A systems genetics approach delineates the role of Bcl2 in leukemia pathogenesis

Abstract

Leukemia is a group of malignancies of the blood forming tissues, and is characterized by the uncontrolled proliferation of blood cells. In the United States, it accounts for approximately 3.5% and 4% of all cancer-related incidences and mortalities, respectively. The current study aimed to explore the role of Bcl2 and associated genes in leukemia pathogenesis using a systems genetics approach. The transcriptome data from BXD Recombinant Inbred (RI) mice was analyzed to identify the expression of Bcl2 in myeloid cells. eQTL mapping was performed to select the potential chromosomal region and subsequently identify the candidate gene modulating the expression of Bcl2. Furthermore, gene enrichment and protein-protein interaction (PPI) analyses of the Bcl2-coexpressed genes were performed to demonstrate the role of Bcl2 in leukemia pathogenesis. The Bcl2-coexpressed genes were found to be enriched in various hematopoietic system related functions, and multiple pathways related to signaling, immune response, and cancer. The PPI network analysis demonstrated direct interaction of hematopoietic function related genes, such as Bag3, Bak1, Bcl2l11, Bmf, Mapk9, Myc, Ppp2r5c, and Ppp3ca with Bcl2. The eQTL mapping identified a 4.5 Mb genomic region on chromosome 11, potentially regulating the expression of Bcl2. A multi-criteria filtering process identified Top2a, among the genes located in the mapped locus, as the best candidate upstream regulator for Bcl2 expression variation. Hence, the current study provides better insights into the role of Bcl2 in leukemia pathogenesis and demonstrates the significance of our approach in gaining new knowledge on leukemia. Furthermore, our findings from the PPI network analysis and eQTL mapping provide supporting evidence of leukemia-associated genes, which can be further explored for their functional importance in leukemia.

1. Introduction

Leukemia is a collection of hematological malignancies characterized by the uncontrolled proliferation and development of leukocytes. In 2018, leukemia was the 15th most commonly detected cancer and 11th leading cause of cancer-related deaths, worldwide [1]. In the United States, leukemia is estimated to account for approximately 3.5% and 4% of all cancer-related incidences and mortalities, respectively. Since 2006, the incidence rate of leukemia has increased annually by an average of 0.6%, whereas mortality rate has decreased by an average of 1.5% per year [1,2]. Over the last few decades, significant progress in understanding the normal physiology of hematopoietic cells and their malignant transformation has been witnessed. Germline mutations and chromosomal abnormalities, such as rearrangements, translocations, and deletions are known to be the most common causes of both acute and chronic subtypes of leukemia. The Philadelphia chromosome translocation (t(9;22)), a well-known phenomenon in chronic myelogenous leukemia (CML) and a subset of acute myelogenous leukemia (AML) conditions results in a chimeric Bcr-Abl protein that has increased tyrosine phosphokinase activity. The elevated tyrosine kinase signaling in turn results in uncontrolled proliferation of leukocytes [3]. Furthermore, constitutive activation of Abl kinase, mediated by Bcr-Abl fusion, promotes the activation of Ras signaling pathway, which enhances the expression of downstream Bcl2 gene [4,5]. Similarly, t(15;17) has been identified as an important translocation in acute leukemic cells. It leads to the fusion of retinoic acid receptor alpha (RARA) and a promyelocytic leukemia transcription factor (PML), resulting in the development of acute promyelocytic leukemia [6]. Transgenic mice coexpressing MRP8-BCL2 and PML-RARA have been shown to exhibit a marked increase in the accumulation of immature myeloid cells [7].

