Inhibition of Nucleocytoplasmic Shuttling
Larry H. Bernstein, MD, FCAP, Curator
LPBI
Molecular PathwaysMolecular Pathways: Anticancer Activity by Inhibition of Nucleocytoplasmic Shuttling
, , , , , and
Clin Cancer Res October 15, 2015 21:4508–4513; Published Online Aug 31, 2015; http://dx.doi.org:/10.1158/1078-0432.CCR-15-0408
A dynamic distribution between nucleus and cytoplasm (nucleocytoplasmic shuttling) is one of the control mechanisms adapted by normal cells to regulate the activity of a variety of molecules. Growing evidence suggests that dysregulation of the nucleocytoplasmic shuttling is involved in promoting abnormal cell survival, tumor progression, and drug resistance, and is associated with poor cancer prognosis. Aberrant nucleocytoplasmic shuttling in cancer cells may result from a hyperactive status of diverse signal-transduction pathways, such as the PI3K–AKT and MAPK pathways, or from alterations in the general nuclear import/export machinery. Among the large number of molecules involved in the shuttling process, exportin XPO1, also known as chromosome region maintenance 1, appears to play a particularly prominent role in pathogenesis of both hematological malignancies and solid tumors. Given the importance of nucleocytoplasmic shuttling in cancer pathogenesis and the rapidly expanding knowledge in this field, attempts have been made to develop compounds able to revert the aberrant nucleocytoplasmic shuttling. A promising new drug, KPT-330 (Selinexor), which belongs to the class of XPO1 inhibitors called selective inhibitors of nuclear export, is now being tested in phase I/II clinical trials. Clin Cancer Res; 21(20); 4508–13. ©2015 AACR.
http://clincancerres.aacrjournals.org/content/21/20/4508/F1.small.gif
Figure 1.
Diagram of nucleocytoplasmic shuttling and the effects of XPO1 inhibition. In the cytoplasm, importin (Imp) forms a complex with cargo protein by recognizing its NLS. The complex passes through the nuclear pore complex (NPC) into the nucleus, where cargo is released upon binding of RanGTP to importin. In the nucleus, XPO1 binds to cargo by recognizing the NES, and together with RanGTP, is exported into the cytoplasm. Following the conversion of RanGTP to RanGDP catalyzed by Ran–GTPase-activating protein (RanGAP), the cargo is dissociated from the complex and released into the cytoplasm. However, in the presence of SINE, XPO1 is inhibited and degraded, and is unable to export its cargo proteins. This leads to nuclear accumulation of important TSPs, including p53, p21, p27, and FOXO, resulting in cell-cycle arrest, apoptosis, antiproliferation, and other antitumor activities. RCC1 is a guanine nucleotide exchange factor for RanGTPase, and guides the exchange of RanGDP to RanGTP in the nucleus.
Physical separation of the nucleus from the cytoplasm by the nuclear envelope is a hallmark of eukaryotic cells. The proper spatiotemporal localization of molecules in these two compartments is crucial for cellular homeostasis, and is regulated by a bidirectional transport system channeled through the nuclear pore complex (NPC). NPC provides a selective portal for movementacrossthenuclearenvelope:smallmolecules(<40kDa)can passively diffuse across the NPC, whereas the transport of larger molecules, including most proteins and RNAs, is a receptor- and energy-dependent process (1). Nucleocytoplasmic transport receptors are termed karyopherins, a family of 20 proteins that mediate the shuttling of proteins from cytoplasm to nucleus (importins) or from nucleus to cytoplasm (exportins) by recognizing specific transport signals in the cargo proteins (2; Fig. 1). The best characterized nucleocytoplasmic transport signals include the classical nuclear localization signal (NLS), required for importin-mediated entry into the nucleus, and the leucine-rich nuclear export signal (NES), required for exportin-mediated exit from the nucleus (3, 4). The Ras-relatednuclearproteinsmallGTPaseconfersdirectionalityto the transport process by regulating cargo loading and unloading by the karyopherins (5).
Nucleocytoplasmic shuttling dysregulation in cancer A dynamic subcellular compartmentalization via nucleocytoplasmic shuttling is one of the regulatory mechanisms used by normal cells to modulate the activity of a variety of molecules. Mislocalization of those molecules may alter their activities, thus disturbing the homeostasis of the cells and causing diseases such as cancer. Growing evidence suggests that dysregulation of nucleocytoplasmic shuttling is involved in promoting cancer cell survival, carcinogenesis, tumor progression, and drug resistance (6). Mislocalization of tumor suppressor proteins (TSP) appears to play a key role in cancer pathogenesis. Given that many TSPs execute their antineoplastic functions within the nucleus, mechanisms that enhance their nuclear export and/or cytoplasmic sequestration effectively result in their functional inactivation (7). Likewise, there is also evidence that the activity of oncoproteins can be influenced by their subcellular localization. All these can result from alterations in the shuttling machinery, which is frequently detected in cancer.
Oncogenic signaling pathways and dysregulated nucleocytoplasmic shuttling of TSPs PosttranslationalmodificationsoftheNLSorNESmotifsinthe cargoes, such as phosphorylation, methylation, and ubiquitylation, can modulate binding affinity of the cargoes to specific karyopherins, thus affecting nucleocytoplasmic shuttling (1). Signaling pathways, such as PI3K–AKT and MAPK, are known toplayrolesinsuchmodifications(8),andaberrantactivationof these pathways canlead to mislocalization and functional alterationsofseveralTSPs,includingcyclin–dependentkinaseinhibitor 1A (CDKN1A and p21Cip1), CDK inhibitor 1B (CDKN1B and p27Kip1), forkhead box O (FOXO) proteins, and TP53 (7). The CDKN1A and CDKN1B act as tumor suppressors in the nucleus through inhibition of cyclin–dependent kinases (CDK) duringcell-cycleprogression(9,10).However,they may acquire oncogenic properties when mislocalized in the cytoplasm, leading to increased cell migration and invasion through the inhibition of Rho proteins and their effector Rho-kinase (11, 12). Phosphorylation of CDKN1A by AKT and PKC inhibits CDKN1A nuclear import, whereas CDKN1B phosphorylation by AKT and ERK enhances CDKN1B nuclear export, thereby contributing to their inappropriate cytoplasmic localization (10–12).Such mislocalization has been observed in esophageal, thyroid, colon, breast, and ovarian cancers, and is related to higher histologic grade, advanced stage of disease, and poorer patient survival (9–12).
