Patent Application
20160281175
The present invention relates to cancer diagnostic methods and means therefor.
Neoplasms and cancer are abnormal growths of cells. Cancer cells rapidly reproduce despite restriction of space, nutrients shared by other cells, or signals sent from the body to stop re-production. Cancer cells are often shaped differently from healthy cells, do not function properly, and can spread into many areas of the body. Abnormal growths of tissue, called tumors, are clusters of cells that are capable of growing and di-viding uncontrollably. Tumors can be benign (noncancerous) or malignant (cancerous). Benign tumors tend to grow slowly and do not spread. Malignant tumors can grow rapidly, invade and destroy nearby normal tissues, and spread throughout the body. Malignant cancers can be both locally invasive and metastatic. Locally invasive cancers can invade the tissues surrounding it by sending out “fingers” of cancerous cells into the normal tissue. Metastatic cancers can send cells into other tissues in the body, which may be distant from the original tumor. Cancers are classified according to the kind of fluid or tissue from which they originate, or according to the location in the body where they first developed. All of these parameters can effectively have an influence on the cancer characteristics, development and progression and subsequently also cancer treatment. Therefore, reliable methods to classify a cancer state or cancer type, taking diverse parameters into consideration is desired. Since cancer is predominantly a genetic disease, trying to classify cancers by genetic parameters is one extensively studied route.
Extensive efforts have been undertaken to discover genes relevant for diagnosis, prognosis and management of (cancerous)disease. Mainly RNA-expression studies have been used for screening to identify genetic biomarkers. Over recent years it has been shown that changes in the DNA-methylation pattern of genes could be used as biomarkers for cancer diagnostics. In concordance with the general strategy identifying RNA-expression based biomarkers, the most convenient and prospering approach would start to identify marker candidates by genome-wide screening of methylation changes.
The most versatile genome-wide approaches up to now are using microarray hybridization based techniques. Although studies have been undertaken at the genomic level (and also the single-gene level) for elucidating methylation changes in diseased versus normal tissue, a comprehensive test obtaining a good success rate for identifying biomarkers is yet not available.
Developing biomarkers for disease (especially cancer)-screening, -diagnosis, and -treatment was improved over the last decade by major advances of different technologies which have made it easier to discover potential biomarkers through high-throughput screens. Comparing the so called “OMICs”-approaches like Genomics, Proteomics, Metabolomics, and derivates from those, Genomics is best developed and most widely used for biomarker identification. Because of the dynamic nature of RNA expression and the ease of nucleic acid extraction and the detailed knowledge of the human genome, many studies have used RNA expression profiling for elucidation of class differences for distinguishing the “good” from the “bad” situation like diseased vs. healthy, or clinical differences between groups of diseased patients. Over the years especially microarray-based expression profiling has become a standard tool for research and some approaches are currently under clinical validation for diagnostics. The plasticity over a broad dynamic range of RNA expression levels is an advantage using RNA and also a prerequisite of successful discrimination of classes, the low stability of RNA itself is often seen as a drawback. Because stability of DNA is tremendously higher than stability of RNA, DNA based markers are more promising markers and expected to give robust assays for diagnostics. Many of clinical markers in oncology are more or less DNA based and are well established, e.g. cytogenetic analyses for diagnosis and classification of different tumor-species. However, most of these markers are not accessible using the cheap and efficient molecular-genetic PCR routine tests. This might be due to 1) the structural complexity of changes, 2) the inter-individual differences of these changes at the DNA-sequence level, and 3) the relatively low “quantitative” fold-changes of those “chromosomal” DNA changes. In comparison, RNA-expression changes range over some orders of magnitudes and these changes can be easily measured using genome-wide expression microarrays. These expression arrays are covering the entire translated transcriptome by 20000-45000 probes. Elucidation of DNA changes via microarray techniques re-quires in general more probes depending on the requested resolution. Even order(s) of magnitude more probes are required than for standard expression profiling to cover the entire 3×109 by human genome. For obtaining best resolution when screening biomarkers at the structural genomic DNA level, today genomic tiling arrays and SNP-arrays are available. Although costs of these techniques analysing DNA have decreased over recent years, for biomarker screening many samples have to be tested, and thus these tests are cost intensive.
Another option for obtaining stable DNA-based biomarkers re-lies on elucidation of the changes in the DNA methylation pattern of (malignant; neoplastic) disease. In the vertebrate genome methylation affects exclusively the cytosine residues of CpG dinucleotides, which are clustered in CpG islands. CpG islands are often found associated with gene-promoter sequences, present in the 5′-untranslated gene regions and are per default unmethylated. In a very simplified view, an unmethylated CpG island in the associated gene-promoter enables active transcription, but if methylated gene transcription is blocked. The DNA methylation pattern is tissue- and clone-specific and almost as stable as the DNA itself. It is also known that DNA-methylation is an early event in tumorigenesis which would be of interest for early and initial diagnosis of disease. In principle screening for biomarkers suitable to answering clinical questions including DNA-methylation based approaches would be most successful when starting with a genome-wide approach.
Shames D et al. (PLOS Medicine 3(12) (2006): 2244-2262) identified multiple genes that are methylated with high penetrance in primary lung, breast, colon and prostate cancers.
Sato N et al. (Cancer Res 63(13) (2003): 3735-3742) identified potential targets with aberrant methylation in pancreatic cancer. These genes were tested using a treatment with a de-methylating agent (5-aza-2′-deoxycytidine and/or the histone deacetylase inhibitor trichostatin A) after which certain genes were increased transcribed.
Bibikova M et al. (Genome Res 16(3) (2006): 383-393) analysed lung cancer biopsy samples to identify methylated cpu sites to distinguish lung adenocarcinomas from normal lung tissues.
Yan P S et al. (Clin Cancer Res 6(4) (2000): 1432-1438) analysed CpG island hypermethylation in primary breast tumor.
Cheng Y et al. (Genome Res 16(2) (2006): 282-289) discussed DNA methylation in CpG islands associated with transcriptional silencing of tumor suppressor genes.
Ongenaert M et al. (Nucleic Acids Res 36 (2008) Database issue D842-D846) provided an overview over the methylation database “PubMeth”.
Microarray for human genome-wide hybridization testings are known, e.g. the Affymetrix Human Genome U133A Array (NCB1 Database, Acc. No. GLP96).
A substantial number of differentially methylated genes has been discovered over years rather by chance than by rationality. Albeit some of these methylation changes have the potential being useful markers for differentiation of specifically defined diagnostic questions, these would lack the power for successful delineation of various diagnostic constellations. Thus, the rational approach would start at the genomic-screen for distinguishing the “subtypes” and diagnostically, prognostically and even therapeutically challenging constellations. These rational expectations are the base of starting genomic (and also other—omics) screenings but do not warrant to obtain the maker panel for all clinical relevant constellations which should be distinguished. This is neither unreliable when thinking about a universal approach (e.g. transcriptomics) suitable to distinguish for instance all subtypes in all different malignancies by focusing on a single class of target-molecules (e.g. RNA). Rather all omics-approaches together would be necessary and could help to improve diagnostics and finally patient management.
Lung cancer is the third most common malignant neoplasm in the EU following breast and colon cancers. Lung cancer presents the second worst 5-year survival figures following pancreas. Thus, although it accounts for 14% of all cancer diagnoses, lung cancer is responsible for 22% of cancer deaths, indicating the poor prognosis of this tumour type and the comparative lack of progress in treatment. Therapy is hampered by the tendency for lung cancer to be diagnosed at a late stage, hence the need to develop markers for early detection. Approximately 80% of lung cancer cases are of the non-small cell type (NSCLC), with squamous cell carcinoma and adenocarcinoma being the most frequent subtypes. A goal of the present invention is to provide an alternative and more cost-efficient route to identify suitable markers for lung cancer diagnostics.
