The clinical impact of IKZF1 mutation in acute myeloid leukemia

Abstract

Genetic heterogeneity poses a great challenge to the understanding and management of acute myeloid leukemia (AML). Knowledge of the IKZF1 mutation in AML specifically is extremely limited. In a previous work, we described the distribution pattern of IKZF1 mutation in AML, but its clinical impact has remained undefined due to the limited number of cases. Herein, we attempt to answer this question in one relatively large cohort covering 522 newly diagnosed AML patients. A total of 26 IKZF1 mutations were found in 20 AML patients (20/522, 3.83%). This condition has a young median age of onset of morbidity (P = 0.032). The baseline characteristics of IKZF1-mutated and wild-type patients were comparable. IKZF1 mutation showed significant co-occurrences with CEBPA (P < 0.001), SF3B1 (P < 0.001), and CSF3R (P = 0.005) mutations, and it was mutually exclusive with NPM1 mutation (P = 0.033). Although IKZF1-mutated AML was more preferably classified into the intermediate-risk group (P = 0.004), it showed one inferior complete remission rate (P = 0.032). AML with high burden of IKZF1 mutation (variant allele frequency > 0.20) showed relatively short overall survival period (P = 0.012), and it was an independent factor for the increased risk of death (hazard ratio, 6.101; 95% CI 2.278–16.335; P = 0.0003). In subgroup analysis, our results showed that IKZF1 mutation conferred poor therapeutic response and prognosis for SF3B1-mutated AML (P = 0.0017). We believe this work improves our knowledge of IKZF1 mutation.

IKZF1 mutation is one rare but recurrent alteration in AML. In a previous work, we described its distribution pattern in AML [1], but the clinical impact of IKZF1 mutation on AML remains undefined. We here address this issue in a cohort of 522 newly diagnosed AML patients (Additional file 1: Fig S1, Patients and methods in supplementary information).

Recurrent IKZF1 mutation, including 12 missense mutations, 4 nonsense mutations, and 10 frame-shift mutations, was found in 20 patients (3.83%). Missense mutation preferred to localize at the exon 5 (91.67%), which mainly influences the DNA binding of IKZF1. A total of 35.7% of nonsense and frame-shift mutations were found to disrupt the DNA-binding domain and caused loss of the dimerization domain, while 64.3% of them only disrupted the dimerization domain (Fig. 1A, Additional file 4: Table S1). As indicated, IKZF1 mutation was recurrent in AML, but its role in AML pathogenesis needed further investigations.

Fig. 1

IKZF1 mutation in AML. (A) The distribution of IKZF1 mutations, which were identified in our cohort, on the protein. The nonsense or frameshift mutation was marked as red, while the missense mutation was marked as blue. (B, C) The OS (B) and RFS (C) of IKZF1^WT and IKZF1^MUT groups in our AML cohort. (D, E) The influence of IKZF1 mutation burden on the prognosis of AML was studied, and the OS (D) as well as RFS (E) of IKZF1^WT, IKZF1^MUT with VAF > 0.20, and IKZF1^MUT with VAF ≤ 0.20 groups are shown. (F) The difference of additional mutations distribution in IKZF1^WT and IKZF1^MUT groups, and the percentage of each gene mutation is exhibited. (G) The distribution of frequent AML-associated gene mutations in IKZF1^WT and IKZF1^MUT groups, and the count as well as percentage of each gene mutation are shown. (H, I) The prognostic role of combined IKZF1 and SF3B1 mutations on AML was investigated, and the OS (H) as well as RFS (I) of AML with different IKZF1 or SF3B1 mutated status are exhibited

To investigate the features of IKZF1^MUT AML, we compared the baseline characteristics of the IKZF1^MUT and IKZF1^WT groups, and the only difference was found in median age. ELN 2017 prognostic stratification predicted the clinical outcome of AML patients well [2]. Compared to the IKZF1^WT group, the IKZF1^MUT group showed a higher frequency of patients in the ELN-intermediate-risk group and a lower frequency in the ELN-low-risk and ELN-high-risk groups, but the CR rate in the IKZF1^MUT group was significantly lower than that in the IKZF1^WT group under our treatment strategy (Table 1). More interestingly, IKZF1^MUT patients showed similar OS and RFS with IKZF1^WT patients (Fig. 1B, C). Though IKZF1 mutation conferred one disadvantaged therapeutic response for AML patients, overall, it finally did not influence their survival time.

