In this retrospective analysis, 297 patients with de novo AML were enrolled. Samples were collected at Zhengzhou University People’s Hospital and the First Affiliated Hospital of Zhengzhou University from January 2017 to July 2021. The last date of follow-up was July 30, 2022. Patients with a history of hematologic disease, therapy-related AML or other malignancies, or acute promyelocytic leukemia (FAB-M3) were excluded. All baseline data were obtained from medical records, including age, sex, white blood cell count (WBC), hemoglobin (HB), platelet (PLT) count, lactate dehydrogenase (LDH), peripheral blood (PB) blast percentage, bone marrow (BM) blast percentage, immunophenotype, karyotype, gene mutations, and French-American-British classification (FAB) subtypes as well as the induction therapy regimen and treatment response. All patients signed an informed con-sent form at the time of treatment, and this study complied with the requirements of our institutional ethics committee (2021-140-02).

Before initial treatment, 2-3 ml of BM aspiration fluid was collected and placed in an EDTA anticoagulant tube. The Navios flow cytometer from Beckman Coulter Company in the United States was used for a 4-color immunolabeling method to detect the expression levels of CD2, CD4, CD7, CD34, CD56, CD38, CD117, CD33, CD13, HLA-DR, and others. CD45/SSC was utilized to set the gate, and the Kaluza software was employed to perform immunophenotypic analysis of BM blasts. Antigen expression was defined as positive when the amount of antigen expression was ≥20%. Representative flow cytometry plots of bone marrow blasts from AML patients in the DNMT3A-mutated/CD7-positive (CD7+) group, the DNMT3A-mutated/CD7-negative (CD7-) group, the DNMT3A-wild-type/CD7+ group, and the DNMT3A-wild-type/CD7- group are displayed in Supplementary Figure 1 .

Molecular biological examinations were performed by next-generation sequencing (NGS). Leukemic mutation genes such as FLT3-ITD, NPM1, KIT, CEBPA, DNMT3A, IDH1, IDH2, TET2, EZH2, RUNX1, ASXL1, PHF6, TP53, SF3B1, SRSF2, U2AF1, BCOR, ZRSR2, NRAS, CBL, SETBP1, ETV6 and JAK2 were analyzed.

Patients received standard induction chemotherapy regimens based on disease and risk stratification (24, 25), with doses slightly adjusted for each patient. The induction regimens for each group of patients are summarized in Supplementary Table 1 . Induction chemotherapy regimens included IA (idarubicin 8-12 mg/m² for 3 days, cytarabine 100-200 mg/m² for 7 days), DA (daunorubicin 40-60 mg/m² for 3 days, cytarabine 100-200 mg/m² for 7 days), MA (mitoxantrone 6-8 mg/m² for 3 days, cytarabine 100-200 mg/m² for 7 days), or CAG+D (decitabine 20 mg/m²/d, d1-5, aclarubicin 20 mg/m²/d, d3-6, cytarabine 25 mg/m²/q12h, d7-14, and G-CSF, 300 μg/d, d1-14). Patients in complete remission (CR) continued to undergo induction chemotherapy with medium- or high-dose cytarabine, while those in partial remission (PR) or nonremission (NR) received a second cycle of induction chemotherapy or salvage therapy. Based on the availability of donors, patients with intermediate and adverse prognoses underwent hematopoietic stem cell transplantation (HSCT).

CR was defined as <5% blasts in BM, no blasts with Auer bodies, no blasts in peripheral blood, no extramedullary leukemia, neutrophil count ≥1.0×10⁹/L, and PLT count ≥100× 10⁹/L. CR with incomplete hematologic recovery (CRi) was defined as meeting all CR criteria, except residual neutropenia (<1.0×10⁹/L) or thrombocytopenia (<100×10⁹/L). PR was defined as the proportion of BM blasts between 5% and 25%, which should be at least 50% lower than before therapy, and the blood cell count must be within the range of CR. NR was defined as failure to achieve CR, CRi or PR after chemotherapy. The primary study endpoints were overall survival (OS), measured as the time from diagnosis to the date of death from any cause or the date of the last follow-up, and relapse-free survival (RFS), measured as the time from the date of CR or CRi to the date of hematologic relapse, death from any cause, or the date of the last follow-up (3).

