The Arabidopsis thaliana T-DNA insertional mutants of atg2-1 (Inoue et al., 2006), atg5-1 (Thompson et al., 2005), atg7-2 (Chung et al., 2010), fyve2-2, and fyve3-1 (Kim et al., 2022) and transgenic plants expressing ProUBQ10:GFP-ATG8A (Kim et al., 2013) and ProVHA-a1:VHAa1-GFP (Dettmer et al., 2006) were previously described, and their progenies were used. ProUBQ10:YFP-ARA7, ProUBQ10:YFP-SYP32 (Geldner et al., 2009), and ProUBQ10:Citrine-2xFYVE (Simon et al., 2014) were obtained from the Arabidopsis Biological Resource Center (ABRC).

cDNA and expression clones were obtained from the ABRC with the following identifiers: Arabidopsis ATG6 (G19464), FYVE2 (G18828), FYVE3 (GC105091), SAR1B (G19464), SAR1C (G11541), pIX-Halo-ATG8F (HALO_SFI_23-A01), pIX-Halo-ATG8I (HALO_SFI_71-H06), PSAT4-DEST-n(1-174)EYFP-C1 (CD3-1089), PSAT5-DEST-c(175-END)EYFP-C1(B) (CD3-1097). To make entry clones containing FYVE3 cDNA lacking the stop codon and TRAF1A, FYVE3 and TRAF1A cDNA were amplified using primers listed in Supplementary Table 1 . The FYVE3 and TRAF1A PCR products were inserted into the pENTR/TOPO and pDONR221 vectors (Thermo Fisher Scientific), respectively. cDNA entry clones containing ATG8E, ATG8F, ATG8I, and ATG18A was generated as described previously (Kim et al., 2022).

To obtain binary vectors for transgenic plants expressing ProUBQ10:GFP-FYVE3 or ProUBQ10:mCherry-FYVE3, recombination of the entry clone containing FYVE3 cDNA with pMDC99-AtUBQ10p-GFP and pMDC99-AtUBQ10p-mCherry (Suttangkakul et al., 2011) was performed via the LR Clonase II reaction (Thermo Fisher Scientific). The binary vectors were introduced into Agrobacterium tumefaciens strain GV3101. To obtain stable transformants, Agrobacterium transformants were used to infect Arabidopsis using the floral dip method (Clough and Bent, 1998).

Seeds were surface-sterilized with 50%(v/v) bleach and then washed using sterile water. The seeds were germinated on 1× Murashige & Skoog (MS) solid medium [1× MS salt including vitamins (Duchefa, Haarlem, the Netherlands) with 1% (w/v) sucrose, 0.25% (w/v) phytagel (Sigma, St Louis, MO, USA)] or liquid medium with gentle shaking (100 rpm). Plants were incubated under long-day conditions (16 h light and 8 h dark) at 21-23°C.

To induce autophagy, the seedlings incubated in the liquid medium for 7 d were transferred to nitrogen-deficient medium (1 × MS micronutrient solution, 3 mM CaCl₂, 1.5 mM MgSO₄, 1.25 mM KH₂PO₄, 5 mM KCl and 1% sucrose, pH 5.7) and further incubated for indicated duration. For drug treatment, seedlings were treated with 0.5 µM AZD 8055 (LC Laboratories), 30 µM wortmannin (Sigma), and 0.5 µM Concanamycin A (Cayman).

Phenotypic analysis of the fixed carbon starvation response was performed as described (Chung et al., 2010). Briefly, after incubating seedlings on 1× MS-suc solid medium (1× MS salt including vitamins, 29 mM mannitol, and 0.25% phytagel) for 14 d, the seedlings were then deprived of light for 10 d. Seedlings resuming growth were counted after 18 d of recovery.

