Ubiquitination and deubiquitination of MCL1 in cancer: deciphering chemoresistance mechanisms and providing potential therapeutic options

MCL1 is an important antiapoptotic member of the BCL-2 family that is distinguishable from other family members based on its relatively short half-life. Emerging studies have revealed the crucial role of MCL1 in the chemoresistance of cancer cells. The antiapoptotic function of MCL1 makes it a popular therapeutic target, although specific inhibitors have begun to emerge only recently. Notably, emerging studies have reported that several E3 ligases and deubiquitinases modulate MCL1 stability, providing an alternate means of targeting MCL1 activity. In addition, the emergence and development of proteolysis-targeting chimeras, the function of which is based on ubiquitination-mediated degradation, has shown great potential. In this review, we provide an overview of the studies investigating the ubiquitination and deubiquitination of MCL1, summarize the latest evidence regarding the development of therapeutic strategies targeting MCL1 in cancer treatment, and discuss the promising future of targeting MCL1 via the ubiquitin–proteasome system in clinical practice.


Introduction
Programmed cell death (PCD), a cell suicide process, is essential for normal organ development and tissue homeostasis. However, overactivation or inactivation of the PCD cascade can lead to pathogeneses, such as Parkinson's disease and tumorigenesis. Apoptosis is the predominant type of PCD, whereas other types have also been discovered in recent years, such as pyroptosis 1 , necroptosis 2 , parthanatos 3 , and ferroptosis 4 . Two distinct pathways can lead to apoptosis, the extrinsic/cell death receptor pathway and the intrinsic/mitochondrial pathway 5 . In the intrinsic pathway, B cell lymphoma 2 (BCL-2) family members have been shown to have a critical role in regulating apoptosis by governing proapoptotic and antiapoptotic intracellular signals 6 . Numerous studies have revealed the subtle regulatory network between BCL-2 family members 7,8 , and their dysregulation can lead to various diseases 9,10 .
Chemotherapy is one of the most widely applied treatments for cancer in humans, and chemoresistance is one of the most severe obstacles that cancer patients face. Apoptosis escape is considered a hallmark of cancer 11 , contributing to the resistance of cancer cells to both chemotherapy 12,13 and radiotherapy 14 . Due to the regulatory role of the BCL-2 family in apoptotic signaling, studies of the BCL-2 family have elucidated the mechanisms underlying chemoresistance and provided several promising therapeutic strategies. Certainly, emerging attempts to target BCL-2 antiapoptotic members will benefit the treatment of cancer.

