Distinct and overlapping mechanisms of resistance to azacytidine and guadecitabine in acute myeloid leukemia
Emily Gruber 1,2 ● Rheana L. Franich1 ● Jake Shortt1,3,4 ● Ricky W. Johnstone 1,2 ● Lev M. Kats 1,2
Received: 7 April 2020 / Revised: 30 June 2020 / Accepted: 2 July 2020
The Author(s), under exclusive licence to Springer Nature Limited 2020
To the Editor:
Cytosine analog hypomethylating agents (HMAs) constitute one of the few approved life-prolonging therapies for myelodysplasia (MDS) and acute myeloid leukemia (AML) in patients who are ineligible for intensive therapy [1]. HMAs are incorporated into DNA during replication and function to covalently ‘trap’ and degrade DNA methyl- transferase enzymes (DNMTs) leading to global loss of DNA methylation. The resultant re-expression of silenced tumor suppressor genes and interferon signaling triggered by reactivation of endogenous retroviruses combine to produce potent anti-leukemic effects [2–4]. Notwithstand- ing their efficacy, the development of acquired resistance and treatment failure is universal, even following initial complete remission [5]. While some studies have begun to explore the underlying molecular basis of HMA resistance [6–9], unbiased genome-wide approaches have not yet been performed.
Clinically utilized HMAs can be broadly classified into two distinct chemotypes—ribonucleosides and deoxyr- ibonucleosides. Although these molecular subtypes have largely been viewed as equivalent therapeutic modalities, Supplementary information The online version of this article (https:// doi.org/10.1038/s41375-020-0973-z) contains supplementary material, which is available to authorized users.
Lev M. Kats [email protected]
1 The Peter MacCallum Cancer Centre, Melbourne 3000 VIC, Australia
2 The Sir Peter MacCallum Department of Oncology, University of Melbourne, Parkville 3052 VIC, Australia
3 Monash Haematology, Monash Health, Clayton 3168 VIC, Australia
4 School of Clinical Sciences at Monash Health, Monash University, Clayton 3168 VIC, Australia multiple lines of evidence both experimental and clinical, hint at overlapping but distinct modes of action and mechanisms of resistance. Whereas deoxyribonucleosides are incorporated into DNA only, ribonucleosides are incorporated into both RNA and DNA and can exert bio- logical effects via interaction with RNA methyltransferases [10]. A direct head-to-head comparison of the ribonucleo- side azacitidine (AZA) and the deoxyribonucleoside deci- tabine (DAC) in AML cell lines revealed differential effects on gene expression, proliferation and cell death [11]. Moreover, in a phase 2 trial of guadecitabine (GDAC; SGI- 110), a ‘next-generation’ dinucleotide decitabine analog, a proportion of MDS and AML patients that were primary refractory or relapsed after AZA treatment responded to GDAC [12]. This observation may be explained by differ- ential pharmacological characteristics of GDAC, which is less susceptible than first generation hypomethylators to inactivation by cytidine deaminase. GDAC was subse- quently demonstrated to be non-inferior to investigator treatment choice (AZA or DAC in the majority of cases) in treatment-naïve AML patients unfit for intensive che- motherapy induction therapy. Interestingly, GDAC also demonstrated an overall survival benefit over the control arm in subjects receiving at least four or six cycles [13]. These and other observations prompted us to explore and compare AZA and GDAC resistance mechanisms using CRISPR/Cas9 technology.
The safety, tolerability, and efficacy of HMA therapies have been attributed to their anti-proliferative properties, rather than the DNA damage-associated cell killing that occurs at higher drug concentrations [2]. Hence, we first sought to carefully define the cytotoxic and anti- proliferative effects of AZA and GDAC across a panel of AML cell lines using AnnexinV/propidium iodide (PI) staining to assess viability and Cell Trace Violet (CTV) dye dilution to trace cell division. As expected, both drugs reduced cell viability in a time- and dose-dependent manner (Fig. 1a and Supplementary Fig. S1A). After three days of daily treatment, AZA had a significantly greater impact on
Fig. 1 Identification of AZA and GDAC resistance mechanisms in THP1 cells using a genome-wide CRISPR screen. a Cell viability of AML cell lines THP1, MV411 and OCI-AML3 was assessed by Annexin V/PI staining following 7 days of daily drug treatment. Error bars represent mean ± SD from three biological replicates. b, c Pro- liferation of THP1 cells treated with AZA (b) or GDAC (c) was analyzed by CTV labeling. Representative FACS histograms and proportion of cells that have undergone rounds of cell division after 5 days of drug treatment are shown. Error bars represent mean ± SD from two experiments. d Scatter plot showing positive-selection associated p-values (as determined using MAGeCK) from the pooled CRISPR screen. Each point represents an individual gene. AZA, 5- azacytidine; GDAC, guadecitabine (SGI-110); AML, acute myeloid leukemia; PI, propidium iodide; CTV, CellTrace violet.
