This study analyzed 80 patients with de novo AML. They were diagnosed between March 2010 and March 2018. The diagnosis was made at each institution by bone marrow aspirate and immunophenotype. Bone marrow samples were obtained at the time of diagnosis and submitted to the Mexican Inter-Institutional Group for Identifying Childhood Leukemia Causes in Mexico City. Data were collected from medical charts, including sex, age, peripheral white blood cell count, percentage of bone marrow blasts at diagnosis, FAB (French-American-British) classification, and treatment protocol. The clinical features of the analyzed patients had been previously reported in Molina et al. (17).

Risk classification at diagnosis was assigned based mainly on morphology; in most public hospitals in Mexico City, cytogenetics and minimal residual disease detection are unavailable. The risk groups were established as follows: standard-risk group: M1, M2, and M4 (with at least 3% of eosinophils); high-risk group: M4 (with less than 3% of eosinophils) and M5. Additionally, in one hospital, more than 5% of blasts in bone marrow on day 15 was used to identify patient with high-risk features. In M3 patients low-risk group includes patients with white blood cell count (WBC) < 10 × 10⁹ and platelets > 40 × 10⁹; for intermedia-risk group WBC count < 10 × 10⁹ and platelets < 40 × 10⁹ and for high-risk group WBC ≥ 10 × 10⁹.

Patients were classified with an intermediate risk when MRD level by flow cytometry was >0.1% after course 1, but fell to <0.1% after course 2. Some child’s parents covered this test in a private laboratory.

The ethics and scientific review boards of the National Institute of Genomic Medicine, Mexico City, Mexico, approved this study (document number 28-2015-1). All human samples and clinical information were approved for the present study. The children’s parents signed the informed consent obtained according to the Declaration of Helsinki.

The DNA was extracted from bone marrow samples with Maxwell^® 16 Blood DNA Purification Kit (Promega, Madison, WI, United States) according to the manufacturer’s recommendations. The DNA purity and concentration were measured with NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States) and Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA, United States).

CEBPA sequencing was performed with the “Myeloid solution” panel by Sophia Genetics (Sophia Genetics SA, Saint-Sulpice, Switzerland). Library preparation and sequencing were performed according to the manufacturer’s protocol. Libraries were pooled and sequenced on a MiSeq System v3 chemistry (Illumina, San Diego, CA, United States). Sequence data were analyzed with the Sophia DDM^® software version 5.2.7.1 (Sophia Genetics SA, Saint-Sulpice, Switzerland). Deep sequencing was greater than 500X for all the target regions. Variant Fraction (VF) was calculated for each mutation by dividing the number of sequencing reads showing the mutation by the total sequencing read at the mutation position.

The statistical analysis was performed as described previously (17). For dichotomic variables, chi-square or two-sided’ Fisher’s exact test was used to compare proportions among different groups. A non-parametric Mann-Whitney U test was applied for continuous variables, a p-value < 0.05 was considered statistically significant. All calculations were performed with the SPSS software package, SPSS v21 (Chicago, IL, United States).

The Kaplan-Meier method (18) was used to assess overall survival (OS). The log-rank test was used to evaluate differences between survival distributions with a 95% confidence interval (CI). The OS time was calculated from the day diagnosis was confirmed to either the last follow-up or death from any cause. Patients who did not experience an event were censored at the last follow-up. Those who did not attend follow-up appointments were censored at the date of the last known contact.

The maternal years of education were used as socioeconomic status (SES) indicator to evaluate if it impacted the inferior OS observed. The patients were further categorized in CEBPA (positive/negative) and death (yes/no) to identify possible associations. As per the categorization used by the Childhood Leukemia International Consortium, SES was assigned as follows: [0–9 years (low SES), 9.1–12.9 years (reference category), ≥13 years of education (high SES)] (19, 20). The exact Fisher’s test was used to assess the difference between groups.

Demographic, clinical, and the main biological characteristics of the patients were previously reported (17). Fifty-five percent of the patients were male. The age at diagnosis was similar in both sexes; the mean age at diagnosis was 9.3 years (range 0.4–17.5). Acute promyelocytic leukemia (APL) M3 was the most prevalent subtype (36.3%), followed by M2 (33.8%). M0, M5, and M6 were the less common subtypes. The mean OS in the study population was 1.95 years.

The patients were treated based on one of the following four protocols: BFM-1998, BFM-2001 (Berlin-Frankfurt-Münster 1998 and 2001), NOPHO-AML93 (Nordic Society of Pediatric Hematology and Oncology), or PETHEMA-APL-05 (Spanish Program of Treatments in Hematology). APL patients were treated according to the PETHEMA-APL-05 protocol. None of the patients received an allogeneic or autologous bone marrow transplant. Treatment information and the impact of treatment on OS have been previously reported (17).