Role of MCL1 in the mitochondrial apoptosis pathway
The BCL2 family contains more than 25 proteins that regulate the intrinsic apoptosis pathway 15 . Based on their structure and function, BCL-2 family members are classified into three groups: antiapoptotic, proapoptotic, and proapoptotic BH3-only proteins. Antiapoptotic proteins that possess BH1-4 domains include MCL1, BCL-2, BCL-W, BCL-B, BCL-XL, and BFL-1/A1; proapoptotic proteins (apoptosis effectors) harboring BH1-4 domains include BAX and BAK; and proapoptotic BH3only proteins (apoptosis activators) include BIM, BID, BAD, NOXA, and PUMA 16 . Biochemical and structural studies have revealed that the BH1, BH2, and BH3 regions of antiapoptotic BCL-2 family proteins are proximal to one another and can form a surface-associated hydrophobic groove, a feature allowing interaction with the BH3 amphipathic helix of BH3-containing proapoptotic proteins and physiological antagonization 15,17 . MCL1 is a crucial antiapoptotic member of the BCL-2 family that was initially identified as an immediate-early gene expressed during the 12-O-tetra-decanoylphorbol-13-acetate (TPA)-mediated differentiation of the ML-1 human myeloid leukemia cell line 18 . MCL1 contains two PEST domains, four BH domains, and a C-terminal transmembrane (TM) domain as well as an unusually long N terminus that is not observed in other BCL-2 family members 19 (Fig. 1). The N terminus of MCL1 harbors numerous regulatory residues responsible for caspase cleavage, phosphorylation, and ubiquitination, and contains a PEST motif, a characteristic sequence of rapidly degraded proteins 20 . In addition, the BH1 domain and Cterminal region from aa 303 to 350 are also important for the rapid degradation of MCL1, contributing to its short half-life together with the PEST sequence 21 . The short MCL1 half-life distinguishes it from many other BCL-2 homologs 22 , and allows cells to rapidly switch between survival and apoptotic states in response to various stress signals. Another characteristic of MCL1 is its distinct tissue distribution and expression patterns compared with those of other BCL-2 family members 23 .
Several studies have provided a relatively detailed mechanism underlying the antiapoptotic function of MCL1. Classically, the MCL1 protein has been shown to anchor itself to the cytoplasmic face of the mitochondrial outer membrane via its C-terminal TM domain 24 , where it structurally interacts with and sequesters proapoptotic proteins to suppress mitochondrial outer membrane permeabilization (MOMP) and cytochrome c release 25,26 (Fig. 2). Conversely, the above interactions can be antagonized by BH3-only proteins, which displace BAK and BAX from MCL1 to activate the mitochondrial apoptosis pathway 6,27 . In addition to the abovementioned classic antiapoptotic function of MCL1, maintenance of MCL1 levels has been shown to be necessary to preserve mitochondrial morphology and support normal mitochondrial bioenergetic activity in cardiomyocytes 28,29 . In addition, an amino-terminal truncated isoform of MCL1 has been reported to be anchored to the inner mitochondrial membrane (IMM) and exposed to the matrix where it retains the normal IMM structure, mitochondrial fusion, ATP production, membrane potential, and respiration 30 . This mitochondrial matrix form of MCL1 can also directly interact with and modulate very long-chain acyl-CoA dehydrogenase, a key enzyme in the mitochondrial fatty acid β-oxidation pathway, to engage in lipid metabolism 31 . Excluding the functions within mitochondria, MCL1 can be translocated into the nucleus to activate Chk1 and maintain genome integrity in response to DNA damage 32,33 . MCL1 also acts together with BCL-2 and BCL-XL as a transcriptional regulator 34 or as a stress sensor to participate in autophagy regulation 35 .

Role of MCL1 in the context of cancer
The amplification and elevated expression of MCL1 has been observed across cancer cell lines and human malignancies. A survey of the expression of antiapoptotic BCL-2 subfamily members in 68 human cancer cell lines revealed that MCL1 mRNA expression was higher than that of other BCL-2 members in lung, prostate, breast, ovarian, renal, and glioma cancer cell lines 36 . Increasing evidence has also shown that MCL1 is highly expressed in multiple cancer subtypes, including hematological malignancies 37 , melanoma 38 , testicular germ cell tumor 39 , hepatocellular carcinoma 40 , breast cancer 22 , urothelial carcinoma 41 etc. Genetic studies have revealed that the MCL1 gene is located within 1q21.2, one of the most frequently amplified chromosome regions, and amplified in more than 10% of cancers 42,43 .
Because MCL1 functions as an efficient brake in the mitochondrial apoptosis pathway, it is understandable why MCL1 expression is preferentially increased in cancer cells to sustain their survival in response to various stresses, such as oncogenic stress, X-rays, chemotherapy, and small-molecule inhibitors [44][45][46] . Indeed, cancer patients with high levels of MCL1 expression have been shown to encounter drug resistance, relapse and poor prognosis outlook. For example, in diffuse large B cell leukemia, AKT-induced aerobic glycolysis promotes MCL1 protein synthesis, thereby maintaining cell survival and resistance to BCL-2 inhibitors 47 . MCL1 is also frequently upregulated in breast cancer 48 , especially in drugresistant triple-negative breast cancer (TNBC) after neoadjuvant chemotherapy, with MCL1 (54%) and MYC (35%) gene coamplifications 49 . Elevated MCL1 expression has also been detected in chemoresistant cell lines and patients with ovarian cancer 50,51 .

