AZD4573 Is a Highly Selective CDK9 Inhibitor That Suppresses MCL-1 and Induces Apoptosis in Hematologic Cancer Cells

Published OnlineFirst November 7, 2019; DOI: 10.1158/1078-0432.CCR-19-1853

CLINICAL CANCER RESEARCH | TRANSLATIONAL CANCER MECHANISMS AND THERAPY

AZD4573 Is a Highly Selective CDK9 Inhibitor That Suppresses MCL-1 and Induces Apoptosis in Hematologic Cancer Cells A C Justin Cidado1, Scott Boiko1, Theresa Proia1, Douglas Ferguson2, Steven W. Criscione1, Maryann San Martin1, Petar Pop-Damkov2, Nancy Su3, Valar Nila Roamio Franklin4, Chandra Sekhar Reddy Chilamakuri4, Clive S. D'Santos4, Wenlin Shao5, Jamal C. Saeh5, Raphael Koch6, David M. Weinstock7, Michael Zinda1, Stephen E. Fawell1, and Lisa Drew1

ABSTRACT ◥ Purpose: Cyclin-dependent kinase 9 (CDK9) is a transcriptional cell death broadly across hematologic cancer models in vitro, and regulator and potential therapeutic target for many cancers. Mul- MCL-1 depletion in a dose- and time-dependent manner was tiple nonselective CDK9 inhibitors have progressed clinically but identified as a major mechanism through which AZD4573 induces were limited by a narrow therapeutic window. This work describes a cell death in tumor cells. This pharmacodynamic (PD) response was novel, potent, and highly selective CDK9 inhibitor, AZD4573. also observed in vivo, which led to regressions in both subcutaneous Experimental Design: The antitumor activity of AZD4573 was tumor xenografts and disseminated models at tolerated doses both determined across broad cancer cell line panels in vitro as well as as monotherapy or in combination with venetoclax. This under- cell line- and patient-derived xenograft models in vivo. Multiple standing of the mechanism, exposure, and antitumor activity of approaches, including integrated transcriptomic and proteomic AZD4573 facilitated development of a robust pharmacokinetic/PD/ analyses, loss-of-function pathway interrogation, and pharmaco- efficacy model used to inform the clinical trial design. logic comparisons, were employed to further understand the major Conclusions: Selective targeting of CDK9 enables the indirect mechanism driving AZD4573 activity and to establish an exposure/ inhibition of MCL-1, providing a therapeutic option for MCL-1– effect relationship. dependent diseases. Accordingly, AZD4573 is currently being Results: AZD4573 is a highly selective and potent CDK9 inhib- evaluated in a phase I clinical trial for patients with hematologic itor. It demonstrated rapid induction of apoptosis and subsequent malignancies (clinicaltrials.gov identifier: NCT03263637).

Introduction CDK9 is crucial for the proper regulation and progression of transcription. Through phosphorylation of serine 2 (pSer2) in the Cyclin-dependent kinases (CDK) are a family of closely related heptapeptide repeats within the C-terminal domain of RNA polymer- serine/threonine kinases known for playing crucial roles in regulating ase II (RNAP2), CDK9 releases RNAP2 from its paused state to enable either cell-cycle progression (CDK1, 2, 4, 6) or gene transcription transcription elongation (10). Acute inhibition of CDK9, therefore, (CDK7, 8, 9, 12, 13; ref. 1). As integral nodes in these regulatory results in transient transcriptional suppression and preferential deple- networks, CDKs have been studied extensively as possible targets for tion of short-lived transcripts and proteins (11), providing a thera- cancer therapy, resulting in the development of nonselective CDK peutic approach to indirectly target key driver oncoproteins with short inhibitors (2) that show signs of clinical activity (3–6). Despite half-lives. Nonselective CDK9 inhibitors, such as dinaciclib and pharmacologically inhibiting multiple CDKs, the observed clinical flavopiridol (alvocidib), have been repurposed to explore this oppor- activity of nonselective CDK inhibitors has been primarily attributed tunity but have not progressed clinically as monotherapies due to to their CDK9 activity (7–9). narrow therapeutic indices (6, 12). However, clinical trials are ongoing to evaluate these molecules in combination with other therapeutics. Despite clinical development of multiple inhibitors with CDK9 1Bioscience, Oncology R&D, AstraZeneca, Boston, Massachusetts. 2DMPK, activity, the precise mode of action driving antitumor effects has yet to 3 Oncology R&D, AstraZeneca, Boston, Massachusetts. Discovery Sciences, be fully elucidated, although their preclinical activity has been attrib- BioPharmaceuticals R&D, AstraZeneca, Boston, Massachusetts. 4CRUK uted predominantly to depletion of MCL-1 protein and subsequent Cambridge Institute, Cambridge University, Cambridge, United Kingdom. – 5Projects, Oncology R&D, AstraZeneca, Boston, Massachusetts. 6Department induction of tumor cell death (13 15). MCL-1 is an antiapoptotic, of Hematology and Medical Oncology, University Medical Center Goettingen, BCL-2 family protein whose high expression has been associated with Goettingen, Germany. 7Department of Medical Oncology, Dana-Farber Cancer increased cancer cell survival that translates to chemotherapy resis- Institute, Boston, Massachusetts. tance and poor patient prognosis (16–18). Note: Supplementary data for this article are available at Clinical Cancer This work describes the mechanism of action and preclinical activity Research Online (http://clincancerres.aacrjournals.org/). of the novel and highly selective CDK9 inhibitor, AZD4573. An Current address for M. Zinda: Repare Therapeutics, Cambridge, MA. agnostic approach utilizing integrated transcriptomic and proteomic analyses revealed MCL1 as one of the top oncogenes most significantly Corresponding Author: Lisa Drew, AstraZeneca, 35 Gatehouse Drive, Waltham, MA 02451. Phone: 1-781-839-4836; E-mail: [email protected] and robustly downregulated at the mRNA and protein level upon acute AZD4573 treatment. Subsequent pharmacologic studies further sup- Clin Cancer Res 2020;XX:XX–XX ported an MCL-1–mediated mechanism, including derivation of a doi: 10.1158/1078-0432.CCR-19-1853 pharmacokinetic (PK)/pharmacodynamic (PD)/efficacy model link- 2019 American Association for Cancer Research. ing CDK9 inhibition, MCL-1 depletion, and induction of apoptosis in

AACRJournals.org | OF1

Cidado et al.

