Notch1 Mutations Are Drivers of Oral Tumorigenesis Free

Head and neck squamous cell carcinoma (HNSCC), accounts for 650,000 new cases worldwide (1–3) with more than 12% distributed in China (4). Oral squamous cell carcinoma (OSCC), the most common subtype of HNSCC, is notorious for poor prognosis (5), which reflects the propensity of OSCC to present as clinically advanced disease upon diagnosis (6, 7). Unfortunately, most patients in China are already in advanced stages when diagnosed and more than 76,000 die each year (4).

Nearly 20% of patients with OSCC harbor multiple premalignant lesions, often identified as leukoplakia (8). It has been postulated that leukoplakia represents an early stage in OSCC, as some lesions evolve to malignant neoplasms (8, 9). Although efforts have been made to predict malignant transformation in oral leukoplakia (10–12), no reliable markers of clinical utility have been identified. OSCC is thought to progress through a series of genetic alterations. Indeed, several signaling pathways are dysregulated in OSCC through genetic and epigenetic alterations, such as those involving TP53, NOTCH1, CDKN2A, CCND1, HRAS, PIK3CA, and FAT1 (13–15). NOTCH1 is particularly noteworthy. In Caucasians, potentially inactivating mutations occurred in 11% to 15% of tumors, and as such, NOTCH1 is the second most frequently mutated gene in HNSCC after TP53 (13–15). Recently, Song and colleagues (16) have shown that in Chinese patients with OSCC, the NOTCH1 mutation frequency was greater than 40% and strongly associated with poor prognosis and shorter survival. Although these data indicate that disruption of NOTCH1 signaling is involved in oral tumorigenesis of both Asian/Chinese and Caucasian populations, this study was performed with conventional PCR to manually amplify NOTCH1 through hundreds of individual reactions and the role of NOTCH1 in malignant transformation of oral leukoplakia was not addressed.

Using new enrichment technology and next-generation sequencing (NGS), we assessed NOTCH1 mutation status at different stages of OSCC progression in Chinese patients. The NOTCH1 mutation frequency was 54% for OSCC and 60% for preneoplastic lesions. Importantly, most patients with leukoplakia with mutated NOTCH1 carried mutations that were also identified in OSCC, indicating an important role of these events in the progression of early neoplasms. Moreover, we compared all known NOTCH1 mutations in Chinese patients with OSCC with those reported in Caucasians to date. Although we found obvious overlaps in critical regulatory domains alterations and identified mutations shared by both cohorts, possible gain-of-function mutations were predominantly seen in the Asian population.

One hundred forty-four tissue samples were collected at the Ninth People's Hospital (Shanghai, China). Forty-nine samples were normal oral mucosa from patients undergoing oral surgery and 45 represented oral leukoplakia biopsies. Fifty OSCC samples were obtained during resection. Among them, 22 paired normal tissues were gained from the adjacent areas at least 1 cm away from cancer. Samples were promptly frozen at −80°C after initial pathologic examination. Frozen tissues were cut into 5-μm sections, stained with H&E, and examined by light microscopy. Lesions with a low neoplastic cellularity (<70%) were additionally microdissected to remove contaminating normal cells before DNA extraction. Patients with OSCC had not been treated with chemotherapy or radiotherapy before their tumor biopsy, so the spectrum of changes we observed largely reflects those of tumors in their naturally occurring state. The histopathologic diagnosis was made by a pathologist on duty according to the World Health Organization criteria (17, 18). Clinical characteristics are shown in Supplementary Tables S1–S4. This study was approved by the Human Research Ethics Committee of Shanghai Jiaotong University and the Administration Office of the Chinese Human Genetic Resources. Informed written consent was obtained from all patients before sampling.

Genomic DNA was isolated from fresh frozen samples by the QIAamp DNA Kit (Qiagen) and quantified with the Nanodrop system (Thermo Scientific).