The members of B-cell leukemia/lymphoma-2 (BCL-2) family of proteins are the key regulators of the intrinsic apoptotic pathway. The BCL-2 family consists of more than 20 proteins, which function as either pro-apoptotic or anti-apoptotic molecules [8]. The founding member of the family, BCL-2 is an anti-apoptotic protein, and contributes to tumorigenesis by blocking mitochondrial-mediated apoptosis and promoting cell survival [8]. Multiple studies have explored the importance of BCL-2 in different subtypes of leukemia and have implicated it as a therapeutic target [9,10]. A number of studies have shown that the overexpression of BCL-2 in AML is associated with poor prognosis, and resistance to chemotherapy and radiation therapy [11,12]. Furthermore, a better understanding of the role of BCL-2 in apoptotic pathway has led to the development of antisense oligonucleotides [13] and novel small molecules [14], which target BCL-2 and induce apoptosis both in human cell lines and murine xenograft models. Using publicly available and their data, Zhou et al. [15] recently confirmed the up-regulation of BCL-2 in AML. Further, the authors identified a correlation between BCL-2 expression and different French-American-British (FAB) classification based subtypes of AML. The therapeutic agent Venetoclax (ABT-199, GDC-0199), is a highly selective BCL-2 inhibitor and is currently widely used for the treatment of AML [10]. Despite its effectiveness, prolonged use of Venetoclax leads to drug resistance or loss of dependence on the targeted protein [10]. Hence, there is a need to better understand the mechanisms underlying the pathogenesis of leukemia and to develop better therapeutic targets.

As one of the largest and best characterized genetic reference population, BXD recombinant inbred (RI) mice that are derived from crosses between C57BL/6J (B6) and DBA/2J (D2) inbred strains have been widely used for studying the genetic basis of various disorders, including cancer [16]. The BXD RI set is comprised of ~150 fully sequenced inbred strains and segregates ~6 million sequence variants scattered across the genome, similar to the number of single nucleotide polymorphisms segregating in the human population [16]. Thus, the BXD RI mice are suitable for defining gene variants linked to various complex genetic diseases. In particular, with the availability of the large-scale omics data (i.e., transcriptomic, proteomic, and metabolomic) from various tissues and conditions, the BXD mice provide a powerful genetic resource to decipher the pathways and regulatory networks using a systems genetics approach [17,18]. Henckaerts et al. [19] have identified several genetic regulators associated with the proliferative capacity of hematopoietic stem cells (HSCs) and progenitor cells using a panel of BXD strains. Bystrykh et al. [20] performed mRNA profiling of the HSCs isolated from BXD RI strains, and identified polymorphic cis-acting stem cell genes, and multiple trans-acting control loci that modulate the expression of a large number of candidate genes involved in HSC turnover.

In this study, we aimed to explore the role of Bcl2 and its key co-regulators in the pathogenesis of leukemia and to delineate the underlying molecular mechanisms using myeloid cell transcriptome data from BXD RI mice. In addition, using a systems genetics approach, we sought to identify Bcl2-correlated genes, associated pathways, and candidate up- or downstream regulators of Bcl2 that may contribute to the pathogenesis of leukemia.

2. Material and methods

2.1. Myeloid cell transcriptomic data

The transcriptomic data corresponding to myeloid cells of BXD RI mice used in the current study was generated by our colleague Dr. Gerald de Haan at the University Medical Center, and is publicly accessible through our GeneNetwork website (http://www.genenetwork.org/). Briefly, RNA isolated from the myeloid cells of 24 BXD strains were profiled using the Illumina Mouse WG-6 v1, v1.1 (GPL6105) gene expression microarray. The raw data was first normalized using the robust-multichip array (RMA) method [21] and then rescaled to a mean of 8 units with a standard deviation of 2 units (2Z + 8 normalization). The normalized data can be accessed through GeneNetwork with the identifier “MCG Myeloid Cells ILM6v1.1 (Apr09) transformed” under the group name “BXD Family” and type “Hematopoietic Cells mRNA”.

2.2. eQTL mapping for Bcl2 expression variation

The traditional QTL mapping method, “interval mapping” was used to identify the genomic loci that regulate molecular traits, also referred as eQTL mapping. Briefly, a total of ~7000 representative genetic markers across the BXD mouse genome were used for association with the Bcl2 mRNA expression variation. This association was evaluated with the likelihood ratio statistic (LRS), which is defined as the ratio of the probability that a molecular trait is associated with the particular genetic marker to the probability that it is not associated. In addition, 1000 permutation tests were used to determine the genome-wide suggestive and significance threshold. All the genetic analyses were performed using the webQTL function on GeneNetwork website.

