To investigate whether ANXA7 participates in lysosome repair, we treated MCF7 breast carcinoma cells and HeLa cervix carcinoma cells with L-Leucyl-L-Leucine O-methyl ester (LLOMe), a lysosomotropic agent which enters cells via receptor-mediated endocytosis and is converted to a membranolytic compound specifically in the endolysosomal compartments (Thiele and Lipsky, 1990). Prior to LLOMe treatment (1 h), lysosomes of MCF7 or HeLa cells were loaded with superparamagnetic colloidal iron dextran (FeDEX) (16 h), enabling magnetic retention and subsequent elution and characterization of FeDEX-loaded lysosomes by immunoblotting (Figure 1A).

The FeDEX-based isolation procedure successfully enriched lysosomal markers, lysosome-associated membrane glycoprotein 2 (LAMP2) and cathepsin B (CTSB), whereas cellular markers for early endosomes (EEA1), the ER (PDI) and mitochondria (TOM20) were vastly depleted (Figures 1B, C, E; Supplementary Figure S1). Notably, LLOMe treatment depleted the lysosomal protease, CTSB, in the lysosomal fraction, suggestive of LLOMe-induced LMP and loss of lysosomal content. Accordingly, treatment with LLOMe reduced the extent to which the pore-forming detergent, digitonin, could release N-acetyl-β-D-glucosaminidase (NAG) from lysosomes in both whole-cell and lysosomal fractions observable as reduced NAG activity (Figures 1D, F). Collectively, the data demonstrate that the procedure successfully isolates lysosomes and treatment with LLOMe effectively induces LMP.

Prompted by our previous studies of ANXAs and their roles in plasma membrane repair (Lauritzen et al., 2015; Boye and Nylandsted, 2016; Boye et al., 2017; Sønder et al., 2019; Bendix et al., 2020; Simonsen et al., 2020; Moreno-Pescador et al., 2022), we wondered if similar repair mechanisms might occur on lysosomal membranes upon damage. To confirm lysosomal damage, we blotted against GAL3 as a sensitive marker of LMP due to its rapid translocation to leaky lysosomes in response to LLOMe treatment (Figure 1B) (Aits et al., 2015). Interestingly, while the lysosomal fraction from LLOMe-treated HeLa and MCF7 cells showed enriched GAL3 levels as expected, we also detected an enrichment of both ANXA7 and ALG-2 (Figures 1G, H; Supplementary Figure S1). Moreover, we detected an enrichment of ANXA1 and ANXA5 in the lysosomal fraction from LLOMe-treated HeLa cells, whereas ANXA2, ANXA4, and ANXA6 were absent in the lysosomal fractions from both cell lines (Figures 1G, H; Supplementary Figure S1). These findings indicate that, despite their conserved structure and functional similarities, ANXAs have distinct cellular functions that are cell type specific. Moreover, the data suggest that endogenous ANXA7 and ALG-2 are recruited to damaged lysosomes in both MCF7 and HeLa cells. Based on this and our prior demonstration of ANXA7’s necessity for ESCRT-III-mediated plasma membrane repair, we chose to focus on investigating the potential role of ANXA7 in lysosome damage and repair.

To further validate our data, we used MCF7 cells stably expressing GAL3 fused to enhanced green fluorescent protein (MCF7 EGFP-LGALS3) as a well-established and sensitive marker to identify and study lysosome damage and repair in live cells (Aits et al., 2015). Cells co-expressing ANXA7 fused to monomeric red fluorescent protein (ANXA7-mRFP) were exposed to LLOMe and imaged for 15 min by time-lapse microscopy. The cells showed strong GAL3 puncta formation upon LLOMe treatment, as expected and previously demonstrated (Aits et al., 2015). In addition, LLOMe treatment triggered pronounced ANXA7 puncta formation at vesicles, becoming positive for GAL3 (Figure 2E; Supplementary Movie S3), in agreement with our lysosomal fractionation data (Figures 1G, H; Supplementary Figure S1). Cell death assays were performed to ascertain that LLOMe did not impact the short-term cell viability in the applied time frame and concentration range, which was not the case (Supplementary Figures S2C–G).

