This Article Abstract Full Text (PDF) Supplemental Data Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via CrossRef Citing Articles via ISI Web of Science (46) Citing Articles via Google Scholar Google Scholar Articles by Singh, A. K. Articles by Sherman, L. A. Search for Related Content PubMed PubMed Citation Articles by Singh, A. K. Articles by Sherman, L. A. Agricola Articles by Singh, A. K. Articles by Sherman, L. A.

GENETICS, GENOMICS, AND MOLECULAR EVOLUTION

Microarray Analysis of the Genome-Wide Response to Iron Deficiency and Iron Reconstitution in the Cyanobacterium Synechocystis sp. PCC 6803^1,[w]

Department of Biological Sciences (A.K.S., L.A.S.) and Computational Genomics and Department of Agronomy (L.M.M.), Purdue University, West Lafayette, Indiana 47907

	ABSTRACT

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 
A full-genome microarray of the (oxy)photosynthetic cyanobacterium Synechocystis sp. PCC 6803 was used to identify genes that were transcriptionally regulated by growth in iron (Fe)-deficient versus Fe-sufficient media. Transcript accumulation for 3,165 genes in the genome was analyzed using an analysis of variance model that accounted for slide and replicate (random) effects and dye (a fixed) effect in testing for differences in the four time periods. We determined that 85 genes showed statistically significant changes in the level of transcription (P 0.05/3,165 = 0.0000158) across the four time points examined, whereas 781 genes were characterized as interesting (P 0.05 but greater than 0.0000158; 731 of these had a fold change >1.25x). The genes identified included those known previously to be Fe regulated, such as isiA that encodes a novel chlorophyll-binding protein responsible for the pigment characteristics of low-Fe (LoFe) cells. ATP synthetase and phycobilisome genes were down-regulated in LoFe, and there were interesting changes in the transcription of genes involved in chlorophyll biosynthesis, in photosystem I and II assembly, and in energy metabolism. Hierarchical clustering demonstrated that photosynthesis genes, as a class, were repressed in LoFe and induced upon the re-addition of Fe. Specific regulatory genes were transcriptionally active in LoFe, including two genes that show homology to plant phytochromes (cph1 and cph2). These observations established the existence of a complex network of regulatory interactions and coordination in response to Fe availability.

Cyanobacteria are (oxy)photosynthetic organisms in which Fe stress has been studied in some detail (Straus, 1994; Behrenfeld and Kolber, 1999). Fe deficiency results in a variety of physiological and morphological changes in cyanobacteria, the most obvious of which is a significant change in cellular pigmentation. The overall changes include: loss of the light-harvesting phycobilisomes (Guikema and Sherman, 1983), changes in the fluorescence and absorption characteristics (Guikema and Sherman, 1983, 1984; Pakrasi et al., 1985a, 1985b), reduction in the number of thylakoids (Sherman and Sherman, 1983), and replacement of proteins with cofactors containing Fe to those with non-Fe cofactors, such as ferredoxin with flavodoxin (Laudenbach and Straus, 1988; Laudenbach et al., 1988; Straus, 1994). Most importantly, a novel chlorophyll (Chl)-binding protein, encoded by isiA, is synthesized (Pakrasi et al., 1985b; Laudenbach and Straus, 1988; Burnap et al., 1993). This IsiA protein resembles CP43 (it is sometimes termed CP43') and has recently been shown to form an 18-mer around the PSI trimer (Bibby et al., 2001; Boekema et al., 2001). This gene is also regulated by salt concentrations (Vinnemeier et al., 1998) and may be the ancestor of the Prochlorococcus spp. Chl a₂/b₂ light-harvesting protein (Ting et al., 2002). Many other changes are associated with Fe deficiency in cyanobacteria and cells continue to grow, although the growth rate is somewhat lower and the cells are smaller. The addition of Fe to the cultures can reverse these changes, and cells return to normal within 18 to 24 h after the addition of Fe (Sherman and Sherman, 1983; Riethman et al., 1988).

Microarray technology permits an assay of global gene expression patterns under a variety of experimental conditions. These arrays are particularly efficient in organisms for which the entire genome has been sequenced, such as Synechocystis sp. PCC 6803, which is now thought to have 3,264 genes (Kaneko et al., 1996; see Cyanobase at http://www.kazusa.or.jp/cyano/cyano.html). Microarrays have been developed for many systems, including for bacteria such as Escherichia coli (Richmond et al., 1999; Tao et al., 1999; Arfin et al., 2000) and for plants such as Arabidopsis (Pérez-Amador et al., 2001; Seki et al., 2001). A series of papers have appeared utilizing Synechocystis sp. PCC 6803 microarrays (Hihara et al., 2001; Suzuki et al., 2001; Gill et al., 2002; Kanesaki et al., 2002). These arrays have been used to monitor changes in different environmental parameters. The arrays that we constructed, in conjunction with the laboratory of Dr. Rob Burnap (Oklahoma State University, Stillwater), contain (in triplicate) cDNAs up to 2 kb of the 3,165 genes annotated in the Kazusa sequence before May 2002.

The substantial pigmentation changes under Fe-deficient growth provide an easy way to determine the cellular response to Fe deficiency or the redevelopment of the normal phenotype. Thus, Fe deficiency is an ideal system in which to study global gene expression in cyanobacteria. In a previous study, we developed a differential expression using customized amplification library for the analysis of global gene expression in the unicellular cyanobacterium, Synechocystis sp. PCC 6803 (Singh and Sherman, 2000). We now extend this study through an analysis of a full-genome microarray of Synechocystis sp. PCC 6803. We identified transcriptional changes in many genes that code for proteins involved in assembly or disassembly processes (e.g. chaperones and proteases) and in the structural proteins (e.g. IsiA or phycobiliproteins). The arrays also enabled us to detect genes involved in the regulation of these processes and for those that encode proteins needed for the acquisition and storage of Fe (Katoh et al., 2000, 2001). In this study, we identify many genes that are transcriptionally regulated during Fe deficiency and after the re-addition of Fe and that provide new insights into optimization of biological processes that enable cells to grow during nutrient limitation.

	RESULTS

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

Array Data and Statistical Analysis

The loop design utilized in this study is discussed in "Materials and Methods" and outlined in Figure 1A. This approach allows comparison among all conditions via the ANOVA model (Churchill, 2002; Yang and Speed, 2002). We were most interested in determining the changes in gene transcription, as a function of time, between the Fe-deficient and -sufficient states at different time points after the reintroduction of Fe. We previously demonstrated that similar physiological and ultrastructural changes occur as we go from Fe sufficiency to Fe deficiency or vice versa, and these results are inherent in the loop design (Sherman and Sherman, 1983; Riethman et al., 1988). Thus, we compared each of the Fe-sufficient states with the Fe-deficient state to identify the major changes in certain classes of genes that were strongly regulated by the presence/absence of Fe. Therefore, we use the terms induced and repressed to reflect the increase and the decrease, respectively, in the level of transcription from Fe sufficiency to deficiency.

