ATG2A acts as a tether to regulate autophagosome-lysosome fusion in neural cells

ABSTRACT

The macroautophagy/autophagy proteins ATG2A and ATG2B transfer lipids for phagophore membrane growth. They also form stable complexes with WDR45 and WDR45B. Our previous study demonstrated that WDR45 and WDR45B mediate autophagosome-lysosome fusion in neural cells. Given the defective autophagosome formation in cells lacking both ATG2s, their role in later autophagy stages is hard to explore. Here, we report that in neuroblastoma-derived Neuro-2a (N2a) cells, knocking down (KD) Atg2a, but not Atg2b, results in significant accumulation of SQSTM1/p62 and MAP1LC3-II/LC3-II, indicating impaired autophagy. Atg2a deficiency does not affect autophagosome formation, but reduces colocalization of autophagosomal LC3 with late endosomal/lysosomal RFP-RAB7, suggesting impaired autophagosome-lysosome fusion. ATG2A interacts with the SNARE proteins STX17, SNAP29, and VAMP8, facilitating their assembly. Overexpression of ATG2A partially rescues the autophagosome-lysosome fusion defects in Wdr45- and Wdr45b-deficient cells. ATG2 and another tether protein, EPG5, function partially redundantly in mediating autophagosome-lysosome fusion. Thus, ATG2A plays a key role in neural autophagy by tethering autophagosomes with lysosomes for fusion.

Abbreviations: AAV: adeno-associated virus; ATG2A^r: RNAi-resistant ATG2A; Baf: bafilomycin A₁; co-IP: co-immunoprecipitation; CQ: chloroquine; DKD: double knockdown; DKO: double knockout; ER: endoplasmic reticulum; KD: knockdown; KO: knockout; MIL: membrane-impermeable Halo ligand; MPL: membrane-permeable Halo ligand; N2a: Neuro-2a; NC negative control; PG: phagophore; PtdIns3K: phosphatidylinositol 3-kinase; PtdIns3P: phosphatidylinositol-3-phosphate; TEM: Transmission electron microscopy; TM: transmembrane domain; WT: wild-type.

Introduction

Macroautophagy, commonly known as autophagy, is an evolutionarily conserved degradation process mediated by lysosomes [1–6]. This process begins with the initiation and nucleation of cup-shaped phagophores (PGs), which subsequently expand and close to form double-membrane autophagosomes. Autophagosomes are then transported to the vacuole (in yeast and plants) or lysosomes (in metazoans) for the breakdown of sequestered materials. Yeast genetic screens identified a group of Atg proteins which play distinct roles at different stages of the autophagy pathway [1,4]. In yeast, the induction of autophagy leads to the initial recruitment of the Atg1 complex to the autophagosome formation site, known as the phagophore assembly site/PAS. This is followed by the recruitment of the Vps34 phosphatidylinositol 3-kinase (PtdIns3K) complex, which generates phosphatidylinositol-3-phosphate (PtdIns3P) locally to promote the targeting of downstream Atg proteins, including the Atg2-Atg18 complex and proteins responsible for Atg8 lipidation, for the elongation of PGs. The Atg2-Atg18 complex is known to retrieve Atg9-containing vesicles from forming autophagosomes [7,8], and tethers the endoplasmic reticulum (ER) to the leading edges of the expanding PGs [9,10]. Recent studies have revealed that Atg2 possesses lipid transfer activity in vitro, which is enhanced by forming a complex with Atg18, suggesting a critical role of Atg2 in transferring phospholipids from the ER to PGs during PG expansion [11]. Despite these findings, the exact molecular functions of Atg2 in autophagy need further investigation.

Mammalian autophagy is a more intricate process compared to yeast autophagy, involving additional steps and specialized mechanisms. This complexity is primarily evident in two aspects: 1) at the formation step, autophagosomes in mammals are generated at multiple sites near the ER and maintain close contact with the ER during their expansion, while yeast autophagy typically occurs in the vicinity of the vacuole; 2) at the fusion step, unlike yeast autophagosomes, which directly fuse with the vacuole, nascent autophagosomes in mammalian cells undergo a series of maturation steps by fusing with endosomes/lysosomes at various stages [1–4]. To facilitate these unique steps, additional factors, such as metazoan-specific EPG proteins identified from C. elegans, are necessary [6,12]. Furthermore, mammalian genomes carry multiple homologs of individual yeast Atg genes. For instance, mammalian cells contain four Atg18 homologs: WIPI1, WIPI2, WDR45B/WIPI3 and WDR45/WIPI4 [13]. Subsequent studies have demonstrated that these proteins have distinct functions in autophagy. WIPI2 mediates the contacts between the ER and PGs [14,15] and recruits the LC3 conjugation system ATG12–ATG5-ATG16L1 to the autophagosome formation sites by binding to PtdIns3P [16]. WDR45 and WDR45B have been reported to regulate autophagy via the AMPK and MTOR signaling pathways and may influence autophagosome size [17]. Our findings with wdr45 and wdr45b knockout (KO) mice have revealed that Wdr45 and Wdr45b are essential for autophagy, specifically in the central nervous system/CNS, but not in other tissues such as liver and spleen [18]. Furthermore, we demonstrated a specific mechanistic role of WDR45 and WDR45B in neural cells: they mediate autophagosome-lysosome fusion by interacting with SNARE proteins and the tether EPG5, thereby facilitating assembly of the SNARE complex [19].

Mammals also have two homologs of yeast Atg2, ATG2A and ATG2B, which function redundantly in autophagy. Knockdown (KD) of either ATG2A or ATG2B does not significantly affect autophagy, but simultaneous depletion of both genes results in abnormal accumulation of ATG9 and blockage in the autophagy pathway [20]. Similar to yeast Atg2, mammalian ATG2A and ATG2B also exhibit lipid transfer activity, suggesting their role in providing phospholipids for the expansion of PGs [21–23]. Furthermore, ATG2 proteins interact with all four WIPI and WDR45 proteins, with stronger binding to WDR45 and WDR45B than to WIPI1 and WIPI2 [24]. The ATG2-WDR45 complex forms a 20-nm-long rod structure [24,25]. Given the function of WDR45 and WDR45B in autophagosome maturation into autolysosomes, it remains unclear whether the Atg2 homologs may also have evolved to fulfil distinct functions at later stages of autophagy.

