Introduction

The elevation of atmospheric CO₂ concentration leads to a decline in the mineral composition of C3 plants ¹. The negative effect of elevated CO₂ on plant mineral composition has been observed worldwide, and alters the content of nutrients that are essential for human nutrition, such as nitrogen (N) and proteins, iron (Fe) or zinc (Zn) ². The rise in atmospheric CO₂ thus poses a major threat to food security in the coming decades. The reasons why elevated CO₂ leads to the degradation of plant mineral composition are far from being well understood. To date, only a few genes with a potential regulatory effect on this mechanism have been identified ^3–7. These elements nevertheless converge towards the fact that the adaptation of plants to future high CO₂ climate can be achieved through the identification and the characterization of genetic mechanisms. In addition to this, several studies suggest that exploring the natural genetic variability of plants represents a major opportunity to understand the mechanisms by which high CO₂ leads to a decline in plant mineral composition ^8–10. Indeed, a significant diversity in the response of mineral composition to high CO₂ has been observed in several plant species. For protein and therefore N content, as well as for Fe or Zn content, substantial variations have been observed between small panels of genotypes from different species ^8–10. This implies the presence of a genetic diversity reservoir, which can facilitate the understanding of the ionome’s response to high CO₂ and subsequently provide an opportunity to alleviate this negative impact. However, in order to identify the genetic determinants of this negative response of the ionome to high CO₂, large-scale approaches are necessary, but are still lacking for the moment. The objective of this work was to fill the aforementioned knowledge gap by using a large collection of natural genotypes of the model plant Arabidopsis thaliana allowing to explore in depth the natural variation of the ionome response to elevated CO₂, and to generate a resource of phenotypic data that can be used in association genetics approaches. To this end, we used several hundreds of accessions from different geographic scales of A. thaliana, and analyzed their mineral composition under contrasted conditions of CO₂ concentration. This allowed us to extract the general trends in the ionome response to high CO₂, and to identify a large set of genes associated with the variation in the mineral composition of plants in response to high CO₂. By combining this information with genome expression data under elevated CO₂, we end up by functionally validating one of these genes for its importance in the reduction of Zn content under elevated CO₂, and therefore by demonstrating the relevance of this resource for future improvement of plant nutrient content under elevated CO₂.

Results

In order to explore the natural variation and identify its underlying genetic basis associated with the negative effect of elevated CO₂ on plant ionome, we used three populations of A. thaliana representing different geographic scales (i.e., the worldwide REGMAP population, the LANGUEDOC regional population and the local TOU-A population from east of France) and displaying different levels of genetic diversity (Fig. 1A). These populations were grown under ambient or elevated CO₂, and we measured in each accession the composition of their ionome in rosettes, including C, N, Na, Fe, Mg, Mn, Zn and Cu content.

Figure 1:

Elevated CO₂ globally decreases ionome content at the population level, whatever the geographic scale

In the three A. thaliana populations, we observed a global and important decrease of the ionome content when plants were grown under elevated CO₂ as compared to ambient CO₂. This was particularly the case for N and Fe, for which the decrease in content was very robust and important in each of the population analyzed (Fig. 1B-D). Zn, Cu and Mg content were also negatively affected to a significant extent by the growth under elevated CO₂ in the REGMAP and in the TOU-A populations (Fig. 1B, C), although not significantly in the LANGUEDOC population (Fig. 1D). More variability for the effect of elevated CO₂ was observed on Mn and Na content, which were decreased in the REGMAP population, but not significantly changed in the TOU-A and LANGUEDOC populations, respectively. In parallel, the C content of these populations increased under elevated CO₂, by very significant factors for the REGMAP and the LANGUEDOC populations. Altogether, these observations demonstrate that elevated CO₂ has on average a strong negative impact on the mineral content of natural genotypes of A. thaliana at the population-scale, whatever their geographic distribution.

