A novel Mcl1 variant inhibits apoptosis via increased Bim sequestration

Members of the Bcl-2 protein family are frequently deregulated in tumors as they critically control cell death induction in mammalian cells. Alterations of these proteins may cause resistance to chemotherapy-induced cell death and immune responses. By serendipity we cloned a variant of the anti-apoptotic Bcl2-family member Myeloid cell leukemia-1 (Mcl1) from human neuroblastoma and leukemia cells. This Mcl1L variant lacks a 45 bp sequence that codes for 15 highly conserved amino acids ranging from Gly158 to Asp172. This region is part of the so called PEST-sequence of Mcl1L and contains two phosphorylation sites (Ser159 and Thr163) that regulate Mcl1L stability. A caspase 3/caspase 8 cleavage site at Asp157 which has been reported to be critical for death-receptor-induced apoptosis and for the conversion of Mcl1L into a pro-apoptotic protein is also missing in this novel variant. Importantly, Mcl1LdelGly158-Asp172 bound significantly more pro-apoptotic Bim compared to Mcl1L and showed increased anti-proliferative and anti-apoptotic activity compared to Mcl1L during death receptor-induced cell death. This suggests that this novel Mcl1L variant efficiently protects tumor cells against extrinsic death signalling and therefore may provide a survival advantage for highly aggressive tumors.


INTRODUCTION
Mcl1 was originally identified in differentiating myeloid cells [1] and has unique structural features among the members of the anti-apoptotic BCL2 family. The C-terminal part (aa 170-300) of Mcl1 shares structural similarities with other anti-apoptotic BCL2 family members, like BclxL. The N-terminal part, however, lacks the characteristic BH4 domain and instead contains two highly conserved proline, glutamic acid, serine and threonine-rich PEST sequences [2]. The second PEST sequence includes also two caspase cleavage sites (Asp127, Asp157) and several phosphorylation sites that are involved in regulating Mcl1 function and stability [3][4][5][6]. Mcl1L expression is controlled by various transcriptional, post-transcriptional and post-translational pathways downstream of growth factor-and cytokine signaling [7][8][9]. Thr163 is the main phosphorylation site in Mcl1 regulating stability, function and association with pro-apoptotic BH3-only proteins. ERK-mediated phosphorylation at Thr163 and Thr92 increases Mcl1stability by binding to Pin-1 [10] as well as its antiapoptotic function. Stress-induced phosphorylation on Ser121 and Thr163 inactivates Mcl1 pro-survival function [11,12] and combined phosphorylation at Thr163 and Ser159 by JNK and GSK3ß destabilizes Mcl1 as well as reduces its interaction with pro-apoptotic Bim [13]. Beside phosphorylation the interaction with distinct BH3only proteins also coordinates Mcl1 expression, function and stability. Mcl1 can bind and thereby inactivate pro-apoptotic Bak [14,15] but this complex can be either disrupted via extrinsic death signaling by tBid or intrinsically by induction of PMAIP1/Noxa, leading to proteasomal degradation of Mcl1 and apoptosis induction via Bak-oligomerisation [15][16][17][18]. Binding and inactivation of Bim and Puma increases Mcl1 levels, protects Mcl1 from degradation and acts anti-apoptotic by sequestration of Bim [17,19,20]. Cleavage of Mcl1 by caspase-3 or 8 during TRAIL-induced apoptosis, however, releases sequestered Bim and causes apoptosis via activation of Bax. The cleavage and inactivation of Mcl1L by caspases represents a second, Bid-independent linkage between extrinsic and intrinsic death pathway [16,21]. Proteasomal degradation of Mcl1 is controlled by different E3-ubiquitin ligases. The most prominent is MULE, which is thought to regulate the constitutive turnover of Mcl1L by binding via its BH3-domain to the hydrophobic pocket of Mcl1L [6]. Two additional E3-ligases have been identified that regulate Mcl1L ubiquitination: During apoptosis execution and triggered by GSK3ß-induced phosphorylation of Mcl1L the E3-ligases SCF FBW7 and ß-TRCP regulate Mcl1 degradation [22,23]. The activity of these E3-ligases is counteracted by the de-ubiquitinase USP9X which removes Lys48-linked polyubiquitine-chains and thereby stabilizes Mcl1L and increases its anti-apoptotic function [18,24].
Beside anti-apoptotic full-length Mcl1L, there is evidence for several pro-apoptotic Mcl1 variants. Proapoptotic variants are generated either by cleavage of Mcl1 by caspase-3 or 8 [25,26] or by alternative splicing. Loss of exon 2 results in the translation of Mcl1s (short, 271aa), a splice variant which only contains the BH3-domain and inactivates Mcl1L thereby acting like a pro-apoptotic BH3-only protein [27,28]. Such a pro-apoptotic variant is also known for BclxL, where alternative splicing generates BclxS [29]. Splicing in exon1, at a non-canonical splice site, leads to Mcl1ES (extra short, 197 aa), which lacks the PEST sequence but binds Mcl1L. This variant does not sequester Bax or Bak and thereby acts pro-apoptotic [30]. In the present study we characterized a novel Mcl1 splice variant, which was cloned from human neuroblastoma and leukemia cells. As this variant lacks important regulatory parts of the PEST sequence we hypothesized that tumor cells expressing this Mcl1 variant may gain survival advantages and escape certain death stimuli.

