pax1-1 partially suppresses gain-of-function mutations in Arabidopsis AXR3/IAA17

Background The plant hormone auxin exerts many of its effects on growth and development by controlling transcription of downstream genes. The Arabidopsis gene AXR3/IAA17 encodes a member of the Aux/IAA family of auxin responsive transcriptional repressors. Semi-dominant mutations in AXR3 result in an increased amplitude of auxin responses due to hyperstabilisation of the encoded protein. The aim of this study was to identify novel genes involved in auxin signal transduction by screening for second site mutations that modify the axr3-1 gain-of-function phenotype. Results We present the isolation of the partial suppressor of axr3-1 (pax1-1) mutant, which partially suppresses almost every aspect of the axr3-1 phenotype, and that of the weaker axr3-3 allele. axr3-1 protein turnover does not appear to be altered by pax1-1. However, expression of an AXR3::GUS reporter is reduced in a pax1-1 background, suggesting that PAX1 positively regulates AXR3 transcription. The pax1-1 mutation also affects the phenotypes conferred by stabilising mutations in other Aux/IAA proteins; however, the interactions are more complex than with axr3-1. Conclusion We propose that PAX1 influences auxin response via its effects on AXR3 expression and that it regulates other Aux/IAAs secondarily.


Background
The phytohormone auxin is central to the regulation of plant growth and development. Processes controlled by auxin include apical dominance, adventitious root formation, tropic responses, vascular patterning and root hair development [1][2][3][4][5][6]. These diverse morphological events are brought about by changes in cell division, expansion and differentiation [7][8][9], and many of them are mediated by the ability of auxin to control gene expression. Several families of genes containing Auxin Response Elements (AuxREs) in their promoters, are rapidly transcriptionally upregulated as a primary response to auxin [10], including members of the Aux/IAA gene family.
The Aux/IAA genes form a large multi-gene family found throughout the plant kingdom [11][12][13][14]. There are 29 members in Arabidopsis and their expression varies with respect to tissue specificity, auxin induction kinetics and sensitivity to auxin dose [11,15]. Aux/IAA proteins are characterised by four highly conserved domains [11]. Cterminal domains III and IV mediate homo-and heterodimerisation between Aux/IAA proteins [16]. Domains III and IV are also found in the Auxin Response Factor (ARF) family of transcription factors, a subset of which activate transcription from AuxRE-containing promoters [17,18]. Conservation of domains III and IV between the Aux/IAAs and ARFs allows combinatorial dimerisation amongst these families. The N-terminal domain I of Aux/IAA proteins functions as a transcriptional repression domain [19]. Thus, dimers between Aux/IAAs and activator ARFs block gene transcription [19][20][21][22].
Most Aux/IAA proteins are extremely unstable, with halflives ranging from 6 to 80 minutes [23][24][25][26]. They interact via domain II with SCF TIR1 (Skp1, Cdc53/cullin, F-box protein TIR1 ) and other auxin receptor F-box-containing E3 ubiquitin-protein ligase complexes, which target them for degradation by the ubiquitin-proteasome pathway [25,27]. Auxin promotes their association with such E3s, increasing their turnover [25,[27][28][29]. Thus transcriptional auxin responses are mediated by the auxin-induced destabilisation of Aux/IAA proteins, which relieves associated activator ARFs from repression and in turn, upregulates transcription from AuxREs [30]. Since Aux/IAA genes contain AuxREs in their promoters, this provides a feedback mechanism for regulating their own expression. Furthermore, it allows Aux/IAAs to control each other's expression creating a complex network of cross regulation amongst the family.
Although the study of the Aux/IAA and ARF families has elucidated many of the early events in auxin signal transduction, there is an ongoing requirement to find upstream regulators and novel targets of Aux/IAA and ARF regulated transcription. Here we present the isolation and characterisation of an extragenic recessive mutant, pax1-1, which partially suppresses the phenotypes of semi-dominant axr3 alleles. The pax1-1 single mutant shows pleiotropic phenotypes consistent with altered auxin responses. Double mutant analyses suggest that PAX1 also interacts genetically with other members of the Aux/IAA family. We propose that PAX1 acts upstream of Aux/IAA genes by regulating their transcription.

