RNA-based FLT3-ITD allelic ratio is associated with outcome and ex vivo response to FLT3 inhibitors in pediatric AML

Internal tandem duplication of FMS-like tyrosine kinase type 3 (FLT3-internal tandem duplication [ITD]) is a recurrent aberration used for risk stratification in acute myeloid leukemia (AML).^1-3  The length of the duplication is variable⁴  and may be associated with the extent of constitutive FLT3 signaling leading to increased cell proliferation.⁵ 

Using fragment length analysis, a relatively high allelic ratio (AR) has been associated with poor outcome.^6-9  No consensus has been reached as to whether AR and/or the length of the ITD have a prognostic impact on outcome or whether ITD-AR analyses should be performed using DNA or RNA.

The current study (1) compares ITD measurements based on RNA and DNA, (2) establishes the role of both ITD-AR and ITD length for further risk classification, and (3) determines whether ex vivo sensitivity to selected FLT3-inhibitors depends on AR and ITD length.

Diagnostic bone marrow samples were obtained from 172 AML patients included in the Dutch-Belgian Pediatric AML Protocol for Children With Newly Diagnosed Acute Myeloid Leukaemia Based on the NOPHO-AML 2004 Study (DB-AML01)¹⁰  (n = 108) or NOPHO-DBH-AML 12 (n = 64) trial. FLT3-ITD-AR was measured on both DNA^6,11  and RNA¹²  and correlated with both clinical outcome and sensitivity to FLT3 inhibitors (gilteritinib and midostaurin).

Thirty-six (21%) patients were positive for the FLT3-ITD mutation at RNA level. One patient was positive at the DNA level only, possibly because of low expression of the FLT3-ITD gene. In 1 patient, an ITD was detected at the RNA level but repeatedly not at DNA level (supplemental Table 1, available on the Blood Web site), possibly because the very low AR (0.01) represents a minor subclone.¹³  ITD-AR of DNA correlated with RNA (AR_spearman = 0.62; 95% confidence interval [CI], 0.22-0.87; Figure 1A). Likewise, ITD length on DNA was highly correlated with RNA (ITDlength_spearman = 0.98; 95% CI, 0.90-1.00; Figure 1B). There was no significant correlation between FLT3-ITD-AR and length (R_spearman = 0.34, P = .063).

In 4 samples, ITD length at the DNA level was approximately 90 base pairs (bps) longer than at the RNA level (supplemental Table 1). DNA and RNA sequencing of 3 of these patients revealed a 90-bp intron between exons 14 and 15 on DNA, which was spliced out and not included in RNA measurements (Figure 1C). Overall, AR was comparable using either DNA or RNA, suggesting no large differences between DNA content and subsequent gene expression, which is in line with recently published data.⁹ 

Survival analysis was performed only on data of patients from the DB-AML01 protocol because of sufficient follow-up time (supplemental Tables 2 and 3). Within the group of FLT3-ITD patients, gender, age, blast percentage, FAB classification, complex karyotype rates, and number of NPM1 and FLT3-TKD mutations (supplemental Table 3) were similar among FLT3-ITD-low and high AR patient groups (data not shown).

FLT3-ITD patients had significantly lower event-free survival (EFS) than FLT3-ITD-wild-type (WT) patients (P_logrank < .001; supplemental Figure 1A). The difference in overall survival (OS) between FLT3-ITD and FLT3-WT AML patients was borderline significant (P_logrank = .066; supplemental Figure 1A).

Using received operating characteristic analysis, the optimal cutoff point for prediction of death (at the end of the follow-up time and after 1 year) was an ITD-AR of 0.5. In patients with FLT3-ITD-AR ≥ 0.5 as measured on RNA, OS was significantly shorter compared with FLT3-ITD-AR < 0.5 (P_logrank = .008; Figure 2A) and compared with FLT3-WT patients (P_logrank < .001). Patients with FLT3-ITD-AR < 0.5 and FLT3-WT patients had similar OS (P_logrank = 1). EFS and relapse-free survival in patients with FLT3-ITD-AR ≥ 0.5 compared with FLT3-ITD-AR < 0.5 were not significantly different (Figure 2B; supplemental Figure 1C). Patients with both high and low ITD-AR (on RNA) had significantly worse EFS compared with FLT3-WT (P_logrank = .016 and P_logrank = .008, respectively; Figure 2B). Based on DNA measurements, no cutoff level could show a significant association with survival (Figures 2C-D).

