Abstract
Quiescent (G0) cells are transient cell-cycle-arrested subpopulations in cancers that include dormant cancer stem cells. While leukemic stem cells/minimal-residual-disease were studied transcriptionally, their translation profile and post-transcriptional mechanisms that control their proteome—and thereby, survival—are unknown. We find G0 leukemic cells are chemoresistant, with altered canonical translation and similar proteome and translatome—rather than transcriptome—to cells isolated post-chemotherapy, implicating post-transcriptional regulation in chemoresistance. Mechanistically, we find DNA-damage-responsive-ATM and stress-activated-p38-MAPK/MK2 alter post-transcriptional mechanisms, regulating mRNA-decay-factor, TTP, to increase AU-rich-element-bearing mRNAs. This permits translation of AU-rich-element-bearing pro-inflammatory-cytokine TNFα, and immune modulators (CD47) that promote survival. Co-inhibiting p38-MAPK/MK2 and TNFα—prior to/alongwith chemotherapy—decreases chemoresistance in vivo and in patient samples. Disrupting TTP regulation reduces TNFα and chemoresistance, revealing TTP as a regulator of inflammation-mediated chemoresistance. These studies uncover a pro-inflammatory subpopulation in cancer that enables chemoresistance via DNA-damage- and stress-regulated post-transcriptional mechanisms, and develop a new combination therapy against chemo-survival.
Statement of Significance These studies reveal the significance of post-transcriptional regulation of inflammatory/immune response in chemoresistance, mediated by AU-rich-elements, mRNA-decay-regulator TTP, and non-canonical translation. These results reveal DNA-damage/stress-induced TTP as a key regulator of inflammation-mediated chemoresistance and developed a new combination therapy to reduce resistance in cell lines, patient samples and in vivo.
Introduction
Quiescent (G0) cells are an assortment of reversibly arrested cells, including dormant stem cells, which are found as a clinically relevant subpopulation in cancers (1-4). Such cells are anti-proliferative, anti-differentiation, and anti-apoptotic, and show distinct properties including resistance to harsh conditions (1, 2, 5-10). G0 cells show specific gene expression that may underlie their resistance and other properties (1, 2, 8-10). Analyses from multiple groups revealed some genes up-regulated at the transcriptional level (1, 8, 11). Altered polyadenylation site selection on mRNAs produces longer 3′-untranslated regions (3′UTRs) in G0 compared to proliferating cells— which increases 3′UTR elements that can mediate post-transcriptional gene expression regulation (12). Our previous data demonstrated that translation mechanisms are distinct in G0 leukemic cells, with decreased canonical translation mechanisms and increase in mRNA translation by alternative mechanisms that involve non-canonical translation initiation factors (13) and 3’UTR mediated specific mRNA translation (14). These data suggest that alternate post-transcriptional mechanisms in G0 cancer cells may regulate a distinct translatome to mediate their resistance. Translated genes, post-transcriptional mechanisms involved, and outcomes on cancer persistence remain to be investigated.
We analyzed the translatome and proteome of chemotherapy-surviving G0 cancer cells, focusing on acute monocytic leukemia (AML), to provide comprehensive information that complement and expand previous transcriptome analyses (1, 2, 8, 11, 15, 16), (Fig. 1A-B), to reveal critical genes that are post-transcriptionally regulated for chemo-survival. G0 can be induced by growth factor-deprivation or serum-starvation and other conditions that isolate dormant cancer stem cells in distinct cell types (1, 6, 7). Our data demonstrate that serum-starvation induced G0 AML cells are chemoresistant—similar to surviving AML cells, isolated after chemotherapy. Chemoresistant cells isolated via serum-starvation, or as surviving cells post-chemotherapy, show inhibition of canonical translation mechanisms, indicating that non-canonical mechanisms express specific mRNAs when these cells are chemoresistant. Consistently, the translatomes and proteomes of serum-starved G0 and chemo-surviving cells show greater similarity than the transcriptome alone. These data indicate a specific post-transcriptional program common between serum-starvation induced G0 cells that exhibit resistance, and chemosurviving cells—which may underlie their shared property of chemoresistance.
We find that chemotherapy and serum-starvation induce DNA damage response (DDR) and p38 MAPK stress signaling that lead to up-regulation of specific genes via post-transcriptional mechanisms (17-19) (20-24). Our data reveal AU-rich elements (AREs) enriched in mRNAs that are up-regulated in G0 and in chemo-surviving cells—due to p38 MAPK activation (22, 25, 26) of MAPKAP2 (MK2) kinase (27-29). MK2 regulates ARE binding mRNA decay and translation repression factor, Zinc finger 36 homolog (ZFP36) or Tristetraprolin (TTP) (27-30). We find that p38 MAPK/MK2 mediated regulation of TTP leads to increased ARE bearing mRNA levels and translation in chemoresistant G0 cells. DNA damage-induced ATM kinase and stress signaling—inhibit canonical translation mechanisms in chemoresistant cells, which permits specific gene translation by non-canonical mechanisms that we observed previously (13, 14). The post-transcriptionally regulated genes expressed in G0 resistant cells, include pro-inflammatory genes of the TNFα/NF к B pathway—which we find increase cell survival genes and are required for chemo-survival. Consistently, decreasing expression of TNFα, by expressing a non-regulatable, active mutant TTP (27-29), reduces chemoresistance. Importantly, correlating with their early induction in serum-starved G0 or chemotherapy-treated cells, pharmacological inhibition of stress signaling and the key translated inflammatory response gene TNFα—prior to or along with chemotherapy—significantly reduces chemoresistance in AML cell lines, patient samples, and in vivo. Our data reveal that DNA damage and stress signaling cause post-transcriptional alterations to produce a specialized gene expression program of pro-inflammatory, immune effectors that elicit chemo-survival.
Results
Serum-starvation or AraC treatment induces a quiescent and chemoresistant state of leukemic cells
To study clinical resistance in cancer, THP1 human AML cells were used as they show significant resistance to cytarabine (31) (cytosine arabinoside, AraC, Fig. S1A), a standard anti-leukemic chemotherapeutic that targets DNA replication and thus proliferating cells (referred to as S+). Our data and others find that serum-starvation of THP1 (13) and other cell lines (1, 8, 11, 32) induces a transient G0 state with known G0 and cell cycle arrest markers expressed (Fig. 1C-D, S1B-C). Such serum-starvation induced G0 cells (referred to as SS) can be returned to the cell cycle upon serum addition (Fig. 1D), verifying that they are quiescent and transiently arrested, unlike senescence or differentiation that are not easily reversed. We find that serum-starvation induced G0 SS cells show resistance to AraC chemotherapy. Serum-grown S+ cells show a dose-dependent decrease in cell viability with AraC as expected, while SS cells persist, indicating their chemoresistance (Fig. 1E).
Chemoresistant cancer cells include cancer stem cells, and are a subpopulation that can be isolated from cancers after treatment with chemotherapy (2, 6-10) that targets and eliminates S+ cells. We find that AraC-surviving THP1 (referred to as AraCS) cells are transiently arrested, like serum-starved SS cells (Fig. 1C-D, S1B); both AraCS and SS cells survive chemotherapy. AraCS cells recover from their transient arrest upon AraC removal, proliferate (Fig. 1D) and give rise to resistance upon subsequent AraC treatment (Fig. 1E), affirming the reversible G0 arrest state of chemoresistant cells, similar to SS cells (1, 2, 6-10).