View larger version (22K): [in this window] [in a new window]   Figure 1. A, Diagrammatic representation of the loop design utilized for identification of differentially expressed genes in response to iron availability. A total of six slides was used with dye swaps between the 0- and 24-h time points and between 3- and 24-h time points. B, Scatter plot that compares the mean spot intensities of 0 h (low iron [LoFe]) and 12 h (+Fe). Data from 0- and 12-h time points were normalized, and their mean signal intensities were plotted. The black line represents equal labeling for the two samples, whereas the dotted (2-fold) and dashed (3-fold) lines identify genes that demonstrated large labeling differences during the hybridization experiment.

The scatter plot in Figure 1B represents the relationship of the average hybridization intensities of LoFe (0 h) versus 12 h plus Fe. This simple procedure permitted an overview of the data and indicated that most of the spots fell along the diagonal and were equally labeled. Those spots that fell off the diagonal were candidates for genes with expression changes and lines indicating 2-fold (dotted) and 3-fold (dashed) changes are shown. Some of the published reports on microarray analysis have used an arbitrary cutoff of a 2-fold change to identify differentially expressed genes. However, it has been shown that changes in gene expression smaller than that of 2-fold can be reliably identified (Arfin et al., 2000; Jin et al., 2001; Long et al., 2001; Yue et al., 2001; Oleksiak et al., 2002; Yang et al., 2002). Jin et al. (2001) reported that changes in the gene expression as small as 1.2-fold can be considered as differentially expressed if a robust statistical method is used. We present an ANOVA modeling approach similar to that reported by Oleksiak et al. (2002) that used the same type of experimental design (see "Materials and Methods"). This enabled us to identify genes that displayed differential expression across the four time points examined. We selected genes for additional consideration using a significance level of 0.05 and at least a 1.25-fold change in transcript level intensity.

One advantage of the LoFe system was the known regulation of the isiAB genes (up-regulated under Fe-deficient conditions) and the phycobilisome genes (down-regulated under Fe-deficient conditions). We used these genes as markers during the early stages of this study to optimize hybridization conditions and spot analysis. The conditions described in "Materials and Methods" demonstrated a 22-fold increase in isiA expression under Fe-deficient conditions with a P value of 6 x 10^–10. The isiA gene represented the largest transcriptional increase among all of the genes, whether we analyzed all significant or interesting genes together or by functional category.

Differential Expression in Response to Fe

Using the criteria described in "Materials and Methods," we identified 85 differentially expressed, statistically significant genes and 731 statistically interesting genes. An additional 50 genes that had statistically significant or interesting P values but violated the normality assumption were included in our considerations after careful examination revealed that the heteroscedasticity of variance causing the departure from normality was due to extreme differences between the Fe-deficient and -sufficient states. Table I shows the number of genes differentially expressed in each functional category as defined in Cyanobase, whereas Table II highlights some specific genes that demonstrated transcriptional changes. The complete statistical analysis for all 3,165 genes can be found in Supplemental Data Table I (see http://www.plantphysiol.org). In addition, the final list of 866 genes examined in the functional analysis can be found in Supplemental Data Table II (see http://www.plantphysiol.org). The genes that were either up- or down-regulated at different times under Fe-deficient conditions or after the addition of Fe are plotted in a Venn diagram (Fig. 2). This figure highlights the transcriptional changes in this system. It is notable that two-thirds of the genes that display differential expression were down-regulated in the Fe-deficient state and that nearly one-third (183/601) had decreased transcription in the Fe-deficient state compared with all three Fe-sufficient states.

View larger version (19K): [in this window] [in a new window]   Figure 2. Venn diagrams that display the number of iron-responsive, differentially expressed genes. Three sample pairs (0/3, 0/12, and 0/24) were compared for increased (A) and decreased (B) transcript levels in response to iron availability. All of the genes had a fold change > 1.25 and P < 0.05. The numbers in the overlapping areas indicate genes that exhibited differential expression in either two or three sample pairs. The numbers in the nonoverlapping areas indicate the number of genes that exhibited differential expression in one sample pair.

We used hierarchical clustering to explore the differential expression as a function of time after the addition of Fe (Fig. 3, A and B). Six of the kinetic categories are diagrammed in Figure 3C and are numbered based on their position from top to bottom within Figure 3A. Category 1 represented the largest group (n = 437) and included approximately 50% of these differentially expressed genes. The genes in this category included those involved in photosynthesis, the biosynthesis of pigments, energy metabolism, regulatory functions, translation, and transport. Interestingly, 190 genes within this category have not yet been assigned a specific function. The expression pattern of photosynthetic genes after the addition of Fe, present in category 1, is further represented in Figure 3B and demonstrated that photosynthesis genes were transcriptionally regulated by Fe (repressed in Fe deficiency and induced upon re-addition of Fe). The second category (n = 74) included genes involved in translational processes, especially genes coding for ribosomal proteins. Category 3 (n = 42) also included some of genes coding for ribosomal proteins and genes involved in transport process. Transcription of genes in categories 2 and 3 increased soon after Fe addition and then reached a plateau or decreased. Categories 4 and 5 demonstrated rather complex kinetics and consisted of many unknown genes. About 80% of the genes in category 4 (n = 26) were those assigned only hypothetical structures or functions at present. A majority of genes in category 5 (46/79) had no known function, although the category also included a few genes involved in regulatory functions or in photosynthesis. Category 6 (n = 202) included those genes (e.g. isiA, isiB, idiA, transport proteins, proteases, and regulatory proteins, such as sigma factors) whose transcript levels decreased rapidly after the addition of Fe.

View larger version (54K): [in this window] [in a new window]   Figure 3. A, Hierarchical cluster display of 866 genes that were used for the functional analysis of the iron-responsive differential expression changes. The expression ratio was calculated relative to time point 0 h (i.e. 0/3, 0/12, and 0/24) and were converted to log scale. These values were then used to cluster genes using Spotfire Decisionsite version 7.0 (Somerville, MA). Euclidean distance was used as measure of similarity between profiles. The various profiles were clustered by using the weighted pair group method with arithmetic mean. Change in transcript levels relative to LoFe: green (down) and red (up). The identification of all genes in this diagram is provided in Supplemental Data Table II (http://www.plantphysiol.org). B, Hierarchical cluster display of genes involved in photosynthetic processes that exhibited fold change of >1.25 and P value of <0.05. C, Schematic representation of various patterns of differentially expressed genes in response to iron availability. The various patterns were generated based on the clustering of genes (860/866) in (A) and numbered accordingly from top to bottom. The x axis represents the time-dependent expression of genes after the addition of Fe, whereas the y axis represents the transcript level on an arbitrary scale. The number of genes in each category is: 1 (n = 437), 2 (n = 74), 3 (n = 42), 4 (n = 26), 5 (n = 79), and 6 (n = 202).

