A novel Ca2+-binding protein that can rapidly transduce auxin responses during root growth

Signaling cross talks between auxin, a regulator of plant development, and Ca2+, a universal second messenger, have been proposed to modulate developmental plasticity in plants. However, the underlying molecular mechanisms are largely unknown. Here, we report that in Arabidopsis roots, auxin elicits specific Ca2+ signaling patterns that spatially coincide with the expression pattern of auxin-regulated genes. We have identified the single EF-hand Ca2+-binding protein Ca2+-dependent modulator of ICR1 (CMI1) as an interactor of the Rho of plants (ROP) effector interactor of constitutively active ROP (ICR1). CMI1 expression is directly up-regulated by auxin, whereas the loss of function of CMI1 associates with the repression of auxin-induced Ca2+ increases in the lateral root cap and vasculature, indicating that CMI1 represses early auxin responses. In agreement, cmi1 mutants display an increased auxin response including shorter primary roots, longer root hairs, longer hypocotyls, and altered lateral root formation. Binding to ICR1 affects subcellular localization of CMI1 and its function. The interaction between CMI1 and ICR1 is Ca2+-dependent and involves a conserved hydrophobic pocket in CMI1 and calmodulin binding-like domain in ICR1. Remarkably, CMI1 is monomeric in solution and in vitro changes its secondary structure at cellular resting Ca2+ concentrations ranging between 10−9 and 10−8 M. Hence, CMI1 is a Ca2+-dependent transducer of auxin-regulated gene expression, which can function in a cell-specific fashion at steady-state as well as at elevated cellular Ca2+ levels to regulate auxin responses.


Introduction
The plant hormone auxin regulates diverse developmental and physiological processes, serving as a morphogen that creates local maxima and gradients [1]. Within plant roots, auxin accumulates at the growing tip, with a maximum at the quiescent center (QC) and the below organized columella cells. It is well documented that this accumulation happens early during embryo development and is essential for root meristem maintenance and root growth [1][2][3][4][5]. Interestingly, this long-term auxin accumulation is translated to gradients of transcription factors PLETHORA (PLT) genes and specific expression pattern of auxin response factors (ARFs) and their Aux/indole-3-acetic acid (IAA) inhibitors that have been shown to maintain the activity of root stem cells [5][6][7][8]. In addition, nontranscriptional responses to auxin have been suspected, but it remains mysterious how they are regulated by auxin gradients at the root tip.
In plants, diverse environmental stimuli induce elevations in cytosolic Ca 2+ from steadystate levels of around 10 −7 M to 10 −6 to 10 −5 M, which are in turn transduced by plethora of Ca 2+ -binding proteins. The majority of these proteins-including calmodulins (CaMs), CaMlike (CML), Ca 2+ -dependent kinases (CDPKs), and calcineurin B-like (CBL)-contain two or more Ca 2+ -binding EF-hand motifs. CaMs, CMLs, CDPKs, and CBLs bind Ca 2+ only when its concentrations are at the μM range and hence function as Ca 2+ sensors that transduce response only when Ca 2+ levels increase from steady-state levels [32,33]. EF-hand proteins that can bind Ca 2+ at concentrations lower than 10 −7 M have not been extensively studied. Although such proteins do not likely function as Ca 2+ sensors, they can function as response transducers directly dependent on their expression levels. Ca 2+ has been named "the missing link in auxin action" [34]. AUX1-dependent auxin influx in root and root hairs induces cyclic nucleotide gated channel 14 (CNGC14)-and TIR1/AFBdependent Ca 2+ signaling within seconds that in turn affects downstream auxin signaling [24][25][26]. CNGC14 function is required in response to gravity stimulus [25], indicating that function involves Ca 2+ signaling. Given the rapid Ca 2+ -associated response that is induced by auxin, the existence of auxin up-regulated Ca 2+ -binding proteins that bind Ca 2+ at its steady-state levels and can therefore transduce Ca 2+ responses upon their expression presents an intriguing mechanism of action In this work, we describe the identification of an auxin-regulated Ca 2+ -binding protein that functions in the root meristem and crucially regulates auxin responses and affects auxininduced changes in cytoplasmic Ca 2+ levels. The expression of this protein is rapidly up-regulated by auxin, and it responds to Ca 2+ at concentrations lower than 10 −7 M and can therefore rapidly transduce auxin signals.

Auxin induces a specific Ca 2+ signal pattern in the root and auxinregulated expression of the Ca 2+ -binding protein CMI1
To study potential changes in cytoplasmic Ca 2+ concentration in the root following auxin treatment, we used Arabidopsis seedlings expressing the fluorescence resonance energy transfer (FRET)-based Ca 2+ indicator Yellow Cameleon 3.6 (YC3.6) [35]. In the control conditions, elevated Ca 2+ concentrations were primarily observed in the root tip, resembling the auxin accumulation pattern (Fig 1A, mock and 1B and S1 Fig, mock). A typical auxin-induced Ca 2+ signal was observed after 100 seconds of auxin (10 μM 1-Naphthaleneacetic acid [NAA]) application and increased after 160 seconds ( Fig 1A, "NAA 100 sec" and "160 sec"). The most pronounced Ca 2+ elevations were observed in the root cap, lateral root (LR) cap (LRC), and vasculature ( Fig 1A, NAA and S1 Fig, NAA). The pattern of the generated Ca 2+ signal was corresponding to auxin response and distribution ( [11,36,37] and Fig 1B). The similarity between auxin-induced Ca 2+ concentration increases ( Fig 1A) and the oscillatory expression pattern of TIR1/AFB auxin receptors-regulated genes [36] is suggestive for mutual interdependency between auxin and Ca 2+ in the root.
Previously, we identified a family of coiled-coil domain Rho of plants (ROP) effectors that we named interactors of constitutively active ROP (ICRs) [38]. Interactor of constitutively active ROP (ICR1) is degraded in an auxin-dependent fashion in the root meristem and affects root growth [38][39][40]. In a screen for ICR1-interacting proteins, we identified a single EF-hand Ca 2+ -binding protein that we designated as Ca 2+ -dependent modulator of ICR1 (CMI1) (At4g27280). CMI1 is expressed at the site of auxin accumulation, and its expression is strongly induced by auxin (Fig 1C-1E and S2 Fig). In Arabidopsis CMI1 is a member of a small protein family, consisting of three members, and was formerly called KIC-related protein 1 (KRP1) [41]. Because the name KRP has originally been used for the cell cycle regulators KIP-related proteins [42], which are unrelated to KRP1, we decided to adhere to the CMI1 nomenclature in this work.

