The Potential Role of PHF6 As an Oncogene: a Genotranscriptomic/Proteomic Meta-Analysis

Tumor Biol. DOI 10.1007/s13277-015-4250-0

ORIGINAL ARTICLE

The potential role of PHF6 as an oncogene: a genotranscriptomic/proteomic meta-analysis

Mohammadreza Hajjari1 & Abbas Salavaty1 & Francesco Crea2 & Young Kee Shin3

Received: 18 August 2015 /Accepted: 13 October 2015 # International Society of Oncology and BioMarkers (ISOBM) 2015

Abstract Epigenetic complexes control various pathways that although the mutation rate of PHF6 is relatively low, within the cells. Their abnormalities can be involved in it is mutated in some tumor types. In addition, our data the initiation and the progression of different types of for 40 epigenetic genes showed that missense and non- cancer. Nucleosome remodeling and deacetylase (NuRD) sense mutations were associated with overexpression and is an epigenetic complex that comprises several subunits under-expression, respectively. Our results suggest that such as PHF6. Although PHF6 is reported as a tumor PHF6 may function as an oncogenic factor in several suppressor in some of the hematopoietic malignancies, its types of cancer. We also hypothesize that PHF6 may also function is still challenging in other cancers. Our study play its role in a tissue-specific manner. Our findings sug- aimed at investigating the role of PHF6 in different types gest further investigations regarding the exact role of of cancer. We conducted a meta-analysis of PHF6 in hu- PHF6 in tumor types. man cancers at genomic, transcriptomic, and proteomic levels. For this purpose, we acquired the data from several Keywords Epigenetics . PHF6 . NuRD . Meta-analysis databases, and tried to statistically integrate and analyze the data in order to find the potential role of PHF6 in different tumors. The results demonstrated that although Introduction PHF6 has been previously known as a tumor suppressor gene, it was remarkably overexpressed in many cancer The expression patterns of human genes are controlled by types such as breast and colorectal cancers. Notably, various epigenetic regulators such as histone acetyl transfer- PHF6 was under-expressed in a few types of cancer, in- ases, histone methyl transferases, and chromatin remodelers cluding esophageal tumors. Moreover, the results indicated [1]. In this context, histone posttranslational modifications serve as docking sites for the reader modules involved in chromatin modification and remodeling [2]. Gene expression Electronic supplementary material The online version of this article deregulation, which is caused by aberrant epigenetic modifi- (doi:10.1007/s13277-015-4250-0) contains supplementary material, which is available to authorized users. cations, has been reported in different tumor types and usually occurs through promoter hypermethylation and chromatin re- * Mohammadreza Hajjari pression of tumor-associated genes [3]. It is shown that cells [email protected]; [email protected] within a field defect typically have an increased frequency of epigenetic modifications, a fact which may be a key factor in 1 Department of Genetics, Faculty of Science, Shahid Chamran the progression of cancer [4]. In addition, epithelial- University of Ahvaz, Ahvaz, Iran mesenchymal transition (EMT), which is an important and 2 Experimental Therapeutics, BC Cancer Research Centre, 675 W 10th known event in the process of cancer progression, is signifi- Avenue, Vancouver, British Columbia V5Z 1L3, Canada cantly regulated by epigenetic mechanisms [5]. Studies of 3 Laboratory of Molecular Pathology and Cancer Genomics, different types of cancer have shown that these aberrant epi- Department of Pharmacy, College of Pharmacy, Seoul National genetic mechanisms are mostly due to a high frequency of University, Seoul, Korea mutation across epigenetic regulators [6]. Tumor Biol.

