Cancer is the leading cause of death in the developing countries. In many developed countries including the United States, cancer has replaced heart disease as the primary cause of death. The number of cancer deaths worldwide is projected to increase from 7.6 million in 2008 to 11.5 million in 2030 (http://www.who.int). It is apparent that cancer has become a global public health issue.

Tremendous progress has been made in our fight against cancer. It is now recognized that cancer is a disease of the genome^{[ 1, 2, 3, 4]}. Alteration or modification of our genome affects key cellular signalling pathways and processes, resulting in uncontrolled cell growth, local invasion and eventually distant metastasis characteristic of cancer^{[ 5]}. Since the discovery that a single point mutation in HRAS oncogene was associated with bladder cancer in 1982^{[ 6]}, many oncogenes and tumor-suppressor genes have been identified and characterized^{[ 7]}. These findings have revolutionized cancer treatment with development of pharmacological inhibitors targeting the cognate proteins of oncogenes^{[ 8, 9]}.

The completion of human genome reference sequence about a decade ago and the advent of high-throughput sequencing technology in the last several years have ushered in a new era of cancer genomics^{[ 9, 10, 11, 12, 13, 14, 15, 16, 17]}. As a result, many new cancer genes or pathways have been discovered, new genomic features revealed and new drugs developed^{[ 5, 14, 18, 19]}. This article will review the technological advance in the analysis of our genome with emphasis on next generation sequencing, summarize the new discoveries by human cancer genomics, and briefly discuss the impact of these findings on the management of human cancer.

In a normal individual, the haploid human genome consists of about 3 billion nucleotides of DNA with a total of 6 billion base pairs in a diploid nucleus distributed unequally among 22 pairs of autosomes and 1 pair of sex chromosomes^{[ 20]}. Only about 2% of our genome (about 23 000 genes) codes for proteins. Less than 1% of our genome also codes for functional RNAs, such as microRNAs, ribosomal RNAs, long intergenic non-coding RNAs and other functional RNAs^{[ 21, 22]}. Up to 8% of our genome is highly conserved in evolution and a proportion of these sequences are important regulatory elements, such as promoters, enhancers and locus control regions. The remaining 90% of our genome does not encode any proteins or functional RNAs, and the majority of these sequences are repetitive elements, such as SINE (short interspersed) or LINE (long interspersed) elements. They can either occur as clusters of tandem repeats or as interspersed repeats. Their function is mostly unknown except for centromere and telomere, which are involved in chromosome packaging, segregation and replication and the maintenance of chromosome ends, respectively. These sequences have been frequently referred to as "junk" DNA or "dark matter".

Not all of the about 23 000 genes in our genome are transcribed. This gene expression profile or transcriptome is dependent on the cell type and stage of differentiation. In a given cell at a given time, only a fraction of the coding genes in our genome are transcribed, generating approximately 300 000 mRNA molecules^{[ 23]}. In contrast, exome refers to the combined DNA sequences of all exons of protein-coding and RNA genes in our genome. It is identical for all types of cells in a given organism regardless of the stage of cell differentiation.

Our genome is unique with significant variability among individuals. This variation includes single base pair changes (single nucleotide polymorphism/variants, SNP/Vs) as well as differences in the length of nucleotide sequences ranging from several to hundreds or millions of base pairs (copy number variants, CNVs)^{[ 24, 25, 26, 27]}. In fact, 1 in every 300 bases in our genome is found to be polymorphic and any 2 individuals will differ at >3 million SNP/V locations. These variations can occur in both coding as well as non-coding regions of the human genome. It is therefore important to bear in mind these normal variations when we try to analyze and interpret genomic alterations in cancer.

Conventional cytogenetics or chromosome analysis is the first and most widely used technology to analyze the human genome. In the field of cancer genomics, Philadelphia chromosome was first recurrent chromosomal abnormality reported in chronic granulocytic leukemia by Peter Nowell and David Hungerford with this method in 1960^{[ 28]}. With the development of chromosome banding techniques in the 1970s, Janet Rowley demonstrated that the t(9;22)(q34;q11) translocation was the cause of Philadelphia chromosome in chronic myelogenous leukemia^{[ 29]}. Many additional chromosomal abnormalities with diagnostic, prognostic and therapeutic applications have been identified by conventional cytogenetics. These abnormalities were subsequently characterized by other molecular methods, such as polymerase chain reaction (PCR) and Sanger sequencing, that form the basis of the current World Health Organization (WHO) classification of pathology and genetics of human cancers. Although routine chromosome analysis covers the entire genome, this technology is limited by its low resolution (about 15-30 Mbp) and can only be used to assess for gain or loss of large regions of the genome and chromosome translocations or rearrangements. To overcome this limitation, fluorescence in situ hybridization (FISH) and comparative genomic hybridization were developed in the 1990s with a resolution of about 50-100 Kbp per DNA probe^{[ 30]}. However, each probe can only interrogate about 0.01% of the genome and spectral overlap of fluorescence signals makes this technology difficult to analyze the entire genome.

Many chip or microarray-based platforms, such as bacterial artificial chromosome comparative genomic hybridization and oligonucleotide SNP arrays, were developed in the last decade or so^{[ 31]}. These technologies can essentially examine the entire genome with millions of unique DNA sequence probes and a resolution of less than 100 Kbp for copy number variations as well as allelic imbalance. Except for whole genome sequencing, SNP array is the only platform that can detect copy neutral loss of heterozygosity (CN-LOH, a process whereby a lost portion of the chromosome is reduplicated from the sister chromatid). Though the most recent array platform can detect mutations and small insertion/deletions (indels), novel SNP/Vs cannot be analyzed by these prefabricated chips with preselected genomic loci. To complement the array platforms, multiplex PCRs were developed at the same time to detect multiple known hotspot mutations in many different cancer genes^{[ 18, 32, 33, 34, 35]}. The SNaPshot platform from Applied Biosystems consists of multiplex PCR and single base extension reactions with fluorescent dideoxy trinucleotides to interrogate more than 50 hotspot mutations from 8-14 key cancer genes, such as RAS, BRAF, AKT1, EGFR, PIK3 CA, MEK1, PTEN, IDH1 and IDH2. Up to 10 single mutations from different amplicons can be examined in a single extension reaction. The most common exon 20 insertion and exon 19 deletion mutations of EGFR gene are often included in the panel. The SNaPshot products are then resolved and analyzed using capillary electrophoresis on the ABI Genetic Analyzers. In contrast, the Sequenom platform employs matrix-assisted laser desorption/ionization time of flight mass spectrometry to rapidly analyze the multiplex PCR products. The Sequenom OncoCarta V1.0 kit can interrogate 238 somatic mutations from 19 different oncogenes including those assayed by SNaPshot platform.

