Glycosylphosphatidylinositol-anchored proteins as chaperones and co-receptors for FERONIA receptor kinase signaling in Arabidopsis

1The Arabidopsis FERONIA (FER) receptor kinase critically controls growth and development, is indispensable for reproduction, and participates in defense-related responses. FER was initially identified as an essential regulator for female fertility; its expression in the female gametophyte is responsible for inducing rupture of an invading pollen tube to release sperm for fertilization. It is also required to prevent supernumerary pollen tube entrance to individual ovules, precluding polyspermy and maximizing seed yield. Thus fer mutant plants are severely female-deficient, producing few seeds. FER is, however, broadly expressed and absent only in pollen; its functions intersect several major plant hormone signaling pathways, including auxin, abscisic acid (ABA), brassinosteroid, and ethylene. FER has also been shown to interact with the peptide hormone rapid alkalinization factor 1 (RALF1). Therefore fer knock-out mutants are pleiotropic, with vegetative phenotypes attributable to defects in growth processes regulated by these hormones. FER also mediates susceptibility to the fungal pathogen powdery mildew and has been implicated in mechano-sensing. Thus it is likely that FER mediates distinct signals under different cellular and developmental conditions and environmental challenges.

2How FER achieves its multiple functionality remains unclear. We showed earlier that FER interacts with ROPGEFs, guanine nucleotide exchange factors that stimulate GDP/GTP exchange in RAC/ROPs, the RHO GTPases of plants, activating them. We demonstrated that FER acts as a cell surface regulator for RAC/ROP-mediated NADPH oxidase-dependent reactive oxygen species (ROS) production to support polarized root hair growth in seedlings, and induce Ca2+-dependent pollen tube rupture and sperm release in the female gametophyte. RAC/ROPs are known to mediate multiple signaling pathways that underlie normal plant growth and development, as well as stress-related responses. ROS are ubiquitous and regulate a broad spectrum of cellular processes as diverse as cell growth and cell death. Utilizing RAC/ROPs and ROS as signal mediators potentially provides almost limitless permutations of how FER-mediated signals might be propagated.

3The diverse functionality of FER could also be provided by its potential ability to interact with multiple ligands. Its extracellular domain shows homology with malectin, a disaccharide-binding protein located in the endoplasmic reticulum (ER) of animal cells. That FER might interact with carbohydrate moieties suggests the potential of mediating cell wall perturbations elicited by a battery of endogenous and environmental conditions such as hormonal changes impacting cell growth and pathogen attacks eliciting cell wall restructuring. RALF1 is one of ∼40 related secreted peptides in Arabidopsis that collectively are ubiquitously present, albeit individually they are all expressed at low levels and their functional roles in plant growth and development remain largely unexplored. If, similar to RALF1, more of these peptide hormones interact with FER, using individual RALFs as signals might be another strategy to achieve its multi-functional roles.

4Glycosylphosphatidylinositol-anchored proteins (GPI-APs) are cell surface-located proteins known to play important roles in regulating a broad range of biological processes including growth, morphogenesis, reproduction, and disease pathogenesis in eukaryotes. They localize to sphingolipid- and cholesterol-enriched domains in the cell membrane where they are believed to play key roles in regulating cell surface signaling dynamics, although much remains to be learned about their precise functional mechanisms. In plants, GPI-APs play indispensable roles throughout development, required for cell wall biosynthesis, embryo viability, organogenesis, reproductive development, and male–female interactive processes crucial for fertilization. LORELEI (LRE) and LRE-like GPI-APs 1, 2, 3 (LLG1, 2, 3) are closely related but differentially expressed. LRE is expressed exclusively in the ovule and loss of LRE function suppresses female fertility. lre mutants display reproductive phenotypes almost identical to those in fer mutants: a majority of lre and fer female gametophytes fail to induce rupture of the invading pollen tubes and their ovules are penetrated by multiple pollen tubes, yet fail to be fertilized because of the lack of sperm release. Here we show that LRE and LLG1 interact physically with FER and that they are crucial for its cell surface signaling capacity. Our results show partnering with related but differentially expressed proteins as a strategy for FER to execute its diverse biological roles; they also elucidate a novel mechanism for how GPI-APs might control cell surface signaling.

