Gas6/AXL Signaling Regulates Self-Renewal of Chronic Myelogenous Leukemia Stem Cells by Stabilizing β-Catenin

This article is featured in Highlights of This Issue, p. 2607

Chronic myelogenous leukemia (CML) is a type of myeloproliferative disorder characterized by uncontrolled clonal expansion of immature granulocytes and their precursors (1). CML is believed to stem from the malignant transformation of bone marrow hematopoietic stem cells (HSC) triggered by a fusion gene BCR-ABL because of a reciprocal chromosomal translocation t(9,22)(q34;q11) (2). Most adult patients with chronic-phase CML respond well to tyrosine kinase inhibitor (TKI) imatinib mesylate (IM; ref. 3). However, acquired resistance to IM conferred by multiple mechanisms [e.g., BCR-ABL mutations (4), leukemia stem cells (LSC; ref. 5)] is an emerging problem. In addition, BCR-ABL–positive pediatric acute lymphoblastic leukemia and adult accelerated phase- or blast-crisis-CML usually respond poor to TKIs (6).

LSCs in CML show common characteristics of cancer stem cells (CSC; refs. 7–9). These traits endow LSCs with resistance to IM. CML patients with complete molecular response, still showed long-term BCR-ABL–negative LSCs (10). Only approaches to effectively eliminate LSCs when combined with IM may help cure CML (7, 8, 11). Therefore, development of rational approaches to eliminate LSCs is imperative.

The self-renewal capacity of LSCs is regulated by intrinsic CSC regulation pathways (e.g., WNT/β-catenin, Hedgehog; refs. 12, 13), metabolism regulators (e.g., ALOX5, SCD; refs. 14, 15), transcription factors (e.g., Foxo3a, HIF1α, γ-catenin; refs. 16–18) and epigenetic regulators (e.g., PRMT5; refs. 19, 20). However, the whole regulation network of LSCs is not fully realized. The complexity is enhanced by the LSC local microenvironment in bone marrow (21). Stromal cells tightly regulate the self-renewal of LSCs via direct contact and by secreting soluble cytokines (22, 23).

AXL is a member of the TAM receptor tyrosine kinases (TAMR), which also includes Mer and Tyro3 (Sky) (24, 25). Two highly homologous ligands for TAMRs are growth-arrest–specific gene 6 (Gas6) and protein S (25). Gas6 has the highest affinity for AXL, whereas protein S predominantly binds to Tyro3 and Mer (26). AXL was originally identified in 1991 from patients with CML (27), then found to be overexpressed in multiple human cancers (24). AXL is involved in acquired resistance to endothelial growth factor receptor inhibitors in lung cancer (28), FLT3-targeted therapy in acute myelogenous leukemia (29) and IM through a bypass mechanism (30).

AXL overexpression was found to induce the epithelial–mesenchymal transition (EMT) and increase tumorigenicity in breast cancer cells (31). The underlying mechanism may involve upregulating β-catenin, snail, slug, and NF-κB triggered by Gas6/AXL activation (31). Recently, AXL was identified as a key regulator for mesenchymal glioblastoma stem-like cells (32). We discovered that among the three TAMR family members, AXL was selectively expressed in primary CML CD34⁺ cells and hypothesized that overexpression of AXL in LSCs and Gas6 by bone marrow stromal cells might increase the self-renewal capacity of LSCs and confer resistance to IM. In this study, we tested this hypothesis and investigated the impact of forced expression of AXL, silencing it via shRNA or inactivating it by small-molecule inhibitors on survival and self-renewal capacity of LSCs in vitro and in CML mice.

Peripheral blood or bone marrow samples (Supplementary Table S1) were obtained from patients with CML and from healthy adult donors in Sun Yat-sen Memorial Hospital, The First Affiliated Hospital of Sun Yat-sen University, Guangdong General Hospital, and The First Affiliated Hospital of Jinan University after informed consent according to the institutional guidelines and the Declaration of Helsinki principles. The isolation and culture of human CD34⁺ cells were performed as described previously (16, 19, 20, 33).

