It is estimated that there will be 437,033 new cases of leukemia and 309,006 deaths associated worldwide in 2018 [1
]. Predictions state that more than 21,450 people will be diagnosed with acute myeloid leukemia (AML) and nearly 10,920 people will die from it alone in the USA in 2019 [2
]. This warrants the search for new therapeutic strategies. The transcription factor β-catenin appears as an actionable drug target in AML, because the expression and activity of β-catenin is linked to disease initiation, unfavorable karyotypes, and poor prognosis. Cells from such patients show enhanced self-renewal capacity, which suggests that β-catenin contributes to a stemness phenotype [3
]. Moreover, β-catenin is expressed at significantly higher levels in AML compared to acute lymphoblastic leukemia and β-catenin was found to be overexpressed in 16/25 and 13/59 primary AML samples [5
]. Furthermore, a recent report illustrates that β-catenin is important for chemotherapy-associated senescence and contributes to aggressive growth patterns of leukemia and lymphoma cells [10
]. However, patient-to-patient variations regarding the addiction of AML cells to β-catenin [4
] demonstrate the need to further characterize how β-catenin affects leukemogenesis.
Secreted factors of the wingless-type MMTV integration site (WNT) family stabilize β-catenin. They reduce the phosphorylation of β-catenin and its subsequent poly-ubiquitylation for proteasomal degradation. Consequently, β-catenin accumulates in the nucleus, where it induces genes in complex with the transcription factor, T cell factor-4 [12
]. In AML cells, a cell-intrinsic activation of β-catenin can render leukemic stem cells independent of niche-derived WNT signals [14
]. Factors that are responsible for the proteasomal degradation of β-catenin include the glycogen synthase kinase-3 (GSK3), the adapter adenomatous polyposis coli, E3 ubiquitin ligases such as seven-in-absentia-homologues-1/-2, F-box/WD repeat-containing protein 1A, and other proteins [15
]. In addition to this proteasomal pathway, β-catenin can be cleaved by caspases [16
Current research strives to identify and exploit epigenetic mechanisms for the therapy of solid tumors and leukemia. A dysregulation of histone deacetylases (HDACs) is seen in leukemic cells and four Histone deacetylase inhibitors (HDACi; LBH589, FK228, vorinostat (SAHA), and belinostat (PXD101)) have been approved by the FDA for the treatment of hematological disorders. Moreover, there are currently at least 15 HDACi in clinical trials [18
]. The development of specific HDACi aims to define and exploit individual functions of HDACs in vivo and to reduce adverse effects of HDACi treatment [18
]. Clinically used HDACi encompass small molecules that inhibit all zinc-dependent HDACs or HDAC subclasses ([18
] and http://www.clinicaltrials.org
). For example, the hydroxamic acid LBH589 blocks zinc-dependent HDACs that belong to class I (HDAC1, -2, -3, and -8), class II (HDAC4, -5, -6, -7, -9, and -10), and class IV (HDAC11). The benzamide MS-275 and the depsipeptide FK228 act only against the class I HDACs HDAC1, -2, and -3 [22
A recent report shows that the anthraquinone oxime-analog BC2059 attenuates the expression of β-catenin in AML cells with and without an internal tandem duplication of the leukemogenic kinase FMS-like tyrosine kinase (FLT3-ITD) through a proteasomal mechanism. This mechanism involves a displacement of the scaffold protein transducin β-like-1 from β-catenin [23
]. This work also indicates that LBH589 and BC2059 combine favorably against AML cells and that this effect is associated with a reduction of β-catenin [23
]. However, it has not been addressed how HDACi decrease β-catenin, whether the attenuation of β-catenin by HDACi is a key factor for the fate of AML cells, and if one of the 11 zinc-dependent mammalian HDACs controls β-catenin expression. One of these HDACs might be the class IIB deacetylase HDAC6, because the tubulin polymerization promoting protein-1 blocks HDAC6 and attenuates β-catenin expression [24
]. Furthermore, HDAC6 interacts with the cancer stem cell marker membrane glycoprotein CD133 and both were reported to stabilize β-catenin acetylation-dependently [25
]. However, a key impact of HDAC6 on an important developmental regulator like β-catenin [12
] is hard to reconcile with the lack of phenotypic abnormalities in HDAC6 knockout mice and with accumulating evidence that an inhibition of HDAC6 does not affect cancer cell growth [26
]. HDACi that preferentially target HDAC3 evoke a destabilization of β-catenin, but not its mRNA expression, in breast cancer cells [30
]. Therefore, HDAC3 could be a key positive regulator of β-catenin stability in leukemic cells.
