Cohesin is a multimeric protein complex that is involved in the cohesion of sister chromatids, post-replicative DNA repair and transcriptional regulation. Here we report recurrent mutations and deletions involving multiple components of the cohesin complex, including STAG2, RAD21, SMC1A and SMC3, in different myeloid neoplasms. These mutations and deletions were mostly mutually exclusive and occurred in 12.1% (19/157) of acute myeloid leukemia, 8.0% (18/224) of myelodysplastic syndromes, 10.2% (9/88) of chronic myelomonocytic leukemia, 6.3% (4/64) of chronic myelogenous leukemia and 1.3% (1/77) of classical myeloproliferative neoplasms. Cohesin-mutated leukemic cells showed reduced amounts of chromatin-bound cohesin components, suggesting a substantial loss of cohesin binding sites on chromatin. The growth of leukemic cell lines harboring a mutation in RAD21 (Kasumi-1 cells) or having severely reduced expression of RAD21 and STAG2 (MOLM-13 cells) was suppressed by forced expression of wild-type RAD21 and wild-type RAD21 and STAG2, respectively. These findings suggest a role for compromised cohesin functions in myeloid leukemogenesis.
At a glance
Recent genetic studies have led to the discovery of a number of new mutational targets in myeloid malignancies, unmasking unexpected roles for deregulated histone modification and DNA methylation in both acute and chronic myeloid neoplasms1, 2. However, knowledge of the spectrum of gene mutations in myeloid neoplasms remains incomplete. We previously reported a whole-exome sequencing study of 29 paired tumor and normal samples of myeloid neoplasms with myelodysplastic features3. Although our major discovery was that frequent spliceosome mutations are uniquely associated with myelodysplasia phenotypes, we also identified hundreds of previously unreported gene mutations3. Most of those mutations affected single individuals only and are probably passenger changes. Therefore, their importance in leukemogenesis remains undetermined. However, through closer inspection of an updated list of mutations, including newly validated single-nucleotide variants, we identified additional recurrent mutations involving STAG2, a core component of the cohesin complex (Online Methods and Supplementary Table 1). In addition, we found that two other functionally related cohesin components, STAG1 and PDS5B, were mutated in single specimens (Supplementary Fig. 1).
Cohesin is a multimeric protein complex that is conserved across species and is composed of four core subunits, SMC1, SMC3, RAD21 and STAG proteins, together with a number of regulatory molecules such as PDS5, NIPBL and ESCO proteins (Fig. 1a)4, 5. Forming a ring-like structure, cohesin is thought to be engaged in the cohesion of sister chromatids during cell division5, post-replicative DNA repair6, 7 and the regulation of global gene expression through long-range cis interactions8, 9, 10, 11, 12. Germline mutations in cohesin components lead to the congenital multisystem malformation syndromes known as Cornelia de Lange syndrome and Roberts syndrome13, 14, 15.
To investigate a possible role of cohesin mutations in myeloid leukemogenesis, we examined an additional 581 primary specimens of various myeloid neoplasms for mutations in nine cohesin or cohesin-related genes that have been implicated in mitosis5 using high-throughput sequencing (Supplementary Table 2). We also investigated copy-number alterations in cohesin loci in 453 samples using SNP arrays (Supplementary Table 3). After excluding known and putative polymorphisms that are registered in the dbSNP or the 1000 Genomes project databases or that were predicted from multiple computational imputations, we identified a total of 60 nonsynonymous mutations involving nine genes in a total of 610 primary samples, which we validated by Sanger sequencing (Fig. 1b and Supplementary Table 4). After conservative evaluation of the probability of random mutational events across these genes, only four genes remained significantly mutated: STAG2, RAD21, SMC1A and SMC3 (P < 0.001) (Supplementary Table 5 and Online Methods). In addition, we detected five deletions in STAG2 (n = 4) and RAD21 (n = 1) (Supplementary Fig. 2a,b and Supplementary Table 6). We also found mutations in these four genes in four of the 34 myeloid leukemia cell lines studied (12%) (Supplementary Table 7).
