Rational Design of a Parthenolide-based Drug Regimen That Selectively Eradicates Acute Myelogenous Leukemia Stem Cells

Abstract

Although multidrug approaches to cancer therapy are common, few strategies are based on rigorous scientific principles. Rather, drug combinations are largely dictated by empirical or clinical parameters. In the present study we developed a strategy for rational design of a regimen that selectively targets human acute myelogenous leukemia (AML) stem cells. As a starting point, we used parthenolide, an agent shown to target critical mechanisms of redox balance in primary AML cells. Next, using proteomic, genomic, and metabolomic methods, we determined that treatment with parthenolide leads to induction of compensatory mechanisms that include up-regulated NADPH production via the pentose phosphate pathway as well as activation of the Nrf2-mediated oxidative stress response pathway. Using this knowledge we identified 2-deoxyglucose and temsirolimus as agents that can be added to a parthenolide regimen as a means to inhibit such compensatory events and thereby further enhance eradication of AML cells. We demonstrate that the parthenolide, 2-deoxyglucose, temsirolimus (termed PDT) regimen is a potent means of targeting AML stem cells but has little to no effect on normal stem cells. Taken together our findings illustrate a comprehensive approach to designing combination anticancer drug regimens.

Introduction

Numerous studies have documented the biological complexity of human tumor cell populations in which genetic, epigenetic, biochemical, and metabolic properties can often be quite heterogeneous (1, 2). As a result, complicated multidrug regimens are often employed to target various components of tumor biology and/or acquired drug resistance (3, 4). However, the rationale behind the design of multidrug regimens is usually inconsistent and often driven by pragmatism more than scientific rigor. Thus, establishing more effective means by which to identify optimal drug regimens is an important priority, particularly with the recent emergence of a broad range of targeted agents.

To enhance the design of combination regimens, we sought to draw upon advanced methods of global cell analysis that allow comprehensive studies of genomic, proteomic, and metabolomic properties. Such strategies have become prevalent in recent years and are now routinely used in nearly all types of biological research. However, to our knowledge, simultaneous use and integration of multiple “omic” methods to design drug combinations for cancer have not been widely reported. Thus, in the present study our goal was to develop a multifaceted approach that would utilize several platforms for comprehensive cell analysis.

As a model system in which to pilot this approach, we chose targeting of human acute myelogenous leukemia (AML).² Previous studies have clearly established the biological heterogeneity of AML as well as sophisticated experimental systems to evaluate the efficacy of candidate regimens (5–8). Furthermore, among human tumors, AML has arguably the best-characterized cancer stem cell population. Eradication of the AML stem cells is thought to be critical for achieving improved clinical outcomes (9–13). Although several strategies have been reported (14–19), as yet it is not clear to what extent any approach will comprehensively eradicate heterogeneous AML stem cell populations. Hence, the studies described here were designed to identify and target fundamental conserved components of AML stem cell biology.

As a foundation for rational design of a multidrug regimen, we chose to evaluate parthenolide (PTL), a naturally occurring small molecule with reported anti-tumor properties in nearly all major forms of human cancer (20, 21). Of particular interest is the activity of PTL as an agent in AML where it can target bulk tumor cells as well as the relatively rare AML stem cell population (22). These properties have made PTL an exciting preclinical compound to study. However, the use of PTL as a single agent to treat cancer still has some potential drawbacks. First, data from the Cancer Therapeutics Response Portal show that certain tumor types still display relative resistance to PTL treatment (23). Second, PTL has relatively unfavorable aqueous solubility and stability (18, 24); thus, there is a strong rationale for strategies to enhance the anti-cancer properties of PTL at as low a dose as possible. Third, given the complexity of most human tumors, it is likely that combinatorial strategies using drugs to target multiple intracellular pathways will yield superior results. Thus, PTL makes an excellent starting point to test our strategy for the rational design of a multidrug regimen.

The first step of our strategy involved comprehensive characterization of the anti-leukemia mechanism of PTL in AML. Although previous studies by our group and others have shown that PTL is a strong inhibitor of the NF-κB pathway and also a potent inhibitor of glutathione metabolism (22, 25–29), there have been very few studies that systematically characterize its anti-cancer mechanism at a multi-omic level. In the present study, we investigated the mechanism of PTL by first using a biotinylated analog to identify its proteomic interactome. In addition, we profiled PTL-induced transcriptomic changes in primary AML cells. Furthermore, we employed isotope labeling experiments to monitor global metabolic flux changes that occur upon treatment of primary AML cells with PTL. By integrating results from these omic studies along with our previous reports, we discovered that PTL directly modulates several components of AML redox balance and induces strong activation of the Nrf2-mediated oxidative stress response as well as up-regulation of the pentose phosphate pathway (PPP), both of which are thought to be compensatory mechanisms that are employed by AML cells in an effort to survive the PTL insult. Therefore, we tested the concept that inhibition of such compensatory events would enhance the cytotoxicity of PTL. Subsequent studies identified the triple drug regimen “PDT” in which parthenolide (P) was combined with 2-deoxyglucose (D) and temsirolimus (T), drugs chosen for their ability to inhibit the PPP and the Nrf-2 mediated anti-oxidant response, respectively (30, 31). Biological analyses of the PDT regimen demonstrated strong toxicity toward primary AML cells including those that are relatively resistant to PTL treatment alone. Importantly, the PDT regimen also displays potent toxicity toward leukemic stem/progenitor cells with very little toxicity toward normal hematopoietic counterparts. Thus, by using a targeted omic approach, we were able to perform rational design of the PDT regimen with a superior leukemia-specific cytotoxicity compared with its parent compound, parthenolide.

Results

Identification of the Proteomic Interactome of PTL in Primary AML Cells

As an initial step toward developing enhanced PTL-based combination regimens, we first sought to thoroughly characterize the proteomic interactome of PTL. The most prevalent chemical feature of PTL is an α-methylene-γ-lactone moiety, which mediates covalent interactions via the Michael addition reaction (Fig. 1A). Hence, PTL's cellular targets should be readily identifiable using biochemical pulldown methods. To this end, we previously employed a biotinylated analog of PTL, known as MMB-biotin (Fig. 1B), to identify candidate target proteins. We validated its use for known targets of PTL including IKKβ and NF-κB p65 (26–28). In the current study we combined the MMB-biotin-based pulldown assay with liquid chromatography-tandem mass spectrometry (LC-MS/MS) to allow the identification of the entire proteomic interactome of PTL in primary AML cells.

FIGURE 1.

Characterization of PTL's proteomic interactome in primary AML cells. A, chemical structure of parthenolide. B, chemical structure of MMB-biotin. C, silver stain of primary AML total cell lysates and subcellular fractions on a SDS-PAGE gel. D, Western blot (WB) showing the purity of subcellular fractions. E, Western blot showing the SA pulldown products from total lysates and all subcellular fractions, probed with streptavidin-HRP. F, a schematic diagram illustrating the methodology used to compare the proteomic interactome and transcriptomic signature of PTL in primary AML cells. G, number of PTL binding targets within the top five most perturbed signaling pathways induced by PTL in primary AML cells. p values were calculated by IPA analysis to test if a particular pathway is significantly represented by the proteomic interactome of PTL.

