Targeting HASPIN kinase disrupts SR protein–mediated RNA splicing and synergizes with BCL-2 inhibitor venetoclax in AML Open Access

Acute myeloid leukemia (AML) is a cancer defined by an expansion of abnormal, undifferentiated hematopoietic cells in the body. AML constitutes one-third of all leukemia cases diagnosed per year.¹ With the current available treatments, 5-year overall survival of patients with AML is just 24%. Furthermore, survival rates decrease but incidence rates increase with advancing patient age. Although recommended treatment strategies are revised annually,^2,3 AML remains a significant clinical burden that demands additional research into underlying molecular mechanisms so that improved therapeutics can be devised.

Hematopoietic stem cell maintenance, proliferation, and differentiation are regulated by families of protein kinases. Receptor and nonreceptor tyrosine kinases initiate and propagate signaling cascades to strongly affect cellular behavior. Similarly, serine/threonine protein kinases directly phosphorylate substrates to regulate function. Signaling modulator and kinase mutations are present in at least two-thirds of AML cases and are the most common genetic lesion in patients with AML.⁴ In addition, concurrent kinase mutations confer worse prognoses, higher risk of disease relapse, and, consequently, lower overall survival rates.² Pathogenic mutations grant proliferative advantages by constitutively activating receptor kinases and associated downstream pathways, such as with mutant fms related receptor tyrosine kinase 3 (FLT3) or KIT proto-oncogene, receptor tyrosine kinase (c-KIT) activating phosphoinositide 3-kinase/protein kinase B (PI3K/AKT) pathways.^4,5 Importantly, leukemia can also develop “addictions” to nonmutated protein kinases.^6,7 Normally, dispensable kinases can become critical vulnerabilities in leukemia after adopting key roles in pathways mediating unchecked proliferation,⁸ metabolic alterations,⁹ and blocked differentiation.¹⁰ Notably, because these dependencies are specific to cancerous genetic environments, they are uniquely amenable to use as therapeutic targets.

AML treatment has not changed significantly for several decades.² Intensive chemotherapy is broadly prescribed to newly diagnosed patients with AML aimed at achieving prolonged remission. However, patients still experience high relapse rates and poor survival after relapse,¹¹ indicating a need for better treatment strategies. Protein kinases are actively being researched as candidates for targeted inhibition therapy in AML.^12,13 Several multireceptor tyrosine kinase inhibitors have been approved for AML, such as FLT3 mutant-specific inhibitors, and have demonstrated clinical success.² Furthermore, selective inhibitors for nonmutated serine/threonine kinases critical to proleukemic processes are actively being applied in clinical trial.^14,15 However, despite improved outcomes, many patients treated with these agents relapse and develop drug resistance.¹¹ Combination therapies are proven to bypass drug resistance and increase treatment efficacy in patients with cancer.^16-18 Effective combinations target multiple dysregulated pathways to produce simultaneous toxic effects that cannot be easily adapted to by cancer cells.¹⁹ Consequently, this strategy demands a growing repertoire of effective molecular inhibitors and a deep understanding of associated disease mechanisms from which new and rational combinations are designed. Currently, only ∼30% of the human kinome is being researched for potential use in targeted treatment strategies,^12,13 indicating that significant potential remains for the discovery of new drug targets. CRISPR screening has been highly effective at identifying novel genetic dependencies in AML.^10,20,21 Domain-targeted CRISPR screens extend the approach by targeting genes at functional domains to enhance screen sensitivity.²² 

Here, we use a human kinase domain–targeted CRISPR screen to discover novel kinase dependencies in human AML. We identify histone H3-associated protein (HASPIN) kinase as a novel kinase dependency. HASPIN has been reported as a mitotic kinase responsible for coordinating spindle fiber attachments, but it has not been investigated in the context of hematopoiesis and leukemia. Using a series of functional and bioinformatic analyses, we highlight the underappreciated role of HASPIN as a messenger RNA (mRNA) splicing regulator. In addition, in vitro testing revealed robust and leukemia-specific therapeutic effects from HASPIN inhibitor modulation both alone and in combination with the B-cell lymphoma 2 (BCL-2) inhibitor venetoclax, highlighting the potential clinical value of HASPIN as a novel target for AML.

See the supplemental Methods for complete description of methods.