We found mutations and deletions of these four genes in a mostly mutually exclusive manner in a variety of myeloid neoplasms, including acute myeloid leukemia (AML) (19/157), chronic myelomonocytic leukemia (CMML) (9/88), myelodysplastic syndromes (MDS) (18/224) and chronic myelogenous leukemia (CML) (4/64). Mutations were rate in classical myeloproliferative neoplasms (MPN) (1/77) (Fig. 1c, Table 1 and Supplementary Table 8). In MDS, mutations were more frequent in refractory cytopenia with multilineage dysplasia and refractory anemia with excess blasts (11.4%) but were rare in refractory anemia, refractory anemia with ring sideroblasts, refractory cytopenia with multilineage dysplasia and ring sideroblasts and MDS with isolated del(5q) (4.2%) (P = 0.044). We also evaluated promoter methylation in 33 cases either with (n = 12) or without (n = 21) cohesin mutations or deletions for which sufficient nonamplified DNA was available using the HumanMethylation450 BeadChip; however, we found no aberrant methylations in cohesin loci, with the exception of hemimethylation of the SMC1A promoter that we found in two female cases (Supplementary Fig. 3).
We confirmed somatic origins for 17 mutations detected in 16 cases for which matched normal DNA was available (Supplementary Table 4). The somatic origins of an additional 23 mutations in STAG2 or SMC1A found in 20 male cases were supported by the presence of reproducible wild-type signals or reads in Sanger and/or deep sequencing of the tumor samples, which were considered to originate from the X chromosome of the residual normal cells (Supplementary Fig. 4). In addition, for 20 mutations, the observed allele frequencies determined by pyrosequencing, deep sequencing or digital PCR showed significant deviations from the expected value for polymorphisms in the absence of apparent chromosomal alterations in a SNP array analysis (P < 0.01) (Supplementary Figs. 5 and 6 and Supplementary Tables 9, 10, 11, 12), suggesting their somatic origins. In addition, 32 of the 33 STAG2 mutations and all of the nine RAD21 mutations were either nonsense (n = 18), frameshift (n = 14) or splice-site (n = 9) changes, which were predicted to cause premature truncation of the protein or abnormal exon skipping (Fig. 1b and Supplementary Figs. 7 and 8). Thus, we considered the majority of the mutations to represent functionally relevant changes, probably of somatic origins (Supplementary Table 13).
Most of the cohesin mutations and deletions were heterozygous, except for the STAG2 and SMC1A mutations on the single X chromosome in male cases (n = 23). In female samples, the STAG2 promoter was hemimethylated through X inactivation regardless of mutation status (Supplementary Fig. 3), and a heterozygous mutation of the unmethylated STAG2 allele would lead to biallelic STAG2 inactivation, as has been previously documented in a female case with Ewing's sarcoma16 and was also confirmed in a single case (CMML-036) in our cohort (Supplementary Fig. 9).
Cohesin mutations frequently coexisted with other mutations that are common in myeloid neoplasms and significantly associated with mutations in TET2 (P = 0.027), ASXL1 (P = 0.045) and EZH2 (P = 0.011) (Fig. 2a). We performed deep sequencing of the mutant alleles in 20 available samples with cohesin mutations, which allowed for accurate determination of their allele frequencies. The majority of the cohesin mutations (15/20) existed in the major tumor populations, indicating their early origin during leukemogenesis. In the remaining five samples, we found cohesin mutations only in a tumor subpopulation, indicating that the mutations were relatively late events (Fig. 2b). Two male cases (MDS-176 and AML-158) harbored two independent subclones with different STAG2 mutations, indicating that STAG2 mutation could confer a strong advantage to pre-existing leukemic cells during clonal evolution (Supplementary Fig. 10). The number of mutations determined by whole-exome sequencing3 was significantly higher in four cases with cohesin mutation or deletion compared to cases with no mutation or deletion of cohesin (P = 0.049) (Supplementary Fig. 11).
