CDK4 regulates lysosomal function and mTORC1 activation to promote cancer cell survival

Abstract

Cyclin-dependent kinase 4 (CDK4) is well-known for its role in regulating the cell cycle, however, its role in cancer metabolism, especially mTOR signaling, is undefined. In this study, we established a connection between CDK4 and lysosomes, an emerging metabolic organelle crucial for mTORC1 activation. On the one hand, CDK4 phosphorylated the tumor suppressor FLCN, regulating mTORC1 recruitment to the lysosomal surface in response to amino acids. On the other hand, CDK4 directly regulated lysosomal function and was essential for lysosomal degradation, ultimately regulating mTORC1 activity. Pharmacological inhibition or genetic inactivation of CDK4, other than retaining FLCN at the lysosomal surface, led to the accumulation of undigested material inside lysosomes, which impaired the autophagic flux and induced cancer cell senescence in vitro and in xenograft models. Importantly, the use of CDK4 inhibitors in therapy is known to cause senescence but not cell death. To overcome this phenomenon and based on our findings, we increased the autophagic flux in cancer cells by using an AMPK activator in combination with a CDK4 inhibitor. The cotreatment induced autophagy (AMPK activation), and impaired lysosomal function (CDK4 inhibition), resulting in cell death and tumor regression. Altogether, we uncover a previously unknown role for CDK4 in lysosomal biology and propose a novel therapeutic strategy to target cancer cells.

Introduction

Cyclin-dependent kinase 4 (CDK4) has a well-established role in cell cycle control (1) and CDK4-cyclin complexes are commonly deregulated in tumorigenesis (2). These complexes are of great interest as therapeutic targets, and the FDA has approved the specific CDK4/6 kinase inhibitors PD0332991 (palbociclib), LEE011 (ribociclib) and LY2835219 (abemaciclib) for treating advanced or metastatic hormone receptor (HR)-positive and HER2-negative breast cancer. Clinical studies using CDK4/6 inhibitors to treat other malignancies are being conducted (3).

Research from our group and others has shown that the role of CDK4 is not limited to the control of the cell cycle. Indeed CDK4 is also a major regulator of energy homeostasis (4–6) through E2F1-RB complex (7), AMPK (8) and IRS2 (9). Importantly, the CDK4 pathway has been shown to cross-talk with the mTOR pathway, which is a major regulator of cell growth and metabolism (10,11). CDK4/6 inhibition attenuates mTOR Complex 1 (mTORC1) activity in some cancer models (12,13), yet the effects of CDK4/6 inhibitors on mTORC1 seem to be cell-type specific since opposite results were observed in other cancer types (14). The exact mechanism underlying the CDK4-mTOR cross-talk in mammals is unknown, although in Drosophila it occurs via the phosphorylation of TSC2 (15). Given that mTOR activity is increased in numerous cancers and participates in the translational regulation of several oncogenic proteins, mTOR inactivation constitutes an attractive strategy for cancer treatment (16).

Lysosomes, considered for years as only the digestive system of the cell, have emerged as key effectors in cell metabolism, due to their role as platforms in the activation of mTOR pathway (17–19). mTORC1 is recruited to the surface of lysosomes in a complex amino acid (AA)-dependent manner (17). Among the multiple regulators of this process, we focused on FLCN, a tumor suppressor which functions as a complex with FNIP. The FLCN-FNIP complex interacts with Rag GTPases in the absence of AAs repressing their activity. When AAs are sensed, FLCN-FNIP complexes dissociate from Rag GTPases eliciting their activation. The activation of Rag GTPases is crucial for mTORC1 recruitment to lysosomes (20). Importantly, mTORC1 activation is also triggered by the accumulation of AAs in the lysosomal lumen (21). Therefore, alterations in the lysosomal function directly impact mTORC1 activity (22,23). Additionally, these organelles play roles in cell survival and cell proliferation, thus becoming emerging targets for cancer therapy (24–26).

In this study, we demonstrate that CDK4 is capable of modulating mTORC1 activity in a direct manner, through the phosphorylation of FLCN, and indirectly, by promoting lysosomal function. When CDK4 inhibitors are used, the lack of lysosomal function induces senescence in triple-negative breast cancer (TNBC) cells and impairs tumor growth in a mouse xenograft model. Moreover, a combination of AMPK activation and CDK4 inhibition was used in an attempt to trigger autophagy in conditions when lysosomes are dysfunctional and resulted in cell death and tumor regression. This finding is of high relevance in TNBC, a highly invasive and aggressive cancer type that does not have a clear therapeutic strategy yet (27).

Materials and Methods

Materials

LY2835219, PD0332991 and LEE011 were purchased from MedChem Express and used at 0.5 μM, unless otherwise indicated. R³ human IGF-1 (I11146, Sigma) was used at 30 ng/ml. Minimal Essential Media (MEM) Amino Acids Solution (50X, Gibco) was used at the indicated concentrations. Rapamycin (LC Laboratories) was kindly provided by Pedro Romero (UNIL) and used at the indicated concentrations. Bafilomycin A1 (BafA1) was purchased from Enzo Life Sciences (ALX-380-030-M001) and used at 0.3 μM. Cell permeable αKG analog DMKG (349631, Sigma) was used at 5mM.

Cell culture and transfection

MDA-MB-231, CCRF-CEM, HTC116, IB115, HT29, SKOV, MCF7 and PC3 cell lines, all obtained from ATCC, were cultured in RPMI 1640 + GlutaMAX containing, 10mM HEPES, 1mM sodium pyruvate (All from Gibco) and 10% fetal bovine serum (FBS) (PAA Laboratories). Thawed cells were allowed one passage to reach exponential growth phase before being used. Cells were used during maximum of ten passages in the experiments performed in this study. PCR-based mycoplasma tests were done routinely using specific primers: 5’-TGCACCATCTGTCACTCTGTTAACCTC-3’ and 3-GGGAGCAAACAGGATTAGATACCCT-5’, the last test was performed in June 2019.

