Immunophenotypic changes following menin inhibition in acute myeloid leukemia

Abstract

Menin inhibition leads to an antileukemic effect through hematopoietic differentiation. Treatment with the menin inhibitor revumenib results in clinical remissions in relapsed or refractory (R/R) acute myeloid leukemia (AML) with either rearrangement of lysine methyltransferase 2A (KMT2A) or mutation in nucleophosmin 1 (NPM1), leading to regulatory approval of this drug. However, determinants of response to revumenib have not been fully elucidated. We examined the immunophenotype of leukemia cells by flow cytometry, in sequential bone marrow specimens from 48 patients with R/R AML treated with revumenib. We observed dynamic changes in the immunophenotype after treatment in 16 of 31 (52%) patients, characterized by a switch from a myeloid/stem-like to a monocytic or myelomonocytic immunophenotype, or vice versa, or by substantial changes in the intensity of antigen expression or in patterns of leukemia-associated immunophenotypes. Morphologic remission with undetectable measurable residual disease (MRD) by flow cytometry following revumenib was associated with improved overall survival, with a median of 23.6 months compared with 20.8 months in patients with morphologic response and detectable MRD, and 3.2 months in non-responders. In summary, treatment monitoring of AML by flow cytometry, following menin inhibition, requires recognition of phenotypic changes associated with differentiation.

Introduction

The homeobox genes (HOX) promote an undifferentiated phenotype in acute leukemia [1]. Multiple recurrent genetic abnormalities in acute leukemia, including rearrangements of the lysine methyltransferase 2A (KMT2Ar) or nucleoporin 98 (NUP98r) genes, or nucleophosmin 1 mutation (NPM1mt), are associated with dysregulated transcription of HOX genes [2]. The interaction of the scaffold protein menin with KMT2A is a vulnerability in these acute leukemia subtypes [3]. Monotherapy with menin inhibitors leads to clinical responses with an overall response rate reaching 60% in the R/R setting [3–8]. The menin inhibitor revumenib was approved by the Food and Drug Administration in November of 2024 for R/R acute leukemia with KMT2Ar in adult and pediatric patients 1 year and older, and later in October of 2025 for R/R AML with a susceptible NPM1mt in adult and pediatric patients 1 year and older who have no satisfactory alternative treatment options. Notably, differentiation syndrome has been observed following menin inhibition, along with reduced transcription of critical leukemogenic genes, such as HOXA9 and MEIS1, and upregulation of differentiation markers, including CD11b, suggesting phenotypic differentiation and an antileukemic effect [5, 6]. However, not all patients respond to this therapy, and predictors of primary resistance have not been identified.

Flow cytometry (FC) is used to detect and quantify measurable residual disease (MRD) in acute myeloid leukemia (AML). Understanding changes in phenotype after differentiation-inducing treatment is crucial for accurate monitoring of response. NPM1mt AML is typically associated with characteristic immunophenotypic alterations, including frequent monocytic differentiation, which at times hinders the detection of residual disease [9]. The leukemia cell of origin could affect response to therapy; for example, a primitive phenotype has been associated with resistance to therapy and monoblastic differentiation with resistance to BCL2 inhibition [10, 11]. To date, immunophenotypic predictors of response to menin inhibitors or phenotypic changes induced by menin inhibition in AML have not been defined. We conducted a retrospective analysis of patients with R/R AML treated with revumenib and correlated their immunophenotypic and clinical findings.

Materials and methods

Study population

We identified 48 patients with R/R AML with KMT2Ar, NPM1mt or NUP98r treated with revumenib monotherapy on clinical trials (NCT04065399, NCT05360160) who achieved a morphologic response following therapy and compared the immunophenotype of their pre-menin inhibitor AML with the immunophenotype of persistent MRD during remission, or relapsed AML following menin-inhibitor therapy, when applicable.

