All reagents were from commercial sources. With tetramethylsilane (TMS) as internal standard, the ¹H NMR was recorded on Bruker AV-300 apparatus by using deuterated solvents. HR-MS was collected on Agilent technologies 6520 Accurate-Mass Q-TOF LC/MS instruments. Every targeted compound was purified via silica gel (60Å, 70–230 mesh) column chromatography.

methyl (2S, 3S, 4R, 5R, 6R)-3, 4, 5, 6-tetrahydroxytetrahydro-2H-pyran-2-carboxylate

Material A (35.2 g, 0.20 mol) was dissolved in CH₃OH (200 ml) and the pH was adjusted to 8 with 1 M/L CH₃ONa at 0°C temperature. The mixture was stirred for 6 h at room temperature maintaining the pH value of 8. Monitored by TLC, the reaction was completed and provided compound B 37.5 g, yield: 90%.

(2R, 3S, 4S, 5R, 6R)-6-(methoxycarbonyl)tetrahydro-2H-pyran-2, 3, 4, 5-tetrayl tetraacetate

To the mixture of acetic anhydride (200 ml) and compound B (37.9 g, 0.18 mol), HClO₄ (1 ml) was added dropwise. The reaction was stirred overnight. By replenishing HClO₄ (0.5 ml), the mixture was reacted for 4–5 h sequentially and affording compound C 14.6 g, yield: 21%.

(2S, 3R, 4R, 5S, 6S)-2-bromo-6-(methoxycarbonyl)tetrahydro-2H-pyran-3, 4, 5-triyl triacetate

Compound C (9.0 g, 0.023 mol) was added to HBr/acetic acid (90 ml) at low temperature, stirring for 1 h at 0°C and 3 h at room temperature. Then the system was extracted by water (40 ml) and dichloromethane (50 ml), and the dichloromethane layer was adjusted to neutral by saturated NaHCO₃ (50 ml). After filtrated, the precipitate was dried and gave Compound D 6.03 g, yield: 66%.

(2S,3R,4R,5S,6S)-2-((5-hydroxy-6-methoxy-4-oxo-2-phenyl-4H-chromen-7-yl)oxy)-6-(methoxycarbonyl)tetrahydro-2H-pyran-3, 4, 5-triyl triacetate

Compound D (5.0 g, 0.013 mol), OA (2.3g, 0.0081 mol) and CaSO₄ (2.3g, 0.0169 mol) were dissolved into quinoline (60 ml), then, Ag₂CO₃ was added. The mixture was reacted in dark at room temperature overnight. After that CHCl₃ (60∼70 ml) was added to the system, whose filtrate was washed with dilute sulphuric acid till there was no temperature rising, and CHCl₃ layer was collected and purified as Compound E 1.8 g, yield: 28%.

Compound E (1.8 g, 0.00388 mol) was dissolved in acetone (36 ml), 1 M/L NaOH was added dropwise for 30–40 min at a low temperature, after acetone was eliminated, the system should be adjusted to pH 3–4 until there was solid precipitating, followed by filtrating and drying, the final product Oroxyloside A (OAG) was obtained 1.1g, yield: 71%. ¹H NMR (300 MHz, DMSO-d₆) δ 3.35–4.09 (sugar H), 3.78 (3H, s, OCH₃), 5.35 (1H, d, J = 6.5 Hz, H), 7.06 (1H, s, H), 7.13 (1H, s, H), 7.60–7.62 (3H, m, H), 8.08–8.11 (2H, m, H), 12.8 (1H, s, 5-OH). HRMS (ESI): m/z, calculated for C₂₂H₂₁O₁₁, 461.1078 (M + H)⁺, found 461.1075, and the purity was over 99% by HPLC (Supplementary Figures S1, S2).

Primary antibodies against Shh, GLI1, BCR-ABL, BCL-XL, Caspase 3, 7 and 9 were purchased from Cell Signaling Technology (Beverly, MA, United States). Primary antibodies against GAPDH, Tubulin α/β, PI3K, BAD, and p-BAD (Ser136) were obtained from Bioworld (Dublin, OH, United States). Primary antibodies against AKT1/2/3, p-AKT1/2/3 (Ser473), and Survivin were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, United States). IRDye^TM 800 conjugated secondary antibodies were obtained from Rockland Inc. (Philadelphia, PA, United States). FITC-conjugated CD13 antibody was bought from eBioscience (San Diego, CA, United States). GLI1 small interfering RNA (siRNA) and GANT61 was purchased from Santa Cruz Biotechnology. MTT (3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyl tetrazolium bromide) was purchased from Fluka Chemical Corp. (Ronkonkoma, NY, United States).