Except for 3 patients, all FLT3-ITD⁺ patients received allogeneic stem cell transplantation. Without these patients, the survival analysis remained similar, suggesting that FLT3-ITD-WT and FLT3-ITD-low AR patients are more successfully salvageable and high AR patients would benefit from therapy able to reduce the burden of FLT3-ITD⁺ cells.

ITD length has been related to the extent of autophosphorylation, which negatively affects OS because of a growth advantage,¹⁴  whereas others report no significant relation between ITD length and outcome.^9,15,16  Based on our data, ITD length ≥48 bp was associated with shorter OS compared with FLT3-ITD patients having an ITD length <48 bp (P_logrank = .029; Figure 2E). EFS and relapse-free survival were both statistically different between patients with relatively long and short ITD (Figure 2F; supplemental Figure 1D). The number of patients harboring FLT3-ITD did not allow multivariate analysis. However, using Cox regression, it could be shown that −2log likelihood of FLT3-ITD-AR is slightly better than length itself and that adding the latter parameter to the model did not significantly improve the model (P_LR = .078). Hence, AR is the most significant prognostic parameter, but, interestingly, ITD length has relevance as well in this study and should be investigated in more detail.

Ten patients had a single NPM1 mutation, whereas 6 patients had mutations in both NPM1 and FLT3-ITD. The effect of NPM1 mutations on the prognostic significance of FLT3-ITD-AR ≥ 0.5 could not be determined because there was only 1 patient with both a FLT3-ITD-AR ≥ 0.5 and mutated NPM1 (supplemental Table 2).

FLT3-ITD is a prognostically relevant molecular aberration.^8,11,17,18  Tyrosine kinase inhibitors are being investigated in clinical trials, with varying results,^19-22  but the literature is inconclusive about the effect of AR on FLT3-inhibitor sensitivity.^23,24  Exposure of primary AML samples to 0.05 µM gilteritinib resulted in significantly decreased clonogenic capacity of FLT3-ITD-AR ≥ 0.5 samples compared with FLT3-WT and FLT3-ITD-AR < 0.5 samples (Figure 1G). This effect was present to a lesser extent in specimens treated with midostaurin, possibly explained by aspecific binding of multiple kinases with suboptimal inhibition of FLT3 by midostaurin, whereas gilteritinib specifically and potently targets FLT3 (Figure 1H). At higher dosages, the nonsignificant differences can be explained by more off-target effects and the cell numbers becoming too low. Moreover, 5 patients with relatively low AR (0.29-0.48) were analyzed according to ITD length, and long ITD responded significantly better to gilteritinib, suggesting that ITD length is also relevant for response to FLT3 inhibition in patients with low FLT3-ITD-AR, whereas in patients with high FLT3-ITD-AR, the ITD length had less effect on response (supplemental Figure 2).

Our data indicate that FLT3-ITD-AR is associated with survival in pediatric AML, which is essential for risk assessment. Although DNA and RNA measurements show good concordance, the FLT3-ITD-AR on RNA is recommended because of superior prognostic value and accurate messenger RNA length measurement. In addition, FLT3-ITD-AR might be a potential marker for selecting patients for treatment with tyrosine kinase inhibitors but should be investigated more thoroughly. Moreover, our findings suggest the relevance of ITD length in both outcome and response to FLT3 inhibitors, which warrants further research.

This work was supported by a grant from the Egbers Foundation.

Contribution: D.G.J.C., B.D., J.P., G.J.S., J.C., and E.S. designed the study and performed experiments; B.I.L.-W. assisted in statistical analysis and reviewed the manuscript; G.J.L.K., V.d.H., J.J.W.M.J., S.Z., G.J.O., C.M.Z., M.M.v.d.H.-E., B.d.M., F.V., K.V. and E.S.J.M.d.B. provided clinical samples and data and reviewed the manuscript; Z.K. and T.C. performed experiments; and D.G.J.C. and J.C. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jacqueline Cloos, Pediatric Oncology/Hematology and Hematology, VU University Medical Center, De Boelelaan 1118, 1081 HV Amsterdam, The Netherlands; e-mail: j.cloos@vumc.nl.