2.3. Candidate gene identification in the eQTL region

Typically, an eQTL region may contain dozens or even hundreds of genes. Therefore, we applied the following criteria to prioritize the candidates: (1) Presence of the genetic variations between BXD parental strains, B6 and D2. This analysis was based on our previous whole genome re-sequencing dataset for these parental strains [1]. Specifically, gene coding region variants, such as stop gain, stop loss, frameshift, and mis-sense variants were considered. (2) cis-eQTL identification. This analysis was performed according to the method described above for all the genes in the eQTL region. The cis-eQTL is defined as the regulating loci located within or near the gene’s physical position. (3) Co-expression with Bcl2 expression. Pearson correlation coefficient was used to evaluate the co-expression of the candidate genes with Bcl2 (p-value < 0.05). (4) Biological functions related to leukemia. The biological functions associated with leukemia related phenotypes were explored for each candidate gene using Mouse Genome Informatics (http://www.informatics.jax.org/) and Rat Genome Informatics (https://rgd.mcw.edu/rgdweb/homepage/) databases. Both of these databases contain gene-phenotype/disease association portals allowing us to identify the known and potential mouse models of human diseases.

2.4. Gene co-expression analysis

Co-expressed genes (genes with similar expression patterns) are known to be involved in similar molecular functions or are part of the same biological pathways; hence they can be used to explore the functions of a gene of interest. We used Pearson correlation coefficient to identify Bcl2 co-expressed genes in the myeloid cell transcriptome data of BXD mice. Genes with a p-value < 0.05 were considered to be significantly co-expressed with Bcl2. We further filtered the Bcl2 co-expressed genes that have association with findings in the literature by examining the r values of the genes that are described by similar terminologies in MEDLINE/PubMed abstracts [22]. Genes with r > 0.3 were considered to have a high literature correlation and were used for further analysis.

2.5. Gene set enrichment analysis

Gene set enrichment analysis is used for determining whether a set of genes (Bcl2 co-expressed genes in this study) are overrepresented in certain annotations, such as Gene Ontology (GO), KEGG pathways, Mammalian Phenotype Ontology (MPO), or customized gene sets. We performed the enrichment analysis of Bcl2 co-expressed genes using WEB-based Gene SeT AnaLysis Toolkit (WebGestalt, http://www.web-gestalt.org). This analysis uses hypergeometric statistical test to result adjusted p values and enrichment ratios. The annotations with a minimum overlap of 5 genes and false discovery rate (FDR) < 0.05 (Benjamini and Hochberg correction) were considered statistically significant.

2.6. Protein-protein interaction (PPI) network construction

The PPI network of Bcl2 co-expressed genes was constructed using the STRING database (www.string-db.org). The interactions in STRING are based on multiple evidences, such as text mining, experiments, co-expression, neighborhood, gene fusion, and co-occurrence. A minimum score of 0.4 (medium confidence) was considered as a threshold for filtering the interactions. The interacting proteins were further analyzed and visualized using Cytoscape [23].

3. Results

3.1. Bcl2 mRNA expression varies extensively in BXD myeloid cells

The transcriptome profiling of myeloid cells from 24 BXD strains was performed using the Illumina Mouse WG-6 v1, v1.1 array. The microarray platform has a total of eight probes targeting the Bcl2 gene body, with six located at the 3′ UTR and two at the 5′ UTR region (Table 1). The probe, ILM5860504 had the lowest mean expression value of 6.56, while that of ILM4810037 had the highest mean expression value of 10.35 (Table 1). The median value for the fold range for Bcl2 expression across 24 BXD strains was ~2.065 (Table 1). To simplify the subsequent analysis, we performed principal component analysis using the expression values of eight Bcl2-targeting probes and used the first principal component to represent the overall expression of Bcl2 in BXD mice, as it explains ~44% of the total variance (Fig. 1A). Furthermore, as shown in Fig. 1B, BXD11 and BXD21 mice showed the lowest and highest expression values for Bcl2, respectively.

Table 1.

Probes on illumina mouse WG-6 v1 array targeting Bcl2 gene.