Next we investigated the possible impact of mutations on cohesin function. We examined the expression of STAG1, STAG2, RAD21, SMC3, SMC1A and NIPBL in 17 myeloid leukemia cell lines with (n = 4) or without (n = 13) known cohesin mutations, as well as in the chromatin-bound fractions of 13 cell lines (Fig. 3a–d and Supplementary Table 14)14, 17, 18, 19. Although we observed an evaluable reduction in RAD21 expression in Kasumi-1 cells that harbored a frameshift alteration in RAD21 (p.Lys330ProfsX6) (Fig. 3a), alterations in P31FUJ (RAD21 p.His208Arg), CMY (RAD21 p.Tyr3X) and MOLM-7 (SMC3 p.Arg661Pro) cells were not accompanied by measurable decreases in the corresponding mutated proteins compared to wild-type cell lines. In contrast, we observed severely reduced expression of one or more cohesin components in KG-1 (STAG2)16 and MOLM-13 (STAG1, STAG2, RAD21 and NIPBL) cells without any accompanying mutations in the relevant genes (Fig. 3a). We found no significant differences in protein expression of the cohesin components in cohesin-mutated and non-mutated cell lines in whole-cell extracts (Fig. 3b). However, expression of one or more cohesin components, including SMC1, SMC3, RAD21 and STAG2, was significantly reduced in the chromatin-bound fractions of cell lines with mutated or reduced expression of cohesin components, including Kasumi-1, KG-1, P31FUJ, MOLM-7 and MOLM-13 cells, compared with the cell lines with no known cohesin mutations or abnormal cohesin expression (P < 0.05), suggesting a substantial loss of cohesin-bound sites on chromatin (Fig. 3c,d and Supplementary Table 14)14.
We next examined the effect of forced expression of wild-type cohesin components on the proliferation of a cohesin-mutated cell line (Kasumi-1) or a cell line with reduced expression of cohesin components (MOLM-13). Forced expression of wild-type RAD21 and/or STAG2, but not of a truncated RAD21 allele, induced significant growth suppression of the Kasumi-1 (with mutated RAD21) and MOLM-13 (with severe reduction of RAD21 and STAG2 expression) cell lines but not the K562 and TF1 (with wild-type RAD21) cell lines, supporting a leukemogenic role for compromised cohesin functions (Fig. 4a–c and Supplementary Fig. 12a–g). To explore the effect of forced expression of RAD21 on global gene expression, we performed expression microarray analysis of RAD21- and mock-transduced Kasumi-1 cells. In agreement with previous experiments with other cohesin and cohesin-related components, the magnitudes of the transcriptional changes induced by forced RAD21 expression were generally small14, 16, 20. However, 63 genes reproducibly and significantly showed a more than 1.2-fold increase (n = 35) or decrease (n = 28) in gene expression (P < 0.05), which was validated by quantitative PCR and/or RNA sequencing for 59 of the 63 genes (Supplementary Fig. 13a–c and Supplementary Tables 15 and 16).
Mutations in the cohesin complex have recently been reported in a cohort of de novo AML and MDS in which four major cohesin components were mutated in 6.0–13.0% of cases21, 22, 23, 24, 25. Less frequent mutations of cohesin components have been described in other cancers, including STAG2 mutations in glioblastoma (4/68), melanoma (1/48) and Ewing's sarcoma (1/24)16. In primary colon cancer samples, in which impaired cohesion and consequent aneuploidy have been implicated in oncogenesis, mutations in SMC1A (4/132), NIPBL (4/132), STAG3 (1/130) and SMC3 (1/130) have been reported26. In contrast, in our cohort of myeloid neoplasms, we found no significant differences in the number of numerical chromosome abnormalities between cohesin-mutated and non-mutated cases, and the 43 cases with cohesin mutations or deletions showed diploid or near-diploid karyotypes, including 23 cases with completely normal karyotypes (Fig. 4d). Therefore, in these euploid cases, cohesin-mutated cells were not clonally selected as a result of aneuploidy. Supporting this finding is the observation that expression of scc1p, a RAD21 homolog, at only 13% of its normal level was sufficient for normal cohesion in yeast27. Furthermore, Kasumi-1 and MOLM-13 cells showed almost normal cohesion of sister chromatids, even though Kasumi-1 cells have a truncated RAD21 allele and MOLM-13 cells have substantially reduced expression of multiple cohesin components (Fig. 4e).
