L-leucyl-L-leucine methyl ester does not release cysteine cathepsins to the cytosol but inactivates them in transiently permeabilized lysosomes Free

Late endosomes and lysosomes, here collectively termed lysosomes, represent the end-point of the endocytic pathway and are primarily involved in degradation of material endocytosed from outside the cell or sequestered from the cytoplasm by autophagy (Huotari and Helenius, 2011). For their degradative function, lysosomes contain over 60 different hydrolases and maintain an acidic pH through the V-ATPase proton pump localized in the lysosomal membrane (Schröder et al., 2010).

Over 40 years ago, it was shown that membrane-permeable α-L-monoamino acid esters cause loss of latency of lysosomal enzyme activity when added to isolated lysosomes; this was explained by lysosomal membrane permeabilization (LMP) due to increased intra-lysosomal osmotic pressure. This effect was apparent with concentrations of α-L-mono-amino acid esters between 0.5 and 20 mM, and occurred within 10–15 min of incubation with lysosomes (Goldman and Kaplan, 1973). Some α-L-mono-amino acid esters were also toxic to cells, in particular the L-leucine methyl ester. Further studies showed that a leucine dipeptide, L-leucyl-L-leucine methyl ester (LLOMe) was even more cytotoxic (Thiele and Lipsky, 1985). LLOMe induces apoptosis in CD8+ T cells and natural killer cells at 100 μM, in myeloid cells at 500 μM and in most non-bone-marrow-derived cell lines at above 1 mM (Thiele and Lipsky, 1990, 1992a; Droga-Mazovec et al., 2008). Sensitivity of cells to LLOMe-induced apoptosis correlates with the cellular content of the cysteine protease cathepsin C (catC; also known as CTSC or dipeptidyl peptidase I) and apoptosis could be reduced by inhibiting catC with the specific catC inhibitor Gly-Phe-diazomethylketone (GF-DMK) (Thiele and Lipsky, 1990, 1992b). CatC is a lysosomal hydrolase that normally acts as an amino-dipeptidase. However, with increasing pH catC favors the transfer of a dipeptidyl residue to an acceptor other than water, such as an unprotonated amino group, and can therefore exert considerable dipeptidyl transferase activity, thereby assembling polypeptides (McDonald et al., 1969; McGuire et al., 1992). Cytotoxicity of LLOMe was shown to depend on catC-mediated polymerization of LLOMe into condensation products with the structure (LL)_n≥3OMe, whose membranolytic activity in red blood cells is strongly increased compared to the that of the dipeptide (Thiele and Lipsky, 1990).

Later studies increasingly suggested that apoptosis induced by LMP-inducing agents, as well as oxidizing agents or TNF is associated with release of lysosomal hydrolases into the cytosol and can be reduced by inhibitors of lysosomal cathepsins (Kågedal et al., 2001; Roberg et al., 1999; Hellquist et al., 1997; Foghsgaard et al., 2001; Guicciardi and Gores, 2000). However, it was in particular the studies using LLOMe that led to the firmly entrenched model postulating a cascade of signaling events involved in the so-called ‘lysosomal apoptotic pathway’. In this model, cysteine cathepsins with endopeptidase activity, predominantly cathepsin B (catB, also known as CTSB) and cathepsin L (catL, also known as CTSL), are released from LLOMe-permeabilized lysosomes to the cytosol. There, they were proposed to initiate apoptotic signaling by cleaving the pro-apoptotic regulator protein BH3-interacting domain death agonist (BID), and by degrading the anti-apoptotic BCL-2 homologs and the caspase inhibitor XIAP (Cirman et al., 2004; Droga-Mazovec et al., 2008; Stoka et al., 2001).

Reagents that can permeabilize lysosomes have attracted considerable interest due to their therapeutic potential, such as killing cancer cells (Repnik et al., 2012; Kirkegaard and Jaattela, 2009). However, for this goal, it is necessary to better understand the molecular mechanisms of their action on cells and on lysosomes. With this in mind, we performed a detailed study of the effects of LLOMe in HeLa cells, which were used in the original studies postulating a concept of cysteine cathepsin-mediated lysosomal apoptotic pathway. Surprisingly, we found no dose-dependent correlation between induction of cell death and LMP or release of internalized lysosomal markers into the cytosol. Moreover, we show that, in addition to targeting the lysosomal membrane, LLOMe also destabilizes soluble lysosomal proteins, especially cysteine cathepsins, leading to their inactivation and degradation. Our results refute the model whereby LLOMe induces release of cysteine cathepsins from lysosomes into the cytosol to initiate apoptosis. Moreover, we show that LLOMe-induced apoptosis is independent of catB and catL activities. We propose a new model for the action of LLOMe on lysosomes whereby LLOMe permeabilizes the lysosomal membrane to protons and small molecular weight compounds, but lysosomal enzymes remain in lysosomes where they are inactivated.

To analyze a dose-dependent effect of LLOMe on viability, HeLa cells were treated continuously with LLOMe in the concentration range of 1–7 mM for 18 h. The levels of attached cells (presumed to be healthy) were quantified by Crystal Violet staining (Fig. 1A). Loss of plasma membrane integrity was evaluated by release of the cytosolic enzyme lactate dehydrogenase (LDH) into the medium (Fig. 1B). The results from both assays showed that LLOMe induced cell death but only at concentrations above 4 mM. In previous studies, BID-cleavage and extensive apoptosis in HeLa cells were observed at 2 mM (Droga-Mazovec et al., 2008) and even as low as 0.4 mM LLOMe (Cirman et al., 2004).

Loss of adherence could be reduced if cells were pre-treated for 2 h with the cysteine cathepsin inhibitor E64d or with the caspase inhibitor Z-VAD(OMe)-FMK, both at 20 μM (Fig. 1A). Inhibition of caspases, which are proteases essential for apoptosis (Kumar, 2007), completely prevented the release of LDH even at the highest LLOMe concentration (7 mM) (Fig. 1B). Therefore, the integrity of the plasma membrane in cells pre-treated with the caspase inhibitor was preserved even if they detached. This is evidence that these high concentrations of LLOMe induce apoptosis, and that when release of LDH does occur in LLOMe-treated cells it is likely due to secondary necrosis, which follows apoptosis. In contrast, there was little effect of E64d on LDH release (Fig. 1B), suggesting that inhibition of cysteine cathepsins in LLOMe-treated cells cannot prevent the induction of apoptosis. These findings were confirmed by observation of cells in transmitted light, which showed that detached cells in the presence of the caspase inhibitor were less damaged than detached cells in the presence of E64d (Fig. S1A).