In this study, we uncovered a critical and previously unrecognized role of ATG2A in autophagy within Neuro-2a (N2a) neural cells. Contrary to the established view that ATG2A and ATG2B function redundantly in autophagy, our findings reveal that ATG2A, but not ATG2B, is indispensable for autophagy in these cells. While the absence of both ATG2A and ATG2B impairs autophagosome formation, silencing Atg2a alone specifically causes defects at the later autophagosome-lysosome fusion step. We demonstrated that ATG2A interacts with WDR45 and WDR45B, thereby tethering autophagosomes to lysosomes. Additionally, ATG2A binds to SNARE proteins and promotes assembly of the SNARE complex, which is essential for autophagosome-lysosome fusion. Our study reveals that the ATG2A-WDR45 or -WDR45B complexes are pivotal for the maturation of autophagosomes into autolysosomes, and thus provides novel insights into the specialized functions of ATG2A in neural cells.

Results

ATG2A, but not ATG2B, is critical for autophagy in neural cells

Interactions of Atg2 with various Atg18 homologs across different species, including yeast Atg18, C. elegans EPG-6 and mammalian WDR45 and WDR45B, have been extensively documented [9,10,24–26]. Our previous study demonstrated that WDR45 and WDR45B specifically modulate autophagosome-lysosome fusion in N2a cells, a mouse neuroblastoma-derived cell line. This finding prompted us to investigate the role of ATG2A and ATG2B in neural cells. We performed siRNA knockdown (KD) of Atg2a alone, KD of Atg2b alone, or double knockdown (DKD) of Atg2a and Atg2b in N2a cells. Compared to cells treated with negative control (NC) siRNA, Atg2a KD and Atg2a Atg2b DKD N2a cells showed markedly elevated amounts of SQSTM1/p62 and LC3-II, indicative of autophagy inhibition, whereas Atg2b KD cells did not exhibit obvious autophagy defects (Figure 1A). Reintroducing RNAi-resistant GFP-ATG2A^r, but not ATG2B^r-GFP, reversed the accumulation of SQSTM1 and LC3-II levels in Atg2a Atg2b DKD cell (Figure 1A). In Atg2a KD cells, SQSTM1 aggregates and LC3 puncta accumulated and they colocalized well with each other (Figure 1B,C). RFP-GFP-LC3 assays revealed that compared to fed cells, the number of red-only puncta increased after 4 h starvation in NC RNAi-treated cells, while siAtg2a caused a significant reduction in the number of red-only LC3 puncta upon starvation (Figure 1D,E).

Figure 1.

ATG2A is important for neural autophagy. (A) Immunoblots of SQSTM1 and LC3 show that levels of SQSTM1 and LC3-II are dramatically increased in siAtg2a and Atg2a Atg2b DKD N2a cells, but not in siAtg2b cells, compared to cells treated with negative control (NC) siRNA. The lower rows show that the expression of ATG2A and ATG2B is efficiently inhibited by siRNAs against Atg2a and Atg2b. Increased levels of SQSTM1 and LC3-II caused by Atg2a Atg2b DKD are alleviated by the re-expression of RNAi resistant GFP-ATG2A^r, but not by ATG2B^r-GFP. SE, short exposure; LE, long exposure. * indicates nonspecific bands. Quantification of SQSTM1 and LC3-II levels (normalized by ACTB levels) is shown as mean ± SEM (n = 3). The level in control cells is set to 1.0. *, p < 0.05; N.S., not significant. (B and C) Immunostaining with anti-SQSTM1 and anti-LC3 antibodies shows that SQSTM1⁺ and LC3⁺ puncta accumulate in siAtg2a N2a cells, while no accumulation is detected in NC siRNA-treated cells (B). Numbers of SQSTM1⁺ and LC3⁺ puncta from 3 biological repeats are quantified (NC, n = 54; siAtg2a, n = 58) and shown as mean ± SEM (C). ****, p < 0.0001. Scale bars: 5 μm; inserts, 2 μm. (D and E) RFP-GFP-LC3 assays demonstrate that after 4 h EBSS starvation (strv), the number of red-only RFP-GFP-LC3 dots in NC-treated N2a cells is dramatically increased, compared to cells at fed conditions. At fed conditions, siAtg2a cells contain more GFP⁺ RFP⁺ LC3 puncta than NC cells. Starvation does not induce an increase in the number of red-only RFP-GFP-LC3 dots in siAtg2a cells. (D). Numbers of red-only puncta from 3 biological repeats (NC fed, n = 53; NC strv, n = 51; siAtg2a fed, n = 55; siAtg2a strv, n = 52) are shown as mean ± SEM (E). ****, p < 0.0001; N.S., not significant. Scale bars: 5 μm; inserts, 2 μm. (F) Immunoblots of SQSTM1 and LC3 show that levels of SQSTM1 and LC3-II are dramatically increased in siAtg2a and Atg2a Atg2b DKD HT22 cells, but not in siAtg2b cells, compared to NC siRNA-treated cells. The lower rows show that expression of ATG2A and ATG2B is efficiently inhibited by siRNAs against Atg2a and Atg2b. * indicates nonspecific bands. Quantification of SQSTM1 and LC3-II levels (normalized by ACTB levels) is shown as mean ± SEM (n = 4). The level in NC cells is set to 1.0. *, p < 0.05; ****, p < 0.0001; N.S., not significant. (G and H) Immunostaining with anti-SQSTM1 and anti-LC3 antibodies shows that SQSTM1⁺ and LC3⁺ puncta accumulate in siAtg2a HT22 cells, but not in cells treated with NC siRNA (G). Numbers of SQSTM1⁺ and LC3⁺ puncta from 3 biological repeats are quantified (NC, n = 45; siAtg2a, n = 49) and shown as mean ± SEM (H). ****, p < 0.0001. Scale bars: 5 μm; inserts, 2 μm. (I) Immunoblots of LC3 show that levels of LC3-II are dramatically increased in shAtg2a and Atg2a Atg2b DKD primary hippocampal neurons, but not in shAtg2b neurons, compared to NC shRNA-treated cells. The lower rows show that expression of ATG2A and ATG2B is efficiently inhibited by shRNAs against Atg2a and Atg2b. Quantification of LC3-II levels (normalized by ACTB levels) is shown as mean ± SEM (n = 4). The level in NC cells is set to 1.0. ***, p < 0.001; ****, p < 0.0001; N.S., not significant.