MDA-MB-231 CDK4 knockout (KO), E2F1 KO and FLCN KO stable cell lines were generated with CRISPR/Cas9 technology. The lentiCRISPR v2 plasmid was a gift from Feng Zhang (MIT; Addgene plasmid # 52961)(28). The pMD2.G plasmid was a gift from Didier Trono (EPFL; Addgene #12259). The pCMV-dR8.91 plasmid and the guide RNA for FLCN were gifts from Christian Widmann (UNIL). The target sequences for the guide RNAs are shown in Supplementary Table S1.

Synthetic oligonucleotides were purchased and cloned into the digested LentiCrispR vector as described in Shalem et al (28). Lentiviral production was based on the standard protocol established by Salmon and Trono (29). The resulting lentivirus (2ml) was then used to infect MDA-MB-231 cells for 72 hours. Infected cells were selected with 5 μg/ml puromycin for five consecutive days. Western blotting was performed to ensure that the protein of interest was no longer expressed.

Plasmid transfection was performed using X-tremeGENE HP. pRK5-FLAG-FLCN was obtained from Addgene (Addgene #72290) (30). MIG-human RagC WT and MIG-human RagC S75N plasmids were kindly provided by Alejo Efeyan (CNIO). pCMV-R24C plasmid was a gift from Mariano Barbacid (CNIO).

For insulin pathway stimulation, cells were first cultured in FBS-free media for 15 h and subsequently treated with IGF-1 for 20 min before lysis or fixation.

For AA stimulation, after a 15 h treatment with FBS-free media with or without CDK4/6 inhibitor, the cells were incubated for two hours in KRBB media containing 111 mM NaCl, 25 mM NaHCO₃, 5 mM KCl, 2.5 mM CaCl2, 1 mM MgCl2, 25 mM HEPES, 25 mM glucose and dialyzed FBS with or without CDK4/6 inhibitor. For the last 20 min, 2× MEM amino acids solution and 2 mM glutamine were added to the cells.

For acute CDK4/6 inhibition (3 hours), LY2835219 was diluted in RPMI complete media at the indicated concentrations.

Western blotting

Western-blotting was performed as previously described (7). The band intensities were quantified using the Fiji image-processing package (31). The following antibodies were used: anti-CDK4 (H22), anti-CDK6 (C21) and anti-RB (C15) from Santa Cruz Biotechnology; anti-phospho Rb-S780 (D59B7), anti-phospho P70S6K-T389 (108D2), anti-4E-BP1 (53H11), anti-phospho AKT-T308 (244F9), anti-phospho AKT- S472/3 (D9E), anti-p70S6 kinase (49D7), anti-LC3B (polyclonal), anti-SQSTM1 (polyclonal), anti-AcetylCoA Carboxylase (ACC) (polyclonal), anti-phospho AcetylCoA Carboxylase (p-ACC), anti-ULK1 (D8H5) and anti-phospho ULK-S757 from Cell Signaling Technology; and anti-α-tubulin (DM1A) from Sigma.

Site-specific rabbit polyclonal antibody against phospho-FLCN was a gift from Kei Sakamoto (NIHS). It was generated by YenZym Antibodies by immunization and affinity purification with a phosphorylated peptide of the human sequence (CQMNSRMRAH-*S-PAEG-amide for pS62 FLCN, the prefix * denotes the phosphorylated residue).

Mass spectrometry

Protein samples were separated on SDS-PAGE gels. Bands corresponding to FLCN-GST or FLCN-FLAG were excised and digested with trypsin. Peptide mixtures were analyzed by HPLC-MS/MS. MS data were analyzed using Mascot software 2.6 (Matrix Science). Scaffold software (version 4.8.4, Proteome Software Inc.) was used to validate MS/MS-based peptide and protein identifications, and to perform dataset alignment. MsViz software (32) was used for comparison of sequence coverage and phosphorylation of FLCN protein. A full description of the methods can be found in Supplementary Methods.

Flow cytometry

Cells were trypsinized and stained with Annexin V-PE (BioLegend, 640907) according to the manufacturer’s protocol. The cells were analyzed on either a Gallios™ (Beckman Coulter) or LSR II (BD Biosciences) flow cytometer. For intracellular Ki67 staining, cells were fixed and permeabilized according to the manufacturer’s protocol (Biolegend). Cells were then incubated with anti-Ki67-FITC (Biolegend) for 30 min on ice. After washing twice with permeabilization buffer, cells were resuspended in PBS prior to analysis. For each sample, at least 10000 events were acquired.

For LysoTracker experiments, fluorescence was analyzed with an ImageStream III Flow Cytometer. After treating the cells, 100 nM Lysotracker Green DND-26 (Life Technologies) was added to live cells for 1h. The cells were then trypsinized and fixed with 4% PFA for 15 min at RT, washed twice with PBS, and resuspended with 2% FBS. Nuclear staining was performed with DAPI. 10000 events were acquired per condition at a magnification power equivalent to 60X.

Lysosomal intracellular activity assay kit (Cell-Based) (Biovision) was used according to the manufacturer’s instructions. 10000 events were acquired with an ACCURI C6 flow cytometer.

All flow cytometry analyses were performed with FlowJo software (Version 7.6.5).

Cathepsin B activity assay kit (fluorometric)

Cathepsin B activity was assessed with a fluorometric kit (ab65300, Abcam). Fluorescence was measured with a Tecan plate reader (Ex/Em = 400/505 nm).

Colorimetric detection of senescence-associated β galactosidase

For Senescence-associated β-galactosidase (SA-β-Gal) analysis, cells in culture were fixed with 2% PFA and 0.2% glutaraldehyde for five minutes at room temperature (RT), washed with PBS and stained with a solution containing 40 mM citric acidNa phosphate buffer, 5 mM K₄ [Fe(CN)₆] 3H₂O, 5 mM K₃ [Fe(CN)₆], 150 mM NaCl, 2 mM MgCl₂ and 1 mg/ml X-gal in distilled water for 15h at 37°C. After staining, the cells were washed twice with PBS and once with methanol, and the plates were allowed to air dry. Bright-field images were obtained with an upright light microscope (Zeiss) with a 20X objective. Positive cells and the total number of cells per field were counted manually.

