To the Editor:

The World Health Organization (WHO) classification of acute myeloid leukemia (AML) has incorporated molecular genetic and cytogenetic aberrations in the definition of most entities [1]. The diagnosis of acute panmyelosis with myelofibrosis (APMF) is still not based on genomic changes but on clinicopathologic features and the exclusion of other myeloid malignancies, in particular AML with myelodysplasia-related changes (AML-MRC) [2]. APMF is a rapidly progressive hyperfibrotic subtype of “AML not otherwise specified” (AML NOS) and accounts for <1% of AML cases [3]. This aggressive disease is characterized by rapid onset of cytopenia and constitutional symptoms in the absence of splenomegaly, previous history of myeloproliferative neoplasm (MPN) or myelodysplastic syndrome (MDS), and exposure to radiation or cytotoxic drugs. Blood smears are devoid of tear-drop-shaped cells but may contain dysplastic platelets and myeloid cells including rare blasts [2, 3]. Histopathologic evaluation of core bone marrow (BM) biopsies is mandatory to recognize the hallmark of this myeloid neoplasm: “panmyelosis”, i.e. trilineage myeloid proliferation with dysplastic erythropoiesis, abundant atypical megakaryocytes, and overt BM fibrosis ≥ grade 2 on a 0–3 scale and ≥20% predominantly CD34-positive precursors/blasts often arranged in clusters (Fig. 1a–d) [3,4,5]. As a consequence of dry tap BM aspiration and rare circulating CD34-positive precursors/blasts, genetic data are reported in a limited number of patients. An abnormal karyotype lacking recurrent aberrations was identified in most APMF cases with available cytogenetic information. Published cytogenetic data were similar to AML-MRC including aberrations of chromosome 5q, 7q/7, 8, and 11q [3, 4]. Two other studies detected different chromosomal aberrations involving chromosomes 3q, 5, 12, 13q, and 22 [6, 7]. In a single APMF patient, an amplification of EVI1/MECOM within a derivative chromosome 8 was described [8]. In another study of four APMF patients no JAK2 V617F, MPL, or CALR mutations were detected [9].

Fig. 1
figure 1

Representative BM trephine biopsy, summary of the OncoScan copy number analysis and Kaplan–Meier analysis of overall survival. a–d BM morphology in a patient with high complexity genetic aberrations: a BM spaces are nearly devoid of fat cells and infiltrated by blast cells and residual dysplastic hematopoietic cells including micromegakaryocytes (arrow, hematoxylin and eosin, ×63). b Diffuse densely increased reticulin fiber meshwork consistent with BM fibrosis grade 3 (Gomori’s reticulin stain, ×63). c Accumulation of CD34-positive blast cells (CD34 immunohistochemistry, ×63). d Nuclear overexpression of the p53 protein in about 30% of nucleated cells (p53 immunohistochemistry, ×100). e Summary of the OncoScan copy number analysis showing the type of CNAs per chromosome in all patients using Nexus Express Software for OncoScan, blue color presents copy number gains and red color copy number losses. f Summary of the OncoScan copy number analysis showing the number of CNAs per patient. Stars inidicate the TP53 mutations in five patients identified via panel sequencing. g Kaplan–Meier analysis of overall survival (OS) of the total cohort. h Kaplan–Meier analysis of OS according to genomic complexity group. i Kaplan–Meier analysis of OS according to TP53 status as defined by: (#) no TP53 aberrations in the OncoScan assay and unknown mutational status by NGS; ($) TP53 aberrations in the OncoScan assay and/or TP53 mutations by NGS

Full size image

The first step of our study was to identify in our files de novo AML patients who presented with relevant fibrosis at initial diagnosis. Among those, we retrospectively re-evaluated cases with available BM biopsies for the presence of clinicopathologic features consistent with the 2016 WHO criteria for APMF. Cases with a previously proven MDS-type cytogenetic profile and prominent dysplasia involving ≥ 50% at least in two lineages as mandatory for AML-MRC were excluded. In addition, no cases with megakaryocytic abnormalities, such as arrangement in dense clusters characteristic for primary myelofibrosis (PMF) were considered. By consequently applying WHO criteria, we retrieved a cohort of 16 patients that we could assign to the APMF category (Fig. 1a–d, Table 1, Supplementary Table 1). Briefly, the core biopsies showed BM fibrosis grade 2–3 associated with a generally rather low frequency of CD34-positive myeloblasts (median 30% in the total cohort), a proliferation of predominantly small dysplastic CD61-positive megakaryocytes and a macroblastic glycophorin C-positive and CD71-positive erythropoiesis with increased proerythroblasts. Using routine methods suitable for formalin-fixed paraffin-embedded (FFPE) material MPL W515L was absent in 5/5 cases, while 1/5 cases harbored a JAK2 V617F mutation with a low (5%) mutant allele burden. Karyotyping by metaphase analysis or FISH had been unsuccessful due to BM dry tap and low circulating CD34+ cells. Thus, we performed molecular studies on genomic DNA extracted from the diagnostic FFPE and EDTA-decalcified BM trephine biopsies by applying the OncoScanTM FFPE assay (Thermo Fisher Scientific, MA, USA) for whole genome tumor profiling [10]. Using stringent settings for gains and losses (Supplementary Materials and Methods) genomic copy number abnormalities (CNAs) were discovered in 94% (15/16) of APMF patients. Based on the number of CNAs we could distinguish two groups within our APMF cohort. Six patients harbored ≤ 3 (mean 2) CNAs and were referred to as low genomic complexity group (Fig. 1e, f, Table 1, Supplementary Table 2). The low complexity group showed mostly single but heterogeneous CNAs (Table 1). In contrast, 10 patients were assigned to a high genomic complexity group since >3 (mean 15) CNAs were identified. As a part of a complex genomic profile, the most frequent CNAs were losses of 17p, 5q, and 7q (in 10, 9, and 9 patients, respectively). Other MDS-related or MPN-related CNAs occurred with lower frequency such as loss of 18p in five patients and losses of 17q, 18q, 11p, and 3p, as well as a gain of 3q in four patients, respectively (Fig. 1e, Table 1, Supplementary Table 2). The gains of chromosome 3q, 8, 12q, 17q, and 21q, as well as loss of 7q were the CNAs common to both groups and have been previously described in APMF [3, 6, 8].

