Pre-emptive Detection and Evolution of Relapse in Acute Myeloid Leukemia by Flow Cytometric Measurable Residual Disease Surveillance

Pre-emptive Detection and Evolution of Relapse in Acute Myeloid Leukemia by Flow Cytometric Measurable Residual Disease Surveillance | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Pre-emptive Detection and Evolution of Relapse in Acute Myeloid Leukemia by Flow Cytometric Measurable Residual Disease Surveillance Sylvie Freeman, Nicholas McCarthy, Gege Gui, Florent Dumezy, Christophe Roumier, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3978470/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 18 Jun, 2024 Read the published version in Leukemia → Version 1 posted 9 You are reading this latest preprint version Abstract Measurable residual disease (MRD) surveillance in acute myeloid leukemia (AML) may identify patients destined for relapse and thus provide the option of pre-emptive therapy to improve their outcome. Whilst flow cytometric MRD (Flow-MRD) can be applied to high-risk AML/ myelodysplasia patients, its diagnostic performance for detecting impending relapse is unknown. We evaluated this in a cohort comprising 136 true positives (bone marrows preceding relapse by a median of 2.45 months) and 121 true negatives (bone marrows during sustained remission). At an optimal Flow-MRD threshold of 0.045%, clinical sensitivity and specificity for relapse was 73% and 89% respectively (51% and 98% for Flow-MRD ≥ 0.1%) by ‘different-from-normal’ analysis. Median relapse kinetics were 0.78 log 10 /month but significantly higher at 0.92 log 10 /month for FLT3 -mutated AML. Computational (unsupervised) Flow-MRD (C-Flow-MRD) generated optimal MRD thresholds of 0.036% and 0.082% with equivalent clinical sensitivity to standard analysis. C-Flow-MRD-identified aberrancies in HLADRlow or CD34 + CD38low (LSC-type) subpopulations contributed the greatest clinical accuracy (54% sensitivity, 93% specificity) and notably, by longitudinal profiling expanded rapidly within blasts in > 40% of 86 paired MRD and relapse samples. In conclusion, flow MRD surveillance can detect MRD relapse in high risk AML and its evaluation may be enhanced by computational analysis. Health sciences/Diseases/Haematological diseases/Haematological cancer/Leukaemia Health sciences/Risk factors Health sciences/Medical research Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Detection of impending relapse in acute leukemias allows selection of patients for pre-emptive therapies that may avoid ultimate treatment failure but also the challenges of cytoreduction and morbidities following hematological relapse. In acute myeloid leukemia, the evidence that sequential MRD testing can provide an early warning of relapse comes predominantly from studies applying real-time quantitative polymerase chain reaction (RT-qPCR) assays( 1 , 2 ). Most of this data is for common PCR MRD leukemia targets PML-RARA , CBFB-MYH11 , RUNX1-RUNX1T1 , and NPM1 mutations although the relapse kinetics of other PCR targets, namely rarer rearrangements and overexpressed WTI have also been examined( 3 – 5 ). Based on the findings of these studies, definitions of treatment failure now include molecular MRD relapse (increase of MRD copy numbers ≥1 log 10 or conversion from MRD negativity to MRD positivity, confirmed in a second sample) and RT-qPCR surveillance for MRD relapse after treatment (3 monthly if bone marrow sampling) is recommended for CBF and NPM1 mutated AML( 2 ). The term MRD relapse now also incorporates conversion from MRD negativity to positivity detected by other methods including flow cytometry and is thus applicable across AML subtypes( 2 ). There remain however unresolved issues that limit the information from sequential MRD results and therefore the clinical benefit and opportunities from MRD detection of impending relapse. These in part arise from the challenges of determining clinical specificity and sensitivity of a result to predict relapse when applying the recommended monitoring schedule. Conversion from MRD negativity to low level MRD positivity may have low predictive accuracy for relapse (low specificity), even when the sample MRD target is a high risk leukemic transcript such as DEK-NUP214 ( 5 ). Although a second consecutive sample showing rising MRD levels can confirm MRD relapse, this risks reducing the interval for treatment decisions in the context of often unpredictable relapse kinetics. Furthermore, the majority of hematological relapses may not be identified by a preceding MRD positive test (low clinical sensitivity) despite using the recommended 3 monthly bone marrow monitoring schedules with established highly sensitive qPCR assays such as applied to CBF AML patients( 6 ). Additionally, as the use of flow cytometry or NGS for surveillance of impending relapse remains exploratory in AML( 2 , 7 ), there is an unmet need to extend MRD relapse detection to all AML patients. Moreover, parallel molecular and flow cytometric serial surveillance may be necessary for some patients such as when there is a risk of clonal evolution( 8 ) including from treatment escape with loss of the molecular MRD target( 9 ). Although flow cytometric MRD has the advantage of rapid turn-around time in addition to a sensitivity of at least 10 − 4 , the current dependence on manual analysis of bidimensional plots can lead to inconsistent quantitation according to expertise( 10 ) and also limit the potential of deeper-immunophenotyping for further optimization. We recently demonstrated the prognostic value of flow cytometric MRD monitoring performed during the first year following allogeneic stem cell transplantation( 11 ). This study now evaluates the clinical sensitivity and specificity of flow cytometric bone marrow MRD surveillance for early relapse prediction. We compare standard MRD detection with a newly developed computational analysis approach, investigating assay performance and aberrant immunophenotypic populations that are most specific for imminent relapse. Methods Samples and Patients Patient samples were from AML patients > 18 years old followed for flow cytometric measurable residual disease (MRD) detection using standard published methods( 2 , 12 – 14 ) by a single reference laboratory (Birmingham, UK) from April 2015 to September 2022. MRD bone marrow monitoring was conducted after completion of chemotherapy or post allogeneic hematopoietic stem cell transplantation (HSCT) in patients who had achieved a complete remission (CR) or CR with incomplete hematologic recovery (CRi). MRD detection analyses were conducted with informed patient consent in accordance with the Declaration of Helsinki and subject to appropriate ethics committee approval. MRD bone marrow samples preceding a hematological relapse were retrospectively identified and included if a paired relapse sample had been received up to 4 months after the MRD sample and the diagnosis of relapse was made subsequent to this relapse sample. AML MRD monitoring control samples were defined as 1) post-treatment PCR negative bone marrows from patients with acute promyelocytic leukemia (APML) maintaining molecular remission (test control cohort) and 2) post treatment / HSCT bone marrows from non-favourable risk AML patients in continuous remission > 6 months after MRD sample without a treatment intervention (validation control cohort). Additionally, pre-transplant bone marrow files from 156 patients entered into the FIGARO trial with previously reported standard flow cytometric MRD results( 14 ) were included for clinical validation of computational analysis. Flow cytometric MRD testing and analysis MRD was assessed by flow cytometry as previously described in a central reference laboratory) ( 13 , 14 ) Details on sample logistics, processing, and analysis strategy are provided in the Supplementary Methods. Routine flow cytometric MRD analysis was performed using a standardized manual gating strategy that screened blasts for different-from-normal (DfN) aberrant immunophenotypes that were established as useful and frequently observed leukemic aberrant immunophenotypes (LAIPs) and also for any previously identified baseline LAIPs when available. Samples were reported as MRD-negative if no baseline and/or different-from-normal LAIP cells could be quantitated above assay threshold (of 0.05%). To confirm the results of standard analysis, and exclude variability arising from subjective interpretation, flow cytometry standard files from MRD testing were analyzed using a computational approach (C-Flow MRD), updated following previous clinical evaluation( 11 , 14 ). Blast cells (CD117+/CD34+) from test samples were clustered together with a 40–50 control BM reference set using the FlowSOM clustering algorithm. Automated decision tree analysis was then applied to define abnormal blasts with an immunophenotype significantly different from the reference set in 7-dimensional space (light scatter and CD45 parameters excluded). C-Flow-MRD results were calculated by summating discrete abnormal blast populations above the limit of detection (LOD), and the assay result reported as the highest value of the two tubes of the AML MRD antibody panel. Analyses of specific progenitor compartments for C-Flow-MRD + blast cells were performed in FlowJo software through progenitor pre-set sub-gating based on optimised thresholds. Statistical analyses Receiver operating curve (ROC) statistics with area under the curve (AUC) were generated for MRD results of the clinical cohorts to summarise the discrimination ability of testing to predict relapse, with values of > 0.75 considered as good. Clinical specificity (true negative results (TN) ÷ [TN + false positives (FP)] x 100), sensitivity (true positive results (TP) ÷ [TP + false negatives (FN)] x 100), balanced accuracy ([%sensitivity + %specificity] ÷ 2) and Youden index ([sensitivity + specificity] – 1) were determined for test performance at specific assay cut-points. Optimal assay cut-points were derived from peaks in the Youden Index or alternatively by the R-based MaxStat package( 15 ), which uses maximally selected rank statistics. Cumulative incidence of relapse (CIR) and treatment related mortality (TRM) from C-Flow-MRD applied to the previously published Figaro pre-transplant MRD sample dataset( 14 ) were calculated using the ‘cumulative incidence of competing events and Gray test analysis’ and ‘Fine-Gray proportional hazard regression for competing events’ functions of the EZR software package v1.61( 16 ). Further details are in Supplementary Methods. Results Flow cytometric MRD performance for detection of relapse during monitoring To evaluate the diagnostic performance of routine flow cytometry MRD for detection of impending relapse in a clinical setting, we interrogated the flow cytometric data of 136 MRD bone marrows preceding a paired relapse sample ( ≤ 4 months, median interval 2.45 months) from 118 patients who experienced relapse during longitudinal flow cytometric MRD monitoring (Table 1 ). MRD results were defined by MRD analysis that did not require a diagnostic sample (‘different-from-normal’ / DFN approach, Methods). Table 1 Characteristics by standard flow cytometric MRD status in pre-relapse samples All n = 136 Flow MRD negative n = 47 Flow MRD positive All n = 89 < 0.1% n = 17 ≥ 0.1% n = 72 Interval between MRD sample and relapse Months, median [IQR] 2.45 [1.74–3.08] 2.77 [2.30–3.43] 2.27 [1.43–2.80] 2.80 [2.07–3.43] 2.12 [1.39–2.71] Diagnostic Genetics Adverse