Diverse phosphate and auxin transport loci distinguish phosphate tolerant from sensitive Arabidopsis accessions

Abstract

Phosphorus (P) is an essential element for plant growth often limiting agroecosystems. To identify genetic determinants of performance under variable phosphate (Pi) supply, we conducted genome-wide association studies on five highly predictive Pi starvation response traits in 200 Arabidopsis (Arabidopsis thaliana) accessions. Pi concentration in Pi-limited organs had the strongest, and primary root length had the weakest genetic component. Of 70 trait-associated candidate genes, 17 responded to Pi withdrawal. The PHOSPHATE TRANSPORTER1 gene cluster on chromosome 5 comprises PHT1;1, PHT1;2, and PHT1;3 with known impact on P status. A second locus featured uncharacterized endomembrane-associated auxin efflux carrier encoding PIN-LIKES7 (PILS7) which was more strongly suppressed in Pi-limited roots of Pi-starvation sensitive accessions. In the Col-0 background, Pi uptake and organ growth were impaired in both Pi-limited pht1;1 and two pils7 T-DNA insertion mutants, while Pi -limited pht1;2 had higher biomass and pht1;3 was indistinguishable from wild-type. Copy number variation at the PHT1 locus with loss of the PHT1;3 gene and smaller scale deletions in PHT1;1 and PHT1;2 predicted to alter both protein structure and function suggest diversification of PHT1 is a key driver for adaptation to P limitation. Haplogroup analysis revealed a phosphorylation site in the protein encoded by the PILS7 allele from stress-sensitive accessions as well as additional auxin-responsive elements in the promoter of the “stress tolerant” allele. The former allele’s inability to complement the pils7-1 mutant in the Col-0 background implies the presence of a kinase signaling loop controlling PILS7 activity in accessions from P-rich environments, while survival in P-poor environments requires fine-tuning of stress-responsive root auxin signaling.

A series of insertion/deletion nucleotide polymorphisms at PHOSPHATE TRANSPORTER1 and PIN-LIKES7 loci confer natural variation in low phosphate tolerance in 200 Arabidopsis accessions.

Introduction

Phosphorus (P) is an essential macronutrient for plant growth and development. However, in most soils, the concentration of phosphate (Pi) is limiting due to its low solubility and mobility (Raghothama, 1999). Pi deficiency commonly impairs plant growth and affects ∼70% of cultivated land globally (Hinsinger, 2001; Cakmak, 2002; López-Arredondo et al., 2014). Thus, P-containing chemical fertilizers are essential to sustain sufficient plant growth, as well as high grain quality and yield in most agroecosystems. Plants have evolved an array of strategies to cope with variable Pi environments, including remodeling of root system architecture (RSA) and metabolic adjustments (Plaxton and Tran, 2011; Shahzad and Amtmann, 2017; Bhosale et al., 2018). Auxin plays an important role in altering RSA. Application of auxin mimics the root’s response to low Pi supply, with a shorter primary root, and increased lateral root and root hair elongation (Bates and Lynch, 1996; Nacry et al., 2005). This phenotype was abolished in the auxin signaling mutants auxin resistant (axr)1-7, axr2-1 and axr4-1, auxin response factor (arf)7, arf19, and transport inhibitor response (tir)1 (Nacry et al., 2005; Perez-Torres et al., 2008; Huang et al., 2018). Plants grown under P-limiting conditions not only accumulate more auxin in roots but are also more responsive to auxin (López-Bucio et al., 2002; Nacry et al., 2005; Perez-Torres et al., 2008).

The molecular mechanisms by which plants respond to Pi limitation are well studied. They consist of MYB transcription factor (TF) PHOSPHATE STARVATION RESPONSE1 (PHR1), microRNA MIR399, E2 ubiquitin conjugate PHOSPHATE2 (PHO2), Pi exporter PHOSPHATE1 (PHO1), and the PHOSPHATE TRANSPORTER1 (PHT1) family—often referred to as the PHOSPHATE (PHO) regulon (Muchhal et al., 1996; Rubio et al., 2001; Hamburger et al., 2002; Aung et al., 2006). Auxin affects Pi starvation signaling by regulating the expression of PHR1, which is the transcriptional master regulator of Pi starvation response (PSR) (Huang et al., 2018). Exogenous auxin application induces PHR1 expression while auxin transport inhibitors suppress it. Pi transporters are responsible for Pi acquisition from the environment and translocation between organs, cell types, or organelles. In Arabidopsis (Arabidopsis thaliana), nine plasma membrane located PHT1 family members have been characterized, and at least eight of them are expressed in P-limited roots (Mudge et al., 2002; Shin et al., 2004). PHT1;1, PHT1;2, PHT1;3, and PHT1;6 are colocated in a gene cluster on chromosome 5 indicating a series of recent gene duplication events (Ayadi et al., 2015). Among PHT1 genes, PHT1;1 shows the highest expression in P-replete organs, suggesting an important role in bulk Pi uptake. The pht1;1 mutant shows a 60% reduction of Pi uptake by P-replete roots. Compared to PHT1;1, both PHT1;2 and PHT1;3 have lower transcript abundance in Pi-rich media but are highly transcribed under Pi starvation (Shin et al., 2004). PHT1;2 and PHT1;3 together contribute ∼30% of the Pi uptake in P-limited roots, while PHT1;1 contributes between 15% and 20% (Ayadi et al., 2015).

Genome-Wide Association Studies (GWAS) are a powerful tool for identifying genetic variants associated with phenotypic traits. Several GWAS have investigated traits related to plant nutrition (Rosas et al., 2013; Kawa et al., 2016; Satbhai et al., 2017; Kisko et al., 2018; Bouain et al., 2019; Jia et al., 2019). PHO1 was associated with root plasticity in heterogeneous environments, impacting the distribution of lateral roots along the primary axis (Rosas et al., 2013). Given the importance of the PHO regulon in regulating Pi acquisition and use in the A. thaliana reference accession Col-0, it is quite surprising that no natural genetic variation in PHT1 or PHO1 transporters has been directly associated with adaptation to variable P environments thus far.

In this study, we investigate the genetic basis of the response to changes in Pi availability among 200 highly diverse ecotypes of the model plant A. thaliana. Focusing on physiological (organ biomass and primary root length) and metabolic traits (organ Pi, shoot anthocyanin, and shoot elemental composition), we test the expectation that variation in PSR is primarily mediated by allelic variation at transporter-encoding loci. We leverage the power of GWAS combined with haplogroup (Hap) structure analysis and functional validation to establish PHT1 and PILS7 as important loci underlying natural variation in low Pi tolerance.