ATG2A and ATG2B have been shown to function redundantly with each other, and siATG2A caused only minor or no autophagy defects in HeLa cells [20]. To confirm the specific function of ATG2A in neural cells, we knocked down ATG2A and/or ATG2B in another neural cell line, HT22, and in two non-neural cell lines, HeLa and 293T. Consistent with the previous reports, siATG2A alone in HeLa or 293T cells did not noticeably alter SQSTM1, LC3-I or LC3-II levels, while simultaneous KD of ATG2A and ATG2B led to substantially increased levels of SQSTM1 and LC3-II (Figure S1A,B). Moreover, the increased SQSTM1 and LC3-II levels were rescued by re-expressing either GFP-ATG2A^r or ATG2B^r-GFP (Figure S1A,B). In contrast, in HT22 cells, Atg2a deficiency caused autophagy defects while Atg2b KD did not (Figure 1F–H). These results mirror our observations in N2a cells.

We further validated the findings in primary cultured neurons using adeno-associated virus (AAV)-delivered shRNA. shAtg2a and Atg2a Atg2b DKD, but not shAtg2b, neurons showed increased levels of LC3-II, indicating impaired autophagic flux (Figure 1I). Collectively, these findings suggest that ATG2A, akin to WDR45 and WDR45B, plays a critical and non-redundant role in regulating autophagy in neural cells.

ATG2A modulates autophagosome-lysosome fusion

We next examined at which step siAtg2a results in an autophagy defect. Transmission electron microscopy (TEM) is the “gold standard” method to study the nature of accumulated LC3 structures. With TEM, we found that after 2 h starvation, double-membrane autophagosomes were detected in control cells, while Atg2a Atg2b DKD mainly caused accumulation of phagophores (Figure 2A,B), which is consistent with previous reports [27]. To our surprise, in siAtg2a cells, large numbers of double-membrane autophagosomes were observed in starved cells, although they were smaller in size compared with control cells (Figure 2A). This observation suggests that suppressing Atg2a alone does not affect autophagosome formation.

Figure 2.

ATG2A regulates autophagosome-lysosome fusion. (A and B) TEM micrographs of NC, siAtg2a and Atg2a Atg2b DKD N2a cells after 4 h EBSS starvation (A). White arrows indicate autophagosomes. Black arrows indicate autolysosomes. White arrowheads indicate phagophores. The lower panels show enlarged images of the boxed regions from the upper panels. NC, negative control siRNA. Strv, starvation. The size of autophagosomes (n = 42 autophagosomes for NC; n = 80 autophagosomes for siAtg2a; n = 27 autophagosomes for DKD) and the numbers of autophagosomes and autolysosomes (n = 100 cell sections for NC; n = 104 cell sections for siAtg2a; n = 107 cell sections for DKD) are quantified and shown as mean ± SEM (B). *, p < 0.05; **, p < 0.01; ****, p < 0.0001. Scale bars: 200 nm. (C-E) Schematic illustration of the Halo-LC3 assay. PG, phagophore. AV, autophagosome. AL, autolysosome (C). After 6 h bafilomycin A₁ (Baf) treatment, MPL⁺ Halo-LC3 puncta are detected in siAtg2a N2a cells, similar to cells treated with NC siRNA, while most Halo-LC3 puncta in Atg2a Atg2b DKD cells are MIL⁺ MPL⁻. The defects were reversed by reintroduction of GFP-ATG2A^r orATG2B^r-GFP (D). The percentages of MIL⁺ MPL⁻, MIL⁺ MPL⁺ and MIL⁻ MPL⁺ Halo-LC3 puncta from 3 biological repeats are quantified (NC, n = 44; siAtg2a, n = 51; DKD, n = 48; DKD GFP-ATG2A^r, n = 49; DKD ATG2B^r-GFP, n = 51) and shown as mean ± SEM (E). Compared to NC cells, ****, p < 0.0001; N.S., not significant. Scale bars: 5 μm. (F and G) After bafilomycin A₁ (Baf) treatment for 6 h, the colocalization of accumulated LC3 dots with RFP-RAB7⁺ vesicles is much lower in siAtg2a N2a cells compared to cells treated with NC siRNA (F). The percentages of LC3⁺ puncta positive for RAB7 from 3 biological repeats are quantified (NC, n = 53; siAtg2a, n = 54) and shown as mean as mean ± SEM (G). ****, p < 0.0001. Scale bars: 5 μm; inserts, 2 μm. (H and I) TEM micrographs of NC siRNA and siAtg2a cells after 6 h chloroquine (CQ) treatment plus EBSS starvation (Strv) during the last 2 h (H). White arrows indicate autophagosomes. Black arrows indicate autolysosomes. The numbers of autophagosomes and autolysosomes are quantified (n = 124 cell sections for NC; n = 128 cell sections for siAtg2a) and shown as mean ± SEM (I). ****, p < 0.0001. Scale bars: 500 nm.

TEM analysis of single sections cannot determine whether the autophagosomes are fully closed. To verify the closure of the autophagosomes, we utilized the Halo-LC3 assay, which differentiates between PGs, nascent autophagosomes and mature autophagosomes [28]. After permeabilizing the plasma membrane, cells expressing a Halo-LC3 fusion protein were incubated with an excess amount of a membrane-impermeable Halo ligand (MIL) conjugated with Alexa Fluor 660. This step labeled Halo-LC3 on accessible early autophagic structures, including unclosed autophagosomes and the outer membranes of closed nascent autophagosomes. Subsequently, the cells were fixed and stained with a membrane-permeable Halo ligand (MPL) conjugated with tetramethylrhodamine/TMR. This second step labeled Halo-LC3 on the inner membranes of nascent autophagosomes and autolysosomes, providing a clear distinction between different stages of autophagosome maturation (Figure 2C). The Halo-LC3 assay was performed in cells treated with bafilomycin A₁ (Baf), which blocks the late stages of autophagy by inhibiting lysosomal degradation and thus allows accumulation of autophagic structures for visualization. We found that similar to control cells, numerous MIL⁻ MPL⁺ LC3 puncta, representing closed autophagic structures, were detected in Baf-treated siAtg2a cells, while in Atg2a Atg2b DKD cells, LC3 puncta were positive for MIL, but not MPL (Figure 2D,E), representing PGs. Reintroduction of GFP-ATG2A^r or ATG2B^r-GFP increased the percentage of MIL⁻ MPL⁺ LC3 puncta in Atg2a Atg2b DKD cells (Figure 2D,E). These results indicate that the LC3⁺ structures in Atg2a KD and DKD cells are autophagic structures at different stages.

