The Arabidopsis adaptor protein AP-3µ interacts with the G-protein β subunit AGB1 and is involved in abscisic acid regulation of germination and post-germination development

Heterotrimeric G-proteins (G-proteins) have been implicated in ubiquitous signalling mechanisms in eukaryotes. In plants, G-proteins modulate hormonal and stress responses and regulate diverse developmental processes. However, the molecular mechanisms of their functions are largely unknown. A yeast two-hybrid screen was performed to identify interacting partners of the Arabidopsis G-protein β subunit AGB1. One of the identified AGB1-interacting proteins is the Arabidopsis adaptor protein AP-3µ. The interaction between AGB1 and AP-3µ was confirmed by an in vitro pull-down assay and bimolecular fluorescence complementation assay. Two ap-3µ T-DNA insertional mutants were found to be hyposensitive to abscisic acid (ABA) during germination and post-germination growth, whereas agb1 mutants were hypersensitive to ABA. During seed germination, agb1/ap-3µ double mutants were more sensitive to ABA than the wild type but less sensitive than agb1 mutants. However, in post-germination growth, the double mutants were as sensitive to ABA as agb1 mutants. These data suggest that AP-3µ positively regulates the ABA responses independently of AGB1 in seed germination, while AP-3µ does require AGB1 to regulate ABA responses during post-germination growth.

Studies on loss-of-function alleles and gain-of-function overexpression lines of G-protein subunits suggest that the G-proteins modulate hormonal and stress responses, and play regulatory roles in many growth and developmental processes (reviewed by Jones and Assmann, 2004;Chen, 2008). agb1 mutants are characterized by aberrant leaf and flower shape, increased production of lateral root primordial, and shorter hypocotyls and siliques (Lease et al., 2001;Ullah et al., 2001Ullah et al., , 2003. Additionally, seed germination and early seedling development of agb1 mutants are hypersensitive to abscisic acid (ABA). Because plants lacking AGB1 have greater ABA hypersensitivity than plants lacking GPA1,AGB1 has been suggested to be the predominant regulator of G-protein-mediated ABA signalling (Pandey et al., 2006). ABA was shown to be bound by GTG1 and GTG2, which are Gα-interacting receptors on the plasma membrane (Pandey et al., 2009). A quantitative proteomics-based analysis of WT and gtg1gtg2 mutants revealed that the majority of ABA-responsive proteins require the presence of GTG proteins (Alvarez et al., 2013), supporting the importance of the G-proteins in ABA signal transduction.
The multiple phenotypes of agb1 mutants suggest that AGB1 is a key factor of several signalling pathways. So far some genetic and/or physical AGB1-interaction partners have been identified and characterized, for example a Golgilocalized hexose transporter SGB1 (Wang et al., 2006), an N-MYC downregulated-like1 (NDL1) (Mudgil et al., 2009), and an acireductone dioxygenase-like protein, ARD1 (Friedman et al., 2011). An interactome analysis revealed the involvement of G-proteins in cell wall modification (Klopffleisch et al., 2011). However, the molecular mechanisms underlying the AGB1-mediated signalling are unclear (Klopffleisch et al., 2011).
In Arabidopsis, each subunit of the AP-3 complex is encoded by a single-copy gene (Bassham et al., 2008). Loss-of function mutants of several subunits of the AP-3 complex have been shown to be the suppressors of zigzag1 (zig1), which is abnormal in both shoot gravitropism and morphology due to the lack of a vesicle trafficking regulator, SNARE VTI11 (Niihama et al., 2009). The AP-3 complex also plays a role in vacuolar function in Arabidopsis, including mediation of the transition between storage and lytic vacuolar identity (Feraru et al., 2010;Zwiewka et al., 2011). However, it is unclear whether the AP-3 complex also has roles in stress and hormonal responses.
Here we show that AP-3µ physically interacts with AGB1 in yeast and in vitro, as well as in planta. Genetic interaction between AP-3µ and AGB1 is also examined using agb1/ap-3µ double mutants.
Yeast two-hybrid analysis A yeast two-hybrid screen using AGB1 as a bait was performed as described previously (Tsugama et al., 2012a). The construct of pGAD-AP-3µ was generated as described in Supplementary Method S1.
