• Open Access

Molecular biology of chronic myeloid leukemia


To whom correspondence should be addressed.

E-mail: ymaru@research.twmu.ac.jp


Detailed information on the crystal structure of the pharmacologically targeted domains of the BCR-ABL molecule and on its intracellular signaling, which are potentially involved in growth, anti-apoptosis, metabolism and stemness, has made the study of chronic myeloid leukemia the most successful field in tumor biology. However, we now face the issue of drug resistance due to deregulation in the quality control of both DNA and protein. BCR-ABL is basically a misfolded protein with intrinsically disordered regions, which not only produces endoplasmic reticulum stress followed by unfolded protein response in some settings, but also conformational plasticity that may affect the structure of the whole molecule. The intercellular signaling derived from the leukemic cell microenvironment may influence the intracellular responses that take place in a manner both dependent on and independent of BCR-ABL tyrosine kinase activity.

Chronic myeloid leukemia (CML) clinically consists of two phases. The chronic phase is characterized by the expansion of myeloid progenitor cells with apparently normal differentiation. Several years later, this is followed by the acute leukemic progression phase called blast crisis. CML is a somatic mutation disease whose diagnostic hallmark is the Philadelphia chromosome, an abnormally short chromosome 22 resulting from a reciprocal chromosome translocation t (9;22)(q34;q11).[1] The translocation juxtaposes the ABL gene on chromosome 9 with the BCR gene on chromosome 22, thereby generating the chimeric BCR-ABL mRNA.[2, 3] The BCR-ABL protein is a constitutively activated form of the ABL tyrosine kinase and has been extensively studied in mouse models. It is currently believed to be the cause of CML.

Treatment of CML using the tyrosine kinase inhibitor (TKI) imatinib can efficiently reduce the leukemic cell load in the chronic phase, but in this context, there is now the issue of drug resistance, which leads to a failure to prolong the remission period before blast crisis. Increased levels of BCR-ABL protein are observed in blast crisis. PP2A agonists can restore G-CSF-stimulated C/EBP-dependent myeloid differentiation capacity by allowing BCR-ABL protein degradation.[4] BCR-ABL is the target of imatinib; however, BCR-ABL with the T315I gatekeeper mutation in the ABL kinase domain, currently the most representative imatinib-resistant mutation, does not have the ability to be its pharmacological receptor. This raises a fundamental question about how homeostasis of protein and DNA is deregulated in CML.

Chronic myeloid leukemia was thought to be a stem cell disease long before the establishment of the cancer stem cell (CSC) concept. Imatinib can inhibit BCR-ABL in CML stem cells, but it fails to kill them because they are not addicted to BCR-ABL.[5] At a CML meeting held in Bordeaux in 2009,([6]) the organizer held a vote on the issue of whether or not CML stem cells should be eliminated for treatment. The majority of attendees agreed with this notion, with the author being in the minority. It can safely be said that the causative molecule for CML needs to be identified and targeted each time pharmacological ineffectiveness occurs (Fig. 1). The idea of a “cure,” the goal of clinical treatment, is discussed by Jerald P. Radich.[7]

Figure 1.

Chronic myeloid leukemia (CML) is a stem cell disease. A concept for possibly curing CML may be repeated elimination of progenitor cell expansion derived from CML stem cells. This could provide an acceptable quality of life (QOL) during the human lifespan. Normally, differentiation occurs in the direction from hematopoietic stem cells (HSC) to granulocyte-macrophage progenitors (GMP). Imatinib resistance may take place at the level of GMP that might differentiate back to HSC. T315I, BCR-ABL mutant at Thr315 to Ile (see details in the text).

Protein quality control plays a pivotal role in many biological settings. Structural alterations of proteins could take place through translation errors, damage by reactive metabolites, such as reactive oxygen species (ROS), and other mechanisms in a stochastic fashion. Proteins generated in pathological settings, such as chromosomal reciprocal translocation, are non-physiological and misfolded from the beginning. This article is based on the supposition that BCR-ABL is still misfolded even if chaperoned by Hsp90 for refolding, and that newly synthesized BCR-ABL proteins are immature and unfolded before they become folded to acquire enzymatic activity.[8] In addition, intrinsically disordered regions can be predicted. They confer regional structural plasticity for functions such as binding sites. A yeast ubiquitin ligase San1 recognizes misfolded nuclear protein by its intrinsically disordered sequence.[9]

Blast crisis has additional cytogenetic abnormalities in more than 80% of patients. The mutator phenotype of BCR-ABL was initially described by Pierre Laneuville, showing that P210 BCR-ABL expression by itself induced new karyotypic abnormalities within a few months.[10] However, it remains unresolved whether T315I was generated by BCR-ABL from the onset of CML or after imatinib treatment. Alternatively, the true cellular alteration (Factor X) responsible for CML coexists with BCR-ABL from the onset and the phantom oncogene BCR-ABL might just be supporting it before it causes blast crisis (Fig. 1). In mouse models, BCR-ABL certainly causes leukemia. However, we have observed that lentivirus-mediated transduction of CD34+ marmoset cells with BCR-ABL failed to induce leukemia even if BCR-ABL expression was maintained for 18 months.[11]

The biology of CML is analogous to that of metastatic tumors in at least two essential aspects: the clinical mode of progression and the acquisition of therapeutic resistance due to stem cell-like properties in some populations. The clinical manifestations of CML involve a clear spectrum of disease progression. In this context, this review should provide some insight by making a comparison between CML and metastatic tumor progression.

