZ and FCP quantification.
Supplemental Figure S6. Section of an electropherogram of the cDNA-AFLP fingerprint of the two S. robusta strains.
Supplemental Figure S7. The nine clusters identified by adaptive qualitybased clustering (De Smet et al., 2002) of 917 expression profiles that were highly reproducible.
Supplemental Figure S8. Expression of the S. robusta MCM5 and histone H3 ortholog as a function of time after reillumination.
Supplemental Figure S9. Clustering of the expression profiles of the cell cycle-modulated, annotated TDFs.
Supplemental Figure S10. Blast2GO pie chart representing the distribution of functional GO labels for biological processes.
Supplemental Figure S11. Expression of the S. robusta orthologs of b-tubulin and the WD40-repeat protein as a function of time after reillumination.
Supplemental Table S1. Representation of the cDNA-AFLP analysis, including all quantified TDFs, annotations, and expression data.
Supplemental Table S2. Distribution of selected genes with respect to their expression cluster.
Supplemental Table S3. Mapping of GO labels to the annotated TDFs, together with their assigned functional categories.
Supplemental Table S4. Mapping of InterPro domains to the annotated TDFs.
Supplemental Table S5. Calculation of x 2 correlation between photosynthetic genes and expression cluster C7-L.