Interrogation of novel CDK2/9 inhibitor fadraciclib (CYC065) as a potential therapeutic approach for AML

Over the last 50 years, there has been a steady improvement in the treatment outcome of acute myeloid leukemia (AML). However, median survival in the elderly is still poor due to intolerance to intensive chemotherapy and higher numbers of patients with adverse cytogenetics. Fadraciclib (CYC065), a novel cyclin-dependent kinase (CDK) 2/9 inhibitor, has preclinical efficacy in AML. In AML cell lines, myeloid cell leukemia 1 (MCL-1) was downregulated following treatment with fadraciclib, resulting in a rapid induction of apoptosis. In addition, RNA polymerase II (RNAPII)-driven transcription was suppressed, rendering a global gene suppression. Rapid induction of apoptosis was observed in primary AML cells after treatment with fadraciclib for 6–8 h. Twenty-four hours continuous treatment further increased efficacy of fadraciclib. Although preliminary results showed that AML cell lines harboring KMT2A rearrangement (KMT2A-r) are more sensitive to fadraciclib, we found that the drug can induce apoptosis and decrease MCL-1 expression in primary AML cells, regardless of KMT2A status. Importantly, the diversity of genetic mutations observed in primary AML patient samples was associated with variable response to fadraciclib, confirming the need for patient stratification to enable a more effective and personalized treatment approach. Synergistic activity was demonstrated when fadraciclib was combined with the BCL-2 inhibitor venetoclax, or the conventional chemotherapy agents, cytarabine, or azacitidine, with the combination of fadraciclib and azacitidine having the most favorable therapeutic window. In summary, these results highlight the potential of fadraciclib as a novel therapeutic approach for AML.


Introduction
Acute myeloid leukemia (AML) is one of the most common hematologic malignancies, characterized by clonal, proliferative, and abnormally or poorly differentiated myeloid cells infiltrating the bone marrow, blood or extramedullary tissues [1][2][3] . Although treatment has progressed, in some AML patients, particularly the elderly 1,4 , outcome is dismal due to the remarkable genetic complexity 5,6 , epigenetic alterations [7][8][9] , and the dynamics of the disease 10,11 . Treatment-related morbidity/mortality and resistance to chemotherapy are major causes of treatment failure 12 . However, longer life expectancy and possibly a cure is achievable if complete remission (CR) is attained 13,14 . Therefore, there is substantial research to improve and personalize therapies, and lessen treatment toxicity [15][16][17][18] .
In this study, fadraciclib was assessed for efficacy against AML cell lines and primary AML samples in vitro either as a single agent or in combination with venetoclax (VEN), cytarabine (AraC), or azacitidine (AZA).
Taken together, fadraciclib potently induces apoptosis, increases the subG0 population, and reduces proliferation as evidenced by reduced metabolic activity in AML cell lines.
Fadraciclib downregulates MCL-1, resulting in rapid induction of apoptosis in AML cell lines AML cell lines were treated with 1 µM fadraciclib, for 4 and 24 h. An inhibitory effect on downstream targets was demonstrated by immunoblotting ( Fig. 2A). Inhibition of CDK2 and CDK9 was evaluated by phosphorylation levels at serine 807/serine 811 of retinoblastoma (Rb) and serine 2 of RNAPII, respectively. Rb was not expressed in OCI-AML3. In MOLM-13, densitometry confirmed a 75% reduction in Rb levels by 4 h (p < 0.0001) and greater than 99% at 24 h, with a corresponding reduction in the phosphorylated form (p-Rb). In MV4-11, only a modest reduction in the levels of Rb and p-Rb at 4 h, with Rb level returning to baseline and a modest increase in p-Rb observed at 24 h ( Fig. 2A, B). At 4 h, in all cell lines, the level of RNAPII decreased by approximately 60% with a corresponding decrease in phosphorylation ( Fig. 2A, C); with similar results at 24 h.
A significant downregulation of MCL-1 occurred at 4 h in all cell lines and was maintained or further decreased at 24 h ( Fig. 2A, D). This induced apoptosis as indicated by the accumulation of cleaved poly (ADP-ribose) polymerase-1 ( Fig. 2A, E). The function of CDK1 was not perturbed as observed by an unchanged level of threonine 320 phosphorylated protein phosphatase 1 alpha (Supplementary Fig. S2A, B).
In particular, at 4 h, the expression of CDKs, especially CDK9, the transcription factor E2F1, and MCL1, were suppressed. PPP1R10 was also significantly downregulated (Fig. 2F, G). Effects on gene expression were less marked at 24 h. Conversely, the DNA damage response regulator gene CDKN1A, encoding p21 Cip1 , was upregulated at 24 h, indicating the intrinsic apoptosis pathway 40 was activated following fadraciclib treatment (Supplementary Fig. S4B).