dilutions. Automated dosing was performed using an Echo 555 Translational Relevance acoustic liquid dispenser (Labcyte). To calculate percent caspase Multiple cyclin-dependent kinase 9 (CDK9) inhibitors, both activation, luminescence values were normalized against maximum nonselective and selective, have been evaluated in clinical trials. (100%, mixture of MCL-1, BCL-2, and BCL-XL inhibitors) and Despite signs of clinical activity and ongoing combination studies, minimum (0%, DMSO) controls. Percent growth inhibition was monotherapy development was ultimately not pursued due to normalized to untreated cells at the time of dosing, whereas percent narrow safety margins. Contributing factors could include a lack viability was normalized against maximum (100%, DMSO) and of compound selectivity for CDK9, the absence of a clear patient minimum (0%, medium only) controls. Dose–response curves were selection strategy, and suboptimal choice of dose and/or schedule. generated and half-maximal effective (EC50) and growth inhibitory The development and full preclinical mechanistic, pharmacoki- (GI50) concentrations calculated using GeneData Screener and Graph- netic, and pharmacodynamic understanding of a highly selective Pad Prism. and potent CDK9 inhibitor, like AZD4573, could mitigate those limitations. Accordingly, pharmacologic evaluation of AZD4573 Washout assay demonstrated that it represses MCL-1 through acute transcription Cells were seeded the day before treatment with the indicated inhibition and results in rapid induction of apoptosis broadly AZD4573 dose. At the indicated times, cells were washed twice in across preclinical hematologic cancer models. This work estab- complete media and plated at 1 103 cells/well in 96-well plates. Cells lished a quantitative relationship between extent and duration of were incubated until 12 or 24 hours after the time of initial dosing, at CDK9 inhibition, depletion of MCL-1, and rate of apoptosis which point Caspase-Glo 3/7 and CellTiter-Glo readouts were per- fi induction that was used to estimate the ef cacious dose range and formed, respectively. hopefully maximize the therapeutic window in clinical studies. RNA sequencing library preparation and analysis cDNA library construction for RNA sequencing (RNA-seq) was performed at Novogene using an NEBNext Ultra mRNA Library Prep response to AZD4573. Application of this mechanistic understanding Kit for Illumina (New England Biolabs) according to the manufac- using short-term treatment and intermittent dosing with AZD4573 turer's protocol. RNA-seq gene expression quantification was con- revealed broad antitumor activity across preclinical hematologic can- ducted using Sailfish against hg38 Ensembl transcripts (version 79), cer models. Collectively, this work supports the clinical evaluation of and differential expression analysis was done using Voom and Limma AZD4573 in patients with hematologic malignancies (NCT03263637). R package. A detailed description of experimental procedures can be found in the Supplementary Methods. The data have been deposited in Materials and Methods NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE129012. Chemicals AZD4573 and AZD5991 were synthesized at AstraZeneca. Vene- Mass-spectrometry proteomics toclax (ABT-199) was purchased from MedKoo. For in vitro assays, Trypsin digestion, Tandem Mass Tag labeling, and LC-MS/MS compounds were solubilized to 10 mmol/L in DMSO. Compounds analysis were conducted as previously described (20). A detailed were dosed to yield a final DMSO concentration of 0.3%. For in vivo description of experimental procedures can be found in the Supple- studies, AZD4573 was prepared in a 2%/30%/68% mix of N,N- mentary Methods. The data have been deposited to the ProteomeX- dimethylacetamide, PEG-400, and 1% (v/v) Tween-80, AZD5991 was change Consortium via the PRIDE partner repository with the dataset formulated in 30% 2-hydroxylpropyl-beta-cyclodextrin (HPBCD) at identifier PXD013796. pH 9, and venetoclax was solubilized in a 10%/30%/60% mix of ethanol, PEG-400, and Phosal 50 PG. Real-time PCR 6 Enzyme assays MV411 cells (1 10 )treatedwithvehicleor0.1mmol/L AZD4573 was screened at 0.1 mmol/L against 468 kinases and AZD4573 were collected at the indicated timepoints. MV-4-11 relevant variants using the KINOMEscan assay (DiscoveRx), and tumors were also harvested, placed in a Lysing Matrix D tube kinome selectivity images were generated using TREEspot Software (Millipore) with Buffer RLT (Qiagen) supplemented with 1% b Tool. SignalChem compound selectivity profiling was utilized to -mercaptoethanol, and mechanically disrupted using the Millipore screen AZD4573 against select CDKs and non-CDK kinases at FastPrep. RNA isolation and cDNA conversion were conducted using RNAeasy Mini (Qiagen) and High Capacity cDNA RT 5 mmol/L (high) ATP concentration for IC50 determination. (Applied Biosystems) kits, respectively, following the manufac- Immunoprecipitation and immunoblotting turers’ protocols. Relative MCL1 mRNA levels were quantified on Cell and tumor lysates and immunoprecipitation (IP) samples were an ABI Prism 7900HT Real-time PCR system (ThermoFisher) using MCL1 prepared and subjected to immunoblotting as described previously Taqman Assays (Applied Biosystems, : Hs01050896_m1, 18s MCL1 using antibodies listed in Supplementary Table S1 (19). rRNA: 4319413E). Relative expression at each timepoint was DD calculated using the - CT method by normalizing cycle threshold In vitro caspase and viability assays values to the 18s rRNA reference. Assays were conducted as previously described (19) and according to the manufacturer's protocols for Caspase-Glo 3/7 and CellTiter-Glo In vivo studies (Promega). For cell panel screens, cells were plated in 384-well plates Subcutaneous xenografts 1 and dosed the following day with 10-point, /2 log dilutions of Studies were conducted at AstraZeneca in accordance with animal 1 compound, whereas combination treatments used 5-point, /2 log research guidelines from the NIH and the AstraZeneca Institutional

OF2 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

AZD4573, a Selective CDK9 Inhibitor, Targets MCL-1

Animal Care and Use Committee and as previously described (19). other CDKs upon short-term treatment of MCF7 cells (Fig. 1D; Injected cell numbers for cell line xenografts are as follows: 5 106 for Supplementary Fig. S1A). MCF7 is a favorable model for assessing MV-4-11, MOLP-8, and OCI-LY10, and 1 107 for all others. AZD4573 selectivity and mechanism of action because a frameshift AZD4573 was dosed i.p.) at either 5 or 15 mg/kg, twice daily with mutation in the CASP3 gene results in the absence of functional a 2-hour split dose (BID q2h), on a 2-day on/5-day off or biweekly caspase-3 protein, preventing potentially confounding apoptosis schedule. Venetoclax was dosed daily at 100 mg/kg by oral gavage. induction and cell death. AZD5991 was dosed i.v. at 60 mg/kg. For combination studies, Kinetic analysis of CASPASE-3 cleavage (“caspase activation”), a venetoclax was administered first, and AZD4573 was administered hallmark of committed apoptosis, and loss of viability in cell lines within 30 min. cultured with AZD4573 over 24 hours showed that potent and selective CDK9 inhibition can kill cancer cells. Though results AML disseminated patient-derived xenografts varied slightly, they highlighted a rapid response to AZD4573 in Study was conducted at Champions Oncology using nine mod- sensitive models and a direct correlation of caspase activation with els (21). Following 4 weeks of AZD4573 therapy, mice (n ¼ 3 per loss of cell viability (Fig. 1E; Supplementary Fig. S1B and S1C). model) were euthanized for end of study analysis on the spleen, bone Conversely, the insensitive cell lines exhibited no caspase activation marrow, and peripheral blood by FACS analysis to detect human norlossofviabilityuntilprolongedtranscription inhibition caused þ þ CD45 /CD33 /CD3 acute myeloid leukemia (AML) blasts. nonspecificcytotoxicity(SupplementaryFig.S1D).Thisobserva- tion underscored the need to determine the requisite duration of T-cell lymphoma patient-derived xenograft target engagement to achieve maximal effect in tumor cells while Study was conducted under Dana-Farber Cancer Institute Animal mitigating unwanted toxicity in normal tissue. Therefore, the AML Care and Use Committee protocol #13-034 and Public Health Service cell line MV-4-11 was pulsed with AZD4573 for various lengths of animal assurance #A3023-01. A full characterization of the model, time,washedout,andthenassessedfor caspase activation and loss DFTL-78024, is available online at www.PRoXe.org. Animal work was of cell viability at 12 and 24 hours, respectively. This study showed performed in NSG mice purchased from Jackson Laboratories as that approximately 6 to 8 hours of CDK9 inhibition was sufficient to published previously (22). When sufficiently engrafted, mice were achieve maximal cell death equivalent to continuous exposure of treated for four cycles of AZD4573 and monitored daily for clinical AZD4573 (Fig. 1F). signs of disease and humanely euthanized when they reached a clinical þ endpoint. Annexin V was assessed in human CD45 tumors isolated CDK9 inhibition by AZD4573 results in rapid turnover of from peripheral blood 6 hours after the second day 2 dose as previously MCL-1 described (23). In previous studies, MCL-1 overexpression rescued CDK9 inhibitor–mediated apoptosis, seemingly validating MCL-1 deple- PK/PD/efficacy model development tion as the mechanism driving the antitumor effects (25, 26). Model parameters are listed in Supplementary Table S2, and full However, overexpression studies relying on supraphysiological methodology and derivation are available in the Supplementary protein levels come with known caveats, especially in systems Methods. requiring a delicate balance between multiple proteins (i.e., apoptosis regulation by BCL-2 family proteins; ref. 27). Therefore, an Data sharing statement agnostic approach utilizing integrated transcriptomic and prote- For original data, please contact [email protected]. omic time-course experiments was employed to elucidate this mechanism. For reasons listed above, the AZD4573-insensitive cell line, MCF7, was used for these experiments to ensure cell death Results did not introduce artifact into the assay. Cells were treated with Identification of a highly selective and potent CDK9 inhibitor, vehicle or AZD4573 for short periods of time (0, 4, and 8 hours) AZD4573 based upon target engagement requirements (Fig. 1F). Biological Understanding the binding of a high-quality probe compound to triplicates were subjected to RNA-seq or LC-MS/MS, which dem- CDK9 helped guide optimization of a selective and potent CDK9 onstrated high reproducibility based upon principal component inhibitor with suitable physicochemical properties that would result analysis (Supplementary Fig. S2A). in a short compound half-life and enable intravenous administra- A cancer-associated gene list emphasizing driver oncogenes, includ- tion in humans. This structure-based drug discovery effort ulti- ing the antiapoptotic BCL-2 and IAP family proteins, was curated from mately led to the identification of AZD4573 (Fig. 1A;ref.24),which multiple sources to identify candidate genes whose rapid mRNA and potently inhibits the CDK9 enzyme (IC50 at KM and high ATP protein depletion could drive AZD4573-mediated cell death (Supple- concentrations are <0.003 mmol/L and <0.004 mmol/L, respectively). mentary Methods and Supplementary Table S3). Through this agnos- Kinase selectivity profiling measured the affinity of AZD4573 across tic approach, the only oncogene and antiapoptotic protein with a set of 468 kinases and relevant variants using the KINOMEscan significantly diminished mRNA and protein levels after 4 and 8 hours platform. At a screening concentration of 0.1 mmol/L, which is more of treatment with AZD4573 compared with vehicle treatment was – than 100-fold above the measured IC50 in a CDK9 biochemical MCL-1 (Fig. 2A and B; Supplementary Tables S4 S6). Conversely, the assay, AZD4573 resulted in 90% reduction of binding to the other antiapoptotic BCL-2 and IAP family proteins were quite stable cognate ligand capture matrix for 16 kinases (Fig. 1B). For 14 of (BFL-1 and BCL-B were not detected) despite reports that XIAP was the top hits, the IC50 for AZD4573 was determined at a physio- rapidly modulated upon CDK9 inhibition (28). Many other onco- logically relevant concentration of ATP, revealing >10-fold selec- genes, despite rapid mRNA turnover, either had prolonged protein tivityforCDK9comparedwith13of14kinasesand>100-fold half-lives or were not detected (Supplementary Table S5). This sup- selectivity compared with 8 of 14 kinases (Fig. 1C). Similarly, ports MCL-1 modulation as a major driver of AZD4573-mediated AZD4573 exhibited >25-fold cellular selectivity for CDK9 over tumor cell apoptosis.