From the FeDEX-based isolation procedure of lysosomes, we detected an enrichment of ANXA7 and ALG-2 in both the lysosomal fractions from LLOMe-treated MCF7 and HeLa cells (Figures 1G, H; Supplementary Figure S1). We have previously shown that ANXA7 and ALG-2 form a complex upon repair at the plasma membrane and that ANXA7 binding to free membrane edges is independent of ALG-2, whereas ANXA7 is required for ALG-2 membrane binding and positioning at the site of repair (Sønder et al., 2019). To validate whether ANXA7 translocation to damaged lysosomes is independent of ALG-2, we investigated the translocation kinetics of ANXA7-mRFP following LLOMe treatment in control and ALG-2 siRNA depleted MCF7 cells. The cells were co-transfected with EGFP-GAL3 to verify lysosomal damage after LLOMe treatment. The siRNA transfected MCF7 cells co-expressing EGFP-GAL3 and ANXA7-mRFP were exposed to LLOMe and imaged for 15 min by time-lapse confocal microscopy. LLOMe treatment induced accumulation of both GAL3 and ANXA7 puncta (Figures 2A, B; Supplementary Movies S1, S2). However, siRNA depletion of ALG-2 (>90%) did not significantly decrease ANXA7 puncta after LLOMe treatment as compared to control siRNA transfected cells, indicating that ANXA7 translocates to damaged lysosomes independently of ALG-2 (Figures 2A–D; Supplementary Figure S3A; Supplementary Movies S1, S2).

To address whether the observed damage response is specific to ANXA7, we also investigated the translocation kinetics of other ANXAs in response to LLOMe treatment. We chose to include ANXA2 and ANXA5 as the most auspicious candidates, as ANXA2 has previously been described to bind damaged endosomes and to facilitate lysosome repair (Scharf et al., 2012; Yim et al., 2022), while we detected an enrichment of ANXA5 in the lysosomal fraction from LLOMe-treated HeLa cells (Figure 1H). As before, LLOMe treatment induced accumulation of GAL3 puncta. However, neither ANXA2-mRFP nor ANXA5 fused to turbo-RFP (ANXA5-tRFP) formed any puncta in response LLOMe treatment (Figures 2F, G; Supplementary Movies S4, S5). Thus, there seems to be a specificity of ANXA7 towards damaged lysosomes in MCF7 cells compared to ANXA2 and ANXA5.

In light of our earlier findings demonstrating the requirement of ANXA7 for ESCRT-III-mediated plasma membrane repair (Sønder et al., 2019), we wondered if ANXA7 likewise promotes ESCRT-III-mediated lysosome repair. We approached this question using our previously generated MCF7-p95ErbB2 cells with CRISPR/Cas9 disrupted ANXA7 gene expression (CRISPR-A7KO) (Sønder et al., 2019), which was used to investigate the role of ANXA7 in plasma membrane repair. First, we treated MCF7-p95ErbB2 CRISPR-A7KO and control (CRISPR-Ctrl) cells with LLOMe or vehicle (diH₂O) (30 min) before fixation and immunostaining with anti-CHMP4B as a marker for the ESCRT-III complex. LLOMe treatment induced massive CHMP4B puncta formation co-occurring with LAMP2 puncta, strongly indicating lysosomal recruitment of the ESCRT-III complex as expected and previously reported (Radulovic et al., 2018; Skowyra et al., 2018; Niekamp et al., 2022; Yim et al., 2022). Surprisingly, we detected no significant difference in the number of CHMP4B puncta between CRISPR-A7KO and CRISPR-Ctrl cells (Figures 3A–C; Supplementary Figure S5A). These data show ANXA7 to be dispensable for CHMP4B buildup and ESCRT-III assembly at the lysosomal membrane in MCF7-p95ErbB2 cells, demonstrating that the damage response of ANXA7 at the plasma membrane and endolysosomal compartments differs mechanistically.

Next, we opted to investigate whether ANXA7 knockout impacted lysosome damage by treating MCF7-p95ErbB2 CRISPR-A7KO and CRISPR-Ctrl cells with LLOMe or vehicle (30 min) prior to fixation and immunostaining with anti-GAL3. LLOMe treatment triggered massive GAL3 puncta formation co-occurring with LAMP2 puncta (Figures 3D, E), highly indicative of lysosomal recruitment of GAL3 as expected and previously reported (Aits et al., 2015). Interestingly, CRISPR-A7KO cells displayed significantly more GAL3 puncta per cell than CRISPR-Ctrl cells after LLOMe treatment (Figures 3D, E), suggesting that ANXA7 knockout sensitizes MCF7-p95ErbB2 cells to LLOMe-induced LMP independently of ESCRT-III recruitment. Importantly, during steady-state conditions (vehicle treatment), no difference in GAL3 accumulation was detected between the cell lines (Figures 3D, E), indicating that ANXA7 specifically functions in response to provoked and acute lysosome damage. Following this discovery, we repeated the setup and included a siRNA targeting CHMP4B to evaluate the impact of disrupted ESCRT-III-mediated lysosome repair on GAL3 puncta formation (Supplementary Figures S4A–C, S5C, D). SiRNA depletion of CHMP4B (>90%) did not significantly alter GAL3 accumulation in the CRISPR-Ctrl cells as compared to control siRNA transfected cells (Supplementary Figures S4A–C, S5C, D), consistent with previous demonstrations by Skowyra et al. (2018). However, we were able to reproduce the increase in GAL3 accumulation upon ANXA7 knockout for both siRNA conditions (Ctrl and CHMP4B#2 siRNA) (Supplementary Figures S4A, B), reaffirming the detrimental impact that ANXA7 knockout has on the lysosomal damage response.

