|Home | About | Journals | Submit | Contact Us | Français|
The novel tumor suppressor Pdcd4 affects tumorigenesis by inhibiting translation. Pdcd4 is phosphorylated and subsequently lost by proteasomal degradation in response to tumor promoting conditions. Here, we describe the development of a reporter cell system to monitor the stability of Pdcd4. The phosphorylation-dependent degradation domain (“target”) or an adjacent (“off-target”) region of Pdcd4 were cloned into a luciferase expression system. The target constructs were responsive to Pdcd4 degrading conditions (e.g. TPA, p70S6K1-overactivation), while the off-target constructs remained stable. The system was optimized for and shown to be reliable in a high-throughput compatible 384-well format. Screening of 15,275 pure compounds resulted in a hit rate of 0.30% (> 50% inhibition of TPA-induced loss of signal, confirmed by re-assay). Among the hits were inhibitors of previously identified critical signaling events for TPA-induced Pdcd4 degradation. One compound was identified to be nonspecific using the off-target control cell line. Screening of 135,678 natural product extracts yielded 42 confirmed, specific hits. Z’ averaged 0.58 across 446 plates. Further characterization of active natural products and synthetic compounds is expected to identify novel Pdcd4 stabilizers that may be useful in targeting translation to prevent or treat cancers.
Programmed cell death 4 (Pdcd4) is a novel tumor suppressor that inhibits tumorigenesis by interfering with translation initiation. Pdcd4 exerts its translation inhibitory function largely by inhibiting the helicase activity of the eIF4A 1,2. Consequently, Pdcd4 is predicted to specifically prevent the translation of mRNAs with highly structured 5’UTRs that require unwinding for efficient translation. As a result of its translation inhibitory effects, Pdcd4 inhibits transformation, migration and invasion in vitro 3–5. In mouse models, transgenic overexpression of Pdcd4 was shown to reduce papilloma multiplicity and incidence in the two-stage skin carcinogenesis model, while knock-down of Pdcd4 led to higher tumor yields in the same model 6,7. Furthermore, mice nullizygous for Pdcd4 were reported to spontaneously develop lymphoma 8. There is also increasing evidence for the tumor suppressive function of Pdcd4 in humans, as recent reports indicate that Pdcd4 is lost in various tumor entities 9–12. The loss of Pdcd4 has been shown to be a prognostic marker for colon cancer progression 13. Mutational inactivation has been excluded as an underlying mechanism for the loss of Pdcd4 14. Instead, Pdcd4 levels are post-transcriptionally regulated by miR-21 9,10 and post-translationally regulated by proteasomal degradation 7,15. The increased proteasomal degradation leading to a reduced protein half-life in response to tumor promoting conditions is regulated by p70S6K1- and/or Akt-dependent phosphorylation of serine 67. Upon phosphorylation, the E3-ubiquitin ligase β-TrCP1 binds to and facilitates the ubiquitylation of Pdcd4, thus, targeting Pdcd4 for proteasomal degradation 7,15. Pdcd4 is also phosphorylated at serine 457 by PKB/Akt which appears to influence its sub-cellular localization rather than its stability 16. Since Pdcd4 appears to be influenced by multiple signaling pathways, it may serve as an important integration point to control translational changes during tumorigenesis. Thus, stabilizing Pdcd4 might prove to be an interesting, novel target for interfering with dysregulated translation during tumor development.
The therapeutic potential for interfering with translation in cancer treatment is supported by the fact that translation inhibitors have been recognized as promising therapeutic entities in tumor therapy in recent years. Rapamycin and its analogs are currently in clinical trials or already approved for use as potent anti-cancer drugs. These compounds inhibit mTOR (mammalian target of rapamycin), a protein serine/threonine kinase central to the phosphoinositide 3-kinase (PI3K) pathway in controlling gene expression and cell proliferation and it has been implicated in tumorigenesis. Inhibition of mTOR prevents activation of both p70S6K1 and eIF4E, resulting in inhibition of translation initiation 17,18. A recently described small molecule inhibitor of eIF4E was also shown to have proapoptotic activity in multiple cancer cell lines 19. Since mTOR and eIF4E inhibitors target rather general processes, both are predicted to lack specificity. In the case of mTOR inhibitors, immunosuppressive side-effects have been a major drawback during their development as anti-cancer drugs. In fact, rapamycin was initially characterized and is still widely used as an immunosuppressant during organ transplants 18,20. Similarly, eIF4E inhibitors can be anticipated to generally inhibit cap-dependent translation. Limiting the translation inhibitory effects to a well-defined, tumor-relevant set of target mRNAs would be a more desirable approach. Since Pdcd4 inhibits eIF4A activity, it is predicted to specifically affect translation of so-called ‘weak’ mRNAs which are characteristic regulators of many proto-oncogenes (e.g. c-myc) 21. Thus, stabilizers of Pdcd4 might prove to be specific in inhibiting tumor-associated changes in translation.
