There is an active quest for biomarkers of high-risk disease in CML. ASXL1 and other cancer gene mutations, some additional cytogenetic alterations and Philadelphia-associated rearrangements (collectively, additional genetic abnormalities or AGAs) appear to have a negative impact on response and survival outcomes, but they are likely to mirror a more genetically unstable disease. SETD2 is a histone methyltransferase responsible for the deposition of the trimethyl mark at histone H3 lysine 36 (H3K36me3), a key epigenetic modification implicated in transcriptional elongation, chromatin architecture, DNA damage repair. SETD2 loss-of-function (LOF) due to inactivating mutations or, more frequently, accelerated proteasomal degradation, has recently been reported in patients (pts) with multi-TKI-resistant chronic phase (CP) or with blast phase (BP) CML, but it can be observed as early as at diagnosis, particularly in CD34+ hematopoietic progenitors. In this study, we aimed to dissect the impact of SETD2 LOF in primary CD34+ progenitors and in CML cellular models where SETD2 was alternatively silenced or overexpressed, using an integrated approach combining liquid chromatography-tandem mass spectrometry (LC-MS/MS), RNA sequencing (RNA-seq), chromatin immunoprecipitation sequencing (ChIP-seq) and single nucleotide polymorphism arrays (SNP-arrays) for discovery, and Western blotting (WB), immunofluorescence (IF), and co-immunoprecipitation (co-IP) for validation. In SETD2-deficient primary CD34+ progenitors from newly diagnosed CP CML pts, nucleofection of a SETD2-expressing construct reduced clonogenic potential of >50%, indicating that SETD2 LOF enhances leukemic cell propagation. Differential transcriptomic profiling in cell line models revealed SETD2-dependent transcriptional regulation of genes involved in DNA repair (MSH2, MSH6), cell cycle control (CDK1), and metabolic homeostasis (PFKP, LDHA, PDK1). Differential interactome profiling by LC-MS/MS identified SETD2 interactions with proteins critically involved in mismatch repair (MSH2, MSH6), cell division (α-/β-tubulin), and glycolysis (PFKP, PFKFB3, PD, LDHA). Notably, SETD2 was also found to interact with key kinases regulating proliferation and stress response, including ERK1/2 and p38 MAPK. All MS-identified interactions were experimentally validated by IF and co-IP in nuclear, cytoplasmic, or cytoskeletal fractions. Furthermore, IF imaging demonstrated the nuclear colocalization of SETD2 with γ-H2AX foci upon hydrogen peroxide and UV-induced genotoxic stress and its recruitment to DNA damage sites, where it spatially overlapped with MSH2/MSH6 complexes. Integration of SNP-array analysis after chronic exposure to DNA damaging agents with ChIP-seq-based genomic mapping of H3K36me3 showed enrichment of breakpoints at SETD2 target sites (SETD2 knocked-down cell line: 29/45 vs 4/45 regions enriched in genomic breakpoints had loss vs gain of H3K36me3, respectively [p=0.0002]; SETD2-deficient cell line: 25/33 vs 1/33 regions enriched in genomic breakpoints had H3K36me3 loss vs gain, respectively [p=0.009]). Notably, we uncovered a novel role for SETD2 LOF in rewiring cellular metabolism, since SETD2 re-expression attenuated the glycolytic shift observed in SETD2-deficient cells, as evidenced by downregulation of glycolytic enzymes and mitochondrial oxidative phosphorylation complexes used by SETD2-deficient cells as compensation of TCA down-regulation. This was functionally validated in both total lysates and isolated mitochondrial fractions. In contrast, SETD2-deficient cells displayed hyperactivation of hypoxia-associated pathways, consistent with pseudohypoxic reprogramming. Finally, SETD2/H3K36me3 deficiency as assessed by a simple WB assay in total leukocytes could be detected in CP CML pts with AGAs at diagnosis, and, importantly, could discriminate non-optimal vs optimal responders to subsequent imatinib therapy (tx). Our findings point to SETD2 LOF as a key cooperating event in CML, that may act since diagnosis to set the stage for TKI resistance and disease acceleration by i) sustaining BCR::ABL1-independent genomic instability that fuels acquisition of AGAs before and despite TKI tx, and ii) inducing metabolic reprogramming towards glycolysis; ultimately enhancing leukemogenicity of CML progenitors. SETD2 LOF may serve as a biomarker of high-risk disease at diagnosis, and its impact on response to 2nd-gen TKIs or asciminib vs imatinib is worth to be explored further.

Despite the greater biological understanding and the new drugs available, acute myeloid leukaemia (AML) patients who are refractory to intensive induction chemotherapy represents an unmet clinical need, especially in young/fit adults who are eligible for bone marrow transplantation. Since venetoclax/azacitidine (ven/aza) was introduced in AML management in 2020, survival of elderly/unfit patients has dramatically improved, especially in those carrying NPM1 or IDH2 mutations. However, the use of ven/aza in young and fit adults remains limited, raising ongoing debate about its potential role beyond patients ineligible to intensive chemotherapy. Here, we discuss three under 60 years chemorefractory AML patients, who, given the concomitant IDH2 mutations, were started to ven/aza as bridge-to-transplant and successfully treated. These cases confirm the extraordinary sensitivity of IDH2-mutated AML to aza/ven even in the refractoriness setting and show that such less-intensive regimen can be driven by genetics offering a promising alternative to intensive salvage chemotherapy, while preserving patient fitness for allo-transplant.

Acute myeloid leukemia (AML) is an aggressive, molecularly heterogeneous malignancy. Genomic data, including expression profiles, are key for precision medicine, aiding patient stratification and target discovery. Artificial intelligence and machine learning can uncover clinical and biological insights from such datasets but require large, well-annotated data, which are often unavailable. Moreover, technical variability limits RNA-seq data integration across studies.We addressed this by (i) establishing a unified real-world RNA-seq AML cohort and (ii) generating synthetic data to expand sample availability. We analyzed one unpublished and three public RNA-seq datasets, forming a cohort of 918 patients reclassified by ICC/ELN 2022. Standard normalization failed to efficiently integrate datasets.For data harmonization, we applied a novel algorithm (“Gu”) combining eight normalization methods with ComBat batch correction, yielding 27 integrated datasets. To select the best one, we tested various strategies, resulting in a unified real-world cohort of n = 894 clinically and molecularly annotated cases. This preserved biological integrity and reflected ICC 2022 classifications.Pathway analysis revealed broader detection of dysregulated networks, reflecting improved stability from the larger cohort. Using this, synthetic data were generated with the "synthpop" R package, using clinical variables and 66 gene expression features, achieving a 300 × augmentation to n = 2682 synthetic samples. Gene expression networks and ICC-based clustering were preserved, with improved subgroup separation. Both real and synthetic datasets replicated ELN risk stratifications.To our knowledge, our integrated, fully-annotated real-world dataset (n = 894) is the largest in AML. Synthetic data expanded this to n = 2682 samples, supporting machine learning, especially for rare but clinically relevant subgroups.

