Similarity of Regulatory Network Between Leukemia Stem Cells and Normal Hemopoietic Stem Cells

Infection International; 2017; 6 (4): 129-140

Original Article Zhi-Jian Li1, Xing-Ling Sui2, Xue-Bo Yang2, Wen Sun2* Similarity of regulatory network between leukemia stem cells and normal hemopoietic stem cells

DOI: 10.1515/ii-2017-0165 Received December 06, 2017; accepted January 18, 2018; published online April 10, 2018

Abstract

To reveal the biology of AML, we compared gene-expression profiles between normal hematopoietic cells from 38 healthy donors and leukemic blasts (LBs) from 26 AML patients. We defined the comparison of LB and unselected BM as experiment 1, LB and CD34+ isolated from BM as experiment 2, LB and unselected PB as experiment 3, and LB and CD34+ isolated from PB as experiment 4. Then, protein–protein interaction network of DEGs was constructed to identify critical genes. Regulatory impact factors were used to identify critical transcription factors from the differential co-expression network constructed via reanalyzing the microarray profile from the perspective of differential co-expression. Gene ontology enrichment was performed to extract biological meaning. The comparison among the number of DEGs obtained in four experiments showed that cells did not tend to differentiation and CD34+ was more similar to cancer stem cells. Based on the results of protein–protein interaction network, CREBBP, F2RL1, MCM2, and TP53 were respectively the key genes in experiments 1, 2, 3, and 4. From gene ontology analysis, we found that immune response was the most common one in four stages. Our results might provide a platform for determining the pathology and therapy of AML.

Keywords: acute myeloid leukemia, bone marrows, CD34+, peripheral blood

1 Introduction

Leukemia, characterized by abnormal proliferation leukocytes and their precursors, is a prevalent malignancy of hematologic organs [1]. Acute myeloid leukemia (AML) is a heterogeneous clonal disorder of hematopoietic progenitor cells and the most common malignant myeloid disorder in adults [2]. It arises in precursors of myeloid, erythroid, megakaryocytic, and monocytic cell lineages [3]. These myeloid cells differentiated into hematopoietic progenitor cells clump together and form the blast cells. Accumulated blast cells in the bone marrow (BM) result in the onset of diseases like anemia, blood clotting, and severe infection leading to the death from AML [4]. The incidence of AML increases significantly with age, with 4 cases per 100,000 people annually in the sixth decade of life to >20 cases per 100,000 in the ninth decade of life [5]. Thus, a better understanding of the molecular mechanisms of AML may be helpful to develop more specific and less toxic therapies for AML. AML has high levels of genetic heterogeneity, and this makes targeted therapies challenging [6]. In order to explore the biological process of this complicated disease, researchers used microarray technologies to

1Shandong Science and Technology Press Co., Ltd., Jinan 250002, China 2Beijing Splinger Medical Research Institute, Beijing 100054, China *Correspondence: Wen Sun, E-mail: [email protected] 130 Zhi-Jian Li et al. measure the expression of numerous genes in an individual sample [7]. For example, Valk et al. [8] speculated several genes such as IL5Ra, IL2Ra, and PIM1 were associated with AML using unsupervised analyses based on gene-expression profiling. Moreover, Mrózek et al. [9] analyzed the leukemic blasts (LBs) from pretreatment marrow or blood of patients with AML and discovered multiple submicroscopic genetic alterations, including internal tandem duplication of the FLT3 gene, mutations in the NPM1 gene, partial tandem duplication of the MLL gene, high expression of the BAALC gene, and mutations in the CEBPA gene. Although microarray studies on AML patients have extracted several novel biomarkers, few studies have deeply compared the expression profiles between normal hematopoietic cells and AML blasts. Different conditions have their own characteristic gene-expression profiles, and these profiles reflect the state of the cells and shed an insight into possible effectors to influence the phenotype. A characteristic of microarray gene-expression profiles is the ability to distinguish disease states from normal conditions [10,11]. This can form the foundation of a disease–gene–drug connection, with the identification of potential candidate therapeutics [12]. Thus, microarray studies comparing AML blasts and normal hematopoietic cells may discover critical genes or functions for the development of specific and less toxic therapies [13]. AML blasts are different in the differentiation stages in patients and a single patient, which poses a challenge for researches comparing AML blasts to normal hematopoietic cells. To find more information about AML, we conducted microarray studies comparing gene profiles between AML blasts and normal hematopoietic cells. Moreover, we examined normal hematopoietic cells at different stages of maturation (immature CD341 cells, unselected BMs, and unselected peripheral blood (PB)) to compensate for the challenge of identifying the optimal normal hematopoietic counterpart for AML comparisons. Specifically, gene-expression profiles of normal hematopoietic cells from healthy donors and LBs from AML patients were compared to identify differentially expressed genes (DEGs); in addition, the DEG numbers of the 4 experiments were compared. Then, protein–protein interaction (PPI) network was constructed to identify critical genes. Regulatory impact factors (RIFs) were used to identify critical transcription factors (TFs) from the differential co-expression network (DCN) constructed via reanalyzing the microarray profile from the perspective of differential co-expression. Gene ontology (GO) functional enrichment analysis was performed to extract biological meaning. Our study can help us to find biomarkers for early detection and diagnosis of AML.

