|Home | About | Journals | Submit | Contact Us | Français|
KRAS is one of the most frequently mutated human oncogenes. In some settings, oncogenic KRAS can trigger cellular senescence, whereas in others it produces hyperproliferation. Elucidating the mechanisms regulating these 2 drastically distinct outcomes would help identify novel therapeutic approaches in RAS-driven cancers. Using a combination of functional genomics and mouse genetics, we identified a role for the transcription factor Wilms tumor 1 (WT1) as a critical regulator of senescence and proliferation downstream of oncogenic KRAS signaling. Deletion or suppression of Wt1 led to senescence of mouse primary cells expressing physiological levels of oncogenic Kras but had no effect on wild-type cells, and Wt1 loss decreased tumor burden in a mouse model of Kras-driven lung cancer. In human lung cancer cell lines dependent on oncogenic KRAS, WT1 loss decreased proliferation and induced senescence. Furthermore, WT1 inactivation defined a gene expression signature that was prognostic of survival only in lung cancer patients exhibiting evidence of oncogenic KRAS activation. These findings reveal an unexpected role for WT1 as a key regulator of the genetic network of oncogenic KRAS and provide important insight into the mechanisms that regulate proliferation or senescence in response to oncogenic signals.
Mutations in KRAS are commonly found in pancreatic, lung, and colon cancers (1). Furthermore, many cancers expressing wild-type KRAS have genetic alterations in genes known to function in the RAS pathway (2, 3). Thus, understanding the mechanisms responsible for KRAS-induced oncogenesis is an important goal in cancer research. While several RAS-effector pathways are well known, aberrant RAS signaling leads to alterations in a vast and only partially understood signaling network involving both forward activation via kinase cascades and negative feedback mediated by transcriptional regulation (4, 5). The importance of dissecting the complexity of this network is demonstrated by the dramatically distinct consequences of oncogenic RAS mutations depending on the cellular context. Activation of the RAS pathway can lead to either proliferation or senescence (6, 7). However, the precise mechanisms governing these distinct outcomes are not fully understood.
An important consequence of signaling downstream of RAS is a change in the expression of a large number of genes. Previous work has identified KRAS-specific gene expression signatures using mouse, zebrafish, or human model systems (8–11). Gene expression signatures are useful tools for identifying the complex signaling networks that drive diverse cellular processes in normal physiology and disease. For example, expression correlations between genes in microarray data have also been utilized to identify transcription factors that act as “master regulators” of oncogenic transformation (12, 13). Such an approach has not been systematically applied to the elucidation of oncogenic KRAS-driven signaling networks.
Negative-selection RNAi screens are a powerful approach for high-throughput functional analysis of genes in mammalian systems (14–17). Recently, RNAi screens have been used to identify STK33, PLK1, and TBK1 as synthetic lethal in cells expressing oncogenic RAS (18–20). As genome-wide screening approaches generally are limited by low signal and high noise and are thus nonsaturating, a more focused approach based on querying the functional significance of defined subsets of genes could yield novel insight into RAS biology. To test this hypothesis, an shRNA-negative selection screen was used to query an oncogenic KRAS gene expression signature and its putative transcriptional regulators. This approach identified what we believe to be a novel role for Wilms tumor 1 (WT1) as specifically required in cells expressing oncogenic KRAS but not in cells expressing wild-type KRAS. The synthetic effect of WT1 loss in cells expressing oncogenic KRAS was tested and confirmed in primary cells, in a genetically engineered mouse model, and in human cell lines. In both mouse and human cells, loss of WT1 activated a senescence program in cells expressing oncogenic KRAS but not in cells expressing wild-type KRAS. Increasing evidence points to a critical role for senescence as a barrier to oncogene-induced tumorigenesis (21). The studies described here highlight WT1 as a key modulator of senescence driven by oncogenic KRAS and establish what we believe is a previously unknown link between a critical oncogenic signaling pathway and a gene otherwise best characterized for its role in development and in tumor suppression.
A negative selection shRNA screen was carried out to identify genes required for Kras-driven oncogenesis. We included in the screen genes previously identified as part of a Kras gene expression signature (11) (n = 89) as well as potential transcriptional regulators identified using the Kolmogorov-Smirnov scanning (KSS) algorithm (13) (n = 35) (Supplemental Table 1; supplemental material available online with this article; doi: 10.1172/JCI44165DS1). KSS uses microarray data to establish a correlation matrix between transcription factors1…n and all other genes. This matrix is then used to identify enrichment of a given gene set within each list of transcription factor–correlated genes. In this analysis, a large compendium (n = 190) of gene expression microarray data from cancer tissues and cell lines was used to establish the correlation matrix (22). Transcription factors with high enrichment of Kras signature genes in their correlated gene lists were included in the screen (see Supplemental Methods). shRNAs against other genes were introduced in the library because they had been previously implicated as Kras effectors (n = 47) or because they were differentially expressed in prior microarray experiments performed on human non–small cell lung cancer (NSCLC) samples (n = 20) or cell lines (n = 23) (see Supplemental Methods and Supplemental Table 2). A primary goal was to identify shRNAs that confer a disadvantage to the proliferation of Kras mutant tumor cells and would thus be selected against in a pooled screen. To maximize the likelihood of identifying such shRNAs, a detection platform with high signal and low background was employed (23) (see Supplemental Methods). The performance of this system was first tested by analyzing the ability to detect negative selection of an shRNA against Kras (Supplemental Figure 1A).