Primers pairs (108) were designed by Fluidigm to cover all 34 exons of the NOTCH1 and exon–intron boundaries (Supplementary Table S5). PCR amplification was performed using a Fluidigm Access-Array microfluidic chip according to the manufacturer's instructions. Each sample was combined with primer pairs in a microfluidic chip, and thermal cycling on a Fluidigm FC1 Cycler was performed. PCR products were then collected using the integrated fluidic circuit (IFC) controller and transferred to a 96-well plate. In a separate PCR, Illumina sequence–specific adaptors and barcodes were attached.

Pooled and indexed PCR products were sequenced on the Illumina MiSeq instrument following standard protocols with the following modifications: Illumina-specific sequencing primers were substituted with a mixture of two Fluidigm-specific primers pairs (FL1 and FL2). The average coverage of each base in the targeted regions was 668-fold, and 97% of targeted bases were represented by at least 10 reads.

Bioinformatic analyses were performed at PGDx. The sequences were aligned to the human genome reference sequence (hg18) using the ELAND algorithm of CASAVA 1.7 software (Illumina). The chastity filter of the BaseCall software of Illumina was used to select sequence reads for the subsequent analysis. The ELANDv2 algorithm of CASAVA 1.7 software (Illumina) was then applied to identify point mutations and small insertions and deletions. Known polymorphisms recorded in dbSNP were removed from the analysis. Potential somatic mutations were filtered and visually inspected as described previously (19).

We first examined the NOTCH1 coding regions and intron–exon boundaries in 22 OSCC tumor–normal pairs (Supplementary Table S1B). Using stringent criteria, we identified 25 nonsynonymous aberrations (Fig. 1A and B and Supplementary Fig. S1A) in 11 (50%) tumors (Fig. 1C). Among these alterations, 24 were missense substitutions and 1 was a frameshift deletion. Twenty-three missense mutations were in the coding region and 1 was at a splice-acceptor site (Fig. 1B). The average mutation rate was 2.27 mutations per patient (Fig. 1B). No nonsense mutations or copy number variations were discovered in this cohort. EGF-like repeats was the most mutated region (Fig. 1D), followed by Ankyrin repeats (ANK) and the recombination signal sequence binding protein (RBP)–associated module (RAM), while fewer mutations were seen in the heterodimerization domain (HD) and Lin12-NOTCH repeats (LNR). We next sequenced NOTCH1 in 28 OSCC tumors without matched normal controls and 49 oral mucosa samples obtained from healthy individuals (Supplementary Table S2). The sequences were aligned to the reference, and all identified point mutations were checked against dbSNP database for known polymorphisms. Common polymorphisms were recorded in dbSNP, and all mutations identified in both OSCC and normal samples were removed from the analysis (Supplementary Tables S6–S8). The remaining mutations (Fig. 1E and Supplementary Fig. S1B) are most likely either somatic alterations or rare polymorphisms. Of 22 nonsynonymous alterations, 21 were single-nucleotide substitutions and 1 was a frameshift mutation. Eighteen point mutations, including 1 nonsense, were at the coding region and 3 at splice sites (Fig. 1F). NOTCH1 mutations were found in 16 (57%) of these patients (Fig. 1G), which was only slightly higher than the mutated NOTCH1 prevalence identified in 22 matched tumor–normal samples, confirming the success of our conservative filtering approach (Fig. 1C). The small difference may be explained by chance alone or by the presence of a few uncommon SNPs alongside somatic alterations. Because mutation distribution in this group (Fig. 1E and H) closely resembled that found in the 22 tumor–normal pairs, we combined all nonsynonymous mutations identified in both groups into one “OSCC-tumors” cohort (Fig. 1I). Altogether, 47 nonsynonymous alterations in 27 (54%) patients were identified in the “OSCC-tumors” set (Fig. 1J and K). The mutation distribution across NOTCH1 functional domains is shown in Fig. 1L. Interestingly, most mutations in the frequently mutated exons (Fig. 1M) were located around the HD and transactivation domain (TAD)/region rich in proline (P), glutamine (E), serine (S) and threonine (T) residues (PEST) domains, where most of the activating mutations were reported in hematologic malignancies (Fig. 1I). Of 43 mutations in the NOTCH1 coding region, 11 mutations in the same amino acid were previously reported in various malignancies, including HNSCC (Fig. 1N), supporting their status as bona fide somatic mutations and suggesting a potential functional role in tumor progression.