For SA-β-Gal analysis of mouse tumors, fresh tissues were frozen embedded in OCT and stored at -80°C. Tumors were cut into 8-μm-thick sections and mounted onto glass slides. After air-drying for 30 min, the sections were fixed with 2% PFA and 0.2% glutaraldehyde and stained as previously. The sections were counterstained with 0.1% Nuclear Fast Red (Sigma).

Electron microscopy

For transmission electron microscopy, a standard fixation, embedding and staining protocol of mice tumors and cells was established. Ultrathin sections were prepared and imaged using a Philips CM100 at 80 kV acceleration voltage equipped with a TVIPS TemCam-F416 digital camera. For further details see the Supplementary Methods.

Animal studies

MDA-MB-231 cells were injected into the fourth mammary gland of eight-week-old female NSG mice (NOD.Cg-Prkdc^scidIl2rg^tm1wjl/Sz strain, The Jackson Laboratory). Tumor growth and body weight were measured twice per week until the tumor size reached 50 mm³ per mouse. Mice were randomized into two groups for the 8-days treatment. LY2835219 (75 mg/kg), formulated in 1% HEC in distilled water, was administrated orally. A-769662 (20mg/kg) was administrated by intraperitoneal injection (IP) in NaCl 0.9%. Body weight and tumor growth were monitored every other day. After 8-days treatment, the animals were anesthetized with isoflurane (3%) and sacrificed by cervical dislocation. Tumors from 10 mice per group were dissected and cut into two pieces; one piece was fixed in PFA for histological staining, and the other piece was snap-frozen in liquid N2 for protein and RNA analysis. Tumors from 5 mice per group were processed for transmission electron microscopy analysis and SA-β-Gal staining. All animal care and treatment procedures were performed in accordance with Swiss guidelines and were approved by the Canton of Vaud, Service de la Consommation et des Affaires Vétérinaires (SCAV) (authorization VD 2797.1).

Quantification and Statistical Analyses

The results are expressed as the means ± standard error of the mean (S.E.M.). Comparisons between 2 groups were performed with unpaired two-tailed Student’s t-tests, and multiple group comparisons were performed by unpaired one-way ANOVA and two-way ANOVA, both followed by Tukey’s test or otherwise indicated. All p-values below 0.05 were considered significant. Statistically significant values are represented by asterisks corresponding to *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001.

Supplementary Methods

Additional details about the methods used in this paper can be found in the Supplementary Methods and in Supplementary Tables S2 to S5.

Results

CDK4 activity is required for mTORC1 localization at lysosomes

CDK4/6 inhibition was previously shown to decrease mTORC1 activity in some models of human cancer (12–14). To determine the cell-type specificity of this cross-talk, we activated mTORC1 in eight human cancer cell lines with IGF-1 in the presence of CDK4/6 inhibitor LY2835219. The efficiency of CDK4/6 inhibition was measured by RB phosphorylation. Treatment with LY2835219 caused a decrease in p70S6K and 4E-BP1 phosphorylation (two well-known targets of mTORC1), both in the unstimulated and in the IGF-1 stimulated conditions (Supplementary Fig. S1A). Despite showing considerable mTORC1 inhibition, some cell lines showed only a mild decrease in AKT phosphorylation in the presence of LY2835219 (HCT116, IB115, MDA-MB-231, SKOV3, PC3), suggesting that the effects observed in mTORC1 activity were at least partially independent of decreased AKT signaling.

We focused our experiments on the TNBC cell line MDA-MB-231 since it was one of the most responsive to CDK4/6 inhibition. The treatment with PD0332991 and LEE011, two other CDK4/6 inhibitors, also resulted in decreased mTORC1 activity in this cell line (Supplementary Fig. S1B-C).

We next found that MDA-MB-231 wild-type (WT) cells treated with LY2835219, or CDK4 knock-out (KO) MDA-MB-231 cells showed impaired translocation of mTORC1 to the lysosomes in response to AAs, a key step for mTORC1 activation (Fig. 1A–B), which correlated with decreased phosphorylation of 4E-BP1, p70S6K and ULK (Fig. 1C-D). Inhibition of glutaminolysis has been shown to prevent lysosomal recruitment and subsequent activation of mTORC1 (33). The effects of LY2835219 could not be rescued, however, with α-ketoglutarate (αKG) stimulation, indicating that CDK4 inhibition or depletion does not affect mTORC1 translocation to the lysosomal surface by impairing glutamine metabolism (Supplementary Fig. S2A-B). Interestingly, MDA-MB-231 cells lacking E2F1 showed normal mTORC1 translocation to the lysosomal surface and increased mTORC1 activation (Supplementary Fig. S2C-F) but were still sensitive to CDK4 inhibition. This suggests that the effects of CDK4 on mTORC1 activity might be independent of E2F1 transcriptional activity.

Figure 1. CDK4 inhibition or depletion prevents the recruitment of mTOR to the lysosomal surface in response to amino acid stimulation.

A, Confocal immunofluorescence analysis showing mTOR colocalization with lysosomes (LAMP1 and mTOR staining) in WT and CDK4 KO MDA-MB-231 cells, with or without AA stimulation and in the presence of 0.5 μM of the CDK4/6 inhibitor LY2835219 or DMSO. B, Quantification of mTOR-LAMP1 colocalization from A using Pearson’s coefficient. At least 40 fields per treatment condition from 3 independent experiments were analyzed. C, Western blot showing levels of phospho-RB and total RB, CDK4, mTORC1 target genes (phospho-p70S6K and total p70S6K, phospho-ULK and total ULK, and 4E-BP1) in WT and CDK4 KO MDA-MB-231 cells with or without AA stimulation and with or without 0.5 μM LY2835219. α-tubulin was used as a loading control. D, Quantification of phospho-p70S6K normalized to α-tubulin from three independent experiments like C.

