Unconventional Initiation of PINK1/Parkin Mitophagy by Optineurin

Cargo sequestration is a fundamental step of selective autophagy in which cells generate a double membrane structure termed an autophagosome on the surface of cargoes. NDP52, TAX1BP1 and p62 bind FIP200 which recruits the ULK1/2 complex to initiate autophagosome formation on cargoes. How OPTN initiates autophagosome formation during selective autophagy remains unknown despite its importance in neurodegeneration. Here, we uncover an unconventional path of PINK1/Parkin mitophagy initiation by OPTN that does not begin with FIP200 binding nor require the ULK1/2 kinases. Using gene-edited cell lines and in vitro reconstitutions, we show that OPTN utilizes the kinase TBK1 which binds directly to the class III phosphatidylinositol 3-kinase complex I to initiate mitophagy. During NDP52 mitophagy initiation, TBK1 is functionally redundant with ULK1/2, classifying TBK1’s role as a selective autophagy initiating kinase. Overall, this work reveals that OPTN mitophagy initiation is mechanistically distinct and highlights the mechanistic plasticity of selective autophagy pathways.


INTRODUCTION
Macroautophagy (hereafter autophagy) is a highly conserved catabolic process which degrades and recycles intracellular components. Autophagy is cytoprotective and plays essential roles in maintaining cellular homeostasis through the removal of harmful materials such as protein aggregates, dysfunctional organelles, and invading pathogens (Dikic and Elazar, 2018;Melia et al., 2020). Cargoes destined for autophagic degradation are sequestered by double-membrane vesicles termed autophagosomes before being delivered to lysosomes for degradation.
Autophagosomes are generated through the formation of a double membrane precursor termed the phagophore that arises on a platform of endoplasmic reticulum termed the omegasome (Axe et al., 2008;Hayashi-Nishino et al., 2009;Uemura et al., 2014).
A general hierarchy of autophagy machinery recruitment during autophagosome formation has been established and typically begins with concurrent recruitment of ATG9A vesicles and the ULK1/2 complex, the latter consisting of the kinase ULK1 or its homologue ULK2, FIP200, ATG13 and ATG101 (Dikic and Elazar, 2018;Itakura and Mizushima, 2010;Kishi-Itakura et al., 2014;Melia et al., 2020;Mizushima et al., 2011;Yang and Klionsky, 2010). The ULK1/2 complex then promotes the recruitment and activation of the class III phosphatidylinositol 3-kinase complex I (hereafter PI3K complex). The PI3K complex consists of ATG14, the lipid kinase VPS34, VPS15, and Beclin, which together function to generate the lipid phosphatidylinositol 3-phosphate (PI(3)P) on phagophore membranes. This enables the recruitment of PI(3)P effector proteins including the WIPIs (WD-repeat protein interacting with phosphoinositides) that promote phagophore expansion. WIPI2 plays a role in recruiting the ATG8 conjugation machinery through binding to the ATG16L1 subunit (Dooley et al., 2014), resulting in the attachment of ubiquitinlike ATG8 family proteins onto phosphatidylethanolamine (PE) on autophagosomal membranes (Ichimura et al., 2000;Mizushima et al., 1998). The similarity of ATG8 conjugation to ubiquitination has led to the process being termed atg8ylation, which describes the conjugation of ATG8s to lipid or protein (Agrotis et al., 2019;Deretic and Lazarou, 2022;Kumar et al., 2021;Nguyen et al., 2021). The ATG8 family consists of six members divided equally across the LC3 and GABARAP subfamilies. Attachment of ATG8s to PE may function to expand autophagosomal membranes, at least in part, via promoting membrane tethering and hemifusion (Nakatogawa et al., 2007) and by acting as a protein platform to recruit autophagy factors (Lamark and Johansen, 2021;Martens and Fracchiolla, 2020). WIPI1 and WIPI4 can function to promote membrane expansion through their interactions with ATG2 proteins that transfer phospholipids from the ER to the growing phagophore (Chowdhury et al., 2018;Kotani et al., 2018;Maeda et al., 2019;Osawa et al., 2019;Valverde et al., 2019;Zheng et al., 2017). ATG9A facilitates ATG2-mediated lipid transfer by enabling ER-phagophore contacts and acting as a lipid scramblase by re-arranging lipids from the cytoplasmic leaflet to the luminal leaflet of the growing phagophore (Ghanbarpour et al., 2021;Gomez-Sanchez et al., 2018;Maeda et al., 2020;Matoba et al., 2020). ATG9A vesicles, which are trafficked to autophagosome formation sites by ARFIP2, the immune disease protein LRBA, and the ATG4 family (Judith et al., 2019;Nguyen et al., 2021), might also provide the membrane seed for initial phagophore formation (Judith et al., 2019;Karanasios et al., 2013;Mari et al., 2010;Sawa-Makarska et al., 2020).
Selective targeting of autophagic cargoes, including the ER, invading pathogens, and mitochondria, necessitates recruitment of autophagy machineries to the surface of cargo to enable selective capture by autophagosomes. Selective recruitment of autophagy machineries is conducted by receptor proteins resident on the surface of cargoes or via adaptor proteins that recognize 'eat me' signals on the surface of cargo in the form of ubiquitin chains (Gatica et al., 2018;Goodall et al., 2022;Mizushima, 2020). Several autophagy adaptors have been identified and amongst the most highly studied are p62, NBR1, NDP52, TAX1BP1, and OPTN (Lamark and Johansen, 2021). These autophagy adaptors appear to use a common mechanism of engaging the autophagy machinery by recruiting the ULK1/2 complex through binding to the FIP200 subunit to initiate the cascade of autophagosome formation (Ravenhill et al., 2019;Turco et al., 2021;Turco et al., 2019;Vargas et al., 2019). In addition to the ULK1 and ULK2 kinases, TBK1 is another kinase that has been linked to selective autophagy pathways (Heo et al., 2015;Heo et al., 2018;Lazarou et al., 2015;Moore and Holzbaur, 2016;Richter et al., 2016;Vargas et al., 2019), although its precise role during selective autophagy driven by each individual autophagy adaptor remains unclear.
Another study reported that an OPTN-ATG9A interaction, that is independent of TBK1, is crucial for PINK1/Parkin mitophagy (Yamano et al., 2020). However, it is unclear if this interaction regulates autophagosome initiation or other downstream steps since ATG9A seems to be involved in multiple stages during autophagosome biogenesis (Gomez-Sanchez et al., 2021). Nevertheless, for autophagy adaptors, the conventional path of selective autophagy initiation identified to date involves direct interaction with FIP200, as has been reported for NDP52 during mitophagy and xenophagy (Ravenhill et al., 2019;Vargas et al., 2019), and for p62 and TAX1BP1 during aggrephagy (Turco et al., 2021;Turco et al., 2019).
OPTN is mutated in human neurodegenerative diseases including amyotrophic lateral sclerosis and glaucoma (Maruyama et al., 2010;Rezaie et al., 2002), and unlike NDP52, OPTN is highly expressed in brain tissue (Lazarou et al., 2015), making it a high priority cargo adaptor to understand. In this study, we uncover OPTN's mechanism for initiating selective autophagy. Our work shows that OPTN follows an unconventional path by utilizing TBK1 to interact with the PI3K complex upstream of the ULK1/2 complex. The ULK1/2 kinase subunits were found to be dispensable for OPTN mediated mitophagy, but can contribute to NDP52 mediated mitophagy in a functionally redundant manner with TBK1, revealing TBK1 as a ULK1/2-like kinase in mitophagy. These results demonstrate the mechanistic divergence of mitophagy initiation by OPTN from NDP52 and other cargo adaptors, revealing an alternative path of selective autophagy initiation.