Mutant isolation
To search for new mutants altered in auxin signalling and more specifically, to find novel genetic interactors with AXR3, we screened x-ray-mutagenised axr3-1 (Columbia) plants for reversion of their strict apical dominance. The mutagenised seed were also homozygous for the gl1-1 mutation (which causes plants to lack leaf trichomes) so that revertant lines could be distinguished from contaminating wild-type plants. One line recovered from the screen showed increased numbers of axillary branches arising from the shoot, compared with axr3-1 controls (Figure 1c and 1d). The dwarf phenotype and dark green, curled leaves caused by the axr3-1 mutation were also partially suppressed in the revertant. In addition, the carpels were relatively longer than the sepals and petals, such that they protruded from the closed flower buds. Revertant plants were outcrossed to wild-type (Columbia) plants and segregation of individuals in the F 2 generation morphologically similar to axr3-1 homozygotes indicated that the suppressor mutation is extragenic to axr3. Thus, the new mutant was named partial suppressor of axr3-1 (pax1-1). Segregation analysis also suggested that the pax1-1 mutation is recessive to wild type and that it is linked to axr3-1 (data not shown).

Morphological phenotypes of pax1-1 seedlings
To gain insight into the wild-type role of PAX1, a detailed analysis of the pax1-1 mutant phenotype in a wild-type genetic background was carried out. Germination is slower in pax1-1 mutants compared with wild-type seeds (data not shown). pax1-1 hypocotyls are longer than wildtype when grown under a long day photoperiod, as a result of an increased growth rate (Figure 1e and 1f, Figure  2a). In contrast, dark grown hypocotyl elongation was not significantly different from wild-type ( Figure 2b). However, the rate of root growth is reduced in pax1-1 mutants (Figure 2c), and pax1-1 roots are abnormally straight, indicating that they have a reduced root wave response (Figure 1e and 1f).
pax1-1 plants produce root hairs at a higher density than wild-type and with altered morphology (Figure 1i and 1j). pax1-1 hairs are approximately 34% shorter than wildtype root hairs (wild type = 379 ± 1.16 μm, pax1-1 = 252 ± 0.96 μm), and develop in aberrant patterns. In wild-type Arabidopsis, the root epidermis is arranged in longitudinal files of cells [40]. Each file is composed entirely of either hair-forming cells (trichoblasts) or non-hair-forming cells (atrichoblasts), and each trichoblast forms a single root hair. However, in pax1-1 some trichoblasts produce multiple hairs, with up to 5 hairs observed per cell (Fig 1j). These hairs arise from a single initiation site. Thus, the increase in root hair density can be attributed at least partially to the development of multiple hairs from some trichoblasts. Also, some hairs form branches during the rapid tip growth phase of hair elongation, resulting in stalked, branched structures. Therefore PAX1 appears to pax1-1 mutant phenotypes in wild-type and axr3 backgrounds The elongated carpel phenotype seen in pax1-1 is also observed in axr3 pax1-1 double mutants. Bar = 0.5 mm. (i) Wild type, (j) pax1-1, (k) axr3-1 and (l) axr3-1 pax1-1 root hairs 3 dag. In pax1-1 and axr3-1 pax1-1, multiple roots hairs arise from a single root epidermal cell. Bar = 250 μm. All axr3-3 pax1-1 phenotypes were qualitatively similar to axr3-1 pax1-1, showing suppression of axr3 phenotypes by pax1-1.
affect both the orientation and the amount of root hair elongation.

Auxin response of pax1-1 mutants
To test the auxin response of pax1-1 roots, seedlings were transferred to medium containing various concentrations of the natural auxin, indole-3-acetic acid (IAA) and root elongation was measured 5 days later. On increasing concentrations of IAA between 10 nM and 10 μM, wild-type root growth was inhibited in a dose-dependent manner ( Figure 3a). pax1-1 root elongation follows a similar dose response to IAA as wild-type.