Most of the epigenetic regulators are parts of the pro- proteomic levels. The data showed that PHF6 may have tein complexes known as epigenetic complexes. The nu- an important role in other types of cancer as well as leu- cleosome remodeling and deacetylase (NuRD) complex is kemia. Since there is a known association between NuRD widely conserved, functions as a transcriptional co-regula- and PRC2 complexes, we specifically investigated the as- tor, and exerts both histone deacetylase and nucleosome sociation between PHF6 and ASXL1. remodeling activities [7, 8]. In addition, NuRD may be involved in gene silencing by DNA methylation [9]. NuRD is a key factor in early stages of embryonic stem Materials and methods cell differentiation and the reprogramming of somatic cells to induced pluripotent stem cells. Also, abnormalities in Mutation and CNV analysis of PHF6 in different cancers several NuRD proteins are associated with cancer and ag- ing [7]. Data from tissue samples with mutations in PHF6 were ob- PHF6 is one of the subunits of NuRD complex. Its tained from the COSMIC database (Supplementary Table S1). gene is located on the X chromosome and encodes a The obtained data were used to analyze the distribution of the protein with four nuclear localization sequences and mutations in different types of cancer. Then, the mutation rates two PHD-type zinc finger domains. These two domains and CNV (copy number variation) gain/loss of PHF6 in dif- are highly conserved domains which are important in ferent tumor types were extracted from cBioPortal (2015) recognizing histone modifications and protein-DNA in- (http://www.cbioportal.org/public-portal/)[21, 22], which is teractions [10, 11]. Transient transfection studies using designed to store and display the somatic mutation tagged PHF6 indicate that this protein has a diffuse information and the related details of genes. Also, the ICGC nuclear staining and significant nucleolar accumulation. Data Portal (https://dcc.icgc.org/) [23] was searched to find The PHD-type zinc finger domains of PHF6 might be the distribution of those types of cancer in which PHF6 is suggestive of PHF6 involvement in transcriptional reg- mutated. ulations [12]. Based on some studies on leukemia, PHF6 is known as The expression analysis of PHF6 RNA in different tumors a potential tumor suppressor gene [13]. Mutations in PHF6 gene are recurrent genetic abnormalities in T cell The number of tissues with up/downregulation of PHF6 RNA acute lymphoblastic leukemia (T-ALL) [14]. Also, PHF6 was achieved from the COSMIC database. Then, the levels of gene is commonly inactivated in T-ALL, which is indica- PHF6 RNA in tumor tissues with a significant change in tive of its tumor suppressive role in this type of cancer. PHF6 were extracted from COSMIC (Supplementary These studies support the idea that PHF6 is a tumor sup- Table S2). Finally, t score, P value, and the mean of PHF6 pressor that plays an important role in the pathogenesis of level in each study were measured (Z score=0 for normal hematologic tumors [15]. expression) by R statistical software (R Development Core A mutational analysis in AML patients showed that Team (2014), http://www.r-project.org). In addition, the the mutations in PHF6 and ASXL1 were associated with Oncomine platform (https://www.oncomine.org/) [24, 25] reduced overall survival [16]. ASXL1 is one of the sub- was searched to obtain the expression of PHF6 RNA in units of polycomb repressive complex 2 (PRC2). different tissues. Then, the over/under-expression of PHF6 NuRD-mediated deacetylation of histone H3K27 spec- was extracted and compared between normal and cancerous ifies the recruitment of PRC2 in embryonic stem (ES) tissues. Furthermore, the Expression Atlas database (http:// cells. Such deacetylation promotes PRC2 recruitment www.ebi.ac.uk/gxa)[26, 27] was used to check the over/ and the subsequent H3K27 trimethylation at NuRD tar- under-expression of PHF6 in common cancer types. The list get promoters. Hence, the knockdown of NuRD com- of common cancer types was extracted from NCI (http://www. plex results in the deregulation of several bivalent cancer.gov). The Cancer Genomic Browser (https://genome- genes, increasing H3K27ac and reducing H3K27me3 cancer.ucsc.edu/) was used to find the association of PHF6 as well [17, 18]. In other words, it seems that NuRD with some clinicopathological parameters in tumor complex collaborates with PRC2 to suppress its target tissues. genes and is involved in tumorigenesis [19]. Although different studies have indicated the tumor sup- The analysis of PHF6 protein in different cancers pressor role of PHF6 in leukemia, its role in other types of cancer is unclear. Thus, we evaluated PHF6 to determine The Human protein atlas (http://www.proteinatlas.org)[28, whether it may have any potential role in other cancers. 29] was used to check the protein expression level of PHF6 Accordingly, we conducted a meta-analysis of PHF6 in between tumor and normal tissues (Supplementary Table S3). human cancers at the genomic, transcriptomic, and Moreover, the TAG database (http://www.binfo.ncku.edu.tw/ Tumor Biol.

