A glutathione-dependent control of the indole butyric acid pathway supports Arabidopsis root system adaptation to phosphate deprivation

Arabidopsis root responses to indole butyric acid (IBA), but not IAA, depend on glutathione levels, and endogenous IBA and glutathione are both required for root system adaptation to phosphate starvation.


Introduction
Root developmental plasticity, combining root growth and root branching, is essential for plants to adapt and optimize their growth in changing environmental conditions, such as nutrient and water availability, rhizosphere microbiome, or soil structure heterogeneity. Root growth relies on the activity of the root apical meristem that regulates histogenesis via the control of cell proliferation. Root branching is a complex organogenesis process that allows the development of new lateral roots (LRs) from regularly spaced pericycle founder cells in Arabidopsis thaliana. These founder cells are originally specified by an oscillatory mechanism occurring in the root tip (Moreno-Risueno et al., 2010;Xuan et al., 2016). In addition to the root system architecture, the development of epidermal root hairs (RHs) is particularly sensitive to changes in environmental conditions, and they contribute to the root system adaptation (Grierson et al., 2014).
Among nutrients, phosphorus has a key role in all living organisms and is one of the main limiting factors for plant growth and crop productivity (Leinweber et al., 2018). Its homeostasis is highly dependent on phosphate uptake by the root cap (Kanno et al., 2016). Phosphate concentration strongly impacts root system development in many plant species including Arabidopsis (López-Bucio et al., 2002, 2003Hodge, 2004). In particular, general low phosphate availability is well known to increase RH length (Bates and Lynch, 1996). Biotic factors also modulate root development. Among them, some rhizospheric PGPR (plant growth-promoting rhizobacteria) such as Mesorhizobium loti are able to stimulate RH elongation, hence favouring soil exploration and nutrient uptake by the plant (Vacheron et al., 2013;Poitout et al., 2017).
Auxins, and particularly the most abundant endogenous auxin, indole acetic acid (IAA), play a pivotal and integrative role in all steps of root system development, under both optimal and stress conditions (Saini et al., 2013;Korver et al., 2018). During LR development, auxin is responsible for the activation of founder cells, the development and organization of the LR primordia (LRPs), and the emergence of the newly formed LRs through the external layers of the primary root. Even earlier, the oscillatory positioning of pericycle founder cells depends on auxin maxima generated via auxin release by dying root cap cells (Moreno-Risueno et al., 2010;Du and Scheres, 2018). Auxin also modulates RH elongation, thus mediating responses to stimuli such as abscisic acid or phosphate deprivation (Nacry et al., 2005;Wang et al., 2017). Recently, a set of works on Arabidopsis and rice evidenced that RH response to phosphate deprivation requires an increase in auxin de novo synthesis in the root cap, and its apico-basal transport to the epidermal cells via the auxin influx facilitator AUX1 Giri et al., 2018;Parry, 2018).
Indole-3-butyric acid (IBA) is a structural derivative of IAA, differing by only two additional carbons in the side chain (Korasick et al., 2013). Although convincing evidence supports a role for IAA in IBA biosynthesis, the mechanisms responsible and enzymes involved are still unknown (Ludwig-Müller et al., 1995;Frick and Strader, 2018). It is now broadly accepted that IBA solely acts as an IAA precursor through its β-oxidative decarboxylation in peroxisomes (reviewed in Frick and Strader, 2018). This enzymatic process involves several enzymes, some shared with other β -oxidation pathways such as PED1, and others apparently specific to the IBA to IAA conversion, such as IBR1, IBR3, IBR10, and ECH2 Frick and Strader, 2018). IBA homeostasis is also regulated via its transport and conjugation, but only a few regulators have been identified. The type G ABC transporters ABCG36 and ABCG37 can efflux IBA from the cells (Strader and Bartel, 2009;Ruzicka et al., 2010), the NPF family member TRANSPORTER OF IBA1 transports IBA across the vacuolar membrane (Michniewicz et al., 2019), while the generic PXA1/COMATOSE transporter is likely to import IBA into the peroxisome. Like other auxins, IBA conjugates with sugars and amino acids. UGT84B1, UGT74D1, UGT74E2, and UGT75D1 can conjugate IBA to glucose (Jackson et al., 2001;Tognetti et al., 2010;Jin et al., 2013;Zhang et al., 2016), whereas GH3.15 is able to conjugate IBA to amino acids (Sherp et al., 2018). IBA-derived IAA plays important roles during root development, including root apical meristem maintenance and adventitious rooting. IBA conversion to IAA is also critical for RH elongation . Finally, recent works have reported the critical role of IBA to IAA conversion in the root cap as a source of auxin for the oscillatory positioning of LR founder cells (De Rybel et al., 2012;Xuan et al., 2015). IBA to IAA conversion taking place in the LRP itself also probably participates in further LR development Michniewicz et al., 2019). Despite all these reported functions in root development, the importance of IBA-derived IAA relative to other IAA sources is still poorly understood and scarcely documented, particularly in the case of changing environmental constraints.
Changes in environmental conditions result in reactive oxygen species (ROS) imbalance, thus affecting the general redox cellular homeostasis (Noctor et al., 2018). For example, phosphate deficiency alters H 2 O 2 and O 2 ·production in roots (Tyburski et al., 2009). ROS modulate many aspects of plant development, including root system development (Singh et al., 2016;Tsukagoshi, 2016). Controlled ROS production by NADPH oxidases is particularly involved in RH elongation and LR development (Carol and Dolan, 2006;Mangano et al., 2016Mangano et al., , 2018Orman-Ligeza et al., 2016). Auxin and ROS pathways interact to control root development, since auxin can trigger ROS production and in turn ROS can affect IAA levels and transport (Orman-Ligeza et al., 2016;Tognetti, 2017;Zwiewka et al., 2019).
Glutathione is a small and stable thiol-containing tripeptide (Glu-Cys-Gly) essential for plant survival and present in large concentrations in cells, up to several millimolar (Noctor et al., 2012). It has many roles, including the detoxification of heavy metals and xenobiotics, sulfur homeostasis, ROS homeostasis, and redox signalling. To ensure these functions, glutathione can act as a precursor (e.g. of phytochelatins), be conjugated to other molecules via the glutathione S-transferase enzyme family, or serve as an electron donor for antioxidant systems (Noctor et al., 2012). Reduced glutathione (GSH) is therefore converted to oxidized glutathione (GSSG) which is reduced back by glutathione reductases (Marty et al., 2009(Marty et al., , 2019. In standard growth conditions, GSH is in large excess relative to GSSG. Glutathione biosynthesis is a two-step reaction, with GSH1 encoding the rate-limiting γ-glutamylcysteine synthetase and GSH2 encoding a glutathione synthetase (Meyer et al., 2012;Noctor et al., 2012). Knock-out mutations in either of these two genes are lethal. However, genetic screens identified several knock-down alleles of GSH1, allowing plants with reduced glutathione levels to be obtained (Howden et al., 1995;Vernoux et al., 2000;Ball et al., 2004;Parisy et al., 2007;Jobe et al., 2012;Shanmugam et al., 2012). The importance of glutathione in root development is illustrated by the absence of root meristem maintenance in the rml1 mutant, together with the impairment of primary root growth and LR development in less severe gsh1 alleles (Vernoux et al., 2000;Bashandy et al., 2010;Koprivova et al., 2010;Marquez-Garcia et al., 2014).
We previously reported that cad2 or pad2 mutants can respond almost normally to exogenous IAA for root development (Bashandy et al., 2010), although auxin transport seems to be affected in glutathione-deficient plants (Bashandy et al., 2010;Koprivova et al., 2010). The role of glutathione in controlling root development is therefore still misunderstood. Given the importance of IBA-derived IAA in regulating root development, we chose to address the role of glutathione in root responses to IBA. We show that glutathione deficiency impairs LR and RH responses to IBA but not IAA. We then try to identify the glutathione-dependent mechanisms required for IBA responses, but every IBA-related pathway examined does not reveal sensitivity to glutathione levels. We finally suggest a physiological role for this glutathione-dependent IBA response, showing that IBA and glutathione pathways are critical for RH responses to phosphate deficiency.