Energy Metabolism

We were most interested in identifying differentially transcribed genes by functional category, and we will present data that are pertinent to fundamental cellular processes in cyanobacteria and plants. A relatively small number of genes involved with basic energy metabolism or central intermediary metabolism demonstrated significant changes, and the net effect of this regulation was to lower the breakdown of Glc and promote the storage of carbohydrates in the form of glycogen. A key feature was the down-regulation in Fe-deficient conditions of three genes in the heart of the glycolysis pathway: phosphofructokinase (pfkA; sll1196; P = 0.003), Glc-6-phosphate isomerase (pgi; slr1349; P = 0.025), and Fru-bisphosphate aldolase (fda; slr0943; P = 7 x 10^–7). In addition, Suc phosphate synthetase (sps; sll0045; P = 4 x 10^–11) transcription was strongly depressed in Fe-deficient conditions, and this may also help route sugars toward glycogen accumulation in these cells (Sherman and Sherman, 1983). Transcription for UDP-Glc dehydrogenase (slr1299; P = 0.006) also decreased, but there were only minor changes in the glgA and glgC genes (encoding proteins involved in glycogen synthesis). There was also enhanced transcription of 6-phosphogluconate dehydrogenase (gnd; sll0329), and this should lead to enhanced activity of the oxidative branch of the pentose phosphate pathway.

The isiAB Region

The induction of the isiA gene is the signature change in cyanobacteria grown in Fe-deficient conditions, and isiA had the largest fold change (22-fold) and a P value of 6 x 10^–10. We identified a set of five genes starting from sll0247 that showed enhanced transcription in the Fe-deficient state relative to Fe-sufficient conditions (Fig. 4). The fold changes for all of the genes in this cluster were substantial, and the P values were all less than 4 x 10^–5. These genes code for proteins of many different functions, including a Chl protein (sll0247; 22-fold, P = 6 x 10^–10), a flavodoxin (sll0248; 14-fold, P = 8 x 10^–11), a putative pantothenate metabolism flavoprotein (sll0250; 2.4-fold, P = 4 x 10^–5), and two genes with no functional designation (sll0249; 14-fold, P = 9 x 10^–10; and ssl0461; 2.2-fold, P = 0.004). Analysis of several cyanobacterial genomes showed that isiA is not contiguous to isiB in all cases, whereas the other four genes are found in two clusters (the homologs of sll0248 and sll0249 are contiguous, as are the homologs of ssl0461 and sll0250; data not shown).

View larger version (14K): [in this window] [in a new window]   Figure 4. The isiA region of the genome in which gene expression was enhanced in LoFe cells. The arrows indicate those five genes that demonstrated an increase in transcription in LoFe cells, and the size of the boxes corresponds to the lengths of the open reading frames.

Pigment Biosynthesis

One of the more striking transcriptional patterns involved the Chl biosynthetic pathway (Table III). None of the genes in the first one-third of the pathway (gltX to hemE) demonstrated statistically significant or interesting changes in transcription. However, the genes coding for six of the next seven enzymatic reactions along the main pathway were down-regulated approximately 1.4- to 6.3-fold in Fe-deficient conditions relative to Fe sufficiency (Table III). Protoporphyrin IX is the common branch point for the synthesis of Chl and heme and is converted to heme by ferrochelatase (plus Fe). Interestingly, transcription of the ferrochelatase gene (hemH, slr0839; P = 0.08) changed very little in response to changes in Fe levels. Finally, the genes coding for the last two enzymes in the pathway were up-regulated in Fe-deficient conditions: protochlorophyllide oxidoreductase (pcr; slr0506 P = 0.014) and Chl synthetase (chlG; slr0056 P = 0.01). In addition, a putative gene involved in bilin synthesis, heme oxygenase (ho1, sll1184 P = 0.02) was down-regulated in Fe-deficient conditions.

Photosynthesis Genes

The photosynthesis genes represent a complicated series of adaptations to the Fe-deficient state (Table II; Fig. 3, A and B). Many of the genes encoding the PSII structural proteins were transcribed at high levels under all conditions and were not transcriptionally regulated by Fe; e.g. psbA2, psbA3, both psbD genes, and psbC. The most significant transcriptional change in PSII was the decline in psbB of approximately 2.4-fold under Fe-deficient conditions (P = 3 x 10^–6). Similarly, all three lumenal proteins that are involved with the regulation of O₂ evolution (psbO, psbU, and psbV) decreased approximately 2-fold in LoFe (P values = 2 x 10^–5, 2 x 10^–4, and 2 x 10^–3, respectively). This fact may be of great importance for the assembly/disassembly of PSII. Transcription of psbH and psbI declined rather significantly under Fe-deficient conditions, with psbH some 3-fold lower relative to Fe-sufficient cells (P = 4 x 10^–6) and psbI some 2-fold lower (P = 3 x 10^–6). PSI demonstrated a somewhat different type of adaptation, and all statistically interesting genes had lower transcript levels in LoFe (Table II). The reaction center genes (psaAB) were transcribed at quite high levels in the Fe-deficient state but rose to even higher levels (1.6–2.0-fold increases; P = 2 x 10^–4 and 0.015, respectively) during normal growth. Subunits that demonstrated some downward Fe regulation in the Fe-deficient state were psaC (2.2-fold, P = 6 x 10^–4), psaE (1.7-fold, P = 5 x 10^–5), psaI (1.5-fold, P = 2 x 10^–4), psaJ (1.7-fold, P = 0.018), psaK (1.4-fold, P = 0.013), and psaL (1.8-fold, P = 4 x 10^–5). The psaC subunit includes a [4Fe-4S] Fe-sulfur cluster, so this 2-fold drop in transcription was understandable. Other photosynthesis complexes were also affected by Fe deficiency. The main ATPase operon, sll1322-sll1328, was down-regulated about 2-fold under Fe-deficient conditions (Table II), although statistically significant or interesting results were not obtained for the operon encoding the - and -subunits. The regulation of the cytochrome (cyt) b₆/f complex was both simple and surprising. The petD gene was up-regulated some 2.5-fold in the Fe-deficient state (P = 1 x 10^–5), whereas the transcription of the other genes encoding the cyt b₆, cyt f, and Rieske Fe-S proteins did not change appreciably. The petG and petM genes, on the other hand, were down-regulated some 1.6- to 2.0-fold in Fe deficiency. Many of the genes encoding subunits of the NADH dehydrogenase complex were transcriptionally down-regulated, whereas three of the genes encoding components of the cyt c oxidase underwent a series of changes (data not shown). Transcription of these genes was high in the Fe-deficient state, decreased soon after the addition of Fe, but then surpassed the levels of gene expression in the Fe-deficient state 24 h after Fe addition.