CEBPA was mutated in 12.5% (10/80) of the cases (CEBPA^POS). A total of 14 different mutations were identified. The mutational profile is shown in Table 1 and Figure 1. The most common mutations were frameshifts located at the N-terminus of the protein 57.14% (8/14). Biallelic CEBPA mutations (CEBPA^BI) were identified in five patients, which accounted for 50% of the CEBPA^POS cases. All identified mutations were uploaded to the ClinVar database (accession numbers for submission SUB11180409 are SCV002104196-SCV002104210)¹.

Five germline variants were also identified: His191His (27.5%), Thr230Thr (5%), Hist195_Pro196 (5%), and Pro204Pro (1.2%). They were classified as benign or likely benign according to the ACMG criteria (21). The variant pGly223Ser located between the TAD2 and BDB domains was identified in one individual (1.2%, VF = 47.5%). It is listed in the ClinVar database (Variation ID 239926) and shows conflict regarding its clinical significance (uncertain significance vs. likely benign). According to the gnomAD v3 population database, its global frequency is 0.0004189 (58 of 138,460 alleles), with a maximum allele frequency of 0.004221 (56 of 13,266) in the Latino subpopulation. It has been identified in adults only² (as of March 2022). The frequency in a local exome database of Mexican individuals is 1.1%. This frequency exceeds the prevalence of a pathogenic variant causing CEBPA-associated familial AML (22). The variant has not been reported in cases of familial AML. It is predicted benign for most in silico algorithms, although no functional studies have been performed. Considering the available information, we classified this variant as likely benign.

Sex distribution, FAB classification, treatment protocol, risk assessment, WBC count, age, and blast percentage in bone marrow at diagnosis were stratified according to the presence of CEBPA mutations. The results are shown in Table 2. A non-significant (NS) higher proportion of male patients was found in the CEBPA^POS subgroup (60 vs. 54.3% in CEBPA^NEG group, p = NS). Statistically significant differences were found in the FAB subtype distribution. M2 was the most common among CEBPA^POS patients (70 vs. 28.6% in CEBPA^NEG p = 0.009). CEBPA^POS patients had significantly higher WBC counts at diagnosis: 50% of CEBPA^POS patients had a WBC count > 100,000, compared with 12.9% in the CEBPA^NEG group (p = 0.004).

A significantly higher mean OS was observed in CEBPA^NEG patients considering 1 year after diagnosis (3.1 years in CEBPA^NEG vs. 1.8 years in CEBPA^POS, p = 0.0001) or 2 years (2.1 years in CEBPA^POS vs. 3.9 years in CEBPA^NEG, p = 0.0258).

The impact of CEBPA mutation on OS was analyzed by comparing three subgroups of patients according to CEBPA status: CEBPA^MONO (individuals with monoallelic mutations), CEBPA^BI, and CEBPA^NEG. The results are shown in Figures 2A–C. The OS significantly decreased in patients with CEBPA mutations compared with the CEBPA^NEG group (p = 0.002). No differences were found in OS between patients with CEBPA^MONO and CEBPA^BI. Similar results were observed after removing from the analysis the three CEBPA^MONO patients with CEBPA mutation VF < 2% (patients M157, M168, and M183). A significant decrease in OS was observed (p = 0.009), as shown in Figure 2B. OS analysis was performed after removing the M3 patients from the CEBPA^NEG mutated group. The result still showed a significantly lower OS in the CEBPA^POS group (p = 0.04; Figure 2C).

No statistically significant association was detected between CEBPA^POS cases and a low SES (p = 0.99) or between a low SES and a poor outcome (death) (p = 0.76; Results of the analysis are shown in Supplementary Table 1).

CEBPA mutations have been found more frequently among FAB M1 and M2 patients (35, 36). This was also true in our series (Table 2); M1 or M2 were present in 80% of the CEBPA^POS patients vs. 40% of the CEBPA^NEG subgroup (p = 0.01). However, a statistically significant difference was observed in the M2 distribution (p = 0.009). CEBPA^POS patients had a significantly higher WBC count at diagnosis; 50% of the CEBPA^POS patients had a WBC count > 100,000, compared with 12.9% in the CEBPA^NEG subgroup. Hyperleukocytosis at diagnosis has been associated with an increased risk for relapse in adults with CEBPA^BI AML (36). Due to the small number of patients in each subgroup, the statistical power is low to compare the distribution of clinical features between CEBPA^BI and CEBPA^MONO.

CEBPA mutations were associated with a significantly lower OS but without significant differences between mono or biallelic CEBPA mutated patients. The mean OS in CEBPA^BI patients was 2 months longer than in CEBPA^MONO (11.2 vs. 9.27 months p = NS).

