Deletion of the RNA-editing enzyme ADAR1 causes regression of established chronic myelogenous leukemia in mice

Authors


  • R.A.S. wrote the manuscript and participated in project design and planning, data generation and analysis; Q.Y. and X.L. conducted mouse transplantation and phenotyping; M.G. conducted flow cytometry studies and analyzed data; J.H. participated in design and analysis; L.J.R. analyzed histology; D.E.L. conducted statistical analysis; C.S. contributed to data analysis; Q.W. conceived project, contributed to data and writing, obtained funding and coordinated the work.

Abstract

Patients with chronic myelogenous leukemia (CML) respond well to tyrosine kinase inhibitors (TKIs) of the Bcr-Abl oncoprotein. However, intolerance and resistance to these agents remains a challenge, and TKIs are unable to eradicate rare leukemia-initiating cells. Leukemia treatment would benefit from a better understanding of molecular signals that are necessary for the survival of leukemia-initiating cells but dispensable for normal hematopoietic stem cells. Leukemia-initiating cells in CML can arise from myeloid progenitor cells, a population that we have reported in normal hematopoiesis to depend on the RNA-editing enzyme adenosine deaminase acting on RNA-1 (ADAR1). We now report that Bcr-Abl transformed leukemic cells were ADAR1-dependent in a conditional ADAR1 knockout mouse model. ADAR1 deletion reversed leukocytosis and splenomegaly, and preferentially depleted primitive Lin-Sca+Kit+ (LSK) leukemic cells but not LSK cells lacking the leukemic oncoprotein. ADAR1 deletion ultimately normalized the peripheral white blood count, eliminating leukemic cells as assessed by PCR. These results uncover a novel requirement for ADAR1 in myeloid leukemic cells and indicate that ADAR1 may comprise a new molecular target for CML-directed therapeutics.

Patients with chronic myelogenous leukemia (CML) respond well to tyrosine kinase inhibitors (TKIs) of the Bcr-Abl oncoprotein, such as imatinib.1 However, patients may not tolerate TKI treatment and approximately 17% of patients become resistant to imatinib because of imatinib-resistant stem cells, Bcr-Abl mutation, or secondary mutations that enable early progenitor cells to self-renew and initiate tumors.2–4 Optimal treatment of CML should kill the CML progenitor cell compartment that can acquire leukemia-initiating capabilities (review5) as well as CML (but not normal) stem cells. We recently reported that adenosine deaminase acting on RNA 1 (ADAR1), a known RNA-editing enzyme, selectively eliminated normal hematopoietic progenitor cells (HPC) but not normal hematopoietic stem cells.6 The dependence of leukemic progenitor- and stem cell-compartments on ADAR1 is unknown. Recently, ADAR1 was reported to be elevated in pediatric acute lymphocytic leukemia and was decreased with clinical response, implying a link between ADAR1 and human leukemogenesis.7

ADAR1 post-transcriptionally modifies RNA (including micro-RNAs) by catalyzing the conversion of adenosine to inosine. The consequence of A to I RNA-editing is to uncouple RNA sequences from that of genomic DNA, such that proteins arising from edited RNAs could differ from those encoded by the genome.8,9 RNA structures, stabilities, and splicing patterns may also be modified by RNA-editing.10,11 ADAR1 has also been shown to edit noncoding sequences12 including viral13 and mammalian microRNAs.14,15 Hyperediting of noncoding RNAs has recently been reported to suppress interferon signaling.16 Despite global analysis indicating thousands of edited RNA sites, only a handful of ADAR-edited targets have been validated.17–19 Loss of ADAR1 is embryonic lethal, with mice dying at 11 to 12 days in conjunction with massive hepatocyte death and defective hematopoiesis in the fetal liver.20–22 Postnatal deletion of ADAR1 in normal hematopoietic cells selectively depleted hematopoietic progenitor cells compared with more primitive cells.6 We now report our use of a conditional ADAR1 knockout mouse model to determine whether Bcr-Abl transformed leukemic cells were ADAR1-dependent. Our data indicate a novel requirement by the leukemic cells for ADAR1.

Material and Methods

Mice

Donor mice used were of mixed background, arising from a cross between SV129 mice bearing floxed ADAR1 mice (prepared by homologous recombination as described previously22) and CreER transgenic mice23 (Jax stock #004453, STOCK Tg(cre/Esr1)5Amc/J) bearing a mixed B6, SV129 and Swiss Webster background. The recipient mice were NOD-SCID IL2Rg (Jax Stock 005557).

