Molecular and functional characterization of a new 3′ end KIT juxtamembrane deletion in a duodenal GIST treated with neoadjuvant Imatinib

Gastrointestinal stromal tumors (GISTs) are the most common mesenchymal tumors of the gastrointestinal tract. GISTs express the receptor tyrosine kinase KIT, and the majority of GISTs present KIT gain-of-function mutations that cluster in the 5' end of the receptor juxtamembrane domain. On the other hand, little information is known about GISTs carrying mutations in the 3′ end of the KIT juxtamembrane domain. Here we report and discuss a clinical case of localized duodenal GIST whose molecular characterization revealed the presence of a new 21 nucleotide/7 amino acid deletion in the 3′ end of KIT juxtamembrane domain (Δ574–580). The patient was treated with Imatinib at standard regimen dose (400 mg/day), and responded well as the original tumor mass reduced, ultimately allowing conservative surgery. In line with these clinical evidences computer simulations, biophysical techniques and in vitro experiments demonstrated that the receptor tyrosine kinase KIT carrying the Δ574–580 mutation displays constitutive phosphorylation, which can be switchedoff upon Imatinib treatment. In addition, results from this study showed that a clinical useful procedure, neoadjuvant treatment, can occasionally be of value for the understanding of the molecular pathogenesis of GIST.


INTRODUCTION
Gastrointestinal stromal tumors (GISTs) are a relatively rare entity, accounting for less than 1% of GI tumors; nonetheless, they represent the most common mesenchymal tumors of the GI tract. GISTs arise from the muscle layer and are usually found in the stomach (60-70%), the proximal small intestine (25-30%), but can occur anywhere along the GI tract, exceptionally in the esophagus [1,2]. In nearly 80-90% of GISTs, the oncogenic driver is a gain-of-function genetic alteration (mutations) in the receptor tyrosine kinase (RTK) KIT and it is now widely believed that GISTs arise from KITexpressing interstitial cells of Cajal or their precursors [3].

Research Paper
Oncotarget 56159 www.impactjournals.com/oncotarget The vast majority of KIT mutations (60-70%) are in-frame deletions (other include missense mutations (20-30%) and internal tandem duplications) clustered in the 5′ end of KIT juxtamembrane (KIT-JM) domain (exon 11) between Q550 and E561. Alterations in the 3′ end distal part of KIT-JM are rarely reported, and these include missense point mutations in codon L576, in-frame deletions, and rare internal tandem duplications of 1 up to more than 20 codons that are more often observed in gastric GISTs and associated with a favorable outcome [4]. Functional and molecular characterization of these rare 3-end variants is lacking. Other alternative mutational "hotspots" in KIT extracellular (exon 9, 18% of cases) and kinase (exon 13 and 17, less than 2%) domains have been identified in the GISTs that are negative for the exon 11 mutation [5,6]. In the absence of KIT mutations, GISTs can harbor mutations of the PDGFRA gene (5-15% of cases), mainly located at exons 12 and 18, which are homologous with KIT exons 11 and 17 [7,8]. The remaining cases (12-15%) lack KIT and PDGFRA mutations (KIT and PDGFRA wild-type GISTs), but these may include BRAF mutations (3%), loss of function of succinate dehydrogenase (SDH) complex (3%), and NF-1 mutations [9].
Seventy five % of GISTs are less than 4 cm of size, and it is recommended that tumors greater than 2 cm of size should undergo surgical resection [10]. In particular, approximately 20-25% of gastric and 40 to 50% of small intestinal GISTs are clinically malignant, metastases commonly develop in the abdominal cavity and liver, and may develop 10 to 15 years after primary surgery necessitating long-term clinical follow-up [2].
Most GISTs are localized and are managed by surgery alone; however, locally advanced or metastatic cases, as well as high-risk operated GISTs, require systemic therapy with Imatinib, a small molecule tyrosine kinase inhibitor. Response to Imatinib correlates with the presence and type of RTK KIT mutations. GISTs with the most common 5′ end KIT exon 11 mutations show the highest response rates, whereas the responsiveness of exon 9 KIT mutations appears to be sensitive to increased drug concentrations [11]. On the other hand, information on Imatinib activity in case of the rare 3′ end KIT-JM mutations is lacking. Neoadjuvant Imatinib is now considered a valuable option for treating KITmutated GISTs [12]; it can render a locally advanced GIST resectable, allow to perform less invasive procedures or to promote preservation of function, especially if the tumor is located in an anatomically difficult position, as in case of the rare subset of GISTs that arise in the duodenum (2-5% of cases) [13]. Furthermore, neoadjuvant treatment has the potential to be of value for tumor biology, providing a way to gain in vivo information on Imatinib sensitivity in case of new mutations of undetermined functionality.
Under this perspective, in this work we report and discuss a clinical case of localized duodenal GIST whose molecular characterization revealed the presence of a new 21 nucleotide/7 amino acid deletion in the 3′ end of KIT-JM domain (∆574-580). The patient was treated with Imatinib at standard regimen dose (400 mg/day), and responded well as the original tumor mass reduced, ultimately allowing conservative surgery. In line with these clinical evidences computer simulations, biophysical techniques and in vitro experiments demonstrated that the TKR KIT carrying the ∆574-580 mutation displays constitutive phosphorylation, which can be switched-off upon Imatinib treatment.

