The Biology of CML Blast Crisis

The classic textbook description of the natural history of chronic myeloid leukemia (CML) is a stereotypical progression from a relatively benign chronic phase through accelerated phase, and into fatal blast crisis. This relentless march can be diverted or aborted by curative (or near-curative) therapies such as tyrosine kinase inhibitors, transplantation, and in an occasional case, interferon. All treatments work far better in chronic phase than advanced disease (accelerated or blast phases). The clinical dilemma is predicting who will go into advanced disease, and when. This is especially important in the age of tyrosine kinase inhibition (TKI) therapy. Most patients with chronic-phase CML will respond favorably to TKI; if transplantation is delayed as a salvage therapy when patients progress, however, survival outcomes plummet from chronic phase (80%) to accelerated phase (50%) to blast crisis (20%). The purpose of this review is to address the biology of blast crisis in the age of tyrosine kinase therapy. What do we know, and need to know, that will help us prevent, detect, abort and treat progression in CML?

Over the years, several schemes have been used to classify chronic, accelerated, and blast phases of CML (Table 1 ). In general these classifications rely on clinical and pathologic features, including bone marrow and peripheral blood blast counts, clonal cytogenetic evolution, splenomegaly, and response to therapy.¹ Like most classification schemes, they are approximations to account for outcomes, but are not driven by the biology of CML. Thus, the new World Health Organization (WHO) classification of blast crisis places the blast cut-off at 20%, rather than the old standard of 30%. Certainly, one would not expect any great biological differences in a case with 18% (thus, “accelerated phase”) and 20% (blast crisis). Nonetheless, these classification schemes are useful in providing uniformity for clinical trials, and offering some prognostic guideposts. However, a somewhat more enlightened view, relying on what is known about the effects of the Philadelphia chromosome (Ph), would be that of a steady progression from chronic phase to advanced disease unless a major intervention aimed at disrupting Bcr-Abl function and/or ablating the CML clone (Figure 1; see Color Figures, page 515).

Blast crisis is fundamentally different than chronic phase in many aspects. The main obvious functional changes that occur with progression of CML are marked changes in proliferation, differentiation, apoptosis, and adhesion.^2,3 These functional changes accompanying progression are accompanied by profound changes in treatment response. Despite the dramatic changes that have occurred in CML therapy, from oral chemotherapy, interferon, transplantation, and now tyrosine kinase inhibition, nothing has changed the stark fact that all treatments work far better in chronic phase than blast crisis. For example, the complete cytogenetic response rate (CCyR) for early chronic-phase patients placed on imatinib is approximately 80%; for accelerated phase and blast crisis, the value falls to roughly 40% and 20%, respectively.^4,5 For allogeneic transplantation, the same trend occurs, with approximately 80% surviving 5 years, compared to approximately 40% and 20% in accelerated and blast phases, respectively.^6,7 Why is success so common in chronic phase, and so elusive in advanced phase disease? The answer may lie in the fundamental biology of progression and how it affects treatment response.

Any model of blast crisis must account for several features of the disease: (1) the initial event in CML pathobiology is the acquisition of the Ph chromosome in stem cells; (2) in chronic-phase Bcr-Abl causes genetic instability, an increase in proliferation, and changes in differentiation and apoptosis; (3) progression to advanced-phase disease is accompanied by a severe block in differentiation and apoptosis; and (4) without the intervention of therapy, progression is inevitable. What are the likely events that cause progression, what normal pathways are subverted on the road to blast crisis, and can these pathways be restored to put patients back into chronic phase?

