Amino Acid Polymorphisms in Strictly Conserved Domains of a P-type ATPase HMA5 are Involved in the Mechanism of Copper Tolerance Variation in Arabidopsis 1

Copper (Cu) is an essential element in plant nutrition, but it inhibits growth of roots at low concentrations. Accessions of Arabidopsis ( Arabidopsis thaliana ) vary in their tolerance to Cu. To understand the molecular mechanism of Cu tolerance in Arabidopsis, we performed QTL analysis and accession studies. One major QTL on chromosome 1 (QTL1) explained 52% of the phenotypic variation in Cu tolerance in roots in a L er /Cvi recombinant inbred population. This QTL regulates Cu translocation capacity and involves a Cu-transporting P 1B-1 -type ATPase, HMA5. The Cvi allele carries two amino acid substitutions in comparison with the L er allele and is less functional than the L er allele in Cu tolerance when judged by complementation assays using a T-DNA insertion mutant. Complementation assays of the ccc2 mutant of yeast using chimeric HMA5 proteins revealed that N923T of the Cvi allele, which was identified in the tightly conserved domain N(x) 6 YN(x) 4 P (where the former asparagine was substituted to threonine), is a cause of dysfunction of the Cvi HMA5 allele. Another dysfunctional HMA5 allele was identified in Chisdra-2 (Chi-2), which showed Cu sensitivity and low capacity of Cu translocation from roots to shoots. A unique amino acid substitution of Chi-2 was identified in another strictly conserved domain, CPC(x) 6 P, where the latter proline was replaced with leucine. These results indicate that a portion of the variation in Cu tolerance of Arabidospsis is regulated by functional integrity of the Cu-translocating ATPase, HMA5, and in particular to the amino acid sequence in several strictly conserved motifs.


