Introduction

The KIT receptor tyrosine kinase proto-oncogene (KIT) encodes a type 3 transmembrane receptor (KIT, CD117) found on the surface of multiple cell types, including hematopoietic stem cells, mast cells, germ cells, melanocytes, and interstitial cells of Cajal [1]. Activation of KIT through the binding of its ligand stem cell factor (SCF) drives diverse signaling pathways involved in cell survival, growth, proliferation, differentiation, and migration. Various point mutations and in-frame insertions/deletions in KIT result in a constitutively active and SCF-independent oncogenic tyrosine kinase that can induce neoplastic transformation in KIT-expressing cell types [2]. Among hematopoietic neoplasms, KIT mutations are found at highest frequency in mast cell disorders, particularly systemic mastocytosis (SM) (>90% of cases), and the core-binding factor acute myeloid leukemias [CBF-AML: AML with t(8;21) and AML with inv(16) or t(16;16)] (30–40% of cases) [3,4,5]. By comparison, KIT mutations are infrequent in non-CBF AML, myelodysplastic syndromes (MDS), myeloproliferative neoplasms (MPN), or myelodysplastic/myeloproliferative neoplasms (MDS/MPN) such as chronic myelomonocytic leukemia (CMML) [6,7,8,9].

One particular subtype of SM, SM with associated hematological neoplasm (SM-AHN), is distinct in requiring diagnostic evidence of both SM and a separate clonal hematologic non-mast cell lineage disease, which are often detected simultaneously in diagnostic bone marrow (BM) biopsy specimens [10]. Like other subtypes of SM, acquired KIT mutations are common in SM-AHN, particularly the hallmark KITD816V variant, which is found in the majority of cases [11,12,13]. Establishing a diagnosis of SM-AHN can be challenging, especially in situations where the AHN component predominates and obscures the presence of SM. In fact, therapy that targets the non-mast cell neoplasm can reveal the presence of previously unsuspected underlying SM [14, 15].

The hematologic neoplasms most frequently associated with SM-AHN are myeloid diseases, including non-CBF-AML, MDS/MPN (especially CMML), and to a lesser extent MDS and MPN [16, 17]. Despite the high frequency of KITD816V mutations in CBF-AML, it is less often associated with concurrent SM [18, 19]. In contrast, recent work suggests a strong correlation between non-CBF-AML and concurrent SM when KITD816V is detected [20]. However, the diagnostic significance of detecting a pathogenic KIT mutation (p.D816V or otherwise) in other myeloid neoplasms, particularly those with a chronic component (i.e., chronic myeloid neoplasms), is not yet well understood. To address this question, we systematically evaluated a series of BM biopsies from patients with known or suspected myeloid neoplasms associated with pathogenic KIT mutations. Our results demonstrate that most myeloid neoplasms with a chronic course (MDS/MPN, MDS, and MPN) harboring the KITD816V mutation are associated with SM, which may be unrecognized both clinically as well as by pathologists conducting routine histologic evaluation of biopsy specimens. We also found a strong correlation between SM and AML with myelodysplasia-related changes (AML-MRC) bearing the KITD816V mutation, whereas other non-CBF, KITD816V-mutated AML subtypes were only rarely associated with SM in our cohort.

Materials and methods

Cohort selection

Institutional Review Board approval was obtained at all participating institutions. The cohort (n = 64) included all patients seen at Brigham and Women’s Hospital/Dana-Farber Cancer Institute (BWH/DFCI) or Massachusetts General Hospital (MGH) during the time period of July 2011 to February 2019 with each of the following: (a) a pathogenic KIT mutation detected at any point during the disease course in peripheral blood (PB), BM, or cerebrospinal fluid by clinical next-generation sequencing (NGS); and (b) BM biopsy materials available for histologic review and further evaluation as described below. KIT mutations were detected on at least 1 of 4 custom NGS panels as follows: Rapid Heme Panel (KIT exons 8–9, 11, and 17; BWH/DFCI) [21]; OncoMap (KIT hotspots including p.D816V; BWH/DFCI) [22]; OncoPanel (KIT, all exons; BWH/DFCI) [23, 24]; or SNaPshot (KIT exons 1–2, 5, 8–15, 17, and 18; MGH) (see also Supplementary Materials and Methods). Single-gene KIT mutation testing performed as part of the workup for SM in patients without suspicion for another hematologic malignancy was not included. Details of one patient in the cohort have been published previously [25, 26].