The human CML K562 cell line was purchased from the cell bank of Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences. The human stromal cell line HS-5 was purchased from the American Type Culture Collection (CRL-11882). Cells were cultured in RPMI-1640 medium (Gibco, Invitrogen Corporation, Carlsbad, CA, United States) supplemented with 10% fetal bovine serum (FBS) (Gibco), 100 U/ml of penicillin, and 100 U/ml of streptomycin in a humidified CO₂ (5%) incubator at 37°C.

For cell co-culture system, HS-5 cells (3 × 10⁵ /well) were first seeded in 6-well plates overnight. K562 cells were placed onto the porous trans-well inserts (0.4 μm, MILLIPORE, United States) which allowed them to share the soluble growth factors from HS-5 cell layers. All treatments were performed after co-culturing for 24 h.

Cell proliferation was evaluated with Ki67 Detection Kit (KeyGen Biotech, Nanjing, China) according to the manufacturer’s instructions. The photograph was taken under light microscope (Nikon Instruments, Lewisville, TX, United States). Ki67 index was the proportion of positive staining cells in total cells.

Anchorage-independent cell growth in vitro was assessed by Soft Agar Colony Formation Assay. Briefly, cells were collected and seeded in six-well plates at 15,000 cells/well in 2×RPMI-1640 medium containing 0.8% agar (Oxoid, Basingstoke, United Kingdom) and 2 × FBS over a 1.2% agar layer. The plates were maintained for 21 days with humidified CO₂ (5%) at 37°C. The colonies were photographed. For each group, all formed colonies were counted out under inverted microscope equipped with a color camera (Nikon Instruments). The percentage was calculated when set control group as 100.

Cell viability was measured by MTT assay as previously reported (Wang Y. et al., 2014). The inhibition ratio (%) was calculated as [(A_control — A_treated/A_control × 100]. A_treated and A_control are the average absorbance of three parallel wells from each group.

DAPI staining was employed to detect cell apoptosis. Briefly, cells were fixed with 4% paraformaldehyde (PFA) for 30 min, washed thrice with cold PBS for 5 min and incubated with 0.1%Triton X-100 for 10 min, followed by staining with DAPI (1 mg/ml) for 15 min. The pictures were taken by fluorescence microscopy (Olympus Corporation, Tokyo, Japan).

Cell apoptosis was also evaluated with Annexin V/PI Cell Apoptosis Detection Kit (KeyGen Biotech) according to the manufacturer’s instructions. The stained cells were subjected to FACSCalibur flow cytometer (BD, San Jose, CA, United States) equipped with Cell-Quest software. The early apoptotic cell was Annexin V positive and PI negative, whereas late apoptotic one was both Annexin V and PI positive.

Cells were collected and washed with PBS. Fractionation of nuclear and cytoplasmic protein was performed with Nuclear/Cytosol Fractionation Kit (BioVision, Mountain View, CA, United States) following the manufacturer’s protocol. Then the nuclear extract was subjected to western blot analysis.

The Western blot was conducted as described previously (Chen et al., 2013). All data were analyzed by Odyssey Scanning System equipped with densitometry detection software (LI-COR, Inc., Lincoln, NE, United States).

Total RNA was extracted by Trizol reagent (Invitrogen, Carlsbad, CA, United States). One microgram of RNA was reverse transcribed using cDNA synthesis kit (TaKaRa, Dalian, China) according to the manufacturer’s instruction. Real-time PCR was conducted using Applied Biosystems 7500 System (Carlsbad, CA, United States) by combining FastStart Universal SYBR Green Master (Roche) with specific primers as follows: GLI1, forward primer, 5′-TTCCTACCAGAGTCCCAAGT-3′, and reverse primer, 5′-CCCTATGTGAAGCCCTATTT-3′; BCR-ABL, forward primer, 5′-TGACCAACTCGTGTGTGAAACTC-3′; and reverse primer, 5′-TCCACTTCGTCTGAGATACTGGATT-3′; GAPDH, forward primer, 5′-CCACCCATGGCAAATTCCATGGCT-3′, and reverse primer, 5′-TCTAGACGGCAGGTCAGGTCCACC-3′. Data were analyzed according to the comparative Ct method, and the expression of target gene was normalized to GAPDH expression levels in each sample.