The comparisons between groups were made using Kruskal-Wallis rank-sum test and one-way ANOVA for continuous variables, which are presented as medians (ranges). Fisher’s exact test or the chi-square test was used to compare categorical variables. The Kaplan−Meier method was used to examine OS and RFS, and the log rank test was used to compare group differences. To examine the parameters influencing OS and RFS, univariate and multivariate Cox proportional hazards regression were performed. All statistical analyses were carried out using SPSS software version 21.0, and graphs were created using GraphPad Prism version 9.0. Two-sided p values <0.05 were considered statistically significant. Additionally, OncoPrint was utilized to identify and visualize gene mutations in AML patients.

This study comprised 297 patients with a median age of 48 years (range: 5-79 years), including 172 men and 125 women. In the study population, the median and range of WBC and PLT counts, Hb levels, and percentage of blasts in BM and PB were 15.88 (0.43-384.87)×10⁹/L, 38.5 (2-579)× 10⁹/L, 76 (35-147) g/L, 62.0% (12.4%-96.0%), and 43% (0-97%), respectively. According to the mutation of DNMT3A and the expression of CD7, the patients were divided into the following four groups: the DNMT3A-mutated/CD7+ group (n=21), the DNMT3A-mutated/CD7- group (n=44), the DNMT3A-wild-type/CD7+ group (n=104), and the DNMT3A-wild-type/CD7-group (n=128). We found that the DNMT3A-mutated/CD7+ group had significantly higher initial WBC counts than the other three groups (all p<0.05, Table 1 ). The age of the DNMT3A-mutated patients was significantly higher than that of the DNMT3A-wild-type patients (p=0.001). The DNMT3A-mutated patients had higher initial PLT counts and BM blast ratios (p<0.001, p=0.030). The rate of the FAB-M5 subtype was higher in DNMT3A-mutated AML patients than in DNMT3A-wild-type AML patients (35.4% vs. 15.1%, p<0.001), and the rate of the FAB-M2 subtype was lower than that in DNMT3A-wild-type patients (38.5% vs. 59.5%, p=0.003). Patients were categorized into favorable, intermediate and adverse risk groups according to the 2022 ELN stratification, with no significant differences among the four groups (p>0.05). Moreover, no differences in any other baseline characteristics were discovered among these four groups (p>0.05, Table 1 ).

Our results showed that the positive expression proportions of CD34, CD117 and CD123 were significantly higher in the CD7+ group than in the CD7- group (p<0.05, Table 2 ); the positive expression proportions of CD4, CD56 and CD19 were significantly lower in the CD7+ group than in the CD7- group (p<0.05; Table 2 ). The differences in the proportions of CD38, HLA-DR, CD15, CD33, CD2, CD64 and CD71 expression between the two groups were compared separately, and the results were not significantly different (p>0.05, Table 2 ).

According to the 2022 ELN genetic risk stratification, we divided AML patients into favorable (n=36), intermediate (n=242), and adverse (n=19) karyotype groups. Within the intermediate karyotype group, the CEBPA bZIP mutation rate was significantly lower in the DNMT3A-mutated group (0.0%) than in the DNMT3A-wild-type group (12.0%); the FLT3-ITD mutation rate (32.2%), NPM1 mutation rate (47.5%) and IDH1/2 mutation rate (28.8%) were significantly higher in the DNMT3A-mutated group than in the DNMT3A-wild-type group (11.5%, 13.1%, 10.9%), the differences were statistically significant (p<0.05, Table 3 ). However, no differences in gene mutation rates were observed between the DNMT3A-mutated group and the DNMT3A-wild-type group in the favorable and adverse karyotype groups ( Supplementary Table 2 ). Our further analysis revealed that the DNMT3A-mutated/CD7+ group still exhibited a higher FLT3-ITD mutation rate compared with the non-DNMT3Amut/CD7+ group ( Supplementary Table 3 ). In addition, we generated an oncoprint to display the distribution of gene mutations in AML patients ( Figure 1 ).