To confirm the result of the yeast two-hybrid screen, pGBK-AGB1 and pGAD-AP-3µ were co-introduced into the Saccharomyces cerevisiae strain AH109. After transformation, at least four colonies grown on SD media lacking leucine and tryptophan (SD/-Leu/-Trp), were streaked on SD/-Leu/-Trp and SD media lacking leucine, tryptophan, and histidine (SD/-Leu/-Trp/-His).
GST-AP-3µ and GST-AP-3µ DN were induced and purified as described in Supplementary Method S3. GST-AP-3µ or GST-AP-3µ DN in the crude extracts was bound to Glutathione Sepharose 4 Fast Flow (GE Healthcare, UK) following the manufacturer's instructions, and the resin was washed four times with phosphate-buffered saline (PBS, 137 mM NaCl, 8.10 mM Na 2 HPO 4 .12H 2 O, 2.68 mM KCl, 1.47 mM KH 2 PO 4 , pH 7.4). After removing the PBS, the resin was resuspended in solution containing purified His-AGB1 and incubated at room temperature for 60 min with gentle shaking. The resin was then washed four times with PBS and resuspended in 20 mM reduced glutathione in 50 mM Tris-HCl pH 8.0. The suspension was incubated at room temperature for 15 min to release GST-AP-3µ or GST-AP-3µ DN . The slurry of the resin was centrifuged for a few minutes at 12 000 g. GST-AP-3µ or GST-AP-3µ DN and His-AGB1 in the supernatant were analysed by immunoblotting using an anti-GST antibody (diluted 4000-fold; GE Healthcare, UK) and HisProbe-horseradish peroxidase (HRP) (diluted 2000-fold; Thermo Fisher Scientific, USA). After the reaction of an anti-GST antibody, HRP-linked rabbit antibodies against goat IgG (diluted 5000-fold; MBL, Japan) were used as second antibodies. Signals were detected with SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific).

Bimolecular fluorescence complementation assay
To express cYFP (the C-terminal half of YFP, yellow fluorescence protein)-fused AP-3µ, the open reading frame (ORF) of AP-3µ was amplified by PCR using pGAD-AP-3µ as template and the following primer pair: 5′-CCGGTCTAGAATGCTTCAATGTATCTTTCTC-3′ and 5′-GGCGCCCGGGTACAACCTGACATCGAACTCACCAGC-3′ (XbaI and SmaI sites underlined). The PCR products were cloned into the SmaI site of pBluescript II SK -. The resultant plasmid was digested by XbaI, and the resultant ORF fragments of AP-3µ were inserted into the SpeI site of pBS-35SMCS-cYFP (Tsugama et al., 2012a), generating pBS-35S-AP-3µ-cYFP. To express nYFP (the N-terminal half of YFP)-fused AGB1, pBS-35S-nYFP-AGB1 (Tsugama et al., 2012a) was used. A mixture of an nYFP construct and a cYFP construct (500 ng each) was used for particle bombardment to co-express proteins of interest in onion epidermal cells. Particle bombardment and fluorescence microscopy were performed as previously described (Zhang et al., 2008). Images were processed using Canvas X software (ACD Systems).

Measurement of germination and greening rates
Germination and greening rates were compared between seed lots that were produced, harvested, and stored under identical conditions. Seeds were sown and grown as already described. Germination was defined here as emergence of the radicle through the seed coat. Cotyledon greening was defined as obvious cotyledon expansion and turning green. Germination and greening rates were scored daily for 9 days after seeds were transferred to the light at 22 °C. The experiments were repeated at least twice. The data shown are the means of all the experiments ± SD.

Semi-quantitative and quantitative real-time reverse-transcription PCR
The expression of AP-3µ mRNA in the wild type and the ap-3µ mutants was tested by semi-quantitative reverse-transcription (RT) PCR. Plants of each genotype were grown for 2 weeks and sampled. Total RNA was prepared using the GTC method (Chomczynski and Sacchi, 1987) and cDNA was synthesized from 4.6 µg of total RNA with PrimeScript Reverse Transcriptase (Takara, Japan) using an oligo(dT) primer. The primers used for the RT-PCR are shown in Supplementary Fig. S1A and the primer sequences are given in Supplementary Table S2. The expressions of the ABA-responsive genes (RAB18, RD29A, and AHG1) in the wild type and the ap-3µ mutants were tested by quantitative real-time RT-PCR. Plants of each genotype were grown for 18 days on 0.8% agar containing 0.5×MS salts 1% (w/v) sucrose, and 0.5 g/l MES, pH 5.8, with 0 or 1.0 µM ABA and sampled. Total RNA was prepared using RNeasy Plant Mini Kit (Qiagen, Netherlands) and cDNA was synthesized from 2 µg of the total RNA with High Capacity RNA-to-cDNA Kit (Applied Biosystems, USA) according to the manufacturer's instructions. The reaction mixtures were diluted 20 times with distilled water and used as a template for PCR. The primer sequences are given in Supplementary Table S2 (Nishimura et al., 2007;Tsugama et al., 2012b). Quantitative real-time RT-PCR was performed using SYBR Premix Ex Taq II (Perfect Real Time) (Takara) and the StepOne Real-Time PCR System (Applied Biosystems).

