Introduction
The ‘Inhibitor of DNA-binding’ (ID) family of helix-loop-helix proteins function as key regulators of lineage specification and cell fate determination in Metazoa [
1-
3]. In mammals, there are four ID family members (ID1-ID4) that function by heterodimerising with and antagonising the activities of several classes of transcription factor. The E-protein family of basic helix-loop-helix transcription factors (E2A/TCF3, E2-2/TCF4 and HEB/TCF12) are the best characterised ID protein targets [
1-
4]. In hematopoietic cells, individual ID proteins perform distinct, but overlapping functions in a lineage- and differentiation-stage-specific manner [
4-
7]. ID proteins have also been causally implicated in the pathogenesis of leukemias and lymphomas; as in many solid tumour types, ID-mediated tumourigenesis is coupled to various oncogene/tumour suppressor pathways in hematopoietic cells [
6]. Compelling evidence from loss- and gain-of-function studies in transgenic mice and cell line models supports a role for ID proteins in hematopoietic malignancies. Individual ID proteins have been ascribed either an ‘oncogene’ or ‘tumour suppressor’ function in primary human hematopoietic malignancies on the basis of expression level, mutational pattern and functional properties. For example, ID1 is a common downstream target of oncogenic tyrosine kinases, exemplified by BCR-ABL in chronic myeloid leukaemia, driving cell proliferation, survival and invasiveness [
6]. High ID1 expression is also associated with a poor-prognosis subgroup of acute myeloid leukaemia [
8]. Deregulated expression of ID2 is a consistent feature of Hodgkin’s lymphoma and appears to function in concert with ABF-1 in sequestering E2A and probably also PAX5 to augment the B-cell-specific gene regulatory programme in Hodgkin’s-Reed/Sternberg cells [
9,
10]. In Burkitt lymphoma by contrast, the function of the ID3 protein is recurrently inactivated through the acquisition of missense mutations in the
ID3 gene, predominantly affecting the helix-loop-helix dimerisation domain [
11-
13]. The
ID4 gene similarly behaves as a tumour suppressor through epigenetic silencing in most cases of acute myeloid leukemia [
14], while in a sub-group of B-cell precursor acute lymphoblastic leukemia, expression of the
ID4 gene is deregulated by the recurrent t(6;14)(p22;q32) chromosomal translocation [
15,
16].
B-cell chronic lymphocytic leukemia (CLL) is the most prevalent type of leukemia in the Western world and it manifests as a clonal expansion of CD5
+, CD19
+, CD23
+ B cells [
17,
18]. In this leukemia type, the status of only the ID4 family member has been evaluated in detail. In the Eμ-TCL1 mouse model of CLL, loss of an
ID4 allele leads to more aggressive disease while hemizygous loss of
ID4 in nontransformed TCL-1-positive B cells enhances cell proliferation [
19]. These findings, together with the observation that
ID4 mRNA and protein expression is universally silenced in primary human CLL [
14], strongly implicate ID4 as a tumour suppressor in this disease [
19]. For the ID3 family member, microarray gene expression profiling data has shown that the expression of this gene is deregulated in CLL. An analysis of published microarray datasets of Zheng and colleagues [
20] reveals a four-fold upregulation of
ID3 gene expression in CLL compared to normal CD5
+ B-cells. An independent study [
21] showed that
ID3 is among the most significantly overexpressed genes in a multivariate gene expression analysis comparing CLL with normal CD19
+ B-cells, consistent with a potential role in CLL pathogenesis.