Briefly, we treated primary AML cells in culture with the MMB-biotin probe or free biotin control for 2 h followed by washing and harvesting steps. We then fractionated MMB-biotin/free biotin-treated cells into cytosolic, membrane, nuclear, and insoluble cytoskeletal fractions to allow identification of PTL binding targets from each distinct subcellular compartment (Fig. 1C). Purity of each fraction was confirmed by showing enrichment of fraction-specific markers (Fig. 1D). Each fraction was then incubated with streptavidin-coated beads to isolate PTL binding targets. We confirmed the specificity of our pulldown by probing each pulldown product with streptavidin-HRP (SA-HRP). As shown in Fig. 1E, lane 1, there are only two bands detected by SA-HRP from pulldown products of free biotin-treated cells: one at ∼75 kDa, which is a previously described nonspecific binding event (labeled as n.s.) (26), and another at ∼12 kDa, which is presumably a monomeric SA dissociated from SA-coated beads (labeled as SA). In contrast, pulldown products from MMB-biotin-treated cells have a spectrum of binding targets between 20 and 100 kDa (Fig. 1E, lane 2). In addition, these targets are largely distributed in the cytosolic and membrane fractions, with much less in the cytoskeletal fraction and the least in the nuclear fraction (Fig. 1E, lanes 3–6). Using LC-MS/MS of each fraction, we subsequently identified 312 binding targets of PTL in primary AML cells (supplemental Table 1). Notably, several previously reported binding targets of PTL including IKKβ, GCLC, GPX1, and TXN were all successfully identified through this approach (25–27), suggesting our method had good specificity and coverage. Subsequently, we employed the Ingenuity Pathway Analysis (IPA) software and identified a list of pathways/biological functions that are significantly represented by this pool of binding targets (supplemental Table 2). Importantly, the Nrf2-mediated oxidative stress response (p < 0.0005) and glutathione metabolism (p < 0.005) pathways are highly enriched, consistent with our previous findings that the ability of PTL to target glutathione metabolism and induce oxidative stress is important for its anti-leukemia activity (22, 25).

Characterization of the Transcriptomic Changes Induced by PTL in AML Cells

We next wanted to characterize PTL-induced transcriptomic changes in primary AML cells. To this end we treated primary AML cells (n = 4) with 7.5 μm PTL or DMSO control for 6 h, isolated total mRNA, and profiled the expression of their transcriptomes using RNA-seq. By comparing the gene expression level between DMSO control and 7.5 μm PTL-treated cells, we identified a total of 2114 significantly dysregulated genes (-fold change ≥1.5, p < 0.05, supplemental Table 3). We then used the IPA software to predict the significantly dysregulated pathways in PTL-treated AML cells (supplemental Table 4). Among the top five most dysregulated pathways are the Nrf2-mediated oxidative stress response pathway (hereafter referred to as the Nrf2 pathway) and the protein ubiquitination pathway (hereafter referred to as the ubiquitin pathway). This finding is consistent with our previous study using a different cohort of primary AML specimens (GEO deposit ID: GSE7538) (32, 33), suggesting these anti-leukemic activities of PTL are universally present irrespective of heterogeneous primary AML specimens.

To test if there is a link between the PTL proteomic interactome and PTL-induced transcriptomic changes, we examined the pool of 312 PTL binding targets for ones that are directly present in the top 5 transcriptionally dysregulated pathways induced by PTL (Fig. 1F). This analysis identified 13 (p < 0.0005) PTL binding targets in the Nrf2 and 14 (p < 0.0005) targets in the ubiquitin pathway (Fig. 1G), suggesting that proteomic interactions of PTL may be linked to PTL-induced transcriptomic changes. Importantly, the up-regulation of the Nrf2 pathway suggests an accumulation of cellular oxidative stress (34), whereas the activation of the ubiquitin pathway usually implies an increase of protein misfolding stress (35). Thus these data essentially suggest two strong anti-leukemic activities of PTL including the induction of cellular oxidative and protein misfolding stress.

Metabolomic Analyses Reveal Increased PPP Activity for NADPH Production in PTL-treated AML Cells

To further investigate the anti-leukemic activity of PTL, we characterized PTL-induced metabolomic changes in primary AML cells using previously established methods (36). As outlined in Fig. 2A, primary AML cells were cultured in media containing uniformly labeled [U-¹³C₆]glucose or dual-labeled [¹³C_1,2]glucose and treated with vehicle control or 7.5 μm PTL. We harvested cells at 0, 4, and 6 h and quantified the relative distribution of the isotopologue of ¹³C-labeled metabolites at each time point.

FIGURE 2.

Metabolomic analyses reveal increased PPP activity for NADPH production in PTL-treated AML cells. A, a schematic diagram showing the workflow used to characterize PTL-induced metabolomic changes in primary AML cells. Results from the uniform labeled [U-¹³C₆]glucose (B) and the dual-labeled [¹³C_1,2]glucose (C) labeling experiments showing quantification of ¹³C-labeled metabolites in glycolysis, PPP, and TCA cycle in primary AML cells after DMSO or 7.5 μm PTL treatment. Data represent the mean ± S.D. (n = 3). For each metabolite, M+x indicates specific isotopologoues of a metabolite labeled with x numbers of the heavy ¹³C isotopic carbons. A circle diagram is also given to each metabolite to illustrate its carbon frame labeled with different numbers of ¹³C carbons (red solid circles represent heavy ¹³C, clear open circles represent ¹²C). D, quantification of NADPH and NADPH/NADP+ ratio in primary AML cells treated with DMSO control or 7.5 μm PTL using a bioluminescent assay. Data represent the mean ± S.D. (n = 3).

As shown in Fig. 2B, compared with vehicle control, 6 h PTL-treated AML cells have increased levels of several important metabolic intermediates including [¹³C]glucose 6-phosphates (M+6, 5.36-fold), [¹³C]ribose phosphate (M+5, 2.80-fold), [¹³C]glyceraldehyde 3-phosphate (M+3, 15.04-fold), and [¹³C]lactate (M+3, 1.83-fold), suggesting primary AML cells respond to PTL insult by increasing flux of glucose through glycolysis and the pentose phosphate pathways (PPP). In addition, we also observed increased levels of citric acid (TCA) cycle intermediates including [¹³C]citrate (M+2, 4.67-fold), [¹³C]malate (M+2, 4.23-fold), and [¹³C]aspartate (M+2, 1.47-fold), further suggesting that PTL-treated AML cells become more glucose-addicted than control-vehicle-treated cells, as they appear to rely on glucose oxidation to fuel energy-generating pathways.

