Background
Chromosomal rearrangements of the lysine methyltransferase 2A (KMT2A), known also as mixed-lineage leukemia (MLL) gene, are associated with high-risk infant, pediatric, adult, and therapy-induced acute leukemia. In infant and early childhood, acute leukemia is the most prevalent cancer and very often can be addressed with available therapeutics. A significant exception are patients genetically defined by MLL-fusions, where for most fusions, a worse prognosis [
1] is underscoring the need for improved treatment options.
MLL associated genomic changes are balanced chromosomal translocations which result in an in-frame fusion of the MLL1 protein with a nuclear protein often involved in transcriptional elongation. So far, more than 130 different chromosomal rearrangements have been identified, but four of the most frequent fusion partners (AF4, AF9, ENL, and AF10) account for more than 70% of all observed rearrangements in patients [
2]. While the diversity of observed fusions in patients suggests many disparate genetic subtypes, a common mode of action has been proposed for the oncogenic function of most frequently observed direct fusion (MLL-X) proteins [
3]. These proteins essentially combine the target gene binding properties of the MLL1 protein with the capacity to trigger efficient transcriptional elongation by RNA polymerase II (RNAPII) recruitment. With the aforementioned properties, the MLL-fusion acts as the dominant transcriptional regulator which disrupts differentiation and promotes leukemogenesis [
4,
5]. Wild-type MLL1 is responsible for the tissue-specific epigenetic regulation of homeotic gene expression in differentiation and development [
6]. Catalytic SET domain is lost in the direct (MLL-X) fusion proteins, while the N-terminal DNA-binding domains and the capability to interact with recruiting co-factors, such as MENIN, are retained. The C-terminal part of different MLL1 fusion proteins is capable of recruiting a large multiprotein machinery (“super elongation complex” (SEC)) involved in activation of RNAPII for transcriptional elongation [
7]. The mechanistic consequence of the SEC complex recruitment is an increased expression of MLL1 target genes leading to impaired differentiation. It has been shown that MLL-fusions exhibit their transforming capacity largely through upregulation of
HOX genes [
8,
9], especially HOXA9 and MEIS1 [
10‐
12]. Normally,
HOXA9 and
MEIS1 are expressed at higher levels in stem cells and early lineage progenitors, and expression levels are downregulated with the process of differentiation [
13]. Aberrant expression of
HOX genes by the fusion induces a differentiation blockade resulting in leukemic cells with stem cell-like characteristics and increased self-renewal properties, growth, and survival advantages [
14‐
16]. Since this differentiation blockade is an essential pathomechanism of MLL-fusion proteins, different therapeutic targets, whose inhibition might lead to terminal differentiation and reversal of the leukemia-initiating cells, have been suggested [
1]. Notably, inhibitors that target core transcriptional proteins are of high interest, since they potentially interfere with the aberrant transcriptional elongation machinery and the leukemic gene expression program. Therefore, inhibitors against the kinase P-TEFb (CDK9/CyclinT1) [
17], the histone methyltransferases DOT1L [
18], and the bromodomain and extra-terminal domain (BET) family of proteins [
19] are currently in clinical testing for AML. Another rather new strategy is the inhibition of the recruitment of the MLL-fusion and associated complex to the target genes. For this propose, inhibitors of the MENIN-MLL interaction have been described and are currently in pre-clinical evaluation [
20‐
22]. Based on a phenotypic screening approach aimed towards HoxA9 regulation, inhibitors of the dihydroorotate dehydrogenase (DHODH) have emerged as an additional new strategy to overcome the differentiation blockade [
23]. Despite initial positive pre-clinical evaluation of inhibitors against those targets in fused models of AML/ALL, first data on clinical activity of P-TEFb, BET, and DOT1L first-generation inhibitors are still awaiting true clinical proof of concept [
19].
