The lysosomal proteome of senescent cells contributes to the senescence secretome

Abstract Senescent cells accumulate in tissues over time, favoring the onset and progression of multiple age‐related diseases. Senescent cells present a remarkable increase in lysosomal mass and elevated autophagic activity. Here, we report that two main autophagic pathways macroautophagy (MA) and chaperone‐mediated autophagy (CMA) are constitutively upregulated in senescent cells. Proteomic analyses of the subpopulations of lysosomes preferentially engaged in each of these types of autophagy revealed profound quantitative and qualitative changes in senescent cells, affecting both lysosomal resident proteins and cargo proteins delivered to lysosomes for degradation. These studies have led us to identify resident lysosomal proteins that are highly augmented in senescent cells and can be used as novel markers of senescence, such as arylsulfatase ARSA. The abundant secretome of senescent cells, known as SASP, is considered their main pathological mediator; however, little is known about the mechanisms of SASP secretion. Some secretory cells, including melanocytes, use the small GTPase RAB27A to perform lysosomal secretion. We found that this process is exacerbated in the case of senescent melanoma cells, as revealed by the exposure of lysosomal membrane integral proteins LAMP1 and LAMP2 in their plasma membrane. Interestingly, a subset of SASP components, including cytokines CCL2, CCL3, CXCL12, cathepsin CTSD, or the protease inhibitor SERPINE1, are secreted in a RAB27A‐dependent manner in senescent melanoma cells. Finally, proteins previously identified as plasma biomarkers of aging are highly enriched in the lysosomes of senescent cells, including CTSD. We conclude that the lysosomal proteome of senescent cells is profoundly reconfigured, and that some senescent cells can be highly active in lysosomal exocytosis.


| INTRODUC TI ON
Lysosomes are membrane-bound intracellular organelles with a relevant role in metabolism and in organelle and protein quality control (Lawrence & Zoncu, 2019). These organelles are characterized by a single limiting membrane and an acidic lumen enriched in resident hydrolases, including numerous proteases. Lysosomes mediate the degradation and recycling of intracellular components delivered to lysosomes through autophagy, and extracellular material captured in endosomes and phagosomes (Levine & Kroemer, 2008;Settembre et al., 2013;Xu & Ren, 2015). Lysosomes also perform other important cellular functions such as Ca 2+ -buffering, intracellular signaling, and direct extracellular secretion of the lysosomal contents. The latter process, known as lysosomal secretion, is a Ca 2+ -dependent mechanism active in osteoclasts, melanocytes, endothelial cells, and cells from the hematopoietic lineage, including lymphocytes, neutrophils, mast cells, and macrophages (Luzio et al., 2014;Schmidt et al., 2009;Settembre et al., 2013;Sheshachalam et al., 2014).
It has long been known that the lysosomal compartment is largely expanded in senescent cells (Robbins et al., 1970). In fact, the characteristic senescence-associated β-galactosidase (SAβGal) activity detected in senescent cells reflects the increased lysosomal mass of senescent cells (Dimri et al., 1995;Kurz et al., 2000;Lee et al., 2006). Similarly, most tested lysosomal hydrolases are also enriched in senescent cells, including α-mannosidase, α-fucosidase, and N-acetylβ-hexosaminidase (Knaś et al., 2012). However, there is a lack of comprehensive information on the proteome of lysosomes of senescent cells, including both lysosome-resident and cargo proteins.
The intracellular proteins degraded in lysosomes in mammalian cells generally enter through two separate routes, macroautophagy and chaperone-mediated autophagy. Macroautophagy (MA) is responsible for capturing multiprotein complexes, such as ribosomes, protein aggregates, and organelles, inside double-membrane vesicles (autophagosomes) that deliver their cargo to lysosomes through vesicular fusion. MA is potentiated during cellular senescence, and its inhibition delays the establishment of senescence and the senescence-associated secretory phenotype (SASP) (Gamerdinger et al., 2009;Young et al., 2009). Interestingly, prolonged activation of MA during senescence leads to the activation of mTOR, thereby facilitating protein synthesis and contributing to the SASP (Herranz et al., 2015;Laberge et al., 2015;Narita et al., 2011). Chaperonemediated autophagy (CMA) constitutes another important mechanism of intracellular protein degradation in lysosomes through direct funneling of protein cargoes across the lysosomal membrane. Substrate proteins are targeted by the chaperone HSC70 to lysosomes and internalized through a translocation complex formed by the lysosomal membrane protein LAMP2A (Kaushik & Cuervo, 2018). Despite the important roles described for CMA in the maintenance of the metastable proteome and in the regulation of cellular processes such as metabolism, cell cycle, transcription, or cell death Park et al., 2015;Schneider et al., 2014;Valdor et al., 2014), little is known about CMA in senescent cells.
Here, we have characterized the autophagic flux through MA and CMA in senescent cells, observing a previously unknown increase in CMA activity, which is parallel to the known increase in MA, albeit with different kinetics. Taking advantage of the possibility of isolating lysosomal subpopulations preferentially engaged in MA or CMA, we present here a comprehensive quantitative proteomic analysis of purified lysosomes from senescent cells, which reveals selective changes in the composition and quantity of resident proteins and cargo proteins undergoing lysosomal degradation. Interestingly, we also found that lysosomal secretion contributes to the SASP in a RAB27A-dependent manner in melanoma senescent cells. These data constitute a useful resource to further elucidate the interplay between autophagy and senescence, and to find new biomarkers and vulnerabilities in senescent cells.