Cloning of a novel variant of human Mcl1 in human cancer cells.
In the course of PCR analyses of Mcl1 mRNA variants in human neuroblastoma cells we amplified an mRNA species that carries a 45 bp deletion in the Mcl1L coding region from SH-EP neuroblastoma cells. For simplicity we termed this Mcl1L del158-172 variant Mcl1 JAM (Just Another Mcl1). PCR-products corresponding to the full length and the Mcl1L JAM variant were detected also in a CEM leukemia cell line (Fig. 1A). The shortened Mcl1 variant lacks a 15 amino acids region ranging from Gly158 to Asp172 that contains a caspase 3/8 cleavage site and the two important regulatory amino acids Ser159 and Thr163 (Fig. 1B). This part of Mcl1 is highly conserved within mammals (Fig. 1C). The phosphorylation of Mcl1L on Ser159 and Thr163 by GSK3β, ERK or JNK is critical for Mcl1 stability and its interaction with pro-apoptotic BH3only proteins [13].
The region Gly158 to Asp172 is the most prolineand glutamate-rich part of the PEST region which acts as a signal sequence for proteasomal degradation and determines the short protein half-life of Mcl1. As the functional consequences of lack of this region are unclear, we next investigated how the function of Mcl1L JAM differs from Mcl1L and if this novel variant affects the physiology and death resistance of human neuroblastoma cells.

Mcl1L JAM is an unstable variant that enhances the anti-proliferative effect of Mcl1L.
Besides playing a key role in the regulation of mitochondrial cell death Mcl1L and the recently described proteolytic fragment snMcl1 were also implicated in the regulation of cell cycle progression. Mcl1L binds to PCNA in the nucleus and thereby inhibits proliferation, whereas snMcl1 reduces CDK1 activity [31,32]. To assess whether Mcl1L JAM differs in its anti-proliferative activity from full length Mcl1L SH-EP cells were infected with a retrovirus vector containing either the coding sequence for Mcl1L full-length or the mRNA variant Mcl1L JAM . The expression of the smaller variant was verified by immunoblot in presence or absence of the proteasome inhibitor Bortezomib. Similar to Mcl1L Bortezomib treatment led to accumulation of the variant Mcl1L JAM suggesting that despite its truncated PEST sequence, Mcl1 JAM is still degraded via the proteasome ( Fig. 2A). Ectopic Mcl1L expression reduced the colony forming capacity of SH-EP neuroblastoma cells to 74.4% compared to mock-infected controls (100%) but did not influence the colony size (Fig. 2B). Ectopic Mcl1L JAM expression, however, further reduced the number of colonies to 66% compared to SH-EP/Ctr cells and also interestingly reduced colony size (Fig. 2B). This suggests that Mcl1L JAM reduces the ability of single cells to form colonies and, in contrast to Mcl1L, also reduces colony size suggesting a more pronounced anti-proliferative effect of this variant.
Lack of Gly158 to Asp172 reduces protein stability.
Mcl1L function, interactions and degradation are critically regulated by phosphorylation of Ser159 and Thr163 within the PEST region (reviewed in [33]