Plant cultures
Seeds were surface sterilized by constant agitation with 0.05% SDS in 70% (v/v) ethanol for 20 min, then washed three times with 95% (v/v) ethanol and dried on sterile paper. Seeds were placed on plates containing 50 ml of half-strength Murashige and Skoog (1/2 MS) medium with 0.5 g l -1 MES and 0.8% (w/v) plant agar (Duchefa Biochemie) without sucrose, unless indicated. For the experiments regarding phosphate deprivation, we used three-tenths strength (3/10) MS supplemented with 500 µM NaCl (phosphate deprivation) or 500 µM NaH 2 PO 4 (control conditions) as previously published (Arnaud et al., 2014). For M. loti experiments, seeds were grown on standard 1/2 MS medium for 5 d, then transferred to new plates containing 0.1 OD of inoculum (Poitout et al., 2017) for an extra 4 d before phenotyping. All plates were incubated vertically at 20 °C with 160 µE m -2 s -1 light intensity and a 16 h light/8 h dark regime.
Phenotypic analyses of the root system LR density was quantified on 10-day-old seedlings by counting the number of visible LRs emerged and dividing by the length of the primary root section between the first and the last visible LRs. LRP density was calculated in the same way on 6-day-old seedlings using a light microscope (Axioscop2, Zeiss) for counting. For gravistimulation experiments, LRP initiation was forced by rotating the vertical plates by 90°. LRP stages were observed 48 h later, using a light microscope (Axioscop2, Zeiss). For quantification of RH lengths, plates were photographed using a camera (DFC425C, Leica) sited on a stereomicroscope (MZ12, Leica). The primary root elongation rate was quantified between day 8 and day 10. Lengths were quantified from pictures using the public domain image analysis program ImageJ 1.52i (https://imagej.nih.gov/ ij/) and its NeuronJ plugin (Meijering et al., 2004).