ULK1 and ULK2 are dispensable for PINK1/Parkin mitophagy
ULK1 has been reported to phosphorylate several autophagy factors that are important for autophagosome formation (Egan et al., 2015;Mercer et al., 2018;Mercer et al., 2021). Given the role of ULK1 and it relative ULK2 as the most upstream kinases of autophagy, we assessed their contribution to PINK1/Parkin mitophagy initiation mediated by either OPTN or NDP52.
Mitophagy was induced by treating cells with oligomycin and antimycin A (OA) and assessed by measuring the degradation of cytochrome c oxidase subunit II (COXII), a mitochondrial DNAencoded protein located within the inner membrane. As can be seen in Figures 1B and 1C, the loss of ULK1/2 did not prevent COXII degradation in GFP-NDP52 expressing ULK1/2 DKO/penta KO cells, and even slightly increased the levels of COXII degradation in GFP-OPTN expressing cells. To determine if the efficiency of mitophagy was affected in the absence of ULK1/2, the mtKeima mitophagy assay was employed (Katayama et al., 2011;Lazarou et al., 2015;Vargas et al., 2019). Consistent with the COXII degradation data in Figures 1B and 1C, there was no mitophagy defect in penta KO cells lacking ULK1/2 undergoing either GFP-OPTN or GFP-NDP52 driven mitophagy ( Figures 1D and 1E). Instead, mitophagy rates appeared to be slightly higher for GFP-OPTN mitophagy ( Figure 1D), and to a lesser degree the same was true for GFP-NDP52 ( Figure 1E). Analysis of mitochondrial recruitment of the early autophagosome marker WIPI2b showed that the loss of ULK1/2 did not decrease the formation of autophagosome precursors (Figures 1F and 1G). Consistent with the mitophagy measurements in Figures 1D and   1E, the loss of ULK1/2 resulted in more robust formation of WIPI2b autophagosome intermediates during GFP-OPTN-dependent mitophagy and to a lesser degree during GFP-NDP52-dependent mitophagy ( Figures 1F and 1G). Autophagosome formation and overall autophagosome morphology in the absence of ULK1/2 also appeared normal by electron microscopy (EM) analysis ( Figure 1H). These results show that despite the role of ULK1 and ULK2 as the most upstream kinases of autophagy, they unexpectedly are not essential for autophagosome formation and mitophagy mediated by either OPTN or NDP52.

