Introduction

Acute myeloid leukemia (AML) is the predominant acute leukemia among adults, characterized by an accumulation of malignant immature myeloid precursors. It represents a heterogeneous group of diseases with different responses to treatment, which can be separated by genetic abnormalities [1]. The MECOM locus in chromosome band 3q26.2 gives rise to two major mRNA and protein species, EVI1 and MDS1/EVI1, of which the former has been characterized for more extensively. The ecotropic viral integration site 1 (EVI1) gene is a proto-oncogene that encodes a zinc-finger DNA-binding protein previously detected in some AML and myelodysplastic syndromes (MDS) [2, 3], but not in normal marrow and cord blood cells. Enhanced expression of EVI1 is associated with a very poor disease outcome [4, 5], despite allogeneic hematopoietic stem cell transplantation [6]. Experimental studies suggest EVI1 blocks cellular differentiation by binding to GATA-1 or other specific DNA sequences controlling gene expression, and may be involved in the pathogenesis of some AMLs [7]. EVI1 influences transcription regulation in response to the myeloid differentiation inducing agent, all-trans-retinoic acid (ATRA) [8]. The EVI1 gene is consistently expressed in acute promyelocytic leukemia (APL) cells either constitutively or after ATRA therapy [9]. ATRA regulates EVI1 expression in blood cells [8].

ATRA is the paradigm of treatment in APL [10]. In addition to allowing transcriptomic activity downstream of retinoid receptor alpha (RARa), ATRA allows its degradation and acts on the pool of leukemic stem cells (LSCs) [11]. Because treatment of APL patients with ATRA is very successful, it has been hypothesize that ATRA might also be effective in treatment of other AML subtypes. Several clinical trials have already evaluated the combination of chemotherapy with ATRA in non-APL AMLs. Large randomized trials failed to observe an advantage to adding ATRA to induction chemotherapy [12,13,14,15,16]. Conversely, the ULM Study Group demonstrated an advantage to receiving ATRA with induction chemotherapy [17], and explorative subgroup analyses revealed better survival for genetic low-risk patients according to European LeukemiaNet (ELN) recommendations [18].

Several recent studies have shown that ATRA may efficiently drive leukemic cells into differentiation and/or apoptosis in a subset of AML patients with an NPM1 mutation [19], a FLT3-ITD [20, 21], and/or an IDH-1 mutation [22]. Furthermore, it was recently demonstrated that enhanced expression of EVI1 in HL60 cells increased the response to ATRA, that the protein GDF15 is part of the ATRA-induced cell cycle block [8], and that the in vitro part of the EVI1-positive cases are sensitive to ATRA by inducing differentiation and cell death and decreasing leukemic engraftment [23].

Here, we retrospectively reviewed the response to ATRA of 13 high-risk AML patients with overexpression of EVI1. Our results showed that responses can be obtained and suggest that addition of ATRA to leukemia treatment might increase complete response (CR) rates or prevent relapses in EVI1-positive AML patients.

Methods

Between February 2016 and January 2017, 13 consecutive patients with high-risk AML presenting EVI1 abnormalities were treated by ATRA. High-risk AML patients were defined either as AML patients in front-line therapy considered unfit for intensive chemotherapy, or as relapsed/refractory AML patients unfit for intensive treatment and/or who relapsed after allogeneic stem cell transplantation (ASCT) or after treatment for secondary AML. ATRA was given at 45 mg/m2/day for two consecutive weeks each month and was administered initially either alone (3 patients) or combined with low-dose cytarabine (20 mg/day for 10 consecutive days with cycles of 28 days) (1 patient), azacitidine (75 mg/m2/day for 7 consecutive days with cycles of 28 days) (4 patients), gemtuzumab ozogamicin (3 mg/m2/day on days 1, 4, and 7) (1 patient), or 6-mercaptopurine (50–100 mg/day) in association with low-dose methotrexate (10–15 mg/week) (4 patients), according to patient situation. Combination therapy was decided according to patient physical condition and patient past therapeutic history. Informed consent was obtained from all patients. The study was approved by the review board protocol of the Hospices Civils de Lyon and conducted in accordance with the Declaration of Helsinki.

Results for cytogenetic analysis were available in 12 patients and were classified according to the ELN recommendations [1]. RNA isolation and EVI1 real-time quantitative polymerase chain reaction (RQ-PCR) conditions for study of EVI1 overexpression were considered as previously recommended [4]. The 3q26 amplified cell line SKOV3 overexpressing EVI1 served as a calibrator for quantification. Only standard curves established by serial dilutions of SKOV3 cDNA aliquots with correlation coefficients larger than 0.9 were taken into account. Equal amplification efficiencies of target and reference genes in both EVI1 samples and SKOV3 at different cDNA concentrations were considered. The relative EVI1 expression was calculated using the delta–delta CT method. EVI1 expression levels were dichotomized based on a cut-off of 0.1 relative to SKOV3. In all patients, no other molecular abnormalities were detected among biological markers systematically investigated at the time of diagnosis or relapse (NPM1, FLT3-ITD, FLT3-TKD, IDH1/2).

