Plant-specific mitochondrial fission machinery connects mitochondrial dynamics and mitophagosome formation for piecemeal mitophagy

Significance

Mitophagy, the selective autophagic degradation of mitochondria, plays a vital role in maintaining mitochondrial homeostasis. Unlike the typically networked mitochondria in mammals, plant mitochondria are often physically discrete. Although wholesale degradation of mitochondria has been reported in plants, the contribution of plant mitochondrial fission activity to mitophagy and its physical relevance remains unexplored. Here, we demonstrate that dysfunction in the plant-specific mitochondrial fission machinery delays the completion of mitophagosomes and leads to the accumulation of megamitochondria sequestered by multiple phagophore precursors marked with ATG8 and the selective autophagy receptor NBR1. Our findings underscore the importance of mitochondrial fission machinery in heat-induced piecemeal mitophagy across evolutionarily diverse organisms, providing molecular insights into mitophagosome biogenesis for plant stress resilience.

Abstract

As the energy center of the cell, mitochondria display enormous metabolic plasticity to meet the cellular demand for plant growth and development, which is tightly linked to their structural and dynamic plasticity. Mitochondrial number and morphology are coordinated through the actions of the mitochondrial division and fusion. Meanwhile, damaged mitochondrial contents are removed to avoid excess toxicity to the plant cells. Mitophagy, a selective degradation pathway of mitochondria through a double-membrane sac named autophagosome (also known as mitophagosome), plays a crucial role in maintaining mitochondrial homeostasis. Typically, wholesale mitophagy requires the elongation of a cup-shaped phagophore along the entire mitochondrion, which finally seals and closes as a mitophagosome. How plant mitophagosome formation and mitochondria sequestration are coordinated remains incompletely understood. In this work, we report an unappreciated role of the plant-specific mitochondrial fission regulator ELM1, together with the dynamin-related protein family DRP3 and the autophagic regulator SH3P2, to coordinate mitochondria segregation for piecemeal mitophagy under heat stress conditions. Dysfunction in mitochondrial fission activity impairs heat-induced mitophagy, leading to an accumulation of interconnected megamitochondria which are partially sequestered by the ATG8-positive phagophore. Furthermore, we show that the ELM1-mediated piecemeal mitophagy also engages the plant archetypal selective autophagic receptor NBR1. Using 3D tomography analysis, we illustrate the morphological features and spatial relationship of the megamitochondria and phagophore intermediates in connection with the mitochondrial fission sites. Collectively, our study provides an updated model of mitophagosome formation for piecemeal mitophagy mediated by the plant-unique mitochondrial fission machinery.

As the major powerhouse in almost all eukaryotic cells, dysfunction in mitochondrial quality control will cause detrimental effects to cells, leading to cellular toxicity or even cell death. Mitochondrial quality control is maintained through efficient biogenesis and the removal of damaged mitochondrial components or entire mitochondria. Extensive studies have demonstrated that autophagy is required for selective degradation of mitochondria, also known as mitophagy (1–3). Autophagy is initiated by forming a membrane sac named phagophore, which finally seals into a double-membrane autophagosome (4, 5). During this process, the ubiquitin-like protein autophagy-related 8 (ATG8, also known as LC3 in mammalian cells) is conjugated to the phagophore membrane through a series of ubiquitin-like process, which is also termed as ATG8 lipidation (6). In addition, proteins harboring an ATG8-interacting motif (AIM) interact with ATG8, which has been implicated to play a central role in linking cargo receptors/substrates to the growing phagophore to ensure selective autophagy in different organisms (7–9). A subset of damaged mitochondrial substrates are ubiquitinated and recognized by selective autophagy receptors (SARs), namely p62, which further recruit the mammalian ATG8 homolog LC3, followed by the phagophore growth on the mitochondrial surface for the sequestration of the damaged mitochondrion by the phagophore(10, 11). Receptor-mediated mitophagy can be processed independent of ubiquitination. In yeast, Atg32 has been first demonstrated as a receptor that binds to Atg8 to activate mitophagy (12, 13). In animal cells, a number of receptors have been reported to participate in targeting the mitochondrial contents for degradation via interaction with LC3. Several mitochondrial outer membrane (OMM) proteins, namely yeast Atg32 homolog BCL2L13 (Bcl-2-like protein 13, also known as BCL-Rambo,), BNIP3 (Bcl-2 interacting protein 3), BNIP3L (BCL2 interacting protein 3 like, also known as NIX), FKBP8/FKBP38 (FK506-binding protein 8/38), and FUNDC1 (FUN14 Domain Containing 1), undergo different posttranslational modifications, which trigger their direct binding to LC3 (14–18). On the other hand, certain stress conditions stimulate the release of the mitochondrial inner contents (proteins or lipids) that directly bind to LC3 or other ATG proteins independent of ubiquitination, which also initiates the development of phagophore along the mitochondrial surface for wholesale mitophagy (19, 20). Alternatively, damaged part of a mitochondrion can be removed by piecemeal mitophagy to maintain the healthy portion (21, 22). Notably, several studies have suggested the roles of mitochondrial fission factors in piecemeal mitophagy for separation of the damaged mitochondrial contents to facilitate their sequestration into the autophagosome (23–25).

Mitochondria display conserved morphological features in both mammals and plants (26, 27). However, unlike the typical network pattern in the majority of the mammalian cells, plant mitochondria are often discrete and highly mobile. Wholesale mitophagy for degradation of an entire mitochondrion has been implicated to function in different plant developmental stages, from seed germination to environmental adaptation (28–30). Nevertheless, the underlying molecular mechanism for mitophagosome formation remains largely unexplored in plants. A few plant-unique mitophagy regulators have been demonstrated to function in the turnover of depolarized mitochondria, including ATG11, Friendly (also known as FRIENDLY MITOCHONDRIA, FMT), FCS-like zinc finger proteins (FLZ) and TRB family proteins (31–35). Our previous study showed that SH3P2, which belongs to the Bin/Amphiphysin/Rvs (BAR) domain and Src homology 3 (SH3) containing protein family, harbors an AIM-like motif for binding with ATG8 to promote autophagy (36, 37). Of note, we observed that upon autophagic induction, constricted mitochondria are sequestered by the SH3P2-positive ring-like structures, indicating a fission process occurs during this event. We further showed that SH3P2 is associated with the plant-specific mitochondrial fission factor Elongated Mitochondria 1 (ELM1), as well as the dynamin-related protein DRP3. Intriguingly, we found that heat stress induces morphological reshaping of mitochondria into tubules and promotes the mitochondrial localization of SH3P2. Furthermore, deficiency in mitochondrial fission activity impairs piecemeal mitophagy and mitophagosome formation, leading to the accumulation of megamitochondria enclosed by ATG8e- and NBR1- positive phagophore fragments. Using live cell imaging and 3D tomography analysis, we revealed the temporal and spatial relationship between dysfunctional mitochondria and developing phagophore intermediates. Taken together, this study unravels an intimate interplay between mitochondrial fission and phagophore assembly for effective piecemeal mitophagy to promote plant heat resilience.