Therefore, in a first aspect, the present invention provides a set of nucleic acid primers or hybridization probes being specific for a potentially methylated region of marker genes being suitable to diagnose or predict lung cancer or a lung cancer type, preferably being selected from adenocarcinoma or squamous cell carcinoma, the marker genes comprising WT1, SALL3, TERT, ACTB, CPEB4. Preferably the set further comprises any one of the markers ABCB1, ACTB, AIM1L, APC, AREG, BMP2K, BOLL, C5AR1, C5orf4, CADM1, CDH13, CDX1, CLIC4, COL21A1, CPEB4, CXADR, DLX2, DNAJA4, DPH1, DRD2, EFS, ERBB2, ERCC1, ESR2, F2R, FAM43A, GABRA2, GAD1, GBP2, GDNF, GNA15, GNAS, HECW2, HIC1, HIST1H2AG, HLA-G, HOXA1, HOXA10, HSD17B4, HSPA2, IRAK2, ITGA4, JUB, KCNJ15, KCNQ1, KIF5B, KL, KRT14, KRT17, LAMC2, MAGEB2, MBD2, MSH4, MT1G, MT3, MTHFR, NEUROD1, NHLH2, NKX2-1, ONECUT2, PENK, PITX2, PLAGL1, PTTG1, PYCARD, RASSF1, S100A8, SALL3, SERPINB5, SERPINE1, SERPINI1, SFRP2, SLC25A31, SMAD3, SPARC, SPHK1, SRGN, TERT, THRB, TJP2, TMEFF2, TNFRSF10C, TNFRSF25, TP53, ZDHHC11, ZNF256, ZNF711, F2R, HOXA10, KL, SALL3, SPARC, TNFRSF25, WT1.
In a further aspect, the present invention provides a method of determining a subset of diagnostic markers for potentially methylated genes from the genes of gene marker IDs 1-359 of table 1, suitable for the diagnosis or prognosis of lung cancer or lung cancer type, comprising
The present invention provides a master set of 359 genetic markers which has been surprisingly found to be highly relevant for aberrant methylation in the diagnosis or prognosis of lung cancer. It is possible to determine a multitude of marker subsets from this master set which can be used to diagnose and differentiate between various lung cancer or tumor types, e.g. adenocarcinoma and squamous cell carcinoma.
The inventive 359 marker genes of table 1 (given in example 1 below) are: NHLH2, MTHFR, PRDM2, MLLT11, S100A9 (control), S100A9, S100A8 (control), S100A8, S100A2, LMNA, DUSP23, LAMC2, PTGS2, MARK1, DUSP10, PARP1, PSEN2, CLIC4, RUNX3, AIM1L, SFN, RPA2, TP73, TP73 (p73), POU3F1, MUTYH, UQCRH, FAF1, TACSTD2, TN-FR5F25, DIRAS3, MSH4, GBP2, GBP2, LRRC8C, F3, NANOS1, MGMT, EBF3, DCLRE1C, KIF5B, ZNF22, PGBD3, SRGN, GATA3, PTEN, MMS19, SFRP5, PGR, ATM, DRD2, CADM1, TEAD1, OPCML, CALCA, CTSD, MYOD1, IGF2, BDNF, CDKN1C, WT1, HRAS, DDB1, GSTP1, CCND1, EPS8L2, PI-WIL4, CHST11, UNG, CCDC62, CDK2AP1, CHFR, GRIN2B, CCND2, VDR, B4GALNT3, NTF3, CYP27B1, GPR92, ERCC5, GJB2, BRCA2, KL, CCNA1, SMAD9, C13orf15, DGKH, DNAJC15, RB1, RCBTB2, PARP2, APEX1, JUB, JUB (control NM 198086), EFS, BAZ1A, NKX2-1, ESR2, HSPA2, PSEN1, PGF, MLH3, TSHR, THBS1, MYO5C, SMAD6, SMAD3, NOX5, DNAJA4, CRABP1, BCL2A1 (ID NO: 111), BCL2A1 (ID NO: 112), BNC1, ARRDC4, SOCS1, ERCC4, NTHL1, PYCARD, AXIN1, CYLD, MT3, MT1A, MT1G, CDH1, CDH13, DPH1, HIC1, NEUROD2 (control), NEUROD2, ERBB2, KRT19, KRT14, KRT17, JUP, BRCA1, COL1A1, CACNA1G, PRKAR1A, SPHK1, SOX15, TP53 (TP53_CGI23_1 kb), TP53 (TP53_both_CGIs_1 kb), TP53 (TP53_CGI36_1 kb), TP53, NPTX1, SMAD2, DCC, MBD2, ONECUT2, BCL2, SERPINB5, SERPINB2 (control), SERPINB2, TYMS, LAMA1, SALL3, LDLR, STK11, PRDX2, RAD23A, GNA15, ZNF573, SPINT2, XRCC1, ERCC2, ERCC1, C5AR1 (NM_001736), C5AR1, POLD1, ZNF350, ZNF256, C3, XAB2, ZNF559, FHL2, IL1B, IL1B (control), PAX8, DDX18, GAD1, DLX2, ITGA4, NEUROD1, STAT1, TMEFF2, HECW2, BOLL, CASP8, SERPINE2, NCL, CYP1B1, TACSTD1, MSH2, MSH6, MXD1, JAG1, FOXA2, THBD, CTCFL, CTSZ, GATA5, CXADR, APP, TTC3, KCNJ15, RIPK4, TFF1, SEZ6L, TIMP3, BIK, VHL, IRAK2, PPARG, MBD4, RBP1, XPC, ATR, LXN, RARRES1, SERPINI1, CLDN1, FAM43A, IQCG, THRB, RARB, TGFBR2, MLH1, DLEC1, CTNNB1, ZNF502, SLC6A20, GPX1, RASSF1, FHIT, OGG1, PITX2, SLC25A31, FBXW7, SFRP2, CHRNA9, GABRA2, MSX1, IGFBP7, EREG, AREG, ANXA3, BMP2K, APC, HSD17B4 (ID No 249), HSD17B4 (ID No 250), LOX, TERT, NEUROG1, NR3C1, ADRB2, CDX1, SPARC, C5orf4, PTTG1, DUSP1, CPEB4, SCGB3A1, GDNF, ERCC8, F2R, F2RL1, VCAN, ZDHHC11, RHOBTB3, PLAGL1, SASH1, ULBP2, ESR1, RNASET2, DLL1, HIST1H2AG, HLA-G, MSH5, CDKN1A, TDRD6, COL21A1, DSP, SERPINE1 (ID No 283), SERPINE1 (ID No 284), FBXL13, NRCAM, TWIST1, HOXA1, HOXA10, SFRP4, IGFBP3, RPA3, ABCB1, TFPI2, COL1A2, ARPC1B, PILRB, GATA4, MAL2, DLC1, EPPK1, LZTS1, TNFRSF10B, TNFRSF10C, TNFRSF10D, TNFRSF10A, WRN, SFRP1, SNAI2, RDHE2, PENK, RDH10, TGFBR1, ZNF462, KLF4, CDKN2A, CDKN2B, AQP3, TPM2, TJP2 (ID NO 320), TJP2 (ID No 321), PSAT1, DAPK1, SYK, XPA, ARMCX2, RHOXF1, FHL1, MAGEB2, TIMP1, AR, ZNF711, CD24, ABL1, ACTB, APC, CDH1 (Ecad 1), CDH1 (Ecad2), FMR1, GNAS, H19, HIC1, IGF2, KCNQ1, GNAS, CDKN2A (P14), CDKN2B (P15), CDKN2A (P16_VL), PITXA, PITXB, PITXC, PITXD, RB1, SFRP2, SNRPN, XIST, IRF4, UNC13B, GSTP1. Table 1 lists some marker genes in the double such as for different loci and control sequences. It should be understood that any methylation specific region which is readily known to the skilled man in the art from prior publications or available databases (e.g. PubMeth at www.pubmeth.org) can be used according to the present invention. Of course, double listed genes only need to be represented once in an inventive marker set (or set of probes or primers therefor) but preferably a second marker, such as a control region is included (IDs given in the list above relate to the gene ID (or gene loci ID) given in table 1 of the example section).