The SETD2 tumour suppressor encodes a histone methyltransferase that specifically trimethylates histone H3 on lysine 36 (H3K36me3), a key histone mark implicated in the maintenance of genomic integrity among other functions. We found that SETD2 protein deficiency, mirrored by H3K36me3 deficiency, is a nearly universal event in advanced-phase chronic myeloid leukemia (CML) patients. Similarly, K562 and KCL22 cell lines exhibited markedly reduced or undetectable SETD2/H3K36me3 levels, respectively. This resulted from altered SETD2 protein turnover rather than mutations or transcriptional downregulation, and proteasome inhibition led to the accumulation of hyper-ubiquitinated SETD2 and to H3K36me3 rescue suggesting that a functional SETD2 protein is produced but abnormally degraded. We demonstrated that phosphorylation by Aurora-A kinase and ubiquitination by MDM2 plays a key role in the proteasome-mediated degradation of SETD2. Moreover, we found that SETD2 and H3K36me3 loss impinges on the activation and proficiency of homologous recombination and mismatch repair. Finally, we showed that proteasome and Aurora-A kinase inhibitors, acting via SETD2/H3K36me3 rescue, are effective in inducing apoptosis and reducing clonogenic growth in cell lines and primary cells from advanced-phase patients. Taken together, our results point to SETD2/H3K36me3 deficiency as a mechanism, already identified by our group in systemic mastocytosis, that is reversible, druggable, and BCR::ABL1-independent, able to cooperate with BCR::ABL1 in driving genetic instability in CML. KEY POINTS: Virtually all CML patients in blast crisis display SETD2 loss of function. SETD2 loss seems to be accomplished at the posttranslational level rather than being the result of genetic/genomic hits or transcriptional repression. Phosphorylation by Aurora kinase A and ubiquitination by MDM2 contribute to SETD2 proteasome-mediated degradation in blast crisis CML patients. Loss of SETD2 results in increased DNA damage.

The SETD2 tumour suppressor encodes a histone methyltransferase that specifically trimethylates histone H3 on lysine 36 (H3K36me3), a key histone mark implicated in the maintenance of genomic integrity among other functions. We found that SETD2 protein deficiency, mirrored by H3K36me3 deficiency, is a nearly universal event in advanced-phase chronic myeloid leukemia (CML) patients. Similarly, K562 and KCL22 cell lines exhibited markedly reduced or undetectable SETD2/H3K36me3 levels, respectively. This resulted from altered SETD2 protein turnover rather than mutations or transcriptional downregulation, and proteasome inhibition led to the accumulation of hyper-ubiquitinated SETD2 and to H3K36me3 rescue suggesting that a functional SETD2 protein is produced but abnormally degraded. We demonstrated that phosphorylation by Aurora-A kinase and ubiquitination by MDM2 plays a key role in the proteasome-mediated degradation of SETD2. Moreover, we found that SETD2 and H3K36me3 loss impinges on the activation and proficiency of homologous recombination and mismatch repair. Finally, we showed that proteasome and Aurora-A kinase inhibitors, acting via SETD2/H3K36me3 rescue, are effective in inducing apoptosis and reducing clonogenic growth in cell lines and primary cells from advanced-phase patients. Taken together, our results point to SETD2/H3K36me3 deficiency as a mechanism, already identified by our group in systemic mastocytosis, that is reversible, druggable, and BCR::ABL1-independent, able to cooperate with BCR::ABL1 in driving genetic instability in CML.

