Influence of ULK1 activity on the memory of ATG101 auto-activation dynamics

macro autophagy, autophagy initiation , protein metamorphosis, ULK1, ATG13, 40 ATG101, ATG9A, auto-activation

Macro-autophagy (called autophagy hereafter) is a conserved process of regulated degradation. It eliminates damaged and unnecessary cellular compo nents by transporting biomolecules to 45 lysosomes, where they are degraded to be recycled. The ULK1 complex responds to upstream nutrient and energy signals and is essential to initiate autophagy by bridging and building the autophagy initiation machinery1. When fully assembled, the ULK complex is composed of four proteins: the ULK1 or -2 kinase, ATG13, ATG101, and focal adhesion kinase family interacting protein of 200 kDa (FIP200) 2-5. ULK kinase activity is a key regulatory event to initiate and 50 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 3 coordinate the autophagy pathway6. For example, ULK1 phosphorylates VPS15 to activate the PI3-kinase complex7, but also its own components like ATG101 with unknown consequences6. The other subunits of the ULK complex are thought to have numerous scaffolding and bridging roles, some of which act independently and upstream of the ULK kinase and its activity. The 55 recruitment of ATG13 and ATG101 to sites of autophagy initiation is an essential and defining early step in both starvation -induced autophagy and selective autophagy3,8,9. The C-terminal 300 residues of ATG13 are predicted to be intrinsically disordered and interacts with FIP200 and the ULK kinase10-12. The function of ATG101 is unclear, but it has been reported to increase the lifetime of its binding partner ATG13 by protecting it from proteasomal degradation and to 60 stabilize basal phosphorylation of ULK1 and ATG13 3,5. Hetero-dimerization of ATG13 and ATG101 via their HORMA domains (named after the Hop1p, Rev7p and MAD2 proteins ) is essential for autophagy, which might in part be due to the reciprocal stabilizing effect 13-15. Notably, both ATG13 and ATG101 are present in higher abundance than the rest of the ULK kinase complex and might have additional roles outside the ULK kinase complex 15. For 65 example, an ATG13-ATG101 dimer lacking an ULK kinase assembles with the early autophagy machinery, such as ATG9A, to promote phagophore growth in fed cells 8,9,16,17. The ATG13- ATG101 dimer interacts with the “HORMA dimer–interacting region” (HDIR) in ATG9A and recruits ATG9A during p62-dependent autophagy, independent of FIP200 and ULK1 13,16,18. Deletion of ATG13 or ATG101 abrogates the ir reciprocal interactions, and leads to a mis -70 localization of ATG9A and phenotypes similar to an ATG9A knock -out16,19. The ATG13- ATG101 dimer plays a critical coordinating role in recruitment of downstream factors to the autophagosome formation site by nucleating the assembly of stable multi -megadalton complexes3,13,15,16,20-23. However, it is currently unclear how the assembly of these complexes is regulated in space and time. 75 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 4 Despite its importance, there are no known ATG101 orthologs in S. cerevisiae, but it is present in lower metazoans. Human ATG101 (218 amino acid s), does not have obvious sequence similarity to any other protein in species that lack this gene . The X-ray structures of human ATG101 show that its fold is promiscuous. ATG101 in isolation a dopts a similar fold as the 80 ‘open’ conformer of the prototypical metamorphic HORMA domain protein MAD2 as defined by the position of the C -terminus into a beta-sheet24 (Figure 1A). The fold of ATG101 does however change upon dimerization with ATG13, which includes the ejection of the C-terminus and its conversion from a beta -strand into an alpha -helix22,24 (Figure 1A). This topological conversion allows the interaction of th e newly formed C-terminal helix with the PI3 -kinase 85 complex22. Together, the ATG13-ATG101 hetero-dimer is analogous to the ‘open’-‘closed’- conformational homo-dimer of MAD225. The topologically asymmetric MAD2 homo-dimer is an obligatory reaction intermediate in the mechanism to accelerate the assembly of the Mitotic Checkpoint Complex, by lowering the activation barrier of the otherwise very slow metamorphosis of ‘open’ MAD2 to the ‘closed’ conformer 26,27. Therefore, the default ‘open’ 90 conformer of MAD2 represents an autoinhibited state and the dimerization accelerates the rate- limiting conversion into the active ‘closed’ conformer in order to induce the formation of its downstream effector complex. The emerging p aradigm is that regulated acceleration of metamorphosis of HORMA domain proteins controls the rate of signalling or assembly of effector complexes27,28. This paradigm might be conserved beyond HORMA domain proteins, 95 as dimerization has been found to modulate interaction kinetics of unrelated metamorphic proteins25,29-32. Using qualitative interaction assays, we had previously shown that the interaction kinetics of ATG13 and ATG101 to ATG9A are remarkably slow, similar to related HORMA domain proteins MAD2 and REV7 13,26,27,33,34. This suggested that both ATG13 and ATG101 too default to an autoinhibited state. Indeed, ATG13 adopts two distinct folds that 100 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 5 could be identified and separated based on their differential surface charge and mutually exclusive interactions 13. Mutants designed to interfere with the metamorphic beh aviour in ATG13 and ATG101 both show strong autophagic defects and do not allow for the assembly of the initiation machinery13. 105 The role of ATG101 is enigmatic, in particular the contribution of its topological conversion to its function and interaction kinetics. In this study , we monitored the interaction kinetics of ATG101 in real time at physiologically relevant concentrations using an in vitro Fluorescence Polarisation (FP) sensor. We observed that the interactions of ATG101 with either ATG13 and ATG9A have remarkably small on-rate constants, which are however dramatically accelerated 110 at increased concentrations of ATG101. We show this is due to transient homo-dimerization of ATG101, which is stabilized after phosphorylation by ULK1 , leading to the activation of ATG101 and rapid assembly of the ATG9A-ATG13-ATG101 complex. This mechanism results in a memory of activation, as A TG101 ‘remembers’ its activation for many hours after dephosphorylation. This suggests that the initial phosphorylation induces a structural 115 conversion into an active state required for dimerization. Indeed, shifts in CD spectra indicate structural changes correlating with the topological conversion of the beta strand into an alpha helix. Although it paradoxically uses the same interaction interface to engage with ATG13, the transient nature of the homo -dimerization promotes, rather than inhibits, the interaction of ATG101 to ATG13 and subsequent interaction with ATG9A. Moreover, only a small amount 120 of activated ATG101 suffices to auto-catalytically activate all ATG101 molecules to accelerate the formation of the key ATG9A-ATG13-ATG101 complex. Overall, this suggests an unusual regulatory mechanism, where the ATG101 dimer serves to present a template for the conversion of subsequent ATG101 molecules to create a responsive positive feedback and initiate the local fast assembly of the autophagy machinery on demand. 125 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 6

The interaction of ATG101 with ATG13 and ATG9A is exceedingly slow. Structures of ATG101 show a striking difference in the topology of its C-terminal 25 residues. Alone, this region folds as a beta -strand incorporated into an extended beta sheet 24. However 130 when bound to ATG13, the terminal beta-strand is dislodged from the extended sheet to form an alpha-helix22,24 (Figure 1A). This C-terminal mobile structural element corresponds with the topologically mobile ‘seat-belt’ in MAD2, which is elementary in defining its metamorphic state and thus interaction spectrum25,35. Structure prediction software has a mixed track record in predicting multiple metamorphic conformers from single polypeptides 36-38. For example, 135 AlphaFold339 is somewhat uncertain about the C -terminal ATG101 region, but changes its prediction from the beta -strand to the helical conformer upon introducing the ATG9A HDIR peptide or ATG13 HORMA domain (Figure S1A ). We note that the C-terminal region is missing in the ATG13-ATG101-ATG9AHDIR structure and therefore does not guide the prediction18. This structural conversion is allosteric ally induced, as this region is not part of , 140 nor close to, the ATG13-ATG101 dimer interface. Since the C-terminal region is not part of the interaction interfaces and seems to prefer to fold as a beta strand by default, it is therefore unclear what comes first: the hetero-dimerization with ATG13 or the conversion to the alpha-helix? In case of the latter, the prediction would be that 145 the ATG13-ATG101 heterodimer would form slowly due t o significant unfolding of ATG101 required, and thereby create a rate-limiting step in complex formation (Figure 1A). To test this, we developed an in vitro fluorescence polarization (FP) assay by fusing an Alexa488 