Pooled vectors carrying shRNAs against target genes were used to produce virus (Supplemental Table 2) and infect mouse lung cancer cell lines LKR10 and LKR13 (24). Cells were allowed to proliferate in vitro or injected subcutaneously into immunodeficient mice (Figure (Figure1A).1A). The relative presence of each shRNA was quantified using fluorescent beads to which an shRNA-specific “barcode” had been attached (see Methods). Negatively selected shRNAs were defined as those with a statistically significant decrease in MFI at time point 1 (T1) and T2 (in vitro) or T3 (in vivo) when compared with T0 (Figure (Figure1B, 1B, Supplemental Figure 1, B–D, and Supplemental Table 3). To control for off-target effects, only genes that met these criteria with 2 or more shRNAs were considered. The 23 genes identified as negatively selected are shown in Table Table1.1. Importantly, the list includes 2 known Kras effectors (Braf and Rac1). Seventeen out of the 23 hits had 2 shRNAs that decreased the level of expression of the target by more than 50% (Figure (Figure1C).1C). Those shRNAs without at least a 50% decrease in expression of the intended target were considered likely to be off target and removed from further analysis. All on-target shRNAs demonstrated a profound impact on proliferation, confirming why they scored in the initial screen (Figure (Figure1D).1D).
To determine which of the previously identified genes might function as synthetic lethals in cells expressing oncogenic Kras, KrasLSL-G12D/+, or wild-type control, mouse embryo fibroblasts (MEFs) were infected with a pool of those shRNAs negatively selected in the initial screen (25). Relative changes over time for each shRNA were quantified as above (Figure (Figure2A).2A). Two shRNAs against 3 genes, Rac1, Phb2, and Wt1, were negatively selected in KrasG12D/+ MEFs but not in wild-type MEFs (Figure (Figure2B2B and Supplemental Figure 2). Rac1 has previously been linked to Ras signaling in lung cancer (26). Phb2 is a mitochondrial chaperone protein that regulates responses to apoptotic stimuli (27). Wt1 is a gene with multiple isoforms and pleiotropic cellular functions that was originally identified as a tumor suppressor linked to the pathogenesis of some cases of Wilms tumor (28). While best characterized as a transcription factor, Wt1 has isoforms that do not bind DNA but have been implicated in RNA processing (29). Wt1 was included in the initial screen because it was identified as a potential transcriptional regulator of the Kras signature. Thus, the role of Wt1 in Kras-driven oncogenesis was further explored.
To confirm that the oncogenic Kras-specific effect of Wt1 loss observed in MEFs was also seen in lung epithelial cells, the pooled secondary screen was repeated in an immortalized mouse lung epithelial cell line (MLE12) that expresses wild-type Kras (30). shRNAs against Wt1 were not negatively selected in MLE12 cells, whereas a parallel experiment in LKR10 and LKR13 demonstrated negative selection for the same 2 shRNAs (Figure (Figure2C).2C). The results suggest that the Wt1 shRNAs have a deleterious effect specific to cells that express oncogenic Kras.
To determine the effect of Wt1 loss on tumorigenesis, 2 shRNAs against Wt1 were used to infect LKR13 cells and perform a xenograft experiment. Wt1 suppression significantly decreased the size of tumors when compared with controls and to a level similar to that caused by suppression of Kras (Figure (Figure2D).2D). Taken together, these results suggest that in both mouse primary fibroblasts and lung epithelial cells, Wt1 expression is required only in the context of oncogenic Kras but is dispensable in cells expressing wild-type Kras.