SiRNA depletion of ANXA7 (>90%) significantly enhanced GAL3 accumulation upon LLOMe treatment in MCF7 cells as compared to control siRNA transfected cells (Figures 3F–H; Supplementary Figure S5B), showing that ANXA7 depletion robustly sensitizes MCF7 cells and their derived cell lines to LLOMe-induced LMP. Furthermore, this demonstrates that the observed increase in GAL3 accumulation is not an outcome of single cell-derived knockout clones or heightened membrane dynamics and invasiveness of the MCF7-p95ErbB2 cell model (Martin et al., 2012; Rafn et al., 2012). Collectively, these data demonstrate that lack of ANXA7 negatively impacts lysosome repair in MCF7 cells in a manner mechanistically distinct from the ESCRT machinery.

To more directly assess the putative role of ANXA7 in lysosome repair, we performed the well-established Magic Red lysosome repair assay (Bright et al., 2016; Skowyra et al., 2018; Yim et al., 2022). In this assay, our CRISPR/Cas9 modified MCF7-p95ErbB2 cells were incubated with Magic Red, a chemical substrate that releases a photostable red fluorophore, cresyl violet, upon catalytic cleavage by lysosomal proteases. Loss of the lysosomal pH gradient upon LLOMe treatment results in reduced enzymatic activity, detectable as a loss of fluorescent signal, which is regained upon lysosome repair and restored pH gradient (Figure 4A), hence providing a quantifiable readout for repair kinetics. Once again, we used siRNA depletion of CHMP4B as a benchmark for impaired ESCRT-III-mediated lysosome repair. However, surprisingly we detected no difference in the reoccurrence of Magic Red puncta upon CHMP4B depletion in either CRISPR-A7KO or CRISPR-Ctrl cells despite efficient knockdown (Figures 4B–D; Supplementary Figures S6A–D). By contrast, knockout of ANXA7 significantly impaired the reoccurrence of Magic Red puncta independent of the siRNA condition (Ctrl or CHMP4B#2 siRNA) (Figures 4B, D). Consequently, the data suggest that ANXA7 is more crucial than the ESCRT-III complex for lysosome repair in MCF7-p95ErbB2 cells under the investigated conditions.

Recruitment and ESCRT-III-mediated lysosome repair have been observed to precede and function independently of lysophagy (Radulovic et al., 2018; Skowyra et al., 2018). To establish whether ANXA7-mediated lysosome repair also functions separately from lysophagy, we investigated the formation of LLOMe-induced ANXA7 puncta after inhibiting lysophagy using the generic autophagy inhibitor 3-methyladenine (3-MA) (Figure 5A). By suppressing the production of phosphatidylinositol-3-phosphate (PtdIns3P), 3-MA effectively hinders autophagosome formation (Seglen and Gordon, 1982; Blommaart et al., 1997; Høyer-Hansen et al., 2007).

To accomplish this, we used MCF7 cells stably expressing enhanced green fluorescent protein fused to LC3B (referred to as LC3) (MCF7 EGFP-LC3), a widely used marker of autophagosomes (Høyer-Hansen et al., 2007). MCF7 EGFP-LC3 cells co-expressing ANXA7-mRFP and pretreated with 3-MA or vehicle (30 min) were exposed to LLOMe and imaged for 15 min by time-lapse confocal microscopy. Treatment with LLOMe induced profound LC3 lipidation, surpassing the effect observed with the potent autophagy inducer rapamycin (Noda and Ohsumi, 1998), clearly indicating extensive induction of autophagy within the treatment window (Figure 5B). Pretreatment with 3-MA robustly suppressed LC3 lipidation in response to both LLOMe and rapamycin, demonstrating the efficacy of 3-MA in inhibiting autophagy (Figure 5B; Supplementary Figure S7).

As previously observed, cells showed strong ANXA7 puncta formation upon LLOMe treatment, and inhibiting autophagy did not significantly affect the average number of ANXA7-positive cells (Figures 5C–E), showing that ANXA7-mediated lysosome repair, like the ESCRT machinery, operates separately from lysophagy. Further supporting this conclusion, ANXA7 and LC3 puncta did not appear to co-occur (Figures 5D, E).

