Myeloid sarcoma shows a high frequency of mutations activating the MAPK/ERK pathway and association with clonal hematopoiesis

Abstract

Myeloid sarcoma (MS) is a mass‐forming extramedullary manifestation of myeloid blasts, either in relation to an underlying acute myeloid leukemia (AML), another myeloid neoplasm (MN) or as a de novo occurrence. Data on the genetic profile of MS are sparse. In this study, 41 MS of 34 patients, including 7 de novo cases and 24 patients with antecedent or synchronous MN, were analyzed with targeted next‐generation sequencing (NGS), RNA‐based fusion detection, and gene expression profiling (GEP). In 10 patients, a MS developed after stem cell transplantation for MN. Additionally, 21 available pre‐transplant bone marrow biopsies (BMB) from 20 patients and 6 post‐transplant BMB from 6 patients were investigated. The most frequently mutated gene was TET2 (41%), followed by NPM1 (38%) and NRAS (35%). Overall, 74% of the cases exhibited mutations affecting the MAPK/ERK pathway. AML‐type fusions were detected in seven MS patients, who were younger than those without fusions (median 49 versus 67 years). Nine of 13 patients with a MN and available pre‐transplant BMB showed additional mutations restricted to the MS, including an additional NRAS mutation in 3/5 cases with AML. Five of seven of patients with pre‐transplant BMB without evidence of a MN revealed clonal hematopoiesis (CH), mostly shared TET2 mutations. Comparative GEP between BM and MS revealed upregulation of the MAPK/ERK pathway in MS and of gene sets relevant for interaction with the microenvironment. In conclusion, MS is characterized by a high incidence of MAPK/ERK pathway mutations and activation, frequent clonal evolution, and association with CH in elderly patients without recurrent AML‐type fusions.

Introduction

Myeloid sarcoma (MS) is a mass‐forming extramedullary manifestation of myeloid blasts with effacement of the tissue [1, 2, 3]. MS can occur throughout the body, but the most frequently affected locations are the skin, soft tissue, lymph nodes, gastrointestinal tract, bone, and the head‐and‐neck region [2]. Most affected patients are between 40 and 70 years, but every age can be affected, and there is a predilection for males [2, 4].

MS are most often observed in association with an acute myeloid leukemia (AML), either concomitant to the primary diagnosis or as a relapse, including after allogeneic stem cell transplant (SCT). Furthermore, MS may represent progression from another myeloid neoplasm (MN), including myelodysplastic syndromes (MDS), myeloproliferative neoplasms (MPN), and myelodysplastic/myeloproliferative neoplasms (MDS/MPN). In a smaller subset, MS can present as an isolated extramedullary mass without evidence of systemic disease [2, 4].

As in AML, the immunohistochemical profile of MS is quite variable, with significant immunophenotypic heterogeneity due to different maturation stages and involved lineages. Commonly expressed markers include CD33, CD43, lysozyme, myeloperoxidase, and CD68, whereas stem cell markers CD34 and CD117 may be more variably expressed, and their absence may result in diagnostic difficulties [2, 4, 5, 6, 7, 8, 9].

In contrast to the extensive literature on AML, genetic information on MS is more limited due to practical restrictions for molecular and cytogenetic studies. Overall, published studies of MS indicate a cytogenetic and molecular profile similar to AML, with a broad range of primary and secondary alterations. In comparison to AML [10], MS has shown a higher frequency of NPM1 and NRAS mutations [11].

Clonal cytogenetic alterations are detected in up to three‐fourths of MS, with widely varying percentages in different studies, likely due to different technical approaches. The most frequent alteration is t(8;21)/RUNX1::RUNX1T1, followed by trisomy 8, inv(16)(p13q22)/CBFB::MYH11 fusion, 11q23/KMT2A rearrangement, and monosomy 7 [7, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22].

Although the few studies comparing genetic findings in MS and matched bone marrow (BM) showed an overall high concordance of the mutation profile between both sites [23, 24], a discordant molecular profile was associated with significantly worse overall survival in one study [25]. In addition, the number of mutations was associated with outcome [24].

A study focusing on NPM1‐mutated AMLs with and without MS development showed a tendency for a more complex karyotype and more MDS‐related gene alterations, as well as fewer mutations in genes regulating DNA methylation in NPM1+ AML with MS development [26]. A high frequency of NFE2 mutations has been described in de novo MS, and data from a mouse model from the same group suggested that NFE2 activity could induce the homing of neoplastic stem cells to an extramedullary site [27, 28].