MCL1 modulation by UPS
Ubiquitin, a 76-residue polypeptide, is a highly stable and conserved protein 52 , and ubiquitin conjugation is achieved through a cascade of multiple enzymatic reactions catalyzed by three classes of enzymes, including E1 ubiquitin-activating enzymes, E2 ubiquitin-conjugating enzymes, and E3 ubiquitin ligases (E3s) 53 . Eight residues (M1, K6, K11, K27, K29, K33, K48, and K63) within ubiquitin serve as attachment sites for the formation of polyubiquitin chains 54 . Proteins modified with K48-linked chains, the most abundant linkage type in cells, are typically degraded by the 26S proteasome 53 . Non-K48 linkages are primarily involved in nondegradative functions, including cellular signaling, intracellular trafficking, DNA damage response, and chromatin remodeling 52,55 , although their roles are not well defined. The ubiquitination process can be reversed and edited by deubiquitinases (DUBs), which cleave ubiquitin linkages from the substrates to alter their stability or activity 56 . All cellular proteins can be ubiquitinated at least once during their lifetime 57 ; therefore, the ubiquitin code permeates every space in the cell and is involved in regulating almost every biological process.
Considering the short half-life and unstable nature of MCL1, posttranslational regulation, especially by the ubiquitin-proteasome system (UPS), is an important mechanism by which high MCL1 expression is maintained in cancer. To date, at least six E3s (Mule, SCF β-TrCP , SCF FBW7 , TRIM17, APC/C Cdc20 , and FBXO4) have been shown to have a role in the ubiquitination of MCL1 58,59 . The first study suggesting that MCL1 could be regulated by the UPS was published in 2002 but lacked detailed experimental insight 60 . In 2003, two research groups successively reported that the rapid turnover of MCL1 is mediated by the proteasome 61,62 . In 2005, the first molecule shown to directly regulate the ubiquitination of MCL1 was identified through biochemical fractionation of cell extracts 63 . MCL1 ubiquitin ligase E3 (Mule) is a 482-kDa HECT domain-containing ubiquitin ligase that was named for its ability to ubiquitinate MCL1 and induce its degradation in vitro 63,64 . Mule specifically acts on MCL1, as no interactions occur between the BH3 domain of Mule and BCL-XL, BCL-2, or BAX 63 .
The ubiquitination-mediated degradation of MCL1 has been shown to be dependent on its phosphorylation 58 . SCF β-TrCP ubiquitinates and destabilizes MCL1 in a GSK3-dependent manner 65 , whereas the phosphorylation of MCL1 in SCF FBW7 -induced MCL1 degradation can be mediated by the phosphokinases GSK3, JNK, p38, CKII, or CDK1 depending on the cellular context 66,67 . TRIM17 is the fourth E3 ligase demonstrated to target MCL1 68 . Interestingly, both the ubiquitination of MCL1 by TRIM17 and the association between MCL1 and TRIM17 have also been shown to be associated with the phosphorylation of MCL1 by GSK3 68 . APC/C Cdc20 has been shown to engage in the ubiquitination of MCL1 and to control MCL1 stability during mitosis 69,70 ; however, existing evidence appears to be insufficient to support the hypothesis that APC/C Cdc20 is a bona fide MCL1 E3 ligase. FBXO4 is the last MCL1 E3 ligase identified and is a specific F-box protein that promotes MCL1 degradation in lung cancer 59 . Intriguingly, Choi et al. 71 showed that an E3 ligase, TRAF6, promoted the K63-linked polyubiquitination of MCL1 on the C terminus and prevented its degradation; these results differ from those of previous studies showing that E3 ligases conjugate MCL1 with K48-linked polyubiquitin chains to promote its degradation.

Deubiquitinases of MCL1
The discovery of ubiquitination events and E3s related to the critical antiapoptotic protein MCL1 revealed one aspect of its posttranslational modulation. Deubiquitination is the reverse process of ubiquitination, and the ubiquitination of MCL1 can be counteracted by specific DUBs. Therefore, the identification of DUBs regulating MCL1 revealed another side of this modulation process (Fig. 3).

