Vibrio cholerae cytotoxin MakA induces noncanonical autophagy resulting in the spatial inhibition of canonical autophagy Free

Macroautophagy (hereafter referred to as autophagy) is a highly conserved intracellular degradation system through which cytoplasmic constituents are sequestered in double-membraned autophagosomes and delivered to lysosomes for clearance. In addition to maintaining cellular homeostasis, autophagy plays a prominent role in the cellular response to microbial pathogens as a mediator of both innate and adaptive immunity (Deretic et al., 2013). From the direct sequestration and elimination of intracellular pathogens (xenophagy) (Levine, 2005), to the regulation of antigen presentation (English et al., 2009; Schmid et al., 2007) and cytokine secretion (Dupont et al., 2011; Zhang et al., 2015), the known roles for autophagy in the detection and elimination of cellular pathogens are numerous. As a result, many pathogens have evolved sophisticated mechanisms to evade or subvert host autophagy (McEwan, 2017; Wu and Li, 2019).

Vibrio cholerae, the causative agent of the acute diarrheal disease cholera, is a highly motile Gram-negative bacterium existing naturally in aquatic environments (Clemens et al., 2017). As an extracellular pathogen, its ability to manipulate host cell function depends on the secretion of numerous exotoxins capable of entering host cells and altering cellular processes in favor of bacterial infection. Recently, we described a novel cytotoxin of V. cholerae, the motility-associated killing factor A (MakA), which we found to act as a potent virulence factor contributing to pathogenicity in the model host organisms zebrafish (Danio rerio) and Caenorhabditis elegans (Dongre et al., 2018). The makA gene was observed in all V. cholerae strains including non-O1 non-O139 V. cholerae (NOVC) isolates. Most NOVC strains lack the cholera toxin gene (ctxAB) and do not cause the typical diarrhoeal disease, cholera, but have recently been emerging as extra-intestinal pathogens (Faruque et al., 1998). Here, we characterize the molecular basis of the interaction between MakA and mammalian host cells. We found that MakA binding to the plasma membrane of host cells promoted endocytosis resulting in the formation of an endolysosomal membrane-rich aggregate at the perinuclear region of intoxicated cells. The aggregate was targeted by the noncanonical autophagy pathway mediating LC3 (herein referring to MAP1LC3B) lipidation on perturbed endosomal membranes. The end result was a spatial inhibition of canonical autophagy pathway by sequestering core autophagic components to the membranous aggregate. The spatial inhibition of autophagy was shown to reduce the unconventional secretion of interleukin-1β (IL-1β), suggesting a novel approach to immune modulation by a bacterial toxin.

MakA was recently identified in V. cholerae as a novel cytotoxin affecting both vertebrate and invertebrate hosts (Dongre et al., 2018), yet the molecular basis for its role as a virulence factor remained unknown. To determine the effect MakA has on human host cells, human colon carcinoma cells (HCT8) were treated with purified MakA at a dose approximately equal to the established median lethal concentration (LC50) (Fig. S1), and RNA sequencing (RNA-seq) analysis was performed to monitor gene expression in a comparison with untreated cells (Fig. 1A). Gene ontology (GO) enrichment analysis of differentially expressed genes identified several biological processes related to lipid metabolism and homeostasis, with the ‘cholesterol biosynthetic process’ being the most significantly modulated (Fig. 1B,C). Maintaining cholesterol homeostasis is essential to sustain cellular function. As such, elaborate regulatory networks exist to sense and maintain cellular cholesterol levels within a narrow range (Luo et al., 2020). To determine whether the observed transcriptional response was related to altered cellular cholesterol levels, HCT8 cells were treated with toxin, and total cellular cholesterol was quantified. Although MakA treatment did result in a slight dose-dependent increase in total cellular cholesterol (Fig. 1D), visualization of free cholesterol revealed a striking toxin-induced redistribution, with large cholesterol aggregates visible at the perinuclear region of intoxicated cells (Fig. 1E). A similar cholesterol redistribution was observed in cells treated with supernatant harvested from cultured V. cholerae (Fig. 1F).

