1. Executive Summary
The literature regarding neurofibromas is briefly reviewed, followed by a review of recent studies regarding potential biomarkers. Several promising methods for biomarker discovery and final assay platform are then presented.
Work to date has provided a large number of candidate biomarkers including gene mutations, up- and down-regulated gene expression, DNA copy number variations, and biochemical changes in proteins secreted by neurofibromas. The number of candidates is greater than could be incorporated into a single diagnostic assay, yet biomarkers have not yet successfully predicted presence, number, or volume of benign tumors. The plethora of biomarkers has not predicted rate of benign tumor growth nor progression to malignancy.
I do not propose a new previously unidentified biomarker, but propose that the current slate of biomarkers be considered in multiplexes. This requires [a] application of methods appropriate for high multiplexing such as Next Generation Sequencing and [b] study of statistically significant numbers of normal and tumor (benign and malignant) biopsies. A Discovery Phase is proposed in which serially obtained biopsies are tested for a panel of candidate biomarkers.
Several methods that are appropriate for multiplex testing are discussed. Following the identification and validation of biomarkers that correlate to the desired tumor progressions, an assay must be developed that can test for the (presumed) multiple biomarkers. This project proposes Next Generation Sequencing (NGS) as the most applicable platform to detect and monitor multiple biomarkers that indicate and predict tumor progression. NGS instrumentation is commercially available and is already being adopted within clinical labs.
2. Introduction and Review of Plexiform Neurofibroma
Neurofibromatosis type I (NF1) is a dominant autosomal disorder with an incidence of one birth in 2500, and a prevalence of about 1:4000. NF1 is caused by mutations of the tumor suppressor gene Neurofibromin 1 (NF1). The gene codes for neurofibromin, a negative regulator of RAS proteins. NF1 patients are at risk of developing benign and malignant tumors. The benign tumors can be dermal (DNF) or plexiform (PNF) neurofibromas. The PNFs are caused by a single autosomal dominant genetic mutation in NF1. PNFs are congenital or appear before the age of 15 in 30-50% of NF1 patients. PNFs may become malignant in about 10% of patients, forming malignant peripheral nerve sheath tumors (MPNSTs). About half of all MPNSTs are sporadic rather than inherited through NF1. A sporadic tumor appears as an isolated feature in a given patient without other signs of NF1.
The expression of NF1 in terms of tumor occurrence, number and size varies between patients, even within the same family. There are no known correlations between type of NF1 mutation and the presence or behavior of PNFs.
In patients with NF1 (that is, one copy of the gene on Chromosome 17 knocked out), tumorigenesis is expected to occur when a somatic mutation knocks out the second copy of NF1. The expected chromosomal loss of heterozygosity (LOH) was seen in cells isolated from PNFs and MPNSTs. Thomas et al. (2012) sought evidence of LOH and somatic NF1 mutations by sequencing DNA and RNA from tumors or cell cultures derived from Schwann cells (the most likely cells of origin in MPNSTs). No association was noted between the type of germline and somatic NF1 lesion within the same individual. Indeed, some sporadic MPNSTs appeared to be wild-type at the NF1 locus (Miller et al., 2006). In another case, a sporadic tumor was shown to be the result of two independent hits that knocked out both chromosomal copies of NF1 (Beert et al., 2012).
Nerve tumors are diagnosed by MRI neurography, CT scans, and Electromyography (EMG). Imaging using a PET/CT scan and image-guided stereotactic biopsy can resolve if the tumor is cancerous. Granted, imaging plays an important role in regular checkups and confirmation of patient tumor burden. The proposed project, however, rests primarily on biomarker tests that can be done remotely. There is one key exception, described below.
Questions remain as to [a] what specifically triggers formation of benign PNFs, [b] what causes existing but quiescent PNFs to grow rapidly, and [c] what causes benign PNFs to transform into malignancy by forming MPNSTs? The project involves proposing specific biomarkers or a new approach for finding such markers, specifically with the intention of developing a diagnostic test applicable to pediatric patients.