There is evidence that non-steroidal anti-inflammatory drugs (NSAIDs), which inhibit the immunomodulatory enzymes cyclooxygenase-1/cyclooxygenase-2 (COX1/COX2), antagonize β-catenin expression and the stemness of leukemic cells [31
]. Moreover, data collected with cells from various solid tumors suggest that a combined application of indomethacin with pan-HDACi as single or bifunctional molecules suppresses cell proliferation and angiogenesis. Remarkably, normal cells are spared by such treatment [35
]. Whether NSAIDs combine favorably with HDACi against leukemic cells is an interesting option that remains to be investigated.
The transcription factor myelocytomatosis oncogene (MYC), which has been implicated in maintaining the stemness of AML cells [13
], can be a target gene of β-catenin in leukemic cells [23
]. Pan- and class I HDACi provoke a downregulation of MYC in cancer cells [19
]. In addition to β-catenin and MYC, the transcription factor Wilms tumor (WT1) is associated with cell differentiation and leukemogenesis [45
], and, like β-catenin and MYC, WT1 is regulated by HDACi [46
]. Furthermore, WT1 can regulate the expression of MYC and of β-catenin positively and negatively [47
]. It is currently not clear if there is a mutual regulation of these factors in leukemic cells and whether HDACi alter a potential interplay between these three transcription factors.
In order to fully develop the therapeutic potential of HDACi, some key questions need to be resolved. For example, 18 human HDACs fall into the classes I, II, III, and IV and it has to be further clarified if a pharmacological inhibition of subsets or even single HDACs can produce anti-leukemic effects. Moreover, further molecular mechanisms and markers have to be identified to use HDACi effectively. We set out to answer these questions. Our data demonstrate that HDAC3 activity is a key factor for the survival and the maintenance of tumor- and stemness-associated transcription factors in AML cells.
We showed that LBH589 evoked growth arrest and apoptosis in leukemic cells. As this effect was also seen with MS-275, FK228 and RGFP966, we concluded that class I HDACs, and particularly HDAC3, were necessary for the survival of leukemic cells. The compromised survival of HDACi-exposed leukemic cells is associated with an accumulation of the replication stress/DNA damage marker ɣH2AX and with the formation of 53BP1 foci. 1 µM RGFP966 evokes replication stress/DNA damage without massive apoptotic DNA damage and z-VAD-FMK cannot abrogate ɣH2AX accumulation in HDACi-treated leukemic cells. From these data we deduced that HDACi-induced replication stress/DNA damage precedes apoptotic DNA fragmentation, which can also trigger the formation of ɣH2AX [57
]. Others also report that a pharmacological interference with HDAC3 triggers replication stress, DNA damage, cell death, and enhanced chemosensitivity of lymphocytic and myeloid leukemia cells [59
]. Such data are consistent with the reported impaired DNA repair capacities of HDACi-treated leukemic cells [21
As HDAC3 has functions that are dependent and independent of its catalytic activity [71
] it is important that RGFP966 phenocopies a genetic depletion of HDAC3 in transformed hematopoietic cells [65
]. Since up to 15 µM RGFP966 specifically inhibits HDAC3 [66
], our experimental settings with 1–10 µM RGFP966 likely revealed specific functions of HDAC3 activity. Hence, targeting HDAC3 seems to be a promising strategy to decrease β-catenin, MYC, and WT1, and to eliminate leukemic cells through replication stress/DNA damage and apoptosis. Our flow cytometry data demonstrated that RGFP966 decreased the numbers of cells in the G1 and G2/M phases dependent on caspases. The S phase populations were not rescued though when z-VAD-FMK was applied with RGFP966. This also applied to the accumulation of ɣH2AX, which was largely unaffected by z-VAD-FMK. Therefore, we concluded that S phase stalling and replication stress/DNA damage induction by RGFP966 was due to a genuine induction of replication stress/DNA damage that halted cell cycle progression. Further studies are underway to determine the targets of HDAC3 that control cellular responses to replication stress and DNA damage and whether cells slip from G2/M phase to G1 phase or die out of G2/M phase through mitotic catastrophe.