| Increased lysosomal biogenesis in senescent cells
To study lysosomal function in senescent cells, we began by confirming the expansion of the lysosomal compartment in cellular models of senescence. Melanoma SK-MEL-103 cells treated with the CDK4/6 inhibitor, palbociclib, displayed SAβGal staining and increased protein levels of the lysosomal membrane proteins LAMP1 and LAMP2 (Figure 1a). This increase was particularly pronounced in the case of LAMP2A, the only spliced variant of the Lamp2 CCL3, CXCL12, cathepsin CTSD, or the protease inhibitor SERPINE1, are secreted in a RAB27A-dependent manner in senescent melanoma cells. Finally, proteins previously identified as plasma biomarkers of aging are highly enriched in the lysosomes of senescent cells, including CTSD. We conclude that the lysosomal proteome of senescent cells is profoundly reconfigured, and that some senescent cells can be highly active in lysosomal exocytosis.

K E Y W O R D S
aging, autophagy, cellular senescence, exocytosis, lysosome, SASP gene required for CMA and a limiting component for this type of autophagy (Cuervo & Dice, 2000a). Luminal lysosomal hydrolases, such as β-glucocerebrosidase (GBA) and cathepsin D (CTSD), were also increased in senescent SK-MEL-103 cells (Figure 1b). Similar findings were observed in hepatocarcinoma Huh7 and osteosarcoma U2OS cells treated with palbociclib ( Figure 1c). To ascertain whether the expansion of lysosomes in senescence was accompanied by de novo lysosomal synthesis, we analyzed the expression of the transcription factor EB (TFEB), the master regulator of lysosomal biogenesis (Napolitano & Ballabio, 2016). Interestingly, TFEB mRNA levels dramatically increased over time after palbociclib addition, preceding an elevation in LAMP1 known to be part of the TFEBtranscriptional program (Figure 1d). We also found marked elevation of lysosomal components that are not under TFEB regulation, such as LAMP2A and LAMP2B (Figure 1e). Of note, transcriptional upregulation of the LAMP2A variant (~sixfold), which acts as a receptor for CMA (Cuervo & Dice, 2000b), was higher than the upregulation of LAMP2B (~2.5-fold), which does not play a specific role in CMA (Cuervo & Dice, 2000b) (Figure 1e). Together, these results indicate that senescent cells enlarge the lysosomal compartment due, at least in part, to de novo biogenesis of lysosomes.