Lack of Gly158 to Asp172 increases steady state expression of pro-apoptotic Bim.
Phosphorylation at Ser159 and Thr163 not only induces destabilization of Mcl1L but was also reported to decrease the ability of Mcl1L to bind and inactivate pro-apoptotic protein Bim [13]. We therefore next studied how deletion of the region Gly158 to Asp172 in Mcl1L  (Fig. 4B). This suggests that increased levels of Noxa lead to the accumulation of more stable Noxa/Mcl1L complex in neuroblastoma cells [34]. A possible explanation for this stabilization might be that Noxa binds into the BH3domain of Mcl1L displaces the ubiquitine-ligase MULE from Mcl1L and thereby reduces Mcl1L turn over [35,36].
In contrast, doxycycline-induced expression of Noxa in SH-EP/tetNoxa-EYFP-Mcl1L JAM cells did not increase the fluorescence intensity or protein steady state levels of ECFP-Mcl1L JAM although Noxa still elevated protein levels of endogenous Mcl1L (Fig. 4C).

Co-immunoprecipiation experiments of Noxa revealed that in SH-EP/tetNoxa-ECFP-Mcl1L cells endogenous
Mcl1L and ECFP-Mcl1L precipitated with Noxa, with increased amounts when Mcl1 degradation is blocked by proteasome inhibition (Fig. 4B). Since BclxL is also an interaction partner of Noxa in neuroblastoma cells [34], we also analysed BclxL in Noxa-precipitates and found low amounts of BclxL bound to Noxa in Mcl1Loverexpressing cells (Fig.4B). This suggests a basal interaction of BclxL and Noxa and if Noxa expression is further elevated part of Noxa is sequestered by Mcl1L. In contrast, in SH-EP/tetNoxa-EYFP-Mcl1L JAM cells only small amounts of EYFP-Mcl1L JAM precipitated with Noxa, whereas endogenous Mcl1L binds to Noxa in the same manner as in ECFP-Mcl1L-overexpressing cells. In these cells, however, significantly increased amounts of BclxL co-purified with Noxa (Fig. 4D)  Co-purification of BclxL was only observed in cells treated with both, doxy and bortezomib (Fig. 5D). This combined effect may be due to the fact that bortezomib further induces Noxa, which partially sequesters Mcl1L and BclxL in neuroblastoma cells [34]. The combined data suggest that Mcl1L JAM efficiently binds and inactivates Bim and thereby changes the interaction of this proapoptotic protein with other Bcl2 proteins. In the next step we therefore analysed whether Mcl1L JAM expression changes the sensitivity to distinct apoptosis-stimuli of extrinsic and intrinsic apoptosis signalling.

Increased expression of Mcl1L JAM protects neuroblastoma cells against death receptormediated apoptotic cell death.
SH-EP/Ctr, SH-EP/Mcl1L or SH-EP/Mcl1L JAM cells were treated either with the chemotherapeutics etoposide and doxorubicin that are expected to mainly trigger the intrinsic apoptotic pathway or with FASL/ CH11 or TRAIL to induce apoptosis via death receptors. Interestingly, both Mcl1L and Mcl1L JAM only inhibited FASL-and TRAIL-induced apoptosis (Fig.6AB), whereas they failed to rescue neuroblastoma and leukemia cells from etoposide or doxorubicin-induced cell death ( Figure  6C and Supplemental figure 1AB). Mcl1L JAM expression significantly reduced cell death from 48.8% to 18.8% after TRAIL treatment and from 23.3% to 3.6% after CH11 treatment. Mcl1L JAM thereby induced significantly higher death resistance than Mcl1L overexpression after 48 hours (P < 0.05). To study long-term survival after treatment we performed clonogenic survival assays which demonstrated that survival of neuroblastoma cells after FAS-receptor activation by CH11 antibody was significantly increased in Mcl1L JAM -expressing cells (201%, P < 0.01) compared to Mcl1L-expressing cells (143%) and SH-EP/Ctr cells (Fig.6C). However, no significant differences were detectable in etoposide or doxorubicin treated cells (Fig.6C). Taken together, these data suggest that deletion of the sequence Gly158 to Asp172 in Mcl1L JAM confers increased resistance to death receptor-induced apoptosis and thereby may provide a survival advantage for tumor cells.  (Fig.3A), suggesting that the lack of the sequence Gly158 to Asp172, although being a critical part of the PEST sequence does not prevent proteasomal degradation.  (Fig.3A), suggesting that the upstream kinase cascades that modulate Mcl1L activity and stability via Ser159/Thr163 phosphorylation are active in these neuroblastoma cells.