Flow cytometric immunophenotyping (FC) for MRD assessment

A 12-color FC for AML with (MRD, sensitivity of 10^–3–10^–4) was performed using a two tube combination containing i) CD7-FITC, CD33-PE, CD19 BV786, CD34- PerCP-Cy5·5, CD13-APC, CD38-APC-R718, CD45-APC-H7, CD117-PE, CD133-V450, CD5-V500, CD4-BV605 and CD54 BV711; ii) HLA-DR-V450, CD34- PerCP-Cy5·5, CD123-APC, CD38-APC-R718, CD45-APC-H7 CD36-FITC, CD56-BV605CD64-PE-Cy7, CD14-BV786, CD117-PE, CD15-V500 and CD25-BV711 [11, 12]. A separate add-on tube containing markers of megakaryocytic differentiation (CD41, CD42b and CD61) was used as needed; at least two markers were required for megakaryocytic designation. The initial diagnostic FC and AML MRD FC panels were nearly identical in the composition of tubes and the markers that were assessed, facilitating MRD detection. Immunophenotypic aberrancy and presence of MRD were determined using a combined antigen/marker approach, employing both the leukemia-associated immunophenotype (LAIP) and deviation from normal approaches, as described previously [13, 14]. In each analysis tube, after removing debris and doublets, a broad initial gate (CD45dim gate) was established around the CD45dim blast and monocyte regions on a standard CD45/SSC plot, with subsequent back-gating used to pinpoint the CD34+ populations, CD117+ populations, and CD33bright/CD4+ monocytic populations, within the CD45/SSC display. (Supplementary File 1). Data from each tube was assessed for cells within both the CD34+ myeloblast gate and the CD117+ populations, as well as the entire CD45dim gate, including monocytes. This method enabled precise measurement of key marker expression levels on CD34+ myeloblasts and other populations within the CD45 dim gate, with additional scrutiny of any anomaly in this gate.

Blasts were grouped into four immunophenotypic categories: myeloid/stem-like (SL), myelomonocytic, monocytic, or megakaryoblastic (Fig. 1A and Supplementary Table 1).

Fig. 1. Pre-treatment AML immunophenotypes and the clinical impact of immunophenotypic changes after treatment with revumenib.

A Examples of myeloid/stem line (top panel); myelomonocytic (middle panel) and monocytic (bottom panel) AML immunophenotypes prior to treatment with revumenib; myeloid blasts are shown in red, and monocytes are shown in dark blue. B Immunophenotypic changes at relapse C Overall survival (OS) stratified by MRD status. MRD-negative patients had superior OS compared to non-responders (p < 0.001); patients achieving a response but with residual MRD (MRD + ) also had superior OS compared to non-responders (p = 0.019); the OS of MRD-negative patients (by flow cytometry) was numerically longer compared to MRD+ patients, but the difference did not attain statistical significance (by log-rank). Abbreviations: MRD + : measurable residual disease detected by multicolor flow cytometry at best response. NE not evaluable.

Cytogenetic and molecular analysis

Chromosome banding analysis was conducted on all baseline and follow-up samples as part of our standard diagnostic evaluation for hematological malignancies. Twenty Giemsa-banded metaphase cells were analyzed, and the findings were reported according to the 2020 International System for Human Cytogenetic Nomenclature (ISCN 2020).

Fluorescence in situ hybridization (FISH) for KMT2A (at chromosome 11q23) and NUP98 (at chromosome 11p15.4) rearrangement was performed using a dual-color breakapart rearrangement probe (Abbott Molecular Inc).

Next-generation sequencing (NGS) was performed as previously described, targeting all exonic or “hotspot” regions of 81 genes commonly mutated in myeloid neoplasms [15].

KMT2Ar molecular MRD assessment

Molecular MRD assessment for KMT2A rearrangement was performed using an RNA-based multiplex quantitative polymerase chain reaction (qPCR) assay (Pupil Bio, Houston, TX). This assay employs 205 primers to interrogate 12 exons of KMT2A (including major and minor breakpoint cluster regions) against 193 exons from 12 common fusion partners (AFDN, AFF1, ELL, EPS15, MLLT1, MLLT10, MLLT11, MLLT3, MLLT6, SEPTIN6, SEPTIN9, USP2), covering approximately 95% of KMT2Ar(+) cases, as described previously [16]. This allows detection of as few as 5 copies, corresponding to ~0.05% VAF from 1 µg of total RNA.” (Supplementary Fig. 1).

