Potential Role of Notch Signalling in CD34+ Chronic Myeloid Leukaemia Cells: Cross-Talk between Notch and BCR-ABL

Notch signalling is critical for haemopoietic stem cell (HSC) self-renewal and survival. The role of Notch signalling has been reported recently in chronic myeloid leukaemia (CML) – a stem cell disease characterized by BCR-ABL tyrosine kinase activation. Therefore, we studied the relationship between BCR-ABL and Notch signalling and assessed the expression patterns of Notch and its downstream target Hes1 in CD34+ stem and progenitor cells from chronic-phase CML patients and bone marrow (BM) from normal subjects (NBM). We found significant upregulation (p<0.05) of Notch1, Notch2 and Hes1 on the most primitive CD34+Thy+ subset of CML CD34+ cells suggesting that active Notch signalling in CML primitive progenitors. In addition, Notch1 was also expressed in distinct lymphoid and myeloid progenitors within the CD34+ population of primary CML cells. To further delineate the possible role and interactions of Notch with BCR-ABL in CD34+ primary cells from chronic-phase CML, we used P-crkl detection as a surrogate assay of BCR-ABL tyrosine kinase activity. Our data revealed that Imatinib (IM) induced BCR-ABL inhibition results in significant (p<0.05) upregulation of Notch activity, assessed by Hes1 expression. Similarly, inhibition of Notch leads to hyperactivation of BCR-ABL. This antagonistic relationship between Notch and BCR-ABL signalling was confirmed in K562 and ALL-SIL cell lines. In K562, we further validated this antagonistic relationship by inhibiting histone deacetylase (HDAC) - an effector pathway of Hes1, using valproic acid (VPA) - a HDAC inhibitor. Finally, we also confirmed the potential antagonism between Notch and BCR/ABL in In Vivo, using publically available GSE-database, by analysing gene expression profile of paired samples from chronic-phase CML patients pre- and post-Imatinib therapy. Thus, we have demonstrated an antagonistic relationship between Notch and BCR-ABL in CML. A combined inhibition of Notch and BCR-ABL may therefore provide superior clinical response over tyrosine-kinase inhibitor monotherapy by targeting both quiescent leukaemic stem cells and differentiated leukaemic cells and hence must be explored.


Introduction
The Notch signalling is evolutionarily conserved pathway, which plays an important role in regulating the process of development [1,2]. Notch signalling is composed of a Notch receptor, ligand and CBF1/Su(H)/Lag-1 family (CSL) DNA binding protein. In mammals, four Notch receptors (Notch1-4) and five Notch ligands (Jagged1 and 2, and Delta-like1, 3 and 4) have been identified [3]. Upon binding of Notch ligands to its receptors, the receptor undergoes at least two proteolytic cleavages, releasing the activated Notch intracellular domain (NICD) into the cytoplasm, which then translocate to the nucleus and subsequently activates the transcription of the Notch target genes. The important downstream target genes of Notch include Hes1, Hes5 and Hey. The activated Notch pathway determines the cell fate and maintains the stem cell state [1,2].
Chronic myeloid leukaemia (CML) is a stem cell disease with the differentiated cells in CML constituting the bulk of leukemic cell mass, whereas the leukemic stem cells are responsible for the disease maintenance [4,5]. It has been shown that IM is highly toxic to the differentiated CML progenitors but not to the leukemic stem cells which remain viable in a quiescent state [6,7]. Therefore, it is possible that CML stem cells survival and self-renewal capacity is related to the same pathway, which regulates the normal HSCs such as Notch signalling pathway.
Well established role for Notch signalling in human T-cell acute lymphoblastic leukaemia (T-ALL) [8,9], and the axon development in Drosophila support the hypothesis of possible interactions between ABL protein kinase and Notch signalling. It has been found that Notch interacts genetically with ABL, as Notch and ABL mutations synergise to cause synthetic lethality in Drosophila axons [10]. In another study, Giniger and colleagues have also found that Delta ligand and Notch provides a guidance signal to the developing axon by regulating the ABL kinase signalling pathway [11].
In CML, Notch signalling has been demonstrated to mediate the disease progression [12] and in K562 CML cell line model Notch signalling inhibited the development of erythroid/ megakaryocytic cells by induction of Hes1 [13] and proliferation of K562 cells [14]. Recently, Yang et al., showed that over-expression of Notch2 inhibits the proliferation of CML cells [15,16]. Hes1 which is the most widely characterised Notch downstream target gene has been shown to immortalize committed progenitors and play a role in transformation of chronicphase CML to blast crisis [9,17,18]. However, the underlying molecular relationship between Notch signalling and CML remains largely unknown. Based on the above findings and the notion that Notch co-operates with several signal-transduction pathways to induce leukaemogenesis, we hypothesized that Notch signalling may be altered in CML, and that Notch might interact with the BCR-ABL fusion protein in CML.
Therefore based on our hypothesis the objectives of this study were (i) To investigate the expression of Notch receptors in CD34+ primary CML at mRNA and protein level, (ii) To investigate the expression of Notch target genes Hes1 and Herp1&2 using PCR to determine the activity of Notch signalling in CD34+ primary CML cells, (iii) To investigate the possible cross-talk between Notch and BCR-ABL in primary CD34+ CML cells as well as in cell line models, and lastly (iv) To validate the Notch-BCR-ABL relationship using CML microarray datasets from GSE database.

