Loss of the JAK2 intramolecular auto-inhibition mechanism is predicted by structural modelling of a novel exon 12 insertion mutation in a case of idiopathic erythrocytosis


Francesco Rodeghiero, Department of Haematology, San Bortolo Hospital, Viale Rodolfi 37, 36100 Vicenza, Italy. E-mail: rodeghiero@hemato.ven.it


We report a novel gain-of-function JAK2 exon 12 insertion mutation in a patient with idiopathic erythrocytosis and low serum erythropoietin level. To date, only rare cases of such mutations have been reported in the JAK2 exon 12. Using computer-based structural modelling we propose that this mutation causes the loss of the JAK2 auto-inhibition step, leading to the constitutive activation of JAK2 tyrosine kinase-dependent activity. Our model-based hypothesis provides a useful approach for the investigation of the phenotype-genotype relationship in myeloproliferative disorders involving JAK2.

JAK2 belongs to the family of ubiquitously expressed cytoplasmic protein tyrosine kinases (PTKs) that play a critical role in mediating intracellular cytokines signalling triggered by erythropoietin (Epo), interleukin-3 and interferon-γ receptors (Ihle, 1995). Receptor-mediated JAK2 activation is normally rapid and transient, whereas aberrant constitutive activity is characteristic of a variety of tumours (Reiter et al, 2005). JAK2V617F exon 14 mutation is associated with polycythemia vera (PV), essential thrombocythemia and idiopathic myelofibrosis (Levine et al, 2005) while JAK2 mutations located in exon 12 have been recently found in patients with idiopathic erythrocytosis (IE) phenotype or in JAK2V617F negative PV cases (Cazzola, 2007; Scott et al, 2007). JAK2 is maintained in a low activity state through various auto-regulatory mechanisms, preventing the proper orientation for catalysis of the activation loop (A-loop) in the kinase domain (JH1; Huse & Kuriyan, 2002). Previous studies confirmed the regulatory role of the catalytically inactive JH2 pseudo-kinase domain, including exons 12–19, on modulating basal enzymatic activity (Luo et al, 1997; Chen et al, 2000; Saharinen et al, 2000, 2003; Saharinen & Silvennoinen, 2002). The crystallographic structure of JAK2 is not yet available, but computer-based homology models and extensive experimental data on mutant JAK2 indicate that the JH2 domain negatively regulates the activity of JAK2 by intramolecular interactions with the adjacent C-terminal JH1 domain (Saharinen et al, 2000, 2003; Lindauer et al, 2001; Saharinen & Silvennoinen, 2002). JH2 induces conformational JAK2 changes by engaging the JH1 domain in an intramolecular interaction that distorts the arrangement essential for the protein catalysis. Auto-inhibitory properties are dependent on three specific inhibitory regions (IR1–3) within JH2 domain that, when deleted, confer increased activity to the PTK (Saharinen & Silvennoinen, 2002). Complete definition of the interaction sites that involve JAK2 domain–domain interfaces is not known, but structural modelling of the combined JH2–JH1 inter-domain arrangement revealed two main interfaces (Lindauer et al, 2001). The former includes the N-terminal α-helix (αC) from each domain (interface 1) whereas the latter interaction site, named interface 2, takes place between the JH1 A-loop and a loop connecting N-terminal strands β4 and β5 of JH2 (Fig 2A).

Figure 2.

 (A) Wild-type three-dimensional (3D) model of the human JAK2 JH2 domain. The interfaces of interaction with the JAK2 JH1 domain are highlighted (Interface 1 and Interface 2). (B) 3D model of the mutated JH2 domain. The pool of the modelled loops corresponding to the 547insL+I540-F547dup8 mutation is represented by multiple coils. The interfaces of interaction with the JAK2 JH1 domain are highlighted. (C) View of the site of interaction between the N-terminal lobe of JH2 (F537 and K539) with the JH2 Interface 2 (V615). Hydrogen bonds are shown as dashed lines. (D) 3D model of the mutated JH2 domain. Site of interaction between the side chain of the arginine residue in the middle of the selected loop (Loop) with the charged aspartate (D620) lying in the JH2 Interface 2. (E) View of the site of interaction between the side chain of arginine residue (RMut) with the charged aspartate D620. Hydrogen bond is shown as a dashed line (distance between atoms = 2.73 Å).

