Diminished AHR Signaling Drives Human Acute Myeloid Leukemia Stem Cell Maintenance

Acute myeloid leukemia (AML) is initiated by driver gene mutations in hematopoietic stem or progenitor cells (HSPC) that give rise to leukemic stem cells (LSC), characterized by extensive self-renewal activity and a limited differentiation capacity that enables their production of large numbers of immature myeloid cells (1–3). Because LSCs drive disease initiation and relapse (4), effective long-lasting antileukemic treatments are dependent on promoting LSC death or terminal differentiation. Meeting this critical goal has been challenging due to the limited understanding of the molecular pathways that underpin the human AML-LSC self-renewal program.

Ng and colleagues, have shown that LSC-containing populations (LSC⁺) carry unique gene expression changes and that LSC⁺ gene signatures predict therapy resistance and lower overall patient survival (5–7). Importantly, the gene expression profile that defines LSC⁺ fractions within patients are not bound by genetic subtype or AML classification (6), and share similarities to the expression profile of normal hematopoietic stem cells (HSC; ref. 5). These studies suggest that there are fundamental molecular mechanisms characteristic of HSCs that may be exploited to drive the active self-renewal and impaired differentiation of LSCs. The challenge, given the reliance of LSCs on mechanisms also utilized for normal HSC function, is to define selective means to disrupt these pathways preferentially in LSCs, so that HSCs are spared.

Here we have explored the AHR pathway as a potential regulatory axis that human LSCs may converge on to promote their oncogenic function. AHR is a ligand-activated transcription factor that is well-studied for its role in signaling that contributes to cellular metabolism of environmental toxins and more recently, is being recognized for its importance in endogenous cell functions. In mouse AHR knockouts, disparate genetic backgrounds have contributed to differential phenotypes and a lack of clarity around AHR's exact function (8). Nonetheless, the elevation of HSPC proliferation in some of these mouse models echoes studies in the human hematopoietic context where small-molecule AHR antagonists were found to promote HSPC expansion ex vivo (9). The same phenomenon of enhanced stem cell maintenance in culture via AHR antagonism was later shown to occur for LSCs from select patient samples (10). AHR activation has also been explored in THP-1 acute monocytic leukemia cells and in coordination with high-dose retinoic acid (RA) in the acute myeloblastic leukemia with maturation cell line context where it has offered some enhancements to in vitro leukemic differentiation and proliferation inhibition (11–14).

Despite the mounting studies indicating intrinsic roles for the AHR pathway in HSC self-renewal/differentiation regulation in normal hematopoiesis, no study has examined the necessary function of AHR signaling in endogenous primary human AML cell behavior, nor in the in vivo regulation of the most primitive LSCs that drive this disease. Thus, the potential of AHR signaling as a prognostic or intrinsic therapeutic target in human AML remains unexplored. In this study, we have explored the hypothesis that AHR repression constitutes a critical mechanism through which LSCs ensure their self-renewal. We identify the native suppression of AHR signaling in leukemia and LSCs in particular. Enforcing AHR pathway activation by administration of the AHR agonist FICZ across xenografted diverse patient AMLs resulted in irreversible antileukemic effects. Importantly, we find that LSCs, but not their normal HSC counterparts, are uniquely responsive to enforced FICZ-mediated AHR signaling. Altogether, our work implicates the AHR pathway as a core prodifferentiation signaling module that human LSCs suppress in vivo to promote their self-renewal and shows that activating this pathway may provide a novel therapeutic strategy for targeting LSCs.

NOD-scid-IL2Rγc^−/− (NSG; Jackson Laboratory) mice were bred and maintained in the Stem Cell Unit animal barrier facility at McMaster University (Hamilton, Ontario, Canada). All procedures received the approval of the Animal Research Ethics Board at McMaster University (Hamilton, Ontario, Canada).

All cord blood (CB) and AML patient samples were obtained with written informed consent and with the approval of the local human subject research ethics board at the University Health Network and McMaster University (Hamilton, Ontario, Canada) in accordance with Canadian Tri-Council Policy Statement on the Ethical Conduct for Research Involving Humans (TCPS). Following Ficoll-Paque separation, mononuclear cells were stored in the vapor phase of liquid nitrogen in 10% DMSO, 40% FCS, and α-MEM. Primary samples were thawed in PBS 10% FBS with 100 μg/mL DNAse (AML only) prior to using in in vitro and in vivo assays.

Primary AML samples were grown in DMEM with 15% FBS, β-mercaptoethanol (50 μmol/L), stem cell factor (SCF; 100 ng/mL; R&D Systems), IL3 (10 ng/mL; R&D Systems), IL6 (20 ng/mL; PeproTech), thrombopoietin (TPO, 10 ng/mL; PeproTech) and FMS like tyrosine kinase 3 ligand (FLT3L, 10 ng/mL; R&D Systems). AML cell lines HL60 (ATCC) and MV-4;11 (ATCC) were grown in Iscove Modified Dulbecco Medium (IMDM) with 10% FBS, and NB4 in RPMI1640 with 10% FBS. Human CB samples were cultured in StemSpan SFEM (StemCell Technologies) with IL6 (20 ng/mL), SCF (100 ng/mL), TPO (20 ng/mL), and FLT3 (100 ng/mL). MOLM-13, OCI-AML3, and Kasumi-1 cell lines (ATCC) were all grown in RPMI + 20% FBS. Cell cultures were not authenticated, however, were thawed and used at low passages for experiments from stocks obtained from ATCC. MOLM-13, OCI-AML3, and Kasumi-1 cell lines were tested for Mycoplasma. All flow cytometry analysis was performed using a BD LSRII flow cytometer (BD Biosciences) and FlowJo Software (v7.6.5). Cell sorting was performed with MoFlo XDP (Beckman Coulter).

