embellish van ring lucky man fake give ground matrimony to a higher degree well-being from loersertydass's blog
Sequencing of neuroblastoma identifies chromothripsis and defects in neuritogenesis genes
Neuroblastoma is a childhood tumour of the peripheral sympathetic nervous system. The pathogenesis has for a long time been quite enigmatic, as only very few gene defects were identified in this often lethal tumour1. Frequently detected gene alterations are limited to MYCN amplification (20%) and ALK activations (7%)2, 3, 4, 5. Here we present a whole genome sequence analysis of 87 neuroblastoma of all stages. Few recurrent amino acid changing mutations were found. In contrast, analysis of structural defects identified a local shredding of chromosomes, known as chromothripsis, in 18% of high stage neuroblastoma6. These tumours are associated with a poor outcome. Structural alterations recurrently affected ODZ3, PTPRD and CSMD1, which are involved in neuronal growth cone stabilization7, 8, 9. In addition, ATRX, TIAM1 and a series of regulators of the Rac/Rho pathway were mutated, further implicating defects in neuritogenesis in neuroblastoma. Most tumours with defects in these genes were aggressive high stage neuroblastomas, but did not carry MYCN amplifications. The genomic landscape of neuroblastoma therefore reveals two novel molecular defects, chromothripsis and van cleef and arpels bracelets copy neuritogenesis gene alterations, which frequently occur in high risk tumours.
a, The number of amino acid changing mutations in 87 primary neuroblastoma (single nucleotide variants (SNVs) in red, deletions in grey, insertions in green and substitutions affecting more than 1 base pair (Sub) in blue). Numbers shown are events after CGAtools CallDiff with somatic scores and not present in dbSNP130, nor in 46 reference genomes released by Complete Genomics. b, Average number of mutations per tumour stage (International Neuroblastoma Staging System (INSS) stage 1, n = 9; stage 2, n = 14; stage 3, n = 5; stage 4, n = 50 and stage 4S, n = 9). Boxes include 50% of data and error bars indicate extremes with a maximum of two times the box size. st., stage. c, Kaplan curves for tumours with high versus low frequency of mutations. The optimal cut off level for the categories was determined by Kaplan scanning (see Supplementary Information and Methods). The significance (log rank test) was corrected for the multiple testing (Bonferroni correction). Number of patients per group is shown in parentheses. d, Age at diagnosis (rank order) versus the number of somatic variants. e, Average number of structural variations per tumour stage (INSS). Group sizes and the definition of the error bars as in Fig. 1b.
a, Circos plot showing structural variations in sample N492. The inner ring represents the copy number variations (red, gain; green, loss) based on coverage of the tumour and lymphocyte genomes. The lines traversing the ring indicate inter and intrachromosomal rearrangements identified by discordant mate pairs from paired end reads. N492 is a chromothripsis sample with an extreme amount of junctions on chromosome 5. b, Circos plot of the affected chromosome 5 in sample N492. c, Kaplan curves of the overall survival for tumours with or without chromothripsis. Numbers of patients per group are shown between brackets.
a, Coverage plots displaying the structural variations in the ATRX gene. The dots indicate summed coverage for bins of 1,000 pairs of the tumour genome, normalized to the coverage in corresponding normal tissue. The intron exon structure of ATRX is shown in red (dark red are exons). b, ATRX mRNA expression of 70 tumours as measured on Affymetrix full exon arrays. Tumours with ATRX deletions are encircled.
a, Diagram of a neurite growth cone depicting the function of the proteins encoded by genes with genomic aberrations in neuritogenesis. Red proteins have defects (for references see Supplementary Table 8). Rac and Rho small GTPases cycle between an inactive GDP bound and active GTP bound conformation, transducing signals from a wide variety of membrane receptors. They are activated by GEFs and inactivated by GAPs. Guanine nucleotide dissociation inhibitors (GDIs) sequester GDP bound GTPases. Proteins with aberrations in more than one tumour are marked with an asterisk (). b, Diagram of genetic defects and clinical parameters of all 87 sequenced neuroblastoma. Each vertical lane summarizes one tumour. Patients are sorted by the presence of genomic aberrations in neuritogenesis genes (Neuritogenesis, n = 19), by MYCN amplification (MYCN, n = 23), and by INSS stage (high to low). Middle panel, amino acid changing mutations and structural variations are indicated for all genes having two or more events and that are involved in neuritogenesis (red, mutated or structural variant; grey, not affected).