The increase of glucose flux into the PPP is particularly interesting, as the PPP is a major pathway generating NADPH, which is important for glutathione homeostasis (37). We previously reported that 4–6-h PTL (7.5 μm) treatment can induce maximum glutathione depletion in primary AML cells (25). Thus we hypothesized that the observed increase of glucose flux into the PPP at ∼6 h is used to generate NADPH to rescue PTL-induced glutathione depletion. To test this hypothesis, we first confirmed the PPP activity increase in PTL-treated AML cells by a second set of labeling experiments using the dual-labeled [¹³C_1,2]glucose. In this particular experiment, because one carbon atom is lost in the form of CO₂ during glucose oxidation steps through the PPP (38), the M+1 isotopologue of [¹³C]glyceraldehyde 3-phosphate is exclusively generated by glucose flux through the PPP, not glycolysis. Thus, the level of the M+1 isotopologue represents the PPP activity. As shown in Fig. 2C (complete data set in supplemental Fig. 1), this M+1 isotopologue of [¹³C]glyceraldehyde phosphate is dramatically increased over time in PTL-treated AML cells, whereas its level in vehicle control-treated cells remains virtually zero (p < 0.001). Next we tested the hypothesis that the increased PPP activity is used to produce more NADPH. To this end, we performed a separate bioluminescent assay to quantify the cellular NADPH and NADP+ levels and found that 7.5 μm PTL treatment indeed significantly elevated both net NADPH level and NADPH/NADP+ ratio in AML cells (Fig. 2D). Together these data strongly demonstrate that AML cells respond to PTL insult with a robust activation of the PPP for NADPH production presumably to detoxify PTL-induced oxidative stress in primary AML cells.

Integration of omics Data Provides Strategies for Rational Design of PTL-based Anti-leukemic Regimens

The metabolomic data above reveals that activating glucose metabolism to shunt more glucose into the PPP for NADPH production is a major metabolic response of AML to PTL insult. This observation is consistent with our transcriptomic study results, which show that PTL dramatically induces Nrf2-mediated oxidative stress response (Fig. 1G). Thus both metabolomic and transcriptomic studies shed light on the robust oxidative stress induced by parthenolide and some of the prominent responses mounted by primary AML cells upon challenge with PTL. In addition, our proteomic study also shows that PTL can directly interfere with many proteins that are present in the Nrf2 pathway (Fig. 1G), and our previous study has reported that PTL can potently deplete the cellular antioxidant glutathione (25), thereby providing two additional mechanistic insights to explain PTL-induced oxidative stress. Thus, through the integration of multiple omic techniques performed in the current study along with previous research, we hypothesize that among the many anti-leukemic activities of PTL, its ability to induce oxidative stress is the most important. Notably, oxidative stress is known to induce protein-misfolding stress as correct folding of proteins is usually redox-sensitive (39, 40). Therefore, we also hypothesize that the increase of protein misfolding stress upon PTL treatment is a secondary consequence of PTL-induced oxidative stress.

Importantly, the above analyses of the anti-leukemic activities of PTL provide us with strategies for rational design of PTL-based anti-leukemic regimens. As illustrated in Fig. 3A, we reasoned that the addition of drugs that can enhance PTL-induced oxidative stress and possibly protein misfolding stress represent a promising strategy to further increase the anti-leukemic activity of PTL. To this end, we tested 2-deoxyglucose (2DG), a hexokinase inhibitor that can block glucose oxidation via the PPP pathway responsible for NADPH production (30), and temsirolimus (TEM), that can effectively abrogate PTL-induced Nrf2-mediated oxidative stress responses (31). We hypothesized that the combined use of PTL + 2DG + TEM (termed the PDT regimen) would achieve superior eradication of primary AML cells due to simultaneous induction of oxidative stress and inhibition of the compensatory anti-oxidant responses intrinsic to AML.

FIGURE 3.

The PDT regimen depletes glutathione and induces ROS without activating Nrf2-mediated oxidative stress response and NADPH production. A, a simplified diagram illustrating the basic rationale of the triple drug regimen PDT. B, amount of total glutathione in primary AML cells after 4 h of treatment. C, intracellular ROS level in primary AML cells after each treatment, quantified by the CM-H₂DCFDA dye. MFI, mean fluorescence intensity. D, QPCR data showing -fold changes of mRNA expression induced by 7.5 μm PTL or the PDT regimen. Each line represents an individual AML specimen (n = 4). ns, not significant. E, Western blot (WB) data showing protein expression of HMOX1 after various treatments for 6 h. F, QPCR data showing -fold changes of mRNA expression induced by 7.5 μm PTL or the PDT regimen. The bars represent the mean ± S.D. (n = 4). G, quantification of NADPH level in primary AML cells after various treatments. UNT, untreated; RFU, relative fluorescence units. Bars represent the mean ± S.D. (n = 3); * indicates p < 0.05; ** indicates p < 0.01.

The PDT Regimen Blocks Both Nrf2 Response and NADPH Production Thereby Increasing the Potency of PTL

To initially evaluate activity of the PDT triple drug regimen in AML cells, we investigated the impact of PDT on several aspects of the AML redox system, including glutathione, oxidative stress, Nrf2 pathway response, and NADPH. Importantly, for these experiments and all subsequent analyses in the current study, we employed a lower dose of PTL (2.5 μm), which is not sufficient to induce significant anti-leukemic activity alone. Using a lower dose permits better detection of additive/synergistic interactions that result in enhanced PTL activity. Furthermore, even though 2.5 μm PTL was used for the drug combinations, the single agent PTL control for these studies was maintained at 7.5 μm, which is generally a potent cytotoxic dose for AML cells (22, 25).

As shown in Fig. 3B, whereas each drug alone had a minimal effect in depleting cellular glutathione, they synergized and depleted >80% of total glutathione in primary AML cells. This degree of glutathione suppression is similar to what we have previously observed using PTL alone at 7.5 μm (25). Thus, the addition of 2DG and TEM restores the ability of a low dose PTL to target glutathione metabolism, a central mechanism in the overall AML-specific toxicity of PTL-based anti-leukemic strategies. Furthermore, we note that a biological consequence of the glutathione inhibition obtained in PDT-treated AML cells is a strong increase in production of reactive oxygen species (ROS), shown by CM-H₂DCFDA staining (Fig. 3C). Compared to 7.5 μm PTL, the PDT treatment induced significantly stronger ROS increase at all time points. And interestingly, unlike a gradually decreased ROS profile in 7.5 μm PTL-treated AML cells, the ROS increase in PDT-treated AML cells was sustained up to 8 h post PDT treatment, suggesting a lack of Nrf2-mediated oxidative stress response, therefore disabling AML cells from managing the elevated ROS. To test this hypothesis, we treated 4 primary AML specimens with the PDT combination for 6 h and performed QPCR analysis on genes in the Nrf2 pathway. For comparison purposes, we also profiled 6 h, 7.5 μm PTL-induced gene expression changes in these cells side by side. As shown in Fig. 3D, our data demonstrate that 7.5 μm PTL induced strong up-regulation of all glutathione pathway components including the cysteine transporter SLC7A11-SLC3A2 complex, the rate-limiting glutathione synthesis enzyme complex GCLC-GCLM, and glutathione reductase (GSR). This is consistent with our previous finding that 7.5 μm PTL can potently deplete glutathione in AML cells and indicates that up-regulation of these genes is a component of AML response to PTL (25). In contrast, despite the strong glutathione depletion shown in Fig. 3B, PDT demonstrated almost no induction of GCLC-GCLM and GSR. We propose that the lack of this compensatory response to drug insult may explain the sustained ROS increase seen in PDT-treated AML cells (Fig. 3C). In addition to the glutathione system, we also investigated the expression of HMOX1, another major effector of the Nrf2-mediated anti-oxidant response. As shown in Fig. 3D, whereas HMOX1 mRNA was strongly up-regulated by 7.5 μm PTL, it is only mildly induced by PDT. Consistent with the mRNA expression data, PTL treatment produced a dramatic increase in HMOX1 protein expression, whereas the PDT regimen only induced a modest increase (see Fig. 5E). Finally, we observed that 7.5 μm PTL induced the expression of thioredoxin system proteins, TXN and TXNRD1, and suppressed the expression of thioredoxin inhibitory protein TXNIP in AML. However, PDT did not alter the expression of these thioredoxin system genes as well (Fig. 3D).

