We have developed a cross-species comparative oncogenomics approach utilizing genomic analyses and functional genomic screens to identify and compare therapeutic targets in tumors arising in genetically engineered mouse models and the corresponding human tumor type.
Malignant Peripheral Nerve Sheath Tumors (MPNSTs) are derived from Schwann cells or their precursors. In patients with the tumor susceptibility syndrome neurofibromatosis type 1 (NF1), MPNSTs are the most common malignancy and the leading cause of death. These rare and aggressive soft-tissue sarcomas offer a stark future, with 5-year disease-free survival rates of 34-60%. Treatment options for individuals with MPNSTs are disappointingly limited, with disfiguring surgery being the foremost treatment option. Many once-promising therapies such as tipifarnib, an inhibitor of Ras signaling, have failed clinically. Likewise, phase II clinical trials with erlotinib, which targets the epidermal growth factor (EFGR), and sorafenib, which targets the vascular endothelial growth factor receptor (VEGF), platelet-derived growth factor receptor (PDGF), and Raf, in combination with standard chemotherapy, have also failed to produce a response in patients.
In recent years, functional genomic screening methods combined with genetic profiling of cancer cell lines have proven useful for identifying essential cytoplasmic signaling pathways and the development of target-specific therapies. In the case of rare tumor types, a variation of this approach known as cross-species comparative oncogenomics is increasingly being used to identify novel therapeutic targets. In cross-species comparative oncogenomics, genetic profiling and functional genomics are performed in genetically engineered mouse (GEM) models and the results are then validated in the rare human specimens and cell lines that are available.
This paper describes how to identify candidate driver gene mutations in human and mouse MPNST cells using whole exome sequencing (WES). We then describe how to perform genome-scale shRNA screens to identify and compare critical signaling pathways in mouse and human MPNST cells and identify druggable targets in these pathways. These methodologies provide an effective approach to identifying new therapeutic targets in a variety of human cancer types.
Malignant peripheral nerve sheath tumors (MPNSTs) are highly aggressive spindle cell neoplasms that arise in association with the tumor susceptibility syndrome neurofibromatosis type 1 (NF1), sporadically in the general population and at sites of previous radiotherapy1,2,3. NF1 patients are born with a wild-type copy of the NF1 tumor suppressor gene and a second NF1 allele with a loss-of-function mutation. This state of haploinsufficiency renders NF1 patients susceptible to a second loss-of-function mutation in their wild-type NF1 gene, which triggers tumorigenesis. When this "second hit" NF1 mutation occurs in a cell in the Schwann cell lineage, the resulting tumor is either a dermal neurofibroma arising in the skin or a plexiform neurofibroma that develops in large nerves or nerve plexuses. Although the pathology of dermal and plexiform neurofibromas is identical, their biologic behavior is quite different-although both dermal and plexiform neurofibromas are benign, only plexiform neurofibromas can undergo transformation and give rise to MPNSTs. In addition to the loss of neurofibromin, the Ras GTPase-activating protein encoded by the NF1 gene, MPNSTs carry mutations of multiple other tumor suppressor genes, including TP534,5,6,7, CDKN2A8,9, and PTEN10, mutations of genes encoding components of polycomb repressive complex 211,12 (PRC2; the SUZ12 and EED genes) and aberrant expression of receptor tyrosine kinases1,2. Mutations of NF1 and the other genes noted above are also present in sporadic and radiation-induced MPNSTs11,12.
While these advances in our understanding of the genomic abnormalities in MPNSTs have been invaluable for understanding their pathogenesis, they have not yet resulted in the development of effective new therapies for MPNSTs. A major barrier impeding the development of new treatments is the fact that MPNSTs are rare cancers. Because of this, it is difficult to obtain the large number of patient samples that are required for global analyses defining key driver mutations such as those undertaken by The Cancer Genome Atlas (TCGA). In our experience, accumulating even a modest number of human MPNST specimens can take years. To overcome such limitations, many investigators studying other rare cancer types have turned to the use of cross-species comparative oncogenomics to identify essential driver gene mutations, define the essential cytoplasmic signaling pathways in their tumor of interest, and identify new therapeutic targets. Since the signaling pathways that are essential for tumorigenesis are highly conserved between humans and other vertebrate species, applying functional genomics approaches such as genome-scale shRNA screens can be an effective means of identifying these new driver mutations, signaling pathways, and therapeutic targets13,14,15,16,17,18,19, particularly when studying rare human tumor types that are available in limiting numbers20.
