Targeted inhibition of STAT/TET1 axis as a therapeutic strategy for acute myeloid leukemia

Effective therapy of acute myeloid leukemia (AML) remains an unmet need. DNA methylcytosine dioxygenase Ten-eleven translocation 1 (TET1) is a critical oncoprotein in AML. Through a series of data analysis and drug screening, we identified two compounds (i.e., NSC-311068 and NSC-370284) that selectively suppress TET1 transcription and 5-hydroxymethylcytosine (5hmC) modification, and effectively inhibit cell viability in AML with high expression of TET1 (i.e., TET1-high AML), including AML carrying t(11q23)/MLL-rearrangements and t(8;21) AML. NSC-311068 and especially NSC-370284 significantly repressed TET1-high AML progression in vivo. UC-514321, a structural analog of NSC-370284, exhibited a more potent therapeutic effect and prolonged the median survival of TET1-high AML mice over three fold. NSC-370284 and UC-514321 both directly target STAT3/5, transcriptional activators of TET1, and thus repress TET1 expression. They also exhibit strong synergistic effects with standard chemotherapy. Our results highlight the therapeutic potential of targeting the STAT/TET1 axis by selective inhibitors in AML treatment.

TET1 is the founding member of a family of DNA methylcytosine dioxygenases (including TET1/2/3), which are known to convert 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC), leading to DNA demethylation. While previous studies suggested that all three TET genes likely function as tumor-suppressor genes in various cancers, the Chen lab and others have provided compelling evidence showing that TET1 plays a critical oncogenic role in the development of myeloid malignancies. Importantly, Tet1 knockout has been reported to have minor effects on normal development including hematopoiesis in mice. Therefore, TET1 is a promising therapeutic target for AML treatment and targeting TET1 signaling is a hot pursuit in the field of cancer/myeloid malignancies. In this manuscript, Jiang et al. timely reported the identification of several novel therapeutic drugs against TET1 signaling. The authors demonstrated the high efficacy and selectivity of these drugs in targeting acute myeloid leukemias (AMLs) with high levels of TET1 expression (i.e., TET1-high AMLs) both in vitro and in vivo. To develop inhibitors targeting TET1 signaling, the authors employed an innovative strategy to screen small-molecule compounds that inhibit the transcription of TET1, rather than inhibiting TET1's enzymatic activity, since TET1 exerts its oncogenic role through both enzymatic-dependent and -independent mechanisms. Using this strategy, the authors identified compounds NSC-311068 and NSC-370284 as selective inhibitors of TET1 signaling by inhibiting TET1 transcription. They showed that both compounds, especially NSC-370284, exhibited potent therapeutic effects in treating TET1-high AMLs. A structural analogue of NSC-370284, namely UC-514321, exhibited even better therapy effects. In addition, the authors delineated the molecular mechanism by which these compounds inhibit the TET1 transcription, by showing that the compounds directly target the transcriptional activators of TET1, STAT3/5. The authors also showed that NSC-370284 or UC-514321 exhibit a synergistic therapeutic effect with standard chemotherapy agents such as daunorubicin. While, these compounds had no noticeable (both acute and long-term) side effects on normal hematopoietic system in vivo, highlighting the low toxicity of these compounds for future potential clinical application.
In summary, this is the first report for the identification of therapeutic agents targeting TET1 signaling, which exhibit high efficacy and selectivity for treating TET1-high AMLs. Moreover, it is novel to show that STAT3/5 are direct transcriptional regulators of TET1 and small-molecule inhibitors targeting STAT3/5 can efficiently suppress TET1expression. This work is of great significance and novelty in both basic and clinical cancer research. The combination of such TET1 inhibitors with standard chemotherapy agents holds potential to improve the clinical treatment of TET1-high AMLs. I only have miner comments as the following: Minor points: 1. While the therapeutic effects of the TET1 transcription inhibiting compounds are quite promising, the authors should provide more pre-clinical data such as PK/PD, which are important to evaluate the potential of clinical appliances.
2. Although the authors have shown that these TET1 transcription inhibiting compounds exhibited no obvious effects on normal hematopoiesis in vivo, it would be helpful to examine the effects of these compounds on normal HSC/HPCs in vitro, if possible compared with TET1-high HSC/HPCs. Colony-forming assay and cell viability (apoptosis) assay can be applied.
Reviewer #2 (Remarks to the Author): Alteration in enzymatic DNA methylcytosine oxidation has been reported in AML. But the therapeutic opportunity has not been explored. This manuscript describes the first effort in suppressing AML by inhibiting the function of TET oxidase which appears to act as an oncogene. The overall concept presented in this manuscript is novel.
