An immune escape screen reveals Cdc42 as regulator of cancer susceptibility to lymphocyte-mediated tumor suppression

Major advances in the understanding of the immune system have fostered the application of immune interventions to the benefit of patients. Among those, cancer immunotherapy is of particular interest, as it aims to specifically search and destroy malignant cells while sparing the surrounding healthy tissues.¹  In the past decade, passive immunotherapy using monoclonal antibodies has been successfully introduced into clinical oncology, and has become standard of care for patients suffering from certain breast cancers, B-cell lymphomas, colorectal cancer, and others. Adoptive cellular immunotherapy in the context of allogeneic hematopoietic stem cell transplantation (ASCT) is the standard treatment for patients with high-risk leukemias, and additional indications are clinically explored.^2,3  In contrast, the clinical role of autologous cellular immunotherapy for cancer still remains to be defined. Current studies explore the application of various formats of cancer vaccines, as well as the adoptive transfer of ex vivo–manipulated tumor-reactive lymphocytes.^4-6 

In the process of optimizing autologous and allogeneic cancer immunotherapies, much attention is paid to the identification of cancer-specific antigens and strategies to improve the activation of an immune response.⁶  Obviously, the effector phase of cancer immunity is of equal importance for therapeutic success. Malignant cells were reported to devise multiple strategies to escape immune recognition and to suppress antitumoral immune responses.⁷  Moreover, cancers may resist death signals imposed by cellular immune effectors using resistance mechanisms that interfere with endogenous signal transduction. Examples are the expression of serpins, which neutralize cytotoxic T-lymphocyte (CTL)–derived granzymes, and FLICE-inhibitory proteins, which interfere with death receptor signaling.^8-10  Recently, it was shown that factors conferring resistance to anticancer drugs and radiation can also protect tumor cells against CTL-induced apoptosis.^11,12  Accordingly, therapeutic targeting of such central mediators of resistance may increase the efficacy of both pharmacologic and immunologic cancer treatments. The definition of such targets requires a detailed understanding of the molecular mechanisms used by cancer cells to resist an immune attack. To this end, we have devised functional screening and expression cloning to reveal Cdc42 as a factor mediating cell-autonomous resistance to CTL-induced tumor suppression and cytotoxic therapies. Cdc42 is a Rho GTPase involved in the regulation of the actin cytoskeleton and growth factor signal transduction, which contributes to oncogenic transformation and cancer invasiveness.^13-15  Here, we show that Cdc42 signals to mitogen-activated protein kinase (MAPK) and Bcl-2 to protect cancer cells against immune-mediated destruction in vitro and in vivo. Pharmacologic inhibition of MAPK/extracellular signal–regulated kinase (ERK) kinase (MEK) activity successfully sensitized established cancers to tumor suppression by adoptively transplanted cellular immune effectors, thus providing a lead for therapeutic modulation of Cdc42-mediated immunoresistance.

HLA-A*0201–transgenic murine embryonic fibroblasts (MEFs) generated from cA2K^b mice backcrossed on a C57BL/6 background were retrovirally transduced to express cooperating oncogenes as published¹¹ ; cA2K^b is a transgenic line expressing a chimeric MHC class I molecule composed of the α1 and α2 domains of HLA-A*0201 and the α3, transmembrane, and cytoplasmic domains of H-2K^b.¹⁶  MEFs were maintained in Dulbecco modified Eagle medium (DMEM) supplemented with fetal bovine serum, l-glutamine, and penicillin and streptomycin (Invitrogen, Carlsbad, CA). Allo-A2K^b–reactive CD8⁺ CTL-lysing targets expressing the HLA-A*0201 antigen, HLA-A*0201/influenza matrix(58-66)–specific CTLlysing targets, and HLA-A*0201/p53(264-272) CTL lysing targets presenting the influenza matrix(58-66) epitope or the human p53(264-272) epitope in the context of HLA-A*0201 were generated by immunizing HLA-A*0201 transgenic mice with the respective peptide epitopes, as published previously.^17,18  The HLA-A*0201⁺ human colorectal cancer cell line HCT116 has been generously provided by B. Vogelstein (Johns Hopkins Kimmel Cancer Center, Baltimore, MD).