PAX1 and AXR3 interact genetically
To characterise in more detail the effect of pax1-1 on the axr3-1 phenotype, we re-crossed pax1-1 into an axr3-1 background. We also constructed double mutants between pax1-1 and a second, weaker semi-dominant allele, axr3-3, in order to assess the allele specificity of suppression. axr3-1 and axr3-3 show severely reduced root growth compared with wild-type ( Figure 4a) [39]. However, in double mutants with pax1-1, root elongation is increased by 32 percent and 23 percent, respectively. This effect is particularly striking considering that the pax1-1 mutation causes a reduction in root length when in a wild-type background. Thus, PAX1 appears to be required for the full inhibition of root growth by axr3 gain-of-function alleles. Other aspects of the root phenotype are also suppressed by pax1-1. In contrast to the axr3 single mutants, where root growth apparently wanders randomly with frequent changes in direction (e.g. kink in root in Figure 1k), the axr3-1 pax1-1 and axr3-3 pax1-1 double mutant roots grow more towards the gravity vector, and produce more root hairs (Figure 1k and 1l).
With respect to hypocotyl elongation, a similar interaction to that observed with root growth occurs. During the first few days of growth axr3-1 hypocotyls elongate at an increased rate compared with wild-type, followed by a period of slower elongation [41]. Five days after germination, axr3-1 and axr3-3 hypocotyls are slightly longer than wild-type, similar to pax1-1 ( Figure 4b). However, axr3-1 pax1-1 hypocotyls are no longer than either of the single mutants and axr3-3 pax1-1 hypocotyls are shorter than either single mutant.

Mechanism of suppression
Two possible mechanisms by which pax1-1 may suppress the axr3-1 phenotype are: (i) by increasing the degradation rate of the over-stable axr3-1 protein, or (ii) by decreasing the level of axr3-1 transcription. To test the first possibility, pax1-1 was crossed with transgenic plants expressing amino-terminal domains I and II of axr3-1 (axr3-1NT) fused to the β-glucuronidase (GUS) reporter, under the control of the soybean heat shock promoter (HS) [25]. This construct is an established reporter for axr3-1 protein stability. Plants were heat shocked at 37°C for 2 h and stained for GUS activity at various intervals following heat shock, to monitor the rate of GUS turnover.
In wild-type plants GUS activity remained constant between 30 min and 2 h following heat shock, showing that the axr3-1 protein is not observably degraded during this time period (Figure 6a). Similar results were observed in a pax1-1 background suggesting that pax1-1 has no effect on axr3-1 turnover (Figure 6b). In another experiment, no difference was observed between wild type and pax1-1 on the turnover of AXR3NT-GUS, in which the wild-type domain II destabilisation sequence is present (data not shown) [25].
To assess whether pax1-1 controls AXR3 transcription, pax1-1 was crossed to transgenic plants containing a transcriptional fusion between the AXR3 promoter and the GUS reporter gene (AXR3::GUS). Seedlings were stained for GUS activity two days after germination. In a wild-type background, GUS activity was observed in the root and hypocotyl (Figure 6c). However, AXR3::GUS expression was significantly reduced in pax1-1 (Figure 6d). GUS staining was almost completely absent from the hypocotyl and was greatly reduced in the root.

PAX1 interacts genetically with Aux/IAA genes
To determine whether PAX1 interacts with other Aux/IAA genes, we created double mutants between pax1-1 and axr2-1, slr-1 and shy2-2 [31,42,43]. These mutants carry semi-dominant mutations in domain II of IAA7, IAA14 and IAA3, respectively [31,33,37]. Like axr3-1, the mutations result in gain-of-function phenotypes due to hyper-stabilisation of the encoded proteins. Analysis of the double mutant phenotypes showed effects that could be classified into two different categories (Figure 4).
In the first class, the two mutations produced an additive effect. For example, the root length of the axr2-1 pax1-1 double mutant is less than that of both the axr2-1 and pax1-1 single mutants, which in turn are both less than wild-type (Figure 4a). This suggests that AXR2/IAA7 and PAX1 function independently in root elongation, although since the axr2-1 mutation is a dominant gain-offunction allele these results must be interpreted with caution. Likewise, slr-1 and pax1-1 had an additive effect on both root and hypocotyl elongation, implying that the two genes act independently during these processes (Figure 4a and 4b). The effects of shy2-2 and pax1-1 on root elongation were also additive.
In the second class of genetic interaction, one of the mutations was epistatic to the other. For example, under our growth conditions, five days after germination axr2-1 and axr2-1 pax1-1 hypocotyls were approximately the same length, with axr2-1 suppressing the long hypocotyl phenotype of pax1-1 (Figure 4b). shy2-2 was also epistatic to pax1-1 with respect to hypocotyl elongation.

