Treating the Patient with High-Risk MDS: Practical Strategies and Emerging Therapeutic Options

Treating the Patient with High‐risk MDS: Practical Strategies and Emerging Therapeutic Options

David A. Sallman, MD Assistant Member Malignant Hematology Department Moffitt Cancer Center Tampa, Florida

My name is David Sallman and I am an assistant member in the Malignant Hematology Department at Moffitt Cancer Center in Tampa, Florida, and it is an honor to present treatment strategies for high‐risk myelodysplastic syndrome.

Disclosures

• Research Funding: Jazz • Research Funding/Speakers’ Bureau: Celgene • Speakers’ Bureau: Agios, Incyte, Novartis

These are my disclosures.

Learning Objectives

• Identify clinical‐, patient‐, and disease‐related conditions that impact selection of therapy in high‐risk MDS patients • Correlate these factors with recently approved agents or ongoing clinical trials of novel agents and combination of agents being investigated for use in patients with high‐risk MDS • Outline practical strategies developed in conjunction with clinical trials for optimizing the use of novel agents in patients with high‐risk, relapsed/refractory MDS

As far as the learning objectives today there are really three. One is to first look at the clinical patient and disease factors which largely inform us on the prognosis of our patients but will also impact the selection of therapy. Additionally, to correlate these factors both with approved agents as well as ongoing clinical trials of novel agents and combinations of both standard therapies with novel agents. Then lastly, to outline some practical strategies where trials can be utilized, again in particularly with novel agents for again high risk, both for frontline and relapsed/refractory patients.

© 2019 MediCom Worldwide, Inc. 3 Treating the Patient with High‐risk MDS: Practical Strategies and Emerging Therapeutic Options Defining Higher Risk MDS (Clinical Prognostic Models) • Higher risk MDS = higher chance of AML transformation and worse overall survival • Historically, IPSS is the most common tool used to define higher risk MDS [ie, int‐2 and high risk (higher) versus int‐1 and low risk (lower)] • One‐third of MDS patients are classified as int‐2 or high risk by IPSS with expected overall survival <1.5 years

First, it is important to find what does higher risk myelodysplastic syndrome mean. Basically, higher risk MDS means a higher chance of leukemic transformation and overall an inferior overall survival. The IPSS or International Prognostic Scoring System has been the most widely utilized system historically. It is somewhat easy to use and that it classifies basically lower risk and higher risk by two categories. Patients who are intermediate‐2 or high are considered higher risk, and those who are intermediate‐1 or low are considered lower risk. Overall, if you look at a population basis, about a third of patients are classified as higher risk with a median overall survival of less than one‐and‐a‐half years.

IPSS‐R Risk Groups

%of Median Time until 25% of patients Risk group Points Patients survival, years develop AML, years Very low ≤1.5 19% 8.8 Not reached Low >1.5 – 3 38% 5.3 10.8 Intermediate >3 – 4.5 20% 3.0 3.2 High >4.5 – 6 13% 1.6 1.4 Very High >6 10% 0.8 0.73

Very low Low Int High Very high

100 100

80 80

60 60

40 40 Patients, % Patients, %

20 20

0 0 024681012 0 2 46 81012 Overall Survival, years Time to AML Evolution, years

Adapted from Greenberg PL, et al. Blood. 1997;89:2079‐2088.

Now since the IPSS there has been multiple newer prognostic systems that have been published on. This is probably the most widely currently used system which is the IPSS‐R or the revised IPSS system. This now stratifies patients into five different survival curves, and you could see very nicely predicts both overall survival as well as leukemic transformation. Patients who are very low and low are sort of stereotypical lower‐risk patients, or high and very high are high‐risk patients, and intermediate does fall in between there.

Intermediate Risk IPSS‐R Prognostic Score Variable Coefficient Score No. Patients Age ≥66 years No 0 143 Yes 0.87 2 152 Peripheral blood blasts ≥2% No 0 247 Yes 0.52 1 48 Red blood cell transfusion No 0 199 Yes 0.51 1 96

54% 73%

Benton C, et al. Am J Hematol. 2018;93(10):1245‐1253.

