Integration of two RAB5 groups during endosomal transport in plants

RAB5 is a key regulator of endosomal functions in eukaryotic cells. Plants possess two different RAB5 groups, canonical and plant-unique types, which act via unknown counteracting mechanisms. Here, we identified an effector molecule of the plant-unique RAB5 in Arabidopsis thaliana, ARA6, which we designated PLANT-UNIQUE RAB5 EFFECTOR 2 (PUF2). Preferential colocalization with canonical RAB5 on endosomes and genetic interaction analysis indicated that PUF2 coordinates vacuolar transport with canonical RAB5, although PUF2 was identified as an effector of ARA6. Competitive binding of PUF2 with GTP-bound ARA6 and GDP-bound canonical RAB5, together interacting with the shared activating factor VPS9a, showed that ARA6 negatively regulates canonical RAB5-mediated vacuolar transport by titrating PUF2 and VPS9a. These results suggest a unique and unprecedented function for a RAB effector involving the integration of two RAB groups to orchestrate endosomal trafficking in plant cells.


Introduction
Eukaryotic cells contain various single-membrane-bound organelles, each of which possesses distinctive constituents and functions. For an organelle to maintain a specific identity and function, the protein and lipid content must be strictly regulated; however, organelles are actively interconnected, exchanging various substances to fulfill complex and diverse cellular activities. The inter-organelle trafficking system mediated by vesicular and/or tubular trafficking intermediates plays an integral role in ensuring proper organelle function. A single round of membrane trafficking involves several sequential steps: budding of vesicles/tubules from a donor organelle, delivery of the transport vesicles, and tethering and fusion of the vesicles with the target membrane. RAB GTPase is an evolutionarily conserved key regulator of the targeting/tethering step and is also responsible for other cellular activities, such as organelle movement and inter-and intracellular signaling (Saito and Ueda, 2009;Stenmark, 2009). These functions of RAB GTPases are fulfilled through nucleotide statedependent binding with specific sets of proteins, collectively called effector proteins. To ensure the organized and specialized functions of each RAB GTPase, interactions with effectors should be temporally, spatially, and combinatorially regulated for each trafficking event (Grosshans et al., 2006;Zerial and McBride, 2001).
The early endosome acts as a communication platform between the intracellular environment and the cell surface and/or extracellular space. RAB5 is a key regulator of a wide spectrum of early endosomal functions in animal cells, including homotypic fusion between early endosomes, endosomal eLife digest Living cells often contain compartments that pass proteins, fats and other biological molecules to one another via a process called membrane trafficking. Endosomes are one of the key platforms of membrane trafficking. These structures accumulate molecules from the outside of the cell, sort them, and then redirect them back to the cell surface or send them to other compartments within the cell where they can be broken down. Proteins known as RAB5s regulate many of the activities of endosomes. Some are found in a wide range of organisms, including animals, fungi, and plants, and are referred to as the "canonical" RAB5 group. Another group of RAB5 proteins are unique to land plants and some green algae.
The existence of two RAB5 groups (i.e. canonical and plant-unique) is a distinctive feature of plant cells. In 2011, researchers showed that a plant-unique RAB5 could interfere with and counteract the activities of a canonical RAB5. However, it remained ambiguous how these proteins could do this.
To resolve this question, Ito et al. -who include several researchers from the 2011 study -set out to find proteins that interact with a plant-unique RAB5 from Arabidopsis thaliana. The experiments identified one partner of a plant-unique RAB5, which was named PUF2. Unexpectedly, further experiments revealed that PUF2 also regulates canonical RAB5. PUF2 was found on the surface of the endosome together with RAB5s and a protein that activates RAB5s. Notably, PUF2 also interacted with the activating factor and the inactive form of canonical RAB5.
Based on these findings, Ito et al. propose that PUF2 acts as a landmark to bring inactive canonical RAB5 close to its activating factor, which helps to activate canonical RAB5. They suggest that the plant-unique RAB5 also competitively binds to the landmark and blocks the canonical RAB5.