Previous Studies of Differential Expression
Many reports in the literature propose candidates based on several technologies. (Miller et al., 2006) used Affymetrix arrays to identify a 159-gene signature to distinguish MPNST cell lines from Schwann cells. Other, more recent, reports have described approaches and genes that may be involved in progression to malignancy. Several of these reports indicate involvement of genes that play roles in other cancers.
The literature describes implicated genes including IRS1/IRS2, FAK, CD34, podoplanin, adrenomedullin and CDKN2A/B. Below, I will review studies that specifically relate to this proposal. These include studies of genes, expressed proteins, SNPs, and microRNAs.
Copy Number Variations
Beert et al. (2011) compared chromosomal copy numbers in benign, atypical, and malignant tumors. Benign tumors had almost no chromosomal copy number aberrations. Copy number loss of the CDKN2A/B gene locus was one of the most common events in the group of MPNSTs, with deletions in low-, intermediate-, and high-grade MPNSTs. These data support the hypothesis that atypical neurofibromas are premalignant tumors, with the CDKN2A/B deletion as the first step in the progression toward MPNST.
Pasmant et al. (2011a) studied human PNF tissues using high-resolution array comparative genomic hybridization. They focused on human chromosomal region 9p21.3, which contains five cancer-susceptibility SNPs. They found that deletions, including the CDKN2A/B-ANRIL locus, were the only recurrent somatic alterations in PNFs. This suggested that the antisense noncoding RNA in the INK-4 locus (ANRIL) locus may be involved in neurofibroma formation. Mußotter et al. (2012) further investigated the 9p21.3 region SNP rs2151280, which correlates with reduced ANRIL expression. They found neither the PNF number nor PNF volume was found to be associated with the T-allele of rs2151280.
Stewart et al. (2012) studied the cytogenetics and SNP array-based copy number analysis for NF1-associated glomus tumors along with germline DNA. They found LOH of chromosome 17q in two of five tumors, although the chromosomal copy number remained the same. This indicates a mitotic recombination event.
Mouse models of NF1 previously did not fully reflect the genetic heterogeneity observed in humans. Recently, however, Kazmi et al. (2013) described a mouse model for progression of neurofibromas to malignancy that found reproducible chromosomal alterations in mouse MPNST cells and focal gains and losses affecting 39 neoplasia-associated genes (including Pten, Tpd52, Myc, Gli1, Xiap, and Bbc3/PUMA). They believe their Po-GGFb3 mouse provides a medically relevant model for PNFàMPNST progression in humans. Kazmi et al. also note that several human genes proposed to promote neurofibromaàMPNST progression (Storlazzi et al., 2006) map to mouse chromosome 11, which was altered in mouse MPNST cells. Noting the number and wide mutation spectrum of neoplasia-associated genes, they concluded that the may be multiple pathways leading to MPNST pathogenesis.
Chen et al. (2012) cultured benign neurofibroma cells and determined the protein secretions (‘secretome’) relative to normal Schwann cells. They found that neurofibroma cultures increased secretion of 22 proteins and decreased secretion of 9 proteins relative to cultured nerve cells. They found that secretion of tumor suppressor protein RARRES1 exclusively in NF1 cell cultures.
Park et al. (2013) found that the anti-apoptotic Bcl-xL protein is upregulated in MPNST tissues compared to PNF tissues from patients with NF1 by immunohistological staining. They concluded that upregulation of Bcl-xL in MPNST-derived Schwann cells may be caused by the NF1 deficiency-mediated elevation in Ras/MAPK signaling.
Pasmant et al. (2011b) used quantitative RT-PCR to study expression of 16 genes and two microRNAs coded for in a region often microdeleted from patients. Samples were DNFs (assumed negative control since they do not progress to malignancy), PNFs, and MPNSTs. Their results indicated that several genes in the microdeletion region, particularly CENTA2 and RNF135, may be involved in increasing the risk of malignancy.