moiety to the C-terminus of ATG101 using Sortase labeling40 (Figure S1B). This allowed us to determine binding isotherms and quantify apparent binding strengths ( KD,app) of ATG101 to ATG13, 150 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 7 ATG101 to ATG9A and ATG13-ATG101 to ATG9A, respectively (Figure S1C-E). This sensor also allowed us to monitor ATG13-ATG101 complex formation in real time (Figure 1B). The time-dependence increase of the FP signal could be fitted with a single exponential function to yield an apparent reaction rate ( kobs) that increased linearly with the concentration of ATG13, indicative of pseudo -first-order kinetics (Figure 1C ). Even a t a bove physiological 155 concentrations of ATG13 (500 nM, using 20 nM ATG101), complex assembly is very slow, with a half-life (t1/2) of approximately 13 hours. The corresponding association rate of 1.47 x 10-5 µM-1s-1 is several orders of magnitude slower than typical protein-protein interactions, and an order of magnitude slower than other known metamorphic HORMA domain proteins MAD227 and REF733, highlighting the extraordinary slow nature of the assembly of the ATG13-160 ATG101 complex (Figure 1C,F). The association of ATG101 with ATG9A was equally slow, indicating that likely an inhibitory, and thus rate limiting, event in ATG101 prevents a canonical interaction to both ATG13 and ATG9A (Figure S1F,G). Importantly however, when the ATG13-ATG101 complex was allowed to pre-form during an 165 overnight pre-incubation, the interaction with ATG9A was complete within 1 minute at near - physiological concentrations (125 nM) (Figure S1J ). The pre -incubation accelerated the interaction to ATG9A by 3-orders of magnitude to 1.5 x 10 -2 µM-1s-1 (Figure S1K). Overall, this suggests that the ATG13-ATG101 dimerization is the limiting step for association with ATG9A and that the interaction with ATG13 lifts the inhibitory effect in ATG101 that prevents 170 prompt complex formation. Concentration controls the unusual activation of ATG101 interaction dynamics. The rate -limiting nature of the ATG101 interaction kinetics prompted us to invert the experiments, and instead vary the concentration of ATG101. We used the FP assay to observe 175 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 8 the interaction kinetics with ATG13 or ATG9A at varying ATG101 concentrations. This yielded two surprising, and at first glance contradictory, observations. Most prominently, we observed that ATG13-ATG101 complex formation was slow until it suddenly accelerated at concentrations above ~5 µM ATG101, taking place within a few minutes (Figure 1D). This was corroborated by quantifying the on -rates, where the typical linear increase of rates is 180 followed by sudden atypical acceleration at elevated ATG101 concentrations (Figure 1E). This switch-like 100-fold increase of on -rate could therefore not be explained by the increase of ATG101 concentration alone. Conversely, we observed that the interaction of ATG101 with ATG9A, in the absence of ATG13, slows down with increasing ATG101 concentration (Figure S1L,M). These results suggest that the association of ATG101 with its binding partners follows 185 an unusual mechanism, which has an obligatory intermediate that is sensitive to the concentration of ATG101 and directly controls both interaction dynamics and strength . The inhibited interaction of ATG101 to ATG9A shows the order of interaction: the formation of the ATG13-ATG101 dimer has to precede the interaction of ATG101 with ATG9A. 190 Figure 1: The interaction of ATG101 with ATG13 is exceedingly slow, but suddenly accelerates at higher ATG101 concentration. a) The structures of ATG101 suggest an activation might be required to interact with .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 9 ATG13 and ATG9A in order to initiate autophagy. b,c) The interaction of ATG101 with ATG13 is exceedingly slow. Time zero is the first time point after mixing 20 nM ATG101Alexa488 with indicated ATG13 concentrations. 195 All panels in b reporting time-dependent changes in Fluorescence Anisotropy signal are single measurements representative of at least three independent technical replicates of the experiment. After single exponential fitting of the curves in b, the apparent first order rate constants (kobs) were plotted as function of ATG13 concentration in c, with kon being the slope of the resulting line. d,e) The slow association of ATG101 to ATG13 accelerates at least two orders of magnitude depending on ATG101 concentration. Fluorescence Anisotropy 200 measurements as performed in b and c, but instead mixing 50 µM unlabeled ATG13 with indicated ATG101 concentrations. ATG101 transiently homodimerizes. These observations indicate that there is an inhibitory element in ATG101 that prevents a 205 straightforward interaction to ATG13, and that this inhibition is overcome at increased concentrations of ATG101. Metamorphic proteins are known to modulate their interaction kinetics by inducing the transition from a (default) autoinhibited -conformer to an interaction competent-conformer via homo-dimerization25,29-32. We have not observed homodimerization of ATG101 during any size exclusion chromatography experiment (using wild -type, nor 210 modified protein (see below)) , although the elution peaks were typically not symmetric, suggesting heterogeneity (Figure S1B). Regardless, we performed a pull -down experiment using differentially tagged ATG101 fusions at 2 or 6 µM (bait and prey, respectively) concentration, and using quick washes with reduced volumes to keep the concentration relatively elevated. This approach revealed that ATG101 can indeed stoichiometrically homo-215 dimerize at concentrations that accelerate ATG13-ATG101 complex formation (Figure 1E and 2A). This dimer uses the canonical HORMA dimerization interface, as the introduction of the L29R and H30R mutations41 abrogated both the ATG101 homo-dimerization as well as the hetero-dimerization with ATG13 (Figure 2 B,C). As we observed a decreased binding to ATG9A at elevated ATG101 concentrations, we wondered if the ATG101 homo-dimer would 220 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 10 be incompatible with ATG9C binding. Indeed, when fusing ATG101 with a GST-tag to promote homodimerization, the interaction with ATG9A was strongly reduced compared to an MBP- tagged ATG101 (Figure 2D). Since ATG101 homo-dimerization, in contrast to ATG13-ATG101 hetero-dimerization, seems 225 to be weak and/or transient, we aimed to stabilize it using the crosslinker BS3. In the presence of the crosslinker, we observed the emergence of a single additional protein species with a mass corresponding to that of a dimer when using the wild -type protein, but not when using the ATG101L29R,H30R dimerization-deficient mutant (Figure 2E and S2A). Using mass photometry, we confirmed that this treatment stabilized the homo -dimer (Figure 2F). We determined the 230 sites of cross -linking using mass -spectrometry (Figure 2G and Table S1) and used th ese to create a model of the dimer using HADDOCK42. The cross-links agreed well with a model that ATG101 homo-dimerization uses the canonical HORMA dimer interface (Figure 2H (left) and S2C). The model thus highlighted that homo-dimerization is competitive with the interaction with ATG13 (Figure 2H, right). 235 The cross-links cannot distinguish if the dimerization is symmetric (meaning with specifically the same conformer, as seen with REV7 in PolZ 43), asymmetric (specific for opposing conformers, as seen for MAD2 25), or agnostic (there is no selectivity, as seen for REV7 in Shieldin33). Regardless, as dimerization typically stabilizes the change in fold in metamorphic 240 proteins, we wondered if the same is true in ATG101. Therefore, we used circular dichroism (CD) measurements of ATG101 and the ATG101L29R,H30R mutant, which showed significant changes in secondary structure upon dimerization (Figure 2I). We used SESCA44,45 which uses a Bayesian statistics approach to determine the likelihood of possible secondary structures of a target protein based on the measured CD spectra (Figure 2J). The proposed models show an 245 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 11 increase in helical content and concomitant decrease of relative beta-sheets in ATG101 upon dimerization. These changes agree with the observed conversion of the C-terminal beta-strand into an alpha-helix in the published ATG101 structures (Figure 1A). Overall, this shows that dimerization initiates a structural and topological conversion in ATG101. 