BCAT1 is abnormally activated in CBF-AML and associated with poor clinical outcome
Initially, we analyzed the expression of BCAT1 across 15 subtypes defined by most frequent genetic variations in 577 AML patients using the datasets derived from The Cancer Genome Atlas (TCGA)(44, 45) and BEAT-AML.(46) We classified the samples based on the quartile expression of BCAT1 (Q1 ~ Q4) and genetic background (Supplemental Table S1). In doing so, cases with top 50% (Q3 + Q4) BCAT1 expression were preferentially observed in subtypes carrying t(8;21) and inv(16) fusions compared to other subtypes (Fig. 1A). In agreement with this, higher expression of BCAT1 was seen in CBF-AML compared to AML with normal karyotypes (NK) and human peripheral blood mononuclear cells (PBMC) (Fig. 1B, Supplemental Table S2). In contrast, such a difference in its isoenzyme, mitochondrial BCAT2, was not observed in AML patients and AML cell lines (Fig. 1B, Supplemental Fig. 1, Supplemental Table S3). These data suggested that high expression of BCAT1 in these two AML subtypes might be ascribed to the t(8;21) and inv(16) fusions.
To further confirm the observed aberrant high expression of BCAT1 in TCGA and BEAT-AML datasets, BCAT1 expression at mRNA and protein levels were validated using quantitative RT-PCR, Western blot and DepMap database(47) in AML cell lines with or without CBFb-MYH11 and RUNX1-ETO fusions. Significantly higher expression of BCAT1 was found in RUNX1-ETO positive cells (Kasumi-1 and SKNO-1) and CBFb-MYH11 positive cells (ME-1), whereas BCAT1 was merely detectable in non-CBF-AML cells (Fig. 1C). Moreover, we found that high expression of BCAT1 was associated with poor prognosis in CBF-AML (Fig. 1D).(48) Taken together, aberrant and specific high expression of BCAT1 driven by CBFb-MYH11 and RUNX1-ETO fusions confers poor outcomes in CBF-AML patients.
Metabolic and transcriptomic profiling reveals abnormal metabolic effects associated with high expression of BCAT1 in CBF-AML
To comprehensively reveal the role of BCAT1 activation in CBF-AML, we performed metabolic and transcriptomic profiling upon down-regulation of BCAT1 in Kasumi-1 (carrying RUNX1-ETO) and ME-1 (carrying CBFb-MYH11) cells (Fig. 2A). By analyzing metabolites directly catalyzed by BCAT1, we found that down-regulation of BCAT1 led to the accumulation of intracellular BCAAs (isoleucine, leucine, valine) and aKG and decreased glutamine in Kasumi-1 and ME-1 cells (Fig. 2B), suggesting that BCAT1 was a primary pro-catabolic factor in BCAA metabolism in CBF-AML cells. Down-regulated BCAAs catabolism led to impaired substrate supply for protein synthesis, TCA cycle, and de novo nucleotide synthesis (Fig. 2B), influencing pathways related to essential energy, carbon, and nitrogen sources for the progression and material accumulation of leukemia.(28, 49–51) To further analyze the biological effects of inhibiting BCAT1 expression, we performed GO enrichment analysis based on the differential metabolites (Supplemental Table S4) and differentially expressed genes (DEGs) (Supplemental Table S5) upon BCAT1 knockdown in ME-1 cells. We found that the sphingolipid pathway, contributing to cell proliferation inhibition, apoptosis induction, and cell cycle arrest in leukemia cells,(52–58) was significantly down-regulated (Fig. 2C). As confirmed by RNA-seq results, DEGs upon BCAT1 knockdown were also enriched in pathways involved in essential cellular processes and overall global metabolism (Fig. 2D). Collectively, these results revealed an indispensable role of BCAT1 in maintaining the homeostasis of amino acid metabolism and regular cellular processes in CBF-AML cells.
RUNX1-ETO directly activates BCAT1 via a RUNX1-dependent manner
To further dissect the mechanism by which BCAT1 was upregulated in RUNX1-ETO AML, we attempted to explore the hypothesis that RUNX1-ETO transcriptionally activated BCAT1. We transformed U937 cell line by introducing Dox-inducible expression of RUNX1-ETO and found that Dox treatment led to increased expression of RUNX1-ETO, resulting in concomitant upregulation of BCAT1 (Fig. 3A). In good concert with this, pharmacological inhibition of RUNX1-ETO by two calpain inhibitors MG-101 and MDL-28170(59, 60) significantly downregulated BCAT1 expression (Fig. 3B, Supplemental Fig. 2A). Moreover, direct binding of RUNX1-ETO at the BCAT1 gene locus was confirmed by RUNX1-ETO ChIP-seq, and three potential enhancers (E1, E2, E3) were identified (Fig. 3C), in consistent with public H3K27ac and H3K4me3 ChIP-seq data from Kasumi-1 cells(61). Notably, MG-101 treatment abolished the binding of RUNX1-ETO at the BCAT1 gene locus, especially the E1 enhancer showing the most significant reduction of enrichment (Fig. 3C). To further establish the transcriptional regulation of BCAT1 by RUNX1-ETO, we cloned the sequence from BCAT1 promoter and E1 enhancer into the pGL3-basic and pGL3-promoter reporter vectors respectively (details in Supplemental Methods). As expected, luciferase activity was significantly enhanced by both the BCAT1 promoter and E1 enhancer upon the induction of RUNX1-ETO expression (Fig. 3D).