Membrane trafficking is a universal system for all living organisms, yet the system seems to be customized among different organisms. These new findings provide further evidence that land plants have evolved a unique mechanism to regulate the activities of their endosomes. The next step is to reconstruct how this system evolved and unravel its relevance to the evolution of plant-specific traits.
isoprenylated at the C-terminus, which are essential for membrane binding and functions of canonical RAB GTPases (Seabra, 1998), ARA6 harbors an extra stretch in the N-terminus, where this protein is N-myristoylated and palmitoylated to target a distinct subpopulation of endosomes from canonical RAB5 with substantial overlap (Haas et al., 2007;Ueda et al., 2001). ARA6 promotes formation of the SNARE complex, which contains plant-specific R-SNARE VAMP727 at the plasma membrane (Ebine et al., 2011), and an endosomal function has also been described; overexpression of the nucleotide-free mutant form of ARA6 results in impaired vacuolar trafficking (Bolte et al., 2004;Bottanelli et al., 2012).
Despite their distinct functions, ARA6 and canonical RAB5 share the common activating factor, VPS9a (Goh et al., 2007). Intriguingly, loss-of-function mutations of ARA6 and canonical RAB5 confer counteracting effects in a vps9a mutant, further highlighting the distinct functions of these two plant RAB5 groups (Ebine et al., 2011), although the molecular mechanism integrating these two groups in endosomal trafficking remains unexplored.
In the present study, we identified and characterized the first effector molecule of ARA6, PLANT-UNIQUE RAB5 EFFECTOR 2 (PUF2). Based on these results, PUF2 is a key integrator of the two RAB5 groups in the unique endosomal trafficking system of plants.

PUF2 interacts with plant-specific RAB5
Consistent with the notion that endosomal trafficking pathways in animals and plants have substantially diverged, close homologs of well-characterized RAB5 effectors, such as EEA1 and Rabaptin-5, do not exist in plants. To identify effector molecules of ARA6, we performed yeast two-hybrid screening using a GTP-fixed mutant of ARA6 (ARA6 Q93L ) as bait. After screening 4.57 Â 10 5 independent clones, we obtained a candidate clone encoding the C-terminal region of At1g24560 ( Figure 1A). At1g24560 encodes a 678-amino-acid protein of unknown function, without clear homologs in animals or yeasts. We designated this protein PUF2. Bacterially expressed and purified full-length PUF2 also bound to GST-tagged ARA6 Q93L but not to GST-ARA6 S47N or -ARA7 Q69L in vitro ( Figure 1B). Surprisingly, PUF2 also interacted with ARA7 S24N ( Figure 1B). Based on a coimmunoprecipitation analysis using a lysate prepared from a transgenic plant expressing PUF2-GFP and an anti-GFP antibody, PUF2-GFP and ARA6 form a complex in planta, whereas canonical RAB5 members ARA7 and RHA1 were not co-precipitated with PUF2-GFP under these experimental conditions ( Figure 1C, left panels). However, when a crosslinker (dithiobis succinimidyl propionate; DSP) was added to the reaction, canonical RAB5 was also co-precipitated ( Figure 1C, right panels). Thus, PUF2 interacts with active GTP-bound ARA6, and PUF2 also weakly and/or transiently forms a complex with GDP-bound canonical RAB5.
Four coiled-coil domains were predicted in the PUF2 protein using a simple modular architecture research tool (SMART, http://smart.embl.de/) (Letunic et al., 2015;Schultz et al., 1998), although PUF2 contained no known functional domain. We subsequently examined whether these coiled-coil regions are responsible for the interaction with ARA6. Truncated PUF2 containing only the fourth coiled-coil region was isolated in the yeast two-hybrid screening. Consistently, truncated PUF2 containing the other coiled-coil region did not interact with ARA6 ( Figure 1D). Furthermore, full-length PUF2 did not interact with ARA6, which may indicate that these regions negatively regulate the interaction between ARA6 and PUF2. We did not detect interactions between any PUF2 constructs and ARA7 (Figure 1-figure supplement 1).

PUF2 preferentially colocalizes with canonical RAB5 on endosomes
To investigate the subcellular localization of PUF2, we observed transgenic plants expressing GFPtagged PUF2 under the regulation of its own regulatory elements (promoter, introns, and terminator), thus retaining the authentic function of PUF2 as described below. As shown in Figure 2A, PUF2 localized to punctate organelles in the cytoplasm, which dilated with the application of wortmannin (Wm; a phosphatidylinositol-3 and -4 kinase inhibitor) and aggregated into so-called BFA bodies with brefeldin A (BFA; an ARF GEF inhibitor) treatment. These drug responses were similar to those of multivesicular endosomes bearing ARA6 and/or ARA7 (Ebine et al., 2011;Grebe et al., 2003; Figure 1. PLANT-UNIQUE RAB5 EFFECTOR 2 (PUF2) isolated as an ARA6 effector. (A) Yeast two-hybrid interaction between the C-terminal region of PUF2 and RAB5s (ARA6 or ARA7). PUF2 388-672 aa was expressed as a fusion protein with a transcriptional activation domain (AD), and ARA6 and ARA7 with the indicated mutations were expressed as fusion proteins with DNA binding domains (BDs) in the yeast strain AH109. Interactions with the HIS3 reporter gene were examined. Q to L mutations are GTP-freezing mutations, and S to N mutations are GDP-freezing mutations. (B) Interaction Figure 1 continued on next page Ito et al., 2012;Jaillais et al., 2008). The endosomal nature of the PUF2-positive compartments was also supported by their accessibility to an endocytic tracer FM4-64 (Figure 2A, arrowheads).