5Gene expression and mutant analyses showed that LLG1 is important for vegetative growth and development. LLG1 is the most prominent LRE family protein expressed in vegetative tissues, where FER expression is also prevalent. The LLG1 promoter::GUS (pLLG1::GUS) expression pattern (Figure 1A; Figure 1—figure supplement 1A) overlapped considerably with that of pFER::GUS in Arabidopsis seedlings (see Supplementary file 1 for a list of constructs used). Two T-DNA-induced knock-out mutants, llg1-1 and llg1-2, were indistinguishable from each other and from fer-4, a previously characterized knock-out fer mutant throughout vegetative development (Figure 1B–E; Figure 1—figure supplement 1B,C). Results described from here on are largely based on llg1-2, the mutant with which this work was initiated; observations made with llg1-1 provided confirmation.

6Similar to fer-4, llg1 plants showed retarded growth starting from 4–5 days after germination and seedlings looked visibly stressed, accumulating higher levels of anthocyanin and appeared more purplish than wild type seedlings (Figure 1C, upper). Under dark-grown conditions, llg1 were also de-etiolated and showed reduced apical hook bending relative to wild type, similar to fer-4 (Figure 1E; Figure 1—figure supplement 1C). Both llg1 and fer-4 remained smaller than wild type throughout growth and at maturity (Figure 1C lower; Figure 1D). Contrary to pronounced reproductive defects shared by fer-4 and lre mutants, llg1 plants had no reproductive phenotype and produced normal amounts of seeds, consistent with negligible LLG1 promoter activity in pollen and ovules (Figure 1—figure supplement 1A).

7llg1 mutants also developed root hair and trichome defects, similar to fer mutants. A large majority of llg1 root hairs collapsed upon emergence and those that emerged remained significantly shorter than wild type root hairs (Figure 2A,B). Trichomes on llg1 leaf epidermis were mostly defective, with a significant number of them having curly and more than three branches relative to those on wild type leaves (Figure 2C). Expression from a genomic LLG1 fragment and from a LLG1 promoter-expressed HA-tagged LLG1 (pLLG1::HA-LLG1) in llg1-2 fully complemented its phenotypes (Figure 2B,C; Figure 2—figure supplement 1), confirming that loss of LLG1 function underlies the fer-like defects in llg1 mutants. Furthermore, fer-4 llg1-2 double mutant seedlings were indistinguishable from their single mutant parents (Figure 2—figure supplement 2). Together these results are consistent with FER and LLG1 functioning in the same pathways and that both proteins are required for these pathways, just as FER and LRE are both required to mediate reproductive success by controlling similar events in pollen tube–ovule interaction.

8Given the role of FER in controlling RAC/ROP-regulated ROS production in seedling roots and that FER and LRE regulate ROS levels in the female gametophyte to mediate sperm release from the pollen tube, we examined whether loss of LLG1 conferred similar defects as those in fer-4 seedlings. llg1 root ROS levels were significantly reduced and did not respond to auxin-stimulated ROS accumulation as wild type did (Figure 3A). Furthermore, unlike wild type root hairs whose elongation was stimulated by auxin, llg1 root hair defects were not mitigated by auxin and emerged root hairs in llg1 remained shorter and substantially less sensitive to auxin stimulation (Figure 3B), similar to fer-4. Both fer-4 and llg1 seedlings were also defective in their epidermal cell pattern (Figure 3C), another auxin- and RAC/ROP-regulated property. Complementary lobes and indents along the surfaces of neighboring cells were considerably suppressed in llg1 and fer-4, giving rise to more box-shaped epidermal cells clearly distinguishable from those of a wild type seedling epidermis with interdigitated cells patterned like a jig-saw puzzle. Moreover, both FER and RAC/ROPs down-regulate ABA signaling. Like fer-4, llg1-2 was similarly hypersensitive to ABA-inhibited seedling development (Figure 4A; Figure 4—figure supplement 1A,B).