Primary chronic myeloid leukaemia samples
Fresh or frozen peripheral blood samples from non-treated patients with chronic myeloid leukaemia (CML) in chronic phase were used in this project. Bone marrow (BM) of normal subjects (NBM) and cord blood from normal subjects were used as controls.

Ethics statement
Bone marrow and cord blood samples were kindly provided by Dr. John Burthem (Clinical Senior Lecturer, University of Manchester, Manchester, UK) and all the samples were ethically approved by University of Manchester committee as described [19]. CML samples were kindly provided by Prof., Tessa Holyoake, (professor of Experimental Hematology, Paul O'Gorman Research Centre, Gartnavel General Hospital, Glasgow, UK) and were ethically approved [20,21].

Isolation of mononuclear cells (MNC)
Mononuclear cells from blood samples were isolated using ficoll-paque (Amersham Pharmacia Biotech, UK) density gradient method under sterile conditions according to the manufacturer's instructions. Samples were diluted 1:1 with hanks balanced salt solution (HBSS) (Sigma-Aldrich, UK) supplemented with 5% newborn calf serum (NCS) (Invitrogen, UK). 20 ml of the diluted blood was then carefully layered onto 10 ml ficoll in a 50 ml falcon tube and centrifuged at 389g for 30 minutes at room temperature (RT). Mononuclear cells were harvested from the interface layer and washed twice with 50 ml HBSS/5%NCS by centrifugation at 389g at RT for 7 minutes. The pellet was then re-suspended in known volume of HBSS/5% NCS for FACS sorting, or processed for storage in liquid nitrogen.

Isolation of haemopoietic progenitor cell populations
Haemopoietic progenitors positive for CD34 were isolated from cord blood from normal subjects and CML sample, using StemSep kit (StemCell Technologies, UK) according to the manufacturers' instructions. The eluted CD34+ve cells were then pooled and viability assessed before cells were pelleted and re-suspended in 100 μl of (1:20 dilution) CD34-APC, (1:20 dilution) Thy-PE and (1:20 dilution) Lin-FITC cocktail and incubated for 20 minutes at 4°C in the dark. Cells were then washed with 2 ml HBSS/5% serum and re-suspended in 1 ml HBSS/5% serum for sorting. Cells were sorted into a 24 well plate using a FACS Vantage (Becton Dickinson, USA) flow cytometer. Sorted cells were then transferred into RNAse free eppendorf tubes. 100 ng/ml Flt3-ligand, 100 ng/ml stem cell factor, 20 ng/ml each of interleukin (IL)-3, IL-6 and granulocyte colony stimulating factor (GCSF) (R&D Systems, UK).

FACS analysis of extra-and intra-cellular Notch1
To study the expression of the extra-cellular Notch1 (ECN1) and intra-cellular Notch1 (ICN1) on the cell surface of K562 /primary CD34+ CML, cells (1x10 6 ) were directly stained with EA1 and b-TAN20 antibody respectively (S1 Table) as described previously [24,25]. As K562 cells are negative for CD34 surface antigen the analysis gate used here included all live K562 cells.

Poly-A PCR
RNA extraction, construction of cDNA from low cell number, cDNA tailing reaction and poly-A PCR was carried out as described previously [22].
Construction of cDNA from high cell numbers. High Capacity cDNA Reverse Transcription Archive Kit (Applied Biosystems, UK) was used for the cDNA synthesis from cell numbers higher than 1x10 5 cells, according to the manufacturer's instructions.
Gene specific PCR. Gene specific PCR was performed in 10 μl reactions consisting of 5 μl PCR Reddy Mix (ABgene, UK), 0.5 μl of forward and reverse primers, (S2 Table) (Sigma-Aldrich, UK) and 1 μl of human genomic DNA diluted 1:500 (Promega, UK) as described previously [22]. Agarose gel electrophoresis was used to resolve and visualise PCR product using 1.5% agarose (Sigma-Aldrich, UK) in TBE. The size of the PCR products was determined by the GeneRuler 100bp ladder (Fermentas, UK). Finally, the resulting gel was observed on a Typhon 8600.