DISCUSSION
The loss of the phosphorylation site at Thr163 suggests altered stability and function of Mcl1L JAM compared to full length Mcl1L since phosphorylation at Thr163 represents the main regulatory phosphorylation site in Mcl1L [33]. Combined phosphorylation at Thr163 und Thr92 through ERK-1 increases Mcl1 stability through association with Pin-1 [10] whereas phosphorylation at Thr163 together with Ser159 by JNK and GSK-3ß decreases Mcl1 stability as well as binding to Bim [12,13]. Actually Mcl1L JAM even showed reduced stability during CHX treatment (Fig.3B)   and therefore more instable than the wild type. Since no ubiquitin residues are affected by the deletion in Mcl1 JAM (Lys5, 40, 136, 194 and 197) [6] we hypothesize that the degradation may be caused by changes in the C-terminus of Mcl1L JAM , which might affect binding to BH3-only proteins, for example Puma which was shown to protect Mcl1L from MULE induced degradation by binding to its BH1 domain, [36] or Bim-interaction [20]. So expression of the shortened variant may protect wild type Mcl1L from its degradation, which is likely to provide a survival advantage to Mcl1L JAM -expressing cancer cells.
We also observed this enhanced stability of Mcl1L when coexpressed with Mcl1L JAM after treatment with CH11 or TRAIL. Mcl1L JAM was degraded after CH11 and TRAIL treatment but delayed phosphorlation and degradation of Mcl1L (Supplemental figure S3). The elevated apoptosis inhibitory function of Mcl1L JAM was limited to extrinsic death signaling, as both Mcl1L and Mcl1L JAM failed to reduce cell death induced by etoposide and doxorubicin treatment ( Fig. 6C and supplemental figure S1). In SH-EP neuroblastoma cells efficient death receptor signalling requires involvement of mitochondria [38]. Our data suggest that the Mcl1L JAM variant specifically interferes with the connection between extrinsic and intrinsic death signaling. Once caspase-8 is activated, it may directly cleave Mcl1L at Asp157 leading to its conversion into a pro-apoptotic Bcl2-protein [39] and to changes in the sequestration of pro-apoptotic Bcl2 proteins [21]. Since Mcl1L JAM lacks Gly158, it may be protected from cleavage. Additionally caspase-8 cleaves and activates Bid, a strong BH3-only protein that neutralizes Mcl1L. Since Mcl1L JAM showed increased affinity to Bim compared to Mcl1L (Fig.  4 and Fig. 5), tBid may not be able to disrupt Mcl1 JAM / Bim complexes, resulting in prolonged inactivation of Bim in Mcl1 JAM -expressing cells. Deletion of the entire Mcl1L N-terminus changes the C-terminal part in a way that an increased binding with Bim is observed [21]. Interestingly, the same is also true for the short 15 amino acid deletion present in Mcl1L JAM (Fig. 5B) suggesting conformational changes in the C-terminal part of this protein variant that result in altered hydrophobic BH3 binding [2]. Consistent with this hypothesis Mcl1L JAM completely failed to sequestrate increased cellular Noxa amounts upon tetracycline-regulated Noxa induction, (Fig. 4D). Instead, Noxa mainly interacted with endogenous BclxL and Mcl1L (Fig 4B).  [37]. These changes in Mcl1 localization also lead to reduced stability and less apoptosis protection against mitochondria-induced cell death. Life cell imaging analyses of SH-EP cells infected with the EYFP-Mcl1L JAM construct uncovered a partial nuclear localization after stabilization of the protein with bortezomib (Supplemental figure S4, upper panel). Subcellular fragmentation experiments also detected large amounts of EYFP-tagged Mcl1jh in the nuclear extracts, also in untreated cells, whereas a small amount of endogenous Mcl1L was also detected in the nucleus after proteasome-inhibition (Supplemental figure S4, lower panel). In line with this report, Mcl1L JAM was not able to inhibit mitochondrial cell death, but increases resistance to death receptor induced apoptosis (Fig 6ABC). This suggests that loss of this short peptide sequence in Mcl1 significantly affects stability, interaction with BH3-only proteins and also death sensitivity to distinct apoptotic signals. Expression of this Mcl1 variant may therefore represent an adaption of tumor cells to avoid extrinsic death signaling and may thereby serve as a diagnostic and/or therapeutic gene in neuroblastoma and other malignancies.