To further characterize the interaction between oncogenic Kras and Wt1, we employed mice carrying a conditional allele of Wt1 (Wt1f/f) (31). Expression of Cre recombinase in Wt1f/f cells leads to in-frame deletion of exons 8 and 9, generating a shortened protein (Wt1Δ/Δ) lacking zinc fingers 2 and 3 that is severely compromised for DNA binding. Wt1f/f mice were crossed to KrasLSL-G12D/+ mice to obtain K-rasLSL-G12D/+;Wt1f/f MEFs. Infection of these MEFs with adenoviral Cre (AdCre) led to expression of oncogenic KrasG12D and loss of wild-type Wt1 (Supplemental Figure 3, A and B). Real-time RT-PCR (rtPCR) analysis revealed the expected decrease in Wt1 full-length mRNA levels in both Wt1Δ/Δ and KrasG12D/+;Wt1Δ/Δ MEFs compared with KrasG12D/+ MEFs (Figure (Figure3,3, A and B). Cell-cycle analysis 7 days after infection demonstrated a marked decrease in the number of cells in S phase in KrasG12D/+;Wt1Δ/Δ compared with Wt1f/f or Wt1Δ/Δ cells (Figure (Figure3C).3C). No differences in the number of apoptotic cells were detected (data not shown). A 3T3 assay confirmed the previous observation that KrasG12D/+ MEFs are hyperproliferative (25). Loss of wild-type Wt1 expression had no effect on MEFs expressing wild-type Kras. In contrast, KrasG12D/+;Wt1Δ/Δ MEFs demonstrated markedly attenuated cumulative population doubling (CPD) (Figure (Figure3D).3D). These results provide strong genetic evidence for an interaction between oncogenic Kras and Wt1 and rule out off-target effects that may have occurred in the shRNA functional screen.
This was surprising, as it has been assumed that oncogenic Kras leads to senescence in MEFs only when the oncogene is expressed at high levels (25). A key protein involved in senescence of both human and mouse fibroblasts is promyelocytic leukemia (PML) (32, 33). A dramatic increase in the localization of Pml to nuclear bodies (PNBs) in KrasG12D/+;Wt1Δ/Δ MEFs was detected (Figure (Figure3F).3F). At this time point, we also observed a significant decrease in the number of BrdU-positive cells and a decrease in the number of cells entering mitosis (Figure (Figure3,3, G and H). Thus, Wt1 is a critical component of the genetic network that allows oncogenic Kras to induce proliferation and bypass senescence in primary cells.
Oncogenic Kras induces senescence in MEFs when it is expressed at high levels (34) and the mechanism of induction of senescence is due to a marked increase of Erk phosphorylation (35, 36). In contrast, expression of oncogenic Kras at physiologic levels neither induces senescence nor increases phosphorylation of Erk (25). Sprouty1 and MKP3, negative regulators of Mek/Erk signaling, have been reported as transcriptional targets of WT1 (37, 38). Therefore, Wt1 deletion might act by altering the intensity of Mek/Erk signaling and thereby altering the cellular response to oncogenic Kras. However, no consistent significant difference in the phosphorylation of Erk was found in KrasG12D/+;Wt1Δ/Δ MEFs compared with KrasG12D/+ MEFs expressing normal levels of Wt1. Time course analysis was performed in 2 independent MEF lines (Supplemental Figure 4A and data not shown). Similar results were seen using a p-Mek–specific antibody (data not shown).
Activation of p16Ink4a or p19Arf is frequently seen in association with senescence in mouse and human cells. Therefore, the level of expression of these proteins in KrasG12D/+;Wt1Δ/Δ MEFs was assessed. We did not detect a difference in the expression of p16Ink4a or p19Arf when comparing KrasG12D/+ to KrasG12D/+;Wt1Δ/Δ MEFs at day 7 after infection. In both genotypes, p16Ink4a and p19Arf were increased to a similar level (Supplemental Figure 4B). As trp53 is also a critical regulator of senescence, we evaluated whether trp53 protein level or activity was altered in KrasG12D/+;Wt1Δ/Δ MEFs. No consistent increase in trp53 protein levels in KrasG12D/+;Wt1Δ/Δ MEFs compared with KrasG12D/+ MEFS could be detected. As with p16Ink4a and p19Arf, trp53 levels were increased to a similar level (Supplemental Figure 4C). Thus, the senescence phenotype induced by Wt1 loss in MEFs expressing oncogenic Kras involves modulation of genetic pathways either downstream or parallel to Erk, p16Ink4a, p19Arf, and trp53. These results demonstrate that Kras can lead to senescence even in the absence of Erk upregulation and that trp53 upregulation by itself is not a sufficient trigger for senescence in the presence of Wt1.
Gene expression microarrays were used to obtain a genome-wide view of the changes induced by loss of wild-type Wt1 activity. RNA for independent lines of wild-type (n = 6), KrasG12D/+ (n = 5), Wt1Δ/Δ (n = 5), and KrasG12D/+;Wt1Δ/Δ (n = 6) MEFs was generated 7 days after AdCre. This time point was chosen to provide enough time to see changes in gene expression in response to Wt1 loss while avoiding analyzing cells that were already phenotypically senescent (as these were not seen until days 10–11). A heat map including all probes with more than a 2-fold change across any of the 4 genotypes demonstrated large gene expression changes in both KrasG12D/+ and KrasG12D/+;Wt1Δ/Δ when compared with wild-type and Wt1Δ/Δ MEFs (Figure (Figure4A4A and Supplemental Table 4). Surprisingly, we were unable to identify many genes that were differentially expressed between KrasG12D/+ and KrasG12D/+;Wt1Δ/Δ MEFs, with commonly accepted statistical cutoffs (2-fold change and 10% false discovery rate [FDR] q value, as determined by Significance Analysis of Microarrays [SAM]).