Glutathione measurements
Total glutathione content of 8-day-old seedlings was determined using the recycling enzymatic assay (Rahman et al., 2006). The method consists of the reduction of GSSG by glutathione reductase and NADPH to GSH. GSH levels are determined by its oxidation by 5,5′-dithio-bis(2nitrobenzoic acid) (DTNB), that produces a yellow compound 5′-thio-2-nitrobenzoic acid (TNB), measurable at 412 nm (Rahman et al., 2006). Briefly, 100 mg of fresh plant material ground in liquid nitrogen was extracted in 0.5 ml of 0.1 M Na phosphate buffer, pH 7.6, and 5 mM EDTA. After microcentrifugation (10 min, 9000 g), total glutathione in 0.1 ml of the supernatant was measured by spectrophotometry in a 1 ml mixture containing 6 mM DTNB (Sigma-Aldrich), 3 mM NADPH, and 2 U of glutathione reductase from Saccharomyces cerevisiae (Sigma-Aldrich). Glutathione-dependent reduction of DTNB was followed at 412 nm. Total glutathione levels were calculated using the equation of the linear regression obtained from a standard GSH curve. GSSG was determined in the same extracts after derivatization of reduced GSH. Derivatization of 100 ml of plant extract was performed in 0.5 ml of 0.5 M K phosphate buffer, pH 7.6, in the presence of 4 ml of 2-vinylpyridine (Sigma-Aldrich) during 1 h at room temperature. After extraction of the GSH-conjugated 2-vinylpyridine with 1 vol. of diethylether, GSSG was measured by spectrophotometry as described for total glutathione.

Confocal analyses
Auxin response analyses using DR5:n3EGFP/DR5v2:ntdTomato (Liao et al., 2015) and the study of the redox state of glutathione with the roGFP2 line (Meyer et al., 2007) were performed by using a confocal laser scanning microscope LSM 700 (Zeiss). Images were acquired in 16 bits using a ×10 EC Plan Neofluar objective (Zeiss). Settings were based on the respective methods previously published (Meyer et al., 2007;Liao et al., 2015) with minor modifications. Excitation of roGFP2 was performed at 488 nm and 405 nm, and a bandpass (BP 490-555 nm) emission filter was used to collect the roGFP2 signal. For background subtraction, the signal was recovered using a BP 420-480 nm emission filter during excitation at 405 nm.
For DR5:n3EGFP/DR5v2:ntdTomato analysis, seeds were grown on standard 1/2 MS medium supplemented or not with 0.5 mM buthionine sulfoximine (BSO; pre-treatment). Seven-day-old seedlings were transferred for 24 h to plates containing the appropriate treatments (10 µM IBA or 50 nM IAA), still combined or not with 0.5 mM BSO, according to the pre-treatment. Regarding the analyses with roGFP2, seeds were grown for 8 d in 1/2 MS medium with or withoutt 10 µM IBA.

Gene expression quantification
Total RNA was extracted using TriZol reagent (GE Healthcare, UK), and the RNA was purified with the RNeasy Plant Mini Kit (Promega, USA), according to the manufacturer's protocols. cDNA was subsequently synthesized using the GoScript™ Reverse Transcription System (Promega, USA). Quantitative real-time PCR was done using Takyon™ No Rox SYBR® MasterMix blue dTTP (Eurogentec, Belgium) and the LightCycler 480 (Roche, Switzerland). Primers used are presented in Supplementary Table S1 at JXB online. All reported results are presented normalized with the ACTIN2 control gene but behave similarly if normalized with ACTIN7 or GAPDH control genes.

Auxin accumulation assays
Root tips (5 mm) from 8-day-old seedlings were excised and incubated in 40 µl of uptake buffer (20 mM MES, 10 mM sucrose, and 0.5 mM CaSO 4 , pH 5.6) for 15 min at room temperature. Root tips were then incubated for 1 h in uptake buffer containing 25 nM [ 3 H]IAA (20 Ci mmol -1 ; American Radiolabeled Chemicals) or [ 3 H]IBA (25 Ci mmol -1 ; American Radiolabeled Chemicals) prior to washing with three changes of uptake buffer. Root tips were then placed in 3 ml of CytoScint-ES liquid scintillation cocktail (MP Biomedicals) and analysed by scintillation counting.

Data replicability and statistical analyses
All the presented experiments illustrate results obtained in at least three independent biological repetitions. For in vitro phenotyping experiments, each biological repetition consisted of two technical replicate plates per condition, each containing ~12 seeds for the mutant and 12 for the appropriate control, sown side by side. Statistical analyses were performed as indicated in the figure legends.