CDK4 regulates mTORC1 activity through phosphorylating FLCN

We hypothesized that CDK4 potentially regulates mTORC1 pathway through phosphorylation of one of its regulators. A bioinformatics search of the proteins of mTORC1 pathway identified proteins containing the putative phosphorylation site for CDK4: [ST]Px[KRP] using phosphosite.org (Supplementary Fig. S3A). The corresponding peptides were used for a kinase competition assay versus RB consensus peptide. The peptides belonging to 4E-BP1, LARP1 and FLCN, triggered a decrease of RB phosphorylation, indicating a competition with RB for CDK4 phosphorylation (Supplementary Fig. S3B). In vitro kinase assays using CDK4/CycD3 recombinant protein and glutathione S-transferase (GST) fusion of full-length FLCN followed by mass spectrometry revealed that FLCN could be phosphorylated by recombinant CDK4/CycD3 at four different sites (S62, S73, T227 and S571) (Fig. 2A). Importantly, S62, S73 and S571 phosphorylation sites were found when FLCN was immunoprecipitated from MDA-MB-231 cells upon AA and IGF-1 stimulation (Supplementary Fig. S3C). Radioactive in vitro kinase assays confirmed that CDK4/CycD3 can phosphorylate the full-length FLCN in vitro (Supplementary Fig. S3D). When FLCN S62, S73 and S571 were mutated to alanine (non-phosphorylatable forms), the intensity of the autoradiography bands was reduced, indicating a decrease in FLCN phosphorylation. This result was even more striking in the triple mutant (Supplementary Fig. S3D).

Figure 2. CDK4 regulates mTORC1 activity through phosphorylating FLCN.

A, Representation of the phosphorylation sites detected by mass spectrometry analysis in the in vitro kinase assay with FLCN-GST and CDK4/CyclinD3. Green rectangles represent protein coverage. Blue dots represent phosphorylation sites, and their size correlates with their abundance. B, Confocal immunofluorescence analysis showing the colocalization of overexpressed FLCN with lysosomes (LAMP1 and FLCN staining) in WT and CDK4 KO MDA- MB-231 cells, with or without AA stimulation and in the presence of 0.5 μM of LY2835219 or DMSO. C, Quantification of FLCN-LAMP1 colocalization from B using Pearson’s coefficient. At least 40 fields per treatment condition from 3 independent experiments were analyzed. D, Western blot of WT and FLCN KO MDA-MB-231 cells, treated with DMSO or LY2835219 for 3h. Phospho-FLCN and total FLCN, phospho-p70S6K and total p70S6K, phospho-RB and total RB, and α-tubulin as a loading control. E, Quantification of phospho-FLCN levels normalized to α-tubulin from D. F, Immunofluorescence of phospho-S6 from WT and FLCN KO MDA-MB- 231 after 20 minutes of AA stimulation, in the presence or absence of LY2935219. G, Quantification of phospho-S6 intensity from F.

Importantly, FLCN is retained at the lysosomes in the presence of AAs upon CDK4 inhibition or depletion (Fig. 2B-C), which is consistent with the absence of mTOR. To further study the effects of CDK4 on FLCN, we generated an MDA-MB-231 FLCN KO stable cell line. We used a phospho-specific antibody directed against phospho-S62 FLCN, one of the phosphorylation sites that we found in CDK4-FLCN in vitro kinase assays. Phosphorylation of FLCN decreases significantly in WT cells treated with LY2835219 for only 3 hours (Fig.2D-E), indicating an acute control of CDK4 over FLCN, potentially resulting in mTORC1 regulation. FLCN KO cells were still sensitive to mTORC1 inactivation by CDK4/6 inhibition, as shown by decreased p70S6K phosphorylation, but to a lesser extent than WT cells (Fig. 2D). Moreover, WT cells showed a 2-fold decrease in the phosphorylation of S6 when treated with LY2835219, whereas FLCN KO cells showed a milder, yet significant, reduction of levels of S6 ribosomal protein phosphorylation. S6 phosphorylation remained significantly higher in LY2835219-treated FLCN KO cells, compared to LY2835219-treated WT cells (Fig. 2F-G). Our results suggest that complementary mechanisms to FLCN phosphorylation account for the effects of CDK4 on mTORC1 activation.

Interestingly, the overexpression of the R24C CDK4 mutant, which abolishes the ability p16(INK4a) to inhibit CDK4 (34), highly increased phospho-p70S6K levels under FBS starvation conditions as compared with untransfected cells (Supplementary Fig. S4A-B). Increased pS62-FLCN/FLCN ratio was also observed with R24C overexpression (Supplementary Fig. S4A and S4C), indicating that the R24C mutation bypasses the effect of starvation on the activation of the mTORC1 pathway.

Additionally, the overexpression of RagC S75N mutant confers insensitivity to mTORC1 activity upon acute treatment with LY2835219 (Supplementary Fig. S4D-E). This reinforces the link between CDK4 and mTORC1, not only through FLCN, but also through other mechanisms of nutrient sensing.

CDK4 inhibition or depletion increases lysosomal mass

Lysosomal biogenesis is a biological process coordinated by the transcription factor TFEB, which is repressed by mTORC1. Under nutrient-rich conditions, TFEB is phosphorylated by mTORC1, causing its retention in the cytoplasm. By contrast, when mTORC1 is inactivated, unphosphorylated TFEB translocates to the nucleus and promotes the transcription of numerous genes encoding lysosomal and autophagic proteins (19). To test whether CDK4 inhibition or depletion affected TFEB target genes, we treated WT and CDK4 KO MDA-MB-231 cells with complete or serum-starvation media, with or without LY2835219. As expected, under starvation conditions, WT MDA-MB-231 cells showed increased expression of genes regulated by TFEB (Fig. 3A). Moreover, we found that CDK4/6 inhibition synergized with starvation to further increase the expression of those genes. CDK4 KO cells also presented increased expression of those genes under basal conditions, but no further increase was found upon CDK4/6 inhibition (Fig. 3A).