Neuroblastoma have a highly variable clinical outcome, with an excellent prognosis for stage 1 and 2 tumours, but a poor outcome for high stage tumours. Stage 4S neuroblastoma are metastasized but nevertheless undergo spontaneous regression. Low stage tumours are marked by numeric changes of chromosomal copy numbers, whereas high stage tumours typically show structural chromosomal defects resulting in, for example, hemizygous deletions of the chromosomal regions 1p36 or 11q and gain of 17q (refs 1, 10 Age at diagnosis above 1.5 is associated with high stage tumours and poor outcome.
We performed whole genome paired end sequencing as used by Complete Genomics13 for 87 untreated primary neuroblastoma tumours of all stages (Supplementary Table 1) and their corresponding lymphocyte DNAs. All samples had a minimal tumour content of 80% as determined by immunohistochemical analysis. Genomes were sequenced at an average coverage of 50 and an average fully called genome fraction of 96.6% (Supplementary Table 2). Compared to the HG18 reference genome we obtained an average of 3,347,592 single nucleotide variants (SNVs) per genome, in accordance with reported frequencies of interpersonal variants. Validation of 1,014 candidate somatic small mutations (SNVs, substitutions, insertions, deletions), including 763 SNVs, established a specificity of 88% and a sensitivity of 85% at a somatic score cut off of 0.1 (Supplementary Fig. 1a). SNVs above this score and all validated SNVs with lower scores were used for further analyses (total 586 genes, Supplementary Table 3). The sequence data identified an average of 12 somatic candidate amino acid affecting mutations per tumour (Fig. 1a and Supplementary Fig. 1b). 1b). 1d), as was also observed in medulloblastoma14. Within high stage neuroblastoma, MYCN amplification status did not correlate to mutation frequency (Supplementary Fig. 1c).
a, The number of amino acid changing mutations in 87 primary neuroblastoma (single nucleotide variants (SNVs) in red, deletions in grey, insertions in green and substitutions affecting more than 1 base pair (Sub) in blue). Numbers shown are events after CGAtools CallDiff with somatic scores and not present in dbSNP130, nor in 46 reference genomes released by Complete Genomics. b, Average number of mutations per tumour stage (International Neuroblastoma Staging System (INSS) stage 1, n = 9; stage 2, n = 14; stage 3, n = 5; stage 4, n = 50 and stage 4S, n = 9). Boxes include 50% of data and error bars indicate extremes with a maximum of two times the box size. st., stage. c, Kaplan curves for tumours with high versus low frequency of mutations. The optimal cut off level for the categories was determined by Kaplan scanning (see Supplementary Information and Methods). The significance (log rank test) was corrected for the multiple testing (Bonferroni correction). Number of patients per group is shown in parentheses. d, Age at diagnosis (rank order) versus the number of somatic variants. e, Average number of structural variations per tumour stage (INSS). Group sizes and the definition of the error bars as in Fig. 1b.
Only very few recurrent mutations were identified. ALK mutations were found in 6% of the tumours, in accordance with frequencies established in large neuroblastoma tumour series (Supplementary Table 4)2, 3, 4, 5. Three tumours carried mutations in TIAM1, a known regulator of cytoskeleton organization and neuritogenesis15. In a parallel study we sequenced four primary neuroblastoma tumours as well as cell lines derived from these tumours and their metastases. et al., submitted). Together with the lack of recurrent mutations, our data indicate that neuroblastoma carry few early somatic tumour driving mutations with amino acid changing consequences.
Analysis of the paired end clones with bracelet van cleef and arpels copy discordant ends can be used to identify candidate structural rearrangements, which together with sequence coverage data can identify somatic structural variants (SVs). Comparison of tumour versus lymphocyte coverage generated ultra high resolution comparative genomic hybridization (CGH) like profiles (Supplementary Fig. 2a). Analysis of the frequency of structural variations per chromosome revealed ten tumours with chromothripsis characteristics6 (see Methods). Chromothripsis is a localized shredding of a chromosomal region and subsequent random reassembly of the fragments. An extreme example of chromothripsis in chromosome 5 is shown in Fig. 2a and 2b (for other cases see Supplementary Fig. 2b). 2c). They were found in 18% of the stage 3 and 4 neuroblastoma, but not in low stage tumours (Fisher exact test P = 0.01). Accordingly, their prognostic impact is not independent of age and stage in multivariate analyses. Chromothripsis related structural aberrations frequently affected genes involved in neuroblastoma pathogenesis and were associated with amplification of MYCN or CDK4 and loss of heterozygosity of 1p (Supplementary Fig. 2c). In one tumour, chromothripsis resulted in amplification and very strong overexpression of MYC (c Myc) (Supplementary Fig. 2d). Chromosome 5 had undergone chromothripsis in three tumours, but no clear tumorigenic target on this chromosome was identified. To identify genetic defects that allowed chromothripsis and subsequent survival of the cell, we searched for defects in DNA damage response pathways in tumours with chromothripsis. The most extreme case of chromothripsis (N492, Fig. 2a and 2b) showed an inactivating deletion in FANCM and another chromothripsis tumour sample (N576) had a missense mutation in FAN1, predicted to be damaging by the polyphen2 program16. These findings might suggest involvement of inactivating events in the Fanconi anaemia signalling pathway to allow chromothripsis17.
