Acute myeloid leukemias (AML) are still an open challenge today. As known, these are diseases involving the bone marrow, with the presence of immature cellular elements. However, these neoplastic cells are capable of homing in tissues other than bone marrow through mechanisms not yet fully explained. Although a more appropriate name for this manifestation would be “extramedullary acute myeloid leukemia tumor”, the condition is commonly known as myeloid sarcoma (MS). This term is strictly considered by World Health Organization for the cases with no concurrent bone marrow (BM) involvement. A more appropriate name for contemporary manifestation in the BM and another site should be referred to as “extramedullary acute myeloid leukemia tumor”. This is a reality that every hematologist must deal with. In a retrospective analysis of 346 AML patients, extramedullary involvement had an incidence of 11% and was found to significantly complicate prognosis and therapeutic strategies ^[1]. On the other hand, the improvements in diagnostic techniques and the recent findings in molecular biology (including FLT3 and its inhibitory drugs) promise to revolutionize the therapeutic workup of this patient setting.

The cancer cells of acute myeloid leukemia can sometimes aggregate in extramedullary sites. This extramedullary tumor, known as myeloid sarcoma (MS, also known as granulocytic sarcoma or chloroma), is defined by the World Health Organization as a tumor mass consisting of myeloid blasts with or without maturation occurring at an anatomic site other than BM. MS (or, in general, extramedullary acute myeloid leukemia tumor, e-AML) is composed of cancer cells in various stages of maturation: myeloblasts, monoblasts, and, rarely, megakaryocytic or erythroid precursors ^[2]. The pathogenesis of e-AML and MS is unknown. It has been associated with the expression of specific cell adhesion molecules, chemokine receptor/ligand interaction, and aberrant FAS-MAPK/ERK signaling. In particular, adhesion molecules, such as CD56 (neural cell adhesion molecule), are more frequently expressed in leukemia cells. CD56 can promote the adhesion of cancer cells through homophilic binding to tissues expressing this molecule (adipose/soft tissue, gastrointestinal, brain, testicular, and skeletal muscle) ^[3][4][5].

In contrast to this hypothesis, however, a large-scale study showed that the occurrence of CD56-positive leukemic blasts was similar in patients with and without e-AML or MS ^[6]. Another adhesion molecule is CD11b (surface β2-integrin member macrophage-1 antigen) expressed on mononuclear cells. This molecule mediate the interaction between cells and the matrix and could determine the frequency of extramedullary localization in forms of AML with monoblastic/myelomonocytic differentiation ^[7][8]. e-AML can have different localizations, and symptoms depend on the involved organ ^[9]. The most affected organs are skin, soft tissues, and lymph nodes, although the testes, bone, peritoneum, and gastrointestinal tract may also be involved ^[2]. MS is rarely be isolated (with a finding of 2 new cases per million adult subjects) but more frequently diagnosed in individuals with a previous or concomitant history of other hematological malignancies. Among these, the most frequent are AML, myeloproliferative neoplasm (MPN), myelodysplastic syndrome (MDS), MPN/MDS, and in the blast phase of chronic myeloid leukemia (CML) ^[10].

Diagnosis starts with the patient’s medical history and clinical suspicion, detailed with instrumental examinations. Among them, computed tomography (CT) is the first imaging modality. CT scans display mild to marked homogeneous post-contrast enhancement. In addition, magnetic resonance imaging (MRI) provides images of the central nervous system (CNS). ¹⁸Fluorodeoxy-glucose positron emission tomography/CT (FDG-PET/CT) is the gold standard imaging modality to evaluate the presence of e-AML and monitor the response to therapy over time ^[11][12].

For this reason, FDG-PET/CT is among the mandatory exams of the NCCN for the evaluation of extramedullary localizations of AML. However, a biopsy of the lesion is essential to establish a diagnosis with certainty. A condition of severe thrombocytopenia (PLT < 50,000/mmc) can make biopsy difficult for many suspected organs, particularly those involving the CNS. In such situations, the regression of the lesion following therapy for AML establishes diagnosis with reasonable certainty. The evaluation of the histological section typically reveals infiltrating myeloid cells at various stages of maturation, with either monocytic or granulocytic maturation, in the same way as in AML. Diagnosis is confirmed using immunohistochemistry, an antibody panel commonly used in AML. Specifically, staining for myeloperoxidase (MPO-expressed in 66% to 96% of MS) is used to help differentiate lesions from lymphoma ^[13]. The other commonly tested antigens are CD43, CD68, lysozyme, myeloperoxidase, CD117, CD11c, CD13, and CD33 ^[10][14][15].