We applied another method to verify the closure of autophagosomes in siAtg2a cells. PtdIns3P plays a pivotal role in the lipidation of LC3, a key step in autophagosome formation. Inhibition of PtdIns3K by wortmannin prevents the accumulation of LC3 on early autophagic structures, but does not affect the presence of SQSTM1 signals (Figure S2A). We treated NC, Atg2a KD and DKD cells with Baf in the presence or absence of wortmannin. Our results showed that in Atg2a KD cells, wortmannin treatment did not disrupt the accumulation of LC3 puncta (Figure S2B,C). Conversely, in Atg2a Atg2b DKD cells, LC3 puncta were abolished by wortmannin treatment, while SQSTM1 puncta persisted (Figure S2B,C).

STX17, an autophagosomal Qa SNARE, along with its transmembrane domain (STX17[TM], comprising amino acids 229–275 of STX17), has been identified as a marker for closed autophagosomes. In control cells, GFP-STX17 and STX17(TM)-GFP primarily exhibited a diffuse localization on the ER under nutrient-rich conditions. In siAtg2a cells, GFP-STX17 and STX17(TM)-GFP formed numerous dot- and ring-like structures that were positive for LC3, while the LC3 dots in Atg2a Atg2b DKD cells were negative for GFP-STX17 and STX17(TM)-GFP (Figure S2D,E). These data suggest that in contrast to the function of ATG2 proteins in generating closed autophagosomes, ATG2A itself may have a unique function downstream of autophagosome formation.

To directly examine the role of ATG2A in autophagosome-lysosome fusion, we inhibited lysosomal degradation in Atg2a KD cells by addition of Baf. The results showed that endogenous LC3 puncta partially colocalized with RFP-RAB7-labeled lysosomes in control cells after EBSS starvation for 4 h. However, the colocalization of LC3 dots with RFP-RAB7-stained structures in Atg2a KD cells was greatly reduced compared to control cells (Figure 2F,G), which suggests a defect in autophagosome fusion with late endosomes and lysosomes. To investigate this, we used TEM to visualize starved cells treated with chloroquine (CQ), which blocks lysosomal degradation. The number of autolysosomes was significantly reduced while the number of autophagosomes was increased in siAtg2a cells compared to control cells (Figure 2H,I). Combined, these data suggest that ATG2A plays a critical role in maturation of autophagosomes to autolysosomes.

ATG2A acts as a tether for fusion of autophagosomes with late endosomes/lysosomes

Autophagosome-lysosome fusion requires coordination among SNARE, tether and RAB proteins [6]. The autophagosomal SNARE complex includes autophagosomal STX17 (Qa), SNAP29 (Qbc) and lysosome-localized VAMP8 (R) [29]. ATG2A was shown to tether liposomes in vitro [25]. We speculated that ATG2A may function as a tether for autophagosome-lysosome fusion. Consistent with our hypothesis, we observed that ATG2A interacted with the SNARE proteins STX17, SNAP29 and VAMP8 in co-immunoprecipitation (co-IP) assays in N2a cells (Figure 3A). To explore the function of these interactions, we detected the formation of SNARE complexes in Atg2a KD cells by co-IP. siAtg2a did not affect the formation of the STX17-SNAP29 complex, but it dramatically suppressed the interaction of STX17 with VAMP8 (Figure 3B,C). We then performed live-cell imaging to visualize the localization of ATG2A in N2a cells during starvation. We used a C-terminal truncation of ATG2A, GFP-ATG2A(Δ1830–1938), to prevent ATG2A from targeting to lipid droplets while preserving its localization on phagophores and function in autophagy [30]. The results showed that GFP-ATG2A(Δ1830–1938) left LC3-positive puncta before they fused with lysosomes, similar to ATG2B-GFP (Figure S3A,B). Using current techniques, it is very challenging to obtain clear fluorescence images of N2a cells. Therefore, it is possible that a low level of ATG2A is recruited to (or remains on) sealed autophagosomes for subsequent fusion with lysosomes. These data indicate that ATG2A has a previously unidentified function in mediating autophagosome-lysosome fusion by promoting and/or stabilizing SNARE assembly.

Figure 3.

ATG2A acts as a tether mediating autophagosome-lysosome fusion, assisted by WDR45 and WDR45B. (A) Endogenous STX17, SNAP29 and VAMP8 are precipitated by GFP-ATG2A in a GFP-Trap assay of N2a cells. (B and C) In a GFP-Trap assay, levels of endogenous VAMP8 precipitated by GFP-STX17 are dramatically decreased in siAtg2a N2a cells compared to cells treated with negative control (NC) siRNA, while levels of endogenous SNAP29 precipitated by GFP-STX17 in siAtg2a cell are unchanged (B). Quantification of SNAP29 and VAMP8 levels (normalized by GFP-STX17 levels) is shown as mean ± SEM (n = 3) (C). The levels in NC cells are set to 1.0. *, p < 0.05; ****, p < 0.0001. (D) Endogenous ATG2A is precipitated by WDR45-GFP and WDR45B-GFP in a GFP-Trap assay of N2a cells. (E and F) In a GFP-Trap assay, levels of endogenous ATG2A precipitated by GFP-STX17 are not obviously changed in wdr45 wdr45b DKO N2a cells compared to control cells (E). Quantification of ATG2A levels (normalized by GFP-STX17 levels) is shown as mean ± SEM (n = 4) (F). The level in control cells is set to 1.0. N.S., not significant. (G and H) In a GFP-Trap assay, levels of endogenous ATG2A precipitated by GFP-SNAP29 are reduced in wdr45 wdr45b DKO N2a cells compared to control cells. The reduction is rescued by reintroduction of WDR45-mCherry (G). Quantification of ATG2A levels (normalized by GFP-SNAP29 levels) is shown as mean ± SEM (n = 3) (H). The level in control cells is set to 1.0. *, p < 0.05. (I and J) In a GFP-Trap assay, much lower levels of endogenous ATG2A are precipitated by GFP-VAMP8 in wdr45 wdr45b DKO N2a cells than in control cells. The reduction is rescued by reintroduction of WDR45-mCherry (I). Quantification of ATG2A levels (normalized by GFP-VAMP8 levels) is shown as mean ± SEM (n = 3) (J). The level in control cells is set to 1.0. **, p < 0.01. (K and L) Immunoblots of ATG2A, SQSTM1 and LC3 show dramatically increased levels of SQSTM1 and LC3-II and decreased levels of ATG2A in wdr45 wdr45b DKO N2a cells. These changes are reversed by reintroduction of wild-type (WT) WDR45 but not its ATG2-binding mutant (K). Quantification of ATG2A, SQSTM1 and LC3-II levels (normalized by ACTB levels) is shown as mean ± SEM (n = 3) (L). The level in control cells is set to 1.0. **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; N.S., not significant. (M and N) Immunostaining with anti-SQSTM1 and anti-LC3 antibodies shows that SQSTM1⁺ and LC3⁺ puncta accumulate in wdr45 wdr45b DKO N2a cells, compared to control (ctrl) cells. These defects are alleviated by reintroduction of wild-type (WT) WDR45 but not its ATG2-binding mutant (M). Numbers of SQSTM1⁺ and LC3⁺ puncta from 3 biological repeats are quantified (ctrl GFP, n = 45; DKO GFP, n = 42; DKO WDR45-GFP, n = 40; DKO WDR45 mutant-gfp, n = 45) and shown as mean ± SEM (N). ****, p < 0.0001; N.S., not significant. Scale bars: 5 μm; inserts, 2 μm. (O) A model for the role of ATG2A in mediating autophagosome-lysosome fusion.