To determine the composition of the MakA-induced cholesterol-rich aggregates, and gain insight into the subcellular localization of MakA, correlative light and electron microscopy (CLEM) was performed with cells treated with Alexa Fluor 568 (A568)-labeled MakA (Fig. 2A). The toxin was accumulated at both the plasma membrane and within the perinuclear aggregate, which appeared to be membrane rich and contained vesicular structures closely resembling endosomes (Klumperman and Raposo, 2014). Immunofluorescence analysis of MakA-treated cells, using antibodies against the early, late and recycling endosomal markers EEA1 (Simonsen et al., 1998) (Fig. 2B; Fig. S2), Rab7a (Feng et al., 1995) (Fig. S2) and Rab11a (Urbé et al., 1993) (Fig. S2), confirmed endosomal membrane enrichment within the membranous aggregate. Aggregates were also shown to contain plasma membrane-localized transmembrane proteins caveolin-1 (Cav1) (Fig. S3A) and Na/K-ATPase α1 (Fig. S3B), but were free of intracellular Golgi (Fig. S3C) or mitochondrial (Fig. S3D) membrane. Furthermore, live-cell imaging of Cos7 cells transiently transfected with the early endosomal marker EGFP-Rab5 (herein referring to Rab5a) (Bucci et al., 1992), and treated with A568-MakA, showed that toxin was present within Rab5⁺ endosomes (Fig. 2C; Movie 1), suggesting that MakA travels from the plasma membrane to the perinuclear membranous aggregate via endocytosis. To confirm endosomal involvement, Cos7 cells were transfected with a dominant-negative Rab5-S34N mutant (Li and Stahl, 1993). Expression of the mutant allele was sufficient to block A568-MakA-induced aggregate formation (Fig. 2D).

Interestingly, MakA-induced endocytosis was found to be pH-dependent. Unlike HCT8 cells, which showed robust aggregate formation after 24 h of MakA treatment (Fig. 1E), Cos7 cells did not develop MakA-induced aggregates until 48 h post toxin treatment. This response time could be decreased by adjusting the starting pH of the culture medium from pH 7.9 to pH 7.2 (Fig. 2E). This pH-dependent aggregate formation could be prevented by treating cells with the dynamin inhibitor dynasore (Macia et al., 2006) to block dynamin-dependent endocytosis (Fig. 2E). Together, these data confirm that the MakA-induced aggregate formation is formed as a result of toxin-induced dynamin- and Rab5-dependent endocytosis, and is composed of plasma membrane-derived, cholesterol-rich endosomal membrane.

RNA-seq identified a significant change in the expression of genes involved in the regulation of autophagy after toxin treatment 1 (Fig. 1B). To determine how MakA affects cellular autophagy, HCT8, HEK293 T and Cos7 cells were treated with toxin, and LC3 lipidation status was evaluated by western blotting (Klionsky et al., 2016). Upon the induction of autophagy, cytosolic LC3 (LC3-I) is conjugated to phosphatidylethanolamine (PE) to form membrane-bound LC3-II (Tanida et al., 2008). A strong dose- and time-dependent increase in LC3 lipidation was observed following MakA treatment in all cell lines tested (Fig. 3A,B). As was reported for MakA-induced endocytosis (Fig. 2E), the timing of LC3 lipidation could be controlled by altering the starting pH of the culture media (Fig. 3C). Interestingly, MakA-induced LC3 lipidation was restricted to within a narrow pH window (pH 7.5 to pH 6.6). Dropping below pH 6.6 prevented LC3 lipidation and inhibited cholesterol-rich aggregate formation (Fig. 3D).

To explore the relationship between aggregate formation and the induction of autophagy, we performed immunofluorescence analysis for endogenous LC3 in Cos7 cells treated with MakA. The majority of the LC3 localized at the perinuclear aggregate after MakA treatment (Fig. 3E). The aggregate was also enriched in mono- and poly-ubiquitylated conjugates (Fig. 4A); a common prerequisite for substrate recognition in selective autophagy (Kim et al., 2008; Pankiv et al., 2007), and the LC3-ubiquitin adaptor protein p62 (also known as SQSTM1) (Fig. 4B). Interestingly, both ubiquitylated conjugates and p62 localized primarily to the periphery of the aggregate, indicating that the aggregate could be targeted by the autophagic machinery after formation. We also observed clustering of lysosomes around the aggregate (Fig. 4C), similar to what has been described for autophagy-mediated degradation of abnormal polypeptide aggregates (aggresomes) (Zaarur et al., 2014).