Expression of CMI1 is regulated by auxin through TIR1/AFB auxin receptors
To gain first indications for the function of CMI1 in plants, we examined the expression pattern and regulation of CMI1 expression and their correlation with cytoplasmic Ca 2+ levels. High CMI1-β-glucuronidase (GUS) and monomeric red fluorescent protein (mRFP)-CMI1 levels were detected in the root meristem of pCMI1::CMI1-GUS and pCMI1>>mRFP-CMI1 plants (Fig 1C), resembling the expression pattern of DR5 promoter-driven auxin reporters [4]. Microarray experiments revealed that induction of CMI1 expression by auxin was reduced in the auxin-resistant 1 (axr1) auxin signaling mutant [43], indicating that auxin induces CMI1 expression by a TIR1/AFB-dependent mechanism ( Fig 1D). Furthermore, our analysis of additional publicly available microarray data revealed that CMI1 was induced by exogenous Confocal images of root expressing the YC3.6-free Ca 2+ sensor prior to auxin treatment (mock) and after treatment with 10 μM NAA ("NAA"). The same is imaged after 100 and 160 sec of NAA treatment. (B) Images of a root with schematic representations of auxin and Ca 2+ signal distribution at the root tip and the SCN. (C) Expression of pCMI1::CMI1-GUS in and transcription-transactivation driven pCMI1>>mRFP-CMI1 primary root meristem. (D) Microarray expression data showing the induction of CMI1 by auxin is reduced in axr1 auxin response mutant background. (E) Expression level and pattern of pCMI1-driven CMI1-GUS in cmi1 mutant background without (mock) and 2 hours following treatment with 10 μM NAA ("NAA"). Scale bars, (A) 20 μm and (C and E) 50 μm. Underlying data for this figure can be found in S1 Data. axr1, auxin-resistant 1; CMI1, Ca 2 + -dependent modulator of ICR1; GUS, β-glucuronidase; IAA, indole-3-acetic acid; NAA, 1-Naphthaleneacetic acid; SCN, stem cell niche; YC3. 6 auxin treatments and suppressed in the axr2-1/iaa7 auxin-insensitive mutant [44]. A quantitative PCR (qPCR) analysis confirmed induction of CMI1 mRNA following auxin treatments (S2 Fig). In agreement, regions of increased expression of CMI1-GUS in the root elongation and maturation differentiation zones of pCMI1::CMI1-GUS/cmi1 plants were observed following treatments with 10 μM IAA (Fig 1E), resembling the pattern of TIR/AFB auxin-induced genes [36]. Taken together, these results indicate that the expression of CMI1 is enhanced in cells and tissues with increased auxin concentration and also regulated by auxin via the TIR1/ AFB auxin receptor system.

CMI1 mediates auxin responses and fine-tunes root growth
Next, we further examined the function of CMI1 and its interconnection with auxin signaling in plants by analyzing the phenotype of a CMI1 loss-of-function mutant. The cmi1 mutant (Cold Spring Harbor Laboratory [CSHL] line GT_24505) carries a transposon insertion at nucleotide 37 in the CMI1 coding region (Fig 2A and 2B). Compared to a wild-type (WT) control, the cmi1 plants have shorter primary roots (Fig 2C and 2D) as a result of a smaller root meristem size, defined as the length of the region between the QC and the initiation of the elongation zone (Fig 2E and 2F). Importantly, the shorter primary root phenotype was complemented by pCMI1::CMI1-GUS (S3A Fig), confirming that the mutant phenotype resulted from the loss of CMI1 function and that the observed expression pattern of pCMI1::CMI1-GUS reflects the expression pattern of the endogenous CMI1 gene.
Next, we determined whether the loss of CMI1 function affects auxin response using the DR5 rev ::GFP auxin response marker [45,46]. In cmi1 mutant roots, the auxin response maximum in the QC was reduced, compared to a WT control (Fig 2G). Quantification of the green fluorescent protein (GFP) fluorescence levels revealed a significant reduction (p � 0.006, t test) in fluorescence level in the QC cells ( Fig 2H). Ectopic expression of GFP-ICR1 was detected in the QC cells of the cmi1 mutant but not in WT roots (Fig 2I and 2J), in line with the reduced auxin response in the QC [39,40]. Hence, CMI1 affects ICR1 levels, indirectly by regulating the auxin response.
The regulation of CMI1 expression by auxin prompted us to examine the possible involvement of CMI1 in further well-characterized auxin responses. The initiation and elongation of root hairs are regulated by TIR1/AFB-Aux/IAA-dependent auxin signaling [47][48][49]. We found that root hairs were longer in the cmi1 mutant compared to WT and cmi1/pCMI1::CMI1GUS plants and were elongated in response to exogenous auxin treatments ( Fig 2K and S3C Fig). TIR1/AFB-Aux/IAA-dependent auxin signaling also affects hypocotyl length [50,51]. Hypocotyls of the cmi1 mutant were significantly longer (p � 0.001, t test), compared to WT and cmi1/pCMI1::CMI1GUS plants. As expected, external IAA treatments induced hypocotyl elongation in WT, cmi1 mutant, and cmi1/pCMI1::CMI1 GUS plants (Fig 2L and S3B Fig). LR formation is regulated by both auxin response and distribution [45,[52][53][54][55]. cmi1 plants exhibited abnormal LR patterning (Fig 2M and S4 Fig), with an average of 7 LRs/cm in cmi1 compared to 4 LRs/cm in control Landsburg erecta (Ler) seedlings. Together, the changes in DR5 rev ::GFP and GFP-ICR1 expression pattern and the macroscopic phenotype of cmi1 mutant plants suggest that CMI1 regulates both the spatial distribution and the level of auxin responses.
Corresponding to the increased auxin response of cmi1 mutants, the DR5::GUS staining was stronger in cmi1 primary root and LR initials compared to WT control (Fig 3A-3J). Notably, similar to the results with the DR5 rev ::GFP, a shift in the DR5::GUS signal toward the columella in cmi1/DR5::GUS seedlings was also observed (Fig 3B, arrow).
To examine a potential impact of CMI1 function on the regulation of auxin transport and PIN polarity, we carried out whole-mount immunostaining of roots with anti-PIN1 and anti-PIN2 antibodies. Importantly, these immunostaining experiments revealed that PIN2 distribution in the cortex is altered in cmi1 (S5A Fig). In 55% of the cells, PIN2 displayed apical localization, and in another 25%, it was nonpolar. In comparison, in Ler (WT) in 90% of the cells, PIN2 displayed basal localization in the cortex, and in only 10% of the cells, it was either apical or nonpolar (S5B Fig).