| Macroautophagy (MA) is increased in senescent cells
Next, we wondered whether the expanded lysosomal compartment in senescent cells was indeed functional. In general, proteins degraded in lysosomes have long half-lives (Dice, 1987), and metabolic labeling pulse and chase experiments in cultured cells, using radiolabeled amino acids and lysosomal inhibitors, can be used as a good assessment of lysosomal degradative function (Kaushik & Cuervo, 2009). Thus, to study the lysosomal degradative capacity of senescent cells, we first treated SK-MEL-103 cells with palbociclib to induce senescence and then added 3 H-leucine to the media for 48 h to radiolabel de novo synthesized proteins (pulse period). After extensive washing, we measured the breakdown of radiolabeled proteins (chase period) as the release of free 3 H-leucine into the culture medium (Figure 2a). We found a significant increase in the degradation rate of long-lived proteins in senescent cells compared to their non-senescent counterparts. Moreover, these differences were ablated upon addition of inhibitors of lysosomal proteolysis consisting of ammonium chloride (NH 4 Cl), a weak base that neutralizes the lysosomal acidic pH required by many lysosomal enzymes, F I G U R E 1 Characterization of the lysosomal system in senescence. (a) Representative SAβGal staining pictures of SK-MEL-103 cells treated with 2 μM palbociclib for 1 week. Scale bar 50 μM. (b) Western blot analysis for LAMP1, LAMP2A, LAMP2, GBA, and CTSD in SK-MEL-103 cells. Arrows indicate precursor and mature CTSD. Ponceau staining is shown as loading control. (c) Western blot analysis for LAMP2A, GBA, and CTSD in Huh7 and U2OS cells treated with 2 μM palbociclib for 1 week. Ponceau staining is shown as loading control. (d, e) mRNA levels of TFEB, LAMP1 (d) and LAMP2A, LAMP2B (e) in SK-MEL-103 cells treated with 2 μM palbociclib at the indicated timepoints. ACTB, 18S rRNA, and B2M were used for input normalization (mean of the three housekeepers). Values are relative to control cells and are expressed as mean ± SD, and statistical significance was assessed by one-way ANOVA and Dunnett's multiple comparisons test (versus control group). *p < 0.05 F I G U R E 2 Upregulation of MA in senescent cells. (a) Scheme of the assay to determine proteolytic rates of long-lived proteins in control and palbociclib-treated cells and analysis of lysosomal contribution to degradation. (b) Protein degradation rates of long-lived proteins in control and 7 days palbociclib-treated SK-MEL-103 cells. Values are mean ± SEM and are expressed as percentage proteolysis. n = 3 in two different experiments. Statistical significance was assessed by two-way ANOVA and Sidak's multiple comparisons test. Differences among time points were significant for p < 0.0001 and between control and palbociclib-treated cells (shown in the legend) for p < 0.05. (c) Percentage of proteolysis of long-lived proteins in the cells in (b) after 24 h of culture without additions (None) or in presence of ammonium chloride and leupeptine (N/L). Statistical significance was assessed by two-way ANOVA and Tukey's multiple comparisons test (versus control group). Differences between control and palbociclib-treated cells and between None and N/L are shown in the figure.
(d) Representative images of control and 7 days palbociclib-treated SK-MEL-103 cells stably transduced with a lentivirus expressing the tandem reporter mCherry-GFP-LC3 to monitor autophagic flux. Insets: higher magnification of merged channels or red channel. Nuclei are highlighted with DAPI. (e) Quantification of autophagic vacuoles (AV, combination of autophagosomes (APG) and autolysosomes (AUT)) (left), APG (mCherry + GFP + vesicles) (middle) and AUT (mCherry + GFP − vesicles) (right) in cells as in (d. Values are expressed as puncta per cell section and are individual values and mean ± SEM. n = 6 experiments with >1200 cells/condition. Statistical significance was assessed by the two-tailed Student's t-test (versus control group). (f-i) Representative immunoblot for the indicated proteins (f) and densitometric quantification of steady-state values (left) or lysosomal degradation for LC3-II (g), p62 (h), and NBR1 (i) in SK-MEL-103 cells at the indicated days after addition of palbociclib (2 μM) to the media. Cells were supplemented with ammonium chloride and leupeptin (N/L) for 4 h where indicated (+). Ponceau red staining is shown as loading control. Quantifications are shown as folds of the cells no supplemented with palbociclib (Control) and values are mean ± SEM (n = 3 experiments). Statistical significance was assessed by the one-way ANOVA and Tukey's multiple comparisons test (versus control untreated). Significant differences by ANOVA are indicated in the graph and differences with time 0 at specific time points in the graph. and leupeptin, a potent inhibitor of cysteine and serine peptidases (the two components abbreviated as N/L) (Figure 2b,c). Therefore, we conclude that the higher rates of protein degradation in senescent cells mostly reflect their increased degradation in lysosomes.
Previous studies have reported increased levels of macroautophagy (MA) in replicative senescence and in oncogene-induced senescence (Gamerdinger et al., 2009;Narita et al., 2011;Young et al., 2009). We monitored MA in SK-MEL-103 cells using a tandem fluorescent LC3 construct (mCherry-GFP-LC3) (Kimura et al., 2007).  Photoswitching of Dendra fluorescence from green to red allows to distinguish the photoconverted reporter (red, pseudocolored here in yellow, in consideration to color-blind readers) as it is delivered to CMA lysosomes over the background cytosolic signal of the newly synthesized reporter (green). Transduction of SK-MEL-103 cells with this reporter revealed a significant increase in the number of fluorescent puncta in senescent cells (Figure 3a,b). We confirmed that the increase in CMA was not related directly to the presence of palbociclib, as addition of the drug for 1 day to control cells or re-addition of palbociclib for 1 day to already senescent cells did not upregulate CMA activity (Figure 3c). These results support that CMA upregulation was not a direct consequence of CDK4/6 inhibition but rather a feature of senescence. In fact, a time-course analysis revealed that  Figure S1A) induced by each of the treatments. Therefore, we conclude that senescent cells upregulate CMA degradative pathways.