FIGURE 5.

The PDT regimen effectively eradicates primary AML. A, viability of 10 primary AML specimens after being treated with various drugs alone or in combination for 24 h. Data are plotted as % of untreated. Bars represent the mean ± S.D. B, viability of 5 PTL-resistant primary AML specimens after treatment with 7.5 μm PTL or the PDT regimen. Data are plotted as % of untreated. Bars represent the mean ± S.D. C, number of colonies in methylcellulose culture per 100,000 primary AML cells treated with the PDT regimen or 7.5 μm PTL. Bars represent the mean ± S.D. In all drug combinations including the PDT regimen, PTL, 2DG, and TEM was used at 2.5 μm, 0.1125 mg/ml, and 2.5 μg/ml, respectively. ** indicates p < 0.01; **** indicates p < 0.0001.

We next looked at Nrf2 downstream genes that are specifically related to NADPH production. Shown in Fig. 3F, whereas 7.5 μm PTL induced strong up-regulation of the PPP enzymes G6PD and PGD, PDT treatment did not induce them. Consequently, whereas 7.5 μm PTL induced a dramatic NADPH increase in AML cells, PDT did not elevate cellular NADPH levels during 8 h of its treatment (Fig. 3G), suggesting a lack of NADPH production from PDT-treated AML cells despite a >80% loss of cellular glutathione (Fig. 3B).

Taken together, these data indicate that the PDT regimen is a potent ROS inducer in primary AML cells, but despite the rapid increase in oxidative stress, PDT-treated cells are unable to mount an effective anti-oxidant (i.e. protective) response. These findings predict that cytotoxicity toward AML cells should be increased for the PDT regimen in comparison to PTL.

PDT Regimen Does Not Affect NF-κB Signaling but Induces Strong Protein Misfolding Stress in Primary AML Cells

PTL is also known as a strong NF-κB inhibitor (20). However, although we confirmed that 7.5 μm PTL induced a loss of p65 electrophoretic mobility shift assay (EMSA) binding activity, a reduction of NF-κB p65 phosphorylation, and a decrease in NF-κB target gene IL-6 expression, the PDT regimen had no effect on any of these three readouts, indicating that the anti-leukemic mechanism of PDT is independent of NF-κB inhibition (Fig. 4, A–C). It further suggests that the anti-leukemic mechanisms of PDT are similar but not identical to PTL alone. Given these differences, we sought to develop a more global knowledge of the overall PDT mechanism and, therefore, conducted a transcriptome analysis. For these studies we profiled 6-h PDT-induced gene expression changes in four primary AML specimens using RNA-seq and then performed a bioinformatic analysis to characterize key signaling pathways that are related to the anti-leukemic activities of PDT.

FIGURE 4.

The PDT regimen does not inhibit NF-κB but induces strong protein misfolding stress in primary AML cells. A, NF-κB EMSA of primary AML cells after various treatments for 6 h. Intensity of EMSA signal in each lane was quantified and normalized to the untreated (UNT) group and expressed as relative intensity values below each lane. B, Western blot (WB) showing level of phospho-p65 (Ser-536) in primary AML cells after various treatments for 6 h. C, QPCR results showing mRNA expression of NF-κB downstream gene IL6 in primary AML cells following various treatments for 6 h. D, GSEA enrichment plots showing up-regulation of both KEGG_PROTEASOME and REACTOME_UNFOLDED_PROTEIN_RESPONSE gene sets in AML cells treated with the PDT regimen. E, heat maps showing up-regulation of genes related to chaperone function, ER stress response, and proteasome activity. -Fold change (FC) and standard deviation (SD) are given for each gene (n = 4). Red indicates increase. Blue indicates decrease. F, QPCR data showing expression of CHOP mRNA in primary AML cells after various treatments for 6 h. In all sub panels, unless explicitly stated, the doses used in all drug combinations including the PDT regimen are 2.5 μm, 0.1125 mg/ml, and 2.5 μg/ml for PTL, 2DG, and TEM, respectively. In C and F, bars represent the mean ± S.D. (n = 3).

Using a collection of KEGG gene sets, we performed Gene Set Enrichment Analysis (GSEA) of our RNA-seq data and found that the “Proteasome” function was strongly up-regulated by the PDT treatment (Fig. 4D, left panel, supplemental Table 5). Because an increase of proteasome activity usually is a cytoprotective response of cells to degrade misfolded proteins (41), these data suggest that the PDT treatment induces protein misfolding stress in primary AML cells. Consistent with this observation, in an independent GSEA analysis using a collection of REACTOME gene sets, we also found that the “Unfolded protein response” pathway is significantly up-regulated upon PDT treatment (Fig. 4D, right panel). And lastly, we also analyzed the RNAseq data with the iPAGE gene expression analysis tool (42). This analysis identified the unfolded protein binding function as being significantly up-regulated in three out of four primary AML specimens treated with PDT (supplemental Fig. 2A).

To confirm the results from these gene expression analyses, we next examined the expression level of individual genes related to proteasome activity, chaperone function, and ER stress response, specifically. As shown in Fig. 4E, a global up-regulation of chaperone proteins (e.g. HSP60, HSP70, HSP90, and DNAJs, etc.) was seen in PDT-treated AML cells. In addition, expression of many proteasome subunits was also increased in PDT-treated AML cells. Moreover, we observed a global increase of well known ER stress markers (SERCA2, PDIs, GRP94, GRP78, and CHOP) upon PDT treatment. Lastly, CHOP up-regulation is typically considered to be a marker for ER stress that readily triggers apoptosis (43). Therefore, we performed more detailed quantitative PCR analysis on the expression of CHOP after PDT treatment. As shown in Fig. 4F, although each drug alone has a small effect on CHOP expression (1.7–3.6-fold), together they induced an impressive 12.7-fold increase in CHOP gene expression, suggesting a synergistic ER stress induction effect by PDT. Together these analyses suggest that the PDT regimen creates a strong protein misfolding stress in primary AML cells.

In addition to up-regulated pathways, our GSEA analysis also identified biological functions that are significantly down-regulated by PDT (supplemental Table 6). Among them are results suggesting down-regulation of DNA replication and fatty acid metabolism (supplemental Fig. 2, B and C). DNA replication is important for proliferation of cancer cells; thus, PDT-induced inhibition of DNA replication is potentially important to its activity toward bulk leukemic cells. On the other hand, fatty acid metabolism is known to be important for leukemia stem cells (44). Thus, inhibition of fatty acid metabolism induced by PDT could contribute its ability to target AML stem/progenitor cells.

