BCR-ABL1-independent PI3Kinase activation causing imatinib-resistance

Background The BCR-ABL1 translocation occurs in chronic myeloid leukemia (CML) and in 25% of cases with acute lymphoblastic leukemia (ALL). The advent of tyrosine kinase inhibitors (TKI) has fundamentally changed the treatment of CML. However, TKI are not equally effective for treating ALL. Furthermore, de novo or secondary TKI-resistance is a significant problem in CML. We screened a panel of BCR-ABL1 positive ALL and CML cell lines to find models for imatinib-resistance. Results Five of 19 BCR-ABL1 positive cell lines were resistant to imatinib-induced apoptosis (KCL-22, MHH-TALL1, NALM-1, SD-1, SUP-B15). None of the resistant cell lines carried mutations in the kinase domain of BCR-ABL1 and all showed resistance to second generation TKI, nilotinib or dasatinib. STAT5, ERK1/2 and the ribosomal S6 protein (RPS6) are BCR-ABL1 downstream effectors, and all three proteins are dephosphorylated by imatinib in sensitive cell lines. TKI-resistant phosphorylation of RPS6, but responsiveness as regards JAK/STAT5 and ERK1/2 signalling were characteristic for resistant cell lines. PI3K pathway inhibitors effected dephosphorylation of RPS6 in imatinib-resistant cell lines suggesting that an oncogene other than BCR-ABL1 might be responsible for activation of the PI3K/AKT1/mTOR pathway, which would explain the TKI resistance of these cells. We show that the TKI-resistant cell line KCL-22 carries a PI3Kα E545G mutation, a site critical for the constitutive activation of the PI3K/AKT1 pathway. Apoptosis in TKI-resistant cells could be induced by inhibition of AKT1, but not of mTOR. Conclusion We introduce five Philadelphia-chromosome positive cell lines as TKI-resistance models. None of these cell lines carries mutations in the kinase domain of BCR-ABL1 or other molecular aberrations previously indicted in the context of imatinib-resistance. These cell lines are unique as they dephosphorylate ERK1/2 and STAT5 after treatment with imatinib, while PI3K/AKT1/mTOR activity remains unaffected. Inhibition of AKT1 leads to apoptosis in the imatinib-resistant cell lines. In conclusion, Ph+ cell lines show a form of imatinib-resistance attributable to constitutive activation of the PI3K/AKT1 pathway. Mutations in PIK3CA, as observed in cell line KCL-22, or PI3K activating oncogenes may undelie TKI-resistance in these cell lines.


Background
Expression of the Philadelphia chromosome (Ph), resulting from fusion of the non-receptor tyrosine kinase ABL1 on chromosome 9 with BCR on chromosome 21, is the hallmark of chronic myeloid leukemia (CML), but is also found in 20-30% of acute lymphoblastic leukemia (ALL) cases. The development of clinically applicable tyrosine kinase inhibitors (TKI) has fundamentally changed the treatment of patients with CML: imatinib mesylate induces hematologic remission in nearly all CML patients [1]. In Ph+ ALL, imatinib is much less effective [2].
Many studies performed to elucidate imatinib-resistance have made use of cells ectopically expressing BCR-ABL1 or of cell lines which gained resistance after prolonged exposure to rising drug concentrations [6,7]. Cell lines that were inherently imatinib-resistant have rarely been used, which is astonishing because imatinibresistant cell lines KCL-22 and SD-1 were described very early, in 1997 [12]. Here, we screened the DSMZ cell lines bank to find imatinib-resistant BCR-ABL1 positive cell lines. Five out of 19 Ph+ cell lines (26%) were resistant to imatinib. We set out to investigate whether these cell lines displayed the known molecular and cellular causes for imatinib-resistance.

Results and Discussion
Imatinib-resistant BCR-ABL1 positive cell lines  Table 1). These cell lines were inherently resistant to imatinib according to the results of proliferation and apoptosis assays, as they had not been preincubated with the TKI.