Response assessment and survival analysis

Responses to revumenib were assessed using the European LeukemiaNet 2022 criteria [17]. Suspicion of differentiation syndrome (DS) was defined according to any of the Montesinos criteria [17, 18]. Binomial univariate and multivariate logistic regression models were used to identify predictors of morphologic remission. Firth-type penalization was added to the logistic regression for DS to address quasi-complete separation in the covariates. Overall survival (OS) was calculated from the revumenib start date to the time of death or last follow-up. The Kaplan–Meier method was used to estimate the OS probability, and the estimate was compared using the log-rank test. This study was approved by the institutional review board and performed in accordance with the Declaration of Helsinki.

Results

Baseline characteristics

The baseline characteristics of AML patients included in this analysis are detailed in Table 1. The median age was 23.5 years (range, 0.8–78 years), with 19 (40%) patients younger than 18 years at the start of treatment. Patients were heavily pre-treated, with a median of 2 (1–8) prior lines of therapy; 36 (75%) had received venetoclax, and 28 (58%) had undergone allogeneic stem cell transplant. Thirty-nine (81%) patients had KMT2Ar, 6 (13%) had NPM1mt, and 3 (6%) had NUP98r. The most common co-occurring mutations were in WT1 (27%), K/NRAS (20%), and PTPN11 (10%) (Fig. 2). Among 47 evaluable patients, twenty (42%) had a myeloid/SL phenotype, 16 (33%) had a monocytic phenotype, 9 (19%) had a myelomonocytic phenotype, and 2 (4%) had a megakaryoblastic phenotype. The latter 2 were children, 4 and 9 years old, both had KMT2A rearrangements caused by t(9;11)(p22;q23) and t(11;17)(q23;q21), respectively. The pretreatment immunophenotype was not evaluable in one patient because the sample was completely necrotic.

Table 1.

Baseline characteristics and treatment response.

Characteristics	N = 48
Median age, years (range)	23.5 (0.8–78)
Age < 18 years, n (%)	19 (40%)
Male, n (%)	23 (48%)
Female, n (%)	25 (52%)
BM blasts, % (range)	58 (4-99)
Extramedullary disease, n (%)	14 (29%)
Prior therapy
Median lines of therapy (range)	2 (1–8)
Venetoclax, n (%)	36 (75%)
Allo-HSCT, n (%)	28 (58%)
Genotypes
KMT2A-rearranged, n (%)	39 (81%)
NPM1-mutated, n (%)	6 (13%)
NUP98-rearranged, n (%)	3 (6%)
Pre-treatment immunophenotype
Myeloid/stem cell, n (%)	20 (42%)
Monocytic, n (%)	16 (33%)
Myelomonocytic, n (%)	9 (19%)
Megakaryoblastic, n (%)	2 (4%)
Not evaluable **	1 (2%)
Best morphologic response
CR, CRh, CRi, or CRp, n (%)	19/46 (41.3%)
CR/CRh	8/46 (17.4%)
MLFS, n (%)	8/46 (17.4%)
ORR, n (%)	27/46 (58.7%)
No response, n (%)	19/46 (41.3%)
Not assessed, n (%)	2/48 (4.2%)
MRD by FC in responders	N = 27
Positive, n (%)	12 (44.4%)
Negative, n (%)	12 (44.4%)
Not evaluable, n (%)	3 (11.1%)

Allo-HSCT allogeneic hematopoietic stem cell transplant, CR complete response, CRh complete response with partial hematologic recovery, CRi complete Remission with incomplete count recovery, CRp complete response with incomplete platelet recovery, FC flow cytometry, MLFS morphologic leukemia-free state, MRD measurable residual disease, ORR overall response rate.

** Immunophenotypic characterization was not possible in one patient due to poor viability of the sample. The concurrent core biopsy specimen showed extensive necrosis.

Fig. 2. Distribution of co-mutations by genotype in patients with relapsed or refractory AML treated with revumenib.

Red fill indicates the presence of a mutation.