CMI1 affects auxin-induced Ca 2+ signaling in a cell type/tissue-specific manner
To further examine a potential role of CMI1 in interconnecting Ca 2+ signaling and auxin function possibility, we compared the auxin-induced cytoplasmic Ca 2+ signals in the LRC, epidermis, and vasculature of WT and cmi1 roots expressing YC3.6. The quantification of Ca 2+ dynamics was carried out by calculating the ratio change in the FRET signals (ΔR/R 0 ) (Fig 4A-4F). Quantitative analyses revealed genotype-specific differences in the tissue specificity and intensity of auxin-induced cytoplasmic Ca 2+ signals between Ler (WT) and cmi1. In the LRC, the auxin-induced cytoplasmic Ca 2+ response displayed a higher amplitude in the WT Ler than in cmi1 mutants. Specifically, none of the cmi1 roots, as opposed to 35% of the analyzed WT roots, showed a ratio change ΔR/R 0 � 0.15. In contrast, 40% of the cmi1 roots displayed  lower Ca 2+ elevations, characterized by ΔR/R 0 ranging between 0 and 0.1, whereas only 5% of the WT roots displayed such low Ca 2+ signals in the LRC (Fig 4A and 4B). In the epidermis, strong increases in Ca 2+ levels (ΔR/R 0 � 0.15) predominated in both the WT and cmi1 backgrounds, with only slightly fewer samples with high Ca 2+ levels detected in cmi1 mutants ( Fig  4C and 4D). In the vasculature, the amplitude of the auxin-induced Ca 2+ signals was comparable in WT and cmi1 mutant. The low-threshold Ca 2+ signals (ΔR/R 0 ranging between 0 and 0.1) predominated in both WT and cmi1 backgrounds, and only 10% more WT roots displayed higher Ca 2+ levels, with ΔR/R 0 values ranging between 0.1 and 0.15 (Fig 4E and 4F). However, there was a striking difference in the shape of the signal and in the kinetics of the signal to reach the maximum amplitude. The WT roots evoked a maximum Ca 2+ response in 120 seconds, whereas in the cmi1 mutants, the Ca 2+ maxima were reached in 230 seconds. Interestingly, in the vasculature, the restoration of basal level of Ca 2+ followed a similar kinetics. Taken together, these results reveal that loss of CMI1 function alters auxin-induced Ca 2+ , epidermis (C and D), and vascular tissues (E and F). Note the reduced Ca 2+ levels and different kinetics in Ca 2+ decrease and increase in the lateral root cap and the vascular tissues, respectively. Error bars are SE. Underlying data for this figure can be found in S1 Data. cmi1, Ca 2+ -dependent modulator of ICR1; wt, wild type. signals, especially in the LRC and vascular cells, and suggest that CMI1 regulates auxin-associated changes in cytoplasmic Ca 2+ levels in a cell/tissue-specific fashion. Moreover, our finding points to an elaborate cell specificity and diversity in complex tissues.
To further examine the function of CMI1, we ectopically expressed mRFP-CMI1 under regulation of the ICR1 promoter (pICR1>>mRFPCMI1), using a transcription-transactivation system [56]. The roots of pICR1>>mRFPCMI1 plants were short and had reduced columella layers, indicating root meristem collapse, and reduced auxin response maxima (Fig 5A-5F). Hence, ectopic expression of CMI1 was associated with reduction in auxin levels or responses and suppression of root growth.
Previously, we demonstrated that inducing elevated levels of ICR1 in the QC by its expression under regulation of the CMI2 (At5g54490) promoter, utilizing the pOp/LhG4 transcription/transactivation system (pCMI2>>GFP-ICR1), resulted in inhibition of root growth [40] and (Fig 5G and 5H). Remarkably, coexpression of GFP-ICR1 and mRFP-CMI1 in pCMI2>>GFP-ICR1/mRFP-CMI1 resulted in suppression of root growth arrest ( Fig 5H). In LR primordium, mRFP-CMI1 was detected in nuclei, cytoplasm, and plasma membrane ( Fig  5I), similar to its distribution in the leaf epidermis pavement cells (S6A Fig) and in agreement with the protein immunoblot with anti-CMI1 antibodies that indicated localization in both soluble and insoluble fractions (S6B Fig). Examination of the subcellular localization of both GFP-ICR1 and mRFP-CMI1 revealed that when coexpressed together with GFP-ICR1 in the same cells, mRFP-CMI1 was excluded from nuclei ( Fig 5J-5L). These data indicated that coexpression of CMI1 and ICR1 in the same cells affected the subcellular distribution of CMI1 and its function.

Ca 2+ promotes the interaction between CMI1 and ICR1
CMI1 is a small 14-kDa protein containing a single EF-hand (S7A Fig). To gain the molecular mechanism of CMI1 action, we characterized its interaction with ICR1 and whether it is Ca 2 + -dependent. CMI1 interacted specifically with ICR1 but not with ICR2 (At2g37080) or ICR4 (At1g78430), the closest homologues of ICR1 ( Fig 6A). To further characterize the interaction between CMI1 and ICR1 and to examine whether it is Ca 2+ -dependent, we performed in vitro pull-down experiments using either anti-CMI1 antibodies to examine whether they coimmunoprecipitate His-ICR1 with His-CMI1 or glutathione S-transferase (GST)-ICR1 to examine whether, when bound to glutathione sepharose beads, it would precipitate His-CMI1. His-ICR1 was immunoprecipitated together with His-CMI1 using anti-CMI1 antibodies in the presence of Ca 2+ (Fig 6B). In contrast, pull-down of His-CMI1 by GST-ICR1 did not take place when the Ca 2+ was chelated with EGTA. The interaction of ICR1 and CMI1 in the pulldown assays was specific, since His-CMI1 was not precipitated by nonfused GST or glutathione beads (S7B and S7C Fig). To further corroborate that the interaction between ICR1 and CMI1 is Ca 2+ -dependent, we created a CMI1 D85N mutant in which a conserved EF-hand Asp required for Ca 2+ binding [57] was mutated to Asn (S7A Fig). Yeast two-hybrid assays showed that CMI1 interacts with ICR1 but not with the cmi1D85N protein ( Fig 6C). Taken together, these results establish that the interaction between ICR1 and CMI1 is Ca 2+ -dependent both in yeast and in vitro.
Here, CD-spec was used to examine changes in CMI1 secondary structure at different free Ca 2 + concentrations. The analysis was carried out in solutions with the following free Ca 2+ concentrations: 10 −10 M Ca 2+ (1 mM EDTA), 2 nM, 20 nM, 0.2 μM, 2 μM, 200 μM, and 2 mM Ca 2+ . Because of technical limitations, the measurements at free Ca 2+ concentrations of 2 nM and 20 nM and 0.2 μM, 2 μM, 200 μM, and 2 mM were carried out on different days. Control measurements in 1 mM EDTA solutions were carried out on both days (Fig 7A and 7B). The CMI1 CD spectra at free Ca 2+ concentrations ranging between 0.2 μM to 2 mM were similar but were all significantly different from the 1 mM EDTA Ca 2+ -free solution ( Fig 7A). The CD spectra of CMI1 in 20 nM free Ca 2+ concentrations were also significantly different from the Ca 2+ -free 1 mM EDTA solution, and also, different spectra were observed at 2 nM free Ca 2+ (Fig 7B). The percentage of α-helices that were calculated based on the CD spectra were around 40% for free Ca 2+ concentrations ranging between 0.2 μM and 2 mM and below 30% for CMI1 in the Ca 2+ -free 1 mM EDTA solution (Fig 7C). Although the percentage of α-helices ( Fig 7D) were lower compared to the measurements presented in Fig 7C, the differences in α-helix content between the 20 nM free Ca 2+ and 1 mM EDTA were around 10%, similar to the differences between the 0.2 μM-2 mM Ca 2+ and the Ca 2+ -free 1 mM EDTA solutions (compare Fig 7C and 7D). The CD-spec analysis suggested that CMI1 can bind Ca 2+ at free Ca 2+ concentrations ranging between 10 −9 and 10 −8 M, which in turn induce secondary structure changes that result in an increase in α-helicity. Steady-state cytoplasmic Ca 2+ concentrations [Ca 2+ ] cyt have been estimated to be around 100 nM [64,65]. Similar values of 50-90 nM of resting [Ca 2+ ] cyt along the root have recently been reported [66]. Thus, it appears well conceivable that CMI1 is bound to Ca 2+ and functional also at steady-state [Ca 2+ ] cyt . Furthermore, since CMI1 expression is up-regulated by auxin and highly expressed when auxin response and Ca 2+ levels are high (Fig 1), it likely functions as an auxin-dependent signaling transducer.
Many Ca 2+ -binding proteins require at least two EF-hands for their function or function as dimers if they contain a non-even number of EF-hands. We therefore hypothesized that CMI1, bearing only a single Ca 2+ -binding EF-hand, might oligomerize in solution. Therefore, the quaternary structure of CMI1 was examined by size-exclusion chromatography multiangle light scattering (SEC-MALS). To eliminate potential effects of the poly-His-tag on solution structure, we performed the analysis on recombinant bacterially expressed and purified recombinant CMI1 from which the poly-His-tag was cleaved. At both concentrations of 2 and 4 mg/ml, CMI1 eluted as a monodisperse species at around 11.5 minutes with a measured molecular mass of 16 kDa, corresponding to the monomeric form of the protein (Fig 7E and  S8 Fig). Hence, we conclude that CMI1 is strictly monomeric in vitro at least to concentrations of 25 μM in high Ca 2+ conditions. This finding, however, does not exclude that CMI1 upon interaction with additional proteins may form oligomeric assemblies.
Possible homo-and heterodimerization of CMI1 was also examined by yeast two-hybrid assays. The analysis was carried out with Clontech LexA yeast two-hybrid yeast strain EGY48, since CMI1 activates gene expression in Gal4-based yeast two-hybrid strains when expressed fused to Gal4 DNA-binding domain (Gal4-DB). Following 24 hours of incubation, a very faint blue color appeared in X-Gal assays of CMI1-LexA-BD/CMI1-Lex-activation domain (AD) compared to a strong blue color in the CMI1-ICR1 and no color in the vector-control assays. Following 48 hours of incubation, the X-Gal assays of the CMI1-BD/CMI-AD assays had a light blue color compared to the strong blue of the CMI1-ICR1 and no color in the negative vector-control assays (S9 Fig). Together, the yeast two-hybrid assays suggest that CMI1 could form dimers in yeast in the absence of ICR1 but also that the high affinity to ICR1 would interfere with this homodimerization. Therefore, the differences in the strength of the interaction  in yeast and the SEC-MALS results strongly suggest that CMI1 very likely interacts with ICR1 as a monomer.