AP-3µ interacts with AGB1
To identify interacting partners of AGB1, we performed a yeast two-hybrid screen of the Arabidopsis leaf library using full-length AGB1 as bait. Even on high-stringency selection media (SD/ QDO), more than 3600 positive clones were obtained. Using yeast colony PCR with an AGG1-or AGG2-specific primer, we found that 60-70% of these clones expressed AGG1 (data not shown). Plasmid inserts from non-AGG1 clones were then amplified by colony PCR using a vector-specific primer pair, and sequenced. Around 400 clones were sequenced, and three of them expressed adaptor protein AP-3µ (GenBank accession: BX814222; At1g56590). Yeast cells could grow on the selection medium when both AP-3µ and AGB1 were present, but not when either of them was absent, indicating that AP-3µ and AGB1 interact with each other in yeast cells (Fig. 1A).
To confirm that AP-3µ binds directly to AGB1, we studied their interactions using a GST pull-down assay. His-AGB1 was detected only when it was reacted with AP-3µ, indicating that AGB1 and AP-3µ interact in vitro (Fig. 1B). Although the C-terminal 18 amino acids of AP-3µ are needed for recruiting cargo into the forming vesicle (Owen and Evans, 1998), they are not needed for the interaction with AGB1 because His-AGB1 was also detected when it was reacted with AP-3µ DN , which lacks the C-terminal 18 amino acids (Fig. 1B right).
In a bimolecular fluorescence complementation assay, YFP fluorescence was recovered in the cytosol and the nucleus when nYFP-AGB1 (AGB1 fused to the N-terminal half of YFP) and AP-3µ-cYFP (AP-3µ fused to the C-terminal half of YFP) were coexpressed (Fig. 1C), suggesting that they interact in the cytosol and nucleus in plant cells.

Subcellular localizations of AP-3µ and AGB1
When co-expressed in onion epidermal cells, GFP-fused AP-3µ (AP-3µ-GFP) was detected in the cytoplasm and nucleus, while mCherry-fused AGB1 (AGB1-mCherry) was detected in the cytoplasm, nucleus, and the plasma membrane (Fig. 2), suggesting the possibility that AP-3µ and AGB1 are co-localized in the cytoplasm and nucleus. This result is consistent with the above-described cytoplasmic and nuclear bimolecular fluorescence complementation between AP-3µ and AGB1 (Fig. 1C). ABA treatment did not affect the patterns of signals of either AP-3µ-GFP or AGB1-mCherry (data not shown).

ap-3µ mutants show ABA hyposensitive phenotypes in seed germination and post-germination growth
To examine the physiological role of AP-3µ in plants, we obtained two different mutant lines, ap-3µ-2 (SALK_064486C) and ap-3µ-4 (CS859652), which carry T-DNA insertions in intron 5 and exon 8 of the AP-3µ gene, respectively ( Supplementary Fig. S1A). Genomic PCR analyses verified that the T-DNA alleles were homozygous ( Supplementary Fig. S1B). RT-PCR using the primer combinations PF2+PR2, confirmed the absence of full-length transcripts (Supplementary Fig. S1A and C).