Protein Structure

We cannot accurately discuss the structure of BCR-ABL unless we have information on the crystallization of BCR-ABL as a whole molecule. It should be noted that the number in the term T315I refers to the 315th T in human-type Ia cABL (Genbank accession no. M14752), but not BCR-ABL.[12] This is somewhat confusing for non-BCR-ABL researchers. For example, in a sophisticated paper by Grebien et al.[13] on ABL structure, the mutant name BCR-ABL I164E is used. The number 164 refers to the human-type Ib cABL that was crystallized. Because the number in type Ia cABL is I145, here all the ABL numbers are simply referred to using T315I.

The catalytic domain of protein kinase is well conserved and consists of the N-lobe and the C-lobe (Fig. 2). The activation loop, especially the conserved Asp-Phe-Gly (DFG) motif, in the N-lobe, and the αC helix in the C-lobe make essential contributions to an on–off switch related to the conformational status of “ready-to-phosphorylate substrates.” This plasticity within this globular domain is executed by tyrosine phosphorylation at 393 in the activation loop.[14, 15] Binding of imatinib to the kinase domain is achieved through six hydrogen bonds, one of which is disrupted by T315I. The second-generation TKI, dasanitib or nilotinib, can partially restore the TKI sensitivity in imatinib-resistant patients, but not in those with T315I. Third-generation TKI ponatinib can make contact with 315I and, therefore, is effective against T315I (Fig. 1).

Figure 2.

Presumed molecular structure of BCR-ABL. Intramolecular inhibitory folding through the interactions between domains is shown with dotted lines, including binding between: (1) the SH3 domain and the SH2 linker; (2) the SH2 and the tyrosine kinase domains; and (3) the myristoyl group in the amino-terminal ABL and the tyrosine kinase domain. The inhibitory folding in the inactive state in cABL is released in BCR-ABL, allowing autophosphorylation in the activation loop at Y393 and the interaction between: (1) the SH2 domain and the N-lobe; and (2) the SH2 ligands and the SH2 domain (see details in the text).

The cABL tyrosine kinase activity is negatively regulated by the myristoyl group attached at G in the N-terminus, which is missing in BCR-ABL but is inserted into the pocket in the C lobe of the cABL kinase domain.[16] The interface between the SH3 domain and the linker region connecting the SH2 and kinase domains also provides negative regulation (Fig. 2). In an inactive state of ABL, the SH2 domain is docked into the C-lobe to prevent it from binding to SH2 ligands. In an activated state of BCR-ABL, the inhibition through the SH3 domain and the linker region is released and then the SH2 domain not only binds the N-lobe via tight interactions between I145 and T272/Y312 but also accepts SH2 ligands,[17] which may include the BCR first exon-encoded sequence (Fig. 2). In fact, I145E mutation abolishes BCR-ABL activity. Therefore, the SH2 ligand-dependent activation and structural changes can be considered to be like the chicken and the egg. In addition to the ATP analogs as represented by imatinib, non-ATP analog inhibitors have been developed on the basis of these allosteric mechanisms.[18]

The BCR first exon-encoded region exhibits high disorder that gives structural plasticity to BCR-ABL (Fig. 3). For example, the SH2-binding domain is essential for transformation by BCR-ABL, presumably by intramolecular folding over to the ABL SH2 domain.[19] Despite the critical contribution of BCR in BCR-ABL activation, unfortunately, neither sensitive nor resistant mutations within the BCR sequence in CML patients have been studied or reported.[20]

Figure 3.

Signaling in BCR-ABL. The IUPred analysis of BCR-ABL along with each domain.[103] Potentially globular regions are indicated by blue boxes and the rest of the BCR-ABL sequences are intrinsically disordered regions. Information on the cABL last exon-encoded sequence (LX) is not shown.[19] Major signaling pathways and their biological activity in chronic myeloid leukemia (CML) and normal hematopoiesis are shown. αapo, anti-apoptosis; Cata, catabolism; eryth, erythropoiesis; myel, myelopoiesis; Pro Stab, protein stability; SH2BIND, the ABL SH2-binding domain; Tran, transformation; Warb, Warburg effect.

The tetramer domain at the N-terminus of the BCR first exon-encoded sequence is essential for converting inactive cABL to constitutively active BCR-ABL. This fragment has been crystallized to reveal a coiled-coil structure that tetramerizes ABL.[21] The mechanism mimics ligand-induced dimerization of receptor tyrosine kinase as represented by EGFR. A TKI-resistant mutation, T790M, is analogous to T315I in BCR-ABL.[22]

The crystal structure of the Dbl homology (DH) domain of Sos that has 67% similarity to the BCR DH domain has been successfully obtained.[23] Different from the DH domain of Dbl, CDC24, or Sos, that of BCR has a disordered portion within this domain (Fig. 3), which may convey plasticity for binding to other molecules.