We further show that the extent of β-catenin processing by caspases tied in with a failure of leukemic cells to recover from HDACi-induced stress. Accordingly, a destabilization of β-catenin with indomethacin combined superiorly with HDACi. This finding could be due to the destabilization of β-catenin by both HDACi and NSAIDs, but NSAIDs exert multiple effects [35
]. However, while several NSAIDs repress transcription factors of the NF-κB family, which can be a driver of tumorigenesis in hematopoietic and solid tumor cells, indomethacin does not inactivate NF-κB, AP1, and cancer-relevant kinases. Nonetheless, indomethacin activates PPARɣ [74
] and an activation of PPARɣ by glitazones can attack the leukemic stem cell pool in CML [75
]. Additional investigations are justified to define the beneficial effects of NSAIDs on leukemic cells and their stem cells. This also applies to solid tumor cells form breast, colon, lung, and prostate, for which bifunctional conjugates of indomethacin with SAHA are potent inhibitors of growth and angiogenesis [35
]. For example, compound 11b, an indomethacin-SAHA fusion molecule that selectively inhibits COX2 and HDAC6 > HDAC8 > HDAC3 > HDAC2 > HDAC1, inhibits the proliferation of androgen-dependent prostate carcinoma cells [35
]. Further studies may find a relevance of HDAC8 inhibition by this compound, because this class I HDAC is also a valid target in AML cells [76
]. The finding that compound 11b is effective against cancer cells and selective for COX2 is promising because COX1 is a constitutively expressed housekeeping enzyme while COX2 is induced upon inflammation [35
]. Independent thereof is the consensus that the anti-tumor effects of NSAIDs are not directly linked to their inhibitory effects on COX1/COX2, but rather on other effects. These include among others the suppression of β-catenin and aberrant WNT signaling [32
Caspases -3, -6, -7, and -8 catalyze the degradation of β-catenin [16
] in solid tumor cells and in vitro. Our data collected with z-VAD-FMK suggest that caspases cleave β-catenin in HDACi-treated leukemic cells. Consistent with this idea, we see β-catenin cleavage products with a similar size as the ones seen by others under apoptotic conditions and caspase activation [16
]. Further experiments are necessary to decipher which caspases or other enzymes process β-catenin in leukemic cells and why only a subset of HDACi-treated leukemic cells cleaves β-catenin.
We demonstrate that MYC and WT1 have an antagonistic relationship, with MYC activating WT1 and WT1 suppressing MYC. We also show a regulation of β-catenin through WT1 in leukemic cells. Data collected with kidney-, solid tumor-, and CML-derived cells coherently show that WT1 can repress the expression of MYC [47
]. However, the opposite effect of WT1 on MYC has also been reported in such cells [54
]. Culture conditions as well as a context-dependent regulation may explain such discrepancies. Moreover, different isoforms of WT1 can determine whether it has a role as an oncogene or a tumor suppressor [49
]. Our data illustrating an accumulation of β-catenin upon a reduction of WT1 in leukemic cells corresponds to an observed negative impact of WT1 on β-catenin signaling in breast cancer cells [51
], Sertoli cells forming the blood-testis barrier [50
], and podocytes within the kidney [52
]. WT1 also impairs β-catenin/WNT signaling via other pathways in vitro and in vivo [53
], but one cannot exclude that there is a cell type-specific interplay between β-catenin and WT1 [54
]. It likewise remains to be clarified whether the mutual control of MYC and WT1 has further implications, for example for their antagonistic effects on the cell cycle regulator p21 [78
]. The concomitant loss of all three transcription factors in HDACi-treated cells demonstrates that an inhibition of HDACs, and particularly of HDAC3, overwrites such regulatory effects and leads to their loss by caspase-dependent and -independent mechanisms. Our data additionally suggest that the reduction of MYC by HDACi may lead to a reduction of WT1 in leukemic cells, but not vice versa.