In the presence of 0.5-2.0 µM ABA, ap-3µ seeds germinated earlier than did wild-type seeds (Fig. 3A-D). The effect of ABA on the post-germination growth of seedlings was analysed by determining the percentages of seedlings with fully expanded green cotyledons (greening rate) at 0.5-2.0 µM ABA. Greening rates of ap-3µ seedlings in the presence of 0.5 and 1.0 µM ABA were higher than those of wild-type seedlings (Fig. 3E-G and Supplementary Fig. S2). On the contrary, agb1 mutants were hypersensitive to ABA during both germination and post-germination growth, as described previously (Pandey et al., 2006). In the presence of 2.0 µM ABA, the wild type and each mutant line were able to germinate, but none of them formed green cotyledons ( Fig. 3D and 3H). In the presence of ABA, which prevents the degradation of the seed storage proteins during germination (Garciarrubio et al., 1997), the basic subunit of 12S globulin, which is a seed storage protein, degraded faster in ap-3µ mutant seedlings than in wild-type seedlings. In contrast, the basic subunit of 12S globulin was most preserved in agb1 mutants ( Supplementary  Fig. S3). These results suggest that the ap-3µ mutants are less sensitive to ABA than the wild type. However, no difference between wild type and ap-3µ-4 mutant was observed in the inhibition of root growth by ABA (Supplementary Fig. S4).
We investigated the expression profiles of RAB18, RD29A, and AHG1, which are ABA-induced marker genes. ABAinduced gene expression was reduced in ap-3µ mutants, as determined by the transcript levels of the marker genes (Fig. 4). No effect of ABA on expression of AP-3µ transcripts was observed. The expression of AGB1 in the wild type did not change upon ABA treatment, while the expression of AGB1 in ap-3µ mutant was upregulated and higher than that in the wild type in the presence of ABA (Fig. 4 left).
ABA also has roles in the responses to environmental stresses, including desiccation and high salinity (Busk and Pagès, 1998;Leung and Giraudat, 1998). However, when seeds and seedlings were exposed to various osmotic stresses (400 mM mannitol, 150 mM NaCl, or 9.2% polyethylene In vitro GST pull-down assay. GST-fused AP-3µ (GST-AP-3µ) or GST-fused AP-3µ DN (GST-AP-3µ DN ), respectively and His-tagged AGB1 (His-AGB1) were expressed in Escherichia coli and used for the analysis. The presence or absence of each protein in the reaction mixture is shown as + or -, respectively. Experiments were performed four times and a representative result is shown. Antibodies used for immunoblotting are shown as IB:His and IB:GST. (C) Bimolecular fluorescence complementation in onion epidermal cells. The ORF of AGB1 was cloned in frame behind the coding sequence of the N-terminal region of YFP (nYFP) to express nYFP-fused AGB1 (nYFP-AGB1), and the ORF of AP-3µ was cloned in frame in front of the coding sequence of the C-terminal region of YFP (cYFP) to express cYFP-fused AP-3µ (AP-3µ-cYFP). Both constructs were introduced into onion epidermal cells. cYFP alone and nYFP alone were used as controls. More than 20 cells were observed and a representative cell is shown. Bars=50 µm (this figure is available in colour at JXB online). glycol), no difference was observed between the wild type and ap-3µ with respect to seed germination, seedling growth, or seedling development (Supplementary Figs. S5, S6, and S7). These data suggest that AP-3µ is not involved in the responses to either osmotic stress or salt stress.
In the presence of 0.25 µM ABA, the germination rates of all the double mutants were similar to the germination rate of agb1-1 mutant (Fig. 5B). In the presence of 0.5 µM ABA, the germination rates of all the double mutants were higher than the germination rate of the agb1-1 mutant (Fig. 5C), suggesting that AP-3µ positively regulates the ABA response independently of AGB1 in seed germination. In the presence of 0.25 µM ABA, the greening rate of DM1-5-3 was significantly higher than the greening rate of agb1-1 mutant only at day 6, while no significant difference was observed between DM2-8-5-5 and agb1-1 mutant in their greening rates at any time points ( Fig. 5E; see Supplementary Fig. S9E for t-test in comparison between agb1-1 mutant and each genotypes). In the presence of 0.5 µM ABA, cotyledon greening was strongly inhibited in both the double mutants and agb1-1 mutant ( Fig. 5F; see Supplementary Fig. S10 for growth phenotypes in the presence of ABA). And the greening rate of DM1-5-3 was significantly but only slightly higher than the greening rate of agb1-1 mutant at day 9, while no significant difference was observed between DM2-8-5-5 and agb1-1 mutant in their greening rates at any time points (Supplementary Fig. S9F). These results suggest that the AP-3µ-dependent alleviation of the effects of ABA is at least partially dependent on AGB1 at the post germination stage.
