Identification of a germline F692L drug resistance variant in cis with Flt3-internal tandem duplication in knock-in mice

Internal tandem duplication (ITD) mutations in the juxtamembrane domain of the fms-like tyrosine kinase 3 (FLT3) gene occur in approximately one quarter of cases of acute myeloid leukemia (AML),1 are associated with constitutive activation of the kinase2 and confer a poor prognosis.3 Our understanding of the molecular consequences of these mutations has benefited from studies of bespoke mouse models.64 Herein we report the identification of the germline variant F692L in cis with the Flt3-ITD allele of the widely studied Flt3^tm1Dgg mouse.4 As this variant is analogous to the human FLT3 F691L “gatekeeper” mutation3 we investigated this finding further. We found that primary AML cells from Npm1cA; Flt3^tm1Dgg double-mutant mice are resistant to sorafenib and quizartinib (AC220), but sensitive to ponatinib. The same pattern of sensitivity was observed when we tested these tyrosine kinase inhibitors against isogenic Ba/F3 cells transfected with the murine Flt3-ITD F692Lor the reverted Flt3-ITD wild-type cDNA, confirming that Flt3-ITD F692L is responsible for recapitulating the resistance pattern of the human F691L mutation. The presence of this Flt3 tyrosine kinase inhibitor-resistant variant needs to be considered when interpreting data using this model, but also makes the model an extremely useful tool for studying tyrosine kinase inhibitor resistance.

The molecular consequences of FLT3-ITD are pleiotropic and include activation of STAT5 via the SRC kinase, phosphorylation of the transcription factor FOXO3A with associated activation of Akt and induction of reactive oxygen species, among others (reviewed by Yeung et al.3). Knock-in mouse models have been instrumental in deciphering the hematopoietic and leukemogenic effects of FLT3-ITD alone64 and in combination with mutations co-occurring with FLT3-ITD in human AML127; as well as in studying the effectiveness of different therapeutic approaches in FLT3-ITD-positive leukemias.1311 As part of these efforts, we studied the powerful collaboration between FLT3-ITD and mutant NPM1, by crossing our conditional Npm1^flox-cA knock-in14 with the Flt3-ITD knock-in mice (Flt3^tm1Dgg) generated by Lee et al.4 Triple transgenic Npm1^flox-cA/+; Flt3^ITD/+; Mx1-Cre⁺ animals rapidly succumbed to AML with a median latency of 49 days, exhibiting full recombination of the conditional Npm1^flox-cA allele without the need to induce Mx1-Cre expression with polyinosinic-polycytidylic acid as previously published.8 To identify or exclude additional cooperating somatic mutations involved in leukemic progression, we performed whole exome sequencing of five cases of Npm1^flox-cA/+; Flt3^ITD/+; Mx1-Cre⁺ AML. Comparison with constitutional (tail) exome sequence from C57BL/6N mice identified a common variant located at Chr5:147349699 (GRCm38), within exon 17 of the Flt3 gene. The variant, Flt3 c.2076 T>A (p.F692L), was present at allele frequencies of 0.55–0.94. Using polymerase chain reaction amplification followed by MiSeq next-generation sequencing we found that these variant allele frequencies correlated strongly with those of the Flt3-ITD allele itself (Figure 1A). As the Flt3-ITD mutation is located at exon 14 and frequently exhibits loss of heterozygosity in our model8 this indicated that the F692L variant is present in cis with the knock-in ITD. We went on to confirm this by genotyping germline DNA (tail) from Flt3^ITD/+ and Flt3^ITD/ITD single transgenic mice from our cohort (Figure 1B). Furthermore, we identified the same variant in germline DNA from the cohorts of two separate research groups who obtained the Flt3-ITD mouse model independently (Online Supplementary Figure S1).

Figure 1.Identification of the Flt3 c.2076 T>A (F692L) variant in cis with Flt3-ITD in knock-in mice. (A) The Flt3 c.2076 T>A exon 17 single nucleotide variant was identified in five of five exomes of Npm1^cA/+; Flt3^ITD/+; Mx1-Cre⁺ acute myeloid leukemia (AML) samples at varying variant allele frequencies (VAF) (left panel). These results were confirmed by polymerase chain reaction-amplification followed by MiSeq sequencing of AML DNA samples (T1–T5) and this also revealed a close correlation between the single nucleotide variant with the ITD VAF (right panel). (B) Sequencing of tail DNA from Flt3^+/+, Flt3^ITD/+ and Flt3^ITD/ITD mice confirmed that the single nucleotide variant is present in cis with the Flt3^tm1Dgg allele and was not particular to the Npm1^cA; Flt3^tm1Dgg double mutant cohort. (C) Protein sequence alignment of mouse and human FLT3 protein demonstrates that the mouse p.F692 corresponds to human p.F691, target of the F691L gatekeeper tyrosine kinase inhibitor resistance mutations.

As the murine F692L variant is equivalent to the human F691L gatekeeper mutation, which confers AML resistance to multiple FLT3 tyrosine kinase inhibitors,153 we proceeded to test whether this was also true for the murine variant. To do this we cultured AML cells from two independent Npm1^cA/+; Flt3^ITD/+; Mx1-Cre⁺ mice8 in X-VIVO medium (Lonza, Switzerland) supplemented with murine inteleukin-3, interleukin-6 and stem cell factor (Miltenyi, Germany). After 72 h in culture, we removed the interleukin-3 and exposed the cells to varying concentrations of quizartinib (AC220), sorafenib and ponatinib for 3 days, using the human FLT3-ITD-positive AML cell line MV4-11 as a positive control. After 3 days of treatment, cell viability was assessed using the CellTiter 96® AQueous assay (Promega, USA). We found that, compared to MV4-11, Npm1^cA/+; Flt3^ITD/+; Mx1-Cre⁺ AML cells exhibited resistance to quizartinib and sorafenib, but not to ponatinib which is known to retain activity against the human FLT3-ITD F691L mutant isoform1615 (Figure 2A).

