Introduction

Acute lymphoblastic leukemia (ALL) in infants ( < 1 year of age at diagnosis) is a rare but highly aggressive type of leukemia in which ~80% of cases is driven by chromosomal translocations of the KMT2A gene (encoding Lysine methyltransferase 2 A). Translocations of the KMT2A gene on chromosome 11q23 lead to fusions of the N-terminus of KMT2A to the C-terminus of one of many known fusion partners. In infant ALL, KMT2A is most recurrently fused to either AFF1 (49% of the cases), MLLT1 (22% of the cases) or MLLT3 (16% of the cases)1,2. The event-free survival (EFS) rates for infant ALL patients without KMT2A translocations is 75-80%, whereas the EFS of KMT2A-rearranged infant ALL patients still remains 40%3,4,5, urging the need for more effective treatment strategies.

The three most recurring KMT2A fusion proteins (i.e. AFF1, MLLT1, MLLT3) hijack the transcriptional elongation machinery, thereby rewriting the transcriptomic and epigenetic landscape of the cell, KMT2A-rearranged ALL cells are often sensitive to epigenetic inhibitors6,7,8. A well-known class of epigenetic-based drugs are histone deacetylase (HDAC) inhibitors, which are able to reverse the acetylation status of histones as well as a variety of non-histone proteins9,10. The most commonly used and extensively studied HDAC inhibitors (HDACi) are non-selective compounds targeting a wide variety of HDAC isoforms or complete classes simultaneously11,12,13. Previous pre-clinical work from our laboratory demonstrated pronounced in vivo anti-leukemic activity of the broad-spectrum HDAC inhibitor Panobinostat (LBH589) in patient-derived xenograft (PDX) mouse models of KMT2A-rearranged infant ALL14. Unfortunately, additional experiments combining Panobinostat with other agents showed high levels of toxicity and often lethality in mice. In line with this, a clinical phase II study in which Panobinostat was combined with melphalan, thalidomide and prednisone (MPT) for the treatment of multiple myeloma (MM) showed severe toxicities resulting in the forced amendment of the protocol15. Also, clinical trials using the pan-HDAC inhibitor Romidepsin for the treatment of cutaneous T-cell lymphoma (CTCL) resulted in a total of 6 deaths whereof 5 cases occurred 16-60 hours after the medication had been administered16. Hence, global inhibition of multiple nonredundant HDACs and the lack of isoform-specific HDAC inhibition often result in difficult to predict side-effects, including (cardio)toxicity, anaemia, neutropenia, and thrombocytopenia17, limiting its application either as monotherapy or in combination with conventional chemotherapy18,19. However, over the past few years, more HDAC-specific inhibitors have been developed, which may overcome the adverse effects observed with non-selective HDAC inhibition.

The human HDAC superfamily comprises 18 proteins grouped into four classes (i.e., class I, IIA/IIB, III, and class IV) according to structure, function, expression pattern and sub-cellular localization, as well as their homology to yeast HDACs20,21. The class I HDACs have been studied most extensively in relation to leukemia, and expression levels of individual HDAC isoforms belonging to this class have been associated with clinical outcome10,22,23. In contrast to class I HDACs, the expression of class IIA HDACs is more heterogeneous, and tends to vary between different types of leukemia10. It has already been shown that class I HDAC inhibition is effective in various leukemic subtypes23,24,25 including KMT2A::AFF1 leukemias where class I HDAC inhibition resulted in reactivation of wildtype KMT2A and overwriting the dominant functions deriving from KMT2A:AFF126,27. Since class IIA HDAC isoforms are less well studied especially in KMT2A-rearranged infant ALL, the present study was designed to determine the dependency of this type of leukemia on each individual class IIA HDAC isoform. Based on these findings we explored the efficacy of selective HDAC inhibition against KMT2A-rearranged ALL in vitro as well as in vivo using xenograft mouse models.

Results

Knock-down of Class IIA HDACs in KMT2A-rearranged ALL cells

To determine the dependency of KMT2A-rearranged ALL cells on each individual class IIA HDAC isoform (i.e. HDAC4, HDAC5, HDAC7, HDAC9), shRNA-mediated RNA interference (RNAi) was used to successfully reduce mRNA levels in KMT2A-rearranged ALL cell line models SEM and ALL-PO (Fig. 1a). Knock-down of mRNA expression was accompanied by down-regulation at the protein level (Fig. 1b). Next, the effects of the knock-downs on cell viability and proliferation were assessed. Depletion of HDAC4, HDAC5, and HDAC7 in particular significantly compromised cell viability already after 24 hours and progressively decreased over time. In contrast, loss of HDAC9 had no effect on cell viability in KMT2A-rearranged ALL cells at any of the evaluated timepoints (Fig. 1c). Dependent on the cell line model and shRNA construct, impaired cell viability upon knock-down of HDAC4, HDAC5, and HDAC7 was a result of apoptosis induction (Fig. 1d), G1 cell cycle arrest (Fig. 1e), or a combination of these. Hence, specific inhibition of HDAC4, HDAC5, and/or HDAC7 may represent an attractive therapeutic strategy.