K562 cells were plated on 6-well plates overnight. Specific siRNA against GLI1 gene was transiently transfected into cells at a final concentration of 100 nmol/L using PepMute siRNA & DNA Transfection Reagent (SignaGen, Ijamsville, MD, United States) according to the manufacturer’s instructions. After 24 h transfection, gene silencing was measured by real-time PCR and western blot analysis.

K562 cells were collected and applied to the glass coverslips after treatment, and then were subjected to immunofluorescence as described previously (Zhao et al., 2015).

Animal studies were carried out in accordance with the regulations of the China Food and Drug Administration (CFDA) of China on Animal Care. This study was reviewed and approved by Animal Care Use Committee of Jiangsu Province (Animal authorization reference number: SYXK2012-0035). Twenty female NOD/SCID immune-deficient mice (aged 6–9 weeks) from Shanghai SLAC Laboratory Animal Company Limited were raised in air-conditioned pathogen-free rooms under controlled lighting (12 h per day) and fed with sterilized laboratory food and water. K562 cells (5 × 10⁶) were injected into mouse via tail vein (16 mice in total). After 1 week, the mice inoculated with K562 were randomly divided into four groups: (a) untreated group as a negative control; (b) OAG monotherapy (30 mg/kg); (c) IM monotherapy (200 mg/kg); (d) OAG combined with IM. The non-injected mice (4 mice) were set as blank control group. OAG was given intravenously once every 3 days, and IM was administered orally once every day. Mice were sacrificed 30 days later and BM, peripheral blood (PB), and spleens were collected. Spleen coefficient was calculated as (W_spleen/W_body) × 100. W_spleen and W_body is spleen weight and body weight of mouse, respectively. The expression of CD13 was examined by flow cytometry (BD) to evaluate K562 cell proliferation in vivo.

Protein expression in spleen was assessed using specific antibodies together with Ultra-Sensitive^TM SP kit (Maixin-Bio, Fuzhou, China) as described previously (Wang H. et al., 2014).

Statistical analysis of multiple group comparisons was performed by one-way analysis of variance (ANOVA) followed by the Bonferroni post hoc test. Comparisons between two groups were analyzed using the two-tailed Student’s t-test. P-value < 0.05 was considered statistically significant.

K562 cells were co-cultured with HS-5 cells as described in part of Section “Materials and Methods.” Then, cell growth and apoptosis of K562 cells was examined in the presence of IM. As shown in Figure 1A, IM could suppress K562 cells growth in a dose-dependent manner, whereas co-cultured cells seemed to be more refractory to IM treatment compared with regular culture as evaluated by Ki67 staining. In addition, the clonogenic assay was carried out to determine the ability of K562 cells to divide and form colonies. As a result, co-cultured K562 cells formed more colonies following IM treatment compared to regular-cultured cells (Figure 1B and Supplementary Figure S3A). Consistently, we observed that nuclear condensation (Figure 1C) and apoptosis events (Figure 1D and Supplementary Figure S3B) of K562 cells induced by IM were significantly attenuated due to co-culturing with HS-5 cells. These data suggested not only could co-culturing promote K562 cell proliferation, but also protect it from apoptosis caused by IM, which at least in part contributed to the acquirement of drug resistance.

To rule out the role of Hh signaling pathway in co-culture-mediated IM resistance, we examined the expression of important proteins of this pathway. As shown in Figures 2A–C, enhanced expression of Shh and GLI1 in whole lysate and nuclear extract was observed in co-cultured K562 cells, which was not affected by the addition of IM. Interestingly, BCR-ABL and its downstream targets such as PI3K, p-AKT, p-BAD, Bcl-xL and Survivin, were also up-regulated in co-culture cells with or without IM treatment, compared to that of regular culture cells (Figure 2A). These data indicated that Hh signaling pathway could be activated by cytokines from HS-5 cells, accompanied with IM resistance. As expected, nuclear translocation of GLI1 in co-cultured K562 cells was blocked by GANT-61 (Hyman et al., 2009), which was a small-molecule inhibitor of GLI1-mediated transcription (Figure 2D). Moreover, silencing of GLI1 or pharmacologic inhibition using GANT-61 was accompanied with BCR-ABL down-regulation at both mRNA and protein level in co-culture system (Figures 2E–G). Together, these results suggested that Hh signaling activation subsequently resulted in aberrant upregulation of BCR-ABL and downstream targets in the co-culture model, which might contribute to IM resistance of K562 cells.