Given the limited published data on the genetics of MS, we performed a comprehensive NGS‐based study on the mutational, gene expression, and gene fusion spectrum of MS in different clinical settings and compared these results to matched BM biopsies (BMB), irrespective of the presence or absence of BM involvement by AML or another MN.

Materials and methods

Sample selection

A total of 41 formalin‐fixed paraffin‐embedded (FFPE) extra‐medullary samples of 34 patients with a diagnosis of MS (including a case of a borderline lesion between MS and histiocytic sarcoma) were retrieved from the archives of the Institute of Pathology and Neuropathology and the division of dermatopathology at the University Hospital of Tübingen, as well as from the institutes of pathology, Technical University Munich, Munich, Paracelsus Medical University Salzburg, Austria, and Biopticka Laboratory, Pilzen, Czech Republic. The cases had been diagnosed between 2001 and 2023. Eleven patients underwent an allogenic SCT (32%). Furthermore, 21 pre‐transplant BM biopsies (BMB) from 20 patients of our cohort from any time point were retrieved and analyzed in parallel. Additionally, six BMB after a SCT at the time of MS diagnosis were also available. All BMB had been formalin‐fixed, decalcified in ethylenediaminetetraacetic acid (EDTA) for 7–12 h at 37 °C and embedded in paraffin wax. The study was conducted in accordance with the Declaration of Helsinki and was approved by the local Ethics Review Committee and the institutional review board panels of the contributing institutions (335/2021BO2). Patients provided informed consent. For older samples, a waiver was granted by the local Ethics Committee.

Histopathological and immunohistochemical assessment

All samples were reviewed and reclassified according to the current International Consensus Classification [3] and the 5th WHO classification [2] by two experienced hematopathologists (DN and FF). The following stains were performed, if they had not already been part of the initial workup: CD34 (DAKO, 1:100), CD117 (Roche, ready‐to‐use), lysozyme (DAKO, 1:8000), CD56 (Medac, 1:300), MPO (DAKO, 1:25,000), CD33 (Novocastra, 1:200), and NPM1 (DAKO, ready‐to‐use). Immunohistochemistry was performed on the Ventana Ultra automated staining System (Ventana Medical Systems, Tucson, AZ, USA) using Ventana reagents, according to the manufacturer's protocol [29].

DNA and RNA isolation

Genomic DNA and RNA were extracted from macrodissected 5 μm paraffin sections using the Maxwell® RSC DNA FFPE Kit, the Maxwell® RSC RNA FFPE Kit, and the Maxwell® RSC Instrument (Promega, Madison, WI, USA) according to the manufacturer's instructions.

Next‐generation sequencing for mutation, fusion, and copy number variation analysis

Targeted mutation analysis, fusion gene detection, and targeted copy number variation (CNV) testing on autosomal chromosomes were performed by next‐generation sequencing (NGS) (Ion GeneStudio S5 prime, Thermo Fisher Scientific, Waltham, MA, USA) using the Oncomine myeloid research assay (Thermo Fisher Scientific). The assay examines 40 genes (covering hotspots in 23 genes and the complete coding sequence in 17 genes), 29 gene fusions recurrent in MN, and targeted copy number alterations (for more detail see supplementary materials and methods). For the detection of NFE2 alterations, an additional NFE2 custom panel covering the complete coding sequence and also including NRAS and KRAS was designed using the Ion AmpliSeq™ Designer from Thermo Fisher Scientific (supplementary material, Table S1).

Gene expression profiling

Gene expression profiling (GEP) was performed using HTG EdgeSeq Technology (HTG Molecular Diagnostics, Inc., Tucson, AZ, USA). The HTG EdgeSeq Oncology Biomarker Panel comprising 2,549 genes associated with oncology and molecular pathways was used according to the manufacturer's manual. Libraries were sequenced on the Ion GeneStudio S5 Prime (Thermo Fisher Scientific, Waltham, MA, USA) (for more details see supplementary materials and methods).

Statistical analysis

Statistical analysis was conducted with the Fisher's exact test or the Mann–Whitney U test; p values <0.05 were considered significant.