By definition, progression to accelerated phase and beyond is associated with genetic instability, since the definition of accelerated phase includes new clonal chromosomal abnormalities. A strong case can be made for Bcr-Abl directly causing genomic instability. The aberrant cellular activities of Bcr-Abl have been well chronicled (reviewed extensively and elegantly in Melo and Barnes⁸), including increased proliferation through the activation of ras,⁹ increased transcriptional activity via STAT recruitment,^10,11 decreases in apoptosis through activation of PI(3)K/AKT¹² and changes in adhesion binding to actin with phosphorylation to cytoskeletal proteins.¹³ While the contributions of these pathways in creating CML are clear, it is not clear that these pathways drive progression from chronic-phase to advanced-phase disease. However, insights that Bcr-Abl causes genetic instability (reviewed beautifully in Skorski¹⁴) may prove to be a unifying theme in starting the basis for progression at the very initiation of CML. When considering the effect of Bcr-Abl on genetic instability, one must confront a seeming paradox: cell lines with activated tyrosine kinases, such as Bcr-Abl, accumulate more DNA damage than those cell lines with these activated kinases, yet these activated cell lines actually repair DNA damage faster.^15,16 However, the combination of more DNA damage and more repair activity may lead to less exact repair; since Bcr-Abl also up-regulates anti-apoptosis genes BCL2 and BCL-Xl, and thus causes G₂/M delay, the stage is set to accumulate DNA damage without the mechanisms to eliminate these cells.¹⁶ Moreover, it appears that Bcr-Abl in and of itself can cause DNA damage by increasing reactive oxygen species (ROS);¹⁷ these ROS can lead to DNA base-pair transversions (GC → TA) and transitions (GC → AT). Taken together, this genetic instability may lay the groundwork for chromosomal aberrations, mutations, and changes in gene expression that hallmark progression.

An increase in Bcr-Abl level is seen in advanced-phase CML, and this increased activity affects several cellular functions. Blast phase is accompanied by an increase in both Bcr-Abl mRNA and protein level. The increase in levels are not extreme (generally approximately 3-fold for both message and protein) but this increase appears to be associated with an increase in activation of signaling (as evidenced by increased Crkl phosphorylation), and in vitro changes of clonality, growth factor independence, proliferation, and block in apoptosis.^18,–20 In mouse models, cells constructed to produce more Bcr-Abl have a more rapid speed of tumor development.¹⁹ 

Since Bcr-Abl level is associated with progression, a logical question is whether this phenomenon also plays a role in imatinib resistance, since blast crisis is relatively resistant to TKI therapy. However, the role of increasing Bcr-Abl with the imatinib resistance usually found in advanced phase disease is unclear. One model using cell lines cultured in imatinib to develop resistance suggested high Bcr-Abl was associated with a shorter time to the development of Abl point mutations. Nonetheless, a model of human CD34⁺ cells transduced with Bcr-Abl suggested that high Bcr-Abl expression was conversely associated with increased sensitivity to imatinib, suggesting that such cells were “addicted” to high levels of Bcr-Abl.^18,–20 Of note is that the mechanism of increased Bcr-Abl mRNA and protein levels in progressive disease is unclear; although some cases of blast crisis have multiple copies of the Ph chromosome, such abnormalities are not the rule.

If Bcr-Abl causes genomic instability, it is logical that unchecked Bcr-Abl will inevitably yield further genomic changes. But why did the Ph chromosome arise in the first place? Were there genetic lesions in the stem cell that created poor DNA maintenance even before the acquisition of the Ph, facilitating the incidence of the disease? Two observations suggest a stem cell disorder preceding the acquisition of the Ph. First, the Bcr-Abl transcript can be found in the peripheral blood of more than 25% of normal “older” (>55 years) individuals.^21,22 Such findings could be either from the increasing instability associated with aging, or which accumulated genotoxic insults that occur over time, and thus coincident with chronologic aging. However, there are no long-term data to test if the Bcr-Abl persists in this population. The second interesting finding is that clonal abnormalities in Ph⁻ cells are seen in approximately 5% of patients with chronicphase CML treated with imatinib, despite the maintenance of CCyR.^23,–25 This occurrence suggests that more primitive stem cells, before those acquiring the Ph, have some genetic instability, and when imatinib selects against the Ph clone, these cells gain a competitive advantage.