The FOXO transcription factor family (FOXO1a, FOXO3a, FOXO4, and FOXO6) represent one of the most relevant targets downstream of the PI3K–AKTpathway,whose hyperactivity leads to inhibition of apoptosis and induction of cel lproliferation(13). When localized in the nucleus, FOXO proteins act as tumor suppressors by upregulating the inhibitors of cell proliferation (CDKN1B and RB family member p130), and of survival (BIM, Fas ligand, and TRAIL; ref. 14). FOXO proteins are inactivated by AKT-mediated phosphorylation that promotes their interaction with exportin XPO1 and facilitates FOXO export to the cytoplasm. (15,16).It has been shown that prostate and renal cancer cell lines with a deficiency of PTEN (a negative regulator of the PI3K–AKT pathway) display a constitutive cytoplasmic mislocalization of inactive FOXO1a (17). Importantly, reconstitution of FOXO1a nuclear localization restores its transcriptional activity in PTENnull cells, leading to cell-cycle arrest and apoptosis (17). Also, forced expression of nuclear FOXO proteins has been shown to induce apoptosis in a wide range of cancer cells, including breast cancer and malignant melanoma, although it could also exert paradoxical oncogenic effects in specific tumor histotypes and genetic context (18–24). In human cancer,TP53 is the most frequently inactivated tumor suppressor and nearly half of the cancer cases harbor its loss-offunction mutations or deletions (25). Inactivation of TP53 can also occur due to aberrant nuclear exclusion of the wild-type protein (26). Iendeed, abnormal cytoplasmic overexpression together with lack of nuclear presence of wild-type TP53 have been observed in many tumor types, including inflammatory breast carcinomas, neuroblastomas, retinoblastomas, colorectal, and ovarian cancers (26–30). Data from human neuroblastoma cell lines show that cytoplasmic entrapment of wild-type TP53 is sufficient to cause the loss of TP53-mediated cell-cycle arrest induced by DNA-damaging agents. In these cell lines, restoration of TP53 nuclear localization results in the recovery of its tumor suppressor function (31). Several potential mechanisms underlying cytoplasmic sequestration of TP53 have been described. These include increased nuclear export due to overexpression or hyperactivation of MDM2, mutations in the TP53 NLSs, and overexpression of cytoplasmic proteins able to bind andtrapTP53,suchastheglucocorticoidreceptorandtheparkinlike ubiquitin ligase protein (27, 32, 33). Although TP53 nuclear export is facilitated by the nuclear export receptor exportin-1 (XPO1), TP53, on the other hand, represses XPO1 expression in response to DNA damage in normal cells(34). Overexpression of XPO1 in cancer cells may disrupt this feedback regulatory loop, leading to overly decrease of nuclear TP53 concentration and inadequate DNA damage response.
Alterations of nuclear export receptor XPO1 in tumors Dysregulated nucleocytoplasmic shuttling in tumors can also be caused by an alteration in the transport machinery. A large number of molecules are involved in the shuttling process; nevertheless, alteration of the exportin XPO1, also known as chromosome region maintenance 1, plays a particularly prominent role in tumor pathogenesis. XPO1 is one of seven exportins expressed in mammalian cells, and it exports approximately 220 different proteins as well as a small subset of RNAs from nucleus to the cytoplasm. Interestingly, XPO1 is the sole nuclear export receptor for a large number of TSPs (see Table 1), with a key role in the control of several cancer-related processes, including cell-cycle progression, apoptosis, metastasis, and drug resistance (1).
Table 1. Selected tumor suppressor proteins that are exported from the nucleus by XPO1, and their roles in cancer
Overexpression of XPO1 has been observed in many types of human solid tumors and hematologic malignancies (35–38). Importantly, XPO1 overexpression has usually been correlated with higher tumor grade, more advanced tumor stage, and poor prognosis (34, 37, 39), suggesting its involvement in tumorigenesis. Exogenous overexpression of XPO1 resulted in the transformation of human normal bronchial epithelial cells, whereas its inhibition significantly delayed the growth of A549 NSCLC xenograft tumors in mice(39).Recent experimental data also support a possible role of XPO1 in carcinogen-induced lung cancer development (39). Moreover, whole-genome sequencing analysis revealed somatic mutations in XPO1 in approximately 4% of patients with chronic lymphocytic leukemia (CLL), all affecting the same glutamic residue (E571;ref.40). The recurrent nature of the E571 mutation suggests that it may represent an “oncogenic driver,” with a causative role in CLL leukemogenesis. A different missense mutation (D624G) in XPO1has also been identified in 1 patient with esophageal squamous cell carcinoma (41). Nevertheless, the pathologic role of these XPO1 mutations and the underlying molecular mechanism remain to be elucidated.
Clinical–Translational Advances Given the critical role of XPO1 in nucleocytoplasmic shuttling and in tumor pathogenesis, its inhibition has emerged as a therapeutic strategy in cancer. The rationale behind the targeting of XPO1 is to increase the nuclear concentration of XPO1 cargoes, in particular of tumor suppressor gene products. It must be noted, however, that XPO1 also plays a role in RNA export and in mitotic processes, such as microtubule nucleation at kinetochores (42). Thus, interference with these XPO1 functions may also contribute to the therapeutic effect of XPO1 inhibition.
The first XPO1 inhibitor discovered was Leptomycin B (LMB), a natural compound isolated from the bacteria Streptomyces species ATS1287 (43). It is an irreversible inhibitor that covalently binds a cysteine residue (C528) in the cargo-binding region of XPO1, preventing cargo interaction with XPO1 (43). LMB has potent antitumor activity in vitro; however, it only induced a transient reduction of tumor biomarkers, such as cancer antigen 125 (CA-125), and human chorionic gonadotrophin in patients with ovarian carcinoma and trophoblastic tumor, respectively, and one stable disease in a sarcoma patient in a phase I trial (44). The severe toxicity profile of LMB has prevented its further clinical development (44). Toxicities have been attributed to off-target effects due to its binding to several cysteine proteases, in addition to the irreversible inhibition of XPO1 (45).
Subsequently, several natural products such as Ratjadone as well as synthetic compounds (KOS-2464, PKF050-638, and CBS9106) have been developed. All these compounds inhibit XPO1 by binding its C528 residue, causing cell-cycle arrest and apoptosis in a time- and dose-dependent manner in a broad spectrum of cancer cells. These inhibitors are more clinically relevant because they are much less toxic than LMB, while maintaining high potency (46–48).