Mice of the bone marrow donors (ADAR1 lox/lox and CreER) were bred in Hillman Cancer Center according to an IACUC approved breeding protocol, and genotyping was done as described previously.22

Bone marrow transduction/transplantation

The retroviral vector MSCV-IRES-eGFP carrying the p210 BCR-ABL cDNA was a gift from Dr. Shaoguan Li, University of Massachusetts Medical School. This Bcr-Abl vector has been used extensively to transduce mouse bone marrow cells that generate a CML phenotype when transplanted.24,25 Although we did not pretreat donor mice with 5-FU before marrow harvest and transduction as in Li et al.,25 we did select early stem/progenitor cells (see below) as the transduction target. The engraftment was determined through the congenetic markers of CD45, (Donor cells were CD45.2, and host cells were CD45.1). For each transplantation, bone marrow cells were collected from two ADAR1 f/f and Cre-ER-positive mice or control wild-type mice at 6 to 8 weeks old. Roughly 2 × 108 total cells were labeled with 100 µl microbeads (CD117/c-kit, Miltenyi Biotec Inc, #130-091-224) and enriched for CD117/c-kit positive primitive cells using MidiMACS LS columns. Up to 1 × 107c-kit-expressing stem/progenitor cells were eluted for retrovirus transduction following prestimulation with the cytokines mSCF, Flt-3 L, and TPO. Two rounds of retrovirus transduction were conducted using the centrifugation method.6 NOD-SCID-IL2Rgnull mice were chosen as recipients because of the lack of a syngeneic match for the ADAR1 f/f and Cre-ER mice. Recipient mice were irradiated (3.5 Gy, 82 cGy/min, 137Cs radiator) 6 hr preceding the transplantation; 1.5 × 105 transduced cells were injected into the tail veins of recipients.

PCR analysis

Genomic analysis of wild-type, floxed and Cre-deleted ADAR1 loci was conducted using sense PCR primer P1 (5′ CGGGATCCCCAAGGTGGAGAATGGTGAGTGGTA) and a mixture of antisense primers P2 (5′ GCTCTAGAAGAGGGCACAGCCACAGCAGGAC, located between loxP sites) and P3 (5′ GCTCTAGAGAATCAAACCCACAAGAGGCCAGTG, located in the 3′UTR). The P1-P2 fragment was preferentially amplified over the P1-P3 fragment in the wild-type and in the floxed genome because of the long stretch of genomic DNA upstream of P3 (see Fig. 1a). Lox sequences render the P1-P2 product slightly larger in the floxed compared with wild-type genome. In the deleted genome (lacking P2 and with P3-P1 in closer approximation) the P1-P3 fragment was amplified. Other PCR primers amplified genes as follows: Bcr-Abl: S-GTCTCCACCCAGGAAGGACT; AS-TGAATTGGAAAGAAGCAGCA, P21: S-AAGCCTTGATTCTGATGTGGGC; AS-TCACCGTCCTGTTTACCCCAG, GFP: S-CACAAGTTCAGCGTGTCCGGCGAGG; AS-GACGTTGTGGTGTTGTAGTTGTACTC. Oligomers for RT-PCR amplification of actin were: S-CTTCTACAATGAGCTGCGTGTG; AS-TACAGGATAGCACAGCCTGGAT.

Figure 1.

Inducible ADAR1 gene deletion mouse model. (a). Strategy for generation of mice with leukemia-specific conditional ADAR1 expression. The conditional knockout mouse derived from the breeding of floxed ADAR1 mouse and Cre-estrogen receptor fusion protein transgenic mice. Marrow from these mice were pooled, transduced with Bcr-Abl, and transplanted into CD45 congenic NOD Scid recipients as illustrated. (b). MSCV vector used to transduce p210 Bcr-Abl into stem/progenitor (c-kit+) cells from Adar f/f Cre-ER mice. (c). Upon tamoxifen induction, floxed ADAR1 alleles undergo recombination and gene deletion. PCR with specific primers can identify floxed and deleted ADAR1 alleles. Positions of the primers used and association with exons and introns of ADAR1 gene are indicated. Dotted arrowheads indicate loxP sites. Primers P1 and P2 amplify floxed ADAR1, and P1 and P3 generate the band indicative of ADAR1 deletion. (d). DNA and RNA samples were prepared from host and donor spleen and bone marrow or from mouse that had developed leukemia roughly 3 weeks after transplantation. Both floxed (donor, top) and wild-type (host, bottom) ADAR1 alleles are evident in transplant recipients. Untreated spleen pooled three samples. p21 served as loading control. (e). After transplantation, a myeloproliferative syndrome developed quickly in the immunodeficient recipients. The white blood cell number in the peripheral blood increased as shown (×109/L). Values represent the average and SD of n = 12, 22, 21, and 4 mice, respectively, for the four time points. **p < 0.01 compared with week 1. (f). Example represents predominance of donor-derived CD45.2 peripheral leukocytes in host blood 2 weeks after transplant. Cells positive for the myeloid marker CD11b predominated in most recipients.

