AGD1, a class 1 ARF-GAP, acts in common signaling pathways with phosphoinositide metabolism and the actin cytoskeleton in controlling Arabidopsis root hair polarity


  • Cheol-Min Yoo,

    1. Plant Biology Division, The Samuel Roberts Noble Foundation Inc., 2510 Sam Noble Parkway, Ardmore, OK 73401, USA
    Search for more papers by this author
  • Li Quan,

    1. Plant Biology Division, The Samuel Roberts Noble Foundation Inc., 2510 Sam Noble Parkway, Ardmore, OK 73401, USA
    Search for more papers by this author
  • Ashley E. Cannon,

    1. Plant Biology Division, The Samuel Roberts Noble Foundation Inc., 2510 Sam Noble Parkway, Ardmore, OK 73401, USA
    2. Department of Chemistry, College of Science and Mathematics, Midwestern State University, 3410 Taft Blvd. Wichita Falls, TX 76308, USA
    Search for more papers by this author
    • Present address: Section of Molecular Cell and Development Biology, University of Texas, Austin, TX 78712, USA.

  • Jiangqi Wen,

    1. Plant Biology Division, The Samuel Roberts Noble Foundation Inc., 2510 Sam Noble Parkway, Ardmore, OK 73401, USA
    Search for more papers by this author
  • Elison B. Blancaflor

    Corresponding author
    1. Plant Biology Division, The Samuel Roberts Noble Foundation Inc., 2510 Sam Noble Parkway, Ardmore, OK 73401, USA
    Search for more papers by this author

(fax +1 580 224 6692; e-mail


The Arabidopsis thaliana AGD1 gene encodes a class 1 adenosine diphosphate ribosylation factor-gtpase-activating protein (ARF-GAP). Previously, we found that agd1 mutants have root hairs that exhibit wavy growth and have two tips that originate from a single initiation point. To gain new insights into how AGD1 modulates root hair polarity we analyzed double mutants of agd1 and other loci involved in root hair development, and evaluated dynamics of various components of root hair tip growth in agd1 by live cell microscopy. Because AGD1 contains a phosphoinositide (PI) binding pleckstrin homology (PH) domain, we focused on genetic interactions between agd1 and root hair mutants altered in PI metabolism. Rhd4, which is knocked-out in a gene encoding a phosphatidylinositol-4-phosphate (PI-4P) phosphatase, was epistatic to agd1. In contrast, mutations to PIP5K3 and COW1, which encode a type B phosphatidylinositol-4-phosphate 5-kinase 3 and a phosphatidylinositol transfer protein, respectively, enhanced the root hair defects of agd1. Enhanced root hair defects were also observed in double mutants to AGD1 and ACT2, a root hair-expressed vegetative actin isoform. Consistent with our double-mutant studies, targeting of tip growth components involved in PI signaling (PI-4P), secretion (RABA4b) and actin regulation (ROP2), were altered in agd1 root hairs. Furthermore, tip cytosolic calcium ([Ca2+]cyt) oscillations were disrupted in root hairs of agd1. Taken together, our results indicate that AGD1 links PI signaling to cytoskeletal-, [Ca2+]cyt−, ROP2-, and RABA4b-mediated root hair development.


The final shape of a plant cell is dictated by the manner in which it grows. Plant cells exhibit either diffuse growth, which involves the deposition of wall material along several points in the expanding cell, or tip growth in which trafficking of membrane vesicles carrying cell wall precursors is restricted to a defined point within the cell (Smith and Oppenheimer, 2005). In higher plants, the latter type of growth is a feature of pollen tubes and root hairs, which both attain cylindrical shapes and reach lengths that are several-fold greater than their width (Cole and Fowler, 2006). Research spanning several years has established that tip growth is driven by interconnected signaling pathways that involve a dynamic cytoskeleton, establishment of ion gradients and membrane trafficking (Šamaj et al., 2006; Cheung and Wu, 2008; Yalovsky et al., 2008). However, mechanisms by which these fundamental cellular processes interact to regulate polar growth in plant cells remain poorly understood.

Arabidopsis root hairs provide a convenient system to uncover mechanisms underlying tip growth control because of their amenability to microscopic observation, their highly predictable growth behavior, and the ease by which root hair mutants can be evaluated (Bibikova and Gilroy, 2002). Indeed, several components of the root hair tip growth machinery have been identified using genetic and cell biological approaches. For instance, the understanding of the role of phosphoinositide (PI) metabolism in tip growth has advanced significantly from genetic studies of Arabidopsis root hairs (Thole and Nielsen, 2008; Heilmann, 2009). In such studies, mutating or overexpressing genes that encode enzymes that are involved in PI metabolism, such as phosphatidylinositol-4-phosphate (PI-4P) phosphatase, phosphatidylinositol 4-OH kinase (PI-4K) and type B phosphatidylinositol-4-phosphate 5-kinase 3 (PIP5K3), resulted in root hairs with a range of morphological defects (Preuss et al., 2006; Kusano et al., 2008; Stenzel et al., 2008; Thole et al., 2008).

Other signaling molecules that have been implicated in root hair polarity are small guanosine triphosphatases (GTPases). In eukaryotic cells, small GTPases function as molecular switches that cycle between an active, GTP-bound, and an inactive, GDP-bound state to regulate various cellular processes (Donaldson and Jackson, 2011). In Arabidopsis, the small GTPase superfamily of proteins belong to four groups namely RAB, RHO, ARF and RAN, which have four to 57 members within each group (Vernoud et al., 2003). One small GTPase that has been linked to tip growth is RABA4b where its targeting to post-Golgi compartments in the apex of Arabidopsis root hairs is strongly linked to sustained growth (Preuss et al., 2004, 2006).