In the methodologies presented here, we describe this approach to performing genomic profiling in human MPNST cell lines and early passage MPNST cultures derived from P0-GGFβ3 mice, a genetically engineered mouse model (GEM) in which Schwann cell-specific overexpression of the growth factor neuregulin-1 (NRG1) promotes the pathogenesis of plexiform neurofibromas and their subsequent progression to MPNSTs21,22,23. The first step in this approach is to identify candidate driver genes in P0-GGFβ3 MPNSTs, human MPNST cell lines, and surgically resected human MPNSTs. To functionally validate the signaling pathways affected by these mutations, we then use genome-scale shRNA screens to identify the genes required for proliferation and survival in human and mouse MPNST cell lines. After identifying the genes required for proliferation and survival, we then identify the druggable gene products within the collection of "hits" using the Drug Gene Interaction Database. We also compare the "hits" in human and mouse MPNST cells, to determine whether the GEM model and human MPNSTs demonstrate similar dependence on the same genes and signaling pathways. Identifying overlaps in the genes required for proliferation and survival and the affected signaling pathways serves as a means of validating the P0-GGFβ3 mouse model at a molecular level. This approach also emphasizes the effectiveness of combining human and mouse screens to identify novel therapeutic targets, where the mouse model can serve as a complement to the human screens. The value of this cross-species approach is particularly apparent when looking for therapeutic targets in rare tumors, where human tumors and cell lines are difficult to obtain.
Prior to the initiation of the studies, have animal procedures and protocols for handling viral vectors reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) and the Institutional Biosafety Committee (IBC). The procedures described here were approved by the Medical University of South Carolina's IACUC and IBC Boards and were performed by properly trained personnel in accordance with the NIH Guide for Care and Use of Laboratory Animals and MUSC's institutional animal care guidelines.
1. WES-Seq Analyses and Identification of Pathogenic Variants
2. Genome-Scale shRNA Screens
NOTE: Several shRNA and CRISPR libraries are available that can be used for genome-scale functional screens with low passage tumor cultures. Here, we describe the use of CELLECTA DECIPHER shRNA libraries as an example. CELLECTA DECIPHER lentiviral shRNA libraries are optimized for RNAi genetic screens in pooled format. Each transcript is targeted by at least 5-6 unique shRNAs and each lentiviral shRNA vector contains a unique genetic barcode flanked by PCR primer sites. These libraries cover the majority of human and mouse disease-relevant genes but do not cover all genes in the genome. Cellecta human library plasmid DNA pools are available in three modules (Human Module I, II, III; targets 15,377 genes) while the mouse library plasmid pools are available in two modules (Mouse Modules I and II; targets 9,145 genes). These libraries are used to perform "drop out" assays in which targeted genes that are required for proliferation and/or survival are differentially expressed at different time points after viral transduction.
3. Perform Cytometer Assays of Cell Numbers and Viability in MPNST Cells Challenged with Candidate Therapeutic Agents
Figure 5 plots display depletion scores of core essential genes (CEGs) labeled as TRUE compared to non-CEGs (labeled as FALSE) in each human cell line that was screened. Points represent log2 of fold depletion scores for individual genes, which are plotted over a boxplot representation of the overall score distribution. Student's t-test was used to test for a significant difference in the mean of depletion scores between the two groups in each cell line. The resulting p-values are indicated in each panel. Note that the average fold depletion scores are significantly higher for the CEGs than the non-CEGs. This is expected as Core Essential Genes are, by definition, consistently required for proliferation and/or survival in most cell types.