Patient clinical history
On March 2015 a 33-year-old man with silent medical history presented at the emergency unit with acute GI tract hemorrhagic anemia. No other clinical signs or symptoms were present. A CT scan showed a lesion in the II duodenal tract and the patient was admitted. Endoscopy showed persistent bleeding from the submucosal ulcerated duodenal mass. Contrast-enhanced abdomen NMR detected an oval mass (3.4 × 2.2 × 3.1 cm) with regular margins, in the medial aspect of the duodenal wall, immediately distal to the papilla and extending to nearby pancreatic head, displaying contact with the inferior cava vein ( Figure 1, top panel, left).
Cytology obtained via echoendoscopy was highly suggestive for GIST, spindle cell type. Molecular typing was not readily possible, clinical conditions prompted intervention and the patient was eager to start treatment. The mass was technically resectable but required pancreatic-duodenectomy, with consequent significant morbidity. Bleeding indicated significant growth. 18 F-FDG-PET/CT scan showed intense metabolic activity (SUV 6.3) (Figure 1, top panel, right), as expected for most GISTs, providing a way for testing Imatinib activity after a short drug trial in the context of a neoadjuvant approach (despite limited information in duodenal GISTs, Imatinib-sensitive mutations involving exon 11 are not uncommon) [14]. Indeed, a multidisciplinary evaluation proposed neoadjuvant treatment at the standard 400 mg/ day Imatinib dose, which was started on early April 2015. Concomitantly, the patient underwent also a complete cardiac evaluation. Routine liver and renal function tests were within normal range.

KIT and PDGFRA molecular analysis
DNA sequencing of the duodenal GIST identified a new 21 codon in frame deletion c1718:1739del21 (p.T574_H580delTQLPYDH) encompassing the internal part of the juxtamembrane zipper region of KIT (JM-Z, 10 amino acids, residues 572-581), an area adjacent to the first trans-phosphorylation sites (Tyr 568 and Tyr 570) (see Supplementary Figure 2). As result of the deletion, the JM-Z consisted of just the 3 amino acids DPK. Interestingly, mutations described in this area include point mutations and internal tandem duplications [4].
No mutations were found in the KIT exons 9, 13 and 17, as well as in PDGFRA exons 12, 14 and 18 hot spots (Supplementary Figure 3).

Patient evaluation after Imatinib treatment
After prolonged discussion with the patient and informed consent obtained, Imatinib was started at the usual daily dose of 400 mg. Patient was monitored for hemoglobin concentration and occult fecal blood. Counts were stable and at day +30 of treatment, metabolic activity was not detected at a second FDG-PET-CT ( Figure 1, middle panel, right), attesting optimal Imatinib sensitivity. The treatment was protracted till conservative surgery could be planned (approximately 8 months later). At that time, CE-NMR documented a mass of 1.3 cm size ( Figure  1, middle panel, left) that allowed conservative surgery. Segmental duodenectomy with end-to-end duodenal reconstruction [14] was performed with optimal outcome, tumor resection was radical. Imatinib was started again as soon as clinical conditions allowed, and this because tumor histology after neoadjuvant treatment cannot reliable predict risk category. The patient was free of recurrence at last contrast-enhanced CT scan of the abdomen (approximately 1.5 years after surgery, Figure 1, bottom panel).