Results

Clinical features

The clinicopathological features are summarized in Tables 1 and 2. The study included 34 patients; 18 men and 16 women (M:F 1.1:1) with a median age at onset of MS of 64.5 years (range 1–90 years), including two children with an age of 1 and 6 years. Four patients had a history of cytotoxic therapy for lymphoma or solid cancer. MS cases were separated in four categories, depending on clinical presentation: (1) ‘metachronous post AML’ with recurrence of known AML as MS in 12/34 patients (35%), (2) ‘synchronous with AML’ in 2/34 (6%) patients with simultaneous manifestation of AML and MS, (3) ‘isolated MS’ in 7/34 (21%) patients without history of a MN, and (4) ‘progression from other MN’ in 10/34 (29%) cases with known history of MDS, MPN, or MDS/MPN. In three (9%) patients, this information was lacking. In seven patients, two extramedullary manifestations were available, resulting in 41 MS samples. In 10 cases with a metachronous appearance after AML, the median time between the first AML manifestation and the MS was 19.5 months (range: 4–223 months). The 10 cases with a progression from a non‐AML MN consisted of 5 MDS/MPN, including 3 cases of chronic myelomonocytic leukemia (CMML), 4 MDS, and 1 patient with primary myelofibrosis (PMF). The median time between the first MN manifestation and the MS was 26 months (range: 5–87 months). Ten MS developed after SCT for their MN; two of these patients already had an initial MS manifestation before SCT. In the remaining eight cases with primary MS after SCT, the median time of manifestation from SCT was 43.5 months (range 4–171 months).

Table 1.

Patient characteristics

Case	Sex	Age (years)	Primary site of MS	Type of manifestation	Prior diagnoses
1 †	F	58	Breast	Metachronous post AML	sAML after PMF
2 †	M	89	Skin	Progression from MDS, MPN, or MDS/MPN	MDS‐IB1/EB
3 ‡	F	66	Pelvis soft tissue	Metachronous post AML	AML
4*	M	50	Thorax soft tissue/skin	Metachronous post AML	AML
6* , †	M	81	Skin/lymph node	Metachronous post AML	AML § , T‐NHL
7 ‡	F	49	Breast	Metachronous post AML	AML
8* , †	F	64	Skin/lymph node	Isolated MS	–
9 †	M	68	Skin	Isolated MS	–
10* , †	M	85	Skin/eye lid	Isolated MS	–
11 †	F	54	Lymph node	Metachronous post AML	AML
12 †	M	67	Skin	Progression from MDS, MPN or MDS/MPN	CMML
13 ‡	F	57	Adrenal gland	Progression from MDS, MPN or MDS/MPN	PMF
14	F	6	Soft tissue head	Isolated MS	–
15 †	M	72	Small intestine	Isolated MS	–
16	F	48	Paravertebral	Isolated MS	–
17	M	56	Lymph node	Metachronous post AML	AML
18* , ‡	M	67	Scrotal/penis	Progression from MDS, MPN or MDS/MPN	CMML
19	F	78	Skin	Unknown	Unknown
20 †	M	74	Spleen	Progression from MDS, MPN or MDS/MPN	MDS/MPN with EB
22 ‡	M	52	Gingiva	Metachronous post AML	AML
23* , †	F	73	Eye/gingiva	Progression from MDS, MPN or MDS/MPN	MDS/MPN
24 †	M	62	Thorax	Metachronous post AML	AML
25 †	F	65	Liver	Progression from MDS, MPN or MDS/MPN	MDS § with fibrosis, ovarian carcinoma
27* , †	F	70	Skin/skin	Metachronous post AML	AML
28	F	57	Neck	Isolated MS	–
29 †	M	51	Middle ear	Concomitant with AML	AML
30 †	M	62	Testis	Progression from MDS, MPN or MDS/MPN	MDS
31 †	M	79	Skin	Concomitant with AML	AML
32 ‡	M	59	Skin	Metachronous post AML	AML
33	M	60	Spinal canal	Progression from MDS, MPN or MDS/MPN	MDS § , DLBCL
34	F	1	Brain	Metachronous post AML	AML
35 †	F	70	Bone	Progression from MDS, MPN or MDS/MPN	CMML § , breast carcinoma
36	M	90	Skin	Unknown	Unknown
37	F	81	Pharynx	Unknown	Unknown

Type of manifestations: metachronous: metachronous after known AML; progression: progression of MDS, MPN, or MDS/MPN to MS without any history of AML; isolated: isolated MS without any history of AML; concomitant: MS with concomitant AML. Cases 5, 21, and 26 were eliminated after pathology review.