One advantage making DNA methylation an attractive target for biomarker development, is the fact that cell free methylated DNA can be detected in body-fluids like serum, sputum, and urine from patients with cancerous neoplastic conditions and disease. For the purpose of biomarker screening, clinical samples have to be available. For obtaining a sufficient number of samples with clinical and “outcome” or survival data, the first step would be using archived (tissue) samples. Preferably these materials should fulfill the requirements to obtain intact RNA and DNA, but most archives of clinical samples are storing formalin fixed paraffin embedded (FFPE) tissue blocks. This has been the clinic-pathological routine done over decades, but that fixed samples are if at all only suitable for extraction of low quality of RNA. It has now been found that according to the present invention any such samples (as any comprising tumor DNA) can be used for the method of generating an inventive subset, including fixed samples. The samples can be of lung tissue or any body fluid, e.g. sputum, bronchial lavage, or serum derived from peripheral blood or blood cells. Blood or blood derived samples preferably have reduced, e.g. <95%, or no leukocyte content but comprise DNA of the cancerous cells or tumor. Preferably the inventive markers are of human genes. Preferably the samples are human samples.
The present invention provides a multiplexed methylation testing method which 1) outperforms the “classification” success when compared to genomewide screenings via RNA-expression profiling, 2) enables identification of biomarkers for a wide variety of diseases, without the need to prescreen candidate markers on a genomewide scale, and 3) is suitable for minimal invasive testing and 4) is easily scalable.
In contrast to the rational strategy for elucidation of biomarkers for differentiation of disease, the invention presents a targeted multiplexed DNA-methylation test which outperforms genome-scaled approaches (including RNA expression profiling) for disease diagnosis, classification, and prognosis.
The inventive set of 359 markers enables selection of a subset of markers from this 359 set which is highly characteristic of lung cancer and a given lung cancer type. Further indicators differentiating between cancer types or generally neoplastic conditions are e.g. benign (non (or limited) proliferative) or malignant, metastatic or non-metastatic tumors or nodules. It is sometimes possible to differentiate the sample type from which the methylated DNA is isolated, e.g. urine, blood, tissue samples.
The present invention is suitable to differentiate diseases, in particular neoplastic conditions, or tumor types. Diseases and neoplastic conditions should be understood in general including benign and malignant conditions. According to the present invention benign nodules (being at least the potential onset of malignancy) are included in the definition of a disease. After the development of a malignancy the condition is a preferred disease to be diagnosed by the markers screened for or used according to the present invention. The present invention is suitable to distinguish benign and malignant tumors (both being considered a disease according to the present invention). In particular the invention can provide markers (and their diagnostic or prognostic use) distinguishing between a normal healthy state together with a benign state on one hand and malignant states on the other hand. A diagnosis of lung cancer may include identifying the difference to a normal healthy state, e.g. the absence of any neoplastic nodules or cancerous cells. The present invention can also be used for prognosis of lung cancer, in particular a prediction of the progression of lung cancer or lung cancer type. A particularly preferred use of the invention is to perform a diagnosis or prognosis of metastasizing lung cancer (distinguished from non-metastasizing conditions).
In the context of the present invention “prognosis”, “prediction” or “predicting” should not be understood in an absolute sense, as in a certainty that an individual will develop lung cancer or lung cancer type (including cancer progression), but as an increased risk to develop cancer or the lung cancer type or of cancer progression. “Prognosis” is also used in the context of predicting disease progression, in particular to predict therapeutic results of a certain therapy of the disease, in particular neoplastic conditions, or lung cancer types. The prognosis of a therapy can e.g. be used to predict a chance of success (i.e. curing a disease) or chance of reducing the severity of the disease to a certain level. As a general inventive concept, markers screened for this purpose are preferably derived from sample data of patients treated according to the therapy to be predicted. The inventive marker sets may also be used to monitor a patient for the emergence of therapeutic results or positive disease progressions.
Some of the inventive, rationally selected markers have been found methylated in some instances. DNA methylation analyses in principle rely either on bisulfite deamination-based methylation detection or on using methylation sensitive restriction enzymes. Preferably the restriction enzyme-based strategy is used for elucidation of DNA-methylation changes. Further methods to determine methylated DNA are e.g. given in EP 1 369 493 A1 or U.S. Pat. No. 6,605,432. Combining restriction digestion and multiplex PCR amplification with a targeted microarray-hybridization is a particular advantageous strategy to perform the inventive methylation test using the inventive marker sets (or subsets). A microarray-hybridization step can be used for reading out the PCR results. For the analysis of the hybridization data statistical approaches for class comparisons and class prediction can be used. Such statistical methods are known from analysis of RNA-expression derived microarray data.
If only limiting amounts of DNA were available for analyses an amplification protocol can be used enabling selective amplification of the methylated DNA fraction prior methylation testing. Subjecting these amplicons to the methylation test, it was possible to successfully distinguish DNA from sensitive cases from normal healthy controls. In addition it was possible to distinguish lung-cancer patients from healthy normal controls using DNA from serum by the inventive methylation test upon preamplification. Both examples clearly illustrate that the inventive multiplexed methylation testing can be successfully applied when only limiting amounts of DNA are available. Thus, this principle might be the preferred method for minimal invasive diagnostic testing.
In most situations several genes are necessary for classification. Although the 359 marker set test is not a genome-wide test and might be used as it is for diagnostic testing, running a subset of markers—comprising the classifier which enables best classification—would be easier for routine applications. The test is easily scalable. Thus, to test only the subset of markers, comprising the classifier, the selected subset of primers/probes could be applied directly to set up of the lower multiplexed test (or single PCR-test). Serum DNA can be used to classify or distinguish healthy patients from individuals with lung-tumors. Only the specific primers comprising the gene-classifier obtained from the methylation test may be set up together in multiplexed PCR reactions.
In summary the inventive methylation test is a suitable tool for differentiation and classification of neoplastic disease. This assay can be used for diagnostic purposes and for defining biomarkers for clinical relevant issues to improve diagnosis of disease, and to classify patients at risk for disease progression, thereby improving disease treatment and patient management.
The first step of the inventive method of generating a subset, step a) of obtaining data of the methylation status, preferably comprises determining data of the methylation status, preferably by methylation specific PCR analysis, methylation specific digestion analysis. Methylation specific digestion analysis can include either or both of hybridization of suitable probes for detection to non-digested fragments or PCR amplification and detection of non-digested fragments.
The inventive selection can be made by any (known) classification method to obtain a set of markers with the given diagnostic (or also prognostic) value to categorize a lung cancer or lung cancer type. Such methods include class comparisons wherein a specific p-value is selected, e.g. a p-value below 0.1, preferably below 0.08, more preferred below 0.06, in particular preferred below 0.05, below 0.04, below 0.02, most preferred below 0.01.
Preferably the correlated results for each gene b) are rated by their correct correlation to lung cancer or lung cancer type positive state, preferably by p-value test or t-value test or F-test. Rated (best first, i.e. low p- or t-value) markers are the subsequently selected and added to the subset until a certain diagnostic value is reached, e.g. the herein mentioned at least 70% (or more) correct classification of lung cancer or lung cancer type.
Class comparison procedures include identification of genes that were differentially methylated among the two classes using a random-variance t-test. The random-variance t-test is an improvement over the standard separate t-test as it permits sharing information among genes about within-class variation without assuming that all genes have the same variance (Wright G. W. and Simon R, Bioinformatics 19:2448-2455, 2003). Genes were considered statistically significant if their p value was less than a certain value, e.g. 0.1 or 0.01. A stringent significance threshold can be used to limit the number of false positive findings. A global test can also be performed to determine whether the expression profiles differed between the classes by permuting the labels of which arrays corresponded to which classes. For each permutation, the p-values can be re-computed and the number of genes significant at the e.g. 0.01 level can be noted. The proportion of the permutations that give at least as many significant genes as with the actual data is then the significance level of the global test. If there are more than 2 classes, then the “F-test” instead of the “t-test” should be used.
Class Prediction includes the step of specifying a significance level to be used for determining the genes that will be included in the subset. Genes that are differentially methylated between the classes at a univariate parametric significance level less than the specified threshold are included in the set. It doesn't matter whether the specified significance level is small enough to exclude enough false discoveries. In some problems better prediction can be achieved by being more liberal about the gene sets used as features. The sets may be more bio-logically interpretable and clinically applicable, however, if fewer genes are included. Similar to cross-validation, gene selection is repeated for each training set created in the cross-validation process. That is for the purpose of providing an unbiased estimate of prediction error. The final model and gene set for use with future data is the one resulting from application of the gene selection and classifier fitting to the full dataset.