QRT-PCR using TaqMan probes
Real time PCR experiments were performed using TaqMan probes as described previously [22,25]. This method was used to measure Notch1, Notch2, and Hes1 in CML patients as well as bone marrow from normal subjects. S3 Table shows the list of primers and TaqMan probes used in real time PCR.
The P-crkl assay CRKL is an important substrate of the BCR-ABL oncoprotein in CML and binds to both BCR-ABL and c-Abl. [26] Crkl tyrosine is phosphorylated in CML cells which is absent in normal haemopoietic cells, [26] hence it has been recognized as a prognostic marker for CML. The levels of phosphorylated crkl (P-crkl) were measured by intra-cellular FACS. Cells (1 x 10 5 ) were harvested and washed once in HBBS/5% FBS, then fixed in fixing reagent (Caltag Laboratories, UK) and incubated at RT for 15 minutes. The cells were then washed once with 3 ml HBBS/ 5% FBS and re-suspended with 25 μl permeabilizing reagent (Caltag Laboratories, UK) and 2.5 μl of P-crkl primary antibody (New England Biolabs, UK). The cells were then gently vortexed and incubated at RT for 40 minutes, before being washed twice with 3 ml HBBS/ 5% FBS. The secondary antibody was added directly at the appropriate dilution and the cells were mixed and incubated at RT for 30 minutes in the dark. The P-crkl results were reported as mean fluorescence intensity (MFI).

Western blotting
Total protein was extracted from K562 cells using RIPA buffer (Sigma-Aldrich, UK) containing phosphatase and protease inhibitors (Sigma-Aldrich, UK). Protein concentration was determined by bio-rad protein assay (Bio-Rad, UK). Proteins were separated on SDS-PAGE and separated proteins were transferred to a nitrocellulose membrane (Sigma-Aldrich, UK). The membrane was incubated in blocking buffer (5% skimmed milk in 1x PBS containing1% (v/v) TWEEN 20) (Sigma-Aldrich, UK) for 1 hr at RT and subsequently stained with primary antibody (S2 Table) for 2h at RT or at 4°C overnight. The membrane was then washed once with water and three times with PBST (1x PBS containing 1% (v/v) TWEEN 20) for 15 minutes at RT prior to adding secondary antibody for one hour at RT. The membrane was washed 4 times at 10 minutes intervals at RT. After applying chemiluminescence substrate (Pierce, UK), the membrane was developed using an autoradiography in a Fuji film FPM800A automated developer.
Analysis of CML CD34+ patients and BM from normal subjects microarray dataset from GSE database Three datasets of microarrays on CD34+ cells isolated from BM from normal subjects and patients with CML were retrieved from GSE database. These datasets were GSE550 (n = 17; Affymetrix Human HG-Focus Target Array), GSE1418 (n = 14; Affymetrix Human HG-Focus Target Array), GSE12211 (n = 12; Affymetrix Human Genome U133A 2.0 Array) [27][28][29]. 7947 genes common to these three datasets were used for analysis. Using Combat function in R, batch effects in these datasets were combatted as described previously [30]. In these datasets, there were normal donors (n = 14), CML patients prior treatment (n = 15), CML patients treated with Imatinib (IM) for a median duration of 28 months (n = 8) and CML patients treated with IM for 7 days (n = 6). Three axis principal component analysis (PCA) was performed on combatted datasets using DUDI.PCA function. Gene set enrichment analysis (GSEA) as previously described, using KEGG pathways was also performed [31]. Gene expression signature of BM from normal subjects was compared with the paired samples from GSE12211 datasets which included samples from patients with CML pre-and 7 days post-IM treatment.

Statistical analysis of datasets
The Mann-Whitney test was used to compare the differences in gene expression between BM from normal subjects and CML biological samples. To assess the statistical difference in drug treated samples, a paired T-test was carried out. All differences with p<0.05 were considered statistically significant. Data were expressed as mean± SEM for all the statistical analysis.