Endosomal localization of PUF2 does not require RAB5 activity
To determine whether the endosomal localization of PUF2 requires RAB5 activity, we observed the localization of PUF2-GFP expressed in mutants of ARA6 and the common activator of all RAB5 members in Arabidopsis, VPS9a (Goh et al., 2007). In ara6-1 mutant plants, PUF2-GFP localized to punctate compartments, which were swollen and aggregated after treatment with Wm and BFA, respectively, and were labeled with FM4-64 as observed in wild-type plants ( Figure 2D, arrowheads). PUF2-GFP also exhibited the same responses to drugs and FM4-64 accessibility in the vps9a-2 mutant ( Figure 2E). Thus, ARA6 and canonical RAB5 activities are not required for the recruitment of PUF2 onto endosomes.

Cooperation between PUF2 and RAB5 is required for plant development
We next examined the effects of the loss of PUF2 function using a puf2 mutant, in which PUF2 was not detected at either the mRNA or protein level ( . Thus, we explored the potential genetic interactions between PUF2 and the two RAB5 groups. The puf2 ara6-1 double mutant was indistinguishable from wild-type, puf2, and ara6-1 plants. By contrast, the puf2 rha1 double mutation resulted in dwarfism ( Figure 3A), suggesting that the functions of PUF2 and canonical RAB5 lie in the same trafficking pathway.
We previously demonstrated that ara6-1 and ara7/rha1 mutations exerted opposing effects on syp22-1, a mutation in a SNARE protein mediating membrane fusion at the vacuole, and vps9a-2; ara7 and rha1 exaggerated the deleterious phenotypes of syp22-1 and vps9a-2, whereas ara6-1 remedied these phenotypes (Ebine et al., 2011). If PUF2 is a downstream effector mediating ARA6 function, then loss of PUF2 function should exert effects similar to that of ara6-1 on syp22-1 and vps9a-2 mutations. However, contrary to this hypothesis, the puf2 syp22-1 double mutant exhibited more severe wavy-rosette and late-flowering phenotypes than the syp22-1 mutant ( Figure 3B), which were restored by introducing the genomic sequence of PUF2 (Figure 3-figure supplement 1E and F). A similar synergistic genetic interaction was also observed between puf2 and vps9a-2; the hemizygous puf2 mutation aggravated the growth phenotypes of vps9a-2, resulting in seedling lethality ( Figure 3C). Conversely, the puf2 mutant harboring the hemizygous vps9a-2 mutation (puf2 vps9a-2+/À) exhibited no discernable abnormality, enabling segregation analysis. The observation of developing seeds in seedpods of the puf2 vps9a-2+/À mutant revealed that approximately onefourth of the seeds exhibited a brownish and shriveled appearance (normal:abnormal = 76:25, n = 101 seeds, p<0.001, chi-square test, c 2 = 67.7) ( Figure 3D), indicating that the double mutation resulted in halted embryogenesis. We subsequently observed embryos collected from puf2 vps9a+/ À seedpods. Although we were unable to distinguish double mutants from their normal siblings at the globular stage, double-mutant embryos exhibited severe developmental retardation at later developmental stages ( Figure 3E). The embryonic lethality of the puf2 vps9a-2 mutant was restored after introducing a genomic fragment containing PUF2 (  Hemizygous mutation of puf2 (puf2+/À) aggravated the vps9a-2 phenotype. Sixteen-day-old seedlings are shown. Bar = 1 cm. (D) Seed pods collected from wild-type, puf2, vps9a-2 and puf2 vps9a-2+/À plants. Approximately a quarter of the seeds collected from the puf2 vps9a-2+/À plants exhibited a brown and deflated appearance (arrowheads). (E) Embryogenesis of the puf2 vps9a-2 double mutant was delayed and terminated at the globular stage. Corresponding images in the upper and lower sets represent normal and abnormal embryos collected from the same seedpods at different developmental stages. Bars = 50 mm. (F) Overexpression of PUF2 rescued defective primary root elongation of the vps9a-2 mutant. Ten-day-old seedlings are presented. Bar = 1 cm. DOI: https://doi.org/10.7554/eLife.34064.008 The following source data and figure supplements are available for figure 3: Similar complementation was also observed for the genomic PUF2 fragment, in which cDNA for GFP was inserted after the start codon or in front of the stop codon, indicating the functionality of GFP-tagged PUF2. An intimate genetic interaction between PUF2 and VPS9a was also observed in this overexpression experiment: overexpressed PUF2 partially rescued the defective growth of vps9a-2 ( Figure 3F, Figure 3-figure supplement 1I and J). As the deleterious phenotype of vps9a-2 reflects the defective activation of canonical RAB5s rather than ARA6 (Goh et al., 2007), this genetic evidence indicates that PUF2 acts in the same trafficking event as canonical RAB5, although PUF2 preferentially interacts with GTP-bound ARA6.