The cell employed different regulatory strategies for the soluble proteins or complexes (Table II; Fig. 3B). The soluble carriers responded as expected—the flavodoxin (isiB) gene was strongly induced in Fe-deficient conditions, whereas most ferredoxins were down-regulated in the Fe-deficient state. An exception was sll0662 (P = 2 x 10^–4), annotated as a bacterial-type ferredoxin, for which the transcript levels increased 3- to 4-fold in the Fe-deficient state compared with the Fe-sufficient states. The phycobilisome genes acted in concert, and virtually all of them were repressed in the Fe-deficient states (Table II). All 12 of the genes encoding phycobilisome complex proteins were statistically significant or interesting and were down-regulated 1.5- to 3.0-fold, as expected from previous spectral data (Guikema and Sherman, 1983, 1984). Among the CO₂ fixation genes, the rbcLS genes were down-regulated 1.6- to 2.2-fold (P = 1 x 10^–6 and 1.6 x 10^–3, respectively), as were the genes encoding proteins involved in the CO₂-concentrating mechanism.

Transport Proteins

The interactions and complexities among cellular systems were never more apparent than within the transport proteins. As shown in Table I, a large number of transport genes were regulated in the Fe-deficient state, including many ABC transporters. The microarray data show that the following specific genes that encode putative transport proteins are up-regulated in the Fe-deficient state: Leu/Ile/Val uptake (livFH; sll0374 and slr1200), cobalt uptake (cbiMO; sll0383 and sll0385), and molybdate uptake (sll0738 and sll0739; Table II; see Supplemental Data Table II at http://www.plantphysiol.org). One putative Fe transport gene that demonstrated Fe regulation was futC (sll1878; 2.3-fold increase, P = 1 x 10^–7). Transcription for the ferrichrome-Fe receptor gene fhuA (sll1409; P = 7 x 10^–8) dropped sharply in the Fe-deficient state, whereas the periplasmic Fe-binding protein slr0513 increased more than 4-fold (P = 3 x 10^–7) in the Fe-deficient state. Interestingly, transcription of the slr0074 gene, an ABC transporter subunit that is related to chloroplast ycf24, also dropped significantly (P = 7 x 10^–8) in Fe deficiency.

Transcription and Translation

We observed transcriptional changes in all five Group 1 and Group 2 sigma factors, described in Cyanobase: Group 1, the primary -factor, slr0653 (2.3-fold, P = 0.0004); and Group 2, the nonessential -factors, sll2012 (2.4-fold, P = 7 x 10^–6); sll0184 (2.3-fold, P = 0.009); sll0306 (1.8-fold, P = 0.01); and sll1689 (–1.4-fold, P = 0.003). Changes in rpoAB transcript accumulation were also found (sll1184, P = 0.004; sll1787, P = 6 x 10^–6, respectively), but the magnitude of the changes was more modest (Table II).

One of the most obvious features of cells grown in the Fe-deficient state is the drop in protein synthesis and cell-doubling time (Guikema and Sherman, 1983, 1984). The corresponding effect on the transcription of the ribosomal proteins was clearly demonstrated in the microarray experiment. Forty-seven of the genes encoding ribosomal proteins had a drop in transcript levels in the Fe-deficient state of approximately 2-fold that were considered to be statistically significant or interesting (see Supplemental Table II at http://www.plantphysiol.org). This included most of the genes in the major cluster of ribosomal protein genes (sll1799, P = 4 x 10^–5 to sll1822, P = 0.01). We also noted that the transcript levels of proteins involved in various translation modification events were also affected by Fe levels (Table II). For example, transcript levels of peptidyl-prolyl cis-trans-isomerase declined significantly in LoFe cells (slr1251, 1.7-fold, P = 3 x 10^–5), whereas polypeptide deformylase increased steeply (slr1549, 2.2-fold, P = 0.007).

Other Regulatory Genes

A number of regulatory genes were differentially expressed during growth in different Fe concentrations. The largest change was in slr0593 (–6.25-fold, P = 2 x 10^–11), the putative cAMP protein kinase regulatory chain. This gene was repressed in the Fe-deficient state and accumulated to high transcript levels after Fe addition. Genes encoding phytochromes represent another important regulatory class that was strongly up-regulated in the Fe-deficient state. These genes included the two-component regulatory system slr0473/slr0474 in which slr0473 (1.4-fold, P = 0.03) is the phytochrome-like gene cph1, and slr0474 (1.9-fold, P = 0.04) is the response regulator rcp1 (closely related to cheY; Yeh et al., 1997; Garcia-Dominguez et al., 2000; Park et al., 2000b; Lamparter et al., 2001). A second putative phytochrome (termed cph2; sll0821; 2.4-fold, P = 5 x 10^–5) has been identified (Park et al., 2000a; Wilde et al., 2002), and this gene was also strongly up-regulated under Fe-deficient conditions (Table II).

Cellular Processes. Cell Division, Chaperones, and Proteases

The transcription of many of the genes involved in basic cellular processes, such as chaperones and proteases, was increased by 2- to 3-fold under Fe-deficient conditions relative to 24 h, and it often decreased rapidly after Fe addition. The proteases that were altered in the Fe-deficient state included three ftsH genes: slll463 (1.5-fold, P = 0.03), slr0228 (1.7-fold, P = 0.0006), and slr1604 (0.5-fold, P = 0.04). Other LoFe-induced proteases included subunits of the ClpP complex: clpC (sll0020, 3.2-fold, P = 0.0002), clpP3 (slr0165, 2.3-fold, P = 5 x 10^–6), clpP4,(slr0164, 1.8-fold, P = 0.0009), and clpX (sll0535, 1.9-fold, P = 0.001). Importantly, slr0008 (–1.7-fold, P = 0.006), the carboxyl-terminal-processing protease (ctpA), decreased significantly in LoFe, consistent with the decrease in PSII assembly.

Chaperones also demonstrated interesting transcriptional changes, including htpG (hsp90, sll0430, 1.7-fold, P = 0.003), dnaJ (slll666, 1.7-fold, P = 0.0002), dnaK (sll1932, – 1.4-fold, P = 0.008), groELS (slr2075, – 1.7-fold, P = 1 x 10^–4; and slr2076, – 1.7-fold, P = 0.03), and groEL-2 (sll0416, – 2.0-fold, P = 5 x 10^–7). In the case of the GroELS proteins, the transcript accumulation increased strongly immediately after the addition of Fe and then settled down to a steady-state level that was lower than under Fe-deficient conditions. Another protein with chaperone activity is slr0374 (2.0-fold, P = 0.003), a protein previously identified as Fe induced (Singh and Sherman, 2000). We have subsequently determined that the gene is part of an operon with slr0373 (3.0-fold, P = 8 x 10^–6), slr0374, and slr0376 (2.6-fold, P = 7 x 10^–6) and that the operon is induced under many stress conditions (Singh and Sherman, 2002).