Figure 3. CDK4 inhibition or depletion increases lysosomal mass.

A, qPCR analysis showing the expression of transcription factor EB (TFEB) regulated genes that are involved in lysosomal or autophagic processes (cathepsins B and D and SQSTM1) in WT or CDK4 KO MDA-MB-231 cells in complete (+FBS) or serum-starvation media (-FBS), treated with or without 0.5μM LY2835219. Each condition was assessed in duplicates for three independent experiments. B, Representative flow cytometry images of WT or CDK4 KO MDA-MB-231 cells treated with or without LY2835219 and stained with LysoTracker Green DND-26 under the indicated conditions. C, Proportions of cells containing lysosomes, estimated by quantification of fluorescence from LysoTracker Green DND-26 stained cells from two independent experiments. At least 10000 events were acquired. Data are subdivided into categories of the number of lysosomes per cell. Significant differences between WT and KO MDA-MB-231 cells are indicated by: * total lysosomal particles (P<0.05); $ lysosomal particles per cell (3-5 particles) (P<0.05); & lysosomal particles per cell (>5 particles) (P<0.05); Two-way ANOVA followed by Tukey’s multiple comparisons test. D, As for panel C, but subdivided into categories of the sizes of lysosomes per cell. Significant differences between WT and KO MDA-MB-231 cells are indicated by: * lysosomal particles <0.5μm² (P<0.05); $ lysosomal particles per cell (<0.5 μm²) (P<0.05). E, Immunofluorescence analysis showing LAMP1 and CDK4 expression in WT and CDK4 KO MDA-MB-231 cells in complete media. F, Quantification of the total volume of LAMP1-positive particles in z-stack images. At least 30 cells in total from three independent experiments were analyzed.

We next used LysoTracker staining and flow cytometry to look at the percentage of lysosome-positive cells upon CDK4 inhibition or depletion. The percentage of LysoTracker positive cells and the size of the LysoTracker positive particles were consistently and markedly increased in the absence of CDK4 activity (Fig. 3B-D). Similarly, when we quantified the size of lysosomal-associated membrane protein 1 (LAMP1)-positive particles using immunofluorescence, we found a significant increase in lysosomal density in CDK4 KO cells compared to WT MDA-MB-231 cells (Fig. 3E-F). Consistently, PD0332991 and LEE011 treatments also increased LysoTracker intensity in MDA-MB-231 cells (Supplementary Fig. S5A). Immortalized MEFs treated with LY2835219 also showed increased LysoTracker intensity (Supplementary Fig. S5B). These results suggest a new function of CDK4 in the control of lysosomal biology.

CDK4 is required for lysosomal function

It is well known that the inhibition of mTORC1 results in the induction of autophagy, a conserved catabolic process that triggers the degradation of intracellular constituents and organelles in the lysosome (35). Serum-starvation conditions, as well as mTOR inhibition by rapamycin, increased the amounts of the autophagosome marker LC3-II (Fig. 4A-B, long exposure). CDK4/6 inhibition similarly increased LC3-II levels. Moreover, CDK4 KO cells showed increased levels of LC3-II in basal conditions and were more sensitive to starvation-mediated autophagic stimuli (Fig. 4A-B). This indicates either an increase of autophagosome biogenesis or an impairment of the autophagic flux in the absence of CDK4. Indeed, these effects could be secondary to mTOR inactivation, since the decreased mTORC1 activity caused by CDK4 inhibition or depletion shown in Fig. 1 could ultimately induce lysosomal biogenesis.

Figure 4. Absence of CDK4 alters lysosomal function.

A, Western blot analysis of autophagosomal markers LC3-I, LC3-II and SQSTM1 in WT and CDK4 KO MDA-MB-231 cells treated 24h with complete media (+FBS) or serum-starvation media (-FBS), presence/absence of 0.5 μM rapamycin and of 0.5 μM LY2835219 and presence/absence of 0.3 μM BafA1, 6 h before the treatment. B, Quantification of protein expression of LC3-II protein levels normalized to α-tubulin from three independent experiments like A (only the conditions without BafA1). C, Quantification of LC3-II protein levels normalized to α-tubulin from three independent experiments like A (only the conditions with BafA1). Statistical analysis was performed using a paired t-test. D, Representative electron micrographs of WT and CDK4 KO MDA-MB-321 cells showing that cells lacking CDK4 have increased autolysosome (arrows) density and size. In CDK4 KO cells, autolysosomes accumulated nondegraded material, such as undegraded autophagosomes (arrowheads), in both complete (+FBS) and serum-starvation media (-FBS). In cell cultures without FBS, an increase in autophagosomes (arrowheads) was also observed in CDK4 KO cells as compared to WT cells. N: nucleus. E-G, Quantification of the autophagosome (E) and autolysosome (F) area per cell, and of the percentage of autolysosomes containing degraded material (G), for the conditions shown in Figure 4D. n = 20 cells per condition. H, Quantification of intracellular lysosomal activity of WT and CDK4 KO MDA-MB-231 cells in complete media. I, Quantification of cathepsin B activity in WT and CDK4 KO MDA-MB-231 cells in complete media.

In WT cells, under starvation conditions and in the presence of rapamycin treatment, bafilomycin A1 (BafA1), which is a potent V-ATPase inhibitor that blocks autophagosomelysosome fusion, further increased LC3-II levels, indicating that rapamycin induces autophagosome biogenesis. In contrast, in CDK4 KO cells or cells treated with CDK4 inhibitor, BafA1 failed to cause any additional increase in LC3-II levels (Fig. 4A-C). These results suggest that CDK4 does not directly participate in autophagosome biogenesis. On the other hand, no abnormal SQSTM1 accumulation was observed after CDK4 inhibition or depletion, despite observing a consistent SQSTM1 increase with BafA1 (Fig. 4A). SQSTM1 protein levels are often negatively correlated with autophagic degradation. However, it has been already observed that the expression of SQSTM1 does not always inversely correlate with autophagic activity, given that they can be restored during prolonged starvation (36).