Results
Expression profile of Notch receptors and Notch target genes in CD34 + cells in the chronic phase of chronic myeloid leukaemia at transcriptional level The expression profile of Notch1-4 receptors was studied in four CML patient samples along with four bone marrow samples from normal subjects (NBM) using the polyA PCR technique. Cells in each sample were fractionated into CD34+ Thy+, CD34+ Thy-, and total CD34+ subsets to enable the study of gene expression in haemopoietic progenitors at different maturation levels and sorted cells were of 95% purity. Fig 1a shows the PCR profile of Notch receptors in both NBM and CML samples. The housekeeping gene GAPDH was used to assess the quality of cDNA and to check the uniformity of DNA content among different samples.
Notch1 was expressed in all the normal samples (n = 4) and in all three haemopoietic CD34+, Thy+, and Thy-subpopulations (n = 4). Interestingly, Thy+ and Thy-, both subsets expressed Notch1 in the CD34+ cells taken from BM of normal subjects. Similar results were also seen in CML samples with no clear evidence of differences in the expression between the CD34+, Thy+, and Thy-subpopulations. Notch2 was weakly expressed in all three CD34+ subpopulations in both NBM and CML samples. Notch3 did not show any expression in either NBM or CML samples (data not shown). Notch4 was irregularly expressed and was seen in two normal and two CML CD34+ samples, respectively (Fig 1a). In order to determine the significant difference between the Notch expression seen in NBM and CML CD34+ populations, quantitative RT-PCR was performed. QRT-PCR showed a 3-4-fold statistically significant (p<0.05) upregulation in Notch1 transcripts in CD34+ Thy+ cell subset compared with NBM. However, no significant difference was observed in Notch1 expression in CD34+ CML and CD34+ Thy-subpopulation compared to NBM (Fig 1b). Interestingly, Notch2 was significantly (p = 0.02) upregulated in all CML subpopulations (n = 4) compared with NBM. There was more than a 100-fold increase in Notch2 expression in the CD34+ Thy+, CD34+ Thy-, and in the total CD34+ cell subsets as compared with NBM samples (Fig 1c).
The expression of Notch target genes Hes1, Herp1, and Herp2 was also studied to assess the active Notch signalling in CML compared with normal CD34+ cells. Results showed that neither Herp1 nor Herp2 was expressed in both NBM and CML patient samples. Interestingly, Hes1 expression was seen in both normal and CML samples with no precise pattern of activity discernible, suggesting that Notch signalling was activated in these samples (Fig 2a). However, quantitative real-time PCR analysis demonstrated a greater than 100-fold significant (p<0.05) increase in the Hes1 expression in all the CML CD34+ cell subsets (n = 4) compared with NBM (n = 4) (Fig 2b).

Differential expression profile of Notch1 in chronic myeloid leukaemia at protein level
Does overexpression of Notch1 mRNA get translated to its protein product in CML patients? To address this question, mononuclear cells from chronic phase CML patients were stained with EA1, a monoclonal antibody [24] that recognises the extracellular domain of Notch1 (ECN1) [25] (S1 Fig).
In CML, Notch1 was expressed in 35% of gated CD34+ primary cells compared with 21% of CD34+ cells in normal cord blood (Fig 2c and S2 Fig). Finally, the expression of Notch1 was also confirmed on the very primitive haemopoietic CD34+ Thy+ and stem cell-enriched CD34+/CD38-progenitor cells, and results showed that Notch1 was expressed in 21.6 ± 2.3% of the CD34+ Thy+ population and 15.0 ± 0% of the CD4+ CD38-population (n = 4) (Fig 2c  and S3 Fig). Interestingly, stem cell-enriched CD34+/CD38-progenitor cells (corresponding to primitive HSC) do not show significant difference in Notch1 expression at protein level between cord blood samples from normal donors and in CML samples from the chronic CML patients (Fig 2c). In conclusion, the above results confirm that Notch1 is upregulated in the most primitive CD34+ Thy+ cells in the chronic phase of CML as compared to normal cord blood, suggesting that Notch signalling might be involved in the survival and/or self-renewal of leukemic stem cells in CML.

