TO THE EDITOR:

FMS-like receptor tyrosine kinase 3 (FLT3) expression is almost exclusively found in the myeloid compartment [1]. The FLT3 protein consists of an extracellular region containing the ligand-binding site, a transmembrane domain, a cytoplasmic juxtamembrane domain, and a tyrosine kinase domain (TKD). Being a proto-oncogene, gain-of-function mutations in FLT3 (FLT3MT) gene create a robust selection pressure on affected myeloid leukemic precursors by alleviating the need for ligand binding [2].

FLT3MT occurs in about 30% of acute myeloid leukemia (AML) cases, with internal tandem duplications (ITD; FLT3ITD) accounting for 80% of all FLT3MT cases. In addition to FLT3ITD, distinct point mutations and deletions within the TKD, and more specifically mutations occurring in the activation loop (AL; FLT3AL) of the TKD can occur. Nonetheless, the latter are less common, making up approximately 7–10% of all AML cases [3]. FLT3MT has been linked to unfavorable prognoses [2], but recently, various FLT3 tyrosine kinase inhibitors (TKIs) of varying selectivity have been developed and their application may mitigate some unfavorable prognostic features of FLT3ITD and FLT3AL in the setting of myeloid neoplasms (MNs) [3, 4]. “Non-canonical” FLT3MT (NC; FLT3NC), i.e., not being ITD or located outside of the AL, have only occasionally been found in AML and related MNs [5]. However, the prognostic impact and clinical implications of such atypical molecular alterations have been far less explored than typical ITD and AL defects [2, 5,6,7]. We conducted a systematic analysis of the molecular landscape of FLT3MT, including FLT3NC, in a cohort of MN patients.

When we conducted a molecular and clinical analysis of 4529 patients with diverse MN subtypes (Tables S1 [excel format] and S2), a total of 248 cases with somatic FLT3MT [ITD, n = 152; AL, n = 66; NC, n = 30] (Fig. 1A) were identified. Of note is that, of 49 FLT3NC, 19 were excluded due to inadequate inclusion criteria (Supplemental Material). While all the FLT3AL were predicted to be deleterious, the FLT3NC included 8 alterations classified according to standard criteria as variants of unknown significance, 9 likely pathogenic, and 9 pathogenic variants (Table S1). Notably, 2 pathogenic/likely pathogenic FLT3NC variants were recurrent (p.K663R [n = 2] and p.N676K [n = 4]), and the other 2 co-occurred in one case (p.A813T and p.V825M) possibly due to somatic clonal mosaicism or compound heterozygosity.

Fig. 1: Characteristics of our patient cohort and FLT3 mutations.
figure 1

A Schematic representation of FLT3 domains and their amino acid location depicting type and number of mutations, disease groups, and variant interpretation according to practice guidelines set forth by the ACMG. Double hits within the same sample are indicated by an asterisk. B Pie chart breakdown of frequency of FLT3 mutations and type of mutations. C Co-mutation analysis comparing FLT3ITD, FLT3AL, and FLT3NC subgroups. D Response to phosphotyrosine kinase inhibitors in cases carrying FLT3NC vs. FLT3AL. E Kaplan–Meier curves showing probability of survival in cases with FLT3AL and FLT3NC AML subgroups divided by phosphothyrosine kinase inhibitors treatment status. Comparisons were considered statistically significant at P values less than 0.05. WT, wild type; MT, mutation; AML, acute myeloid leukemia; ITD, internal tandem duplication; AL, activation loop; NC, non-canonical; TKI, tyrosine kinase inhibitor; MDS, myelodysplastic syndrome, MPN, myeloproliferative neoplasms.

Full size image

In line with their strong leukemogenic potential, FLT3MT was enriched in AML (16% vs. 0.8% of non-AML entities, e.g., MDS/MPN, MDS, or MPN, P < 0.0001). Among FLT3 lesions found, FLT3ITD was present in 10% of AML and 0.3% of non-AML; FLT3AL in 4% of AML and 0.2% of non-AML, while FLT3NC was detected in 1.3% of AML and 0.3% non-AML. In a reverse analysis of 218 cases with FLT3ITD/AL, 205 (94%) had AML, but among 30 cases with FLT3NC, 11 (37%) were diagnosed with non-AML entities. Although FLT3MT was generally rare in the non-AML category, the proportion of FLT3NC alterations was significantly higher in these MNs vs. AML (48% vs. 8.5%, P < 0.0001; Fig. 1B). Consistent with this observation, we found that fewer non-canonical defects were present in high-risk AML disease subtypes (67% vs. 94% FLT3ITD/AL; P = 0.001).

We next assessed co-occurring mutations and found that, as expected, NPM1 mutations were significantly less likely to be associated with FLT3NC than FLT3ITD (20% vs. 46%; P = 0.008). DNMT3A mutations were recurrently found with both FLT3ITD and FLT3AL as compared to FLT3NC (36% and 33% vs. 10%, P = 0.005 and P = 0.023, respectively) due to their strong association with AML (Fig. 1C). Of note is that the very low frequency of TP53 mutations in our cohort of FLT3 (2/248) precluded comparisons in FLT3ITD vs. FLT3NC and FLT3AL. Median VAFs of both FLT3ITD and FLT3AL did not differ from those of FLT3NC (21% vs. 31%; P = 0.08). AML patients with either FLT3NC or FLT3AL mutations did not differ in terms of overall survival (36 vs. 23 mo., P = 0 .66; Fig. S1) or laboratory parameters at diagnosis (Table S3). When we explored the effects of TKI therapy in the various FLT3 genomic configurations, of 17 patients receiving TKIs (available clinical outcome data) 5 were FLT3NC (4 AML, 1 MDS/MPN). All 5 showed response to treatment, supporting an early promise of the efficacy of type I TKIs (midostaurin and gilteritinib) in these atypical patients (Fig. 1D–E) [8].