To further confirm that LLOMe induces apoptosis, we monitored caspase-3-like activity by measuring cleavage of the fluorogenic substrate Ac-DEVD-AMC (Fig. 1C). At 5 mM LLOMe, considerable caspase-3-like activation was observed as early as 3 h after treatment and reached maximum activity at 10 h after treatment (Fig. 1Ci). Caspase-3-like activation was prevented by pre-treatment of cells with Z-VAD(OMe)-FMK. In contrast, caspase-3-like activity was ∼80% reduced, but was not prevented by pre-treatment with E64d (Fig. 1Cii). At increasing concentrations of LLOMe, caspase-3-like activity was detectable only above 4 mM and increased up to 7 mM (Fig. 1Ciii).

Next, we stained cells with Acridine Orange, which accumulates in compartments with low pH, to determine the concentration of LLOMe that causes lysosomal membranes to lose the proton gradient (Repnik et al., 2016c). As early as 10 min after treatment, the proton gradient of acidic vesicles was lost at concentrations as low as 0.25 mM LLOMe (Fig. 1D), well below the apoptotic threshold of 4 mM LLOMe. Moreover, loss of retention of Acridine Orange in LLOMe-treated cells was similar in magnitude to the level seen when cells were treated with the proton pump V-ATPase inhibitor bafilomycin A1 (Fig. 1D). This argues that all acidic lysosomes were permeabilized by LLOMe and therefore that LMP was complete.

Because protons appear to be released from lysosomes after LLOMe treatment, we next asked whether small molecules are also released, and if so, what is the upper-size limit for molecules that can exit lysosomes. For this, cells were incubated with fluorescent markers of different molecular masses [given as the relative molecular mass or molecular weight (MW)] in order to be internalized by endocytosis and reach lysosomes. We then monitored the retention or release of these molecules by lysosomes following treatment with LLOMe (Repnik et al., 2016b). In control cells the markers gave a punctate pattern, indicating lysosomal localization (Fig. 1E). After treatment of cells with 1 or 5 mM LLOMe for 30 min, puncta representing sulforhodamine B (559 MW) or 44,000 (4.4K) MW dextran conjugated to tetramethylrhodamine (TRITC) largely disappeared. This argues that the lysosomal membrane was sufficiently destabilized to allow the release of small macromolecules from lysosomes. In contrast, 10,000 (10K) MW dextran conjugated to Alexa Fluor 546 or 70,000 (70K) MW dextran conjugated to Texas Red remained localized in lysosomes after both 1 and 5 mM LLOMe treatment. Thus, concentrations of LLOMe well below those causing apoptosis can rapidly permeabilize lysosomes to molecules up to at least 4.4K MW, but not to molecules of 10K MW or above.

If molecules of 10K MW are unable to exit lysosomes after LLOMe treatment, it seems unlikely that much larger cathepsins (20–30 kDa) could do so, as is postulated by the current model (Turk and Turk, 2009). To determine the localization of cysteine cathepsins in LLOMe-treated cells we used the cell permeable pan-cathepsin activity-based probe BMV109, which becomes fluorescent after it covalently binds to cysteine cathepsins (Withana et al., 2016). Cells labeled with BMV109 for 1 h showed puncta overlapping with the LAMP-1 signal, whereas pre-treatment of cells with 20 µM E64d for 1 h gave no labeling; this confirms the specificity of BMV109 signal for active cysteine cathepsins (Fig. 2A, left). By analogy to pre-loading cells with fluorescent lysosomal markers, cells were labeled with BMV109 for 1 h and then treated with 1 or 5 mM LLOMe for 15 min. Puncta representing cysteine cathepsin-positive lysosomes were preserved, indicating that cysteine cathepsins remained in lysosomes (Fig. 2A, middle). However, when cells were treated first with 1 or 5 mM LLOMe and then incubated with BMV109 and LLOMe for another hour, they showed no labeling, which argues that cysteine cathepsins were no longer active (Fig. 2A, right).

To clarify this intriguing observation, we used a microtiter assay to measure total cysteine cathepsin activity in cells upon LLOMe treatment, and, in addition, to probe for cysteine cathepsin activity in the cytosol (Repnik et al., 2016a). The endopeptidase activity of cysteine cathepsins, mostly catB and catL in HeLa cells, was monitored by measuring the cleavage of the fluorogenic substrate Z-FR-AMC (referred to from now on as ‘catB/L activity’). Appropriate conditions for cell lysis using digitonin were determined by measuring LDH or catB/L release to the supernatant, or by measuring catB/L or β-galactosidase activity in permeabilized cells, all at increasing digitonin concentrations (Fig. S2). We found that 30 μg/ml digitonin selectively permeabilized the plasma membrane, while 200 μg/ml digitonin efficiently permeabilized all cell membranes, including lysosomes. The assay was performed using four different strategies (see Fig. 2B). To measure the total cellular catB/L activity the substrate was added directly to cells lysed with 200 μg/ml digitonin (strategy 1). Alternatively, cells were lysed with 30 μg/ml digitonin to selectively permeabilize the plasma membrane, but not lysosomes (strategies 2a–2c). When the catB/L substrate (649 MW) was added to cells with the selectively permeabilized plasma membrane (strategy 2a), it could reach the cytosol as well as the lysosomal lumen provided the lysosomal membrane was sufficiently damaged by preceding treatment of cells by LLOMe. To demonstrate the release of catB/L into the cytosol, the organelle-free extract released from cells lysed with 30 μg/ml digitonin was transferred to new wells and only then was the catB/L substrate added (strategy 2b). In addition, after removal of the organelle-free extract, the remaining cell material (denoted ‘extracted cells’) was additionally lysed with 200 μg/ml digitonin before the catB/L substrate was added in order to monitor the catB/L activity that remained in membrane-bound organelles.