In chronic myeloid leukemia (CML) patients treated with tyrosine kinase inhibitors (TKIs), molecular response (MR) milestones have been identified that harbor prognostic significance. Major molecular response (MMR)—defined as 3-log reduction in BCR::ABL1 transcript levels from the standardized baseline on the International Scale (IS)—was the very first of such MR milestones to be recognized: It emerged at the time of the IRIS trial as a “safe haven” protecting from loss of response and progression.1 According to the ELN recommendations, achievement of MMR after 12 months of TKI treatment defines an optimal response whereas failure to achieve MMR or loss of MMR represents a warning, requiring the therapeutic strategy to be carefully evaluated for continuation or change.2 Emerging TKI-resistant point mutations in the kinase domain (KD) of BCR::ABL1 may underlie warning responses in a proportion of patients, tilting the balance toward treatment change. Previous studies have interestingly shown that a peculiar kinetics of BCR::ABL1 transcripts fluctuating around 0.1% IS (“unstable MMR”) may be observed during TKI therapy in some patients and have suggested it may be an early indicator for BCR::ABL1 KD mutation testing.3 However, further data are needed to support inclusion of unstable MMR among the indications for mutation testing in clinical recommendations. The aim of this collaborative EUTOS (European Treatment and Outcome Study for CML) study was to take advantage of next generation sequencing (NGS) to investigate the role of BCR::ABL1 KD mutations in CML patients with unstable MMR on TKI therapy. In order to assess the clinical relevance of low frequency variants and the value of NGS-based analysis in this setting, we established a dedicated bioinformatic pipeline including an in-house developed software using a statistically derived dynamic threshold for mutation calling, which enabled us to maximize discrimination between true variants and sequencing errors. Importantly, we showed that therapy changes in patients with confirmed BCR::ABL1 mutations during unstable MMR resulted in higher probability of deep molecular response (DMR) achievement. A total of 91 CML patients with unstable MMR on TKI therapy were analyzed by NGS, with unstable MMR defined as BCR::ABL1 transcript levels in the range of 0.1% IS ±0.16 (the uncertainty of the measurement) for a minimum of 6 months. Two groups were identified according to molecular response trends (Table S1). Group A (n = 15/91) included patients on TKI therapy (1st line, n = 13; 2nd line, n = 1; 5th line, n = 1) displaying unstable MMR after an initial DMR. In regard to ELN recommendation, all patients were classified as warning.2 Group B (n = 76/91) consisted of patients (1st line, n = 59 [8/59 patients maintained an unstable MMR also during subsequent lines of therapy]; 2nd and/or subsequent line, n = 17) who had achieved no better response than unstable MMR. According to ELN recommendations, these patients have optimal response even when BCR::ABL1 levels fluctuate around the level of 0.1% IS. Detailed characteristics of the patients are summarized in Table S2. Follow-up samples were also analyzed. Altogether, 159 unstable MMR samples were analyzed by NGS (median, two samples per patient; range, 1–11) during the course of unstable MMR period. Sample preparation, BCR::ABL1 transcript quantification, NGS analysis, Limit of Blank (LoB), Limit of Detection (LoD), and Limit of Quantification (LoQ) estimation and statistical analysis are described in detail in Supplementary data including Tables S3–S5, Figures S1–S3. Having optimized a bioinformatic pipeline minimizing NGS errors at low mutation frequencies (Supplementary data; Figure S4), we investigated the rate and kinetics of BCR::ABL1 KD variants in patients with unstable MMR. We chose to perform the analysis in duplicate (starting from reverse transcription of RNA to cDNA) to assess reproducibility of mutation results at lower transcript levels, and we checked for mutation kinetics in follow-up samples. Variant frequencies (VAFs) of variants detected in both duplicates were similar and we provided their mean. We found that underlying BCR::ABL1 KD mutations can be detected both in patients who display unstable MMR after an initial DMR (group A) and in patients with persistent unstable MMR as their best response (group B)—although mutations were more frequently observed in the former than in the latter group. The median VAF in both duplicates was 80% (range 4.6%–100%) in group A. In group B the median of VAF in both duplicates was 9.1% (range 4.4%–30%). Mutations were indeed found in 9/15 (60%) as against 12/76 (16%) patients, respectively (Figure 1); in 7/15 (47%) and 6/76 (8%) patients, the mutations identified could be recognized as poorly sensitive to the administered TKI based on IC50 data or literature reports. However, 3/9 mutations in group A and 7/12 mutations (one at high level) in group B were detected in only one of the two replicates. Except for two mutations (F311L and M244V), all low-level mutations (below 20% VAF) and also one high-level mutation detected in only one replicate were not confirmed in follow-up samples, despite no change in therapy. One possible explanation is that these mutations might have been artifacts. Even when using an optimized pipeline of NGS data analysis, in case of a single assessment some low-level variants may happen to pass the LoQ reaching the statistically significant VAF for calling. In contrast, variants passing this level in both duplicates are highly likely to be true mutations. However, it cannot be excluded that at low BCR::ABL1 transcripts and low mutation frequency, true mutations may happen to be picked in one replicate but not in another. Nevertheless, the fact that they disappeared in subsequent samples indicates that such few mutant molecules would have no clinical relevance anyway.4 To assess the frequency of low-level variants in relation to BCR::ABL1 transcript levels, 499 samples from 151 patients with transcript levels in the range of 0.1%–1% IS and 287 samples from 143 patients with transcript levels >1% IS routinely analyzed by NGS between 2017 and 2021 were compared. In addition, samples from 30 healthy donors were used for estimation of NGS error rates at each nucleotide position of the KD. Frequency of detected clinically relevant mutations did not differ between the samples with BCR::ABL1 transcript level >1% IS and with 0.1%–1% IS. However, low-level variants were significantly more frequent in samples with BCR::ABL1 transcript levels 0.1%–1% IS. The results are described in detail in Supplementary data including Tables S6 and S7. To evaluate the clinical benefit of mutational analysis during unstable MMR, the probability of DMR achievement was compared between patients with unchanged therapy and with therapy change (TKI switch or TKI dose increase) (Figure S5). In group A, the therapy change increased the probability of DMR re-achievement regardless of whether the patients had no BCR::ABL1 mutations (n = 3) or had confirmed mutations (n = 5). In contrast, DMR was not re-achieved in two patients with mutations and no therapy change. A decreased probability of DMR re-achievement was found also for patients without mutations and no therapy change (n = 5). The differences were not significant due to low number of patients. Similar observations were found for group B: a higher probability of DMR achievement was observed in patients who changed the therapy, both in case of confirmed mutations (n = 5) and in case of no mutations (n = 31). The probability of DMR was slightly lower for patient without mutations and without therapy change (n = 38), but was not significantly different compared to patients with therapy change (Figure S5). In conclusion, TKI switch or dose increase in patients with detected mutations during unstable MMR led to higher probability of DMR achievement compared to no therapy change. Based on these results, we propose unstable MMR as a novel trigger for BCR::ABL1 KD mutation analysis. Loss of DMR and BCR::ABL1 increase to the levels of unstable MMR is most probability associated with the development of TKI-resistant mutations; thus, once 0.1% IS BCR::ABL1 transcript level is reached, samples should undergo NGS analysis. For patients whose response to TKI therapy fluctuates around MMR with no further improvement, the likelihood of an underlying mutation is much lower, and NGS analysis might be performed 1–2 times per year during unstable MMR, and promptly in case of BCR::ABL1 transcript increase above