While interrogating the FLT3NC landscape, among 49 patients, we encountered 3 patients with germline FLT3NC that occurred at much higher rates among our cohort than in the general population (p.R311W, p.L262F, p.A291P; Fig. 2A, B, Table S4). Interestingly, all 3 variants were located within the extracellular ligand-binding domain [9]. The index case EC1, was a 57-year-old woman diagnosed with severe aplastic anemia (AA), who harbored a germline FLT3NC centrally located within the protein binding pocket (p.R311W, VAF = 50%, gnomAD 6.57 × 10−6). This alteration is predicted to reduce ligand-receptor interaction (Fig. 2C). Patient’s family history included unexplained anemia in two siblings. Remarkably, in line with the classical post-AA leukemogenesis, she subsequently evolved to MDS with monosomy 7 and later progressed to AML upon acquisition of FLT3AL (p.D835V, VAF = 14%) resulting in a biallelic clone in combination with a RUNX1 (p.P398L, VAF = 19%) and an NRAS mutation (p.G13D, VAF = 24%) [10]. This may possibly serve as an illustrative example of maladaptive clonal somatic gene rescue (SGR) (Fig. 2D, Fig. S2) analogous to MNs developing in the context of congenital bone marrow failure. Classic examples of such a scenario include the evolution of somatic CSF3R lesions in patients with severe congenital neutropenia due to germline variants in ELANE or the appearance of somatic mutations in eIF6, TP53, or del(20q) in patients with Shwachman–Diamond syndrome [11, 12]. To that end, the remaining 2 patients with germline FLT3MT also showed instructive clinical features. EC2 and EC3, both developing MNs, and presented with cytopenia.

Fig. 2: Dynamics of FLT3 clones and protein simulation.
figure 2

A Gene structure depicting the location of the three germline FLT3 variants. B Bar graph showing the frequency of each variant in the general population versus disease population at our institution. C In silico prediction of three germline FLT3 variants. Simulated using Glide from Schrodinger Inc. and visualized with PyMOL. D Clonal dynamics and example of somatic genetic rescue.

Full size image

Furthermore, EC2 was a 53-year-old woman diagnosed with MDS/MPN overlap, harboring a germline FLT3NC (p.L262F, VAF = 47%, gnomAD 3.99 × 10−6) with a somatic JAK2 mutation (p.V617F, VAF = 25.0%), which could be hypothesized to function as maladaptive SGR lesion. The patient’s family history was significant for unexplained thrombocytopenia in her mother and leukemia in her maternal grandfather. This variant was predicted to impair interactions between domains 3 and 4 of the extracellular region of the FLT3 receptor (Fig. 2C). EC3 was a 58-year-old man with MDS harboring a germline FLT3NC (p.A291P, VAF = 60%, gnomAD 1.59 × 10−5) developing AML with del(7q). The structural model showed that this mutation might allosterically modulate the conformation of the ligand binding domain via the constraint of proline (Fig. 2C).

The cases presented, indicating that the hypomorphic germline FLT3MT may result in maladaptive clonal SGR, prompted us to consider the opposite scenario: FLT3MT could point towards the presence of other hypomorphic germline genetic lesions in genes which could induce selection pressure and SGR via acquisition of FLT3 lesions. As potential candidates for screening, we selected genes related to severe congenital neutropenia (e.g., ELANE, WAS, etc.) and other TKRs (e.g., CSF1R, CSF2RA, CSF2RB, KIT). Indeed, 4 cases with potentially hypomorphic germline mutations in CSF3R were found to have 5 somatic FLT3MT (3 AL, 2 ITD; Table S5), including an AML case with germline CSF3R p.L619S; 2 AML cases with germline CSF3R p.E808K (one of them with 2 somatic FLT3MT); and a 71-year-old patient with MDS and germline CSF3R p.E835K who presented with neutropenia. All 3 variants in CSF3R have been identified as germline and have been associated with myeloid diseases [13,14,15].

Our study suggests that albeit somatic FLT3NC are overrepresented in less advanced MNs, once patients progress to AML, no distinct mutational or cytogenetic signatures or prognostic features can be assigned to FLT3NC. However, a few informative cases suggest the potential efficacy of TKIs in FLT3NC MNs. Most importantly, we have inadvertently identified unique cases of germline FLT3NC and FLT3-mediated SGR, in which background hypo proliferative pressure induces somatic escape via oncogenic transformation. This observation indicates that activating FLT3MT may represent a maladaptive SGR of hypomorphic germline FLT3MT in a multi-hit fashion, or alternatively, they might also shift to hyperproliferation via activating mutations in other TKRs or cytogenic lesions. These findings provide new insights and future research perspectives on the molecular complexity of FLT3MT in MNs.