⁵¹Cr release cytotoxicity assays (5 hours) were carried out as described.¹⁹  Apoptosis was detected by cell-cycle analysis following staining with propidium iodide (PI). Cells with maintained mitochondrial transmembrane potential Δψ_m were quantitated flow cytometrically following staining with the dye TMRE (Invitrogen). To assay proliferative survival, 5000 adherent MEFs were incubated in 96-well plates with CTL effectors at the indicated effector-target (E/T) ratios for 4.5 hours. Following removal of CTLs, MEFs were harvested and replated in 35-mm dishes for a 7-day culture period. The resulting colonies were fixed, stained, and counted.¹¹ 

A cDNA encoding a GTPase-defective Cdc42 was generated by polymerase chain reaction (PCR) using pRK5-MYC.Cdc42 (Q61L; provided by P. Aspenström Uppsala University, Uppsala, Sweden) as template and cloned into the retroviral vector plasmids pMxIG (provided by T. Kitamura, University of Tokyo, Tokyo, Japan) and pQCxIN (Clontech, Mountain View, CA). All inserts were verified by sequencing. The following primary antibodies were used: actin (C4; MP Biomedicals, Irvine, CA); Bcl-2 (C2), Bcl-xL (H5), and Mcl-1 (all from Santa Cruz Biotechnology, Santa Cruz, CA); and Cdc42 (MAB3707; Chemicon, Temecula, CA). All phosphoepitope-specific antisera were purchased from Cell Signaling Technology (Danvers, MA). ABT-737 was generously provided by Abbott Laboratories (Abbott Park, IL); and PD98059, staurosporine, and etoposide were purchased from Calbiochem (San Diego, CA) and Sigma-Aldrich (St Louis, MO), respectively.

A cDNA library was constructed from the leukapheresis product of a patient with chronic myeloid leukemia using the pCMV-Script XR Library Construction Kit (Stratagene, Cambridge, United Kingdom). Total RNA was isolated and poly(A) + mRNA was prepared by means of the Oligotex mRNA Kit (Qiagen, Hilden, Germany). The mRNA was reverse-transcribed using an oligo-dT primer containing an EcoRI site at its 5′ end and ligated to adaptors. Ligation products were then digested with XhoI and ligated into the EcoRI and XhoI sites of pMxIG, and recombinant plasmids were electroporated into Escherichia coli XL10-Gold Ultracompetent Cells (Stratagene). The resulting library of 3.8 × 10⁵ cells represented 95% inserts with sizes between 0.5 and 4.0 kb. Replication-defective retroviral vectors were generated as published¹¹  to transduce oncogene-transformed cA2K^b MEFs under conditions ensuring a multiplicity of infection of 1. All studies using human material were approved by the local ethics committee (Landesärztekammer Rheinland-Pfalz, Mainz, Germany); the patient provided written informed consent in accordance with the Declaration of Helsinki.

Irradiated (1.5 Gy [150 rad]) nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice received subcutaneous injections of 10⁶ cA2K^b MEFs (Cdc42 MEFs in left flank, control MEFs in right flank). Following the outgrowth of palpable fibrosarcomas (days 7-8), mice were treated by single tail-vein injections of 5 × 10⁷ splenocytes from C57BL/6 or cA2K^b mice resuspended in saline, and a subcutaneous injection of 6 × 10⁵ U interleukin-2 (Medicopharm, Nussdorf, Germany) dissolved in saline and incomplete Freund adjuvant. Tumor sizes were measured bidimensionally using a calliper. Differences between the resulting growth curves were calculated using analysis of variance (ANOVA) and Student t test. To obtain splenocyte preparations devoid of CD8⁺ T cells or CD49⁺ natural killer (NK) cells, splenocytes were depleted by means CD8 or CD49 microbeads, respectively, and a magnetic-activated cell sorter (MACS) system (Miltenyi Biotec, Bergisch-Gladbach, Germany) following the manufacturer's instructions. Daily intraperitoneal injections of PD98059 (5 mg/kg; Calbiochem) dissolved in DMSO (Sigma-Aldrich) and diluted in saline were administered for MEK inhibition in vivo. All animal studies were conducted in compliance with institutional guidelines and were approved by the responsible regulatory authority (Landesuntersuchungsamt Rheinland-Pfalz).

