PILS proteins provide a homeostatic feedback on auxin signaling output

More than a century of intensive auxin research revealed the crucial role of auxin signaling for growth and developmental processes, such as embryogenesis, root and shoot development, reproductive development and plant responses to the environment (Weijers and Wagner, 2016; Weijers et al., 2018; Gallei et al., 2020; Pernisova and Vernoux, 2021; Anfang and Shani, 2021). Auxin activity is dependent on a variety of auxin transporters that modulate the direction of auxin flow, which is crucial for establishing auxin gradients and for keeping the intracellular auxin homeostasis. Among the most prominent transporters are the PIN-FORMED (PIN) proteins that transport auxin from cell to cell (known as intercellular auxin transport) or across the endoplasmic reticulum (ER) membrane (known as intracellular auxin transport) (Adamowski and Friml, 2015; Zwiewka et al., 2019). The role of the PIN proteins with a long intracellular loop (called canonical) is well established. They localize at the plasma membrane (PM) and define the polar auxin transport that establishes gradients and modulates a plethora of growth and developmental processes (Adamowski and Friml, 2015; Zwiewka et al., 2019). In contrast to the canonical PINs, the developmental roles of the PIN proteins with a short intracellular loop, such as PIN5 and PIN8 (called non-canonical), or with an intermediate loop, such as PIN6 (called semi-canonical), are less understood. These non-canonical auxin transporters localize to the ER (Mravec et al., 2009; Dal Bosco et al., 2012; Ding et al., 2012; Sawchuk et al., 2013; Ganguly et al., 2014; Simon et al., 2016; Ditengou et al., 2018). The PIN-LIKES (PILS) are predicted to be structurally similar to the PIN proteins and belong to another family of putative auxin transporters at the ER (Barbez et al., 2012). The PILS and non-canonical PIN proteins may perform a similar function in transporting auxin at the ER but, importantly, they evolved independently, at least in the plant lineage (Feraru et al., 2012; Bogaert et al., 2022). The PIN5- and PILS-based compartmentalization of auxin correlates with higher auxin conjugation rates (Mravec et al., 2009; Barbez et al., 2012; Dal Bosco et al., 2012; Ding et al., 2012). Auxin conjugates are thought to play important roles as storage forms for the active plant hormone indole-3-acetic acid (IAA) and, hence, intracellular auxin transport at the ER was suspected to play a role in cellular auxin homeostasis (Mravec et al., 2009; Barbez et al., 2012; Dal Bosco et al., 2012; Ding et al., 2012). On the other hand, PILS proteins define the nuclear abundance of auxin, presumably by limiting auxin diffusion from the cytosol into the nucleus, which affects the differential (asymmetric) growth during apical hook development as well as the overall shoot and root organ growth rates (Barbez et al., 2012; Beziat et al., 2017; Feraru et al., 2019). Moreover, PILS proteins integrate internal and environmental signals, such as brassinosteroid, light and temperature, into auxin-dependent growth programs (Beziat et al., 2017; Feraru et al., 2019; Sun et al., 2020), arguing against a purely homeostatic role.

In this article, we introduce a forward genetic screen that identified gloomy and shiny pils (gasp) mutants that define PILS6 protein abundance. We found that GASP1 encodes a non-characterized E3 ubiquitin ligase from the RING/U-box family. gasp1 mutants show altered auxin signaling output, which appears to be compensated by altered PILS6 abundance at the ER membrane. Our data propose that auxin exerts a posttranslational feedback regulation on the PILS proteins, thereby contributing to cellular auxin homeostasis.

Moderately high temperature induces PILS6 protein turnover, which consequently mediates auxin-dependent root thermomorphogenesis (Feraru et al., 2019; Fonseca de Lima et al., 2021), indicating that post-translational mechanisms define PILS activity and thereby plant adaptation. To further address these uncharted aspects of plant development, we performed a forward genetic screen, using a constitutive PILS6 expression line fused to GFP (p35::PILS6-GFP; hereafter named PILS6^OE), and screened the progeny of about 5000 M1 ethyl methanesulfonate (EMS)-mutagenized PILS6^OE seeds (Fig. 1A). We germinated the PILS6^OE seedlings under standard growth conditions (21°C) for 4 days and, subsequently, shifted the plates for 24 h to 29°C. Afterwards, we evaluated the temperature-sensitive PILS6-GFP fluorescence intensity, using an epifluorescence microscope. After re-screening, we identified 21 mutants that showed either reduced (8) or enhanced (13) PILS6-GFP fluorescence intensity under these conditions. We accordingly named these mutants gloomy and shiny pils (gasp) (Fig. 1A).