The newer additions of XPO1 inhibitors are a group of small molecule compounds called selective inhibitors of nuclear export (SINE), including KPT-115, KPT-127, KPT-185, KPT-251, KPT276, KPT-330 (Selinexor), and KPT-335 (Verdinexor). Unlike LMB, SINEs bind reversibly the C528 residue in XPO1, with virtually no off-target activity. KPT-330 (Selinexor) has an IC50 of about 20nmol/L for XPO1inhibition, but has minimal activity (>10 mmol/L) against 114 other proteins, including enzymes, receptors, transporters, ion channels, and other cysteinyl-active site kinases and proteases (45).
SINE compounds have been shown to inhibit nuclear export of many TSPs with key roles in genomic stability and DNA repair (TP53, TP73, and BRCA1), cell-cycle control (pRB1, CDKN1A, and CDKN1B), and apoptosis [FOXO proteins, adenomatous polyposiscoli (APC) protein, nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (IkBa)] in cancer cell lines and tumor biopsies (49–56; Fig. 1). Notably, preclinical data suggest that the cytotoxic effect of XPO1 inhibitors in different tumor types may specifically rely on nuclear entrapment of certain XPO1 cargoes with particular biologic activities. For example, the activity of SINE compounds in some breast cancer cell lines seems to depend largely ontheinhibitionofsurvivinshuttling(49).Survivin,amember of the inhibitors of apoptosis protein family, is highly expressed in breast cancer cells, and its cytoplasmic levels have been shown to be an independent predictor of poor prognosis, whereas its nuclear levels were associated with a favorable prognosis (55). Survivin nuclear export exclusively relies on XPO1, and a significant depletion of cytoplasmic survivin can be induced by XPO1 inhibition. The proapoptotic activity of SINEs in breast cancer cell lines largely depends on inhibition of survivin shuttling, and restoring survivin levels by ectopic overexpression is sufficient to impair the proapoptotic effect of SINEs (49). In TP53 wild-type NSCLC cells, however, the antiproliferative effects of SINEs mainly depends on the increase of nuclear TP53, because concomitant silencing of TP53 expressionorits inhibition by pifithrin-alpha causes SINE resistance. Interestingly, XPO1 inhibition also displays a cytotoxic activity in a TP53 mutated lung cancer cell line. In this case, the activity of SINEs seems dependent on TP73, a member of the TP53 faat is also involved in DNA damage induced cell-cycle arrest and apoptosis and regulates TP53 dependent genes in TP53-deficient cells (50).1 fusion oncoprotein. ABL protooncogene 1, nonreceptor tyrosine kinase (ABL1) acts in the nucleus as a tumor suppressor in normal cells, whereas BCR– ABL1 in the CML cells is constantly exported by XPO1 to the cytoplasmt
Altered nucleocytoplasmic shuttling may contribute to drug resistance as well, and in this regard, XPO1 inhibitors have shown synergistic anticancer activity when used in combination with chemotherapy or targeted therapeutic agents. Aberrant nuclear export of topoisomerase II is one of the mechanisms of resistance to doxorubicin and etoposide in myeloma cells, and such resistance can be reverted, the concomitant inhibition of XPO1(57). In chronic myeloid leukemia (CML) cells with t(9;22) chromosome translocation, LMB is able to revert their acquired resistance to imatinib that targets BCR–ABLo exert its mitogenic and antiapoptotic activities(57).
Interestingly, when BCR–ABL1 protein is forced into the nucleus byXPO1 inhibition, it retains the proapoptotic functions like that of ABL kinase. Further, imatinib and LMB combination induce IC50 m ore than 5 to 20 mmol/L. Continuous (72 hour) exposure of non-neoplastic cells to SINEs at nanomolar concentrations only induces cell-cycle arrest without apoptosis (45, 53). The differential effect of SINEs on normal and neoplastic cells is not yet fully understood. However, one plausible explanation is that restoration of TSPs in the nucleus by XPO1 inhibition triggers the apoptotic pathways in response to extended DNA damage accumulated in the neoplastic cells.
So far, KPT-330 (Selinexor) is the only XPO1 inhibitor in the phase I/II clinical trials. Compared with LMB, Selinexor showed a much better toxicity profile. Most adverse events in patients with solid tumors and hematologic malignancies were reversible grade 1 and 2, primarily nausea, anorexia, and fatigue. Among 106 patients evaluable for response, an overall disease control rate of 49% with some partial responses were observed in colorectal, melanoma, ovarian, and cervical cancer (58–61). Preliminary results of an ongoing phase II clinical trial evaluating the activity of single-agent Selinexor in patients with heavily pretreated, progressive gynecologic cancers, showed promising antitumor activity across ovarian, endometrial, and cervical cancers. The disease control rate was up to 52% (33 of 63 patients), with several patients remaining on study for up to 12 months (61). Durable responses and disease stabilization with single-agent Selinexor werealso observedinhematologicmalignancies across all disease subtypes, with some patients remaining on study for over 1 year (59, 60). Currently, 36 clinical trials with Selinexor were registered at the Clinical Trials.Gov database (http://clinicaltrials.gov/ct2/home).
Conclusions Nucleocytoplasmic shuttling has been identified to have a role in cancer pathogenesis, and this has led to the development of new therapeutic strategies to revert its alterations. Particularly, the inhibition of XPO1, leading to nuclear retention and functional reactivation of TSPs, is the most advanced therapeutic strategy. Numerous new drugs have been developed and among them, Selinexor is currently being evaluated in phaseI/II human clinical trials with promising preliminary results in hematologic malignancies and solid cancers. Although interfering with nucleocytoplasmictransportmachinerycouldbedetrimentaltoallcells,the new XPO1inhibitors have beenshownto preferentially suppress or eliminate tumor cells, relatively sparing normal cells. Despite significant progress, several crucial questions remain unresolved. Patient selection appears challenging, as we do not know which cargoes are important, and these may vary from tumor type to tumor type and even from patient to patient. Furthermore, therapeutic efficacy of XPO1inhibitorsishampered by intrinsic and acquired resistance, as evidenced by the preliminary results from phase I/II clinical trials. Elucidation of the resistant mechanisms will be necessary for the development of sound combination strategies. Nonetheless, growing data clearly show that targeting the nucleocytoplasmic shuttling is a worth strategy to pursue.