In this report, we describe the development, optimization and validation of a novel, cell-based high-throughput screening (HTS) assay for the identification of stabilizers of tumor suppressor Pdcd4. Since this assay measures an increase in signal compared to TPA control, toxic compounds will not be identified as false positives. An off-target control was developed and used to eliminate nonspecific effects. We further show that the assay robustly and reproducibly identifies predicted Pdcd4 stabilizers from a variety of sources.
DMSO, TPA (12-O-Tetradecanoylphorbol-13-acetate), rapamycin, LY294002 and anti-β-actin antibody were obtained from Sigma (St.Louis, MO). Anti-luciferase antibody, anti-rabbit and anti-mouse HRP-coupled antibodies came from Millipore (Billerica, MA). Peptide-purified anti-Pdcd4 antibody was described previously 14. Nitrocellulose membranes and enhanced chemiluminescence (ECL) reagents were from Amersham Biosciences. White 384 well plates and Steadylite Plus luciferase reagents were from Perkin-Elmer (Boston, MA). pGL3-control vector was obtained from Promega (Madison, WI) and pFB-neo from Stratagene (La Jolla, CA). p70S6K1 expression plasmids were kindly provided by J. Blenis 22. All media, supplements and antibiotics for cell culture were purchased from Invitrogen (Carlsbad, CA) unless otherwise noted.
Libraries of pure natural products and synthetic compounds were provided by the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis National Cancer Institute (Frederick, MD). Natural product extracts were obtained from the Natural Products Support Group, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis, National Cancer Institute (Frederick, MD). A library of known protein kinase inhibitors was obtained from BioMol (Plymouth Meeting, PA). A library of pharmaceutical active compounds (LOPAC1280) was purchased from Sigma (St.Louis, MO). Other synthetic compound libraries were obtained from a variety of commercial and academic suppliers. Pure compound libraries were stored frozen at 1–5 mM in DMSO. Natural product extracts were also stored frozen in DMSO, at 1–10 mg/ml.
Four fragments of Pdcd4 (encoding amino acids 64–80, 39–91, 16–142, 108–206) were amplified by PCR using a random human cDNA sample as template. HindIII and NarI restriction sites were added to the Pdcd4 specific amplicons. The resulting extended Pdcd4 fragments were fused to the luciferase expression cassette of the pGL3-control vector. Subsequently, the Pdcd4-luciferase fusion constructs were amplified by PCR, introducing EcoRI and BamHI restriction sides before and after the fusion product, respectively. The resulting Pdcd4-luciferase constructs were then inserted into a modified pFB-neo plasmid in which the neomycin-resistance had been replaced by a blasticidin resistance cassette. All DNA constructs were confirmed by sequence analysis.
For transient expression, various Pdcd4-luc plasmids, TK-Renilla luciferase plasmid and expression plasmids of either mutated (inactive) or wild type p70S6K1 were co-transfected into HEK293 cells using Fugene 6 (Roche) according to the manufacturer’s protocol.
For the generation of stable cell lines, Pdcd4-luc plasmids and pVSV-G packaging plasmid were co-transfected into GP2–293 cells using the same protocol. Packaged virus was collected using established methods (Clontech, Mountain View, CA). Subsequently, retroviral infection of HEK293 cells was performed to create cells stably expressing either Pdcd4(16/142)luc, Pdcd4(39/91)luc or Pdcd4(108/206)luc. Clones stably expressing the Pdcd4-luc constructs were selected by supplementing regular growth medium with 6 μg/mL blasticidin. Expression of the Pdcd4-luc constructs was verified using western blot analysis and luciferase activity measurement after selection for two weeks. Monoclonal cell lines were established by single cell dilutions in 96-well plates.