250 Figure 2: ATG101 transiently homo-dimerizes using the canonical HORMA dimer interface. a) ATG101 can homo-dimerize. Pull-down experiment using 2 μM MBP-ATG101 as bait and 6 μM Strep-ATG101. b,c) ATG101 interaction with ATG13 (panel b) and homo-dimerization (panel c) require the same interface. Pull- down experiments using 2 μM MBP-ATG13 (panel b) or 2 μM MBP-ATG101 (panel c) as bait and 6 μM His-255 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 12 ATG101. Introduction of the L29R and H30R (reference 41) mutations abrogated both the homo- as well as the hetero-dimerization with ATG13. d) Inducing homo-dimerization via a dimeric GST-fusion prevents the interaction of ATG101 to ATG9A. Pull-down experiment using 1 μM ATG9A with biotinylated amphipol as bait and 3 μM GST- or MBP-ATG101. Quantification in Fig S2B of three independent technical replicates of the experiment. e) Incubation of ATG101 with cross-linker BS3 results in the appearance of a distinct second protein 260 with a mass corresponding to an ATG101 dimer. Input is shown in lanes with no BS3 (0 mM). f) ATG101 homo- dimerization is stabilized by crosslinking. Mass photometry experiment using ATG101 (150 nM), comparing untreated and after BS3 (1 mM) treatment and subsequent purification. g,h) Crosslinking with BS3 coupled with mass spectrometry confirmed that ATG101 homo-dimerizes via the canonical HORMA dimer interface. The cross-links determined in g were used to predict three structural models using HADDOCK42 in h and S2C of the 265 ATG101 homo-dimer. The resulting models show that ATG101 uses the same interface for both homo- (h, left) and hetero-dimerization (h, right). Cross-linked residues in the dimer interface are shown as black in g and yellow beads on string in h). i,j) Dimerization of ATG101 induces structural changes. CD spectra (panel i) recorded using 3.5 µM ATG101 or the ATG101L29R,H30R dimerization defective mutant, where analysed and modelled using SESCA44,45 (panel j), where the proposed models show an increase in helical content and 270 concomitant decrease of beta-sheets in ATG101 upon dimerization. ULK1 activity induces ATG101 structural conversion and homodimerization. Homo-dimers of metamorphic proteins usually contain at least one converted or intermediate ‘unbuckled’ conformer25,29-32, therefore we wondered if structural changes in ATG101 are 275 required to induce homo -dimerization. Residues Ser11 and Ser203 of ATG101 are known phosphorylation sites of ULK1 with unknown function6. Ser203 is part of the C -terminal topologically flexible region, and could hypothetically directly destabilize the beta-strand and thereby promote the formation of the converted conformer ( Figure 1A). Indeed, AlphaFold3 changes its prediction to the helical conformer upon introducing the phosphorylation at Ser203 280 (Figure S1A). To test if phosphorylation affects ATG101 structure and its ability to homo - dimerize, we added the generic phosphatase inhibitor Okadaic Acid during the last 5 hours (of the insect cell culture expressing ATG101 , resulting in the purification of phosphorylated .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 13 protein (Figure S3A). Using mass-photometry we confirmed that this phosphorylation results in ATG101 dimerization in contrast to unphosphorylated ATG101 at 150 nM (Figure 2F and 285 S3B). Next, we co-expressed ATG101 and ULK1 by mixing their respective baculo-viruses (in a 5:1 ratio) to the expression culture, which also resulted in the phosphorylation and dimerization of purified ATG101 (Figure 3A and S3C). This is a direct result of ULK1 activity, as the phosphorylation of ATG101 using purified full-length recombinant ULK1 kinase yielded an ATG101 homo-dimer with an apparent Kd between 100 and 250 nM (Figure 3B,C). Using 290 mass-spectrometry, we confirmed that ATG101 is indeed phosphorylated by ULK1 at Ser11 and Ser203 (Table S2). A similar stabilization of the ATG101 homo-dimer was observed using the Ser11 and Ser203 phospho-mimetic mutants (Figure S3D). 295 Figure 3: ULK1 activity induces ATG101 homodimerization. a) Co-expression of ULK1 kinase and ATG101 resulted in the stabilization of the ATG101 homo-dimer. Mass-photometry experiment showing size distribution of 250 nM purified ATG101 co-expressed with or without ULK1. b) ATG101 can be phosphorylated by purified ULK1. ATG101 was incubated with ULK1 (1:100 ratio) and Mg2+/ATP for 30 min at 27 °C. Phosphorylation was subsequently removed by the addition of lambda-phosphatase for 30 min at 27 °C. c) Phosphorylation 300 stabilizes ATG101 homodimerization. Mass-photometry experiment showing size distribution at various concentrations of purified ATG101 in vitro phosphorylated by ULK1. Influence of ULK1 activity on ATG101 auto-activation dynamics. The results so -far have created a paradox: ATG101 homo-dimerization - despite being 305 competitive with the ATG13-ATG101 complex, as well as weaker and transient – seems to .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 14 boost heterodimer formation. We therefore wondered what the relevance and function of the ATG101 homo-dimerization could be. Given that the phosphorylation of ATG101 by ULK1 induces a structural conversion in ATG101, we wondered if it could influence the interaction kinetics of ATG101. When performing a pulldown experiment that monitors ATG13-ATG101 310 heterodimerization, we observed that phosphorylated ATG101 would interact stoichiometrically with ATG13 within a few minutes while untreated ATG101 did not (Figure 4A,B and S4A). Similar results were observed with the phospho -mimetic ATG101 mutants (Figure 4C and S4B). Next, we quantified the interaction kinetics of the ATG13 -ATG101 complex using the FP assay. Using low concentrations of phosphorylated ATG101 ( 20 nM), 315 the interaction with ATG13 was unchanged and slow (1.47 x 10 -5 µM-1s-1) (Figure S4C,D). However, upon incrementally increasing the concentration of phosphorylated ATG101, w e again observed a similar non -linear stark increase in ATG13-ATG101 complex assembly (Figure 4D,E ). This observed effect was a similar acceleration as seen before using unphosphorylated ATG101 at elevated concentrations, but now occurred at much lower 320 concentrations. Similarly, the ATG9A-ATG13-ATG101 complex forms within a few minutes when using the phosphorylated ATG101, while the untreated ATG101 does not allow prompt complex formation (Figure 4F,G and S4E). S imilar results were obtained when using the phospho-mimetic ATG101 mutants (Figure 4H and S4F). 325 Interestingly, despite the typical incomplete phosphorylation by ULK1, we invariably observed complete and rapid binding of both phosphorylated and unphosphorylated ATG101 to ATG13 and ATG9A (Figure 4B,G). Upon mixing 400 nM stoichiometric total amounts of ATG101 and ATG13, the presence of 10% of phosphorylated ATG101 sufficed to accelerate the assembly of the complex (Figure 4I and S4G). Overall, this shows that phosphorylation activates ATG101, 330 and although phosphorylated ATG101 can interact with ATG13-ATG9, but is not a requirement .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 15 for the interaction. After being primed by phosphorylation, ATG101 will auto -catalytically activate other ATG101 molecules to promote quick association of ATG101 regardless of phosphorylation status. 335 Figure 4: Activated ATG101 auto-catalytically promotes ATG9A-ATG13-ATG101 complex formation. a,b) Phosphorylation activates ATG101 to promote a quick interaction with ATG13. Panels a (SDS-PAGE) and b (Phos-tagTM gel) show pull-down experiments after 10 min incubation of 0.2 μM GST-ATG101 as bait and 0.4 μM His-ATG101. Quantified in S4A from three independent experiments. c) Introducing phospho-mimetic 340 mutants S11D and S203D in ATG101 facilitate prompt dimerization of ATG13-ATG101. Pull down experiment after 10 min incubation of 0.2 µM of ATG13 as bait and 0.4 µM ATG101 as prey. Quantification in S4B of three independent technical replicates. d,e) Phosphorylation activates ATG101 to promote a quick interaction with ATG13. Time zero is the first time point after mixing 50 μM unlabelled ATG13 with indicated phosphorylated ATG101Alexa488 concentrations. Shown in d are single measurements representative of at least three independent 345 technical replicates of the experiment. After single exponential fitting of the curves, the apparent first order rate constants (kobs) were plotted as function of ATG13 concentration in e, with the results from panel 1e shown as .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 16 reference. f,g) Phosphorylation activates ATG101 to promote a quick assembly of the ATG9A-ATG13-ATG101 complex. Panels f (SDS-PAGE) and g (Phos-tagTM gel) show pull-down experiments after 10 min incubation of 0.2 μM MBP-ATG9A as bait and 0.4 μM His-ATG101 and GST-ATG13, and the result is quantified in S4E 350 from three independent experiments. h) Introducing phospho-mimetic mutants S11D and S203D in ATG101 facilitate quick assembly of the ATG9A-ATG13-ATG101 complex. Pull down experiment after 10 min incubation of 0.2 μM MBP-ATG9A as bait and 0.4 μM His-ATG101 and GST-ATG13 as prey. Quantification in S4F of three independent technical replicates. i) After being primed by phosphorylation, ATG101 will auto- catalytically activate other ATG101 molecules to promote quick association of ATG101 regardless of 355 phosphorylation status. Pulldown experiment monitoring ATG13-ATG101 complex formation using varying relative amounts of phosphorylated ATG101 separated on a 15 % SDS-PAGE. ATG101 and ATG13 were mixed at a final concentration of 400 nM each. Quantification in S4G of three independent technical replicates. Memory of activation of ATG101 is retained many hours after dephosphorylation. 