Fadraciclib induces apoptosis and decreases MCL-1 expression in primary AML cells, regardless of KMT2A-PTD status
Published results indicate AML cell lines bearing KMT2A rearrangements (KMT2A-r) are sensitive to fadraciclib 29 . Five primary AML samples harboring KMT2A-partial tandem duplication (PTD) and five primary AML samples with KMT2A wild-type (WT) were tested with fadraciclib as a single agent (Table. 1). All samples were treated with 0.5 and 1 µM fadraciclib for 24 h, then the drug was washed out and cells were cultured in fresh media for up to 72 h.
A concentration-dependent increase in %active caspase-3 was observed in both KMT2A-PTD and KMT2A WT patient samples following fadraciclib treatment (Fig. 3B). However, due to genetic heterogeneity of AML samples (Table. 1), this difference did not reach statistical significance.

Pulsed dosing with fadraciclib in primary AML cells results in apoptosis and reduces cell viability, but is less effective than continuous dosing
To determine if pulsed dosing achieved an optimal response, primary AML cells were treated for up to 8 h with fadraciclib at a concentration of 1 μM. This resulted in a significant increase in %annexinV at both 6 and 8 h (p = 0.023 and p = 0.0411, respectively) and %active caspase-3 at 6 h (p = 0.0351) as compared with NDC ( Supplementary Fig. S6A, B). As fadraciclib exerts a rapid effect, we investigated whether a 6-h pulse was as effective as 24-h continuous treatment at inducing apoptosis. Fadraciclib showed a higher %active caspase-3 after continuous treatment compared to the pulsed treatment ( Supplementary Fig. S6C). MCL-1 levels were significantly reduced after 24-h continuous treatment compared to both the NDC and the 6-h pulsed treatment (p = 0.0167 and p = 0.0304, respectively; Supplementary Fig. S6D).
Increased apoptosis of primary human AML samples in combination studies of fadraciclib + VEN, fadraciclib + AraC, and fadraciclib + AZA For combination studies, six primary AML samples were selected based on the presence of the most common favorable (e.g., NPM1) and unfavorable mutations (e.g., FLT3-internal tandem duplication (ITD) and KMT2A-PTD), in order to assess the efficacy of fadraciclib and its synergistic activity when combined with VEN, AraC, or AZA, towards these mutations.
Cell cycle analysis revealed %subG0, but no other cell cycle phase, marginally increased in agreement with the increase in %annexinV and %active caspase-3 observed in FAD2 + VEN2, FAD2 + AraC2 and FAD2 + AZA2 combination ( Supplementary Fig. S9A-C) as compared with each single drug treatment. Overall results indicate the more complex the molecular genetic lesions or complexity of karyotype, the less efficacious the combination therapy was (Supplementary Table S4). Fadraciclib + AZA combination therapy has a therapeutic window in in vitro cell proliferation assays Finally, cell proliferation assays using CellTrace Violet staining were performed. Representative histograms for AML and normal CD34 + samples are shown (Supplementary Fig. S10). When considering the combination therapies, none of the low concentration drug combinations resulted in a significant anti-proliferative effect (increased percentage of undivided population, %Undivided %undivided) compared to either NDC or single drug therapies in either AML or normal CD34 + samples ( Fig. 4A-F).
However, differences occurred with the high concentration drug combinations. Specifically, FAD2 + VEN2, had significant anti-proliferative effects when compared to VEN2, but might be explained by the effect of FAD2 (Fig. 4A). This combination had no antiproliferative effect on normal CD34 + cells (Fig. 4D). Although FAD2 + AraC2, has significantly increased anti-proliferative effect as compared to FAD2 in both AML and normal CD34 + samples, this is fully accounted for by the known cytotoxic effects of AraC (Fig. 4B, E). For FAD2 + AZA2, the anti-proliferative effect in AML samples was significantly increased compared to both FAD2 and AZA2 (%Undivided % undivided of 68.7 ± 29.1% for FAD2 + AZA2 compared with 28.5 ± 10.2% in FAD2, p = 0.0004; and 33.5 ± 17% in AZA2, p = 0.0024), as well as NDC (Fig. 4C), indicating synergism of this combination with a combination index (CI) of 0.75. No anti-proliferative effect was demonstrated for this combination on normal CD34 + cells (Fig. 4F).
These findings indicate a therapeutic window and provide evidence for further exploration of combination treatment with fadraciclib and AZA as a promising treatment approach.