USP9X
The first DUB targeting MCL1, USP9X, was identified among the immunoprecipitants interacting with MCL1 72 . USP9X prolongs the half-life of MCL1 in an enzymatic activity-dependent manner by specifically cleaving the degradative K48-linked chains on MCL1 to prevent its proteasomal degradation 72 . Structurally, USP9X interacts with the N terminus of MCL1 but not with the BH domain, and the USP9X/MCL1 interaction is relatively weak compared with that between MCL1 and other BCL-2 members, such as NOXA 72 , indicating that this interaction may be easily disrupted. Indeed, subsequent studies showed that NOXA overexpression triggers a decrease in the USP9X/MCL1 interaction, leading to increased ubiquitination and degradation of MCL1 and the induction of apoptosis 73,74 . Schwickart et al. 72 also showed that maintaining the MCL1 residues Ser155, Ser159, and Thr163 in an unphosphorylated state is crucial for the interaction of USP9X and MCL1. Conversely, MCL1 phosphorylation at these sites by GSK3β disrupts the USP9X/MCL1 interaction, triggers its recognition by E3 ligases and promotes rapid degradation. It appears that phosphokinases regulate the state of MCL1 to determine whether it is degraded by its E3 ligases or stabilized by USP9X. Interestingly, USP9X exhibits biased expression in the brain and immune system 75 , and its function is dependent on the cancer subtype context. USP9X acts as an oncogene in multiple myeloma, lymphoma and nonsmall cell lung cancer 72,74 but acts as a tumor suppressor in pancreatic ductal adenocarcinoma and colorectal cancer 76,77 . Moreover, USP9X has also been reported to suppress colorectal cancer progression by stabilizing SCF FBW7 , an MCL1 E3 ligase 77 .

Ku70
Lupus Ku autoantigen p70 (Ku70), a DNA doublestrand break (DSB) repair protein, has been shown to have an important role in the nonhomologous end-joining (NHEJ) pathway 78 . Intriguingly, in addition to its familiar function, Ku70 has been reported to possess intrinsic deubiquitination activity, cleaving degradative K48-linked polyubiquitin chains from MCL1 and promoting its stabilization 79 . The role of Ku70 as a DUB has been further confirmed by evidence showing that Ku70 directly deubiquitinates MCL1 in a dose-dependent manner and hydrolyzes polyubiquitin chains into monoubiquitin units in vitro 79 . Ku70 specifically stabilizes MCL1 but none of the other antiapoptotic BCL-2 family members 79 . Interestingly, the N-and C termini of Ku70 have distinct functions as the N terminus (aa 1-535) exhibits DSB repair activity, and the C terminus (aa 536-609) is required for the deubiquitination of MCL1 and the antiapoptotic activity of Ku70 79 . In addition, IR enhances MCL1 nuclear translocation, increases the Ku70/MCL1 interaction and reduces the USP9X/MCL1 interaction in a dose-dependent manner 79 . This compensation strategy may be part of the mechanism underlying radioresistance in lung cancer.

USP13
USP13 was identified as a bone fide MCL1 DUB based on its ability to directly remove the ubiquitin chains modified on MCL1 in vitro 80 . Zhang et al. 80 showed that both USP13 and MCL1 are genomically amplified in numerous cancer types and reported a positive correlation between USP13 and MCL1 in ovarian cancer patient tissues. However, a significant positive correlation between USP13 and MCL1 is not observable in lung cancer patient tissues 80 , and our data even showed a negative correlation between USP13 and MCL1 in nine ovarian cancer cell lines 51 . Previous studies have also shown that USP13 can deubiquitinate essential tumor suppressors (Vps34 complexes, p53, and PTEN) to function as a tumor suppressor in ovarian, breast and bladder cancers [81][82][83] . Thus, the significance of USP13-MCL1 regulation in the detailed context of cancer must be validated, and the possibility of targeting USP13 as a cancer treatment option must be explored.