BCR-ABL1 mutations, BCR-ABL1 expression, imatinib transporters
Point mutations in the kinase domain of BCR-ABL1 are the main cause of imatinib-resistance in the chronic phase of CML [13]. Although second generation BCR-ABL1 inhibitors (nilotinib, dasatinib) are effective in most BCR-ABL1 mutated cases, all 5 imatinib-insensitive cell lines identified here were also resistant to nilotinib suggesting that resistance might not be caused by BCR-ABL1 mutations (Table 1). In accordance with this notion, genomic sequencing showed no sequence alterations in the kinase domain of the resistant cell lines ( Table 1).
The DNA-binding protein Ikaros is a major regulator of lymphoid development [14]. Deletion of Ikaros is found in the majority of BCR-ABL1-positive ALL and of CML in progression to lymphoid blast crisis [15,16]. Public genomic array data indicate hemizygous loss of the 7p12 region in cell line NALM-1, including IKZF1 and the neighbouring gene Dopa decarboxylase (DDC) http://www.sanger.ac.uk/cgi-bin/genetics/CGP/ 10kCGHviewer.cgi?chr=7&dna=NALM-1. Genomic PCR analysis confirmed loss of IKZF1 in this cell line, but not in cell lines SD-1, SUP-B15 and MHH-TALL-1 (Additional File 1). However, the majority of Ph+ ALL with IKZF1 aberrations do not show deletion of the whole gene, but instead intragenic loss of various IKZF1 exons, leading to the expression of mRNA variants that mimic "normal" splice variants [15,16]. A recent publication correlates expression of the Ikaros variant Ik6 with high BCR-ABL1 mRNA levels and imatinib-resistance in Ph+ ALL [17]. We could not confirm this correlation among Ph+ ALL and CML cell lines: Ik6 was expressed in 2/19 BCR-ABL1 positive cell lines, one being imatinib sensitive (BV-173) and one resistant (SUP-B15) ( Table 1). Neither cell line SUP-B15 nor most other TKI-resistant cell lines showed particularly high BCR-ABL1 expression levels according to quantitative RT-PCR analysis ( Table 1). The only exception was cell line KCL-22 with about 2-fold higher BCR-ABL1 expression levels, both at the mRNA (Table 1) and the protein level (not shown). While supporting the notion that a causative correlation might exist between the high expression of the mutated kinase and imatinibresistance for cell line KCL-22, these results also showed that in 4/5 cell lines TKI-resistance was not the consequence of BCR-ABL1 overexpression (Table 1).
Thus, neither BCR-ABL1 mutations nor overexpression of the kinase were the general cause for imatinibresistance in these cell lines. Further analyses showed that also dysregulation of drug transporters was improbable: unlike imatinib, nilotinib is neither imported via hOCT-1, nor exported via ABCB1 [5,18]. All five imatinib-resistant cell lines were nilotinib-resistant (Table 1). Therefore, it appeared unlikely that imatinib-resistance was caused by deregulated transport proteins. Finally, the finding that both imatinib and nilotinib induced dephosphorylation of signal transducer and activator of transcription 5 (STAT5) in the TKI-resistant cell line SUP-B15 as shown in Figure 2 further excludes resistance being due to low intracellular drug levels. Both drugs were transported into the cells which responded by dephosphorylating STAT5 while retaining viability.

SRC kinases
SRC kinases had been described to play an important role in BCR-ABL1 positive ALL ( Figure 3) [9,[19][20][21]. Interestingly, 4/5 imatinib-resistant Ph+ cell lines were from patients with pre-B ALL, T-ALL, or CML in B-cell blast crisis (Table 1). Among lymphoid Ph+ cell lines 5/7 were imatinib-resistant, including TOM-1, a pre-B cell line classed "semiresistant" displaying normal IC50 values in the thymidine uptake assay while remaining relatively unresponsive to higher concentrations (Table 1). Therefore, we applied dasatinib to elucidate whether activity of SRC kinases was important for the growth of imatinibresistant cells. Dasatinib is a dual BCR-ABL1 and SRC kinase inhibitor, as evidenced by its ability to inhibit phosphorylation of SRC and STAT5 in TKI-responsive JURL-MK2 cells (Additional File 2). However, two of three imatinib-resistant cell lines tested (NALM-1, SUP-B15) were resistant to dasatinib in the proliferation assay (Table 1). Furthermore, TKI-resistant SUP-B15 cells did not express an active, phosphorylated SRC kinase and dasatinib did not affect RSP6 phosphorylation in this cell line (Additional File 2). These results are not consistent with the notion that SRC kinases are the cause of imatinib-resistance in these cell lines.
In TKI-sensitive cells, imatinib induced dephosphorylation of all three proteins ( Figure 4). In TKI-resistant cell lines, treatment with TKI reduced phosphorylation of STAT5 (5/5 cell lines) and of ERK1/2 (4/5 cell lines) but did not comparably affect phosphorylation of RPS6 (Figure 2 Figure 4). This observation allowed three conclusions: (i) cells that survive in the presence of imatinib are not necessarily completely unresponsive to the drug; (ii) activation of ERK1/2 and the JAK/STAT5 pathway is not obligatory for short-term proliferation of Ph+ positive cell lines; (iii) TKI-resistance is correlated with -if not actually caused by -the constitutive and imatinibresistant activity of the PI3K/AKT1/mTOR pathway.

