USP6 Confers Sensitivity to IFN-Mediated Apoptosis through Modulation of TRAIL Signaling in Ewing Sarcoma

This article is featured in Highlights of This Issue, p. 1817

Sarcomas are a diverse class of malignancies that represent a significant challenge in oncology. Ewing sarcoma is the second most common bone sarcoma, and typically affects individuals in the first two decades of life (1). Although patients with localized disease experience 5-year survival rates of 75%, patients with metastatic disease face a dismal survival probability of approximately 20%. Thus, there is an urgent need to identify biomarkers that can predict recurrence and response to therapy, and develop strategies to combat metastatic disease.

The key etiologic agent in Ewing sarcoma is a translocation product that fuses the EWS RNA-binding protein with an Ets family transcription factor, most commonly FLI1 (2). Sustained EWS–FLI1 activity is required for transformation, and significant efforts have been aimed at identifying its critical targets. Multiple effectors that contribute to pathogenesis have been identified, both in cultured cells in vitro and in murine models. Furthermore, therapeutics have been developed against some of these effectors, including IGF, VEGF, and EWS-FLI1 itself (3, 4). However, their clinical efficacy has been limited, underscoring the need to identify novel targets and approaches for Ewing sarcoma treatment.

Our research focuses on the ubiquitin-specific protease 6 (USP6) oncogene, which is translocated in multiple benign mesenchymal tumors, including primary aneurysmal bone cyst (ABC), and nodular fasciitis (5, 6). USP6 translocations were also identified in fibroma of tendon sheath and giant–cell rich granuloma (7, 8). In all cases, translocation resulted in promoter swapping and high-level expression of wild-type USP6. USP6 expression is normally highly restricted in adult human tissues, with significant levels observed only in testes (9). Although USP6 was first cloned in 1992 (10), until recently, little was known regarding its molecular functions, either physiologically or during tumorigenesis. We have shown that when ectopically expressed in candidate cells of origin for ABC and nodular fasciitis (i.e., fibroblasts and preosteoblasts), USP6 induces formation of tumors that recapitulate key clinical, histologic, and molecular features of the human tumors (11–13), with its catalytic activity as a deubiquitylating enzyme being essential (12). Our more recent work revealed that USP6 promotes tumorigenesis through multiple pathways, including Jak1/STAT3, Wnt/β-catenin, and NFκB (11, 14, 15). Within the Jak1/STAT3 pathway, Jak1 itself is the critical target of USP6 (15). Deubiquitylation of Jak1 by USP6 rescues it from proteasomal degradation, leading to greatly elevated levels of the kinase, and sensitizing cells to Jak1 agonists such as IL6 (15).

Although translocation-driven overexpression of USP6 plays a key role in benign neoplasms, its role in malignant entities where it is not the oncogenic driver remains unexplored. It is often incorrectly cited that USP6 is widely expressed in cancer cell lines. However, this erroneous conclusion is based on an early study in which Northern probes cross-reacted with the highly related, widely expressed USP32 gene (10). Later reverse transcription-quantitative PCR (RT-qPCR) of primary tumors with USP6-specific primers indicated that its expression is far more restricted: high USP6 expression appears to occur predominantly in tumors of mesenchymal origin (16). Yet, to date, there have been few publications exploring USP6 functions in malignant cells, with most in HeLa cells (14, 17–20).

We sought to investigate functions of USP6 in Ewing sarcoma, one of the malignancies shown to express high levels (16). We show that USP6 triggers a gene signature reflective of response to IFN, a Jak1 agonist that functions in immunity. USP6 renders Ewing sarcoma cells exquisitely sensitive to exogenous IFNs: not only is STAT1-mediated gene expression dramatically potentiated in USP6-expressing cells by IFN treatment, but Type I IFN is selectively cytotoxic to USP6-positive but not USP6-negative Ewing sarcoma cells. IFN-induced death is mediated by TRAIL, a potent proapoptotic ligand. This work represents one of the first studies to examine USP6 functions in malignant cells, and suggest that it might serve as a prognostic indicator for response of Ewing sarcoma to IFN treatment.