The ionome of Arabidopsis thaliana natural accessions displays a high range of phenotypic diversity in response to elevated CO₂

To explore the effect of elevated CO₂ in each accession, we calculated the relative change in nutrient composition of A. thaliana accessions from the three populations in response to elevated CO₂. In agreement with the results previously mentioned, we observed that the median relative change of most nutrient content at the population-level was negatively affected by elevated CO₂ (Fig. 2). But the most striking observation was the genetic diversity of ionome response observed in these populations. Indeed, while most the natural accessions were negatively affected by elevated CO₂ (with a negative relative ratio of their nutrient content between ambient and elevated CO₂), a considerable number of accessions were rather not affected by elevated CO₂, or even positively affected, therefore showing an improved nutrient composition under elevated CO₂. For macronutrients like N, the relative change of concentration between ambient and elevated CO₂ varied from 20% to -50%, and for micronutrients like Cu, Fe or Zn, the relative change of concentration between ambient and elevated CO₂ varied from 100% to -60% (Fig. 2). In addition, some differences among nutrients were observed between populations. For instance, a smaller dispersion of Fe relative change in the LANGUEDOC population, against a higher distribution of Mn relative change.

Figure 2:

In order to explore the behavior of the different elements in response to elevated CO₂ and to observe the structure of phenotypic variation, we performed a principal component analysis (PCA) of the relative change in the 8 elements for the accessions from the three populations. The accessions from all populations seem to have globally similar responses to elevated CO₂, as suggested by the overlap of the three populations in the two first principal components (Fig. 3A). The first component of the PCA described a clear antagonistic trend between C content and the change of other mineral elements (Fig. 3B), suggesting that most of the variation between accessions in term of mineral response (almost 40%) could be driven by one or a few mechanisms resulting in an inverse variation between the whole ionome and C change (Fig. 3B). Interestingly, the second component, explaining almost 15% of the variation among accessions in term of mineral response, was mainly driven jointly by change in N and C concentration. Altogether, these results show that there is a marked and large variability among accessions in their mineral concentration in response to elevated CO₂, illustrated by accessions negatively affected by elevated CO₂ and others positively affected by elevated CO₂. In order to explore specific behavior of sub-populations, we clustered the accessions from the REGMAP panel via a k-means approach. This multivariate clustering resulted in the partitioning of accessions in three groups (Fig.4 – Suppl. Table 1). Cluster 1 displayed the most negative pattern of ionome response to elevated CO₂. Inversely, accessions included in Cluster 2 displayed a globally positive response, with the highest relative change for almost all mineral elements, except for C content. These accessions did not appear to be clustered geographically with respect to their collection origin in the REGMAP panel (Suppl. Fig. 1), which is in line with the high genetic diversity of response to elevated CO₂ observed at smaller geographical scales (Fig.2). Finally, Cluster 3 displayed a resilient pattern, with accessions showing a globally attenuated response to elevated CO₂. Interestingly, the large phenotypic diversity of the ionome observed in the three populations in response to high CO₂, as well as the presence of contrasted subpopulations in the REGMPA panel, suggests the presence of genetic determinants associated with this response.

Figure 3:

Figure 4:

Genetic architecture of the ionome response to elevated CO₂, and identification of genetic determinants

We run Genome-Wide Association (GWA) mapping to describe the genetic architecture of the ionome response to elevated CO₂, and to fine-map candidate genes underlying the detected quantitative trait loci (QTLs). We focused here on the phenotypic data collected on the REGMAP population, and used the sequencing data available for this population ¹¹. We included in this analysis the level of each mineral under ambient and under elevated CO₂, as well as the relative change between ambient and elevated CO₂ for each element. We also included a trait corresponding for each accession to the coordinate on the first and on the second PCA axes (PCA1 and PCA2) explaining collectively more than 50% of ionomic variation (Fig. 3). Therefore, these values correspond to traits driving and summarizing a large part of the ionome variation under elevated CO₂. This resulted as a whole in running GWA mapping on 30 different single-trait GWAS. The overall approach was first validated by observing expected results for traits phenotyped under ambient CO₂. For instance, we observed a very strong peak for the Na content at the locus of the HKT1 gene (Suppl. Fig. 2A), which is known to be involved in the natural genetic variation of Na content in Arabidopsis thaliana ¹², or a strong peak for the N content at the locus of the NIA1 gene (Suppl. Fig. 2B), encoding for an isoform of the nitrate reductase required for the first step of nitrate reduction and associated with natural genetic variation of N content in A. thaliana ¹³.