To address the role of NOTCH1 in the transition from premalignant lesions to malignant neoplasms, we sequenced NOTCH1 in 45 Chinese patients with oral leukoplakia. This population represents a unique model to study this question, because the proportion of primary tumors harboring NOTCH1 mutations is much higher than in Caucasians. Patients were subgrouped as leukoplakia with low malignant potential (mild dysplasia) or high malignant potential (moderate or severe dysplasia; Supplementary Table S3). Because corresponding normal tissues were not available for these samples, we used the same approach as used for analysis of unmatched OSCC tumors. Thirty-nine nonsynonymous substitutions (Fig. 2A and B and Supplementary Fig. S2) were identified in 27 (60%) of 45 patients (Fig. 2C). Of 36 alterations in the coding region, 33 were missense and three nonsense substitutions, with an average mutation rate of 1.4 mutations per patient (Fig. 2B). Overall mutation prevalence and distribution across the gene in oral leukoplakia (Fig. 2D) were consistent with those found in patients with OSCC. Seven mutations affecting the same codon were previously reported in various malignancies and 32 were novel (Fig. 2E). Whereas NOTCH1 mutation status was not associated with higher tumor stage (Supplementary Fig. S3A), in patients with leukoplakia we noticed a slight progressive pattern in NOTCH1 mutations prevalence from mild to moderate and severe dysplastic lesions (44% and 69%, respectively; Fig. 2F). Although this difference was not significant (P = 0.08), the high frequency of NOTCH1 mutations in patients with mild dysplasia suggests that NOTCH1 mutagenesis is an early event in OSCC tumor progression.

There were several mutation “hotspots,” including 4 cases with mutation at codon 571D>A (8.9%), 3 patients with mutations at codon 1115L>P and mutations at codons 1531L>P and 517 P>L/S each in two tumors (Fig. 2A). Because of a lack of germline DNA, it is possible that some of these “hotspots” may be rare germline polymorphisms. However, a comparison between mutations found in leukoplakia and OSCC samples revealed that all 4 “hotspots” in leukoplakia were overlapping with mutations found in OSCC (Fig. 2G). Thus, it is reasonable to assume that these “hotspots” are most likely somatic aberrations. Surprisingly, of 27 patients with leukoplakia with mutated NOTCH1, 16 (59.3%) carried the same 9 mutations that were also seen in 10 (20%) of 50 patients with OSCC (Fig. 2G and H). This observation indicates that in some cases NOTCH1 mutagenesis may underlie a progression of oral premalignant lesions into the primary carcinoma. Furthermore, of 16 patients with leukoplakia with overlapping mutations, 12 were diagnosed with high-grade dysplasia, whereas only 4 patients had mild dysplasia upon diagnosis (Fig. 2I), suggesting an evolutionary advantage during tumor progression.