360 Presumably the activation of ATG101 is reversible in order to achieve a responsive system. Moreover, since phosphorylation is not strictly required for homo-dimerization nor the interaction with ATG13 or ATG9A, we wondered what would happen if we removed the phosphorylation. Therefore, we prepared four ATG101 samples, where we either left ATG101 untreated (A), purified ATG101 after a partial ULK1 treatment (B), or subsequently removed 365 all phosphorylation using the generic lambda phosphatase and allowed ATG101 to revert to the inactivated state by waiting either 30 minutes (C) or overnight (D) (Figure 5A). We used these ATG101 samples to assess the kinetics of ATG9A -ATG13-ATG101 complex assembly by performing a pulldown experiment after 10 minutes after mixing them with separately purified ATG13 and ATG9A (Figure 5B-D). As before, no complex formed in this short time frame 370 with the untreated ATG101 (A), while an incompletely phosphorylated ATG101 (B) would promote complete complex formation. Surprisingly however, although all phosphorylation was removed (samples C and D) , ATG101 remained activated even after waiting overnight . Only the amount of ATG13 slowly reduced, yielding a stoichiometric 3:3:3 complex (Figure 5D). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 17 Indeed, the homo-dimerization of ATG101 is persistent long after it had been dephosphorylated 375 (Figure 5E). Overall, these experiments show that the phosphorylation is the trigger of the structural change that allows the dimerization and the increased interaction kinetics, is auto - catalytically propagated to non -phosphorylated ATG101 molecules. At this point, the phosphorylation becomes redundant for maintaining the structural change and homo - dimerization, yielding in a memory of activation as ATG101 ‘remembers’ its activation for 380 many hours after dephosphorylation. Figure 5: Memory of activation of ATG101 is retained many hours after dephosphorylation. a) Experimental design to test for the memory of activation after removal of phosphorylation of ATG101. b-d) 385 Memory of activation of ATG101 is retained after dephosphorylation. Panels b (SDS-PAGE) and c (Phos-tagTM gel) show pull-down experiments after 10 min incubation of 0.2 μM MBP-ATG9A as bait and 0.4 μM His- ATG101 and GST-ATG13. ATG101 treatment is described in panel a. Quantification from three independent experiments is shown in d. e) Dimerization of ATG101 is retained after dephosphorylation. Mass-photometry experiment showing size distribution of 200 nM purified ATG101 before and after phosphorylation by ULK1, 390 and subsequent dephosphorylation. f) Model of the influence of ULK1 activity on ATG101 auto-activation .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 18 dynamics, leading to quick assembly of the ATG9A-ATG13-ATG101 complex. Homologous to the MAD2- template model, ATG101 activation is likely auto-catalytically propagated by lowering activation energy of transition (see inset) allowing for ‘copying’ the state from the activated ‘template’ ATG101. This mechanism would support a responsive local initiation of autophagy. 395

Following the induction of autophagy, a complex array of proteins promptly converges to create its initiation machinery. The ATG13-ATG101 dimer coordinates the recruitment of a variety of downstream factors to the autophagosome formation site 3,13,15,16,20-23. Although some ATG13-400 ATG101 dimer can be found pre-assembled in fed cells, it only induces autophagy at limited levels15. It is likely that their interactions with its various binding partners, which might be mutually exclusive and have a multitude of functions, is regulated at multiple levels. This likely includes the regulation of the metamorphic behaviour of ATG13 and ATG101, as mutants in predicted metamorphic regions in ATG13 and ATG101 both do not allow for the assembly of 405 the initiation machinery and show strong autophagic defects13. To shed light on the regulatory mechanisms that dictate complex assembly , it is pertinent that the rate -limiting steps are identified as these are ultimately what determine when and where the initiation machinery assembles. In this manuscript , we have shown that the enigmatic core autophagy protein ATG101 interacts exceedingly slowly with its interaction partners. ATG101 can spontaneously 410 interact, but our quantifications show that it is 4 -orders of magnitude slower than typical protein-protein interactions and therefore requires over 24 hours to complete at physiologically relevant concentrations. This timing is not conducive to responsive autophagosome formation, which is estimated to take 5 to 10 minutes in mammals46,47. We show that this unusual behavior of ATG101 is likely the result of the large activation energy required for the structural 415 conversion of ATG101, where the most C-terminal beta-strand needs to be dislodged from an extended beta-sheet, and refold s as an alpha -helix. This structural metamorphosis, which is .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 19 stabilized by transient homo -dimerization, is obligatory to convert from its default autoinhibited conformer to an active autophagy promoting conformer (Figure 5F). We postulate that this rate -limiting event provides an opportunity to regulate when and where 420 autophagy is initiated. The kinase ULK1, a well-characterized interaction partner of ATG101 in the cell 2-5, is the presumable candidate to perform this regulatory task, as we show that it phosphorylates ATG101 resulting in homo-dimerization of ATG101. Crucially, this dimerization induces the structural conversion of ATG101 which is essential to accelerate ATG101 activation dynamics. Further structural analysis will be required to confirm the 425 (AlphaFold3) prediction that the position of the phosphorylation stabilizes ATG101 in the converted fold. Variations of this mechanism have been observed in related metamorphic HORMA domain proteins 48, where kinases initiate the activation of otherwise inhibited signaling nodes by accelerating their structural conversion 27,28. Once converted, these subsequently accelerate additional conversions via transient homo-dimerization25,29-32. This 430 cascading ‘templating’ mechanism26 would thus be auto -catalytic and thereby efficiently inducing effector complex formation in a spatio-temporally controlled manner. The observation that only a minor portion of ATG101 needs to be phosphorylated to induce all molecules to interact with ATG9A and ATG13, strongly suggests that ATG101 also allows a similar auto- catalytic regulatory ‘templating’ mechanism (Figure 5F). The ATG101 homo -dimerization 435 would lower the activation barrier by stabilizing a ‘reaction intermediate’ that catalyses the structural conversion and thus the interaction to ATG13 and ATG9A (Figure 5F, inset). We show that stabilization of homo-dimer is not directly dependent on phosphorylation, but rather activates the transition of ATG101 into a form that can homo-dimerize and interact with ATG13. The ATG101 homo-dimer cannot interact with ATG9A, securing the order of events: the 440 transient nature of the homo-dimer promotes binding to ATG13 first and together they assemble the ATG9A-ATG13-ATG101 complex . This complex assembles the re st of the initiation .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 20 machinery, where for example the newly ejected alpha-helix can serve as a receptor for further autophagy initiation factors, such as the PI3 -kinase complex 22. Regulating interactions via interfaces exclusively accessible after structural refolding would prevent premature 445 recruitment of initiation factors, a mechanism that finds precedent in other HORMA domain proteins49, including ATG1313. Upon silencing autophagy, the initiation machinery is likely disassembled. The conversion of ATG101 back to the default inhibited state is equally slow, which constitutes the long memory of activation even after dephosphorylation (Figure 5F, inset). We expect that currently unknown external factors will accelerate this conversion to 450 silence autophagy. Overall, t his extra-ordinary mechanism would create responsive feedforward mechanism to control autophagosome biogenesis in time and space.

We are grateful to Stefanie Aspe r for expert technical assistance. We thank the MPI of 455 Multidisciplinary Sciences for support , and in particular the live-cell imaging facility (MPI- NAT) for access to the plate reader and Mass Photometer. We thank the Stein Lab (MPI-NAT) for sharing the SortaseA plasmid. We also acknowledge the help of Monica Raabe, Ralf Pflanz and Sabine König from the Proteomics Facility (MPINAT) for assistance with sample handling. We are grateful to the Griesinger laboratory (MPI-NAT) for access to the CD spectrometer. We 460 thank the current members of the Faesen Lab for critical comments, discussions and helpful suggestions. Funding for this work was provided by the Max-Planck Society (to HG and HU) via the Max -Planck Research Group Leader program (to AF) and by the German Research Foundation (Deutsche Forschungsgemeinschaft; DFG) via SFB1190 (project P19 to AF) and FA 1752/3-1 (project number 542770124 to AF). 465 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 21 Author contributions: AP, SS and ACF designed in vitro experiments and analyzed results. AP and SS set up recombinant expression systems, established purification protocols and purified proteins. AP and SS performed in vitro experiments. GN helped with the analysis of CD data and assisted with SESCA calculations and analysis. OD performed mass-spectrometry 470 and analysis, supervised by HU. ACF supervised project and wrote the manuscript. Competing interests: Authors declare that they have no competing interests. Data and materials availability: All data are available in the main text or the supplementary 475 materials.

