A regulator of nutritional copper signaling in Chlamydomonas is an SBP domain protein that recognizes the GTAC core of copper response element

Results

Complementation of crr1. We used an indexed cosmid library carrying Chlamydomonas genomic DNA to complement the crr1 strains (see Supporting Materials for details). DNA from three microtiter plates, 9, 33, and 81, supported the restoration of Crr1 function in one or two Arg+ colonies per Petri dish in both sets of transformations of crr1 strains. Five transformants were tested for cosegregation of growth on -Cu medium and Arg prototrophy by backcross to crr1arg7 or (from plate 81) were confirmed to be photosynthetically competent in -Cu medium on the basis of fluorescence rise and decay kinetics, which suggests restoration of CYC6 expression (e.g., Fig. 1B). Representative transformants (12 corresponding to plate 9 and 6 to plate 33) were further analyzed for copper-responsive CYC6 and CPX1 expression. The pattern of expression of these copper-deficiency response marker genes confirmed that each transformant was likely to be a genuine rescued strain. The complementing cosmids from microtiter plates 9 and 33 were specifically localized by testing DNA preparations from columns and rows of each plate for their ability to rescue strain crr1arg7cw15. Approximately 20 candidate CRR1 colonies were obtained for each positive column or row, and, of these, 4 representatives were tested from each set specifically for copper-responsive CYC6 and CPX1 expression. The complementing cosmids 9F1, 9A3, and 33E8 were isolated and shown to rescue the crr1 phenotype.

Restriction mapping indicated that the inserts in cosmids 9F1 and 9A3 overlap ≈90% with each other and share a common 13-kb section of DNA with 33E8, suggesting that the complementing cosmids represented a single region of the genome. The minimal complementing fragment was localized to a 6-kb BamHI fragment (Fig. 1 A; also see Fig. 3) and was confirmed to rescue both growth and photosynthetic competence of strain crr1–2arg7 (data not shown). Transformants derived from rescue of crr1-1 and crr1-2 by the 6-kb BamHI fragment showed normal copper-responsive expression of each of the marker genes CYC6, CPX1, and CRD1 and the 3-kb form of the CTH1 mRNA (Fig. 2A).

Previously, we had shown that several target genes of the copper response pathway were up-regulated by growth in low O₂ or by provision of high concentrations of nonessential transition metals like cobalt or nickel ions (13, 15, 21). The hypoxia- or nickel-responsive expression of these genes required the CuREs and was absent in the crr1 mutant, suggesting that the same signal transduction components were responsible for the response to low O₂ and nickel/cobalt ions. Therefore, we tested and confirmed restoration of nickel- and hypoxia-responsive expression of target genes in representative rescued colonies (Fig. 2 B and C, respectively).

Sequence analysis (see below and Supporting Materials) indicated that the BamHI fragment did not include the entire 3′ UTR of the CRR1 gene, and a 9-kb EcoRI fragment, which included the entire gene (Fig. 3), was shown to rescue the crr1 phenotype. Nevertheless, the expression of target genes was indistinguishable from wild-type strains in transformants carrying the 3′ truncated (:B) vs. intact (:E) CRR1 gene (e.g., Fig. 2 B and C).

Crr1 Sequence Analysis. Southern hybridization analyses and sequencing of the cloned complementing DNA and cDNAs derived from the locus revealed that the CRR1 locus represents a single gene with three small exons that produced a 5.8-kb mRNA encoding a 1,232 residue polypeptide (Fig. 3). A significant portion of the polypeptide consists of low complexity regions (shown in light gray), but we noted three regions that spoke to function: An SBP domain was identified by both blast analysis (7 × 10^-17 on August 2002) and via a search at the Pfam site (1.3 × 10^-35 on February 2004) as was a repeat of three ankyrin motifs, and visual inspection of the sequence revealed five sets of Cys-containing motifs in the C-terminal region of the protein, CxxCxC at 942, 1146, and 1185, and CxxC at 1164 and 1220, plus a number of His, Met, and additional Cys residues.

