Bi-directional signaling by membrane-bound KitL induces proliferation and coordinates thymic endothelial cell and thymocyte expansion

The ligand for the c-Kit receptor, KitL, exists as a membrane-associated (mKitL) and a soluble form (sKitL). KitL functions outside c-Kit activation have not been identified. We show that co-culture of c-Kit– and mKitL–expressing NIH3T3 cells results in signaling through mKitL: c-Kit–bound mKitL recruits calcium-modulating cyclophilin ligand (CAML) to selectively activate Akt, leading to CREB phosphorylation, mTOR pathway activation, and increased cell proliferation. Activation of mKitL in thymic vascular endothelial cells (VECs) induces mKitL- and Akt-dependent proliferation, and genetic ablation of mKitL in thymic VECs blocks their c-Kit responsiveness and proliferation during neonatal thymic expansion. Therefore, mKitL–c-Kit form a bi-directional signaling complex that acts in the developing thymus to coordinate thymic VEC and early thymic progenitor (ETP) expansion by simultaneously promoting ETP survival and VEC proliferation. This mechanism may be relevant to both normal tissues and malignant tumors that depend on KitL–c-Kit signaling for their proliferation. c-Kit receptor–Kit ligand complex signaling is known to activate c-Kit and is essential for tissue development. Here, Buono et al. show that membrane-bound KitL signaling induces proliferation via CAML-Akt-CREB pathway activation, establishing a role for bidirectional signaling in tissue expansion.

T he c-Kit receptor and its ligand KitL form a signaling complex that plays important roles in hematopoiesis, fertility, pigmentation, digestion, and nervous system function 1 . Furthermore, activating mutation in c-Kit is observed in several malignancies, including acute myeloid leukemia, mastocytosis, gastrointestinal stromal tumors and melanoma, and c-Kit inhibitors are being developed for cancer therapy 2 .
KitL is the only known c-Kit ligand, and exsists in both a membrane-associated (mKitL) and soluble form (sKitL). Whereas sKitL is generated through juxtamembrane proteolytic cleavage, mKitL is generated by skipping of the exon that contains the proteolytic cleavage site 3 . Genetic experiments have established that mKitL and sKitL each carry out unique physiological roles: Genetic deletion of the sKitL proteolytic cleavage site resulted in loss of mast cells from the skin and peritoneum, and increased radiosensitivity 4 . In contrast, selective mKitL ablation demonstrated that mKitL expressed by thymic vascular endothelial cells (VECs) and cortical thymic epitelieal cells (cTECs) plays an important role in the survival of c-Kit-expressing early thymic progenitors (ETPs) 5 . Importantly, upon loss of mKitL from thymic stromal cells similar decreases in the number of thymocytes, thymic epithelial cells and VECs are observed 5 , indicating the presence of homeostatic mechanisms that preserve the proportionality of thymic cell types.
During development the induction of the mouse thymus occurs around embryonic day 11.5 (e11.5), followed by diversification of cortical (cTECs) and medullary thymic epithelial cells (mTECs), and vascularization around e13.5 6,7 . The vascularized thymus expands rapidly until postnatal day 12 (P12) when it reaches its adult size 8 . Several signaling molecules, including interleukin (IL-)7, Dll4, Ccl19, Ccl25, Cxcl12, BMP4, and Wnt4, have been identified as important for the expansion and differentiation of thymocytes, whereas TEC specification involves Shh, BMP4, Fgf, and Wnt signaling 9,10 . However, little is known about how thymic VECs are specified or how thymocyte and stromal cell expansion is coordinated.
Given that mKitL depletion eliminates both c-Kit signaling in thymocyte progenitors and mKitL in thymic VECs and TECs this raised the possibility that mKitL transduces a signal upon mKitL-c-Kit interaction that promotes the expansion of mKitLexpressing cells. We therefore tested whether engagement of mKitL by c-Kit elicits signaling in mKitL-expressing cells. We find that stimulation of mKitL by cell-associated or soluble c-Kit activates the Akt/mTOR/CREB pathway and increases cell proliferation. Finally, loss of mKitL in thymic VECs decreases their perinatal proliferation. Therefore, c-Kit and mKitL constitute a bi-directional signaling complex that can coordinate cell proliferation and survival in the developing thymus.