Introduction
Copper (Cu) is an essential element for higher plants and plays key roles in a series of major biological systems, such as respiration, photosynthesis and ethylene signaling (see Capaldi, 1990;Maksymiec, 1997;Clemens, 2001). In most of these systems, Cu is involved in electron transfer reactions that are mediated by Cu-containing proteins such as plastocyanin, which functions in the electron transfer system of photosystem I, and cytochrome c oxidase (E.C. 1.9.3.1), which catalyzes terminal oxidation in the mitochondrial electron transfer chain. The ability of Cu to mediate high rates of electron transfer is also a potential cause of toxicity for plant cells (especially in growing roots). Less than 5 μM Cu inhibits growth of roots in wheat (Triticum aestivum;Taylor et al., 1991, Parker et al., 1998 and Arabidopsis (Arabidopsis thaliana; Toda et al., 1999). This sensitivity of growing roots is a potential risk in agriculture that arises from the use of Cu-containing fungicides and fertilizers. For example, bordeaux mixture, which contains Cu (without other synthesized organic chemicals) is used in both conventional and organic agriculture (e.g. Semu and Singh, 1995). In addition, Cu accumulation in soil is expected to arise from continuous application of organic fertilizers such as pig manure, since supplemental Cu is used to improve growth rates of pigs (e.g. Coffey et al., 1994). Thus, establishment of breeding programs to develop Cu tolerant germplasm could be important for use in sustainable agriculture systems. Molecular breeding (e.g. marker assisted selection) is a promising approach that could be used if mechanisms of variation in Cu tolerance are clarified at the molecular level.
Mutant studies in yeast have revealed that free Cu concentrations in the cytosol are strictly regulated by a complex Cu homeostasis system, consisting of Cu transporters (e.g. Ccc2p;Fu et al., 1995), Cu binding proteins (e.g. Cox17p;Glerum et al., 1996) and Cu chaperone (Ccsp;Culotta et al., 1997). Functional analyses of Arabidopsis homologues suggest that plants possess similar homeostasis mechanisms (e.g. COPT1; Kampfenkel et al., 1995, CCH;Himelblau et al., 1998, RAN1;Hirayama et al., 1999, CCS;Chu et al., 2005;HMA5;Andrés-Colás et al., 2006). Because free Cu in the cytoplasm is regulated by these systems (Clemens, 2001), variation in Cu tolerance among varieties might be regulated by the differential capacity of each component of the homeostatic system. In fact, delivery of Cu to plastocyanin and Cu/Zn superoxide dismutase via P-type ATPases that regulate free Cu in the cytosol (Abdel-Ghany et al., 2005) has been suggested as one mechanism of variation in Cu tolerance in shoots of Arabidopsis. Although several genes that regulate Cu homeostasis in roots have been identified [e.g. methollothionine synthase in the metal tolerant plant, Silene vulgaris (Moench) Garcke (van Hoof et al., 2001); HMA5 (Andrés-Colás et al., 2006)], their involvement in determining the variation in Cu tolerance among varieties has not been verified.
Studies of natural variation within Arabidopsis provide a useful approach to understand the mechanisms of variation in target traits, (see for review; Koornneef et al., 2004). Several critical genes regulating traits such as freezing tolerance (Alonso-Blanco et al., 2005), growth and flowering (Balasubramanian et al., 2006) have been successfully identified by this approach. We have applied this approach to identify chromosome 3 with a LOD score of 3.2 (designated as QTL3), its contribution to the variation in phenotypic Cu tolerance was smaller than QTL1. QTL1 explained 52% of the variation in phenotypic Cu tolerance, while 15% of the variation was explained by QTL3. These QTLs could possibly account for the observed bimodal distribution of Cu tolerance across the RILs.
To further characterize the genetic architecture of Cu tolerance in the Ler/Cvi RI population, epistatic interacting loci pairs were searched by a complete pair-wise search using the Epistat program (Chase et al., 1997), which allows the detection of epistatic interacting loci pairs even if each locus was not detected significantly by the composite interval mapping (CIM) method. The QTL1 region (linked to GD.160C) interacted with two other chromosome regions, namely chromosome 1 (linked to DF.93C; "epistasis B" in Fig. 2A) and 5 (linked to CC.262C; "epistasis C" in Fig. 2A), while the QTL3 region (linked to GD.296-Col) interacted with another chromosome 1 region (linked to DF.73L; "epistasis A" in Fig. 2A), respectively (Table I). Mean of allele combinations indicates that negative additive effect of Cvi allele at QTL1 is partially cancelled by Ler allele on chromosome 5 ( Fig. 2 and Table I). On the other hand, there were 6 RILs that showed transgressive segregation (0 of sensitive, 6 of tolerant; LSD = 29, P < 0.05). This might be partially explained by one of the combined allelic combinations (LC) of "epistasis C" ( Table I).

Characterization of QTL1 in Cu tolerance
QTL1 showed the largest R 2 value (0.52) indicating that this locus contains the most important genetic factor determining variation in Cu tolerance among the Ler/Cvi RILs.
To identify the gene(s) accounting for QTL1, a candidate gene search and physiological characterization were performed using RILs contrasting in Cu tolerance and with different alleles (i.e. Cvi and Ler alleles at QTL1). QTL1 was flanked by two genetic markers, namely GD.160C and HH.375L ( Fig. 2A). Although the physical positions of these genetic markers in the Ler/Cvi RILs have not been reported, the markers can be positioned using the genetic markers shared by Ler/Col whose physical position is assigned to the genomic DNA sequence of Col (Alonso-Blanco et al., 1998). Genetic markers DF.408C-Col and CH.215L flank QTL1 and are also closely flanked by genetic markers GD.160C and HH.375L, which are shared with Ler/Col mapping population. These genetic markers are closely related to other genetic markers, namely ATHGENEA and m305, which correspond to two genes on chromosome 1, At1g60840 and At1g64180, respectively (TAIR; http://www.arabidopsis.org/servlets/sv) (Fig. 2B). From these estimations, we believe that the major gene(s) regulating QTL1 locate(s) between the At1g60840 and At1g64180 region. To further characterize the QTL1 locus, we developed four dCAPs markers at At1g62660, At1g62810, At1g63010 and At1g63760, which were used to evaluate homozygous Cvi and Ler alleles at given positions, and re-perform the CIM analysis. By this analysis, QTL1 is flanked by At1g60840 and At1g63760 (correspond with HH.375L; Fig. 2B) and it is closely linked to At1g63010 with 0.50 R 2 value. This indicates that the major gene causing QTL1 in the Ler/Cvi RILs is localized in this physical position. This region contains a gene previously reported to regulate Cu homeostasis in roots, namely HMA5, which functions as a Cu translocator (At1g63440; Andrés-Colás et al., 2006). To test the possibility that this gene could be the cause of QTL1, we analyzed metal translocation capacity of Cvi and Ler alleles of QTL1. RILs with the Cvi allele for QTL1 showed lower Cu translocation capacity (i.e. lower shoot/total ratio of Cu content) than those carrying the Ler allele. These differences in Cu translocation capacity were related to Cu tolerance as assessed by RRL ( Fig. 3; R 2 = 0.82). There were no significant differences in Zn and Mn translocation capacity between RILs (Fig. 3). This profile of metal translocation was similar to that of the KO line of HMA5 (see below Supplemental Fig. S5). From these results, we inferred that the Cu sensitivity of the Cvi allele at QTL1 was correlated with low Cu translocation capacity.

