Targeted Sequencing Panel Research Articles

Automated Machine Learning-Based Diagnosis and Molecular Characterization of Acute Leukemias using Flow Cytometry Data

Abstract Current flow cytometric analysis of blood and bone marrow samples for the diagnosis of acute leukemias relies heavily on manual intervention in both the processing and analysis steps, introducing significant subjectivity into the resulting diagnosis and increasing diagnostic turn-around time. Additionally, concurrent molecular characterization of these samples via cytogenetics and targeted sequencing panels can take multiple days, thereby delaying patient diagnosis and treatment. Attention-based multi-instance learning models are machine learning models that can make accurate predictions and generate interpretable insights regarding the classification of a sample from multiple events/cells; while these models have been developed for anatomic pathology applications, they have yet to be applied to flow cytometry data. By utilizing 1,820 flow cytometry samples from 2019-2022 at Brigham and Women’s Hospital, we developed attention-based multi-instance machine learning models for automated diagnosis of acute leukemia, including differentiation of acute myeloid leukemia (AML) from B-lymphoblastic leukemia/lymphoma (B-ALL). Additionally, using concurrent cytogenetic and targeted sequencing data from 674 acute leukemia samples, machine learning models for prediction of molecular aberrancies from flow cytometry data were developed. Machine learning models were created using the TabNet deep learning architecture and VIME self-supervised training algorithm, which are state-of-the-art approaches towards machine learning from tabular data including flow cytometry data. Attention-based multi-instance models demonstrated strong performance for the automated diagnosis of acute leukemia versus non-leukemia samples (AUROC 0.869), as well as the separation of AML from B-ALL samples (AUROC 0.971). These models also accurately predicted cytogenetic aberrancies among AML samples, including t(15;17);PML::RARA (AUROC 1.00), as well as mutations including NPM1 (AUROC 0.725). These models do not require any manual intervention including compensation or gating, and additionally provide quantitative scores for the relative importance of different flow cytometry events and markers for the diagnosis of a particular sample. These importance scores can be integrated into flow cytometry analysis software for visualization and interpretation by hematopathologists. In this study, we have demonstrated the capability of machine learning models to provide automated diagnoses of acute leukemia, as well as accurately predict cytogenetic and molecular aberrancies in blood and bone marrow samples using flow cytometry data. This automated workflow can significantly decrease diagnostic turn-around time and ultimately improve patient outcomes.

Clonal Hematopoiesis Associated with Rheumatoid Arthritis

Introduction: Clonal hematopoiesis (CH) becomes common with age and can play a role in the pathophysiology of inflammatory diseases by modifying immune cell function. CH is characterized by the presence of somatic mutations in genes such as DNMT3A and TET2 (CH of indeterminate potential, CHIP), and as mosaic chromosomal alterations (mCAs). While CHIP, particularly in the context of TET2 mutations, can promote inflammation in cardiovascular diseases, the association between clonal hematopoiesis and autoimmunity has not been comprehensively studied to date. Rheumatoid arthritis (RA) is one of the most common autoimmune diseases primarily affecting joints. Seropositive RA is defined by the presence of anti-cyclic citrullinated peptide antibodies (ACPA) and/or rheumatoid factor (RF), whereas seronegative RA is a more heterogeneous inflammatory disease entity involving patients with clinical RA manifestations in the absence of ACPA and RF. Understanding of the association between CH and inflammatory processes may yield novel insights into prevention and management of autoimmune diseases. Methods: To comprehensively evaluate the association between CH and autoimmune diseases, we identified CHIP variants in whole exome sequencing (WES) data from the FINRISK study (N=10 129). We also detected mCAs and CHIP in the nationwide FinnGen study (N=340 173), where we manually identified CHIP carriers as outliers in SNP intensity plots at known CHIP loci in DNA microarray data. Only known CHIP loci with variant prevalence comparable to UK Biobank WES data and significant positive associations with age and hematological malignancies were included for downstream analyses. To further characterize the link between CHIP and RA phenotypes, we performed targeted panel sequencing of 69 myeloid driver genes in 675 RA patient blood samples (91% collected at diagnosis) and 60 healthy controls. We studied the association between CH and RA in multivariable models excluding participants with prevalent hematologic malignancies and adjusting for sex, age, smoking, and principal components of ancestry. We performed uni- and multivariable analyses to evaluate the link between CHIP and RA patient phenotypes. Results: We identified CHIP in 381 (3.8%) FINRISK participants. As expected, the prevalence of CHIP increased with age, and CHIP variants were associated with prevalent and incident hematologic malignancies. C-reactive protein values were higher in participants with CHIP (P<0.001). In multivariable models, CHIP status (OR 1.98, 1.13-3.45, P=0.017) was more common in participants with questionnaire-based history of RA (N=151) (Fig 1A). In the FinnGen study, CHIP carrier status was associated with prevalent autoimmune diseases (OR 1.26, 1.05-1.53, P=0.015). We validated the association between prevalent RA (N=10 990) and CHIP carrier status (N=2 393) in the FinnGen study (OR 1.58, 1.18-2.13, P=0.002) (Fig 1A). In particular, CHIP (OR 2.17, 1.40-3.36, P<0.001), and GNB1 variants (OR 4.71, 1.87-11.82, P=0.001) were associated with the seronegative RA, whereas DNMT3A was associated with seropositive RA (OR 1.83, 1.08-3.09, P=0.024). We observed no association between autosomal mCAs and RA. Using targeted panel sequencing, we found CHIP in 22% (146/675) of RA patients. RA patients with CHIP variants were older than those without a CHIP variant (median, 69 vs 63 years, P<0.001). Erythrocyte sedimentation rate (ESR) was higher in those with CHIP (median, 42.5 vs 26.0 mm/h, P=0.043). RA patients with TET2 variants were more likely to have seronegative disease compared with TET2 wild-type patients (proportion of seronegative patients, 56% vs 33%, OR 2.62; P=0.008) (Fig 1B). The association remained significant (OR 2.38, 1.14-4.96, P=0.02) when adjusting for age, sex, and smoking. Evaluation of longitudinal disease outcomes by CHIP status are currently ongoing. Conclusion: CHIP variants were more common in participants with prevalent RA in two population-level cohorts, and the association was specific to seronegative RA in the large FinnGen cohort. In newly diagnosed RA patients, TET2 mutations were more common in seronegative than in seropositive RA. While the causality and the functional basis of these observations remain to be explored in future studies, our results provide additional evidence for the association between CHIP and inflammation in various disease contexts, including autoimmune diseases.

Transcriptomic without Clinical Response to Menin Inhibition As a Mechanism of Upfront Resistance in Samples from Mutliply Relapsed Patients with KMT2A-Rearranged Leukemia

Inhibitors of the Menin-KMT2A interaction are emerging as promising agents for the treatment of KMT2A-rearranged ( KMT2A-r) acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL). Menin is required for the recruitment of the KMT2A-fusion to a subset of its target genes. Early phase clinical trials have shown highly encouraging results in patients with relapsed and refractory leukemias with KMT2A rearrangements and NPM1 mutations. In a recently reported early phase clinical trial evaluating the Menin inhibitor revumenib, about 50% of patients failed to show a clinical response despite robust downregulation of known Menin:KMT2A target genes, suggesting on target activity of the drug. This was the most common pattern of upfront resistance. We sought to understand upfront resistance (in contrast to the reported induced resistance via MEN1 mutations) in patient derived xenograft models. To this end, we evaluated the efficacy of Menin-inhibition in KMT2A-r leukemias with high-risk features: high-risk KMT2A fusion partners in AML ( MLLT10, MLLT4), infant ALL, and serial samples from patients with multiply relapsed ALL. We found that AML samples with high-risk fusion partners banked at initial diagnosis were sensitive to Menin-inhibition in xenografts. In contrast, three highly pretreated samples from patients with KMT2A-AFF1 ALL showed only a slight, or no decrease in latentcy in the Menin-inhibitor arm. Interestingly, Menin inhibition showed excellent in vivo efficacy against leukemia samples obtained from the same three patients earlier in their disease course. Therefore, prior chemotherapy (+/- immunotherapy) selects for Menin-inhibitor resistant cells without prior exposure to Menin inhibitor. Transcriptomic analysis (RNA-Seq) documents sustained on-target efficacy of the Menin inhibitor on a transcriptional level in resistant cells. None of the samples contained a MEN1 mutation at any time point. Targeted panel sequencing at initial diagnosis and relapse documented clonal evolution and the emergence of multiple co-mutations (particularly RAS pathway mutations and TP53 mutations), however, all co-mutations tested failed to induce resistance as a single alteration. It is possible that the cooperation of multiple emerging mutations (rather than a single mutation) is sufficient to drive AML in a KMT2A-fusion independent manner, thus mediating Menin resistance. A transcriptomic response (i.e. downregulation of KMT2A-fusion target genes) without a clinical response was the most common pattern of upfront resistance in the recently reported phase I/II clinical trial of the Menin inhibitor revumenib. We conclude that it will be critical to use Menin-inhibitors upfront or in early lines of therapy before substantial clonal evolution has occurred.

