GATA2 zinc finger 1 mutations are associated with distinct clinico-biological features and outcomes different from GATA2 zinc finger 2 mutations in adult acute myeloid leukemia

Mutations of the GATA binding protein 2 (GATA2) gene in myeloid malignancies usually cluster in the zinc finger 1 (ZF1) and the ZF2 domains. Mutations in different locations of GATA2 may have distinct impact on clinico-biological features and outcomes in AML patients, but little is known in this aspect. In this study, we explored GATA2 mutations in 693 de novo non-M3 AML patients and identified 44 GATA2 mutations in 43 (6.2%) patients, including 31 in ZF1, 10 in ZF2, and three outside the two domains. Different from GATA2 ZF2 mutations, ZF1 mutations were closely associated with French-American-British (FAB) M1 subtype, CEBPA double mutations (CEBPAdouble-mut), but inversely correlated with FAB M4 subtype, NPM1 mutations, and FLT3-ITD. ZF1-mutated AML patients had a significantly longer overall survival (OS) than GATA2-wild patients and ZF2-mutated patients in total cohort as well as in those with intermediate-risk cytogenetics and normal karyotype. ZF1 mutations also predicted better disease-free survival and a trend of better OS in CEBPAdouble-mut patients. Sequential analysis showed GATA2 mutations could be acquired at relapse. In conclusion, GATA2 ZF1 mutations are associated with distinct clinico-biological features and predict better prognosis, different from ZF2 mutations, in AML patients.

Somatic GATA2 mutations mainly cluster in the two zinc finger (ZF) domains, which can occupy GATA DNA motif in thousands of genes 9 . The patterns of somatic GATA2 mutations differ among myeloid diseases. ZF1 mutations predominate in AML, and ZF2 mutations are frequently identified in CML blastic phase 3 . GATA2 mutations are strongly associated with CEBPA double mutations (CEB-PA double-mut ) 9,10,12 . However, discrepancies exist among different reports regarding prognostic impact of GATA2 mutations in AML patients 10,13 . We hypothesize that mutations in different domains of GATA2 may have distinct impact on clinico-biological features and outcomes in AML patients, like IDH2 mutations in which IDH2 R172 is associated with gene mutations and clinical outcomes different from other IDH mutations 14 . However, little is known about this issue till now.
In this study, we investigated the clinical and prognostic relevance of mutations in different GATA2 domains in a large cohort of 693 unselected de novo non-M3 AML patients. To our knowledge, this is the first study to show GATA2 ZF1 mutations are associated with distinct clinical features, gene mutations, and outcomes different from ZF2 mutations. Longitudinal follow-ups were also performed in 419 samples from 124 patients to evaluate the dynamic changes of the mutations. Furthermore, we analyzed the global gene expression profiles in 328 patients to interrogate the possible molecular pathways associated with mutations in different GATA2 domains.

Subjects
We consecutively enrolled 693 newly diagnosed de novo non-M3 AML patients at the National Taiwan University Hospital (NTUH) from 1994 to 2011. Diagnosis and classification of AML were made according to the FAB Cooperative Group Criteria and the 2016 WHO classification 15 . To focus on a more homogeneous group of patients with de novo AML, those with antecedent hematological diseases, history of cytopenia, and family history of myeloid neoplasms or therapy-related AML were excluded 16 . Survival analyses were performed in 469 (67.7%) patients who received standard chemotherapy. This study was approved by the Institutional Review Board of the NTUH, and written informed consents were obtained from all participants in accordance with the Declaration of Helsinki.

Cytogenetics
Chromosomal analyses were performed as described previously 17 . Karyotypes were classified using Medical Research Council (MRC) risk groups 18 .

Functional annotation analysis of GATA2 mutationregulated genes
We analyzed the differentially expression genes associated with GATA2 mutations by the knowledge-based Ingenuity Pathway Analysis (IPA) (Qiagen, Redwood City, CA) software for associated functions. We also used Gene Set Enrichment Analysis (GSEA) software to investigate systematic enrichments of GATA2 mutation-governed expressional profile in biological functions 33 . Statistical significance of the degree of enrichment was assessed by a 1000-time random permutation test.

Statistical analysis
The discrete variables were compared using the χ 2 tests, but if the expected values of contingency tables were <5, Fisher's exact test was used. Mann-Whitney U tests were used to compare continuous variables and medians of distributions. Overall survival (OS) was measured from the date of first diagnosis to the date of last follow-up or death from any cause. Disease-free survival (DFS) was measured from the date of diagnosis until treatment failure, relapse from CR, or death from any cause, whichever occurred first. To ameliorate the influence of hematopoietic stem cell transplantation (HSCT) on survival, DFS and OS were censored at the time of HSCT in patients receiving the treatment 34 . Multivariate Cox proportional hazard regression analysis was used to investigate independent prognostic factors for OS and DFS. A P value <0.05 was considered statistically significant. All statistical analyses were performed with the SPSS 18 (SPSS Inc., Chicago, IL, USA) and StatsDirect (Cheshire, England, UK).