Smoking and excessive alcohol consumption are the major risk factors for OSCC (20). Notably, high alcohol-content liquor is traditionally consumed in China compared with predominantly wine or beer in Caucasians (16). Although exposure to tobacco and alcohol was associated with high frequency of TP53 mutations in HNSCC (20, 21), the association with NOTCH1 mutagenesis has not been reported. Because history of tobacco and alcohol consumption was not available for all patients in normal cohort, we compared the number of patients who either smoked, consumed alcohol, or both between the 50 patients with OSCC and 45 patients with oral leukoplakia. Although the percentage of smokers or alcohol consumers was not significantly different between the two cohorts, the percentage of patients with smoking and alcohol consumption history was higher among the patients with OSCC compared with patients with leukoplakia (28% and 13.3%, respectively; P = 0.07; Supplementary Fig. S3B). We next compared the amount of patients who either smoked, consumed alcohol, or both between the 27 patients with OSCC and 27 patients with leukoplakia harboring nonsynonymous mutations in NOTCH1. No significant difference in the percentage of patients with either smoking or alcohol consumption history was observed between the two groups, whereas a 3-fold increase in patients with tobacco and alcohol history was seen in OSCC, compared with patients with leukoplakia with similar background (11.1% and 33%, respectively; P = 0.05; Supplementary Fig. S3C). The most common base-pair substitutions in nonsmoking OSCC and leukoplakia patients were C–T (30.8% and 30.6%) and G–A (27% and 19.4%, respectively; Supplementary Fig. S3D). In patients with OSCC and leukoplakia with a smoking background, the A to G transition mutations were much more prominent in the preneoplastic lesions. Unfortunately, we were unable to determine the association between NOTCH1 mutation status and survival, as the follow-up time from tumor diagnosis until death was available for only 12 of 50 patients with OSCC.

A collective analysis of all patients (leukoplakia and OSCC) revealed the most common mutated NOTCH1 exons in oral lesions (Fig. 2J). Although the NOTCH1 mutation spectrum in oral lesions was fundamentally different from that seen in leukemia, 21 (27.3%) of 77 mutations found in this study were located around the HD domain (exons 26–28) and 8 (10.4%) were scattered throughout the TAD/PEST region (exon 34), where most gain-of-function mutations were reported in hematopoietic malignancies (22, 23). Consistent with previous reports (13, 15), 37 (48%) of all mutations found in oral lesions were in the EGF-like repeats. Interestingly, 15 (40.5%) of these mutations, including 2 “hotspots,” were within or around the “ligand-binding” domain (Fig. 2G), suggesting a loss-of-function potential for these mutations. These data suggest a dual biologic role for NOTCH1 as either a cancer-promoting or a tumor suppressor gene in OSCC.

To better understand the NOTCH1 mutation spectrum in Asian patients with OSCC, we combined mutations found in this study with nonsynonymous substitutions identified by Song and colleagues (Fig. 3A, top; Supplementary Table S9; ref. 16). Moreover, to evaluate the differences in NOTCH1 mutations spectrum between Chinese and Caucasian oral tumors, we gathered NOTCH1 nonsynonymous variants in Western population from all NGS articles known to date (13–15, 24) and the recently reported TCGA data (Fig. 3A, bottom; Supplementary Table S10). Our combined dataset contained 101 Chinese and 595 Caucasian patients. The HPV status in 50 Chinese patients was not defined, but HPV infection does not appear to be a significant risk factor for OSCC, and the prevalence of HPV-positive oral cavity cancers is relatively low (5.9% or lower; refs. 25, 26). Among Caucasian tumors, 93% were HPV-negative and more than 70% were OSCC (Supplementary Table S11). Although the mutation prevalence was slightly higher in Chinese patients with OSCC (Fig. 3B), the overall NOTCH1 mutation frequency was 3-fold greater than in Caucasians, 48.5% versus 15.8%, respectively (P = 0.0002; Fig. 3C). Interestingly, we found a disproportionate prevalence of nonsense (23.4% and 9.2%; P = 0.009) and frameshift (18.7% and 4.6%; P = 0.003) mutations in Caucasian versus Chinese patients, respectively (Fig. 3B).