The ideal technology to interrogate the entire genome for all kinds of known and unknown abnormalities with a resolution at the single nucleotide level is DNA sequencing. The traditional dideoxy chain termination method developed by Fred Sanger in the 1970s can only read about 1 000 nucleotides in a single reaction, less than one-millionth of our genome^{[ 36]}. Even with method improvement and automation, it still took about 10 years and a staggering 2.7 billion US dollars to sequence the first human reference genome^{[ 9, 12, 13]}. Apparently, Sanger sequencing can only be used practically and economically to analyze the known hotspot mutations, rather than the entire human genome. In the latter half of last decade, next generation sequencing or massively parallel sequencing (NGS or MPS) technologies were invented^{[ 10, 11, 17, 37, 38, 39]}. These technologies, along with powerful computers and informatics, were rapidly adapted by many laboratories to explore large genomic regions or the entire human genome for abnormalities in human cancer and other congenital disorders. Now it only takes several days to a few weeks and a few thousand US dollars to sequence partial or the entire human genome. Since the completion of the first human cancer genome sequence by NGS in 2008, more than 900 genomes from more than 25 different cancer types have been sequenced thanks to individual and large-scale collaborative efforts such as The Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov/) and the International Cancer Genome Consortium (ICGC) (http://icgc.org/).

NGS technologies allow massively parallel sequencing of millions of genomic DNA fragments simultaneously. There are several different NGS platforms available commercially^{[ 10, 11, 37, 38]}. The Roche 454 GS FLX Titanium and Junior systems (Roche Applied Sciences, Penzberg, Germany) use the sequencing-by-synthesis strategy based on the principle of pyrosequencing reaction. The Life Technology ABI/5500 SOLiD^TM system (Life Technologies, Carlsbad, CA) uses sequencing-by-ligation technology involving iterative rounds of oligonucleotide ligation extension. In contrast, the more recent Life Technology Ion Torrent personal genome machine (PGM) and Proton systems employ sequencing-by-synthesis rather than sequencing-by-ligation strategy. The detection method involves ultra-sensitive pH meter that measures hydrogen ions released after nucleotide incorporation during DNA synthesis. The Illumina HiSeq2000/2500 and MiSeq systems (Illumina, San Diego, CA) also use sequencing-by-synthesis strategy with reversible dye terminator and imaging. Other newer platforms, such as Helicos from BioSciences HeliScope, SMRT from Pacific Biosciences and Oxford Nanopore Technologies, allow real-time direct sequencing of single molecules without amplification. Regardless the types of platform, the workflow of NGS involves parallel sequencing with different chemical reactions, base calling, sequence alignment and variant calling.

NGS technologies can be employed to sequence the whole genome (whole genome sequencing, WGS), the exome (whole exome sequencing, WES), or the transcriptome (whole transcriptome sequencing, WTS or RNA-Seq)^{[ 17]}. WGS can interrogate the entire genome for all types of abnormalities including complex genomic rearrangements. However, since more than 90% of the abnormalities identified by WGS are of unknown biological or clinical significance now, WGS is best suited for discovery of new cancer genes and investigation of new mechanisms. WES started with the exome only. With more advanced capturing techniques, WES now covers a much broader range of the genome including all coding exons, microRNA genes, the untranslated regions, genes of the newly discovered unannotated transcripts and other functional RNA genes. It is more cost-effective and less time-consuming with higher depth of coverage than WGS. However, WES will not detect chromosome rearrangements or fusion genes. In the near future, the combination of WES and WTS, which also provides quantitative information of gene expression, might be the best practical strategy to explore the human cancer genome for clinical application. An alternative to WGS/WES/WTS for most clinical laboratories currently, however, is the deep sequencing of a panel of genes that are recurrently mutated and have prognostic or predictive significance in a given cancer^{[ 40, 41, 42, 43, 44, 45, 46]}. For example, the Washington University Cancer Mutation Profiling (WUCaMP) assay examines 25 cancer genes using 454 GS Junior NGS platform to detect clinically actionable somatic mutations in human cancers^{[ 45]}. The University of Washington UW-OncoPlex assay interrogates 194 clinically relevant genes using Illumina HiSeq2000 platform for somatic point mutations as well as other abnormalities including small indels, internal tandem duplications, gain or loss of gene copy number, and chromosomal rearrangements^{[ 44]}. Frampton et al from Foundation Medicine also developed and validated a clinical cancer genomic profiling test using Illumina HiSeq 2000 platform^{[ 43]}. In their assay, 4 557 exons of 287 cancer-related genes were interrogated with high sensitivity (95%-99% across genomic alterations) and specificity (positive predictive value>99%). The disadvantages of targeted sequencing are the relatively high initial costs in designing a custom gene panel and the high cost associated with adding additional genes to the panel.

Numerous acquired genomic abnormalities have been identified in the human cancer genome^{[ 1, 2, 4, 6, 16, 17, 47, 48]}. These include single nucleotide changes or point mutations, small indels, copy number alterations (CNAs) such as gain or loss, and structural abnormalities. Single nucleotide changes within the coding region of a gene may result in missense mutation (amino acid substitution), nonsense mutation (truncation of the gene product) or splice-site mutation. Small indels may result in disruption of the reading frame of a gene (out-of-frame indels). CNAs may involve a single gene or multiple genes, causing aberrant expression of the affected gene(s). Structural abnormalities include inter-chromosomal and intra-chromosomal translocations or rearrangements, generating fusion gene(s) with altered functionality. These acquired genomic abnormalities are the underpinning of carcinogenesis and cancer evolution through activation of cancer genes.

Many of the classical cancer genes encode proteins that affect genome stability, cell division, proliferation or apoptosis^{[ 4, 5]}. In contrast, many of the new cancer genes discovered by large scale and WGS affect global processes such as signal transduction pathways, epigenetic regulation, RNA splicing, metabolism, etc (see Table 1 for a partial list and http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/ for details). Here is a brief summary of pathways or cellular processes and new cancer genes discovered in the last decade by cancer genomics.