We describe a new insertion mutation of JAK2 located in exon 12, the first encoding exon of JH2 domain, in a patient with IE phenotype negative for the JAK2V617F mutation. This new mutation was found by testing 40 consecutive IE and 4 PV patients, negative for V617F; in four of them (all IE) N542-E543del, already described (Cazzola, 2007), was detected. The patient with the new mutation, a male aged 61 years, was referred on January 1990 for erythrocytosis. Haemoglobin and haematocrit were greatly increased (230 g/l and 68% respectively) in the presence of normal leucocyte (7·5 × 109/l) and platelet (179 × 109/l) counts. Physical examination was negative. An increased red cell mass [53·9 ml/kg, normal range (NR), <35·0] was found. Epo-independent endogenous erythroid colonies (EECs) were obtained. Serum Epo level was below the normal range (<5 IU/ml, NR 5–25 IU/ml). A bone marrow biopsy showed marked hyperplasia of the erythroid lineage. As diagnosis the patient underwent regular phlebotomies (on the average one every 4 weeks). Between May 2001 and the time of writing, he has been under regular oral anticoagulation for atrial fibrillation. On November 2005 the patient was found to be negative for JAK2V617F. At last follow-up (October 2007) the subject remained JAK2V617F negative and showed a normal haematocrit (47%), haemoglobin level (149 g/l), leucocyte (6·3 × 109/l) and platelet (146 × 109/l) counts. Liver and spleen remained unchanged.

DNA was extracted from isolated granulocytes, erythroid colonies (ECs) and EECs and JAK2 exon 12 was amplified. Denaturing high performance liquid chromatography (DHPLC)-based screening (Fig 1A) and subsequent direct sequencing revealed the presence of a heterozygous complex mutation at the 3′ end of the exon 12. The same mutation was detected on cDNA from the patient’s peripheral blood cells total RNA. ECs showed a low level mutation (three mutated colonies every 30 analysed), while all seven patient-derived EECs were positive for the novel mutation (Appendix SI).

Figure 1.

 Identification of a novel JAK2 exon 12 mutation. (A) Abnormal DHPLC chromatogram of a sample from the mutated patient against the wild-type (wt) homoduplex control. Retention time (min) was plotted versus the Absorbance Intensity (mV) and the run was performed at 53°C. (B) Lane 1: DNA ladder 100 bp; Lane 2: patient’s DNA amplified with mutation specific primer (AS); Lane 3: healthy control DNA amplified with AS primer; Lane 4: mix 1 reaction control; Lane 5: patient’s DNA amplified with wt primer; Lanes 6: healthy control DNA amplified with wt primer; Lane 7: mix 2 reaction control. (C) Sequencing electropherogram of mutated fragment obtained by AS-PCR. (D) Comparison of JAK2_wt and JAK2_exon 12 mutant nucleotide sequences reveals a 27-bp insertion starting after coding position 547 (547insL+I540-F547dup8; GenBank accession no. NP_O60674).

Allele-specific polymerase chain reaction (AS-PCR) was performed to confirm the identity of the involved mutated nucleotides. Using the wild-type (wt) primer for amplification, definite bands of 200 bp were detected on agarose gel for the patient and healthy control; a single band of 225 bp was observed from AS-PCR only when using patient sample DNA (Fig 1B; Appendix SI). A sequencing electropherogram of the mutated allele showed a 5′ splice-site g>t transition (IVS12+1g>t) leading to the generation of a leucine codon (TTA), together with a 25-bp duplication of the terminal exon 12 nucleotides, resulting in a 27-bp in-frame insertion (547insL+I540-F547dup8, Fig 1C and D). Nucleotides insertion led to nine amino acids elongation of the predicted protein consisting of 1141 amino acids compared with 1132 in wt JAK2 (Fig 1D). The novel mutation lies at the beginning of the JH2 domain, occurring at a highly conserved amino acid region (http://toolkit.tuebingen.mpg.de/prot_blast).

We constructed a three-dimensional (3D) model of the JAK2 JH2 domain to investigate the intramolecular interactions on predicted mutated protein. For structural prediction we used as a template the X-ray structure at 2·6 Å resolution of the auto-inhibited intact human zap-70 protein (Protein Data Bank, entry 2ozo). This template was found using the HHpred server for homology detection and modelling (http://toolkit.tuebingen.mpg.de/hhpred). The amino acid sequence identity of the JAK2 JH2 domain and zap-70 was predicted to be 25%; however, root mean square deviation of the distance between superimposed α-carbons of the target and the template proteins was 1·23 Å, well below the resolution of the template protein.