The clinical manifestations of CML involve massive expansion of myeloid progenitor cells without differentiation block. Detailed quantitative PCR analysis of sorted hematopoietic stem cell (HSC), granulocyte-macrophage progenitor (GMP) or common myeloid progenitor (CMP) population of cells from CML patients revealed that the level of BCR-ABL mRNA was high in HSC in the chronic phase, but high levels of expression were also observed in GMP and CMP in blast crisis.[24] An increase in GMP was also observed in a mouse CML model.[25]

BCR-ABL exerts a variety of biological activities, such as cell proliferation, anti-apoptosis, cell migration and unique metabolism (Fig. 3). The critical pathways elicited by the constitutively active ABL TK include Ras-MAPK (Erk), PI3K, RhoA-Rac, and transcription factors Myc and Stat5. Tyrosine phosphorylation of Shc and Y177 in BCR, whether intact BCR protein or BCR sequence within BCR-ABL, activates Ras by recruiting the Grb2-SOS complex, leading to MAPK (Erk) and subsequent D1 cyclin activation. Retroviral transfer of active MEK into mouse Lin-cKit + Sca1+ (LSK) cells, which is highly enriched in HSC, and transplantation into irradiated RAG2-KO mice was found to give rise to a myeloproliferative disorder with enlarged spleen containing Gr1 + Mac1 + myeloid cells.[26] Activated Ras is reported to result in ROS production.[19] ROS production in BCR-ABL-transformed cells was initially shown by James D. Griffin's lab.[27] In physiological circumstances, such as cytokine signaling by interleukin-3 (IL-3), granulocyte colony-stimulating factor (G-CSF) or granulocyte-macrophage colony-stimulating factor (GM-CSF), ROS is also produced and functions as a second messenger. G-CSF-stimulated ROS production appears to involve Nox2. Although the activity of conventional PKC (α, β, γ) is reported to be unchanged in BCR-ABL-expressing cells, the activity of a novel form, PKCδ, was increased and its dominant-negative form or its specific chemical inhibitor, but not those against conventional PKC, suppressed transformation by BCR-ABL.[28]

The PI3K-Akt pathway is activated indirectly by Ras, but the phosphorylated YXXM motif found in BCR-ABL directly binds the p85 subunit, whose mutation failed to change PI3K activity in BCR-ABL-expressing cells. cCbl is heavily tyrosine-phosphorylated in BCR-ABL-expressing cells and recruits the p85 subunit of PI3K to exhibit high PI3K activity.[29] However, cCbl is found to be dispensable for BCR-ABL-dependent transformation in a transplantation CML model.

Although mutations are frequently found in Akt and phosphatase and tensin homolog (PTEN) in a variety of tumors, those in myeloid neoplasm, including blast crisis in CML, are rare. However, a transplantation model of artificial expression of constantly activated myristoylated Akt or conditional PTEN-KO mice showed both myeloid and lymphoid malignancy with HSC driven into the cell cycle for differentiation and, therefore, exhausted without changing the level of ROS.[30] Restoration of the stem cell phenotype by rapamycin back to cells from these artificial conditions indicates the presence of the PI3K-Akt-mTOR pathway. Akt also regulates FOXO family transcription factors, some of which are essential in stem cell biology (see Microenvironment). Other Akt substrates of biological significance include Bad, MDM2 and p27. Phosphorylation of BAD impairs their pro-apoptotic activity derived from mitochondria, MDM2 inducing degradation p53, and p27 de-inhibiting negative regulation of the cell cycle.

Generally, a catabolic pathway is activated in tumor cells and the glucose transporter GLUT1, whose expression is stimulated by Akt and hypoxia-inducible factor 1 (HIF-1), participates in glycolysis and glucose addiction. Activation of glycolysis, which is usually executed in hypoxic conditions, is observed in tumor cells even in normoxic conditions (Warburg effect). This is supported by recent metabolome analysis of tumor versus non-tumor tissues in colon cancer.[31] Glucose-dependent ROS production has been shown in BCR-ABL-expressing cells, which is mediated by the PI3K-Akt pathway.[32] In tumor cells transformed by activated tyrosine kinase oncogenes, PK M2 is allosterically downregulated by phosphotyrosine-containing peptides.[33] In acute myeloid leukemia cells expressing FGFR fusion oncogenes, which intriguingly include BCR-FGFR1, the clinical features of which resemble CML, pyruvate dehydrogenase (PDH) kinase is inactivated by tyrosine phosphorylation.[34] Inhibition of both PKM2 and PDH blocks pyruvate entry into the tricarboxylic acid cycle and eventually promotes the Warburg effect.

The Akt-mTOR pathway stimulates the expression of active HIF-1α, mimicking hypoxic responses in normoxic conditions.[35] This explains the increased expression of vascular endothelial growth factor (VEGF) in CML. Nuclear factor kappa B (NFkB) activation is observed in CML blasts and is not only required for transformation by BCR-ABL but also linked to possible production of growth factors in CML cells. Inhibition of NFkB by the IKKβ inhibitor AS602868 results in apoptosis of T315I-expressing cells.[36] Mechanistically, Raptor-dependent reciprocal activation between mTOR and IKK and partial contribution of Ras are proposed.[37] Bidirectional activation between NFkB and HIF-1α or autocrine mechanisms involving TNFα and endogenous ligands for toll-like receptor 4 (TLR4) are also assumed in vivo (see Microenvironment).