the uncertainty of the measurement. The reasons of achievement of unstable MMR as the best response may be more often related to patient compliance, lower doses of TKIs, or clonal hematopoiesis. The loss of deep MR represents a risk factor of BCR::ABL1 resistant mutation development. Patients, whose measurable residual disease fluctuates at levels of MMR, but not deeper, are in lower risk of TKI-resistant mutation acquisition. We suggest the use of our novel bioinformatic tool (freely available at nextdom.uhkt.cz) specifically optimized for BCR::ABL1 KD mutation calling and calibrated on the calculation of error rates all over the KD of the BCR::ABL1 transcript. Based on our observations, mutation testing in unstable MMR samples should be done in duplicate. Moreover, whenever total BCR::ABL1 levels are in the range 0.1%–1% IS, dynamics of low-level mutations (below 20% VAF) should be checked in at least one subsequent sample before treatment change. AB performed mutational analysis, established the bioinformatical pipeline, evaluated data, drafted methodological part of the manuscript, and created Tables and Figures of the paper; SdS and CM performed mutational analysis; VP participated on mutational analysis; PP developed NextDOM; JK participated on mutational analysis; PS converted NextDOM to open access tool and performed statistical analysis; HK evaluated and provided clinical data; DS evaluated and provided clinical data; HZ was responsible for BCR::ABL1 transcript level monitoring; AH supervised the work and the project; SS coordinated the study, interpreted results, and revised the paper; KMP designed the study, interpreted results, and wrote the paper. This work was funded by Project Grant NU21-07-00225 from the Czech Health Research Council, by the European Treatment and Outcome study (EUTOS) for CML, by MH CZ – DRO (IHBT – 00023736), and by Ministerstvo Školství, Mládež a Tělovýchovy (MEYS CZ) (BBMRI.cz no. LM2023033). Computational resources were supplied by the project “e-Infrastruktura CZ” (e-INFRA CZ ID:90140) supported by MEYS CZ. The authors are grateful to the patients for providing their samples for this study. AH, KMP, and SS received support by Novartis through the European Treatment and Outcome Study (EUTOS) for CML. The data that support the findings of this study are available from the corresponding author upon reasonable request. Data S1. Supporting Information FIGURE S1. Scheme of LoQ calculation based on sequenced ABL1 kinase domain of 30 healthy donors. (A) After data normalization the data were divided into individual nucleotide changes and using the Poisson distribution the limits “of specific nucleotide change” were counted; (B) After data normalization the number of reads of each position were processed with Poisson distribution to obtain the limits “of each nucleotide position.” FIGURE S2. The threshold data set cover the range of amino acids A28-D504. The amplicons prepared for sequencing are delineated with full arrow: for the patient sample (PT—upper part) it is 1493 bp, for the healthy donor (HD) it is 1429 bp. The threshold data set was generated for the region 272–1704 (1433 bp), that is, A28-D504 (ref. seq. NM_005157) and is depicted with dashed arrow. FIGURE S3. The limits of quantification of the most frequently detected TKI-resistant mutations (n = 29). Red bars show the most frequently used uniform thresholds levels in published NGS studies—1% and 3%. The T315I mutation is highlighted with red color. FIGURE S4. The NextDOM pipeline. (A) Scheme depicts how the software NextDOM utilizes the data generated in software NextGENe, furthermore the data are normalized and evaluated using either LoD or LoQ data set. As a final outcome, the statistically significant variants are provided (p-value ≤.05). (B) The interphase of NextDOM comprises of windows for paths to files and of instructions for using either LoD or LoQ data set. FIGURE S5. Cumulative achievement of DMR of patients with therapy change and with unchanged therapy in relation of mutation detection during unstable MMR. (A) Analysis of patients from Group A. (B) Analysis of patients from Group B. Blue curve—patients with therapy change during unstable MMR with no mutation detected; dashed blue curve—patients with therapy change during unstable MMR with mutation detected; red curve—patients no therapy changes during unstable MMR with no mutation detected; dashed red—patients no therapy changes during unstable MMR with mutation detected. TABLE S1. Characteristics of CML patients with unMMR during TKI therapy. Group A includes patients who displayed unMMR after an initial DMR. Group B includes patients who never achieved any better response than unMMR. TABLE S2. Patient characteristics. Group A includes patients (n = 15) on TKI therapy who had initially achieved deep molecular response, but BCR::ABL1 transcripts had subsequently increased to the levels of unMMR. Group B consists of patients (n = 76) who never achieved better response to TKIs than unMMR TABLE S3. The LoQ and thresholds of significant mutation detection. TABLE S4. The values of LoD and LoQ for each specific nucleotide change. (A) Thirty samples of healthy donors were sequenced and for each type of nucleotide change the LoDs were calculated. (B) The values of LoQs were set to fivefold the LoDs. TABLE S5. Limits of quantification of each nucleotide position. LoQs were calculated for 1248 positions of the KD, and for each position they were compared to the value of LoQ of the specific nucleotide changes. Then the higher value was chosen for the final threshold data set. The second column shows the total number of each individual nucleotide in the 1248 bp region. The third and the fifth columns show the number of nucleotides with higher LoQ for each nucleotide positions for transitions and for transversion, respectively. TABLE S6. Frequency of low- and high-level variants detected by NGS in samples with BCR::ABL1 below or above 1% IS. TABLE S7. ddASO PCR analysis of low-level variants. The table presents the results of ddASO PCR analysis of five low-level variants, which were detected by NGS only in one duplicate, in five samples. Only in one case the variant was confirmed by ddASO PCR. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Introduction SETD2 is a tumor suppressor that trimethylates histone H3 at Lys36 (H3K36Me3), a key mark for transcription and DNA damage repair (DDR). We previously reported SETD2 nongenomic loss of function in blast crisis (BC)-CML, but not in chronic phase at diagnosis. The aim of this study was to assess in CML models whether the contribution of SETD2 loss to disease progression occurs through the impairment of DDR fidelity and the perturbation of the epigenetic landscape. Methods and Results To investigate the activation and proficiency of homologous recombination (HR) of double-strand breaks (DSBs) and of mismatch repair (MMR), we used chronic exposure to UV rays or a single exposure to hydrogen peroxide. Both induced DNA damage in LAMA 84 cells siRNA-depleted of SETD2. Compared with parental cells, SETD2-silenced cells lost the ability to activate the ATM-dependent repair pathway. ATM inactivation impaired H2AX localization to the sites of DNA damage in vitro and prevented the formation of RAD51 (HR) and MSH2-MSH6 (MMR) repair foci. As a confirmation, forced expression of SETD2 in KCL22 cells was able to restore DDR, as demonstrated by ATM and H2AX phosphorylation and RAD51 and RAD54 expression after either UV or hydrogen peroxide exposure. ChIP-seq showed that forced SETD2 re-expression rescued H3K36Me3 in coding sequences but not promoters, as expected, and identified a series of genes that are targeted by SETD2 trimethylation in CML and may thus be more prone to DSBs and mutations in vivo. In addition, H3K36Me3/RNAPolII colocalization identified genes whose transcription is SETD2/H3K36Me3-dependent (eg, AUNIP, CEP170, PNKP, and KIF22, which are involved in mitosis, cytokinesis, and DNA damage repair). Significant changes in expression of the same genes were detected in our SETD2 on/off models by RNA-seq. Conclusion We have dissected some potential roles for SETD2 loss of function in CML transformation and demonstrated that SETD2/H3K36Me3 deficiency is a novel, BCR::ABL1-independent mechanism of genetic and genomic instability in CML. We have also shown that SETD2/H3K36Me3 loss perturbs the expression of key cellular genes, including known oncogenes and tumor suppressors, and may compromise their sequence integrity after DNA damage. Acknowledgement Supported by AIRC IG2019 grant (23001).

