MLL rearrangements are a genetic hallmark of acute leukemia patients, which exhibit a particular poor outcome. To date, more than 70 MLL rearrangements have been described at the molecular level.1 For the most frequently diagnosed MLL rearrangements, for example, AF4-MLL, MLL-AF9, MLL-AF10 or MLL-ENL, it has been shown that these fusion proteins are sufficient for acute leukemia onset in murine model systems.2 However, these models had a latency time of 4–12 months for the disease phenotype to become overt. This argues in favor of pre-leukemic clones that carry the fusion genes, but need, in addition, complementing mutations to develop a malignant disease.

In case of MLL rearrangements—specifically MLL-AF4—it has been shown that copy number aberrations could not be detected through whole genome high-resolution technologies,3 favoring the possibility that single nucleotide activating mutations, such as mutated Fms-like tyrosine kinase 3 (FLT3) or RAS, are more important than previously reported.4, 5, 6 Therefore, we investigated the incidence of RAS and FLT3 mutations in a large cohort of MLL-rearranged leukemia patients (n=144). This cohort of MLL-associated leukemias was split into three subgroups. The t(4;11) group, which was of major interest includes 21 t(4;11) patients displaying complex genomic rearrangements (>2 fusions alleles; 11 pediatric, 10 adult), whereas 79 t(4;11) patients had a balanced translocation (31 pediatric, 48 adult). The control group of 44 patients comprised cases with MLL-AF9 (n=22; 19 pediatric, 3 adult), MLL-ENL (n=14; 13 pediatric, 1 adult), MLL-AF10 (n=4; 4 pediatric), MLL-ELL (n=2; 2 pediatric) and, finally, 1 case with MLL-AF6 (adult) and one with MLL PTD (adult), respectively. All patient material has been made available by the I-BFM and German multicenter acute lymphoblastic leukemia (ALL) study groups.

Genomic DNA of these patients was investigated by amplifying the corresponding genomic regions of the N-RAS and K-RAS genes (exons 2 and 3), and of the FLT3 receptor gene (exons 13–15 and exon 20) and, subsequently, by sequencing the resulting amplimers. FLT3 is predominantly expressed in hematopoietic stem/progenitor cells. In addition, FLT3 is expressed at considerable levels in most clinical samples from acute myeloid leukemia (AML) and B-cell precursor ALL patients. Internal tandem duplication of the juxtamembrane domain of the FLT3 gene is frequently observed in AML. AML patients bearing a particular FLT3 mutation (D835; FLT3-TKD) tend to have poor prognosis, suggesting that FLT3-TKD mutations have an important role in AML.7 Otherwise, it has been published that FLT3-TKD mutations were found in approximately 15% of patients with de novo mixed-lineage leukemia (MLL) rearrangements.5 However, these patients were screened by Southern blot experiments, without gaining information about the involved MLL fusion partner.

In the cohort of 144 MLL-rearranged leukemia samples we analyzed, none of the investigated patients—regardless of whether diagnosed with balanced, complex t(4;11) rearrangements or other MLL fusions—displayed a genomic FLT3 mutation. This is in line with data reported in the literature that mutant FLT3 is mostly diagnosed in hyperdiploid leukemia cases.8

The reason for the absence of mutant FLT3 receptor genes in MLL-rearranged leukemia might be that a mutant FLT3 receptor gene confers a selective disadvantage for MLL-rearranged cells. Published data demonstrates that MLL-rearranged leukemia cells seem to overexpress wild-type FLT3 receptor and that stimulation of MLL-rearranged leukemia cells with FLT3-ligand (FL) causes quiescence, leading to chemoresistance.9 As a result, MLL-rearranged leukemia cells get protected against chemotherapy when residing in the bone marrow and binding to FL-expressing stroma cells. This also explains why a high FLT3 expression in leukemia cells directly predicts a high relapse rate and poor outcome.10 Thus, MLL-rearranged leukemia seems to be addicted to wild-type FLT3 receptor, and not to mutant FLT3. This becomes obvious when recent findings are considered: FLT3–internal tandem duplication mutations cause oncogenic signaling from the endoplasmatic reticulum, which—different from wild-type FLT3 receptors—causes a direct phosphorylation of STAT5, and subsequently, an upregulation of SOCS proteins. SOCS proteins block cytokine signaling from the cell surface, by inhibiting the family of Janus kinases (JAK1-3, TYK2), which results in a “shielding effect” against external cytokine signals.11 However, MLL-rearranged cells depend on such external signals deriving from the expressed FL on bone marrow stroma cells. Thus, inhibition of cell surface signaling through a mutant FLT3 receptor represents a counter-productive event that seems to be selectively omitted in MLL-rearranged leukemia, which could be a reason for the absence of FLT3 mutations in our study.