Results

SH3P2 Forms a Complex with the Plant-Specific Mitochondrial Fission Factor ELM1.

As SH3P2 has been shown to participate in autophagy and interact with ATG8 family protein (36, 37), we wondered whether SH3P2 participates in mitophagy for selective degradation of mitochondria. A previous study has shown that plant mitophagy can be induced after uncoupler treatment (31). We therefore crossed SH3P2-GFP with a mitochondrial reporter (Mito-mCherry) and subjected to confocal analysis with uncoupler treatment. Notably, we observed that upon uncoupler treatment, SH3P2 proteins localize to mitochondria and occasionally deposited on the mitochondrial constriction site (SI Appendix, Fig. S1A). We therefore conducted a time-lapse imaging analysis to reveal their dynamic association (Movies S1 and S2). As shown in Fig. 1A and SI Appendix, Fig. S1B, we observed a mitochondrial fragment is pinched off and engulfed by a cup-shaped structure labeled by SH3P2-GFP, while the opening of the phagophore is ultimately sealed toward to the mitochondrial constriction site. After phagophore closure, the SH3P2-GFP-positive mitophagosome together with the mitochondria fragment was detached from the remaining mitochondrial body and moved away (SI Appendix, Fig. S1B and Movie S3). The observation of dual localizations of SH3P2 on mitochondrial constriction sites and mitophagosomes suggests a possible link between SH3P2 and mitochondria. To identify the mitochondria-related proteins for SH3P2, we performed a GFP-trap coupled with mass spectrometry using GFP-tagged SH3P2 transgenic plants. Surprisingly, the plant-specific mitochondrial fission regulator ELM1 and the dynamin family protein DRP3B were identified (38). ELM1 was initially identified as a mitochondrial fission factor by genetic screening for the elongated mitochondria in Arabidopsis. Both DRP3B and its homolog DRP3A have been implicated to function as the counterpart of DRP1 in animals for mitochondria and peroxisome fission (39–41).

Fig. 1.

SH3P2 localizes to mitophagosome-like structure and associates with the plant-specific mitochondrial fission factor ELM1. (A) Subcellular analysis of SH3P2-GFP with mitochondria upon mitochondria uncoupler treatment. 5-d-old Arabidopsis seedlings expressing SH3P2-GFP/Mito-mCherry were treated with 50 μM DNP for 1 h, followed by 0.5 h recovery before confocal observation. Area indicated by the dashed square is subjected to montage analysis and shown at the Bottom. Arrowheads indicate the constriction site of a mitochondrion enclosing by the SH3P2-positive phagophore-like structure. Quantification of the Mander’s colocalization coefficients between SH3P2-GFP and Mito-mCherry is shown on the Right. M1, fraction of SH3P2-GFP signal that overlaps with Mito-mCherry signal. M2, fraction of Mito-mCherry signal that overlaps with SH3P2-GFP signal. Bars indicate the mean ± SD of 10 replicates. (Scale bar, 5 µm.) (B) BiFC analysis of SH3P2 and ELM1 in tobacco leaf cells. SH3P2-nYFP and ELM1-cYFP were cotransformed in tobacco leaf cells, or with empty nYFP and cYFP, respectively, as control. IDH1-mRuby was coexpressed to indicate the transformed cells and mitochondria. (Scale bar, 50 µm.) Statistical analysis of colocalization between YFP-positive foci and IDH1-mRuby signal per cell in the BiFC assay is shown on the Right. At least three different leaves were selected, and more than 15 images were taken for analysis. One-way ANOVA was performed to analyze the significant difference between the indicated groups. ****P value <0.0001; ns, no significant difference. (C) Immunoprecipitation (IP) analysis of Flag-tagged SH3P2 with ELM1-GFP, DRP3A-YFP, and DRP3B-YFP. Cell lysates from Arabidopsis protoplasts transiently expressing SH3P2-5xFlag with ELM1-GFP, DRP3A-YFP, DRP3B-YFP, or GFP were subjected to the GFP-trap assay, followed by immunoblot using anti-GFP and anti-Flag antibodies, respectively. Similar results were obtained from three different independent experiments. (D) IP analysis of Flag-tagged SH3P2 with ELM1 truncations. Cell lysate from Arabidopsis protoplasts transiently expressing SH3P2-5xFlag with ELM1-GFP, ELM1-NT-GFP, ELM1-CT-GFP, or GFP were subjected to the GFP-trap assay, followed by immunoblotting using anti-GFP and anti-Flag antibodies, respectively. Similar results were obtained from three different independent experiments. (E) Recruitment assay of SH3P2-GFP with CNX-mCherry-tagged ELM1 or its truncations in Arabidopsis protoplasts. CNX-RFP was used as a control. Quantification of the Mander’s colocalization coefficients between SH3P2-GFP and CNX-mCherry tagged ELM1 variants. M1, fraction of SH3P2-GFP signal that overlaps with CNX-mCherry tagged ELM1 variants’ signal is shown on the RIGHT. M2, fraction of CNX-mCherry tagged ELM1 variants’ signal that overlaps with SH3P2-GFP signal. Bars indicate the mean ± SD of 10 replicates. (Scale bar, 10 µm.)