Selective inhibitors of nuclear export show that CRM1/XPO1 is a target in chronic lymphocytic leukemia
Rosa Lapalombella1,*, Qingxiang Sun2,*, Katie Williams1, Larissa Tangeman1, Shruti Jha1, Yiming Zhong1, Virginia Goettl1, Emilia Mahoney1, et al.
Blood Nov 29, 2012; 120 (23):4621 – 4634 DOI: http://dx.doi.org/10.1182/blood-2012-05-429506
The nuclear export protein XPO1 is overexpressed in cancer, leading to the cytoplasmic mislocalization of multiple tumor suppressor proteins. Existing XPO1-targeting agents lack selectivity and have been associated with significant toxicity. Small molecule selective inhibitors of nuclear export (SINEs) were designed that specifically inhibit XPO1. Genetic experiments and X-ray structures demonstrate that SINE covalently bind to a cysteine residue in the cargo-binding groove of XPO1, thereby inhibiting nuclear export of cargo proteins. The clinical relevance of SINEs was explored in chronic lymphocytic leukemia (CLL), a disease associated with recurrent XPO1 mutations. Evidence is presented that SINEs can restore normal regulation to the majority of the dysregulated pathways in CLL both in vitro and in vivo and induce apoptosis of CLL cells with a favorable therapeutic index, with enhanced killing of genomically high-risk CLL cells that are typically unresponsive to traditional therapies. More importantly, SINE slows disease progression, and improves overall survival in the Eμ-TCL1-SCID mouse model of CLL with minimal weight loss or other toxicities. Together, these findings demonstrate that XPO1 is a valid target in CLL with minimal effects on normal cells and provide a basis for the development of SINEs in CLL and related hematologic malignancies.
Introduction
Multicellular organisms have evolved a complex and overlapping array of proteins/pathways that function to “guard the genome” and prevent genesis of neoplastic clones. These proteins, referred to as tumor suppressor proteins (TSPs) and growth regulatory proteins (GRPs), act primarily in the nucleus. CRM1/XPO1 (chromosome region maintenance 1 protein, also called exportin1 or XPO1 in humans) is the best-characterized nuclear exporter, and transports more than 200 proteins and certain RNA species from the nucleus to the cytoplasm.1,2 XPO1 binds to a diverse array of protein cargos through their canonical leucine-rich nuclear export signals (NESs) domain. The NESs are 10- to 15-residue motifs containing 4 or 5 spaced hydrophobic amino acids, which form combined α-helix-loop or all loop structures that bind to the hydrophobic groove of XPO1.1,3⇓⇓–6 XPO1 and cargo form a ternary export complex with RanGTP in the nucleus, which is then translocated through the nuclear pore complex.2,7 In the cytoplasm, cargo is released from XPO1 through the combined action of GTPase regulators RanGAP and RanBP1. XPO1 cargo proteins include numerous TSPs and GRPs, such as p53, FoxO3a, and the endogenous inhibitor of NF-κB, IκB. By exporting these proteins from the nucleus of normal cells, XPO1 prevents them from acting in the absence of DNA damage or other oncogenic insults.8,9 More than 14 distinct TSP/GRP pathways have been identified to be exported by XPO1 in an exclusive fashion to date, and many of these coexist in different types of cancer that continue to be defined.10
Elevated expression or dysfunction of the XPO1 have been reported in various hematologic and solid tumors, and have been correlated with poor prognosis and resistance to therapy.4,8⇓⇓–11 For example, mutation of the TSP nucleophosmin (NPM1) has been reported in a specific subgroup of cytogenetically normal acute myeloid leukemia (AML)12 in which, a gain-of-function mutation in the C-terminus of the NPM1 creates a novel NES and leads to greatly enhanced and unregulated binding to XPO1. In NPM1-mutated AML (NPM1c), enhanced XPO1-mediated transport of NPM1 removes it from the nucleus (and nucleolus), rendering it oncogenic; thus, NPM1c is believed to be a leukemia initiation mutation in this subset of AML.13 This example attests to the importance of nuclear-cytoplasmic transport in the development of leukemia.12,14 Similarly, activated oncogenic signaling pathways can lead to inappropriate phosphorylation and other posttranslational modifications of TSPs and GRPs, rendering the modified proteins susceptible to XPO1-mediated nuclear export.15 Thus, XPO1 is a nodal point by virtue of its nonredundant gate-keeping function, exclusively controlling the directional exodus of TSPs/GRPs from the nucleus to the cytoplasm.
Chronic lymphocytic leukemia (CLL) is the most prevalent type of adult leukemia and is incurable with current therapies. Unlike chronic myeloid leukemia or hairy cell leukemia, CLL does not have a common translocation or mutation that drives the pathogenesis of the disease. CLL tumor cells are highly dependent on the microenvironment where cytokines (eg, CD40L, BAFF, IL-4, IL-6), and contact (eg, stromal cells) promote cell activation and proliferation, and also resistance to spontaneous and drug-mediated apoptosis. Many of these microenvironment-activated pathways merge with TSPs exported by XPO1. XPO1 is therefore a highly attractive molecular target to explore in CLL, because it impacts multiple antitumor and growth suppressive signaling pathways that are dysregulated in this disease.
We therefore hypothesized that a selective XPO1 inhibitor would show efficacy with an acceptable therapeutic index in CLL and other diseases. Indeed, XPO1 inhibition in normal cells (ie, possessing an intact genome) leads to transient cell cycle arrest without cytotoxicity, followed by fast recovery after the drug is removed.16,17 To date, efforts to clinically pharmacologically inhibit XPO1 have been unsuccessful because of off-target effects.18⇓⇓–21 A selective XPO1 antagonist may allow targeting of the TSPs axes in tumor cells.
In this report, we describe the design, via in silico docking methods that were based on an earlier structure activity relationship study,22 of small molecule drug-like selective inhibitors of nuclear export (SINEs) that irreversibly bind to and block XPO1, and demonstrate that XPO1 is a valid target in CLL with minimal effects on normal cells providing the basis for the development of SINEs in CLL and related hematologic malignancies.
……
Results
KPT-185 binds in the NES-binding groove of XPO1
SINEs, developed by Karyopharm, are small molecules designed in silico to covalently modify a cysteine (Cys528) and participate in numerous noncovalent interactions in the NES-binding groove of human XPO1. The lead compounds KPT-185 and KPT-251 share similar warheads but present distinct pharmacokinetic properties in vivo because of differences in their side chains (Figure 1A).