Among the eight mutants having reduced PILS6-GFP fluorescence signal intensity, we identified the gasp1-1;PILS6^OE mutant that showed very weak PILS6-GFP fluorescence signal after 24 h exposure to 29°C (Fig. S1). Notably, when grown under standard temperature of 21°C, gasp1-1 mutation caused a dramatic (85%) reduction of PILS6-GFP fluorescence intensity when compared with wild-type backgrounds (Fig. 1B,C; Fig. S1). This finding indicates that the gasp1-1 mutation affects PILS6-GFP protein abundance independently of moderately high temperature. In accordance with its negative effect on the fluorescence intensity of PILS6-GFP, gasp1-1 mutation alleviated the short root phenotype of PILS6^OE by 15% (Fig. 1D,E). Therefore, we identified gasp1-1 mutant as a suppressor of PILS6 under standard growth conditions.

To identify the causal GASP1 gene, we established a pool of gasp1-1;PILS6^OE individuals isolated from an F2 backcross (gasp1-1;PILS6^OE crossed to PILS6^OE). Accordingly, we re-sequenced the genome of this pooled mutant population as well as a pool of non-mutagenized PILS6^OE control seedlings using the Illumina and DNBseq^TM platform. By comparing the sequencing results of the two samples, we identified a single nucleotide polymorphism (SNP) in the uncharacterized protein-coding gene AT3G05545, which belongs to the RING/U-box superfamily protein (Kraft et al., 2005; Stone et al., 2005). The gasp1-1 mutation causes a C-to-T mutation, resulting in a proline (P)-to-leucine (L) amino acid substitution at the position 274 (Fig. 2A; Fig. S2A). P274L is not in a conserved region of GASP1, but a leucine substitution of a proline residue may have dramatic structural and functional consequences (Vilsen et al., 1989; Molnar et al., 2016).

To confirm that gasp1-1 mutation in AT3G05545 gene is indeed responsible for the suppression of PILS6^OE, we isolated a second mutant allele (SALK_091345; hereafter called gasp1-2) from the Salk collection of t-DNA insertion lines (Alonso et al., 2003) (Fig. S2A). GASP1 transcripts were not detectable in the gasp1-2 allele, indicating a full knockout of GASP1 (Fig. S2B). When crossed to PILS6^OE (gasp1-2;PILS6^OE), gasp1-2 reduced PILS6-GFP fluorescence intensity and, consequently, rescued total root length (Fig. 2B-E). Next, we crossed gasp1-1;PILS6^OE to the gasp1-2;PILS6^OE allele as well as to the PILS6^OE control line. In contrast to the control cross, the allelic test between gasp1-1 and gasp1-2 showed that the PILS6-GFP intensity and PILS6^OE phenotypes remained suppressed in the F1 generation (Fig. 2F-H; Fig. S2C). Altogether, we concluded that defects in the GASP1 are responsible for the phenotypes observed in the gasp1-1;PILS6^OE.

To assess the specificity of GASP1, we crossed gasp1-2 to the PILS5 overexpression line (p35S::PILS5-GFP; PILS5^OE). Similar to PILS6^OE, PILS5^OE-induced reduction in main root growth was also suppressed in gasp1-2;PILS5^OE (Fig. S2D,E). PILS5^OE also represses dark-grown hypocotyl growth (Barbez et al., 2012; Beziat et al., 2017), which was as well alleviated by the gasp1-2 mutation (Fig. 2I,J). In agreement, the PILS5-GFP signal intensity was strongly reduced in gasp1-2;PILS5^OE dark-grown hypocotyls (Fig. 2K,L), showing that GASP1 affects at least two PILS proteins, in distinct tissues and growth conditions.

To directly address whether GASP1 indeed affects PILS5 and PILS6 protein abundance, we subsequently used quantitative western blots. In accordance with the reduced PILS5/6-GFP fluorescence intensity, gasp1 mutants displayed reduced PILS5 and PILS6 protein levels in the dark-grown hypocotyls and light-grown seedlings, respectively (Fig. 3A; Fig. S3A,B). We conclude that GASP1 defines the abundance of PILS proteins, such as PILS5 and PILS6.