AML cell lines were seeded at a density of 1–3 × 10⁵ cells/mL in medium supplemented with 0.1% DMSO vehicle or 6-formylindolo[3,2-b]carbazole (FICZ; R&D Systems) dissolved in DMSO. HL-60 cells were cultured in DMSO or FICZ (200–1,000 nmol/L) and counted at 3, 5, 7, and 10 days. At day 10, DMSO/FICZ was removed and cells were replated with fresh medium. MV4;11 (FAB M5) cells were cultured in medium supplemented with 1, 3, or 5 μmol/L FICZ or 0.1% DMSO control and grown for a total of 9 days. At day 4, cells were counted and replated at 1 × 10⁵ cells/mL in freshly supplemented medium. MOLM-13 (FAB M5a) and OCI-AML3 (FAB M4, NPM1 mutated) cells were cultured in medium supplemented with 500 nmol/L or 2.5 μmol/L FICZ or 0.1% DMSO control and grown for a total of 7 days. At day 3, cells were counted and replated at 1 × 10⁵ cells/mL. Kasumi-1 (FAB M2, t8;21) cells were cultured in 2.5 μmol/L or 5 μmol/L FICZ or 0.1% DMSO control and grown for 7 days. At day 3, cells were counted and replated at 3 × 10⁵ cells/mL. NB4 cells treated with FICZ were similarly cultured as MV4;11 cells but final cell counts were performed after 5 days. For RA experiments, NB4 cells were plated at a density of 1 × 10⁵ cells/mL in medium varying in doses of all-trans RA (125–1,000 nmol/L; Sigma-Aldrich) and grown for 5 days to generate a dose–response curve. Alternatively, NB4 cells were plated at a density of 1 × 10⁵ cells/mL in medium supplemented with (i) 0.1% DMSO and 100–125 nmol/L RA or 3 μmol/L FICZ and 100–125 nmol/L RA for 3, 5 and 7 days; and (ii) 1,000 nmol/L RA for 48 hours and then 0.1% DMSO or 3 μmol/L FICZ for 3, 5, and 7 days. At days 3 and 5 cells were counted and replated at 1 × 10⁵ cells/mL.

Total cellular RNA was isolated with TRIzol LS reagent (Invitrogen) according to the manufacturer's instructions and cDNA was synthesized using qScript cDNA Synthesis Kit (Quanta Biosciences). qRT-PCR was done in triplicate with PerfeCTa qPCR SuperMix Low ROX (Quanta Biosciences) with gene-specific probes (Universal Probe Library, UPL, Roche) and primers (Supplementary Table S1). The mRNA content of samples compared by qRT-PCR was normalized on the basis of the amplification of GAPDH.

Immunocytochemistry was performed on primary AML samples that were transduced with shMSI2 or shControl lentivirus. Live transduced cells were purified by FACS and then fixed, stained and imaged for CYP1B1 protein levels as described by Rentas and colleagues (15).

Whole-cell lysates were prepared by lysing cells in RIPA buffer (50 mmol/L NaCl, 1% NP-40, 0.5% DOC, 0.1% SDS, 50 mmol/L Tris pH 8.0, 1 mmol/L EDTA) with Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific) and quantified using the Bradford assay (Bio-Rad). Protein samples were normalized to 1 μg/μL in 1× NuPAGE LDS sample buffer (Thermo Fisher Scientific) with β-mercaptoethanol (Sigma-Aldrich) and boiled for 5 minutes at 95°C prior to electrophoresis. Protein was transferred onto polyvinylidene difluoride membrane (LI-COR Biosciences), blocked with 5% BSA in 1× TBS-T for 30 minutes at room temperature and then incubated overnight at 4°C with primary antibodies rabbit-anti-MSI2 (EP1305Y, Abcam), rabbit-anti-CYP1B1 (EPR14972, Abcam), rabbit-anti-α-tubulin (11H10, Cell Signaling Technology), or mouse-anti-β-actin (AC-74, Sigma). Following membrane washing, secondary antibodies IRDye 680 goat-anti-rabbit (LI-COR Biosciences) and IRDye 800 goat-anti-mouse (LI-COR Biosciences) were added for 1 hour at room temperature and then imaged with the Odyssey Classic Imager (LI-COR Biosciences).

Cell death was measured with flow cytometry using Annexin V 350 or Annexin V 647 (BD Biosciences) and 7-AAD (BD Biosciences) in Annexin V binding buffer (BioLegend). For cell-cycle analysis, HL60 cells were cultured with and without FICZ for 10 days then treated with 10 μmol/L BrdU (BD Biosciences) for 3 hours at 37°C, washed, fixed, and stained following BD Pharmingen BrdU Flow Kit manufacturer recommendations (BD Biosciences).