To confirm that autophagy was induced in response to aggregate formation, dynamin-dependent MakA endocytosis was inhibited with dynasore to prevent aggregate assembly (Fig. 2E). Blocking aggregate formation was sufficient to prevent MakA-induced LC3 lipidation (Fig. 4D), confirming autophagy induction was dependent on MakA-induced aggregate formation. Furthermore, treatment of autophagy-deficient (Atg5^−/−) mouse embryonic fibroblasts (MEFs) with MakA did not induce LC3 lipidation (Fig. 4E), yet A568-MakA still assembled into perinuclear aggregates (Fig. 4F), confirming that autophagy is not required for aggregate formation. Together, these data demonstrate that the autophagy induction observed after MakA treatment is a result of aggregate formation.

To further characterize the nature of autophagy induction, MakA-induced LC3 lipidation was assessed in MEFs deficient for the unc-51-like kinase (ULK)-interacting protein FIP200 (also known as RB1CC1; the key component of the ULK1/2-ATG13-FIP200 complex). FIP200 is essential for autophagosome initiation in canonical autophagy (Hara et al., 2008), but dispensable for many noncanonical autophagy pathways (Florey et al., 2011; Jacquin et al., 2017; Martinez et al., 2015). MakA-induced LC3 lipidation was maintained in the absence of FIP200 (Fig. 4G), indicating activation of a noncanonical autophagy pathway. To explore further, we employed HEK293 cells deficient for ATG16L1, and stably expressing ATG16L1α, ATG16L1β or an ATG16L1 C-terminal deletion mutant (ATG16L1 1-249). Individual ATG16L1 isoforms/mutants have been recently described to play distinct roles in the regulation of LC3 lipidation on different intracellular membranes. ATG16L1α, ATG16L1β and ATG16L1(1-249) can support LC3 lipidation in canonical autophagy, but ATG16L1(1-249) cannot support LC3-associated phagocytosis (LAP). Only the β isoform supports VPS34-independent LC3 lipidation on perturbed endosomes (Fletcher et al., 2018; Lystad et al., 2019). MakA-induced LC3 lipidation was blocked in ATG16-knockout (KO) cells and restored in cells expressing the beta-isoform of ATG16L1, but not in cells expressing the α isoform or ATG16L1(1-249) (Fig. 4H). These results suggested that MakA induces neither canonical autophagy nor LAP, but specifically induces LC3 lipidation on perturbed endosomal membranes (Lystad et al., 2019), in line with the endolysosomal composition of the membrane, and lack of double-membrane structures observed in transmission electron microscopy images (Fig. 2A). Interestingly, MakA-induced membranous aggregates were found to be devoid of galectin 3 (LGALS3) (Fig. S4), a sensor of endomembrane damage implicated in the subsequent autophagic response (Chauhan et al., 2016).

ATG16L1, in complex with ATG5 and ATG12, localizes to the site of autophagosome initiation and functions as an E3-like enzyme to catalyze the conjugation of PE to LC3 (Fujita et al., 2008; Mizushima et al., 2003). ATG16L1 and ATG12, which are normally dispersed throughout the cytosol, were found to be enriched at MakA-induced aggregates after toxin treatment (Fig. 5A,B). With no observable change in the total protein levels of ATG16L1 or ATG12 after toxin treatment (Fig. 5C), we aimed to determine how sequestration of the E3-like complex might impact canonical and noncanonical autophagy pathways dependent on the same machinery. To assess canonical autophagy function, we explored the effect of MakA on autophagosome formation induced by mammalian target of rapamycin (mTOR) inhibition. In vehicle-treated cells, ATG16L1 and ATG12 were dispersed throughout the cytosol (Fig. 6A). After torin1 treatment, both proteins localized to punctate structures marking the sites of autophagosome formation (Fig. 6A). In MakA-treated cells, ATG16L1 and ATG12 were enriched at the aggregate before torin1 treatment. After treatment, we observed significantly fewer ATG16L1 and ATG12 puncta than in vehicle-treated cells (Fig. 6A,B). Furthermore, puncta that did appear were predominantly associated with residual aggregates, marked by mono- and poly-ubiquitylated conjugates (Fig. 6C; Movie 2). These data suggest that initiation site formation is inhibited in MakA-treated cells. To confirm this, Cos7 cells were transiently transfected with EGFP-LC3, pre-treated with MakA to induce aggregate formation, and treated with torin1 to induce autophagy. Cells were imaged every 30 s for 1 h to observe the total number and location of EGFP-LC3 puncta. MakA treatment not only resulted in a significant decrease in the number of EGFP-LC3 puncta, but the puncta which were quantified were predominantly associated with the existing aggregate (Fig. 6D; Movie 3, Fig. S5). These results confirm that the presence of MakA-induced endolysosomal aggregates leads to the inhibition of canonical autophagy.

