Deletion analysis of AGD1 reveals domains crucial for plasma membrane recruitment and function in root hair polarity

Root hairs are single-cell projections that emerge from the epidermis of the main root body. These cylindrical cellular projections increase the total surface area of the root, thereby maximizing the uptake of soil resources needed to support the optimal growth and development of the plant (Brown et al., 2012, 2013; Salazar-Henao and Schmidt, 2016). Root hairs expand via tip growth, a process in which membranes and cell wall material are trafficked toward the extreme cell apex (Rounds and Bezanilla, 2013). A wealth of information on mechanisms underlying tip growth control in plants has come from studies on root hair development in the model plant Arabidopsis thaliana. For example, detailed microscopic analyses of epidermal cells of post-embryonic Arabidopsis roots enabled the identification of discrete growth phases during root hair development. The first phase (phase 1) is characterized by the appearance of a small bulge at the basal (root tip-directed) end of root hair-forming trichoblast cells. This phase is followed by slow expansion of the root hair bulge as it transitions to tip growth (phase 2). In the third phase, tip growth proceeds at a rapid rate leading to the tubular cells typical of root hairs (phase 3) (Dolan et al., 1994). A fourth phase involves the termination of tip growth, characterized by gradual shrinking of the cytoplasm and protrusion of the vacuole to the apical dome (phase 4) (Grierson et al., 2014). The location of root hairs on the root surface has also allowed for convenient forward genetic screens, leading to the isolation of mutants exhibiting a range of root hair polarity defects (Schiefelbein and Somerville, 1990; Parker et al., 2000; Grierson et al., 2001). Detailed studies of these mutants have led to the discovery of a number of proteins that orchestrate tip growth in plants. These include proteins that regulate cytoskeletal dynamics, endoplasmic reticulum morphology, cell wall formation, phosphoinositide (PI) metabolism, lipid modification, monomeric small guanine nucleotide binding protein (small GTPase) cycling, and reactive oxygen species (ROS) signaling (Favery et al., 2001; Seifert et al., 2002; Foreman et al., 2003; Böhme et al., 2004; Carol et al., 2005; Hemsley et al., 2005; Bernal et al., 2008; Thole et al., 2008; Yoo et al., 2008; Lai et al., 2014).

Through a root hair mutant screen in Arabidopsis, we identified the ADP-ribosylation factor-GTPase activating protein (ARF-GAP) called AGD1 (ARF-GAP domain 1) (Vernoud et al., 2003) as a regulator of root hair growth directionality, based on the observation that agd1 mutants exhibited root hairs with wavy/spiral tip growth instead of the straight growth typical of wild-type plants. Furthermore, agd1 mutants had a large percentage of root hairs with two tips originating from one initiation site, in contrast to the wild type, which had most of its root hairs with only one tip per initiation point (Yoo et al., 2008). Genetic and cell biological studies indicated that AGD1 functions in root hair development through signaling pathways that overlap with those that modulate cytoskeletal remodeling and PI metabolism (Yoo et al., 2012; Yoo and Blancaflor, 2013). However, the precise molecular function of AGD1 in specifying root hair polarity needs further investigation.

ARF-GAPs, the class of proteins to which AGD1 belongs, are best known for their role in regulating cycling of the ARF class of small GTPases. Small GTPases switch between an active and inactive configuration, a process that is essential for their signaling function. When bound to GTP, the small GTPase is active and associates with membranes, where it interacts with a range of effector proteins that mediate cytoskeletal organization and protein trafficking. In yeast, the process of GTP hydrolysis on ARF1 is required for coat disassembly and packaging of cargo during secretory membrane transport (D'Souza-Schorey and Chavrier, 2006). Whereas the switch to the activated, GTP-bound state is facilitated by guanine nucleotide exchange factors (GEFs), GAPs promote hydrolysis of GTP to GDP by converting the small GTPase into its inactive state. Although there is accumulating evidence showing that the GDP-bound forms of small GTPases can have important biological functions, GAPs are generally referred to as small GTPase inactivators (East and Kahn, 2011; Müller and Goody, 2017).

Based on sequence homology, AGD1 is most similar to the ARF-GAP with coiled-coil, ankyrin repeat and PH domain (ACAP) family of mammalian ARF-GAPs. Among the 15 ARF-GAPs in Arabidopsis, four are ACAPs or class 1 plant ARF-GAPs (Vernoud et al., 2003). In addition to the GAP catalytic domain, which contains a conserved zinc-binding motif, ACAPs have a Bin1-amphiphysin-Rvs167p/Rvs161p (BAR) and PH domain, and two to three ankyrin (ANK) repeats. The BAR and PH domains function in membrane curvature induction and PI binding, respectively, while the ANK domains enable protein-protein interactions (Vernoud et al., 2003; Gillingham and Munro, 2007; Inoue and Randazzo, 2007). There is accumulating evidence that the additional domains in ACAPs and other multi-domain ARF-GAPs allow them to combine their catalytic GAP activities with other effector scaffolding functions (Donaldson and Jackson, 2011; East and Kahn, 2011). For instance, ACAP2 has been shown to function both as an effector for RAB35 small GTPase and as an inactivator of the ARF6 small GTPase during neurite outgrowth (Kobayashi and Fukuda, 2012, 2013).