AACRJournals.org Clin Cancer Res; 2020 OF3

Cidado et al.

A B

Percent control 0% 0.1% 0.1−1% 1−5% 5−10% 10−35% >35%

C D Phospho Corresponding IC (μmol/L) 10,000 endpoint CDKs 50

Selectivity pSer2-RNAP2 CDK9 0.014 1,000 50 pPP1α T320 CDK1 0.37 >100x

100 pSer5-RNAP2 CDK7 1.1 10-100x pRb S780 CDK4/6 1.1

Fold change over 10

CDK9 high [ATP] IC pRb S807/811 CDK4/6 >10 <10x

1 pRb T821 CDK2 >10 2 1 b b JNK1 CDK7 CDK5 CDK4 CDK6 CDK1 CDK2 CDK3 DYRK2 CDK12 CK1g CK1g GSK3 GSK3 Enzyme

E F 125 Caspase activation 125 Caspase activation Viability Loss of viability 100 100

75 75

50 50 % Activity % Activity

25 25

0 0 05101520 25 12468No wash Time (h) Time of washout (h)

Figure 1. Identiﬁcation of a highly selective and potent CDK9 inhibitor, AZD4573. A, Chemical structure of the CDK9 inhibitor, AZD4573. B, TREEspot image of 468 kinases screened with AZD4573 using the KINOMEscan platform. Percent control corresponds to the amount of kinase bound to its cognate ligand following AZD4573 treatment and relative to vehicle control. C, Biochemical selectivity of 0.1 mmol/L AZD4573 at 5 mmol/L ATP for CDK9 compared with 14 top kinase

hits from KINOMEscan.AZD4573IC50 values for each kinase were determined andplottedasfoldchangeovertheCDK9IC50. Selectivity thresholds are depicted by dashed lines and shaded regions. D, Cellular selectivity for multiple CDKs was determined in MCF-7 cells using putative phosphorylation endpoints for the respective kinase. Lysates of MCF-7 cells treated with a dose response of AZD4573 for 6 hours were immunoblotted and quantiﬁed by densitometry to

derive IC50 values. E, MV-4-11 cells treated with 0.1 mmol/L AZD4573 were assessed for caspase activation and loss of viability at multiple time points relative to controls. F, MV-4-11 cells treated with 0.1 mmol/L AZD4573 were washed out at the indicated times and assessed for caspase activation at 12 hours (red) or viability at 24 hours (blue).

OF4 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

AZD4573, a Selective CDK9 Inhibitor, Targets MCL-1

A 10.0 10.0

MCL-1 MCL-1 7.5 7.5 XIAP

XIAP value

value P P BCL-2 5.0 Bcl-w 5.0

Bcl-w Adj. Adj.

BCL-2 Adj.

10 10

BCL-XL

2.5 2.5

–Log –Log BCL-XL

0.0 0.0 -4 04 -5 0 5 Log2 fold change Log2 fold change B 10.0

7.5 MCL-1

7.5

value

P value P 5.0 MCL-1

5.0

Adj. Adj.

10 10 2.5 2.5

Bcl-w –Log –Log XIAP XIAP Bcl-w BCL-XL BCL-2 BCL-XL 0.0 BCL-2 0.0 -2 -1 012 -3 -2 -1 0 1 23

Log2 fold change Log2 fold change C [AZD4573] (nmol/L)

1,000 3,000 MW (kD) pSer2-RNAP2 -220

MCL-1 -40

BCL-2 -28

BCL-XL -30

Cleaved -19 caspase-3 -17

Vinculin -124

D % pSer2-RNAP2 % BCL2 Protein % MCL1 RNA % BCLXL Protein 125 % MCL1 protein Fold caspase activation 250 Relative fold increase in caspase activation 100 200

75 150

50 100

25 50 Relative expression (%)

0 0 0246810 Time (h)

Figure 2. Acute inhibition of CDK9 with AZD4573 induces apoptosis through rapid depletion of MCL-1. A and B, Differential transcriptomic (A) and proteomic (B) analysis of MCF-7 cells treated with 0.1 mmol/L AZD4573 for 4 hours (left) or 8 hours (right). Changes in transcript and protein levels are defined as not significant (gray, FDR > 0.05), significant (yellow, FDR < 0.05), and significant with a strong effect (orange, FDR < 0.05 and two-fold change). Other key, measurable antiapoptotic proteins are highlighted in the volcano plots. C, Lysates from MV-4-11 cells treated with a range AZD4573 concentrations for 6 hours were immunoblotted for the indicated proteins. D, Protein and RNA lysates from MV-4-11 cells treated with 0.1 mmol/L AZD4573 for 24 hours were harvested at the indicated times. MCL1 transcript level was measured by RT-qPCR, whereas pSer2-RNAP2, MCL-1, BCL-2, BCL-XL, and cleaved caspase-3 protein levels were evaluated by immunoblot. Transcript and protein levels are normalized to the pretreatment time point.

AACRJournals.org Clin Cancer Res; 2020 OF5

Cidado et al.