Seedlings frozen in liquid nitrogen were homogenized using a TNPI lysis buffer [50 mM Tris-Cl, 150 mM NaCl, 0.5% (w/v) sodium dodecyl sulfate (SDS), 0.5% (v/v) Triton-X 100, 5% (v/v) glycerol, 2 mM dithiothreitol, 1 mM phenylmethyl sulfonyl fluoride, and 10 mM iodoacetamide, pH 8.0]. Lysates were clarified by centrifugation at 16,000 × g at 4°C for 10 min, and the supernatant was mixed with 5 × SDS- polyacrylamide gel electrophoresis (PAGE) sample buffer [200 mM Tris-HCl (pH 6.8), 25% (v/v) glycerol, 10% (w/v) SDS, 10% (v/v) 2-mercaptoethanol]. The samples were boiled at 95°C for 10 min to ensure denaturation of protein samples and loaded into SDS-PAGE. The separated proteins were then transferred onto the Immobilon-P polyvinylidene fluoride membranes (Millipore, St Louis, MO, USA).

To prepare membrane proteins, seedlings expressing GFP-FYVE3 were homogenized with ice-chilled pestle and mortar in TNPI buffer [50 mM Tris-Cl, 150 mM NaCl, 1 mM phenylmethyl sulfonyl fluoride, 10 mM iodoacetamide, pH 8.0]. After centrifugation at 500 × g at 4°C for 5 min, 5% of the supernatant was set aside as total extract, and 95% of the supernatant was further separated into soluble and pellet fractions by centrifugation at 20,000 × g at 4°C for 15 min.

Western blot analysis was performed as previously described (Kim et al., 2022). Anti-GFP (Sigma-Aldrich 11814460001), Anti-H3 (Abcam ab1791), Anti-BiP (Agrisera AS09481), and Anti-UDP-glucose pyrophosphorylase (UGPase; Agrisera AS05 086) antibodies were used. The protein band intensity of the immunoblots was quantified by ImageJ (National Institute of Health).

To evaluate RNA abundance, seedlings were grown on 1 x MS medium for 10 d, frozen with liquid nitrogen, and homogenized in TRIsure agent (Bioline, London, UK). The RNA extract was treated with DNase I (New England Biolabs), and cDNA was synthesized using RevertAid H Minus reverse transcriptase (Thermo Fisher Scientific) with oligo(dT)₂₀ primers. Quantitative real-time PCR (qRT-PCR) was performed using Step One Plus Real-Time PCR System (Applied Biosystems) and EvaGreen (SolGent, Daejeon, Korea). The relative transcript abundance of target genes was calculated by the ΔΔCT method using the UBC9 cDNA as an internal control. All primer sequences are provided in Table S1 .

The entry clone containing FYVE3 cDNA was recombined with the destination vector pDEST22 (Thermo Fisher Scientific) via the LR Clonase II reaction. The resulting pDEST22-FYVE3 and pDEST32 expression vectors for bait proteins fused to SAR1B, SAR1C, ATG8A, ATG8E, ATG8F, ATG8I, ATG18A and FYVE2 (Kim et al., 2022) were used for transformation of the AH109 strain and Y187 strain through the Yeastmaker Yeast Transformation System 2 (Clontech, Mountain View, CA, USA). To determine protein interaction, yeast colonies isolated from Trp- and Leu- deficient medium were transferred to Trp-, Leu-, and His- deficient medium containing 0 and 1 mM 3-amino-1,2,4-triazole.

The expression vectors PSAT5-c(175-END)EYFP-C1(B)-SAR1B/ATG8A/ATG18A/ATG6 were described previously (Kim et al., 2022). To generate the expression vectors PSAT4-n(1-174)EYFP-C1-FYVE3 and PSAT5-c(175-END)EYFP-C1(B)-FYVE2/TRAF1A, entry clones containing FYVE3, FYVE2, and TRAF1A cDNA were recombined with the destination vectors PSAT4-n(1-174)EYFP-C1 and PSAT5-c(175-END)EYFP-C1(B), respectively, via the LR Clonase II reaction. The BiFC assay using Arabidopsis leaf protoplasts was performed as previously described (Kim et al., 2022).