Switching of small GTPases from the GTP-bound to the GDP-bound configuration and vice versa require the action of other regulatory proteins. Activation is facilitated by guanine nucleotide exchange factors (GEFs), whereas GTPase-activating proteins (GAPs) enhance GTPase activity leading to fast and localized inactivation of the small GTPase (Bos et al., 2007). The importance of small GTPases in root hair growth is further supported by studies that modified the activity of the small GTPase. For instance, when a constitutively active (GTP-locked)-ROP (for RHO Of Plants) was expressed in plants, it triggered the formation of root hairs with various polar growth defects (Molendijk et al., 2001; Jones et al., 2002; Bloch et al., 2005).

From previous work, we identified a recessive Arabidopsis mutant (agd1) that had wavy root hairs and a large percentage of two root hairs originating from a single initiation site on the trichoblast (Yoo et al., 2008). This mutant was disrupted in a gene encoding a class 1 ARF-GAP called AGD1 for ARF-GAP Domain-containing protein (Vernoud et al., 2003). In mammalian cells, the ACAP subfamily of ARF-GAPs is closely related to AGD1. ACAP is an acronym for ARF-GAP containing a Coiled coil domain, Ankyrin repeats and pleckstrin homology (PH) domain. The coiled coil domain was later found to be a Bin1-Amphiphysin-Rvs167p/Rvs161p (BAR) domain (Kahn et al., 2008) that has been implicated in small GTPase binding, membrane curvature sensing, protein dimerization and cytoskeletal interaction (Habermann, 2004). The PH domain of ACAP binds to PIs and this process was shown to be essential for stimulating GAP activity (Jackson et al., 2000). Indeed, recombinant AGD3 (also called VAN3/SCF; Sieburth et al., 2006), another member of the class 1 ARF-GAPs in Arabidopsis, was shown to bind to PI-4P (Koizumi et al., 2005; Carland and Nelson, 2009). Because of the multidomain structure of many mammalian and yeast ARF-GAPs, they have been proposed to facilitate the spatial and temporal assembly of signaling complexes to trigger a specific cellular response in addition to their classical role as inactivators of ARF-GTPases (Donaldson and Jackson, 2011). However, details about the signaling and regulatory roles of the ACAP-type plant ARF-GAPs are not fully understood.

Like the mammalian ARF-GAPs (Randazzo et al., 2007), AGD1 has been implicated in remodeling the cytoskeleton because mutations in AGD1 induced bundled actin and microtubules at the growing root hair tip that mirrored the phenotype of a mutant disrupted in a gene encoding a kinesin microtubule motor protein (Sakai et al., 2008; Yoo et al., 2008). The connection between AGD1 and cytoskeletal function is further substantiated by observations that the morphological defects of agd1 root hairs resembled wild type root hairs treated with cytoskeletal disrupting drugs (Bibikova et al., 1999). However, the mechanisms by which AGD1 facilitates cytoskeletal-driven polar growth of root hairs remain unresolved.

To better understand how AGD1 fits within the network of molecular players that control root hair growth, we conducted genetic interaction studies between agd1 and other root hair developmental mutants, and isolated genetic modifiers of agd1. Because AGD1 is predicted to interact with PIs through its PH domain (Vernoud et al., 2003), and our previous work showed that agd1 root hairs have altered actin dynamics (Yoo et al., 2008), we focused on evaluating double mutants to AGD1 and genes encoding enzymes involved in PI metabolism (Stenzel et al., 2008; Thole et al., 2008), and ACT2, which encodes a root hair-expressed actin isoform (Nishimura et al., 2003). These genetic studies together with the identification of cow1 (Bohme et al., 2004; Vincent et al., 2005) as an agd1 enhancer, support the hypothesis that AGD1 functions in common signaling pathways with PIs to facilitate remodeling of the actin cytoskeleton and directional membrane trafficking during root hair development. Furthermore, live cell imaging suggest that AGD1 restricts the distribution of tip growth components such as RABA4b, PI-4P, [Ca2+]cyt, and ROP2 to specific cellular domains to maintain the default, straight growth behavior of root hairs.


AGD1 functions in converging signaling pathways with phosphoinositides and actin in controlling root hair development

The PH domain of AGD1 (Vernoud et al., 2003) suggests that PIs and AGD1 might act cooperatively in signaling pathways that regulate root hair polarity. To determine the genetic basis for AGD1 and PI interaction, we analyzed root hair morphology of agd1 rhd4 and agd1 pip5k3 double mutants. RHD4 encodes a PI-4P phosphatase, which when mutated resulted in the root hairs with various defects (Thole et al., 2008). PIP5K3 on the other hand encodes a type B phosphatidylinositol-4-phosphate 5-kinase 3 that catalyzes the formation of phosphatidylinositol 4,5-bisphosphate (PI4,5P2) from PI-4P. A mutation in PIP5K3 resulted in plants with root hairs that were shorter than wild type (Kusano et al., 2008; Stenzel et al., 2008). We found that agd1 rhd4 double mutants were similar to rhd4 in regard to overall root hair morphology and length (Figure 1a–e) indicating that rhd4 is epistatic to agd1. In contrast, root hairs of agd1 pip5k3 double mutants were significantly shorter than wild type, pip5k3 or agd1 (Figure 1e–g). In addition, agd1 pip5k3 double mutants exhibited two root hairs originating from a single initiation site; a feature that is typical of agd1 but not pip5k3 (Figure 1g,h).

Figure 1.