Table 1 Baseline characteristics of our AML cohort

To interpret the contrast phenomena and define the prognostic role of IKZF1 mutation more clearly, we analyzed the influence of its VAF, mutational type, and mutational count on the duration of survival. We performed maximally selective log-rank statistics in OS based on VAF and found that IKZF1^MUT patients with a high IKZF1 VAF burden (VAF > 0.20) showed significantly poorer OS than those with low VAF or IKZF1^WT, but the RFS did show any statistically significant difference (Fig. 1D, E, Additional file 5: Table S2). We found that neither the type nor the number of mutations influenced OS or RFS in IKZF1^MUT patients (Additional file 1: Fig S1C–F). In this way, a high burden of IKZF1 mutation might predict poor prognosis in AML.

To exclude the impact of additional factors on OS, we performed univariate and multivariate analyses that included baseline characteristics and genetic alterations. In univariate analysis, we identified 20 factors that had a significant influence on OS in our AML cohort, including IKZF1 mutations with high VAF. In multivariate analysis, we strongly indicated that IKZF1 mutation with high VAF was one independent risk factor for the death of AML (HR, 6.101; 95% CI 2.278–16.335; P = 0.0003) (Additional file 6: Table S3).

We also analyzed the relationships among IKZF1 mutation and other gene mutations. IKZF1 mutation exhibited concurrences with CEBPA, SF3B1, and CSF3R mutations, but it was mutually exclusive with NPM1 mutation (Fig. 1F, G). We also performed subgroup survival analysis. The prognostic role of CEBPA^bZIP−inf [3,4,5], SF3B1, and CSF3R mutations was revealed in our cohort (Additional file 2: Fig S2). IKZF1 mutation did not influence OS or RFS in CSF3R^WT and CSF3R^MUT (Additional file 3: Fig S3A, B, Additional file 7: Table S4). In IKZF1^MUT patients, CEBPA^bZIP−inf mutation (83.3%) was more common than non-CEBPA^bZIP−inf mutation (16.7%). IKZF1 mutation conferred a relatively low CR in the CEBPA^WT/non-CEBPA^{bZIP−inf−MUT} group, but not in the CEBPA^{bZIP−inf−MUT} group (Additional file 8: Table S5), and it influenced OS and RFS in the CEBPA^WT/non-CEBPA^{bZIP−inf−MUT} group but not in the CEBPA^{bZIP−inf−MUT} group (Additional file 3: Fig S3C, D). IKZF1^WT/SF3B1^MUT AML patients exhibited a CR rate of 50%, and the therapeutic response was even worse in IKZF1^MUT/SF3B1^MUT AML. None of these patients achieved CR at any point during the regimen (Additional file 9: Table S6). IKZF1 mutation combined with SF3B1 mutation conferred extremely poor OS on AML, but the RFS of IKZF1^MUT/SF3B1^MUT AML patients was unavailable because no patient reached CR (Fig. 1H, I).

Compared with foreign cohorts (OHSU [6], 1.35%; TCGA [7], 0.5%; TARGET [8], 4.21%), the frequency of IKZF1 mutation was relatively high (3.83%). This may be because patients were of different races or it may be because of differences in sequencing depth. IKZF1 deletion, caused by -7/monosomy 7, was detected in 3.20% of our patients. Unlike in ALL [9], IKZF1 mutation and deletion were equally dominant in AML [10]. Missense mutation accounted for nearly half of IKZF1 mutations, and it almost affected the DNA-binding domain in AML, while its DNA-binding domain and dimerization domain involvement was relatively balanced in ALL [9]. IKZF1 aberration conferred poor prognosis in ALL [11], but only a high burden of IKZF1 mutation predicted poor OS in AML because IKZF1 mutation with VAF < 10% accounted for 35% of all IKZF1^MUT patients, and IKZF1 mutation contributed less to disease than other mutations did in this group of patients. CEBPA mutation was the most common co-mutation that occurred alongside IKZF1 mutation in AML [1, 12].