The unique aspects of Atg5 (Figure 1A), outside of its role in canonical autophagy, have strong ex vivo and important in vivo phenotypes (Deretic and Wang, 2023). Inactivation of Atg5 in myeloid lineage renders mice excessively susceptible to acute infection with M. tuberculosis attributed to excessive inflammation (Wang et al., 2023; Castillo et al., 2012; Watson et al., 2012; Kimmey et al., 2015; Golovkine et al., 2023; Kinsella et al., 2023), which did not extend to other phases of infection as tested here in a murine model of latent tuberculosis (Mccune and Tompsett, 1956; Mccune et al., 1956; Scanga et al., 1999, Figure 1—figure supplement 1A–D). In a previous study, we could only partially explain the cellular parameters associated with acute infection, which included degranulation in neutrophils from Atg5^fl/fl Lyz2^Cre (also referred to in prior literature as Atg5^fl/fl LysM-Cre⁺) mice (Wang et al., 2023). Furthermore, markers of LyHYP (lysosome hypersensitivity phenotype) (Wang et al., 2023) in ATG5 knockout (KO) cells, such as galectin 3 (Gal3) response, could not be explained by the exocytic phenomena (Wang et al., 2023). We thus sought to uncover additional processes (Figure 1A) affected by ATG5 at the intracellular level. To include factors beyond the known processes, we analyzed proteomic data obtained with APEX2-ATG5 (Wang et al., 2023) and compared APEX2-ATG5^WT with APEX2-ATG5^K130R (an ATG5 mutant deficient in conjugation to ATG12) in cells treated with LLOMe, an agent routinely used to cause lysosomal damage (Jia et al., 2020b; Aits et al., 2015; Thiele and Lipsky, 1990; Eriksson et al., 2020; Bonet-Ponce et al., 2020; Tan and Finkel, 2022, Figure 1B, Figure 1—figure supplement 2A, B). The proximity biotinylation (APEX2)-based proteomic data indicated the presence in the vicinity of APEX2-ATG5 of several protein complexes that control important membrane trafficking pathways including retromer (Figure 1B). Beside the retromer VPS subunits, there were very few other VPS proteins (Figure 1—figure supplement 2C). VPS proteins were observed in ATG5 proteomic data by others (Baines et al., 2022, Figure 1—figure supplement 2D). VPS26 and VPS29 showed statistically significant increase in proximity to ATG5 in cells subjected to LLOMe treatment (Figure 1—figure supplement 2A–C). Using APEX2-SGALS1 proximity biotinylation LC–MS/MS as a control, we did not observe enrichment of retromer subunits in cells treated with LLOMe (Figure 1—figure supplement 1 S2E).

Association of endogenous ATG5 with endogenous retromer subunits (VPS26A, VPS29 and VPS35) was tested in coimmunoprecipitation (co-IP) experiments (Figure 1C–E). The interaction between retromer and ATG5 was detected in these experiments and showed increased association in cells subjected to lysosomal damage with LLOMe (Figure 1C–E). Thus, ATG5, a key component of the known atg8ylation E3 ligases (Mizushima, 2020; Deretic and Klionsky, 2024), is found in protein complexes with the retromer and this association is enhanced upon lysosomal damage.

During lysosomal damage a number of processes are set in motion to repair, remove, and regenerate/replenish lysosomes (Jia et al., 2022; Jia et al., 2020b; Jia et al., 2018; Nakamura et al., 2020; Jia et al., 2019; Eapen et al., 2021) including membrane atg8ylation-dependent processes of repair by ESCRT machinery (Corkery et al., 2024) and by lipid transfer via ATG2 (Tan and Finkel, 2022; Cross et al., 2023). We have reported that ATG5 KO cells display elevated Gal3 response to lysosomal damage agents including LLOMe (Wang et al., 2023). Gal3 is one of the well-characterized sentinel galectins alerting cellular homeostatic systems to lysosomal damage (Jia et al., 2022; Jia et al., 2020b; Aits et al., 2015; Jia et al., 2018; Jia et al., 2020a). Based on the observed interactions between ATG5 and retromer, we tested whether retromer, like ATG5 (Wang et al., 2023), affected Gal3 response to lysosomal damage. We used the previously established quantitative high content microscopy (HCM) approach, which provides unbiased machine-driven image acquisition and data analysis (Kumar et al., 2021a; Claude-Taupin et al., 2021; Jia et al., 2020b; Jia et al., 2018; Jia et al., 2020a). In the experiments herein HCM was based on >500 valid primary objects/cells per well and a minimum of 5 wells per sample, with the independent biological replicates being n ≥ 3 (≥3 separate plates) as described (Wang et al., 2023). Upon treatment with LLOMe, HeLa^VPS35-KO cells had increased Gal3 puncta relative to HeLa^WT cells, comparable to HeLa^ATG5-KO (Figure 2A, B, Wang et al., 2023). The Gal3 phenotype in VPS35 knockout cells was partially complemented by expression of GFP-VPS35 (Figure 2C, D). The observed elevated Gal3 recruitment to damaged lysosomes in the absence of VPS35 cannot be explained by increased pools of LAMP2 organelles in the cells, since the overall complement of lysosomes (quantified by LAMP2 puncta/cell) did not increase in LLOMe treated HeLa cells (Figure 2—figure supplement 1A, B). Another marker of lysosomal damage, ubiquitination response (Papadopoulos et al., 2017), was elevated in both VPS35 and ATG5 deficient cells (Figure 2E, F). This phenotype was confirmed in another cell line, Huh7 (Figure 2—figure supplement 1C, D). Thus, VPS35 and ATG5 defects have a similar effect on lysosomes, previously characterized and termed lysosomal hypersensitivity phenotype/LyHYP (Wang et al., 2023).