The Triple Drug Regimen PDT Can Effectively Target AML Stem and Progenitor Populations

Given that the PDT triple drug regimen has a unique ability to induce ROS and ER stress and simultaneously inhibits many cytoprotective antioxidant responses of AML, we hypothesized that the PDT combination would demonstrate strong anti-leukemic activity. To test this hypothesis, we treated 10 primary AML specimens with single, dual, and triple drug combinations for 24 h and measured the viability of cells after each treatment. As presented in Fig. 5A, our data showed that the viability of primary AML cells was modestly affected when treated with each drug alone at a sublethal dose: 2.5 μm PTL (87.7 ± 13.8% viable), 0.1125 mg/ml 2DG (67.5 ± 19.5% viable), and 2.5 μg/ml TEM (85.6 ± 13.2% viable). Dual drug combinations yielded somewhat more efficient targeting: PTL+2DG (45.3 ± 23.0% viable), PTL+TEM (63.3 ± 16.1% viable), and 2DG+TEM (37.1 ± 14.7% viable). However, treatment with triple drug regimen PDT showed the most effective eradication of AML cells (19.5 ± 10.2% viable), demonstrating that the combination of all three drugs is substantially more toxic to primary AML cells than single and dual drug treatments (Fig. 5A). Importantly, we have also used PDT to treat primary AML cells that are especially resistant to single agent PTL. As shown in Fig. 5B, although 7.5 μm PTL treatment only resulted in ∼25% median cytotoxicity, PDT was much more effective, inducing ∼75% cell death, demonstrating that the PDT triple drug regimen can be superior to a single agent PTL.

Although targeting of bulk leukemia cells is certainly desirable, for AML the most critical targets are the more primitive stem and progenitor cell populations, which are thought to propagate the disease (10–13). To directly determine if the PDT combination can functionally impair progenitor populations, we treated primary AML cells with various combinations of PTL, 2DG, and TEM drugs for 24 h in culture and then plated them for colony-formation assays (to detect progenitor cell activity). Our results showed that PDT treatment greatly reduced the colony-forming potential of primary AML cells and was consistently superior to a 7.5 μm PTL alone (Fig. 5C). Together, these data clearly indicate that, in addition to its toxicity to bulk leukemia cells, the PDT regimen can also strongly impair the more primitive progenitor cell population of primary AML.

The PDT Regimen Spares Normal Hematopoietic Stem and Progenitor Cells

PTL was originally characterized as an anti-AML agent based in part on its ability to selectivity target malignant but not normal hematopoietic cells (22). Thus, it was important to determine the relative activity of the PDT regimen toward normal cells as well. To this end, we isolated mononuclear cells (MNCs) from normal bone marrow donor specimens and performed experiments analogous to the AML studies described in Fig. 5. Our results showed that all single, dual, and even the triple drug regimen PDT (94.3% viable) had only minimal toxicity toward total MNCs (Fig. 6, A and B), indicating no overt toxicity to bulk normal bone marrow cells. More importantly, the CD34+ population, which contains normal hematopoietic stem and progenitor cells (45), also showed no significant toxicity to PDT treatment (93.8%, Fig. 6, B and C). To readout functional abilities, we also performed a colony formation assay and demonstrated no significant targeting of normal progenitor cells (Fig. 6D).

FIGURE 6.

The PDT regimen is not toxic to normal hematopoietic cells. A, viability of total normal MNC cells after treatment with various drugs alone or in combination for 24 h. Data are plotted as % of untreated. Bars represent the mean ± S.D. (three individual MNC specimens). B, representative flow cytometry plots showing the impact of PDT treatment to both total MNCs and CD34+ MNCs. Viable MNCs were determined by double negative staining of 7-AAD and annexin V. Percentage of CD34+ cells were quantified within the viable MNCs. C, viable CD34+ normal MNCs after being treated with various drugs alone or in combination for 24 h. Data are normalized to untreated control. Bars represent mean ± S.D. (three individual MNC specimens). D, number of colonies in methylcellulose culture per 1 × 10⁶ normal MNCs treated with or without the PDT regimen. Bars represent the mean ± S.D. (n = 3). CFU, colony forming unit; UNT, untreated. In all different drug combinations including the PDT regimen, PTL, 2DG, and TEM were used at 2.5 μm, 0.1125 mg/ml, and 2.5 μg/ml, respectively.

The PDT Regimen Is Selectively Toxic to Leukemia Stem Cells

Finally, to measure the ability of the PDT regimen to target the leukemia stem cell population of primary AML, we treated primary AML cells with various combinations of PTL, 2DG, and TEM drugs for 24 h ex vivo and transplanted them into immune-deficient NSG (NOD.Cg-Prkdc^scid II2rg^tm1Wjl/SzJ) mice (to detect leukemia-initiating cell activity). We found that the ability of PDT-treated AML cells to engraft and repopulate in NSG mice decreased significantly (7.9 ± 1.9%) compared with either untreated (61.7 ± 4.2%), PTL alone (58.4 ± 7.5%), or dual drug combinations: PTL+2DG (27.0 ± 8.0%) and PTL+TEM (22.8 ± 13.3%) (Fig. 7A). In contrast, the same treatment with the PDT triple drug regimen resulted in no effect to the engraftment ability of normal mononuclear cells, indicating a minimal toxicity of the PDT regimen to normal hematopoietic stem cells (Fig. 7B). These results strongly indicate that the PDT triple drug regimen designed by our rational approach represents a potent and selective means to selectively target both bulk and stem/progenitor populations of primary AML.

FIGURE 7.

The PDT regimen preferentially targets LSCs but spares HSCs. A, engraftment potential of AML cells treated with various drug combinations in NSG mice. Each dot represents an individual mouse. Lines represent mean ± S.D. (n = 10). UNT, untreated. B, engraftment potential of normal MNCs treated with or without the PDT regimen in NSG mice. Each dot represents an individual mouse. Lines represent mean ± S.D. (n = 10). In all sub panels, the PDT regimen is composed of 2.5 μm PTL, 0.1125 mg/ml 2DG, and 2.5 μg/ml TEM. ns, not significant. C, a working model describing the anti-leukemia mechanism of the PDT regimen. ** indicates p < 0.01; *** indicates p < 0.001.

Discussion

The primary goal of this study was to demonstrate the rational design of a novel drug combination regimen to target human AML stem and progenitor cells. This is a challenging objective because of the well known cellular and molecular heterogeneity within different subpopulations of primary AML cells (1, 2). In particular, functionally defined AML stem cells are biologically distinct from bulk leukemic blast cells and are often less responsive to many forms of conventional therapy (10–13). Hence, an improved therapy must target broadly conserved properties among AML subpopulations, including the stem cell subsets. In addition to the heterogeneity of AML, a lack of systematic methodologies to design complex yet effective drug combinations is another major challenge in the drug discovery field.

In the current study, we addressed these challenges by employing and integrating multiple methods of global cell analysis to inform rational design of effective anti-leukemic regimens. To pilot our approach, we chose PTL, a compound that has already been shown to be effective against AML stem cells (22), as a foundation for designing a novel drug combination with superior anti-leukemic properties. Our global cell analysis methods involve multiple omic studies designed to thoroughly characterize the anti-leukemic mechanisms of PTL. To begin, we used a biotinylated analog of PTL to provide the first comprehensive characterization of the PTL proteomic interactome. Next, we profiled PTL-induced gene expression changes in primary AML cells at the transcriptome level. By analyzing the results from proteomic and transcriptomic studies together, we found that although the overall repertoire of PTL binding candidates is relatively broad, there is a striking prevalence for proteins directly involved in the Nrf2-mediated stress response and the protein ubiquitination pathways, which are significantly dysregulated by PTL. These findings suggest a strong link between the proteomic interactome and the transcriptomic response of PTL in AML cells. Most importantly, these data provide an unbiased means by which to pinpoint key anti-leukemic activities of PTL, including induction of oxidative and protein misfolding stress.