Peripheral blood samples were obtained from patients with CML and from healthy adult donors. Mononuclear cells isolated by histopaque gradient centrifugation were resuspended in DMEM-low glucose medium (Gibco) supplemented with 10% FBS and 1% penicillin–streptomycin. The cells were kept at 37°C in a humidified incubator with 5% CO₂ for 48 hours. After discarding the nonadherent cells, the adherent cells were cultured for 2 to 4 weeks and used as mesenchymal stromal cells (MSC) in subsequent experiments (34).

Whole cell lysates were prepared in RIPA buffer (35). The membranes were scanned by using the Odyssey infrared imaging system (LI-COR). Detailed information for antibodies and their sources is described in Supplementary Materials and Methods.

Human and mouse Gas6 immunoassay involved use of the Quantikine ELISA Kit (R&D Systems; ref. 36). Cell culture supernatants or plasma from CML patients or healthy individuals were used for detecting the secreted Gas6.

Lentivirus were produced by transfection in 293T cells with control shRNA (Scramble) or specific shRNA targeting AXL together with the pCMV-dR8.2 packing construct and the pCMV-VSVG envelope construct. CML or NBM CD34⁺ cells (1 × 10⁶ cells/mL) were transduced by spinoculation (1,500 g, 90 minutes, 32°C) with virus-containing supernatants two rounds. Cells were harvested 48 hours later for further analysis (16, 20).

Apoptosis of primary CD34⁺ cells in CML patients or NBM was determined as described previously (35).

CFSE-labeled cells were cultured with XL880 and R428 for 96 hours, cells were harvested and stained with Annexin V-PE, and quiescent cell apoptosis (CD34⁺ CFSE^max Annexin V⁺) was analyzed by using with BD FACS Aria II flow cytometer (8, 16, 20).

Colony-formation cell (CFC)/replating and LTC-IC assay were performed as reported previously (16, 20, 37). Detailed information is described in Supplementary Materials and Methods.

The retrovirus was produced by transient transfection with the MSCV-BCR-ABL-IRES-EGFP construct in Plat-E cells (9). Bone marrow cells from 6- to 8-week-old donor male C57BL/6 mice primed with 5-fluorouracil (5-FU) were stimulated with cytokines in vitro, then transplanted by tail-vein injection into irradiated (550 cGy) recipient female C57BL/6 mice after transduction two rounds with the MSCV-BCR-ABL-IRES-EGFP retrovirus (16, 20). All the mice developed CML in approximately 2 to 3 weeks.

The detailed information for the treatment of CML mice and detection of LSK, LT-HSC and ST-HSC is described in Supplementary Materials and Methods.

Primary CD34⁺ cells from 2 CML patients (1 CP-CML patient and 1 BP-CML patient) were cultured with XL880 (1.0 μmol/L) for 72 hours, then cells (1 × 10⁶ cells/mouse) were transplanted by tail-vein injection into sublethally irradiated (300 cGy) 8-week-old NOD-scid-IL2Rg^−/− (NSI) which were generated by TALEN-mediated Il2rg gene targeting in NOD/SCID mice (Guangzhou Institutes of Biomedicine and Health, Chinese Academy of Sciences; refs. 20, 38). Mice were euthanized after 12 weeks, then bone marrow cells and spleen cells were harvested; the engraftment of human cells was analyzed by using BD FACS Aria II flow cytometer.

All experiments were performed 3 times, and results are presented as mean ± SD. GraphPad Prism 5.0 (GraphPad Software) was used for statistical analysis. Data for 2 groups were compared by Student t test and for multiple groups by one-way ANOVA, post hoc intergroup comparisons, Tukey test. A P value of <0.05 was considered statistically significant.