In addition to AGD1, another Arabidopsis ACAP-type ARF-GAP called AGD3 has been studied using cell biological and genetic approaches. AGD3, also known as VASCULAR NETWORK DEFECTIVE 3 (VAN3) and SCARFACE (SCF), functions in vascular development based on the observation that agd3 mutants have discontinuous leaf venation patterns (Koizumi et al., 2005; Sieburth et al., 2006). AGD3-GFP fusions were shown to localize to a subpopulation of the trans-Golgi network (TGN)/early endosomes (EE) and the PM, consistent with observations in mammalian ACAPs (Jackson et al., 2000; Koizumi et al., 2005; Naramoto et al., 2009, 2010). The localization of AGD3-GFP is also consistent with the reported endocytic defects of agd3 mutants (Naramoto et al., 2010). Unlike AGD3, however, the subcellular localization of AGD1 is not known. Such information is important because some in vitro studies in mammalian cells have shown that ARF-GAPs have very little differences in ARF substrate specificity. As such, the targeted localization of an ARF-GAP is likely a crucial factor in determining its ARF substrate in vivo and, therefore, in explaining a large part of its cellular functions (East and Kahn, 2011).

Here, we used the Arabidopsis root hair system to study the localization of a functional AGD1 tagged with GFP to obtain new insight into how an ACAP-type ARF-GAP modulates cell development in plants. We show that AGD1 is recruited to specific PM domains, and that spatiotemporal changes in its localization are correlated with the growth status of root hairs. Based on our results, we propose that differential recruitment of AGD1 to distinct root hair PM domains represents a mechanism to facilitate developmental transitions of root hairs from one phase to another. Furthermore, deletion analysis of AGD1-GFP using an inducible promoter system revealed that the PH domain by binding to specific pools of PIs is the minimal unit that directs AGD1 to the root hair PM, a process that is required for its full functionality.

When an AGD1-GFP construct under the control of its native promoter (AGD1pro:AGD1-GFP) was expressed in the agd1 mutant (Yoo et al., 2008), we found that the wavy root hair growth and two root hair tips from one initiation point defects of the mutant were rescued, indicating that the AGD1-GFP construct was functional (Fig. 1A–F). Consistent with the role of AGD1 in root hair development, the AGD1-GFP signal was enriched in root hair-forming trichoblasts (Fig. 1G). Close examination of AGD1-GFP uncovered three patterns of localization in roots. First, AGD1-GFP signal marked the ends of the cells in the primary root meristem and transition zone, a pattern indicative of PM localization. Second, diffuse AGD1-GFP signal within the cell interior also indicated cytoplasmic localization of the construct. Third, AGD1-GFP was associated with punctate bodies of varying sizes (Fig. 1G,H). In root hair cells, some bodies moved slowly while others moved rapidly throughout the cell (Movie 1).

The AGD1-GFP bodies observed in primary roots and root hairs were reminiscent of fluorescent foci typical of the endomembrane system. Furthermore, previous studies of other Arabidopsis ARF-GAPs showed that they localized to Golgi and post-Golgi compartments leading us to hypothesize that the AGD1-GFP puncta might also represent known endomembrane organelles (Jensen et al., 2000; Koizumi et al., 2005; Song et al., 2006; Min et al., 2007; Liljegren et al., 2009; Naramoto et al., 2009; Stefano et al., 2010).To test this hypothesis, we generated Arabidopsis lines expressing both AGD1-GFP and mCherry markers for the TGN/EE, late endosomes/prevacuolar compartment (LE/PVC) and Golgi (Nelson et al., 2007; Geldner et al., 2009), and asked if the AGD1-GFP bodies colocalized with any of the mCherry endomembrane foci. Surprisingly, confocal imaging of growing root hairs revealed that AGD1-GFP bodies did not overlap with compartments decorated by all of the endomembrane mCherry markers available to us (Fig. 2A–D). Imaging of dual-labeled root cells verified that dynamic green and red bodies corresponding to AGD1-GFP and endomembrane-mCherry fusions, respectively, did not overlap (Movies 2 and 3).

Further verification that AGD1-GFP foci were distinct from post-Golgi compartments was obtained by examining roots treated with Brefeldin A (BFA). Upon exposure to 100 µM BFA, TGN/EE-mCherry markers in the root hair formed fluorescent agglomerates, a behavior typical of post-Golgi organelles (Robinson et al., 2008). However, AGD1-GFP remained in discrete foci that did not overlap with the red-emitting BFA bodies (Fig. 2E). We also treated roots with wortmannin, a chemical that triggers homotypic fusion of the LE/PVC in plant cells. When seedlings were exposed to wortmannin, the PVC markers ARA6 (also known as RABF1)-mCherry and RABF2A-mCherry formed structures that were reminiscent of the fused PVCs reported previously (Wang et al., 2009, 2010). By contrast, AGD1-GFP remained in discrete foci that did not overlap with the red-emitting fused PVC bodies (Fig. 2F).

In addition to its localization to unknown foci and the cytoplasm, we found that AGD1-GFP marked the PM of developing root hairs. AGD1-GFP signal in root hair PM was first detected at a domain of the trichoblast where a small root hair bulge formed (Fig. 3A). This observation was in contrast to other proteins known to mark the site of root hair emergence such as the Rho of Plants (ROP) family of small GTPases, which have been reported to accumulate at the PM of the root hair initiation site prior to bulge formation (Molendijk et al., 2001; Jones et al., 2002). Indeed, no AGD1-GFP signal at the PM of root hair initiation sites could be observed in files of trichoblasts in the root transition and elongation zones (Fig. S1A,B). By contrast, in wild-type plants, a ROP2 (also known as ARAC4)-enhanced yellow fluorescent protein fusion (ROP2-EYFP) (Xu and Scheres, 2005) accumulated at the PM of the root hair initiation site prior to the appearance of a root hair bulge (Fig. S1C,D).