Chemoresistant G0 cells induced by SS or AraC have similar translatomes and proteomes that recapitulate gene expression profiles of in vivo chemoresistant leukemic and G0 models
We asked whether the underlying gene expression is similarly regulated in both SS and AraC cells, leading to their common property of chemoresistance. We previously observed canonical translation shutdown that allowed non-canonical translation of specific genes in SS cells (13, 14), indicating post-transcriptional changes in such cells. Therefore, to differentiate translationally-regulated from transcriptionally-regulated genes associated with chemoresistance, we profiled AraCS cells and SS cells—and compared them with S+ cells at the proteome, translatome and transcriptome levels using: multiplexed quantitative proteomics (14), microarray analysis of heavy polysome-associated mRNAs (13, 14, 33) for the translatome, and total RNA microarray for the transcriptome (Fig. 1A-B, S1D-F). Consistently, we find that AraCS and SS cells show significantly similar gene expression profiles in their proteomes and translatomes rather than transcriptomes (Fig. 1F). These data suggest that although these chemoresistant G0 cells are isolated via two different methods: therapy survival or serum-starvation, they exhibit a common set of specific genes expressed at the protein level, which could underlie their common characteristic of chemoresistance. These data indicate the relevance of examining the translatome and post-transcriptional regulation of gene expression, in addition to the transcriptome. Time-course translatome analysis revealed that early serum-starved G0 (4 h and 1 day SS) cells are distinct from late, 2-4 days G0 cells (2-4 days SS are more similar in their translatomes, R2=0.81, Fig. S1E-F). This is consistent with G0 as a continuum of assorted, arrested states (1), with temporal differences in underlying gene expression in early G0 compared to more homogeneity at late G0.
SS and AraCS cells provide sufficient material to perform concurrent translatome, proteome and transcriptome profiling for post-transcriptional regulation analyses, compared to limited cells from in vivo resistance models. To test whether our G0 leukemic cells are relevant models to study chemoresistance and G0, gene expression profiles of AraCS and SS cells were compared to published transcriptome profiles of in vivo chemoresistance models, including leukemic stem cells, dormant leukemic cells and chemosurviving cells, as well as to a published G0 model that were isolated by different methods: leukemia stem cells (LSC) from AML (16), dormant leukemic cells (LRC), and minimal residual disease (MRD) from chemotherapy surviving patient samples with acute lymphocytic leukemia (ALL) (15), as well as SS G0 fibroblasts (G0 HFF) (1). Importantly, we find that these published transcriptome signatures for in vivo chemoresistance and G0 models were significantly up-regulated in our SS and AraCS cells (referred to as resistant G0 leukemic cells), compared to S+ cells (Fig. 1G, S1G). These data indicate that our resistant G0 leukemic cells are relevant models to study post-transcriptional regulation in chemoresistance as they have similar gene expression profiles to known transcriptional profiles from in vivo chemoresistance models.
Altered translation with inhibition of canonical translation mechanisms in resistant G0 leukemic cells
As previously observed in SS cells (13, 34-39), we find overall protein synthesis is reduced at least 2-fold in AraCS, compared to S+ cells (Fig. 2B, S1D). Mechanistically, both rate-limiting steps in canonical translation initiation: recruitment of initiator tRNA, and mRNA cap recognition to recruit mRNAs to ribosomes are inhibited in G0 leukemic cells (Fig. 2A-C). Recruitment of initiator tRNA by eIF2 can be blocked by eIF2α phosphorylation as a stress response (13, 34-39). We find that two eIF2 kinases, PKR and PERK, are activated and increase eIF2α phosphorylation (Fig. 2A, 2C) to inhibit canonical translation initiation in G0 leukemic cells. Our previous data revealed low mTOR activity (14), leading to dephosphorylated and active eIF4EBP (4EBP) (Fig. 2A, 2C) that inhibits cap-dependent translation initiation (40, 41). Low mTOR activity reduces translation of terminal oligopyrimidine tract (TOP) mRNAs such as ribosomal protein mRNAs (40, 42, 43), which we find decreased in SS and AraCS cells (Fig. 2A, 2D). Decreased canonical translation can enable post-transcriptional regulation of specific genes, as observed previously (13, 14) and lead to survival of G0 leukemic cells.
Global translatome analysis shows that inflammatory response genes are selectively translated in resistant G0 cancer cells
We measured the number of genes upregulated at the transcriptome, translatome and proteome levels in resistant G0 leukemic cells, compared to S+ cells. A significantly greater number of genes were upregulated in the translatome (580 genes, Table S1) and proteome (716 genes), compared to the transcriptome (318 genes) as shown in Fig. 2E. Importantly, 70% of upregulated genes (418 out of 580) were increased in the translatome (Fig. 2F) but not in the transcriptome, indicating a common set of specific genes that are post-transcriptionally/translationally expressed. To investigate the biological function of these differentially expressed genes, gene ontology (GO) analysis was performed with each gene set. Gene categories up-regulated in G0 translatomes include inflammatory response, immune modulators (receptors, antigen presentation and processing genes), cell adhesion, cell migration, lipid biosynthesis and cholesterol pathway genes (Fig. 2G, S2E). Down-regulated genes include RNA processing and ribosome genes, consistent with the limited, specialized gene expression here, as well as decreased DNA repair genes that would permit DNA damage and stress signaling (Fig. 2G). To identify translationally up-regulated genes, we measured the change in ribosome occupancy, RO, which is the ratio of polysome-associated mRNA levels to total mRNA levels of each gene (Fig. 2F, Venn diagram, heat map). We find 180 genes are translationally up-regulated above RNA level changes in G0 (ΔRO = RO G0 – RO proliferating > 1.5-fold). These genes include antigen processing and presentation genes (44) (HLA-G, HLA-E) and immune receptors (CD47, Fig. 2F-G, S2I) (45-47) that regulate anti-tumor immune response and are associated with leukemic stem cells and resistance (48, 49).
We asked if this specific gene expression profile in resistant G0 leukemic cells is conserved in G0 cells of other tumors and in normal cells. Therefore, global translatome profiling was conducted in G0 cells from four different cells lines: breast cancer (MCF7, Fig. S2B-D), liver cancer (HEP-G2), and osteosarcoma (U2OS) as well as non-cancerous fibroblasts (HFF) (Fig. S2A, D). Their translatomes were compared with resistant G0 leukemic cells, using GSEA and DAVID tools (Fig. 2H-I, S2E-F). We find that 580 signature genes of resistant G0 leukemic cells (Table S1) were highly conserved at the translatome level in G0 cells of these other cell types (Fig. 2H). As expected for these arrested cells, genes related to cell cycle, ribosome biogenesis, and DNA replication were commonly down-regulated (Fig. 2I, S2E). Importantly, inflammatory response genes were commonly up-regulated in cancer G0 cells but not normal G0 fibroblasts and do not significantly overlap with senescence-associated secretory pathway (SASP) (Fig. 2I, S2G) (50, 51), indicating a specific role in chemoresistant cancer cells. These results indicate that specialized post-transcriptional mechanisms are enabled in these conditions of decreased canonical translation and lead to common expression of specific genes; in particular, of inflammatory response genes in chemoresistant G0 cancer cells.