Transient exposure to AZD4573 induces apoptosis upon tional repression results in general cytotoxicity or cytostasis (Sup- depletion of MCL-1 plementary Fig. S3A), consistent with data from the time-course Effect of acute AZD4573 treatment on MCL-1 was measured in the assays (Fig. 1E; Supplementary Fig. S1B–S1D). sensitive MV-4-11 cell line. AZD4573 treatment for 6 hours reduced The same panel of cancer cell lines was also assessed for 6-hour ¼ pSer2-RNAP2 and MCL-1 in a dose-dependent manner (IC50 14 caspase activation when treated with the selective MCL-1 inhibitor, nmol/L), and upon sufficient MCL-1 depletion, cleaved caspase-3 AZD5991, one of multiple MCL-1 inhibitors in clinical develop- levels increased. Although caspase activation only occurred in sensitive ment (19, 33). There was a strong correlation between the activity cell lines, suppression of pSer2-RNAP2 and MCL-1 occurred in all observed with AZD4573 and AZD5991 (r2 ¼ 0.76; Fig. 3C), providing assayed cell lines, which shows that some cancers rely on survival additional evidence that depletion of MCL-1 is the primary mecha- factors other than MCL-1 (Fig. 2C; Supplementary Fig. S2B–S2D). nism driving AZD4573-mediated cell killing. Therefore, when MV-4- Cleaved CASPASE-3 levels decreased at higher doses in the MV-4-11 11 cells were treated with a combination of AZD4573 and AZD5991 and MOLP-8 models, but this is likely due to the more rapid onset of in vitro and assayed for caspase activation, no combination benefit was cell death at those concentrations and, therefore, earlier loss of signal observed across the dose matrix (Supplementary Fig. S3B). Further- that is consistent with the relative kinetics of apoptosis induction for more, measurement of MCL-1 and BCL-XL protein levels for a subset those cell lines (Supplementary Fig. S1B–S1D). of AZD4573-sensitive and -insensitive cell lines revealed that higher Temporal effects of AZD4573 on relevant biomarkers were examined levels of BCL-XL corresponded to decreased sensitivity to AZD4573 to understand the relationship between rapid biomarker modulation, (Supplementary Fig. S3C). This is consistent with a lack of BCL-XL caspase activation, and loss of cell viability. When MV-4-11 cells were depletion upon acute AZD4573 treatment (Fig. 2D) and previous treated with AZD4573, pSer2-RNAP2 was reduced >80% within 1 hour reports of the MCL-1:BCL-XL expression ratio predicting response to and MCL1 mRNA and protein levels were reduced by >80% by 2 and CDK9 inhibitors and BH3 mimetics (13, 34). 4 hours, respectively (Fig. 2D; Supplementary Fig. S2E). Cleaved caspase-3 levels began to increase between 4 and 6 hours. Importantly, Intermittent dosing of AZD4573 drives regression of MV-4-11 levels of other antiapoptotic proteins, BCL-2 and BCL-XL, remained tumor xenografts relatively unchanged over the same dose range (Fig. 2C) and time frame To determine if observed in vitro activity translated to in vivo (Fig. 2D), consistent with the transcriptomic and proteomic findings. efficacy, AZD4573 was evaluated in a hematologic subcutaneous Because caspase-3 cleavage can occur through nonapoptotic xenograft model using an intermittent dosing schedule that was mechanisms (29), experiments were performed to ensure selected to provide the requisite acute target inhibition identified from AZD4573-mediated caspase activation and subsequent cytotoxicity in vitro studies (6 hours). Mice bearing MV-4-11 tumors were resulted from engagement of the intrinsic apoptosis pathway. Indi- treated weekly with 5 or 15 mg/kg AZD4573, dosed IP BID q2h on vidual knockdown of apoptosis effector molecules, BAK and BAX, in consecutive days (5/5 and 15/15 mg/kg), resulting in a tolerated dose- the diffuse large B-cell lymphoma (DLBCL) cell line OCI-LY10 only dependent response. The 15/15 mg/kg dose led to regressions for all partially reduced caspase activation following 6 hours of AZD4573 mice that were sustained out to more than 125 days (Fig. 4A; treatment, but simultaneous knockdown completely suppressed cas- Supplementary Fig. S4A). Notably, when this study was repeated to pase activation (Supplementary Fig. S2F). In MV-4-11 cells, AZD4573 explore other biweekly schedules, similar robust and durable single- rapidly decreased mitochondrial outer membrane potential and agent activity was observed (data not shown). increased caspase activation and phosphatidylserine exposure as AZD4573 treatment was also compared head-to-head with measured by tetramethylrhodamine ethyl ester staining, Caspase-3/7 AZD5991 in the sensitive MV-4-11 model and relatively insensitive Glo, and Annexin V staining, respectively, consistent with MCL-1 OCI-AML3 model. Both the 15/15 mg/kg AZD4573 regimen and inhibition kinetics and triggering of intrinsic apoptosis (19). These a single i.v. dose of 60 mg/kg AZD5991 resulted in regressions of events were closely followed by loss of cellular ATP, signaling the loss MV-4-11 tumors and minimal tumor growth inhibition (TGI) of cell viability (Supplementary Fig. S2G). OCI-AML3 tumors (Supplementary Fig. S4B and S4C). These strik- ingly similar effects in vivo for the two inhibitors provide further AZD4573 induces rapid cell death across a diverse panel of evidence of the primarily MCL-1–dependent mechanism of action for hematologic cancer cell lines acute AZD4573 treatment. With MCL-1 reportedly essential for the survival of various tumor types, particularly leukemias and lymphomas (30, 31), A mechanistic PK/PD/efficacy model relates CDK9 inhibition, AZD4573 was screened in a diverse panel of cancer cell lines to MCL-1 depletion, and induction of apoptosis in response to assess the breadth of anticancer activity. Caspase activation and loss AZD4573 treatment of viability were measured following 6 and 24 hours of treatment Using a series of studies with MV-4-11 tumor xenografts, a PK/PD/ with AZD4573, respectively. Greater activity was observed with efficacy model was developed to quantify the dynamic relationship AZD4573 across hematologic cancer cell lines (geometric mean between AZD4573 PKs and tumor pSer2-RNAP2 and MCL-1 PDs ¼ m ¼ m fi EC50 0.166 mol/L, GI50 0.023 mol/L) compared with solid (Fig. 4B). Model parameter values were estimated by tting the model ¼ m ¼ in vivo tumor cell lines (geometric mean EC50 6.871 mol/L and GI50 to compiled PK/PD study individual animal data (Supplemen- 0.774 mmol/L), although a subset of solid tumor cell lines was tary Table S2), and an example model fit is shown in Fig. 4B. The free sensitive (Fig. 3A and B; Supplementary Table S7). Importantly, AZD4573 concentration that results in half-maximal inhibition of there was good concordance between caspase and viability data. pSer2-RNAP2 production rate was estimated to be 0.011 to 0.021 m in vitro Nonselective CDK9 inhibitors also exhibited robust activity across mol/L, which was consistent with the IC50 derived from both hematologic and solid tumor lineages, although longer-term studies (Fig. 1D). The kinetics of MCL-1 mRNA and protein mod- readouts were employed (32). Likewise, when viability was assessed ulation were also consistent with in vitro data, demonstrating an after 72-hour continuous AZD4573 exposure, equipotency was estimated mRNA and protein half-life of 5 and 17 minutes, observed across all cell lines suggesting that prolonged transcrip- respectively.

OF6 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

AZD4573, a Selective CDK9 Inhibitor, Targets MCL-1

A * B * 100 100

10 10

( m mol/L) 1 1 50 ( m mol/L)

0.1 50 0.1

MV-4-11 GI MV-4-11 EC50 0.01 0.01 GI50 Caspase EC

0.001 0.001 L a g n L a g n U GI M L i GI er ML CL C n G her M GU A MM BCL T AL r Ski AML A Sk B Lu OvCa Ot BrC Lun OvCa Oth DL DLBCL MCL/BL MCL/BL Hemes Solids Hemes Solids

C 100

( m mol/L) 1 50

0.1

0.01 r 2 =0.76 AZD5991 EC

0.001 0.001 0.01 0.1 1 10 100

AZD4573 EC50 (mmol/L)

Figure 3. AZD4573 induces rapid cell death across a diverse panel of hematologic cancer cell lines. A, AZD4573 was screened for caspase activation at 6 hours across a panel of hematologic (n ¼ 110) and solid (n ¼ 105) tumor cell lines. Each dot represents the EC50 of individual cell lines, and cell lines are grouped by indication. The geometric mean EC50 with 95% conﬁdence interval is depicted within each indication. The dotted line represents the EC50 value for MV-4-11 as a comparison, and an represents a signiﬁcant P value < 0.0001. B, A similar screen was conducted for viability at 24 hours after AZD4573 treatment in hematologic (n ¼ 99) and solid (n ¼ 86) tumor cell lines. GI50 values are plotted with that of MV-4-11 represented by the dotted line. C, A Spearman correlation between the caspase EC50 values for AZD4573 and AZD5991, the MCL-1 inhibitor, in hematologic cancer cell lines (n ¼ 98).

The model was extended to quantify the relationship between extent (Fig. 4B). Each 15 mg/kg split dose was modeled to reduce pSer2- of MCL-1 protein suppression and rate of cell death induction in vivo RNAP2 below 20% of baseline levels for approximately 4 hours. This as measured by reduction in MV-4-11 tumor xenograft volume or extent and duration of pSer2-RNAP2 reduction caused suppression of increase in cleaved caspase-3 in PK/PD study tumor samples. The MCL-1 below the determined Mcl50 for approximately 4 hours, which parameter values describing this relationship were estimated by fitting subsequently resulted in approximately 55% MV-4-11 tumor volume the model to sets of mean tumor volumes from multidose level efficacy reduction that was sufficient to sustain progressive reduction in tumor studies (Supplementary Table S2). The rate of tumor cell death volume on a biweekly dosing schedule (Fig. 4B). Application of this induction was modeled as a saturable first-order process that exhibited derived exposure–effect relationship to other preclinical models, a steep apoptotic response to the suppression of tumor MCL-1. The including OCI-LY10, estimated equivalent parameters and predicted estimated tumor MCL-1 protein level associated with half-maximal the in vivo response (Supplementary Fig. S4D; Supplementary rate of cell death induction (Mcl50) was consistent across studies and Table S2). estimated to be approximately 25% of baseline. The estimated max- fi In vitro in vivo imum rst-order rate constant for cell death induction (Kmax) was also sensitivity to AZD4573 accurately predicts consistent across studies and estimated to be 0.16 to 0.19 hr 1. efficacy in hematologic tumor xenograft models Satisfactory model fits to observed tumor cleaved caspase-3 kinetic Having established a detailed understanding of the relationship data were also obtained using the same Mcl50 and Kmax values to between AZD4573 exposure, MCL-1 suppression, and apoptosis describe the relationship between MCL-1 protein suppression and the induction in the MV-4-11 xenograft model, in vivo efficacy analysis rate of caspase activation from the pool of tumor proCASPASE-3. was expanded to include ten more hematologic models. Based on Modeling suggested that each daily split dose results in the death of a in vitro 6-hour caspase activity parameters used to define sensitivity consistent fraction of xenograft tumor cells, underscored by the similar (EC50 45 nmol/L based on modeled MV-4-11 EC99; maximum overall efficacy for 2-day on/5-day off and biweekly dosing schedules caspase activation 60%), certain models were predicted to be

AACRJournals.org Clin Cancer Res; 2020 OF7

Cidado et al.