Computational results
The calculated free energy of binding values listed in upper part of Table 1 show that the newly reported ∆574-580 mutant receptor is endowed with an affinity toward Imatinib (∆G bind = −8.58 kcal/mol) slightly lower than that of the activating but Imatinib-responsive ∆559 isoform (∆G bind = −9.15 kcal/mol , ∆∆G bind = −0.57 kcal/mol). Contextually, this deletion mutant seems to be provided with better affinity to the inhibitor compare to the WT receptor (∆G bind value is −8.19 kcal/mol, [15]). From the structural point of view, no significant perturbation of the overall 3D structure of the Imatinib/∆574-580 complex is detected compared to the Imatinib responsive KIT ∆559 isoform ( Figure 2). On the other hand, the T670I is found very resistant to the TK inhibitor, as expected (∆G bind = −6.38 kcal/mol).
In order to gain further insight into the binding mode of Imatinib to ∆574-580 KIT mutant receptor, the energetic contribution (∆G bind,RES ) for those residues which afford a substantial contribution to the binding was calculated. As shown in the left panel of Figure 3, the major contribution to ∆574-580 KIT/Imatinib binding stems from the combined action of four hydrogen bonds involving the side chains of residues E640, T670, C673, and D810 and further stabilizing interactions provided by two clustered hydrophobic regions (HRI and HRII) of the receptor including residues: A621, Y672, and L799 (HRI) and V643, L783, and H790 (HRII), respectively. The strength of the hydrogen bonds network is ensued by the persistence and optimal average dynamics length (ADL) of these interactions, monitored during all the equilibrated MD simulation. Indeed, the ADL values for each residue involved fall in the typical range of the permanent hydrogen bond interactions (ADL E640 = 1.93 ± 0.01 Å; ADL T670 = 1.91 ± 0.03 Å; ADL C673 = 2.15 ± 0.04 Å; ADL D810 = 2.05 ± 0.01 Å). Such efficient intermolecular interaction scheme accounts for the most substantial favorable energetic term to the total drug binding free energy (∆G bind = −8.58 kcal/mol, Table 1), as these residues per se afford a stabilizing contribution of -8.23 kcal/mol (Figure 3, right panel).
In the case of the Imatinib/KIT ∆559 complex, no meaningful differences are detected at the individual residue binding level (Figure 3, right panel), and the identified interactions are in agreement with those described in previous works for these prototypical Imatinib-responsive KIT mutant [15][16][17][18][19][20]. Conversely, in the case of the Imatinib/KIT T670I assembly, the overall reduction of drug affinity cannot be attributed either to a single residue or to a particular cluster of binding site residues. Instead, a general decrease of the binding energy for all residues involved is predicted. This analysis confirms that the conformation adopted by the newly reported ∆574-580 KIT mutant and, in particular, of its inhibitor binding site, is comparable to the one characterizing the Imatinib-sensitive ∆559 KIT receptor; concomitantly, the quantitative examination of the essential molecular determinants for Imatinib binding provides a sensible explanation of the efficiency of this inhibitor against this new KIT juxtamembrane mutant.
The lower half of Table 1 lists the ∆G bind value obtained in silico for the ∆574-580 KIT mutant. The calculated ∆G bind values for the reference KIT variants ∆559 and T670I are also reported for comparison. As seen from these numbers, and from the related ∆∆G bind values in the last row of the top half of Table 1, the ∆574-580 KIT isoform shows an affinity for ATP (∆G bind = −24.02 kcal/mol) utterly comparable to that of the ∆559 (∆G bind = −24.20 kcal/mol, ∆∆G bind = −0.18 kcal/mol) and the T670I (∆G bind = −25.14 kcal/mol, ∆∆G bind = +0.94 kcal/mol) mutants, supporting the activating character of these mutations [15][16][17]19].