GASP1 belongs to the RING/U-box superfamily and plays a role as an E3 ubiquitin ligase (Kraft et al., 2005; Stone et al., 2005). RING E3 ubiquitin ligases typically mediate the ubiquitylation of target proteins, where K48-linked ubiquitylation recruits these targets for degradation via the 26S proteasome (Joazeiro and Weissman, 2000; Smalle and Vierstra, 2004; Vierstra, 2009). To investigate whether GASP1 could directly modulate PILS protein abundance at the ER membrane, we generated a transgenic line overexpressing GFP-GASP1 fusion. 35S::GFP-GASP1 lines displayed weak but ubiquitous signal in the root (Fig. S3C). In agreement with its predicted localization (https://suba.live/suba-app/factsheet.html?id=AT3G05545; Hooper et al., 2017), GFP-GASP1 was detectable in the nucleus, but also showed cytosolic localization (Fig. 3B). Although GFP-GASP1 appeared to be enriched in the perinuclear regions of light-grown seedlings, we did not detect pronounced association with the ER and, accordingly, GFP-GASP1 did not show co-localization with PILS3-RFP (Fig. 3B,C). In addition, GASP1 did not interact with PILS3 or PILS5 proteins in a yeast mating-based split-ubiquitin system (Fig. S3D). Even though we cannot fully rule out a direct interaction in planta, we assume that the putative E3 ubiquitin ligase GASP1 indirectly affects the protein abundance of PILS5 and PILS6. Hence, we propose that GASP1 may define the ubiquitylation and subsequent proteasomal degradation of cytosolic and/or nuclear proteins that are upstream regulators of PILS5 and PILS6 proteins. To test whether the degradation of the PILS proteins is affected by the disturbance of the ubiquitin-26S proteasome pathway, we subsequently used MG132 and Bortezomib (BTZ) to pharmacologically interfere with the proteasome function in Arabidopsis (Marshall et al., 2015; Yu et al., 2015; Wang et al., 2016). Seedlings of PILS6^OE were treated for 3 h with MG132 or BTZ, which caused a significant increase of PILS6-GFP fluorescence in roots when compared with the DMSO-treated control seedlings (Fig. 3D,E), indicating that PILS6 protein abundance is regulated in a 26S proteasome-sensitive manner. Importantly, another ER membrane-localized protein, the RING E3 ligase DER1 (AT4G29330; Kirst et al., 2005), a component of the ERAD pathway, translationally fused to mScarlet (35S::DER1-mScarlet) is not affected by proteasome inhibition, indicating some specificity of this effect (Fig. 3F,G). The pharmacological inhibition of the 26S proteasome also increased the PILS6-GFP fluorescence in gasp1-1 and gasp1-2 mutants (Fig. 3E; Fig. S3E-G), indicating that other molecular components contribute to the effect of MG132 and BTZ on PILS6 abundance. Collectively, our data shows that, although the 26S proteasome activity is required for PILS6 degradation, PILS protein abundance is not directly controlled by the E3 ubiquitin ligase GASP1.

We next addressed whether the reduced PILS5 and PILS6 protein abundance correlates with the expected increased nuclear auxin signaling output in gasp1 (Barbez et al., 2012; Beziat et al., 2017; Feraru et al., 2019; Sun et al., 2020). To visualize the auxin signaling output in gasp1-2, we crossed gasp1-2 with the auxin response marker DR5::GFP. Although DR5::GFP signal intensity was not distinguishable in the root tip of gasp1-2 mutant and wild type (Fig. S4A,B) we, unexpectedly, observed reduced DR5::GFP signal in the upper vascular tissues of light-grown roots as well as dark-grown hypocotyls of gasp1-2 mutant (Fig. 4A-D). In agreement, auxin responsive genes, such as IAA1, IAA5, IAA7, SAUR19 and SAUR63, showed reduced expression in the light-grown seedlings and in the dark-grown hypocotyls of gasp1 mutants (Fig. 4E,F). This finding suggests that GASP1 is required to maintain auxin signaling output in roots and shoots, which is independent from its effect on PILS5 and PILS6 abundance (Fig. 4E,F). The overexpression of PILS proteins also limits nuclear auxin signaling, but the repression of auxin signaling output was not additive in gasp1-1;PILS6^OE, gasp1-2;PILS6^OE and gasp1-2;PILS5^OE (Fig. 4E,F), suggesting that the effect on PILS abundance balances the auxin response. This finding hints at a molecular mechanism in which the PILS abundance could relate to a homeostatic feedback mechanism on auxin signaling output.