Models for utilizing gene methylation profile to predict the class of future samples can also be used. These models may be based on the Compound Covariate Predictor (Radmacher et al. Journal of Computational Biology 9:505-511, 2002), Diagonal Linear Discriminant Analysis (Dudoit et al. Journal of the American Statistical Association 97:77-87, 2002), Nearest Neighbor Classification (also Dudoit et al.), and Support Vector Machines with linear kernel (Ramaswamy et al. PNAS USA 98:15149-54, 2001). The models incorporated genes that were differentially methylated among genes at a given significance level (e.g. 0.01, 0.05 or 0.1) as assessed by the random variance t-test (Wright G. W. and Simon R. Bioinformatics 19:2448-2455, 2003). The prediction error of each model using cross validation, preferably leave-one-out cross-validation (Simon et al. Journal of the National Cancer Institute 95:14-18, 2003), is preferably estimated. For each leave-one-out cross-validation training set, the entire model building process was repeated, including the gene selection process. It may also be evaluated whether the cross-validated error rate estimate for a model was significantly less than one would expect from random prediction. The class labels can be randomly permuted and the entire leave-one-out cross-validation process is then repeated. The significance level is the proportion of the random permutations that gave a cross-validated error rate no greater than the cross-validated error rate obtained with the real methylation data. About 1000 random permutations may be usually used.
Another classification method is the greedy-pairs method described by Bo and Jonassen (Genome Biology 3(4):research0017.1-0017.11, 2002). The greedy-pairs approach starts with ranking all genes based on their individual t-scores on the training set. The procedure selects the best ranked gene gi and finds the one other gene gi that together with provides the best discrimination using as a measure the distance between centroids of the two classes with regard to the two genes when projected to the diagonal linear discriminant axis. These two selected genes are then removed from the gene set and the procedure is repeated on the remaining set until the specified number of genes have been selected. This method attempts to select pairs of genes that work well together to discriminate the classes.
Furthermore, a binary tree classifier for utilizing gene methylation profile can be used to predict the class of future samples. The first node of the tree incorporated a binary classifier that distinguished two subsets of the total set of classes. The individual binary classifiers were based on the “Support Vector Machines” incorporating genes that were differentially expressed among genes at the significance level (e.g. 0.01, 0.05 or 0.1) as assessed by the random variance t-test (Wright G. W. and Simon R. Bioinformatics 19:2448-2455, 2003). Classifiers for all possible binary partitions are evaluated and the partition selected was that for which the cross-validated prediction error was minimum. The process is then repeated successively for the two subsets of classes determined by the previous binary split. The prediction error of the binary tree classifier can be estimated by cross-validating the entire tree building process. This overall cross-validation included re-selection of the optimal partitions at each node and re-selection of the genes used for each cross-validated training set as described by Simon et al. (Simon et al. Journal of the National Cancer Institute 95:14-18, 2003). 10-fold cross validation in which one-tenth of the samples is withheld can be utilized, a binary tree developed on the remaining 9/10 of the samples, and then class membership is predicted for the 10% of the samples withheld. This is repeated 10 times, each time withholding a different 10% of the samples. The samples are randomly partitioned into 10 test sets (Simon R and Lam A. BRB-ArrayTools User Guide, version 3.2. Biometric Research Branch, National Cancer Institute).
Preferably the correlated results for each gene b) are rated by their correct correlation to lung cancer or lung cancer type positive state, preferably by p-value test. It is also possible to include a step in that the genes are selected d) in order of their rating.
Independent from the method that is finally used to produce a subset with certain diagnostic or predictive value, the subset selection preferably results in a subset with at least 60%, preferably at least 65%, at least 70%, at least 75%, at least 80% or even at least 85%, at least 90%, at least 92%, at least 95%, in particular preferred 100% correct classification of test samples of lung cancer or lung cancer type. Such levels can be reached by repeating c) steps a) and b) of the inventive method, if necessary.
To prevent increase of the number of the members of the subset, only marker genes with at least a significance value of at most 0.1, preferably at most 0.8, even more preferred at most 0.6, at most 0.5, at most 0.4, at most 0.2, or more preferred at most 0.01 are selected.
In particular preferred embodiments the at least 50 genes of step a) are at least 70, preferably at least 90, at least 100, at least 120, at least 140, at least 160, at least 180, at least 190, at least 200, at least 220, at least 240, at least 260, at least 280, at least 300, at least 320, at least 340, at least 350 or all, genes.
Since the subset should be small it is preferred that not more than 60, or not more than 40, preferably not more than 30, in particular preferred not more than 20, marker genes are selected in step d) for the subset.
In a further aspect the present invention provides a method of identifying lung cancer or lung cancer type in a sample comprising DNA from a patient, comprising providing a diagnostic subset of markers identified according to the method depicted above, determining the methylation status of the genes of the subset in the sample and comparing the methylation status with the status of a confirmed lung cancer or lung cancer type positive and/or negative state, thereby identifying lung cancer or lung cancer type in the sample.
The methylation status can be determined by any method known in the art including methylation dependent bisulfite deamination (and consequently the identification of mC—methylated C—changes by any known methods, including PCR and hybridization techniques). Preferably, the methylation status is determined by methylation specific PCR analysis, methylation specific digestion analysis and either or both of hybridisation analysis to non-digested or digested fragments or PCR amplification analysis of non-digested fragments. The methylation status can also be determined by any probes suitable for determining the methylation status including DNA, RNA, PNA, LNA probes which optionally may further include methylation specific moieties.
As further explained below the methylation status can be particularly determined by using hybridisation probes or amplification primer (preferably PCR primers) specific for methylated regions of the inventive marker genes. Discrimination between methylated and non-methylated genes, including the determination of the methylation amount or ratio, can be performed by using e.g. either one of these tools.
The determination using only specific primers aims at specifically amplifying methylated (or in the alternative non-methylated) DNA. This can be facilitated by using (methylation dependent) bisulfite deamination, methylation specific enzymes or by using methylation specific nucleases to digest methylated (or alternatively non-methylated) regions—and consequently only the non-methylated (or alternatively methylated) DNA is obtained. By using a genome chip (or simply a gene chip including hybridization probes for all genes of interest such as all 359 marker genes), all amplification or non-digested products are detected. I.e. discrimination between methylated and non-methylated states as well as gene selection (the inventive set or subset) is before the step of detection on a chip.
Alternatively it is possible to use universal primers and amplify a multitude of potentially methylated genetic regions (including the genetic markers of the invention) which are, as described either methylation specific amplified or digested, and then use a set of hybridisation probes for the characteristic markers on e.g. a chip for detection. I.e. gene selection is performed on the chip.
Either set, a set of probes or a set of primers, can be used to obtain the relevant methylation data of the genes of the present invention. Of course, both sets can be used.
The method according to the present invention may be performed by any method suitable for the detection of methylation of the marker genes. In order to provide a robust and optionally re-useable test format, the determination of the gene methylation is preferably performed with a DNA-chip, real-time PCR, or a combination thereof. The DNA chip can be a commercially available general gene chip (also comprising a number of spots for the detection of genes not related to the present method) or a chip specifically designed for the method according to the present invention (which predominantly comprises marker gene detection spots).
Preferably the methylated DNA of the sample is detected by a multiplexed hybridization reaction. In further embodiments a methylated DNA is preamplified prior to hybridization, preferably also prior to methylation specific amplification, or digestion. Preferably, also the amplification reaction is multiplexed (e.g. multiplex PCR).
The inventive methods (for the screening of subsets or for diagnosis or prognosis of lung cancer or lung cancer type) are particularly suitable to detect low amounts of methylated DNA of the inventive marker genes. Preferably the DNA amount in the sample is below 500 ng, below 400 ng, below 300 ng, below 200 ng, below 100 ng, below 50 ng or even below 25 ng. The inventive method is particularly suitable to detect low concentrations of methylated DNA of the inventive marker genes. Preferably the DNA amount in the sample is below 500 ng, below 400 ng, below 300 ng, below 200 ng, below 100 ng, below 50 ng or even below 25 ng, per ml sample.