Cytoplasmic Src family proteins participate in BCR-ABL signaling. Hck works in concert with BCR-ABL to phosphorylate Stat5. In Stat5a-deficient mice, GM-CSF-induced myeloid cell growth was abrogated. Jak2 is activated by cytokines like IL-3, GM-CSF and IL-6. Jak2-KO cells failed to respond to IL-3, GM-CSF and IFNγ.[38] The Stat5 target genes include anti-apoptotic genes, such as Bcl-2 and Bcl-XL. Cytochrome c release is impaired by the blocking of these gatekeepers. Even after cytochrome c is released, it was claimed that BCR-ABL inhibits binding between Apaf-1 and Caspase9,[39] exerting anti-apoptotic effects in multiple modes.

Reciprocal activation between Jak2 and BCR-ABL results in protein stability in BCR-ABL and Lyn-mediated stimulation of the PI3K-Akt pathway.[40] A non-ATP-competitive inhibitor for Jak2, ONO44580, has also been shown to be effective in T315I without directly targeting T315I.[41]

Myc, a regulator of G1-S transition, is shown to cooperate with and be required for transformation by BCR-ABL.[19] The Myc induction is mediated by BCR-ABL-activated Jak2. Conditional Myc-KO shows poor myelopoiesis and Myc transgenic mice show high self-renewal of HSC.[42] In cooperation with HIF-1α, Myc also upregulates GLUT1 for catabolic pathway activation.

The function of the BCR first exon-encoded sequence is controversial.[19] A highly purified GST-tagged BCR first exon-encoded portion of BCR from baculovirus showed strong autophosphorylation activity.[43] Mutation of C332 resulted in prominent abrogation of the kinase activity in baculovirus as well as mammalian cells.[19] Radziwill et al.[44] show that, at the plasma membrane, BCR phosphorylates and associates with AF-6 to downregulate Ras signaling. The serine-threonine-rich region has the ability to bind not only the ABL SH2 domain (Fig. 3) but also 14-3-3τ, 14-3-3β and CKII. Through 14-3-3β, this region may interact with Raf. The growth-promoting and anti-apoptotic activity of CKII, the expression of which is increased in many tumors, led to the development of its inhibitor DMAT, which shows an anti-leukemic effect on transgenic BCR-ABL cells.[45]

P210BCR-ABL activates Rac1, which is an activating component of Nox2 in myeloid cells. The DH domain of BCR, which is retained in P210, is capable of exerting guanine nucleotide exchange factor activity for RhoA.[46] Replacement of the BCR DH domain with that of Dbl or CDC24 fails to confer full biological activity on BCR-ABL.[47] Thus, the DH domain provides ROCK-dependent survival and transformation, although PI3 kinase activation is also located upstream of RhoA-ROCK.[48] We and another group have demonstrated the interaction between the DH domain and xeroderma pigmentosum group B protein (XPB), an essential component of the TFIIH complex involved in nucleotide excision repair (NER) (see Mutator phenotype).[19]

Quality Control

Quality control of cellular proteins is a double-edged sword. It participates in refolding of unfolded or misfolded proteins, but also leads to failed refolding or proteolytic degradation. At least two quality control systems are well known: proteasome and lysosome (Fig. 4). Typically, ubiquitination confers acute and specific protein degradation in proteasome, whereas lysosomal catabolism by autophagy, except for Hsc70-mediated cargo, is performed in a non-specific and chronic manner.

Figure 4.

A proposed protein quality control scheme in chronic myeloid leukemia. Basically, the BCR-ABL protein is a pathologically misfolded protein, even if it becomes folded or mature after translation to acquire enzymatic activity. BCR-ABL immediately after translation may be unfolded or immature and censored by Bag1, ubiquitinated by CHIP, and subjected to proteasomal degradation unless folded properly and chaperoned by Hsp90. Once folded, BCR-ABL can bind ATP or imatinib. Macro-autophagy cascade is suppressed by the PI3K-Akt-mTOR pathway.[104] The cascade is simplified and shown only with representative molecules. See details in the text. ER, endoplasmic reticulum; ERAD, ER-associated degradation; UPR, unfolded protein response.

Many tyrosine kinases have been observed to associate with Hsp90 and cCbl-mediated degradation,[49, 50] including BCR-ABL, EGFR, VEGFR1 and c-KIT. Imatinib inactivates the activity of BCR-ABL, but stabilizes its protein structure. We have found that dose-dependent suppression of BCR-ABL-expressing leukemic growth by Hsp90 inhibitors like geldanamycin (GA) is blunted by an increasing dosage of imatinib and that BCR-ABL protein degradation by GA is counteracted by imatinib.[8] Mutations of c-Kit lead to its own upregulation and can also be targeted by imatinib. Imatinib has been shown to abrogate the binding between Hsp90 and c-KIT. Secondary mutations in c-KIT are also sensitive to Hsp90 inhibition.[51] In retrovirus-mediated CML model mice carrying T315I, an Hsp90 inhibitor, IPI-504, induced BCR-ABL protein degradation and decreased the number of CML stem cells more effectively than in wild-type BCR-ABL.[52] JAK2 mutated at the region homologous to a region of ABL was found to be more susceptible to an Hsp90 inhibitor, AUY922, than non-mutated JAK2.[53] These findings suggest that additional mutations in the kinase domain might increase the level of misfolding to surpass the refolding or chaperoning activity of Hsp90.