Results

Arabidopsis thaliana accessions display highly diverse responses to Pi withdrawal

The accessions used in this study were selected based on a previous study (Li et al., 2010) to minimize genetic redundancy and family relatedness of accessions and to ensure maximum genetic diversity within the population (Supplemental Table S1). We first assessed the impact of varying Pi supply on critical growth parameters using developmentally synchronized seedlings (“Materials and methods”; Supplemental Figure S1). We observed reductions in organ biomass, organ Pi concentration, and primary root growth as well as anthocyanin accumulation in P-limited shoots (Figure 1;Supplemental Tables S2 and S3) and altered shoot elemental composition (Supplemental Figure S2A; Supplemental Table S2). Within these general trends, accessions showed large qualitative and quantitative differences in the degree of PSR (Figure 1;Supplemental Tables S2 and S3). For instance, on average, Pi withdrawal resulted in a 42% reduction of shoot fresh weight, while it only conferred a 12% reduction of root fresh weight. Pi concentrations in P-limited organs ranged from 0.73 to 4 µmol g⁻¹ fresh weight (FW) in shoots, and from 0.6 to 5.3 µmol g⁻¹ FW in roots (Figure 1, C and D). Total P (i.e. the sum of inorganic and organic P) concentration in P-limited shoots ranged from 3 to 30 µmol g⁻¹ FW (Supplemental Figure S2A; Supplemental Table S2). Large variation in shoot anthocyanin concentration reflected differences in P-limited shoot P status (Figure 1F). The variation in Pi concentration of P-limited organs was largely due to higher Pi acquisition in P-replete condition, with Pi concentrations ranging from 5.5 to 39.6 µmol g⁻¹ FW in P-replete shoots, and from 3.7 to 19.1 µmol g⁻¹ FW in P-replete roots (Figure 1, C and D; Supplemental Table S2). Total P concentration in P-replete shoots ranged from 12 to 55 µmol g⁻¹ FW (Supplemental Figure S2A). An inhibition of primary root growth, often described as a generalized response of A. thaliana roots to Pi limitation (Gutiérrez-Alanís et al., 2018), was not universal across accessions. Similar to the findings by Chevalier et al. (2003), some accessions did not arrest their primary root growth upon Pi withdrawal and 30 accessions even showed increased root growth (Figure 1, A and E; Supplemental Tables S2 and S3). In contrast, a reduction in shoot biomass was observed across all accessions in the P-limited treatment (Figure 1B). Thus, shoot −P/+P biomass ratio was positively correlated with Pi concentration in P-limited shoots (r = 0.29, P = 3.21E-4; Supplemental Figure S2B). Accessions with the lowest shoot total P levels in P-limited conditions generally also had the highest iron concentration in leaves (Supplemental Figure S2A; Supplemental Table S2). Counter-intuitively, −P/+P root biomass ratio and Pi concentration in P-limited roots were negatively correlated (r = −0.21, P = 2.52E-3). This is most likely due to dilution of the root Pi pool by lateral root growth, resulting in lower root and higher shoot Pi concentration. Together, these data show that P resources accumulated during seedling establishment are crucial to support RSA changes during Pi limitation while iron accumulation in shoots restricts P-limited root growth. Albeit significant, correlations were very low overall, again indicating strong variability between accessions.

Figure 1.

Arabidopsis thaliana accessions vary significantly in their physiological and metabolic response to Pi withdrawal. A and B, Violin plots of root (A) and shoot (B) biomass in P-replete and P-limited seedlings of 200 genetically diverse accessions. C and D, Violin plots of root (C) and shoot (D) Pi concentration in P-replete and P-limited seedlings. E, Effective primary root length in P-replete and P-limiting condition. F, Anthocyanin concentration in P-limited shoots. Anthocyanin concentrations in P-replete shoots were below the assay’s detection limit. G, Root and shoot biomass ratios (fresh weight of P-limited versus P-replete organs). For reference, Col-0 data are highlighted in red. ND, not determined. For each accession, the mean of two to three independent replicates was used (Supplemental Table S2). The dot in the middle of the Violin plot indicates the mean of all accessions and the vertical bar represents mean ± se. Statistical significance between treatments was determined by one-way ANOVA. **P < 0.01, ***P < 0.001.

Overall, the considerable variation of quantified traits across accessions allowed for a highly resolved genetic analysis of underlying determinants by GWAS.

GWAS reveals candidate genes involved in a more efficient PSR

To identify genes regulating individual PSR traits, we performed GWAS using single nucleotide polymorphism (SNP) data from the RegMap panel and the 1001 Genome Project (“Materials and methods”; Alonso-Blanco et al., 2016; Horton et al., 2012). Using the 1001 Genome SNP panel, we identified 154 significant SNPs (−log₁₀(P) ≥ 7) that showed strong genetic association with Pi concentration in P-limited roots and shoots, anthocyanin concentration in P-limited shoots, and effective primary root length under P-replete conditions (Supplemental Figure S3; Supplemental Table S4). Using the RegMap panel SNPs, we identified seven significant SNPs for two PSR traits (−log₁₀(P) > 6.4), including one genomic region associated with root biomass ratio (−P/+P). SNPs in the same genomic region were also associated with root biomass ratio with the 1001 Genome SNP panel but fell just below the selection threshold (−log₁₀(P) of 6.95 and 6.83, respectively; Supplemental Figure S3; Supplemental Table S4). These significant SNPs resided in 70 candidate genes (Supplemental Figure S3; Supplemental Table S4). To narrow down the candidate list, we cross-referenced the expression profile of these candidate genes using published RNA-seq data (Linn et al., 2017, Supplemental Table S5). A combination of SNP P-value, SNP impact, trait-of-interest, and RNA-seq expression profile of associated genes was considered to select two loci for further analyses: Five SNPs with significant association to Pi concentration in P-limited shoots were located on chromosome 5 (Figure 2, A–C). This locus contains four PHT1 genes (PHT1;6, PHT1;1, PHT1;3, and PHT1;2), with the latter three genes in the 12-kb region surrounding the lead SNP. We will therefore refer to this locus as the PHT1 locus. In agreement, reverse transcription-quantitative polymerase chain reaction (RT-qPCR) confirmed that all three genes were significantly induced in P-limited Col-0 roots (2-fold induction of PHT1;1, 85-fold induction of PHT1;2, 354-fold induction of PHT1;3; Figure 2D). Another locus on chromosome 5 was associated with Pi concentration in P-limited roots, a trait that is of great interest due to its negative correlation with −P/+P root biomass ratio and its wider implications for RSA (Supplemental Figure S2B). This locus contains two Pi starvation-responsive genes: one encodes a putative auxin efflux carrier family protein, PIN-LIKES7 (PILS7), and the other one encodes the amino acid transporter protein AMINO ACID VACUOLAR TRANSPORTER3 (AVT3) (Figure 2, E–G). A third gene of unknown function, AT5G66000, showed no transcriptional response to Pi withdrawal (Figure 2H;Supplemental Table S5). RT-qPCR confirmed that PILS7 is suppressed and AVT3 is induced in P-limited over P-replete Col-0 roots (Figure 2H). We will refer to this locus as PILS7 because nine out of 11 significant SNPs are in the genomic sequence of PILS7 (Supplemental Table S4). Considering the important roles of both Pi and auxin transport for P status, organ growth, and RSA (Perez-Torres et al., 2008; Bhosale et al., 2018), we focused on PHT1 and PILS7 loci for further analysis of causal mechanisms.

Figure 2.

Genome-wide association reveals loci responsible for variation in organ Pi levels in P-limited A. thaliana accessions. A, Manhattan plots for association with Pi concentration in P-limited shoots. The dashed horizontal line indicates the Bonferroni-adjusted significance threshold (–log₁₀(p) = 7.0). SNPs located within 5 kb of the lead SNP are labeled as red dots. B, Quantile–Quantile plot (Q–Q plot) for Pi concentration in P-limited shoots. C, Magnification of the genomic region surrounding the “PHT1” locus (12.3 kb). SNPs above the Bonferroni threshold are marked as red dots, gene models in this genomic region are shown below the x-axis. D, Transcript abundance of the candidate genes at the “PHT1” locus in P-replete (black bars) and P-limited (gray bars) roots of Col-0. E and F, Manhattan plot (E) and Q–Q plot (F) for genetic association with Pi concentration in P-limited roots, annotated as in (A) and (B). G, Close-up of the genomic region surrounding the “PILS7” locus (12.3 kb). Annotations are the same as in (C). H, Transcript levels of the candidate genes at the “PILS7” locus in P-replete and P-limited roots of Col-0. See (D) for detailed annotation. In D and H, each dot represents a biological replicate comprising eight seedlings grown vertically on a plate. Data are mean ± se. Statistical significance was determined by one-way ANOVA, **P < 0.01, ***P < 0.001.