a, Circos plot showing structural variations in sample N492. The inner ring represents the copy number variations (red, gain; green, loss) based on coverage of the tumour and lymphocyte genomes. The lines traversing the ring indicate inter and intrachromosomal rearrangements identified by discordant mate pairs from paired end reads. N492 is a chromothripsis sample with an extreme amount of junctions on chromosome 5. b, Circos plot of the van cleef arpels bracelet replica affected chromosome 5 in sample N492. c, Kaplan curves of the overall survival for tumours with or without chromothripsis. Numbers of patients per group are shown between brackets.
Full genome paired end sequencing allowed us to identify structural variants specifically perturbing single genes (see Methods and Supplementary Fig. 4 for selection procedure). We detected a total of 451 genes harbouring structural variants (306 genes without the events on chromothripsis chromosomes, Supplementary Tables 5 and 6). The structural variants often consisted of deletions of one or a few exons, inversions or translocations deleting part of a gene. One tumour showed an intrachromosomal rearrangement activating FOXR1 transcription (Supplementary Fig. 2e), which we recently identified as a recurrent but rare event in neuroblastoma18. Similar to the findings for amino acid changing mutations, there was a strong relation between the frequency of structural variations and the tumour stage (one way ANOVA P = 0.03; Fig. 1e), which extends the well established relationship between tumour stage and structural chromosomal defects in neuroblastoma10, 11, 12. Breakpoints identifying deletions were supported by changes in coverage plots. Most of the structural variants affected only one allele of a gene (Supplementary Table 5). This indicates that the tumour driving mechanism of these defects is haploinsufficiency, possibly combined with epigenetic attenuation of the non affected allele. On average, genes with structural variants resulting in loss of coverage indeed showed a reduced expression in tumours with these defects, as compared to tumours with normal alleles (Supplementary Fig. 2f). As an additional validation, we generated SNP arrays of 52 of the sequenced tumours. Although the SNP data have a much lower resolution than the sequence coverage plots, they supported the deletions and gains of sufficient size. This is especially evident on plots of chromothripsis samples (Supplementary Fig. 2g).
To identify relevant genes and pathways that contribute to neuroblastoma pathogenesis, we generated one list of all genes with amino acid changing mutations (n = 586), mutations in splice junctions (n = 37) and structural variations (n = 451). The total of 1,041 genes with alterations were analysed by two approaches. First, we analysed the most frequently affected genes (Supplementary Table 7). Four genes belonged to the MYCN amplicons (MYCN, MYCNOS, DDX1 and NBAS) and except for MYCN probably play no role in pathogenesis. Three genes, PTPRD, ODZ3 and ATRX, showed structural variants in five tumours each (Fig. 3a and Supplementary Fig. 3a) and 61 genes showed alterations in two to four tumours (Supplementary Table 7). This strongly indicates that at least the defects in PTPRD, ATRX and ODZ3 did not accumulate due to for example, the genomic length of the genes, but that they were selected for during the process of tumorigenesis.
a, Coverage plots displaying the structural variations in the ATRX gene. The dots indicate summed coverage for bins of 1,000 pairs of the tumour genome, normalized to the coverage in corresponding normal tissue. The intron exon structure of ATRX is shown in red (dark red are exons). b, ATRX mRNA expression of 70 tumours as measured on Affymetrix full exon arrays. Tumours with ATRX deletions are encircled.
The X chromosome encoded ATRX gene was affected by structural variants in five tumours (Fig. 3a). In two male patients this resulted in complete inactivation of the gene. Frequent ATRX defects were recently found in pancreatic neuroendocrine tumours19. ATRX is a chromatin remodelling protein involved in exchange of H3.3 in GC rich repeats and mutations of this gene are associated with X linked mental retardation20. Exon mRNA profiles of part of the sequenced series showed that the three samples included with ATRX structural variations had the lowest ATRX mRNA expression of all samples and showed a specific collapse of the signal in the deleted regions, illustrating the inactivating nature of the ATRX defects (Fig. 3b and Supplementary Fig. 3b).