In control cells not treated with LLOMe, catB/L activity was detected after lysis with 200 μg/ml digitonin (Fig. 2C, strategy 1), but not with 30 μg/ml digitonin (Fig. 2C, strategy 2a), consistent with the observation that 30 μg/ml digitonin permeabilizes the plasma membrane but not lysosomes. In contrast, in cells treated with 1, 3 or 5 mM LLOMe for only 10 min we observed the same level of catB/L activity as in cells lysed with 200 or with 30 μg/ml digitonin (strategies 1, 2a); although the activity was reduced, especially at 1 mM. To understand this observation, we needed to discriminate between two possibilities: (1) that catB/L activity was detected because the fluorogenic substrate permeates into lysosomes through damaged lysosomal membranes, or, (2) that cysteine cathepsins are released into the cytosol. Therefore, we analyzed organelle-free extracts (strategy 2b). In these, hardly any catB/L activity could be detected, irrespective of whether or not cells were treated with LLOMe (Fig. 2C, strategy 2b). Consistent with this, the catB/L activity remained associated with extracted cells (Fig. 2C, strategy 2c). Taken together, these data argue strongly that, in response to LLOMe, although lysosomes become permeable to protons (Fig. 1D) and small molecules of 4.4K MW (Fig. 1E), cysteine cathepsins are not released into the cytosol and remain within lysosomes.

The in situ labeling of active cysteine cathepsins with the activity-based probe BMV109 suggested that LLOMe modulates cysteine cathepsin activity (Fig. 2A). More sensitive measurement of cysteine cathepsin activity using the fluorogenic substrate Z-FR-AMC also indicated that by 10 min after treatment with LLOMe the total catB/L activity was reduced, particularly at 1 mM LLOMe (Fig. 2C). Therefore, we next investigated how catB/L activity is modulated depending on LLOMe concentration (Fig. 2D). For this, catB/L activity was measured after adding the substrate directly to lysed cells (strategy 1, strategy 2a). In cells treated with 0.5–7 mM LLOMe for 30 min, roughly the same amount of catB/L activity was observed after total cell membrane lysis (strategy 1) as after lysis of only the plasma membrane (strategy 2a); this confirms that lysosomal membranes are permeabilized by LLOMe over a broad range of concentrations. This finding also corroborated the results obtained by staining cells with Acridine Orange (Fig. 1D). Notably, at above 0.25 mM LLOMe, total cell catB/L activity was clearly reduced when measured 30 min after treatment. In the range of 0.25–4 mM LLOMe, the remaining catB/L activity followed an inverted bell curve with the least activity at 1 mM LLOMe; this indicates that above 1 mM LLOMe, the inhibitory effect on cysteine cathepsin activity decreases with increasing LLOMe concentration.

To investigate this counterintuitive observation, we examined the effects of LLOMe on total catB/L activity over a broader time range (Fig. 2E). For this, cells were treated with 1, 3 or 5 mM, and catB/L activity was measured at 5, 15, 30, 60 and 90 min after treatment. Cells were lysed with 200 μg/ml digitonin (strategy 1) or 30 μg/ml (strategy 2a). This kinetic experiment demonstrated rapid loss of total catB/L activity, whereby at least half of the activity observed in control cells was lost within 30 min after LLOMe treatment. In addition, although initial loss of catB/L activity occurred most rapidly with 1 mM, by 90 min after incubation, more or less complete loss of catB/L activity was observed at all three LLOMe concentrations.

Since cells were treated continuously with LLOMe, we wondered whether loss of catB/L activity would also be observed after a pulse of LLOMe treatment followed by wash-out and incubation in LLOMe-free medium. CatB/L activity was reduced to the same extent in cells treated with LLOMe continuously for 1 h, as in cells treated with 1 mM LLOMe for 10 min, or 5 mM LLOMe for 5 min, and then cultured in LLOMe-free medium for 1 h (Fig. 2F). This argues that LLOMe causes the reduction in total catB/L activity rapidly after treatment and that the process continues even if LLOMe is washed out.

Because in permeabilized lysosomes the pH rises, and cysteine cathepsins are fairly unstable at neutral pH (Turk et al., 2012), we examined whether the elevation in pH alone could be responsible for the loss of catB/L activity in cells treated with LLOMe. For this, cells were incubated with 30 nM bafilomycin A1 (Fig. 3A). While catB/L activity was almost abolished 1 h after the treatment with 1 mM LLOMe (Fig. 2E), it took around 6 h to reach a similar loss of activity with 30 nM bafilomycin A1 (Fig. 3A). Thus, neutralization of the lysosomal pH during LMP by itself cannot account for the rapid loss of catB/L activity.

Since LLOMe appears to selectively target lysosomes within minutes, we asked whether the proton gradient is required for selective accumulation of LLOMe in lysosomes. For this, cells were first treated with bafilomycin A1 and then with LLOMe before measuring the catB/L activity (Fig. 3B). Surprisingly, catB/L activity was lost almost to the same extent with or without bafilomycin A1 pre-treatment, strongly arguing that LLOMe does not require an acidic lysosomal pH to cause loss of catB/L activity. In addition, catB/L activity measured in LLOMe-treated cells after lysis with 30 μg/ml digitonin (strategy 2a) was similar to the total catB/L activity measured after lysis with 200 μg/ml digitonin (strategy 1); this confirms that lysosomal membranes were indeed permeabilized and further argues that LLOMe does not depend on acidification to induce LMP.

CatC is crucial for LLOMe membranolytic activity and cytotoxicity (Thiele and Lipsky, 1990), so we next asked whether LLOMe also has a direct effect on catC activity. The aminodipeptidase activity of catC was measured by determining the level of the fluorogenic substrate H-GF-AMC (referred to as ‘catC activity’). The bulk of catC activity was indeed lost following treatment with LLOMe, and pre-treatment of cells with bafilomycin A1 only had a small effect (Fig. 3C), much like with catB/L activity (Fig. 3B). Compared to catB/L activity, catC activity proved even more resistant to an elevated pH mediated by pre-treatment with bafilomycin A1, since its activity was reduced by half compared to that in control cells only after 6–12 h (Fig. 3D).

To better understand the rapid loss of cysteine cathepsin activity upon LLOMe treatment, we performed immunoblotting of whole-cell extracts. Compared to control cells, the protein levels for the mature forms of catB, catL and catC were visibly reduced already 15 min after treatment with 1, 3 or 5 mM LLOMe, and even more so after 1 h (Fig. 3E). This indicates that LLOMe-mediated loss of cysteine cathepsin activities is not due to changes in enzyme kinetics but rather due to a considerable decrease in the total amount of enzymes. In contrast, with bafilomycin A1 alone, the only enzyme whose protein level was strongly reduced was catL; catB and catC levels were reduced much less with bafilomycin A1 than with LLOMe treatment (Fig. 3E). As shown above, bafilomycin A1 alone causes loss of catB/L and catC activities at a much slower rate than LLOMe (Fig. 3A–D). Collectively, these results argue that the mechanism of LLOMe acting on cysteine cathepsins is not mediated solely by the elevated pH that accompanies LMP. We tentatively concluded that LLOMe can structurally destabilize cysteine cathepsins, which subsequently become susceptible to proteolytic degradation.