Loss-of-function alleles of PHT1;1, PHT1;2, and PILS7 in Col-0 affect plant growth and organ Pi levels

To characterize the impact of genes in the PHT1 and PILS7 loci on acclimation of the Col-0 reference genotype to low Pi supply, T-DNA insertion mutants of the five PSR genes (PHT1;1, PHT1;2, PHT1;3, PILS7, AVT3) were tested for their PSR (Supplemental Figure S4; “Material and methods”). In agreement with a previous study (Shin et al., 2004), the pht1;1 mutant showed a significant reduction in shoot fresh weight compared to that of wild-type in both Pi conditions (Figure 3A). Pi concentrations in P-replete pht1;1 roots and shoots were significantly reduced by 35% and 70% compared to wild-type, respectively, reaffirming the prominent role of PHT1;1 in Pi uptake (Figure 3, C and D). The pht1;2 mutant had the opposite effect on growth as its shoot biomass was significantly higher than that of wild-type under P-replete conditions, as was root biomass under P-limiting conditions (Figure 3, A and B). However, there was no significant difference in organ Pi accumulation between pht1;2 and wild-type (Figure 3, C and D). Loss of function in Pi transporters causes retarded plant growth (Nagarajan et al., 2011; Remy et al., 2012), so enhanced organ biomass of the pht1;2 mutant suggests that PHT1;2 may either be a Pi exporter, expressed in cell types associated with Pi translocation or have functions beyond Pi transport. The pht1;3 mutant showed no trait difference compared to wild-type. Across accessions, the PHT1 locus was associated with Pi concentration in P-limited shoots (Figure 2); however, in the Col-0 background, there was no significant difference in shoot Pi concentrations between P-limited pht1 mutants and the control. Knockout of PHT1;1 seemed to impact P status of P-replete seedlings instead (Figure 3, C and D). In line with the GWAS results, the pils7-1 allele caused higher Pi accumulation in P-limited roots (25%; Figure 3D). Both P-replete PILS7 loss-of-function mutants showed significant reductions in root biomass (32%) and root Pi concentration (20%; Figure 3, B and D). The avt3 mutant did not display any P status-dependent physiological changes compared to wild-type. These data suggest that PILS7 is associated with Pi concentration in P-limited roots, and that impaired PILS7 activity leads to reduced Pi uptake and/or Pi translocation from root to shoot.

Figure 3.

pht1;1 and pils7 mutants show impaired growth, organ Pi accumulation and root Pi acquisition. A and B, Fresh weight of shoots and roots of 14-d-old P-replete and P-limited seedlings. C and D, Pi concentration in shoots and roots of 14-d old seedlings. Experiments in (A)–(D) were performed in two separate batches, one for PHT1 locus mutants and one for PILS7 locus mutants. Each dot represents a biological replicate comprising eight seedlings grown vertically on a plate. Data are mean ± se. E, Pi acquisition by P-replete and P-limited roots. Results are from two independent experiments with three replicates of five seedlings each, with pht1;1, pht1;3, and avt3 only included in one experiment. Data are mean ± se. Asterisks indicate significant differences from Col-0 under each Pi treatment (one-way ANOVA and Tukey’s HSD test, P < 0.05).

pht1;1 and pils7 mutant alleles are impaired in Pi acquisition

To determine whether changes in organ Pi accumulation were a result of altered Pi uptake by roots, we conducted a Pi depletion assay using PHT1 and PILS7 locus mutants and compared those to the pho2-2/ubc24-1 (SAIL_47_E01) (Aung et al., 2006) and phr1-2 (SALK_067629C) mutants (Nilsson et al., 2007). Consistent with the earlier reports, the pho2-2 mutant exhibited significantly enhanced Pi uptake, while uptake tended to be lower in the phr1-2 mutant compared to wild-type (Figure 3E). Similar to results by Shin et al. (2004), Pi uptake capacity of P-limited pht1;1 was reduced to 70% of that of wild-type under P-limited conditions, but we did not observe significant differences compared to wild-type in P-replete condition. The pht1;2 and pht1;3 mutants behaved like wild-type (Figure 3E), as did the avt3 mutant. Both pils7 mutant alleles showed impaired Pi uptake under P limitation, with P-limited pils7-1 and pils7-2 having only 71% and 70% of the wild-type uptake capacity, respectively (Figure 3E). Similar Pi uptake capacity between pht1;1, pils7 mutants and wild-type under P-replete conditions clashes with the observation of a lower Pi concentration in P-replete pht1;1 and pils7 roots (Figure 3D). This could be other Pi transporters compensating for the pht1;1 knockout in the short term (Pi depletion from the media in Figure 3E was measured after 8 h), but not in the long term (accumulative effect in Figure 3D was measured 7 d after transfer). In P-replete pils7 mutants this effect could either be achieved by impairing PHT1;1 function, or by altering root-to-shoot Pi translocation.

Large-scale rearrangement of the PHT1 locus corresponds with lower Pi concentration in P-limited shoots of Hap 2 accessions

Next, we assessed how the allelic variation in PHT1;1, PHT1;2, or PHT1;3 identified by GWAS is causal of the variation in shoot Pi concentration across the 200 accessions under low Pi supply. Using the RegMap panel’s 250K SNP data, we performed a haplotype analysis on the genomic region encompassing PHT1;1, PHT1;3, and PHT1;2 (Figure 4A). The two haplogroups showed significant differences in Pi concentrations of P-limited shoots, with Pi levels in Hap1 accessions higher than those in Hap2 (P = 0.00632; Figure 4B;Supplemental Table S6). SNP patterns of those accessions that were sequenced by the 1001 Genome Project revealed many variants segregating among accessions from Hap2 compared to those of Hap1 as well as the Col-0 reference allele (Figure 4A). For both the PHT1;1 and PHT1;2 coding regions, one SNP shared by representative accessions of Hap2 led to a conservative amino acid change in each (Figure 4A). Strikingly, representative Hap2 accessions featured deletions in promoters, exons, and introns (gray bars in Figure 4A, dotted red boxes in Supplemental Figure S4, A and C). Consequences of these SNPs and deletions for the amino acid composition of PHT1;1 and PHT1;2 proteins from Hap2 accessions and their predicted membrane topology are presented in Supplemental Figure S5: Exon 2 deletions in Hap2 alleles of PHT1;1 and PHT1;2 cause a loss of the second last extracellular loop at the transporters’ C-termini which will dramatically alter overall membrane topology. Given that the C-terminus contains two phosphorylation sites (S514 and S520 of PHT1;1) that affect PHOSPHATE TRANSPORTER TRAFFIC FACILITATOR-mediated PHT1 exit from the endoplasmic reticulum (ER), as well as the predicted ER exit site itself (Bayle et al., 2011), these deletions are expected to dramatically affect PHT1 activity and/or regulation. The PHT1;1 allele is likely to be further functionally compromised due to two deletions in transmembrane domain VII and the preceding cytoplasmic loop (dotted gray boxes in Supplemental Figure S5B). In Hap2 accessions, PHT1;2 would be the only remaining, fully functional Pi transporter of the PHT1 locus. Even more striking, the entire PHT1;3 gene is missing from Hap2 accessions (Figure 4A). Expression profiling of PHT1;3 in roots of select Hap1 and Hap2 accessions across Pi treatments confirmed that the transcript could not be detected in the latter (Figure 5A). Alignment of whole-genome sequencing reads for these accessions against the Col-0 PHT1 locus confirmed the deletion of the PHT1;3 gene in Hap2 accessions (Figure 5B).

Figure 4.