Results c-Kit signals through mKitL. To test the hypothesis that mKitL has signaling capacity we expressed c-Kit in NIH3T3 cells by lentiviral transduction (Fig. 1a), generating NIH-Kit cells. Upon co-culture of NIH-Kit cells with wild-type NIH3T3 cells, where mKitL is endogenously present (Fig. 1b), we observed a strong upregulation of the Ki67 proliferation marker in the wild-type NIH3T3 cells, not observed upon co-culture with NIH3T3 cells transduced with the control Venus expression vector (NIH-Venus) (Fig. 1c-e; Supplementary Fig 1). Inhibition of c-Kit signaling with Imatinib did not decrease proliferation of NIH3T3 cell in NIH-Kit co-cultures, indicating that c-Kit activation in NIH-Kit cells did not indirectly contribute to NIH3T3 proliferation ( Supplementary Fig 2a-c). This was supported by the ability of NIH3T3 cells expressing kinase-dead c-Kit (NIH-KitK623M cells) 11 to induce proliferation similarly to NIH-Kit cells ( Supplementary Fig 2d-f). Incubation of NIH3T3 cells with a recombinant fusion between the c-Kit extracellular domain (ECD) and the immunoglobulin G (IgG) constant domain (Kit-Fc) also promoted cell cycle progression, as measured by Ki67 expression (Fig. 1f-h) and bromodeoxyuridine (BrdU) incorporation (Fig. 1i). These results demonstrated that exposure of mKitL-expressing NIH3T3 cells to the c-Kit ECD, either presented on a neighboring cell surface or in solution, is sufficient to elicit a proliferative response. We screened major signaling pathways for activation downstream of mKitL, and observed that incubation with Kit-Fc induced transient serine 133 (S133) phosphorylation of CREB (Fig. 2a) and serine 235/236 (S235/ S236) phosphorylation of Rps6 (Fig. 2b) in NIH3T3 cells, whereas activating phosphorylation of Erk1/2 or p38α was not observed (Fig. 2c, d). CREB S133 phosphorylation was accompanied by nuclear accumulation of CREB phospho-S133 (Fig. 2e). A blocking antibody against mKitL abolished Kit-Fc-induced CREB S133 phosphorylation (Fig. 2e, f) and nuclear translocation (Fig. 2e) induced by Kit-Fc, validating that the c-Kit ECD induces CREB S133 phosphorylation through direct interaction with mKitL.
mKitL signaling activates the Akt pathway. Next, to globally identify signaling pathways activated downstream of KitL we performed phospho-proteome analysis of Kit-Fc and IgG-Fc treated NIH3T3 cells after 10 min of incubation, as this time point showed optimal induction of CREB and Rps6 phosphorylation after Kit-Fc incubation (Fig. 2a). A total of 427 proteins showed significantly altered phosphorylation (P < 0.05, t-test) in Kit-Fc-treated cells (Supplementary Data 1). Pathway analysis of the mKitL-regulated phospho-proteome showed extensive perturbation of mTOR signaling, eIF2 signaling and eIF4/p70S6kinase signaling (Supplementary Data 2), indicating activation of the Akt/mTOR pathway 12 , and consistent with the observed induction of Rps6 S235/S236 phosphorylation after mKitL activation (Fig. 2b). Examination of specific phosphorylation states in NIH3T3 cells using intracellular flow cytometry confirmed increased Akt S473, mTOR S2448, and Rps6 S235/S236 phosphorylation ( Fig. 3a-c), as well as increased S133 phosphorylation of CREB (Fig. 3d), after NIH-Kit co-culture. To determine the role of Akt and CREB in the proliferative response to mKitL activation NIH3T3 cells were co-cultured with NIH-Kit or NIH-Venus cells in the presence or absence of inhibitors of CREB or Akt. The previously observed proliferative response, as measured by Ki67 upregulation, was blocked by inhibition of either Akt or CREB (Fig. 3e, f). We therefore conclude that binding of the c-Kit ECD to mKitL leads to the activation of Akt and its downstream effectors mTOR and CREB in mKitL expressing cells, promoting their proliferation.
