Targeting the PI3K/AKT pathway via GLI1 inhibition enhanced the drug sensitivity of acute myeloid leukemia cells

Combination targeted therapy is commonly used to treat acute myeloid leukemia (AML) patients, particularly in refractory/relapse (RR) population. However, concerns have been raised regarding the safety and patient tolerance of combination chemotherapy. It is critical to choose the appropriate treatment for precision therapy. We performed genome-wide RNA profiling using RNA-Seq to compare the RR group and the complete remission (CR) group (a total of 42 adult AML patients). The Hedgehog (Hh) and PI3K/AKT pathways were upregulated in the RR population, which was further confirmed by western blot and/or qPCR. Overexpression of GLI1 in AML cells led to increased AKT phosphorylation and decreased drug sensitivity, which was attenuated by GLI1 inhibition. By contrast, neither the expression of GLI1 nor apoptosis in response to Ara-C treatment of AML cells was significantly affected by PI3K inhibition. Furthermore, co-inhibition of GLI1 and PI3K induced apoptosis of hematopoietic stem/progenitor cells (HSPCs), which raised serious concerns about the side effects of this treatment. These results indicated that GLI1 inhibition alone, but not combined inhibition, is sufficient to enhance AML drug sensitivity, which provides a novel therapeutic strategy for AML treatment.

The Hh pathway plays an important role in cell proliferation, differentiation, apoptosis and migration. It has been shown to cross-talk with other signalling pathways (such as PI3K/AKT, RAS, NOTCH and others) and has been identified as a major target in the treatment of hematological malignancies [11][12][13] . In chronic myeloid leukemia (CML), inhibition of the Hh pathway could reduce the number of leukemia stem cells and reverse drug resistance to imatinib 14 . Recently Zahreddine et al. also demonstrated that a key transcription factor of the Hh pathway, GLI1, affects drug resistance by inducing glucuronidation, which could be reversed by a Hh inhibitor 15 . In addition, several clinical studies have reported that the expression of GLI1 is closely related to the prognosis of acute myeloid leukemia (AML) 16 .
Meanwhile, sustained activation of the PI3K/AKT pathway has been identified in different types of leukemia, and this effect promotes cell cycle progression, inhibits cell apoptosis and facilitates invasion and metastasis of cancer cells [17][18][19] . Our group found that PTEN-regulated PI3K/AKT activation is closely related to the progression and drug resistance of adult AML 20 . In addition, it has been shown that combined inhibition of the Hh and PI3K/ AKT pathways has synergistic anti-leukemia effects in AML and in chronic lymphocytic leukemia 21-23 . However, whether there is cross-talk between the Hh and PI3K/AKT pathways in AML drug resistance/ relapse and the potential risk of combined inhibition of these pathways in AML patients are still unknown. The possibility of using Hh and PI3K double inhibition in clinical AML treatment requires further study.
By comparing the clinical samples of AML-RR and AML-CR patients using RNA-Seq, we showed that both the Hh and PI3K/AKT pathways were upregulated in the AML-RR group. Moreover, overexpression of GLI1 decreased drug sensitivity in AML cells, which could be reversed by a GLI1 inhibitor. In addition, GLI1 inhibition blocked the PI3K/ AKT pathway, whereas PI3K inhibition had no effect on GLI1 expression or AML cell drug sensitivity. This study describes a novel mechanism of Hh and PI3K/AKT pathway interaction in chemo-resistance and relapse of AML patients, which provides guidance to develop an efficient and safe combination targeted therapy for AML-RR patients.