2 Methods

2.1 Acquisition of gene-expression profile

The microarray expression profile of GSE9476 [13] was downloaded from Gene Expression Omnibus (http:// www.ncbi.nlm.nih.gov/geo/) database, which is based on the platform of GPL96 [HG-U133A] Affymetrix Human Genome U133A Array. In the gene-expression profile of GSE9476, there were 64 samples, including 38 normal hematopoietic samples from 38 healthy donors and 26 LBs from 26 AML patients. Normal hematopoietic samples included CD34+ selected cells (n = 18), unselected BMs (n = 10), and unselected PB (n = 10). In order to determine if AML-specific expression changes were generalizable across different populations of AML patients, microarray studies compared expression profiles between AML blasts and normal hematopoietic cells at variety of different stages of maturation (immature CD341 cells, unselected BMs, and unselected PB).

2.2 Data pretreatment and identification of DEGs

Before identifying DEGs, we first preprocessed the raw data. To begin with, the probe-level data in CEL files were transferred into expression measures. Then, background was corrected and quartile data were norma- lized by the robust multi-array average algorithm. Afterward, the probes were mapped to the genomics to explore the relationship between them and human gene symbols. The probes were filtered if they did not Potential genes in AML 131 have the corresponding gene symbols. If multiple probes were mapped to a gene symbol, the average value of multiple probes was calculated to be the expression level of the given gene. Then, linear models for microarray data (LIMMA) package [14] and t-test were used in our study to identify DEGs between AML and healthy samples. The false discovery rate (FDR) was calculated to correct the raw P-values using the Benjamini–Hochberg approach [15]. FDR ≤0.05 was regarded as the threshold to identify DEGs, and these genes were defined as candidate genes with AML-specific expression changes.

2.3 Reanalyzing the microarray profile from the perspective of differential co-expression to identify differential co-expression genes (DCGs) and gene links

Many genes together can accomplish biological function, and highly co-expressed genes participate in similar biological processes and pathways [16]. Thus, in our study, two link-based quantitative methods, differential co-expression profile and differential co-expression enrichment, were used to identify DCGs. Differential co-expression profile works on the filtered set of gene co-expression value pairs. Each pair of them is made up with two co-expression values. The values were calculated under two different conditions separately. The subset of co-expression value pairs associated with a particular gene, in two groups for the two conditions separately, was expressed as two vectors X and Y (n is the co-expression neighbor for a gene), which are given in the following:

X = (xi1, xi2, …, xin) (1)

Y = (yi1, yi2, …, yin) (2)

Differential co-expression (dC) of gene i was measured based on the following formula:

22 2 ()xyii11−+()xyii22−+… +−()xyin in dC ()i = (3) n n

To evaluate the significant dC of a gene, permutation test was performed. In the permutation test, the disease and normal samples were randomly permuted, new Pearson correlation coefficients (PCCs) were calculated, gene pairs were filtered based on the new PCCs, and new dC statistics were calculated. The experiment was repeated 1000 times. A large number of permutation dC statistics were used for the formation of an empirical null distribution. Then, the P-value for each gene was estimated. In differential co-expression enrichment experiment, the correlation pairs are divided into three parts: pairs with same signs (N1), pairs with different signs (N2), and pairs with differently signed high co-expression values (N3). Two subsets of DCLs (K1 and K2) were obtained by processing the first two parts with the “Limit

Fold Change” model separately. A total of K = N3 + K1 + K2 DCLs were determined from a total of N gene links.