The ULK1/2 complex subunits, FIP200 and ATG13, are important for PINK1/Parkin mitophagy
We next asked whether the other subunits of the ULK1/2 kinase complex, including FIP200 and ATG13, are required for PINK1/Parkin mitophagy. Consistent with their essential role in general mammalian autophagy (Hara et al., 2008;Kaizuka and Mizushima, 2016), knockout of FIP200 or ATG13 in penta KO cells ( Figures S2A and S2B), blocked PINK1/Parkin mitophagy mediated by either GFP-OPTN or GFP-NDP52, as measured by COXII degradation (Figures 2A and 2B). Next, the recruitment of ATG13 and FIP200 foci to mitochondria was analyzed in penta KO cells lacking FIP200 or ATG13 respectively. In the absence of FIP200, neither GFP-OPTN or GFP-NDP52 could recruit ATG13 foci to mitochondria upon mitophagy induction ( Figure 2C). Similarly, in the absence of ATG13, GFP-OPTN could not trigger mitochondrial recruitment of FIP200 foci ( Figure 2D). In contrast, GFP-NDP52 was capable of recruiting FIP200 foci to damaged mitochondria in the absence of ATG13 ( Figure 2D), which is consistent with the ability of NDP52 to directly bind FIP200 (Ravenhill et al., 2019;Vargas et al., 2019). These results show that the FIP200 and ATG13 subunits of the ULK1/2 complex are both important for OPTN-and NDP52dependent PINK1/Parkin mitophagy. Notably, the inability of OPTN to stably recruit FIP200 in the absence of ATG13 indicates that OPTN uses a mechanistically distinct form of mitophagy initiation from NDP52.
In the analyses of GFP-NDP52 mediated recruitment of ATG13 and FIP200 ( Figures 2C and 2D), we noted that GFP-NDP52 failed to form mitochondrial foci in the absence of FIP200. This result led us to hypothesize that NDP52-FIP200 interactions can form a positive feedback loop to allow further recruitment and therefore foci formation of NDP52 on damaged mitochondria. We tested this hypothesis by reconstituting NDP52 recruitment in vitro using recombinant proteins and a fluorescence microscopy-based bead assay (Abert and Martens, 2019). Recombinant mCherry (mCh)-NDP52 was robustly recruited to cargo mimetic GST-4xUb beads, and this then allowed the recruitment of FIP200-GFP ( Figure 2E), providing further confirmation that NDP52 directly interacts with FIP200 (Ravenhill et al., 2019;Shi et al., 2020a;Vargas et al., 2019). However, in the absence of FIP200, the levels of mCh-NDP52 recruited to the GST-4xUb beads was significantly reduced (Figures 2F and 2G), consistent with the failure to form mitochondrial foci in cells lacking FIP200 ( Figure 2C). The presence of FIP200 therefore enhances NDP52 recruitment to cargo, supporting the existence of a FIP200-dependent recruitment loop for NDP52.

