- Email: [email protected]
Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes
J. Comp. Path. 2014, Vol. -, 1e12
Available online at www.sciencedirect.com
ScienceDirect www.elsevier.com/locate/jcpa
NEOPLASTIC DISEASE
Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes M. Zamani-Ahmadmahmudi*,†, A. Najafi‡ and S. M. Nassirix * Department of Clinical Science, Faculty of Veterinary Medicine, Shahrekord University, Shahrekord, † Department of Clinical Science, Faculty of Veterinary Medicine, Shahid Bahonar University of Kerman, Kerman, ‡ Molecular Biology Research Center, Baqiyatallah University of Medical Sciences and x Department of Clinical Pathology, Faculty of Veterinary Medicine, University of Tehran, Tehran, Iran
Summary Lymphoma is one of the most common malignancies in dogs. Canine lymphoma is similar to human nonHodgkin’s lymphoma (NHL) with shared clinical presentation and histopathological features. This study reports the construction of a comprehensive gene regulatory network (GRN) for canine diffuse large B-cell lymphoma (DLBCL), the most common type of canine lymphoma, and performs analysis for detection of major functional modules and hub genes (the most important genes in a GRN). The canine DLBCL GRN was reconstructed from gene expression data (NCBI GEO dataset: GSE30881) using the STRING and MiMI interaction databases. Reconstructed GRNs were then assessed, using various bioinformatics programmes, in order to analyze network topology and identify major pathways and hub genes. The resultant network from both interaction databases had a logically scale-free pattern. Gene ontology (GO) analysis revealed cell activation, cell cycle phase, immune effector process, immune system development, immune system process, integrinmediated signalling pathway, intracellular protein kinase cascade, intracellular signal transduction, leucocyte activation and differentiation, lymphocyte activation and differentiation as major GO terms in the biological processes of the networks. Moreover, bioinformatics analysis showed E2F1, E2F4, PTEN, CDKN1A, PCNA, DKC1, MNAT1, NDUFB4, ATP5J, PRKDC, BRCA1, MYCN, RFC4 and POLA1 as the most important hub genes. The phosphatidyl inositol signalling system, P53 signalling pathway, Rac CycD pathway, G1/S checkpoint, chemokine signalling pathway and telomere maintenance were the main signalling pathways in which the protein products of the hub genes are involved. Ó 2014 Elsevier Ltd. All rights reserved. Keywords: dog; gene regulatory network; hub genes; lymphoma
Introduction Canine lymphoma is one of the most common malignancies in dogs and most commonly has multicentric anatomical distribution (MacEwen, 1990). Canine lymphoma shares clinical and histopathological features with human non-Hodgkin’s lymphoma (NHL) (Vail and MacEwen, 2000; McCaw et al., 2007). Histopathological classification of canine lymphoma
Correspondence to: M. Zamani-Ahmadmahmudi (e-mail: zamani_2012@ alumni.ut.ac.ir). 0021-9975/$ - see front matter http://dx.doi.org/10.1016/j.jcpa.2014.11.008
has been performed using three classification schemes the updated Kiel scheme, the Revised European and American Lymphoma (REAL) scheme and the World Health Organisation (WHO) scheme. These classifications consider epidemiological, clinical, clinicomorphological, genetic and phenotypic parameters. The most common form of canine lymphoma is diffuse large B-cell lymphoma (DLBCL) (Ponce et al., 2010). Cancer develops through dysregulation of normal cellular processes including apoptosis, cell mitosis, cell cycling, cell differentiation and DNA repair. Ó 2014 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
2
Neoplastic transformation results in activation of oncogenes, deactivation of tumour suppressor genes and/or genome instability (Croce, 2008; EmmertStreib et al., 2014) and various cell processes are involved in cancer (Hanahan and Weinberg, 2000). Cancer involves intricate interactions between different signalling components such as genes, proteins and metabolites, and as such interactions do not follow a simple chain-like pattern, it has proven challenging to understand the molecular interactions underlying oncogenesis (Emmert-Streib et al., 2014). One of the most useful approaches in providing a global overview of cancer pathways has been reconstruction of ‘gene regulatory networks’ (GRNs) using various algorithms or interaction databases (Blais and Dynlacht, 2005; Bansal et al., 2007; Emmert-Streib et al., 2014). GRNs can yield valuable information about major gene ontology modules and are able to identify major genes (‘hub genes’) involved in development of a specific tumour (Basso et al., 2005, 2010; Emmert-Streib et al., 2014). Through reconstruction and analysis of a breast cancer GRN, Emmert-Streib et al. (2014) characterized the major cellular processes contributing to this form of neoplasia, including cell cycling, cell adhesion, translation, organelle fission, the immune response and mitosis. Using a reconstruction of a transcriptional network, Agnelli et al. (2011) identified the most critical genes associated with poor prognosis in patients with multiple myeloma. These methods have not yet been applied widely to the investigation of canine cancer; however, microarray data exist for canine mammary gland tumours, canine lymphoma and canine haemangiosarcoma (http://www.ncbi.nlm.nih.gov/gds/?term¼canine% 20lymphoma). The aim of the current study was to reconstruct a canine DLBCL GRN and evaluate it using different analytical methods. Specifically, the study defined major functional modules (GO), hub genes (highly connected nodes or the most important genes in a GRN) and the functional biological processes associated with the identified hub genes.
Materials and Methods Data Collection and Primary Processing
Gene expression data were obtained from the NCBI Gene Expression Omnibus GEO dataset (GSE30881, platform: GPL3738) (Mudaliar et al., 2013). Data included the gene expression profile of 23 canine DLBCLs and normal lymph nodes from 10 healthy dogs. Data were downloaded in the CEL file format and converted to expression values by the Affy package (Gautier et al., 2004) in R program,
version 3.0.2 (http://www.r-project.org/). As part of this process, data were transformed logarithmically. Then, the data (probe sets) were entered into the geWorkbench 2.5.1 package (Floratos et al., 2010) and non-useful data were filtered based on two criteria: data without Enterz identification were omitted and markers with multiple probset identifications were filtered, while probsets with the highest mean expression value were retained. Upregulation or downregulation of the genes (P <0.01) was determined by use of a t test provided in the geWorkbench package (correction method: just alpha, group variance: equal). k-means Clustering and Principle Component Analysis
Genes were clustered through two algorithms including k-means clustering and principle component analysis (PCA). Using k-means clustering, data were partitioned into specified numbers of groupings (k) in which each observation belonged to the cluster with the nearest mean. PCA reduced dimensionality through generating few linear combinations of all data that are named ‘principle components’. Principle components can extract clusters from the original data. Both analyses were performed in MATLAB 7.8.0 (R2009a) (MathWorks, Natick, Massachusetts, USA). Network Reconstruction and Analysis
The GRN was constructed using two major interaction databases, including MiMI (Michigan Molecular Interactions) (Tarcea et al., 2009) and Search Tool for the Retrieval of Interacting Genes/Proteins (STRING; Franceschini et al., 2013). Briefly, the list of genes was submitted online or through a related plug-in within Cytoscape software (Cline et al., 2007) and then interactions between genes were retrieved and displayed as an interactions network. MiMI collects real interaction data from interaction databases such as the Swiss Protein Database (SwissProt), the Human Protein Reference Database (HPRD), the Biomolecular Interaction Network Database (BIND), the Database of Interacting Proteins (DIP), Reference Sequence (RefSeq) and the International Protein Index (IPI). Interaction data from STRING originate from four major sources including high-throughput experiments, genomic context, co-expression and previous knowledge. Constructed networks were imported to Cytoscape and analyzed using various plug-ins for different aspects of network features. The networks obtained from these databases were analyzed by NetworkAnalyzer plug-in (Assenov et al., 2008) in order to determine topological parameters and centrality measures, including the network
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
3
Gene Analysis in Canine Lymphoma
diameter, radius, density, centralization, heterogeneity, clustering coefficient, connected components, the number of nodes/edges, the number of self-loops, the shortest path length and the characteristic path length. This plug-in also computed node degree distribution, neighbourhood connectivity distribution, shared neighbour distribution, betweenness centrality, closeness centrality and stress centrality distribution. Detection of Functional Modules Using Gene Ontology Analysis
Overrepresented GO terms (especially biological pathways, BPs) were detected in the whole network using the BINGO plug-in (version 2.4) (Maere et al., 2005) in the Cytoscape software. This plug-in identified statistically significant functional modules in a given gene set and displayed them as a Cytoscape network. The Biological Networks gene ontology Tool (BINGO) used hypergeometric testing for statistical analysis and the Benjamini and Hochberg false discovery rate (FDR) correction for multiple testing correction. P <0.05 was considered significant. Detection of Hub Genes
Hub genes (genes with a high number of edges) were explored using the CytoHubba software (release 16) (http://hub.iis.sinica.edu.tw/cytoHubba/index.html). Hubba nodes (genes) were detected by several ranking methods as well as closeness, betweenness, radiality,
stress, degree, edge percolated component (EPC), density of maximum neighbourhood component (DMNC), maximal clique centrality (MCC), maximum neighbourhood component (MNC), bottleneck (BN) and eccentricity. The top 10 genes with the highest score were selected for each parameter. To demonstrate the role of hub genes in biological processes and cancer pathways, functional annotation clustering was also conducted using the Database for Annotation, Visualization and Integrated Discovery (DAVID; Huang et al., 2009a, 2009b). Ontology parameters were adjusted as follows: classification stringency of medium, similarity term overlap of 3.0, similarity threshold of 0.05, multiple linkage threshold of 0.05 and an enrichment threshold of 1.0.
Results Global Properties of Genes with Significant Expression Change
Data filtration using the geWorkbench package extracted 16,970 genes from a total of 43,035 genes. t test analysis revealed 2,087 genes with significantly different expression between neoplastic and healthy samples (Supplementary Table 1). k-means clustering and PCA assays were performed for evaluation of global expression profiles of upregulated or downregulated genes. k-means clustering divided the genes into 16 major clusters with genes in each cluster possessing a similar expression pattern in studied individuals (Fig. 1, Supplementary
Fig. 1. k-means clustering of genes with significantly different expression between samples from dogs with neoplasia and healthy dogs. Analysis indicates 16 different clusters. Expression of each gene is plotted in 33 animals. The first 23 animals have neoplasia and the last 10 animals are healthy. Vertical axes indicate gene expression values.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
4
M. Zamani-Ahmadmahmudi et al.
Fig. 2. PCA for the gene expression profile of canine lymphoma reveals two major clusters that clearly indicate neoplastic and healthy samples. Each cluster is defined by a red ellipse.
Table 2). In addition, the PCA analysis grouped samples into two major clusters that clearly separated neoplastic (n ¼ 23) and healthy samples (n ¼ 10) (Fig. 2). PCA results were confirmed by generating a clustergram in MATLAB (Fig. 3). Gene Regulatory Network Reconstruction
A GRN was reconstructed using the MiMI and STRING interaction databases for genes with significant expression change. The MiMI database delivered a network with 710 nodes (genes) and 1,835
edges (interactions), while STRING yielded a network with 1,022 nodes and 5,803 edges (Supplementary Cytoscape File 1). General topological analysis for both networks was performed by use of NetworkAnalyzer (Table 1). The number of connected components was 25 in the MiMI network and 13 in the STRING network, indicating that a number of subnetworks comprised a network. A lower number of connected components implies strong connectivity; however, the networks identified in the present study also had strong connectivity because 658 nodes (1,806 edges) in the MiMI network and 987
Fig. 3. Clustergram of genes with significant expression changes in the canine dataset. On the left side of the image is a scale bar indicating expression fold (from minus threefold [downregulation] to threefold [overexpression]). The clustergram separates the samples from dogs with neoplasia (1e23) and healthy dogs (24e33).