In another aspect the present invention provides a subset comprising or consisting of nucleic acid primers or hybridization probes being specific for a potentially methylated region of at least marker genes selected from a set of nucleic acid primers or hybridization probes being specific for a potentially methylated region of marker genes being suitable to diagnose or predict lung cancer or a lung cancer type, preferably being selected from adenocarcinoma or squamous cell carcinoma, the marker genes comprising WT1, SALL3, TERT, ACTB, CPEB4 or any other subset selected from one of the following groups
a set of at least 50%, preferably at least 60%, at least 70%, at least 80%, at least 90%, 100% of the markers of anyone of the above a) to t). The present inventive set also includes sets with at least 50% of the above markers for each set since it is also possible to substitute parts of these subsets being specific for—in the case of binary conditions/differentiations—e.g. good or bad prognosis or distinguish between lung cancer or lung cancer types, wherein one part of the subset points into one direction for a certain lung cancer type or cancer/differentiation. It is possible to further complement the 50% part of the set by additional markers specific for diagnosing lung cancer or determining the other part of the good or bad differentiation or differentiation between two lung cancer types. Methods to determine such complementing markers follow the general methods as outlined herein.
Each of these marker subsets is particularly suitable to diagnose lung cancer or lung cancer type or distinguish between certain cancers, samples or cancer types in a methylation specific assay of these genes.
The inventive primers or probes may be of any nucleic acid, including RNA, DNA, PNA (peptide nucleic acids), LNA (locked nucleic acids). The probes might further comprise methylation specific moieties.
The present invention provides a (master) set of 360 marker genes, further also specific gene locations by the PCR products of these genes wherein significant methylation can be detected, as well as subsets therefrom with a certain diagnostic value to detect or diagnose lung cancer or distinguish lung cancer type(s). Preferably the set is optimized for a lung cancer or a lung cancer type. Lung cancer types include, without being limited thereto, adenocarcinoma and squamous cell carcinoma. Further indicators differentiating between disease(s), including the diagnosis of any type of lung cancer or lung tumor, or between tumor type(s) are e.g. benign (non (or limited) proliferative) or malignant, metastatic or non-metastatic. The set can also be optimized for a specific sample type in which the methylated DNA is tested. Such samples include blood, urine, saliva, hair, skin, tissues, in particular tissues of the cancer origin mentioned above, in particular lung tissue such as potentially affected or potentially cancerous lung tissue, or serum, sputum, bronchial lavage. The sample my be obtained from a patient to be diagnosed. In preferred embodiments the test sample to be used in the method of identifying a subset is from the same type as a sample to be used in the diagnosis.
In practice, probes specific for potentially aberrant methylated regions are provided, which can then be used for the diagnostic method.
It is also possible to provide primers suitable for a specific amplification, like PCR, of these regions in order to perform a diagnostic test on the methylation state.
Such probes or primers are provided in the context of a set corresponding to the inventive marker genes or marker gene loci as given in table 1.
Such a set of primers or probes may have all 359 inventive markers present and can then be used for a multitude of different cancer detection methods. Of course, not all markers would have to be used to diagnose a lung cancer or lung cancer type. It is also possible to use certain subsets (or combinations thereof) with a limited number of marker probes or primers for diagnosis of certain categories of lung cancer.
Therefore, the present invention provides sets of primers or probes comprising primers or probes for any single marker subset or any combination of marker subsets disclosed herein. In the following sets of marker genes should be understood to include sets of primer pairs and probes therefor, which can e.g. be provided in a kit.
Set a, WT1, DLX2, SALL3, TERT, PITX2, HOXA10, F2R, CPEB4, NHLH2, SMAD3, ACTB, HOXA1, BOLL, APC, MT1G, PENK, SPARC, DNAJA4, RASSF1, HLA-G, ERCC1, ONECUT2, APC, ABCB1, ZNF573, KCNJ15, ZDHHC11, SFRP2, GDNF, PTTG1, SERPINI1, TNFRSF10C and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers are in particular suitable to detect lung cancer and to distinguish between normal lung tissue (non-cancerous) from lung tumor tissue.
Set b, WIT1, PITX2, SALL3, F2R, DLX2, TERT, HOXA10, MSH4, NHLH2, GNA15, PENK, RASSF1, BOLL, HOXA1, ONECUT2, ABCB1, SPARC, MT1G, HSPA2, SFRP2, PYCARD, GAD1, C5orf4, C5AR1, GDNF, ZDHHC11, SERPINE1, NKX2-1, PITX2, C5AR1, ZNF256, FAM43A, SFRP2, MT3, SERPINE1, CLIC4, TNFRSF10C, GABRA2, MTHFR, ESR2, NEUROG1, PITX2, PLAGL1, TMEFF2, PTTG1, CADM1, S100A8, EFS, JUB, ITGA4, MAGEB2, ERBB2, SRGN, GNAS, TJP2, KCNJ15, SLC25A31, ZNF573, TNFRSF25, APC, KCNQ1, LAMC2, SPHK1, DNAJA4, APC, MBD2, ERCC1, HLA-G, CXADR, TP53, ACTB, KL, SMAD3, HIST1H2AG, CPEB4 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers are also suitable to detect lung cancer and to distinguish between normal lung tissue and lung tumor tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set c, WT1, DLX2, SALL3, TERT, TNFRSF25, ACTB, SMAD3, CPEB4 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers are suitable to detect lung cancer and to distinguish between normal lung tissue (non-cancerous) from lung tumor tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set d, WT1, DLX2, SALL3, TERT, PITX2, TNFRSF25, KL, ACTB, SMAD3, CPEB4 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers are in particular suitable to detect lung cancer and to distinguish between normal lung tissue (non-cancerous) from lung tumor tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set e, WT1, PITX2, SALL3, DLX2, TERT, HOXA10, RASSF1, SPARC, IRAK2, ZNF711, DNAJA4, HLA-G, CXADR, TP53, ACTB, CPEB4 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers are also suitable to detect lung cancer and to distinguish between normal lung tissue (non-cancerous) from lung tumor tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set f, WT1, PITX2, SALL3, F2R, TERT, HOXA10, RASSF1, SPARC, IRAK2, ZNF711, DRD2, DNAJA4, CXADR, TP53, ACTB, CPEB4 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to detect lung cancer and to distinguish between normal lung tissue (non-cancerous) from lung tumor tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set g, WT1, ACTB, DLX2, PITX2, SALL3, HOXA10, TERT, CPEB4, HLA-G, SPARC, RASSF1, DNAJA4, CXADR, TP53, IRAK2, ZNF711 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung carcinoma, in particular using blood samples, e.g. to distinguish blood from healthy persons from tumor samples, including tumor tissue sample or blood from tumor patients. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set h, F2R, ZNF256, CDH13, SERPINB5, KRT14, DLX2, AREG, THRB, HSD17B4, SPARC, HECW2, COL21A1 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and distinguish the grade of differentiation of poor, moderate and well predictions. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set i, KL, HIST1H2AG, TJP2, SRGN, CDX1, TNFRSF25, APC, HIC1, APC, GNA15, ACTB, WT1, KRT17, AIM1L, DPH1, PITX2, PITX2, KIF5B, BMP2K, GBP2, NHLH2, GDNF, BOLL and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and distinguish between malign states (in particular adenocarcinoma and squamous cell carcinoma) together with lung tissue against healthy blood or serum samples. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set j, WT1, DLX2, SALL3, TERT, PITX2, HOXA10, F2R, CPEB4, NHLH2, SMAD3, ACTB, HOXA1, BOLL, APC, MT1G, PENK, SPARC, DNAJA4, RASSF1, HLA-G, ERCC1, ONECUT2, APC, ABCB1, ZNF573, KCNJ15, ZDHHC11, SFRP2, GDNF, PTTG1, SERPINI1, TNFRSF10C and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose, lung cancer and distinguish between malign states selected from adenocarcinoma and squamous cell carcinoma from healthy lung tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set k, HOXA10, NEUROD1 and/or either HOXA10 or NEUR001 can be used to diagnose lung cancer and further to distinguish between adenocarcinoma from squamous cell carcinoma.