Chromosomal changes in addition to the Ph are the norm in blast crisis. A number of these chromosomal abnormalities are relatively common, though it is unclear if this is because their occurrence drives progression, if they have a selective advantage in the progression environment, or if they are simply more frequent because of structural elements making their recombination more common. Isochromosome i(17q) is a relatively common occurrence (approximately 20% of cases of blast crisis cases) and is interesting since it causes the loss of a copy of the p53 tumor suppressor gene. However, the remaining p53 allele does not seem to be mutated in these cases, so a direct link of p53 inactivation and progress to blast crisis is not so clear.²⁶ However, reduction of the total p53 level may upset the complex integration of genetic repair and apoptosis and contribute to progression.²⁷ Alternatively, there may be critical but yet unknown genes on 17q contributing to progression.

Trisomy 8 is also common in blast crisis (approximately 40%), Since MYC is located at 8q24, it is tempting to speculate that here MYC is driving progression. There are several lines of evidence linking MYC to progression. In vitro inhibition of c-Myc with antisense oligonucleotides, or dominant-negative constructs, can inhibit Bcr-Abl transformation or leukemogenesis.²⁸ Myc is often overexpressed in blast crisis compared with chronic phase,^29,30 while in patients with AML with trisomy 8, c-Myc is down-regulated, but other genes on chromosome 8 are upregulated.³¹ Curiously, trisomy 8 is a common feature of cases of clonal evolution in patients with CML treated with imatinib who are in cytogenetic remission (thus, these clones have trisomy 8, but not the Ph). These cases with trisomy 8 seem to have a benign course, suggesting that trisomy 8 in and of itself may not be leukemogenic.²⁴ 

Translocations of known oncogenes in blast crisis occur rarely (<5%). The most notable of these recurrent translocations are t(3;21) and t(7;11), involving the AML-1/EVI-1 and NUP98/HOXA9 genes, respectively.^32,33,EVI-1 and HOXA9 are both transcription factors, and their aberrant expression in the context of these fusion proteins causes differentiation arrest in the case of AML-1/EVI-1, and increased proliferation in the case of NUP98/HOXA9. In mouse models, BCR-ABL and AML-1/EVI-1 co-expression creates a disease produces a picture of myeloid leukemia, whereas mice that received transplants of transfected with NUP98/HOXA9 develop a myeloproliferative disease that evolves into an acute leukemia.^34,35 It is not clear, however, how often these genes are dysregulated in blast crisis without these specific translocations.

Last, in blast crisis there are additional chromosomal abnormalities that involve the Ph. “Double” Ph are seen in more than 30% of blast crisis patients.³⁶ Additional loss of chromosome 9 (der9) occurs in 5% to 10% of chronic phase patients; there are some data that these patients respond poorly to interferon, likely since this deletion eliminates critical interferon receptor genes. It is not clear if the der 9 abnormality affects prognosis for patients treated with imatinib or transplantation.^37,38 

Given that the evolution from chronic to blast phase is fairly stereotypical, it has been tempting to consider the involvement of tumor suppressors or oncogenes, especially those found in acute myeloid leukemia. Alas, the data do not support this simple hypothesis. ras was the first activated oncogene studied in acute myeloid leukemia (AML); it is mutated only rarely in CML blast crisis.³⁹ The reason for this rarity became clear with the understanding of Bcr-Abl signaling through ras; thus, no selective advantage would be gained by further activating the signal cascade. Similarly, other common mutations in AML, such as FLT3, are rare in CML blast crisis. The tumor suppressor p53, inactivated in many solid tumors, is occasionally involved in blast crisis (~20–30%); it is not known if there are aberrations in genes involved in p53-mediated function.⁴⁰ In lymphoid blast crisis (but not myeloid), a homozygous deletion of exon 2 of INK4A/ARF occurs commonly (approximately 50%).⁴¹ This deletion eliminates both p16 and p19, two proteins that normally check G₁/S cell-cycle progression and up-regulate p53.