Interaction of CMI1 with ICR1 involves a conserved hydrophobic pocket in CMI1 and a CaM binding-like domain (CBLD) in ICR1
Having established that CMI1 could function as a Ca 2+ sensor, we sought to obtain more insights into the molecular details of its structure and function. The 3D structure of CMI1 was predicted using homology modeling based on the structure of KIC, which belongs to the same subfamily single EF-hand Ca 2+ -binding proteins [41,67]. The predicted structure of CMI1 suggests the formation of two helix-loop-helix domains, one which binds Ca 2+ and one which does not. This structural feature likely enables CMI1 to function as a monomer with regard to Ca 2+ binding (Fig 8A and 8B). The CMI1 structure with Ca 2+ bound is predicted to form a hydrophobic pocket (Fig 8A, residues highlighted in yellow). Modeling of CMI1 in complex with the CaM-binding domain (CBD) of the KIC interactor kinesin-like CaM-binding protein (KCBP) [41,67] revealed that three Leu residues in the putative hydrophobic pocket of CMI1 -namely, L59, L92, and L100-can potentially serve as interacting side chains with a Trp residue in a domain that would be structurally related to a CBD, which we therefore designated as CBLD (Fig 8B).
To test the hypothesis that L59, L92, and L100 form a hydrophobic pocket, we exchanged L to A in each of the respective Leu residues and tested the interaction of this modified CMI1 versions with ICR1 in yeast. As expected, neither cmi1 mutants L59A nor L92A nor L100A interacted with ICR1 in yeast two-hybrid assays (Fig 8C), strongly suggesting that the three Leu residues are part of a hydrophobic pocket required for protein-protein interaction. To ensure that the mutations do not cause structural modifications in the hydrophobic pocket, an aligned structural model of CMI1 and its hydrophobic pocket Leu mutants was created. This model confirmed that the mutations have not caused structural changes within the core CMI1 that contains the hydrophobic pocket, confirming the requirement for the Leu residues (S10 Fig). It is very likely that the structural differences that are observed in the N-and C-terminal domains result from the low confidence of the model in these regions.
There are two Trp residues in the C-terminal end of ICR1 at positions 266 and 338 that could be part of a potential CBLD. To map a potential CMI1-interaction domain in ICR1, we generated a series of N-and C-terminal deletion mutants of ICR1 and examined their interaction with CMI1 in yeast two-hybrid assays. These analyses revealed that the 44 C-terminal residues of ICR1 (icr1-301-344) are necessary and sufficient for interaction with CMI1 ( Fig 8D). Slightly stronger yeast growth, which was apparent at dilutions of 1:10 3 and 1:10 4 , that resembled the full-length ICR1 was observed between an ICR1 C-terminal fragment encompassing residues 151 to 344 (icr1-151-344) and CMI1 (Fig 8D). Hence, the C-terminal 44-residue domain of ICR1 can function as a CBLD, but possibly, other residues also contribute to the interaction.
Next, we examined the interaction between CMI1 and ICR1 harboring the single amino acid substitutions W266A and W338A. In yeast two-hybrid assays, ICR1W266A still interacted with CMI1 whereas ICR1W338A did not (Fig 8E). Similar results were obtained when W266 and 338 were mutation to Gln (Q) (S11 Fig). Together, these results suggest that W338 is the CMI1, Ca 2+ -dependent modulator of ICR1; ICR, interactor of constitutively active ROP; M, million; MRE, mean residue ellipticity; SEC-MALS, size-exclusion chromatography multiangle light scattering.
https://doi.org/10.1371/journal.pbio.3000085.g007 primary Trp residue in ICR1 CBLD that is most crucial for interaction with residues in the hydrophobic pocket of CMI1.
Next, we examined the localization of CMI1 and ICR1 and the potential influence of interaction with ICR1 on CMI1 localization in plants. When expressed in plants, ICR1-mCherry localized to microtubules (MTs), as indicated by its colocalization with the MTs marker Tubulin alpha-6 (TUA6)-GFP (S12A-S12C Fig). When expressed by itself in Arabidopsis under control of its own promoter, CMI1 was observed at the plasma membrane, throughout the cytoplasm, and in nuclei (S6 Fig). Imaging of leaf epidermis pavement cells showed the mRFP-CMI1 is indeed localized to the plasma membrane as well as to nuclei and cytoplasm (S6A Fig). Furthermore, protein immunoblot with anti-CMI1 antibodies indicated that CMI1 is localized in soluble and insoluble fractions in different tissues (S6B and S6C Fig). However, when ICR1 and CMI1 were transiently coexpressed in Nicotiana benthamiana leaf epidermis cells, ICR1-mCherry and GFP-CMI1 were localized to MTs (Fig 9A-9C). The colocalization of both mCherry-ICR1 and GFP-CMI1 was sensitive to the anti-MT drug oryzalin, confirming that they were both localized to MTs (S12D-S12F Fig). In contrast to GFP-CMI1, neither GFP-CMI1D85N (mutated in the Ca 2+ binding EF-hand) nor GFP-CMI1L59A (mutated in the hydrophobic pocket) were recruited to MTs (Fig 9D-9F and 9G-9I). Likewise, when GFP-CMI1 was coexpressed with ICR1W338A-mCherry, it was not recruited to MTs, although ICR1W338A-mCherry was observed on MTs (Fig 9J-9L). Taken together, the coexpression assays in plants reinforced the combined conclusions derived from the results of the structural modeling and interaction assays, demonstrating that also in plant cells, the interaction between ICR1 and CMI1 is very likely Ca 2+ -dependent and involves a hydrophobic pocket in CMI1 and a C-terminal CBLD involving W338 in ICR1. These results also provide the possibility that CMI1 modulates the function of ICR1 and/or fulfills alternative functions in its ICR-bound and non-ICR-bound form.