Post-transcriptional up-regulation of ARE-bearing mRNAs is mediated by phosphorylation of TTP in resistant G0 leukemic cells
To identify RNA motifs that mediate post-transcriptional regulation, the untranslated regions (UTRs) of differentially expressed genes in G0 leukemic cells were examined. Three UTR elements were found: two 5′UTR elements and a key post-transcriptional regulatory 3′UTR AU-rich element. A GC-rich motif was enriched on 5′UTRs of translationally up-regulated genes (ΔRO > 1.5) and an AU rich motif, on 5′UTRs of translationally suppressed genes (ΔRO < 1.5), indicating that mRNAs with structured 5′UTRs are highly translated in G0 cells (Fig. S3A-B). Importantly, 3′UTRs of genes in the up-regulated translatome have higher AU-rich elements (AREs) scores, compared to down-regulated translatome (Fig 3A). Over 30% of the translatome signature of G0 leukemic cells bear AREs, including key proinflammatory cytokines/chemokines, such as TNFα (Figs. 3B-D, Table S2) (25). AREs regulate mRNA stability and translation to control expression of critical growth factors and immune genes (52, 53). AREs are involved in mediating specific mRNA decay; however, depending on their associated RNA binding proteins and cellular conditions that signal and modify these interactions, AREs can promote mRNA stability and translation, as in the case of TNFα and other mRNAs in G0 and other conditions (14, 28, 52, 54-56). Consistently, we find that most ARE binding proteins that are RNA decay or translation repression factors (Fig. S3C, S3F) (57, 58), are decreased in SS and AraCS cells. Additionally, we find decrease of the exosome RNA decay complex (Fig. S3D) that is involved in ribosome processing (59) and ARE mRNA decay (28, 29, 60), and of proteasome factors (61) (Fig. S3E) that are implicated in ARE mRNA decay. A key ARE mRNA decay factor, TTP, was surprisingly increased in G0 leukemic cells (Fig. S3F, 3B-C, 3E-F). TTP phosphorylation is established to increase its levels (28), and results in inability of phospho-TTP to down-regulate ARE mRNA expression; such ARE mRNAs are stabilized and translated (29, 30). Accordingly, we find that TTP is phosphorylated in SS and AraCS cells (Fig. 3F), accounting for its increased levels (Fig. S3F, 3B-C, 3E-F). To test whether phosphorylation of TTP was required for expressed ARE mRNAs in SS and AraCS cells, we over-expressed non-regulatable active mutant TTP that has its key phosphorylation sites (Ser 52, 178) mutated to alanine (TTP-AA). TTP-AA has been shown to dominantly maintain ARE mRNA decay activity and reduce pro-inflammatory cytokines like TNFα in immune cells (28-30). Over-expression of myc-tagged TTP-AA compared to control vector significantly reduced TNFα mRNA in both THP1 and K562 AraCS cells (Fig. 3G), overturning the decay inactivity of endogenous phospho-TTP. These data indicate that inactivation of ARE mRNA decay by TTP phosphorylation (27-29) is a key regulator of pro-inflammatory gene/TNFα expression in chemoresistant G0 cells. These results are consistent with our findings of increased levels and translation of ARE bearing mRNAs—including many pro-inflammatory cytokines (Fig. 2G, 3A-C)—due to decreased ARE decay activity in SS and AraCS cells (Fig. 3F-G, S3C-F).
Activation of p38 MAPK/MK2 phosphorylates TTP to promote expression of ARE-bearing mRNAs in resistant G0 leukemic cells
To investigate how TTP is phosphorylated in resistant G0 leukemic cells, we examined key signaling molecules involved in DNA-damage response (DDR) or stress (20) since chemotherapies like AraC trigger DNA damage-mediated induction of ATM kinase and stress, which activate p38 MAPK (17-20). P38 MAPK signaling is also activated by G0 and serum-starvation (62). As expected, AraC treatment induced rapid phosphorylation and activation of ATM (Fig. 3H-I) (17-20). Importantly, we find that ATM-mediated DDR leads to phosphorylation and activation of p38 MAPK and its downstream effector, MAPKAPK2 (MK2) (22, 26) on AraC treatment (Fig. 3H-I). MK2 phosphorylates TTP to up-regulate inflammatory gene expression in immune cells (27-29). To examine if MK2 kinase could phosphorylate TTP in resistant G0 leukemic cells, two different inhibitors of p38 MAPK/MK2 were tested. Inhibition of p38 MAPK, using the clinically tested p38 MAPKα/β inhibitor, LY2228820 (LY) (22, 63), or a pan p38 MAPK inhibitor that targets all isoforms, BIRB796 (BIRB) (64), blocks phosphorylation of MK2, which prevents MK2-mediated TTP phosphorylation and reduces TNFα in AraCS cells (Fig. 3J). These results suggest that p38 MAPK/MK2 phosphorylates TTP, resulting in expression of ARE mRNAs such as TNFα in G0 leukemic cells (Fig. 3H). To test if p38 MAPK/MK/TTP regulates TNFα expression via its ARE, a Firefly luciferase reporter bearing the 3′ UTR ARE of TNFα was tested, with Renilla luciferase as co-transfected control. Luciferase activity increased by 2-fold in AraCS cells compared to S+ cells but not when p38 MAPK was inhibited with LY (Fig. 3K). These data suggest that p38 MAPK/MK2/TTP axis up-regulates specific gene expression via AREs in G0 leukemic cells. These results are consistent with our findings of increased ARE bearing, pro-inflammatory factor mRNA levels and translation—due to decreased ARE mRNA decay by the ATM-p38 MAPK-MK2-TTP axis (Fig. 3H) that is induced by stress and DNA damage signaling due to serum-starvation and chemotherapy (26).
Transiently activated p38 MAPK/MK2 in early G0—at early times of serum-starvation and chemotherapy treatment—promotes chemoresistance
We noted that p38 MAPK and MK2 were transiently phosphorylated, at early times upon induction of G0 cells with SS or AraC treatment (Fig. 3I and 4A-B). To investigate whether early activation of p38 MAPK/MK2 impacted AraC resistance, p38 MAPK was pharmacologically inhibited before (or along with) as well as after treatment with AraC—and then chemo-survival was measured using multiple assays, including cell death and two cell viability assays (Fig. 4C-E). Inhibition of p38 MAPK/MK2 with BIRB (Fig. 4D) or LY (Fig. 4E), one day after treatment with AraC, does not show any significant reduction in survival of AraC-resistant cells. Critically, inhibition of p38 MAPK/MK2 at early time points—4h-1 day prior to AraC treatment or along with AraC treatment—increased apoptosis and reduced survival of 70% of AraC-resistant cells (Fig. 4D-E). As a control, p38 MAPK inhibition alone does not affect viability of S+ cells that are not treated with AraC (Fig. 4D-E). These results suggest that p38 MAPK/MK2 is rapidly activated upon AraC treatment to turn on downstream survival genes at early times. Therefore, to overcome AraC resistance effectively, p38 MAPK/MK2 should be targeted at early time points before it induces downstream survival pathways that would dampen the effect of anti-inflammatory therapeutics. Consistently, sequential treatment with p38 MAPK inhibitor, LY, followed by AraC showed a dramatic reduction (50-70%) in chemo-survival in multiple AML cell lines but not in non-cancerous cells (Fig. 4F, CD34+), as these pathways (MAPK, inflammatory genes) are not up-regulated in non-cancerous G0 (Fig. 2I). When treated with p38 MAPK inhibitor alone, viability of S+ cells in multiple AML cell lines remained unchanged, indicating the synergism of AraC and p38 MAPK inhibitors on resistant leukemic cells (Fig. 4F). Interestingly, p38 MAPK inhibition eliminated resistant cells more significantly at increasing concentrations of AraC (Fig. 4G). Treatment with high concentrations of AraC would increase the number of cells induced into the resistant G0 state with induction of p38 MAPK; therefore, inhibition of this pathway is more effective at high concentrations of AraC. Conversely, even low concentration of p38 MAPK inhibitor, BIRB, was sufficient to reduce chemoresistance (Fig. S4B). Activation of p38 MAPK/MK2 can induce resistance by arresting the cell cycle (22, 25, 26); however, p38 MAPK/MK2 inhibition did not affect the cell cycle in resistant G0 leukemic cells (Fig. S4A), indicating a distinct mechanism of resistance by p38 MAPK/MK2 in G0. These data uncover early and rapid activation of a p38 MAPK/MK2 pathway that enables chemo-survival of G0 leukemic cells—in part, through inhibition of TTP activity that enables expression of pro-survival, pro-inflammatory genes like TNFα (Fig. 5A). Pharmacological inhibition of this early-activated p38 MAPK/MK2 pathway prior to and along with chemotherapy reduces resistance; in contrast, post-chemotherapy inhibition of p38 MAPK/MK2 is not effective, as at later time points their downstream effectors of cell survival genes are already on.