TBK1 is essential for OPTN-mediated mitophagy, but functions redundantly with ULK1/2 during NDP52-mediated mitophagy
TBK1 has been reported to play a role in PINK1/Parkin mitophagy (Heo et al., 2015;Heo et al., 2018;Lazarou et al., 2015;Moore and Holzbaur, 2016;Richter et al., 2016;Vargas et al., 2019), but its role in mitophagy mediated individually by OPTN or NDP52 has not been explored. This is an important question because the two cargo receptors are not co-expressed in all cell types including neurons. In addition, given the finding that the ULK1/2 kinases are not essential for OPTN or NDP52 mediated mitophagy (Figure 1), further investigation into TBK1 was warranted.
First, we assessed the role of TBK1 kinase activity in recruiting the ULK1/2 complex by analyzing mitochondrial recruitment of the ATG13 subunit. Upon mitophagy induction, inhibition of TBK1 kinase activity using BX795 blocked mitochondrial recruitment of ATG13 in penta KO cells undergoing GFP-OPTN mediated mitophagy, but not in cells utilizing GFP-NDP52 for mitophagy ( Figure 3A). This result indicates that TBK1 kinase activity is required for initiation of OPTN but not NDP52 mediated mitophagy. For further validation, and to address whether the kinase activity or the physical presence of TBK1 is necessary, TBK1 knockouts were generated in the penta KO background ( Figure 3B), and used to analyze mitophagy via COXII degradation, and ULK1/2 complex recruitment via imaging of ATG13. In the absence of TBK1, GFP-OPTN failed to mediate COXII degradation (Figures 3C and 3D) and ATG13 recruitment to mitochondria (Figures 3E and 3F). In contrast, COXII degradation was unaffected by the loss of TBK1 during GFP-NDP52 dependent mitophagy ( Figures 3C and 3D), although mitochondrial recruitment of ATG13 was significantly reduced ( Figures 3E and 3F). Taken together, these results demonstrate that TBK1 is essential for the recruitment of the ULK1/2 complex and initiation of OPTN-mediated mitophagy, but is largely dispensable for NDP52 mediated mitophagy, although it appears to regulate the efficiency of ULK1/2 complex recruitment and therefore mitophagy initiation by NDP52, as recently reported (Vargas et al., 2019).
Neither of the ULK1/2 or TBK1 kinases were essential for NDP52-mediated mitophagy ( Figures   1-3). However, given that some structural similarities between TBK1 and the ULK1/2 complex have been reported (Shi et al., 2020b), we asked whether TBK1 and ULK1/2 might function redundantly during NDP52 mediated mitophagy. Mitophagy rates using the mtKeima assay were analyzed in ULK1/2 DKO/penta KO cells expressing either GFP-OPTN or GFP-NDP52 in the presence or absence of TBK1 inhibitor BX795. GFP-OPTN mediated mitophagy was completely inhibited upon addition of BX795 irrespective of the presence or absence of ULK1/2 (Figures 4A) supporting the absolute requirement of TBK1 kinase activity for OPTN dependent mitophagy. In contrast, TBK1 kinase inhibition with BX795 was only able to prevent NDP52 dependent mitophagy when ULK1/2 were absent ( Figure 4B). This demonstrates that TBK1 kinase activity compensates for the loss of ULK1/2 during NDP52 mediated mitophagy.
To further address the putative functional redundancy between TBK1 and ULK1/2 during NDP52 mediated mitophagy initiation, and to ensure no off-target effects of BX795 on other kinases, TBK1 was knocked out in the ULK1/2 DKO/penta KO background ( Figure 4C). The combined deletion of TBK1 and ULK1/2 resulted in GFP-NDP52 losing its ability to drive COXII degradation ( Figures 4D and 4E), demonstrating that indeed, ULK1/2 and TBK1 are functionally redundant in initiating NDP52 mediated mitophagy. Analysis of ATG13 recruitment to mitochondria showed that consistent with its direct binding to FIP200, GFP-NDP52 was still able to support recruitment of ATG13 in the absence ULK1/2 and TBK1 ( Figure 4F), demonstrating that a core complex consisting of FIP200, ATG13, and likely ATG101, remains intact (Jung et al., 2009;Shi et al., 2020b). However, the recruitment of WIPI2b, a PI(3)P effector protein, was completely abolished ( Figures 4G and 4H), pointing toward a defect in the activation or recruitment of the PI3K complex. Our results therefore indicate that TBK1 is functionally redundant with ULK1/2 in promoting the activation of PI3K complex, which is consistent with the role of ULK1/2 in enhancing activity of the PI3K complex during starvation induced autophagy (Mercer et al., 2018;Mercer et al., 2021;Wold et al., 2016).

OPTN initiates mitophagy through an unconventional mechanism that is dependent on TBK1
A recent structural analysis revealed that peptides of OPTN and FIP200 interact in vitro (Zhou et al., 2021), indicating that OPTN follows the canonical mode of selective autophagy activation of directly recruiting the ULK1/2 complex via FIP200. However, our results show that OPTN is unable to stably recruit FIP200 in the absence of ATG13 ( Figure 2D), arguing against FIP200 binding being the primary mitophagy initiation mechanism of OPTN. To directly address whether FIP200 binding, and therefore ULK1/2 complex recruitment, is the most upstream event of OPTN mitophagy initiation, we decided to block the next step of mitophagy involving PI3K complex recruitment. The rationale being that if the ULK1/2 complex is the most upstream event, then OPTN should retain the ability to recruit the ULK1/2 complex irrespective of downstream mitophagy steps. Knockout lines of ATG14, an essential subunit of the PI3K complex involved in autophagy, were generated in the penta KO background ( Figure 5A). Analysis of COXII degradation confirmed that ATG14 is essential for PINK1/Parkin mitophagy mediated by either GFP-OPTN or GFP-NDP52 (Figures 5B and 5C). We also confirmed that the loss of ATG14 blocked activation of the PI3KC3 complex as shown by the lack of mitochondrial recruitment of the PI(3)P effector protein WIPI2b ( Figure 5F).
Next, the recruitment of the ULK1/2 complex was analyzed by assessing ATG13 and FIP200 foci on mitochondria in the presence or absence of ATG14. During GFP-NDP52 mediated mitophagy, both ATG13 and FIP200 were recruited to damaged mitochondria irrespective of the presence or absence of ATG14 (Figures 5D, 5E and S4B), aligning with NDP52's ability to directly bind and recruit FIP200 upstream of the PI3K complex (Ravenhill et al., 2019;Vargas et al., 2019).
However, in direct contrast, cells undergoing GFP-OPTN mediated mitophagy failed to recruit ATG13 and FIP200 in cells lacking ATG14 ( Figures 5D, 5E and S4B). This result argues for an unconventional mechanism of mitophagy activation by OPTN, in which the PI3K complex is upstream of the ULK1/2 complex. Combined with the finding that TBK1 is essential for OPTN but not NDP52 mediated mitophagy, we conclude that mitophagy initiation by OPTN is mechanistically distinct from NDP52.

