Re-Evaluation of Reportedly Metal Tolerant Arabidopsis thaliana Accessions

Santa Clara, Limeport, and Berkeley are Arabidopsis thaliana accessions previously identified as diversely metal resistant. Yet these same accessions were determined to be genetically indistinguishable from the metal sensitive Col-0. We robustly tested tolerance for Zn, Ni and Cu, and genetic relatedness by growing these accessions under a range of Ni, Zn and Cu concentrations for three durations in multiple replicates. Neither metal resistance nor variance in growth were detected between them and Col-0. We re-sequenced the genomes of these accessions and all stocks available for each accession. In all cases they were nearly indistinguishable from the standard laboratory accession Col-0. As Santa Clara was allegedly collected from the Jasper Ridge serpentine outcrop in California, USA we investigated the possibility of extant A. thaliana populations adapted to serpentine soils. Botanically vouchered Arabidopsis accessions in the Jepson database were overlaid with soil maps of California. This provided no evidence of A. thaliana collections from serpentine sites in California. Thus, our work demonstrates that the Santa Clara, Berkeley and Limeport accessions are not metal tolerant, not genetically distinct from Col-0, and that there are no known serpentine adapted populations or accessions of A. thaliana.


Introduction
Arabidopsis thaliana is widely used as a model organism in plant sciences for its high number of seeds per plant, short life cycle, small genome, and self-fertility. The first A. thaliana accession was described in the sixteenth century [1] and careful curation of collections has resulted in 5683 currently available accessions at the Arabidopsis Biological Resource Center (ABRC; abrc.osu.edu). These were collected from locations that differ in temperature, day length, precipitation levels, salinity, altitude, and biotic pressures. Local adaptation of A. thaliana populations to abiotic and biotic stresses have been widely investigated and demonstrated [2][3][4][5]. Accessions collected from locations at the extremes of the abiotic environments occupied by A. thaliana typically display adaptation to that environment. For example, accessions from locations with extreme winter temperatures exhibit low temperature tolerance [2] and alleles conferring salt tolerance are significantly more frequent in A. thaliana populations situated in marine coastal environments or near unusually saline soils [3]. metal tolerance. We now can readily sequence the whole genomes of these lines to clarify the relationships between these reportedly metal tolerant accessions and the worldwide diversity of A. thaliana.
We began a study of natural variation in heavy metal tolerance, due to our interest in identifying within species variation in this aspect of serpentine adaptation in the genetic model A. thaliana. To this end we carefully described the metal tolerance of Limeport, Santa Clara and Berkeley and characterized the genetic differences between them by whole genome resequencing. All deposited stocks for each of the accessions were obtained from the Arabidopsis Biological Resource Center (ABRC). All stocks were grown on multiple concentrations of Zn, Cu, or Ni and growth was assessed at multiple days after germination in multiple independent grow outs with multiple technical replicates for each. None of these accessions display any resistance to any concentrations of the toxic heavy metals tested when compared to Col-0. Whole genome resequencing did not uncover genetic variation missed by the 149 SNPs used previously. Thus, detailed phenotypic and genotyping characterization of these indicated that all three accessions are certainly re-deposited derivatives of the standard laboratory accession Col-0. No vouchered botanical collections of Arabidopsis were found from the Jasper Ridge serpentine. Furthermore, no additional evidence of any accessions with serpentine adaptation was found by comparing the Jepson database of botanical collections to serpentine soil distributions. We propose that these previously identified metal-adapted accessions do not provide any evidence of heavy metal adaptation in A. thaliana and that further study of these materials will not provide insight into the mechanisms of serpentine adaptation or edaphic selection at mine sites in plants.