Accession numbers
Sequences from this article can be found in the Arabidopsis Genome Initiative database with the following accession numbers:

Glutathione deficiency specifically alters LR and RH responses to IBA but not IAA
Since IBA plays important roles during root development, we investigated root responses to both IAA and IBA in several gsh1 weak alleles and in plants treated with BSO, a specific chemical inhibitor of GSH1.
We first wanted to know the precise glutathione levels in the different genotypes in our growth conditions, thus we quantified endogenous glutathione in whole 8-day-old seedlings (Fig. 1A). We find the same amount of total glutathione (i.e. ~25% of the wild-type content) in cad2 and pad2 mutants, which was expected. We show that 1 mM exogenous BSO also reduces endogenous glutathione levels of wild-type plants to approximately the same levels. In addition, we report that exogenous IBA treatment does not impact endogenous glutathione levels.
Supplementary Fig. S1A and B shows that cad2 and pad2 mutants display the same primary root growth as the wild type in our growth conditions, and that the primary root responds normally to both IAA and IBA. In the same way, BSO treatment (1 mM) affects neither primary root growth nor its response to IAA or IBA.
As expected, both IAA and IBA also induce LR density in the mature zone of the root in 10-day-old seedlings (Fig. 1B). LR density in both cad2 and pad2 mutants responds normally to IAA treatment but displays hyposensitivity to exogenous IBA. As expected, the addition of BSO phenocopies cad2 and pad2 mutants (Supplementary Fig. S2A). Finally, we could revert cad2 hyposensitivity to IBA by adding exogenous GSH, while it has no effect on the sensitivity of wild-type plants to IBA ( Supplementary Fig. S2B). Because the final emerged LR density depends both on LR initiation and subsequent LRP development, we also addressed the LRP density in the same conditions. Figure 1C reveals that both LRPs and emerged LR densities display IBA hyposensitivity in plants with low glutathione levels, suggesting that glutathione affects LR development upstream of LRP development. Another way to investigate LRP development is to force and synchronize LRP initiation by gravistimulation (Péret et al., 2012). We therefore gravistimulated glutathione-deficient plants, in the absence or presence of exogenous IBA (10 µM). We observe that exogenous IBA treatment slightly slows down wild-type LRP development upon gravistimulation in our growth conditions (Fig. 1D). We also observe that cad2, pad2, and BSO-treated plants respond to IBA in a similar way to the wild type. This suggests that, once established, the regulation of LRP development by IBA is independent of glutathione, at least in its later stages. Taken together, these results tend to support a role for glutathione in IBA-dependent specification or activation of founder cells necessary for the initial establishment of LRPs.
Finally, we monitored RH elongation responses to IBA and IAA in glutathione-deficient plants (Fig. 1E, F). As for LRs, we notice that cad2 and pad2 mutants are hyposensitive to the IBA-dependent induction of RH growth but respond normally to IAA. Similarly, 1 mM BSO treatment also reduces the RH response to IBA but not IAA. We can therefore conclude that glutathione is also required for the IBA-dependent induction of RH growth.

Glutathione levels affect auxin signalling in the basal part of the meristem
We have shown that glutathione deficiency alters RH elongation and LRP density in response to IBA. We know that root tip-derived IAA transits through the LR cap to regulate both founder cell positioning and RH growth in the basal part of the meristem. We therefore investigated auxin response in the root tip, by using the DR5:GUS auxin signalling reporter (Ulmasov et al., 1997). In standard growth conditions, we only reveal GUS staining in the quiescent centre and the columella, and BSO treatment by itself does not change the DR5:GUS expression pattern ( Fig. 2A). As expected, exogenous IAA treatment for 24 h induces auxin response in the whole root tip and root epidermis, and the presence of BSO again has no effect on this distribution. In response to IBA, GUS staining increases in the distal meristem and strongly appears specifically in the trichoblast epidermal cell files in the basal part of the meristem, up to the differentiation zones. In the presence of BSO, this IBA-dependent strong signal in the basal meristem almost disappears while the signal in the distal meristem remains strong.
In order to confirm these results, we used another auxin signalling marker that allows the quantification of the auxin response, the DR5:n3EGFP/DR5v2:ntdTomato double reporter line (Liao et al., 2015). We quantified both green fluorescent protein (GFP) and Tomato fluorescence in the distal and basal parts of the meristem (Fig. 2B). In agreement with the DR5:GUS reporter line, both IBA and IAA treatments increase auxin signalling in the distal and basal parts of the meristem. The presence of BSO does not significantly alter auxin signalling, either in standard conditions or upon exogenous IAA treatment. However, we can confirm that BSO severely represses IAA signalling in the basal part of the meristem in response to exogenous IBA, while it has limited effect in the distal part.
Hence we show that glutathione is specifically required for IBA-derived IAA signalling in the basal part of the meristem, where LR founder cells are specified and RH growth is determined. Moreover, we show that glutathione does not affect IAA signalling components since auxin signalling reporters respond normally to exogenous IAA when glutathione content is depleted.