Generation of the theoretical 3D structure was performed using the SwissModel automated comparative protein modelling server and DeepView V3.7 software. The sequence alignment of the JAK2 JH2 and 2ozo kinase domains was obtained with hhpred program package (http://toolkit.tuebingen.mpg.de/hhpred) and by combining available informations from PSIpred V2.0 and PredictProtein secondary structure predictors. Similarly to other modelled JAK2 pseudo-kinase domains, based on previously solved protein kinase structures and alternative modelling methods (Lindauer et al, 2001; Saharinen et al, 2003; Levine et al, 2005) the rebuilt JH2 model consisted of a N-terminal lobe of five-stranded antiparallel β-sheet and the αC followed by a mainly α-helical C-terminal lobe (Fig 2A). Loop modelling is one of the most difficult aspects in 3D structure prediction, for this reason several loop conformations corresponding to the 547insL+I540-F547dup8 variant were generated by scanning the DeepView loop_database. The 10 best fitting loops on the basis of general consensus rules were modelled to obtain a Nuclear Magnetic Resonance-like bundle of legitimate loop conformations (Fig 2B). The previously described procedure was performed to consider the probable conformational variability of the loop as a result of the high exposure to the solvent of this region.

V615 is predicted at the C-terminal end of strand β4, at the JH2–JH1 interface 2. The range of the interactions between the JH2 loop and the JH1 in the interface 2 would imply that A-loop translocation from an inactive to an active conformation of JH1 is inhibited by JH2 in a physiological manner. Internal rearrangement of the phosphorylated JH1, leading to the swing of its N-terminal lobe on the C-terminal lobe is influenced by the interactions with JH2 domain at this site (Lindauer et al, 2001). Based on the predicted possible loop conformations we hypothesize that the loop may approach to the interface 2, creating a new electrostatic interaction. The nature of the proposed interaction is based on the length of the new amino acids string and its physico-chemical properties. In fact, the positively charged arginine located in the middle of the newly created loop, was very close to the negative acidic residue D620 of JH2 in some of the chosen conformations. The proximity of the complementary-charged amino acids in the mutated JH2 model allows a salt-bridge ionic pair formation, taking place between the arginine side-chain of the mutated loop (RMut) and the aspartate residue D620 (Fig 2D and E). In wt protein kinase, D620 lies at the beginning of the first JH2 inhibitory region (IR1, spanning from G619 to I670), suggesting that a molecular perturbation at this site could lead to an alteration of the intramolecular auto-inhibition mechanism of JAK2 activity, allowing A-loop to assume an active conformation.

We further identified two conserved sites of interactions in wt and mutated JH2 domain between the N-terminal F537 and K539 with V615 backbone atoms (Fig 2C). The 9-amino acid insertion, occurring at the strand β1, preserves the F537-V615 and K539-V615 network of interactions. However, allelic variants involving JAK2 exon 12 often precisely involve F537 and K539 amino acids (Cazzola, 2007). The present study highlighted the existence of intradomain polar interactions between this region (F537 and K539) and the V615 residue belonging to the second JH2–JH1 interface. This could suggest a similar effect of other JAK2 exon 12 mutant-types resulting in the loss of the JAK2 auto-inhibition mechanism. The IE phenotype of our patient is similar to the previously described patients with JAK2 exon 12 mutations.

In conclusion, we propose that this novel exon 12 insertion mutation in the JH2 domain interferes with the interface 2 interdomain interaction site by creating a more favourable connection between mutated JH2 and JH1 domains. Extensive reported experimental data on mutant JAK2 (JH2 deleted) are available for comparison with the homology modelling predictions, but experimental solution of the JAK2 JH1–JH2 structure is needed to confirm our model-based hypothesis. However, our molecular model is in keeping with the known inhibitory role of the JH2 on the PTK activity and provides a useful approach for the investigation of the phenotype–genotype relationship in JAK2-related myeloproliferative diseases.


We are grateful to S. Finotto for clinical data. This work was supported by the Associazione Vicentina per le Leucemie, i Linfomi e il Mieloma (AVILL-AIL), Vicenza. EA and MB were recipient of a grant from the ‘Fondazione Progetto Ematologia’, Vicenza.