A growing body of evidence has suggested that cohesin mediates long-range chromosomal cis interactions28 and regulates global gene expression11, 12. For example, two cohesin subunits, Rad21 and Smc3, have been implicated in the transcriptional regulation of the hematopoietic transcription factor Runx1 in zebrafish10. Furthermore, an up to 80% downregulation of Nipped-B, a NIPBL homolog in Drosophila, does not affect chromosomal segregation but does cause impaired regulation of gene expression20. We also previously demonstrated that only mild loss (17–28%) of cohesin binding sites within the genome results in deregulated global gene expression14, 18, 19. These observations suggest the possibility that cohesin mutations participate in leukemogenesis through the deregulated expression of genes that are involved in myeloid development and differentiation.
In conclusion, we report frequent mutations in cohesin components that involve a wide variety of myeloid neoplasms. Genetic evidence suggests that aneuploidy may not be the only leukemogenic mechanism, at least in vivo, and that deregulated gene expression and/or other mechanisms, such as DNA hypermutability, might also operate in leukemogenesis. Given the integral functions of cohesin for cell viability, genetic defects in cohesin might be potential targets in myeloid neoplasms14, 29.
Patients and samples.
Twenty-nine cases analyzed by whole-exome sequencing were described previously3. Anonymized genomic DNA from an additional 581 patients with different myeloid neoplasms were collected from collaborating institutes and used for the analyses described below. All the analyses were performed after written informed consent was obtained. This study was approved by the ethics boards of the University of Tokyo, University Hospital Mannheim, University of Tsukuba, the Munich Leukemia Laboratory, Showa University, Tokyo Metropolitan Ohtsuka Hospital and Chang Gung Memorial Hospital.
The CMS, CMY, UTP-DSAL-1, MOLM-1, MOLM-7, HEL, SS9;22 and TS9;22 cell lines were provided by Y. Hayashi. 293gp and 293gpg cells were provided by R.C. Mulligan. P31FUJ and CMK-86 cells were purchased from the Health Science Research Resources Bank (Osaka, Japan). 293T, KG-1, K562 and F-36P cells were obtained from RIKEN BioResource Center Cell Bank (Tsukuba, Japan), and Kasumi-1, HL-60, MOLM-13 and TF-1 cells were from the American Type Culture Collection. Chromosome spreads were performed for the CMS, Kasumi-1 and MOLM-13 cell lines as previously described14, except that cells were treated with colcemid (100 μg/ml) and hypotonically swollen in 75 mM KCl for 20 min.
The whole-exome sequencing of the 29 paired samples of myelodysplasia was previously described3, through which we identified a total of 497 candidate single-nucleotide variants and insertions/deletions (indels), of which 268 and 167 were determined by Sanger sequencing as true positives and negatives, respectively, with 62 mutations unconfirmed. In the present study, we updated the list of somatic mutations by rigorously validating the remaining 62 unconfirmed mutations by Sanger sequencing and also by deep sequencing (Supplementary Table 1).
Mutation analysis of cohesin components.
In total, 534 tumor DNA samples from a variety of myeloid neoplasms were analyzed for possible mutations in nine components of the cohesin complex, STAG1, STAG2, SMC1A, SMC3, RAD21, PDS5B, ESCO1, ESCO2 and NIPBL, using high-throughput sequencing of pooled exons amplified from pooled genomic DNA samples. In an additional 47 samples, mutations in STAG2, RAD21, SMC1A and SMC3 were examined by deep sequencing after enrichment for these targets using a SureSelect custom kit (Agilent) designed to capture all of the coding exons from the target genes, performed as previously described with minor modifications in the algorithm for mutation call30.