Table 1

Table 1

Table 1 Pathways and New Cancer Genes Discovered by Cancer Genomics Pathways or Processes Example of New Cancer Genes Representative Cancer Types Signal Transduction Receptor Tyrosine Kinase Signalling EGFR, ERBB2, MET, ALK, JAK2, RET, ROS, FGFR1, FGFR2, PDGFRA, CRKL, CALR, RHOA Breast, Brain, Lung, Myeloproliferative Neoplasm, T Cell Lymphoma MAPK Signalling KRAS, NRAS, BRAF, MAP2 K1, MAP3 K1, MAP2 K4, NF1 Breast, Lung, Melanoma, Hairy Cell Leukemia PI3K Signalling PIK3 CA, AKT1, AKT3, PTEN, PIK3 R1 Breast, Brain, Colorectal, Endometrial Notch Signalling NOTCH1, NOTCH2, NOTCH3 Head and Neck Squamous Cell, T Cell ALL mTOR Signalling STK11, TSC1, TSC2 Colorectal, Bladder Wnt/β-Catelin Signalling APC, CTNNB1 Gastrointestinal TGFβ Signalling SMAD2, SMAD4, TGFBR2 Colorectal NF-βB Signalling MYD88 Diffuse Large B Cell Lymphoma SLIT/ROBO Signalling ROBO1, ROBO2, SLIT2 Pancreatic KEAP1/NRF1 Signalling KEAP1 Lung Semaphorin Signalling SEMA3 A, SEMA3 E, SEMA5 A Pancreatic RAC/PAK Signalling RAC1, RAC2, CDC42, ELMO1, DOCK2, PREX2 Melanoma, Esophageal Epigenomic Regulation DNA Methylation DNMT3A AML, MDS DNA Hydroxymethylation TET2 AML, MDS Histone (Lysine) Methyl-transferase (KMT) EZH2, MLL, MLL2, MLL3, NSD1, NSD3, SETD2 AML, MDS, Lung, Gastric, Prostate, Large B Cell Lymphoma Histone (Lysine) Deme-thylase (KDM) JARID1 A, UTX, KDM5 A, KDM5 C, KDM6 A Myeloma, Renal Cell, AML and Others Histone Acetyltransferase CREBP, EP300 Diffuse Large B Cell Lymphoma Histone Deubiquitinase BAP1 Mesothelioma, Renal Clear Cell SWI/SNF Chromatin Complex PBRM1, SMARCA1, SMARCA4, ARID1 A, ARID1 B, ARID2 Melanoma, Renal and Ovarian Clear Cell, Hepatocellular and Pancreatic Chromatin Compaction CHD1, CHD2, CHD3, CHD4 CLL, Prostate and Endometrial RNA Splicing Machinery SF3 A1, SF3 B1, SF1, SFRS1, SFRS7, ZRSR2, SRSF2, PRPF40 B, U2 AF1, U2 AF2, U2 AF65 MDS, CLL, Lung, Breast, Pancreatic Metabolism IDH1, IDH2 Brain, AML and Others Genome Stability Genome Integrity TP53, MDM2, MSH, MLH, ATM, POLE Melanoma, Bladder, Hepatocellular, and Many Others Telomere Length TERT promoter mutation Melanoma, Bladder, Hepatocellular, and Many Others Transcription Factor Lineage Dependency MITF, NKX2 -1, SOX-2, EGR, ETV1, CDX2 Lung, Esophageal Others MYC, RUNX1, GATA3, FOXA1, NKX3 .1, SOX9, NEF2 L2, MED12 Breast, Lung, Prostate Protein Ubiquitination SPOP, PBXW7, WWP1 Endometrial, Head and Neck, Bladder, Gastrointestinal Protein Translation FAM46 C Myeloma Protein Homeostasis XBP1 Myeloma RNA Abundance DIS3 Myeloma Cell Cycle CCND1, CCNE1, CDKN2 A, CDKN2 B, CDKN1 B Uterine Serous Cell Apoptosis MCL1, BCL2 A1, BCL2 L1 Breast, Lung, Brain, Colorectal, Melanoma

Table 1 Pathways and New Cancer Genes Discovered by Cancer Genomics

4 .1 Signal transduction pathways

Signal transduction pathways play an important role in normal cell proliferation and survival. In the 1990s, mutations were discovered in key genes encoding members of the receptor tyrosine kinase as well as other signalling pathways in many cancers, such as HER-2 in breast cancer, c-KIT in gastrointestinal stromal tumors, and ABL in chronic myelogenous leukemia^{[ 4, 49]}. Small molecule inhibitors and/or monoclonal antibodies targeting these mutant proteins, such as imatinib and trastuzumab, were subsequently developed and successfully used to treat patients with cancers harbouring the mutant proteins. These successful stories led to unbiased sequencing to survey dozens of cancer genes from dozens of patients by Sanger method in the early 2000s. These effort revealed mutations in EGFR in non-small cell lung cancer, BRAF V600E in melanomas, PIK3 CA in breast, colorectal and endometrial cancers, and JAK2 V617F in myeloproliferative neoplasms^{[ 1, 2, 4, 16, 50, 51]}. Some of these findings have already had an impact on drug development and clinical treatment, such as erlotinib/gefitinib for non small cell lung cancer, vemurafenib for metastatic melanoma and jakafi for primary myelofibrosis.

The wide-spread use of NGS in the last several years have revealed new recurrent mutations in genes involved in several signal transduction pathways, and some of these pathways were not previously suspected to drive cancer (Table 1). For example, MAP3 K1 and MPA2 K4 genes encoding serine/threonine kinases in the MAPK signalling pathways were mutated in breast cancer^{[ 52]}; RAC1, ELMO1 and DOCK2 genes in the RAC/PAK signalling pathways were mutated in melanoma and esophageal cancers, respectively^{[ 53, 54]}; NYD88 gene in the NF-κB signalling pathway was mutated in diffuse large B cell lymphoma^{[ 55]}; RHOA is mutated in the majority of angioimmunoblastic T cell lymphoma^{[ 56]}; and surprisingly, ROBO and SLIT genes involved in axon guidance in neurons were mutated in about 20% of pancreatic adenocarcinomas^{[ 54, 57]}. Most recently, recurrent mutations in the calcium-binding protein calreticulin gene CALR were found in myeloproliferative neoplasms that are negative for JAK2 and MPL mutations^{[ 58]}. These studies also confirmed the classical cancer genes and extended their role in new cancers. For example, the common BRAF V600E mutation in melanomas is also present in hairy cell leukemia^{[ 55, 59]}.

Besides point mutations or small indels, CNAs and genomic rearrangements may also activate cancer genes^{[ 60]}. MCL1 and BCL2 L1 encoding important anti-apoptotic proteins were found to be amplified in a wide range of cancers, including breast, lung, colorectal and melanoma. Rearrangements involving ALK, RET and ROS1 have been reported in non small cell lung cancer. BRAF rearrangements have also been observed in pediatric pilocytic astrocytomas and melanomas.

4 .2 Epigenomic regulation

One of the most exciting discoveries by cancer genomics has been the critical role of epigenomic changes in tumorigenesis^{[ 61, 62, 63, 64]}. Reversible modification of chromatin is a complex process involving more than 40 genes many of which are mutated in a variety of human cancers. Histone (lysine) methyltransferases (KMTs), histone (lysine) demethylases (KDMs), histone deubiquitinase, and histone acetyltransferases (HATs) modify histone 3 posttranslaltionally, and their coding genes are frequently mutated in acute myeloid leukemia, myelodysplasia, renal, gastric, prostate and other solid tissue tumors. Mutations in genes coding for the SWI/SNF complexes, such as PBRM1 and ARID1A, which regulate chromatin structure through ATP-dependent nucleosome remodelling, are also seen in renal cell, ovarian clear cell, prostate, hepatocellular carcinomas and melanomas. Another unexpected chromatin-related target is the chromodomain-helicase-DNA-binding ( CHD) gene family. CHD proteins regulate chromatin compaction during stem cell differentiation, and mutations of CHD1 through CHD4 have been observed in prostate, endometrial, and brain tumors.