In addition to the various roles ascribed to individual ID proteins in regulating cell cycle/cell growth, differentiation, invasiveness, angiogenesis and metastasis in tumours of diverse histological origin, these proteins have also been widely documented to play a key role in regulating cell survival [
1-
4]. However, the behavior of individual ID proteins in functioning as either positive or negative regulators of cell viability is highly cell type-dependent, as illustrated by their contrasting functions in mediating cell survival or cell death in different solid tumour types in response to cytotoxic drugs [
22-
24] (and references therein). Since the primary phenotypic ‘defect’ in CLL cells is their impaired ability to undergo programmed cell death, and this has major implications for cytotoxic drug therapy [
17,
18], it was pertinent to determine whether ID proteins perform a functional role in regulating cell survival in this leukemia, particularly in response to cytotoxic drug treatment. We report here that the ID2 and ID3 proteins impart pro-survival functions in CLL cells cultured
in vitro. In a more physiologically-relevant
in vitro co-culture system, vascular endothelial cells rescue CLL cells from spontaneous and drug-induced cell death via an ID protein-coupled redox-dependent mechanism.
Discussion
Previous studies have shown that ID2 and ID3 are crucial mediators of B-lymphocyte development and are also involved in the regulation of B-cell viability (reviewed in [
4,
39]), functioning to promote either B-cell death or B-cell survival, depending on the particular context. In CLL, our bioinformatics analysis showed that the expression profiles of
ID2 and
ID3 are associated with distinct pathobiological features of this disease. ID2 expression was down-regulated in CLL versus normal B cells in most microarray datasets. Consistent with this, high expression of
ID2 was associated with a more favourable clinical outcome. By contrast,
ID3 expression was consistently up-regulated in CLL versus normal B cells but exhibited no significant association with clinical end-points, at least when analysed in available datasets. Both
ID genes displayed a distinct expression profile amongst the different molecular sub-types of CLL that have been defined previously [
25] and MIC analysis showed that the two
ID genes are coupled to gene regulatory networks that are largely non-overlapping in CLL. However, gene set and pathway enrichment analysis suggested that both
ID genes function in many of the same biological processes and pathways by regulating the expression (directly or indirectly) of a distinct set of target genes. This analysis invoked regulation of apoptosis/cell death in which both
ID genes play a major role in CLL. It should be noted that because of the non-directional nature of MIC-inferred regulatory interactions, not all of the genes identified by MIC analysis are necessarily regulated by
ID2/
ID3 (rather than
vice versa). Nonetheless, given the large number of apoptosis genes that were identified from pathway enrichment analysis (see Figure
4), it is plausible that ID2 and ID3 each regulate the expression of a sizable number of genes (both pro- and anti-apoptotic) involved in cell survival of CLL.
Recent whole-genome and exome sequencing of CLL has revealed that, in contrast to the reported high frequency of recurrent ID3 mutations observed in Burkitt lymphoma and (less commonly) in some other B-lymphoma types [
11-
13], mutations in ID genes do not occur at a significant frequency in CLL [
18] (and references therein). Indeed, of the several hundred CLL cases so far sequenced in different laboratories, only a single instance of a mutated ID gene has so far been reported (a missense mutation - E48V) affecting ID2 [
40], the functional significance of which is unknown. Consistent with the bioinformatics analysis (above), our data show that, at the protein level, the expression of ID2 and ID3 is extremely heterogeneous amongst different CLLs and, from siRNA knock-down experiments, both proteins appear to perform pro-survival functions in both spontaneous and drug-induced cell death in this leukemic cell type. Although based on a very small cohort of CLL samples, we also observed a possible association between low ID3 protein expression (before drug treatment) and
in vitro drug resistance for both fludarabine and chlorambucil. On exposure to these drugs, this was reflected by a pattern of down-regulation of ID3 expression in the more drug-sensitive and up-regulation in the most drug-resistant samples, consistent with the pro-survival function of this ID protein. Albeit with different kinetics, chlorambucil also elicited down-regulation of at least one of the two ID proteins in chlorambucil-sensitive CLLs and up-regulation of both ID proteins in the CLL that was most resistant to this drug. Finally, the drug, ethacrynic acid, which in contrast to fludarabine and chlorambucil, acts at least in part through inhibition of the Wnt/β-catenin signalling pathway [
30] that is also known to be a key regulator of
ID gene expression [
2,
3] (and was also inferred from pathway enrichment analysis of
ID2 MIC ‘targets’ in CLL – see Additional file
6: Table S4). Perhaps unsurprisingly, ethacrynic acid elicited dramatic changes in ID2/ID3 protein expression that were characterised by marked up-regulation of ID2/ID3 levels in those CLLs that were the most resistant to this drug, again consistent with a pro-survival function for the ID2/ID3 proteins in the ethacrynic acid cytotoxic response.