Is an NF1 mutation assay of any medical value? The NF1 gene is localized to chromosome band 17q11.2, and is comprised of 60 exons and 350 kb of genomic DNA. The Sanger COSMIC database provides information on NF1 mutations observed in at least 32 different tissues across 127 reports. More than 500 different NF1 mutations have been identified to date, of which most are unique to a particular kindred. A qPCR test has recently been reported (Terribas et al., 2013). Miller et al. (2006) found at least some sporadic MPNSTs that lacked NF1 mutations, indicating that malignancy can arise independent of an NF1 loss. NF1 expression has also been found in MPNSTs, indicating that NF1 patients do not all require a second gene hit to progress to malignancy.
In a CDMRP 2012 Investigator Vignette entitled “Investigating a Novel Tumor Suppressor That Cooperates with NF1 in MPNST Development,” Dr. Karen Cichowski of Brigham and Women’s Hospital said she “performed a screen and identified a gene known as SUZ12 within this region that appeared to function as a tumor suppressor and cooperates with NF1. Our genetic collaborator Eric Legius then found mutations in the SUZ12 gene that occur in tumor types from all NF1 patients, from tumors from all NF1 patients… We are looking right now specifically for what genes are impacted by the combined mutations in NF1 and this other gene.” Pasmant et al. (2011b) flagged SUZ12 as a candidate, as noted above.
There seems to be a paucity of diagnostic clinical trials at this time. There are 84 NF1-related clinical trials, of which 22 are related to plexiform neurofibromas, listed as of 2/12/13 at ClinicalTrials.gov. There are total 27 trials related directly to plexiform neurofibromas. These include 21 therapeutic trials, one imaging trial, and observation trials. One is NCT00340522, which involves making tissue microarrays. In another (NCT00006435), samples are collected for later study. The latter collect tissue only if surgery is done as part of Standard of Care; no additional surgeries or procedures are added to the patient’s regimen. Another, NCT01218139, will screen tumors by microarray comparative genome hybridization and full exome sequencing (this study has not been updated since 2010).
We may summarize prior art as follows: No single biomarker distinguishes all tumors, benign or malignant. Studies have uncovered numerous genes whose expression differs between nontumor, PNF, and MPNST. A diagnostic assay need not measure a causative agent, as long as a target correlates monotonically with the medical result. An approach to determining risk of tumor growth or malignancy may be to test a panel of genes, proteins, microRNAs, or other biomolecules. Or a different approach altogether may be needed.
Here we review recent reports that employs methods useful to the proposal.
Luciferase Knockin Expression
Burd et al. (2013) developed a luciferase knockin mouse (p16LUC), which reports expression of p16INK4a, a tumor suppressor. This allows ready visualization of tumors at an early stage. Background fluorescence, however, increased overall as the knockin mice aged. Tumors were clearly marked visually. The tumors themselves did not express the marker, but the surrounding stromal tissues did. Activation of p16INK4a was noted in the emerging neoplasm and surrounding stromal cells. This indicates that tissue surrounding a nascent tumor responds to tumor changes in status. It is not clear what biomolecule(s) respond, and a proposed research project involves comparing stromal tissue surrounding each type.
The fluorescent detection of tumors is intriguing, but requires the luciferase knockin gene. Therefore, the part of Burd et al. (2013) that pertains to the project is their examination of stromal tissue expression changes. In general, studies have focused on changes within the actual tumor of interest. The neighboring cells may be as or even more sensitive to changes, possibly with biomarkers that are as early or earlier than from the tumor directly.
Approaching oncology from a different angle, Chernet & Levin (2013) measured voltage in X. laevis embryos to detect tumor cells. They studied the role of transmembrane voltage potential gradients in tumorigenesis. Expression of mammalian oncogenes can induce formation of tadpole tumor structures, whose membranes are depolarized. These could be detected by fluorescent voltage reporter dyes before becoming morphologically apparent. Hyperpolarization of these membranes through one or more ion channel types can reduce incidence of these tumor precursors. The authors state, “these data identify a convenient new non-invasive marker for diagnosis,” although they do not explain how one might implement optical sensing noninvasively on humans. It is possible, however, to induce cell hyperpolarization by pharmacological or genetic methods to prevent tumor formation. Several ion channel drugs are currently approved for human use and might potentially be repurposed.