RNA isolation and RT-PCR

RNA was isolated from spleen and bone marrow cells thawed from liquid nitrogen. Trizol reagent (Invitrogen, Inc.) was used for total RNA isolation according to the manufacturer's instructions. Reverse transcription of 1.0 µg total RNA was carried out per manufacturer's instructions for the cDNA synthesis kit (BioRad, Inc).

Flow cytometry analysis

Mononuclear cells from blood, spleen, and bone marrow of NOD-SCID-IL2Rg recipient mice transplanted with c-kit-enriched cells from ADAR1 f/f/Cre-ER crossed mice, which were transduced with retrovirus expressing Bcr-Abl onco-protein, were stained with various combinations of CD11b (clone M1/70), Ter119, Ly-6G/6C (RB6-8C5), B220 (RA3-6B2), CD3e (145-2C11), CD117 (2B8) and Sca-1 (E13-161.7), as well as with CD45.1 (104) and CD45.2 (A20) to distinguish recipient and donor cells, respectively. All antibodies and staining reagents were from BD Biosciences (San Jose, CA). Flow cytometric analysis was performed on a CyAn® device equipped with 405nm violet, 488 nm argon and 633 HeNe lasers (Beckman Coulter; Brea, CA). Data were analyzed using FlowJo® Software (TreeStar; Palo Alto, CA).

Pathological analysis

Tissues from spleen and bones were taken from the control- and tamoxifen-treated leukemic mice at indicated time points after administration of these agents. The samples were fixed with formalin, and the bones of femur and vertebra were subjected to de-calcification before sectioning. H&E stained sections of paraffin-embedded tissue were used for pathologic analysis.

Tamoxifen induction of Cre activity

Tamoxifen (Sigma T5648, Sigma Aldrich, St. Louis, MO) was dissolved in corn oil and orally fed to the mice at a final concentration of 25 mg/ml. Mice were treated with 5 mg/mouse/day for 3 days consecutively in short-term experiments. Corn oil vehicle alone was used in controls. In the long-term treatment, single injections of 2.5 mg/mouse/week were administered for 2 months.

Statistical Analysis

Two-sided unpaired t-tests were used to compare changes in WBC among tamoxifen-treated and vehicle-treated mice. This test was also used to compare vehicle and TM-exposed cells in the GFP− and GFP+ cells within the peripheral blood, marrow, spleen, and in the LSK populations, and the weight of spleens exposed to TM or to vehicle control.

Results

It has not been known if ADAR1 plays a vital role in leukemogenesis. We have noted expression of ADAR1 in patient samples of CML blasts, in some cases at levels higher than in nonleukemic CD34+ progenitor cells (Supporting Information Fig. S1). Along with the report that ADAR1 decreased in treatment of childhood B-ALL,7 the presence of ADAR in CML cells, this supports investigation of a potential function of ADAR1 in CML cells.