More and more basic transcriptional machinery factors are now being explored for therapeutic applications with success and promises, such as BRD4 inhibitors (best example), integrator complex, TAF250 and so on. The inhibitors reported in this work actually targeted STATs but was identified by using TET1 as readout in a screen. If these compounds had reliable efficacy to AML cells in vivo and not much side effects, they would be promising in principle. However, the specificity to TET1 transcription inhibition has to be further studied. The question whether the compounds deliver their effects really through inhibiting TET1 transcription is not fully addressed. Targeting Stats would influence transcription of many genes. Additionally, AML is highly heterogenous in its genetics, therefore the mechanism that makes the compounds specifically effective for the AML with high TET1 expression would need to be better characterized.
A specific suggestion is to do inducible ectopic expression of TET1 in AML cells treated with UC-514321 and examine whether this would mitigate the effect of the compound, restoring cell viability.
The compounds appeared to work well in vivo with the MLL-AF9 and AML-ETO9a models. But how is their effect with other important models?
The title appears somewhat confusing as the compounds bind STAT3/5 as STAT inhibitors rather than inhibit TET1 transcription directly.
Reviewer #3 (Remarks to the Author): In this manuscript, the authors focus on TET1 as a target in AML. Based on a chemical library screen database, they have identified several compounds that they suggest decrease TET1 expression in a STAT3-or STAT5-dependent manner, and exert a therapeutic effect in in vitro and murine models of AML.
The overall concept of this study is very interesting and provocative, and these compounds certainly do show activity in the experiments shown. However, the data leave open many fundamental questions regarding the mechanism by which these compounds are exerting their effect, and much of the data can be explained by non-specific toxicity, especially as at least one of the compounds is structurally related to a known topoisomerase II inhibitor. In particular, the following points should be considered: 1. The rationale for the strategy to identify the initial lead compounds should be clarified. In the NCI60 panel, only six of the lines are hematopoietic. Since TET1 may have a tumor suppressor role in solid tumors (as noted in the Introduction), it is not clear why a TET1 inhibitor would be expected to show decreasing viability proportional to TET1 expression across all 60 lines. This is important, again, inasmuch as the question of non-specific toxicity needs to be considered carefully. It may be worthwhile to show the primary data from this analysis.
2. The authors make the case that NSC-311068 and NSC-370284 decrease the expression of TET1 mRNA. However, at the time point (48 hours) and concentration (300 nM) that they examine (Figure 1c), there is >90% loss of viability of sensitive cells, making the significance of this finding difficult to determine. This likely also applies to Figure 2c and d. Many mRNAs and proteins with short half lives will decrease after cells are exposed to toxic agents, but that does not mean that these decreases are specific or driving the change in viability.
3. The suggestion that STAT transcription factors may be involved in the mechanism of action of these compounds derived from a mutational analysis related to clones resistant to these compounds. The authors than extrapolated from pathway analysis that the "Jak/STAT5" pathway may be involved. However, it is not apparent that any of the mutations shown in Supplemental Table 3 would overcome the effect of a direct STAT inhibitor (the proposed mechanism). 4. Figure 3 (and subsequent figures) shows data on STAT DNA binding. It would be worthwhile to show the levels of STAT3 and STAT5 phosphorylation in the relevant cells, and how they change, if at all (particularly at early time points), with these compounds. 5. If the mechanism of action of these compounds is the inhibition of STAT3 and/or STAT5 transcriptional function, it is not at all clear why the other STAT inhibitors they tested failed to show a similar effect. Conversely, since many other genes necessary for proliferation and survival are STAT-dependent, it is not clear why these compounds would only be effective in TET1-high AML cells. The authors should also note if there is any evidence that these compounds inhibit signatures of STAT-dependent genes in these AML cells, as would be expected if they exert a STAT-dependent mechanism.
6. The authors show that transduction of MLL_AF9 into bone marrow progenitor cells confers sensitivity to the compounds of interest ( Figure 5). Since this transduced gene is presumably no longer under the control of a STAT-dependent promoter, this would seem to undermine the central hypothesis regarding the mechanism of action of the compound. 7. A plausible mechanism for why the resistant subclones are sensitive to daunorubicin should be proposed. It is not at all obvious why this should be so.
8. If the underlying mechanism for these compounds effects is that they downregulate TET1, then ectopic expression of TET1 should rescue cells from their effects. This experiment would seem worthwhile to perform.