RD-ES and TC-71 were from Dr. Frederic Barr (National Cancer Institute, Bethesda, MD) and Dr. Lee Helman (Children's Hospital of Los Angeles, Keck School of Medicine, Los Angeles, CA), respectively. CHLA-10 and SK-N-MC were from Dr. Irfan Asangani (Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA). Lines expressing USP6 in a doxycycline-inducible manner were generated as described previously (12). Cells were tested for mycoplasma every 3 to 6 months, and prophylactically maintained in Mycoplasma Removal Agent (MP Biomedicals #09350044) for 2 weeks after thawing. All experiments used cells maintained for fewer than 20 passages after thawing. Cell line identity was confirmed by short tandem repeat analysis just prior to manuscript submission.

Validated CRISPR target sequences for Jak1, STAT1, and STAT3 were from published sequences (21). Target gRNAs [Jak1 (CACCGTCCCATACCTCATCCGGTAG), STAT1 (CACCGTCCCATTACAGGCTCAGTCG), and STAT3 (CACCGAGATTGCCCGGATTGTGGCC)] were subcloned into LentiCRISPRv2 (Addgene #52961) as described previously (21). USP6/RD-ES cells were transfected with CRISPR constructs and subjected to puromycin selection. Clones were screened by immunoblotting.

Doxycycline was obtained from ClonTech (#8634–1). Jak Inhibitor I (CAS 457081-03-7; #420099) and PS-1145 (P6624) were obtained from Sigma-Aldrich. Lipofectamine 2000 was obtained from Life Technologies. IFNα, IFNβ, and IFNγ were obtained from PBL Assay Science (#11410-2 and #11200-1) and PeproTech (#300-02), respectively. ZVAD (FMK001) and IETD (FMK007) were obtained from R&D Systems. TRAIL (catalog no. 752904) and anti-TRAIL (catalog no. 308202) were obtained from BioLegend. Caspase-3/7 (#G8090) and caspase-9 (#G8210) activation kits were purchased from Promega, and assays were performed on Molecular Devices SpectroMax. Annexin V staining kit was obtained from eBioscience (#88-8007-72), and samples were analyzed on BD Biosciences Accuri C6 and LSR II machines.

Cell lysis was performed as described previously (12). Jak1 (cs-3332), pSTAT1 (cs-9167), pSTAT3 (cs-9145), TRAIL (cs-3219), PARP (cs-9542), caspase-8 (cs-9746), and Bid (cs-2002) were obtained from Cell Signaling Technology. HA (sc-805), STAT1 (sc-346), STAT3 (sc-482), and p65 (sc-372) were obtained from Santa Cruz Biotechnology. Erk antibody was obtained from Dr. John Blenis, Weill Cornell Medical College, New York, NY. Quantification was performed using the Image Studio Lite. TRIzol was used for RNA isolation, and qPCR was performed using SYBR Green (catalog no. 436765, Thermo Fisher Scientific). Erk, STAT3, and p65 were used as protein-loading controls as described previously (22–25); their levels were comparable across conditions as shown.

RNA was isolated from USP6/RD-ES cells treated with or without doxycycline and IFNα for 24 hours. RNA sequencing (RNA-seq), alignment, processing, and repository deposit was performed by the University of Pennsylvania Next-Generation Sequencing Core (GSE107307). The CDF files for the Affymetrix U133A and U133 Plus 2.0 arrays were edited to remove probes from the USP6 probe set (206405_x_at) that cross-reacted with USP32 or other genes. This refined USP6-specific probe set comprised Probe 4, 8, 9, and 11. Publicly available Ewing sarcoma datasets [GSE7007 (26) and GSE37371 U133A (27)] were sorted by USP6 expression. Patients with the highest USP6 levels were compared to those with the lowest (5 per group). For the germ cell tumor dataset GSE10615 (28), samples were segregated as seminomas (USP6^high) versus yolk sac tumors (USP6^low). Gene Set Enrichment Analysis (GSEA) was performed as described previously, using the “Hallmarks” molecular signature database. Gene expression analysis in nodular fasciitis was compared with other USP6-nonexpressing mesenchymal tumors, as described previously (15).

DNA methylation datasets for five Ewing sarcoma cell lines (CADO-ES1, SK-NMC, A673, RD-ES, and SK-ES-1) were procured from the Cancer Cell Line Encyclopedia (CCLE; https://portals.broadinstitute.org/ccle; refs. 29, 30), and relative CpG methylation for various genes was plotted using GraphPad. Methylation probe IDs for the USP6 promoter were obtained from MExpress (31) and used in conjunction with GSE89041 (32).