Activation of p38 MAPK/MK2 promotes chemoresistance via inactivation of TTP and subsequent up-regulation of TNFα
We investigated the mechanism by which p38 MAPK/MK2 promotes chemo-survival in G0 leukemic cells. As shown in Fig. 3G-K, p38 MAPK/MK2 phosphorylates and inactivates TTP, leading to increased production of TNFα. TNFα activates the NFкB pathway that increases anti-apoptotic gene expression as a stress response to promote cell survival (65-67). Importantly, our translatome analysis shows a significant increase in TNFα, TNFα receptors (Fig. S5A) and NFкB signaling (Fig. 5B), including anti-apoptotic BCL family members (Fig. S5B) (67-69), indicating that TNFα-mediated inflammatory response may cause chemoresistance. To examine the role of the TTP-TNFα axis in AraC resistance, we expressed non-phosphorylatable TTP mutant (TTP-AA) that is active in ARE mRNA decay and inhibits TNFα (Fig. 3G). Importantly, we find that TTP-AA expression reduces survival of AraC-resistant cells (Fig. 5C). These data suggest that TTP phosphorylation and inactivation—by p38 MAPK/MK2—enables chemo-survival through expression of ARE-bearing pro-inflammatory genes like TNFα.
Inhibition of TNFα/NFкB—prior to or at the same time as chemotherapy—decreases resistance, correlating with transient, early increase in expression of TNFα/NFкB in early G0
TTP may have other targets apart from TNFα. To verify that TTP regulates chemoresistance via targeting TNFα, we created a stable cell line expressing an inducible shRNA against TNFα. As shown in Fig. 5D, induction of TNFαdepletion prior to AraC treatment effectively reduced AraC resistance, compared to after AraC treatment, while no effect was observed with TNFα depletion alone without AraC. In contrast, addition of recombinant TNFα enhanced survival of AraC-treated cells (Fig. 5D). TNFα-mediated chemoresistance is not due to arrested cell cycle as TNFα treatment without subsequent AraC does not alter the cell cycle (Fig. S5C). These data suggest that phosphorylation of TTP and subsequent expression of TNFα, which are induced by p38 MAPK/MK2, are responsible for survival of G0 leukemic cells.
TNFα can also be inhibited pharmacologically with the inhibitor Pirfenidone (PFD, Fig. 5A) that can block TNFα translation, reduce inflammatory factors, and is clinically used for inflammatory disease like fibrosis (22, 70, 71). PFD decreases TNFα at the translatome and proteome levels in G0 leukemic cells (Fig. 5E). Our observation of early activation of p38 MAPK/MK2 (Fig. 4A-B) suggested that TNFα could be rapidly up-regulated upon G0 induction. Time-course translatome analysis affirmed that TNFα is most significantly increased (16-fold) at the earliest time point of 4 h after serum-starvation or AraC treatment (Fig. 5B), causing early expression of NFкB target genes including cell survival genes (Fig. 5B and S5B). Consistently, PFD treatment at least 18 hours prior to or along with AraC or serum-starvation significantly reduced viability of G0 leukemic cells but failed to reduce resistance when added after AraC treatment (Fig. 5F, S5D). PFD treatment alone does not affect the viability of untreated S+ cells, indicating that the cytotoxic effect of PFD is specific to G0 leukemic cells (Fig. 5F). Similar results were observed in MCF7 cells, where PFD reduced doxorubicin resistance (Fig. S5H). These data indicate that activation of the inflammatory pathway is an early event in G0 cells, which leads to resistance, and needs to be inhibited early to preclude downstream survival regulators.
PFD treatment reduced chemotherapy survival in multiple AML cell lines but not all leukemia types (Fig. 5G, AML cell lines versus chronic myelogenous leukemia K562). Since PFD could affect targets other than TNFα to reduce resistance (22, 70, 71), we tested the effect of PFD along with shTNFα depletion together, prior to AraC chemotherapy, in comparison to individual treatments: an additive effect would not be observed if PFD affects chemosurvival through TNFα and not via other targets. Co-treatment with PFD and TNFα shRNA prior to AraC did not show an additive effect of reducing survival compared to PFD treatment alone, indicating that the effect of PFD was mediated through TNFα inhibition (Fig. S5E). Similarly, we find that inhibition of TNFα-induced NFКB, with NF К B inhibitor, BAY11-7082 (72) to block downstream anti-apoptotic response, prior to or along with AraC or serum-starvation, decreases survival; treatment after chemotherapy or serum-starvation had no effect as observed with upstream TNFα inhibitor PFD (Fig. 5H). These data suggest that the TNFα/NF К B inflammatory pathway is upregulated as an early survival pathway in G0 cells, which increases chemo-survival. In accord with the transient increase of inflammatory genes at early times in these G0 cells after chemotherapy or serum-starvation, we find that treatment with inflammation inhibitors (PFD or BAY11-7082) or TNFα shRNA prior to and continued with chemotherapy significantly reduces chemoresistance—while treatment post-chemotherapy, fails to reduce resistance.
Co-inhibition of p38 MAPK and TNFα sensitizes resistant leukemic cells to AraC treatment by reducing TNFα and activating a pro-apoptotic JNK pathway
Although chemoresistant cells are sensitive to individual inhibition of either TNFα or p38 MAPK by PFD or LY respectively, a substantial proportion of cells still survived (Fig. 4F, 5G). Therefore, we asked if co-inhibition of p38 MAPK and TNFα with LY and PFD respectively, could eliminate the remaining resistant cells. We find that individual treatment, with low concentrations of LY or PFD (half of the concentrations used in Fig. 4-5), prior to or along with AraC, reduces approximately 50% of surviving leukemic cells (Fig. 6A-B). Importantly, the combination of PFD and LY2228820 prior to AraC treatment–called PLA therapy–eliminates about 90% of chemoresistant cells in multiple AML cell lines (Fig. 6A-C). Furthermore, PLA therapy decreased colony formation capacity of leukemic cells on methylcellulose by 10-fold, compared to AraC-treatment alone (Fig. 6D). These data indicate a severe loss of stem cell capacity of leukemic cells treated with PLA therapy. In contrast, in the absence of AraC treatment, the combination of PFD and LY2228820 did not affect cell viability, apoptosis and colony formation capacity, indicating the synergistic effect between AraC and anti-inflammatory drugs (Fig. 6A-D). Despite the fact that stromal niche cells have been shown to protect leukemic cells from chemotherapy (73), we find that AML cells co-cultured with stromal cells remained sensitive to PLA therapy (Fig S5F).
