CO(2)-cAMP-responsive cis-elements targeted by a transcription factor with CREB/ATF-like basic zipper domain in the marine diatom Phaeodactylum tricornutum.

Expression controls of the carbon acquisition system in marine diatoms in response to environmental factors are an essential issue to understand the changes in marine primary productivity. A pyrenoidal β-carbonic anhydrase, PtCA1, is one of the most important candidates to investigate the control mechanisms of the CO(2) acquisition system in the marine diatom Phaeodactylum tricornutum. A detailed functional assay was carried out on the putative core regulatory region of the ptca1 promoter using a β-glucuronidase reporter in P. tricornutum cells under changing CO(2) conditions. A set of loss-of-function assays led to the identification of three CO(2)-responsive elements, TGACGT, ACGTCA, and TGACGC, at a region -86 to -42 relative to the transcription start site. Treatment with a cyclic (c)AMP analog, dibutyryl cAMP, revealed these three elements to be under the control of cAMP; thus, we designated them, from 5' to 3', as CO(2)-cAMP-Responsive Element1 (CCRE1), CCRE2, and CCRE3. Because the sequence TGACGT is known to be a typical target of human Activating Transcription Factor6 (ATF6), we searched for genes containing a basic zipper (bZIP) region homologous to that of ATF6 in the genome of P. tricornutum. Gel-shift assays using CCRE pentamers as labeled probes showed that at least one candidate of bZIP proteins, PtbZIP11, bound specifically to CCREs. A series of gain-of-function assays with CCREs fused to a minimal promoter strongly suggested that the alternative combination of CCRE1/2 or CCRE2/3 at proper distances from the minimal promoter is required as a potential target of PtbZIP11 for an effective CO(2) response of the ptca1 gene.

Marine diatoms are the most successful group of algae in the ocean and are responsible for about onefifth of global carbon fixation (Tréguer et al., 1995;Falkowski et al., 2000). The maximum CO 2 concentration dissolved in seawater under the present atmosphere at 20°C is calculated to be below 15 mM, which is far lower than the K m [CO 2 ] of the carbon fixation enzyme Rubisco in diatoms (Badger et al., 1998). This implies that CO 2 in seawater would not support sufficient primary production; instead, an alternative inorganic carbon source is additionally required for it. It has been well established that many microalgae including marine diatoms possess CO 2 -concentrating mechanisms (CCMs), which actively take up both CO 2 and HCO 3 2 from the surrounding media and efficiently supply CO 2 from accumulated HCO 3 2 to Rubisco (Kaplan et al., 1980;Colman and Rotatore, 1995;Matsuda and Colman, 1995;Johnston and Raven, 1996;Matsuda et al., 2001;Moroney and Ynalvez, 2007;Spalding, 2008;Trimborn et al., 2008). Besides the CCMs, it has been reported in one species of marine diatom that a C 4 -like metabolism also operates in the CO 2 acquisition system (Reinfelder et al., 2004). However, general occurrences of a C 4 -type system in diatoms are still in question (Roberts et al., 2007;Matsuda et al., 2011).
The activity of CCMs is suppressed under high-CO 2 conditions and induced (or derepressed) under normal air (and/or further lower CO 2 ) conditions in a few hours (Kaplan et al., 1980;Miller et al., 1990;Colman and Rotatore, 1995;Matsuda and Colman, 1995;Johnston and Raven, 1996;Matsuda et al., 2001;Kucho et al., 2003;Vance and Spalding, 2005). This indicates that algal cells can perceive changes in ambient CO 2 concentration as a signal and acclimate to new CO 2 conditions in a short period. Mechanisms of acclimation to CO 2 and its relationship with other abiotic signals in major oceanic primary producers such as diatoms are a critical matter in order to understand the fate of aquatic primary productivity. Notwithstanding, molecular level studies of them are still in their infancy.
On the other hand, knowledge about the regulations of putative CCM components is relatively limited in eukaryotic algae. In the green alga Chlamydomonas reinhardtii, it is known that the transcriptional regulation of a periplasmic carbonic anhydrase (CA), Cah1 (Fujiwara et al., 1990), and mitochondrial CA, mtCA, are highly responsive to ambient CO 2 concentration (Eriksson et al., 1998). The transcriptional activity of a promoter region of Cah1 is known to be under bidirectional controls in response to CO 2 , being ensured by enhancer and silencer regions, which reside back to back on the Cah1 promoter. This enhancer region is further found to possess two critical elements designated as enhancer-element consensus (Kucho et al., 2003), which is the target of a class of transcription factors, enhancerelement consensus-binding proteins (Yoshioka et al., 2004). Although the upstream signal cascade toward this promoter is not known in C. reinhardtii, a zinc finger protein, CCM1/Cia5, has been found to govern the upregulation of most low-CO 2 -inducible genes and thus is considered to be a master regulator of the transcriptional response (Fukuzawa et al., 2001;Xiang et al., 2001).
The CCM and controls on expression of the genes involved in response to changes in ambient [CO 2 ] are also observed in many marine eukaryotic photoautotrophs (Johnston and Raven, 1996;Badger et al., 1998;Sü ltemeyer 1998;Thoms et al., 2001;Matsuda et al., 2001;Colman et al., 2002). In marine diatoms, physiological evidence has shown that active uptake systems for CO 2 and HCO 3 2 operate in both centric and pennate species (Rotatore and Colman, 1992;Burkhardt et al., 2001;Matsuda et al., 2001Matsuda et al., , 2011Chen et al., 2006;Trimborn et al., 2008). Since both the capacity to take up dissolved inorganic carbon (DIC) and a cell's affinity for DIC were shown to be induced upon transferring cells of the marine diatom Phaeodactylum tricornutum from high CO 2 to air , DIC uptake and internal CO 2 flow toward Rubisco seem to be strengthened under CO 2 limitation. As one of the putative key components for this acclimation process, chloroplastic CAs have been isolated and genes have been cloned (Satoh et al., 2001;.