These results suggested that the phenotype observed upon Wt1 loss is the result of subtle changes across genes in 1 or more pathways rather than large fold changes of expression across a few genes. To further assess this possibility, the Molecular Signatures Database (MSigDB), a large collection of curated gene sets, was used to query the KrasG12D/+ vs. KrasG12D/+;Wt1Δ/Δ data set using Gene Set Enrichment Analysis (GSEA) (39). Fifty-seven gene sets were enriched in KrasG12D/+ MEFs with an FDR of less than 25% (Supplemental Table 5). This included a gene set of glutamine metabolism genes, “HSA00251_GLUTAMATE METABOLISM” (KEGG curated glutamate pathway), as well as experimentally derived gene sets downregulated by glutamate or leucine starvation (“PENG_GLUTAMINE_DOWN” and “PENG_LEUCINE_DOWN”) (Figure (Figure4B4B and Supplemental Figure 5A). Notably, 3 experimentally derived, independent gene sets for Myc pathway genes were upregulated in KrasG12D/+ MEFs compared with KrasG12D/+;Wt1Δ/Δ MEFs (normalized enrichment score [NES] –1.8, –1.79, and –1.72; all with P < 0.05, Figure Figure4C4C and Supplemental Figure 5B). Enrichment of Myc target genes was further confirmed using 4 additional gene sets not included in MSigDB (40) (Supplemental Table 6). These 4 gene sets were also found to be upregulated in KrasG12D/+ MEFs compared with KrasG12D/+;Wt1Δ/Δ (NES –1.97, –1.77, –1.75, and –1.72; all with P < 0.001) (Supplemental Figure 5C). Myc is known to cooperate with Kras to drive oncogenesis and RAS effector pathways and also stabilize MYC by phosphorylation (41). A role for Myc in regulating glutamine metabolism has also recently been demonstrated (42, 43). The GSEA results suggest that Wt1 may function to modulate the output of Myc, perhaps by specifically regulating those genes involved in glutamine metabolism.
If WT1 target genes are important regulators of RAS-driven oncogenesis, it would be expected that patients with activated oncogenic KRAS signaling would benefit from lower expression of WT1-regulated genes. In other words, human lung cancer patients with activated RAS but with decreased expression of WT1-regulated genes (low WT1 gene signature) would have a better prognosis than those human lung cancer patients with increased expression of these genes (high WT1 gene signature). To test this hypothesis, a list of genes most consistently associated with Wt1 loss in MEFs was identified using the prediction analysis of microarrays (PAM) algorithm (44). PAM identified a gene signature consisting of 100 genes (62 with decreased expression in KrasG12D/+;Wt1Δ/Δ MEFs and 38 with increased expression in KrasG12D/+;Wt1Δ/Δ MEFs) that accurately distinguished samples that express KrasG12D/+ in the context of either wild-type Wt1 or Wt1 deletion (Figure (Figure4D4D and Supplemental Table 7).
This gene set of WT1-regulated genes was then used to analyze a recently published large cohort of lung cancer gene expression data (45). As KRAS genotype is not available for this data set, an oncogenic KRAS gene expression signature was first derived using a smaller, independent data set of lung tumor gene expression for which the corresponding KRAS genotype is known (Supplemental Table 6) (46). This oncogenic KRAS signature was then used on the larger cohort in order to stratify patient samples as having high, low, or intermediate KRAS signature status (Figure (Figure4E).4E). After cross-species mapping (Supplemental Table 8), the MEF-derived “Wt1 high” and “Wt1 low” signature was used to stratify patients in the human lung cancer KRAS signature high and low groups into those with high WT1 or low WT1 gene signatures (see Supplemental Methods and Supplemental Tables 9 and 10). Importantly, the high WT1 or low WT1 gene signature was able to separate lung cancers that express KRAS signature genes into poor and good prognosis groups, respectively. In contrast, the high WT1 or low WT1 gene signature had no prognostic significance in samples without evidence of KRAS signature activation (Figure (Figure4E).4E). These results further support the hypothesis that genes whose expression is modulated by WT1 are specifically relevant to human lung cancers carrying oncogenic KRAS.