TAG/DS/) [30] was used to search for the putative oncogenic including TSGene, Network of Cancer Genes, Cancerrxgene, domains of PHF6 gene. For this purpose, the sequence of and COSMIC were searched to find out their association with PHF6 was retrieved from the NCBI (http://www.ncbi.nlm. cancer (Supplementary Table S4). nih.gov/). The mutation distribution and the overall percentage of over-/under-expression of all the 40 genes were drawn from the COSMIC database (Feb. 2015) (http://cancer.sanger.ac.uk/ The genotranscriptomic comparison of PHF6 with other cancergenome/projects/cosmic/)[20], which is the catalogue members of epigenetic complexes of somatic mutations in cancer. These data are available in the Supplementary Table S5. The comparison of different We compiled a list of eight epigenetic complexes [6], which epigenetic genes involved in various types of cancer showed altogether consist of 40 genes. Then, several databases that the overall percentage of overexpressed tissues was over

Fig. 1 The comparison of different tumors based on the percentage of a the mutation rate, b CNV rate, and c the overexpression/under-expression of PHF6 gene. The data are drawn from the COSMIC database Tumor Biol. five folds more than the overall percentage of under-expressed Results tissues for 9 of 40 genes. These nine notable genes consist of MBD3, PHF6, JARID2, PHF19, ASXL1, EZH2, UHRF1, PHF6 locus analysis In 16,769 tumor samples (COSMIC RAG2,andMLL3 (KMT2C) (Supplementary Table S5). Based database, Feb. 2015), including 34 different types, 162 tissues on these data, a heatmap was achieved by the R statistical were recorded as having mutations in PHF6. Interestingly, software, using pheatmap package (https://cran.r-project.org/ different types of mutations, except in-frame insertion and web/packages/pheatmap/index.html) and its default clustering deletion types (0 %), were found in tumors (Supplementary method and distance function, separately for 40 genes and 9 Table S5). We found that the most of the mutations were notable genes. These two heatmaps helped with the analysis of missense types (40.12 %). The nonsense mutations had also the association between mutation distributions and over/ the ratio of 19.14 % among the mutation frequencies in tis- under-expressions as well as the analysis of the similarities sues. As shown in Fig. 1a, the mutation rate of PHF6 in between different genes. various cancers is relatively low. The data showed that the hematopoietic and lymphoid tumors have the highest mutation rate. The majority of these The association between PHF6 and ASXL1 mutations were related to acute lymphoblastic T cell leukemia. It is also noticeable that the mutations were distributed evenly The COXPRESdb database (http://coxpresdb.jp/)[31] was in the PHF6 protein structure. used to check the overall coexpression of PHF6 and ASXL1. We also analyzed the amplification profile of the PHF6 Also, the Coremine database (www.coremine.com)was locus in human neoplasms using the COSMIC database. The searched to find the interactors of PHF6.Coremineisa result was that PHF6 has relatively low rates of CNV gain and database that functions as a text mining program. loss in most of the tumor tissues except urinary tract cancer (Fig. 1b). Furthermore, the ICGC data portal showed that the Statistical analysis rate of mutations in PHF6 is low in different samples, which is in concordance with the data extracted from COSMIC All statistical analyses were done by the R statistical software database. (R Development Core Team (2014), http://www.r-project.org) The genomic analysis of PHF6 locus through cBioportal and Graphpad. In all of the analyses, the P value<0.05 was showed that 19 cancer tissues have an intermediate rate of used as the significance threshold. In this study, the t test and mutations. Various types of alteration were found in cancer the R-plot analysis were done by the R software. Moreover, tissues except prostate (gene was amplified in 8.2 % of cases) the two-way ANOVAand the contingency table analysis were and leukemia (gene was mutated in 3.1 and 3.2 % of cases in done by Graphpad. different studies).