CRISPR screen was performed using the human kinase domain–focused CRISPR knockout (KO) library¹⁰ (Addgene 117725; a gift from Christopher Vakoc) and 2 t(8;21) AML cell lines. Kasumi-1 and SKNO-1 were transduced with LentiV_Cas9_Puro to generate Cas9-expressing cell lines. Cells equal to 300 to 500 times in library size (3051 single-guide RNAs [sgRNAs]) were transduced with the pooled, green fluorescent protein (GFP)-tagged library sgRNAs at low multiplicity of infection (MOI) (∼0.3) to ensure singular transductions and sufficient sgRNA coverage. GFP-positive cells were isolated by fluorescence-activated cell sorting (FACS). Genomic DNA was harvested at initial (3 days after transduction) and final (13 doublings) time points using Qiagen DNeasy Blood and Tissue Kit. Cell lines were screened in duplicate. Sequencing libraries were constructed by 2-step polymerase chain reaction amplification, barcoded with Illumina Nextera XT V2 primers, pooled, and sequenced by Illumina HiSeq 4000. Raw sequencing reads were trimmed and processed using Model-based Analysis of Genome-wide CRISPR/Cas9 Knockout (MAGeCK) analysis pipelines.

To identify new and therapeutically viable kinase targets for AML, we used a GFP-tagged, domain-targeted sgRNA library consisting of 3051 sgRNAs targeting 482 kinases (Figure 1A). We prioritized kinases with a minimum average log₂(fold change) of −1.0 and top 10 ranking by CRISPR score in each of 2 AML cell lines (Figure 1B; supplemental Table 1). Seven kinases were shared between the most critical dependencies (supplemental Figure 1A). Encouragingly, several established AML kinase dependencies were contained within this list.^5,23 To explore biological pathways enriched with AML kinase dependencies, we performed preranked gene set enrichment analysis (Figure 1C). Cell cycle, TP53 regulation, RNA metabolism, and progenitor status gene sets were most enriched with kinase dependencies, highlighting recurrent vulnerabilities in AML. We decided to focus our subsequent experiments on HASPIN, which scored highly in both cell lines. Five unique, HASPIN-targeting sgRNAs had significant log₂-fold depletions in both cell lines relative to known dependencies and positive controls (Figure 1D). Finally, to validate our CRISPR screen findings, we performed a competitive growth assay^10,22 (Figure 1E). HASPIN targeting sgRNAs had reduced growth rates relative to negative controls confirming the HASPIN dependency observed initially.

HASPIN is known as a mitotic protein kinase responsible for ensuring correct spindle fiber attachment to the centromere during metaphase.^24,25 HASPIN is a dependency in several solid tumors,^26-28 but it has not been evaluated in leukemia. Nevertheless, HASPIN has the highest median expression in hematopoietic cell lines (supplemental Figure 1B), supporting its potential functional relevance in blood cancers. Moreover, Haspin KO mice are both fertile and viable,^29,30 suggesting the status of HASPIN as a genetic dependency is cancer specific.

To assess the functional role of HASPIN in AML, we transduced Kasumi-1 cells with 2 unique, HASPIN-targeted short hairpin RNAs (shRNAs) and confirmed protein depletion by western blot (Figure 2A). HASPIN loss resulted in a consistent 50% reduction in Kasumi-1 proliferation rate (Figure 2B). To determine the cause of the reduction in growth, we performed cell cycle analysis. HASPIN depletion caused a notable cell cycle arrest into the G0 to G1 phase with corresponding reductions in the S and G2 to M phases (Figure 2C). HASPIN inhibition has also been associated with accumulation of S and G2/M phase cells^24,25 and cell death.³¹ To determine whether apoptosis contributed to proliferation reductions, we performed Annexin-V/7-AAD staining. HASPIN depletion produced a moderate increase in apoptosis, suggesting most observed growth reductions were due to cell cycle arrest or checkpoint delays (supplemental Figure 2A).