Set l, WT1, PITX2, SALL3, F2R, TERT, HOXA10, RASSF1, SPARC, IRAK2, ZNF711, DRD2, DNAJA4, CXADR, TP53, ACTB, CPEB4, DLX2, TNFRSF25, KL, SMAD3 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and distinguish between cancerous lung tissue from healthy lung tissue.
Set m, TNFRSF25, SALL3, RASSF1, TERT, SPARC, F2R, HOXA10, ZNF711, PITX2 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and distinguish between cancerous lung tissue from healthy lung tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set n, SALL3, PITX2, SPARC, F2R, TERT, RASSF1, HOXA10, CXADR, KL and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and distinguish between cancerous lung tissue from healthy lung tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set o, SALL3, SPARC, PITX2, F2R, TERT, RASSF1, HOXA10, KL and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and distinguish between cancerous lung tissue from healthy lung tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set p, SALL3, PITX2, SPARC, F2R, HOXA10, DRD2, ACTB, DNAJA4, CXADR, KL and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and to distinguish between normal lung tissue (non-cancerous) from lung tumor tissue.
Set q, SALL3, SPARC, PITX2, F2R, TERT, RASSF1, HOXA10, TNFRSF25, DNAJA4, TP53, CXADR, KL and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer and to distinguish between normal lung tissue (non-cancerous) from lung tumor tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set r, SPARC, SALL3, F2R, PITX2, RASSF1, HOXA10, TERT, KL, TNFRSF25 and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer, distinguish between adenocarcinoma, healthy lung tissue and squamous cell carcinoma. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set s, SALL3, SPARC, PITX2, F2R, TERT, RASSF1, HOXA10, KL, TNFRSF25, CXADR and 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer, distinguish adenocarcinoma and squamous cell carcinoma from healthy (benign) lung tissue. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Set t, HOXA10, RASSF1, F2R and sets with at least 50%, preferably at least 60%, at least 70%, at least 80% or at least 90% of these markers can be used to diagnose lung cancer, distinguish between adenocarcinoma and squamous cell carcinoma. The distinction or diagnosis can be made by using any sample as described above, including serum, sputum, bronchial lavage.
Also provided are combinations of the above mentioned subsets a) to t), in particular sets comprising markers of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 or more of these subsets, preferably for the lung cancer type or preferably complete sets a) to t). One preferred set comprises gene markers WT1, SALL3, TERT, ACTB and CPEB4. These markers are common in a set for the diagnosis of lung cancer and suitable to distinguish normal from lung cancer samples. This set preferably is supplemented by the marker genes DLX2, TNFRSF25 or SMAD3. Furthermore, the inventive set may comprise any one of the markers ABCB1, ACTB, AIM1L, APC, AREG, BMP2K, BOLL, C5AR1, C5orf4, CADM1, CDH13, CDX1, CLIC4, COL21A1, CPEB4, CXADR, DLX2, DNAJA4, DPH1, DRD2, EFS, ERBB2, ERCC1, ESR2, F2R, FAM43A, GABRA2, GAD1, GBP2, GDNF, GNA15, GNAS, HECW2, HIC1, HIST1H2AG, HLA-G, HOXA1, HOXA10, HSD17B4, HSPA2, IRAK2, ITGA4, JUB, KCNJ15, KCNQ1, KIF5B, KL, KRT14, KRT17, LAMC2, MAGEB2, MBD2, MSH4, MT1G, MT3, MTHFR, NEUROD1, NHLH2, NKX2-1, ONECUT2, PENK, PITX2, PLAGL1, PTTG1, PYCARD, RASSF1, S100A8, SALL3, SERPINB5, SERPINE1, SERPINI1, SFRP2, SLC25A31, SMAD3, SPARC, SPHK1, SRGN, TERT, THRB, TJP2, TMEFF2, TNFRSF10C, TNFRSF25, TP53, ZDHHC11, ZNF256, ZNF711, F2R, HOXA10, KL, SALL3, SPARC, TNFRSF25, WT1 or any combination thereof, in particular preferred are markers ACTB, APC, CPEB4, CXADR, DLX2, DNAJA4, F2R, HOXA10, KL, PITX2, RASSF1, SALL3, SPARC, TERT, (either TNFRSF10C or TNFRSF25 or both), WT1 or any combination thereof, even more preferred are markers HOXA10, PITX2, RASSF1, SALL3, SPARC, TERT or any combination thereof, in a marker set according to the present invention, in particular as additional markers for any one of sets a) to t), especially the marker set of markers WT1, SALL3, TERT, ACTB and CPEB4.
According to a preferred embodiment of the present invention, the methylation of at least two genes, preferably of at least three genes, especially of at least four genes, is determined. Specifically if the present invention is provided as an array test system, at least ten, especially at least fifteen genes, are preferred. In preferred test set-ups (for example in microarrays (“gene-chips”)) preferably at least 20, even more preferred at least 30, especially at least 40 genes, are provided as test markers. As mentioned above, these markers or the means to test the markers can be provided in a set of probes or a set of primers, preferably both.
In a further embodiment the set comprises up to 100000, up to 90000, up to 80000, up to 70000, up to 60000 or 50000 probes or primer pairs (set of two primers for one amplification product), preferably up to 40000, up to 35000, up to 30000, up to 25000, up to 20000, up to 15000, up to 10000, up to 7500, up to 5000, up to 3000, up to 2000, up to 1000, up to 750, up to 500, up to 400, up to 300, or even more preferred up to 200 probes or primers of any kind, particular in the case of immobilized probes on a solid surface such as a chip.
In certain embodiments the primer pairs and probes are specific for a methylated upstream region of the open reading frame of the marker genes.
Preferably the probes or primers are specific for a methylation in the genetic regions defined by SEQ ID NOs 1081 to 1440, including the adjacent up to 500 base pairs, preferably up to 300, up to 200, up to 100, up to 50 or up to 10 adjacent, corresponding to gene marker IDs 1 to 359 of table 1, respectively. I.e. probes or primers of the inventive set (including the full 359 set, as well as subsets and combinations thereof) are specific for the regions and gene loci identified in table 1, last column with reference to the sequence listing, SEQ ID NOs: 1081 to 1440. As can be seen these SEQ IDs correspond to a certain gene, the latter being a member of the inventive sets, in particular of the subsets a) to t), e.g.
Examples of specific probes or primers are given in table 1 with reference to the sequence listing, SEQ ID NOs 1 to 1080, which form especially preferred embodiments of the invention.
Preferably the set of the present invention comprises probes or primers for at least one gene or gene product of the list according to table 1, wherein at least 10%, at least 15%, at least 20%, at least 25%, at least 30%, at least 35%, at least 40%, at least 45%, at least 50%, at least 55%, at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, especially preferred at least 100%, of the total probes or primers are probes or primers for genes of the list according to table 1. Preferably the set, in particular in the case of a set of hybridization probes, is provided immobilized on a solid surface, preferably a chip or in form of a microarray. Since—according to current technology—detection means for genes on a chip allow easier and more robust array design, gene chips using DNA molecules (for detection of methylated DNA in the sample) is a preferred embodiment of the present invention. Such gene chips also allow detection of a large number of nucleic acids.
Preferably the set is provided on a solid surface, in particular a chip, whereon the primers or probes can be immobilized. Solid surfaces or chips may be of any material suitable for the immobilization of biomolecules such as the moieties, including glass, modified glass (aldehyde modified) or metal chips.
The primers or probes can also be provided as such, including lyophilized forms or being in solution, preferably with suitable buffers. The probes and primers can of course be provided in a suitable container, e.g. a tube or micro tube.
The present invention also relates to a method of identifying lung cancer or lung cancer type in a sample comprising DNA from a subject or patient, comprising obtaining a set of nucleic acid primers (or primer pairs) or hybridization probes as defined above (comprising each specific subset or combinations thereof), determining the methylation status of the genes in the sample for which the members of the set are specific for and comparing the methylation status of the genes with the status of a confirmed lung cancer or lung cancer type positive and/or negative state, thereby identifying the lung cancer or lung cancer type in the sample. In general the inventive method has been described above and all preferred embodiments of such methods also apply to the method using the set provided herein.