We have solved the 2.1 Å X-ray structure of XPO1 bound to KPT-185 (Figure 1B, Table 1). The crystals contained KPT-185 bound to the ternary complex of Saccharomyces cerevisiae XPO1 (ScXPO1), S cerevisiae RanBP1 (ScRanBP1), and human RanGTP (Table 1). Because ScXPO1 has a threonine residue Thr539 in place of the reactive Cys528 in human XPO1, Thr539 was mutated to cysteine to enable covalent modification by KPT-185 and the T539C mutant of XPO1 is named ScXPO1*. Atomic coordinates and structure factors have been deposited in the protein data bank under RCSB ID code rcsb074384 and PDB ID code 4GMX.
The overall structure of the KPT-185-ScXPO1*-Ran-RanBP1 complex is similar to the previously reported inhibitor-free ScXPO1-Ran-RanBP1 structure (Cα rmsds of ∼ 0.65 Å).7 The ring-shaped XPO1 protein contains 21 tandem HEAT repeats (designated H1-H21), each composed of a pair of antiparallel helices A and B. The N-terminal half of XPO1 wraps around RanŸGppNHp, which in turn wraps around RanBP1 with its C-terminal extension (Ran residues 177-216; Figure 1B). KPT-185 binds in the NES-binding groove, which is located on the central, convex side of the XPO1 ring (Figure 1B-C, supplemental Figure 1A). In the absence of inhibitor, the NES-binding groove of ScXPO1 is closed in the ScXPO1-Ran-RanBP1 complex (Figure 1D, supplemental Figure 1B).7 In our structure, the NES groove has opened to accommodate KPT-185 (Figure 1E). Interestingly, interactions between KPT-185 and XPO1 are almost entirely of hydrophobic nature (Table 2). The methoxy, carbonyl, and ester groups of KPT-185 do not seem to make any polar contacts with XPO1. The tri-fluoromethyl group of KPT-185 is buried deep in the XPO1 groove, whereas its methoxy group reaches toward the groove opening (Figure 1C). Fluorine atoms F1, F2, and F3 are buried in the XPO1 groove through numerous hydrophobic interactions with several XPO1 sidechains including Ile555, Met556, and Val559 that line the floor of the NES-binding groove (Figure 1C, Table 2).
The regions of human/mouse and yeast XPO1 proteins that form the NES-binding grooves share 81% sequence identity and almost all XPO1 residues involved in NES and inhibitor-binding are strictly conserved suggesting that mammalian and yeast XPO1 grooves likely bind ligands in very similar fashion (Figure 2A) that of XPO1 bound with the NES from protein kinase A inhibitor (PKIα).3,5,6 XPO1 helices at the grooves of both ScXPO1-KPT-185 and mouse XPO1-PKIαNES structures superimpose with a Cα rmsd of 1.2 Å. Examination of the grooves show the KPT-185–bound XPO1 groove to be narrower and deeper that the NES-bound groove (Figure 2B-C). A slight reorientation helix H11A and several sidechain rearrangements accompany the structural shift from NES to inhibitor binding (Figure 2D-E). Structures of inhibitor-free, inhibitor-bound, and NES-bound XPO1 grooves clearly indicate that the NES-binding groove is conformationally quite plastic. Interestingly, the trifluoromethyl phenyl of KPT-185 penetrates much deeper into the groove than the NES sidechains, possibly contributing to the potency of the compound in outcompeting nuclear export cargos (Figure 2B-E).
To investigate the effects of KPT-185 on XPO1-cargo complexes, we performed pull-down inhibition assays using purified recombinant human XPO1, and molar excesses of RanGTP and NESs from HIV1-REV and snurportin-1 (SNUPN) immobilized on glutathione sepharose (Figure 2F). Human XPO1 was preincubated with either leptomycin B (LMB)28,29 or KPT-185. Both LMB and KPT-185 inhibited the formation of XPO1-cargo complex. Furthermore, a similar SINE KPT-251 was also able to inhibit XPO1 mediated HIV-Rev nuclear export in U2OS cells stably expressing RevGFP (supplemental Figure 2A). A large panel (50) of in vitro protein binding assays was performed to evaluate the potential interaction of KPT-251 and KPT-185 with other proteins. At a concentration of 10μM, both compounds show exquisite specificity for XPO1 and no detectable binding to other proteins, including the cysteine proteases believed to be the cause of poor tolerance to LMB (data not shown). Given these results, SINEs are considered to be highly selective agents (> 100-fold compared with inhibition of XPO1-mediated HIV Rev transport of 100nM, supplemental Figure 2A). The effect of SINEs on XPO1 interaction in HeLa cells cotransfected with Rev target with blue fluorescent protein (BFP) and either wild-type or XPO1(C528S) mutant human XPO1 tagged with yellow fluorescent protein (YFP) was assessed.22,30 When coexpressed, a significant fraction of both wild-type XPO1-YFP (Figure 2G) or XPO1(C528S)-YFP (Figure 2H) colocalized with Rev in the nucleoli, suggesting an interaction between the 2 proteins. On treatment with KPT-185 or KPT-251, the Rev-dependent nucleolar localization of wild-type XPO1-YFP but not XPO1(C528S)–YFP is abolished confirming that the Cys528 in XPO1 is required for SINEs to disrupt the XPO1 binding to Rev cargo.
XPO1 inhibition induces selective cytotoxicity in CLL cells
Expression of XPO1 in primary CLL cells9,31 and control normal B cells was examined. Immunoblot analysis showed XPO1 to be overexpressed in CLL cells compared with normal B cells at the protein (Figure 3A-B) and mRNA level (Figure 3C). As XPO1 is a recycled transporter, even modest increases in its levels might have a marked effect on the subcellular localization of cargo proteins.
The effect of inhibiting either its expression or its activity was investigated using siRNA or SINE compounds. CLL cells were transiently transfected with a XPO1 siRNA and its effect on cell death was evaluated. Analyses of XPO1 expression by real-time RT-PCR and immunoblot (supplemental Figure 3A-B) showed that the gene knockdown, although modest, resulted in a significant reduction in cell viability relative to the missense control (supplemental Figure 3C).