Here, we analyzed how inhibitors of some emerging therapeutic targets impact the differentiation blockade induced by the MLL-fusion in a comprehensive benchmark study. A better understanding of the differentiation effects could facilitate the further development and clinical translation of these novel agents. Therefore, in our study, we analyzed OTX015 (BET inhibitor) [
24], Brequinar (DHODH inhibitor) [
25], EPZ-5676 (DOT1L inhibitor) [
26], and BAY 1251152 (novel first-in-class selective CDK9/P-TEFb inhibitor) [
27], all representing clinical-stage small molecules (Table
1). Since MENIN-MLL inhibitors are not yet in clinical development, we additionally tested BAY-155, a novel potent and selective inhibitor derived from an in house program (further information see Additional file
1: Table S1) [
28]. All different inhibitors were benchmarked for their capabilities to overcome the differentiation blockade, potential overlaps in transcriptional activities, selectivity for the MLL-fusion, and their combination potential.
Table 1
Inhibitors used in this study. Chemical structures of inhibitor used in this study tackling Menin-MLL1 interaction, BRD4/2/3, DOT1L, CDK9, and DHODH active sites, with respective biochemical IC50, rationale and current developmental status
Materials and methods
Cell lines
HL-60 cells were obtained from NCI 60-Panel. Jurkat and MV4-11 cells were obtained from ATCC. OCI-AML5, RS4;11, SEM, ML-2, MOLM-13, MOLM14, NOMO-1, OCI-AML2, KOPN-8, EOL-1, and OCI-AML3 cells were obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ, Braunschweig, Germany). All used cells were cultured in the appropriate media and conditions.
Inhibitors
All inhibitors used in this study were synthesized in-house (Bayer AG). BAY-155 was synthesized according to the methods outlined in patent application WO2017207387A1. Inhibitor concentrations for EPZ-5676, Brequinar, and OTX015 used in this in vitro study are lower as plasma concentrations measured in clinical studies [
24,
26,
29]. Plasma concentrations of BAY 1251152 in humans are not yet reported.
Cell proliferation
Cells were seeded in the optimal growth medium at 4000–5000 cells/well in a 96 MTP and cultured 18–24 h before inhibitor treatment. Upon treatment with the indicated inhibitor, cells were cultured for 24 h, 96 h, and 168 h and effects on proliferation were determined using alamarBlue Cell Viability Reagent (Thermo Fisher Scientific, Waltham, MA, USA).
Flow cytometry
Four thousand cells per well were seeded 24 h before they were treated with the indicated inhibitor in a 96 MTP. After 4 or 7 days of treatment, cells were washed with PBS and stained with CD11b - APC (BioLegend, San Diego, California, USA) and DAPI (Thermo Fisher Scientific, Waltham, Massachusetts, USA) or AnnexinV – FITC (BioLegend, San Diego, California, USA) and PI solution (Sigma-Aldrich St. Louis, Missouri, USA) using the FACS Canto II (BD Biosciences, Heidelberg, Germany) and data was analyzed with FACSDiva software.
Cell cycle analysis
Cells were washed with PBS and fixed overnight at − 20 °C with 70% ethanol. Fixed cells were stained with PI solution (Sigma-Aldrich St. Louis, MO, USA) solution containing RNase A (Qiagen, Hilden, Germany). Fluorescence was measured with FACS Canto II (BD Biosciences, Heidelberg, Germany) flow cytometer and data was analyzed with FACSDiva software.
Wright-Giemsa staining
Approximately 10,000 of cytospin prepared cells were air dried, fixed in 100% methanol for 1 min, stained in 100% in Wright-Giemsa staining solution (Sigma-Aldrich St. Louis, Missouri, USA) for 90 s, washed two times in deionized water, and air dried.