Not all aspects of LyHYP, previously observed in ATG5 knockouts (Wang et al., 2023), were seen in VPS35 KO cells. In the case of ATG5 inactivation (Huh7^ATG5-KO and HeLa^ATG5-KO), the recruitment of ALIX, an ESCRT component involved in early stages of lysosomal repair (Jia et al., 2020b; Skowyra et al., 2018; Radulovic et al., 2018) is abrogated (Wang et al., 2023), whereas in VPS35 KO cells (Huh7^VPS35-KO and HeLa^VPS35-KO) ALIX was efficiently recruited to lysosomes upon LLOMe treatment (Figure 2G, H, Figure 2—figure supplement 1E, F).

The elevated Gal3 and ubiquitin markers of increased lysosomal damage in cells with inactivated VPS35 or ATG5 suggest that both may play a related role in lysosomal resilience to membrane damage, which complements the previously observed effects of aberrant VPS35/retromer on lysosomal morphology and proteolytic capacity (Cui et al., 2019; Daly et al., 2023). The dissimilarities between ATG5 and retromer effects on the ALIX recruitment component can be explained by the previously described perturbed function of the ATG conjugation machinery specifically in ATG5 KO cells affirming the unique additional roles of ATG5 (Wang et al., 2023). Nevertheless, the striking similarities between ATG5 KO and VPS35 KO effects on heightened Gal3 and ubiquitin responses to lysosomal membrane damage suggest that the two systems (membrane atg8ylation and retromer) closely intersect.

To avoid the confounding, and possibly indirect, effects of lysosomal damage on endolysosomal sorting and trafficking in cells treated with LLOMe, it was necessary to test the effects of ATG5 KO on retromer function under basal conditions with unperturbed lysosomes. Our co-IP analyses with endogenous proteins (Figure 1C–E) indicated that ATG5 can be found in protein complexes with retromer subunits even without the lysosomal damage. We further confirmed this in co-IPs showing that GFP-VPS35 and YFP-VPS29 were found in complexes with endogenous ATG5 in resting cells, whereas an isotype IgG control did not pulldown retromer components (Figure 3A). Reverse co-IPs with GFP-VPS35 and YFP-VPS29 detected endogenous ATG5 and ATG12–ATG5 complexes in GFP/YFP immunoprecipitates (Figure 3B).

Having established that ATG5 and retromer interact under basal conditions, we tested the effects of ATG5 KO on localization of a well-established cargo for retromer-dependent sorting, GLUT1 (Steinberg et al., 2013; McGough et al., 2014). We used Huh7 cells, where we generated a panel of membrane atg8ylation and VPS35 CRISPR knockouts (Figure 3—figure supplement 1A), because they have a well-defined localization of GLUT1 at the plasma membrane or as intracellular profiles (Figure 3C), fully amenable (and comparatively more reliable than in HeLa cells) to unbiased quantification by HCM. ATG5 KO in Huh7 cells resulted in redistribution of GLUT1 from the plasma membrane to intracellular punctate profiles (Figure 3D, Figure 3—figure supplement 1B). This phenotype was identical to the one observed in VPS35 KO (Figure 3D, Figure 3—figure supplement 1B). The increased intracellular GLUT1 puncta overlapped with the lysosomal marker LAMP2 in both ATG5 KO and VPS35 KO cells (Figure 3C, E, Figure 3—figure supplement 1C). A similar albeit morphologically less distinct phenotype was observed in HeLa cells (Figure 3—figure supplement 1D–H). GLUT1 is sorted by retromer via VPS26’s interactor SNX27, with GLUT1 (and other similar cargo) being captured by the PDZ domain of SNX27 (Gallon et al., 2014; Steinberg et al., 2013). In WT cells, SNX27 was found in abundant intracellular profiles (Figure 3F). In both ATG5 KO and VPS35 KO cells, GLUT1 and SNX27 relocalized to the same intracellular compartment (Figure 3F, G; Figure 3—figure supplement 1I, J). We carried out lysosomal purification using the well-established LysoIP method (Jia et al., 2020b; Eapen et al., 2021; Jia et al., 2020a; Wyant et al., 2018), and quantified levels of GLUT1 and SNX27 in lysosomal preparations that were positive for LAMP2 and devoid of GM130 (Golgi) and PDI (ER) (Figure 3H, I). Both GLUT1 and SNX27 were enriched in lysosomal preparations from ATG5 KO and VPS35 KO cells relative to the WT cells (Figure 3H, I). These experiments biochemically identify the lysosomes as a compartment to which GLUT1 and SNX27 relocalize in ATG5 KO and VSP35 KO cells. The above findings show that a loss of ATG5 affects retromer-controlled trafficking events.