Response to therapy was evaluated by bone marrow (BM) aspirates. In responding patients, cytological analysis was completed by minimal residual disease (MRD) monitoring using multiparametric flow cytometry immunophenotyping and/or molecular biology as previously described [24]. Descriptive statistics were used to characterize patients and their disease.

Results

Patient characteristics at the time inclusion are reported in Table 1. Their outcomes are summarized in Table 2. The median age was 65 years (range 49–81). Three of the 13 patients were newly diagnosed. All other patients had previously received one line (8 patients) or two lines of treatment including anthracycline plus cytarabine-based intensive chemotherapy (8 patients) eventually followed by ASCT (2 patients), or low-intensity therapy with cycles of hypomethylating agents (2 patients). Most of the patients presented with morphological signs of cytologic dysplasia, of which five had secondary AML. At the time of inclusion, Eastern Cooperative Oncology Group (ECOG) performance status (PS) was ≥ 2 in six patients. Regarding cytogenetics, five patients displayed a normal karyotype, while seven had unfavorable-risk cytogenetics and one had cytogenetic failure.

Table 1 Patient characteristics at the time of inclusion
Table 2 Patient outcomes

Six patients did not achieve CR (Fig. 1a). Five (#1, #3, #8, #9, #10) died within 5 months following treatment initiation (median 3 months; range 1–5). One (#12) is still alive after 1.5 year with sequences of treatment (patient decision) using ATRA alone and then ATRA plus azacitidine. Among those six non-responding patients, three died before any evaluation could be performed, but three were evaluated for response: two patients (#3 and #12) showed decreased bone marrow blast infiltration after 5 months and 3 months of treatment, respectively, and one patient (#1) showed a decreased peripheral blast percentage, from 63 to 18%, after 1 month of therapy.

Fig. 1
figure 1

Evolution on ATRA therapy in patients who did not achieve CR (a) and in patients who achieved CR (b). 6MP 6-mercaptopurine, AE adverse event, AraC low-dose cytarabine, ATRA all-trans-retinoic acid, Aza azacitidine, BM bone marrow, chemo chemotherapy, Chim chimerism, CR morphological complete remission, flow flow cytometry, GO gemtuzumab ozogamicin, GvH graft-versus-host reaction, MRD minimal residual disease, MTX methotrexate, neg negative, PegIFN peginterferon, PB peripheral blood, PD patient decision, Pt patient

Seven patients achieved CR (Fig. 1b). Almost all of them had an initial low bone marrow leukemic burden (median 10%; range 5–55%). Patient #2, who maintained for a while a low minimal residual disease (MRD) on ATRA plus gemtuzumab ozogamicin and then ATRA alone, achieved, after a frank relapse, a cytological CR with four cycles of ATRA plus azacitidine. Patient #4 achieved CR after four cycles of ATRA plus low-dose cytarabine, while three patients (#5, #6, #13) obtained CR after 2–6 cycles on ATRA plus azacitidine, and one patient (#7) after 2 months on ATRA plus 6-mercaptopurine and methotrexate. Molecular remission was achieved in two patients (#4, #5) after treatment for 11 months and 6 months, respectively. At the time of analysis, four patients (#6, #7, #11, #13) had relapsed, two of which after stopping treatment because of major cutaneous rash (#7) or by patient decision (#13). Three patients (#2, #4, #5) are still in continuous CR, although treatment was stopped in two of them.

Overall, treatment was well tolerated. Toxicity attributed to ATRA only included cutaneous rash (grade 3) in one patient (#7) and cutaneous and mucosal dryness (grade 2) in another patient (#2).

Multicolor/multidimensional flow cytometry was used to characterize the different leukemia compartments and evaluate the importance of the immature CD34+CD38 cell population compared to the more mature CD34+CD38low and CD34+CD38+ leukemic cell subpopulations. Although the LSC population is neither uniform nor static within individual patients, the CD34+CD38 cell population, which certainly remains enriched in leukemia cells with self-renewal capacities, decreased significantly in patients responding to ATRA therapy (Fig. 2).