HEK293 cells were maintained in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS), 100 U/mL penicillin, 100 μg/mL streptomycin and 2 mM L-glutamine. Stable HEK293-Pdcd4-luc cell lines were maintained in regular growth medium supplemented with 3 μg/mL blasticidin. All cells were maintained at 37 °C under 5% CO2 atmosphere in a humidified incubator.
Cells were trypsinized, concentrated by centrifugation, resuspended in medium, and added to 384-well plates (40 μL/well) using a μFill microplate liquid dispenser (BioTek, Winooski, VT). Seeding and treatment conditions were optimized as indicated below. Cells were allowed to attach for 18 h. For screening, 5 μL of 100 nM TPA in medium (containing 1% (v:v) DMSO) was added to each well immediately followed by test samples. “DMSO” control wells received 5 μL of 1% DMSO in medium. Test samples were diluted in medium to 10 × final concentration in 384-well polypropylene dilution plates (multiple sources) and 5 μL transferred to assay plates containing cells and TPA. Control wells received 5 μL of PBS containing the same DMSO concentration as the diluted samples. Final concentrations for screening were 5–10 μM for pure compounds and 5 μg/mL for natural product extracts. Final DMSO concentrations were 0.1–0.6% (v:v). All dilutions and transfers were performed on a Biomek FX with a 384-channel pipet head (Beckman Coulter, Fullerton, CA).
Luminescence measurements were carried out using the Steadylite Plus luciferase reagent from Perkin Elmer according the manual. Briefly, 50 μL lysis/detection reagent was added per well (384-well plate) and luminescence was measured 10–15 min after addition using the PHERAstar luminescence reader (BMG LABTECH, Inc., Durham, NC).
For Western blot analysis, cells were sonicated and then lysed on ice for 30 min in lysis buffer [50 mmol/L Tris-HCl, 1% NP40, 150 mmol/L NaCl, 1 mmol/L EDTA, 10% glycerol, 1 mmol/L PMSF, protease inhibitor mix, phosphatase inhibitor cocktail 2 (Sigma)].
Lysates containing 50 μg protein were separated on SDS gels, and analyzed using Western blot analysis. Proteins were visualized using specific primary and appropriate secondary antibodies followed by ECL detection.
The activities of compounds were calculated using the following formula:
Compounds were considered to be hits if the activity, i.e. rescue of luciferase signal from TPA-induced reduction relative to ΔDMSO-TPA (= RLUDMSO − RLUTPA), was higher than 50%. For confirmation of activity, primary hits were re-assayed in quadruplicates. Hits were considered to be confirmed if the activity remained > 50% at a confidence interval of 95%.
For estimation of the reliability of the assay, Z’ values were calculated for each plate comparing DMSO and TPA (10 nM) controls (8 wells each) 23. A Z’ value > 0.4 was chosen as cut-off in the high-throughput screen.
For exclusion of nonspecific effects primary hits were tested in the off-target cell line in quadruplicate as well. Since TPA treatment generally did not yield changes relative to DMSO treatment, effects were calculated relative to TPA-only treatment according to the following formula:
Compounds yielding significant increases (> 120%) in the off-target cells were considered to be nonspecific. Coefficients of variation were used to determine repeatability and reproducibility of the assay. Primary hits that were not confirmed in re-testing were considered false positives.
Apparent IC50 values were calculated using SigmaPlot (SPSS, Inc., Chicago) 4-parameter logistic nonlinear regression analysis. Each assay plate was assessed separately. Unless otherwise noted, all data are presented as mean ± standard deviation.
For the identification of Pdcd4 stabilizing compounds, fragments of Pdcd4 that span the ‘phosphorylation-dependent degradation (PDD)’-domain containing both the p70S6K1 phosphorylation as well as the β-TrCP1-recognition motifs (Fig. 1A) were fused to a luciferase-encoding reporter construct (pGL3-control). Specifically, fragments coding for aa 64–80, aa 39–91 and aa 16–142 were used as target contructs, whereas a fragment coding for adjacent aa 108–206 was used as off-target construct (i.e. the resulting construct does not contain any part of the PDD-domain) (Fig. 1B). Transient co-expression of the resulting reporter vectors with wild type p70S6K1 yielded a reduction of the luciferase signal relative to co-transfection with an inactive mutant p70S6K1 to 25% and 33% for Pdcd4(16/142)luc and Pdcd4(39/91)luc, respectively. Pdcd4(64/80)luc (86%) and Pdcd4(108/206)luc (111%) did not respond to overexpression of wild type p70S6K1 (Fig. 2A).