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
Gene Analysis in Canine Lymphoma Table 1 Simple topological parameters of GRNs from MiMI and STRING databases Parameter
MiMI
STRING
Number of nodes 710 1022 Network diameter* 13 10 Network radius† 1 1 Network centralization‡ 0.107257 0.086990839 Network densityx 0.007291 0.01112259 Network heterogeneityǁ 1.427738 1.219856226 Number of connected components{ 25 16 Number of shortest paths# 432376 (85%) 973236 (93%) Characteristics path length** 4.832391 3.745058752 Average number of neighbours 5.169014 11.35616438 *
The highest distance between two nodes, where the distance is minimum number of edges that connected two nodes. † The smallest distance between two nodes. ‡ How does the network topology resemble a star structure? (A value between 0 and 1). x How densely is the network populated with edges? (A value between 0 and 1). ǁ This parameter indicates the affinity of a network to contain hub markers. { Number of subnetworks that constitute a network. # Shortest path: minimum numbers of edges form a path. Shortest paths defined as distance. ** Averaged shortest path length.
nodes (5,781 edges) in the STRING network participated in one component, while 52 nodes (29 edges) in the MiMI network and 35 nodes (22 edges) in the STRING network constituted 24 and 15 connected components, respectively. In other words, pairwise connected nodes falsely increased the number of connected components (Supplementary Cytoscape File 1). Network centralization, density, and average number of neighbours are parameters that describe neighbourhood status in networks (http://med. bioinf.mpi-inf.mpg.de/netanalyzer/index.php). Expectedly, both networks showed a biological scale-free pattern, as many nodes had low numbers of interactions and a few nodes were highly connected (Figs. 4A, B; Supplementary Cytoscape File 1). Neighbourhood connectivity distribution followed an assortive pattern, which means that nodes had a tendency to connect with nodes with similar connectivity. An assortive model implicates a hierarchical clustering for studied probsets (Figs. 4C, D). Gene Ontology Analysis
GO analysis was conducted for identifying important BPs involved in the pathogenesis of canine DLBCL. The full list of statistically significant GO terms (BPs) that was common between both networks is provided in Supplementary Table 3. Respectively, 545 and 334 significant GO terms were detected in
5
the MiMI and STRING networks and 300 of these GO terms were common to both networks. Some of the important GO terms included cell activation, cell cycle phase, immune effector process, immune system development, immune system process, integrin-mediated signalling pathway, intracellular protein kinase cascade, intracellular signal transduction, leucocyte activation and differentiation, lymphocyte activation and differentiation, positive regulation of ab T-cell activation and positive regulation of ab T-cell differentiation. Hub Genes
Various algorithms found hub genes in both GRNs (Tables 2 and 3). Some of the hub genes in the MiMI-derived GRN included E2F1, E2F4, MNAT1, NDUFB4, ATP5J, PRKDC, BRCA1, CDKN1A and PCNA. Some of the important hub genes in the STRING-derived GRN were PCNA, CDKN1A, MYCN, POLA1, PTEN, RFC4, EGFR, DKC1 and POLA1. Additionally, functional analysis revealed the role of the hub genes in biological processes, especially cancer pathways. Ontology analysis was performed using the DAVID database (Table 4), which indicated that most of the biological processes were associated with the cell cycle, mitosis, proliferation, differentiation, death, apoptosis, DNA damage and DNA repair. Furthermore, a number of genes involving in cancer biology (e.g. E2F1, CDKN1A, RAF1, ABL1, PCNA, PRKDC, CHEK1, SMC1A, CYCS, TSC2, ITGA7 and CREB3L4) were identified in KEGG, REACTOME and other pathway databases (Table 5). Functional analysis of the hub genes revealed that these genes were mainly involved in the P53 signalling pathway, Rac CycD pathway, G1/S checkpoint, chemokine signalling pathway, phosphatidylinositol signalling system and telomere maintenance. As cancer pathways in the KEGG database are limited to some specific types of cancers, including colorectal, pancreatic, thyroid, bladder, prostate, endometrial, non-small cell lung cancer, renal cell carcinoma, melanoma, basal cell carcinoma, glioma, acute myeloid leukemia and chronic myeloid leukemia, pathways were only retrieved for these tumours and not for other cancers such as lymphoma or breast cancer (Table 5). PIK3CG, CDKN1A, KITLG, ABL1 and PTEN were found in different cancer pathways. PIK3CG, CDKN1A and PTEN play a critical role in melanoma, glioma and endometrial and prostate cancer. Ontology findings revealed E2F1, EGFR and RAF1 as significant components in pancreatic cancer and non-small cell lung cancer. PIK3CG, CDKN1A and ABL1 were also found as hub markers in chronic myeloid leukaemia (CML).
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
6
Fig. 4. Node degree distribution in (A) the MiMI network and (B) the STRING network. Neighbourhood connectivity distribution in (C) the MiMI network and (D) the STRING network. Both networks have a scale-free pattern because many nodes have a low number of neighbours (A and B). In addition, an assortive pattern is evident for both networks (C and D).
Discussion The aim of the present study was to clarify key biological processes, pathways and hub genes in canine DLBCL by analyzing major GO terms (especially biological processes) in the GRNs obtained from the MiMI and STRING databases. The network was then investigated in more depth in order to identify functional modules (subnetworks) and their related genes. Both GRNs were ‘logic networks’, as they had a biological scale-free pattern and assortive model of neighbourhood connectivity distribution. Several pathways were highlighted from these analyses. One was the PI3K/AKT pathway, one of the critical and common pathways in human lymphoma biology (Jardin et al., 2010; Qiao et al., 2013; Xu et al., 2013; Zhang et al., 2013). Four hub genes (i.e. PTEN, PIK3CG, PLCB4 and INPP4B) identified in the study are involved in the PI3K/AKT pathway. Mutation, amplification and translocation of various components of this pathway occur in different cancers. This pathway was reported to possess prognostic power in patients with DLBCL, as patients with PI3K/AKT/mTOR activation were characterized as having more severe symptoms, a
poor response to cancer therapy and poor survival. Activity of the PI3K/AKT/mTOR pathway can be inhibited by rituximab combined with rapamycin (Xu et al., 2013), and inhibition of this pathway and its downstream regulator NF-kB was shown to promote apoptosis in three human Burkitt’s lymphoma cell lines (Qiao et al., 2013). PTEN, a key gene in this pathway, serves as a tumour suppressor, and its dysregulation is recognized in some cancers, including endometrial carcinoma, prostatic carcinoma, melanoma, glioma and NHL. Somatic mutation of PTEN was reported in NHL and Burkitt’s lymphoma cell lines (Sakai et al., 1998; Butler et al., 1999). PTEN induces its tumour suppressor function through accumulation of the phosphatidylinositol-3, 4, 5triphosphate (PIP3) in the PI3K/AKT signalling pathway (Engelman et al., 2006). Mutation of PTEN also triggers the onset of T-cell lymphoma in mice, in which lymphomagenesis and overexpression of C-MYC occur when there was deficiency of PTEN protein (Liu et al., 2010). Zhang et al. (2013) suggested that PIK3CD was an oncogene that might be a useful therapeutic target in DLBCL, as a single point mutation in the catalytic domain of PIK3CD
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
7
Gene Analysis in Canine Lymphoma Table 2 Hub genes detected using 12 different algorithms in the network obtained from the MiMI database MCC
DMNC
Gene NDUFB4 ATP5J RPLP0 EIF3A NDUFB5 RPS4X RPL8 EIF2S2 ATP5G1 NDUFS8
MNC
EPC
Radiality
Stress
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13
EIF5B NDUFB5 NDUFS8 NDUFA10 NDUFB7 NDUFA11 NDUFA1 ND2 NDUFA13 NDUFA2
1.124 1.119 1.119 1.119 1.119 1.119 1.119 1.119 1.119 1.119
E2F1 E2F4 RPS5 RPS10 RPS17 NDUFB4 RPS4X RPS3 CYCS RPS26
35 33 26 25 25 24 24 24 24 24
E2F1 E2F4 MNAT1 SMC1A POLR2E POLR2B RPS5 POLR2H RPS26 POLR2F
111.97 109.14 107.83 106.31 105.46 105.44 104.88 104.80 104.77 104.76
E2F1 E2F4 PRKDC BRCA1 MNAT1 PCNA PRKCD ABL1 CHEK1 SNRPA
10.170 10.101 9.980 9.918 9.900 9.829 9.805 9.778 9.751 9.740
E2F1 E2F4 EGFR TIAM1 PRKDC MNAT1 ATP5C1 ABL1 PRKCD COX5B
483784 418270 250102 181610 146578 111758 110216 110024 109822 99,496
Closeness Gene E2F4 E2F1 MNAT1 PRKDC BRCA1 TSC2 PCNA EGFR SNRPA CDKN1A
Betweenness
Clustering
Eccentricity
BN
Degree
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
270.62 269.91 235.42 233.36 232.69 226.03 225.90 223.71 221.84 219.58
E2F1 E2F4 EGFR TIAM1 PRKDC GIPC1 LYN BRCA1 ABL1 PRKCD
51,034.37 49,588.15 29,403.27 14,397.02 14,110.83 13,369.13 11,789.21 11,529.26 10,660.97 10,277.8
ALG5 GTF3A WDR77 CREB3L4 BRIP1 SNAPC4 CCT8 UNG IRS4 NAP1L4
1 1 1 1 1 1 1 1 1 1
PBK BRCA1 PRKDC E2F4 PCNA E2F1 CHEK1 EFHC1 ITGA7 ALG5
0.132 0.132 0.132 0.132 0.132 0.132 0.132 0.116 0.116 0.116
E2F1 E2F4 EGFR GIPC1 TIAM1 LYN PRKDC RAF1 ATP5C1 MNAT1
178 98 96 37 34 33 31 29 29 26
E2F4 E2F1 EGFR CYCS RPS26 RPL8 RPS5 RPS10 NDUFS7 ATP5J
81 67 34 28 27 26 26 26 26 25
Each gene received a score for each algorithm.