To have similar experimental condition to Murphy et al. (1995), Arabidopsis seeds sown on plates with-half strength Murashige and Skoog medium (Murashige and Skoog, 1962; Phytotechlab, Shawnee Mission, KS; catalog no. M404) containing 0.7% agar (Sigma-Aldrich, St. Louis, MO; catalog no. A1296) in the presence or absence of different concentration of nickel (II) nitrate (Acros Organics, Morris Plain, NJ; catalog no. AC223155000), zinc (II) nitrate (J.T Baker, Phillipsburg, NY; catalog no. 4344), or copper (II) chloride (Sigma-Aldrich, St. Louis, MO; catalog no. 467847). Seeds were stratified on the plates at 4°C for 72h. Seeds were germinated and plants grown in vertical orientation on light racks in an environmentally controlled room (24°C/20°C, 12h/12h light/dark). Each plate contained five seeds each of two accessions. The accessions planted per plate were distributed randomly and each accessiontreatment combination was replicated five independent times. Comparisons between accessions and treatments were done by ANOVA using Tukey (P<0.05). The connecting letters reports summarizing the Tukey test for each metal-day combination are available in the supplemented data section (S1-S9 Tables). Botanical collection data from the Jepson Interchange database (http://ucjeps.berkeley.edu/interchange/) and serpentine soils map data were overlaid using the R statistical package by Brian Anacker at the University of California, Davis.

DNA extraction and next-generation sequencing reads mapping
For DNA extraction, A. thaliana lines were grown in the green house for several weeks and leaves were collected and frozen in liquid nitrogen. Frozen leaves were used to extract DNA for genomic sequencing using the CTAB method [16]. Illumina libraries (v4; Illumina San Diego, CA) were constructed from DNA extracted from each Arabidopsis accession. Barcoded libraries were pooled on a single lane and 275,794,696 100bp paired-end sequences were generated. To call SNPs, reads corresponding to each library were separated and the data from each accession aligned to the latest version of the A. thaliana reference genome (obtained from http:// www.phytozome.net version 9.1) [17]. Alignments were carried out using the BWA short reads aligner program (version 0.62) [18]. To expedite NGS reads mapping, we first generated the suffix array co-ordinates for each sequenced read using the BWA aln command with parameter option: "-t 8". The suffix array co-ordinates enabled the generation of sequence alignments to the A. thaliana reference genome, using the BWA sampe command with the parameters: -P -r "@RG\tID: SAMPLEID\tSM: SAMPLENAME\tPL:Illumina". The output alignments generated in the SAM NGS alignment format, by the BWA program is converted to the compressed BAM NGS alignment format using the view command in the SAMtools package (version 0.1.18) [19], with parameters: "-b -h -S". The BAM alignment files were then sorted and indexed using the SAMtools commands sort and index respectively, and with default parameters.
SNP calling was performed using a combination of the SAMtools mpileup comand and the BCFTools view program. The parameters for mpileup were "-B -Q 20 -P Illumina -C50 -uf". The output of the mpileup program was redirected from the UNIX standard output device to the BCFTools view program, using the standard UNIX/Linux pipe tool. The selected parameters for the BCFTool view command were "-vcg". The varFilter command in the SAMtools vcfutils.pl utility script (parameter option: "-D100") was used to remove SNPs derived from repetitive origins. This command also filtered SNPs with quality score less than 10 and within three bases of any gaps detected in the alignment. The UNIX grep utility program (with parameter: -v "INDEL") was then used to parse the output variant call file, to remove all predicted insertion/deletion (INDEL) variants.
The SnpSift filter command in the SnpEff software suite (version 3.1) [20] was used to filter the remaining SNPs in a stepwise manner. All SNPs with quality scores less than 20 and nonhomozygous SNP positions were eliminated using the command string "(QUAL>

SNP Effect Annotation
To annotate consequences of each SNP on gene coding capacity, we used the SnpEff program with parameter options: "eff -c", and the pre-built "athalianaTair10" SnpEff database for the A. thaliana genome (version 10), to compute the SNP impact. For SNPs overlapping the open reading frames of genes, we appended the gene function description, using the gene annotation file for A. thaliana, which we obtained from the Phytozome website (version 9.1).
To compare SNPs called in the various A. thaliana accessions, we used the BEDTools software package (version 2.17.0) [21]. To detect common (overlapping) SNPs for a pair of Arabidopsis accessions, we used the BEDTools intersect command with parameter "-wa". Accession-specific SNPs were detected using the BEDTools subtract command with parameter option "-A". To facilitate the comparison of the subset of SNPs in the coding regions of the genome, we used the BEDTools intersect command with parameter option "-wa" to obtain the SNPs overlapping the latest version of the Arabidopsis reference genome annotation (.gff3) file.