In order optimally to explore a requirement for ADAR1 in CML, we generated a novel mouse model in which ADAR1 was deleted specifically in Bcr-Abl-transduced donor cells following transplant and after tamoxifen induction. First, the floxed ADAR1 mice (ADAR1 f/f) were crossed with Cre-inducible transgenic mice bearing a Cre DNA recombinase/estrogen receptor fusion protein (Cre-ER mice). Progeny carrying both ADAR1 f/f and Cre-ER genes were selected. Floxed ADAR1 alleles were intact and functional in these mice until tamoxifen was administered to activate Cre-ER, resulting in ADAR1 deletion. The bone marrow progenitor/stem cells from these ADAR1 f/f/Cre-ER mice were then transduced with retrovirus expressing p210 Bcr-Abl onco-protein, followed by transplantation into NOD-SCID-IL2Rg immunodeficient recipients (Fig. 1a). Because GFP was encoded in cis with the Bcr-Abl gene, donor leukemic cells could be monitored in NOD-SCID-IL2Rg recipients by flow cytometry for green fluorescence (Fig. 1b). Donor cells also expressed the congenic marker CD45.2, enabling them to be distinguished from CD45.1 recipient cells. In addition, the ADAR1 gene locus was divergent between the recipient- and donor-derived leukemia cells, enabling us to monitor the presence of donor-derived leukemic cells in recipient tissues by detection of the floxed ADAR1 locus. Tamoxifen-directed rearrangement of the floxed ADAR1 locus is depicted in Figure 1c.

Rapid leukemic cell death after tamoxifen-induced loss of ADAR1

In five independent transplantation experiments, a total of 34 immunodeficient mice (NOD-SCID-Il2Rg) were successfully engrafted with ADAR1 f/f/Cre-ER or wild-type bone marrow hematopoietic stem/progenitor cells that had been pooled from 6- to 8-week-old mice and transduced with the Bcr-Abl retrovirus. Between 2 and 3 weeks after transplantation, mice engrafted with the Bcr-Abl-transduced marrow cells manifested a phenotype compatible with CML, consisting of a peripheral blood leukocytosis with a left shift including circulating myeloid precursors, marrow hypercellularity with myeloid predominance, and splenomegaly associated with increased myeloid cells and megakaryocytes. This phenotype and time course was consistent with prior reports of leukemic marrow transplantation with the vector we used.25 Figure 1d demonstrates discrete expression of floxed versus normal Adar1 DNA in donor and host cells and also confirms Bcr-Abl and GFP DNA in oncogene-transformed donor cells. Bcr-Abl transcripts were detectable in spleen and marrow of mice harboring leukemia. Peripheral white blood cell counts were elevated and continued to increase over time up to 4 weeks post-transplant (Fig. 1e). Flow cytometry analysis (Fig. 1f) showed that donor-derived (CD45.2) cells engrafted the host, generally with a myeloid phenotype and predominated in peripheral blood.

To delete ADAR1 in the leukemic cells, mice received a single cycle of oral tamoxifen (5.0 mg/day/mouse for 3 consecutive days) beginning 2 to 3 weeks after transplantation. Tamoxifen-induced excision of ADAR1 in the leukemic cells could be monitored by PCR because of the unique size of the deleted allele.

PCR testing of peripheral blood for tamoxifen-mediated deletion of ADAR1 confirmed that the Cre-ER-mediated deletion of the ADAR1 alleles was specifically detected after tamoxifen addition but was not found in leukemic cells not exposed to tamoxifen (Fig. 2a). Because of the rapid death of leukemic cells after TM, this population may be underrepresented in the PCR pool leading to a disproportionately weaker “deleted” band in this figure. Retained floxed bands after 2 weeks of treatment arise from nondeleted cells that escaped death during this period.

Figure 2.

Rapid loss of leukemic cells after ADAR1 deletion. Leukemia mice were treated with either tamoxifen or corn oil vehicle three weeks after transplantation once leukemia was established. (a). Demonstration of specific ADAR1 gene deletion in tamoxifen-treated mice. PCR bands associated with host, donor-floxed, and donor-deleted ADAR1 are indicated. Analysis is at 2 weeks post-tamoxifen treatment. A relapse in WBC subsequently occurred in a subset of mice (relapse) and was associated with outgrowth of nondeleted donor cells; 100-bp marker on left denotes sizes of 300 to 500. (b). Peripheral white blood count (×109/L) in mice beginning 3 weeks post-transplant. The increase in WBC in vehicle-treated (red) versus tamoxifen-treated (black) mice is shown. The effect of adding tamoxifen to mice transplanted with Bcr-Abl-transduced normal mouse marrow (e.g. without floxed ADAR1) is shown as a green line representing average and SD of three mice tested. (c). TM-induced ADAR1 deletion significantly lowered the burden of leukemic but not of nonleukemic donor cells (n = 5 vehicle-treated and 6 TM-treated mice). Representative flow cytometric profiles are shown above the graph.