Impact of the puf2 mutation on vacuolar transport Canonical RAB5 acts during the transportation of soluble cargos such as 12S globulin and GFP-CT24 to protein storage vacuoles in Arabidopsis seeds, a process that also involves SYP22 (Ebine et al., 2011;Ebine et al., 2014;Ebine et al., 2008). If PUF2 indeed acts in the same trafficking pathway as canonical RAB5 and SYP22, then mutations in PUF2 would synergistically affect the transport of vacuolar cargos in syp22-1. We verified this hypothesis by examining the accumulation of storage protein precursors in a puf2 syp22-1 double mutant. Although processing of storage proteins was not markedly affected in the puf2 mutant ( Figure 4A), precursor proteins with larger molecular masses were observed in the puf2 syp22-1 double mutant ( Figure 4A, arrowhead). Immunoblot analysis using an anti-12S globulin antibody also revealed the accumulation of 12S globulin precursors in puf2 syp22-1 and puf2+/À vps9a-2, as observed in rha1 syp22-1 and vps9a-2 mutants ( Figure 4B and Synergistic enhancement of the trafficking defect in vps9a-2 by puf2 was also observed in vegetative tissue. It is possible to monitor the constitutive transport of PIN2-GFP from the plasma membrane to the vacuole by the accumulation of GFP fluorescence in the vacuole of root epidermal cells of dark-grown wild-type and puf2 mutant plants ( Kleine-Vehn et al., 2008;Tamura et al., 2003; Figure 4D). Weaker fluorescence was observed in vps9a-2 plants than in wild-type and puf2 plants, indicating partial impairment of the endocytic pathway as previously reported (Inoue et al., 2013). This trafficking defect was substantially enhanced by hemizygous puf2 mutation; fluorescence in the vacuole was barely detected, and PIN2-GFP accumulated at punctate or ring-shaped compartments, which likely represent endosomal compartments in the puf2+/À vps9a-2 mutant ( Figure 4D

PUF2 also interacts with VPS9a
The observation that PUF2 acts in the same trafficking pathway as canonical RAB5, even though this protein was isolated as an ARA6 effector, prompted us to determine whether PUF2 integrates the functions of plant-unique and canonical RAB5 groups at endosomes. Therefore, we examined the interactions between PUF2 and other RAB5-related molecules and observed that PUF2 also interacted with VPS9a in a yeast two-hybrid assay ( Figure 5A). Deletion analysis revealed that this interaction was mediated by the N-terminal coiled-coil region of PUF2 and was distinct from the interaction with GTP-bound ARA6 mediated by the C-terminal coiled-coil region ( Figure 1D, Figure 5A, Figure 5-figure supplement 1). The in planta interaction between PUF2 and VPS9a was also confirmed by performing co-immunoprecipitation analysis ( Figure 5B). This interaction was direct: bacterially expressed and purified GST-tagged PUF2 37-127 , which contained the N-terminal coiled-coil region, pulled down purified HA-tagged VPS9a, while PUF2 461-639 , which contained the C-terminal coiled-coil region, did not ( Figure 5C). Furthermore, VPS9a and PUF2 colocalized at a subpopulation of endosomes bearing ARA6 ( Figure 5D, arrowheads).