Figure 2.The Flt3 c.2076 T>A (F692L) variant confers resistance to quizartinib and sorafenib. (A) Two independent leukemic cell lines, derived from Npm1^cA/+; Flt3^ITD/+ mice, were cultured in X-VIVO-20 medium (Lonza) (supplemented with 5% fetal calf serum, 10 ng/mL murine interleukin-6 and 50 ng/mL murine stem cell factor) and incubated with Flt3 tyrosine kinase inhibitors over a range of concentrations. IC₅₀ values for each murine leukemia and for the control cell line MV4;11, in response to each treatment are displayed. (B) MSCV-Flt3-ITD-CFP retroviral constructs used to stably transduce CellSensor® irf-bla Ba/F3 cells. Both the Flt3-ITD^F692L and the reverted Flt3-ITD^F692F cDNAs transformed the Ba/F3 cells to interleukin-3 (IL-3) independent growth with growth kinetics similar to those of IL-3-stimulated cells transduced with an MSCV-CFP (empty) vector (see also Online Supplementary Figure S2). (C) Stably transduced mutant Ba/F3 cell lines and the FLT3-ITD-positive AML cell line MV4;11 were incubated with Flt3 inhibitors at a range of concentrations. The MSCV-Flt3-ITD^F692L-CFP construct conferred Ba/F3 resistance to both quizartinib and sorafinib, but not to ponatinib; cells transduced with the MSCV-Flt3-ITD^F692F-CFP construct remained sensitive to all three inhibitors. IC₅₀ values for Flt3ITD^p.F692L (blue) or the reverted Flt3-ITD^F692L cDNA (red) transduced cells are displayed inset for each drug. All assays were performed in triplicate.

To exclude the possibility that our murine leukemias and MV4-11 displayed different resistance patterns for reasons other than the F692L mutation, we amplified the full open-reading frame of the Flt3 cDNA from bone marrow mRNA of homozygous Flt3^ITD/ITD mice and cloned this into an MSCV-CFP retroviral backbone (kind gift from Dr. Brian Huntly, University of Cambridge, UK) (Figure 2B). We then also generated a reverted c.2076A>T (Flt3-ITD) construct in the same MSCV-CFP retroviral backbone using Gibson assembly (see Online Supplementary Methods). The two retroviruses were each used to transduce interleukin-3-dependent CellSensor® irf-bla Ba/F3 cells (Invitrogen, Carlsbad, USA), which were subsequently sorted to >95% purity (Online Supplementary Methods and Online Supplementary Figure S2A). Both the Flt3-ITD and Flt3-ITD cDNA-transfected cells displayed comparable interleukin-3-independent growth patterns (Figure 2B) and Forster resonance energy transfer assays showed concomitant activation of the JAK/STAT pathway, recapitulating the effects of human FLT3-ITD mutants (Online Supplementary Figure S2B). Next, we treated the transformed Ba/F3 cell lines as described above for the primary AML cells. This demonstrated that cells transformed with Flt3-ITD were resistant to quizartinib and sorafenib, but not ponatinib, unlike wild-type Flt3 cells that were sensitive to all three compounds (Figure 2C). These data, taken together with our results with the primary AML cells, confirm that the murine F692L mutation mimics the properties of the human F691L mutant FLT3-ITD and confers a similar profile of resistance to FLT3 inhibitors.16

Here, we show that the Flt3^tm1Dgg knock-in mice carry the Flt3 F692L resistance variant in cis with the human W51 Flt3-ITD mutation. The variant corresponds to the human F691L and shares at least some of its drug sensitivity/resistance profile. The origin of the c.2076A>T (p.F692L) variant within the Flt3^tm1Dgg allele is not clear. Examination of the strategy and design of the original gene targeting construct revealed that this did not extend to exon 17, which contains codon F692. It does, therefore, appear unlikely that the c.2076T>A single nucleotide variant was introduced into the mouse Flt3 locus during gene targeting.4 Furthermore, the presence of the variant in three out of three independent Flt3^tm1Dgg mouse colonies tested rules out the possibility that the variant was acquired uniquely in our cohort. The single nucleotide variant was, therefore, either present in the original embryonic stem cells used to generate the knock-in mice or was acquired in the germline during the expansion of the original colony.

Notably this single nucleotide variant is not present in any of the 48 sequenced mouse strains according to the mouse genome informatics (MGI) database (http://www.informatics.jax.org/javawi2/servlet/WIFetch?page=snpQF). Nevertheless, regardless of its origin, the presence of this variant needs to be taken into account when interpreting the findings of the many studies using Flt3 mice137 particularly those investigating therapeutic approaches.1311 Additionally, our findings make the Flt3^tm1Dgg model particularly useful for studying novel tyrosine kinase inhibitors such as ponatinib and PLX3397, which retain activity against the human F691L gatekeeper mutation.15 Moreover, our findings emphasize the importance of careful characterization of genetically modified mice before drawing conclusions about the pathologies they model, particularly when targeted therapeutic interventions are being investigated. Although the genetic variant influencing drug response was not present in wild-type mice in our study, in other cases it may represent a consistent human - mouse difference, as was recently described for lenalidomide and a mouse-specific variant in Cereblon (Crbn) that makes murine cells insusceptible to the drug.17 Given the availability and relative affordability of next-generation sequencing, we propose that genomic characterization, such as exome sequencing, of frequently-used or novel mouse models should be applied to identify coding variants in order to avoid misinterpretation of findings from these otherwise highly valuable biological reagents.