| Proteomics of senescent lysosomes
We wondered whether the increased lysosomal activity of senescent cells is accompanied by changes in the proteins undergoing degradation (delivered into lysosomes by CMA or MA), and in constitutive resident lysosomal proteins (both lumen, membrane integral, and membrane-associated proteins). For this, we isolated lysosomes from control and palbociclib-induced senescent SK-MEL-103 cells using a previously described method based on differential centrifugation and flotation in gradients of discontinuous  (Kaushik et al., 2008). Analysis of the lysosomal fractions (CMA and MA) isolated from control and senescent cells revealed comparable total protein levels and specific activity of enzymes such as hexosaminidase ( Figure S2A,B), in further support that lysosomes from senescent cells were able to efficiently degrade the increased amount of cargo delivered by MA and CMA to these compartments. Furthermore, we did not find significant differences between the lysosomal fractions isolated from control and senescent cells in lysosomal recovery (percentage of total cellular hexosaminidase activity recovered in the lysosomal fractions), lysosomal purity (enrichment of hexosaminidase activity in the lysosomal fractions relative to total cellular activity) or in the integrity of the lysosomal membrane (percentage of hexosaminidase activity detected outside lysosomes because of breakage of their membranes) ( Figure S2C-E). Integrity of the lysosomal fraction (breakage <15%) is an essential requirement to accurately assay CMA and MA, and consequently, only isolations showing hexosaminidase release below those levels were used for these experiments ( Figure S2F). Interestingly, comparison of the proteolytic capacity of the luminal content of lysosomes isolated from senescent cells, once corrected per amount of protein, was not higher than in lysosomes isolated from control cells ( Figure S2F), suggesting that the increase in MA and CMA flux in senescent cells was not a consequence of faster luminal proteolysis but rather of higher delivery of cargo to lysosomes.
After validating our isolation method, we analyzed the lysosomal proteome (cargo and resident proteins) by mass spectrometry (MS). Of note, MS of lysosomes allows for the detection of resident lysosomal proteins, and it can also detect peptides derived from lysosomal substrates that have been degraded inside the lysosomes during the hours immediately prior to isolation. We treated control and senescent SK-MEL-103 cells with N/L for 16 h to block lysosomal proteolysis before lysosomal isolation in order to distinguish lysosomal substrates from resident lysosomal proteins. This allowed us to classify the proteins that increase in the lysosomes upon N/L treatment as lysosomal substrates, and the proteins that do not change or decrease upon N/L as resident lysosomal proteins, as previously described (Schneider et al., 2014) ( Figure 4b). We performed three biological replicates on separate days, and each followed an independent purification process.
After proteomic analysis by MS, we identified more than 3,400 F I G U R E 3 Upregulation of CMA in senescent cells. (a) Representative images of control and 7 days palbociclib-treated SK-MEL-103 cells stably transduced with a lentivirus expressing a KFERQ-dendra reporter to monitor CMA activity. Cells were photoswitched, and fluorescence in the red channel (pseudocolored in yellow, in consideration for color-blind readers) was monitored using high-content miscroscopy. Insets: higher magnification of red channel images. Nuclei are highlighted with DAPI. (b) Quantification of changes in the mean number of puncta per cell section quantified with high-content microscopy. All values are individual values and mean ± SEM of six separate wells for each condition, and quantifications were done in at least 2500 cells per condition. Statistical significance was assessed by the two-tailed Student's t-test. (c) Quantification of changes in the mean number of puncta per cell section of control and 7 days palbociclibtreated SK-MEL-103 cells 1 day after the addition of a single doses of palbociclib. All values are individual values and mean ± SEM of six separate wells for each condition, and quantifications were done in at least 2500 cells per condition. Statistical significance was assessed by the two-way ANOVA and Sidak's multiple comparisons test (versus control group). (d) Time course of CMA activity calculated in SK-MEL-103 cells expressing the KFERQ-dendra reporter at the indicated times after addition of pablobiclib. Values are expressed as puncta per cell section and are mean ± SEM of three separate wells for each condition, and quantifications were done in >2500 cells per condition. Statistical significance was assessed by the one-way ANOVA and Dunnett's multiple comparisons test. (e) Representative images of NIH 3T3 cells stably transduced with a lentivirus expressing a KFERQ-dendra reporter to monitor CMA activity 7 days after the indicated treatments. Cells were photoswitched and imaged with high-content microscopy as in (a). Merged images of fluorescence in the green and red (pseudocolored in yellow) channels are shown. Nuclei are highlighted with DAPI. (f) Quantification of changes in the mean number of puncta per cell section in cells in (e). All values are individual values and mean ± SEM. Quantifications were done in >600 cells/condition in 3 independent experiments. Statistical significance was assessed by the one-way ANOVA and Dunnett's multiple comparisons test. (g) Representative images of primary human fibroblasts (IMR90 cells at population doubling level (PDL) 20) transduced with a lentivirus expressing a KFERQ-dendra reporter to monitor CMA activity 7 days after exposure to palbociclib. Cells were photoswitched and imaged by high-content microscopy. Bottom: Quantification of changes in the mean number of fluorescent puncta per cell. All values are mean ± SEM and individual values of >50 cells/condition in 3 independent experiments. Statistical significance was assessed by the unpaired t-test. (h) Similar studies as in (g) but using primary mouse embryonic fibroblasts (MEFs) and high-content microscopy. Red channel (top) and merged green and red channels (bottom) are shown. Bottom: Quantification of changes in the mean number of fluorescent puncta per cell (left) or in the percentage of cellular area occupied by fluorescent puncta (right). All values are mean ± SEM and individual values. Quantifications were done in >100 cells/condition in 10 independent wells. Statistical significance was assessed by the unpaired t-test. (i) Similar studies as in (g) but using a neuroblastoma cell line. Examples of two cells for each condition in the red channel (pseudocolored in yellow) are shown. Bottom: Quantification as in (h) was done in >1000 cells/condition in 10 independent wells. Statistical significance was assessed by the unpaired t-test. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. proteins in the CMA lysosomes, and a similar amount was detected in the MA lysosomes. Principal component analysis (PCA) revealed a consistent clustering of the different biological entities.
The first component that accounted for 54% of the differences was the subtype of lysosome, that is, CMA or MA. The second component (11% of variation) separated samples according to the state of cells, that is, non-senescent (CTRL) or senescent (SEN) ( Figure 4c). We conclude that the autophagic pathway (CMA or MA) and the cellular state (non-senescent or senescent) are associated with a specific protein composition of lysosomes.