Evaluation of Cvi and Ler HMA5 alleles by in planta and yeast complementation assays
Homozygous HMA5-KO lines, SALK_040252 carrying a T-DNA insertion at the first intron showed hypersensitivity to Cu, while showing similar growth to that of WT (Col-0) in control solution (Fig. 4) shared the same alleles at epistatic interacting loci, but different alleles at QTL1 (i.e., Cvi or Ler). As shown in Fig. 4, the parental accession Cvi grew poorly in Cu toxic solutions, similar to HMA5-KO. Both F 1 plants derived from crosses between HMA5-KO and RILs carry Cvi or Ler alleles at QTL1 grew similar in control solution, whereas HMA5-KO×Cvi F 1 showed short roots like Cvi (RRL of F 1 ; 25.6%, Cvi; 20.5%) and that of HMA5-KO×Ler F 1 was better than Cvi (RRL of F 1 ; 72.3%, Ler; 92.7%) in Cu toxic solution (Fig. 4). These results indicate that the differences in the HMA5 allele are a cause of QTL1. HMA5 expression was similar between the parental lines or two RILs used for the complementation test (Supplemental Fig. S2), suggesting that differential Cu tolerance between Cvi and Ler alleles was not a result of differential gene expression, but might instead be due to a decrease in the ability of the Cvi protein to function.
Sequencing analysis identified two amino acid substitutions in the deduced amino acid sequence of Cvi-and Ler-HMA5 ( Fig. 5A and 6A). The first substitution was at the 178 th amino acid residue [Ler:Ser (TCG) to Cvi:Lys (TTG)], while the second substitution was found at 923 rd amino acid residue [Ler:Asn (AAC) to Cvi:Thr (ACC)] (Fig. 5A). To test the effect of these substitutions, we conducted yeast (Saccaromyces cerevisiae: Δccc2 mutant) complementation assays with chimeric and authentic HMA5 proteins. As described previously, the ccc2 mutant (Δccc2) cannot grow under iron-limited conditions because in the absence of the Ccc2 protein (a Cu-transporting P-type ATPase), Cu cannot be delivered to the multicopper oxidase Fet3p, which is required for high-affinity iron uptake at the plasma membrane. We Ler-HMA5 cDNA (LerLer) partially complemented the ccc2 mutation on agar plates, while that of Cvi-HMA5 (CviCvi) did not (Fig. 5B). Growth of ccc2 mutant carrying Ler-HMA5 (LerLer) in liquid medium was about half that of the parental strain (BJ2168), but significantly greater than with Cvi-HMA5 (CviCvi; Fig. 5C). In this condition, replacement of 178S to 178L (designated as CviLer in Fig. 5) caused no significant change compared to Ler-HMA5. On the other hand, replacement of 923T to 923N gave complementation in the Cvi-HMA5 type HMA5 protein (Fig. 5). These results indicate that the difference at the 923 rd amino acid residue (N to T) could account for the phenotypic differences associated with the Ler-and Cvi-HMA5 alleles and was in turn detected by the major QTL in the Ler/Cvi RI population.