Isothermal titration calorimetry (ITC)
Isothermal titration calorimetry (ITC) experiments were to validate in silico prediction of KIT mutants affinity for Imatinib via determination of the drug binding thermodynamics (i.e., ∆G bind,ITC and its major enthalpic (∆H bind,ITC ) and entropic (-T∆S bind,ITC ) components, the The secondary structure of the protein is in ribbon representation (red, α-helices; magenta, β-sheets; gray, turns and coil, while Imatinib is portrayed via its van der Waals surface. Na + and Clions and counterions are shown as purple and green spheres, respectively. Water molecules are depicted as transparent, light blue spheres. Hydrogen atoms are omitted for clarity. Superposition of MD equilibrated snapshots of ∆559 (green) and the ∆574-580 (blue) KIT in complex with Imatinib (center) and details of the corresponding Imatinib binding site (right).
Oncotarget 56162 www.impactjournals.com/oncotarget dissociation constant K d,ITC , and the binding stoichiometry n). Figure 4 shows the titration curves for Imatinib binding to ∆574-580, ∆559 and T670I KIT isoforms, while the relevant numerical results obtained from data fitting with a binding model that assumes one set of identical binding sites are displayed in Table 2.
As seen in Table 2, the stoichiometry of Imatinib binding to all KIT variants is approximately 1, confirming binding of one inhibitor molecule per RTK. The derived K d,ITC confirm strong Imatinib binding to both deletion mutants (359 nM and 127 nM for ∆574-580 and ∆559 KIT, respectively). The ineffectiveness of Imatinib is obvious for T670I, with a K d,ITC of approximately 30 μM. This analysis clearly demonstrate that Imatinib binds to the inactivated forms of the newly reported deletion mutant ∆574-580 with affinity comparable to that of the activating yet responding ∆559 isoform, and the loss of binding affinity of the inhibitor for the notoriously resistant T670I KIT variant. Finally, a comparison of the experimentally derived ∆G bind,ITC values (Table 2) with the computer-predicted ones (∆G bind , Table 1) reveals that the two data sets are in excellent agreement, ultimately yielding a direct validation of the in silico results presented above.

Biological effects of the Δ574-580 mutation on KIT receptor activity
Next, the effect of Imatinib on the activation of the ∆574-580 KIT mutant was examined. Again, the same study was conducted in parallel on the reference ∆559 and T670I KIT variants for comparison. Accordingly, the HEK293T cells previously transfected with the mutated KIT expression constructs were cultured in absence and in presence of 1 and 5 µM Imatinib, respectively. As shown in Figure 5A, all proteins are expressed and phosphorylated, and both the mature (145 kDa) and partially glycosylated (125 kDa) forms of the receptor could be detected by immunoblotting of the lysates. The new ∆574-580 KIT mutant reveals constitutive phosphorylation of both 145 and 125 kDa forms by the Western blot with anti-phospho-KIT antibodies. This behavior matches that observed for the well-defined ∆559 and T670I KIT variants, thereby confirming the constitutive KIT receptor-activating role of the ∆574-580 deletion, in agreement with the computational predictions reported in Table 1.
Also, again in accordance with in silico calculations/ ITC determinations the newly reported KIT deletion .94 According to its definition (∆∆G bind = ∆G bind,∆559 -∆G bind,mut ), negative values calculated for ∆∆G bind indicate that the considered amino acid substitution at a given position of KIT is unfavorable in terms of imatinb/ATP affinity, whereas positive ∆∆G bind values indicate that the considered mutation is favorable. Oncotarget 56163 www.impactjournals.com/oncotarget variant is sensitive to Imatinib treatment at both concentrations ( Figure 5A). Moreover, it exhibits a profile similar to that of the prototypical Imatinib-sensitive ∆559 KIT, in that the relative KIT phosphorylation levels are strongly reduced compared to the untreated cells (about 75%, Figure 5B). On the other hand, the T670I KIT mutant does not show any reduction of the KIT phosphorylation level in the presence of Imatinib, in line with its intrinsic Imatinib-resistant profile.
Furthermore, the effects of the new reported deletion on the typical signaling pathway triggered by KIT were investigated by Western blotting in absence and in presence of Imatinib. As shown in Figure 5C, phosphorylation of ERK1/2 and AKT is detectable in untreated cell expressing ∆559 or ∆574-580, thereby confirming the constitutive activation character of the new deletion KIT mutant. Interestingly, at both Imatinib concentrations (1 µM and 5 µM), the inhibition promoted by the inhibitor switches off the ∆574-580 KIT pathway, as both AKT and ERK1/2 phosphorylation is reduced (Figures 5D and 5E). As a similar effect is seen for the well-known Imatinib-sensitive ∆559 KIT isoform, these evidences support the Imatinibsensitive behavior of this new KIT juxtamembrane variant, as anticipated by computer-assisted predictions and confirmed by isothermal titration calorimetry experiments.