The Complex Structural Variant Chromothripsis Can be Defined on Targeted Sequencing Panels, Allowing Direct Clinical Translation in Order to Improve Multiple Myeloma Prognostication

Introduction: the complex structural variant (SV) chromothripsis is an independent predictor of progression-free (PFS) and overall survival (OS) in newly diagnosed multiple myeloma (NDMM); predictive in multivariate analysis when considering ISS, IgH translocations and TP53 status (Rustad BCD 2020). Chromothripsis is not included in current prognostic scores, despite a prevalence of 24% in NDMM. While gold-standard definition requires whole genome sequencing (WGS), we've previously shown that copy-number (CN) signatures are also able to accurately define chromothripsis in whole exome sequencing (WES, Maclachlan Nat Com 2021). As neither WGS nor WES are currently used in clinical medicine, we applied CN signatures to targeted sequencing, aiming for direct clinical translation of this key biological feature into clinical practice. Memorial Sloan Kettering (MSK) now offers the MSK-Heme-IMPACT targeted sequencing panel to all MM patients having a bone marrow biopsy (BMBx), as well as offering retrospective analysis for current patients having leftover DNA from prior BMBx. Methods: We analyzed 420 samples having either myTYPE or MSK-Heme-IMPACT targeted sequencing at MSK prior to July 2023. We excluded those with whole genome duplication, applied CN signatures, then for samples from NDMM performed multivariate analysis for PFS prediction by backwards stepwise regression. In samples with CN signatures predictive for chromothripsis and with DNA available from the same BMBx, we performed parallel WGS in order to check for chromothripsis. WGS sequencing was at 60-80x for tumor, 30-40x for normal, with analysis including Battenberg for CN analysis, with SvABA, BRASS and GRIDSS for SV calling. Additional parallel WGS is in progress for 45 samples with NDMM prior to DVRd or DKRd, in order to examine chromothripsis prediction and PFS impact in the context of highly effective induction. Results: From 420 samples, 48 were collected at the precursor stage, 205 with NDMM, and 167 with relapsed disease. 39 patients had 2 assays performed, while the remainder were unique patient samples. Within NDMM, the range of induction regimens received included lenalidomide and dexamethasone with bortezomib (VRd), carfilzomib (KRd) or daratumumab (DRd), along with quadruplet therapy (DVRd or DKRd). Targeted sequencing data detected all 6 key CN features observed in WGS and WES analysis. Similar to WES, there was a lower contribution from very high breakpoint counts per 10mB and longer lengths of oscillating CN segments, due to the lower resolution of targeted sequencing data. De novo extraction of CN signatures defined 1 signature containing multiple features consistent with chromothripsis; high breakpoint count per 10mB, more jumps between adjacent CN segments, longer lengths of oscillating CN segments and a predominance of small CN segments. When examining samples from NDMM, contribution from this signature was predictive of PFS in multivariate analysis when considering age, ISS, t(4;14), TP53 status and gain1q21 (hazard ratio > 4, p=0.003). 33 samples had a predictive probability for chromothripsis > 0.6 (the cut-off in WGS / WES having the highest sensitivity and specificity for prediction), with also having further DNA available. 30/33 (91%) were confirmed to harbor chromothripsis on manual inspection of WGS data. 19/30 (63%) with chromothripsis were BMBx from NDMM, while 11/30 (37%) were from relapsed disease. Discussion: the complex SV chromothripsis is a key genomic risk factor predicting for poor PFS and OS in MM, which is currently not being captured in routine practice or in clinical trials. Here, we demonstrate that chromothripsis can be predicted from targeted sequencing data, confirmed by gold-standard WGS analysis. Ongoing studies are expanding our cohort of NDMM with concurrent targeting sequencing and WGS, to define the sensitivity and specificity of targeted-sequencing-based CN signature prediction of chromothripsis. The additional cohort largely comprises patients treated with daratumumab-quadruplet induction therapy, aiming to confirm that chromothripsis and CN signatures retain prognostic significance in the context of this therapy. With this data is in hand, we will develop a CN signature fitting tool, similar to mmsig for mutational signatures (Rustad Comm Biol 2021), to allow easily accessible incorporation of chromothripsis assessment into clinical practice.

Clinical and Molecular Characterization of Triple-Negative Essential Thrombocythemia: Data from the Prospective Spanish Registry of Essential Thrombocythemia

Introduction Triple-negative Essential Thrombocythemia (TN-ET) accounts for up to 10-20% of patients with a histological diagnosis of ET. Despite being considered an indolent disease, it remains a diagnostic and therapeutic challenge as it can be confused with non-clonal thrombocytosis and its prognostic factors and management are not standardized. Objectives The aim of this study was to evaluate the utility of a Next Generation Sequencing (NGS) panel targeting recurrently mutated genes in myeloid pathology for detecting genetic variants in patients diagnosed with TN-ET and to assess the significance of these variants in the risk of thrombosis and hemorrhage, progression to Acute Leukemia (AL) or Myelofibrosis (MF), and overall survival. Materials and Methods Data from 151 TN-ET patients with confirmed biopsy and NGS sequencing from the Spanish Registry of Essential Thrombocythemia of the Spanish Group of Philadelphia-Negative Myeloproliferative Neoplasms was obtained. This registry includes information on diagnosis, molecular biology, clinical characteristics, and molecular biology. In addition, 39 patients whose molecular diagnosis was performed at the Molecular Biology Laboratory of the Hematology Department at Hospital Universitario 12 de Octubre, Madrid, Spain were included. This last panel included the genes: ANKRD26, ASXL1, ATG2B, BCOR, BCORL1, CALR, CBL, CEBPA, CSF3R, CUX1, DDX41, DNMT3A, EGLN1, EPAS1, EPOR, ETNK1, ETV6, EZH2, FLT3, GATA2, GNAS, GNB1, GSKIP, IDH1, IDH2, JAK2, KDM6A, KIT, KMT2A, KRAS, MBD4, MPL, NF1, NFE2, NPM1, NRAS, PHF6, PPM1D, PRPF8, PTPN11, RAD21, RIT1, RRAS, RUNX1, SAMD9, SAMD9L, SETBP1, SF3A1, SF3B1, SH2B3, SMC1A, SMC3, SRSF2, STAG2, TERC, TERT, TET2, THPO, TP53, U2AF1, VHL, WT1, ZRSR2. Statistical analyses were performed using SPSS v.29.0. Results A canonical mutation was detected in 20 out of the 190 studied patients, ( JAK2 V617F n=13, CALR n=3 and MPL n=4) being discarded from subsequent studies. Among the other 170 samples, 206 genetic variants were detected in a total of 86 patients (median 2, range 1-9). A total of 39 pathogenic (P) or probably pathogenic (PP) variants were found in 28 patients, whereas 59 uncertain significance (US) and 108 benign (B) variants were found in 44 and 54 patients, respectively. The distribution of genes in which P/PP/US variants were detected is shown in Figure 1. Patients with P /PP mutations had a statistically higher age compared to the remaining (50 vs. 64.8 years; p<0.001). The prevalence of males was higher in the group where P/PP mutations were detected (57.14% vs. 28.87%; p=0.004). No differences were found regarding cardiovascular comorbidity or symptomatology at diagnosis, but the P/PP group presented a higher frequency of thrombosis prior to diagnosis (28.57% vs. 5.64%; p=0.001). There were no differences in hemoglobin, leukocyte, or platelet counts at diagnosis according to the presence of P/PP variants. The presence of P/PP mutations was associated with a higher risk of progression to AL and a lower overall survival. A higher risk of thrombosis or progression to MF was not found (Figure 2). No differences were found in the risk of thrombosis, transformation, or overall survival between patients with or without the presence of US, PB or B variants. Conclusions The use of NGS technology is useful for the diagnosis and prognostication of TN-ET. The detection of P/PP mutations in patients diagnosed with TN-ET is associated with a higher risk of transformation to AL and lower overall survival, but not with higher risk of thrombosis or progression to MF. It is necessary to evaluate the individual risk of each of these mutations and the potential prognostic role of variants of US.