We first examined the expression of AXL and other TAMRs members in primary CD34⁺ cells purified with immunomagnetic columns from CML patients and NBM from healthy donors. The mRNA level of AXL, but not other TAMR members (e.g., Mer and Tyro3), was significantly higher in the primary CD34⁺ cells from CML patients than from NBM (Fig. 1A). The levels of phospho-AXL and AXL were higher in CML CD34⁺ cells than NBM CD34⁺ cells as determined by Western blot analysis (Fig. 1B). Flow cytometry results showed that there was surface expression of AXL in CML CD34⁺ cells (Fig. 1C). We next examined whether BCR-ABL regulated AXL expression in CML CD34⁺ cells. The levels of phospho-AXL and AXL were not changed after treatment with IM (Fig. 1D). However, BCR-ABL knockdown by lentiviral shRNA in the primary CML CD34⁺ cells dramatically decreased the AXL mRNA levels (Fig. 1E). These data suggest that AXL expression is dependent on BCR-ABL protein level rather than its tyrosine kinase activity.

Forced overexpression of human AXL in K562 cells reduced sensitivity to IM, suggesting that increased AXL confers resistance to IM (Supplementary Fig. S1A and S1B). The percentage of side population (SP) cells and Aldefluor⁺ cells were also significantly increased in K562 overexpressing AXL as compared with the control (Supplementary Fig. S1C and S1D). In addition, AXL overexpression upregulated the expression of stemness-associated genes ALDHA1, OCT4, SOX2 and NOTCH1 but not LIN28A and NANOG (Supplementary Fig. S1E).

To further define the role of AXL in CML CD34⁺ cells, we transduced the primary CD34⁺ cells purified from CML patients or NBM with scramble shRNA or lentivirus AXL shRNA for 48 hours, then treated them with IM for 24 hours. The knockdown effect of AXL in CML CD34⁺ cells and NBM CD34⁺ cells was confirmed by qRT-PCR (Fig. 1F and Supplementary Fig. S2A). Given the gene homology of AXL with Mer and Tyro3, we examined the levels of these receptors in cells treated with shRNA against AXL by Western blot analysis. The results showed that the effect of AXL knockdown was of selective (Fig. 1G). AXL silencing significantly increased apoptosis in CML CD34⁺ cells (Fig. 1H and I) but not in NBM CD34⁺ cells (Supplementary Fig. S2B and S2C). The effect of AXL knockdown on CML CD34⁺ cells was augmented with combined IM treatment (Fig. 1I and Supplementary Fig. S2C), in which the BCR-ABL activity as reflected by its phosphorylation (Y245) was completely inhibited by IM (Fig. 1J). Because the self-renewal capacity is an important trait of CSCs, we determined whether silencing AXL by its specific lentivirus shRNA impaired the CFC/replating ability of CML stem cells. The primary CD34⁺ cells of human CML and NBM were transduced with lentiviral shAXL for 72 hours and underwent three rounds of 14-day serially replating of CFC culture in methylcellulose medium. When AXL was silenced, the CFC/replating ability was significantly decreased in human CML CD34⁺ but not in NBM CD34⁺ cells (Fig. 1K and Supplementary Fig. S2D). Thus, AXL knockdown reduced CML CD34⁺ cell survival and growth, and IM treatment further suppressed CML CD34⁺ cell growth.

We next used a widely used human BCR-ABL gene-driven CML mouse model (Fig. 2A) to examine the in vivo effect of AXL knockdown on CML LSCs (14, 16, 20, 33). We knocked down AXL with shRNA in spleen GFP⁺ cells from the first generation of CML mice (Fig. 2B), then transplanted these spleen cells into sublethally irradiated recipient C57BL/6 mice. The results showed that AXL knockdown alone or combination with IM potently blocked splenomegaly (Fig. 2C) and markedly decreased the proportion of BCR-ABL–expressing (i.e., GFP⁺) leukemia cells, myeloid cells (Gr-1⁺ Mac-1⁺) in bone marrow cells (Fig. 2D and E) and splenic cells (Supplementary Fig. S3A and S3B) of CML mice.