AML, acute myeloid leukemia; CMML, chronic myelomonocytic leukemia; DLBCL, diffuse large B‐cell lymphoma; F, female; M, male; MDS, myelodysplastic syndrome; MDS‐IB1/EB, myelodysplastic syndrome with increased blasts/excess blasts; MDS/MPN, myelodysplastic/myeloproliferative neoplasm; MPN, myeloproliferative neoplasm; MS, myeloid sarcoma; PMF, primary myelofibrosis; sAML, secondary acute myeloid leukemia; T‐NHL, T‐cell non‐Hodgkin lymphoma.

^*Cases with two myeloid sarcomas.

^† Cases with additional pre‐transplant bone marrow biopsy.

^‡ Cases with additional post‐transplant bone marrow biopsy.

^§ Cases with prior cytotoxic therapy – WHO: ‘post cytotoxic therapy’ and ICC: ‘therapy‐related’.

Table 2.

Summary of patient and immunohistochemical characteristics

Characteristic	Value
Number of patients	34
Age	Median 64.5 years (range 1–90 years)
Sex	M:F 1.1:1
Appearance
Metachronous post AML	12/34 (35%)
Progression from MDS, MPN or MDS/MPN	10/34 (29%)
Isolated myeloid sarcoma	7/34 (21%)
Concomitant AML	2/34 (6%)
Unknown	3/34 (9%)
Immunohistochemistry
CD33 and/or lysozyme	34/34 (100%)
CD33	31/31 (100%)
Lysozyme	26/30 (87%)
NPM1 cytoplasmic	9/14 (64%)
CD117	18/29 (62%)
CD56	12/22 (55%)
MPO	10/28 (36%)
CD34	11/34 (32%)

The most commonly affected site was the skin, in 11/33 (33%) cases (Figure 1); followed by two cases each in the breast and lymph nodes (each 2/33, 6%).

Figure 1.

Example of a myeloid sarcoma within the skin (A–G, case 32) and bone marrow infiltration of an acute myeloid leukemia (H–M, case 31). (A) The skin biopsy shows a dense infiltrate within the deep dermis and subcutis [hematoxylin and eosin staining (H&E), ×10 original magnification]. (B) The cells are medium‐sized with vesicular chromatin and small nucleoli (H&E, ×400). (C–G) The cells are positive for CD33 (C, immunoperoxidase ×200), heterogeneously for CD56 (D, immunoperoxidase ×400), weak for CD117 (E, immunoperoxidase ×400), and heterogeneously for CD34 (F and G, immunoperoxidase ×400). (H) The bone marrow shows a highly cellular infiltrate (H&E, ×10). (I) The neoplastic cells have irregular nuclei, small nucleoli, and irregular cell contours (H&E, ×400). (J–M) The cells are positive for CD33 (J, immunoperoxidase ×200), heterogeneously for CD56 (K, immunoperoxidase ×400), weak and patchy for CD117 (L, immunoperoxidase ×400), and negative for CD34 (M, immunoperoxidase ×400).

Morphology and immunophenotype

The MS generally showed dense infiltrates of a monotonous proliferation of medium to large cells with sometimes finely dispersed, sometimes open chromatin and small to medium‐sized nucleoli. All MS were positive for CD33 and/or lysozyme (34/34, 100%), confirming their myeloid origin (Figure 1). The most reliable marker was CD33 (31/31, 100%), followed by lysozyme with 26/30 (87%) cases positive. In contrast, stem cell markers CD34 (32%) and CD117 (62%) only stained a subset of cases, as did CD56 (55%) and MPO (36%). NPM1 showed an aberrant cytoplasmic positivity in 9/14 (64%) analyzed cases. Case 19 exhibited morphological and phenotypic features of a borderline lesion between MS and histiocytic sarcoma.

Genetic landscape of MS

The DNA was successfully analyzed in all 62 samples (41 MS and 21 BMB); whereas RNA‐based fusion detection was successful in 38 of 41 MS samples (93%) and in 20 of 27 BMB (74%).

The median read depth of the DNA analysis was 3,421 reads (range 259–13,963 reads); the median total mapped fusion reads of the RNA analysis was 178,188 reads (range 10,260–2,839,892 reads). The median coverage of all alterations was 1,998 reads (range 20–35,008 reads).