Similarly to CCMs in green algae and cyanobacteria, CAs at close proximity to Rubisco must play a crucial role in marine diatoms to create an efficient flux of CO 2 to Rubisco under CO 2 -limiting conditions. Two b-type CAs, PtCA1 and PtCA2, found as chloroplast-localized enzymes in P. tricornutum, were localized in the pyrenoid, a large protein body in the stroma, and it is known that the C-terminal amphipathic helices of PtCAs play a role in the formation of a protein complex in the pyrenoid (Tanaka et al., 2005;Kitao and Matsuda, 2009;Tachibana et al., 2011). Transcription of the ptca1 and ptca2 genes is regulated by CO 2 , and the response of ptca1 to changes in [CO 2 ] has been shown to be stronger than that of ptca2 (Harada et al., , 2006. The function of PtCA1 has yet to be determined, but this enzyme is one of the most interesting candidates for elucidating the molecular mechanisms of the CCM and acclimation to CO 2 in diatoms. In previous studies, the promoter region of ptca1 (Pptca1) was isolated, and it was shown that the core regulatory region is composed of two putative cAMP-response elements (CRE1 and CRE2) and a putative p300-binding element (p300be; Harada et al., 2006); additionally, a putative SKiNhead-1 (Skn-1)-binding element was found downstream of the p300be (in this article). CRE and p300be were related to cAMP response elements in mammalian systems (Montminy et al., 1986;Deutsch et al., 1988;Rikitake and Moran 1992;Tanaka et al., 1997), and Skn-1 was activated by binding with cAMP-responsive element-binding protein 1 in the nematode Caenorhabditis elegans (Walker et al., 2000). It was indeed shown in P. tricornutum that transcription of ptca1 was repressed even under CO 2limiting conditions in the presence of the cAMP analog, dibutyryl cAMP (dbcAMP; Harada et al., 2006). This repressive effect of dbcAMP was negated by impairing the sequences around CREs and p300be (Harada et al., 2006), thus strongly suggesting that the core regulatory region of Pptca1 is under the repressive control of cAMP as a second messenger for CO 2 signaling, but critical cis-elements for this process have yet to be determined in the Pptca1 promoter.
The perception of CO 2 signals via cAMP has been reported in cyanobacteria, a pathogenic fungus, and mammals. The activity of soluble adenylyl cyclase (sAC) in rat testis is regulated by HCO 3 2 and pH, and it functions as a HCO 3 2 sensor (Chen et al., 2000). In the pathogenic fungus Candida albicans, the activity of sAC was related directly to the regulation of filamentation by CO 2 /HCO 3 2 (Klengel et al., 2005). The activity of CyaB1, one of the sACs in Anabaena PCC7120, is also regulated by HCO 3 2 (Cann et al., 2003). In the cyanobacterium Synechocystis PCC6803, it was shown that another sAC, Cya1, is regulated directly by CO 2 rather than HCO 3 2 (Hammer et al., 2006). However, there are no data yet to support the relationship between sACs and CCM regulation in cyanobacteria, while it is strongly suggested that a product of the photorespiratory pathway, 2-phosphoglycolate, directly enhances the binding of CmpR to the promoter of the cmpABCD operon to ensure the CO 2 response of this promoter (Nishimura et al., 2008). In eukaryotic algae, besides the above-described information from C. reinhardtii and P. tricornutum, little is known about the sensing mechanism for ambient CO 2 concentrations, the subsequent signal transduction pathways, and transcriptional controls of promoters in response to changes in CO 2 concentration.
In this study, critical cis-elements required for the CO 2 -cAMP response of the ptca1 promoter were identified, and transcription factors that may target the responsive elements were investigated.

Determination of Critical CO 2 -Responsive Elements in the Core Regulatory Region of Pptca1
Our previous study has shown that the core regulatory region governing the CO 2 response of Pptca1 resided 280 bp downstream relative to the transcription start site. To determine the precise CO 2 -responsive elements of Pptca1, a linker scan and the following one-base replacement assays were carried out on this core region. A series of linker-scan substitutions by the NotI restriction site (Fig. 1A) were introduced into the putative core regulatory region (280 to +61 relative to the transcription start site) of Pptca1 (Fig. 1B). These fragments were ligated to the b-glucuronidase gene, uidA, and were transformed into P. tricornutum by microparticle bombardment. As shown in the previous study , transcript levels of uidA closely reflected the levels of GUS activity in P. tricornutum. The results in Figure 2 show the apparent loss of the CO 2 response in the manipulated promoters LS1 to LS3 and LS5 to LS7. In contrast, other linkerscan promoters exhibited a clear CO 2 response equivalent to that observed in the intact promoter ( Fig. 2, LS-). Since the impaired sites of LS1 to LS3 and LS5 to LS7 corresponded, respectively, to regions around CRE1 and p300be, it was likely that these two elements are required for the CO 2 response, as was assumed previously , while other sequences including the Skn-1 and the CRE2 elements were not required for the CO 2 response of Pptca1. Levels of GUS activities expressed in transformants varied significantly depending upon clones. This is presumably due to position effects and the copy number of the promoter::uidA constructs integrating into different parts of the genome. However, GUS activity ratios (air to 5% CO 2 ) were relatively stable, which enabled us to assess the CO 2 response with some accuracy (Fig. 2). GUS activity ratios are thus presented for the GUS promoter assays hereafter.
We followed these initial results by determining the precise CO 2 -responsive elements around the putative CRE1 and p300be elements. Single-nucleotide ex-changes within purines or pyrimidines (A for C or T for G and vice versa) in regions around CRE1 (Fig. 3A) and p300be (Fig. 3B) revealed that the sequences 5#-ACGTCA-3# (268 to 263) and 5#-GTGACGC-3# (247 to 241), which did not completely overlap with CRE1 and p300be, were critical for the CO 2 response (Fig. 3). This implied that the originally described CO 2 -responsive elements, CRE1 and p300be, were not adequately defined. Raw activity data for Figure 3 are available in Supplemental Figure S4.