We then employed a comprehensive metabolomic analysis to further characterize the anti-leukemic mechanism of PTL. Both our uniformly labeled [U-¹³C₆]glucose and dual-labeled [¹³C_1,2]glucose experiments clearly demonstrate a strong increase in the PPP activity upon PTL treatment, presumably for the production of NADPH. Notably, NADPH is also known to be produced by other metabolic steps including the conversion of malate into pyruvate catalyzed by the NADP-dependent malic enzyme 1 (ME1) and the conversion of citrate into α-ketoglutarate catalyzed by the isocitrate dehydrogenase 1 enzyme (IDH1) (46). In our current study, although we cannot exclude the possibility that these processes may also contribute to NADPH generation, given the strong increase of the PPP activity shown in Fig. 2, B and C, and supplemental Fig. 1, we propose that the increased PPP activity is a major factor contributing to elevated NADPH production.

In addition to the strong increase in the PPP, our metabolic analysis also shows a global increase of glycolysis, lactate production, and OXPHOS. Given that the glycolysis, PPP, and OXPHOS pathways are all interlinked by metabolic intermediates, we propose that the increase in OXPHOS and glycolysis intermediates is a likely result of the PPP feeding back to glycolysis and eventually into the TCA cycle. Importantly, the increase in the PPP is relevant because we also observed a clear NADPH increase upon PTL treatment. In addition, we showed in our dual-labeled [¹³C_{1, 2}]glucose experiment, that the M+1 isotopologue of [¹³C]glyceraldehyde-3-phosphate (could only be generated by the PPP) is specifically elevated in PTL-treated AML cells but not in untreated cells (Fig. 2C and supplemental Fig. 1). These studies demonstrate that the most dominant metabolic response of AML cells to PTL treatment is the up-regulation of the PPP for NADPH production, which is known to detoxify oxidative stress (37). Thus, together our metabolomic data further supports induction of oxidative stress as a central component of PTL's anti-leukemic activity. Consequently, we used oxidative stress as the focal point in the rational design of PTL-based drug combinations.

With a detailed understanding of the PTL mechanism in hand, we further integrated the analyses from all omic studies and identified two key opportunities for developing an enhanced PTL-based drug regimen: 1) the Nrf2-mediated oxidative stress response pathway (readily evident from analyses of both PTL's proteomic interactome data and PTL-induced transcriptome change data) and 2) the PPP (metabolomic data showed a dramatic increase of PPP activity as a major cytoprotective response of AML to PTL insult). We hypothesized that the addition of drugs to simultaneously inhibit both the Nrf2-mediated anti-oxidant response and the PPP would enhance the selective anti-leukemic activity of PTL. To this end we designed a triple drug regimen termed PDT that is composed of parthenolide, 2-deoxyglucose (to inhibit NADPH production from the PPP), and temsirolimus (previously shown to inhibit Nrf2-mediated responses in AML cells (31)). We showed that this PDT regimen displayed potent toxicity toward both bulk and stem/progenitor populations of primary AML cells but very limited toxicity toward the normal hematopoietic cells. Notably, the PDT regimen was effective even for AML specimens that were relatively resistant to PTL alone. Thus, not only was PDT more active toward AML cells, but it was also more broadly effective among varying AML specimens.

From a mechanistic perspective, the anti-leukemic activity of PDT is associated with its strong ability to induce oxidative stress without activating the compensatory Nrf2 and PPP responses in primary AML cells. We postulate this activity at least partly explains its selectivity because multiple studies have suggested that an aberrant redox homeostasis is a common characteristic of primary AML cells compared with normal hematopoietic cells. For example, we previously showed that CD34+ primary AML cells (enriched for stem and progenitor AML cells) and CD34+ normal hematopoietic cells have substantial differences in levels of reduced glutathione and relative response to PTL insult. Intriguingly, the steady state level of reduced glutathione is higher in normal cells and is also restored more quickly upon oxidative insult (25). Thus, it appears that AML cells exist in an intrinsically higher oxidative state and also have less capacity to manage further oxidative stress. Our findings are supported by other studies which demonstrate that oxidative stress is a common property of AML and increased oxidative stress appears to be a key event during leukemic transformation (47–49). Thus it is likely that AML stem cells are more sensitive to PDT-induced oxidative stress due to their higher basal level of ROS relative to normal hematopoietic stem cells. Furthermore, as shown in the present study, by disabling compensatory responses of AML (Nrf2 and PPP), the activity of agents such as PTL can be substantially enhanced.

Interestingly, the addition of temsirolimus seems to inhibit parthenolide's effect in reducing NF-κB p65 phosphorylation. We think this is likely caused by a cross-talk between oxidative stress and NF-κB signaling. Both the current study and previous studies have shown that temsirolimus can induce oxidative stress by suppressing antioxidant responses, and oxidative stress can activate NF-κB signaling (31, 50). Thus we think that the addition of temsirolimus rescued PTL-induced NF-κB inhibition through suppression of antioxidant response and subsequent induction of oxidative stress that finally reactivated the NF-κB signaling. As a consequence, PDT-induced cell death in primary AML cells is independent of NF-κB inhibition. Although primary AML cells are known to have constitutively active NF-κB activity that can be targeted to enhance cell death (22, 51), our data suggest that it is also possible to induce severe cell death in primary AML cells by generating other forms of cytotoxic stimuli such as protein misfolding stress.

We show in the current study that in addition to generating oxidative stress, PDT can also induce strong protein misfolding stress in both the cytoplasm and ER of primary AML cells. Protein misfolding stress in the cytoplasm is usually detoxified by chaperone proteins, which refold misfolded proteins, and proteasomes that degrade and recycle misfolded proteins (52, 53). On the other hand, protein misfolding can also happen in the ER where nascent protein synthesis takes place. The accumulation of misfolded proteins in the ER can trigger the “death arm” of the unfolded protein response, thereby leading to apoptosis (54). In our study we observed global up-regulation of chaperone proteins, proteasome subunits, and ER stress markers in PDT-treated primary AML cells, indicating a strong ability of PDT to induce protein misfolding stress in AML cells. Importantly, studies have suggested that leukemia cells might already be under higher protein misfolding stress compared with normal hematopoietic cells. For example, Dai et al. (55) have reported that heat shock proteins are often up-regulated in cancer cells and are particularly high in hematopoietic malignancies. Our own studies have also found that the expression of HSP60, HSP70, and HSP90 were consistently higher in multiple primary AML cells compared with CD34+ normal hematopoietic cells (data not shown). Consistent with these observations, HSP90 inhibitors 17-(Allylamino)-17-demethoxygeldanamycin (17-AAG) and 17-Dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG) have both been shown to be effective against AML cells (56). Together, these data suggest that PDT-induced protein misfolding stress might be another mechanism underlying its potency and selectivity against primary AML specimens.