ATG5 is a part of the membrane atg8ylation machinery (Deretic and Lazarou, 2022; Mizushima, 2020; Deretic and Klionsky, 2024), also known under the term ‘LC3 lipidation’ (Kabeya et al., 2000), which functions within the canonical autophagy pathway but is also engaged in a wide array of noncanonical processes (Deretic and Lazarou, 2022; Kaur et al., 2023; Corkery et al., 2023; Boyle et al., 2023; Durgan and Florey, 2022; Deretic et al., 2024). We tested whether other known components of the atg8ylation apparatus affect retromer. The KOs in Huh7 cells of the components of the prototypical atg8ylation E3 ligase (Deretic and Lazarou, 2022; Mizushima, 2020) included: ATG5 and ATG16L1, as well as the previously characterized Huh7 lines knocked out for E1 and E2 enzymes, ATG7 (Wang et al., 2023) and ATG3 (Jia et al., 2022). These KOs all caused entrapment of GLUT1 in intracellular punctate profiles (Figure 4A, B) as well as an increase in GLUT1 colocalization with lysosomes (GLUT1⁺ LAMP2⁺ profiles; Figure 4C, D). We tested whether retromer component levels were altered in atg8ylation mutants and found no change in total cellular levels of VPS26, VPS29, and VPS35, tested in ATG3, ATG5, and ATG7 KO Huh7 cells (Figure 4—figure supplement 1A, B) as well as in ATG5 KO HeLa cells (Figure 4—figure supplement 1C, D). Thus, the entire ATG16L1-dependent atg8ylation apparatus is required to maintain proper retromer-dependent sorting.

Furthermore, comparison of the previously characterized HeLa^Hexa-KO cells (Nguyen et al., 2016), with all principal mATG8s inactivated, with their isogenic parental HeLa^WT cells, showed increased GLUT1 on lysosomes in HeLa^Hexa-KO (GLUT1⁺ LAMP2⁺ profiles; Figure 4E, F). We observed a similar phenotype in separate triple knockouts (TKOs) of LC3 subfamily and GABARAP subfamily of mATG8s (Nguyen et al., 2016) compared to HeLa^Hexa-KO (Figure 4—figure supplement 1E). Thus, at least two or more mATG8s from two different mATG8 subclasses (LC3s and GABARAPs) and the whole membrane atg8ylation machinery were engaged in and required for proper GLUT-1 sorting.

Membrane atg8ylation and mATGs play roles in diverse processes (Deretic and Lazarou, 2022) including canonical autophagy (Morishita and Mizushima, 2019). We tested whether the participation of atg8ylation in canonical autophagy, previously reported to affect retromer-dependent cargo sorting under nutrient-limiting conditions (Roy et al., 2017; Carosi et al., 2024), is the reason for reduced trafficking of GLUT1 in our experiments. We used KOs in Huh7 cells of ATG13 and FIP200/RB1CC1, two obligatory components of the canonical autophagy initiation machinery (Morishita and Mizushima, 2019), to test whether canonical autophagy under the basal conditions used in our study was responsible for the effects on GLUT1 sorting. We found that in the Huh7 cells knocked out for ATG13 (Wang et al., 2023) or FIP200 (Jia et al., 2022), there was no increase in intracellular GLUT1 puncta in contrast to the increase observed with the isogenic ATG5 and VPS35 mutants (Figure 5A, B). This was mirrored by no increase of GLUT1⁺LAMP2⁺ profiles in ATG13 and FIP200 KO cells whereas colocalization between GLUT1 and LAMP2A increased in ATG5 and VPS35 mutants (Figure 5C, D). To further confirm that autophagy initiation mutant cells retained the ability to perturb GLUT1 trafficking due to defective atg8ylation, we knocked down ATG5 in FIP200 KO cells (Figure 4—figure supplement 1F) and found that GLUT1 puncta and GLUT1⁺LAMP2⁺ profiles increased even in the FIP200 KO background with the effects nearing those of VPS35 knockout (Figure 5E–G, Figure 4—figure supplement 1G), with the difference between VPS35 KO and ATG5 KD being attributable to residual ATG5 levels in cells subjected to siRNA knockdowns. Thus, we conclude that atg8ylation but not canonical autophagy affects retromer function under basal conditions and that a functional atg8ylation apparatus is required for proper sorting of the retromer cargo GLUT1.

Rab7 is considered to be an interactor of the retromer (Harrison et al., 2014). Rab7 translocates in cells lacking the retromer component VPS35 to lysosomes (Jimenez-Orgaz et al., 2018) including endolysosomal domains with lysosomaly positioned mTORC1 regulatory machinery (Kvainickas et al., 2019). We thus tested whether absence of ATG5 and components of the atg8ylation apparatus affected Rab7 localization. ATG5, VPS35 as well as knockouts of all major components of the atg8ylation in Huh7 cells, displayed increased Rab7⁺ cytoplasmic profiles (Figure 6A, B) and overlap between Rab7 and lysosomes (LAMP2) (Figure 6C, Figure 6—figure supplement 1A). This was confirmed biochemically using lysosomal purification by LysoIP and immunoblotting for Rab7: there was an increase of Rab7 in lysosomal preparations from ATG5 and VPS35 KO cells (Figure 6D, E).

This effect extended to other components of the atg8ylation machinery, as KOs in ATG3, ATG7, and ATG16L1 caused a similar effect (Figure 6A–C, Figure 6—figure supplement 1A). In contrast, KOs in genes specific for canonical autophagy (ATG13 and FIP200) did not alter intracellular distribution of Rab7 (Figure 6F–H, Figure 6—figure supplement 1B). Rab7 distribution was affected by ATG5 knockdown in FIP200 KO cells (Figure 6—figure supplement 1C–F). These experiments mirror the effects of ATG5 and atg8ylation on GLUT1 trafficking, showing that ATG5 and atg8ylation machinery but not canonical autophagy is required for proper localization of Rab7.