Results
To ensure that the biological materials were as similar to each other as possible, all accessions (see Materials and Methods) were grown in the same greenhouse and fresh mature seeds collected by hand. For the three reportedly metal tolerant Arabidopsis accessions (Berkeley, Santa Clara, and Limeport), we ordered all available depositions of the stocks. Seeds were desiccated and stored in microcentrifuge tubes with a hole punched in the top to maintain high seed quality and low moisture levels. Thus, all genotypes had similar seed qualities and were at similar seed ages to minimize the impact of differences in early seedling establishment on plates from confounding the analysis of metals resistance or growth rates. In addition, this produced enough seed from each accession obtained from the stock center to carry out all treatments on the same batch of seed.
None of the tested accessions, including Santa Clara, are Ni resistant Santa Clara was originally described as collected from serpentine soil in Santa Clara County, CA (http://www.arabidopsis.org). This accession has been described as a Ni tolerant ecotype of A. thaliana [6,7]. However, Santa Clara did not show any differences with Col-0 in comparative genomic analysis using 149 SNP markers [12].
To confirm the Ni tolerance of Santa Clara [6,7], both Santa Clara stocks (CS28722 and CS8069) were obtained from the stock center and seed produced. The collected seeds were grown on agar plates supplemented with a range of Ni concentrations. The reportedly Ni-sensitive accessions Limeport, Berkeley and Col-0 [6,7] were grown in the same conditions as control genotypes. Primary root measurements were taken at 15 days after germination (dag; Fig  1) as well as 10 dag and 20 dag (S1 Fig).
Root growth at a range of Ni concentrations was measured to ensure that we observed plants at physiologically relevant Ni toxicities. At concentrations above 150 μM, Ni is highly toxic for root growth resulting in inhibition over 90% (Fig 1). At 100 μM Ni and 125 μM Ni growth inhibition for all the accessions tested was between 50% and 70%, respectively (Fig 1). At 50 μM Ni and 75 μM Ni, root growth inhibition was approximately 20% and 30%,  (Fig 1). Santa Clara, Col-0, Berkeley and Limeport did not show any significant differences in primary root length under any concentration of Ni nor in control conditions. Thus, whether roots were 90% stunted or only 20% shorter than controls none of these accessions were discernably resistant, or differentially sensitive to this plant toxin relative to Col-0 controls. This indicates that Ni has the same inhibition effect in all the accession tested, and that Santa Clara is not Ni resistant.
None of the tested accessions, including Limeport, are Zn resistant The Limeport accession was collected from soil within a Zn mining site located in Friedensville, PA (http://www.arabidopsis.org) and was described as Zn resistant in the study that also reports the collection [6]. However, Limeport was genetically similar to the Zn sensitive accession Col-0 when tested at 149 SNPs [12] and contains the same stop codon in HMA3 associated with heavy metal sensitivity as Col-0 [10]. To reassess Zn tolerance of the Limeport accession we acquired both Limeport stocks (CS28464 and CS8070) available in the stock center. Mature seeds were sown on agar plates supplemented with a variety of concentrations of Zn. The reportedly Zn-sensitive Santa Clara, Berkeley and Col-0 [6] accessions were grown in parallel as controls. Primary root lengths were measured at 15 dag (Fig 2)  Zn induced~40% root growth inhibition and Zn concentrations over 400 μM were extremely toxic causing root growth inhibition over 90% (Fig 2). We conclude Limeport is not Zn resistant despite reported collection from a Zn mining site.