Fig. 2 The overexpression of PHF6 in different types of cancer. The data are drawn from the COSMIC database based on the Z score. PHF6 levels in tumor tissues with significant changes in the expression level have been analyzed by t test Tumor Biol.

than the under-expressed ones (P value=0.0057). We analyzed the expression level of PHF6 in tumors with a significant change in PHF6 RNA, compared with normal tissues, in order to get the mean Z score (Fig. 2). The distribution analysis of mutations in PHF6 drawn from the COSMIC database showed that 67.7, 16.1, and 16.2 % of nonsense mutations were related to lymphoid neoplasm, hematopoietic neoplasm, and other carcinomas, respectively. Moreover, 24.2, 19.7, and, 56.1 % of missense mutations were related to lymphoid neoplasm, hematopoietic neoplasm, and other different carcinomas, respectively (Fig. 3). The Fisher test showed that the distribution of mutation types is signifi- Fig. 3 The distribution of mutations in PHF6 among different cancer P tissues. The data are drawn from the COSMIC database cantly different in various types of cancer ( value<0.0001). The analysis of PHF6 expression in normal tissues, Altogether, the low rate of mutations and the copy using Oncomine database, indicated that PHF6 has the number variation in tumor samples may suggest that the highest expression level in lymph nodes compared to other alterations at the PHF6 locus are unlikely to impact the tissues (Fold change, 3.013; P value, 6.67E−5) cancer progression. (Supplementary Figure S1). The comparison of PHF6 expression between normal and cancerous tissues also indi- The analysis of the expression level of PHF6 in different cated that PHF6 was notably overexpressed in most of the types of cancer tumors compared to their corresponding normal tissues. However, PHF6 was notably under-expressed in esopha- Although PHF6 has been previously known as a tumor suppres- geal tumors (Supplementary Figure S2). sor gene, our data derived from the Cosmic database indicated The analysis of the over/under-expression of PHF6,using that PHF6 was notably overexpressed in some types of cancer. Expression Atlas database, indicated that PHF6 was The percentage of tissues overexpressing PHF6 is notable in overexpressed in most of common types of cancer, while it some tumor types (Fig. 1c). The two-way ANOVA showed that was under-expressed in some others such as acute myeloid the percentage of overexpressed tumors is significantly higher leukemia (P value<0.05) (Supplementary Figure S3).

Table 1 Comparison of data indicating the overexpression of PHF6 RNA in different cancers. The data are P values derived from three different databases. The combined P value is also calculated through chi-square distribution (last column)

Cancer Cosmic Oncomine Gene expression Atlas Combined P value

Breast cancer 2.20e−16 1.06e−16 8.09e−6 CV: 8.090000000326e−6 CP <0.001 Colorectal cancer 3.57e−12 8.78e−16 0.0032767 CV: 0.003276700003570878 CP <0.001 Cervical cancer 0.001112 1.56e−4 No data CV: 0.001268 CP <0.001 Ovarian cancer 2.53e−8 0.017 No data CV: 0.0170000253 CP <0.001 Kidney cancer 0.002092 No data 0.0144069 CV: 0.0164989 CP <0.001 Chronic lymphocytic leukemia No data 2.4e−24 0.0094184 CV: 0.0094184 CP <0.001 Liver cancer 2.94e−55.68e−6 No data CV: 0.00003508 CP <0.001 Lung cancer 2.2e−16 1.45e−51.3e−8 CV: 0.00001451300000022 CP <0.001 Stomach cancer 0.0002431 0.003 No data CV: 0.0032431 CP <0.001 Prostate cancer No data 1.79e−42.45e−6 CV: 0.00018145 CP <0.001

CV chi-square critical value, CP cumulative probability Tumor Biol.