We next used transmission electron microscopy (TEM) to analyze the ultrastructure of lysosomes in WT and CDK4 KO MDA-MB-231 cells incubated with serum-starvation media to trigger autophagy. Serum-starvation induced autophagosome and lysosome formation in MDA-MB-231 WT cells. CDK4 KO MDA-MB-231 cells displayed higher densities of autophagosomes and lysosomes. Moreover, TEM analysis revealed that the lysosomes in CDK4 KO cells were filled with electron-dense undigested material (Fig. 4D- G, Supplementary Fig. S5C). LY2835219 treated WT cells mimicked CDK4 KO cells (Supplementary Fig. S5D-H). Consistently, the intracellular lysosomal activity and cathepsin B activity were decreased in the CDK4 KO cells (Fig. 4H-I). Overall, these results suggest that CDK4 is fundamental for the activity of lysosomes and that its absence impairs autophagic flux at the lysosomal degradation step.

Dysfunctional lysosomes, but not mTORC1 inhibition, induce senescence

Lysosomal processes are associated with aging and longevity (37), and the increase of lysosomal content is a characteristic of senescence progression (38). This led us to investigate the fate of cells lacking CDK4 activity. Analysis of Ki-67 levels showed that CDK4/6 inhibition significantly decreased the proliferation rate of cells in complete medium (Fig. 5A). Yet, when the cells were serum-starved, no differences were observed in response to LY2835219 (Fig. 5A). Apoptosis, measured by Annexin V staining, was not significantly induced by LY2835219 in complete medium (Fig. 5B). A slight increase in apoptosis was seen when CDK4/6 inhibition was combined with serum starvation, but the overall percentage of apoptotic cells was extremely low (Fig. 5B). Interestingly, eight days of treatment with LY2835219 resulted in higher senescence, as measured by senescence-associated betagalactosidase (SA-βGal) staining. CDK4 KO cells also showed increased senescence (Fig. 5C-E). No differences were observed under serum starvation conditions or upon rapamycin treatment, indicating that the induction of senescence in the absence of CDK4 was not due to mTORC1 inhibition (Fig. 5C-E).

Figure 5. Dysfunctional lysosomes, but not mTORC1 inhibition, induce senescence.

A, Proliferation of WT MDA-MB-231 cells cultured in either complete (+FBS) or serumstarvation (-FBS) media, with or without LY2835219, shown as the percentage of Ki67- positive cells. B, Apoptosis under the same conditions, shown as the percentage of Annexin V-positive cells. C, mTOR-independent induction of senescence by LY2835219, visualized by colorimetric senescence-associated β-Galactosidase (SA-β-Gal) staining of WT and CDK4 KO MDA-MB-231 cells cultured in -/+ FBS media for the last 16 h after 8 days of DMSO, 0.5 μM LY2835219 or 0.5μM rapamycin treatment. D, Quantification of the percentage of SA-β-Gal-positive WT and CDK4 KO MDA-MB-231 cells from triplicates, at least five fields per replicate. E, Quantification of the fold induction of SA-β-Gal induced by LY2835219 or rapamycin treatment from triplicates, at least five fields per replicate. F, qPCR analysis showing expression of genes usually upregulated in senescence, in WT MDA-MB- 231 cells treated for 8 days with DMSO, LY2835219, or rapamycin in complete media. Triplicates were analyzed for each condition.

Consistent with the SA-βGal data, LY2835219 treatment induced the expression of most of the senescence-related genes evaluated (Fig. 5F). However, CDKN1A and RBL2, which are p53-regulated genes, did not respond to CDK4/6 inhibition in the CDKN2A and p53 mutant MDA-MB-231 cell line. This reinforces the idea that the induction of senescence by CDK4 inhibition or depletion is more likely related to the lysosome than to any effect on p53. With the exception of RKHD3 and IGFBP5, mTOR inhibition by rapamycin also failed to induce the expression of genes related to senescence (Fig. 5F).

These results suggest that the lysosomal dysfunction induced by CDK4 inhibition or depletion is the direct cause of the senescent phenotype in these cells, independent of mTORC1 inhibition.

The CDK4 inhibitor LY2835219 alters lysosomal function, attenuates mTORC1 activity, and decreases tumor growth in a breast cancer xenograft mouse model

To investigate the effects of CDK4/6 inhibition on lysosomal function in vivo, we used a breast cancer xenograft model by injecting MDA-MB-231 cells into the mammary glands of NSG mice. Intratumoral decrease in RB phosphorylation in response to LY2835219 confirmed that CDK4/6 was indeed inhibited in the tumors (Fig. 6A-B). Consistent with previous reports (39), LY2835219-treated mice showed halted tumor growth with reduced cell proliferation (Fig. 6C-E), compared to mice treated with vehicle. The lack of immune cell infiltration in the tumors suggested that the effects were independent of the immune system (Supplementary Fig. S6A).

Figure 6. The CDK4/6 inhibitor LY2835219 alters lysosome morphology in a breast cancer xenograft mouse model.