SUMMARY:
We are very grateful to the reviewers for their positive appraisals of our manuscript and very much appreciate their thoughtful and constructive suggestions and comments. Following their suggestions and comments, we have performed additional experiments and data analysis, and have also carefully revised the manuscript accordingly. Summarized below are major experiments and data analysis conducted for the paper revision: a) A series of additional experiments have been carried out to further strengthen the mechanism studies of NSC-370284. We first created an inducible Tet1 construct, and found ectopic expression of Tet1 sufficiently reversed the inhibitory effects of NSC-370284 and UC-514321 and restored viability of the AML cells. To understand why NSC-370284 exhibited a selective effect on targeting TET1-high AMLs, we analyzed the affinity of STAT3/5 on the promoter regions of TET1 and other STAT target genes, e.g., HIF2α, by ChIP-qPCR, and found that the STAT proteins have a stronger enrichment on the promoter of TET1 than on that of HIF2α in TET1-high AML cells and NSC-370284 could dramatically inhibit the binding of STAT proteins to TET1 promoter. Further, we detected no significant influences of NSC-370284 on STAT activity (i.e., phosphorylation). Our results indicate that different from typical STAT inhibitors that target STAT kinase activity, NSC-370284 (and its derivative, UC-514321) may exert its function mainly through interfering with the association between STAT protein and the DNA regions which have relatively higher basal affinity to STATs, like the TET1 promoter, in AML cells. b) We were able to explore into the mechanisms underlying NSC-370284 resistance in more details as well. Choosing JAK1 A893G , one of the mutations associated with NSC-370284 resistance as a representative, we showed ectopic expression of this mutant largely reversed the inhibitory effect of NSC-370284 on MLL-AF9-AML cell viability. It is likely the mutations in genes associated with the JAK/STAT signaling that were identified in our drug-resistant clones may overcome NSC-370284-mediated inhibitory effect on activation of the JAK/STAT signaling and AML cell viability/growth, and thereby confer drug resistance to AML clones. c) Two more AML models, i.e., MLL-AF10 AML and FLT3-ITD/NPM1 mut AML, have been included to show the broad therapeutic effects of our lead compounds in treating AML. d) Additional cell viability assays have been conducted to assess the effects of NSC-370284 and UC-514321 in normal hematopoietic stem/progenitor cells (HSPCs; c-Kit + BM cells) in vitro. Our data showed that while the compounds significantly inhibited the viability and increased apoptosis of Tet1-high AML cells, they showed minor effects on normal HSPCs, suggesting a good therapeutic window for the application of our compounds in treating TET1-high AMLs. e) More preclinical data, including maximum tolerated dose (MTD), LD50 and in vivo preclinical pharmacokinetic (PK) analysis, have been added. f) More analysis of the RNA-seq data of NSC-370284-treated or NSC-370284-resistant AML cells have been included, to show the genes and enriched gene clusters associated with daunorubicin (DNR) sensitivity.
All the text in the manuscript that we changed or added during the revision is marked in red. Deleted text has been removed entirely from the file. Our detailed responses are described below. We hope that the revised version of the manuscript would fulfill the requirements of the reviewers and editor, and will be suitable for the publication in Nature Communications.
The following are our point-by-point responses to the comments/suggestions from the reviewers:

Responses to the comments of reviewer 1:
General Remarkably, NSC-370284 or UC-514321 treatment dramatically suppressed the viability of AML cells, but not that of normal HSPCs (Supplementary Fig. 8a). In addition, the compounds significantly increased apoptosis in AML cells, but not in normal HSPCs (Supplementary Fig.  8b). Thus, these two compounds showed no obvious toxicity on normal HSPCs. This is consistent with endogenous Tet1 expression pattern, as AML cells with MLL-AF10 or FLT3-ITD/NPM1 mut have relatively higher Tet1 expression levels, as compared with normal HSPCs (Supplementary Fig. 8c). The above results have been included as the new Supplementary Fig.  8.

Responses to the comments of reviewer 2:
General Comments from Reviewer 2: Alteration in enzymatic DNA methylcytosine oxidation has been reported in AML. But the therapeutic opportunity has not been explored. This manuscript describes the first effort in suppressing AML by inhibiting the function of TET oxidase which appears to act as an oncogene. The overall concept presented in this manuscript is novel. More and more basic transcriptional machinery factors are now being explored for therapeutic applications with success and promises, such as BRD4 inhibitors (best example), integrator complex, TAF250 and so on. The inhibitors reported in this work actually targeted STATs but was identified by using TET1 as readout in a screen. If these compounds had reliable efficacy to AML cells in vivo and not much side effects, they would be promising in principle. However, the specificity to TET1 transcription inhibition has to be further studied. The question whether the compounds deliver their effects really through inhibiting TET1 transcription is not fully addressed. Targeting Stats would influence transcription of many genes. Additionally, AML is highly heterogenous in its genetics, therefore the mechanism that makes the compounds specifically effective for the AML with high TET1 expression would need to be better characterized.

Response:
We appreciate the reviewer's overall positive appraisal of our manuscript and we are very grateful to him/her for the constructive comments and suggestions. We have taken the reviewer's advice and performed additional experiments to further address potent therapeutic effects of the candidate compounds and to better characterize the molecular mechanism underlying their selectivity on targeting TET1-high AML cells. These new results have been added into this revised manuscript.

Comment 3 from Reviewer 2: The title appears somewhat confusing as the compounds bind STAT3/5 as STAT inhibitors rather than inhibit TET1 transcription directly.
Response: We appreciate reviewer's comment, and have changed the title to "Targeted inhibition of STAT/TET1 axis as a potent therapeutic strategy for acute myeloid leukemia".

Responses to the comments of reviewer 3:
General Comments from Reviewer 3: In this manuscript, the authors focus on TET1 as a target in AML. Based on a chemical library screen database, they have identified several compounds that they suggest decrease TET1 expression in a STAT3-or STAT5-dependent manner, and exert a therapeutic effect in in vitro and murine models of AML.
The overall concept of this study is very interesting and provocative, and these compounds certainly do show activity in the experiments shown. However, the data leave open many fundamental questions regarding the mechanism by which these compounds are exerting their effect, and much of the data can be explained by non-specific toxicity, especially as at least one of the compounds is structurally related to a known topoisomerase II inhibitor.