We next followed the time course of AGD1-GFP PM localization as root hairs transitioned through the various developmental phases. We found that AGD1-GFP continued to mark the apical-most PM domains of expanding root hair bulges, and as they transitioned to tip growth (Fig. 3B–E; phase 1–2) (Dolan et al., 1994). However, as root hairs proceeded to rapid tip growth (phase 3), AGD1-GFP localization to the apical-most PM domain became less intense and eventually dissipated. Distinct AGD1-GFP signal was confined to the PM of the root hair flanks while gradually weakening as root hairs continued to rapidly elongate (Fig. 3F–H; phase 3). As the root hairs matured and tip growth ceased, intense AGD1-GFP fluorescence returned to the apical-most PM domain (Fig. 3I; phase 4) (Grierson et al., 2014). Finally, AGD1-GFP signal at all PM domains of the root hair dissipated as the cytoplasm dispersed and the vacuole protruded toward the tip (Fig. 3J).

We quantified changes in AGD1-GFP PM localization by obtaining the ratio of PM tip fluorescence to PM fluorescence along the lateral flanks of root hairs. In the case of root hair bulges, PM fluorescence along the more apical (shoot-directed) PM was obtained following the method of Stanislas et al. (2014) (Fig. 3K, inset). Plotting root hair growth rate and the AGD1-GFP tip/flank fluorescence ratio as a function of time in a double Y graph revealed that a higher tip/flank ratio was associated with periods of slower root hair growth (Fig. 3K). Slower root hair growth occurred during the onset of bulge formation and as the root hair transitioned to tip growth (phase 1 and 2) (Dolan et al., 1994), and then again as the root hairs matured (phase 4) (Grierson et al., 2014). During periods of rapid tip growth (phase 3), the tip to flank AGD1-GFP PM fluorescence ratio was significantly lower (Fig. 3K). A scatter plot revealed a strong inverse relationship between root hair growth rate and AGD1-GFP PM fluorescence with a Pearson's correlation coefficient of −0.615 (Fig. 3L). Our results support that AGD1 location at distinct PM domains is correlated with root hair growth status.

Whereas AGD1-GFP localization at the apical-most PM domain was associated with slower root hair growth rate during phase 1, 2 and 4, signal at the PM tip dissipated and was confined to the lateral PM in rapidly elongating root hairs at phase 3 (Fig. 3). We therefore hypothesized that when AGD1 is localized to the apical-most PM domain of the root hair tip, it functions in fine-tuning tip growth, most likely as a negative regulator. To test this hypothesis, we prematurely terminated the growth of root hairs at phase 3 by subjecting seedlings expressing AGD1-GFP to a hyperosmotic shock using mannitol. Prior to mannitol application, young (∼20–30 µm in length) and old (>50 µm in length) root hairs of lines expressing the AGD1-GFP fusion at phase 3 showed the typical flank PM localization and lack of GFP signal at the apical-most PM domain (Fig. 4A). Liquid growth medium supplemented with 500 mM mannitol was then injected into the semi-solid gel that supported the seedling roots, and roots were immediately imaged. Within 1 min of mannitol application, we found intense AGD1-GFP signal at the apical-most PM domain of root hairs at phase 3 that coincided with shrinkage of the tip cytoplasm and stoppage of tip growth (Fig. 4A). We observed that many root hairs at phase 3 were able to resume tip growth within 5–10 min after exposure to mannitol. Resumption of tip growth was accompanied by dissipation of AGD1-GFP fluorescence at the apical-most PM domain and relocalization of the fusion protein to the lateral PM, an observation reminiscent of changes in the localization of the fusion protein as root hairs transitioned from slow tip growth in phase 2 to rapid tip growth in phase 3 (Fig. 4B,C; compare to Fig. 3E–H).

The correlation between AGD1-GFP localization to the apical-most root hair PM apex and root hair growth status was also observed in other treatments that inhibited tip growth. For example, like mannitol, treatment with sodium chloride (NaCl) led to relocalization of AGD1-GFP to the apical-most PM domain of phase 3 root hairs within 1 min. However, unlike in mannitol, phase 3 root hairs exposed to NaCl were not able to quickly resume tip growth. As such, localization to the apical-most PM domain persisted in these nongrowing root hairs (Fig. 4D). Taken together, our results support that recruitment of AGD1 to the apical-most PM domain of root hair is associated with root hairs that either grow slowly or terminate tip growth due to an external stimulus.

Previous characterization of the agd1 mutant focused on the wavy growth patterns of elongating root hairs at phase 3 (Yoo et al., 2008, 2012). Although the increased frequency of two root hairs emerging from one initiation point in agd1 mutants is consistent with the localization of AGD1 at the PM of root hair bulges observed here (Fig. 1B and Fig. 3A–C) (Yoo et al., 2008, 2012), it is possible that other phenotypes related to root hair initiation might have been overlooked. We therefore asked whether agd1 exhibited any planar polarity defects, given that AGD1-GFP localized at the PM during early root hair bulge formation (Fig. 3). In root hair development, planar polarity refers to the position of the root hair initiation site, which is typically directed toward the basal end of the trichoblast (Fischer et al., 2006; Stanislas et al., 2015). We found that there was a slight, but statistically significant, apical (shoot-directed) shift in root hair position in agd1 mutants compared to wild type (Fig. 5A,B). Like other root hair phenotypes, this mild apical shift in planar polarity was reversed in agd1 lines expressing the AGD1pro:AGD1-GFP construct (Fig. 5B). Our results indicate that AGD1 has a subtle role in defining the position of root hair initiation sites.