To functionally identify resistance factors against CTL-mediated tumor suppression, we devised an expression cloning strategy (Figure 1A). In brief, oncogene-transformed HLA-A*0201–transgenic cA2K^b MEFs were retrovirally transduced to express a cDNA library and green fluorescent protein (GFP). Clones surviving coculture with allo-A2K^b–reactive CD8⁺ CTLs (allo-A2), which lyse targets expressing the HLA-A*0201 antigen,^17,18  were expanded and analyzed for GFP expression. Vector-encoded cDNA sequences of CTL-resistant and GFP⁺ clones were retrieved by PCR and sequencing using vector-specific primers. Here, we report clone A4, which expressed a library sequence spanning the entire open reading frame of the Homo sapiens cell division cycle 42 (GTP-binding protein, 25 kDa; Cdc42), transcript variant 1, mRNA (GenBank accession no. NM_001791²⁰ ; Figure 1B). Clone A4 exhibited marked resistance to allo-A2 CTL-induced cytotoxicity as compared with parental cA2K^b MEF (Figure 1C).

To study whether Cdc42 in fact mediated protection against CTL-induced cytotoxicity, we constructed a bicistronic retroviral vector for stable expression of a constitutively active Cdc42 (Q61L) mutant²¹  and GFP. Following transduction of cA2K^b MEFs, the GFP⁺ population (cA2K^b-Cdc42 MEF) was isolated by fluorescence-activated cell sorting. This approach was taken to avoid antibiotic selection of drug-resistant clones. Such cA2K^b-Cdc42 MEFs were highly resistant to cytotoxicity induced by allo-A2 CTLs (Figure 2A). In addition, cA2K^b-Cdc42 MEFs cocultured with allo-A2 CTLs exhibited enhanced clonogenic survival as compared with cA2K^b-control MEF (Figure 2B). This protection was not explained by Cdc42-mediated down-modulation of MHC class I, as all MEF populations exhibited equal HLA-A*0201 expression (Figure S1A, available on the Blood website; see the Supplemental Materials link at the top of the online article). Similar results were obtained using the HLA-A*0201–positive human colorectal cancer cell line HCT116. HCT116 cells expressing Cdc42 (Q61L) were more resistant to immunologic tumor suppression by allo-A2 CTLs in vitro (Figure 2C).

To extend these observations, we made use of an in vivo model recapitulating the graft-versus-tumor (GVT) effect of ASCT and adoptive lymphocyte transfer. In this model, the transplantation of unprimed splenocytes derived from C57BL/6 mice suppressed the growth of HLA-A*0201–transgenic C57BL/6 fibrosarcomas, which were established in NOD/SCID mice. The tumor suppressive activity of transplanted splenocytes mainly relied on HLA-A*0201–reactive CD8⁺ lymphocytes and CD49⁺ NK cells (Figure 3A,C) acting via perforin-dependent pathways.²²  Following subcutaneous injection in NOD/SCID mice, cA2K^b-Cdc42 and cA2K^b-control MEFs readily formed HLA-A*0201–transgenic fibrosarcomas. The growth rate of these tumors was not altered by the expression of Cdc42 (Figure 3D). As expected, transplantation of allogeneic C57BL/6 splenocytes significantly suppressed the growth of cA2K^b-control fibrosarcomas, when compared with syngeneic cA2K^b splenocytes, which were tolerant to HLA-A*0201–expressing tumor cells (Figure 3A). In contrast, cA2K^b-Cdc42 fibrosarcomas largely resisted tumor suppression by transplanted allogeneic C57BL/6 splenocytes in vivo (Figure 3B). In summary, constitutively active Cdc42 protected cancer cells against cytotoxicity and tumor suppression by alloreactive CTLs in vitro and CTLs and NK cells in vivo.

Next, we asked whether protection of cancer cells by Cdc42 was restricted to cytotoxicity induced by alloreactive CTLs. To this end, cA2K^b-Cdc42 and cA2K^b-control MEFs were loaded with peptides representing the HLA-A*0201–restricted influenza matrix(58-66) or human p53(264-272) epitopes, followed by coculture with HLA-A*0201/influenza matrix(58-66)–specific CTLs (A2 Flu) and HLA-A*0201/p53(264-272) CTLs (A2 p53), respectively.^17,18  Again, cA2K^b-Cdc42 MEFs exhibited significant protection against cytotoxicity induced by cellular immune effectors of different specificities (Figure 4A). This argued for an activity of Cdc42 interfering with essential CTL effector mechanisms.