Inhibition of CDK2 and CDK9 was evaluated by phosphorylation of Rb 51,52 and RNAPII 53,54 . We found the effects of fadraciclib on Rb were cell line specific ( Fig. 2A, B), whereas, the endogenous level of RNAPII was decreased in all cell lines ( Fig. 2A, C). A transcriptional downregulation of RNAPII is the most plausible explanation for the global gene suppression observed ( Fig. 2F and Supplementary Figs. S3-S5). This is consistent with a previous study stating that a downregulation of POLR2A encoding the major subunit of RNAPII was observed in leukemic blasts following treatment with alvocidib, which potently inhibits CDK9 55 . A significant decrease in short half-life MCL-1 at gene (Fig. 2G) and protein ( Fig. 2A, D) levels highlights a potential target inhibited by fadraciclib.
A rapid decrease in MCL-1 level perturbs a balance in the BCL-2 family, activating the intrinsic apoptosis pathway (Fig. 5B). AML cells are more dependent on MCL-1 than normal hematopoietic cells 56 , highlighting the potential of fadraciclib as a therapeutic approach for AML. Downregulation of E2F1 (Fig. 2G) may explain the G1 cell cycle arrest seen in OCI-AML3 and MV4-11 following fadraciclib treatment ( Supplementary Fig. S1). Interestingly, Erk1/2 were suppressed in all cell lines tested ( Supplementary Fig. S2A, D), which potentially promotes a reduction in MCL-1 stability and half-life 57 . In various cancer cell lines, siRNA-mediated PPP1R10 knockdown results in tumor suppressor PTEN release from nuclear sequestration 58 . Hence, the significant decrease in PPP1R10 gene expression observed (Fig. 2G) potentially promotes an induction of apoptosis. Using leukemic blasts from adult patients with refractory AML in a phase 1 clinical trial (NCT00470197), treatment with the pan-CDKi alvocidib, resulted in a downregulation of various genes 55 . Among these, downregulation of POLR2A, encoding the major subunit of RNAPII, and E2F1 was seen. In addition, a decrease in phosphorylation of RNAPII and expression of MCL-1 was observed in leukemia blasts of some patients treated with alvocidib who achieved a CR 59 . Favorably, we found that the function of CDK1, which plays an important role in mitosis 60,61 , was not perturbed, suggesting better immune function with less impact on normal hematopoietic cells following fadraciclib treatment.
Recently, a phase 1 clinical trial of fadraciclib in solid tumor patients was completed (NCT02552953) 42 . Doselimiting toxicities were reversible and a biologically effective dose of 192 mg/m 2 /day, which corresponds to the drug concentration of 6-7 μM in vitro was established 42 .
Over the past few years, many targeted therapies for patients with AML have emerged with efficacy dependent on molecular subtype of AML 62 . This highlights the important of assessing the efficacy of fadraciclib on specific AML subtypes based on its mechanism of action of inhibiting CDK9. As part of the Super Elongation Complex (SEC), CDK9 is important for transcriptional elongation, in particular in KMT2A-r AML 63 , it is therefore rational to evaluate the efficacy of fadraciclib in this AML subtype. Preliminary results indicate KMT2A-r AML cell lines are more sensitive to fadraciclib 29 . Patient samples, however, display a high diversity of genetic mutations (Table. 1). We therefore observed a more variable fadraciclib response with KMT2A-PTD status not being a predictor for better treatment response. This is consistent with previous results showing that the CDK9-specific inhibitor atuveciclib inhibits the proliferation of seven AML cell lines, regardless of KMT2A-r status 64 . In addition, atuveciclib displayed potent in vitro activity in 80% of AML patient samples harboring KMT2A WT, including those with mutant NPM1 or FLT3-ITD.
Considering the combination studies, primary AML samples were selected based on the most common favorable (e.g., NPM1) and unfavorable mutations (e.g., FLT3-ITD and KMT2A-PTD). Initially, we hypothesized that the treatment response would be relevant to the risk profiles of the individual genetic lesions in primary AML samples. However, we found that complexity of the molecular genetic lesions and karyotypes have more influence. The more complex the molecular landscape, the less efficacious the combination therapy (Supplementary Table S4). The treatment response variability associated with the high diversity of genetic mutations A Fadraciclib downregulates E2F1, resulting in cell cycle arrest at G1 phase. Downregulation of RNAPII is the most plausible explanation for the global gene suppression observed. A rapid decrease in short halflife MCL-1 level perturbs a balance in the BCL-2 family. A decrease in the Erk1/2 activity observed in all cell lines tested potentially promotes a reduction in MCL-1 stability and half-life. p38 MAPK is markedly increased in MOLM-13 cell line only, implying