GWA mapping revealed a polygenic architecture for each phenotypic trait, although its complexity largely differs among traits. For instance, very few and neat peaks of association were detected Na and Mn content under elevated CO2, or of Fe and Cu relative change between ambient and elevated CO2 (Fig.5A, Suppl. Fig. 3). On the other hand, a more complex genetic architecture with the detection of a large number of QTLs was observed for traits related to N or C content (Fig.5A, Suppl. Fig. 3). For each of the traits that have been analyzed under elevated CO₂ or corresponding to the relative change of their content between ambient and elevated CO₂, we isolated the 50 SNPs with the most significant p-value, hereafter named top SNPs (Fig. 5A, Suppl. Tables 2 and 3). In order to identify the overlap between the genetic architecture of each trait, we looked whether some of the top SNPs were shared among traits. While the large majority of SNPs were specific to one trait, 30 and 21 SNPs were shared between two traits for the content under elevated CO₂ or for the relative change between ambient and elevated CO₂, respectively (Fig. 5B and C, Supplemental Tables 2 and 3). In addition, 8 and 2 SNPs were shared between three traits for the content under elevated CO₂ or for the relative change between ambient and elevated CO₂, respectively (Fig. 5B and C, Supplemental Tables 2 and 3). Most of the shared SNPs were associated with micronutrients (Fe, Mn, Zn and Mg content) and with N and/or with the first component of the PCA axis. An interesting QTL located on chromosome 1 was notably associated with 6 traits, displaying SNPs shared between Mn, Zn and N relative change and SNPs shared between Mn, N and PC1 content under elevated CO₂ (Fig. 5A, Suppl. Tables 2 and 3). Another QTL located on chromosome 3 encompasses SNPs shared between Fe, Zn and PC1 content under elevated CO₂ (Fig. 5A, Suppl. Table 2).

Figure 5:

We next identified for each trait a list of the genes located at ±25 kb from the top 50 SNPs, which corresponds to the rough estimate of the decay of linkage disequilibrium identified in A. thaliana at the worldwide scale ¹⁴. This resulted in a list of genes for each element, ranging from 154 to 422 genes depending on the element (Suppl. Tables 2 and 3). Among others, several genes associated with top 50 SNPs were identified as obvious candidates of the effect of elevated CO₂ on plant nutrition and ionome content. This was the case of ZINC INDUCED FACILITATOR 1 (ZIF1, AT5G13740) and ZIF-LIKE1 (AT5G13750), linked with SNPs identified for Zn content under elevated CO₂, and involved Zn sequestration mechanisms ¹⁵. We also noticed the link between SNPs identified for Zn relative change and TIP2;2 (AT4G17340), known to be involved in Zn root-to-shoot translocation ¹⁶. Concerning N relative change, some of the top 50 SNPs were linked to ASN1 (AT3G47340), which is an actor of N status and remobilization ^{17, 18}. The H+/CATION EXCHANGER 1 (CAX1) gene (AT2G38170), involved in the response to Mn deficiency, was also linked to SNPs associated with Mn content under elevated CO₂ ¹⁹. Some of the top 50 SNPs identified for Fe relative change were linked to MCO2 (AT5G21100) and MCO3 (AT5G21105) genes, which have been recently characterized as actors of the regulation of Fe homeostasis ²⁰. Finally, it is interesting to note that the QTL located on chromosome 3 mentioned above displaying significant shared SNPs identified for Fe, Zn and PC1 content under elevated CO₂ was associated among other genes with ISU2 (AT3G01020), coding for one of the Fe-S clusters in Arabidopsis thaliana, which are known to be essential for photosynthesis and metabolism ²¹. Altogether, this demonstrated that genes identified through this approach represent a large and valuable reservoir of candidates to study and to counteract the effect of elevated CO₂ on plant nutrition and ionome content.