ODZ3 and PTPRD were also hit by structural variations in five tumours each (Supplementary Fig. 3a). One tumour showed homozygous inactivation of ODZ3 (see legends of Supplementary Fig. 3a). In addition, ODZ2 and ODZ4, two highly homologous members of the conserved ODZ family, were together affected three times. PTPRD and ODZ genes encode transmembrane receptors expressed in the developing nervous system and localizing to axons and axonal growth cones21. Targeted silencing of ODZ homologues in Drosophila, Caenorhabditis elegans and mouse caused severe axon guidance defects9. Overexpression of ODZ2 in neuroblastoma cells enhanced neuritogenesis22. PTPRD is a member of the LAR subfamily of receptor protein tyrosine phosphatases. Transgenic mouse models strongly implicate the LAR subfamily receptors in neuritogenesis8. PTPRD defects in neuroblastoma were reported previously23. 3a). Interestingly, CSMD1, which showed structural variants in three tumours, is also a transmembrane protein expressed on nerve growth cones7. As the frequencies of PTPRD and ODZ3 defects exclude that they were found by chance, we propose that the function of these genes and of ODZ2, ODZ4 and CSMD1 in neuronal growth cones might hold a clue to their function in neuroblastoma pathogenesis.
The second analysis that we performed for the list for 1,041 affected genes was a gene ontology study to identify enrichment of genes with defects in specific molecular processes. This finding urged us to further investigate GTPase regulating genes in the list. TIAM1 was mutated in three tumours (see Supplementary Table 4). It functions as a guanine nucleotide exchange factor (GEF) for the small GTPase Rac and is, together with Rac, central to regulation of cellular polarity and neuritogenesis24, 25. The W1285S mutation creates a premature stop codon in the carboxy terminal pleckstrin homology domain required for Rac activation, whereas the other mutations were predicted to be damaging by polyphen2 analysis16. Rac is activated by GEFs and inactivated by GTPase activating proteins (GAPs)26 (Fig. 4). We identified a total of eight alterations in six GEFs specific for Rac (including TIAM1), but none in GAPs specific for Rac (Supplementary Tables 7, 8 and 10 for functional consequences). Whereas activation of Rac1 stimulates neuritogenesis, activation of its small GTPase antagonist RhoA promotes axon retraction and growth cone collapse (Fig. 4a)15. Strikingly, we detected seven alterations in five GAPs for RhoA, but only one GEF specific for RhoA (ARHGEF12) showed a translocation with unknown functional consequences (Fig. 4a, Supplementary Tables 7, 8 and 10 for functional consequences). Of note, transgenic mice with ATRX mutations causing mental retardation in humans showed abnormal dendritic spine formation with increased TIAM1 phosphorylation and Rac1 signalling27.
a, Diagram of a neurite growth cone depicting the function of the proteins encoded by genes with genomic aberrations in neuritogenesis. Red proteins have defects (for references see Supplementary Table 8). Rac and Rho small GTPases cycle between an inactive GDP bound and active GTP bound conformation, transducing signals from a wide variety of membrane receptors. They are activated by GEFs and inactivated by GAPs. Guanine nucleotide dissociation inhibitors (GDIs) sequester GDP bound GTPases. Proteins with aberrations in more than one tumour are marked with an asterisk (). b, Diagram of genetic defects and clinical parameters of all 87 sequenced neuroblastoma. Each vertical lane summarizes one tumour. Patients are sorted by the presence of genomic aberrations in neuritogenesis genes (Neuritogenesis, n = 19), by MYCN amplification (MYCN, n = 23), and by INSS stage (high to low). Middle panel, amino acid changing mutations and structural variations are indicated for all genes having two or more events and that are involved in neuritogenesis (red, mutated or structural variant; grey, not affected).
We conclude that alterations with significant frequencies (PTPRD and ODZ genes) affect transmembrane receptors that function in neuronal growth cone guidance and maintenance. In addition gene ontology analysis of the 1,041 genes showed significant enrichment of GTPase regulating genes. Alterations in GEFs for Rac and GAPs for Rho significantly deviate from a random distribution, implicating inhibition of Rac1 and activation of RhoA in impairing neuritogenesis in neuroblastoma (Fig. 4a).