Membrane atg8ylation has multiple presentations that include not only canonical autophagy but also encompass processes such as CASM (Durgan and Florey, 2022). We thus asked whether induction of CASM (Cross et al., 2023) could affect GLUT1 trafficking. Using LLOMe as one of the inducers of CASM (Cross et al., 2023), we detect increased membrane atg8ylation (LC3 puncta formation) in response to lysosomal damage (Figure 7A, B). When cells were treated with LLOMe (Cross et al., 2023) and another CASM inducer, monensin (Goodwin et al., 2021; Florey et al., 2015; Fletcher et al., 2018), GLUT1 trafficking to the plasma membrane was negatively affected and instead GLUT1 accumulated intracellularly and trafficked to lysosomes (Figure 7C–E, Figure 7—figure supplement 1A). Thus, lysosomal perturbation and induction of CASM may affect retromer-dependent sorting of GLUT1.

It was previously reported that under glucose starvation conditions, LC3A recruits TBC1D5, a GAP for Rab7 (Popovic et al., 2012; Borchers et al., 2021; Stroupe, 2018), away from this small GTPase (Roy et al., 2017) and a similar phenomenon could be occurring during CASM. However, under CASM-inducing conditions, no changes were detected (Figure 7—figure supplement 1B–D) in interactions between TBC1D5 and LC3A or in levels of VPS35 in LC3A co-IPs, a proxy for LC3A-TBC1D5-VPS29/retromer association. This suggests that CASM-inducing treatments and additionally bafilomycin A1 do not affect the status of the TBC1D5-Rab7 system. Moreover, expression of constitutively active Rab7^Q67L did not promote trafficking of GLUT1 to plasma membrane but rather caused accumulation of intracellular GLUT1 puncta (Figure 7F–H).

A complication of interpreting CASM as a membrane atg8ylation process responsible for effects on retromer-dependent sorting is that CASM is elicited by lysosomal perturbations using lysosomal pH modifying agents (Goodwin et al., 2021; Durgan and Florey, 2022; Florey et al., 2015; Fletcher et al., 2018), and lysosomal stress could be an independent factor preventing proper sorting of GLUT1. Endolysosomal membrane damage (LLOMe) or perturbations of luminal pH (monensin, bafilomycin A1) negatively affected GLUT1 sorting and caused it to accumulate in lysosomes (Figure 7C–E). Since bafilomycin A1 does not induce CASM (Cross et al., 2023) but disturbs luminal pH, we conclude that it is the less acidic luminal pH of the endolysosomal organelles, and not CASM, that is responsible for interference with the proper sorting of GLUT1. To additionally test this, we compared ATG16L1 full length (ATG16L1^FL) and ATG16L1^E230 (Rai et al., 2019) for complementation of the GLUT1 sorting defect in ATG16L1 KO cells (Figure 7I, J). ATG16L1^E230 (Rai et al., 2019) lacks the key domain to carry out CASM via binding to V-ATPase (Goodwin et al., 2021; Hooper et al., 2022; Xu et al., 2019; Fischer et al., 2020; Xu et al., 2022) but retains capacity to carry out atg8ylation. Both ATG16L1^FL and ATG16L1^E230 complemented mis-sorting of GLUT1 (Figure 7I, J). Collectively, these data indicate that it is not specifically an absence of CASM/VAIL but the absence of membrane atg8ylation in general that promotes GLUT1 mis-sorting.

The above experiments suggest the role of acidification and integrity of endolysosomal compartments in GLUT1 sorting. Membrane atg8ylation is important for lysosomal repair and homeostasis with specific downstream mechanisms. These processes include: (1) ATG2A, which transfers lipids during lysosomal repair (Tan and Finkel, 2022) and is engaged on damaged lysosomes upon membrane atg8ylation by LC3A (Cross et al., 2023), and (2) ESCRT-based lysosomal membrane repair (Jia et al., 2020b; Skowyra et al., 2018; Radulovic et al., 2018; Zhen et al., 2021), with membrane atg8ylation specifically recruiting an ESCRT regulator ALG2 (Corkery et al., 2024) and ESCRT-I component VPS37A (Javed et al., 2023). Hence, we tested whether these established systems, involved in maintenance of lysosomal membrane integrity, represented by ATG2 and VPS37, can explain the mechanism underlying the observed effects of membrane atg8ylation on retromer-dependent trafficking. When U2OS ATG2A/ATG2B double KO cells were compared to parental U2OS WT cells, we observed aberrant GLUT1 trafficking, presented as accumulation of cytoplasmic GLUT1 puncta and increased colocalization of GLUT1 with lysosomes (LAMP2A) (Figure 8A–C, Figure 8—figure supplement 1A). A similar effect was observed when the previously characterized VPS37A knockout cells (Javed et al., 2023) were tested (Figure 8D–F, Figure 8—figure supplement 1B). These effects were detected with or without induced lysosomal damage (Figure 8A–F).

We interpret these data as evidence that endolysosomal homeostasis and functionality, which is at all times maintained by membrane atg8ylation, even under basal conditions, contributes to proper function of retromer-dependent sorting of its cognate cargo such as GLUT1 (Figure 8G).