To determine whether WT1 loss was deleterious to human cell lines expressing oncogenic KRAS, 2 shRNAs against human WT1 were used to knock down WT1 in a panel of human NSCLC lines with known RAS mutations status (Figure (Figure5A).5A). As expected, all KRAS mutant cells expressed high levels of KRAS protein (Supplemental Figure 6). Strikingly, a significant decrease in the cell viability of cell lines harboring KRAS mutations was observed in response to WT1 loss (Figure (Figure5B).5B). In contrast, a significantly lesser effect was seen in response to WT1 knockdown in NSCLC cell lines carrying wild-type KRAS. To extend this observation to other cell types in which KRAS mutations are present, we used a pair of human pancreatic cell lines with mutant (ASPC1) or wild-type (HPAFII) KRAS (Figure (Figure5C).5C). Consistent with the results in NSCLC, WT1 knockdown by 2 independent shRNAs lowered cell viability of ASPC1 cells to a significantly greater extent than HPAFII cells.
The genotype-specific effect of WT1 loss on cell viability was not due to increased apoptosis, as annexin V staining was not consistently different in cell lines expressing oncogenic Kras compared with those expressing wild-type Kras in response to WT1 knockdown (Figure (Figure5D).5D). In contrast, cell proliferation as measured by BrdU incorporation was significantly decreased in RAS mutant compared with RAS wild-type cell lines (Figure (Figure5E).5E). Mutant KRAS NSCLC cell lines adopted a senescence-like appearance in response to WT1 knockdown, and this was confirmed by staining with SA-βgal (Figure (Figure5,5, F and G). Thus, in response to WT1 loss, NSCLC cell lines expressing oncogenic KRAS showed a striking cell cycle arrest and senescence phenotype. Taken together, these results in human cells further support the role of WT1 as a regulator of proliferation in cells carrying RAS mutation.
To assess the role of Wt1 in Kras-driven tumor initiation and progression in vivo, KrasLSL-G12D/+ (n = 10) and KrasLSL-G12D/+;Wt1f/f (n = 13) mice were generated and AdCre was delivered to the lung epithelium to induce simultaneous deletion of Wt1 and activation of oncogenic Kras. Sixteen weeks after AdCre infection, deletion of Wt1 had a significant effect on tumor volume assessed both by microcomputed tomography (microCT) and histology (Figure (Figure6,6, A and B). While some large tumors remained in KrasLSL-G12D/+;Wt1f/f mice, PCR analysis of laser-capture microdissected tissue confirmed that in most cases those tumors retained at least 1 copy of Wt1 (Supplemental Figure 7). Analysis of tumor area excluding tumors with retention of a wild-type Wt1 allele (tumor area > 0.5 mm2) revealed a greater than 50% decrease of tumor area in KrasLSL-G12D/+;Wt1f/f compared with KrasLSL-G12D/+ mice (Figure (Figure6C).6C). Kras activation in the KrasLSL-G12D/+ tumor model led to the formation of multiple small adenomas, only a fraction of which would progress to adenocarcinoma. Analysis of total tumor number in KrasLSL-G12D/+ versus KrasLSL-G12D/+;Wt1f/f mice showed no significant differences (Figure (Figure6D).6D). However, binning of the total number of tumors in both genotypes into quartiles by size and assessing the percentage difference of the median size for each quartile revealed a gradual increase in the percentage difference from first to fourth quartile (Figure (Figure6E).6E). Median differences at each quartile were statistically significant between the 2 groups (Figure (Figure6E).6E). To assess whether this was due to changes in tumor cell proliferation, we stained lungs from KrasLSL-G12D/+ (n = 4) and KrasLSL-G12D/+;Wt1f/f (n = 4) mice with the proliferation marker Ki67. Tumors from KrasLSL-G12D/+;Wt1f/f mice had a significant decrease in the amount of Ki67-positive cells when compared with those from KrasLSL-G12D/+ mice (Figure (Figure6F).6F). These results strongly support a critical role for Wt1 in tumor progression of Kras-driven lung cancer in vivo.
In this work, WT1 was identified as what we believe to be a novel modulator of oncogenic KRAS signaling. In both mouse and human cells, WT1 regulates the proliferative potential of oncogenic KRAS, and loss of WT1 drives cells expressing oncogenic KRAS toward a senescence program. In a primary screen and in 2 follow-up screens to identify genes that specifically function in mutant Kras cells, Wt1 shRNAs were negatively selected in the context of oncogenic Kras. These results were validated using a genetic approach in mice as well as utilizing shRNA-mediated knockdown in human cells. Surprisingly, these results demonstrate that physiologic levels of expression of oncogenic KRAS in the absence of WT1 lead to senescence. Activation of RAS signaling either by mutation of RAS proteins or loss of NF-1 can lead to senescence (5). Despite the potential importance of senescence as a barrier to tumorigenesis, the precise mechanisms accounting for the activation of senescence in response to RAS remain poorly understood. Furthermore, previous analysis of oncogenic KRAS signaling in primary cells suggested that perhaps senescence is an artifact of KRAS overexpression, leading to supraphysiologic activation of ERK, and that senescence does not occur under “physiologic” levels of oncogenic KRAS (25). In contrast, here we demonstrate that in mouse primary cells, senescence in response to oncogenic KRAS can occur despite a lack of upregulation of ERK signaling in specific contexts, i.e., in the absence of Wt1. The importance of Wt1 expression was further demonstrated using human lung cancer cell lines and a mouse model of lung cancer. Wt1 deletion inhibits proliferation of Kras-driven lung tumors, thus inhibiting the progression of lung cancer. In addition, decreased expression of WT1-regulated genes is associated with improved prognosis in patients whose tumor gene expression profiles are consistent with activation of oncogenic KRAS. Taken together, these results using a combination of mouse and human cells demonstrate a surprising and unexpected role for WT1 as a critical regulator of the cellular mechanisms leading to either proliferation or senescence in response to oncogenic KRAS.