This prompted us to address whether the diminished auxin signaling output observed in gasp1 mutants could reflect an auxin impact on PILS proteins abundance. To test this, we used L-Kynurenin (KYN) to pharmacologically interfere with auxin biosynthesis (He et al., 2011). KYN applications indeed phenocopied gasp1 mutants and decreased the PILS6-GFP fluorescence as well as protein abundance (Fig. 4G-I; Fig. S4C). Conversely, 100 nM IAA treatment increased both PILS6-GFP fluorescence and protein abundance (Fig. 4G-I; Fig. S4C). These experiments suggest that high and low auxin signaling outputs increase and decrease PILS6 abundance, respectively. In contrast, the ER membrane marker DER1-mScarlet did not show any response to the treatments with either KYN or IAA (Fig. S4D,E), suggesting certain specificity of this auxin response. This data proposes that auxin exerts a homeostatic feedback on its own signaling rate by controlling the abundance of PILS intracellular auxin transporters.

Our forward genetic screen performed to identify regulators of the PILS6 response to moderately high temperature yielded 21 gasp mutants that either decreased (gloomy) or increased (shiny) PILS6 protein abundance. In this study, we investigated gasp1-1 that repressed PILS6-GFP under standard temperature and found that GASP1 may function as a modulator of auxin signaling rates. GASP1 encodes an uncharacterized E3 ubiquitin ligase, from the H-type, that belongs to the RING/U-box superfamily protein, which supposedly mediates substrate specific ubiquitylation (Kraft et al., 2005; Stone et al., 2005). The gasp1 mutants display severe reduction in auxin signaling output but did not obviously increase phenotypic trait variations, being largely indistinguishable from wild-type seedlings. It is therefore conceivable that homeostatic auxin responses may balance the molecular responses in gasp1.

The biological role of GASP1 remains largely unknown, but we show here that the severely reduced auxin signaling output in gasp1 mutants is in part compensated by enhanced turnover of at least PILS5 and PILS6 proteins. Our data show that GASP1 does not directly interact with PILS proteins, such as PILS3 or PILS5, heterologously expressed in yeast. Considering that E3 ligases are typically negative regulators of their clients, the reduced PILS5 and PILS6 abundance in gasp1 mutants also questions a direct impact of the GASP1 on the PILS proteins. We propose that the GASP1 impact on auxin signaling output indirectly affects PILS5 and PILS6 turnover. We, accordingly, show that sub- and supra-optimal levels of auxin decrease and increase PILS6 abundance at the ER, respectively. It remains to be seen how precisely auxin levels determine the turnover of PILS proteins. Such a response could involve the canonical TIR1/AFB auxin receptors and downstream signaling events, altering the yet-to-be-defined PILS degradation mechanisms. Alternatively, auxin availability may structurally affect PILS proteins, which could alter their interaction with the degradation machinery.

PILS proteins define the nuclear abundance and signaling of auxin, which appears to be highly responsive to internal and environmental signal perturbations (Beziat et al., 2017; Feraru et al., 2019; Sun et al., 2020). Here, we propose a working model where an auxin impact on PILS abundance provides homeostatic feedback (Fig. S4F), enabling auxin signaling output to maintain its own cellular homeostasis. Altogether, we envision that a PILS-dependent feedback mechanism provides robustness to plant growth and development.

Arabidopis thaliana ecotype Col-0 (wild-type), p35S::PILS5-GFP (PILS5^OE; AT2G17500; Barbez et al., 2012), p35S::PILS6-GFP (PILS6^OE; AT5G01990; Barbez et al., 2012) and pDR5rev::GFP (Benkova et al., 2003) have been previously described. p35S::GFP-GASP1 (AT3G05545), p35S::PILS3-RFP (AT1G76520) and 35S::DER1-mScarlet (AT4G29330) were generated in this study. gasp1-2 (SALK_091345) was obtained from NASC (Alonso et al., 2003); gasp1-1 was identified in this study.

Seeds were sterilized in 70% ethanol (1-2 min sterilization in paper bags, followed by 30 min drying) and plated, usually on one single line, uniformly spaced, in the upper part of Petri dishes containing 50 ml solidified Murashige and Skoog (MS) agar medium (0.8% agar, 0.5×MS and 1% sucrose, pH 5.9), then stratified for 3 days in the dark at 4°C and grown on vertically oriented plates in a plant cabinet equipped with above-placed cool-white fluorescent bulbs set at about 140 μmol/m⁻²s⁻¹, long-day photoperiod (16 h light/8 h dark) at 21°C. This way of plating ensures that all seedlings in the plate are exposed to the same light intensity and humidity, which results in low variability. For moderately high temperature (HT)-related experiments, we used the growth conditions described in Feraru et al. (2019). Seedlings were grown on plates (in pairs), for 4 days at 21°C (standard temperature) and subsequently shifted for 24 h to a cabinet displaying similar settings, excepting the temperature was 29°C (moderately high temperature). The control plates remained in the cabinet equipped with standard conditions.