Peripheral blood counts were determined periodically after treatment. A marked decline in white blood cell numbers was observed in all of the tamoxifen-treated ADAR f/f-CE (Bcr-Abl) mice within 1 week after treatment and remained low over the following week (shown in black, Fig. 2b), whereas white blood cell counts in vehicle-treated controls increased (shown in red, Fig. 2b). In wild-type mice (with wild-type, nonfloxed ADAR1) that were transplanted with Bcr-Abl-transduced wild-type hematopoietic stem/progenitors, tamoxifen treatment did not reverse Bcr-Abl-induced leukocytosis (Fig. 2b, green line). These results indicated that the antileukemic effect after tamoxifen exposure specifically resulted from ADAR1 excision. However, one cycle of tamoxifen treatment did not eliminate ADAR1 from all of the leukemic cells, as indicated by the persistent floxed ADAR1 band in Figure 2a. This finding indicated that repetitive dosing with tamoxifen might be necessary for a durable anti-leukemic effect (see below).

Because a subset of cells in donor marrow were not transduced with Bcr-Abl (e.g. were GFP-negative), it was possible to compare the effect of tamoxifen-mediated ADAR1 deletion between leukemic (GFP+) and nonleukemic (GFP−) cells. As shown in Figure 2c, loss of ADAR1 had a more significant effect on leukemic than nonleukemic cells.

The decrease in peripheral blood cell counts was paralleled by a significant reversal of splenomegaly (Fig. 3a) and a decrease in total splenocytes (Fig. 3b). Spleen weights for tamoxifen-treated mice approximated those of nontransplanted, nonleukemic wild-type mice (Fig. 3c). On histologic examination, spleens from tamoxifen-treated Bcr-Abl transduced mice showed the normal predominance of small lymphocytes (Fig. 3d) with varying degrees of fibrosis seen in some of the treated spleens (Supporting Information Fig. S2). In contrast, the large spleens from untreated Bcr-Abl-positive mice showed extensive replacement of splenic lymphoid cells by hematopoietic elements, with increased megakaryocytes and a predominant population of myeloid precursors and mature granulocytes, findings characteristic of involvement by chronic myelogenous leukemia (Fig. 3d). Notably, deletion of ADAR1 significantly reduced the Bcr-Abl transformed subset of the donor population infiltrating the spleen while minimally affecting the nontransformed (GFP-negative) subset (Fig. 3e) over the 2-week treatment period.

Figure 3.

ADAR1 deletion reduces the leukemic cell burden in the spleen. (a). Gross appearance of the spleen in vehicle and in tamoxifen-treated mice harvested at 2 weeks post treatment. (b). Cellularity per spleen in untreated (n = 6) and tamoxifen-treated (n = 8) mice at 2 weeks. (c). Spleen weight in Bcr-Abl transplanted mice treated with vehicle or with tamoxifen to delete ADAR1 as compared to the spleen weight of untreated normal mice. Spleens are harvested at 2 weeks post-treatment. N = 6, 8, and 3 for vehicle control, tamoxifen-treated and wild-type mice, respectively. p = 0.003 for vehicle versus tamoxifen treatment. (d). H&E stained sections of spleens from Bcr-Abl transduced mice, control (left) or tamoxifen treated to stimulate ADAR1 deletion (right); fields are each 210 microns wide. The controls show extensive extramedullary hematopoiesis with myeloid predominance and increased megakaryocytes consistent with splenic involvement by chronic myelogenous leukemia; these features are absent from the treated spleens which show a predominant population of small lymphocytes. (e). Decrease in the Bcr-Abl+ (GFP+) but not in the nonleukemic GFP- subsets of donor cells infiltrating spleen. Representative flow plots on left, n = 6 (untreated), 8 (tamoxifen-treated) for graph on right.