Sequence variation in PHT1 and PILS7 loci is associated with natural variation in organ Pi concentrations in P-limited seedlings. A and C, Genomic sequence surrounding the PHT1 (A) and the PILS7 locus (C) in five representative accessions from the two most distinct haplogroups. Gene models (shown in green at the top) represent those in Col-0. Colored vertical lines show single base pair substitutions with the letter of the nucleotide shown next to the line; black vertical lines indicate single base pair deletions; gray horizontal bars indicate larger deletions. Text and arrows below each part indicate nucleotide and amino acid substitutions shared by the five Hap2 accessions; the nonconservative amino acid change in PILS7 is labeled in red. SNPs shared uniquely by either Hap1 or Hap2 accessions are indicated by hash and vertical lines at the top and bottom of the alignment, respectively. The large deletion within the PHT1 locus encompassing PHT1;3 is highlighted by a red box. Pictures were generated from http://signal.salk.edu/atg1001/3.0/gebrowser.php. B and D, Boxplots of normalized Pi concentration in P-limited shoots (B) and roots (D) of accessions forming two distinct haplogroups with respect to the PHT1 (B) and PILS7 locus (D). The lower and upper box edges correspond to the first and third quartiles, the horizontal line indicates the median, the whiskers extend to minimum and maximum values within 1.5× interquartile ranges. Statistical significance was determined by one-way ANOVA (P < 0.05). LP: P-limited (6.5 μM).

Figure 5.

The PHT1;3 gene is absent from the genome of representative Hap2 accessions. A, Pi-dependent PHT1;3 expression in four representative accessions of the two most distinct PHT1 haplogroups. Each dot represents a biological replicate comprising 10 seedlings grown vertically on a plate. Data are mean ± se. In P-replete conditions, PHT1;3 expression was detected in two replicates of Ag-0, Bay-0, Wt-5, and only one replicate in Gu-0. Note that some of the accessions chosen here differ from those shown in Figure 4A (for Hap2, PHW-33 and UKSE06-278 were not sequenced by 1001 Genome Project). B, Read coverage of sequenced PHT1 loci of three accessions from each of the two distinct haplotypes. Hap1 and Hap2 accessions are marked in blue and orange, respectively. ND, not detected.

In conclusion, copy-number variation with loss of PHT1;3 and major rearrangement of the remaining two PHT1 genes in Hap2 accessions dramatically reduces Pi acquisition resulting in lower shoot Pi concentration and higher sensitivity to Pi limitation. Differences in PHT1 gene content and sequence variation may reflect adaptations of haplogroups to Pi availability in their habitats with Hap2 most likely originating from a P-rich environment.

Allelic variation at the PILS7 locus is associated with root Pi concentration in P-limited accessions

For the PILS7 locus, the Pi concentration in P-limited roots of Hap1 accessions was significantly lower than that of Hap2 accessions (Figure 4D;Supplemental Table S6). Due to the negative correlation between Pi concentration in P-limited roots and root biomass ratio (Supplemental Figure S2B), Hap1 accessions are likely to be more tolerant to Pi withdrawal than those of Hap2. To identify causal sequence polymorphisms, we compared 1001 Genome Project derived PILS7 locus SNP information for select accessions in two contrasting haplogroups: Across the genomic region, Hap2 accessions harbor 22 common SNPs that are absent from those of Hap1 (Figure 4C). Of the SNPs in the PILS7 coding region, eight reside in exons and seven in introns of the Hap2 allele (Figure 4C). Of the eight exonic SNPs, only one leads to a nonsynonymous change from alanine (Ala) to threonine (Thr). The other seven SNPs are silent mutations (Figure 4C). Sanger sequencing of genomic PILS7 sequences PCR-amplified from representative Hap1 accession HSm and Hap2 accession Liarum confirmed these SNP locations (Supplemental Figure S6). It furthermore revealed extensive insertions and deletions within the promoter and the first intron of these two PILS7 alleles (blue boxes in Supplemental Figure S6).

There is very little information available on functional domains within PIN-LIKES (PILS) proteins, but transmembrane helices are highly conserved among PILS family members, and thus may have central roles in auxin carrier function. The cytosolic loops display a lesser degree of conservation and may have regulatory functions (Barbez et al., 2012). The nonsynonymous amino acid change is located in the longest cytosolic loop of PILS7 (Supplemental Figure S7). Substitution of the Ala residue at position 197 with Thr in group 2 accessions may change the regulation of PILS7, possibly through protein phosphorylation at Thr¹⁹⁷. This posttranslational modification could alter PILS7 activity or turnover, subsequently affecting auxin sequestration in the ER and nuclear auxin signaling (Beziat et al., 2017; Feraru et al., 2019). Changes in auxin gradients would then impact lateral root and root hair formation, and either directly or indirectly impact on Pi uptake and/or Pi translocation.

SNPs and indels in the promoter and first intron could alter the expression and/or splicing of PILS7, resulting in altered PILS7 protein abundance. Promoter analysis using PlantPAN version 3.0 (Chow et al., 2019) identified a key TF binding region (Supplemental Figure S6; Supplemental Table S7) that was unique to the HSm (Hap1) PILS7 allele. It featured binding sites for TFs of the APETALA2 (AP2)/ETHYLENE RESPONSE FACTOR (ERF) and ARF families. AP2/ERF TFs have been associated with auxin-sensitive abiotic stress signaling in roots promoting the transcription of ARF-family AUXIN/INDOLE ACETIC ACID repressors in response to desiccation and osmotic stress (Shani et al., 2017). All of the ARF TFs predicted to bind to the HSm PILS7 promoter (ARF2, ARF4, ARF5, ARF6, ARF7, ARF8, and ARF11; Supplemental Table S7) have been associated with auxin-controlled root hair as well as primary and lateral root development (Choi et al., 2018; Dastidar et al., 2019; Santos Teixeira and Ten Tusscher, 2019; Yin et al., 2020). While ARF5 and ARF11 stimulated root hair elongation, ARF2 and ARF4 acted as repressors (Choi et al., 2018). ARF7—together with ARF19—targets the PHR1 promoter which features three auxin-response elements that confer auxin-stimulated lateral root formation and increased Pi uptake in P-limited A. thaliana seedlings (Huang et al., 2018). These findings suggest increased auxin-sensitive abiotic stress responsiveness of the HSm (Hap1) but not the Liarum (Hap2) PILS7 allele. We, therefore, measured PILS7 expression in roots of select Hap1 and Hap2 accessions under P-replete and P-limiting conditions (Supplemental Figure S8). While across accessions, expression was higher in P-replete and lower in P-limited roots, PILS7 transcripts were significantly less abundant in P-limited roots of Hap2 accessions by about two-fold. Thus, the additional TF binding site in the Hap1/HSm allele may help sustain PILS7 expression in P-limited roots. To test whether differential PILS7 expression was associated with expression changes in genes associated with Pi uptake or translocation, PHT1;1, PHT1;4, PHO1, and MIR399D transcript profiles were also determined (Supplemental Figure S8). The four genes showed the typical expression profile reported for P-limited Col-0 roots, with strong induction of MIR399D and PHT1;4, moderate induction of PHT1;1 and no change in PHO1 expression. None of these genes showed differences in expression between haplogroups, indicating that observed changes were PILS7 specific.

Taken together, extensive allelic variation across the entire PILS7 genomic sequence leads to altered PILS7 abundance and possibly altered post-translational regulation which would affect root auxin signaling under stressful versus nonstressful conditions and cause the observed natural variation in root Pi concentration and organ growth. It is unlikely that altered transcript expression of PHO regulon components is responsible for trait variation.