CMI1 function in auxin-regulated Ca2+ response
The specificity of Ca 2+ response has been proposed to be regulated by the differential response of Ca 2+ sensors to fluctuations in cellular Ca 2+ levels, referred to as the Ca 2+ code/language, and the specificity of Ca 2+ sensors toward specific targets [33,68]. A third regulatory mechanism is the expression regulation of Ca 2+ -binding proteins in time and space. In the root, the uneven distribution of auxin regulates cell-specific gene expression patterns [36]. Hence, auxin-regulated Ca 2+ binding proteins are candidates for carrying out cell-specific responses. CMI1, whose expression is regulated directly through TIR1/AFB-Aux/IAA auxin receptorscoreceptors, is an example of Ca 2+ -binding protein whose cell-specific expression is regulated by auxin (Fig 1). Results of this work also indicate that CMI1 can bind Ca 2+ at free Ca 2+ concentrations ranging between 10 −9 and 10 −8 M, which are 1 to 2 orders of magnitude lower than the estimated steady-state [Ca2+] cyt . Furthermore, Ca 2+ levels are high around the root tip, where auxin levels are high and CMI1 is expressed (Fig 1). Taken together, these data indicate that CMI1 should be constantly bound to Ca 2+ . Thus, rather than functioning as a Ca 2+ sensor, CMI1 is likely a cell-specific transducer of the auxin response. dilution with icr1-151-344 C-terminal or longer fragments. (E) ICR1 Trp residue W338 but not W266 is required for the interaction between CMI1 and ICR1. (C-E) Numbers above panels denote dilutions of the yeast cells. CBD, calmodulin-binding domain; CMI1, Ca 2+ -dependent modulator of ICR1; ICR, interactor of constitutively active ROP; KCBP, kinesin-like CaM-binding protein; -LT, Leu-, Trp-deficient medium; -LTH, Leu-, Trp-, His-deficient medium. https://doi.org/10.1371/journal.pbio.3000085.g008 Studies carried out over the last several years have shown that rapid responses to auxin, which take place within minutes, depend on TIR1/AFB-Aux/IAA auxin receptors-coreceptors [1,[26][27][28]. For these rapid auxin responses, the regulation of CMI1 expression by auxin and its biochemical properties, which allow it to function irrespective of the changes in Ca 2+ levels, are intriguing. CMI1 is regulated by TIR1/AFB-Aux/IAAs and, upon its expression, can rapidly transduce nontranscriptional responses. The phenotype of cmi1 mutant plants is associated with impaired auxin responses (Figs 2  and 3). The interaction assays indicate that CMI1 interaction with ICR1 requires CBLD but that, at the same time, its interaction is highly specific, since CMI1 did not interact with other members of the ICR family, including ICR4, which is highly similar to ICR1. The binding specificity makes identification of CMI1-binding proteins by prediction extremely difficult, if not impossible. We have recently identified additional CMI1 interacting protein in highthroughput screens. However, the analysis of these proteins is beyond the scope of this work.

The function of CMI1 and its interaction with ICR1
Ectopic expression of CMI1 resulted in root meristem collapse and inhibition of root growth (Fig 5). It appears therefore that the tight regulation of CMI1 expression is critical for plant development. Interestingly, coexpression of CMI1 with ICR1 suppressed the root growth inhibition phenotype (Fig 5), suggesting that ICR1 may function in restricting CMI1 signaling. ICR1 can affect the subcellular localization of CMI1 (Figs 5 and 9). In the root meristem, CMI1 levels are highest in the QC, where ICR1 is posttranslationally degraded [39,40], suggesting that CMI1 function in QC is different than in surrounding tissues. Auxin has complex roles in regulation of the interaction between CMI1 and ICR1 and, in turn, CMI1 function because it induces both the expression of CMI1 and the degradation of ICR1 [39,40]. Auxin or other factors may also regulate the cell-specific levels of other CMI1-interacting protein(s).
Although KIC and CMI1 do not share common binding partners, it is interesting that they both interact with MT-binding proteins (this work and [41]). Unlike KCBP, which is a kinesin with enzymatic activity, ICR1 is a coiled-coil domain protein that does not contain additional known catalytic or structural domains and likely functions as a scaffold [38]. KIC inhibits interaction of KCBP with MTs and its ATPase activity [41,67]. Data in this work indicate that ICR1 can recruit CMI1 to MTs, yet it is unknown whether CMI1 affects ICR1 interaction with MTs. Unfortunately, in vitro assays to test the effect of CMI1 on ICR1 MT binding were unsuccessful because of the requirement to include Ca 2+ in the reaction medium, which in vitro leads to MTs destabilization.

Structure-function of CMI1
Despite having a single Ca 2+ -binding EF-hand, CMI1 most likely interacts with ICR1 and possibly with other target proteins as a monomer. The 3D structure of CMI1 homologue KIC revealed existence of two EF-hands: a canonical Ca 2+ -binding EF-hand and a second EF-hand that lacks essential residues for divalent ion binding. Upon Ca 2+ binding, both the Ca 2+ -binding and Ca 2+ independent EF-hands form an open conformation, creating the hydrophobic pocket that can accommodate the KCBP CBD [67]. Our structural modeling, structure-function assays, and the SEC-MALS results indicate that like KIC, CMI1 has a canonical Ca 2 + -binding EF-hand and a Ca 2+ -independent EF-hand that enable it to function as a monomer. Possibly, the oligomerization of ICR1, which contains a long coiled-coil domain [38], may induce accumulation of CMI1 molecules at discrete cellular domains.
The identification of the hydrophobic pocket in CMI1 and its interaction with ICR1 via a CBLD suggest that CMI1 may interact with other proteins that contain a CBLD. Given that the interaction of CMI1 with the C-terminal ICR1 CBLD is weaker than its interaction with longer fragments of ICR1, it is likely that the binding specificity between the proteins is determined by additional residues in ICR1. Thus, it is difficult to predict which CBLD-containing proteins would interact with CMI1.