CTLs eliminate their specific targets by inducing apoptotic cell death. It was shown that CTLs can activate apoptotic caspases through several mechanisms, including the “mitochondrial” pathway, which is also triggered by anticancer drugs or radiation.^11,23  Against this background, we assessed whether cA2K^b-Cdc42 MEFs were resistant to drug-induced apoptosis. Indeed, expression of Cdc42 protected cA2K^b MEFs against dissipation of the mitochondrial transmembrane potential Δψ_m (Figure 4B top panels) and apoptotic DNA fragmentation (Figure 4B bottom panels) triggered by the protein kinase inhibitor staurosporine and the topoisomerase II inhibitor etoposide. These findings suggested that Cdc42 mediated immunoresistance by inhibition of cancer cell–intrinsic apoptosis signaling, and not by interfering with the interaction between CTLs and cancer cells. As Cdc42-expressing cells maintained their Δψ_m, apoptotic signal transduction was possibly intercepted at or upstream of the permeabilization of the mitochondrial outer membrane (MOM).

The integrity of the MOM is regulated by the Bcl-2 family of proteins. Antiapoptotic family members, such as Bcl-xL, Mcl-1 or Bcl-2 itself, counteract the proapoptotic “multidomain” proteins Bax and Bak, which permeabilize the MOM through a mechanism, which is incompletely understood.^24,25  Investigating the expression of antiapoptotic Bcl-2, Bcl-xL, and Mcl-1 in extracts derived from cA2K^b fibrosarcomas, we observed an increase of endogenous Bcl-2 expression in cA2K^b-Cdc42 tumors (Figure 5A). Using reverse transcription and quantitative PCR, no elevation of Bcl-2 RNA expression was detectable in cA2K^b-Cdc42 MEFs as compared with control MEFs (data not shown). To the contrary, expression levels and stability of endogenous Bcl-2 were markedly increased in Cdc42-expressing cA2K^b MEFs cultured in the presence of the translation inhibitor cycloheximide (Figure 5B). Thus, Cdc42 most likely inhibited apoptosis by posttranscriptional modulation of Bcl-2.

To study whether Cdc42-mediated protection of cA2K^b MEFs in fact required the activity of antiapoptotic Bcl-2 proteins, we made use of the pharmacologic apoptosis sensitizer ABT-737. This compound was rationally developed as a mimetic of “BH3-only” proteins, such as Bim, Bid, or Bad, which act as inhibitors of antiapoptotic Bcl-2 proteins.^26,27  Pretreatment of cA2K^b-control MEFs with ABT-737 failed to further enhance etoposide-induced apoptosis. In contrast, ABT-737 completely reversed the apoptosis resistance of cA2K^b-Cdc42 MEFs (Figure 5C). Recently, ABT-737 was shown to act exclusively on Bcl-2 and Bcl-xL, but not Mcl-1.^28-30  As endogenous Bcl-xL levels of cA2K^b-Cdc42 MEFs remained unchanged (Figure 5A,B), these results confirmed that Cdc42-mediated inhibition of apoptosis indeed depended on Bcl-2 activity.

Mammalian Cdc42 is a GTP-binding protein, which was initially described as a regulator of actin dynamics and cytoskeletal architecture in response to integrins and extracellular matrix signals.¹⁵  Recently, it was found to be involved in additional signal transduction pathways triggered by cytokine receptors and growth factor receptor tyrosine kinases.¹³  Cdc42 affects MAPK, which targets several regulators of apoptosis, including Bcl-2. Analyzing p42/44 mitogen-activated protein kinase (ERK) in extracts derived from cA2K^b-Cdc42 and cA2K^b-control fibrosarcomas, we observed activating phosphorylation of ERK in those tumors expressing Cdc42 (Figure 6A). Also, Cdc42-expressing HCT116 cells exhibited enhanced ERK phosphorylation in response to epidermal growth factor (EGF) stimulation (Figure S1D). As ERK-dependent phosphorylation of Bcl-2 was described to enhance its antiapoptotic activity,^29,31  we assessed whether treatment with PD98059, a pharmacologic MEK inhibitor acting upstream of ERK in MAPK signal transduction, could overcome Cdc42-dependent immunoresistance. In control experiments, PD98059 did not influence the suppression of clonogenic survival of cA2K^b-control MEFs by allo-A2 CTLs. In contrast, cA2K^b-Cdc42 MEFs pretreated with PD98059 proved more susceptible to tumor suppression and cytotoxicity by alloreactive CTLs in vitro (Figures 6B,S1B).