However, changes in the activity of the tumor suppressor PP2A may be involved in the pathogenesis of CML progression and may provide a new drug target. PP2A activity is involved in regulating proliferation, survival, and differentiation, and is involved in the proteosomal degradation of Bcr-Abl. Increasing Bcr-Abl levels (induced in vitro or in human leukemia) increases the expression of the phosphoprotein SET, a negative regulator of PP2A.^30,42 Thus, progression may set up a feedback loop whereby increasing Bcr-Abl increases SET, decreasing PP2A, which further ensures the persistence of Bcr-Abl. In in vitro and mouse models, restoration of PP2A activity by the activator forskolin appeared to decrease Bcr-Abl leukemic potential, thus suggesting a potential therapeutic target to arrest or regress CML progression.⁴² 

A block in myeloid differentiation occurs in progression, contributing to the accumulation of immature blasts. A search for mutation differentiation genes has been relatively unrewarding. For example, the CCAAT/enhancer binding protein alpha (CEBPA) is a member of the basic leucine zipper family of transcription factors essential to the control of granulocytic differentiation, regulating several genes necessary for orderly differentiation, including CSFR3. Inhibition of CEBPA in t(8;21) AML occurs through repression on CEBPA by the fusion AML1-ETO protein. In blast crisis CML, high levels of Bcr-Abl induce the MAPK phosphorylation of the poly (rC)–binding protein hnENPE2, which then causes the translation block of CEBPA mRNA.^43,44 However, complete loss of CEBPA, experimental performed by tranducing Bcr-Abl into primitive murine cells engineered to have CEPBA deleted, confers an erthroid leukemia, rather than a myeloproliferative disease.⁴⁵ This implies that some residual CEPBA activity must be present, even in blast crisis, and that disturbances in proliferation in CML progression may be from subtle changes in the regulation of factors critical in controlling the flow of hematopoietic differentiation.

As noted above, changes in basic cell operations such as proliferation, differentiation, and apoptosis occur in CML disease evolution, but gross genetic changes affecting the “usual suspects” are relatively uncommon. How then can we begin to study the genetic changes associated with progression? Gene expression microarrays allow for the simultaneous study of mRNA expression of thousands of genes. This technology has been used to classify, build prognostic models and discover new genes in various solid tumors and leukemia, including CML. These studies are somewhat difficult in CML, however, because of the nature of chronic and blast phases. Some studies have focused on selected “stem cells” in order to avoid possible different expression signatures that could be associated with different cell states (i.e., blasts in blast crisis vs more differentiated cells in chronic phase). The problem with this strategy is that CD34⁺ cells from chronic phase are very infrequent. Other studies have looked at unselected samples, suggesting that the true biology of progression takes place in the entire hematopoietic system. Thus, it is not surprising that different gene lists are generated in the study of progression and resistance in CML. Nonetheless, the functional changes of progression—changes in differentiation, apoptosis, cell adhesion, and inflammatory response—remained as common themes across these studies. Several interesting genes associated with progression have a biological “track record” that makes them of particular interest to study. For example, SOCS2, thought to normally play a part in negative regulation of proliferation, has been found to up-regulated in advanced phases of CML.^30,46 Elastase (ELA2) and BMI2 have been found in microarray studies to be implicated in progression, and these may be associated with disease response in chronic-phase patients.^30,47 

Data from several sources seem to be converging on the Wnt/β-catenin pathway as critical for the evolution of CML. The Wnt/β-catenin signaling pathway has come front and center in normal and abnormal hematopoesis. Wnt/β-catenin signaling is thought to be important in cell self-renewal, and mutations in β-catenin have been found in various epithelial solid tumors.^48,–50 Aberrant Wnt/β-catenin signaling has thus been demonstrated in CML and TALL.^51,52 In CML, activation of the WNT/β-catenin pathway was observed in primary cell samples from patients with CML, with levels increasing with progression.⁵² β-catenin activation in CML seemed to reside predominately in granulocyte-macrophage progenitor cells, and the self-renewal of these progenitor could be reduced by over-expression of axin, with modulates free β-catenin.

The aberrant regulation of transcription factors can affect scores of downstream genes and hence are an efficient method to cause functional changes in cells.