In the newly identified PITT pathway, rapid generation of PtdIns4P by phosphatidylinositol-4 kinase type 2α (PI4K2A) at damaged lysosomes facilitates the recruitment of oxysterol-binding protein-related proteins (ORPs), which in turn interact with ER-anchored vesicle-associated membrane protein-associated proteins (VAPs) to orchestrate extensive MCSs between the ER and lysosomes (Radulovic et al., 2022; Tan and Finkel, 2022). Robust ER-to-lysosome transfer of PS and cholesterol through established MCSs promotes subsequent lysosome repair. Interestingly, different members of the ANXA family are becoming recognized for their implication in MCS formation (Enrich et al., 2021). The putative function of ANXAs in MCS formation is hypothesized to originate from FFAT-like motifs (two phenylalanines in an acidic tract) within the ANXAs, which amongst others, would permit interactions with ER-localized VAPs. Phospho-FFAT motifs have been identified in ANXA5 and ANXA8 (Di Mattia et al., 2020; Enrich et al., 2021), while homologies to FFAT motifs have been identified in ANXA1, ANXA6, and ANXA11 (Rentero et al., 2018; Enrich et al., 2021). Accordingly, ANXA6 has been implicated in MCS formation and ER-to-late endosome/lysosome transfer of cholesterol (Meneses-Salas et al., 2020). However, given the extensive interaction between ANXAs and ANXA binding partners, their implication in MCS formation should not be restricted solely to family members containing phospho- or FFAT-like motifs. Supportive hereof, ANXA7 and ANXA11 were identified as binding partners when using ANXA6 as a bait (Enrich et al., 2021).

In light of this, we wondered whether ANXA7 might be part of orchestrating the formation of ER-lysosomal MCSs to facilitate PITT-mediated lysosome repair. As for the PITT pathway, we demonstrate that ANXA7 mediates lysosome repair independently of ESCRT-III (Figures 3A, B), highlighting a novel repair role of ANXA7 differing from its known function in ESCRT-III-mediated plasma membrane repair (Sønder et al., 2019). We approached the question by performing quantitative mass spectrometry (MS)-based shotgun lipidomic analysis of FeDEX-purified lysosomes from MCF7-p95ErbB2 CRISPR-A7KO and CRISPR-Ctrl cells treated with LLOMe or vehicle (diH₂O) (1 h). We identified around 300 or more different lipid species belonging to 27 lipid classes distributed in the categories glycerolipids (GLs), glycerophospholipids (GPLs), sphingolipids (SLs), and sterol lipids (Figure 6B; Supplementary Figure S8A). We absolutely quantified them and expressed their quantities in molar percentages (mol%) of all identified lipid species to determine the stoichiometry between lipids in the samples.

The FeDEX-based isolation procedure once again successfully enriched lysosomal markers, LAMP2 and CTSB, whereas cellular markers for early endosomes (EEA1), plasma membrane (ATP1A1), ER (PDI) and mitochondria (TOM20) were vastly depleted (Figure 6A; Supplementary Figure S8D). Moreover, the lysosomal fractions obtained from vehicle-treated cells were abundant in cholesterol, sphingomyelin, and the endolysosome-specific lipid bis (monoacylglycero) phosphate (BMP) (Figures 6F,G; Supplementary Figure S8C) (Möbius et al., 2003; Tharkeshwar et al., 2017), in agreement with previously reported lipid profiles of lysosomes obtained from various mammalian cells (van Meer et al., 2008; Bilgin et al., 2017; Tharkeshwar et al., 2017; Nielsen I. et al., 2020; Scrima et al., 2022; Stahl-Meyer et al., 2022).

LLOMe treatment did not significantly alter the mol% values of most lipid classes in the whole-cell fraction in either of the two cell lines (Figure 6C; Supplementary Figure S8C); exceptions include BMP and certain lysoglycerophospholipid (lysoGPL) classes, which decreased (BMP) or increased (lysoGPL) significantly upon LLOMe treatment in line with previously reported effects of other LMP-inducing drugs (Nielsen I. et al., 2020; Fogde et al., 2022). In contrast, LLOMe treatment significantly altered several lipid classes in the lysosomal fractions (Figure 6C). In agreement with damaged lysosomes initiating the PITT pathway by forming MCSs to the PC-rich ER and subsequently transferring ER-synthesized PS (Radulovic et al., 2022), the LLOMe treatment increased PC and PS in the obtained lysosomal fractions (Figures 6C–E). Furthermore, even though not quantifiable due to the lack of a standard in our lipidomics pipeline, we detected ions matching multiple PtdIns4P species almost exclusively after LLOMe treatment, particularly prominently in the lysosomal fractions (Figure 6I; Supplementary Figure S8B). We observed a non-significant reduction of cholesterol upon LLOMe treatment in the lysosomal fraction (Figure 6F), even though the ER-to-lysosome transfer of cholesterol was previously proposed as a crucial event in the PITT pathway (Radulovic et al., 2022; Tan and Finkel, 2022). The expected increase in lysosomal cholesterol was possibly masked by the large increase in the highly abundant PC (Figure 6D). Notably, LLOMe treatment also triggered a significant reduction in sphingomyelin and an increase in ceramide in the isolated lysosomes (Figures 6C, G, H), indicating that sphingomyelin was hydrolyzed at the surface of damaged lysosomes to form ceramide, as previously proposed as a mechanism of lysosome repair (Niekamp et al., 2022). All the above-described lipid changes induced by the LLOMe treatment occurred to a similar extent in lysosomes isolated from MCF7-p95ErbB2 CRISPR-A7KO and CRISPR-Ctrl cells (Figures 6C–I; Supplementary Figures S8A–C). In summary, our lipidomics data strongly suggest that MCF7-p95ErbB2 cells can promote lysosome repair through the newly identified PITT pathway (Radulovic et al., 2022; Tan and Finkel, 2022), while also revealing that ANXA7 is expendable for the formation of ER-lysosomal MCSs and subsequent lipid transfer.