| Lysosomal resident proteins in senescent cells
Lysosomes from senescent cells presented remarkable changes in the levels of resident proteins (i.e., those that are not degraded and therefore do not accumulate upon N/L treatment). There was a general direct correlation between the changes observed in CMA and MA lysosomes upon senescence (4d,e). For example, we identified 435 resident proteins that were enriched in both CMA and MA lysosomes from senescent cells compared to lysosomes from control cells (Figure 4d,e; Table S1). The top highest enriched  Figure S3). This is consistent with previous observations reporting the enhanced association of mTOR with the lysosomes of senescent cells (Narita et al., 2011). Other proteins were markedly reduced in both lysosomal populations in senescent cells, as it was the case of RhoGTPases, known to participate in lysosomal positioning ( Figure S3). Amino acid metabolism was also increased in senescent CMA and MA lysosomes ( Figure S4). We also identified changes specific for each of the lysosomal subgroups. For example, CMA lysosomes in senescent cells showed a marked increase in enzymes involved in fatty acid metabolism, and in proteins that participate in phagocytosis and in vesicular fusion, whereas cytoskeleton and endocytosis-related proteins were reduced ( Figures S3 and S4A).
MA lysosomes from senescent cells displayed increased levels in enzymes involved in metabolism of amino acids and micronutrients, proteins in the vesicular transport group, as well as mitophagy components, whereas we identified a marked decrease in signaling proteins normally present in the surface of control cells lysosomes ( Figures S3 and S4B).

| Lysosomal protein substrates in senescent cells
We next focused on lysosomal protein substrates (i.e., those that are degraded and therefore accumulate upon N/L treatment). We refer to the ratio between N/L and vehicle conditions as degradation ratio (which encompasses lysosomal delivery and proteolysis inside lysosomes) and for each substrate protein we generated a degradation ratio in senescent (SEN) and control (CTRL) cells. Finally, these two degradation ratios were divided (SEN vs. CTRL) to obtain the "degradation fold change" (Table S2). For CMA lysosomes, we found 333 proteins that displayed enhanced lysosomal degradation in senescent cells and 225 proteins whose lysosomal degradation was reduced ( Figure 6a). For MA lysosomes, we found 701 proteins with increased lysosomal degradation rates in senescent cells and 335 proteins that displayed lower degradation rates in these lysosomes ( Figure 6a; Table S2). Since we did not find differences in the proteolytic capacity of the luminal proteases per se once the lysosomal membrane was disrupted ( Figure S2F), we interpret that the differences in degradation rate of specific proteins mostly reflect their rate of delivery to lysosomes. The top proteins with increased or decreased degradation in each type of lysosomes in senescent cells are shown in Figure 6b. Besides changes in degradation rate, we also identified proteins no longer degraded by MA (37) or CMA (83) in senescent cells, as well as proteins degraded through these pathways (165 for MA and 47 for CMA) only in senescent cells but not in control cells ( Figure S5). We conclude that senescence entails extensive changes in lysosomal substrates, which are likely consequence of changes in their delivery rates.
To understand the consequences of these quantitative and qualitative changes in autophagy-mediated degradation in senescence, we first focused on the lysosomal substrates for MA. We analyzed our dataset by gene set enrichment analysis (GSEA) using the c5 Gene Ontology (GO) terms database. Interestingly, we found a significant enrichment in proteasomal machinery gene sets ( Figure 6c; Table S2). In relation to this, MA is known to contribute to proteasome degradation under stress conditions such as starvation (Cohen-Kaplan et al., 2016;Cuervo et al., 1995). We also found an increased degradation rate via MA for translation initiation factors, particularly for several subunits of the eukaryotic translation initiation factor 3 (eIF3) and components of amino acid metabolism pathways including tRNA synthetases (Figure 5c; Table S2). This suggests that senescent cells have a high turnover of the protein translation and quality control machineries, probably due to aberrant protein accumulation. Among the group of proteins not normally degraded by MA, but that become MA substrates in senescent cells, we found a large number of proteins related to the organization of the actin and microtubule cytoskeletons ( Figure S5A), which could be behind the reorganization of the cytoskeleton recently described to occur in senescent cells (Moujaber et al., 2019).
Among the protein categories undergoing less degradation by MA during senescence, we found endoplasmic reticulum (ER) components and iron-binding proteins (Figure 6d; Figure S5B; Table S2).
This suggests that the ER turnover in senescent cells by MA is lower as compared to control cells, which is consistent with the secretory nature of senescent cells. Also, in line with the changes in secretion in senescent cells, we identified a subset of proteins involved in constitutive exocytosis (through Golgi transport vesicles that continuously dock in the plasma membrane) that become MA substrates in senescent cells, whereas a different protein subset involved in regulated exocytosis (mediated by secretory vesicles only released in response to a extracellular signal) usually degraded by MA was spared from degradation in senescent cells ( Figure S5A,B). These findings support a possible regulatory role for MA in the secretory function and SASP of senescent cells.
We next studied the senescence-induced changes in the degradation of protein substrates for CMA. Interestingly, GSEA also showed an enrichment in proteasome components under the category of "Regulation of cellular amino acid metabolism" in the substrates degraded more via CMA in senescence, along with zincbinding proteins (Figure 6e; Table S2). Degradation of specific proteasome subunits by CMA has also been described to play a regulatory role in overall proteasome content and activity (Cuervo et al., 1995;Juste et al., 2021;Schneider et al., 2015). We also analyzed gene sets that displayed significantly downregulated degradation by CMA and found an enrichment in cell-cell junction organization components ( Figure 6f; Table S2). Lysosomes are known to regulate cell-cell junctions (Nighot & Ma, 2016), and cell senescence has been associated with disruptions in tight and gap junctions (Krouwer et al., 2012;Xie et al., 1992). Our results suggest that alterations in cellular junctions might be controlled by CMA lysosomes. Lastly, we noticed a large pool of mitochondrial proteins that become CMA substrates in senescent cells, whereas another subset of mitochondrial proteins is spared from degradation through this pathway (Figure 6f; Figure S5C,D). Although CMA cannot degrade mitochondria as a whole organelle, recent studies have demonstrated regulatory degradation of nuclear-encoded mitochondrial proteins by CMA before they reach mitochondria (Schneider et al., 2014). We propose that the observed changes in CMA activity and in the type of proteins degraded through this pathway in senescent cells may contribute in part to the remodeling of the mitochondrial proteome described in these cells (Catherman et al., 2013;Sabbatinelli et al., 2019;Sullivan et al., 2021).