Acute myeloid leukemia (AML) is an aggressive hematologic neoplasia with a complex polyclonal architecture. Among driver lesions, those involving the FLT3 gene represent the most frequent mutations identified at diagnosis. The development of tyrosine kinase inhibitors (TKIs) has improved the clinical outcomes of FLT3-mutated patients (Pt). However, overcoming resistance to these drugs remains a challenge. To unravel the molecular mechanisms underlying therapy resistance and clonal selection, we conducted a longitudinal analysis using a single-cell DNA sequencing approach (MissionBioTapestri® platform, San Francisco, CA, USA) in two patients with FLT3-mutated AML. To this end, samples were collected at the time of diagnosis, during TKI therapy, and at relapse or complete remission. For Pt #1, disease resistance was associated with clonal expansion of minor clones, and 2nd line TKI therapy with gilteritinib provided a proliferative advantage to the clones carrying NRAS and KIT mutations, thereby responsible for relapse. In Pt #2, clonal architecture was less complex, and 1st line TKI therapy with midostaurin was able to eradicate the leukemic clones. Our results corroborate previous findings about clonal selection driven by TKIs, highlighting the importance of a deeper characterization of individual clonal architectures for choosing the best treatment plan for personalized approaches aimed at optimizing outcomes.

The recent introduction of targeted therapies for Systemic Mastocytosis (SM) has raised the awareness about the importance of timely and accurate diagnosis- that, outside reference centers, is hampered by the rarity of this malignancy, by the heterogeneity of clinical symptoms and presentation and by the need to integrate the specific expertise of different healthcare professionals. Molecular testing plays a key role in the diagnosis, classification and management of SM: detection of KIT D816V activating mutation, underlying approx 95% of SM cases, is among the diagnostic criteria and quantitation of the allele burden (AB) helps in subtype classification, provides prognostic information and might serve, in future, for minimal residual disease monitoring. Screening for KIT D816V requires sensitive PCR-based methods, since disease burden, especially in the indolent forms, may be very low. The importance of reliable and accurate molecular testing in SM prompted the Rete Italiana MAstocitosi (RIMA) to undertake, with the sponsorship of the GIMEMA Working Party on Chronic Myeloproliferative Neoplasms, an initiative aimed to i) map the status of KIT D816V mutation testing in Italy; ii) foster the creation of a network of specialized, reference laboratories available to support clinicians in the diagnosis of SM. A survey was conducted among 35 molecular biology labs at major hematological centers all across Italy to assess whether and with which methodology KIT D816V mutation testing was offered. Twenty-nine labs (83%) declared they were routinely performing KIT mutation testing, either by Sanger sequencing (n=4), NGS (n=4), digital PCR (with either home-brew assays or commercial kits; n=17), semi-quantitative real time PCR (either home-brew or commercial; n=2), quantitative ARMS-PCR (home-brew; n=1), or qualitative ASO-PCR (home-brew; n=1). After a pilot experience that involved 7 labs in 2023, a second control round of proficiency testing was conducted among 10 labs (Milan, Monza, Bergamo, Vicenza, Pisa, Pescara, Rome, Bari, Palermo, Nuoro) selected among those who reported to be using a quantitative assay on a digital PCR platform (BioRad QX200 or QX600, n=8; Thermo Fisher QuantStudio Absolute Q, n=1; Qiagen QIAcuity One, n=1). KIT D816V-mutated and wild-type DNAs isolated from the HMC-1.2 and HL-60 cell lines, respectively, were mixed in variable proportions to mimic different ABs (5%; 0.5%; 0.3%; 0.1%; 0.05%; 0.02%; 0%) and used to prepare 10 identical batches of 7 blinded vials. Dilutions were first externally assessed and validated by the UK Wessex Genomics Laboratory Service (WGLS) using an accredited droplet dPCR assay with an LoD=0.01% and then shipped to the participating labs, where they were analyzed according to local procedures. Labs were asked to score each sample as positive or negative for KIT D816V and to provide a quantitative estimate of the AB. Agreement between measurements was assessed using Weighted Deming Linear Regression and Bland-Altman bias analyses. The vial containing no KIT D816V DNA was scored negative by 9/10 labs. The 0.02% vial was scored positive for the D816V by 5 labs, borderline by one lab and negative by 4 labs (in accordance with their reported LoDs of 0.1%, 0.1%, 0.06% and 0.04%). The remaining dilutions were correctly scored positive by all labs. In these samples, quantitation of AB showed excellent correlation and concordance among labs and between labs and the WGLS, with Pearson R correlation coefficients ranging from 0.9997 to 0.9999 and Bland-Altman plots yielding a mean bias ranging from -0.17 to 0.02. Overall, our experience shows that the rarity of the disease and the lack of awareness about the diagnostic challenges in detecting KIT D816V result into routine use of an array of heterogeneous and sometimes inadequately sensitive methods (like Sanger sequencing and NGS, that may yield false-negative results leading to missed/delayed SM diagnosis in a not negligible proportion of patients). Our control round demonstrates that digital PCR, regardless of the platform, provides a reliable and reproducible option for KIT D816V mutation detection and quantitation, although further efforts should be made to enhance LoD in some cases. Our experience also shows that lab networking and exchange of positive controls and expertise plays a critical role in guaranteeing high standards of molecular testing in rare diseases like SM. Supported by Istituto Gentili.