To verify the significance of the interaction between PUF2 and VPS9a in vacuolar trafficking, we tested the effect of overexpression of the N-terminal coiled-coil region of PUF2 on vacuolar transport of sporamin. When expressed in Arabidopsis suspension cultured cells, sporamin tagged with Venus predominantly accumulated in vacuoles, which was inhibited by co-expression of dominantnegative ARA7 (ARA7 S24N ) ( Figure 5-figure supplement 1B and D). The N-terminal coiled-coil of PUF2 exerted a similar effect to ARA7 S24N ( Figure 5-figure supplement 1B and D). This effect  should be attributed to titration of VPS9a by the PUF2 N-terminus, because the deleterious effect was suppressed by co-expressing VPS9a-tagRFP, but not by tagRFP alone (Figure 5-figure supplement 1C and D). Co-expression of GFP, GFP-tagged full-length PUF2, and the C-terminal coiled-coil region of PUF2 where PUF2 interacts with ARA6 did not exhibit an inhibitory effect on vacuolar Figure 5. Interaction between PUF2 and the RAB5 activating factor VPS9a. (A) Yeast two-hybrid interaction between PUF2 and VPS9a. PUF2 containing each coiled-coil region was expressed as a fusion protein with the AD, and VPS9a was expressed as a fusion with the BD in the yeast strain AH109. Interactions were tested using the HIS3 reporter gene. Black boxes indicate coiled-coil regions. (B) Plant lysates prepared from transgenic plants expressing PUF2-GFP (black arrowhead) or free GFP (white arrowhead) were subjected to immunoprecipitation with an anti-GFP antibody, followed by immunoblotting using the indicated antibodies. (C) The interaction between PUF2 and VPS9a was detected using an in vitro binding assay. Yeast lysate containing VPS9a-HA was subjected to a pull-down assay with GST-fused truncated PUF2 protein. VPS9a-HA interacted with PUF2 37-127 aa containing the N-terminal coiled-coil region but not with PUF2 461-639 aa containing the C-terminal coiled-coil region. VPS9a-HA did not interact with GST. Loading was performed with 1% input. trafficking of sporamin. These results indicated functional significance of the interaction of PUF2 with VPS9a at the N-terminal region in vacuolar transport in Arabidopsis.

PUF2 recruits VPS9a to endosomes
To investigate the effects of the puf2 mutation on VPS9a localization, we expressed VPS9a-GFP in the null alleles of the vps9a mutant vps9a-1 (Goh et al., 2007) and the puf2 vps9a-1 double mutant under regulation of the VPS9a promoter, which complemented VPS9a function by comparable expression levels of VPS9a-GFP ( Figure 6-figure supplement 1A). In root epidermal cells of puf2 vps9a-1, the endosomal population of VPS9a-GFP was significantly reduced compared to that in vps9a-1 cells [14.3 ± 1.9 puncta per cell slice in vps9a-1 (mean ±SD, n = 6 independent images, each of which contained five cells) versus 5.4 ± 2.1 puncta in puf2 vps9a-1 (mean ±SD, n = 6 independent images, each of which contained five cells), p<0.01, Student's t-test] ( Figure 6A). The remaining VPS9a-GFP-positive dots in puf2 vps9a-1 plants were sensitive to BFA treatment; however, endosome dilation was rarely observed after Wm treatment ( Figure 6A, middle and right panels), although VPS9a-positive puncta retained their endosomal identity as demonstrated by ARA6 localization ( Figure 6B). We did not observe a noticeable difference in VPS9a-GFP fluorescence in lateral root cap cells between the vps9a-1 and puf2 vps9a-1 mutants ( Figure 6-figure supplement 1B), likely reflecting distinct PUF2 requirements in different tissues.
According to the results described above, PUF2 promotes the recruitment of VPS9a onto the endosome. This was further confirmed using a transient expression system in protoplasts, in which fluorescently tagged VPS9a (VPS9a-tagRFP) failed to localize on endosomes with dispersed distribution in the cytosol ( Figure 6C, upper panels). When PUF2-GFP was coexpressed with VPS9a-tagRFP, VPS9a was recruited to PUF2-positive compartments ( Figure 6C, lower panels), whereas coexpression of another endosomal protein, GFP-VAMP727, did not affect the cytosolic localization of VPS9a ( Figure 6-figure supplement 1C). Thus, PUF2 facilitates the localization of VPS9a onto endosomes.

