Research ArticleCancer

Homoharringtonine (omacetaxine mepesuccinate) as an adjunct for FLT3-ITD acute myeloid leukemia

See allHide authors and affiliations

Science Translational Medicine  05 Oct 2016:
Vol. 8, Issue 359, pp. 359ra129
DOI: 10.1126/scitranslmed.aaf3735

More than a FLT-ing success in leukemia

Acute myeloid leukemia is a difficult disease to treat under the best of circumstances, and the subtype containing internal tandem duplication of fms-like tyrosine kinase 3 (FLT3-ITD) tends to be particularly challenging. Lam et al. performed a high-throughput drug screen and identified homoharringtonine as a candidate treatment for this type of leukemia and then confirmed its effectiveness in cancer cells, in mouse models, and in patients. The treatment showed promising results in a phase 2 clinical trial, which included elderly patients and those who have failed all previous treatments, paving the way for further development of this drug.


An in vitro drug-screening platform on patient samples was developed and validated to design personalized treatment for relapsed/refractory acute myeloid leukemia (AML). Unbiased clustering and correlation showed that homoharringtonine (HHT), also known as omacetaxine mepesuccinate, exhibited preferential antileukemia effect against AML carrying internal tandem duplication of fms-like tyrosine kinase 3 (FLT3-ITD). It worked synergistically with FLT3 inhibitors to suppress leukemia growth in vitro and in xenograft mouse models. Mechanistically, the effect was mediated by protein synthesis inhibition and reduction of short-lived proteins, including total and phosphorylated forms of FLT3 and its downstream signaling proteins. A phase 2 clinical trial of sorafenib and HHT combination treatment in FLT3-ITD AML patients resulted in complete remission (true or with insufficient hematological recovery) in 20 of 24 patients (83.3%), reduction of ITD allelic burden, and median leukemia-free and overall survivals of 12 and 33 weeks. The regimen has successfully bridged five patients to allogeneic hematopoietic stem cell transplantation and was well tolerated in patients unfit for conventional chemotherapy, including elderly and heavily pretreated patients. This study validated the principle and clinical relevance of in vitro drug testing and identified an improved treatment for FLT3-ITD AML. The results provided the foundation for phase 2/3 clinical trials to ascertain the clinical efficacy of FLT3 inhibitors and HHT in combination.


Acute myeloid leukemia (AML) is a group of heterogeneous diseases with distinct clinicopathologic, cytogenetic, and genetic features, whose common feature is an abnormal increase in myeloblasts. Induction chemotherapy followed by consolidation or allogeneic hematopoietic stem cell transplantation (HSCT) are the standard therapeutic approaches. These methods are based on a series of randomized control trials performed since the early 70s, when groups of patients were treated uniformly according to trial protocols. However, the outcome has been unsatisfactory, and despite improvement of supportive care and antimicrobial agents, long-term remission is only achieved in 30 to 40% of patients. Most patients died of disease progression or toxicity of repeated but futile chemotherapy (1).

Disease heterogeneity in malignancies implies that different strategies have to be adopted for different patients. In this regard, personalized medicine has been proposed to target genetic lesions unique to specific diseases (2). To look for new therapeutic targets of specific diseases, laboratory models consisting of high-throughput drug screening, genomics, and gene expression analyses have been reported in cancer cell lines (35). Recently, in vitro drug screening that guided treatment for individual patients (6, 7) in AML has been reported, underscoring the possibility of personalizing treatment for this disease.

FLT3-ITD (fms-like tyrosine kinase 3–internal tandem duplication) occurs in nearly 30% of AML and is associated with high relapse rates and hence inferior disease-free and overall survivals (OSs). The pathogenetic mechanisms of FLT3-ITD have been extensively reviewed (8, 9). Multikinase inhibitors targeting FLT3, including sorafenib, midostaurin, and quizartinib (formerly AC220), have been evaluated. Responses to monotherapy were typically short-lived, rendering it difficult to bridge patients to allogeneic HSCT. Combinations of conventional chemotherapy with FLT3 inhibitors are also being evaluated in phase 2/3 clinical trials (10). A potential caveat of this approach is the repeated exposure of patients to genotoxic chemotherapy that might enhance emergence of resistant leukemic clones (11). Azacitidine consolidation after sorafenib monotherapy increased the duration of response, but the benefit was seen predominantly in patients with myelodysplastic syndrome that has transformed into AML and acquired FLT3-ITD (12).

In the development of an in vitro drug-screening platform in this study, the antileukemia profile of homoharringtonine (HHT), a protein translation inhibitor also known as omacetaxine mepesuccinate, was found to correlate with that of FLT3 inhibitors and suppress FLT3-ITD AML preferentially. Synergism was demonstrated between HHT and FLT3 inhibitors in vitro and in vivo. A phase 2 clinical study based on these findings demonstrated clinical efficacy of HHT and sorafenib combination (S + HHT) in relapsed or refractory FLT3-ITD AML. These results served to validate the principle and clinical relevance of in vitro drug screening, in addition to identifying HHT as an important partner with FLT3 inhibition in the treatment of FLT3-ITD AML.


Culture systems for primary AML samples

Primary AML myeloblasts cultured in Iscove’s modified Dulbecco’s medium (IMDM) containing 10% fetal bovine serum (IMDM10) showed a significant decrease in cell number (P < 0.0001) and increase in apoptosis (P < 0.0001) on day 3 compared to input (Fig. 1, A and B). Two other culture media were examined, including IMDM10-containing medium conditioned by murine stromal cell lines [M2-10B4 and Sl/Sl fibroblasts, as previously described (13)] that have been engineered to produce recombinant human cytokines [stem cell factor (SCF), granulocyte colony-stimulating factor (G-CSF), interleukin-3 (IL-3), and FLT3 ligand] [conditioned medium (CM)] and IMDM containing 10% patient-derived autologous human serum (HS). Myeloblasts were better at maintaining cell number in CM (P < 0.0001) and exhibited significantly less apoptosis in both HS and CM (P < 0.0001) (Fig. 1, A and B). More myeloblasts were found in S and G2/M phases and fewer in the G0/G1 phase in CM compared with IMDM10 or HS, indicating more proliferation (Fig. 1C). Differentiation of postculture myeloblasts was evidenced in some samples by the decrease in CD34+ population (fig. S1A) and the increase in cytoplasm and presence of cytoplasmic vacuolation and perinuclear halo (fig. S1B) in all systems.

Fig. 1. In vitro drug-screening platform for primary AML.