For a gene (gi), the total number of links (ni) and DCLs, particularly (ki), associated with it were counted. The binomial probability model was used to estimate the significance of the gene being a DCG, and these DCGs were used to construct DCN.

− x xni i     n K K pg()= x    1 −  i ∑ c     (4) n i     xk= i N N

2.4 Identification of TFs from DCN

As documented, TFs play a key regulatory role in controlling gene expression, but it is a challenge to extract the significant TFs from gene-expression data due to their low and sparse expression. In order to solve this problem, an RIF metric was used to identify critical TFs based on gene-expression data [17]. RIF was computed as follows: 132 Zhi-Jian Li et al.

jn= de 1  22 RIF(TFi) =× er−×er (5) ∑ ()1j ij ()2j 2ij n   de ji=

where nde is the number of DCGs; e1 (e2) is the expression value of DCGj in condition 1 (condition 2), and r1ij (r2ij) represents the correlation of TFi and DCGj in condition 1 (condition 2).

2.5 PPI network for DEGs

Human PPIs have the potential for understanding disease mechanisms and signaling cascade, smaller scale, more specific interaction maps [18]. PPI networks are relevant from a systems biology point of view, as they may help to uncover the generic organization principles of functional cellular networks, when both spatial and temporal aspects of interactions are considered [19]. Biological General Repository for Interaction Datasets is an open access archive of genetic and protein interactions that are curated from the primary biomedical literature for all major model organism species [20]. In this study, we applied the Biological General Repository for Interaction Datasets tool to identify the PPIs for DEGs, and then the PPI network was visualized using the Cytoscape software (http://cytoscape.org/) [21].

2.6 Functional enrichment analysis

GO analysis has been widely utilized as a functional enrichment research for large-scale genes [22]. Database for Annotation, Visualization and Integrated Discovery is a soft tool that assists in the interpreta- tion of genome-scale datasets by facilitating the transition from data collection to biological meaning [23]. In our analysis, we used the online tool Database for Annotation, Visualization and Integrated Discovery to conduct GO enrichment analysis. GO terms with P-value <0.01 were considered to be significant. P-value was calculated as follows:

ab+  cd+       a   c  (6) P =  n    ac+ 

Here n is the number of background genes; a’ is on behalf of the number of one gene set in the gene lists; a’ + b denotes the number of genes in the gene list ,including at least one gene set; a’ + c represents the number of one gene list in the background genes; and a’ is replaced with a such that a = a’ - 1.

3 Results

3.1 Identification of DEGs

Based on the FDR ≤0.05, DEGs were identified. The numbers of DEGs identified in the four experiments of comparison between LB and unselected BM (experiment 1), LB and CD34+ isolated from BM (experiment 2), LB and unselected PB (experiment 3), LB and CD34+ isolated from PB (experiment 4) are shown in Figure 1. From this figure, we found that the most DEGs were in experiment 3 and the least DEGs were in experiment 2. The number of DEGs in CD34+ isolated from BMs and CD34+ isolated from PB was smaller than that in corresponding BMs and PB, respectively. The comparison among the number of DEGs obtained in the four experiments showed that LB cells did not tend to differentiation and CD34+ was more similar to cancer Potential genes in AML 133 stem cells. The heatmaps of the top 20 DEGs obtained in the four experiments are shown in Figures 2–5, respectively. As shown in Figure 2, the top 5 ranked DEGs in experiment 1 were GYPA, LCK, ITK, BCL11B, and RHAG. In Figure 3, the top 5 ranked DEGs were TAL1, RHOH, MYH10, TYROBP, and CFD. In Figure 4, the top

Fig. 1: Number of DEGs in different experiments. LB vs BM: the comparison between LB and unselected BM. LB vs CD34+ BM: the comparison between LB and CD34+ isolated from BM. LB vs PB: the comparison between LB and unselected PB. LB vs CD34+ PB: the comparison between LB and CD34+ isolated from PB. DEG, differentially expressed gene; LB, leukemic blast; BM, bone marrow; PB, peripheral blood.

Fig. 2: Heatmap of the top 20 DEGs from the experiment of comparison between LB and unselected BM. DEG, differentially expressed gene; LB, leukemic blast; BM, bone marrow.