WIPI and WDR45 proteins facilitate the lipid transfer activity of ATG2A, possibly by enhancing tethering between vesicles [21]. Thus, we investigated whether WDR45 WDR45B help ATG2A to bridge autophagosomes with lysosomes. As previously reported [24], both WDR45 and WDR45B bound to ATG2A, and WDR45 showed a much stronger binding activity than WDR45B, in N2a cells (Figure 3D). We found that in wdr45 wdr45b double knockout (DKO) cells, ATG2A precipitated comparable levels of STX17 (Figure 3E,F), but the binding of ATG2A to SNAP29 and VAMP8 was significantly reduced in DKO cells compared to controls (Figure 3G–J). Moreover, the reduction in the level of ATG2A trapped by GFP-SNAP29 or GFP-VAMP8 was rescued by reintroducing WDR45 (Figure 3G–J). Furthermore, the autophagy defects in wdr45 wdr45b DKO cells, including increased SQSTM1 and LC3-II levels, accumulation of SQSTM1 and LC3 puncta, and compromised colocalization of RAB7 and LC3, were rescued by re-expression of wild-type (WT) WDR45, but not the WDR45 mutant (N61K D84G) which is deficient in ATG2 binding (Figure 3K–N and S3C-E) [31]. Thus, these data indicate that WDR45 and WDR45B function in assisting ATG2A to mediate fusion between autophagosomes and lysosomes.

Interestingly, we observed that the ATG2A levels were decreased in wdr45 wdr45b DKO cells (Figure 3E,G,I,K,L), as previously reported in HEK293T cells [32]. Overexpression of WT WDR45, but not the ATG2-binding mutant, rescued ATG2A levels in wdr45 wdr45b DKO (Figure 3K,L). Similarly, levels of WDR45 and WDR45B were decreased in siAtg2a cells (Figure S3F,G). These results suggest that, similar to non-neural cells, formation of ATG2A-WDR45 or -WDR45B complexes helps stabilize these proteins in neural cells. Reduced ATG2A levels may contribute to impaired autophagosome-lysosome fusion in cells lacking Wdr45 and Wdr45b.

The above results suggest that the autophagy defect in cells deficient in Wdr45 and Wdr45b may be due to the reduced ATG2A-SNARE interaction. ATG2A retains weak binding to SNAREs in wdr45 wdr45b DKO cells, which may be due to the interaction of ATG2A with Rab7 (Figure S3H). We examined whether overexpression of ATG2A ameliorated the defect in wdr45 wdr45b DKO cells. Using TEM, we found that increased levels of ATG2A rescued the autophagy defect in wdr45 wdr45b DKO cells, as indicated by the recovery of autolysosome numbers in ATG2A-overexpressing DKO cells (Figure S3I,J). Moreover, we observed that the accumulation of ATG9A puncta in wdr45 wdr45b DKO cells was rescued by overexpression of GFP-ATG2A, but not by ATG2B-GFP (Figure S3K). These results suggest that ATG2A collaborates with WDR45 or WDR45B at the autophagosome maturation step, and the requirement for WDR45 and WDR45B may be partially overcome by elevated levels of ATG2A.

EPG5 is critical for efficient assembly of SNARE complexes, which mediate autophagosome-late endosome or lysosome fusion [33]. To examine the contribution of ATG2A and EPG5 as tethers in autophagosome maturation, we analyzed the autophagy defects in Atg2a and Epg5 single KD and DKD cells. Compared to single KD cells, the accumulation of SQSTM1 and LC3-II was much more dramatic in Atg2a and Epg5 DKD cells (Figure S3L,M). In addition, EPG5 was immunoprecipitated by GFP-ATG2A (Figure S3N), which suggests that ATG2A and EPG5 May be present in the same SNARE-tether complex. Taken together, our results indicate that two tether proteins, ATG2A and EPG5, function redundantly to facilitate autophagosome-lysosome fusion.

Discussion

Initially, yeast Atg2 was found to regulate the retrieval of Atg9 proteins [7,8]. In mammalian cells, ATG9 also accumulates at LC3⁺ structures in cells lacking both ATG2A and ATG2B, but not when either gene is individually depleted [20]. Recent studies have shown that Atg2 mediates ER-PG contacts [9,34]. Both yeast Atg2 and mammalian ATG2 possess lipid transfer activity, which is facilitated by Atg18 and WIPIs and WDR45s, respectively, possibly through enhancing the membrane tethering of ATG2 with PtdIns3P-containing liposomes [11,21–23]. Hence, ATG2 has been proposed to mediate bulk delivery of glycerophospholipids from the ER to the PG for PG expansion with the assistance of WIPIs and WDR45s.

In this study, we demonstrate a novel function of ATG2A, but not ATG2B, in tethering and fusion of autophagosomes with late endosomes and lysosomes (Figure 3O). In Atg2a KD, but not Atg2b KD, N2a cells the autophagic flux is dampened. Distinct from Atg2a Atg2b DKD cells, closed autophagosomes are formed and autophagy is blocked at the autophagosome-lysosome fusion step in Atg2a single KD cells. ATG2A interacts with SNARE proteins, including STX17, SNAP29 and VAMP8. Atg2a deficiency does not affect the formation of Qabc SNARE complexes, but greatly reduces their interaction with VAMP8. The ATG2A-WDR45 complex forms a 20-nm-long rod structure [25]. This is reminiscent of the HOPS complex, an autophagosome-lysosome tether that possesses a 30-nm elongated seahorse-like structure [35]. In vitro liposome tethering assays demonstrated that binding of WDR45 on one end of ATG2 enables it to bridge PtdIns3P-containing vesicles with PtdIns3P-free vesicles [25]. ATG2 tethers large unilamellar vesicles/LUVs much less efficiently than small unilamellar vesicles/SUVs [21,22]. We found that without WDR45 and WDR45B, the interaction of ATG2A with SNAP29 and VAMP8 is suppressed. This suggests that the autophagosome-lysosome tether function of ATG2A is also facilitated by WDR45 and WDR45B, possibly because WDR45 and WDR45B mediate the targeting of ATG2A to the two organelles and/or act as scaffolding proteins to stabilize ATG2A and SNARE interactions (Figure 3O). Autophagosome-lysosome fusion requires PtdIns3P on autophagosomes [6,36], which may be the prerequisite for retargeting the ATG2A-WDR45 or -WDR45B complex. In addition, we found that autophagosomes in siAtg2a cells, similar to in wdr45 wdr45b DKO cells, are smaller in size, which provides in vivo evidence for the lipid transfer activity of ATG2 during PG expansion.