For pooled-DNA sequencing, all target exons (n = 232) encompassing 89,323 nucleotides were PCR amplified using a set of primers having common NotI adaptor sequences on their 5′ ends, digested with NotI, ligated using T4 ligase and sonicated to approximately 200-bp fragments using an ultrasonicator (Covaris); these fragments were used for the generation of sequencing libraries according to a modified pair-end protocol from Illumina. The libraries were then sequenced using HiSeq 2000 (Illumina) with a standard 100-bp paired end–reads protocol. On average, 99.5% of the target bases were analyzed at the depth of 12,000 per pool or 1,000 per sample. Data processing and variant calling were performed as previously described3 with minor modifications. First, each read from a given DNA pool was aligned to the set of target sequences using BLAT31 with the -fine option. The mapping information in a .psl format was transformed into a .sam format using the my_psl2sam script, which was further converted into the .bam format using SAMtools32. Among the successfully mapped reads, reads were removed from further analysis that either mapped to multiple sites, mapped with more than four mismatched bases or had more than ten clipped bases. Next, the Estimation_CRME script was run to eliminate strand-specific errors and exclude PCR-derived errors. Then, a strand-specific mismatch ratio was calculated for each nucleotide variation for both strands using the bases corresponding to 11−50 cycles. By excluding the top five cycles showing the highest mismatch rates, strand-specific mismatch rates were recalculated, and the smaller value between both strands was adopted as the nominal mismatch ratio. In addition, the nucleotide variations that were present across multiple pools were removed based on permutations across different pools using the Permut_Rm_com script because it is probable that such variations result from systemic sequencing errors. Finally, after excluding variations found in the dbSNP database, the database from the 1000 Genomes project or our in-house SNP database, the variants whose mismatch rate exceeded 0.009 were adopted as candidate mutations. Each candidate mutation was validated by Sanger sequencing of the 12 original individual DNAs from the corresponding DNA pools.
The functional impact of each amino acid substitution was evaluated by computer prediction using SIFT33, PolyPhen-2 (ref. 34) and Mutation Taster35. The significance of nonsilent mutations in each cohesin component was evaluated assuming a uniform distribution of the background mutations within the coding regions, which was estimated to be ~0.3 Mb−1 on the basis of a previous whole-exome sequencing of myelodysplasia3.
Determination of variant allele frequencies.
Variant allele frequencies were evaluated by deep sequencing of PCR amplicons, pyrosequencing36, 37 and/or digital PCR (Fluidigm CA, US)38, 39, 40 of the variants using nonamplified DNA. For amplicon sequencing, genomic fragments harboring the variants of interest were PCR amplified using Not1-tagged primers. Ninety-two randomly selected SNP loci that do not contain repetitive sequences were amplified using normal genomic DNA as a template, which served as the control. Touch-down PCRs using high-fidelity DNA polymerase KOD-Plus-Neo (TOYOBO, Tokyo) were performed, and an equimolar mixture of all PCR products was prepared for deep sequencing using HiSeq2000 or Miseq (Illumina), as described above, with a 75-bp or 100-bp pair end–read option. To calculate the allele frequency of each variant, all reads were mapped to the target reference sequence using BLAT31, followed by differential enumeration of the dichotomic variant alleles. For indels, individual reads were first aligned to each of the wild-type and altered sequences and then assigned to the one with better alignment in terms of the number of matched bases.
Array-based copy-number and methylation analyses.
Genomic DNA from 453 bone marrow samples with myeloid neoplasms was analyzed using GeneChip SNP genotyping microarrays as previously described using CNAG/AsCNAR software41, 42. The results of the SNP array karyotyping for 290 of the 453 cases have been previously published3, 41, 42, 43, 44. The promoter methylation of each cohesin component gene was analyzed using the HumanMethylation450 BeadChip (Illumina), as previously described30, 45, in which methylation status was evaluated by calculating the ratio of methylation-specific and demethylation-specific fluorophores (β value) at each CpG site using iScan software (Illumina).
Protein expression of cohesin components in whole-cell extracts and chromatin-enriched fractions.
Whole-cell extracts of myeloid cell lines were separated into soluble supernatant and chromatin-containing pellet fractions and analyzed by SDS-PAGE and protein blot analysis for the expression of different cohesin components as previously described12, 14. Antibodies used for protein blot analysis are described in Supplementary Table 18.
Gene expression and cell proliferation assays.