AXL knockdown or combination with IM greatly decreased the proportions of LSK cells (Lin⁻Sca-1⁺c-Kit⁺), LT-HSCs (LSK Flt3⁻CD150⁺CD48⁻) and ST-HSCs (LSK Flt3⁻CD150⁻CD48⁻) in bone marrow cells (Fig. 2F–I) and splenic cells (Supplementary Fig. S3C–S3E) of CML mice. Similarly, the proportions of granulocyte-macrophage progenitors (GMPs, Lin⁻Sca-1⁻c-Kit⁺CD34⁺FcγRII/III^hig) and common myeloid progenitors (CMPs, Lin⁻Sca-1⁻c-Kit⁺CD34⁺FcγRII/III^low) in bone marrow cells (Fig. 2J–L) and splenic cells (Supplementary Fig. S3F and S3G) were significantly reduced in AXL knockdown mice. IM treatment alone did not significantly reduce the proportions of LSKs, LT-HSCs, ST-HSCs, GMPs and CMPs in bone marrow cells and splenic cells of CML mice (Fig. 2F–L and Supplementary Fig. S3C–S3G), which agreed with previous reports (7, 8). AXL knockdown significantly prolonged survival of CML mice (Fig. 2M). In conclusion, silencing AXL eliminated LSCs in vivo and prolonged survival of CML mice.

Next, we determined the effect of the small-molecule AXL inhibitors XL880 (39) and R428 (40) on survival of primary CML CD34⁺ cells. The data showed that XL880 treatment increased the apoptosis in CML CD34⁺ cells (Fig. 3A, top). Combined treatment with XL880 and IM enhanced the apoptosis in CML CD34⁺ cells, suggesting that XL880 sensitized CML CD34⁺ cells to apoptosis in response to IM. XL880 alone or with IM did not induce apoptosis in CD34⁺ cells from NBM (Supplementary Fig. S4A and S4B). Similar results were obtained when the primary CML CD34⁺ cells were exposed to R428, a more selective AXL inhibitor (Fig. 3A, bottom). In these settings, both XL880 and R428 treatment blocked the AXL tyrosine phosphorylation in CML CD34⁺ cells (Fig. 3B).

To further identify the effectiveness of AXL inhibition on quiescent LSCs, we labeled primary CD34⁺ cells of human CML patients or NBM with carboxyfluorescein diacetate succinimidyl ester (CFSE), then incubated with XL880 (1.0 μmol/L) for 96 hours. XL880 treatment markedly increased apoptosis in undivided CML CD34⁺ cells but not undivided NBM HSCs (Fig. 3C and Supplementary Fig. S4C and S4D). R428 treatment exhibited similar effect compared with XL880 in undivided CML CD34⁺ cells (Fig. 3D).

We next evaluated whether inhibiting AXL by XL880 and R428 impaired the self-renewal ability of CML stem cells by CFC/replating and LTC-IC assay. In the presence of XL880 and R428, the CFC/replating ability was greatly decreased in the primary CML CD34⁺ cells but not in NBM CD34⁺ cells (Fig. 3E and F and Supplementary Fig. S4E). Similarly, XL880 greatly decreased the number of LTC-IC–derived colony-forming units for CML bone marrow cells as compared with NBM cells (Fig. 3G and Supplementary Fig. S4F). Combined treatment with XL880 and IM enhanced the decreased CML LTC-IC frequency as compared with either agent alone (Fig. 3G). Thus, pharmacologic inhibition of AXL may reduce the survival and self-renewal ability of human CML LSCs.

To examine the in vivo effect of pharmacologic inhibition of AXL on CML LSCs, we used the CML mouse model again mentioned before. Mice were treated with XL880, IM or both (Fig. 4A). Splenomegaly was ameliorated in mice that received XL880 or both XL880 and IM (Fig. 4B). XL880-treated CML mice showed significantly prolonged survival, and XL880 with IM extended survival as compared with each alone (Fig. 4C).

XL880 alone or with IM markedly decreased the proportion of BCR-ABL–expressing (GFP⁺) leukemia cells, myeloid cells (Gr-1⁺ Mac-1⁺) in bone marrow cells (Fig. 4D and E) and splenic cells (Supplementary Fig. S5A and S5B) of CML mice. XL880 alone or with IM greatly decreased the proportions of LSK cells (Lin⁻Sca-1⁺c-Kit⁺), LT-HSCs (LSK Flt3⁻CD150⁺CD48⁻), and ST-HSCs (LSK Flt3⁻CD150⁻CD48⁻) in bone marrow cells (Fig. 4F–I) and splenic cells (Supplementary Fig. S5C–S5E) of CML mice.