Results
The Hh pathway is activated in AML-RR patients. Forty-two AML patients were recruited in this study. Based on their clinical history, the patients were classified into two groups: the refractory/relapse (AML-RR) group and the complete remission (AML-CR) group (Tables 1 and 2). RNAs were isolated from mononuclear cells (MNCs) of these patients and analysed using RNA-Seq. To identify significant differences in the transcriptomic profiles of the two sample groups (AML-RR and AML-CR), we generated a pair wise scatter matrix by CummeRbund. This method compares and correlates the RPKM profile of all expressed genes in the two sample groups, and it also shows the density distribution of the RPKM for the expressed genes. The RPKM of all expressed genes ranged from 0.003 to 3000 (log 10 RPKM − 2.5 to 3.5), with the majority of the genes ranging from 1 to 100 (log 10 RPKM 0 to 2). The majority of genes involved in the Hh pathway, such as GLI, SMO, SHH, and PTCH, (Fig. 1a-c, Table 3) were upregulated in the AML-RR group compared with the AML-CR group. qRT-PCR and western blot further confirmed the upregulation of GLI1, a key transcript of the Hh pathway, at the mRNA ( Fig. 1d, P < 0.01) and protein (Fig. 1e,f, P < 0.05) levels in the AML-RR group. These results were consistent with the RNA-Seq results, which indicated that the Hh pathway may be related to the progression and prognosis of AML.
The PI3K/AKT pathway is also activated in AML-RR patients. As previously noted, we also found that the majority of genes involved in the PI3K/AKT pathway, such as AKT1, AKT2, AKT3, and BTK ( Fig. 2a-c, Table 3) were upregulated in the AML-RR group compared with the AML-CR group. Western blot validated the increased phosphorylation of AKT in the AML-RR group (Fig. 2d,e, P < 0.01). Together, these results suggested that the PI3K/AKT pathway is activated in the AML-RR group. Overexpression of GLI1 upregulated p-AKT and this effect is abolished by a GLI1 inhibitor in acute myeloid leukemia cell lines. To explore the role of GLI1 and the relationship of Hh and the PI3K/ AKT pathway in AML relapse and drug resistance, GLI1 was overexpressed in both HL60 and NB4 cells by a lentiviral vector carrying EGFP for selection. Seventy-two hours after infection, cells were sorted by FACS to establish new cell lines, which were named HL60/GLI1 and NB4/GLI1. The upregulation of GLI1 in HL60/GLI1 and NB4/GLI1 was also confirmed by both RT-PCR ( Fig. 3a,b) and western blot (Fig. 3c,d). Interestingly, overexpression of GLI1 increased the p-AKT level. GANT61, a GLI1 inhibitor, could reverse the increase in p-AKT by GLI1 overexpression, suggesting that GLI1 functions upstream of p-AKT. LY294002, a PI3K inhibitor, did not inhibit the expression of GLI1 ( Fig. 3a-d). These results indicate that GLI1 could regulate the expression of p-AKT and had a unidirectional regulatory effect on the PI3K/AKT pathway.
Inhibition of GLI1 could inhibit AML cell growth. Ara-C was assessed for its ability to inhibit AML cell growth and apoptosis. Ara-C inhibited the growth of HL60 and NB4 cells in a dose-dependent manner with an IC 50 of 4 μ M for the HL60 cells (Fig. 4a) and an IC 50 of 25 μ M for the NB4 cells (Fig. 4b). A time course experiment was used to optimize the timing and dose of Ara-C. HL60 and NB4 cells were treated with different concentration up to 48 h. The growth inhibition rate did not increase significantly when cells were treated with Ara-C at concentrations higher than 10 μ M at both 24 h and 48 h (Fig. 4a-c). We found that the cell growth inhibition rates in both HL60/GLI1 and NB4/GLI1 cells were significantly lower than the wild-type cells. However, GANT61 treatment could compensate for this difference, suggesting that GLI1 is both sufficient and necessary to decrease drug sensitivity of Ara-C. Consistent with the previous results showing that LY294002 does not affect GLI1 expression, LY294002 inhibited neither HL60/GLI1 nor NB4/GLI1 cell growth (Fig. 4d,e). Besides, there was no difference in growth inhibition rate between the treatment of GANT61 or LY294002 in wild-type cells ( Supplementary Fig. S4a). A major safety concern for AML is that the chemotherapy drugs also lead to severe bone marrow depression, which is related to the inhibition of proliferation in human HSPCs. Here, we tested the effect of GANT61 (20 μ M), LY294002 (20 μ M), and GANT61 + LY294002 (20 μ M) on HSPCs. Surprisingly, the cell growth inhibition rate of the HSPCs in the GANT61 + LY294002 (20 μ M) treatment group was higher than 90% ( Fig. 4f), which raises the serious safety concern that the combination treatment may impair the human hematopoietic system.