To identify molecular changes associated with HASPIN depletion, we performed RNA sequencing on Kasumi-1 cells expressing either HASPIN targeting or control shRNAs and differential gene expression analysis (supplemental Figure 2B). We identified 493 significantly differentially expressed genes (302 upregulated, 191 downregulated; Figure 2D; supplemental Table 2). Gene set enrichment analysis revealed significant negative enrichments of gene sets associated with cell cycle regulation, stemness, and cellular proliferation after HASPIN knockdown (Figure 2E; supplemental Figure 2C). Conversely, we found significant positive enrichment of gene sets associated with TP53, differentiation, and other tumor-suppressive transcription factors (supplemental Figure 2C). Accordingly, in a hematopoietic cell-type enrichment analysis, significantly downregulated genes were strongly associated with progenitor cell types (hematopoietic progenitor cell (HPC), common myeloid progenitor (CMP), megakaryocyte-erythroid progenitor (MEP), granulocyte-monocyte progenitor (GMP)), whereas upregulated genes were associated with more differentiated myeloid cell types (monocytic, myelocytic; supplemental Figure 2D). Therefore, HASPIN seems to regulate transcriptional programs critical for leukemia survival either directly or as a byproduct of other mechanisms.

HASPIN has been reported to phosphorylate histone H3 at threonine 3 (H3T3) to regulate mitosis in nonhematopoietic cells.^24,32-34 We analyzed H3T3 phosphorylation in AML cells and observed moderate reductions on HASPIN depletion (supplemental Figure 2E). This result supports the role of HASPIN in mitotic regulation in AML. Interestingly, HASPIN is also active during interphase,²⁵ indicating its possible nonmitotic roles. We, therefore, explored whether HASPIN phosphorylates other key cellular regulators by integrating mass spectrometry analyses of peptides differentially phosphorylated on semispecific HASPIN inhibition reported by Maiolica et al³⁵ in 2014 with an atlas of serine/threonine kinase substrate specificities published by Johnson et al³⁶ in 2023. Of the 4095 differential phosphosites affected by HASPIN inhibition, 3845 residues mapped onto the substrate atlas (Figure 3A). Matched residues were assigned kinase percentile scores from the atlas, which represent their specificity as a given kinase’s substrate relative to all other residues in the atlas. A total of 583 matched residues scored as 70th to 90th percentile HASPIN substrates (Figure 3B). Next, we performed gene ontology analysis on proteins in the top 3 deciles of HASPIN substrates to ascertain pathways potentially regulated by HASPIN kinase activity (Figure 3C). RNA splicing was uniquely enriched with 90th percentile substrates. Several serine/Arginine-rich splicing factor (SRSF) family SR proteins, spliceosome components, and other notable RNA-binding proteins (RBPs) were among the top splicing-related substrate proteins, supporting the possible involvement of HASPIN in splicing.

To validate our findings, we probed for changes in phosphorylated and total SR protein levels on HASPIN depletion (Figure 3D). We observed significant reductions in phospho–SR protein levels in predicted HASPIN kinase substrates such as SRSF2, SRSF5, and SRSF6 without significant change in corresponding total protein levels. HASPIN knockdown did not cause complete loss of phospho–SR protein signal because other splicing regulatory kinases are likely able to cross-regulate splicing factor substrates.^37,38 To investigate this possibility, we calculated what fraction of predicted, RNA splicing–enriched HASPIN substrate phosphosites (Figure 3C) were also high confidence substrates for known splicing regulatory kinases (Figure 3E). Half of HASPIN-specific substrates were shared with other splicing regulatory kinases, suggesting that HASPIN potentially cross-regulates target RBPs with other kinases and functions similarly. Next, we investigated how mechanistically HASPIN phosphorylates splicing factors. We compared our inferred 90th percentile HASPIN substrates with HASPIN protein-protein interaction data from published studies³⁵ and the BIOGRID database. We observed minor overlaps between each set (Figure 3F), suggesting that HASPIN may not interact with these substrates directly and may require adaptor or scaffold proteins to coordinate modifications. Accordingly, STRING protein-protein interaction analysis returned a robust network between HASPIN interactors and inferred substrates (Figure 3G), supporting the adaptor-dependent model for the HASPIN kinase functionality.³⁵ Cluster analysis identified 3 major clusters of proteins enriched in cell cycle, RNA splicing, and transcription with HASPIN substrates present in each. Altogether, our results describe HASPIN as a novel splicing regulatory kinase that may phosphorylate substrates while associated with protein complexes.