Consistent with the previous fundings that depletion of WDR45 and WDR45B leads to decreased levels of ATG2A, but not ATG2B, in HEK293T cells [32], we also observed reduced amounts of ATG2A in wdr45 wdr45b DKO N2a cells. Additionally, ATG2A ATG2B DKO results in dramatically lower WDR45 levels in HEK293T cells [32] and we found similar results in siAtg2a neural cells. These observations suggest that the formation of ATG2A-WDR45 or -WDR45B complexes may be required for stabilization of these proteins. The decreased levels of ATG2A in wdr45 wdr45b DKO cells may contribute to the autophagosome-lysosome fusion defects in these cells. Indeed, we found that overexpression of ATG2A rescued both the smaller autophagosome size and fusion defects in wdr45 wdr45b DKO cells. This suggests that a large amount of ATG2A overcomes the lipid transfer and membrane tethering defects caused by depletion of WDR45 and WDR45B. Similarly, high expression of the ATG2A lipid transfer domain can rescue the autophagy defect in ATG2A ATG2B DKO cells [23]. A conserved LIR motif is present in ATG2A and ATG2B, and mutation of the motif affects autophagosome formation [37], which indicates that GABARAP binding also contributes to the autophagosome retention of ATG2. On the lysosomal side, binding to SNARE proteins and RAB7 May help ATG2A targeting in addition to WDR45 and WDR45B.

siAtg2a, but not siAtg2b, affects autophagosome maturation in N2a cells, while simultaneous depletion of Atg2a and Atg2b leads to defective autophagosome closure. This suggests that ATG2B may have a more significant role in forming closed autophagosomes. It is also possible that ATG2A and ATG2B function redundantly in the process of phagophore expansion and/or closure. We cannot rule out the possibility that yeast Atg2 and Atg18 have a similar function in autophagosome maturation, which may be masked by their upstream role in autophagosome formation.

Our results also indicate that like WDR45 and WDR45B, ATG2A alone is specifically required for autophagy in neural cells [18]. Autophagosome maturation involves multi-stage fusion of autophagosomes with vesicles derived from endolysosomal compartments [6], which are distinct in different cell types [38–40], possibly due to the varied composition and spatiotemporal distribution of endolysosomal compartments. In neurons, there is a spatial gradient of endolysosomal compartments with increased degradative activity from the distal regions to the soma [41,42]. Neuronal autophagosomes are mainly formed at the distal tip of the axon and undergo sequential maturation accompanied by retrograde transport to the soma for lysosomal fusion and degradation [43,44]. Thus, we speculate that for the precise fusion of autophagosomes with diverse endolysosomal compartments in neurons, more tethering factors are required, including ATG2A, EPG5, and perhaps the HOPS complex. It is also possible that these proteins form a complex to modulate the fusion process. The autophagy defect in epg5 KO mice is also more prominent in the central nervous system [45]. Mutations in the human WDR45 and WDR45B genes are linked with the human neurological diseases beta-propeller protein-associated neurodegeneration/BPAN [46–48] and intellectual disability/ID [34,49,50], respectively. We previously demonstrated that overexpression of ATG2A partially rescued the smaller autophagosomes in wdr45 wdr45b DKO cells. In this study, we further discovered that ectopic expression of ATG2A also ameliorates the autophagosome-lysosome fusion defects in Wdr45/45-deficient cells. These findings highlight the potential therapeutic role of ATG2A in addressing autophagic dysfunctions associated with WDR45 and WDR45B deficiencies.

Our findings underscore the critical role of ATG2A in autophagosome maturation, particularly in neural cells. ATG2A, but not ATG2B, is essential for the tethering and fusion of autophagosomes with late endosomes or lysosomes. The ability of ATG2A to modulate lipid transfer and membrane tethering highlights its crucial role in maintaining autophagic integrity. These insights contribute to our understanding of the molecular mechanisms governing autophagosome maturation.

Materials and methods

Plasmids

RFP-GFP-LC3 was kindly given by Dr. Tamotsu Yoshimori (Osaka University) [51]. GFP-ATG2A was generated by cloning human ATG2A into the pEGFP-C1 vector (Clontech, 6084–1), and ATG2B-GFP was generated by cloning rat Atg2b into the pEGFP-N1 vector (Addgene, 172281; deposited by Antony K. Chen). STX17(TM)-GFP was generated by insertion of human STX17 aa 229–275 into pEGFP-N1. WDR45 mutant (N61K D84G)-GFP were constructed by PCR-based mutagenesis from WDR45-GFP. GFP-ATG2A(Δ1830–1938), RNAi-resistant GFP-ATG2A^r, ATG2B^r-GFP and GFP-EPG5^r, were generated by PCR-based mutagenesis from original plasmids. The other plasmids used in this study, namely GFP-LC3, SNAP-LC3, BFP-LC3, RFP-RAB7, GFP-RAB7, LAMP1-BFP, WDR45-GFP, WDR45-mCherry, GFP-EPG5, GFP-STX17, GFP-SNAP29 and GFP-VAMP8, have been described previously [14,19,33].

Antibodies

The following primary antibodies were used in this study: rabbit anti-ATG2A (Cell Signaling Technology, 15011), rabbit anti-ATG2B (Sigma, HPA019665), rabbit anti-LC3B (Sigma, L7543; for immunoblotting), mouse anti-LC3B (MBL, M152–3; for immunostaining), rabbit anti-STX17 (MBL, PM076), rabbit anti-SNAP29 (Abcam, ab138500), rabbit anti-VAMP8 (Abcam, ab76021), rabbit anti-ATG9 (MBL, PD042), mouse anti-GFP (Roche, 11814460001), rabbit anti-mCherry (Genetex, GTX59788), mouse anti-Flag (Sigma, F180), rabbit anti-MYC (Huaxingbio, HX1821), mouse anti-GAPDH (Proteintech, 60004–1-Ig), and mouse anti-ACTB (Proteintech, 60008–1-Ig). Mouse anti-EPG5 was generated as previously reported [33].

