FLT3mutation incidence and timing of origin in a population case series of pediatric leukemia

Certain, but not all, chromosome translocations in childhood leukemia are known to be present at birth. This phenomenon of prenatal origin was initially presumed from twin studies where it was observed that mono-amniotic twins always harbored the same translocations (reviewed in [1]). In addition, several studies have shown that specific mutations found at diagnosis in children with leukemia were present at birth, that is, "backtracked" to neonatal Guthrie Cards (blood spots used for newborn screens) (reviewed in [2]). The MLL rearrangement in infant ALL "backtracks" in nearly all cases and TEL-AML1 is found on 75% of Guthrie cards matched to leukemia cases with the translocation [3–5]. The E2A-PBX1 fusion generated by the t(1;19) translocation is a likely exception, with a postnatal origin [6], along with possibly others. These results collectively support a "two hit" model of leukemia, with one hit early in life or in utero, and another at a later date in temporal proximity to leukemia diagnosis.

The FLT3 gene is located on chromosome 13q12 and encodes a Type III membrane receptor kinase that regulates normal hematopoiesis. Mutations in FLT3 in AML occur in approximately 5-15% in children and 25-35% in adults, and account for the most common single gene defect in AML (reviewed in [7]). Several studies have indicated that children and adults with AML and the FLT3 mutation have a very poor prognosis. FLT3 mutations have also been documented in adult and pediatric ALL. An initial report demonstrated a 14% frequency of FLT3 mutations among childhood ALL overall, with mutations concentrated among the cytogenetic subgroups high hyperdiploidy (> 50 chromosomes in diagnostic karyotype) and MLL-translocation [8]; more recent studies have indicated a lower overall frequency in childhood ALL (in the 1-8% range) while consistently demonstrating a higher incidence among those with MLL rearrangement and high hyperdiploidy [9–13]. The absolute incidence of FLT3 mutations in pediatric leukemia is of interest in part because of the existence of several promising FLT3 inhibitors currently under development (reviewed in [14]) such inhibitors are more effective in the presence of FLT3 activation.

Internal Tandem Duplications (ITDs) and activation loop mutations are two unique FLT3 mutations that have been characterized. ITDs, which occur on exon 14, are insertions of repeated base pairs that range from 3-400 base pairs each. Activating loop mutations occur on exon 20; they are most commonly missense point mutations that occur at codon 835/836. Point mutations at codon 840, 841 and 842 have been described as well as insertion of base pairs between codons 841 and 842. In adult AML, ITDs comprise the majority of mutations, while in pediatric AML, ITD mutations are less common. It has been proposed that FLT3 mutation may be a late event in leukemia, given that FLT3 mutation status are often changed between paired diagnosis and relapse of adult and childhood patients with AML [15–17] or ALL [18]. The present study was conducted to formally assess this timing. We screened a large population case series of pediatric leukemia for FLT3 mutations of both types and determined whether FLT3 ITD mutations were present at birth by examining DNA on the corresponding Guthrie cards.

The research presented here was reviewed and approved by the UCSF Committee on Human Research, protocol # H10806-17300-12, and all study personnel completed appropriate human subjects training courses. Research material was derived from the Northern California Childhood Leukemia Study (NCCLS), and epidemiology study based at UC Berkeley.

Eight indel mutations were identified consisting of 8 internal tandem duplications and 1 unconventional deletion. All indel mutants occurred in a 102 base pair region within exon 14 of the FLT3 gene, except for one case (0126) whose duplication included one base in the intronic region between exon 14 and 15 (Figure 1 and Additional File 1, Figure S1). Indels ranged from a 9 base pair deletion to a 90 base pair insertion and all preserved the open reading frame. Of the 9 indel mutations found, 6 Guthrie cards were available for backtracking. Each of these 6 Guthrie cards was used with their respective mutant specific primers (Table 2) in an attempt to amplify the DNA and determine if the mutation was present in the birth blood. None of the DNA from the 6 Guthrie cards DNA was amplified using the mutant specific primers at a sensitivity of 1 cell per 6,700 tested in 40 ng DNA (1.5 × 10^-4) suggesting that the indel mutations were either not present in the DNA from the Guthrie cards, or were present but at a lower frequency than detectable by our assay.