The reported HASPIN phosphorylation sites are mainly localized to the regulatory serine/arginine-rich (RS) domains of these SR protein splicing factors.^35,36 Phosphorylation at these RS domains is critical for regulating SR protein splicing activity.^39-46 We, therefore, investigated how HASPIN depletion affects RNA splicing in AML. To begin, we analyzed global mRNA splicing changes using replicate multivariate analysis of transcript sequences after shRNA-mediated HASPIN depletion in Kasumi-1 cells. The following are the 5 types of splicing events reported by the replicate multivariate analysis of transcript sequences: alternative 3′ and 5′ splice site, mutually exclusive exons, retained intron, and skipped exon. We detected the following significant splicing events: 285 alternative 3′ splice sites, 228 alternative 5′ splice sites, 963 mutually exclusive exons, 401 retained intron, and 1859 skipped exons (Figure 4A; supplemental Figure 3A). Differentially spliced genes were enriched in DNA repair, mRNA metabolism, and cell cycle pathways (Figure 4B). Furthermore, differentially spliced genes had low overlap with differentially expressed genes after HASPIN knockdown, suggesting 2 distinct functional effects resulting from protein depletion (Figure 4C). To evaluate the functional consequences of HASPIN-related splicing dysregulation, transcript isoforms were matched to control or knockdown conditions and assigned functional biotypes consisting of protein coding, noncoding, or unannotated (Figure 4D). HASPIN depletion caused reciprocal biotype changes, where protein coding becomes noncoding or unannotated and vice versa, and preserved biotypes, suggesting broad alterations in coding potential and functional transcript diversity. Transcripts also switched into nonsense-mediated decay and vice versa, but these constituted fewer events overall (supplemental Figure 3B). We next identified specific examples of transcripts undergoing functional alterations in leukemia-dependent pathways (Figure 4E; supplemental Figure 3C). BCL2-like 11 (BCL2L11, BIM) is a BH3-only proapoptotic protein. On HASPIN depletion, we observed a switch from a nonsense-mediated decay BCL2L11 transcript to BIM_L, an isoform more active than the canonical BIM_EL isoform,^47,48 through greater exon 2C inclusion indicating potentially increased apoptotic responses. Reverse transcription polymerase chain reaction analysis validated this increased exon inclusion in BCL2L11 transcripts after HASPIN depletion (Figure 4E).

RBP activity is modulated by regulatory kinases that influence localization, splicing site selection, and kinetics.³⁸ To determine whether HASPIN depletion functionally affects RBP splicing behavior, we performed RBP motif enrichment analysis using RNA Map Analysis and Plotting Server 2 (rMAPS2). Motifs for major splicing factors such as SRSF1 and heterogeneous nuclear ribonulcear protein C (HNRNPC) were strongly enriched throughout differentially spliced events, suggesting deviation from normal behavior (Figure 4F). In addition, RBP motifs for several major splicing factor protein families including potential HASPIN substrates SRSF2, SRSF5, and SRSF6 were significantly enriched in dysregulated events, recapitulating our integrated proteomics analysis (supplemental Figure 3D). Together, our results demonstrate that HASPIN depletion dysregulates splicing in AML-critical pathways potentially through alteration of RBP behavior.

AML is a heterogeneous disease defined by a variety of initiating and supporting chromosomal aberrations and mutations. We initially discovered HASPIN dependency using 2 t(8;21) AML cell lines. To determine whether HASPIN dependency generalizes to other AML subtypes, we queried published CRISPR screens performed in AML cell lines.^10,21 HASPIN is a significant dependency in several non-t(8;21) AML cell lines, such as THP-1, OCI-AML3, and MOLM-13, which we independently replicated by competitive growth assay (Figure 5A-B; supplemental Figure 4A). Next, we analyzed HASPIN in AML patient data sets to evaluate its clinical relevance. In both The Cancer Genome Atlas Acute Myeloid Leukemia (TCGA-LAML) (Figure 5C) and BEAT-AML (Figure 5D) data sets, patients with KMT2A lysine methyltransferase (MLL) structural variants or t(8;21) translocations had the highest median HASPIN expression. Furthermore, higher HASPIN expression correlated with worse overall survival in patients with AML (Figure 5E). Multivariate survival analysis of clinical and genetic covariates in high or low HASPIN-expressing patients with AML suggested that advanced diagnosis age, “poor” genetic risk group, and oncogenic mutations increase risk (Figure 5F). However, these covariates did not significantly coincide with either patient cohort (supplemental Figure 4B), suggesting HASPIN expression was the main prognostic factor for survival in our analysis. Together, these findings suggest that HASPIN may be a clinically significant dependency in several non-t(8;21) AML cell types.

