Cyclin D1-CDK4 activity drives sensitivity to bortezomib in mantle cell lymphoma by blocking autophagy-mediated proteolysis of NOXA

The MCL cell lines Mino, Jeko-1, Rec-1, Jvm2, and Granta-519 were obtained from the German Collection of Microorganisms and Cell Cultures GmbH (Germany). NIH3T3 cells stably transfected with human CD40 ligand were a kind gift from Dr. Martina Seiffert from the German Cancer Research Center, DKFZ (Germany). All cell lines were tested for mycoplasma contamination and recently authenticated by short tandem repeat profiling. Cell lines as well as the primary MCL cells were cultivated in RPMI-1640 (Biochrom, Germany) supplemented with 20% fetal calf serum, 0.1 g/l penicillin-streptomycin (Gibco, Germany), and 2 mM L-glutamine (Biochrom, Germany).

Cycloheximide, 3-methyladenine, hydroxychloroquine, Spautin-1, orlistat, BEZ235, Cerulenin, and hydrogen peroxide were purchased from Sigma-Aldrich Chemie GmbH (Germany). Liensinine was obtained from ChemFaces (China) and bortezomib, carfilzomib, and palbociclib from Selleck Chemicals (USA).

Cells were lysed, sonicated, and boiled to obtain cellular proteins according to standard protocols. Western blots were performed using a SDS-PAGE Gel Electrophoresis system. Proteins were blotted on Amersham Protran 0.1 NC nitrocellulose membranes, except for LC3 which was blotted on Amersham Hybond P 0.2 PVDF membranes (GE Healthcare, USA). The primary antibodies used were anti-NOXA antibody (Calbiochem, USA), anti-cyclin D1 antibody (Santa Cruz Biotechnology, USA), anti-MCL1, RB1, β-actin, α-tubulin, ATG5, ATG7, LC3, CDK4, PUMA, BAX, BAK, and GAPDH (Cell Signaling, USA).

After treatment, samples containing 5 × 10⁵ to 1 × 10⁶ cells were stained with Cyto-ID Green Detection Reagent (Enzo Life Sciences, USA) according to manufacturer’s instructions for 30 min at 37 °C in the dark. Afterwards, cells were stained with AnnexinV-APC (BD Pharmingen, Germany) and subsequently analyzed by flow cytometry. The Cyto-ID fluorescence intensity of AnnexinV-APC negative cells was detected and compared by histogram overlays for the different treatments.

RNA was extracted using RNeasy Mini Kit (QIAGEN N.V., Netherlands) and transcribed to cDNA according to standard protocols. NOXA (PMAIP1) expression was analyzed using TaqMan Gene Expression Assay Hs00560402_m1 (Applied Biosystems, USA) on 7900 HT Fast Real-Time PCR System (Applied Biosystems) according to the manufacturer’s instructions. TBP (Hs00427620_m1) was used for normalization of NOXA mRNA expression.

For gene silencing, we used siGENOME siRNA Reagents - Human SMARTpool siRNA (Dharmacon, UK). Sequences targeted by the siRNAs are shown in Additional file 1: Table S1. As control, siRNA (cosi) non-targeting siRNA#1 (Dharmacon, UK) was used. For electroporation, the Nucleofector™ II/2B, the Cell Line Nucleofector® Kit V, and the program X-001 (Lonza Group Ltd., Basel, Switzerland) were used. Knockdown efficacy was analyzed by Western blot.

Cell death was assessed by staining the cells with AnnexinV-FITC (BD Pharmingen, Heidelberg, Germany) and propidium iodide (PI) (Sigma-Aldrich, Steinheim, Germany) and subsequent analysis by flow cytometry. Combination index (CI) values were determined with the CalcuSyn Software (Biosoft, UK), whose algorithm is based on the Chou and Talalay’s method [25]. A combination index smaller than 0.9 indicates a synergistic effect between two substances.

Analysis of cell cycle distribution was carried out using BrdU staining as previously described [26].

The half-life of NOXA protein was determined by treating the cells with 20 μg/ml cycloheximide and harvesting cell pellets at different time points for subsequent analysis by Western blot.

Data are obtained from at least three independent experiments and expressed as standard deviation (SD) of the mean. Statistics were calculated using GraphPadPrism 4.0 software (GraphPad Software, La Jolla, CA, USA). Changes in paired samples were analyzed using two-sided paired t test, and results were considered statistically significant when p < 0.05 (*p < 0.05; **p < 0.01; ***p < 0.001).