Primary samples were cultured in vitro for 3 days in IMDM10 (IM), CM, and autologous HS (10%). (A) Fold change in viable cell numbers compared to day 0 (annexin V/7AAD). (B) Percentages of annexin V+ population. For (A) and (B): IMDM10, N = 36; CM, N = 35; HS, N = 24. (C) Percentages of cells in G0/G1 and S/G2/M phases are shown. IMDM10, N = 9; CM, N = 21; HS, N = 16. Cryopreserved samples were used in these experiments. The boxes extended from the 25th to the 75th percentiles and the whiskers from minimum to maximum values. (D) Ninety-six primary AML samples screened (CM, N = 93; HS, N = 32) against 25 drugs were categorized as minimal, lethal, and variable, defined by ≥50% inhibitory effect in <10, ≥95, and 10 to 94% samples, respectively, represented by axitinib, HHT, and sorafenib (from left to right). (E) A representative dot plot showing the correlation of antileukemia effect between drugs. Coefficient (r) and P value of Pearson correlation are shown. (F and G) Correlations among (F) JAK and (G) FLT3 inhibitors in CM (upper half) and HS (lower half). (H) Correlation of dasatinib with lestaurtinib and vorinostat in CM. (I) Correlation of azacitidine with ponatinib and TG101209 in CM. (J) List of drugs in descending order of r for their correlation with HHT, compared to the correlation with AraC. The “+” symbol indicates reported inhibitory activity against the FLT3 receptor. (K) Representative graphs for (J). Each dot and line represent one patient sample. All r not reaching statistical significance (P ≥ 0.05) in (F), (G) and (J) were italicized, whereas r ≥0.5 in both culture systems were bolded.

Drug-response profile of primary AML samples

Ninety-six primary AML samples were obtained from 81 patients at diagnosis (n = 56) or relapse (n = 40), 50 being fresh and 46 cryopreserved. They were screened with a panel of 25 drugs (table S1) in CM (N = 93) (fig. S2) and HS (N = 32) (fig. S3). For an individual leukemia sample, drugs that inhibited leukemia growth by ≥50% on day 3 at the doses tested as compared with vehicle [0.1% dimethyl sulfoxide (DMSO)] were considered effective. For the whole batch of samples, drugs effective in less than 10% of the samples were defined as “minimal” and those effective in more than or equal to 95% samples were defined as “lethal.” Drugs that showed variable effectiveness in 10 to 94% samples were defined as “variable” (Fig. 1D). The entire spectrum of ~3125 drug-sample-condition data points was shown in fig. S4, where the area under the curve derived from dose-response curves in figs. S2 and S3 represented the growth suppression effect of each drug on each sample. The choice of culture medium could alter the dose-response dynamics, as evidenced for instance by the stronger antileukemia effect of JAK (Janus kinase) inhibitors (ruxolitinib, tofacitinib, TG101209, and lestaurtinib) observed in CM than in HS. Despite this, relative responses among samples were significantly correlated between CM and HS systems in 20 drugs (range of Pearson’s r = 0.371 to 0.805, P < 0.05) (fig. S5). For the remaining five drugs (vorinostat, rapamycin, axitinib, pazopanib, and arsenic trioxide), there was no significant correlation.

The results were further validated by the observation that drugs with similar mechanisms of action showed comparable effects on leukemia growth inhibition in both CM and HS, as exemplified by JAK and FLT3 inhibitors (Fig. 1, E to G). Although the effects of epidermal growth factor receptor, vascular endothelial growth factor receptor, and BCR-ABL1 inhibitors on leukemia growth were modest (figs. S2 and S3), significant correlation between their antileukemia effects was also observed in both culture systems (P < 0.05 except for nilotinib versus ponatinib in CM) (fig. S6, A to C). Significant correlation of antileukemia effects between some drugs with apparently distinct targets was also observed, including that of dasatinib with lestaurtinib and with vorinostat (CM, both P < 0.0001, Fig. 1H; HS, both P < 0.0001, fig. S6D); azacitidine with ponatinib and with TG101209 (CM, both P < 0.0001, Fig. 1I; HS, P = 0.0009 and P < 0.0001, fig. S6E); as well as HHT with FLT3 inhibitors (P < 0.0001 for all except sorafenib in HS, P = 0.0030) (Fig. 1, J and K). On the contrary, cytarabine (AraC), a prototype drug used in the treatment of AML, exhibited little correlation with the tyrosine kinase inhibitors tested in the panel.

To analyze the drug-response patterns of the 96 AML samples, we performed unsupervised hierarchical clustering on the basis of the samples’ drug sensitivity profiles in both culture systems. Four distinct clusters of drugs in CM (groups A to D) and three in HS (groups E to G) were identified on the basis of distinct patterns of antileukemia effects (Fig. 2A and fig. S7). In both culture systems, bortezomib and AraC were clustered together (groups A and E). Similarly, JAK inhibitors ruxolitinib and tofacitinib (groups C and F), as well as FLT3 inhibitors (sorafenib and ponatinib in both CM and HS; quizartinib and lestaurtinib in HS) and HHT (groups D and G) were also clustered. TG101209, a JAK inhibitor that also targets FLT3, was clustered with HHT and FLT3 inhibitors (group D). The relative drug responses within each cluster were examined in detail and correlated with the clinicopathologic characteristics of the samples (figs. S8 and S9). The source [peripheral blood (PB) or bone marrow (BM)], cryopreservation, disease stage (diagnostic or relapse), or karyotype of AML samples had no significant impact on the sample clustering.

Fig. 2. Drug-response clustering analysis.

(A) Unsupervised hierarchical clustering of drugs classified as variable (V) and lethal (L) on the basis of their antileukemia effect into subgroups. (B) Comparison of response to sorafenib [wild-type (WT), N = 46; ITD, N = 60] (left) and other drugs (WT, N = 46; ITD, N = 47) (right) between FLT3-WT and FLT3-ITD samples in CM. The difference in mean Z score (ΔZ, ITD minus WT) and P values calculated by Mann-Whitney U test are shown. Note that additional samples were included in experiments with sorafenib so that the total number of samples exceeded 93. (C) Clustering of 93 AML samples based on their responses to cluster D drugs in CM into four classes (I to IV). Each row represents a patient sample and each column represents a drug. Negative Z score indicates higher sensitivity. FLT3: WT or ITD mutation. (D) Frequency of mutations for the 15 most frequently mutated (≥5 samples) genes in 93 samples detected by Illumina MiSeq targeted sequencing on a panel of 54 genes. (E) Gene-drug associations represented by a volcano plot showing P values (one-way ANOVA) and effects represented by difference in Z scores between mutant and WT groups. The horizontal line represents the cutoff P value at a false discovery rate of 20% (P < 0.05, Benjamini-Hochberg procedure). Sizes of the dots represent the number of mutated samples. (F and G) Comparison of response to (F) ponatinib in FLT3-ITD–mutated and/or NPM1-mutated samples and to (G) HHT in FLT3-ITD–mutated and/or DNMT3A–mutated samples.

