Background

Chromosomal translocations involving the KMT2A (MLL) gene constitute the cytogenetic hallmark of acute lymphoblastic leukemia (ALL) diagnosed in infants (< 1 year of age), giving rise to an aggressive malignancy with high relapse rates and low event-free survival (EFS) chances of 30–40% [1, 2]. Hence, currently available treatment regimens for KMT2A-rearranged infant ALL are inadequate and require more effective therapeutic options to improve clinical outcome.

KMT2A translocations result in the fusion of KMT2A to one of its many translocation partner genes, [3] generating chimeric transcripts encoding highly oncogenic KMT2A fusion proteins. Among infant ALL patients, KMT2A is most recurrently fused to either AFF1 (AF4), MLLT1 (ENL), or MLLT3 (AF9) [3]. Functionally, wild-type KMT2A plays an essential role in definitive hematopoiesis [4] regulating gene expression through histone 3 lysine 4 (H3K4) methyltransferase activity, mediated by its Su(Var)3–9, Enhancer-of-zeste, Trithorax (SET) domain [5]. In contrast, KMT2A fusion proteins lose their SET domain, but instead recruit the histone 3 lysine 79 (H3K79) methyltransferase DOT1L through binding motifs encoded by the translocation partner genes [6,7,8]. Binding of DOT1L to KMT2A fusion proteins causes inappropriate H3K79 hypermethylation at KMT2A target genes, leading to an altered transcriptomic landscape that strongly favors leukemia development [6,7,8,9].

Interestingly, the mutational landscape of KMT2A-rearranged infant ALL is remarkably silent, with only 1.3–2.5 leukemia-specific, non-silent mutations in the dominant clone per patient, [10, 11] suggesting that the KMT2A translocation may well be the sole oncogenic lesion driving this aggressive type of leukemia [12]. Therefore, targeting DOT1L, through which KMT2A fusion proteins exert their oncogenic effects, represents an attractive therapeutic strategy. Accordingly, the development of the DOT1L inhibitor EPZ004777 and its successor EPZ5676 (pinometostat) [13, 14] were expected to become key to successful treatment of KMT2A-rearranged acute leukemias. However, despite promising preclinical results, subsequent clinical trials revealed that good initial responses in patients treated with pinometostat readily led to non-responsiveness due to acquired resistance and poor pharmacokinetics (PK) [15, 16]. Previously, it was reported that acquired resistance to pinometostat in KMT2A-rearranged acute leukemic cell lines is associated with enhanced drug efflux mediated by the elevated expression of the multidrug resistance transporters ABCB1 and ABCG2 [17].

Despite this, targeting DOT1L remains a promising avenue for treating KMT2A rearranged leukemias, and novel small-molecule DOT1L inhibitors with improved PK profiles have already been identified [40,41,42] and has been exploited as a therapeutic target for AML. In SEM a subpopulation of 20% of CD33-positive cells was identified, which was increased in SEMPINO_RES to about 40%, yet CD33-positive cells again were largely absent in SEMPINO−INT (Fig. 5D,F). Similarly, in RS4;11 a CD33-positive subpopulation of 9% was increased upon pinometostat resistance to 14% in RS4;11PINO_RES#1 and up to 50% in RS4;11PINO_RES#2 (Fig. 5D,F).

Together these data indicate that under prolonged pressure of DOT1L inhibition, KMT2A-rearranged ALL cells seem to initiate a reprogramming process that involves the acquisition (or selection) of myeloid-like characteristics.

Drug screens reveal minimal cross resistance, and sensitization towards venetoclax after acquired pinometostat resistance

Finally, acquired pinometostat resistance led to the upregulation of the multidrug efflux pump ABCB1 (Fig. 2F and Figure S6A,B), associated with multidrug resistance and previously reported as the mechanism of pinometostat resistance in KMT2A-rearranged acute leukemia cell lines [17]. However, our data challenges the concept that elevated ABCB1 expression alone is the mechanism of resistance to DOT1L inhibition as reported previously [17]. Despite significant ABCB1 upregulation, we still observe comparable inhibition of H3K79 methylation in SEM and SEMPINO_RES (Fig. 1D and E), indicating that the amount of pinometostat and/or its retention in SEMPINO_RES cells is sufficient to effectively reduce H3K79me2 levels, overriding the impact of ABCB1 upregulation.

