PAX5 P80R mutation identifies a novel subtype of B-cell precursor acute lymphoblastic leukemia with favorable outcome

TO THE EDITOR:

B-cell precursor acute lymphoblastic leukemia (BCP-ALL) is a rare aggressive malignancy in adults. BCP-ALL is frequently characterized by recurrent chromosomal translocations that deregulate proto-oncogenes or result in fusion genes encoding chimeric oncoproteins.¹  Gene expression profiling studies have consolidated the notion that specific chromosomal aberrations delineate unique leukemia subtypes with distinctive biological and clinical features.²  Therefore, their identification is critical for optimal therapeutic management. The most clinically relevant example is the use of tyrosine kinase inhibitors in patients with the Philadelphia chromosome (Ph) or related BCR-ABL1 fusion, which is found in ∼25% to 30% of adult BCP-ALL. However, apart from Ph-positive ALL, a significant proportion of adult BCP-ALL lacks recurrent classifying chromosomal aberrations, contributing little to current treatment stratification strategies.³  During the last decade, genome-wide copy-number analyses have uncovered numerous genomic microdeletions in BCP-ALL that recurrently target tumor suppressor and lymphoid factor genes like CDKN2A, IKZF1, and PAX5.⁴  These alterations are the result of erroneous RAG-mediated recombination events that accumulate in transformed progenitor B cells.^5,6  By contrast, point mutations are relatively infrequent in BCP-ALL as compared with other tumors, including other hematopoietic malignancies. Remarkably, both microdeletions and point mutations are usually thought to be secondary oncogenic events, whereas the biological and clinical features of BCP-ALL are frequently dictated by the primary cytogenetic aberrations.

To refine oncogenic subtypes and elucidate primary genetic alterations in adult BCP-ALL, we performed transcriptome analysis by RNA-sequencing (RNA-seq) in a discovery cohort of 170 BCP-ALL cases that were virtually all negative for the common classifying cytogenetic aberrations (details of the study cohorts are provided in supplemental Figure 1, available on the Blood Web site). Unsupervised hierarchical clustering of gene-expression data identified well-defined clusters of BCP-ALL cases (Figure 1A). Four clusters corresponded to known oncogenic subtypes, namely ZNF384,^7,8  ERG/DUX4,^7-10  BCR-ABL1-like,¹¹  and KMT2A/MLL, as demonstrated by relevant in-frame fusion transcripts or IGH fusions in most cases (including KMT2A fusions, which were cryptic on cytogenetics). An additional cluster comprised 14 cases, which lacked any fusion gene. Mutational analysis of RNA-seq data revealed the presence of a unique PAX5 mutation, c.239C>G, p.P80R, in all cases. PAX5 is a transcription factor from the paired box family that controls B-cell identity and development.^12,13 PAX5 is also a haploinsufficient tumor suppressor gene in BCP-ALL^14,15  that is targeted by various deletions and mutations in ∼30% of BCP-ALL.^4,16,17  Deletions and mutations of PAX5 have been considered so far as secondary oncogenic events because they were found in various BCP-ALL subtypes and possibly in minor subclones. However, we did not identify PAX5^P80R mutation in any case from other clusters nor in association with any classifying alteration, suggesting that this specific PAX5 mutation may define a distinct BCP-ALL subtype.

We analyzed a recurrence cohort consisting of 419 additional, unselected BCP-ALL from patients enrolled in trials conducted by the Group for Research on Adult Acute Lymphoblastic Leukemia (GRAALL)^18,19  (supplemental Figure 1). By Sanger sequencing, we identified 16 additional PAX5^P80R cases that, as anticipated, did not harbor any classifying cytogenetic aberration. Thus, the prevalence of the PAX5^P80R subtype in the population of adult patients with newly diagnosed Ph-negative BCP-ALL enrolled in GRAALL trials was 5.3% (29 out of 551).