A, Western blot for RB and phospho-RB in xenograft protein extracts from both groups. B, Quantification of A, as arbitrary levels of phospho-RB normalized to total RB. C, Volume of MDA-MB-231 tumors from mice (n=8-9 mice per condition). D, Immunohistochemistry with Ki67 antibody and DAPI for tumor sections from DMSO- or LY2835219-gavaged mice. E, Percentage of Ki67-positive cells from D. Three images per section were analyzed for 8 or 9 mice. F, Immunohistochemistry with LAMP1 antibody and DAPI of tumor sections from mouse xenograft models treated with either LY2835219 or DMSO. G, Quantification in arbitrary units of the area of LAMP1 staining normalized to the number of nuclei per field. Three images per section were analyzed for 8 or 9 mice. H, Western blot for the mTOR target protein p70S6K and phospho-p70S6K from tumor lysates, following gavage with either DMSO or LY2835219. I, Quantification of H, showing arbitrary levels of phospho-p70S6K normalized to total p70S6K. J, Representative electron micrographs of xenografts, showing that LY2835219 treatment notably increases the density of autolysosomes (arrows) and that these autolysosomes accumulate nondegraded materials, such as undegraded autophagosomes (arrowheads). N: nucleus. K-M, Quantification for conditions shown in J, as the percentage of autophagosome area per cell (K), percentage of autolysosome area per cell (L) and the percentage of autolysosomes containing degraded material per cell (M). n = 30 cells per condition. N, SA-β-Gal staining of tumor cell senescence in tumor xenograft sections taken from mice gavaged with LY2835219 or DMSO.

The tumors of LY2835219-treated mice had increased expression of the lysosomal marker LAMP1 (Fig. 6F-G) and of TFEB target genes (Supplementary Fig. S6B). In addition, LY2835219 treatment also decreased the activity of mTORC1 (Fig. 6H-). TEM analysis proved that tumors from LY2835219 treated mice had higher densities of autophagosomes and lysosomes, and that those lysosomes accumulated non-digested material (Fig. 6J-M, Supplementary Fig. S6C). Furthermore, we showed that the tumors of these mice had typical features of senescence (Fig. 6N, Supplementary Fig. S6D). Together, these results further demonstrate that CDK4 plays an essential role in the regulation of lysosomal function in vivo and that alterations in lysosomal function lead to tumor cell senescence in a mouse model of breast cancer.

The CDK4 inhibitor LY2835219 in combination with the AMPK activator A769662 induces cell death in breast cancer cells and tumors

Lysosomes are essential for autophagy. We speculated that triggering autophagy in LY2835219-treated mice, which have impaired lysosomal function, would trigger the collapse of the system and result in cell death. In vivo, the allosteric AMPK activator A769662, which is a potent inducer of autophagy, in combination with LY2835219, resulted in tumor regression, in contrast to the treatment with LY2835219 or A769662 alone (Fig. 7A-D). Indeed, almost 50% of the tumors decreased their size upon co-treatment, although no major differences were found in proliferation (Fig. 7E-F). In contrast, cleaved-Caspase-3 staining revealed a six-fold induction of apoptosis in mice co-treated with A769662 and LY2835219 (Fig. 7E-G), suggesting that increased apoptosis was underlying the decrease in tumor size in these mice. CDK4 inhibition resulted in decreased RB phosphorylation, but no further decrease was observed in co-treated tumors (Fig. 7H-I). As for mTORC1 activity, LY2835219 treatment, as well as AMPK activation, decreased p70S6K phosphorylation (Fig. 7H and 7J). Interestingly, CDK4 inhibition triggered AMPK activation, as measured by the phosphorylation of ACC, a known target of AMPK (Fig. 7H and 7K).

Figure 7. The CDK4 inhibitor LY2835219 in combination with the AMPK activator A769661 induces cell death in breast cancer cells and tumors.

A, Tumor volume of breast cancer xenografts in NSG mice was monitored throughout the whole experiment. Treatment started on day 21 and lasted for 8 days (n=9-10). B, Representative images from tumors at the day of sacrifice, after the corresponding treatment. C, Increment of tumor volume per mouse: Vol day29/Vol day21. D, Representation of percentage of tumors, which increases more than 1.5-fold, from 1 to 1.5-fold or less than 1-fold. E, Ki67 and Cleaved Caspase-3 immunostaining in tumor sections after the corresponding treatment. F, Quantification from E: percentage of Ki67 positive cells. G, Quantification from E: Cleaved- Caspase-3 area per field, comparing LY2835219 and A769662+LY2935219. H, Western blot analysis of phosphorylation of RB, p70S6K and ACC proteins in tumor samples. I, Quantification from H: phospho-RB normalized to total RB. J, Quantification from H: phospho-p70S6K normalized to total p70S6K. K, Quantification from H: phospho-ACC, normalized to total ACC.

In vitro, CDK4 inhibition and AMPK activation as single treatments failed to induce cell death after one-week treatment in MDA-MB-231 cells as shown by the low levels of Annexin V-positive cells (Supplementary Fig. S7A). Only the combination of LY2835219 and A769662 increased notably the percentage of Annexin V-positive cells (Supplementary Fig. S7A).

TEM analysis of the tumors revealed that the percentage of the autophagosome and lysosome area per cell was increased in the single treatments, as well as in the A769662 and LY2835219 co-treated group (Supplementary Fig. S7B-D). Mice co-treated with both A769662 and LY2835219 still displayed a significant decrease in lysosomes with digested material (Supplementary Fig. 7B and 7E). The underlying mechanism of this apparent paradox could be that A769662 induced cell death only in cells in which LY2835219 treatment impaired lysosomal function. Indeed, some of the co-treated cells displayed mixed morphological features of apoptotic cell death (highly condensed chromatin, shrinkage of the cytoplasm) and autophagic cell death (numerous autophagosomes and autolysosomes, focal swelling of the perinuclear membrane (Supplementary Fig. S7B (indicated by *)).

Taken together, these results show that a combination treatment using A769662 and LY2835219 provides a better outcome than LY2835219 alone, inducing cell death and tumor regression in the MDA-MB-231 breast cancer xenograft model.