Oroxyloside A was synthesized as described in Section “Materials and Methods” part, and the route was shown in Figure 3A. We firstly evaluated the toxicity of OAG on K562 cells by MTT assay and Annexin V/PI double-staining assay. As a result, OAG exhibited very low cytotoxic effect on cells (Figure 3B), and we choose the concentration of 40, 80, and 160 μM for the following experiments since no severe apoptosis events occurred, especially in co-culture system(Figure 3C and Supplementary Figure S4A).

To assess the synergy effect of OAG with IM in co-culture system, we examined the growth and apoptosis of K562 cells treated with IM alone or combined with OAG. As shown in Figures 3D,E, the combination of OAG with IM significantly decreased the percentage of colonies to 13.56 ± 4.13%, compared to that of K562 cells treated IM alone (67.80 ± 5.93%) under the condition of co-culturing with HS-5 cells. In addition, OAG synergized with IM to cause K562 cell apoptosis in co-culture model as evaluated by DAPI and Annexin V/PI staining (Figures 3F,G and Supplemental Figure S4B). Collectively, the weakly toxic concentrations of OAG could, at least in part, overcome BM microenvironment-induced K562 cell resistance to IM.

As mentioned above, Hh pathway activation and subsequent BCR-ABL up-regulation could be involved in the IM resistance of CML, thus we ask whether OAG could affect Hh signaling in co-culture system. Shh, as a Hh ligand, could activate signaling pathway through paracrine mechanism in CML (Long et al., 2011). Figure 4A showed that the amount of Shh was decreased from 1.94 ± 0.02 to 1.00 ± 0.01 ng/ml by OAG in a dose-dependent manner. Besides, both GLI1 and BCR-ABL mRNA expression was promoted under co-culturing condition, whereas OAG treatment suppressed their up-regulation (Figure 4B). Consistently, treatment with OAG resulted in protein down-regulation such as Shh, Shh-N, GLI1, and BCR-ABL, as well as GLI1 nuclear export in co-cultured cells (Figures 4C–E). Moreover, the addition of OAG combined with IM not only down-regulated the level of BCR-ABL as well as its related proteins, but it also induced cleaved caspases, a well-known events for cell apoptosis, in co-cultured K562 cells (Figures 4F,G). These data draw a possible mechanism that OAG could suppress the activation of Hh/BCR-ABL axis, thereby facilitating IM to cause cell death in drug-resistant CML cells.

To further support that OAG sensitized CML cells to IM in vivo, we inoculated K562 cells into female NOD/SCID mice by intravenous injection as described in Section “Materials and Methods” part. Mice with graft were treated with or without IM (200 mg/kg), together with or without OAG (30 mg/kg). All animals were raised for 4 weeks, and their body weight was monitored every 3 days. As shown in Figure 5A, mice transplanted K562 cells showed decreased body weight with days. However, there was no further reduction of body weight in mice treated with drugs compared to those with vehicle treatment, suggesting that drug itself could be safe for chemotherapy. As expected, the amount of white blood cells (WBC) was significantly increased from 1.26 ± 0.06 to 4.68 ± 0.25 × 10⁹/L due to K562 cell transplantation, whereas it was notably decreased to 1.57 ± 0.11 × 10⁹/L by the treatment in combination of IM and OAG (Figure 5B). In addition, spleens were weighed to calculate the spleen coefficient. K562 cell graft significantly raised spleen coefficient to 0.68 ± 0.05, which was back to normal range in the presence of IM combined with OAG (Figure 5C). Next, the expression of CD13, a surface marker of K562 cells, was analyzed by FACS. As shown in Supplementary Figure S5 and Figure 5D, grafted mice treated with OAG and IM together showed significant reduction in CD13 positive cells, suggesting that the combination treatment could eliminate most of K562 cells in vivo.

Consistent with in vitro data, K562 cell inoculation elevated blood Shh for about twofold, whereas OAG alone and its combination with IM could diminish the level of Shh in blood of mice with K562 cell graft (Figure 6A). Furthermore, the relative mRNA levels of GLI1 and BCR-ABL in PB and BM were significantly inhibited by combination treatment in inoculated mice (Figures 6B,C). In addition, the expression of Shh, Shh-N, GLI1, and BCR-ABL in spleens of control group was up-regulated compared with non-inoculated group, which was significantly suppressed by combination treatment (Figures 6D–F). Taken together, the results above suggested that the synergistic effect of IM and OAG against CML cells in vivo could be attributed to suppressing Hh pathway and BCR-ABL expression.