Phagocytosis assay
After 7 days of treatment with the indicated inhibitor, cells were washed once with PBS and quantified. Ten thousand viable cells were resuspended in fresh media along with fluorescein-labeled heat-killed Escherichia coli BioParticles (Molecular Probes, Eugene, OR, USA) (100,000 units), incubated at 37 °C for 30 min and stained with CD11b - APC (BioLegend, San Diego, CA, USA ) and DAPI. Phagocytosis capability was measured with FACS Canto II (BD Biosciences, Heidelberg, Germany). Immunofluorescence of cytospin preparations was measured on LSM700 microscope (ZEISS, Oberkochen, Germany) using CD11b (APC), DAPI, and E.coli particles (FITC).
Gene expression
Total RNA was isolated using RNeasy-Plus Mini kit (Qiagen, Hilden, Germany). RNA (1 μg) was reverse transcribed using SuperScript III First-Strand Synthesis SuperMix (Life Technologies, Carlsbad, CA, USA) and obtained cDNA was used for qRT-PCR at the TaqMan 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) utilizing TaqMan Fast Advanced Master Mix (Life Technologies). Commercial primers used in this study are listed in Additional file
2: Materials and methods. RNA-seq study: cells were treated for 8 h (P-TEFb—0.05 μM, OTX015—1 μM), 24 h (BAY-155—2 μM, Brequinar—2 μM, DMSO—0.1%) and 96 h (EPZ-5676—3 μM, DMSO—0.1%) prior to RNA extraction using RNeasy-Plus Mini kit (Qiagen). Obtained RNA was used for library preparation (Illumina, San Diego, CA, USA. TruSeq Stranded mRNA Kit) and obtained libraries were sequenced (Illumina, HiSeq2500 HTv4, SR, dual-indexing, 50 cycles).
Data analysis and statistical methods
RNA-seq reads were aligned to hg38 using STAR aligner. Gene expression was quantified using RSEM. Samples with less than 10 million reads aligning to the genome were excluded; protein-coding genes with more than 10 reads in more than three samples were used for the analysis (total samples
N = 305; genes
N = 15,007). DESeq2 was used to find genes differentially expressed upon treatment by inhibitors in either each cell line or in the group of sensitive cell lines, while controlling for differences between the cell lines. GSEA analysis was run on the pre-ranked list based on logFC in expression for each compound. To remove cell line-specific differences in PCA, average expression in the DMSO sample was subtracted for each corresponding cell line. Top 1000 variable genes were selected based on median absolute deviation. Data is available at GEO (
https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE125437.
Immunoblotting
Western blot analysis was performed on cell lysates from at least 100,000 cells. Forty micrograms of whole cell protein extract was separated on 4–20% Tris-Glycine gels, transferred to 0.2-μm nitrocellulose membranes, and probed with anti-HEXIM1 (Bethyl, Montgomery, TX, USA) and β-ACTIN (Cell Signaling, Beverly, MA, USA) antibodies.
Discussion
The concept of differentiation therapy emerged in the late 1970s when retinoic acid (RA) cAMP, sodium butyrate, arsenic trioxide, and cytokines were proposed to treat acute promyelocytic leukemia (APL). Since then, several clinical studies have shown treatment benefits by using all-trans RA in combination with arsenic trioxide resulting in > 90% complete remission [
32]. Nevertheless, effects are restricted to a specific chromosomal translocation
t [
15,
17] driving APL comprising 10% of all AML patients [
16]. Therefore, new strategies tackling the differentiation blockade and self-renewal capacity of AML/ALL cells with different genetic alterations were proposed and are currently under clinical evaluation [
33,
34].
In our comprehensive study in MLL-fused AML/ALL models, we have used inhibitors against CDK9 (BAY 1251152), DOT1L (EPZ-5676), BRD2/3/4 (OTX015), MENIN-MLL interaction (BAY-155), and DHODH (Brequinar). All these proteins have been associated with differentiation in AML/ALL [
23,
31,
35‐
40], but since inhibitors for those protein targets have all been used so far under isolated experimental conditions, a direct comparison of their differentiation capacity was not possible. Therefore, we profiled those inhibitors head-to-head for gene expression effects in a large cell line panel. We further examined cellular responses such as inhibition of proliferation, apoptosis induction, cell cycle arrest, and phagocytosis as functional differentiation readout. Based on our results, we found clear differences in the differentiation capacity and specificity for MLL-fused AML/ALL cell lines of examined inhibitors (Fig.