If cysteine cathepsins in LLOMe-treated cells remain in lysosomes after LMP and are degraded there, then inhibition of the proteolytic system in the cytosol represented by proteasomes should not prevent the loss of cysteine cathepsins. To test this hypothesis, cells were incubated with proteasome inhibitors MG132 or bortezomib for 30 min before they were treated with 1, 3 or 5 mM LLOMe for 1 h. MG132 had a significant effect on catB degradation; in contrast the effect of bortezomib on cysteine cathepsin degradation was minimal (Fig. 3F). Importantly, it has been reported that MG132 is also an efficient inhibitor of cysteine cathepsins (Ito et al., 2009), which we confirmed in HeLa cells using our microtiter assay for total catB/L and catC activity measurements (strategy 1) (Fig. 3G). In contrast, bortezomib showed little effect on catB/L activity and no effect on catC activity. Both inhibitors showed a strong effect in degradation assays for long-lived proteins (data not shown), confirming that both efficiently inhibited proteasomes. Therefore, we conclude that selective inhibition of proteasomes does not affect degradation of cysteine cathepsins, which is consistent with our observation that cysteine cathepsins are not released in the cytosol and that their degradation takes place inside lysosomes.

LLOMe had striking effects on the activities and protein levels of the three cysteine cathepsins. We therefore investigated whether other lysosomal hydrolases are also affected by LLOMe treatment. For this, the activities of three glycosidases were monitored using specific fluorogenic substrates. After a 6 h treatment with 1, 3 or 5 mM LLOMe, α-D-galactosidase activity was reduced by up to 20%, β-N-acetyl-galactosidase activity by up to 40% and β-D-galactosidase activity by up to 60% (Fig. S3A). As observed with cysteine cathepsins, LLOMe-mediated loss of β-D-galactosidase activity was initiated rapidly after LLOMe treatment and continued in LLOMe-free medium (Fig. S3B). This loss of activity also occurred in cells with neutralized lysosomal pH (Fig. S3C).

We observed that inhibition of cysteine cathepsins by MG132 largely prevented their degradation after treatment of cells with LLOMe (Fig. 3F,G). Since the membranolytic activity of LLOMe was shown to be enhanced by catC-mediated polymerization of LLOMe (Thiele and Lipsky, 1990), we asked whether inhibition of catC in particular would also affect the LLOMe-mediated loss of cysteine cathepsin activity and/or protein levels. In order to determine conditions for selective inhibition of catC in HeLa, we evaluated the inhibitory profiles of the cysteine cathepsin inhibitors E64d, GF-DMK and leupeptin (Fig. 4A,B). The inhibitors were incubated with cells at several concentrations for 30 min or 2 h before cells were lysed with 200 μg/ml digitonin (strategy 1) to analyze catB/L and catC activities by using fluorogenic substrates.

The results were surprising (Fig. 4A,B) as they indicated that E64d is a potent inhibitor of not only catB/L activity, as expected, but also of catC activity. Thus, E64d could not be used to selectively inhibit catB/L without also inhibiting catC activity to some extent. In contrast, GM-DMK selectively inhibited catC. From these data, we selected the more rigorous inhibitory conditions for further experiments (Fig. 4C,D).

First, we investigated the effect of catC inhibition on LLOMe-mediated loss of catB/L activity (Fig. 4E). When catC was inhibited there was an almost complete preservation of catB/L activity after LLOMe treatment. The effect of catC inhibition decreased with increasing LLOMe concentration, which could suggest that high concentrations of LLOMe can cause some loss of catB/L activity independently of catC activity. In contrast, preferential catB/L inhibition had little effect on LLOMe-mediated loss of catC activity (Fig. 4F). Similarly, LLOMe-mediated loss of β-D-galactosidase activity was also strongly dependent on catC activity (Fig. S3D).

In addition, we analyzed the effect of the pre-treatment with these inhibitors on protein levels of catB, catL and catC after LLOMe treatment (Fig. 4G). Immunoblotting confirmed that loss of cysteine cathepsins at the protein level was considerably reduced only under conditions that inhibited catC, and not by preferential inhibition of catB/L or by neutralizing lysosomal pH with bafilomycin A1.

Our results that cysteine cathepsins are not released from lysosomes by LLOMe contradict the current model of LLOMe-induced apoptosis (Cirman et al., 2004; Droga-Mazovec et al., 2008). However, our data on the inhibitory profiles (Fig. 4A–C) raised the possibility that the protective effect of E64d against LLOMe-induced apoptosis observed in these earlier studies was mediated by inhibition of catC rather than by selective inhibition of endopeptidase cysteine cathepsin activity. Thus, we separately investigated the effect of preferential inhibition of catB/L, selective inhibition of catC or inhibition of all cysteine cathepsin activity on different adverse effects induced in cells treated with LLOMe (Fig 4H–J; Fig. S3D). Importantly, our results reveal that inhibition of catC, but not of catB/L activities protects cells from LMP, and from loss of cysteine cathepsin and lysosomal glycosidase activities, as well as from caspase activation and cell death.

The results until now showed that 1 mM LLOMe, which is well below the apoptotic threshold, causes rapid and complete LMP in conjunction with rapid and substantial loss of cysteine cathepsin activities and protein levels. In spite of these severe insults, cells did not undergo cell death but continued to grow, even in the continued presence of LLOMe in the medium. Therefore, we next analyzed lysosomal pH and the content of cysteine cathepsins following longer incubation periods to try to understand how cells are able to survive.

Staining with LysoTracker or Acridine Orange over a 24 h period following 1 mM LLOMe treatment showed that after initial loss of acidic pH, cells progressively regained their ability to accumulate weak basic dyes (Fig. 5A,B), which indicates that both the integrity of lysosomal membranes, as well as the acidic pH were restored. The recovery process had started within 30 min after LLOMe treatment, and at 4 h after treatment cells were in fact able to accumulate more dye than control cells; this is likely due to the increased size (volume) of lysosomes [as suggested by electron microscopy (EM) data below]. Cycloheximide, an inhibitor of protein synthesis, did not prevent re-acidification of cells (Fig. 5B), indicating that membrane recovery was largely independent of de novo protein synthesis.