Whether additional mutations in genes other than BCR::ABL1 harbor predictive or prognostic value and whether they should be incorporated in routine diagnostic workups for newly diagnosed CP CML pts is a controversial issue. Most of the currently available data pointing to a potential negative impact of ASXL1 and/or epigenetic genes on response to therapy come from NGS analyses of selected and/or non-homogeneously treated cohorts of pts, with only 2 studies having so far focused on pts enrolled in prospective clinical trials (TIDEL II and TIGER, respectively). We undertook a translational study aimed to investigate the impact of molecular profiling at diagnosis on the likelihood to achieve deep MR and stable TFR in CP CML pts enrolled in the GIMEMA/HOVON CML1415 (‘SUSTRENIM‘), an international prospective clinical trial that randomized pts to receive either NIL or IM with switch to NIL in case of no optimal response (NCT02602314). All the pts achieving ≥MR4 by 36 months (mo) and maintaining it up to 48 mo of therapy qualified for the TFR phase. SUSTRENIM enrolled a total of 448 pts (IM, n=220; NIL, n=228); median follow-up is 56 mo (range, 42-60). One hundred and twenty-four pts who signed the informed consent for participation in this translational study had peripheral blood samples collected at diagnosis and subject to whole exome sequencing (WES). This subset of pts did not differ from the whole population in terms of age, gender, Sokal and ELTS scores, TFR eligibility; 73 of them were randomized to NIL and 67 to IM. Exome capture was performed with the egSEQ Exome Panel (Edinburgh Genetics), followed by 2×150bp paired-end sequencing on a NovaSeq X Plus (Illumina) (48Gb/sample; average sequencing depth, 300X). Raw sequence reads were aligned on the GRCh38 human reference genome and variants called using Dragen v3.9.5 pipeline. Variants labeled as somatic were retained for downstream annotation using VarSeq v2.5.0. For the purpose of the present report, we focused on somatic single nucleotide variants and frameshift indels in ASXL1 and in a comprehensive set of 123 genes compiled using publicly available databases of epigenetic factors and leukemia-associated genes. Strict filtering criteria were adopted, for uniformity with previous NGS studies in CML, to select for variants predicted to be nonsynonymous or to disrupt essential splice donor or acceptor sites and to be either loss of function or damaging according to ≥4 functional prediction tools. Pts with and without mutations were compared using Wilcoxon rank sum test, Fisher's exact test or Pearson's Chi-squared test; cumulative incidences of progression-free survival (PFS) were compared with the Fine and Gray test. ASXL1 mutations were detected in 8/124 (7%) pts. These pts were slightly younger (median, 44 vs 58 years, p=0.059); no correlation was observed with gender, Sokal or ELTS, transcript. Depth of MR at 3, 12, 24, 36 and 48 mo was not significantly different between pts with and without mutations, either stratifying into MR levels at each timepoint or categorizing responses as < or ≥MR3, MR4, M4.5 at a given timepoint (of note: ≥MR3 at 3 mo, 0% vs 13%, p=0.6; ≥MR3 at 12 mo, 80% vs 74%, p>0.9; ≥MR4 at 24 mo, 60% vs 60%, p>0.9; ≥MR4 at 36 mo, 100% vs 75%, p=0.6; ≥MR4 at 48 mo, 100% vs 83%, p>0.9 for pts with and without ASXL1 mutations, respectively). A similar percentage of ASXL1-mutated and -unmutated pts became eligible for TFR as per protocol (43% vs 42%, p>0.9). PFS did not differ between pts with or without ASXL1 mutations (p=0.95). Mutations in epigenetic and leukemia-associated genes were detected in 17/124 (14%) pts. Mutated genes were IKZF1, IDH1, IDH2, DNMT3A, BCOR, BCORL1, NSD2, PHF2, KDM1B. As above, no significant differences were observed between pts with and without mutations in terms of MR depth at different timepoints, TFR eligibility, PFS. In conclusion, presence of an ASXL1 mutation (or of other epigenetic or leukemia-associated gene mutations) did not impact on the depth of MR or on TFR eligibility either in pts treated with NIL or in pts treated with IM with proactive switch to NIL in case of no optimal response. Although algorithms for risk stratification and tailored treatment based on molecular profiling at diagnosis are eagerly awaited, our data warn against the premature incorporation of ASXL1 mutations among high risk features in treatment recommendations, and highlight the need for further studies in prospective series of uniformly treated pts.

Context The FLT3-ITD mutation is among the most recurrent alteration and predicts a dismal prognosis in acute myeloid leukemia (AML). The tyrosine kinase inhibitor midostaurin (Mido) has been approved in association with chemotherapy for newly diagnosed FLT3-mutated AML patients. While Mido demonstrates clinical effectiveness, it falls short in eliminating leukemic stem cells (LSCs), leading to limited responses and eventual resistance development. The hypoxic niche, where LSC physiologically resides, acts as a protective environment, sustaining various pro-survival mechanisms crucial for AML development and treatment response (Bruno et al. Int J Mol Sci. 2021). Objective To investigate the metabolic and transcriptomic changes in FLT3-mutated AML cells treated with Mido under conditions mimicking the hypoxic niche. Our ultimate objective is to identify a novel combination approach to eradicate LSCs. Design MV-4-11 and MOLM-13 cell lines (FLT3-ITD), were treated with 100 nM Mido for 24 h in normoxia and hypoxia (1%O2). Metabolic perturbations were explored by glutamate, lactate and ROS-Glo reagents (Promega), while transcriptomic alterations were analysed by gene expression profiling (GEP, ClariomS, Affymetrix). GSEA was performed to identify deregulated pathways. Pre-clinical efficacy of combination treatment has been proven by monitoring cellular viability and proliferation. Results Mido exhibited antileukemic effectiveness in both normoxic and hypoxic conditions, with MV4-11 being the most sensitive model. As compared with vehicle, Mido triggered a pronounced decline in both extracellular and intracellular glutamate levels, indicating an elevated reliance of treated cells on glutamine metabolism for mitochondrial respiration, irrespective of oxygen availability. Lactate and ROS levels, typically increased by hypoxia, were not affected by Mido. Transcriptomic analysis unveiled that Mido downregulated signatures of the mTORC1 pathway and suppressed the expression of biosynthetic enzymes crucial for synthesizing non-essential amino acids. Integration of metabolic and transcriptomic data highlights the activation of the autophagy network as a potential prosurvival mechanism adopted by FLT3-ITD mutated cells. Consistent with this finding, concurrent treatment with the autophagy inhibitor chloroquine synergistically enhanced the antileukemic effects of Mido, leading to decreased viability and proliferation. Conclusion These results underscore the potential of combining autophagy inhibitors with Mido as a rational strategy to target FLT3-ITD LSCs residing within hypoxic niches.