Several transcription factors appear to be involved in CML progression. Jun B knock-out mice develop a myeloproliferative disorder similar to CML.^30,53 Jun B has been shown to be down-regulated in CML progression.²⁶ Genes controlled by the transcription factors MZF1 and δEF1 appear to be deregulated in CML progression.^30,53 MZF1 is a member of the Kruppel family of zinc finger proteins originally cloned from a cDNA library from a blast crisis patient with CML⁵⁴ and plays a critical role in hematopoietic stem cell differentiation, including modulation of CD34 and c-myb expression,^55,56 and MZF1^−/− knock-out mice display an increase in hematopoietic progenitor proliferation, which continues in long-term culture conditions.⁵⁵ Both MZF1 and δEF1 have been shown to influence cadherin expression.^57,–59 Moreover, MDFI, an inhibitor of myogeneic basic helix-loop-helix transcription factors found overexpressed in CML progression,³⁰ both interacts with axin and influences Jun signaling, thus perhaps linking these two pathways implicated in CML progression.^60,61 

How does one integrate the findings that progression in CML is stereotypical, but no consistent gene differences are found, despite the large catalog of genetic lesions found in the disease (Table 2 )? Imagine that in the first step in CML, the acquisition of the Ph (Bcr-Abl) causes genetic instability, changes in proliferation, differentiation and apoptosis that cause an expansion of the hematopoietic compartment (Figure 2A, 2B; see Color Figures, page 515515). The longer that Bcr-Abl is active before the initiation of therapy, the more time the cell is exposed to genomic instability. Potentially many self-renewing clones may evolve in these CML progenitors, some with point mutations, or other lesions. While there are likely many “hot spots” that are more prone to collecting genetic events (doublestrand DNA breaks, point mutations, etc.), it is likely that the ensuing genetic damage is widespread across the genome. Many genetic lesions thus will occur in genes irrelevant to hematopoietic biology, and thus may have no functional consequences. Some mutations will occur in networks with ample redundancy, or in networks that have no selective advantage (e.g., mutations that promote apoptosis are literally a dead end). Under these assumptions, only a limited number of mutations will occur in pathways that can be so disturbed, and give the mutated clone a selective advantage. Thus, the genetic bedlam accompanying genetic instability is effectively funneled into recurrent functional themes underlying progression: increases in proliferation, blocks in differentiation, inhibition of apoptosis, etc. Since changes in scores of genes in these pathways may have the same consequence, perhaps it is not surprising that the search for the “progression gene(s)” has been quite difficult.

Complete cytogenetic response (CCR) to imatinib is the rule in chronic-phase CML; in blast crisis there is an initial response in most patients, but a CCR is relatively unusual, and if it occurs, generally short-lived. Resistance is often mediated by point mutations in the Abl tyrosine kinase domain, which may be associated with a relief of kinase inhibition, or alter the signaling pathways of Bcr-Abl.^62,–64 Resistance associated with Abl mutations is uncommon in newly diagnosed chronic-phase CML, and increases in frequency as time from diagnosis, or phase, progresses.^63,65 Last, point mutations can be found in blast crisis samples never exposed to tyrosine kinase inhibitors.⁶⁴ How does the acquisition of point mutations, resistance, and progression to blast crisis relate?

The acquisition of point mutations may result from the genetic instability associated with Bcr-Abl, perhaps through the intermediates of ROS.⁶⁶ As noted above, the groundwork for the clinical observation of progression conceptually begins immediately as the stem cell acquires the Ph. Thus, the temporal relationship of frequency of point mutations with time and phase of CML is intuitive. Point mutations have been found in the CD34⁺ cells derived from patients in CCR.^67,68 Thus, even in the context of an excellent response, residual CML stem cells remain, and there is either an expansion of a resistant clone, or the development of a clone, while on therapy. Moreover, there is further evidence that progression may be slowly occurring in patients who have an excellent response. Gene expression arrays were performed on CML patients who achieved a CCR, but then had a relapse into clinical chronic phase, most with point mutations. Many of these cases had gene expression profiles more similar to advanced-phase disease.