Fig. 1: HDAC4, HDAC5 and HDAC7 compromises cell viability upon RNAi knock-down in vitro.
figure 1

a Relative mRNA expression 72 hours post puromycin selection in two KMT2A-rearranged ALL cell line models SEM and ALL-PO (n = 3). Two shRNA constructs used per class IIA HDAC, except for HDAC7. b HDAC isoform protein expression 72 hours post puromycin selection. Representative western blots shown. c Fold change viable cells (n = 3) for both cell line models assessed using daily 7-AAD viability assay starting 24 hours till 96 hours post puromycin selection. d Apoptosis analysis for SEM and ALL-PO (n = 3) 96 hours post puromycin selection upon HDAC isoform depletion using Annexin-V/7-AAD assay. e Cell cycle distribution analysis based on cell DNA content flowcytometry 96 hours post puromycin selection (n = 2). Statistics were determined using Student’s t-test, * p < 0.05, ** p < 0.01, *** p < 0.005, **** p < 0.0001 and ns for no significant p. Error bars represent S.E.M.

Full size image

HDAC4/5-selective inhibitor LMK-235 recapitulates the knock-down phenotype

To determine whether the effects of HDAC knock-down on cell viability could be mimicked by pharmacological inhibition, several class II specific HDAC inhibitors targeting the whole class or only one or two isoforms were tested on a panel of KMT2A-rearranged ALL cell lines (n = 5) in vitro using 4-day cell viability assays. To date, there are no specific HDAC7 inhibitors available. However, a small molecule inhibitor specifically targeting both HDAC4 and HDAC5, i.e. LMK-235), does exist. For all tested HDAC inhibitors (Table 1) complete cell viability inhibition was observed in all cell line models at varying concentrations (Fig. 2a). The most consistent results at the lowest IC50 values of nanomolar concentrations were observed for the selective HDAC4/5 inhibitor LMK-235 (Fig. 2b) which was selected for further investigation. In contrast to previous observations with broad-spectrum HDAC inhibitors14,28 no specificity towards KMT2A-rearranged ALL was observed for LMK-235 or any other specific class II HDAC inhibitor, as wildtype KMT2A BCP-ALL cell lines responded equally as well (Supplementary Fig. 3a-b). Exposure of KMT2A-rearranged ALL cells to 100 nM or 200 nM of LMK-235 did not consequently reduce HDAC4/HDAC5 mRNA expression (Fig. 3a) but led to substantial down-regulation of both HDACs on protein level (Fig. 3b). The greatest reduction was observed using 200 nM LMK-235 and was therefore used for downstream experiments. This resulted in a drastic reduction of cell viability over time in various KMT2A-rearranged ALL cell line models treated with 200 nM LMK-235 (Fig. 3c). As previously seen, the reduced cell viability was caused by increased apoptosis (Fig. 3d), G1 cell cycle arrest (Fig. 3e), or a combination of these. The largest increase in apoptotic cells was seen for SEM, while the other cell line models showed a greater increase in G1 cell cycle distribution, indicating the mechanism behind cell viability inhibition upon LMK-235 treatment differs among cell line models.

Table 1 Targets class IIA HDAC inhibitors
Full size table
Fig. 2: Several class IIA HDAC inhibitors tested in KMT2A-rearranged ALL cell line models.
figure 2

a Dose response curves for the class IIA HDAC inhibitors LMK-235, BRD4354, MC1568, ACY1215, TMP269 TMP195 and CAY10603 in several KMT2A-rearranged ALL cell line models (n = 5) using 4-day MTT assay. b IC50 values for each class II HDAC inhibitor based on the dose response curves displayed under A). Error bars represent S.E.M.