ADAR1 dependence of the LSK leukemia cell subset

In Bcr-Abl induced mouse leukemia, cells bearing the phenotype Lin-Sca1+Kit+ (LSK) are enriched for leukemic stem cells.26–29 Durable antileukemic effects are most likely if LSK early progenitor/stem cells are targeted rather than more mature blast and progenitor cells alone. Relapse is generally ascribed to the inability to eliminate leukemic stem cells. To determine whether ADAR1 deletion induced death of LSK cells, mature progenitors (and their progeny) or both, we compared the bone marrow LSK cell populations of vehicle-treated and tamoxifen-treated mice after ADAR1 f/f Cre-ER leukemic cell transplantation. LSK subsets were measured in six untreated and in six tamoxifen-treated mice (Supporting Information Fig. S3) and leukemic (GFP-positive) and nonleukemic donor cells were compared. Figure 4a shows an example of the LSK content in these subsets as manifested by the percentage of Lin − cells that were positive for Sca-1 and c-kit. As summarized in Figure 4b, tamoxifen-mediated ADAR1 deletion resulted in a marked decrease in leukemic, GFP-positive LSK cells (p = 0.004), whereas there was no change in the LSK profile in the nonleukemic, GFP-negative donor cell population. Similarly, in the marrow as a whole, the proportion of GFP-positive cells decreased significantly after tamoxifen treatment (p = 0.03, Fig. 4d) whereas nonleukemic CD45.2+ GFP-negative donor cells did not change significantly. We conclude that both immature (LSK) and total leukemic cells were more vulnerable to death from ADAR1 deletion than their ADAR1-floxed counterparts that were not transformed with Bcr-Abl.

Figure 4.

ADAR1 dependence of primitive leukemic cells. (a). Representative flow plots demonstrating a decrease in stem cell-enriched (LSK) subset of donor leukemic (GFP+) but not nonleukemic (GFP-) cells in marrow 2 weeks after treatment. Lineage negative cells were isolated and then analyzed for Sca+Kit+ subset. (b). Graphical summary (n = 6 per group, average and SD). (c). ADAR1-dependence of leukemic cells in bone marrow. Mice (2–3 weeks post-transplant) were treated with tamoxifen (tamoxifen) or vehicle control and marrow was harvested 2 weeks later for assessment of viable Bcr-Abl transduced (GFP+) and nontransduced (GFP−) donor cells in marrow. Flow plots (above) are representative of results graphed in (d) (average and SD, n = 6 for each group).

Long-term suppression of leukemia in mice

Although our data supported a requirement for ADAR1 in LSK cells, flow cytometry and PCR analysis conducted on WBC 2 weeks after tamoxifen demonstrated that the single 3-day treatment of leukemic mice with tamoxifen did not completely eradicate the leukemia cells (Figs. 2a and 2b), presumably because of incomplete Cre-ER-mediated deletion. We were interested in establishing whether bolstering Cre activity would lead to more durable suppression of leukemic cells. In order to determine whether eradication of ADAR1 in the leukemic cell pool resulted in long-term remission, we changed the tamoxifen treatment protocol to one of sustained, lower-dose tamoxifen treatment. Leukemic mice were given tamoxifen at 2.5 mg/week/mouse for 2 months. This dose was first shown to be effective in deleting ADAR1 from leukemic cells within 48 hr in vitro. Tamoxifen treatment began at 2 to 3 weeks post-transplant (when peripheral leukocytosis was evident). Within a week of tamoxifen addition, the peripheral WBC dropped significantly and remained at a near-normal level throughout the ensuing 2 months. In contrast, the WBC increased in vehicle-treated mice (Fig. 5a, p = 0.034, tamoxifen vs. nontamoxifen). The survival of vehicle-treated mice out to 8 weeks is longer than usual with p230 Bcr-Abl models, and could reflect lower oncogene transduction efficiency of the donor pool used to seed vehicle and tamoxifen-treated transplants, or could arise because of the mixed strain background of the donor mice providing the transduced marrow.

Figure 5.

Targeting of ADAR1 achieves long-term remission of CML. (a). Peripheral WBC in mice receiving Bcr-Abl transfected marrow and treated with continuous low-dose tamoxifen or vehicle over 2 months as described in text. Dotted lines denote WBC from mice receiving corn oil vehicle (N = 4); solid black lines indicate WBC from mice receiving low-dose weekly tamoxifen (N = 3). Treatment began 2 to 3 weeks post-transplant. (b). PCR analysis of peripheral blood for floxed and deleted ADAR1 in animals treated for 2 months with vehicle or tamoxifen. By 2 months, absence of floxed or deleted ADAR1 leukemic donor cells and complete restoration of the host, endogenous ADAR1 PCR pattern was evident. (c). Genetic analysis of blood, spleen and bone marrow harvested from untreated mice or mice treated for 8 weeks with Tamoxifen, as indicated. Normal host organs were harvested from an untransplanted SCID mouse. DNA from untreated mice was processed from blood (8 weeks) or organs (2 weeks) after leukemia was established. (d). Lack of detectable Bcr-Abl in RT-PCR of RNA from indicated organs of treated mice.