Clinicopathologic evaluation

Clinical data, including pathologic diagnoses, patient symptoms, physical exam findings, laboratory results, and disease course and treatment, were obtained from the electronic medical record. When possible, each respective NGS sample was paired with histologic findings from the most proximate BM biopsy. Original pathology reports and slides were reviewed, and case materials including blocks were recovered from institutional archives when required for further immunohistochemistry (IHC). For patients with established SM, pathologic diagnoses were confirmed by re-review of original histologic sections and relevant IHC. For patients without a documented history of SM, available BM biopsy specimens, including when possible those obtained within 6 months of detection of any pathogenic KIT mutation, were reassessed for SM using an IHC panel composed of CD25, KIT (CD117), and/or mast cell tryptase (MCT) if such studies were not performed as part of the original diagnostic workup (see “Immunohistochemistry” section below). Materials from the latter group were independently reviewed by two pathologists with expertise in hematopathology and mast cell neoplasia who were blinded to the diagnosis (JCA; JLH). When appropriate, pathologic diagnoses were revised according to the 2017 WHO Classification of Tumors of Haematopoietic and Lymphoid Tissues (revised 4th edition) [27]. As a control cohort, thirty BM biopsies from patients diagnosed with either CMML (n = 15) or MDS (n = 15) without pathogenic KIT mutations, randomly selected from the same study period, were included in the blinded review.

Immunohistochemistry

IHC for MCT [mouse monoclonal clone AA1 (DAKO Corp, Carpenteria, CA), titer 1:4000, retrieval 10 min with Enzyme 1 (Leica Microsystems, Buffalo Grove, IL)], and CD25 [mouse monoclonal clone 4C9 (Leica), titer 1:50, retrieval 30 min with Bond Epitope Retrieval Solution 2 (Leica)] were performed on the Leica immunostainer (BOND III) using a 60 min primary antibody incubation period and the BOND Polymer Refine DAB Detection system (Leica). IHC for KIT [rabbit polyclonal (DAKO), titer 1:300, no retrieval] was performed manually using a 50 min primary antibody incubation period with detection performed using Rabbit PowerVision (Leica) for 30 min; staining was developed with the DAB+ kit (DAKO). For all studies, slides were counterstained manually with either hematoxylin or methyl green, dehydrated through solvents, and cover slipped.

Statistical analysis

When appropriate, Fisher’s exact test (two-tailed) was used to assess conditional distribution probabilities associated with the presence or absence of SM (https://www.graphpad.com/quickcalcs/). P values < 0.05 were considered statistically significant.

Results

Pathogenic KIT mutations

Our final cohort (n = 64) included all patients in whom a pathogenic KIT mutation was detected at one or more timepoints by clinical NGS sequencing and for whom BM biopsy materials were available to evaluate for involvement by SM. The canonical p.D816V KIT variant was the most common KIT mutation detected within our patient cohort (n = 52, 81%), including one patient with an additional p.D816Y KIT mutation. Samples from the remaining 12 patients exhibited one or more pathogenic non-D816V KIT mutations without a concurrent KITD816V mutation.

KIT D816V mutation in myeloid neoplasms

Forty of the 52 patients with a KITD816V mutation were diagnosed with a myeloid neoplasm other than isolated SM (Fig. 1a, Supplementary Table 1). This included 22 patients with chronic myeloid neoplasms (MDS, MPN, or MDS/MPN) and 18 patients with AML. Based on comprehensive IHC testing for CD25, KIT, and/or MCT and expert evaluation, nearly all (n = 20, 91%) chronic myeloid neoplasms with a KITD816V mutation fulfilled criteria for SM-AHN, whereas only 8 of 18 (44%) AML cases fulfilled criteria for SM-AHN (p = 0.0021), as described below.