The mutational landscape of MS is summarized in Figure 2 (for details see supplementary material, Table S2). All MS showed at least one alteration, and there were in total 127 pathogenic variants in 29 genes and 7 different AML‐type gene fusions. Twenty‐seven (79%) MS had only gene mutations, three (9%) only a fusion (including the two children) and four (12%) showed gene mutations and a fusion. AML‐type fusions occurred in younger patients (median 49 years, range 1–72 years versus 67 years, range 51–90 years). KMT2A and RUNX1 were involved in three and two fusions, respectively, but always with different partners. The fusions were significantly more often detectable in patients under 60 years (p = 0.007) and were mutually exclusive with NPM1 mutations, as expected (p = 0.0286).

Figure 2.

Mutational profiles of 34 myeloid sarcomas. The genes are arranged according to the frequencies of mutations. Only genes mutated in at least one sample are shown. Hatched boxes indicate copy number variations. AML, acute myeloid leukemia; ITD, internal tandem duplication; MDS, myelodysplastic syndrome; MDS/MPN, myelodysplastic/myeloproliferative neoplasm; MPN, myeloproliferative neoplasm; MS, myeloid sarcoma; SNP, single‐nucleotide polymorphism. The bar chart on the right demonstrates the frequencies of genetic alterations. Orange: mutations; grey: CNV; grey/orange overlay: both mutations and CNV.

The median number of mutations per case was 4 (range 0–9 mutations), with significantly fewer mutations in patients <60 years (2 mutations/case, range 0–7); with 4 mutations per case (range 1–9 mutations) in patients ≥60 years (p = 0.033).

The most frequently mutated gene in MS was TET2 (14/34, 41%), followed by NPM1 (13/34, 38%), NRAS (12/34, 35%), EZH2 (7/34, 21%), IDH1, KRAS (each 6/34, 18%), NF1, TP53 (each 5/34, 15%), and ASXL1, STAG2, DNMT3A, SRSF2, FLT3, and RUNX1 (each 4/34, 12%). All MS with cytoplasmic staining for NPM1 also showed an NPM1 mutation. All other genes were mutated in <10% of cases. Multiple mutations were observed in TET2 (three cases), DNMT3A, RUNX1, CEBPA, IKZF1, MPL, NF1, and PRPF8 (one MS each). KRAS and NRAS mutations were detectable in 18/34 MS (53%) and were mutually exclusive. Two of four patients (50%) with a history of cytotoxic therapy revealed a TP53 mutation, compared to 3/30 (10%) without previous history of malignancy. No NFE2 mutations were identified in 27 analyzed MS of 24 patients (supplementary material, Table S3).

In seven patients more than one MS sample was analyzed. All MS from the same patient were clonally related, with four patients (cases 4, 6, 18, and 27) showing an identical profile in both samples. Three patients (cases 8, 10, and 23) exhibited evidence of clonal evolution between the two MS specimens.

Copy number variation analysis

Targeted CNV analysis was performed in 33 of 34 cases on MS based on the results of the Oncomine myeloid assay (Figure 2). In 17/33 (52%) cases, CNV alterations were demonstrated, whereas 16 cases remained without CNV in the analyzed regions. In total, there were 57 alterations in 16 genes (TP53, DNMT3A, SH2B3, PRPF8, IKZF1, TET2, ETV6, RB1, CALR, FLT3, KRAS, NF1, ASXL1, ABL1, BRAF, and EZH2): 45 heterozygous deletions, 6 homozygous deletions, 2 focally heterozygous deletions, 1 focally homozygous deletion, and 3 amplifications.

The most affected gene was TP53 with 11 heterozygous and 1 homozygous deletion. In three cases, there was also a TP53 mutation in addition to the heterozygous deletion. DNMT3A showed five heterozygous and 1 homozygous deletion in two cases, additional to mutations. The IKZF1 gene was focally deleted in exons 5–8 in two cases on one allele, and in one case on both alleles. In a fourth case, the gene was completely deleted. TET2 had three heterozygous deletions, all in cases with additional mutations.

Selected matched BMB were also analyzed for CNV; in all of them, the MS showed an increase in alterations as further evidence of clonal evolution (data not shown).

MS shows frequent acquisition of MAPK/ERK pathway mutations and an association with clonal hematopoiesis (CH) in elderly patients

The 21 pre‐transplant BMB (of 20 patients) included 7 samples without evidence of a MN, whereas 14 samples showed a MN, including AML (n = 6), MDS (3), MPN (1), and MDS/MPN (4) (Figure 3).

Figure 3.