We tested whether ATG5 affects cation-independent mannose 6-phosphate receptor (CI-MPR). For this, we employed the previously developed methods (Figure 8—figure supplement 2A) of monitoring retrograde trafficking of CI-MPR from the plasma membrane to the TGN (Simonetti et al., 2019; Calcagni’ et al., 2023; Seaman, 2004; Kvainickas et al., 2017; Zhang et al., 2023). In the majority of such studies, CI-MPR antibody is allowed to bind to the extracellular domain of CI-MPR at the plasma membrane and its localization dynamics following endocytosis serves as a proxy for trafficking of CI-MPR. We used ATG5 KOs in HeLa and Huh7 cells and quantified by HCM retrograde trafficking to TGN of antibody-labeled CI-MPR at the cell surface, after being taken up by endocytosis and allowed to undergo intracellular sorting, followed by fixation and staining with TGN46 antibody. There was a minor but statistically significant reduction in CI-MPR overlap with TGN46 in HeLa^ATG5-KO that was comparable to the reduction in HeLa cells when VPS35 was depleted by CRISPR (HeLa^VPS35-KO) (Figure 8—figure supplement 2B, C). Morphologically, endocytosed Ab-CI-MPR appeared dispersed in both HeLa^ATG5-KO and HeLa^VPS35-KO cells relative to HeLa^WT cells (Figure 8—figure supplement 2D). Similar HCM results were obtained with Huh7 cells (WT vs. ATG5KO; Figure 8—figure supplement 2E, F). We interpret these data as evidence of indirect action of ATG5 KO on CI-MPR sorting via membrane homeostasis, although we cannot exclude a direct sorting role via retromer. We favor the former interpretation based on the magnitude of the effect and the controversial nature of retromer engagement in sorting of CI-MPR (Seaman, 2021; Simonetti et al., 2019; Buser and Spang, 2023; Cui et al., 2019; Kvainickas et al., 2017).

Finally, we addressed the strength of effects on GLUT1 trafficking observed with ATG5 KO vs. KOs in other membrane atg8ylation genes, which were significant in all experiments but presented effects lower in magnitude than ATG5 KO (Figures 4 and 6). One possibility was that since ATG3, ATG5, and ATG7 all participate in the same process of membrane atg8ylation, that our KO mutants had unequal gene inactivation levels. This was ruled out by comparing effects of ATG3 KO, ATG5 KO, and ATG7 KO cells for their effects on LC3 puncta formation in a well-defined system of starvation-induced autophagy. The results of these analyses indicated that all three KOs had indistinguishable effects on LC3 puncta formation during canonical autophagy induction by starvation (Figure 8—figure supplement 3A, B). Thus, the stronger effects of ATG5 KO on retromer vis-à-vis KOs in other atg8ylation genes reflected additional action of ATG5. In the absence of ATG5, less VPS26 and VPS35 could be pulled down in co-IPs with YFP-VPS29 (Figure 8—figure supplement 2C–F). We conclude that ATG5, in addition to contributing to the retromer-dependent sorting via membrane atg8ylation and its downstream effector mechanisms which maintain healthy endolysosomal organelles, has additional effects on the retromer.

HEK293T and HeLa cells were from ATCC (American Type Culture Collection). Huh7 cells were from Rocky Mountain Laboratories. HeLa HEXA and TKO cell lines and their parental cell line were from Michael Lazarou. U2OS and its ATG2A/B DKO derivative were from Fulvio Reggiori.

Atg5^fl/fl LysM-Cre^-, Atg5^fl/fl LysM-Cre⁺ mice were previously described (Castillo et al., 2012; Manzanillo et al., 2012).

All mice were housed in AAALAC-accredited Animal Research Facility (ARF) of the University of New Mexico Health Sciences Center (UNM-HSC) and institutionally approved husbandry conditions and approved breeding protocols were followed 23-201351-B-HSC. M. tuberculosis-infected animals were housed in a separate ABSL3 suite within the UNM HSC ARF facility and all staff followed strict ABSL3, BSL3, and animal protocols approved by the UNM HSC Biosafety Committee and the Institutional Animal Care and Use Committee. The protocol number was 23-201379-HSC. The study was compliant with all relevant ethical guidelines for animal research.

Antibodies from Abcam were ATG5 (1:2000 for western blot (WB), ab108327), ATG7 (1:2000 for WB, ab52472), GFP (1:2000 for WB; 1:300 for immunoprecipitation (IP), ab290), mCherry (1:1000 for WB, ab183628), VPS35 (1:500 for WB; 1:1000 for IF, ab10099), VPS29 (1:500 for immunofluorescence (IF), ab10160), GLUT1 (1:3000 western blot (WB), 1:1000 for immunofluorescence (IF), ab115730), CI-MPR(2G11) (2 μg/ml for antibody uptake assay, ab2733), RAB7 (1:2000 western blot (WB), and 1:500 for immunofluorescence (IF), ab137029).

Antibodies from Biolegend were ALIX (1:400 for immunofluorescence (IF), #634502) and Galectin-3 (1:200 for Immunofluorescence (IF), #125402).

Antibodies from Proteintech were VPS26A (1:500 for western blot (WB), 12804-1-AP), VPS35 (1:500 for western blot (WB), 1:1000 for immunofluorescence (IF), 10236-1-AP), and TBC1D5 (1:1000 for western blot (WB), 17078-1-AP).