To observe fluorescence signals from Arabidopsis plants or leaf protoplasts, either an LSM 510 or AxioObserver LSM 800 confocal microscope (Carl Zeiss) was used. For LSM 510, a 488 nm laser and BP500-530IR emission filter were used to detect GFP signals. For LSM 800, a 488 nm laser was used for excitation and fluorescence was detected at the 410-546 nm range to acquire GFP, citrine, and YFP signals. To detect mRFP and mCherry, a 587 nm laser was used for excitation, and fluorescence was detected at the 595-700 nm range.

Quantitative analysis of the confocal microscopy images was conducted ImageJ (NIH) as previously described (Kim et al., 2022). Data were presented as mean ± standard error. Student’s t-test was used to measure significance via VassarStats (http://www.vassarstats.net/).

ATG2 (At3g19190), ATG5 (At5g17290), ATG6 (At3g61710), ATG7 (At5g45900), ATG8A (At4g21980), ATG8E (At2g45170), ATG8F (At4g16520), ATG8I (At3g15580), ATG18A (At3g62770), RABF2b/ARA7 (At4g19640), FYVE2/CFS1 (At3g43230), FYVE3 (At1g29800), SAR1B (At1g56330), SAR1C (At4g02080), SYP32 (At3g24350), TRAF1A (At5g43560), UBC9 (At4g27960), and VHA-a1 (At2g28520).

To gain insight into the FYVE3 protein, we studied its physical interaction with other proteins. An Arabidopsis interactome study identified yeast two-hybrid (Y2H) interactions of FYVE3 with ATG8D (Arabidopsis Interactome Mapping Consortium, 2011). To identify additional interactors, we performed Y2H assay using FYVE3 as a prey and select ATG proteins, FYVE2, and SAR1 isoforms as baits. Our Y2H assay demonstrated interaction of FYVE3 with select ATG8 isoforms such as ATG8A and ATG8F but not with ATG8E and ATG8I ( Figure 1A ).

The positive Y2H interactions were further tested using in planta interaction assays using bimolecular fluorescence complementation (BiFC) in Arabidopsis leaf protoplasts. The BiFC interaction of NYFP-FYVE3 with CYFP-ATG8A was verified ( Figure 1B ), whereas no considerable fluorescence was reconstituted from negative controls of NYFP-FYVE3 paired with either CYFP or CYFP-TRAF1A (Qi et al., 2017) ( Figure 1B ). In addition, interactions of FYVE3 with SAR1B, ATG18A, ATG6, and FYVE2 were detected in BiFC experiments ( Figure 1B ), although these interactions were not apparent in Y2H ( Figure 1A ). We previously showed that FYVE2 interacts with SAR1B and ATG18A (Kim et al., 2022). Thus, FYVE3 may be brought into proximity of SAR1B, FYVE2, and ATG18A, for example, by their simultaneous binding to a scaffold protein. These data suggested that FYVE3 associates with ATG8A, FYVE2, SAR1B, ATG18A, and ATG6 in plant cells.

To investigate the subcellular distribution and trafficking of FYVE3 proteins, we prepared Arabidopsis transgenic plants expressing FYVE3 fused to either GFP or mCherry. The membrane fractionation experiment indicated that GFP-FYVE3 was primarily observed in the soluble fraction ( Figure S2 ). We then determined the subcellular distribution of mCherry-FYVE3 by co-localizing it with various organelle markers ( Figure 2 ). Root cells expressing mCherry-FYVE3 had diffuse fluorescence with relatively scarce punctate signals. The mCherry-FYVE3 puncta often overlapped with the puncta of the PI3P biosensor citrine-2xFYVE ( Figures 2A, G ), the late endosome marker YFP-ARA7 ( Figures 2B, G ), and the autophagic marker GFP-ATG8A ( Figures 2C, G ). However, relatively few mCherry-FYVE3 puncta overlapped with markers for the TGN ( Figures 2D, G ) and Golgi stacks ( Figure 2E, G ).