Comparison of mutational profile in 24 cases with additional bone marrow biopsy. The cases are grouped according to the status of the first bone marrow biopsy. AML, acute myeloid leukemia; BM, bone marrow; CMML, chronic myelomonocytic leukemia; EB, excess of blasts; IB, increase of blasts; ITD, internal tandem duplication; MDS, myelodysplastic syndrome; MDS/MPN, myelodysplastic/myeloproliferative neoplasm; MPN, myeloproliferative neoplasm; MS, myeloid sarcoma; PMF, primary myelofibrosis; SCT, stem cell transplantation; SNP, single‐nucleotide polymorphism; (t), cases with prior cytotoxic therapy – WHO: ‘post cytotoxic therapy’ and ICC: ‘therapy‐related’.

Three BMB showed no mutation, all with an AML‐type fusion in the MS. However, fusion analysis was possible in only one of these BMB. Mutations were detected in 18 BMB, including 5 without evidence of MN (cases 6, 8, 9, 10, and 32), thus classified as CH, and all 13 samples with a MN. CH was significantly more common in patients ≥60 years (p = 0.0386).

Of note, all BMB with mutations revealed a clonal relationship with the MS, including the pretransplant biopsies of MS cases which occurred after allogeneic SCT. Figure 4 shows a comparison of the distribution of mutations in MS and BM, highlighting the occurrence of CH with mutations in TET2, ASXL1, and STAG2 in non‐infiltrated BMB.

Figure 4.

Comparison of mutational profiles of bone marrow biopsies (BMB) and myeloid sarcoma in 20 patients with both available samples. Cases are grouped according to the status of the BMB: BMB without evidence of myeloid neoplasm, BMB with MDS, MPN, or MDS/MPN, including cases with excess of blasts, and BMB with AML.

Four of five cases with CH in the pre‐transplant BMB showed evidence of genetic evolution in the corresponding MS. An NPM1 mutation was gained in 3/5 cases (60%), followed by NRAS and KRAS in two cases each (40%) and BRAF and U2AF1 in one case each (20%). BM mutations absent from the MS indicative of branching evolution were found in three cases and affected PHF, TET2, and STAG2.

In the 13 pre‐transplant BMB with mutations and manifest MN, clonal evolution was observed in 9/13 cases, with additional mutations restricted to the MS in all 9 matched cases and other mutations only present in the BMB in 3 patients. The median gain was two mutations per patient (range 1–5 mutations). The most striking difference was the acquisition of NRAS mutations in the MS in comparison to the BMB in 4/13 patients (31%) overall and 3/5 patients (60%) with AML. Additional mutations of EZH2, CEBPA, PRPF8, IDH1, and RUNX1 were found in two patients (15%), and of NPM1, IKZF1, ETV6, and WT1 in one patient each (7%). In three patients, mutations in NF1, TET2, and CEBPA were only observed in the BM.

MS after stem cell transplantation

A clonal relationship to the pre‐transplant MN was confirmed in all but one (n = 6) patients with an available pre‐transplant sample. In this latter patient (case 7), the recipient origin of the MS 171 months after the SCT was confirmed by the identification of an NF1 polymorphism in both pre‐ and post‐transplant samples, but the RUNX1::USP42 fusion in the MS could not be analyzed in the pre‐transplant BMB for technical reasons. We also studied six post‐transplant BMB obtained at the time of diagnosis of the MS. Of note, only 1/6 BMB after SCT showed relapse of the initial MN (case 18, CMML), confirmed by the presence of an identical EZH2 mutation. No mutations were detected in the other five BMB with donor origin hematopoiesis.

Comparative gene expression analysis

Gene expression analysis was performed successfully on 19 MS samples and 3 matched BMB, as well as 8 additional AML samples without MS. A total of 154 genes showed a significant difference between MS and AML: 62 genes were upregulated in the AML group and 92 genes in the MS group (supplementary material, Figure S1). Principal component analysis (Figure 5) demonstrated a good separation between both the MS and the matched AML BM samples, as well as additional AML BM samples, despite the high heterogeneity of the tissue of origin of the MS. Only one MS (MS 14) of a 6‐year‐old child with KATA6::CREBBP translocation clustered within the AML group. Unsupervised clustering also resulted in a quite good separation of the two groups also (heatmap; supplementary material, Figure S2). Gene set enrichment analysis of differentially expressed genes identified two main groups (supplementary material, Figure S3). (1) Gene sets relevant for interaction with the extracellular microenvironment, especially for integrin binding, fibronectin binding, and extracellular matrix structural constituents, in part reflecting the different microenvironment and (2) the MAPK/ERK pathway and its regulation cascades, supporting the findings of the mutational analysis.