Fig. 3: Heatmap of the top 20 DEGs from the experiment of comparison between LB and CD34+ isolated from BM. DEG, differentially expressed gene; LB, leukemic blast; BM, bone marrow. 134 Zhi-Jian Li et al.

Fig. 4: Heatmap of the top 20 DEGs from the experiment of comparison between LB and unselected PB. DEG, differentially expressed gene; LB, leukemic blast; PB, peripheral blood.

Fig. 5: Heatmap of the top 20 DEGs from the experiment of comparison between LB and CD34+ isolated from PB. DEG, differentially expressed gene; LB, leukemic blast; PB, peripheral blood.

5 ranked DEGs were CXCR2, MGAM, TNFRSF10C, KCNJ15, and CYP4F3. In Figure 5, the top 5 ranked were NAP1L3, TSPYL5, MYH10, PFKM, and PBX1.

3.2 Impact analysis of TFs

The top 5 ranked TFs in experiment 1 were STAT4, FOXO3, TAL1, GATA1, and DDIT3. In experiment 2, the top 5 ranked TFs were TCF3, TGIF1, GATA1, CDC5L, and GATA2. In experiment 3, the top 5 ranked TFs were NR3C1, STAT3, GATA3, MYCN, and STAT1. In experiment 4, the top 5 ranked TFs were TP53, HOXA5, HLF, MECOM, and NR3C1.

3.3 PPI network of DEGs

The PPI network of DEGs in experiment 1 is shown in Figure 6. From Figure 6, we can find that DEGs such as CREBBP, FXO1, NCOA2, NCOA1, IFI16, and NR3C1 are critical. The PPI network of DEGs in experiment 2 is shown in Figure 7. From the results, we can conclude that DEGs including CHUK, F2RL1, IKBKB, PYCARD, TMED2, DACT1, COPS5, and IKBKG are important in the PPI network. The PPI network of DEGs in experiment 3 is displayed in Figure 8, which indicates that DEGs such as MCM2, MCM6, MCM7, CREBBP, SFPQ, IKBKG, DPM1, FADD and URI1 might be significant in disease progression. The PPI network for experiment 4 is shown in Figure 9. From the results, we can conclude that SIRT1, CHUK, GPSM2, TP53, IKBKB, MCM2, PYCARD, MCM6, CBX5, IKBKG, and MYC are important. Potential genes in AML 135

Fig. 6: PPI networks of DEGs from the experiment of comparison between LB and unselected BM. PPI, protein–protein interaction; DEG, differentially expressed gene; LB, leukemic blast; BM, bone marrow.

Fig. 7: PPI networks of DEGs from the experiment of comparison between LB and CD34+ isolated from BM. PPI, protein–protein interaction; DEG, differentially expressed gene; LB, leukemic blast; BM, bone marrow. 136 Zhi-Jian Li et al.

Fig. 8: PPI networks of DEGs from the experiment of comparison between LB and unselected PB. PPI, protein–protein interaction; DEG, differentially expressed gene; LB, leukemic blast; BM, bone marrow.

Fig. 9: PPI networks of DEGs from the experiment of comparison between LB and CD34+ isolated from PB. PPI, protein–protein interaction; DEG, differentially expressed gene; LB, leukemic blast; BM, bone marrow. Potential genes in AML 137

3.4 GO functional enrichment analysis

Considering P<0.05, GO enrichment analysis in experiment 1 indicated that DEGs were significantly enriched in immune response (P = 1.59E-21), leukocyte activation (P = 4.67E-18), and cell activation (P = 8.94E-18). In experiment 2, DEGs mainly participated in defense response (P = 7.44E-10), leukocyte migration (P = 5.86E-09), and immune response (P = 5.88E-09). In experiment 3, the results showed that the biological functions enriched by DEGs were related to immune response (P = 2.98E-21), leukocyte activation (P = 1.34E-14), and cell activation (P = 6.33E-13). In experiment 4, we concluded that the biological terms were associated with chromosome organization (P = 1.10E-06), immune response (P = 2.27E-06), and chromatin organization (P = 2.45E-06). From this analysis, we found that the function of immune response was common to all these four stages.