Concluding remarks
The roles of Ca 2+ as a second messenger in responses to diverse environmental stimuli have been extensively studied and discussed. Yet, the steady-state distribution of Ca 2+ in the root and its close association with auxin distribution (this work and [24][25][26]) suggest that Ca 2+ also has important functions in regulation of root development. This work identified CMI1 as a cell-specific transducer of the auxin-Ca 2+ developmental signaling in the root.

Molecular cloning
The plasmids used in this study are listed in supplemental information S1 Table. pICR1>>GF-P-ICR1 and pCMI2>>GFP-ICR1 plants were previously described [39,40]. pCMI1::CMI1-GUS (pSY1804) was constructed by amplifying a 2,526-bp fragment containing the 2,040-bp promoter, 78-bp 5 0 -UTR, and the 408-bp CMI1 ORF, in which the TGA stop codon was changed to TAA (Leu). The resulting fragment was digested with EcoRI and SalI and cloned into pEN-TRY1a. The resulting plasmid pSY1802 was recombined with pMDC162 using LR clonase to obtain pSY1804. pCambia2300-RFP-CMI1 (pSY1351) was generated by cloning mRFP upstream to the CMI1 ORF into pCambia2300. Transactivation CMI1 promoter plasmid (pSY1806) was constructed as follows: a 2,040-bp fragment of the CMI1 promoter was amplified, digested with Sal1, and subcloned into pLhG4Bj36 upstream of the chimeric transcription factor LhG4 to create plasmid pSY1805. pSY1805 was then digested with NotI, and the resulting fragment containing pCMI1::LhG4-terminator was subcloned into pART27 plant binary plasmid to obtain pSY1806. To obtain the mRFP-CMI1 Op reporter plasmid, the mRFP-CMI1 fragment from pSY1351 was digested with HindIII and XhoI and subcloned into pOp to obtain pSY1807. Subsequently, pSY1807 was digested with NotI, and the resulting fragment containing 10XOp::mRFP-CMI1 was subcloned into the plant binary vector pMLBART to obtain pSY1808. pGAD-CMI1 was created as follows: the coding sequence of CMI1 was amplified from cDNA and subcloned into pGET (Thermo Fisher Scientific). It was then digested with BamHI and SalI, and the resulting fragment was ligated to pGAD vector to obtain pSY1565.
The generation of plasmids for yeast 2-hybrid and plant colocalization assays of site-directed and deletion mutants of CMI1 and ICR1 and plasmids for expression of CMI1 in Escherichia coli were carried out as follows. For site-directed mutagenesis (SDM), primers were designed using the QuikChange Primer Design tool found at Agilent website (https://www.genomics. agilent.com/primerDesignProgram.jsp). SDM was perform with Pfu-Ultra DNA polymerase (S2 Table) followed by digestion with DpnI (S2 Table) to eliminate unwanted templates. In two cases, when the SDM did not provide the desired mutants (cmi1 L92A ), an alternative approach of a three-step overlap extension PCR reaction using Phusion DNA polymerase (S2 Table) was performed. From this point, the cloning steps were the same as described below.
Genes of interest were cloned with flanking ends of attB1/2 recombination sites using a two-step reaction of Phusion high-fidelity DNA polymerase (S2 Table). In cases when several DNA fragments were observed in the PCR reaction products, the relevant band was extracted using QIAEX II gel extraction kit (QIAGEN) or Wizard SV Gel and PCR Clean-Up System (Promega) (S2 Table). The attB1/2 flanking genes were transferred into pDONR221 using the BP clonase reaction (S2 Table). All clones were verified by sequencing.
For yeast 2-hybrid, constructs were transferred by recombination from pEntry221 and then by recombination to pDEST22 (prey) or pDEST32 (bait) using the LR clonase (S2 Table). Bait and prey plasmids were transformed into PJ69-4a yeast strain. Presence of respective plasmids was verified by yeast colony PCR (S2 Table).
Cloning for protein expression in E. coli. A PCR product of CMI1 with flanking BamH1 and Not1 sites was subcloned into pJET1.2 using the CloneJET PCR cloning kit (S2 Table). The resulting plasmid was digested with BamH1 and Not1, and the CMI1 fragment was subcloned into pET21d.H8.Nia.yBRFc.T.GSTrc digested also with BamH1 and Not1 to isolate the pET21d-His 8 -TEV fragment. The resulting plasmid pSY2408 (pET21d_His8-TEV-CMI1) was designed to express His 8 -TEV-CMI1 fusion protein that enables purification of CMI1 on a metal chelate Ni-column and cleavage of the His 8 -tag by TEV protease.

Plant material and growth conditions
The Arabidopsis transgenic lines used in this study are listed in supplemental information S3 Table. Long-day grown (16 hours light/8 hours dark, 22˚C) Arabidopsis Columbia-0 (Col-0) and Ler ecotypes were used for stable expression, mutant phenotypic analysis, protein localization, and Ca 2+ measurement. Arabidopsis, mRFP-CMI1, and pCMI1>>mRFP-CMI1 and pICR1>>mRFP-CMI1 plants were generated by crossing pCMI1-LhG4 to pop-mRFP-CMI1 and pICR1-LhG4 to pop-mRFP-CMI1. The cmi1 mutant (CSHL_GT24505) is in the Ler background. To analyze DR5::GFP rev and pICR1>>GFP-ICR1 expression in the cmi1 mutant background, DR5::GFP rev and pICR1>>GFP-ICR1 were crossed into WT Ler and cmi1 backgrounds. M3 generation WT or cmi1 homozygote mutant plants that harbored the erecta phenotype and expressed either DR5::GFP rev or pICR1>>GFP-ICR1 were selected for further analysis. Quantification of fluorescent signals was performed using Image J. For DR5::GFP rev quantification, we used 12-16 images of independent root tips when 2-4 QC cells are in the center. Cell layers 1-6 were defined from QC to the last columella layer, and GFP signal intensity was measured in the same area (below the QC cells) in each layer using Image J. The average of the GFP intensity is presented in the graph, and the bars are the standard error (SE) (Fig  6). To quantify the ectopic expression of GFP-ICR1 in the QC cells of cmi1 mutant, 18-20 roots of each WT (Ler) or cmi1 plants were imaged when QC cells (2-4 cells) are visible in the center. The number of QC cells in which a GFP-ICR1 signal was detected was used to calculate the percentage of the roots with or without ectopic expression. Complementation of cmi1 was performed by crosses with pCMI1::CMI1-GUS plants. The analysis was performed using nonsegregating lines from the fourth and fifth generations. For Ca 2+ imaging, the pUBQ10::YC3.6 Yellow Cameleon [35] was transformed into Ler WT and cmi1 plants. Several independent transgenic lines were used for the Ca 2+ imaging.