To further investigate the relationship of SH3P2 with ELM1 and DRP3, we first examined the subcellular localization of SH3P2 with ELM1 by transient expression in Arabidopsis protoplasts (SI Appendix, Fig. S2A). Occasionally, we detected partial overlapping between ELM1-labeled puncta with those of SH3P2-RFP. Then, we performed BiFC (Bimolecular fluorescence complementation) analysis for SH3P2 and ELM1 by transient expression in tobacco leaf cells. However, only few puncta were observed, probably due to a weak or transient association between SH3P2 and ELM1 (Fig. 1B). It was noted that some signals of IDH1-mRuby are not on the mitochondria. As IDH1 is a cytosolic protein that requires subsequent transport into the mitochondrial matrix, overexpression of IDH1 might increase its cytosolic localization. Next, we carried out a GFP-trap assay using GFP-tagged ELM1, DRP3A, or DRP3B with Flag-tagged SH3P2. As shown in Fig. 1C, SH3P2 is associated with ELM1, DRP3A, and DRP3B. As the functional domains of ELM1 have not been characterized, to further assess the association between SH3P2 and ELM1, we generated two ELM1 truncations, including ELM1-NT (1-230aa) and ELM1-CT (231-427aa), for subcellular localization analysis with SH3P2. We found that ELM1-NT still displays a mitochondrial-like ring pattern, with few puncta overlapped with those positive with SH3P2 (SI Appendix, Fig. S2A). In contrast, ELM1-CT is retained in the cytosol. When DRP3A-RFP is coexpressed with the ELM1 truncations, in comparison to the tight association between DRP3A-RFP and ELM1-GFP, neither the N-terminus nor the C-terminus of ELM1 is overlapped with DRP3A-RFP (SI Appendix, Fig. S2B). We further conducted a GFP-trap assay for ELM1 truncations and Flag-tagged SH3P2, showing that SH3P2 has higher affinity to ELM1 or ELM1-NT than ELM1-CT (Fig. 1D). This was further supported by a recruitment assay (37) using the ER-anchored chimeras of ELM1 and its truncations. We found that SH3P2-GFP is well colocalized with CNX-mCherry-ELM1 and partially associated with CNX-mCherry-ELM1-NT (Fig. 1E). However, no obvious overlapping was detected between SH3P2-GFP and CNX-mCherry-ELM1-CT, or the CNX-RFP control (Fig. 1E).

Mitochondrial Fission Deficiency Leads to the Formation of Megamitochondria and Delays Phagophore Closure.

It has been suggested that ELM1 localizes to the outer mitochondrial membrane (OMM) and acts upstream of DRP3 for DRP3 recruitment to mediate mitochondrial fission, as DRP3 fails to localize to the mitochondria in the elm1 mutant (38). As SH3P2 is a cytosolic BAR domain containing protein that can induce membrane remodeling (42), we hypothesized that SH3P2 might be associated with ELM1 to facilitate mitochondrial fission and subsequent mitophagy. To exploit the possible hierarchical relationship between SH3P2 and ELM1, we introduced SH3P2-GFP and a mitochondrial marker (Mito-mCherry) in the wild-type (WT) or the elm1 mutant background for subcellular analysis. Of note, it seemed there are more overlapping signals between SH3P2-GFP and Mito-mCherry in elm1, particularly on the swollen mitochondria (Fig. 2A). Meanwhile, we generated transgenic plants expressing SH3P2-GFP/mCherry-ATG8e in elm1 to examine the mitophagosomal structures. Noticeably, we observed a swollen mitochondrion stained by a mitochondrial tracking dye (MitoBlue) was surrounded by a cup-shaped phagophore labeled by mCherry-ATG8e, while SH3P2 seemed to localize on the phagophore and in the mitochondrion (Fig. 2B). To gain more insights into the dynamic behavior of the phagophore with the mitochondria, we performed time-lapse imaging analysis using SH3P2-GFP/Mito-mCherry to monitor the development of the phagophore-like structure along the mitochondrion. In comparison with that in the WT background, it appeared that the development of cup-shaped phagophore structures along the mitochondrion is significantly delayed in elm1 (Fig. 2 C and D and Movie S4). It is worth noting that mitochondrial constriction becomes more evident when the phagophore expands, indicating it is unlikely for the entire mitochondrion to be sequestered by the phagophore.

Fig. 2.

Dysfunction of ELM1 leads to SH3P2 localization to swollen mitochondria. (A) Subcellular analysis of SH3P2 with the mitochondrial marker Mito-mCherry in WT and elm1 backgrounds. 5-d-old WT or elm1 seedlings expressing SH3P2-GFP/Mito-mCherry were used for confocal analysis. (Scale bar, 10 µm.) Quantification of the Mander’s colocalization coefficients between SH3P2-GFP and Mito-mCherry is shown on the Right. M1, fraction of SH3P2-GFP signal that overlaps with Mito-mCherry signal. M2, fraction of Mito-mCherry signal that overlaps with SH3P2-GFP signal. Bars indicate the mean ± SD of 10 replicates. (B) Subcellular analysis of SH3P2-GFP, mitochondria, and mCherry-ATG8e. 5-d-old elm1 seedlings expressing SH3P2-GFP/mCherry-ATG8e were stained by a mitochondrial tracking dye (MitoBlue) and subjected to confocal observation. (Scale bar, 10 µm.) Normalized fluorescence intensity profiles of GFP, mCherry for the area labeled by the white dashed lines are shown on the Right. Box and whisker plots at the Bottom show the percentage of mitochondria labeled by MitoBlue in the elm1 mutant in association with SH3P2-GFP only, mCherry-ATG8e only, or both. 10 different images containing > 10 megamitochondria from at least three seedlings after treatment were used for this analysis. (C) Time-lapse image analysis showed the formation of an SH3P2-GFP-positive phagophore-like structure engulfing a swollen mitochondrion. 5-d-old Arabidopsis elm1 seedlings expressing SH3P2-GFP and Mito-mCherry were subjected to confocal observation. The area indicated by the dashed square was enlarged and subjected to montage analysis as shown below. (Scale bar, 10 µm.) (D) Statistical analysis of the closure duration for the SH3P2-GFP-positive phagophore-like structures in WT and elm1 backgrounds. In WT, the SH3P2-GFP-positive phagophores close within an average of 60 s, whereas the SH3P2-GFP-positive phagophores remain in an unclosed status after 160 s in elm1. 10 independent cells from at least three different seedings for each background were analyzed. One-way ANOVA was performed to analyze the significant difference between indicated groups. ****, P value <0.0001.