Detection of Mcl1L JAM by PCR.
Mcl1L JAM expression was analysed via RT-PCR. For PCR detection the coding sequence of MCL1 was amplified from cDNA and a nested PCR was performed using 5' AAGAGGAGCTGGACGGGTAC and 3' TGGCTTTGTGTCCTTGGC which amplifies part of the PEST region. PCR products were analysed on 2% agarose gels.

Immunoprecipitation and Immunoblotting.
For immuoprecipitation 1x10 7 cells were lysed in PBS containing 1% IGEPAL, phosphatase-and protease-inhibitors. For immunoprecipitation 1 µg of anti-rabbit Bim antibody (Cell Signaling Technology Inc., Boston, MA, USA), 2.5 µg of anti-mouse Noxa (Abcam, Cambridge, UK), or mouse or rabbit immunoglobulin, as a negative control, were covalently coupled to Tachisorb TM Immunoadsorbent (Calbiochem, Nottingham, UK) or Affi-Prep Protein A Support (BioRad Laboratories, Munich, Germany) using dimethylpimelidate dihydrochloride/ Borax buffer. Antibody-bead complexes were added to 500 µg lysate and incubated at 4°C for 6 hours. Tachisorb TM -/Protein A-immunocomplexes were washed four times in PBS/IGEPAL-buffer, resuspended in SDSsample buffer and subjected to SDS-PAGE and blotting. Equal amounts of total-protein and cleared supernatants were loaded as controls. Immunoblot analysis was performed as previously described [45] using primary antibodies directed against human Bim and Mcl1 (BD-Pharmingen, USA, pMcl1(Ser159/Thr163) and BclxL (Cell Signaling Technology Inc., Boston, USA), Noxa (Alexis Biochemicals, San Diego, CA, USA), GFP (Sigma, Vienna, Austria) and GAPDH (Acris antibody GmbH, Herford, Germany). The membranes were then washed and incubated with horseradish-peroxidaseconjugated anti-mouse or anti-rabbit secondary antibodies (Amersham Biosciences, Buckinghamshire, UK). The blots were developed by enhanced chemiluminescence (GE-Healthcare, Vienna, Austria) and measured with an AutoChemi detection system. Densitometry analysis was performed using LabWorks software (UVP, Cambridge, UK).

Colony forming assay (CFA).
To determine the ability of SH-EP cells to form colony units, 2x10 3 cells were seeded into a 6well and cultured up to 7 days. For chemotherapeutic treatment 4x10 4 cells were seeded into 6wells and treated for 72 hours with chemotherapeutic agents (CH11, etoposide, doxorubicin). Afterwards medium was removed and cells were fixed and stained with 0.2% crystal violet in 50% methanol. Cell density was measured photometrically after discoloration with 0.5% SDS in 50% ethanol. Untreated/ mock-infected cells were set as 100%.

Flow cytometry analyses.
Apoptosis was assessed by staining the cells with propidium-iodide (PI) using a CytomicsFC-500 Beckman Coulter as previously described [46]. In short: 2x10 5 cells were harvested and resuspended in hypotonic PI solution for 2-4 hours at 4°C. Stained nuclei in the sub-G1 marker window were considered to represent apoptotic cells. Statistical analysis was performed using GraphPad Prism 4.0 software.

Live cell fluorescence microscopy.
For live cell analyses cells were grown on LabTek Chamber Slides TM (Nalge Nunc International, Rochester, NY, USA) coated with 0.1 mg/ml collagen. Images were collected with an Axiovert200M microscope with a 63x-oil objective (Zeiss, Vienna, Austria). Mitochondria staining was performed using 300 nM CMXRos (Invitrogen, Carlsbad, USA).

Statistical analysis.
Statistical significance of differences between controls and treated cells were calculated using unpaired t-test. All statistical analyses were performed using Graph Pad Prism 4.0 software.