 Root hair phenotypes of agd1 rhd4 and agd1 pip5k3 double mutants.
(a, b) Wild type and agd1 root hairs.
(c) rhd4-1 has altered root hair morphology (arrowheads, inset).
(d) Root hairs of double agd1 rhd4 mutants.
(e) Quantification of root hair length in various double mutants. Data are means (±SD) from >300 root hairs. Means with different letters are significantly different as determined by Tukey’s tests (< 0.005).
(f) pip5k3 has root hairs that are shorter than wild type.
(g) The average root hair length of double agd1 pip5k3 mutants is significantly less than pip5k3 or agd1 and plants have two tips originating from a single initiation site (arrowheads, inset).
(h) agd1 pip5k3 has a similar percentage of root hairs with two tips originating from one initiation point as agd1. Values (means ± SE) with different letters are significantly different by Tukey’s tests (< 0.005) ( 30 roots sampled). Bars = 50 μm.

We also generated double mutants to agd1 and act2. Single act2 mutants were reported to have root hair morphological abnormalities such as irregular diameters and swollen bases (Ringli et al., 2002) that partially resembled some of the defects of agd1 (Yoo et al., 2008). Here, act2 single mutants had mostly irregular root hair diameters with thicker bases and root hairs that were shorter than wild type or agd1 (Figure 2a–d). In double agd1 act2 mutants, basal swelling became more pronounced compared to agd1 and act2 single mutants particularly in root hairs close to the root-hypocotyl junction (Figure 2e,f). Furthermore, unlike agd1, which typically had two tips originating from a single initiation point, agd1 act2 double mutants often had three or more tips originating from the same initiation point (Figure 2g,h). Several root hairs also developed into short, thick stubs that formed on two initiation points within the same trichoblast (Figure 2i), and the average root hair length of agd1 act2 double mutants was less than that of agd1 or act2 (Figure 2d).

Figure 2.

 Root hair phenotypes of agd1 act2 double mutants.
(a–c) Wild type, act2 and agd1 root hairs. Root hairs of act2 are irregular in diameter (arrow). (b) Occasionally swell at the base (arrowheads).
(d) Quantification of root hair length. Values are means of 100–200 root hairs from 15 to 20 independent seedlings for each genotype. Means (±SD) with different letters are significantly different (< 0.005; Tukey’s test).
(e–i) Root hair morphology of agd1 act2 mutants. Some root hairs of agd1 act2 double mutants are wavy but basal swelling is more pronounced (arrows, e and f). Double mutants also have more root hairs with three or more tips originating from a single initiation point (arrows, g, h) and two initiation points from the same trichoblast (arrows, i). Bars = 50 μm.

Cow1 enhances agd1 root hair defects

We mutagenized agd1 seed with ethyl methane sulfonate (EMS) and isolated a mutant with increased number of root hair tips originating from one initiation point. This enhancer mutant, designated agd1 eag5 (for enhancer of agd1:Figure 3a) was back-crossed with agd1. This cross resulted in an F2 population that segregated 3:1 for the agd1 eag5 phenotype, which indicated that eag5 is a recessive, single locus, loss-of-function mutant. When agd1 eag5 mutant was back-crossed to the wild type, the F2 progeny contained wild type, agd1, agd1 eag5 phenotypes and an additional fourth root hair phenotype, which we later identified as single eag5 mutants. Unlike the original agd1 eag5 enhancer mutant, which had root hairs with up to five tips originating from one initiation site (Figure 3a), single eag5 had thicker root hairs that were longer than those of agd1 eag5 and had two tips from one initiation site (Figure 3b). Mapped-based cloning of eag5 identified a missing G (373rd base pair from ATG start) in the cDNA of the At4g34580 gene resulting in a premature stop codon (Figure S1). At4g34580 was shown to encode a sec14p domain phosphatidyl inositol (PtdIns) transfer protein (PITP), previously designated as COW1 (for Can Of Worms1 or AtSFH1p;Bohme et al., 2004; Vincent et al., 2005). Another mutant allele of cow1 (SALK_002124; Figure S1) had similar root hair phenotypes as single eag5 (Figure 3c), and a cross between SALK_002124 and agd1 was able to reconstitute the original agd1 eag5 root hair phenotype (Figure 3d), which indicated that EAG5 is COW1. Confocal microscopy of single eag5 mutants expressing a GFP filamentous actin (F-actin) reporter (Wang et al., 2008) revealed extensive alterations in their root hair actin cytoskeleton. In some eag5 root hairs, a ring of transverse actin bundles was observed in the sub-apical region. These transverse actin bundles were never observed in wild type (Figures 3e,f). Furthermore, distinct actin bundles extended into the extreme apex of eag5 root hairs as evident from increased fluorescent labeling at the tip (Figure 3g).

Figure 3.

cow1 enhances agd1 root hair defects.
(a) agd1 eag5 has root hairs with multiple tips from one initiation point (arrowheads).
(b) Single eag5 mutants have short root hairs with two tips originating from one initiation point (arrowheads).
(c) Root hairs of cow1 resemble those of eag5.
(d) Root hairs of agd1 cow1 double mutants resemble that of agd1 eag5.
(e, f) Actin organization is altered in eag5 (arrowheads). Bent arrows in e and f indicate computer generated cross-section at various distances from the root hair tip.
(g) Quantification of actin bundling in wild type and eag5 root hairs. The average fluorescence intensity from the extreme apex and base of more than 50 root hairs ± standard deviation (SD) from 10 seedlings were measured using image j, as indicated by the white boxes in the inset. A higher apical to basal ratio is indicative of more extensive tip actin bundling. Asterisk indicates statistically significant difference according to a Student’s t-test (< 0.01). Bar in (d) is for (a-d) = 50 μm. Bar in (e) is for (e-f) = 20 μm.