So we, along with the MD Anderson and Cleveland Clinic group, have looked at this question. This is a relatively recent publication, where we try to focus just at the intermediate‐risk group and with clinical components could we further stratify patients. So basically this created a simple system of age greater or equal to 66 years of age. An increased peripheral blood blast count as well as RBC transfusion dependence, and basically if you had a score of 2 or higher, that then put you into the intermediate adverse category and you could see nice separation of curves as far as overall survival. A relatively easy system to utilize.

Genes Recurrently Mutated in MDS

Tyrosine Kinase Pathway Transcription Factors Others

JAK2 KRAS BRAF RUNX1 GATA2 TP53 NPM1 ETV6 NRAS NOTCH? RTKs MAML? CBL WT1 PHF6 ZSWIM4? PTPN11 BCOR UMODL1? Epigenetic Dysregulation Splicing Factors ZRSF2 IDH EZH2 U2AF1 1 & 2 DNMT3A SF3B1

SETBP1 PRPF40B U2AF2 TET2 UTX ASXL1 SRSF2 PRPF8 SF1 ATRX SF3A1

Courtesy of Bejar R.

I think more importantly than that is the identification of somatic mutations, the most critical work starting back in 2011 has really brought forth the paradigm change in how we think about patients. I think firstly and what we will discuss is its impact on prognosis, but more importantly leading to personalize therapy for our patients. Basically the size of the circle on this slide represent the frequency of the mutation, so you can see mutations in both epigenetic dysregulation as well as splicing factors represent the most common mutations. They occur each in about 50% of MDS patients. There are some critical transcription factors TP53 which is something kind of dear to my heart and we will talk to it in detail which is a group with really the poorest outcomes, and then unfortunately some of the signaling or tyrosine kinase pathways are relatively uncommon which at least up until recently has led to a lack of targeted agents, at least from a molecular perspective. But all of this is going to inform again both prognosis and treatment selection.

IWG‐PM Collaborative MDS Sample Compilation

MDS sample data collected from 18 centers in Data Summary Europe, the United States, and Asia Clinical features ‐ Age and sex ‐Blast % ‐Karyotype ‐Hemoglobin ‐ Platelet count ‐ Neutrophil count

3562 Overall survival data MDS ‐ Available for 3359 ‐ 3.6 years follow‐up ‐ 1780 deaths ‐ Median OS 2.65 years

Treatment Status Gene Mutations

Courtesy of Bejar R; IWG‐PM.

There are many papers that have looked at the impact of these mutations on prognosis. I think one of the most important collaborations is called the IWG‐PM. My center at Moffitt Cancer Center is part of this, but this is over 18 centers across the US, Europe, and Asia, and this is data from several years ago presented by Rafael Bejar. This is data in over 3000 patients. This has continued to accrue. Again, looking at what the impact of mutations for these patients are.

Prognosis of Mutations in MDS

Kaplan‐Meier curve of overall survival in years for the 2504 patients with sequence results for SF3B1 and all six adverse genes (TP53, CBL, EZH2, RUNX1, U2AF1, and ASXL1) Bejar R, et al. ASH 2015.

And so what we are able to show is mutations of one of six genes including TP53, CBL, EZH2, RUNX1, U2AF1, ASXL1, each could predict for an inferior overall survival. Again, this is independent of clinical prognostic models where only mutations of SF3B1 and the absence of an adverse risk mutation could predict for an improved overall survival. I think ideally the goal of our International Working Group is to create an optimal molecular model and this is something that we should see in the years to come.

Overall Survival by Mutation Number Overall SurvivalOverall (%)

Bejar R, et al. Blood. 2015;126:907.; Bejar R; IWG‐PM.

So even specific mutations have an inferior prognosis, actually the more simple thing is just to add up the number of mutations. So each additional mutation actually further leads to inferior outcomes. Both from a leukemia free survival which this is a paper from Papaemmanuil and colleagues on the left as well as work from our IWG group where again you can see a really poor outcome in patients that have three or more mutations. Again, this is independent of the specific mutation.