Although EGF-like was the most mutated domain in Chinese (49.4%) and Western (77%) populations, the mutation distribution among NOTCH1 functional domains significantly varied (Fig. 3D and E). On the basis of the distribution spectrum, we identified three distinct functional clusters in Chinese and Caucasian patients (Fig. 3F). In Chinese tumors, cluster I partially spreads across the “ligand-binding” domain (exons 8–11) and contains 9 (11%) of 83 mutations. Cluster II overlaps with the Abruptex region (exons 19–23) and consists of 19 (23%) mutations, while cluster III covers the negative regulatory region (NRR) and the area at the boundary between the extracellular and transmembrane domains and consists of 36 (41.4%) mutations (Fig. 3F, top). By comparison, in Caucasian patients, 35 (40%) of 87 nonsynonymous substitutions are gathered into cluster I (exons 6–9; Fig. 3F, bottom). Cluster II (6 mutations) overlaps with the C-terminus of the Abruptex region, and 15 (17.3%) mutations (including 8 nonsense) are found in the RAM/ANK area of the NOTCH1 intracellular domain (NICD).

Among 87 single-nucleotide variants found in Caucasian patients with mutated NOTCH1, there were several recurrent mutations identified by at least two independent research groups in different OSCC cohorts (Fig. 3G). Four mutations identified in Caucasian tumors, G481S/V, Q1214*, and Q1080*, were also seen in Chinese patients with OSCC. However, another set of two mutations, E424K and D142H, were found in patients with leukoplakia (Fig. 3G), suggesting their potential tumorigenic role in both groups.

In this study, we examined NOTCH1 mutation status in OSCC and oral leukoplakia from Chinese patients and observed a mutation prevalence of 54% and 60%, respectively. In Asians, more than a third of mutations are located within the EGF-like repeats, particularly around the “ligand-binding” and Abruptex regions. Mutations in these areas may significantly affect NOTCH1 activity. The extracellular domain of NOTCH1 contains 36 EGF repeats. Repeats 11 to 13 constrain the “ligand-binding” domain and are necessary for direct interactions between NOTCH1 with the Delta/Serrate/Lag-2 family (27, 28). Because NOTCH1 signaling relies on these interactions, mutations in this region may inhibit ligand–receptor interactions and subsequently signal transduction. Although little is known about the contribution of EGF repeats to NOTCH1 function, it was shown that the integrity of the Abruptex (EGF repeats 24–29) is required for suppression of NOTCH1 activity (29) and that mutations within this region enhance NOTCH1 signaling (30). Although EGF repeats are involved in ligand-mediated receptor activation, mutations in NRR, another frequently mutated area in Asian tumors, can lead to ligand-independent activity (31). The membrane-proximal NRR acts as a receptor activation switch, and it is one of most mutated regions found in T-cell acute lymphoblastic leukemia (32). Taken together, the NOTCH1 mutation spectrum in Asian tumors suggests that significant proportion of alterations may result in a gain-of-function. A low prevalence of nonsense and frameshift mutations among Asian tumors further supports this notion. In Caucasian patients, the vast majority of mutations clustered around the “ligand-binding” domain, indicating that averting NOTCH1–ligand interaction may be the most prevalent cause for NOTCH1-related oral tumorigenesis in the Western populations. The next most mutated area was the RAM/ANK region of the NICD. Together with the DNA-binding factor CSL and coactivator proteins of MAML family, NICD is a core component of NOTCH1 nuclear complexes (33). The RAM region constitutes a high-affinity binding module for CSL, and the ANK domain, together with the CSL, creates a composite binding site required for MAML recruitment into these complexes (33, 34). Therefore, mutations in the RAM/ANK module may disrupt proper nuclear complexes assembly and subsequently prevent transcription of NOTCH1-dependent genes. Overall, almost 60% of mutations identified in Caucasians were located in areas associated with loss-of-function phenotype, concurrent with a recent report (13). Consequently, while there is an evidence of some clustering overlap between Caucasian and Asian tumors, the overall spectrum of NOTCH1 mutations is profoundly different between these cohorts.