The pharmacological activator of PP2A induces proteasomal degradation of T315I.[4] Considering the collaboration between PP2A and PIN1 to induce changes in serine-proline configuration in the tau protein implicated in Alzheimer's disease,[54] PP2A might affect protein structure and act on the same pathway as Hsp inhibition (Fig. 4).

We assume that Hsp90 acts as a chaperone for a folded or mature form of BCR-ABL. Imatinib-bound or ATP-bound BCR-ABL may be stabilized in its folded form[8] and no longer need stabilizing support from Hsp90 (Fig. 4). In some conditions, ROS cleaves and inactivates Hsp90, resulting in BCR-ABL degradation. This can be canceled by N-acetyl cysteine (NAC). Consistent with this, we have previously shown that NAC can promote leukemic growth of P210BCR-ABL-expressing cells.[55]

In addition to BCR-ABL, Hsp90 clients of biological significance include HIF-1α and mutated forms of p53, the most frequent additional genetic event in blast crisis, Akt and mTOR. It is noteworthy that, even under normoxic conditions, BCR-ABL can upregulate active HIF-1α, which is usually activated by hypoxia (see Signaling section).

Hsc70 might participate in triage decisions for unfolded proteins immediately after translation. We have shown that Hsp90 inhibition swings the balance in favor of the unfolded form of BCR-ABL, which binds Hsc70.[8] If the refolding process cannot be accomplished, a triage decision is reached and unfolded or immature BCR-ABL is censored by Bag1 and ubiquitinated by CHIP for proteasomal degradation (Fig. 4).

Constitutive elevation of Hsp70, which is usually inducible under stress, such as nutrient shortage and oxidative stress, is found in many tumors. It is assumed that it contributes to stabilization of a variety of abnormal proteins in tumor cells. A selective Hsp70 inhibitor, 2-phenylethynesulfonamide, was found to disrupt Hsp70 and Bag1M and to induce caspase-independent tumor cell death accompanied by protein aggregation, impaired autophagy and lysosomal function.[56] Specific dual silencing of Hsp70 and Hsc70 by siRNA was also found to result in tumor cell apoptosis.[57]

Autophagy appears to be a double-edged sword for tumor cells. The front edge suppresses tumor progression by reducing genomic instability. Beclin +/− mice have been shown to develop spontaneous tumors with aneuploidy, necrosis and inflammation. However, mTOR negatively regulates autophagy (Fig. 4) and mTOR inhibitors are now in clinical use for the treatment of metastatic renal cell carcinoma (Fig. 5). As stated in the Signaling, the mTOR inhibitor rapamycin can block cell proliferation in imatinib-resistant CML patients.[58] Resveratrol also de-inhibits autophagy, causing autophagic cell death in K562 cells by activating AMPK, which blocks mTOR, and by inducing expression of p62 via JNK.[59] This is an autophagy stimulation strategy that presumably occurs in cells that survive in an autophagy-independent fashion and is apparently based on the front edge. When autophagy is activated by trehalose before EGFR-targeted diphtheria toxin treatment in tumor cells, the dying cells are shown to release HMGB1 without necrosis and membrane lysis.[60] HMGB1 is an alarmin that constitutes a danger signal (see Microenvironment).

Figure 5.

A schematic representation of tumor metastasis by analogy to chronic myeloid leukemia (CML). Shown on the left is hypoxic responses and NFkB activation in normoxic conditions in metastatic renal cell carcinoma (mRCC) mostly deficient in VHL.[105] This affects the tumor microenvironment as represented by resident cells such as osteoclast progenitor cells expressing VLA4, which binds VCAM1 that is expressed in tumor cells in an NFkB-dependent manner. A similar situation is found in CML on the right. However, CML cells express VLA4, which binds VCAM1 that is expressed in stromal cells. In both cases, the VLA4–VCAM1 interaction has anti-apoptotic effects on tumor cells and metastatic progression in the bone (see details in the text). VEGF, vascular endothelial growth factor. PIGF, placental growth factor; VHL, von Hippel–Lindau.

In CML, in contrast, the autophagy inhibitor chloroquine is in clinical trials because imatinib-treated BCR-ABL activates autophagy in stem cells for their survival.[61] BCR-ABL-specific activation of autophagy in imatinib-treated growing cells is preceded by endoplasmic reticulum (ER) stress; the mechanism involved is explained by a sudden decrease of survival signal derived from BCR-ABL.[61] However, the combination of TKI and autophagy inhibitor showed dramatic inhibition in LTC-IC assay, which reflects the potency of primitive CML HSC, with CD34+ cells from CML patients, in the presence or absence of a growth factor cocktail. BCR-ABL proteins in CML stem cells can be degraded by Hsp90 inhibitors.[52] Even in the presence of Hsp90 inhibitors, ATP or imatinib can stabilize BCR-ABL, which can no longer be censored by Bag1, but retains the nature of a misfolded protein.[8] Therefore, the survival signal might be the autophagy induced by ER stress that is elicited by the persistent presence of misfolded BCR-ABL; this could be intensified by imatinib-bound stabilization of the misfolded BCR-ABL. In the misfolded PML-RAR chimeric protein in acute promyelocytic leukemia (APL) with a reciprocal chromosome translocation t(15;17) the misfolded protein promotes secondary misfolding of N-CoR, which is accumulated in ER to cause ER stress.[62] Although this idea is interesting, it remains to be investigated whether sequential misfolding takes place from the starter BCR-ABL in CML.