Although agb1 mutants have an increased number of lateral roots (Ullah et al., 2003), the numbers of lateral roots were not significantly different between the wild type and ap-3µ-4 mutant in the presence of 0 and 2 µM ABA. Similarly, the numbers of lateral roots were not different between agb1-1 mutant and agb1/ap-3µ double mutants ( Supplementary Fig. S11), suggesting that AP-3µ is not involved in regulating lateral root formation. Although lateral root formation can be controlled by auxin (Fukaki et al., 2007 for review) and AGB1 is known to be involved in the auxin-dependent control of lateral root formation (Ullah et al., 2003), the ap-3µ mutants and the wild type did not differ in their responses to an auxin, indole acetic acid, and an auxin-transport inhibitor, N-(1-naphthyl)phthalamic acid (data not shown). These results suggest that AP-3µ is not involved in the control of lateral root growth by auxin.

Mutants of AP-3δ subunit and clathrin heavy chain (CHC) show ABA-hyposensitive phenotypes in post-germination growth
The ap-3δ and chc1 mutants harbour T-DNA insertions in exon 1 of the AP-3δ gene and exon 24 of the CHC1 gene, respectively ( Supplementary Fig. S12). Genomic PCR analyses confirmed that the T-DNA plants were homozygous (Supplementary Fig.   Fig. 2. Subcellular localizations of AP-3µ and AGB1. GFP-fused AP-3µ (AP-3µ-GFP) and mCherry-fused AGB1 (AGB1-mCherry) (A) or GFP alone and mCherry alone (B) were transiently co-expressed in onion epidermal cells under the control of 35S promoter. More than 10 cells were observed and a representative cell is shown in each panel. Bars=50 µm (this figure is available in colour at JXB online). S12). RT-PCR using primers specific for AP-3δ confirmed the absence of transcripts in ap-3δ (Supplementary Fig. S12A) and RT-PCR using primers specific for CHC1 confirmed the absence of transcripts in chc1 (Supplementary Fig. S12B). In the presence of 1.0 µM ABA, the rates of seed germination in ap-3δ and chc1 were significantly but only slightly different from that in the wild type (Fig. 6B). However, in the greening test, only 23% of wild-type seedlings developed green cotyledons on day 10 at 1.0 µM ABA, whereas about 43% of the ap-3δ mutant seedlings and 50% of the chc1 mutant seedlings developed green cotyledons (Fig. 6D). These results suggest that AP-3δ and CHC, as well as AP-3µ, function in the ABA response during post-germination growth.

AP-3µ interacts with AGB1 and negatively regulates AGB1
We have shown that AP-3µ both physically and genetically interacts with AGB1 and regulates the ABA-dependent seed germination and cotyledon greening. Because AGB1 Fig. 3. Seed germination and post-germination development of ap-3µ mutants are hyposensitive to ABA. (A-D) Germination rates of the wild-type (WT) seeds and agb1-1, agb1-2, ap-3µ-2, and ap-3µ-4 mutant seeds in the presence of 0 (A), 0.5 (B), 1 (C), or 2 µM ABA (D). Germinated seeds were counted at the indicated time points. (E-H) Percentages of seedlings with fully expanded green cotyledons (greening rates) of WT and agb1-1, agb1-2, ap-3µ-2, and ap-3µ-4 mutants in the presence of 0 (E), 0.5 (F), 1 (G), or 2 µM ABA (H). Seedlings with fully expanded green cotyledons were counted at the indicated time points. The experiment was repeated three times and data were averaged. n=20/genotype for each experiment. Error bars represent SD. *P<0.05, **P<0.005 as determined by t-test in comparison between wild type and each mutant.
is a negative regulator of ABA responses (Pandey et al., 2006), and because AP-3µ-dependent positive regulation of ABA responses during post-germination growth requires AGB1 (Fig. 5 and Supplementary Fig. S9), AP-3µ is thought to be an upstream negative regulator of AGB1 in the suppression of the inhibition of post-germination growth by ABA (Fig. 7). Although no information about the physical interaction between AGB1 and AP-3µ was available in Arabidopsis G-Signalling Interactome Database (AGIdb, http://bioinfolab.unl.edu/AGIdb), our results strongly support the idea that AP-3µ participates in the AGB1-mediated signalling.