Approximately 10,000 seeds of 35S::PILS6-GFP (PILS6^OE) were soaked (gently shaking) for 10 h in 0.1 M phosphate buffer, pH 7, containing 0.3% (v/v) EMS. Before mutagenesis, the seeds were soaked for 5 min in water containing 0.05% Triton X-100, then rinsed three times with water. After mutagenesis, the seeds were rinsed seven times with water, then dispersed as desired in 0.1% agarose and transferred to soil by pipetting. From ∼5400 mutagenized plants (M1), we harvested 360 pools (each pool containing about 15 M1), and screened more than 80,000 M2 seedlings under an Olympus stereomicroscope for individual seedlings with weaker or stronger fluorescence than PILS6^OE control. The seedlings having different fluorescence intensity than the control were picked up, propagated, and confirmed in the next generation as gasp mutants.

For sequencing of gasp1-1, we crossed gasp1-1;PILS6^OE to PILS6^OE and selected in F2 the individuals showing the gasp1-1;PILS6^OE phenotype. A pool of seedlings weighing 100 mg was used to extract genomic DNA using the DNeasy Plant Mini Kit (Qiagen), according to the manufacturer's instructions. A sample of more than 1.5 µg (>13 ng/µl sample concentration) was sent for sequencing at BGI Tech Solutions, along with a similar sample containing the PILS6^OE control. The samples were sequenced using the DNBseq platform and the standard bioinformatics analysis was performed by BGI Tech Solutions, which identified the different SNPs between each sample and Arabidopsis thaliana genome from the TAIR database, followed by SNP calling, annotation and statistics. To identify the gasp1-1 mutation, we compared the list of SNPs identified in the gasp1-1;PILS6^OE with the list of SNPs identified in the PILS6^OE sample. After elimination of common SNPs between the two samples and of those heterozygous and synonymous SNPs, we identified one single typical EMS mutation.

For root and hypocotyl length measurements, seedlings were grown on vertically oriented plates in the light (root) or dark (hypocotyl). Plates were scanned with an Epson Perfection V700 scanner and the length was measured using ImageJ 1.41 software (http://rsb.info.nih.gov/ij/).

A Leica TCS SP5 confocal microscope was used for fluorescence imaging. Unless otherwise stated, 5-day-old seedlings were used. When treated before imaging, the seedlings were either submerged in MS liquid medium (MG132, BTZ) or transferred on plates (IAA, KYN) containing the desired concentration of the drug or similar amount of solvent and kept in the plant cabinet for the duration specified in the text or figure legend. The mean gray value of the fluorescence intensity was quantified in a defined rectangle region of interest (ROI), marked on the images, using the ‘Quantify’ tool of Leica software (LAS AF Lite).

To generate p35S::GFP-GASP1, the GASP genomic fragment was cloned into the pDONR221 using the primers B1_GASP_FP and B2_GASP^STOP_RP, listed in Table S1. The resulting entry clone was subsequently transferred to the gateway-compatible destination vector pK7WGF2 (Karimi et al., 2002). Transformed lines were selected on 100 mg/l kanamycin in F1 and 25 mg/l kanamycin in F2. For the split ubiquitin assay, we amplified the PILS3, PILS5 and GASP1 coding sequence without stop codon using PILS3_FP and PILS3^NOSTOP_RP, PILS5_FP and PILS5^NOSTOP_RP, B1_GASP_FP and B2_GASP^NOSTOP_RP listed in Table S1. The fragments were firstly cloned into the pDONR221. Subsequently, we recombined the baits (PILS3 and PILS5) and prey (GASP1) into pMetYC-DEST (Grefen et al., 2007) and PNX35-DEST (Grefen and Blatt, 2012), respectively. We used Gibson Assembly (New England Biolabs) to generate 35S::DER1-mScarlet. The coding sequence of DER1 (DER1_FP and DER1_RP), the 35S promoter (35S_FP and 35S_RP) and mScarlet-i tag (mScarlet_FP and mScarlet_RP) were amplified by PCR using Q5 High-Fidelity DNA Polymerase (New England Biolabs). The fragments were then cloned into a linearized (EcoRV-HF-NEB) pPLV03 vector using Gibson Assembly. The transformed lines were selected on 15 mg/l phosphinothricin (Basta). The 35S::PILS3-RFP plasmid generated previously (Barbez et al., 2012) was transformed into Col-0 plants and the transformed lines were selected on 20 mg/l hygromycin.