We investigated the molecular mechanism by which PLA therapy enhanced chemosensitivity. First, TNFα depletion reduces chemoresistance (Fig. 5D). LY destabilizes TNFα mRNAs by TTP dephosphorylation (28) (Fig. 3J), while PFD was shown to suppress TNFα selectively at the translation level (71); consistently, PFD treatment in chemoresistant G0 leukemic cells reduced TNFα at the translatome level but not at the RNA level (Fig. S5G). Therefore, in PLA therapy, phosphorylation of p38 MAPK target MK2 is decreased by LY, diminishing TTP phosphorylation and reducing TNFα mRNA levels (Fig. 3G-J) (28)—and additionally, due to inhibition of TNFα translation by PFD, TNFα remains more effectively blocked, compared to individual drug treatments (Fig. 6E). Second, we investigated c-Jun N terminal kinase (JNK) that can promote apoptosis depending on stimulus and cell type (74). Phosphorylation of JNK was enhanced in cells treated with PLA therapy, compared to single-drug treatments (Fig. 6E). JNK inhibitor, JNK-IN-8 partially reversed apoptosis of PLA therapy-treated cells, but did not affect the viability of cells not treated with PLA (Fig. 6F). Together, these results suggest that PLA therapy reduces TNFα and promotes a pro-apoptotic JNK pathway, leading to apoptosis of chemoresistant cells.
PLA therapy reduces chemoresistance in primary AML cells ex vivo and in vivo
To test the anti-leukemic activity of PLA therapy in primary AML (75), primary cells from AML patients as well as two murine AML models driven by Hoxa9/Meis1 or MLL/AF9 (SI Methods), were used. When either p38 MAPK or TNFα was inhibited prior to AraC treatment, moderate apoptosis of chemoresistant cells was observed in primary AML cells (Fig. 7A-B). Importantly, co-inhibition of p38 MAPK and TNFα by PLA therapy (pre-treatment before AraC) significantly reduced AraC resistance in fourteen out of fifteen AML patient samples as well as in primary cells from two AML mouse models ex vivo (Fig. 7A-B). In contrast, the viability of normal CD34+ cells from healthy donors was not affected by treatment with LY or PFD (Fig. 4F, 7A-B), consistent with clinical studies that have shown that PFD and LY have acceptable safety and tolerance (63, 70). To further investigate the therapeutic potential of PLA therapy in vivo, human AML cells expressing luciferase (MOLM13-Luc, SI Methods) were intravenously or subcutaneously injected into NSG mice. After confirmation of engraftment by measuring tumor volume or bioluminescent imaging (BLI), the mice were treated with PLA therapy or AraC for two weeks. Consistent with ex vivo results (Fig. 6B), PLA therapy significantly decreased the leukemic burden and tumor volume by 6-fold, compared to AraC treatment alone (Fig. 7C-D). Next, we tested PLA therapy in C57BL/6 mice engrafted with primary Hoxa9/Meis1 leukemia cells expressing luciferase. These mice were treated with PLA therapy or AraC. Consistently, BLI shows that PLA therapy eliminated 78% or 96% of chemoresistant cells in a dosage-dependent manner (Fig. 7E-F) and extended mice survival by 7 days (Fig. 7F). In the absence of AraC treatment, the combination of PFD and LY2228820 did not affect leukemic burden, suggesting that cytotoxic effects of this combination are limited to AraC-resistant cells, rather than proliferating cells (Fig. 7G). Together, these results suggest PLA therapy has potential for improving AraC-mediated apoptosis in AML.
Discussion
G0 cells are a transiently arrested, clinically relevant subpopulation in cancers (1, 2, 5-10). Our previous data and others, revealed altered gene expression mechanisms in G0 leukemic cells, at the post-transcriptional (8, 12) and translational levels (13, 14, 32). This would lead to a distinct gene expression profile to enable G0 cell survival in harsh conditions. G0 cells are resistant to stress conditions like serum-starvation, with transient inhibition of apoptosis, and proliferation (1, 11, 32), which are features required for cells to survive chemotherapy. Importantly, we find that serum-starved leukemic SS cells exhibit chemoresistance (Fig. 1E); consistently, true chemo-surviving AraCS cells are transiently arrested and chemoresistant (Fig. 1D-E, S1B-C). In accord, we find that SS cells are similar in translatome and proteome to AraCS cells (Fig. 1F), indicating that consistent with their common features of G0 arrest and chemosurvival, they show similar post-transcription gene expression. Published transcriptional signatures of G0 fibroblasts, and of leukemic stem cells and other in vivo chemoresistance leukemic models (1, 2, 8, 11, 15, 16), are also highly expressed in SS and AraCS cells (Fig. 1G, S1G). Thus, the common G0 resistance gene expression profile observed in AraCS and SS G0 cells—distinct from neighboring proliferating tumor cells—likely comprises genes that control survival and resistance. These data revealed that in addition to known transcriptional profiles, altered post-transcriptional and translation mechanisms in G0 resistant cells contribute to their unique gene expression profile that underlies their chemoresistance.
Our findings reveal the importance of DNA damage and stress signaling that can initiate a pro-inflammatory response that causes survival instead of cytotoxicity (Fig. 4-6, S4-S6). Differential genomic instability in cancers would lead to distinct subpopulations within a tumor with disparate DDR and stress signaling (17-19) that we find, enables their chemotherapy survival via pro-inflammatory cytokines. Cytokines upregulated in SS and AraCS cells do not significantly match SASP (50, 51) (Fig. S2G). This is consistent with differences between G0 and senescence (1)—with low mTOR activity (Fig. 2A-C, I, S2E) (14) and mutated p53 in these G0 cells—unlike high mTOR activity and p53 in senescence (32). These data indicate that a quiescence- and resistance-specific set of pro-inflammatory genes are expressed in these resistant cells. These include inflammatory cytokines (Fig. 2F-G) that promote downstream NFКB activated pro-survival target genes (65-67) including BCL family members of anti-apoptotic genes (67-69) (Fig. 5A-B, S5A-B). Treatment with reagents against these resistance-enabling immune regulators after chemotherapy is not very effective as the downstream survival effectors have already been induced; thus, targeting their upstream cytokine regulators would not be effective at this later time (Fig. 4A-B, 4D-E, 5F-H, S5D, 6B). Thus, treatment with reagents that block these resistance pathways prior to (and continued with) or along with chemotherapy, enables the most effective reduction of resistance, as they prevent further enrichment of such resistant cells by blocking induction of pro-survival signaling.
AraC is a nucleotide analog and inhibits replication, which triggers DDR (31). Increasing the concentration of AraC would cause increased DDR and downstream p38 MAPK signaling (17-19) and should lead to more cells expressing this inflammatory pathway that enables resistance—and thus pushes more cells into the inflammatory phase that can be targeted by inhibitors. Consistently, increased AraC treatment leads to more cells in the inflammatory phase that can be targeted by LY, leading to more significant loss of resistance (Fig. 4G). Non-cancerous cells do not show this pathway (Fig. 2I) and are not affected by inhibitors (Fig. 4F, 7A). These data suggest that certain chemotherapies and stresses like serum-starvation induce stress signaling (Fig. 4A-C) and enrich for resistant G0 cells—in addition to pre-existing subpopulations with genomic instability that trigger DDR and stress (17-19). Importantly, this resistance mechanism can be blocked to significantly decrease resistance to AraC, not only in different AML cell lines (Fig. 4C-G, 6B) but also in vivo in AML xenograft and mouse models (Fig. 7C-G) as well as in multiple patient-derived AML samples—without affecting normal cells (Fig. 7A)— supporting their potential applicability as a therapeutic target against chemoresistance in AML.