Thawed primary AML samples were counted and plated in a methylcellulose-based hematopoietic colony formation medium (Colony Gel, ReachBio), supplemented with DMSO or 750 nmol/L FICZ or 750 nmol/L SR1. Colonies were scored on days 10–14. For secondary CFU assays, the cells were recovered from the methylcellulose by diluting and washing with PBS and then plated in fresh complete medium without DMSO or FICZ supplementation. Secondary colonies were scored on days 10–14. Human CB samples were plated as described with a density of 1 × 10³ cells per 35-mm plate. AML-CFU assays were also performed on flow-sorted GFP⁺ shMSI2 or shControl GFP⁺ cells. In all cases, cell suspensions were plated in duplicate and loose colonies consisting of 10 or more cells were counted.

Human primary AML cells and AML cell lines grown in suspension were collected and cytospun with Cytospin 3 (Thermo Shandon) at 800 RPM for 5 minutes onto glass slides. After drying, slides were stained for morphology using the Kwik-Diff Stains (Thermo Fisher Scientific) and washed with water. Slides were then mounted with Histomount Mounting Solution (Thermo Fisher Scientific) and imaged with the ScanScope (Aperio) at ×20 magnification.

Production, knockdown, and validation of shMSI2- and shControl-expressing lentiviral particles were performed as described previously (15). Prior to primary AML transduction with shMSI2 or shControl lentivirus, AML samples were flow sorted for CD34 marker expression (581; BD Biosciences) to enrich for LSCs. AML cells were transduced at an MOI of 50 for 24 hours in X-VIVO (Lonza) with 20% BIT (StemCell Technologies) and 2 mmol/L l-glutamine (Invitrogen) supplemented with growth factors IL6 (20 ng/mL), SCF (100 ng/mL), FLT3 (100 ng/mL), and TPO (20 ng/mL). Posttransduction, cells were validated for GFP expression (monitored sample of cells 3 days after transduction), washed, and then transplanted intrafemorally at 50,000–200,000 cells (cell dose matched for each experiment between shMSI2 and shControl) per sublethally irradiated (315 cGy) NSG mouse at 6–8 weeks of age. Three months posttransplant, mice were sacrificed and bone marrow from tibias, femurs, and pelvis was harvested, crushed with a mortar and pestle, filtered, and red blood cell lysed using ammonium chloride buffer (StemCell Technologies). Human AML engraftment was analyzed by blocking reconstituted mouse bone marrow with mouse Fc block (BD Biosciences) and human IgG (Sigma), followed by staining with fluorochrome-conjugated antibodies against human CD45 (HI30) and CD33 (P67.6, BD Biosciences) and LIVE/DEAD Fixable Violet (Invitrogen) or 7-AAD (BD Biosciences).

Prior to NSG mouse xenotransplantation, primary AML samples were treated in vitro overnight in IMDM supplemented with 15% BIT, growth factors IL3 (20 ng/mL), SCF (100 ng/mL), FLT3 (50 ng/mL), and granulocyte colony-stimulating factor (20 ng/mL) and DMSO or 750 nmol/L FICZ in the case of pretreatment or media alone when no pretreatment was done. Sublethally irradiated mice were given a sample-dependent dose of cells, ranging from 1 to 5 × 10⁶ cells, via intrafemoral injection. Cohorts of mice were treated in vivo with one of the three treatment regimens where in all cases vehicle was 0.5% DMSO and 30% captisol (ligand) in PBS and FICZ was dissolved in vehicle. In one approach, 7 days posttransplant, mice were given intraperitoneal injection of 1 mL of vehicle or 100 or 250 μg/kg of FICZ. Intraperitoneal injections were performed three times a week on nonconsecutive days over 4 weeks. In alternate approaches at 7 days or 4 weeks posttransplant, mice were given intraperitoneal injections of vehicle or 250 μg/kg FICZ for 5 consecutive days for 4 weeks. On the day following the final injection, mice were sacrificed and right femur, bone marrow, and spleen were collected for analysis of human leukemic engraftment as described above.

Primary AML cells were initially stained with anti-CD34 PE (581, BD Biosciences) antibody and LIVE/DEAD Fixable Violet (Invitrogen) and then fixed with the Cytofix/Cytoperm kit (BD Biosciences) according to the manufacturer's instructions. Fixed and permeabilized cells were immunostained with anti-MSI2 rabbit monoclonal IgG antibody (EP1305Y, Abcam) and detected by Alexa-488 goat anti-rabbit IgG antibody (Invitrogen) as described by Rentas and colleagues (15).

Microarray Innovations in LEukemia (MILE; ref. 16) normalized expression data was obtained from Gene Expression Omnibus database with accession number GSE13159 and normalized expression data from LSC+/− populations (6) was obtained using accession number GSE76009. Differential gene expression analysis was performed using R Studio version 1.0.44, Bioconductor, GEO2R, and limma package. For MILE dataset, comparisons between individual AML subtypes [normal karyotype, complex karyotype, t(11q23)/MLL, inv(16)/t(16;16), t(15;17), t(8;21), MDS, CML] were made relative to nonleukemia healthy bone marrow controls. The t-statistic was used for both MILE and LSC datasets to filter for individual probe IDs. A list of significantly downregulated genes from MSI2 OE versus control transduced CD34⁺ cord blood HSPCs was obtained from Rentas and colleagues (15).

iRegulon (17) was used to retrieve the top 107 AHR-predicted targets as described previously (15). The data were analyzed using gene set enrichment analysis (GSEA; ref. 18) with ranked data as input with parameters set to 1,000 gene-set permutations.