To evaluate the therapeutic potential of HASPIN dependency in leukemia, we performed several in vitro experiments using the specific HASPIN kinase inhibitor, CHR-6494.³¹ First, we generated CHR-6494 dose-response curves for Kasumi-1 AML cells and healthy human CD34⁺ progenitor cells derived from patients (Figure 6A). Kasumi-1 AML cells were ∼15-fold more sensitive to HASPIN inhibitor than healthy CD34⁺ cells by relative CHR-6494 50% inhibitory concentration (IC₅₀) values. To determine whether other AML subtypes were susceptible to CHR-6494, we tested a panel of different leukemia cell lines (Figure 6B). Relative to the 750 nM CHR-6494 IC₅₀ of healthy CD34⁺ cells, we observed 200 to 600 nM CHR-6494 IC₅₀s in KG1-a, OCI-AML3, THP-1, NB4, K562, and HL-60. CHR-6494 IC₅₀s in Kasumi-1, SKNO-1, MOLM-13, and MV4-11 were 10 to 40 nM, which is substantially lower than other leukemia cell lines and healthy CD34⁺ cells. Last, we reveal that CHR-6494 reduced splicing factor phosphorylation in AML cells (supplemental Figure 5A-B), suggesting that part of the inhibitor effect is related to the role of HASPIN in splicing. These data demonstrate HASPIN inhibition to be effective in a variety of AML subtypes with especially high efficacy in MLL-translocated and t(8;21) AMLs.

Next, we wondered whether CHR-6494 could be used in novel combination therapies for AML. We chose to combine CHR-6494 with BCL-2 inhibitor venetoclax as it is US Food and Drug Administration approved and used in combination with current chemotherapeutics. Furthermore, in a published drug-perturbation CRISPR screen, HASPIN KO AML cells experienced dramatic reductions in growth when also treated with venetoclax (Figure 6C).⁵⁰ To test CHR-6494 and venetoclax in combination, we treated AML cells with dose-response matrices and calculated several synergy scores for each. In Kasumi-1, CHR-6494 and venetoclax dramatically potentiated each other’s effects with maximal zero interaction potency synergy scores at 10 to 100 nM and 3 to 300 nM, respectively (Figure 6D; supplemental Figure 6C). The combination yielded similarly synergistic responses in HL-60 and NB4 and additive responses in MOLM-13 and MV4-11 (supplemental Figure 6A-C). Venetoclax resistance is a significant complication in single-agent venetoclax strategies or in certain de novo primary AML cases.⁵¹ Therefore, we next tested whether our combination therapy could be used in venetoclax-resistant AML. THP-1 and OCI-AML3 were chosen to model venetoclax-resistant AML^52,53 and exhibited venetoclax IC₅₀s upward of 5 μM (supplemental Figure 6D). Surprisingly, when treated with CHR-6494 and venetoclax together, both OCI-AML3 and THP-1 were discovered to be effectively resensitized to venetoclax (Figure 6E-F; supplemental Figure 6C). In THP-1 cells, 100 to 300 nM of CHR-6494 was sufficient to reach 50% inhibition with 10 nM of venetoclax.

To extend our findings to the patient context, we treated several different primary AML samples (supplemental Table 3) with CHR-6494 alone and in combination with venetoclax. CHR-6494 alone reduced viability of 4 of 5 primary AMLs with IC₅₀ values between 200 and 600 nM (supplemental Figure 7A). To evaluate drug synergy, we focused on PT 1 which was resistant to either single-agent treatment (supplemental Figure 7A-B). Excitingly, CHR-6494 and venetoclax potentiated strong synergistic responses from PT 1 and overcame observed resistances (supplemental Figure 7C-D). Combination treatments yielded mainly additive responses in other primary AMLs (data not revealed). Together, our data emphasize the potential clinical utility of HASPIN inhibition for several AML subtypes alone and in a novel combination with BCL-2 inhibition.