Demonstration of cross-talk between Notch and BCR-ABL in primary CD34+ chronic myeloid cells in vitro
The above results confirm the overexpression of Notch and its target gene Hes1 in primary CD34+ CML cells. Moreover, it has been demonstrated that enhanced kinase activity of BCR-ABL and altered expression of Notch1 synergises to induce acute leukaemia in a transgenic model for CML [32]. Therefore, we initially assessed the interaction between Notch and BCR-ABL in primary CD34+ CML cells and then proceeded to establish these interactions in CML cell line models.
(i) Imatinib mesylate (IM) inhibits BCR-ABL activity in chronic phase CML CD34+ primary cells. After establishing the cross-talk between Notch and BCR-ABL in the CML cell line models, we wished to delineate the interaction between Notch and BCR-ABL in the primary CD34+ CML patient cells.
Increased kinase activity of BCR-ABL in CD34+ CML primary cells was first confirmed by a P-crkl expression using P-crkl-FITC conjugated antibody (S4 Fig). The CD34+ CML primary cells were isolated from frozen blood samples of CML patients (n = 6) whose Notch activity was not confirmed. Therefore, to assess the activity of Notch signalling in these cells, Hes1 gene expression was measured in the CD34+ cells in all CML samples by real-time PCR. Results showed that Hes1 expression was significantly (p<0.01) upregulated in CD34+ cells in all six CML samples compared with control (normal CD34+ cells from NBM) group (Fig 3). The efficacy of IM on primary CD34+ CML was tested by FACS-based P-crkl assay. Samples from patients with chronic phase CML (n = 5) were enriched for CD34+ and cultured for 24h in serum-free medium (SFM) supplemented with growth factor cocktail comprising 100 ng/mL Flt3-ligand, 100 ng/mL stem cell factor, and 20 ng/mL each of interleukin (IL)-3, IL-6 and granulocyte-colony stimulating factor (G-CSF). Primary CD34+ CML cells were then treated with 10 μM IM for 72h, and the inhibitory effect of IM on CD34+ CML primary cells was assessed by the P-crkl assay, and K562 cells were used as a positive control. Results showed that expression of P-crkl was clearly reduced in CD34+ cells from three CML samples as compared with the control group (Fig 4a). The results also demonstrated that BCR-ABL activity was inhibited by IM in CD34+ cells by marked reduction of crkl phosphorylation. These results are in good agreement with Chu et al. (2004), who showed that IM-induced inhibition of crkl phosphorylation in CML CD34+ cells in a dose-dependent manner [33]. However, two out of five CD34+ CML primary cells showed some degree of P-crkl expression at 72h post-IM treatment (Fig 4b), suggesting that these two samples (CML-2 and CML-4) are resistant to IM treatment.
(ii) IM upregulated Notch target gene Hes1 expression in CD34+ CML cells. Next, we investigated the effect of BCR-ABL inhibition on Notch signalling in CD34+ CML primary cells. All CML samples were enriched for CD34+ cells using magnetic CD34 selection. After  culturing in the presence of IM, the percentage of CD34 cells was assessed by FACS, and RNA was extracted directly from cultured cells if they were >90% CD34+, or FACS sorted if they were <90% CD34+. The Notch target gene Hes1 transcript levels were measured using qRT-PCR. There was a significant (p<0.05)~4-fold increase (n = 4) in Hes1 gene expression following IM treatment (Fig 4c). Interestingly, two CML samples (CML-2 and CML-4), which exhibited resistance to IM, revealed no significant (p>0.05) difference in Hes1 expression compared with the control group (Fig 4d). From the above results, it was concluded that CD34-+ CML primary cells sensitive to IM showed further activation of Notch.
(iv) GSI increases the kinase activity of BCR-ABL in CD34+ CML primary cells. To further explore the cross-talk between Notch and BCR-ABL, the effect of the Notch inhibitor GSI on BCR-ABL activity was investigated in CD34+ CML primary cells. Cells were cultured in growth factor cocktail for 72h in the presence of 10 μM GSI before measuring BCR-ABL activity by P-crkl assay. To calculate the change in P-crkl expression, the mean fluorescence intensity (MFI) of GSI-treated and untreated CD34+ cells was first determined by subtracting the MFI of P-crkl-stained cells from the MFI of isotype control in each condition. The MFI of GSItreated cells was then compared with the MFI of untreated cells, and the change in P-crkl expression was reported as a percentage.
Interestingly, FACS data showed that GSI treatment increased the P-crkl expression in CD34+ cells between 18%-42% as compared with the control group (Fig 5c). Importantly, an increase in the P-crkl expression was seen in all the CD34+ CML primary cells that showed downregulation of Hes1 mRNA post-GSI treatment, suggesting that the increase in P-crkl expression is Notch dependent. The increase in P-crkl in CD34+ CML cells (n = 6) was statistically significant (p<0.01), as shown in Fig 5d. Evidence of cross-talk between Notch signalling and BCR-ABL in chronic myeloid leukaemia cell line model (i) K562 cell line as a model for BCR-ABL and Notch cross-talk. K562 cells are known to show high BCR-ABL activity, which can be inhibited by IM and expression of active Notch signalling [33]. Moreover, K562 cells express the Notch downstream target gene Hes1 [14,16] at levels that can be inhibited by GSI; hence, it is a good cell line model to study the interaction between BCR-ABL and Notch. First, we confirmed the expression of BCR-ABL and Notch1 in K562 cells (Fig 6a and 6b). We then validated for optimal P-crkl expression at different dilutions of a P-crkl antibody (S5a- S5c Fig) and at different passages (S5d-S5f Fig). GSI has been shown to inhibit Notch signalling in normal CD34+ cells and in T-ALL cell lines [22,34]. To investigate the effect on BCR-ABL activity following Notch inhibition, K562 cells were cultured with 10 μM GSI for 24h. The cells were then harvested, and an intracellular P-crkl expression was performed. Hes1 expression was assessed by real-time PCR to confirm the inhibition of Notch activity. Real-time PCR results showed significant (p<0.01) downregulation of transcriptional target gene Hes1 in the GSI-treated K562 cells (Fig 6c). The FACS analysis showed a dramatic increase in P-crkl expression in the GSI-treated cells compared with the control group (Fig 6d). To study the effect of IM-induced BCR-ABL inhibition on Notch signalling, cells were cultured in the presence of an increasing concentration of (0.1, 0.5, 1, 5 and 10 μM) IM for 48h. Following confirmation of significant dose-dependent P-crkl inhibition (Fig 7a-7c), real-time PCR was performed for Hes1 expression. Results showed that Hes1 was    (Fig 7d). Furthermore, we also evaluated the effects of histone deacetylase (HDAC) inhibitor valproic acid (VPA) in the K562 cell line model. Treatment of VPA significantly (p = 0.01) inhibited the Notch target gene Hes1 (Fig 7e) and induced increased phosphorylation of crkl (Fig 7f) in the K562 cell line model.
(ii) ALL-SIL cell line as a model for ABL-Notch cross-talk. We further assessed the ABL-Notch cross-talk in the ALL-SIL cell line model since the reported activity of ABL and Notch signalling in ALL-SIL cells [35,36] also makes this cell line a good In Vitro model to evaluate the cross-talk of ABL and Notch. The FACS-based P-ckrl assay showed the expression of P-crkl as a marker for the ABL kinase activity in ALL-SIL cells (Fig 8a). To study if the ABL activity can be switched off and to assess whether the P-crkl assay can be utilised as an IM sensitivity assay, the effect of IM was investigated on ALL-SIL cells.
ALL-SIL cells were cultured in the presence of 10 μM IM for 48h. The cells were then stained with P-crkl primary antibody and PE secondary antibody. Results showed that ABL kinase is active in ALL-SIL cells, and this activity is evident by the phosphorylation of P-crkl in the absence of IM. Treatment of ALL-SIL cells with IM resulted in clear reduction of P-crkl expression (Fig 8a). To further investigate the effect of ABL inhibition on Notch signalling, real-time PCR was performed to see the effect of IM on Hes1. Results showed that Hes1 was significantly (p<0.01) upregulated in ALL-SIL cells at 48h posttreatment with 10 μM IM (Fig 8b).