The cytotoxic effect of KPT-185 against primary CLL cells was next evaluated and compared with that induced by LMB. CLL cells were incubated with increasing concentrations of KPT-185 or LMB (ranging from 0.01μM to 10μM) for 24, 48, 72, and 96 hours. KPT-185 induced significant time and dose-dependent cytotoxicity (Figure 3D) as measured by MTS conversion (EC50 < 500nM). Cell death was observed as early as 24 hours and continued to increase up to 96 hours; a 72-hour time point was chosen for all the subsequent experiments. Cell death was confirmed by annexin/PI flow cytometry in cells treated with 1μM KPT-185 (data not shown). KPT-185 induced superior cytotoxic effect compared with LMB (supplemental Figure 3D). Consistent with the irreversible mechanism of action of SINEs, short exposures (as little as 1 hour) were sufficient to promote apoptosis (supplemental Figure 3E). An isomer of KPT-185,10 was used to confirm that the cytotoxic effect seen on CLL cells is because of the specific inhibition of XPO1 and not to an off-target effect (supplemental Figure 3F-G). KPT-251 and KPT-185 on CLL cells were next compared and found to be equally effective in inducing apoptosis of CLL cells (Figure 3E). The effects of KPT-185 on PBMC and normal B cells from healthy volunteers were then assessed. KPT-185 produced only modest apoptosis (estimated EC50 > 40μM, Figure 3F) at 72 hours in normal PBMCs and B cells compared with CLL cells (EC50 ∼500nM; Figure 3G), suggesting that transformed B cells are more sensitive to KPT-185 treatment than normal B cells. As CLL (and normal B cells) are not cycling, these data indicate that apoptosis induction by SINE is not cell cycle–dependent. Although significant cytotoxicity was observed with KPT-185 treatment, the variability in patient CLL cell response was marked (5% to 90% at 0.5μM). Traditional CLL prognostic factors were therefore examined to determine whether they would predict response to KPT-185. Cytogenetic 12q, 11q, or 13q abnormalities did not confer differential sensitivity to KPT-185 induced cell death, whereas 17p deletions (associated with reduced p53 expression) were associated with reduced overall sensitivity (Figure 3H). Interestingly, when 17p deletions were divided into those with unmutated and mutated IVGH, only the latter subset of del(17p) samples showed reduced sensitivity to KPT-185 induced death (Figure 3I). Therefore, IVGH mutational status was examined for differences in response to KPT-185, as it has a strong influence on not just chemotherapy response but also on progression-free survival associated with standard therapies used to treat CLL.32 In contrast to other therapies in CLL, a significant increase in sensitivity to KPT-185 in patient cells with unmutated IVGH was found compared with those with mutated IVGH (Figure 3J). Interestingly, although the presence of del(17p) in patients with IVGH unmutated status did not alter the cytotoxicity levels of KPT-185, the presence of the same deletions in patients with IVGH mutated status significantly reduced the cytotoxicity effect of KPT-185. These data suggest that KPT-185 may have more clinical activity in the unfavorable IVGH unmutated CLL subset, and may also be active in CLL with 17p deletions that have IVGH unmutated disease. Cytotoxicity induced by SINEs was determined to be caspase-dependent, as evidenced by cleavage of the caspase-3 substrate PolyADP ribose polymerase (PARP) and inhibition of cytotoxicity by the caspase inhibitors Q-VD-OPH and BOC-D-FMK (Figure 3K-L).
SINEs-specifically inhibit nuclear export
Human CLL cells exhibit dysregulated growth and TSP pathways such as constitutive active AKT33 and NF-κB,34 as well as functional loss of p53 activity.35 As the nuclear export of factors involved in each of these pathways is mostly mediated by XPO1, treatment of CLL with KPT-185 could restore normal regulation of these pathways by forcing nuclear retention of FoxO3a (counters AKT/PI3K), IκB (counters NF-κB), and p53, thereby inducing the death of CLL cells. As shown in Figure 4, treatment of CLL cells with KPT-185 led to strong accumulation of these proteins in the nucleus in a time-dependent manner with the maximum effect observed at 12 hours as revealed by confocal microscopy (Figure 4A, supplemental Figure 4A-C). Results were confirmed by immunoblot analysis (Figure 4B) of lysates derived from DMSO or KPT-185 treated CLL cells.
IκB is a potent endogenous inhibitor of NF-κB, a transcription factor with inflammatory, antiapoptotic activity that is constitutively active in CLL.34 KPT-185 induced IκB nuclear accumulation, allowing it to complex with nuclear NF-κB and reduce the DNA binding capacity of NF-κB (Figure 4C). Interestingly, KPT-185–enforced nuclear retention of IκB leads also to depletion of NF-κB p50 and p65 (Figure 4D) therefore reducing NF-κB function in CLL. Among its many functions, NF-κB has been shown to up-regulate Mcl1 the most critical survival factors for CLL cells.36 Interestingly, KPT-185–enforced nuclear retention of IκB leads to Mcl1 depletion in CLL cells (Figure 4E-F). Similarly, additional NF-kB target genes such as Bcl-xL were also reduced after treatment of CLL cells with KPT-185 (supplemental Figure 4D).
SINEs antagonize microenvironment stimuli
CLL tumor cells are known to receive a variety of survival signals from the microenvironment that confer them resistance to spontaneous apoptosis as well as to chemotherapy.37 Therefore, the ability of KPT-185 to induce cytotoxicity of CLL cells in the presence or absence of soluble factors known to reduce the spontaneous apoptosis associated with CLL cells (TNF, IL-6, and IL-4) or induce activation of key signaling pathways (CD40L and BAFF) was examined. As shown in Figure 5A through F, each of these factors significantly reduced the spontaneous apoptosis associated with CLL cells and cotreatment with SINEs abrogated this protection. Interestingly, the cytotoxic effect elicited by KPT-185 was enhanced in CPG-activated cells. The survival benefit of CLL in vivo is not only influenced by soluble factors such as those previously discussed, but also by cocontact with a variety of cells composing the bone marrow and lymph node microenvironment.38Therefore, the efficacy of SINEs in the presence of stromal protection was investigated using the human marrow-derived fibroblast cell line HS-5 that enables long-term survival of primary human B cells and B-CLL cells ex vivo.39 Direct treatment of the HS-5 stromal cells with SINEs for 72 hours had no effect on viability (supplemental Figure 5A). CLL patient cells were incubated with either DMSO, KPT-185, or KPT-251 for 12 hours before washing and plating in flasks with or without HS-5 for a total of 60 hours (Figure 5G). Alternatively CLL cells with or without HS-5 were continuously treated with KPT-185 or KPT-251 for 48 hours (Figure 5H). As expected, coculture of untreated CLL cells on the HS-5 stromal cell line resulted in reduction of spontaneous apoptosis (supplemental Figure 5B), and cells treated with KPT-185 or KPT-251 without HS-5 coculture exhibited apoptosis (Figure G-H). However, the prosurvival effect of HS-5 was unable to effectively prevent SINEs induced apoptosis; in fact, the cytotoxic effect mediated by KPT-185 was enhanced under stromal coculture conditions (Figure 5G). These results provide important evidence that KPT-185 may evade the protective effects of the CLL cell microenvironment counteracting multiple oncogenic and growth potentiating signals and therefore providing an advantage over other therapeutics used in the treatment of this disease.