The Pro80 residue is included in the paired box DNA binding domain (PD) of PAX5 (Figure 1B). Structural modeling suggests that Pro80Arg mutation could potentially disrupt DNA binding by perturbing the PD N-terminal helical domain. Alternatively, it might increase DNA binding through electrostatic interactions with the DNA phosphodiester backbone, leading to an increase in nonspecific DNA binding. To gain insights into the functional consequences of PAX5^P80R mutation, we analyzed the gene expression signature of PAX5^P80R cases. Using gene set enrichment analysis (Figure 1C), we observed a negative enrichment of the PAX5 target genes activated during early stages of B-cell development,²⁰  in accordance with a loss-of-function effect of the mutation. However, the transcriptional program of PAX5 was not completely abrogated despite biallelic genomic alteration, suggesting that the P80R mutant retained some function. Interestingly, pathway analysis also demonstrated significant enrichment of MYC target genes. Collectively, these data suggest that the effects of P80R mutation may involve both partial loss of function of PAX5 on its target genes and activation of an aberrant transcriptional program.

We analyzed additional genomic alterations in the 30 PAX5^P80R cases (Figure 1D). The karyotype frequently displayed structural rearrangements of chromosomal arms 9p (n = 15/30) and/or 7p (n = 11/30) such as dic(9;20) and der(7;9) (supplemental Table 1). PAX5 gene is located at 9p13.2, and homozygosity of P80R was indeed observed in 19 cases. All heterozygote remaining cases with available data (n = 9/11) harbored a second PAX5 mutation (3 frameshifts, 3 missenses, 1 nonsense, and 2 splice variants) leading to PAX5 biallelic alteration. IKZF1 located at 7p12.2 was also frequently lost. Strikingly, all the PAX5^P80R cases harbored mutations in signal transduction factor genes, especially in the RAS pathway. Copy-number data showed remarkably few additional focal alterations (median 1 per case, range 0-8) except focal deletions of the second CDKN2A allele located at 9p21, resulting in biallelic loss. The transcriptional signature of the PAX5^P80R subtype showed significant enrichment in MYC and E2F target genes, the latter in agreement with the loss of repression by CDKN2A, and in genes activated by the mTORC1 complex, in agreement with RAS-mediated activation of mTOR signaling (Figure 1C). Altogether, the PAX5^P80R subtype was characterized by a distinct transcriptional signature in relation with a unique combination of genomic alterations targeting PAX5, CDKN2A, and RAS.

Clinical features and outcome associated with the PAX5^P80R BCP-ALL subtype were analyzed within a cohort of 312 patients with Ph-negative BCP-ALL enrolled in the GRAALL-2003 and GRAALL-2005 trials^18,19  (supplemental Figure 1). The patients and initial disease characteristics (supplemental Table 2) were not different in the PAX5^P80R group, with the exception of a lower frequency of IKZF1 intragenic deletion (0% vs 27.0%, P = .025). PAX5^P80R patients had a good peripheral blood prednisone response at day 8 (100% vs 83.9%, P = .14), and they had a significantly better bone marrow blast clearance at day 15 (92.9% vs 56.5% in the remaining cohort, P = .010). At the end of induction, all PAX5^P80R patients achieved complete remission (100% vs 94.3%, P = 1.0) and good molecular response, none of them having minimal residual disease (MRD) higher than 10⁻³ (MRD1 ≥10⁻³, 0% vs 31.2%, P = .036). The median follow-up for outcome analyses was 5.3 years. When censoring at allogeneic stem cell transplants, PAX5^P80R patients had a significantly higher probability of 5-year event-free survival (EFS; 81.8% vs 42.7%; hazard ratio = 0.20, 95% confidence interval 0.05-0.83; P = .026) and 5-year overall survival (OS; 90.9% vs 47.0%; hazard ratio = 0.12, 95% confidence interval 0.02-0.87; P = .036) (Figure 2A-B). This difference remains without censoring at allogeneic transplantation (Figure 2C-D; supplemental Figure 2). Therefore, PAX5^P80R characterizes a distinct genetic subtype with a favorable prognosis, at least in the context of a high-intensity, pediatric-inspired chemotherapy regimen.

In conclusion, we report here the first oncogenic subtype defined by a point mutation in ALL, which is also the first good-risk genetic subtype in adult BCP-ALL. In addition to the PAX5^P80R mutation, this subtype is typically associated with inactivation of the second PAX5 allele, biallelic deletion of CDKN2A, and alterations of the RAS pathway. Systematic screening at diagnostic workup may be easily implemented in order to refine risk-adapted treatment stratification.