Incorporation of Mutation Data and IPSS‐R

Nazha A, et al. Leukemia. 2016;30(11):2214‐2220.

Can we incorporate mutation data into our clinical prognostic model? This is a nice paper from the Cleveland Clinic Group where they look at the impact of each mutation in combination with the IPSS‐R and indeed one of three genes p53, SF3B1 and EZH2 could further improve this molecular model. I think notably what the curves in the bottom were able to show is that this was sort of dynamic. Patients could be reassessed over time or with therapy or a relapsed/refractory settings and the model was still prognostically relevant.

TP53 Mutant

TP53 Wildtype

P < .0001 Overall Survival

Time (days) Median OS 162 days MT versus NR in WT (HR 2.64; CI 2.00 to 6.31)

Sallman D, et al. Leukemia. 2016;30(3):666‐673.

Now focusing a little bit at p53, which we are going to kind of hear throughout the talk. This is again the group that does the worst. The data here shows what has been seen by multiple groups where the TP53 mutations do predict for an inferior overall survival. But we asked whether or not just the presence or absence, we know that is relevant, but what about the burden, and burden can be associated with the variance allele frequency,

MCC Cohort King’s College Cohort

TP53 VAF > 40% TP53 VAF > 40% TP53 VAF < 20% TP53 VAF < 20% P = .01 P = .01 Overall Survival Overall Survival

Time (days) Time (days) Median OS 124 days > 40% and NR in < 20% Median Survival 272 days > 40% and 1372 days in 20% (HR 3.5; C.I. 1.24 to 6.50) (HR 4.94, CI 1.43 to 9.21

Sallman D, et al. Leukemia. 2016;30(3):666‐673.

and so what we were able to show is patients that had a high allele burden had significantly poor outcomes. You can see this median survival is in days versus patients that had a lower allele frequency and importantly we were able to validate these findings with a cohort from Kings College. This work has also been similarly shown as part of a recent IWG publication.

Treatment Algorithm: Higher‐Risk MDS

Favorable Comorbidities AHSCT Donor functional status Unfavorable Start HMA AHSCT candidate Start HMA No donor Continue HMA Combo Trial

Primary or secondary failure

Investigational

Modified from NCCN Clinical Practice Guidelines in Oncology. MDS. Version1.2020.

Now switching gears a little bit to talk about the treatment algorithm for higher‐risk MDS patients, again the most recent update that is available is version 1.2020. Overall hypomethylating agents, both azacitidine or decitabine, are our standard frontline option. Azacitidine at a seven‐day schedule is considered a category 1 recommendation, this is because this is the only agent that has improved overall survival in the phase 3 setting, although decitabine is also a first‐line recommendation. Considerations and special circumstances can be for intensive chemotherapy, relatively a rare subgroup for that. Then the question of whether or not to proceed directly to transplant or not is dependent on a number of factors including the patient's current performance status, bone marrow blast count, whether or not they have a donor that can be relatively quickly identified, and so in general, most patients are started on a hypomethylating agent while donor identification, let us say if they are less than 75 years of age is ongoing. If they overall are favorable, so they do not have significant comorbidities, they have adequate functional status and adequate donor, typically after two to four cycles of hypomethylating agent the patient is then transplanted assuming that they have not had progressive disease. Whereas patients that do not meet those characteristics or again do not have a donor or really are not qualified, again, in general HMA is continued indefinitely until progression or at least six to nine cycles, which is the standard for patients to be able to achieve response. I think what you are going to see on subsequent slides is the overall outcomes for standard HMA are still quite poor in this patient group and so we overall strongly favor a trial and there are multiple HMA combo trials that I think are quite exciting for this high‐risk group. Now upon progression again this is a very poor risk group despite clinical or molecular features, really these patients should all be considered for investigational agents at that time.