Simultaneous induction of multiple types of stress that can converge to synergistically promote apoptosis is an advantage of using drug combinations to treat cancer. From this perspective, we propose that the oxidative stress and protein misfolding stress induced by PDT are likely interconnected and collaborate to create amplified apoptotic signaling in primary AML cells. This reasoning is largely based on the fact that proper protein folding is a redox-sensitive process subjected to redox modifications including S-glutathionylation, carbonylation, etc. (39, 40). Thus, excessive exposure of the AML proteome to sustained oxidative stress will likely lead to dramatic protein misfolding stress increase as a consequence.

Finally, we summarize the proposed mechanism of PDT-mediated cytotoxicity in the diagram shown in Fig. 7C. Briefly, our data show that PTL mediates a potent inhibition of glutathione metabolism (25), which triggers activation of Nrf2-mediated oxidative stress response as well as increased utilization of the PPP for NADPH production. 2DG blocks the PPP, reduces the production of NADPH, and thus inhibits the conversion of oxidized glutathione (GSSG) to reduced glutathione (GSH). In parallel, TEM blocks translocation of Nrf2 to the nucleus, thereby inhibiting the synthesis of anti-oxidant machineries (i.e. GCLM, GPX1, TXN), which leads to further ROS increase (31). Although not directly examined in the present study, it is likely that the increase in ROS promotes ER stress via a redox-mediated mechanism (39, 40). And finally, PDT-induced oxidative and protein misfolding stress synergize to induce potent cell death in primary AML cells. Importantly, because AML cells are likely more susceptible than normal cells to increases of ROS and protein misfolding stress, we propose that these mechanisms also explain the selectivity of the PDT regimen.

Experimental Procedures

Human Specimens

AML specimens were obtained from apheresis product, peripheral blood, or bone marrow of AML patients who gave informed consent for sample procurement on the University of Colorado tissue procurement protocol. Normal bone marrow and normal cord blood specimens were obtained from volunteer donors who gave informed consent on a research review board-approved protocol at the University of Colorado. Total MNCs were isolated from normal bone marrow or normal cord blood donor specimens by standard Ficoll procedures (GE Healthcare). If needed, total MNCs were further enriched for CD34-positive cells using the MACS CD34 enrichment kit (Miltenyi Biotec). See supplemental Table 7 for additional details on human specimens.

Cell Culture and Drug Treatments

Primary human AML cells and MNCs were cultured and treated at 1 million cells per ml in Iscove's modified Dulbecco's medium (IMDM; Life technologies)-based serum-free media at 37 °C, 5% CO₂ incubator as previously described (25). Cells were preincubated in serum-free media for 1 h before treatment with parthenolide (Enzo), temsirolimus (Pfizer), and 2-deoxyglucose (Sigma) at desired concentrations. For all drug combinations, drugs were added at the same time.

Preparation of Subcellular Fractions and Total Cell Lysates

Subcellular fractionation of primary AML cells was performed using the Qproteome Cell Compartment kit (Qiagen) per manufacturer's protocol. Total cell lysates were made by performing cell lysis in Buffer F (10 mm Tris-HCl, pH 7.5, 50 mm NaCl, 30 mm sodium pyrophosphate, 50 mm NaF, 5 μm ZnCl2, 1% Triton X-100, freshly add 1 mm PMSF, 1× proteinase inhibitor cocktail, and 0.1 mm Na₃OV₄) at 10–20 million per ml.

Streptavidin Pulldown Assay

Total cell lysates and subcellular fractions from MMB-biotin-treated cells were incubated with SA beads (Thermo Fisher Scientific) at 4 °C for 2 h to pull down binding targets of PTL. After incubation, beads were sequentially washed 1 time with PBS, 3 times with high salt wash buffer (500 mm NaCl in 0.1 m pH 5.0 NaOAc), 3 times with low pH wash buffer (0.1 m pH 2.8 glycine-HCl), and 1 last time with PBS. Finally, the SA beads were boiled for 10 min in 2× SDS-PAGE sample buffer to elute all pulldown products for Western blot analysis or digested with trypsin to allow subsequent LC-MS/MS-based identification of targets.

LC-MS/MS-based Identification of the Proteomic Interactome of PTL

After SA pulldown, on-bead trypsin digestion was performed to generate peptide fragments of targets. These fragments were analyzed by nano-spray LC-MS/MS using Magic C18 AQ reverse-phase liquid chromatography resin (Michrom BioResources) custom packed into a 5-cm × 75-μm fused silica column that was coupled in-line to an ion trap mass spectrometer (Finnigan LTQ, Thermo Fisher). MS/MS-acquired data were searched against human amino acid sequences within the NCBI protein database using the MASCOT software (Matrix Science). This method identified a pool of candidate proteins from total cell lysates and all four subcellular fractions of MMB-biotin-treated primary AML cells. To exclude nonspecific binding events, pulldown products from biotin control-treated total cell lysates were also identified through the same LC-MS/MS platform. The final list of 312 proteins (supplemental Table 1) contains targets that are uniquely identified in the MMB-biotin-treated cells but not biotin control-treated cells.

Quantitative Real-time PCR

Total mRNA was isolated with the RNeasy plus mini kit (Qiagen) according to the manufacturer's instructions. mRNA purity and quantity were determined with NanoDrop (Thermo Fisher Scientific). mRNA samples were reverse-transcribed into cDNA using the iScript One-Step RT-PCR kit (Bio-Rad). Quantitative real-time PCR was performed with LightCycler480 real-time PCR using LightCycler 480 SYBR Green I Master Mix reagent (Roche Applied Science). Primer sequences used for QPCR analysis are listed in supplemental Table 8.

RNA-seq

The TruSeq RNA Sample Preparation Kit V2 (Illumina) was used for next generation sequencing library construction per the manufacturer's protocols. Amplicons are ∼200–500 bp in size. The amplified libraries were hybridized to the Illumina single end flow cell and amplified using the cBot (Illumina). Single end reads of 100 nucleotides were generated for each sample and aligned to the organism specific reference genome. Raw reads generated from the Illumina HiSeq2500 sequencer were de-multiplexed using configurebcl2fastq.pl version 1.8.4. Quality filtering and adapter removal was performed using Trimmomatic version 0.32 with the following parameters: “SLIDINGWINDOW:4:20 TRAILING:13 LEADING:13 ILLUMINACLIP:adapters.fasta:2:30:10 MINLEN:15”. Processed/cleaned reads were then mapped to the UCSC hg19 genome build with SHRiMP version 2.2.3 with the following setting: “–qv-offset 33 -all-contigs.”