Immunophenotypic changes and response to therapy

Responses to revumenib were similar to those observed in the multicenter study, with an overall response rate of 59% and a complete remission (CR) or a CR with partial hematologic recovery of 17%. (Table 1) [4, 6]. Response to revumenib was similar across baseline immunophenotypes, with no impact on OS by these subgroups (Supplementary Fig. 2). However, we found a high frequency of changes in immunophenotype following treatment with revumenib, either at the time of MRD assessment in patients with MRD positivity in morphologic remission or at relapse. Specifically, among 12 patients who attained a morphologic remission with MRD positivity by multicolor FC, 8 (67%) had aberrant blasts with an immunophenotypic shift compared with the pre-treatment immunophenotype (Fig. 3). The phenotypes of baseline and relapse were available for comparison in 14 patients who experienced relapse after achieving morphologic remission. In this subset, 9 (64%) of 14 showed an immunophenotypic shift at relapse, including 6 with an altered state of differentiation (Fig. 1B). Among the 27 patients who attained a morphologic response with revumenib, 12 (44%) had no detectable MRD by multicolor FC, which was associated with numerically longer OS compared to those with MRD-positive responses (median OS 23.6 months vs 20.8 months) (Fig. 1C). The immunophenotype before treatment, including myeloid, myelomonocytic, monocytic, or megakaryocytic differentiation, did not predict the probability of achieving a morphologic remission (Fig. 4A and Supplementary Table 2). The overall response rates were 50%, 55%, 63%, and 50% for myeloid/stem-like, myelomonocytic, monocytic, and megakaryoblastic cases, respectively (p = 0.9) (Supplementary Table 3). Patients who had received venetoclax had lower odds of response to revumenib, possibly associated with impairment of apoptotic machinery. Those who had undergone allogeneic stem cell transplant had an increased rate of remission, in line with emerging preclinical data suggesting a role of menin inhibitors in potentiating the graft-vs-leukemia effect [19, 20]. Twenty-five patients had a suspicion of DS, and the baseline immunophenotype did not predict the likelihood of developing DS (Fig. 4B and Supplementary Table 4). A small number of patients in each subgroup limits this analysis.

Fig. 3. Changes in immunophenotype following revumenib in patients who attained a morphologic response with measurable residual disease.

Expression of antigens measured by flow cytometry in leukemic blasts or monocytes compared to their normal counterparts (e.g., normal CD34+ myeloid progenitors and mature monocytes).

Fig. 4. Predictors of response and differentiation syndrome following revumenib in relapsed or refractory acute myeloid leukemia patients.

Univariate and multivariate analyses predicting the odds of composite complete remission (CRc) (A) and differentiation syndrome (B). UVA, univariate analysis; MVA, multivariate analysis EMD extramedullary disease, BM blast % bone marrow blast percentage at diagnosis, # Prior Tx number of previous lines of therapy, Prior Ven prior venetoclax, Prior HSCT prior hematopoietic stem cell transplantation.

Flow cytometry and other disease burden measures

We compared the results of FC and FISH in patients with KMT2Ar or NUP98r (Table 2). Focusing on discordant results, two of 11 (18%) evaluable patients with undetectable MRD by FC had detectable KMT2A rearrangements in >90% of cells, suggesting differentiation of leukemic cells. Among 4 evaluable patients with MRD+ with a phenotype similar to the original AML and available FISH and/or karyotype, all had undetectable rearrangements by FISH; however, all of these patients had MRD by FC of less than 1%, which falls below the level of detection for FISH. Among evaluable patients with MRD+ disease and immunophenotypic shift (n = 7), 2 (29%) had undetectable disease by FISH. In comparison, all the others (71%) had a large cytogenetic burden detected, disproportionate to the small amount of disease by FC, suggesting at least partial differentiation.

Table 2.

Details of flow cytometry and cytogenetics data in patients with discordant results.