Performance of a Liquid Biopsy Diagnostic Prediction Model for EBV-Positive Burkitt Lymphoma in Sub-Saharan Africa

Burkitt Lymphoma (BL) is classified into EBV-positive BL (EBL) and EBV-negative BL (sporadic BL), the former being more prevalent in SSA (Al-Khreisat et al 2023). Conventional diagnosis of BL depends on the morphological assessment of invasive tissue biopsy followed by immunohistochemistry (IHC) and fluorescent in situ hybridization (FISH) to reveal the pathognomonic t(8;14), or t(8;22) translocations. This is associated with significant diagnostic delay in regions with limited pathology services (Mawalla et al 2023), often necessitating initiation of treatment based on clinical or histopathological findings without immunohistochemistry (Ogwang et al 2011), with ensuing mismanagement contributing to poor survival outcomes observed in SSA compared to High-Income Countries (HIC) (Sung et al 2021). Here, we sought to explore whether liquid biopsy could be used as an initial diagnostic aid for EBL in settings without easy access to pathology using different EBV DNA parameters and the MYC-Immunoglobulin ( MYC-Ig) gene rearrangement. This prospective study collected whole blood samples prospectively from 221 participants (90 BL and 131 non-BL) enrolled in the Aggressive Infection-Related East Africa Lymphoma (AI-REAL) project (Legason et al 2022) across 5 sites in Tanzania and Uganda. The revised BL diagnostic algorithm (Naresh et al 2021) was used as a gold-standard diagnosis. CfDNA was extracted using the QIAamp Circulating Nucleic Acid Kit (Qiagen, USA), according to the manufacturer's instructions. DNA libraries were prepared with the Thru PLEX Tag-Seq HV kit (Takara Bio Inc. Japan), and hybridization capture was performed using a custom, next-generation sequencing (NGS) panel targeting genes commonly mutated in BL and three EBV genes: EBER1, EBER2, and EBNA2. All sequencing was performed in-country (MiSeq, Illumina, CA, USA) and analyzed in a custom pipeline. Using custom scripts, EBV DNA copies per cell and EBV DNA fragment size ratio were calculated as the proportion of EBV fragments of length 180 - 200 base pairs (bp) divided by the proportion of autosomal fragments within the same size window. The diversity in the fragment size distribution of the EBV and autosomal DNA was calculated and presented as EBV entropy and autosomal entropy. Translocations were called using IgCaller (Nadeu et al 2020) and Genomic Rearrangement Identification Software Suite (GRIDSS) (Cameron et al 2017). Univariate analysis was performed to assess the association between predictor variables (clinical parameters, EBV parameters, and MYC translocation) and a diagnosis of EBL. Ten-fold cross-validation was performed on a liquid biopsy model consisting of all the liquid biopsy variables with the clinical variables included in the model as covariates. Performance metrics for the liquid biopsy model as a diagnostic aid were calculated, and variable importance was determined to assess each variable's relative influence on the model. In our cohort, EBL diagnosis was negatively associated with tumor site (neck) and not significantly associated with tumor site (jaw). A shorter duration of symptoms was associated with a diagnosis of EBL. Presence of MYC translocation t(8;14), higher EBV fragment size ratio, proportion, entropy, and copies per cell for EBER1, EBER2, and EBNA2 were all associated with a diagnosis of EBL.The combined liquid biopsy diagnostic prediction model had an estimated AUC of 90%, with sensitivity between 85%-90%, specificity between 75% - 80%, and positive and negative predictive values of 85% (Figure 1). MYC translocation t(8;14) was the most important diagnostic variable, followed by EBV entropy (Figure 2). This study demonstrates the feasibility of liquid biopsy as a non-invasive diagnostic aid for EBL in settings without easy access to pathology. To strengthen the application of this technology into clinical practice, further analysis is being conducted for weighting and assigning of scores for each predictor variable. This would assist clinicians in making a precise diagnosis, complementing existing diagnostic modalities for EBL, and reducing the rate of misdiagnosis arising from diagnosis based on clinical suspicion and tissue morphology without immunohistochemistry.

Lymphoma Cell Sorting to Enrich Low Degree Infiltration Clonal B-Cells for Diagnostic Delineation of Molecular Genetic Background By NGS

Introduction: According to WHO 2022 classificationthe vast majority of B-cell lymphoma entities continue to be defined histologically and by their immunophenotype. However, genetics help to define these entities and provide prognostic information. NGS targeted panel sequencing with short turn-around time has emerged as powerful diagnostic tool to uncover relevant genes frequently mutated in B-cell lymphoma. Yet its sensitivity limits its application to lymphoma cases with clone sizes >5%. In case of clinical symptoms and need for therapy small B-cell clones (<5%) can be diagnostically challenging if tissue for histopathology is difficult to obtain and immunophenotype is not characteristic for a specific lymphoma entitiy such as splenic B-cell lymphoma. Aim: To enrich small B-cell clones (<5%) by cell sorting to >15% to enable lymphoma NGS panel sequencing for potential genetic characterization and risk stratification. Methods: Peripheral blood or bone marrow samples (n=43) that containedsmall B-cells clones (<5%) as identified by flow cytometric routine diagnostics were stained for cell sorting by a combination of monoclonal antibodies against 9 B-cell antigens, including CD5, CD19 and CD45. The gating strategy excludes debris and doublets and defines B- and T-cells out of CD45-positive events. By application of cell sorting (SRT Cytoflex, Beckman Coulter) we aimed at >500,000 cells consisting of B- and T-cells, containing >15% clonal B-cells to ensure sufficient NGS sensitivity but minimize sort time. DNA was extracted from sorted samples and lymphoma NGS panel sequencing (66 genes ~ 1,500x coverage, min. 400x) was performed (NovaSeq 6000, Illumina). For technical and biological validation, native sample material was sequenced in parallel for a subset of patient samples (n=30). Results: In total43samples with <5% clonal B-cells were applied for cell sorting and clonal B-cells were enriched to a median of 17% (range, 3-97%) in a median of 800,000 cells (range, 127,400-3,780,000). In a pilot study to optimize the sorting procedure 8 of 13 sorted samples (62%) revealed quantitatively sufficient and good quality DNA and could be successfully analyzed by NGS. In 4 of these 8 samples (50%) mutations were identified. Optimized processing regarding cell sorting and DNA extraction was applied to further 30 samples all of which could be successfully sequenced. Overall, in 50% of all samples with NGS analysis after cell sorting (19/38) one or more mutations (total number of mutations: n=30) associated with B-cell lymphoma were detected. Mutations comprised genes frequently affected in B-cell lymphoma: MYD88 (30% of detected mutations, 9/30), CXCR4 (20%, 6/30), NOTCH1 (7%, 2/30), NOTCH2 (3%, 1/30), KMT2D (7%, 2/30), TP53 (13%, 4/30), BRAF (7%, 2/30), KLF2 (10%, 3/30) and DNMT3A (3%, 1/30). All of the cases negative for NGS after sorting (n=15) were negative also by NGS on matched native sample material without prior enrichment of clonal B-cells. Out of the 15 cases with positive NGS results after sorting NGS on native sample material revealed mutations in only 4 (27%) cases (only 1 case with detection of all mutations) while in 11 (73%) cases no mutations were detected. In the 15 cases positive after sorting median variant allele frequency (VAF) was 8.8% (range 3.4-38.3%). As anticipated, in the 4 cases with mutations detected also without sorting VAF was lower in native sample material (median 3.1%, range 1.98-4.9%) as compared to sorted sample material (median 8.9%, range 6.8-23.2%, p=0.0412). Conclusion: Cell sorting can overcome the diagnostic challenge of low degree infiltration B-cell lymphoma that require molecular genetics for further classification and risk prediction prior to therapy. Analysis of 30 samples and comparison with matched native material revealed that enrichment targeting 15% of clonal B-cells in a total of >500.000 cells are desirable for successful and sensitive NGS panel sequencing. In 50% of such cases mutations are detectable after sorting. Cell sorting to enrich clonal B-cells thus is a powerful tool to enable NGS panel sequencing for low percentage mature B-cell lymphoma and should be considered for routine diagnostics of B-cell lymphoma.

Nanopore Cas9-Targeted Long-Read Sequencing - a Fast and Flexible Diagnostic Tool for the Identification of B-Cell Acute Lymphoblastic Leukemia Associated Gene Rearrangements

B-cell acute lymphoblastic leukemia (B-ALL) is characterized by a multitude of recurrent genetic alterations, which drive malignant transformation. Besides their essential role in leukemia biology these aberrations also serve as predictors of treatment response and survival, making their reliable detection a prerequisite for risk stratification in state-of-the-art clinical trials. Currently, the genetic characterization of B-ALL in most diagnostic laboratories still relies on an integrative approach combining conventional cytogenetics, reverse transcription polymerase chain reaction (RT-PCR) and fluorescence in situ hybridization (FISH) assays as well as single nucleotide polymorphism (SNP) array analysis. However, these conventional diagnostic tools fail to identify a considerable number of genetic alterations, particularly cryptic rearrangements such as IGH::DUX4, IGH::EPOR or MEF2D involving fusions. Although next generation DNA- and RNA-based sequencing (NGS) assays have significantly improved the reliable detection of fusion genes, the need for large sample numbers to achieve acceptable turnaround times and cost effectiveness impairs their application in small- to medium-sized diagnostic laboratories. To address these issues, we developed a targeted sequencing assay, based on the commercially available combination of Cas9-mediated target enrichment and Nanopore long-read sequencing, which facilitates the identification of B-ALL associated gene rearrangements on the DNA level in a single sample format. The targeted panel design included an initial review of the literature to define the breakpoint-spanning regions involved in fusion gene formation followed by the design of specific guide RNAs (gRNA) to facilitate Cas9-mediated enrichment of these regions of interest (ROIs). We first conducted ROI enrichment and sequencing with a limited number of gRNAs targeting only ETV6, KMT2A and DUX4 fusions. Then we progressively extended the pool of gRNAs with each sequencing run until it comprised 91 gRNAs covering most B-ALL-associated rearrangements. This stepwise approach allowed us to identify negative effects of increasing panel size on overall enrichment efficacy and simplified the identification and replacement of poor performing gRNAs. Our initial cohort consisted of 12 well-characterized samples harboring 13 gene rearrangements, including ETV6::RUNX1, KMT2A::AFF1, MEF2D::BCL9, ZMIZ1::ABL1, SSBP2::CSF1R, PDGFRB::EBF1, IGH::EPOR, IGH::DUX4, IGH::CRLF2, IGH::CEBPE, IGH::MYC, and IGH::BCL2. To ensure easy implementation in diagnostic workflows, we used genomic DNA, isolated according to standard protocols. Cas9-based target enrichment and long-read sequencing on MinION flow cells (v9.4.1) were performed according to protocols provided by Oxford Nanopore and each sample was sequenced for 72 hours. On average, each run yielded 3.72±1.21 Gb, a mean read length of 7.40±1.00 kb with an N50 value of 12.87±1.71 kb and an average read quality (Q) score of 14.72±0.52. For data processing and analysis, we adapted the existing nextflow nf-core bioinformatics pipeline, nanoseq, to include publicly available structural variation callers and quality control tools. Our approach correctly identified 92.3% (12/13) of the rearrangements with a mean of 42 (range 2-97) breakpoint-supporting reads (BSR) per alteration. Only the IGH::MYC translocation was missed as the breakpoint was located outside of our predefined ROIs for MYC. To corroborate these results, we sequenced an additional cohort of 14 B-ALL samples, including ZNF384::DUX4, KMT2A::MLLT10, KMT2A::USP2, PAX5::SOX5, EP300::ZNF384, P2RY8::CRLF2, and IGH::DUX4 positive cases and were able to correctly identify 100% (14/14) of the fusion genes with an average of 83 (range 7-367) BSR. In conclusion, our newly developed targeted sequencing panel correctly identified 96.3% (26/27) analyzed gene rearrangements in B-ALL, including alterations evading standard diagnostic methodologies such as IGH::DUX4 or IGH::EPOR. Importantly, the assay allows individual analysis in a single sample format, thereby representing a valuable tool for B-ALL diagnostics in small to medium-sized laboratories.