To analyze how these genes identified by GWA mapping are regulated by elevated CO₂, we performed RNA-seq from shoots and roots grown under ambient and elevated CO₂.

Differentially expressed genes (DEG) associated to the effect of elevated CO₂ were identified from shoots and roots (Suppl. Table 4). We crossed the list of shoots or roots elevated CO₂- DEG with the list of genes identified by GWA mapping for each element, which resulted in a list of 182 genes identified by GWA mapping and differentially regulated by elevated CO₂ in shoot or in roots (Suppl. Table 5), making them strong candidates to be involved in the response of the mineral composition of plants to elevated CO₂. Most of these genes were deregulated by elevated CO₂ in shoot (Fig. 6A, B). In shoot or in roots, these genes mainly showed an association with C-, Mg- or Zn-related traits (Fig. 6A, B). Several of these genes, identified by GWA mapping and whose expression is deregulated in response to high CO₂, were known for their role in nutrient homeostasis. This was the case for the ASN1 and DUR3 genes, encoding an asparagine synthase and a urea transporter involved in N metabolism and remobilization, both associated here with a N-related peak of association, and whose expression is modulated by high CO₂ in leaves (Fig. 6C). We also observed in the leaves an interesting profile for several genes related to C metabolism and photosynthesis. This was the case for the BGP3 gene, involved in chloroplast development, or for the carbonic anhydrase CA1, both showing a decreased expression in response to high CO₂ and both associated with a peak in C-related GWA mapping under elevated CO₂ (Fig. 6C). In roots, the gene most deregulated in response to high CO₂ was AT1G64710, encoding a GroES-type alcohol dehydrogenase, which interestingly is also deregulated in leaves (Fig. 6D). We also observed in the roots that the expression of the TIP2;2 gene, associated here with a peak detected for Zn relative change GWA mapping, was deregulated in response to elevated CO₂ (Fig. 6D).

Figure 6:

To go further, we selected one of the association peaks identified by GWA mapping, and sought to functionally validate the importance of this QTL in response to elevated CO₂, in order to demonstrate the value of our data set and GWA mapping analyses. To do so, we selected an association peak located on chromosome 4 and associated with Zn relative change (Fig. 7A). More precisely, this association peak displayed the SNPs with the most significant p- values and more largely three SNPs that fell into the top 10 SNPs of the trait corresponding to the Zn relative change between ambient and elevated CO₂. The SNPs corresponding to the alternative alleles were associated to an increase of Zn content under elevated CO₂ (Fig. 7B). These SNPs are located very close to the TIP2;2 (AT4G17340) gene, which has been recently characterized as an actor of Zn root-to-shoot translocation ¹⁶. We thus selected a set of accessions from haplotype 0 (reduced Zn content under elevated CO₂) or haplotype 1 (increased Zn content under elevated CO₂), and analyze TIP2;2 expression in the roots under elevated CO₂. This analysis revealed a haplotype-specific difference in TIP2;2 expression under elevated CO₂, with accessions from haplotype 1 showing a reduced TIP2;2 expression in the roots compared to those from haplotype 1 (Fig. 7C), correlated with a higher Zn content in the shoot (Fig. 7B). To validate the effect of TIP2;2 expression of Zn content under elevated CO₂, we used the tip2;2-1 knock-out mutant and compared its Zn content under elevated CO₂ to this of the WT. We observed that the tip2;2-1 mutant line displayed a significant higher Zn content in the shoot under elevated CO₂ compared to the WT, confirming that TIP2;2 expression determines Zn content under elevated CO₂ (Fig. 7D). Altogether, these results demonstrated that these data sets generated in this study and the associated analyses are a valuable resource to identify genes able to counteract the general negative effect of elevated CO₂ on the mineral composition of plants.