IBA-derived IAA responses are specifically affected by glutathione deficiency
Because glutathione is a critical regulator of cellular redox homeostasis, we addressed root responses to IBA in other redox-related mutants. We chose to analyse the gr1 mutant, affected in cytosolic and peroxisomal glutathione reduction (Marty et al., 2009), the cat2 mutant, affected in H 2 O 2 detoxification (Queval et al., 2007), and the ntra ntrb double mutant, affected in the thioredoxin-dependent thiol reduction system (Reichheld et al., 2007). Quantification of both GSH and GSSG (Fig. 3A) shows that gr1, ntra ntrb, and cat2 mutant plants have higher total glutathione concentrations than wild-type plants. The application of exogenous IBA has almost no effect on both GSH and GSSG levels. As expected, cat2 and gr1 mutants also display a higher glutathione oxidation status. In addition, the use of the roGFP2 probe allowed us to confirm that the presence of IBA does not generate any imbalance in the glutathione redox status in root tissues ( Fig. 3B; Supplementary  Fig. S3). In contrast to cad2, none of the other mutants display any LR density hyposensitivity to IBA (Fig. 3C). These results suggest that root responses to IBA specifically depend on glutathione overall levels rather than glutathione redox status or any general redox imbalance.

IAA transport from the distal to the basal meristem is not targeted by glutathione
We have presented data showing that the IBA-derived IAA response in the basal meristem is dependent on glutathione levels. AUX1 and PIN2 are the IAA transporters that ensure the apico-basal IAA flux in the root cap and the root tip epidermis. We previously published our results showing that strong BSO treatment induces a long-term decrease in the expression of AUX1 and PIN2 (Bashandy et al., 2010).
In order to determine if AUX1 and/or PIN2 are the targets of glutathione to modulate root responses to IBA, we examined LR density in aux1 and pin2 mutants. As shown in Fig. 4A, aux1 and pin2 mutants still display hyposensitivity to IBA upon BSO treatment. In other words, the glutathione regulation still occurs in both aux1 and pin2 mutants, revealing that neither AUX1 nor PIN2 is regulated by glutathione to control root responses to IBA. Because AUX1 and PIN2 are members of multigene families, we wanted to ensure that other members of AUX/LAX or PIN families are not replacing AUX1 and PIN2 in the respective mutants. We therefore addressed IBA responses with or without BSO treatment in the presence of specific inhibitors of both families (Fig. 4B). We observed that BSO still leads to LR hyposentitivity to IBA both in the presence of N-1-naphthylphthalamic acid (NPA), that inhibits PIN-dependent auxin efflux, and in the presence of 1-naphthoxyacetic acid (NOA), a specific inhibitor of AUX/LAX influx facilitators. All together, these data suggest that the glutathione-dependent control of root responses to IBA does not affect IAA transport from the root apex to the basal meristem.

Looking for glutathione targets in IBA pathways.
Since IAA transport was not the target of glutathione, we investigated IBA homeostasis in plants with low glutathione content. Interestingly, gsh1 mutants display pleiotropic phenotypes opposite to the phenotypes of mutant alleles in PDR8/PEN3/ABCG36 for IBA sensitivity, but also for sensitivity to Pseudomonas and cadmium treatments (Kim et al., 2007;Strader and Bartel, 2009;Lu et al., 2015). ABCG36, together with its homologue PDR9/ABCG37, is responsible for IBA efflux from the cells, which prompted us to investigate the IBA import into plant cells in the cad2 mutant. As shown in Fig. 5A, we did not detect any impairment in [ 3 H]IBA accumulation in cad2 excised root tips, in contrast to the expected reduction of uptake in a gain-of-function allele of PDR9 (pdr9-1). This suggests that IBA uptake from the medium is not perturbed by glutathione deficiency. This is consistent with the normal response of primary root growth to IBA in such conditions ( Supplementary Fig. S1A, B).
In order to better investigate IBA transport, we analysed the glutathione-dependent LR and RH response to IBA in abcg36/pdr8 and abcg37/pdr9 mutants ( Fig. 5B; Supplementary  Fig. S4). As expected, the pdr8 mutant displays hypersensitivity to exogenous IBA. However, both mutants are still clearly resistant to IBA in the presence of BSO. Because of putative redundancy between these two proteins, we generated the pdr8 pdr9 double mutant that displays a pdr8-like hypersensitivity of LR density to exogenous IBA. As for single mutants, LR density and RH length in the double mutant are induced by IBA, but BSO is still able to decrease this response. This result means that IBA efflux transporters ABCG36 and ABCG37 are not the targets of glutathione-dependent regulation.
We know that IBA homeostasis is also regulated via conjugation with glucose, and several glycosyltransferases are able to catalyse such a reaction (Jackson et al., 2001;Tognetti et al., 2010;Jin et al., 2013;Zhang et al., 2016). Among them, we assayed ugt74d1 and ugt74e2 mutants. Again, we could observe that LR and RH responses to IBA are still BSO sensitive in both mutants, although ugt74d1 LR density is hypersensitive to IBA compared with the wild type ( Fig. 5C; Supplementary  Fig. S4). Because of putative redundancy between these two glycosyltransferases, we generated a ugt74e2 ugt74d1 double mutant. However, the response to IBA is still reduced in the presence of BSO in the double mutant, suggesting that these glycosyltransferases are not the targets of glutathionedependent regulation.
Finally, we also investigated the enzymatic pathway involved in the IBA to IAA conversion in the peroxisome. Figure 6A shows that ibr1, ibr10, and ibr1 ibr3 ibr10 mutants are fully insensitive to exogenous IBA in our growth conditions, thus making it impossible to genetically address the putative dependency of IBR1 and IBR10 on glutathione levels. However, we noticed that in ech2 and ibr3 mutants, LRs are still hyposensitive to IBA in the presence of BSO. Again, this suggests that ECH2 and IBR3 are not regulated in a glutathionedependent manner. In order to detect an eventual defect in gene expression in glutathione-deficient plants, we analysed the expression level of genes involved in IBA to IAA conversion (Fig. 6B). Surprisingly, ECH2, IBR1, IBR3, IBR10, AIM1, and PED1 genes were all moderately (10-50% increase) but consistently up-regulated in the cad2 mutant compared with the wild type. In any case, this does not allow us to identify any target that could be transcriptionally down-regulated upon glutathione depletion.
To conclude, we carefully examined most of the known components of IBA homeostasis and response pathways, but none of them seems to be the target of glutathione-dependent regulation.