Search for Further Upstream Elements That May Work
Cooperatively with Elements around CRE1 and p300be The intact promoter, which has both ACGTCA and GTGACGC motifs, showed a clear CO 2 response (Fig.  4A, i). On the other hand, impairment of either one or both of these motifs by NotI insertion resulted in the loss of CO 2 response (Fig. 4A, ii-iv). Interestingly, however, addition of the upstream sequence of Pptca1 (21,292 to 281) to the region upstream of the short promoter sequence (280 to +61), in which one of two critical CO 2 -responsive motifs was impaired by the NotI insertion, conferred a clear CO 2 response on Pptca1 (Fig. 4A, vi and vii). However, such synergistic effects on the CO 2 response did not occur when both of these motifs were impaired by NotI insertion (Fig. 4A, viii). These results indicated the occurrence of additional critical elements upstream of 280 relative to the transcription start site, which could work cooperatively with the two newly found motifs.
In order to detect the upstream elements, which were anticipated from the results of Figure 4A to work cooperatively with the two elements, the long Pptca1 constructs with singly mutated motifs (Fig. 4A, vi and vii) were truncated from upstream at positions 2484, 2225, 2115, and 280 relative to the transcription start site and promoter activity was compared with those of intact promoters of each length (Fig. 4B). All transformants with these truncated constructs exhibited a clear CO 2 response, with GUS activity ratios between 1.8 and 7.0 ( Fig. 4B, ii, iii, v, vi, viii, and ix), except that uidA was totally derepressed when truncations were made to 280 (Fig. 4A, xi and xii). Results of GUS assays on cells containing the intact promoters of each length are also shown as controls (Fig. 4B, i, iv, vii, and x). These results clearly demonstrated the occurrence of the additional CO 2 -responsive element, which works with either of two downstream elements, at the region 2115 to 280 relative to the transcription start site of Pptca1. Raw activity data for Figure 4 are available in Supplemental Figure S4. critical elements, 5#-TGACGC-3#, at 247 to 242 (Fig.  5A). These two motifs are of complementary relationship with the other critical element, 5#-ACGTCA-3#, at 268 to 263 (Fig. 5A). When either the 286 to 281 or the 247 to 242 sequence relative to the transcription start site was replaced by the NotI sequence, the CO 2 response of the promoter (290 to +61) was well conserved (Fig. 5B, ii and iv). Furthermore, treatment with dbcAMP repressed these mutated promoters even under the air condition, mimicking the response under 5% CO 2 (Fig. 5B, ii and iv). However, impairment of the 268 to 263 sequence relative to the transcription start site or impairments of more than two motifs of these three inverted sequences resulted in full derepression of these manipulated promoters under 5% CO 2 (Fig. 5B, iii and v-viii). dbcAMP treatment also became completely ineffective in regulating these manipulated promoters; that is, the uidA reporter was derepressed independently of either CO 2 or dbcAMP (Fig. 5B, iii and v-viii). According to these Figure 1. Linker-scan constructs. A, The promoter region of ptca1 from 280 to +61 relative to the transcription start site was packaged into the transformation vector pPha-T1 upstream of the GUS reporter gene uidA. The NotI restriction site (gcggccgc) was introduced between 271 and 28 relative to the transcription start site by inverse PCR on pPha-T1 with Pptca1 (280 to +61) as the template. The resulting PCR products were self-ligated. B, The position of the introduced NotI site was slid toward downstream by 4 bp from LS1 to LS15. CRE1 and CRE2, p300be, and a putative Skn-1 site are indicated at the top of the sequence and highlighted in gray.
results, three inverted elements were designated, from 5# to 3#, as CO 2 -cAMP-Responsive Element1 (CCRE1), CCRE2, and CCRE3 ( Fig. 5A), and it was strongly suggested that CCRE2 is the most important element for the CO 2 -cAMP response, as a single impairment of this sequence totally abolished the function of Pptca1 (Fig. 5B, iii). Raw activity data for Figure 5 are available in Supplemental Figure S4.

Search for Putative Transcription Factors That May Target CCREs
The CCRE1 element was identical to the cis-element known to be the typical target sequence of human Activating Transcription Factor6 (ATF6; Wang et al., 2000). This prompted us to search for homologs of human ATF6 in the genome database of P. tricornutum. Candidates were screened on the basis of homology of the DNA-binding site, NxxSAxxSR, which is well conserved in the mammalian ATF6 family (Vinson et al., 2002), and the putative Leu-zipper motif, whose primary structure is constituted of a heptad repeat of Leu residues (Fig. 6A). As a result, eight genes were found as candidates encoding basic zipper (bZIP) proteins that may target CCREs in Pptca1 (Fig. 6A). Seven of these bZIP sequences, PtbZIP7, -8, -11, -12, -13, -15, and -16, were recently described and grouped by Rayko et al. (2010), although PtbZIPx had not previously been reported. The critical amino acid sequence for the bZIP domain deduced from these eight genes revealed significant similarity (Fig. 6A), although the sequence similarity of other regions of the open reading frames of these putative proteins was minimal (Supplemental Fig. S1). These eight candidates were expressed in Escherichia coli BL21(DE3), and PtbZIP11, -15, and -16 were successfully expressed as soluble proteins, whereas the other candidates were found in insoluble fractions (Supplemental Fig. S2). For PtbZIP13, we could not detect induction in the E. coli system (Supplemental Fig. S2), so it was not studied further.