For further experiments, Pdcd4(16/142)luc and Pdcd4(39/91)luc were chosen as target and Pdcd4(108/206)luc as off-target constructs. Stable HEK293 cell lines were generated expressing either of these 3 constructs. The two clones of each construct with the highest luciferase expression were selected for further analysis (Fig. 2B, black arrows). Pdcd4(16/142)luc cell lines (clone 5: 53%; clone 8: 54%) and Pdcd4(39/91)luc cell lines (clone 1: 42%; clone 8: 42%) displayed a strongly decreased luciferase signal in response to 8h TPA (10 nM) treatment. The Pdcd4(108/206)luc cell lines (clone 2: 79%; clone 7: 91%) were only mildly affected by TPA (Fig. 2C).
In summary, Pdcd4(39/91)luc was chosen as a sensitive Pdcd4-stability monitoring tool and Pdcd4(108/206)luc proved to be a valid specificity control. In all further experiments, Pdcd4(39/91)luc (clone 8) and Pdcd4(108/206)luc (clone 2) cell lines were used.
After initial characterization of constructs and cell lines, optimization of the assay focused on identification of conditions to maximize the difference in signal between DMSO and TPA controls and to provide a robust (high Z’) HTS-compatible 384-well assay. Systematic variation of cell numbers only minimally affected the TPA-induced loss of luciferase signal. Thus, optimal cell number determined by the highest Z’ value was 1000 cells/well (Z’ = 0.70) (Fig. 3A). Varying the concentration of TPA revealed that maximal reduction was achieved at 10 nM (56.2% of DMSO; Z’ = 0.68) (Fig. 3B). The TPA effect was maximal at 8h incubation as was Z’ calculated for the difference between DMSO and TPA (Fig. 3C). The slightly less pronounced effect at 4 and 6h is consistent with previous observations for endogenous Pdcd4 where maximal Pdcd4 downregulation was shown at TPA exposure ≥ 8h 7.
For control purposes, the PI3K inhibitor LY294002 and the mTOR inhibitor rapamycin were tested as potential Pdcd4 stabilizers based on their reported effectiveness in stabilizing Pdcd4 7. The TPA-induced decrease in Pdcd4(39/91)luciferase signal was attenuated by co-incubation with LY294002 or rapamycin (Fig. 4A). Treatment of cells with 10 nM TPA resulted in a similar loss of both endogenous Pdcd4 and the luciferase-target fusion protein (Fig. 4B). Treatment with rapamycin (100 nM) or LY294002 (10 μM) completely rescued both proteins. In contrast, luciferase fused to the off-target construct remained stable and was unaffected by the inhibitors (Figure 4B). Dose-response analysis for LY294002 and rapamycin using the screening assay yielded IC50s of < 0.1 nM and 2 μM respectively (Fig. 4C). Based on its higher efficiency, rapamycin was chosen as the recovery control on each plate. Varying the addition of positive controls between 2h pre- and 2h post-TPA-addition did not change the recovery of the luciferase signal (data not shown). Thus, rapamycin as well as the test compounds were added within 15 min after TPA addition during the HTS. DMSO concentrations up to 0.7% had no effect on the assay (data not shown).
Systematic variation of assay parameters thus confirmed optimal conditions for cell number, (1000 cells/well), TPA concentration (10 nM), and incubation time (8 h). Other variables such as read time, time and order of addition, etc. had minimal effect on the assay.