IBA and glutathione control RH responses to phosphate deprivation
We analysed the RH elongation rate in response to phosphate starvation and to M. loti. We observed that both stimuli increase RH length in wild-type plants (Fig. 7A, B). Both cad2 and ibr1 ibr3 ibr10 mutants display a wild-type-like RH response to M. loti, suggesting that neither glutathione nor IBA participates in RH elongation response to PGPR (Fig. 7A).
In contrast, we observed that the cad2 mutant is clearly hyposensitive to phosphate deprivation, and that the RH elongation response to phosphate starvation is almost fully abolished in the ibr1 ibr3 ibr10 mutant (Fig. 7B). We also confirmed the decreased glutathione levels in the cad2 mutant while the ibr1 ibr3 ibr10 mutant does not have an altered glutathione content and redox status (Supplementary Fig. S5). Interestingly, phosphate deprivation does not significantly affect total glutathione content in plants but increases the GSH/GSSG ratio.
We examined if cad2 and ibr1 ibr3 ibr10 mutants affect the general response to phosphate deficiency or specifically the response of RHs to this abiotic stress. We therefore quantified the expression of marker genes known to be induced in response to phosphate deprivation. Figure 7C shows that the induction of the expression of marker genes in response to phosphate starvation is not abolished in cad2 and ibr1 ibr3 ibr10 mutants. We only noticed a slight down-regulation of SPX1 in the cad2 mutant. As SPX1 encodes a negative regulator of phosphate starvation responses, this could explain why other responsive genes are slightly up-regulated in cad2. We can conclude that both IBA conversion to IAA and glutathione are required to trigger the root hair response to phosphate starvation, without affecting general responses to this abiotic stress.

Glutathione regulation of IBA homeostasis or conversion to IAA in the root cap
In this work, we first report that RH and LR responses to exogenous IBA, but not IAA, are impaired by glutathione deficiency. More precisely, the glutathione-dependent control of IBA response affects early steps of LR development, either specification of founder cells or their activation to divide and form an LRP. We also show that low glutathione levels alter IBAdependent auxin signalling in the basal meristem, where RH elongation occurs and LR founder cells are specified. Finally, we show herein that transporters known to be involved in the apico-basal flux of IAA to the basal meristem, namely AUX1 and PIN2, are not the targets of the glutathione-dependent control. All these data strongly suggest that glutathione modulates IBA homeostasis or IBA to IAA conversion in the root cap, although it does not exclude that the same regulation can also occur in additional tissues.
All the phenotypes reported here concern root responses to exogenous IBA, and one can wonder if the glutathionedependent regulation normally occurs in physiological conditions, modulating pathways relying on endogenous IBA. Because of our scarce knowledge of IBA metabolism and the absence of genetic tools concerning IBA biosynthesis, it is rather difficult to address such a question. However, we accumulated several clues that, taken together, strongly support the importance of glutathione for the control of endogenous IBA pathways. First, we observe in cad2 mutant seedlings grown under standard conditions (i.e. without exogenous IBA) a general increase in the expression of all genes involved in IBA to IAA conversion (Fig. 6B). This might reveal a feedback mechanism that would report an excess of IBA or a depletion in IBA-derived IAA in low glutathione conditions. Secondly, we observe that both ugt and pdr double mutants are hypersensitive to exogenous IBA (Fig. 5B, C), which is consistent with the already reported overaccumulation of endogenous free IBA in corresponding single mutants (Strader and Bartel, 2009;Tognetti et al., 2010;Ruzicka et al., 2010). Interestingly, it appears that BSO has more effect in both ugt and pdr mutant genotypes than in wild-type plants, reducing IBA response by 20-25% in wild-type plants, but by 28-33% in pdr and ugt double mutants (Fig. 5B, C). This supports a role for glutathione in modulating responses to endogenous IBA. Finally, we show that both cad2 and the ibr1 ibr3 ibr10 triple mutant display hyposensitivity in their RH response to phosphate, but not to M. loti. In addition, neither is impaired in the general plant response to phosphate deprivation, but only in the RH elongation response. This set of arguments proves that both the endogenous IBA pathway and glutathione mediate RH response to phosphate starvation, and strongly suggests that the glutathione-dependent regulation of the endogenous IBA pathway is modulating the RH response.