The following secondary antibodies were used in this study: goat anti-rabbit IgG(H+L)-HRP (EASYBIO, BE0101), goat anti-mouse IgG(H+L)-HRP (EASYBIO, BE0102), goat anti-mouse-rhodamine/TRITC (Jackson, 115-025-00), goat anti-rabbit-rhodamine (Jackson, 111-025-003), goat anti-rabbit-fluorescein/FITC (Jackson, 111-095-003), and goat anti-mouse-fluorescein (Jackson, 115-095-003).

Cell lines

Neuro-2a (N2a; ATCC, CCL-131) cells were grown with DMEM/F12 (Invitrogen, 11320082) supplemented with 10% fetal bovine serum (FBS; Hyclone, SH30084.0) and penicillin-streptomycin, under conditions of 37°C and 5% CO₂. HT22 (Pricella, CL-0697), HeLa (ATCC, CCL-2) and 293T (ATCC, CRL-11268) cells were cultured in DMEM (Hyclone, SH30022.01B), with 10% FBS with penicillin-streptomycin. All cell lines were verified as mycoplasma-free and none were listed in the International Cell Line Authentication Committee/ICLAC database as misidentified or cross-contaminated. The wdr45 wdr45b DKO N2a cells were generated as previously reported [19]. To induce autophagy, cells were treated with EBSS (Thermo Fisher, 24010043) for the indicated time. To block the activity of PtdIns3K, cells were exposed to 2 μM wortmannin (Life Technologies, PHZ1301) for 4 h. To inhibit autophagy, cells were treated with either 200 nM bafilomycin A₁ (Baf; Sigma, B1793) or 50 µM chloroquine (Sigma, C6628) for 6 h.

Transfection and RNA interference

Cells were transfected with the indicated plasmids using Lipofectamine 2000 (Life Technologies, 12566014). For RNA interference, cells were transfected with negative control (NC) or siRNA oligos using Lipofectamine RNAi MAX (Life Technologies, 13778150), and collected 72 h post-transfection.

siRNAs oligos were purchased from GenePharma. The sequences of siRNAs used in this study were listed below:

NC, 5’-UUCUCCGAACGUGUCACGUTT-3’;

Mouse Atg2a, 5’-GCUACUUGCUACAGCACUA-3’;

Mouse Atg2b, 5’-GCUGGAACCUACAAUUUAU-3’;

Mouse Epg5, 5’-GGCUGACUGACUUAGAAAU-3’;

Human ATG2A, 5’-GCAUUCCCAGUUGUUGGAGUUCCUA-3’;

Human ATG2B, 5’-AGGUCUCUCUUGUCUGGCAUCUUUA-3’.

Hippocampal neuron culture and transfection

Primary hippocampal cultures were prepared from E17.5 mouse embryos and grown on glass coverslips coated with 1 mg/ml poly-D-lysine (Sigma, P7280) at 37°C and 5% CO₂, for 14–21 days. Briefly, hippocampi were dissected in dishes filled with HBSS, and dissociated by treatment with trypsin (0.25%; Thermo Fisher, 15090046) plus DNase (1 mg/ml; Sigma, D4527-40KU) at 37°C for 15 min. Subsequently, neurons were plated in Neurobasal medium (Thermo Fisher, 21103049) supplemented with B27 (Thermo Fisher, 1750404), GlutaMax (Thermo Fisher, 35050061) and 10% FBS (Hyclone, SH30070.03). After 4 h, the medium was replaced with Neurobasal medium containing B27, GlutaMax and 1% FBS. After every 3 days of culture, half the volume of the medium was changed with Neurobasal medium supplemented with B27 and GlutaMax. AAV-mediated infection was performed at days in vitro/DIV 7, and experiments were conducted at DIV 12–14.

Adeno-associated virus (aav)-mediated knockdown (KD)

For AAV-mediated KD of Atg2a or Atg2b, AAV vectors were used encoding shRNA specific for Atg2a or Atg2b, or a non-targeting control shRNA. The shRNA sequences were cloned into pAAV2-U6-EF1A-copGFP vector, which allows for the expression of shRNA under the control of the U6 promoter and GFP under the control of the CMV promoter for visualization. The following sequences were used:

NC,

5’- TGCGAATACGCCCACGCGATCTCGAGATCGCGTGGGCGTATTCGCA-3’;

Mouse Atg2a,

5’- GCTACTTGCTACAGCACTATCTCGAGATAGTGCTGTAGCAAGTAGC-3’;

Mouse Atg2b,

5’- GCTGGAACCTACAATTTATTCTCGAGAATAAATTGTAGGTTCCAGC-3’.

AAV particles were produced by co-transfecting HEK293T cells with the shRNA plasmids and packaging plasmids (pHelper [HGT1] and rep/cap, kindly provided by Dr. Tian Ruilin at South University of Science and Technology, China) using Lipofectamine 2000. Ninety-six h after transfection, all cells and medium were collected and mixed with 3 ml chloroform and 7.6 ml 5 M NaCl. After vortexing for 5 min, samples were then centrifuged at 3000 g, 4°C. The aqueous phase was transferred into a new 50-ml conical tube, and 9.4 ml of 50% (v:v) PEG 8000 (Biosharp, BS138) were added. The mixture was incubated on ice for 1 h, then centrifuged at 3000 g for 30 min at 4°C. The pellets were resuspended with 1.4 ml HEPES buffer (50 mm, pH 8.0) with 14 μl DNase (2000 U/ml; New England Biolabs, M0303S) and 1.4 μl RNase A (10 μg/μl; Tiangen, rt405–02) and incubated at 37°C for 20 min. Chloroform was then added at 1:1 (v:v), and the samples were vortexed and centrifuged at 3000 g for 5 min at 4°C. The upper aqueous phase was collected, re-mixed with chloroform, centrifuged and collected for another 3 times. Samples were loaded onto Amicon Ultra Centrifugal Filters (Merck Millipore, UFC510024) and centrifuged at 14,000 g for 5 min at room temperature. The filters were then washed with 400 μl Dulbecco’s phosphate-buffered saline/DPBS (Thermo Fisher, C14190500) for 4 times. Finally, the filter was placed upside down in a new tube and centrifuged at 1000 g for 2 min at room temperature to harvest AAV.