Table 3 Hub genes detected using 12 different algorithms in the network derived from the STRING database MCC
DMNC
Gene NOP56 DKC1 DDX54 NOL10 NMD3 RBM19 MPHOSPH10 POLQ C13orf34 GNL2
PCNA MYCN POLA1 RFC4 PSMD14 CDKN1A GMPS NHP2 DKC1 TCP1
EPC
Radiality
Stress
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13
C13orf34 RBM19 CDCA7 KIF20B NOP14 THUMPD2 GTSE1 TTC27 KCTD19 GINS2
1.172 1.163 1.157 1.144 1.142 1.107 1.104 1.097 1.086 1.080
PCNA RFC4 NHP2 DKC1 GMPS ABCE1 NOP56 PSMD14 RSL24D1 POLA1
117 107 100 92 88 87 79 78 78 72
RFC4 PCNA NHP2 NOP56 RSL24D1 ABCE1 DKC1 PSMD14 GMPS POLR1C
304.41 303.79 301.54 301.47 301.18 301.02 300.58 299.17 297.35 295.30
PCNA MYCN POLA1 RFC4 CDKN1A KITLG PTEN PSMD14 ABL1 NPM1
8.2911 8.24066 8.19169 8.093 8.09008 8.07692 8.06522 8.05133 8.02355 8.02355
PCNA MYCN POLA1 CDKN1A PTEN RFC4 ABL1 GMPS LYN TFRC
1515402 1358060 1038576 718350 707432 700784 617368 604372 501280 492414
Closeness Gene
MNC
Betweenness
Clustering
Eccentricity
BN
Degree
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
612.22 585.88 577.95 571.73 554.27 552.28 546.45 546.43 545.02 544.87
MYCN PCNA POLA1 PTEN DNAH8 CDKN1A ABL1 TFRC POMC LYN
64,829.06 56,636.91 50,847.86 37,527.92 32,123.96 29,903.31 28,208.08 24,610.5 22,725.36 22,265.08
MYL7 DNAH3 NAALADL2 CARS2 PROP1 INPP4B COL27A1 ACRV1 ADORA2B PTH2R
1 1 1 1 1 1 1 1 1 1
NAP1L1 ABCB1 NOP56 PLCB4 DKC1 LARP7 COMP HAT1 DRG1 MRPL13
0.162 0.162 0.162 0.162 0.162 0.162 0.162 0.162 0.162 0.162
POLA1 EGFR CDKN1A PIK3CG PTEN COPS5 PCNA DNAH8 POMC TFRC
74 42 38 34 29 28 26 24 24 23
PCNA RFC4 NHP2 DKC1 GMPS ABCE1 MYCN PSMD14 POLA1 NOP56
120 107 100 92 88 88 80 80 80 79
Each gene received a score for each algorithm.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
8
Table 4 Gene ontogeny analysis of the hub genes detected in the canine DLBCL GRN Term GO:0006974wresponse to DNA damage stimulus GO:0033554wcellular response to stress GO:0006281wDNA repair GO:0006259wDNA metabolic process GO:0042770wDNA damage response, signal transduction GO:0010608wposttranscriptional regulation of gene expression GO:0006396wRNA processing GO:0051329winterphase of mitotic cell cycle GO:0051325winterphase GO:0008283wcell proliferation GO:0051726wregulation of cell cycle GO:0000278wmitotic cell cycle GO:0022403wcell cycle phase GO:0022402wcell cycle process GO:0000079wregulation of cyclin-dependent protein kinase activity GO:0007049wcell cycle GO:0042127wregulation of cell proliferation GO:0045859wregulation of protein kinase activity GO:0043549wregulation of kinase activity GO:0051338wregulation of transferase activity GO:0008284wpositive regulation of cell proliferation GO:0010035wresponse to inorganic substance GO:0043066wnegative regulation of apoptosis GO:0043069wnegative regulation of programmed cell death GO:0060548wnegative regulation of cell death GO:0008285wnegative regulation of cell proliferation GO:0050730wregulation of peptidyl-tyrosine phosphorylation GO:0001775wcell activation GO:0042981wregulation of apoptosis GO:0043067wregulation of programmed cell death GO:0010941wregulation of cell death GO:0043065wpositive regulation of apoptosis GO:0043068wpositive regulation of programmed cell death GO:0010942wpositive regulation of cell death GO:0006917winduction of apoptosis GO:0012502winduction of programmed cell death
Genes
Fold enrichment
P value
MNAT1, CDKN1A, LYN, UNG, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1, RPS3 MNAT1, CDKN1A, LYN, UNG, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1, RPS3 MNAT1, UNG, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1 MNAT1, UNG, CYCS, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1 BRIP1, CHEK1, ABL1, SMC1A, BRCA1
7.2536193
5.16E-07
4.780212
2.81E-05
7.1450704
3.00E-05
4.455863
3.17E-04
14.091667
4.03E-04
EIF5B, NDUFA13, PRKDC, GIPC1, RPS4X, RPS5, PRKCD POLR2H, POLR2F, POLR2E, RPS17, WDR77, SNRPA, SMC1A, POLR2B E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, ABL1 E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, ABL1 E2F1, EGFR, MNAT1, PCNA, RAF1, PRKCD E2F1, EGFR, MNAT1, CDKN1A, TSC2, BRIP1, CHEK1, SMC1A, BRCA1 E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A, BRCA1 EGFR, MNAT1, CDKN1A, CHEK1
7.4799368
3.03E-04
3.2975015
0.009395
15.322977
5.47E-06
14.889308
6.46E-06
3.1027523 6.1305136
0.041054 8.83E-05
5.4843243
1.90E-04
4.9014493
4.06E-04
3.9905605
7.05E-04
16.701235
0.001669
2.9054983
0.006104
2.578399
0.020238
3.2676329 3.1577965 3.0304659 2.7230274 3.2995122 1.9107345 1.8841226
0.06337 0.070007 0.07879 0.106154 0.224946 0.459516 0.466802
EGFR, MNAT1, CDKN1A CDKN1A, TSC2, CHEK1 EGFR, LYN, PRKCD
1.8788889 1.8736842 9.9470588
0.468253 0.469701 0.035445
EGFR, LYN, PRKDC, PRKCD EGFR, MNAT1, CDKN1A, TIAM1, CYCS, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 EGFR, MNAT1, CDKN1A, TIAM1, CYCS, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 EGFR, MNAT1, CDKN1A, TIAM1, CYCS, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, ABL1, BRCA1, RPS3
3.1423926 2.8043118
0.129717 0.007649
2.7766831
0.008142
2.7664622
0.008332
3.6703876
0.010832
3.6449577
0.011184
3.6281992
0.011423
4.2275 4.2143302
0.012671 0.012829 (Continued)
E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A, BRCA1 EGFR, MNAT1, CDKN1A, E2F4, LYN, TSC2, CHEK1, RPS4X, BRCA1 EGFR, MNAT1, CDKN1A, TSC2, CHEK1 EGFR, MNAT1, CDKN1A, TSC2, CHEK1 EGFR, MNAT1, CDKN1A, TSC2, CHEK1 EGFR, MNAT1, CDKN1A, LYN, RPS4X EGFR, MNAT1, CDKN1A EGFR, MNAT1, CDKN1A EGFR, MNAT1, CDKN1A
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
9
Gene Analysis in Canine Lymphoma Table 4 (continued ) Genes
Fold enrichment
P value
EGFR, CDKN1A, LYN, PRKDC, SMC1A, BRCA1 CDKN1A, LYN, PRKDC, BRCA1 CDKN1A, LYN, PRKDC, BRCA1 E2F1, RPS26, NDUFA13
3.676087 2.4574024 2.2268313 1.8685083
0.021864 0.215105 0.259248 0.471147
E2F1, MNAT1, E2F4, NDUFA13, PRKDC, BRIP1, CREB3L4, ABL1, BRCA1 EGFR, LYN, TIAM1, TSC2, ITGA7, RAF1, GIPC1
1.1445008
0.516764
0.8503592
0.837679
EGFR, LYN, TSC2, PCNA, NDUFA13, GIPC1 LYN, TSC2, PCNA, NDUFA13, GIPC1 EGFR, TSC2, PCNA, NDUFA13, GIPC1 LYN, TSC2, PCNA, NDUFA13, GIPC1 E2F1, MNAT1, GTF3A, E2F4, SNAPC4, NDUFA13, PRKDC, BRIP1, CREB3L4, ABL1, BRCA1, RPS3 LYN, RPS4X, BRCA1
1.5337868 1.4794401 2.7429035 1.4659731 1.0402153
0.340044 0.426981 0.104069 0.433927 0.597635
2.7835391
0.286452
1.243611
0.393199
Term GO:0009628wresponse to abiotic stimulus GO:0009725wresponse to hormone stimulus GO:0009719wresponse to endogenous stimulus GO:0051253wnegative regulation of RNA metabolic process GO:0006355wregulation of transcription, DNAdependent GO:0007166wcell surface receptor-linked signal transduction GO:0008104wprotein localization GO:0015031wprotein transport GO:0034613wcellular protein localization GO:0045184westablishment of protein localization GO:0045449wregulation of transcription GO:0051247wpositive regulation of protein metabolic process GO:0051252wregulation of RNA metabolic process
E2F1, RPS26, MNAT1, E2F4, NDUFA13, PRKDC, BRIP1, CREB3L4, ABL1, BRCA1
occurs in patients with DLBCL. Inositol polyphosphate 4-phosphatase-II (INPP4B) also acts as a tumour suppressor because of its regulatory role in the PI3K signalling pathway. INPP4B knockdown resulted in AKT activation and accelerated cell proliferation. Downregulation of INPP4B in tumours was accompanied by loss of PTEN function (Fedele et al., 2010). Another key pathway in the analysis was the p53 signalling pathway. Related hub genes included E2F1, CDKN1A, PCNA, PIK3CG, PTEN and GTSE1. The role of E2F1 and E2F4 in the development of sporadic Burkitt’s lymphoma (sBL) has been investigated (Molina-Privado et al., 2009, 2012). E2F1 is highly expressed by sBL cells and inhibition of its expression suppresses tumour formation. E2F1 and C-MYC may work together in sBL. Consistent with these findings, overexpression of E2F1 was shown in this study (Supplementary Table 1). PCNA as a tumour proliferation activity marker was strongly correlated with grading score in NHL (Korkolopoulou et al., 1993, 1994; Rabenhorst et al., 1996) and PCNA, together with c-AgNORs and myc p62, can efficiently estimate proliferation activity in NHL (Korkolopoulou et al., 1993). Furthermore, significant correlations were found between P53 and PCNA expression in human patients with NHL (Korkolopoulou et al., 1994). CDKN1A (p21), another component of the p53 pathway, was shown to be downregulated in adult T-cell leukaemia/lymphoma by an epigenetic event induced by the PI3K/AKT signalling pathway (Watanabe et al., 2010). Downregulation of CDKN1A and activation of the PI3K/AKT
signalling pathway were also found in the present study. Expression and function of p21 is dependent on p53 protein, and absence of the both markers was shown in 67.7% of 253 people with NHL (Chilosi et al., 1996). Telomeres are repeated nucleotide sequences that preserve chromosomes from harmful effects. Telomerase and alternative lengthening of telomeres (ALT) are considered as important mechanisms that maintain telomere length. Telomere shortening originates from mutations in genes encoding telomerase or telomere-binding protein as well as mutations in DKC1, NOP10, TERT, TINF2 and TERC (Calado and Young, 2008). In the present study, RFC4, DKC1, POLA1 and PCNA were listed as genes involved in telomere maintenance. Dysregulation of RFC4 and 11 other genes, including IGF2BP3, UPP1, CDKN2B, CDKN2A, RMI1, SLBP, AUTS2, FUT8, ACY1, BCL2 and PDXK, has been detected in human DLBCL (Jardin et al., 2010). DKC1 and AKT2 expression levels were also found to be downregulated by miR-150, which functioned as a tumour suppressor. Additionally, miR-150 exerted its effects by amplifying the protein levels of Bim and p53. In this context, downregulation of miR-150 and subsequent activation of the PI3K/AKT pathway, as well as telomerase, was confirmed in malignant lymphoma (Watanabe et al., 2011). The role of POLA1 in telomere maintenance has also been investigated (Fujita et al., 2013). The results of the present study implicated E2F1, CDKN1A, ABL1 and RAF1 as gene components incorporated in two cell cycle pathways, including G1 and Rac CycD pathways. The G1/S cell cycle
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
10
Table 5 Pathway analysis of the hub genes detected in the canine DLBCL GRN Genes
Fold enrichment
P value
E2F1, CDKN1A, ABL1 E2F1, CDKN1A, PCNA E2F1, CDKN1A, RAF1
7.823956 13.34675 9.453947
0.049278 0.018038 0.034759
LYN, TIAM1, RAF1, PRKCD PIK3CG, PLCB4, INPP4B, PTEN E2F1, CDKN1A, E2F4, PCNA, PRKDC, CHEK1, ABL1, SMC1A EGFR, PIK3CG, MYL7, COMP, PTEN LYN, RAF1, PRKCD EGFR, PIK3CG, CDKN1A, KITLG, ABL1, PTEN E2F1, EGFR, RAF1 EGFR, PIK3CG, PTEN PIK3CG, CDKN1A, PTEN PIK3CG, CDKN1A, PTEN PIK3CG, CDKN1A, PTEN E2F1, EGFR, CDKN1A, RAF1 PIK3CG, CDKN1A, ABL1 E2F1, EGFR, RAF1 LYN, RAF1, PRKCD POLR2H, POLR2F, POLR2E PIK3CG, PLCB4, PTEN
2.219797 10.18018 6.641633
0.258506 0.006049 1.46E-04
4.684909 3.991366 3.445122 4.32398 10.86538 11.95767 8.464419 10.61033 9.883382 7.533333 5.765306 5.307121 11.87119 1.566338
0.017892 0.167306 0.023173 0.147411 0.028505 0.003846 0.010074 0.005388 0.007002 0.055547 0.09163 0.098415 0.023052 0.553964
POLR2H, POLR2F, POLR2E, CREB3L4
11.34856
0.004041
PIK3CG, CDKN1A, PTEN, GTSE1 RFC4, DKC1, POLA1, PCNA
5.290741 12.13571
0.031942 0.003408
Term h_g1 pathway: cell cycle: G1/S check point* h_p53pathway: p53 signalling pathway* h_Rac CycD pathway: influence of Ras and Rho proteins on G1 to S transition* hsa04062: chemokine signalling pathway† hsa04070: phosphatidylinositol signalling system hsa04110: cell cycle hsa04510: focal adhesion hsa04664: Fc epsilon RI signalling pathway hsa05200: pathways in cancer hsa05212: pancreatic cancer hsa05213: endometrial cancer hsa05214: glioma hsa05215: prostate cancer hsa05218: melanoma hsa05219: bladder cancer hsa05220: chronic myeloid leukaemia (CML) hsa05223: non-small cell lung cancer (NSCLC) P00010: B-cell activation‡ P00023: general transcription regulation‡ P00031: inflammation mediated by chemokine and cytokine signalling pathway‡ P00055: transcription regulation by b ZIP transcription factor‡ P00059: p53 pathway‡ REACT_7970: telomere maintenancex *
BIOCARTA databases. hsa identifier belonging to KEGG database. ‡ PANTHER database. x REACTOME database. †
checkpoint controls transition from the G1 phase into the DNA synthesis S phase. Different stimulatory or inhibitory factors regulate this complex transition. For example, CDK2-Cyclin E and CDK4/6-Cyclin D accelerate the G1-S transition, while p21, p27 and p57 reduce the transition through inhibition of cyclineCDK complexes (Nojima, 2004). Previous studies have shown that canine DLBCL rarely expresses BCL6 (Richards et al., 2013) and this was supported by the findings of the present study, which did not detect BCL6 as a hub gene. To the best of our knowledge, this has been the first study to reconstruct and explore canine lymphoma GRNs. However, the study suffered from a major limitation. Although microarray data from human cancers are very extensive and informative, microarray data related to animal cancers are rare and incomplete. There are only two canine lymphoma datasets, with sample sizes of 33 and 58 patients. Analysis of datasets with larger sample numbers yields more realistic and accurate networks and subsequently detects more subtle hub genes.