Gel-Shift Assay of the CCRE Sequences by PtbZIP11, PtbZIP11, -15, and -16 were purified by nickel-Sepharose column chromatography (Supplemental Fig. S3), and the binding capacity of these candidates Figure 2. GUS reporter assay of P. tricornutum cells transformed with intact Pptca1 and the linker-scan constructs. Schematic drawings of native Pptca1 (280 to +61; LS-) and the linker-scan constructs 1 to 15 (LS1-LS15) are indicated at the left. GUS activity per total protein in lysates was determined in three clones of each transformant grown in 5% CO 2 (white bars) or air (black bars). Values of each bar are means 6 SD from three separate cultures. Ratios of GUS activity in airgrown cells to that in 5% CO 2 -grown cells are shown to the right of the bars. Asterisks indicate ratios below 2.0. PNP, p-Nitrophenol.
to a pentameric repeat of the TGACGT motif, (TGACGT) 5 , was examined. When 5 to 75 mg mL 21 PtbZIP15 and -16 was mixed with 125 mM (TGACGT) 5 , which was labeled by Cy5, the band of free probe started to disappear (Fig. 6B). However, in these assays, clear shifted bands were not detected; instead, some smeared signals were apparent at the top part of the gels (Fig. 6B). In contrast, a band shift accompanied by a disappearance of free probe was clearly observed by mixing more than 5 mg mL 21 PtbZIP11 with 125 mM (TGACGT) 5 (Fig. 6B).
In order to confirm the binding specificity between CCRE motifs and PtbZIP11, additional gel-shift assays were carried out using nonlabeled competitor sequences. When 25 mg mL 21 PtbZIP11 was mixed with 125 mg mL 21 of either of the two types of CCRE pentamers, (TGACGT/C) 5 , clear band shifts were observed (Fig. 6C). However, these shifted bands began to disappear with the addition of nonlabeled specific competitors for (TGACGT) 5 or (TGACGC) 5 at 10-fold higher concentrations for the labeled probe, and band shifts were completely abolished by the addition of 100-fold specific competitors (Fig. 6C). In sharp contrast, the addition of nonspecific competitor to concentrations 100-fold higher than the labeled probe did not inhibit the band shifts of either (TGACGT) 5 or (TGACGC) 5 (Fig. 6C).
Expression Profiles of ptbZIP11 during Acclimation from High CO 2 to Air Since the mode of regulation of Pptca1 was demonstrated previously to be repression under high-CO 2 conditions mediated by cAMP and derepression under normal air conditions (Harada et al., 2006), PtbZIP11 is most likely to act as a repressor of transcription. The expression of ptbZIP11 was analyzed during cell acclimation from 5% CO 2 to air by quantitative (q)PCR and was related to that of ptca1 (Fig.  6D). It was apparent that ptbZIP11 mRNA constitutively accumulated independently of ambient CO 2 concentrations (Fig. 6D), whereas that of ptca1 was completely repressed in 5% CO 2 but was fully derepressed under air conditions (Fig. 6D). Transcript . One-base replacements introduced to regions around the putative CRE1 and putative p300be elements, and effects on GUS activity. A, One-base replacements introduced to regions around CRE1 (271 to 260) are shown in the left half. GUS reporter assays performed on transformants with each construct of Pptca1::uidA are shown in the right half. B, Onebase replacements introduced to regions around p300be (252 to 242) are shown in the left half. GUS reporter assays done on transformants with each construct of Pptca1::uidA are shown in the right half. GUS activities in lysates of each transformant grown in 5% CO 2 (white bars) are expressed as relative values to those in air-grown cells (black bars). GUS reporter assays were performed on two to three clones obtained from each transformant. Values of each bar are means 6 SD obtained from three separate cultures. Ratios of GUS activity in air-grown cells to that in 5% CO 2 -grown cells are shown to the right of the bars. Asterisks indicate ratios below 2.0. . CO 2 response of Pptca1 with an introduced NotI site within putative CO 2 -responsive elements. A, Two different lengths of Pptca1 (280 to +61 and 21,292 to +61 relative to the transcription start site) were ligated to the uidA reporter gene in the pPha-T1 vector, and fragments of newly identified CO 2 -responsive elements, ACGTCA and/or GTGACGC, were substituted with the NotI restriction site. GUS activities in lysates of each transformant grown in 5% CO 2 (white bars) are expressed as relative values to those in air-grown cells (black bars). GUS reporter assays were made on two to three clones obtained from each transformation. Values of each bar are means 6 SD of three separate cultures. Ratios of GUS activity in air-grown cells to that in 5% CO 2grown cells are shown to the right of the bars. B, The upstream truncated fragments (2484 to +61, 2225 to +61, 2115 to +61, and 280 to +61 relative to the transcription start site) of Pptca1 were ligated to the uidA reporter gene in the pPha-T1 vector, and either the ACGTCA or GTGACGT motif was disrupted by the NotI restriction site. The intact promoter of each length is also shown as a control. GUS activities in lysates of each transformant grown in 5% CO 2 (white bars) are expressed as relative values to those in air-grown cells (black bars). GUS reporter assays were made on two to three clones obtained from each transformation. Values of each bar are means 6 SD obtained from three separate cultures. Ratios of GUS activity in air-grown cells to that in 5% CO 2 -grown cells are shown to the right of the bars. levels of ptbZIP11 were temporarily increased upon transfer of high-CO 2 cells to air, but levels under high CO 2 were equivalent to the basal air levels in cells acclimated for 12, 24, and 48 h (Fig. 6D). The relative expression level of ptbZIP11 was about 2 orders of magnitude lower than that of ptca1 in air-grown cells (Fig. 6D).

A Gain-of-Function Assay Using the CCRE Elements
In order to understand the topology of the putative transcription factors bound on the newly found CCRE elements to the basal transcriptional complex, a range of CCRE sequences were ligated to the minimal promoter (mP) of Pptca1 (210 to +61), and GUS reporter assays were carried out to assess the activity of the promoter. The distance between CCRE2 and mP was changed from 3 to 52 bp (Fig. 7, construct group a). The mP region alone drove GUS expression constitutively (Fig. 7, control). Interestingly, when CCRE2 was positioned immediately upstream (23 bp) of mP, it conferred a clear CO 2 response on it (Fig. 7, construct group a, i). However, by extending the interval to 10 or 12 bp, CCRE2 became inert in conferring the CO 2 response (Fig. 7, construct group a, ii and iii), while the function was recovered again when the interval was further extended to between 14 and 22 bp (Fig. 7, construct group a, iv-vii). At distances greater than 32 bp, the effect of the single CCRE2 was lost completely (Fig. 7, construct group a, viii-x).