TBK1 directly binds to the PI3K complex
We next investigated whether OPTN can directly engage the PI3K complex by reconstituting mitophagy initiation in vitro. Ubiquitin coated cargo mimetic beads were incubated with a TBK1 phospho-mimetic form of OPTN, termed OPTN S2D (S177D, S473D (Chang et al., 2021)), that was used to assess recruitment of the PI3K complex. Phosphomimetic OPTN S2D failed to recruit fluorescently labeled PI3K complex (mCherry-ATG14) to GST4xUb beads ( Figure 6A), arguing against a direct binding model for OPTN mitophagy initiation. However, the PI3K complex was recruited upon the addition of GFP-TBK1 ( Figure 6A). This led us to assess whether TBK1 can directly bind to the PI3K complex. Indeed, GFP-TBK1 alone tethered to GFP-Trap beads was sufficient to recruit the PI3K complex ( Figure 6B). In vitro reconstitution of NDP52 mediated mitophagy initiation using ubiquitin coated beads was conducted next. The analysis showed that that NDP52 can directly recruit the ULK1 complex ( Figure 6C), and this was sufficient to then recruit the PI3K complex ( Figure 6D), indicating that purified ULK1 and PI3K complexes alone are sufficient for the two to directly interact. Another in vitro reconstitution in which the PI3K complex was immobilized onto RFP-Trap beads via mCherry-ATG14 confirmed that a direct biochemical interaction exists between the ULK1 and PI3K complexes ( Figure S5A).
The positioning of the PI3K complex upstream of the ULK1/2 complex during OPTN mediated mitophagy raised the question of how the ULK1/2 complex might be recruited by OPTN. Given that the PI3K and ULK1 complexes directly bind in vitro ( Figure S5A) and have been reported to interact in cells (Park et al., 2016), it was decided to explore this area further. Using human cells, it has been proposed that the N-terminal HORMA domain within ATG13 interacts with ATG14 (Park et al., 2016), while in yeast it plays a role in the recruitment of ATG14 (Jao et al., 2013). We therefore assessed whether the HORMA domain of ATG13 is required for recruitment of the ULK1/2 complex during OPTN mediated mitophagy. As can be seen in Figure 6E, deletion of the HORMA domain within ATG13 (ATG13 D1-198 ) prevented its recruitment to mitochondria during OPTN, but not NDP52, mediated mitophagy ( Figure 6E). The ability of NDP52 to recruit ATG13 D1-198 is explained by NDP52's direct binding to FIP200 which interacts with the Cterminus of ATG13 (Jung et al., 2009;Shi et al., 2020b). Collectively, our results demonstrate that OPTN initiates autophagosome formation via direct interaction between TBK1 and the PI3K complex, whereas NDP52 initiates mitophagy via direct interaction with the ULK1/2 complex ( Figure S5C), while the HORMA domain of ATG13 plays a linking role between the PI3K and ULK1/2 complexes ( Figure S5C). The linking role played by ATG13 explains how OPTN's unconventional mechanism of mitophagy initiation can proceed by first engaging with the PI3K complex through TBK1.