Considering that the biological impact of mutations with VF < 2%, classified as VUS, that may belong to a small subclone, would be different from those present in a higher proportion in the dominant clone, we evaluated the effect of CEBPA mutations in OS after removing the three patients with low allelic fractions. The result was similar to the one obtained taking into account the whole CEBPA^POS group (Figure 2B).

Several studies have assessed the clinical impact of CEBPA biallelic mutations on the survival of pediatric patients to determine if they are useful biomarkers of favorable-risk AML, as observed in adults. The findings of the most representative studies are summarized in Table 3. In contrast with our results, none of those studies have reported a negative effect of CEBPA mutations on OS; they found no impact of CEBPA mutations or a positive effect mainly for CEBPA^BI mutations.

Heterogeneity in relapse rates and survival outcomes has been reported in CEBPA^POS patients despite the favorable impact of CEBPA^BI mutations on OS (36). Concurrent mutations in other genes and a differential impact of CEBPA mutations according to their location would contribute to the observed heterogeneity (15, 37). We searched for concurrent mutations in FLT3, NPM1, CSF3R among CEBPA^POS patients; however, statistical tests could not be run due to the small sample size.

FLT3 was mutated with VF > 30% in 3 out of 10 CEBPA^POS patients (one CEBPA^BI and two CEBPA^MONO). We had previously reported that FLT3 mutations have a negative impact on clinical outcomes in Mexican pediatric AML patients, and OS is significantly lower in patients with FLT3 mutations than in FLT3^NEG (17, 37). Akin et al. reported that 2 out of 3 patients who carried concurrent CEBPA and FLT3 mutations died during treatment (38). A negative effect of FLT3 mutations on the CEBPA^BI positive effect has been observed in adult patients. However, a similar frequency of FLT3-ITD between mono and biallelic CEBPA mutated patients with no impact on OS has been found in children (14).

None of the CEBPA^POS patients harbored NPM1 mutations. This supports the idea that CEBPA and NPM1 mutations are mutually exclusive (39). The frequency of NPM1 in our series (1.2%) was much lower than previously reported (6–8%) (40–43).

CSF3R was mutated in two CEBPA^POS individuals; p.Thr618Ile with VF = 48.6% was identified in one of them. This activating mutation occurs in the membrane-proximal region of the colony-stimulating factor 3 receptor. The mutation produces ligand-independent receptor activation, a hallmark of chronic neutrophilic leukemia (44). Concurrent CSF3R mutations in pediatric CEBPA^BI AML were associated with significantly poorer relapse-free survival than wild-type CSF3R; however, OS was not significantly different (45). A similar impact of CSF3R mutations on CEBPA^BI groups has been reported in adults (46). A pathogenic mutation in WT1 with VF = 87% was also identified in one CEBPA^BI patient. Ho et al. also found three WT1-mutated patients in the CEBPA^BI subgroup, two of them died of progressive disease during induction (47). WT1 mutations are independent poor prognostic factors with a 5-year OS of 35% and EFS of 22% in children (48). No KIT mutations were identified in the CEBPA^POS patients.

In the present research, a low OS of the cohort was noted compared with reports of other parts of the world. It is well recognized that in developing countries, survival rates for this disease are low (∼40%). Several other biological and non-biological factors not evaluated in the present study could be contributing to this poor outcome. For instance, a late presentation, malnutrition, high treatment-related mortality, low SES, and high treatment abandonment rates have been recognized (49). The SES analysis performed, based on the mother years of study as an SES indicator, no statistically significant association was detected between CEBPA^POS positive cases and a low SES (p = 0.99) or between a low SES and a poor outcome (death) (p = 0.76; Results of the analysis are shown in Supplementary Table 1). SES does not seem to be a confounding factor affecting the results of this study.

This study has several limitations that must be considered to interpret the findings. Cytogenetic/FISH characterization is not routinely performed in most Mexican public institutions; it was not performed in these patients impairing the risk stratification and precluding further analysis aimed to evaluate the impact of CEBPA mutations on the normal karyotype subgroup. This study is a retrospective analysis of a heterogeneous group of patients treated according to four different protocols in eight different institutions. Although this study is the most extensive series of Mexican AML patients analyzed so far, the number of patients is small. Thus, it would reduce the statistical power to detect additional differences in the distribution of clinical features between positive and negative CEBPA patients.

However, these limitations do not seem to explain the significant reduction in OS in the CEBPA^POS. The lack of cytogenetic information at diagnosis and the use of different treatment regimens are common to the whole cohort of patients, not affecting the CEBPA^POS patients exclusively. The mutational status of CEBPA gene was unknown to the physician at the moment of choosing treatment protocols and during the whole course of the disease; therefore, a differential bias in clinical management affecting only the CEBPA^POS group is unlikely. CEBPA^POS patients were treated in 4 different hospitals. Therefore, a possible “hospital-effect” on OS of CEBPA^POS patients is also unlikely.