We also noted a possible association between high ID3 protein levels and resistance to spontaneous cell death in the absence of drug treatment, again consistent with a pro-survival function for this ID protein, although this observation should again be interpreted with caution since the analysis was based on a very small cohort of CLLs. However, these findings are in accord with published microarray gene expression studies; datamining reveals that
ID3 mRNA levels are significantly higher in the IGHV-mutated subset of CLL [
41], that is reportedly more resistant to
in vitro spontaneous cell death, than IGHV-unmutated CLLs [
42]. Moreover, low
ID3 mRNA expression is part of a distinct gene expression signature associated with ATM-mutated CLL [
43]. ATM-mutated CLL (low
ID3) define a sub-group of patients with an unfavourable clinical course [
43] and IGHV-mutated CLL (high
ID3) define a more favourable prognostic sub-group (reviewed in [
44]).
Knock-down of ID2/ID3 expression dramatically reduced cell viability in the MEC1 cell line model. But as with the ID expression data, the effect of siRNA knock-down in primary CLL was heterogeneous. A statistically significant effect of ID2/ID3 knock-down on cell viability in the absence of fludarabine treatment was observed in three of the four CLLs examined. The exception (CLL18) expressed the lowest levels of ID2 and ID3, consistent with the observed diminished effect of ID knock-down in these cells. Following fludarabine treatment, the effect of ID knock-down on cell viability was even more heterogeneous, reaching statistical significance in only one CLL, but with an appreciable effect in a further two.
Recent studies have shown that expression of the E-protein bHLH transcription factor, E2A/TCF3, that is one of the key targets of ID proteins, is elevated relative to normal B cell subsets in CLL and also promotes cell survival [
45];
E2A mRNA knock-down leads to reduced CLL cell viability [
45]. A similar effect of
E2A mRNA knock-down has also been described in prostate cancer cells where it was shown to cause down-regulation of ID gene expression [
46]. Since we have shown that loss of ID2/ID3 expression leads to loss of viability in CLL cells, the cell death reported to accompany loss of E2A expression in CLL cells [
45] may well be mediated via loss of ‘downstream’ ID protein expression.
The interaction between CLL cells and the bone marrow/lymph node stromal cell environment
in vivo is known to profoundly affect CLL cell viability and drug sensitivity, and this can be recapitulated
in vitro by co-culture with accessory bone marrow stromal cells [
47] or with vascular endothelial cells [
37,
48]. We found that depletion of GSH using PEITC abrogated HUVEC-mediated rescue of CLL cells from both spontaneous and fludarabine-induced cell death, implicating a redox-dependent pro-survival mechanism imparted by HUVEC cells, similar to that reported for bone marrow stromal cells [
33]. Although off-target effects of PEITC, at the concentrations used in our study, have not been noted in previous studies on CLL [
33,
38] this cannot however be completely ruled out. Consistent with recently published gene expression microarray data [
37], co-culture with HUVEC cells led to up-regulation of ID2 and or ID3 protein expression in CLL cells, an affect that was modulated by PEITC depletion of GSH or by direct addition of GSH or L-cysteine to an extent commensurate with rescue from cell death. These observations are in accord with the pathway enrichment analysis of MIC-inferred regulatory interactions (see Additional file
6: Table S4) which identified the oxidative stress response as a pathway shared by both ID2 and ID3. Although we did not directly determine intracellular GSH levels in our study, the role of GSH in survival of CLL (either in isolation or co-culture with accessory cells) has previously been established by other laboratories from direct measurement of intracellular GSH [
33]. It should be noted that the data in this part of our investigation was based on analysis of only two CLLs. However, given that these CLLs were quite different (one sensitive, the other highly resistant to fludarabine) and both gave broadly consistent results, the data could be considered to be ‘representative’. With this caveat in extrapolating the findings to CLL too generally, our data support a model in which HUVEC co-culture imparts its protective effect on CLL cells at least in part by increasing intracellular GSH levels, which in turn leads to increased expression of the redox-responsive pro-survival ID2 and ID3 proteins.