In this report, by CRISPR screen, HASPIN was identified as a critical AML dependency where depletion restricted growth and dysregulated transcription and cell cycle. Furthermore, by integrated data mining and further experimental studies, RNA splicing factors were demonstrated as HASPIN kinase substrates. Accordingly, splicing was broadly affected by HASPIN disruption. HASPIN inhibitor treatments were effective and synergized with BCL-2 inhibitors in tested AML cells. These findings establish HASPIN as a valuable dependency in AML, highlight one of its understudied molecular functions, and provide a foundation for continued design of potentially useful therapeutic inhibition strategies.

HASPIN is known as a critical mitotic kinase responsible for recruiting the chromosomal passenger complex to centromeres through phosphorylated histone 3 threonine 3-survivin interactions.^24,25 Accordingly, in AML, HASPIN knockdown reduced H3T3 phosphorylation. The consequences of losing this mitotic modification likely contribute to the functional HASPIN dependency in AML. However, HASPIN is active during interphase,²⁵ suggesting additional functionality to explore. Using an atlas of serine/threonine kinase substrate specificities,³⁶ we disambiguated potential off-target effects in the data set of Maiolica et al³⁵ in 2014 and revealed splicing factors such as HNRNPs, RNA-binding motif proteins (RBMs), and, most significantly, SRSFs to be high confidence HASPIN kinase substrates. Both HASPIN-targeting shRNAs and inhibitor induced significant reductions in target SR protein phosphorylation, further supporting this relationship. Structural analyses of HASPIN substrate recognition regions suggest that it prioritizes basic regions comparable to lysine-rich histone tails,^54,55 potentially explaining preferences for RBPs with similarly basic arginine-rich regulatory domains. However, how HASPIN phosphorylates RBP substrates is still unclear. Maiolica et al³⁵ initially proposed that HASPIN relies on interactor-substrate complexes to mediate phosphorylation. Using additional substrates and the BIOGRID database, we identified a larger HASPIN interactor-substrate network that supports this adaptor-dependent model. Furthermore, CLK1 and CLK2 are known SR protein kinases. They are also potential HASPIN substrates, making indirect RBP regulation through HASPIN phosphorylating other splicing regulatory kinases, another plausible mechanism. In addition, SR protein phosphorylation can be affected by different stages of the cell cycle,^45,56,57 and splicing factor activity is critical for cell cycle regulation.^58-60 Because HASPIN depletion induces a cell cycle arrest and causes differential splicing factor phosphorylation, possibly altering behavior, functional crossover may be occurring where both effects feed into each other and contribute to the overall dependency in AML. Further study will necessitate in-depth, specific proteomic analysis of HASPIN and putative splicing-related substrate proteins to characterize these pathways.

RNA splicing is frequently dysregulated in hematological malignancies.⁶¹ Mutant RBPs or aberrant regulatory kinase activity establish tumorigenic splicing profiles,^62-64 such as splicing factor RBM10 regulating inhibitor of apoptosis protein isoforms in AML to repress proapoptotic responses.²⁰ Because HASPIN likely phosphorylates RBPs and is an AML dependency, we hypothesized that, in addition to mitosis, the role of HASPIN in sustaining AML is regulating splicing factor activity. HASPIN knockdown induced significant differential splicing in transcripts related to DNA repair, cell cycle regulation, and mRNA metabolism and broadly altered transcript coding potentials with predicted functional consequences. Hematopoietic factors implicated in AML with functionally critical splice isoforms, such as PML,^65,66 were significantly affected. Splicing factor kinase inhibition alters substrate RBP phosphorylation, which represses AML through subsequent splicing changes.^20,64 RBP motif enrichment analysis suggests that HASPIN loss affects RBP activity in AML. Importantly, motifs for putative HASPIN substrates identified in our proteomic analysis, such as SRSF family proteins, were most significantly affected, supporting this model. Together, these results describe an alternative role for HASPIN as a splicing regulator and support further efforts to characterize molecular mechanisms underpinning its role in AML.