To evaluate the reproducibility of the assay in an HTS-format, 4 plates containing 1280 compounds of the commercially available LOPAC1280 (Library of Pharmacologically Active Compounds, Sigma, St. Louis, MO) were screened at 10 μM under optimal assay conditions on three days in triplicate each day. Figure 5 shows the relative activities (normalized to ΔDMSO/TPA) of two runs on two different days. The resulting slope of the curve was 0.86 and the correlation coefficient was 0.75. Variability for controls (DMSO, TPA, TPA + rapamycin) from plate-to-plate (CVs 6.4 to 8.1%), from day-to-day (CVs 4.9 to 7.5%) and within a plate (CVs 7.0 to 7.7%) was low. Plate-to-plate and day-to-day variability (summarized in Table 1) were higher for the compounds in the LOPAC1280 library (CVs 10.8 and 17.2%) and for 2111 randomly selected crude natural product extracts (CVs 6.9% and 17.3%)
These observations and an average Z’ value of 0.58 across 446 plates run under screening conditions demonstrate that the assay is highly reliable, reproducible and well-suited for HTS purposes.
After confirmation of the performance of the assay, various libraries containing a total of 15275 pure natural and synthetic compounds were screened at 5 or 10 μM. 47 confirmed hits were identified. One of these was shown to be nonspecific because it showed activity of 134.1 ± 5.4% in the off-target cell line. Thus, the final hit rate was 0.30% for confirmed, specific hits. Table 2 includes activities of several confirmed hits and depicts likely mechanisms of action for their activity as Pdcd4 stabilizers. Among the hits are mTOR inhibitors (rapamycin and analogs 18), general protein kinase inhibitors (7-hydroxystaurosporine 24, H-89, PP1 25), inhibitors of PKC signaling (GF-109203X 26), ERK inhibitors (hypothemycin 27) and previously described antagonists of phorbol ester activity (bryostatin 4 28). In addition to pure compounds, 135,678 natural product extracts were analyzed. For these samples, 43 were confirmed by re-assay. One of these was eliminated based on activity in the off-target cell line. Active samples included both organic and aqueous extracts from marine organisms, fungi, and plants.
Thus, the assay appears well-suited for high-throughput screening of both pure compounds and complex, natural product extracts, to reliably identify specific Pdcd4 stabilizers.
Expression of a fusion protein of a Pdcd4 fragment and luciferase (Pdcd4(39/91)luc, Fig. 1) was used to develop an HTS assay for inhibition of TPA-induced Pdcd4 turnover. The assay has proven to be robust and reproducible with regard to cell number, TPA concentration, and incubation times (Fig. 3). TPA treatment significantly and reproducibly reduced the luciferase signal (Z’ averaged > 0.5 throughout screening). Inhibitors of pathways leading to Pdcd4 degradation caused the signal to increase, albeit not to the level of untreated (DMSO) control cells. Application of the assay to libraries of pure natural products and synthetic compounds and to libraries of natural product extracts confirmed reproducibility of results with “real” samples.
As shown in Table 1, the CVs of all controls (DMSO, TPA, Rapamycin + TPA) were within a range from 4.7 to 8.1% for day-to-day, plate-to-plate and within plate repeatability, confirming the robustness of the assay. The CVs for the LOPAC1280 and the 2111 natural extracts were higher then those of the controls reflecting an increased variability that may also result from necessary dilution steps when screening libraries. The day-to-day variability was consistent and independent of the library source, and was higher than the plate-to-plate variability within a single day.
As a result of these observations, each plate was treated as an independent experiment. Decisions regarding identification of active samples were made based on controls in the same plate.
Previous work with Pdcd4 suggests that activation of PI3K/mTOR/p70S6K1 and MEK/ERK signaling result in destabilization of Pdcd4 7,15. During assay development, two inhibitors that target the PI3K/mTOR pathway were used for validation of the assay. The inhibitors were rapamycin (mTOR inhibitor) and LY294002 (PI3 kinase inhibitor). The IC50 values (estimated from data in Figure 4C) for rapamycin (< 0.1 nM) and LY294002 (~2 μM) are consistent with expected activities. Rapamycin was chosen as a positive control for the screening assay since it consistently gave a more complete recovery of activity at lower doses compared to LY294002. In addition to these compounds, a number of known, well-characterized compounds represented in the compound libraries were identified and confirmed as hits in the screening assay. Table 2 summarizes the activity data for several compounds and the target(s) likely to be involved in Pdcd4 stabilization. It should be noted that many protein kinase inhibitors once thought to be relatively specific, in fact have fairly broad inhibitory activity 25. For example PP1, identified by many suppliers as a src family kinase inhibitor also inhibits other protein kinases, including p70S6K1 and ERKs 25. It is therefore not surprising to see PP1 as a Pcdc4 stabilizer. While the hit compounds listed in Table 2 would be predicted to stabilize Pdcd4 in this assay, detection of these compounds alone does not validate the assay. However, these results demonstrate that the assay is capable of detecting modulators of multiple signaling pathways reported to be associated with Pdcd4 stability. Interestingly, a recently described AP-1 inhibitor, bryostatin 4 30, was identified as stabilizing Pdcd4. Pdcd4 has been characterized previously to inhibit AP-1 activity 3, thus, bryostatin 4 might exert its AP-1 inhibitory function by stabilizing Pdcd4.