cytogenetics or TP53 mutated 46 (35.7%) 14 (34.1%) 32 (36.4%) 9 (52.9%) 23 (32.4%) FLT3 mutated 29 (22.5%) 13 (31.7%) 16 (18.2%) 4 (23.5%) 12 (16.9%) MDS-related mutations 22 (17.1%) 6 (14.6%) 16 (18.2%) 1 (5.9%) 15 (21.1%) Other NPM1 mutated (FLT3 wild type) 32 (24.8%) 5 8 (19.5%) 1 24 (27.3%) 4 3 (17.6%) 4 21 (29.6%) 0 Unknown 7 6 1 0 1 Treatment Stage Post chemotherapy 48 (35.3%) 10 (21.3%) 38 (42.7%) 5 (29.4%) 33 (45.8%) Post allograft 88 (64.7%) 37 (78.7%) 51 (57.3%) 12 (70.6%) 39 (54.2%) Relapse major blast Immunophenotype CD34+ 109 (80.1%) 44 (93.6%) 65 (73.0%) 12 (70.6%) 53 (73.6%) CD34-CD117+ 21 (15.4%) 3 (6.4%) 18 (20.2%) 5 (29.4%) 13 (18.1%) Monocytic 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) Other 6 (4.4%) 0 (0%) 6 (6.7%) 0 (0%) 6 (8.3%) 89 samples were MRD positive resulting in a diagnostic sensitivity for subsequent relapse in 2 to 4 months of 64% for a result of MRD ≥ 0.05% and 51% for MRD ≥ 0.1%. We then assessed the specificity of a flow cytometric MRD test to predict relapse using two reference sets of MRD bone marrow samples from non-relapsing patients (clinical negatives). In the test set compromising 55 APML follow-up bone marrows (PCR negative, no subsequent relapse), 9% had MRD ≥ 0.05%. A similar frequency of clinical false positives (8%) was observed in the second ‘validation’ reference set, consisting of 66 monitoring MRD bone marrows identified from higher risk AML patients with sustained off-treatment remission > 6 months after the sample, Supplementary Table 1). From the integrated dataset of pre-relapse and all reference samples, MRD test specificity by ‘true negatives’ (no relapse in the absence of intervention > 6 months from test) was 92% for a result of ≥ 0.05%. From receiver operating characteristic curve analysis, area under the curve was 0.84 for identification of patients with imminent relapse by routine flow cytometric MRD, with an optimal MRD clinical cut-off of 0.045% (sensitivity 73%, specificity 89%) in this retrospective cohort (Fig. 1 A). Relapse Kinetics Kinetics of relapse can inform appropriate monitoring interval and choice of intervention but data as yet has been limited to PCR MRD targets. In this large sample cohort from 118 high risk AML patients experiencing relapse the median bone marrow increment was 0.78log 10 /month (IQR, 0.58–1.18), comparable to previous results for NPM1 mutated AML( 17 ). We observed similar results between different high risk genetic subgroups although there was a significantly higher median increment in patients with FLT3 mutations at diagnosis 0.92log 10 /month (IQR, 0.66–1.42) compared to patients with adverse risk cytogenetics or TP53 mutations (P = 0.036) (Fig. 1 B). Evaluation of MRD relapse detection using computational flow cytometric analysis We developed a computational flow cytometric analysis tool incorporating a clustering algorithm and machine learning for standardised multidimensional MRD analysis (C-Flow-MRD) (Methods, Supplementary Methods) that required no baseline sample or expert annotation for MRD quantitation. This was applied to further assess the robustness and accuracy of immunophenotypic aberrancies for relapse prediction. The technical sensitivity of C-Flow-MRD from dilution experiments of different leukemic aberrant immunophenotypes (LAIPs) was below 0.01% for most aberrancies (range 0.002–0.02%) (Supplementary Fig. 1). The previously reported adverse prognostic impact of pretransplant MRD on relapse risk was recently confirmed in patients who entered the FIGARO trial, a randomized trial of reduced-intensity conditioning regimens. The clinical validity of C-Flow-MRD was tested on this cohort of 156 transplanted patients. The prognostic performance of results generated by C-Flow-MRD was equivalent to routine Flow-MRD (Supplementary Fig. 2). Aberrant Immunophenotypes with indeterminate leukemic potential reduce clinical specificity We applied C-Flow-MRD to the integrated dataset of pre-relapse and all reference samples. Results showed an equivalent clinical sensitivity to standard MRD but reduced specificity (Fig. 2 A). C-Flow-MRD detected immunophenotypic aberrancies based on their absence from the training sample set (normal lymphoid neoplasia staging BMs) and this provided a high specificity in the APML ‘non-relapse’ MRD bone marrow samples (82% and 91% at MRD cut-offs of 0.04% and 0.08% respectively). However the clinical false positive rate from C-Flow-MRD increased in the validation non-relapse cohort (specificity of 79% at MRD cut-off of 0.08%) (Supplementary Fig. 3). Since this indicated that some ‘different-from-normal’ blasts immunophenotypes, not represented in our large set of lymphoma staging BMs (minimum 38 per antibody combination), are poor predictors of AML relapse, we extended the C-Flow-MRD training set to include 10 APML control samples with outlier false positive results (detectable above 0.036%). We hypothesised that these false positive results represented candidate aberrant immunophenotypes with indeterminate potential. Excluding these for C-Flow-MRD testing on non-APML monitoring BMs resulted in an improved specificity, that was most apparent for the lower thresholds of MRD detection (74% vs 55% at 0.04% threshold) but maintained comparable sensitivity for relapse (Fig. 2 B). Aberrant immunophenotypes identified by C-Flow-MRD in the outlier false positive APML samples included CD117 + progenitors with higher expression of both CD33 and HLADR or decreased HLADR with relative CD33 overexpression and B lymphoid progenitors with asynchronous weaker CD38 expression (Supplementary Fig. 4). Defining these as ‘false positives’ supports their exclusion by standard DfN analysis. Hierarchy for relapse prediction by MRD is dependent on progenitor compartment Depending on the AML genetic subtype, leukemic cells may be variably distributed between CD34 + and CD34- subpopulations, however data is lacking for how this impacts MRD detection. We examined this by C-Flow-MRD to enable extended, objective phenotypic characterisation of aberrancies between CD34 + and CD34- blasts. In our cohort of high-risk AML patients, clinical sensitivity for relapse by C-Flow-MRD was substantially from CD34 + aberrancies, however for 8 of 90 analysed pre-relapse samples, MRD was only detectable in the CD34-117 + compartment. CD34-CD117 + blast based MRD added up to 9% clinical sensitivity to CD34 + blast based MRD with MRD cut-offs ranging from 0.04–0.1% and had comparable clinical specificity (close to 90% when MRD was > 0.08%). Notably only 3 of 8 patients with CD34-CD117 + restricted C-Flow-MRD had NPM1 mutated AML (all 3 were also FLT3 -ITD co-mutated). All relapse samples from these patients gained CD34 + aberrant populations, although CD34-CD117 + remained the predominant population (> 95% of abnormal blasts) in 6. Relapse-initiating cells may be enriched in certain hematopoietic stem and progenitor (HSPC) compartments associated with functional resistance through leukemic stem cell properties (CD34 + CD38low) or immune evasion (such as HLADR downregulated HSPC). In our cohort that was skewed to post-allogeneic transplant relapses, C-Flow-MRD-detected aberrancies in the HLADRlow compartment provided the greatest effectiveness for clinical sensitivity and specificity (44% and 98% respectively above a cut-off of 0.02%) (Fig. 3 A). Aberrancies of CD34 + CD38low cells (LSC-type) contributed an additional 10% sensitivity (specificity of 93% above a cut-off of 0.02%) (Fig. 3 A) notwithstanding the absence of certain LSC-specific markers (such as CD45RA, CLL-1 and CD123). We then examined incremental clinical sensitivity contributed by DfN aberrancies in other progenitor compartments that constitute established LAIPs using ELN core MRD markers. Of these, aberrancies in the CD7 + compartment provided the greatest increment in sensitivity for incipient relapse. Two thirds of pre-relapse samples were identified as C-Flow-MRD positive by one or more of the above three aberrancy categories whilst the cut-off of 0.02% maintained a specificity of ≥90% in the combined non-relapse samples (Fig. 3 A). Most of the other tested aberrant subpopulations co-aggregated with these three (Fig. 3 B). Evolution in blast composition in months preceding relapse Non-genetic modes of evolution may potentially contribute to progression to relapse and could be reflected by changes in the abundance of specific immunophenotypic aberrant populations within the progenitor compartment. We investigated this by C-Flow-MRD in 86 paired pre-relapse (MRD timepoint) and relapse samples. Consistent with previous observations( 18 , 19 ), LSC or HLADR-low type aberrant subpopulations were abundant in blasts across relapsed patients (Fig. 4 A). Additionally, in over 40% of relapse samples with these populations, we observed rapid expansion (≥ 10 fold) within blasts during the pre-relapse interval (median 2.45 months) (Fig. 4 B). Overall, rapid expansion in LSC and /or HLADR-low type aberrant blasts occurred for 59% (51/86) of paired MRD and relapse samples. Prevalence was similar across the higher risk genetic categories for HLADR-low subpopulations; however 60% of relapses in patients with MDS-like mutations had a rapid expansion of LSC-like subpopulations compared to 32% with FLT3 -ITD mutated AML. Other categories of detectable leukemic aberrant immunophenotypes had significantly lower abundance and relative expansion across patients. For example, CD11b + type progenitor aberrancies, despite high MRD test specificity for relapse, remained as a relatively minor blast subpopulation (< 10% of blasts) in most patients. We observed instability of certain aberrant populations between MRD and relapse, defined by a ≥ 10 fold reduction within blasts, in 21 of 86 relapse samples; 71% of the 21 had adverse cytogenetics and/or FLT3 mutations. Instability was most frequently observed with CD7 + type aberrancies. In all cases of instability, 1 or more other detectable MRD aberrant blast subpopulations expanded within relapse blasts (examples provided Supplementary Fig. 5). Discussion Although there is a clinical need for pre-emptive relapse detection in high-risk AML patients, there has been a gap in evidence for the predictive capability of serial off-treatment flow cytometric MRD testing. To address this, we evaluated flow cytometric MRD assay clinical performance in a large set of pre-relapse and non-relapse AML MRD samples. This evaluation was then extended to a potential tool for standardised interpretation of flow cytometric MRD, computational ‘different-from-normal’ flow cytometric MRD by an unsupervised analysis pipeline (C-Flow-MRD). Finally we applied the objective approach of C-Flow MRD to construct a hierarchy of aberrant progenitor compartments according to clinical specificity / sensitivity for relapse and examine their evolution profile in paired relapse samples. Our results point to standard flow cytometric MRD as a highly specific predictor of relapse (98% ≥0.1%, 92% ≥0.05%) in higher risk AML for monitoring off-treatment by ‘different-from-normal’ analysis. This may obviate the requirement for a confirmatory repeat testing of a positive test as recommended based on experience from monitoring by PCR. Although over 60% of incipient relapses were detected by flow cytometric MRD ≥ 0.05%, this reduced to 50% when the higher current ELN threshold of 0.1% was applied. It is likely that extending monitoring interval beyond 2–3 months will further reduce clinical sensitivity. Our data underlines