Beyond their well-established roles in regulating intracellular degradation, autophagy-related proteins (ATGs) have been implicated as regulators of a wide variety of non-autophagic cellular processes (Cadwell and Debnath, 2018). One such process relevant to bacterial infection is the control of protein secretion (secretory autophagy) (Ponpuak et al., 2015). Secretory autophagy involves the transport of proteins via the autophagosome to the plasma membrane for extracellular release. Among the proteins that have a release that is known to be, at least partially, regulated by the autophagic machinery, is the pro-inflammatory cytokine IL-1β (Dupont et al., 2011; Kimura et al., 2017; Zhang et al., 2015; Öhman et al., 2014); a known mediator of the innate immune response to V. cholerae infection (Haines et al., 2005; Kayagaki et al., 2011; Queen et al., 2015; Toma et al., 2010). To determine whether sequestration of the autophagic machinery by MakA-induced aggregates would alter IL-1β secretion, human monocytes (THP-1) were pre-treated with increasing concentrations of toxin to induce aggregate formation and stimulated with lipopolysaccharide (LPS) to induce cytokine secretion. Unlike the V. cholerae accessory toxins hemolysin and multifunctional-autoprocessing repeats-in-toxin (MARTX) (Toma et al., 2010), MakA treatment was limited in its ability to induce IL-1β secretion, but did result in a significant dose-dependent inhibition of the IL-1β secretion in response to LPS stimulation (Fig. 6E). Importantly, MakA did not inhibit LPS-induced secretion of IL-8 (Fig. 6E), which is secreted through the autophagy-independent canonical ER-Golgi pathway and is insensitive to autophagy inhibition (Iula et al., 2018). This confirms that sequestration of the autophagic machinery at MakA-induced aggregates is capable of inhibiting canonical autophagy and autophagy-mediated processes.

Autophagy plays a critical role in host defense against many pathogens. As a result, evolution of pathogens has led to diverse mechanisms for evasion and subversion of host autophagy. In this study, we determine the molecular basis of the interaction between the recently identified V. cholerae toxin MakA and host cells, revealing a novel mechanism of autophagy modulation. After endocytosis into the host cell, MakA promotes the formation of an endolysosomal membrane-rich aggregate that becomes the target of the noncanonical autophagy pathway mediating LC3 lipidation in perturbed endosomal membranes. Ubiquitylation and p62 recruitment were observed at the membranous aggregate yet the aggregate was found to be devoid of LGALS3, a marker of endomembrane damage, suggesting MakA-induced LC3 lipidation of endosomal membranes occurs through a LGALS3-independent mechanism (Chauhan et al., 2016). We found that MakA specifically induces the noncanonical autophagy pathway, which was shown to be dependent on the membrane binding region and the WD-repeat containing C-terminal domain of ATG16L1 (Fletcher et al., 2018; Lystad et al., 2019). Sequestration of the E3-like ATG12-ATG5-ATG16L1 complex at the membranous aggregate resulted in the inhibition of canonical autophagy, as well as autophagy-mediated cytokine secretion. How the E3-like complex is recruited to the aggregate needs further investigation, but may be ubiquitin dependent because ubiquitin has been shown previously to bind the WD-repeat domain of ATG16L1 to enhance the recruitment of ATG16L1 to damaged endosomes during bacterial infection (Fujita et al., 2013).

Compared to multiple intracellular pathogens that are targeted by and subvert xenophagy, it is less clear why a noninvasive bacterium such as V. cholerae has evolved to regulate autophagy (McEwan, 2017; Wu and Li, 2019). Nonetheless, its importance is emphasized by the fact that several V. cholerae toxins have evolved to engage host cell autophagy. The pore-forming exotoxin V. cholerae cytolysin (VCC) has been shown to induce autophagy (Elluri et al., 2014; Gutierrez et al., 2007), the α/β-hydrolase effector domain of MARTX toxin has phosphatidylinositol 3-phosphate (PI3P)-specific phospholipase A1 (PLA1) activity capable of inhibiting autophagosome formation by reducing intracellular PI3P levels (Agarwal et al., 2015), and cholera toxin has been shown to inhibit autophagy through increasing cyclic AMP production (Shahnazari et al., 2011). MakA is unique in that it induces noncanonical host-cell autophagy, resulting in a spatial inhibition of canonical autophagy, with autophagosome formation restricted to within close proximity of the toxin-induced perinuclear aggregate. How, or if, these different autophagy-modulating toxins of V. cholerae may work in concert remains unknown, but our data suggest that the combined disruption of autophagy could serve to dampen the innate immune response to bacterial infection. Alternatively, the MakA-mediated spatial inhibition of autophagy could serve to shut off the host response to other secreted toxins to promote their effect. The MakA protein is a factor found in all pathotypes of V. cholerae, and it may be of particular relevance for the pathogenesis of V. cholerae that are causing not only the diarrhoeal disease cholera, but are responsible for severe extra-intestinal infections and bacteremia.