We find key signaling pathways (Fig. S6), mediated by AraCS and SS treatments (Fig. 4A-B, 5A-B), which alter post-transcriptional and translational gene expression to enable resistance. These include: 1. DNA damage ATM (17-19) and stress activated p38 MAPK signaling that in turn promotes (Fig. 4A-B) MK2 (22, 26); MK2 post-transcriptionally upregulates ARE bearing mRNAs (27-29) (Fig. 3G-K, 4A-B), including proinflammatory cytokines like TNFα (65, 66) that activate downstream anti-apoptosis signals (Fig. 5A-B, S5A-B) (67-69) to promote resistance (Fig. 5-7). 2. ATM mediated suppression of mTOR activity (17, 18) to inhibit canonical translation via 4EBP dephosphorylation that enables non-canonical translation (14) (Fig. 2A-C); this results in specific translation of pro-inflammatory cytokines (Fig. 3A-D) and immune modulators (44) (HLA-G, CD47, Fig. 2F-G, S2I) (45-47) that regulate antitumor immune response and are associated with resistance (48, 49). 3. UPR stress signaling that can be induced by p38 MAPK (62) and DNA damage (23, 24); this inhibits canonical translation via PERK phosphorylation of eIF2α and enables non-canonical specific mRNA translation (Fig. 2A-D, S2H)— which also increases inflammation and blocks apoptosis to enable chemoresistance (23, 24). Blocking the p38 MAPKα/β pathway with LY (22, 63), in combination with the anti-inflammatory PFD (22, 70, 71) that blocks downstream TNFα expression (70, 71) (Fig. 5A, E, S5G)—prior to (and continued with) AraC chemotherapy—lead to effective loss of chemoresistance in multiple AML cell lines (Fig. 6B), in tumors in vivo in AML mouse models (Fig. 7C-G), and in patient samples (Fig. 7A), validating their ability to reduce resistance and tumors in vitro and in vivo. LY destabilizes TNFα mRNA by TTP dephosphorylation (Fig. 3J) (28), while PFD suppresses TNFα selectively at the translation level (71) (Fig. S5G) and thus enables PLA combination therapy to more effectively sustain reduction in resistance than the individual drugs. Apart from its effect on TNFα translation, PFD blocks inflammation regulator (21, 76) p38 MAPKγ that can be increased upon p38MAPKα/β inhibition, preventing feedback reactivation of inflammation, and enabling PLA combination therapy to remain more efficacious than the individual drugs. Consistently, in the combination treatment with LY and PFD, LY decreases phosphorylation of MK2 by p38 MAPK, reducing TTP phosphorylation and TNFα RNA levels (Fig. 3G-J) (28); additionally, TNFα translation is inhibited by PFD, resulting in chemo-survival being more effectively blocked, compared to individual drug treatments (Fig. 6B, E). In addition, we find that upon inhibition of p38 MAPK and TNFα, the JNK pathway (74) is activated to promote apoptosis (Fig. 6E-F).Therefore, the combination of PFD and LY suppresses the inflammatory and stress response more effectively as it blocks upstream and downstream steps in the pathway to circumvent feedback reactivation or alternate pathways, which leads to the enhanced decrease in chemoresistant cell survival and cancer persistence in vitro and in vivo (Fig. 6, 7). These data indicate that blocking pro-inflammatory effectors—that are induced by chemotherapy mediated DNA damage and stress signaling—leads to increased chemosensitivity and decreased resistant cell survival.
EIF2α phosphorylation and mTOR inhibition reduces both rate limiting steps of canonical translation initiation (Fig. 2A-C). ATM and p38 MAPK can lead to activation of PKR and UPR stress-activated PERK, which phosphorylate eIF2α to decrease canonical tRNA recruitment and translation (Fig. 2B-C), leading to specific mRNA translation (33, 38, 42). These include cell cycle arrest factor, p27 KIP1(77) (Fig. S1C, S2A-B, p27), and GADD34 (Fig. S2H)—that are translated when canonical translation mechanisms are reduced (33, 42) as in G0 (13, 14). ATM activates AMPK which inhibits mTOR (17-19), leading to dephosphorylated, active 4EBP (Fig. 2B-C). 4EBP blocks canonical translation via eIF4E inhibition (35, 40), and permits non-canonical translation mechanisms of specific genes, including immune modulators like TNFα and CD47 (Fig. 2F-G, S2I, 3A-C, 5A-C,-D, S5A) (14), which activate survival genes (Fig. S5B).
Our findings revealed that these inflammatory genes upregulated in G0, have AREs and other UTR sequences that regulate mRNA levels and translation (Fig. 3A-C, S3A). The ATM-p38 MAPK-MK2 axis stabilize these ARE bearing pro-inflammatory cytokine mRNAs by phosphorylating ARE binding mRNA decay factor, TTP to prevent its mRNA decay activity on pro-inflammatory cytokine mRNAs like TNFα (Fig. 3D-K, 4A-B, 4F, 5A-C). This is consistent with previous studies on the role of AREs in cancers (14, 25, 28, 52, 54-56). All AML cell lines tested respond to the combination treatment that inhibits the p38 MAPK/MK2/TTP/TNFα axis leading to reduced TTP phosphorylation and inhibited TNFα (Fig. 6B, 6E-F) (28-30). These data suggest that the phospho-TTP and thus TTP activity, may be a key regulator of pro-inflammatory gene mediated chemoresistance. In support, overexpression of TTP-AA—that cannot be phosphorylated and is a dominant active form that restores ARE mRNA decay (28-30)—causes decrease in TNFα expression, and leads to reduced chemoresistance in both cell lines (Fig. 3G, 5C). These data suggest that phospho-TTP level is an important regulator of inflammatory response mediated chemoresistance, which can be harnessed as a marker and target against clinical resistance. Consistently, published in vivo leukemia resistance models show similar increased TTP and ARE bearing genes including inflammatory genes expressed (15, 78), as in our studies (Fig. 3A-F); additionally, inhibition of these pathways curtails chemoresistance in vivo in primary AML patients and tumor models (Fig. 7). Together, these pathways that are up-regulated in resistant cells (Fig. 4A, 5A) via chemotherapy and stress induced signaling—decrease canonical translation and permit non-canonical translation and post-transcriptional regulation of specific genes (Fig. S6)—to induce a pro-inflammatory response that promotes chemo-survival of G0 cancer cells.
Materials & Methods
Methods
including details on plasmids, cell viability assays, flow cytometry, protein analysis, drugs, and motif analysis are described in Supplemental Information.
Cell Culture & Cell lines
THP1 cells were cultured in Dulbecco’s modified Eagle medium (RPMI)1460 media supplemented with 10% fetal bovine serum (FBS), 2 mM L-Glutamine, 100 µg/mL streptomycin and 100 U/ml penicillin at 37°C in 5% CO2. SS THP1 cells were prepared by washing with PBS followed by serum-starvation at a density of 2 × 105 cells/mL and AraCS cells, by treatment with 5 µM AraC for 3 days or 9 days. MCF7, HFF, HEPG2 and U2OS cells were cultured in Dulbecco’s modified Eagle medium (DMEM) media with 10% FBS, 2 mM L-Glutamine, 100 µg/mL streptomycin and 100 U/ml penicillin, as done previously (13, 14). MCF7 cells were serum-starved or treated with 150 uM Doxorubicin. THP1 (TIB-202), MV4:11 (CRL-9591), K562 (CCL243), HFF (SCRC-1041), MCF7 (HTB-22), U2OS (HTB-96) and HEPG2 (HB-8065) were obtained from ATCC. MOLM13 (ACC554), NOMO1 (ACC542) and MONOMAC6 (ACC124) were obtained from DSMZ. Cell lines kindly provided by David Scadden (79) and MOLM13-GFP-Luc by Monica Guzman (80) were tested for Mycoplasma (Promega) and authenticated by the ATCC Cell Authentication Testing Service (79).