CAML interacts with the mKitL ICD. To further address the mechanism by which KitL transduces signaling we used the KitL intracellular domain (ICD) as bait in a yeast 2-hybrid screen, and identified several putative KitL-ICD interacting proteins (Supplementary Table 1). Of these calcium-modulating cyclophilin ligand (CAML) was phosphorylated upon mKitL activation (Supplementary Data 1), and CAML was previously shown to participate in TCR signaling 13 . Knockdown of CAML in NIH3T3 cells ( Supplementary Fig 3) did not affect mKitL expression (Fig. 4a) or cell surface presentation (Fig. 4b). However, upon CAML knockdown Akt S473 and CREB S133 phosphorylation was decreased, whereas the Erk activation state was unaffected (Fig. 4c). Co-immunoprecipitation of CAML and mKitL confirmed their interaction in NIH3T3 cells (Fig. 4d, Supplementary  Fig 4). Consistent with CAML-dependent activation of Akt and CREB mediating the proliferative response to mKitL activation, we observed that CREB activation after Kit-Fc stimulation was abolished by CAML knockdown (Fig. 4e), as was the proliferative response to mKitL activation as measured by Ki67 expression (Fig. 4f) and cell proliferation (Fig. 4g). CAML is therefore required for induction of signal transduction and cell proliferation upon activation of mKitL signaling. mKitL signaling promotes thymic VEC proliferation. In the thymus immigrating c-Kit+ ETPs initially remain in contact with cortical VECs (Supplementary Fig 5a), and VEC-associated mKitL supports their survival 5 . The frequency of c-Kit+ thymocytes is higher in perinatal, expanding thymus compared to fully expanded 4-week-old thymi (Fig. 5a-d; Supplementary Fig 5b;  Supplementary Fig 6a), as is the proliferation of thymic VECs ( Fig. 5e, f; Supplementary Fig 6a). We therefore tested if c-Kit expressed on thymocytes contribute to VEC proliferation during thymus expansion by analyzing mice, in which mKitL expression had been genetically ablated specifically in VEC. This was done by crossing mice with a loxP-flanked Kitl exon 7 (Kitl LEx7 allele), which encodes the trans-membrane domain of mKitL, to the VEC-specific Tie2-Cre driver, generating Tie2ΔEx7 mice (Kitl-LEx7/LEx7 ; Tie2-Cre tg/+ genotype) 5 . Analysis of perinatal thymi from Tie2ΔEx7 and control mice (Kitl LEx7/LEx7 genotype; LEx7 mice) showed a significant reduction of Ki67-expressing, proliferating VEC upon VEC-specific mKitL depletion (Fig. 6a, b; Supplementary Fig 6a). To determine whether this effect was c-Kit-mediated, thymic VECs from LEx7 and Tie2ΔEx7 mice were co-cultured with NIH-Venus or NIH-Kit cells. Proliferation of control VECs was higher in NIH-Kit, compared to NIH-Venus, co-culture, and this effect was absent in the case of Tie2ΔEx7  Finally, as observed for NIH3T3 cells, Akt inhibition led to a significant decrease in c-Kit-induced thymic VEC proliferation (Fig. 6e). We were not able to detect c-Kit expression on thymic VECs in vivo ( Supplementary Fig 5a,b), nor did we observe any effect of Imatinib on VEC proliferation in NIH-Kit co-culture ( Supplementary Fig 5c-e), indicating that also for VECs c-Kit activation is not required for the proliferative response elicited through mKitL.

Discussion
Evidence that niche cells can depend on the stem-or progenitor cells they support has been obtained from the Drosophila ovary, where Cap niche cells maintain germline stem cells (GSCs)       15 . We here show that thymic VEC proliferation is regulated by c-Kitmediated activation of mKitL signaling, and that this pathway is rate limiting for VEC proliferation during postnatal thymus expansion. In conjunction with the finding that mKitL expression on cortical thymic VECs is required for the survival of c-Kitexpressing thymocyte progenitors 5 , this indicates mutual dependence of thymic progenitors and their niche cells. Importantly, crosstalk between mKitL+ thymic VECs and c-Kit+ thymocytes occurs simultaneously via a single molecular complex and requires cell-cell contact, enabling strict spatial control and selectivity of signaling. Bi-directional signaling has previously been observed for Ephephrin complexes, with their most important roles being in patterning of the vasculature and central nervous system 16,17 . Ephrin expression in neurons has been shown to inhibit proliferation of the neuronal stem and progenitor cells that generate them 18 , and reciprocal regulation of EphB2/B3 and EphrinB1/B2 by Wnt signaling in the intestine restricts progenitor cell proliferation to the crypt 19 , both examples of negative feedback signaling that contributes to tissue homeostasis. In contrast, mKitL-c-Kit bi-directional signaling supports the simultaneous expansion of interacting mKitL and c-Kitexpressing cells, here exemplified by c-Kit+ thymocyte progenitors and their supporting mKitL+ vascular endothelial stromal cells. Such a bi-directional signal would be suitable to drive coordinated cellular expansion in developing and regenerating tissues. Relevant to this, local KitL signaling has been proposed to sustain hematopoietic stem cells (HSCs) not only under steady state conditions 20 , but also during the expansion of facultative HSCs niches in the spleen during extramedullary hematopoiesis 21 . Furthermore, reprogramming of stromal cells by cancer cells to enhance their ability to support tumor progression has been observed in both hematopoietic 22 and solid cancers 23,24 , and c-Kit is widely expressed in malignant cancers 1 , and supports the maintenance of cancer stem cells in several tumor types 25 , including lung 26 and ovarian cancers 27 . The molecular mechanisms that mediate stromal adaptation include PTEN downregulation 23 and Hmga2 overexpression 28 , both of which activate Akt signaling [29][30][31] , and upregulation of HSF1 24 , a transcription factor activated by mTOR-mediated phosphorylation 32 . Bidirectional mKitL-c-Kit signaling may therefore contribute to the interaction between cancer cells (including cancer stem cells) and stromal cells that generates reactive stroma through Akt/ mTOR pathway activation. Combined with the importance of c-Kit-KitL interaction for the development and maintenance of multiple tissues 1 bi-directional c-Kit-mKitL signaling may therefore be important for both normal and malignant cell-cell interactions.