Several recent studies have utilized loss-of-function RNAi screens to identify genes that are selectively required in RAS-dependent human cell lines (18–20). Notably, WT1 loss was not identified as a synthetic-lethal interaction in the context of RAS in these studies. Negative selection screens are limited in their discriminative ability due to the noise inherent in any shRNA quantification method. Thus, in general, these screens are not saturating. As our screen was focused on fewer genes, this might explain why WT1 was identified whereas it was not in previously published work that included WT1 as part of the queried shRNA library (19). Another notable feature of the screen reported here is that it was performed in genetically defined murine cells. Previous work has demonstrated the utility of cross-species comparisons to leverage and filter gene expression profiles (11). This study provides further support for the utility of this cross-species approach by demonstrating that a signature derived from mouse primary cells with defined genetic mutations can be translated to lung cancer patients with oncogenic KRAS dependency as a predictor of overall survival. Thus, these results provide support for the use of functional screens using mouse shRNA libraries in order to study key molecular events in human oncogenesis.
While WT1 is best known as a tumor suppressor, considerable evidence points to a potential role for this protein as an oncogene in other tumor types (28). For example, almost all cases of the rare sarcoma desmoplastic small round blue cell tumor (DSRBCT) carry an EWS-WT1 translocation (47, 48). In addition, overexpression of WT1 has been reported in leukemias, breast, and lung cancer (49–51). WT1 has been demonstrated to play a role in regulation of apoptosis in a manner that is modulated by the serine protease HtrA2/Omi (52). Loss of HtrA2/Omi leads to increased WT1 protein levels, which confers resistance to apoptosis by cytotoxic drugs. Interestingly, in human lung cancer cell lines and in murine lung tumors, we detected very little apoptosis in response to WT1 depletion, suggesting that WT1 loss may have distinct consequences compared with WT1 upregulation. On the other hand, we observed that WT1 loss leads to senescence specifically in tumor cells harboring RAS mutations. Therefore, WT1 may function at least in part by ameliorating the oncogenic stress conferred by RAS mutations in these experimental systems. WT1 has also been shown to regulate expression of Snail at the transcriptional level, thus playing a key role in the differentiation toward mesenchymal lineages in embryonic stem cells and in the epicardium (53). Given these pleiotropic effects in normal and diseased cells, WT1 would appear to be a good candidate for a gene that can have “context-dependent” oncogenic or tumor suppressor effects (54). The complexity of the effect of WT1 overexpression and loss in normal and cancer cell physiology is perhaps partly due to the fact that WT1 has multiple isoforms with distinct biological functions. While some isoforms of WT1 are able to bind DNA and clearly have a role as transcription factors, other isoforms do not bind DNA but appear to participate in RNA processing. Further work will be needed to determine the specific targets of WT1 that are critical for regulating Ras-driven oncogenesis and senescence.
In conclusion, this work has identified what we believe to be a novel link between oncogenic KRAS signaling and expression of WT1. We used genomic data to inform a focused genetic screen with the goal of identifying regulators of oncogenesis. Our results demonstrate that the prooncogenic effects of WT1 are closely linked to elements of the KRAS signaling pathway. Further elucidation of this established link may provide avenues for therapeutic intervention in a wide range of human cancers. The general approach described here, combining computational analysis of expression signatures with cross-species functional genomics and genetics, should be applicable to other oncogenes and may provide insights into the complexity of the cancer genome.