Figure 7:

Discussion

The natural variation of ionome response to elevated CO₂ in Arabidopsis thaliana displays a high degree of genetic variation

In the present work, we analyzed the diversity of ionome response to elevated CO₂ present in the natural variation of Arabidopsis thaliana. In agreement with several other phenotypic traits related to phenology and disease resistance {Brachi, 2013 #1232;Huard-Chauveau, 2013 #1247;Roux, 2022 #1248}, we observed a wide range of responses at complementary geographical scales, from accessions with a ionome strongly negatively affected by high CO₂ to accessions with a ionome benefiting from high CO₂. This confirms for the first time on large and complementary sets of natural genotypes what has been observed by meta-analysis on isolated groups of plants worldwide ^{2, 9}. The global analysis of the distribution of each mineral element studied suggests firstly a trend where the whole ionome would evolve in a unified manner in response to high CO₂, and in an opposite manner to C. This is in line with a number of studies that have proposed that the accumulation of carbohydrates due to the stimulation of photosynthesis by high CO₂ would be the cause of the decrease in plant mineral composition ^22–25. However, the reading of the genetic architecture performed here by a genome-wide association genetics approach suggests that the majority of the genetic mechanisms underlying the negative effect of elevated CO₂ on the ionome are specific to each mineral element. Some specific cases, such as the QTL detected on chromosome 1 and associated with the natural genetic variation of 6 traits among the 20 considered, will certainly deserve a more in-depth analysis.

By clustering globally distributed accessions according to their ionome sensitivity to high CO₂, we were able to observe that the geographic origin of the accessions likely did not determine their response to CO₂. This suggests that inherent genetic factors, more than those due to local adaption, direct the response of plants to elevated CO₂. This seems consistent since the CO₂ elevation applied here to natural Arabidopsis thaliana variants does not correspond to any environment experienced by plants yet, at least for several tens of millions of years ^{26, 27}.

In this context of brutal and highly impactful environmental change, the presence of cryptic genetic variation often explains the appearance of relatively rapid adaptive mechanisms ^{28, 29}. Although not formally tested here, it would be interesting to examine whether the variation in the ionome in response to elevated CO₂ shows evidence of cryptic variation. In any case, the presence of high phenotypic diversity in these natural populations of A. thaliana demonstrates very clearly the possibility of taking advantage of this genetic variation to understand and alleviate the negative response of plant mineral composition to high CO₂.

GWA mapping of ionome variation under elevated CO₂ identified a large number of genes to understand and mitigate the negative effect of high CO₂ on plant mineral composition