Clinical relevance of in vitro drug response

Predictably, tyrosine kinase inhibitors with activity against FLT3, including ponatinib, sorafenib, lestaurtinib, and quizartinib, were more effective in FLT3-ITD than FLT3-WT myeloblasts (Fig. 2B and table S2), corroborating the clinical observation that FLT3 inhibitors were generally more effective in FLT3-ITD AML (14, 15). HHT was also more effective in FLT3-ITD than FLT3-WT samples (Fig. 2B, right panel), consistent with the clustering analysis results. The effects of drugs clustered in group D (sorafenib, HHT, TG101209, ponatinib, and azacitidine) in CM, encompassing FLT3 inhibitors and HHT, on AML were examined in detail (Fig. 2C). In decreasing order of in vitro drug sensitivities, primary samples were classified into four distinct classes (I to IV). The classification was correlated with the prevalence of FLT3-ITD, being 8 of 8 (100%) in class I, 24 of 35 (68.6%) in class II, 13 of 40 (32.5%) in class III, and 1 of 10 (10%) in class IV [two-tailed Fisher’s exact test for any two classes, P < 0.005, except for I versus II (P = 0.090) and III versus IV (P = 0.246)].

Extended molecular profiling of samples

Mutational analysis of 54 myeloid genes based on an Illumina MiSeq myeloid panel was performed on 93 of 96 primary samples. Patterns of gene mutations were similar to those reported by recent large-scale studies (1618), with FLT3, NPM1, and DNMT3A being the most frequently mutated genes (Fig. 2D). The higher frequencies of FLT3-ITD, NPM1, and DNMT3A mutations in this cohort reflected the mutation spectrum among our archived samples, which were generally more abundant in and might contain an overrepresentation of highly proliferative AML subtypes such as FLT3-ITD–mutated AML. Only those mutations with variant allele frequency ≥ 20% were included to ensure their abundance and relevance in the samples. Correlations between single-gene mutations and drug sensitivity were performed using analysis of variance (ANOVA) and represented by a volcano plot (Fig. 2E). In addition to the correlation between FLT3 inhibitors (sorafenib and lestaurtinib) and HHT in FLT3-ITD, NPM1, RUNX1, and DNMT3A mutations were also correlated with enhanced response to ponatinib, dasatinib, and HHT, respectively (fig. S10, A to C). On the other hand, AML with TP53, ASXL1, and NRAS mutations were relatively insensitive to AraC, HHT, and ruxolitinib, respectively (Fig. 2E). When concurrent mutations of FLT3-ITD, NPM1, and DNMT3A were considered by two-way ANOVA, the responses of FLT3-ITD AML to ponatinib (Fig. 2F) and HHT (Fig. 2G) were more apparent in the presence of concurrent NPM1 and DNMT3A mutations. JAK inhibitors, including tofacitinib, ruxolitinib, and TG101209, also showed antileukemia effects in FLT3-ITD AML only in the presence of concurrent DNMT3A mutations (fig. S10, D to F). The complete drug-response data for 96 samples with 25 drugs at different doses and the molecular profile data set are available in table S3.

Preferential effect of HHT in FLT3-ITD AML

The observation that HHT clustered with FLT3 inhibitors and its effects correlated with the presence of FLT3-ITD mutation prompted us to investigate the possibility of its application in FLT3-ITD AML. The preferential effects of HHT on FLT3-ITD AML were tested in an isogenic system using a murine B-lymphoid Ba/F3 cell line lentivirally transduced with green fluorescent protein (GFP), FLT3-WT, and FLT3-ITD. FLT3-ITD Ba/F3 cells were IL-3–independent and addicted to constitutively active FLT3 signaling, and hence, sensitive to sorafenib (Fig. 3A, left panel). They were modestly but significantly more sensitive to the growth inhibitory effect of HHT than Ba/F3-GFP or Ba/F3-WT, both of which were dependent on IL-3 rather than FLT3 signaling (P < 0.01 at 7.5 and 10 nM; Fig. 3A, middle panel). The respective median inhibitory concentration of HHT in Ba/F3-GFP, Ba/F3-WT, and Ba/F3-ITD cells were 11.9, 13.9, and 6.73 nM. The increased sensitivity of Ba/F3-ITD to HHT was not accounted for by a general sensitivity to cytotoxic agents, because AraC was equally inhibitory to all three lines (Fig. 3A, right panel), or any difference in proliferation rates, which were similar in all three lines (Fig. 3B). When FLT3-ITD Ba/F3 cells were grown in the presence of murine IL-3, they became addicted to IL-3 and less sensitive to sorafenib or HHT, attesting to the specificity of HHT toward FLT3-ITD signaling (Fig. 3A, left and middle panels). Mechanistically, both sorafenib and HHT inhibited phosphorylated extracellular signal–regulated kinase (pERK) signaling in FLT3-ITD Ba/F3 cells only in the absence of IL-3, when Ba/F3 became dependent on FLT3 signaling (fig. S11A). These observations supported the proposition that HHT was effective preferentially in FLT3-ITD AML and corroborated the findings in AML cell lines and primary myeloblasts.

Fig. 3. Preferential effect of HHT and its synergism with sorafenib in FLT3-ITD AML.

(A) Dose-response curves of Ba/F3-GFP, Ba/F3-WT, and Ba/F3-ITD cells treated with sorafenib, HHT, and AraC. Ba/F3-GFP and Ba/F3-WT cells were cultured with murine IL-3 (mIL-3, 2 ng/ml), and Ba/F3-ITD cells were cultured either with or without mIL-3. n = 4 to 5. (B) Growth kinetics of the Ba/F3 lines in (A). (C) Response of FLT3-WT (N = 46) and ITD (N = 47) primary AML samples to HHT and AraC in CM. Each dot represents one patient sample. The horizontal bars indicate the mean of each group. (D) Dose-response curves of eight AML and one CML blastic phase (K562) cell lines to 3-day in vitro treatment with HHT, CHX, or AraC; n = 3 to 7. (E) Extent of apoptosis induced in five AML cell lines after 24-hour treatment with HHT; n = 3 to 5. (F) Extent of apoptosis induced in MV4-11 after 24-hour treatment with sorafenib and/or HHT; n = 3. Two-tailed paired t test comparing vehicle/single agent to combination treatment. Data are shown as means ± SEM in (A) and (D) to (F). (G) Left: Cell viability of MV4-11 treated in vitro with HHT and/or sorafenib (n = 3) compared to vehicle, measured by PrestoBlue assay. Synergism was shown by positive EOBA (middle) and combination indices < 1 at molar ratio (sorafenib/HHT) of 1:1.90 (right).