To investigate whether cells also regain normal levels of cysteine cathepsins when treated continuously with LLOMe, we performed immunoblotting of whole-cell extracts up to 48 h following 1 mM LLOMe treatment (Fig. 6A). As shown by this analysis, the protein levels of catB, catL and catC decreased progressively until 6 h after treatment. However, cysteine cathepsin protein bands reappeared by 12 h, and, in the case of catB and catL, these were characterized by immature, unprocessed forms, and single-chain forms of cathepsins, consistent with the presence of newly synthesized molecules. Analysis of catB/L activity following 1 mM LLOMe treatment showed that despite some protein remaining, cells were almost completely devoid of catB/L activity between 1 and 9 h after LLOMe treatment (Fig. 6B). At 12–24 h following the treatment, catB/L activity rapidly increased (Fig. 6C). Importantly, catB/L activity was not regained in cells pre-treated with brefeldin A, an inhibitor of transport from endoplasmic reticulum to the Golgi, or with cycloheximide, which confirms that the reappearance of catB/L activity was dependent on de novo protein synthesis and transport to lysosomes. Moreover, similar results were obtained for β-D-galactosidase activity (Fig. S3E), suggesting that re-acquisition of other lysosomal hydrolases that are affected by LLOMe treatment is also dependent on de novo protein synthesis.

We hypothesized that because of a destabilizing effect of LLOMe on lysosomal hydrolases, and in particular cysteine cathepsins, lysosomal proteolytic capacity would be impaired. In order to test this, we measured degradation of long-lived proteins during the first 10 h of incubation with LLOMe. In parallel, cells were incubated with either bafilomycin A1 or the cysteine cathepsin inhibitor E64d to determine the fraction of degradation that presumably takes place in lysosomes (Fig. 6D). We observed that 1 mM LLOMe reduced degradation of long-lived proteins to the same extent as the two inhibitors of lysosomal degradation. Moreover, pre-treatment of cells to selectively inhibit catC activity was able to prevent LLOMe from reducing lysosomal degradation, which confirms our earlier observations that LLOMe-mediated loss of lysosomal hydrolases depends on catC activity (Fig. 4E,G).

Finally, we focused on the structure of lysosomes, taking advantage of a well-characterized and abundant lysosomal membrane protein, LAMP-1. This protein showed no quantitative changes in response to 1 mM LLOMe treatment when evaluated by immunoblotting (Fig. 6A). In addition, LAMP-1 localization was analyzed by whole-mount immunofluorescence labeling (Fig. S4). Following continuous culture with 1 mM LLOMe, LAMP-1-positive vesicles became enlarged and localized to the perinuclear region, which corroborates the results with LysoTracker Red DND-99 staining (Fig. 5A). However, by 48 h after treatment, the pattern of LAMP-1 labeling reversed and resembled the appearance of lysosomes seen in cells not treated with LLOMe.

To analyze the size of lysosomes in more detail, ultrastructural analysis was performed by transmission electron microscopy (TEM) (Fig. 7A,B). For this, we incubated cells with 5 nm colloidal gold particles stabilized with bovine serum albumin for 4 h to allow their uptake by fluid-phase endocytosis, followed by a 2-h chase to allow gold particle accumulation in lysosomes. Then cells were treated with LLOMe and analyzed at indicated time points. Individual lysosomes containing internalized gold particles already appeared enlarged at 30 min after treatment with 1 mM LLOMe. We never observed colloidal gold particles in the cytosol, indicating that they remain localized to lysosomes. At 6 h, and even more so at 24 h after treatment, lysosomes were strikingly enlarged and the ultrastructure of the lumen became heterogeneous. There was an extensive increase in cargo domains not containing internalized gold particles (Fig. 7B), suggesting that this material accumulated during the post-treatment period, which is consistent with our results showing that lysosomal degradation is reduced after LLOMe treatment (Fig. 6D). The fraction of the total lysosomal volume relative to the total volume of the cytoplasm was estimated by point-counting stereological analysis (Fig. 7C). The volume fraction of lysosomes had already doubled by 1 h after LLOMe treatment and by 24 h this increased further to a 3-fold larger volume fraction than in control cells. In most cells, after treatment with LLOMe, the relative lysosomal volume was reduced to pre-treatment levels after 48 h. We therefore conclude that de novo synthesized cysteine cathepsins are delivered to enlarged LAMP-1-positive lysosomes and consequently enable degradation of the accumulated cargo, leading to the eventual decrease in the volume of the lysosomal compartment.

The current model of LLOMe-induced apoptosis postulates that when LLOMe is added to cells it accumulates in, and permeabilizes lysosomes, which leak cysteine cathepsins into the cytosol, where their proteolytic activities initiate apoptotic signaling (Cirman et al., 2004; Droga-Mazovec et al., 2008). Our data here argue against this concept and call for a new model, which needs to take into account the following key observations: (1) the lowest concentrations of LLOMe that induce LMP (above 250 μM in HeLa cells) are much lower than those needed to induce apoptosis (above 4 mM in HeLa cells); (2) LMP occurs within 10 min after LLOMe is added to cells and affects all lysosomes; (3) internalized molecular markers of at least 4.4K MW are released from permeabilized lysosomes into the cytosol, but molecules of 10K MW and 5 nm colloidal gold particles are retained in lysosomes; (4) cysteine cathepsins are not released into the cytosol, but are structurally destabilized inside lysosomes, leading to loss of their activity and their degradation; (5) adverse effects of LLOMe on cells do not depend on catB/L activity; (6) LMP is transient such that lysosomal membranes anneal and lysosomes re-acidify within hours after continuous incubation with LLOMe, a process that does not need de novo protein synthesis; and (7) following LLOMe treatment, lysosomal degradation is reduced for many hours and de novo synthesis of lysosomal hydrolases is needed to restore it to normal levels.

Based on previous reports and our own data, we propose a new model to describe the effects of LLOMe on lysosomes, which is summarized in Fig. 8.

LLOMe added to cells passively crosses cell membranes into the cytoplasm and accumulates in lysosomes either because of protonation in the acidic lumen or alternatively because of ester hydrolysis, which generates free leucine dipeptides (Ransom and Reeves, 1983; Reeves, 1979). As suggested previously by experiments where cells were pre-treated with ammonium chloride, an acidic pH is not a prerequisite for generation of LLOMe polymers (Thiele and Lipsky, 1990). This implies that low pH is also not a prerequisite for LLOMe to accumulate in lysosomes at concentrations sufficient to drive catC-mediated LLOMe polymerization. In fact, the rate of lysosomal ester hydrolysis is highest in the neutral pH range (Goldman and Kaplan, 1973). In line with these observations, our study showed that LLOMe-mediated loss of lysosomal enzyme activities was reduced very little in cells that were pre-treated with bafilomycin A1 to block the lysosomal proton gradient (Figs 3B,C and 4G; Fig. S3C).

