|Home | About | Journals | Submit | Contact Us | Français|
Merkel cell carcinoma (MCC) is the most aggressive skin cancer. Recently, it was demonstrated that human Merkel cell polyomavirus (MCV) is clonally integrated in ~80% of MCC tumors. However, direct evidence for whether oncogenic viral proteins are needed for the maintenance of MCC cells is still missing. To address this question, we knocked down MCV T-antigen (TA) expression in MCV-positive MCC cell lines using three different short hairpin RNA (shRNA)-expressing vectors targeting exon 1 of the TAs. The MCC cell lines used include three newly generated MCV-infected cell lines and one MCV-negative cell line from MCC tumors. Notably, all MCV-positive MCC cell lines underwent growth arrest and/or cell death upon TA knockdown, whereas the proliferation of MCV-negative cell lines remained unaffected. Despite an increase in the number of annexin V-positive, 7-amino-actinomycin D (7-AAD)-negative cells upon TA knockdown, activation of caspases or changes in the expression and phosphorylation of Bcl-2 family members were not consistently detected after TA suppression. Our study provides the first direct experimental evidence that TA expression is necessary for the maintenance of MCV-positive MCC and that MCV is the infectious cause of MCV-positive MCC.
Merkel cell carcinoma (MCC) is a highly aggressive neuroendocrine skin cancer. Although it is rare, its reported incidence is increasing (19). MCC is associated with UV exposure and affects primarily elderly and immune-suppressed patients (5, 11, 17, 26). The susceptibility of MCC to immune surveillance is similar to that of known virus-induced cancers and suggests that MCC has an infectious trigger (9). Recently, a new human polyomavirus, termed Merkel cell polyomavirus (MCV), was discovered to be clonally integrated into MCC tumor genomes (14). While MCV integration occurs at distinct sites in MCC tumors from different individuals, primary tumors and corresponding metastases have identical integration sites, consistent with the occurrence of MCV infection and integration prior to clonal expansion and metastasis (14, 37). A number of studies have confirmed that MCV is present in 69 to 85% of MCC tumors collected from Europe and the United States (4, 15, 21, 41). Surveys of control non-MCC skin, hematolymphoid, and neuroendocrine tumors are generally negative for MCV, although incidental low-level infection can be detected (4, 14, 22, 33, 34, 39, 42, 44).
All polyomaviruses encode alternatively spliced large T (LT) and small T (sT) antigen transcripts that share exon 1 of the T-antigen (TA) locus. Additional multiply spliced TA transcripts have been described for different polyomaviruses, including the 17kT and 57kT antigens in simian virus 40 (SV40) and MCV, respectively (40, 46). Research on viral proteins encoded by the TA locus has been central to uncovering cell signaling networks important in cancer biology (10, 38). The targeting of cellular proteins, such as retinoblastoma protein (Rb), p53, and protein phosphatase 2A (PP2A), by TAs contributes to polyomavirus-induced cell transformation (for reviews, see references 1 and 2). MCV TAs that are expressed in MCC tumors lack a putative p53 binding domain because of tumor-associated T-antigen deletion mutations (37, 40). Other conserved tumor suppressor-targeting motifs, including the Rb binding domain (LXCXE motif), the J domain (HPDK) in LT/57kT, and a putative PP2A interaction domain in sT, remain intact (40).
Current data point toward MCV as the infectious cause for most Merkel cell cancers: the virus is associated with MCC tumors and, when present, expresses T antigen in tumor cells but not in healthy surrounding tissues (7, 20, 39). MCV is specific to MCC and is not detected at significant levels in other cancers or in healthy skin examined to date, despite widespread circulation of MCV among human populations (8, 23, 29, 42). Clonal analysis of MCC tumors also supports the correct temporal relationship for causality; i.e., MCV infection occurs prior to MCC tumor development (18). If MCV is a direct cause of MCC tumorigenesis, it is expected that MCC tumors will require MCV protein expression to maintain the tumor phenotype—the so-called oncogene addiction.
To address this question, we generated four new MCC cell lines that were examined together with previously established MCC and non-MCC cell lines. Using two independent methods, we show that short hairpin RNA (shRNA) targeting of the MCV T antigens initiates cell cycle arrest and cell death only in MCV-positive MCC cells. Thus, MCV TA expression is necessary to maintain the oncogenic phenotype of MCV-positive MCC cell lines.