Antibodies from Sigma-Aldrich were ATG3 (1:1000 for western blot (WB), #A3231), Ubiquitin (FK2, 1:500 for immunofluorescence (IF), 04-263), and mouse Anti-FLAG M2 (1:500 for immunofluorescence (IF), #F1804).

Antibodies from Cell Signaling were α-Tubulin (DM1A) (1:3000 for western blot (WB); #3873) and LAMP1 (1:3000 for immunofluorescence (IF), #9091).

Other antibodies used in this study were from the following sources: beta-Actin (1:500 for western blot (WB), sc-47778) and GAPDH (1:500 for western blot (WB), sc-47724) from Santa Cruz Biotechnology; LAMP2 (1:1000 for immunofluorescence (IF), H4B4) from DSHB of University of Iowa; ATG5 (1: 500 for western blot (WB), ASA-B0113) from Novateinbio, SNX27 (1:1000 for western blot (WB), MA5-27854) from Invitrogen.

Secondary antibodies labeled with Alexa Fluor 488, 568, 647 (1:500 for immunofluorescence (IF)) and IgG-HRP (1:10,000 for western blots (WB)) were from Thermo Fisher Scientific. IgG Polyclonal Antibody Goat anti-mouse IRDye 680 (LI-COR, 925-68020), and Goat anti-rabbit IRDye 800 (LI-COR, 926-32211) secondary antibodies were from LI-COR Biosciences.

Bafilomycin A1 (BafA1, InvivoGen; 13D02-MM), Monensin (Sigma, M5273), LLOMe (Sigma, L7393), Lipofectamine 2000 (Thermo Scientific, 11668019); Triton X-100 (OmniPur, 9410-OP), saponin (Sigma, S4521-25G). DMEM (Gibco, #11995040), and Penicillin–Streptomycin (1000 U/ml; Gibco, #15140122). OptiMEM from Life Technologies, Puromycin dihydrochloride (Sigma, P9620), Hygromycin B (Sigma, H0654).

Plasmids used in this study, such as ATG5 were generated by first cloning inserts into pDONR221 (Gateway Technology cloning vector, Thermo Scientfic) using a BP cloning reaction and the expression vectors were made utilizing LR cloning reaction (Gateway, Thermo Fisher) in appropriate (pDEST) destination vectors for immunoprecipitation assay. Addgene clones were: eGFP-Rab7 WT (Addgene, #12605), eGFP-Rab7^Q67L (Addgene, #28049), and eGFP-Rab7^T22N (Addgene, #28049). Additional plasmids were pDEST-3X flag (from Terje Jonansen), Flag-Atg16l1^FL , and Flag-Atg16l1^E230 (synthetic clone prepared by Thabata Duque). Plasmid transfections were performed using the Lipofectamine 2000/3000 Transfection Reagent (Thermo Fisher Scientific, #11668019).

The siRNAs were from Horizon Discovery (formerly known as Dharmacon): siGENOME Non-Targeting Control siRNA (Identifier: D-001810-01-05, Target sequence: UGGUUUACAUGUCGACUAA); siGENOME human ATG5 SMARTpool siRNA (Identifier: M-004374-04-0005). ATG5 was a pool of four different siRNAs targeting a single gene with individual siRNA sequences: GGAAUAUCCUGCAGAAGAA; CAUCUGAGCUACCCGGAUA; GACAAGAAGACAUUAGUGA; and CAAUUGGUUUGCUAUUUGA.

For this study, a total of 68 mice were used (35 Atg5^fl/fl LysM-Cre⁺ and 33 Atg5^fl/fl LysM-Cre^-). Early in the course of the study, one mouse died due to malocclusion and thus could not be included in the final analysis.

Inoculum was prepared by diluting M. tuberculosis Erdman frozen stock 1:50 in PBS/0.01% Tween for a final amount of ~7.38e⁶ CFU/ml. Inoculum was serially diluted five times at 1:10 each time in PBS/Tween. 50 µL aliquots of the third, fourth, and fifth dilutions were plated on 7H11 agar plates to determine actual inoculum CFUs (Actual inoculum for this study: 6.45e⁶ CFU/ml). Remaining inoculum was added to Glas-Col inhalation System and mice were infected via aerosol according to the following machine settings:

Glas-Col Cycle Settings:

M. tuberculosis Erdman diluted 1:50, targeting 200 CFU/mouse.

Immediately following aerosol infection, three C57BL/6 mice, infected in parallel with the experimental cohort, were euthanized to determine lung deposition CFUs. 5 × 200 µl aliquots of neat, homogenized tissue were grown for 2–3 weeks at 37°C and 5% CO₂. Initial deposition: ca. 100 CFUs per lung. This procedure was used for all subsequent CFU determinations in this study.

After a period of 2.5 weeks bacterial growth was assessed by determining lung CFUs (2 Cre⁺ and 1 Cre⁻ sacrificed).

Subsequently, mice were treated PO with antibiotics (0.1 g/l INH and 0.15 g/l RIF) in drinking water for 8 weeks and bacterial clearance by chemotherapy was assessed by determining lung CFUs (11 Cre⁺ and 10 Cre⁻ sacrificed).

The remaining mice were subjected to antibiotic washout for 7 weeks plus a spontaneous reactivation period with no treatment for 3 weeks at which time lung CFUs were determined (8 Cre⁺ and 8 Cre⁻ sacrificed).