HMA5 polymorphism and Cu tolerance among accessions of Arabidopsis
We assessed the Cu tolerance of 103 Arabidopsis accessions (JA series and parental accessions of RILs), and found that Cu tolerance among RRL ranged between 16.7 to 88.6% (RRL). To determine whether variation in HMA5 is involved in the mechanism of Cu tolerance variation among these accessions, we compared the deduced amino acid sequence of HMA5 among Cu tolerant-and sensitive-accessions.
Sequence analysis of DNA from selected Cu tolerant and sensitive accessions (total of 40) identified 7 polymorphic sites at various positions, which can be grouped into 7   Table S1). Almost all of accessions that belonged to the haplotype 1 showed Cu tolerance (i.e. greater than the average of 103 accessions) (Fig. 6A) and greater Cu translocation capacity than HMA5-KO (Supplemental Fig. S5). One Cu sensitive accession (Bu-14) belonged to haplotype 1, but its sensitivity was complemented by HMA5-KO and the shoot/total ratio of Cu in Bu-14 was significantly higher than that of HMA5-KO (Fig. 6C). This observation further supports that Ler HMA5 allele, haplotype 1, is a Cu tolerant allele in Arabidopsis.
The remaining 22 accessions consisted of 9 sensitive (including Cvi-1) and 13 tolerant accessions. These accessions showed various amino acid polymorphisms at 7 different positions, which were grouped into 6 other haplotypes ( Fig. 6A and B).
Haplotypes 2-5 were shared by both tolerant and sensitive accessions, suggesting these substitution(s) may not alter HMA5 functional capacity and Cu tolerance (Fig. 6B). In fact, the phenotype of the most Cu sensitive accession of haplotype 2 (Hn-0), 3 (Wil-3) and 4 (Tu-0) were complemented by HMA5-KO in the F 1 generation (Fig. 6C). Also, their Cu translocation capacities were significantly greater than that of the HMA5-KO ( Fig. 6C). This indicates that Cu sensitivity of these accessions is not caused by decreased HAM5 function and not associated with reduced Cu translocation. On the other hand, haplotype 6 and 7 were specific to sensitive accessions, Cvi-1 and Chi-2, respectively ( Fig. 6A HMA5-KO (F 1 progeny 20.4%, Fig. 6C), but was complemented by Col-0 (F 1 progeny 98.3%, Supplemental Fig. S4), indicating that Chi-2 allele is another Cu sensitive allele of HMA5. In this case, the Cu translocation capacity of Chi-2 was significantly lower than other accessions and was the same level as that of the HMA5-KO (Fig. 6C), while Zn and Mn translocation capacity were the same among all accessions (Supplemental

Characterization of amino acid substitutions on a motif and domain map of HMA5
Previous studies of mutations in Cu-transporting ATPases indicated that substitution(s) in or adjacent to essential motifs in conserved domains caused disruption of protein function and in turn a mutant phenotype (e.g. ran1 and paa1 mutations in Arabidopsis; Hirayama et al., 1999;Shikanai et al., 2003, Supplemental  among the P 1B-1 -ATPases of a variety of organisms (see 20 organisms in Fig. 7B and C, in which conserved percentages ranged from 5 to 45%; Fig. 7A). On the other hand, both of the less functional HMA5 alleles (Cvi-1 and Chi-2; Haplotypes 6 and 7) carried a substitution in an amino acid that is 100% conserved among reported P 1B-1 -ATPases substitution (unique to Chi-2) and the N923T substitution (unique to Cvi-1) are located in the amino acid motifs [CPC(x) 6 P] (the latter P to L) and [N(x) 6 YN(x) 4 P] (the former N to T), which are located in the 6 th and 7 th transmembrane domains, respectively (Fig.   7).