Since multidrug efflux pumps are associated with multidrug resistance [43, 44], we assessed whether SEMPINO_RES cells had become more resistant to current chemotherapeutics for KMT2A-rearranged infant ALL [1, 2] and whether we could identify agents to which SEMPINO_RES cells had become more sensitive by performing drug library screens (Table S2, Fig. 6A,B).

Fig. 6
figure 6

Moderate levels of cross-resistance and substantial sensitization towards venetoclax in pinometostat-resistant KMT2A-rearranged ALL cells. A. Drug response curves for prednisolone, dexamethasone, vincristine, daunorubicin, cladribine, cytarabine, and L-asparaginase as determined by 4-day MTT assays in SEM and SEMPINO_RES cells with n = 4 biological replicates, each comprising n = 3 technical replicates, and presented as the mean ± standard error of the mean (SEM). B. Comparison of the IC50 in SEM and SEMPINO_RES cells for a total of 679 compounds tested in drug library screens. C. Showing the top 10 agents with the highest fold-changes in IC50 values in SEMPINO_RES as compared to SEM, indicating enhanced resistance in SEMPINO_RES cells. D. The top 10 drugs with the lowest fold-changes in IC50 values in SEMPINO_RES as compared to SEM, indicating enhanced sensitivity in SEMPINO_RES cells. E. Drug response curves for venetoclax as determined by 4-day MTT assays in indicated cell lines models (in duplicate), presented as the mean ± standard error of the mean (SEM)

Full size image

This revealed an increased resistance to the glucocorticoids dexamethasone and prednisolone (the liver-activated form of prednisone), vincristine, daunorubicin, and L-asparaginase, and increased sensitivity to cytarabine and for instance to the BCL-2 inhibitor venetoclax (Fig. 6A-E). Interestingly, cytarabine typically represents a drug commonly used in the treatment of myeloid leukemias, and the combination of venetoclax and cytarabine has successfully been tested in AML patients [45,46,47].

Discussion

The currently accepted dogma of KMT2A-fusion driven leukemogenesis dictates the requirement of DOT1L-mediated activation of KMT2A target genes through H3K79 methylation [6, 9, 48, 49]. Therefore, targeting DOT1L [13, 18, 19] represents an attractive therapeutic option for patients diagnosed with KMT2A-rearranged acute leukemia, despite the first-in-class DOT1L inhibitor pinometostat showing dissatisfying results in adult patients [15]. While next generation DOT1L inhibitors with improved pharmacokinetic profiles are in development, [18, 19] we reasoned that the mechanisms by which KMT2A-rearranged acute leukemia cells evade DOT1L inhibition may provide novel insights into the biology of these unique malignancies. Therefore, and in a similar fashion as published by Campbell and colleagues [17], we efficiently induced acquired pinometostat resistance in various cell line models, demonstrating how readily KMT2A-rearranged acute leukemia cells become resistant to DOT1L inhibition. The study of Campbell et al., mainly focused on examples of possible mechanisms of pinometostat resistance including increased expression of drug efflux transporters and activation of the PI3K/AKT and RAS/RAF/MEK/ERK pathways [17]. In contrast, we here specifically focused on the behavior and epigenetic regulation of DOT1L-associated KMT2A fusion-driven target genes and on how the transcriptomic landscape changes in KMT2A-rearranged ALL cells that are able to evade leukemic cell death during prolonged inhibition of DOT1L-mediated H3K79 methylation.

In concordance with the previous finding by Campbell et al., we found increased expression of the multi-drug efflux pump ABCB1 in our pinometostat-resistant SEMPINO_RES cells. However, despite the elevated levels of ABCB1 expression, pinometostat continued to inhibit H3K79 methylation in SEMPINO_RES. This strongly indicates that the increased levels of ABCB1 are insufficient to prevent pinometostat from exerting its inhibitory effects, and therefore cannot be the sole mechanism of acquired pinometostat resistance.

Interestingly, while cell viability of SEMPINO_RES cells was no longer affected by pinometostat-induced inhibition of DOT1L-mediated H3K79 methylation, these cells remained dependent on the physical presence of DOT1L protein. This may indicate that recently described biological functions of DOT1L that are independent of H3K79 methylation [50,51,52] are also important for KMT2A-rearranged acute leukemia cells. Thus, in addition to its enzymatic methylatransferase activity, DOT1L clearly has a scaffold function in assembling transcriptionally competent complexes. Therefore, therapeutic degradation of DOT1L instead of solely inhibiting its catalytic activities might be beneficial in the treatment of KMT2A-rearranged acute leukemia.