Our previous findings underlined the critical role of the interaction between RUNX1-ETO and RUNX1 in regulating downstream pathways. Indeed, direct binding of RUNX1 at the same region with RUNX1-ETO was observed at the BCAT1 gene locus (Fig. 3C). To further confirm the role of RUNX1 in the transactivation of BCAT1, we knocked down RUNX1 using small interfering RNA (siRNA) in both Kasumi-1 and RUNX1-ETO-induced U937 cells (Fig. 3E, Supplemental Fig. 2B). In doing so, binding of RUNX1-ETO was significantly attenuated at both BCAT1 promoter and E1 enhancer in RUNX1-ETO positive U937 cells, and a concomitant downregulation of BCAT1 expression upon knocking down RUNX1 was observed in Kasumi-1 cells (Fig. 3E-F). Moreover, knocking down RUNX1 led to a dramatic reduction of transcriptional activity of both BCAT1 promoter and E1 enhancer, indicated by decreased luciferase activity (Fig. 3G). Additionally, we also observed typical co-activators involved in the formation of the activation complex (e.g., EP300, PRMT1)(16, 62, 63) at the same region with RUNX1-ETO (Supplemental Fig. 2C), and the inhibition of EP300 led to a significant decrease in BCAT1 expression (Supplemental Fig. 2D). Overall, these results indicated that RUNX1-ETO transcriptionally upregulated BCAT1 in a RUNX1-dependent manner.
Partial nucleus retention of CBFb-MYH11 activates BCAT1 in a RUNX1-dependent manner
Regarding how the CBFb-MYH11 fusion protein exerts its function, two molecular mechanisms were proposed that CBFb-MYH11 can sequester wild-type RUNX1 into the cytoplasm or keep wild-type RUNX1 in the nucleus.(15, 64, 65) To identify which scenario accounts for the activation of BCAT1, we first looked into RUNX1 ChIP-seq data and identified direct binding at the BCAT1 promoter and E1 enhancer. Besides, direct binding of CBFb-MYH11 (overlapped bindings of CBFb and MYH11 were considered as binding of CBFb-MYH11)(13) was also observed at the BCAT1 promoter and E1 enhancer (Fig. 4A). Consistently, the transcriptional activity of the BCAT1 promoter and E1 enhancer was enhanced by CBFb-MYH11, indicated by increased luciferase activity (Fig. 4B). Based on these results, we proposed that at least partial retention of both CBFb-MYH11 and wild-type RUNX1 in the nucleus served as a potential transcriptional activator on cis-regulatory elements of the BCAT1 locus. In line with this reasoning, overexpression of CBFb-MYH11 led to upregulated expression of BCAT1 in U937 cells (Fig. 4C). Conversely, when the interaction between CBFb-MYH11 and RUNX1 was abolished by RUNX1/CBFb-MYH11 interaction inhibitor RO5-3335(66) in ME-1 cells, BCAT1 expression was dramatically compromised at both protein and mRNA levels (Fig. 4D). Also, inhibiting RUNX1 resulted in attenuated expression of BCAT1 (Fig. 4E). Moreover, in U937 cells where overexpression of CBFb-MYH11 was induced, knocking down endogenous RUNX1 led to a significant reduction of CBFb-MYH11 binding at BCAT1 gene locus (Fig. 4B) and CBFb-MYH11-mediated transcriptional activation of BCAT1 promoter and E1 enhancer (Fig. 4E). Furthermore, EP300, a co-activator of RUNX1, showed both co-localization and co-regulatory effects with RUNX1 (Supplemental Fig. 3-A-B). Collectively, these results suggested that wildtype RUNX1 transcriptionally activated BCAT1 by forming complex with CBFb-MYH11 and some other co-activators on the BCAT1 gene locus in CBFb-MYH11 positive AML.
CBFb-MYH11 and RUNX1-ETO reinforced promoter-enhancer interaction to activate BCAT1 transcription
The physical interaction between the promoter and the enhancers is critical for transcription of target genes.(67) To examine whether the BCAT1 enhancers physically interact with its promoter, we performed chromosome conformation capture (3C) and circular chromosome conformation capture (4C) in Kasumi-1 and ME-1 cells. Strong chromatin interaction between E1, E2, and E3 enhancers and the BCAT1 promoter was captured in both Kasumi-1 and ME-1 cells (Fig. 5A). Furthermore, such interaction was abolished when RUNX1-ETO or CBFb-MYH11 were depleted by specific inhibitors MG-101 or RUNX1/CBFb-MYH11 interaction inhibitor RO5-3335, respectively (Fig. 5A). As the E3 enhancer showed the most significantly reduced interaction between the promoter after inhibitor treatment, for the further validation, we selected the E3 enhancer as the viewpoint. In agreement with the previous results, the chromatin interaction between E3 enhancer and BCAT1 promoter captured in both Kasumi-1 and ME-1 cells by 4C and can also be attenuated by either RO5-3335 or MG-101 treatment (Fig. 5B). We then sought to test whether these chromatin interactions at the BCAT1 gene locus exist in CBF-AML patient blasts. For this purpose, we retrieved the published promoter-capture Hi-C data of AML patient blasts, including both CBF-AML and non-CBF-AML.(43) The interactions between the BCAT1 promoter and its enhancers were more robust in CBF-AML patients, as compared to those in non-CBF-AML patients (Fig. 5C, left panel). In line with this, BCAT1 expression was much higher in CBF-AML than in non-CBF-AML (Fig. 5C, right panel).