4 Discussion

Hematopoietic stem cells (HSCs) observed in CD34+ are of an undifferentiated, primitive form. Seminal studies about AML initially suggested that leukemia stem cells (LSCs) might be closely related to normal HSCs. To solve this problem, microarray studies compared the expression profiles between AML blasts and normal hematopoietic cells. Four different types of normal hematopoietic cells were analyzed, ranging from immature CD34-expressing cells to unselected PB (unselected BMs, CD34+ isolated from BMs, unselected PB, and CD34+ isolated from PB). By comparing the DEGs between the four experiments, we observed that the highest number of DEGs was in experiment 3, which indicated that genes in LB did not tend to differentiation. The number of DEGs in CD34+ isolated from BMs and in CD34+ isolated from PB was less than that in corresponding BMs and PB, respectively, which indicated that CD34+ was more similar to LSCs. TAL1 was one of the top 5 DEGs in experiment 2 (CD34+ isolated from BMs), which belonged to the basic helix–loop–helix (HLH) family of TFs that bind as heterodimers with theE2A and HEB/HTF4 gene products to a nucleotide sequence motif termed the E-box [24]. It plays a key role in controlling the development of primitive and definitive hematopoiesis during mouse development [25]. It is also a critical regulator of hematopoietic and vascular development and is misexpressed in the majority of patients with T-cell acute lymphoblastic leukemia (ALL) [26]. Tumor- specific activation of the TAL1 gene occurs in ~25% of patients with T-cell ALL [27]. Thus, TAL1 might be closely connected with the development and progression of AML. RIF metric was used to identify critical TFs from gene-expression data. The top 5 ranked TFs in experiment 4 were TP53, HOXA5, HLF, MECOM, and NR3C1. Moreover, TP53 was one of the core genes in the PPI network. Inactivation of the ATM or TP53 gene is frequent in B-cell lymphocytic leukemia and leads to aggressive disease [28]. Researches showed that methylation of CpG and CCWGG motifs in the promoter of TP53 could represent a novel mechanism leading to functional impairment of this tumor suppressor gene in ALL [29]. Although other genes in the PPI network of experiment 4 have not been reported to be related to AML, we speculated that they were important in AML. CREBBP was seen as the core gene in the PPI network of experiment 1. During the last few years, numerous transcriptional regulators have been found to interact with CREB1-binding protein (CBP) and p300 [30]. CBP and p300 are widely expressed and are believed to regulate gene expression in most cell types [31]. They are believed to participate in the activities of hundreds of different TFs [32]. CREB can activate transcription of target genes by recruiting the coactivator CBP[33]. Vizmanos et al. reported that MORF–CREBBP fusion by t(10;16)(q22;p13) is a recurrent event in AML. They performed their experiment on an 84-year-old patient diagnosed with AML M5b, in which the t(10;16)(q22;p13) was the only cytogenetic aberration. Ultimately, this supports that t(10;16)(q22;p13) is a recurrent pathogenic translocation in AML [34]. In human colorectal cancer cell line, CBP/p300 has an evolutionarily conserved role as a buffer regulating TCF-beta-cat/Arm binding [35]. Double and triple immunofluorescence analyses revealed that CBP was present in a subset of promyelocytic leukemia (PML) bodies [36]. From these reports, we concluded that CREBBP might be related to AML. 138 Zhi-Jian Li et al.

In experiment 2, there were many genes involved in the PPI network, and F2RL1 was the core gene. F2RL1 gene (proteinase-activated receptor 2, also designated as PAR-2 or PAR2) is located in 5q13, a genomic area that is linked to the development of allergy [37], and is expressed in various tissues, including cancer lesions [38]. It has been suggested as a mediator of inflammatory, mitogenic, and fibrogenic responses to injury [39]. It is activated by trypsin-like serine proteases and can promote cell migration through an ERK1/2-dependent pathway. Ge et al. [40] reported that autocrine activation of PAR-2 by secreted proteases may contribute to the migration of metastatic tumor cells through beta-arrestin-dependent ERK1/2 activation. It contributes to tumor cell motility and metastasis [41]. Since there are many studies that report the relationship between F2RL1 gene and cancer, we speculated that F2RL1 gene might play an important role in AML progression. MCM2 was the most critical gene in the PPI network of DEGs in experiment 3. MCM2 was included in the MCM complex composed of six subunits (MCM2, MCM3, MCM4, MCM5, MCM6, and MCM7), which are all structurally related and highly conserved in eukaryotes [42]. The MCM proteins 2–7 are required for both the initiation and elongation steps of chromosomal DNA replication [43]. The Mcm complex contains both a single bipartite nuclear localization signal (NLS) and a nuclear export signal (NES) [44]. MCM2 is a neural stem cell marker; research performed by Lee et al. revealed the significance of MCM2 expression in invasive tumor [42]. The expression of MCM2 in colonocytes retrieved from the fecal surface can be used to identify colorectal cancer [45]. MCM2 can be taken as a sensitive marker for premalignant lung cells, due to the fact that it is present in cells at the surface of metaplastic lung lesions, which are more likely to be exfoliated into sputum [46]. Moreover, MCM2 is regulated via a P53-independent pathway and is a useful biomarker of proliferating cells [47]. Therefore, we speculated that MCM2 might be related to AML.