Probe ID	Location (Chr, Mb)	Description	Mean	Median	Minimum	Maximum	Range (fold)
ILM3800044	Chr1: 106.538347	3′ UTR	8.23	8.23	7.58	8.73	2.21
ILM5690068	Chr1: 106.538852	3′ UTR	7.83	7.88	7.16	8.41	2.39
ILM2470138	Chr1: 106.538901	3′ UTR	7.47	7.50	6.94	8.01	2.10
ILM5860504	Chr1: 106.543161	3′ UTR	6.56	6.56	6.35	6.81	1.38
ILM1570736	Chr1: 106.543200	3′ UTR	7.01	7.00	6.79	7.29	1.41
ILM4810037	Chr1: 106.712234	3′ UTR	10.35	10.40	9.39	11.06	3.19
ILM6650164	Chr1: 106.712978	5′ UTR	7.12	7.14	6.65	7.66	2.03
ILM730132	Chr1: 106.712994	5′ UTR	7.51	7.49	7.07	7.91	1.79

Fig. 1.

Bcl2 expression across BXD mice. Principal component analysis was performed using the expression values of eight Bcl2-targeting probes. (A) Percentages of total variance explained by each principal component. (B) Distribution of the first principal component across 24 BXD strains.

3.2. eQTL mapping identifies a Bcl2 regulating genetic locus on chromosome 11

To identify genetic loci that regulate Bcl2 mRNA expression, we performed eQTL mapping using the first principal component values. With the genome-wide suggestive and significant threshold of 10.81 and 18.3, respectively, as determined by 1000 permutation tests, a significant eQTL was mapped on chromosome 11 with a peak LRS of 24.08 (Fig. 2). The identified eQTL encompasses a 4.5 Mb genomic region from 97 to 101.5 Mb. Since the physical location of Bcl2 is at 106.5 Mb on chromosome 1, we defined the locus as a trans-acting eQTL, suggesting that Bcl2 expression variation in BXD mice is modulated by upstream regulatory genes rather than by alterations in its own sequence.

Fig. 2.

eQTL mapping of Bcl2 in BXD-mice myeloid cells. The x-axis denotes the chromosomal positions in megabases on the mouse genome and y-axis indicates the LRS score. eQTL mapping was performed with “interval mapping” method on GeneNetwork portal (https://www.genenetwork.org/). The genome-wide suggestive and significant thresholds were determined with 1000 permutation tests.

3.3. Top2a is a candidate upstream regulator of Bcl2 expression

The eQTL region mapped on chromosome 11 included ~200 potential candidate genes. To further prioritize the candidate genes regulating Bcl2 expression, we employed different filtering strategies. First, based on the sequence variants between the BXD parental strains, B6 and D2, we shortlisted 25 genes that harbored nonsynonymous SNPs, which are segregated in the BXD family. eQTL mapping performed for thê200 potential candidates, identified 5 genes showing cis-regulation with LRS > 15. Further, we identified 42 genes that were significantly correlated (p < 0.05) with Bcl2 expression. Lastly, 9 genes among the candidates were identified to be associated with leukemia-related phenotypes based on Mouse Genome Informatics and Rat Genome Informatics databases. Fig. 3A shows the genes that are shared across and exclusive to these four filtering criteria. There were four genes that were common to any of the three filtering criteria. Among these, three genes (Cdk12, Med1, and Wnk4) were significantly correlated with Bcl2, harbored nonsynonymous variants, and exhibited cis-regulation (Fig. 3A), while Top2a was significantly correlated with Bcl2 (Fig. 3B), exhibited cis-regulation (Fig. 3C), and more importantly, has been implicated in leukemia related phenotypes. Hence, because of the association of Top2a with biological functions related to leukemia, it was considered as the best candidate regulating Bcl2 expression among the potential candidate genes identified within the mapped locus.

Fig. 3.

Identification of candidate upstream regulator for Bcl2 expression. (A) Venn diagram showing the number of overlapped genes across the four categories—mis-sense variants between B6 and D2, cis-eQTL, correlation with Bcl2, and biological functions associated with leukemia related phenotypes. (B) Dot plot showing the Pearson correlation (r = −0.578, p = 0.003) between Bcl2 and Top2a. (C) Interval mapping showing the genome-wide Top2a regulating loci in BXD-mice myeloid cells. The x-axis denotes the chromosomal position in megabases on the mouse genome and y-axis indicates the LRS score. The genome-wide suggestive and significant thresholds were determined with 1000 permutation tests.