1.0 HR: 0.58 (95% CI: 0.43‐0.77; 0.9 log‐rank P = .0001) 0.8 0.7 0.6 0.5 24.5 mos 0.4 15 mos 0.3 Azacitidine 0.2

Proportion Surviving CCR 0.1 0 0 5 10152025303540 Time From Randomization (Mos)

Fenaux P, et al. Lancet Oncol. 2009;10:223‐232.

Again why is azacitidine at category 1 recommendation? This is one of our most famous curves in high‐risk MDS from the AZA‐001 trial which clearly showed that azacitidine could improve overall survival in comparison to patients that received conventional care regimens with a median overall survival of 24.5 months versus 15 months, and this publication is now about a decade old.

How Do HMAs Perform in Real Life Setting

• A retrospective analysis of 636 HR‐MDS in the MDS Clinical Research Consortium database Median OS from (6 tertiary centers, no single center diagnosis 17.0 accounted for >39%) months (95%CI: • 69.6% INT‐2, 30.4% high IPSS. 15.8, 18.4) • Median follow‐up 15.7 months (95% CI: 14.6, 16.8) • Median time from diagnosis to HMA initiation 0.95 months (95% CI: 0.86, 1.06) • 67.9% azacitidine, 32.1% decitabine • Median number of cycles 5.0 (IQR: 3.0, 8.0) • 72.2% received ≥4 cycles

CR in only 20%‐30% with 50% of patients having any response Zeidan AM, et al. Leukemia. 2016;30(3):649‐657.

But in real life how do these patients truly do, let us say with standard frontline hypomethylating agents? This is some work from the MDS clinical research consortium that we are a part of, again has six centers across the US, and this has been shown by other groups as well where the median survival for these patients has been around 15 to 17 months and not really been able to reach that 24‐month metric from the original publication. This goes along with response rates, CR rate, in general about 20% maybe as high in 30% in some cohorts with only 50% of patients in total having some response whether that be hematologic improvement or mirror responses. So again I think this data supports that we really need to improve on this.

Bejar R, et al. Blood. 2016;124(17):2705‐2712.

Are there any predictors that hypomethylating agents may work better for your patient? I think this is controversial in that there are not well‐validated genes predicting for improved response. One is presence of TET2 mutation in the absence of an ASXL1 mutation was shown by this publication by again Rafael Bejar in Blood that predicted for an improved overall response, although no difference in survival. We have similarly seen this data in particularly patients that do have that ASXL1 mutation can have very low response rate, particularly complete remission rates.

HMA Outcomes Based on Mutations

Sallman D, et al. ASH 2018.

In just looking at outcomes, again this is data that has been shown by multiple groups that this p53 mutant group despite overall similar response rate still does significantly inferior compared to wild‐type patients undergoing frontline HMA, and if you look at the survival again really significantly lower than wild type.

Molecular Targeted Therapy in MDS/AML

PTX‐200

Sallman D, et al. Clin Lymphoma Myeloma Leuk. 2017;17(10):613‐620.

So now switching gears to more novel therapies. How can we approach these patients to improve outcomes? This is a review slide that we have made and you could see that there are number of potential therapies and targets, again this is more from a molecularly targeted angle that are potential options. I think, fortunately some of these, you can see, for example, FLT3 is a very rare mutation in MDS, although of course this has been a significant new treatment option in the AML spectrum, but there are number of therapies and some of these we will talk about on subsequent slides that are showing significant promise.

HMA Failure Studies

• Targeted studies ‒ H3B‐8800 (spliceosome modulator) – 81 patients treated with on‐target effects, good safety, low activity to date (1 HI‐P; EHA 2018) ‒ Ivosidenib (IDH1 inhibitor) – ORR was 92% (11/12) with 5 CRs (42%); P2 underway (ASH 2018) ‒ Enasidenib (IDH2 inhibitor) – ORR was 59% (10/17); P2 +/‐ azacitidine underway (ASH 2016) • Non‐targeted studies ‒ Venetoclax (Bcl‐2 inhibitor) – multiple P1/P2 studies in frontline and R/R MDS, data pending ‒ Nivolumab (PD1 inhibitor) with ORR of 0% (n=15); ipilimumab (CTLA‐4 inhibitor with 22% ORR (n=9); ASH 2016 ‒ Multiple trials with Bi‐specific T cell Engager (BiTE) and cellular therapies including chimeric antigen receptor (CAR) T cell therapies (most common targets including CD33, CD123, and NKG2D ligands)