Figure 6A presents a Venn diagram of the "hits" for three human MPNST cell lines. We typically find that a large number of genes are shared between multiple lines; these hits are a high priority as they represent genes encoding proteins that are likely to be essential for the proliferation and/or survival of a large subset of MPNSTs. Note also that there are a number of genes that are hits in only one cell line. We encounter this commonly and it should not be taken as an indication that the screens are of poor quality. The genes that are common hits between multiple lines are then assessed using the Drug Gene Interaction Database to identify genes within this subset that encode proteins that are druggable with existing agents. We then select several of these and perform an initial validation by knocking down the expression of the corresponding mRNA with shRNAs. Since some shRNAs will have off-target effects, we always test multiple shRNAs targeting the same transcript. Figure 6B shows a representative result in which we have transduced MPNST cells with a non-targeting control and multiple shRNAs targeting BCL6. Cell numbers were then determined at varying times after transduction. Note that several of the BCL6 shRNAs markedly reduced cell numbers; as shown in the accompanying immunoblot, the degree of decrease in cell numbers correlates with the degree of BCL6 knockdown. Figure 6C shows a representative growth curve for an early passage P0-GGFβ3 MPNST culture.
Figure 1: Workflow for performing whole exome sequencing of MPNST tissue or early passage MPNST cells. Schematic illustrates the general workflow of variant detection present in tumor-derived early passage cultures. Isolate DNA from early passage cultures (1) and submit quality DNA to the sequencing core according to their submission protocols (2). The sequencing core will check the quality of the submitted DNA and perform all necessary sample and genome library preparations. The core facility will provide users with FASTQ sequencing files with quality control metrics (3). Users will upload the FASTQ files to a genome alignment and variant caller program of their choice. (4) Annotated variants are filtered on user-defined criteria to remove non-relevant variants. Representative data shown compare resected human MPNST tumor sample vs. a cell line derived from the tumor26. (5) Perform functional classification analysis with PANTHER. Abbreviation: MPNST = Malignant Peripheral Nerve Sheath Tumor. Please click here to view a larger version of this figure.
Figure 2: Workflow for performing viral transduction of the shRNA libraries into MPNST cells and isolating genomic DNA from the cells at time point 1 and time point 2. (A) Target cells are infected at a low MOI of 0.3 with barcoded lentiviral particles and selected for 72 h. Cells are passaged for 5-7 population doublings (approximately 7 days). Cell pellets at Day 0 and Day 7 are stored at -80 °C for genomic DNA isolation. Day 0 is referred to as Time point 1 (T1) and Day 7 is referred to as Time point 2 (T2). (B) Genomic DNA isolation begins with resuspension of cell pellets in resuspension buffer that are then split into two 15 mL tubes. To facilitate cell lysis, 10% SDS is added to each tube and sonicated for 25 cycles of 30 s on/30 s off. Following sonication, phenol/chloroform is added to each tube and vortexed vigorously for 45-60 s. Tubes are then centrifuged. (C) A clear upper phase is pipetted off and added to a clean tube with the addition of sodium acetate/isopropanol and mixed well. Tubes are centrifuged again. This time, the supernatant is discarded, and the pellet is dislodged with the addition of 70% ethanol. Combine the resuspended pellets into one tube and spin at maximum speed in a benchtop centrifuge. Discard the supernatant and resuspend the pellet in distilled water. Please click here to view a larger version of this figure.
Figure 3: Workflow for amplifying barcode sequences from lentiviral shRNA vectors in preparation for quantification of barcodes with next-generation sequencing. (A) Representation of how to set up tubes for the first nested PCR reaction (7 tubes: one for negative control, one for positive control, and the remaining 4 tubes for genomic DNA. The last tube will serve as the master mix tube). Following the first PCR reaction, combine the genomic DNA tubes into one tube and mix. (B) The products from the first nested PCR reaction will serve as templates for the second nested PCR reaction. Following the second PCR, combine genomic DNA tubes into one tube and mix. Please click here to view a larger version of this figure.
Figure 4: Workflow for purifying the amplified shRNA barcodes and sequencing barcodes to quantify their representation at time point 1 and time point 2. (A) Pour a 3.5% agarose gel. Because the volume of the pooled DNA exceeds the limit for one well, tape 4-6 teeth of a gel comb together to produce one large well. Prepare PCR products with 6x loading dye and load the positive control, negative control, and pooled DNA into the gel. Following electrophoresis, a large band at approximately 250 base pairs should appear in the pooled DNA lane. Using a clean scalpel, excise the entire band, and then cut into 4 gel slices. Solubilize the gel pieces and then combine them into two spin columns to elute DNA. Combine the eluted DNA into one tube. (B) The DNA is purified through a second purification step. Add Binding Buffer 2 to the tube of pooled DNA, and then pipette onto a spin column. Wash the membrane and then elute the DNA in distilled water. (C) Submit the purified DNA to the sequencing core. Please click here to view a larger version of this figure.