This study analyzing human cell lines was conducted according to the principles expressed in the Declaration of Helsinki. The study was approved by the Institutional Review Board of Würzburg University Hospital (Ethikkommission der Medizinischen Fakultät der Universität Würzburg; sequential study number 124/05) and the University of Pittsburgh Cancer Institute (protocol 96-099). All patients provided written informed consent for the collection of samples and subsequent analysis.
Table Table11 provides information on the origin and in vitro features of the MCC cell lines used in this study. We established four permanently growing cell lines derived from patients with histologically confirmed MCC (BroLi, WaGa, MS-1, and MaTi). BroLi, WaGa, MS-1, MKL-1 (40), and MKL-2 (42a) were positive for MCV DNA by PCR. Clonal integration of the MCV genome was confirmed by Southern blotting for MKL-1, MKL-2, and MS-1 (data not shown). UISO (35), MCC13 (27), and MaTi are MCV-negative MCC cell lines. FM88 is a melanoma cell line (3), and Jurkat (16) is a T-cell line. All cell lines were grown in RPMI 1640 supplemented with 10% fetal bovine serum (FBS) (except for MS-1 at 20% FBS), 100 U/ml penicillin, and 0.1 mg/ml streptomycin. For lentiviral infections, cell counting, or flow cytometric analysis MKL-1, MKL-2, BroLi, MS-1, and MaTi cells were dissociated by incubation in a trypsin-EDTA solution.
Cells were lysed in protein extraction buffer (either 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate [SDS], 150 mM NaCl, and 50 mM Tris-HCl [pH 8.0] or 0.6% SDS, 1 mM EDTA, 10 mM Tris-HCl [pH 8.0], 2 mM NaF, and 2 mM NaVO3) supplemented with a protease inhibitor cocktail (Roche Diagnostics). Samples were resolved by SDS-polyacrylamide gel electrophoresis and were transferred to nitrocellulose membranes. Following 1 h of blocking with Tris-buffered saline (TBS) containing 0.05% Tween 20 and 5% powdered skim milk, membranes were incubated overnight with a primary antibody, washed three times with TBS with 0.05% Tween 20 (TBS-T), and then incubated with a peroxidase-coupled secondary antibody. Bands were detected using a Western Lightning Plus-ECL chemiluminescence detection kit (Perkin-Elmer). To detect MCV TA proteins, monoclonal antibodies CM2B4, recognizing MCV LT (39), and CM8E6, recognizing both the LT and sT proteins (25), were used as reported previously. Rabbit polyclonal antibodies to cleaved poly(ADP-ribose) polymerase (PARP), cleaved caspase-3, cleaved caspase-9, Bcl-2, phospho-Bcl-2 (Ser70), phospho-Bcl-2 (Ser56), Bcl-xL, Mcl-1, Bad, phospho-Bad (Ser112), Bax, Bik, Bim, Bmf, and Puma (Cell Signaling) and a mouse monoclonal antibody to p53 DO-1 (Santa Cruz) were used at 1:1,000. As an internal control, a monoclonal antibody to α-tubulin was used at 1:2,500.