Human breast adenocarcinoma MCF7 and human cervix adenocarcinoma HeLa cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium containing GlutaMAX (Termo Fisher Scientific, 61870) and supplemented with 6% (v/v) heat-inactivated fetal bovine serum (FBS) (Termo Fisher Scientific, 10270-106) and 0.25% (v/v) penicillin-streptomycin (Life Technologies, 11811-031). Cells were kept at 37°C in a humidified atmosphere with 5% CO₂ as a sub-confluent monolayer and were routinely tested for the presence of mycoplasma (Minerva Biolabs Inc., Venor®GeM Classic). The MCF7 S1 clone and its derivatives expressing EGFP-LC3B or EGFP-LGALS3 have been previously described (Jäättelä et al., 1995; Høyer-Hansen et al., 2007; Aits et al., 2015). MCF7-p95ErbB2 cells are single cell clones of the MCF7 S8 cell line stably expressing a truncated version of ErbB2 (p95ErbB2) under the transcriptional control of a tetracycline-controlled transactivator (Egeblad et al., 2001). These clones were further modified using CRISPR/Cas9 technology to generate an ANXA7 knockout (CRISPR-A7KO) and control (CRISPR-Ctrl) cell line (Sønder et al., 2019). MCF7-p95ErbB2 cells were cultivated in RPMI 1640 medium supplemented with 1 μg/mL tetracycline (Sigma-Aldrich, T7660), G418 Sulfate (Life Technologies, 11811-031) and 1 μg/mL puromycin (Sigma-Aldrich, P7255) in addition to FBS and penicillin-streptomycin. All experiments were performed 3–5 passages after p95ErbB2 induction by removing tetracycline from the cell culture medium.

Live-cell imaging was performed in cell-imaging medium (CIM) consisting of RPMI 1640 medium without phenol red (Termo Fisher Scientific, 32404014) supplemented with 2 mM GlutaMAX (Thermo Fisher Scientific, 35050038), 6% (v/v) FBS, 0.25% (v/v) penicillin-streptomycin and 25 mM HEPES (Sigma-Aldrich, H3375).

FeDEX-mediated lysosomal isolation was performed in Dulbecco’s modified Eagle’s medium (DMEM) containing GlutaMAX (Termo Fisher Scientific, 31966021) supplemented with 6% (v/v) FBS and 0.25% (v/v) penicillin-streptomycin to reduce FeDEX precipitation.

LMP was induced using LLOMe (Santa Cruz Biotechnology, SC-285992) dissolved in autoclaved deionized water (diH₂O). The applied concentration and incubation period are specified in the figure legends of the individual experiments. Autophagy was suppressed by exposing cells to 10 mM 3-MA (Sigma-Aldrich, M9281) dissolved in CIM for 30 min, whereas a 2 h treatment with 200 nM rapamycin (Sigma-Aldrich, 553210) dissolved in DMSO served as a positive control of autophagy induction.

Lysosomes were isolated using FeDEX particles (40 kDa) (Bilgin et al., 2017). FeDEX was prepared in solution as previously described (Bilgin et al., 2017) and no more than a week before experiments. Lysosomes were loaded with FeDEX using a Pulse-Chase approach, where cells were incubated overnight (ON) in cell culture medium containing 0.66% (v/v) FeDEX solution followed by a 3-h chase in clean medium. Vehicle or 2 mM LLOMe was added to the cell culture medium during the last hour of the chase period. Cells were harvested by scraping, spun down at 500 × g for 2 min at 4°C and washed in ice-cold Dulbecco’s phosphate-buffered saline (DPBS) (Thermo Fisher Scientific, 14190094). Cells were subsequently lysed mechanically in a hypotonic SCA buffer (10 mM HEPES, 210 mM mannitol, 70 mM sucrose, pH 7.5) supplemented with 1 M DTT, ×1 Complete mini EDTA-free protease inhibitor cocktail (Roche, 4693124001) and 1 µL Benzonase^® Nuclease (≥250 units/µL) (Sigma-Aldrich, E1014) (SCA+++ buffer) using a 10 mL syringe and G25 needle (HeLa: ×5 strokes, MCF7: ×15 strokes). The crude cell lysate was spun down at 800 × g for 10 min at 4°C, and the supernatant was transferred to a new Eppendorf tube and spun down at 800 × g for 10 min at 4°C. The cleared supernatant was loaded onto an equilibrated MS column (Miltenyi Biotec, 130-042-201) mounted on a magnetic stand (Miltenyi Biotec, 130-042-109) placed at 4°C, and the column was washed repeatedly with SCA buffer before eluting the lysosomal fraction using a plunger.