One used to search for tumor biomarkers is serological analysis of recombinant cDNA expression libraries of human tumors with autologous serum (SEREX). This approach has been applied to melanomas, sarcomas, neuroblastomas, and many other cancers. Yet it does not appear in the literature applied to neurofibromas. The approach is as follows (Chen, 2004):
One recent application to colon cancer antigens is the study of Hanafusa et al. (2012), in which they identified the new antigen candidate TEKT5. The result of the method is to combine serological analysis with antigen cloning techniques to identify human tumor antigens eliciting high-titer IgG response. Therefore, candidates are specifically intended as immunoassay targets. Background information can be found at the Cancer Immunity website http://cancerimmunity.org/serex/methodology/#steps.
DNA Copy Number Variations (CNV) as an Assay Platform
Gene copy numbers can be elevated in cancer cells. For example, the EGFR copy number can be higher than normal in non-small cell lung cancer and correlated to patient response and time to progression (Cappuzzo et al., 2005). Studies have correlated CNVs to susceptibility to cancers of the breast, testis, lung, bladder, and other organs. In one recent example, Liu et al. (2013) compared tumor and matching normal thyroid tissue-derived DNA using SNP arrays. They found differential amplifications of two chromosomal segments, and validated a five-gene panel that distinguished adenomas from carcinomas. This approach can provide an assay of diagnostic and/or prognostic value.
Prostate cancer has a medical situation similar to NF1 progression in that physicians seek biomarkers that will predict which cancers will remain slow growing and which will fulminate and metastasize. Yu et al. (2012) used Affymetrix SNP chips to examine CNVs from tumors, benign tissue adjacent to tumors, and blood from prostate cancer patients. They found that genomic abnormalities in all three samples, either benign or malignant tissues, were predictive of relapse and the kinetics of relapse. From 85 samples, the blood-based CNVs correctly predicted 81% of cases for relapse and 69% of cases for short PSA doubling time.
A similar approach can be applied to NF1 by comparing CNVs from PNFs, MPNSTs, and adjacent stromal tissue along with blood samples. Beert et al. (2011) compared chromosomal copy numbers in benign, atypical, and malignant tumors as discussed above. This promising work can be expanded for a potential basis of a diagnostic assay, as discussed below.
Next Generation Sequencing (NGS)
Recent developments in NGS instrumentation and methods have provided the ability to determine the entire DNA sequence of entire genomes. As costs drop, clinical labs are considering the use of NGS for routine diagnostic procedures. Recent examples of clinical use include detection of chromosomal aneuploidy in fetal DNA, identification of rare genetic variants in Mendelian disorders and detection of mutations in cancer genes such as TP53 (Link et al., 2011). Somatic mutations in lung cancer have been identified by NGS, which can now be applied to single cells and free circulating DNA (Daniels et al., 2012).
3. Proposed Biomarker Project and New Experimental Work
The proposed project includes a combination of approaches. To date, the field has amassed a large number of promising candidate biomarkers, no single one of which has proven to be diagnostic or prognostic. The project should address the following:
- Detect both NF1 and sporadic tumors.
- Predict risk of PNF growth rate.
- Predict risk of PNF progression to malignancy.
- Diagnostic test applicable to pediatric patients.
The proposed project involves application of [a] methods that extend published and ongoing work to determine and validate one or a multiplex of biomarkers, and [b] approaches that are being applied to other biomedical problems for a proposed assay.
The project assumes that no single biomarker will answer any one of the above points, much less all of them. It is instead assumed that each phase of tumor growth (presence, growth rate, and progression to malignancy) may be predicted or detected by a panel or multiplex of biomarkers.
The studies briefly described above have provided evidence of numerous biomarker candidates, yet none of these provided a satisfactory test. Two major issues with candidates to date are the small patient sample size and the fact that no single mutation or marker is present in all cases. This indicates that as a minimum, the current slate of reported biomarker candidates must be tested as part of a multiplex. Criteria include testing on a single assay platform and consideration of the relevant mutation spectrum.
