Genetic drivers of vulnerability and resistance in relapsed acute lymphoblastic leukemia

Relapsed acute lymphoblastic leukemia (ALL) is associated with very poor outcomes despite modern therapies in both children and adults (1–3). In PNAS, Oshima et al. elucidate both the mutational landscape as well as patterns of clonal evolution of pediatric relapsed ALL and identify potential therapeutic targets in this challenging illness (4). Although prior studies have performed SNP array analysis (SNPa), whole-exome sequencing (WES), whole-genome sequencing (WGS), and/or mRNA sequencing (RNA-seq) at diagnosis and relapse (Table 1), the majority of prior analyses have focused on relapsed pediatric B-ALL specifically. Here, the authors performed WES of 55 pediatric ALL patients (33 T-cell ALL and 22 B-cell precursor ALL) at diagnosis, remission, and relapse and complemented these data with RNA-seq of a validation cohort of 49 paired diagnosis/relapse B-ALL patient samples. This analysis has allowed for reevaluation of the pattern of clonal evolution in relapsed ALL and revealed that the majority of relapsed ALLs (∼85%) contain only some of the genetic lesions present in the major clone at diagnosis. In contrast, only in 4% of cases did the relapsed clones contain all mutations present at diagnosis plus additional secondary relapse-specific lesions. Together, these data identify that linear evolution is rarely involved in tumor progression in ALL and that relapsed ALL originates primarily as derivatives of ancestral subclones related to, but distinct from the main leukemic population present at diagnosis. This conclusion extends previous observations of the clonal basis of relapsed ALL by Mullighan et al. (5), which used SNPa to reveal that cells responsible for relapse were often minor subpopulations of the cells responsible for initial disease (Table 1).

Table 1.

Genomic analysis of relapsed ALL

Population	Diagnosis, remission, relapse (scenarios studied)	Genomic analysis technique*	Patients (N)	Selected findings	Ref.
Pediatric B-ALL	Diagnosis, relapse	SNP arrays	61	Relapsed ALL most frequently represents selection of clones ancestral relative to the diagnostic ALL clone	5
Pediatric t(12;21) B-ALL	Diagnosis, relapse	SNP arrays	18	Recurrent NR3C1 and BMF mutations and mismatch repair defects	18
Pediatric B-ALL	Diagnosis, relapse	SNP arrays	20	EBF1 and IZKF1 deletions enriched at relapse	19
Pediatric high hyperdiploid B-ALL	Diagnosis, relapse	WES	19	CREBBP mutations and KRAS mutations enriched at relapse	8
Pediatric B-ALL	Diagnosis, remission, relapse	Targeted gene sequencing	60	Recurrent somatic mutations in SETD2, CREBBP, MSH6, KDM6A, and MLL2 at relapse	20
Pediatric B-ALL	Diagnosis, remission, relapse	WES/WGS	20	Enrichment of NT5C2, CREBBP, WHSC1, TP53, USH2A, NRAS, and IKZF1 mutations in relapsed disease	21
Pediatric T- and B-ALL	Relapsed	WES	138	NT5C2 mutations at relapse	7
Pediatric B-ALL	Diagnosis, relapse	RNA-seq	10	NT5C2 mutations at relapse	6
Pediatric T- and B-ALL	Diagnosis, remission, relapse	WES and	55 (WES)	Enrichment of NT5C2, TP53, NR3C1, CREBBP, and MLL2 as well as N/KRAS mutations at relapse	4
		RNA-seq	49 (RNA-seq)

^*WES, whole-exome sequencing; WGS, whole-genome sequencing.

In addition to evaluating the clonal basis of relapsed ALL, the authors identified several new genes enriched at relapse including ZFHX3, USP9X, CACNA1H, EPHA3, SHROOM3, USP7, RPGR, HTR3A, MED12, ODZ3, and IL17RA. Mutations associated with relapsed ALL appear to fall into several categories. First, there are mutations functionally related to the mechanism of action of chemotherapy used in initial treatment and/or previously linked to chemotherapy resistance. These include mutations in NT5C2, TP53, NR3C1, CREBBP, and MLL2. Ultradeep sequencing of diagnostic samples failed to identify these mutations in most cases, which suggests they were acquired at relapse and/or present in very small subclones at diagnosis. Given the need to develop novel targeted therapeutics for relapsed ALL, identification of therapeutically targetable genetic alterations enriched in relapse is critical. To this end, prior work from this group and others identified that mutations in the 5′-nucleotidase enzyme NT5C2 enhance inactivation of nucleoside-analog chemotherapeutic agents (6, 7). The frequency of NT5C2 mutations at relapse and the fact that NT5C2 mutations confer enhanced enzymatic activity argue for efforts to develop small-molecule inhibitors of mutant NT5C2.