Modification of DNA nucleotides, particularly within the CpG islands, also plays an important role in carcinogenesis. WGS survey revealed that approximately 25% of AMLs carry inactivating mutations in DNMT3 A, an enzyme that catalyzes the addition of methyl groups to CpG dinucleotides^{[ 65]}. In AML with mutated DNMT3 A, the promoter region of many genes involved in cancer shows decreased methylation. The newly identified ten/eleven translocation ( TET) DNA hydroxylase converts 5-methycytosine to 5-hydroxylmethylcytosine at CpG islands. Mutations in TET2 have also been found in acute myelogenous leukemia (AML), myelodysplastic syndrome (MDS) and myeloproliferative neoplasms^{[ 66]}. These epigenomic changes affect multiple target genes simultaneously, providing an efficient mechanism to alter cellular processes and induce cancer formation.

4 .3 Genomic stability

Mutations of genes involved in monitoring and maintaining the integrity of genome allow cells to evade normal apoptosis and continue replication despite the presence of extensive DNA damage. TP53 is probably the best example and is mutated in a variety of cancers including lung, endometrial, ovarian and breast. Mutations in double-strand break DNA repair gene ATM are seen in familial pancreatic and sporadic breast cancers. More recently, somatic mutation in the portion of POLE encoding the exonuclease domain was found in endometrial and colorectal cancers with unusually high mutation rates (>100 per Mbp)^{[ 67, 68]}. POLE is an enzyme responsible for the synthesis of the leading strand during DNA replication, and its exonuclease domain with proofreading capability is critical for high-fidelity copying of the DNA template during the S-phase of the cell division cycle.

The telomeric ends of chromosomes gradually shorten over the normal life span, leading to cell senescence or death. Many cancer cells overexpress telomerase encoded by TERT to maintain the long telomeres and their replicative potential. The mechanism was unclear until recently that two studies reported somatic mutations in the TERT promoter region in a variety of human cancers and a large number of cancer cell lines^{[ 69, 70, 71]}. The majority of these mutations occur at two positions in the TERT promoter region, resulting in generation of a de novo consensus binding motif for the EST transcription factors and upregulation of TERT expression^{[ 62]}.

4 .4 RNA splicing machinery

Another important discovery by cancer genomics is the deregulation of mRNA processing in cancer cells. Mutations in genes that encode the RNA splicing machinery, such as U2 AF1, ZRSR2, SRSF2 and SF3 B1, were first noted in chronic lymphocytic leukemia and myelodysplastic syndrome^{[ 72, 73, 74]}. Similar mutations were subsequently found in solid tumors including lung, breast and pancreatic cancers^{[ 75, 76]}. It has been postulated that aberrant RNA splicing may contribute to tumorigenesis by altering the expression of wild-type protein targets.

4 .5 Metabolism

A surprise discovery by cancer genomics is the identification of IDH1 and IDH2 mutations in brain tumors and acute myeloid leukemia^{[ 77, 78, 79]}. Both IDH1 and IDH2 convert isocitrate to α-ketoglutarate (KG) in the tricarboxylic acid cycle. Mutated IDH1/2 lose their normal function to produce α-KG, and gain a new activity producing D2-hydroxylglutarate. This "oncometabolite" is a potent inhibitor of many α-KG-dependent enzymes, such as TET family DNA demethylases, KDM-family histone demethylases, KMT family methyltransferase, and many other enzymes. Inhibition of these enzymes results in aberrant epigenomic modification as well as potential deregulation of other important cellular processes.

Many additional pathways or cellular processes and cancer genes have been revealed by cancer genomics, illustrating the power of NGS in cancer gene discovery^{[ 2, 4, 16, 37, 47, 48, 52, 53, 55, 62, 76, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89]}. The challenge now is to understand the biological significance of these cancer genes and to develop new effective small molecule inhibitors targeting the mutated cancer gene products for clinical applications^{[ 75, 90, 91, 92]}.

4 .6 Other new features of the human cancer genome

NGS has also revealed many new features of the human cancer genome. Mutation rates vary drastically between tumors of different origins^{[ 4, 16, 62, 76, 82, 93]}. The mutation rate in chronic lymphocytic leukemia is about one per Mbp, while the mutation rate in melanomas is about 15 per Mbp. In general, lower mutation rates are observed in hematological tumors, sarcomas and pediatric cancers than in solid epithelial tumors. The highest mutation rates are seen in tumors associated with exposure to carcinogens, such as lung cancer. In addition, mutation rates vary tremendously within a cancer type, often owing to the degree of exposure to the environmental carcinogens or the particular genes mutated. For example, tumors with mutations in mismatch repair genes typically have the highest mutation rates.

The pattern of mutations also differs across cancer types^{[ 76, 82]}. This difference appears to reflect the underlying mechanisms of mutagenesis. Lung cancers have a high proportion of G to T transversions, which is attributable to exposure to polycyclic aromatic hydrocarbons from tobacco smoke. Melanomas have a high proportion of C to T transitions in dipyrimidines caused by ultraviolet-induced DNA damage and misrepair from sun exposure or tanning practice. Gastrointestinal cancers have a high frequency of transition mutations at the CpG dinucleotide islands that may be due to the increased methylation levels in these tumors. In addition, cervical, bladder, some head and neck as well as breast cancers frequently have mutations at cytosines of the TpC dinucleotides, characteristic of mutations caused by the APOBEC family of cytidine deaminases.

Large scale (whole chromosome or chromosome arm) and focal CNAs of chromosomes are frequent changes in human cancers^{[ 60]}. Many focal amplifications and deletions have been localized to peak regions containing a median of 6-7 genes by NGS. These focal events are important in tumorigenesis, but the driver genes remained to be assigned. Besides simple balanced chromosomal translocations, more complex chromosomal rearrangements have also been revealed by NGS, such as chromothripsis and chromoplexy. Chromothripsis or chromosomal shattering is a catastrophic phenomenon that produces dozens or even hundreds of rearrangements within one or two chromosomes^{[ 94]}. This process occurs in about 2-3% of human cancers, probably due to erroneous chromosome segregation during mitosis. It is a single event that plays an important role in initial tumor progression. Chromoplexy or chromosomal chains/weave or braid is a process whereby a closed chain of chromosomes is formed by copy neutral rearrangements that consist of 3 to more than 40 breakpoint junctions or deletion bridges distributed across multiple chromosomes^{[ 80, 95]}. Chromoplexy tends to occur at regions of chromatin that are transcriptionally active. It occurs more frequently than chromothripsis and plays an important role in later cancer evolution. These findings demonstrate that human cancer genome is much more complex than we have anticipated prior to NGS.

Genomic information has been widely incorporated into the daily practice of cancer management for decades. Philadelphia chromosome or chromosomal translocation t(9;22)(q34;q11) is pathognomonic of chronic myelogenous leukemia. Demonstration of t(15;17)(q22;q21) or PML-RARA fusion gene is essential to initiate all trans retinoic acid (ATRA) treatment for patients with acute promyelocytic leukemia. In the last decade or so, cancer genomics has significantly changed our view about cancer and the way we manage cancer patients. The ultimate goal of cancer genomics is personalized or precision cancer medicine, i.e., diagnostics and therapeutics tailored to the individual cancer patient based on the genomic information.