9RALF1 treatment of llg1-2 and fer-4 seedlings showed that these mutants were comparably less sensitive than wild type plants to RALF1-mediated growth responses. At the end of a 2-day treatment period, wild type seedlings were evidently shorter than their mock-treated counterparts, while treated fer-4 and llg1-2 seedlings as a population remained comparable in their sizes to their untreated counterparts (Figure 4B; Figure 4—figure supplement 2A,B). Since growth was not uniformly suppressed among fer-4 and llg1 mutants (see Figure 2—figure supplement 2A–C; Figure 4—figure supplement 2), actual root growth during the 2 days of RALF1 treatment was therefore measured to better quantify the response to RALF1. While wild type seedling growth during RALF1 treatment was significantly retarded relative to the control seedlings, RALF1-treated fer-4 and llg1-2 seedlings grew comparably with their mock-treated counterparts (Figure 4C; Figure 4—figure supplement 2C). Sensitivity to RALF1-regulated gene expression was also examined. When two representative RALF1-stimulated and two RALF1-suppressed genes were examined, their expression levels in llg1-2 and fer-4 were also less sensitive to the impact of RALF1 than in wild type seedlings (Figure 4D).

10These observations together indicate that FER and LLG1/LRE share the same function in several hormone- and RAC/ROP GTPase-mediated responses and they are both required for at least a subset of FER-mediated pathways.

11We demonstrated previously that FER interacts with RAC/ROPs in a multi-component signaling complex. Using the same protein pull-down strategy with ROP2, an Arabidopsis RAC/ROP, as bait, we observed that LLG1 was also pulled-down by ROP2 and in a guanine nucleotide-dependent manner, favored by GDP (Figure 5A), as was LRE (Figure 5; Figure 5—figure supplement 1). Distinct families of effectors are targeted by activated RAC/ROPs to mediate downstream pathways. We observed that GTP-saturated ROP2 preferentially pulled-down the N-terminal fragment of the Arabidopsis RbohD-encoded NADPH oxidase (Figure 5B), indicating that it is also a RAC/ROP effector, as previously demonstrated for rice RAC1 and RbohB. FER was identified as a ROPGEF interacting protein and GEF-RHO GTPase interaction is well established as has been demonstrated for ROPGEFs and RAC/ROPs in plants. Results reported here therefore imply a tetrameric FER-LLG1/LRE-ROPGEF-RAC/ROP signaling complex mediated by preferential ROPGEF binding to GDP-bound inactive RAC/ROPs, and once activated, the GTP-bound activated RAC/ROPs recruit NADPH oxidases to mediate downstream ROS-dependent processes. Emerging root hairs in fer-4 and llg1 seedlings began to extrude cytoplasm prior to their collapse (Figure 5C), consistent with a weakening cell wall as a result of reduced ROS production that ultimately led to their collapse.

12Functional interactions between FER and LLG1 occur beyond acting as components of the RAC/ROP signaling apparatus. In wild type vegetative organs, FER promoter-expressed FER-GFP localized to the cell membrane and intracellular signal was negligible (Figure 6A;; ). On the other hand, the intracellular FER-GFP signal was considerably more pronounced in llg1 hypocotyl (Figure 6B) and roots (Figure 6; Figure 6—figure supplement 1) than in wild type tissues, often appearing in reticulate arrays and perinuclear regions, patterns reminiscent of the ER. Intracellular FER-GFP signal was prevalent in emerging root hairs prior to or during their rupture (Figure 6C), and colocalized with ER-Tracker-stained membrane patches (Figure 6—figure supplement 1C,D).

13In the ovules, FER-GFP appears most prominently at the filiform apparatus (Figure 6D;; ), a synergid cell membrane-enriched cell wall region at the entrance to the female gametophyte. The concentrated localization of FER-GFP at the filiform apparatus was highly compromised in lre ovules. A significantly higher percentage of mutant ovules relative to wild type showed female gametophyte FER-GFP either totally retained in the synergid cell cytoplasm (Figure 6F) or distributed between the filiform apparatus and inside the synergids (Figure 6E; Figure 6—figure supplement 2B). FER-GFP retained inside the synergids sometimes appeared as cytoplasmic patches (Figure 6F; Figure 6—figure supplement 2A), reflecting a dense synergid cell endomembrane system (see, for example, ) that supports secretion of cell wall materials to construct the filiform apparatus.