The ≈78-residue SBP domain (see Fig. 8, which is published as supporting information on the PNAS web site) is named for a novel Zn-dependent DNA-binding domain, which was discovered originally in two nuclear proteins SBP1 and SBP2 from Antirrhinum majus that bound the promoter of a floral meristem identity gene called SQUAMOSA (24, 25). The SBP domain occurs in a number (16 in Arabidopsis, >12 in Chlamydomonas, and >6 in rice) of otherwise unrelated proteins in the plant kingdom that function in fundamental developmental processes, especially associated with reproduction (26–29). The consensus DNA sequence bound by the SBP domain contains a GTAC core (23, 29), and this point was immediately relevant in terms of Crr1 function because the Chlamydomonas CuRE had been identified as having a critical GTAC core sequence by extensive mutagenesis of two such elements associated with CYC6 and one element associated with CPX1 (13). The SBP domain also contains a nuclear localization signal, which has been shown to target a reporter construct to the nucleus (23), and a recent study showed that a GFP fusion with Arabidopsis SPL14 was targeted to the nucleus (30). The presence of the SBP domain in Crr1, therefore, is consistent with its function as a transcription factor.

Southern analysis of the crr1-1 and crr1-2 strains did not reveal any major alteration of the CRR1 locus in mutants relative to the wild-type and the locus, corresponding to position -1092 to +4839 of the BamHI fragment, was amplified from each strain and analyzed by sequencing (see Methods). The crr1-2 and crr1-1 strains were both found to carry single nucleotide deletions at positions 2834 and 3343, respectively, leading to frameshifts between the SBP domain and the ankyrin repeats (at codons 780 and 895) and, hence, premature termination of translation (Fig. 3).

Expression of CRR1. RNA blot hybridization indicated that the locus produces a low abundance, ≈6 kb mRNA (Fig. 4). In the wild-type strains 2137, CC125, and CC425, the abundance of the RNA is always slightly (0.7× to 0.8×) lower in RNA prepared from -Cu vs. +Cu cells, as confirmed by real-time PCR measurement (Table 1), but it is unlikely that this difference is significant in terms of the dramatic (up to 10³-fold for CYC6) metal-responsive regulation of the Crr1 target genes. The abundance of the CRR1 mRNA is not changed much in the crr1 mutants relative to wild-type (e.g., if it were degraded by nonsense mediated decay) (Table 1), and this observation is consistent with the position of the frame-shift mutations in the third and fourth exons of the gene. The strains (:B transformants) carrying the 3′ truncated CRR1 gene produce an appropriately truncated mRNA of 4.8 kb, whose pattern of expression resembles that of the full-length endogenous gene (Fig. 4).

RT-PCR and 5′ RACE analysis of CRR1 mRNAs indicated the presence of two populations of mRNAs in both the total RNA preparation and also the polyadenylylated RNA preparation, and these mRNAs represented molecules in which the first intron was either spliced or not (diagrammed in Fig. 3; see also Fig. 9, which is published as supporting information on the PNAS web site, for data). The abundance of the unspliced mRNA was equivalent to or greater than the abundance of the spliced mRNA. The unspliced mRNA encodes an in-frame termination codon within the first intron, which could lead to nonsense mediated decay of the mRNA or, if translation were reinitiated at the next methionine, to a slightly shorter protein lacking the first 50 amino acids, including the AHA motif. To assess whether alternative splicing of the first intron might be biologically significant, we tested the ratio of spliced to unspliced CRR1 mRNAs under conditions where Crr1 target genes are regulated, including copper-, zinc-, or iron-deficiency vs. sufficiency, in normoxic vs. hypoxic, or with nickel supplementation (see Fig. 9). The ratio of the spliced to the unspliced form was not affected under any condition, and we conclude that the first intron is spliced slowly relative to other steps of mRNA processing. Slow splicing events have been observed previously for Chlamydomonas RNAs (31).

Crr1-SBP Domain Binds the CuREs. Genetic analysis had suggested that the CRR1 locus functions through the CuREs of the target genes of the nutritional copper-signaling pathway (unpublished results and ref. 16). To assess whether Crr1 interacts biochemically with the CuREs, we tested the putative DNA-binding SBP domain of the protein in an electrophoretic mobility shift assay with labeled CuREs from the CYC6 and CPX1 genes (Fig. 5). The His-6-tagged SBP domain of Crr1 could bind each of the two CuREs from the CYC6 gene (Fig. 5A for the distal CuRE and data not shown for the proximal CuRE) and the CuRE from the CPX1 gene. DNA binding was shown to be specific for the wild-type vs. the mutated element by competition with unlabeled DNA: a 10-fold higher concentration of unlabeled mutant oligonucleotide was required compared to the wild-type sequence to achieve the same degree of competition (Fig. 5). An irrelevant nucleotide sequence from the Arabidopsis CAULIFLOWER promoter was also not an effective competitor in the gel-shift assay (data not shown).