5 .1 Diagnosis, classification and prognostication

Histomorphology and immunohistochemistry are the mainstay to diagnose and classify human cancers. Conventional and molecular cytogenetics such as FISH and PCR also play an important role in confirming the morphologic diagnosis of cancers and their further subclassification. In fact, recurrent chromosomal translocations or rearrangements are required to define specific cancer subtypes in the most recent WHO classification of human cancers. A few cancer genes discovered by cancer genomics have also been incorporated into our current cancer classification scheme. For example, JAK2 V617F mutation is a major diagnostic criterion for polycythemia vera. NPM1 and CEBPA mutations define a new provisional category of AML with normal cytogenetics. The BRAF V600E mutation is present in almost all cases of hairy cell leukemia, making it a useful diagnostic marker for this disease entity^{[ 55, 59]}. A dramatic example of diagnostic application of NGS has been the identification of cryptic PML-RARA fusion gene from a patient with clinically suspected acute promyelocytic leukemia, while conventional cytogenetics and other molecular tests failed to demonstrate the presence of PML-RARA fusion gene or transcript. Although it took a few weeks to complete, the result eventually spared the patient from undergoing more aggressive therapy such as allogeneic bone marrow transplantation which is associated with a higher mortality rate^{[ 96]}. In addition, NGS may be used to determine the origin of metastatic carcinoma of unknown primary^{[ 97]}.

Chromosomal abnormalities remain the strongest prognostic predictor in hematolymphoid as well as solid tumors. Integration of somatic mutations identified by cancer genomics has shown promise in creating more sophisticated prognostic models. For example, hypermutation of immunoglobulin heavy chain gene is associated with a better prognosis, while SF3 B1 mutation is associated with more rapid disease progression and lower overall survival in chronic lymphocytic leukemia^{[ 73]}. DNMT3 A mutation in AML with normal cytogenetics correlates with poorer overall survival^{[ 65]}. The status of internal tandem duplication in FLT3 and mutations in NPM1, MLL and CEBPA genes help stratify patient treatment and predict treatment outcome.

Recent comprehensive analysis of genomic abnormalities in human cancers has revealed that a subset of targetable mutations in driver cancer genes, CNAs and rearrangements are common across different human cancers^{[ 47, 60, 76, 81]}. These findings have led to the suggestion that human cancers may be more logically classified according to a particular genomic profile rather than their anatomic site for better and more efficient management of cancers with similar genomic alterations^{[ 98]}, but further validation is required.

5 .2 Treatment, response and relapse

For years, surgical resection followed by chemotherapy is the standard care for solid tumors, while patients with hematolymphoid neoplasm generally receive chemotherapy with or without bone marrow transplantation. In 1998, the U.S. Food and Drug Administration approved the use of anti-human EGFR antibody trastuzumab to treat Her2/ neu-positive breast cancer. This was followed by development of many small-molecule inhibitors and monoclonal antibodies against druggable driver cancer genes and a new paradigm of targeted therapy. Imatinib and other second generation tyrosine kinase inhibitors have been successfully used to treat chronic myelogenous leukemia. EGFR tyrosine kinase inhibitors erlotinib or gefitinib are effective for non-small cell lung cancer with activating EGFR mutations. Though short-lived, vemurafenib induces remission of metastatic melanoma with BRAF V600E mutation. ALK-positive advanced non-small cell lung cancer is responsive to ALK inhibitor crizotinib, but not to erlobinib or gefitinib. More recently, jakafi has been approved to treat primary myelofibrosis. Hundreds of small molecules in the pipeline are being tested in various clinical trials and some of them are expected to be approved in the near future^{[ 8, 61, 90, 91, 98, 99]}.However, the success of these therapies depends on the identification of the mutated cancer genes or pathways, and cancer genomics has and will play an indispensible role in this therapeutic decision-making process.

One of major obstacles in targeted therapy of human cancers is the development of acquired drug resistance. For most targets, acquired resistance usually occurs due to secondary mutations within the target gene, such as T315I mutation in BCR-ABL1 that is resistant to all currently available kinase inhibitors for chronic myelogenous leukemia. Intratumoral genomic heterogeneity or secondary mutation within the non-target cancer genes also plays a role in acquired resistance. Colorectal and non-small cell lung cancers with wild-type KRAS are sensitive to anti-EGFR antibodies, but they always develop resistance within several months after initiation of therapy due to the emergence of KRAS mutation. It is likely that the resistant clone is present in a small subpopulation of cells within the tumor prior to initiation of therapy. In contrast, resistance to vemurafenib treatment for metastatic melanoma is due to a new mutation in MEK1 that was not present at diagnosis. Identification of the resistant subclone at diagnosis or during therapy by NGS may predict whether the patient will relapse and when combination therapy will be initiated^{[ 99, 100, 101]}.

More recently, NGS of plasma cell-free tumor DNA (ctDNA) has emerged as a promising tool to monitor tumor burden, treatment response and development of drug resistance^{[ 101, 102, 103, 104, 105]}. Leary et al showed that WGS of ctDNA from patients with colorectal and breast cancer can detect genomic abnormalities specific to patients. Dawson and colleagues demonstrated that WGS or targeted sequencing of ctDNA has higher sensitivity and specificity than other tumor biomarkers to monitor tumor burden and treatment response in patients with metastatic breast cancer. Analysis of ctDNA by NGS can also detect emergence of drug-resistant subpopulation that will allow initiation of combination therapy to delay or prevent disease progression in patients with breast and colorectal cancers. Further studies are needed to determine the performance characteristics and its clinical applications.

The advent of NGS technologies has dramatically increased our understanding of basic cancer biology. With the decreasing cost and availability of benchtop sequencers, NGS has already been adopted by many clinical laboratories to explore mutations of actionable and druggable driver cancer genes to guide treatment and to monitor treatment response of human cancers. In the next decade or so, completion of the mutational atlas and creation of a functional encyclopedia of altered pathways in human cancer will make personalized cancer medicine truly a reality.