The plant class 1 ARF-GAPs are the most complex as they have three other domains in addition to the catalytic GAP domain (Vernoud et al., 2003). However, the contribution of these domains to AGD1 function and subcellular localization has never been addressed. We therefore generated a series of AGD1-GFP constructs with single protein domains deleted. The constructs were placed under the control of an estradiol-inducible promoter (pER8) (Zuo et al., 2000) and expressed in the agd1 mutant background to simultaneously test functionality of the domain deletion constructs and impact of the deleted domains on AGD1 localization. For unknown reasons, we were unable to isolate agd1 lines expressing AGD1-GFP fusions with only the BAR domain deleted (pER8:AGD1ΔBAR-GFP). The constructs we were able to analyze included PH (pER8:AGD1ΔPH-GFP), GAP (pER8:AGD1ΔGAP-GFP) and ANK (pER8:AGD1ΔANK-GFP) domain deletions as well as full-length AGD1 (pER8:AGD1-GFP) (Fig. 6A). Non-induced agd1 mutants expressing pER8:AGD1-GFP and all deletion domain constructs showed the root hair polarity defects observed in untransformed agd1, including the increased frequency of two root hairs originating from a single initiation point (Fig. 6B; Fig. S2A). By contrast, and consistent with results obtained with the AGD1pro:AGD1-GFP-complemented agd1 lines (see Fig. 1A–F), induction with estradiol rescued the root hair defects of agd1 expressing the pER8:AGD1-GFP construct, indicating functionality of the full-length AGD1 fusion under the pER8 inducible system (Fig. 6B; Fig. S2B). However, root hair defects of agd1-expressing pER8:AGD1ΔPH-GFP, pER8:AGD1ΔANK-GFP and pER8:AGD1ΔGAP-GFP were not rescued by estradiol application (Fig. 6B; Fig. S2C–E).

Upon examination of GFP signals in root hairs of the estradiol-induced seedlings, we found that pER8:AGD1-GFP lines showed AGD1-GFP patterns at the root hair PM that mirrored the spatiotemporal changes observed in the AGD1pro:AGD1-GFP-complemented agd1 lines. Like AGD1pro:AGD1-GFP lines, estradiol-induced pER8:AGD1-GFP lines showed fluorescent signal that marked the PM of bulging root hairs, the lateral PM of root hairs undergoing rapid tip growth, and the PM at the apical-most PM domain of root hairs that ceased growth (Fig. 6C). AGD1-GFP signal resulting from inducible expression of the pER8 promoter was generally weaker compared to the AGD1pro:AGD1-GFP lines. Furthermore, the unidentified bodies that were typically observed in AGD1pro:AGD1-GFP-expressing lines were not as numerous as in estradiol-induced pER8:AGD1-GFP lines (Fig. 6C).

We next examined the localization of the three AGD1-GFP domain-deleted constructs after estradiol induction. We found that AGD1ΔPH-GFP no longer marked the PM and was exclusively in the cytoplasm during all phases of root hair development (Fig. 6D). By contrast, AGD1ΔGAP-GFP and AGD1ΔANK-GFP still marked the PM of bulging root hairs and the PM along the flanks of root hairs that proceeded to tip growth, a pattern that roughly mirrored that of full-length AGD1-GFP (Fig. 6E,F). Compared to full-length AGD1-GFP and AGD1ΔGAP-GFP, the signal of AGD1ΔANK-GFP was generally weaker. The AGD1ΔGAP-GFP and AGD1ΔANK-GFP constructs did not complement the agd1 root hair shape defects, and their localization along the PM flanks of phase 3 root hairs was often asymmetric (Fig. 6E,F). The AGD1ΔPH-GFP and AGD1ΔANK-GFP constructs did not label the unidentified bodies decorated by full-length AGD1-GFP. Although the AGD1ΔGAP-GFP construct also labeled numerous fluorescent aggregates, these foci were smaller and more heterogeneous in size compared to those marked by full-length AGD1-GFP.

The localization of the AGD1-GFP deletion constructs indicated that the PH domain is required for PM localization. However, because we were unable to isolate plant lines expressing the inducible AGD1ΔBAR-GFP fusion, we could not conclude if the BAR domain contributed to the recruitment of AGD1 to the root hair PM. To address this question, we generated a GFP construct to the AGD1 PH domain alone (pER8:PH-GFP) and expressed it in the agd1 mutant. Upon induction with estradiol, we found that PH-GFP had diffuse signal in the root hair cytoplasm that was more intense than the cytosolic signal observed in full-length AGD1 and the other AGD1-deletion constructs. Despite the strong cytoplasmic signal, PH-GFP could be detected in the PM as the root hair progressed through the different developmental phases (Fig. 7A–F). However, unlike full-length AGD1-GFP or the GAP and ANK AGD1-GFP deletion constructs, PH-GFP was often observed to maintain PM localization at the apical-most PM domain of root hairs (Fig. 7E,F). Like the GAP and ANK deletion constructs, PH-GFP did not complement the agd1 root hair defects. Localization of PH-GFP was also often observed to be unevenly localized to the PM along the lateral flanks as root hairs developed (e.g. Fig. 7D). PH-GFP tip/flank fluorescence was quantified in a similar manner to that for AGD1-GFP (see Fig. 3K). Like full-length AGD1-GFP, there was an inverse relationship between PH-GFP PM localization and root hair growth rate (Fig. 7G). However, a scatter plot and Pearson's correlation analysis indicated a weaker inverse relationship between PH-GFP PM localization and root hair growth rate (Pearson's correlation coefficient=−0.284) when compared with full-length AGD1-GFP (Fig. 7H).