Other techniques that can confirm or improve diagnostic capacity are immunohistochemistry, flow cytometry, and fluorescence in situ hybridization (FISH). A percentage of cases ranging from 25% to 47% are initially mislabeled as melanoma, thymoma, Ewing sarcoma, or malignant lymphoproliferative disorders (e.g., Hodgkin lymphoma, histiocytic lymphoma, mucosa-associated lymphoid tissue lymphoma, large-cell lymphoma) or poorly differentiated carcinoma ^[14][16][17].

e-AML could also be confused with a blastic plasmacytoid dendritic cell neoplasm (BPDCN), a rare and aggressive hematologic malignancy of the BM and blood that can affect organs such as lymph nodes, skin, spleen, and CNS. Skin lesions are characteristic of this disease; therefore, they must be distinguished from e-AML ^[18].

NCCN guidelines recommend that patients affected by extramedullary disease should be considered for a screening lumbar puncture evaluation, followed by an empiric dose of intrathecal chemotherapy ^[19].

The European Society for Hematology (ESH) classifies MS into four subgroups ^[10]:

-

Extramedullary involvement with concurrent newly diagnosed AML;

-

Extramedullary relapse of AML, including in the post bone marrow transplant setting;

-

Blast phase/transformation of a myeloproliferative neoplasm or chronic myelomonocytic leukemia;

-

Isolated MS in association with a normal bone marrow biopsy and blood film, as well as the absence of any history of myeloid neoplasia.

Extramedullary leukemic involvement associated with concurrent AML or at relapse occurs more frequently in monoblastic leukemia with translocations involving 11q23, acute myelomonocytic leukemia with eosinophilia, inv(16) (p13;q22) or t(16;16) (p13;q22) ^[20][21], and in AML with maturation and t(8;21) (q22;q22). Significantly, these cases of MS, among the first described years ago, involving genital organs and frequently associated with AML, formerly known as FAB M2 ^[22], are characterized by an excellent response to systemic chemotherapy ^[23].

Several factors influence the choice of best therapy for e-AML. The choice depends on the involved organ, the synchronous presence of medullary involvement of the AML, and any previous lines of treatment for the underlying disease (prior allo-HSCT); furthermore, as already discussed, the choice of therapy is contingent upon the presence of targetable abnormalities (e.g., FLT3 and IDH1/2 mutations). For localized forms of the disease, one therapeutic possibility to relieve symptoms may be represented by surgical decompression. MS appears to be sensitive to ionizing radiation, so involved-field radiotherapy (RT) should be considered for all patients with isolated MS refractory to systemic therapy ^[24]. However, no improvement in five-year survival was found in a retrospective study analyzing patients with extramedullary disease localization who were treated between 1994 and 2006 with systemic therapy either associated or not with RT ^[25]. For the above reasons, intensive systemic chemotherapy represents the gold standard of e-AML treatment in eligible patients, whether or not associated with bone marrow localization of AML ^[19]. These patients are treated with classic anthracycline or cytarabine-based therapeutic regimens. In an analysis by Tsimberidou et al., complete remission was achieved by 65% of patients with e-AML, whereas 5% reached a partial remission, with a median overall survival of 20 months (range 1–75) ^[26].

On the other hand, for patients not eligible for intensive care, hypomethylating agents, e.g., azacitidine (HMA) and deoxyribonucleic acid methyltransferase inhibitor (DNMTi) decitabine, appear to have been able to induce complete remission after 2–4 cycles of therapy in few selected clinical cases ^[27]. Among target therapies, limited data support the use of sorafenib (an FLT3 inhibitor) ^[28] or drugs such as enasidenib or ivosidenib in the context of the form with IDH2 or IDH1 mutations, occurring in 30% and 15% of cases, respectively ^[29][30].