Is Allogeneic Hematopoietic Stem Cell Transplant (HCT) in First Remission Beneficial to Acute Myeloid Leukemia (AML) Patients with Isocitrate Dehydrogenase(IDH) Mutations?

Introduction Somatic mutations in the isocitrate dehydrogenase 1 and 2 genes (IDH1 and IDH2) are relatively frequent in acute myeloid leukemia (AML) but with heterogeneous prognoses. The prognostic impact of this AML subtype with IDH mutations may be influenced by the comutational status with other somatic mutation, the specific types of IDH1/2 mutation (ie, IDH1 R132, IDH2 R140, and IDH2 R172). Allogeneic stem cell transplant allogeneic HCT represents a curative option for subgroups of AML patients (pts), but not for all subtypes. Transplant referral is made based on the ELN risk group classified into favorable, intermediate, and adverse risk. Although many AML pts with IDH1/2 mutations are referred to allogeneic hematopoietic stem cell transplantation (HCT) and transplanted, however it is not fully elucidated whether allogeneic HCT brings definite survival benefit to this group of the pts. Statistical attention is to be paid to the time to HCT from CR1 achievement. Before HCT, the case should be counted as a case of no HCT, while after HCT this case can be accounted as the case received HCT using the Mantel-Byar test (MBT). To handle immortal time bias, the Mantel-Byar method was used to depict outcomes during the follow-up period. Kaplan-Meier curves were avoided as they can introduce immortal bias, considering the period of time before treatment as part of the treatment arm. Instead, the MBT and landmark methods were employed to address this issue. Allogeneic HCT was treated as a time-dependent variable to prevent immortal bias. The aim of the present study is to identify if AML pts with IDH mutations obtain survival benefit from allogeneic HCT in CR1 using HCT as a time-dependent covariate with using Mantel-Byar test. Patients & Method We conducted a retrospective study that included 1,228 adult pts who were diagnosed with AML and achieved CR1 in 5 centers (Toronto, Vancouver, and Montreal, Canada; Hwasun and Seoul, South Korea) from January 1997 to January 2020. Next-generation sequencing was performed in each institution according to their own standard targeted sequencing panel. We considered allogeneic HCT in CR1 as a time-dependent covariate and used a Mantel-Byar test to avoid immortal bias. Relapse-free survival (RFS) was selected for the primary endpoint, defined as the time from CR1 to relapse or death from any cause. EZR software version 1.41 was used throughout the study for statistical analysis. Results Out of 1228, a total of 238 pts (19.4%) was found to have IDH1/2 mutation, among whom 92 pts (38.6%) had isolated IDH1 mutations, 140 pts (58.8%), isolated IDH2 mutations, and 6 pts (2.5%) having coexistence of IDH1 and IDH2 mutations. With a median follow-up duration of 20.8 months (mo) (1-194.3 mo) among survivors after achieving CR1, 91 pts died while 115 pts experienced relapse. One hundred twenty-one pts received HCT in CR1, and 44 pts relapsed after HCT. The 2-year overall survival of those 238 pts was 58.9% (51.3-65.6), and 2-yr RFS was48.9% (41.6-55.8) Between the HCT (n=121) vs non-HCT group (n=117), there were no differences in sex, cytogenetic risk or other somatic mutation except NPM1 mutation which was more frequently detected in the non-HCT group (p =0.002). As expected, the HCT group was younger (55 years (15-74 yrs)) than non-HCT group (62 years (27-82 yrs), p<0.001). First, the RFS and OS was compared between HCT vs non-HCT groups considering HCT as non-time dependent covariate. The 2-year RFS rate was higher in the HCT group (61.0%) compared to the non-HCT group (37.0%, p= 0.0002). However, it was not translated into superior OS in the HCT group: 2-year OS rate, 64.4% in the HCT group vs 53.7% in the non-HCT group, p=0.1). We have conducted time-dependent analysis taking taken HCT in CR1 into account as a time-dependent (td) covariate using MBT. The trend of superior RFS in HCT group was reproduced in MBT: HCT in CR1 was found to improve RFS (Hazard ratio [(HR) of 0.62, (95% C.I. [0.42-0.92]; p=0.02)] as shown in Fig.1B utilizing Simon-Makuch plot. However, Multivariate analysis (MVA) for RFS did not confirm HCT in CR1 as an independent prognostic factor for RFS, while Following factors were found to be associated with RFS adversely: age > 60 years, presence of DNMT3A mutation, and presence of IDH1 mutation. Conclusion Allogeneic HCT in CR1 could bring some survival benefit on RFS in AML pts with IDH1/2 mutation, but not confirmed in MVA. Further study is warranted to reach a clearer conclusion with larger number of

Recent Patterns of Radiotherapy Use in Multiple Myeloma: Clinical Outcomes and Possible Radiologic, Biochemical, and Genomic Correlates of Response

Background/Objectives: Systemic therapies for multiple myeloma (MM) have advanced considerably, changing the landscape of MM treatment paradigms. Yet, the use of radiotherapy (RT) has remained heterogeneous, and even controversial, due to minimal data on outcomes. With the ultimate goal of guiding the design of prospective trials incorporating RT, we initiated a study of our institutional experience treating MM with RT since 1/1/2000. Here we report a preliminary feasibility analysis of an initial sample cohort, identifying patterns of RT use, outcomes, and the impact of RT on radiographic and biochemical markers, with genomic characterization for more recently treated patients. Materials/Methods: 506 pathologically confirmed MM patients who received RT to 1190 sites between 1/1/2000 and 6/1/2022 were identified. Patient, disease, and treatment characteristics were analyzed for 50 consecutive patients treated in 2019 (recent sampling but with reasonable follow for analysis) and tested for association with cumulative incidence of local and distant failure (LF, DF) using univariable and multivariable analysis. Genomic data was obtained via next generation sequencing using an institutional targeted sequencing panel (Memorial Sloan Kettering IMPACT). Results: Among the 50 patients analyzed (median 63 years), 90 lesions were treated with RT, 33% with concurrent systemic therapy, to a median dose of 20 Gy (8-46 Gy) over a median of 5 fractions (1-25). RT Indications were pain (56%), critical structure involvement (25%), peri-operative (9%), salvage/consolidation (8%), and bridging therapy (2%). Median size of RT-treated lesions was 4.2 cm (1.4-7.9) and included non-vertebral bones (62%), spine (24%), and extramedullary sites (14%). The median number of lines of pre-RT therapy was 7 (1-14) and 51% had >9 lesions on imaging, 47% involving both medullary and extramedullary sites. Pre-RT PET imaging was performed in a majority of patients (62%). With a median follow-up of 12.4 months (0.5-46), LF occurred in 5% of treated sites and 89% had DF, most commonly in both medullary + extramedullary (51.4%) sites. Of the patients with DF, 93% had sustained local control of the irradiated site. Absolute decreases 1-week to 1-month post-RT were observed in marrow plasma cell percentage, (median 4.0%), M spike (0.30 g/dL), total protein (0.3 g/dL), K:L ratio (0.01), lesion size (1.5cm), and lesion SUV (3.1) but in this limited sample, none were significantly associated with disease control. A cohort of 62 RT-treated MM patients from 2016-2022 had genomic data available; most common tumor mutations were in TP53 (35%), HIST1 (34%), NRAS (34%), and KRAS (23%). Conclusion: In this pilot analysis of a sampling cohort of RT-treated MM, we report on patterns of utilization, outcomes, and biochemical and radiographic correlates. Ultimately, we aim to characterize the role of RT in the modern era of systemic therapy to guide the design of future prospective trials and to inform novel approaches for incorporating RT into the treatment paradigm. At the meeting, we will present data from a larger cohort of MM patients, with further sub-analyses on the relevance of baseline imaging to subsequent irradiated lesions, as well as toxicity and outcomes data with the use of radiotherapy in conjunction with newer systemic agents (bi-specific antibodies and chimeric antigen receptor T-cell therapy).