The narrow pH window in which MakA is shown to induce aggregate formation could provide site specificity within the human gastrointestinal tract due a natural pH gradient between the stomach and colon (Cook et al., 2012). The observed pH range at which MakA induced aggregate formation in vitro closely matches that of the small intestine (pH 6.15 to pH 7.88), the preferential site of colonization for V. cholerae. Thus, we hypothesize that MakA could contribute to V. cholerae pathogenicity through localized innate immune suppression at the colonization site. This pH dependence also raises an important caveat to modeling V. cholerae infection in mice, as mice have a lower intestinal pH than humans (McConnell et al., 2008). With a mean intestinal pH below 5, bacterial toxins that have evolved to function within a pH range relevant to the human intestinal tract would be inactive in mice. In agreement with this hypothesis, a ΔmakA V. cholerae mutant with reduced pathogenicity in C. elegans and zebrafish (Dongre et al., 2018) was not limited in its ability to colonize mice, nor did it impact cytokine secretion (Fig. S6). Thus, care must be taken when translating V. cholerae infection data from mice to humans.

Juxtanuclear aggregate formation has been reported in the cellular response to various pathogens including Clostridioides difficile (Miura et al., 2011), Salmonella enterica (López-Montero et al., 2016; Mesquita et al., 2012), prions (Kristiansen et al., 2005) and large cytoplasmic DNA viruses (e.g. iridovirus, vaccinia virus and African swine fever virus), which assemble virus factories for replication (Wileman, 2006). Although the composition and assembly of each aggregate differ slightly, many have been reported to engage host autophagy (Boellaard et al., 1991; Heiseke et al., 2010; López-Montero et al., 2016; Wileman, 2006). MakA-induced endomembrane-rich aggregates most closely resemble the lysosomal membrane glycoprotein-positive (LGP⁺) aggregates induced in response to Salmonella infection of fibroblasts (López-Montero et al., 2016). Similar to MakA-induced aggregates, LGP⁺ aggregates are endomembrane-rich and, although reportedly devoid of ubiquitin, engage the autophagic machinery for degradation. The aggregates simultaneously degrade intraphagosomal bacteria within their vicinity, and those bacteria that do not come within close proximity of the aggregate are spared from autophagic degradation (López-Montero et al., 2016). This further supports our hypothesis that autophagic targeting of cellular aggregates can result in a spatial restriction of autophagy. It also suggests that the cytoplasmic aggregate generation observed in response to multiple pathogens may represent a common mechanism of subverting host autophagy.

The wild-type Vibrio cholerae O1 El Tor strain A1552 (Yildiz and Schoolnik, 1998) and A1552 ΔmakA (Dongre et al., 2018) were used in this study. Bacterial supernatants were isolated under non-cholera toxin producing conditions. Briefly, bacteria were grown to an optical density at 600 nm (OD_600nm) of 2.0 in Luria-Bertani (LB) broth. Bacterial culture (1 ml) was collected and centrifuged at 8000 g for 5 min. The supernatant was isolated and filtered through a 0.45-μm PVDF membrane filter (Millipore, Merck Chemicals and Life Science). The resulting culture supernatant was used for the treatment of HCT8 cells.

HCT8 (ATCC), THP-1 (ATCC) and MEFs [Atg5^−/−,  a kind gift from Noboru Mizushima, Tokyo Medical and Dental University, Japan (Kuma et al., 2004); and FIP200^−/−, a kind gift from Jun-Lin Guan, University of Cincinnati, USA (Gan et al., 2006)] were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin and non-essential amino acids at 37°C with 5% CO₂. Cos7 (ATCC), HeLa (ATCC), HEK293 ATG16-KO (Lystad et al., 2019) and HEK293 T (ATCC) cells were cultured in Dulbecco's modified Eagle’s medium (DMEM) (Sigma-Aldrich) supplemented with 10% FBS, 1% penicillin/streptomycin and non-essential amino acids at 37°C with 5% CO₂. Cells were routinely tested for mycoplasma contamination using a LookOut Mycoplasma PCR detection kit (Sigma-Aldrich). Cell lines were authenticated by ATCC, with no further authentication carried out after purchase.