The PH domain is known to bind PIs. In doing so, proteins that have them are recruited to specific cell membrane regions by binding to PIs that are enriched in that particular membrane (Lemmon et al., 1996). We therefore hypothesized that AGD1, through its PH domain, could be targeted to the root hair PM by binding to specific pools of PIs that are enriched in root hair cells. To test this hypothesis, we first asked whether the AGD1 PH domain alone is capable of directly binding PIs. We cloned the AGD1 PH domain and generated a glutathione-S-transferase (GST)-AGD1-PH domain construct (GST-PH). Purified GST-PH, or GST alone, was used in lipid overlay assays on nitrocellulose strips blotted with a range of phospholipids. We found that GST-PH, but not the GST only controls, bound to PI monophosphates, namely phosphatidylinositol 3-phosphate [PI(3)P], phosphatidylinositol 4-phosphate [PI(4)P] and phosphatidylinositol 5-phosphate [PI(5)P] (Fig. 7I).

Direct binding of the AGD1 PH domain to PI monophosphates prompted us to next ask if the localization of specific PIs in root hair cells mirrored that of PH-GFP. To address this question, we followed the localization of in vivo PI(3)P (YFP-2XFYVE) and PI(4)P (YFP-FAPP1) markers (Vermeer et al., 2009) during root hair development. We found that YFP-2XFYVE localized to endosomes and the tonoplast, similar to a previous report (Vermeer et al., 2006), but did not label the PM in phase 1 and phase 2 root hairs (Fig. 7J). By contrast, YFP-FAPP1 marked the PM of root hairs in a similar manner to PH-GFP. Like PH-GFP and full-length AGD1-GFP, YFP-FAPP1 fluorescence was not enriched at the PM of the future root hair initiation site (Fig. 7K). A clear enrichment of YFP-FAPP1 PM fluorescence only became obvious when there was a visible root hair bulge (Fig. 7L). The PI(4)P biosensor continued to mark the PM as the root hair bulge enlarged and proceeded to broadly label the tip and flank PM as the root hairs proceeded to phase 3 (Fig. 7M). As root hairs rapidly elongated, YFP-FAPP1 signal became strongest along the apical 10 µm of the tip, similar to previous reports (Fig. 7N) (Vermeer et al., 2009; Yoo et al., 2012). The enrichment of YFP-FAPP1 along a broad region of the PM apical dome of phase 3 root hairs resembled that of PH-GFP localization (see Fig. 7E,F). Like PH-GFP, YFP-FAPP1 PM localization and root hair growth rate had an inverse relationship (Fig. 7O,P). The similar patterns of PH-GFP and YFP-FAPP1 enrichment during root hair development suggest that enriched pools of PI(4)P recruits AGD1 to the PM. Furthermore, our results on the localization of AGD1ΔPH-GFP (Fig. 6D) and PH-GFP (Fig. 7A-F) support the conclusion that the PH domain is the minimal unit needed for targeting AGD1 to the root hair PM.

We used the root hair system of Arabidopsis to gain a deeper understanding of the role of an ACAP-type ARF-GAP in specifying plant cell polarity. Plant root hairs are tubular extensions of root epidermal cells that develop through a series of distinct phases (Dolan et al., 1994; Grierson et al., 2014). Like pollen tubes, fungal hyphae and neurons, root hairs expand via tip growth, leading to a cellular geometry with an extreme aspect ratio in which the length of the cell is several times its width. For tip-growing plant cells to maintain this extreme aspect ratio, material for constructing the cell wall has to be transported over long distances and targeted to the apical-most PM of the cell (Chebli et al., 2013). The convenience of the root hair system for live cell microscopic and genetic studies has led to the identification of several key proteins that are part of the tip-growth machinery in plants. In many cases, localization of these proteins to the root hair tips using GFP fusions has validated their importance for root hair development (Kusano et al., 2008; Stenzel et al., 2008; Thole et al., 2008; Park et al., 2011; Park and Nebenführ, 2013). Here, we provide evidence that the ARF-GAP AGD1 is another root hair tip-localized protein. Based on its localization patterns at the root hair apex, we propose that AGD1 is an important component of the secretory apparatus that defines polarized root hair growth through its timely depletion or displacement from, and recruitment to, specific PM domains.