that the stress response and apoptosis are highly induced. B Fadraciclib downregulates MCL-1, which activates the intrinsic apoptosis pathway, followed by cell death. Indeed, a rapid decrease in MCL-1 level perturbs a balance in the BCL-2 family, which leads directly to activator BH3-only proteinsbinding BAK and BAX, resulting in their homo-oligomerisation and MOMP. Following this, pro-apoptotic proteins within the mitochondrial intermembrane space, for example, Cyt c, DIABLO and HTRA2 are released. Cyt c binds to APAF-1 to form the apoptosome. Once formed, the apoptosome can then recruit and activate the inactive pro-caspase-9. Following this, caspases-3, 6, and 7 are activated and multiple proteolytic events occur. observed reiterates the results of previous experiment showing the variable fadraciclib response, regardless of KMT2A-PTD status. Taken together, our results highlight the importance of genetic and molecular profiling prior to initiating therapy, to identify personalized therapeutic combinations most likely to benefit individual patient. Regarding the 2017 ELN risk stratification 50 , next generation sequencing (NGS), exome sequencing and genome wide assays 65 , are being developed to replace single gene assays. This will facilitate better prognostication and more personalized therapy in the future 66 . A short delay in commencing therapy in newly diagnosed AML had no effect on outcome after accounting for other prognostic covariates 67 , providing a potential window to perform NGS.
As a global effect, an increase in cell death in primary human AML cells was shown for apoptosis, active caspase-3 assays, and cell cycle analysis when fadraciclib was combined with VEN, AraC, or AZA. In primary AML cells, combining the MCL-1 inhibitor S63845 with VEN results in greater efficacy than either inhibitor alone, with a more potent activity against leukemic rather than normal hematopoietic progenitors 68 . In an AML patientderived xenograft mouse model bearing FLT3-ITD, the CDK9 inhibitor A-1592668 combined with VEN provided a significant survival advantage over single treatments 69 . In a lymph node mimicking microenvironment, fadraciclib in combination with VEN for 24 h efficiently induced apoptosis of primary CLL cells 70 , a disease where MCL-1 plays a role in disease progression and fludarabine resistance 71 . Ongoing phase 1 clinical trials are evaluating the safety and tolerability of fadraciclib in combination with VEN in patients with relapsed/refractory AML or myelodysplastic syndromes (MDS) (NCT04017546); and relapsed/refractory CLL (NCT03739554).
When considering combining CDK9 inhibitors with AraC, a recent phase 2 study (NCT01349972), demonstrated a significantly higher efficacy of FLAM (alvocidib, cytarabine plus mitoxantrone) as compared with the standard "7 + 3" regimen in terms of a CR rate 26,28 . However, no increase in overall or event-free survival rates were observed. Potentially, this may be due to the higher toxicity of the combination regimen based on our preliminary data showing the impact of the fadraciclib + AraC on normal hematopoietic cells.
In preliminary experiments, pre-treatment with alvocidib reduced the IC50 of AZA in MV4-11 cell line 72 . In the same manner seen in fadraciclib + AZA combination presented here, alvocidib + AZA also increased %active caspase-3 as compared with single treatments. In a MOLM-13 xenograft model, the combination showed greater tumor growth inhibition compared to single agent treatments. Our results, demonstrated a potential therapeutic window with the FAD2 + AZA2 combination, as measured by %Undivided %undivided of AML cells in cell proliferation assay (Fig. 4C). These interesting results identify an attractive therapeutic strategy warranting consideration of clinical development in AML.
In summary, our studies indicate the preclinical efficacy of the CDK2/9 inhibitor fadraciclib as a single agent or in conjunction with frontline AML chemotherapeutics, highlighting its potential as a novel therapeutic approach. Our work implicates the importance of molecular genetic profiling as a critical step prior to initiating therapy for AML, highlighting the need for a more personalized medicine approach to improve outcomes for patients.
Healthy donor and diagnostic primary AML samples were taken in accordance with the Declaration of Helsinki, and Ethics Committee approval (15/WS/0077). Demographic data, cytogenetic abnormalities and genetic lesions of AML patients are shown in Table. 1. Primary samples were thawed and cultured overnight in serumfree medium II (Stem Cell Technologies) supplemented with hIL-3, hIL-6, hSCF, and hFLT3L (all at 10 ng/mL) (Stem Cell Technologies).