It is well established that CTCF is a critical factor forming the long-range chromatin interaction, such as topologically associated domains (TADs).(68, 69) In this regard, motif scanning confirmed significant enrichment of CTCF motif at BCAT1 promoter and enhancer regions (Supplemental Fig. 4A). Correspondingly, analysis based on published ChIP-seq data confirmed that BCAT1 promoter and E3 enhancer were flanked by binding of CTCF and a subunit of the cohesion complex, Rad21 (Supplemental Fig. 4B). Together, our findings indicated that both CBFβ-MYH11 and RUNX1-ETO are potentially involved in the CTCF-mediated promoter-enhancer interaction in CBF-AML.
BCAT1 knockdown leads to impaired CBF-AML cell fitness in vitro and in vivo
Having established the critical role of BCAT1 in BCAA metabolism and molecular mechanism of aberrant expression of BCAT1 in CBF-AML cells, we sought to understand the anti-leukemic effects of inhibiting BCAT1. Knocking down BCAT1 resulted in enhanced apoptosis (Fig. 6A, Supplemental Fig. 5A), arrested G1/S cell cycle (signified by decreased cyclins)(70) (Fig. 6B, Supplemental Fig. 5B), as well as upregulated cell differentiation (Fig. 6C, Supplemental Fig. 5C) in three CBF-AML cell lines. Besides, the blunted self-renewal ability was indicated by attenuated colony-forming ability in three CBF-AML cell lines in vitro (Fig. 6D, Supplemental Fig. 5D). To further investigate the effects of inhibiting BCAT1 in vivo, sub-lethally irradiated mice were transplanted with doxycycline (Dox)-inducible BCAT1 knockdown CBF-AML cell lines (luciferase-labeled) (Fig. 6E). Successful engraftment was confirmed by in vivo luminescence at day 7 (Fig. 6F), followed the next 2 weeks by Dox treatment to induce the knockdown of BCAT1 expression in vivo (Fig. 6G). In doing so, a significantly reduced leukemia burden was observed, as evidenced by decreased overall luminescence and percentage of leukemia blast in the mouse bone marrow (Fig. 6F-H). In line with this, the prolonged survival, alleviated splenomegaly and a relatively slight weight loss 30 days after transplantation were observed compared to mice in the control group (Fig. 6I-J). Collectively, these results suggested that inhibiting BCAT1 demonstrated anti-leukemic efficacy in vitro and attenuated leukemia progression in vivo, representing a potential therapeutic vulnerability for CBF-AML.
Gabapentin treatment effectively retards CBF-AML leukemogenesis in vivo
As a structural analog of leucine, Gabapentin (Gbp) specifically and competitively inhibits the transaminase activity of BCAT1, and its efficacy and safety have been proved in clinical practice.(17, 21, 71) However, there is a paucity of studies confirming its in vivo anti-leukemic effect, particularly in the context of CBF-AML characterized by high BCAT1 expression. In this regard, we first examined the effect of Gbp treatment on CBF-AML cell lines in vitro. By downregulation of BCAT1 expression, we first evaluated the in vitro effects of Gbp to CBF-AML by treating CBF fusion positive and negative AML cell lines with Gbp at its widely used concentration,(21) 20 mM, for 48 hours. Gbp treatment led to more significant growth inhibition in CBF-AML cells (Kasumi-1 and ME-1) compared to non-CBF-AML cell lines (U937, MV411, OCI-AML3 and THP-1) (Fig. 7A, Supplemental Fig. 6A).
We next tested the anti-leukemic effect of Gbp on CBF-AML in vivo by establishing xenograft mouse model using CBF-AML cell lines (Fig. 7B). All the mice were successfully engrafted with leukemia cells 10 days after i.v. injection of CBF-AML cell lines, followed by Gbp treatment (100 mg/kg) every five days by intraperitoneal injection.(72) Gbp treatment significantly reduced leukemia burden, indicated by reduced radiance, decreased expression of RUNX1-ETO or CBFb-MYH11 in the whole bone marrow, and milder splenomegaly (Fig. 7C-E). As expected, Gbp treatment led to prolonged median survival (Fig. 7F). In conclusion, these results proposed that BCAT1 is a potential therapeutic target and Gbp holds great promise to develop new therapies for CBF-AML.