Little is known about how USP6 functions in the context of malignant cells where it is not the oncogenic driver, with only a handful of reports largely restricted to HeLa (14, 17–20). As mentioned, USP6 expression in neoplasms is far more restricted than initially believed: screening of a broad panel of primary samples demonstrated that high expression was predominantly confined to tumors of mesenchymal origin, including Ewing sarcoma (16).

To explore what functions USP6 might have in Ewing sarcoma, gene expression patterns were investigated in primary patient samples. Most large Ewing sarcoma patient datasets utilize Affymetrix microarrays, which use probesets consisting of 11 distinct probes against a given gene. However, most probes in the USP6 probeset (206405_x_at) cross-reacted with USP32 or other genes. Therefore, GSEA analysis was refined to use only the USP6-specific subset of probes, comparing Ewing sarcoma tumors with the highest versus lowest levels of USP6 expression. From two independent patient datasets, IFNα (type I) and IFNγ (type II) responses emerged among the top signatures associated with high USP6 expression (Fig. 1A; see Supplementary Table S1 for expanded GSEA results). In addition, the IL6/Jak/STAT3 pathway was potently activated, which we previously showed to be induced by USP6 in our model of ABC.

To determine whether USP6 directly induces these gene signatures, we sought to perform mechanistic studies in immortalized Ewing sarcoma cells. However, none of the commonly used Ewing sarcoma lines expressed appreciable USP6 levels (data not shown). We found extensive CpG methylation across the USP6 promoter in all immortalized Ewing sarcoma cell lines examined, while comparatively low methylation was observed in primary Ewing sarcoma tumors (Supplementary Fig. S1A). CpG methylation heatmaps revealed significant silencing of USP6 in multiple Ewing sarcoma cell lines relative to genes known to be highly expressed in Ewing sarcoma, such as Myc and EZH2 (Supplementary Fig. S1B). How USP6 becomes methylated upon cell immortalization is unknown, but regardless, this necessitated expression of USP6 ectopically. We generated clonal and pooled stable lines expressing varied levels of USP6 in a doxycycline-inducible manner in the patient-derived Ewing sarcoma cell line, RD-ES (Fig. 1B). USP6 induced dose-dependent upregulation of the Jak1 kinase (which we recently reported to be stabilized by direct deubiquitylation by USP6) and phosphorylation of STAT3 (Fig. 1B), similar to what we observed in our models of ABC and nodular fasciitis (15). In addition, we observed robust phosphorylation of STAT1, the key STAT family member that mediates IFN signaling (Fig. 1B). RNA-seq was performed comparing the pooled cell line, USP6/RD-ES, in the presence versus absence of doxycycline. As in primary Ewing sarcoma samples, IFNα and IFNγ responses emerged as the top "hits," followed by IL6/Jak/STAT3 activation (Fig. 1C; Supplementary Table S1). Together, these results demonstrate that not only is USP6 associated with an IFN response in Ewing sarcoma in vivo, but that it is sufficient to activate this pathway. They also validate that our RD-ES cell model faithfully reflects physiologic USP6 functions in Ewing sarcoma patient samples.

We next explored whether USP6 was associated with an IFN response in other tumor types. An IFN signature was also induced in nodular fasciitis, which is driven by translocation-driven overexpression of USP6 (Fig. 1D; Supplementary Table S1). In addition, USP6 expression was associated with an IFN response in germ-cell tumors. Yolk sac tumors (i.e., germ-cell tumors arising from cells lining the yolk sac that are normally destined to become ovaries or testes) uniformly express low USP6 levels, whereas seminomas (i.e., germ cell tumors arising from the germinal epithelium of the testes) exhibit high levels (Fig. 1E). GSEA comparing seminomas to yolk sac tumors revealed that high USP6 expression was again correlated with IFN and Jak-STAT3 signatures (Fig. 1F). Together, these results indicate that USP6 may be more broadly associated with an IFN response in human tumors.