From these findings we propose that defects in neuritogenesis regulating genes form an important category of tumour driving events in neuroblastoma. For a preliminary analysis of tumours with these defects, we selected the genes with recurrent defects in tumours that function in neuronal growth cones (PTPRD, ODZ3, ODZ2, CSMD1) or regulation of these processes through Rac/Rho signalling (TIAM1, DLC1, ARHGAP10, ATRX). The 19 tumours with defects in these genes were almost all stage 3 and 4 tumours diagnosed above 1.5
Neuroblastoma is a childhood tumour of the peripheral sympathetic nervous system. The pathogenesis has for a long time been quite enigmatic, as only very few gene defects were identified in this often lethal tumour1. Frequently detected gene alterations are limited to MYCN amplification (20%) and ALK activations (7%)2, 3, 4, 5. Here we present a whole genome sequence analysis of 87 neuroblastoma of all stages. Few recurrent amino acid changing mutations were found. In contrast, analysis of structural defects identified a local shredding of chromosomes, known as chromothripsis, in 18% of high stage neuroblastoma6. These tumours are associated with a poor outcome. Structural alterations recurrently affected ODZ3, PTPRD and CSMD1, which are involved in neuronal growth cone stabilization7, 8, 9. In addition, ATRX, TIAM1 and a series of regulators of the Rac/Rho pathway were mutated, further implicating defects in neuritogenesis in neuroblastoma. Most tumours with defects in these genes were aggressive high stage neuroblastomas, but did not carry MYCN amplifications. The genomic landscape of neuroblastoma therefore reveals two novel molecular defects, chromothripsis and van cleef and arpels bracelets copy neuritogenesis gene alterations, which frequently occur in high risk tumours.
a, The number of amino acid changing mutations in 87 primary neuroblastoma (single nucleotide variants (SNVs) in red, deletions in grey, insertions in green and substitutions affecting more than 1 base pair (Sub) in blue). Numbers shown are events after CGAtools CallDiff with somatic scores and not present in dbSNP130, nor in 46 reference genomes released by Complete Genomics. b, Average number of mutations per tumour stage (International Neuroblastoma Staging System (INSS) stage 1, n = 9; stage 2, n = 14; stage 3, n = 5; stage 4, n = 50 and stage 4S, n = 9). Boxes include 50% of data and error bars indicate extremes with a maximum of two times the box size. st., stage. c, Kaplan curves for tumours with high versus low frequency of mutations. The optimal cut off level for the categories was determined by Kaplan scanning (see Supplementary Information and Methods). The significance (log rank test) was corrected for the multiple testing (Bonferroni correction). Number of patients per group is shown in parentheses. d, Age at diagnosis (rank order) versus the number of somatic variants. e, Average number of structural variations per tumour stage (INSS). Group sizes and the definition of the error bars as in Fig. 1b.
a, Circos plot showing structural variations in sample N492. The inner ring represents the copy number variations (red, gain; green, loss) based on coverage of the tumour and lymphocyte genomes. The lines traversing the ring indicate inter and intrachromosomal rearrangements identified by discordant mate pairs from paired end reads. N492 is a chromothripsis sample with an extreme amount of junctions on chromosome 5. b, Circos plot of the affected chromosome 5 in sample N492. c, Kaplan curves of the overall survival for tumours with or without chromothripsis. Numbers of patients per group are shown between brackets.
a, Coverage plots displaying the structural variations in the ATRX gene. The dots indicate summed coverage for bins of 1,000 pairs of the tumour genome, normalized to the coverage in corresponding normal tissue. The intron exon structure of ATRX is shown in red (dark red are exons). b, ATRX mRNA expression of 70 tumours as measured on Affymetrix full exon arrays. Tumours with ATRX deletions are encircled.
a, Diagram of a neurite growth cone depicting the function of the proteins encoded by genes with genomic aberrations in neuritogenesis. Red proteins have defects (for references see Supplementary Table 8). Rac and Rho small GTPases cycle between an inactive GDP bound and active GTP bound conformation, transducing signals from a wide variety of membrane receptors. They are activated by GEFs and inactivated by GAPs. Guanine nucleotide dissociation inhibitors (GDIs) sequester GDP bound GTPases. Proteins with aberrations in more than one tumour are marked with an asterisk (). b, Diagram of genetic defects and clinical parameters of all 87 sequenced neuroblastoma. Each vertical lane summarizes one tumour. Patients are sorted by the presence of genomic aberrations in neuritogenesis genes (Neuritogenesis, n = 19), by MYCN amplification (MYCN, n = 23), and by INSS stage (high to low). Middle panel, amino acid changing mutations and structural variations are indicated for all genes having two or more events and that are involved in neuritogenesis (red, mutated or structural variant; grey, not affected).