Elemental composition differs in the two contrasting haplogroups associated with PHT1 and PILS7 loci

Shoot elemental composition data for each accession (Supplemental Figure S2A; Supplemental Table S2) offered the opportunity to investigate the interaction of P with other nutrients across the contrasting P-related haplogroups. Across accessions, the PHT1 locus was associated with variation of Pi concentration in P-limited shoots (Figure 4B). The elemental profiles showed that total P (sum of inorganic and organic P) levels in P-limited shoots were similar between haplogroups. However, P-replete Hap1 accessions had higher leaf Pi and total P concentration than Hap2 accessions (Supplemental Figure S9A). This may suggest that due to their three functional PHT1 paralogs, Hap1 accessions are able to build up higher organic P pools to support growth under Pi limiting conditions.

For the PILS7 locus, P-limited Hap1 accessions had lower root Pi concentration (Figure 4D). These accessions also had higher Pi concentration in P-replete organs and higher root biomass irrespective of Pi supply (Supplemental Figure S9B). Total elemental composition analysis revealed that their P-replete shoots also had higher total P levels (Supplemental Figure S9B), again suggesting higher mobilization capacity upon Pi withdrawal. Unlike PHT1 locus-associated haplogroups, contrasting PILS7 Hap accessions also differed in their leaf iron and copper content. In P-limited environments, root architecture is also modified by iron and copper availability (Ward et al., 2008; Perea-Garcia et al., 2013). Higher iron content in P-limited leaves of Hap2 accessions is consistent with lower shoot Pi and total P levels and confirms their higher sensitivity to Pi withdrawal (Supplemental Figure S9B). Irrespective of P status, copper concentration is always higher in leaves of Hap2 accessions which can be an indicator of altered PIN1-mediated auxin distribution (Yuan et al., 2013). High copper concentrations cause primary root length inhibition via auxin depletion of the root apical meristem, a phenotype similar to that seen here for pils7 mutants (Figure 3B).

Despite similar nutrient allocation profiles, overlap in accessions between Hap1 and Hap2 of PHT1 and PILS7 loci was low—with only 3 out of the 23 and 19 accessions shared in Hap1, and 8 out of 30 and 29 accessions shared in Hap2, respectively (Supplemental Table S6). This would indicate that there has been no common selection for these genetic marks.

The PILS7 allele of Hap2 accessions fails to complement PILS7 knockout in Col-0

To further assess the impact of the contrasting Hsm and Liarum alleles on PILS7 function, their genomic sequences were used for the complementation of the pils7-1 mutant (Supplemental Figure S10A). The CaMV 35S promoter-driven coding sequence of the Col-0 allele of PILS7 was also transformed into the pils7-1 mutant background for comparison. The latter construct resulted in strong (at least 435-fold) PILS7 overexpression compared to Col-0 (Supplemental Figure S10B). 35S::PILS7^Col-0 overexpression led to poor seedling and lateral root development as well as stronger anthocyanin accumulation in shoots (Supplemental Figure S11, E and F). Ectopic expression of PILS7 which—under its native promoter—is much more strongly expressed in Col-0 roots than shoots (Supplemental Figure S11G), thus appears to severely impair root auxin distribution, perception, or signaling, leading to growth impairment.

Across all progeny obtained for the two Hap alleles, the Hap1 allele PILS7^HSm led to distinctly higher expression than the Hap2 allele PILS7^Liarum (Supplemental Figure S10, C and D). With each T1 line representing an independent T-DNA insertion, this could already be an indication of differences in the relative promoter strengths of these two alleles. For each allele, we chose two complementation lines for further characterization (highlighted in Supplemental Figure S10, C and D). While pils7-1 mutants expressing the Hap1 allele PILS7^HSm to similar levels as the wild-type allele in Col-0 were able to restore organ biomass and Pi concentration of the pils7-1 mutant back to Col-0 levels, the Hap2 allele PILS7^Liarum failed to complement the pils7-1 mutant when expressed at levels similar to the wild-type allele or at >150-fold higher levels (Figure 6; Supplemental Figure S10D). Despite strong overexpression, the latter complementation line did also not lead to the retarded growth phenotype observed in the 35S::PILS7^Col-0 lines.

Figure 6.

Natural allelic variation in PILS7 impacts Pi-dependent growth and organ Pi allocation. A and B, Shoot (A) and root (B) fresh weights of 14-d-old Col-0, pils7-1 mutant and complementation lines carrying HSm or Liarum PILS7 alleles. C and D, Organ Pi concentration of seedlings shown in (A) and (B). HSm and Liarum are accessions from Hap1 and Hap2 accessions, respectively (Figure 4C). For pils7-1 complementation, two individual lines for each haplotype allele were selected. Data are from two independent experiments with three biological replicates. Each dot represents a biological replicate comprising ten seedlings, the lower and upper box edges correspond to the first and third quartiles, the horizontal line indicates the median, the whiskers extend to minimum and maximum values within 1.5× interquartile ranges. Lines carrying the same PILS7 allele were compared with Col-0 as one group. Asterisks indicate significant differences from Col-0 under each Pi-supply condition (one-way ANOVA and Tukey’s honestly significant difference (HSD) test, *P < 0.05, **P < 0.01, ***P < 0.001).

These results demonstrate that the two contrasting alleles not only have different promoter strengths, but also result in functionally distinct PILS7 proteins. Genetic differences between them are likely the result of adaptation to local P environments with Hap1 PILS7 alleles providing improved auxin signaling in roots. This promotes more vigorous (lateral) root growth and higher Pi uptake capacity enabling plants to actively seek out and exploit P-rich topsoil patches.

Discussion

Complexity of genome-wide associations with key PSR traits

In this study, we performed a GWAS on a number of traits associated with acclimation to low Pi availability using A. thaliana accessions of high genetic diversity (Li et al., 2010). Unlike other studies that germinate seeds on media with contrasting Pi levels, our experimental design aimed at identifying key determinants of more efficient Pi acquisition and utilization in the presence of Pi, as seedlings were established on P-replete media prior to transfer to either low or high Pi media. The selected traits showed variation across accessions, but only five traits showed significant association with SNPs (Figure 2;Supplemental Figure S3; Supplemental Table S4). One explanation for the limited number of associations with some traits might be that these are controlled by a large number of genetic variants, each with only a modest contribution to the total phenotypic variation. These minor-effect loci are only detectable when the size of study population is big enough (Visscher et al., 2017). The fact that we found strong genetic determinants of root but not shoot biomass ratio in P-limited over P-replete plants came as a surprise, given that root growth relies on exported assimilate from shoots, and in nonstressed plants, a strong genetic coupling between root and shoot growth has been found (Bouteillé et al., 2012). The requirement to respond to environmental challenges would have made this relationship more complex over time. The resulting complex genetic architecture of shoot growth can be a major challenge for GWAS (Bouteillé et al., 2012; Marchadier et al., 2019). Nutrient limitation causes the strongest allocation responses, with large increases in root biomass at the expense of stem and leaf biomass and no significant difference between species from nutrient-poor and nutrient-rich habitats (Poorter et al., 2012). The lack of variability in shoot biomass reduction across P-limited A. thaliana accessions found in this study supports this hypothesis.