Figure 5: Representative examples of the distribution of Core Essential Genes after analysis as described in the protocol. In this example, three human MPNST cell lines (S462, T265, and 2XSB) cells were screened with Cellecta DECIPHER shRNA libraries. For each human MPNST cell line, a boxplot was created to compare gene-level depletion scores for genes in the list of Core Essential Genes (CEG; True box plot)25 to that of genes that are not found in the list of CEGs (False box plot). Individual data points are layered on top of each box plot. P-values are from a standard t-test comparing gene-level depletion scores of CEGs to non-CEGs. Abbreviations: CEG = core essential gene; MPNST = Malignant Peripheral Nerve Sheath Tumor. Please click here to view a larger version of this figure.
Figure 6: Validation of screen results. (A) Representative Venn diagram of overlapping hits in three human MPNST cell lines. (B) S462 human MPNST cells transduced with a non-targeting (NT) lentiviral vector and lentivirus expressing four different BCL6 shRNAs (shRNA1, shRNA2, shRNA3, and shRNA4). Cells were transduced with lentivirus and then treated with a selection agent (puromycin) for 3 days. Cell numbers were then assessed over the next seven days. (C) Western blot analysis showing protein levels of BCL6 following transduction with NT,shRNA1, shRNA2, shRNA3, and shRNA4 lentivirus. Please click here to view a larger version of this figure.
Table 1: Plate layout for lentiviral titering. Please click here to download this Table.
Table 2: Initial setup of first PCR reactions. Please click here to download this Table.
Table 3: Preparation of master mix for first PCR reaction. Please click here to download this Table.
Table 4: Cellecta first PCR parameters. Please click here to download this Table.
Table 5: Initial setup of second PCR reaction. Please click here to download this Table.
Table 6: Cellecta second PCR parameters. Please click here to download this Table.
The detailed methods presented here were developed to study peripheral nervous system neoplasia and MPNST pathogenesis. Although we have found these methods to be effective, it should be recognized that there are some potential limitations to the methods we describe here. Below, we discuss some of those limitations and potential strategies for overcoming them in other model systems.
We have found that whole exome sequencing effectively identifies mutations of interest in P0-GGFβ3 mice. It should be recognized, though, that whole exome sequencing itself has limitations. First, whole exome sequencing is not an effective approach to identifying fusion gene products. This is because the majority of chromosomal breaks and subsequent fusion predominantly involve intergenic regions and introns as those regions represent the majority of the genome. We have instead found that RNA-Seq with paired-end reads much more effectively identifies fusion genes. There is also the question of how effectively whole exome sequencing identifies relatively large regions of chromosomal loss.
Although several algorithms have been developed for identifying such losses, the term “whole exome sequencing” is itself misleading because the capture of the exome even in good runs often misses up to 5-10% of exonic regions. Because of this, we routinely complement whole exome sequencing with other approaches such as array comparative genomic hybridization (aCGH). After identifying gains and losses, we examine the genes within these intervals and compare them to the known driver mutations that are associated with their human counterparts. However, the mouse genome is more stable than the human genome27. Consequently, mouse tumors typically do not show chromothripsis analogous to what is seen in human neoplasms. The pattern in mouse tumors is instead much simpler, tending towards whole chromosome or chromosome gains or losses with relatively few focal deletions that tend to occur under strong selective pressure22,23.
There are some potential pitfalls that we have encountered when performing genome-scale shRNA screens. One of the most common problems that we encounter is the relatively poor transduction of the lentiviral vectors into the target cells. We most often find that the problem arose from improper titering of the packaged lentiviral pools. Because early passage mouse tumor cell cultures are a limiting resource, many investigators will instead attempt to titer their lentivirus using another established cell line that is more readily available. The problem with this approach, though, is that the efficiency of lentiviral transduction can vary considerably from cell type to cell type. It is for this reason that we recommend titering lentivirus on the actual cells that will be used in the experiment. We have also encountered problems with relatively low viral titers. That problem most often reflects poor transfection of 293T cells when producing the packaged virus.