The inventive marker set, including certain disclosed subsets and subsets, which can be identified with the methods disclosed herein, are suitable to diagnose lung cancer and distinguish between different lung cancer forms, in particular for diagnostic or prognostic uses. Preferably the markers used (e.g. by utilizing primers or probes of the inventive set) for the inventive diagnostic or prognostic method may be used in smaller amounts than e.g. in the set (or kit) or chip as such, which may be designed for more than one fine tuned diagnosis or prognosis. The markers used for the diagnostic or prognostic method may be up to 100000, up to 90000, up to 80000, up to 70000, up to 60000 or 50000, preferably up to 40000, up to 35000, up to 30000, up to 25000, up to 20,000, up to 15000, up to 10000, up to 7500, up to 5000, up to 3000, up to 2000, up to 1000, up to 750, up to 500, up to 400, up to 300, up to 200, up to 100, up to 80, or even more preferred up to 60. The inventive set of marker primers or probes can be employed in chip (immobilised) based assays, products or methods, or in PCR based kits or methods. Both, PCR and hybridisation (e.g. on a chip) can be used to detect methylated genes.
The inventive marker set, including certain disclosed subsets, which can be identified with the methods disclosed herein, are suitable to distinguish between lung cancer from normal tissue, in particular for diagnostic or prognostic uses.
The inventive marker set, including certain disclosed subsets, which can be identified with the methods disclosed herein, are suitable to distinguish between adenocarcinoma from squamous cell carcinoma, in particular for diagnostic or prognostic uses.
The present invention is further illustrated by the following examples, without being restricted thereto.
Samples from solid tumors were derived from initial surgical resection of primary tumors. Tumor tissue sections were derived from histopathology and histopathological data as well clinical data were monitored over the time of clinical management of the patients and/or collected from patient reports in the study center. Anonymised data and DNA were provided.
The invention assay is a multiplexed assay for DNA methylation testing of up to (or even more than) 360 methylation candidate markers, enabling convenient methylation analyses for tumor-marker definition. In its best mode the test is a combined multiplex-PCR and microarray hybridization technique for multiplexed methylation testing. The inventive marker genes, PCR primer sequences, hybridization probe sequences and expected PCR products are given in table 1, above.
Targeting hypermethylated DNA regions in the inventive marker genes in several neoplasias, methylation analysis is performed via methylation dependent restriction enzyme (MSRE) digestion of 500 ng of starting DNA. A combination of several MSREs warrants complete digestion of unmethylated DNA. All targeted DNA regions have been selected in that way that sequences containing multiple MSRE sites are flanked by methylation independent restriction enzyme sites. This strategy enables pre-amplification of the methylated DNA fraction before methylation analyses. Thus, the design and pre-amplification would enable methylation testing on serum, urine, stool etc. when DNA is limiting.
When testing DNA without pre-amplification upon digestion of 500 ng the methylated DNA fraction is amplified within 16 multiplex PCRs and detected via microarray hybridization. Within these 16 multiplex-PCR reactions 360 different human DNA products can be amplified. From these about 20 amplicons serve as digestion & amplification controls and are either derived from known differentially methylated human DNA regions, or from several regions without any sites of MSREs used in this system. The primer set (every reverse primer is biotinylated) used is targeting 347 different sites located in the 5′UTR of 323 gene regions.
After PCR amplicons are pooled and positives are detected using strepavidin-Cy3 via microarray hybridization. Although the melting temperature of CpG rich DNA is very high, primer and probe-design as well as hybridization conditions have been optimized, thus this assay enables unequivocal multiplexed methylation testing of human DNA samples. The assay has been designed such that 24 samples can be run in parallel using 384well PCR plates.
Handling of many DNA samples in several plates in parallel can be easily performed enabling completion of analyses within 1-2 days.
The entire procedure provides the user to setup a specific PCR test and subsequent gel-based or hybridization-based testing of selected markers using single primer-pairs or primer-subsets as provided herein or identified by the inventive method from the 360 marker set.
MSRE digestion of DNA (about 500 ng) was performed at 37° C. over night in a volume of 30 μl in 1× Tango-restriction enzyme digestion buffer (MBI Fermentas) using 8 units of each MSREs AciI (New England Biolabs), Hin 6 I and Hpa II (both from MBI Fermentas). Digestions were stopped by heat inactivation (10 min, 75° C.) and subjected to PCR amplification.
An aliquot of 20 μl MSRE digested DNA (or in case of preamplification of methylated DNA—see below—about 500 ng were added in a volume of 20 μl) was added to 280 μl of PCR-Premix (without primers). Premix consisted of all reagents obtaining a final concentration of 1× HotStarTaq Buffer (Qiagen); 160 μM dNT-Ps, 5% DMSO and 0.6 U Hot Firepol Taq (Solis Biodyne) per 20 μl reaction. Alternatively an equal amount of HotStarTaq (Qiagen) could be used. Eighteen (18) μl of the Pre-Mix including digested DNA were aliquoted in 16 0.2 ml PCR tubes and to each PCR tube 2 μl of each primer-premix 1-16 (containing 0.83pmol/μl of each primer) were added. PCR reactions were amplified using a thermal cycling profile of 15 min/95° C. and 40 cycles of each 40 sec/95° C., 40 sec/65° C., 1 min20 sec/72° C. and a final elongation of 7 min/72° C., then reactions were cooled. After amplification the 16 different multiplex-PCR amplicons from each DNA sample were pooled. Successful amplification was controlled using 10 μl of the pooled 16 different PCR reactions per sample. Positive amplification obtained a smear in the range of 100-300 bp on EtBr stained agarose gels; negative amplification controls must not show a smear in this range.
Microarrays with the probes of the 360 marker set are blocked for 30 min in 3M Urea containing 0.1% SDS, at room temperature submerged in a stirred choplin char. After blocking slides are washed in 0.1×SSC/0.2% SDS for 5 min, dipped into water and dried by centrifugation.
The PCR-amplicon-pool of each sample is mixed with an equal amount of 2× hybridization buffer (7×SSC, 0.6% SDS, 50% formamide), desaturated for 5 min at 95° C. and held at 70° C. until loading an aliquot of 100 μl onto an array covered by a gasket slide (Agilent). Arrays are hybridized under maximum speed of rotation in an Agilent-hybridization oven for 16 h at 52° C. After removal of gasket-slides microarray-slides are washed at room temperature in wash-solution I (1×SSC, 0.2% SDS) for 5 min and wash solution II (0.1×SSC, 0.2% SDS) for 5 min, and a final wash by dipping the slides 3 times into wash solution III (0.1×SSC), the slides are dried by centrifugation.
For detection of hybridized biotinylated PCR amplicons, streptavidin-Cy3-conjugate (Caltag Laboratories) is diluted 1:400 in PBST-MP (1×PBS, 0.1% Tween 20; 1% skimmed dry milk powder [Sucofin; Germany]), pipetted onto microarrays covered with a coverslip and incubated 30 min at room temperature in the dark. Then coverslips are washed off from the slides using PBST (1×PBS, 0.1% Tween 20) and then slides are washed in fresh PEST for 5 min, rinsed with water and dried by centrifugation.
In many situations DNA amount is limited. Although the inventive methylation test is performing well with low amounts of DNA (see above), especially minimal invasive testing using cell free DNA from serum, stool, urine, and other body fluids is of diagnostic relevance.
Samples can be preamplified prior methylation testing as follows: DNA was digested with restriction enzyme FspI (and/or Csp6I, and/or MseI, and/or Tsp5091; or their isoschizomeres) and after (heat) inactivation of the restriction enzyme the fragments were circularized using T4 DNA ligase. Ligation-products were digested using a mixture of methylation sensitive restriction enzymes. Upon enzyme-inactivation the entire mixture was amplified using rolling circle amplification (RCA) by phi29-phage polymerase. The RCA-amplicons were then directly subjected to the multiplex-PCRs of the inventive methylation test without further need of digestion of the DNA prior amplification.