Response:
We appreciate the reviewer's positive appraisal of our manuscript. Following the reviewer's constructive suggestions and comments, we have conducted additional experiments to further strengthen the functional evaluation and mechanistic studies of the chemical compounds. New results have been included in Figs. 1, 4, 5, Supplementary Figs.1, 4, 5, 9, 10 . In fact, high expression of TET1 could be found not only in AML, but also in many other tumors, e.g., uterine cancer, glioma, etc., and especially, in testicular germ cell malignancies. This indicates potential oncogenic role of TET1 in various cancers. We have included the TCGA expression pattern of TET1 as new Supplementary Fig. 1. The original drug response data has been enclosed as the new Supplementary Table 2.

Comment 2 from Reviewer 3:
The authors make the case that NSC-311068 and NSC-370284 decrease the expression of TET1 mRNA. However, at the time point (48 hours) and concentration (300 nM) that they examine (Figure 1c), there is >90% loss of viability of sensitive cells, making the significance of this finding difficult to determine. This likely also applies to Figure 2c and d. Many mRNAs and proteins with short half lives will decrease after cells are exposed to toxic agents, but that does not mean that these decreases are specific or driving the change in viability.
Response: This is a good point. In order to rule out the possibility of non-specific toxicity, we reduced the dose of NSC-311068 and NSC-370284 to 25 nM, and analyzed gene expression and cell viability 24 hours after treatment. The low dose, short-term treatments again resulted in a significant down-regulation of TET1 transcription, accompanied with a very minor decrease in the viability of MONOMAC-6, THP-1 and KOCL-48 cells (see new Fig. 1e-f). Thus, it is unlikely that the inhibitory effects of NSC-311068 and NSC-370284 on TET1 expression were due to nonspecific toxicity. Shown in Figs. 2c-d are results of in vivo sample analysis. Briefly, we conducted secondary BMT, treated the leukemic mice with the chemical compounds or control, collected BM samples at their end points and performed statistical analysis of survival data and conducted qPCR and Western blot assays to evaluate the effects of drug treatment on Tet1 expression. The dose of NSC-311068 and -370284 treatment (i.e., 2.5 mg/kg, i.p., once per day, for 10 days) has been proven to be within the safe range, based on the acute and long-term toxicity profiling, the maximum tolerated dose, LD50 and in vivo preclinical pharmacokinetic (PK) analysis shown in Supplementary Tables 7-9, and Supplementary Figs. 9-11. Thus, it is unlikely the inhibitory effects of NSC-311068 and NSC-370284 on TET1 expression were due to nonspecific toxicity.

Comment 3 from Reviewer 3:
The suggestion that STAT transcription factors may be involved in the mechanism of action of these compounds derived from a mutational analysis related to clones resistant to these compounds. The authors than extrapolated from pathway analysis that the "Jak/STAT5" pathway may be involved. However, it is not apparent that any of the mutations shown in Supplemental Table 3 would overcome the effect of a direct STAT inhibitor (the proposed mechanism).