Primary AML patient samples and human monocytes
All human samples (de-identified) were handled in accordance with IRB protocols to SV (2015P000998/MGH), approved by the Partners Human Research Committee Institutional Review Board /MGH IRB, to DAS, and to TG (DF/HCC 13-583), approved by DF/HCC Office for Human Research Studies. AML samples used in this study were obtained by DAS including: MGH15 - bone marrow 60% blasts, karyotype 46, XX, t(9;11)(p22;q23)[20/20]; MGH22 - peripheral blood, 60% blasts, karyotype 46,XX,t(3;21)(q26;q22),t(9;22)(q34;q11.2)[18]/46,XX[2]; and MGH25 - bone marrow, 90% blasts, karyotype 46,XX[20] and by JL-S and TG including bone marrow samples: EQ1899, CI2095, PO2038, LA2053, NC1866, GO1122, CM2164, MV2192, VD2160, XD2101, VL2317, and OA2500. Bone marrow or peripheral blood mononuclear cells were isolated from de novo AML patients by ficoll density gradient centrifugation and cryopreserved with DMSO in a liquid nitrogen tank. Thawed cells were maintained in RPMI media with 10% FBS for several days before drugs treatment and analyses. Human CD34+ monocytes (2M-101) were obtained from Lonza. Primary cells from MLL-AF9, HoxA9-Meis1 mouse models were provided by DS (3). Mouse primary cells were maintained in RPMI media with 10% FBS, 2 mM L-Glutamine, 100 μg/mL streptomycin, 100 U/ml penicillin, 5 ng/ml murine IL-3 and 25 ng/ml murine Stem Cell Factor (SCF).
In vivo AML mouse models
AML mouse models have been shown to predict therapy response accurately (81). C57BL/6 and NSG were obtained from MGH Cox-7 Gnotobiotic animal facility of the AAALAC-accredited Center for Comparative Medicine and Services at MGH. C57BL/6 or NSG mice were injected intravenously or subcutaneously with HoxA9-Meis1 or MOLM13 cells expressing luciferase (80, 82). IVIS imaging system (Perkin Elmer) were used to confirm engraftment of AML cells. Mice were intraperitoneally injected with 200 µl of luciferase substrate D-Luciferin (15 mg/ml) and anesthetized. Images were taken 5 or 10 minutes after D-Luciferin injection. After confirmation of engraftment by IVIS imaging, mice were randomly assigned to two groups and treated with pirfenidone (100 mg/kg, intraperitoneally), LY2228820 (20 mg/kg, intraperitoneally), AraC (30 mg/kg, intraperitoneally) or saline according at indicated combinations and dosages. Tumor volumes were measured by IVIS imaging at indicated time points.
Polysome profiling with microarray
Sucrose was dissolved in lysis buffer containing 100 mM KCl, 5 mM MgCl2, 100 µg/ml cycloheximide, 2 mM DTT and 10 mM Tris-HCl (pH 7.4). Sucrose gradients from 15% to 50% were prepared in ultracentrifuge tubes (Beckman) as previously described (13, 83-85). Cells were treated with 100 µg/mL cycloheximide at 37°C for 5 minutes before collecting them. Harvested cell were rinsed with ice-cold PBS having 100 µg/mL cycloheximide and then were resuspended in lysis buffer with 1% Triton X-100 and 40 U/mL murine (New England Biolabs) for 20 minutes. After centrifugation of cell lysates at 12,000 x g for 20 minutes, supernatants were loaded onto sucrose gradients followed by ultracentrifugation (Beckman Coulter Optima L90) at 34,000 × rpm at 4°C for 2 hours in the SW40 rotor. Samples were separated by density gradient fractionation system (Teledyne Isco). RNAs were purified by using TRIzol (Invitrogen) from heavy polysome fractions and whole cell lysates. The synthesized cDNA probes from WT Expression Kit (Ambion) were hybridized to Gene Chip Human Transcriptome Array 2.0 (Affymetrix) and analyzed by the Partners Healthcare Center for Personalized Genetic Medicine Microarray and BUMC facilities. Gene ontology analysis for differentially expressed translatome or proteome was conducted by DAVID 6.7 tools (86) (87). Molecular signatures enriched in AraCS or SS were identified by Gene Set Enrichment Analysis (GSEA) (88).
Data Availability
Raw datasets will be submitted to GEO public repository at final submission and will be available publicly as well as on request to the authors. TBD GEO accession number.
Statistical analyses & data availability
are described in Supplemental Information.
Statement on Human Data
All human samples (de-identified) were handled in accordance with IRB protocols to SV (2015P000998/MGH), approved by the Partners Human Research Committee Institutional Review Board /MGH IRB, to DAS, and to TG (DF/HCC 13-583), approved by DF/HCC Office for Human Research Studies. Details in Supplemental Information.
Statement on animal data
nod-scid-gamma (NSG), C57Black/6 mice, 10-12 weeks, are obtained from MGH Cox-7 Gnotobiotic animal facility of the AAALAC-accredited Center for Comparative Medicine and Services at MGH. These facilities are supervised by veterinarians in the Center for Comparative Medicine and the MGH Subcommittee for Animal Research (SRAC) and maintained according to the protocol approved by SRAC, and provide services for breeding, regular health checks, histopathology and macropathology.
Author Contributions
SL conducted the research and bioinformatic analysis; SST, SIAB & DL contributed data; SL, YK, DM, BTN, ID-G, DTM, M-KC, DS, SM, & DAH provided in vivo models & in vivo data; MAM, RR & RG did immune data; AL, NJH, & ML provided mRNA folding energies & patient gene signatures; MB & WH conducted proteomics; DAS & JL-S provided patient samples; SV supervised the project & wrote the manuscript.
List of Supplemental files
1. Supplemental Information
Methods (related to main text, main figures 1-7 & supplemental figures S1-S6) Supplemental References (related to supplemental information, methods section) Supplemental Figure legends S1-S6, related to main figures 1-7
2. Supplemental Figures S1-S6 (related to main figures 1-7)
3. Supplemental Tables 1-2 (related to main figures 1-7 & supplemental Figures S1-S6)
List of Supplemental Information
Methods (related to main text, main figures 1-7 & supplemental figures S1-S6) References (related to supplemental information, methods section)
Supplemental Figure legends S1-S6, & Supplemental Table legends, related to main figures 1-7 Supplemental Figures S1-S6 (related to main figures 1-7)
Methods (related to main text, main figures 1-7 & supplemental figures S1-S6) Plasmids
TRIPZ plasmids expressing shRNA against human TNFα (V2THS_111606), and miR30a primiR sequences used as control (RHS4750), were obtained from Open Biosystems and MGH cancer center, respectively. Stable cell lines were constructed as described by Open Biosystems. The stable cells expressing shRNA against TNFα were induced with 1 µg/mL doxycycline at indicated time points to knockdown TNFα. Luciferase reporters to test ARE expression were previously described (1). Cells were treated with 10 ng/ml recombinant TNFα (R&D Systems) to activate NF КB pathway. Myc-tagged TTP-AA (2, 3) was a gift from Nancy Kedersha and Shawn Lyons from Paul Anderson’s lab.
MTS assay
MTS assay, a colorimetric quantification of viable cells was conducted as described by the manufacturer, Promega. A volume of 100 µl cells was placed in a 96-well plate after drugs treatment. A volume of 20 µl MTS reagent (CellTiter 96® Aqueous Non-Radioactive Cell Proliferation Assay) was added to each well followed by incubation at 37°C for 1 hour. Absorbance was measured at 490 nm by using a microplate reader.
Caspase 3/7 assay
After drugs treatment, cell death was measured by using caspase-glo® 3/7 assay kit (Promega) according to the protocol provided by the manufacturer. The equal volume of caspase-glo reagent was added to cells, and samples were gently mixed with pipetting. The plates were incubated at room temperature in the dark for 2 hours. The luminescence of each sample was measured in a luminometer (Turner BioSystems).