If we focus briefly on the HMA failure population and then on targeted studies there are I would say the three most interesting ones to date, one has been H3B‐8800, this is a spliceosome modulator with recent data presented at EHA by Dr. Steensma from the Dana‐Farber group and overall showed on‐target effects and safety, but unfortunately really low activity at least as of today with only one patient obtaining a hematologic improvement. Although I think the concept of inducing what we know is synthetic lethality and targeting splices of mutations still needs significant more investigation. The IDH1 and 2 inhibitors of course are either ivosidenib and enasidenib which have approval in relapsed/refractory AML as well as frontline indication with ivosidenib. These mutations do occur in MDS, although at significantly lower levels probably, about 5 to maybe 5% to 10% at the most in total, but responses are very similar to what we see in AML. Again small cohorts but high overall response rate including complete remissions and there are multiple. There are phase 2 studies underway in this group and hopeful that this will be also an option in the future for MDS patients with these mutations.

As far as some non‐targeted studies venetoclax which has brought forth the paradigm shift in elderly AML that is undergoing multiple studies both in relapsed/refractory and frontline MDS. Checkpoint inhibitors such as a PD1 inhibitor and nivolumab and ipilimumab as a CTLA inhibitor had been investigated in MDS, although to date the PD‐1 have had very low response rates if any, although some improved responses have been seen with CTLA‐4 or combination checkpoint studies or combination from the HMA therapy, but really not a clear signal as of today's date. Then I think there is quite exciting newer immune therapies including bi‐specific therapies, CAR T therapies, the most common targets being CD33, CD123, NKG2D ligands, I think these therapies are being investigated, MDS are quite preliminary but do hold promise for these patients.

Ongoing Frontline MDS Studies

Combination AZA or DAC + venetoclax (P1b and P2) AZA + pevonedistat (phase 3) AZA + PDL‐1/CTLA‐4 inhibitors (phase 1/2) AZA + APR‐246 (TP53 mutant, phase 3)* AZA + CD47 inhibitor (phase 1b/expansion)* AZA + ivosidenib (IDH1) or enasidenib (IDH2, P2) CPX‐351 HR‐MDS (HMA naïve, phase 1/1b)*

*Not currently approved.

So if we look now at frontline MDS studies, so the combination of azacitidine or decitabine with venetoclax again which has really changed elderly AML. These studies are ongoing in myelodysplastic syndrome and hopefully data will be available soon as far as response rates. AZA and pevonedistat, which is a NED‐8 inhibitor previously a publication in Blood, has shown higher response rates in myelodysplastic syndrome patients and there is currently a randomized phase 3 study that is underway. Again, the checkpoint inhibitors with multiple combinations with azacitidine are underway. Again, potentially in AML showing some early signal at least as far as combination. On the next couple of slides were are going to focus two. One is on an agent called APR‐ 246 and as we as a CD47 inhibitor called 5F9. We are going to focus on those in great detail. Lastly, AZA and ivosidenib or AZA and enasidenib are also undergoing investigation in MDS. Lastly, CPX‐351 which again has approval in secondary AML which again is AML from a myelodysplastic syndrome, is undergoing early testing again with the goal particularly for transplant eligible patients.

Detection of wild‐type p53 M237I p53 xenograft conformational epitope

p53 R175H

p53 R175H + APR‐246

…restores wt p53 APR‐246 binds …and triggers cell cycle conformation & covalently to p53… arrest and apoptosis activity…y

Aprea unpublished data.; Kaar JL, et al. Prot. Sci. 2010;19:2267‐2278.; Zhang Q, et al. Cell Death Disease. 2018;9:439.; Furukawa H, et al. Cancer Sci. 2018;109(2):412‐421.