Mutations in CREBBP and P300, which encode acetyltransferases, have also been previously linked to glucocorticoid resistance in ALL (8, 9). Although previously suggested to confer loss of function, continued annotation of mutations in CREBBP and P300 actually demonstrates that these mutations usually occur as heterozygous mutations at specific residues in the region encoding the enzymatic domain in each of these genes. Thus, investigation of the mechanistic effects of hot-spot mutations in CREBBP and P300 are clearly needed. Similarly, it will be important to study exactly how alterations in CREBBP or P300 are associated glucocorticoid resistance. Finally, increasing efforts to study how deletion and loss-of-function mutations create therapeutic vulnerabilities in relapsed ALL will be important. For example, frequent deletions of the CDKN2A/B locus may also involve deletion of the adjacent gene 5-methylthioadenosine phosphorylase (MTAP) and render cells sensitive to PRTM5 inhibition (10, 11). Thus, efforts to link recurrent copy number (CN) changes enriched in ALL relapse, such as CDKN2A/B, IZFK1, and EBF1 deletions, with therapeutic vulnerabilities may be important.

In addition to previously described relapse-associated mutations and CN changes, the authors also noted enrichment of mutations in genes involved in MAP kinase signaling (8, 12), including NRAS (24%), KRAS (20%), and PTPN11 (4.5%) mutations. Altogether, 48.5% of all cases had a mutation known to activate MAP kinase signaling at relapse. In addition, mutations of N/KRAS were found in 34% of diagnostic high-risk ALL cases here. These data suggest the importance of MAPK pathway activation in high-risk B-ALL as well as relapsed B- and T-ALL. However, given that RAS mutations are known to be frequently subclonal in leukemias (13, 14), consideration of the allelic frequency of these mutations in ALL will be important. Consistent with this concern, the authors discovered that some patients with RAS mutations at diagnosis sometimes relapsed with RAS wild-type clones or with a RAS-mutant clone that was not present at diagnosis.

Given the enrichment of RAS mutations in relapsed ALL identified here, the authors further evaluated a potential relationship between activation of MAP kinase signaling and therapeutic resistance in ALL. They first generated a T-ALL isogenic model by overexpressing constitutively activated NOTCH1 in bone marrow cells from KrasG12D knockin or Kras wild-type mice. Interestingly, these Kras mutant cells had increased resistance to methotrexate (MTX) yet demonstrated enhanced sensitivity to vincristine compared with wild-type counterparts. Moreover, MTX resistance seemed to correlate with the extent of MAPK activation, as expression of combined KRASG12D plus Q61R mutations was associated with even greater MTX resistance than expression of KRASQ61R alone. In contrast, treatment with other commonly used drugs in ALL, including daunorubicin, dexamethasone, and cytarabine, did not cause differential effects in KRAS mutant relative to KRAS wild-type cells. These data suggest that the mixed pattern of positive and negative selection of RAS mutations observed at relapse appears to be related to the selective pressure of the chemotherapeutic agents administered during initial therapy.

Several interesting questions are raised by these observations. First, it will be important to understand why RAS mutations are mechanistically linked to MTX resistance while being associated with vincristine sensitivity. Moreover, given that the therapeutic studies described above were performed in T-ALL models, evaluation of these findings in the context of B-ALL will also be important. Finally, RAS mutations have also previously been linked to prednisolone resistance (15) as well, and the MEK inhibitor trametinib was shown in at least one study to be synergistic with prednisolone (12, 16).

Given these findings and the data here revealing collaborative cytotoxic effects of MTX and MEK inhibition, combining MEK inhibitors with conventional therapies may be important to evaluate clinically in the treatment of RAS-mutant ALL.

The insights from this study and prior literature on the genomic analysis of relapsed ALL further highlights that, in most cases, relapse in ALL represents evolution from an ancestral clone present as a subclone at diagnosis. Furthermore, it is clear that relapses in pediatric ALL are not associated with marked increases in genomic alterations. Some mutations appear to be selected due to their direct involvement in chemotherapy resistance, whereas in other cases there is a more complex relationship between the relapsed clone and therapy. Moreover, these data continue to argue for the need to correlate treatment response in ALL with genotype in the clinical setting. Further efforts to identify BCR-ABL–like ALL patients as well as mutations in NT5C2, NR3C1, NRAS, KRAS, PTPN11, and BAM in routine clinical practice are clearly needed given their important therapeutic implications. It will be essential to know whether the recurrent mutations identified in pediatric ALL, with resulting effects on chemotherapy response, are also relevant for adults with ALL. Last, it is important to note that therapeutic approaches for relapsed ALL are rapidly changing with the introduction of anti-CD19–targeting agents (such as chimeric antigen receptor T cells targeting CD19 and blinatumomab), as well as exciting agents targeting CD22. Evaluation of genetic drivers of resistance to these newly introduced agents is ongoing (17) and will hopefully elucidate further novel therapeutic approaches in relapsed ALL.