Based on these results, we reasoned whether pharmacologic MEK inhibition was a strategy to overcome Cdc42-mediated immunoresistance of cancer in vivo. To this end, cA2K^b-Cdc42 and cA2K^b-control fibrosarcomas were established in NOD/SCID mice. Daily intraperitoneal injections of PD98059 had no impact on the growth rate of these tumors, either in the presence or absence of cellular immune effectors (Figures 6C,S1C). Interestingly, treatment with PD98059 significantly delayed the growth rate of cA2K^b-Cdc42 fibrosarcomas in mice undergoing adoptive immunotherapy by transplantation of allogeneic C57BL/6 splenocytes (Figure 6C). These findings suggested that PD98059 sensitized cA2K^b-Cdc42 fibrosarcomas toward immune-mediated tumor suppression by modulating Cdc42 signaling to ERK. In support, cA2K^b-Cdc42 fibrosarcoma extracts prepared from NOD/SCID mice treated with PD98059 exhibited reduced ERK phosphorylation and Bcl-2 expression as compared with their vehicle-treated counterparts (Figure 6D). Thus, pharmacologic inhibition of Cdc42 signaling to the MEK-ERK pathway reduced Bcl-2 expression and sensitized established cancers to adoptive cellular immunotherapy.

The combination of target selectivity with long-term surveillance could make cellular immunotherapy an ideal cancer treatment. Accordingly, numerous approaches are explored to break tolerance to established cancers. Recent advances in the understanding of immune regulation have greatly added to the instrument, which can be applied to elicit an antigen-specific immune response in vivo. Further, the adoptive transfer of ex vivo–manipulated CTLs with improved target specificity is studied as a means to circumvent immunologic tolerance to cancer. A common prerequisite for these strategies is the assumption that once a cancer cell has successfully been targeted by lymphocytes, it will readily succumb to their various effector mechanisms. However, cancer cells can very well resist immune-mediated demise using cell-intrinsic resistance mechanisms, which parallel those conferring resistance to anticancer drugs and radiation.^8,10-12  Hence, effective cancer immunotherapy may require sensitization of intrinsically resistant cancer cells to immune-mediated cytotoxicity. With the advancement of molecularly targeted therapies, the implementation of such a strategy has become feasible. Consequently, suitable targets have to be defined, which can be addressed to improve the efficacy of immunologic cancer treatments.

Against this background, we have devised functional screening and expression cloning to reveal Cdc42, a ubiquitously expressed Rho family GTPase, as mediator of resistance to CTL-mediated growth suppression of MEFs and HCT116 colorectal cancer cells. Cdc42 acts as a molecular switch that, upon activation by various cell surface receptors, cycles from the GDP-bound inactive state to the GTP-bound active state. Downstream effector pathways of Cdc42 regulate the actin cytoskeleton, migration, and cell-cycle progression.^13-15  Numerous studies have implied Cdc42 in oncogenic transformation and tumor progression. Cdc42 function was required for full transformation of fibroblasts by oncogenic Ras,³²  Cdc42 was expressed in multiple cancers and induced invasiveness of malignant cells.^33,34  This was in keeping with the frequently observed overexpression of Cdc42 at the tumor invasive front and in particularly aggressive cancers.^35-37 