JOSD1
JOSD1 was initially observed to exhibit membranerelated functions 84 , and a subsequent study revealed its antiviral effects via modulation of SOCS1 85 . We performed an in vivo screen to identify essential DUBs that may contribute to the acquired chemoresistance of gynecological cancer, and the results revealed JOSD1 to be a key antiapoptotic DUB whose accumulation leads to acquired chemoresistance in both cervical and ovarian cancer. Our following studies confirmed that JOSD1 acts as a bona fide DUB that cleaves K48-linked polyubiquitin chains on MCL1 and protects it from UPS-mediated degradation. Interestingly, previous studies primarily revealed the role of JOSD1 on the cell membrane and proposed that the deubiquitination activity of JOSD1 is dependent on its monoubiquitination status 84 . In our in vitro deubiquitination assay, we observed that JOSD1 functioned as an MCL1 DUB, and that its deubiquitination activity was independent of its monoubiquitination 86 . Moreover, the colocalization of JOSD1 and MCL1 in the cytoplasm indicates that the DUB activity of JOSD1 may be more important in the cytoplasm than on the membrane, although relatively high JOSD1 expression has been observed on the membrane. In future studies, the JOSD1-MCL1 regulatory axis should be validated in other cancer types.

DUB3
As mentioned above, each previously reported MCL1 DUB has a largely context-specific effect on MCL1 stability. Furthermore, Zhang et al. 80 also indicated that several DUBs may affect the stability of MCL1, although they selected USP13 for further studies due to its high amplification rates. Therefore, we performed an unbiased DUB screening to identify the strongest regulator of MCL1. DUB3 was confirmed to be the predominant factor that regulates the stability of MCL1 51 . By analyzing the expression levels of MCL1 and the other DUBs in nine ovarian cancer cell lines, we observed that only DUB3, and not USP9X or USP13, showed a significant positive correlation with MCL1, further supporting our hypothesis that DUB3 is the crucial DUB that modulates MCL1 stability 51 . Moreover, since DUB3 is a veritable oncogene due to its ability to stabilize Snail, Geminin, Cdc25A, NRF2, BRD4, and Cyclin A 87-93 , targeting DUB3 may be a promising strategy in cancer treatment with potentially fewer side effects.

Targeting MCL1 in cancer treatment
The specific structure of MCL1 had rendered the design of specific MCL1 inhibitors difficult 94 . Fortunately, the emergence of some specific MCL1 inhibitors holds promise for interfering with MCL1, and some BH3 mimetics targeting MCL1 are already being investigated in clinical practice [95][96][97][98] . Targeting upstream transcriptional regulators and DUBs of MCL1 has provided another approach to target MCL1. Proteolysis-targeting chimeras (PROTACs) also show a promising future for targeting MCL1 (Fig. 4).