After washout and reactivation periods, dexamethasone was administered by IP injection 5 times/week at 0.08 mg/mouse/day to induce immunosuppression. Mice were immunosuppressed in this manner for 6 weeks after which point lung CFUs were determined (14 Cre⁺ and 13 Cre⁻ sacrificed).

Knockout cells (HeLa^ATG5-KO, HeLa^VPS35-KO Huh7^ATG5-KO, Huh7^ATG3-KO, Huh7^ATG7-KO, Huh7^VPS35-KO) were generated by CRISPR/Cas9-mediated knockout system. The lentiviral vector lentiCRISPRv2 carrying both Cas9 enzyme and a gRNA targeting ATG5 (gRNA-puro: AAGAGTAAGTTATTTGACGT), ATG3 (gRNA-Hygro: GTGAAGGCATACCTACCAAC), ATG7 (gRNA-puro: CTTCCGTGACCGTACCATGC), VPS35 (gRNA-hygro: GCTCACCGTGAAGATGGACC), or Scramble (gRNA-hygro: GTGTAGTTCGACCATTCGTG) were transfected into HEK293T cells together with the packaging plasmids psPAX2 and pCMV-VSV-G at the ratio of 5:3:2. Two days after transfection, the supernatant containing lentiviruses was collected. Cells were infected by the lentiviruses with 8–10 μg/ml polybrene. 36 hr after infection, the cells were selected with puromycin (1–10 μg/ml) or hygromycin (100–500 µg/ml) for 1 week in order to select knockout cells. All knockouts were confirmed by Western blot. Selection of single clones was performed by dilution in 96-well.

Cells in 96-well plates were fixed in 4% paraformaldehyde for 5 min. Cells were then permeabilized with 0.1% saponin in 3% bovine serum albumin (BSA) for 30 min followed by incubation with primary antibodies for 2 hr and secondary antibodies for 1 hr. Hoechst 33342 staining was performed for 3 min. High content microscopy with automated image acquisition and quantification was carried out using a Cellomics HCS scanner and iDEV software (Thermo Fisher Scientific). Automated epifluorescence image collection was performed for a minimum of 500 cells per well. Epifluorescence images were machine analyzed using preset scanning parameters and object mask definitions. Hoechst 33342 staining was used for autofocusing and to automatically define cellular outlines based on background staining of the cytoplasm. Primary objects were cells, and regions of interest (ROIs) or targets were algorithm-defined by shape/segmentation, maximum/minimum average intensity, total area and total intensity, etc., to automatically identify puncta or other profiles within valid primary objects. Each experiment (independent biological repeats; n ≥ 3) consists of machine-identified 500 valid primary objects/cells per well, ≥5 wells/sample. All data collection, processing (object, ROI, and target mask assignments) and analyses were computer driven independently of human operators. HCM also provides a continuous variable statistic since it does nor rely on parametric reporting cells as positive or negative for a certain marker above or below a puncta number threshold.

CI-MPR antibody was either incubated with cell for 1 hr in 37°C incubator or pulse-chase with the following procedures: cells were preincubate at 4°C for 15 min, followed by 30 min incubation with antibody at 4°C, three times wash with PBS, and 30 min incubation in 37°C incubator. The cells were then fixed and subject to following procedure.

For co-IP, cells transfected with 8–10 μg of plasmids were lysed in ice-cold NP-40 buffer (Thermo Fisher Scientific) supplemented with protease inhibitor cocktail (11697498001; Roche) and 1 mM PMSF (93482; Sigma-Aldrich) for 30 min on ice. Lysates were centrifuged for 10 min at 10,000 × g at 4°C. Supernatants were incubated with (2–3 μg) antibodies overnight at 4°C. The immune complexes were captured with Dynabeads (Thermo Fisher Scientific), followed by three times washing with 1× PBS. Proteins bound to Dynabeads were eluted with 2× Laemmli sample buffer (Bio-Rad) and subjected to immunoblot analysis.

For immunoblotting, lysates were centrifuged for 10 min at 10,000 × g at 4°C. Supernatants were then separated on 4–20% Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad) and transferred to nitrocellulose membranes. Membranes were blocked in 3% BSA for 1 hr at RT and incubated overnight at 4°C with primary antibodies diluted in blocking buffer. They were then incubated with an HRP-conjugated secondary antibody, and proteins were detected using ECL and developed using ChemiDoc Imaging System (Bio-Rad). Analysis and quantification of bands were performed using ImageJ software.

Plasmids for LysoIP were purchased from Addgene. Cells were plated in 10 cm dishes in DMEM and 10% fetal bovine serum and transfected with pLJC5-TMEM192-3xHA or pLJC5-TMEM192-2XFLAG constructs at 75–85% confluency. After 24 hr of transfection, cells with or without treatment were quickly rinsed twice with PBS and then scraped in 1 ml of KPBS (136 mM KCl, 10 mM KH₂PO₄, pH 7.25 adjusted with KOH) and centrifuged at 3000 rpm for 2 min at 4°C. Pelleted cells were resuspended in 1000 µl KPBS and reserved 50 µl (for the whole cell lysate) before further processing. The remaining cells were gently homogenized with 20 strokes of a 2-ml homogenizer. The homogenate was then centrifuged at 3000 rpm for 2 min at 4°C and the supernatant was incubated with 100 µl of KPBS prewashed anti-HA magnetic beads (Thermo Fisher) on a gentle rotator shaker for 15 min. Immunoprecipitants were then gently washed three times with KPBS and eluted with 2× Laemmli sample buffer (Bio-Rad) and subjected to immunoblot analysis.