A previously described list of genes (15) in close proximity to AHR-bound regions identified from TCDD-treated MCF7 chromatin immunoprecipitation sequencing (ChIP-seq) data was combined with the iRegulon identified AHR target gene list to generate a complete list of 1,887 AHR target genes. Genes showing >1.5-fold downregulation or upregulation and FDR < 0.05 from MILE and LSC datasets were compared with the AHR target gene list. The total number of AHR-bound regions was then graphed for each of the gene sets up- and downregulated. P values were generated with Fisher exact test between gene lists.

Heatmaps displaying log₂FC for GSEA leading edge genes and AHR target gene lists were generated using R Studio with gplots (heatmap.2) and RColorBrewer packages. The UpSetR (19) package was used to plot gene intersections between lists of significantly downregulated (FDR < 0.05 and >1.5-fold) AHR target genes from MILE and LSC datasets.

Survival data was obtained from The Cancer Genome Atlas (TCGA) for AML (AML TCGA) and z-score normalized to normal skin tissue from the same patients (20, 21). This represented a total of N = 173 RNAseq samples from all patients in the study with complete data. Patients were subsetted on the basis of having lower AHR and ZFP36L1 expression levels in a patient's blasts as compared with their comparative skin cells (AHR^low/ZFP36L1^low) and the Kaplan–Meier method was used to generate survival curves from this data, where the P value was determined by a right-censored log-rank test as initially formulated by Mantel and Cox as described previously (22, 23).

Expression data for ZFP36L1 was obtained from Bloodspot (www.bloodspot.eu) and evaluated in the normal hematopoiesis with AML dataset, using 211965_at probes. Populations analyzed were as follows: AML, acute myeloid leukemia; HSC, hematopoietic stem cell; MPP, multipotent progenitor; CMP, common myeloid progenitor; GMP, granulocyte monocyte progenitor; MEP, megakaryocyte-erythroid progenitor; PMN, polymorphonuclear cells; Mono, monocyte.

All statistical analysis for in vitro and in vivo studies was performed using GraphPad Prism Software (v5.0). Unpaired Student t tests or Mann–Whitney tests were performed with P < 0.05 as the cutoff for statistical significance. All data are presented as mean ± SEM.

In exploring differential gene expression that may contribute to LSC function, we examined published transcriptional profiling from the validated LSC⁺ and LSC-devoid (LSC⁻) fractions of 78 patient samples (6). Interestingly, within this dataset, we observed that MSI2, a regulator of asymmetric division and numb translation in mouse AML (24, 25) and which we have also shown mediates repression of AHR signaling in human HSCs to promote their expansion (15), is transcriptionally elevated 1.6-fold in LSC⁺ versus LSC⁻ samples (Supplementary Fig. S1A; ref. 6). We also find that MSI2 knockdown in primary AML LSC⁺ cells from 4 primary samples decreases leukemia reconstitution 2.5-fold (Supplementary Fig. S1B and S1C) consistent with MSI2 being required for LSCs. As MSI2 can function to repress the AHR pathway, these findings highlight the potential for dysregulated AHR signaling to underlie elements of AML pathophysiology. Indeed upon MSI2 knockdown in NB4 and THP-1 AML cell lines and primary samples, the levels of CYP1B1, a canonical downstream effector of AHR signaling and indicator of pathway activity, were significantly increased (Fig. 1A–C). These results provide evidence of AHR signaling derepression upon MSI2 knockdown across diverse AML samples and provide the impetus to explore further the potential role for AHR signaling dysregulation in AML and LSCs.

Examining evidence for AHR signaling competency in AML we find that AHR mRNA levels were detectable and unchanged between patient LSC⁺ and LSC⁻ populations (Fig. 1D). Despite this, CYP1B1 was significantly downregulated in LSC⁺ fractions (Fig. 1E), indicating the likelihood of differential AHR pathway signaling between self-renewing and non-self-renewing populations of AML. To expand our analysis and test for broad AHR pathway attenuation in LSC⁺ fractions, we performed GSEA with a list of predicted and validated AHR targets derived by iRegulon (17). This analysis found a significant negative enrichment score, indicating AHR pathway suppression in LSC⁺ populations versus LSC⁻ (Fig. 1F). We extended the investigation of AHR signaling status beyond intratumor dynamics to bulk leukemia and healthy bone marrow by examining the MILE Study Group global gene expression profiling dataset (18). GSEA revealed that most AML subtypes exhibited varying degrees of negative AHR target gene enrichment while complex karyotype AML and MDS had positive enrichment scores (Supplementary Fig. S2A). The leading edge genes driving a significant negative enrichment score in LSC⁺ populations clustered more closely with t(8;21) and t(15;17) subtypes and overall represent genes downregulated in all AML subtypes compared with healthy bone marrow (BM; Supplementary Fig. S2B). Expanding the list of iRegulon AHR target genes to contain AHR ChIP-seq identified targets (26; “AHR combined-targets”) shows significant overlap with downregulated genes (1.5-fold, FDR < 0.05) within leukemia (LSC+ versus LSC−) and across multiple leukemia subtypes relative to healthy bone marrow (Fisher test, P < 0.05), whereas overlap with upregulated genes is not significant (Fig. 1G; Supplementary Tables S2 and S3). This demonstrates the downregulation of AHR targets is enriched in the blasts of most AML subtypes and within LSCs.