In order to compare the expression data of PHF6 in differ- cancer. The cancer genomic browser showed that the PHF6 ent types of cancer and evaluate its potential as an oncogenic level has a negative correlation with genes which were differ- factor, we showed the tumors which have been reported to entially expressed under HDAC inhibitors (data not shown). overexpress PHF6 and verified by at least two databases. We calculated the combined P value by chi-square test The analysis of PHF6 protein level and characteristics through the chi-square calculator (http://stattrek.com/online- in different cancers calculator/chi-square.aspx)(Table1). The analysis of the Human Protein Atlas demonstrated The association of PHF6 expression that PHF6 had a high expression level in most of the with clinicopathological parameters tumor tissues such as lymphoma, glioma, colorectal, and cervical cancer tissues (Fig. 4). The analysis also indi- The expression of PHF6 showed a correlation with the ex- cated that, based on immunofluorescence experiments, pression of genes involved in EMT in different types of PHF6 protein is mostly localized within the nucleus

Fig. 4 The expression level of PHF6 protein in cancerous tissues. a The b The immunohistochemistry of PHF6 in different tumor tissues (The graphs are based on the number of patients with various PHF6 staining data and images are drawn from the Human Protein Atlas) intensity (the data and images are drawn from the Human Protein Atlas). Tumor Biol.

The analysis of COXPRESdb database revealed a significant correlation of expression pattern between PHF6 and ASXL1 (Supplementary Figure S4). The COXPRESdb is a database that provides co-regulated gene relationships to estimate gene functions. The Coremine database also sug- gested that ASXL1 was a notable interactor of PHF6 (significance, 4.48E−5).

Discussion

The plant homeodomain finger 6 (PHF6) is a subunit of NuRD complex that contains two PHD-type zinc finger domains [32]. Although PHF6 has been known as a tumor suppressor [33, 34], our results demonstrate that PHF6 is Fig. 5 The localization of PHF6 protein within the nucleus and the nucleoli. The data are drawn from the Human Protein Atlas based on overexpressed in several tumors. Since overexpression is immunofluorescence experiments one of the features of oncogenes in cancer cells [35], it is conceivable that PHF6 may function as an oncogenic factor and the nucleoli (Fig. 5). The summary of identified in several types of cancer. TAG domains with TAG database analysis showed that Based on our results, PHF6 is notably overexpressed in the scores of oncogene and the tumor suppressor gene several types of cancer such as colon, breast, stomach, glioma, identity of PHF6 were 0.39 and 0.24, respectively. kidney, ovary, liver, prostate, and cervix tumors. By contrast, there are some studies indicating that PHF6 is notably under- expressed in some tumors such as esophagus cancer. It is of The association between PHF6 and ASXL1 note that there was an inconsistency between the results drawn from various databases for a few cancers including melanoma The R-plot drawn for 40 genes indicated that the genes and leukemia. It seems that these types of cancer should be PHF6, MBD3, and KDM1A, which are the subunits of studied more in the future, considering the different types and NuRD complex, had the highest similarity with ASXL1, samples for each study. PHF19,andEED, which are the subunits of PRC2 com- In addition, proteomic analysis of PHF6 demonstrates that plex, respectively (P value=0.025) (Fig. 6). The R-plot PHF6 has a high expression level in most of the cancerous showed that the overexpression and under-expression had tissues such as lymphoma, glioma, colorectal, and cervical the most association with missense and nonsense, respec- cancer tissues. We hypothesize that PHF6 may exert its role tively (P value<0.05). Besides, the R-plot drawn for nine in a tissue-specific manner. The transcriptomic and proteomic notable genes indicated that PHF6 and MBD3 had the analyses of PHF6 altogether indicate that while PHF6 may be highest similarity with ASXL1 and PHF19, respectively. overexpressed and function as an oncogenic factor in some