5 Conclusion

In this study, to understand the relationship between LSCs and normal HSCs, comparisons between LB and unselected BMs, LB and CD34+ isolated from BMs, LB and unselected PB, and LB and CD34+ isolated from PB were performed. According to our results, we believe that our work offers novel insights into the expression differences between AML blasts and normal hematopoietic cells. We concluded that LSCs are closely related to normal CD34+ HSCs. Genes such as TAL1, TP53, CREBBP, and F2RL1 are related to the biology of AML. However, our results were obtained only using bioinformatics methods but were not verified by means of animal experiments or patient tissues. Thus, further study is needed to validate our obtained results.

Acknowledgments: This research received no specific grants from any funding agency in public, commercial, or not-for-profit sectors.

Competing interests: The authors declare that they have no competing interests.

References

[1] McGuire S., World Cancer Report 2014. Geneva, Switzerland: World Health Organization, International Agency for Research on Cancer, WHO Press, 2015, Adv. Nutr., 2016, 7, 418-419. [2] Estey E., Döhner H., Acute myeloid leukaemia, Lancet, 2006, 368, 1894-1907. [3] O’Donnell M.R., Abboud C.N., Altman J., Appelbaum F.R., Coutre S.E., Damon L.E., et al., Acute myeloid leukemia, J. Nat. Compr. Cancer Network, 2011, 9, 280-317. [4] Distler O., Gay S., Neumann E., Müller-Ladner U., Minimum information about a microarray experiment: comment on the editorial by Firestein and Pisetsky, Arthritis Rheum., 2003, 48, 861; author reply 862. [5] Craig C.M., Schiller G.J., Acute myeloid leukemia in the elderly: conventional and novel treatment approaches, Blood Rev., 2008, 22, 221-234. [6] Haferlach T., Molecular genetic pathways as therapeutic targets in acute myeloid leukemia, Hematology, 2008, 2008, 400. Potential genes in AML 139