Rapid selective permeabilization of lysosomal membranes is one of the most striking effects of LLOMe. We show that with increasing concentration of LLOMe, LMP is increasingly independent of catC activity (Fig. 4H), which was shown to mediate LLOMe polymerization (Thiele and Lipsky, 1990). Although the exact mechanism of membrane destabilization by LLOMe is not known, an increased membranolytic activity of polymerized LLOMe can be correlated with its increased hydrophobicity. Hydrophobicity is also increased in a free leucine dipeptide, which is generated from LLOMe after hydrolysis of the ester bond. Poly-leucine chains might not be much larger than six leucine residues (Thiele and Lipsky, 1990) and are thus too short to span the lysosomal membrane. According to the partitioning simulations, leucine peptides of six or fewer residues rapidly adsorb to the membrane surface, but are unlikely to insert deeper into the membrane (Ulmschneider et al., 2010). This would lead to accumulation of LLOMe polymers as well as monomers, in particular free leucine dipeptides, at the luminal leaflet of the phospholipid bilayer. Consequently, the bilayer would expand asymmetrically until, at above a certain threshold value, the bilayer curvature strain would cause abrupt destabilization of the bilayer structure (Heerklotz, 2008). This scheme would be consistent with many other reports showing that LLOMe affects lysosomal stability rapidly after LLOMe treatment (Aits et al., 2015; Maejima et al., 2013; Thiele and Lipsky, 1992a; Uchimoto et al., 1999). Notably, in our experiments all lysosomes in the cells were permeabilized (Fig. 1D,E), which argues for a complete LMP. This contradicts an earlier, albeit less convincing report, that LLOMe caused limited LMP (Droga-Mazovec et al., 2008).

Since LLOMe causes rapid loss of cysteine cathepsin activities (Figs 2A,C–F, 3E), which cannot be explained by an elevated lysosomal pH alone (Fig. 3A,D), we assume that LLOMe directly interacts with lysosomal hydrolases, especially cysteine cathepsins. Like LMP, this effect of LLOMe is enhanced by catC activity (Fig. 4E), suggesting that polymerized LLOMe is more potent in destabilizing the enzymes. Similar to the interaction of leucine peptides with lipid bilayers (Ulmschneider et al., 2010), LLOMe polymers may either accumulate on the surface of hydrolases or protrude deeper into their hydrophobic interior. This way, LLOMe could destabilize intramolecular interactions and cause the molecules to unfold. Unfolded molecules would then become susceptible to proteolytic degradation, which could explain why cathepsins appear less abundant in immunoblots (Figs 3E,F, 4G and 6A).

Slower loss of cysteine cathepsins at 5 mM compared to 1 mM LLOMe (Fig. 2E) could reflect differences in the kinetics of LMP. This would be similar to glycyl-phenylalanine-β-naphthylamide-induced LMP, whose rate is concentration-dependent (Penny et al., 2014). We hypothesize that at 1 mM LLOMe, LMP is delayed until a threshold level of polymerized forms is generated by cathepsin C activity. During this delay, polymerized forms of LLOMe accumulating inside lysosomes also target luminal hydrolases, including cysteine cathepsins, such that by the time LMP occurs little activity remains. At 5 mM LLOMe, LMP occurs faster because it is only partially dependent on catC-mediated polymerization of LLOMe (Fig. 4H) and because the polymerization of LLOMe by catC is probably faster due to the higher LLOMe concentration. At this high concentration, luminal hydrolases might endure little damage in the short period before LMP. However, after LMP, cysteine cathepsins would remain exposed to the destabilizing effects of LLOMe because they continue to be restricted to lysosomes, where catC-dependent polymerization of LLOMe also continues to takes place. This explanation also fits with the observation that although cysteine cathepsin activities were lost faster at 1 mM LLOMe, the protein levels of cathepsins were higher than at 5 mM (Fig. 3E). We assume that these remaining protein levels represent structurally destabilized and inactive cysteine cathepsins, which are degraded slower at 1 mM than at 5 mM LLOMe because of particularly abrupt and extensive loss of the proteolytic capacity after treatment with the low concentration.

In contrast to what is widely assumed, our study provides strong evidence through multiple techniques that cysteine cathepsins are not released from LLOMe-permeabilized lysosomes into the cytosol. By monitoring the retention of small internalized lysosomal markers, it was previously shown that in cells treated with glycyl-phenylalanine-β-naphthylamide, a compound with a mechanism of action similar to LLOMe (Goldman and Kaplan, 1973; Thiele and Lipsky, 1990), lysosome-loaded sulforhodamine B was released into the cytosol, whereas larger 10K MW dextran was not (Steinberg et al., 2010; Penny et al., 2014). This is consistent with our results using LLOMe (Fig. 1E). It is important to note that studies that suggested release of cathepsins into the cytosol mostly analyzed isolated lysosomes or cytoplasmic extracts following subcellular fractionation (Cirman et al., 2004; Maejima et al., 2013). It is reasonable to assume that cells treated with LLOMe may be more prone to artifactual release of lysosomal hydrolases during the subsequent lengthy and harsh analytical procedures, which includes challenge with detergents, including digitonin, or mechanical stress, including centrifugation at high speed. We consider non-invasive imaging assays, such as release or retention of internalized fluorescent lysosomal markers to be a more reliable approach to evaluate the size-exclusion characteristics of permeabilized lysosomal membranes. In addition, localization of cysteine cathepsins to lysosomes before and after LMP was also confirmed by non-invasive in situ labeling of cysteine cathepsins with the activity-based probe BMV109 (Fig. 2A).

During continued incubation of cells in the presence of sub-apoptotic concentrations of LLOMe, the lysosomal limiting membrane is rapidly annealed, independently of de novo protein synthesis, and the proton gradient is re-built (Fig. 5). This is consistent with earlier findings showing that after a 10-min incubation with glycyl-phenylalanine-β-naphthylamide, which was sufficient to completely neutralize the pH of the entire population of lysosomes in a cell, lysosomes re-acidified within minutes after this reagent was washed out (Steinberg et al., 2010). We hypothesize that in this rapid recovery phase, the lysosomal membrane anneals in a process facilitated by a rearrangement of leucine peptides within the phospholipid bilayer in a biophysically stable manner (Heerklotz, 2008). It has also recently been suggested that the autophagic machinery is recruited to damaged lysosomes and that this process may contribute to clearance or repair of damaged lysosomes (Aits et al., 2015; Maejima et al., 2013; Papadopoulos et al., 2017).