A 1,500

1,250 Vehicle (n = 5) ) 3 5/5 mg/kg (n = 5) 1,000 15/15 mg/kg (n = 5)

750

500

250 Tumor volume (mm

0 25 50 75 100 125 150 Days after implant

B Pharmacokinec model Pharmacodynamic models Eﬃcacy models

inhibits Cleaved Peripheral Producon Producon Producon Procasp-3 caspase-3 V2 Cu/(IC50 + Cu) Kcleave Kcasp3

Cl2 Cl IP 3 Plasma Tumor pSer2- MCL-1 MCL-1 inhibits fu = 0.1 Kp Dose V1 p tumor RNAP2 mRNA protein Distribuon

Cl Kout,pser2 Kout,mRNA Kout,Mcl1 K Live death

tumor Dead1 Dead2 Dead3 cells Ktr Ktr Ktr Kgrow Loss of tumor volume

Plasma PK Tumor PK pSer2-RNAP2 MCL-1 protein Cleaved caspase-3 25 100 100 140 160 15/15 mg/kg 140 10 10 120 20 100 120 5/5 mg/kg 100 15 1 1 80 80 0.1 0.1 60 10 15/15 mg/kg 60 model 40 40 0.01 0.01 5 5/5 mg/kg 20 MCL-1 (% baseline) 20 2 x 5 mg/kg IP 5 mg/kg 2 x

Caspase-3 (fold baseline) model pSer2-RNAP2 (% baseline) Tumor concentraon (μmol/L) concentraon Tumor Plasma concentraon (μmol/L) concentraon Plasma 0.001 0.001 0 0 0 02468 02468 02468 02468 04812 Time (hours) Time (hours) Time (hours) Time (hours) Time (hours)

MCL-1 protein Tumor volume Plasma PK Tumor PK pSer2-RNAP2 5/5 mg/kg; biweekly 100 100 140 160 120 140 300 ) 5/5 mg/kg; 2d on, 5d 10 10 3 100 120 oﬀ 100 1 1 80 10/10 mg/kg; 80 biweekly 60 30 0.1 0.1 60 15/15 mg/kg; 40 40 biweekly 0.01 0.01

20 MCL-1 (% baseline) 20 15/15 mg/kg; 2d on, Tumor volume (mm Tumor volume 2 x 15 mg/kg IP 15 mg/kg 2 x pSer2-RNAP2 (% pSer2-RNAP2 baseline) 5d oﬀ Tumor concentraon (μmol/L) concentraon Tumor Plasma concentraon (μmol/L) concentraon Plasma 0.001 0.001 0 0 3 02468 02468 02468 02468 18 25 32 39 Time (hours) Time (hours) Time (hours) Time (hours) Study day

Figure 4. A mechanistic PK/PD/efficacy model relates CDK9 inhibition, MCL-1 depletion, and caspase activation to in vivo response with AZD4573 treatment. A, An MV-4-11 subcutaneous xenograft model was treated with three cycles of vehicle or AZD4573 (5 mg/kg or 15 mg/kg, BID q2h, 2 days on/5 days off). Tumor volumes are presented as geometric mean SEM (n ¼ 5). Shaded area indicates treatment period. B, The top portion depicts a schematic of the established PK/PD/efficacy model. A three-compartment PK model was used to describe the observed plasma and tumor AZD4573 concentration data. Tumor pSer2-RNAP2 and MCL-1 PD was modeled using a series of linked indirect response models. Unperturbed tumor growth was modeled as a first-order process, and MCL-1 was modeled as inhibiting the induction of apoptosis in the MV-4-11 cells. A series of transit compartments were used to bridge the time delay between cell death (hours) and observed tumor shrinkage (days). On the bottom, graphs show the model fit (lines) to observed PK and PD data (individual data points) where darker lines indicate population mean with lighter lines indicating 10% to 90% percentiles expected based on estimated interindividual variability. Efficacy graphs show the model fit to observed tumor volume data. All modeling data are based upon MV-4-11 subcutaneous tumor xenograft studies.

OF8 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

AZD4573, a Selective CDK9 Inhibitor, Targets MCL-1

relatively insensitive (KG1a, OCI-M1, and OCI-AML3), borderline like BIM from BCL-2, which can then become sequestered by sensitive (Karpas422), or mostly sensitive (MV-4-11, SU-DHL-4, MCL-1 (36). Similar BCL-2 family protein dynamics were observed AMO-1, NOMO-1, EOL-1, MOLP-8, and OCI-LY10). In these mod- here. In vitro treatment with venetoclax revealed a rapid (3 hours) els, intermittent dosing of AZD4573 for 2 or 3 cycles produced a range increase in MCL-1 for both models (Fig. 6C). Similarly, SU-DHL-4 of responses, from tumor regressions to modest or no TGI (Fig. 5A; tumor xenograft PD samples treated with a single dose of venetoclax Supplementary Figs. S4B and S4C; S5A–S5I). The eight models pre- exhibited an increase in MCL-1 protein by 6 hours after dose with dicted to show in vivo sensitivity to AZD4573 did achieve either maximal levels achieved by 24 hours (Fig. 6D). In OCI-AML3 cells regressions or extensive TGI, whereas minimal response was observed treated with venetoclax and coimmunoprecipitated with either anti– in the three models predicted to be relatively insensitive. MCL-1 or anti-BIM antibodies, Bim was displaced from BCL-2 and then sequestered by MCL-1 (Fig. 6E), likely stabilizing the protein and AZD4573 demonstrates antitumor activity in disseminated creating an MCL-1 dependency, lending further credence to the leukemia and lymphoma patient-derived xenograft models reported combination mechanism. In addition to subcutaneous cell line xenograft models, in vivo studies were expanded to include i.v. disseminated patient-derived xenograft (PDX) models of AML and T-cell lymphoma. Nine AML Discussion PDX models were evaluated for the ability of AZD4573 to decrease Seminal work with nonselective CDK inhibitors showed that tumor burden in the bone marrow. By the end of study, five of the nine induction of tumor cell death required activity against CDK9 (37), models exhibited greater than 50% reduction of leukemic blasts in the which resulted in the rapid depletion of prosurvival factors, like bone marrow (Fig. 5B). The angioimmunoblastic T-cell lymphoma MCL-1 (7, 38). Despite signs of clinical activity, development of these PDX model, DFTL-78024, which exhibited strong MCL-1 dependence molecules as monotherapies was suspended due to narrow therapeutic based upon BH3 profiling (Supplementary Fig. S5J; ref. 35), was also windows (2). However, trials investigating their combination potential þ evaluated. Treatment with AZD4573 led to significantly more CD45 are currently ongoing. Recently, more selective CDK9 inhibitors have tumor cells undergoing apoptosis, as measured by increases in been developed to overcome the noted limitations of nonselective Annexin V staining following the second dose on day 2, which CDK inhibitors (15, 39, 40). This work describes the novel CDK9 ultimately led to a significant survival benefit compared with vehicle inhibitor, AZD4573, that was optimized for high selectivity and (P value ¼ 0.0018; Fig. 5C). Together, these data establish that potency against CDK9 as well as physicochemical properties AZD4573 is efficacious in models of disseminated disease. enabling intravenous infusion and a short PK half-life. These attributes allow for intermittent dosing of AZD4573, resulting in AZD4573 combinations with venetoclax are tolerated and periods of acute yet sufficient CDK9 inhibition that drives rapid efficacious tumor cell apoptosis without causing prolonged transcription inhi- Despite broad monotherapy activity for AZD4573 across hema- bition that could erode the therapeutic index. AZD4573 is currently tologic tumor models, several models exhibited little to no response in phase I clinical trials for evaluation in patients with hematologic (Supplementary Fig. S5A–S5C).Furthermore,insomesensitive malignancies (NCT03263637). models, AZD4573 resulted in extensive TGI, yet tumors regrew There are various studies that have successfully associated immediately following cessation of dosing, suggesting either the CDK9 inhibition with MCL-1 depletion and subsequent induction presence of an underlying refractory cell population or insufficient of tumor cell death (41). However, many come with study design tumor cell death during the dosing period. To produce more caveatssuchasprolongedCDK9inhibition (15, 42), a narrow durable responses, combinations with the selective BCL-2 inhibitor, focus on one indication and/or a limited number of preclinical venetoclax, were explored given the observed benefit for AZD5991 models (8, 43), and use of nonselective compounds (14, 44) that combinations with venetoclax (19). make it difficult to draw firm conclusions. This study overcame AZD4573 monotherapy treatment in the SU-DHL-4 (GCB-DLBCL) those limitations by utilizing the selective CDK9 inhibitor, tumor xenograft model exhibited robust antitumor activity, although AZD4573, and applying multiple approaches to interrogate the overall regressions were not achieved (Supplementary Fig. S5F). In vitro, mechanism of action of CDK9 inhibition. First, with the agnostic ¼ MCL1 SU-DHL-4 cells were sensitive to single-agent AZD4573 (EC50 16 and integrated transcriptomic and proteomic analysis, was nmol/L; max caspase activation ¼ 79%) and relatively insensitive to identified as the most robustly and significantly modulated cancer- ¼ ¼ venetoclax (EC50 94 nmol/L; max caspase activation 24%). associated gene in cells treated acutely with AZD4573. Interest- Combination treatment, however, increased the extent of caspase ingly, several transcripts and proteins were upregulated after ¼ ¼ activation (EC50 10 nmol/L; max caspase activation 118%), which transient AZD4573 exposure, which could result from CDK9 was also observed in vivo (Fig. 6A and B, left). Monotherapy treatment inhibition driving stress responses or repressing another repres- with venetoclax was minimally efficacious, consistent with the in vitro sor(45).Althoughinclusionofmorethanonecelllineinthese response, and single-agent AZD4573 again exhibited extensive TGI but multiomic experiments would have been preferred for hit valida- not regressions. The combination of AZD4573 with venetoclax, how- tion, focused evaluation of MCL-1 depletion in multiple other ever, produced highly durable regressions in 100% of treated animals, models revealed a high degree of consistency across tumor types with all 8 mice remaining tumor-free out to day 63. Similar effects were and ranges of AZD4573 sensitivity. This work also highlighted the observed in the OCI-AML3 model, which was intrinsically resistant to requirement of an intact intrinsic apoptotic pathway to achieve both monotherapies (Fig. 6A and B, right). Notably, minimal body AZD4573-induced cell death, a high concordance between the weight loss was observed, highlighting that this could be a tolerated activity observed in vitro and in vivo for AZD4573 and the novel combination regimen (Supplementary Fig. S5K and S5L). MCL-1 inhibitor AZD5991 (19), and a lack of benefitforthe The combination benefit in cancer cell line models insensitive to combination of AZD4573 and AZD5991. Collectively, these data inhibition of BCL-2 or both BCL-2 and MCL-1 was striking. Previous support MCL-1 depletion as the primary mechanism driving studies have shown that venetoclax displaces proapoptotic molecules tumor cell death upon acute CDK9 inhibition.