PAX1 and gibberellic acid response
Several aerial features of the pax1-1 phenotype are suggestive of defects in gibberellic acid (GA) response. These include slow germination, short stature, dark green leaves (Figure 1a and 1b), protruding pistils (Figure 1g and 1h) and features indicative of slower phase change [44][45][46][47]. Plants pass through several distinct phases of development, marked by morphological changes in the lateral organs produced by the shoot apical meristem. In Arabidopsis, the first two rosette leaves are relatively round and each subsequent leaf becomes progressively more elongated [48]. pax1-1 produces rounder leaves than wild type, suggesting that it remains for longer in more juvenile phases of development. To test this, other markers of phase change were compared in pax1-1 and wild-type plants. Juvenile leaves form trichomes on their adaxial (dorsal) surfaces but not their abaxial (ventral) surfaces, and the transition from the juvenile to the adult phase of vegetative development is marked by the onset of trichome production on the abaxial surfaces of the leaves [46]. Under our growth conditions, wild-type plants produced 5.80 ± 0.09 juvenile leaves, lacking abaxial trichomes, whereas pax1-1 plants produced 8.60 ± 0.37 ( Table 1). As the rate of pax1-1 leaf initiation is reduced, this suggests that the transition to the adult growth phase is delayed both developmentally and temporally. Furthermore, floral transition also occurs both developmentally and temporally later than wild type, although this differ-ence is mostly due to the longer juvenile phase of pax1-1 (Table 1).
To test more directly for defects in GA response, pax1-1 plants were assayed for hypocotyl elongation on medium supplemented with GA 3 . Wild-type hypocotyl elongation was stimulated in a dose-dependent manner by exogenous GA 3 (Figure 3b). However in pax1-1, GA-induced growth is reduced compared with wild type.

Genetic mapping
From the F 2 of the outcross of the original pax1-1 axr3-1 double mutant to wild type (Columbia) PAX1 was estimated to map approximately 20 cM proximal to AXR3 on chromosome 1. To refine the map position of the PAX1 locus, pax1-1 (Columbia) plants were crossed to plants of the Landsberg erecta (Ler) ecotype and the F 2 generation was scored for segregation of polymorphic molecular markers between the two ecotypes. Segregation analysis revealed that PAX1 maps to a 420 kb region between markers SNP82 (3 recombinants/964 chromosomes) and cer474010 (3 recombinants/964 chromosomes). The region appears to be somewhat recombinationally suppressed, with an estimated 677 kb per cM. Attempts to identify the gene by transformation rescue were thwarted by the discovery that the mutant phenotype is unstable, reverting to wild type at a low frequency that was substantially enhanced by the transformation process (data not shown).
The delimited region includes 113 genes, of which only ARF19 is a clear PAX1 candidate, given its predicted role in the regulation of Aux/IAA gene expression. However, DNA sequencing of the ARF19 locus, including 1.5 kb upstream of the predicted translation start, from pax1-1, revealed no mutations. Furthermore arf19 insertion mutants were reported to have no phenotype [49,50], and trans-heterozygotes between an arf19 mutant and pax1-1 are phenotypically wild-type (data not shown) suggesting that PAX1 and ARF19 are not allelic. Thus, PAX1 is likely to be a previously unknown component of the Aux/IAA regulatory network.

Discussion PAX suppresses axr3 gain-of-function alleles
In this paper we present the isolation and characterisation of a new Arabidopsis mutant, pax1-1, which partially suppresses the phenotype of axr3 gain-of-function alleles. Virtually every aspect of the axr3 mutant phenotype analysed is suppressed at least partially by pax1-1. This is particularly compelling in cases where pax1-1 confers a more wild-type phenotype on axr3 mutants, despite having the opposite effect in a wild-type background. Examples of such are during root and hypocotyl elongation, and in the outgrowth of axillary inflorescences. These results suggest that PAX1 may encode a general positive regulator of AXR3 action. PAX1 does not appear to act at the level of AXR3 protein stability since axr3-1NT-GUS and AXR3NT-GUS translational fusions were turned over at similar rates in pax1-1 compared to wild type. In contrast, expression of an AXR3::GUS reporter was down-regulated in a pax1-1 mutant background suggesting that in wild-type plants, PAX1 regulates AXR3 transcription positively. Consistent with this idea, suppression of the axr3 phenotypes is not allele specific, as might be expected if the effect was at the protein level. This suggests a model for the suppression of axr3-1 and axr3-3 phenotypes by pax1-1. Such phenotypes are caused by AXR3 hyperstabilisation leading to the accumulation of increased protein levels. Therefore, reduced AXR3 transcription in pax1-1 would result in lower levels of AXR3 protein accumulation, and thus weaker axr3 phenotypes. Although this model is attractive, we have been unable to detect reliable differences from wild-type in the steady state levels the endogenous AXR3 mRNA in the pax1-1 mutant background (JJ and HMOL unpublished results). This might be because of the differences are tissue specific and therefore less easy to detect by RT-PCR than by histochemical GUS staining. However, it is also possible that the AXR3::GUS reporter does not accurately reflect the expression of the endogenous gene, as is often the case with promoter-GUS reporters. Whilst this would argue against a model of axr3-1 suppression by specific transcriptional down-regulation of AXR3, our results none the less suggest that the pax1-1 phenotype may be mediated by changes in transcription of auxin-regulated genes, since