After LLOMe treatment, lysosomes remain depleted of cysteine cathepsins, and to a lesser extent of other lysosomal proteases (Fig. 6A,B; Fig. S3A). Lysosomal cargo accumulates in enlarged, hydrolysis-defective lysosomes (Figs 6D and 7A–C). The normal organization of the endocytic compartments is restored only after lysosomes regain their degradative function, which is dependent on de novo synthesis of hydrolases (Figs 6C; Fig. S3E).

Similar to observations in a previous report (Méthot et al., 2007), we observed that E64d, a potent inhibitor of cysteine cathepsins with endopeptidase activity (such as catB and catL) is also a strong inhibitor of catC (Fig. 4B). Moreover, we showed that in LLOMe-treated cells, selective inhibition of catC alone with the inhibitor GF-DMK reduces caspase activation and apoptosis, whereas preferential inhibition of catB/L activity with a low dose of E64d, or leupeptin, does not (Fig. 4I,J). Our results imply that conclusions about the essential role of cysteine cathepsins in proteolytic signaling of LLOMe-induced apoptosis reported previously (Cirman et al., 2004; Droga-Mazovec et al., 2008) are incorrect because the possibility that the cysteine cathepsin inhibitor E64d could inhibit also catC was not considered. Moreover, consistent with our conclusions, other recent studies have observed that LMP is not correlated with the induction of apoptosis (Aits et al., 2015; Maejima et al., 2013).

LLOMe is a compound with many striking features and our study adds to our knowledge of its actions by unraveling mechanisms of LLOMe-mediated effects and their consequences for the treated cells. We show that, within minutes of being added to cells, LLOMe causes LMP, and loss of cysteine cathepsins and other lysosomal hydrolase activities, with a consequent impairment of lysosomal degradation. Remarkably, after this spike of destabilizing action, which is most pronounced at sub-apoptotic concentrations, cells are able to fully recover even in the presence of LLOMe. We believe that this mechanism could lead to novel, or improve already existing, approaches for LLOMe-based modulation of cell functions. For example, LLOMe could be used for release of endocytosed compounds into the cytosol or for transient knockdown of lysosomal hydrolases, especially cysteine cathepsins.

LLOMe, Z-VAD(OMe)-FMK [benzyloxycarbonyl-Val-Ala-Asp(OMe)-fluoromethylketone], H-GF-AMC (H-Gly-Phe-7-amido-4-methylcoumarin) (Bachem); Z-FR-AMC (benzyloxycarbonyl-Phe-Arg-7-amido-4-methylcoumarin), E64d [2S,3S-trans-(ethoxycarbonyloxirane-2-carbonyl)-L-leucine-(3-methylbutyl) amide] (Enzo Life Sciences); GF-DMK (Gly-Phe-diazomethylketone) (MP Biomedicals); β-N-acetylglucosaminidase substrate 4-methyl umbelliferone(MU)-acetyl-β-D-glucosaminide (Calbiochem, CA); Acridine Orange, bafilomycin A1, brefeldin A, cycloheximide, Crystal Violet, α-D-galactosidase substrate MU-α-D-galactopyranoside, β-D-galactosidase substrate MU-β-D-galactopyranoside, 4.4K MW dextran–TRITC (Sigma); bortezomib (Merck, NJ); digitonin, Ac-DEVD-AMC (N-acetyl-Asp-Glu-Val-Asp-7-amido-4-methylcoumarin), MG132 (Cayman Chemical); LysoTracker Red DND-99, sulforhodamine B, 10K MW dextran–Alexa Fluor 546, 70K MW dextran–Texas Red (Thermo Fisher Scientific, MA, USA); iABP smart cathepsin probe kit (BMV109) (Vergent Bioscience, MN). The following antibodies were used: mouse monoclonal against catL (clone CPL33/1) at 1:2000 (Abcam), rabbit polyclonal against catB at 1:1000 (PA5-14255), mouse monoclonal against α-tubulin (clone TU-01) at 1:10,000 (both ThermoFisher Scientific), mouse monoclonal against catC (clone D-6) at 1:200, mouse monoclonal against β-actin (clone C-4) at 1:1000 (Santa Cruz Biotechnology), mouse antibody against LAMP-1 (clone H4A3, hybridoma supernatant) at 1:1000 for western blotting and at 1:70 for immunocytochemistry (Developmental Studies Hybridoma Bank (DSHB, deposited by J.T. August and J.E.K. Hildreth), horseradish peroxidase-conjugated secondary antibodies at 1:5000 (GE Healthcare), and goat anti-mouse-IgG secondary antibody conjugated to Alexa Fluor 488 at 1:200 (Molecular Probes, Thermo Fisher Scientific).

HeLa cells were purchased from the European Collection of Cell Cultures (ECACC, product number 93021013). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 2 mM stable L-glutamine (complete medium) (all Lonza, Switzerland) and 10% fetal bovine serum (FBS; Sigma), at 37°C in a humidified atmosphere with 5% CO₂. For Crystal Violet staining and LDH release assays, cells were cultured in DMEM without Phenol Red (Lonza) and supplemented with 5% FBS and 2 mM stable L-glutamine. For all microtiter assays, cells were seeded in 96-well plates at 0.015×10⁶ cells per well. Cells were grown overnight before they were treated.

The assay was performed as described by Feoktistova et al. (2016). Cells were washed once with PBS and then fixed with 4% paraformaldehyde in 200 mM HEPES for 15 min. Cells were washed once with dH₂O, before a 15-min incubation with 0.05% Crystal Violet in 20% ethanol, followed by six rounds of washing with dH₂O. Afterwards 100 µl of methanol was added to solubilize the dye and the absorbance was measured at 595 nm in a Wallac Victor 2 (Perkin Elmer) plate reader. Experiments were performed in triplicate.

LDH assay buffer [200 mM Tris-HCl pH 8.0, 0.05% 2-p-iodophenyl-3-p-nitrophenyl-5phenyl tetrazolium chloride (INT), 0.01% 1-metoxy-5-methylphenazinium methyl sulfate, 0.12% nicotinamide adenine dinucleotide (NAD) and 0.7% lactic acid (all Sigma)] was prepared as described by Chan et al. (2013). Cell culture supernatant was harvested and centrifuged at 200 g, for 2 min. Cleared supernatant was mixed with LDH aasay buffer at 1:1 volume ratio. After a 30-min incubation absorbance was measured at 570 nm in a Wallac Victor 2 plate reader. Experiments were performed in triplicate.