Animal experiments were performed in accordance with the animal protocols that were approved by the Ethical Committee of Animal Experiments of Nanjing Agricultural University (permit SYXK [Su] 2017-0007).

The streptomycin-treated adult mouse model was used as described previously (Wang et al., 2017). Briefly, 5-week-old CD-1 mice were provided with drinking water containing 0.5% (w/v) streptomycin and 0.5% aspartame for 12 h before inoculation. For single colonization, ∼10⁸ cells of wild type (SmR, lacZ⁻) or ΔmakA mutant (SmR, lacZ⁺) were intragastrically inoculated into each mouse. Fecal pellets were collected at the indicated time points, resuspended in LB broth, serially diluted and spread on plates containing X-Gal. Colony-forming units were counted after 24 h.

EEA1 [3288, immunofluorescence (IF), 1:100], Rab7 (9367, IF, 1:100), Rab11 (5589, IF, 1:100), LC3B [2775, western blot (WB), 1:1000; IF, 1:200], LAMP1 (9091, IF, 1:200), ATG16L1 (8089, IF, 1:100), GM130 (12480, IF, 1:200), ATG12 (mouse specific) (2011, WB, 1:1000; IF, 1:100), ATG12 (human specific) (2010, WB, 1:1000; IF, 1:100), Cav1 (3267, IF, 1:400), galectin-3 (87985, IF, 1:400) and Na/K-ATPase α1 (23565, IF, 1:100) antibodies were purchased from Cell Signaling Technology. Anti-β-actin antibody (A2228, WB, 1:10,000) was purchased from Sigma-Aldrich. Anti-p62/SQSTM1 antibody (PM045, WB, 1:1000; IF, 1:200) was purchased from MBL international. Mono- and poly-ubiquitylated conjugates (FK2) antibody (BML-PW8810, IF, 1:100) was purchased from Enzo Life Sciences. Goat anti-rabbit-IgG conjugated to horseradish peroxidase (HRP) (31460, WB, 1:10,000) and goat anti-mouse-IgG conjugated to HRP (31430, WB, 1:10,000) antibodies were purchased from Thermo Fisher. Alexa Fluor 488-, 594- and 647-conjugated secondary antibodies for immunofluorescence were purchased from Jackson ImmunoResearch.

The EGFP-Rab5 construct was generated by PCR amplification of human Rab5a from cDNA, which was then subcloned into the pEGFP-C2 (Clontech) vector using XhoI/BamHI restriction sites. EGFP-Rab5 S34N construct was generated via site-directed mutagenesis. The EGFP-LC3 plasmid has been described previously (Laraia et al., 2019). Transfection of DNA constructs was performed using X-tremeGENE HP transfection reagent (Roche), according to the manufacturer's directions.

MakA expression and purification was carried out as previously described (Dongre et al., 2018). Alexa Fluor 568 labeling of MakA was performed using an Alexa Fluor 568 protein labeling kit (Thermo Fisher Scientific), according to the manufacturer's instruction.

Cell viability of HCT8 cells treated with MakA for 48 h was quantified using PrestoBlue (Thermo Fisher Scientific) or MTS (Promega) cell viability assays, according to the manufacturer's instruction. PrestoBlue fluorescence (Ex/Em, 560/590 nm) and MTS absorbance (490 nm) were measured on an Infinite M200 microplate reader (Tecan). Data from both assays were normalized to untreated cells and expressed as a percentage of the control.

HCT8 cells were treated with MakA (500 nM) for 48 h, and total RNA was extracted using an RNeasy Mini Kit (Qiagen). Library preparation and sequencing was completed by Novogene, with a paired-end protocol and read length of 150 nt (PE150), resulting in a total output of 20 million reads per sample. All reads outputs were checked for compliance with Illumina quality standards (Ewels et al., 2016; Wingett and Andrews, 2018). Raw data were cleaned up with Trimmomatic 0.36 (Bolger et al., 2014) to remove sequences originating from Illumina adaptors and low quality reads. Files were aligned with the human reference genome hg38 (Rhead et al., 2010) using TopHat/2.1.1 (Trapnell et al., 2009) with the default options. Once the alignment was completed, Samtools 1.8 (Li et al., 2009) was used to select and sort the reads that were aligned in proper pair. The reads counting per gene was performed with HTSeq 0.9.1 (Anders et al., 2015), using stranded mode and exon as a feature. Differential expression analysis was performed using edgeR (Robinson et al., 2010). Genes with adjusted P<0.05 and log₂ fold changes more than two were considered differentially expressed and were functionally characterized using the Database for annotation, visualization and integrated discovery (Huang et al., 2009a,b) and Gene set enrichment analysis (GSEA) (Mootha et al., 2003; Subramanian et al., 2005).