For the green fluorescent protein (GFP)-selectable knockdown system, a shRNA sequence (sense strand, 5′-GAT CCA CAA GCT CAG AAG TGA CTT CTC TAT TCA AGA GAT AGA GAA GTC ACT TCT GAG CTT GTG GAT TTT TTT G-3′) designed to target nucleotides 391 to 419 (GenBank accession number EU375803) present in all 4 T-antigen mRNAs was cloned into the lentiviral vector KH1 (43) by using SmaI-XbaI sites to generate TA.GFP. For control purposes, a scrambled KH1 shRNA construct provided by Monique Verhaegen (43) was used. For the puromycin-selectable knockdown system, shRNA sequences designed to target nucleotides 222 to 242 (sense strand, 5′-CCG GAA GAG AGG CTC TCT GCA AGC TCT CGA GAG CTT GCA GAG AGC CTC TCT TTT TTG-3′) and nucleotides 173 to 193 (sense strand, 5′-CCG GAA CTC CTT CTG CAT ATA GAC ATC AAG AGT GTC TAT ATG CAG AAG GAG TTT TTT TG-3′) were individually cloned at AgeI and EcoRI sites of the pLKO.1 lentiviral vector to generate T1.puro and T2.puro, respectively. The control plasmid pLKO.1, which has a scrambled nontargeting short hairpin RNA sequence, was obtained from Addgene. Lentiviruses for the KH1 GFP constructs were produced in HEK293T cells using pRSV rev, pHCMV-G, and pMDLg/pRRE helper constructs. Lentiviruses for the puromycin pLKO constructs were produced in HEK293FT cells (Invitrogen) using pPax2 and pMD2.G helper constructs. Two or 3 days following transfection, virus supernatants were harvested and filtered through 0.45-μm-pore-size filters. For infection, virus-containing supernatants were supplemented with 1 μg/ml Polybrene and were then added to the target cells for 4 h (WaGa cells, which are sensitive to Polybrene) or overnight (other cell lines). Following incubation with virus, cells were washed twice with medium. Since KH1 also encodes GFP, the infection rate could be determined by flow cytometry (FACScanto; BD Biosciences). Equal virus titers for TA shRNA and the scrambled viruses were ensured by serial dilutions of the viruses applied to MCC13 cells, giving equal infection efficiencies as measured on day 4 following infection. Infectious titers for puromycin-selectable shRNA vectors determined in UISO cells ranged from 5 × 10−6 to 20 × 10−6 CFU/ml. This indicates that growth failure in MCV-positive cell lines is not due to a low titer of shRNA virus. In order to improve infection rates, it was necessary to dissociate spheroidal growing cells as well as to include 1 μg/ml Polybrene; both measures are accompanied by some degree of death of MCC cells, as indicated by a major sub-G1 population on day 7 after infection with Sc.GFP (see Fig. Fig.6B6B).
GFP expression by KH1-infected cells was used to compare the behavior of infected and uninfected cells: on day 1 following infection, the cells were mixed with approximately 20% Polybrene (1 μg/ml)-treated, uninfected cells. Starting from day 4 postinfection, the frequency of GFP-positive cells in this mixture was determined in a time course.
At 48 h after lentiviral infection, 104 infected UISO or MCC13 cells or 2.5 × 104 infected cells from MCV-positive cell lines (MKL-1, MKL-2, WaGa, and MS-1) were seeded in 96-well plates with 150 μl growth medium containing puromycin (1 μg/ml). From day 3 postinfection, the growth was monitored in the presence of puromycin using the WST-1 assay reagent (Roche) according to the instruction manual. Due to the low efficiency of lentiviral infection in WaGa cells, the proliferation assay was performed after puromycin selection starting at day 6 postinfection. WST-1 is a tetrazolium salt that, when in contact with metabolically active cells, is cleaved to formazan by mitochondrial dehydrogenases. MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium] is a similar tetrazolium reagent that can be used to determine the metabolic activity in cell cultures. The MTS assay (Promega) was used for Sc.GFP- and TA.GFP-infected cells, which contained at least 80% infected GFP-positive cells on day 4 postinfection and were not subjected to any kind of selection.
The annexin V-phycoerythrin (PE) apoptosis detection kit I (BD Pharmingen, San Diego, CA) was used to identify apoptotic cells by flow cytometry. According to the manufacturer's instructions, the cells were double stained with PE-labeled annexin V and the DNA-intercalating agent 7-amino-actinomycin D (7-AAD). Analysis was performed on a FACSCanto flow cytometer (BD Biosciences, Heidelberg, Germany). In order to analyze the infected cells exclusively, they were gated on GFP. Subsequently, the early apoptotic cells were identified as the 7-AAD− annexin V+ cells.
To determine the proportion of cells in different phases of the cell cycle, analysis of cellular DNA content was performed. Dissociated single cells were pelleted and resuspended in 0.5 ml phosphate-buffered saline (PBS) supplemented with 1% fetal calf serum. Five milliliters of ice-cold ethanol (100%) was added, followed by overnight incubation at 4°C. Fixed cells were pelleted; resuspended in 1 ml PBS supplemented with 1% fetal calf serum, 0.05 mg/ml propidium iodide (PI), and 0.1 mg/ml RNase A; and incubated for 1 h at 37°C. Analysis was performed on a FACSCanto flow cytometer. For bromodeoxyuridine (BrdU) incorporation, MCV-positive cell lines were infected with the lentiviral shRNA vector Sc.puro or T1.puro. To eliminate uninfected cells, puromycin (1 μg/ml) was added to the culture medium for 4 days starting at day 2 postinfection. At day 8 postinfection, cells were labeled with 10 μM BrdU for 3 h. BrdU incorporation was assayed using an anti-BrdU monoclonal antibody (BD Biosciences), followed by PI staining as described above. Dead cells were gated out in this assay.