PUF2 is required for endosomal fusion induced by Wm
Reduced endosomal localization of VPS9a in puf2 may lead to inefficient activation of RAB5 on the endosomal membrane. This hypothesis was verified by examining Wm-induced endosomal fusion (Wang et al., 2009) in the puf2 mutant. Although the distribution of ARA6-GFP and mRFP-ARA7 was not markedly affected in DMSO-treated puf2 mutant cells ( Figure 6D, left panels), the diameters of the dilated endosomes induced by Wm treatment were significantly reduced by approximately 30% [1.98 ± 0.06 mm in wild-type (n = 83 dilated endosomes) versus 1.60 ± 0.03 mm in mutant cells (n = 206 dilated endosomes), mean ±SD, p<0.01, Student's t-test] ( Figure 6D, right panels). A histogram of the diameters of the dilated endosomes also showed a shift in the peak to the smaller population in the puf2 mutant ( Figure 6-figure supplement 1D). Thus, recruitment of VPS9a to the endosomal membrane by PUF2 is required for efficient RAB5 activation, which is likely necessary for Wm-mediated endosomal fusion.
Together, these results suggest that PUF2 promotes canonical RAB5-mediated endosomal transport by assembling VPS9a and GDP-bound canonical RAB5 on the endosomal membrane, thereby facilitating the activation of canonical RAB5 on the endosome. GTP-bound ARA6 negatively regulates this process by titrating PUF2, as shown by the effects of different concentrations of ARA6 Q93L on binding between GST-ARA7 S24N and PUF2. As the amount of ARA6 Q93L increased, the amount of PUF2 pulled down by GST-ARA7 S24N decreased (57.5 ± 34.1% band intensity, n = 4 independent experiments, p<0.05; Figure 7C and D, Figure 7-figure supplement 1A). ARA6 S47N and ARA7 Q69L , non-interactive RAB5 partners of PUF2, did not hamper the interaction between PUF2 and ARA7 S24N (Figure 7-figure supplement 1B). Thus, ARA6 in its active state interferes with the assembly of GDP-bound ARA7, PUF2, and likely VPS9a by competitively binding to PUF2 to diminish endosomal transport mediated by canonical RAB5.

Discussion
The existence of two RAB5 groups with distinct functions, specifically the canonical and plant-unique RAB5 groups, is a unique characteristic of the plant membrane trafficking system. These RAB5 groups share the upstream regulator VPS9a, containing the VPS9 domain, which is also conserved in activating factors for RAB5 in animal systems (Carney et al., 2006;Goh et al., 2007;Ishida et al., Figure 7. PUF2 integrates functions of two plant RAB5 groups. (A) Seed pods collected from wild-type, ara6-1 vps9a-2, and ara6-1 vps9a-2 puf2+/À plants. Arrowheads indicate brown and deflated seeds in a seed pod from the ara6-1 vps9a-2 puf2+/À plant. (B) Fourteen-day-old seedlings with the indicated genotypes. Bar = 1 cm. (C) Interaction between ARA7 S24N and PUF2 pre-incubated with ARA6 Q93L . The assay was performed as described for Figure 1B, but PUF2 was pre-incubated with 0 pmol, 25 pmol or 50 pmol ARA6 Q93L before mixing with GST-ARA7 S24N . Loading was performed with Figure 7 continued on next page 2016). In animal cells, several VPS9 domain-containing proteins with distinct domain structures regulate Rab5 activity in different endocytic trafficking processes. In plant cells, by contrast, the single activating factor VPS9a coordinates two plant RAB5 groups during vacuolar and endosomal transport, whose molecular mechanisms remain unknown. In the present study, we identified a novel ARA6 effector, PUF2, which acts as a keystone to integrate the functions of two plant RAB5 groups together with VPS9a on the endosomal membrane. PUF2 captures inactive canonical RAB5 and VPS9a on the endosomal membrane, leading to efficient activation of canonical RAB5 and the enhancement of vacuolar transport from endosomes to the vacuole. ARA6, the plant-unique RAB5 in Arabidopsis, diminishes this process by competitively binding to PUF2 in its GTP-bound state against GDP-bound canonical RAB5, thus balancing the distinct canonical RAB5-and ARA6-mediated endosomal transport pathways in plant cells ( Figure 7E).