	DISCUSSION

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 
In this study, we utilized DNA microarray technology to profile the genes that were differentially expressed during growth of Synechocystis sp. PCC 6803 in Fe-deficient versus Fe-reconstituted medium. The results paint a rather detailed picture of how this cyanobacterium responds to growth in Fe-deficient versus Fe-reconstituted medium. Transcription of the protein synthesis machinery was decreased substantially, transcription of genes encoding proteins involved in protein modification, assembly, or degradation was altered, and glycolysis genes were turned down in the Fe-deficient state. These changes resulted in a cell that was smaller, had fewer photosynthetic membranes, and utilized less energy. Glc utilization was slowed and glycogen synthesis enhanced—this leads to the accumulation of glycogen granules as was documented by electron microscopy (Sherman and Sherman, 1983). The results indicated that the cell responds immediately to the presence of Fe by synthesizing proteins needed for energy production and pigment biosynthesis. It then devotes a significant period to the assembly of the translational machinery, and these genes are eventually turned off. Those genes involved with cell maintenance under Fe-deficient conditions are efficiently repressed upon the presence of Fe.

When cells are grown under Fe-deficient conditions, a variety of physiological and morphological phenomena occur, the most obvious of which is a significant change in cellular pigmentation and the reorganization of photosynthetic complexes (Guikema and Sherman, 1983, 1984; Pakrasi et al., 1985a, 1985b; Laudenbach and Straus, 1988; Laudenbach et al., 1988; Straus, 1994). Data presented here showed that the major PSII genes were not the targets of transcriptional control. The transcription of the PSII reaction center genes remained high in the Fe-deficient state and changed little after the addition of Fe. Instead, significant changes occurred in genes, which suggested that posttranscriptional regulation played a significant role under these conditions. For example, psbB, psbH, psbI, and those coding for oxygen-evolving proteins that are located in the thylakoid lumen (psbO, psbU, and psbV) were down-regulated in the Fe-deficient state. It has been shown recently that a Synechocystis sp. PCC 6803 psbH mutant, which resulted in the absence of PsbH, destabilized PSII (Komenda et al., 2002). In addition, the 3.8-Å crystal structure of PSII indicates that both PsbH and PsbI are localized near the interface of the D2 and CP47 (PsbB) proteins (Zouni et al., 2001; Heathcote et al., 2002). Thus, the lowering of the transcript levels of psbB, psbH, and psbI is consistent with a tendency to disassemble or to destabilize PSII. These changes were also coordinated by the regulation of two key enzymes involved in PSII assembly processes. The growth of Synechocystis sp. PCC 6803 in the Fe-deficient conditions resulted in transcriptional increases for the ftsH (slr0228) gene. This gene has been suggested to be involved in D1 protein degradation in cyanobacteria and in Arabidopsis (Lindahl et al., 2000; Mann et al., 2000; Hauühl et al., 2001; Bailey et al., 2002). Similarly, transcription of the gene encoding ctpA, the carboxy-terminal protease involved in D1 protein processing, was reduced significantly in response to LoFe. These results were consistent with the cellular tendency to enhance PSII instability.

A similar regulation was also observed for PSI,and the genes that were regulated by Fe tend to destabilize PSI. For example, down-regulation under Fe-deficient conditions of PsaE, a stromal protein, and PsaK, an intrinsic PSI component, was important because these proteins are required for the stable assembly of PSI (Xu et al., 2001). Similarly, PsaI (important for the normal organization of the PsaL subunit that is required for PSI trimer formation in cyanobacteria) was also down-regulated (Fromme et al., 2001; Xu et al., 2001). In addition, the PSI psaAB reaction center genes were down-regulated some 2-fold under Fe-deficient conditions. All of these changes are aimed at destabilizing the PSI complex or at ensuring that fewer functional PSI complexes can be assembled, especially in their trimeric form. Thus, the cell was attempting to balance the stoichiometry among the different complexes; the net result is a PSI:PSII ratio that is much closer to 1:1 than the 3:1 found in cells grown under Fe sufficiency (Riethman et al., 1988).

When cells were transferred to Fe-deficient conditions, Chl and phycobilin content declined and remained at a basal level sufficient for a limited photosynthetic apparatus. Genes encoding phycobilisome proteins were reduced to an appropriate lower level commensurate with the presence of a lower level of phycobilin pigments. However, transcript levels of genes encoding proteins involved in pigment biosynthesis showed interesting differences. Intermediates needed for heme biosynthesis were maintained, whereas intermediates after protoporphyrin IX were decreased. However, any of the later intermediates can be rapidly converted into Chl due to an increase in enzyme availability of the oxidoreductase and the Chl synthetase. This pattern may suggest a fine-tuning of the metabolic intermediates to ensure that Chl levels are commensurate with the Chl-binding proteins that are synthesized.

Genes encoding phytochromes represent an important regulatory class that was strongly up-regulated in the Fe-deficient state. Both cph1 and cph2 were also dark induced and light repressed (Park et al., 2000a). Although it has been postulated that Cph1 (and also Cph2) are involved in the regulation of events during the light/dark or dark/light transitions, no strong phenotype was obtained with a cph1 knockout mutant (Garcia-Dominguez et al., 2000). However, a clue to function came from an experiment in which a cph2 mutant and a cph1/cph2 double mutant responded differently than wild type and moved toward a blue light source. One interpretation of this result was that Cph2 is needed for an inhibition of positive mobility toward blue light. Our results suggested that the putative phytochromes might have a function during the cellular response to Fe stress. We will analyze mutants in these genes to determine how the mutants grow and respond under LoFe conditions.

A large number of transport genes, including many ABC transporters, were regulated by Fe. Such results were in agreement with the idea that Fe-deficient cells do everything possible to bring ions and other metabolites into the cell to develop a modified metabolic state and that the Fe inducibility of various transport genes resulted from a strong cellular response to nutrient limitation. A particular interesting Fe-regulated protein is IdiA (Michel et al., 1996, 1999, 2001). The protein (now ascribed to slr1295 and not slr0513) is thought to have a protective role for the acceptor side of PSII during Fe stress, and it has been shown to be located at the thylakoid lumen in Synechocystis sp. PCC 6803 (Tölle et al., 2002). Quite interestingly, we found that slr1295 was not significantly Fe regulated (P = 0.167), whereas slr0513 was one of the genes with the largest, and most significant, fold increase (approximately 4.5-fold, P = 3 x 10^–7) between the Fe-deficient state and 24 h after Fe addition. The transcription of slr1295 also was shown to be cold repressible by some 3-fold (Suzuki et al., 2001) and light inducible (Gill et al., 2002). Thus, this protein may be most important for oxidative protection in Synechocystis sp. PCC 6803 due to environmental factors in addition to Fe starvation.