DISCUSSION
Resistance to tyrosine kinase inhibitors is a major problem in cancer targeted therapy. Gastrointestinal stromal tumors are not exception to this, in that the majority (nearly 85%) of GISTs patients present primary activating mutations in the gene encoding the mast/stem cell growth factor receptor KIT, a type III RTK that plays an essential role in the regulation of cell survival and proliferation, hematopoiesis, stem cell maintenance, gametogenesis, mast cell development, migration and function, and in melanogenesis. Most of the reported KIT mutations in GISTs (missense, deletion and deletion/insertion variants) cluster in the receptor juxtamembrane domain (residues 550-586), encoded by the exon 11 of the related KIT gene. According to the COSMIC database, the most often reported KIT-JM mutations are point substitutions at positions 557 (e.g., W557R/G/S/C, V559D/A/G, and V560D/A/G), thereby affecting the 5′ end of the KIT-JM domain. At the same time, the efficacy of the tyrosine kinase inhibitor Imatinib in GISTs has been linked to the capacity of this small molecule to inhibit these aberrant receptor-activating mutant KIT proteins. Alterations at the other extreme of the KIT-JM (3′ end) are much less frequent, and these include missense point mutations in codon L576, in-frame deletions, and rare internal tandem duplications of 1 up to more than 20 codons. The latter KIT variants are more often observed in gastric GISTs, and are associated with a favorable outcome [4].
In this work we presented and discussed a newly discovered 3′ end KIT-JM 7-residues deletion mutant ∆574-5870 (i.e., delTQLPYDG), found in a 33-year-old male patient with a rare duodenal GIST that was difficult to resect at time of presentation. Neoadjuvant Imatinib at  Oncotarget 56164 www.impactjournals.com/oncotarget standard regimen (400 mg/day), reduced the tumor mass to an extent that conservative surgery could be practiced. Molecular analysis revealed that no other mutations were present either in the KIT exons 9, 13 and 17, or in the PDGFRA exons 12, 14 and 18.
In line with the in vivo observations, computerbased simulations predicted that the new ∆574-580 KIT variant was indeed Imatinib-responsive, with an inhibitor affinity comparable to that of the known activating and Imatinib-responding ∆559 KIT isoform. In fact, the calculated ∆G bind for this mutant is −8.58 ± 0.08 kcal/ mol, corresponding to K d or IC 50 values (obtained from the fundamental relationship ∆G bind = RTln K d = RT ln IC 50 ) of 516 nM, while the corresponding values of these quantities for the ∆559 KIT mutants are −9.15 ± 0.12 kcal/ mol and 198 nM, respectively (Table 1). At the same time, a validation test performed on the Imatinib refractory T670I KIT mutant yielded a ∆G bind value of -6.38 ± 0.11 kcal/mol, corresponding to a K d > 21 µM (Table 1).
Isothermal titration calorimetry (ITC) experiments ( Figure 4) performed on purified, unphosphorylated KIT constructs fully supported the in silico predictions, leading to ∆G bind,ITC and K d,ITC values −8.80 kcal/mol and 359 nM for the ∆574-580 KIT and −9.41 kcal/mol and 127 nM for the ∆559 KIT, respectively ( Table 2). The experimental values of ∆G bind,ITC and K d,ITC for the T670I KIT mutant (−6.18 kcal/mol and > 29 µM, Table 2) was also found in agreement with the corresponding computed values (Table 1), thereby ultimately validating the computational strategy and results adopted in the present study.
Cell-based assays next confirmed the constitutive activation of the KIT receptor bearing the ∆574-580 mutation; concomitantly, the capacity of Imatinib to inhibit this mutant KIT phosphorylation was also verified. Importantly, the effect of ∆574-580 KIT inhibition by Imatinib shared similar effects to those observed in cells transfected with the ∆559 KIT mutant, for which a strong phosphorylation of both AKT and ERK1/2 was detected ( Figure 5).
In summary, in the present study we reported and characterized a new 7-residue deletion mutation ∆574-580 in the rarely involved 3′end of tyrosine kinase receptor Oncotarget 56165 www.impactjournals.com/oncotarget KIT juxtamembrane region, found in a 33-year-old male patient diagnosed with localized duodenal GIST who successfully underwent neoadjuvant imatinib treatment at standard dose (400 mg/day). In line with the in vivo observation of imatinib optimal sensitivity as attested by early complete metabolic response at FDG-PET, the results achieved by a combination of experimental and computational techniques indicate that the new ∆574-580 KIT mutant does not influence the overall sensitivity of KIT mutant toward Imatinib. Molecular simulations reveal that the deletion of 7 amino acids in the juxtamembrane domain of the receptor does not perturb the overall protein structure as well the binding region in which Imatinib is encased (Figures 2 and 3). Furthermore, the predicted affinity values of this KIT variant for ATP (Table 1) are in line with those calculated for two renown activating KIT mutations, i.e., the ∆559 and T670I KIT receptor [15][16][17]19]. The molecular rationale yielded by the computational procedure is validated through ITC-based receptor/drug binding experiments and immunoblotting analysis of the effect of this new deletion on the intrinsic KIT activity and its response to Imatinib. Indeed, phosphorylation of the ∆574-580 KIT mutant is inhibited by Imatinib in a way similar to that observed for the Imatinib-sensitive ∆559 KIT control, whereas the kinase activity and thus receptor phosphorylation of the T670I, the prototypical Imatinib resistant KIT variant, is not inhibited at drug concentrations between 1 and 5 μM. The positive response to the kinase inhibitor of the ∆574-580 KIT form is further confirmed by the analysis of its signaling pathway in presence and in absence of Imatinib, according to which both AKT and ERK1/2 phosphorylation is indeed impaired at both Imatinib concentrations considered ( Figure 5).
In addition, results from this study showed that a clinical useful procedure, neoadjuvant treatment, can occasionally be of value for the understanding of the molecular pathogenesis of GIST.