In previous work, we had indicated that the GTAC core of the CuRE was absolutely essential for copper-responsive gene expression in vivo because alteration of any nucleotide in the core to any other nucleotide completely abolished gene expression (13). This finding was recapitulated in the in vitro assay, where the single T to A substitution in the GTAC core (Fig. 5A) was as effective as the scrambling of all four nucleotides (Fig. 5B for CPX1 CuRE and data not shown for scrambled CYC6 CuRE). The Crr1 SBP domain is also able to bind the putative binding site of the Arabidopsis SBP domain in the AP1 promoter, and the SBP domain from Arabidopsis SPL1 is able to bind the CuRE (23), which confirms that the GTAC core is a key recognition element for the SBP domain.

Consistent with the structural model, which indicated the presence of two bound zinc ions in the SBP domain, preincubation or renaturation of the protein in the presence of EDTA inhibited the DNA binding activity of the recombinant protein, whereas renaturation in the presence of zinc could restore activity to protein that had been treated previously with EDTA (Fig. 10, which is published as supporting information on the PNAS web site).

Discussion

Crr1 is required for up-regulation of CYC6, CPX1, CRD1, and CTR genes and also for down-regulation of the CTH1 locus. Both types of responses rely on transcriptional activation through CuREs associated with the target genes. For up-regulation, Crr1-dependent activation produces templates for synthesis of the respective polypeptides, whereas for down-regulation of Cth1, Crr1-dependent promoter activity interferes with the production of mRNA templates (19). Crr1 is, therefore, a key regulator of copper homeostasis in Chlamydomonas, and it works by transcriptional activation. Three biochemical activities are required for transcriptional activation of the copper-deficiency response: recognition of the CuRE on target genes, copper sensing, and transcriptional activation. These activities may reside on a single molecule or on different molecules that interact with each other. Genetic analysis has revealed only a single regulatory locus for nutritional copper signaling and, therefore, we consider and discuss a model where Crr1 carries all three activities (Fig. 6).

CuRE and the GTAC Core. Sequence and functional analysis of Crr1 confirms the presence of a DNA-binding domain with selectivity for the CuRE (Figs. 3 and 5), and, therefore, we conclude that Crr1 houses at least one of the three required properties of a regulator of nutritional copper signaling. Mutational analysis of the promoters on the CYC6 and CPX1 genes had indicated the importance in vivo of a GTAC core in the CuRE, where every position in the core was critical for CuRE activity (13). The DNA-binding activity of the Crr1-SBP domain on the GTAC core of the CuRE now provides the biochemical basis for this in vivo sequence requirement.

The prototypical SBP domain was discovered on the basis of its ability to bind the SQUAMOSA promoter in Antirrhinum majus (24). Subsequently, a sequence comparison of the SQUA promoter, the promoter of the homologous AP1 gene in Arabidopsis and the promoters of other MADS-box genes that are expressed specifically during floral development, suggested a consensus SBP domain-binding site T-CGTACAA- (29). However, specific in vivo targets of each SBP domain protein are not known, and, hence, an in vivo mutational analysis of the binding site was not possible. The discovery of an SBP domain protein as a regulator of target genes with well defined cis-regulatory sequences (13) and the recovery of the GTAC motif in a target site selection study (23) now allows us to authenticate the DNA-binding site of the SBP domain. We conclude that all positions in the GTAC core (but not the flanking nucleotides) are critical. Nevertheless, other nucleotides must contribute to selectivity because the core by itself is insufficient for in vivo CuRE or HyRE (hypoxia responsive element) activity (13, 21).

Copper Sensing. A provocative aspect of the Crr1 sequence was the presence of two candidate metal-binding domains: the SBP domain and the C-terminal Cys-rich sequence, which opens the door to the possibility that the copper sensing activity also resides on Crr1. The SBP domain consensus sequence includes 10 highly conserved His and Cys residues that are potential metal-binding ligands (see Fig. 8). Eight of these residues were shown to bind two Zn ions in an NMR study of two bacterially expressed SBP domains from Arabidopsis, and mutagenesis confirmed the importance of the zinc-liganding cysteine residues for the DNA-binding activity of the SBP domain (23, 25). The two remaining His residues could potentially be involved in metal-sensing. The C-terminal Cys-rich sequence (16 Cys and 3 His in 120 C-terminal residues) is unique to Crr1, and its similarity to Drosophila metallothionein 2 (see Fig. 11, which is published as supporting information on the PNAS web site) suggests that it may participate in some aspect of metal-sensing. We are considering two, not necessarily mutually exclusive, models for metal sensing based on models developed in other laboratories (2).