It is possible to obtain false positive hits when performing genome-scale shRNA screens. Because of this, once we have identified the potentially druggable targets that are of the most interest to us, we always validate the results of our shRNA screens. Typically, we will use two different approaches to validate high-interest targets. First, we knock down gene expression using two or more shRNAs distinct from those used in the initial screen and determine the effect that this has on tumor cell proliferation and survival. Second, we obtain the drug(s) identified in the Drug Gene Interaction Database and determine the effect that this has on tumor cell proliferation and survival. We use both approaches in tandem because we have encountered circumstances in which the shRNAs work and the drug does not. In at least some of those instances, examination of the whole exome sequence dataset has shown that the targeted protein is produced by a gene that has a mutation potentially affecting drug-protein interactions.
The approaches outlined above will provide the investigator with an applicable means of identifying potential driver mutations that occur in rare neoplasms, functionally identifying the signaling pathways required for proliferation and survival, and prioritizing targets for therapeutic development. We hope that other investigators will find these approaches useful for identifying key therapeutic targets in other human cancers. The reader should be aware, though, that there are other functional genomic approaches that can be used to identify genes involved in tumor pathogenesis and genes encoding potential therapeutic targets. As an example, CRISPR libraries are available that can be used in a manner analogous to that we describe for shRNA libraries. Functional screens can also be performed in vivo to identify genes promoting tumorigenesis. As an example of this, the Sleeping Beauty transposon-based somatic mutagenesis system has been previously used to target Schwann cells and their precursors, resulting in the identification of several hundred genes involved in MPNST pathogenesis28. As these systems approach functional genomics in distinct ways, we recommend that the investigator carefully consider the goals of their planned experiments and base their selection of a functional genomics methodology on those goals.
The authors have nothing to disclose.
This work was supported by grants from the National Institute of Neurological Diseases and Stroke (R01 NS048353 and R01 NS109655 to S.L.C.; R01 NS109655-03S1 to D.P.J.), the National Cancer Institute (R01 CA122804 to S.L.C.), and the Department of Defense (X81XWH-09-1-0086 and W81XWH-12-1-0164 to S.L.C.).
Bioruptor Sonication System | Diagenode | UCD-600 | |
CASAVA 1.8.2 | |||
Cbot | Illumina, San Diego, CA | N/A | |
Celigo Image Cytometer | Nexcelom | N/A | |
Cellecta Barcode Analyzer and Deconvoluter software | |||
Citrisolve Hybrid | Decon Laboratories | 5989-27-5 | |
Corning 96-well Black Microplate | Millipore Sigma | CLS3603 | |
Diagenode Bioruptor 15ml conical tubes | Diagenode | C30010009 | |
dNTP mix | Clontech | 639210 | |
Eosin Y | Thermo Scientific | 7111 | |
Elution buffer | Qiagen | 19086 | |
Ethanol (200 Proof) | Decon Laboratories | 2716 | |
Excel | Microsoft | ||
FWDGEX 5’-CAAGCAGAAGACGGCATACGAGA-3’ | |||
FWDHTS 5’-TTCTCTGGCAAGCAAAAGACGGCATA-3’ | |||
GexSeqS (5’ AGAGGTTCAGAGTTCTACAGTCCGAA-3’ | HPLC purified | ||
GraphPad Prism | Dotmatics | ||
Harris Hematoxylin | Fisherbrand | 245-677 | |
Illumina HiScanSQ | Illumina, San Diego, CA | N/A | |
Paraformaldehyde (4%) | Thermo Scientific | J19943-K2 | |
PLUS Transfection Reagent | Thermo Scientific | 11514015 | |
Polybrene Transfection Reagent | Millipore Sigma | TR1003G | |
PureLink Quick PCR Purification Kit | Invitrogen | K310001 | |
Qiagen Buffer P1 | Qiagen | 19051 | |
Qiagen Gel Extraction Kit | Qiagen | 28704 | |
RevGEX 5’-AATGATACGGCGACCACCGAGA-3’ | |||
RevHTS1 5’-TAGCCAACGCATCGCACAAGCCA-3’ | |||
Titanium Taq polymerase | Clontech | 639210 | |
Trimmomatic software | www.usadellab.org |