Fig. 1: Systemic mastocytosis in patients with myeloid neoplasms and KIT mutations.
figure 1

a Forty patients in the cohort had a KITD816V mutation detected by next-generation sequencing in the setting of an acute or chronic myeloid neoplasm. Twenty of twenty-two (91%) patients with a chronic myeloid neoplasm (chronic myelomonocytic leukemia (CMML), myelodysplastic/myeloproliferative neoplasm, unclassifiable (MDS/MPN-U), myelodysplastic syndrome (MDS) with single lineage dysplasia (SLD), multilineage dysplasia (MLD), or excess blasts-1 (EB-1), myeloproliferative neoplasm (MPN) including polycythemia vera (PV), primary myelofibrosis (PMF), and MPN unclassifiable (MPN-U)) also had systemic mastocytosis (SM), while only 8 of 18 (44%) patients with acute myeloid leukemia (AML; AML with myelodysplasia-related changes, AML-MRC; AML with mutated NPM1, AML-NPM1; AML, not otherwise specified, AML, NOS; core binding factor-associated AML, CBF-AML) also had SM (p = 0.0021). b Eleven patients in the cohort had a pathogenic KIT mutation other than the p.D816V variant in the setting of a myeloid neoplasm, all of which were AML; two patients exhibited evidence of SM, both CBF-AML.

Full size image

KIT D816V mutations in chronic myeloid neoplasms

The 20 patients with SM-AHN where the AHN component was a chronic myeloid neoplasm (MDS-single lineage dysplasia, n = 2; MDS-multilineage dysplasia, n = 2; CMML, n = 9; MDS/MPN, unclassifiable (MDS/MPN-U), n = 4; polycythemia vera, n = 1; primary myelofibrosis, n = 1; MPN, unclassifiable, n = 1) included three patients with a preceding (>1 year prior) diagnosis of SM [aggressive SM (ASM), n = 2; indolent SM (ISM), n = 1], two patients with a preceding (>1 year prior) diagnosis of the myeloid neoplasm, and 15 patients with concurrently diagnosed SM-AHN (Supplementary Table 1). The KITD816V mutation was detected in the first NGS test for all of these patients, and these initial time points were nearly always (n = 17, 85%) obtained within 6 months of the SM-AHN diagnosis. In all but one case, at least one additional pathogenic non-KIT mutation was also identified in the first NGS test (Fig. 2). In both PB and BM specimens, the KITD816 variant allele frequency (VAF) was observed at lower or equivalent (within 5%) levels compared with the non-KIT mutations (Supplementary Table 1).

Fig. 2
figure 2

Comutation plot of all pathogenic mutations by next-generation sequencing from 20 patients in whom a KITD816V variant was detected, at closest timepoint to initial diagnosis of systemic mastocytosis with associated hematological neoplasm (SM-AHN), where AHN = chronic myelomonocytic leukemia (CMML); myelodysplastic syndrome/myeloproliferative neoplasm, unclassifiable (MDS/MPN-U); myelodysplastic syndrome (MDS); myeloproliferative neoplasm (MPN). Each column represents a single patient.

Full size image

BM biopsies from the two remaining patients with a chronic myeloid neoplasm and KITD816V mutation (MDS/MPN-U and MDS-excess blasts-1) lacked clinicopathologic evidence of SM following immunohistologic investigation and expert review. The KITD816V mutation was detected in the first NGS test for one of these patients. For the other patient, the KITD816V mutation was not present at the time of the initial NGS test (which identified IDH2R140Q [45.4%], ASXL1G642fs* [46.7%], CSF3RT618I [6.3%], and NRASG12D [29.4%] driver mutations) but emerged later, following treatment with a hypomethylating agent and an IDH inhibitor, as the dominant clone (subsequent NGS test; KITD816V [29.2%], IDH2R140Q [1%], ASXL1G642fs* [10%]).

Histopathologic investigation of 30 randomly-selected MDS or CMML cases without a pathogenic KIT mutation (p.D816V or otherwise) showed no evidence of concurrent SM (data not shown).

KIT D816V mutations in acute myeloid leukemia

Eighteen patients in the KITD816V cohort were diagnosed with AML at some point during their disease course; diagnoses included AML-MRC (n = 9), AML with mutated NPM1 (n = 5), AML, not otherwise specified (AML, NOS, n = 2), CBF-AML (n = 1), and therapy-related AML (n = 1) (Fig. 1a, Supplementary Table 1). Of the nine patients ultimately diagnosed with AML-MRC, the diagnosis of AML-MRC was based on a prior history of MDS (n = 3) or MDS/MPN (n = 3), AML-MRC-defining karyotype (n = 1) or based only on the presence of AML-MRC-defining morphologic dysplasia (n = 2). Of these nine patients, seven (78%) had known (n = 3) or were concurrently diagnosed with (n = 4) SM at the time of AML-MRC diagnosis. In contrast, only one (11%) of nine patients diagnosed with non-MRC AML had evidence of SM (p = 0.0152).