None of the tested accessions including, Berkeley and Limeport, are Cu resistant The Berkeley accession was collected near the UC Berkeley library in CA (http://www. arabidopsis.org) and was described as Cu resistant in the study that also describe their collection [6]. Like Limeport and Santa Clara, Berkeley was reported to be genetically similar to the Cu sensitive Col-0 at 149 SNP positions [12]. To check if Berkeley is Cu resistant, both Berkeley stocks (CS28067 and CS8068) available in the stock center were grown as described previously. Fresh seed was planted on agar-solidified media containing a variety of concentrations of Cu. The other three accessions Santa Clara, Col-0 and Limeport were grown in the same Cu toxic conditions as controls. Primary root lengths were measured at 15 dag (Fig 3) as well as 10 and 20 dag (S3 Fig). No discernable differences in root length could be detected between Berkeley, Santa Clara, Limeport and Col-0 at any Cu concentration for any duration of growth tested. Again, a physiologically relevant range of Cu toxicities were achieved in these experiments. We observed that 20 μM Cu caused~20% root growth inhibition while concentrations over 40 μM Cu were highly toxic inhibiting root growth by~60% (Fig 3). Thus, Berkeley was Cu sensitive as Limeport, Santa Clara and Col-0. It was previously showed that Santa Clara, Berkeley, Limeport and Col-0 were genetically similar using 149 SNP markers [12]. Given that thousands of accessions were genotyped it is possible that these similarities are spurious. It is also possible that mutations of adaptive significance, not included among the 149 polymorphic positions are present in these lines. The genomes of each accession were resequenced to distinguish between a spurious similarity at 149 positions, and genetic identity between Col-0, Santa Clara, Limeport and Berkeley. Alignment of reads to the A. thaliana Col-0 reference genome permitted calling of SNPs relative to Col-0. SNPs were compared between each accession pair ( Table 1). Any SNPs present in two accessions were flagged as "Common SNPs". By comparing the called SNPs in Santa Clara, Limeport and Berkeley to the SNPs called in the resequenced Col-0 we were able to identify all putative sites of divergence for these accessions, and distinguish them from errors in the reference genome or error prone positions. We identified between 855 and 892 SNPs between each accession and the digital reference genome. These are orders of magnitude less variation than identified between accessions [22][23][24][25]. When comparing each line pair-wise, we observed that the numbers of shared SNPs were similar whether we compared between the two stock numbers for the same accession and/or between accessions for Santa Clara, Berkeley or Limeport (Table 1). Surprisingly, the sequence of Col-0 was similarly different from the reference genome and the same SNPs were often present in Col-0 as in the other accessions. This indicated that the source of variation within Limeport, Santa Clara and Berkeley was similar to that in Col-0 and likely within the margins of To obtain better estimates of the base substitution rate between lines, we identified all SNPs present in both stocks of an accession and not contained in Col-0. Fig 4 summarizes the shared alleles between the four genotypes as a Venn diagram, and Table 2 contains SNP counts for the most informative categories. If these lines were resistant to different metals, the alleles responsible for these differences would have to be unshared, so we also determined the SNPs present in both sequencing experiments for a genotype and absent in any other accessions. These are  referred to as the unique SNPs for each accession (Table 2). These comparisons indicated that the variation between Santa Clara, Limeport, Berkeley and Col-0 is low (Table 2). Of these, 653 are shared between all accessions (Fig 4), 20 times more than the number of unique SNPs per accessions (Table 2). Further, the number of common SNPs private to any three or two-way comparison is much lower than the common SNPs shared among all the accessions tested. Given that Col-0 itself exhibits 24 SNPs private to that line this indicates that the variability between Limeport, Santa Clara and Berkeley is likely to be insignificant and close to the error rate of sequencing.
Nonsynonymous variation is highly unlikely to affect metal tolerance in Santa Clara, Limeport and Berkeley genomes We showed that Berkeley, Santa Clara, Limeport and Col-0 are nearly identical by whole genome sequencing and suffer similar inhibition of root growth in response to Ni (Fig 1 and S1  Fig), Zn (Fig 2 and S2 Fig) and Cu (Fig 3 and S3 Fig). Since only one sequence change is needed to provide metal resistance we scanned the putative changes for non-synonymous changes to genes associated with antioxidant mechanisms and metal transport. The unique non-synonymous SNPs for each accession were identified. Only one non-synonymous SNP was unique to Limeport (S10 Table), four to Santa Clara (S11 Table) and two to Berkeley (S12 Table). None of these fell in genes associated with known metal tolerance mechanisms, consistent with the observation of no alterations to metal resistance in any of the tests performed here (Figs 1-3 and S1-S3 Figs).