We also located mCherry-FYVE3 puncta relative to GFP-FYVE2 ( Figure 2F ). GFP-FYVE2 were mostly localized at cytoplasmic puncta, some of which were also decorated with mCherry-FYVE3. Although mCherry-FYVE3 puncta were fewer than GFP-FYVE2 puncta, the former mostly overlapped with the latter ( Figures 2F, G ). As FYVE2 was shown to localize on the autophagic membrane and, to a lesser extent, on the endosomal membrane (Kim et al., 2022), these data indicated that mCherry-FYVE3 puncta mainly represent either autophagic or endosomal compartments that are abundant in PI3P, whereas the diffuse signal corresponds to a cytosolic pool of mCherry-FYVE3.

The subcellular distribution of FYVE3 is consistent with their protein interaction with select ATG8 isoforms, suggesting that FYVE3 binds to ATG8 on phagophores during autophagosome biogenesis. As fluorescent fusions of FYVE3 are largely co-localized with autophagic puncta, we then examined whether these proteins are delivered together to the vacuole for degradation and detected as autophagic bodies. Observation of the autophagic bodies requires the treatment of GFP-ATG8 transgenic plants with concanamycin A (ConA), an inhibitor of vacuolar proton pumps. Transgenic plants expressing both GFP-ATG8A and mCherry-FYVE3 were treated with ConA, and their root cell images were compared with DMSO-treated controls ( Figure 3A ). In ConA-treated roots, 81% of the intravacuolar mCherry-FYVE3 bodies emitted a GFP-ATG8A signal ( Figure 3A ). ConA-treated roots expressing GFP-FYVE2 and mCherry-FYVE3 also showed a moderate level of co-localization in the vacuole ( Figure 3B ). Similarly, ConA-treated root cells expressing mCherry-FYVE3 alone showed intravacuolar puncta resembling autophagic bodies ( Figure S3A ). These data indicated that fluorescent fusions of FYVE3 are degraded in the vacuolar lumen, together with ATG8A and FYVE2.

To define the molecular mechanism underlying the vacuolar trafficking of FYVE3 fluorescent fusions, we tested the effect of ConA and the PI3K inhibitor wortmannin (Wm) treatment on the subcellular localization of GFP-FYVE3 ( Figure 3C ). As expected, autophagic body-like puncta were observed in the vacuole of ConA-treated wild-type (WT) seedlings expressing GFP-FYVE3 ( Figure 3C , WT, + ConA). Accumulation of the GFP-FYVE3 bodies was suppressed by Wm treatment ( Figure 3C , WT, + ConA, Wm), suggesting that GFP-FYVE3 is also targeted to the vacuole via PI3K-dependent trafficking. To test whether core ATG genes are involved in this trafficking, we introgressed the GFP-FYVE3 transgene into atg2-1 and atg5-1 background by a genetic cross. atg5-1 is defective in ATG8 lipidation (Chung et al., 2010), whereas atg2-1 accumulates a high level of lipidated ATG8 (Kang et al., 2018). When we observed their progenies by confocal microscopy, no autophagic body-like GFP-FYVE3 puncta were observed in ConA-treated atg5-1 and atg2-1 vacuoles ( Figure 3C ), demonstrating that canonical autophagy mediates the vacuolar targeting of GFP-FYVE3.