Variation in copy number and protein topology of Pi transporters as potential sources of adaptation to low P conditions

Pi transporters of the PHT1 family are essential for Pi acquisition and Pi translocation, however, studies on their function have so far focused on the A. thaliana accession Col-0. Here, we detected an association between the PHT1 locus on chromosome 5 and Pi concentration in P-limited shoots across 200 A. thaliana accessions. The locus identified on chromosome 5 contains four PHT1 paralog genes, most likely derived from a series of duplication events (Poirier and Bucher, 2002). From an evolutionary standpoint, gene duplication events in rate-limiting ion transporter families, such as the PHT transporters, are sometimes associated with increased dosage, but many are subjected to stronger purifying selection in the long term (Hudson et al., 2011). However, so called “fate determining mutations” can sub or neofunctionalize duplicates and reduce selection pressure whilst maintaining the original functional copy (Innan and Kondrashov, 2010; Fournier-Level et al., 2011; Carretero-Paulet and Fares, 2012). Copy-number variants have been detected in Arabidopsis accessions (Long et al., 2013; Bush et al., 2014; Göktay et al., 2021; Jiao and Schneeberger, 2020; Zmienko et al., 2020). The most recent study of 1,135 whole-genome sequenced accessions from the 1001 Genomes Project identified copy-number variants associated with 18.5% of protein-coding regions, in particular regions of tandem duplications (Zmienko et al., 2020). Loss of PHT1;3 has been captured as CNV_18358. In Hap2 accessions, the loss of PHT1;3 and substantial deletions in promoter and exon regions of PHT1;1 and PHT1;2 associated with progressive loss of function could be an adaptation to environments with reliable Pi availability (Figures 4, A and 5; Supplemental Figure S5). Our finding that these accessions have lower Pi and total P levels in P-limited shoots than those of Hap1 confirm their reduced Pi uptake capacity and higher sensitivity to Pi starvation (Supplemental Figure S9). In high P environments, the PHT1;3 gene might not be under the same selection pressure as in low P environments, and its loss in Hap2 accessions does not impact in situ performance (Supplemental Figure S9A). The extra control loop that prevents hyperaccumulation of Pi in variable P environments by phosphorylating the C-terminus of excess PHT1 proteins and retaining them in the ER (Bayle et al., 2011) is not needed in habitats with more readily available Pi that incur only a moderate expression of PHT1 protein in the first place (Supplemental Figure S5). In Col-0, PHT1;2 and PHT1;3 are considered redundant (Ayadi et al., 2015) but only PHT1;3 has been lost in Hap2 accessions. We found that unlike other Pi transporter mutants in Col-0, pht1;2 had higher organ biomass than P-replete wild-type, and higher root biomass in P-limited conditions (Figure 3). Its retention may therefore be due to its positive impact on plant growth. In wheat, the expression of Pi transporters, and in particular TaPHT1;2, in response to Pi limitation differed between P-acquisition-efficient and -inefficient cultivars, which also showed marked differences in organ- and tissue-specific PSR traits (Aziz et al., 2014; de Souza Campos et al., 2019). In summary, Hap2 accessions carry genome modifications that are likely to reduce overall PHT1 transporter abundance at the plasma membrane as a reflection of adaptation to their local environment.

Hormonal signaling during PSR-induced changes in RSA

Several phytohormones are involved in PSR, for example, auxin, jasmonic acid, and ethylene (Borch et al., 1999; Perez-Torres et al., 2008; Khan et al., 2016; Bhosale et al., 2018). Recent studies have shown that auxin signaling is crucial for Pi starvation-induced modifications of RSA (Bhosale et al., 2018; Huang et al., 2018). The PILS family of auxin transporters comprises seven members (PILS1–PILS7) (Barbez et al., 2012). Individual members of this family have recently been functionally characterized as ER-localized auxin carriers that sequester auxin in the ER (Feraru et al., 2019), which in turn promotes auxin conjugation and dampens nuclear auxin signaling. PILS2–PILS7 transcript abundance increased with external auxin application (Barbez et al., 2012). Overexpression of PILS1 or PILS3 led to shoot developmental defects and dwarf plants. Knockout of PILS2 and PILS5 promoted hypocotyl, primary and lateral root growth (Barbez et al., 2012). This and other studies support a role of PILS proteins as negative regulators of plant growth and development (Barbez et al., 2012; Beziat et al., 2017; Feraru et al., 2019). In contrast, our results suggest that PILS7 is a positive regulator of organ growth and Pi allocation as well as Pi acquisition by P-limited roots (Figure 3). Lack of complementation of the pils7-1 mutant by constitutive PILS7 expression suggests a highly dose-dependent, stress- and/or cell-specific role. Given its role in nuclear auxin depletion, PILS7 function could be associated with short-distance auxin transport and signaling during abiotic stress (Korver et al., 2018). Its function could be to establish the cytokinin-dependent auxin minimum needed to promote root cell differentiation and/or auxin oscillations required for lateral root formation (De Rybel et al., 2010; Di Mambro et al., 2017). Similar to pils7 mutants (Figure 6), Hap2 accessions have lower PILS7 expression (Supplemental Figure S8) and higher Pi concentration in P-limited roots (Figure 4D;Supplemental Figure S9B). The PILS7 protein of Hap2 is furthermore predicted to carry an extra phosphorylation site in its central cytoplasmic loop (Supplemental Figure S7). These genomic modifications render the Hap2 PILS7 allele incapable of rescuing the pils7-1 mutant in the Col-0 background (Figure 6). The negative correlation between P-limited root Pi concentration and root biomass ratio (−P/+P) (Supplemental Figure S2B) would suggest that Hap2 accessions come from P-rich habitats and are more sensitive to Pi limitation. This is supported by their reduced capacity to take up Pi in replete condition and the higher iron accumulation in P-limited shoots (Supplemental Figure S9B). Selection pressure to sustain stress-responsive PILS7 promoter activity in these habitats may have been low. The phosphorylation site in the Hap2 PILS7 protein may be part of an additional kinase/phosphatase signaling loop to help regulate auxin transporter activity in response to other environmental or developmental clues. Hap1 accessions are more stress tolerant and maintain higher organ Pi levels in P-replete conditions to support root growth upon Pi withdrawal (Supplemental Figure S9B). The Hap1 PILS7 allele can complement the pils7-1 mutant in the Col-0 background (Figure 6). Unlike the Col-0 allele, its promoter is targeted by stress responsive CBFs/ERFs and early ARF-dependent auxin signaling modules for lateral root development (Santos Teixeira and Ten Tusscher, 2019) that help to sustain PILS7 expression upon Pi withdrawal (Supplemental Figure S8). The regulatory elements involved would suggest that—in stress-tolerant Hap1 accessions—PILS7 is part of the TIR1- and ARF19-dependent signaling cascade that stimulates the first asymmetric divisions in pericycle cells to promote lateral root formation upon Pi withdrawal (Perez-Torres et al., 2008). How exactly PILS7 activity impacts on nuclear auxin levels and ARF-dependent auxin and PSR signaling to promote root hair and lateral root growth remains to be elucidated.

Conclusion

The results of this study revealed that higher Pi acquisition, Pi translocation from shoot to root and higher investment in root biomass are critical for successful adaptation to a low Pi environment. A switch in PHT1 isoform use, together with altered transcriptional and post-translational regulation of PHT1 isoforms and PILS7 are tightly associated with these traits. Interactions between these two loci are complex, however, with only a limited number of either Pi limitation tolerant or sensitive accessions sharing both genetic marks. The initial SNP association led to the identification of more substantial genomic variation in alleles of individual accessions that allowed us to identify additional aspects in the regulation of known players (PHT1 isoforms) and another player (PILS7) as key determinants of P efficiency that can inform plant selection and improve fertilizer use in agronomic production systems.

Materials and methods

Plant materials and growth conditions

The 200 Arabidopsis (A. thaliana) accessions were kindly provided by Justin Borevitz (Research School of Biology, The Australian National University, Canberra, Australia). In order to identify differences in PSR without interference from seed quality, accessions were propagated in the same temperature-controlled glasshouse and seeds were harvested from individual plants showing the expected growth habit according to the germplasm details provided by The Arabidopsis Information Resource (www.arabidopsis.org). Accessions requiring vernalization (Supplemental Table S1) were transiently transferred to a temperature-controlled cabinet for cold treatment. Names and identities of accessions as well as vernalization information are provided in Supplemental Table S1.