Acknowledgment This project was supported by grant No. 15674 from the Shahid Bahonar University of Kerman.
Conflict of Interest Statement The authors report no conflict of interest.
Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jcpa.2014.11. 008.
References Agnelli L, Forcato M, Ferrari F, Tuana G, Todoerti K et al. (2011) The reconstruction of transcriptional networks reveals critical genes with implications for clinical outcome of multiple myeloma. Clinical Cancer Research: An Official Journal of the American Association for Cancer Research, 17, 7402e7412.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
Gene Analysis in Canine Lymphoma
Assenov Y, Ramırez F, Schelhorn S-E, Lengauer T, Albrecht M (2008) Computing topological parameters of biological networks. Bioinformatics, 24, 282e284. Bansal M, Belcastro V, Ambesi-Impiombato A, di Bernardo D (2007) How to infer gene networks from expression profiles. Molecular Systems Biology, 3, 78. Basso K, Margolin AA, Stolovitzky G, Klein U, DallaFavera R et al. (2005) Reverse engineering of regulatory networks in human B cells. Nature Genetics, 37, 382e390. Basso K, Saito M, Sumazin P, Margolin AA, Wang K et al. (2010) Integrated biochemical and computational approach identifies BCL6 direct target genes controlling multiple pathways in normal germinal center B cells. Blood, 115, 975e984. Blais A, Dynlacht BD (2005) Constructing transcriptional regulatory networks. Genes and Development, 19, 1499e1511. Butler MP, Wang SI, Chaganti RS, Parsons R, DallaFavera R (1999) Analysis of PTEN mutations and deletions in B-cell non-Hodgkin’s lymphomas. Genes Chromosomes and Cancer, 24, 322e327. Calado RT, Young NS (2008) Telomere maintenance and human bone marrow failure. Blood, 111, 4446e4455. Chilosi M, Doglioni C, Magalini A, Inghirami G, Krampera M et al. (1996) p21/WAF1 cyclin-kinase inhibitor expression in non-Hodgkin’s lymphomas: a potential marker of p53 tumor-suppressor gene function. Blood, 88, 4012e4020. Cline MS, Smoot M, Cerami E, Kuchinsky A, Landys N et al. (2007) Integration of biological networks and gene expression data using Cytoscape. Nature Protocols, 2, 2366e2382. Croce CM (2008) Oncogenes and cancer. New England Journal of Medicine, 358, 502e511. Emmert-Streib F, de Matos Simoes R, Mullan P, HaibeKains B, Dehmer M (2014) The gene regulatory network for breast cancer: integrated regulatory landscape of cancer hallmarks. Frontiers in Genetics, 5, 15. Engelman JA, Luo J, Cantley LC (2006) The evolution of phosphatidylinositol 3-kinases as regulators of growth and metabolism. Nature Reviews Genetics, 7, 606e619. Fedele CG, Ooms LM, Ho M, Vieusseux J, O’Toole SA et al. (2010) Inositol polyphosphate 4-phosphatase II regulates PI3K/AKT signaling and is lost in human basal-like breast cancers. Proceedings of the National Academy of Sciences of the United States of America, 107, 22231e22236. Floratos A, Smith K, Ji Z, Watkinson J, Califano A (2010) geWorkbench: an open source platform for integrative genomics. Bioinformatics, 26, 1779e1780. Franceschini A, Szklarczyk D, Frankild S, Kuhn M, Simonovic M et al. (2013) STRING v9.1: proteinprotein interaction networks, with increased coverage and integration. Nucleic Acids Research, 41, D808eD815. Fujita T, Asano Y, Ohtsuka J, Takada Y, Saito K et al. (2013) Identification of telomere-associated molecules by engineered DNA-binding molecule-mediated chro-
11
matin immunoprecipitation (enChIP). Scientific Reports, 3, 3171. Gautier L, Cope L, Bolstad BM, Irizarry RA (2004) Affy e analysis of Affymetrix GeneChip data at the probe level. Bioinformatics, 20, 307e315. Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell, 100, 57e70. Huang DW, Sherman BT, Lempicki RA (2009a) Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Research, 37, 1e13. Huang DW, Sherman BT, Lempicki RA (2009b) Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nature Protocols, 4, 44e57. Jardin F, Jais J-P, Molina T-J, Parmentier F, Picquenot JM et al. (2010) Diffuse large B-cell lymphomas with CDKN2A deletion have a distinct gene expression signature and a poor prognosis under R-CHOP treatment: a GELA study. Blood, 116, 1092e1104. Korkolopoulou P, Oates J, Kittas C, Crocker J (1994) p53, c-myc p62 and proliferating cell nuclear antigen (PCNA) expression in non-Hodgkin’s lymphomas. Journal of Clinical Pathology, 47, 9e14. Korkolopoulou P, Patsouris E, Pangalis G, Tsenga A, Elemenoglou J et al. (1993) A comparative assessment of proliferating cell nuclear antigen, c-myc p62, and nucleolar organizer region staining in non-Hodgkin’s lymphomas: a histochemical and immunohistochemical study of 200 cases. Human Pathology, 24, 371e377. Liu X, Karnell JL, Yin B, Zhang R, Zhang J et al. (2010) Distinct roles for PTEN in prevention of T-cell lymphoma and autoimmunity in mice. Journal of Clinical Investigation, 120, 2497e2507. MacEwen EG (1990) Spontaneous tumors in dogs and cats: models for the study of cancer biology and treatment. Cancer Metastasis Reviews, 9, 125e136. Maere S, Heymans K, Kuiper M (2005) BiNGO: a Cytoscape plugin to assess overrepresentation of gene ontology categories in biological networks. Bioinformatics, 21, 3448e3449. McCaw DL, Chan AS, Stegner AL, Mooney B, Bryan JN et al. (2007) Proteomics of canine lymphoma identifies potential cancer-specific protein markers. Clinical Cancer Research, 13, 2496e2503. Molina-Privado I, Jimenez-P R, Montes-Moreno S, Chiodo Y, Rodrıguez-Martınez M et al. (2012) E2F4 plays a key role in Burkitt lymphoma tumorigenesis. Leukemia, 26, 2277e2285. Molina-Privado I, Rodrıguez-Martınez M, Rebollo P, Martın-Perez D, Artiga M-J et al. (2009) E2F1 expression is deregulated and plays an oncogenic role in sporadic Burkitt’s lymphoma. Cancer Research, 69, 4052e4058. Mudaliar MAV, Haggart RD, Miele G, Sellar G, Tan KAL et al. (2013) Comparative gene expression profiling identifies common molecular signatures of NF-kB activation in canine and human diffuse large B cell lymphoma (DLBCL). PLoS One, 8, e72591.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
12
M. Zamani-Ahmadmahmudi et al.
Nojima H (2004) G1 and S-phase checkpoints, chromosome instability, and cancer. Methods in Molecular Biology, 280, 3e49. Ponce F, Marchal T, Magnol JP, Turinelli V, Ledieu D et al. (2010) A morphological study of 608 cases of canine malignant lymphoma in France with a focus on comparative similarities between canine and human lymphoma morphology. Veterinary Pathology, 47, 414e433. Qiao Q, Jiang Y, Li G (2013) Inhibition of the PI3K/AKTNF-kB pathway with curcum in enhanced radiationinduced apoptosis in human Burkitt’s lymphoma. Journal of Pharmacological Sciences, 121, 247e256. Rabenhorst SH, Burini RC, Schmitt FC (1996) Proliferating cell nuclear antigen (PCNA) in non-Hodgkin’s lymphomas: correlation with working formulation and Kiel classification in formalin-fixed paraffin-embedded material. Pathology, 28, 12e16. Richards KL, Motsinger-Reif AA, Chen H-W, Fedoriw Y, Fan C et al. (2013) Gene profiling of canine B-cell lymphoma reveals germinal center and postgerminal center subtypes with different survival times, modeling human DLBCL. Cancer Research, 73, 5029e5039. Sakai A, Thieblemont C, Wellmann A, Jaffe ES, Raffeld M (1998) PTEN gene alterations in lymphoid neoplasms. Blood, 92, 3410e3415. Tarcea VG, Weymouth T, Ade A, Bookvich A, Gao J et al. (2009) Michigan molecular interactions r2: from interacting proteins to pathways. Nucleic Acids Research, 37, D642eD646.
Vail DM, MacEwen EG (2000) Spontaneously occurring tumors of companion animals as models for human cancer. Cancer Investigation, 18, 781e792. Watanabe A, Tagawa H, Yamashita J, Teshima K, Nara M et al. (2011) The role of microRNA-150 as a tumor suppressor in malignant lymphoma. Leukemia, 25, 1324e1334. Watanabe M, Nakahata S, Hamasaki M, Saito Y, Kawano Y et al. (2010) Downregulation of CDKN1A in adult T-cell leukemia/lymphoma despite overexpression of CDKN1A in human T-lymphotropic virus 1-infected cell lines. Journal of Virology, 84, 6966e6977. Xu Z-Z, Xia Z-G, Wang A-H, Wang W-F, Liu Z-Y et al. (2013) Activation of the PI3K/AKT/mTOR pathway in diffuse large B-cell lymphoma: clinical significance and inhibitory effect of rituximab. Annals of Hematology, 92, 1351e1358. Zhang J, Grubor V, Love CL, Banerjee A, Richards KL et al. (2013) Genetic heterogeneity of diffuse large Bcell lymphoma. Proceedings of the National Academy of Sciences of the United States of America, 110, 1398e1403.