Considering that ELM1 itself is a mitochondrial outer membrane protein, it is possible that SH3P2 is recruited by ELM1 to the mitochondrial surface. However, unlike DRP3, which fails to localize to the mitochondria when ELM1 is absent (38), our data suggest that mitochondrial localization of SH3P2 is not abolished in the elm1 background. It was previously noted that the T-DNA in the elm1 mutant is inserted into the ELM1 N-terminus, leading to a premature stop codon (38). Given that SH3P2 is associated with the N-terminus of ELM1 (Fig. 1 D and E), we suspect that an N-terminal fragment of ELM1 might be expressed in the elm1 mutant for association with SH3P2. Using RT-PCR analysis for the elm1 mutant, we detected an RNA fragment corresponding to the ELM1 N-terminus (1–122aa), albeit the expression level is declined (SI Appendix, Fig. S2C). To test the possible association between this ELM1 N-terminal fragment and SH3P2, we performed a GFP-trap assay, showing that ELM1 (1–122) is still associated with SH3P2 (SI Appendix, Fig. S2D). Furthermore, when we coexpressed the mNeonGreen-tagged ELM1 (1–122) with SH3P2-RFP in Arabidopsis protoplasts, ELM1 (1–122)-mNeonGreen was still detected on mitochondria labeled by COX4-mTurquise2, with few puncta overlapping with those of SH3P2-RFP (SI Appendix, Fig. S2E). Meanwhile, when ELM1 (1–122)-mNeonGreen was coexperssed with DRP3A-RFP, DRP3A-RFP primarily displayed a cytosolic pattern (SI Appendix, Fig. S2F). To further validate the above observation, we generated transgenic plants expressing ELM1-GFP and ELM1 (1–122)-mNeonGreen for subcellular observation. Consistently, we observed that ELM1 (1–122)-mNeonGreen localizes on swollen mitochondria labeled by MitoTracker, while ELM1-GFP was detected on relatively normal mitochondria (SI Appendix, Fig. S2G).

To further examine the ultrastructure of the SH3P2-positive swollen mitochondria in the elm1 mutant, we performed an immuno-gold electron microscopy (EM) labeling analysis in the SH3P2-GFP/elm1 root cells. Consistent with the confocal observation, swollen mitochondria are frequently observed (Fig. 3). As shown in Fig. 3A, gold particles against GFP antibodies were detected on a mitochondrion with two swollen ends connected by a narrow tube, seemingly undergoing fission. In comparison with the WT, more gold particles were detected on the mitochondria in the elm1 mutant (SI Appendix, Fig. S3).

Fig. 3.

Ultrastructure analysis of the abnormal mitochondria in the elm1 mutant. (A and B) Immunogold labeling of abnormal megamitochondria structures in SH3P2-GFP/elm1 root cells. 5-d-old Arabidopsis elm1 roots expressing SH3P2-GFP were subjected to high-pressure freezing, followed by anti-GFP gold particle labeling before observation under the electron microscope. Stars indicate the megamitochondria. Enlarged area indicated by the dashed square from A and B is shown in (A′ and B′) respectively, and gold particles are indicated by arrows.

Heat Stress Reshapes Tubular Mitochondria Into Megamitochondria and Promotes Mitochondrial Localization of SH3P2.

Given that ELM1 affects SH3P2 localization to mitochondrial and acts upstream of DRP3 to mediate mitochondrial fission, we next sought to examine the subcellular localization of SH3P2 in the DRP3-deficient background. Previous studies have shown that dysfunction of DRP3A causes a defect in mitochondrial fission, which is even more severe in the drp3adrp3b double mutant (40, 41). Indeed, in comparison with WT and drp3a-1, we found that mitochondria labeled by MitoTracker became more networked in two drp3adrp3b double mutants (40, 41), denoted as drp3adrp3b-1 and drp3adrp3b-3, respectively (SI Appendix, Fig. S4A). We then introduced SH3P2-GFP and Mito-mCherry into drp3a-1 for subcellular analysis. Similar to those in the elm1 mutant, SH3P2-positive puncta were also observed on swollen mitochondria labeled by Mito-mCherry in the drp3a-1 mutant (SI Appendix, Fig. S4B). We further introduced SH3P2-GFP in the drp3adrp3b-3 double mutant, together with mCherry-ATG8e to visualize the autophagosome structures. Notably, both SH3P2-GFP and mCherry-ATG8e were occasionally detected on large ring-like structures that could be readily observed in the bright field, resembling those found in the elm1 mutant (SI Appendix, Fig. S4C). To further evaluate the behavior of SH3P2 and mitophagy in the mitochondrial fission deficient condition, we generated a double mutant of elm1 and drp3a and transformed with SH3P2-GFP for crossing with RFP-ATG8e and Mito-CFP. As shown in Fig. 4 A, a, phagophore structures labeled by SH3P2-GFP and mCherry-ATG8e tend to expand along the Mito-CFP-positive swollen mitochondrion. Additionally, a dumbbell-shaped mitochondrion, likely undergoing fission, is partially enclosed by a phagophore positive with RFP-ATG8e, while SH3P2-GFP is primarily detected within the mitochondria but not on the phagophore (Fig. 4 A, a). As SH3P2-GFP does not always overlap with RFP-ATG8e on the mitochondria, it raises a possibility that SH3P2 localizations to the mitochondria and the phagophores might be regulated separately.

Fig. 4.