If we look at the mechanism of action of APR‐246, this is a drug that is a P53 reactivator. It spontaneously converted to MQ. MQ can then bind this sort of inactive or unfolded protein. This leads to a restoration of wild‐type confirmation, so you can see in the immunohistochemistry on the bottom whereas with the presence of the mutation you really get no binding to DNA whereas with APR‐246 you get this now wild‐type function, and then this can induce normal downstream activity of P53 such as cell cycle arrest in apoptosis.

Study Design

• TP53 mutant (mTP53) HMA‐naïve MDS and AML (≤30% blasts)

And so based on this and some early data showing increased activity with azacitidine, we have run a frontline high‐risk MDS in oligoblastic AML study. This allowed patients up to 30% blasts. The patients got APR‐246 over three dose levels from 50 to 100 mg/kg in body weight and the phase 2 that was equivalent to 4500 mg, and if you see the schedule, patients received APR‐246 over four days. They did get bone marrows pre/post as lead‐in and then subsequently APR‐246 was given over the same schedule. It is an IV six‐hour infusion along with standard seven‐day azacitidine.

AZA + APR‐246 Treatment Duration and Response

Sallman D, et al. ASH 2018.

This slide shows the activity from our data cut‐off at ASH of last year. You can see that our overall response rate was 95% with a true complete remission rate of 70%. Of course, this was somewhat of an early data cut‐off, and again compared to historical this is quite exciting data. This trial is ongoing. We did see one patient that had an objective response even after the lead in phase in the phase 1B portion, and again based on all of these data in detail this has led to a randomized phase 3 study which is underway in the US and will be opening in France for this combination versus azacitidine alone. Again, the goal to move the needle for P53 mutant patients.

© 2019 MediCom Worldwide, Inc. 24 Treating the Patient with High‐risk MDS: Practical Strategies and Emerging Therapeutic Options 5F9 is a Novel Macrophage Immune Checkpoint Inhibitor Targeting CD47 Control mAb: No Phagocytosis SIRP⍺ CD47

“Eat me” signal Anti‐CD47 mAb: Phagocytosis

5F9 Macrophages Cancer cells

• 5F9 enables macrophages to phagocytose cancer cells by blocking the binding of the “don’t eat me” signal CD47 to its receptor SIRPα • Normal cells are not phagocytosed as they do not express “eat me” signals, except for aged red blood cells • 5F9 is a first‐in‐class anti‐CD47 antibody

Another novel therapy is 5F9 which is a macrophage immune checkpoint specifically targeting CD47 or what is known as a “don’t eat me” signal and so cancer cells express this “don’t eat me” signal and so you can see in this video on the top basically the orange cells are macrophages and the green cells are cancer cells, and there is really no phagocytosis. Whereas in the presence of anti‐CD47 monoclonal antibody there is rapid induction of phagocytosis, and this nicely allows for selective elimination of cancer cells in that the vast majority of normal cells do not express “eat me” signals. The one notable exception are old red blood cells which is sort of our body's normal way to clear this and so we do see some early on target anemia associated with 5F9.

5F9005 Study Design: 5F9 Alone or in Combination with Azacitidine in AML and MDS

5F9 Monotherapy Primary objectives Safety Run‐in Cohort (N=10) 1. Safety of 5F9 alone or with AZA Relapsed/ 5F9: 1, 30 mg/kg* 2. Efficacy of 5F9 in R/R AML/MDS and refractory twice weekly 5F9+AZA in untreated AML/MDS (R/R) AML or MDS Secondary objectives 1. PK, PD and immunogenicity of 5F9 Untreated AML 5F9 + AZA Combo 2. Additional measures of efficacy ineligible for Safety Evaluation (N=6) Expansion (N=30) (DOR, PFS, OS) induction 5F9: 1, 30 mg/kg* 5F9: 1, 30 mg/kg* Exploratory objectives chemotherapy or weekly weekly 1. To assess CD47 receptor occupancy, untreated MDS AZA: 75 mg/m2 D1‐7 AZA: 75 mg/m2 D1‐7 markers of immune cell activity, and intermediate to molecular profiling in AML/MDS very high risk by *Dose ramp up from 1 to 30 mg/kg by week 2, IPSS‐R then 30 mg/kg maintenance dosing