why a single MRD negative test should be interpreted with caution, notwithstanding the relatively short intervals of serial testing, due to rapid kinetics in a proportion of patients. The median log increase estimate per month in this high-risk cohort is comparable to that reported for NPM1 mutated patients by PCR derived results (0.7 log 10 per month)( 17 ). Consistent with clinical observations, results here show the highest median monthly log increase was apparent in the FLT3 mutated subgroup even compared to those with adverse cytogenetics. More sensitive targeted deep sequencing-based assays for FLT3 -ITD mutated variants (threshold of 0.01 − 0.001%) have the technical capacity to allow earlier MRD detection, but as a counterpoint, instability of these mutations risks false negatives, and the assays currently have a higher cost and slower turnaround time than flow cytometric MRD. Importantly this study demonstrates the potential of computational unsupervised pipelines for analysis of flow cytometric MRD, not only to facilitate standardised interpretation but also to objectively distinguish aberrant immunophenotypes that have the greatest accuracy for pre-emptive relapse detection. Our findings highlight that exclusion of what we term candidate ‘aberrant immunophenotypes with indeterminate potential’ (AIP-IP) depends on the range of controls used as a reference or ‘training set’ and impacts test clinical specificity most at low MRD levels. It is uncertain as yet whether such aberrant immunophenotypes are restricted to cells with clonal mutations although we would speculate that epigenetic mechanisms may contribute. Once these AIP-IPs are identified, in our pipeline they can be incorporated into the unsupervised training sample set, analogous to the process that informs expert manual MRD analysis. Indeed the higher specificity of the latter in our study may be explained by its empirical exclusion of the predominant AIP-IPs identified by C-Flow-MRD. Unsupervised analysis by C-Flow-MRD allowed broad characterisation of immunophenotypic aberrant blast subpopulations. Leveraging this for the paired MRD and relapse samples, we examined blast composition longitudinally through disease progression in order to identify aberrant HSPC subtypes most likely to escape disease control. We observed positive selection of aberrant HSPC subpopulations with weak / negative expression of CD38 (on CD34+) and/or HLADR– a finding in ~ 45% of this high-risk AML cohort by the metric of > 10-fold relative expansion within the blast compartment in the months preceding relapse. MRD detection of these subpopulations also had the highest accuracy for relapse. There are limitations to these results as aberrant subpopulation evolution is likely to depend on preceding treatments; the majority of paired MRD and relapse samples were post-transplant. Additionally, our assay had a restricted number of markers and was limited to blasts expressing CD34 and/or CD117. However extending this approach to other paired pre-relapse and relapse sample sets in AML and MDS may map immunophenotypic compartments enriched for leukemic cells with competitive advantage versus neutral evolution. This provides a potential tool for targeted multiomic profiling of relapse cell subpopulations to decipher the properties of fitness advantage and rapid relapse kinetics according to treatment. Our findings also have implications with regards future optimisation of recently proposed strategies to improve MRD detection depth by mutational screening on sorted HSPC populations( 20 , 21 ). In summary this study supports the utility of flow cytometric MRD for pre-emptive relapse surveillance in high-risk AML and provides a platform for the application of unsupervised MRD analysis to standardise and enhance evaluation of emerging leukemia. Declarations Author contributions Conception and design: Sylvie D Freeman, Nicholas McCarthy, Christophe Roumier, Adriana Plesa, Florent Dumezy, Gege Gui Collection and assembly of data: Sylvie D Freeman, Alexandra Adams, Georgia Andrew, Sarah Green, Naeem Khan, Nicholas McCarthy. Data analysis and interpretation: Nicholas McCarthy, Sylvie D. Freeman, Madeleine Jenkins, Gege Gui, Christophe Roumier, Adriana Plesa, Florent Dumezy Manuscript writing: Sylvie D Freeman, Nicholas McCarthy Final approval of manuscript: All authors Accountable for all aspects of the work: All authors Acknowledgements We would like to thank Blood Cancer UK for research support. This research was supported in part by the Intramural Research Program of the NIH and NHLBI. Disclosure of Conflicts of Interest NM, FD, AP, CR, GA, SG, AA, MJ, AP, have no conflicts to declare. SDF declares research funding from Jazz and BMS; Speakers Bureau with Jazz, Pfizer and Novartis; advisory committee with MPAACT. CH and GG, The National Heart, Lung, and Blood Institute receives research funding for the laboratory of Dr. Hourigan from the Foundation of the NIH AML MRD Biomarkers Consortium. References Schuurhuis GJ, Heuser M, Freeman S, Bene MC, Buccisano F, Cloos J, et al. Minimal/measurable residual disease in AML: a consensus document from the European LeukemiaNet MRD Working Party. Blood. 2018;131(12):1275–91. Heuser M, Freeman SD, Ossenkoppele GJ, Buccisano F, Hourigan CS, Ngai LL, et al. 2021 Update on Measurable Residual Disease (MRD) in Acute Myeloid Leukemia (AML): A Consensus Document from the European LeukemiaNet MRD Working Party. Blood. 2021. Ommen HB, Hokland P, Haferlach T, Abildgaard L, Alpermann T, Haferlach C, et al. Relapse kinetics in acute myeloid leukaemias with MLL translocations or partial tandem duplications within the MLL gene. Br J Haematol. 2014;165(5):618–28. Ommen HB, Schnittger S, Jovanovic JV, Ommen IB, Hasle H, Ostergaard M, et al. Strikingly different molecular relapse kinetics in NPM1c, PML-RARA, RUNX1-RUNX1T1, and CBFB-MYH11 acute myeloid leukemias. Blood. 2010;115(2):198–205. Ommen HB, Touzart A, MacIntyre E, Kern W, Haferlach T, Haferlach C, et al. The kinetics of relapse in DEK-NUP214-positive acute myeloid leukemia patients. Eur J Haematol. 2015;95(5):436–41. Puckrin R, Atenafu EG, Claudio JO, Chan S, Gupta V, Maze D, et al. Measurable residual disease monitoring provides insufficient lead-time to prevent morphologic relapse in the majority of patients with core-binding factor acute myeloid leukemia. Haematologica. 2021;106(1):56–63. Dohner H, Wei AH, Appelbaum FR, Craddock C, DiNardo CD, Dombret H, et al. Diagnosis and Management of AML in Adults: 2022 ELN Recommendations from an International Expert Panel. Blood. 2022. Hollein A, Meggendorfer M, Dicker F, Jeromin S, Nadarajah N, Kern W, et al. NPM1 mutated AML can relapse with wild-type NPM1: persistent clonal hematopoiesis can drive relapse. Blood Adv. 2018;2(22):3118–25. Schmalbrock LK, Dolnik A, Cocciardi S, Strang E, Theis F, Jahn N, et al. Clonal evolution of acute myeloid leukemia with FLT3-ITD mutation under treatment with midostaurin. Blood. 2021;137(22):3093–104. Lawrence L. Making MRD Assessment Work for AML. 2023. Loke J, McCarthy N, Jackson A, Siddique S, Hodgkinson A, Mason J, et al. Posttransplant MRD and T-cell chimerism status predict outcomes in patients who received allografts for AML/MDS. Blood Adv. 2023;7(14):3666–76. Tettero JM, Freeman S, Buecklein V, Venditti A, Maurillo L, Kern W, et al. Technical Aspects of Flow Cytometry-based Measurable Residual Disease Quantification in Acute Myeloid Leukemia: Experience of the European LeukemiaNet MRD Working Party. Hemasphere. 2022;6(1):e676. Freeman SD, Hills RK, Virgo P, Khan N, Couzens S, Dillon R, et al. Measurable Residual Disease at Induction Redefines Partial Response in Acute Myeloid Leukemia and Stratifies Outcomes in Patients at Standard Risk Without NPM1 Mutations. J Clin Oncol. 2018;36(15):1486–97. Craddock C, Jackson A, Loke J, Siddique S, Hodgkinson A, Mason J, et al. Augmented Reduced-Intensity Regimen Does Not Improve Postallogeneic Transplant Outcomes in Acute Myeloid Leukemia. J Clin Oncol. 2021;39(7):768–78. Hothorn T, Lausen B. On the exact distribution of maximally selected rank statistics. Computational Statistics & Data Analysis. 2003;43(2):121–37. Kanda Y. Investigation of the freely available easy-to-use software 'EZR' for medical statistics. Bone Marrow Transplant. 2013;48(3):452–8. Ivey A, Hills RK, Simpson MA, Jovanovic JV, Gilkes A, Grech A, et al. Assessment of Minimal Residual Disease in Standard-Risk AML. N Engl J Med. 2016;374(5):422–33. Zeng AGX, Bansal S, Jin L, Mitchell A, Chen WC, Abbas HA, et al. A cellular hierarchy framework for understanding heterogeneity and predicting drug response in acute myeloid leukemia. Nat Med. 2022;28(6):1212–23. Christopher MJ, Petti AA, Rettig MP, Miller CA, Chendamarai E, Duncavage EJ, et al. Immune Escape of Relapsed AML Cells after Allogeneic Transplantation. N Engl J Med. 2018;379(24):2330–41. Stasik S, Burkhard-Meier C, Kramer M, Middeke JM, Oelschlaegel U, Sockel K, et al. Deep sequencing in CD34 + cells from peripheral blood enables sensitive detection of measurable residual disease in AML. Blood Adv. 2022;6(11):3294–303. Dimitriou M, Mortera-Blanco T, Tobiasson M, Mazzi S, Lehander M, Hogstrand K, et al. Identification and surveillance of rare relapse-initiating stem cells during complete remission post-transplantation. Blood. 2023. Additional Declarations Yes there is potential conflict of interest. Supplementary Files SupplementMethodsandTables.pdf SupplementaryFigures.pdf Cite Share Download PDF Status: Published Journal Publication published 18 Jun, 2024 Read the published version in Leukemia → Version 1 posted Editorial decision: revise 20 Mar, 2024 Review # 1 received at journal 19 Mar, 2024 Review # 2 received at journal 07 Mar, 2024 Reviewer # 2 agreed at journal 03 Mar, 2024 Reviewer # 1 agreed at journal 02 Mar, 2024 Reviewers invited by journal 02 Mar, 2024 Editor assigned by journal 01 Mar, 2024 Submission checks completed at journal 23 Feb, 2024 First submitted to journal 22 Feb, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3978470","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":275991617,"identity":"0c5ad5f8-d054-4847-9554-122ca764cd2b","order_by":0,"name":"Sylvie Freeman","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3UlEQVRIiWNgGAWjYBAC+Qb+DxAWew8bTJAZrxaDAzxQFs8ZYrUwwLRI5BCvhe3Dh4pt8uaSb489+JnDIM/fwGNsgE8L0C+HZ844c9tw5+y8dMPebQyGMw7wGCfgteYADzMzb9ttxg23c8wkeLcxMG5g4DE+QFDL33+37TfcPGMm+Xcbgz1xWhgbbiduuMFjJg20JRGkBa/DDA7zMDP2HLudvOFMXpq07DaJ5BmH2Yrxe7+9h5nhR81t2w3Hzx6TfLvNxra/vXmzBF6HocWBBKFYGQWjYBSMglFADAAArXhE0l2NKZMAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0003-1869-180X","institution":"University of Birmingham","correspondingAuthor":true,"prefix":"","firstName":"Sylvie","middleName":"","lastName":"Freeman","suffix":""},{"id":275991618,"identity":"d64bc58a-4ac4-4b5e-a20a-e320e79667a0","order_by":1,"name":"Nicholas McCarthy","email":"","orcid":"","institution":"University Hospitals Birmingham National Health Service Foundation Trust","correspondingAuthor":false,"prefix":"","firstName":"Nicholas","middleName":"","lastName":"McCarthy","suffix":""},{"id":275991619,"identity":"cd1300fb-d200-494c-9e62-217349f38eb3","order_by":2,"name":"Gege Gui","email":"","orcid":"","institution":"National Heart, Lung, and Blood Institute, National Institutes of Health","correspondingAuthor":false,"prefix":"","firstName":"Gege","middleName":"","lastName":"Gui","suffix":""},{"id":275991620,"identity":"f602c675-c673-47ca-904e-b94bb225e228","order_by":3,"name":"Florent