cell viability than GDAC in all three cell lines tested (Supplementary Fig. S1A). In contrast, after seven days GDAC was significantly more potent, especially in MV411 and OCI-AML3 cells where it reduced cell viability to <20% at 5 μM, compared with >50% for AZA at the same concentration (Fig. 1a). Notably, however, lower HMA concentrations (50–600 nM of AZA and 37.5–300 nM of GDAC) that more closely reflect physiological exposures and are sufficient to deplete DNMT1, decreased prolifera- tion with only negligible induction of cytotoxic cell death (Fig. 1a–c, Supplementary Fig. S1A–C) [2, 11].
To identify genes that mediate resistance to the anti- proliferative effects of AZA and/or GDAC we performed a pooled genome-wide CRISPR/Cas9 knock-out screen in THP1 cells. The screen was performed using low doses of GDAC or AZA (150 nM and 300 nM, respectively) that we posit reflect the epigenetic (rather than direct cytotoxic) activities of these drugs, and we confirmed that continuous long-term treatment caused only minor cytotoxicity (Supplementary Fig. S1D). Importantly, these doses are in the clinically-relevant range of what leukemic cells are exposed to in vivo during standard HMA therapy [14]. We collected cells after initial sgRNA library transduction and then following 45 days of continuous GDAC, AZA, or vehicle treatment. Notably, at day 45 the proliferation of library-transduced THP1 cells was the same in all three conditions, suggesting that GDAC and AZA-resistant clones had been selected (Supplementary Fig. S1D). The frequency of individual sgRNAs was quantified using next- generation sequencing and the MAGeCK algorithm (see Supplementary Methods) was used to rank candidate genes that were enriched in drug-treated conditions com- pared with vehicle at day 45. This analysis identified deoxycytidine kinase (DCK) and uridine cytidine kinase 2 (UCK2) as high confidence hits driving single-agent resis- tance to GDAC or AZA, respectively; and solute carrier family 29 member 1 (Augustine blood group) (SLC29A1) as a high confidence hit for dual GDAC/AZA resistance
Fig. 2 Loss of SLC29A1, DCK and UCK2 mediate distinct resis- tance patterns to AZA, GDAC, and cytarabine. a Competition assays using AML cell lines transduced with individual sgRNAs (2 sgRNAs/gene) and cultured in DMSO, AZA or GDAC. Error bars represent mean ± SD from two experiments. b, c Mice xenografted with sgRNA-transduced luciferase-expressing THP1 cells were treated with AZA, GDAC or vehicle. Bioluminescence imaging was per- formed on day 0 and 14 of treatment. Representative images (b) and signal quantification (c) is shown. Error bars represent mean ± SD, significance is analyzed using non-parametric one-tailed student’s t- test. *p < 0.05, **p < 0.01.
(Fig. 1d and Supplementary Table 1). DCK phosphorylates deoxyribonucleosides, UCK2 is a pyrimidine ribonucleo- side kinase and SLC29A1 (also known as ENT1) is a member of the equilibrative nucleoside transporter family. Examination of normalized sgRNA counts confirmed con- sistent trends for three or more independent guides per gene (Supplementary Fig. S1F). Strikingly, approximately 60% of the GDAC culture was comprised of cells transduced with DCK or SLC29A1 sgRNAs; and ~50% of the AZA culture contained SLC29A1 or UCK2 guides (Supplemen- tary Fig. S1E). DCK, UCK2 or SLC29A1 knockout cells were not enriched by continuous culture in the absence of HMA treatment (Supplementary Fig. S1F), confirming that these genes do not increase the proliferation of THP1 cells. No other resistance candidates were identified in our screen, with all other genes demonstrating a lack of correlation between independent sgRNAs (Supplementary Table 1).