SINEs do not alter T cell or NK cell viability but negatively influence IL-6 and IL-10 production
CLL is associated with immune suppression that is often augmented by therapeutics used to treat the disease. The influence of KPT-185 on T cell and natural killer (NK) cell viability and function was therefore investigated. The viability of naive T cells, CD3 activated T cells, and NK cell was minimally influenced by SINE treatment (Figure 6A-B). Cytokine production by CD3-activated T cells demonstrated no difference in tumor necrosis factor-α (TNF) production, whereas production of both IL-6 and IL-10 was diminished by KPT-185 (Figure 6C-E). Neither antibody-dependent cellular cytotoxicity (Figure 6F) nor direct cytotoxicity (Figure 6G) mediated by NK cells was affected by KPT-185 or by KPT-251. Collectively, these studies suggest that SINEs have minimal effects on normal immune cells with respect to viability or NK cell–mediated killing but may impact both inflammatory (IL-6) and immunosuppressive (IL-10) cytokines linked to CLL pathogenesis.
SINEs prolong survival in a mouse model of CLL
The in vivo significance of SINE inhibition of XPO1 was studied using the Eμ-TCL1-SCID transplant model of CLL.40 Eμ-TCL1 mice develop disease very similar to that of CLL patients including activation of the AKT pathway, elevated Igκ+ B cells, splenomegaly, and infiltration of malignant B-lymphocytes to the liver, lungs, and kidney.27 CD19+ leukemia cells from these mice were engrafted into SCID mice.40 These cells were also tested to confirm the expression of XPO1 and the sensitivity to SINEs and fludarabine. Unlike KPT-185, with poor systemic pharmacokinetic (PK) properties including minimal oral bioavailability in mice, KPT-251 displayed improved PK in mice and good oral availability, allowing in vivo experiments with oral administration (Table 3). Considering that both compounds present similar selectivity and induce similar levels of in vitro cytotoxicity of CLL (Figure 4E) and murine TCL1+ cells (Figure 7A) the in vivo experiment was conducted using KPT-251.
Mice were treated (14 days after engraftment) with vehicle, 75 mg/kg KPT-251 5 d/wk for 2 weeks by oral gavage and then QoDx3/wk until the end of the study. Fludarabine 34 mg/kg 5 d/wk every 4 weeks intraperitoneally was used as control because TCL1 leukemic cells have been shown to have wild-type p53 and respond to fludarabine both in vitro and in vivo.27 Dose and time schedules were chosen based on PK data derived from CD1 mice receiving a single dose of KPT-251 (Tables 4–5). The primary end point of the study was overall survival. Mice treated with KPT-251 showed a significant improvement in survival over both vehicle and fludarabine treated mice (Figure 7B). The secondary end point was progression free survival (PFS), defined as increase in circulating CLL (CD19+/TCL1+) cells to > 20 000/μL. KPT-251 showed a significant improvement in PFS compared with both vehicle and fludarabine (Figure 7C). In addition, KPT-251 was well tolerated in mice, resulting in moderate loss in body weight, (≤ 10%) that was reversed by the end of the study (Figure 7D). An analysis of peripheral blood lymphocytes (PBLs) at week 8 showed that KPT-251 significantly prevented an increase in circulating CLL cells compared with both vehicle and fludarabine (Figure 7E).
To further validate KPT-251 in mice with leukemic phase, 20 additional C.B-17 SCID mice were engrafted with CD19+ TCL1 leukemia cells and treatment was initiated 70 days (week 10) after engraftment. Mice were treated with vehicle or 75 mg/kg KPT-251 (QoD×3/wk). Mice treated with KPT-251 had a significant survival advantage over vehicle-treated controls (Figure 7F). Moreover, KPT-251 significantly prevented an increase in circulating CLL cells compared with vehicle (Figure 7G).
To determine the in vivo relevance of the in vitro pharmacodynamic studies in primary human CLL cells, 27 additional Eμ-TCL1-SCID mice were left untreated until disease developed, as defined by circulating PBLs ≥ 30 000/uL. Mice were than randomized to receive a single dose of KPT-251 or vehicle control (9 mice/group). Three mice for each group were sacrificed at 1, 3, or 5 days posttreatment, and protein and mRNA expression were analyzed in tumor cells isolated from mice. PBLs count was also monitored at the time of treatment and when the mice were sacrificed. Figure 7H shows that a single dose of KPT-251 significantly prevented an increase in circulating CLL cells compared with vehicle for all the analyzed time points. More importantly, the reduced PBL count correlates also with an increased level of p53, FoxO3a, and IκB in the nuclei of KPT-251 but not in vehicle-treated cells (3 days, Figure 7I). Similar to the results in vitro, Mcl1 was also down-modulated after KPT-251 treatment in vivo (supplemental Figure 6). In summary, the effects of SINEs observed in vitro also were observed with a single dose of KPT-251 in vivo. These data show that KPT-251 represents a novel therapeutic agent that targets XPO1 in the Eμ-TCL1-SCID CLL model and provide support for clinical development in CLL and related lymphoproliferative disorders.