The proportions of granulocyte-macrophage progenitors (GMPs, Lin⁻Sca-1⁻c-Kit⁺CD34⁺FcγRII/III^hig) and common myeloid progenitors (CMPs, Lin⁻Sca-1⁻c-Kit⁺CD34⁺FcγRII/III^low) in bone marrow cells (Fig. 4J–L) and splenic cells (Supplementary Fig. S5F and S5G) were significantly reduced in XL880-treated mice. Taken together, pharmacologic inhibition of AXL eliminated LSCs and sensitized CML LSCs to IM in vivo.

We investigated the effect of ex vivo treatment with XL880 in human CML CD34⁺ cells on their ability to be engrafted in NOD-scid-IL2Rg^−/− (NSI) mice (Supplementary Fig. S6A). XL880 treatment reduced the engraftment of human CML CD45⁺ cells in bone marrow (Supplementary Fig. S6B) and spleens (Supplementary Fig. S6C) at 12 weeks after transplantation. The engrafted human CML CD45⁺ cells sorted by flow cytometry from NSI murine bone marrow cells expressed BCR-ABL, and XL880 treatment markedly decreased BCR-ABL mRNA levels (Supplementary Fig. S6D). The proportions of engrafted human CML CD33⁺ and CD14⁺ myeloid cells in NSI murine bone marrow cells were decreased with XL880 treatment (Supplementary Fig. S6E and S6F). XL880 treatment may selectively target human primitive CML cells with in vivo engraftment capacity.

Wild-type AXL executes its function dependently by binding its ligand Gas6. Given that AXL was overexpressed in CML LSCs, we wondered whether CML CD34⁺ cells produced extra Gas6 to form an autocrine regulation loop. We evaluated Gas6 mRNA level in primary CML and NBM CD34⁺ cells and found that Gas6 gene transcription was similar in the primary CD34⁺ cells from CML patients (Fig. 5A). The data did not support a Gas6/AXL autocrine regulation loop in CML LSCs.

The stemness of LSCs is tightly regulated by the components of its surrounding niches (22). We next examined the plasma concentration of Gas6 in CML patients and found increased plasma Gas6 level in CML patients as compared with healthy donors (Fig. 5B). We detected the cellular human Gas6 expression in the primary MSCs from CML patients and healthy individuals by qRT-PCR, the results showed that the Gas6 mRNA levels were higher in CML MSCs than normal MSCs (Fig. 5C). Furthermore, the secreted Gas6 levels as determined by ELISA were also increased in the cell culture supernatants of CML MSCs compared with normal MSCs (Fig. 5D). These data together suggest the cellular and secreted Gas6 levels appears increased in CML MSCs. Co-cultured with CML MSCs for 10 days significantly promoted growth of CML CD34⁺ cells (Supplementary Fig. S7A). The number of CFCs for CML CD34⁺ cells co-cultured with MSCs was significantly elevated in comparison with those without co-culture with MSCs (Fig. 5E).

To determine the reason for the increased Gas6 level in CML, we seeded human leukemia K562 cells on the feeder layer of murine BMDSCs OP9 cells, then evaluated murine Gas6 secretion in the culture medium by ELISA. The level of murine Gas6 was higher in co-culture than with OP9 or K562 cells alone. Of note, murine Gas6 but not human soluble Gas6 level was increased in the co-culture system, suggesting that CML cells might educate OP9 cells to secrete Gas6 (Fig. 5F). Indeed, with the number of K562 cells increased in the system of a constant number of OP9 cells, the soluble murine Gas6 level was increased, CML cells indeed educated OP9 cells to dose-dependently secrete Gas6 (Supplementary Fig. S7B). These data suggest that BMDSCs secreted Gas6, which activated the AXL membrane receptor on leukemia cells to form a paracrine regulation loop.