译 文

在发展中国家,癌症是死亡的一个主要原因。在许多发达国家,包括美国,癌症已经代替心脏疾病,成为死亡的首要原因。世界范围内因癌症而死亡的人数预计将从2008年的760万人上升至2030年的1 150万人 (http://www.who.int)。很明显,癌症已经成为一个全球性的公共卫生问题。

我们对癌症的抗争在一定程度上已经取得了巨大进展。目前认为,癌症是一种基因组疾病^{[ 1, 2, 3, 4]}。基因组的改变或修改会影响关键细胞信号的通路与程序,导致细胞生长失控、局部侵入以及癌症所特有的最终远处转移^{[ 5]} 。1982年,人们发现 HRAS致癌基因中的单点突变与膀胱癌有一定关联^{[ 6]}。根据此项发现,人们对许多致癌基因与肿瘤抑制基因进行了确定及表征^{[ 7]}。随着癌症基因同源蛋白相应抑制药物的发展,以上的这些发现都对癌症的治疗有着翻天覆地的影响^{[ 8, 9]}。

早在10年前,人们就已经完成了人类基因组参考序列的研究。近几年,高通量测序技术的出现使癌症基因组学的研究进入了一个新纪元^{[ 9, 10, 11, 12, 13, 14, 15, 16, 17]}。许多新型癌症基因或通路的发现,揭示了一些新型基因组的特点,从而对一些新型药物进行了开发^{[ 5, 14, 18, 19]}。这篇综述将对目前基因组分析的技术革新进行介绍,并着重对下一代基因测序的相关内容进行描述,并对人类癌症基因组的新发现进行总结,对这些发现在人类癌症管理中的影响进行简单讨论。

在一个正常的个体中,单倍体人类基因组大约由 30 亿个核苷酸DNA 组成,二倍体核中总共有 60 亿个碱基对,不均匀地分布在22 对常染色体和 1 对性染色体中^{[ 20]}。仅有大约2%的人类基因组(大约23 000个基因)针对蛋白进行编码。仅有不到1%的基因组对功能性RNA进行编码,诸如microRNA、核糖体RNA、长链非编码RNA以及其他种类的功能性RNA^{[ 21, 22]}。有8%的基因组在进化过程中高度保守,这些序列的其中一部分,诸如启动子、增强子和位点控制区域是重要的调控元件。剩余90%的基因组不对任何蛋白或功能性RNA进行编码,这类序列大部分是重复子,诸如短散布(SINE)或长散布(LINE)元件。它们可以产生串联重复簇或散布重复簇。除了着丝粒和端粒,大部分功能还未被人们所了解,其分别涉及到染色体的包装、分离、复制以及染色体末端的维护。这些通常被称之为“垃圾”DNA或“暗物质”。

不是所有23 000多种人类基因组里的基因都能被转录。这类基因的表达谱或转录物组主要取决于细胞的类型与分化的阶段。在给定的时间与给定的细胞中,基因组中仅有一小部分代码基因被转录,从而产生大约300 000个mRNA分子^{[ 23]}。相比之下,外显子组是指所有的基因组蛋白编码外显子与RNA基因DNA序列的总和。不管出于分化的哪个阶段,在所给定的有机体中,其对所有类型细胞来说,外显子组都是一样的。

人类的每一个基因组都是独一无二的,在不同的个体中都有着巨大的差异。这种变异包括单一碱基对的变化[单核苷酸多态性/变异(SNP/V)]以及核苷酸序列范围从几个到几百个或几百万个碱基对长度上的差异[拷贝数变异(CNV)]^{[ 24, 25, 26, 27]}。事实上,在我们的基因组中,每300个碱基就会有1个是多态的情况发生。2个个体会在>300万个SNP/V位点上存在差异。这些变异常在人类基因组编码与非编码区域中发生。因此,当我们在试图分析并解释癌症基因组变化的时候,应当熟记这些正常的变异情况,这一点对于相关研究是非常重要的。

我们常常运用常规的细胞遗传学或染色体分析方法来分析人类基因组。这是人类基因组分析中最早,并且是运用最为广泛的一项技术。在癌症基因组学领域,Peter Nowell与David Hungerford在1960年运用此方法首次指出,费城染色体是慢性粒细胞性白血病重复性染色体异常的情况^{[ 28]}。随着20世纪70年代染色体显带技术的发展,Janet Rowley指出,t(9;22)(q34;q11) 转移是慢性粒细胞性白血病费城染色体产生的原因^{[ 29]}。通过传统细胞遗传学,我们还发现许多其他具有诊断、预后以及治疗意义的染色体异常情况。这些异常情况已经通过其他一些分子生物学技术,如聚合酶链反应(PCR)与Sanger测序进行表征和确定,并成为目前世界卫生组织(WHO)人类癌症病理及遗传学分类的基础。虽然常规的染色体分析可以覆盖整个基因组,但是这一技术受到低分辨率(大约15~30 Mbp)的限制,仅仅可用于评估基因组与染色体转移或重组大部分区域的获得或缺失。为了避免这一限制,20世纪90年代对荧光原位杂交技术(FISH)以及比较基因组杂交进行了开发,其达到了大约每DNA探针50~100 Kbp的分辨率^{[ 30]}。然而,每个探针仅能检测0.01%基因组以及荧光信号光谱的重叠,这使得这一技术很难对整个基因组进行分析。

许多芯片或微阵列平台,如细菌人工染色体比较基因组杂交、寡核苷酸的SNP阵列,在最近10年中有一定程度的开发^{[ 31]}。这类技术在本质上能够通过使用数以万计特异的DNA序列探针检测整个基因组的CNV和等位失衡,分辨率可以达到<100 Kbp。除了对整个基因组的测序,SNP阵列是唯一能检测杂合性拷贝无特征缺失的平台(CN-LOH;在这一过程中,染色体的部分缺失从姐妹染色单体中被复制)。虽然最新的阵列平台可以用来检测突变以及小型插入/缺失(得失位)。但是新型的SNP/V不能够通过这些已知基因组位点的预制芯片来进行分析。为了互补这一阵列平台,多重PCR同时得以开发,其可用来检测许多不同的癌症基因中多重已知的热点突变^{[ 18, 32, 33, 34, 35]}。Applied Biosystems中的SNaPshot平台由多重PCR与荧光双脱氧三核苷酸单碱基延伸反应组成,可以用来研究8~14个关键癌症基因,如 RAS、 BRAF、 AKT1、 EGFR、 PIK3 CA、 MEK1、 PTEN、 IDH1与 IDH2的50个热点突变。在不同的扩增子中,多达10个单一突变可以在单碱基延伸反应中进行检测。SNaPshot平台也常包括最为常见的 EGFR基因外显子20插入与外显子19缺失突变。SNaPshot平台的产物通过ABI Genetic Analyzers毛细管电泳分离并进行分析。相比之下,Sequenom平台使用基质辅助激光解吸/电离飞行时间质谱技术,可快速分析多重PCR产物。Sequenom OncoCarta V1.0试剂盒可以用于分析19个不同的致癌基因中238个体细胞癌症基因突变,其中包括所有SNaPshot平台检测的突变。

针对所有已知与未知的异常类型对整个基因组进行研究,在分辨率为单一核苷酸水平的情况下,最为理想的检测技术就是DNA测序。由Fred Sanger在20世纪70年代研究开发的传统双脱氧链终止法能够读取一个单一反应中大约1 000个核苷酸,这一大小小于人类基因组的一百万分之一^{[ 36]}。即使检测方法在不断地改进,并向自动化方向发展,科学家们仍然用了大约10年时间,花费了27亿美金,才完成了对第1份人类参考基因组的测序^{[ 9, 12, 13]}。很显然,Sanger测序仅仅只能有效地、经济地、合理地用于分析已知热点突变,而不能用于整个人类基因组的测序。最近5年已出现了下一代测序或大规模平行测序(NGS或MPS)技术^{[ 10, 11, 17, 37, 38, 39]}。这些技术随着电脑信息化的发展已被许多实验室用于人类癌症与其他先天性疾病异常情况的大型基因组区域或整个人类基因组的研究中。目前,仅仅只需要几天或是几周时间,花费几千美元,就能对部分序列或是整个人类基因组进行研究。首个人类癌症基因组序列分析已在2008年完成。到目前为止,在广大工作人员及机构,如the Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov/)、the International Cancer Genome Consortium (ICGC) (http://icgc.org/)的共同努力下,人们已运用NGS技术对来自超过25个不同癌症类型的900多个基因组进行了测序。