Another intriguing aspect of our model of acquired pinometostat resistance is the observation that SEMPINO_RES cells remained vulnerable to knockdown of the KMT2A::AFF1 fusion gene. This may suggest that inappropriate recruitment of DOT1L to loci otherwise not associated with H3K79me2 may not represent the sole KMT2A fusion-mediated attribute driving leukemogenesis and/or leukemia maintenance. If so, the identification of such DOT1L-independent oncogenic properties may well uncover important therapeutic targets and more effective treatment options for KMT2A-rearranged acute leukemias.

As shown, acquired resistance to pinometostat led to marked transcriptional downregulation of putative KMT2A-fusion target genes, which was accompanied by reductions in H3K79me2, as well as loss of binding of KMT2A and AFF1, and chromatin condensation at the corresponding loci. The complete loss of PROM1/CD133, which was shown to be transcriptionally regulated via KMT2A::AFF1-mediated H3K79me2/3 enhancer–promoter interactions, [29] in SEMPINO_RES is highly remarkable, since the expression of PROM1/CD133 was reported to be essential for leukemic cell growth in KMT2A-rearranged ALL [29, 30]. Consequently, targeting PROM1/CD133-positive cells has been proposed as a therapeutic option for KMT2A-rearranged ALL, although the expression of PROM1/CD133 on both fetal and adult hematopoietic stem cells (HSCs) may compromise the specificity of such an approach [29, 30, 53, 54]. Moreover, PROM1/CD133 is expressed in most, but not all, KMT2A-rearranged acute leukemia patients, and its presence seems to reflect the immunophenotype and/or cell of origin of the leukemia, as HSCs and early progenitors typically express PROM1/CD133, while more differentiated B-cell progenitors do not [29, 54, 55]. As acquiring resistance to DOT1L inhibition was accompanied by a complete loss of PROM1/CD133 expression, this may suggest that prolonged exposure to pinometostat triggered the differentiation towards (or selection of) a more mature immunophenotype. On the other hand, our data revealed that prolonged exposure of KMT2A-rearranged ALL cells to pinometostat seems to initiate a reprogramming process that involves the acquirement (or selection) of myeloid-like characteristics. Co-expression of myeloid CD markers, including CD33, [36] represents a familiar phenomenon in KMT2A-rearranged infant ALL with prognostic relevance [37, 38]. Moreover, a recent single-cell multiomics study by Chen and co-workers revealed the presence of pre-existing lymphomyeloid primed progenitors and myeloid blasts in diagnostic samples derived from KMT2A-rearranged B-ALL patients [56]. From this perspective, prolonged inhibition of DOT1L seems to favor KMT2A-rearranged leukemia cells that completely lack PROM1/CD133 but do display LILRB4/CD85k and CD33 expression. Interestingly, both LILRB4/CD85k and CD33 are therapeutic targets in AML [57,58,59,60,61,62,63] and have shown potential as therapeutic vulnerabilities in KMT2A-rearranged ALL. Targeting LILRB4/CD85k with antibody-conjugates [57] or anti-LILRB4 CAR-T cells [64, 65] and/or CD33 with gemtuzumab ozogamicin, could prevent resistance to DOT1L inhibitors in KMT2A-rearranged ALL. Moreover, combining BCL-2 inhibition by venetoclax with agents targeting DOT1L, LILRB4/CD85k, and/or CD33 may enhance the efficacy of these drug combinations. Venetoclax was found to synergize with DOT1L inhibitors [25, 66] and is being evaluated in clinical trials for pediatric KMT2A-rearranged leukemias [67, 68].

Taken together, we present an in vitro model of acquired resistance to DOT1L inhibition in KMT2A-rearranged ALL, revealing selective loss of epigenetic regulation and gene expression of KMT2A-fusion target genes, accompanied by upregulation of myeloid-like characteristics. This study may not only impact the development of novel DOT1L inhibitors, but also reveal key characteristics of KMT2A-rearranged ALL cells that are able to evade therapy, providing therapeutic targets to prevent that.