Neuroblastoma have a highly variable clinical outcome, with an excellent prognosis for stage 1 and 2 tumours, but a poor outcome for high stage tumours. Stage 4S neuroblastoma are metastasized but nevertheless undergo spontaneous regression. Low stage tumours are marked by numeric changes of chromosomal copy numbers, whereas high stage tumours typically show structural chromosomal defects resulting in, for example, hemizygous deletions of the chromosomal regions 1p36 or 11q and gain of 17q (refs 1, 10 Age at diagnosis above 1.5 is associated with high stage tumours and poor outcome.
We performed whole genome paired end sequencing as used by Complete Genomics13 for 87 untreated primary neuroblastoma tumours of all stages (Supplementary Table 1) and their corresponding lymphocyte DNAs. All samples had a minimal tumour content of 80% as determined by immunohistochemical analysis. Genomes were sequenced at an average coverage of 50 and an average fully called genome fraction of 96.6% (Supplementary Table 2). Compared to the HG18 reference genome we obtained an average of 3,347,592 single nucleotide variants (SNVs) per genome, in accordance with reported frequencies of interpersonal variants. Validation of 1,014 candidate somatic small mutations (SNVs, substitutions, insertions, deletions), including 763 SNVs, established a specificity of 88% and a sensitivity of 85% at a somatic score cut off of 0.1 (Supplementary Fig. 1a). SNVs above this score and all validated SNVs with lower scores were used for further analyses (total 586 genes, Supplementary Table 3). The sequence data identified an average of 12 somatic candidate amino acid affecting mutations per tumour (Fig. 1a and Supplementary Fig. 1b). 1b). 1d), as was also observed in medulloblastoma14. Within high stage neuroblastoma, MYCN amplification status did not correlate to mutation frequency (Supplementary Fig. 1c).
a, The number of amino acid changing mutations in 87 primary neuroblastoma (single nucleotide variants (SNVs) in red, deletions in grey, insertions in green and substitutions affecting more than 1 base pair (Sub) in blue). Numbers shown are events after CGAtools CallDiff with somatic scores and not present in dbSNP130, nor in 46 reference genomes released by Complete Genomics. b, Average number of mutations per tumour stage (International Neuroblastoma Staging System (INSS) stage 1, n = 9; stage 2, n = 14; stage 3, n = 5; stage 4, n = 50 and stage 4S, n = 9). Boxes include 50% of data and error bars indicate extremes with a maximum of two times the box size. st., stage. c, Kaplan curves for tumours with high versus low frequency of mutations. The optimal cut off level for the categories was determined by Kaplan scanning (see Supplementary Information and Methods). The significance (log rank test) was corrected for the multiple testing (Bonferroni correction). Number of patients per group is shown in parentheses. d, Age at diagnosis (rank order) versus the number of somatic variants. e, Average number of structural variations per tumour stage (INSS). Group sizes and the definition of the error bars as in Fig. 1b.
Only very few recurrent mutations were identified. ALK mutations were found in 6% of the tumours, in accordance with frequencies established in large neuroblastoma tumour series (Supplementary Table 4)2, 3, 4, 5. Three tumours carried mutations in TIAM1, a known regulator of cytoskeleton organization and neuritogenesis15. In a parallel study we sequenced four primary neuroblastoma tumours as well as cell lines derived from these tumours and their metastases. et al., submitted). Together with the lack of recurrent mutations, our data indicate that neuroblastoma carry few early somatic tumour driving mutations with amino acid changing consequences.
Analysis of the paired end clones with bracelet van cleef and arpels copy discordant ends can be used to identify candidate structural rearrangements, which together with sequence coverage data can identify somatic structural variants (SVs). Comparison of tumour versus lymphocyte coverage generated ultra high resolution comparative genomic hybridization (CGH) like profiles (Supplementary Fig. 2a). Analysis of the frequency of structural variations per chromosome revealed ten tumours with chromothripsis characteristics6 (see Methods). Chromothripsis is a localized shredding of a chromosomal region and subsequent random reassembly of the fragments. An extreme example of chromothripsis in chromosome 5 is shown in Fig. 2a and 2b (for other cases see Supplementary Fig. 2b). 2c). They were found in 18% of the stage 3 and 4 neuroblastoma, but not in low stage tumours (Fisher exact test P = 0.01). Accordingly, their prognostic impact is not independent of age and stage in multivariate analyses. Chromothripsis related structural aberrations frequently affected genes involved in neuroblastoma pathogenesis and were associated with amplification of MYCN or CDK4 and loss of heterozygosity of 1p (Supplementary Fig. 2c). In one tumour, chromothripsis resulted in amplification and very strong overexpression of MYC (c Myc) (Supplementary Fig. 2d). Chromosome 5 had undergone chromothripsis in three tumours, but no clear tumorigenic target on this chromosome was identified. To identify genetic defects that allowed chromothripsis and subsequent survival of the cell, we searched for defects in DNA damage response pathways in tumours with chromothripsis. The most extreme case of chromothripsis (N492, Fig. 2a and 2b) showed an inactivating deletion in FANCM and another chromothripsis tumour sample (N576) had a missense mutation in FAN1, predicted to be damaging by the polyphen2 program16. These findings might suggest involvement of inactivating events in the Fanconi anaemia signalling pathway to allow chromothripsis17.