Materials and methods
Ethics statement
The study protocol including consent procedures was approved by the UK local NHS Ethics Committee (protocol reference: 08/H0302/90). Peripheral blood samples were obtained from CLL patients, together with ‘anonymised’ patient data after informed written consent in accordance with the principles expressed in the Declaration of Helsinki. All records (including signed consent forms) were maintained in a secure database at the Ipswich Hospital NHS Trust, Suffolk, UK.
Datamining of microarray gene expression data
For analysis of
ID2/
ID3 gene expression in CLL versus normal B cells, normalized microarray gene expression datasets were obtained from the NCBI Gene Expression Omnibus database [
49]. Samples representing normal B and CLL cells were curated from each dataset and, after log
2 transformation of expression values, differential expression was analysed using the limma package [
50] in Bioconductor R.
P values for the significance of differential expression were corrected for false discovery rate [
51].
Two microarray datasets with publicly available clinical follow-up data were fRMA-normalised [
52] and downloaded from
InSilico DB [
53]: GSE39671 [
54] with annotation data on time to first treatment for 130 patients, and GSE22762 [
55] with annotation data on both time to first treatment and survival time for 107 patients. Kaplan-Meier plots were constructed using GenePattern [
56] by partitioning samples according to
ID gene expression into
ID2/
ID3-high (upper 50%) and
ID2/
ID3-low (lower 50%) patient groups. The statistical significance of differences in Kaplan-Meier plots was determined by log-rank test.
Consensus clustering analysis was performed essentially as described previously [
25] except that a composite dataset comprised of 871 CLLs from 14 individual datasets (omitting GSE15777) was used. Briefly, CLL samples from each dataset were curated in
InSilico DB [
53] and the fRMA-normalised datasets were downloaded and merged using the ‘COMBAT’ algorithm with the inSilicoMerging R/Bioconductor package [
57]. The combined dataset was marker center-normalized and analysed with the ‘ConsensusClusterPlus’ package in R/Bioconductor [
58] using Euclidean distance and Ward2 agglomerative methods with 1000 iterations. The optimum number of cluster groups (seven) was ascertained from the delta area plot where there was minimal relative decrease in the consensus cumulative distribution function (CDF). Gene signatures representing each cluster group (gene sets significantly up-regulated in each cluster group) were generated using the GenePattern
limma package [
50,
56] applying a Bonferroni-corrected
P value threshold of 0.01 and were analysed for significant overlap with KEGG pathway and Oncogenic signatures databases using the GSEA on-line database [
59].
To identify candidate target genes that are regulated by ID2/ID3 in CLL, we employed the ‘maximal information-based nonparametric exploration’ (MINE) statistics package in Bioconductor, R [
26] to compute maximum information coefficient (MIC) scores and other metrics as a measure of the statistical dependency between expression of
ID2/
ID3 and all other genes in the composite 871 CLL microarray dataset (above). A threshold cut-off MIC score corresponding to a Bonferroni-corrected
P value (calculated using 3×10
7 permutations) of 0.01 was applied to identify the most statistically significant candidate ‘target’ genes. The resulting gene lists were analysed for over-representation of Gene Ontology (biological process) and pathway gene sets using the GeneCodis database [
60]. Candidate apoptosis target genes were further investigated for protein-protein interactions using the ‘String’ (v9.1) database [
28] and for literature-validated regulatory interactions using the ‘Unified Human Interactome’ (UniHI) database [
29]. A network graph was constructed and visualised using Cytoscape v2.8 [
61].