A plethora of genes that currently have no known functional designation were regulated by Fe availability. Such findings provide the impetus for future microarray experiments with cells grown under different environmental conditions and with the many knockout mutants that we can produce. A first start at functional analysis of these genes has come from cluster analysis of the genes that were up- or down-regulated under specific conditions. For example, slr0374 was identified as a Fe-responsive gene in our previous work (Singh and Sherman, 2000); then, we determined that slr0374 and the neighboring genes (slr0373, slr0374, and slr0376) represented an operon (Singh and Sherman, 2002). All three genes are strongly transcribed under LoFe conditions compared with Fe-sufficient conditions and appeared to be involved in the periplasm in some type of chaperone activity of membrane protein assembly (A.K. Singh and L.A. Sherman, unpublished data). Thus, the microarray data identified genes that represent the focus of more intensive research with genetics and physiology to determine gene function, their chromosomal organization, and their mode of regulation. We will continue such efforts, with an emphasis on the assembly/disassembly of the photosynthetic complexes and especially on the interaction of the novel IsiA protein with both PSII and PSI.

	MATERIALS AND METHODS

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

Strain and Growth Conditions

Glassware used in LoFe medium preparation was treated with EDTA, and the LoFe BG-11 medium was made as follows. Ferric ammonium citrate present in normal BG-11 medium was replaced with ammonium citrate for the LoFe medium, and four of the BG-11 stock solutions (NaNO₃, ammonium citrate, K₂HPO₄, and Na₂CO₃) were passed through Chelex-100 (Bio-Rad Laboratories, Hercules, CA) columns to eliminate trace amounts of Fe. Synechocystis sp. strain PCC 6803 cells were subcultured in LoFe media at least 6 d before experimental use. Cells were grown phototrophically in LoFe medium at 30°C under a light intensity of 20 to 30 µE m^–2 s^–1. The culture was bubbled vigorously by air. Recovery of Fe-deficient cultures was accomplished by addition of 6 mg of ferric ammonium citrate per liter of medium, a concentration equal to that present in the normal BG-11 medium. Cells were removed during recovery at 0 h (iron-deficient culture) or at 3, 12, and 24 h after the addition of iron (reconstituting cultures) and harvested by centrifugation at 4,000g in a refrigerated centrifuge. Cells were either frozen and stored at –80°C or immediately used for RNA isolation.

Construction of DNA Array

The complete description of array construction has been described elsewhere (Postier et al., 2003). In brief, we utilized a two-stage PCR process to amplify 3,165 genes identified on the Synechocystis sp. PCC 6803 genome on the Kazusa Web site before May 2002. The first stage of PCR was based on a bipartite primer that contained both gene-specific and universal sequences to amplify the individual genes using genomic DNA as template. The gene-specific sequences were of variable length to allow uniform annealing temperatures, and longer genes (230 of 3,165) were truncated at the 3' end to 2 kb. The second stage PCR involved the amplification of first stage products with primer corresponding to the universal sequence. PCRs were performed using Platinum Pfx polymerase (Invitrogen, Carlsbad, CA) in 96-well plates using a 100-µL reaction mix containing 1x enhancer; 0.15 mM dATP, dGTP, dCTP, and dCTP; 1 mM MgSO₄; 1 µM primer; and 2 units of Platinum Pfx polymerase in 1x reaction buffer. After each amplification, 5 µL of PCR products was run in 1.2% (w/v) agarose gel to assess the quality of desired products. After successful amplification of all 3,165 genes, each plate of PCR products was purified using Multiscreen-PCR plates (Millipore, Bedford, MA) following the manufacturer's instructions. The DNA was eluted in 50 µL of water and transferred into UV-transparent 384-well plates (Corning Incorporated, Corning, NY). The yield of each purified product was measured with a Spectramax 384 plus spectrophotometer (Molecular Devices Corporation, Sunnyvale, CA). These products were then dried in a Speed Vac (Savant Instruments, Holbrook, NY). Using a Bio-robot 3000 (Qiagen, Valencia, CA), we first resuspended each product in water to give a concentration of 1 µg µL^–1, and then an equal volume of 2x microspotting solution (Telechem International, Sunnyvale, CA) was added. Ten microliters of each product was transferred to printing source plates (384-well plates, Genetix Limited, Hampshire, UK). PCR products were printed on superamine slides (Telechem International) with 16-microquill 2000 pins (Majer Precision, Tempe, AZ) using an Omnigrid arrayer (Gene Machines, San Carlos, CA). All PCR products were printed in triplicate on each slide. Many authors have commented on the importance of sufficient replicates in the analysis of microarray data, and the technical replicates were valuable in our statistical analysis (Arfin et al., 2000; Kerr and Churchill, 2001a, 2001b; Long et al., 2001; Wolfinger et al., 2001; Yang and Speed, 2002). The diameter of each spot was about 125 µm with a spacing of 250 µm between center to center of any given two spots. After printing was completed, slides were baked at 80°C for 2 h and stored in the dark at room temperature.

RNA Isolation and RNA Gel-Blot Analysis

Total RNA from Synechocystis sp. strain PCC 6803 was isolated using the procedure described by Reddy et al. (1990) with modifications as in Colon-Lopez et al. (1997). RNA was isolated from cells that were harvested from cultures collected after LoFe growth and 3, 12, and 24 h after the addition of iron to the iron-deficient cells. At each time point, cells were mixed with 1/20 volume of stop solution (200 mM Tris-HCl [pH 8.0], 20 mM EDTA, and 20 mM sodium azide), pelleted, and stored at –80°C.

Preparation of Fluorescently Labeled Probes

Fluorescently labeled cDNA probes were prepared from the total RNA by reverse transcription of total RNA in the presence of aminoallyl-dUTP followed by coupling either with Cy3 or Cy5 monofunctional dye (Amersham Pharmacia Biotech, Piscataway, NJ). The importance of random hexamer priming for bacterial RNA was discussed by Arfin et al. (2000), and we have optimized the priming of total RNA by using a mixture of hexamers, octamers, nanomers, and decamers. The reverse transcription reaction was performed in a 50-µL volume containing 10 µg of total RNA; 5 µg of random primers mix (containing 6:2:1:1 [w/v] hexamer:octamer:nanomer:decamer); 10 mM dithiothreitol; 0.5 mM each of dATP, dCTP, and dGTP; 0.3 mM dTTP; 0.2 mM aadUTP; and 400 units of SuperscriptII reverse transcriptase (Invitrogen) in 1x reaction buffer. RNA and primers were heated at 65°C for 5 min and chilled on ice before the remaining reaction components were added. The reverse transcription was performed for 2 h at 44°C. After reverse transcription, the RNA was degraded by incubating at 65°C for 15 min after the addition of 17 µL of 0.5 M EDTA (pH 8.0) and 17 µL of 1 N NaOH. The reaction was neutralized by the addition of 34 µL of 1 M Tris (pH 7.5). Four hundred microliters of water was added to the neutralized samples, and the mix was placed in a Microcon YM-10 microconcentrator (Millipore) and spun for 30 min in a bench-top microcentrifuge at 11,000g. The cDNA retained by the Microcon was washed twice with 400 µL of water after centrifugation at 11,000g. After the last wash, 50 µL of water was added in Microcon tubes, mixed properly by pipetting, and cDNA was transferred in a new tube. To maximize the recovery of cDNA, a second washing with 50 µL of water was also carried out. Both the eluates were combined and dried in Speed Vac. The aminoallyl labeled cDNAs were resuspended in 10 µL of water and then 10 µL of 0.2 M NaHCO₃ (pH 9.0) was added. This sample was transferred in a tube containing either Cy3 or Cy5 monofunctional dye and incubated for 1 h at room temperature in the dark. The uncoupled dyes were quenched by the addition of 4.5 µL of 4 M hydroxylamine after incubation for 15 min at room temperature in the dark. The unincorporated dyes from Cy3 and Cy5 reactions were separately removed using a Qia-Quick PCR purification column according to the manufacturer's instructions (Qiagen). The column eluates were combined, dried in a Speed Vac and resuspended in 40 µL of TE (10 mM Tris [pH 8.0] and 1 mM EDTA).

