To the editor:

In childhood and young adolescent B-cell precursor acute lymphoblastic leukemia (B-ALL) PAX5 is one of the most frequent _targets of genetic alterations comprising deletions, intragenic amplifications (PAX5AMP), and point mutations as well as rearrangements (PAX5-r) with multiple partner genes [1]. PAX5-r account for 2–3% of all newly diagnosed B-ALL cases and result in the expression of fusion oncoproteins [1, 2].

Patients with PAX5-r tend to have higher relapse and poorer overall survival (OS) rates compared to good-risk genetic groups [3,4,5]. Recent studies indicate that PAX5-r is associated with the IKZF1plus copy number alteration (CNA) profile and a rather poor event-free survival (EFS) [3, 5]. In small cohorts of infant B-ALL, PAX5-r patients had a worse outcome than those with other non-KMT2A genetic subtypes [6, 7].

Most, but not all, cases with PAX5-r belong to the PAX5-altered (PAX5alt) subtype identified by gene expression profiling [1]. PAX5alt is associated with an intermediate to poor prognosis with a strong dependence on IKZF1 codeletion [1, 8, 9]. Despite sharing a distinctive expression signature, the underlying genetic landscape of PAX5alt is heterogeneous and various types of PAX5 lesions, which differently affect disease biology and outcomes, are merged into a single group [1, 9]. Moreover, PAX5 is fused to a multitude of different partner genes, and since most of these fusions have been detected only in a few cases [1], the impact of individual PAX5-r on outcomes remains to be determined.

In this international study, we collected cases with PAX5::AUTS2 B-ALL diagnosed over the past decades in patients aged 0–18 years, without any time or study protocol restrictions, to evaluate the prognostic relevance of this rare genetic subtype. In accordance with the Declaration of Helsinki, patients were enrolled in respective clinical trials with written informed consent from their parents or legal guardians, and the use of surplus diagnostic material for research purposes was approved by the institutional review boards of the participating centers.

Patients with a confirmed PAX5::AUTS2 fusion detected by RT-PCR or a next-generation sequencing (NGS) approach were eligible for inclusion in the study. Additionally, since PAX5::AUTS2 frequently results from unbalanced der(9)t(7;9)(q11;p13) rearrangements, cases with breakpoints in both PAX5 and AUTS2 detected by single nucleotide polymorphism (SNP) array analysis were included without additional molecular genetic verification, likewise cases identified by optical genome mapping (OGM). All PAX5::AUTS2 fusion transcripts were in frame or at least predicted to be. CNA profiling for IKZF1 deletion and IKZF1plus was performed by SNP array, multiplex ligation-dependent probe amplification (MLPA), digitalMLPA or OGM following standard procedures (Supplementary Methods).

The Kaplan–Meier method was used to determine EFS and OS rates and the analysis was performed in R (version 4.2.2) statistical environment. Adverse events were defined as relapse at any site, the development of a second malignant neoplasm (SMN), or death from any cause. OS was defined as the time from diagnosis to the date of last follow-up or death. The cumulative incidence of relapse (CIR) was calculated using the Kalbfleisch and Prentice method and compared with the Gray’s test considering death and SMN as competing events. Multivariate analysis was conducted using a Cox proportional hazards regression model.

We identified 50 patients diagnosed with PAX5::AUTS2 B-ALL, including 16 previously published cases (Supplementary Table S1). The main demographic and clinical features showed a male predominance (60% vs 40%), a median age of 2.0 years (range 0.6–15.5 years), including eight infants (≤1 year), and highly variable white blood cell counts (WBC) ranging from 1.2–537.6 × 109/L (median 40.1 × 109/L) (Table 1).

Table 1 Demographic, clinical, genetic characteristics and outcome of PAX5::AUTS2 patients.
Full size table

We detected IKZF1 deletions in 69.6% (32/46 with available data) of PAX5::AUTS2 cases, most displaying also CDKN2A/B (90.0%, 27/30) and/or PAX5 (87.1%, 27/31) deletions (Supplementary Table S1). Based on this deletion pattern, 62.2% (28/45) showed the IKZF1plus CNA profile [10] (Table 1), which is higher than reported for childhood PAX5alt (20–30% IKZF1-deleted, 20% IKZF1plus) or PAX5AMP (13% IKZF1plus) cases [1, 9, 11].

According to National Cancer Institute (NCI) criteria (age ≥10 years and/or WBC ≥50 × 109/L; infants), 60.0% (30/50) of patients had high-risk (HR) status. Of the patients with available data, 24.4% (10/41) showed a poor prednisone response (Table 1), which is ~15% higher than in the average population of childhood B-ALL [10]. Measurable residual disease (MRD) data assessed by flow cytometry (FCM) on day 15 were only available for 23 patients, and according to AIEOP-BFM definitions, 30.4% (7/23) showed HR disease with ≥10% residual blast cells [12]. At the end of induction therapy (EOI; days 28-42), PCR- and/or FCM-MRD measurements showed that 79.5% (35/44) of patients were MRD-positive: 31.8% (14/44) ≥5×10−4 and 47.7% (21/44) <5 × 10−4. At the end of consolidation therapy (EOC; mainly day 78; range 71–117 days), 29.7% (11/37) of patients still presented with MRD, but except for two (≥5 × 10−4), with levels ≤1 × 10−4 (Table 1 and Supplementary Table S1).