The pLKO.1s lentiviral vector (gift of S. Stewart, Department of Cell Biology and Physiology, Washington University, St. Louis, Missouri, USA) was digested with EcoR1 to insert an oligonucleotide containing Xma and Nhe restriction sites. This vector was then digested with Xma and Nhe, and each of 100 oligonucleotides corresponding to the Luminex FlexMAP set (Luminex Corp.) was cloned into this site. The EcoRI site upstream of Xma was maintained and used to clone the shRNA hairpins together with the Age1 site upstream of the stuffer. Five shRNA hairpins against each of the genes of interest were designed using a publicly available algorithm ( http://www.broad.mit.edu/science/projects/rnai-consortium/trc-shrna-design-process). Then, 631 shRNAs against 162 candidate genes were cloned into the barcoded lentiviral vector (Supplemental Table 2) and sequence verified. Transfection-quality DNA was produced using a QIAGEN 96-well DNA kit. Vector DNA was quantitated and equimolar concentrations were pooled so that each “barcode” was represented only once in each pool. As the Luminex system detects 100 distinct barcodes, each pool had a maximum limit of 100 shRNAs (Supplemental Table 2). Pooled DNA was used to generate lentivirus and infect 2 mouse lung tumor cell lines (LKR10 and LKR13) at a low MOI so that on average each cell was infected with only 1 shRNA vector.
LKR10 and LKR13 cells were infected in triplicate with 8 pools of approximately 100 shRNAs each for 2 days. Cells were selected with puromycin for 3 days and either passaged every 3–4 days or injected subcutaneously into immunodeficient mice and grown for 21 days. DNA was harvested at indicated time points for further analysis.
Lentivirus was produced by transfection into 293FT cells as previously described (55), filtered, and applied directly to cells for infection at an MOI lower than 1.
Cells were plated into 96-well plates and treated with 3-[4,5-dimethlythiazol-2-yl]-2,5-diphenyl tetrazolium bromide (MTT). Experiments were read on the indicated days according to manufacturer’s instructions (Cell Proliferation Kit I; Roche). All data were normalized to day 0 of experiment.
Cell lines and MEFs were plated at 5 × 105 and 15 × 104 cells per 60 cm2 plates, respectively, and stained with annexin V following the manufacturer’s protocol (BD) at indicated time points. Annexin V analysis was done using FlowJo flow cytometry analysis software.
Cell lines and MEFs were plated at 5 × 105 and 15 × 104 cells per 60 cm2 plates, respectively, and treated with 10 μM BrdU for 4 hours previous to analysis, fixed, and prepared for flow cytometry as previously described (56). Cell-cycle analysis was done using FlowJo flow cytometry analysis software.
25 × 103 LKR13 infected with specified shRNAs was suspended in 100 ml of serum-free DMEM and injected subcutaneously into the 2 lower flanks of athymic Balb/cnu/nu mice (Charles River). One week after injection, tumor dimensions were measured every 3 days and tumor volume was calculated using the following: volume= π/6 × (length) × (width)2.
KrasLSL-G12D mice were crossed over 2 generations to Wt1f/f mice. MEFs were isolated from E13.5 embryos and genotyped by PCR. Early passage MEFs (p3–p4) were used for all experiments. All MEF experiments were done using 10% FCS/DMEM.
3T3 assays were as described (57). MEFs were infected with adenoviral GFP (AdGFP) or AdCre for 24 hours, allowed to grow for 1 day, and plated for 3T3 assays. SA-βgal staining assays were as described (58) and performed at day 11 after AdGFP or AdCre infection. For human NSCLC cell lines, SA-βgal staining was done 11 days after infection with specific shRNAs as described in ref. 59. At least 5 random fields taken at ×20 power using an inverted microscope were assessed to include a minimum number of 200 total cells per sample.
Cells were scraped and lysed in Mg2+ Buffer (for phospho-protein detection), NP-40 buffer (for p53 detection), or Laemmli buffer (for pml detection) supplemented with protease inhibitor cocktail (Roche), 25 mM sodium fluoride, and 1 mM sodium orthovanadate (Sigma-Aldrich). Protein samples were resolved by SDS-PAGE, transferred to Immobilon-FL membranes (Millipore), and incubated in blocking buffer (Li-Cor) for 1 hour prior to addition of primary antibody. Antibodies used were anti-mouse p53 (CM5; Novocastra), anti-PML (05-1708; Upstate), anti-p16 (M-156; Santa Cruz Biotechnology Inc.), anti-p21 (C-19; Santa Cruz Biotechnology Inc.), anti–K-ras (C-19; Santa Cruz Biotechnology Inc.), anti-Nras (C-20; Santa Cruz Biotechnology Inc.), anti-Hras (C-20; Santa Cruz Biotechnology Inc.), anti-Lamin A/C (636; Santa Cruz Biotechnology Inc.), anti-ERK (9102; Cell Signaling), and anti–p-ERK (9106; Cell Signaling). A WT1 antibody against the C terminus region of mouse Wt1 (GSDVRDLNALLPAVSSLGC) was produced and used for WT1 protein detection. Ras-GTP was pulled down with the Raf-1 Ras-binding domain (RBD) (17–218; Upstate). Following 3 TBS-T washes, infrared fluorescent-labeled secondary antibodies (IRDye 680 anti-rabbit or IRDye 800 anti-mouse; LI-COR) were incubated at room temperature for 1 hour and the membranes were scanned with the Odyssey Imaging Scanner (LI-COR). β-Actin antibody (clone 1A4; Sigma-Aldrich) was used as a loading control. p19Arf immunoblotting was performed (Ab-80; Abcam) as previously described (60).