| Lysosomal secretion of SASP factors by senescent cells
Considering the high levels of lysosomes in senescent SK-MEL-103 cells, we wondered if senescent cells could be able to perform lysosomal secretion (Settembre et al., 2013). Lysosomal secretion results in the exposure of lysosomal membrane proteins at the plasma membrane. We determined the presence of LAMP1 and LAMP2 by FACS in alive, non-permeabilized, control, and senescent cells. Consistent with lysosomal secretion, senescent cells presented high levels of LAMP1 and LAMP2 at the plasma member (Figure 7a). To study the mechanism of lysosomal exocytosis in senescence, we focused on RAB27A, a small GTPase that plays a key role in non-canonical vesicle secretion, including lysosomes (Bahadoran et al., 2001;Haddad et al., 2001;Johnson et al., 2013;Wilson et al., 2000) and extracellular vesicles (EV) (Ostrowski et al., 2010;Peinado et al., 2012).
Interestingly, our proteomic analysis detected RAB27A as a resident protein in both MA and CMA lysosomes (levels remained unchanged upon N/L treatment; Figure S6A). First, we used a short hairpin RNA targeting RAB27A in SK-MEL-103 cells and confirmed efficient reduction of RAB27A at the protein level ( Figure S6B). We verified that RAB27A-KD cells can efficiently undergo senescence, as evidenced by positive SAβGal staining seven days after palbociclib treatment ( Figure 7b). Finally, we checked for the presence of lysosomal proteins in the conditioned medium (CM) of senescent cells, either wild type (WT) or RAB27A-KD. We found that the lysosomal lumen proteins CTSD and GBA were increased in the CM from senescent cells, while they were undetectable in the CM of RAB27A-KD senescent cells (Figure 7c). To corroborate that RAB27A specifically affects non-canonical exocytosis, we also analyzed the levels of ferritin, which accumulates in senescent cells (Masaldan et al., 2018) and can be secreted via the canonical ER-Golgi secretory pathway (Ghosh et al., 2004). Accordingly, ferritin heavy chain 1 (FTH1) was increased in the CM of senescent cells, regardless of the presence or absence of shRAB27A ( Figure 7c). As an additional control, the levels of CTSD, GBA, and FTH1 in whole-cell lysates were similar in WT and RAB27A-KD senescent cells, and higher than in WT nonsenescent cells (Figure 7c). These findings were replicated using a pool of siRNAs targeting multiple sequences of RAB27A ( Figure S6C).