Dumezy","email":"","orcid":"","institution":"CHRU de lille","correspondingAuthor":false,"prefix":"","firstName":"Florent","middleName":"","lastName":"Dumezy","suffix":""},{"id":275991621,"identity":"8e90c7b5-6fae-47e2-9f00-4726f54acbcd","order_by":4,"name":"Christophe Roumier","email":"","orcid":"","institution":"Lille University Hospital, Université Lille Nord de France","correspondingAuthor":false,"prefix":"","firstName":"Christophe","middleName":"","lastName":"Roumier","suffix":""},{"id":275991622,"identity":"2b2d1061-f850-4886-bd2b-4067adab7d69","order_by":5,"name":"Georgia Andrew","email":"","orcid":"","institution":"Laboratory of Myeloid Malignancies, Hematology Branch, National Heart, Lung, and Blood Institute, National Institutes of Health","correspondingAuthor":false,"prefix":"","firstName":"Georgia","middleName":"","lastName":"Andrew","suffix":""},{"id":275991623,"identity":"1c3f5f23-e6c3-4a1d-a7f7-5c8a9a7903a4","order_by":6,"name":"Sarah Green","email":"","orcid":"","institution":"Institute of Immunology and Immunotherapy, University of Birmingham, United Kingdom","correspondingAuthor":false,"prefix":"","firstName":"Sarah","middleName":"","lastName":"Green","suffix":""},{"id":275991624,"identity":"3e6c855b-9ab3-44d3-b06d-45eea2c7eec3","order_by":7,"name":"Madeleine Jenkins","email":"","orcid":"","institution":"Imperial College","correspondingAuthor":false,"prefix":"","firstName":"Madeleine","middleName":"","lastName":"Jenkins","suffix":""},{"id":275991625,"identity":"c598d650-5b06-414d-842f-6b41cee15553","order_by":8,"name":"Alexandra Adams","email":"","orcid":"","institution":"University of Edinburgh","correspondingAuthor":false,"prefix":"","firstName":"Alexandra","middleName":"","lastName":"Adams","suffix":""},{"id":275991626,"identity":"8e890f66-394a-496a-8f31-d72eb0b908c2","order_by":9,"name":"Naeem Khan","email":"","orcid":"","institution":"University of Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Naeem","middleName":"","lastName":"Khan","suffix":""},{"id":275991627,"identity":"c2a83286-dea2-48d2-97e9-a566c347e662","order_by":10,"name":"Charles Craddock","email":"","orcid":"https://orcid.org/0000-0001-5041-6678","institution":"Queen Elizabeth Hospital","correspondingAuthor":false,"prefix":"","firstName":"Charles","middleName":"","lastName":"Craddock","suffix":""},{"id":275991628,"identity":"4f52998b-e141-4c5e-9dda-b411a83b2169","order_by":11,"name":"Christopher Hourigan","email":"","orcid":"","institution":"National Heart, Lung and Blood Institute, National Institutes of Health","correspondingAuthor":false,"prefix":"","firstName":"Christopher","middleName":"","lastName":"Hourigan","suffix":""},{"id":275991629,"identity":"e5caa5af-ed1e-4e7d-8a2c-68a14a99589e","order_by":12,"name":"Adriana Plesa","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Adriana","middleName":"","lastName":"Plesa","suffix":""}],"badges":[],"createdAt":"2024-02-22 11:52:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3978470/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3978470/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41375-024-02300-z","type":"published","date":"2024-06-18T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":52036417,"identity":"369f15d2-dfad-4c1a-b58e-e34bf085a2bf","added_by":"auto","created_at":"2024-03-05 17:07:46","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":90601,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA\u003c/strong\u003e. \u003cem\u003eClinical performance of flow cytometric MRD monitoring off treatment\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eReceiver operating curve (ROC) statistics for pre-emptive detection of relapse by standard ‘different-from-normal’ MRD analysis, calculated from the combined cohort (136 true positives [by relapse within 4 months] and 121 true negatives [by sustained off-treatment remission after the sample]). Clinical specificity and sensitivity shown for routine assay thresholds (0.05% and ELN 0.10%) and optimum thresholds (derived from Maxstat package).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB\u003c/strong\u003e. \u003cem\u003eRelapse kinetics of high-risk AML\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eRelapse kinetics for 137 paired MRD and relapse bone marrows from 119 high risk AML patients. Relapse leukemic aberrant immunophenotypes were examined in MRD samples in parallel to standard ‘different-from-normal’ MRD analysis and kinetics calculated from the highest MRD values. Summary of relapse kinetics is shown for the overall sample cohort and specific diagnostic genetics subgroups; there was a significant difference (p=0.036) in relapse kinetics between the \u003cem\u003eFLT3\u003c/em\u003e-mutated and adverse cytogenetics /\u003cem\u003eTP53\u003c/em\u003e subgroups by non-parametric Mann-Whitney U testing.\u003c/p\u003e","description":"","filename":"Slide1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3978470/v1/7f177781ccc38f6896f786e2.jpg"},{"id":52036416,"identity":"a0d8099e-41c5-49ed-b782-b3c4fe40dff1","added_by":"auto","created_at":"2024-03-05 17:07:46","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":72075,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA. \u003c/strong\u003e\u003cem\u003eClinical performance of computational flow cytometric MRD.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eReceiver operating curve (ROC) statistics for pre-emptive detection of relapse by C-Flow-MRD (‘different-from-normal’ unsupervised analysis against reference normal staging LPD BMs), calculated from the combined cohort (90 true positives [by relapse within 4 months] and 121 true negatives [by sustained off-treatment remission after the sample]). Clinical specificity and sensitivity shown for threshold MRD values (0.036% and 0.082%) derived from the test cohort.\u003c/p\u003e\n\u003cp\u003eLegend. C-Flow-MRD, computational (unsupervised) flow cytometric measurable residual disease;\u003c/p\u003e\n\u003cp\u003eLPD, lymphoproliferative disease; BM, bone marrow.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB. \u003c/strong\u003e\u003cem\u003eClinical specificity of flow cytometric C-Flow-MRD is impacted by range of controls\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAs C-Flow-MRD detected ‘different-from-normal’ phenotypes (normal = normal staging LPD BM controls) in 10 of 55 APML ‘true negatives’, the clinical specificity of C-Flow-MRD was compared with (LPD+ FP APML) and without (LPD) inclusion of false-positive APML BMs in control set across different % MRD values. Vertical lines represent C-MRD assay threshold values\u003cem\u003e \u003c/em\u003e(0.036% and 0.082%) derived from the test cohort.\u003c/p\u003e\n\u003cp\u003eLegend. C-Flow-MRD, computational (unsupervised) flow cytometric measurable residual disease;\u003c/p\u003e\n\u003cp\u003eLPD, lymphoproliferative disease; BM, bone marrow; APML, acute promyelocytic leukaemia.\u003c/p\u003e","description":"","filename":"Slide2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3978470/v1/3e0c39ffa1b6c7eb6681b6ab.jpg"},{"id":52036419,"identity":"a4a0526b-c3f8-47ca-8cb3-671454d757c6","added_by":"auto","created_at":"2024-03-05 17:07:46","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":112027,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA \u003c/strong\u003e\u003cem\u003eHierarchy of\u003c/em\u003e \u003cem\u003eaberrant progenitor compartments for\u003c/em\u003e \u003cem\u003eeffectiveness in relapse prediction\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eSpecific aberrant progenitor compartments were ranked serially for diagnostic performance by Youden’s index (YI: sensitivity+specificity-1). Panel shows sequential YI with optimal MRD threshold achieving \u0026gt;95% specificity for each aberrancy type, followed by sensitivity, specificity and overall performance (balanced accuracy/B accuracy) for relapse prediction achieved by serial combination. HLADRlow, followed by LSC and CD7+ type aberrancies (detected above a threshold of 0.02%) contributed most to assay clinical sensitivity.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB. \u0026nbsp;\u003c/strong\u003eAberrant phenotype co-occurrence in pre-relapse MRD samples, Frequencies shown are the % of samples positive by phenotype (row heading) also positive by phenotype (column heading), or positive by no other phenotype (‘Only’). Positivity is defined by C-Flow-MRD detection above threshold of 0.02% (0.01% for CD11b+) within these tested phenotypic compartments of blasts (CD34+ and/or CD117+). Applied gates and fluorescent intensity thresholds are shown in supplementary methods.\u003c/p\u003e\n\u003cp\u003eLegend. C-Flow-MRD, computational (unsupervised) flow cytometric measurable residual disease.\u003c/p\u003e","description":"","filename":"Slide3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3978470/v1/bc9b438397f3f03990c80d78.jpg"},{"id":52036926,"identity":"55876bf0-a917-4e5f-9d2d-536d3851151a","added_by":"auto","created_at":"2024-03-05 17:15:46","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":88743,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eRelapse evolution of aberrant progenitor types within blasts\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA.\u003c/strong\u003e Frequency of displayed aberrant phenotypic subtype within relapse sample blasts. Overall frequency for CD34+ and CD34- compartments shown at end of graph. Box plots indicate the median and interquartile range for relapse samples where aberrancy is detectable as assessed objectively by computational unsupervised analysis (whiskers=5-95\u003csup\u003eth\u003c/sup\u003e percentiles). Inset table displays detected versus non-detected frequency for each aberrant subtype (% of 76 relapses).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB\u003c/strong\u003e. Distribution of fold-changes in specific aberrant subtype frequency within blasts from pre-relapse MRD timepoint to relapse in paired samples (n=86). Fold changes in total CD34+ and CD34- aberrancies shown at end of graph.\u003c/p\u003e","description":"","filename":"Slide4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3978470/v1/c934f4914d9ba4004af1a982.jpg"},{"id":58637835,"identity":"49885957-0761-4849-9728-261e19284054","added_by":"auto","created_at":"2024-06-19 07:07:49","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":988041,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3978470/v1/e3e2ee48-f6c3-4e68-8a51-91fa8f294a1e.pdf"},{"id":52036421,"identity":"413b749b-5be5-430c-9e35-365cbcdc2dd8","added_by":"auto","created_at":"2024-03-05 17:07:47","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":179027,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"SupplementMethodsandTables.