14To further understand how loss of LLG1 affects FER-GFP localization, we transformed protoplasts derived from wild type and llg1 plants, where the localization of 35S promoter-expressed FER-GFP (35S::FER-GFP) could be clearly discerned and dose-dependence on the introduced transgenes could be quantitatively assessed. When transformed using the same amount of input FER-GFP DNA, llg1-2-derived protoplasts showed considerably stronger intracellular signals than their counterpart wild type cells (Figure 7; Figure 7—figure supplement 1 and Figure 7—figure supplement 2), consistent with intracellular FER-GFP signal being prominent in llg1 plants (Figure 6). The cytoplasmic FER-GFP signal in llg1 protoplasts often appeared in reticulate structures, reminiscent of the ER, and colocalized with a co-expressed ER marker (RFP-ER) (Figure 7C,D; Figure 7—figure supplement 2B,C). Cell surface localization of FER-GFP was not obliterated from llg1 protoplasts, but could clearly be resolved from its ER locations in cells where cortical ER was prominent (Figure 7D). Co-transfection of llg1 protoplasts by LLG1, epitope- or fluorescent protein-tagged LLG1 all restored cell membrane localization of FER-GFP (Figure 7A,B; Figure 7—figure supplement 2A,D). On the other hand, expression of LLG1ΔC, deleted of its C-terminal signature sequence for GPI-anchor modification, was considerably less effective in restoring FER-GFP to the plasma membrane of llg1 cells (Figure 7B). Together with observations in llg1 seedlings (Figure 6), these results indicate that FER depends on GPI-anchored LRE and LLG1 for efficient cell membrane location.

15Results thus far support the notion that LLG1 and LRE are critical for the biological functions of FER by facilitating localization of FER to the cell membrane (Figures 6, 7) where they also associate with the RAC/ROP complex (Figure 5) to mediate downstream processes (Figures 3, 4). To understand how LLG1 and LRE could attain these two properties, we explored whether they might interact directly with FER (Figures 8, 9). Protein–protein interaction assays showed that protoplast-expressed FER was pulled-down by LLG1 and LRE in vitro (Figure 8B) and FER co-immunoprecipitated with LLG1 that was co-expressed in protoplasts (Figure 8C), suggesting direct interactions between these proteins in plant cells. Moreover, protoplasts-coexpressed FER-GFP and LLG1-HA were co-pulled down by MBP-RALF1 (Figure 8D), indicative of RALF1 interaction with the FER-LLG1 complex.

16Bimolecular fluorescence complementation (BiFC) assays were used to further examine FER-LLG1 interaction in plant cells (Figure 9; Figure 9—figure supplement 1). LLG1 and a truncated version of FER deleted of its kinase domain (FERΔK) (Figure 8A), which could be expressed at higher levels than full-length FER, were reciprocally fused to the N- and C-terminal halves (VN, YC, respectively) of Venus. These split Venus halves were also fused behind the FER signal peptide for targeting to the secretory pathway. Co-expression of cognate pairs of these split Venus in agroinfiltration transformed tobacco leaf epidermis reconstituted Venus yellow fluorescence along the cell surface (Figure 9A,B), reflecting FER-LLG1 interaction on the leaf cell membrane, and consistent with FER-GFP and GFP-LLG1 being predominantly located to the cell surface when expressed alone in similar assays (Figure 9—figure supplement 1A).