In the copper-deficient cell, the zinc-binding site of the SBP is occupied and the protein is located in the nucleus, where it interacts through the ankyrin repeats directly or indirectly with a component of the transcription apparatus, and activates the target genes (Fig. 6A). In one model, the structure of the SBP domain is modified, owing to direct interaction of the conserved His residues in SBP with copper and resulting in rearrangement of the zinc-binding site (Fig. 6B), and in another model from intramolecular communication with the copper-bound form of the C-terminal Cys-rich region (Fig. 6C). The structural change could block DNA binding and/or interaction with coactivators through activation domains, or it could prevent nuclear localization, in either case preventing activation of target genes. The displacement of zinc by copper is precedented by mammalian metallothioneins, which release zinc in the presence of copper, and the intramolecular communication between a DNA-binding domain and a regulatory metal site is precedented by Mac1 (10, 32). An intramolecular interaction between a metal-binding Cys cluster and a transcriptional activation domain is also proposed for MTF-1 (33).

It is also possible that there are multiple layers of metal-responsive regulation of Crr1 function, and metal binding to the C terminus may modify protein localization or degradation. Unlike other essential nutrients, the transition metals are toxic in excess, and, therefore, it is not uncommon for nature to call on multiple layers of regulation, including transcription factor trafficking and degradation, to maintain homeostasis (2).

Nickel and O₂ Sensing. Crr1 function is modified not only by copper, but also by nickel and cobalt ions and oxygen supply. Nickel and cobalt ions function antagonistically to copper so that the target genes are turned on by nickel ions, even in fully copper-replete cells (15). Nickel and cobalt are not essential elements for Chlamydomonas and the response, therefore, is clearly artificial, but it provides a useful pharmacological tool to understand the mechanism of copper sensing by Crr1. Comparison of the nickel- and cobalt-induced responses of the target genes suggested that these ions were interfering with copper sensing because the pattern of expression of the target genes resembled closely the pattern noted in the copper-deficiency state. The most likely scenario is that nickel or cobalt bind at the copper-sensing site with an affinity or an off-rate constant that precludes displacement by copper, and, hence, Crr1 would be locked in a state where it is constitutively active. This idea is precedented by structural studies of the prokaryotic ArsR-SmtB family of metalloregulators, which indicate that the same site can accommodate different metal ions with variation in coordination geometry (34, 35). In this context, the conserved but unliganded histidines in the zinc-bound form are of interest.

The hypoxic activation of the target genes is clearly separate from nutritional copper signaling, and it represents a physiologically relevant mechanism. For instance, the CPX1 gene responds much more strongly to hypoxia than does the CYC6 gene, whereas the opposite is true in copper deficiency, and although Crr1 and a CuRE are required for the hypoxic response, a separate HyRE is also required (21). Therefore, we concluded that the hypoxia signaling is not a simple recapitulation of the copper-deficiency response. We proposed instead that hypoxia signaling represents a physiological intersection with the nutritional copper network. One possibility is that hypoxic growth conditions affects aspects of metal metabolism that indirectly impact Crr1 function. Another attractive possibility is that hypoxia alters the redox state of the C-terminal Cys residues leading to a modification of Crr1 function.

The distinction between the hypoxia- and copper-deficiency responses is attributed to the presence of a HyRE on the CPX1 gene but not the CYC6 gene. The HyRE also contains a GTAC core, but it cannot function as a CuRE (13). Perhaps the HyRE represents a binding site for another SBP-domain transcription factor, indicating that other domains in the transcription factor contribute to binding site selectivity in vivo.

SPB-Domain Proteins in Plants. The SPB domain is a plant-specific DNA-binding domain occurring in 16 proteins in the Arabidopsis genome. A subset of the Arabidopsis SPLs contain motifs found in Crr1: an AHA motif (found in SPL1, 7, 12, 14, and 16) that is important for activity in other transcription activators (36) at the N terminus, a region of homology linked to the SBP domain that we refer to as an “extended” SBP domain (in SPL1, 3, 7, and 12), and three ankyrin repeats (in SPL1, 12, 14, and 16). It is possible that some of these Crr1-type SPLs are regulated by trace metal nutrients like copper. A recent study suggested a role for a copper transporter of the COPT family for pollen sac development in Arabidopsis, but the target copper protein is unknown (37). It may be relevant to consider metal-containing transcription factors as targets for micronutrient dependent phenomena in plant development.