Further target therapies concern B-cell lymphoma 2 (BCL-2) inhibitors, i.e., venetoclax. In this context, there is no solid scientific evidence yet. Sporadic case reports include e-AML patients treated with venetoclax both in the course of disease refractory to the first line of therapy and in the extramedullary localization of AML arising after HSCT ^[31][32]. Furthermore, it is interesting to note that there are reports of potential penetration by venetoclax through the cerebrospinal membrane being helpful in treating AML involvement of the CNS ^[33]. These results certainly require more consistent confirmatory data.

Regarding the immunotherapeutic approach, success is poor. A proportion of 10% of leukemic cells of the MS overexpress PD-1 ^[34][35], and immune checkpoints that generally have the task of controlling the degree of inflammation and preserving damage induced by T lymphocytes on healthy tissues and in neoplastic conditions represent an escape from the tumor against the organism’s response to the disease. However, this feature’s clinical and therapeutic significance remains a question to be further investigated ^[36]. Once complete remission is achieved, consolidation with HSCT is the best therapeutic strategy. The rate of post-allo-HSCT 5-year OS for patients with MS ranges from 47 to 53% ^[37], comparable to the OS of patients with AML undergoing HSCT (25–40%) ^[38]. HSCT or donor lymphocyte infusion (DLI) is effective against eAML, probably causing the graft-versus-leukemia (GvL) effect guaranteed by the engrafted immune system (graft-versus-leukemia effect following hematopoietic stem cell transplantation for leukemia). In the case of disease relapse, even after a first bone marrow transplant, a second allo-HSCT can be contemplated for younger patients. However, prospective studies with cytogenetic and molecular data analysis from patients affected by eAML are needed to guarantee an appropriate risk stratification and create a guide for upfront therapy and, eventually, consolidative strategies ^[39].

Since its discovery by two independent groups in 1991 ^[40][41], the scientific community has realized that FMS-like tyrosine kinase 3 (FLT-3) could play a pivotal role in the study and knowledge of human neoplastic pathologies. The FLT3 gene, belonging to the class III receptor tyrosine kinase family, encodes a human protein of 993 amino acids and is expressed as a receptor in the placenta, gonads, and brain. FLT3 receptor activation by its ligand (FL) leads to rapid receptor autophosphorylation and induces the activation of several intermediate signal-transduction mediators, determining cell proliferation and expansion ^{[42][43][44][45][46]}.

These data demonstrate an essential role for FLT3 in developing multipotent stem and B cells.

Furthermore, FLT3 has played a cardinal role in the study and management of AML since 1996, when Yokota et al., using reverse transcriptase-polymerase chain reaction (RT-PCR), found, in 5 AML patient samples (17%), a different transcript with a primer combination that could amplify the transmembrane (TM) domain through the juxtamembrane (JM) domain (FLT3-ITD mutation). These repeat sequences disrupted the JM domain’s autoinhibitory activity, resulting in constitutive tyrosine kinase activation ^[47]. Subsequently, a further mutation was identified involving the activation loop of FLT3, a component of the tyrosine kinase domain (FLT3-TKD mutation) ^[48][49][50].

Larger-scale evaluations have shown that FLT3 mutations are found in approximately 30% of de novo diagnosed AML cases and are divided between ITDs (about 25%) and point mutations in the TKD (7–10%) ^[51].