17In transformed Arabidopsis protoplasts for BiFC analysis, typically about half of the BiFC positive cells displayed prominent cell membrane-localized signal with little notable intracellular signals (Figure 9B, upper panel). The other half harbored prominent Venus signal inside the cell, some appearing in reticulate structures, along with signal on the cell surface or cell cortex (Figure 9B, lower panel; Figure 9—figure supplement 1C). When the split Venus pairs were co-expressed with RFP-ER, reconstituted intracellular yellow fluorescence colocalized with the ER marker (Figure 9D,E), reflecting FER-LLG1 interaction in the ER. The notable BiFC signal within the ER most probably had resulted from retarded trafficking of the interacting proteins out of this compartment. An Arabidopsis ARF1 mutation, ARF1(Q71L), has been shown to block ER to Golgi trafficking and trap secretory proteins in the ER. When co-expressed with the split Venus pair of FERΔK and LLG1, ARF1(Q71L) dramatically increased the number of BiFC-positive cells (to >90%) showing intracellular yellow fluorescence. The reconstituted Venus signal often appeared perinuclear and in highly dynamic reticulate structures, typical of the ER (Figure 9C; Videos 1, 2). These observations further support FER-LLG1 interaction occurred already in the ER where these proteins would first encounter each other early in the secretory process.

18Having determined that LLG1 is crucial for FER localization to the cell membrane and that FER and LLG1 interact physically (Figures 6–9), we sought further evidence for these interactions and their biological significance. We observed that FERΔexJM, FER deleted of the extracellular juxtamembrane region (exJM, amino acid residues 365–447) (Figure 8A) was less efficiently pulled-down by LLG1 (Figure 10A), while the 83 amino acid exJM fragment was adequate to pull-down LLG1 (Figure 10B). Yeast two-hybrid assays also indicated that exJM interacted with LLG1 and LRE (Figure 10C; Figure 10—figure supplement 1). Furthermore, when pFER::FERΔexJM-GFP was transformed into Arabidopsis, FERΔexJM-GFP showed prominent intracellular localization in structures reminiscent of the ER (Figure 10D,E; Figure 10—figure supplement 2A). These intracellular FERΔexJM-GFP signals also colocalized with the ER-tracker dye (Figure 10E) and with RFP-ER when co-expressed in protoplasts (Figure 10—figure supplement 2B). These observations together indicate that LLG1 and LRE target the exJM in FER for binding and that FER depends on these interactions for proper delivery to the cell membrane. They also imply that the phenotypes in llg1 and lre, together spanning the developmental spectrum of fer phenotypes, could have largely resulted from diminished FER localization to its proper functional location in the cell membrane, thus mimicking loss of FER.

19Results reported here demonstrate that related GPI-APs LLG1 and LRE are functional counterparts that physically interact with FER and essential for its cell surface signaling capacity. Partnering with and relying on this differentially expressed protein pair clearly would provide versatility for the almost constitutively expressed FER to control when and where signaling can be deployed. Moreover, modulation of LLG1 and LRE expression and their post-translational modifications, which involve numerous GPI-anchor biosynthetic and restructuring steps, would provide many opportunities for FER signaling activity to be fine-tuned to meet a broad range of developmental needs and environmental challenges. Given the spectrum of phenomena already known to be regulated by FER, LLG1 and LRE could be of core importance to how FER treads among several signaling pathways to mediate growth and developmental processes regulated by multiple hormones, reproduction, and defense, each potentially controlled by distinct signals. FER is a member of a family of 17 related Arabidopsis receptor kinases, which are also conserved in other plant species; LRE-like proteins are also present in other plants. It is conceivable that pairing with their cognate LRE-like proteins also underlies the signaling functions of other FER-related receptor kinases. GPI-APs, including, for example, the COBRA family proteins and arabinogalactan proteins, play crucial roles in plant growth and reproduction. Demonstrating that FER depends on LLG1 and LRE for efficient localization to the cell membrane (Figures 6, 7) elucidates not only a strategy whereby FER acquires its cell surface signaling capacity but also a mechanism with which GPI-APs might enable the signaling capacity of a broader set of cell surface receptors.