Background The SETD2 tumor suppressor gene encodes a histone methyltransferase that safeguards transcription fidelity and genomic integrity via trimethylation of histone H3 lysine 36 (H3K36Me3). SETD2 loss of function has been observed in solid and hematologic malignancies. We have recently reported that most patients with advanced systemic mastocytosis (AdvSM) and some with indolent or smoldering SM display H3K36Me3 deficiency as a result of a reversible loss of SETD2 due to reduced protein stability. Methods Experiments were conducted in SETD2-proficient (ROSA KIT D816V ) and -deficient (HMC-1.2) cell lines and in primary cells from patients with various SM subtypes. A short interfering RNA approach was used to silence SETD2 (in ROSA KIT D816V cells), MDM2 and AURKA (in HMC-1.2 cells). Protein expression and post-translational modifications were assessed by WB and immunoblotting. Protein interactions were tested by using co-immunoprecipitation. Apoptotic cell death was evaluated by flow cytometry after annexin V and propidium iodide staining, respectively. Drug cytotoxicity in in vitro experiments was evaluated by clonogenic assays. Results Here, we show that the proteasome inhibitors suppress cell growth and induce apoptosis in neoplastic mast cells by promoting SETD2/H3K36Me3 re-expression. Moreover, we found that Aurora kinase A and MDM2 are implicated in SETD2 loss of function in AdvSM. In line with this observation, direct or indirect targeting of Aurora kinase A with alisertib or volasertib induced reduction of clonogenic potential and apoptosis in human mast cell lines and primary neoplastic cells from patients with AdvSM. Efficacy of Aurora A or proteasome inhibitors was comparable to that of the KIT inhibitor avapritinib. Moreover, combination of alisertib (Aurora A inhibitor) or bortezomib (proteasome inhibitor) with avapritinib allowed to use lower doses of each drug to achieve comparable cytotoxic effects. Conclusions Our mechanistic insights into SETD2 non-genomic loss of function in AdvSM highlight the potential value of novel therapeutic targets and agents for the treatment of patients who fail or do not tolerate midostaurin or avapritinib.

BACKGROUND AND AIMS - Genomic instability is a hallmark of chronic myeloid leukemia (CML) cells since the chronic phase (CP) of the disease, and results in BCR::ABL1 mutations and/or additional genetic and genomic aberrations that may drive resistance to tyrosine kinase inhibitors (TKIs) and progression to blast crisis (BC). Genomic instability is also a feature of CML stem and progenitor cells and may contribute to their persistence. The SETD2 tumor suppressor codes for a protein that trimethylates histone H3 at lysine 36 (H3K36me3). In solid tumors, SETD2 loss of function has been shown to impair H3K36me3-mediated recruitment of DNA damage response components. We have recently reported that non genomic loss of function of SETD2 is a feature of BC CML and results from premature proteasome-mediated degradation of the SETD2 protein triggered by Aurora kinase A phosphorylation and MDM2 ubiquitination. In the present study, we aimed to assess SETD2/H3K36me3 status in CD34+ progenitors of CP CML patients (pts) and whether SETD2/H3K36me3 deficiency may play a role in genomic instability in CML models. METHODS - Western blotting (WB) was used to assess SETD2 protein expression and H3K36me3 as a surrogate marker of SETD2 function in the CD34+ cell fraction isolated from the bone marrow of 20 newly diagnosed CP CML pts and from a pool of healthy donors (HD). SETD2 forced expression in CD34+ progenitors from newly diagnosed CP CML pts and in the SETD2-deficient KCL22 cell line was performed by nucleofection. SETD2 knock-down in the SETD2-proficient LAMA84 cell line was performed by RNAi. Clonogenic capacity was evaluated by clonogenic assays. Chromatin immunoprecipitation sequencing (ChIP-seq) for H3K36me3 was performed on an Illumina HiSeq2000 with a min 50 million 150-bp single-end reads per replicate. High resolution karyotyping wias perfomed with Cytoscan HD arrays. DNA damage and DNA repair activation were assessed in primary samples and cell lines by WB and immunofluorescence (IF) using antibodies specific for phospho-H2AX, mismatch repair (MMR) and homologous recombination repair (HR) proteins. RESULTS - WB demonstrated a marked down-modulation of SETD2 expression, paralleled by H3K36me3 deficiency, in the CD34+ cells of all newly diagnosed CP CML pts as compared to the total mononuclear fraction and to CD34+ cells from HDs. To investigate whether SETD2 loss affects the activation and proficiency of HR and MMR, we used chronic exposure to UV rays or a single exposure to hydrogen peroxide (1mM for 30 and 60 min). Both induced DNA damage in SETD2 siRNA-depleted LAMA84 cells. Compared to parental cells, cells silenced for SETD2 failed to activate the ATM-dependent repair pathway. Moreover, we found that ATM inactivation prevents H2AX foci formation, associated with a loss of RAD51 and RAD54 (HR) and MSH6 (MMR) repair foci. To confirm the hypothesis that SETD2 is a tumor suppressor implicated in maintaining genomic stability in CML, we transfected SETD2-deficient KCL22 cells with an ectopic SETD2 plasmid. SETD2 forced expression was able to restore DNA damage response, as demonstrated by WB and IF detection of ATM, p95 and H2AX phosphorylation, BRCA1, BRCA2 and CtIP expression and finally RAD51 and RAD54 localization on HR repair foci observed after UV or hydrogen peroxide exposure. High-resolution karyotyping after DNA damage showed increased rate of DNA breakpoints in SETD2-deficient cells, preferentially occuring at loci where H3K36me3 marks were lost as assessed by ChiP-seq. In line with the effects observed in cell line models, forced expression of SETD2 in CD34+ cells from 5 CP CML pts was found to restore proliferation control, since a >50% reduction in clonogenic potential was observed after nucleofection. CONCLUSIONS - Our findings demonstrate that SETD2 is a bona fide tumor suppressor in CML progenitors and establish a functional link between SETD2 loss of function and genomic instability. SETD2 inactivation may thus play a pivotal role in the harmful cascade of events that may foster drug resistance and ultimately lead to disease progression. Since SETD2 loss of function in CML is mediated by post-translation mechanisms, hence is reversible, therapeutic strategies aimed at interfering with these mechanisms may restore proliferation control and interrupt this cascade. Supported by AIRC IG 2019 (23001) and by Italian Ministry of Health, “Bando Ricerca Finalizzata 2016”, project GR-2016-02364880.