In contrast to the results for construct group a, vii to x, in which promoters partially or totally lost CO 2 responsiveness, CCRE2 at intervals 21, 31, 41, and 51 from mP showed a clear CO 2 response when CCRE3 was added downstream of CCRE2 (Fig. 7, construct group b, i-iv). Similarly, the presence of CCRE1 upstream of CCRE2 also increased the repressive activity of CCRE2 at longer intervals from mP (Fig. 7, construct group c). Notwithstanding, the promoters completely lost their CO 2 response in the absence of CCRE2 at any interval, even in the presence of both CCRE1 and CCRE3 (Fig. 7, construct group d). Raw activity data for Figure 7 are available in Supplemental Figure S4.

DISCUSSION
In our previous study, a repressive effect of both CO 2 and cAMP to Pptca1 was clearly demonstrated both in qPCR assays of the endogenous ptca1 transcript and in Pptca1::uidA reporter assays (Harada et al., 2006). This previous study also identified CRE1 and CRE2 and a putative p300be occurring downstream of 270 in Pptca1. Partial deletion of one of the CREs or the putative p300be resulted in derepression of Pptca1 in high CO 2 or in the presence of dbcAMP in air. In this Figure 5. Primary structure of the putative core regulatory region of Pptca1, and confirmation of the function of three inverted repeats in the CO 2 and cAMP responsiveness of Pptca1. A, The three inverted sequences, TGACGT/C, newly found in Pptca1. B, The core regulatory region (290 to +61 relative to the transcription start site) of Pptca1 was ligated to the uidA reporter gene in the pPha-T1 vector, and parts or all three inverted sequences were substituted with the NotI restriction site with different combinations (left half). GUS activities in lysates of each transformant grown in 5% CO 2 (white bars) and acclimated to air in the presence of dbcAMP (gray bars) are expressed as relative values to those in air-grown cells (black bars). Values of each bar are means 6 SD obtained from three separate cultures. Ratios of GUS activity in air-grown cells to that in 5% CO 2 -grown cells are shown to the right of the bars. study, the newly found elements, CCRE1 to CCRE3, were clearly shown to be CO 2 -responsive elements that repressively control the transcription of ptca1 in high-CO 2 conditions via the second messenger cAMP. We have further demonstrated that CCRE2 clearly overlapped with CRE1, but CCRE3 is new, since it did Figure 6. Structures, binding property to CCRE, and expression profile of putative human ATF6 superfamily transcription factors in P. tricornutum. A, Putative bZIP regions of eight candidates, which may belong to the human ATF6 superfamily (PtbZIP7, -8, -11, -12, -13, -15, -16, and -x), were aligned with that of the human ATF6 protein. The well-conserved DNA-binding sequences, which may target the CCRE motif, are indicated by black dots and highlighted. The putative repeats of Leu in the ZIP region are indicated by asterisks and highlighted. B, A concentration of 125 mM CCRE1 pentamer labeled by Cy5 was mixed with different concentrations of PtbZIP11, -15, and -16. The reactant was separated on 6% polyacrylamide gels, and bands were detected by fluorography. C, A concentration of 125 mM labeled CCRE1 or CCRE3 pentamer was mixed with 25 mg mL 21 PtbZIP11 in the absence or presence of competitors. A specific competitor (SC), which is the nonlabeled CCRE1 or CCRE3 pentamer, or a nonspecific competitor (NC), which is 5#-CGATTCAAATG-GAGAGTGGGTGAGCAAGGGCGAG-3#, was applied at concentrations 10 or 100 times more than the labeled probe. White and black arrowheads indicate free probe and shifted bands, respectively. D, qPCR of the ptca1 and ptbZIP11 transcripts was carried out on total RNAs extracted from cells grown in 5% CO 2 , air, and cells acclimated from 5% CO 2 to air for 3, 6, 12, 24, and 48 h. Values are expressed as relative abundance to the amount of Cytoplasmic Glyceraldehyde-3-Phosphate Dehydrogenase2 (GapC2) transcript. not match the putative p300be, which turned out no longer to be important. Skn-1 and CRE2 are also found to not be functional for CO 2 response in this study. The occurrence of CCRE1 was overlooked in the previous study due to an apparently sufficient repressive effect with only CCRE2 and CCRE3 in response to high CO 2 or dbcAMP. However, it was clearly shown in this study that CCRE1 complemented the absence of CCRE3.
CCRE1 to CCRE3 occurred with intervals of 12 to 15 nucleotides, which implies that each CCRE region locates by about two pitch turns of double-stranded DNA; thus, the plane of a base pair hybrid, which is targeted by a transcription factor, presumably faces to one side of the double helical structure of DNA in CCREs. CCRE2 is likely the most critical component among CCREs, due to its requirement for the CO 2 response of Pptca1 (Fig. 5B).
It is interesting that CCRE1 and CCRE3 were nearly perfect inverted sequences of the CCRE2 element (Fig.  5A), strongly suggesting the pivotal requirement for CCRE2 to work cooperatively with CCRE1 and/or CCRE3. Synergistic effects between CRE-binding protein (CREB) and other transcription factors, which Figure 7. Gain-of-function assay using CCREs at altered distances from the mP region of Pptca1. The CCRE2 sequence was fused to mP of Pptca1 (210 to +61 relative to the transcription start site) at 10 different distances from 3 to 52 bp from mP (left half). GUS activities in lysates from each transformant grown in 5% CO 2 (white bars) are expressed as relative values to those in air-grown cells (black bars). Values of each bar are means 6 SD obtained from three separate cultures. Ratios of GUS activity in air-grown cells to that in 5% CO 2grown cells are shown to the right of the bars. Asterisks indicate ratios that exceed 2.0. target other cis-elements, have frequently been observed in cAMP-responsive transcriptional controls in mammalian cells (Hyman et al., 1989;Nitsch et al., 1993;Roesler et al., 1995). However, to our knowledge, the repression of a promoter by alternative combinations of three cAMP-responsive elements has not been observed previously.