For the split ubiquitin assay, the yeast strains THY.AP4 and THY.AP5 were transformed with bait and prey constructs, respectively, using a modified protocol from Gietz and Woods (2002). Approximately 100 µl of fresh yeast were scraped from YPD plates, resuspended in 200 µl sterile H₂O, centrifuged for 5 min at 2000 g and the supernatant removed. The yeast was afterwards resuspended in 200 µl yeast transformation buffer (40% PEG 3350, 200 mM LiAc, 100 mM DTT), added 10 µl single stranded carrier DNA and 1 µg of plasmid DNA, and mixed by pipetting up and down. We incubated the yeast for 15 min at 30°C and for 45 min at 45°C, subsequently plated on Synthetic Defined (SD) medium, and incubated for 4 days at 28°C. A pool of transformed colonies was mated as described in Grefen et al. (2007). The selected diploid colonies were afterwards incubated on plates contacting selective medium (SD-Trp, -Leu, -Ade, -His, -Ura) at 21°C, under light (16 h light/8 h dark) and dark (plates were wrapped in aluminum foil to mimic dark) conditions. Growth was recorded up to 9 days after plating.

To identify gasp1-1, we amplified the genomic sequence with B1_GASP1_FP and B2_GASP1^STOP_RP and sequenced the sequence around the mutation using the primer GASP1_FP6 listed in Table S1. To genotype gasp1-2, we used a combination of gasp1-2 and t-DNA insertion-specific primers listed in Table S1.

RT-qPCR analysis was performed as described in Feraru et al. (2019). We used the InnuPREP Plant RNA Kit (Analytic Jena) to extract total RNA according to the manufacturer's recommendation. The RNA samples were treated with InnuPREP DNase I (Analytic Jena). To synthesize cDNA, 1 μg of RNA and the iSCRIPT cDNA Synthesis Kit (Bio-Rad) were used. RT-qPCR was carried out in a C1000 Touch Thermal Cycler equipped with the CFX96 Touch Real-Time PCR Detection System (Bio-Rad) and using the Takyon qPCR Kit for SYBER assay (Eurogentec), according to the manufacturer's recommendation. Gene expression was normalized to the expression of reference gene ACTIN 2, which showed similar expression among genotypes. We used RealTimeDesign qPCR Assay Design Software from Biosearch Technologies to design the qPCR primers or we used primers already published (Inoue et al., 2016; Schlereth et al., 2010). The primers used for qPCRs are listed in Table S1.

We used 5-day-old dark-grown hypocotyls for the experiment related to PILS5^OE and 6-day-old total seedlings for the experiments related to PILS6^OE. For IAA and KYN treatments, the seedlings were grown for 5 days on nylon mesh on MS plates, then transferred with the underlying mesh to the plates supplemented with IAA, KYN or similar amount of DMSO solvent, and harvested after 24 h. Protein extraction was performed as described in Moulinier-Anzola et al. (2020). Each sample contained 20 mg seedlings. The frozen plant material was ground and then extracted in 150 μl buffer [65 mM Tris (pH 6.8), 8 M urea, 10% glycerin, 2% SDS, 5% β-mercaptoethanol and 0.25% bromophenol blue]. The samples were then heated at 65°C for 5 min and spun down before loading. Anti-GFP (Roche, 11814460001, 1:1000), monoclonal anti-α-Tubulin (Sigma-Aldrich, T6074, 1:100,000) and goat anti-mouse IgG (Jackson ImmunoResearch, 115-035-164, 1:40,000) antibodies were used for detection of PILS5^OE, PILS6^OE and Tubulin.

We used Excel to organize data and GraphPad Prism to generate graphs and perform statistical analysis. The experiments were performed at least three times or in three replicates, but only the data from one representative experiment is shown.

We are grateful to E. Benkova and J. Friml for sharing pDR5rev::GFP, BOKU-VIBT Imaging Center and M. Debreczeny for access and advice, and L. Mach and R. Strasser for sharing equipment.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200929