Of the seven patients with SM-AHN where the AHN component was AML-MRC, three had NGS performed at multiple time points including before and after development of AML-MRC. In these three patients, the KITD816V mutation decreased in VAF or became undetectable at the time of AML diagnosis, whereas the other driver mutation VAFs remained stable (Supplementary Fig. 1). In two of these cases, additional driver mutations in genes encoding epigenetic modifiers and signaling proteins were detected at AML diagnosis. Three additional patients with SM-AHN where the AHN component was AML-MRC had only one NGS timepoint. Detailed NGS data with VAF were only available for two of them, and in both cases, KITD816V was present at a lower VAF compared with other non-KIT mutations (Supplementary Table 1), suggesting that the KITD816V was present in a subclone. The seventh patient in this group had gastrointestinal mast cell infiltrates without malabsorption (ISM) in addition to maculopapular cutaneous mastocytosis (MPCM) but did not exhibit BM involvement by SM at any point in the disease course. The KITD816V mutation, which had been previously detected by PB qualitative allele-specific polymerase chain reaction (PCR) at time of MPCM diagnosis, was detected by PB NGS at a lower VAF (3%) compared with an IDH2R140Q mutation (36%) at the time of AML-MRC diagnosis; that KIT mutation was not detected in subsequent marrow samples showing persistent AML-MRC and IDH2R140Q.

The two AML-MRC patients without SM exhibited scattered abnormal mast cells with CD25 co-expression, but the overall findings failed to definitively meet current WHO diagnostic criteria for SM (Supplementary Table 1). Pretransformation molecular data is only available for one of these patients, and in that case, the KITD816V mutation was not present at the time of the initial diagnosis of MDS and was only detected in a secondary AML that arose following allogeneic transplant.

Of the remaining nine patients with AML (non-MRC), eight did not have any clinicopathologic evidence of SM. In three patients (two AML with mutated NPM1, one AML-CBF), the KITD816V mutation was present at the time of AML diagnosis. In one patient (AML, NOS), the KITD816V mutation was detected subsequent to the initial diagnosis and then persisted until the time of death. In four cases (two AML with mutated NPM1, one AML, NOS, one therapy-related AML), the KITD816V mutation was only detected in the postallogeneic transplant setting; in these patients, death occurred within 4 months of detection of the mutation. Scattered abnormal mast cells with CD25 co-expression were noted in the biopsy from the patient with therapy-related AML, but the overall findings were insufficient for a current WHO diagnosis of SM (Supplementary Table 1). The final patient in this group, with a history of ASM, developed AML with mutated NPM1 during SM-directed therapy; at the time of AML diagnosis, there was no evidence of SM in the BM and the KIT mutation was no longer detectable.

Systemic mastocytosis without an associated hematological neoplasm

The remaining 12 patients within the KITD816V cohort were diagnosed with SM only, without histologic evidence of an AHN following expert review. Eight of these 12 cases had no additional mutations detected by clinical NGS sequencing. In the remaining four cases, including two cases of mast cell leukemia, additional pathogenic mutations in DNMT3A, TET2, and/or ATM were detected (data not shown).

Non-D816V KIT mutations

Twelve patients from the cohort produced samples containing one or more pathogenic non-D816V KIT mutations without a concurrent KITD816V mutation. One patient was diagnosed with mast cell leukemia (KITD816Y). The other 11 patients included five CBF-AML with t(8;21) [p.418_419insFF (n = 1); p.V559D and p.D816Y (n = 1); p.Y823C (n = 1); p.D816Y (n = 1); p.D816Y and p.D816H (n = 1)]; two CBF-AML with inv(16) or t(16;16) [p.417_419delTYDinsI (n = 1); p.419delD (n = 1)]; two AML, NOS [p.D816Y (n = 1); p.D820Y (n = 1)]; one AML-MRC (p.419delD); and one acute promyelocytic leukemia (p.N822K) (Fig. 1b, Supplementary Table 2). Only two of these patients had clinicopathologic evidence of SM, both of whom had CBF-AML (p.418_419insFF; p.D816Y and p.V559D); the patient with p.418_419insFF exhibited well-differentiated SM without aberrant CD25 expression in the mast cell aggregates [28]. One patient with AML, NOS had scattered abnormal mast cells with CD25 co-expression, not fulfilling current WHO criteria for SM (Supplementary Table 1). All KIT mutations detected within the non-D816V KIT mutation cohort were present at the first NGS timepoint. At least one additional non-KIT pathogenic mutation was identified in samples from 8 of these 12 patients (Supplementary Fig. 2). The maximal non-D816V KIT VAF ranged from subclonal to within 10% of the levels of other non-KIT mutations.