Expression and purification of proteins Table 1. Buffers used for protein purification 480 Buffer Composition A 150 mM NaCl, 20 mM HEPES pH 7.5, 5 %. Glycerol, 0.5 mM TCEP B 150 mM NaCl, 20 mM HEPES pH 7.5, 5 %. Glycerol, 0.5 mM TCEP, 0.03 % DDM ATG101 (insect cells) ATG101 was cloned as an N-terminal GST, 6xHis-MBP, or Strep fusion construct from Hi5 cells using the biGBac expression system. 5 ml of V2 baculovirus was infected in 400-500 ml Hi5® cell cultures at 1x10E6/ml density. Cells were harvested at 72h of infection. Cell pellets 485 were washed and resuspended in buffer A supplemented with 1 mM PMSF and 10 µg/50 mL DNAseI. The resuspended cells were lysed by sonication using the Branson Sonifier at an amplitude of 25 %, with a 5 min cycle at a pulse of 5 seconds on and 10 seconds off. The lysate .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 22 was cleared by centrifugation at 15,000 rcf for 30 minutes. The cleared lysate was filtered using a 0.2 µM filter and passed over a 5m L GSTrap, MBPTrap, or StrepTrap column (Cytiva) for 490 separation by affinity chromatography. Columns were pre -equilibrated with buffer A and unbound lysate after sample application with buffer A. Tagged ATG101 was then eluted in buffer A, supplemented with either 20 mM reduced glutathione at pH 7.0, 20 mM maltose, or 5 mM desthiobiotin depending on the column used. Eluted protein was concentrated in an Amicon-Ultra-15 Centrifugal Filter with a 10 kDa MWCO (Millipore). Sample was applied 495 for further purification using Size Exclusion Chromatopgraphy on a Superdex 75 Increase 10/300 GL column equilibrated with buffer A. Peak fractions corresponding to ATG101 at the optimal retention volume were collected and again concentrated in an Amicon -Ultra-4 Centrifugal Filter 10 kDa MWCO . Protein aliquots were flash-frozen in liquid nitrogen and stored at -80 °C. 500 ATG101 (E. coli) ATG101 with an N-terminal 6xHis Tag fusion was produced in E. coli cells. 100 µg plasmid was transformed in 50 µL Rosetta2 cells using heat-shock. Cells were grown in 500 ml cultures to reach an O.D. of 0.6 after which 0.2 mM IPTG was added. Cells were grown further for 16 505 h at 16 °C and harvested by centrifugation. Downstream purification steps remain the same as above except using a 5 mL HiTrap column for affinity chromatography. Column equilibration and unbound lysate washing was carried out using buffer A supplemented wi th 10 mM Imidazole. Tagged protein was eluted using buffer A supplemented with 250 mM Imidazole. Size Exclusion Chromatography and storage of protein was as above. 510 ATG9A .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 23 ATG9A with an N-terminal 6xHis-MBP tag was expressed in Hi5 insect cells using the biGBac expression system50,51. Cell pellet was resuspended in lysis buffer A supplemented with 1% DDM, 1 mM PMSF and 10 µg/50 mL DNAseI. Cells were lysed on ice for at least 1 h by stir-515 mixing. The lysate was then diluted 3 times with DDM-free buffer A and stir-mixed for another 30 min before clarifying by centrifugation at 15,000 rcf for 30 min. Columns were pre - equilibrated with buffer A and unbound lysate after sample application with buffer A. Tagged ATG9 was eluted in buffer A, supplemented with 20 mM maltose . Peak fractions were concentrated using a n Amicon-Ultra-15 Centrifugal Filter with a 100 kDa MWCO (Millipore) 520 followed by size exclusion chromatography using a Superose 6 10/300 or 16/600 column (Cytiva) pre -equilibrated with buffer B. The purified protein was concentrated using an Amicon-Ultra-4 Centrifugal Filter with a 10 0 kDa MWCO (Millipore), flash-frozen in liquid nitrogen, and stored at - 80 °C. 525 ATG13 ATG13 full-length and HORMA (1-200) constructs were cloned as N-terminal GST or 6xHis- MBP fusion constructs from Hi5 cells using the biGBac expression system. 5 ml of V2 baculovirus was infected in 400 -500 ml Hi5® cell cultures at 1x10E6/ml density. Cells were harvested at 72 h of infection. Cell pellets were washed and resuspended in buffer A 530 supplemented with 1 mM PMSF and 10 µg/50 mL DNAseI. The resuspended cells were lysed by sonication using the Branson Sonifier at an amplitude of 25 %, with a 5 min cycle at a pulse of 5 s on and 10 s off. The lysate was cleared by centrifugation at 15,000 rcf for 30 minutes. The cleared lysate was filtered using a 0.2 µM filter and passed over a 5mL GSTrap, MBPTrap, or StrepTrap column (Cytiva) for separation by affinity chromatography. Columns were pre -535 equilibrated with buffer A and unbound lysate after sample application with buffer A. Tagged ATG13 was then eluted in buffer A, supplemented with either 20 mM reduced glutathione at .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 24 pH 7.0 or 20 mM maltose depending on the column used. Eluted protein was concentrated in an Amicon-Ultra-15 Centrifugal Filter with a 10 kDa MWCO (Millipore). Sample was applied for further purification using Size Exclusion Chromatography on a Superose 6 Increase 10/300 540 GL column equilibrated with buffer A. Peak fractions corresponding to ATG13 at the optimal retention volume were collected and again concentrated in an Amicon -Ultra-4 Centrifugal Filter 10 kDa MWCO. Protein aliquots were flash-frozen in liquid nitrogen and stored at - 80 °C. 545 ULK1 ULK1 with an N -terminal GST-tag fusion was expressed in Hi5 cells using the biGBac expression system. Harvested cells were lysed by stir-mixing in buffer A supplemented with 1% DDM, 2 mM PMSF, 100 µM Leupeptin for 30 min - 1 h on ice. The lysate was then diluted 3 times with buffer A and clarified by centrifugation at 15,000 rcf for 30 min at 4 °C. The protein 550 was purified from the lysate by affinity chromatography using a 5 mL GSTrap column (Cytiva) pre-equilibrated with buffer A and eluted using buffer A supplemented with 20 mM reduced glutathione at pH 7.0, as elution buffer. The purified protein was concentrated, flash-frozen in liquid nitrogen and stored at -80 °C. 555 GST, MBP and Strep pulldown assays GST, MBP and Strep pulldown experiments were performed using GSH Sepharose beads, MBP beads (Amylose resin) (NEB), or StrepTactin Superlow Plus (Qiagen) beads pre -equilibrated by washing in buffer B. Concentrations of proteins mentioned in the figure legends were incubated separately before being added to 20 µL of corresponding bead resin. Beads were 560 spun down at 500 rcf for 15 s. The supernatant was removed, and beads were washed twice with 300 μL buffer, with each wash & spin -down cycle not more than 30 s . The supernatant .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 25 was removed completely, and 10µL of 4x Laemmli sample loading buffer was added to the beads. Samples were run on either 12 % SDS–PAGE gel made in house or a 4-15 % TGX-gel (BioRad). Bands were visualized with Coomassie Brilliant Blue (CBB) staining. Detailed 565 information regarding concentration, temperature, and other variations in the pulldown is provided in the figure legends. Cross-linking Mass Spectrometry To cross-link the ATG101 dimer, two approaches were taken. First, 5 µM of ATG101 was pre-570 incubated for 1 h at 20 °C, followed by a 30-min incubation at 20 °C with the indicated amount of BS3 (Thermo Scientific), and subsequent quenching by addition of 1M Tris -HCl, pH 7.0 (final concentration 25 mM) for 15 min. Second, 2 µM of MBP -ATG101 was incubated for 1 h at 20 °C with 6 µM of Strep -ATG101. The formed dimer -complex was pulled down and washed to remove excess unbound protein (as in the procedure for in vitro pull-down assay 575 described above). Proteins were then separated by SDS -PAGE using a 4 –12 % gradient gel (BioRad). Upshifted bands corresponding to the cross -linked complexes were excised from the gel and in-gel digested with trypsin (Shevchenko et al.et al., 2006). In brief, samples were reduced with 580 10 mM dithiothreitol and alkylated with 55 mM iodoacetamide and subsequently digested with trypsin (sequencing grade, Promega) at 37 °C for 18 h. Extracted peptides were dried in a SpeedVac vacuum concentrators (Thermo Scientific) and dissolved in a buffer composed of 5 % acetonitrile and 0.1 % trifluoroacetic acid. Samples were analyzed by liquid chromatography electrospray ionization mass spectrometry (LC -ESI-MS). For this, peptides were online 585 separated on a Dionex UltiMate 3000 UHPLC system (Thermo Scientific) equipped with an in-house-packed analytical C18 column (75 µm x 300 mm, ReproSil-Pur 120 C18-AQ, 3 µm, .