Figure 5.

Principal component analysis of myeloid sarcoma (MS) and acute myeloid leukemia (AML) in the bone marrow. The numbers correspond to the case numbers. The AMLs labeled with capital letters are additional AML samples without MS. MS 14 from a 6‐year‐old child with a known KATA6::CREBBP translocation is the only MS that clusters clearly within the AMLs.

Discussion

In this study, the mutational and gene expression profile of a series of MS was analyzed and compared with matched BMB in a subset of cases. Despite the marked genetic heterogeneity of MS and its overall similarity to AML, some genetic features of MS set it apart from AML with BM manifestation. Some of these differences point to biological mechanisms with a potential role in the extramedullary manifestation of MS, whereas others probably reflect acquired resistance mechanisms and are indicators of poor prognosis. The parallel investigation of matched BM biopsies revealed a high frequency of clonal evolution in MS.

Most prominently, alterations of the MAPK/ERK pathway, especially NRAS mutations in 35% of our cases, are prevalent in MS, and in concordance with published frequencies and in contrast to AML (31% versus 15%) [10, 11, 16, 19, 30]. Considering all examined genes belonging to the MAPK/ERK pathway [31] (KIT, FLT3, CBL, NRAS, KRAS, BRAF, and PTPN11), 74% (25/34) of the MS revealed mutations in these genes, similar to results from other series ranging from 49% to 84.6% [11, 16, 19]. In AML, this pathway is mutated in 15–20%, significantly lower than in MS [32, 33, 34, 35]. Of note, RAS and BRAF mutations were acquired in the MS in 55% of the cases with MN or CH, indicating that secondary activation of this pathway is associated with an increased propensity for extramedullary dissemination. The frequent activation of the MAPK/ERK pathway in MS, as compared to AML, was confirmed independently by GEP.

Another gene with a significantly higher frequency of mutations (21%) in MS in our series was the histone methyltransferase EZH2, which is infrequently (<5% in most series) mutated in de novo AML [36, 37, 38]. The reason for this increased incidence in our series is possibly twofold: on one hand, EZH2 mutations are more common in MDS/MPN and primary myelofibrosis (PMF) [36, 38]. Five of our MS cases arose in the background of CMML or PMF, with three CMML already carrying an EZH2 mutation. More importantly, however, loss of function mutations and low levels of EZH2 expression promote chemotherapy resistance and may be acquired in relapse [39, 40]. Of note, 2/7 EZH2 mutations in this series were only present in the MS, but not in the matched BMB, as evidence of clonal evolution. Whether EZH2 mutations are a risk factor for extramedullary dissemination or mainly reflect treatment resistance needs to be analyzed in further studies.

NPM1 mutations were identified in 38% of MS samples (41% after exclusion of two childhood cases), which lies in the range of published data on MS and is not significantly increased in comparison to AML with mutated NPM1 (25–30%) [10, 11, 15, 16, 17, 30, 41]. NPM1 mutations were acquired in the MS in 5 cases in the background of CH or a chronic MN. Of note, acquisition of NPM1 mutations is virtually always associated with rapid evolution to manifest AML [42, 43]. From a practical point of view, cytoplasmic staining for NPM1 proved to be a reliable indicator for the presence of an NPM1 mutation in MS.

In a recent study focusing on the mutational profile of NPM1+ primary MS versus NPM1+ AML, Ramia de Cap et al found more mutations in spliceosome (U2AF1) and histone modifying (ASXL1) genes in MS, and fewer alterations in DNA methylation genes [26]. In our subgroup of NPM1 mutated MS, DNA methylation genes including TET2, DNMT3A, IDH1, and IDH2 were mutated in 46%, 8%, 23%, and 15%, respectively, and spliceosome genes including SF3B1, SRSF2, ZRSF2, and U2AF1 were mutated in 0%, 8%, 8%, and 8%, respectively. Of interest, the only case of isolated NPM1+ MS in our series exhibited ASXL1 and U2AF1 mutations. Twelve of 13 (92%) NPM1‐mutated and 6/7 EZH2‐mutated MS also shared mutations in the MAPK/ERK pathway, suggesting cooperation among these genetic alterations.