NGS技术可以实现数以万计基因组DNA片段同时大规模的平行测序。目前已有几个不同的商业化NGS平台^{[ 10, 11, 37, 38]}。在焦磷酸测序反应这一原则的基础上,Roche 454 GS FLX Titanium与Junior systems (Roche Applied Sciences, Penzberg, Germany)使用的是边测序边合成策略。Life Technology ABI/5500 SOLiDTM系统(Life Technologies, Carlsbad, CA)使用的是边测序边连接技术,其中涉及到寡核苷酸连接延伸的反复反应。而近期的Life Technology Ion Torrent personal genome machine (PGM)与Proton系统应用的是边测序边合成策略。该检测系统运用了超敏感pH计,可以检测在DNA合成期间核苷酸结合之后释放的氢离子。Illumina HiSeq2000/2500与MiSeq系统(Illumina, San Diego, CA)也同样运用的是边测序边合成的策略,其需要与可逆性染料末端终止物和成像配合使用。其他较新的平台,如BioSciences HeliScope的Helicos、Pacific Biosciences and Oxford Nanopore Technologies的SMRT都是运用实时直接测序方法对单一分子在没有扩增条件的情况下进行测序。不管使用的是哪一种平台,NGS技术的工作流程都涉及到不同化学反应、碱基识别、序列比对与变异体识别等。

NGS技术可用于检测整个基因组[整个基因组测序(whole genome sequencing, WGS)]、外显子组[整个外显子组测序(whole exome sequencing, WES)]或是转录物组[整个转录物组测序(whole transcriptome sequencing, WTS或RNA-Seq)]的序列^{[ 17]}。WGS可以对所有异常类型的整个基因组,包括复杂基因组重组进行研究。然而,由于超过90%经WGS确定的异常情况都具有未知的生物学或临床意义,因此WGS是目前研究新型癌症基因、新型机制最好的方法。WES仅以外显子组开始。随着越来越多先进捕获技术的发展,WES可覆盖基因组的范围更加广泛,其中包括所有的编码外显子、microRNA基因、非翻译区域、新发现未注释转录基因以及其它功能性RNA基因。这要比WGS更为经济,所需要的时间也较少,并且能够覆盖更大的范围。然而,WES并不能检测染色体的重组或融合基因。在不久的将来,WES与WTS的联合运用不仅可以提供有关基因表达的大量信息,还可成为研究人类癌症基因组用于临床的一个最佳实际操作策略。目前,可以替代大部分临床实验室WGS/WES/WTS的一个选择是对特定的癌症有预后或预测意义的一组基因进行深度测序,寻找突变^{[ 40, 41, 42, 43, 44, 45, 46]}。例如the Washington University Cancer Mutation Profiling (WUCaMP)试验对25个癌症基因进行检测,运用454 GS Junior NGS平台对临床可操作的体细胞基因突变进行了检测^{[ 45]}。The University of Washington UW-OncoPlex试验对194个临床相关基因,运用Illumina HiSeq2000平台对体细胞基因突变以及其他异常情况,包括小型插入缺失、内部串联重复、基因拷贝数的获得与缺失以及染色体的重组进行研究^{[ 44]}。Foundation Medicine的Frampton等人对Illumina HiSeq 2000平台的临床癌症基因组谱进行了研究与确认^{[ 43]}。这些试验结果显示,在287个与癌症有关的基因中,有4 557个外显子的检测具有高敏感性(95%~99%基因组改变)与特异性(阳性预测值>99%)。靶向测序的不足之处是在设计自定义基因组时具有相对较高的初始成本,并且添加其它有关基因至基因组过程中的费用较高。

我们已经对许多人类癌症基因组的获得性基因组异常进行了确定^{[ 1, 2, 4, 6, 16, 17, 47, 48]}。这些异常包括单核苷酸的改变或点突变、小型得失位、获得或缺失的拷贝数变化(CNA)以及相关结构异常。在基因编码区域内的单核苷酸变化会导致错义突变(氨基酸代替)、无义突变(基因产物断裂)或剪接位点突变。小型得失位可能会导致基因阅读框的分裂(失帧得失位)。CNA可能会涉及到单一基因或多重基因,导致受影响基因的异常表达。结构异常包括染色体间与染色体内的转移或重组,从而生成具有不同功能性的融合基因。获得性基因组异常及癌症基因的活化是癌病变与癌症发展的基础。

许多传统癌症基因对影响基因组稳定性及细胞分裂、增殖或凋亡的蛋白进行编码^{[ 4, 5]}。相比之下,许多大规模检测及WGS检测所发现的新型癌症基因会影响信号转导通路、表观遗传调控、RNA剪接及代谢等过程(见表1,另可参阅http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/上的详细内容)。在此,对通路或细胞过程以及近10年通过癌症基因组学发现的新型癌症基因进行简要的概述。

4.1 信号转导通路

信号转导通路在正常细胞增殖与生存中起了重要的作用。在20世纪90年代,人们在受体蛋白酪氨酸激酶的关键基因编码与许多其他癌症中的信号通路中发现突变,如乳腺癌中的HER-2、肠胃肿瘤中的c-KIT以及慢性髓细胞性白血病中的ABL^{[ 4, 49]}。小型分子抑制剂和/或以这些突变蛋白为靶向的分子抗体,如伊马替尼与曲妥单抗等,已有一定程度的发展并成功用于治疗具有突变蛋白癌症的病患中。由于这些成功的案例,在21世纪早期,临床实验室已运用Sanger方法对许多患者癌症基因进行测序,如非小细胞肺癌中 EGFR、黑色素瘤中 BRAF V600E、乳腺癌、直肠癌以及子宫内膜癌中 PIK3 CA、骨髓增生性肿瘤中 JAK2 V617F的突变情况^{[ 1, 2, 4, 16, 50, 51]}。其中一些发现已经对相关药物的发展与临床治疗产生了一定的影响,如运用埃罗替尼/吉非替尼治疗非小细胞肺癌,运用威罗菲尼治疗转移性黑色素瘤,运用磷酸鲁索替尼jakafi治疗骨髓纤维性病变。