Development of direct MCL1 inhibitors
As we introduced above, BH3-only proteins displace BAK and BAX from MCL1 to induce apoptosis; thus, drugs mimicking BH3-only proteins might overcome chemoresistance. Conformational analyses showed that the hydrophobic groove of MCL1 is more rigid than that of other BCL-2 family members, and that MCL1 requires a high affinity for its endogenous ligands, which renders the design of MCL1 inhibitors challenging and explains why less progress has been achieved in their development compared with inhibitors targeting other BCL-2 family members 6 . A-1210477 is the first bona fide BH3 mimetic that has been demonstrated to target MCL1, as it selectively binds to MCL1 and disrupts the MCL1-BIM interaction 99 . In 2016, Kotschy et al. 95 developed the first specific MCL1 inhibitor, S63845, which showed a 20-fold higher affinity than the previously reported A-1210477. The effect of S63845 was shown to be ideal for hematological tumors in subsequent studies 100,101 , although the effects on solid tumors were less pronounced 51,95 . Subsequently, AMG 176, a first-in-class orally bioavailable inhibitor of MCL1, was reported 96 . AMG 176 has two primary advantages over S63845: it can be administered orally and is effective against B cell, monocytic, and neutrophilic leukemia. Soon after the development of AMG 176, another MCL1 inhibitor named VU661013 was also reported 98 .
The sudden emergence of effective MCL1 inhibitors has broadened the horizon for strategies used to target MCL1, and studies have also confirmed the efficacy of using MCL1 inhibitors in combination with other BCL-2 family inhibitors 98,[102][103][104] . Several MCL1 inhibitors, including S63845 95 , AMG 176 96 , AZD5991 97 , S64315 105 , and AMG 397 (https://clinicaltrials.gov), are already being investigated in the clinical stage, and the future of their application in the treatment of chemoresistance is encouraging. Notably, current MCL1 inhibitors show a higher affinity for human MCL1 than mouse MCL1, leading to complications for testing MCL1 inhibitors in pre-clinical mouse models. Fortunately, the recent generation of a humanized MCL1 mouse model provides a solution to more accurately evaluate the tolerability and efficacy of new MCL1 inhibitors in pre-clinical animal experiments 106 .

Targeting transcriptional regulators of MCL1
As discussed above, the rigid structure and rapid turnover of MCL1 make the design of specific inhibitors against this protein difficult. Therefore, before the recent success of the development of specific MCL1 inhibitors, targeting the regulation of MCL1 was considered an easier strategy than developing an effective BH3 mimetic 107 . One alternative strategy to target MCL1 is by facilitating its transcriptional repression. CDK9 has an important role in regulating gene transcription and has become a potential therapeutic target in cancer treatment 108,109 . Inhibition of CDK9 triggers the suppression of short-lived transcripts and proteins, such as MCL1, to induce tumor cell death 110,111 , and is thus used in concert with BH3 mimetic drugs for cancer treatment 112 . Currently, the selective CDK9 inhibitors dinaciclib, voruciclib, AZD4573 and alvocidib are already under clinical development for the treatment of hematological malignancies 110,111,113 . Another category of MCL1 transcriptional regulator comprises the signal transducer and activator of transcription (STAT) transcription factors. Activation of either STAT1 114 or STAT3 115 can promote MCL1 transcription. FMS-like tyrosine kinase-3 signaling also activates MCL1 expression through its STAT5-docking domains 116 . Consistently, both suppression of STAT3 by an antisense oligonucleotide inhibitor (AZD9150) 117 and inhibition of STAT3 phosphorylation by hydrazinocurcumin 118 lead to decreased MCL1 transcription. Combined usage of sorafenib (a multikinase inhibitor that suppresses STAT3 phosphorylation) and ABT-737 also synergize to induce glioma cell apoptosis by the inhibition of both MCL1 and BCL-2 119 . c-MYC controls the transcription of MCL1 directly and also regulates BCL-XL by regulating the eIF4E gene, in which context the inhibition of c-MYC sensitizes gastric cancer cells to the histone deacetylase inhibitor (HDACi) suberoylanilide hydroxamic acid 120 .