Impact of dysfunctional HMA5 on the growth of Arabidopsis in Cu contaminated soil
We have shown that HMA5 is a critical factor in the Cu tolerance of roots grown in hydroponic culture. To test whether this has impact on other growth conditions, we grew contrasting Cu tolerance RILs, and a set of HMA5-KO (Col background) and Col on soil artificially contaminated with Cu. In control and lower contaminated Cu soil, no differences were observed among all tested accessions. However, growth of sensitive RILs (i.e. Cvi HMA5 allele) and HMA5-KO was more affected by Cu on highly contaminated Cu soil than tolerant RILs or Col (Fig. 8). These results indicated that HMA5 is a critical gene for Cu tolerance in Cu contaminated soil. Molecular biological studies, mainly in yeast, have led to the development of hypotheses concerning the mechanisms for Cu sensitivity in relation to Cu homeostasis (Puig et al., 2007). The roles of similar mechanisms that might contribute to natural variation in plant tolerance to Cu, especially for root growth, have remained unclear. We have performed QTL and accession analyses using RRL as a specific index of Cu tolerance and found that amino acid polymorphisms in a single major gene can account for the Cu sensitivity of some accessions. We found that the Ler genotype at QTL1 confers Cu tolerance (greater than Cvi) due to its higher capacity for Cu translocation from roots to shoots. This capacity for metal translocation was not observed for Mn and Zn (Fig. 3). This is consistent with the proposed function of HMA5 (our candidate gene collocating in QTL1), which detoxifies Cu in the root by translocation of Cu from roots to shoots (Andrés-Colás et al., 2006). The conclusion that HMA5 plays critical role in QTL1 was supported by a series of complementation assays that utilized yeast mutant ccc2 (Fig. 5) and the Arabidopsis HMA5-KO (Fig. 4).

Discussion
HMA5 belongs to a large family of genes encoding P 1B -type cation transporting ATPases that transport a variety of cations. Several amino acid residues are highly conserved among or within sub-family members (Argüello, 2003). Interestingly, previous mutant studies with Cu and other metals indicate that substitutions in highly conserved amino acid negatively affect the transport capacity of these proteins. In fact, Arabidopsis mutants with defects in Cu-transporting ATPases, namely ran1-1 substitutions at amino acid residues that are conserved in Cu-transporting ATPases. In our study, the less functional HMA5 alleles, Cvi-1 and Chi-2, which have lower capacity for Cu translocation (Fig. 6C), carry amino acid substitutions in the consensus amino acid residues shared by all P 1B-1 -type cation transporting ATPases (Argüello, allele showed a reduced capacity to complement an yeast ccc2 mutant. Our yeast complementation tests (Fig. 5) suggest that the amino acid substitution located in the N(x) 6 YN(x) 4 P motif (N923T, found in CviCvi and LerCvi) was responsible for the Cvi phenotype ( Fig. 4 and 6C). This is further supported by the observation that the other substitution, S178L, is also found in tolerant accessions (Fig. 7A). Substitutions in strictly conserved amino acids in the CPC(x) 6 P motif (Fig. 7B)  In the present study, we have applied an association analysis between HMA5, a candidate gene for QTL1, and Cu tolerance among natural accessions. We have sequenced 40 accessions that contained almost all of the sensitive accessions identified from a larger population (103 accessions). This allowed us to identify seven haplotypes in the amino acid polymorphism of HMA5, including two sensitive alleles from Cu hyper sensitive accessions (i.e. Cvi and Chi-2; haplotypes 6 and 7). Because these two alleles were both unique among sequenced accessions, we could not identify the cause of Cu hyper sensitivity of these dysfunctional alleles directly by this approach. As     length was measured using a video microscope as described by Toda et al (1999).

Arabidopsis accessions
Average root length values were calculated using the top three root lengths in the each treatment. Relative root length (RRL) was then calculated following the formula; RRL= mean root length in Cu solution/means of the root length in control solution (%). This experiment set was replicated three times and the averages of those three RRL measurements were used for QTL analyses and accession study.