AACRJournals.org Clin Cancer Res; 2020 OF9

Cidado et al.

A 1,000 500 * *

100 TGI

* 0 * *

-50 Regression Change from baseline (%)

-100 1 L3 o P8 10 M m 411 Y KG1a CIM1 AMO1 as422 o EOL1 O IA N MOL MV SUDHL4 OC OCIL Karp

B 1,500 1,000

100

-50 Change from baseline (%) -100

3 5 0 5 4 35 38 28 2 2 -224 24 -22 -2 -2 -22 G G TG TG-2229 TG C CTG- CTG-2242C CT C CTG-2226CT CTG

C 100 Vehicle (n = 5) 100 Vehicle (n = 3) * AZD4573 (n = 5) AZD4573 (n = 3) 80 80 cells

+ 60 60 * 40

40 20 % hCD45 % Survival 20 0 Neg. Pos. Annexin V 0 0 20406080100 Day of treatment

Figure 5. AZD4573 demonstrates antitumor activity in disseminated leukemia and lymphoma PDX models. A, Other subcutaneous tumor xenograft models from AML, multiple myeloma (MM), and DLBCL cancer cell lines were treated with AZD4573 at 15 mg/kg, BID q2h, 2 days on/5 days off for three cycles ( represents models receiving only two cycles). The percent change from tumor volume at time of first dose is depicted. Red bars denote models predicted from in vitro caspase sensitivity parameters to show signs of efficacy in vivo. B, Human disseminated AML PDX models grown in NOG-EXL mice were treated with four cycles of vehicle or AZD4573 (15 mg/kg, BID q2h, 2-day on/5-day off; n ¼ 3). FACS analysis was used to detect leukemic blasts in the bone marrow (human CD3/CD33þ) and presented as mean percent change SEM from pretreatment counts. C, Kaplan–Meier survival curves for the DFTL-78024 model treated with four cycles of AZD4573 (n ¼ 5; left). PD response to vehicle or AZD4573 treatment assessed 6 hours after a single 15 mg/kg dose. Tumors from 3 mice per group were assessed for Annexin V staining in human CD45þ tumor cells via flow cytometry (right).

OF10 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

AZD4573, a Selective CDK9 Inhibitor, Targets MCL-1

Figure 6. A SU-DHL-4 OCI-AML3 125 AZD4573 AZD4573 combinations with venetoclax are 125 AZD4573 Venetoclax ﬁ Venetoclax tolerated and ef cacious. A, SU-DHL-4 (left) 100 100 AZD4573 + AZD4573 + and OCI-AML3 (right) cells treated in vitro Venetoclax Venetoclax 75 with a dose-response of AZD4573, veneto- 75

50 clax, or the combination were assessed for 50 caspase activation. B, SU-DHL-4 (left) and 25 OCI-AML3 (right) subcutaneous xenograft 25 0 Caspase activation (%) models were treated with three cycles of Caspase activation (%) 0 vehicle, venetoclax (100 mg/kg QD), -25 -25 0.001 0.01 0.1 1 AZD4573 (15 mg/kg, BID q2h, 2-day on/ 0.001 0.01 0.1 1 Dose (mmol/L) 5-day off), or the combination. Shaded area Dose (mmol/L) indicates treatment period. Tumor volumes are presented as geometric mean SEM B 2,000 2,500 Vehicle (n = 5) Vehicle (n = 5) n ¼ n ( 5). C, Lysates from SU-DHL-4 (left) ) ) AZD4573 ( = 5) n 3 3 1,600 2,000 AZD4573 ( = 5) Venetoclax (n = 5) and OCI-AML3 (right) cells treated with Venetoclax (n = 5) AZD4573 + Venetoclax (n = 5) AZD4573 + Venetoclax (n = 5) 0.1 mmol/L venetoclax, 0.1 mmol/L AZD4573, 1,200 1,500 or the combination for various times were immunoblotted for the depicted proteins. D, 800 1,000 SU-DHL-4 tumor lysates from mice treated 500 with a single dose of vehicle or venetoclax 400 Tumor volume (mm Tumor volume (mm (100 mg/kg) for the indicated times (n ¼ 3) 0 0 were immunoblotted for the indicated pro- 20 40 60 80 100 15 20 25 30 35 40 45 50 teins. E, OCI-AML3 cells were treated with Days after implant Days after implant 0.1 mmol/L venetoclax in vitro for the indi- C Venetoclax AZD4573 Venetoclax AZD4573 Combinaon Combinaon cated times. MCL-1 and Bim were immuno- 100 nmol/L 100 nmol/L 100 nmol/L 100 nmol/L precipitated from whole-cell lysates (input) MW (kD) Time (h) 036036036 Time (h) 036036036MW (kD) and immunoblotted for the indicated -220 pSer-RNAP2 pSer-RNAP2 -220 protein. -40 MCL-1 MCL-1 -40

BCL-2 -28 BCL-2 -28

BCL-XL -30 BCL-XL -30

-23 -23

Bim -15 Bim -15 -12 -12

Cleaved -19 Cleaved -19 caspase-3 -17 caspase-3 -17

Vinculin -124 Vinculin -124

Venetoclax D Vehicle Venetoclax E 100 nmol/L Time (h) Time (h) 0 3 6 24 48 MW (kD) 0123 MW (kD) -23 MCL-1 -40 Bim -15 BCL-2 -28 IP: Mcl1 -12

BCL-XL -30 MCL-1 -40

-23 BCL2 -28 Bim -15 BCL-XL -30 -12 IP: Bim -23 Vinculin - 124 Bim -15 -12

MCL-1 -40

BCL-2 -28

BCL-XL -30

Input -23 Bim -15 -12

Vinculin -124

Despite the strong correlation between AZD4573 and AZD5991 AZD4573 is likely rapidly depleting another factor necessary for cancer activity in vitro, there were some hematologic cancer cell lines sensitive cell survival where MCL-1 inhibition alone is insufﬁcient to induce to AZD4573 but not AZD5991. Although prolonged AZD4573 expo- apoptosis. sure would expectedly cause turnover of numerous proteins that could A quantitative AZD4573 exposure/effect relationship was also cause general cytotoxicity, these were short, 6-hour assays. Therefore, established that further substantiates MCL-1 as the downstream target

AACRJournals.org Clin Cancer Res; 2020 OF11

Cidado et al.