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3978470/v1/b654906dbc87e6f413ac52da.pdf"},{"id":52036420,"identity":"c2793116-8804-469a-bff4-6df19bb6e3b4","added_by":"auto","created_at":"2024-03-05 17:07:46","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":969628,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3978470/v1/5a0cf818c3462108b071ca15.pdf"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential conflict of interest.","formattedTitle":"Pre-emptive Detection and Evolution of Relapse in Acute Myeloid Leukemia by Flow Cytometric Measurable Residual Disease Surveillance","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDetection of impending relapse in acute leukemias allows selection of patients for pre-emptive therapies that may avoid ultimate treatment failure but also the challenges of cytoreduction and morbidities following hematological relapse. In acute myeloid leukemia, the evidence that sequential MRD testing can provide an early warning of relapse comes predominantly from studies applying real-time quantitative polymerase chain reaction (RT-qPCR) assays(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Most of this data is for common PCR MRD leukemia targets \u003cem\u003ePML-RARA\u003c/em\u003e, \u003cem\u003eCBFB-MYH11\u003c/em\u003e, \u003cem\u003eRUNX1-RUNX1T1\u003c/em\u003e, and \u003cem\u003eNPM1\u003c/em\u003e mutations although the relapse kinetics of other PCR targets, namely rarer rearrangements and overexpressed WTI have also been examined(\u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Based on the findings of these studies, definitions of treatment failure now include molecular MRD relapse (increase of MRD copy numbers \u0026ge;1 log\u003csub\u003e10\u003c/sub\u003e or conversion from MRD negativity to MRD positivity, confirmed in a second sample) and RT-qPCR surveillance for MRD relapse after treatment (3 monthly if bone marrow sampling) is recommended for CBF and \u003cem\u003eNPM1\u003c/em\u003e mutated AML(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). The term MRD relapse now also incorporates conversion from MRD negativity to positivity detected by other methods including flow cytometry and is thus applicable across AML subtypes(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). There remain however unresolved issues that limit the information from sequential MRD results and therefore the clinical benefit and opportunities from MRD detection of impending relapse. These in part arise from the challenges of determining clinical specificity and sensitivity of a result to predict relapse when applying the recommended monitoring schedule. Conversion from MRD negativity to low level MRD positivity may have low predictive accuracy for relapse (low specificity), even when the sample MRD target is a high risk leukemic transcript such as \u003cem\u003eDEK-NUP214\u003c/em\u003e(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Although a second consecutive sample showing rising MRD levels can confirm MRD relapse, this risks reducing the interval for treatment decisions in the context of often unpredictable relapse kinetics. Furthermore, the majority of hematological relapses may not be identified by a preceding MRD positive test (low clinical sensitivity) despite using the recommended 3 monthly bone marrow monitoring schedules with established highly sensitive qPCR assays such as applied to CBF AML patients(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Additionally, as the use of flow cytometry or NGS for surveillance of impending relapse remains exploratory in AML(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e), there is an unmet need to extend MRD relapse detection to all AML patients. Moreover, parallel molecular and flow cytometric serial surveillance may be necessary for some patients such as when there is a risk of clonal evolution(\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e) including from treatment escape with loss of the molecular MRD target(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Although flow cytometric MRD has the advantage of rapid turn-around time in addition to a sensitivity of at least 10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e, the current dependence on manual analysis of bidimensional plots can lead to inconsistent quantitation according to expertise(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e) and also limit the potential of deeper-immunophenotyping for further optimization.\u003c/p\u003e \u003cp\u003eWe recently demonstrated the prognostic value of flow cytometric MRD monitoring performed during the first year following allogeneic stem cell transplantation(\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). This study now evaluates the clinical sensitivity and specificity of flow cytometric bone marrow MRD surveillance for early relapse prediction. We compare standard MRD detection with a newly developed computational analysis approach, investigating assay performance and aberrant immunophenotypic populations that are most specific for imminent relapse.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eSamples and Patients\u003c/h2\u003e \u003cp\u003ePatient samples were from AML patients\u0026thinsp;\u0026gt;\u0026thinsp;18 years old followed for flow cytometric measurable residual disease (MRD) detection using standard published methods(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e) by a single reference laboratory (Birmingham, UK) from April 2015 to September 2022. MRD bone marrow monitoring was conducted after completion of chemotherapy or post allogeneic hematopoietic stem cell transplantation (HSCT) in patients who had achieved a complete remission (CR) or CR with incomplete hematologic recovery (CRi). MRD detection analyses were conducted with informed patient consent in accordance with the Declaration of Helsinki and subject to appropriate ethics committee approval.\u003c/p\u003e \u003cp\u003eMRD bone marrow samples preceding a hematological relapse were retrospectively identified and included if a paired relapse sample had been received up to 4 months after the MRD sample and the diagnosis of relapse was made subsequent to this relapse sample.\u003c/p\u003e \u003cp\u003eAML MRD monitoring control samples were defined as 1) post-treatment PCR negative bone marrows from patients with acute promyelocytic leukemia (APML) maintaining molecular remission (test control cohort) and 2) post treatment / HSCT bone marrows from non-favourable risk AML patients in continuous remission\u0026thinsp;\u0026gt;\u0026thinsp;6 months after MRD sample without a treatment intervention (validation control cohort).\u003c/p\u003e \u003cp\u003eAdditionally, pre-transplant bone marrow files from 156 patients entered into the FIGARO trial with previously reported standard flow cytometric MRD results(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e) were included for clinical validation of computational analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eFlow cytometric MRD testing and analysis\u003c/h2\u003e \u003cp\u003eMRD was assessed by flow cytometry as previously described in a central reference laboratory) (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e) Details on sample logistics, processing, and analysis strategy are provided in the Supplementary Methods. Routine flow cytometric MRD analysis was performed using a standardized manual gating strategy that screened blasts for different-from-normal (DfN) aberrant immunophenotypes that were established as useful and frequently observed leukemic aberrant immunophenotypes (LAIPs) and also for any previously identified baseline LAIPs when available. Samples were reported as MRD-negative if no baseline and/or different-from-normal LAIP cells could be quantitated above assay threshold (of 0.05%).\u003c/p\u003e \u003cp\u003eTo confirm the results of standard analysis, and exclude variability arising from subjective interpretation, flow cytometry standard files from MRD testing were analyzed using a computational approach (C-Flow MRD), updated following previous clinical evaluation(\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Blast cells (CD117+/CD34+) from test samples were clustered together with a 40\u0026ndash;50 control BM reference set using the FlowSOM clustering algorithm. Automated decision tree analysis was then applied to define abnormal blasts with an immunophenotype significantly different from the reference set in 7-dimensional space (light scatter and CD45 parameters excluded). C-Flow-MRD results were calculated by summating discrete abnormal blast populations above the limit of detection (LOD), and the assay result reported as the highest value of the two tubes of the AML MRD antibody panel. Analyses of specific progenitor compartments for C-Flow-MRD\u0026thinsp;+\u0026thinsp;blast cells were performed in FlowJo software through progenitor pre-set sub-gating based on optimised thresholds.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analyses\u003c/h2\u003e \u003cp\u003eReceiver operating curve (ROC) statistics with area under the curve (AUC) were generated for MRD results of the clinical cohorts to summarise the discrimination ability of testing to predict relapse, with values of \u0026gt;\u0026thinsp;0.75 considered as good. Clinical specificity (true negative results (TN) \u0026divide; [TN\u0026thinsp;+\u0026thinsp;false positives (FP)] x 100), sensitivity (true positive results (TP) \u0026divide; [TP\u0026thinsp;+\u0026thinsp;false negatives (FN)] x 100), balanced accuracy ([%sensitivity + %specificity]\u0026thinsp;\u0026divide;\u0026thinsp;2) and Youden index ([sensitivity\u0026thinsp;+\u0026thinsp;specificity] \u0026ndash; 1) were determined for test performance at specific assay cut-points. Optimal assay cut-points were derived from peaks in the Youden Index or alternatively by the R-based MaxStat package(\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e), which uses maximally selected rank statistics. Cumulative incidence of relapse (CIR) and treatment related mortality (TRM) from C-Flow-MRD applied to the previously published Figaro pre-transplant MRD sample dataset(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e) were calculated using the \u0026lsquo;cumulative incidence of competing events and Gray test analysis\u0026rsquo; and \u0026lsquo;Fine-Gray proportional hazard regression for competing events\u0026rsquo; functions of the EZR software package v1.61(\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). Further details are in Supplementary Methods.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eFlow cytometric MRD performance for detection of relapse during monitoring\u003c/h2\u003e \u003cp\u003eTo evaluate the diagnostic performance of routine flow cytometry MRD for detection of impending relapse in a clinical setting, we interrogated the flow cytometric data of 136 MRD bone marrows preceding a paired relapse sample (\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026le;\u003c/span\u003e\u0026thinsp;4 months, median interval 2.45 months) from 118 patients who experienced relapse during longitudinal flow cytometric MRD monitoring (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). MRD results were defined by MRD analysis that did not require a diagnostic sample (\u0026lsquo;different-from-normal\u0026rsquo; / DFN approach, Methods).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics by standard flow cytometric MRD status in pre-relapse samples\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;136\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eFlow MRD negative\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;47\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c6\" namest=\"c4\"\u003e \u003cp\u003eFlow MRD positive\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;89\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.1%\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;17\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e\u0026thinsp;0.1%\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;72\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInterval between MRD sample and relapse\u003c/p\u003e \u003cp\u003eMonths, median [IQR]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.45\u003c/p\u003e \u003cp\u003e[1.74\u0026ndash;3.08]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.77\u003c/p\u003e \u003cp\u003e[2.30\u0026ndash;3.