BCR-ABL1-resistant cell lines show constitutive activation of mTORC1
The PI3K/AKT1/mTOR/p70S6kinase (p70S6K) pathway is a BCR-ABL1 downstream target and implicated in the survival of leukemic cells ( Figure 3) [23,24]. A major difference between TKI-sensitive and -resistant cell lines was seen with respect to the phosphorylation level of the p70S6K substrate RPS6: incubation with imatinib inhibited RPS6 phosphorylation in TKI-responsive, but  not -or to a much lesser degree -in TKI-resistant cell lines (Figure 2 Figure 4). p70S6K is an exclusive substrate of mTOR complex 1 (mTORC1). Rapamycin inhibits this complex, but not mTORC2 [25]. Recent studies suggest that targeting mTOR might become an efficient anti-cancer therapy [25]. Rapamycin arrests Ph+ K-562 cells in the G1 phase of the cell cycle and induces apoptosis in primary CML cells [26]. Antileukemic effects of rapamycin in patients with TKI-resistant CML have been shown [27]. These results prompted us to test whether rapamycin inhibits constitutive RPS6 phosphorylation, whether it reduces cell growth of TKI-resistant CML cell lines and -most importantly -whether the combination of rapamycin and imatinib induces apoptosis in imatinib-resistant cells.
Rapamycin effected dephosphorylation of RPS6 in imatinib-sensitive and imatinib-resistant cell lines ( Figure 5). Rapamycin alone did not induce apoptosis in imatinibresistant cell lines, as evidenced by annexin-V staining (Figure 6A). However, in 6/6 cell lines, rapamycin (10 nM, for 24 h) reduced [ 3 H]-thymidine uptake, which was paralleled by an increase in the percentage of G1-phase cells ( Table  2). For multiple myeloma, it has been shown that an antiproliferative drug, the CDK4/6 inhibitor PD0332991 can sensitize cells to a second agent, a cytotoxic drug (bortezomib) [28]. Therefore, we speculated that rapamycin and imatinib might cooperate in a similar way, rapamycin acting as growth inhibitor and imatinib as cytotoxic agent. The combination of rapamycin plus imatinib had the same inhibitory effect on phosphorylation of RPS6 and of STAT5 in TKI-resistant cells as imatinib alone had in TKIsensitive cells ( Figure 5). However, the combination of imatinib and rapamycin did not lead to a significant increase of apoptotic cells in imatinib-resistant cells, compared to the effects of each drug alone ( Figure 6A). Thus, inhibition of mTORC1 was insufficient to restore responsiveness in TKI-resistant cell lines.