Primary FLT3-ITD AML samples were also preferentially more sensitive to HHT (CM, Fig. 3C; HS, fig. S11B). In contrast, such a trend was not observed with other drugs, such as AraC, bortezomib, and non–FLT3-targeting tyrosine kinase inhibitors (Fig. 3C and fig. S11, C and D). Chemosensitivity profiling was also performed in six AML and one CML blast phase cell lines (table S4). FLT3-ITD cell lines MV4-11 and MOLM-13 were more sensitive to FLT3 inhibitors (sorafenib and quizartinib), and K562 was more sensitive to BCR-ABL inhibitors (nilotinib and dasatinib). Furthermore, HHT preferentially inhibited the growth of FLT3-ITD cell lines, whereas AraC showed no differential effect (Fig. 3D, left and right panels). The FLT3-ITD cell lines were also among the cell lines most sensitive to cycloheximide (CHX), another protein translation inhibitor (Fig. 3D, middle panel).

HHT induced significant apoptosis in MV4-11, MOLM-13, and Kasumi-1 (P < 0.05), but not in KG-1 and K562 at nanomolar range (Fig. 3E and fig. S11E, AraC). It also accentuated the apoptotic effects of sorafenib in MV4-11 (Fig. 3F) and MOLM-13 (fig. S11F). HHT synergized with sorafenib and other FLT3 inhibitors in suppressing leukemia growth in Ba/F3-ITD cells and FLT3-ITD AML cell lines (Fig. 3G and fig. S12, A to C), as shown by positive excess over bliss additivism (EOBA) (19) and combination indices <1 (20). Conversely, no significant synergism between sorafenib and HHT (S + HHT) could be demonstrated in the FLT3-ITD–negative cell line Kasumi-1, KG-1, or K562 (fig. S12D). Among the 37 AML samples (18 WT and 19 ITD) treated in vitro with the combination of HHT and sorafenib at fixed doses in CM, FLT3-ITD AML samples were more sensitive to growth inhibition (fig. S13, A and B).

HHT targeting short-lived proteins and high protein synthesis rate

Mechanistically, HHT has been shown to inhibit formation of the first peptide bond in polypeptide synthesis (21, 22). To explain the preferential effect of protein translation inhibitors on FLT3-ITD AML, protein synthesis rates in FLT3-WT and ITD AML were correlated with their susceptibility to HHT inhibition. Relative protein synthesis rates were measured by O-propargyl-puromycin (OP-Puro) incorporation assay (23) (Fig. 4A). Treatment with CHX (at 100 μg/ml) for 30 min was used as the positive control. Ba/F3-ITD cells showed a significantly higher rate of protein synthesis compared with parental Ba/F3 cells (P = 0.0295; Fig. 4B). Furthermore, MV4-11 and MOLM-13 were among the cell lines with the highest basal protein synthesis rates (Fig. 4C). HHT also exhibited a greater extent of protein synthesis inhibition in MV4-11 (sensitive cell line) than in KG-1 and K562 (relatively insensitive cell lines) after treatment for 24 hours (Fig. 4D). The amount of phosphorylated 4E-BP1 (p4E-BP1) was also used as a surrogate to indicate active protein synthesis inducible by mammalian target of rapamycin (mTOR) activation (24). Phosphorylation of 4E-BP1 removes its inhibition on eIF4E, favoring protein synthesis initiation. Both FLT3-ITD cell lines (MV4-11 and MOLM-13) (Fig. 4E) and FLT3-ITD primary AML samples (Fig. 4, F and G) had high amounts of p4E-BP1 protein.

Fig. 4. Mechanism of action of HHT in FLT3-ITD AML.

(A) Histograms showing the fluorescence signal of OP-Puro conjugated with Alexa Fluor 488 (OP-Puro-AF488) incorporated in the MV4-11 cell line after 30-min treatment with vehicle (0.1% DMSO), CHX (100 μg/ml), and HHT (100 nM and 1 μM). The mean fluorescence intensity ratio reflected protein synthesis rate. (B and C) Relative basal protein synthesis rates of (B) Ba/F3 and Ba/F3-ITD; n = 3 and (C) different cell lines, each normalized to MV4-11; n = 4 to 7. (D) Changes in protein synthesis rates after 30-min and 24-hour treatment with HHT in cell lines with high (MV4-11) and low (KG-1 and K562) protein synthesis rates relative to vehicle (0.1% DMSO); n = 3. (E and F) Western blots showing amounts of p4E-BP1 among (E) different cell lines and (F) representative primary AML samples, as an indicator of mTOR complex 1 activation. (G) Relative p4E-BP1 band intensities in FLT3-WT (N = 13) and FLT3-ITD (N = 13) cryopreserved AML samples, each normalized to its actin content. All samples were normalized to sample 367.4 for each blot. The horizontal bars indicate the mean. Mann-Whitney U test was used to calculate the P value. (H and I) Western blots showing amounts of phosphorylated FLT3 (pFLT3) and its downstream signaling proteins in (H) MV4-11 treated with vehicle, HHT (100 nM), or CHX (100 μg/ml) for 2, 4, and 6 hours and (I) MV4-11 treated with vehicle, sorafenib (Sor), and/or HHT for 6 hours, with actin being the loading control.

Addiction to a high protein synthesis rate in specific AML subtypes might confer susceptibility to protein synthesis inhibitors through depletion of short-lived proteins. In MV4-11, amounts of FLT3 as well as other signaling proteins with short half-lives were reduced by HHT and CHX in a time-dependent manner (Fig. 4H). Sorafenib alone reduced the amount of pFLT3 protein and downstream molecules, including pSTAT5 (phosphorylated signal transducer and activator of transcription 5) and pERK1/2, without much effect on the amounts of total proteins in MV4-11 (Fig. 4I). HHT at a pharmacological dose [~66 nM (25)] reduced total FLT3 and pFLT3 protein amounts, as well as pSTAT5 and pERK in MV4-11. S + HHT combination could further reduce the amounts of these phosphorylated proteins. Similar results were observed by flow cytometric measurements of total and phosphorylated proteins (fig. S14).

The effect of HHT on sorafenib-resistant FLT3-ITD AML models

Sorafenib inhibited the growth of Ba/F3-ITD cells but was ineffective in FLT3-ITD–D835V, FLT3-ITD–D835F, and FLT3-ITD–F691L–expressing lines. HHT inhibited the growth of all these cell lines with similar efficacies (fig. S15A). A sorafenib-resistant MOLM-13 cell line (M13-S50) carrying the FLT3-TKD (tyrosine kinase domain) F691L (gatekeeper) mutation was also sensitive to HHT (fig. S15, B and C). On the other hand, HHT-resistant MOLM-13 cells (M13-H10) were sensitive to the antileukemia (fig. S15C) and FLT3-inhibitory effects of sorafenib (see Materials and Methods for details on the generation of these drug-resistant lines). For both resistant lines, the combination of sorafenib and HHT resulted in a greater inhibitory effect on leukemia growth (fig. S15D) and pSTAT5 protein (fig. S16).