DISCUSSION
During PINK1/Parkin mitophagy, OPTN and NDP52 enable the recruitment of early autophagy machineries that initiate the de novo formation of autophagosomes on the surface of damaged mitochondria (Lazarou et al., 2015;Vargas et al., 2019). The mechanism of NDP52-mediated mitophagy initiation was recently revealed and involves NDP52 binding to the C-terminal part of FIP200 via an N-terminal SKICH domain (Vargas et al., 2019). The interaction between NDP52 and FIP200 enables recruitment and activation of the ULK1/2 complex through local clustering (Vargas et al., 2019). An analogous mechanism was identified during xenophagy for NDP52 (Ravenhill et al., 2019), and for TAX1BP1 and p62 during aggrephagy (Turco et al., 2021;Turco et al., 2019). The discovery that the LIR region of OPTN can also bind to the Claw domain of FIP200 in vitro indicated that OPTN also initiates selective autophagy in the same way. However, OPTN has also been reported to bind ATG9A in cells (Yamano et al., 2020), and to date it has remained unclear exactly how OPTN initiates selective autophagy. The importance of understanding OPTN's initiation mechanism is highlighted by the fact that it is mutated in amyotrophic lateral sclerosis (ALS) (Maruyama et al., 2010), is involved PINK1/Parkin mitophagy linked to Parkinson's disease (Kitada et al., 1998;Valente et al., 2004;Youle, 2019), and is abundantly expressed in brain tissue where there is little to no NDP52 (Lazarou et al., 2015).
By delineating the initiation mechanism of OPTN in isolation during PINK1/Parkin mitophagy, we have revealed an unconventional pathway of selective autophagy initiation that is independent of FIP200 binding as the most upstream event. We propose a model in which OPTN's binding partner TBK1 (Li et al., 2016;Richter et al., 2016), recruits/activates the PI3K complex via direct binding, which is linked to ULK1/2 complex via the HORMA domain within ATG13 (Figures 6E and S5C), thereby bringing the ULK1/2 complex to the surface of mitochondria. Upon recruitment of the ULK1/2 complex, OPTN's interaction with FIP200 (Zhou et al., 2021) may become stabilized, and in combination with the recruitment of ATG9A vesicles via direct interaction with OPTN (Yamano et al., 2020), the formation of phagophore membranes is triggered ( Figure 6F).
The presence of ATG9A vesicles at this early stage likely provides lipids for generating the phagophore (Judith et al., 2019;Karanasios et al., 2013;Mari et al., 2010;Sawa-Makarska et al., 2020;Yamamoto et al., 2012). Whether the complex interplay of OPTN mediated interactions is mutually exclusive to NDP52 mitophagy initiation remains to be determined. While our data supports that engagement of the PI3K complex occurs first via a direct interaction with TBK1, the recruitment of the PI3K and ULK1/2 complex to autophagosome formation site is likely to be more complex and dynamic. Indeed, previous work has identified oscillatory recruitment of ATG13 and OPTN, as well as positive feedback loops between PI3K and ULK1/2 complex recruitment (Karanasios et al., 2013;Zachari et al., 2019). Whether TBK1 mediated phosphorylation of the PI3K and ULK1/2 complexes regulates the dynamics of initiation ( Figure   S5D), including OPTN binding to FIP200, warrants further investigation.
Given the essential role played by ULK1/2 kinases in starvation autophagy initiation (Chan et al., 2007;Hosokawa et al., 2009;Jung et al., 2009), it was reasonable to conclude that they should be essential for PINK1/Parkin mitophagy. Instead, we discovered that ULK1/2 are dispensable for mitophagy initiated by either OPTN or NDP52. More specifically, ULK1/2 are completely dispensable for OPTN mediated mitophagy, but are functionally redundant with TBK1 during NDP52 mediated mitophagy (Figure 4), although mitophagy through TBK1 appears to be slightly faster ( Figure 1). A PINK1/Parkin independent form of mitophagy induced by ivermectin treatment also does not require ULK1/2 and is instead driven by TBK1 (Zachari et al., 2019), while other autophagy pathways independent of ULK1/2 kinase activity have also been identified (Alers et al., 2011;Hieke et al., 2015). It is possible that TBK1 compensates for the loss of ULK1/2 in the ULK1/2 independent pathways. We propose that TBK1 can function as a selective autophagy initiating kinase, akin to ULK1/2, but TBK1's role in selective autophagy will depend on which autophagy adaptor is involved. For example, TBK1 can play a dominant role during OPTN mediated selective autophagy through stable direct binding (Li et al., 2016;Richter et al., 2016).
While the ULK1/2 kinases were not essential for PINK1/Parkin mitophagy (Figure 1), the ATG13 and FIP200 subunits were found to play an important role (Figure 2), highlighting the importance of key subunits of the ULK1/2 complex. During NDP52 mediated mitophagy, FIP200 was found to promote the recruitment of NDP52 both in cells and in vitro (Figure 2) in an apparent positive feedback loop. The discovery that membrane binding drives allosteric activation of ULK1/2 complex recruitment by FIP200 and NDP52 likely also contributes to the positive feedback loop in cells (Shi et al., 2020a).
In yeast, the HORMA domain of Atg13 was shown to be important for Atg14 recruitment to autophagosome formation sites during autophagy (Jao et al., 2013). In mammalian systems, the PI3K complex has also been reported interact with the ULK1 complex and this interaction is mediated by direct binding between ATG13 and ATG14 via the HORMA domain in ATG13 (Park et al., 2016). OPTN and NDP52 mediated mitophagy is consistent with the HORMA domain of ATG13 playing a role in connecting the PI3K and ULK1/2 complexes ( Figure 6) which is crucial for ULK1/2 complex recruitment during OPTN mediated mitophagy. It would be interesting to investigate the role of the HORMA domain within ATG101, an ULK1/2 complex subunit that has been shown to dimerize with ATG13 (Qi et al., 2015;Suzuki et al., 2015). ATG101 could conceivably allow the simultaneous binding of the ULK1/2 complex to ATG9A and the PI3K complex to enhance autophagosome formation (Ren et al., 2022).
Overall, our discovery that OPTN and NDP52 utilise distinct mechanisms to drive mitophagy initiation has broad implications for selective autophagy. It indicates that within the same type of selective autophagy, multiple mechanisms of initiation can be undertaken depending on which autophagy adaptors are present, and that can vary across different cell types and tissues. This finding can become important when it comes to targeting a particular selective autophagy pathway for therapeutic purposes. For example, strategies to enhance PINK1/Parkin mitophagy in neurons that express high levels of OPTN will differ from that in tissues that express little to no OPTN, such as heart and small intestine (Lazarou et al., 2015). It would also be important to dissect whether the mechanism of mitophagy initiation driven by OPTN is more suited to the cell biology of a neuron. Digging deeper into the precise mechanisms of selective autophagy initiation mediated by each autophagy adaptor during different forms of selective autophagy is likely to yield therapeutic strategies to target damaged mitochondria, invading pathogens, and aggregating proteins in human disease.  Figure  S1A). (F and G) penta KO and ULK1/2 DKO/penta KO cells expressing BFP-Parkin and GFP-OPTN or GFP-NDP52 were treated with OA for 1 h and immunostained for WIPI2 and B17.2L (F) and the number of mitochondrial WIPI2 puncta per cell was quantified (G) (Untreated samples shown in Figure S1B Figure S3A). Data in (D and F) are mean ± s.d. from three independent experiments. *P<0.05, **P<0.005, ***P<0.001, ****P<0.0001 (one-way ANOVA). ns: not significant. Scale bars: overviews, 10 µm; insets, 2 µm.   Figure S4A and S4C). Data in (C) and (E) are mean ± s.d. from three independent experiments. *P<0.05, **P<0.005, ***P<0.001, ****P<0.0001 (one-way analysis of variance (ANOVA)). ns: not significant. Scale bars: overviews, 10 µm; insets, 2 µm.