Loss of ELM1 and DRP3 affects SH3P2 localization to mitochondria and mitophagy upon heat stress. (A) SH3P2 and ATG8e were both trapped on swollen mitochondria when loss of ELM1 and DRP3A. 5-d-old elm1drp3a seedlings expressing SH3P2-GFP/Mito-CFP/RFP-ATG8e were used for confocal analysis. Arrows indicate the swollen mitochondria labeled by Mito-CFP and SH3P2-GFP partially sequestered by the RFP-ATG8e-positive phagophores. (Scale bar, 10 µm.) Box and whisker plots on the Right show the percentage of megamitochondria labeled by Mito-CFP in the elm1drp3a mutant associated with SH3P2-GFP only, RFP-ATG8e only, or both. 10 different images containing > 10 megamitochondria from at least three seedlings after treatment were used for this analysis. (B) Phenotypic analysis of elm1, elm1drp3a, and drp3adrp3b-3 mutants under HS conditions. 5-d-old Arabidopsis WT, elm1, elm1 drp3a, and drp3adrp3b-3 seedlings were subjected to 37 °C treatment for 3 d, then recovered at 22 °C for 3 d. Similar results were obtained from three different independent experiments. (C and D) Fractionation analysis of SH3P2 subcellular distribution upon HS treatment. 5-d-old Arabidopsis seedlings expressing SH3P2-GFP were subjected to HS treatment (37 °C for 3 h and recovery at 22 °C for 2 h) prior to Percoll fractionation. Quantification of the relative intensity ratio for the SH3P2 to mitochondrial proteins is shown in (D). Similar results were obtained from three different independent experiments. Bars represent the mean (±SD), and asterisks (*) indicate significant difference from HS treatment of the control. Two-way ANOVA was performed to analyze the significant difference between the indicated groups. *P value <0.05; **P value <0.01. T, total; C, cytosol; P, pellet; M, mitochondria; OMM, outer mitochondrial membrane. (E and F) Proteinase K analysis of SH3P2 subcellular localization in WT, elm1, or elm1drp3a background. Crude mitochondrial fractions were extracted from WT, elm1, and elm1drp3a mutants expressing SH3P2-GFP, respectively, and subjected to immunoblotting with anti-GFP, anti-IDH, or anti-VDAC antibodies. IDH and VDAC were used to indicate the mitochondria matrix and outer mitochondrial membrane, respectively. Quantification of the relative intensity for the protein degradation levels after Proteinase K treatment is shown in (F). Similar results were obtained from four different independent experiments. Bars represent the mean (±SD).

It has been previously shown that dysfunctional ELM1 and DRP3 both impair plant resistance to heat stress (HS) conditions (43), indicating the importance of mitochondrial dynamics in response to temperature fluctuation. We therefore subjected WT, elm1, elm1drp3a, and drp3adrp3b-3 seedlings to heat stress (37 °C) for phenotypical analysis. Notably, during the recovery period, the leaves of elm1, elm1drp3a, and drp3a3b-3 turned yellow, while those of WT remained green (Fig. 4B). The drp3adrp3b-3 mutant also displays a severe growth defect, which might be explained by the additional function of DRP3 proteins in peroxisome division (40, 41). Next, we tested how HS treatment alters the mitochondrial morphology and the localization of SH3P2. We first examined the subcellular localization of SH3P2-GFP/Mito-mCherry in WT before/after heat treatment. Upon heat stress, we observed more elongated mitochondria that were partially decorated by the signals of SH3P2-GFP (SI Appendix, Fig. S5A). Additionally, SH3P2-positive ring-like structures were induced. As shown in SI Appendix, Fig. S5B, an SH3P2-positive punctum is likely enclosing the mitochondrial fragment. In contrast, signals of SH3P2-GFP were primarily detected in the swollen mitochondria in elm1 after HS treatment. We then carried out a fractionation assay to isolate mitochondria before and after heat stress treatment. In agreement with the above observations, SH3P2-GFP proteins are more abundant on mitochondria upon HS treatment (Fig. 4 C and D). Furthermore, we isolated the mitochondria fractions from WT, elm1, and elm1drp3a mutants to examine the mitochondrial localization of SH3P2 using a Proteinase K protection assay (Fig. 4E). Treatment with Proteinase K in different concentrations first caused the degradation of SH3P2-GFP in WT regardless of HS treatment, but relatively less extent for the OMM-localized membrane marker VDAC and the matrix-localized cytosolic marker IDH, suggesting that SH3P2 is primarily associated with the outer mitochondrial membrane (Fig. 4 E and F). Consistently, the levels of VDAC and IDH were both increased in elm1 and elm1drp3a after HS treatment, while SH3P2-GFP proteins were still degraded faster than VDAC and IDH. We also noted a reduction in IDH when subjected to higher concentrations of Proteinase K, although it was less pronounced compared to that observed in SH3P2-GFP or VDAC. This might be attributed to the contamination from ruptured mitochondria during mitochondrial fractionation.

To better visualize the overall population of mitochondria, we introduced a mitochondrial outer membrane marker TOM7.2 into SH3P2-GFP/elm1 for HS treatment. Intriguingly, tubular mitochondria labeled by mRuby-TOM7.2 in the elm1 mutant were reshaped into interconnected megamitochondria after heat stress (Fig. 5 A and B and SI Appendix, Fig. S5C). Notably, these megamitochondria are still retained during the recovery period, indicating a defect in turnover. Additionally, the majority of the SH3P2 signal resides within the megamitochondria. During the recovery, SH3P2-GFP was also detected on the phagophore-like structure at the rim of the megamitochondrion that seemingly undergoes division, which resembles to the process of piecemeal mitophagy (Fig. 5 A and C and SI Appendix, Fig. S5C). We further conducted a time-lapse imaging analysis to dissect the dynamic relationship between SH3P2 and the megamitochondria. As shown in Fig. 5D and Movie S5, alone the development of SH3P2-positive phagophore, the megamitochondria was reshaped into a dumbbell, with the middle section constricted in a manner resembling the mitochondrial division process.

Fig. 5.

Heat stress induces overaccumulation of megamitochondria in elm1. (A). Subcellular analysis of SH3P2 and mitochondrial outer membrane protein TOM7.2 in the elm1 mutant. 5-d-old Arabidopsis elm1 seedlings expressing SH3P2-GFP/mRuby-TOM7.2 were subjected to HS treatment (37 °C for 3 h), or with additional recovery (22 °C for 1 h) (HS+recovery) before confocal analysis. (Scale bar, 10 µm.) (B) Quantification of the megamitochondria number in elm1 before/after HS treatment from (A). At least three different roots were selected, and 10 cells were captured. Student’s t test was performed to analyze the significant difference between the indicated groups. ****P value <0.0001. (C) Quantification of the Mander’s colocalization coefficients between SH3P2-GFP and Mito-mCherry from (A). M1, fraction of SH3P2-GFP signal that overlaps with Mito-mCherry signal. M2, fraction of Mito-mCherry signal that overlaps with SH3P2-GFP signal. Bars indicate the mean ± SD of 10 replicates. (D) Time-lapse analysis of SH3P2-GFP/mRuby-TOM7.2 in the elm1 mutant. 5-d-old Arabidopsis elm1 seedlings expressing SH3P2-GFP/mRuby-TOM7.2 were subjected to HS treatment (37 °C for 3 h) and recovery (22 °C for 1 h) before confocal analysis. Normalized fluorescence intensity profiles of GFP, mRuby for the area labeled by the dashed boxes are shown on the Right. Below showed the montage analysis for area indicated by the dashed square. (Scale bar, 10 µm.)