T-DNA insertion lines were obtained from the Nottingham Arabidopsis Stock Center (pht1;2/SALK_110194C; pht1;3/GK-557C09; pils7-1/GK-768F05; pils7-2/SALK_069485; avt3/SALK_010447C). Genotyping was carried out using primer combinations listed in Supplemental Table S8A. T-DNA insertion sites in either the first or second exon of each mutant were confirmed by Sanger sequencing (Supplemental Figure S4). Transcript abundance was determined via RT-qPCR (Supplemental Figure S4). Previously published mutants, pht1;1-2 (SALK_088586C, Shin et al., 2004), phr1-2 (SALK_067629C, Nilsson et al., 2007) and pho2-2/ubc24-1 (SAIL_47_E01, Aung et al., 2006), and Col-0 (N70000) were used as controls in the phenotyping experiments.

Plants for genotyping and propagation were grown in soil with 0.5 L coarse Vermiculite, 0.33 L Perlite, 33 g Nutricote controlled-release fertilizer, 28 g ammonium nitrate, 25 g water-holding granules, 15 g trace elements, and 7 g garden lime added per kilogram of standard potting mix (Van Schaik’s BioGro, Mount Gambier, Australia) under a 16-/8-h light–dark cycle with 120 µmol m⁻² s⁻¹ light intensity, at 22°C/19°C (light/dark), and 55% relative humidity.

For the accession screen as well as phenotyping of T-DNA mutants and transgenic lines, seeds were sterilized with chlorine gas for 2 h, and then stratified at 4°C for 2 d in the dark. Seedlings were germinated and grown on 10-cm square Petri dishes filled with 50 mL agar-solidified Murashige and Skoog (MS) medium (Murashige and Skoog, 1962). After sowing of seeds, the Petri dishes were placed in a near-vertical position. The environmental settings were the same as for soil-grown plants. The MS medium had the following composition: 0.61 g L⁻¹ MS Modified Basal Salt mixture (M407; Phytotech Laboratories, Lenexa, Kansas), 20.6 mM NH₄NO₃, 18.8 mM KNO₃, 1 mM KH₂PO₄, 0.1% (w/v) MES, and 0.9% (w/v) Difco™ Granulated Agar (LOT 6173985). For Pi depletion, 1 mM KH₂PO₄ was replaced by 1 mM KCl. The solution was adjusted to pH 5.8 using 5 M KOH. The residual Pi concentration of the agar used was 6.5 µM.

Accessions were established on MS medium, before seedlings with 2-cm-long primary roots were transferred to either P-replete (1 mM Pi) or P-limited (6.5 µM Pi) medium and assessed after 7 d of growth (Supplemental Figure S1). Using seedlings of similar size across accessions aimed at reducing the bias arising from maternal effects around seed quality and/or inherent genetic differences in germination. Following an initial growth study, accessions were put into eight groups defined by the number of days after sowing when the primary root length reached ∼2 cm (Supplemental Table S1). The seedlings were established for 4 + x days in P-replete medium, with x equaling the group number. To characterize T-DNA mutants in the Col-0 background, as well as PILS7 overexpression and pils7-1 complementation lines, seedlings were established in P-replete medium until the primary root length reached ∼2 cm, and then transferred to either P-replete or P-limited medium and grown for another 7 d prior to harvesting root and shoot material.

Tissue collection

In the accession screen, one plate containing ten seedlings constituted one biological replicate. Most accessions had three biological replicates per treatment and genotype, and a few accessions only had two biological replicates due to poor germination (Supplemental Table S3). For each plate, seedlings were separated into root and shoot for harvesting. Five individual shoots and ten roots were combined into one sample for measuring Pi and anthocyanin (shoots only) concentrations. For the characterization of transgenic lines, one plate containing eight seedlings constituted one biological replicate. Each genotype had three biological replicates per treatment and fresh weights were recorded for all samples, prior to shock-freezing in liquid N₂ and stored at −80°C.

Primary root length measurement

Primary root length was determined as described earlier (Linn et al., 2017). Root images were analyzed in the ImageJ software using the SmartRoot plugin (Lobet et al., 2011). The effective primary root length was calculated by subtracting root length before transfer from root length at final harvest. The effective primary root lengths of P-replete or P-limited seedlings were used for GWAS.

Determination of Pi and anthocyanin concentration

To determine Pi and anthocyanin concentration, the frozen plant samples were ground and extracted with 1% (v/v) acetic acid at 4°C in the dark. Pi concentration was measured using the colorimetric ammonium molybdate assay as described earlier (Jost et al., 2015). Anthocyanin concentration in leaf samples was determined using a pH-differential method as described previously (Wrolstad et al., 2005).

Total P and elemental composition analysis

Accessions were grown as described (Supplemental Figure S1). Three shoot replicates were pooled to generate sufficient dry weight for acid digestion. The method for elemental analysis was adapted from Foroughi et al. (2014). Dry shoot material (∼10 mg) was digested with 300 µL of HCl:HNO₃ (3:1) at 70°C for 3 h. Tomato (Lycopersicon esculentum) leaf reference material (Sigma Aldrich, St Louis, MO, USA; NIST1573A) was used to validate the method accuracy. The digested samples were adjusted to a final volume of 10 mL of Milli-Q water and quantified by inductively coupled plasma mass spectrometry.

Correlation analysis

The average of each measured trait was used for the correlation analysis, with two to three biological replicates for each accession (Supplemental Table S2). Correlation coefficients between the traits were calculated using the “cor” function for Pearson’s correlation in R (www.r-project.org). P-values were calculated using “cor_pmat’ function in the ggcorrplot package (Version 0.1.3) in R.

Pi depletion assay

Seedlings were grown on P-replete MS medium for 7 d and transferred to P-replete or P-limited medium for another 7 d. Seedlings were then transferred to 2.5 mL of liquid P-replete MS medium in 24-well plates (Greiner CELLSTAR, Frickenhausen, Germany; M9312), with five seedlings in each well. Aliquots of 200 µL MS medium were sampled prior to and 8 h after seedling addition. The Pi concentration of the medium was measured as described above to calculate the amount of Pi absorbed by the plants.

Statistical analysis of the measured traits

To account for possible batch effects, the best linear unbiased prediction (BLUP) of the phenotypic data was obtained, and the linear mixed effect function “lmer” in the lme4 package of R (version 3.5.3) was used to fit the model (Borevitz et al., 2002). The model for the phenotypic trait was Y_ij= u + Group_i + Genotype_j + e_ij, where u is the total mean, Group_i is the random group effect of the ith group, Genotype_j is the random genetic effect of jth genotype, e_ij is a random error. The genotypic (breeding) value for each accession was computed as the BLUP of the genotype effect.

Genome-wide association analysis

Out of the 200 accessions used in this study, 194 were covered by the RegMap panel and 104 by the 1001 Genome Project (Horton et al., 2012; Alonso-Blanco et al., 2016). BLUP values for each trait were used as phenotypic input for the GWAS analysis. GWAS was performed on the easyGWAS website (https://easygwas.ethz.ch) using the Efficient Mixed-Model Association eXpedited algorithm that accounts for population structure (Yu et al., 2006; Kang et al., 2010; Grimm et al., 2016). SNPs with a minor allele frequency of ˂0.05 were excluded from the analysis. The effective number of independent SNPs was calculated using a method described by Li et al. (2012). The effective number of independent SNPs for this study was calculated as 461,582 and 126,433 for the 1001 Genome Project and RegMap panel, respectively. A significance threshold of α = 0.05 was used after Bonferroni correction for multiple testing. Manhattan plots were generated using the qqman package in R version 3.5.3. The location of genes closest to these significant SNPs was visualized by PhenoGram (Wolfe et al., 2013).