CLL patients, cell isolation and culture conditions
Peripheral blood samples from 14 CLL patients were obtained from patients attending the Hematology out-patient clinic at Ipswich Hospital NHS Trust (Suffolk, UK). Patients were selected having a white cell count of >45×10
9/L in order to ensure a high representation of CLL cells. Patient characteristics are listed in Additional file
7: Table S5. Peripheral blood mononuclear cells were isolated by density gradient centrifugation using Ficoll-Paque PLUS (GE Healthcare, Little Chalfort, UK) according to the manufacturer’s protocol. The MEC1 cell line [
31] (DMSZ, Braunschweig, Germany) and primary CLL cells were cultured in IMDM supplemented with 10% fetal bovine serum (FBS) and 45 μg/ml gentamicin (all from PAA, Pasching, Austria) at 37°C and 5% CO
2 in a humidified incubator. Primary CLL cells were seeded at a density of 1.5-5×10
6 cells/ml, while MEC1 cells were maintained at a cell density between 0.5-1×10
6 cells/ml.
Human umbilical vein endothelial cells (HUVEC, TCS Cellworks, Buckingham, UK) were cultured in human large vessel endothelial cell growth medium (TCS Cellworks) on poly-lysine-coated tissue culture flasks. For co-culture experiments, HUVECs were seeded at 60% density and incubated for 24 hrs before the medium was removed and CLL cells were seeded on top of the monolayer at 1.5×106 cells/ml in IMDM/10% FBS. Cells were maintained under these co-culture conditions for 24 hrs, prior to drug addition and then cultured in the presence of drugs for another 48 hrs. The ‘double conditioned medium’ (CM), was harvested from a co-culture of HUVEC and CLL cells after 48 hrs, filter-sterilized and stored at 4°C until required. Glutathione (GSH), Phenethyl-isothiocyanate (PEITC), L-cysteine, fludarabine, chlorambucil and ethacrynic acid were purchased from Sigma-Aldrich (St-Louis, MO, USA).
Lentivirus production and infection of MEC1 cells
Lentiviral vectors encoding siRNA targeting either control or the
ID2 and
ID3 genes were purchased from Applied Biological Materials (ABM) Inc. (Richmond, BC, Canada). The vector backbone (piLenti-siRNA-GFP) contains convergent U6 and H1 promoters producing double-stranded siRNA molecules. The sequences targeted by these siRNAs were as follows:
-
ID2R-siRNA: 5’-TGTGGACGACCCGATGAGC-3’,
-
ID2Y-siRNA: 5’-ATCGACTACATCTTGGACCTGCAGATCGC-3’,
-
ID2G-siRNA: 5’-CCCACTATTGTCAGCCTGCATCACCAGAG-3’,
-
ID2B-siRNA: 5’-TCTGAGTTAATGTCAAATGACAGCAAAGC-3’,
-
ID3R-siRNA: 5’-ACTCAGCTTAGCCAGGTGGAAATCCTACA-3’,
-
ID3Y-siRNA: 5’-ATCGACTACATTCTCGACCTGCAGGTAGT-3’,
-
ID3G-siRNA: 5’-ACCTTCCCATCCAGACAGCCGAGCTCACT-3’,
-
ID3B-siRNA: 5’-CCGGAACTTGTCATCTCCAACGACAAAAG-3’,
-
Negative control siRNA: 5’-GGGTGAACTCACGTCAGAA-3’.