Although several studies have demonstrated tip-directed protein localization in Arabidopsis root hairs using fluorescent protein fusions, not many have followed details of their spatiotemporal changes during root hair development, or in response to treatments that suppress tip growth. In studies in which such experiments were conducted, tip-directed functional YFP-cellulose synthase-like 3 (YFP-CSLD3) and YFP-myosin-17 (YFP-XIK) were monitored in root hairs treated with the actin disrupting drug latrunculin B (LatB) (Park et al., 2011; Park and Nebenführ, 2013). A rapid loss in tip-directed YFP-CSLD3 and YFP-XIK localization was observed within minutes of LatB treatment that coincided with a decline in root hair growth (Park et al., 2011; Park and Nebenführ, 2013). In the case of YFP-CSLD3, its accumulation at the tip and resumption of root hair growth was restored upon LatB washout (Park et al., 2011). The spatiotemporal patterns of CSLD3 and XIK localization during root hair development are examples of proteins that positively regulate tip-focused expansion. Their role as positive regulators of tip growth is consistent with xik mutants having short root hairs, while those of csld3 rupture at the tip soon after initiation (Wang et al., 2001; Bernal et al., 2008; Kusano et al., 2008; Stenzel et al., 2008; Park and Nebenführ, 2013). By contrast, our data indicate that AGD1 partly functions to restrict growth since its depletion at the apical-most PM domain coincides with rapid tip growth. Evidence to this effect can be seen by enhanced AGD1-GFP signal at the apical-most PM domain when root hair growth is at its slowest (i.e. during phase 1, 2 and 4), or when tip growth of rapidly growing phase 3 root hairs are prematurely terminated by hyperosmotic shock. However, the slower root hair growth rate of agd1 mutants does not fully support the hypothesis that localization of AGD1 to the apical-most region of the root hair limits growth (Yoo et al., 2008). An alternative explanation is that the onset of rapid tip growth results in AGD1 being displaced from the apical-most PM domain of the root hair leading to its accumulation along the lateral PM flanks. The data presented here do not allow us to distinguish between these possibilities and further studies will be needed to better understand how AGD1 dynamics translate to root hair growth status.

To acquire their highly cylindrical shapes and sustain the mostly symmetrical growth typical of tip-growing plant cells, pollen tubes and root hairs direct the nonvectorial internal turgor pressure toward the apical region of the cell through gradients in the stiffness and extensibility of their cell walls. The establishment of these gradients in cell wall stiffness is dictated by endocytic and exocytic trafficking of vesicles to the apical region (Chebli et al., 2013). In pollen tubes, microindentation studies and mathematical modeling have shown that apical cell wall stiffness at the tip is less than that of the cylindrical shank allowing turgor-driven expansion of the cell to be confined to the apical-most dome (Zerzour et al., 2009; Fayant et al., 2010). Given their similar growth habits, it is likely that gradients in cell wall biomechanical properties reported for pollen tubes also occur in root hairs. In this regard, depletion of AGD1 at the tip PM as root hairs transition from slow growth in phase 2 to rapid growth in phase 3 suggest that it participates in secretory processes that maintain cell wall softness at the tip while reinforcing cell wall stiffness along the subapical cylindrical shanks (Chebli et al., 2013). Without functional AGD1, it is reasonable to assume that vesicular trafficking at the root hair tip would be altered. A consequence of such a defect in tip-directed trafficking is the development of lateral gradients in cell wall stiffness across opposite flanks of the root hair cell, which in turn would lead to breaking of axial symmetry (Chebli et al., 2013). Broken axial symmetry is evident from the fact that root hairs of agd1 mutants at phase 3 exhibit wavy/spiral growth and occasional swollen tips, a scenario that can also be explained, in part, by disrupted cytoskeletal organization (Yoo et al., 2008, 2012). Furthermore, our data indicate that enrichment of AGD1 at the apical-most PM domain of root hairs at phase1 and 2 serves to maintain growth at a slow enough rate to ensure stability of the trafficking apparatus as root hairs emerge. The mild apical shift in planar polarity and the formation of two root hair tips from one initiation point in agd1 mutants support this hypothesis.

The precise mechanism by which AGD1 translates processes occurring at the root hair PM to the establishment of cell wall biomechanical gradients is not known. However, given that AGD1 is an ARF-GAP, a root hair tip-localized ARF-GTPase is likely involved. Although the ARF substrate for AGD1 has yet to be identified, a study showing that overexpression of constitutively active ARF1 was associated with root hair polarity defects in Arabidopsis support this possibility (Xu and Scheres, 2005). Furthermore, the observation that BFA, an inhibitor of ARF-GEFs, is capable of reversing the wavy root hair phenotypes of the agd1 mutant indicates that an ARF-GTPase molecular switching mechanism operates in maintaining normal root hair axial symmetry (Yoo et al., 2008). In addition to ARFs and AGD1, potential components of the ARF-GTPase machinery that drive root hair development are the ARF-GEFs GNOM (also known as GN) and BEN1/MIN7 (Nomura et al., 2006; Tanaka et al., 2009; Naramoto et al., 2010, 2014). Whereas ben1/min7 mutants were recently shown to exhibit root hair shape defects (Sparks et al., 2016), GNOM and GNOM-like (GNL) ARF-GEFs are required for root hair planar polarity (Fischer et al., 2006). Moreover, GNOM was shown to be involved in root hair tip growth and GNOM-GFP localizes to trichoblasts during early bulge formation, where it is a direct BFA target (Richter et al., 2011; Stanislas et al., 2014).