Nine point mutations in the FLT3 gene were identified with 8 located at codon 835 and 1 located at codon 842. The latter was discovered using Denaturing High Pressure Liquid Chromatography (DHPLC) while the remaining by PCR-RFLP; all were sequenced for confirmation. One patient exhibited two mutations in codon 835 (patient #0803, Table 1). We cloned the PCR product using TA cloning kit (Invitrogen) and sequenced 8 clones - the mutations were present on opposite alleles meaning that two oncogenic mutations existed in this patient. Guthrie cards were available for 4 of these patients, and backtracking experiments were successfully performed on 3 patients at a similar sensitivity as the indel mutations (1 cell in 40 ng DNA), and the three patients were negative for mutant sequence. The requisite reaction sensitivity for one of the patient samples with available cards was not obtained. In sum, between indel and point mutation samples, 9 cards were tested and all were negative for presence of FLT3 mutation.

Using NCCLS bone marrow samples, we have screened for FLT3 mutations in the largest sample of pediatric leukemias yet reported. Our results confirm the previously reported occurrence of FLT3 mutations in both pediatric ALL and AML although the incidence was lower than that of reported for some previous patient series. The highest rates of FLT3 mutation were recorded in population series of leukemias in Japan (9% of 162 ALL patients) [28] and Sweden, with 8% of ALL and 21% of AMLs (in children up to age 17 yrs) [9]. Lower frequencies were found in other population series in Greece (2.3% among 86 ALL patients), the UK (3.5% of 86 ALL patients), and Japan (1% of ALL 95 patients) [12]. Our study is the largest FLT3 screen in a pediatric population, and was performed in a population-based series of 517 cases. The rate of FLT3 among AML patients was 12.3%, which may be lower than some pediatric series since our study includes younger children only (< 15 years). The ALL rate of 2.0% found in here is comparable with the larger more recent reports. It is unlikely that we have missed pathogenetic FLT3 mutations in exons 11 and 15 since we incorporated a DHPLC screen of mutations. An additional notable difference between the current study and other studies is the apparent lack of a significant association between high hyperdiploidy and FLT3 mutation among our ALL patients, although this lack of significance may be due to small numbers. The association between RAS mutations and high hyperdiploidy is extremely strong in our patient sample set [29] and therefore we can confirm a RAS pathway association. It is unclear whether a lower prevalence of FLT3 mutations among hyperdiploid cases in California (compared to, for instance Paulsson et al., [11]) is due to differences in etiology of this subtype based on genetic or environmental characteristics of patients in California.

Our results suggesting that the FLT3 mutation is not present at birth is consistent with the hypothesis that FLT3 mutations are a second stage mutation. However, we cannot rule out that the mutations were present in some children but at a level beneath the sensitivity of the assay, or sequestered in the bone marrow and not in blood circulation. Our results corroborate that of Burjanivova, et al., who did not find evidence of prenatal origin of FLT3 in two AML patients [30]. The "two hit" model of leukemogenesis related to FLT3 proposed by Gilliland and Griffin, hypothesizes that two specific types of mutations are required in leukemia: one mutation that promotes proliferation and another mutation that stops differentiation [31]. Mutation of FLT3, like other tyrosine kinases is associated with proliferation and may complement mutations that impair cell differentiation such as the deletion of B-cell transcription factors or translocation-associated fusion genes. NRAS mutations and FLT3 both promote myeloproliferation, so their presence together is not necessary for leukemogenesis. The MLL gene rearrangement blocks differentiation and its association with FLT3 supports this model as well. None of our FLT3-mutant patients exhibited an MLL, TEL-AML1, AML1-ETO or other common translocation, so we were unable to examine a translocation concurrently with FLT3 in a backtracking experiment. However, strong prior evidence that high hyperdiploidy is a prenatal event [32–35], combined with no evidence of prenatal origin of FLT3 mutation in five high hyperdiploid patients here, helps to place FLT3 mutation postnatally.

In conclusion, FLT3 mutations are not common in our California childhood leukemia population, where RAS mutations are far more common [29]. Our investigation provided no evidence that FLT3 mutation occurs before birth, compatible with a hypothesis of postnatal origin for these mutations.