3.4. Bcl2 co-expressed genes are involved in hematopoietic system regulation and cancer related pathways

A total of 819 genes had significant genetic (p < 0.05) and literature correlation (r > 0.3) with Bcl2. To understand the functional role of Bcl2 in leukemia, we performed gene set enrichment analysis using Bcl2-coexpressed genes. Our results demonstrated that the Bcl2 co-expressed genes are mainly involved in modulating the functions and pathways associated with hematopoietic system (Fig. 4). Hence, deregulation of Bcl2 and its co-variants may result in abnormal hematopoietic system physiology [24] and immune cell physiology [25,26]. Of the top sixty MPOs with an FDR < 1E-09 (Supplementary Table 1), sixteen were found to be directly associated with hematopoietic system related functions (Fig. 4A). Furthermore, the enrichment analysis resulted in a total of 117 significant KEGG pathways with an FDR < 0.05 (Supplementary Table 2). Among these, more than 90% pathways were found to be directly or indirectly involved in tumorigenesis. Designations for these pathways including ‘pathways in cancer’, ‘proteoglycans in cancer’, ‘central carbon metabolism in cancer’, ‘transcriptional mis-regulation in cancer’, and ‘choline metabolism in cancer’ are shown in Fig. 4B. Additionally, a number of signaling pathways, such as, TNF, TGFβ, B cell receptor, IL-17, and p53 signaling were found to be significantly enriched by the Bcl2-coexpressed genes. It is worth noting that both chronic myeloid leukemia (FDR <4.20E-06) and acute myeloid leukemia (FDR <7.34E-05) pathways were significantly enriched by Bcl2-coexpressed genes, which strongly suggests that Bcl2 is a leukemia causal gene and it may be involved in leukemia pathogenesis through the co-expressed genes.

Fig. 4.

Bubble charts of the MPOs (A) and KEGG pathways (B) significantly overrepresented by Bcl2 co-expressed genes. The x-axis represents the enrichment ratio and y-axis represents the ontologies/pathways. The size of the dots represents the number of genes and the color indicates the FDR value.

3.5. PPI network analysis revealed direct interaction between Bcl2 and genes with hematopoietic system related functions

To further understand the functional role of Bcl2 in leukemia, we constructed PPI network using Bcl2-coexpressed genes. Fig. 5 shows the PPI network of 27 genes that directly interact with Bcl2. Among these, 17 genes, including Bag3, Bak1, Bcl2l11, Bmf, Bnip3, Casp2, Casp8, Cd34, Ctsc, Ddit3, Gimap5, Mapk1, Mapk8, Mapk9, Myc, Ppp2r5c, and Ppp3ca have been implicated in hematopoietic system related functions. Of these, Casp8, Mapk8, and Bcl2l11 interacted strongly with Bcl2 (with confidence score of 0.8) and additionally interacted with at least 10 other proteins. Thus, these proteins may have important role in regulating the Bcl2-related functions.

Fig. 5.

Protein-protein interaction network of Blc2 co-expressed genes. A total of 819 Bcl2 co-expressed genes were submitted to STRING database (www.string-db.org). The network shown here includes 27 genes that directly interact with Bcl2. Node size or node color (red) indicates increasing degree, and edge thickness indicates increasing interaction score.

4. Discussion

In biological systems, genes usually, but not independently, modulate the phenotypic variance through interactions with other genes or signaling pathways. Therefore, uncovering the genetic regulatory networks for the gene of interest is of vital importance to explain its biological function. Systems genetics approach is a powerful tool to leverage this gap by integrating the large-scale omics data and various statistical methods [27]. In the current study, we used a systems genetics approach to identify the possible mechanisms underlying leukemia pathogenesis.