Sensitivity of MCL cells to inhibitors of the ubiquitin-proteasome pathway (UPS), such as bortezomib, has been associated with the BH3 only protein NOXA [24, 27, 28]. A recently published preclinical study also demonstrated a link between cyclin D1 expression and bortezomib sensitivity in multiple myeloma cell lines [29]. However, it is not well established if bortezomib sensitivity is dependent on the kinase activity of the cyclin D1/CDK4 complex and if NOXA is affected by this activity in MCL cells. To address this issue, we first incubated MCL cell lines Mino, Jeko-1, and Granta-519 with or without palbociclib, a potent and selective CDK4/6 inhibitor and co-treated the cell lines with bortezomib. In all three MCL cell lines investigated, co-treatment with palbociclib partially rescued cells from bortezomib-induced cell death (Fig. 1a, left). Co-treatment of bortezomib with palbociclib led to a reduced induction of NOXA protein in the MCL cell lines and in patient-derived primary MCL cells (Fig. 1a, right). Importantly, palbociclib treatment inhibited CDK4-dependent phosphorylation of RB1 in Mino, Jeko-1, and Granta-519 (Additional file 2: Figure S1). In order to prevent excessive cell death caused by the extended cultivation duration during the co-treatment schedule, we co-cultivated the primary MCL cells with CD40 ligand-expressing fibroblasts and different cytokines. This resulted in an acquired therapy resistance of the primary MCL cells which did not exhibit cell death after bortezomib treatment with or without palbociclib co-treatment (data not shown). The development of therapy resistance mechanisms after co-culturing primary MCL cells is known and involves the activation of survival pathways and the increased expression of anti-apoptotic Bcl-2 family proteins [30, 31]. In order to analyze if the antagonism on the bortezomib effect after palbociclib treatment was dependent on inhibition of cyclin D1-driven CDK4 activity, we performed siRNA-mediated knockdown of either cyclin D1 or CDK4. We observed a comparable reduction in bortezomib sensitivity in MCL cell lines Mino and Jeko-1 after knockdown of cyclin D1 or CDK4 (Fig. 1b, left and Additional file 3: Figure S2). In addition, NOXA induction after bortezomib treatment was also antagonized by knockdown of cyclin D1 or CDK4 (Fig. 1b, right). Double knockdown of cyclin D1 and CDK4 did not increase the antagonism on bortezomib-induced cell death (Additional file 4: Figure S3), indicating that in this context, both of these proteins exert their function through the same pathways.

Collectively, these data demonstrate that cyclin D1/CDK4 activity is required for effective accumulation of NOXA protein and thus also for complete execution of bortezomib-induced cell death in MCL.

We next asked if inhibition of CDK4 activity might also reduce sensitivity to other clinically relevant proteasome inhibitors such as carfilzomib. In contrast to bortezomib, which not only reversibly blocks the chymotrypsin- and caspase-like activity of the proteasome but also binds different serine proteases, carfilzomib is more specific and irreversibly binds the N-terminal threonine active sites of the proteasome [32, 33]. As shown in Fig. 1c, the effect of palbociclib on carfilzomib-induced cell death and NOXA accumulation in MCL cell line Mino was identical to that observed with bortezomib. Importantly, this could also be shown for other compounds which have been identified to be effective in MCL cells via upregulation of NOXA, such as the fatty acid synthase inhibitor orlistat (Fig. 1c; [24]) as well as for hydrogen peroxide (Fig. 1c; [34]). Similar results were also observed in the MCL cell line Jeko-1 (Additional file 5: Figure S4). In order to confirm that the inability to induce cell death after co-treatment of bortezomib with palbociclib is dependent on NOXA, we performed knockdowns of NOXA and its anti-apoptotic binding partner MCL1. NOXA knockdown rescued bortezomib-induced cell death (Fig. 1d). Furthermore, knockdown of MCL1 reversed the palbociclib-mediated antagonism on bortezomib-induced cell death, whereas NOXA knockdown further increased the antagonism on cell death induction (Fig. 1d).

Together, these observations indicate that cyclin D1-driven CDK4 activity is required for effective upregulation of NOXA and concomitant induction of cell death not only upon treatment with proteasome inhibitors but also other NOXA protein inducing agents.

Previous studies have shown that inhibition of the proteasome leads to induction of NOXA transcript [28, 35]. However, the predominant mechanism by which NOXA protein is enhanced seems to be protein stabilization by targeting the rapid NOXA protein turnover both in MCL and in melanoma cells [24, 36]. To uncover the mechanism by which CDK4 inhibition diminishes NOXA protein in bortezomib-treated cells, we first quantified NOXA RNA levels. We found only minor differences of NOXA transcript induction upon bortezomib in cells pre-incubated with or without palbociclib, indicating that transcriptional regulation of NOXA upon bortezomib is not the predominant effect of CDK4 inhibition on NOXA regulation (Additional file 6: Figure S5). Palbociclib antagonism on bortezomib-induced cell death might also be mediated by alterations in cell cycle distribution. We therefore performed a knockdown of RB1 to allow cells to progress to s-phase while CDK4 is inhibited. Even though cells progressed to s-phase (Additional file 7: Figure S6, left panel), palbociclib antagonized bortezomib-induced cell death in RB1 knockdown cells in the same extent as cells transfected with control siRNA (Additional file 7: Figure S6, middle panel). Furthermore, bortezomib treatment led to a similar decrease of s-phase cells as palbociclib treatment (Additional file 7: Figure S6, left panel). Consequently, we conclude that it is unlikely that cell cycle distribution does mediate palbociclib effects on bortezomib-induced cell death.