In this study, eight genes presumably encoding putative transcription factors were retrieved in the genome database of P. tricornutum, based on the homology of bZIP domains to that of the human ATF6 family (Fig.  6A). These proteins showed a striking similarity in their putative DNA-binding domains, but sequences other than bZIP domains showed little similarity to each other or to any known sequences (Supplemental Fig. S1). It is thus suggested from the primary structure that these putative bZIP proteins could be categorized to the ATF6 superfamily, but their interaction partners and functions are likely to be unique to diatoms.
Structures and functions of ATF family proteins have been studied mostly in mammalian cells. In general, transcription-activating domains reside at their N termini, although the domain structures of the N-terminal regions are totally different depending on each ATF subclass (Rehfuss et al., 1991;Nomura et al., 1993;Ribeiro et al., 1994;van Dam et al., 1995;Haze et al., 1999;Hai and Hartman, 2001;Hashimoto et al., 2002). Most mammalian ATF family proteins have a bZIP domain at their C terminus, but ATF6, exceptionally, has a bZIP region within the central part of the sequence (Haze et al., 1999;Hai and Hartman, 2001). Human ATF6 is a transmembrane protein localized on the endoplasmic reticulum membrane and is transferred to the nucleus in response to endoplasmic reticulum stresses by proteolysis of both sides of the transmembrane domain by the proteases, S1P and S2P (Haze et al., 1999). PtbZIPs reported in this study also contained more central bZIP domains (Supplemental Fig. S1), although none of them possessed transmembrane regions, suggesting that their initial functional location in cells would also be unique to diatoms.
Recently, putative bZIP sequences were analyzed by Rayko et al. (2010) within the genomes of the two model diatoms, P. tricornutum and Thalassiosira pseudonana. Diatom bZIP sequences were divided into six groups and constituted a stramenopile-specific family, which showed no notable homology with those from other species (Rayko et al., 2010). ATF family candidates found in this study mainly belong to group 1 to 3 diatom bZIPs, although none of them (except PtbZIP7 and PtbZIPx) contained Per-Arnt-Sim domains. PtbZIPx could not be categorized into any of the subgroups reported by Rayko et al. (2010), but this putative transcription factor possesses a Per-Arnt-Sim domain overlapping with the bZIP domain (data not shown), strongly suggesting its relation to group 2 diatom bZIPs.
ATF family proteins form homodimers or heterodimers via their Leu-zipper domain upon binding to cis-elements (Hai and Curran, 1991;Liu et al., 2006). On the other hand, the number of Leu or Ile residues in putative Leu-zipper regions in PtbZIPs was variable, and some of the proteins, PtbZIP8, -12, -13, -15, and -x, may not even be able to form a dimer due to the lack of enough Leu repeats (Fig. 6A). Nonetheless, their DNA-binding motif is well conserved, strongly suggesting their function as transcription factors that are closely related to the bZIP family.
The binding and specificity of PtbZIP15 and -16 to (TGACGT) 5 were not clear (Fig. 6B), although an apparent decrease in free probe strongly suggested the affinity of these proteins to the element (Fig. 6B). Thus, PtbZIP11 is so far the only candidate transcription factor that directly binds to CCREs. However, from the results shown in Figure 6B, there is a possibility that other directly binding transcription factors besides PtbZIP11, which target CCREs with different affinities and may recruit different binding partners to CCRE regions, also participate in the regulatory system.
The constitutive expression under changing CO 2 conditions of the ptbZIP11 transcription factor gene ( Fig. 6D) strongly suggested that the function of PtbZIP11 is activated under high-CO 2 conditions by posttranslational processes in response to the increment of cytosolic cAMP. This is also the case for human ATF6, which is known to be expressed constitutively and whose function is regulated at the posttranslational level by proteolytic digestion (Haze et al., 1999). Accordingly, upon exposure of P. tricornutum cells to limited-CO 2 conditions, these repressive factors could be quickly deactivated by the reverse processes of the posttranscriptional modification to enable the derepression of Pptca1 within 3 h (Fig. 6D).
The data from the gain-of-function assays strongly suggest that spatial configurations between transcription factors on CCRE2 and mP regions on the double helical strands of DNA critically control the affinity of the basal transcriptional complex on the mP region, and that CCRE1 or CCRE3 reveals a function to ensure this regulation only in the presence of CCRE2. In other words, when CCRE2 is close enough to the mP region, PtbZIPs on CCRE2 would stably immobilize the basal transcription complex or stabilize the repressor on mP and repress transcription (Fig. 8A), presumably mediated by some other transcription factors, which is conceptually depicted in Figure 8, as a mediator complex. But this system probably requires specific topology of these complexes on the DNA strands, which would be why repression was lost when CCRE2 was located on the twisted face of the DNA strand at relatively longer distances from mP (10-12 bp and more than 32 bp; Fig. 8B). The construct shown in Figure 7 (construct a, x) has the same distance between CCRE2 and mP as that in the intact Pptca1, but construct a, x, exhibited no repression. It is thus probable that, at distances greater than 10 to 12 bp, the DNA strand downstream of CCRE2 would need to bend to interact with the basal transcriptional complex, as demonstrated previously in NeP1 binding to the F1 silencer element of the chicken lysozyme gene (Arnold et al., 1996; Fig. 8C), and CCRE1 or CCRE3 may work with CCRE2 by stabilizing this DNA bending (Fig. 8D). This assumption is supported by the gain-of-function assays using construct groups b to d in Figure 7, which show that the repressive function of CCRE2 at locations longer than 32 bp from mP under high CO 2 was resumed when CCRE2 was combined with either CCRE1 or CCRE3.