In order to understand the genetic mechanisms underlying the effect of high CO₂ on plant mineral composition, and to enable future breeding approaches, we adopted an association genetics approach. This led to the identification of a large number of candidate genes associated to the variation of nutrients under elevated CO₂. Several genes in this list can easily attract attention. In particular, we can note the identification of ASN1 and DUR3 genes in two of the loci associated with N content variation under elevated CO₂. ASN1, and to a lesser extent DUR3, play an important role in the remobilization and the reallocation of N within the plant, and their manipulation can lead to variation in N use efficiency ^{17, 18, 30}. This is interesting because for the moment, root N uptake and N assimilation seemed to be the key targets of the negative effect of high CO₂ on plant N content ^{4, 31}, but these results suggest that remobilization of N may also be involved. We also identified the CA1 gene, coding for a carbonic anhydrase, in the vicinity of a QTL associated with C variation under high CO₂. CA1 is involved in the regulation of stomatal opening by elevated CO₂ ³², and the α carbonic anhydrase family of which CA1 belongs is involved in the regulation of photosynthetic efficiency, although CA1 shows no significant effect under standard conditions ³³. It would be therefore interesting to assess the role of CA1 natural genetic variation under elevated CO₂. If CA1 regulates the C variation of the ionome under elevated CO₂, this could, according to our observations, significantly influence the global mineral composition of plants. Interestingly, the genes identified by GWA mapping in the ionome response to high CO₂, including those mentioned above, showed substantial variation at the gene expression level. We ended this study with the functional validation of an association peak identified by GWA mapping for the relative change of Zn content between ambient and elevated CO₂. Zn is an essential element for a large number of metabolic processes in humans, and Zn deficiency, found in up to one third of the world’s population, leads to severe health problems. We demonstrated that TIP2;2 gene expression varied in response to CO₂ in a haplotype-specific manner. Consistent with these results, we show that manipulating TIP2;2 expression with a knock-out mutant can modulate the Zn loss observed under high CO₂. A recent study demonstrated that TIP2;2 was responsible for Zn retention in the roots {Wang, 2022 #1177}. It therefore seems consistent that natural accessions with the lowest expression levels of this gene are those with the highest Zn content in aerial parts, due to low retention in their roots. This example illustrates the potential of the resource we have generated here towards the development of biofortified plants. The development of biofortified plants represents a considerable challenge in view of the current problems of malnutrition on a global scale, and this challenge becomes even more important in a context of rising atmospheric CO₂ ³⁴. This reservoir of data and genes will certainly contribute to the understanding of the mechanisms underlying the general negative effect of CO₂ on mineral composition, and to the development of crop plants adapted to forthcoming high-CO₂ climate.

Methods

Data and code availability

Data and R notebooks containing the analyses performed in this article can be found at https://src.koda.cnrs.fr/groups/ipsim/sirene-team. RNA-seq data generated for this study are available at https://www.ebi.ac.uk/biostudies/arrayexpress/studies using the accession no xxx.

Plant Material

A subset of the REGMAP panel, the LANGUEDOC panel and the TOU-A panel were used in this study. These populations were previously described here {Brachi, 2013 #1232;Frachon, 2017 #1249;Horton, 2012 #1231}. These populations were grown on Jiffy-7 peat pellets (Jiffy Products International, NL) under ambient (∼420 ppm) or elevated (900 ppm) CO₂ in the growth chambers of the Microcosms experimental platform at the Montpellier European Ecotron CNRS. Growth conditions were 6-h/22-h light (22°C) / dark (20°) photoperiod, with 200 µmol m^-2 s^-1 light intensity and 65% of hygrometry. Plants were watered twice a week with a growth solution containing KH₂PO₄ 1 mM, MgSO₄ 1 mM, K₂SO₄ 250 µM, CaCl₂ 250 µM, Na-Fe-EDTA 100 µM, KNO₃ 10 mM, KCl 50 µM, H₃BO₃ 30 µM, MnSO₄ 5 µM, ZnSO₄ 1 µM, CuSO₄ 1 µM, (NH₄)₆Mo₇0₂₄ 0,1 µM, as described by ³⁵. The entire rosettes were collected three weeks after sowing. The tip2;2-1 mutant line corresponds to the SALK_152463 allele ¹⁶.

Ionome analysis

From 3 to 5 replicates per accession were used for each ionome analysis. Total C and N content was obtained from dried shoot tissue using an Elementar Pyrocube analyzer. Cu, Fe, Mg, Mn, Na and Zn content was obtained from dry shoot tissue mixed with 750 µl of nitric acid (65% [v/v]) and 250 µl of hydrogen peroxide (30% [v/v]). After one night at room temperature, samples were mineralized at 85°C during 24 hours. Once mineralized, 4 ml of milliQ water was added to each sample. Mineral contents present in the samples were then measured by microwave plasma atomic emission spectroscopy (MP-AES, Agilent Technologies).

Removal of outlier observations

Prior to GWAS and multivariate analyses such as PCA or clustering, mineral composition measures were pre-processed to remove technical outliers. For a given element and CO₂ condition, the values positioned more than 5 median absolute deviations away from the median were removed from the dataset.