Targeting DUBs upstream of MCL1
The rapid discovery of several DUBs stabilizing MCL1 has provided another opportunity to indirectly target MCL1 by suppressing DUBs (Table 1). Indeed, USP9X knockdown decreases the level of MCL1 protein, sensitizes MCL1resistant colon carcinoma and leukemia cells to ABT-737 72 , and sensitizes radioresistant cells to apoptosis induction 121 . The shRNA-mediated silencing of USP9X increases the sensitivity of chronic myelogenous leukemia cells to imatinib and other apoptotic stimuli 122 . BIX-01294, a euchromatic histone-lysine N-methyltransferase 2 inhibitor, has been shown to suppress the USP9X-MCL1 axis and induce apoptosis in bladder cancer cells 123 . WP1130, a secondgeneration tyrphostin derivative, was first identified in screens for AG490-like molecules 124 . Interestingly, WP1130 has also been confirmed to be a small-molecule inhibitor of USP9X activity 125 , and has been shown to trigger apoptosis in myeloid leukemia cells by destabilizing MCL1 122,124,125 .
The discovery of Ku70 stabilizing MCL1 in lung cancer is also promising for the development of corresponding inhibitors. Chen et al. 126 identified a novel small molecule, MI-223, that strongly synergizes with DNA replication stress agents to combat lung cancer by disrupting the MCL1/Ku70 interaction and inhibiting homologous recombination activity. In another comparative study, Wang et al. 79 showed that compared with that in cells individually treated with Ku70 shRNA or USP9X shRNA, the level of MCL1 in Ku70/USP9X double-knockdown cells was almost undetectable, increasing the sensitivity of lung cells to treatment with staurosporine or the BCL-2 inhibitor ABT-737. Thus, the development of an inhibitor targeting both Ku70 and USP9X, or at least a combination of both Ku70 and USP9X inhibitors, is a feasible approach for further studies. Interestingly, Zhang et al. 80 showed that pharmacological inhibition of USP13 by a smallmolecule inhibitor, spautin-1, markedly downregulated MCL1 protein levels and sensitized ovarian and lung cancer cells to ABT-263, a selective antagonist of BCL-2 and BCL-XL.
We selected DUB3 as an ideal target in ovarian cancer due to its oncogenic functions in various cancer subtypes 51,[87][88][89][90][91][92][93] . After screening several small-molecule inhibitors, we observed that PaTrin-2 showed the most effective suppression of DUB3 mRNA expression 51 . Mechanistically, PaTrin-2 acts as a pseudosubstrate of O 6 -methylguanine-DNA methyltransferase (MGMT) 127 , and the inhibition of DUB3 transcription may result from the inhibition of its transcription factors by MGMT interference 128 . Functional study results verified that PaTrin-2 effectively reduced the viability of ovarian cancer cells with high MGMT-DUB3-MCL1 expression levels 51 . Most surprisingly, we observed that HDACis could activate DUB3, although the underlying mechanisms remain poorly understood. This "side effect" results in the successful use of PaTrin-2 and HDACis in combination for the treatment of ovarian cancer 51 . Of course, further clinical studies are still necessary to validate this effect. The rapid development of gene therapy has provided flexible methods for modulating the DUBs that stabilize MCL1. However, DUB3 is excluded from this approach because of its high level of homology with several members of the ubiquitin-specific peptidase 17-like family (USP17L1, USP17L3, USP17L4, and USP17L8) since "offtarget" effects would be difficult to avoid. For this reason, JOSD1 appears to be the most ideal gene therapy target. Using siRNA-mediated JOSD1 knockdown, we confirmed that the loss of JOSD1 could markedly increase the apoptosis of gynecological cancer cells 86 . Based on this evidence, adeno-associated virus (AAV)-mediated depletion of JOSD1 by shRNA was applied to several ovarian cancer patient-derived xenograft models, and the results showed that AAV treatment targeting JOSD1 significantly suppressed the growth of the xenografts 86 . The results of our study thus provide a promising strategy for the use of therapeutic gene-targeting methods to target DUBs upstream of MCL1.