Morphologic and Genetic Features of Unclassifiable Chronic Myeloid Neoplasms: Five-Year Experience of a Canadian Provincial Referral Center

Introduction: The challenge of diagnosing and definitively classifying myeloproliferative neoplasms (MPNs) and myelodysplastic/myeloproliferative neoplasms (MDS/MPNs) remains a conundrum in the field of hematology. While the World Health Organization has outlined diagnostic criteria for various categories, there remain cases that elude specific classification, falling under the umbrella of MPN, unclassifiable (MPN-U) and MDS/MPN, unclassifiable (MDS/MPN-U). In this study, we aimed to compare the morphological and genetic profiles of cases classified as MPN-U or MDS/MPN-U against more definitively classified cases (including chronic myelomonocytic leukemia (CMML), essential thrombocythemia (ET), MDS/MPN with ring sideroblasts and thrombocytosis (MDS/MPN-RS-T), and primary myelofibrosis (PMF), including prefibrotic PMF (pre-PMF)). Methods: Morphologically unclassifiable myeloid neoplasm cases referred for pathologist review to our laboratory in the largest tertiary care center in British Columbia, Canada were identified through pathology database query over a five-year period. Bone marrow materials were reviewed and reanalyzed with standardized collection of qualitative and semiquantitative data by two hematopathologists. Morphological parameters studied included the myeloid to erythroid ratio, bone marrow cellularity, lineage dysplasia, megakaryocyte histotopography, bone marrow blast frequency, fibrotic changes, and bone marrow vascularity, among others. Clinical data, cytogenetic findings, and results of a targeted sequencing panel assessing a set of approximately 50 genes recurrently mutated in myeloid malignancies were collected by electronic medical record review. A cohort of contemporarily collected myeloid neoplasms with definitive pathologic diagnoses was reviewed for comparison. Statistical analyses were performed using python3. Results: The cohort comprised 29 cases classified as MPN-U, 12 cases of MDS/MPN-U, and 5 cases each of ET, PMF, pre-PMF, CMML, and MDS/MPN-RS-T. MPN-U cases were further subcategorized based on the apparent reason for difficulty in classification, including: overlapping features with another MPN; late stage/overt fibrosis; early stage without fully-developed morphologic features; and coexisting atypical morphologic findings. MPN-U with overlapping features had highly variable reticulin fibrosis, while, as expected, late stage entities had a high degree of reticulin fibrosis and early stage entities showed less overt fibrosis (Figure, Left, demonstrating select morphologic features of evaluated chronic myeloid neoplasms). MPN-U as a whole appeared to show less prominent fibrosis-associated findings than MDS/MPN-U and to, in general, show a greater degree of neovascularization, similar to changes observed in cases with PMF. Splicing factor and epigenetic modifier mutations appeared to be overrepresented in MDS/MPN-U relative to MPN-U, notably excluding MPN-U with atypical morphologic features (Figure, Right, demonstrating the distribution of pathogenic variants identified in evaluated myeloid neoplasms; results are indicated as presence (1) or absence (0) of a variant for each gene). Notably, no CALR mutations were identified in the MDS/MPN-U category or early stage / atypical morphology MPN-U, though they were enriched in overtly fibrotic MPN-U entities. Discussion/Conclusion: In this cohort we observed, as expected, significant overlap in cases classified as MPN-U and MDS/MPN-U with better delineated entities, though patterns of morphologic change and mutational spectra suggest potential directions for further investigation of additional criteria that may guide accurate classification of these neoplasms. These findings reflect the heterogeneity of MPNs described in the literature and the complexity of their genetic underpinnings, particularly in the context of concurrent dyshematopoiesis; suggest enhanced bone marrow vascularization may be an informative differential diagnostic criterion in evaluation of these neoplasms, though further study is required; and emphasize the need for broader genetic testing and perhaps further subtyping of MPN-U cases for a more nuanced analytical approaches for these disorders.

BIOM-58. ZERO-WASTE MOLECULAR DIAGNOSTICS FROM BIOPSY NEEDLE WASHES

Abstract BACKGROUND Brain tumor classification has increasingly integrated molecular diagnostics into official criteria. However, biopsy tissue is limited due to the small quantity acquired in eloquent areas of the brain and is, therefore, usually earmarked for histology. We have identified that the fluid used to wash biopsy needles contains excess tumor material that can be diagnostically useful. Characterization of this non-zero-sum waste material provides an opportunity for improvement in the diagnosis and treatment for brain tumor patients. METHODS For suspected or established high grade brain tumor biopsies (n = 37 across a multi-institutional cohort at submission), neurosurgeons were instructed to save fluid used to wash biopsy needles. Samples were processed and characterized in a variety of methods to establish diagnostic utility including (1) cell-free DNA isolation and targeted panel sequencing (n = 13), (2) isothermal amplification of cell-free DNA from raw wash water for rapid diagnostics (n = 2), (3) cell-line generation (n = 24) and, (4) cell-free mutant protein capture and quantification (n = 9). RESULTS (1) 13 wash samples were considered for cf-tDNA extraction and Illumina TruSight Oncology 500 panel sequencing (with extracted DNA ranging from 20ng-40ug). In 2 returned cases (11 pending), concordance was 100% with clinical targeted panel sequencing where panels overlapped. (2) One case suspected of harboring H3K27M was subjected to targeted, loop-mediated isothermal amplification and Oxford Nanopore real-time sequencing, confirming an H3.3K27M diagnosis (11.8% allele fraction) within 1hr35min of receipt of sample from the operating room. (3) 5 cases resulted in primary cell culture generation (4) A novel single-molecule mutant protein quantification technique was performed on 9 cases. RESULTS from all analyses and concordance with outstanding sequencing will be presented. CONCLUSION We demonstrate that biopsy byproducts contain enough high-quality tumor-derived material to perform a variety of useful diagnostics with the potential to have immediate translational impact on standard of care.

Can Whole Genome and Whole Transcriptome Sequencing Replace Standard Procedures in CLL Diagnostics?

Background: The advantages of genome-wide sequencing approaches over conventional methods in CLL diagnostics are a matter of debate and exact guidelines for the application of next-generation sequencing in a diagnostic context are currently missing. Aim: Compare the accuracy of whole genome sequencing (WGS) and whole transcriptome sequencing (WTS) in determining the presence of point mutations, IGHV mutational status (IGHVms), and chromosomal aberrations to conventional procedures in a clinical setting. Patients and Methods: The cohort comprised 317 CLL patients. Diagnosis was established following WHO guidelines. WGS (100x, 2x151bp) and WTS (5x10 7 reads, 2x101bp) data were generated on NovaSeq instruments. For routine diagnostics, sequencing was carried out by Sanger or targeted panel sequencing (1500x, 2x101bp). Chromosome banding analysis (CBA) and FISH were available for 231 patients. WGS data were analyzed for chromosomal aberrations and copy number alterations using Manta and GATK. IGHVms was determined by Sanger sequencing and/or fragment analysis for 248 patients. Analysis of IGHVms was carried out using IgCaller (WGS) and MiXCR (WTS). Results: Comparing copy number variants (CNV) and balanced structural variants (SV) between CBA and WGS revealed a high concordance. In detail, 297 CNV were detected by both techniques while 32 and 52 were detected only by CBA (small subclones) or by WGS, respectively. 33 balanced SV were detected by both while 8 and 3 were detected only by CBA and WGS, respectively. A complex karyotype (≥ 3 abnormalities) was observed by both techniques in 55 cases, while 5 and 10 cases were categorized as complex by CBA only or WGS only, respectively. Survival analysis of cases with 3 to 5 chromosomal aberrations predicted by routine or WGS revealed similar hazard ratios that were positively correlated with the number of aberrations. For comparing SNV calls from routine and WGS, we investigated genes with validated mutations in at least ten patients in routine diagnostics. This criterion was met by ATM, MYD88, NOTCH1, SF3B1, and TP53. We observed high concordance between routine and WGS data ( ATM 99.4%, MYD88 100%, NOTCH1 97.8%, SF3B1 100%, TP53 99.1%). Five mutations were detected in routine only and three in WGS only (routine only/WGS only: ATM 1/0, MYD88 0/0, NOTCH1 2/2, SF3B1 0/0, TP53 2/1) and found to be caused by low variant allele frequency (5 miscalls, VAF < 5%) and/or filtering out of low quality variants (5 cases). IGHV rearrangements analyzed by Sanger (248 patients), WGS (242/248 patients), and WTS (236/248 patients) data were compared for matches in the rearranged IGHV. IGHVms calling employed 98% of sequence identity as cutoff. Calling the rearranged IGHV was concordant in 93%, 95%, and 92% when comparing Sanger/MiXCR, Sanger/IgCaller, and MiXCR/IgCaller calls, respectively. For IGHVms, the corresponding percentages of concordance were 91%, 86%, and 83%. Mismatches in calling the IGHVms were found in borderline cases. Detection of stereotyped IGHV was fully concordant. In addition to reliably replicating routine results, we find that WGS/WTS data can also outperform routine procedures. In our hands, the highest hazard ratio (4.94) in survival analysis distinguishing mutated CLL (mCLL) from unmutated CLL (uCLL) is observed using WTS for IGHVms determination and 97% sequence identity as cutoff (208 patients, up to 30 years follow up). Survival for borderline cases (97.00-97.99% sequence identity) was statistically indistinguishable from uCLL in routine diagnostics as well as WGS/WTS. Furthermore, using WGS, CN-LOH regions and deletions below the detection limit of CBA (10 MB) can be detected. We observed such regions in samples with mutated TP53 (5 CN-LOH), ATM (2 CN-LOH, 4 deletions < 10MB), and deletions of minimally overlapping region chr8:21,700,000-23,600,000, whose loss was associated with poor survival (3/29 deletions < 3.1 MB). Conclusions: WGS/WTS data reliably replicate results obtained by standard procedures in a diagnostic setting. Sequencing offers advantages such as a unified workflow, better sensitivity for small deletions, CN-LOH detection, genome-wide analyses of coding and non-coding mutations, among others. Challenges of short read based sequencing regard the detection limit for point mutations and specificity/sensitivity in detecting chromosomal aberrations.