PCA and Clustering

Principal Component Analysis was performed using the R ade4 package after the prior scaling of the variables to a z-score. Clustering of the REGMAP panel based on the relative changes of the mineral composition of each accession has been done using a k-means clustering with the R kmeans function. For this step, the variables were also scaled to a z-score. The number of clusters in the k-means algorithm was chosen by the elbow method on the criteria of cluster homogeneity (within-sum of squares).

GWAs

Genome-Wide Association mapping was performed using the R statgenGWAs package. Genotype data was prepared using the codeMarkers function, removing duplicated SNPs and filtering for a minimum allele relative frequency of 0.04. Associations were performed by the runSingleTraitGwas function, that implements the EMMA algorithm. Population structure was modeled via a kinship matrix built from the Astle method. Manhattan plots were drawn using the manPlotFast function of the ramwas R package.

RNA-seq experiments

Plants were grown in hydroponics to have access to the roots in addition to the shoot, as previously described in ⁴. Shoot or root from 5 plants were pooled into one biological replicate, flash frozen in liquid nitrogen, and stored at -80◦C. RNA of three biological replicates were extracted from shoot or root tissues using Direct-zol RNA Miniprep (Zymo Research, CA, USA), according to the manufacturer recommendations. RNA-sequencing libraries were done from shoot or root total RNA using standard RNA-Seq protocol method (Poly-A selection for mRNA species) by the Novogene company. RNA-sequencing was performed using Illumina technology on a NovaSeq6000 system providing PE150 reads. The quality control and adapter trimming of raw paired-end fastq files was done with fastp and its default parameters. Mapping to the TAIR10 reference genome was performed with STAR, and using the following options:

--outSAMtype BAM SortedByCoordinate

--outFilterMismatchNmax 1

--outFilterMismatchNoverLmax 0.15

--alignIntronMin 30

--alignIntronMax 5000

Quantification of the bam files against the TAIR10 GFF3 annotation file was done using htseq- count with options:

-f bam --type gene -r pos

--idattr=Name --stranded=no

Normalization and differential expression were performed using DIANE R package ³⁶, with no fold change constraint, and an adjusted p-value threshold (FDR) of 0.05. Lowly expressed genes with an average value across conditions under 25 reads were excluded from the analysis.

Quantitative real-time PCR

Plants were grown in hydroponics to have access to the roots, as previously described in ⁴. Root tissue from 5 plants were pooled into one biological replicate, flash frozen in liquid nitrogen, and stored at -80◦C. RNA were extracted from shoot or root tissues using TRIZOL (Invitrogen, USA), according to the manufacturer recommendations, and DNAse treated using RQ1 (Promega, USA). Reverse transcription was achieved from 1 μg of total RNA with M-MLV reverse transcriptase (RNase H minus, Point Mutant, Promega, USA) using an anchored oligo(dT)20 primer. Accumulation of transcripts was measured by qRT-PCR (LightCycler 480, Roche Diagnostics, USA) using the SYBR Premix Ex TaqTM (TaKaRa, Japan). Gene expression was normalized using UBQ10 and ACT2 as internal standards. Results are presented as the expression relative to UBQ10. Sequences of primers used in RT-qPCR for gene expression analysis are listed in Supplemental Table 6.

Acknowledgements

This work was supported by the I-Site Montpellier Université d’Excellence (MUSE; project ECO2THREATS), the CNRS through the Mission for Transversal and Interdisciplinary Initiatives (MITI) 80 PRIME program, and the Plant Biology and Breeding department of INRAE. O.C. was recipient of a PhD fellowship from the CNRS. T.M. was recipient of a PhD fellowship from INRAE and Région Occitanie. We thank Jiping Liu from Cornell University for the gift of tip2;2- 1 mutant line. This study benefited from the CNRS human and technical resources allocated to the Ecotron Research Infrastructure and from the state allocation “Investissement d’Avenir” AnaEE France ANR-11-INBS-0001. A CC-BY public copyright license has been applied by the authors to the present document.