a, Circos plot showing structural variations in sample N492. The inner ring represents the copy number variations (red, gain; green, loss) based on coverage of the tumour and lymphocyte genomes. The lines traversing the ring indicate inter and intrachromosomal rearrangements identified by discordant mate pairs from paired end reads. N492 is a chromothripsis sample with an extreme amount of junctions on chromosome 5. b, Circos plot of the van cleef arpels bracelet replica affected chromosome 5 in sample N492. c, Kaplan curves of the overall survival for tumours with or without chromothripsis. Numbers of patients per group are shown between brackets.
Full genome paired end sequencing allowed us to identify structural variants specifically perturbing single genes (see Methods and Supplementary Fig. 4 for selection procedure). We detected a total of 451 genes harbouring structural variants (306 genes without the events on chromothripsis chromosomes, Supplementary Tables 5 and 6). The structural variants often consisted of deletions of one or a few exons, inversions or translocations deleting part of a gene. One tumour showed an intrachromosomal rearrangement activating FOXR1 transcription (Supplementary Fig. 2e), which we recently identified as a recurrent but rare event in neuroblastoma18. Similar to the findings for amino acid changing mutations, there was a strong relation between the frequency of structural variations and the tumour stage (one way ANOVA P = 0.03; Fig. 1e), which extends the well established relationship between tumour stage and structural chromosomal defects in neuroblastoma10, 11, 12. Breakpoints identifying deletions were supported by changes in coverage plots. Most of the structural variants affected only one allele of a gene (Supplementary Table 5). This indicates that the tumour driving mechanism of these defects is haploinsufficiency, possibly combined with epigenetic attenuation of the non affected allele. On average, genes with structural variants resulting in loss of coverage indeed showed a reduced expression in tumours with these defects, as compared to tumours with normal alleles (Supplementary Fig. 2f). As an additional validation, we generated SNP arrays of 52 of the sequenced tumours. Although the SNP data have a much lower resolution than the sequence coverage plots, they supported the deletions and gains of sufficient size. This is especially evident on plots of chromothripsis samples (Supplementary Fig. 2g).
To identify relevant genes and pathways that contribute to neuroblastoma pathogenesis, we generated one list of all genes with amino acid changing mutations (n = 586), mutations in splice junctions (n = 37) and structural variations (n = 451). The total of 1,041 genes with alterations were analysed by two approaches. First, we analysed the most frequently affected genes (Supplementary Table 7). Four genes belonged to the MYCN amplicons (MYCN, MYCNOS, DDX1 and NBAS) and except for MYCN probably play no role in pathogenesis. Three genes, PTPRD, ODZ3 and ATRX, showed structural variants in five tumours each (Fig. 3a and Supplementary Fig. 3a) and 61 genes showed alterations in two to four tumours (Supplementary Table 7). This strongly indicates that at least the defects in PTPRD, ATRX and ODZ3 did not accumulate due to for example, the genomic length of the genes, but that they were selected for during the process of tumorigenesis.
a, Coverage plots displaying the structural variations in the ATRX gene. The dots indicate summed coverage for bins of 1,000 pairs of the tumour genome, normalized to the coverage in corresponding normal tissue. The intron exon structure of ATRX is shown in red (dark red are exons). b, ATRX mRNA expression of 70 tumours as measured on Affymetrix full exon arrays. Tumours with ATRX deletions are encircled.
The X chromosome encoded ATRX gene was affected by structural variants in five tumours (Fig. 3a). In two male patients this resulted in complete inactivation of the gene. Frequent ATRX defects were recently found in pancreatic neuroendocrine tumours19. ATRX is a chromatin remodelling protein involved in exchange of H3.3 in GC rich repeats and mutations of this gene are associated with X linked mental retardation20. Exon mRNA profiles of part of the sequenced series showed that the three samples included with ATRX structural variations had the lowest ATRX mRNA expression of all samples and showed a specific collapse of the signal in the deleted regions, illustrating the inactivating nature of the ATRX defects (Fig. 3b and Supplementary Fig. 3b).