Inhibitors
Ten micromolar stock solutions of fadraciclib (Cyclacel), AZA (Stratech), and VEN (Stratech) were prepared in DMSO. 10 mM AraC (Sigma-Aldrich) was prepared in water. All stocks were stored at −20°C and dilutions freshly prepared in cell culture media.

Gene expression analysis
RNA was extracted using RNA Easy Micro kits (Qiagen) and converted to cDNA using high-capacity cDNA reverse transcription kit (Applied Biosystems). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed using Fluidigm Biomark technology and data were collected as per manufacturer's instructions. Data was analyzed using the 2 −ΔΔCT compared to no drug control (NDC) as previously described 73 . Internal sample control was ensured by subtracting the average of six housekeeping genes ATP5F1B, B2M, CYC1, RNF20, TYW1, and UBE2D2 from the Ct value of each gene of interest. Mean and standard deviation of fold change in expression were calculated. For Primer sequences see Supplementary Table S1.

Protein extraction and quantification
In all, 1-5 × 10 6 cells were harvested and washed twice in ice-cold PBS and pelleted at 300 × g for 5 min. Then, the pellets were lyzed in ice-cold solubilization buffer containing phosphatase and protease inhibitors at a concentration of 1 × 10 6 cells per 10 μl solubilization buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Nonidet P40, 10% Glycerol, cOmplet ULTRA 1 tablet and PhosSTOP phosphatase inhibitor 1 tablet per 10 mL of solubilization buffer). This suspension was left on ice and vortexed every 15 min for 1 h, then centrifuged at 14,000 × g for 10 min to pellet all cell debris. The supernatant was transferred to a fresh eppendorf and stored at −80 o C until required.
Protein concentration was quantified using the Quick Start Bradford Dye reagent and compared to a standard curve prepared using Quick StartTM BSA protein. The colorimetric protein assay was measured by visual absorbance on the Spectramax M5 plate reader at 595 nm at using SoftMax Pro software.

Protein analysis
Cell lysates were prepared and 30 µg of protein was resolved. Western blotting was performed using the NuPAGE electrophoresis system (Invitrogen) as per manufacturer's instructions, and protein detection via Odyssey Fc Imaging System. Densitometry was performed using Image Studio Lite version 5.2. Antibodies and experimental conditions are outlined in Supplementary  Table S2.

Flow cytometry
For proliferation analysis, primary AML and normal CD34+ cells were stained with CellTrace Violet (Thermo Fisher Scientific) as per manufacturer's protocol. For establishing a maximum point of fluorescence staining, cells were cultured with Colcemid (100 ng/mL, Sigma-Aldrich) to determine non-dividing cells. All primary cells were treated with fadraciclib, VEN, AraC, AZA or fadraciclib in combination with VEN, AraC or AZA in physiological growth factor conditions. Annexin V/7aminoactinomycin D (7-AAD) or DAPI (4',6-diamidino-2-phenylindole) (BD Biosciences) staining to assess apoptosis by flow cytometry using 1 × 10 5 cells per condition. Propidium iodide (PI) staining buffer to assess cell cycle progression as per manufacturer's protocol. Fixation/Permeabilization Solution Kit (BD Biosciences) was used prior to MCL-1 or active caspase-3 staining. For staining reagents used, see Supplementary Table S3.

Combination studies
CompuSyn software (ComboSyn, Inc) was used to investigate the synergism in the cell proliferation assays. The software was based on the Chou-Talalay method for drug combination based on the median-effect equation derived from the mass-action law principle 74 . The CI provides a quantitative definition for additive effect (CI = 1), synergism (CI < 1) or antagonism (CI > 1) in drug combinations.