Acute Myeloid Leukemia (AML) Patients with Myelodysplasia (MDS) Related-Gene Mutations in First Complete Remission (CR1) Found to Have Benefit from Allogeneic Hematopoietic Stem Cell Transplant (HCT)

Introduction: Acute myeloid leukemia (AML) with myelodysplasia-related gene mutations is a new disease entity introduced in the International Consensus Classification of Myeloid Neoplasms 2022 (ICC 2022). Myelodysplasia-related gene mutations include ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1, or ZRSR2 . Allogeneic hematopoietic stem cell transplant (HCT) represents a curative option for subgroups of AML patients, but not for all. In a prior study (Song et al.) reported the survival benefit of allogeneic HCT in 59 patients with AML with secondary type mutations (STM) including SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, and STAG2. Allogeneic HCT could abrogate adverse prognosis in AML with STMs, and reduce its mortality risk by 58% (Song et al. BMT 2022). In the present study, we aim to evaluate the frequency of this new disease entity of AML with MDS-related mutations and evaluate the prognostic impact of allogeneic HCT in patients with MDS-related mutations achieving CR1. For comparison of the benefit from HCT, the mantel-Byar test was applied to take into account time-dependent covariate of HCT in CR1 to avoid immortal bias. Patients and Methods: We conducted a retrospective study that included 1,228 adult patients who were diagnosed with AML from January 1997 to January 2020and achieved CR1 in 5 centers (Toronto, Vancouver, and Montreal, Canada; Hwasun and Seoul, South Korea). Next-generation sequencing was performed as their standard clinical practice using targeted sequencing panel in each institution. As defined in ICC 2022, AML with myelodysplasia-related gene mutations was defined as the case of AML with ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1, or ZRSR2 mutation(s). We considered allogeneic HCT in CR1 as a time-dependent covariate and used a Mantel-Byar test to avoid immortal bias. We identified 278 pts (22.6%) who had at least one MDS-related gene mutation at diagnosis. Relapse-free survival (RFS) at 2 years was selected for the primary endpoint, defined as the time from CR1 to relapse or death from any cause. EZR software version 1.41 was used throughout the study for statistical analysis. Results: The median follow-up duration among survivors after achieving CR1 was 16.5 months (mo) (Range: 1.3-217.9 mo). Between the HCT vs non-HCT group, there were no differences in sex, cytogenetic risk or other somatic mutation except NPM1 mutation which was more frequently detected in non-HCT group (p <0.001). As expected, younger pts were more frequently noted in HCT group, p=0.001). First, considering HCT as non-time dependent covariate, we have compared RFS and OS between HCT vs non-HCT groups. The 1 year RFS rate was superior in the HCT group (77.7%) to that in non-HCT group (49.9%, p<0.0001; Fig.1A). A similar trend was noted for the 1-year OS rate presenting higher OS rate in the HCT group (83.8%) than in the non-HCT group (72.7%, p=0.0003). Next, when HCT in CR1 was taken into account as a time-dependent (td) covariate, Mantel-Byar analysis showed a consistent finding of superior RFS in HCT group to non-HCT group: time dependent covariate of HCT in CR1 was found to be favorable (Hazard ratio [(HR) of 0.45, (95% C.I. [0.31-0.65]; p=0.00003)] as shown in Fig.1B which is a Simon-Makuch plot. In multivariate analysis (MVA) for RFS, time-dependent covariate of HCT in CR1 retained its statistical significance, thus concluding that HCT in CR1 is an independent prognostic factor for RFS (HR: 0.44, 95% CI [0.29-0.68], p =0.0002) together with other variables found to be adverse on RFS including monosomal karyotype (p=0.03), IDH1 mutation (p=0.01), and age > 60 years (p=0.002); while CEBPA (p=0.02) and NPM1 (p=0.049) mutations were found to be favorable on RFS in MVA. Conclusion: In our cohort of 1228 AML patients, we have identified 278 AML patients with myelodysplasia-related gene mutations (22.6%). We have confirmed that allogeneic HCT in CR1 is strongly recommended in a subgroup of AML with myelodysplasia-related gene mutations based on the favorable impact of HCT in CR1 on RFS from the present study. This result supports the urgent referral of this patient group for allogeneic HCT in CR1.

Parallel Genomic Analysis from Paired Bone Marrow and Peripheral Blood Samples of 200 Cytopenic Patients

Background: Analysis of bone marrow is the gold standard in patients with unexplainable cytopenia for diagnosing (or excluding) myelodysplastic neoplasm (MDS). Besides morphology also molecular genetics and cytogenetics are currently assessed from BM as both are required for diagnosis and risk stratification. Substitution of BM analysis with peripheral blood (PB) analysis would be a substantial progress especially in the elderly to allow for a less invasive strategy. Aims: Validate, whether PB reliably reflects molecular and cytogenetic aberrations of BM by evaluating somatic mutations and copy number variations (CNVs) in concurrent PB and BM samples. Methods: We analyzed paired BM and PB samples (time difference of sampling: <15 days) of 200 cytopenic patients. All samples were subjected to targeted panel sequencing analyzing mutations (assessed as positive at variant allelic frequencies/VAFs >1%) in 40 genes associated with myeloid neoplasms (coverage: 1500x). If available, BM chromosome analysis (CA) and BM FISH analysis were compared to NGS CNV analysis from PB using a human CNV backbone hybridization panel (coverage: 50x). Results: In 69 patients (35%) no mutations (MUTs) were detected in either BM or PB (Figure 1A). For the remaining 131 patients at least one MUT was found in BM (median MUTs/patient: 2; range 1-10). In 130/131 patients MUTs were detected in both BM and PB (same number of MUTs: n=121; higher number of MUTs in BM than in PB: n=9) while in only one patient harboring a DNMT3A MUT as the sole MUT in BM (VAF: 9%) no MUT was detected in PB. This resulted in a negative predictive value of 98.6% (69/70). Overall, we detected 364 MUTs in 29 of the analyzed genes (BM MUTs: n=362; PB MUTs: n=349) and a concordance between BM and PB of 95% (347/364; Figure 1A). The median VAF was lower in PB compared to BM across all MUTs (median VAF: 20% vs. 12%; p<0.001). All MUTs detected in BM at VAF ≥13% (n=229) were also detected in corresponding PB. Next, we compared CNVs detected by PB NGS to BM karyotype (KT). For 119/200 patients PB NGS including a CNV-backbone panel was available. BM CA revealed a normal KT in 71 and an aberrant KT in 48 patients. In total, 126 aberrations were detected in BM CA including 7 balanced aberrations that can technically not be detected using the CNV backbone panel. CNV analysis from PB revealed 68 aberrations (of 119 detectable BM CNVs; 54%) with 64 being concordant with BM CA. In total, in 88/119 patients (74%) concordant results were found in BM and PB (70/71 with normal and 18/48 with aberrant BM KT). In 3 patients (2.5%) different CNVs were detected (1/3: no CNV in BM CA), while in 28 (23.5%) either fewer CNVs (n=8; all complex KTs) or no CNVs (n=20) were detected in PB compared to BM. The majority (76%) of BM CNVs not detected by PB NGS with available BM FISH data were small BM clones (clone size ≤30%) and therefore not reliably captured by NGS PB with a technical sensitivity of 30%. Interestingly, 4 of the discordant patients with aberrant BM KT did not harbor any somatic MUT in BM and PB. Regarding the IPSS-R cytogenetic risk group, 95 (80%) patients fell into the same category, while 22 (18%; all with aberrant BM KT) were assigned to a better risk category by NGS PB than based on BM CA data and 2 patients (2%) to a worse risk category by NGS PB (Figure 1B). Conclusion: In cytopenic patients mutations can be reliably detected from PB showing a concordance with BM of individual mutations of 95% and a consistency regarding overall mutation detection between PB and BM in 99.5% of patients. Cytogenetic risk group assignment based on NGS CNV analysis from PB compared to chromosome analysis from BM resulted in the same risk group in 80% of patients. However, to further decrease the number of discrepant cases, the technical limitations (sensitivity, inability to detect balanced CNVs) of the CNV backbone method might be overcome by increasing the sequencing depth and/or using CD34-enriched PB samples and adding probe sets. Alternatively, FISH on PB smears could be used for CNV detection. Overall, our data indicate a high degree of overlap between PB and BM regarding mutational and cytogenetic analysis in patients with unexplained cytopenia and suggest PB as reliable surrogate especially in patients without any genetic aberrations.