The total cholesterol content of HCT8 cells treated with MakA was determined using the Amplex Red Cholesterol Assay kit (Molecular Probes) as previously described (Eskelinen et al., 2004). Cells were washed with ice-cold PBS and lysed (50 mM Tris-HCl pH 8, 2 mM CaCl₂, 80 mM NaCl and 1% Triton X-100). For cholesterol quantification, 25 µl of lysate was mixed 1:1 with reaction buffer and then with an equivalent volume of Amplex Red working solution (300 µM Amplex Red reagent, 2 U/ml HRP, 2 U/ml cholesterol oxidase and 0.2 U/ml cholesterol esterase). Reactions were incubated for 30 min at 37°C, and fluorescence was measured using a Synergy H4 microplate reader (BioTek) (Ex/Em, 535/595 nm). Cholesterol content was normalized to total protein, as determined using a Bradford assay (Bio-Rad Protein Reagent Dye).

Free cholesterol was visualized using Filipin (Sigma-Aldrich). For staining, cells were fixed in 3% paraformaldehyde in PBS for 1 h at room temperature, washed three times with PBS containing 1.5 mg/ml glycine, and incubated with 0.05 mg/ml Filipin in PBS containing 10% FBS for 30 min at 37°C. When combined with antibody staining, cells were incubated with primary antibody immediately after Filipin staining, without any additional permeabilization. Primary antibody was diluted in 0.05 mg/ml Filipin in PBS containing 10% FBS.

Cos7 cells were plated on gridded 35 mm dishes (MatTek Corporation, P35G-1.5–14-CGRD), treated with 250 nM A568-MakA for 48 h in DMEM, and initially fixed with paraformaldehyde (4%) and glutaraldehyde (0.5%) for 1 h at room temperature. For confocal microscopy, the cells were imaged on a Leica SP8 inverted confocal system equipped with a HC PL APO 63×/1.40 oil immersion lens.

After confocal microscopy, the cells were fixed in 0.05% malachite green oxalate, 2.5% glutaraldehyde and 0.1 M sodium cacodylate buffer, and post-fixed with 0.8% K3FC(CN)₆ and 1% OsO₄ (14 min). Samples were further stained with 1% tannic acid and 1% uranyl acetate (EMS), dehydrated with a graded series of ethanol, infiltrated with resin (Spurrs, TAAB Laboratories) and then left in 100% resin for 1 h at room temperature and later polymerized overnight at 60°C. All sample preparation steps were performed with the PELCO Biowave pro+ (Ted Pella). After polymerization, blocks were trimmed down to the regions previously identified on the confocal microscope, based on the grid markings imprinted on the resin. Sections were cut on an Ultracut UCT ultramicrotome (Leica) and collected on formvar-coated slot grids. Samples were observed using a Thermo Scientific Talos L120C transmission electron microscope. Image overlay of immunofluorescence images and electron micrographs was performed manually using Adobe Photoshop.

For fixed cell immunofluorescence, cells were grown on poly-L-lysine-coated coverslips and fixed either in 3% paraformaldehyde in PBS for 10 min at room temperature or with ice-cold methanol for 15 min at −20°C. Cells were then washed three times with PBS containing 1.5 mg/ml glycine, permeabilized in 0.25% Triton X-100 in PBS for 5 min, and again washed three times with PBS (permeabilization step was skipped with methanol fixation or when immunofluorescence was preceded by Filipin staining). Cells were blocked with 5% donkey serum for 30 min followed by a 1-2 h incubation with primary antibody at room temperature. Cells were then washed three times with PBS and incubated with Alexa Fluor-conjugated secondary antibodies for 30 min at room temperature. Cells were again washed three times with PBS and mounted on slides using ProLong Diamond antifade mountant (Thermo Fisher Scientific). For live-cell imaging, cells were seeded on poly-L-lysine-coated eight-well cover glass-bottom chamber slides (Sarstedt) and incubated for 24 h. Imaging was performed in DMEM/RPMI without Phenol Red (Sigma-Aldrich).