Discussion

We show here for the first time that CDK4 regulates lysosomal function and mTORC1 activity in cancer cells. We demonstrate that CDK4, through phosphorylation of FLCN, facilitates the migration of mTOR to the lysosomes in order to be activated. The abrogation of mTOR activation by depletion or chemical inhibition of CDK4 consequently resulted in a substantial increase in the numbers of lysosomes and autophagosomes. This proves that CDK4 is necessary for the dissociation of FLCN from the lysosomes, and for the subsequent recruitment and activation of mTORC1. However, this increase was not correlated with greater lysosomal activity, which suggested that CDK4 has an mTOR-independent role in this process. mTORC1 activation is initiated at the lysosome and requires the presence of AAs in the cytosol and in the lysosomal lumen (21). Importantly, we prove that CDK4 inhibition or depletion leads to the accumulation of intra-lysosomal undigested material, which probably results in impaired AA recycling and blunted mTORC1 activation. We show here an additional mechanism in which CDK4 favors mTORC1 activation by promoting the digestion of proteins in the lysosome, in turn providing metabolic intermediates that sustain cell growth and survival via downstream effectors. This pathway would be particularly crucial during prolonged starvation conditions, such as the typical environment of some tumors.

Cells initially respond to nutrient deprivation by inactivating their energy-consuming processes, such as protein or lipid biosynthesis, and by activating catabolism. At the same time, other mechanisms are activated to recycle molecules to provide the cell with enough substrates and metabolic intermediates to survive – an important function of autophagy. In the long term, autophagy reactivates the mTORC1 pathway by replenishing the lysosomes with digested proteins and AAs (40). CDK4 inhibition or depletion could therefore mimic a starvation signal. This hypothesis is in agreement with previous studies showing that CDK4/6 inhibitors induce autophagy (41,42). We also observed an increase in the number of autophagosomes upon CDK4 inhibition. However, when analyzing the ultimate fate of the autophagosomes, we found that they accumulate due to the impairment of lysosomal degradation upon CDK4 inhibition or depletion. Indeed, others have demonstrated that the inhibition of lysosomal activity causes decreased fusion with autophagosomes and vice-versa (43–45). In this study, we show that inhibiting or depleting CDK4, in addition to inducing autophagy, likely through mTORC1 inactivation, impairs the autophagic flux at the lysosomal degradation step. This observation may explain the increased susceptibility of CDK4 KO cells to autophagic stimuli. We also found that CDK4 inhibition or depletion increased the expression of many lysosomal genes. It is well known that mTORC1 negatively regulates lysosomal biogenesis (46). This could be secondary to mTORC1 inhibition, or due to a compensatory mechanism to create new lysosomes, as the existing ones are dysfunctional.

CDK4/6 inhibitors have been shown to accumulate into lysosomes, a phenomenon called lysosomal trapping (47). However, the use of CDK4 KO cells in our work demonstrates that CDK4 rather stimulates lysosomal function and that the lysosomal trapping of the drug is secondary to the lysosomal impairment induced by CDK4 inhibition. In addition, CDK4/6 inhibition has been found to result in proteasomal activation (48). A negative-feedback exists between proteasomal activity and autophagic flux (49). Thus, the impairment of autophagic flux at the lysosomal degradation step can result in proteasomal activation and vice versa.

Despite the known requirement for lysosomes in cell cycle progression (50,51), little is known about the relationship between lysosomes and senescence. Importantly, in eukaryotes, autophagy impairment via lysosomal dysfunction has been described to be an important characteristic of oxidative stress-induced senescence (52). Our findings suggest that the ultimate fate of cells that lack CDK4 is the activation of senescence due to lysosomal dysfunction. The absence of senescence when cells are treated with rapamycin further demonstrates that mTORC1 inactivation is a consequence of lysosomal impairment due to CDK4 inhibition or depletion. However, CDK4 inhibitor-induced lysosomal dysfunction was not sufficient to prompt cell death in the tumors of the treated mice. Instead, cancer cells were arrested but were still alive, which explained that the tumor burden was not decreased but only stabilized (Fig. 6). We reasoned that further forcing the autophagic flux would create additional stress that could kill these tumor cells. Autophagic induction in conditions where the lysosomal function is impaired abrogates the restoration of metabolic intermediates in the cytoplasm and the synthesis of macromolecules. Altogether, this might create a situation in which cells collapse and enter into apoptosis. With this aim, we used the AMPK activator A769662, which is known to increase autophagy, in combination with the CDK4 inhibitor. Indeed, co-treatment resulted in increased cancer cell death and therefore tumor regression (Fig.7). This is a major finding that represents a paradigm switch regarding the use of CDK4 inhibitors for the treatment of cancer. Strikingly, and consistent with our findings, CDK4 inhibitors are not efficient as a single drug in the clinical practice and are often used in combination with other drugs (53–55).

We show here that the effects of CDK4 in MDA-MB-231 cells are likely to be independent of E2F1, the transcription factor modulated by CDK4 during the cell cycle. It has been reported, however, that E2F1 regulates lysosomal positioning and activates mTORC1 by promoting its recruitment to the lysosomal surface (56,57). This suggests a dual and complementary role for CDK4 in the regulation of the mTORC1 pathway: first, through regulation of the lysosomal function; and second, through FLCN phosphorylation.

It was previously described that CDK4/6 inhibitors display anti-tumor activity only in RB-positive cells (58). In addition, it was unclear whether CDK4/6 inhibition had an effect on TNBCs. Our findings are consistent with other studies showing that CDK4/6 inhibitors can still have some effects on RB-negative cells (59) and that CDK4/6 inhibitors do have anti-tumor effects in TNBCs. Indeed, depending on the cell type, CDK4/6 inhibition triggers either a quiescence or senescence response, not necessarily via the canonical RB-E2F pathway (reviewed in (60); (61)).

Overall, the present study demonstrates a new role for CDK4 in the regulation of lysosomal function, which ultimately leads to senescence in cancer cells and mTORC1 inactivation. In addition, we highlight the importance of lysosomes in cancer and we propose that CDK4/6 inhibitors could be used in combination with other drugs to target lysosomal function as a novel anticancer strategy.

Statement of significance.

Findings uncover a novel function of CDK4 in lysosomal biology which promotes cancer progression by activating mTORC1, targeting this function offers a new therapeutic strategy for cancer treatment.