• A 5F9 priming dose (1 mg/kg) and dose ramp up was utilized to mitigate on target anemia

• 5F9 monotherapy safety was confirmed in R/R AML/MDS patients prior to 5F9+AZA combination

Sallman D, et al. EHA 2019.

And so this is the design of the trial. I presented initial data at EHA this past June of 2019 where there was a safety lead in looking at single agent, but after that there is combination of standard azacitidine. The 5F9 has a priming dose. This is to mitigate that on‐target anemia and then over the course of two weeks is ramped up to the full dose of 30 mg/kg again with standard azacitidine and present some initial data on this. Preclinically, there is some nice synergistic data of this combination improving response rates and survival in vivo models, and a previous phase 1 presentation showed that at least 5F9 alone was safe in this patient group.

Anti‐Leukemic Activity is Observed with 5F9 Monotherapy and in Combination with AZA in AML and MDS

R/R 1L 1L 5F9+AZA Best Overall AML/MDS AML MDS Response 5F9 mono 5F9+AZA 5F9+AZA N=10 N=14 N=11 ORR 1 (10%) 9 (64%) 11 (100%) CR 0 5 (36%) 6 (55%) CRi 0 2 (14%) ‐ PR 00 0 4 (36%)

MLFS/ in Bone Marrow Blast (%) 1 (10%) 2 (14%) 2 with marrow marrow CR CR+HI Best Relative Change from Baseline Hematologic ‐‐1 (9%) improvement (HI) SD 7 (70%) 5 (36%) 0 Patient PD 2 (20%) 0 0 2 patients not shown due to missing values Response assessments per 2017 AML ELN criteria and 2006 IWG MDS criteria; <5% blasts imputed as 2.5% Patients with at least one post‐treatment response assessment are shown “‐” not applicable

• 5F9 monotherapy has an ORR of 10% in R/R AML/MDS • 5F9+AZA has an ORR of 100% in MDS, 64% in AML which compares favorably to AZA monotherapy ORR median time to response is more rapid (1.9 months) than AZA alone

Sallman D, et al. EHA 2019.

This table and graph shows the early efficacy data for this patients. You can see as a monotherapy there was one patient that had an objective response, but I think quite exciting if we look at both AML and MDS patients, although these cohorts were small at data cut‐off, had significantly high response rate of 64% in AML but actually 100% in MDS patients. This also went along with a 50% or 55% complete remission rate in AML and MDS respectively, with majority of patients having significant hematologic improvement. You can see in the waterfall plots on the right that basically all you know about one patient had significant bone marrow blast reduction and notably these responses are occurring rapidly with a median response time of 1.9 months and again although these data are early, are quite exciting, and at data cut‐off none of the responding patients had progressed. This study is also ongoing.

Acknowledgements Celgene GFM Moffitt Cancer Center Kyle MacBeth Thomas Cluzeau Alan List Sean Yoder Pierre Fenaux Rami Komrokji Jodi Kroeger Eric Padron Sheng Wei Jeffrey Lancet Najla Ali MDS CRC Forty Seven Amy McLemore Jinming Song Amy Dezern Mark Chao Seongseok Yun Kendra Sweet David Steensma Kathy McGraw Michelle Jennings Mikael Sekeres Aprea Lisa Nardelli Yainet Sanchez Gail Roboz Eyal Attar Guillermo Garcia‐Manero

And so at that I just like to thank my collaborators all at Moffitt Cancer Center, some of our key collaborators that have been on the studies that I reported about today as well as the MDS Clinical Research Consortium and then funding from the MDS Clinical Research Consortium, the MDS Foundation as well as the Dresner Foundation which has supported some of these work.