Currently, several HASPIN inhibitors exist, including 5-iodotubercidin and CHR-6494.³¹ Until recently,^67,68 none have been tested in leukemia. Our study serves as an exploration of the efficacy and potential therapeutic utility of HASPIN inhibition in AML. Although novel HASPIN inhibitors are being developed,⁶⁹ we chose CHR-6494 for its increased specificity and successful use in solid tumors. In Kasumi-1, CHR-6494 treatment achieved 50% reduction in cell viability with ∼15-fold less drug compared with healthy CD34⁺ HSPCs, which support findings that HASPIN is dispensable for normal development and function.^29,30 Except for U937, tested AML cell lines were more sensitive to CHR-6494 than healthy CD34⁺ cells, but t(8;21) (Kasumi-1, SKNO-1) and MLL-translocation (MOLM-13, MV4-11) AMLs were exceptionally responsive. Interestingly, patients with TCGA-LAML and BEAT-AML with these translocations had the highest average HASPIN expression levels, suggesting subtype-specific mechanisms that are more HASPIN reliant than others, but how this might occur is currently unclear. Future studies should consider how primary translocations or mutations in AMLs interact with HASPIN and consequent dependency. Nevertheless, patients with AML with t(8;21) and MLL translocation represent potentially ideal recipients for HASPIN inhibitor treatment.

Venetoclax is a US Food and Drug Administration–approved BCL-2 inhibitor to treat AML.¹⁷ Despite this, patients with AML can acquire venetoclax resistance causing disease relapse.⁷⁰ Analysis of venetoclax-treated CRISPR screens suggested that HASPIN KO contributes to a synthetic lethal relationship in AML. Indeed, in tested AMLs, CHR-6494 and venetoclax increased treatment potency 5- to 10-fold over individual agent treatments when combined. Remarkably, this extended to venetoclax-resistant THP-1 and OCI-AML3, wherein CHR-6494 was sufficient to re-establish venetoclax sensitivity. The opposite was also true, such as in HL-60, where CHR-6494 was significantly potentiated by venetoclax. Importantly, CHR-6494 reduced cell viability in several primary AML samples. Furthermore, combination with venetoclax resensitized resistant primary AML, providing strong preclinical support for the potential utility of HASPIN. In AML, venetoclax resistance hinges on utilization of BCL-2 alternatives, such as MCL-1, metabolic alterations, or TP53 mutation status.⁵¹ Kasumi-1, THP-1, HL-60, and NB4 possess TP53 mutations leading to loss of function. However, all 4 responded robustly to the combination treatment. Mechanisms underlying the combined effect of venetoclax and CHR-6494 are unclear. CRISPR screens in venetoclax-resistant AMLs identified RBPs as major mediators of drug resistance.²⁰ In our analysis, BH3-only factors such as BCL2L11 undergo isoform alteration into more active proapoptotic forms after loss of HASPIN and possible accompanying RBP activity. Other BH3-only factors such as PUMA were also upregulated on HASPIN knockdown. Moreover, DNA damage response and cell cycle genes upstream of apoptosis were upregulated despite nonfunctional TP53, suggesting HASPIN may activate TP53-independent pathways to mediate venetoclax response.

In conclusion, we identified HASPIN as a promising, novel kinase dependency in AML, expanded on the underappreciated role of HASPIN in splicing and demonstrated therapeutic efficacy of HASPIN inhibition across several AML subtypes alone or in combination with BCL-2 inhibitors.

The authors thank Christopher Vakoc for invaluable advice regarding CRISPR screening. The authors thank the University of California, San Diego Moores Cancer Center Flow Cytometry Core Facility staff for assistance with FACS analysis and the Institute for Genomic Medicine core facility staff for assistance with quality control and associated next-generation sequencing of CRISPR screen library preparations. The visual abstract was created in part with BioRender.com.

This work was supported by the National Institutes of Health (NIH), National Cancer Institute (NCI) Grant R01CA104509 (D.-E.Z.) and the NIH/NCI T32 Contemporary Approaches to Cancer Signaling and Communication Training Grant CA009523 (M.L.). The Flow Cytometry Core Facility is partially supported by NIH/NCI P30CA023100.

Contribution: M.L., S.A.S., and D.-E.Z. conceived the project; M.L. performed the research, experiments, and analysis, wrote all necessary code, performed all informatics analysis, and prepared and edited the manuscript; M.Y. and A.G.D. performed wet laboratory experiments; E.D.B. provided patient-derived primary acute myeloid leukemia samples and technical assistance; and D.-E.Z. oversaw the study and the preparation of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Dong-Er Zhang, University of California San Diego, 9500 Gilman Dr, La Jolla, CA 92037; email: d7zhang@ucsd.edu.