5c).
We observed that BAY-155 and EPZ-5676 treatment led to anti-proliferative effects, transcriptional changes, and differentiation exclusively in the MLL-fused AML models. This data confirms a driver function of Menin and DOT1L especially in the MLL-fusion-induced de-differentiation and increased self-renewal activity via aberrant transcriptional activation of master regulators (e.g., HOXA9, MEIS1, and MYB). Inhibiting expression of those stemness-associated master regulators by inhibition of Menin or DOT1L triggers expression of differentiation-associated genes. This could explain our observation of a higher number of upregulated genes after inhibitor treatment in contrast to the described activating function of those proteins. Menin is required for the recruitment of the MLL-fused protein, which co-recruits the elongation complex (AF4, P-TEFb, ENL, DOT1L, and BRD4) causing extension of H3K4me3 and H3K79me3 marks on transcribed gene bodies. DOT1L is the essential H3K79 methyltransferase, which creates extended H3K79 methylation and overwrites normal epigenetic regulation pattern [
41]. As consequence, productive elongation of MLL-fusion target genes by RNAPII is promoted resulting in transcriptional reprograming and loss of cellular identity [
42]. In a clinical phase I study, EPZ-5676 was evaluated in AML patients and a significant reduction of H3K79me2 on
HOXA9 and
MEIS1 was observed [
26]. This observation also correlates with our gene expression analysis and previous reports. Interestingly, while comparing the effects of BAY-155 and EPZ-5676, it appears that blocking the recruitment of MLL-fusion complex is a more efficient way to induce transcriptional changes, differentiation, and cell killing than inhibiting DOT1L. Tackling the Menin-MLL interaction in MLL-fused AML/ALL induces overall very similar transcriptional changes as with inhibition of the DOT1L methyltransferase activity. Nevertheless, Menin-MLL inhibition resulted in significantly faster anti-proliferation and differentiation effects. Faster effects after the inhibition of the Menin-MLL interaction can be partially explained by the kinetics of the MLL-fusion as an oncogenic driver. The Menin-MLL interaction is mechanistically further upstream than the methylation activity of DOT1L [
43]. Therefore, Menin-MLL inhibition leads to an overall reduced recruitment of ENL and other elongation factors (like DOT1L), which then leads to the observed suppression of HOXA10, MEIS1, and MYB, and upregulation of CD11b [
44]. For DOT1L, it has been reported that both genetic and pharmacological targeting results in delayed (4–10 day) effects on transcriptional regulation and cell viability in AML [
41,
45], which can be explained by the slow turnover rate of pre-existing H3K79 methylation [
46]. Interestingly, we could detect proliferation and differentiation synergisms of BAY-155 and EPZ-5676 combination. This might be explained by the possibility that inhibition of Menin-MLL or DOT1L alone does not fully inhibit all MLL-fusion activities. Possibly, Menin independent recruitment or other SEC member (e.g., ENL) activities might promote transcriptional elongation independently from H3K79me [
17]. Pharmacological inhibition of the Menin-MLL interaction appears to be selective to the MLL-fused AML/ALL with differentiation induction and anti-proliferation potential; however, this treatment option still awaits clinical evaluation.