AGD1 stabilizes phosphatidylinositol-4-phosphate lateral plasma membrane domains to sustain straight growth in root hairs

Evidence for genetic interactions between AGD1 and PI signaling prompted us to inquire whether the distribution of PI pools in root hairs of agd1 mutants was perturbed. To address this question, a PI-4P biosensor (YFP-FAPP1; Vermeer et al., 2009) was expressed in agd1. Using spinning-disc confocal microscopy we found that PI-4P pools were highly enriched along the lateral sub-apical plasma membrane of wild type root hair tips. The plasma membrane PI-4P pools were maintained symmetrically and moved forward as the root hair elongated (Figure 4a and Video Clip S1). In wavy root hairs of agd1, however, PI-4P plasma membrane pools became asymmetrically distributed along the sub-apical flanks of the root hair. PI-4P plasma membrane pools preferentially accumulated along the future concave side of agd1 root hairs and such asymmetry became visible prior to the change in growth direction of the cell (Figure 4b and Video Clip S2).

Figure 4.

 Redistribution of PI-4P domains in wild type and agd1 root hairs.
(a) Distinct YFP-FAPP1 signals move forward on the lateral plasma membrane of the root hair tip as it grows (arrows).
(b) In agd1 root hairs, YFP-FAPP1 preferentially accumulates along the future concave side of the root hair (arrows). Bars = 10 μm. Time-lapse movie sequences are presented as Video Clips S1 and S2.

Tip [Ca2+]cyt oscillations are dampened in root hairs of agd1

Tip-growing cells are characterized by tip-focused [Ca2+]cyt gradients and oscillations, which are important in maintaining tip growth (Gu et al., 2005; Monshausen et al., 2008). [Ca2+]cyt signaling typically acts downstream of PIs (Munnik and Nielsen, 2011), and given genetic evidence that PIs interact with AGD1, we next asked if [Ca2+]cyt oscillations were perturbed in agd1 root hairs. We used sensitized fluorescence resonance energy transfer (FRET) imaging of yellow cameleon (YC) 3.60 (Rincon-Zachary et al., 2010) to study tip [Ca2+]cyt dynamics. We found that a majority (62%) of growing wild type root hairs displayed the typical two to four peak per minute [Ca2+]cyt oscillations reported previously (Monshausen et al., 2008; Figure 5a,b) while 38% exhibited occasional gaps between maximal [Ca2+]cyt peaks (Figure 5c). Although the predominant polarity defect of agd1 root hairs was wavy growth, some root hairs exhibited mild waving while others were partially swollen but straight (Figure 5d). In this class of agd1 root hairs, only 22.5% exhibited [Ca2+]cyt oscillations similar to wild type whereas 62.5% had mostly large gaps between [Ca2+]cyt peaks, and 15% had severely dampened [Ca2+]cyt peaks (Figures 5e–g). Disruption of [Ca2+]cyt dynamics was most pronounced in agd1 root hairs that were both swollen and wavy, and those that had two tips (Figure 5h). In this class of root hairs, 66% displayed large gaps between maximal [Ca2+]cyt peaks while 28% showed no prominent [Ca2+]cyt peaks (Figures 5i–k). Representative time-lapse movies of [Ca2+]cyt dynamics in root hairs are presented as Video Clips S3–S5.

Figure 5.

 Tip [Ca2+]cyt oscillations are dampened in root hairs of agd1.
(a) Tips of wild type root hairs are characterized by elevated [Ca2+]cyt at the tip.
(b) Peaks of high [Ca2+]cyt (asterisks) are spaced regularly in a majority of wild type root hairs.
(c) In some wild type root hairs, gaps between [Ca2+]cyt peaks were observed (arrows).
(d–k) Polarity defects in agd1 range from mild waving (d) to root hairs with two tips (h). [Ca2+]cyt at the tips of these different classes of agd1 root hairs was still elevated (d, h) and in some measurements, distinct [Ca2+]cyt peaks were observed (asterisks e and i). In a majority of agd1 root hairs, gaps between the [Ca2+]cyt peaks were more prevalent (arrows f and j). Some agd1 root hairs had reduced or missing [Ca2+]cyt peaks (g, k). Numbers in parenthesis indicate the percentage of root hairs that exhibited the specific [Ca2+]cyt signatures shown in each panel. Bars = 10 μm. Time-lapse sequences of [Ca2+]cyt changes of root hairs depicted in panels (a), (d) and (h) are presented as Video Clips S3–S5.

AGD1 binds to phosphoinositides

To determine whether AGD1 binds to PIs, recombinant AGD1 with the maltose binding protein (MBP) tag without the BAR domain was purified using affinity chromatography. Like AGD3/VAN3, it was first necessary to delete the AGD1 BAR domain due to unstable expression in bacteria (Koizumi et al., 2005; Carland and Nelson, 2009). The purified AGD1-MBP was used to probe membrane strips prespotted with 15 different lipids. Membrane strips were then incubated with anti-MBP antibodies to detect bound protein. We found that AGD1-MBP bound to phosphatidylinositol-3-phosphate (PI-3P), PI-4P and phosphatidylinositol-5-phosphate (PI-5P). The strongest signal from the protein–lipid binding assay was observed for PI-5P. No signal was detected in negative At1g68200-MBP controls (Figure S2).