We therefore studied NOXA protein stability using cycloheximide pulse-chase experiments. In accordance with previous studies [24, 37], bortezomib significantly prolonged NOXA half-life in MCL cell line Mino (Fig. 2a), whereas palbociclib alone had no effect on NOXA protein stability. Interestingly, the stabilizing effect of bortezomib was almost completely antagonized when co-treated with palbociclib indicating that CDK4 inhibition activates a proteasome-independent degradation of NOXA.

Several groups have shown that CDK4 activity [7‐11] can suppress autophagy, which represents another central degradation mechanism delivering cytosolic proteins, aggregates, or organelles to lysosomes [38]. In line with this, we observed an increase of the autophagy marker LC3-II in MCL cell line Mino upon co-treatment of bortezomib with palbociclib or after treatment with palbociclib alone (Fig. 2b, upper panel). LC3-II protein levels can be elevated either because the completion of autophagy is impaired which leads to an accumulation of autophagosomes or because there is an actual increase of the autophagic flux [39]. To differentiate this, we treated the MCL cell line Mino with palbociclib and the downstream autophagy inhibitor hydroxychloroquine which inhibits the acidification of lysosomes, completely blocks autophagy in its final step and therefore leads to the accumulation of autophagosomes [40]. Hydroxychloroquine treatment induced LC3-II protein levels which, importantly, were further increased by co-treatment with palbociclib (Fig. 2b, middle panel). We can thereby conclude that inhibition of CDK4 activity results in increased autophagic flux. In addition, direct staining of autophagic vacuoles with the autophagy Cyto-ID Green dye and subsequent cytometric analysis corroborated our results that we obtained by analyzing LC3-II protein expression. CDK4 inhibition with palbociclib alone or after co-treatment with bortezomib induced an increase in green fluorescence, indicative of elevated levels of autophagic vesicles (Fig. 2b, lower panel). Similar results were observed in MCL cell line Jeko-1, where palbociclib treatment led to an increase in green fluorescence (Additional file 8: Figure S7A). Furthermore, co-treatment of palbociclib with hydroxychloroquine showed higher green fluorescence than hydroxychloroquine alone (Additional file 8: Figure S7A), corroborating our result that CDK4 inhibition results in increased autophagic flux (Fig. 2b, middle panel). Short-term treatment of MCL cell line Mino with Palbociclib, however, did not led to an increase in autophagic vacuoles (Additional file 8: Figure S7B), indicating that it requires more time to activate intermediate signals that induce autophagy.

To test whether the antagonizing effect of palbociclib on bortezomib-induced NOXA stability is mediated by activating an autophagy-lysosome pathway (ALP)-dependent degradation of NOXA, we performed RNAi-mediated knockdowns of autophagy-related proteins (ATG5/7) and investigated NOXA protein stability. Indeed, compared to control siRNA genetic blockade of autophagy by pre-incubation with ATG5/7 siRNA prolonged NOXA half-life in the MCL cell line Mino co-treated with palbociclib (Fig. 2c), reaching a protein stability comparable to that observed in cells treated with bortezomib alone (Fig. 2a). Importantly, in cells co-treated with palbociclib, genetic or pharmacologic inhibition of autophagy with Spautin-1, liensinine, or 3-methyladenine (3-MA) also restored the effect of bortezomib on cell death induction in MCL cell lines Mino and Jeko-1 (Fig. 2d, left and Additional file 9: Figure S8) and on NOXA induction in MCL cell line Mino (Fig. 2d, right). These results indicate that inhibition of CDK4 antagonizes bortezomib efficacy via an autophagy-mediated degradation of NOXA.