Validation of Notch-BCR-ABL cross-talk using CML CD34+ patients and NBM donor microarray datasets from GSE database system
To corroborate our In Vitro IM treatment data, we also performed a three-axis principal component analysis (PCA) on the three retrieved GSE microarray datasets. PCA analysis showed that patients treated with IM for a prolonged duration (28 months) clustered with NBM donors, and importantly, patients treated with IM for seven days showed a gene expression profile closely related to chronic phase CML (Fig 9a). However, at seven days post-IM treatment, some deviation in gene expression from chronic phase CML was observed, suggesting early changes in gene expression occur in CD34+ cells from chronic phase CML following a sevenday IM therapy.
To assess the role of Notch signalling in chronic phase CML and following BCR-ABL inhibition with IM therapy, we compared the gene expression profile of CD34+ cells from NBM donors with paired samples of CD34+ cells from chronic phase CML patients pre-and seven days post-IM therapy. Interestingly, we found that Notch signalling was upregulated both preand seven days post-IM therapy at false discovery rate (FDR) value of 0.005 and 0.1 as compared with NBM donors. Furthermore, in comparison with normal CD34+ cells from NBM, cell cycle and DNA replication pathways were significantly upregulated in CD34+ cells from chronic phase CML patients. Although these pathways were significantly downregulated in CD34+ cells from CML patients at seven days post-IM therapy, Notch signalling remained upregulated in both pre-and seven days post-IM therapy (Fig 9b). Similar to our In Vitro results, we observed that Hes1 was significantly upregulated pre-IM therapy and further upregulated at seven days post-IM treatment. Importantly, we also observed that JAG-1 and JAG-2, which are Notch target ligands, were also significantly upregulated post-IM (Fig 10). Thus, gene expression profile analysis from the retrieved datasets further corroborated our findings of activated Notch signalling during both hyperactivation and inhibition of BCR-ABL activity.