CCREs are almost identical sequences that align in inverted directions within 50 bp of Pptca1 (Fig. 5A). The functions of CCREs are likely specific for their directionality, because CCRE3 could not repress Pptca1 even when the element was attached immediately upstream of mP in the absence of CCRE2 (Fig. 7, construct d, i). It is probable that the repression complex formed on CCREs would be asymmetric, whose function is thus highly dependent on the direction of the CCRE sequence. Such a direction-dependent complex should show its asymmetric property strongly when the location is artificially approximated to mP. In the intact Pptca1, CCREs need to work with alternative combinations of CCRE1/2 or CCRE2/3. Combination did not work between CCRE1 and CCRE3; thus, the partners probably need to be complementary with each other. According to the structural model of the CREB-CRE complex, a bZIP dimer occupies one CRE element (Schumacher et al., 2000). This is also presumed to be the case for PtbZIP11. Each of the two CCRE combinations would thus be expected to form two complexes of CCRE-PtbZIP11 on Pptca1 (Fig. 8D). Alternatively, it is also possible that these combinations of CCREs form stem loops on each DNA strand when DNA is single stranded (Fig. 8E). However, epigenetic properties enabling such structural controls of Pptca1 upon transcriptional switching still remain to be explored.

CONCLUSION
In this study, the precise sequences of CCREs in Pptca1 and a possible transcription factor targeting them, PtbZIP11, were determined. Presumed topological models of Pptca1 required for CO 2 -cAMP-responsive transcription were proposed in this study, and the model suggested that a repeated number and directionality of CCREs (i.e. distances between CCREs and the relative location of CCREs from mP) are the critical factors in controlling Pptca1 activity. Our preliminary study further found that other CO 2 -responsive genes also possessed CCREs in their promoters but have different numbers and arrangements, and these promoters behave differently from CO 2 compared with that of Pptca1 (Y. Matsuda, unpublished data). The model we propose in this study thus potentially provides an elemental example to understand the mechanisms of structural control of a promoter region, which lead to switching of transcriptional activity in response to changes in ambient CO 2 in marine diatoms.

Cells and Culture Conditions
The marine diatom Phaeodactylum tricornutum (UTEX 642) was obtained from the University of Texas Culture Collection and was grown in artificial  Figure 7, direct harboring of a single CCRE2 to mP conferred a repression in high CO 2 (Fig. 7, construct a, i). The PtbZIP complex anchors into CCRE2 and immobilizes the basal transcription complex, presumably mediated by some repressive complex depicted as a mediator. B, Extension of the distance between mP and the single CCRE2 abolishes the repressive interaction (Fig. 7, construct a, ii and iii), presumably due to changes in the relative configuration between the PtbZIP complex and the basal transcription complex. C, The PtbZIP complex regains repression (Fig. 7, construct a, iv-vi) when it is turned back to the proper face of the DNA strand. The DNA strand needs to be bent at longer distances between CCRE2 and mP, which would be ensured by the strength of interaction between the PtbZIP complex, the putative repressive mediator, and the basal transcription complex. D and E, In the endogenous Pptca1, CCRE2 cooperates with either CCRE1 or CCRE3, which would ensure a stronger interaction with the basal transcription complex to be immobilized. Multiple PtbZIPs may strengthen the interaction (D), or a partial unfolding of the DNA strand at CCRE1/2 or CCRE2/3 may make the angle of the CCRE-PtbZIP complex pliable and ensure the interaction with the basal transcription complex (E). In the cartoons, the case of CCRE2 and CCRE3 is depicted. seawater, which was supplemented with half-strength Guillard's "f" solution (F/2ASW;Guillard and Ryther, 1962;Harrison et al., 1980) under continuous illumination of 50 mmol m 22 s 21 photosynthetic photon flux density under constant aeration with 5% CO 2 or ambient air at 20°C. In acclimation experiments, 5% CO 2 -grown cells were harvested by centrifugation at 3,500g for 5 min under room temperature, washed twice with CO 2 -free F/2ASW, transferred to F/2ASW aerated with air, and allowed to acclimate for 1 to 2 d. In some experiments, 1.0 mM dbcAMP (Sigma-Aldrich) was added during the acclimation to air.

Preparation of Transformation Constructs
A series of upstream truncated constructs of Pptca1 (21,292 to +61, 2484 to +61, 2225 to +61, 2115 to +61, 290 to +61, and 280 to +61 relative to the transcription start site) were amplified by PCR using Pptca1 (21,292 to +61) as a template. These constructs were phosphorylated and inserted into the bluntended site of transformation vector pFcpApGUS , which is a derivative of the general P. tricornutum transformation vector pPha-T1 equipped with a bleomycin-resistant gene cassette (Zaslavskaia et al., 2000).
A set of linker scan constructs with a NotI restriction site as the linker was synthesized by a modified inverse PCR method using the vector plasmid pFcpApGUS containing Pptca1 (280 to +61) as a template (Fig. 1A). Primer sets were designed to generate a NotI restriction site (gcggccgc) at desired locations by self-ligation (Fig. 1A). Fifteen linker-scan constructs 280 downstream from the transcription start site were generated (Fig. 1B, LS-and LS1-LS15).
In some experiments, NotI substitutions were introduced into 267 to 260 and/or 247 to 240 relative to the transcription start site using the aforementioned vector plasmid pFcpApGUS containing upstream truncated Pptca1s of five different lengths (21,292 to +61, 2484 to +61, 2225 to +61, 2115 to +61, and 280 to +61) as templates. In some experiments, NotI substitutions were introduced into 287 to 280, 269 to 262, and/or 248 to 241 relative to the transcription start site using the vector plasmid pFcpApGUS containing upstream truncated Pptca1 (290 to +61) as a template.
To generate one-base replacement constructs, PCR was carried out targeting the sequences 271 to 260 and 252 to 242 using the vector pFcpApGUS containing Pptca1 (280 to +61). To generate gain-of-function constructs for modeling Pptca1 structure, LS7 was used as a template, and the distance between ACGTCA (268 to 263) and mP was altered by a modified inverse PCR method. Primer sets were designed to alter the distance between ACGTCA and mP from 3 to 52 bp by self-ligation. Primer sequences used to generate these transformation constructs are summarized in Supplemental  Table S7.