A full-length RAD21 cDNA (BC050381) was provided by S. Sugano. A full-length STAG2 cDNA was obtained from total cDNA derived from bone marrow cells and cloned into pBluescript. The truncated mutant of RAD21 was subcloned by PCR. Flag-tagged RAD21 or STAG2 cDNAs were constructed into the retrovirus vector pGCDNsamIRESEGFP (provided by M. Onodera)46 or a tetracycline-inducible lentiviral vector, CS-TRE-Ubc-tTA-IRESPuro. The wild-type RAD21, the mutant RAD21 and/or a mock-induced retroviral vector were generated as previously described3 and transduced into Kasumi-1, K562 and TF1 cells, which were sorted by GFP marking using a MoFlo FACS cell sorter (Beckman Coulter) or a BD FACSAria cell sorter (BD Biosciences) 48–96 h after retroviral transduction. The wild-type RAD21, the wild-type STAG2 and a mock-induced lentiviral vector were generated as described previously47, transduced into MOLM-13 cells and selected by 1 μg/ml puromycin. Gene expression was induced by 1 μg/ml doxcycline. For cell growth assays, the cells were inoculated into 96-well culture plates in RPMI 1640 medium supplemented with 5% FCS (and 5 ng/ml GM-CSF for TF1 cells), and cell growth was monitored in three independent experiments by MTT assay using the Cell Counting Kit-8 (Dojindo Co.).
Expression microarray analysis.
RNA was extracted from Kasumi-1 cells that were either mock transduced or transduced with wild-type RAD21 and analyzed in triplicate using the Human Genome U133 Plus 2.0 Array (Affymetrix) according to the manufacturer's protocol. For data analysis, raw array signals were first extracted from .CEL files using dChip Software48. After background correction and normalization across the six array data sets, the standardized signal value was obtained for each probe set in each of triplicate array experiments, which were compared between mock-transduced and wild-type RAD21–transduced cells. Two independent microarray experiments were performed. To identify transcriptionally altered genes, we used the criteria of fold change greater than ±1.2 and P < 0.05 (two-tailed paired t test) in two independent experiments.
RNA sequencing of RAD21-transduced Kasumi-1 cells and subsequent data analyses were performed as previously described3 with minor modifications. For quantifications of expression values from the RNA sequencing data, we used a slightly modified version of RKPM (reads per kb of exon per million mapped reads) measures49. After removing the sequencing reads that were inappropriately aligned or that had low mapping quality, the number of bases on each exonic region for each RefSeq gene50 was counted. Then the number of bases was normalized per kb of exon and per 100 million aligned bases. Finally, the expression value of each gene was determined by taking the maximum values among the RefSeq genes corresponding to the gene symbol.
We measured RAD21 expression by differentially enumerating endogenous and exogenous RAD21 sequence reads, which were discriminated by the absence and presence of the Flag sequence, respectively. After normalization by the number of total reads for each sample, the raw differential read counts were further calibrated against the read counts containing the stop codon in RAD21.
The significance of the difference in frequency of cohesin component mutations between disease subtypes was tested by one-tailed Fisher's exact test. The coexistence of mutations was tested by two-tailed Fisher's direct method. The significance of the difference in the total number of somatic mutations between cohesin-mutated or -deleted and non-mutated or -deleted samples was tested by Mann-Whitney U test. Differences in the number of numerical abnormalities in cytogenetics between two groups with and without cohesin mutations or deletions was assessed by one-sided χ2 test.
dbSNP, http://www.ncbi.nlm.nih.gov/projects/SNP/; the 1000 Genomes Project, http://www.1000genomes.org/; the UCSC Genome Browser; http://genome.ucsc.edu/cgi-bin/hgGateway/; hg19, http://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/; RefSeq genes, http://www.ncbi.nlm.nih.gov/RefSeq/; CNAG/AsCNAR, http://www.genome.umin.jp/; dChip, http://www.dchip.org/; the Integrative Genomics Viewer, http://www.broadinstitute.org/igv/; SIFT, http://sift.jcvi.org/; PolyPhen-2, http://genetics.bwh.harvard.edu/pph2/; Mutation Taster, http://www.mutationtaster.org/.