F I G U R E 6
Mass spectrometry analysis of lysosomal substrate proteins. (a) Proteins degraded at higher or lower rates in CMA and MA lysosomes from senescent cells. Substrate proteins were defined as those that accumulate significantly upon N/L treatment, that is, log2 fold change N/L versus vehicle >0.21 (fold >1.1) and p value <0.05. Red, substrate proteins that display increase degradation rates in senescence were defined as those either exclusively degraded in senescent cells, or degraded in both, but with a degradation fold change ≥0.6. Dark green, substrate proteins degraded with a similar rate in both conditions (degradation fold change <0.6 and >−0.6). Light green, substrate proteins that display lower degradation in senescence were defined as those exclusively degraded in control, or with a degradation fold change ≤−0.6. (b) Top lysosomal substrate proteins displaying higher or lower rates of lysosomal degradation in CMA and MA lysosomes.  Figure S6D). We wondered whether RAB27A inhibition could affect the mRNA levels of CCL2 and SERPINE1 in senescent cells. However, the mRNA levels of these SASP factors were highly elevated in senescent cells regardless of the inhibition or not of RAB27A ( Figure S6E). The translation of many cytokines in senescent cells depends on the characteristic high levels of mTORC1 and its downstream effector phospho-4EBP (Herranz et al., 2015;Laberge et al., 2015). Again, knockdown of RAB27A mRNA with siRNAs did not affect the high levels of phospho-4EBP1 in senescent cells ( Figure S6F). Overall, our results indicate that the secretion of some SASP components in senescent melanoma cells occurs, at least in part, via secretory lysosomes in a RAB27A-dependent manner. F I G U R E 7 Analysis of lysosomal secretion in senescence. (a) Flow cytometry analysis of LAMP1 and LAMP2 in the surface of alive, non-permeabilized SK-MEL-103 cells, treated or not with 100 nM doxorubicin or 5 μM palbociclib for 1 week. Top, values for three independent biological replicates (n = 3), mean ± SD. Due to sample limitation, n = 2 in cells treated with doxorubicin and stained for LAMP1. Statistical significance was determined using the unpaired t-test.

| Plasma markers of aging present in the lysosome of senescent cells
It has been recently reported that 25 SASP proteins shared by several types of senescence (Basisty et al., 2020) are also present in a catalogue of human plasma proteins significantly upregulated with aging (Tanaka et al., 2018). We wondered if these 25 "SASP-andaging" proteins could result from lysosomal exocytosis by senescent cells. We generated a ranked list of all the proteins enriched in senescent lysosomes compared to control cells, without taking into account whether the proteins are resident or substrates (Table S3).
Interestingly, among the 25 "SASP-and-aging" markers previously defined (Basisty et al., 2020), 9 proteins (36%) were enriched in the lysosomes of senescent cells (CMA or MA lysosomes or both) ( Table 1). Most of these 9 proteins are typical resident lysosomal proteins including four proteases, namely cathepsins CTSB, CTSD, CTSZ, and metalloproteinase MMP2 (Table 1). These observations suggest that some serum biomarkers of aging may reflect the process of lysosomal secretion by senescent cells.

| DISCUSS ION
One of the main features of senescent cells is their enlarged lysosomal compartment. Here, we have confirmed the previously de-  (Lawrence & Zoncu, 2019), but to the best of our understanding, this is the first report of pronounced changes in these lysosome-signaling pathways in senescent cells. On this respect, we found higher association of several mTOR components to lysosomes in senescent cells, consistent with previous observations reporting an association between mTOR and autolysosomes in senescence, which facilitates protein synthesis (Narita et al., 2011). In contrast to this increase in mTOR components, our proteomic analysis revealed significant reduction in other major pathways that signal from the lysosome such as phospholipase D, Ras, or Hippo pathways ( Figure S3B). Future studies on the functional consequences of these changes could provide novel insights on the contribution of lysosomal signaling to the senescent phenotype.
The activation of MA has been reported in replicative senescence and also during oncogene-induced senescence, where it plays a key role in SASP induction (Gamerdinger et al., 2009;Narita et al., 2011;Young et al., 2009). In the senescence paradigm used TA B L E 1 Lysosomal proteins upregulated in senescent cells in our study that are age-upregulated serum SASP factors (Basisty et al., 2020)  In this regard, studies in yeast have shown that eIF3 can assemble into a large supercomplex, the translatome, which contains elongation factors, tRNA synthetases, 40S and 60S ribosomal proteins, chaperones, and the proteasome (Sha et al., 2009). Further studies are needed to elucidate the link between protein synthesis and degradation in senescence. MA in senescent cells may also mediate cellular structural changes, such as remodeling of the cytoskeleton and turnover of the proteasome. Of note, impaired proteasome function has been tightly correlated to senescence and aging (Chondrogianni et al., 2003;Chondrogianni & Gonos, 2004, 2005Stratford et al., 2006). we favor more a regulatory function for the changes in autophagy during senescence. For example, increased lysosomal degradation of specific proteasome subunits, such as PSMD5, may reduce its cellular levels and thus prevent the previously described ameliorating effect of PSMD5 on replicative senescence (Lu et al., 2014).