Statistical analysis (by the Mann-Whitney test) was performed using GraphPad Prism software (GraphPad Software, La Jolla, CA).
The MCV TA locus expresses alternatively spliced transcripts encoding the LT, sT, and 57kT proteins (39, 40), which share an identical N-terminal amino acid sequence encoded by exon 1 of LT (Fig. (Fig.1).1). We examined MCV TA expression in a panel of established MCV-positive (MKL-1 and MKL-2 [42a, 40]) and MCV-negative (UISO  and MCC-13 ) MCC cell lines, as well as in four newly established MCC cell lines (BroLi, MaTi, WaGA, and MS-1). WaGa, BroLi, and MS-1 are positive for MCV DNA, whereas MaTi is negative (Table (Table11).
Immunoblotting for an exon 2 epitope with antibody CM2B4 confirmed LT protein expression in MKL-1, MKL-2, BroLi, WaGa, and MS-1 cells (Fig. (Fig.22 A). LT bands range in size from 40 kDa (MKL-2) to 60 kDa (MS-1), consistent with the different tumor-derived truncating LT mutations that were determined for these cell lines (Table (Table1).1). However, the apparent molecular weights are higher than the predicted molecular weights, and several cell lines (MKL-1, BroLi, and WaGa) show multiple LT bands, suggesting posttranslational LT modification and aberrant splicing events. Similar LT banding patterns were replicated by immunoblotting with CM8E6 (Fig. (Fig.2B),2B), a monoclonal antibody recognizing an exon 1 epitope common to all TA isoforms, including the sT protein (18 kDa), which is not affected by truncating mutations (Fig. (Fig.1).1). MCV-negative UISO, MCC13, and MaTi cells do not express any MCV TA isoforms, although a prominent nonspecific signal is present at 75 kDa in all cell lines immunoblotted with antibody CM8E6.
To determine if MCV TA expression is required for MCV-positive cell line survival, two independent strategies to knock down all TA mRNAs by targeting exon 1 were developed; the results of these experiments were mutually confirmatory. Locations of knockdown targeting sequences are shown in Fig. Fig.1.1. The first strategy targeted MCV TA exon 1 using the shRNA lentiviral vector KH1, which also encodes green fluorescent protein (shTA.GFP) (43). Flow cytometry analysis of cells infected with shTA.GFP or its scrambled control (shSc.GFP) virus demonstrated general infection rates ranging from 75 to 95%; one exception, however, was the MCV-negative cell line MaTi, in which less than 50% of cells were infectible. In the second approach, puromycin-selectable lentiviral shRNA vectors (shT1.puro and shT2.puro), which allowed for enrichment of infected cells prior to further analysis, were used. Immunoblotting confirmed the efficient knockdown of LT and sT protein expression in MCV-positive MCC cells by both strategies compared to expression with the respective scrambled controls (Fig. (Fig.33 A and B).
The impact of shTA.GFP infection on cell growth was analyzed in mixed cultures of infected and uninfected cells (distinguished by the presence or absence of green fluorescence) for the different cell lines. In all 5 MCV-positive cell lines (MKL-1, WaGa, BroLi, MKL-2, and MS-1), a gradual loss of green fluorescent cells was observed over time with shTA.GFP infection, but not with shSc.GFP infection (Fig. (Fig.44 A). This is consistent with decreased proliferation and/or survival upon TA knockdown. MTS assays performed on MCV-positive shTA.GFP-infected cells confirmed the decrease in metabolic activity 8 to 15 days after infection compared with the activity of cells infected with shSc.GFP (data not shown). In contrast, none of the 5 MCV-negative cell lines (MCC13, MaTi, UISO, FM88, and Jurkat) infected with shTA.GFP or shSc.GFP showed a decrease in cell proliferation over time, as indicated by stable ratios of green fluorescent to nonfluorescent cells (Fig. (Fig.4A4A).