Discussion
The development of cancer is a multistep process generally involving dysfunction of multiple tumor suppressing proteins that are either silenced or compartmentally localized to the cytoplasm where they are ineffective at detecting genomic damage and, when appropriate, promoting cell death. XPO1 is a major nuclear export protein involved in externalizing multiple TSP, and is overexpressed or mutated in a variety of cancers including CLL. XPO1 cargos include numerous targets including tumor suppressors, and cell cycle inhibitors such as p53, FoxO, topo IIα, and IκB.41 The increased export of these proteins from the nucleus has been implicated in cancer disease progression and drug resistance. It has been shown that blocking XPO1–mediated nuclear export of any or all of these proteins by siRNA or XPO1 inhibitors may restore apoptotic pathways and tumor cell sensitivity to chemotherapeutic drugs such as doxorubicin,42 etoposide,42 cisplatin,43 and imatinib mesylate. Therefore XPO1 export inhibitors have the potential to be used as both single agents and in combination with current chemotherapeutic drugs.
CLL is characterized by disrupted apoptosis caused by aberrant activation of several signaling/transcriptional pathways that promote survival (eg, PI3K/AKT, Wnt/β-catenin and NF-κB). Therefore, a therapeutic strategy simultaneously targeting multiple death and antioncogenic pathways disrupted in this disease may have broad application for many subsets of patients.
Consequently, XPO1 represents a highly attractive molecular target in CLL, because it impacts multiple signaling pathways that are dysregulated in this disease. Published data have established XPO1 as a validated cancer target in solid tumors8,9,11,18,44; however, previous attempts to pharmacologically manipulate XPO1 have been unsuccessful because of off-target effects. Several irreversible non–drug-like inhibitors that bind covalently to Cys 528 in the NES-binding groove of human XPO1 have been reported, including the natural products leptomycin B (LMB), ratjadone C, anguinomycin, goniothalamin, along with the small molecule drug-like N-azolylacrylates.22,45 Recently, a novel reversible oral XPO1 inhibitor with XPO1 degrading activity (CBS9106) has also been reported.46
LMB is the most extensively studied XPO1 inhibitor, and is a widely used biologic tool to define XPO1-mediated protein export. LMB was shown to be active preclinically in several solid tumor and hematologic tumor models18,19,21,42 but was associated with a low therapeutic index in mouse studies because of off-target gastrointestinal effects, as well as profound dose-limiting anorexia, fatigue, and gastrointestinal effects when introduced in a phase 1 study when given intravenously.20 In this trial neither target validation of XPO1 inhibition nor etiology of nausea/emesis and fatigue were adequately addressed.19,20 Semisynthetic derivatives of LMB with improved pharmacologic properties nearly eliminated the toxicities in mice, suggesting that at least some of the LMB toxicities were not mediated by XPO1 inhibition.18
Our work documents the creation of novel, orally bioavailable selective and irreversible inhibitors of XPO1-mediated nuclear export that bear a favorable therapeutic index to transformed tumor cells compared with normal cells. The SINEs compounds show exquisite specificity for XPO1 and no detectable binding to other proteins, including the cysteine proteases believed to be the cause of poor tolerance to LMB. The high resolution crystal structure of XPO1 bound to KPT-185 validates conjugation of KPT-185 to the cysteine in the cargo-binding groove of XPO1 (Cys528) and explains its potency in inhibiting XPO1-cargo interactions and nuclear export. More importantly, KPT-251 has pharmacokinetic and pharmacodynamic properties, including oral bioavailability that are superior to LMB and allow its use in vivo. SINEs induce apoptosis of CLL cells with a favorable therapeutic index, with enhanced killing of genomically high-risk CLL cells that are typically unresponsive to traditional therapies. The amount of cytotoxicity induced by SINEs did not correlate with the level of expression of XPO1. SINEs restored normal regulation to the majority of the dysregulated pathways in CLL by forcing the nuclear retention of key TSPs such as FOXO, IκB, and p53 both in vitro and in vivo. Among its many functions, NF-κB has been shown to play a role in the up-regulation of Mcl1 the most significant antiapoptotic protein associated with normal as well as malignant B lymphocytes.36 High levels of Mcl1 mRNA and protein have been found in CLL, which are inversely correlated with in vitro response to chemotherapeutic agents or with the failure of CLL patients to respond to fludarabine, chlorambucil, and rituximab therapy in vitro and in vivo.47⇓–49 Therapeutically, down-regulation of Mcl1 protein expression by antisense oligonucleotides or through indirect Mcl1 transcription and translation inhibitors results in cell death during in vitro culture or in vivo therapy.47,50,51 Interestingly, KPT-185 induces depletion of Mcl1 message and protein in CLL cells, probably because of the inactivation of NF-κB and the sensitivity of patients’ samples to KPT-185 correlates with the amount of down-modulation of Mcl1. Similarly, a significant reduction of Mcl1 mRNA was also observed on XPO1 down-modulation using siRNA strategy. Based on our data showing that KPT-SINEs modify nuclear level of TSPs, such as p53, important to resistance to traditional therapies, the interaction of KPT-SINEs treatment with traditional p53-dependent therapies used in CLL, (ie, fludarabine, chlorambucil) warrants future study.
KPT-SINE has been previously shown to selectively kill acute leukemia cells compared with PBMCs and CD34+ progenitor cells in vitro.52 Data presented here further support this observation indicating that SINEs possess tumor-cell selectivity, with only weak effects on normal PBMCs. Moreover, KPT-185 evades the protective effects of the CLL cell microenvironment providing an advantage over other therapeutics used in the treatment of this disease. The mechanism by which KPT-185 antagonizes survival stimuli is not known and warrants further study. In a mouse model of CLL, KPT-251 reduces leukemic cell counts, slows disease progression, and improves overall survival with minimal weight loss or other toxicities. It should be emphasized that KPT-251 can be given over many months to mice indicating that is well tolerated. Similar to the results in vitro, Mcl1, and XIAP mRNA were also down-modulated in mice receiving a single dose of KPT-251, although no differences were observed at protein level. Together, these findings show that XPO1 is a useful target in CLL cells with minimal effects on normal cells, and provide a basis for development of SINEs in CLL and related hematologic malignancies. We believe that the future development of low-toxicity, small-molecule XPO1 inhibitors may provide a new approach to treating cancer.
Authorship
Contribution: R.L., Q.S., Y.M.C., and J.C.B. designed the experiments, analyzed the data, wrote the paper, and reviewed and approved the final version of the paper; and K.W., L.T., S.J., Y.Z., V.G., E.M., C.B., S.G., A.F., R.M., A.J.J., D.L., X.M., D.D., V.S., S. Shechter, D.M., S. Shacham, and M.K. planned and contributed to components of the experimental work presented (chemistry, biologic, or animal studies), reviewed and modified versions of the paper, and approved the final version of the paper.
Leave a Reply