20The FER-LLG1/LRE partnership also underscores the importance and the versatility of GPI-APs as regulators for cell signaling activities. GPI-APs are secreted to the outer leaflet of the cell membrane. LLG1 and LRE binding to the exJM domain of FER (Figure 9) is consistent with their being tethered to the cell membrane via the GPI anchor, thus facilitating interactions with FER immediately along the outer cell surface. GPI anchors are assembled and transferred to proteins destined for lipid modification in the ER. GPI-APs are known to undergo structural remodeling in the ER for efficient trafficking to the Golgi in vesicles that are distinct from those for other secretory proteins, already marking them for delivery to specialized sphingolipids- and cholesterols-enriched cell membrane microdomains. With its signal peptide, FER could be secreted by default to the cell membrane. That FER and LLG1 already interact in the ER (Figure 9) and this interaction underlies efficient FER localization to the cell membrane (Figures 6, 7 and 10) are consistent with at least a fraction of FER exiting the ER as a FER-LLG1/LRE complex (Figure 11). That FER and LLG1/LRE remain associated with the RAC/ROP signaling complex along the cell membrane (Figures 5, 9) also provides additional support for FER-LLG1/LRE being transported from the ER to the cell membrane as a complex. In complexing with FER in the ER and ‘chaperoning’ its delivery to the cell membrane, LLG1 and LRE ensure delivery of FER to GPI-AP-destined micro membrane environments for its proper functional location and assembly of the RAC/ROP signaling apparatus (Figure 11). In the remaining part of the FER-ROPGEF-RAC/ROP signaling complex (Figure 5) and being required for several RAC/ROP regulated processes (Figure 4; ), LLG1 and LRE apparently also function as an integral component of the FER signal reception apparatus. They might directly participate in signal perception by FER and/or regulate how FER interacts with various possible binding targets, such as RALF1 or other hormones and cell wall carbohydrates. They also conceivably provide a landmark for the assembly of the FER signaling apparatus and regulation of its activity such as by maintaining its stability or inducing its recycling from the cell membrane, roles known to be played by GPI-APs. To what extent the LRE family protein controls the delivery of receptor kinases as a class or more specifically those closely related to FER remains to be determined.

21Results presented here also provide evidence for the notion that LLG1/LRE acts as a co-receptor to mediate at least those FER-regulated processes examined here. On the phenotypic level, growth and developmental defects in llg1 (Figures 1, 2) and reproductive defects in lre together span the spectrum of vegetative and reproductive phenotypes in fer mutants. The non-additive phenotype of fer-4 llg1-2 double mutant relative to each of its parent single mutants (Figure 2—figure supplement 2) provides further support that FER and LLG1 function in and both are required for the same pathways. Biochemically, both LLG1 and LRE physically interact with FER on the cell membrane (Figure 9) and exist as components of the FER-ROPGEF-RAC/ROP-NADPH oxidase signaling pathway (Figure 5), consistent with their serving with FER as signal mediators on the cell surface to RAC/ROPs. This is further supported biologically by loss of LLG1 or LRE inducing the same signaling defects as in fer null mutants, including processes regulated by auxin, ABA, RALF1, and ROS (Figures 3, 4; ). Besides LLG1/LRE, RALF1 is thus far the only other molecule reported to interact with the FER extracellular domain, although the precise RALF1 target site on FER remains unknown. That RALF1 interacts with co-expressed FER and LLG1 (Figure 8D) and that both FER and LLG1 are needed to mediate RALF1 signaled responses (Figure 4B–D) indicate that the FER-LLG1 complex indeed has the capacity to serve as a co-receptor for this first known ligand of FER. Given the already known participation of FER in multiple hormone and defense signaling pathways, the FER-LLG1/LRE complex could equally be a surface co-regulator for multiple signals. Moreover, FER is also broadly speculated to interact with cell wall carbohydrates by virtue of its extracellular homology with the disaccharide-binding malectin. Therefore, the extracellular interactions engaged by FER are likely to be complex and influenced by multiple factors whose presence fluctuates depending on cellular and environmental conditions. The FER-LLG1/LRE partnership discovered here laid the ground work towards a more comprehensive understanding of how FER attains its multiple biological role; this will however require first deciphering other biochemical interactions maintained by FER.