[7] Dalman M.R., Deeter A., Nimishakavi G., Duan Z.H., Fold change and p-value cutoffs significantly alter microarray interpretations, BMC Bioinf., 2012, 13(Suppl. 2), S11. [8] Valk P.J., Verhaak R.G., Beijen M.A., Erpelinck C.A., van Doorn-Khosrovani SBvW, Boer J.M., et al., Prognostically useful gene-expression profiles in acute myeloid leukemia, N. Engl. J. Med., 2004, 350, 1617-1628. [9] Mrózek K., Marcucci G., Paschka P., Whitman S.P., Bloomfield C.D., Clinical relevance of mutations and gene-expression changes in adult acute myeloid leukemia with normal cytogenetics: are we ready for a prognostically prioritized molecular classification? Blood, 2007, 109, 431-448. [10] Smalley J.L., Gant T.W., Zhang S.D., Application of connectivity mapping in predictive toxicology based on gene-expression similarity, Toxicology, 2010, 268, 143-146. [11] Lamb J., Crawford E.D., Peck D., Modell J.W., Blat I.C., Wrobel M.J., et al., The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease, Science, 2006, 313, 1929-1935. [12] Gullans S.R., Connecting the dots using gene-expression profiles, N. Engl. J. Med., 2006, 355, 2042. [13] Stirewalt D.L., Meshinchi S., Kopecky K.J., Fan W., Pogosova-Agadjanyan E.L., Engel J.H., et al., Identification of genes with abnormal expression changes in acute myeloid leukemia, Genes Chromosomes Cancer, 2008, 47, 8-20. [14] Smyth G.K., Limma: linear models for microarray data, In: R. Gentleman, V.J. Carey, W. Huber, R.A. Irizarry, S. Dudoit (eds.), Bioinformatics and computational biology solutions using R and Bioconductor, Springer, Springer, New York, NY, 2005. 397-420. [15] Benjamini Y., Hochberg Y., Controlling the false discovery rate: a practical and powerful approach to multiple testing, J. R. Stat. Soc. S B (Methodol.)., 1995, 57, 289-300. [16] Allocco D.J., Kohane I.S., Butte A.J., Quantifying the relationship between co-expression, co-regulation and gene function, BMC Bioinf., 2004, 5, 18. [17] Reverter A., Hudson N.J., Nagaraj S.H., Pérez-Enciso M., Dalrymple B.P., Regulatory impact factors: unraveling the transcriptional regulation of complex traits from expression data, Bioinformatics., 2010, 26, 896-904. [18] Stelzl U., Worm U., Lalowski M., Haenig C., Brembeck F.H., Goehler H., et al., A human protein-protein interaction network: a resource for annotating the proteome, Cell., 2005, 122, 957-968. [19] Ge H., Walhout A.J., Vidal M., Integrating ‘omic’ information: a bridge between genomics and systems biology, TRENDS Genet., 2003, 19, 551-560. [20] Chatr-aryamontri A., Breitkreutz B.-J., Heinicke S., Boucher L., Winter A., Stark C., et al., The BioGRID interaction database: 2013 update, Nucleic Acids Res., 2013, 41, D816-D823. [21] Smoot M.E., Ono K., Ruscheinski J., Wang P.L., Ideker T., Cytoscape 2.8: new features for data integration and network visualization, Bioinformatics, 2011, 27, 431-432. [22] Ashburner M., Ball C.A., Blake J.A., Botstein D., Butler H., Cherry J.M., et al., Gene ontology: tool for the unification of biology. The Gene Ontology Consortium, Nat. Genet., 2000, 25, 25-29. [23] Huang da W., Sherman B.T., Lempicki R.A., Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources, Nat. Protoc., 2009, 4, 44-57. [24] Huang S., Brandt S.J., mSin3A regulates murine erythroleukemia cell differentiation through association with the TAL1 (or SCL) transcription factor, Mol. Cell. Biol., 2000, 20, 2248-2259. [25] Reynaud D., Ravet E., Titeux M., Mazurier F., Rénia L., Dubart-Kupperschmitt A., et al., SCL/TAL1 expression level regulates human hematopoietic stem cell self-renewal and engraftment, Blood, 2005, 106, 2318-2328. [26] Huang S., Qiu Y., Shi Y., Xu Z., Brandt S.J., P/CAF‐mediated acetylation regulates the function of the basic helix–loop–helix transcription factor TAL1/SCL, EMBO J., 2000, 19, 6792-6803. [27] Hsu H.-L., Wadman I., Baer R., Formation of in vivo complexes between the TAL1 and E2A polypeptides of leukemic T cells, Proc. Natl. Acad. Sci., 1994, 91, 3181-3185. [28] Stankovic T., Hubank M., Cronin D., Stewart G.S., Fletcher D., Bignell C.R., et al., Microarray analysis reveals that TP53-and ATM-mutant B-CLLs share a defect in activating proapoptotic responses after DNA damage but are distinguished by major differences in activating prosurvival responses, Blood, 2004, 103, 291-300. [29] Agirre X., Vizmanos J.L., Calasanz M.J., García-Delgado M., Larráyoz M.J., Novo F.J., Methylation of CpG dinucleotides and/ or CCWGG motifs at the promoter of TP53 correlates with decreased gene expression in a subset of acute lymphoblastic leukemia patients, Oncogene, 2003, 22, 1070-1072. [30] Chan H.M., La Thangue N.B., p300/CBP proteins: HATs for transcriptional bridges and scaffolds, J. Cell Sci., 2001, 114, 2363-2373. [31] Blobel G.A., CREB-binding protein and p300: molecular integrators of hematopoietic transcription, Blood, 2000, 95, 745-755. [32] Vo N., Goodman R.H., CREB-binding protein and p300 in transcriptional regulation, J. Biol. Chem, 2001, 276, 13505-13508. [33] Goodman R.H., Smolik S., CBP/p300 in cell growth, transformation, and development, Genes Dev., 2000, 14, 1553-1577. [34] Vizmanos J.L., Larráyoz M.J., Lahortiga I., Floristán F., Alvarez C., Odero M.D., et al., t(10; 16)(q22; p13) and MORF-CREBBP fusion is a recurrent event in acute myeloid leukemia, Genes Chromosomes Cancer., 2003, 36, 402-405. [35] Li J., Sutter C., Parker D.S., Blauwkamp T., Fang M., Cadigan K.M., CBP/p300 are bimodal regulators of Wnt signaling, EMBO J., 2007, 26, 2284-2294. 140 Zhi-Jian Li et al.