Mitochondrial Fission Activity Is Required for Efficient Piecemeal Mitophagy.

To further assess how mitophagy is affected in the mitochondrial fission defective mutants, we introduced YFP-ATG8e into elm1, elm1drp3a, and drp3adrp3b-3 mutants, respectively. Subcellular analysis reveals that YFP-ATG8e is primarily dispersed in the cytoplasm with a punctate pattern, but localizes on dilated unclosed phagophore structures in elm1drp3a and drp3adrp3b-3 mutants (SI Appendix, Fig. S6A). However, when we performed the GFP turnover assay to evaluate the degradation rate of YFP-ATG8e, it seemed that the GFP core was still processed in the mutants (SI Appendix, Fig. S6B). Next, we performed the ATG8 lipidation assay to monitor autophagic flux, showing that levels of nonlipidation and lipidation forms (ATG8 and ATG8–PE) were both increased in elm1drp3a and drp3adrp3b-3 mutants with/without HS treatment (SI Appendix, Fig. S6C).

The above observations suggest that dysfunction in mitochondrial fission disturbs mitophagosome completion, leading to the unclosed mitophagosome intermediates decorated by ATG8–PE. It is also possible that other types of autophagy are processed, thus not all lipidated ATG8 is correlated to mitophagosome-related structures. To further dissect the YFP-ATG8e-positive mitophagosome structures for heat-induced piecemeal mitophagy, we transformed the mitochondrial matrix protein IDH1-mRuby in WT and elm1drp3a double mutant, respectively, followed by crossing with YFP-ATG8e or YFP-ATG8e/elm1drp3a. Upon heat stress, we observed that more YFP-ATG8e-positive puncta are associated with the mitochondria labeled by IDH1-mRuby, while elm1drp3 mutant exhibits a higher ratio of association between YFP-ATG8e and IDH1-mRuby (Fig. 6 A and B). Of note, more than one phagophore intermediates labeled by YFP-ATG8e coalesce on the IDH1-positive megamitochondria, which however is barely detected in the WT background (Fig. 6 A and C). In addition, ring-like structures were also observed in elm1drp3, suggesting that complete mitophagosomes are still formed (Fig. 6C). Indeed, time-lapse imaging analysis revealed that during the expansion of the YFP-ATG8e-positive cup-shaped phagophore, a small YFP-ATG8e punctum was incorporated, ultimately resulting in phagophore closure into a ring (Fig. 6D and Movie S6).

Fig. 6.

Mitochondrial fission activity is required for efficient mitophagosome formation. (A–C) Subcellular analysis of mitophagosome formation in WT and elm1drp3b upon heat stress. 5-d-old Arabidopsis WT and elm1drp3b seedlings expressing YFP-ATG8e and IDH1-mRuby were subjected to HS treatment (37 °C for 3 h) and recovery (22 °C for 1 h), respectively, followed by confocal observation. (Scale bar, 10 µm.) Quantification of the percentage of mitochondria (IDH1-mRuby) in association with YFP-ATG8e is shown in (B). Quantification of the types of mitophagosome structures labeled by YFP-ATG8e is shown in (C). 5 images from at least three seedlings of each genotype were used for the analysis. Bars represent the mean (±SD), and asterisks (*) indicate significant difference from HS treatment of the control. Two-way ANOVA was performed to analyze the significant difference. *P value <0.05; **P value <0.01; ***P value <0.001; ****P value <0.0001; ns, no significant difference. (D) Time-lapse image analysis showed the development of a YFP-ATG8e-positive phagophore-like structure into a closed mitophagosome. 5-d-old Arabidopsis elm1drp3a seedlings expressing YFP-ATG8e and IDH1-mRuby were subjected to confocal observation. (Scale bar, 10 µm.) (E) Colocalization analysis of YFP-ATG8e and mRuby-NBR1 on the megamitochondria. 5-d-old Arabidopsis elm1drp3a seedlings expressing YFP-ATG8e and mRuby-NBR1 were subjected to confocal observation. Quantification of the Mander’s colocalization coefficients between YFP-ATG8e and mRuby-NBR1 is shown on the Right. M1, fraction of YFP-ATG8e signal that overlaps with mRuby-NBR1 signal. M2, fraction of mRuby-NBR1 signal that overlaps with YFP-ATG8e signal. Bars indicate the mean ± SD of 10 replicates. (Scale bar, 10 µm.) (F) Immunoblot analysis of mitochondrial turnover upon HS treatment (37 °C for 3 h) and recovery (22 °C for 1 h) in WT, atg5, elm1, elm1 drp3a, and drp3adrp3b-3 mutants, respectively. The protein samples were subjected to immunoblotting with anti-IDH, anti-VDAC, anti-TOM, or anti-NBR1 antibodies. cFBPase was used as a loading control. (G) Quantification of the relative intensity for the indicated proteins in (F) from four different independent experiments. Bars represent the mean (±SD), and asterisks (*) indicate significant difference from HS treatment of the control. Two-way ANOVA was performed to analyze the significant difference. *P value <0.05; **P value <0.01; ****P value <0.0001; ns, no significant difference.