July 10th, 2014 ½ Received, Accepted, November 29th, 2014
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
Available online at www.sciencedirect.com
ScienceDirect www.elsevier.com/locate/jcpa
NEOPLASTIC DISEASE
Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes M. Zamani-Ahmadmahmudi*,†, A. Najafi‡ and S. M. Nassirix * Department of Clinical Science, Faculty of Veterinary Medicine, Shahrekord University, Shahrekord, † Department of Clinical Science, Faculty of Veterinary Medicine, Shahid Bahonar University of Kerman, Kerman, ‡ Molecular Biology Research Center, Baqiyatallah University of Medical Sciences and x Department of Clinical Pathology, Faculty of Veterinary Medicine, University of Tehran, Tehran, Iran
Summary Lymphoma is one of the most common malignancies in dogs. Canine lymphoma is similar to human nonHodgkin’s lymphoma (NHL) with shared clinical presentation and histopathological features. This study reports the construction of a comprehensive gene regulatory network (GRN) for canine diffuse large B-cell lymphoma (DLBCL), the most common type of canine lymphoma, and performs analysis for detection of major functional modules and hub genes (the most important genes in a GRN). The canine DLBCL GRN was reconstructed from gene expression data (NCBI GEO dataset: GSE30881) using the STRING and MiMI interaction databases. Reconstructed GRNs were then assessed, using various bioinformatics programmes, in order to analyze network topology and identify major pathways and hub genes. The resultant network from both interaction databases had a logically scale-free pattern. Gene ontology (GO) analysis revealed cell activation, cell cycle phase, immune effector process, immune system development, immune system process, integrinmediated signalling pathway, intracellular protein kinase cascade, intracellular signal transduction, leucocyte activation and differentiation, lymphocyte activation and differentiation as major GO terms in the biological processes of the networks. Moreover, bioinformatics analysis showed E2F1, E2F4, PTEN, CDKN1A, PCNA, DKC1, MNAT1, NDUFB4, ATP5J, PRKDC, BRCA1, MYCN, RFC4 and POLA1 as the most important hub genes. The phosphatidyl inositol signalling system, P53 signalling pathway, Rac CycD pathway, G1/S checkpoint, chemokine signalling pathway and telomere maintenance were the main signalling pathways in which the protein products of the hub genes are involved. Ó 2014 Elsevier Ltd. All rights reserved. Keywords: dog; gene regulatory network; hub genes; lymphoma
Introduction Canine lymphoma is one of the most common malignancies in dogs and most commonly has multicentric anatomical distribution (MacEwen, 1990). Canine lymphoma shares clinical and histopathological features with human non-Hodgkin’s lymphoma (NHL) (Vail and MacEwen, 2000; McCaw et al., 2007). Histopathological classification of canine lymphoma
Correspondence to: M. Zamani-Ahmadmahmudi (e-mail: zamani_2012@ alumni.ut.ac.ir). 0021-9975/$ - see front matter http://dx.doi.org/10.1016/j.jcpa.2014.11.008
has been performed using three classification schemes the updated Kiel scheme, the Revised European and American Lymphoma (REAL) scheme and the World Health Organisation (WHO) scheme. These classifications consider epidemiological, clinical, clinicomorphological, genetic and phenotypic parameters. The most common form of canine lymphoma is diffuse large B-cell lymphoma (DLBCL) (Ponce et al., 2010). Cancer develops through dysregulation of normal cellular processes including apoptosis, cell mitosis, cell cycling, cell differentiation and DNA repair. Ó 2014 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
2
Neoplastic transformation results in activation of oncogenes, deactivation of tumour suppressor genes and/or genome instability (Croce, 2008; EmmertStreib et al., 2014) and various cell processes are involved in cancer (Hanahan and Weinberg, 2000). Cancer involves intricate interactions between different signalling components such as genes, proteins and metabolites, and as such interactions do not follow a simple chain-like pattern, it has proven challenging to understand the molecular interactions underlying oncogenesis (Emmert-Streib et al., 2014). One of the most useful approaches in providing a global overview of cancer pathways has been reconstruction of ‘gene regulatory networks’ (GRNs) using various algorithms or interaction databases (Blais and Dynlacht, 2005; Bansal et al., 2007; Emmert-Streib et al., 2014). GRNs can yield valuable information about major gene ontology modules and are able to identify major genes (‘hub genes’) involved in development of a specific tumour (Basso et al., 2005, 2010; Emmert-Streib et al., 2014). Through reconstruction and analysis of a breast cancer GRN, Emmert-Streib et al. (2014) characterized the major cellular processes contributing to this form of neoplasia, including cell cycling, cell adhesion, translation, organelle fission, the immune response and mitosis. Using a reconstruction of a transcriptional network, Agnelli et al. (2011) identified the most critical genes associated with poor prognosis in patients with multiple myeloma. These methods have not yet been applied widely to the investigation of canine cancer; however, microarray data exist for canine mammary gland tumours, canine lymphoma and canine haemangiosarcoma (http://www.ncbi.nlm.nih.gov/gds/?term¼canine% 20lymphoma). The aim of the current study was to reconstruct a canine DLBCL GRN and evaluate it using different analytical methods. Specifically, the study defined major functional modules (GO), hub genes (highly connected nodes or the most important genes in a GRN) and the functional biological processes associated with the identified hub genes.
Materials and Methods Data Collection and Primary Processing
Gene expression data were obtained from the NCBI Gene Expression Omnibus GEO dataset (GSE30881, platform: GPL3738) (Mudaliar et al., 2013). Data included the gene expression profile of 23 canine DLBCLs and normal lymph nodes from 10 healthy dogs. Data were downloaded in the CEL file format and converted to expression values by the Affy package (Gautier et al., 2004) in R program,
version 3.0.2 (http://www.r-project.org/). As part of this process, data were transformed logarithmically. Then, the data (probe sets) were entered into the geWorkbench 2.5.1 package (Floratos et al., 2010) and non-useful data were filtered based on two criteria: data without Enterz identification were omitted and markers with multiple probset identifications were filtered, while probsets with the highest mean expression value were retained. Upregulation or downregulation of the genes (P <0.01) was determined by use of a t test provided in the geWorkbench package (correction method: just alpha, group variance: equal). k-means Clustering and Principle Component Analysis
Genes were clustered through two algorithms including k-means clustering and principle component analysis (PCA). Using k-means clustering, data were partitioned into specified numbers of groupings (k) in which each observation belonged to the cluster with the nearest mean. PCA reduced dimensionality through generating few linear combinations of all data that are named ‘principle components’. Principle components can extract clusters from the original data. Both analyses were performed in MATLAB 7.8.0 (R2009a) (MathWorks, Natick, Massachusetts, USA). Network Reconstruction and Analysis
The GRN was constructed using two major interaction databases, including MiMI (Michigan Molecular Interactions) (Tarcea et al., 2009) and Search Tool for the Retrieval of Interacting Genes/Proteins (STRING; Franceschini et al., 2013). Briefly, the list of genes was submitted online or through a related plug-in within Cytoscape software (Cline et al., 2007) and then interactions between genes were retrieved and displayed as an interactions network. MiMI collects real interaction data from interaction databases such as the Swiss Protein Database (SwissProt), the Human Protein Reference Database (HPRD), the Biomolecular Interaction Network Database (BIND), the Database of Interacting Proteins (DIP), Reference Sequence (RefSeq) and the International Protein Index (IPI). Interaction data from STRING originate from four major sources including high-throughput experiments, genomic context, co-expression and previous knowledge. Constructed networks were imported to Cytoscape and analyzed using various plug-ins for different aspects of network features. The networks obtained from these databases were analyzed by NetworkAnalyzer plug-in (Assenov et al., 2008) in order to determine topological parameters and centrality measures, including the network
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
3
Gene Analysis in Canine Lymphoma
diameter, radius, density, centralization, heterogeneity, clustering coefficient, connected components, the number of nodes/edges, the number of self-loops, the shortest path length and the characteristic path length. This plug-in also computed node degree distribution, neighbourhood connectivity distribution, shared neighbour distribution, betweenness centrality, closeness centrality and stress centrality distribution. Detection of Functional Modules Using Gene Ontology Analysis
Overrepresented GO terms (especially biological pathways, BPs) were detected in the whole network using the BINGO plug-in (version 2.4) (Maere et al., 2005) in the Cytoscape software. This plug-in identified statistically significant functional modules in a given gene set and displayed them as a Cytoscape network. The Biological Networks gene ontology Tool (BINGO) used hypergeometric testing for statistical analysis and the Benjamini and Hochberg false discovery rate (FDR) correction for multiple testing correction. P <0.05 was considered significant. Detection of Hub Genes
Hub genes (genes with a high number of edges) were explored using the CytoHubba software (release 16) (http://hub.iis.sinica.edu.tw/cytoHubba/index.html). Hubba nodes (genes) were detected by several ranking methods as well as closeness, betweenness, radiality,
stress, degree, edge percolated component (EPC), density of maximum neighbourhood component (DMNC), maximal clique centrality (MCC), maximum neighbourhood component (MNC), bottleneck (BN) and eccentricity. The top 10 genes with the highest score were selected for each parameter. To demonstrate the role of hub genes in biological processes and cancer pathways, functional annotation clustering was also conducted using the Database for Annotation, Visualization and Integrated Discovery (DAVID; Huang et al., 2009a, 2009b). Ontology parameters were adjusted as follows: classification stringency of medium, similarity term overlap of 3.0, similarity threshold of 0.05, multiple linkage threshold of 0.05 and an enrichment threshold of 1.0.
Results Global Properties of Genes with Significant Expression Change
Data filtration using the geWorkbench package extracted 16,970 genes from a total of 43,035 genes. t test analysis revealed 2,087 genes with significantly different expression between neoplastic and healthy samples (Supplementary Table 1). k-means clustering and PCA assays were performed for evaluation of global expression profiles of upregulated or downregulated genes. k-means clustering divided the genes into 16 major clusters with genes in each cluster possessing a similar expression pattern in studied individuals (Fig. 1, Supplementary
Fig. 1. k-means clustering of genes with significantly different expression between samples from dogs with neoplasia and healthy dogs. Analysis indicates 16 different clusters. Expression of each gene is plotted in 33 animals. The first 23 animals have neoplasia and the last 10 animals are healthy. Vertical axes indicate gene expression values.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
4
M. Zamani-Ahmadmahmudi et al.
Fig. 2. PCA for the gene expression profile of canine lymphoma reveals two major clusters that clearly indicate neoplastic and healthy samples. Each cluster is defined by a red ellipse.
Table 2). In addition, the PCA analysis grouped samples into two major clusters that clearly separated neoplastic (n ¼ 23) and healthy samples (n ¼ 10) (Fig. 2). PCA results were confirmed by generating a clustergram in MATLAB (Fig. 3). Gene Regulatory Network Reconstruction
A GRN was reconstructed using the MiMI and STRING interaction databases for genes with significant expression change. The MiMI database delivered a network with 710 nodes (genes) and 1,835
edges (interactions), while STRING yielded a network with 1,022 nodes and 5,803 edges (Supplementary Cytoscape File 1). General topological analysis for both networks was performed by use of NetworkAnalyzer (Table 1). The number of connected components was 25 in the MiMI network and 13 in the STRING network, indicating that a number of subnetworks comprised a network. A lower number of connected components implies strong connectivity; however, the networks identified in the present study also had strong connectivity because 658 nodes (1,806 edges) in the MiMI network and 987
Fig. 3. Clustergram of genes with significant expression changes in the canine dataset. On the left side of the image is a scale bar indicating expression fold (from minus threefold [downregulation] to threefold [overexpression]). The clustergram separates the samples from dogs with neoplasia (1e23) and healthy dogs (24e33).