Enhanced antileukemia effect of S + HHT in vivo

To confirm the effects of HHT and sorafenib in vivo, sublethally irradiated NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were transplanted with MV4-11 or MOLM-13 cells transduced with the luciferase reporter gene by tail vein injection and treated daily with vehicle, sorafenib (5 mg/kg per day), HHT (0.5 mg/kg per day), or the combination, starting on day 14 (MV4-11) or day 9 (MOLM-13) after injection. Mice treated with sorafenib and HHT showed little engraftment of MV4-11 up to 42 days, followed by a rebound on day 70 after transplantation (Fig. 5, A and B). Median survival of mice treated with combination S + HHT was better than those receiving vehicle, sorafenib, or HHT monotherapy, although only the comparison with the vehicle group reached statistical significance (Fig. 5C). Similarly, animal survival was improved with combination treatment of NSG transplanted with MOLM-13 (fig. S17). Effects of HHT on FLT3-ITD Ba/F3 engrafting into NSG mice have also been examined. HHT up to 4 mg/kg per day, a dose clearly inhibitory to FLT3-ITD cell lines MOLM-13 and MV4-11, had no significant effects on Ba/F3-ITD (fig. S17, D and E), suggesting that the cytokine milieu in NSG mice might confer the ability to engraft and grow independent of FLT3 to Ba/F3-ITD cells, resulting in resistance to HHT, similar to that observed in vitro.

Fig. 5. In vivo effect of sorafenib and HHT combination.

(A) Bioluminescence intensities and (B) images of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) mice transplanted with MV4-11 carrying a luciferase reporter gene before (day 14) and after treatment with vehicle (n = 7), HHT (n = 7), sorafenib (n = 8), and their combination (n = 8). (C) Survival curves for the same mice. Log-rank test was used to calculate significance in the differences between each curve and vehicle.

Superior outcome of S + HHT over sorafenib monotherapy in FLT3-ITD AML

A total of 24 FLT3-ITD AML patients were treated in a phase 2 trial of combination treatment with sorafenib and HHT (S + HHT), administered continuously until intolerance/disease progression or allogeneic HSCT. There were 8 men and 16 women, at a median age of 50 (21 to 76) years, who had relapsed/refractory leukemia (naïve to tyrosine kinase inhibitors, N = 17; relapsed from or refractory to sorafenib and/or ponatinib, N = 5) or were unsuitable for induction chemotherapy because of advanced age and comorbidities (N =2) (table S5). For 19 patients naïve to tyrosine kinase inhibitors, S + HHT resulted in complete remission (CR) in 2 cases (10.5%) and CR with incomplete hematologic recovery (CRi) in 13 cases (68.4%) after a median of 22 (18 to 55) days of treatment. Five of them were bridged to allogeneic HSCT. At a median follow-up of 7.1 (2.2 to 20.5) months, nine patients relapsed and four patients were still in remission (including three patients after HSCT). One 76-year-old woman with relapsed FLT3-ITD AML achieved CRi with S + HHT and remained in remission for 546 days, despite being taken off treatment after 161 days (five cycles) of HHT and 446 days of sorafenib because of episodes of community-acquired pneumonia. Among the remaining four patients who did not achieve CR/CRi, one showed a reduction of BM blasts from 16 to 6% after one course of S + HHT [nCRi (near CRi)] and was bridged to allogeneic HSCT. The other three patients succumbed to refractory disease. The five patients previously relapsed from or refractory to sorafenib and/or ponatinib monotherapy all achieved CR/CRi with S + HHT after a median of 22 (20 to 26) days of treatment. However, the median duration of CR/CRi was only 18 (9 to 76) days. The outcomes of the whole cohort are summarized in Fig. 6A and Table 1. Overall, 20 of 24 patients (83.3%) achieved CR/CRi at a median time of 22 (18 to 55) days, with median leukemia-free survival (LFS) and OS of 88 days (9 to 510 days) and 228 days (53 to 615 days), respectively. The median LFS and OS were 118 and 259 days when excluding patients with previous exposure to FLT3 inhibitors. In comparison with a historical cohort of FLT3-ITD patients treated with sorafenib monotherapy (12, 26) (table S6), time to best BM response was significantly shorter (P = 0.0235; Fig. 6B), and the LFS (P = 0.0002) and OS (P = 0.0327) were significantly longer (Fig. 6, C and D). For 19 evaluable and responding patients, FLT3-ITD allelic burden decreased from 73.0 ± 5.4% before S + HHT to 16.0 ± 5.8% (P < 0.0001) at first BM response (Fig. 6E), an extent significantly greater than that in the historical cohort after sorafenib monotherapy (P = 0.0044; Fig. 6F). Toxicity of the combination therapy was mainly hematological, whereas nonhematological toxicity was limited to rash and fever in association with sorafenib (Table 1). The combination treatment was well tolerated, with three patients beyond 65 years of age tolerating the treatment without major complications. The emergence of FLT3-TKD mutations at the gatekeeper site (F691) and activation domain (D835) at leukemia relapse was examined. Among the nine relapsed patients who were naïve to FLT3 inhibitors when treated with S + HHT, emergence of D835 mutation was identified in four patients (allelic frequencies: 72, 82, 5, and 48%). D835 mutation was also found in a pretreatment DNA sample of a responding patient (allelic frequency, 32.4%), who is still in remission (247 days). Moreover, two patients showed complete loss of FLT3-ITD clones at relapse.

Fig. 6. S + HHT phase 2 clinical trial.

(A) Flow chart showing the outcome of 24 FLT3-ITD AML patients in response to S + HHT. Four relapsed patients were re-treated with S + HHT and three achieved remission. (B) Time to achieve best BM response (N = 21) compared to historical cohorts (12, 26) treated with sorafenib monotherapy (S) as induction (N = 27, including patients who received sorafenib monotherapy as induction followed by consolidation with azacitidine). Boxes, 25th to 75th percentiles; whiskers, 10th to 90th percentiles; vertical lines, median. (C) LFS (N = 15, 6 censored) and (D) OS (N = 15, 7 censored) of patients receiving S + HHT (vertical dotted line in orange, median survival). Five patients with previous exposure to sorafenib/ponatinib were excluded. P value calculated by log-rank test comparing with S (vertical dotted line in blue, median survival; N = 14) is shown. (E) FLT3-ITD allelic burden (%) before treatment (P; N = 22), at remission (C; N = 20), and at relapse/loss of response (LR; N = 14) stages for patients receiving S + HHT. Two patients (orange) relapsed with loss of FLT3-ITD clones, and three (gray) did not respond. (F) Percentage decrease in ITD burden at remission in both cohorts. (G) PB smear of patients after S + HHT, showing dysplastic neutrophils with hypogranularity (top two panels), nuclear hypolobation (lower left), and abnormal nuclear configuration (right two panels). Scale bar, 10 μm. (B and F) Mann-Whitney U test. CR, complete remission; CRi, CR with insufficient hematological recovery; nCRi, near CRi (reduction of blasts without complete clearance); NR, nonremission; BMT, BM transplantation; TRM, treatment-related mortality.