Fig. 6 The heatmap showing the association between mutation distributions and over/under- expressions for 40 genes Tumor Biol. types of cancer, it may be under-expressed in some others. Compliance with ethical standards Based on our results, it is also conceivable that PHF6 may Conflicts of interest None function in a tissue-specific manner. The proteomic analysis of PHF6 moreover shows that PHF6 is mostly localized within the nucleus and the nucleoli. These data may be informative of the role of PHF6 in the regulation of transcription in a tissue- PHF6 specific manner. The genomic analysis of demonstrates References that it is mutated in several types of cancer such as prostate, bladder, uterine, ovarian, and liver cancers. These mutations 1. Wee S, Dhanak D, Li H, Armstrong SA, Copeland RA, Sims R, are mostly amplifications and point mutations, which may et al. Targeting epigenetic regulators for cancer therapy. Ann N Y result in the overexpression of PHF6. Acad Sci. 2014;1309:30–6. doi:10.1111/nyas.12356. Considering the 40 desired epigenetic genes, the distri- 2. Patel DJ, Wang Z. Readout of epigenetic modifications. Annu Rev bution of missense mutations has the most resemblance Biochem. 2013;82:81–118. doi:10.1146/annurev-biochem- 072711-165700. with the distribution of overexpression. Furthermore, the 3. Bert SA, Robinson MD, Strbenac D, Statham AL, Song JZ, Hulf T, distribution of nonsense mutations has the most resem- et al. Regional activation of the cancer genome by long-range epi- blance with the distribution of under-expression. The genetic remodeling. Cancer Cell. 2013;23:9–22. doi:10.1016/j.ccr. analysis of substitution missense and nonsense distribu- 2012.11.006. tion in different types of cancer reveals interesting facts 4. Bernstein C, Nfonsam V, Prasad AR, Bernstein H. Epigenetic field defects in progression to cancer. World J Gastrointest Oncol. that support our hypothesis indicating the function of 2013;5:43–9. doi:10.4251/wjgo.v5.i3.43. PHF6 in a tissue-specific manner. Substitution nonsense 5. Kiesslich T, Pichler M, Neureiter D. Epigenetic control of mutations of PHF6 aremostlyassociatedwithlymphoid epithelial-mesenchymal-transition in human cancer. Mol Clin and hematopoietic neoplasms, while substitution missense Oncol. 2013;1:3–11. doi:10.3892/mco.2012.28. mutations are mostly associated with different carcino- 6. Huether R, Dong L, Chen X, Wu G, Parker M, Wei L, et al. The PHF6 landscape of somatic mutations in epigenetic regulators across 1, mas. Although the function of in some of lymphoid 000 paediatric cancer genomes. Nat Commun. 2014;5:3630. doi: and hematopoietic neoplasms such as T-ALL and AML 10.1038/ncomms4630. has been known already, its function is still unknown in 7. Alqarni SSM, Murthy A, Zhang W, Przewloka MR, Silva APG, most of carcinomas. These data suggest that PHF6 deserves Watson AA, et al. Insight into the architecture of the NuRD com- more focus in future experimental studies on cancer. plex: structure of the RbAp48-MTA1 subcomplex. J Biol Chem. 2014;289:21844–55. doi:10.1074/jbc.M114.558940. Some studies indicate that about half of the disease- 8. Dos Santos RL, Tosti L, Radzisheuskaya A, Caballero IM, Kaji K, associated mutations of PHF6 are located within its second Hendrich B, et al. MBD3/NuRD facilitates induction of extended PHD domain (ePHD2), indicating the functional pluripotency in a context-dependent manner. Cell Stem Cell. – importance of such domain [32]. However, we found that 2014;15:102 10. doi:10.1016/j.stem.2014.04.019. 