The elimination of β-catenin does not affect MYC and we see no mutual regulation of β-catenin and MYC in leukemic cells. Accordingly, we found that MYC was decreased in HDACi-treated cells irrespective of whether they retained β-catenin. While this lack of coregulation between β-catenin and MYC is unexpected [11
], others also found that an attenuation of β-catenin by BC2059 did not affect MYC levels in leukemic cells [23
]. Moreover, this work shows that a similar reduction of β-catenin by LBH589 or BC2059 leads to either a clear or no reduction of MYC. Hence, LBH589 triggers processes that decrease MYC largely independent of β-catenin. It is also possible that the wild-type or mutant status of β-catenin has an impact on its relevance in cells and onto MYC expression. Our data suggest that β-catenin is wild-type in the leukemic cell lines that we analyzed. This finding agrees with the notion that only a subset of AML cases overexpresses β-catenin [4
]. Moreover, an elimination of β-catenin restricts the proliferation of a limited number of AML cells [11
] and variable combination indices resulted from LBH589/BC2059 cotreatment schedules in primary AML cells [23
]. Furthermore, a dysregulation of β-catenin in leukemic cells does not always imply their addiction to β-catenin [11
]. This also holds true for colon cancer cells with mutant β-catenin [56
] and β-catenin does not affect a lethal growth of lymphoma in mice [10
]. Likewise, if β-catenin was the only critical factor for leukemic cell growth, indomethacin would be far more useful for leukemia treatment.
Our finding that single elimination of β-catenin, MYC, and WT1 does not kill MV4-11 cells may equally hint that HDACi trigger complex apoptosis cascades that cannot be mimicked by a depletion of one stemness-associated transcription factor. Furthermore, a remaining expression of β-catenin, MYC, and WT1 may have ensured the survival of the investigated cells. Maybe a combined inhibition and a more prolonged reduction of these factors halts cell proliferation. Such an idea is supported by the finding that a CRISPR-Cas9-based deletion of β-catenin requires over 10 days to eliminate HEL cells [39
]. Since β-catenin cleavage products exert biologically important functions [16
], it is equally plausible that RNAi-mediated depletion cannot mimic the effect of HDACi-induced processing of β-catenin in response to a treatment with HDACi. Moreover, other HDACi-triggered mechanisms, including replication stress and DNA damage [21
], may be more relevant inducers of cell death. This idea agrees with the finding that HU also triggers the processing of β-catenin.
Remarkably, MYC is of predictive value for therapy success of AML patients who are treated with HDACi [19
] and we show that MYC is a target of RGFP966. Whether RGFP966 is useful in the clinic remains to be resolved. In rodents, repetitive applications of RGFP966 were safe and produced beneficial effects without toxicity to normal tissues [65
]. Our data illustrate that a reduction of MYC, irrespective of whether β-catenin is cleaved or not, is a good marker for apoptosis induction by HDACi. In addition to the HDAC1/HDAC2-dependent transcriptional regulation of MYC in pancreatic carcinoma cells [81
], our data collected with RGFP966 suggest that HDAC3 supports the expression of MYC in leukemic cells. Consistent with these data, others found that HDAC3 is necessary for the growth of MYC-driven lymphoma [65
]. Nonetheless, we cannot rule out that HDAC1 and HDAC2 maintain the expression of β-catenin, MYC, and WT1 in leukemic cells. Moreover, faster migrating products of MYC in LBH589-treated HL60 AML cells suggest a possible proteolysis of MYC [23
]. Irrespective thereof, the data that we collected with z-VAD-FMK suggest that HDACi decrease MYC independent of caspase activation.
Regarding WT1, HDACi trigger a loss of the WT1 protein through an induction of the E2 ubiquitin conjugase UBCH8 and a shutdown of the WT1
]. As recent data demonstrate caspase-dependent processing of WT1 upon replication stress and DNA damage in leukemic cells [55
], caspases may as well degrade WT1 upon an inhibition of class I HDACs in leukemic cells. The details on how HDACi alter the expression, stability, acetylation, and other posttranslational modifications of β-catenin, MYC, and WT1 as well as the physiological consequences thereof should be analyzed in future research.
Taken together, our results show that a loss of β-catenin, MYC, and WT1 are molecular markers for beneficial effects of HDACi. The data also suggest combining class I HDACi with indomethacin against leukemic cells.