RABA4b- and ROP2-GTPase targeting to the root hair tip is disrupted in agd1

Because RABA4b localization in wild type root hairs were shown to be dependent on [Ca2+]cyt (Preuss et al., 2006), we investigated whether the tip localization of an enhanced yellow fluorescent protein (EYFP)-RABA4b- construct (Preuss et al., 2004) was altered in root hairs of agd1. In root hairs of wild type seedlings, EYFP–RABA4b was enriched at the tips and such enrichment was maintained as the root hair elongated (Figure 6a and Video Clip S6; Preuss et al., 2004). In contrast, tip targeting of EYFP–RABA4b was compromised in root hairs of agd1. The most obvious defect observed in agd1 root hairs was a constant shifting in the position of EYFP–RABA4b that followed the direction of root hair growth (Figure 6b). In many cases, tip EYFP–RABA4b would dissipate and reform as the root hair changed direction (Video Clip S6). The dissipation of EYFP–RABA4b was most pronounced in root hairs that had two tips (Figure 6c) and root hairs that had swollen bases (Figure 6d). Restoration of the EYFP–RABA4b gradient at the root hair tip often coincided with the resumption of growth (Figure 6d and Video Clip S6). Disruption of the EYFP–RABA4b tip gradient was quantified by obtaining the ratio of the length of time during the growth of the root hair when EYFP–RABA4b signal was absent at the tip to the total elapsed time of the entire movie sequence. Thus, a larger ratio would be indicative of more extensive dissipation of EYFP–RABA4b signal at the tip. Indeed, a large percentage of agd1 root hairs exhibited extended periods of EYFP–RABA4b disruption (Figure 6e) resulting in a significantly higher average ratio compared with wild type root hairs (Figure 6f).

Figure 6.

 Delivery of RABA4b secretory vesicles to the root hair tip is disrupted in agd1.
(a) EYFP–RABA4b is enriched at the tips of wild type root hairs (arrowhead).
(b) In wavy root hairs, the EYFP–RABA4b gradient (arrowhead) follows the change in root hair growth direction.
(c) In a root hair with two tips, the EYFP–RABA4b gradient periodically dissipates (asterisks) and then reforms (arrows).
(d) In a root hair with a swollen base the EYFP–RABA4b is absent (asterisk) but reforms when focused tip growth resumes (arrows). Time-lapse movie sequences are presented as Video Clip S6. Bars = 20 μm.
(e, f) Quantification of EYFP–RABA4b dynamics in wild type and agd1 root hairs. Ratio values are means from time-lapse movies of 25–86 root hairs ± SE. Asterisks indicates statistically significant difference according to a Student’s t-test (< 0.01).

Another component of the tip growth machinery is ROP2, which has been shown to localize to the apical plasma membrane of Arabidopsis root hairs and control the organization of fine actin at the tip (Jones et al., 2002; Xu and Scheres, 2005). Because actin dynamics in root hairs of agd1were altered (Yoo et al., 2008), we asked whether patterns of ROP2 localization were disturbed in root hairs of agd1. Consistent with previous findings, wild type root hairs displayed strong localization of EYFP–ROP2 (Xu and Scheres, 2005) to the apical plasma membrane that was maintained as the root hair elongated (Figure 7a). Like wild type, EYFP–ROP2 localized to the apical plasma membrane of growing agd1 root hairs. However, in contrast with wild type, EYFP–ROP2 plasma membrane signal in agd1 gradually shifted to one side of the root hair apex. The accumulation of EYFP–ROP2 along defined plasma membrane domains of the cell often preceded the change in direction of agd1 root hairs (Figure 7b,c).

Figure 7.

 Plasma membrane targeting of ROP2 to the tips of root hairs is altered in agd1.
(a) A wild type root hair shows enrichment of EYFP–ROP2 at the apical plasma membrane (arrows).
(b, c) In agd1 root hairs, plasma membrane localized EYFP–ROP2 shifts position prior to the change in root hair growth direction (arrows). Bars = 10 μm.

Internalization of the endocytic tracer FM 1-43 is not inhibited in agd1 root hairs

Because ARF-GTPases are involved in clathrin-mediated endocytosis (Chen et al., 2011), we asked whether the uptake of the fluorescent endocytic tracer (N-(3-triethylammoniumpropyl)-4-(4-[dibutylamino]styryl) pyridinium dibromide (FM 1-43) is impaired in agd1 root hairs. Root hairs were treated for 5 min with FM 1-43 and washed for 5 min prior to imaging. Similar to a previous study, FM 1-43 prominently stained the plasma membrane of wild type and agd1 root hairs within 3 min of washing (Ovecka et al., 2005; Figure 8a). By 10 min, FM 1-43 was internalized by the root hair cell and distinct labeling at the extreme apex of both wild type and agd1 root hairs was observed. In both wild type and agd1, FM 1-43 plasma membrane labeling became progressively weaker 15 min or more after washing but apical labeling remained distinct (Figure 8a). Root hairs were then observed 2 h after washing and we found that FM 1-43 incorporated into punctate bodies in both wild type and agd1 (Figure 8b), which again was consistent with previous observations (Ovecka et al., 2005). To quantify internalization of FM 1-43 dye, we obtained the ratio of average plasma membrane fluorescence to average internalized fluorescence at the apical 5 μm of the root hair apex (Figure 8c). We found that FM 1-43 uptake in growing wild type root hairs was similar to that of agd1 root hairs (Figure 8d). Although, uptake of FM 1-43 was not inhibited in agd1 root hairs, the internalized apical labeling of FM 1-43 shifted with the direction of root hair growth (Figure 8a), which was reminiscent of EYFP–RABA4b dynamics (see Figure 6 and Video Clip S6).

Figure 8.

 Uptake of FM 1-43 is not inhibited in agd1 root hairs.
(a) Pulse treatment of growing roots hairs with FM 1-43. Dye is incorporated into the plasma membrane at 3 min (arrows) after washing and internalization along the apical cytoplasm occurs at later time points (arrowheads).
(b) Two hours after washing, FM 1-43 accumulates in punctate endosomal-like compartments (arrows). Bars in (a) and (b) = 10 μm.
(c) FM 1-43 internalization was quantified by measuring the average fluorescence intensity in the apical cytoplasm (1) and the plasma membrane (2) (arrow).
(d) The fluorescence ratio of internalized FM 1-43 to plasma membrane FM 1-43 (1:2 in c). Values are means from 15 to 20 elongating root hairs from 20 independent seedlings ± standard error (SE). Differences between wild type and agd1 ratios at all times points were not significant according to a Student’s t-test (< 0.05).