Our study presents evidence for a critical role of the PH domain in directing AGD1 to the root hair PM based on results showing that an AGD1-GFP fusion with the PH domain deleted (AGD1ΔPH-GFP) was cytosolic, while a GFP fusion to only the PH domain (PH-GFP) still localized to the PM. The importance of the PH domain for TGN/EE localization was also reported for the AGD3 protein (Naramoto et al., 2009). Our results show that the localization of AGD1 to the root hair PM is required for its full functionality as estradiol induction of AGD1ΔPH-GFP expression failed to complement the agd1 root hair phenotypes. However, GAP and ANK domain deletions, despite also maintaining their association with the root hair PM, did not complement the agd1 root hair phenotypes. In mammalian cells, the ANK and PH domains are required for efficient catalytic activity of multi-domain ARF-GAPs (Ha et al., 2008). The PH domain in particular binds to phosphatidylinositol 4,5-bisphophate [PI(4,5)P2], leading to more efficient ARF-GAP catalytic activity (Ha et al., 2008). Thus, it is possible that, like the GAP domain-deleted fusion, the ANK and PH domain deletions have reduced or no longer possess functional GAP activity. If this is the case, it appears that both the recruitment of AGD1 to the PM and its GAP activity are essential for normal root hair development. We also show that the AGD1 PH domain alone is sufficient for PI monophosphate binding. Like full-length AGD1, the PH domain alone binds to PI(3)P, PI(4)P and PI(5)P (Yoo et al., 2012). Because spatiotemporal patterns of PH-GFP localization to the root hair PM mirrored those of a PI(4)P biosensor (Vermeer et al., 2009), we propose that this PI species is the major phospholipid that recruits AGD1 to the PM. It is worth noting that unlike full-length AGD1-GFP, the PH-GFP fusion persisted at the apical-most PM domain of many phase 3 root hairs. Although one could argue that the persistence of PH-GFP at the root hair PM apex is a consequence of the altered root hair shapes and slower tip growth due to the inability of the construct to complement agd1, one cannot discount the possibility that the ANK and GAP domains could be involved in modulating in planta binding of the PH domain to specific pools of PI(4)P. This level of interdomain regulation could represent a mechanism for dissociating AGD1 from the apical-most PM domain as root hairs transition to rapid phase 3 tip growth.

The foci decorated by AGD1-GFP did not colocalize with any of the endomembrane markers tested. This result was surprising given that AGD3, a closely related class 1 ARF-GAP, closely localizes to the TGN/EE (Koizumi et al., 2005; Naramoto et al., 2009, 2010). The fact that AGD1-GFP foci are rare or absent in the inducible pER8:AGD1-GFP-expressing lines, but yet are able to complement the agd1 root hair defects, suggests that AGD1 at the PM, rather than those at the foci, are crucial for maintaining polar root hair growth. Although our observations imply that the AGD1-GFP bodies are possibly mislocalization artifacts, we cannot discount the possibility that they represent unknown organelles. Additional tools, including colocalizing AGD1-GFP with other red-emitting endomembrane markers not represented in our study and immunoelectron microscopy of high pressure-frozen root samples, are needed to determine if these AGD1-GFP foci reflect the authentic localization of native AGD1.

In summary, our work sheds new light on the role of an ACAP-type ARF-GAP in plants and highlights the advantages of the root hair system as a model for addressing basic cell biological questions. Based on our results, we propose that differential recruitment of AGD1 to specific PM domains during root hair development is a process that is pivotal for its in vivo function. When associated with the PM, AGD1 restricts growth in bulging root hairs and those that are at an early stage of development. Maintaining tip growth at this slower, restricted pace stabilizes the trafficking apparatus to ensure normal root hair emergence. By dissipating from the tip PM, this AGD1-imposed growth restriction is released, allowing root hairs to grow at a more rapid rate while maintaining symmetrical growth. Finally, our study shows that tight control of spatiotemporal patterns of AGD1 localization is facilitated, in part, through the association of its PH domain with pools of tip-enriched PI(4)P.

The AGD1 promoter was amplified using the primers AGD1_pro F and AGD1_proR. The polymerase chain reaction (PCR) product was cloned to the pDONRP4-P1R vector (Invitrogen) by the Gateway BP reaction. For the AGD1 open reading frame (ORF), the primers AGD1F and AGD1R were used to amplify the AGD1 gene and clone it to the pENTR D-Topo vector (Invitrogen). Arabidopsis genomic DNA isolated from 10-day-old seedlings was used as a template for the PCR reaction. The AGD1pro:AGD1-GFP fusion construct was generated by incorporating the AGD1 promoter and AGD1 ORF entry clones into the R4pGWB450 vector through the LR reaction (Nakagawa et al., 2008).

AGD1-GFP deletion constructs were generated using the estradiol-inducible expression vector pER8 (Zuo et al., 2000). Complementary DNA (cDNA) of AGD1 was amplified from 10-day-old seedlings by PCR using primers AGD1-XhoI-F/AGD1-SpeI-R and cloned into pER8. The g3GFP DNA fragment was amplified using AGD1pro:AGD1-GFP as a template and fused to the SpeI site at the C-terminus of AGD1 to generate pER8:AGD1-GFP. To generate the AGD1 deletion constructs, a similar strategy was used. In the case of pER8:AGD1ΔPH-GFP, for example, the N-terminal BAR domain and the C-terminal GAP-ANK domains of AGD1 were amplified using primers AGD1-XhoI-F/268-EcoRI-R and 457-EcoRI-F/AGD1-SpeI-R, respectively. The two DNA fragments were sequentially cloned into the pBluescript II vector. The resulting DNA was then cloned into the XhoI/SpeI sites of the pER8 vector, and the g3GFP fragment was fused into the SpeI site at the C-terminus. Primers used to generate the constructs are shown in Table S1.

The agd1-1 mutant (Yoo et al., 2008) was transformed with AGD1_pro:AGD1-GFP and inducible pER8 constructs using the floral dip method (Clough and Bent, 1998). Antibiotic-resistant seedlings with GFP signal were selected using an SZX12 stereofluorescence microscope (Olympus America).