No splenomegaly was detected in mice after chronic tamoxifen treatment. Examination after 3 months of treatment revealed splenic histology comparable to that of untreated wild-type controls (Supporting Information Fig. S4).

We sampled blood and tissue out to 8 weeks for PCR to see whether genomic and expression data supported leukemic cell loss as manifested in flow measurements and the leukemic cell burden. Continued treatment with low-dose TM decreased the level of floxed- and deleted-ADAR1 bands that were detectable by PCR, ultimately to barely detectable levels (Fig. 5b and “Blood” lanes, Fig. 5c). Residual floxed ADAR1 DNA was detectable in spleen and bone marrow after 2 months of treatment (Fig. 5c), indicative of Cre-resistant donor cells. However, substantial depletion of leukemic donor cells was supported by the reduction of GFP and Bcr-Abl signals to near-undetectable levels. These findings attest to a dramatic decrease in the CML leukemic burden, consistent with leukemic cell death following successful ADAR1 deletion.

Discussion

The above data demonstrate for the first time that the RNA-editing enzyme ADAR1 is required for leukemia cell survival in vivo and that targeting ADAR1 results in rapid leukemic cell loss. By transplanting Bcr-Abl marrow cells in which ADAR1 could subsequently be conditionally knocked out, it was possible to monitor the effect of ADAR1 deletion after leukemia had been established. Genetic models have been used to establish several other candidate genes involved in CML pathogenesis, including arachidonate 5-lipoxygenase (ALOX5),30 Smo,26,29 Myb,31 PTEN,32 and beta-catenin.33 In these instances, however, the collaborating gene was knocked out before Bcr-Abl transduction and leukemic cell transplantation so that genetic modulation could not specifically target established disease. One study has demonstrated regression of myeloid leukemia after Stat5 excision.34 There is no evidence for editing of Stat5 message, and it remains to be established whether ADAR1 and Stat5 collaborate in maintaining leukemia. Our ability to delete ADAR1 after the appearance of myeloproliferative CML-like disease in mice uncovered the rapid kinetics of leukemic cell clearance after ADAR1 loss.

Maintenance of the leukemic state in CML requires a variety of epigenetic and post-transcriptional adaptations, including methylation, translational control by miRNAs and by RNA binding proteins, miRNA recruitment as ds-RNA decoys, and modulation of mRNA turnover rates.35–39 A role for the RNA-binding and editing enzyme ADAR1 could broaden the repertoire of post-transcriptional regulators in CML biology.

RNA-editing enzymes have not hitherto been shown to be required for a malignant state, although highly-expressed ADAR1 observed in childhood B-ALL reportedly decreased during remission.7 Upregulation of the ADAR2 gene has been observed in transformed cells40 whereas downregulation of ADARs was reported in brain cancers.41 ADAR1 could act at many levels including alteration of micro-RNAs or their target sites; modulation of interferon or of PKR signaling; alteration of protein sequence or of DNA regulatory sequence and functions.20,42–45 Hyperedited transcripts suppress the interferon response.20 We have not been able to capture upregulation of interferon targets in samples from tamoxifen-treated mice, but detection may be compromised by the rapidity of cell death after Cre-mediated deletion of ADAR1 (data not shown).

Despite elegant delineation of RNA editing sites in silico, the experimentally validated targets of ADAR1 are few, and whether they comprise editing targets in the rare population of CML stem/progenitor cells is currently unknown (e.g. ADAR1 editing targets FlnA, Blcap, and Cyfip2 have been identified but in neuronal cultures). It would be ideal to validate the specific editing targets in the CML stem/progenitor cells that revert to genomically-encoded sequences after ADAR1 deletion. Quantitating an increase in the percentage of nonedited transcripts in the ADAR1 deleted cell population in the animal model is infeasible; however, given the challenge of separating leukemic cells sustaining floxed from deleted ADAR1 and ongoing death of the ADAR1 deleted cells.