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
Gene Analysis in Canine Lymphoma Table 1 Simple topological parameters of GRNs from MiMI and STRING databases Parameter
MiMI
STRING
Number of nodes 710 1022 Network diameter* 13 10 Network radius† 1 1 Network centralization‡ 0.107257 0.086990839 Network densityx 0.007291 0.01112259 Network heterogeneityǁ 1.427738 1.219856226 Number of connected components{ 25 16 Number of shortest paths# 432376 (85%) 973236 (93%) Characteristics path length** 4.832391 3.745058752 Average number of neighbours 5.169014 11.35616438 *
The highest distance between two nodes, where the distance is minimum number of edges that connected two nodes. † The smallest distance between two nodes. ‡ How does the network topology resemble a star structure? (A value between 0 and 1). x How densely is the network populated with edges? (A value between 0 and 1). ǁ This parameter indicates the affinity of a network to contain hub markers. { Number of subnetworks that constitute a network. # Shortest path: minimum numbers of edges form a path. Shortest paths defined as distance. ** Averaged shortest path length.
nodes (5,781 edges) in the STRING network participated in one component, while 52 nodes (29 edges) in the MiMI network and 35 nodes (22 edges) in the STRING network constituted 24 and 15 connected components, respectively. In other words, pairwise connected nodes falsely increased the number of connected components (Supplementary Cytoscape File 1). Network centralization, density, and average number of neighbours are parameters that describe neighbourhood status in networks (http://med. bioinf.mpi-inf.mpg.de/netanalyzer/index.php). Expectedly, both networks showed a biological scale-free pattern, as many nodes had low numbers of interactions and a few nodes were highly connected (Figs. 4A, B; Supplementary Cytoscape File 1). Neighbourhood connectivity distribution followed an assortive pattern, which means that nodes had a tendency to connect with nodes with similar connectivity. An assortive model implicates a hierarchical clustering for studied probsets (Figs. 4C, D). Gene Ontology Analysis
GO analysis was conducted for identifying important BPs involved in the pathogenesis of canine DLBCL. The full list of statistically significant GO terms (BPs) that was common between both networks is provided in Supplementary Table 3. Respectively, 545 and 334 significant GO terms were detected in
5
the MiMI and STRING networks and 300 of these GO terms were common to both networks. Some of the important GO terms included cell activation, cell cycle phase, immune effector process, immune system development, immune system process, integrin-mediated signalling pathway, intracellular protein kinase cascade, intracellular signal transduction, leucocyte activation and differentiation, lymphocyte activation and differentiation, positive regulation of ab T-cell activation and positive regulation of ab T-cell differentiation. Hub Genes
Various algorithms found hub genes in both GRNs (Tables 2 and 3). Some of the hub genes in the MiMI-derived GRN included E2F1, E2F4, MNAT1, NDUFB4, ATP5J, PRKDC, BRCA1, CDKN1A and PCNA. Some of the important hub genes in the STRING-derived GRN were PCNA, CDKN1A, MYCN, POLA1, PTEN, RFC4, EGFR, DKC1 and POLA1. Additionally, functional analysis revealed the role of the hub genes in biological processes, especially cancer pathways. Ontology analysis was performed using the DAVID database (Table 4), which indicated that most of the biological processes were associated with the cell cycle, mitosis, proliferation, differentiation, death, apoptosis, DNA damage and DNA repair. Furthermore, a number of genes involving in cancer biology (e.g. E2F1, CDKN1A, RAF1, ABL1, PCNA, PRKDC, CHEK1, SMC1A, CYCS, TSC2, ITGA7 and CREB3L4) were identified in KEGG, REACTOME and other pathway databases (Table 5). Functional analysis of the hub genes revealed that these genes were mainly involved in the P53 signalling pathway, Rac CycD pathway, G1/S checkpoint, chemokine signalling pathway, phosphatidylinositol signalling system and telomere maintenance. As cancer pathways in the KEGG database are limited to some specific types of cancers, including colorectal, pancreatic, thyroid, bladder, prostate, endometrial, non-small cell lung cancer, renal cell carcinoma, melanoma, basal cell carcinoma, glioma, acute myeloid leukemia and chronic myeloid leukemia, pathways were only retrieved for these tumours and not for other cancers such as lymphoma or breast cancer (Table 5). PIK3CG, CDKN1A, KITLG, ABL1 and PTEN were found in different cancer pathways. PIK3CG, CDKN1A and PTEN play a critical role in melanoma, glioma and endometrial and prostate cancer. Ontology findings revealed E2F1, EGFR and RAF1 as significant components in pancreatic cancer and non-small cell lung cancer. PIK3CG, CDKN1A and ABL1 were also found as hub markers in chronic myeloid leukaemia (CML).
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
6
Fig. 4. Node degree distribution in (A) the MiMI network and (B) the STRING network. Neighbourhood connectivity distribution in (C) the MiMI network and (D) the STRING network. Both networks have a scale-free pattern because many nodes have a low number of neighbours (A and B). In addition, an assortive pattern is evident for both networks (C and D).
Discussion The aim of the present study was to clarify key biological processes, pathways and hub genes in canine DLBCL by analyzing major GO terms (especially biological processes) in the GRNs obtained from the MiMI and STRING databases. The network was then investigated in more depth in order to identify functional modules (subnetworks) and their related genes. Both GRNs were ‘logic networks’, as they had a biological scale-free pattern and assortive model of neighbourhood connectivity distribution. Several pathways were highlighted from these analyses. One was the PI3K/AKT pathway, one of the critical and common pathways in human lymphoma biology (Jardin et al., 2010; Qiao et al., 2013; Xu et al., 2013; Zhang et al., 2013). Four hub genes (i.e. PTEN, PIK3CG, PLCB4 and INPP4B) identified in the study are involved in the PI3K/AKT pathway. Mutation, amplification and translocation of various components of this pathway occur in different cancers. This pathway was reported to possess prognostic power in patients with DLBCL, as patients with PI3K/AKT/mTOR activation were characterized as having more severe symptoms, a
poor response to cancer therapy and poor survival. Activity of the PI3K/AKT/mTOR pathway can be inhibited by rituximab combined with rapamycin (Xu et al., 2013), and inhibition of this pathway and its downstream regulator NF-kB was shown to promote apoptosis in three human Burkitt’s lymphoma cell lines (Qiao et al., 2013). PTEN, a key gene in this pathway, serves as a tumour suppressor, and its dysregulation is recognized in some cancers, including endometrial carcinoma, prostatic carcinoma, melanoma, glioma and NHL. Somatic mutation of PTEN was reported in NHL and Burkitt’s lymphoma cell lines (Sakai et al., 1998; Butler et al., 1999). PTEN induces its tumour suppressor function through accumulation of the phosphatidylinositol-3, 4, 5triphosphate (PIP3) in the PI3K/AKT signalling pathway (Engelman et al., 2006). Mutation of PTEN also triggers the onset of T-cell lymphoma in mice, in which lymphomagenesis and overexpression of C-MYC occur when there was deficiency of PTEN protein (Liu et al., 2010). Zhang et al. (2013) suggested that PIK3CD was an oncogene that might be a useful therapeutic target in DLBCL, as a single point mutation in the catalytic domain of PIK3CD
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
7
Gene Analysis in Canine Lymphoma Table 2 Hub genes detected using 12 different algorithms in the network obtained from the MiMI database MCC
DMNC
Gene NDUFB4 ATP5J RPLP0 EIF3A NDUFB5 RPS4X RPL8 EIF2S2 ATP5G1 NDUFS8
MNC
EPC
Radiality
Stress
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13
EIF5B NDUFB5 NDUFS8 NDUFA10 NDUFB7 NDUFA11 NDUFA1 ND2 NDUFA13 NDUFA2
1.124 1.119 1.119 1.119 1.119 1.119 1.119 1.119 1.119 1.119
E2F1 E2F4 RPS5 RPS10 RPS17 NDUFB4 RPS4X RPS3 CYCS RPS26
35 33 26 25 25 24 24 24 24 24
E2F1 E2F4 MNAT1 SMC1A POLR2E POLR2B RPS5 POLR2H RPS26 POLR2F
111.97 109.14 107.83 106.31 105.46 105.44 104.88 104.80 104.77 104.76
E2F1 E2F4 PRKDC BRCA1 MNAT1 PCNA PRKCD ABL1 CHEK1 SNRPA
10.170 10.101 9.980 9.918 9.900 9.829 9.805 9.778 9.751 9.740
E2F1 E2F4 EGFR TIAM1 PRKDC MNAT1 ATP5C1 ABL1 PRKCD COX5B
483784 418270 250102 181610 146578 111758 110216 110024 109822 99,496
Closeness Gene E2F4 E2F1 MNAT1 PRKDC BRCA1 TSC2 PCNA EGFR SNRPA CDKN1A
Betweenness
Clustering
Eccentricity
BN
Degree
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
270.62 269.91 235.42 233.36 232.69 226.03 225.90 223.71 221.84 219.58
E2F1 E2F4 EGFR TIAM1 PRKDC GIPC1 LYN BRCA1 ABL1 PRKCD
51,034.37 49,588.15 29,403.27 14,397.02 14,110.83 13,369.13 11,789.21 11,529.26 10,660.97 10,277.8
ALG5 GTF3A WDR77 CREB3L4 BRIP1 SNAPC4 CCT8 UNG IRS4 NAP1L4
1 1 1 1 1 1 1 1 1 1
PBK BRCA1 PRKDC E2F4 PCNA E2F1 CHEK1 EFHC1 ITGA7 ALG5
0.132 0.132 0.132 0.132 0.132 0.132 0.132 0.116 0.116 0.116
E2F1 E2F4 EGFR GIPC1 TIAM1 LYN PRKDC RAF1 ATP5C1 MNAT1
178 98 96 37 34 33 31 29 29 26
E2F4 E2F1 EGFR CYCS RPS26 RPL8 RPS5 RPS10 NDUFS7 ATP5J
81 67 34 28 27 26 26 26 26 25
Each gene received a score for each algorithm.
