438 children with de novo AML were treated in the ELAM02 trial and RUNX1 gene status was screened in 386 of them. RUNX1 abnormality was found in 8% (29 of 386) of the cases, with 24 mutations and 5 deletions.
Clinical, cytological and cytogenetic characteristics (Table I)
Main clinical, cytological and cytogenetic characteristics of children RUNX1 mutated and deleted (RUNX1m/del) versus RUNX1 wild type (RUNX1wt) are reported in Table I.
There were no differences between the two groups regarding sex, age, white blood cells count or central nervous system involvement.
RUNX1 m/del AMLs were more likely to be AML-FAB M0 (5/29). RUNX1 mutations were associated with a normal karyotype (10/24) in contrary to deletions. In our cohort, RUNX1 mutation or deletion was exclusive with AML-FAB M5 and KMT2A (11q23) rearrangement and rarely associated with Core Binding Factor (CBF) abnormalities as t(8;21)(q22;q22).
Most of RUNX1m/del patients were classified in the intermediate risk group (69%) and global repartition was comparable with the rest of the cohort. The 2 patients classified in standard group were the 2 with CBF abnormalities.
Type, location, and allele frequency of RUNX1 alterations (Table II, Figs. 1 and 2)
In our cohort we identified 30 RUNX1 mutations in 24 patients, 6 of them cumulating 2 mutations; we counted 11 frameshift mutations, 5 nonsense mutations, 2 duplications, 1 insertion and 10 missense mutations. Numbers of mutations were in RUNT Homology domain. 8 mutations were already described in literature by Brown et al15, including one in a family with germline mutation. In addition to mutated patients, 5 patients presented a RUNX1 deletion (between 50kb and 1,6 Mb, all involving the RUNT domain).
All RUNX1 alterations are represented in Fig. 1 and listed in Table II.
In our cohort, 16 (67%) mutated patients had an elevated variant allele frequency (> 30%). For these patients, RUNX1 status in complete remission was assessed to determine if RUNX1 mutation was germline, which is considered as leukemia predisposing factor. When VAF < 30%, mutation was considered somatic.
Among these 16 patients, 4 could not be evaluated, and only 1 had constitutive mutation. He is still in complete remission.
Otherwise, we observed a harmonious dispersion of VAF, independently of leukemia aggressivity. (Fig. 2)
Among 5 deletions, only one patient had a constitutive RUNX1 deletion, already described by Preudhomme et al.16 This patient was transplanted but died after relapse.
Using the 2015 American College of Medical Genetics and Genomics (ACMG) and Association for Molecular Pathology (AMP) guidelines, we could classify RUNX1 mutations as benign/likely benign, pathogenic/likely pathogenic, and variant of unknown significance (VUS). Even if this classification was established for constitutive mutations and might be less informative in the case of somatic mutation specifically found in leukemoid cells, it is interesting to define the functional impact of RUNX1 alterations. 17
All RUNX1 alterations are classified as pathogenic or likely pathogenic except for 2. The first one is known as benign variant of RUNX1 gene and the second one is unknown. These 2 patients are cured and were treated in favorable group because of CBF alteration.
Molecular co-occurrence (Table III, Fig. 4)
Molecular co-occurrences of genes mutations are summarized in Table III and represented in Fig. 4.
RUNX1 mutated patients had higher number of co-mutations (2.71 on average) compared to the rest of the cohort (1.43).
The most frequent class of co-mutated genes involved control kinase signaling (50%) especially FLT3-ITD, NRAS, FLT3-TKD and KRAS or WT1. RUNX1 alteration was also associated with EZH2 and BCOR mutations (but groups were very small). We found no CEBPA, NPM1, TET2, SETBP1, RAD21, CBL mutations in RUNX1 mutated patients. Except for 1 patient with RUNX1 deletion who had a co-mutation in U2AF1, RUNX1 mutated patients had no alteration in splicing factor (as SRSF2 or SF3B1).
Impact of RUNX1 mutation and deletion on clinical outcome (Table I, II and Fig. 3)
Among the 29 RUNX1m/del patients, 83% achieved complete remission (CR) after two courses of chemotherapy, which is not different from 92% of the entire cohort.
We observed a significant poorer outcome for RUNX1m/del patients compared to RUNX1wt. At 5 years, EFS was estimated at 32.5% (95% confidence interval :16.8–62.8) for RUNX1m/del patients versus 61.4% (CI = 56.2–67.2) for the entire cohort; and OS was 33.6% (CI = 18.6–60.8) compared to 75.7% (CI = 71.3–80.4) (Table I). Hazard ratios for EFS and OS were respectively 2.2 (CI = 1–4,7; p-value = 0.003) and 3,3 (CI = 1.4–7.5; p < 0.0001).
Compared by risk groups, RUNX1m/del patients still had a poorer outcome than patients in adverse risk group (5-y OS: 33.6% for RUNX1m/del versus 66.2% for adverse risk group) (Fig. 3).
There were 8 deaths out of 12 patients (67%) cumulating RUNX1m with FLT3-ITD or adverse cytogenetic risk in ELAM02 versus 44 out of 128 (34%) with adverse risk or FLT3-ITD without RUNX1m/del (p = 0.056), suggesting an even worse prognosis when RUNX1 abnormality is combined with FLT3-ITD rearrangement or other adverse risk factors.
There was no difference in relapse or chemoresistance rate between RUNX1m/del and RUNX1wt patients even if we observed higher mortality. Causes of deaths are listed in Table II including leukemia, infection or post-transplant toxicity.