The removal of exons 12 to 15 of ADAR1 through Cre-deletion removes the catalytic domain of ADAR1. Loss of RNA-editing (which has been the predominant activity associated with ADAR1) could fully account for the death of Cre-deleted leukemic cells. However, we cannot rule out other mechanisms through which sustained ADAR1 is required for cell survival. In particular, deletion of exons 12 to 15 does not result in the formation of a truncated protein.21 Gain of function through truncated ADAR1 is therefore unlikely to account for cell death. But loss of other ADAR1 domains distinct from the catalytic domain could be contributory. For instance, one recent report indicated that the Z-DNA/RNA-binding region of ADAR1 bound to ribosomal RNA and could affect translation.46 We have initiated structure–function studies of ADAR1 to dissect the linkage of ADAR1 and survival (Sharma et al., manuscript in preparation) as well as the generation of a mouse mutated in the ADAR1 RNA-editing domain for use in crosses that will show whether CML survival requires ADAR1 RNA editing. Elaboration of the mechanism of CML cell dependence on ADAR1 will inform the development of therapeutics that exploits this ADAR1 requirement. Of note, we cannot rule out the possibility that the ADAR1 locus spanning exons 12-15 also contains uncharacterized cryptic transcripts whose function is necessary for cell survival.

Clinical recurrence of CML after imatinib treatment arises in part from the inability of tyrosine kinase inhibitors to eliminate the primitive CML stem cell compartment.47 Dual Src/Bcr-Abl inhibitors, while more potent against early progenitors, do not eliminate the most primitive, leukemia-initiating cells,48 indicating that curative treatment may require drug targets beyond kinases. Moreover, a recent report has demonstrated Bcr-Abl kinase-independence of CML leukemia-initiating cells.49 We have found that expression of ADAR1 is independent of Bcr-Abl kinase activity (data not shown). Therapeutic strategies targeting ADAR1 may therefore eliminate leukemic cells that survive treatments directed at the Bcr-Abl kinase. It will be of interest to determine whether ADAR1 deletion is additive or synergetic with tyrosine kinase inhibitors in suppressing leukemia and whether a non-TKI-sensitive leukemic subset is ADAR1-dependent in secondary transplant experiments.

It is important to establish genetic pathways that are required by CML but not by normal stem cells. The Lin-Sca+Kit+ (LSK) compartment of CML has been shown to contain the leukemia-initiating capability of CML and has been shown to depend on genes that are dispensable for normal LSK function.30 Our data indicate that primitive leukemic cells with the LSK phenotype are more sensitive to the loss of ADAR1 than primitive normal stem/progenitor LSK cells. This is consistent with our prior report of an ADAR1 requirement in the wild-type mature (lineage-positive) but not early (LSK) stem progenitor cells.6 Hartner et al. have also have reported stable total LSK after ADAR1 deletion in normal cells, but found that the cycling stem/progenitor subset was depleted after ADAR1 excision.20 It is possible that CML LSK cells are preferentially targeted by ADAR1 loss because they have increased cycling kinetics relative to normal LSK. We observed that while ADAR1 deletion lowered the leukemic cell burden in the peripheral blood, spleen, marrow and the LSK marrow compartment, this effect was most notable in the leukemic LSK population (e.g. comparison of Figs. 2c and 4b). The sensitivity of these cells to ADAR1 loss could account for our finding that ongoing Cre-mediated ADAR1 deletion reversed and then suppressed CML for 2 months. However, the LSK compartment is heterogeneous and includes cells with both short-term and long-term reconstitution capability. It is not yet clear whether both short-term and long-term LSK cells are ADAR1 dependent and future studies will be required to unequivocally link ADAR1 deletion to the elimination of all leukemia-initiating cells. It will be useful to compare the ability of discrete leukemic LSK subsets46 to cause leukemia after secondary transplantation with or without ADAR1 deletion concurrent with transplant.

The rapid clearance of CML cells from marrow and peripheral blood after ADAR1 deletion indicated that more mature CML cells in addition to LSK cells required ADAR1. The ability of ADAR1 inhibition to suppress both early and more mature leukemic cells could comprise a distinct advantage over treatments directed at only one of these compartments. A caveat is that even though leukemic cells are selectively eliminated by ADAR1 deletion, lineage-positive normal cells are also sensitive to ADAR1 loss.6 Development of ADAR1 inhibition for use ex vivo or as treatment will need to take into account the risk to normal progenitors.

Acknowledgements

The authors thank Pei Zhou for assistance with PCR and RT-PCR analyses.

Ancillary