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 26 Dr. Maisch GmbH) and coupled to an Exploris 480 mass spectrometer (Thermo Scientific). The latter was operated in data -dependent mode with the following settings: MS1 resolution, 120 000, normalized AGC target, 300%, scan range 350 -1550 m/z; cycle time 3 s ; MS2 590 resolution, 30 000, normalized collision energy for HCD fragmentation, 30 %, maximal injection time, 128 ms, normalized AGC target, 75%, isolation window, 1.6. Only precursors with charge state of 3-8 were selected for MS2. To verify protein composition of the samples, Thermo raw files were searched first with MaxQuant 2.6.7.0 against a database encompassing UniProt reference proteomes of Homo sapiens and Trichoplusia ni as well as the sequences of 595 MBP- and Strep-tagged ATG101 and protein contaminants commonly found in MS samples. Cross-linked peptides were identified in a subsequent search using pLink3.0.16 46 and a database containing only trypsin, MBP- and Strep-tagged ATG101 sequences. FDR was set to 1% at spectral level. The crosslinks were visualized using xiNET47. 600 Sortase-mediated fluorescence labelling of ATG101 To make a fluorescently labelled peptide for enzymatic conjugation by SortaseA, 1 mol GGGC tetrapeptide (dissolved in 100 % DMSO, ordered from GenScript) was mixed with 2 mol Alexa-488® C-5 Maleimide dye (dissolved in 100 % DMSO, ThermoFisherTM scientific) for 2 h at 20 °C in a dark environment. The reaction was quenched by adding 10 mM DTT. The 605 resulting labelled peptide GGGC-Alexa488 was flash-frozen stored at -80 °C. To fluorescently label ATG101, 30-40 µM of purified ATG101 in buffer A was mixed with 1 µM SortaseA enzyme (purified in house), 10 mM CaCl 2 and 150 µM GGGC -Alexa488 labelled peptide. The reaction mixture was incubated at 27 °C for 1 -2 h in a thermoshaker. Labelled 610 protein obtained was separated from enzyme and excess peptide by SEC with UV -490nm monitoring. Absorbance of collected protein was measured at 280 and 490 nm using .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 27 NanoDropTM one to determine labelling efficiency. Labelled protein was flash -frozen and stored at -80 °C for further use. 615 In vitro phosphorylation of ATG101 For phosphorylation by ULK1, ATG101 purified from E. coli was used as substrate. In-house purified ULK1 at a final concentration of 3 -5 µM, was mixed with ATG101 in buffer A supplemented with 0.5 mM ATP and 10 mM MgCl 2 and incubated at 25 °C for 30 min. Phosphorylated ATG101 was separated from ULK1 and excess ATP by SEC. Phosphorylated 620 protein was flash -frozen and stored at -80 °C for further use. Analysis of phosphorylated protein by gel electrophoresis was either done by separation on SuperSepTM PhosTag (Fujifilm) gels or normal 12 or 15 % SDS-PAGE gels as mentioned in the captions, followed by staining with Pro-Q DiamondTM Phosphostain (Thermo Fisher) as per manufacturer’s protocol. 625 Mass Spectrometry to confirm phosphorylation sites ATG101 purified from E.coli was phosphorylated in vitro using ULK1 kinase as described below. The samples further purified by SEC for kinase and ATP removal were loaded on a 12% SuperSepTM PhosTag12.5% gel (Fig.). The upshifted band signifying phospho -ATG101 was checked to confirm previously known ULK1 phosphorylation sites by mass spectrometry. For 630 this, in-gel digestion was performed as described above for cross -linked samples followed by an enrichment of phospho -peptides using Titanium Dioxide TopTips (Glygen) and a subsequent LC -ESI-MS analysis. MASCOT 2.3.02 was used as a search engine. Carbamidomethylation of cysteine was set as fixed modification; oxidation of methionine and phosphorylation of serine/threonine/tryptophan was considered as variable modifications. 635 Peptides corresponding to Ser11 and Ser203 are shown in Supplementary Table S2. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 28 Fluorescence polarization assay All FP -based binding measurements were performed using the BioTek Synergy Neo2 microplate reader (Agilent). For kinetic experiments measuring association rates, labelled 640 ATG101 (or phosphorylated ATG101) or labelled ATG13 -101 was prepared in buffer B in concentrations indicated in the figure legends and added to a 96 -well assay plate. MBPATG9A or MBPATG13 full-length at concentrations indicated in the figures was added into ATG101 or ATG13-101 containing wells and measurements recorded. The final reaction volume in each well was 100 µL. For longer measurements, a time course was set for 12 h, readings taken 645 every 45 s. Gain was set to 50/50. For shorter measurements, a time course was set for 30 min or 1 h, readings taken every 10 s. Gain was set to 50/50. Fluorescence polarization was selected as the readout, using filter cubes Dual FP (top) 480/520 (bottom) (Agilent). Temperature at the time of all recordings was 22 °C ± 1 °C. Data analysis for kinetic experiments was performed by fitting obtained curves to a one -phase association equation using GraphPad Prism ®. 650 Observed rate contstants kobs from this equation were plotted against the concentrations to fit into a linear regression. Association rates k on were given by the slope of the equation in units µM-1 h-1. For end-point saturation measurements, 10 nM of labelled ATG101 or 5 nM of ATG13 -101 655 was pre-incubated for 16 h with serial dilutions of MBPATG9A or MBPATG13 full-length. An end-point measurement of fluorescence -polarization was performed using the same gain settings and filter cubes as mentioned above. Data obtained were plotted against the concentrations and fitted with a one-phase specific binding equation using GraphPad Prism ® to obtain the equilibrium dissociation constants (Kd). 660 SDS-PAGE densitometric quantification .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 29 For SDS-PAGE quantification of relative stoichiometries , the density of each protein band scanned by a gelscanner (Epson) was calculated using GelAnalzyer (version 19). Densities were divided by the molecular weight of the protein and ratios were calculated by normalizing 665 to the bait protein. Change in stoichiometries were visualized as bar plots. The results (n=3) are shown as mean + standard deviation calculated with GraphPad Prism (version 9.0.0). Mass Photometry Mass Photometry measurements were performed using a OneMP mass photometer (Refeyn 670 Ltd, Oxford, UK). Data was acquired using the AcquireMP software (Refeyn Ltd. v2.3). For measurement, a drop of immersion oil was applied to the objective lens. Silicon gasket wells to hold the samples were then fixed on a clean dust-free cover slip placed on the lens stage. To find focus, filtered and degassed buffer A was pipetted into one gasket well and the focal point locked. Each sample at concentrations indicated in figure legends were pipetted add mixed nto 675 a gasket well, and data were acquired with an acquisition time varying between 30 – 60s. The timing was adjusted to get a good number of landing events while avoiding saturation. DiscoverMP software (Refeyn Ltd. v2.3) was used to analyze the data. CD spectrometry and prediction of secondary structures 680 All CD measurements were performed using the JASCO J815 spectrometer (JASCO inc.). 4 µM protein in buffer A was added to a rectangular quartz cell cuvette (0.1 cm pathlength) (JASCO inc.). A buffer blank was recorded before each measurement and automatical ly subtracted from the sample measurement. Spectra were recorded in a wavelength range of 260 – 190 nm with a data pitch of 1 nm. The scanning mode was set to continuous with a speed of 685 50 nm/min. Three repeats were collected for each spectrum. SESCA v097 was used for the analysis of the spectra. Before estimating the percentage of secondary structures, the spectrum .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 30 was processed using SESCA process. Only data points within wavelengths of 250nm -205nm were chosen due to voltage limitations and the CD spectrum was converted to 1000 mean residue ellipticity units. SESCA Bayes was used for the structure estimations using the DS-dT 690 basis set (1500 iterations where performed and a fraction of 0.1 was discarded). Heatmaps were generated using SESCA projection from the Bayesian estimations. Bacterial and virus strains DH10EMBacY™ Geneva Biotech Cat#10361012 NEB®5-alpha New England Biolabs Cat#C2987H Rosetta2™ Novagen Cat#71402 695 Cell lines Sf9 insect cells (Spodoptera frugiperda) Invitrogen Cat#10503433 High Five insect cells (Trichopulsia Ni) Invitrogen Cat#10747474 Culture media LB medium BioChemica Cat#23143289 LB-agar Roth Cat#X969.2 ESF 921 medium Expression Systems Cat#96-001-01 Sf-900 medium Gibco Cat#12658-019 Chemicals, peptides, and recombinant proteins 700 Dodecyl-β-D-maltoside (DDM) Carl Roth Cat#69227-93-6 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 31 Triton X-100 Merck Cat#9036-19-5 Amylose resin NEB Cat#E8021L Pierce™ Glutathione Agarose Thermo Fisher Scientific Cat#16101 Strep-Tactin Superflow Plus IBA LifeSciences Cat#30004 SERV A FastLoad 1 kb DNA Ladder SERV A Cat#39317.01 SERV A DNA Stain G SERV A Cat#39803 Protein Dual Color Standards BioRad Cat#1610374 Gel Loading Dye, Purple (6X) NEB Cat#B7024S BS3 (Bis(sulfosuccinimidyl)suberat) Thermo Fisher Scientific Cat#21580 Maltose Monohydrat Merck Cat#6363-53-7 L-Glutathion Merck Cat#70-18-8 X-TREMEGENE 9 DNA Tranfection Merck Cat#6365779001 