Table 1 Clinicopathologic and molecular characteristics of patients with APMF
Full size table

DNA for mutational analysis (Supplementary Information) was only available in 5/16 patients (Table 1). Three of these patients were assigned to the high complexity group while two patients were included in the low complexity group. Molecular profiling revealed TP53 mutations in all of these five patients. Taken together with the copy number results, 12/16 patients harbored TP53 abnormalities (i.e. loss of 17p and/or TP53 mutation) including 3 with inactivation of both TP53 alleles (Table 1).

Further mutations were detected in the epigenetic regulators DNMT3A and TET2 (in 2/5 samples), signaling molecule CBL (1/5 sample), and co-repressor BCOR (in 2/5 samples). Four patients did not show TP53 abnormalities in the OncoScan assay but were not evaluable for TP53 mutation analysis. Thus, they were considered as “TP53 unknown”.

As compared with other AML entities, the genetic alterations underlying APMF are not well characterized. An interesting finding that has not yet been described is the presence of TP53 aberrations in the large majority of APMF patients. In accordance, we frequently observed high levels of genomic complexity suggesting a high genomic instability universally associated with an adverse outcome in AML [11]. Patients of the large high complexity group (10/16 patients) exhibited a genetic profile overlapping with AML MRC. TP53 alterations are rare in de novo AML (about 8–14%) but are closely associated with the presence of a complex karyotype and/or monosomal karyotype [11, 12]. In a recent study, AML with TP53 mutations and chromosomal aneuploidies, such as deletions and monosomies of chromosomes 5, 7, 12, and 17, as well as trisomy 8 were considered a distinct category [12]. Patients in this subgroup were older and had the worst overall survival of all AML subgroups [11]. Accordingly, the term “AML with TP53 mutations and chromosomal aneuploidy” could also be assigned to most of our APMF patients. The detection of a JAK2 V617F mutation in one patient does not exclude de novo AML but may suggest a rapidly transformed MPN without a clinically evident chronic phase [13, 14]. MPN patients with TP53 mutations have a poor prognosis with a high risk of transformation [14]. Very high risk PMF overlap at least in loss of 17p with APMF but the most frequent CNAs reported in PMF affect different chromosomal regions: 20q, 17q, 7p, 9p, 13q, or 1q [15]. Therefore, our results could be helpful in distinguishing APMF from PMF.

Despite the molecular heterogeneity, all patients of our APMF cohort had a dismal outcome with a median survival of 5.4 months (range 1.8–11.3 months, Fig. 1g–i, Supplementary Table 1). No patient was treated by allogeneic hematopoietic stem cell transplantation. Three patients received standard induction chemotherapy, one patient thalidomide and another patient decitabine. Statistical analyses comparing low versus high complexity cases or those with known or unknown TP53 abnormalities did not reveal any significant differences with regard to overall survival and hematologic pretreatment parameters, except with regard to platelet counts and number of CNAs (Fig. 1, Supplementary Table 1). Patients negative for TP53 abnormalities in the OncoScan assay and unknown TP53 mutation status had significantly higher platelet counts (p = 0.045) and lower numbers of CNAs (p = 0.006) than those with TP53 abnormalities. It is worth noting that the superior survival of 11.3 months was observed in a patient of the high complexity group with a TP53 exon 10 splice site mutation receiving hypomethylating therapy (decitabine). Interestingly, the frequency of TP53 splice mutations (4/5, 80%) in our cohort is high, compared to the splice mutation frequency of around 5% in myeloid neoplasms (11/201, see Suppl.) estimated from the IARC TP53 database. Whether this observation can be confirmed and may even be linked to the fibrotic phenotype needs to be shown in a larger cohort.

The mechanisms that contribute to the typical morphologic and clinical features, and the rapidly progressive disease that entailed the designation as APMF still remain to be elucidated. Previously, we have observed an increased inflammatory T-cell rich background in APMF core biopsies [5]. Inflammatory microenvironment changes of the niche may contribute to genomic instability of hematopoietic stem cells and disease evolution. Cytokines and chemokines released from inflammatory cells and abundant megakaryocytes may contribute to the fibrotic modulation of the BM that is a hallmark of APMF. Although the number of our patients is small and the spectrum of methods that we could apply is limited, the data presented here provide further insights into the molecular basis of APMF and highlight the high prevalence of TP53 abnormalities and chromosomal aneuploidy.