MEFs were trypsinized at day 9 after AdCre or AdGFP infection, plated onto slides, and allowed to grow overnight. Cells were fixed using 4% PFA, processed, and stained with antibodies against pml (05-1708, Upstate) and BrdU (555627; BD). For quantitation, at least 5 random fields taken at ×40 power were assessed to include a minimum number of 400 total cells per sample. Ki67 staining was performed as previously described (61).
RNA was isolated 7 days after lentiviral infection using TRIzol reagent (Invitrogen) following the manufacturer’s specifications. cDNA was synthesized with a DyNAmo cDNA synthesis kit (F470; New England Biolabs), and rt-PCR was performed using SYBR Green (Applied Biosystems) (see Supplemental Methods for primer sequences).
RNA was isolated using TRIzol 7 days after infection of MEFs with AdCre. RNA was further prepared by passage over an RNeasy column. cDNA synthesis, biotinylation of cRNA, and hybridization to mouse genome 430 2.0 array was performed according to the manufacturer’s instructions (Affymetrix). Microarray data was normalized with dChip software ( http://biosun1.harvard.edu/complab/dchip/). Low signals (below 50) were filtered out using the PreprocessDataset module in GenePattern ( http://www.broad.mit.edu/cancer/software/genepattern/). Differentially expressed genes in each group were identified using SAM ( www-stat.stanford.edu/~tibs/SAM/) and Prediction Analysis for Microarrays ( http://www-stat.stanford.edu/~tibs/PAM/). Genes with an FDR below 5% and fold change over 1.5 were included in the respective up/down lists. GSEA ( http://www.broadinstitute.org/gsea/) (39) was used to compare our data sets with gene sets derived from previous studies. Raw data are available in Gene Expression Omnibus (GEO GSE15325).
Mice were anesthetized with 2% isoflurane and scanned using a GE Healthcare microCT scanner. Each scan was performed for 7 minutes with a resolution of 97 μm. Images were visualized using GE MicroView software.
Animals were intercrossed to generate KrasLSL-G12D/+ or KrasLSL-G12D/+;Wt1f/f animals. Mice were in a mixed 129/Sv and FVB background. Genotyping of mice was done on DNA extracted from tail clippings as described previously (25, 31). All animal experiments were approved by the Stanford University School of Medicine Committee on Animal Care (APLAC).
Micrographs of H&E slides from multiple lung sections for each sample were obtained and pictures were taken at ×20 magnification. Bioquant software was used to montage the entire lung sections and to calculate tumor area for each sample. Tumor area was calculated by using the manual measurement feature in the Bioquant software. Tumor area was presented as mm2.
LKR10 and LKR13 were a gift of Julien Sage (Stanford School of Medicine, Stanford, California, USA) and were grown in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin. MLE12 cells were purchased from ATCC and grown according to vendor’s specifications. Human non–small cell lung cancer lines used either had wild-type RAS alleles (NCI-H522, NCI-H1437, NCI-H1568, NCI-H1650, NCI-H1975, and NCI-H2126) or were mutant for RAS (NCI-H23, NCI-H358, NCI-H441, NCI-H460, NCI-H727, NCI-H1299, NCI-H2009, and NCI-A549). All cell lines were grown according to ATCC specifications.
A 2-tailed t test was used for comparisons of different groups. Error bars correspond to either SD or SEM. P < 0.05 was considered significant.
We thank Sheila Stewart for providing us with the pLKO.1s vector. We thank Julien Sage, Steve Artandi, Laura Attardi, Kevin Haigis, Erica Jackson, and all the members of the Sweet-Cordero lab for helpful discussions. We thank the Stanford High-Throughput Biosciences Center and David Solow-Cordero for assistance in the execution of the pooled shRNA screen. We thank the members of the RNAi Platform at the Broad Institute and The RNAi Consortium (TRC) for advice and technical support. S. Vicent was supported by fellowships from the Fundación Alfonso Martín Escudero and the Tobacco-Related Disease Research Program of California (TRDRP). This work was supported in part by the NIH and the National Cancer Institute (I-K08CA98096-01 and 1R01CA138256-01 to E.A. Sweet-Cordero). E.A. Sweet-Cordero was also supported by grants from the Sidney Kimmel Foundation, the Rita Allen Foundation, the Joan’s Legacy Foundation, and the Harold Amos Medical Faculty Development Program. W.C. Hahn was supported in part by the National Cancer Institute (R33 CA128625) and the Starr Cancer Consortium (I1-A11).
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2010;120(11):3940–3952. doi:10.1172/JCI44165.
See the related Commentary beginning on page 3804.