Over the years, studies have highlighted the crucial role of these mutations in the context of newly diagnosed AML, as well as their ability to actively characterize the prognosis and, above all, their potential as a therapeutic target. For the FLT3-TKD mutation, the prognostic impact is not yet well defined. Despite being based on small samples of patients, the literature data do not confirm a relevant correlation between the presence of this mutation and clinical outcome ^[45]. Risk stratification guidelines for AML have been drawn up by several organizations, including the World Health Organization (WHO), the National Comprehensive Cancer Network (NCCN), and the European Leukemia Network (ELN) ^[19][52], specifying that the presence of FLT3-ITD is a genetic alteration that identifies a well-defined subgroup of patients. The NCCN guidelines classify patients with these mutations as having poor prognoses ^[19]. The same is true for the ELN guidelines, in which the FLT3 mutation plays a prominent role, specifically as an adverse prognostic indicator of disease. The concept of allelic ratio (AR) represents the number of ITD-mutated alleles compared with the number of the wild-type alleles and therefore is not only influenced by the amount of leukemia versus normal cells in the tested sample but also by the percentages of cells with 0, 1, or 2 mutated alleles. Numerous studies have concluded that a high ratio (AR > 0.5) is associated with worse prognoses and patient outcomes. This concept made it possible to stratify patients into three risk classes (favorable, intermediate, or adverse) depending on the presence or absence of the nucleophosmin member 1 (NPM1) gene mutation and to favorably mediate the prognostically adverse effect of FLT3-ITD ^[53]. In contrast, Sakaguchi’s group analyzed 147 patients with FLT3-ITD gene mutation-positive AML, stratifying them, according to ELN indications, into high and low AR, depending on the presence or absence of the NPM1 mutation. They found that FLT3-ITD^low AR was not associated with favorable outcomes (overall survival (OS), 41.3%). Furthermore, only patients who underwent allogeneic stem cell transplantation (allo-HSCT) in the first complete remission (CR1) had a more favorable outcome than those who did not (relapse-free survival (RFS) p = 0.013; OS p = 0.003). It was also underlined by multivariate analysis that allo-HSCT in CR1 is the only prognostic factor capable of giving a better OS and progression-free survival (PFS), demonstrating that prognosis was unfavorable in NPM1-mutated AML with FLT3-ITD^low AR when allo-HSCT was not carried out in CR1 ^[54]. Regarding the setting of refractory/relapsed AML patients, there is a risk of onset of leukemic clones with multiple adverse-risk genetic mutations (including FLT3-ITD) or the presence, at relapse, of an allelic burden higher than that at diagnosis, capable of negatively influencing the prognosis, unlike the diagnosis, where there is a greater probability of facing a polyclonal disease ^{[55][56][57][58][59][60]}. In this context, FLT3-ITD mutations are newly detected at relapse more often than FLT3-TKD mutations (8% and 2%, respectively) ^[56], with a prognosis worse than that in patients maintain the wild-type form of the gene ^[61].

In e-AML, the study of molecular markers is crucial for categorizing the disease and guiding clinicians to the best therapeutic choice for patients. Ali Ansari-Lari, M. et al. analyzed 24 e-AML specimens from 20 patients in a study protocol. Clinical information was available for 15 out of 20 patients. e-AML was diagnosed in the setting of AML in nine cases (60%), CML in three cases (20%), and precursor B (Pre-B) ALL in one patient (6.7%), with appearance ranging between 3 months and 21 years after a leukemia diagnosis. In three instances, the e-AML was diagnosed concurrently with leukemia (two AML, one CML). However, six patients had no leukemia bone marrow involvement (it should be emphasized that three cases did not have a concurrent bone marrow aspirate or biopsy). From a molecular point of view, no FLT3 D835 mutations were identified in the 20 cases examined, whereas FLT3-ITD mutations were identified in three of the 20 cases (15%). Contrary to what happens in FLT3-mutated AML, their data were too few to identify a reliable prognostic significance ^[62]. Mutation in the FLT3-ITD gene would favor the infiltration of leukemic cells into the visceral organs, simultaneously reducing the BM homing of leukemic cells by deregulating CXCR4 signaling ^[63]. In the context of extramedullary localizations, as already seen, there are several molecules associated with more significant infiltration of leukemic cells in other organs. As described, CD56 plays an essential role as an adhesion molecule. It is responsible for homing of these cells in several tissues and is highly expressed in the breast, testicular tissue, ovary, and gut. Furthermore, numerous studies have shown that miRNAs play a crucial role in hematopoiesis and hematological malignancies. In this context, FLT3 can modulate the expression of different miRNAs, with both downregulation (miR-451 and miR-144) and upregulation (miR-155, miR-10a, and miR-10b) mechanisms. The expression of these non-coding endogenous small molecules appears to play a role in these conditions, with mechanisms yet to be explored ^[64].

This evidence explains how FLT3 mutations occur in a significant subgroup of patients with e-AML. However, as already discussed, if FLT3-ITD mutations appear to be associated with increased relapse risk, adverse disease-free survival, and overall survival, drawing a definitive prevalence is not possible due to the limited sample size of analyzed patients.