Whole-cell and lysosomal fractions were diluted 1:20 in SCA buffer, treated with vehicle (diH₂O) or 250 μg/mL digitonin (Sigma-Aldrich, D141), and briefly vortexed before being aliquoted to a non-transparent, flat-bottomed 96-well plate (Greiner Bio-One, 655090). Next, 0.3 mg/mL 4-Methylumbelliferyl N-acetyl-β-D-glucosaminide (4-MUF-NAG) (Sigma-Aldrich, M2133) dissolved in 0.2 M citrate buffer (pH 4.5) was added, and the release of fluorogenic 4-Methylumbelliferyl was measured using a Varioskan Flash plate reader (Thermo Fisher Scientific) and SkanIt^® Research Edition Software for Varioskan^® Flash (v2.4.5). The plate was allowed to acclimate inside the reader for 4 min before measuring the fluorescent signal at 448 nm every 45 s for 20 min at 37°C.

Whole-cell and cellular fractions were lysed in Laemmli sample buffer (125 mM Tris (pH 6.7), 20% (v/v) glycerol, 140 mM SDS, 0.3 µM bromophenol blue) supplemented with 0.1 M DTT, cOmplete™, Mini Protease Inhibitor Cocktail (Roche, 4693124001) and PhosSTOP™ (Roche, 4906837001). Lysates were boiled, and proteins were separated by SDS-PAGE using a 4%–15% TGX™ Precast Protein Gel (BioRad, 4561086/5671085) and blotted onto a nitrocellulose membrane (BioRad, 1704158/59) using a Trans-Blot TurboTM transfer system. Novex Sharp Protein Standard (Thermo Fisher Scientific LC5800) was used to evaluate the molecular weights of proteins from the gels. Following blocking in PBS containing 0.1% (v/v) Tween-20 (PBST) and 5% (w/v) BSA, membranes were cut in appropriate sizes for detection of proteins of interest and incubated ON at 4°C with primary antibodies in 0.1% (v/v) NaN₃/5% (w/v) BSA/PBST (Supplementary Table S1). Next, membranes were incubated for 30 min with matching secondary HRP-conjugated antibodies (Supplementary Table S1) and the immunoreactivity was detected using a luminescent image reader (Fujifilm, LAS-4000) following incubation with Clarity™ Western enhanced chemiluminescent substrate (BioRad, 1705061).

Expression plasmids with ANXA2 with a monomeric RFP C-terminal tag and ANXA5 with a turbo RFP C-terminal tag have been constructed by subcloning cDNA sequences of human ANXA2 (OriGene RG205081) or ANXA5 (OriGene RG205619) into the respective mammalian vectors, pCMV6-AC-mRFP (OriGene PS100041) or pCMV6-AC-tRFP (OriGene PS100034) using the restriction enzymes AsiSI and MluI. ANXA7-mRFP was constructed previously using the same approach (Sønder et al., 2019). EGFP-LGALS3 was constructed by subcloning human LGALS3 (isoform 1) with an N-terminal enhanced GFP-tag into the lentiviral CDH-EF1-MCS-IRES-Neo vector as previously described (Aits et al., 2015). Cells were transiently transfected with the described plasmids using Lipofectamine LTX (Invitrogen, 15338100) according to the manufacturer’s instructions, omitting the Plus Reagent, and used for live-cell imaging experiments the day following transfection.

Reverse siRNA transfection was performed using Oligofectamine Transfection Reagent (Invitrogen, 12252011) according to the manufacturer’s instructions. The cell culturing medium was replaced following 24 h, and experiments were performed 72 h after transfection. AllStar Negative Ctrl siRNA (Qiagen, 1027281), ALG-2 [5′-CAC GAC AUC CUC AUU CGA A (dT) (dT)-3′], ANXA7 [5′-GCU UAU UCU AGG AUG GCU A (dT) (dT)-3′], CHMP4B#1 [5′-CGC UAU GCC GCC AAG GCC A (dT) (dT)-3′] and CHMP4B#2 siRNA [5′-CAA GAA CUU GCA GAG GAG A (dT) (dT)-3′] were used at 25 nM.