As the selective autophagic receptor NBR1 has been well characterized in binding ubiquitinated substrates for degradation through autophagy under heat stress condition (44), we wondered whether piecemeal mitophagy also engages NBR1. Intriguingly, when YFP-ATG8e and mRuby-NBR1 were coexpressed in elm1drp3a, NBR1 displays a similar pattern as ATG8e at multiple sites on the megamitochondria (Fig. 6E). It is plausible to speculate that NBR1 might participate in mitophagy for the removal of the mitochondria by targeting the ubiquitinated mitochondrial proteins, leading to the recruitment of ATG8-positive phagophore along the damaged mitochondrial area for piecemeal mitophagy. Indeed, when we examined the overall ubiquitination levels by immunoblotting, we found that both elm1drp3a and drp3adrp3b-3 showed a strong ubiquitination signal even without HS treatment (SI Appendix, Fig. S6D). Next, we sought to measure the mitochondrial protein turnover rates, as well as the levels of NBR1, in WT, elm1, elm1drp3a, and drp3adrp3b-3 backgrounds, respectively. As shown in Fig. 6 F and G, it seemed that the reduction level of representative mitochondrial proteins after HS treatment in elm1, including IDH, VDAC, and TOM9, is comparable to that in the WT. However, their turnover rates were suppressed in both elm1drp3a and drp3adrp3b-3. On the other hand, NBR1 protein level is only slightly increased in drp3adrp3b-3 mutant when compared with other backgrounds. It is possible that NBR1 might not function as a specific receptor for mitophagy, implying that other mitophagy receptors might be involved.

To further assess the mitochondrial fission events in connection with the mitophagosome formation in a higher resolution, we subjected the SH3P2-GFP/elm1drp3a root cells to high-pressure freezing (HPF) and employed 3D electron tomography (ET) to reconstruct the spatial arrangement of mitochondria and phagophore structures. Three representative examples are shown in Fig. 7. In Fig. 7A, a megamitochondrion accompanied by a phagophore-like structure was captured, with the outgrowth of the phagophore intermediates toward the constriction site. In Fig. 7B, two phagophores surround the megamitochondrion, which is interconnected with the tubular-shaped mitochondrion that comprising organized cristae structures. Additionally, we observed several protrusions with normal cristae developing from a swollen mitochondrion, likely due to the loss of fission activity (SI Appendix, Fig. S7A). A complete mitophagosome engulfing an abnormal mitochondrial structure is also captured and depicted in Fig. 7C. Of note, inward budding vesicle-like structures comprising cytoplasmic ribosomes were detected in the lumen of the mitochondrion, indicating that these structures are likely originated from the cytoplasm. In agreement with the above model, when the samples were subjected to immuno-gold EM labeling against anti-ATG8 antibodies, we observed that some abnormal megamitochondria are encircled by complete or incomplete mitophagosome structures, while vesicle-like structures are also detected within the megamitochondria (SI Appendix, Fig. S7B).

Fig. 7.

3D tomography analysis reveals piecemeal sequestration of swollen mitochondria by autophagosomal-like structures in the elm1drp3a mutant. (A and B) Serial tomography slices and reconstructed 3D model illustrate the phagophore-like structures (white triangles) develop along the swollen mitochondria that are likely undergoing fission. (C) Serial tomography slices and reconstructed 3D models showing an abnormal mitochondrion enclosed by a double-membrane autophagosomal-like structure. Vesicular structures in the mitochondrion were labeled by triangles in different colors (purple, magenta, and green). The cup-shaped membrane (white), mitochondrial outer membrane (purple), mitochondrial cristae (cyan), damaged mitochondrial protein aggregates (yellow), and ribosomes (luminous yellow spheres) are represented in the 3D models. (Scale bar, 500 nm.)

Since SH3P2 is likely recruited by ELM1, we next sought to investigate whether SH3P2 deficiency also affects mitophagy and mitochondrial morphology. There are two RNAi lines reported for SH3P2 (36, 45, 46). Immunoblotting analysis using SH3P2 antibodies confirmed a reduction of the SH3P2 level in these two lines (SI Appendix, Fig. S8A). We then subjected these two lines to HS treatment for measuring the turnover rate of representative mitochondrial proteins and NBR1. Of note, while all the tested mitochondrial proteins declined after HS treatment in the WT, they retained the same levels in the two RNAi lines, suggesting loss of SH3P2 suppresses their turnover (SI Appendix, Fig. S8 B and C). To gain more insights into the impact of SH3P2 on mitochondrial morphology, we used the root tips from 5-d-old WT and two RNAi lines with/without HS treatment for HPF and freeze substitution. As shown in SI Appendix, Fig. S8 D and E, elongated dumbbell-shaped mitochondria were frequently observed in both RNAi lines. Particularly, after HS treatment, irregular mitochondria with multiple constriction sites emerged, which are resemble to that in the elm1drp3a mutant (SI Appendix, Fig. S7A). In agreement with the above observation, quantification analysis showed that both mitochondrial size and length were increased in two SH3P2 RNAi lines in comparison with that in the WT (SI Appendix, Fig. S8 F and G).

Discussion

In this work, we have unraveled an unappreciated mitophagosome formation model for piecemeal mitophagy in connection with the mitochondrial fission machinery in Arabidopsis plants (SI Appendix, Fig. S9). By subcellular analysis, we demonstrated that segregation of mitochondrial fragment is accompanied by phagophore expansion prior to mitophagosome completion (Fig. 1A and SI Appendix, Fig. S1B and Movies S1–S3). We further showed that the plant-unique autophagic regulator SH3P2 forms a complex with the plant-specific mitochondrial fission factor ELM1, together with the conserved dynamin DRP3. Upon mitophagy induction, SH3P2 is recruited to the mitochondrial constriction site prior to phagophore expansion. Our functional and genetic analysis suggests that SH3P2 likely acts downstream of ELM1 (Figs. 2 and 5 and SI Appendix, Fig. S2). It has been reported that SH3P2 interacts with DRP2 to induce vesicle tubulation for clathrin-coated vesicle formation (42). Regarding the membrane tubulation activity of SH3P2, it is plausible to speculate that SH3P2 cooperates DRP3 to facilitate the mitochondrial membrane deformation for mitochondrial fission. Similar to the mammalian system, it seemed that plant mitochondrial fission activity is dispensable for phagophore initiation, as swollen mitochondria are still partially sequestered by the ATG8-positive phagophore structures in the fission-defective mutants, albeit the completion rate is significantly delayed (Figs. 2D and 6D). Therefore, it is likely that ELM1-mediated mitochondrial fission serves as a contractile force for efficient mitochondrial segregation into the phagophore. We also noticed that a mitochondrial-localized ELM1 N-terminal fragment is produced in the elm1 mutant to associate with SH3P2 (SI Appendix, Fig. S2). Overexpression of this fragment leads to accumulation of swollen mitochondria, indicating a dominant negative effect of the ELM1 N-terminus on mitochondrial morphology. It would be valuable to analyze the dynamics of SH3P2 using an independent null elm1 mutant line. Moreover, given that mitophagosome completion and mitophagy flux is not completely blocked in elm1 and drp3 mutants (Fig. 6 D–G), it raises the possibility that alternative mitochondrial fission factors or pathways might be engaged. Indeed, mitochondrial fission factors independent of ELM1 and DRP3 have been reported in plants to participate in the division of other organelles, such as peroxisomes and chloroplasts (47–50). Thus, it is unnecessary to conclude that ELM1 and DRP3 function exclusively in piecemeal mitophagy. Future studies are required to explore the molecular network governing plant mitochondrial fission and their contributions to mitophagy or general organelle turnover.