Table 3 Hub genes detected using 12 different algorithms in the network derived from the STRING database MCC
DMNC
Gene NOP56 DKC1 DDX54 NOL10 NMD3 RBM19 MPHOSPH10 POLQ C13orf34 GNL2
PCNA MYCN POLA1 RFC4 PSMD14 CDKN1A GMPS NHP2 DKC1 TCP1
EPC
Radiality
Stress
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13 9.22E+13
C13orf34 RBM19 CDCA7 KIF20B NOP14 THUMPD2 GTSE1 TTC27 KCTD19 GINS2
1.172 1.163 1.157 1.144 1.142 1.107 1.104 1.097 1.086 1.080
PCNA RFC4 NHP2 DKC1 GMPS ABCE1 NOP56 PSMD14 RSL24D1 POLA1
117 107 100 92 88 87 79 78 78 72
RFC4 PCNA NHP2 NOP56 RSL24D1 ABCE1 DKC1 PSMD14 GMPS POLR1C
304.41 303.79 301.54 301.47 301.18 301.02 300.58 299.17 297.35 295.30
PCNA MYCN POLA1 RFC4 CDKN1A KITLG PTEN PSMD14 ABL1 NPM1
8.2911 8.24066 8.19169 8.093 8.09008 8.07692 8.06522 8.05133 8.02355 8.02355
PCNA MYCN POLA1 CDKN1A PTEN RFC4 ABL1 GMPS LYN TFRC
1515402 1358060 1038576 718350 707432 700784 617368 604372 501280 492414
Closeness Gene
MNC
Betweenness
Clustering
Eccentricity
BN
Degree
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
Gene
Score
612.22 585.88 577.95 571.73 554.27 552.28 546.45 546.43 545.02 544.87
MYCN PCNA POLA1 PTEN DNAH8 CDKN1A ABL1 TFRC POMC LYN
64,829.06 56,636.91 50,847.86 37,527.92 32,123.96 29,903.31 28,208.08 24,610.5 22,725.36 22,265.08
MYL7 DNAH3 NAALADL2 CARS2 PROP1 INPP4B COL27A1 ACRV1 ADORA2B PTH2R
1 1 1 1 1 1 1 1 1 1
NAP1L1 ABCB1 NOP56 PLCB4 DKC1 LARP7 COMP HAT1 DRG1 MRPL13
0.162 0.162 0.162 0.162 0.162 0.162 0.162 0.162 0.162 0.162
POLA1 EGFR CDKN1A PIK3CG PTEN COPS5 PCNA DNAH8 POMC TFRC
74 42 38 34 29 28 26 24 24 23
PCNA RFC4 NHP2 DKC1 GMPS ABCE1 MYCN PSMD14 POLA1 NOP56
120 107 100 92 88 88 80 80 80 79
Each gene received a score for each algorithm.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
8
Table 4 Gene ontogeny analysis of the hub genes detected in the canine DLBCL GRN Term GO:0006974wresponse to DNA damage stimulus GO:0033554wcellular response to stress GO:0006281wDNA repair GO:0006259wDNA metabolic process GO:0042770wDNA damage response, signal transduction GO:0010608wposttranscriptional regulation of gene expression GO:0006396wRNA processing GO:0051329winterphase of mitotic cell cycle GO:0051325winterphase GO:0008283wcell proliferation GO:0051726wregulation of cell cycle GO:0000278wmitotic cell cycle GO:0022403wcell cycle phase GO:0022402wcell cycle process GO:0000079wregulation of cyclin-dependent protein kinase activity GO:0007049wcell cycle GO:0042127wregulation of cell proliferation GO:0045859wregulation of protein kinase activity GO:0043549wregulation of kinase activity GO:0051338wregulation of transferase activity GO:0008284wpositive regulation of cell proliferation GO:0010035wresponse to inorganic substance GO:0043066wnegative regulation of apoptosis GO:0043069wnegative regulation of programmed cell death GO:0060548wnegative regulation of cell death GO:0008285wnegative regulation of cell proliferation GO:0050730wregulation of peptidyl-tyrosine phosphorylation GO:0001775wcell activation GO:0042981wregulation of apoptosis GO:0043067wregulation of programmed cell death GO:0010941wregulation of cell death GO:0043065wpositive regulation of apoptosis GO:0043068wpositive regulation of programmed cell death GO:0010942wpositive regulation of cell death GO:0006917winduction of apoptosis GO:0012502winduction of programmed cell death
Genes
Fold enrichment
P value
MNAT1, CDKN1A, LYN, UNG, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1, RPS3 MNAT1, CDKN1A, LYN, UNG, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1, RPS3 MNAT1, UNG, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1 MNAT1, UNG, CYCS, PCNA, PRKDC, BRIP1, CHEK1, ABL1, SMC1A, BRCA1 BRIP1, CHEK1, ABL1, SMC1A, BRCA1
7.2536193
5.16E-07
4.780212
2.81E-05
7.1450704
3.00E-05
4.455863
3.17E-04
14.091667
4.03E-04
EIF5B, NDUFA13, PRKDC, GIPC1, RPS4X, RPS5, PRKCD POLR2H, POLR2F, POLR2E, RPS17, WDR77, SNRPA, SMC1A, POLR2B E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, ABL1 E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, ABL1 E2F1, EGFR, MNAT1, PCNA, RAF1, PRKCD E2F1, EGFR, MNAT1, CDKN1A, TSC2, BRIP1, CHEK1, SMC1A, BRCA1 E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A, BRCA1 EGFR, MNAT1, CDKN1A, CHEK1
7.4799368
3.03E-04
3.2975015
0.009395
15.322977
5.47E-06
14.889308
6.46E-06
3.1027523 6.1305136
0.041054 8.83E-05
5.4843243
1.90E-04
4.9014493
4.06E-04
3.9905605
7.05E-04
16.701235
0.001669
2.9054983
0.006104
2.578399
0.020238
3.2676329 3.1577965 3.0304659 2.7230274 3.2995122 1.9107345 1.8841226
0.06337 0.070007 0.07879 0.106154 0.224946 0.459516 0.466802
EGFR, MNAT1, CDKN1A CDKN1A, TSC2, CHEK1 EGFR, LYN, PRKCD
1.8788889 1.8736842 9.9470588
0.468253 0.469701 0.035445
EGFR, LYN, PRKDC, PRKCD EGFR, MNAT1, CDKN1A, TIAM1, CYCS, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 EGFR, MNAT1, CDKN1A, TIAM1, CYCS, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 EGFR, MNAT1, CDKN1A, TIAM1, CYCS, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, PRKDC, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, ABL1, BRCA1, RPS3 CDKN1A, TIAM1, NDUFA13, ABL1, BRCA1, RPS3
3.1423926 2.8043118
0.129717 0.007649
2.7766831
0.008142
2.7664622
0.008332
3.6703876
0.010832
3.6449577
0.011184
3.6281992
0.011423
4.2275 4.2143302
0.012671 0.012829 (Continued)
E2F1, EGFR, MNAT1, CDKN1A, E2F4, CHEK1, PBK, ABL1, SMC1A, BRCA1 EGFR, MNAT1, CDKN1A, E2F4, LYN, TSC2, CHEK1, RPS4X, BRCA1 EGFR, MNAT1, CDKN1A, TSC2, CHEK1 EGFR, MNAT1, CDKN1A, TSC2, CHEK1 EGFR, MNAT1, CDKN1A, TSC2, CHEK1 EGFR, MNAT1, CDKN1A, LYN, RPS4X EGFR, MNAT1, CDKN1A EGFR, MNAT1, CDKN1A EGFR, MNAT1, CDKN1A
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
9
Gene Analysis in Canine Lymphoma Table 4 (continued ) Genes
Fold enrichment
P value
EGFR, CDKN1A, LYN, PRKDC, SMC1A, BRCA1 CDKN1A, LYN, PRKDC, BRCA1 CDKN1A, LYN, PRKDC, BRCA1 E2F1, RPS26, NDUFA13
3.676087 2.4574024 2.2268313 1.8685083
0.021864 0.215105 0.259248 0.471147
E2F1, MNAT1, E2F4, NDUFA13, PRKDC, BRIP1, CREB3L4, ABL1, BRCA1 EGFR, LYN, TIAM1, TSC2, ITGA7, RAF1, GIPC1
1.1445008
0.516764
0.8503592
0.837679
EGFR, LYN, TSC2, PCNA, NDUFA13, GIPC1 LYN, TSC2, PCNA, NDUFA13, GIPC1 EGFR, TSC2, PCNA, NDUFA13, GIPC1 LYN, TSC2, PCNA, NDUFA13, GIPC1 E2F1, MNAT1, GTF3A, E2F4, SNAPC4, NDUFA13, PRKDC, BRIP1, CREB3L4, ABL1, BRCA1, RPS3 LYN, RPS4X, BRCA1
1.5337868 1.4794401 2.7429035 1.4659731 1.0402153
0.340044 0.426981 0.104069 0.433927 0.597635
2.7835391
0.286452
1.243611
0.393199
Term GO:0009628wresponse to abiotic stimulus GO:0009725wresponse to hormone stimulus GO:0009719wresponse to endogenous stimulus GO:0051253wnegative regulation of RNA metabolic process GO:0006355wregulation of transcription, DNAdependent GO:0007166wcell surface receptor-linked signal transduction GO:0008104wprotein localization GO:0015031wprotein transport GO:0034613wcellular protein localization GO:0045184westablishment of protein localization GO:0045449wregulation of transcription GO:0051247wpositive regulation of protein metabolic process GO:0051252wregulation of RNA metabolic process
E2F1, RPS26, MNAT1, E2F4, NDUFA13, PRKDC, BRIP1, CREB3L4, ABL1, BRCA1
occurs in patients with DLBCL. Inositol polyphosphate 4-phosphatase-II (INPP4B) also acts as a tumour suppressor because of its regulatory role in the PI3K signalling pathway. INPP4B knockdown resulted in AKT activation and accelerated cell proliferation. Downregulation of INPP4B in tumours was accompanied by loss of PTEN function (Fedele et al., 2010). Another key pathway in the analysis was the p53 signalling pathway. Related hub genes included E2F1, CDKN1A, PCNA, PIK3CG, PTEN and GTSE1. The role of E2F1 and E2F4 in the development of sporadic Burkitt’s lymphoma (sBL) has been investigated (Molina-Privado et al., 2009, 2012). E2F1 is highly expressed by sBL cells and inhibition of its expression suppresses tumour formation. E2F1 and C-MYC may work together in sBL. Consistent with these findings, overexpression of E2F1 was shown in this study (Supplementary Table 1). PCNA as a tumour proliferation activity marker was strongly correlated with grading score in NHL (Korkolopoulou et al., 1993, 1994; Rabenhorst et al., 1996) and PCNA, together with c-AgNORs and myc p62, can efficiently estimate proliferation activity in NHL (Korkolopoulou et al., 1993). Furthermore, significant correlations were found between P53 and PCNA expression in human patients with NHL (Korkolopoulou et al., 1994). CDKN1A (p21), another component of the p53 pathway, was shown to be downregulated in adult T-cell leukaemia/lymphoma by an epigenetic event induced by the PI3K/AKT signalling pathway (Watanabe et al., 2010). Downregulation of CDKN1A and activation of the PI3K/AKT
signalling pathway were also found in the present study. Expression and function of p21 is dependent on p53 protein, and absence of the both markers was shown in 67.7% of 253 people with NHL (Chilosi et al., 1996). Telomeres are repeated nucleotide sequences that preserve chromosomes from harmful effects. Telomerase and alternative lengthening of telomeres (ALT) are considered as important mechanisms that maintain telomere length. Telomere shortening originates from mutations in genes encoding telomerase or telomere-binding protein as well as mutations in DKC1, NOP10, TERT, TINF2 and TERC (Calado and Young, 2008). In the present study, RFC4, DKC1, POLA1 and PCNA were listed as genes involved in telomere maintenance. Dysregulation of RFC4 and 11 other genes, including IGF2BP3, UPP1, CDKN2B, CDKN2A, RMI1, SLBP, AUTS2, FUT8, ACY1, BCL2 and PDXK, has been detected in human DLBCL (Jardin et al., 2010). DKC1 and AKT2 expression levels were also found to be downregulated by miR-150, which functioned as a tumour suppressor. Additionally, miR-150 exerted its effects by amplifying the protein levels of Bim and p53. In this context, downregulation of miR-150 and subsequent activation of the PI3K/AKT pathway, as well as telomerase, was confirmed in malignant lymphoma (Watanabe et al., 2011). The role of POLA1 in telomere maintenance has also been investigated (Fujita et al., 2013). The results of the present study implicated E2F1, CDKN1A, ABL1 and RAF1 as gene components incorporated in two cell cycle pathways, including G1 and Rac CycD pathways. The G1/S cell cycle