Discussion
Notch signalling controls cell fate decisions, and stem cell renewal and differentiation [2,9]. The emerging evidence demonstrates the role of activated Notch signalling in haematopoietic malignancies [37]. Deregulation of Notch signalling is also implicated in the development of chronic B-cell lymphoid leukaemia [9]. Interestingly, activation of mutations of the Notch genes have been identified in more than 50% of human T-cell acute lymphoblastic leukaemia and in a subset of non-Hodgkin lymphomas [9]. Therefore, genetic or pharmacological manipulation of Notch signalling is a novel potential strategy for the treatment of many cancers. It is therefore important to understand how the genetic abnormality that causes cancer is related to Notch signalling. This study summarises the preclinical evidence, linking Notch signalling to BCR-ABL in chronic phase CML.
In the present study, the role of Notch signalling was evaluated in CML. Notch1 and Notch2 were significantly upregulated in the most primitive CD34+ Thy+ and total CD34+ subset in the chronic phase of CML. In particular, activation of Notch signalling in the CD34+ Thy+ subset indicates it may have a role in leukemic stem cell expansion similar to that of normal stem cells [38]. Hes1 is upregulated in the CD34+ Thy+, CD34+ Thy-, and in the total CD34+ cell subsets. Moreover, the expression of Notch1 in the CD34+ Thy+ and CD34+ CD38-cell subsets is interesting as these populations are enriched for leukemic stem cells in CML [39].
To understand the relationship between Notch and BCR-ABL, we used inhibitors of both Notch and ABL signalling and found that the Notch and BCR-ABL pathways antagonise each  other in primary CD34+ cells isolated from chronic phase CML as well as in cell line models. Inhibition of BCR-ABL activity in primary CD34+ CML cells resulted in significant upregulation of Hes1. Hes1 functions as a transcriptional repressor by interacting with histone deacetylase HDAC1 [40,41]. Interestingly, HDAC inhibitor VPA treatment in the K562 cell line model also resulted in the inhibition of Notch signalling/Hes-1 and overexpression of BCR-ABL activity, further confirming the antagonistic interaction between the Notch signalling pathway and BCR-ABL in the blastic phase of CML, which was seen with GSI treatment of K562 cells.
For the first time, these results show antagonistic interactions between the BCR-ABL and Notch signalling pathways in CD34+ chronic CML cells. Interestingly, a similar effect was not observed in IM-resistant samples. Since IM does not target Notch directly and does not influence γ-secretase activity In Vitro [42], these observations suggest that upregulation of Hes1 was a BCR-ABL-mediated effect. Resistance of CML stem cells to IM has been reported [43]. The IM resistance observed here in total CD34+ cells from two CML patients could be due to either BCR-ABL mutations on the CD34+ cells or the presence of most primitive CD34+/CD38-cells at a high percentage in CML samples compared with IM-sensitive CML samples used in this study.
GSI treatment of CD34+ primary CML cells resulted in Hes1 downregulation; however, it failed to show any effect in two CML samples (CML-1 and CML-6). Interestingly, CD34+ cells from those two CML patients who did not respond to GSI showed significant upregulation of Hes1 when BCR-ABL activity was inhibited by IM. It is therefore possible that BCR-ABL may act as a Notch repressor, and its inhibition activates downstream proteins, which then activates the transcription factor RBP-J directly and results in Notch activation while bypassing the Notch receptor [44,45]. The other possibility is that GSI non-responding cells may have an activating mutation in the Notch pathway as explained in T-ALL [46]. In T-ALL cells, inhibition of aberrant Notch signalling by GSI leads to decreased proliferation [47] and increased sensitivity to apoptosis [22], suggesting that Notch contributes to the transformation of the cells. Therefore, activating Notch mutations may also occur in CML and may be responsible for the active Notch signalling in CML.
Interestingly, Hes1 transcription factor mediator of Notch signalling was shown to maintain hematopoietic and neuronal stem cell self-renewal as well as fetal T-cell and malignant CML progenitor immaturity [17,48]. Very recently, Sukanya et al. showed that CCN3 regulates Notch signalling in chronic myeloid leukaemia [49]. In the present study, supported by In Silico analyses, we have shown for the first time the direct antagonistic effect of Notch/Hes-1 and BCR-ABL, and we have shown that inhibition of Notch signalling directly suppresses the overexpression of Hes-1 in CML primary cells, which provides a rationale for the development of combinational therapeutic strategies targeting Notch and BCR-ABL in chronic phase CML. In T-ALL, Notch positively regulates the phosphatidylinositol 3-kinase (PI3K)/AKT pathway [50]. Furthermore, it has been shown that Notch1 induces upregulation of the PI3K-AKT pathway via Hes1, which negatively controls the expression of PTEN [51]. In the present study Hes1 was up-regulated in the ALL-SIL T-ALL cell line, a cell line that has constitutive Notch and ABL kinase activities, following the exposure to IM as demonstrated. It is possible that Hes1 upregulation following IM exposure in CML cells may also activate the PI3K/AKT pathway and confer anti-apoptotic signals to CML cells regardless of the BCR-ABL repressed activity (Fig 11a). However, further