Full size image
Fig. 3: HDAC4/5-selective inhibitor LMK-235 recapitulates the knock-down phenotype.
figure 3

a mRNA expression levels of HDAC4 and HDAC5 upon 100 and 200 nM LMK-235 treatment for 96 hours in three KMT2A-rearranged ALL cell line models. b HDAC isoform protein expression upon 100 and 200 nM LMK-235 treatment for 96 hours in three KMT2A-rearranged ALL cell line models. GAPDH was used as loading control. c Fold change viable cells upon 200 nM LMK-235 treatment assessed using daily 7-AAD viability assay starting 24 hours till 96 hours. d Apoptosis analysis upon LMK-235 treatment (200 nM) for 96 hours using Annexin-V/7-AAD flowcytometry analysis. e Cell cycle distribution analysis upon 96 hours LMK-235 treatment (200 nM) in three KMT2A-rearranged ALL cell line models based on cell DNA content flowcytometry. f Dose response curves for LMK-235 in KMT2A-rearranged primary patient samples from infants (n = 3) and non-leukemic whole bone marrow samples (n = 2) using 4-day MTT assay. g Cell viability upon 10, 100 and 1000 nM LMK-235 treatment using 4-day MTT assay in KMT2A-rearranged primary patient samples from infants (n = 3) and non-leukemic whole bone marrow samples (n = 2). Error bars represent S.E.M.

Full size image

Next, we evaluated whether LMK-235 was also effective against primary KMT2A-rearranged infant ALL patient samples. Dose response curve of these primary patient samples (n = 4) showed complete cell viability inhibition with IC50 values ranging between 40-100 nM (Fig. 3f-g). Importantly, with IC50 values of ~600 nM, whole bone marrow samples derived from non-leukemic individuals (n = 2) were significantly less affected by LMK-235. Important to note is that these KMT2A-rearranged primary patient cells hardly proliferate in vitro and show high apoptotic levels within 96 hours post culture start, limiting the resolution to detect effects on apoptosis. Nevertheless, a minor increase in apoptotic cells was seen already upon 48 hours induced by LMK-235 exposure (Supplementary Fig. 4). The LMK-235 doses for primary patient samples were adjusted to 200 nM and 300 nM due to their slightly lower sensitivity compared to cell lines.

In vivo efficacy of LMK-235 in a xenograft mouse model of KMT2A-rearranged ALL

Before studying the anti-leukemic effects of LMK-235 in vivo using a xenograft mouse model, pharmacokinetics of this compound was analysed in mice (n = 5 per condition) dosed with 20 mg kg-1 day-1 LMK-235 either administered daily orally or via i.p. injections for a total of 35 days. Oral dosing showed complete clearance in the blood within 4 hours (Supplementary Fig. 5a) at day 1 as well as after 29 days of treatment (Supplementary Fig. 5b). LMK-235 i.p. dosing at day 1 also showed rapid clearance within 4 hours but the mice retained plasma levels of 10–18 nM up to 24 hours post i.p. injection (Supplementary Fig. 5c). After 29 days of daily i.p. dosing an increase was observed in the plasma level concentration whereby an average concentration of 100 nM was maintained for 8 hours post i.p. injection (Supplementary Fig. 5d). This was similar to the IC50 values observed in vitro and resulting in anti-leukemic effects.

To study the anti-leukemic effects of LMK-235 on KMT2A-rearranged ALL cells in vivo, we established a xenograft mouse model enabling leukemia tracking by bioluminescence imaging. The KMT2A-rearranged ALL cell line SEM was modified to stably express the reporter proteins eGFP and luciferase (Supplementary Fig. 6a). After transduction, this modified cell line termed SEM-GFP-luc, was successfully sorted based on eGFP fluorescence and the signal remained stable for at least 11 weeks of in vitro culture, including a freeze/thaw cycle (Supplementary Fig. 6b). Additionally, luciferase activity was successfully tested and remained active for at least 10 weeks post transduction (Supplementary Fig. 6c). Importantly, SEM-GFP-luc displayed a comparable sensitivity towards LMK-235 as the parental cell line SEM (Supplementary Fig. 6d). Immunodeficient NRG-S mice were transplanted with SEM-GFP-luc cells, and after engraftment was confirmed by bioluminescence imaging on day 5, drug treatment was initiated on day 7 post transplantation (Fig. 4a, Supplementary Fig. 7a). After a three-week treatment schedule (5-days on 2-days off) with 20 mg kg-1 day-1 LMK-235, experimental animals were humanely sacrificed, and relevant tissues harvested for niche-specific efficacy analysis. No toxicities were observed based on necropsy and body weight during treatment duration (Supplementary Fig. 7b). Based on bioluminescence imaging, no reduction in overall disease burden was shown (Fig. 4b-c). LMK-235 did not result in reduced splenomegaly compared to controls, which is an initial indication for disease burden reduction (Fig. 4d). This was confirmed by the fraction of CD19-positive cells in the bone marrow, spleen, and peripheral blood as determined by flow cytometric analysis (Fig. 4e-g). Analysis of extracted bone marrow samples from control and treated animals showed that neither the mRNA nor protein expression levels of HDAC4 and HDAC5 were affected in mice treated with LMK-235 (Supplementary Fig. 7c-d), suggesting that the obtained drug concentration in vivo was insufficient to be effective.