ID	MRD by FC Status (%)		IP Switch	Genotype	FISH (%)		Karyotype
1	Neg		No	KMT2Ar	Pos	96%	46,XX,t(9;11)(p22;q23)[17]/46,XX[3]
2	Neg		No	KMT2Ar	Neg	0	46,XY[10]
3	Neg		No	KMT2Ar	Neg	0	46,XX [20]
4	Neg		No	KMT2Ar	ND	ND	Hypocellular
5	Neg		No	KMT2Ar	Neg	0	Hypocellular
6	Neg		No	KMT2Ar	Neg	0	46,XX [20]
7	Neg		No	KMT2Ar	Neg	0	Hypocellular
8	Neg		No	KMT2Ar	Neg	0	Hypocellular
9	Neg		No	KMT2Ar	Pos	97%	46,XX,t(9;11)(p22;q23)[14]/46,XX[6]
10	Neg		No	KMT2Ar	Neg	0	ND
11	Neg		No	KMT2Ar	Neg	0	ND
12	Neg		No	NUP98r	Neg	0	46,XY[20]
13*	Pos	Rare (<0.01%)	Yes	KMT2Ar	Neg	0	46,XX[20]
14	Pos	0.02%	No	KMT2Ar	Neg	0	46,XX[20]
15	Pos	0.89%	No	KMT2Ar	Neg	0	46,XY[20]
16	Pos	1.5%	No	KMT2Ar	ND	ND	Hypocellular
17	Pos	0.1%	No	NUP98r	Pos	29%	46,XY[20]
18*	Pos	0.8%	Yes	KMT2Ar	ND	ND	47,XY,t(9;11)(p22;q23),+21[18]
19*	Pos	2.8%	Yes	KMT2Ar	Pos	49%	47,XY,t(9;11)(p22;q23),+21[19]//46,XX[1]
20*	Pos	0.01%	Yes	KMT2Ar	Pos	81%	46,XX,t(11;19)(q23;p13.1),del(12)(p13p11.2)[6]/46,XX,t(1;14)(q21;q24),t(11;19)(q23;p13.1)[1]
21*	Pos	0.02%	Yes	KMT2Ar	Neg	0	ND
22*	Pos	0.2%	Yes	KMT2Ar	Pos	93%	46,XX,ins(10;11)(p11.2;q13q23)[13]/46,idem,del(11)(q14q22)[7]
23*	Pos	0.2%	Yes	KMT2Ar	Pos	31%	46,XX,t(9;11)(p22;q23)[5]/47,idem,+21[13]/46,XX[2]

MRD measurable residual disease, FC flow cytometry, IP immunophenotype, FISH fluorescence in situ hybridization, Neg negative, ND not done, Pos positive.

*MRD+ with immunophenotypic shift.

Case 15 had a partial response only and is therefore not represented in Fig. 3.

Molecular assessment for KMT2Ar was performed retrospectively in 3 patients with available banked samples (pts 7, 10, and 21 in Table 2). These patients had undetectable KMT2Ar by FISH) (Supplementary Fig. 3). Among these patients, patients 7 and 10 had detectable KMT2Ar by qPCR (estimated variant allelic frequencies of 0.9–8.1% and 0.016–0.12%, respectively). In contrast, patient 21 had an inconclusive result, with a very low-level detectable transcript below the assay’s limit of detection/quantification. Patients 7 and 10 had undetectable AML MRD by FC, and patient 21 had 0.02% abnormal CD34+ myeloid blasts detected by FC, which were different from the pre-treatment AML phenotype, which was myelomonocytic.

Discussion

Our findings highlight the dynamic immunophenotypic changes and hematopoietic differentiation induced by the menin inhibitor revumenib. In this cohort of adult and pediatric patients, we observed a switch in immunophenotype in 52% of cases, typically from a myeloid/stem-like to a monocytic or myelomonocytic profile, or significant alterations in antigen expression intensity and leukemia-associated immunophenotype patterns.

Overall, this suggests that phenotypic changes following treatment with revumenib in AML patients need to be recognized to determine response to therapy and assess MRD accurately. Given that FC frequently relies on “deviations from normal” for detection and quantification of MRD, defining a response to differentiating agents like menin inhibitors could be challenging, as marrow recovery or differentiation may resemble residual disease and resistance due to atypical immunophenotypes associated with regeneration [21]. Therefore, longer treatment may be necessary in some cases, and repeat assessment in 1–2 weeks could facilitate better discrimination.