9. Zhang Y, Ng HH, Erdjument-Bromage H, Tempst P, Bird A, mutations are almost evenly distributed in different domains Reinberg D. Analysis of the NuRD subunits reveals a histone of PHF6 protein. We believe that more studies should be done deacetylase core complex and a connection with DNA methylation. in order to find the impact of various mutations on the PHF6 Genes Dev. 1999;13:1924–35. doi:10.1101/gad.13.15.1924. function. 10. Chao MM, Todd MA, Kontny U, Neas K, Sullivan MJ, Hunter AG, Our results moreover reveal that three subunits of NuRD et al. T-cell acute lymphoblastic leukemia in association with Börjeson-Forssman-Lehmann syndrome due to a mutation in complex show high similarities in the distribution of mutation PHF6. Pediatr Blood Cancer. 2010;55:722–4. doi:10.1002/pbc. and the over/under-expression with three subunits of PRC2 22574. complex. These data are supportive of the previously known 11. Patel JP, Levine RL. How do novel molecular genetic markers associations between NuRD and PRC2. Among these three influence treatment decisions in acute myeloid leukemia? – PHF6 ASXL1 Hematol Am SociHematol Educ Prog. 2012;2012:28 34. doi:10. subunits, shows high similarities with (a subunit 1182/asheducation-2012.1.28. of PRC2 complex). Our results also demonstrate a significant 12. Lower KM, Turner G, Kerr BA, Mathews KD, Shaw MA, Gedeon correlation of expression pattern between PHF6 and ASXL1. AK, et al. Mutations in PHF6 are associated with Börjeson- Accordingly, the identification of the interactions between Forssman-Lehmann syndrome. Nat Genet. 2002;32:661–5. doi: PHF6 and ASXL1 and in a broader range between NuRD 10.1038/ng1040. 13. Kunze K, Gamerdinger U, Leßig-Owlanj J, Sorokina M, Brobeil A, and PRC2 complexes may help elucidate the exact role of Tur MK, et al. Detection of an activated JAK3 variant and a Xq26.3 PHF6 in different types of cancer. In addition, the identifica- microdeletion causing loss of PHF6 and miR-424 expression in tion of the exact role of PHF6 may introduce PHF6 as a myelodysplastic syndromes by combined targeted next generation potential biomarker for the prognosis and the early diagnosis sequencing and SNP array analysis. Pathol Res Pract. 2014;210: 369–76. doi:10.1016/j.prp.2014.02.006. of different types of cancer. In addition, this identification can PHF6 14. Wang Q, Qiu H, Jiang H, Wu L, Dong S, Pan J, et al. Mutations of help designing pharmaceuticals that specifically target PHF6 are associated with mutations of NOTCH1, JAK1 and rear- or its suppressor in different tumors. rangement of SET-NUP214 in T-cell acute lymphoblastic leukemia. Tumor Biol.