The path of assay development is assumed to progress from tumor biopsy during the Biomarker Discovery phase to blood or urine in the assay Development phase. There are several NF1 research organizations around the world that may cooperate in the acquisition of biopsies. The few ongoing clinical trials as noted above do not collect large numbers of biopsies. Many studies do not collect any biopsies outside of those already dictated by Standard of Care guidelines. New studies must be funded and initiated to provide a statistically significant number of tissues of each type. These include wild type Schwann cells, PNFs, MPNSTs, and adjacent stromal cells.
The proposed project assay outputs will be:
- Presence/Absence of indications of a sporadic or NF1 tumor.
- Presence/Absence of benign tumor growth indications.
- Presence/Absence of indications of progression to malignancy.
These outputs may require confirmation in the form of an imaging scan of the patient.
Because the developed assays will be used on pediatric patients, tissue biopsy is considered unacceptable even if it would provide the most accurate result. Blood or urine are the ideal samples for testing, but these have a limitation: such fluids will give an average result, and will not indicate which specific tumor will grow faster or progress to malignancy. Upregulation of a marker is far more likely to be detected from blood than downregulation, especially if the tumor contribution to the total biomarker presence is diluted by other organ sources. For example, of the 159 candidates identified by Miller et al. (2006), they listed 13 as consistently overexpressed in MPNSTs. This, as a first pass, can reduce the total candidates to a ‘reasonable’ number.
Determination of Biomarkers
The literature review above describes some but not all of the studies which have generated potential biomarkers that could identify presence and growth potential of tumors. The project expects continuation and extension of these efforts, including the tumor status-specific Secretome (Chen et al., 2012), microarray-derived candidates (Miller et al., 2006), genes from mouse models (Kazmi et al., 2013), quantitative RT-PCR (Pasmant et al., 2011b), CNV (Beert et al., 2011), and other studies. Diagnostic markers may also be found by methods that detect correlative (but not necessarily causative) biomarkers such as SEREX.
In particular, the project urges that a dedicated NGS project be implemented to sequence a malignant tumor biopsy and some non-malignant biopsies from the same patient. Initially, a Whole Genome Sequencing (WGS) project should examine biopsies from at least the four NF1-related tissues (wild type Schwann cells, PNFs, MPNSTs, and adjacent stromal cells) to define the genomic bases of these tissues. This must be performed on serial samples to seek differences that define PNFs that grow at different rates, and define which PNFs will progress to malignancy. It is understood that recruitment may be problematic as it requires multiple and serial biopsies from pediatric patients.
The WGS concept includes in-depth sequencing of the sample tissue. This is particularly important in the case of a heterogeneous tissue. Within each tumor, the cell population includes cells that are and are not progressing to growth and/or malignancy. The WGS output will provide in-depth sequencing of the entire mixed cell population. This allows detection of rare copies of a differential mutation or CNV at an early stage of progression.
The WGS approach applied to biopsies for Discovery need not be used for the clinical assay. As discussed above, the final assay will use blood or possibly urine. If the medically efficacious assay requires only a limited number of targets, then a more efficient approach is to use targeted genomic sequencing which will examine only the selected fewer gene targets. NGS can produce millions of sequences per sample. Because WGS aims at the entire genome, any given gene may have only a few sequencings. A targeted approach generates the same millions of sequences, but they are divided only by the smaller number of desired targets of interest.
Another method of sensitive detection of CNVs is digital PCR. One method is digital drop PCR (ddPCR), in which a single bulk reaction is divided into thousands or millions of oil droplets. The ddPCR method can detect ratios of altered/normal copy number as low as 1.2 (Whale et al., 2012). This means that the number of detected cells exhibiting CNV can be relatively low relative to wild type cell expressions. This allows early detection of tumor status change. Instruments and reagents are commercially available from companies such as Fluidigm and RainDance Technologies. Several genes may be multiplexed in ddPCR and very small changes in CNV can be detected.
Specific studies for NGS are:
- Examine which NF1 tumors may or may not become malignant. That is, sequence a malignant tumor biopsy and non-malignant biopsies from the same patient.
- Study stromal tissue in the presence and absence of tumor. Is the response different for benign quiescent vs. benign growing vs. malignant tumors?