Immunostaining

Cells were grown on coverslips (Fisherbrand, 12-545-83), and then fixed with 4% paraformaldehyde/PFA for 30 min. Permeabilization was carried out using 10 μg/ml digitonin (Sigma, D141) for 20 min at room temperature. Subsequently, cells were blocked with 1% bovine serum albumin (BSA; Aladdin, A116563) for 60 min, and then incubated with the specified primary antibodies (diluted in 1% BSA) overnight at 4°C. After rinsing with PBS (140 mm NaCl, 2.7 mm KCl, 10 mm Na₂HPO₄, 1.8 mm KH₂PO₄, pH 7.5), cells were stained with fluorescently-labeled secondary antibodies for 1 h at room temperature. Coverslips were then mounted with DAPI in 50% glycerol and images were acquired with a confocal microscope (Zeiss, LSM 980 Meta plus Zeiss Axiovert zoom) with a 63×/1.40 oil-immersion objective lens (Zeiss, Plan-Apochromatlan) and a point detector (Zeiss, GaAsP PMT). Zeiss ZEN software (blue edition) was utilized for image acquisition.

Live-cell imaging

For live-cell imaging, cells were seeded on glass-bottom dishes (NEST, 706001) and transfected with the indicated plasmids. Following 24 h transfection, cells transfected with SNAP-LC3 were incubated with 60 nM SNAP-Cell 647 (New England Biolabs, S9102S) for 30 min before imaging. Then cells were starved with EBSS and imaged at 37°C and 5% CO₂ using a confocal microscope (Zeiss, LSM 980 Meta plus Zeiss Axiovert zoom) with a 63×/1.40 oil-immersion objective lens (Zeiss, Plan-Apochromatlan) and a point detector (Zeiss, GaAsP PMT). Image sequences were captured every 15 s for 1 h.

Co-immunoprecipitation (co-IP) and immunoblotting

For co-IP experiments, cells were transfected with the specified plasmids for 24–48 h or with siRNAs for 72 h. After transfection, cells were lysed using a lysis buffer containing 20 mm HEPES, pH 7.5, 150 mm NaCl, 1 mm EDTA, and 0.1% Triton X-100 (Sigma, X100) supplemented with a protease inhibitor cocktail (Roche, 11836170001). The lysates were kept on ice for 30 min, followed by centrifugation at 15,871 g for 10 mins at 4°C. The supernatants were then transferred to new tubes and incubated with GFP-Trap (Shenzhen KT Life Technology, KTSM1334) agarose beads at 4°C for 1 h. The bound proteins were eluted with SDS sample buffer and subjected to immunoblotting analysis.

For the immunoblotting procedure, cells were lysed in a buffer containing 20 mm HEPES, pH 7.5, 150 mm NaCl, 1 mm EDTA, 0.1% Triton X-100, and a protease inhibitor cocktail. The lysates were incubated on ice for 30 min, and then centrifuged at 15,871 g for 10 min at 4°C. The supernatants were collected and mixed with SDS sample buffer before being subjected to SDS-PAGE. The proteins were transferred onto hydrophobic PVDF membranes with a 0.45-µm pore size (Millipore, IPVH00010). Indicated primary and secondary antibodies were used for protein detection and signals were visualized using the 5200SF imaging system (Tanon, 5200SF).

Transmission electron microscopy (TEM)

For TEM analysis, cells were initially fixed in 2.5% glutaraldehyde in PBS for 2 h at room temperature. Following this, samples were post-fixed with a solution containing 1% OsO₄ and 1% potassium ferrocyanide for 45 min. Then cells were rinsed with double-distilled water (ddH₂O) and treated with thiocarbohydrazide/TCH solution for 30 min at room temperature. After another rinse with ddH₂O, the cells were placed in 1% OsO₄ for an additional 45 min at room temperature. Cells were washed again with ddH₂O before undergoing dehydration through a graded series of ethanol solutions. The dehydrated cells were embedded in epoxy EMBED-812 resin (Electron Microscopy Sciences, 14900). Ultrathin sections were prepared and stained with uranyl acetate and lead citrate. The samples were examined using a 120 kV electron microscope (FEI, Tecnai Spirit) at 100 kV at room temperature with a CCD camera (EMSIS, MoradaG3) using RADIUS (EMSIS) software.

Halo-LC3 assays

The Halo-LC3 assays were conducted following the method outlined previously [28]. Briefly, cells were transfected with Halo-LC3 for 24 h. Subsequently, cells were treated with 20 μM digitonin in KHM buffer (110 mm potassium acetate, 20 mm HEPES, 2 mm MgCl₂, pH 7.5) at 37°C for 2 min. Then cells were incubated with the membrane-impermeable ligand (MIL) HaloTag® Alexa Fluor® 660 ligand (Promega, G847A) at 37°C for 15 min. Post-incubation, the cells were fixed in 4% paraformaldehyde for 20 min, rinsed with PBS, and incubated with the membrane-permeable ligand (MPL) HaloTag® tetramethylrhodamine/TMR ligand (Promega, G8251) for 30 min at room temperature. Finally, the samples were mounted and examined using a confocal microscope (Zeiss, LSM 980 Meta plus Zeiss Axiovert zoom) with a 63×/1.40 oil-immersion objective lens (Zeiss, Plan-Apochromatlan) and a point detector (Zeiss, GaAsP PMT). Zeiss ZEN software (blue edition) was utilized for image acquisition.

Quantification and statistical analysis

Immunoblotting and coIP results are representative of three independent experiments. Sample size was estimated with preliminary experiments. Cells or images were chosen randomly for statistical analysis. Colocalization of two puncta was assessed manually, with more than 80% overlap between two fluorescent signals being classified as colocalization. Levene’s test was used to test the equality of variance. The normality of the data was assessed by the Shapiro-Wilk test. Specific statistical parameters, including n, SEM and SD, are detailed in the corresponding figure legends. Statistical significance was determined by one-way ANOVA analysis, with a p value below 0.05 considered significant.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Funding Statement

The work was supported by the Guangdong Innovative and Entrepreneurial Research Team Program [2021ZT09Y104]; National Natural Science Foundation of China [32222021]; National Natural Science Foundation of China [32170753]; National Key Research and Development Program of China [2021YFA1300800]; the Guangdong Program [2021QN02Y378]; R&D Program of Guangzhou Laboratory [GZNL2024A01008]; the Shenzhen Science and Technology Program [ZDSYS20220402111000001]; the Shenzhen Talent Program [KQTD20210811090115021].

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data will be available upon reasonable request.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15548627.2025.2479427