Flow cytometry and cell cycle analysis
Cell proliferation was determined by flow cytometry of cells labeled with propidium iodide and bromodeoxyuridine (BrdU). The cells were incubated with 10 µM BrdU for 90 minutes at 37°C in 5% CO2 before harvesting. Collected cells were fixed in ice cold 70% ethanol overnight. Cells were washed in PBS and treated with 2 M HCl for 30 min. Cells were incubated for 1 hour with anti-BrdU antibody conjugated to FITC (eBioscience) in the dark, washed and stained with propidium iodide. Samples were filtered through a nylon mesh filter and cell cycle analysis performed on the flow cytometry (4).
Western blot analysis
Cells were collected and resuspended in lysis buffer containing 40 mM Tris-HCl (pH 7.4), 6 mM MgCl2, 150 mM NaCl, 0.1% NP-40, 1 mM DTT and protease inhibitors (Roche). Samples containing 80 µg of protein were loaded onto 10% or 12% SDS-PAGE (Bio-Rad), transferred to PVDF membranes and processed for immunoblotting. Antibodies against p27 (06-445) and tubulin (05-829) were obtained from Millipore. Antibodies against HES1 (sc-25392), eIF2α (sc-11386) and phospho-4EBP1 (sc-1809) were from Santacruz. Antibodies against phospho-ATM (ab81292), phospho-PKR (ab32036) and phospho-IRE1 (ab124945) were from Abcam. Antibodies against phospho-PERK (649401) were from Biolegend. Antibodies again TNFα (3707), phospho-p38 MAPK (4511), phospho-MK2 (3007), phospho-eIF2α (9721), TTP (71632) and 4EBP1 (9452) were from Cell Signaling Technology.
Apoptosis analysis
Leukemic cells were treated with indicated drug combinations. Annexin V FITC/PI staining was performed with FITC Annexin V Apoptosis Detection Kit I (BD Pharmingen). Flow cytometry analysis and FlowJo software were used to quantitate the percentages of apoptotic cells.
Colony forming assay
After treatment with indicated drug combinations, the same number of cells were plated in methylcellulose-based media with human recombinant cytokines (stem cell technology, MethoCultTM H4435). Number of colonies was quantitated in each plate after 10 days.
Mass Spectrometry
Multiplex quantitative proteomics analysis was conducted, as previously(5), from S+, SS and AraC treated THP1 leukemic cells.
Inhibitors
Pirfenidone (10 to 300 µg/ml (6-9)) was obtained from Chemietek. AraC (1 to 10 µM (10, 11)), LY2228820 (0.03 to 2 µM (12-15)), BIRB796 (BIRB, 5µM (16-20)), and JNK-IN-8(21) were from Selleckchem. KU55933 (10 µM (22)), BAY 11-7082 (10 µM (23)) and D-luciferin were from Cayman Chemical and Doxorubicin (10 to 500 nM (24)) was from Tocris Bioscience.
Motif, AREs, RNA binding proteins & Ribosome occupancy analysis
The Multiple Em for Motif Elicitation (MEME) software was used to search for cis-elements enriched in 5’ UTR of translationally regulated genes (25). Human 5’ UTR sequences were retrieved from UCSC table browser (26). In a discriminative mode, 5’ UTR sequences of translationally up- or down-regulated genes were used as the primary sequences and 5’ UTR sequences of translationally unchanged genes, the control sequences. Motifs were found in the given strand with 6-30 nt motif width. We compared polysome-associated mRNAs with their total RNA levels in serum-starved and AraCS cells to generate the change in ribosome occupancy (RO)(27-29)—which is the ratio of the level of mRNA that is associated with heavy polysomes compared to the total mRNA level of each gene (Fig. 2F, venn diagram, heat map). ARE Score algorithm (30) was used to assess scores of AU-rich elements quantitatively. The list of RNA binding protein genes were obtained from RBPDB database (31).
Statistical analyses
All experiments in every figure used at least 3 biological replicates except for microarray, mass spectrometry, and patient sample data. Each experiment was repeated at least 3 times. No statistical method was used to pre-determine sample size. Sample sizes were estimated on the basis of availability and previous experiments (32, 33). No samples were excluded from analyses. P values and statistical tests were conducted for each figure. Statistical analyses were conducted using R or Excel. Two-tailed unpaired t-test or Wilcoxon rank sum test was applied to assess statistical significance. SEM (standard error of mean) values are shown as error bars in all figures. Means were used as center values in box plots. P-values less than 0.05 were indicated with an asterisk. E-values were used for the statistical significance in the motif analysis.
Supplemental Figures (Related to main figures 1-7)
Supplemental Table S1. Related to main Figures 1-7. Genes up-regulated at the translatome level in both AraC and SS cells.
Supplemental Table S2. Related to main Figures 1-7. Genes bearing AU-rich elements (AREs) and up-regulated at the translatome level.
Acknowledgments
We thank Partners Healthcare Center for Personalized Genetic Medicine & BUMC facilities for microarray data; N. Kedersha, S. Lyons & P. Anderson for plasmids & antibody; T. Graubert for patient samples; D. Bloch, S. Wu, A. Naar, M. Guzman, N. Bardeesy, D. Scadden, S. Ramaswamy, & C. Benes for reagents.
Footnotes
Financial statement The study is funded by Cancer Research Institute, D. & M-E Ryder, Leukemia & Lymphoma Society, Smith Family Foundation, R01 GM100202 grants to SV & R01 CA185086. SL was funded by Fund for Medical Discovery fellowship & by a postdoctoral fellowship from the Basic Science Research Program through the National Research Foundation of Korea (2015017218). DAS is funded by R01 CA115772. DAH is funded by Howard Hughes Medical Institute and 2R01 CA129933.
Statement on conflict of Interest Authors declare no conflict of interest.
References
- (1).↵
- (2).↵
- (3).↵
- (4).↵
- (5).↵
- (6).↵
- (7).↵
- (8).↵
- (9).↵
- (10).↵
- (11).↵
- (12).↵
- (13).↵
- (14).↵
- (15).↵
- (16).↵
- (17).↵
- (18).↵
- (19).↵
- (20).↵
- (21).↵
- (22).↵
- (23).↵
- (24).↵
- (25).↵
- (26).↵
- (27).↵
- (28).↵
- (29).↵
- (30).↵
- (31).↵
- (32).↵
- (33).↵
- (34).↵
- (35).↵
- (36).
- (37).
- (38).↵
- (39).↵
- (40).↵
- (41).↵
- (42).↵
- (43).↵
- (44).↵
- (45).↵
- (46).
- (47).↵
- (48).↵
- (49).↵
- (50).↵
- (51).↵
- (52).↵
- (53).↵
- (54).↵
- (55).
- (56).↵
- (57).↵
- (58).↵
- (59).↵
- (60).↵
- (61).↵
- (62).↵
- (63).↵
- (64).↵
- (65).↵
- (66).↵
- (67).↵
- (68).
- (69).↵
- (70).↵
- (71).↵
- (72).↵
- (73).↵
- (74).↵
- (75).↵
- (76).↵
- (77).↵
- (78).↵
- (79).↵
- (80).↵
- (81).↵
- (82).↵
- (83).↵
- (84).
- (85).↵
- (86).↵
- (87).↵
- (88).↵
References
- (1).
- (2).
- (3).
- (4).
- (5).
- (6).
- (7).
- (8).
- (9).
- (10).
- (11).
- (12).
- (13).
- (14).
- (15).
- (16).
- (17).
- (18).
- (19).
- (20).
- (21).
- (22).
- (23).
- (24).
- (25).
- (26).
- (27).
- (28).
- (29).
- (30).
- (31).
- (32).
- (33).
- (34).
- (35).