In this paper, we present new insights into the nature of the root hair polarity defects of agd1. Arabidopsis root hair development involves intersecting signaling pathways with the emergence of a single root hair from one initiation point in the trichoblast as one of the earliest events in this process. Soon after initiation, the machinery to maintain the targeted delivery of secretory vesicles to the extreme apex commences so that tip growth is sustained until the root hair matures (Bibikova and Gilroy, 2002). The root hair phenotypes (i.e. reduced length, two tips originating from one initiation point and wavy growth) of agd1 mutants indicate that AGD1 functions in most stages of root hair development. Furthermore, double mutant analyses suggest that directional membrane trafficking to sustain root hair growth requires the coordinated action of AGD1 and PIs. Genetic interaction studies also indicate that the stage of root hair development defines the nature of such interactions. For example, the similar percentage of two tips originating from a single initiation point in agd1 pip5k3 double mutants and agd1, show that agd1 is epistatic to pip5k3 but only with regard to the control of root hair initiation. With regard to root hair length, root hairs of agd1 pip5k3 double mutants were significantly shorter compared with agd1 or pip5k3. This synergistic reduction in root hair length point to overlapping functions for AGD1 and PIP5K3 in maintaining tip growth after initiation.

AGD1 and PIP5K3 could be linked through the dynamic regulation of endogenous PI-4P or PI4,5P2 levels (Kusano et al., 2008; Stenzel et al., 2008). In pollen tubes, the production of PI-4P and PI4,5P2 can act synergistically to regulate secretion of pectin, and the conversion of PI-4P to PI4,5P2 is crucial for maintaining normal pollen tube morphology (Ischebeck et al., 2010). In this regard it is noteworthy that transient overexpression of some enzymes in the PI pathway triggered morphological changes in pollen tubes and root hairs that resembled the root hair phenotypes of agd1 (Stenzel et al., 2008; Ischebeck et al., 2010). These observations support the notion that AGD1 regulates tip growth through a PI-dependent pathway. The observation that rhd4 was epistatic to agd1 indicates that RHD4 functions earlier during root hair development than AGD1. RHD4 and AGD1 might coordinate polar root hair growth through the timely depletion of PI-4P levels (Thole et al., 2008).

The link between AGD1 and PI signaling is also supported by the identification of cow1 as an enhancer of agd1. In yeast, sec14p PITPs are known to mediate the energy-independent transfer of PtdIns monomers between membrane lipid bilayers in vitro (Mousley et al., 2007). As for agd1, eag5 (cow1) single mutants had an extensively bundled actin cytoskeleton at the root hair tip. Consistent with these findings, the loss of fine actin in root hairs of AtSFH1p, another mutant allele of cow1, was reported previously (Vincent et al., 2005). Therefore, actin disruption at the root hair tip induced by mutations in COW1 and AGD1, and the enhanced root hair defects of agd1 eag5 (cow1), support the notion that the sec14p PITP encoded by COW1 acts cooperatively with AGD1 in modulating actin dynamics during root hair initiation. In yeast, it was proposed that PITPs and ARF-GAPs are part of common secretory pathways wherein PITPs generate a suitable lipid environment that stimulates ARF-GAP activity (Wong et al., 2005; Mousley et al., 2007). Interestingly, the multiple root hair tips of agd1 eag5 (cow1) double mutants resembled the phenotype of scn1, which is disrupted in a gene encoding a ROP – guanine dissociation inhibitor (GDI), a protein that prevents ROP activation by GEFs (Carol et al., 2005). Taken together with the observation that ROP2 was mistargeted in agd1 root hairs, and that multiple root hair tips were also observed in agd1 act2 double mutants, COW1 and AGD1 might function cooperatively with SCN1 in the spatial control of root hair initiation through mechanisms that involve actin remodeling (Carol et al., 2005).

Although mutant analyses presented here support a connection between PI metabolism and AGD1 in modulating root hair development, it is important to combine such studies with the assessment of the spatial distribution of PIs (Heilmann, 2009). Here, we found that the sub-apical plasma membrane localization of PI-4P pools was altered in agd1 root hairs. The sub-apical PI-4P domains observed resembled other components that control pollen tube tip growth, including (PI-4P) 5 kinases (Klahre and Kost, 2006; Ischebeck et al., 2008, 2011; Zhao et al., 2010), which suggested that, in root hairs, PI-4P might serve as a substrate for the formation of PI4,5P2. The altered stability of PI-4P sub-apical domains in agd1 root hairs suggests either altered PI-4P biosynthesis or differential breakdown of PI-4P. The latter scenario can occur by the action of PI-4P phosphatases, phospholipase C or PIP5K3, which leads to the production of PI4,5P2 (Dowd et al., 2006; Stenzel et al., 2008; Thole et al., 2008). Because overexpressing PIP5K3 triggers polarity defects in Arabidopsis root hairs that partly mirror the agd1 mutant phenotype (Stenzel et al., 2008), it is possible that altered PI4,5P distribution contributes to disrupted actin dynamics in agd1 root hairs (Yoo et al., 2008). This notion is supported by a recent report that demonstrated that PI4,5P2 produced by PI-4P 5 kinases impacts fine actin structures in the pollen tube apex via the ROP/RAC pathway (Ischebeck et al., 2011). The mistargeting of ROP2 in the root hair apex of agd1 might be a consequence of altered PI4, 5P2 signaling as is the case for pollen tubes (Ischebeck et al., 2011).