[36] von Mikecz A., Zhang S., Montminy M., Tan E.M., Hemmerich P., Creb-binding protein (Cbp/P300) and RNA polymerase II colocalize in transcriptionally active domains in the nucleus, J. Cell Biol., 2000, 150, 265-274. [37] Bouzigon E., Ulgen A., Dizier M.-H., Siroux V., Lathrop M., Kauffmann F., et al., Evidence for a pleiotropic QTL on chromosome 5q13 influencing both time to asthma onset and asthma score in French EGEA families, Hum. Genet., 2007, 121, 711-719. [38] Ikeda O., Egami H., Ishiko T., Ishikawa S., Kamohara H., Hidaka H., et al., Signal of proteinase-activated receptor-2 contributes to highly malignant potential of human pancreatic cancer by up-regulation of interleukin-8 release, Int. J. Oncol., 2006, 28, 939-946. [39] Wang J., Zheng H., Hollenberg M.D., Wijesuriya S.J., Ou X., Hauer-Jensen M., Up-regulation and activation of proteinase-activated receptor 2 in early and delayed radiation injury in the rat intestine: influence of biological activators of proteinase-activated receptor 2, Radiation Research., 2003,160 (5) :524. [40] Ge L., Shenoy S.K., Lefkowitz R.J., DeFea K., Constitutive protease-activated receptor-2-mediated migration of MDA MB-231 breast cancer cells requires both β-arrestin-1 and-2, J. Biol. Chem., 2004, 279, 55419-55424. [41] Shi X., Gangadharan B., Brass L.F., Ruf W., Mueller B.M., Protease-activated receptors (PAR1 and PAR2) contribute to tumor cell motility and metastasis11NIH grants CA85405 (BM Mueller), HL16411 (W. Ruf), and HL60742 (W. Ruf), Mol. Cancer Res., 2004, 2, 395-402. [42] Lee J.-K., Seo Y.-S., Hurwitz J., The Cdc23 (Mcm10) protein is required for the phosphorylation of minichromosome maintenance complex by the Dfp1-Hsk1 kinase, Proc. Natl. Acad. Sci., 2003, 100, 2334-2339. [43] Lee J.-K., Hurwitz J., Processive DNA helicase activity of the minichromosome maintenance proteins 4, 6, and 7 complex requires forked DNA structures, Proc. Natl. Acad. Sci., 2001, 98, 54-59. [44] Braun K.A., Breeden L.L., Nascent transcription of MCM2-7 is important for nuclear localization of the minichromosome maintenance complex in G1, Mol. Biol. Cell., 2007, 18, 1447-1456. [45] Davies R.J., Freeman A., Morris L.S., Bingham S., Dilworth S., Scott I., et al., Analysis of minichromosome maintenance proteins as a novel method for detection of colorectal cancer in stool, Lancet, 2002, 359, 1917-1919. [46] Tan D.-F., Huberman J.A., Hyland A., Loewen G.M., Brooks J.S., Beck A.F., et al., MCM2-a promising marker for premalignant lesions of the lung: a cohort study, BMC Cancer., 2001, 1, 6. [47] Kodani I., Shomori K., Osaki M., Kuratate I., Ryoke K., Ito H., Expression of minichromosome maintenance 2 (MCM2), Ki-67, and cell-cycle-related molecules, and apoptosis in the normal-dysplasia-carcinoma sequence of the oral mucosa, Pathobiology., 2002, 69, 150-158.