Inhibition of GLI1 could enhance acute myeloid leukemia cell drug sensitivity.
To test drug sensitivity, both wild-type and GLI1 overexpressing cells were treated with Ara-C only, Ara-C + GANT61 (20 μ M), and Ara-C + LY294002 (20 μ M) separately for 48 hours, and cell apoptosis was assessed using Annexin V assays (Fig. 5a,d). Consistent with the cell growth inhibitory results, GANT61 significantly enhanced the drug sensitivity of HL60 and NB4 cells, whereas LY294002 failed to do so (Fig. 5b,c). While, there was no difference in cell death rate between the treatment of GANT61 or LY294002 in wild-type cells ( Supplementary Fig. S4b). Moreover, we found that the apoptosis rate of HSPCs with both inhibitors was significantly higher than that in the control group, which raises a serious concern about the safety of the combination treatment (Fig. 5e,f).
Other groups have shown that the AKT1 promoter possesses two GLI1 binding sites (BS1 and BS2) and proved the expression of AKT1 was regulated at the transcriptional level by GLI1 31 . Our results suggested that for AML patients with both GLI1 and AKT activation, the GLI1 inhibitor alone is sufficient, which will significantly decrease the unnecessary side effects of the combination therapy using Hh and PI3K inhibitors. The GLI1 inhibitor would provide a potential targeted therapy to AML cells, which have been reported in other hematonosis 32 . Additional studies are needed to identify the feasibility of clinical application by using GLI1 inhibitor in refractory/relapse population of AML patients.
Our study characterizes the Hh and PI3K/AKT interaction in AML drug sensitivity, which may provide a theoretical basis for further research of targeted therapy.

Materials and Methods
Cell culture and transfection. HL60 and NB4 cells were obtained from the Cell Resource Centre (Xiangya Medical College, Central South University, Hunan, China). The cell lines were maintained in RPMI-1640 medium (Corning Inc., Corning, NY, USA) containing 10% foetal bovine serum (Corning Inc.) and 1% antibiotic solution of penicillin/streptomycin (Sigma, MO, USA) in a 37 °C incubator with a humidified atmosphere of 5% CO 2 (Supplementary Fig. S2). To overexpress GLI1, AML cells were infected with lentivirus containing the GLI1 open reading frame (MOI: 50-100). After 72 h of infection, HL60 and NB4 lines were allowed to recover for 24 h with fresh media and were referred to as HL60/GLI1 and NB4/GLI1 cells (Supplementary Fig. S3). To evaluate the response to drug treatment, 1 × 10 6 HL60/GLI1 and NB4/GLI1 cells were seeded in 6-well culture plates and incubated with either 20 μ M GANT61 (Adooq Bioscience, A13252) or 20 μ M LY294002 (Adooq Bioscience, A10547) for 72 h prior to subsequent measurements. Patient characteristics. Liquid bone marrow samples from 42 patients between 18 and 64 years old, who were diagnosed with AML according to the 2008 WHO criteria and treated at Xiangya Hospital of Central South University, Hunan, China, were included in the AML group (Tables 1 and 2, Supplementary Fig. S1). For comparison, 15 healthy candidate donors were included as normal controls. The experimental protocols were approved by the ethical committee of Xiangya Hospital, Central South University. The informed consent was obtained from all the research subjects. All the bone marrow samples were collected into sterile tubes with anticoagulant (heparin sodium). Mononuclear cells (MNCs) were enriched by density centrifugation over Ficoll-Paque (Sigma, St. Louis, MO, USA) and stored at − 80 °C.    Annexin V APC assay. In each well, 1 × 10 6 cells were seeded in 6-well culture plates and treated with Ara-C at its IC 50 and GANT61/LY294002 (20 μ m). After 24 h of culture, cells were analysed by flow cytometry with an Annexin V-APC kit (Beyotime Institute) to determine the cell death rate.

Statistical analysis.
All experiments were repeated independently at least three times, and the data were represented as the mean ± SD of the vehicle controls. All groups were compared using the Statistical Package for the Social Sciences (SPSS) version 20.0. The differences between the populations were assessed by analysis of covariance (ANCOVA). The diagrams were generated by GraphPad Prism 6 software. Significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Ethical approval. The experimental protocols and methods were performed in accordance with relevant guidelines and regulations. The informed consent was obtained from all the research subjects.