Prototypic CH mutations affecting TET2, DNMT3A, and ASXL1 were detected in 16/34 (47%) MS, in concordance with published data [10]. All five matched BMB with detectable CH were clonally related to the MS, with shared TET2 mutations in 4/5 cases. These results underline that CH probably is an important mechanism in the pathogenesis of MS in elderly patients and may explain some of the cases arising de novo. Since two recent publications demonstrated recurrent NFE2 mutations especially in the setting of primary MS and suggested that these mutations promote the homing of neoplastic stem cells to an extramedullary site [27, 28], we additionally analyzed NFE2 with a custom panel. However, no NFE2 mutations were found in the present cohort of 27 MS samples, as well as in 38 MS cases analyzed by Anekpuritanang et al, indicating that this is not a common alteration [44]. The frequency of TP53 mutations in our MS series was comparable to AML in general, with an enrichment in therapy‐related MS.

CNV were analyzed for the regions covered by the Oncomine myeloid assay, identifying numerical alterations in 52% of analyzed cases. The most frequently altered gene was TP53 in 12/33 (36%) cases, which is in line with the frequency in AML [10]. Further alterations affected chromosome 7, especially deletion of 7q [45], in three cases of our study with deletions of BRAF (two cases), and BRAF and EZH2 (one case). Cases in which the corresponding BMB was analyzed showed acquisition of numerical aberrations in the MS as a sign of clonal evolution.

The gene expression analysis performed on a subset of cases achieved a good discrimination of MS and AML with an unsupervised clustering approach and principal component analysis. The MAPK/ERK pathway and its regulatory genes were a major discriminating feature between MS and AML, confirming the genetic findings. The analysis also revealed differential expression of genes relevant for growth factor and chemokine receptor binding, possibly reflecting adaptation to the different microenvironment. Since microdissection was not performed, extracellular matrix components and structural genes probably reflect the influence of the background tissue. However, given the high heterogeneity of the tissues of origin of the investigated MS, the good separation from AML is remarkable and indicates some common properties of MS independent of the genetic background.

Since we analyzed MS without restriction to specific disease groups, such as de novo cases or MS associated with AML in the BM, the encountered clinical spectrum is broad. The high frequency of cases arising in the background of an underlying chronic MN and after allogeneic SCT is notable and may explain why MS generally is considered to herald a poor prognosis. Of interest, 5/6 cases with MS after allogeneic SCT showed donor hematopoiesis in the BMB, indicating that MS in this setting may arise from leukemic stem cells persisting in extramedullary sanctuary sites. Due to the significant clinical and genetic heterogeneity of our cohort and the lack of follow‐up data for some cases, we abstained from correlating the mutational profile with outcome, which represents a limitation of our study.

Although morphology and immunophenotyping were not the focus of this study, the MS showed a broad phenotypical spectrum. CD34 was positive in only 32% of the MS and CD117 in 62%, supporting the myelomonocytic character of many MS. The MS in this study were positive for CD56 in 55%, whereas AML overall only about 20% show CD56 expression [46, 47, 48]. Higher frequencies of CD56 positivity are found in cases with monoblastic differentiation, with up to 51% in leukemic transformed CMML [46, 49].

In conclusion, clonal evolution with mutations and activation of the MAPK/ERK pathway, NPM1 and EZH2 mutations, and an association with CH or antecedent chronic MNs, especially in elderly patients, are common features of MS. After allogeneic SCT, isolated MS is a recurrent form of relapse.

Author contributions statement

FF: study design; T‐CS, LM, IB: carried out experiments; DN, T‐CS, MO, LM, EB, SF, JS‐H, LB, KS, IB, FF: data collection; DN, T‐CS, LM, EB, LQ‐M, IB, FF: data analysis and interpretation; DN, T‐CS, IB, FF: literature search; DN, EB: generation of figures; DN, LM, LQ‐M, IB, FF: writing of the manuscript; FF: supervision. All authors were involved in writing the paper and had final approval of the submitted and published versions.

Supporting information

Supplementary materials and methods.

Figure S1. Differential gene expression analysis of 2,506 analyzed genes with the most significantly different genes between myelosarcoma and acute myeloid leukemia in the bone marrow

Figure S2. Heatmap of myelosarcoma and acute myeloid leukemia

Figure S3. Gene set enrichment analysis of the most discriminating genes, shown in Figure S2

Table S1. AmpliSeq custom NFE2 panel

Table S2. List of mutations and fusions detected by targeted next‐generation sequencing with the Oncomine myeloid assay

Table S3. List of mutations detected by targeted next‐generation sequencing with the additional NFE2 panel

Data availability statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.