Table 1. Treatment response of 24 FLT3-ITD patients in the phase 2 clinical trial of S + HHT combination treatment.

NA, not available; Nil, not done; HFSR, hand-foot-skin reaction.

View this table:

Morphologic features of PB and BM at CR/CRi

The PB of patients achieving CRi after S + HHT showed dysplasia in granulocytic lineage, mainly in the form of hypogranularity and abnormal nuclear configuration (Fig. 6G), observed 1 to 2 weeks after treatment in 9 of 16 patients (56.3%) assessed. BM assessment in 14 of 19 patients in CR/CRi revealed a moderate to markedly hypocellular marrow with stromal changes and relative predominance of plasma cells (fig. S18). This was different from the normal and often increased cellularity observed in FLT3-ITD patients at remission induced by sorafenib monotherapy (26).

Correlation between clinical and in vitro drug response in patients treated by S + HHT

Among the 24 patients in the clinical trial, treatment-naïve samples were available from 18 of them. In six patients, there was an insufficient number of myeloblasts for storage. Of the 24 patients, 20 showed CR/CRi, 1 showed nCRi, and 3 were refractory. Samples from nine patients who have achieved initial remission were obtained at subsequent leukemia relapse. Paired samples from seven patients obtained before S + HHT treatment and during subsequent leukemia progression were tested for their in vitro drug sensitivity. Myeloblasts at the time of leukemia progression were significantly less sensitive to combination S + HHT treatment in vitro, compared with those that were treatment-naïve (P < 0.05; Fig. 7, A and B). Among patients with CR/CRi, treatment-naïve myeloblasts from those who showed longer LFS, arbitrarily set at 100 days, were significantly more sensitive to sorafenib (P = 0.0426) and HHT (P = 0.0426) in vitro than those with shorter LFS (Fig. 7C).

Fig. 7. Clinical correlation of in vitro drug response.

(A and B) Seven pairs of treatment-naïve (N) and relapse (R) samples obtained from patients who received S + HHT were treated with S + HHT in vitro in CM. Each line in (A) represented the response of treatment-naïve (left) and relapse (right) samples of a patient to S + HHT. Two-tailed paired t test was used. (C) Comparison of in vitro response to sorafenib (left) and HHT (right) between treatment-naïve samples from patients who showed <100 days (N = 8) and ≥100 (N = 6) days LFS. Mann-Whitney U test was used. UPN, unique patient number.


Here, we developed and validated a drug-screening platform for primary AML samples to test the hypothesis that in vitro drug sensitivity of leukemia cells could be used to guide personalized treatment for relapsed/refractory patients. We found that HHT exhibits preferential antileukemia effects against FLT3-ITD AML, synergizes with multikinase inhibitors against FLT3, and targets the cells’ addiction to protein synthesis. These observations were confirmed in primary AML samples, AML cell lines, Ba/F3 cells carrying FLT3-ITD, and FLT3-ITD AML patients receiving S + HHT in a phase 2 clinical trial. The study may provide a basis for future clinical trials in FLT3-ITD AML and offer guidance for the design of personalized treatment for AML.

The culture systems provided a reliable tool to evaluate drug responses in AML. CM contained human cytokines, including SCF, G-CSF, IL-3, and FLT3 ligand, as well as other stromal factors that induced proliferation of myeloblasts reminiscent of their proliferative states in patients. Autologous HS recapitulated the patient-specific cytokine milieu, likely as a result of unidentified factors in the serum, thereby preserving the pathologic characteristics of myeloblasts in vivo. Similar observations have been reported in tumor explants of head and neck and colorectal cancers (27). Direct contact with stromal cells (28, 29) or an arylhydrocarbon receptor antagonist (30) can also enhance maintenance of primary AML cells in vitro. How these systems could be adopted in drug screening to predict clinical responses to treatment is presently unclear. The optimal culture system for drug screening in primary AML samples may depend on the specific drugs and AML subtypes tested, as shown by the discrepant responses between CM and HS systems in some drugs tested in the platform. A critical issue in the design of personalized treatment is whether the drug response of blasts in vitro can be correlated with clinical response. Here, FLT3-ITD AML samples showed greater responses to FLT3 inhibitors than FLT3-WT samples in vitro, an observation also seen in most clinical trials. Furthermore, samples obtained from S + HHT treatment–naïve patients also showed a greater sensitivity in vitro to sorafenib and HHT than samples obtained from patients at the time of drug-resistant relapse. Samples from patients with longer LFS after S + HHT also showed greater response to sorafenib and HHT. Whether the results could be generalized to other drugs would have to be further evaluated.

Extended molecular profiling of 93 samples and its correlation with in vitro drug responses generated important information about therapeutic targets. In addition to FLT3-ITD, mutations of DNMT3A, RUNX1, and NPM1 correlated with enhanced response to HHT, dasatinib, and azacitidine/ponatinib, respectively. In particular, RUNX1 mutant AML is a new disease category in the revised World Health Organization classification (31) and is associated with inferior clinical prognosis. The therapeutic mechanisms of dasatinib in RUNX1 mutant AML and its potential clinical value are currently unclear and would have to be carefully evaluated. Furthermore, the results suggested that multiple gene mutations might have combined effects on drug response. Concurrent NPM1 and DNMT3A mutations clearly affected the drug responses in FLT3-ITD AML. These observations might provide biomarkers for future genotype-based clinical trials.