Haematologica. 2011;96:1808–14. doi:10.3324/haematol.2011. 25. Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh 043083. D et al. ONCOMINE: a cancer microarray database and integrated 15. Van Vlierberghe P, Patel J, Abdel-Wahab O, Lobry C, Hedvat CV, data-mining platform. Neoplasia (New York, NY). 2004;6:1-6. Balbin M, et al. PHF6 mutations in adult acute myeloid leukemia. 26. Fonseca NA, Marioni J, Brazma A. RNA-Seq gene profiling—a Leuk Off J Leuk Soc Am Leuk Res Fund UK. 2011;25:130–4. doi: systematic empirical comparison. PLoS ONE. 2014;9:e107026-e. 10.1038/leu.2010.247. doi:10.1371/journal.pone.0107026. 16. Patel JP, Gönen M, Figueroa ME, Fernandez H, Sun Z, Racevskis J, 27. Petryszak R, Burdett T, Fiorelli B, Fonseca NA, Gonzalez-Porta M, et al. Prognostic relevance of integrated genetic profiling in acute Hastings E, et al. Expression Atlas update—a database of gene and myeloid leukemia. N Engl J Med. 2012;366:1079–89. doi:10.1056/ transcript expression from microarray- and sequencing-based func- NEJMoa1112304. tional genomics experiments. Nucleic Acids Res. 2014;42:D926– 17. Reynolds N, Salmon-Divon M, Dvinge H, Hynes-Allen A, 32. doi:10.1093/nar/gkt1270. Balasooriya G, Leaford D, et al. NuRD-mediated deacetylation of 28. Berglund L, Björling E, Oksvold P, Fagerberg L, Asplund A, H3K27 facilitates recruitment of polycomb repressive complex 2 to Szigyarto CA-K, et al. A genecentric Human Protein Atlas for direct gene repression. EMBO J. 2012;31:593–605. doi:10.1038/ expression profiles based on antibodies. Mol Cell Proteomics. – emboj.2011.431. 2008;7:2019 27. doi:10.1074/mcp.R800013-MCP200. 18. Voigt P, Tee WW, Reinberg D. A double take on bivalent pro- 29. Uhlén M, Björling E, Agaton C, Szigyarto CA-K, Amini B, moters. Genes Dev. 2013;27:1318–38. doi:10.1101/gad.219626. Andersen E, et al. A human protein atlas for normal and cancer 113. tissues based on antibody proteomics. Mol Cell Proteomics. 2005;4:1920–32. doi:10.1074/mcp.M500279-MCP200. 19. Zhu W, Wei F, Wang XI, Wang H. 427: NuRD complex co-operates 30. Chen JS, Hung WS, Chan HH, Tsai SJ, Sunny SH. In silico iden- with PRC2 to suppress gene expression. Eur J Cancer. 2014;50: tification of oncogenic potential of fyn-related kinase in hepatocel- S102–S3. doi:10.1016/S0959-8049(14)50382-6. lular carcinoma. Bioinformatics. 2013;29:420–7. doi:10.1093/ 20. Forbes SA, Beare D, Gunasekaran P, Leung K, Bindal N, ’ bioinformatics/bts715. Boutselakis H, et al. COSMIC: exploring the world s knowledge 31. Okamura Y, Aoki Y, Obayashi T, Tadaka S, Ito S, Narise T, et al. of somatic mutations in human cancer. Nucleic Acids Res. – COXPRESdb in 2015: coexpression database for animal species by 2014;43:D805 D11. doi:10.1093/nar/gku1075. DNA-microarray and RNAseq-based expression data with multiple 21. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, quality assessment systems. Nucleic Acids Res. 2014;43:D82–D6. et al. The cBio Cancer Genomics Portal: an open platform for ex- doi:10.1093/nar/gku1163. ploring multidimensional cancer genomics data. Cancer Discov. 32. Liu Z, Li F, Ruan K, Zhang J, Mei Y, Wu J, et al. Structural and – 2012;2:401 4. doi:10.1158/2159-8290.CD-12-0095. functional insights into the human Börjeson-Forssman-Lehmann 22. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO syndrome-associated protein PHF6. J Biol Chem. 2014;289: et al. Integrative analysis of complex cancer genomics and clinical 10069–83. doi:10.1074/jbc.M113.535351. profiles using the cBioPortal. Science signaling. 2013;6:pl1-pl. doi: 33. Ariës IM, Gutierrez A. An X-linked tumor suppressor in T-ALL. 10.1126/scisignal.2004088. Blood. 2015;125:3–4. doi:10.1182/blood-2014-11-607838. 23. Zhang J, Baran J, Cros A, Guberman JM, Haider S, Hsu J et al. 34. Perez-Garcia A, Ambesi-Impiombato A, Hadler M, Rigo I, LeDuc International cancer genome consortium data portal-a one-stop shop CA, Kelly K, et al. Genetic loss of SH2B3 in acute lymphoblastic for cancer genomics data. Database. 2011;2011:bar026-bar. doi:10. leukemia. Blood. 2013;122:2425–32. doi:10.1182/blood-2013-05- 1093/database/bar026. 500850. 24. Rhodes DR, Kalyana-Sundaram S, Mahavisno V,Varambally R, Yu 35. Jeong HM, Kwon MJ, Shin YK. Overexpression of cancer- J, Briggs BB, et al. Oncomine 3.0: genes, pathways, and networks associated genes via epigenetic derepression mechanisms in gyne- in a collection of 18,000 cancer gene expression profiles. Neoplasia cologic cancer. Frontiers Oncol. 2014;4:12. doi:10.3389/fonc. (New York NY). 2007;9:166–80. doi:10.1593/neo.07112. 2014.00012.