To assess whether Gas6 regulates CML responses in vitro, we found that Gas6 stimulation induced AXL activation (Fig. 5G) and increased growth of CML cells (Fig. 5H). The protective role of stromal cells or Gas6 on CML cells was then investigated in the co-culture system with or without AXL-Fc chimeric protein treatment. AXL-Fc contains the extracellular domain of the AXL protein and the Fc portion of human IgG1. AXL-Fc binds Gas6 and prevents it from interacting with AXL (39). We counted the number of K562 cells when incubated in the presence or absence of AXL-Fc chimeric proteins without or with co-culture with human HS5 BMDSCs or Gas6 stimulation. AXL-Fc treatment counteracted the protective role of BMDSCs (Fig. 5I) or Gas6 (Fig. 5J).

In addition, Gas6 stimulation increased the self-renewal ability of CML CD34⁺ cells detected by CFC/replating assay (Fig. 5K). Next, we evaluated whether blocking the Gas6/AXL paracrine loop inhibited the self-renewal of human CML CD34⁺ cells. After transduced with human AXL lentiviral shRNA for 48 hours, the CML CD34⁺ cells were co-cultured with human HS5 cells for another 72 hours for CFC/replating assay. In parallel, HS5 cells were infected with human Gas6 lentiviral shRNA for 48 hours, then co-cultured with the primary CML CD34⁺ cells for 72 hours for CFC/replating assay. Silencing AXL in primary CML CD34⁺ cells or Gas6 in stromal cells markedly inhibited the CFC/replating ability of primary CML CD34⁺ cells in the co-culture system (Fig. 5L–O). Similarly, blocking the Gas6/AXL paracrine loop diminished the self-renewal of human CML CD34⁺ cells co-cultured with human CML MSCs (Supplementary Fig. S7C–S7F).

Collectively, the Gas6/AXL paracrine loop positively regulated the self-renewal capacity of human CML LSCs, hinting that targeting the Gas6/AXL paracrine loop may be of benefit for treatment of CML.

The WNT/β-catenin pathway regulates the self-renewal capacity of CML LSCs (12, 41). We examined whether AXL affected β-catenin level. Levels of β-catenin and its downstream target genes (e.g., c-MYC and LEF1) were greatly increased in AXL- overexpressing K562 cells (Fig. 6A, left and middle). Given that DVL (Dishevelled) is a positive upstream regulator in the WNT/β-catenin pathway, we set to examine DVL. Among the three isoforms of DVL, only DVL2 and DVL3 were detectable in K562 cells. The AXL-overexpressing K562 cells showed an increased DVL3 but not DVL2 (Fig. 6A, left and middle). Similar results were obtained in Gas6-stimulated K562 cells (Fig. 6A, right). Conversely, AXL knockdown in the primary CML CD34⁺ cells dramatically inhibited WNT/β-catenin pathway (Fig. 6B). However, the level of BCR-ABL was unaltered, suggesting that BCR-ABL was not likely to be regulated by AXL in the primary CML CD34⁺ cells (Fig. 6B). Similarly, AXL inhibition by XL880 reduced levels of β-catenin and downstream genes in primary human CML CD34⁺ cells (Supplementary Fig. S8A).

Next, we ascertained whether AXL overexpression activates TCF4/LEF1-dependent gene transcription. The data indicated that AXL overexpression increased TCF4/LEF1-dependent luciferase activity (Supplementary Fig. S8B). AXL overexpression also increased the levels of β-catenin target genes (e.g., AXIN2, c-MYC, LEF1 and CCND1; Supplementary Fig. S8C). In a separate approach, the addition of Gas6 increased TCF4/LEF1-dependent luciferase activity (Supplementary Fig. S8D). Conversely, treatment with AXL inhibitor XL880 in K562 cells decreased luciferase activity (Supplementary Fig. S8E), as well as the transcription of these β-catenin target genes in human primary CML CD34⁺ cells (Supplementary Fig. S8F).