Although ABA is known to be involved in acquiring tolerances to osmotic stress and salt stress, no difference was observed between the wild type and ap-3µ in osmotic stress or salt stress treatments (Supplementary Figs. S5,S6,and S7). These data suggest that AP-3µ is not involved in the responses to either osmotic stress or salt stress. Osmotic stresses can retard plant growth independently of ABA, because osmotic stresses inhibit cellular water uptake. In the case of salt stress, ion toxicity can also inhibit plant growth. It is possible that those ABA-independent plant growth inhibitions were much more significant than the ABA-mediated plant growth inhibition in our experiments in which the plants were subjected to osmotic/salt stresses.
In mammals, G-protein-coupled receptors are internalized to desensitize in response to excessive and/or continuous stimuli (Lefkowitz, 2004). An animal G-protein-coupled receptor, β2 adrenergic receptor, has been suggested to be internalized via clathrin-mediated endocytosis when it binds its ligand (Ferguson et al., 1996;Schmid et al., 2006;McMahon and Boucrot, 2011 for review). The classical function of clathrinmediated endocytosis in the regulation of signal transduction is to terminate the signal by physically removing activated receptors from the cell surface (Sorkin and von Zastrow, 2009;Scita and Di Fiore, 2010). The internalization of ligand-receptor complexes into endosomes and then lysosomes may lead to their degradation, which results in termination of signalling. In plants, the internalization of AtRGS1 (regulator of G-protein signalling 1), which is the prototype of a seven-transmembrane receptor fused with an RGS domain, was reported (Urano et al., 2012). AtRGS1 is known to be internalized when cells are treated with sugars such as d-glucose. Endocytosis of AtRGS1 physically uncouples the GTPase-accelerating activity of AtRGS1 from GPA1, permitting sustained activation of G-protein signalling on the plasma membrane (Urano et al., 2013 for review). It is unclear whether the internalization of AtRGS1 is dependent on clathrin. Because AP-3µ is a component of a clathrin complex and interacts with AGB1, it will be interesting to examine whether AP-3µ is involved in the internalization of AtRGS1. Alternatively, it is possible that AGB1 is a direct target of the clathrin-mediated endocytosis. However, in either the presence or the absence of ABA, no difference was observed in the patterns of GFP-fused AGB1 (GFP-AGB1) signals between the wild type and ap-3µ-4 mutant ( Supplementary Fig. S13). It is possible that AP-3µ is involved in AGB1 internalization, but at least it could not be detected in this transient expression experiment. The level of AGB1, which negatively regulates ABA responses, might be higher in the absence of AP-3µ than in its presence, and this may be why the ap-3µ mutants showed hyposensitivities to ABA (Figs. 3 and 4). To our knowledge, this study is the first article reporting possibility of internalization of β subunit of G-protein in plants. However, further studies are required to elucidate whether AP-3µ is involved in endocytosis of AGB1 and other components of G-protein signalling. The finding that the numbers of lateral roots were not significantly different between the wild type and ap-3µ-4 mutant in either the absence or the presence of ABA ( Supplementary Fig. S11), indole acetic acid, or N-(1naphthyl)phthalamic acid (data not shown) suggests that AP-3µ does not function in regulating lateral root formation or in the control of lateral root growth by auxin. Therefore, the interaction between AP-3µ and AGB1 seems not to be involved in the control of lateral root formation and growth. In addition to the involvement in ABAdependent inhibition of post-germination growth, the interaction between AP-3µ and AGB1 may be required in other processes. AGB1 mediates developmental processes and hormone responses. In addition to showing altered sensitivities to ABA and auxin, agb1 mutants show altered sensitivities to gibberellin (Chen et al., 2004), brassinosteroid (Chen et al., 2004;Tsugama et al., 2013), and jasmonic acid (Trusov et al., 2006).