Another approach in AML therapy conceived in the past years is blocking of multiple transformation pathways which are dependent on the P-TEFb function via BET and CDK9 inhibition. Both targets were shown to be critical for AML/ALL cell viability mainly through regulating MYC, MYB, and MCL1 levels [
17,
37,
47]. While genetic and pharmacological BRD4 inhibition was linked to cell differentiation [
47], a direct inhibition of CDK9 activity results in differential responses. Our study results confirm strong cell killing activity of both inhibitors and transcriptional inhibition of CDK9/BET regulated target genes [
17,
48]. In our study, only BET but not CDK9 inhibition resulted in cell differentiation on transcriptional and morphological level. However, early transcriptional profiling of OTX015 did not show any significant positive effects on AML/ALL differentiation associated pathways. When applied for several days at higher concentrations OTX015 induces differentiation effects independent from the MLL-fusion, which hints to differentiation as secondary to primary gene expression effects. One explanation for the delayed effect of OTX015 on differentiation might be the direct downregulation of transcription factors MYB and MYC. It has been reported that their ectopic expression is inhibiting differentiation in a number of cell lines and primary cells [
49,
50]. Additionally, OTX015 modulates the largest number of genes, even at the very early time point tested, of all inhibitors, which indicates a substantial effect on the global gene expression network. Those expression changes resulted in differentiation effects only in a limited number of cells but overall resulted in very robust anti-proliferation effects. Strong global effects on transcription might also be the reason for the inability of CDK9 inhibition to induce differentiation. Inhibition of proliferation and apoptosis induction is the dominant effect of CDK9 inhibitor, and cells are killed before a potential interference with the MLL-fusion leads to differentiation. Currently, BAY 1251152 undergoes phase I clinical evaluation with no final report yet. Initial pharmacodynamics data analysis shows dose-dependent reduction of
MYC,
PCNA, and
MCL-1 levels, all being relevant for cancer cell survival [
51]. Interestingly, OTX015 clinical trial performed in AML patients harboring a number of diverse driving mutations resulted in partial blast clearance and recovery of platelets. However, severe thrombocytopenia as dose-limiting effect was observed in patients with incomplete
bone marrow failure [
24]. Altogether, our cellular analysis for OTX015 and BAY 1251152 support the clinical observations and suggest that interfering with P-TEFb function via BET and CDK9 inhibition leads primary to strong anti-proliferation and apoptosis induction effects which are MLL-fusion independent.
Lastly, DHODH, an enzyme in the de novo synthesis of nucleotides, was shown to be critical for the self-renewal and proliferation capacity in a wide variety of AML models [
23,
52]. Our data is significantly extending those findings by connecting the described differentiation phenotypes of Brequinar with global gene expression profiling and functional AML differentiation. Interestingly, tackling de novo pyrimidine biosynthesis leads to a pronounced effect on global gene expression but also to a very specific response in AML/ALL relevant pathways which is not restricted to MLL-fused models. Moreover, DHODH inhibition by Brequinar undergoes a phase I clinical reevaluation in AML patients after encouraging pre-clinical observations suggesting its role in differentiation [
23,
52]. Furthermore, we have observed that Brequinar effect on gene expression is similar to the effects of BAY-155 and EPZ-5676 in MLL-fused models inducing more terminal differentiation. Brequinar in combination with BAY-155 or EPZ-5676 leads also to significant anti-proliferation and differentiation synergism, whereas combining Brequinar with OTX015 and BAY 1251152 induces exclusively anti-proliferation synergy. While nucleotide shortage induces stress and therefore explains proliferation inhibition and cell cycle arrest, it is also reported to drive HEXIM1 expression [
30]. Our data provides for the first time a direct link between HEXIM1- and Brequinar-induced nucleotide stress leading to AML/ALL differentiation. In summary, our novel findings extend the understanding of Brequinar-mediated AML/ALL differentiation and explore some of possible combinations. Altogether, based on our results, inhibiting Menin-MLL together with DOT1L might allow for a more efficient and MLL-fusion-specific induction of differentiation and apoptosis. In contrast, BAY 1251152, OTX015, and Brequinar are significantly affecting also differentiation independent pathways (e.g., RNA metabolism/translation). This might limit their combination potential since expected treatment tolerability could be lowered.
In conclusion, these new findings enhance our understanding on the activity of used inhibitors of those emerging therapeutic targets in MLL-fusion-driven leukemia. Our novel findings give some valuable insights into their differentiation induction potential, which is a possible underestimated contribution of their therapeutic activities in AML/ALL.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.