Botanical samples of A. thaliana in California
Given the general interest in serpentine adaptation, A. thaliana in general, and the Jasper ridge serpentine in particular, we considered that evidence of A. thaliana serpentine adaptation might be found using the online Jepson collection http://ucjeps.berkeley.edu/. We expect that A. thaliana growing in California serpentine sites would be evident in the online Jepson collection. No botanically vouchered Arabidopsis collections made on the Jasper ridge serpentine were present in the Jepson database (Fig 5). Of the hundreds of data based collections of A. thaliana made in California, only one was taken from within 400 meters of serpentine soil. The one accession closest to a serpentine soils site (http://ucjeps.berkeley.edu/cgi-bin/new_detail. pl?accn_num=RSA781117) was taken from a gravel trail that runs through the known serpentine site at Salmon Falls in El Dorado County in North Central California. Thus, the digitized botanical collections failed to identify any other record of an A. thaliana population growing on serpentine soil in Santa Clara [6] and provides no evidence for serpentine adaptation for any of the hundreds of A. thaliana populations recorded in California.

Discussion
A. thaliana displays local adaptation to a variety of environments across a wide geographic range [2][3][4][5]. The genetic divergence between accessions is correlated with the distance between collection sites. Accessions that grow in close locations are genetically similar, but accessions that grow in distant location are genetically different between each other [12,26]. It was previously observed, that Santa Clara, Berkeley, Limeport and Col-0 are genetically indistinguishable but collected from distant and distinct habitats [12]. Using whole genome sequencing, we observed negligible variability between Santa Clara, Limeport, Berkeley and Col-0 (Tables 1  and 2, Fig 4). Santa Clara was described as collected from serpentine soil. This is the only A. thaliana taken from an extreme soil environment with any indication that it is able to survive normally toxic levels of Ni [6,7]. In our reanalysis, we failed to detect any greater tolerance to any concentration of Ni than was seen in Col-0 (Fig 1). Furthermore, no genetic variation between Santa Clara and Col-0 was likely to affect metal tolerance under untested conditions (S11 Table). It is important to note that we utilized the Santa Clara accessions available from the Arabidopsis Biological Resource Center, whereas Agrawal et al. [7] cited the Lehle Seeds company as their source for Santa Clara. Limeport was collected from Zn rich mine soil and was characterized as a Zn tolerant genotype [6]. Across Ni, Zn, and Cu doses and treatment times, Limeport responded identically to the Ni, Zn and Cu-sensitive Col-0 demonstrating that Limeport is not Zn tolerant (Fig 2). Further, no genetic variation associated with metal tolerance was found when Limeport was compared to Col-0 (S10 Table). Berkeley was initially described as Cu-tolerant [6], but then later described as Cu sensitive by the same authors [27]. Although, both articles grew Arabidopsis on vertically oriented chromatography paper saturated with Murashige -Skoog medium and supplemented with different metal concentrations, the authors explained that the different response of Berkeley to Cu occurred for "significant differences in the growth assay" [6,27]. We demonstrated not only that Berkeley is genetically near-identical to Col-0 but also that Berkeley is Ni, Zn, and Cu sensitive just like Col-0 when exposed to increasing toxic concentrations of the metals (Fig 3).
Based on the sequencing data alone, we conclude that Santa Clara, Limeport, Berkeley and Col-0 are the same accessions. The identical physiological behaviors of these lines under different concentrations of toxic metals and control conditions are consistent with this finding. We strongly suggest that these accessions not be used to study natural variation in metal adaption or any other study. A. thaliana does show differences in Ca/Mg ratio tolerance [28,29] as well genetically modified lines are tolerance to abiotic stresses including heavy metals [30][31][32][33]. Nevertheless, based on collection locations for A. thaliana present in the Jepson database and their non-alignment with serpentine soil sites, we suggest that A. thaliana is unlikely to display natural variation across this edaphic range. A. thaliana is not the right model organism to study genetic adaptation to these unusual and fascinating soils. Detailed mechanistic studies of the myriad serpentine adapted taxa in experimentally tractable genera such as Aquilegia, Mimulus, Thlaspi, Alyssyum, or Noccaea [34][35][36][37][38][39] remains the best route to understand the basis of serpentine adaptation in plants.   Table. Connecting letters report for copper treatment at day 20. (DOCX) S10 Table. All non-synonymous changes unique to the Limeport accessions. (DOCX) S11 Table. All non-synonymous changes unique to the Santa Clara accessions. (DOCX) S12 Table. All non-synonymous changes unique to the Berkeley accessions.