- Combine the work of Burd et al. (stromal tissue) with Yu et al. (CNV of tumor, adjacent, and blood samples) to predict rate of benign tumor growth and progression to MPNST. The goal is to use the validated blood test.
- The diagnostic test may be performed more efficiently using a targeted DNA sequencing approach that focuses only on validated genomic regions of interest. This may be the one method capable of what might be a highly multiplexed assay. The goal again is detection of tumor DNA in the peripheral blood of patients. In the case of non-metastatic tumors, detection of cell-free DNA may require such methods. Commercial kits for isolation of cell-free DNA are available.
There is no single proposed biomarker that will fulfill all desired criteria. Similarly, there may be no single assay or test platform that will be applicable for each medical question (that is, detection and rate of benign tumor growth and risk of malignancy). One must examine candidate assay targets, whether biomarkers or CNVs, to determine medical efficacy. In addition, each medical question may have its own optimal test platform that must be demonstrated. The platforms most likely to be appropriate are:
- NGS to detect a wide mutation spectrum.
- Multiplex PCR, whether qPCR or digital PCR (dPCR).
- CNV determinations using dPCR.
- Immunoassay using candidates derived from a SEREX approach or the Secretome.
- Fluorescent method such as that used for measuring membrane voltage.
Prior to final validation, one may not know if the final assay must detect DNA, RNA, proteins, or other target. Assays exist for the first four platforms listed above. Each is already commercially available and in use for infectious diseases, cancers, and genetic disorders. The development required is that of identifying and validating the specific biomarkers that are shown to correlate to medical efficacy in the Discovery phase.
An immunoassay may be used to detect some or all of the 22 genes upregulated in PNFs (Chen et al., 2012). Detection of upregulation (signal above background) from a blood sample will be more sensitive than a downregulation (signal reduction below a medically validated threshold). This also requires that the Secretome be secreted into the bloodstream and not only into tissue culture medium. The blood testing sensitivities reported by Yu et al. (2012) for prostate cancer of 81% relapse and 69% PSA doubling time predictions are good starts, but a medically efficacious assay must improve on this start. Blood testing may seek cell-free nucleic acids (DNA, miRNA), circulating tumor cells (CTCs), proteins, or other biomolecules. Kits and methods for these are comemrcially available and in current use by clinical labs.
The large number of candidates generated to date may grow once additional methods as discussed above are applied. Ideally, testing will be performed from a blood or urine sample rather than a biopsy. The latter can only test a single tumor at a time, which may not be the medically relevant tumor (i.e., the one that progresses in growth or to malignancy) in the case of NF1. Indicated high risk (of growth or malignancy) can best be pinpointed to a single tumor using MRI or other imaging method.
Assuming that no single biomarker will identify tumor progression risk, the developed assay must be a multiplex. Some platforms are especially appropriate for multiplexes. NGS, for example, specifically assumes the sequencing and hence detection of a large number of different DNA fragments. Limiting its detection by targeted sequencing will allow detection and quantitation of each sequence variation present within a sample. Testing for CNV by microarrays likewise assumes the presence and detection of a large number of genomic analytes. A smaller number of multiplexed analytes may be detected in one reaction by the various types of PCR (digital or otherwise). Immunoassays can also be multiplexed for simultaneous detection of a few analytes.
What are the issues of multiplexing? The sensitivity and specificity of each analyte must be characterized in the presence of other analytes. A method such as NGS simply reads out each DNA sequence, providing the user with thousands of individually obtained sequences. Software then determines how often a particular DNA target is sequenced, and how many of those sequences contain mutations. Other methods such as multiplex single-tube PCR may provide a complex result as the reagents interact. These problems plagued researchers years ago, but highly multiplexed assays, including microarrays, are already in the clinical lab. Multiple fluorescent probes and intensities allow for testing several targets simultaneously.
Most likely, no diagnostic company will submit an NF1 assay to the FDA. The test resulting from this project will probably be a lab-developed test (LDT) and thus requires a flexible platform. At this time, there are numerous companies that sell the instruments for developing these tests. Alternatively, there are CLIA-certified labs such as Ambry Genetics that perform services such as NGS. One will find that tools developed for other cancers and diseases are available for adaptation to NF1.