No striking homologs of PUF2 have been observed in sequence databases of non-plant systems, suggesting that PUF2 was uniquely acquired by the plant lineage. However, PUF2 shares several characteristics with the animal Rab5 effector protein Rabaptin-5/RABEP1, and both proteins contain four coiled-coil regions. Furthermore, both PUF2 and Rabaptin-5/RABEP1 interact with GTP-bound Rab5 members at one of the coiled-coil regions and also interact with their activating factors at another coiled-coil region (Horiuchi et al., 1997). Rabaptin-5 directly binds to the active form of another Rab GTPase, Rab4, at a distinct coiled-coil region from Rab5, suggesting coordination of the two endosomal Rab GTPase functions by this molecule. Although these data suggest functional similarities between PUF2 and Rabaptin-5, PUF2 appears to preferentially interact with inactive canonical RAB5, and GTP-bound active ARA6 represses the action of canonical RAB5 via competitive binding to PUF2. Furthermore, PUF2 interacts with ARA6 and VPS9a at its C-terminal and N-terminal coiled-coil regions, respectively, whereas Rabaptin-5 interacts with Rab5 and Rabex-5 at its fourth and third coiled-coil regions, respectively (Mattera et al., 2006;Vitale et al., 1998). Rabaptin-5 promotes Rab5 activation by alleviating the autoinhibition of Rabex-5 (Delprato and Lambright, 2007;Lippé et al., 2001), resulting in additional Rab5 activation to form an active RAB5 domain on the endosomal membrane. Thus, the mechanisms utilized by PUF2 and Rabaptin-5 are completely different. The downregulation of a Rab GTPase by another Rab GTPase via competitive binding with an effector and a common GEF has not previously been reported. These findings demonstrate that plants have developed specific mechanisms to regulate RAB5 GTPases by acquiring two plantunique machinery components, ARA6 and PUF2, to participate in the underlying mechanism of endocytic pathway regulation in plants.
Although efficient recruitment of VPS9a to the endosome requires PUF2, it is not yet clear how PUF2 is targeted to the endosomal membrane. In the case of Rabaptin-5, Rabaptin-5 interacts with several proteins in addition to Rab5, Rab4, and Rabex-5, including Rabphillin-3 (a Rab3 effector: Ohya et al., 1998), g-adaptin, and GGA (Golgi-localizing, g-adaptin ear homology domain, ARFbinding proteins) (Mattera et al., 2003). The similarity between PUF2 and Rabaptin-5 at the amino acid sequence level is low, and their modes of actions are divergent, but PUF2 may also have other interacting partners because proteins with multiple coiled-coil regions frequently interact with multiple proteins. In the future, it would be interesting to isolate molecules responsible for the upstream regulation of plant RAB5 groups by recruiting PUF2 onto the endosomal membrane. Further screening and characterization of other binding partners of PUF2 is needed. We recently observed an effector of canonical RAB5 in Arabidopsis, EREX, which appears to act only during embryogenesis and early developmental stages (Sakurai et al., 2016), suggesting canonical RAB5 utilizes several distinct effector proteins at different developmental stages. ARA6 may also exert its function by interacting with multiple effector proteins, and the identification and analysis of other ARA6 effectors will allow us to understand the precise molecular mechanisms underlying ARA6 functioning. Such approaches will reveal how plants have pioneered plant-specific endosomal trafficking pathways and provide valuable information regarding the general mechanisms involved in membrane trafficking system diversification during the evolution of eukaryotic organisms. Yeast two-hybrid interaction assays Individual interaction assays were performed using GAL4 Two-Hybrid Phagemid Vector Kits (Agilent Technologies, Santa Clara, California, USA) and the AH109 strain (Takara Clontech, Kusatsu, Siga, Japan). The colonies were cultured in selective medium without leucine and tryptophan (designated as '+H'), diluted to OD600 = 0.5, and spotted onto plates containing media without leucine, tryptophan, and histidine (designated as '-H') as well as +H plates. More than three independent colonies were tested for each interaction.