Auxins fine-tune of root architecture in response to phosphate availability
Although well known to play important roles in plant development, the function of IBA as a source of IAA relative to other IAA sources, such as de novo synthesis or conjugated forms, is still very mysterious and we might wonder 'Why do plants need more than one type of auxin?' (Simon and Petrášek, 2011;Fig. 6. IBA conversion to IAA does not display sensitivity to glutathione. (A) Emerged lateral root density of 10-day-old wild-type (Col-0), ibr1, ibr3, ibr10, ibr1 ibr3 ibr10, or ech2 plants grown on standard 1/2 MS medium, or in the presence of different combinations of 0.5 mM BSO and 10 µM IBA (n≥15). (B) Expression levels (qRT-PCR) of IBR1, IBR3, IBR10, ECH2, AIM1, and PED1 genes in cad2 relative to wild-type 8-day-old plants grown on standard 1/2 MS medium. Data represent three independent replicates with 20 plants per replicate. Histograms represent the mean, and error bars represent the SD. Asterisks indicate a significant difference, based on a two-tailed Student t-test (*P<0.01; **P<0.001). Frick and Strader, 2018). A previous work has reported that IBA regulates both plant development and resistance to water stress, suggesting an important function for IBA in adapting plant development to abiotic stress (Tognetti et al., 2010). However, to our knowledge, our work is the first to clearly demonstrate that an IBA-dependent pathway is necessary for IAA to adapt some developmental pathway, namely RH elongation, to a specific abiotic stress (phosphate starvation).
In their recent work, Bhosale et al. (2018) already revealed the increase in IAA levels in the root cap in response to low external phosphate, necessary for RH elongation. However, because they observed a significant increase in TAA1 (TRYPTOPHAN AMINOTRANSFERASE OF ARABIDOPSIS 1) gene expression in such conditions, they hypothesized that the increase in IAA levels in the root cap is due to the induction of an indole-3-pyruvic acid (IPyA)-dependent pathway, the main one responsible for de novo IAA synthesis in Arabidopsis (Mashiguchi et al., 2011). Intriguingly, RH elongation response to external low phosphate is almost completely abolished in both ibr1 ibr3 ibr10 and taa1 mutants ( Fig. 7B; fig. 2B and C in Bhosale et al., 2018). One could imagine that two independent routes of IAA synthesis, namely the TAA1-dependent de novo synthesis and the IBA to IAA conversion, are both critical and induced in the root cap in response to phosphate deprivation. IBA can itself be derived from IAA and we can imagine that an increase in IBA to IAA conversion would require an increase in IBA levels, and therefore more IAA to supply the IBA stock. However, thinking that a root cap increase in IAA to increase IBA in order to increase IAA makes almost no sense. Thus, we suggest that external low phosphate induces TAA1 expression in the whole plant to increase global available auxin levels on the one hand, and on the other hand activates the local IBA to IAA conversion in the root cap to ensure a local appropriate developmental response.
However, we must remember that we are working in vitro with homogenous media containing a given concentration of (B) RH length in 8-day-old wild-type (Col-0), cad2, or ibr1 ibr3 ibr10 plants grown on 3/10 MS medium supplemented with 500 µM sodium phosphate (control) or 500 µM sodium chloride (minusP). (C) Expression levels (qRT-PCR) of PHT1;4, RNS1, and SPX1 low-phosphate-responsive genes, relative to the ACT2 control gene, in 8-day-old wild-type (Col-0), cad2, or ibr1 ibr3 ibr10 plants grown on 1/3 MS medium supplemented with 500 µM NaH 2 PO 4 (control) or 500 µM NaCl (minusP). Data represent three independent replicates with 20 plants per replicate and are normalized relative to the wild-type value in control conditions. Histograms represent the mean, and error bars represent the SDC. Asterisks indicate a significant difference, based on a twotailed Student t-test (*P<0.01; **P<0.001). phosphate. A very elegant recent work has reported that, when subjected to a heterogenous environment, using a Dual-flow-RootChip, a single root increases RH length in the medium having the highest phosphate concentration (Stanley et al., 2018). This observation contrasts with the induction of RH growth by low phosphate in homogenous in vitro media. RH adaptation to phosphate probably results from the crosstalk between systemic information based on overall available phosphorus in the plant, and local information based on phosphate availability in the environment, that we cannot differentiate in our in vitro homogenous conditions. Such dual control has been extensively documented for root responses to another important nutrient, nitrogen (Ruffel and Gojon, 2017). We might therefore envisage that both sources of IAA are differentially regulated by systemic and local signals. This might ensure a very acute dosage of free IAA to optimize root adaptation to the plant's metabolism and environment. In such a context, glutathione could be a central regulator of local IBA to IAA adaptation to external phosphate in the root cap.