mediating the induction of apoptosis. Reduction in pSer2-RNAP2 understanding enabled the development of a quantitative PK/PD/ occurs in a concentration- and time-dependent manner that precedes a efficacy model that can aid in the selection of an optimized clinical dose proportionate but delayed depletion of MCL-1 protein, and caspase and schedule to maximize the therapeutic index. These findings activation occurs once MCL-1 levels reach a critical value in AZD4573- collectively supported progression of AZD4573 as a clinical candidate sensitive cancer cells. The relationship between PD response and rate for the treatment of patients with hematologic malignancies. of cell death induction, as assessed by reduction in tumor volume, in vivo in vitro was similar to that observed in studies. At tolerated Disclosure of Potential Conflicts of Interest AZD4573 doses in mice, the extent and duration of MCL-1 suppres- fi J. Cidado, S. Boiko, T. Proia, D. Ferguson, S.W. Criscione, P. Pop-Damkov, sion was suf cient to drive MV-4-11 subcutaneous tumor xenografts W. Shao, J.C. Saeh, S.E. Fawell, and L. Drew are employees/paid consultants for into complete regression when using twice-weekly dose schedules. A AstraZeneca. D.M. Weinstock is an employee/paid consultant for Travera, Ajax, and mechanistic PK/PD/efficacy model was developed that relates Bantam Pharmaceuticals, reports receiving commercial research grants from Astra- AZD4573 exposure to the extent and duration of pSer2-RNAP2 and Zeneca, Abbvie, Aileron, Daiichi Sankyo, and Verastem, holds ownership interest MCL-1 suppression necessary to induce caspase activation and reduce (including patents) in Ajax and Travera, and reports receiving other remuneration from Genentech/Roche. M. Zinda is an employee/paid consultant for, reports tumor size. The derived parameter values for the in vivo model were in vitro receiving commercial research grants from, and holds ownership interest (including consistent with the observed target engagement requirement patents) in AstraZeneca. No potential conflicts of interest were disclosed by the other (6 hours) in MCL-1–dependent cell lines. Given the applicability in authors. multiple tumor models, this PK/PD/efficacy model was used to inform design of the current phase I trial. With this greater mechanistic Authors’ Contributions understanding and quantitative modeling of AZD4573, the hope is to Conception and design: J. Cidado, S. Boiko, D. Ferguson, S.W. Criscione, S.E. Fawell, optimize the dose and schedule to maximize the therapeutic window. L. Drew Venetoclax is now approved for treatment of adult patients with Development of methodology: J. Cidado, S. Boiko, D. Ferguson, P. Pop-Damkov, chronic lymphocytic leukemia based upon high response rates with L. Drew Acquisition of data (provided animals, acquired and managed patients, provided monotherapy. In other hematologic cancers, the response rates were facilities, etc.): J. Cidado, S. Boiko, T. Proia, M. San Martin, N. Su, V.N. Roamio much lower, but combinations with standards of care yielded better Franklin, C.S. D'Santos, R. Koch, D.M. Weinstock results and led to the approval of venetoclax in combination with Analysis and interpretation of data (e.g., statistical analysis, biostatistics, hypomethylating agents or low-dose cytarabine for the treatment of computational analysis): J. Cidado, S. Boiko, T. Proia, D. Ferguson, select patients with AML who cannot receive intensive induction S.W. Criscione, N. Su, C. Sekhar Reddy Chilamakuri, C.S. D'Santos, R. Koch, chemotherapy (46, 47). Despite these promising responses and excit- S.E. Fawell Writing, review, and/or revision of the manuscript: J. Cidado, S. Boiko, T. Proia, ing approvals, acquired resistance to venetoclax is beginning to D. Ferguson, N. Su, W. Shao, J.C. Saeh, R. Koch, D.M. Weinstock, M. Zinda, emerge, and one mechanism identified preclinically is compensation S.E. Fawell, L. Drew by increased levels of MCL-1 (48). Although this could result from Administrative, technical, or material support (i.e., reporting or organizing data, selection of subclones with preexisting MCL-1 overexpression, our constructing databases): J. Cidado, D. Ferguson, J.C. Saeh, M. Zinda, L. Drew data suggest that venetoclax can directly induce stabilization of MCL-1 Study supervision: J. Cidado, T. Proia, W. Shao, J.C. Saeh, S.E. Fawell, L. Drew by favoring its binding of BIM. Using models intrinsically resistant to venetoclax, this study showed that adding AZD4573 to the venetoclax The Fusion Lumos Orbitrap mass spectrometer was purchased with the support from a Wellcome Trust Multi-user Equipment Grant (Grant #108467/Z/15/Z). regimen resulted in significant and tolerated combination benefit, which is in line with venetoclax combination studies using other CDK9 and MCL-1 inhibitors (14, 49, 50). This suggests that the combination Acknowledgments could be used to increase depth and duration of response as well as The costs of publication of this article were defrayed in part by the payment of page prevent or overcome the emergence of resistance. charges. This article must therefore be hereby marked advertisement in accordance In conclusion, this study has identified AZD4573, a potent and with 18 U.S.C. Section 1734 solely to indicate this fact. selective CDK9 inhibitor, that induces cancer cell death across preclinical hematologic cancer models at tolerated, intermittent doses Received June 11, 2019; revised September 27, 2019; accepted November 4, 2019; primarily through transient depletion of MCL-1. This mechanistic published first November 7, 2019.

References 1. Lim S, Kaldis P. Cdks, cyclins and CKIs: roles beyond cell cycle regulation. kinase inhibitor, administered weekly in subjects with advanced malignancies. Development 2013;140:3079–93. J Transl Med 2013;11:259. 2. Asghar U, Witkiewicz AK, Turner NC, Knudsen ES. The history and future of 6. Flynn J, Jones J, Johnson AJ, Andritsos L, Maddocks K, Jaglowski S, et al. targeting cyclin-dependent kinases in cancer therapy. Nat Rev Drug Discov 2015; Dinaciclib is a novel cyclin-dependent kinase inhibitor with signiﬁcant clinical 14:130–46. activity in relapsed and refractory chronic lymphocytic leukemia. Leukemia 3. Tong WG, Chen R, Plunkett W, Siegel D, Sinha R, Harvey RD, et al. Phase I and 2015;29:1524–9. pharmacologic study of SNS-032, a potent and selective Cdk2, 7, and 9 inhibitor, 7. Chen R, Wierda WG, Chubb S, Hawtin RE, Fox JA, Keating MJ, et al. Mechanism in patients with advanced chronic lymphocytic leukemia and multiple myeloma. of action of SNS-032, a novel cyclin-dependent kinase inhibitor, in chronic J Clin Oncol 2010;28:3015–22. lymphocytic leukemia. Blood 2009;113:4637–45. 4. Mahadevan D, Plummer R, Squires MS, Rensvold D, Kurtin S, Pretzinger C, et al. 8. MacCallum DE, Melville J, Frame S, Watt K, Anderson S, Gianella-Borradori A, A phase I pharmacokinetic and pharmacodynamic study of AT7519, a cyclin- et al. Seliciclib (CYC202, R-Roscovitine) induces cell death in multiple myeloma dependent kinase inhibitor in patients with refractory solid tumors. Ann Oncol cells by inhibition of RNA polymerase II-dependent transcription and down- 2011;22:2137–43. regulation of Mcl-1. Cancer Res 2005;65:5399–407. 5. Nemunaitis JJ, Small KA, Kirschmeier P, Zhang D, Zhu Y, Jou YM, et al. A ﬁrst- 9. Gregory GP, Hogg SJ, Kats LM, Vidacs E, Baker AJ, Gilan O, et al. CDK9 in-human, phase 1, dose-escalation study of dinaciclib, a novel cyclin-dependent inhibition by dinaciclib potently suppresses Mcl-1 to induce durable apoptotic