The seemingly paradoxical effects of these anti-neoplastic therapies are considered to be based on whether or not the target cells are addicted to oncogenes. CML stem cells are not addicted to BCR-ABL. Cancer stem cells are highly potent in terms of the ability to start colonization in metastatic organs. However, given that cells can migrate to distant organs before oncogene activation.[63] CSC may also be less dependent on oncogenes. Therefore, the back edge of autophagy, with an oncogenic property, needs to be blocked at the level of stem cells or cells that are addicted for survival to autophagy but not oncogenes. Ulk-1 was found to be a client of the Hsp90-cdc37 complex chaperone in K562 cells.[64] Therefore, Hsp90 inhibition eventually inactivated autophagy (Fig. 4). GA and novobiocin interact with the N-terminal or C-terminal region of Hsp90 to exert negative effects. Novobiocin reduced CML CD34+ cells in colony-forming unit assays that represent committed progenitors. In a murine retroviral transplantation model of CML, an Hsp90 inhibitor, IPI-504, decreased the number of leukemic LSK cells both in vitro and in vivo.[52]

Stem Cells

Tyrosine kinase inhibitor failed to kill BCR-ABL-expressing CML stem cells that were not addicted to it. PanHDAC inhibitors combined with TKI were shown to induce apoptosis of primitive HSC in CML patients and in a mouse model.[65] However, given the broad spectrum of genes targeted and the modification of non-histone proteins, the mechanisms behind these findings remain to be uncovered. This includes the deacetylation and inactivation of Hsp90 by HDAC6.

Critical features of physiological HSC are quiescence or growth arrest, glycolytic metabolism and autophagy, self-renewal, and repopulating ability. Here, cancer or leukemic stem cells are defined as cells capable of achieving engraftment to recapitulate the original tumor in a serial manner. HSC have been shown to be maintained by hypoxia and the Tie2-Ang-1 system.[66, 67] Hypoxia also stimulates autophagy through an increase of the AMP/ATP ratio that activates AMPK, which takes place independently of HIF-1α.

The levels of actual oxygen tension in bone marrow from normal volunteers and AML patients are 7.2% and 6.1%, respectively.[68] The physiological level of pO2 is assumed to be even lower in stem cell niches, as suggested by favorable ex vivo cultures of HSC under 1–3% oxygen. Tie2+ LSK cells that represent quiescent HSC were positive for the hypoxic cell marker pimonidazole in vivo.[66] Proline hydroxylase activity of PHD2 is highly dependent on oxygen. Its substrate HIF-1α is hydroxylated in this way in normoxic conditions, which allows HIF-1α to bind the ubiquitin E3 ligase complex von Hippel–Lindau (VHL) and to be subsequently ubiquitinated for proteasomal degradation (Fig. 5). In hypoxia, de-inhibition of HIF-1α activates transcription of 100–200 genes. For example, the ABC transporter BCRP-1 is expressed in HSC, which is experimentally useful for identifying the HSC population using Hoechst dye-efflux. The essential role of regulated expression levels of HIF-1α in the maintenance of quiescence has been shown by experiments using VHL-KO and conditional KO of HIF-1α.[66] Given that Hsp90 inhibitors can promote HIF-1α degradation in a VHL-independent fashion and that Hsc70 plays an essential role in stabilizing cell cycle inhibitor p27 in stem cells,[69] we need more information on the quality control of proteins in stem cells. In an MMTV-Py MT mouse breast cancer model, CD90+ CD24+ cells behaved like stem cells in metastatic lung colonization, which was dependent on Wnt signaling.[70] The Wnt-β-catenin signaling causes self-renewal in stem cells by activating expression of Myc and D1 cyclin. BCR-ABL stabilizes β-catenin by tyrosine phosphorylation.[71] The intact BCR has been reported to bind the β-catenin-Tcf complex and, thereby, inhibit its nuclear entry. Although β-catenin is also localized in centrosomes and regulates mitotic spindles, we could hardly observe β-catenin interaction with either BCR-ABL or BCR in mitotic aster assembly assays (see Mutator phenotype).[72]

Chronic myeloid leukemia (CML) stem cells are reported to fail to home in on their niches when adhesion molecule CD44 is disrupted. The absolute number of CML HSC was similar to that of normal HSC. Gene expression profiling of chronic phase CML HSC revealed a resemblance with that of normal CMP with, for example, downregulated adhesion molecules, such as CD44 or NR4A1, implicated in quiescence, which may facilitate detachment from the niches.[73] Normal stem cells are sensitive to CXCL12 (SDF-1), the expression of which is controlled by HIF-1α, in terms of cell migration. Transplanted HSC may home in on stem cell niches along the CXCL12 gradient. The stimulation of its receptor CXCR4 activates both CDC42 and Rac. In BCR-ABL-expressing cells, the tyrosine kinase domain activates CDC42 and Rac through phosphorylation of Vav1, and this desensitizes the responses to CXCL12.[74]


Here, we discuss cell–cell communications but not signaling within tumor cells. Surrounding cells include endothelial cells, macrophages and so-called bone marrow stromal cells (Fig. 5).