Furthermore, we have demonstrated that the canonical plant selective autophagy receptor NBR1 participates in elimination of damaged mitochondria (Fig. 6E and SI Appendix, Fig. S6D). Arabidopsis NBR1 functions as a hybrid counterpart of mammalian NBR1 and p62/SQSTM1 for targeting ubiquitinated cargoes in different types of selective autophagy (44, 51–53), but whether it contributes to plant mitophagy remains unclear. In mammalian cells, p62/SQSTM1 has also been implicated to mediate piecemeal mitophagy by engaging the OMM protein SAMM50 (21, 22, 54). Intriguingly, we found NBR1 proteins are dispatched on multiple sites on the surface of the mitochondria together with ATG8 proteins. In a classical model, mitophagosome formation is initiated from a phagophore membrane, which expands by fusion with other membrane sources and closes into a complete mitophagosome (6). The observation of more than one phagophore intermediate on the swollen mitochondria allowed us to propose that the coalescence of several phagophore precursors likely represent another assembly model for mitophagosome formation, in which that damaged mitochondrial parts might undergo ubiquitination and then be recognized by mitophagy receptor like NBR1, followed by recruitment of ATG8-positive phagophores for piecemeal mitophagy. Alternatively, it is possible that other uncharacterized mitochondrial proteins directly recruit ATG8 in coordination with NBR1 to facilitate piecemeal mitophagy.

Our 3D ET data also support a membrane coalescence model for ELM1-mediated piecemeal mitophagy in plant cells, in which mitochondrial protrusion is sequestered by more than one phagophore precursor (Fig. 7 A and B). The mitochondrial network tends to segregate the damaged part, forming protrusions that still attached to the healthy mitochondrial body. Alternatively, the fission machinery might on the other way protect the healthy parts from degradation. Another possibility is that the discontinuous phagophores might be due to insufficient membrane supply to seal the megamitochondria. The mitochondrial fission site has been linked to lipid transfer by establishing membrane contact for mitochondria with other membranes. For example, ATG9 vesicles function as one important membrane source for lipid transfer during the phagophore initiation and development (55–61). In addition, ATG9 vesicles have been reported to traffic to mitochondria to facilitate mitophagy (56, 62, 63). We indeed observed defects in mitochondrial fission induces some vesicle-like structures in close contacts with the mitochondria (Fig. 7C). Further efforts are required to investigate the identities and functions of these mitochondrial-related vesicles. Furthermore, we also noted that SH3P2-positive diffuse or punctate signals within the mitochondria in the mitochondrial fission deficient mutants (Fig. 5 A and D). Recent studies have also suggested that remodeling of the mitochondrial inner membrane is required for mitophagy (25). Given that mitochondrial turnover and morphology are perturbed in the SH3P2 RNAi lines (SI Appendix, Fig. S8), it is possible that SH3P2 might play an additional role in the mitochondria. Future studies are necessary to further investigate the potential roles of SH3P2 in remodeling the mitochondrial inner membrane for piecemeal mitophagy.

In summary, our results demonstrated that mitochondrial fission activity is essential for both housekeeping mitophagy and heat stress-induced mitophagy by promoting the mitochondrial segregation and mitophagosome closure. In comparison to wholesale mitophagy, piecemeal mitophagy operates as an efficient adaptive mechanism that can be leveraged to energetically remove damaged mitochondrial parts in plant cells to cope with the stressful environment.

Materials and Methods

An extended section with comprehensive information on materials and methods is provided in SI Appendix. This includes details on plant growth and treatment conditions, plasmid construction, and plant materials. Protocols for transient expression, bimolecular fluorescence complementation assay, mitochondria fractionation and protease treatment assay, confocal imaging, protein extraction and immunoblotting, Co-IP, RT-PCR, EM analysis of resin-embedded cells, and electron tomography analysis are also described. In addition, methods for quantification, statistical analysis, and modeling are included.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Dynamic analysis of SH3P2-GFP-positive mitophagosome formation in Arabidopsis root cells upon DNP treatment. Video corresponds to montage shown in Fig. 1A.

Movie S2.

Dynamic analysis of SH3P2-GFP-positive mitophagosome formation in Arabidopsis root cells upon DNP treatment. Video corresponds to montage shown in Fig. S1B,a.

Movie S3.

Dynamic analysis of SH3P2-GFP-positive mitophagosome formation in Arabidopsis root cells upon DNP treatment. Video corresponds to montage shown in Fig. S1B,b.

Movie S4.

Time-lapse imaging showing the formation of a SH3P2-GFP-positive phagophore-like structure engulfing an abnormal mitochondrion. Video corresponds to montage shown in Fig. 2C.

Movie S5.

Time-lapse imaging of SH3P2-GFP/mRuby-TOM7.2 in elm1 mutant. Video corresponds to montage shown in Fig. 5D.

Movie S6.

Time-lapse imaging of mitophagosome formation in elm1 mutant. Video corresponds to montage shown in Fig. 6D.

Movie S7.

Three-dimensional reconstructed electron tomography and model for the segregating megamitochondrial structure shown in Fig. 7A.

Movie S8.

Three-dimensional reconstructed electron tomography and model for the segregating megamitochondrial structure shown in Fig. 7B.

Movie S9.

Three-dimensional reconstructed electron tomography and model for the complete mitophagosome shown in Fig. 7C.

Movie S10.

Three-dimensional reconstructed electron tomography and model for the mitophagosome structure shown in SI Appendix Fig. S7A.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.