A possible limitation of the assay is that toxic compounds will not be detected even if they might stabilize Pdcd4 at sub-toxic concentrations. For example, the general kinase inhibitor staurosporine which was represented in the tested libraries would have been predicted to be active, especially in light of the activity identified for 7-hydroxystaurosporine. However, the activity of staurosporine (31 ± 12%) remained below the hit threshold. Staurosporine also reduced the luciferase signal in the off-target cells to 62 ± 4%, indicative of mild toxicity and/or inhibition of protein expression. Similarly, other compounds potentially stabilizing Pdcd4 might appear as false negatives due to significant cytotoxicity under the assay conditions. This is further corroborated by screening results of natural product extracts. Out of > 135,000 extracts tested only 42 were confirmed to be active and specific. Many natural product extracts have previously shown significant levels of cytotoxicity in a similar assay 31.
A series of assays to determine the activity of the PI3K/Akt/mTOR pathway has recently been developed and published 30. As a measure of mTOR activity Carlson et al. included an assay that measures phosphorylation of Pdcd4 at Ser457, supposedly by p70S6K. Ser457 phosphorylation was recently reported to control Pdcd4 localization 16. While that assay provides a useful tool for identification of inhibitors of the PI3K/Akt/mTOR pathway, our system is designed to identify compounds affecting Pdcd4. As noted above, Pdcd4 stability is dependent on phosphorylation of a different site, ser67, 7,15 which is targeted in this assay. In addition to identifying inhibitors of PI3K/Akt/mTOR signaling, the assay presented here is predicted to detect modulators of other regulators of Pdcd4 degradation, including ubiquitylation and proteasomal degradation. Furthermore, in focusing on the phosphorylation site directly involved in mediating degradation of Pdcd4, we expect to identify modulators of site-specific phosphorylation events contributing to Pdcd4 destabilization. This approach will require subsequent secondary assays to stratify hits according to their mechanisms of action. One approach to investigation of mechanism(s) of action of hits vis a vis proteasomal degradation would be development of an alternative luciferase fusion construct with a different protein that uses the same proteasomal degradation pathway as Pdcd4. As a side note, while Carlson et al. used insulin and IGF as activators to address feed-back signaling events within the PI3K/Akt/mTOR pathway, we focused on tumor promoting conditions to identify compounds that could be useful as future tumor therapeutics targeting Pdcd4 stability to inhibit tumorigenesis. The two assays thus have different, albeit potentially complementary, purposes.
Natural product extracts comprise complex mixtures of compounds, so they can provide significant challenges to screening assays. Nevertheless, natural products and derivatives have been and continue to be important sources of novel therapeutics29,30, including the majority of current cancer chemotherapeutics29 The ability of this screen to identify active natural product extracts from a variety of taxonomic groups provides access to wide chemical diversity and can be expected to yield novel compounds and, possibly, novel mechanisms of action. The process of extract fractionation is being initiated. As far as we are aware, this is the first systematic, high-throughput investigation of natural product extracts for modulators of Pdcd4 stability.
In addition to detecting known compounds with well-characterized mechanisms of action, 37 other compounds were identified. For some of these, interesting biological activities have been reported, including possible antitumor activities. Further investigation will focus on characterizing the mechanisms of action for these compounds and on the purification and characterization of novel natural products that affect Pdcd4 stability. Compounds identified from this screen will be useful for probing Pdcd4 function and regulation as well as providing starting points for potential therapeutic intervention.
Thanks to Matthew Young and John Beutler for critically reading the manuscript.
This project has been funded in whole or in part with federal funds from the National Cancer Institute, National Institutes of Health, under contract HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.
This research was supported in part by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.