Leukoplakia is considered to be an intermediate step in OSCC histopathologic progression (35, 36), a multistep process that involves accumulation of alterations from normal epithelium to invasive cancer. However, because some low-grade dysplasias progress to HNSCC, while some high-grade lesions regress to a normal state, dysplasia grade remains a poor predictor of malignant progression. Twenty years ago, we reported that the incidence of TP53 mutations in noninvasive lesions was 19% and increased to 43% in invasive carcinomas (37). With the advent of deep sequencing and the identification of NOTCH1 common mutations in Chinese tumors, we took the opportunity to detect the prevalence of NOTCH1 mutations in oral dysplasias. Although a higher incidence of mutated NOTCH1 was seen in lesions with high malignant potential, patients with leukoplakia diagnosed with mild dysplasia also demonstrated a substantial frequency of NOTCH1 alterations, confirming a crucial role for NOTCH1 in early premalignant transition. Moreover, 59.3% of patients with leukoplakia with mutated NOTCH1 carried mutations overlapping with OSCC, further supporting an important role of these events in the progression of early neoplasms. Although these data suggest that NOTCH1 mutagenesis may be sufficient for the malignant transformation in some Chinese patients, the risk of invasive transformation associated with NOTCH1 mutations in Asian population is yet to be determined. Similar to TP53, NOTCH1 mutations may exist in benign oral lesions for many years without progression to malignancy (38), and it is difficult to obtain longitudinal collection of leukoplakias and subsequent OSCCs that developed in the same patients. Nevertheless, larger confirmatory studies using longitudinal samples are warranted to further validate the role of NOTCH1 mutations in OSCC progression. Loss of heterozygosity (LOH) on 3p and/or 9p was reported to predict progression for oral leukoplakias; however, only 20% of leukoplakia cases with LOH progressed to malignant disease (12). Consequently, concurrent evaluation of NOTCH1 mutagenesis alongside LOH status may further stratify the risk and improve diagnostic sensitivity.

Major OSCC-associated risk factors are tobacco and alcohol consumption, and their effect is considered to be multiplicative rather than additive (39). Higher prevalence of individuals with smoking and alcohol consumption history among the patients with OSCC parallels epidemiologic evidence that alcohol acts as a cocarcinogen, enhancing the carcinogenicity of tobacco smoke (20), and confirms that the joint effect between tobacco and alcohol use increases the head and neck cancer risk (40). An increase in TP53 and RB mutations in OSCC has been associated with these carcinogens (21, 41, 42) and affects survival (43). Our notion that the percentage of patients with OSCC with mutated NOTCH1 who smoked and drank was significantly higher than patients with leukoplakia with similar background suggests that NOTCH1 mutations in OSCC tumors can be attributed, in part, to DNA damage from cigarette smoke and alcohol consumption. However, NOTCH1 mutagenesis cannot be entirely attributed to tobacco and alcohol use, because a substantial proportion of patients with OSCC and leukoplakia harboring mutated NOTCH1 neither smoked nor drank, suggesting that exposure to other carcinogens and environmental factors also plays a role in NOTCH1 mutagenesis. Although nonsmoking patients with OSCC and leukoplakia display a similar pattern, the spectrum of base-pair substitutions between patients with OSCC and leukoplakia with smoking history is clearly different. Consistent with complex carcinogen exposure (42), in OSCC and leukoplakia lesions with smoking background, the most frequent mutations were scattered with a very interesting peak in A to G transitions, especially in the preneoplastic lesions, and a slightly higher G to A transition frequency in nonsmokers, as noted in lung cancer (44). In accordance with Song and colleagues (16), we also did not observe an association between tumor stage and NOTCH1 mutation status, further supporting the role of NOTCH1 in OSCC malignant transformation as an early driver event.