In humans, the role of BCL2 in leukemia pathogenesis is well known. The anti-apoptotic protein has been shown to cause leukemia by suppressing mitochondrial-mediated apoptosis and promoting cell survival [8]. Furthermore, there are reports demonstrating the association of BCL2 with chemotherapy resistance [11]. Owing to its expression and active role in tumorigenesis, BCL2 has also been implicated as a therapeutic target in humans [15]. It has been long recognized that biological processes are regulated through a complex network of proteins and other macromolecular interactions rather than a single protein acting independently [28]. Hence, to better understand the role of Bcl2 in the pathogenesis of leukemia, we identified genes that are significantly correlated with Bcl2 expression and studied their possible involvement in leukemia through gene set enrichment and PPI network analyses. The enrichment analysis clearly highlighted the association of co-expressed genes with multiple functions related to hematopoietic system physiology and various cancer related pathways. These genes were found to be significantly enriched in a number of relevant MPOs, such as ‘abnormal blood cell physiology’, ‘abnormal leukocyte physiology’, ‘abnormal hemopoiesis’, and ‘abnormal T cell differentiation’. Multiple studies have established the importance of Bcl2 in maintaining hematopoiesis and immune cell differentiation. For instance, suppression of the intrinsic apoptotic pathway by Bcl2-overexpression has been shown to cause accumulation of murine HSCs in the aorta-gonad-mesonephros and fetal liver [29]. Furthermore, higher Bcl2 expression has been observed in differentiated T-cells rather than in naïve cells [30], while increased Bcl2 has been reported to enhance T-cell survival [31]. The results from the current analysis suggest that Bcl2 modulates the hematopoietic and blood cell physiology related functions through the co-expressed genes.

Furthermore, pathway enrichment analysis revealed multiple inflammatory and signaling pathways related to cancer, including TNF, NF-κB, p53, T cell receptor, IL17, and TGFβ signaling to be overrepresented by the co-expressed genes. The relationship between inflammation and cancer is well known, and the NF-κB pathway is recognized as a major bridge between these two pathological conditions. High concentrations of TNF-α and its receptors (TNFR) have been identified in chronic lymphocytic leukemia (CLL) patients [32]. Additionally, TNF-α has been shown to act as an autocrine tumor growth factor in CLL. A recent study by Durr et al. [33] demonstrated that stimulation of TNF receptor-1 with TNF-α enhanced NFκB activity and viability of CLL cells. Another study [34] analyzed myeloid leukemia mouse models and showed that AML leukemia-initiating cells exhibit constitutive NF-κB activity, which was maintained through autocrine TNF-α secretion, forming an NF-κB/TNF-α positive feedback loop. The role of NF-κB and TNF signaling is not only important in leukemia but also in other types of blood cancers. A report on a rare type of blood cancer, myelofibrosis, in which the bone marrow is replaced with fibrous scar tissue, stresses that the autocrine TNF signaling favors malignant cells over their normal counterparts in a TNFR2 dependent manner [35]. Thus, our analysis suggests that Bcl2 may contribute to leukemia pathogenesis by modulating NF-κB/TNF-α signaling via the co-expressed genes. The TP53 gene is mutated in multiple cancers, including in leukemia [36]. However, in a large number of AML cases, TP53 is unaltered, suggesting a different mechanism underlying the alteration of the p53 pathway [37]. Our results showed 20 Bcl2-coexpressed genes to be enriched in p53-signaling pathway, suggesting genome/transcriptome level changes in these genes could be one possible mechanism underlying the alteration of p53 pathway in AML. Among the inflammatory molecules, the contribution of IL17 in cancer has been emphasized by a growing body of evidence [38]. The IL-17 signaling promotes chemokine and cytokine secretion, significantly altering the tumor microenvironment, which in turn results in tumor progression. The pathway analysis results from the current study showed the involvement of 19 Bcl2-coexpressed genes to be enriched in IL-17 signaling. Thus, Bcl2 may be involved in modulating IL-17 signaling and promoting the pathogenesis of leukemia via the co-expressed genes. TGFβ signaling pathway is an important regulator of proliferation and differentiation of various hematopoietic cell types, and has been implicated in a variety of bone marrow related disorders, including leukemia [39]. A few of the SMAD proteins that are known as receptor-regulated Smads (R-Smads) including SMAD1 are directly phosphorylated and activated by TGFBR1. Activated SMAD1 forms a complex with SMAD4, which then gets accumulated in the nucleus and modulates target gene expression [40]. Both Tgfbr1 and Smad1 are co-expressed with Bcl2 and were enriched in TGFβ signaling pathway, suggesting that Bcl2 may be involved in the pathogenesis of leukemia through activating TGFβ/SMAD signaling. Interestingly, in addition to multiple inflammatory and immune response related pathways, a few of the Bcl2-coexpressed genes were found to be enriched in AML (15 genes) and CML (18 genes) KEGG pathways, further strengthening the role of Bcl2 in the pathogenesis of leukemia.