Beert E, Brems H, Renard M, Ferreiro JF, Melotte C, Thoelen R, De Wever I, Sciot R, Legius E, Debiec-Rychter M. “Biallelic inactivation of NF1 in a sporadic plexiform neurofibroma.” Genes Chromosomes Cancer. 2012 Sep;51(9):852-857.
Beert E, Brems H, Daniëls B, De Wever I, Van Calenbergh F, Schoenaers J, Debiec-Rychter M, Gevaert O, De Raedt T, Van Den Bruel A, de Ravel T, Cichowski K, Kluwe L, Mautner V, Sciot R, Legius E. “Atypical neurofibromas in neurofibromatosis type 1 are premalignant tumors.” Genes Chromosomes Cancer. 2011 Dec;50(12):1021-1032. doi: 10.1002/gcc.20921. Epub 2011 Aug 24.
Burd et al., Monitoring Tumorigenesis and Senescence In Vivo with a p16INK4a-Luciferase Model, Cell (2013), http://dx.doi.org/10.1016/j.cell.2012.12.010.
Cappuzzo F, Hirsch FR, Rossi E, Bartolini S, Ceresoli GL, Bemis L, Haney J, Witta S, Danenberg K, Domenichini I, Ludovini V, Magrini E, Gregorc V, Doglioni C, Sidoni A, Tonato M, Franklin WA, Crino L, Bunn PA Jr, Varella-Garcia M. “Epidermal growth factor receptor gene and protein and gefitinib sensitivity in non-small-cell lung cancer.” J Natl Cancer Inst. 2005 May 4;97(9):643-655.
Chen HL, Seol H, Brown KJ, Gordish-Dressman H, Hill A, Gallo V, Packer R, Hathout Y. “Secretome Survey of Human Plexiform Neurofibroma Derived Schwann Cells Reveals a Secreted form of the RARRES1 Protein.” Int J Mol Sci. 2012;13(7):9380-9399.
Chen YT. Identification of human tumor antigens by serological expression cloning: an online review on SEREX. Cancer Immun 2004 [updated 2004 Mar 10; cited 2004 Apr 1]. URL: http://www.cancerimmunity.org/serex
Chernet BT, Levin M. “Transmembrane voltage potential is an essential cellular parameter for the detection and control of tumor development in a Xenopus model.” Disease Models & Mechanisms 6, (2013) doi:10.1242/dmm.010835 (preprint of 8 Feb issue).
Daniels M, Goh F, Wright CM, Sriram KB, Relan V, Clarke BE, Duhig EE, Bowman RV, Yang IA, Fong KM. “Whole genome sequencing for lung cancer.” J Thorac Dis. 2012 Apr 1;4(2):155-163.
Hanafusa T, Mohamed AE, Domae S, Nakayama E, Ono T. “Serological identification of Tektin5 as a cancer/testis antigen and its immunogenicity.” BMC Cancer. 2012 Nov 14;12:520-527.
Kazmi SJ, Byer SJ, Eckert JM, Turk AN, Huijbregts RP, Brossier NM, Grizzle WE, Mikhail FM, Roth KA, Carroll SL. “Transgenic Mice Overexpressing Neuregulin-1 Model Neurofibroma-Malignant Peripheral Nerve Sheath Tumor Progression and Implicate Specific Chromosomal Copy Number Variations in Tumorigenesis.” Am J Pathol. 2013, published online Jan 14.
Link DC, Schuettpelz LG, Shen D, Wang J, Walter MJ, Kulkarni S, et al. “Identification of a novel TP53 cancer susceptibility mutation through whole-genome sequencing of a patient with therapy-related AML.” JAMA 2011;305:1568-1576.
Liu Y, Cope L, Sun W, Wang Y, Prasad N, Sangenario L, Talbot K, Somervell H, Westra W, Bishop J, Califano J, Zeiger M, Umbricht C. “DNA Copy Number Variations Characterize Benign and Malignant Thyroid Tumors.” J Clin Endocrinol Metab. 2013 Jan 23.