The Presence of Somatic Mutations on Specific Pathways Such As Chromatin Modifier, Spliceosome, and Myeloid Transcription Factor at Diagnosis Was Found to Have Benefit from Allogeneic Hematopoietic Stem Cell Transplantation (HCT) in 1,228 Patients with Acute Myeloid Leukemia in First Complete Remission (CR1)

Introduction Allogeneic stem cell transplant (HCT) brings survival benefit to certain subgroups of acute myeloid leukemia (AML) patients, although it may not be suitable for all cases. Decision for HCT referral is based on the ELN risk grouping, which classifies pts into favorable, intermediate, and adverse risk categories. While HCT is recommended for pts in the intermediate and adverse risk, those in the favorable risk are deferred to receive HCT. However, currently, there is a lack of data directly demonstrating benefits of HCT in specific mutation groups. Patients and Method In this retrospective study, out of 1477 newly diagnosed AML pts 1,228 pts achieved CR1 and had undergone baseline next-generation sequencing (NGS) at diagnosis from January 1997 to January 2020. The data were collected from five different centers located in Toronto, Vancouver, and Montreal, Canada, as well as Hwasun and Seoul, South Korea. NGS was performed in each institution with each institution's standard targeted sequencing panel. To assess the effectiveness of HCT, particular attention will be given to the time from achieving complete remission 1 (CR1) to HCT. To handle immortal time bias, the Mantel-Byar method was used to depict outcomes during the follow-up period. Kaplan-Meier curves were avoided as they can introduce immortal bias, considering the period of time before treatment as part of the treatment arm. Instead, the MBT and landmark methods were employed to address this issue. Allogeneic HCT was treated as a time-dependent variable to prevent immortal bias. Gene mutations were grouped into 6 biologic pathways, including DNA methylation, chromatin modifiers, cohesion complex, activated signaling, tumor suppressor, spliceosome, and myeloid transcription factor (TF) mutations. In the multivariate (MVA) analysis, any mutations with a frequency of >5% were included as well as TP53 which has very grave prognosis. The primary objective of this study is to identify specific groups of AML pts who can derive significant benefits from receiving HCT in CR1, treating HCT as a time-dependent (td) covariate in the analysis. Statistical analyses were conducted using EZR version 1.41. Results The median duration of follow-up among survivors after achieving CR1 was 22.6 months (range: 1-217.9 months). Among the 593 pts, who underwent allogeneic HCT in CR1, the HCT group was composed of younger pts, with a median age of 50 years compared to 56 years in the non-HCT group (p < 0.001). Also, there were fewer pts with NPM1 mutation in the HCT group, accounting for 20.7% compared to 32.4% in the non-HCT group (p < 0.001). Apart from these characteristics, the two groups were similar in other parameters. The 2-year overall survival (OS) was higher in the HCT group, at 66.4%, compared to 54.6% in the non-HCT group (p < 0.0001). Similarly, the 2-year RFS was higher in the HCT group, with a rate of 64.8%, compared to 52.1% in the non-HCT group (p =1.82e-15). Furthermore, HCT as a time-dependent covariate was found to be favorable (Hazard ratio [HR] 0.76, 95% C.I. [0.63-0.91]; p=0.003), as demonstrated in the Simon-Makuch plot (p = 0.003). In the MVA, HCT as a time-dependent covariate retained its favorable impact on RFS (HR 0.67, [0.56-0.82]; p<0.0001). The MVA showed several prognostic factors besides HCT such as NPM1 and CEBPA mutations as favorable factors for RFS, being classified under the European Leukemia Network (ELN) favorable risk group. Conversely, FLT3-ITD, DNMT3A, IDH1, KIT, and TP53 mutations, monosomal karyotype, age > 60 years, and white blood count (WBC) at diagnosis > 55,600/mm^3 were associated with worse RFS. In terms of RFS, the greatest benefit of HCT was observed in the pts having mutations in ASXL1 (HR 0.34 [0.15-0.78]), BCOR (HR 0.29 [0.14-0.61]), IDH1/2 (HR 0.62 [0.42-0.92]), and RUNX1 (HR 0.17 [0.08-0.40]) but no significant benefit in those with KRAS (HR 1.59), NRAS (HR 1.21), PTPN11 (HR 1.23), and WT1 (HR 1.30). According to the biological pathway of somatic mutations, the greatest benefit of HCT in terms of RFS was observed in pts with mutations in the chromatin modifier (HR 0.37 [0.22-0.62]), spliceosome (HR 0.32 [0.16-0.65]), and myeloid TF pathway (HR 0.52 [0.34-0.79]). Conclusion Allogeneic HCT exerts a greater benefit on RFS in AML pts with mutations in the chromatin modifier, spliceosome, and myeloid TF pathways. Further study is warranted to replicate this finding.

NKX101, an Allogeneic Off-the-Shelf CAR NK Therapy Targeting NKG2D-Ls, Has Potent In Vitro Cytotoxicity Against Patient-Derived AML Leukemic Stem Cell (LSC) and Non-Leukemic Stem Cell (Non-LSC) Blasts

Introduction Acute myeloid leukemia (AML) is a malignancy of immature myeloid cells characterized by rapid proliferation of abnormal myeloblasts. Treatment options for patients with AML are limited, especially for relapsed/refractory (r/r) disease. Leukemic cells with stem cell features, known as leukemic stem cells (LSCs), have been implicated as the origin of relapse in minimal residual disease positive (MRD +) AML (Khaldoyanidi et al. 2022, Kamel et al. 2022). Additionally, absence of NKG2D-ligand (NKG2D-L) expression has been reported to be associated with stemness and the LSC population in AML (Paczulla et. al, 2019). NKX101 is an allogeneic, off-the-shelf, healthy donor-derived chimeric antigen receptor (CAR) NK cell therapy candidate engineered to express an NKG2D-L targeting CAR as well membrane bound IL-15. In a clinical trial, NKX101 has shown promise for the treatment of r/r AML, including generating complete responses (CR) with MRD negativity (MRD -). In this study, we utilize a flow cytometry-based approach to (i) evaluate NKG2D-L cell surface expression on normal hematopoietic cells vs. primary AML blasts, (ii) assess the expression pattern of NKG2D-Ls on AML blast subsets, and (iii) determine the potency of NKX101 cytotoxic killing of both LSC and non-LSC AML blasts. Methods NKX101 cells were generated from peripheral blood leukopaks from healthy donors. Cryopreserved bone marrow mononuclear cells (BMMCs) were obtained from AML patients in accordance with approved IRB protocols (n=20). BMMCs from healthy donors were used as controls (n=10). AML somatic mutation status was determined using an Illumina TruSight® Myeloid targeted sequencing panel. NKX101 cytotoxicity against patient derived AML blasts was assessed using a flow cytometry-based assay that simultaneously measured specific cytotoxicity AML blast/LSC markers. NKG2D-L protein expression on LSC and non-LSC subsets was determined using flow cytometry with fluorescently labeled monoclonal antibodies specific to MICA, MICB, and ULBP1-6. Results Utilizing patient-derived primary samples, we demonstrate that NKG2D-Ls are expressed on AML blast cells and expressed at higher levels on AML blast cells compared to healthy normal hematopoietic cells (p = 0.002). Furthermore, we show that NKG2D-Ls are expressed at similar levels on both LSC and non-LSC AML blast populations. Target cell specific flow cytometry-based cytotoxicity assays revealed that NKX101 cells potently kill patient-derived AML blasts in a dose-dependent manner. Moreover, NKX101 cells had similar cytotoxic potency on both LSC and non-LSC AML blasts subsets. Conclusions In summary, we show that NKG2D-Ls are expressed on the surface of AML blasts at levels above those observed in normal hematopoietic cells. NKG2D-Ls are expressed on both non-LSC AML blasts and on the LSC population implicated as the origin of relapse in AML. Additionally, the ability of NKX101 to potently kill AML LSC blasts suggests that NKG2D-L targeting may be a viable mechanism for eliminating LSC in the blood and bone marrow of patients with r/r AML. Taken together, these data support NKG2D-Ls as promising therapeutic targets for AML and support further investigation of NKX101 CAR NK cell therapy for the treatment of r/r AML.