Haplotype analysis

Haplotype analysis was performed as described previously (Li et al., 2014). Briefly, for the 194 accessions from the RegMap panel, SNPs located in the PHT1 loci (from PHT1;1 to PHT1;3) and PILS7 genes including a 3-kb promoter region were extracted (Horton et al., 2012). These SNPs were used as the input for fastPHASE version 1.4.0 (Scheet and Stephens, 2006). The results were analyzed and visualized in R version 3.5.3.

Analysis of public sequencing data

Raw sequencing data of accessions (Ag-0, Wt-5, Do-0, Kelsterbach-4, and Sorbo) were downloaded from the NCBI Sequence Read Archive (Leinonen et al., 2011, https://www.ncbi.nlm.nih.gov/sra/?term=SRP056687). Sequencing adapters and low-quality reads were trimmed with Trimmomatic version 0.32 (Bolger et al., 2014). The trimmed reads were mapped to the A. thaliana reference accession Col-0 genome (TAIR version 10) using HISAT2 version 2.1.0 and sorted using Samtools version 1.6 (Li et al., 2009; Kim et al., 2015). The aligned sequences of Bay-0 (TAIR version 10) were downloaded from the 1001 Genome project data center (Alonso-Blanco et al., 2016, http://1001genomes.org/projects/JGIHeazlewood 2008/). Aligned sequences were visualized using the Integrative Genomics Viewer (Thorvaldsdottir et al., 2013).

Plasmid construction and plant transformation

To generate 35S::PILS7 overexpression lines, binary plasmids were constructed using GATEWAY cloning technology (ThermoFisher Scientific, Waltham, MA, USA; Karimi et al., 2007). The coding sequence without the PILS7 stop codon was amplified from Col-0. Transgenic plants were selected on MS medium containing 50 µg mL⁻¹ kanamycin.

For complementation of the pils7-1 mutant, the Gibson Assembly Cloning Kit (New England Biolabs, Ipswich, MA, USA) was used for all constructs (Gibson et al., 2009). The PILS7 gene, along with a 1,928-bp promoter fragment according to the Col-0 reference genome, was amplified from HSm and Liarum genomic sequences. Primers used for cloning and sequencing of PILS7 genomic sequences from these two accessions are listed in Supplemental Table S8C. The amplified genomic fragments were assembled into the binary vector pCAMBIA1300 (Hajdukiewicz et al., 1994) linearized with EcoRI and HindIII (New England Biolabs). Transgenic plants were selected on 20 µg mL⁻¹ hygromycin-containing MS medium (Harrison et al., 2006).

All binary vector constructs were verified by sequencing (primers listed in Supplemental Table S8C) and transformed into Agrobacterium tumefaciens strain GV3130. The floral dipping technique was used to introduce all of the above constructs into the pils7-1 mutant (Clough and Bent, 1998).

Promoter analysis

To identify binding motifs for A. thaliana TFs, promoter sequences of HSm and Liarum PILS7 alleles obtained from amplified genomic fragments (see ‘Plasmid construction and plant transformation' section above) were used as input for the promoter analysis tool from PlantPAN version 3.0 (Chow et al., 2019). Binding motifs located on the sense strand of indels that discriminated between Hap alleles were chosen for downstream analyses.

RNA Isolation and RT-qPCR

Total RNA was isolated from root and shoot samples using the Spectrum Plant Total RNA kit with on-column DNaseI digest according to the manufacturer (Sigma-Aldrich, St Louis, MO, USA). The Tetro cDNA Synthesis Kit (Bioline London, UK) was used for cDNA synthesis using 1 µg of total RNA as input. Quantitative PCR was performed in a total reaction volume of 10 µL on the QuantStudio™ 12K Flex Real-Time PCR system (Applied Biosystems, Waltham, MA, USA). UBIQUITIN CONJUGATING ENZYME9 (UBC9, AT4G27960) and UBC21 (AT5G25760) were used as reference genes. Relative expression level was calculated using the 40-ΔCt method (Bari et al., 2006). Primers used for RT-qPCR are listed in Supplemental Table S8B.

Statistical analysis

Statistical analyses were performed in R version 3.5.3 using analysis of variance (ANOVA), followed by Tukey’s pairwise multiple comparisons of means. Unless stated otherwise, differences were considered significant at P < 0.05, detailed statistical reports can be found in Supplemental Table S9.

Accession numbers

Sequence data for the genes characterized in this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: AT4G28610 (PHR1), AT3G23430 (PHO1), AT2G33770 (PHO2), AT2G34202 (MICRORNA399D, MIR399D), AT5G43350 (PHT1;1), AT5G43370 (PHT1;2), AT5G43360 (PHT1;3), AT5G65980 (PIN-LIKES 7, PILS7), AT5G65990 (AMINO ACID VACUOLAR TRANSPORTER 3, AVT3), and AT5G66000 (unknown protein).

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. Experimental setup for accession screen.

Supplemental Figure S2. Shoot elemental composition and trait correlations in response to Pi availability.

Supplemental Figure S3. Location of genes is significantly associated with five key PSR traits.

Supplemental Figure S4. Characterization of T-DNA insertion mutants for “PHT1” and “PILS7” loci genes in Col-0.

Supplemental Figure S5. Impact of indels on PHT1;1 and PHT1;2 protein sequences in Hap2 accessions.

Supplemental Figure S6. Genomic sequence variation in PILS7 alleles from contrasting haplotypes.

Supplemental Figure S7. Impact of the amino acid sequence variation in contrasting haplogroups on PILS7 protein topology.

Supplemental Figure S8. Expression of PILS7 and key PSR genes in ten accessions from two distinct haplogroups.

Supplemental Figure S9. Natural variation in PHT1 and PILS7 loci corresponds to root fresh weight, organ Pi, shoot total P, iron, and copper concentrations.

Supplemental Figure S10. Generation and selection of PILS7 overexpression and pils7-1 complementation lines.

Supplemental Figure S11. Overexpression of PILS7 in the pils7-1 background does not restore seedling growth and root Pi levels.

Supplemental Table S1 . Information on A. thaliana accessions screened in this study.

Supplemental Table S2 . Summary of physiological and metabolic traits quantified in this study.

Supplemental Table S3 . Raw data of fresh weight, primary root length, Pi, and anthocyanin concentrations.

Supplemental Table S4 . List of GWAS candidate genes identified.

Supplemental Table S5 . Expression profile of GWAS candidate genes in RNA-seq data set of P-replete and P-limited Col-0 seedlings.

Supplemental Table S6 . Haplotype analysis of genomic sequences of PHT1 and PILS7 loci.

Supplemental Table S7 . Cis-element analysis of PILS7 promoters from HSm and Liarum accessions.

Supplemental Table S8. List of primers used in this study.

Supplemental Table S9. Statistical reports for this study.

 

R.J. and J.W. conceived the project. C.Y., X.W., and Q.C. characterized the phosphate starvation response of accessions. D.C. conducted the ICP-MS analyses. C.Y. carried out phosphate and anthocyanin assays, performed GWAS analyses, genotyped T-DNA mutants, generated transgenic germplasm, and characterized lines on a molecular and physiological level. C.Y., A.F.L., J.W. and R.J. interpreted results and drafted the manuscript. All authors reviewed the article.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys/pages/General-Instructions) is James Whelan (J.Whelan@latrobe.edu.au).