Fig. 4: LMK-235 monotherapy treatment is unable to reduce overall disease burden in vivo.
figure 4

a Schematic of the experimental design. SEM-GFP-luc cells were injected into NRG-S mice; 7 days post-transplantation treatment was initiated after engraftment was confirmed by in vivo imaging. Mice were treated by i.p. injection with either LMK-235 (20 mg kg-1 day-1) or vehicle for total of 3 weeks with a 5-days on 2-days off schedule. Experimental end-point was reached directly after the last treatment cycle. b Graphical depiction of overall disease burden determined by the average radiance (photons/s/cm2/sr) emitted per mouse monitored every week until experimental end-point. Representative graphical depiction of bioluminescence imaging (n = 3, ordered vertically) for all experimental animals. Rows define experimental timepoints. c Quantification of overall disease burden determined by the average radiance (photons/s/cm2/sr) emitted per mouse monitored every week until experimental end-point. d Spleen weight (grams) upon experimental end-point. e Disease development was monitored by quantifying the fraction of human CD19-positive cells in the bone marrow f spleen and g peripheral blood, analysed after experimental animals were euthanised at the experimental end-point using flowcytometry. Statistical difference was determined by Student’s t-testing. Error bars represent S.E.M.

Full size image

Due to the limiting drug solubility of LMK-235, and therefore reaching the maximum achievable dose, we tested two additional class IIA HDAC inhibitors targeting all isoforms of this class, i.e., MC1568 and TMP195, to explore whether these agents may be effective using a similar experimental setup as described for LMK-235. SEM-GFP-luc displayed a comparable sensitivity towards MC1568 and TMP195 as the parental cell line SEM (Supplementary Fig. 8a-b). However, although the body weight loss of the mice was acceptable (Supplementary Fig. 8c-d), the experiments for both inhibitors was terminated after the second treatment cycle due to severe pharmacological toxicity for MC1568 (bloated and inflamed intestines and fluid in abdominal region) and a complete lack of anti-leukemic efficacy as determined by bioluminescence imaging for both MC1568 as well as TMP195 (Supplementary Fig. 8e-f). This suggested that based on the maximum achievable or tolerable dose, targeting of HDAC4/5 or class IIA HDACs is insufficient to obtain anti-leukemic effects.

In vivo efficacy of LMK-235 and Venetoclax combination treatment

Although LMK-235, MC1568, and TMP195 monotherapy did not seem to invoke anti-leukemic effects in our xenograft mouse model of KMT2A-rearranged ALL at maximum tolerable dosages, we found LMK-235 to be highly synergistic with the BCL-2 inhibitor venetoclax in vitro. In three different KMT2A-rearranged ALL cell line models (i.e., SEM, ALL-PO, KOPN-8) combinations of LMK-235 and venetoclax resulted in ZIP-synergy scores of up to 25 (Supplementary Fig. 9a-b) and substantial reductions in IC50 concentrations with the highest reduction seen for KOPN-8 resulting in a reduction from 130 nM to 16 nM (Supplementary Fig. 9c). Therefore, combination treatment was tested in vivo in a xenograft mouse model, with the rationale that the observed synergy may reduce the effective dose required for LMK-235 towards concentrations achievable in mice. Briefly, during and after a three-week treatment schedule (5-days on 2-days off) with 20 mg kg-1 day-1 LMK-235 and 100 mg kg-1 day-1 venetoclax, disease progression was monitored by weekly assessments of huCD19/huCD45-positivity in the peripheral blood and bioluminescence imaging until the humane endpoint was reached (Fig. 5a). Vehicle controls and LMK-235 treated animals showed a similar growth pattern of huCD45/huCD19+ cells in the peripheral blood, while mice treated with venetoclax as well as with LKM-235/venetoclax combination therapy clearly showed inhibition of leukemia progression in this niche. For the combination therapy, the percentage of huCD45/huCD19+ cells seemed to stabilize 5 weeks post transplantation, while for venetoclax monotherapy the percentage of human blasts slowly but steadily increased (Fig. 5b). This observation was confirmed by bioluminescence imaging (Fig. 5c-d). Additionally, the bioluminescence images from 6 weeks post transplantation were divided into four different niches (Supplementary Fig. 10a), namely head, thorax, spleen, and bone marrow. Although not significant, a trend was seen where LMK-235/venetoclax combination treatment showed a greater signal reduction in the head, thorax, and bone marrow compared to venetoclax monotherapy (Supplementary Fig. 10b). This may suggest modest effects of efficacy enhancement between LMK-235 and venetoclax, especially since LMK-235 monotherapy showed no effect at all. However, the observed reductions in human leukemic cell percentages and bioluminescence in specific disease reservoirs induced by LMK-235/venetoclax treatment as compared to venetoclax monotherapy were not sufficient to prolong survival in treated mice (Fig. 5e).