Our integrated meta-analysis revealed a much higher NOTCH1 mutations prevalence among Asian patients. Although the etiology remains to be determined, it may include an inherited genetic background, exposure to different carcinogens and environmental factors (45). The impact of race/ethnicity on molecular alterations in cancer is a subject of important attention, as more studies reveal that groups sharing a common biologic background also acquire distinct carcinogenic molecular pathways. In lung adenocarcinoma, activating mutations in EGFR, important factors to determine small-molecule inhibitor response, commonly occur in Asians, around 50%, versus only 20% in Caucasians (46–48). Moreover, a higher NOTCH1 mutation frequency was reported in esophageal adenocarcinoma from Western patients when compared with Chinese (49), suggesting distinct tumorigenic mechanisms in each population. A significantly higher prevalence of nonsense and indel mutations in Caucasians compared with Asian patients supports the hypothesis that, in Caucasians, oral tumorigenesis more prominently relies on inactivation of NOTCH1 signaling than in the Chinese.

Interestingly, 10 recurrent mutations affecting the same codon were identified in Caucasians by at least two research groups. Although each group (13–15, 24) used a unique sample identification number, there is a chance that some samples were shared between these groups and that recurrent mutations are result of the resequencing of the same clinical sample. Although we cannot eliminate this possibility, given the low amount of overlapping variants (6.5%), the probability of parallel resequencing is very small. Three nonsense mutations result in premature translation termination, while all missense variants are located within or around the “ligand-binding” domain, suggesting an inactivating role for these alterations. Because these mutations were not identified in Chinese patients or any other tumor type, it is tempting to speculate that these variants play a crucial role in OSCC tumorigenesis and may represent “core signature” alterations among Caucasian patients. Moreover, three missense and two nonsense mutations identified in Caucasians were also seen in Chinese patients with OSCC or leukoplakia. Yet again, two of the three missense variants identified in both ethnic groups, G481S and E424K, fall within the “ligand-binding” domain. These findings further support our observation that the disruption of the “ligand-binding” domain architecture plays a prominent role in the oral tumorigenesis of both Chinese and Caucasian populations.

Although we used stringent analytic criteria to identify alterations present in the tumors and not intrinsic in the population, we acknowledge that some mutation candidates need to be further validated in paired normal–tumor samples. However, taken together, our findings provide a new insight into the role of NOTCH1 in OSCC evolution, showing for the first time that premalignant lesions display mutations that may drive malignant transformation and are likely to precede p53 inactivation in OSCC progression (37). Large sequencing studies struggle with the identification of newly mutated genes as true biologic drivers versus just passenger mutations, and our work reminds us that the study of preneoplastic lesions has been overlooked as another way to define key drivers of cancer progression in actual patients. Our work also emphatically emphasizes the need to extensively test tumors for genetic alterations from distinct ethnic and geographic areas.

V.E. Velculescu is on the Board of Directors of Personal Genome Diagnostics and has ownership interest (including patents) in the same. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Izumchenko, M. Brait, N. Agrawal, V.E. Velculescu, W.-W. Jiang, D. Sidransky

Development of methodology: E. Izumchenko, N. Agrawal, V.E. Velculescu, D. Sidransky

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Izumchenko, K. Sun, W. Koch, W.-W. Jiang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Izumchenko, S. Jones, N. Agrawal, C.L. McCord, S.V. Angiuoli, D. Sidransky

Writing, review, and/or revision of the manuscript: E. Izumchenko, K. Sun, M. Brait, N. Agrawal, W. Koch, V.E. Velculescu, D. Sidransky

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Izumchenko, K. Sun, D.R. Riley, D. Sidransky

Study supervision: E. Izumchenko, M. Brait, V.E. Velculescu, W.-W. Jiang, D. Sidransky

This work was supported by NIH grants: NCI Early Detection Research Network grant U01CA84986 (to D. Sidransky), SPORE P50 DE019032 (to D. Sidransky) and CA121113 (to V. Velculescu) and the National Natural Science Foundation of China #30872887 (to W.W. Jiang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.