ODZ3 and PTPRD were also hit by structural variations in five tumours each (Supplementary Fig. 3a). One tumour showed homozygous inactivation of ODZ3 (see legends of Supplementary Fig. 3a). In addition, ODZ2 and ODZ4, two highly homologous members of the conserved ODZ family, were together affected three times. PTPRD and ODZ genes encode transmembrane receptors expressed in the developing nervous system and localizing to axons and axonal growth cones21. Targeted silencing of ODZ homologues in Drosophila, Caenorhabditis elegans and mouse caused severe axon guidance defects9. Overexpression of ODZ2 in neuroblastoma cells enhanced neuritogenesis22. PTPRD is a member of the LAR subfamily of receptor protein tyrosine phosphatases. Transgenic mouse models strongly implicate the LAR subfamily receptors in neuritogenesis8. PTPRD defects in neuroblastoma were reported previously23. 3a). Interestingly, CSMD1, which showed structural variants in three tumours, is also a transmembrane protein expressed on nerve growth cones7. As the frequencies of PTPRD and ODZ3 defects exclude that they were found by chance, we propose that the function of these genes and of ODZ2, ODZ4 and CSMD1 in neuronal growth cones might hold a clue to their function in neuroblastoma pathogenesis.
The second analysis that we performed for the list for 1,041 affected genes was a gene ontology study to identify enrichment of genes with defects in specific molecular processes. This finding urged us to further investigate GTPase regulating genes in the list. TIAM1 was mutated in three tumours (see Supplementary Table 4). It functions as a guanine nucleotide exchange factor (GEF) for the small GTPase Rac and is, together with Rac, central to regulation of cellular polarity and neuritogenesis24, 25. The W1285S mutation creates a premature stop codon in the carboxy terminal pleckstrin homology domain required for Rac activation, whereas the other mutations were predicted to be damaging by polyphen2 analysis16. Rac is activated by GEFs and inactivated by GTPase activating proteins (GAPs)26 (Fig. 4). We identified a total of eight alterations in six GEFs specific for Rac (including TIAM1), but none in GAPs specific for Rac (Supplementary Tables 7, 8 and 10 for functional consequences). Whereas activation of Rac1 stimulates neuritogenesis, activation of its small GTPase antagonist RhoA promotes axon retraction and growth cone collapse (Fig. 4a)15. Strikingly, we detected seven alterations in five GAPs for RhoA, but only one GEF specific for RhoA (ARHGEF12) showed a translocation with unknown functional consequences (Fig. 4a, Supplementary Tables 7, 8 and 10 for functional consequences). Of note, transgenic mice with ATRX mutations causing mental retardation in humans showed abnormal dendritic spine formation with increased TIAM1 phosphorylation and Rac1 signalling27.
a, Diagram of a neurite growth cone depicting the function of the proteins encoded by genes with genomic aberrations in neuritogenesis. Red proteins have defects (for references see Supplementary Table 8). Rac and Rho small GTPases cycle between an inactive GDP bound and active GTP bound conformation, transducing signals from a wide variety of membrane receptors. They are activated by GEFs and inactivated by GAPs. Guanine nucleotide dissociation inhibitors (GDIs) sequester GDP bound GTPases. Proteins with aberrations in more than one tumour are marked with an asterisk (). b, Diagram of genetic defects and clinical parameters of all 87 sequenced neuroblastoma. Each vertical lane summarizes one tumour. Patients are sorted by the presence of genomic aberrations in neuritogenesis genes (Neuritogenesis, n = 19), by MYCN amplification (MYCN, n = 23), and by INSS stage (high to low). Middle panel, amino acid changing mutations and structural variations are indicated for all genes having two or more events and that are involved in neuritogenesis (red, mutated or structural variant; grey, not affected).
We conclude that alterations with significant frequencies (PTPRD and ODZ genes) affect transmembrane receptors that function in neuronal growth cone guidance and maintenance. In addition gene ontology analysis of the 1,041 genes showed significant enrichment of GTPase regulating genes. Alterations in GEFs for Rac and GAPs for Rho significantly deviate from a random distribution, implicating inhibition of Rac1 and activation of RhoA in impairing neuritogenesis in neuroblastoma (Fig. 4a).
From these findings we propose that defects in neuritogenesis regulating genes form an important category of tumour driving events in neuroblastoma. For a preliminary analysis of tumours with these defects, we selected the genes with recurrent defects in tumours that function in neuronal growth cones (PTPRD, ODZ3, ODZ2, CSMD1) or regulation of these processes through Rac/Rho signalling (TIAM1, DLC1, ARHGAP10, ATRX). The 19 tumours with defects in these genes were almost all stage 3 and 4 tumours diagnosed above 1.5
The Wall