Application of PROTACs
The UPS is an evolutionarily conserved apparatus in eukaryotes that is responsible for more than 80% of cellular proteolysis 129 . However,~80% of proteins in human cells cannot be targeted pharmacologically and are defined as "undruggable" 130 . PROTACs are heterobifunctional molecules capable of bringing a target protein into spatial proximity with an E3 ubiquitin ligase and forming a target-PROTAC-ligase ternary complex to mediate target protein degradation by the UPS 131,132 . The first oral PROTAC drug ARV-110, which targets the androgen receptor for degradation in metastatic castration-resistant prostate cancer patients, was approved by the FDA for phase I clinical trials in 2019 133 . As expected, a recent study developed novel MCL1targeting PROTACs that effectively bring MCL1 into proximity of the E3 ligase CUL4A-DDB1-CRBN, which labels MCL1 with ubiquitin chains for proteasomal degradation at nanomolar concentrations, resulting in activation of the mitochondrial apoptosis pathway 134 . Importantly, recent studies have shown that BCL-XL PROTAC degraders can direct BCL-XL to CRBN and VHL E3 ligase-mediated degradation while sparing platelets because of the weak expression of CRBN and VHL E3 ligases in human platelets 135,136 . PROTAC technology exhibits advantages of reducing on-target toxicity, which provides insight into minimization of undesirable cardiotoxicity, hepatotoxicity, and hematological toxicity caused by MCL1 depletion via exploration of specific E3 ligases that are rarely expressed in cardiomyocytes, hepatocytes, hematopoietic stem cells, lymphocytes, and neutrophils 28,29,137,138 . These studies have launched the era of the use of PROTACs to target MCL1 in MCL1dominant cancers, although further supporting studies are urgently needed.

Limitations for targeting MCL1
Targeting MCL1 has shown great potential in cancer treatment. However, it is noteworthy that MCL1 also has an important role in early embryonic development 139 and in the survival of multiple normal cell lineages 140 . MCL1 is highly expressed in the myocardium and is crucial for mitochondrial homeostasis and the induction of autophagy in cardiomyocytes 141 . Deficiency of MCL1 in murine hearts leads to rapid and fatal cardiomyopathy 28,29 , and inhibition of MCL1 in human cardiomyocytes results in a severe contractile defect 142 . Furthermore, deletion of MCL1 triggers the loss of hematopoietic stem cells, lymphocytes, and neutrophils 138,[143][144][145] . MCL1 also contributes to the maintenance of hepatic integrity in murine livers 146 , and its absence in murine hepatocytes causes chronic liver damage 137 and hepatocarcinogenesis 147 . Based on these studies, MCL1 depletion may introduce potential unwanted cardiotoxicity, hepatotoxicity, and hematological toxicity, especially in combination with other cytotoxic drugs. Thus, there is a need to identify a therapeutic window in which cancer cells are more sensitive than normal cells to the loss of MCL1. Interestingly, several groups have found that the loss of a single allele of MCL1 can kill c-MYC-driven lymphoma cells, an alteration that is well tolerated in normal cells 45,148,149 . This anti-tumor effect while sparing normal cells by partial MCL1 inhibition may allow the establishment of a therapeutic window for MCL1 inhibitors. In addition, improving therapeutic methods to specifically deliver the inhibitors to cancer tissues will be another option to reduce side effects on normal tissues.

Concluding remarks
Chemoresistance is a severe concern regarding the poor prognosis of cancer patients, and MCL1, an antiapoptotic BCL-2 family member, has become a popular target for cancer treatment due to its important effects. The stalled development of specific MCL1 inhibitors prompted studies of E3 ligases and DUBs modulating MCL1 and the development of their corresponding inhibitors. Considering the rapid turnover of MCL1 and its rigid structure, indirectly targeting MCL1 (e.g., via DUB inhibitors) may represent a promising alternative. Given the number of known MCL1 DUBs and those not yet identified, targeting only one MCL1 DUB seems unlikely to ensure a beneficial effect because the change in MCL1 protein levels may be compensated by other DUBs. Moreover, since DUBs are usually not monospecific for MCL1, offtarget toxicity and loss of therapeutic effects are possible in applications targeting DUBs. However, if the dominant DUB that controls MCL1 stability in a specific context can be identified, it would be a promising target. The additional tumor-suppressing substrates of DUB should also be considered. Thus, the reliability of targeting the USP9X-MCL1 and USP13-MCL1 axes should be further explored in specific cancer subtypes. Another potentially large liability of DUB inhibitors is that none have entered the clinical stage except for VLX1570, which targets UCHL5 and USP14, but this clinical trial has been suspended 150 . However, the development of DUB inhibitors is still very young, and numerous DUB inhibitors are in the pre-clinical stage 150 . Therefore, targeting DUBs upstream of MCL1 shows great potential in future research.