Unsuspected SM

For the majority of patients with SM described in this study, the diagnosis of SM was already known or suspected before the corresponding KIT mutation result was available, based on clinicopathologic findings. However, for nearly one-third (9 of 30) of SM-AHN patients in this study, there was no clinical or pathologic suspicion of a systemic mast cell disorder, and a diagnosis of SM-AHN was not rendered upon initial pathologic review (Table 1). In seven of these cases, a mast cell infiltrate was not initially appreciated on hematoxylin and eosin (H&E)-stained sections and was instead revealed by IHC performed during case workup or as part of this study (Fig. 3). The average degree of marrow involvement by lesional mast cells in these patients was 10% of the cellularity (range, <5–30%). In two additional cases, the mast cell infiltrate was only identified on deeper tissue sections (Fig. 4).

Table 1 Clinicopathologic characteristics of patients with unsuspected systemic mastocytosis.
Full size table
Fig. 3: Bone marrow biopsy from a patient with chronic myelomonocytic leukemia and systemic mastocytosis.
figure 3

a The abnormal mast cells were not appreciated on initial evaluation, given extensive involvement by chronic myelomonocytic leukemia (hematoxylin and eosin (H&E), 20×). b–e Additional immunostaining revealed large clusters and loose aggregates of spindled, CD25-positive mast cells (b H&E; c CD117; d mast cell tryptase; e CD25; all 200×). Some degranulated forms are not highlighted by mast cell tryptase (d).

Full size image
Fig. 4: Bone marrow biopsy from a patient with myelodysplastic syndrome and systemic mastocytosis.
figure 4

a The abnormal mast cells were not present on initial levels of this bone marrow biopsy (hematoxylin and eosin (H&E), 20×). b, c Deeper levels revealed an abnormal paratrabecular aggregate of spindled cells associated with a lymphoplasmacytic inflammatory infiltrate (H&E; b 20×; c 200×). d–f Additional immunostaining revealed aberrant CD25 expression in the mast cell aggregates (d CD117; e mast cell tryptase; f CD25; all 200×). Some degranulated forms are not highlighted by mast cell tryptase (e).

Full size image

Review of the clinical records of these patients revealed one patient with a prior history of MPCM; this patient had KITD816V detected by PB digital droplet PCR at the time of MPCM diagnosis but did not undergo BM biopsy with NGS until 13 months later for cytopenias. Splenomegaly was present in 7 of 9 patients, elevated tryptase was present in all tested patients, and four patients had “C” findings [29]. Despite the associated myeloid neoplasm, karyotype was normal in 7 of 8 tested patients. All cases except one exhibited additional non-KIT driver mutations by NGS testing. The KITD816V VAF ranged from substantially lower to equivalent to the VAFs of other non-KIT mutations, implying that the KIT mutated clone may be either subclonal or dominant.

Discussion

SM-AHN can be challenging to recognize and establish as a diagnosis due to variability in clinical presentation and histologic appearance, which collectively spans the entire spectrum of hematologic neoplasia. Moreover, many clinical and laboratory findings of SM (e.g., splenomegaly and elevated serum tryptase level) and certain associated morphologic features (e.g., BM fibrosis) also occur frequently in non-mast cell hematologic disease. Thus, in the absence of clinical or pathologic suspicion, SM may remain unrecognized in patients with other hematologic neoplasms.

KIT mutations have long been associated with AML, particularly CBF-AML, where they have important prognostic and therapeutic implications. While KITD816V-mutated CBF-AML is usually not associated with SM, non-CBF-AML with KITD816V was recently reported to correlate strongly with the presence of concurrent SM [20]. However, the implication of detecting a KIT mutation in the setting of other myeloid neoplasms, especially those with a chronic course, has not been thoroughly investigated. We took advantage of the inclusion of KIT in our clinical NGS panels to systematically ask whether the presence of a KIT mutation (p.D816V or otherwise) in myeloid neoplasms correlated with concurrent SM.