The above model is in keeping with a mathematic model of chronic phase CML kinetics based on quantitative PCR of Bcr-Abl in 163 patients.⁶⁹ The best-fitting model had an initial rapid drop of 3 logs of disease, followed by a slow decline. The first phase is likely from the disappearance of differentiated cells, while the slower decline represents the elimination of progenitor cells. Based on relapse rates of patients in CCR, and on the data that most patients who stop imatinib relapse quickly, the model predicted that a relatively stable pool of CML stem cells persist during therapy. This agrees with in vitro data demonstrating that existing TKI does not inhibit or kill putative CML stem cells. Why CML stem cells would not be affected by TKI is not clear.

Why doesn’t TKI cure blast crisis patients? The initial response of blast crisis patients suggest that blasts are over-dependent on Bcr-Abl driven differentiation and apoptosis blockade; disruption of Bcr-Abl thus provides “pharmaceutical judo” whereby cells stumble and pour through differentiation and death. However, the remaining stem cells, still primed with aberrant pathways, either proliferate in the context of Bcr-Abl resistance or acquire Bcr-Abl independent activation.

Not well. While we have made some headway in understanding blast crisis, we are still far from our goals of being able to predict progression and effectively treat it. We are similarly unaccomplished in effectively treating blast crisis.

The fundamental emphasis of CML treatment should be to not allow blast crisis to occur. The foundation of the basic strategy of CML treatment and monitoring was covered in the previous two articles in this section. Thus, effective Bcr-Abl inhibition must be instituted early in chronic-phase disease, where efficacy of inhibition can be defined by cytogenetic and PCR monitoring. Patients who fail outcome milestones should be promptly considered for alternative secondary therapy. In patients which allogeneic related or unrelated donors, transplantation is an effective strategy, as pretreatment with imatinib does not appear to affect outcome.^70,71 Alternatively, patients may be offered secondary therapy.

Many groups are attempting to move findings of array studies of progression and response, noted above, into the arena of prognostic tests to better risk-classify cases. This is important since if the approximately 30% of chronic-phase patients who will not respond or inevitably relapse can be identified at diagnosis, they can be offered secondary TKI or transplantation at the outset.

Unfortunately, a very small minority of patients with CML will present in advanced-phase disease, while others will relapse after initial TKI therapy. What then? As noted in the previous article by Dr. Shah, second TKI agents can produce a response in advanced-phase disease, with CCyR occurring in 20% to 30% of patients. Unfortunately, these responses are generally very short-lived (with median duration of responses only approximately 3 months), and thus should be best considered as a bridge to transplantation. As noted above, disease-free survival for allogeneic transplantation ranges from approximately 40% to 50% for accelerated phase to approximately 20% for blast crisis, highlighting the need for early monitoring to predict and abort evolution to advanced phase disease. Other drug agents are being developed for blast crisis and resistant CML (e.g., the aurora kinase inhibitors), but these also do not appear to be curative;^72,73 again, they may be useful in providing a bridge to transplantation. In this regard, it is critical to emphasize that clinicians need to establish HLA typing and donor options before the patient does poorly. Reliance on an “eleventh hour” attempt at salvage with transplantation is predictably unrewarding.

With luck and hard work, effective drugs and strategies for chronic-phase CML will continue to flourish, and evolution to blast crisis will become a historic artifact, used as a model of progression in the lab, but absent from the landscape of clinical practice.

Acknowledgments

Authorship of this review is a privilege, but in no way relates to my contribution to the subject compared with the many giants of this field. In this regard, I greatly regret not being able to cite all the major contributions to this field, and I apologize if I have inadvertently omitted significant work of any of my colleagues. Many thanks to the editors of this review, who tolerated and gently corrected my understanding, biases and weak attempts at humor. I give many, many, thanks to Janine Bajus for her endless patience and understanding. Last, I highly recommend an excellent recent review of blast crisis by Melo and Barnes,⁸ which delves deeper into the molecular biology of progression than suitable for necessary succinct review.