PAX1 and other Aux/IAAs
Double mutant analysis demonstrates that PAX1 interacts genetically with other members of the Aux/IAA family in addition to AXR3. However, unlike its interaction with AXR3, where virtually all phenotypes are suppressed, a more complex set of interactions is observed with the other Aux/IAAs tested. pax1-1 shows combinations of epistatic and additive phenotypes with axr2-1, slr-1 or shy2-2. Furthermore, the type of interaction pax1-1 has with different Aux/IAA mutants, varies in different organs.
One explanation for the more complex set of interactions observed is that the effects of PAX1 on other members of the Aux/IAA family are indirect. Since Aux/IAA genes regulate each others' transcription, alterations in the expression level of one Aux/IAA gene (i.e. AXR3) could have downstream and feedback effects on the transcription of other family members. Therefore, the primary targets of PAX1 may include AXR3 transcription, whereas the effect on other Aux/IAAs may be secondary. Thus in a pax1-1 background, the phenotypes may result from widespread alterations in the balance of the Aux/IAA-ARF network, triggered by primary changes in just a few genes.

pax1-1 is defective in auxin-regulated development
A prediction of our model is that pax1-1 should have auxin-related phenotypes in a wild-type background. Indeed, many aspects of the pax1-1 phenotype are reminiscent of defects in auxin transport or signal transduction. For example, root, hypocotyl and stem elongation are regulated by auxin, and mutations in components of auxin signalling, such as Aux/IAA and ARF family members, lead to perturbations in these processes [39,51,52]. A mutant in ARF2 flowers late, suggesting that auxin may also control floral transition [53]. Another auxin-mediated process is root waving, with transport and signalling mutants displaying abnormal patterns of waving [37,54].  Furthermore, exogenous auxin affects root hair elongation and morphology, whilst many auxin response mutants show altered root hair growth [3,55,56].
Although the data discussed above suggest that PAX1 is involved in auxin response, pax1-1 roots show a wild-type growth response to exogenous IAA. This is consistent with the idea that the Aux/IAA-ARF network is differently configured but not globally down-regulated in the mutant, so that some phenotypes are suggestive of auxin resistance, but others are not. Analysis of the loss of function phenotypes of individual Aux/IAAs and ARFs demonstrates that the network is very robust with significant functional redundancy. If the PAX1 gene acts to modulate the network, it is therefore likely to affect more than one network member.

PAX1 and GA response
Another aspect of the pax1-1 phenotype is apparent defects in GA response. GA promotes germination and phase change, and increases stem and floral organ elongation [44][45][46][47]. pax1-1 plants show decreased germination, delayed phase change, reduced stem length, and altered floral organ elongation. Consistent with PAX1 functioning in GA responses, mutant hypocotyls treated with GA are resistant to its growth-promoting effects.
These data implicate PAX1 in both auxin and GA responses. Auxin is required for GA signalling, by regulating the GA-induced degradation of DELLA growth repressor proteins [57]. Furthermore, GA mediated destabilisation of the DELLA protein RGA is reduced in the auxin resistant mutant, axr1-12. AXR1 encodes a regulator of an E3 ubiquitin-protein ligase responsible for Aux/IAA turnover, and thus AXR1 may control DELLA protein levels through the destabilisation of Aux/IAAs [25,58]. In such a case, effects of PAX1 on Aux/IAA expression levels could be sufficient to alter DELLA protein turnover. In addition, GA metabolism may be affected since AXR3 and other Aux/IAAs are thought to regulate transcription of GA metabolism genes directly [59]. Alternatively, PAX1 might control GA signalling and/or metabolism independently of its effects on Aux/IAA protein levels.