The vacuolar targeting of GFP-FYVE3 was validated by the immunoblot analysis using anti-GFP antibodies, similar to the GFP-ATG8 and GFP-FYVE2 cleavage assays (Kim et al., 2022). In WT, a very faint band corresponding to GFP-FYVE3 was detected, whereas free GFP released from GFP-FYVE3 was readily observed and its intensity increased by ConA ( Figure 3D ) and dark ( Figure S3B ) treatments, indicating that the free GFP band is a partial degradation product in the vacuole. In atg5-1 and atg2-1, the GFP-FYVE3 band intensity was increased, whereas ConA and dark treatments did not affect the intensity of protein bands (indicated by open diamonds in Figures 3D and S3B ) that migrated slightly slower than free GFP bands (solid arrowheads in Figures 3D and S3B ). Combined with confocal microscopy data ( Figure 3C ), these immunoblot results indicated that GFP-FYVE3 is delivered to the vacuole for degradation via canonical, ATG2- and ATG5-dependent autophagy pathway. Consistently, nitrogen starvation increased the ratio of free GFP to GFP-FYVE3 in WT but not in an atg7-2 background ( Figure S3C ). The intensity of the slower-migrating protein band appeared to increase by nitrogen starvation ( Figure S3C ). Although we do not know the nature of this protein band, the same phenomenon was observed in atg7-2 mutants expressing the autophagy marker GFP-ATG8 (Spitzer et al., 2015). These data indicated that the vacuolar trafficking of GFP-FYVE3 showed a typical pattern of the autophagy marker.

To define the molecular requirement for the formation of GFP-FYVE3 puncta, we examined the effect of atg5 and atg2 mutations and Wm treatment. The abundance of GFP-FYVE3 puncta decreases and increases in atg5 and atg2, respectively ( Figure S3D ). Again, Wm effectively diminished the accumulation of GFP-FYVE3 puncta in atg2 ( Figure S3D ). The PI3P-dependent accumulation of GFP-FYVE3 puncta in atg2 cytoplasm is very similar to that of GFP-ATG8A (Kang et al., 2018), and may result from arrested elongation/maturation of phagophores in atg2.

To investigate the effect of atg2 mutation on mCherry-FYVE3 to the autophagic membrane, we co-localized mCherry-FYVE3 with GFP-ATG8A in WT and atg2 background. atg2 root cells over-accumulated cytoplasmic mCherry-FYVE3 puncta, which extensively overlapped with GFP-ATG8A puncta ( Figure 3E ). Thus, ATG2 is essential for the targeting of mCherry-FYVE3 to the vacuole, but is dispensable for mCherry-FYVE3 recruitment to the phagophore.

To test the potential functions of FYVE3 in autophagy, we employed fyve3-1, a T-DNA insertional allele (Kim et al., 2022). RT-PCR analysis indicated that fyve3-1 is a knock-out allele, and there is no apparent transcriptional feedback between FYVE2 and FYVE3 genes ( Figure S1B ). Because of the FYVE and SYLF domains shared by FYVE2 and FYVE3, we initially postulated a genetic redundancy of FYVE2 and FYVE3. In fyve2 mutant background, phagophores are over-accumulated and autophagic flux is partially compromised (Kim et al., 2022). However, no significant change in autophagic flux was measured in fyve3 single mutants, and fyve2 fyve3 double mutants did not show further reduction in autophagic flux, when compared with fyve2 single mutants (Kim et al., 2022; Figure S4 ). These data were inconsistent with a simple model that FYVE2 and FYVE3 are duplicates. To further investigate the genetic relationship between these two genes, we quantified the abundance of autophagic organelles in N-starved WT, fyve2, fyve3, and fyve2 fyve3 double mutant cells expressing the autophagic marker GFP-ATG8A. Confocal microscopy analysis confirmed previous observation that fyve2 root cells over-accumulated GFP-ATG8A puncta in the cytoplasm, compared with WT ( Figure 4A ; Kim et al., 2022). As previously reported (Kim et al., 2022), Wm effectively blocked the formation of GFP-ATG8A puncta in fyve2 ( Figure 4A ). In contrast, fyve3 single mutants did not accumulate GFP-ATG8A puncta ( Figure 4A ). Notably, less than 10% of GFP-ATG8A puncta were detected in fyve2 fyve3 double mutants, relative to fyve2 mutants ( Figure 4A ). Complementation experiments revealed that the expression of mCherry-FYVE3 transgene in fyve2 fyve3 mutants restored the accumulation of GFP-ATG8A puncta ( Figure S5 ). These results indicate that FYVE3 is responsible for the over-accumulation of immature autophagic structures in fyve2.