The list of significantly downregulated genes that are AHR combined targets in LSCs or each AML subtype (relative to normal BM) were next compared with ascertain the degree of overlap. Of the 40 significantly downregulated AHR target genes in LSCs, we found 21 were uniquely repressed in LSCs (Supplementary Fig. S3), indicating that their downregulation does not reach the level of significant repression in blasts as it does in LSCs. To test the validity of the 21 LSC-AHR target genes as actively downregulated AHR pathway genes, we examined their expression in CB HSPCs overexpressing MSI2 (15). This analysis showed that the profile of their expression in LSC clustered most closely with that upon MSI2 OE and less so with the various AML subtype blast populations potentially due to MSI2 OE driving a higher degree of repression of these genes through AHR antagonism (Fig. 1H). The LSC–AHR gene signature encompasses the downregulation of differentiation-promoting or known/implicated tumor suppressors in AML such as ZFP36L1 (27), CDKN1A, and TLE1 (Supplementary Table S3; ref. 28). Altogether, our analysis reveals AHR pathway attenuation is captured in the unfractionated blasts of multiple AML subtypes relative to healthy BM, and that within AML, LSC⁺ populations further enrich for a subset of downregulated LSC-AHR target genes representing potential players in maintaining the LSC self-renewal gene expression program.

To explore whether AHR pathway activation via exposure to an AHR agonist would provide antileukemic effects we cultured the AML cell line HL-60 in the presence of the well-validated endogenous high-affinity AHR ligand FICZ (29). FICZ-activated signaling of the AHR pathway as measured by upregulation of CYP1B1 transcript occurred in a dose-dependent manner (200–750 nmol/L) and showed potency as an antileukemic small molecule, yielding significant reductions in cell proliferation (1.2- to 1.8-fold; Fig. 2A and B). FICZ treatment did not significantly alter cell-cycle dynamics (Supplementary Fig. S4A) but promoted significant, dose-dependent increases in apoptosis and led to enhanced CD11b expression, indicative of its capacity to promote differentiation (Fig. 2C and D). Moreover, the detrimental effects induced by FICZ were irreversible as upon removal of FICZ from culture medium, HL-60 cells remain impaired up to 10-fold in their capacity to proliferate (Fig. 2E).

Interestingly, NB4 cells of the acute promyelocytic subtype and containing the t(15;17) translocation, appeared resistant to FICZ even at doses as high as 5 μmol/L (Supplementary Fig. S4B). Prolonged exposure to high RA concentrations overcomes the commitment block imposed by PML-RARA to induce growth arrest and differentiation (30). As such, we tested FICZ treatment in NB4 cells in combination with a low dose (125 nmol/L) of RA that has no effects on proliferation (Supplementary Fig. S4C) to gauge potential synergies. In contrast to RA alone, NB4 cells cultured in 125 nmol/L RA with 3 μmol/L FICZ indeed underwent reduced proliferation over 7 days and displayed highly differentiated cell morphology compared with DMSO (Supplementary Fig. S4D and S4E). Moreover, a 48-hour pulse of high-dose RA (1 μmol/L) followed by FICZ treatment led to significant reductions in cell proliferation, differentiated cell morphology, and increased Annexin V⁺ staining relative to controls (Supplementary Fig. S4F–S4H), indicating NB4 cells given RA transiently are unable to resume growth in the presence of FICZ. Thus, even for NB4 cells that are highly reliant on impaired RARA signaling to maintain impaired leukemic differentiation, FICZ can significantly sensitize the capacity of RA to relieve this block.

To test the effects of FICZ on a broader spectrum of AML subtypes, we applied it to the MV4;11, MOLM-13, OCI-AML3, and Kasumi-1 cell lines. FICZ treatment at higher doses (1, 3 μmol/L) was first tested on MV4;11 cells and reduced proliferation by 2.4-fold (Fig. 2F). We then used FICZ doses identified as non- or moderately inhibitory to cell death in the cases of MOLM-13, OCI-AML3, and Kasumi-1 cells. At these doses, in every case, we observed a dose-responsive impairment in either proliferation (Fig. 2G; Supplementary Fig. S4) and/or enhancement in myeloid differentiation (Fig. 2H and I), confirming that across a broad array of genetic/morphologic subtypes of AML cell lines, FICZ is also very effective at impairing leukemic cells and promoting commitment.