Hybridization, Washing, and Scanning

Before hybridization, the slides were washed once in 0.2% (w/v) SDS and twice in water for 5 min each at room temperature. Thereafter, slides were transferred in hot water for 5 min and washed in 0.2% (w/v) SDS, followed by two washes in water for 5 min each. The slides were spun dried and prehybridized in a mixture of 25% (w/v) formamide, 5x SSC, 0.1% (w/v) SDS, and 1% (w/v) bovine serum albumin for 45 min at 42°C in a CLONTECH hybridization chamber (CLONTECH, Palo Alto, CA). The slide was briefly rinsed with water and spun dried. Hybridization was carried in a total volume of 80 µL consisting of 25% (w/v) formamide, 5x SSC, 0.1% (w/v) SDS, 1% (w/v) bovine serum albumin, 0.1 mg of salmon sperm DNA, and Cy3- and Cy5-labled probes. The labeled cDNA in hybridization buffer was heated at 95°C for 2 min and quickly transferred to an oven maintained at 42°C. The slide was placed in a CMT hybridization chamber (Corning) and transferred to the oven at 42°C. After 10 min of incubation, hybridization solution containing labeled probes was placed on the slide and covered by a coverslip. The whole assembly was placed in a water bath maintained at 42°C. After 18 to 20 h of hybridization, slides were washed in 2x SSC and 0.1% (w/v) SDS, which was preheated at 42°C. After 5 min of incubation, the slides were further washed in 0.1x SSC and 0.1% (w/v) SDS for 10 min at room temperature. Finally, slides were rinsed in 0.1x SSC and then briefly in water. The slides were spun dried and immediately scanned. The scanning was performed with a Scanarray 4000 scanner (Packard BioChip Technologies, Billerica, MA) for Cy3 (532 nm) and Cy5 (635 nm) at a resolution of 10 µm per pixel generating two separate TIFF images. Images were often acquired at various laser and photomultipler tube settings.

Experimental Design

The effect of iron deficiency on gene expression in Synechocystis sp. PCC 6803 was studied with cultures grown in iron-deficient medium for 10 generations (0 h). Iron was then added back to the normal levels found in the BG-11 medium, and RNA was isolated at three subsequent time points (3, 12, and 24 h) because these were times of important physiological events (Riethman et al., 1988). We used a loop design that included a "dye swap" such that RNA from each time point was labeled with both Cy3 and Cy5 and was used for hybridization (Fig. 1A; Kerr and Churchill, 2001a, 2001b; Churchill, 2002; Oleksiak et al., 2002; Yang and Speed, 2002). Most importantly, this design ensures that the iron effect is not completely confounded with other potential sources of variation. It should be noted that the three replicate slides for each time point and the three replicate printings per slide represent technical variation. Biological variation was sampled by extracting RNA from three separate experiments and pooling them before all of the hybridizations.

Data Acquisition and Statistical Analysis

Spot intensities of the images were quantified using Quantarray 3.0 (Packard BioChip Technologies). A predefined grid containing a defined circle fitting the size of spots was placed on each image and manually adjusted to ensure optimal spot recognition. Spots were individually quantified using the adaptive method, and the mean intensities corresponding to each spot were exported into a separate Excel spreadsheet for each array. Data for the six slides in this experiment were then uploaded into SAS. Testing has demonstrated that Quantarray is very reliable and similar to results from Imagene 6.0 (Moody et al., 2002).

An ANOVA modeling approach was used to analyze the microarray data (Kerr and Churchill, 2001a, 2001b; Wolfinger et al., 2001; Churchill, 2002; Oleksiak et al., 2002; Wayne and McIntyre, 2002; Drenth et al., 2003). The model Y_ijklm = µ + t_i + d_j + _k + _k(_l) + _ijklm was fit, where Y is the intensity of the spot after correction for the local background signal, normalization, and log transformation. The parameter µ was the overall mean of the normalized values for that gene. Fixed effects for time (t) and dye (d) were fit, but the interaction was omitted because it was confounded with the effect of slide. The random effects of slide () and of replicates within slide [_k(_l)] were included. To test the null hypothesis that a particular gene's expression level was not different over time, an F test of the effect of time for each gene was conducted, and a P value was calculated. We examined the model for conformation to the assumption of normality of the residuals by testing the null hypothesis that the residuals for each gene were normally distributed using the Shapiro-Wilkes Test. Additional contrasts were examined. The effect at time 0 (the iron-deficient state) was compared separately with 24 h after iron addition, time 0 was compared with time 3, time 3 was compared with time 12, and time 12 was compared with time 24. All analyses were performed in SAS (SAS Institute, Inc, Cary, NC).

We used a Bonferroni significance level of 0.05/3,165 or 0.000015797788 as a criterion for rejecting the null hypothesis of a significant time effect. Because type I and type II errors are inversely related, with decreases in false positives (type I) being associated with increases in false negatives (type II), and because the Bonferroni correction will be overly conservative as tests are correlated (Westfall and Young, 1993; Doerge and Churchill, 1996; McIntyre et al., 2000), we used a second overly liberal threshold of 0.05. In addition, we considered the test for dye effects and normality of the residuals. If the test of the null hypothesis of difference across times was rejected at 0.0000158 and we had no evidence for dye effects or departure from normality of the residuals, we declared the gene differentially expressed over time and concluded that the changes in gene expression were statistically significant (these 85 genes are highlighted in red in Supplemental Data Table I, http://www.plantphysiol.org). If the P value for the test of differences over time was less than or equal to 0.05 but larger than 0.0000158 and we had no evidence for dye effects or departure from normality of the residuals, we considered the gene interesting (these 781 genes are highlighted in yellow in Supplemental Data Table I, http://www.plantphysiol.org). When dye effects were present or residuals showed evidence of departure from normality, we advise caution in the interpretation of the results. Genes that show a P value for the test of differential expression that exceeds the Bonferroni criteria or the liberal threshold but that have significant dye effects are indicated as dye-red or dye-yellow (Supplemental Data Table I, http://www.plantphysiol.org); genes that depart from normality are highlighted in green (Bonferroni) or blue (0.05; Supplemental Data Table I, http://www.plantphysiol.org). Once the analysis was completed, we examined the results carefully, and we focused our discussion on statistically significant and interesting genes that exhibited a fold change of at least 1.25x. This included the 85 statistically significant and 731 statistically interesting genes. A total of 10 genes with possible dye effects were included in the functional analysis (six significant and four interesting) based on independent northern-blot confirmation or cluster analysis where other genes in a transcriptional cluster were found to be significant or interesting. Genes that departed from normality of the residuals, but showed some evidence of differential expression, were individually analyzed, and 40 were included in the functional analysis. In all of these cases, the departure from normality was due to extreme differences in the response in the iron-deficient state (time 0) compared with the other three time points and the resulting heteroscedasticity of the error. There were a total of 866 genes examined with a P < 0.05. Our objective is to identify genes that demonstrate differential expression for further experimentation. Thus, we bracket our interpretation of the results with a conservative (Bonferroni) threshold and a liberal 0.05 criterion. The raw P values are available for inspection by other investigators in Supplemental Data Table I (http://www.plantphysiol.org) so that individuals may examine the evidence themselves.