近几年NGS技术得到了广泛使用,人们运用该技术发现了若干信号转导通路涉及的基因新型重复突变,其中一些通路事先并未被预测到与癌症会存在一定关联(见表1)。例如 MAP3K1与 MPA2K4基因,在MAPK信号通路中对丝氨酸/苏氨酸激酶编码,在乳腺癌中存在突变^{[ 52]};RAC/PAK信号通路的 RAC1、 ELMO1与 DOCK2基因分别在黑色素瘤与食道癌中发生突变^{[ 53, 54]};NF-κB信号通路中的 NYD88基因在弥慢性大B细胞淋巴瘤中发生突变^{[ 55]};在大部分血管免疫母T细胞淋巴瘤中会发生 RHOA突变^{[ 56]};出人意料的是,神经元轴突导向的 ROBO和 SLIT基因会在20%左右的胰腺癌中发生突变^{[ 54, 57]}。不久前,人们发现在 JAK2与 MPL突变阴性的骨髓增生瘤中,钙结合蛋白的钙网织蛋白( CALR)基因重复突变^{[ 58]}。这些研究都证实了这些传统癌症基因在新型癌症中的作用。例如,常见的黑色素瘤 BRAF V600E突变也会存在于多毛细胞白血病中^{[ 55, 59]}。

除了点突变或小型得失位之外,CNA与基因组重组也同样能够活化癌症基因^{[ 60]}。人们发现 MCL1与 BCL2L1编码的抗细胞凋亡蛋白可以在许多癌症中被扩增,包括乳腺癌、肺癌、直肠癌及黑色素瘤。有很多有关非小细胞肺癌中涉及 ALK、 RET与 ROS1重组的报道。 BRAF重组也同样在小儿毛细胞型星形细胞瘤与恶性黑色素瘤中有所发现。

表1

表1

表1 癌症基因组学所发现的通路与新型癌症基因 通路或程序 新型癌症基因举例 代表性癌症类型 信号转导 受体酪氨酸激酶信号 EGFR、 ERBB2、 MET、 ALK、 JAK2、 RET、 ROS、 FGFR1、 FGFR2、 PDGFRA、 CRKL、 CALR、 RHOA 乳腺癌、脑癌、肺癌、骨髓增生瘤、T细胞淋巴瘤 MAPK信号 KRAS、 NRAS、 BRAF、 MAP2 K1、 MAP3 K1、 MAP2 K4、 NF1 乳腺癌、肺癌、黑色素癌、多毛细胞白血病 PI3K信号 PIK3 CA、 AKT1、 AKT3、 PTEN、 PIK3 R1 乳腺癌、脑癌、结肠癌、子宫内膜癌 Notch信号 NOTCH1、 NOTCH2、 NOTCH3 头颈部鳞状细胞癌、T细胞ALL mTOR信号 STK11、 TSC1、 TSC2 结肠癌、膀胱癌 Wnt/β-catelin信号 APC、 CTNNB1 胃肠癌 TGFβ信号 SMAD2、 SMAD4、 TGFBR2 结肠癌 NF-κB信号 MYD88 弥漫性大B细胞淋巴瘤 SLIT/ROBO信号 ROBO1、 ROBO2、 SLIT2 胰腺癌 KEAP1/NRF1信号 KEAP1 肺癌 脑信号蛋白信号 SEMA3 A、 SEMA3 E、 SEMA5 A 胰腺癌 RAC/PAK信号 RAC1、 RAC2、 CDC42、 ELMO1、 DOCK2、 PREX2 黑色素癌、食管癌 表观基因组学调控 DNA甲基化 DNMT3 A AML、MDS DNA羟甲基化 TET2 AML、MDS 组蛋白(赖氨酸)转甲基 酶(KMT) EZH2、 MLL、 MLL2、 MLL3、 NSD1、 NSD3、 SETD2 AML、MDS、肺癌、胃癌、前列腺癌、大B细胞淋巴瘤 组蛋白(赖氨酸)脱甲基 酶(KDM) JARID1 A、 UTX、 KDM5 A、 KDM5 C、 KDM6 A 骨髓瘤、肾脏细胞癌、AML等 组蛋白乙酰转移酶 CREBP、 EP300 弥漫性大B细胞淋巴瘤 组蛋白去泛素化酶 BAP1 间皮瘤、肾脏透明细胞癌 SWI/SNF染色质复合物 PBRM1、 SMARCA1、 SMARCA4、 ARID1 A、 ARID1 B、 ARID2 黑色素癌、肾脏与卵巢透明细胞癌、肝细胞癌与胰腺癌 染色质压缩 CHD1、 CHD2、 CHD3、 CHD4 CLL、前列腺与子宫内膜癌 RNA接合体系 SF3 A1、 SF3 B1、 SF1、 SFRS1、 SFRS7、 ZRSR2、 SRSF2、 PRPF40 B、 U2 AF1、 U2 AF2、 U2 AF65 MDS、CLL、肺癌、乳腺癌、胰腺癌 代谢 IDH1、 IDH2 脑癌、AML与其他 基因组稳定性 基因组完整性 TP53、 MDM2、 MSH、 MLH、 ATM、 POLE 黑色素癌、膀胱癌、肝细胞癌与其他 端粒长度 TERT启动子突变 黑色素癌、膀胱癌、肝细胞癌与其他 转录因子 世系依赖性 MITF、 NKX2 -1、 SOX-2、 EGR、 ETV1、 CDX2 肺癌、食管癌 其他 MYC、 RUNX1、 GATA3、 FOXA1、 NKX3 .1、 SOX9、 NEF2 L2、 MED12 乳腺癌、肺癌、前列腺癌 蛋白质泛素化 SPOP、 PBXW7、 WWP1 子宫内膜癌、头颈部癌、膀胱癌、胃肠癌 蛋白质翻译 FAM46 C 骨髓瘤 蛋白质内平衡 XBP1 骨髓瘤 RNA充足 DIS3 骨髓瘤 细胞周期 CCND1、 CCNE1、 CDKN2 A、 CDKN2 B、 CDKN1 B 子宫浆液细胞 凋亡 MCL1、 BCL2 A1、 BCL2 L1 乳腺癌、肺癌、脑癌、结肠癌、黑色素癌

表1 癌症基因组学所发现的通路与新型癌症基因

基因组信息已被广泛纳入癌症管理工作几十年了。费城染色体或染色体转移t(9;22)(q34;q11)是慢性髓细胞性白血病的一种特异表现形式。t(15;17)(q22;q21)或 PML- RARA融合基因的出现对急性早幼粒细胞白血病全反式维甲酸(ATRA)的治疗是很有必要的。在最近的10年中,癌症基因组学已经极大地改变了我们对癌症的看法以及管理癌症患者的方式。癌症基因组学最基本的目的是使癌症医学个体化或精确化,如根据每个癌症患者的基因组信息制定个人的诊断及治疗方案。

NGS技术的出现大大增加了我们对基础癌症生物学的认识。随着台式测序仪成本的降低和可用性的提高,NGS已被许多临床实验室用于研究可行的可药物治疗的驱动癌症基因突变,从而来指导相关治疗,并监测相应癌症的治疗反应。在未来的10年左右,人类癌症基因突变图谱和通路改变的功能性总结的完成将使个体化癌症医学真正地成为现实。

(收稿日期:2014-01-22)

(本文编辑:董悦颖)