The second generation packaging plasmids psPax2 and pMD2.G were purchased from Addgene (deposited by Prof. Didier Trono, Lausanne, Switzerland). Human embryonic kidney (HEK 293 T) cells were cotransfected with 2 μg siRNA lentiviral expression vector together with 1.3 μg psPax2 and 0.7 μg pMD2.G using the calcium phosphate method and then used for a 48 hr co-culture with 5×105 MEC1 cells followed by selection for 14 days with puromycin (1.5 μg/ml). Stable pools of transduced MEC1 cells were then expanded in IMDM/10% FBS without puromycin.
siRNA transfection of primary CLL cells was performed using a HiPerfect transfection kit (Qiagen, Hilden, Germany), according to the manufacturer’s guidelines. Briefly, 1×10
6 CLL cells were seeded in 200 μl IMDM/10%FBS and transfected with 60nM siRNA (final concentration in 600 μl) and 5 μl HiPerfect transfection reagent in 100 μl IMDM. After 6 hrs incubation, 300 μl of fresh IMDM/10%FCS containing 45 μg/ml gentamicin were added. Knockdown efficiency was assessed by western blotting after 72 hrs. Pre-designed, chemically-modified, siRNA oligoribonucleotides (Stealth™) targeting
ID2 or
ID3, as well as a negative control (Stealth™ medium GC Duplex) were purchased from Invitrogen (Carlsbad, CA, USA). The sequences of the
ID siRNAs were as follows:
Cell viability assay
Cell viability was assessed using the 3-(4,5-dimethilthiazol-2yl)-2,5-diphenyl tetrazolium bromide (MTT) assay. MEC1 cells were incubated with 0.5 mg/ml MTT for 2 hrs at 37°C, 5% CO2, while primary CLL cells required 4 hrs of incubation. Cells were then harvested and recovered by centrifugation and incubated in 100–200 μl dimethylsulphoxide at 37°C for 20 minutes before optical density was recorded in a microplate spectrophotometer at 560 nm. The percentage of cell viability assessed by MTT assay was used to determine IC50 concentrations from dose–response curves at 72 hrs following treatment with 4 different drug concentrations.
Western blotting
Cells were lysed, electrophoresed on polyacrylamide (15%)-SDS gels followed by western blotting, essentially as described previously [
62]. After incubation with primary rabbit polyclonal antibodies against either ID2 (sc-489) or ID3 (sc-490) (Santa Cruz Biotechnology, Dallas, TX, USA), membranes were washed in 0.05% Tween-20 in TBS and incubated for 1 hr with polyclonal goat horseradish peroxidase–conjugated secondary antibody (AbCam, Cambridge, UK). Following extensive washes, bound antibodies were visualized by enhanced chemiluminescence (ECL, Millipore, Billerica, MA, USA) on X-ray films. Membranes were then stripped by incubation with stripping buffer (25 mM glycine pH2, 1% SDS) for 30 minutes, washed and reprobed with primary rabbit polyclonal antibody against GAPDH (Sigma-Aldrich), followed by secondary antibody detection as described above. Protein bands on western blot films were quantified by densitometric scanning and analysis using ‘ImageJ’ software. Heatmap images were created using the ‘HeatMapViewer’ module in the GenePattern software suite [
56].
Ad hoc statistical analysis
For MTT assay data, continuous variables were compared using the Student’s
t-test while correlation between continuous variables was performed by determination of Pearson’s correlation coefficient using the ‘VassarStats’ online statistical calculator [
63]. All statistical tests were two-sided. Boxplots were generated by using the ‘BoxPlotR’ online tool [
64]. Statistical analysis of gene overlap by hypergeometric distribution was performed using the ‘phyper’ algorithm in Bioconductor R.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SW JAA & JDN conceived and designed the experiments. JDN performed datamining and bioinformatics analysis, analysed the data and wrote the manuscript. JAA coordinated specimen collection. SW performed all laboratory experiments, analysed the data and wrote the manuscript. All authors read and approved the final manuscript.