AP-3 complex and clathrin are involved in ABA regulation of post-germination development
AP-3 exists in Arabidopsis as a complex (Zwiewka et al., 2011). The CHC is also associated with AP-3β (Zwiewka et al., 2011). AP-3β-GFP was found to localize predominantly in the cytoplasm (Feraru et al., 2010). AP-3µ is present in the cytoplasm and nucleus ( Fig. 2A). Each component of the AP-3 complex plays similar roles in regulating biogenesis and the functions of vacuoles in plants (Feraru et al., 2010;Zwiewka et al., 2011). Furthermore, ap-3µ, ap-3δ, and ap-3β all suppress the shoot gravitropism abnormality of the zig1/vti11 mutant, which is defective in protein trafficking to the vacuoles (Sanmartín et al., 2007;Niihama et al., 2009). Similar phenotypes of the mutants defective in the different subunits of the same AP-3 complex suggest that these proteins act in the same process, possibly in the same complex. Also, the post-germination growth of the ap-3µ, ap-3δ, and chc1 mutants were hyposensitive to ABA (Fig. 6D), supporting the idea that each subunit of AP-3 complex acts in the same process, probably mediating clathrin-based trafficking. However, the hyposensitivity to ABA during post-germination growth was greater in the ap-3µ mutants than in the ap-3δ and chc1 mutants (Fig. 6D) and the rates of seed germination at 1 µM ABA in ap-3δ and chc1 were significantly but only slightly different from that in the wild type (Fig. 6B). One possible explanation for these observations is that the homologue genes are redundant. The Arabidopsis genome encodes two CHCs that have 97% amino acid sequence identity (Kitakura et al., 2011). The homologues of AP-3δ in other AP complexes may compensate for the loss of AP-3δ. Another possibility is that, although each subunit of the AP-3 complex acts in the same process in the ABA response during post-germination growth, AP-3µ is the predominant regulator in the process.
To our knowledge, this study is the first report on the involvement of AP-3 complex and clathrin in the regulation of post-germination growth by ABA. Further studies are needed to understand how the AP-3 complex and clathrin are involved in the ABA regulation of post-germination growth.

Supplementary material
Supplementary data are available at JXB online.
Supplementary Table S1. Primer pairs used for genomic PCR.
Supplementary Table S2. Primer pairs used for RT-PCR analyses. Supplementary Fig. S1. Identification of ap-3µ T-DNA insertional mutants. Supplementary Fig. S2. ap-3µ mutants are hyposensitive to ABA in post-germination growth. Supplementary Fig. S3. The degradation of seed storage proteins occurs faster in ap-3µ mutants than in the wild type in the presence of ABA. Supplementary Fig. S4. No difference between wild type and ap-3µ-4 mutant was observed in the inhibition of root growth by ABA. Supplementary Fig. S5. Responses of ap-3µ mutants to osmotic and salt stresses. Supplementary Fig. S6. Germination rates of wild-type seeds and agb1-1 and ap-3µ-4 mutant seeds in the presence of 400 mM mannitol or 9.2% polyethylene glycol. Fig. 7. Schemes of AP-3µ modes of action. Contrary to AGB1, AP-3µ positively regulates the inhibition of seed germination and post-germination growth by ABA. AGB1 and AP-3µ function independently in ABA regulation of seed germination, but AP-3µ is a negative regulator of AGB1 in ABA regulation of postgermination growth. AP-3µ seems to function in these processes with other subunits of AP-3 complex mediating clathrin-based trafficking. AP-3, AP-3 complex; CHC, clathrin heavy chain. Germination rates (A and B) and greening rates (C and D) of wild type and ap-3µ-4, ap-3δ, and chc1 mutants in the absence of ABA (A and C) or in the presence of 1.0 µM ABA (B and D) over time (days after stratification). The experiment was repeated three times for wild type and ap-3µ-4 and ap-3δ mutants, and twice for chc1 mutant. Data were averaged; n=70/genotype for each experiment. Error bars represent SD. *, P<0.05, **P<0.005 as determined by t-test in comparison between wild type and each mutant. Supplementary Fig. S7. Greening rates of wild type and agb1-1 and ap-3µ-4 mutants in the presence of 400 mM mannitol or 9.2% polyethylene glycol. Supplementary Fig. S8. Generation of agb1/ap-3µ double mutants. Supplementary Fig. S9. T test for germination rates and greening rates in comparison between agb1-1 mutant and each agb1/ap-3µ double mutants. Supplementary Fig. S10. agb1/ap-3µ double mutants display ABA-hypersensitive phenotype in post-germination growth similar to that of agb1 mutants. Supplementary Fig. S11. Numbers of lateral roots of wild type, agb1-1, ap-3µ-4, and agb1/ap-3µ double mutants in the absence or in the presence of ABA. Supplementary Fig. S12. T-DNA insertional mutants of AP-3δ and CHC1. Supplementary Fig. S13. Subcellular localization of AGB1 in wild type and ap-3µ mutant.