Transformation of P. tricornutum
Air-grown cells of P. tricornutum were harvested at midlogarithmic phase (optical density at 730 nm = 0.3-0.4). Approximately 5 3 10 7 cells were spotted as a plaque of 2.5 cm diameter on the surface of the F/2ASW agar plate. Five hundred micrograms of tungsten microcarrier (Bio-Rad; M17; approximately 1.1 mm particle size) was coated with 1.0 mg of plasmid DNA containing 1.0 M CaCl 2 and 16 mM spermidine. The PDS-1000/He Biolistic Particle Delivery System (Bio-Rad) was used for microprojectile bombardment of the microcarrier. The bombardment was done at 10.7 MPa to the cells in the chamber under negative pressure of 91.4 kPa with a target distance of 6 cm. Bombarded cells were maintained on the plate for 1 d in the dark and were then suspended in 2 mL of F/2ASW. This cell suspension was plated on agar plates containing 100 mg mL 21 Zeocin (Takara Bio) in F/ 2ASW and allowed to form colonies for 3 to 4 weeks under continuous illumination.

Quantification of GUS Activity
Cells at different acclimation stages were harvested, disrupted by sonicator (Ultrasonic disruptor model UD-201; TOMY Seiko Co. Ltd.) in 0.5 to 1.0 mL of GUS extraction buffer (50 mM sodium phosphate [pH 7.0], 10 mM b-mercaptoethanol, and 0.1% [v/v] Triton X-100] in an ice bath, and then centrifuged at 15,000g for 5 min at 4°C. Lysate (18 mL) was added to 882 mL of GUS reaction buffer (10 mM p-nitrophenyl b-D-glucuronide in the GUS extraction buffer) and incubated at 37°C. The reaction was terminated by the addition of 0.5 M Na 2 CO 3 every 10 min after starting the reaction, and the absorbance of p-nitrophenol released from the substrate was measured at 405 nm. Protein concentrations in lysate were measured by the Bradford method (Bio-Rad) according to the manufacturer's protocol.

Expression and Purification of PtbZIPs
Open reading frames of ptbZIPs were inserted into pET-21a vector designed to fuse the His hexamer to the C termini of products. Primer sets for the amplification of ptbZIPs are shown in Supplemental Table S6. PtbZIPs were expressed in Escherichia coli strain BL21(DE3) in 200 mL of Luria-Bertani medium supplemented with 1.0 mg mL 21 ampicillin and 0.5 mM isopropyl-1thio-b-D-galactopyranoside. The E. coli cells were lysed by sonication in 10 mL of binding buffer (50 mM Tris-HCl [pH 8.0], 500 mM NaCl, and 5 mM imidazole). Crude lysates were applied to a 1-mL nickel-Sepharose column (Ni-Sepharose 6 Fast Flow; GE Healthcare Biosciences) repeatedly three times at 4°C, then washed with 12 mL of binding buffer (50 mM Tris-HCl [pH 8.0], 500 mM NaCl, and 5 mM imidazole) and 12 mL of washing buffer (50 mM Tris-HCl [pH 8.0], 500 mM NaCl, and 50 mM imidazole). After confirming the absorbance of eluate at 280 nm to be below 0.05, 12 mL of elution buffer (50 mM Tris-HCl [pH 8.0], 500 mM NaCl, and 300 mM imidazole) was applied to the column, and the eluates were collected for every 1 mL. The eluates were dialyzed with 1.2 L of imidazole-deprived buffer (50 mM Tris-HCl [pH 8.0] and 500 mM NaCl) using cellulose membranes (Viskase Companies). Concentrations of protein in eluate were determined by the Bradford method (Bio-Rad).

Gel-Shift Assay
As probes, pentamers of double-stranded 5#-TGACGT-3# or 5#-TGACGC-3# were labeled with Cy5 at their 5# termini. A concentration of 2.5 mM labeled probe was added to 100 mL of annealing buffer (10 mM Tris-HCl [pH 7.5], 50 mM NaCl, and 5 mM EDTA), incubated at 96°C for 10 min, and then gradually cooled to 30°C for 45 min. The labeled probes at a final concentration of 125 mM were mixed with different concentrations of purified PtbZIPs in binding buffer (30 mM Tris-HCl [pH 7.5], 0.6 mM EDTA, 0.6 mM dithiothreitol, 50 mM KCl, 1 mM MgCl 2 , and 12% glycerol) and incubated for 30 min at 20°C. In some experiments, 1.25 or 12.5 mM nonlabeled DNA probes, which have either the same sequence to labeled probes or noncompetitive sequence (5#-CGATT-CAAATGGAGAGTGGGTGAGCAAGGGCGAG-3#) to the labeled probe, were added in the binding buffer. Samples were mixed with 3 mL of 50% glycerol, separated on a nondenaturing 6% polyacrylamide gel in 0.53 Trisborate/EDTA, and Cy5 signal was observed by a fluorescence imager (Molecular Imager FX; Bio-Rad).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Sequence alignment of human ATF6 and eight PtbZIPs.
Supplemental Figure S2. Confirmation of the expression of PtbZIPs in E. coli strain BL21(DE3).
Supplemental Table S1. Primer sequences used to synthesize upstream truncated constructs.
Supplemental Table S2. Primer sequences used to synthesize linker-scan constructs.
Supplemental Table S3. Primer sequences used to synthesize one-base replacement constructs around CRE1.
Supplemental Table S4. Primer sequences used to synthesize one-base replacement constructs around p300be.
Supplemental Table S5. Primer sequences used for the insertion of the NotI sequence into CCREs.
Supplemental Table S6. Primer sequences for the amplification of ptbZIPs.
Supplemental Table S7. Primer sequences used to synthesize Pptca1 modeling constructs.