Extraction of Metabolites and Metabolomic Analysis

DMSO- or PTL-treated cells were pelleted and immediately extracted in ice-cold lysis/extraction buffer (methanol:acetonitrile:water 5:3:2) at 2 million cells per ml. Samples were then agitated at 4 °C for 30 min and centrifuged at 10,000 × g for 15 min at 4 °C. Protein and lipid pellets were discarded, whereas supernatants were stored at −80 °C before metabolomic analyses. Metabolomic analyses were performed as previously reported (57, 58). 20 μl of each sample was injected into an UHPLC system (Ultimate 3000, Thermo Fisher Scientific) and separated during a 3-min isocratic gradient at 250 μl/min (mobile phase: 5% acetonitrile, 95% 18 mΩ H₂O, 0.1% formic acid) on a Kinetex C18 column (150 × 1-mm inner diameter, 1.7-μm particle size; Phenomenex). The UHPLC system was coupled online with a QExactive system (Thermo Fisher Scientific) scanning in full MS mode (2 μscans) at 70,000 resolution in the 60–900 m/z range, 4-kV spray voltage, 15 sheath gas and 5 auxiliary gas operated either in negative and positive ion mode. Calibration was performed before each analysis against positive or negative ion mode calibration mixes (Thermo Fisher Scientific) to ensure sub-ppm error on the intact mass. Upon conversion of .raw files into .mzXML format using MassMatrix, metabolite, and isotopologue assignments were performed using the software Maven following the rationale described by Buescher et al. (38). Assignments were further confirmed by chemical formula determination from isotopic patterns and accurate intact mass and retention times against over 650 standards, including commercially available glycolytic and Krebs cycle intermediates, amino acids, glutathione intermediates, and nucleoside phosphates (Sigma, IROATech).

Cell Viability Assay

After each treatment cells were washed with ice-cold FACS buffer (PBS with 0.5% FBS) and then stained for 15 min at 4 °C in FACS buffer containing antibodies against human CD34 (BD Biosciences). After staining, cells were washed with ice-cold FACS buffer and then stained in annexin-V binding buffer (10 mm HEPES/NaOH, pH 7.4, 140 mm NaCl, 2.5 mm CaCl₂) containing annexin-V (BD Biosciences) and DAPI (Sigma). In some cases viability was measured by annexin-V and 7-Aminoactinomycin D (7-AAD) (Sigma) staining instead. For cells that did not need CD34 staining, after FACS buffer wash, cells were directly stained with viability dyes. Stained cells were analyzed immediately on a LSRII flow cytometer (BD Biosciences). Viable cells were scored as annexin-V and DAPI/7-AAD double negative cells.

Colony Formation Assay

Primary AML cells and normal MNCs were plated in human methylcellulose complete media (R&D Systems) at 2.5 × 10⁴ and 5 × 10⁴ cells per ml, respectively. Colonies were counted and scored after 3 weeks of culture at 37 °C, 5% CO₂ incubator.

Ex Vivo Treatment and Xenograft Engraftment Assay

Overnight-treated cells were washed and resuspended in FACS buffer at 50 million cells per ml. About 5 million cells per mouse were injected intraperitoneally into NSG mice. At ∼7–10 weeks post injection, all groups of mice were sacrificed, their bone marrow cells were harvested, and the total percentage of human CD45+ cells was quantified to determine engraftment potential.

Silver Stain and Western Blot

Silver stain was performed using the Pierce Silver Stain kit (Thermo Fisher Scientific). After gel transfer, blots were probed with SA-HRP (Thermo Fisher Scientific). To detect specific antigens, blot were probed with primary antibodies against IKBα, histone H3, caspase-3, phosphor-p65-Ser-536 (Cell Signaling), IL6Rα, tubulin, actin (Santa Cruz), and HMOX1 (Stressgen) on a shaker at 4 °C, overnight, followed by 2 h of room temperature incubation with HRP-conjugated secondary antibodies (Santa Cruz). Chemo-luminescence was recorded using the automated Gel Doc XR+ system (Bio-Rad) or x-ray films (Thermo Fisher Scientific).

NAPDH/NADP+ Measurement

Quantification of NADPH and NADP+ were performed using the NADP/NADPH-Glo Assay kit (Promega) according to the manufacture's protocol.

Glutathione Measurement

Glutathione quantification was performed using the Glutathione Colorimetric Assay kit (Biovision) according to the manufacturer's protocol.

Cellular ROS Measurement

Cells were incubated at 37 °C in ROS-staining buffer (PBS+2%FBS) containing the CM-H₂DCFDA probe (1μm) (Life Technologies) for 30 min. After incubation, cells were washed two times with ROS-staining buffer before FACS analysis.

NF-κB EMSA Assay

NF-κB DNA binding activity in cell nuclear extracts was measured using electrophoretic mobility shift assay as previously described (59). Briefly, an oligonucleotide corresponding to the consensus sequence (5′-AGTTGAGGGGACTTTCCCAGGC-3′) to NF-κB was radiolabeled and incubated with nuclear extract. The DNA-protein complexes were resolved in 5% native PAGE in low ionic strength buffer. The gels were dried, and protein binding was visualized by autoradiography.

Pathway Analysis Using IPA, GSEA, and iPAGE

The list of 312 PTL-binding targets (supplemental Table 1) and the list of 2114 significantly dysregulated genes induced by PTL (supplemental Table 3) were used to perform the IPA (Qiagen) core analysis with default settings. GSEA analysis (60) was run on the c2.cp.kegg.v5.0 or the c2.cp.reactome.v5.0 genesets using 1000 permutations of the genesets. Gene sets with <15 genes or >500 genes were excluded. Gene sets with a false discovery rate ≤0.25 and a nominal p ≤ 0.05 were considered significant. iPAGE method was used to identify over and under-represented pathways (42) using RNA-seq datasets of PDT-induced AML samples. Briefly in iPAGE, we quantized continuous expression data into equally populated bins. iPAGE then calculated the mutual information (MI) between a vector of expression values and a binary vector of pathway memberships for every pathway. The significance of the calculated mutual information values was then assessed through a randomization-based statistical test. We then used hyper-geometric distribution to determine the level with which the significantly informative pathways are over-represented or under-represented in each expression bin or cluster. We used the resulting p values to draw a heat map, in which rows represent significant pathways, and columns correspond to expression bin/clusters. In the heat map, red entries correspond to pathway over-representations, whereas blue entries correspond to under-representations.

Statistics

Unless otherwise indicated, statistical analyses were performed using two-tailed (non-directional), type three (unequal variance) Student's t test. A p value of <0.05 indicates significance.

Study Approval

All experimental materials and procedures involving human and animals in this study were reviewed and approved by University of Colorado under the approval number of 12-0173 and B-103413(09)1E, respectively.

Author Contributions

S. P. and C. T. J. designed the study, analyzed the data, and wrote the paper. S. P. carried out MMB-biotin pulldown assay, target identification, and IPA analysis. F. K. H. performed the LC-MS/MS characterization. S. P., A. D., T. N., and K. C. H. carried out the metabolomic assay and data analyses. S. P. carried out the glutathione, ROS, QPCR, Western blot, and NADPH assays related to Fig. 3. S. P. and M. M. carried out the NFκB EMSA assay, p-p65 WB, and QPCR assays to determine the activity of PDT to NFκB signaling and ER stress. S. P. and J. M. A. carried out the RNAseq analysis. S. P. carried out the in vitro toxicity assay, cfu assay, and engraftment assay to determine the selective antileukemic activity of PDT. B. M. S., B. A., and N. K. helped with the engraftment assay and flow cytometry analysis. V. K. Y. and S. D. carried out the iPAGE analysis. J. A. G., D. A. P., and C. S. provided primary AML specimens. P. A. C. provided MMB-biotin.