Whole-exome sequence data have been deposited in the DNA Data Bank of Japan (DDBJ) repository under accession number DRA000433. RNA sequencing data have been deposited in the DDBJ repository under accession number DRA001013. Microarray data have been deposited in the Gene Expression Omnibus under accession number GSE47684.
Gene Expression Omnibus
- Unraveling the molecular pathophysiology of myelodysplastic syndromes. J. Clin. Oncol. 29, 504–515 (2011). , &
- Molecular genetics of adult acute myeloid leukemia: prognostic and therapeutic implications. J. Clin. Oncol. 29, 475–486 (2011). , &
- Frequent pathway mutations of splicing machinery in myelodysplasia. Nature 478, 64–69 (2011). et al.
- Chromosomal cohesin forms a ring. Cell 112, 765–777 (2003). , &
- Cohesin: its roles and mechanisms. Annu. Rev. Genet. 43, 525–558 (2009). &
- Postreplicative formation of cohesion is required for repair and induced by a single DNA break. Science 317, 242–245 (2007). et al.
- The cohesin complex is required for the DNA damage-induced G2/M checkpoint in mammalian cells. EMBO J. 28, 2625–2635 (2009). &
- Cohesin, gene expression and development: lessons from Drosophila. Chromosome Res. 17, 185–200 (2009).
- Effects of sister chromatid cohesion proteins on cut gene expression during wing development in Drosophila. Development 132, 4743–4753 (2005). et al.
- Cohesin-dependent regulation of Runx genes. Development 134, 2639–2649 (2007). et al.
- Cohesins functionally associate with CTCF on mammalian chromosome arms. Cell 132, 422–433 (2008). et al.
- Cohesin mediates transcriptional insulation by CCCTC-binding factor. Nature 451, 796–801 (2008). et al.
- Cohesinopathies, gene expression, and chromatin organization. J. Cell Biol. 189, 201–210 (2010). &
- HDAC8 mutations in Cornelia de Lange syndrome affect the cohesin acetylation cycle. Nature 489, 313–317 (2012). et al.
- RAD21 mutations cause a human cohesinopathy. Am. J. Hum. Genet. 90, 1014–1027 (2012). et al.
- Mutational inactivation of STAG2 causes aneuploidy in human cancer. Science 333, 1039–1043 (2011). et al.
- An Smc3 acetylation cycle is essential for establishment of sister chromatid cohesion. Mol. Cell 39, 689–699 (2010). et al.
- Transcriptional dysregulation in NIPBL and cohesin mutant human cells. PLoS Biol. 7, e1000119 (2009). et al.
- Genome-wide DNA methylation analysis in cohesin mutant human cell lines. Nucleic Acids Res. 38, 5657–5671 (2010). et al.
- Regulation of the Drosophila enhancer of split and invected-engrailed gene complexes by sister chromatid cohesion proteins. PLoS ONE 4, e6202 (2009). et al.
- Clonal evolution in relapsed acute myeloid leukaemia revealed by whole-genome sequencing. Nature 481, 506–510 (2012). et al.
- Clonal architecture of secondary acute myeloid leukemia. N. Engl. J. Med. 366, 1090–1098 (2012). et al.
- The origin and evolution of mutations in acute myeloid leukemia. Cell 150, 264–278 (2012). et al.
- The Cancer Genome Atlas Research Network. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N. Engl. J. Med. 368, 2059–2074 (2013).
- Clonal diversity of recurrently mutated genes in myelodysplastic syndromes. Leukemia 27, 12785–1282 (2013). et al.
- Chromatid cohesion defects may underlie chromosome instability in human colorectal cancers. Proc. Natl. Acad. Sci. USA 105, 3443–3448 (2008). et al.
- Systematic reduction of cohesin differentially affects chromosome segregation, condensation, and DNA repair. Curr. Biol. 20, 957–963 (2010). , , , &
- Cohesins form chromosomal cis-interactions at the developmentally regulated IFNG locus. Nature 460, 410–413 (2009). et al.