Fig. 5: In vivo efficacy of LMK-235/venetoclax combination treatment.
figure 5

a Schematic of the experimental design. SEM-GFP-luc cells were injected into NRG-S mice and 4 days post-transplantation treatment was initiated. Mice were treated with either vehicle, LMK-235 (20 mg kg-1 day-1), Venetoclax (100 mg kg-1 day-1) or a combination for a total of 3 weeks with a 5-days on 2-days off schedule. Experimental end-point was reached after mice succumbed to leukemia. b Disease development was monitored by quantifying the fraction of human CD19/CD45-positive cells weekly in the peripheral blood. c Overall disease burden was quantified by determining the average radiance (photons/s/cm2/sr) emitted per mouse monitored weekly until experimental end-point. d Graphical depiction of bioluminescence imaging (ordered vertically) for all experimental animals. Rows define experimental timepoints. Red crosses represent deceased mice. e Survival analysis illustrated by Kaplan-Meyer plot. Statistical significance for survival was determined using log-rank testing between venetoclax monotherapy and LMK-235/venetoclax combination therapy. Statistical significance for disease burden between venetoclax monotherapy and LMK-235/venetoclax combination therapy was determined using Mann Witney-U testing. Error bars represent S.E.M.

Full size image

Discussion

In the present study we showed that knock-down of the class IIA HDAC isoforms HDAC4, HDAC5, and HDAC7 negatively affects the viability of KMT2A-rearranged ALL cells. The largest induction of apoptosis was observed as a consequence of downregulating HDAC7. Interestingly, Barrios et al. reported that HDAC7 exerts a leukemia-suppressing phenotype in KMT2A::AFF1+ pro-B ALL, where low expression of HDAC7 was associated with a poor clinical outcome. In that study, enforced overexpression of HDAC7 impaired proliferation and induced apoptosis in the KMT2A::AFF1+ cell line SEM-K229. Hence, the level of HDAC7 expression in KMT2A-rearranged ALL cells must be tightly balanced, as both upregulation as well as downregulation of HDAC7 results in apoptosis induction. Moreover, HDAC isoform expression levels and their influence on clinical outcome seem to vary between ALL subtypes. While low HDAC7 expression is related to a poor prognosis in KMT2A-rearranged ALL, high-level expression of HDAC7 represents an adverse prognostic factor in childhood ALL22,29. In contrast, overexpression of HDAC9 appears to be associated with a poor clinical outcome in BCP-ALL in general10,22, while HDAC9 knock-down did not influence the viability of KMT2A-rearranged ALL cell lines in our current study.

Previously increased sensitivity towards broad-spectrum HDAC inhibitors was observed in KMT2A-rearranged ALL as compared to wild-type KMT2A28. However, this is not the case for class IIA HDAC inhibition. It has been described that class I HDAC inhibition is able to activate wild-type KMT2A and overwrites the dominant functions exerted by the KMT2A fusion protein. This may well explain the enhanced sensitivity for broad-spectrum HDAC inhibitors26,30 but not towards more specific class II HDAC inhibitors.

Despite the development of a wide variety of HDAC inhibitors, no HDAC7-specific inhibitor is currently available. However, a selective small molecule simultaneously inhibiting both HDAC4 and HDAC5, i.e., LMK235, is at hand. In concordance with shRNA-mediated down-regulation of HDAC4 and HDAC5, exposure to LMK-235 reduced cell viability and induced apoptosis in primary KMT2A-rearranged infant ALL cells in vitro. Unfortunately, as a single agent LMK-235 failed to exert anti-leukemic effects in vivo in a xenograft mouse model of KMT2A-rearranged ALL. This was most likely due to rapid clearance of the drug as observed in treated mice, as well as to its maximum solubility limiting the administrable dose. Likewise, no anti-leukemic effects were observed in similar experiments with the class IIA HDAC inhibitors MC1568 and TMP195. In contrast to reported mouse experiments with MC156831,32,33,34, we observed severe toxicity in mice with this drug resulting in preliminary experiment termination. TMP195 did not induce toxicity but the experiments with this drug were preliminary terminated due to a complete lack of efficacy. Rapid HDAC inhibitor clearance in vivo has been described previously in literature. For example, preclinical pharmacokinetics of the class I specific HDAC inhibitor MS-275 (Entinostat) indicated a half-life of about 1 hour in mice, while a phase I trial in patients with solid tumors showed a much longer half-life of about 39-80 hours35. Although not much is known about the half-life of LMK-235 in preclinical studies nor in clinical trials, a similar phenomenon might be true for LMK-235 and it could well be that the compound is more stable in humans. Additionally, most preclinical studies showing efficacy for LMK-235 have been using a variety of different mouse strains instead of an immunodeficient mouse model. Possibly, the type of mouse strain might influence the rate of drug clearance36,37,38.