Cells were grown on 170 μm glass coverslips and fixed in 4% (v/v) paraformaldehyde (Ampliqon, 43226.1000) for 15 min. Fixed samples were quenched in 50 mM NH₄Cl for 10 min to reduce the background signal before being permeabilized in ice-cold methanol for 10 min at −20°C. Samples were subsequently blocked for at least 1 h in buffer 1 [1% (w/v) BSA, 0.3% (v/v) Triton X-100, DPBS] containing 5% (v/v) goat serum and incubated for 2 h with primary antibodies in buffer 1. Primary antibodies raised against human CHMP4B (Proteintech, 13683-1-AP, RRID:AB_2877971, 1:200), GAL3 (Sigma-Aldrich, MABT51, RRID:AB_10806498, 1:300) and LAMP2 (Developmental Studies Hybridoma Bank, H4B4, RRID:AB_528129, 1.2 μg/mL) were used. Following incubation with primary antibodies, samples were incubated for 1 h with Goat anti-Rat IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ Plus 488 (Invitrogen, A48262, RRID:AB_2896330, 1:1,000) and Donkey anti-Mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 568 (Invitrogen, A10037, RRID:AB_2534013, 1:1,000) in buffer 2 [0.25% (w/v) BSA, 0.1% Triton X-100 (v/v), DPBS]. Nuclei were stained with 2.5 μg/mL Hoechst 33342 (Sigma-Aldrich, #B2261) for 10 min in buffer 3 [0.05% (v/v) Tween-20, DPBS]. Coverslips were mounted on microscope slides using ProLong™ Gold Antifade Mountant (Invitrogen, P36934).

Images were captured using a LSM 700 confocal microscope (Zeiss) equipped with a Zeiss Plan-Apochromat ×63/1.40 Oil DIC M27 objective and 405 nm, 488 nm, and 555 nm lasers and Zeiss imaging software (v7). CHMP4B and GAL3 puncta were quantified by detecting puncta objects based on fluorescent intensity and pixel size within a defined cell using the CellProfiler analysis software (v4.2.1). Manders’ overlap coefficient was calculated using the JACoP plugin with manual thresholding in the ImageJ software (v2.9.0) (Bolte and Cordelières, 2006; Schneider et al., 2012).

Cells grown in MatTek imaging-culture dish were kept at 37°C in CIM. Live-cell images were captured using an inverted microscope Eclipse Ti-E (Nikon) paired with an UltraVIEW VOX Spinning Disk [PerkinElmer and Volocity software (v6.3.1)] or a Crest X-Light V3 Spinning Disk [CrestOptics and NIS-Elements Advanced Research software (v5.42.03)]. The Eclipse Ti-E was equipped with a Nikon CFI Apochromat ×40 WI λS, NA 1.25, WD 0.18 objective or a Zeiss C Plan-Apochromat ×63/1.2 Water Korr objective and 488 nm and 555 nm lasers. MCF7 EGFP-LGALS3 cells transiently transfected with ANXA2-mRFP or ANXA5-tRFP were imaged for 12 min at six frames per minute (fpm), whereas ANXA7-mRFP transfected cells were imaged for 0–3 min at one fpm, 4–6 min at two fpm, 7–10 min at six fpm and 11–15 min at one fpm to reduce photobleaching of the ANXA7-mRFP construct. MCF7 EGFP-LC3 cells transiently transfected with ANXA7-mRFP and siRNA transfected MCF7 cells transiently transfected with EGFP-LGALS3 and ANXA7-mRFP were imaged for 0–3 min at one fpm, 4–9 min at two fpm, 10–15 min at six fpm and 16–21 min at two fpm. LLOMe was added to a final concentration of 5 mM following the initial frame in all time-lapse videos.

MCF7-p95ErbB2 CRISPR-A7KO and CRISPR-Ctrl cells were transfected with AllStar Negative Ctrl or CHMP4B#2 siRNA according to the description in “Plasmids and siRNA transfections” and reseeded in a SCREENSTAR 96-well microplate (Greiner Bio-One, 655866) 24 h prior to the assay. Cells were incubated with 0.6 µM Sir-tubulin (Spirochrome, SC002) and 0.75 µM Nuclear Violet™ (AAT Bioquest, LCS1) for 1 h at 37°C before initiating the assay. Both dyes were present in the CIM throughout the entire assay. Magic Red (Immunochemistry Technology, 937) was reconstituted in DMSO according to the manufacturer’s instructions and stored at −80°C in single-use aliquots. For each experiment, aliquots were thawed immediately before use and diluted 1:2,500 in CIM, after which the cells were incubated with Magic Red solution for 3 min at 37°C before initiating the assay. Images were captured every minute at two sites in each well using an ImageXpress Micro Confocal High-Content Imaging System equipped with a CFI Apochromat LWD Lambda S 40XC/1.15 WI objective and 405/20 nm, 555 nm and 638/17 nm lasers and MetaXpress High Content Image Acquisition & Analysis Software (v6.6). Two frames were captured, and the acquisition paused. LLOMe was added to a final concentration of 1.5 mM, and imaging was resumed for 3 min. Following 3 min, the acquisition was paused, and cells were washed in clean CIM and administered CIM containing all three dyes before resuming imaging for another 9 min. Cells were kept at 37°C and supplied with 5% CO₂ throughout all incubation periods and the entire assay.