AKT1, mediator of imatinib-induced apoptosis
As shown in this study, 2/3 BCR-ABL1 downstream signalling cascades -the JAK2/STAT5 and the ERK1/2 pathways -are druggable by TKI in imatinib-resistant  Figure 3 BCR-ABL1 signalling cascades. BCR-ABL1 induces activation of JAK2/STAT5, RAS/RAF/ERK and PI3K/AKT1/mTOR signalling pathways. SRC kinase family members may interact with BCR-ABL1 leading to mutual activation. The scheme has been adopted [11,18,19]. RPS6 is a downstream target of AKT1/mTOR/ p70S6K. cell lines (Figure 4). The PI3K/mTOR pathway was not comparably inactivated by imatinib, as assessed by RPS6 phosphorylation (Figure 2 Figure 4). These results imply that TKI-resistance is caused by constitutive TKI-unresponsive activation of the PI3K/mTOR pathway. However, rapamycin -despite efficiently dephosphorylating RPS6 -failed to induce apoptosis, whether alone or in combination with imatinib ( Figure 6A). Therefore, we concluded that another member of the PI3K pathway, upstream of mTOR might confer resistance, inhibiting imatinib-triggered apoptosis. It has been shown in another experimental setting that the inhibition of the serine-threonine kinase AKT1 sensitizes tumor cells to apoptotic stimuli [29]. AKT1 stimulates proliferation by activation of mTORC1, and suppresses apoptosis by phosphorylation of proapoptotic proteins like BCL2associated agonist of cell death (BAD) (Figure 3). We inhibited AKT1 with Akt inhibitor IV, as evidenced by dephosphorylation of RPS6 (Additional File 3). Inhibition of AKT1 triggered apoptosis in imatinib-sensitive and -resistant cell lines ( Figure 6B). These data suggest that AKT1, rather than mTOR is the PI3K pathway member that should be inhibited to trigger apoptosis in TKI-resistant cells.

Role of PI3Ka in imatinib-resistance in Ph+ cell lines remains elusive
In this study we show that imatinib-resistance of Ph+ cell lines may be ascribed to the TKI-insensitive activation of the PI3K/AKT1/mTOR pathway. Although other BCR-ABL1-triggered signalling cascades (ERK1/2, JAK2/ STAT5) proved to be imatinib-responsive, inhibition of these pathways did not affect the viability of cells. In contrast to imatinib, wortmannin (PI3K inhibitor), OSU-03102 (PDK1 inhibitor) and rapamycin (mTOR inhibitor) inhibited the PI3K/AKT1/mTOR pathway, suggesting that the TKI-resistance observed in the Ph+ cell lines might be caused by a PI3K-activating oncogene other than BCR-ABL1 itself (Additional File 4). To identify this oncogene we looked for mutations and aberrant expression of genes known to mediate activation of PI3K, such as RAS, CBL and p85 (Figure 3). In addition, PI3K itself was a candidate for genetic alterations causing constitutive activation of the PI3K/AKT1 pathway. RAS mutations occur quite frequently in hematologic malignancies (5% K-RAS, 12% N-RAS) [30]. However, none of the TKI-resistant cell lines showed mutations of the most affected regions of the genes (amino acids 12, 13 and 60 in K-RAS, amino acids 12, 13, and 61 in N-RAS; data not shown), a finding which was scarcely unexpected because RAS mutations would not only stimulate PI3K, but also ERK1/2 in an imatinib-insensitive manner ( Figure 3). However, ERK1/2 was silenced by imatinib in 4/5 cell lines (Figure 2 Figure 4).
The PI3K subunit p85b (PIK3R2) and the Casitas B-Cell lymphoma gene (CBL) belong to those seven genes identified as core components for coordinating the oncogenic functions of BCR-ABL1 [21]. Phosphorylation of CBL recruits the p85 subunit of PI3K leading to activation of PI3K/AKT1/mTOR pathway [31]. Quantitative RT-PCR did not reveal major differences in the expression of CBL and p85 between imatinib-sensitive and -resistant cell lines (data not shown). Besides, we did not detect alterations in exons 7-9 of CBL, described as transforming mutations in myeloid malignancies (data not shown) [31,32].
Class I PI3Ks are heterodimeric proteins consisting of a catalytic and a regulatory adaptor subunit (e.g. p85β). To find out which specific PI3K might be involved in imatinib-resistant activation of AKT1/mTOR, we applied inhibitors with differing specificities for the various PI3K catalytic subunits. [ 3 H]-thymidine incorporation data suggested that PI3Kα (PIK3CA), but not PI3K β or PI3Kγ play a role in the imatinib-resistance of the cell lines tested (Figure 7). Mutations occurring in the catalytic subunit PIK3CA result in constitutive activation and oncogenicity [33]. The majority of PIK3CA mutations occur either in the helical (exon 10) or in the kinase domain (exon 21) of the gene [34]. Thus, we sequenced the respective regions of PIK3CA (database number ENST000000263967) in all imatinib-resistant cell lines. We did not find mutations in the kinase domain, but cell line KCL-22 carried a heterozygous point mutation in the helical domain, leading to the amino acid change PI3Kα E545G (Figure 8). PI3Kα E545 mutations have been observed in clinical samples of solid tumors and the E545A mutation has been shown to constitutively activate the PI3K pathway [33].
These data suggest that also the PI3Kα E545G mutation that we identified in cell line KCL-22 may be responsible for the constitutive activity of the PI3K/ AKT1 pathway conferring TKI-resistance to the cells. Deep sequencing might help to elucidate whether activating mutations in oncogenes other than BCR-ABL1 or PIK3CA, or loss of tumor suppressor genes trigger the PI3K in cell lines NALM-1, SD-1, SUP-B15 and MHH-TALL1, thus causing TKI-resistance.