Methods
Cell lines. NIH3T3 cells (ATCC CRL-6442) were cultured in IMDM with heat inactivated foetal calf serum and penicillin/streptomycin (Gibco) at 37°C and 5% CO 2 . Standard culture was in 10% FCS with serum percentage adjusted is indicated according to the type of experiment performed.
For CAML knockdown 5 × 10 5 NIH3T3 were seeded in 6-well plates and infected overnight with hairpin-expressing lentivirus. Medium was replaced and cells kept or 2 additional days before cell passage and puromycin selection (1 μg/ ml). CAML expression was analysed by western blot and cells with optimal CAML knockdown (NIH-CAML) and control infected cells (NIH-Con) expanded under puromycin selection. Information regarding the lentiviral vector and CAML shRNA sequences can be found in Supplementary Table 3. NIH-Venus and NIH-Kit were generated by infection of NIH3T3 cells with empty pRRL-Venus and pRRL-Venus-c-Kit, respectively, the latter expressing fulllength murine c-Kit. Venus expressing cells were isolated by FACS. Cultures were regularly monitored for GFP expression. For phopho-flow cytometry were then collected following trypsinization and resuspended in 50 μl of cytofix/cytoperm (BD) and left for 30 min in ice before washing in BD wash/perm buffer. The antibodies (anti-Ki67-PE; anti-phospho-Akt-APC; anti-phospho-mTOR-eF450; anti-phospho-Rps6-PECy7; anti-phospho-CREB and goat anti-rabbit AlexaFluor488) were incubated overnight at 4°C. The cells were then washed in BD wash/perm buffer and FACS analysed.
For Ki67 and BrdU readout NIH3T3 cells were grown in the presence of 2% FBS for 24 h and the medium changed before applying 200 ng/ml IgG-Fc or Kit-Fc for 8 or 24 h prior to analysis, as indicated. For BrdU incorporation a BrdU Flow kit (Becton Dickinson) was used according the manufacturer's instructions. BrdU was added 2 h prior to analysis. Cells were then collected by trypsinization and resuspended in 50 μl of cytofix/cytoperm (Becton Dickinson) and left for 30 min in ice before washing in BD wash/perm buffer. Cells were then centrifuged at 500×g for 5 min and the pellet re-suspended in DNaseI solution for 1 h at 37°C. After For ELISA assays protein lysates were prepared according to the manufacturer's instructions (abcam), and CREB phospho-S133 (ab176659), Erk1 phospho-T202/ Y204 + Erk2 phospho-T185/pY187 (ab126445), and p38α phospho-T180/pY182 (ab126453) measured along with the total of the same proteins using the indicated ELISA kits. The readout was carried out by using a CLARIOstar® ELISA reader (BMG LABTECH).
For antibody neutralization NIH3T3 cells were grown in the presence of 0.5% FBS for 24 h. At this time medium was changed and goat-IgG (2 μg) or goat-antimouse-SCF (2 μg) was added to the medium for 1 h before stimulating with IgG-Fc or Kit-Fc for 10 min. The cells were then washed in cold PBS and protein lysates prepared.
Cell proliferation assay. For measurement of cell proliferation 1 × 10 4 NIH-Con or NIH-CAML cells were plated in 10 cm dishes and cultured in IMDM/2% heat inactivated serum/ PS for 7 days. IgG-Fc or Kit-Fc stimulation was applied every other day (starting from day 0) and cells counted at days 3 and 7.