Antibodies and reagents
The monoclonal and polyclonal antibodies used in this study are listed in Table S1. The polyclonal B17.2L antibody (a kind gift from Prof. Mike Ryan (Monash University)) was previously generated in rabbits using recombinant full length B17.2L antigen (Lazarou et al., 2007). Penta KO cells were described previously (Lazarou et al., 2015). identified, gene editing was detected using Sanger sequencing. The CRISPR-targeted regions were amplified via PCR and PCR products were sequenced using sequencing primers that anneal to the amplified regions (Table S3). For analysis purpose, the same regions in the parental line were also sequenced. Synthego ICE v2 CRISPR Analysis Tool (synthego.com/products/bioinformatics/crispr-analysis) was used to compare the sequencing data from the control and the knockout cells, allowing the identification of gene edits in the knockout cells (Table S2). For ULK1 KO cell lines (in ULK1/2 DKO/penta KO and TBK1/ULK1/2 TKO/penta; Table S2), sequencing the amplified PCR products appeared to be difficult, we therefore cloned them into a pGEM4Z vector prior to sequencing analysis (see Table S3 for details of genotyping primers). Although the sequencing reads were not of high quality in some parts, there was a clear edit of 10 bp deletion at the CRISPR target region which resulted in frameshift and premature stop codon (Table S2). Nevertheless, there was no protein detected by immunoblotting ( Figure 1A). Where antibodies were not available, putative KO clones were first identified by three primer PCR (Yu et al., 2014)
Stable cell lines were generated using retroviral system as described previously (Nguyen et al., 2021). First, a retroviral vector (pCHAC-mito-Keima and all pBMN constructs) containing an ORF of interest and retroviral packaging plasmids (VSV-G and Gag-Pol) were transfected into HEK293T cells using Lipofectamine LTX (Life Technologies) for 15 h. The next day, the transfection media was replaced with fresh growing DMEM to allow the assembly and secretion of retroviruses. After 24 h incubation, supernatant containing retroviruses was collected, filtered and applied onto the desired HeLa cells for 24-48 h in the presence 8 mg/ml polybrene (Sigma).
Cells were then recovered and expanded in growing DMEM for 5-7 days prior to sorting by FACS to match protein expression levels where possible.

Mitophagy treatment and Immunoblotting
Translocation and mitophagy experiments were achieved by incubating cells with 10 µM Oligomycin (Calbiochem), 4 µM Antimycin A (Sigma) and 10 µM QVD (ApexBio) in full growth medium for different time points as indicated in figure legends. For the analysis of BX795 treatment, the indicated samples were pre-treated for 30 min with 1 µM BX795 in growth media before inducing mitophagy in the presence of 1µM BX795.
For Western blot analysis, cells from 6 well plates were washed with 1x PBS, harvested using cell scrapers and lysed in 1× LDS sample buffer (Life Technologies) supplemented with 100 mM dithiothreitol (DTT; Sigma). Samples were heated at 99 °C with shaking for 7-10 min.