Whole-cell extracts were prepared using caspase buffer (50 mM HEPES pH 7.2, 100 mM NaCl, 10% sucrose, 0.1% CHAPS, 1 mM EDTA) (Stennicke and Salvesen, 1997), supplemented with 0.1% Triton X-100. Protein concentration in lysates was determined by using BCA reagent (Thermo Fisher Scientific). Caspase activity was measured for 50 μg of protein in a final volume of 100 μl caspase buffer supplemented with 5 mM DTT and 13 μM Ac-DEVD-AMC substrate. Substrate hydrolysis was measured for at least 15 min at an excitation of 380 nm and emission of 460 nm in a Wallac Victor 2 (Perkin Elmer) plate reader. The velocity of the initial reaction was calculated. Experiments were performed in triplicate.

The assay was as described previously (Repnik et al., 2016c). For microscopic analysis, cells were seeded in four-chamber glass-bottom wells (Lab-Tec, Nunc). After treatment, LysoTracker Red DND-99 was added to cells at 50 nM for 10 min. Cells were analyzed with an Olympus FluoView 1000 inverted confocal laser scanning microscope using Olympus UPlanSApo ×60 (NA 1.35) and ×100 (NA 1.4) oil immersion objectives and Olympus FluoView software. For flow cytometry quantification of staining, cells were seeded on a 24-well plate at 0.1×10⁶ cells per well. After treatment, Acridine Orange was added to cells at 1 μg/ml for 10 min. Cells in suspension and adherent cells were collected and analyzed with a FACSCalibur™ flow cytometer (BD Biosciences). Data were analyzed with Flowing Software v. 2.5.1. (University of Turku, Finland) and represent geometric mean. Experiments were performed in duplicate.

Cells were seeded in four-chamber glass-bottom wells (Lab-Tec, Nunc, Thermo Fisher Scientific) at 0.1×10⁶ cells per well. Cells were incubated with sluforhodamine B at 0.2 μg/ml (Carraro-Lacroix et al., 2011), or 4.4K MW dextran–TRITC, 10K MW dextran–Alexa Fluor 546 or 70K MW dextran–Texas Red at 0.5 μg/ml, for 4 h, followed by a 2-h chase in complete medium without markers. Cells were analyzed with the Olympus microscope as described above.

Cells were detached with 0.02% EDTA (Versene, Lonza) and seeded on poly-L-lysine-coated glass coverslips. For in situ labeling of cysteine cathepsins, BMV109 was added to cell culture at 1 µM and incubated for 1 h. Cells were washed twice with complete medium and then fixed with 4% PFA in 200 mM HEPES pH 7.4, for 15 min. After washing with PBS, residual fixative was quenched with 0.1% glycine in PBS for 10 min. Cell membranes were permeabilized with 0.1% Triton X-100 in PBS for 10 min and then incubated with 1% bovine serum albumin (BSA) in PBS for 30 min to block non-specific labeling. Antibodies were diluted in 1% BSA in PBS and incubated on cells for 1 h at room temperature. Nuclei were stained with 0.5 µg/ml 4′,6-diamidine-2′-phenylindole dihydrochloride (DAPI, Sigma) and mounted on glass slides with Mowiol 4-88 mounting medium. Samples were analyzed with the Olympus FluoView 1000 inverted confocal laser scanning microscope as described above.

The microtiter plate assay was set up based on earlier reports (Hulkower et al., 2000; Foghsgaard et al., 2001) and performed as described previously (Repnik et al., 2016a). Cells were seeded in a 96-well black transparent-bottom plate (Brand) at 0.015×10⁶ cells per well and grown overnight before the treatment. To measure enzyme activity cells were washed with PBS, and then lysed with 50 μl acetate buffer (50 mM Na-acetate pH 5.6, 150 mM NaCl, 0.5 mM EDTA) supplemented with 30 µg/ml digitonin to selectively lyse the plasma membrane, or with 200 µg/ml digitonin to lyse all cell membranes. Unless described differently, after a 12-min incubation with digitonin on ice, 50 μl of the fluorogenic substrate, prepared in acetate buffer supplemented with 10 mM DTT, was added at a final concentration of 30 µM. Substrate hydrolysis was measured for at least 15 min at an excitation of 380 nm and emission of 460 nm in a Wallac Victor 2 plate reader. The velocity of the initial reaction was calculated. Experiments were performed in triplicate.

The assay was adapted from that described in Engedal et al. (2013). Cells were cultured in complete medium supplemented with 0.125 µCi/ml L-[¹⁴C]-valine (PerkinElmer) for 24–48 h, followed by a chase in complete medium containing 10 mM non-radioactive L-valine for 16–18 h. Afterwards cells were treated for 10 h as described in the Results. Two measurements of radioactivity were performed for each sample – the acid-soluble fraction of the supernatant and the total of the acid-insoluble fraction of the supernatant and the cells. Ultima Gold LSC Cocktail (PerkinElmer) was added to each sample and radioactivity was measured in a liquid scintillation counter (Tri-Carb 1900TR; PerkinElmer) for 5 min per sample. The percentage of degradation of long-lived proteins was defined as the fraction of acid soluble radioactivity relative to total radioactivity.

All other experimental procedures were performed as described previously: western blotting (Borg et al., 2014), making 5 nm colloidal gold particles (Repnik et al., 2016b), transmission electron microscopic analysis (Løvmo et al., 2017) and point-counting stereological analysis (Lucocq and Hacker, 2013).

We thank Russ Hodge (Max Delbrück Center for Molecular Medicine) for an inspiring discussion on scientific thinking and writing, Maximiliano Gutierrez (Francis Crick Institute) and Fahri Saatcioglu (University of Oslo, UIO) for critical reading of the manuscript, Jens Wohlmann (University of Bonn) for discussion on experimental procedures, Anne Simonsen (University of Oslo) for assistance with the degradation assay of long-lived proteins, Catherine Heyward (University of Oslo) for advice on fluorescence imaging and Ingrid Kjos for advice on western blotting. We thank NorMIC imaging platform and EM laboratory at the Department of Biosciences, University of Oslo, and Alexandre Corthay and Flow Cytometry Core Facility at the Oslo University Hospital.