We speculated that in addition to triggering an IFN response by itself, USP6 might render RD-ES hypersensitive to exogenous IFN due to the elevated Jak1 levels. Indeed, dramatic enhancement and prolongation of STAT1/3 activation in USP6/RD-ES cells was observed with Type I and II IFNs (IFNα/IFNβ and IFNγ, respectively; Fig. 2; Supplementary Fig. S2). Treatment of parental RD-ES cells with IFNα or IFNγ induced phosphorylation of STAT1 and STAT3, which peaked within 30 minutes and gradually declined by 8 hours. In contrast, STAT1 and STAT3 phosphorylation was augmented and prolonged in USP6/RD-ES, with significant activation persisting at 8 hours (Fig. 2A and B; Supplementary Fig. S2). Interestingly, we also noted that type I IFNs induced downregulation of USP6 (Fig. 2A; discussed below).

In addition to prolonging STAT1/3 activation, USP6 heightened sensitivity to low-dose IFN (Fig. 2B; Supplementary Fig. S2). At doses ranging from 10 to 1,000 U/mL, USP6/RD-ES cells showed elevated STAT1/3 phosphorylation compared with parental RD-ES. The ability of USP6 to enhance and/or prolong STAT activation was confirmed in three additional patient-derived Ewing sarcoma lines, TC-71, CHLA-10, and SK-N-MC (Supplementary Figs. S3), indicating that its effects are widely observed in Ewing sarcoma, and are not a peculiarity of the RD-ES line.

Strikingly, we noticed that with prolonged treatment, type I IFN was selectively cytotoxic to USP6-expressing but not parental RD-ES cells. IFNβ exhibited the greatest cytotoxicity, followed by IFNα, then IFNγ, as monitored by PARP cleavage and Trypan blue exclusion (Fig. 3A and B). Annexin V staining confirmed that IFNβ-induced death occurred through apoptosis (Fig. 3C; Supplementary Fig. S4A). IFNβ induced apoptosis more effectively than IFNα at doses up to 2,500 U/mL (Supplementary Fig. S4B), likely due to its greater affinity for the type I IFN receptor. Furthermore, USP6 conferred sensitivity to IFNβ in a dose-dependent manner, as the extent of death correlated with the level of USP6 expression (Fig. 3D). USP6 also sensitized TC-71 cells to IFNβ-induced apoptosis (Fig. 3E and F). However, USP6 minimally enhanced death in CHLA-10 cells and SK-N-MC cells (Supplementary Fig. S4C); the mechanism underlying this differential response is discussed below. Notwithstanding, these results indicate that USP6 can dictate the magnitude of response to IFN, and can greatly sensitize ES cells to the apoptotic potential of IFN.

We next sought to dissect the mechanism of IFN-induced apoptosis. Depending on cell type, IFN can trigger extrinsic apoptosis, which occurs through ligand binding to cell surface receptors, or intrinsic apoptosis, which occurs through mitochondrial dysregulation. These pathways can be distinguished by their requirement for distinct caspase proteases (see Supplementary Fig. S5A for pathway summary). Extrinsic apoptosis requires cleavage/activation of caspase-8, followed by caspase-3/7; intrinsic apoptosis entails cleavage of the mitochondrial protein, Bid, and caspase-9 activation, which also triggers caspase-3/7 activation. However, in some circumstances extrinsic apoptosis induced by IFN can feed into the mitochondrial route, and trigger cleavage/activation of Bid and caspase-9.

IFN−induced death of USP6/RD-ES was blocked by the caspase-8–specific inhibitor, IETD (Fig. 4A), and was accompanied by caspase-8 cleavage (Fig. 4B), implicating the extrinsic pathway. However, IFNβ also induced Bid cleavage (Fig. 4B) and caspase-9 activation (Fig. 4C), indicating engagement of the mitochondrial route. Activation of Caspase-3/7 was also observed, and could be blocked by caspase-8 inhibitor (Fig. 4D). In sum, these data indicate that IFNβ-induced death of USP6/RD-ES cells occurs through a cell surface–mediated, extrinsic route that entails mitochondrial dysregulation.