Protein expression and antibody generation
Expression in E. coli Rosetta (DE3) and purification of recombinant His 6 -CMI1, His 6 -ICR1, and GST-ICR1 were carried out according to standard protocols using Ni-NTA (Qiagen) and Glutathione sepharose (GE) resins, as previously described [38,69]. His 8 -TEV-CMI1 was purified over Ni-NTA (Quiagen). Eluted fractions were passed through HiPrep 26/10 desalting column (GE Healthcare) with the extraction buffer (50 mM sodium phosphate buffer [pH 7.5], 300 mM NaCl, and 1 mM DTT) to ensure flushing of imidazole presence from the elution buffer (50 mM sodium phosphate buffer [pH 7.5], 300 mM NaCl, 1 mM DTT, and 250 mM imidazole). Eluted fractions were incubated overnight with His 6 -tagged TEV protease at 4˚C followed by purification over a second Ni-NTA. The untagged CMI1 was collected from the flow-through and concentrated with Amicon Ultra-15 with molecular weight cutoff (MWCO) of 3 kDa (Millipore) at 4,000g and 4˚C to a final volume of approximately 500 μl. The concentrated protein samples were filtrated through Millex 0.22-μm syringe filter (MILLIPORE) and uploaded onto a gel filtration column of HiLoad 16/600 Superdex 200 pg (GE Healthcare) and eluted with a gel filtration column buffer (60 mM MOPS [pH 7.2], 200 mM KCl, and 2 mM DTT). Purified proteins were concentrated using Amicon Ultra-15 with MWCO of 3 kDa at 4,000g and 4˚C, divided into aliquots, batch frozen in liquid nitrogen, and kept at −80˚C until further use.
Anti-CMI1 antibodies were raised in rabbits. Ni-NTA-purified His 6 -CMI was further purified by SDS-PAGE. The His 6 -CMI1 band was eluted from the gels and were used for rabbit immunization.
For coimmunoprecipitation of His 6 -ICR1 and His 6 -CMI1 with anti-CMI1 antibodies, His 6 -CMI1 and His 6 -ICR1, 1 μg of each, were incubated with shaking in 300 μL of Ca 2 + /EGTA reaction buffer for 1 hour at RT. Subsequently, 1 μL of anti-CMI1 antibodies was added, and the mixture was further incubated with shaking for 2 hours at RT. A Protein A bead (10 μL; Adar Biotech #1016-5) slurry in Ca 2+ /EGTA reaction buffer was added, and the mixture was further incubated with shaking for 1 hour at RT. Subsequently, the beads were washed 3X with 1 mL ice-cold Ca 2+ /EGTA reaction buffer and resuspended in SDS-PAGE sample buffer, and proteins were resolved by SDS-PAGE. Proteins were detected by immunoblots decorated with mouse anti-poly-His monoclonal antibodies (Sigma H-1029) and goat anti-mouse horse radish peroxidase (HRP)-conjugated secondary antibodies (BioRad).

CD-spec
Protein samples were dialyzed overnight in buffers contained 10 mM Tris-H 2 SO 4 (pH 7.5), 25 mM KCl, and 200 μM DTT. Buffers also contained CaCl 2 and EDTA at different concentrations to obtain the desired Ca 2+ -free concentrations ( Table 1). All protein samples and buffers were filtrated before use through a Millex 0.22-μm syringe filter (MILLIPORE) or Stericup 0.22-μm vacuum filtration system (MILLIPORE), respectively. Protein concentration was determined using a Bradford assay standard curve for BSA. Cuvette path length was 0.1 mm, and samples' concentrations were 60 μM. Measurements were performed using a Chirascan CD spectrometer (Applied Photophysics), ranging between 180 nm and 260 nm at 21˚C. Using the Pro-Data Viewer software (https://www.photophysics.com), each spectrum was averaged from five repeated scans. Then, raw data were corrected by subtracting the contribution of the buffer to the signal, and subtracted data were smoothed (5-nm window) and exported to Excel. In Excel, data converted from observed ellipticity to mean residue ellipticity (MRE) units using the following equation: observed ellipticity ðmillidegreeÞ pathlength ðmmÞ � concentration of protein ðMÞ � number of residues All measurements were repeated at least thrice. The α-helical content of sampled proteins was extracted from MRE values at 222 nm using the following equation [71]: The α-helical content was averaged from the three repetitions, and SE was calculated as well.

SEC-MALS
The SEC-MALS buffer containing 10 mM Tris-HCl (pH 7.5), 25 mM KCl, 200 μM DTT, and 2 mM CaCl 2 was double-filtrated through Stericup 0.22-μm vacuum filtration system (Millpore) and then through Whatman Anodisc 0.02-μm Filter Membranes (GE Healthcare). Protein samples were filtrated through Whatman Anotop 10 Plus 0.02-μm syringe filter (GE Healthcare), and their concentration was determined using a Bradford assay standard curve for BSA. Protein samples were injected into a Shodex KW404-4F column (Shodex) equilibrated overnight with the buffer above. The Agilent 1200 Series HPLC System (Agilent Technologies) was coupled with a DAWN HELEOS II light scattering detector (Wyatt Technology) and an Optilab rEX refractive index detector (Wyatt Technology). Molecular mass analyses were performed using the ASTRA software (https://www.wyatt.com/products/software/astra.html). Data were exported from the ASTRA software in order to build the graphs in Excel.

Homology modeling of CMI1
Amino acids sequences of KIC and CMI1 were underwent a multiple sequence alignment (MSA) using the MUSCLE algorithm (https://www.ebi.ac.uk/Tools/msa/muscle). The MSA results were converted into PIR format with the necessary adjustments to the solved crystal structure of KIC-KCBP complex (PDB ID code 3H4S) [67]. A pairwise alignment of KIC with CMI1 was extracted from the MSA PIR format and run using the Modeller 9.19 (https://

Plant protein colocalization assays
Colocalization assays were performed using transient expression of tested proteins by transforming A. tumefaciens GV3101 pMP90 cells harboring the respective plasmids into the abaxial side of N. benthamiana leaf epidermis essentially as previously described [38,72] with the following modifications. In cases when expression levels were too low for detection, Agrobacterium expressing the silencing suppressor protein p19 form tomato bushy stunt virus [73] were cotransformed, added at a dilution OD 600 of 0.05. Following transformation, plants were maintained in a growth room for 48 hours prior to imaging.

RNA extraction and qPCR
For auxin induction experiments, seedlings were grown vertically on 0.5X Murashige Skoog (MS) supplemented with 1% sucrose for 5 days. Before treatment, seedlings were transferred to liquid 0.5X MS with 1% sucrose for 2 hours in a growth chamber. The 0.5X MS medium was then replaced with fresh 0.5X MS medium (mock) or 0.5X MS medium containing 1 μM of NAA. Following 2 hours of incubation, seedlings were frozen in liquid nitrogen, and total RNA was extracted using the RNAeasy kit (Qiagen). PP2A was used as a reference gene. The qPCR data were normalized to the reference gene. Three biological replicates with four technical replicates were carried out for each treatment. The qPCR program was as follows: 10 minutes at 95˚C, followed by 40 cycles of 15 seconds of denaturation at 95˚C, 1 minute annealing, and elongation at 60˚C. The results were analyzed using the StepOne software.