The immunophenotypic shifts observed in our study following treatment with revumenib are consistent with the expected epigenetic rewiring induced by this class of therapies, which disrupts the menin-KMT2A interaction and promotes myeloid differentiation [22–24]. These changes align with the known biological effects of menin inhibitors, which induce upregulation of differentiation markers, as reported in preclinical and clinical studies [5, 25]. Similar immunophenotypic alterations have been observed with other differentiation-inducing therapies, such as IDH inhibitors, where differentiation syndromes and surface marker shifts are well-documented [26], and ATRA/ATO in acute promyelocytic leukemia [27]. Indeed, in a recent study, relapse following IDH inhibition was associated with upregulated expression of stemness markers and downregulation of major histocompatibility complex class II expression by RNA sequencing with a shift in immunophenotype by FC assessment [28]. These parallels highlight that the observed phenotype switches are an anticipated consequence of epigenetic therapies, underscoring the need for careful interpretation of FC-based MRD in the context of menin inhibitor or any differentiating agent, to distinguish differentiation from residual leukemia.

Our findings also underscore the complementary roles of FC and molecular MRD assessments in the context of menin inhibitor therapy. The differentiation-inducing effects of revumenib, akin to those observed with ATRA/ATO in acute promyelocytic leukemia, can lead to residual NPM1mt in maturing myeloid cells early after treatment, which may not signify persistent leukemia but rather a therapeutic response [27]. In such cases, typically early after initiation of a differentiating-inducing therapy, substantial levels of residual NPM1mt are detectable by PCR or NGS but no abnormal phenotype by  FC, as corroborated by immunohistochemical (IHC) evidence of aberrant cytoplasmic NPM1 localization within differentiating cells. An example of this phenomenon is illustrated in Fig. 5.

Fig. 5. Immunophenotypic and morphological analysis of cytoplasmic NPM1 by immunohistochemistry in differentiating cells post-therapy.

Baseline bone marrow shows sheets of immature mononuclear cells consistent with blasts (A hematoxylin and eosin (H&E), 100x). The corresponding flow cytometric immunophenotyping shows a large population of aberrant cells (purple arrow) that are CD117 + , CD34-, CD123 + , HLA-DR- (C top panel). A post-treatment bone marrow shows maturing myeloid cells, with no significant increase in blasts (B H&E, 400x) and the corresponding flow cytometric analysis (C bottom panel) was negative for AML MRD; note the empty space corresponding to the location of the blast population in the baseline sample (purple arrow). Notably, NPM1 IHC performed on the post-treatment sample shows the presence of cytoplasmic staining (mutant pattern) in differentiating myeloid cells (yellow arrows) with indented and segmented nuclei and ample cytoplasm (D immunohistochemistry with hematoxylin counterstain, 1000x).

Collectively, these findings highlight the importance (albeit with some limitations) of FC in the clinical setting for these genotypes, given this method’s broad accessibility and rapid turnaround time. However, we also emphasize the need to integrate FC and molecular results to interpret residual disease accurately. Using IHC for the detection of NPM1 mutations offers a unique opportunity to visualize morphology and the types of cells retaining mutant NPM1, thereby providing greater clarity in many clinical scenarios. However, IHC is not clinically available for KMT2Ar, NUP98r, or for many other alterations, posing a challenge for assessing residual disease in these contexts.

Alternatively, as observed in this study, residual leukemic blasts after treatment with menin inhibitors may exhibit an immunophenotypic change compared with baseline, which could pose a diagnostic challenge when using the LAIP method for disease detection. Of note, evaluating MRD in AML with monocytic differentiation by FC, which represents a substantial subset of KMT2Ar- and NPM1mt AML, poses significant challenges. Differentiating immature monocytes (monoblasts or promonocytes) from their mature counterparts is often complex and, in many instances, unfeasible using immunophenotyping alone. This difficulty stems from the lack of highly specific markers for immature monocytes; while a complete absence of CD14 expression may suggest immaturity, this is observed in only a minority of cases. Other indicators, such as reduced CD13, CD36, and/or HLA-DR levels alongside elevated CD15 expression, are commonly associated with immature monocytes. However, these changes lack specificity because they can also be seen in reactive or dysplastic monocytes [13]. Therefore, interpretation of FC findings should be complemented with other MRD detection tests, such as next-generation sequencing, to better quantify leukemia burden and monitor improvement or resistance to therapy. A limitation of this study is the lack of molecular correlation for residual NPM1mt and KMT2Ar for all cases examined. Nevertheless, recognition of the immunophenotypic changes associated with multicolor FC remains crucial, as this approach remains the most commonly available method for MRD detection in AML. Finally, our results show that achieving response and MRD-negative remission following treatment with revumenib in R/R AML patients is associated with improved OS.