We next examined the effect FICZ has on a group of genetically diverse primary patient AML samples (Supplementary Table S4). To first validate that primary samples activate AHR in response to FICZ, we tested CYP1B1 levels after 12–48 hours of exposure and found them elevated in the large majority of samples (Fig. 3A and B). In all four samples also tested for the expression of CYP1A1, another well-established effector of AHR pathway activation, we observed a significant FICZ-induced elevation that in every case was higher in magnitude than for CYP1B1. Across our analysis, 3 of 12 AMLs (AML-10, AML-1, AML-11) showed negligible effects on AHR activation; however, as only CYP1B1 was measured for these samples, we cannot rule out that AHR activation as measured by CYP1A1 actually occurred. AML samples undergo rapid cell death in culture, challenging interpretations of proliferation and apoptosis measurements. However, we did note that of three samples tested for differentiation changes, two showed a significant elevation in CD14 (Fig. 3C). To provide a conclusive window into the effects of FICZ on primitive AML cells in vitro, we assayed its effectiveness on leukemic progenitors by testing their capacity to generate colonies in the presence of FICZ. We noted a significant, up to 2.7-fold, reduction in primary AML colony output in 5 of 6 samples tested with FICZ stimulation, indicating a striking targeting of AML progenitors in most samples (Fig. 3D). AML-7, the only sample for which the number of progenitors did not respond to FICZ, showed 3.4-fold reduction in CFU potential upon secondary replating, an effect similarly observed when AML-3 colonies were replated (Fig. 3E). It is important to note that for all 6 of the samples responding with negative impacts to progenitors, AHR activation at the level of CYP expression was conclusively verified (Fig. 3A, dark gray bars). In addition, we tested two AML samples with the AHR antagonist SR1 and show only slight increases in colony formation in comparison with DMSO (31). This demonstrates that under the CFU assay conditions used, there is a negligible activation of AHR signaling, leaving ample capacity for additional FICZ-mediated stimulation of the pathway (Fig. 3F). Altogether, we find that FICZ yields potent antileukemic effects across diverse sets of leukemia cell lines and patient samples, and in the instance of APL can synergize with the standard-of-care therapeutic to enhance antileukemic phenotypes.

FICZ has been tested in mice to treat numerous disease models where it has been given over a wide dose range with virtually no toxicity demonstrated (32). To define a FICZ treatment that may be effective at impairing in vivo leukemic growth and the early phases of reconstitution where self-renewal is relied on most heavily, we first incubated four primary AML samples overnight with 750 nmol/L FICZ or vehicle prior to transplantation into NSG mice. After allowing one week for recovery, mice were injected with three doses of 100 μg/kg FICZ or vehicle/week over 4 weeks (Fig. 4A). In FICZ-treated mice, we observed a decreased leukemic engraftment from one of the samples tested (AML-7) and a trend toward reduced disease burden from another (AML-4; Fig. 4B; Supplementary Fig. S5A). The effective dose of antileukemic agents tested in the xenograft setting can be dependent on the level of leukemic burden in the animal, the potential of which we tested next for FICZ. First, we reduced the input cell dose for AML-4 to establish a lower level of engraftment and again carried out the in vivo injections. In this case, we found that leukemic cells in FICZ-treated mice were now substantially reduced in comparison with vehicle-treated animals (Fig. 4B, AML-4 exp. 2) demonstrating that a high leukemic load could alter sensitivity to FICZ at the 100 μg/kg dose. Importantly, here too, the level of CD34⁺ cells was also preferentially reduced within the remaining FICZ-treated grafts as compared with control, indicating a detrimental impact on phenotypically primitive AML cells (Fig. 4C). To assess targeting of functional LSCs, we performed secondary transplantations of AML-4 grafts and found that engraftment derived from FICZ-treated primary BM was reduced by 7.5-fold in comparison with control secondary grafts (Fig. 4D). These findings are suggestive of FICZ effectiveness at the primitive cell level when dose:AML cell number ratios are optimized. To this point, we next tested whether an elevated FICZ dose could exhibit a broader range of effectiveness. By administering 250 μg/kg FICZ in vivo three times/week over 4 weeks, we found across three AML samples a more consistent impairment in leukemogenic engraftment, with the most substantial decreases ranging from 2.5- to 10-fold (Fig. 4E).

Finally, we initiated experiments that involved in vivo treatment of established leukemic grafts (Fig. 4F). We first tested the effects of beginning treatment early posttransplant (7 days) and followed this with injections of 250 μg/kg FICZ 3 times/week for 1 month. Engraftment of sample AML-12 treated in this manner was reduced by approximately 3.5-fold in the FICZ cohort mirroring similar levels of leukemic burden reduction in the above pretreatment experiments (Fig. 4G). As FICZ is rapidly metabolized, to provide the most stringent test of its effects as possible we transplanted two distinct samples (AML-3, AML-29) and allowed in vivo leukemic growth to progress for one or 8 week(s), respectively, which typically yields heightened engraftment levels ranging from 20% to 70%, then treated with FICZ daily for one month. Not surprisingly, as observed with the earlier experiments where engraftment levels were also very high, leukemic burden in these experiments was not significantly reduced at endpoint (Fig. 4H). Importantly, both of these samples, however, showed striking reductions in the percentage of primitive CD34⁺ cells in the FICZ versus DMSO-treated grafts (Fig. 4I). Moreover, in the AML-3 grafts, we observed increases in the percentage or mean fluorescence intensity of myeloid differentiation antigens demonstrating an in vivo FICZ-induced promotion of differentiation (Fig. 4J).