By contrast, our RAS gene analyses revealed mutations in MLL-rearranged patients. The majority of point mutations was identified in pediatric patients (n=31) bearing a balanced t(4;11) translocation (n=8; 1 × N-RAS, 7 × K-RAS), whereas adult patients with balanced t(4;11) translocations (n=48) had significantly less mutated RAS genes (n=4; 2 × N-RAS, 2 × K-RAS; P=0.035). Four additional RAS mutations were identified in pediatric patients with MLL-AF9 (n=1; 1 × K-RAS), MLL-ENL (n=2; 1 × N-RAS, 1 × K-RAS) and MLL-AF10 (n=1; 1 × K-RAS). As we only had six adult cases in this subgroup of MLL-rearranged leukemia patients, statistical evaluation was not eligible. Regarding the three different subgroups of the investigated cohort (Figure 1), no statistical significance could be observed. All identified N-RAS point mutations—except one G12C exchange—resulted in missense G12D mutations. In one pediatric t(4;11) patient, an N-RAS G12D/G13D double mutation was identified. By contrast, the spectrum of K-RAS mutations was more diverse, including G10A, G12A, G12D, G12V and G13D missense mutations. Of interest, the group of patients with complex MLL rearrangement (n=21; 11 × pediatric, 10 × adult) did not exhibit any RAS mutation. All these data were summarized in Figure 2.

Figure 1
figure 1

The cohort of 144 patients is split into 21 t(4;11) leukemia patients that displayed a complex rearrangement of the MLL gene (with 3–4 fusion genes), 79 leukemia patients that displayed a balanced t(4;11) translocation and 44 leukemia patients that displayed several distinct MLL rearrangements. The number of pediatric and adult patients is displayed for each subgroup. We identified a total of 12 pediatric patients with K- or N-RAS mutations. A similar situation was found in four adult t(4;11) patients that displayed again K- or N-RAS mutations. Of three investigated relapse samples, only one patient retained its RAS mutation.

Full size image
Figure 2
figure 2

Summary of all data obtained in patient analyses. UPN, genetic rearrangement, age at diagnosis, identified mutation and sequence chromatogram is displayed. For most patients with RAS mutations, no remission samples were available, because these patients are still in remission or no information was available for these patients.

Full size image

To understand the unusual bias of mutant RAS genes in the group of infant/pediatric t(4;11) patients, we tried to analyze the matched relapse samples. For most t(4;11) patients, this material was not available, because the patients are either still in remission (n=5) or no information for a relapse could be provided (n=8). Thus, we obtained only three matched relapse samples (one adult and two pediatric cases with balanced t(4;11) translocation).

To our surprise, two patients lost their mutated RAS allele upon relapse, whereas one pediatric case still retained its mutation (Figure 2). This clearly argues in favor for the presence of different subclones in the diagnostic tumor sample, which are (1) only supportive during disease onset, but presumably not necessary for disease maintenance, or (2) that mutated RAS genes are recognized by the host immune system, and thus effectively cleared by T-cell clones, or (3) that RAS mutations identified in diagnostic samples are just coincidence.

In conclusion, we found N- and K-RAS mutations in 26% of pediatric patients that exhibit a balanced t(4;11) translocation. We also demonstrate that RAS mutations can be readily diagnosed in adult t(4;11) leukemia patients, however, in a much lower frequency (8%). In view of this, it is interesting that a recent study demonstrated how ectopic expression of MLL-AF4, MLL-AF5 or MLL-LAF4 fusion proteins led to activated ELK-1 protein, a downstream target of the RAS/RAF signaling pathway.12 Thus, these fusion proteins are per se able to activate the RAS/RAF signaling cascade by themselves and, therefore, RAS mutations are actually not necessary. By contrast, in a humanized MLL-AF10 model, the additional expression of mutant K-RAS led to an AML M5 leukemia disease phenotype, whereas MLL-AF10 expression alone did not.13 This indicates that a certain threshold of RAS signaling may be necessary to establish a leukemic disease in the murine system.

Considering the overall frequency of RAS mutations in our analyses, the above mentioned mouse data are hardly comparable to the human system. The fact that we found a significant higher mutation rate in pediatric t(4;11) patients (26% pediatric versus 8% adult) could be interpreted in a way that RAS mutations are a potential second hit for early leukemia onset. However, we disagree with this notion because a 26% frequency of additional mutations is not sufficient to postulate a two-hit model for ALL childhood leukemia. Therefore, other interpretations are necessary. By now, we can only speculate about a biological function, but there is a significant difference between mutant and physiological RAS signaling. Downregulation of phosphoinositide-3 kinase signaling and activation the ATM/ATR-induced DNA damage response system may cause a delay of tumor development.14 In addition, an additional RAS mutation might be recognized by the mother’s immune system, and thus preventing tumor outgrowth already in utero.

To this end, further work will be necessary to find a satisfactory explanation for these observations, and to understand at the molecular mechanisms triggered by mutated RAS signaling. Understanding the importance of oncogenic signaling for the biology of MLL-rearranged leukemia may bear the potential to identify novel drug targets that can be therapeutically addressed in future therapy regimens. This might be true for signaling events deriving from highly expressed FLT3, as inhibition of this receptor by PKC412 and CEP-701 resulted in a selective killing of childhood leukemia cells.15