To validate the results of our primary screen we analyzed the impact of individual gene knockouts on drug resistance in an expanded panel of AML cell lines (THP1, MV411, OCI-AML3). Two sgRNAs per gene were cloned into a GFP-expressing sgRNA vector and transduced individually into Cas9 expressing AML cells. The efficiency of indivi- dual sgRNAs was confirmed using western blotting or the Inference of CRISPR Edits (ICE) algorithm (Supplementary Fig. S2A, B). The effect of each sgRNA on cell prolifera- tion in the presence or absence of drug was then determined in a competitive proliferation assay by tracking the per- centage of GFP+ cells over time using FACS (Fig. 2a). In close concordance with data derived from our whole-genome screen, SLC29A1-targeting sgRNAs provided a significant proliferative advantage in the presence of both AZA and GDAC, whereas sgRNAs directed against UCK2 and DCK mediated resistance to AZA and GDAC, respec- tively, but did not impart cross-protection. We further determined the LD50 for each drug and cell line combination and confirmed that SLC29A1/UCK2/DKC loss reduced the sensitivity of AML cell lines to HMAs by 10–100 fold (Supplementary Fig. S2C). Loss of SLC29A1 and DCK (but not UCK2) also decreased the sensitivity of AML cells to cytarabine, a non-HMA nucleoside analog that is commonly used in the clinic as salvage post AZA or DAC failure (Supplementary Fig. S2D).
We next tested the in vivo significance of our findings. To generate a tractable in vivo AML model, we transduced Cas9/sgRNA-expressing THP1 cells with a retroviral con- struct encoding firefly luciferase and puromycin resistance gene. We transplanted the puromycin selected THP1Cas9/ sgSCR/luc cells into NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) immune-compromised mice and confirmed that the cells engraft and generate a disseminated disease in recipient animals, which is rapidly lethal (data not shown). We subsequently transplanted puromycin selected THP1Cas9/ sgRNA/luc cells into NSG recipients and monitored disease development by bioluminescence imaging. Notably, SLC29A1, UCK2 and DCK knockout cells were able to proliferate at rates that were comparable to control cells, confirming that these genes are dispensable for growth in vivo as well as in vitro (Fig. 2b). Once disease was established, animals with comparable bioluminescence signal were assigned to receive AZA, GDAC or vehicle treatment. In recipients transplanted with control cells, both AZA and GDAC delayed disease progression (Fig. 2b, c). In contrast, THP1 cells transduced with the SLC29A1 sgRNA were resistant to both AZA and GDAC treatment in vivo, whereas cells transduced with the DCK2 guide were resistant to GDAC but sensitive to AZA, and cells transduced with the UCK2 guide were partially resistant to AZA, but sensitive to GDAC (Fig. 2b, c).
After cellular uptake via nucleoside transporters, HMAs must be phosphorylated prior to being incorporated into nucleic acids. DAC analogs including GDAC, which have a deoxyribose core, can then be incorporated directly into DNA during replication, whereas the ribose moiety of AZA must first be reduced by ribonucleotide reductase. Many of the mechanisms of resistance to nucleoside analogs are related to altered drug metabolism and indeed earlier studies have implicated proteins in the pyrimidine metabolism pathway in HMA resistance in vitro [6, 7]. Herein, we used unbiased genome-wide CRISPR/Cas9 technology and identified SLC29A1, DCK, and UCK2 as non-redundant rate-limiting steps in HMA transport and activation. SLC29A1 and DCK deletion conveyed high-level cross resistance to cytarabine, the conventional chemotherapeutic ‘backbone’ of most con- ventional AML treatment regimens. The lack of additional hits in our pooled screen strongly suggests that loss of these genes provides cells with a significant growth advantage in the presence of HMAs over other loss-of-function genetic events. It should be noted however, that other mechanisms that produce lower levels of HMA resistance in vitro, may nonetheless be clinically relevant. Importantly, SLC29A1, DCK and UCK2 are not required for proliferation of AML cells in vivo. Notably, Gu et al. recently reported loss of UCK2 and DCK expression in MDS patients that initially respond but subsequently relapse on AZA and DAC therapy, respectively [15]. We further analyzed gene expression data from a recent study of MDS and chronic myelomonocytic leukemia (CMML) patients [9] and found similar trends of downregulated expression of UCK2 and SLC29A1 in patients refractory to AZA treatment compared with those that respond (Supplementary Fig. S2E). While further correlative studies on larger patient cohorts are required, our findings, and those of Saunthararajah and colleagues [15] suggest that alternating HMA chemotypes, or indeed HMA combination strategies may be warranted to predict or circumvent clinical resistance to nucleoside analog therapy (see Supplementary Table 2 for further references). We propose that patients most likely to benefit from such strategies could be identified by assessing DNA methylation levels (or surrogate markers of DNA damage in the case of cytarabine treatment) in malig- nant cells at baseline, following response and at progression.