Like AGD3/VAN3, we found that AGD1 could bind to PI-4P, which suggested that this PI species might also be an AGD1 ligand (Koizumi et al., 2005; Carland and Nelson, 2009). However, it was surprising to find that the major signal for bound AGD1-MBP was with PI-5P. Although PI-5P has been detected in plants, its biosynthesis remains unclear. So far, no PI-kinase that can produce PI-5P has been identified in plants and this PI monophosphate species might be formed by a yet to be determined phosphatase (Heilmann, 2009; Munnik and Nielsen, 2011). Thus, the biological significance of PI-5P binding to AGD1 is unclear. The Arabidopsis trithorax-like factor, ATX1, which is an epigenetic factor and chromatin modifier, was shown to bind PI-5P and is proposed to function as a receptor for this lipid messenger (Alvarez-Venegas et al., 2006). Additional studies are needed to further verify AGD1-PI-5P binding using liposome assays (Gagne and Clark, 2010) and demonstrate if PI binding is necessary for AGD1’s ARF-GAP activity. However, given the observation that fluorescently tagged- PI-5P decorated the plasma membrane of initiating root hairs (Alvarez-Venegas et al., 2006), it is possible that this PI species might function in polar tip growth. Using comparative genomics, PI-5P was proposed to function in exocytosis from late endosomal compartments to the plasma membrane (Lecompte et al., 2008). Our preliminary results show that AGD1-GFP decorates endosomal-like compartments in root hairs similar to that found in previous transient expression studies in leaf epidermal cells (Yoo et al., 2008) as well as the plasma membrane of initiating root hairs (our unpublished observations).

The intricate nature of AGD1’s position within the cellular machinery that drives directional growth in root hairs is further manifested by its impact on tip [Ca2+]cyt oscillations. However, it is unclear whether the actin cytoskeleton sets up the [Ca2+]cyt gradient or if [Ca2+]cyt triggers F-actin changes that drive tip growth. There is evidence in the literature that supports both scenarios. In pollen tubes for example, it was shown that actin disrupting drugs evoked an increase in [Ca2+]cyt which suggested that the actin cytoskeleton might regulate Ca2+ channel activity directly and therefore set-up the tip [Ca2+]cyt gradient (Wang et al., 2004). In contrast, more recent work also in pollen has shown that the rise in [Ca2+]cyt precedes actin depolymerization, favoring the hypothesis that [Ca2+]cyt regulates F-actin possibly by influencing the activity of actin regulatory proteins (Cardenas et al., 2008). How might dissipated [Ca2+]cyt oscillations explain the polarity defects of agd1 root hairs? Given that one of the most obvious cellular defects of agd1 root hairs is the bundling of the tip actin cytoskeleton (Yoo et al., 2008), one possible target of [Ca2+]cyt oscillations are proteins that modulate the higher order structure of actin such as VILLIN (VLN). Because actin filament severing activity of root hair-expressed VLN is increased by elevated Ca2+ (Khurana et al., 2010; Zhang et al., 2010, 2011), it is conceivable that dampening of Ca2+ influx could lead to abnormal formation of actin bundles observed in root hair tips of agd1 (Yoo et al., 2008).

We showed that the polar growth defects of agd1 root hairs can be caused by the redirection and dissipation of secretory vesicle delivery to the tip. RABA4b secretory vesicles were recently shown to carry complex cell wall polysaccharides (Kang et al., 2011). Thus, the skewed distribution of RABA4b could result in asymmetric deposition of cell wall precursors and lead to differential growth in agd1 root hairs. Consistent with the actin defects of agd1 root hairs, tip localization of EYFP–RABA4b was shown to dissipate when wild type root hairs were treated with the actin disrupting drug, latrunculin B (Preuss et al., 2004). Moreover, disturbing the tip [Ca2+]cyt gradient with A23187 abolished tip targeting of EYFP–RABA4b in Arabidopsis root hairs, and PI-4K, which interacts with RABA4b at the root hair tip, binds to calcineurin-like Ca2+ sensors (Preuss et al., 2006). Taken together, these results continue to implicate AGD1 in signaling cascades that link cytoskeletal reorganization and [Ca2+]cyt signaling to RABA4b and PI-mediated secretion in growing root hairs.

In contrast to root cells of agd3/van3 mutants (Naramoto et al., 2010), uptake of FM 1-43 was not inhibited in agd1 root hairs. It is worth noting that agd3/van3 knockouts have a strong overall plant growth defect due to altered vascular development (Koizumi et al., 2005; Sieburth et al., 2006). Therefore, tracer dye incorporation into agd3/van3 root cells might be more severely compromised compared to agd1. For our FM 1-43 uptake assays, we used agd1 root hairs that although clearly compromised in directional growth, were still elongating. These endocytic tracer studies reinforce our live cell imaging assays that suggested that AGD1 might have a more prominent role in stabilizing the growth machinery of root hairs to specific cellular domains for directed exocytosis rather than endocytic uptake of material from the plasma membrane.

In summary, our studies offer new insights into how AGD1 regulates root hair development and highlight advantages of using Arabidopsis root hairs to better understand how ACAP-type ARF-GAPs modulate polar cell growth. Extensive studies in other eukaryotic systems have provided evidence that ARF-GAPs regulate diverse cellular processes not only through the timely inactivation of ARF-GTPases but also as scaffolding effectors that coordinate the activities of various signaling cascades within the cell (Spang et al., 2010; Donaldson and Jackson, 2011). Our results indicate that AGD1 could indeed function as a signaling hub that links PIs with elements of the cytoskeleton to control directional membrane trafficking in the highly polarized root hair cell.