work is required to investigate the activation of the PI3K pathway by Notch signalling, as reported in T-ALL, in CML cells. Although IM has been shown to inhibit BCR-ABL activity in CD34+ chronic phase CML cells, only a mild increase in apoptosis was demonstrated in these cells [33]. Moreover, it has been shown that IM treatment activated the PI3K/ Akt/ mammalian target of rapamycin (mTor) anti-apoptotic pathway in chronic phase CML patients as well as in BCR-ABL+ Lama cells [52] and proposed that the IM-induced compensatory PI3K-Akt/mTor activation may represent a novel mechanism for the persistence of BCR/ABL-positive cells in IM treated CML patients [52]. The IM induced activation of Notch signalling in the present study may suggest that, blocking the BCR-ABL activity by IM is may not enough to switch off the PI3K/AKT/mTor anti-apoptotic activity. It is also possible that the antagonistic effects between Notch and BCR-ABL signalling seen in this Both BCR-ABL and Notch activate the PI3K/AKT-mTOR pathway that may trigger the survival and proliferation signals to CML cells. Blocking BCR-ABL kinase activity may not be sufficient to induce apoptosis as this may switch to survival signals to the PI3K pathway activated by Notch in CML cells. In this model, IM may be upregulating Notch by modulating Wnt component GSK3β and/or Dishevelled. IM-induced activation of GSK3β or inhibition of Dishevelled stabilises ICN in the cytoplasm, which in turn activates the PI3K/AKT-mTOR signalling by upregulation of Hes1 which abolish the inhibitory effect of PTEN on PI3K pathway. (b) Cooperative model of activated Notch and BCR-ABL signalling in chronic phase CML. Both Notch and BCR-ABL are activated in chronic phase CML and may activate survival signalling pathways to inhibit apoptosis in CD34+ CML cells. In vitro inhibition of Notch by GSI-induced BCR-ABL activity to keep the same level of survival signals is required for CML cell. Treatment with IM leads to activation of Notch signalling to maintain the same level of survival signals needed by CML cells to inhibit apoptosis. The net effect is maintenance of balanced levels of survival signals that protect CD34+ CML cells from apoptosis in the chronic phase of CML (GSI: γ-secretase inhibitor, IM: imatinib mesylate).
doi:10.1371/journal.pone.0123016.g011 study may also involve other mechanism such as Wnt signalling pathway [12,53]. Importantly, haematopoietic progenitor cells have been reported to secrete Wnt [54] and it is possible that this pathway may therefore be active in CML. IM has been shown to inhibit Wnt signalling in CML cells [55] and in the murine myeloid progenitor cell line 32Dcl3 [56] in a way that may involve inhibition of Dishevelled and activation of GSK3β, both of which are key players in the canonical Wnt signalling pathway. Dishevelled has been reported to bind to Notch and downregulate Notch signalling in Drosophila [57]. In contrast, it has been shown in cell line models that GSK3β positively modulates Notch signalling by protecting the intracellular domain of Notch1 (ICN1) from proteasome degradation [58]. Taken together, we suggest that, IM may activate Notch signalling by modulating the Wnt components Dishevelled and GSK3β (Fig 11a) or may activate Notch signalling by blocking the inhibitory action of BCR-ABL on its downstream substrate GSK3β.
Based on our results, we propose a cooperative model for Notch-BCR-ABL activity. Coexistence of activated Notch and BCR-ABL In Vivo in chronic phase CML suggests a cooperative interaction between the two signalling pathways and that both BCR-ABL and Notch signalling are equally critical for CML cell survival and resistance to apoptosis (Fig 11b). However, one should bear in mind that in the present study, we used pharmacological inhibitors to study the antagonistic effect of Notch and BCR-ABL instead of siRNA approach. The main difference between the two approaches is that RNA interference-mediated silencing removes the target mRNA from the cell whereas pharmacological inhibition only blocks the function of a protein, but the protein still is present. This might have implications-for example, the drug-inhibited protein may lack a certain activity but may still interact with some binding partners or assemble into macromolecular complexes and may also have many off-target effects. Nonetheless, our future study will be focused to further corroborate these findings using an in-depth RNAi approach in a large group of CD34+ CML patients.
In conclusion, preclinical data presented here strongly suggest a link between Notch and BCR-ABL in CML. Microarray dataset analysis revealed that inhibition of BCR-ABL with IM led to significant upregulation of Hes1, JAG-1, and JAG-2. This change in Notch signalling profile post-IM therapy clearly shows a response to IM. Therefore, we suggest that assessment of Notch signalling pre-and post-IM therapy in chronic phase CML patients may prove to be a useful biomarker in determining the primary IM resistance. Furthermore, our results suggest that blocking of Notch signalling by GSI, or using potential anti-Notch pharmacological small molecule inhibitors such as PF-03084014 [59], RO4929097 [60,61], GSI-953 [62], and anti-Notch mAb [63] along with IM may have a role in the treatment of CML. In particular, targeting Notch and BCR-ABL simultaneously may prove superior to tyrosine kinase monotherapy in advanced CML disease, and thus, such a combined approach should be explored.