If NOXA stability is targeted by UPS and ALP-mediated proteolysis, we would expect an enhanced half-life of NOXA upon inhibition of both the proteasome and the autophagy pathway as well as an increase of cell death as compared to proteasome inhibition alone. To address this issue, we used a cell line with reduced sensitivity to proteasome inhibitors, namely Jeko-1 (Fig. 1), and performed cycloheximide pulse-chase experiments. For evaluation of NOXA protein stability, cells were incubated in presence or absence of the UPS inhibitor bortezomib, the autophagy inhibitor 3-MA, and the combination of both. Interestingly, inhibition of autophagy alone had no effect on NOXA protein half-life (Fig. 3a) indicating that ALP has a minor role for NOXA degradation in cells with active UPS. However, compared to inhibition of the proteasome alone, combined inhibition of the ubiquitin-proteasome and the autophagy-lysosome pathways led to a significant increase of NOXA stability as shown by Western blot analysis (Fig. 3a, upper panel) and corresponding densitometric quantification (Fig. 3a, lower panel). Importantly, this elevated NOXA accumulation upon combined inhibition of autophagy and proteasome was accompanied by a significantly enhanced cell death as compared to proteasome inhibition alone (Fig. 3b, lower panel). Analogous to the results obtained with 3-MA, combination of bortezomib with the autophagy inhibitors liensinine or hydroxychloroquine also resulted in an elevated NOXA accumulation as well as enhanced cell death compared to proteasome inhibition alone (Fig. 3b). In line with this, autophagy inhibitors also potentiated bortezomib-induced cell death in another bortezomib-resistant cell line, namely Rec-1 (Additional file 10: Figure S9).

Together, these results indicate that bortezomib-mediated accumulation of NOXA is partially hampered by a parallel degradation of NOXA via autophagy. Therefore, combined inhibition of both UPS and autophagy potentiates both stabilization of NOXA protein and induction of cell death in MCL cells.

Previous work of our group showed that MCL cells can be efficiently treated using the fatty acid synthase inhibitor (FASNi) orlistat [24, 41]. This cell death relies on the efficient stabilization of NOXA protein. The FASNi orlistat has been shown to block palmitate synthesis [42]. Interestingly, many studies have shown that palmitate levels are linked to ER stress and autophagy regulation [43‐45]. Therefore, we investigated if bortezomib effects could be potentiated by orlistat co-treatment in analogy to the observation with autophagy inhibitors (Fig. 3).

Co-treatment of bortezomib with the fatty acid inhibitor orlistat led to a blockade of autophagic degradation in MCL cell line Mino (Fig. 4a), which is characterized by an elevated LC-3 II protein expression and a concomitant accumulation of the P62 protein which is exclusively degraded by autophagy [46]. Consequentially, orlistat prolonged bortezomib-mediated NOXA half-life in bortezomib-resistant cell line Jeko-1 as shown by Western blot analysis (Fig. 4b, upper panel) and corresponding densitometric quantification (Fig. 4b, lower panel). This enhanced protein stability resulted in super-induction of NOXA protein and an almost complete loss of viability in four MCL cell lines characterized by different sensitivity to bortezomib monotherapy (Fig. 4c). The same results were obtained with carfilzomib (Additional file 11: Figure S10) indicating that the potentiating effect was not due to a differential bortezomib uptake or substance cross-interaction.

Changes in expression levels of other apoptotic proteins apart from NOXA after combination treatments were not observed (Additional file 12: Figure S11). Treatment of the MCL cell line Jeko-1 with different combinations of orlistat and bortezomib or carfilzomib resulted in synergistic cell death even at low concentrations of bortezomib and carfilzomib (Additional file 13: Figure S12). Cell death induced by these combinations was dependent on caspase activity and could be blocked by pre-incubation with the pan-caspase inhibitor Z-VAD-FMK (Additional file 11: Figure S10). Importantly, similar effects were also observed in patient-derived primary MCL cells. Combination of orlistat together with bortezomib led to an enhanced NOXA protein expression and significantly potentiated the cytotoxic effect of bortezomib alone (when applied at 7 nM; Fig. 4d upper panel). Even a low bortezomib concentration of 5 nM which had only minor effects in monotherapy led to a significant increase of cell death when combined with orlistat (Fig. 4d, lower panel). Of note, this combinatory treatment had only minor effects on peripheral blood mononuclear cells (PBMNC) from healthy donors (Fig. 4e).

Collectively, these data demonstrate that the combination of bortezomib with a FASNi provides a potent tool to effectively induce MCL cell death via super-induction of NOXA presumably by parallel inhibition of the proteasome as well as the autophagic degradation machinery.

After demonstrating that cyclin D1-CDK4 activity is important for response to bortezomib in MCL cells (Fig. 1), we next asked if this also holds true for the more effective combination of bortezomib with orlistat. Indeed, knockdown of cyclin D1 or CDK4 in MCL cell line Jeko-1 blocked accumulation of NOXA upon combinatory treatment as effective as direct NOXA knockdown (Fig. 5a). Analogous to bortezomib alone, pharmacological inhibition of CDK4 by palbociclib prior to combination treatment was sufficient to abrogate NOXA accumulation (Fig. 5b). Again, consistent with the effects observed in cells treated with bortezomib alone, genetic or pharmacologic blockade of cyclin D1/CDK4 also partially rescued the more effective apoptosis upon co-treatment with orlistat.

Together, these results demonstrate that effective NOXA accumulation and induction of cell death after combined proteasome and fatty acid synthase inhibition is dependent on cyclin D1/CDK4 activity in MCL cells.