To further dissect the signaling mechanisms underlying apoptosis, we examined the roles of Jak1-STAT and NFκB, both of which have been shown to participate in IFN-mediated death (22, 23). A pan-Jak family inhibitor completely blocked apoptosis of USP6/RD-ES, whereas the NFκB inhibitor was ineffective (Fig. 4E). Reporter assays confirmed that the NFκB inhibitor was functional (Supplementary Fig. S5B). To confirm the requirement of the Jak1/STAT pathway, CRISPR-mediated knockouts of Jak1, STAT1, and STAT3 were generated in USP6/RD-ES cells (Fig. 4F; Supplementary Fig. S5C). Figure 4F shows that Jak1 deletion significantly reduced death. Deletion of both STAT1 and STAT3 was required to obtain robust inhibition of death, indicating that they play distinct roles in the apoptotic response, consistent with their ability to function as homo- and heterodimers in response to IFN. These genetic and pharmacologic approaches demonstrate that Jak1, STAT1, and STAT3 are required for IFNβ-mediated apoptosis of USP6/RD-ES.

IFN can induce expression of the proapoptotic ligands, FasL and TRAIL. Our RNA-seq data indicated that TRAIL, but not Fas, was synergistically induced by IFN in USP6/RD-ES relative to parental cells. RT-qPCR confirmed that IFNs had little or no effect on TRAIL expression in RD-ES (Fig. 5A). However, TRAIL mRNA levels were dramatically increased in USP6/RD-ES treated with IFNβ. Induction was also observed, but to a much lesser degree, with IFNα and IFNγ (Fig. 5A), correlating with the extent of death induced by each (Fig. 3). In contrast, FasL expression was not significantly affected by USP6 (Supplementary Fig. S6). We also examined expression of the five TRAIL receptors, both the active (DR4/DR5) and inactive decoy (TNFRSF10C/D and OPG) forms, whose balance has been shown to a play an important role in sensitization of cancer cells to TRAIL-induced apoptosis (33). USP6 did not alter expression of receptors in a manner consistent with sensitization to death (Supplementary Fig. S6).

Figure 5C confirms that TRAIL protein was strongly induced upon IFNβ treatment in USP6/RD-ES in a doxycycline-dependent manner. Induction of TRAIL transcription and protein was also confirmed in the USP6/TC-71 Ewing sarcoma cell line (Fig. 5B and C). Neutralizing anti-TRAIL antibody inhibited IFNβ-induced apoptosis of both of USP6/RD-ES and USP6/TC-71 cells, as measured by PARP cleavage and Annexin V staining (Fig. 5D and E). Furthermore, CRISPR-mediated deletion of TRAIL completely abrogated death of USP6/RD-ES by IFNβ (Fig. 5F). These data confirm that TRAIL plays a dominant role in mediating IFNβ-induced apoptosis of USP6-positive Ewing sarcoma cells.

As described above, the various Ewing sarcoma lines exhibited differential sensitivities to IFNβ-induced apoptosis in the presence of USP6: RD-ES and TC-71 were very sensitive, while CHLA-10 and SK-N-MC were largely unresponsive (Fig. 3; Supplementary Fig. S4). To determine whether this was due to disparate induction of TRAIL in these lines, qRT-PCR was performed. However, we found that TRAIL transcription was also synergistically induced in insensitive Ewing sarcoma lines (Fig. 5B). We then explored whether the differential responsiveness might arise from varied expression of TRAIL receptor. Strikingly, we found sensitivity to IFNβ in the presence of USP6 correlated precisely with expression of the TRAIL receptor DR4: DR4 levels were highest in RD-ES and TC-71, and largely undetectable in the insensitive Ewing sarcoma lines (Fig. 5G).

As mentioned above, type I IFNs induce downregulation of USP6 protein (Fig. 2,3,4). We noted that TRAIL also triggered USP6 downregulation, in a time- and dose-dependent manner (Fig. 6A and B). TRAIL acted more rapidly, with USP6 downregulation observed within 4 hours, whereas IFNβ required 12 to 18 hours (Figs. 2A and 6B). Because caspases play a key role in TRAIL signaling, we tested whether they mediate USP6 downregulation. Both the pan-caspase inhibitor ZVAD and the caspase-8 inhibitor IETD completely blocked IFNβ- and TRAIL-induced USP6 downregulation (Figs. 4A and 6C, respectively). Together, these results reveal a negative feedback mechanism whereby USP6 induces TRAIL transcription, which then signals through DR4 to trigger caspase-dependent downregulation of USP6 (see Model Fig. 6D). Notably, this identifies type I IFNs and TRAIL as the first physiologic agonists to regulate USP6.