KIT and PDGFRA molecular analysis
Mutation analysis of KIT exons 9, 11, 13, and 17, as well as PDGFRA exons 12, 14, and 18, was performed using direct sequencing of PCR products as described previously [21,22]. Tumor cells comprised 98% of the cells in the sample used for gene sequencing. Three independent PCR experiments were performed to confirm sequencing results.
Each protein/ligand system was solvated, gradually heated to 25°C, equilibrated and subjected molecular dynamics (MD) data collection runs to perform drug/ protein free energy of binding analysis [28][29].
All simulations were carried out using the Pmemd modules of Amber 16 [30], running on our Mose25 CPU/ GPU calculation cluster (see the Supporting Information appendix for full computational details).
Cloning an KIT mutants purification for experimental Imatinib binding studies ∆559, T670I, and ∆574-580 mutant KIT constructs were produced according to the established methodology described by Gajiwala et al. [31]. All mutant plasmids were sequenced to verify the success of mutagenesis experiments. The purified c-Kit protein was not phosphorylated as judged by routine mass spectrometry (data not shown). Full details are provided in the Supporting information appendix.

Isothermal titration calorimetry (ITC) experiments
Calorimetric titrations were carried out on a MicroCal PEAQ-ITC calorimeter (Malvern, UK) at 25°C. Before titrations, all KIT mutant proteins used were buffer-exchanged into an identical lot of HBS buffer (10 mM Hepes pH7.5, 150 mM NaCl). Gel filtration was used to control buffer heat dilution effects. Each protein sample was thoroughly degassed before ITC experiments, and these were run in triplicate (see Supporting information appendix for details).

Construction and transfection of mutated KIT
∆559, T670I and the new variant ∆574-580 mutant KIT receptors were obtained as following a methodology reported in our previous work [15][16][17]19]. Briefly, an expression vector carrying wild-type (WT) human complementary DNA (cDNA) for KIT (kind gift of Professor Y. Yarden, Weizmann Institute, Rehovot, Israel) was used to generate all mutated forms of KIT via sitedirected mutagenesis using the commercial QuickChange Site-Directed Mutagenesis kit (Promega, Madison, WI), following manufacturer's instruction. We then constructed mutant ∆559, in which amino acid 559 is removed from the juxtamembrane region by deleting nucleotides 1696 -1698 from the portion of the exon 11-derived WT