Experimental procedures

Generation of double mutants

Rhd4-1 (Thole et al., 2008), pip5k3-4 (SALK_026683; Stenzel et al., 2008) and act2-3 (SALK_048987; Nishimura et al., 2003) were used for generating double mutants. All single mutants were obtained from the Arabidopsis Biological Resource Center (ABRC). The progeny from crosses were examined and the selected lines were tested for the presence of the double mutation by PCR-based genotyping.

Map-based cloning of an agd1 enhancer

Seeds of agd1-1 were mutagenized with 0.2% EMS and plants were grown in 20 pools of M1 plants. About 20 000 M2 seed were plated in Petri dishes that contained the Arabidopsis medium described below; after 6 days, root hairs were examined using a stereomicroscope. Seedlings that exhibited enhanced root hair defects were isolated as putative agd1 enhancer mutants. One mutant (agd1 eag5) that displayed more than two root hairs originating from one initiation point was selected for further study.

For map-based cloning, eag5 single mutants were crossed to Landsberg and the resulting F2 seedlings were phenotyped. Publicly available simple sequence length polymorphism and cleaved amplified polymorphic sequence markers were used to map the eag5 loci (Lukowitz et al., 2000). When it was determined that eag5 had a mutation in the At4g34580 gene, a T-DNA knockout line (SALK_002124) was obtained from the ABRC. SALK_002124 was crossed with agd1-1, and the resulting F2 progeny was screened for reconstitution of the agd1 eag5 phenotype. For studying actin organization, eag5 was transformed with the GFP-fimbrin 1 actin binding domain2 (ABD2)-GFP reporter described previously (Wang et al., 2008). Methods for imaging and quantification of actin disruption in eag5 are described in Yoo et al. (2008).

Growth conditions and evaluation of root hair phenotypes

Seeds of agd1 and various double mutants were surface sterilized and planted on 48 × 64 mm coverslips layered with Murashige and Skoog (MS) salts as detailed in Dyachok et al. (2009). To evaluate root hair phenotypes, 4-day-old seedlings were examined with a stereo-microscope and photographed with a Nikon Insight camera (Nikon Corporation, Melville, NY, USA) or imaged with a Leica TCS SP2 AOBS confocal microscope (Leica Microsystems, Exton, PA, USA) equipped with a × 63 water-immersion objective. For the latter method, seedlings were stained with10 μm propidium iodide. Propidium iodide was excited with the 543-nm line of the Argon laser and emission detected at 617 nm. Root hair images were generated from a projection of more than 200 optical sections taken at 0.2 μm intervals.

Imaging [Ca2+]cyt oscillations, ROP2, RABA4b and PI-4P in living root hairs

The agd1-1 mutant was transformed with 35S:YC3.60 and [Ca2+]cyt was imaged using the FRET sensitized emission approach described in Rincon-Zachary et al. (2010). Root hairs of wild type and agd1-1 plants that expressed EYFP–ROP2 (Xu and Scheres, 2005), YFP-FAPP1, (Vermeer et al., 2009) and EYFP–RABA4b (Preuss et al., 2004) were imaged with an UltraView ERS spinning-disc confocal microscope (Perkin-Elmer Life and Analytical Sciences, Waltham, MA, USA) equipped with a ×63 water-immersion objective (Numerical aperture 1.40). EYFP was excited using the 488-nm line of the argon-krypton laser, and emission was detected at 510 nm. Root hairs were imaged by collecting optical sections at 1 μm intervals. Analyses of EYFP–ROP2, EYFP–RABA4b and EYFP–FAPP1 localization in growing root hairs were conducted on images of projected stacks of optical sections acquired every 30 sec over a period of 30 min to 1 h. Examples of single optical sections that were projected for image analysis of EYFP–FAPP in wild type and agd1 are shown in Figure S3.

Quantification of FM 1-43 uptake

Wild type and agd1 roots were pulsed labeled with 0.5 μm FM 1-43 as described (Ovecka et al., 2005). Mean fluorescence intensity of FM 1-43 signal at the apical 5 μm root hair plasma membrane and internalized dye (excluding the plasma membrane) was measured using image j ( as described in Naramoto et al., 2010. Root hairs that stopped elongating after FM 1-43 washout were excluded from the analysis.

AGD1 protein expression, purification and lipid binding assays

Complementary DNA (cDNA) that encoded the AGD1 protein (without the BAR domain) was amplified and ligated into the pMAL-c2X vector that contained the MBP tag (New England Biolabs, Ipswich, MA, USA). The plasmid was then transformed into Escherichia coli Rosetta pLYS strain for protein expression. After the AGD1 protein was eluted from beads, it was concentrated by Amicon Ultra Centrifugal Filter Units (Millipore, Billerica, MA, USA) and processed by sequential affinity chromatography-Resource S 1 ml column and High Performance MBP Trap 1 ml column (GE Health, Piscataway, NJ, USA).

AGD1-lipid binding assay was performed using hydrophobic membranes prespotted with lipids (Echelon Biosciences, Salt Lake City, UT, USA). After incubating membranes with AGD1-MBP or At1g68200-MBP controls, membranes were probed with anti-MBP antiserum at a dilution of 1:10 000 (New England Biolabs, Ipswich, MA, USA). Positive binding was detected using an Anti-Rabbit IgG horse radish peroxidase conjugate at a dilution of 1:5000 (Promega, Madison, WI, USA). The Lumi-Light PLUS Western Blotting kit (Roche, Indianapolis, IN, USA) was used to develop the membrane blot.


This work was supported by the Noble Foundation, the Oklahoma Center for the Advancement of Science and Technology (OCAST PSB-003) and the National Science Foundation Major Research Instrumentation Program (NSF DBI-0722635). We thank Drs Ben Scheres, Teun Munnik, and Erik Nielsen for sharing published material. The ABRC is acknowledged for providing seed stocks. We thank Dr Ping Xu for critical review of the manuscript.