Immunofluorescence imaging was performed on two microscope systems: (1) a Zeiss Cell Observer spinning disk confocal microscope (ANDOR iXon Ultra) equipped with a 63× immersion oil objective lens (Plan-Apochromat 1.40 Oil DIC M27) and a temperature-controlled hood maintained at 37°C and 5% CO₂; and (2) a Leica SP8 FALCON inverted confocal system equipped with a HC PL APO 63×/1.40 oil immersion lens and a temperature-controlled hood maintained at 37°C and 5% CO₂. DAPI and Filipin were excited using a 405 nm diode laser, and EGFP/Alexa488, mCherry/Alexa594 and Alexa647 fluorescence were excited using a tuned white light laser. Scanning was performed in line-by-line sequential mode.

Image processing was restricted to brightness/contrast adjustment using ImageJ/FIJI distribution (Schindelin et al., 2012; National Institutes of Health). Fluorescence intensity profiles were generated using the plot profile command in ImageJ. Live-cell particle tracking was performed using the spot tracking function (Chenouard et al., 2013) of the open-source bioimage processing software, Icy (de Chaumont et al., 2012).

Cells were rinsed with PBS, lysed in ice-cold lysis buffer [20 mM Tris-HCl (pH 8), 300 mM KCl, 10% Glycerol, 0.25% Nonidet P-40, 0.5 mM EDTA, 0.5 mM EGTA, 1 mM PMSF and 1× complete protease inhibitor (Roche)], passed six times through a 21G needle, and cleared by centrifugation (25 min at 18,213 g at 4°C). Protein concentrations were determined using a Bradford assay (Bio-Rad protein reagent) and lysates were normalized. Lysates were then mixed with 4× sample buffer and boiled for 10 min before separation by SDS-PAGE, and transferred to a nitrocellulose membrane (Bio-Rad) using a Trans-Blot Turbo transfer system (Bio-Rad). After a 1 h block in Tris-buffered saline with 0.1% Tween 20 (TBST) with 5% skimmed milk (at room temperature), membranes were incubated with primary antibody overnight at 4°C. Membranes were then washed with TBST and incubated for 1 h at room temperature, with the appropriate HRP-conjugated secondary antibody in blocking buffer. Protein detection was carried out using chemiluminescence (Bio-Rad) and imaged using a ChemiDoc Imaging System (Bio-Rad).

THP-1 cells were treated with MakA in a dose-dependent manner for 12 h, and challenged with LPS (1 µg/ml) from Escherichiacoli 0111:B4 (Sigma-Aldrich) for 4 h. Supernatants collected after treatment were filtered with a 0.2 µm Minisart sterile syringe filter (Sartorius Stedim Biotech), and cytokines IL-8 (Abcam, ab214030) and IL-1β (Abcam, ab214025) were measured using an ELISA, according to the manufacturer’s instruction. For mouse IL-1β measurements, heart blood was collected at 24 h post infection by cardiac puncture, and allowed to clot overnight at 4°C. Serum was collected by centrifugation at for 15 min at 4°C, and stored at −20°C. IL-1β was measured using an IL-1β ELISA kit (Jiangsu Meimian Industrial). Four independent experiments were performed.

Data are shown as mean±s.d. Statistical significance was determined by one/two-way ANOVA or by Student's t-tests (two-tailed, unpaired), as indicated in the corresponding figure legends, using GraphPad Prism v.7 (*P<0.05; **P<0.01; ***P<0.001; ns, not significant).

No statistical methods were used to predetermine sample size. A standard sample size of n≥3 was chosen to properly observe variance and ensure reproducibility. At least three biological replicates were carried out for all experiments, where possible. All attempts at replication were successful, with no data excluded.

We thank Noboru Mizushima for sharing the Atg5^−/− MEFs; Jun-Lin Guan for sharing the FIP200^−/− MEFs; and Nicholas Ktistakis for help with the distribution of the FIP200^−/− MEFs. The computations were performed on resources provided by the Swedish National Infrastructure for Computing through the Uppsala Multidisciplinary Center for Advanced Computational Science, under Project SNIC 2017-7-258. We acknowledge the Protein Expertise Platform at Umeå University for construct design and cloning; the Biochemical Imaging Center (BICU) at Umeå University and the National Microscopy Infrastructure (VR-RFI 2016-00968) for providing assistance with microscopy; and the facilities and technical assistance of the Umeå Core Facility Electron Microscopy (UCEM) at the Chemical Biological Centre (KBC), Umeå University.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.252015.reviewer-comments.pdf