QTL analyses
QTL analyses, consisting of the composite interval mapping (CIM) and complete pair-wise search for epistasis, were conducted as described by Kobayashi et al., 2007b. Briefly, genetic linkage maps for each population were constructed with Mapmaker/EXP version 3.0b (Lander et al., 1987: obtained from http://www.broad.mit.edu/genome_software/) using segregation data for each RIL obtained from databases (167 markers in the Ler/Cvi RILs: obtained from NASC; http://arabidopsis.info/). In addition to these public markers, we constructed and used a new dCAPS marker in the QTL1 (http://helix.wustl.edu/dcaps/dcaps.html. Fig. 2B  Doerge 1994) with 1000 permutations at the permutation significance level for α = 0.05.
Epistatic interactions between any two molecular markers were determined using a complete pair-wise search method with a significance threshold of P < 0.0005 using EPISTAT (Chase et al., 1997). This method can identify significant epistatic interacting marker pairs even if the marker has not been detected as a significant QTL by the CIM method. Broad-sense heritability (h b 2 ) was estimated using RRL values derived from repeated experiments by the following formula; h b 2 = σ 2 g / {(σ 2 e /r) + σ 2 g } where σ 2 g is the genetic variance, σ 2 e is the environmental variance and r is the number of data employed (n = 3).

Determination of translocation of Cu and other ions
Three sets of 50 seedlings of each line were pre-grown for 10 days in control solution containing P i , then transferred to a solution containing 1.3 μM Cu for 2 d. Seedlings were then harvested, rinsed in distilled water, and roots and shoots were separated using a blade. Content of 63 Cu, 55 Mn and 66 Zn were measured by inductively coupled plasma-mass spectrometry (ELAN 6000; PerkinElmer Japan, Tokyo, Japan) as described by Kobayashi et al. (2007b). Translocation capacity was defined as metal content in the shoot/total metal content (shoot + root) and determined for the parental accessions in the Ler/Cvi RI population (Ler-2 and Cvi-1),

DNA sequencing and protein sequence analyses
Genomic DNA was isolated from leaves as described by Kobayashi et al. (2007b).
The HMA5 region was sequenced by direct sequencing procedure using the ABI BigDye Terminater System (ver 3.1) and an ABI PRISM3100 DNA sequencer according to the manufacture's manual. Primers used for this process are shown in Supplemental Table S3. Using the Col-0 sequence as a reference, the complete coding sequence (CDS) of HMA5 was obtained for each accession. Amino acid sequences of previously characterized P 1B-1 -ATPase genes were obtained from a P-type ATPase database (http://www.patbase.kvl.dk./), and sequences reported by Argüello (2003) and Williams and Mills (2005). Multiple amino acid alignment was performed using the

Yeast complementation test for Cvi and Ler HMA5 proteins
There are two amino acid substitutions between Ler-2 and Cvi-1 at 178 th and 923 rd (Ler-2 S and N; Cvi-1, L and T). cDNA encoding authentic and chimeric protein, downstream of the GAL1 promoter (i.e driven by galactose but not glucose). Chimeric cDNA were made from cDNA of each accession using appropriate sites for the restriction endo nucleases. The vectors were transformed to the ccc2 mutant and BJ2168 wild type yeast, respectively, which we used previously to characterize BnRAN1 (a kind of Cu transporting ATPase in Brassica napus; Southron et al., 2004). Using pYES3-BnRAN1 as a control, growth capability on Fe-limited medium (which reflects CCC2 complementation capacity) was estimated as described by Southron et al. (2004).
Each experiment was performed twice and representative results were shown.

Complementation test of HMA5 alleles in planta
F 1 seeds were obtained from crosses between individual accessions and HMA5-KO or Col-0. Both HMA5-KO and Col-0 were used as the female parent and the other accessions were used as the male parent. F 1 seeds from these crosses were used in root elongation assays as described above and then rescued for genotyping to confirm cross fertilization.     Table I)