However, based on our in vivo mice experiments, the current generation of isoform-specific class II HDAC inhibitors are not adequate as single agent for the treatment of KMT2A-rearranged ALL and inhibitors with a more favorable pharmacokinetics profile and stability are needed. The mode of action of most class II HDAC inhibitors have not been described in literature, hence it could be of interest to investigate its functioning mechanism and potentially improve these or develop inhibitors with an alternative mode of action. Alternatively, efforts can be made to increase solubility, improve pharmacokinetics and targeting specific leukemic niches by using drug delivery systems. Major progress has been made the past few years in the use of nanotechnology-based platforms as drug delivery system for HDAC inhibitors39,40,41. A Phase I/II trial was initiated to test the Panobinostat nanoparticle formulation MTX110 in treating participants with newly diagnosed diffuse intrinsic pontine glioma (ClinicalTrials.gov Identifier: NCT03566199, NCT04264143) and recurrent medulloblastoma (ClinicalTrials.gov Identifier: NCT04315064). Another potential alternative to improve HDAC inhibitor stability is to make use of stabilized peptide HDAC inhibitors rather than small molecule inhibitors, which have been shown to be successful in vitro as well as in vivo42,43.

HDAC inhibition monotherapy in general is unlikely to proceed in clinical use for this highly aggressive type of infant ALL but may eventually be successful in the background of a combination treatment. Promising in vitro data showed synergism between LMK-235 and the BCL-2 inhibitor venetoclax, which significantly reduced IC50 values for LMK-235 that can be obtained in vivo. Although, LMK-235/Venetoclax combination treatment did not result in enhanced survival compared to venetoclax monotherapy, a trend towards lower leukemic infiltration in the head, thorax and bone marrow niches was seen. It has been described by others that LMK-235 is able to pass the blood-brain-barrier37,44, and thus might explain the reduction of leukemic infiltration of the head niche resulting in reduced central nervous system (CNS) infiltration. CNS infiltration remains one of the biggest challenges in KMT2A-rearranged infant ALL as it is present in up to 10% of the patients at diagnosis4 and in 30-40% of the cases at relapse45,46 resulting in an even worse prognosis than already known for this type of leukemia3,4.

Taken together, our findings show that isoform specific HDAC inhibition is successful with great anti-leukemic effects in vitro in KMT2A-rearranged cell line models as well as infant primary patient samples. Based on our data, HDAC-specific inhibition as monotherapy will be unlikely to be an effective therapeutic approach in infant KMT2A-rearranged ALL. In combination therapy niche specific clearance might be supported by specific HDAC inhibition, especially when efforts are made in the development of novel inhibitors with greater stability and improved pharmacokinetic profiles than observed thus far in our in vivo xenograft models.

Methods

Cell culture

Cell lines were maintained as previously described28. SEM, ALL-PO, RS4;11, BEL-1 and KOPN-8 are KMT2A-rearranged B-ALL cell line models carrying the KMT2A::AFF1 gene fusion, except for KOPN-8 which carries a KMT2A::MLLT1 fusion. Nalm-6 (ETV6::RUNX1), 697 (TCF3::PBX1), SupB15 (BCR::ABL1), REH (ETV6::RUNX1) and MHHCALL2 (high hyperdiploid) represent B-cell progenitor ALL (BCP-ALL) cell lines without KMT2A translocations. ALL-PO was a gift from Dr. Cazzaniga (University of Milano-Bicocca, Milano MI, Italy), and BEL-1 was a gift from Dr. Ruoping Tang (University Laboratory, Paris, France). All the other cell lines were originally purchased from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). Regular DNA fingerprinting and mycoplasma testing was performed to ensure cell line quality and integrity.

Primary samples derived from KMT2A-rearranged positive infant ALL patients and were obtained as part of the international collaborative INTERFANT studies3,4, and informed consent was obtained according to the Declaration of Helsinki. Whole bone marrow (BM) samples were derived from children suspected to suffer from leukemia or other cancer types but who turned out to have healthy BMs. Sample processing occurred as described previously47. Leukemic blast percentages in the primary patient samples was at least 90% and non-leukemic BM controls contained less than 1% blasts, as confirmed by May-Grünwald-Giesma (Merck) counterstained cytospins.