One of the well-known clinical manifestations of CML is splenomegaly. Murine HSC from bone marrow and spleen are shown to be similar in their capacity for repopulating and self-renewal. The extramedullary growth of leukemic cells in spleen is analogous to metastatic tumor growth in distant tissues, where metastatic organs prepare the microenvironment for coming tumor cells. The organotropism is determined by the properties of both seed (tumor cells) and soil (microenvironment).[75] Expansive growth of CML progenitor cells in CML spleen is considered to be one of the extremes of organ-specific metastasis.

Growth factors such as VEGF, transforming growth factor beta (TGFβ) and tumor necrosis factor alpha (TNFα) are required for the establishment of the metastatic microenvironment.[75] For example, TGFβ is shown to regulate Akt activation to control Foxo3a function in CML HSC.[76] Lung metastasis is abrogated in TNFα knockout mice.[77] mRNA levels of VEGF, placental growth factor (PlGF) and IL-6, but not TNFα, are upregulated in CML bone marrow (Fig. 5) and anti-PlGF antibody is reported to prolong survival of CML mice and inhibit lung metastasis.[78, 79] BCR-ABL-expressing leukemic cells by themselves produce VEGF, but stromal cells stimulated by CML cells via VCAM1-VLA4 (α4β1) signaling can also express PlGF. Its sole receptor is VEGFR1, which promotes the migration of myeloid cells. Bone marrow-derived VEGFR1-expressing myeloid cells are claimed to participate in the formation of a pre-metastatic niche in the lungs. Mating between transgenic CML mice and VEGFR1 TK-KO failed to change the CML phenotype.[75]

Bone marrow transplantation (BMT) of BCR-ABL-expressing cells to PlGF-KO mice prolonged the survival of the mice. VLA4 expression is upregulated in normal CD34+ stem cells. VLA4 on CML cells binds to VCAM1 on bone marrow stromal cells and causes NFkB activation to induce expression of VEGF, PlGF and IL-6 (Fig. 5). VCAM1 in metastatic tumor cells binds VLA4 in monocytic osteoclast progenitor cells.[80] The binding awakens indolent tumor cells to allow colonization in the post-metastatic microenvironment in both lung and bone.

NFkB signaling plays a critical role in TNFα and TLR4 signaling in tumor metastasis. Both TNFα-KO and TLR4-KO backgrounds provide a poor microenvironment for metastatic tumor cells.[75] Given that S100A8 is one of the most abundant proteins in myeloid cells and has been shown to be an endogenous ligand for TLR4 by us as well as by a German group, high turnover of CML progenitor cells may give high and sustained NFkB activation in bone marrow. The contribution of this inflammation-like state cannot be ignored. IL-6 is secreted from cells that are activated by lipopolysaccharide (LPS) or endogenous ligands for TLR4 that is expressed in LSK and GMP cells.[81] IL-6 has recently been shown to affect myeloid differentiation in CML.[82] NFkB induces Cox2, the key enzyme for prostaglandin (PG) synthesis. Once activated, the Cox2-PGE2-EP2-NFkB-Cox2 auto-amplification loop is established. Epidemiological studies reveal not only tumor-suppressive effects of NSAID in many tumors, including leukemias, but also a correlation between Cox2 overexpression and metastasis.[83] Indeed, in CML, inflammatory myeloid cells by themselves are tumor cells. It is asserted that the acquisition of metastatic potential can be explained by tumor cell education in distant tissues by inflammatory stimuli and homing back to the primary site.[84] Although it is not clear that this self-seeding process involves genetic alterations, CML cells might undergo self-education by passing back and forth between bone marrow and spleen.

The two extremes of cell death are apoptosis and necrosis, as can be judged by their distinct morphological changes. Turnover of 100 billion leukocytes per day involves spontaneous apoptosis. Imatinib-treated CML cells also show apoptosis. Anti-tumor chemotherapy causes fortuitous necrosis, which is sometimes accompanied by tissue destruction with hemorrhage. Intracellular proteins, such as HMGB1 and Hsp, which are called danger-associated molecular pattern (DAMP), are released to activate TLR to elicit a danger signal in the innate immune system.[85] Programmed necrosis is also reported in chemotherapy, or TGFα-treated cells, in which RIPK1 and 3 activation causes apoptosis-inducing mitochondrial ROS production.[86]

Mutator Phenotype

Deterioration of DNA homeostasis is found in imatinib-resistant or blast crisis phase cells, including epigenetic changes, such as aberrant DNA methylation[87] and genetic changes involving specific genes like p53 and BCR-ABL itself, as represented by either T315I, which retains its tyrosine kinase activity, or 35INS, which lacks it due to a functional deletion of the kinase domain and the downstream sequences.[88] The 35INS change might be recognized as doubly misfolded BCR-ABL resulting in ER stress (Fig. 4). We have observed that BCR-ABL lacking the kinase domain is rather resistant to Hsp90 inhibitors (Tsukahara and Maru, unpublished).