4–15% Mini -PROTEAN® TGX ™ Precast Protein Gels BioRad Cat#4561083 96-well microplate (non-binding) Corning CoStar Cat#655906 0.45um Syringe Filter AMSTAT Cat#C0000629 Amicon® Ultra -15 concentrator 100 kDa Milipore Cat#UFC910024 Amicon® Ultra -15 concentrator 10 kDa Milipore Cat#UFC903024 T5 exonuclease Epicentre Cat#T5E4111K BSA Roth Cat#8076 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 32 Phusion® High -Fidelity DNA Polymerase NEB Cat#M0530 Q5® High-Fidelity DNA Polymerase NEB Cat#M0491 Taq DNA Polymerase NEB Cat#M0267 Taq DNA Ligase NEB Cat#M0208 BamHI NEB Cat#R0136S HindIII NEB Cat#R0104S Benzonase® Merck Cat#E1014 DpnI NEB Cat#R0176S Alexa Fluor™ 488 C5 Maleimide dye Invitrogen Cat#A10235 GGGC tetrapeptide GenScript Custom order Calcium Chloride Roth Cat#275P.1 IPTG Roth Cat#2316.5 SuperSep TM Phos-Tag Precast gel FujiFilm Cat#193-16571 Pro Q TM Diamond Phosphoprotein Gel Stain Thermo Fisher Scientific Cat#P33300 Recombinant DNA pColi-6x-His-ATG101-LPETGG This study pLIB-2xStrepII-ATG101-LPETGG This study pColi-6x-His-ATG101-L29R-H30R-LPETGG This study pLIB-6x-His-MBP-ATG9 Nguyen, Lugarini et al., 2023 pBIG-6x-His-MBP-ATG13HORMA-ATG101 Nguyen, Lugarini et al., 2023 pBIG-GST-ATG13HORMA(1-200) -ATG101 Nguyen, Lugarini et al., 2023 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 33 pLIB-6x-His-MBP-ATG13 Nguyen, Lugarini et al., 2023 pLIB-GST-ULK1 Nguyen, Lugarini et al., 2023 pLIB-6x-His-MBP-ATG13HORMA (1-200) Nguyen, Lugarini et al., 2023 pLIB-GST- ATG13HORMA (1-200) Nguyen, Lugarini et al., 2023 pLIB-6x-His-MBP-ATG101 Nguyen, Lugarini et al., 2023 pLIB-GST-ATG101 Nguyen, Lugarini et al., 2023 Commercial assay kits QIAprep Spin Miniprep Kit Qiagen Cat#27106 Plasmid Midiprep Kit Invitrogen Cat#K210015 QIAquick PCR purification Kit Qiagen Cat#28106 705 Software and algorithms JalView Barton Group (University of Dundee) 2.11.1.5 PyMol Schrödinger, LLC 2.5.2 AlphaFold 2 DeepMind (Alphabet Inc.) 2 Adobe Creative Cloud Adobe Inc. 5.8.0.592 Graphpad PRISM GraphPad Software Inc. 9.0.0 ImageJ NIH, USA 1.53t ChimeraX 1.4 UCSF 1.4 Snapgene GSL Biotech 5.3.2 Microsoft® Office Microsoft Corporation 16.64 SESCA Dr. Gabor Nagy V o97 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 34 BeStSel ELTE Eötvös Loránd University, Budapest, Hungary Columns and resins used for chromatography MBPTrap™ HP (1 ml; 5 ml) Cytiva Cat#29048641; Cat#28918778 HiTrap TALON crude (1 ml; 5ml) Cytiva Cat#29048565; Cat#28953766 GSTrap™ HP (1 ml; 5 ml) Cytiva Cat#17528101; Cat#17528201 StrepTrap HP (5ml) Cytiva Cat#28907546 HiLoad 16/600 Superdex 75 pg Cytiva Cat#28989333 HiLoad 16/600 Superdex 200 pg Cytiva Cat#28989335 HiLoad 16/600 Superose 6 pg Cytiva Cat#29323952 Superdex 75 Increase 10/300 GL Cytiva Cat#29148721 Superdex 200 Increase 5/150 GL Cytiva Cat#28990945 Superose 6 Increase 10/300 GL Cytiva Cat#29091596 Superose 6 Increase 5/150 GL Cytiva Cat#29091597

710 1 Turco, E., Fracchiolla, D. & Martens, S. Recruitment and Activation of the ULK1/Atg1 Kinase Complex in Selective Autophagy. J Mol Biol 432, 123 -134, doi:10.1016/j.jmb.2019.07.027 (2020). 2 Ganley, I. G. et al. ULK1.ATG13.FIP200 complex mediates mTOR signaling and is essential for autophagy. J Biol Chem 284, 12297-12305, doi:10.1074/jbc.M900573200 715 (2009). 3 Hosokawa, N. et al. Atg101, a novel mammalian autophagy protein interacting with Atg13. Autophagy 5, 973-979 (2009). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 35 4 Jung, C. H. et al. ULK-Atg13-FIP200 complexes mediate mTOR signaling to the autophagy machinery. Mol Biol Cell 20, 1992 -2003, doi:10.1091/mbc.E08 -12-1249 720 (2009). 5 Mercer, C. A., Kaliappan, A. & Dennis, P. B. A novel, human Atg13 binding protein, Atg101, interacts with ULK1 and is essential for macroautophagy. Autophagy 5, 649- 662, doi:10.4161/auto.5.5.8249 (2009). 6 Egan, D. F. et al. Small Molecule Inhibition of the Autophagy Kinase ULK1 and 725 Identification of ULK1 Substrates. Mol Cell 59, 285 -297, doi:10.1016/j.molcel.2015.05.031 (2015). 7 Mercer, T. J. et al. Phosphoproteomic identification of ULK substrates reveals VPS15- dependent ULK/VPS34 interplay in the regulation of autophagy. EMBO J 40, e105985, doi:10.15252/embj.2020105985 (2021). 730 8 Dalle Pezze, P. et al. ATG13 dynamics in nonselective autophagy and mitophagy: insights from live imaging studies and mathematical modeling. Autophagy 17, 1131- 1141, doi:10.1080/15548627.2020.1749401 (2021). 9 Zachari, M. et al. Selective Autophagy of Mitochondria on a Ubiquitin -Endoplasmic- Reticulum Platform. Dev Cell 50, 627 -643 e625, doi:10.1016/j.devcel.2019.06.016 735 (2019). 10 Shi, X. et al. ULK complex organization in autophagy by a C -shaped FIP200 N - terminal domain dimer. J Cell Biol 219, doi:10.1083/jcb.201911047 (2020). 11 Yamamoto, H. et al. The Intrinsically Disordered Protein Atg13 Mediates Supramolecular Assembly of Autophagy Initiation Complexes. Developmental Cell 38, 740 86-99, doi:10.1016/j.devcel.2016.06.015 (2016). 12 Hieke, N. et al. Expression of a ULK1/2 binding-deficient ATG13 variant can partially restore autophagic activity in ATG13 -deficient cells. Autophagy 11, 1471 -1483, doi:10.1080/15548627.2015.1068488 (2015). 13 Nguyen, A. et al. Metamorphic proteins at the basis of human autophagy initiation and 745 lipid transfer. Mol Cell 83, 2077-2090 e2012, doi:10.1016/j.molcel.2023.04.026 (2023). 14 Qi, S., Kim, D. J., Stjepanovic, G. & Hurley, J. H. Structure of the Human Atg13 - Atg101 HORMA Heterodimer: an Interaction Hub within the ULK1 Complex. Structure 23, 1848-1857, doi:10.1016/j.str.2015.07.011 (2015). 15 Wallot-Hieke, N. et al. Systematic analysis of ATG13 domain requirements for 750 autophagy induction. Autophagy 14, 743 -763, doi:10.1080/15548627.2017.1387342 (2018). 16 Kannangara, A. R. et al. BioID reveals an ATG9A interaction with ATG13-ATG101 in the degradation of p62/SQSTM1 -ubiquitin clusters. EMBO Rep 22, e51136, doi:10.15252/embr.202051136 (2021). 755 17 Karanasios, E. et al. Autophagy initiation by ULK complex assembly on ER tubulovesicular regions marked by ATG9 vesicles. Nat Commun 7, 12420, doi:10.1038/ncomms12420 (2016). 18 Ren, X. et al. Structural basis for ATG9A recruitment to the ULK1 complex in mitophagy initiation. Sci Adv 9, eadg2997, doi:10.1126/sciadv.adg2997 (2023). 760 19 Saitoh, T. et al. Atg9a controls dsDNA-driven dynamic translocation of STING and the innate immune response. Proc Natl Acad Sci U S A 106, 20842 -20846, doi:10.1073/pnas.0911267106 (2009). 20 Chen, M. et al. Structure and activation of the human autophagy -initiating ULK1C:PI3KC3-C1 supercomplex. Nat Struct Mol Biol , doi:10.1038/s41594 -025-765 01557-x (2025). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 36 21 Sawa-Makarska, J. et al. Reconstitution of autophagosome nucleation defines Atg9 vesicles as seeds for membrane formation. Science 369, doi:10.1126/science.aaz7714 (2020). 22 Kim, B. W. et al. The C -terminal region of ATG101 bridges ULK1 and PtdIns3K 770 complex in autophagy initiation. Autophagy 14, 2104 -2116, doi:10.1080/15548627.2018.1504716 (2018). 23 Rao, Y ., Perna, M. G., Hofmann, B., Beier, V . & Wollert, T. The Atg1-kinase complex tethers Atg9 -vesicles to initiate autophagy. Nat Commun 7, 10338, doi:10.1038/ncomms10338 (2016). 775 24 Michel, M. et al. The mammalian autophagy initiator complex contains 2 HORMA domain proteins. Autophagy 11, 2300 -2308, doi:10.1080/15548627.2015.1076605 (2015). 25 Mapelli, M., Massimiliano, L., Santaguida, S. & Musacchio, A. The Mad2 conformational dimer: structure and implications for the spindle assembly checkpoint. 780 Cell 131, 730-743, doi:10.1016/j.cell.2007.08.049 (2007). 26 De Antoni, A. et al. The Mad1/Mad2 complex as a template for Mad2 activation in the spindle assembly checkpoint. Curr Biol 15, 214-225, doi:10.1016/j.cub.2005.01.038 (2005). 27 Faesen, A. C. et al. Basis of catalytic assembly of the mitotic checkpoint complex. 785 Nature 542, 498-502, doi:10.1038/nature21384 (2017). 28 Gu, Y ., Desai, A. & Corbett, K. D. Evolutionary Dynamics and Molecular Mechanisms of HORMA Domain Protein Signaling. Annu Rev Biochem , doi:10.1146/annurev - biochem-090920-103246 (2022). 29 Porter, L. L. & Looger, L. L. Extant fold-switching proteins are widespread. Proc Natl 790 Acad Sci U S A 115, 5968-5973, doi:10.1073/pnas.1800168115 (2018). 30 Harkiolaki, M., Gilbert, R. J., Jones, E. Y . & Feller, S. M. The C-terminal SH3 domain of CRKL as a dynamic dimerization module transiently exposing a nuclear export signal. Structure 14, 1741-1753, doi:10.1016/j.str.2006.09.013 (2006). 31 Stewart, K. L., Dodds, E. D., Wysocki, V . H. & Cordes, M. H. A polymetamorphic 795 protein. Protein Sci 22, 641-649, doi:10.1002/pro.2248 (2013). 32 Kuloglu, E. S., McCaslin, D. R., Markley, J. L. & V olkman, B. F. Structural rearrangement of human lymphotactin, a C chemokine, under physiological solution conditions. J Biol Chem 277, 17863-17870, doi:10.1074/jbc.M200402200 (2002). 33 Susvirkar, V . & Faesen, A. C. Shieldin complex assembly kinetics and DNA binding by 800 SHLD3. Commun Biol 6, 384, doi:10.1038/s42003-023-04757-7 (2023). 34 Nguyen, A. & Faesen, A. C. The role of the HORMA domain proteins ATG13 and ATG101 in initiating autophagosome biogenesis. FEBS Lett 598, 114 -126, doi:10.1002/1873-3468.14717 (2024). 