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
M. Zamani-Ahmadmahmudi et al.
10
Table 5 Pathway analysis of the hub genes detected in the canine DLBCL GRN Genes
Fold enrichment
P value
E2F1, CDKN1A, ABL1 E2F1, CDKN1A, PCNA E2F1, CDKN1A, RAF1
7.823956 13.34675 9.453947
0.049278 0.018038 0.034759
LYN, TIAM1, RAF1, PRKCD PIK3CG, PLCB4, INPP4B, PTEN E2F1, CDKN1A, E2F4, PCNA, PRKDC, CHEK1, ABL1, SMC1A EGFR, PIK3CG, MYL7, COMP, PTEN LYN, RAF1, PRKCD EGFR, PIK3CG, CDKN1A, KITLG, ABL1, PTEN E2F1, EGFR, RAF1 EGFR, PIK3CG, PTEN PIK3CG, CDKN1A, PTEN PIK3CG, CDKN1A, PTEN PIK3CG, CDKN1A, PTEN E2F1, EGFR, CDKN1A, RAF1 PIK3CG, CDKN1A, ABL1 E2F1, EGFR, RAF1 LYN, RAF1, PRKCD POLR2H, POLR2F, POLR2E PIK3CG, PLCB4, PTEN
2.219797 10.18018 6.641633
0.258506 0.006049 1.46E-04
4.684909 3.991366 3.445122 4.32398 10.86538 11.95767 8.464419 10.61033 9.883382 7.533333 5.765306 5.307121 11.87119 1.566338
0.017892 0.167306 0.023173 0.147411 0.028505 0.003846 0.010074 0.005388 0.007002 0.055547 0.09163 0.098415 0.023052 0.553964
POLR2H, POLR2F, POLR2E, CREB3L4
11.34856
0.004041
PIK3CG, CDKN1A, PTEN, GTSE1 RFC4, DKC1, POLA1, PCNA
5.290741 12.13571
0.031942 0.003408
Term h_g1 pathway: cell cycle: G1/S check point* h_p53pathway: p53 signalling pathway* h_Rac CycD pathway: influence of Ras and Rho proteins on G1 to S transition* hsa04062: chemokine signalling pathway† hsa04070: phosphatidylinositol signalling system hsa04110: cell cycle hsa04510: focal adhesion hsa04664: Fc epsilon RI signalling pathway hsa05200: pathways in cancer hsa05212: pancreatic cancer hsa05213: endometrial cancer hsa05214: glioma hsa05215: prostate cancer hsa05218: melanoma hsa05219: bladder cancer hsa05220: chronic myeloid leukaemia (CML) hsa05223: non-small cell lung cancer (NSCLC) P00010: B-cell activation‡ P00023: general transcription regulation‡ P00031: inflammation mediated by chemokine and cytokine signalling pathway‡ P00055: transcription regulation by b ZIP transcription factor‡ P00059: p53 pathway‡ REACT_7970: telomere maintenancex *
BIOCARTA databases. hsa identifier belonging to KEGG database. ‡ PANTHER database. x REACTOME database. †
checkpoint controls transition from the G1 phase into the DNA synthesis S phase. Different stimulatory or inhibitory factors regulate this complex transition. For example, CDK2-Cyclin E and CDK4/6-Cyclin D accelerate the G1-S transition, while p21, p27 and p57 reduce the transition through inhibition of cyclineCDK complexes (Nojima, 2004). Previous studies have shown that canine DLBCL rarely expresses BCL6 (Richards et al., 2013) and this was supported by the findings of the present study, which did not detect BCL6 as a hub gene. To the best of our knowledge, this has been the first study to reconstruct and explore canine lymphoma GRNs. However, the study suffered from a major limitation. Although microarray data from human cancers are very extensive and informative, microarray data related to animal cancers are rare and incomplete. There are only two canine lymphoma datasets, with sample sizes of 33 and 58 patients. Analysis of datasets with larger sample numbers yields more realistic and accurate networks and subsequently detects more subtle hub genes.
Acknowledgment This project was supported by grant No. 15674 from the Shahid Bahonar University of Kerman.
Conflict of Interest Statement The authors report no conflict of interest.
Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jcpa.2014.11. 008.
References Agnelli L, Forcato M, Ferrari F, Tuana G, Todoerti K et al. (2011) The reconstruction of transcriptional networks reveals critical genes with implications for clinical outcome of multiple myeloma. Clinical Cancer Research: An Official Journal of the American Association for Cancer Research, 17, 7402e7412.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
Gene Analysis in Canine Lymphoma
Assenov Y, Ramırez F, Schelhorn S-E, Lengauer T, Albrecht M (2008) Computing topological parameters of biological networks. Bioinformatics, 24, 282e284. Bansal M, Belcastro V, Ambesi-Impiombato A, di Bernardo D (2007) How to infer gene networks from expression profiles. Molecular Systems Biology, 3, 78. Basso K, Margolin AA, Stolovitzky G, Klein U, DallaFavera R et al. (2005) Reverse engineering of regulatory networks in human B cells. Nature Genetics, 37, 382e390. Basso K, Saito M, Sumazin P, Margolin AA, Wang K et al. (2010) Integrated biochemical and computational approach identifies BCL6 direct target genes controlling multiple pathways in normal germinal center B cells. Blood, 115, 975e984. Blais A, Dynlacht BD (2005) Constructing transcriptional regulatory networks. Genes and Development, 19, 1499e1511. Butler MP, Wang SI, Chaganti RS, Parsons R, DallaFavera R (1999) Analysis of PTEN mutations and deletions in B-cell non-Hodgkin’s lymphomas. Genes Chromosomes and Cancer, 24, 322e327. Calado RT, Young NS (2008) Telomere maintenance and human bone marrow failure. Blood, 111, 4446e4455. Chilosi M, Doglioni C, Magalini A, Inghirami G, Krampera M et al. (1996) p21/WAF1 cyclin-kinase inhibitor expression in non-Hodgkin’s lymphomas: a potential marker of p53 tumor-suppressor gene function. Blood, 88, 4012e4020. Cline MS, Smoot M, Cerami E, Kuchinsky A, Landys N et al. (2007) Integration of biological networks and gene expression data using Cytoscape. Nature Protocols, 2, 2366e2382. Croce CM (2008) Oncogenes and cancer. New England Journal of Medicine, 358, 502e511. Emmert-Streib F, de Matos Simoes R, Mullan P, HaibeKains B, Dehmer M (2014) The gene regulatory network for breast cancer: integrated regulatory landscape of cancer hallmarks. Frontiers in Genetics, 5, 15. Engelman JA, Luo J, Cantley LC (2006) The evolution of phosphatidylinositol 3-kinases as regulators of growth and metabolism. Nature Reviews Genetics, 7, 606e619. Fedele CG, Ooms LM, Ho M, Vieusseux J, O’Toole SA et al. (2010) Inositol polyphosphate 4-phosphatase II regulates PI3K/AKT signaling and is lost in human basal-like breast cancers. Proceedings of the National Academy of Sciences of the United States of America, 107, 22231e22236. Floratos A, Smith K, Ji Z, Watkinson J, Califano A (2010) geWorkbench: an open source platform for integrative genomics. Bioinformatics, 26, 1779e1780. Franceschini A, Szklarczyk D, Frankild S, Kuhn M, Simonovic M et al. (2013) STRING v9.1: proteinprotein interaction networks, with increased coverage and integration. Nucleic Acids Research, 41, D808eD815. Fujita T, Asano Y, Ohtsuka J, Takada Y, Saito K et al. (2013) Identification of telomere-associated molecules by engineered DNA-binding molecule-mediated chro-
11
matin immunoprecipitation (enChIP). Scientific Reports, 3, 3171. Gautier L, Cope L, Bolstad BM, Irizarry RA (2004) Affy e analysis of Affymetrix GeneChip data at the probe level. Bioinformatics, 20, 307e315. Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell, 100, 57e70. Huang DW, Sherman BT, Lempicki RA (2009a) Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Research, 37, 1e13. Huang DW, Sherman BT, Lempicki RA (2009b) Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nature Protocols, 4, 44e57. Jardin F, Jais J-P, Molina T-J, Parmentier F, Picquenot JM et al. (2010) Diffuse large B-cell lymphomas with CDKN2A deletion have a distinct gene expression signature and a poor prognosis under R-CHOP treatment: a GELA study. Blood, 116, 1092e1104. Korkolopoulou P, Oates J, Kittas C, Crocker J (1994) p53, c-myc p62 and proliferating cell nuclear antigen (PCNA) expression in non-Hodgkin’s lymphomas. Journal of Clinical Pathology, 47, 9e14. Korkolopoulou P, Patsouris E, Pangalis G, Tsenga A, Elemenoglou J et al. (1993) A comparative assessment of proliferating cell nuclear antigen, c-myc p62, and nucleolar organizer region staining in non-Hodgkin’s lymphomas: a histochemical and immunohistochemical study of 200 cases. Human Pathology, 24, 371e377. Liu X, Karnell JL, Yin B, Zhang R, Zhang J et al. (2010) Distinct roles for PTEN in prevention of T-cell lymphoma and autoimmunity in mice. Journal of Clinical Investigation, 120, 2497e2507. MacEwen EG (1990) Spontaneous tumors in dogs and cats: models for the study of cancer biology and treatment. Cancer Metastasis Reviews, 9, 125e136. Maere S, Heymans K, Kuiper M (2005) BiNGO: a Cytoscape plugin to assess overrepresentation of gene ontology categories in biological networks. Bioinformatics, 21, 3448e3449. McCaw DL, Chan AS, Stegner AL, Mooney B, Bryan JN et al. (2007) Proteomics of canine lymphoma identifies potential cancer-specific protein markers. Clinical Cancer Research, 13, 2496e2503. Molina-Privado I, Jimenez-P R, Montes-Moreno S, Chiodo Y, Rodrıguez-Martınez M et al. (2012) E2F4 plays a key role in Burkitt lymphoma tumorigenesis. Leukemia, 26, 2277e2285. Molina-Privado I, Rodrıguez-Martınez M, Rebollo P, Martın-Perez D, Artiga M-J et al. (2009) E2F1 expression is deregulated and plays an oncogenic role in sporadic Burkitt’s lymphoma. Cancer Research, 69, 4052e4058. Mudaliar MAV, Haggart RD, Miele G, Sellar G, Tan KAL et al. (2013) Comparative gene expression profiling identifies common molecular signatures of NF-kB activation in canine and human diffuse large B cell lymphoma (DLBCL). PLoS One, 8, e72591.
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008
12
M. Zamani-Ahmadmahmudi et al.
Nojima H (2004) G1 and S-phase checkpoints, chromosome instability, and cancer. Methods in Molecular Biology, 280, 3e49. Ponce F, Marchal T, Magnol JP, Turinelli V, Ledieu D et al. (2010) A morphological study of 608 cases of canine malignant lymphoma in France with a focus on comparative similarities between canine and human lymphoma morphology. Veterinary Pathology, 47, 414e433. Qiao Q, Jiang Y, Li G (2013) Inhibition of the PI3K/AKTNF-kB pathway with curcum in enhanced radiationinduced apoptosis in human Burkitt’s lymphoma. Journal of Pharmacological Sciences, 121, 247e256. Rabenhorst SH, Burini RC, Schmitt FC (1996) Proliferating cell nuclear antigen (PCNA) in non-Hodgkin’s lymphomas: correlation with working formulation and Kiel classification in formalin-fixed paraffin-embedded material. Pathology, 28, 12e16. Richards KL, Motsinger-Reif AA, Chen H-W, Fedoriw Y, Fan C et al. (2013) Gene profiling of canine B-cell lymphoma reveals germinal center and postgerminal center subtypes with different survival times, modeling human DLBCL. Cancer Research, 73, 5029e5039. Sakai A, Thieblemont C, Wellmann A, Jaffe ES, Raffeld M (1998) PTEN gene alterations in lymphoid neoplasms. Blood, 92, 3410e3415. Tarcea VG, Weymouth T, Ade A, Bookvich A, Gao J et al. (2009) Michigan molecular interactions r2: from interacting proteins to pathways. Nucleic Acids Research, 37, D642eD646.
Vail DM, MacEwen EG (2000) Spontaneously occurring tumors of companion animals as models for human cancer. Cancer Investigation, 18, 781e792. Watanabe A, Tagawa H, Yamashita J, Teshima K, Nara M et al. (2011) The role of microRNA-150 as a tumor suppressor in malignant lymphoma. Leukemia, 25, 1324e1334. Watanabe M, Nakahata S, Hamasaki M, Saito Y, Kawano Y et al. (2010) Downregulation of CDKN1A in adult T-cell leukemia/lymphoma despite overexpression of CDKN1A in human T-lymphotropic virus 1-infected cell lines. Journal of Virology, 84, 6966e6977. Xu Z-Z, Xia Z-G, Wang A-H, Wang W-F, Liu Z-Y et al. (2013) Activation of the PI3K/AKT/mTOR pathway in diffuse large B-cell lymphoma: clinical significance and inhibitory effect of rituximab. Annals of Hematology, 92, 1351e1358. Zhang J, Grubor V, Love CL, Banerjee A, Richards KL et al. (2013) Genetic heterogeneity of diffuse large Bcell lymphoma. Proceedings of the National Academy of Sciences of the United States of America, 110, 1398e1403.
July 10th, 2014 ½ Received, Accepted, November 29th, 2014
Please cite this article in press as: Zamani-Ahmadmahmudi M, et al., Reconstruction of Canine Diffuse Large B-cell Lymphoma Gene Regulatory Network: Detection of Functional Modules and Hub Genes, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.11.008