Response:
We thank the reviewer for the comment. Indeed, the mechanism underlying drug resistance is often complicated and remains a challenge in the field of cancer research. A number 6 of the genes identified from our drug-resistant AML clones with recurrent mutations, e.g., MSH3, Notch, etc., have been reported to be associated with the JAK/STAT signaling (Tseng-Rogenski et al., Gastroenterology, 2015;Cheng et al., Cellular Signalling, 2010;Gagarin et al., J Mol Cell Cardiol, 2005;Huang et al., Cell Cycle, 2008). It is likely that mutations in such genes may overcome NSC-370284-mediated inhibitory effect on activation of the JAK/STAT signaling and AML cell viability/growth, and thereby confer drug resistance to the AML clones. To test this, we chose JAK1 as a representative and cloned a construct carrying the JAK1 A893G mutant that was detected in our drug-resistant THP-1 cells (see Supplementary Table 4). As expected, forced expression of JAK1 A893G mutant largely reversed the inhibitory effect of NSC-370284 on AML cell viability (Supplementary Fig. 4c). It is out of the scope of this paper for us to test pathological effects of all the mutations, but investigating these drug resistance-associated mutations would definitely be an interesting research topic in the future.
Comment 4 from Reviewer 3: Figure 3 ( Response: We appreciate the reviewer's comment. We have now also analyzed the inhibitory effect of NSC-370284 on the expression of HIF2α, a well known target of STAT5, and found that the inhibitory effect was not as obvious as that on TET1 (Supplementary Fig. 4i). The basal association of STAT5 with the HIF2α promoter was low, as compared to that with the TET1 promoter region; and the interruption by NSC-370284 on such association was much less obvious (Supplementary Fig. 4j). The very weak, if any, basal affinity of STAT5 with many of its target genes' promoters (e.g., BCL-x, HIF2) without cytokine stimulation (e.g., IL-3 or EPO) has been reported before (Nelson et al., J Biol Chem, 2004;Fatrai et al., Blood, 2011). Therefore, our results revealed a particularly strong association between STAT3/5 and the TET1 promoter ( Fig. 3e-g, 3k; Supplementary Fig. 4i-j). Our results indicate that different from typical STAT inhibitors that target STAT kinase activity, NSC-370284 (and its derivative, UC-514321) may exert its function mainly through interfering with the association between STAT protein and the DNA regions which have relatively higher basal affinity to STATs, like the TET1 promoter, in AML cells. The new data and above discussion have been added into the revised manuscript.  Supplementary Fig. 8c, is highly likely still STATdependent and thus can be significantly inhibited by our compounds, which in turn leads to the significant inhibitor effect of the compounds on the viability of MLL-AF9-transduced bone marrow progenitor cells as shown in new Figs. 5i-m. Therefore, such data actually support our central hypothesis regarding the mechanism of action of our compound.
Comment 7 from Reviewer 3: A plausible mechanism for why the resistant subclones are sensitive to daunorubicin should be proposed. It is not at all obvious why this should be so.

Response:
We appreciate the reviewer's suggestion. Through analysis of the RNA-seq data of the NSC-370284 resistant clones and parental cells, we found that several gene clusters that are known to be associated with drug response, especially response to topoisomerase II inhibitors such as DNR, are enriched in NSC-370284 resistant cells. These gene clusters include JAK/STAT signaling, G2M checkpoint, MYC targets and E2F targets, etc. (Supplementary  Table 10a). It was reported that JAK/STAT pathway inhibitors, e.g. AG490, could sensitize tumor cells to topoisomerase II inhibitors 1 . A potential mechanism might be through targeting E2F. It was shown that E2F1 overexpression significantly sensitized cancer cells to various topoisomerase II inhibitors 2 . Our RNA-seq data showed increased E2F1 levels in NSC-370284 resistant AML clones, relative to parental cells. This is consistent with the E2F1-inducing effect of other STAT inhibitors, e.g. AG490 3 . Also consistent is the enrichment of E2F targets in control samples relative to NSC-370284 treated samples (Supplementary Table 10b). Therefore, very likely the upregulation of E2F1 in NSC-370284 resistant clones at least partially explains why the resistant clones, compared to the parental AML cells, are more sensitive to DNR treatment. The new data and above discussion have been included in the revised manuscript now.
Comment 8 from Reviewer 3: If the underlying mechanism for these compounds effects is that they downregulate TET1, then ectopic expression of TET1 should rescue cells from their effects. This experiment would seem worthwhile to perform.
Response: As described in our response to Comment 1 from Reviewer 2, we have created a pLenti-puro vectored Tet1 construct, and transduced it into THP-1 and MONOMAC-6 cells and showed that ectopic expression of Tet1 sufficiently reversed the inhibitory effects of NSC-370284 and UC-514321 and restored viability of the AML cells (Fig. 5n-o). This new data together with other relevant data from our paper provide compelling evidence supporting our proposed mechanism underlying the selective effects of the compounds on TET1-high AML cells.