Lentiviral transductions

Lentivirus constructs were packaged in HEK-293T cells by transient transfection in combination with the packaging constructs psPax2 (Addgene #12260) and VSV-g (Addgene #14888). Transfections were carried out using the X-tremeGENETM 9 DNA Transfection Reagent (Roche) according to the manufacturer’s recommendations in Opti-MEM. The virus was harvested 48 hours after transfection and concentrated using vivaspin-20 columns (Sigma-Aldrich). Lentiviral short-hairpin RNA (shRNA) expressing pLKO.1-Puro vectors were designed and purchased from the Erasmus Center for Biomics (Erasmus MC, The Netherlands). Target sequences are listed in Supplementary Table 1. The GFP-luc construct to stably express the reporter proteins eGFP and luciferase was a kind gift of Dr. Vaskar Saha (University of Manchester, Manchester, UK)48. Leukemic cells were transduced via spinfection by incubation with a single dose of lentiviral supernatant supplemented with 4 µg/ml Polybrene (Bio-Connect) for shRNA-mediated knock-down experiments. Puromycin selection started 24 hours post-transduction with 1–1.5 µg/ml (Sigma Aldrich) for 72 hours, whereafter density gradient centrifugation with Lymphoprep (Nycomed Pharma) was used to remove dead cells as previously described47. SEM transduced with GFP-luc construct was sorted based on eGFP signal using the SH800S Cell sorter (Sony Biotechnology).

Viability and cell cycle assay

Cell growth and viability was assessed using 7-amino-actinomycin D (7-AAD) (Biolegend) and AnnexinV (BD Pharmingen) cell staining followed by flow cytometric analysis using the FlowJo software (VX10) measured with CytoFLEX Flow Cytometer (Beckman Coulter). Representative gating strategy for cell count and apoptosis analysis can be found in Supplementary Fig. 1a-b. Cell-cycle distribution was analysed as previously described49 using ModFit LT software (Verity Software House). Representative gating strategy shown in Supplementary Fig. 1c.

Quantitative reverse-transcription PCR

Total RNA was isolated using the RNeasy Mini Kit (Qiagen) and cDNA was generated by using the SensiFast cDNA Synthesis Kit (Bioline). Quantitative reverse-transcription PCR (qRT-PCR) was performed with iQ SYBR Green Supermix (Bio-Rad) according to manufacturer’s protocol. Expression levels were normalized against the reference gene beta-2-Microglobulin (B2M) and calculated according to the 2-ΔΔCT method50. Primer sequences of target genes as well as the reference gene are listed in Supplementary Table 2.

Immunoblotting

Protein was isolated concurrently to RNA using the RNeasy Mini Kit (Qiagen), as described previously49, resolved on precast SDS-polyacrylamide gels (TGX, Bio-Rad), and transferred to nitrocellulose membranes using a Transblot Turbo Transfer System (Bio-Rad). Membranes were blocked with 5% milk (Elk, Campina) and subsequently probed with antibodies against HDAC4 (#7628S, 1:2000), HDAC5 (#20458S, 1:2000)) and HDAC7 (#33418S, 1:2000) purchased from Cell Signaling; HDAC9 (#ab109446, 1:2000) from Abcam. GAPDH (#2118S, 1:1000) or β-actin (#3700S, 1:50.000) from Cell Signaling were used as loading controls. Visualization occurred after incubation with IRDye 680RD/800CW conjugated secondary antibodies using the Odyssey imaging system (LICOR Biotechnology). Immunoblots were quantified by analysing the intensity of the target band compared to the loading control using Image Studio Lite software (V5.2). If not otherwise indicated, representative blots of multiple independent experiments are shown. All uncropped western blots can be found in Supplementary Fig. 2.

In vitro drug exposures

LMK-235 (HY-18998), Venetoclax (HY-15531), MC1568 (HY-16914), TMP195 (HY-18361), TMP269 (HY-18360), ACY-1215 (HY-16026), BRD4354 (HY-112719) and CAY10603 (HY-18613) were purchased from MedChemExpress and dissolved in DMSO. Briefly, cells were seeded semi-automatically in 384-wells (Corning Inc.) using the Multidrop (Thermo Fisher Scientific), followed by the administration of the drugs at different concentrations using the Tecan D300 Digital Dispenser (Tecan). Cell viability was assessed using 4-day thiazolyl blue tetrazolium bromide (MTT) assays (Sigma-Aldrich) as previously described51. Data was normalized to DMSO controls, allowing a maximum concentration of ≤0.5% (v/v).