A proliferation assay applying the tetrazolium compound Wst-1 was used to assess the fate of cells infected with puromycin-selectable vectors. Consistent with the results of GFP lentivirus knockdown experiments, MCV-positive MKL-1, WaGa, MKL-2, and MS-1 cells infected with shT1.puro or shT2.puro failed to grow (Fig. (Fig.4B),4B), while cells of the MCV-negative cell lines UISO (Fig. (Fig.4B)4B) and MCC13 (data not shown) proliferated with kinetics similar to that of the respective shSc.puro-infected controls. In an attempt to differentiate between sT and LT effects, two different shRNA viruses targeting the TA intron 1 region were tested (Fig. (Fig.1).1). Both failed to suppress sT expression; therefore, we cannot distinguish LT from sT effects, although experiments to dissect the contributions of each of these isoforms to cell survival are ongoing.
To distinguish between induction of cell death and reduced proliferation, we performed annexin V staining and cell cycle analysis. On day 7 after infection with shTA.GFP, TA knockdown in GFP-positive MKL-1 and BroLi cells was associated with a small but distinct fraction of cells (6%) displaying early apoptotic features (7-AAD− annexin V+ cells) (Fig. (Fig.55 A). For WaGa cells, we did not observe a significant increase in 7-AAD− annexin V+ early apoptotic cells upon T-antigen knockdown (Fig. (Fig.5A).5A). To study the involvement of apoptosis, we examined caspase activation. shTA.GFP-induced TA knockdown did not increase the cleavage of PARP, caspase-3, or caspase-9 (Fig. (Fig.5B).5B). As a control, cleaved forms of caspases as well as PARP protein were generated after doxorubicin treatment, indicating intact caspase-initiated apoptotic pathways in MCC cells (Fig. (Fig.5B).5B). Comparable results were obtained by analysis of shT1.puro-infected MCC cells (data not shown).
All MCC cell lines are characterized by a prominent cell fraction in G0/G1 (e.g., 85% of MKL-1 cells, 83% of WaGa cells, and 88% of BroLi cells) under standard cell culture conditions. Subsequent to shTA.GFP knockdown, but not after infection with the shSc.GFP control construct, the numbers of WaGa cells in S and G2/M are reduced (Fig. (Fig.66 A and B), an effect that is present but less pronounced for MKL-1 and BroLi cells (Fig. (Fig.6A).6A). Accumulation of cells in G1 is also observed upon TA knockdown by infection of MKL-1, WaGa, MKL-2, and MS-1 cells with shT1.puro (Fig. (Fig.6C6C and data not shown). Moreover, decreased BrdU labeling of these cells is indicative of reduced S-phase entry from G0/G1. (Fig. (Fig.6C6C).
To investigate the mechanism for decreased MCV-positive cell survival after TA knockdown, we examined expression levels for a panel of pro- and antiapoptotic proteins, i.e., p53, Bad, Bax, Bik, Bim, Bmf, Puma, Bcl-2, Bcl-xL, and Mcl-1 (Fig. (Fig.77).
Modest increases in Bcl-xL and Bax levels were observed upon TA knockdown in MKL-1 cells (Fig. 7A and B); however, this observation was not consistent with those for other cell lines analyzed (Fig. (Fig.7C).7C). p53 protein was detected in all MCC cell lines except MKL-2. While p53 protein levels were marginally decreased in the MCV-positive cell lines MKL-1, MS-1, and WaGa after TA knockdown, we did not observe specific stabilization or activation of p53, suggesting that the p53 pathway is unlikely to cause the observed TA knockdown-induced growth inhibition or cell death (Fig. (Fig.7D).7D). Finally, protein expression of Bcl-2, Mcl-1, Bad, Bik, Bim, and Puma was not altered in MKL-1 cells after TA knockdown with shT1.puro or shT2.puro.
Approximately 20% of human cancers worldwide have been linked to infection (28). MCV as a new cause for human cancer is particularly difficult to establish because (i) MCC is relatively rare (5), (ii) MCV infection is near-ubiquitous among adults (10, 23, 29, 42), and (iii) not all MCC tumors are infected with the virus (4, 14, 15, 21, 41). Nonetheless, direct detection of viral proteins is strictly associated with and specific for MCC tumors (39). While MCV infection is widespread, MCV-positive MCC patients have markedly higher titers of antibody to late viral antigens than MCV-negative MCC patients, consistent with persistent antigen stimulation (29, 42). Finally, clonal MCV integration and the presence of specific TA mutations for tumor-derived MCV provide unambiguous evidence that MCV was present prior to tumor cell genesis and that the virus is not an incidental or passenger infection (14, 37).