In conclusion, menin inhibition in AML is associated with immunophenotypic changes detectable by flow cytometry, which should be recognized and, when possible, correlated with molecular methods for optimal disease monitoring

Author contributions

SL: conception and design, data collection, analysis and interpretation, initial manuscript draft, final manuscript preparation, study supervision, and approval of final manuscript. AF: data collection and analysis, final manuscript preparation and approval of final manuscript. TJJ: data collection and analysis, initial manuscript draft, final manuscript preparation and approval of final manuscript. GEH: data collection and analysis, final manuscript preparation and approval of final manuscript. AB: data analysis, final manuscript preparation and approval of final manuscript. SAW: data collection, analysis and interpretation, final manuscript preparation, and approval of final manuscript WW: data collection, analysis and interpretation, final manuscript preparation, and approval of final manuscript GT: data collection, analysis and interpretation, final manuscript preparation, and approval of final manuscript AEQ: data collection, analysis and interpretation, final manuscript preparation, and approval of final manuscript AB: data collection, final manuscript preparation, and approval of final manuscript; BC: data collection, final manuscript preparation, and approval of final manuscript; CDD: data collection, final manuscript preparation, and approval of final manuscript; GMB: data collection, final manuscript preparation, and approval of final manuscript; KT: data collection, final manuscript preparation, and approval of final manuscript; NJS: data collection, final manuscript preparation, and approval of final manuscript; HAA: data collection, final manuscript preparation, and approval of final manuscript; TK: data collection, final manuscript preparation, and approval of final manuscript; FR: data collection, final manuscript preparation, and approval of final manuscript; ND: data collection, final manuscript preparation, and approval of final manuscript; EJ: data collection, final manuscript preparation, and approval of final manuscript; GB: data collection, final manuscript preparation, and approval of final manuscript; MA: final manuscript preparation, and approval of final manuscript; LJM: resources, final manuscript preparation, and approval of final manuscript; HMK: resources, final manuscript preparation, and approval of final manuscript; GCI: conception and design, data collection, analysis and interpretation, initial manuscript draft, final manuscript preparation, study supervision, and approval of final manuscript.

Data availability

Data are available from the corresponding authors upon reasonable request.

Competing interests

SL reports consultancy/ad board funding from: Guidepoint, QualWorld, Gerson Lehrman Group, AlphaSight, Arima, Qiagen, Opinion Health, Abbvie, Daiichi Sankyo, BluePrint Medicine, Caris Diagnostics, Recordati, and Servier; Kura Oncology, Syndax; Grant and contract support from: Astellas, Amgen; Honoraria for lectures from: Pathology Education Partners, VJ HemeOnc, College of American Pathologists, OncLive, ICCS, MD Education, NCCN, MashUp Media, NCTN, Aptitude Health; travel and accommodations support from: Pathology Education Partners, VJ HemeOnc, College of American Pathologists, ICCS, MD Education, NCCN, MashUp Media, Binay Foundation, and stock from: Abbvie. HK reports consulting fees from AbbVie, Amgen, Ascentage Pharma Group Inc., Bristol-Myers Squibb, Curis, Daiichi Sankyo, ImmunoGen, Jazz Pharmaceuticals, Novartis, Pfizer Brazil, Shenzhen TargetRx, Stemline, and Takeda, and honoraria from AbbVie, Amgen, Amphista, Ascentage Pharma Group Inc., Astellas, Infinity BiologiX, Ipsen, KAHR Medical, and Labcorp. GCI received research funding from Kura Oncology, Cullinan Oncology, Syndax Pharmaceuticals, Astex, Novartis, Sumitomo and consultancy or advisory board fees from NuProbe, Pupil Bio, AbbVie, Novartis, Sanofi, AstraZeneca, Syndax and Kura Oncology.

Ethics approval and consent to participate

The authors confirm that all methods were performed in accordance with the relevant institutional guidelines and regulations. The authors confirm that approval was obtained from MDACC IRB (PA17-0485). No live vertebrate studies were performed. The authors confirm that informed consent was obtained from all study participants. No identifiable images are included in this study.

Supplementary information

The online version contains supplementary material available at 10.1038/s41375-026-02905-6.