Conclusion
Genetic analysis of the pax-
Seedlings were transplanted to Levington F2 compost (Fisons, UK) 8-10 days after germination, and grown to maturity under the conditions described above.

Mutant isolation
An M 1 population of 50 000 axr3-1 gl1-1 plants was generated by x-ray mutagenesis of the seed (8 kR dose). The M 2 was harvested as 30 seed pools, and 1500 plants from each pool were screened for suppression of the axr3-1 shoot phenotype. The pax1-1 single mutant was obtained from the F 2 of an outcross between axr3-1 pax1-1 plants and wild-type (Col). pax1-1 homozygotes were backcrossed twice to wild type and for each backcross, the F 2 generation was analysed for segregation of the mutant phenotypes. 55 and 94 F 2 plants were analysed from the first and second backcrosses, respectively, and all of the phenotypes cosegregated. pax1-1 homozygotes from the second backcross were used in all experiments except for the genetic crosses and in Figure 3a, where the original pax1-1 single mutant line was used.

Double mutant isolation
To verify the genotypes of double mutants, each line was backcrossed to both pax1-1 and wild type. In test crosses to wild type, the phenotype of all F 1 plants was qualitatively identical to the dominant mutant. F 1 plants from backcrosses to pax1-1 were all morphologically similar to the respective double mutant.
When double mutants between pax1-1 (Col) and shy2-2 (Ler) were constructed, double mutants lacking the erecta (er) mutation were selected for phenotypic analysis. As a control, shy2-2 homozygotes lacking the er mutation were also selected from the F 2 of the cross. To verify that these plants were homozygous for the wild-type ER allele, they were backcrossed to shy2-2 (Ler) and the F 1 generation was scored for the absence of the er phenotype.

Phenotypic analysis
In all experiments except those shown in Figure 2, only seedlings germinating within 3 days after sowing were analysed. Root and hypocotyl growth was measured using Lucia G (version 3.52a, 1991, Laboratory Imaging, Nikon UK Limited, Kingston, UK) image analysis software with a video camera input. Timing of floral transition was scored as the number of days after germination (dag) when floral buds first became visible, and the number of leaves at floral transition was scored at least 7 days later. Leaf senescence was recorded every 2 to 3 days so that the number of leaves counted reflected the total number of leaves produced. To characterise shoot branching phenotypes plants were analysed every 2 to 3 days for the time at which the first flower opened, and the numbers of secondary inflorescences were counted 14 days after this. Only axillary buds that were at least 3 mm long were counted as inflorescences.

Hormone response assays
For the auxin growth response assay, seedlings were germinated on ATS plates and then transferred after 3 days, to new plates supplemented with IAA. The position of each root tip was marked and after a further 5 days, the amount of new root growth was measured.
To test the hypocotyl growth response to GA, seedlings were germinated on ATS plates and then transferred after 4 days to new plates supplemented with GA 3 . The position of each hypocotyl apex was marked and the amount of new hypocotyl growth was measured after a further 4 days.
A genomic fragment containing 2.0 kb 5' to the AXR3 translation start was cloned into pBI101.3 (Clontech, UK) upstream of the GUS reporter gene, at BamHI and blunted EcoRI/HindIII sites, to create the plasmid pAXR3::GUS. This 2.0 kb promoter region was assumed to be sufficient to confer wild-type AXR3 expression since transgenic Arabidopsis plants containing the axr3-1 cDNA fused directly downstream of this promoter displayed phenotypes qualitatively similar to axr3-1 mutants. pAXR3::GUS was transformed into wild type (Col) by vacuum infiltration [61]. Multiple independent transgenic lines showed the same qualitative pattern of GUS expression. One of these lines was crossed into pax1-1.

Authors' contributions
MT generated the double mutants, carried out the morphological characterization of single/double mutants, hormone response assays and analysis of AXR3::GUS expression, contributed toward the mapping and design of the study, and drafted the manuscript. JJ performed the analysis of axr3 protein turnover, ARF19 allelism tests, and contributed toward the mapping and experimental design. PS carried out the mutant screen, isolated the pax1-1 single mutant and participated in the design of the study. DR generated the AXR3::GUS reporter line. OL participated in the design and coordination of the study and helped to draft the manuscript. All authors read and approved the final manuscript.