GATA2 mutations in patients with AML
Excluding two single-nucleotide polymorphisms (A164T, M400T) 35 and eight missense mutations (N114T, M223I, P250A, A256V, L315P, C319F, V369A, S429T) with unknown biologic significance (because they were not reported previously and could not be verified because of lack of matched bone marrow samples in CR), we identified 44 distinct GATA2 mutations in 43 (6.2%) of 693 patients (Fig. 1). Forty GATA2 mutations were missense mutations. The other four were in-frame deletion or duplication: p.Ser201*(c.598_599insG) in two, p. Thr387_Gly392del (c.1160_1177delCCATGAAGAAGGA AGGGA) and G210dup (c.631_632insGCG) in one each. With regard to the functional sites, 31 mutations were clustered in the highly conserved N-terminal ZF domain (ZF1 domain), and other 10 mutations were within Cterminal ZF domain (ZF2 domain). The remaining three mutations scattered outside of the ZF domains. The most common mutations were A318V (n = 4), followed by L321F and A318T (n = 3 each). p.Ser201* (c.598_599insG), N297S, A318G, G320V, L321H, and K324E occurred in two patients each. All other mutations were detected in only one patient each (Table 1). Only one patient had two GATA2 mutations (patient no. 20). All mutations were heterozygous. The mutant burden ranged from 4.89 to 52% with a median of 39.07% in ZF1 mutations, and from 10.74 to 50.26% with a median of 36.16% in ZF2 mutations.

Impact of different GATA2 domains mutations on treatment response and clinical outcomes
Of the 469 AML patients, including 27 GATA2 ZF1mutated and nine GATA2 ZF2-mutated patients, undergoing conventional intensive induction chemotherapy, 352 (75.1%) patients achieved a CR. The CR rate was 85.2% in ZF1-mutated patients and 60% in ZF2-mutated patients ( Table 2). The relapse rate was similar between the two groups.
Sequential studies of GATA2 mutations in AML patients GATA2 mutations were serially studied in 419 samples from 124 patients who had ever obtained a CR and had available samples for study, including 19 patients with and 105 patients without GATA2 mutations at diagnosis (Table 4). Among the 19 GATA2-mutated patients who had paired samples, all lost the original GATA2 mutations at remission. Five of the six patients regained the original GATA2 mutations at first relapse, but one (no. 27) lost the Median (range) f Only the 469 patients, including 27 with GATA2 ZF1 domain mutations, nine with GATA2 ZF2 domain mutations, and 431 without, who received conventional intensive induction chemotherapy and then consolidation chemotherapy if CR was achieved, as mentioned in the text, were included in the analysis mutation. In the former five patients, the mutation burden, compared to that at diagnosis, was increased in one patient (no. 25), decreased in two (nos. 13 and 16), and stable in the remaining two (nos. 5 and 9). One patient (no. 9) retained the co-occurring ASXL1 mutations at CR status. Among the 105 patients who had no GATA2 mutations at diagnosis, four patients (nos. 44, 45, 46, and 47) acquired novel GATA2 mutations at relapse (Table 4).

GATA2 expression and biological functions associated with GATA2 mutations
We analyzed the microarray dataset of 328 patients studied to assess the impact of GATA2 mutations on gene expression and biological functions. By comparing the mRNA expression profiles between patients with and without GATA2 mutations, we found GATA2 expression levels were higher in those with GATA2 mutations (P = 0.003). More specifically, both ZF1 and ZF2 mutations correlated with higher GATA2 expression level compared to GATA2 wild-type. GATA2 mutations were associated with significant differential expression of 159 probes (t-test, P < 0.05 and >2-fold change). IPA analysis revealed different molecular networks between the GATA2 ZF1 and ZF2-mutated group (Supplementary Figure 3). We also performed the GSEA analysis to identify biological functions associated with genes significantly enriched in GATA2-mutated AML, compared with GATA2-wild AML. Three-hundred and thirteen patients with wild-type GATA2, 12 patients with GATA2 ZF1 mutations, and three patients with GATA2 ZF2 mutations were analyzed. We identified significant underrepresentation of genes hyper-methylated in AML (P = 0.006; normalized enrichment score (NES) = −1.49; Supplementary Figure  4A) and genes related to apoptosis (P = 0.042; NES = −1.33) in the ZF1-mutated patients compared to GATA2 wild-type patients. ZF2-mutations were associated with  Figure  4B). Comparing with ZF2-mutated AML, we identified significant overrepresentation of genes related to myeloid leukocyte differentiation (P = 0.042; NES = 1.36) and underrepresentation of genes hyper-methylated in AML (P = 0.029; NES = −1.37) in the ZF1-mutated AML.