To explore the mechanism by which AXL regulates β-catenin, we performed time–chase experiments. AXL overexpression decelerated the turnover of β-catenin protein (Fig. 6C). Similar results were obtained with Gas6 stimulation (Fig. 6D). K562 cells stably overexpressing AXL and Gas6-stimulated K562 cells showed increased phospho-(S9)-GSK3β, phospho-AKT and phospho-ERK1/2 levels (Fig. 6E, left and middle). Conversely, pharmacologic inhibition of AXL decreased the levels in human primary CML CD34⁺ cells (Fig. 6E, right). Because phospho-(S9)-GSK3β inhibited the activity of GSK3β and decreased the degradation of β-catenin in primary CML CD34⁺ cells (Fig. 6E, right), we hypothesized that AXL activation might phosphorylate GSK3β at S9. To identify the regulator responsible for GSK3β S9 phosphorylation, we used specific inhibitors against MEK/ERK1/2 or PI3K/AKT pathway. The PI3K inhibitor LY294002 but not MEK inhibitor U0126 abrogated the Gas6-induced accumulation of β-catenin (Fig. 6F), which suggests that AKT but not ERK1/2 mediated the phosphorylation of GSK3β S9 in primary CML CD34⁺ cells.

To examine the effect of Gas6/AXL on AKT activation in the absence of BCR/ABL, we knocked down BCR-ABL in primary CD34⁺ cells with lentivirus BCR-ABL shRNA, and then the cells were stimulated with Gas6, the results showed that AKT was activated with Gas6 stimulation in BCR-ABL knockdown cells (Fig. 6G), which is indicative of the critical role of Gas6/AXL on AKT.

Finally, we determined whether AXL exerted its function through β-catenin. The CML CD34⁺ cells were transduced with Scramble (Scr) or AXL lentivirus for 48 hours, then the cells underwent transfection with pcDNA3 (Vector) or pcDNA3-β-catenin (β-catenin) for another 48 hours. Ectopic expression of β-catenin rescued the AXL knockdown-mediated decrease in CFC formation in CML CD34⁺ cells (Fig. 6H).

Taken together, Gas6/AXL ligation leads to AKT-dependent stabilization of β-catenin protein in CML CD34⁺ cells.

LSCs are important resources of TKI resistance and CML relapse. In the present study, we discovered that among three of the TAMR family members, AXL was selectively overexpressed in primary CML CD34⁺ cells. Inhibiting AXL by shRNA or small-molecule inhibitors reduced the survival and self-renewal capacity of human CML LSCs. AXL knockdown and pharmacologic inhibition of AXL prolonged the survival of CML mice and eliminated LSCs in mice. AXL inhibition reduced long-term engraftment of human CML CD34⁺ cells in NSI mice. A Gas6/AXL paracrine loop conferred the self-renewal capacity of human CML CD34⁺ cells. Gas6/AXL ligation stabilizes β-catenin in an AKT-dependent manner in CML CD34⁺ cells. Our findings improve the understanding of LSC regulation and validate a pharmacologic target for eliminating LSCs. Our study represents the first description of Gas6/AXL as a critical regulator of the self-renewal capacity of CML LSCs conferring IM resistance.

Among all members of the TAMR family, only AXL has been reported to be involved in the EMT in glioblastoma and breast cancer (31, 32, 42). Consistently, we identified AXL overexpression in CML LSCs. AXL-overexpressing in CML cells increased the proportions of SP and Aldefluor⁺ cells and stemness-related genes including ALDHA1. In agreement with our findings, erlotinib-resistant non–small cell lung carcinoma cell populations contained a greater proportion of Aldefluor⁺ cells than parental erlotinib-responsive cells with concurrent increase in AXL expression (43).

Overexpressed AXL may be abundantly accessed by its ligand Gas6. However, identifying any gain-of-function mutations or genomic amplification in AXL in CML LSCs needs further investigation.

Little is known about the underlying mechanism of AXL overexpression in LSCs. Our results showed that AXL mRNA and protein levels were higher in human primary CML CD34⁺ cells as compared with NBM counterparts. Furthermore, AXL transcriptional expression appears dependent on BCR-ABL protein level rather than its tyrosine kinase activity. It was previously reported that transcription factors (e.g., AP1) and miRNAs (e.g., miR-34a) lead to AXL overexpression (40). TIG1 in breast cancer cells can increase AXL protein stabilization by physical interaction (44). Further study is needed to identify the mechanism of increased AXL expression in LSCs.