Analysis of MCC tumors revealed a strong selective pressure within tumors to silence independent DNA replication from the integrated viral genomes in MCC cancer cells. MCV LT encodes carboxyl-terminal origin-binding and helicase domains that are required for viral DNA replication (25, 31), but most MCC-derived MCV LT DNA sequences harbor stop codon mutations truncating these domains (40). These tumor-specific mutations do not affect amino-terminal Rb1 interaction and DnaJ domains, although they may eliminate a putative p53-binding domain. The MCV sT protein, encoded by an alternative reading frame 5′ to the hypermutable LT region, remains unaffected by tumor-specific mutations.
The observational evidence for MCV causality in MCC is directly supported by the experimental studies presented. By means of shRNA knockdown, we demonstrate that MCV-positive MCC cell lines are “addicted” to expression of the viral TAs. We achieved efficient TA knockdown in five different MCV-positive MCC cell lines, using three different exon 1 target sequences and two independent selection methods in order to rule out off-target effects or artifacts due to the method used. In each case, MCV-positive cells initiated growth arrest and/or underwent cell death with the TA exon 1-specific vectors but not with scrambled shRNA vectors. MCV-negative cell lines, however, were unaffected by MCV TA-targeting shRNA, further indicating that this is not likely to be an off-target effect. These results show that MCV TA is required for MCC cell survival among those tumors infected with the virus. Since exon 1 is common to all TA isoforms, all early MCV transcripts were inhibited in our study. Dissecting the contributions of each TA isoform (e.g., LT, sT, and 57kT) to the transformed MCC phenotype will be important for future investigations.
The dependency of the transformed phenotype on the expression of TAs has been shown to be time dependent. In transgenic mice with inducible expression of SV40 TAs in the submandibular gland, hyperplasia was reversed upon the silencing of TA expression after 4 months but persisted when expression was shut down after 7 months (12). Similarly, adenovirus-transformed hamster cells have been reported to lose the previously integrated transforming viral DNA while retaining the oncogenic phenotype (30). It is conceivable that such a phenomenon may also apply to some of the MCV-negative cases of MCC; however, for MCV-positive cell lines, we show clear dependence on MCV TA expression.
Since cell death induced by TA knockdown is not associated with caspase activation, PARP cleavage, major shifts in phosphatidylserine location, or alterations in p53 or Bcl-2 family protein expression, it lacks important features of classical apoptosis. Alternatively, cell death may occur through autophagy, a process of cellular self-degradation involving the lysosomal machinery that has recently attracted increasing attention in cancer research (6). Necrosis, a process distinct from either apoptosis or autophagy, may also cause cell death (for a review, see reference 24). The observed cell cycle arrest upon T-antigen knockdown can be more directly explained by alterations of the Rb-E2F pathway: Rb family proteins are the master regulators of S-phase entry, and they are the prominent cellular targets for polyomavirus T antigens (13, 45).
Despite the fact that MCV was discovered only recently, both observational data and the experimental studies described here demonstrate that it is the likely infectious trigger for most human MCC. While polyomaviruses have been studied extensively in animal cancer models, MCV is the first polyomavirus for which strong evidence supports a causal role in human cancer (32). Identifying specific cellular pathways targeted by putative MCV TA oncoproteins can lead directly to new, more effective, and less toxic therapies for this human cancer.
We are grateful to Monique Verhaegen for the KH1 and the KH1-scrambled constructs, to Akio Soeda for providing antibodies, to John Kirkwood for access to Merkel cell tumors through the University of Pittsburgh Cancer Institute Skin Cancer SPORE, to Stefan Gaubatz for critical discussions, and to Billy Bang and Frank Lowe for help with the manuscript.
J. C. Becker was supported by the DFG (KFO124), and R. Houben was supported by the Wilhelm-Sander Stiftung (grant 2007.057.1) and the IZKF Würzburg (B157). This work was supported by NIH CA136363 and CA120726, the Al Copeland Foundation, and University of Pittsburgh EXPLORER grants to P. S. Moore and Y. Chang, who are American Cancer Society Research Professors.
Published ahead of print on 5 May 2010.