Miller SJ, Rangwala F, Williams J, Ackerman P, Kong S, Jegga AG, Kaiser S, Aronow BJ, Frahm S, Kluwe L, Mautner V, Upadhyaya M, Muir D, Wallace M, Hagen J, Quelle DE, Watson MA, Perry A, Gutmann DH, Ratner N. “Large-scale molecular comparison of human schwann cells to malignant peripheral nerve sheath tumor cell lines and tissues.” Cancer Res. 2006 Mar 1;66(5):2584-2591.
Mußotter T, Kluwe L, Högel J, Nguyen R, Cooper DN, Mautner VF, Kehrer-Sawatzki H. “Non-coding RNA ANRIL and the number of plexiform neurofibromas in patients with NF1 microdeletions.” BMC Med Genet. 2012 Oct 26;13:98-104.
Park HJ, Lee SJ, Sohn YB, Jin HS, Han JH, Kim YB, Yim H, Jeong SY. “NF1 deficiency causes Bcl-xL upregulation in Schwann cells derived from neurofibromatosis type 1-associated malignant peripheral nerve sheath tumors.” Int J Oncol. 2013 Feb;42(2):657-666.
Pasmant et al. 2011a. Role of Noncoding RNA ANRIL in Genesis of Plexiform Neurofibromas in Neurofibromatosis Type 1. J Natl Cancer Inst 2011;103:1713–1722.
Pasmant et al. 2011b. Identification of Genes Potentially Involved in the Increased Risk of Malignancy in NF1-Microdeleted Patients. Mol Med. 2011 Jan-Feb;17(1-2):79-87.
Stewart DR, Pemov A, Van Loo P, Beert E, Brems H, Sciot R, Claes K, Pak E, Dutra A, Lee CC, Legius E. “Mitotic recombination of chromosome arm 17q as a cause of loss of heterozygosity of NF1 in neurofibromatosis type 1-associated glomus tumors.” Genes Chromosomes Cancer. 2012 May;51(5):429-437. doi: 10.1002/gcc.21928. Epub 2012 Jan 17.
Storlazzi CT, Brekke HR, Mandahl N, Brosjö O, Smeland S, Lothe RA, Mertens F. “Identification of a novel amplicon at distal 17q containing the BIRC5/SURVIVIN gene in malignant peripheral nerve sheath tumours.” J Pathol 2006, 209:492-500
Terribas E, Garcia-Linares C, Lázaro C, Serra E. “Probe-Based Quantitative PCR Assay for Detecting Constitutional and Somatic Deletions in the NF1 Gene: Application to Genetic Testing and Tumor Analysis.” Clin Chem. 2013 Feb 5. [Epub ahead of print]
Thomas L, Spurlock G, Eudall C, Thomas NS, Mort M, Hamby SE, Chuzhanova N, Brems H, Legius E, Cooper DN, Upadhyaya M. “Exploring the somatic NF1 mutational spectrum associated with NF1 cutaneous neurofibromas.” Eur J Hum Genet. 2012 Apr;20(4):411-419. doi: 10.1038/ejhg.2011.207. Epub 2011 Nov 23.
Whale AS, Huggett JF, Cowen S, Speirs V, Shaw J, Ellison S, Foy CA, Scott DJ. “Comparison of microfluidic digital PCR and conventional quantitative PCR for measuring copy number variation.” Nucleic Acids Res. 2012 Jun;40(11):e82. doi: 10.1093/nar/gks203. Epub 2012 Feb 28.
Yu YP, Song C, Tseng G, Ren BG, LaFramboise W, Michalopoulos G, Nelson J, Luo JH. “Genome abnormalities precede prostate cancer and predict clinical relapse.” Am J Pathol. 2012 Jun;180(6):2240-2248.
 They list 162 genes with consistently differential results, plus 10 additional genes with conflicting results, in their Supplemental Table 2.
 Patients undergo MR perfusion scan with gadopentetate dimeglumine and fludeoxyglucose F 18 positron emission tomography (FDG-PET); NCT00060008. A new MRI study was registered in January, 2013.