Overall, 46.0% (23/50) of PAX5::AUTS2 patients experienced a relapse with a median time-to-event of 1.6 years (range 0.6-4.1 years) with the majority (65.2%, 15/23) occurring within two years after diagnosis (Fig. 1A; Supplementary Table S1). Most patients had isolated bone marrow (BM; 60.9%, 14/23) relapses, but CNS disease was detected in 34.8% (8/23) of patients mostly combined with BM recurrence (Table 1). In total, 38.0% (19/50) of patients had CNS disease: 22.0% (11/50) only at diagnosis, 6.0% (3/50) at diagnosis and relapse, and 10.0% (5/50) only at relapse. Among infants, 87.5% (7/8) had CNS involvement, with 62.5% (5/8) already at diagnosis (Supplementary Methods), one patient at both time points, and two additional cases only at relapse (Supplementary Table S1). CNS disease was more frequently detected in infants than in children [87.5% (7/8) vs 40.0% (12/30), Fisher exact test p = 0.0031].

Fig. 1: Course of disease and outcome of PAX5::AUTS2 patients.
figure 1

A Swimmer plot illustrating the clinical course of each individual patient. Relapses (orange dots), bone marrow transplantation (BMT; blue triangles), and death (red crosses) as well as age groups <18 months (light brown bars) and ≥18 months (gray bars) are indicated. B–C Cumulative incidence of relapse (CIR) and Kaplan-Meier survival curves of event-free (EFS) and overall survival (OS), (B) of all PAX5::AUTS2 cases (n = 50), (C) based on age group; red, <18 months; blue, ≥18 months. Gray’s test p-value for CIR, Log-ranks test p-values for EFS and OS. D Forrest plot showing results of multivariate Cox regression analysis. HR hazard ratio, CI confidence interval, p Gray’s test p-value, WBC white blood cell count, NCI-HR National Cancer Institute high-risk, MRD-EOI (measurable residual disease) determined by PCR and/or flow cytometry at the end of induction (EOI) therapy.

Full size image

The 5-year CIR for all patients was 48.0 ± 7.8% (Fig. 1B) and of the patients who relapsed, 17.4% (4/23) experienced a second BM relapse within 3–9 months of the first (Fig. 1A). Among patients experiencing a relapse, 30.4% (7/23) died, three of progressive disease and four of treatment-related mortality (TRM), and in total, 20.0% (10/50) of patients died (n = 3 leukemia, n = 4 TRM, n = 3 other/unknown reasons) (Fig. 1A, Table 1 and Supplementary Table S1).

With a median follow-up of 4.8 years (range 0.3–13.0 years), we observed 5-year EFS and OS rates for all patients of 47.9 ± 7.6% and 76.2 ± 7.1%, respectively, (Fig. 1B). Notably MRD levels (≥5 × 10−4 vs <5 × 10−4 vs negative) at EOI did not significantly impact CIR, EFS or OS (Supplementary Fig. S1A), suggesting that in PAX5::AUTS2 patients EOI-MRD negativity does not predict a favorable outcome. Patients with ≥5 × 10−4 EOI-MRD had a worse EFS compared to cases with <5 × 10−4/negative EOI-MRD, but the result did not reach statistical significance (p = 0.057) (Supplementary Fig. S1B). Furthermore, we did not find any differences in 5-year EFS and OS between IKZF1plus and non-IKZF1plus patients (46.6 ± 9.9% vs 60.6 ± 14.0%, p = 0.67; 78.7 ± 8.5% vs 93.8 ± 6.1%, p = 0.89) indicating that also IKZF1plus has no predictive value (Supplementary Fig. S1C).

Given that patients were enrolled in various clinical trials with differing risk stratification criteria and treatment regimens, we compared the outcomes between earlier clinical trials and the two most contemporary ones (i.e., AIEOP-BFM ALL 2017, ALLTogether-1) but did not find any significant improvement (Supplementary Fig. S1D).

When comparing the outcomes of children and infants, we observed a higher CIR and consequently a poorer EFS for infant patients (Supplementary Fig. S2A). As several patients were just over 1 year of age (Table 1), we explored whether age influenced the outcome. Our analysis according to different age groups revealed that patients aged <18 months had a significantly higher 5-year CIR (68.5 ± 12.7% vs 36.2 ± 9.4%, p = 0.006) and worse outcome (Fig. 1C) than other age groups (Supplementary Figs. S2B–D). In multivariate analysis, EOI-MRD ≥5 × 10−4 (hazard ratio 15.55, p = 0.001) and age (<18 months vs older; hazard ratio 16.65, p = 0.001) had an independent impact on EFS (Fig. 1D).

Collectively, our study demonstrates that childhood PAX5::AUTS2 B-ALL is characterized by a high frequency of CNS involvement and is a relapse-prone subtype with poor outcomes. The disease mainly affects infants and toddlers with over 80% being less than three years old, which differs from the childhood PAX5alt group, which is more common in older patients (children 1–18 years, n = 94, median 9.5 years; ST1 cohort [1]) [1, 8, 9]. The observed EFS and OS survival rates for PAX5::AUTS2 B-ALL were lower than those in the overall PAX5alt group but similar to those of patients with PAX5AMP [1, 8, 9, 11], underlining the importance of focusing outcome analysis on genetically homogeneous entities with unique alterations.

Our data also support the notion that PAX5-r is in general recurrent in infants and along with KMT2A-r may represent a further subgroup of infant B-ALL with a potentially dismal outcome, while NUTM1-r has a favorable prognosis [6, 7, 13].

The finding that PAX5::AUTS2 patients enrolled in contemporary trials still exhibit high relapse rates, emphasizes the need to further explore innovative _targeted treatment options [3]. Moreover, strategies to mitigate the risk of relapse and treatment-related mortality associated with salvage therapy are needed. Frontline use of the bi-specific T-cell engager blinatumomab, which has been proven to be safe and effective also in infants [14, 15] may be a worthwhile consideration.