Unravelling the glutathione-dependent regulation of the IBA pathway
Although we explored most of the known components of IBA pathways ( Supplementary Fig. S6), we were not able to identify how glutathione regulates IBA homeostasis or conversion to IAA. This can first be explained by the difficulties of working with IBA which is present in low amounts in plants compared with IAA, and is therefore difficult to detect and quantify (Frick and Strader, 2018). Moreover, many components of the IBA pathways remain to be identified. Undoubtedly, the development of new tools, such as IBA-specific reporters or probes, or mutants in IBA biosynthesis, would be very helpful to decipher the important roles played by this auxin in plants. The other problem we faced is the complex multifaceted roles of glutathione in cells, that prevented us from trying to decipher the mechanism from the glutathione starting point. The only clue we get is that IBA hyposensitivity is specific to plants with reduced amounts of glutathione but does not occur in plants affected in other redox pathways such as glutathione or thioredoxin reduction (Fig. 3). This suggests that IBA regulation by glutathione may act through glutathionylation or via the activity of thiol reductases that specifically depend on glutathione, such as glutaredoxins. One hypothesis would be that IBA, or a precursor, could itself be glutathionylated, which would be responsible for its storage or transport. In Arabidopsis, the Tau class of glutathione S-transferase (GSTU) has been shown to glutathionylate some fatty acids, ranging from short to long acyl chains (Dixon and Edwards, 2009). Similarly, GSTU19 and GSTF8 have been proposed to glutathionylate 12-oxophytodienoic acid (OPDA), hence allowing its translocation from the chloroplast to the peroxisome where it is converted to jasmonic acid (Davoine, 2006;Dixon and Edwards, 2009). Interestingly, among the 28 members of the Tau class, encoding cytosolic enzymes, several GSTU genes are transcriptionally regulated by IBA in the Arabidopsis root tip (Xuan et al., 2015) and/or by phosphate deficiency in Arabidopsis roots (Lin et al., 2011), and could be good candidates. It would be interesting to assay the IBA responses of RHs and LRs in corresponding mutants.
Because IBA to IAA conversion occurs in the peroxisome, we might also envisage that the glutathione-dependent regulation is not specific to IBA pathways but affects the peroxisomal machinery. One of the main peroxisomal functions is the β -oxidation of fatty acids, which is essential to supply energy during seed germination. We never observed any defect in seed germination in cad2 or pad2 mutants, in contrast to mutants in peroxisomal functions which generally strongly affect seed germination. However, we cannot exclude that the glutathione-dependent regulation only occurs in the root cap and therefore mainly affects IBA to IAA conversion although regulating the peroxisome machinery. Among the PEROXIN proteins, many participate in the peroxisomal matrix protein import machinery (Cross et al., 2016). PEX5 is a central cargo protein that recognizes proteins targeted with the specific PTS1 (Peroxisomal Targeting Signal 1) signal peptide and import them into the peroxisome. PEX7 recognizes proteins harbouring another signal peptide (PTS2), and then binds to PEX5 for import into the peroxisome. Interestingly, previous work on human and Pichia pastoris PEX5 revealed that its activity and oligomerization depend on a redox switch affecting a conserved N-terminal cysteine Apanasets et al., 2014). However, such regulation is not specific to glutathione but rather depends on the redox status and the content of ROS of peroxisomes.
Finally, the last hypothesis would be that glutathione regulates IBA homeostasis or conversion to IAA via an as yet unidentified or not assayed component. Quantification of IBA to IAA conversion would be helpful in order to confirm this hypothesis. Concerning the IBA to IAA β -oxidation pathway in the peroxisome, PED1 has been reported to harbour a redoxsensitive switch affecting a conserved cysteine and that participates in activating the enzyme when reduced (Pye et al., 2010). In contrast to most of the other enzymes involved in the conversion pathway, PED1 is targeted to the peroxisome through a PTS2 signal peptide. It is interesting to notice that pex5-1, in contrast to pex5-10, is not altered in PTS1-dependent import but only in PTS2-dependent import of proteins into the peroxisome (Woodward and Bartel, 2005;Khan and Zolman, 2010). pex5-1, although not displaying any phenotype during germination, is highly affected in IBA responses. This might reveal that IBA to IAA conversion is highly dependent on a PTS2-targeted protein, and PED1 could therefore be a good candidate to assay. However, PED1 is not specific to the IBA to IAA pathway since it acts in almost all β -oxidation pathways occurring in the peroxisome. We could imagine a local glutathione-dependent regulation of PED1 activity specifically in the root cap.
In conclusion, our work points out the important role of glutathione in both controlling IBA to IAA conversion required for root development and in mediating developmental responses to nutrient deprivation. This opens up novel perspectives in understanding how redox homeostasis can signal and integrate environmental constraints to trigger developmental adaptations via the regulation of morphogenetic hormonal pathways.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. Glutathione does not affect primary root response to IBA. Fig. S2. Glutathione regulates LR responses to IBA. Fig. S3. IBA does not alter redox status of GSH in root tips. Fig. S4. RH response to IBA in mutants affected in IBA homeostasis. Fig. S5. Glutathione content in response to low phosphate. Fig. S6. Auxin-related tools used in this study. Table S1. List of primers used for qPCR analyses.