35 Sironi, L. et al. Crystal structure of the tetrameric Mad1 -Mad2 core complex: 805 implications of a 'safety belt' binding mechanism for the spindle checkpoint. EMBO J 21, 2496-2506, doi:10.1093/emboj/21.10.2496 (2002). 36 Wayment-Steele, H. K. et al. Predicting multiple conformations via sequence clustering and AlphaFold2. Nature 625, 832-839, doi:10.1038/s41586-023-06832-9 (2024). 37 Chakravarty, D., Lee, M. & Porter, L. L. Proteins with alternative folds reveal blind 810 spots in AlphaFold -based protein structure prediction. Curr Opin Struct Biol 90, 102973, doi:10.1016/j.sbi.2024.102973 (2025). 38 Perrakis, A. & Sixma, T. K. AI revolutions in biology: The joys and perils of AlphaFold. EMBO Rep 22, e54046, doi:10.15252/embr.202154046 (2021). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 37 39 Abramson, J. et al. Addendum: Accurate structure prediction of biomolecular 815 interactions with AlphaFold 3. Nature 636, E4, doi:10.1038/s41586 -024-08416-7 (2024). 40 Popp, M. W. & Ploegh, H. L. Making and breaking peptide bonds: protein engineering using sortase. Angew Chem Int Ed Engl 50, 5024-5032, doi:10.1002/anie.201008267 (2011). 820 41 Suzuki, H., Kaizuka, T., Mizushima, N. & Noda, N. N. Structure of the Atg101-Atg13 complex reveals essential roles of Atg101 in autophagy initiation. Nat Struct Mol Biol 22, 572-580, doi:10.1038/nsmb.3036 (2015). 42 Honorato, R. V . et al. The HADDOCK2.4 web server for integrative modeling of biomolecular complexes. Nat Protoc 19, 3219-3241, doi:10.1038/s41596-024-01011-0 825 (2024). 43 Rizzo, A. A. et al. Rev7 dimerization is important for assembly and function of the Rev1/Polzeta translesion synthesis complex. Proc Natl Acad Sci U S A 115, E8191- E8200, doi:10.1073/pnas.1801149115 (2018). 44 Nagy, G., Igaev, M., Jones, N. C., Hoffmann, S. V . & Grubmuller, H. SESCA: 830 Predicting Circular Dichroism Spectra from Protein Molecular Structures. J Chem Theory Comput 15, 5087-5102, doi:10.1021/acs.jctc.9b00203 (2019). 45 Nagy, G. & Grubmüller, H. How accurate is circular dichroism-based model validation? Eur Biophys J Biophy 49, 497-510, doi:10.1007/s00249-020-01457-6 (2020). 46 Mizushima, N. et al. Dissection of autophagosome formation using Apg5 -deficient 835 mouse embryonic stem cells. J Cell Biol 152, 657 -668, doi:10.1083/jcb.152.4.657 (2001). 47 Fujita, N. et al. An Atg4B mutant hampers the lipidation of LC3 paralogues and causes defects in autophagosome closure. Mol Biol Cell 19, 4651-4659, doi:10.1091/mbc.e08- 03-0312 (2008). 840 48 Tromer, E. C., van Hooff, J. J. E., Kops, G. & Snel, B. Mosaic origin of the eukaryotic kinetochore. Proc Natl Acad Sci U S A 116, 12873 -12882, doi:10.1073/pnas.1821945116 (2019). 49 Rosenberg, S. C. & Corbett, K. D. The multifaceted roles of the HORMA domain in cellular signaling. J Cell Biol 211, 745-755, doi:10.1083/jcb.201509076 (2015). 845 50 Altmannova, V ., Blaha, A., Astrinidis, S., Reichle, H. & Weir, J. R. InteBac: An integrated bacterial and baculovirus expression vector suite. Protein Sci 30, 108-114, doi:10.1002/pro.3957 (2021). 51 Weissmann, F. et al. biGBac enables rapid gene assembly for the expression of large multisubunit protein complexes. Proc Natl Acad Sci U S A 113, E2564 -2569, 850 doi:10.1073/pnas.1604935113 (2016). .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 38 Supplemental figures 855 Figure S1: The interaction of ATG101 with ATG13 and ATG9A is exceedingly slow (related to figure 1). a) AlphaFold3 predicted models of ATG101 (left panel), with the ATG9-HDIR peptide (centre) and with Ser203 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 39 phosphorylation (right). Confidence of the predictions are indicated by respective pTM scores. Subset on the bottom left of the first panel shows the ‘open’ conformation of MAD2. b) Size exclusion chromatogram and 860 SDS-PAGE of fluorescently labelled ATG101 used in the FP sensor throughout this study. (c-e) Fluorescence anisotropy-based saturation binding experiments of interactions within the ATG9A-ATG13-ATG101 complex. Cartoons on the top left of each panel indicate the interaction measured. 10 nM of ATG101Alexa488 (in c and d) or ATG13-ATG101Alexa488 (e) were incubated with varying concentrations of ATG13 or ATG9A. Equilibrium dissociation constants (Kd) for each reaction were obtained by fitting a one-site specific binding equation in 865 Graphpad Prism. f,g) The interaction of ATG101 with ATG9 is slow. Time zero is the first time point after mixing 20 nM ATG101Alexa488 with indicated ATG9A concentrations. Experiment is a single measurements representative of at least three independent technical replicates of the experiment. After single exponential fitting of the curves in f, the apparent first order rate constants (kobs) were plotted as function of ATG9A concentration in g, with kon being the slope of the resulting line. j,k) Preformed ATG13-ATG101 complex 870 interacts 3-orders of magnitude faster with ATG9A. Time zero is the first time point after mixing 10 nM ATG101Alexa488–ATG13 with indicated ATG9A concentrations. Experiment is a single measurements representative of at least three independent technical replicates of the experiment. After single exponential fitting of the curves in j, the apparent first order rate constants (kobs) were plotted as function of ATG9A concentration in k, with kon being the slope of the resulting line. l,m) The interaction of ATG101 to ATG9A is 875 progressively inhibited at higher ATG101 concentrations. Time zero is the first time point after mixing 50 µM ATG9A with indicated ATG101 concentrations. Experiment is a single measurements representative of at least three independent technical replicates of the experiment. After single exponential fitting of the curves in l, the apparent first order rate constants (kobs) were plotted as function of ATG9A concentration in m. 880 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 40 Figure S2: ATG101 transiently homo-dimerizes (related to figure 2). a) Incubation of ATG101 mutated in the dimer interface (L29R H30R; reference 41) with cross-linker BS3 does not create in proteins larger then monomeric ATG101. Input is shown in lanes with no BS3 (0 mM). b) Quantification of three independent technical replicates of the experiment in 2C. c) Proposed models of conformationally symmetric or asymmetric 885 ATG101 homo-dimers generated using AlphaFold3 and HADDOCK 2.4. .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 41 Figure S3: Phosphorylation by ULK1 initiates ATG101 homo-dimerization (related to figure 3). a) Purified ATG101 is phosphorylated if okadaic acid is added during the last few hours of the insect cell 890 expression culture. b) Mass Photometry experiment showing the dimerization of purified ATG101 (150 nM) upon adding Okadaic acid to the insect cell expression culture. c) Phos-tagTM gel showing that ATG101 can be reversibly phosphorylated by ULK1 co-expression. d) Mass Photometry experiment showing the effect of mutating the ULK1 phosphorylation sites in ATG101 (150 nM). 895 .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 42 Figure S4: Activated ATG101 auto-catalytically elicits ATG9A-ATG13-ATG101 complex formation (related to figure 4). a,b) Quantification of three independent technical replicates of the experiment in 4A and 4C, respectively. c,d) Interaction of phosphorylated ATG101 at low concentrations with ATG13 is still exceedingly slow. Time zero is the first time point after mixing 20 nM phosphorylated ATG101Alexa488 with 900 indicated ATG13 concentrations. Time-dependent changes in Fluorescence Anisotropy signal are single measurements representative of at least three independent technical replicates of the experiment. After single exponential fitting of the curves in c, the apparent first order rate constants (kobs) were plotted as function of .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint 43 ATG13 concentration in d, with kon being the slope of the resulting line. e-g) Quantification of three independent technical replicates of the experiment in 4F, 4H, and 4I, respectively. 905 Supplementary Table 1. List of crosslinks found in ATG101. Crosslinked Residue 1 Crosslinked Residue 2 1 36 78 2 36 143 3 40 143 4 78 36 5 84 107 6 107 84 7 107 147 8 143 36 9 143 40 10 143 143 11 143 147 12 147 107 13 147 143 14 147 147 910 Supplemental Table 2: Phosphorylation sites of ATG101 by ULK1 Sequence Prob Modification Observed TIC (R)SEVLEVsVEGR(Q) 100% Phospho (+80) 64,22,997 402924 (R)SEVLEVsVEGR(Q) 100% Phospho (+80) 64,22,978 1132720 (R)SEVLEVsVEGR(Q) 100% Phospho (+80) 64,22,996 835967 (K)ISFQITDALGTsVTTTmR(R) 100% Phospho (+80), Oxidation (+16) 67,99,918 1369680 (K)ISFQITDALGTsVTTTmR(R) 100% Phospho (+80), Oxidation (+16) 67,99,917 989470 (K)ISFQITDALGTsVTTTmR(R) 100% Phospho (+80), Oxidation (+16) 1.019,4835 373058 (K)ISFQITDALGTsVTTTmR(R) 100% Phospho (+80), Oxidation (+16) 1.019,4834 247343 (K)ISFQITDALGTsVTTTmR(R) 100% Phospho (+80), Oxidation (+16) 1.019,4843 638139 (K)ISFQITDALGTsVTTTmR(R) 100% Phospho (+80), Oxidation (+16) 1.019,4840 175734 Coverage .CC-BY-NC-ND 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted June 28, 2025. ; https://doi.org/10.1101/2025.06.27.661946doi: bioRxiv preprint