Confocal imaging
Confocal imaging was performed using Zeis780-NLO confocal laser scanning microscope (Zeiss, Jena, Germany) with 40X air and 20X, 40X, and 63X water immersion objectives with NAs of 0.75, 0.8, 1.0, and 1.2, respectively. Proteins tagged with eGFP or GFP were visualized by excitation with an argon laser at 488 nm. Emission was detected with a spectral GaAsP detector set between 499 nm and 552 nm. Proteins tagged to mCherry or mRFP were visualized by excitation with an argon laser at 561 nm and spectral GaAsP detector set between 579 nm and 632 nm.

Microarray experiments
Arabidopsis seedlings were grown hydroponically for 6 days and subjected to auxin treatments as previously described [75]. Roots were collected after 30 minutes of exposure to either 20 nM IAA ("low auxin"), 20 μM IAA ("high auxin"), or conditioned media ("mock"). Affymetrix ATH1 arrays were hybridized with probes generated from total RNA of four biologically independent samples per treatment. Data shown are mean signal values with standard deviation.

Arabidopsis sample preparation and Ca 2+ imaging
Experiments were carried out essentially as previously described [35]. Surface-sterilized Arabidopsis Ler WT or cmi1 (3 independent lines for each) seeds expressing UBQ10-YC3.6 were plated on 0.5X-strength MS medium (Duchefa) containing 1% (w/v) sucrose, solidified with 0.8% agar (Duchefa) (pH 5.8), and stratified for 2 days in the dark at 4˚C. The plates were transferred to a growth chamber (16 hours 22˚C: 8 hours 18˚C, light: dark; 120-150 μmol m −2 s −1 light intensity), and seeds were grown vertically for 5-7 days. Single Arabidopsis seedlings (5-7 days old) were placed inside a custom-made flow-through chamber (or perfusion chamber) containing imaging buffer (5 mM KCl, 10 mM MES, and 10 mM CaCl 2 (pH 5.8), adjusted with Tris). The seedling was fixed inside the chamber with cotton wool soaked in the imaging buffer as previously described [35,76]. The chamber was placed on the stage of an inverted ZEISS Axio observer (Carl Zeiss Microimaging GmbH, Goettingen, Germany) equipped with an emission filter wheel (LUDL Electronic Products, Hawthorne, NY, USA) and a Photometrics cool SNAPHQ2 CCD camera (Photometrics, Tucson, AZ, USA). A Zeiss Plan-APOC-HROMAT 20/0.8 dry objective of the microscope was used for imaging. A xenon short-arc reflector lamp (Hamamatsu) with a 440-nm filter provided the excitation. Emission filters used were 485 nm (CFP) and 535 nm (YFP). A peristaltic pump was used for buffer circulation inside the flow-through chamber with a flow rate of 1.5 ml min −1 . YFP and CFP images were taken at 6-s intervals using the METAFLUOR software (Meta Imaging series 7.7; Molecular Devices, Downingtown, PA, USA). After monitoring the root in the buffer (continuous flowthrough) for 2 minutes, the buffer was replaced by a buffer containing 10 μM NAA (Sigma Aldrich) for 7 minutes.

Ca 2+ imaging data analysis
Offline calculation of the FRET ratio was performed using ImageJ64 software (http://rsb.info. nih.gov/ij/) with the RATIOPLUS plug-in. The intensities of CFP and YFP were measured from single CFP and YFP images as pixel intensity in arbitrary units. The ratio between YFP emission and CFP emission was calculated after background subtraction. We calculated the change in ratio R t -R 0 or ΔR, where R 0 is the basal ratio before application of the stimulus and R t is the ratio at a specific time point. We normalized the ΔR to the basal ratio value (ΔR:R 0 ) and plotted ratio graphs for each measurement. We aligned all the graphs to their first response point and plotted averaged ratio graphs.

Analysis of Ca 2+ responses
The Ca 2+ peaks were divided into low, middle, and high threshold peaks, depending on the ratio change presented as height of the peak from the base. Peaks with a ratio change of 0 to 0.1 were considered as low-threshold peaks, a ratio change of 0.1 to 0.15 as intermediate, and the peaks with ratio changes higher than 0.15 were considered as high threshold peaks. Percentage was then calculated. Fifteen to 17 seedlings were analyzed for each genotype. The average ratio graphs were calculated from six to seven measurements.

High-resolution Ca 2+ imaging
High-resolution imaging was performed as previously described [76], with a Leica DMI 6000B inverted microscope equipped with a Leica TCS SP5 laser scanning device and HDy, using the Leica confocal software (Leica Application Suite-Advanced Fluorescence 2.6.0.7266; Leica Microsystems, Wetzlar, Germany). For excitation, an argon laser with a 458-nm line was used.
The CFP and FRET emissions were collected at 473-505 and 526-536 nm, respectively. Images were acquired with a 25x objective (HCW RAPO L 25.0 × 0.95 water). Image acquisition was conducted as follows: scanning speed (400 Hz), image dimension (512 × 512), pinhole (2-4 airy unit), and line average (4). YFP and CFP images were acquired as a time series in a 6-s interval. Offline calculation of the FRET ratio was performed using ImageJ RATIOPLUS plugin.

Data analysis and statistics
The expression pattern of pCMI1::CMI1-GUS was tested in 12 transgenic lines; we present here line 1, which has a WT phenotype and segregates 1:3. The induction of CMI1 expression with auxin was reproduced more than 10 times in the lab by different people.
The microarray experiments were done with four biologically independent samples per treatment.
Ca 2+ imaging was carried out with three independent lines both of WT and cmi1 background; 15-17 seedlings were analyzed for each genotype. The average ratio graphs were calculated from six to seven measurements. Error bars represent SE.
The root length was measured in 38 and 41 seedlings for the WT and cmi1, respectively. For the root meristem measurements, 18 and 15 images of the roots with propidium iodide (PI) staining were used.
Complementation assays of cmi1 with pCMI1::CMI1-GUS were analyzed on 30-40 seedlings from each phenotype. This root assay was done 6 times with similar results.
Hypocotyl length was measured using image J in 30-50 seedlings of each genotype. This experiment was repeated more than five times with similar results.
The measurements of root hairs were performed using Image J on 20 seedlings from each genotype sampling 20-30 root hairs from each seedling.
Auxin response marker intensity was measured using Image J in different cell layers of 10-12 images of root tips expressing DR5::GFP. Identical confocal settings were used for the imaging of all samples, and the fluorescence intensity was measured for identical square areas so that it could be compared and normalized.
GFP-ICR1 abundance in the QC cells was examined in 14-20 root tips of WT or cmi1 expressing pICR1>>GFPICR1. For quantification, QC cells expressing GFP-ICR1 were counted.
Immunostaining with anti-PIN1 and PIN2 antibodies was reproduced three times each with 10-15 roots for each genotype.
Protein-protein interactions in yeasts were tested 4 to 10 times (each time with four different colonies for the assay).
In vitro assays were done 3 times with similar results.
Negative results with cmiD85N-ICR1 were reproduced in three independent experiments with four replicates each time.
All experiments in N. benthamiana were repeated at least three times; the representative images were used for preparation of the figures.
The means and the SEs were calculated using Excel. For statistical analysis when two groups were compared, we used the t test to calculate the p-value (using IBM SPSS 23 Statistics). p-Values of the relevant tests are noted in the figure legends.