Screening of primary AML samples identified HHT as an effective agent for treatment of FLT3-ITD AML. The responses to HHT in AML samples clustered with those of FLT3 inhibitors, and synergism was demonstrated upon combination treatment in FLT3-ITD AML both in an in vitro and in a xenograft model. Mechanistically, global protein synthesis inhibition by HHT made proteins with short half-lives particularly vulnerable, including FLT3 and its downstream signaling proteins. Distinct from FLT3 inhibitors that block FLT3 phosphorylation, this mode of action may explain its synergism with FLT3 inhibitors and ability to overcome sorafenib resistance. These results translated into improved clinical outcome for patients, with 20 of 24, mostly with relapsed or refractory leukemia, achieving CR/CRi. The regimen was tolerable and nonhematological toxicities at the doses used were mild and mostly associated with sorafenib rather than HHT. The longer duration of remission also allowed more patients to be bridged to HSCT. Direct comparison with a historical cohort on sorafenib monotherapy was limited by the fact that those patients were generally more heavily pretreated (table S6). Drug resistance was seen both in a xenograft model, with a rapid emergence of leukemia growth after day 70 in the MV4-11–engrafted model, and in S + HHT–treated patients, among whom most responding patients without HSCT relapsed. Two patients lost FLT3-ITD clones at relapse, suggesting that at least in some patients, S + HHT could eradicate the FLT3-ITD clones and leukemia relapse could arise from the FLT3-WT clones. However, loss of FLT3-ITD at leukemia relapse has also been shown after conventional chemotherapy. Potential mechanisms of drug resistance to FLT3 inhibitors in FLT3-ITD AML have been reviewed (8). In our cohort, emergence of D835 activation loop domain mutation was identified in 6 of 14 relapsed patients, suggesting that at least in some patients, the combination treatment has exerted pressure to select clones known to be resistant to sorafenib (26). Two patients carrying FLT3-D835 mutations also responded to S + HHT treatment. One patient relapsed with a D835-negative clone, whereas the other has remained in remission. Nevertheless, drug resistance to HHT has not been well defined. It is presently unclear whether the resistant leukemic subclones have developed a mechanism of drug efflux or adaptation of leukemic signaling that overcame protein synthesis inhibition by HHT.

Morphologic features of the BM under S + HHT treatment might shed light on the therapeutic mechanism. At remission, most patients on S + HHT treatment showed hypocellular marrow in contrast to the cellular marrow often observed in patients treated with sorafenib monotherapy. Whether HHT at the present dose used would have different myelosuppressive effects on differentiated hematopoietic cells from neoplastic clones or on normal hematopoiesis is currently unclear. On the other hand, dysgranulopoiesis was also observed in more than half of the evaluable patients after treatment, supporting the proposition that S + HHT has caused aberrant differentiation of myelopoiesis.

HHT has been used for decades primarily in Mainland China for the treatment of AML. When used as monotherapy, the clinical response and outcome were unsatisfactory, with a CR/CRi rate of 10 to 20% (32). HHT is also used as part of multiagent chemotherapeutic regimens because of its extremely low cost and safe toxicity profile, with CR/CRi about 40 to 60% after combination treatment (33). Biomarkers predictive of treatment response to HHT are currently lacking. Results from this study have bridged this gap in knowledge, showing FLT3-ITD being the AML subtype most sensitive to HHT. In this context, protein translation inhibitors might be considered “broad-spectrum tyrosine kinase inhibitors.” S + HHT incorporated the benefits of high efficiency and low toxicities, allowing repeated courses of treatment even in patients who are unfit for conventional chemotherapy. The role of HHT monotherapy in FLT3-ITD AML is currently unclear. Worldwide, HHT is known as omacetaxine mepesuccinate, and it is U.S. Food and Drug Administration–approved for the treatment of CML resistant to tyrosine kinase inhibitors (34). Tosedostat, an aminopeptidase inhibitor that perturbs protein turnover and inhibits protein synthesis by depleting intracellular amino acids, is also being evaluated for the treatment of AML (35). Its effects in FLT3-ITD AML in combination with FLT3 inhibitors should be further explored.

The present study reported a prototype drug-screening platform that was technically and clinically validated. In vitro drug screening in AML has been reported previously (6, 7), but data showing its clinical relevance have been scarce. The present study demonstrated that carefully performed drug screening with multiple culture systems and unbiased clustering of drug response as well as molecular profiling could generate important and clinically relevant information. However, in the present study, about 40% of tested drugs showed little effect on the majority of AML samples. The list of drugs should be continuously evaluated and revised, taking into consideration new drugs that have emerged from recent clinical trials. The value of this platform to predict subsequent clinical outcome has not been ascertained, limited by the small cohort in this clinical trial and the fact that only 3 of 24 patients have failed to respond. These shortcomings notwithstanding, the present platform might be extended to guide the design of effective treatment and develop biomarkers to predict the best treatment for specific AML subtypes.


Study design

This study primarily tested the hypothesis of whether in vitro drug sensitivity testing of primary BM/PB samples from AML patients could be translated into clinical application. In vitro testing of primary samples showed a preferential effect of HHT on FLT3-ITD AML. This phenomenon was reproduced in vitro in Ba/F3 cells carrying FLT3-ITD, as well as in cancer cell lines and in vivo using mouse xenotransplantation. Sample size (N) and replicates (n) of all experiments are indicated in the corresponding figure legends. To ensure validity of the measurement of drug screening, those with signal-to-background ratio < 4 or Z′ factor < 0.5 (36) were excluded from analyses. Data points involving technical errors including contamination and mistakes in dispensing were also excluded. Extended molecular profiling by targeted gene sequencing was also performed on the primary AML samples to explore possible correlation between drug response and gene mutations.

To validate the findings from in vitro drug screening, a phase 2 clinical trial was conducted to see whether HHT could potentiate the therapeutic effect of sorafenib in FLT3-ITD AML patients. No blinding, randomization, or placebo control was included in the trial. Response rate, time to response, response depth in terms of FLT3-ITD allelic burden at remission, LFS, and OS were the endpoint measurements and were compared with results of our previous published cohort of sorafenib monotherapy (12, 26).

Patients and samples

Blood (PB) or BM samples were obtained from AML patients at diagnosis or relapse. Mononuclear cell fraction was isolated by density-gradient centrifugation (Ficoll-Paque PLUS/Lymphoprep) and cryopreserved in liquid nitrogen until use. Samples with 23 to 99% myeloblasts as reported by hematopathologists were included in this study, in which 69.8% of samples contained ≥70% blasts. Preparation of HS and CM and their culture protocols are shown in Supplementary Materials and Methods.

Drug screening of primary AML cells

Twenty-five drugs were selected on the basis of their mechanisms of action, potential antileukemia effects, and availability for off-label clinical application or clinical trials. They included multikinase inhibitors, differentiation agents, mTOR inhibitor, histone deacetylase inhibitor, hypomethylating drug, protein synthesis inhibitor, proteasome inhibitor, and chemotherapeutic drug (table S1). They were tested at a dose range of 104-fold dilution starting from 1 or 10 μM. Round-bottom 96-well plates containing 10 μl of drugs or vehicle (0.1% DMSO) at 10× concentrations were prepared in advance according to the designed layout (fig. S19) and stored at −80°C until use. Before culture, the plates were equilibrated at 37°C, and 90 μl of the corresponding medium with cells was added to each well. The readout of leukemia cell growth and its validation are shown in Supplementary Materials and Methods.

