Therapy of pediatric ALL: from Bowie to Obama

In this issue of Blood, van der Veer et al report that the Ph-like/BCR-ABL1-like phenotype and IKZF1 deletions, but not CRLF2 alterations, are independent negative prognostic factors in a large study encompassing 4 cohorts of childhood acute lymphoblastic leukemia (ALL).¹ 

Changing therapy based on identifying new prognostic factors has been a major theme in pediatric ALL treatment. In the past 5 years, genomic profiling of ALL has identified several recurrent genomic alterations with prognostic import. These include IKZF1 deletions² ; the entity of Ph-like or BCR-ABL1-like ALL with a gene expression profile similar to that of BCR-ABL1–positive ALL and a range of genetic alterations activating kinase signaling pathways^2,3 ; dysregulation of the cytokine receptor gene CRLF2⁴ ; and Janus kinase (JAK) 2 (or occasionally JAK1) point mutations, which occur almost exclusively in cases with CRLF2 genomic alterations.^5,6 

Literature concerning the potential prognostic significance of these alterations continues to grow, and van der Veer and colleagues have added significant new information. However, the field remains confusing. The data on IKZF1 are the most consistent. Since our original description of IKZF1 deletions and adverse prognosis in 2009,²  the majority of studies have found that IKZF1 deletions were an independent adverse prognostic factor in pediatric ALL. However, the genetic lesions listed above are often overlapping, and many studies have not studied important genomic features, including all CRLF2 alterations (IGH-CRLF2, P2RY8-CRLF2, CRLF2 F232C, and CRLF2 expression) and the mutational status of IKZF1 and JAK1/2. Moreover, outcome differences are influenced by risk group, treatment, and ethnicity.⁴  In addition, the definition of Ph-like ALL varies^2,3 : we described Ph-like ALL using formal gene expression prediction algorithms⁷  and a highly overlapping group identified by clustering of the most variably expressed genes. Van der Veer and colleagues used a historic signature of Ph⁺ ALL and hierarchical clustering to identify Ph-like cases. Both identify poor prognosis cases, but with incomplete overlap.^1-3,7  Ongoing sequencing efforts are expanding the range of kinase-activating rearrangements and mutations in Ph-like ALL. These alterations are largely limited to Ph-like cases, and the majority are known or predicted to be responsive to currently available tyrosine kinase inhibitors (TKIs).

So, what does all this mean, and which alterations should be identified to refine risk stratification and tailoring of therapy? We think that focusing on the prognostic import of individual lesions is missing the forest for the trees. Pediatric leukemia doctors have spent decades identifying different risk factors and using them to define risk groups and allocate patients to therapies of different intensities. This strategy has contributed to major improvements in outcome, such that 5-year survival for pediatric ALL now exceeds 90%. However, the rate of improvement is decreasing, and the strategy of refining prognostic factors and optimizing conventional chemotherapy is unlikely to bring cure rates to the desired 100%. Our goal should be to identify patients with specific genomic lesions that can be targeted therapeutically, not on cataloging an ever-expanding lexicon of risk factors that even the cognoscenti cannot agree upon and have trouble remembering.

There are only a few true revolutionary changes that have occurred in pediatric ALL therapy in the past 20 years, one of which is the dramatic improvement in cure rates for Ph⁺ ALL treated with ABL1 TKIs, first reported in 2009.⁸  These results have been stable over time and confirmed by others.⁹  Together, these results indicate that TKIs can be safely combined with chemotherapy in childhood ALL, and that stem cell transplant is not essential to cure most children with Ph⁺ ALL. These data provide a compelling rationale to implement logically directed TKI therapy in Ph-like ALL cases harboring defined kinase-activating alterations that may be targeted with currently available TKIs.

The critical issue is how to implement detection of the Ph-like phenotype, IKZF1 deletions, CRLF2 deregulation, and other lesions driving Ph-like ALL to rapidly direct patients to receive appropriate targeted therapy. Ph-like ALL cases lacking CRLF2 deregulation commonly have rearrangements deregulating ABL1, PDGFRB, and JAK2, among other kinases. These fusions are sensitive to TKIs in vitro and in ALL xenografts,⁷  and there are emerging data of clinical responses of refractory childhood Ph-like ALL to appropriate TKI therapy (eg, EBF1-PDGFRB–positive ALL¹⁰ ). We need to develop and test targeted therapies tailored to patients harboring specific genetic alterations, not simply refine conventional therapies. It is time to test ABL1 class TKIs in patients with ABL1 or PDGFRB fusions (and perhaps others). There is a strong biological rationale to test JAK2 TKIs (such as ruxolitinib) in patients with JAK2 fusions, but there are not yet enough data to know how to optimally combine this agent with chemotherapy. Efforts should focus on gathering these data quickly. An unresolved question is why IKZF1 deletions are associated with such poor outcomes. Is restoring normal IKZF1 activity also a logical therapeutic strategy? More preclinical work is needed to answer this critical question.

Our patients challenge us to move from describing adverse prognostic factors to developing and testing new targeted therapies. That is a change we can all believe in.