- Harnessing synthetic lethal interactions in anticancer drug discovery. Nat. Rev. Drug Discov. 10, 351–364 (2011). &
- Integrated molecular analysis of clear-cell renal cell carcinoma. Nat. Genet. doi:10.1038/ng.2699 (24 June 2013). et al.
- BLAT—the BLAST-like alignment tool. Genome Res. 12, 656–664 (2002).
- The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079 (2009). et al.
- Predicting the effects of coding non-synonymous variants on protein function using the SIFT algorithm. Nat. Protoc. 4, 1073–1081 (2009). , &
- A method and server for predicting damaging missense mutations. Nat. Methods 7, 248–249 (2010). et al.
- MutationTaster evaluates disease-causing potential of sequence alterations. Nat. Methods 7, 575–576 (2010). , , &
- Pyrosequencing sheds light on DNA sequencing. Genome Res. 11, 3–11 (2001).
- Emerging kinetics of BCR-ABL1 mutations and their effect on disease outcomes in chronic myeloid leukemia patients with imatinib failure. Leuk. Res. 37, 43–49 (2013). et al.
- Studying copy number variations using a nanofluidic platform. Nucleic Acids Res. 36, e116 (2008). , &
- Mathematical analysis of copy number variation in a DNA sample using digital PCR on a nanofluidic device. PLoS ONE 3, e2876 (2008). , &
- High-resolution characterization of a hepatocellular carcinoma genome. Nat. Genet. 43, 464–469 (2011). et al.
- A robust algorithm for copy number detection using high-density oligonucleotide single nucleotide polymorphism genotyping arrays. Cancer Res. 65, 6071–6079 (2005). et al.
- Highly sensitive method for genomewide detection of allelic composition in nonpaired, primary tumor specimens by use of affymetrix single-nucleotide-polymorphism genotyping microarrays. Am. J. Hum. Genet. 81, 114–126 (2007). et al.
- Genomewide screening of DNA copy number changes in chronic myelogenous leukemia with the use of high-resolution array-based comparative genomic hybridization. Genes Chromosom. Cancer 45, 482–494 (2006). et al.
- Gain-of-function of mutated C-CBL tumour suppressor in myeloid neoplasms. Nature 460, 904–908 (2009). et al.
- Tissue-specific demethylation in CpG-poor promoters during cellular differentiation. Hum. Mol. Genet. 20, 2710–2721 (2011). et al.
- Potent vaccine therapy with dendritic cells genetically modified by the gene-silencing–resistant retroviral vector GCDNsap. Mol. Ther. 13, 301–309 (2006). , , , &
- Isolation, characterization, and genetic complementation of a cellular mutant resistant to retroviral infection. Proc. Natl. Acad. Sci. USA 103, 15933–15938 (2006). et al.
- Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc. Natl. Acad. Sci. USA 98, 31–36 (2001). &
- Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nat. Methods 5, 621–628 (2008). , , , &
- NCBI Reference Sequences (RefSeq): current status, new features and genome annotation policy. Nucleic Acids Res. 40, D130–D135 (2012). , , &
This work was supported by Grants-in-Aid from the Ministry of Health, Labor and Welfare of Japan and KAKENHI (23249052, 22134006 and 21790907; S.O.), the Industrial Technology Research Grant Program from the New Energy and Industrial Technology Development Organization (NEDO; S.O.) (08C46598a), NHRI-EX100-10003NI Taiwan (L.-Y.S.), the project for development of innovative research on cancer therapies (p-direct; S.O.) and the Japan Society for the Promotion of Science through the Funding Program for World-Leading Innovative R&D on Science and Technology, initiated by the Council for Science and Technology Policy (CSTP; S.O.). We thank Y. Hayashi (Gunma Children's Medical Centre), R.C. Mulligan (Harvard Medical School), S. Sugano (The University of Tokyo), M. Onodera (National Center for Child Health and Development, Japan) and L. Ström (Karolinska Institute) for providing materials. We thank Y. Yamazaki for cell sorting. We also thank Y. Mori, M. Nakamura, N. Mizota and S. Ichimura for their technical assistance and M. Ueda for encouragement.
- Supplementary Text and Figures (3,413 KB)
Supplementary Figures 1–13 and Supplementary Tables 1–18