Discussion
To the best of our knowledge, this is the first study to explore differences in clinical and biological implications between the GATA2 ZF1 and ZF2 mutations in AML patients. We found that mutations in different domains were associated with distinct clinical features, co-occurring mutations and outcomes (Supplementary Table 3).

GATA2-wild type, n=292
GATA2 ZF2 mut, n=9 GATA2 ZF1 mut, n=27 The GATA2 mutation landscape in adult de novo AML differs from that in blastic crisis of CML 3 , familial MDS/ AML 4 , and pediatric AML 5 . In adult AML, ZF1 mutations predominate, while ZF2 mutations are reported sporadically 10,36,37 . In concordance with the findings, twothirds of the 44 distinct GATA2 mutations in our study were located in the ZF1 domain. We also reported two novel missense mutations in ZF2 domain (L359V and G366R) that had not been reported before in adult de novo AML patients, but ever identified in blastic crisis of CML.
AML with CEBPA double-mut has been included as a definite entity in the 2016 WHO Classification of Myeloid Neoplasms 15 . It is well established that GATA2 mutations frequently co-occur with CEBPA double-mut with an incidence of 18-41% 9,10,12 and the two proteins show direct protein-protein interaction 38 . Further study revealed GATA2 ZF1 mutants, but not the ZF2 L359V that is commonly seen at the progression of CML to blast crisis, had reduced capacity to enhance CEBPA-dependent activation of transcription 9 . Based on this functional study and the frequent co-occurrence of CEBPA double-mut and ZF1 mutations, but not ZF2 mutations, in AML patients, it is possible that GATA2 ZF1 mutations and CEBPA double-mut interact together to induce leukemogenesis. In addition, we found ZF1 mutations were associated with lower incidences of NPM1 mutations and FLT3-ITD than wild-type GATA2, different from ZF2 mutations as ZF2-mutated patients had similar incidences of these two mutations to those in GATA2-wild patients. GATA2 ZF1 and ZF2 mutations may induce AML through different oncogenic mechanisms and have distinct impact on clinical outcomes. Truly, in this study, we demonstrated that patients with GATA2 ZF1 mutations had a significantly longer OS than ZF2-mutated patients in total cohort, as well as in patients with intermediate-risk cytogenetics and normal karyotype.
The prognostic impact of GATA2 mutations in CEB-PA double-mut patients was conflicting 12,13,37,39 . Greif et al. and Theis et al. found that GATA2 mutations did not impact clinical outcome in CEBPA double-mut patients. On the contrary, GATA2 mutations correlated with improved survival among CEBPA double-mut patients in other reports 12,13 . In a study of Theis et al., 31 (74%) of GATA2 mutations were detected in ZF1 domain, and 11 (26%) in ZF2 domain. They did not show different clinical outcomes with respect to GATA2 ZF1 and ZF2 mutations in a cohort with both CEBPA double-mut and CEBPA single-mut patients 39 . We were the first to investigate the prognostic implication of GATA2 ZF1 mutations in CEBPA double-mut patients and showed its association with a better DFS and a trend of longer OS than wild-type GATA2 among the CEBPA double-mut subgroup.
The poor prognostic impact of GATA2 ZF2 mutations was also witnessed in blast crisis CML patients as in de novo AML patients shown in this study 4 . The reason that ZF1 and ZF2 mutations had different survival impacts on de novo AML patients might be partially explained by their difference in association with CEBPA double-mut , and by different oncogenic mechanisms. Further studies are warranted to explore the underlying mechanisms of the differences.
The study also recruited the largest number of de novo AML patients for sequential analyses of GATA2  In conclusion, GATA2 ZF1 mutations, but not ZF2 mutations, are closely associated with CEBPA double-mut , and inversely correlated with NPM1 mutations and FLT3-ITD. The two GATA2 ZF domain mutations have different impacts on OS in AML patients. GATA2 ZF1 mutations also affect clinical outcome in CEBPA double-mut The results of serial studies in 101 patients without GATA2 mutation at both diagnosis and relapse were not shown in this table b Interval between the two successive statuses patients. Incorporation of GATA2 ZF1, not ZF2 mutations, allows further refinement of the WHO Classification in the specific entity of AML with CEBPA double-mut .