Independent Validation of Microarray Results with Northern Blots

Microarray experiments provide information on the expression profiling of thousands of genes, and it is critical to have an independent measure for at least a subset of the results. Several factors such as contamination with other genes, dust or scratches on the cDNA spots, and high background can lead to false profiling. In the present study, we have utilized technical replicates on each array, multiple arrays, and a statistical analysis to identify potential problems, and then used northern blots to validate the results. In one experiment, we selected nine genes at random from among the unknown category plus a gene (sll0249) in the isiA region to compare expression patterns as obtained from northern blots versus those from the microarray. In general, there was an excellent qualitative correspondence between the two techniques, although there were some quantitative differences (data not shown).

Distribution of Materials

Upon request, all novel material described in this publication will be made available in a timely manner for noncommercial research purposes, subject to the requisite permission from any third party owners of all or parts of the material. Obtaining any permission will be the responsibility of the requestor.

	ACKNOWLEDGMENTS

 
We would like to thank Hong Li for many discussions during the course of this work and Elsie Grace for her efforts on the data management. We would also like to thank Dr. Rob Burnap and his lab (especially Brad Postier) for the very fruitful collaboration that led to the construction of this microarray.

Received March 21, 2003; returned for revision April 21, 2003; accepted May 12, 2003.

	FOOTNOTES

 
Article, publication date, and citation information can be found at www.plantphysiol.org/cgi/doi/10.1104/pp.103.024018.

¹ This work was supported by the National Science Foundation (grant no. MCB–0084457 to Robert Burnap and L.A.S. for the construction of the microarray), by the Department of Energy (grant no. DE–FG92–99ER20342 to L.A.S.), and by the U.S. Department of Agriculture-Initiative for Future Agriculture and Food Systems (grant no. NOO14–94-1–0318 to L.M.M.).

^[w] The online version of this article contains Web-only data. The supplemental material is available at http://www.plantphysiol.org.

^* Corresponding author; e-mail lsherman{at}bilbo.bio.purdue.edu; fax 765–496–1496.

	LITERATURE CITED

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 
Arfin SM, Long AD, Ito ET, Tolleri L, Riehle MM, Paegle ES, Hatfield GW (2000) Global gene expression profiling in Escherichia coli K12. J Biol Chem 275: 29672–29684[Abstract/Free Full Text]

Bailey S, Thompson E, Nixon PJ, Horton P, Mullineaux CW, Robinson C, Mann NH (2002) A critical role for the Var2 FtsH homologue of Arabidopsis thaliana in the photosystem II repair cycle in vivo. J Biol Chem 277: 2006–2011[Abstract/Free Full Text]

Behrenfeld MJ, Kolber ZS (1999) Widespread iron limitation of phyto-plankton in the south pacific ocean. Science 283: 840–843[Medline]

Bibby TS, Nield J, Barber J (2001) Iron deficiency induces the formation of an antenna ring around trimeric photosystem I in cyanobacteria. Nature 412: 743–745[CrossRef][Medline]

Boekema EJ, Hlfney A, Yakushevska AE, Plotrowski M, Keegstra K, Berry S, Michel K-P, Pistorius EK, Krulp J (2001) A giant chlorophyll-protein complex induced by iron deficiency in cyanobacteria. Nature 412: 745–748[CrossRef][Medline]

Burnap RL, Troyan T, Sherman LA (1993) The highly abundant chlorophyll-protein complex of iron-deficient Synechococcus sp PCC7942 (CP43') is encoded by the isiA gene. Plant Physiol 103: 893–902[Abstract]

Churchill GA (2002) Fundamentals of experimental design for cDNA microarrays. Nat Genet 32: 490–495[CrossRef][ISI][Medline]

Colon-Lopez MS, Sherman DM, Sherman LA (1997) Transcriptional and translational regulation of nitrogenase in light-dark- and continuous-light-grown cultures of the unicellular cyanobacterium Cyanothece sp strain ATCC 51142. J Bacteriol 179: 4319–4327[Abstract/Free Full Text]

Doerge RW, Churchill GA (1996) Permutation tests for multiple loci affecting a quantitative character. Genetics 142: 285–294[Abstract]

Drenth JPH, te Morische RHM, Smink R, Bonifacino JS, Jansen JBMJ (2003) Germline mutations in PRKCSH are associated with autosomal dominant polycystic liver disease. Nat Genet 33: 1–3[CrossRef][ISI][Medline]

Earhart C (1996) Uptake and metabolism of iron and molybdenum. In FC Neidhardt, ed Escherichia coli and Salmonella. ASM Press, Washington, DC, pp 1075–1090

Fox TC, Guerinot ML (1998) Molecular biology of cation transport in plants. Annu Rev Plant Physiol Plant Mol Biol 49: 669–696[CrossRef][Medline]

Fromme P, Jordan P, Kraus N (2001) Structure of PSI. Biochim Biophys Acta 1507: 5–31[Medline]

Garcia-Dominguez M, Muro-Pastor MI, Reyes JC, Florencio FJ (2000) Light-dependent regulation of cyanobacterial phytochrome expression. J Bacteriol 182: 38–44[Abstract/Free Full Text]

Gill R, Katsoulakis E, Schmitt W, Taroncher-Oldenburg G, Misra J, Stephanopoulos G (2002) Genome-wide dynamic transcriptional profiling of the light-to-dark transition in Synechocystis sp strain PCC 6803. J Bacteriol 184: 3671–3681[Abstract/Free Full Text]

Guikema JA, Sherman LA (1983) Organization and function of chlorophyll in membranes of cyanobacteria during iron starvation. Plant Physiol 73: 250–256[Abstract/Free Full Text]

<