Fabarius et al.[89] found that imatinib induces centrosomal deterioration. Polo-like kinase 1 (Plk1), the master regulator of mitosis, is overexpressed in a BCR-ABL-dependent manner.[90] Its inhibitor, BI2536, induces its degradation in proteasome, suggesting stabilization by Hsp. Although imatinib can reduce the expression level of Plk1, changes in the fine tuning that controls mitotic spindles may have deleterious effects on DNA homeostasis. We have shown that LZTS2 transport from centrosome to midbody is dependent on the Plk1-docking protein Mklp-1, and that katanin transport in the same pathway is dependent on LZTS2, which inhibits the microtubule-severing activity of katanin.[72] LZTS2 can suppress transformation by vABL.[72]

The acquisition of the mutator phenotype of BCR-ABL has been studied from the perspective of both “attack” and “defense” (Fig. 6). Genetic mutations are supposed to originate from irreparable damage due to “attacks,” such as overproduction of ROS and/or deficient homeostasis in DNA.

Figure 6.

Relationship between attackers and defenders in DNA and protein homeostasis. The relationship is affected by extrinsic factors, such as irradiation and chemicals, and intrinsic factors, including reactive oxygen species (ROS) generation, gene expression, and BCR-ABL tyrosine kinase activity-dependent or activity-independent protein modifications. It is noteworthy that the BCR-ABL protein by itself is misfolded from the beginning (see details in the text). mit, mitochondria; pol, polymerase.

What is the source of the “attacker” mutagen? Extrinsic mutagens include alkylating reagents, such as busulfan, which was previously used for CML treatment. Intrinsic mutagens include all the proteins involved in DNA homeostasis that are altered in terms of both quality and quantity by transcription or protein modification. PAX5-mediated B cell-specific activation-induced cytidine deaminase (AID) expression clearly but only partly explains the mutator phenotype in B lymphoid crisis because it is originally involved in somatic hypermutation and class switch recombination of Ig genes; however, double KO of PAX5 and AID still leads to the development of imatinib resistance in myeloid crisis.[91] It is intriguing that TLR4-induced NFkB activation is also required along with PAX5 for AID expression (see Microenvironment).[92]

Another intrinsically active “attacker” is ROS, such as superoxide or hydrogen peroxide H2O2. The presence of a feed-forward cycle in ROS production between Nox and mitochondria has been suggested.[93] Targets of ROS include not only DNA but also proteins and lipids. For example, direct modification of the IKK complex and p65RelA has been shown in NFkB signaling. Even redox-modulated protein misfolding of HLA-B27 implicated in spondyloarthropathy and p53 in Alzheimer's disease are reported.[94] Therefore, the resulting accumulation of oxidative misfolding products may serve as a passive “attacker.”

TLR4 activation in HSC and GMP induces not only upregulation of Nox2 and iNOS but also the mitochondrial anti-oxidant enzyme MnSOD, suggesting mitochondrial ROS generation. TNFα secretion upon TLR4 activation also participates in ROS production. Moreover, coupling between TLR4 and Nox4 has been shown.[75] Irradiation and chemotherapy cause Nox4-mediated persistent increase in ROS in murine LSK cells, resulting in their senescence.[95] ATM-dependent sustained DNA damage response to irreparable lesions causes NFkB-dependent secretion of IL-6 and IL-8, both of which play a positive role in tumor progression (see Microenvironment).[96] H2O2 is converted from superoxide. All forms of peroxiredoxins (Prx) are upregulated in a variety of tumors, possibly for eliminating the growth factor receptor-activated generation of H2O2. PrxI was initially isolated from K562 cells and shown to bind TLR4 to induce secretion of IL-6 and TNFα.[97] Collectively, the DNA damage-induced danger signal establishes the tumor microenvironment.

“Defenders” include protein quality controllers and DNA repair proteins that are regulated by these controllers. In BCR-ABL-expressing cells, the catalytic subunit of DNA-PK, which is involved in non-homologous end-joining (NHEJ), is documented to be downregulated. Both chromosomal and centrosomal aberrations caused by the deterioration of the Fanconi anemia pathway in BCR-ABL-expressing CD34+ cells can be rescued by BRCA1, the expression levels of which are decreased, or by proteasome inhibitors.[98] Furthermore, tyrosine phosphorylation of Rad51 and the Werner syndrome helicase (WRN) by BCR-ABL, and increased expression of DNA polymerase β and alteration in the mismatch repair system are reported.[19, 99] We have shown that the DH domain of P210BCR-ABL tightly binds XPB and inhibits its function.[19] XPB is a component of the TFIIH complex that plays an essential role in NER and transcription. XPB also controls the expression levels of C1D.[100] C1D can activate DNA-PK by serving as DNA ends. The yeast homolog of C1D binds to the hinge region of condensin and is considered to be involved in condensation, not only in the mitotic phase, but also in interphase, by cleaning away proteins and mRNA from the unwound ssDNA.[101] cABL can promote degradation of proteins that bind to damaged DNA in a kinase-independent manner.[102] Imatinib-bound BCR-ABL might influence this phenomenon in a dominant negative fashion after entry into the nucleus.


This work was partly supported by a Grant-in-Aid for Scientific Research from the Japanese Government (No. 21117008).

Disclosure Statement

The authors have no conflict of interest to declare.