We explored the interactions among the Bcl2-coexpressed genes to further expand our understanding on the function of Bcl2 in leukemia pathogenesis. The resulting PPI network was filtered to obtain a sub-network of genes that directly interacted with Bcl2. Our results showed that many of the direct interactors of Bcl2 were involved in hematopoietic system related functions. A total of 27 genes had direct interaction with Bcl2, among which, Caspase (Casp) 8, BCL2-like 11 (Bcl2l11), mitogen-activated protein kinase (Mapk) 8, myelocytomatosis oncogene (Myc), apoptotic peptidase activating factor 1 (Apaf1), and Mapk1 further interacted with at least 10 other proteins in the network. Hence, these genes may be considered relatively more important as they can modulate the function of other proteins in the network as well as may regulate multiple pathways. While all these genes are directly or indirectly involved in apoptosis, Casp8, Bcl2l11, and Apaf1 have a crucial role in apoptosis, whereas, Mapk8 and Mapk1 are kinase proteins involved in various signaling pathways, and Myc is a proto-oncogene. Both Casp8 and Bcl2l11 interact with at least 50% of the proteins in the PPI network, possibly affecting the function of most of the directly interacting Bcl2 genes. Casp8 encodes a member of the cysteine-aspartic acid protease and its role in leukemia is well known; certain mutations in its sequence abolish caspase-8-mediated apoptosis in AML patients [41]. In contrast, activation of Casp8 by a specific inhibitor of the proteasome induces apoptosis in adult T-cell leukemia cell lines [42]. Our PPI network showed strong interaction (confidence score of 0.8) between Casp8 and Bcl2. Casp8 is known to interact with multiple Bcl2-family proteins containing Bcl-2 homology (BH) domains [43]. Jianbei et al. [44] have shown that OSW-1, a member of the saponin family possessing cytotoxic effect against several types of malignant cells, induced apoptosis in mammalian cells through Casp-8-dependent cleavage of the anti-apoptotic protein, Bcl-2. Bcl2l11 is an apoptotic activator and its downregulation promotes development of stem cell leukemia/lymphoma [45]. It also suppresses Myc-induced mouse B cell leukemia and acts as a tumor suppressor, at least in B lymphocytes [46]. In addition, Bcl2l11 has been shown to mediate synergistic killing of B-ALL cells by BCL-2 and MEK inhibitors [47]. Apaf1 is an apoptotic factor that contains caspase recruitment domain, and promotes intrinsic apoptosis by cleaving and activating pro-caspase 9 [48]. The activated Casp9 further activates downstream effector caspases, such as Casp-3, − 6, and − 7. Consequently, these caspases trigger a signaling cascade that affects various pro-apoptotic and antiapoptotic proteins [5,49]. Reduction in Apaf1 expression has been observed to result in poor prognosis of CLL patients [50]. The PPI network in the current study showed interaction between Apaf1 and Bcl2 as well as multiple caspases. Our study showed significant enrichment of MAPK signaling pathway by Bcl2-coexpressed genes. Mapk1 and Mapk8, also part of the enriched MAPK pathway phosphorylate various transcription factors, including Myc, and c-Fos and proteins related to cell migration and proliferation [51]. c-Myc plays a major role in hematopoiesis and is required for the balance between self-renewal and differentiation of HSCs [52]. Thus, MAPKs including Mapk1 and Mapk8 may be involved in the phosphorylation and activation of different apoptotic proteins, as well as transcription factors involved in the regulation of genes important for the pathogenesis of leukemia.