Of the 22 chronic myeloid neoplasms (MDS, MPN, or MDS/MPN) with a KITD816V mutation detected by our clinical NGS panels over an ~8-year period, 20 (91%) exhibited concurrent SM on BM biopsy examination. In contrast, thorough pathologic investigation of 30 randomly-selected MDS or CMML cases without a pathogenic KIT mutation (p.D816V or otherwise) showed no evidence of concurrent SM. These results indicate that the detection of KITD816V by NGS in patients with chronic myeloid neoplasms is a powerful predictor of SM-AHN.

In all 20 patients with SM-AHN where the AHN component was a chronic myeloid neoplasm in our cohort, the KITD816V mutation was present at the time of initial NGS testing at subclonal or equivalent levels to other pathogenic driver mutations. Notably, the DNA source for 10 of these 20 initial NGS tests was PB, suggesting that the KITD816V mutation was present in circulating myeloid cells. Indeed, prior studies have demonstrated that at least a subset of AHN cells also share the KITD816V mutation with the neoplastic mast cells. In one study of 48 patients with SM-AHN, the KITD816V mutation was found in the AHN cells of nearly all patients with CMML (89%), although less frequently in the AHN cells of MPN (20%) or AML (30%) [30]. Given the presence of KIT mutations within the AHN component, it is not possible to use the KIT VAF to predict SM burden in SM-AHN. Prior investigations have shown that the KITD816V mutation is a temporally late (subclonal) event in clonal evolution following mutations in TET2, SRSF2, and ASXL1. Additional studies of granulocyte-macrophage colony-forming progenitor cell units from patients with SM-AHN suggest that the KITD816V mutation is also an ontologically late event in many cases, occurring only in more differentiated cells driving the SM phenotype [31]. The KITD816V mutation was invariably present at the first NGS timepoint in the current study, along with TET2, SRSF2, and/or ASXL1 mutations detected in 18 of 20 cases (90%). Due to the lack of comprehensive serial samples, we are unable to properly assess the timing of KIT mutation acquisition in this study; however, VAF data from our cohort appears to support the subclonal and therefore temporally later acquisition of KIT mutations in SM-AHN.

In our cohort of 18 patients with KITD816V-mutated AML, 17 (94%) were non-CBF forms of AML. Fritsche-Polanz et al. identified SM in 8 of 101 patients with AML, of which seven had a KITD816V mutation. Interestingly, 6 of the 7 patients with a KITD816V mutation had an antecedent history of either MDS or CMML [32]. Similarly, in our KITD816V-mutated AML cohort, SM was much more commonly identified in cases of AML-MRC (7 of 9, including four with known preceding MDS or MPD/MPN; 78%) compared with non-MRC AML (1 of 9, 11%; p = 0.0152). Another study investigated patients with KITD816V-mutated non-CBF AML and SM [20]. Of 40 such patients, the majority (24; 60%) were diagnosed with AML-MRC following a prior SM-AHN, where the AHN was MDS/MPN-U, CMML, MDS, or MPN. In addition, 11 of the 40 patients (28%) were diagnosed with de novo AML. In our cohort, no patients with de novo AML (n = 7; four AML with mutated NPM1, two AML-NOS, and one CBF-AML) or therapy-related AML (n = 1) and a KITD816V mutation were diagnosed with SM. Interestingly, in 4 of these 8 patients, the KIT mutation only appeared in the setting of posttransplant relapse with a survival of less than 4 months in all patients following detection of the mutation. Similarly, one of the AML-MRC patients without SM acquired the KITD816V mutation at relapse after allogeneic transplantation, with patient death 3 months later; in all other AML-MRC cases, the KIT mutation was present at the time of initial diagnosis. Taken together, the findings suggest that KITD816V is far more frequently associated with SM in the setting of AML-MRC than other non-CBF-AML subtypes, where acquisition of KITD816V appears to often be a late event associated with relapsed AML without concurrent SM.