Although it has long been recognized that USP6 plays a key etiologic role in several benign neoplasms, its functions in the biology of malignant entities is poorly understood. Analysis of primary tumor samples by Oliveira and colleagues revealed that among human malignancies, highest USP6 expression was most commonly observed in mesenchymal cancers, including Ewing sarcoma. The current study is the first to explore functions of USP6 in Ewing sarcoma. We have found that USP6 expression is associated with an IFN signature in primary Ewing sarcoma tumors. Furthermore, USP6 is sufficient to trigger this response when inducibly expressed in cultured Ewing sarcoma cells. USP6 also confers exquisite sensitivity of Ewing sarcoma cells to exogenous IFNs. Strikingly, Type I IFNs (particularly IFNβ) induce TRAIL-mediated apoptosis of USP6-positive but not USP6-negative Ewing sarcoma cells, in a DR4-dependent manner (see Fig. 6D for results summary).

To date there are notably few studies on USP6, and thus nothing is known of how its expression is regulated, how its activity modulated, or what normal physiologic processes it participates in. We identify type I IFNs and TRAIL as the first physiologic agonists to induce posttranslational modification of USP6. We show that TRAIL triggers the caspase-dependent processing and downregulation of USP6, and that type I IFN can also trigger this downregulation through induction of TRAIL signaling. We speculate that this negative feedback loop (wherein USP6 serves to amplify IFN-mediated induction of TRAIL, which then elicits downregulation of USP6) may play an important role during normal physiology to restrict TRAIL-induced functions, which include not only apoptosis but also inflammation (34, 35).

Along this vein, a key area for future pursuit is determining the consequences of USP6-mediated IFN signaling in Ewing sarcoma pathogenesis. Numerous studies have indicated that IFNs can either promote or antagonize tumor progression across broad tumor types (36–38). This complexity can be ascribed to its ability to act not only on tumor cells, but also on immune cells and other cells in the tumor microenvironment. In some scenarios, IFNs can promote an inflammatory microenvironment that enhances proliferation and metastasis of tumor cells (38). In others, IFNs can stimulate immune infiltration and thereby promote tumor-cell killing. Thus, future studies will determine whether activation of IFN signaling by USP6 acts in a pro- or antitumorigenic manner in Ewing sarcoma in vivo. Notably, previous studies have shown that IFNs and TRAIL largely function in an antitumorigenic capacity in Ewing sarcoma, both in vitro and in murine xenografts (39–41). Results have been somewhat variable, with IFN being sufficient to block proliferation and induce death in some studies, but requiring cotreatment with other agents in others. Our work provides a potential mechanism by which Ewing sarcoma cells acquire sensitivity to the apoptotic effects of IFNs.

Standard of care for patients with Ewing sarcoma has progressed minimally over the past two decades. General cytotoxic chemotherapy is typically inefficacious in patients with disseminated or recurrent disease. Therefore, essential goals have been to develop novel therapies to prevent and treat recurrent/disseminated disease, and to identify biomarkers that can predict response to therapy. Type I IFN has previously been explored as a potential therapeutic for several cancers, but its use is currently restricted to advanced cases of melanoma (42). However, its broader use has been limited by severe systemic side effects due to its potent immunostimulatory activity (42). Our current results may help alleviate this issue, because USP6 greatly sensitizes cells to low-dose IFN. Thus, reduced IFN doses could be utilized that would retain tumoricidal activity while minimizing systemic side effects. Furthermore, because USP6 appears to be associated with an IFN response in other cancers, our findings may be applicable to other malignancies in which USP6 is overexpressed.

No potential conflicts of interest were disclosed.

Conception and design: I.C. Henrich, M.M. Chou

Development of methodology: I.C. Henrich, M.M. Chou

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I.C. Henrich, A.M. Oliveira, M.M. Chou

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I.C. Henrich, M.M. Chou

Writing, review, and/or revision of the manuscript: I.C. Henrich, M.M. Chou

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I.C. Henrich, R. Young, M.M. Chou

Study supervision: I.C. Henrich, M.M. Chou

Other (performed/carried out experiments): R. Young, L. Quick

The authors thank Gerd A. Blobel for critical reading of the manuscript. This work was funded by NIH/NCI grants CA168452 and CA178601 (to M.M. Chou) and TG 32GM008076 (to I.C. Henrich).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.