Animal experiments

All animal experiments, if not stated differently, were carried out according to Dutch legislation and approved by the Animal Ethics Committee (Approval Number: AVD3990020173066 & AVD39900202216167) at the Princess Máxima Center, Utrecht, The Netherlands. The pharmacokinetics study for LMK-235 was approved by the Animal Ethics Committee (Approval Number: AVD301002016407) at the Netherlands Cancer Institute, Amsterdam, The Netherlands.

The pharmacokinetics study for LMK-235 was carried out using immunodeficient NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG) mice. For in vivo drug efficacy studies, immunodeficient NOD.Cg-Rag1tm1Mom Il2rgtm1Wjl Tg(CMV-IL3,CSF2,KITLG)1Eav/J (NRG-S) female mice were transplanted intravenously with the KMT2A-rearranged ALL reporter cell line SEM-GFP-luc (105 cells/mouse, aged 6-20 weeks), and leukemia progression was assessed weekly by intravital imaging of human CD45+ (huCD45+) and/or human CD19+ (huCD19+) leukemic blasts in peripheral blood and relevant tissues using flow cytometry. Representative gating strategy is shown in Supplementary Fig. 1d. Both PE-Mouse Anti-Human CD45 (#555483, 1:50) as well as APC Mouse Anti-Human CD19 (#555415, 1:50) were purchased from BD Biosciences. Group sizes were estimated based on power calculation and consultancy with a biostatistician for statistical power. Transplanted mice were randomly allocated to different treatment arms based on body weight and/or bioluminescence (BLI) signal using RandoMice52. For intra-vital imaging, mice were injected subcutaneous with luciferase substrate (150 mg/kg) (XenoLight D-Luciferin - K+ Salt Bioluminescent Substrate, Perkin Elmer) and imaged using the IVIS Spectrum Imaging System (Perkin Elmer). Images were analysed with the Living Image software (Perkin Elmer) and whole-body radiance (total photon flux) was determined using the region of interest tool as closely to the body outline as possible Treatment with LMK-235 (20 mg kg-1 day-1), MC1568 (50 mg kg-1 day-1) and TMP195 (50 mg kg-1 day-1), all purchased from MedChemExpress, was initiated 3 days post-transplantation and administered intraperitoneally 5 times a week for a total of 15 days. Venetoclax (100 kg-1 day-1) treatment was administered via oral gavage 5 times per week for a total of 15 days. Intraperitoneally administered drugs and vehicle were dissolved in 10% DMSO (Sigma-Aldrich), 40% PEG-300 (Merck), 5% Tween-80 (Sigma-Aldrich) and 45% saline (Gibco). Oral gavage administered drugs and vehicle were dissolved in 60% phosal 50 propylene glycol (PG) (MedChemExpress), 30% PEG-400 (Merck) and 10% ethanol (Sigma-Aldrich). Mice were humanely euthanized after treatment halt or upon reaching the experimental end-point (defined as >85% huCD45 + /huCD19+ cells in PB for the used xenograft model) or clinically overt leukemia, and tissue samples were acquired for further investigation. In none of the experiments these limits were exceeded and we have complied with all relevant ethical regulations for animal use.

Pharmacokinetic profiling

Pharmacokinetic profiling for LMK-235 was performed as followed. Sample pre-treatment was accomplished by mixing 5 µL mouse plasma, 5 µL human plasma and 60 µL formic acid in acetonitrile, followed by centrifugation. The clear supernatant was diluted 1:4 and 40 µL suspension was injected into the liquid chromatography triple quadrupole mass spectrometry (LC-MS/MS) system and analysed using an API4000 detector (Sciex). LMK-235 was detected in positive ionization mode (MRM: 295.3/105.2). LC separation was achieved using a Zorbax Extend C18 column (100 × 2.0 mm: ID). Mobile phase A and B comprised 0.1% formic acid in water and methanol, respectively. The flow rate was 0.4 ml/min and a linear gradient from 10%B to 95%B in 2.5 min, followed by 95% B for 2 min. Re-equilibration was performed at 20%B for 10 min and used for elution.

Statistics and reproducibility

Statistical significance of independent experimental replicates was determined using two-sided Student’s t-tests. In vitro drug combination analysis was performed using the Synergy Finder 2.0 web application53. Survival differences between control and treated mice were analysed using log-rank testing and Student’s t-test for differences in leukemic burden. Statistical significance was defined as P values < 0.05. Software packages utilized were GraphPad Prism 8 (GraphPad Software) and Microsoft Excel. Individual datapoints are shown in graphs to identify the number of independent replicates per experiment.