22Plant growth followed previously described conditions. Tissue culture-grown plants were maintained on B5 medium supplemented with 1% sucrose and solidified by 0.7% agar. Seeds were cold-treated at 4°C for 2 days before being transferred to 22°C for germination and growth under 16/8 hr light/dark cycles, or in total darkness for dark-grown seedlings. For growth to maturity, seeds were either sown directly on soil, or 10-day-old tissue culture-grown seedlings were transferred to soil, and maintained in a growth chamber at 20–22°C under 16/8 hr light/dark cycles. Arabidopsis thaliana Col-0 was used as control for llg1-2 (SALK_086036) and llg1-1 (SAIL_47_G04). Both llg1 mutants behaved similarly throughout growth and development and did not display discernable reproductive defects. Homozygous fer-4 and lre-5 were as previously described. Double fer-4 llg1-2 was generated by a genetic cross. RALF1-regulated growth used Escherichia coli-produced His6-RALF1 and followed previously described conditions. Growth for RALF1 treatment for RT-PCR analysis followed ).

23Arabidopsis was transformed by floral dip. Transient transformation assays were carried out by agroinfiltration of Nicotiana tabacum var SR1 grown at ∼25°C in a growth room. A wound was made in the abaxial epidermis and about 1 ml of bacteria (at 0.1–0.4 OD600) was injected into these spots using a 1 ml syringe. Transient transfection of Arabidopsis protoplasts from 4-week-old soil-grown wild type and llg1-2 plants, and of tissue culture-grown wild type Arabidopsis protoplasts followed procedures in ) and ), respectively. Unless otherwise indicated, DNA amounts used for protoplast transfection were: 1 μg of pFER::FER-GFP; varying amounts of 35S::LLG1 or 35S::LLG1 derivatives (indicated in figures); 1–2 μg of the ER marker 35S::RFP-ER; 4 μg of each split Venus half and 5 μg of 35S::ARF1(Q71L). Empty vector (Bluescript vector SK) DNA was used to equalize the amount of DNA used in comparative assays.

24All recombinant DNA procedures followed standard and PCR-based methodology. A list of constructs is shown in Supplementary file 1; domain maps for some are shown in Figure 8. Plant genomic DNA was used for PCR analysis of T-DNA inserts in transformed plants. RNA for expression analysis by RT-PCR was isolated from 10-day- old seedlings following the manufacture's protocol (PrepEase RNA isolation kit; USB/Affymetrix, Santa Clara, CA). Histochemical staining for GUS activity followed the standard procedure. Primers for RT-PCR of RALF1-regulated genes are: (1) BR6OX2: forward, GAG ACA TCA AGA TTG GCA ACG; reverse, GTA AGG TGA ACA CTT AAG ATGG; (2) GA3OX-1: forward, CAA GTA TTT CGC GAT GAT CTT GG; reverse, G ATA CTC TTT CCA TGT CAC CG; (3) CML38: forward, ATG AAG AAT AAT ACT CAA CCT C; reverse, GCG CAT CAT AAG AGC AAA CTC; (4) ERF6: forward, ATG GCT ACA CCA AAC GAA GTA TC; reverse, AAC AAC GGT CAA TTG TGG ATA ACC.

25Plant phenotype and data analyses mostly followed. Root hairs located between 1.5 and 3.5 mm from the primary root tip of 4-day-old seedlings were examined. For auxin treatments, 1-naphthaleneacetic acid (NAA) was added at concentrations indicated in the figures. ABA treatment followed that in ); hormone was added directly to seed germination plates. For RALF1 treatments, E. coli-produced His6-RALF1 was purified according to ). Two-day-old light-grown seedlings were treated with His6-RALF1 for 2 days according to ) at concentrations indicated in the figures. Root lengths were measured at the beginning and end of treatments to obtain growth during treatment. Epidermal cell analysis was carried out as described. ROS in the primary roots and root hairs were detected by H2DCF–DA (2′, 7′-dichlorodihydro-fluorescein diacetate; Sigma/Aldrich, St. Louis, MO) and ROS fluorescence intensity within a fixed region of interest (ROI) was quantified using Image J.

26FER-GFP localization in ovules was acquired as described in ). The synergid cells from one ovule to another are not identical in shape, size, or relative orientation with the rest of the ovule parts. For comparative quantitative analysis of data between wild type and mutant ovules, signals from the entire recognizable filiform apparatus and synergid cells were quantified and relative signal distribution between the filiform apparatus and the synergid cell cytoplasm was compared between wild type and mutant ovules.