Fig. 2
figure 2

Strategy used for gating the blast cell population and CD34+CD38 cell subpopulations [red: immature CD45low/SSC blast cells; blue: CD34+ cells, green: lymphocytes, blue cyan: monocytes, violet: granulocytes, black: CD45+ cells (erythroblasts)]; CD34+ cells were then separated into different stem cell fractions based on their CD38 antigen expression: a first cell population expressing a large amount of the CD34 antigen and lack of CD38 (CD34+CD38) (yellow), a second cell population characterized by a large amount of CD34 antigen and by a low density of CD38 antigen (CD34+CD38low) (green), and a third cell population characterized by a high density of CD38 antigen and of CD34 antigen (CD34+CD38+) (blue). Blast cell morphology and immunostaining in patient #7 before ATRA introduction (55% of bone marrow blasts) (a), and after 2 months of ATRA therapy (< 5% blasts on bone marrow smears and MRD at 6% detected by flow cytometry) (b)

Discussion

Enhanced expression of EVI1 (MECOM) occurs in approximately 10% of AML patients and is associated with a very poor disease outcome. Here we show that, despite treatments and patient heterogeneity, a substantial part of EVI1-positive AML patients can respond to ATRA therapy. These first clinical results tend to confirm the potential efficacy of ATRA-based therapy with an induction of differentiation and a significant reduction of survival previously described in leukemic cells from patients with EVI1 overexpression [23]. In mice models, ATRA treatment was also associated with less leukemic engraftment suggesting a decreased clonogenic capacity with a direct effect of ATRA on LSCs overexpressing EVI1 [23]. EVI1 is able to modulate the ATRA response of several genes, and has been shown to reinforce the ATRA response in the majority of AML cases [8]. In the present study, we showed that ATRA alone or combined with various other treatments could have a beneficial effect in AML patients with overexpression of EVI1. CR and even molecular remission were obtained in patients with unfavorable prognostic factors (age, comorbidities, unfavorable cytogenetics, prior therapeutic lines). All patients were considered high-risk and primary unfit for intensive chemotherapy or secondary unfit after having received one or two lines of intensive or low-intensity therapy. These complete responses were mainly obtained in patients with a low leukemic burden. Only one patient with initially more than 50% of bone marrow blasts achieved CR. However, most of responses were short. This could be likely attributed to a suboptimal therapy, poor treatment compliance in a generally frail patient population and/or to a lost treatment efficacy after repeated treatment administrations. Overexpression of EVI1 is associated with a poor prognosis in AML probably due to its important role in the maintenance of LSCs [25]. Because of their quiescent state LSCs are not targeted by any antimitotic chemotherapy, and are at the origin of relapses conferring a poor outcome [26]. Targeting LSCs remains a main challenge in the treatment of AML. In APL, the association of ATRA with arsenic trioxide has demonstrated an LSC pool exhaustion [27] and is thereby able to cure most of the patients [28]. In our study, multiparameter flow cytometry analysis tended to show a diminution of the LSC pool, indicating that ATRA might act by targeting LSCs. Although no morphological cell differentiation was observed under ATRA therapy, response to therapy was accompanied in immunostaining by a massive expression of CD38, which is known as a regulator of induced cell differentiation and growth arrest [29]. However, complementary treatment strategies are required to improve ATRA responsiveness in EVI1-positive AML. Recently, a combined effect of ATRA and bromodomain inhibitor JQ1 has been demonstrated on non-APL AML cells with a synergistic growth inhibition resulting from differentiation or apoptosis [30]. Similarly, inhibition of the SUMO pathway appeared as a promising strategy to sensitize patients with non-APL AML to retinoids [31], and the combination of ATRA with the anti-CD38 daratumumab to increase cytotoxicity among AML blasts in vitro and overall survival in murine engraftment models of AML [32]. Preliminary studies have suggested that the epigenetic and transcriptional state of leukemia cells determines their susceptibility to ATRA [33]. LSCs are in a different epigenetic state than the total bulk of the AML. Inhibitors of epigenetic-modifying enzymes might sensitize AML cells to ATRA therapy by driving these LSCs to maturity [34]. Hypermethylation of the promoter RARα is common in AML and causes a decrease in its expression [35]. EVI1-positive AML patients have been shown with a distinct methylation profile [36] and downregulation of EVI1 results in epigenetic alterations [37]. Therefore, the association of ATRA and demethylating agents seems logical and could be synergistic in the AML with overexpression of EVI1.

Although promising, results of our retrospective study should be interpreted with caution. There are first several important limitations mainly due to the small number of patients and the heterogeneity of combination therapies. Secondly, some drugs used in combination with ATRA have previously been shown to be effective in controlling AML for variable periods of time, even in high-risk patients, and their potential relative contribution to transitory favorable results cannot be firmly eliminated.

Conclusions

Although our study showed important limitations, it brings major information confirming in humans a prior in vitro study that demonstrated a potential effect of ATRA in AML displaying EVI1 abnormalities [23]. Combination treatments may likely up-regulate ATRA activity resulting in increased overall survival. In this setting, we are currently testing this approach in a phase 3 randomized study comparing ATRA plus azacitidine versus azacitidine alone in AML patients with EVI1 overexpression unfit for intensive chemotherapy. Combination of ATRA with ‘3 + 7’ chemotherapy could also be proposed for patients with EVI1 overexpression considered fit for intensive treatment.