The diagnostic criteria for SM in the setting of an AHN require that either the major criterion (≥15 mast cells in aggregates) and at least one minor criterion (>25% spindled/atypical mast cells; presence of KIT codon 816 mutation; aberrant co-expression of CD25 by mast cells), or at least three minor criteria, are met [27]. Of note, persistently elevated serum total tryptase is not a valid minor criterion in the setting of an AHN. Cases diagnosed based solely on minor criteria typically show a diffuse interstitial pattern and are challenging to recognize in BM biopsies stained with H&E alone [33,34,35]. Although Giemsa staining can be a helpful adjunct to H&E staining for recognition of mast cells in routine cases (and is performed routinely on all BM cases at our institutions), neoplastic mast cells may be difficult to identify on Giemsa stain due to degranulation. In the course of our study, we identified four cases with subtle (well <5% of the cellularity in most cases) abnormal populations of interstitially-distributed mast cells that showed CD25 co-expression and at least subset (~5–10%) spindled morphology, but not definitively >25% of BM mast cells, on the BM core (and not apparent on the aspirate smears). This included one case of AML, NOS with a KITD820Y mutation, two cases of AML-MRC with KITD816V mutation, and one case of therapy-related AML with a KITD816V mutation. None of these patients exhibited splenomegaly or “C” findings, and tryptase was not measured in any case. The significance of these abnormal mast cell populations is uncertain but suggests a propensity for KIT mutations to drive occasionally more subtle SM-like phenotypic changes that fall short of current WHO diagnostic criteria for SM-AHN.

Notably, inclusion criteria for this study required evaluation on an NGS platform, as opposed to single-gene KIT assays such as allele-specific PCR or digital droplet PCR. Indeed, the reason that there are relatively few patients with SM alone in this NGS-based cohort is that, since 2016, patients at our institutions who are suspected or known to have SM without suspicion for AHN typically undergo PB analysis for the KITD816V mutation by ultrasensitive digital droplet PCR rather than NGS [36]. Although a definitive conclusion is beyond the scope of this study, our findings suggest that detection of KITD816V in the PB using a less sensitive testing technique (i.e., NGS) should raise suspicion for an underlying SM-AHN, as it may reflect the presence of circulating myeloid cells bearing pathogenic KIT mutations. On the other hand, had a more sensitive testing technique been applied to our target population, it is conceivable that we may have detected additional patients with lower levels of SM involvement, or encountered a threshold KITD816V VAF below which the association between KITD816V and concurrent SM was no longer apparent, either due to the detection of minor/transient subclones or the inability of low-level SM-like disease to meet current diagnostic criteria.

In this study, we identified nine patients (seven with a chronic myeloid neoplasm; two with AML-MRC) in whom the SM component of the SM-AHN diagnosis was not recognized upon initial BM evaluation. The challenge of identifying low-level or patchy involvement by the SM component in SM-AHN is well-established [14], but in many of our cases the SM component was not suspected even after both the clinician and pathologist became aware of the KITD816V mutation. This is perhaps not surprising, as KIT mutations have long been identified in CBF-AML without necessarily being indicative of associated SM. However, in the setting of MDS, MPN, MDS/MPN, or AML-MRC, our findings indicate that detection of the KITD816V should further prompt investigation for SM, even in the absence of overt histologic evidence. Given the availability of therapies that specifically target the SM component of SM-AHN [37,38,39], establishing this diagnosis has substantial clinical significance, particularly in patients more with aggressive forms of SM.

In summary, our data indicate that identification of KITD816V in patients diagnosed with chronic myeloid neoplasms or AML-MRC should prompt additional testing for underlying SM. In non-CBF, non-MRC AML, acquisition of the KITD816V mutation appears to often be a late event, occurring in the posttransplant setting, and is not frequently associated with SM. However, as the co-occurrence of SM in cases of KITD816V-mutated, non-CBF de novo AML has also been reported [20], careful clinicopathologic evaluation for SM is also indicated in this setting. Although the numbers are small, SM was not common in patients with non-D816V KIT mutations in our cohort. Nearly all (7 of 8) CBF-AML cases bearing a KIT mutation in our cohort had non-D816V KIT mutations in either exons 8 or 17, similar to prior studies [40, 41], and only two (both non-D816V KIT mutated) were associated with SM. One possible explanation is that the capacity of non-D816V KIT mutations to induce mast cell differentiation and modulate the phenotype of affected clones is simply lower than that of KITD816V. Further studies are needed to investigate this possibility.