Microscopy
Transgenic plants expressing GFP-, Venus-, and/or mRFP-tagged proteins were mounted in 1/ 2 Â MS liquid medium and observed under an LSM710, LSM780 (Carl Zeiss, Oberkochen, Germany), or a microscope (model BX51; Olympus, Shinjuku, Tokyo, Japan) equipped with a confocal scanner unit (model ORCA-AG; Yokogawa Electric, Musashino, Tokyo, Japan). At least three different seedlings from three independent transgenic lines were observed for microscopy. For drug treatments, 5 day-old seedlings were soaked in 1/2 Â MS liquid medium containing 3.3 mM wortmannin (Sigma-Aldrich, St. Louis, Missouri, USA) or 50 mM brefeldin A (Sigma-Aldrich) for two hours or one hour, respectively. For FM4-64 labeling, 5-day-old seedlings were treated with 4 mM FM4-64 (Thermo Fisher Scientific, Waltham, MA) at 23˚C for 30 min. The Feret's diameters of Wm rings were measured using ImageJ software (National Institutes of Health, Maryland, Washington, DC). Transient expression of GFP-and/or tagRFP-tagged proteins in Arabidopsis suspension cultured cells was performed as described previously (Ueda et al., 2001;Ueda et al., 2004). The Arabidopsis Col-0 suspension cultured line (Deep) was described in Mathur et al. (1998). The amounts of plasmids used for transformation were as follows: 10 mg for GFP, PUF2-GFP, PUF2 1-211 -GFP, PUF2 388-672 -GFP, and Sporamin-Venus subcloned in pHTS13, 30 mg for tagRFP, VPS9a-tagRFP, and tagRFP-ARA7 S24N subcloned in pHTS13, and 1 mg for GFP-VAMP727 subcloned in pUC18 Uemura et al., 2004). Ten different cells that were successfully transformed were observed. Wholemount visualization of embryos was performed as previously described (Aida et al., 1997). To monitor the vacuolar targeting of PIN2-GFP, transgenic plants were initially grown vertically for 4 days under constant light and subsequently incubated in the dark for 48 hr. Vacuolar accumulation of PIN2-GFP or sporamin-Venus was measured using Fiji (Schindelin et al., 2012). At least five different seedlings form each line were observed.
In vitro pull-down assay GST-RAB5s (0.2 nmol), pre-bound to glutathione sepharose 4B resin (GE Healthcare), were incubated in buffer A [20 mM Tris-HCl pH, 7.5, 150 mM NaCl, 5 mM MgCl 2 , and 0.05% Tween-20] containing 50 mM EDTA and 100 mM GTPgS or GDP and then incubated with PUF2 (0.266 nmol) in buffer A for 30 min at room temperature. The beads were washed three times with buffer A containing 10 mM GTPgS or GDP, and bound proteins were subjected to immunoblotting analysis. For an in vitro competition assay, the ARA6 Q93L , ARA6 S47N , or ARA7 Q69L were mixed with PUF2 and incubated in buffer A containing 10 mM GTPgS (for ARA6 Q93L and ARA7 Q69L ) or GDP (for ARA6 S47N ) prior to mixing with GDP-preincubated GST-ARA7 S24N . HA-tagged VPS9a was expressed in the yeast strain YPH414 (MATa Dpep4:TRP1 ura3 lys2 ade2 trp1 his3 leu2) under the control of the GAL1 promoter. Yeast cells were collapsed by vortexing with glass beads in PBS with protease inhibitor cocktail (GE Healthcare). The collected lysates (475 mg) were mixed with GST or GST-tagged truncate PUF2 (0.2 nmol), which were pre-bound to glutathione-Sepharose 4B resin (GE Healthcare), and incubated for 60 min at 4˚C in PBS buffer containing 0.05% Tween-20 and protease inhibitor cocktail (GE Healthcare). The beads were washed three times with the same PBS buffer, and the bound proteins were subjected to immunoblotting analysis. At least three independent experiments were performed.

Immunoprecipitation
T3 plants expressing PUF2-GFP were grown vertically on MS medium plates for 16 days, and 0.6 g of each sample was collected and ground in 1 ml of extraction buffer [50 mM HEPES-KOH, pH 7.5, 0.4 M sucrose, 5 mM MgCl 2 , protease inhibitor cocktail (Roche, Basel, Switzerland)] using sea sand. The lysates were centrifuged at Â1000 g to remove debris. For analysis using a chemical cross-linker, 1 mM DSP (Thermo Fisher Scientific) was added to the supernatants and incubated for 30 min at 4˚C, followed by quenching using 50 mM Tris-HCl, pH 7.5, and 0.5% CHAPS was added to the lysates to solubilize the membranes. Next, 750 ml of each sample was incubated with 50 ml of anti-GFP micro beads (Miltenyi Biotec, Bergisch Gladbach, Germany) for 30 min at 4˚C. The samples were loaded onto microcolumns attached to the magnetic field of a micro-MACS separator (Miltenyi Biotec) and washed four times with extraction buffer supplemented with 0.5% CHAPS. Immunoprecipitates were eluted according to the manufacturer's instructions, followed by immunoblotting. At least three independent experiments were performed.