We further identified a 4.5 Mb genomic region on chromosome 11 that included ~200 potential candidate regulators for Bcl2 expression through eQTL mapping. We used a multi-criteria filtering approach and identified Top2a as the best candidate regulating Bcl2 expression. Our analysis showed that Top2a was significantly correlated with Bcl2 expression, showed cis-regulation, and was involved in leukemia related functions. The relationship of Top2a with Bcl2 further strengthened owing to the correlations between Top2a and Bcl2 family of proteins, including Bcl2l1 (r = 0.55, p = 0.006), Bcl2l11 (r = 0.50, p = 0.012), Bcl2l13 (r = 0.41, p = 0.049), Bcl2l2 (r = 0.59, p = 0.003), Bik (r = 0.41, p = 0.049), Bnip1 (r = 0.55, p = 0.005), and Bok (r = −0.46, p = 0.024). Among these, Bcl2l11 strongly interacts with Bcl2 and has also been implicated in hematopoietic system related functions. Top2a encodes a DNA topoisomerase, an enzyme that alters the DNA topology and is involved in processes, such as chromosome condensation, chromatid separation, DNA replication, and transcription. Analysis of TOP2A at the DNA and RNA level in acute lymphocytic leukemia (ALL) patients found it to be amplified by 2.5–8-fold in 72% of the considered cases, indicating that TOP2A amplification is a frequent event in ALL. Furthermore, the gene amplification was correlated with the overexpression of TOP2A in all the cases [53]. Our results showed a significant correlation between Top2a and Bcl2 expression, which corroborates with the findings from the studies on solid tumors. A study by Olivera et al. [54] on breast cancer subtypes indicated that TOP2A and HER-2 gene amplifications are associated with low levels of p53 and Bcl2 as well as with larger tumor size, positive lymph node status, and higher level of apoptotic and proliferative indexes. Further, the knockdown of Top2a has been shown to affect the expression of apoptosis-related proteins, such as Bcl2 and Bax, in colon cancer cells [55]. Additionally, multiple studies have demonstrated the correlation between decreased levels of TOP2A and drug resistance in selected cell lines [56], including in leukemia cells [57] and patient samples [58]. In addition, alteration in DNA-topoisomerase stability and reduced phosphorylation of TOP2A has been thought to be an underlying cause of drug resistance in leukemia cells by a few studies [59,60]. In agreement with these findings, our pathway enrichment analysis revealed the overrepresentation of Top2a along with other Bcl2-coexpressed genes in platinum drug resistance. A relatively recent study suggested that TOP2A contributes to stem-cell transcriptome regulation and also primes developmental genes for subsequent activation upon differentiation [61]. The authors showed that TOP2α preferentially binds to the promoter regions and exons of the genes in mouse embryonic stem cells. Furthermore, genome-wide correlation analysis at promoters revealed a strong correlation of TOP2α binding with the active histone mark H3K4me2, RNA Pol II occupancy and active transcription, indicating a global regulatory role of Top2a. Additionally, genes that are regulated by Top2a were found to be involved in signaling pathways, biosynthesis processes, and transcriptional regulation. Thus, based on the previous findings and the results from the current study, Top2a may be involved in the regulation of genes associated with HSCs/leukemia through Bcl2; however the underlying mechanism needs to be further explored.

Overall, by taking the advantages of the BXD family of mice for both genomic and myeloid cell transcriptomic data sets, we identified a single genome-wide significant eQTL on chromosome 11 that regulates Bcl2 expression in the current study. Further, multi-criteria filtering for the genes identified in the mapped region, revealed Top2a as the upstream candidate regulator modulating Bcl2 expression. The co-expression and PPI network analysis demonstrated that Bcl2 participates in hematopoietic system regulation through multiple signaling related pathways and by directly interacting with the hematopoietic system related genes, such as Bag3, Bak1, Bcl2l11, Bmf, Mapk9, Myc, Ppp2r5c, and Ppp3ca. Thus, the findings from the current study provide better insights into role of Bcl2 in leukemia pathogenesis and demonstrate the significance of the systems genetics approach in understanding leukemia-associated mechanisms better.

Data availability:

The myeloid cell transcriptomic data of the BXD mice used in this study can be accessed through our GeneNetwork (http://www.genenetwork.org) with the accession number of GN144.