Based on autophagy reporter analysis using fyve3 single and fyve2 fyve3 double mutants, we hypothesized that FYVE3 might serve a specific function during autophagosome formation. This hypothesis was tested using a different mutant combination of atg2 and fyve3 ( Figure 4B ). Compared with fyve2, atg2 mutants accumulate even more autophagic puncta, and autophagic flux is compromised almost completely (Kang et al., 2018; Kim et al., 2022). Unlike fyve2 fyve3 double mutants, atg2 fyve3 double mutant cells showed over-accumulation of GFP-ATG8A puncta as atg2 single mutants did (compare Figure 4B with Figure 4A ). As fyve3 did not suppress the over-accumulation of GFP-ATG8A puncta in atg2, FYVE3 appears to play a specific role in the formation of GFP-ATG8A puncta in fyve2. Of note, treatment with Wm only partially suppressed the GFP-ATG8A puncta in atg2 fyve3 double mutants ( Figure 4B ).

Further genetic analysis, including atg2 fyve2 fyve3 triple mutants, also confirmed that FYVE3 is specifically required for the accumulation of GFP-ATG8A puncta in fyve2 ( Figure 5 ). Under N starvation, atg2 fyve2 fyve3 triple mutants accumulated as many GFP-ATG8A puncta as atg2, atg2 fyve2, and atg2 fyve3 mutants ( Figure 5A ), masking the suppressive effects of fyve3 on fyve2 mutation. A similar trend was observed when seedlings were treated with AZD8055 (AZD; Figure 5B ), which induces autophagy by inhibiting Target Of Rapamycin kinase (Soto-Burgos and Bassham, 2017; Dauphinee et al., 2019).

Accumulation of autophagic puncta can result from either induction of autophagy or retarded biogenesis of autophagosomes. Autophagic flux analysis is necessary to distinguish these two possibilities. As fyve3 alone did not significantly affect autophagic flux (Kim et al., 2022; Figure S4 ), we assessed the effect of fyve3 on autophagic flux in atg2 and fyve2 mutant background. To quantify autophagic flux in different mutants, we used the GFP-ATG8A cleavage assay (Chung et al., 2010; Shin et al., 2014), where anti-GFP immunoblot analysis of seedlings expressing GFP-ATG8A is performed, and autophagic flux is determined from the protein band intensity of free GFP moiety, relative to that of GFP-ATG8A. GFP-ATG8A transgenic seedlings in WT, atg2, fyve2, and fyve3, or their double/triple mutant background were exposed to either N starvation ( Figure 5C ) or AZD ( Figure 5D ) and subjected to GFP-ATG8A cleavage assay. In both conditions, autophagic flux was fully inhibited in any genetic background containing atg2 and also greatly interfered in fyve2 ( Figures 5C, D ). In contrast, fyve2 fyve3 double mutant was indistinguishable from WT and fyve3, confirming that fyve3 suppresses fyve2 ( Figures 5C, D ).

Finally, the physiological function of FYVE3 gene expression was tested by phenotypic analysis of fyve3 single and fyve2 fyve3 double mutants. Autophagy-deficient mutants are hypersensitive to carbon starvation (Hanaoka et al., 2002; Chung et al., 2010). Our fixed-carbon starvation experiments showed that fyve2-2 seedlings showed hypersensitivity to light deprivation, whereas fyve3-1 did not ( Figure 5E ). Importantly, fyve2-2 fyve3-1 double mutants showed a WT-like sensitivity, demonstrating that the suppression effect of fyve3 on fyve2 is physiologically significant.