Conclusion
In this study an unexpectedly high number (5/19, 26%) of Ph+ ALL and CML cell lines tested imatinib-resistant. The unresponsiveness of the cell lines was not attributable to known causes as BCR-ABL1 mutations or activation of SRC kinases. While the BCR-ABL1-triggered Table 2 Effect of rapamycin on proliferation of BCR-ABL1positive cell lines JAK2/STAT5 and ERK1/2 pathways were inhibited by imatinib, the resistant cell lines stand out by the constitutive activation of the PI3K/AKT1/mTOR pathway. The mTOR inhibitor rapamycin inhibited cell growth, but did not induce apoptosis and did not sensitize resistant cells to imatinib. Instead, inhibition of AKT1 induced apoptosis in TKI-resistant cell lines. Cell line KCL-22 carries a heterozygous mutation in the helical domain of PIK3CA, a site critical for activation of the gene. These results suggest that activating mutations in the PI3K itself or in PI3K-stimulating oncogenes might be the molecular cause for TKI resistance.

Human cell lines
The cell lines applied in this study were taken from the stock of the cell bank (DSMZ -German Collection of Microorganisms and Cell Cultures) or were provided by originators. Detailed references and cultivation protocols have been described previously [35].

Quantitative real-time PCR analysis
Quantitative PCR was performed on a 7500 Applied Biosystems (Darmstadt, Germany) real-time PCR system using the manufacturer's protocol. RNA was prepared using the RNeasy Mini kit (Qiagen, Hilden, Germany). For mRNA quantification, reverse transcription was performed using the SuperScript II reverse transcriptase kit (Invitrogen, Karlsruhe, Germany). Expression of BCR-ABL1 e1-a2 and b2-a2 fusion mRNAs, of CBL and of p85β (PIK3R2) were assessed using the SYBR GREEN PCR Master Mix (Applied Biosystems) with GAPDH as internal control. BCR For detection of Ikaros splice variant 6 (Ik6), we performed PCR using the following primers: Ikaros exon 2 forward: 5'-ATG GAT GCT GAT GAG GGT CAA GAC-3'; Ikaros exon 8 reverse: 5'-GAT GGC TTG GTC CAT CAC GTG G-3'. The PCR was performed with an annealing temperature of 62°C. Splice variants were detected by electrophoresis on a 1.2% agarose gel and verified by sequencing of the PCR products.

Western blot analysis
Samples were prepared as described previously [38]. The anti STAT5 monoclonal antibody (mAb) was purchased from BD Transduction Laboratories (Heidelberg, Germany). Anti pSTAT5, anti pRPS6 and anti pSrc (Tyr416) antisera as well as the monoclonal antibody directed against RPS6 were purchased from Cell Signalling (New England Biolabs, Frankfurt, Germany). Anti FYN and anti LYN antisera were obtained from Santa Cruz (Heidelberg, Germany). The anti GAPDH mAb was purchased from Abcam (Cambridge, UK). Specific bands on nitrocellulose membranes were visualized with the biotin/streptavidinhorseradish peroxidase system (Amersham, Freiburg, Germany) in combination with the "Renaissance Western Blot Chemoluminescence Reagent" protocol (Perkin Elmer, Waltham, MA, USA).