Acknowledgements This work was supported by a grant in aid from the Cancer Council of Victoria to JS and LMK. LMK was supported by a fellowship from the Victorian Cancer Agency, JS was supported by a fellowship from the Australian Medical Research Future Fund, RWJ was supported by a fellowship from the National Health and Medical Research Council of Australia and EG was supported by the Australian Postgraduate Award.
Compliance with ethical standards
Conflict of interest The laboratory of JS has received research funding from Astex Pharmaceuticals. JS, LMK and RWJ served on the sci- entific advisory board of Celgene Corp.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
1. Quintás-Cardama A, Santos FPS, Garcia-Manero G. Therapy with azanucleosides for myelodysplastic syndromes. Oncogene. 2010;7:433–44.
2. Tsai H-C, Li H, Van Neste L, Cai Y, Robert C, Rassool FV, et al. Transient low doses of DNA-demethylating agents exert durable antitumor effects on hematological and epithelial tumor cells. Cancer Cell. 2012;21:430–46.
3. Chiappinelli KB, Strissel PL, Desrichard A, Li H, Henke C, Akman B, et al. Inhibiting DNA methylation causes an interferon response in cancer via dsRNA including endogenous retroviruses. Cell. 2015;162:974–86.
4. Roulois D, Loo Yau H, Singhania R, Wang Y, Danesh A, Shen SY, et al. DNA-demethylating agents target colorectal cancer cells by inducing viral mimicry by endogenous transcripts. Cell. 2015;162:961–73.
5. Fenaux P, Mufti GJ, Hellstrom-Lindberg E, Santini V, Finelli C, Giagounidis A, et al. Efficacy of azacitidine compared with that of conventional care regimens in the treatment of higher-risk mye- lodysplastic syndromes: a randomised, open-label, phase III study. Lancet Oncol. 2009;10:223–32.
6. Qin T, Jelinek J, Si J, Shu J, Issa J-PJ. Mechanisms of resistance to 5-aza-2’-deoxycytidine in human cancer cell lines. Blood. 2009;113:659–67.
7. Sripayap P, Nagai T, Uesawa M, Kobayashi H, Tsukahara T, Ohmine K, et al. Mechanisms of resistance to azacitidine in human leukemia cell lines. Exp Hematol. 2014;42:294–306.e2.
8. Valencia A, Masala E, Rossi A, Martino A, Sanna A, Buchi F, et al. Expression of nucleoside-metabolizing enzymes in myelo- dysplastic syndromes and modulation of response to azacitidine. Leukemia. 2014;28:621–8.
9. Unnikrishnan A, Papaemmanuil E, Beck D, Deshpande NP, Verma A, Kumari A, et al. Integrative genomics identifies the molecular basis of resistance to azacitidine therapy in myelodys- plastic syndromes. Cell Rep. 2017;20:572–85.
10. Cheng JX, Chen L, Li Y, Cloe A, Yue M, Wei J, et al. RNA cytosine methylation and methyltransferases mediate chromatin organization and 5-azacytidine response and resistance in leu- kaemia. Nat Commun. 2018;2018:1–16.
11. Hollenbach PW, Nguyen AN, Brady H, Williams M, Ning Y, Richard N, et al. a comparison of azacitidine and decitabine activities in acute myeloid leukemia cell lines. PLoS ONE. 2010;5:e9001.
12. Sebert M, Renneville A, Bally C, Peterlin P, Beyne-Rauzy O, Legros L, et al. A phase II study of guadecitabine in higher-risk myelodysplastic syndrome and low blast count acute myeloid leukemia after azacitidine failure. Haematologica. 2019;104:1565–71.
13. Roboz GJ, Döhner H, Gobbi M, Kropf PL, Mayer J, Krauter J, et al. Results from a global randomized phase 3 study of guade- citabine (G) vs treatment choice (TC) in 815 patients with treat- ment naïve (TN) AML unfit for intensive chemotherapy (IC) ASTRAL-1 study: analysis by Number of cycles. Blood. 2019;134:2591–1.
14. Issa JPJ, Roboz G, Rizzieri D, Jabbour E, Stock W, O’Connell C, et al. Safety and tolerability of guadecitabine (SGI-110) in patients with myelodysplastic syndrome and acute myeloid leukaemia: a multicentre, randomised, dose-escalation phase 1 study. Lancet Oncol. 2015;16:1099–110.
15. Gu X, Tohme R, Tomlinson B, Hasipek M, Durkin L, Schuerger C, et al. Resistance to decitabine and 5-azacytidine emerges from adaptive responses of the pyrimidine metabolism network. bioRxiv 2020.02.20.958405; https://doi.org/10.1101/2020.02. 20.958405.