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.27\u003c/p\u003e \u003cp\u003e[1.43\u0026ndash;2.80]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.80\u003c/p\u003e \u003cp\u003e[2.07\u0026ndash;3.43]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2.12 [1.39\u0026ndash;2.71]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDiagnostic Genetics\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdverse cytogenetics or \u003cem\u003eTP53\u003c/em\u003e mutated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e46\u003c/p\u003e \u003cp\u003e(35.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14\u003c/p\u003e \u003cp\u003e(34.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e32\u003c/p\u003e \u003cp\u003e(36.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9\u003c/p\u003e \u003cp\u003e(52.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e23\u003c/p\u003e \u003cp\u003e(32.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eFLT3\u003c/em\u003e mutated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e29\u003c/p\u003e \u003cp\u003e(22.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e13\u003c/p\u003e \u003cp\u003e(31.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003cp\u003e(18.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003cp\u003e(23.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e12\u003c/p\u003e \u003cp\u003e(16.9%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMDS-related mutations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e22\u003c/p\u003e \u003cp\u003e(17.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6\u003c/p\u003e \u003cp\u003e(14.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003cp\u003e(18.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003cp\u003e(5.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15\u003c/p\u003e \u003cp\u003e(21.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003cp\u003e\u003cem\u003eNPM1\u003c/em\u003e mutated\u003c/p\u003e \u003cp\u003e\u003cem\u003e(FLT3\u003c/em\u003e wild type)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e32\u003c/p\u003e \u003cp\u003e(24.8%)\u003c/p\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e8\u003c/p\u003e \u003cp\u003e(19.5%)\u003c/p\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e24\u003c/p\u003e \u003cp\u003e(27.3%)\u003c/p\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003cp\u003e(17.6%)\u003c/p\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e21\u003c/p\u003e \u003cp\u003e(29.6%)\u003c/p\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnknown\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTreatment Stage\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePost chemotherapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48\u003c/p\u003e \u003cp\u003e(35.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10\u003c/p\u003e \u003cp\u003e(21.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e38\u003c/p\u003e \u003cp\u003e(42.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5\u003c/p\u003e \u003cp\u003e(29.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e33\u003c/p\u003e \u003cp\u003e(45.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePost allograft\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e88\u003c/p\u003e \u003cp\u003e(64.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e37\u003c/p\u003e \u003cp\u003e(78.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e51\u003c/p\u003e \u003cp\u003e(57.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12\u003c/p\u003e \u003cp\u003e(70.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e39\u003c/p\u003e \u003cp\u003e(54.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRelapse major blast Immunophenotype\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCD34+\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e109\u003c/p\u003e \u003cp\u003e(80.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e44\u003c/p\u003e \u003cp\u003e(93.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65\u003c/p\u003e \u003cp\u003e(73.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12\u003c/p\u003e \u003cp\u003e(70.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e53\u003c/p\u003e \u003cp\u003e(73.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCD34-CD117+\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e21\u003c/p\u003e \u003cp\u003e(15.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003cp\u003e(6.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18\u003c/p\u003e \u003cp\u003e(20.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5\u003c/p\u003e \u003cp\u003e(29.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13\u003c/p\u003e \u003cp\u003e(18.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMonocytic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e(0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e(0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e(0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e(0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e(0%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6\u003c/p\u003e \u003cp\u003e(4.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e(0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003cp\u003e(6.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e(0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6\u003c/p\u003e \u003cp\u003e(8.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e89 samples were MRD positive resulting in a diagnostic sensitivity for subsequent relapse in 2 to 4 months of 64% for a result of MRD\u0026thinsp;\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e\u0026thinsp;0.05% and 51% for MRD\u0026thinsp;\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e\u0026thinsp;0.1%.\u003c/p\u003e \u003cp\u003eWe then assessed the specificity of a flow cytometric MRD test to predict relapse using two reference sets of MRD bone marrow samples from non-relapsing patients (clinical negatives). In the test set compromising 55 APML follow-up bone marrows (PCR negative, no subsequent relapse), 9% had MRD\u0026thinsp;\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e\u0026thinsp;0.05%. A similar frequency of clinical false positives (8%) was observed in the second \u0026lsquo;validation\u0026rsquo; reference set, consisting of 66 monitoring MRD bone marrows identified from higher risk AML patients with sustained off-treatment remission\u0026thinsp;\u0026gt;\u0026thinsp;6 months after the sample, Supplementary Table\u0026nbsp;1). From the integrated dataset of pre-relapse and all reference samples, MRD test specificity by \u0026lsquo;true negatives\u0026rsquo; (no relapse in the absence of intervention\u0026thinsp;\u0026gt;\u0026thinsp;6 months from test) was 92% for a result of \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e\u0026thinsp;0.05%. From receiver operating characteristic curve analysis, area under the curve was 0.84 for identification of patients with imminent relapse by routine flow cytometric MRD, with an optimal MRD clinical cut-off of 0.045% (sensitivity 73%, specificity 89%) in this retrospective cohort (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eRelapse Kinetics\u003c/h2\u003e \u003cp\u003eKinetics of relapse can inform appropriate monitoring interval and choice of intervention but data as yet has been limited to PCR MRD targets. In this large sample cohort from 118 high risk AML patients experiencing relapse the median bone marrow increment was 0.78log\u003csub\u003e10\u003c/sub\u003e/month (IQR, 0.58\u0026ndash;1.18), comparable to previous results for \u003cem\u003eNPM1\u003c/em\u003e mutated AML(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). We observed similar results between different high risk genetic subgroups although there was a significantly higher median increment in patients with \u003cem\u003eFLT3\u003c/em\u003e mutations at diagnosis 0.92log\u003csub\u003e10\u003c/sub\u003e/month (IQR, 0.66\u0026ndash;1.42) compared to patients with adverse risk cytogenetics or \u003cem\u003eTP53\u003c/em\u003e mutations (P\u0026thinsp;=\u0026thinsp;0.036) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eEvaluation of MRD relapse detection using computational flow cytometric analysis\u003c/h2\u003e \u003cp\u003eWe developed a computational flow cytometric analysis tool incorporating a clustering algorithm and machine learning for standardised multidimensional MRD analysis (C-Flow-MRD) (Methods, Supplementary Methods) that required no baseline sample or expert annotation for MRD quantitation. This was applied to further assess the robustness and accuracy of immunophenotypic aberrancies for relapse prediction. The technical sensitivity of C-Flow-MRD from dilution experiments of different leukemic aberrant immunophenotypes (LAIPs) was below 0.01% for most aberrancies (range 0.002\u0026ndash;0.02%) (Supplementary Fig.\u0026nbsp;1). The previously reported adverse prognostic impact of pretransplant MRD on relapse risk was recently confirmed in patients who entered the FIGARO trial, a randomized trial of reduced-intensity conditioning regimens. The clinical validity of C-Flow-MRD was tested on this cohort of 156 transplanted patients. The prognostic performance of results generated by C-Flow-MRD was equivalent to routine Flow-MRD (Supplementary Fig.\u0026nbsp;2).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eAberrant Immunophenotypes with indeterminate leukemic potential reduce clinical specificity\u003c/h2\u003e \u003cp\u003eWe applied C-Flow-MRD to the integrated dataset of pre-relapse and all reference samples. Results showed an equivalent clinical sensitivity to standard MRD but reduced specificity (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). C-Flow-MRD detected immunophenotypic aberrancies based on their absence from the training sample set (normal lymphoid neoplasia staging BMs) and this provided a high specificity in the APML \u0026lsquo;non-relapse\u0026rsquo; MRD bone marrow samples (82% and 91% at MRD cut-offs of 0.04% and 0.08% respectively). However the clinical false positive rate from C-Flow-MRD increased in the validation non-relapse cohort (specificity of 79% at MRD cut-off of 0.08%) (Supplementary Fig.\u0026nbsp;3). Since this indicated that some \u0026lsquo;different-from-normal\u0026rsquo; blasts immunophenotypes, not represented in our large set of lymphoma staging BMs (minimum 38 per antibody combination), are poor predictors of AML relapse, we extended the C-Flow-MRD training set to include 10 APML control samples with outlier false positive results (detectable above 0.036%). We hypothesised that these false positive results represented candidate aberrant immunophenotypes with indeterminate potential. Excluding these for C-Flow-MRD testing on non-APML monitoring BMs resulted in an improved specificity, that was most apparent for the lower thresholds of MRD detection (74% vs 55% at 0.04% threshold) but maintained comparable sensitivity for relapse (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). Aberrant immunophenotypes identified by C-Flow-MRD in the outlier false positive APML samples included CD117\u0026thinsp;+\u0026thinsp;progenitors with higher expression of both CD33 and HLADR or decreased HLADR with relative CD33 overexpression and B lymphoid progenitors with asynchronous weaker CD38 expression (Supplementary Fig.