Mice with targeted deletion of Cdc42 die at early embryonic stage, thus precluding the study of mammalian Cdc42 at an organismal level.³⁸  However, the recent generation of conditional Cdc42-deficient and Cdc42GAP-deficient mice enabled the study of loss or gain of Cdc42 function in several cellular contexts. MEFs generated from these 2 mouse models featured defects in adhesion, migration, and establishment of cell polarity, thus confirming a critical role for the “fine-tuning” of Cdc42 signaling.^39,40  Interestingly, surviving Cdc42GAP-deficient mice exhibited a phenotype reminiscent of premature aging. Enhanced Cdc42 activity in those mice either resulted in genomic instability, increased apoptosis, or cellular senescence, which was partially rescued by heterozygous loss of p53.⁴¹  Taken together, these findings are most likely explained by a replicative stress response to hyperactive Cdc42, which is similar to the one evoked by oncogenes.⁴²  In turn, Cdc42-deficient MEFs exhibited defective MAPK signaling and impaired cell-cycle progression from G₁ to S phase, demonstrating an important role of Cdc42 in growth factor receptor signal transduction.³⁹ 

In our experimental system, cA2K^b MEFs were transformed by the cooperating oncogenes Myc and H-Ras, and thus had already bypassed oncogene-induced senescence and apoptosis. Accordingly, these cells tolerated the expression of the Cdc42 (Q61L) mutant, which was reported to interfere with in vitro growth of primary fibroblasts or 3T3 cells.¹³  Despite resulting in enhanced ERK activation, constitutively active Cdc42 failed to accelerate proliferation of Myc/H-ras–transformed cA2K^b MEFs in vitro and in vivo. Further, pharmacologic MEK inhibition had no antiproliferative effect on these cells. Hence, Myc/H-ras–transformed cA2K^b MEFs apparently proliferated independently of Raf-MEK-ERK signaling and Cdc42. Nevertheless, we failed to establish cA2K^b MEFs with stable shRNA knock-down of endogenous Cdc42, suggesting that it was still required for additional cellular functions.

Cdc42 plays a key role in the regulation of actin dynamics.¹⁵  Recently, overexpression of ephrin-A1 and scinderin leading to reorganization of the actin cytoskeleton was reported to mediate resistance to CTL-induced tumor cell killing in vitro.⁴³  However, cA2K^b-Cdc42 MEFs exhibited no apparent changes in cell morphology as compared with control cells, and were also protected against pharmacologic induction of cell death. Interestingly, we observed highly increased ERK activation in cA2K^b-Cdc42 MEFs, which correlated with resistance to cytotoxicity induced by CTL or proapoptotic drugs. MAPK signaling as well as constitutively active Cdc42 mutants have been shown to protect cells from death receptor-induced apoptosis.^44,45  However, the CTL populations applied in our study predominantly act via granule-dependent effector mechanisms.^11,22  Here, we have observed Cdc42-induced posttranscriptional stabilization of antiapoptotic Bcl-2, a known target of MAPK signaling.^29,31  Pharmacologic inhibition of MEK reduced Bcl-2 levels and sensitized cA2K^b-Cdc42 MEFs to lymphocyte-induced tumor suppression in vitro and in vivo. Hence, Cdc42 conferred immunoresistance by MAPK-dependent inhibition of apoptosis, which may complement immune evasion through MAPK-induced secretion of immunosuppressive factors as observed in melanoma cells harboring the B-Raf^V600E mutation.⁴⁶ 

We thank L. A. Sherman, P. Aspenström, S. W. Lowe, T. Kitamura, B. Vogelstein, and Abbott Laboratories for providing mice and reagents; S. Hoffarth, A. Hohberger, E. Antunes, R. Engel, A. Konur, and the staff of the Zentrale Versuchstiereinrichtung (central animal facility) for help with experiments; and all members of the Schuler and Theobald laboratories for support.

This work was supported by grants from the Deutsche Krebshilfe (no. 107993) and Deutsche Forschungsgemeinschaft (SCHU 1541/3-1).

National Institutes of Health

Contribution: C.A.M., P.S.H., S.T., and C.W. carried out research; C.W. and M.T. contributed unique reagents; M.T. and C.H. reviewed the data and the manuscript; and M.S. supervised the project, designed research, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Martin Schuler, Department of Medicine (Cancer Research), West German Cancer Center, University Hospital Essen, Hufelandstrasse 55, 45147 Essen, Germany; e-mail: martin.schuler@uk-essen.de.