Alternatively the preamplified DNA which is enriched for methylated DNA regions can be directly subjected to fluorescent-labelling and the labeled products can be hybridized onto the microarrays using the same conditions as described above for hybridization of PCR products. Then the streptavidin-Cy3 detection step has to be omitted and slides should be scanned directly upon stringency washes and drying the slides. Based on the experimental design for microarray analyses, either single labeled or dual-labeled hybridizations might be generated. From our experiences we successfully used the single label-design for class comparisons. Although the preamplification protocol enables analyses of spurious amounts of DNA, it is also suited for performing genomic methylation screens.
To elucidate methylation biomarkers for prediction of meta-stasis risk on a genomewide level we subjected 500 ng of DNA derived from primary tumor samples to amplification of the methylated DNA using the procedure outlined above. RCA-amplicons derived from metastasized and non-metastasized samples were labelled using the CGH Labeling Kit (Enzo, Farmingdale, N.Y.) and labelled products hybridized onto human 244 k CpG island arrays (Agilent, Waldbronn, Germany). All manipulations were according the instructions of the manufacturers.
Hybridizations performed on a chip with probes for the inventive 360 marker genes were scanned using a GenePix 4000A scanner (Molecular Devices, Ismaning, Germany) with a PMT set-ting to 700V/cm (equal for both wavelengths). Raw image data were extracted using GenePix 6.0 software (Molecular Devices, Ismaning, Germany).
Microarray data analyses were performed using BRB-ArrayTools developed by Dr. Richard Simon and BRB-ArrayTools Development Team. The software package BRB Array Tools (version 3.6; in the www at linus.nci.nih.gov/BRB-ArrayTools.html) was used according recommendations of authors and settings used for analyses are delineated in the results if appropriate. For every hybridization, background intensities were subtracted from foreground intensities for each spot. Global normalization was used to median center the log-ratios on each array in order to adjust for differences in spot/label intensities.
P-values (p) used for feature selection for classification and prediction were based on the univariate significance levels (alpha). P-values (p) and mis-classification rate during cross validation (MCR) were given along the result data.
DNA methylation analysis of 96 DNA samples derived from both normal and lung-tumour tissue of 48 patient samples and 8 DNA samples isolated from peripheral blood (PB) of healthy individuals were analysed for methylation deviations in the inventive set of 359 genes.
From this analysis DNA-methylation-biomarkers suitable for distinction of tumour and normal lung DNA as well as DNA-methylation-profiles from blood DNA of healthy controls were deduced. Diagnostic and prognostic markers subsets are suitable for diagnostic testing and presymptomatic screening for early detection of lung cancer were determined, in DNA derived from lung tissue, but also in DNA extracts from patients other than lung, like sputum, serum or plasma.
DNA Methylation testing results and data analyses of chip results as well as qPCR validation of a subset of markers derived from chip-based testing are provided.
DNA Samples analysed were from blood of 8 healthy individuals (PB), 19 tumours (AdenoCa, adenocarcinoma) and 19 normal lung tissue (N) of adenocarcinoma patients and 29 tumours (SqCCL, squamous cell carcinoma) and 29 normal lung tissue (N) of squamous cell carcinoma patients.
For DNA methylation testing 600 ng of DNA were digested and data derived from DNA-microarray hybridizations analysed using the BRB array tools statistical software package. Class comparison, and class prediction analysis were performed with respect to sample groups as listed above or for delineation of biomarkers for tumour samples both AdenoCa and SqCCL were treated as one tumour sample group (TU).
The design of the test enables methylation testing on DNA directly derived from the biological source. The test is also suitable for using a DNA preamplification upon MSRE digestion (as outlined above). Thus using the methylation specific preamplification of minute amounts of DNA samples, biomarker testing is feasible on small samples and limited amounts of DNA. Thus multiplexed PCR and methylation testing is easily performed on preamplified DNA obtained from these DNA samples. This strategy would improve also testing of serum, urine, stool, synovial fluid, sputum and other body fluids using the conceptual design of the methylation test.
The possibility of preamplification enables also differential methylation hybridization of the preamplified DNA itself. This option is warranted by the design of the test and the probes. Thus using the probes of the methylation test (or the array) for hybridization of labelled DNA after enrichment of either the methylated as well as the unmethylated DNA fractions of any DNA sample, can be used for methylation testing omitting the multiplex PCR.
In addition the biomarkers described herein could be applied for methylation testing using alternative approaches, e.g. methylation sensitive PCR and strategies which are sodium-bisulfite DNA deamination based and not based on MSRE digestion of DNA. These sets of methylation markers are suitable markers for disease-monitoring, -progression, -prediction, therapy-decision and -response.
These list of methylation markers were found significant (p<0.005) between TU and N using “unpaired” statistical testing of DNA methylation of 48 tumour samples versus 48 healthy lung tissue samples. Significant markers with 2 fold difference of signal intensities of both classes with p<0.005 are listed.
Class prediction using different statistical methods for elucidating marker panels enabling best correct classification of TU and N (p<0.005).
<1e−07
<1e−07
<1e−07
<1e−07
<1e−07
<1e−07
<1e−07
<1e−07
<1e−07
Using “4 pairs of methylation markers” derived from greedy pairs class prediction with supportive vector machines enables 92% correct classification of TU and N.
Using “5 pairs of methylation markers” derived from greedy pairs class prediction with supportive vector machines enables 95% correct classification of TU and N.
Using “16 methylation markers” derived from the Recursive Feature Elimination method for class prediction with Diagonal Linear Discriminant Analysis enables 100% correct classification of TU and N.
Due to some differences in data importing/normalisation repeated collation of data for statistics (using BRB v. 3.8) a genelist with minor differences (compared to example 12e) has been calculated form data, and is as given below:
To distinguish PB, N, and TU is of interest when minimal invasive testing for lung cancer has to be performed using serum- or plasma from peripheral blood. The markers distinguishing PB, N and TU will be best suited therefore. Using “16 methylation markers” derived from the Recursive Feature Elimination method for class prediction with Diagonal Linear Discriminant Analysis enables 91% correct classification.
Distinguishing the grade of differentiation of the tumours could be also achieved by DNA methylation marker testing. Although the correct classification is only about 60% in this example, the lung tumour groups “AdenoCa” and “SqCCL” can be split and used separately for determining the grade of tumour-differentiation for better performance.
Using Binary Tree prediction (applicable for elucidation of markers for more than 2 classes) provides several sets of predictors which enable classification of PB, AdenoCa, SqCCL, N. These marker sets could be used alternatively for classification.
Quantitative PCR with primers for markers elucidated by microarray analysis were run on MSRE-digested DNAs from the same sample groups as analyzed on microarrays. Marker sets for SYBRGreen qPCR were from Example 10f and Example 10d.
Negative amplification (no Cp-value generated upon 45 cycles of PCR amplification with SYBR green) were set to Cp=45; all qPCR-Cp-values were subtracted from 45.01 to obtain transformed data directly comparable to microarray data,—thus the higher the value the more product was generated (resembles a lower Cp-value. Statistical testing of the transformed data was performed in the same manner as the microarray data using BRB-AT software.
Class comparison and different strategies/methods for class prediction using the qPCR enables correct classification of different sample groups. Although qPCR conditions were not optimized but run under our standard conditions, successful classification of groups with markers deduced from microarray-analysis confirms reliability of methylation markers.
The prediction rule is defined by the inner sum of the weights (wi) and expression (xi) of significant genes. The expression is the log ratios for dual-channel data and log intensities for single-channel data.
A sample is classified to the class N if the sum is greater than the threshold; that is, Σiwi xi>threshold.
The threshold for the Compound Covariate predictor is −172.255
The threshold for the Diagonal Linear Discriminant predictor is −15.376
The threshold for the Support Vector Machine predictor is 0.838
Cross-Validation ROC curve from the Bayesian Compound Covariate Predictor. The area under the curve is 0.944 (
Mean Percent of Correct Classification, n=96:
| Number | Date | Country | Kind |
|---|---|---|---|
| 09450020.4 | Jan 2009 | EP | regional |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 13146901 | Jul 2011 | US |
| Child | 15096848 | US |