\u0026nbsp;4). Defining these as \u0026lsquo;false positives\u0026rsquo; supports their exclusion by standard DfN analysis.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eHierarchy for relapse prediction by MRD is dependent on progenitor compartment\u003c/h2\u003e \u003cp\u003eDepending on the AML genetic subtype, leukemic cells may be variably distributed between CD34\u0026thinsp;+\u0026thinsp;and CD34- subpopulations, however data is lacking for how this impacts MRD detection. We examined this by C-Flow-MRD to enable extended, objective phenotypic characterisation of aberrancies between CD34\u0026thinsp;+\u0026thinsp;and CD34- blasts. In our cohort of high-risk AML patients, clinical sensitivity for relapse by C-Flow-MRD was substantially from CD34\u0026thinsp;+\u0026thinsp;aberrancies, however for 8 of 90 analysed pre-relapse samples, MRD was only detectable in the CD34-117\u0026thinsp;+\u0026thinsp;compartment. CD34-CD117\u0026thinsp;+\u0026thinsp;blast based MRD added up to 9% clinical sensitivity to CD34\u0026thinsp;+\u0026thinsp;blast based MRD with MRD cut-offs ranging from 0.04\u0026ndash;0.1% and had comparable clinical specificity (close to 90% when MRD was \u0026gt;\u0026thinsp;0.08%). Notably only 3 of 8 patients with CD34-CD117\u0026thinsp;+\u0026thinsp;restricted C-Flow-MRD had \u003cem\u003eNPM1\u003c/em\u003e mutated AML (all 3 were also \u003cem\u003eFLT3\u003c/em\u003e-ITD co-mutated). All relapse samples from these patients gained CD34\u0026thinsp;+\u0026thinsp;aberrant populations, although CD34-CD117\u0026thinsp;+\u0026thinsp;remained the predominant population (\u0026gt;\u0026thinsp;95% of abnormal blasts) in 6.\u003c/p\u003e \u003cp\u003eRelapse-initiating cells may be enriched in certain hematopoietic stem and progenitor (HSPC) compartments associated with functional resistance through leukemic stem cell properties (CD34\u0026thinsp;+\u0026thinsp;CD38low) or immune evasion (such as HLADR downregulated HSPC). In our cohort that was skewed to post-allogeneic transplant relapses, C-Flow-MRD-detected aberrancies in the HLADRlow compartment provided the greatest effectiveness for clinical sensitivity and specificity (44% and 98% respectively above a cut-off of 0.02%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Aberrancies of CD34\u0026thinsp;+\u0026thinsp;CD38low cells (LSC-type) contributed an additional 10% sensitivity (specificity of 93% above a cut-off of 0.02%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA) notwithstanding the absence of certain LSC-specific markers (such as CD45RA, CLL-1 and CD123). We then examined incremental clinical sensitivity contributed by DfN aberrancies in other progenitor compartments that constitute established LAIPs using ELN core MRD markers. Of these, aberrancies in the CD7\u0026thinsp;+\u0026thinsp;compartment provided the greatest increment in sensitivity for incipient relapse. Two thirds of pre-relapse samples were identified as C-Flow-MRD positive by one or more of the above three aberrancy categories whilst the cut-off of 0.02% maintained a specificity of \u0026ge;90% in the combined non-relapse samples (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Most of the other tested aberrant subpopulations co-aggregated with these three (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eEvolution in blast composition in months preceding relapse\u003c/h2\u003e \u003cp\u003eNon-genetic modes of evolution may potentially contribute to progression to relapse and could be reflected by changes in the abundance of specific immunophenotypic aberrant populations within the progenitor compartment. We investigated this by C-Flow-MRD in 86 paired pre-relapse (MRD timepoint) and relapse samples. Consistent with previous observations(\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e), LSC or HLADR-low type aberrant subpopulations were abundant in blasts across relapsed patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). Additionally, in over 40% of relapse samples with these populations, we observed rapid expansion (\u0026ge;\u0026thinsp;10 fold) within blasts during the pre-relapse interval (median 2.45 months) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). Overall, rapid expansion in LSC and /or HLADR-low type aberrant blasts occurred for 59% (51/86) of paired MRD and relapse samples. Prevalence was similar across the higher risk genetic categories for HLADR-low subpopulations; however 60% of relapses in patients with MDS-like mutations had a rapid expansion of LSC-like subpopulations compared to 32% with \u003cem\u003eFLT3\u003c/em\u003e-ITD mutated AML.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eOther categories of detectable leukemic aberrant immunophenotypes had significantly lower abundance and relative expansion across patients. For example, CD11b\u0026thinsp;+\u0026thinsp;type progenitor aberrancies, despite high MRD test specificity for relapse, remained as a relatively minor blast subpopulation (\u0026lt;\u0026thinsp;10% of blasts) in most patients. We observed instability of certain aberrant populations between MRD and relapse, defined by a\u0026thinsp;\u0026ge;\u0026thinsp;10 fold reduction within blasts, in 21 of 86 relapse samples; 71% of the 21 had adverse cytogenetics and/or \u003cem\u003eFLT3\u003c/em\u003e mutations. Instability was most frequently observed with CD7\u0026thinsp;+\u0026thinsp;type aberrancies. In all cases of instability, 1 or more other detectable MRD aberrant blast subpopulations expanded within relapse blasts (examples provided Supplementary Fig.\u0026nbsp;5).\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eAlthough there is a clinical need for pre-emptive relapse detection in high-risk AML patients, there has been a gap in evidence for the predictive capability of serial off-treatment flow cytometric MRD testing. To address this, we evaluated flow cytometric MRD assay clinical performance in a large set of pre-relapse and non-relapse AML MRD samples. This evaluation was then extended to a potential tool for standardised interpretation of flow cytometric MRD, computational \u0026lsquo;different-from-normal\u0026rsquo; flow cytometric MRD by an unsupervised analysis pipeline (C-Flow-MRD). Finally we applied the objective approach of C-Flow MRD to construct a hierarchy of aberrant progenitor compartments according to clinical specificity / sensitivity for relapse and examine their evolution profile in paired relapse samples.\u003c/p\u003e \u003cp\u003eOur results point to standard flow cytometric MRD as a highly specific predictor of relapse (98% \u0026ge;0.1%, 92% \u0026ge;0.05%) in higher risk AML for monitoring off-treatment by \u0026lsquo;different-from-normal\u0026rsquo; analysis. This may obviate the requirement for a confirmatory repeat testing of a positive test as recommended based on experience from monitoring by PCR. Although over 60% of incipient relapses were detected by flow cytometric MRD\u0026thinsp;\u0026ge;\u0026thinsp;0.05%, this reduced to 50% when the higher current ELN threshold of 0.1% was applied. It is likely that extending monitoring interval beyond 2\u0026ndash;3 months will further reduce clinical sensitivity. Our data underlines why a single MRD negative test should be interpreted with caution, notwithstanding the relatively short intervals of serial testing, due to rapid kinetics in a proportion of patients. The median log increase estimate per month in this high-risk cohort is comparable to that reported for \u003cem\u003eNPM1\u003c/em\u003e mutated patients by PCR derived results (0.7 log\u003csub\u003e10\u003c/sub\u003e per month)(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Consistent with clinical observations, results here show the highest median monthly log increase was apparent in the \u003cem\u003eFLT3\u003c/em\u003e mutated subgroup even compared to those with adverse cytogenetics. More sensitive targeted deep sequencing-based assays for \u003cem\u003eFLT3\u003c/em\u003e-ITD mutated variants (threshold of 0.01\u0026thinsp;\u0026minus;\u0026thinsp;0.001%) have the technical capacity to allow earlier MRD detection, but as a counterpoint, instability of these mutations risks false negatives, and the assays currently have a higher cost and slower turnaround time than flow cytometric MRD.\u003c/p\u003e \u003cp\u003eImportantly this study demonstrates the potential of computational unsupervised pipelines for analysis of flow cytometric MRD, not only to facilitate standardised interpretation but also to objectively distinguish aberrant immunophenotypes that have the greatest accuracy for pre-emptive relapse detection. Our findings highlight that exclusion of what we term candidate \u0026lsquo;aberrant immunophenotypes with indeterminate potential\u0026rsquo; (AIP-IP) depends on the range of controls used as a reference or \u0026lsquo;training set\u0026rsquo; and impacts test clinical specificity most at low MRD levels. It is uncertain as yet whether such aberrant immunophenotypes are restricted to cells with clonal mutations although we would speculate that epigenetic mechanisms may contribute. Once these AIP-IPs are identified, in our pipeline they can be incorporated into the unsupervised training sample set, analogous to the process that informs expert manual MRD analysis. Indeed the higher specificity of the latter in our study may be explained by its empirical exclusion of the predominant AIP-IPs identified by C-Flow-MRD.\u003c/p\u003e \u003cp\u003eUnsupervised analysis by C-Flow-MRD allowed broad characterisation of immunophenotypic aberrant blast subpopulations. Leveraging this for the paired MRD and relapse samples, we examined blast composition longitudinally through disease progression in order to identify aberrant HSPC subtypes most likely to escape disease control. We observed positive selection of aberrant HSPC subpopulations with weak / negative expression of CD38 (on CD34+) and/or HLADR\u0026ndash; a finding in ~\u0026thinsp;45% of this high-risk AML cohort by the metric of \u0026gt;\u0026thinsp;10-fold relative expansion within the blast compartment in the months preceding relapse. MRD detection of these subpopulations also had the highest accuracy for relapse. There are limitations to these results as aberrant subpopulation evolution is likely to depend on preceding treatments; the majority of paired MRD and relapse samples were post-transplant. Additionally, our assay had a restricted number of markers and was limited to blasts expressing CD34 and/or CD117. However extending this approach to other paired pre-relapse and relapse sample sets in AML and MDS may map immunophenotypic compartments enriched for leukemic cells with competitive advantage versus neutral evolution. This provides a potential tool for targeted multiomic profiling of relapse cell subpopulations to decipher the properties of fitness advantage and rapid relapse kinetics according to treatment. Our findings also have implications with regards future optimisation of recently proposed strategies to improve MRD detection depth by mutational screening on sorted HSPC populations(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn summary this study supports the utility of flow cytometric MRD for pre-emptive relapse surveillance in high-risk AML and provides a platform for the application of unsupervised MRD analysis to standardise and enhance evaluation of emerging leukemia.