Magic Red puncta were quantified by detecting puncta objects based on fluorescent intensity and pixel size within a defined cell. Top-hat filtering was included to reduce artifacts from fluorescent accumulation at object borders. The analysis output was the number of Magic Red puncta per cell at each timepoint normalized to the number of Magic Red puncta per cell in the first frame of the treatment condition.

Cells were treated with LLOMe at different concentrations (0–5 mM) and time intervals (0–120 min). To assess the dead/live ratio, cells were stained for 5 min at 37°C with a propidium iodide (Sigma-Aldrich, P4864, 0.25 μg/mL) and Hoechst 33342 (Sigma-Aldrich, 861405, 6.25 μg/mL) dye mix. The fluorescent intensities were measured using a Celigo^® Imaging Cytometer (Brooks Life Science Systems) and associated software (v4.1.3.0). The cytometer was equipped with a proprietary F-theta lens and 377/50 nm and 531/40 nm lasers.

A modified Bligh and Dyer protocol was used for lipid extraction (Bligh and Dyer, 1959). Briefly, purified samples were resuspended in 200 µL 155 mM ammonium bicarbonate. 1 mL chloroform/methanol 2:1 (v/v) and 12.5 µL internal lipid standard mix was added (see Supplementary Table S2 for a list of all internal lipid standards used in this study). Samples were shaken at 2,000 rpm for 15 min at 4°C and centrifuged at 2,000 × g for 2 min at 4°C after which the upper phase was removed and 100 µL methanol and 50 µL 155 mM ammonium bicarbonate was added. Samples were shaken at 2,000 rpm for 1 min at 4°C and centrifuged at 2,000 × g for 2 min at 4°C. The lower phase was transferred to a new tube and solvents were evaporated for 75 min in a vacuum centrifuge. Samples were resuspended in 100 µL chloroform/methanol 1:2 (v/v), shaken at 2,000 rpm for 5 min at 4°C and spun down at 15,000 × g for 5 min at 4°C. Samples were loaded in positive (13.3 mM ammonium acetate in isopropanol) or negative [0.2% methyladenine (v/v) in 100% methanol] solvent. Samples were directly infused and analyzed in a positive and negative mode using the Orbitrap Fusion Tribrid (Thermo Fisher Scientific, San Jose, CA, United States) coupled to a TriVersa NanoMate (Advion Biosciences, NY, United States). The mass spectra were analyzed using a python-based software called LipidXplorer (Herzog et al., 2012). Using an in-house built R-based software, LipidQ (https://github.com/ELELAB/lipidQ), the absolute quantities were analyzed. The intensity values of the lipids were compared to their internal lipid standard to determine the absolute molar quantities of the lipids (see Supplementary Table S2 for a list of the internal lipid standards used in this study).

Lipid nomenclature was described previously (Nielsen I. O. et al., 2020). R statistical software [R Core Team (2021)] version 1.4.1717 was used to analyze the data. Lipids with a median quantity of all replicates below 0.0001 mol% were excluded from the analysis. To statistically analyze every lipid species or class, the “limma” package (Ritchie et al., 2015) was used. A linear model was fitted based on the quantity of the lipids. Log2-transformed fold change values and associated p-values were calculated. The Benjamini-Hochberg method was used to account for the false discovery rate (FDR) by correcting for multiple testing and adjusted p-values (q-values) were calculated. Lipids were considered statistically significantly changed if the q-values were below 0.05.

Statistical analyses were conducted using an unpaired two-tailed t-test with Welch’s correction (α = 0.05) in the GraphPad Prism software (v9.0.0). All analyses were performed on the mean values from at least three independent experiments, except for the Magic Red assay. For the Magic Red assay, we quantified the area under the curve (AUC) from the last time point of LLOMe treatment until the endpoint, and the statistical analysis was based on the mean AUC from four independent experiments. Outliers were identified using the robust regression and outlier removal (ROUT) method (Q = 1%) for the number of CHMP4B and GAL3 puncta per cell in Figures 3A, B, D–G. Significance levels are denoted as follows: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001.