MA-lys
Coordinated changes during senescence in protein degradation and synthesis may allow senescent cells to transition through the different stages of senescence.
Senescent cells can strongly affect the microenvironment and promote tissue remodeling through the secretion of proinflammatory cytokines, chemokines, growth factors, and proteases, known as the senescence-associated secretory phenotype or SASP (Coppé et al., 2008). Some cell types have the capacity to perform lysosomal exocytosis as part of their normal physiology, in the case of osteoclasts, melanocytes, endothelial cells, and cells from the hematopoietic lineage, including lymphocytes, neutrophils, mast cells, and macrophages (Blott & Griffiths, 2002;Mostov & Werb, 1997;Settembre et al., 2013;Stinchcombe et al., 2004). We observed that a significant number of previously reported SASP proteins (Basisty et al., 2020), which have also been described as human aging biomarkers in plasma (Tanaka et al., 2018), are enriched in the lysosomes of senescent cells. In fact, most of these proteins are lysosomal, such as the aspartyl protease, CTSD. Our finding that secretion of these lysosomal components in melanoma cells is dependent on RAB27A, a protein that we found associated with both lysosomal populations analyzed in this study, highlights this GTPase as a potential future target to modulate SASP in aging. Of note, we also observed that some cytokine characteristic of the SASP are also secreted in a RAB27A-dependent manner. In particular, a cytokine array screening identified that the secretion of CCL2, CCL3/4, CXCL12, MIF, and Overall, we have observed that senescent cells upregulate CMA together with MA, and that the protein content in the lysosomes changes significantly upon senescence induction. We have also shown that lysosomal secretion may contribute to the SASP.
Finally, our comprehensive catalogue of lysosomal proteins can be used to further elucidate the interplay between autophagy and senescence, and to identify novel in vivo biomarkers secreted by senescent cells.

| Intracellular protein degradation assay
Measure of intracellular degradation was performed in SK-MEL-103 cells by metabolic labeling with 2 μCi/ml 3 H-leucine (PerkinElmer) for 48 h at 37°C and pulse-chase experiments as described previously (Auteri et al., 1983). After labeling, cells were extensively washed and maintained in medium with an excess of unlabeled leucine.
Aliquots of the medium were taken immediately after washing and at different times for 24 h and were precipitated in trichloroacetic acid.
Proteolysis was measured as the amount of acid-precipitable radioactivity transformed in acid-soluble radioactivity at each time point.

| Macroautophagy flux
Macroautophagic flux was measured in protein lysates using immunoblot for LC3-II, p62, or NBR1 in cells treated or not with lysosomal protease inhibitors (10 mM ammonium chloride and 100 μM leupeptin). Flux was calculated as the increase in levels of LC3-II in protease inhibitors-treated cells relative to untreated cells.

| Cytokine array
Cytokine levels in conditioned medium (CM) were analyzed using the Proteome Profiler Human Cytokine Array (R&D Systems; #ARY005B), following the manufacturer's instructions. Pixel density was determined using the ImageStudio Software.

| Analysis of mRNA levels
Signal from each antibody was normalized using housekeeping proteins (ACTB or TUBG1) as indicated.

| Mass spectrometry
Lysosome samples were solubilized to a final concentration of 5% or <−1 were defined as regulated. FDR was estimated to be <5% by Benjamini-Hochberg. GSEA was performed using the GSEA software v2.0.6 from the Broad Institute. Differentially expressed proteins were ranked according to their log2 ratio and used as input for the enrichment analysis. All fields were set to default, and only gene sets significantly enriched at a FDR q-values <0.25 were considered. GO analysis was performed using the STRING functional protein association networks (https://strin g-db.org/) and Reactome  Autofluorescence signal from unstained samples was obtained and subtracted for every sample. Data were analyzed using FlowJo v10 software.

| SAβ Gal activity assay
SAβGal staining was performed using the Senescence β-Galactosidase Staining Kit (Cell Signaling) following the manufacturer's instructions. Briefly, cells were fixed at room temperature for 2 min with a solution containing 2% formaldehyde and 0.2% glutaraldehyde in PBS, washed three times with PBS, and incubated overnight at 37°C with the Staining Solution containing X-gal in DMF (pH 6.0).

| Statistical methods
At least three independent replicates were assayed to ensure reproducibility in cell culture experiments. Statistical significance was assessed as appropriate using either the Student's t-test (two-tailed, paired, or unpaired), the Fisher exact test, or the one-way ANOVA

ACK N OWLED G M ENT
We thank Dr. Susmita Kaushik for experimental advice and critical feedback.

CO N FLI C T O F I NTE R E S T
A.M.C. is co-founder of Selphagy Therapeutics (now under LifeBiosciences, Boston, MA, USA) and consults for Generian Therapeutics and Cognition. M.S. is shareholder and advisor of Senolytic Therapeutics, Inc., Rejuveron Senescence Therapeutics, AG, and Altos Labs, Inc; and is shareholder of Life Biosciences, Inc.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.