It is interesting to note that when a sample responded to FICZ (AML-4 at 100 μg/kg and AML-3 at 250 μg/kg) CYP1B1 was >2-fold upregulated in the FACS-isolated AML grafts at in vivo treatment end, but not upregulated when a sample was nonresponsive (AML-3 at 100 μg/kg; Supplementary Fig. S5B). Although more samples will need to be tested to make definitive conclusions, this trend suggests that the greater the increase in AHR agonism or the activation of AHR beyond a certain threshold in vivo, the more pronounced the leukemic impairment may be. Along these lines, where both a low and high FICZ dose was tested (AML-4), we did find that the increased FICZ dose further promoted AHR agonism and led to more effective impairments of leukemic engraftment. While MSI2 is likely among several repressors of AHR activity in leukemia, AML-4 and AML-7, which were sensitive to even low-dose in vivo AHR ligand treatment, had the highest levels of MSI2 and lowest of CYP1B1 compared with samples less responsive (Supplementary Fig. S5C). These findings highlight the importance of future exploration into the potential for high MSI2-expressing AMLs, which display more significant AHR pathway suppression, to be particularly sensitive to AHR stimulation. All together our in vivo FICZ treatment data provides evidence for irreversible antileukemic effects in vivo and highlights the potential of AHR pathway stimulation to target LSCs.

Although FICZ is potent in its capacity to repress proliferation and promote differentiation in the leukemic context, its propensity to do this selectively with minimal impairments to normal HSPCs would further strengthen its candidacy as a future clinically relevant AML therapeutic strategy. To address this, we treated CD34⁺ CB cells with the same 750 nmol/L concentration of FICZ administered to AML samples. FICZ-stimulated CB cells exhibited AHR pathway activation as measured by CYP1B1 upregulation and a 1.3-fold increase in proliferation over 3 days of culture; however, there were no differences in the proportions of primitive cells or mature myeloid cell types generated over this time (Fig. 5A–D). We also noted a slight increase in cell apoptosis was observed in 1 of the 2 CB samples tested (Fig. 5D). With the exception of increasing BFU-Es in one CB sample, FICZ induced no major differences in progenitor outputs, in either first or secondary replating CFU assays (Fig. 5E and F). Under the same high dose (250 μg/kg) in vivo FICZ administration strategy employed for treating AML xenografts, mice transplanted with CD34⁺ CB cells showed virtually no alteration in the level of either the human CD45⁺ graft or the proportion of CD33⁺, CD14⁺, or CD19⁺ cells within these grafts following FICZ administration (Fig. 5G and H). Finally, we noted no alterations in the primitive CD34⁺ compartment within treated CB grafts for one CB sample while for a second sample, the frequency of CD34⁺ cells was increased (Fig. 5I). Together, these results indicate that at identical doses tested on a range of primary AML samples, all of which were sensitive to FICZ treatment, normal CB cells are minimally affected.

As CYP1B1 levels were elevated by FICZ treatment to similar degrees in both AML and CB CYP1B1 upregulation itself cannot explain the selective prodifferentiative, antiproliferative effects of AHR stimulation on AML cells only. While CYP1B1 remains an effective readout of AHR pathway activation in both contexts, the unique epigenomic landscape of primitive AML versus CB cells could mean that targets accessible to and activated by AHR in the leukemic and normal situation differ. To uncover mechanisms underlying the differential FICZ response, we turned to the list of AHR targets preferentially downregulated in LSCs (Fig. 1G). Of these, ZFP36L1 presented an interesting FICZ target because of its decreased expression in normal and malignant primitive hematopoietic cells relative to more mature myeloid cells (Fig. 6A) and its published effectiveness in promoting myeloid differentiation and inhibiting proliferation (27, 33, 34). Following FICZ stimulation of HL-60, MV4;11 and NOMO-1 cells, for which we showed FICZ-induced impairments (Fig. 2), we observed a significant 1.5- to >5-fold upregulation of ZFP36L1 after FICZ treatment (Fig. 6B). In addition, ZFP36L1 was only upregulated by FICZ in primary AML samples and showed no appreciable elevation in primitive CB cells (Fig. 6C). Of interest, ZFP36L1 shows a trend toward poor prognosis irrespective of AML genotype in a subset of patients with AML blast–specific decreases in AHR and ZFP36L1 expression (Fig. 6D). These results highlight ZFP36L1 as one potential candidate underlying at least some of FICZ's differential effects in this setting.

Our global analysis of AHR targets indicates that the AHR pathway is intrinsically repressed across multiple AML subtypes relative to healthy BM and is repressed within the LSC compartment. By excluding significantly downregulated AHR target genes shared by unfractionated AML cells of different subtypes, we identified 21 highly downregulated LSC-specific genes. The inclusion of known tumor suppressors and validated AHR targets CDKN1A (35) and TLE1 (28) and genes that promote differentiation like ZFP36L1 fits with our evidence of reduced cell-cycle entry and differentiated cell morphology upon FICZ treatment and adds confidence that the AHR pathway represents a core prodifferentiation signaling module LSCs converge on and suppress to promote their self-renewal and sustained proliferation. How LSCs suppress AHR signaling is still an open question and there are likely a number of upstream regulators across diverse leukemias contributing to this dampening. We suggest here that the upregulation of MSI2 in LSC⁺ populations may be one mechanism that facilitates this process given its known roles in attenuating the AHR pathway (15), and our demonstration that MSI2 knockdown impairs leukemogenesis and upregulates canonical AHR targets in AML.