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConception and design:\u0026nbsp;Sylvie D Freeman, Nicholas McCarthy, Christophe Roumier, Adriana Plesa, Florent Dumezy, Gege Gui\u003c/p\u003e\n\u003cp\u003eCollection and assembly of data: Sylvie D Freeman, Alexandra Adams, Georgia Andrew, Sarah Green, Naeem Khan, Nicholas McCarthy.\u003c/p\u003e\n\u003cp\u003eData analysis and interpretation: Nicholas McCarthy, Sylvie D. Freeman, Madeleine Jenkins, Gege Gui, Christophe Roumier, Adriana Plesa, Florent Dumezy\u003c/p\u003e\n\u003cp\u003eManuscript writing: Sylvie D Freeman, Nicholas McCarthy\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFinal approval of manuscript: All authors\u003c/p\u003e\n\u003cp\u003eAccountable for all aspects of the work: All authors\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe would like to thank Blood Cancer UK for research support.\u003c/p\u003e\n\u003cp\u003eThis research was supported in part by the Intramural Research Program of the NIH and NHLBI.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDisclosure of Conflicts of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNM, FD, AP, CR, GA, SG, AA, MJ, AP, have no conflicts to declare. SDF declares research funding from Jazz and BMS; Speakers Bureau with Jazz, Pfizer and Novartis; advisory committee with MPAACT.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCH and GG, The National Heart, Lung, and Blood Institute \u0026nbsp;receives research funding for the laboratory of Dr. Hourigan from the Foundation of the NIH AML MRD Biomarkers Consortium.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSchuurhuis GJ, Heuser M, Freeman S, Bene MC, Buccisano F, Cloos J, et al. Minimal/measurable residual disease in AML: a consensus document from the European LeukemiaNet MRD Working Party. Blood. 2018;131(12):1275\u0026ndash;91.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeuser M, Freeman SD, Ossenkoppele GJ, Buccisano F, Hourigan CS, Ngai LL, et al. 2021 Update on Measurable Residual Disease (MRD) in Acute Myeloid Leukemia (AML): A Consensus Document from the European LeukemiaNet MRD Working Party. Blood. 2021.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOmmen HB, Hokland P, Haferlach T, Abildgaard L, Alpermann T, Haferlach C, et al. Relapse kinetics in acute myeloid leukaemias with MLL translocations or partial tandem duplications within the MLL gene. Br J Haematol. 2014;165(5):618\u0026ndash;28.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOmmen HB, Schnittger S, Jovanovic JV, Ommen IB, Hasle H, Ostergaard M, et al. Strikingly different molecular relapse kinetics in NPM1c, PML-RARA, RUNX1-RUNX1T1, and CBFB-MYH11 acute myeloid leukemias. Blood. 2010;115(2):198\u0026ndash;205.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOmmen HB, Touzart A, MacIntyre E, Kern W, Haferlach T, Haferlach C, et al. The kinetics of relapse in DEK-NUP214-positive acute myeloid leukemia patients. Eur J Haematol. 2015;95(5):436\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePuckrin R, Atenafu EG, Claudio JO, Chan S, Gupta V, Maze D, et al. Measurable residual disease monitoring provides insufficient lead-time to prevent morphologic relapse in the majority of patients with core-binding factor acute myeloid leukemia. Haematologica. 2021;106(1):56\u0026ndash;63.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDohner H, Wei AH, Appelbaum FR, Craddock C, DiNardo CD, Dombret H, et al. Diagnosis and Management of AML in Adults: 2022 ELN Recommendations from an International Expert Panel. Blood. 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHollein A, Meggendorfer M, Dicker F, Jeromin S, Nadarajah N, Kern W, et al. NPM1 mutated AML can relapse with wild-type NPM1: persistent clonal hematopoiesis can drive relapse. Blood Adv. 2018;2(22):3118\u0026ndash;25.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchmalbrock LK, Dolnik A, Cocciardi S, Strang E, Theis F, Jahn N, et al. Clonal evolution of acute myeloid leukemia with FLT3-ITD mutation under treatment with midostaurin. Blood. 2021;137(22):3093\u0026ndash;104.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLawrence L. Making MRD Assessment Work for AML. 2023.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLoke J, McCarthy N, Jackson A, Siddique S, Hodgkinson A, Mason J, et al. Posttransplant MRD and T-cell chimerism status predict outcomes in patients who received allografts for AML/MDS. Blood Adv. 2023;7(14):3666\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTettero JM, Freeman S, Buecklein V, Venditti A, Maurillo L, Kern W, et al. Technical Aspects of Flow Cytometry-based Measurable Residual Disease Quantification in Acute Myeloid Leukemia: Experience of the European LeukemiaNet MRD Working Party. Hemasphere. 2022;6(1):e676.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFreeman SD, Hills RK, Virgo P, Khan N, Couzens S, Dillon R, et al. Measurable Residual Disease at Induction Redefines Partial Response in Acute Myeloid Leukemia and Stratifies Outcomes in Patients at Standard Risk Without NPM1 Mutations. J Clin Oncol. 2018;36(15):1486\u0026ndash;97.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCraddock C, Jackson A, Loke J, Siddique S, Hodgkinson A, Mason J, et al. Augmented Reduced-Intensity Regimen Does Not Improve Postallogeneic Transplant Outcomes in Acute Myeloid Leukemia. J Clin Oncol. 2021;39(7):768\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHothorn T, Lausen B. On the exact distribution of maximally selected rank statistics. Computational Statistics \u0026amp; Data Analysis. 2003;43(2):121\u0026ndash;37.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKanda Y. Investigation of the freely available easy-to-use software 'EZR' for medical statistics. Bone Marrow Transplant. 2013;48(3):452\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIvey A, Hills RK, Simpson MA, Jovanovic JV, Gilkes A, Grech A, et al. Assessment of Minimal Residual Disease in Standard-Risk AML. N Engl J Med. 2016;374(5):422\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZeng AGX, Bansal S, Jin L, Mitchell A, Chen WC, Abbas HA, et al. A cellular hierarchy framework for understanding heterogeneity and predicting drug response in acute myeloid leukemia. Nat Med. 2022;28(6):1212\u0026ndash;23.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChristopher MJ, Petti AA, Rettig MP, Miller CA, Chendamarai E, Duncavage EJ, et al. Immune Escape of Relapsed AML Cells after Allogeneic Transplantation. N Engl J Med. 2018;379(24):2330\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStasik S, Burkhard-Meier C, Kramer M, Middeke JM, Oelschlaegel U, Sockel K, et al. Deep sequencing in CD34\u0026thinsp;+\u0026thinsp;cells from peripheral blood enables sensitive detection of measurable residual disease in AML. Blood Adv. 2022;6(11):3294\u0026ndash;303.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDimitriou M, Mortera-Blanco T, Tobiasson M, Mazzi S, Lehander M, Hogstrand K, et al. Identification and surveillance of rare relapse-initiating stem cells during complete remission post-transplantation. Blood. 2023.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"leukemia","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"leu","sideBox":"Learn more about [Leukemia](http://www.nature.com/leu/)","snPcode":"41375","submissionUrl":"https://mts-leu.nature.com/cgi-bin/main.plex","title":"Leukemia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-3978470/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3978470/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMeasurable residual disease (MRD) surveillance in acute myeloid leukemia (AML) may identify patients destined for relapse and thus provide the option of pre-emptive therapy to improve their outcome. Whilst flow cytometric MRD (Flow-MRD) can be applied to high-risk AML/ myelodysplasia patients, its diagnostic performance for detecting impending relapse is unknown. We evaluated this in a cohort comprising 136 true positives (bone marrows preceding relapse by a median of 2.45 months) and 121 true negatives (bone marrows during sustained remission). At an optimal Flow-MRD threshold of 0.045%, clinical sensitivity and specificity for relapse was 73% and 89% respectively (51% and 98% for Flow-MRD\u0026thinsp;\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e\u0026thinsp;0.1%) by \u0026lsquo;different-from-normal\u0026rsquo; analysis. Median relapse kinetics were 0.78 log\u003csub\u003e10\u003c/sub\u003e/month but significantly higher at 0.92 log\u003csub\u003e10\u003c/sub\u003e/month for \u003cem\u003eFLT3\u003c/em\u003e-mutated AML. Computational (unsupervised) Flow-MRD (C-Flow-MRD) generated optimal MRD thresholds of 0.036% and 0.082% with equivalent clinical sensitivity to standard analysis. C-Flow-MRD-identified aberrancies in HLADRlow or CD34\u0026thinsp;+\u0026thinsp;CD38low (LSC-type) subpopulations contributed the greatest clinical accuracy (54% sensitivity, 93% specificity) and notably, by longitudinal profiling expanded rapidly within blasts in \u0026gt;\u0026thinsp;40% of 86 paired MRD and relapse samples.\u003c/p\u003e \u003cp\u003eIn conclusion, flow MRD surveillance can detect MRD relapse in high risk AML and its evaluation may be enhanced by computational analysis.\u003c/p\u003e","manuscriptTitle":"Pre-emptive Detection and Evolution of Relapse in Acute Myeloid Leukemia by Flow Cytometric Measurable Residual Disease Surveillance","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-05 17:07:42","doi":"10.21203/rs.3.rs-3978470/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"revise","date":"2024-03-20T11:30:47+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-03-19T22:39:10+00:00","index":1,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-03-07T07:46:33+00:00","index":2,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-03-04T02:18:00+00:00","index":2,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-03-02T20:53:14+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewersInvited","content":"","date":"2024-03-02T20:35:49+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-03-01T13:18:17+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-02-23T13:50:57+00:00","index":"","fulltext":""},{"type":"submitted","content":"Leukemia","date":"2024-02-22T11:49:47+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"leukemia","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"leu","sideBox":"Learn more about [Leukemia](http://www.nature.com/leu/)","snPcode":"41375","submissionUrl":"https://mts-leu.nature.com/cgi-bin/main.plex","title":"Leukemia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"8e521dcd-6ad5-454a-829d-3cc14418018e","owner":[],"postedDate":"March 5th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":29094982,"name":"Health sciences/Diseases/Haematological diseases/Haematological cancer/Leukaemia"},{"id":29094983,"name":"Health sciences/Risk factors"},{"id":29094984,"name":"Health sciences/Medical research"}],"tags":[],"updatedAt":"2024-06-19T07:07:44+00:00","versionOfRecord":{"articleIdentity":"rs-3978470","link":"https://doi.org/10.1038/s41375-024-02300-z","journal":{"identity":"leukemia","isVorOnly":false,"title":"Leukemia"},"publishedOn":"2024-06-18 04:00:00","publishedOnDateReadable":"June 18th, 2024"},"versionCreatedAt":"2024-03-05 17:07:42","video":"","vorDoi":"10.1038/s41375-024-02300-z","vorDoiUrl":"https://doi.org/10.1038/s41375-024-02300-z","workflowStages":[]},"version":"v1","identity":"rs-3978470","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3978470","identity":"rs-3978470","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}