27For protein pull-downs, bait proteins (MBP-LLG1, MBP-LRE1, MBP-ROP2, MBP-exJM, His6-LLG1, MBP-RALF1) were produced in E. coli and bound to amylose or talon resins as previously described. Plant proteins (FER-HA, FER-GFP, FERΔexJM-GFP, HA-LLG1, RbohD(N)-HA) were expressed in protoplasts (<10 μg DNA per transfection) and extracted at ∼12 hr after transfection in pull-down buffer (binding buffer: 40 mM Tris–HCl, pH 7.5, 100 mM NaCl, 1 mM Na2-EDTA; plus 5% glycerol, 5 mM MgCl2, 1 mM PMSF, protease inhibitor mixture [Calbiochem, San Diego, CA] at 1:100 dilution, and 0.4% Triton X-100 to facilitate solubilization). Protoplast protein extracts or E. coli-produced target proteins were applied to bait protein-bound resins and incubated at 4°C for 2 hr with gentle mixing. The resin was washed three times in binding buffer. Proteins remained bound to the resin were eluted by mixing with SDS/PAGE loading buffer, boiled for 5 min, and applied to SDS/PAGE (7.5% for FER; 12.5–17.5% for LLG1 and LRE) for immunoblot analysis. Protein blots were stained by Ponçeau S Sigma-Aldrich for sample loading comparison, followed by immunostaining. Primary (anti-HA and anti-GFP) and secondary antibodies for chemiluminescence detection were from Santa Cruz. Signals were acquired by the PXi imaging system (Syngene, Cambridge, UK).

28MBP-ROP2 pull-down of protoplasts-expressed HA-LLG1 and HA-LRE followed the previously described procedure. Pull-down of RbohD(N)-HA was carried out similarly with MBP-ROP2 resin pretreated by 10 mM GTP or GDP for 2 hr and pull-down carried out with 10 mM GTP or GDP in the buffer.

29For co-immunoprecipitation, 35S::HA-LLG1 and 35S::FER-GFP were co-expressed in transfected protoplasts. Mock samples were transfected with 35S::FER-GFP and empty SK vector DNA. Proteins were extracted as described above. Anti-HA antibody was used at 1:100 dilution for each immunoprecipitation in 1 ml reactions, incubated at 4°C for 3 hr, followed by the addition of 50 μl of protein G resin suspension (Santa Cruz Technology, Dellas, TX.). After binding for 1 hr, the resin was washed five times in binding buffer. Proteins remained bound to the resin were eluted in SDS/PAGE loading buffer, boiled for 5 min, and applied to SDS/PAGE for immunoblot analysis as described above.

30For BiFC, the split-VENUS system was used for Arabidopsis protoplast transfection and agroinfiltration of tobacco leaf epidermis.

31For yeast two-hybrid assays, products and procedures from Stratagene were used for vectors, yeast growth, selection, and β-galactosidase activity.

32Seedling, inflorescence, and trichome images were acquired on an Olympus SZ61 microscope. Epifluorescence and DIC microscopy were carried out on a NIKON Eclipse E800 microscope equipped with a SPOT camera (Molecular Diagnostic). Filters from Chroma were used: green fluorescence, Ex460-500/DM505/BA510-560; red fluorescence, Ex546(10)/ DM565LP/EM590LP; yellow fluorescence, Ex490-510/DM515/BA520-550. Confocal imaging was carried out on a Zeiss Meta510 or a Nikon A1. Comparative studies were based on identical imaging conditions and followed procedures described in ). Use of ER-Tracker Red (BODIPY TR Glibenclamide; Life-Technologies) to stain ER followed ), ), and the manufacturer's protocol.

33All data presented are representative of at least three independent experiments with comparable results. Quantitative data are presented as averages ± SEM of replicated experiments or samples, or as averages ± SD from a representative experiment as indicated in the figure legends. Sampling sizes are indicated in the figures or figure legends. Student's t-tests were used for p value calculations. p<0.05 is considered significant (indicated by * in figures); most experimental and control data were significantly different with p values ranging from 10−2 to 10−5 (indicated by ** in figures).