OF12 Clin Cancer Res; 2020 CLINICAL CANCER RESEARCH

AZD4573, a Selective CDK9 Inhibitor, Targets MCL-1

responses in aggressive MYC-driven B-cell lymphoma in vivo. Leukemia 2015; 29. Garrido C, Kroemer G. Life's smile, death's grin: vital functions of apoptosis- 29:1437–41. executing proteins. Curr Opin Cell Biol 2004;16:639–46. 10. Sanso M, Fisher RP. Pause, play, repeat: CDKs push RNAP II's buttons. 30. Koss B, Morrison J, Perciavalle RM, Singh H, Rehg JE, Williams RT, et al. Transcription 2013;4:146–52. Requirement for antiapoptotic MCL-1 in the survival of BCR-ABL B-lineage 11. Lam LT, Pickeral OK, Peng AC, Rosenwald A, Hurt EM, Giltnane JM, et al. acute lymphoblastic leukemia. Blood 2013;122:1587–98. Genomic-scale measurement of mRNA turnover and the mechanisms of action 31. Kelly GL, Grabow S, Glaser SP, Fitzsimmons L, Aubrey BJ, Okamoto T, et al. of the anti-cancer drug flavopiridol. Genome Biol 2001;2:RESEARCH0041. Targeting of MCL-1 kills MYC-driven mouse and human lymphomas even 12. Lanasa MC, Andritsos L, Brown JR, Gabrilove J, Caligaris-Cappio F, Ghia P, et al. when they bear mutations in p53. Genes Dev 2014;28:58–70. Final results of EFC6663: a multicenter, international, phase 2 study of alvocidib 32. Parry D, Guzi T, Shanahan F, Davis N, Prabhavalkar D, Wiswell D, et al. for patients with fludarabine-refractory chronic lymphocytic leukemia. Leuk Res Dinaciclib (SCH 727965), a novel and potent cyclin-dependent kinase inhibitor. 2015;39:495–500. Mol Cancer Ther 2010;9:2344–53. 13. Booher RN, Hatch H, Dolinski BM, Nguyen T, Harmonay L, Al-Assaad AS, et al. 33. Kotschy A, Szlavik Z, Murray J, Davidson J, Maragno AL, Le Toumelin-Braizat MCL1 and BCL-xL levels in solid tumors are predictive of dinaciclib-induced G, et al. The MCL1 inhibitor S63845 is tolerable and effective in diverse cancer apoptosis. PLoS One 2014;9:e108371. models. Nature 2016;538:477–82. 14. Dey J, Deckwerth TL, Kerwin WS, Casalini JR, Merrell AJ, Grenley MO, et al. 34. Soderquist RS, Crawford L, Liu E, Lu M, Agarwal A, Anderson GR, et al. Voruciclib, a clinical stage oral CDK9 inhibitor, represses MCL-1 and sensitizes Systematic mapping of BCL-2 gene dependencies in cancer reveals molecular high-risk diffuse large B-cell lymphoma to BCL2 inhibition. Sci Rep 2017;7: determinants of BH3 mimetic sensitivity. Nat Commun 2018;9:3513. 18007. 35. Ng SY, Yoshida N, Christie AL, Ghandi M, Dharia NV, Dempster J, et al. 15. Narita T, Ishida T, Ito A, Masaki A, Kinoshita S, Suzuki S, et al. Cyclin-dependent Targetable vulnerabilities in T- and NK-cell lymphomas identified through kinase 9 is a novel specific molecular target in adult T-cell leukemia/lymphoma. preclinical models. Nat Commun 2018;9:2024. Blood 2017;130:1114–24. 36. Siu KT, Huang C, Panaroni C, Mukaihara K, Fulzele K, Soucy R, et al. BCL2 16. Wuilleme-Toumi S, Robillard N, Gomez P, Moreau P, Le Gouill S, Avet-Loiseau blockade overcomes MCL1 resistance in multiple myeloma. Leukemia 2019;33: H, et al. Mcl-1 is overexpressed in multiple myeloma and associated with relapse 2098–102. and shorter survival. Leukemia 2005;19:1248–52. 37. Cai D, Latham VM Jr, Zhang X, Shapiro GI. Combined depletion of cell cycle and 17. Glaser SP, Lee EF, Trounson E, Bouillet P, Wei A, Fairlie WD, et al. Anti- transcriptional cyclin-dependent kinase activities induces apoptosis in cancer apoptotic Mcl-1 is essential for the development and sustained growth of acute cells. Cancer Res 2006;66:9270–80. myeloid leukemia. Genes Dev 2012;26:120–5. 38. Krystof V, Baumli S, Furst R. Perspective of cyclin-dependent kinase 9 (CDK9) as 18. Balko JM, Giltnane JM, Wang K, Schwarz LJ, Young CD, Cook RS, et al. a drug target. Curr Pharm Des 2012;18:2883–90. Molecular profiling of the residual disease of triple-negative breast cancers after 39. Yin T, Lallena MJ, Kreklau EL, Fales KR, Carballares S, Torrres R, et al. A novel neoadjuvant chemotherapy identifies actionable therapeutic targets. CDK9 inhibitor shows potent antitumor efficacy in preclinical hematologic Cancer Discov 2014;4:232–45. tumor models. Mol Cancer Ther 2014;13:1442–56. 19. Tron AE, Belmonte MA, Adam A, Aquila BM, Boise LH, Chiarparin E, et al. 40. Lucking U, Scholz A, Lienau P, Siemeister G, Kosemund D, Bohlmann R, et al. Discovery of Mcl-1-specific inhibitor AZD5991 and preclinical activity in Identification of atuveciclib (BAY 1143572), the first highly selective, clinical multiple myeloma and acute myeloid leukemia. Nat Commun 2018;9:5341. PTEFb/CDK9 inhibitor for the treatment of cancer. ChemMedChem 2017;12: 20. Papachristou EK, Kishore K, Holding AN, Harvey K, Roumeliotis TI, 1776–93. Chilamakuri CSR, et al. A quantitative mass spectrometry-based approach to 41. Morales F, Giordano A. Overview of CDK9 as a target in cancer research. monitor the dynamics of endogenous chromatin-associated protein complexes. Cell Cycle 2016;15:519–27. Nat Commun 2018;9:2311. 42. Rahaman MH, Yu Y, Zhong L, Adams J, Lam F, Li P, et al. CDKI-73: an orally 21. Verma B, Tati S, Wesa A. Champions Oncology acute myeloid leukemia bioavailable and highly efficacious CDK9 inhibitor against acute myeloid models for ex vivo and in vivo study. Champions Oncology. 2018. Available leukemia. Invest New Drugs 2019;37:625–35. from: https://championsoncology.com/wp-content/uploads/2018/11/Champions- 43. O' Reilly E, Dhami SPS, Baev DV, Ortutay C, Halpin-McCormick A, Morrell R, Oncology-AML-White-Paper.pdf. et al. Repression of Mcl-1 expression by the CDC7/CDK9 inhibitor PHA-767491 22. Townsend EC, Murakami MA, Christodoulou A, Christie AL, Koster J, DeSouza overcomes bone marrow stroma-mediated drug resistance in AML. Sci Rep TA, et al. The public repository of xenografts enables discovery and randomized 2018;8:15752. phase II-like trials in mice. Cancer Cell 2016;30:183. 44. Gojo I, Zhang B, Fenton RG. The cyclin-dependent kinase inhibitor flavopiridol 23. Koch R, Christie AL, Crombie JL, Palmer AC, Plana D, Shigemori K, et al. induces apoptosis in multiple myeloma cells through transcriptional repression Biomarker-driven strategy for MCL1 inhibition in T-cell lymphomas. Blood and down-regulation of Mcl-1. Clin Cancer Res 2002;8:3527–38. 2019;133:566–75. 45. Zhang H, Pandey S, Travers M, Sun H, Morton G, Madzo J, et al. Targeting CDK9 24. Pike KB, Barlaam BC, Hawkins J, De Savi C, Vasbinder MM, Hird A, et al. reactivates epigenetically silenced genes in cancer. Cell 2018;175:1244–58. Polycyclic amide derivatives as CDK9 inhibitors. Patent WO2017001354. 2017 46. Valentin R, Grabow S, Davids MS. The rise of apoptosis: targeting apoptosis in Jan 5. hematologic malignancies. Blood 2018;132:1248–64. 25. Baker A, Gregory GP, Verbrugge I, Kats L, Hilton JJ, Vidacs E, et al. The CDK9 47. DiNardo CD, Pratz K, Pullarkat V, Jonas BA, Arellano M, Becker PS, et al. inhibitor dinaciclib exerts potent apoptotic and antitumor effects in preclinical Venetoclax combined with decitabine or azacitidine in treatment-naive, elderly models of MLL-rearranged acute myeloid leukemia. Cancer Res 2016;76: patients with acute myeloid leukemia. Blood 2019;133:7–17. 1158–69. 48. Leverson JD, Cojocari D. Hematologic tumor cell resistance to the BCL-2 26. Polier G, Ding J, Konkimalla BV, Eick D, Ribeiro N, Kohler R, et al. Wogonin and inhibitor venetoclax: a product of its microenvironment? Front Oncol 2018;8: related natural flavones are inhibitors of CDK9 that induce apoptosis in cancer 458. cells by transcriptional suppression of Mcl-1. Cell Death Dis 2011;2:e182. 49. Bogenberger J, Whatcott C, Hansen N, Delman D, Shi CX, Kim W, et al. 27. Moriya H. Quantitative nature of overexpression experiments. Mol Biol Cell Combined venetoclax and alvocidib in acute myeloid leukemia. Oncotarget 2015;26:3932–9. 2017;8:107206–22. 28. Xie S, Jiang H, Zhai XW, Wei F, Wang SD, Ding J, et al. Antitumor action of CDK 50. Caenepeel S, Brown SP, Belmontes B, Moody G, Keegan KS, Chui D, et al. AMG inhibitor LS-007 as a single agent and in combination with ABT-199 against 176, a selective MCL1 inhibitor, is effective in hematologic cancer models alone human acute leukemia cells. Acta Pharmacol Sin 2016;37:1481–9. and in combination with established therapies. Cancer Discov 2018;8:1582–97.

AACRJournals.org Clin Cancer Res; 2020 OF13

AZD4573 Is a Highly Selective CDK9 Inhibitor That Suppresses MCL-1 and Induces Apoptosis in Hematologic Cancer Cells

Justin Cidado, Scott Boiko, Theresa Proia, et al.

Clin Cancer Res Published OnlineFirst November 7, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-19-1853

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2019/11/07/1078-0432.CCR-19-1853.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2020/01/13/1078-0432.CCR-19-1853. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.