Immunofluorescence assay and confocal image analysis
For immunofluorescence, 48 h before treatment cells were seeded on HistoGrip (ThermoFisher) coated glass coverslips for. All following steps were performed at room temperature. Samples Automated 3D image segmentation approach with 3D ROI manager (v3.93) and FeatureJ (v3.0.0) plugins for FIJI (v1.52p) was employed to process and analyse all 3D image data. As described previously (Padman et al., 2019), the image analysis process includes three steps: object detection, object measurement and analysis. To account for the lack of foci in the untreated conditions,

Mito-Keima autophagosome-lysosome fusion mitophagy assay
Cells were seeded into 24 well plates one day before the experimental day. Following mitophagy induction at indicated time points, cells were washed with 1x PBS, detached with trypsin (Life Technologies) and harvested in normal growth media. Samples were then centrifuged at 4 °C and resuspended in ice cold sorting buffer containing 10 % v/v FBS and 0.5 mM EDTA in 1x PBS).
Samples were analysed using the FAVSDiva software on a LSR Fortessa X-20 cell sorter (BD Biosciences

GFP-labelled ULK1 complex was expressed in
The pGST2-GST-TEV-OPTN(S177D S473D) and pGST2-GST-TEV-NDP52 constructs were generated by introducing TEV cleavage site to corresponding vectors (Chang et al., 2021)  Genes coding for protein sequences of human ATG14, BECN1, VPS34, VPS15 were codon optimized for the Sf9 insect cell expression system, and synthetic genes were purchased from GenScript. All four ORFs were insertent to a GoldenBac plasmid (pGB) via Golden Gate approach by the Vienna BioCenter Core Facilities (VBCF) Protech Facility. To express mCherry labelled PI3KC3-c1 1 L culture of Sf9 cells growing in Sf921 medium at 1-1.5 mil/ml cells/volume were infected with 1 ml of Virus 1 (V1). Baculovirus was obtained by transfection of Sf9 cells with a policystronic construct coding for the mCherry labelled PI3KC3-C1 complex (Table S4). After infection cells were monitored and harvested by centrifugation when they reached a viability of 95-98%. Cell pellets were washed with PBS, flash frozen in liquid nitrogen, and stored at -80 °C until purification. To purify mCherry labelled PI3KC3-c1 the cell pellet corresponding to 1L culture was thawed and resuspended in 50 ml lysis buffer (50 mM HEPES pH 7.5, 300 mM NaCl, 0.5 % CHAPS, Benzonase, 1 mM MgCl2, 1 mM DTT, CIP protease inhibitor (Sigma), cOmplete EDTA-free protease inhibitor cocktail (Roche). Cells were additionally disrupted with Dounce homogenizer. Lysates were cleared by centrifugation at 25 000 rpm for 45 min at 4°C in a Ti45 rotor (Beckman). Supernatant was incubated with 5 ml of Glutathione Sepharose 4B beads slurry (Cytiva) for 1h at 4°C. Then beads were washed twice with 50 mM HEPES pH 7.5, 30 0mM NaCl, 0.5% CHAPS, 1 mM DTT, twice with 50 mM HEPES pH 7.5, 700 mM NaCl, 1 mM DTT and finally twice with 50 mM HEPES pH 7.5, 300 mM NaCl, 1 mM DTT. To elute the protein complex the beads were incubated overnight with Precission (3C) protease at 4°C. The supernatant containing cleaved protein was filtered through a 0.2 μm syringe filter, concentrated down to 0.5 ml and applied onto a Superose 6 column (10/300, Cytiva) pre-equillibrated with a buffer containing 50 mM HEPES pH 7.5, 200 mM NaCl, 1 mM DTT. Fractions containing pure proteins were pooled, concentrated, snap frozen in liquid nitrogen, and stored at −80°C. The protocol is available at dx.doi.org/10.17504/protocols.io.8epv59mz4g1b/v1.
For the experiments shown in Figures 6A and 6B For quantification of mCh-NDP52 recruitment to the beads in the presence and absence of FIP200-GFP ( Figure 2I) using ImageJ software eight lines were drawn across each bead and the maximum brightness value along each the line was taken. Next, the average brightness of an empty area of each picture was measured (background fluorescence) and subtracted from the maximal fluorescence for each bead. The average values for each sample were averaged between 3 independent replicates and plotted with the relative standard errors. For the quantification described above, statistical analysis was performed. Statistical significance of the difference between 2 samples was established by 2 samples unpaired t test. Significant differences are indicated with * when p value ≤ 0.05, ** when p value ≤ 0.01, *** when p value ≤ 0.001. The Protocol for this is available online (dx.doi.org/10.17504/protocols.io.6qpvr6nwpvmk/v1)

Acknowledgments:
We would like to thank the laboratory of Noboru Mizushima (The University of Tokyo) for sharing

Declaration of interests: Sascha Martens is member of the scientific advisory board of Casma
Therapeutics. Michael Lazarou is a member of the scientific advisory board of Automera and a consultant for Casma Therapeutics.
Data availability: All data are provided within the manuscript and supplement.