Multiple types of cells, including fibroblasts, endothelial cells, osteoblasts, liver cells, and cancer cells can secrete Gas6 (45). In bone marrow, Gas6 is believed to be secreted by fibroblasts and other stromal cells to maintain hematopoietic homeostasis (46). Our results support that Gas6 ligation to AXL to promote survival and self-renewal capacity of LSCs. Conversely, Gas6 knockdown in different stromal cells (e.g., HS5, primary CML MSCs) attenuated the self-renewal of LSCs. Similarly, the addition of soluble AXL-Fc into the co-culture of LSCs and HS5 BMDSCs blocked the pro-survival and pro-self-renewal effect of Gas6 on CML LSCs. Our data did not support that LSCs generated Gas6 for autocrine regulation. This finding disagrees with a previous report that bulk AML cells enhanced Gas6 secretion to form an autocrine loop (45). The difference may be due to variant cell types. Of interest, CML cells educated BMDSCs to secrete Gas6, which can increase the self-renewal ability of CML CD34⁺ cells.

Overexpression of AXL in CML cells increased the protein level of β-catenin and facilitated its transcriptional activity, as reflected by the expression of its downstream target genes. Stimulation of Gas6 in CML cells provoked similar effects. Conversely, silencing AXL decreased β-catenin protein and its transcriptional activity. In function, the critical role of β-catenin in CML LSCs was revealed as forced expression of β-catenin rescued the AXL knockdown-mediated decrease in CFC formation in CD34⁺ cells.

BCR-ABL tyrosine kinase can directly phosphorylate β-catenin at Y86 and Y654 residues to increase β-catenin protein stability (47). However, the phosphorylation at Y654 of β-catenin was not altered in AXL-overexpressing CML cells or when starved K562 cells were stimulated with Gas6 (data not shown), which suggests that AXL might not increase β-catenin protein stability in such a direct manner. Therefore, we turned our attention to the canonical cascade of WNT/β-catenin. After inhibiting AXL by shRNA or small-molecule inhibitor XL880 in CML CD34⁺ cells, phospho-(S9)-GSK3β was markedly decreased, which was helpful to explain the decreased β-catenin levels. Regarding the mediator to render the decreased phospho-(S9)-GSK3β, our results pinpointed its upstream regulators—AKT but not ERK1/2. AKT is able to phosphorylate β-catenin at S552 to facilitate its nucleus accumulation and transcriptional activity (48). Consistently, AKT is required for Gas6/AXL to protect cells from apoptosis in ovarian cancer (49). Because LY294002 is a PI3K inhibitor with toxicity, future studies should define the potential application value of more specific inhibitors of the PI3K/AKT pathway in eliminating LSCs.

In conclusion, overexpression of AXL or ligation of Gas6/AXL increases the self-renewal capacity of CML LSCs in vitro and in vivo. LSCs may educate BMDSCs to secrete Gas6, assumably forming a vicious circle (proposed model, Fig. 6I). Our findings may provide a rationale for a clinical trial of IM combined with AXL inhibitor foretinib, which is tolerant during a phase II clinical trial, in resistant and refractory CML (50).

No potential conflicts of interest were disclosed.

Conception and design: Y. Jin, J. Pan

Development of methodology: Y. Jin, C. Liu, J. Zhou

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Jin, D. Nie, Y. Lu, Y. Li, C. Liu, J. Zhou, J. Pan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Jin, J. Zhou, J. Pan

Writing, review, and/or revision of the manuscript: Y. Jin, J. Pan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Jin, J. Pan

Study supervision: J. Pan

Other (provided specimens of CML patients): J. Li, X. Du

This study was supported by grants from National Natural Science funds (grant nos. U1301226, 81373434, 81025021, and 91213304 to J. Pan; 81473247 and 81673451; to Y. Jin), Guangdong Natural Science Funds for Distinguished Young Scholars (Grant no. 2016A030306036 to Y. Jin the Research Foundation of Education Bureau of Guangdong Province, China (grant no. cxzd1103; to J. Pan).

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.