Cells were plated onto coverslips in 6-well plates. After treatment, cells were fixed in 4% paraformaldehyde for 5 min followed by permeabilization with 0.1% saponin in 3% BSA for 30 min. Cells were then incubated with primary antibodies for 2 hr and appropriate secondary antibodies Alexa Fluor 488 or 568 (Thermo Fisher Scientific) for 1 hr at room temperature. Coverslips were mounted using Prolong Gold Antifade Mountant (Thermo Fisher Scientific). Images were acquired using a confocal microscope (META; Carl Zeiss) equipped with a 63 3/1.4 NA oil objective, camera (LSM META; Carl Zeiss), and AIM software (Carl Zeiss).

The previously described HeLaFlp-In-APEX2-ATG5-WT and HeLaFlp-InAPEX2-ATG5-K130R cells (Wang et al., 2023) were incubated in complete medium supplemented with 500 µM biotin–phenol (AdipoGen) with or without 2 mM LLOMe for 30 min. A 1-min pulse with 1 mM H₂O₂ at room temperature was stopped with quenching buffer (10 mM sodium ascorbate, 10 mM sodium azide and 5 mM Trolox in PBS). All samples were washed twice with quenching buffer, and twice with PBS for 1 min. For LC–MS/MS analysis, cell pellets were lysed in 500 µl ice-cold lysis buffer (6 M urea, 0.3 M NaCl, 1 mM EDTA, 1 mM EGTA, 10 mM sodium ascorbate, 10 mM sodium azide, 5 mM Trolox, 1% glycerol and 25 mm Tris–HCl, pH 7.5) for 30 min by gentle pipetting. Lysates were clarified by centrifugation and protein concentrations were determined using Pierce 660 nm protein assay reagent. Streptavidin-coated magnetic beads (Pierce) were washed with lysis buffer. A total of 1 mg of each sample was mixed with 100 µl of streptavidin beads. The suspensions were gently rotated at 4°C overnight to bind biotinylated proteins. The flow-through after enrichment was removed and the beads were washed in sequence with 1 ml IP buffer (150 mM NaCl, 10 mM Tris–HCl, pH 8.0, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100) twice; 1 ml 1 M KCl; 1 ml of 50 mM Na₂CO₃; 1 ml 2 M urea in 20 mM Tris–HCl (pH 8.0); and 1 ml IP buffer. Biotinylated proteins were eluted and processed for mass spectrometry. Protein samples on magnetic beads were washed four times with 200 μl of 50 mM Triethyl ammonium bicarbonate (TEAB) with a 20-min shake time at 4°C in between each wash. Roughly 2.5 μg of trypsin was added to the bead and TEAB mixture and the samples were digested over night at 800 rpm shake speed. After overnight digestion the supernatant was removed, and the beads were washed once with enough 50 mM ammonium bicarbonate to cover. After 20 min at a gentle shake the wash is removed and combined with the initial supernatant. The peptide extracts are reduced in volume by vacuum centrifugation and a small portion of the extract is used for fluorometric peptide quantification (Thermo scientific Pierce). One microgram of sample based on the fluorometric peptide assay was loaded for each LC–MS analysis.

Peptides were desalted and trapped on a Thermo PepMap trap and separated on an Easy-spray 100 μm × 25 cm C18 column using a Dionex Ultimate 3000 nUPLC at 200 nl/min. Solvent A = 0.1% formic acid, Solvent B = 100% acetonitrile 0.1% formic acid. Gradient conditions = 2% B to 50% B over 60 min, followed by a 50–99% B in 6 min and then held for 3 min than 99% B to 2% B in 2 min and total run time of 90 min using Thermo Scientific Fusion Lumos mass spectrometer. The samples were run in DIA mode; mass spectra were acquired using a collision energy of 35, resolution of 30 K, maximum inject time of 54ms and an AGC target of 50 K, using staggered isolation windows of 12 Da in the m/z range 400–1000 m/z.

DIA data were analyzed using Spectronaut 14.10 (Biognosys Schlieren, Switzerland) using the directDIA workflow with the default settings. Briefly, protein sequences were downloaded from Uniprot (Human Proteome UP000005640), ATG5 from Uniprot and common laboratory contaminant sequences from https://thegpm.org/crap/. Trypsin/P specific was set for the enzyme allowing two missed cleavages. Fixed Modifications were set for Carbamidomethyl, and variable modification were set to Acetyl (Protein N-term) and Oxidation. For DIA search identification, PSM and Protein Group FDR was set at 1%. A minimum of 2 peptides per protein group were required for quantification. Proteins known to be endogenously biotinylated were excluded from consideration.

Data in this study are presented as means ± SEM (n ≥ 3). Data were analyzed with either ANOVA with Tukey’s HSD post hoc test, or a two-tailed Student’s t-test. For HCM, n ≥ 3 (independent experiments carried out on different 96-well plates). For each well in 96-well plates, ≥500 valid primary objects/cells were imaged and analyzed, with ≥5 wells per plate per sample. Statistical significance: p ≥ 0.05 (not significant); <0.05 (significant). For HCM, sample size was based on a historic power analysis (published studies), with large effect size (differences and variability derived from published work), power 80%, β 20%, and α 5%, assuming normal distribution and favoring type II false-negative errors over type I false-positive errors. Band intensity in immunoblots, n = 3 (biological replicates); no power analysis was performed.