The introduction of tyrosine kinase inhibitors (TKIs) has changed the treatment paradigm of chronic myeloid leukemia (CML), leading to a dramatic improvement of the outcome of CML patients, who now have a nearly normal life expectancy and, in some selected cases, the possibility of aiming for the more ambitious goal of treatment-free remission (TFR). However, the minority of patients who fail treatment and progress from chronic phase (CP) to accelerated phase (AP) and blast phase (BP) still have a relatively poor prognosis. The identification of predictive elements enabling a prompt recognition of patients at higher risk of progression still remains among the priorities in the field of CML management. Currently, the baseline risk is assessed using simple clinical and hematologic parameters, other than evaluating the presence of additional chromosomal abnormalities (ACAs), especially those at "high-risk". Beyond the onset, a re-evaluation of the risk status is mandatory, monitoring the response to TKI treatment. Moreover, novel critical insights are emerging into the role of genomic factors, present at diagnosis or evolving on therapy. This review presents the current knowledge regarding prognostic factors in CML and their potential role for an improved risk classification and a subsequent enhancement of therapeutic decisions and disease management.

Systemic Mastocytosis (SM) is a rare hematological neoplasm, but its incidence is probably underestimated because of the heterogeneity of clinical symptoms and presentation that may cause diagnostic difficulties especially outside reference centers. Detection of the activating D816V KIT mutation in the bone marrow or peripheral blood is one of the minor criteria for the diagnosis of SM, but it poses some challenges since in the indolent forms (the most frequent) the mast cell burden is very low. For this reason, the National Comprehensive Cancer Network and the European Competence Network on Mastocytosis (ECNM) recommend a high-sensitivity PCR-based assay, such as Allele-Specific Oligonucleotide-Real Time Quantitative PCR (ASO-qPCR) or ddPCR, while Sanger and Next Generation Sequencing are considered not adequate. Accurate quantitation of allele burden (AB) is also desirable since it offers diagnostic (an AB³10% qualifies as a B-finding according to the latest WHO classification) and prognostic information and might be used to monitor response to tyrosine kinase inhibitors. Despite the widespread routine availability of real time and digital PCR technologies, a map of labs across Italy offering KIT D816V testing with adequate sensitivity is lacking, nor have cooperative efforts aimed to check lab performance and reproducibility of diagnostic results ever been performed. Based on these premises, RIMA undertook, with the sponsorship of the GIMEMA Working Party on Chronic Myeloproliferative Neoplasms, a project aimed to: i) build a nationwide network of competent reference laboratories performing KIT D816V mutation testing; ii) promote harmonization of local procedures and ensure adequate proficiency and cross-comparability of results. The pilot phase of this project involved 7 labs (L) spread across the country (Milan, Bologna, Verona, Florence, Rome, Naples, Pavia) serving one or more reference clinical centers for the diagnosis and management of SM. First, a survey aimed to assess the techniques and procedures in use in each participating lab was conducted. Five labs (L2-L6) used a commercial ddPCR assay (KIT p.D816V c.2447A>T, Assay ID dHsaCP2000023; Bio-Rad); one lab (L7) used a home brew ASO-qPCR assay; one lab (L1) used a commercial semi-quantitative real time PCR-based assay (PlentiPlex Mastocytosis D816V kit; Pentabase). All labs performed analyses in duplicate or triplicate using DNA as input material, with amounts variable from 50 to 250ng/replicate. Reported limit of detection (LoD) was 0.0003% for the home brew method and 0.01% for the commercial methods. Subsequently, a round robin test to evaluate and compare performance in terms of sensitivity and inter-lab reproducibility was carried out. Mutated and wild-type DNA isolated from HMC-1.2 and HL-60 cells, respectively, was mixed and diluted to mimic different D816V ABs, from 10% down to 0.01%. Dilutions were externally assessed by the UK Wessex Genomics Laboratory Service, using an accredited ddPCR assay with a limit of detection (LoD) of 0.01%. Of note, the sixth dilution (D6) was scored as slightly below their LoD (0.008%; 3 positive droplets only). Identical batches of blinded vials were then distributed and analyzed in parallel by the 7 participating labs according to their own routine protocols. Positivity/negativity as scored by L1-7 and AB values as scored by L2-7 are detailed in Table 1. All the evaluated methods proved highly accurate in the detection and, for ddPCR and ASO-qPCR, in the quantitation of the KIT D816V (R 2 between 0.988 and 0.998). A very high degree of concordance was achieved, for all dilutions, across different labs and methods (CV between 0.07 and 0.8). D6 was scored borderline by L1 and positive by all other labs. This pilot experience demonstrates that the evaluated PCR-based methods may reliably identify and quantitate the KIT D816V mutation down to an AB of 0.01% with a high degree of cross-comparability of results, thus assisting clinicians in the diagnosis and monitoring of SM patients. To the best of our knowledge, this is the first cooperative effort aimed to assess the performance and reproducibility of KIT D816V mutation testing in SM. Involvement of additional Italian labs is already planned, and further implementation within the ECNM will be proposed. Definition of common SOPs, uniform sample requirements and web-based reporting following the GIMEMA LabNet model will be pursued.

ddPCR proved highly sensitive and accurate. Total hands-on time was approximately 2 hrs, and time from sample to results was 2 days. Therefore, ddPCR may be integrated into diagnostic algorithms of CML (and Ph+ ALL) patients as a convenient first-level screening tool for mutations impacting TKI selection.

Protein kinases (PKs) play crucial roles in cellular proliferation and survival, hence their deregulation is a common event in the pathogenesis of solid and hematologic malignancies. Targeting PKs has been a promising strategy in cancer treatment, and there are now a variety of approved anticancer drugs targeting PKs. However, the phenomenon of resistance remains an obstacle to be addressed and overcoming resistance is a goal to be achieved. Chronic myeloid leukemia (CML) is the first as well as one of the best examples of a cancer that can be targeted by molecular therapy; hence, it can be used as a model disease for other cancers. This review aims to summarize up-to-date knowledge on the main mechanisms implicated in resistance to PK inhibitory therapies and to outline the main strategies that are being explored to overcome resistance. The importance of molecular diagnostics and disease monitoring in counteracting resistance will also be discussed.