Next-generation sequencing

From the 93 samples screened, genomic DNA was extracted using QIAamp DNA Blood Mini Kit (Qiagen) and analyzed by MiSeq next-generation sequencing with TruSight Myeloid Sequencing Panel (Illumina). The panel targeted 54 genes, covering the full coding sequence of 15 genes and exonic hotspots for the other 39 genes (table S7). Workflows of MiSeq sequencing library preparation, variant calling, and annotation, as well as detection of FLT3-ITD by ITDseek, have been previously described (37, 38). Complex insertions and deletions were detected by an in-house designed algorithm INDELseek on a Cray XC30 supercomputer.

In vitro cell line study

Human leukemia cell lines (MOLM-13, MV4-11, KG-1, K562, Kasumi-1, NB4, OCI-AML3, THP-1, and ML2) originated from Leibniz Institut Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ) and American Type Culture Collection and were maintained according to their protocols. Ba/F3-ITD cells were transduced with pLJM1-EGFP (Addgene, no. 19319) as an empty vector (Ba/F3-GFP), or carrying FLT3-WT or FLT3-ITD, with puromycin selection. Experiments on OP-Puro incorporation and drug-resistant models were performed on Ba/F3 cells transduced with pLKO.1-blast (Addgene, no. 26655) carrying FLT3-ITD, FLT3-ITD–D835V, FLT3-ITD–D835F, and FLT3-ITD–F691L. Culture conditions for these cell lines and generation of drug-resistant lines as well as standard laboratory procedures, including flow cytometry, Western blots, and xenotransplantation, are shown in Supplementary Materials and Methods.

Study approval

The animal study was approved by the Committee on the Use of Live Animals for Teaching and Research (CULATR 3501-14) at the University of Hong Kong. For the clinical trial, written informed consent was received from participants before inclusion in the study. The study was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (HKU) (UW 10-393) according to the declaration of Helsinki. Protocols of the clinical trial and FLT3 genotyping are shown in Supplementary Materials and Methods.

Statistical analysis

Unless otherwise specified, data are expressed as means ± SEM and evaluated and compared by either Student’s t test/Mann-Whitney U test (numerical data) or Fisher’s exact test (categorical data). Survival analysis was performed using the Kaplan-Meier method. Differences in survival were determined using log-rank test. P values of less than 0.05 were considered statistically significant and are shown either in the respective figures or table S8. Statistical evaluation of drug responses and molecular profile are shown in Supplementary Materials and Methods.


Materials and Methods

Fig. S1. In vitro culture of primary AML cells.

Fig. S2. Dose-response profile in CM.

Fig. S3. Dose-response profile in HS.

Fig. S4. Distribution of area under the curve in both systems.

Fig. S5. Correlation of antileukemia effects between CM and HS.

Fig. S6. Correlation of antileukemia effects between drugs.

Fig. S7. Unsupervised hierarchical clustering analysis of drug-response profiles of primary AML samples.

Fig. S8. Relative response of AML samples within drug clusters in CM.

Fig. S9. Relative response of AML samples within drug clusters in HS.

Fig. S10. Effect of gene mutations on drug response.

Fig. S11. Preferential effect of HHT on FLT3-ITD AML.

Fig. S12. In vitro effect of S + HHT on cell lines.

Fig. S13. Effect of S + HHT on FLT3-WT and ITD AML samples.

Fig. S14. Effect of sorafenib and/or HHT on the amount of protein in MV4-11.

Fig. S15. Drug-resistant cell line models.

Fig. S16. Western blot of MOLM-13 drug-resistant lines after treatment with sorafenib and/or HHT.

Fig. S17. Xenotransplantation model.

Fig. S18. BM findings before and after S + HHT treatment.

Fig. S19. Plate layout for drug screening.

Fig. S20. Platform validation.

Table S1. Drug library.

Table S2. Comparison of drug response between FLT3-WT and ITD samples.

Table S3. Drug-response and molecular profile data set of 96 primary AML samples (provided as an Excel file).

Table S4. Drug-response profile of AML cell lines to 25 drugs (provided as an Excel file).

Table S5. Clinicopathologic characteristics of 24 patients in the phase 2 clinical trial of S + HHT.

Table S6. Comparison of clinical information between the present S + HHT and the historical sorafenib monotherapy cohort.

Table S7. Panel of 54 genes analyzed by next-generation sequencing of 568 amplicons.

Table S8. P values of statistical tests in figures and supplementary figures (provided as an Excel file).

Table S9. List of antibodies and other staining molecules used for flow cytometry and Western blots.

Table S10. Primer sequences.

References (39, 40)


  1. Acknowledgments: We thank all patients for providing their blood and BM samples in this study. We are grateful that S.F. Yip [Tuen Mun Hospital, Hong Kong], H. Liu (Pamela Youde Nethersole Eastern Hospital, Hong Kong), H. Lee and V. Mak (Princess Margaret Hospital, Hong Kong), C. K. Lau (Tseung Kwan O Hospital, Hong Kong), S. Y. Lin (United Christian Hospital, Hong Kong), J. Lau (Queen Elizabeth Hospital, Hong Kong), and W. Li (Prince of Wales Hospital, Hong Kong) have referred patients to the clinical trial. C. Eaves (Terry Fox Laboratory, BC Cancer Agency, Vancouver, British Columbia, Canada) provided the engineered mouse stromal cell lines. Kasumi-1/ML2 and THP-1 lines were gifts from L.-C. Chan and V. S.-F. Chan. BD LSRFortessa analyzer and PerkinElmer IVIS Spectrum in vivo imaging were provided by the Faculty Core Facility, Faculty of Medicine, The University of Hong Kong. Sorafenib used in the clinical trial was partly provided by Bayer Health Ltd. as a gift. Funding: S.S.Y.L. was supported by the Croucher Foundation, and A.Y.H.L. was supported by Li Ka Shing Faculty of Medicine and by an endowment from the Li Shu Fan Medical Foundation. The study was supported by the S.K. Yee Medical Foundation (210213, 213212, and 209214), Hong Kong Blood Cancer Foundation, Lee Hysan Foundation, and the Collaborative Research Fund (CityU9/CRF/13G). Author contributions: Most of the study design, experiments, data acquisition, and analyses were conducted by S.S.Y.L. and E.S.K.H. B.-L.H., W.-W.W., C.-Y.C., N.K.L.N., C.-H.M., H.G., H.-W.I., and C.-C.S. helped perform some experiments, data acquisition, and analyses. The clinical trial was designed and conducted by A.Y.H.L. The entire study was guided and supervised by A.Y.H.L., with the help of A.M.S.C. at the initial stage. The manuscript was written by S.S.Y.L. and A.Y.H.L. and reviewed and agreed by all coauthors. Competing interests: Y.-L.K. was involved in consultation for Bayer unrelated to this study. The other authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article