Arabidopsis seeds were sterilized with 95% (v/v) ethanol and 20% (v/v) bleach. Seeds were planted on 0.5% phyta-agar supplemented with half strength Murashige and Skoog (MS) salts, 1% sucrose, and the vitamins pyridoxine-HCl (0.5 mg/ml), nicotinic acid (0.5 mg/ml) and thiamine (1 mg/ml), layered onto 48×64 mm cover slips as detailed in Dyachok et al. (2016). Coverslips were placed in 90 mm Petri dishes and positioned at a 20° angle from the vertical, under a 14 h-light–10 h-dark cycle at 22°C to allow primary roots to grow along the surface of the coverslips. Five days after planting, coverslips containing the seedlings were transferred to the stage of an UltraView ERS spinning disc confocal microscope (Perkin-Elmer Life and Analytical Sciences) or an SP8 X point scanning confocal microscope (Leica Microsystems) equipped with 40× or 63× water-immersion objectives. GFP was excited using the 488-nm line of the argon laser, and emission was detected at 510 nm. For lines expressing the pER8-inducible constructs, MS medium containing 10–20 µM estradiol was applied directly to seedling roots using a 10 ml syringe. Seedlings were observed 24 h after induction.

Time-lapse sequences and single time point images were captured using the spinning disc confocal microscope running Improvision Volocity acquisition software version 5.4.2 (Perkin Elmer) or the SP8 X running LAS AF software (Leica Microsystems). Root hair images were obtained at a fixed focal plane that spanned the median of the cell. For time-lapse microscopy, images were captured every 1 s for 1 min. To analyze expression patterns in primary roots of AGD1pro: AGD1-GFP lines, a series of optical sections were taken at 0.5 µm intervals and merged using the isosurface rendering function of Volocity 5.4.2 visualization software (Perkin Elmer).

Root hair growth rates from lines expressing AGD1-GFP, PH-GFP and YFP-FAPP1 were obtained by capturing single optical confocal images from trichoblasts every 10 min for 80 min. Images corresponding to the first time points were captured from root hair bulges that were 2–5 µm in length. Root hair growth rate expressed as µm/min was calculated by first measuring the displacement of the root hair tip overtime for every 10 min time point and dividing this number by 10. AGD1-GFP, PH-GFP and YFP-FAPP1 fluorescence was quantified from the same root hairs in which growth rate was measured. A 2 pixel line ∼10 µm in length was drawn along the root hair cell edges that corresponded to the PM of the tip and flanks (see inset in Fig. 3K), and average fluorescence intensity was acquired using Image J (https://imagej.nih.gov/ij/). The ratio of average tip and flank fluorescence intensity was obtained, and presented as relative fluorescence.

AGD1pro: AGD1-GFP lines were crossed with plants expressing the WAVE-mCherry collection of membrane markers (Geldner et al., 2009) or directly transformed with mCherry-G-rk Golgi (Nelson et al., 2007) and ARA6-mCherry constructs. The mCherry-Golgi and WAVE lines were obtained from the Arabidopsis Biological Research Center (Columbus, OH). The ARA6-mCherry construct was made by subcloning into a pCAMBIA1390 vector that had the cauliflower mosaic virus (35S) promoter. Full-length ARA6 was amplified from genomic DNA from 10-day-old seedlings using the primers ARA6-SalI-F and ARA6-BamHI-R (Table S1). Plants emitting both green and red fluorescence were selected under a stereo fluorescence microscope and propagated for seed. Seedlings were imaged using the sequential scanning mode of the SP8 X confocal microscope.

The PH domain, which spanned 867–1281 base pairs of the AGD1 gene, was PCR amplified from cDNA obtained from 10-day-old seedlings with gene-specific primers AGD1-PH-BamH1-F and AGD1-PH-Sal1-R (Table S1). The PCR product was cloned into the pGEX5X-1 GST vector at the BamH1 and Sal1 sites. For protein expression, the construct was transformed into BL21 cells and grown in 5 ml Luria-Bertani (LB) broth (Thermo Fisher Scientific) overnight at 37°C. The culture was induced with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) (Thermo Fisher Scientific) and incubated for an additional 3 h. Cells were then lysed and proteins purified by incubation in GST resin (Qiagen). Proteins were resolved on a polyacrylamide gel and detected by Coomassie staining. For the lipid overlay assays, ∼2.5 µg GST or GST-PH proteins were incubated with the PI membranes (Echelon Biosciences) at 4°C overnight, and bound proteins were detected with anti-GST HRP-conjugated antibody (Abcam).

All values are expressed as mean±standard error of the mean (s.e.m.). Measurements were obtained from at least four independent experiments with 10 seedlings per experiment. One-way analysis of variance (ANOVA) followed by post hoc Tukey's Honestly Significant Difference (HSD) test (SPSS version 23) were used to determine statistical differences among means. For determining differences in root hair planar polarity, the nonparametric Kolmogorov–Smirnov test was used to test the null hypothesis that distribution of relative root hair position between two genotypes (e.g. wild type versus agd1) are similar. The null hypothesis was rejected if P<0.01. A total of 178 cells from ∼30 seedlings were measured. To determine the relationship between root hair growth rate and AGD1-GFP, PH-GFP or YFP-FAPP1 fluorescence, Pearson's correlation analysis was used. The inverse relationship between root hair growth rate and tip/flank fluorescence was determined to be statistically significant if P≠0 based on 120 pairwise measurements.

We thank Soonil Kwon for assistance with the statistical analysis.