That our work highlights AHR pathway attenuation as having a pro-LSC function is especially important given that AHR has been shown to have tumor suppressive and oncogenic roles, depending on cell and context specificity. In glioma, breast cancer, and multiple myeloma, tumor cells secrete AHR ligands that act in an autocrine fashion to support tumor cell maintenance and/or act on the immune system to inhibit immune responses (36–38). Scoville and colleagues have also shown bulk AML cells can secrete AHR ligands in vitro that impair NK-cell function (39). In contrast, the concept of activating AHR signaling to directly inhibit neoplastic growth is an emerging theme that may hold promise for cancer types such as liver and prostate in which a tumor-suppressive role for the AHR pathway has been shown (40–42), and supports our exploratory studies of AHR pathway stimulation as a novel means to directly inhibit AML progression at the level of LSCs.

While primary sample AML-4 had a favorable response to low-dose (100 μg/kg) in vivo FICZ treatment, other patient AML cells were less susceptible and only began to respond upon administration of a higher dose (250 μg/kg). Although beyond the scope of this study, it will be of interest to evaluate whether underlying genetics or subtype classification may contribute to prediction of future best-responders to FICZ-induced AHR activation. It is interesting to note that the two samples (AML-4 and AML-7) most responsive at the FICZ doses tested here have MSI2 upregulated relative to the other AML samples (Supplementary Fig. S5C), a finding that points to the potential interest in exploring whether MSI2-high leukemias may be more susceptible to AHR ligand therapy. Because AHR pathway suppression appears to occur under most circumstances irrespective of genetic subtype, future efforts placed on understanding how this pathway can be targeted in AML are of clinical interest. Finally, our findings with the use of FICZ as a RA adjuvant in NB4 differentiation and apoptosis promotion present it as a potential option in future multimodal therapies and suggest that it could be advantageous in reducing the dose of existing APL therapeutics. To this point, the selective capacity of FICZ to promote differentiation of primitive AML cells may foreshadow its potential in other combination treatment strategies. Here, synergy of FICZ, employed to achieve effective LSC targeting, in accordance with debulking chemotherapeutics to target blast cells, could allow for robust killing of the entire leukemic clone.

FICZ possesses several important attributes as an AHR agonist that is well suited to therapeutic application as it is rapidly metabolized and well tolerated when administered in vivo. Indeed, in mice given comparable doses to those we have used, impacts on the immune system have been constrained to T cells, are only mildly immunosuppressive and imparting of no detrimental changes to long-term hematopoiesis (32). Furthermore, we have shown that FICZ exerts limited effects on normal CB cells both in vitro and in vivo in comparison with AML cells (Fig. 5). The preferential effect on AML cells could be due to the elevated expression of AHR regulators in the AML context achieving a more extensive repression of AHR signaling. As has been shown for other signaling axes, such heightened repression of AHR signaling in leukemia could indicate a greater dependence on this pathway for maintenance of LSCs and the leukemic state (32, 43). A nonmutually exclusive explanation is that specific AHR targets are activated only in the leukemic and/or LSC context whose restraint is vital for continued leukemic propagation. Cell context and/or ligand-specific AHR target transactivation is a well described phenomenon (44–46). We speculate that differential ligand-induced conformational changes in the receptor and/or distinct interactions with downstream coactivators/repressors in combination with the unique molecular landscape of AML cells allows for unique downstream targets to be activated upon FICZ stimulation of AHR as compared with their normal counterparts. Our demonstration that ZFP36L1 may represent one such target for a subset of leukemias is supportive of future efforts directed toward understanding the FICZ-induced transcriptome more broadly across AML as well as dissecting the contribution of these transcriptional changes to the prodifferentiation inhibitory effects of FICZ on human AML cells.

No potential conflicts of interest were disclosed.

Conception and design: M. Ly, S. Moreira, N. Holzapfel, K.J. Hope

Development of methodology: M. Ly, N. Holzapfel, K.J. Hope

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Ly, S. Rentas, A. Vujovic, J. Xu, N. Holzapfel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Ly, S. Rentas, A. Vujovic, N. Wong, N. Holzapfel, S. Bhatia, D. Tran, K.J. Hope

Writing, review, and/or revision of the manuscript: M. Ly, S. Rentas, N. Wong, S. Moreira, D. Tran, M.D. Minden, J.S. Draper, K.J. Hope

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Ly, N. Wong, K.J. Hope

Study supervision: K.J. Hope

Other (provided primary samples, clinical information, and reviewed manuscript): M.D. Minden

We acknowledge Minomi Subpanditha and Zoya Shapovalova for FACS, Brad Doble, Karun Singh, and Sheila Singh for critical assessment of this work, and all Hope laboratory members for experimental support/advice. This research was supported through an Ontario Institute of Cancer Research Investigator Award to K.J. Hope, Canadian Institute of Health Research (CIHR) grant (#130499 to J.S. Draper), Ontario Graduate Scholarship (to M. Ly), Canadian Blood Services Graduate Fellowship (to S. Rentas), CIHR MD/PhD Scholarship (to N. Holzapfel), CIHR PhD Scholarship (to J. Xu), and Ontario Graduate Scholarship (to D. Tran).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.