The Genomic and Transcriptomic Landscape of Myeloid Sarcoma and Associated Acute Myeloid Leukemia

Background: Myeloid sarcoma (MS) is a distinct form of acute myeloid leukemia (AML) found in ~10% of patients, in which mass-forming myeloid blasts proliferate in extramedullary sites. MS may present as an isolated lesion as the primary manifestation of disease, co-occur with AML, or, more commonly, arise in the setting of AML relapse. It is associated with poor outcomes, and treatment options are sparse. While increased knowledge of AML and its biology have improved the care and survival of AML patients, those with MS continue to have limited successful treatment options, as MS has typically not been included in AML genetic and genomic profiling efforts, and patients are excluded from clinical trials. Leukemia cutis (LC) is a form of extramedullary AML characterized by cutaneous involvement, often with an infiltrative and non-mass forming pattern of growth. Still, historically the literature has not made a clear distinction between MS and LC. Although recent small pilot studies have suggested differences in AML-associated mutations between medullary AML and MS, knowledge of the complete molecular and cellular composition of MS and comprehensive comparison with AML is lacking. While molecular characterization of the extramedullary disease site is encouraged in the updated clinical recommendations, no definitive guidance exists for changes in the management and care of AML patients with MS, given the substantial gap in our current knowledge and understanding of extramedullary AML. Methods: Here, we performed RNA-Seq (n=96), single-cell (sc) RNA-Seq (n=33), and mutation profiling using either whole exome sequencing (WES, n=6) or a 409-gene targeted sequencing panel (n=48) of MS or LC patient samples and bone marrow (BM) aspirates from AML patients with and without extramedullary disease to present the first comprehensive characterization of MS and LC. Results: Our WES data showed thatMS and LC evolve independently in the extramedullary site from concurrent medullary disease with distinct dominant clones, characterized by higher median tumor mutation burden (1.66 vs. 064 mutations per megabase, respectively; p=0.0059) and increased copy number aberrations, which both validated in an MS Tet2 HR mouse model. Notably, we observed location-specific molecular evolution in patients with multiple concurrent MS sites (Figure 1A) and patients with recurring MS in WES and targeted panel data. While targetable mutations (i.e., IDH1/2 mutations and FLT3-ITD/TKD) occur, the inter-site variability of molecular aberrations may explain the short-lived responses of AML patients with MS to IDH/FLT3 inhibitors. Transcriptional profile differences between MS and associated medullary disease included up-regulation of extracellular matrix organization and cell-substrate adhesion and down-regulation of inflammatory response pathways in both bulk and scRNA-seq data (Figure 1B). Further, involved BM malignant cells of MS patients compared to patients without MS up-regulated specific pathways, including interferon-gamma response observed in both bulk and scRNA-seq, as well as chromatin organization and splicing. We further revealed that MS patients without overt BM involvement show remodeling of the BM immune microenvironment, with increased fractions of hematopoietic stem and progenitor cells that down-regulate HLA class II genes. Comparing LC and MS in the skin revealed distinct differences in their expression profiles, including high expression of genes involved in cell migration and immune response in LC, suggesting LC and MS represent distinct disease entities. Lastly, we demonstrated high expression of BCL2 in extramedullary disease (p=0.0074) and presented initial evidence that treatment with BH3-mimetics may be beneficial for the treatment of isolated MS. Conclusions: Myeloid sarcoma shows distinct molecular evolution and microenvironmental composition compared to medullary disease, including location-specific evolution necessitating personalized management consideration. Leukemia cutis is distinct from MS tumorous lesions and should be considered a separate entity. High BCL2 expression in extramedullary disease might represent a promising targetable vulnerability in patients with isolated MS.

Targeted NGS Panel Guides Risk-Adapted Treatment Intention in Newly Diagnosed Myeloma Patients

Introduction While overall survival has improved, patients with high-risk multiple myeloma (MM) have poor outcomes. Fluorescent in situ hybridisation (FISH) is widely used to identify high-risk genomic features at diagnosis, but a significant minority of patients classified with standard-risk disease progress rapidly and in retrospect may have warranted more intense frontline therapy. Recently, we implemented the Myeloma Genome Project Panel (MGPP) into clinical workflows - a myeloma-specific targeted sequencing panel capturing 228 genes, 6 translocation regions and 56 copy number abnormalities (CNAs) in a single assay [PMID: 35522533]. Given that MGPP captures many prognostic regions beyond the limited number assessed by FISH, there is potential to better identify high-risk patients. However, the utility of MGPP in supporting clinical risk designation and therapy choice when compared with FISH in newly diagnosed MM is unknown. Methods A prospective cohort study was conducted to compare the impact of MGPP vs. FISH on baseline risk stratification and treatment intention, in 55 unselected newly diagnosed MM patients. For MGPP, libraries were generated from 100ng DNA from bone marrow (BM) derived CD138 + MM cells and matched germline peripheral blood (Roche KAPA HyperCap Target Enrichment/Hyperplus kit). Reads were aligned to hg38, and somatic mutations, CNAs and translocations called with a custom pipeline. ISO-accredited FISH analysis was performed in NHS laboratories. In a multicentre multidisciplinary team (MDT) meeting, four blinded haematologists independently reviewed each clinical case presented twice with either FISH or MGPP, in a random order. Clinicians designated risk status and treatment intentions using a structured proforma. Therapeutically actionable variants could be selected as per the MyDRUG trial [ClinicalTrials.gov: NCT03732703]. A consensus MDT outcome for high-risk or more intensive therapy required at least 3 out of 4 concordant decisions. Results Data is presented here for 23 of 55 patients. The mean age was 68.9 years, 57% were transplant eligible, with R-ISS-1 (35%), R-ISS-2 (26%) and R-ISS-3 (39%). Firstly, MGPP had greater technical success than FISH, reporting successfully in 100% vs. 83% cases from the same diagnostic BM samples respectively (n=4 assay failures with FISH). Secondly, MGPP reported more genomic features than FISH, detecting 231 vs. 18 somatic variants at 127 vs. 7 regions, respectively. Thirdly, MGPP captured equivalent loci at greater resolution, detecting subclonal CNAs missed by FISH (Figure 1). In a head-to-head comparison with FISH, these additional genomic insights from MGPP influenced clinical treatment decisions. A clinician-led MDT reclassified 9 (39%) patients designated as standard-risk by FISH to high-risk status when presented with MGPP (Table 1). Reclassified patients were enriched for prognostically significant MYC translocations (17%) and subclonal 17p loss (22%), 1p loss (17%) and 1q gain (13%) CNAs that were missed by FISH. Altered risk designation influenced frontline treatment intention; clinicians opted for intensification of induction therapy in 13% (FISH) vs. 48% (MGPP) high-risk patients, and considered tandem autologous transplant as an option in 0% (FISH) vs. 23% (MGPP) transplant-eligible patients. FISH identified a therapeutically actionable variant in 3 (13%) patients [venetoclax, n=3]. MGPP identified 13 (57%) additional patients with opportunity for targeted therapy [MEK inhibitor, n=11; BRAF V600E inhibitor, n=2], which the MDT considered as a treatment option in all cases if available in trial. The full dataset and analysis will be presented at time of the conference. Conclusions We demonstrate that MGPP impacts clinical risk designation and treatment intention in newly diagnosed MM. MGPP had greater technical assay success than FISH and captured a greater range and depth of somatic variants. Importantly, these additional insights led a clinical MDT to reclassify 39% patients as high-risk, consider intensification regimen in 39% cases and potential for actionable targeted therapy in 57% patients. Our results support MGPP as a robust alternative to FISH in supporting risk-adapted treatment decisions. We will implement MGPP in additional large prospective cohorts over the next year, to better understand its prognostic utility and feasibility in routine clinical workflows.

Residual circulating tumor DNA after adjuvant chemotherapy effectively predicts recurrence of stage II-III gastric cancer.

Circulating tumor DNA (ctDNA) is a promising biomarker for predicting relapse in multiple solid cancers. However, the predictive value of ctDNA for disease recurrence remains indefinite in locoregional gastric cancer (GC). Here, we aimed to evaluate the predictive value of ctDNA in this context. From 2016 to 2019, 100 patients with stage II/III resectable GC were recruited in this prospective cohort study (NCT02887612). Primary tumors were collected during surgical resection, and plasma samples were collected perioperatively and within 3 months after adjuvant chemotherapy (ACT). Somatic variants were captured via a targeted sequencing panel of 425 cancer-related genes. The plasma was defined as ctDNA-positive only if one or more variants detected in the plasma were presented in at least 2% of the primary tumors. Compared with ctDNA-negative patients, patients with positive postoperative ctDNA had moderately higher risk of recurrence [hazard ratio (HR) = 2.74, 95% confidence interval (CI) = 1.37-5.48; P = 0.003], while patients with positive post-ACT ctDNA showed remarkably higher risk (HR = 14.99, 95% CI = 3.08-72.96; P < 0.001). Multivariate analyses indicated that both postoperative and post-ACT ctDNA positivity were independent predictors of recurrence-free survival (RFS). Moreover, post-ACT ctDNA achieved better predictive performance (sensitivity, 77.8%; specificity, 90.6%) than both postoperative ctDNA and serial cancer antigen. A comprehensive model incorporating ctDNA for recurrence risk prediction showed a higher C-index (0.78; 95% CI = 0.71-0.84) than the model without ctDNA (0.71; 95% CI = 0.64-0.79; P = 0.009). Residual ctDNA after ACT effectively predicts high recurrence risk in stage II/III GC, and the combination of tissue-based and circulating tumor features could achieve better risk prediction.