AML patients were divided into two groups based on their DNMT3A mutation status; the median OS time was 22 months for DNMT3A-mutated patients and 32 months for DNMT3A-wild-type patients, with no statistically significant difference (p>0.05, Figure 2A ); similarly, the median RFS of the DNMT3A-mutated group was not significantly different from that of the DNMT3A-wild-type group (10 months vs. 22 months, p>0.05, Figure 2B ). AML patients were divided into two groups according to CD7 expression; the median OS time was 24 months in the CD7+ group and 51 months in the CD7- group, and the difference in OS between the two groups was not significantly different (p>0.05, Figure 2C ); the median RFS time was 20 months and 19 months in the CD7+ and CD7- groups, respectively, and the difference was also not significantly different (p>0.05, Figure 2D ).

The median OS time for patients in the DNMT3A-mutated/CD7+ group was 6 months, whereas that in the DNMT3A-mutated/CD7- group was 51 months (p=0.003, Figure 2E ); similarly, the median OS in the DNMT3A-mutated/CD7+ group was significantly shorter than that in the DNMT3A-wild-type/CD7+ and DNMT3A-wild-type/CD7- groups (6 months vs. 36 months, p=0.001; 6 months vs. 32 months, p=0.001, Figure 2E ). In addition, the median RFS was significantly shorter in the DNMT3A-mutated/CD7+ group than in the other three groups (3 months vs. 42 months, p=0.003; 3 months vs. 28 months, p<0.001; 3 months vs. 19 months, p<0.001, Figure 2F ). Therefore, we reclassified AML patients into the DNMT3A-mutated/CD7+ group (n=21) and the other (non-DNMT3Amut/CD7+) group (n=276) based on whether they had concomitant DNMT3A mutation and positive CD7 expression. Comparing the OS and RFS of patients in the two groups separately, we found that the DNMT3A-mutated/CD7+ group had significantly shorter OS and RFS (p<0.001, Figures 2G, H ).

Among the entire cohort of 297 patients, CR (CR/CRi) was achieved in 233 patients (78.5%) after two cycles of induction chemotherapy.

CR was achieved in 12 patients (57.1%) in the DNMT3A-mutated/CD7+ group compared to 221 patients (80.1%) in the non-DNMT3Amut/CD7+ group, and the CR rate was significantly lower in the DNMT3A-mutated/CD7+ group than in the non-DNMT3Amut/CD7+ group (p=0.003). Of the 233 patients with CR, 97 experienced disease recurrence, including 9 (75.0%) in the DNMT3A-mutated/CD7+ group and 88 (39.8%) in the non-DNMT3Amut/CD7+ group. Patients in the DNMT3A-mutated/CD7+ group presented a higher risk of recurrence (p=0.032, Table 4 ). Consistent with this, we found that DNMT3A-mutated/CD7+ patients still showed a low CR rate and a high relapse rate among patients who received intensive induction therapy ( Supplementary Table 4 ).

We analyzed various prognostic factors affecting OS and RFS in AML patients. Univariate Cox regression analysis showed that age >60 years, WBC count ≥100 × 10⁹/L, no HSCT, and DNMT3A-mutated/CD7+ were risk factors for poorer OS and RFS in patients with newly diagnosed AML. Multifactorial Cox regression analysis showed that DNMT3A-mutated/CD7+ remained significantly associated with shorter OS and RFS ( Table 5 ). In addition, we further analyzed the factors affecting prognosis in patients who did not undergo HSCT and those who received intensive induction therapy, and multifactorial Cox regression analysis still suggested that DNMT3A-mutated/CD7+ was an independent risk factor for AML patients ( Supplementary Table 5 ).