Immunofluorescence. NIH3T3 cells were plated onto Millicell EZ 8-well Glass slides and cultured for 24 h in 0.5% FBS. Medium was changed and the cells incubated for 2 additional hours. IgG-Fc or Kit-Fc was added for 10 min before washing the cells and fixing in acetone for 5 min. The slides were then air dried and re-hydrated for 30 min in PBS at room temperature and blocked in 5% BSA in PBS for 10 min at room temperature. Anti CREB phospho-Ser133 primary antibody was diluted in PBS and slides stained for 2 h and then washed several times in PBS before adding the secondary antibody for 45′ (goat anti-rabbit AlexaFluor488) at room temperature. Slides were then counterstained with DAPI (0.2 μg/ml) and mounted with DXP mounting medium (Fisher Scientific). Images were acquired with a Zeiss 780 confocal inverted microscope and processed using Fiji software.
Thymic section preparation and imaging. Freshly isolated thymi were washed in PBS and included in OCT for snap freezing. The OCT blocks were sliced at 7 μm thickness and sections stored at −80°C. The sections were dried at room temperature for 5 min before fixing in acetone for 5 min at 4°C and left at room temperature for 5 min to ensure acetone evaporation. The sections were then rehydrated for 30 min in PBS at room temperature and blocked in 5% BSA in PBS for 10 min at room temperature. Fluorochrome-conjugated or unconjugated primary antibodies were diluted in PBS and sections stained for 1 h and then washed several times in PBS before adding the secondary antibody for 45 min at room temperature. Sections were then counterstained with DAPI (0.2 μg ml −1 ) and mounted with DXP mounting medium (Fisher Scientific). Images were acquired with a Zeiss 780 confocal inverted microscope and processed using Fiji software.
Protein lysates preparation and western blot. NIH3T3 were lysed in RIPA buffer (150 mM sodium chloride, 1.0% NP-40 or Triton X-100, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 50 mM Tris-HCl, pH 8.0) supplied with cOmplete Protease Inhibitor Cocktail (Roche) and PhosSTOP -Phosphatase Inhibitor Cocktail Tablets (Roche). Cells were kept for 20 min in ice and an aliquot set aside to determine the protein concentration with Qubit protein reagent. 6X protein loading buffer (30% glycerol, 0.5 M Tris-HCl, pH 8.0, 6 mM EDTA, 10% SDS, 10% β-mercaptoethanol, 0.01% bromophenol blue) was then added and samples sonicated for 5 min in an ultrasonic bath (Qsonica; Amplitude 50; 30 s ON, 30 s OFF) before boiling for 5 min at 100°C. Twenty micrograms of total protein lysate was loaded onto NuPAGE 4-12% Bis-Tris Protein Gels (Thermo Fisher). Protein transfer onto PVDF membrane was performed using Trans-Blot SD Semi-Dry Transfer Cell. Membranes were washed in TBS/0.001% Tween-20. Primary antibodies were diluted in 5% milk in TBS/0.001% Tween-20 and incubated overnight at 4°C on a rocking platform. Secondary antibodies were diluted in 5% milk in TBS/0.001% Tween-20 and incubated for 1 h at RT on a rocking platform. Signal was developed by using ECL western blot substrate (Pierce) and acquired onto CL-XPosure™ Film (Thermo Scientific). Co-immunoprecipitation. NIH3T3 cells from a 10 cm dish were lysed in RIPA buffer supplied with cOmplete Protease Inhibitor Cocktail by gently scraping, and lysates clarified by centrifugation at 13,000 g for 10 min at 4°C. The supernatant was collected and an aliquot stored as input. The rest of the material was incubated with polyclonal anti-CAML antibody (Santa Cruz Biotechnology, San Diego, CA; Supplementary Table 2) for 3 h at 4°C on a rotating wheel. Forty microliters of Dynabeads protein-G, washed twice in 1 ml of RIPA buffer with protease inhibitors on the rotating wheel at 4°C for 10 min, was added and incubation continued for additional 45 min, followed by washing. The protein-antibody-beads complexes were then isolated on a magnetic rack and re-suspended in RIPA buffer with protease inhibitors and protein loading buffer. The eluted material was used to load two replica gels that were probed with anti-CAML and the anti-mKitl antibody, respectively. Primary antibody was detected with protein A-HRP (CST).
Co-culture of primary thymic cells. Thymic stromal cells were prepared from 4 to 6-week-old LEx7 and Tie2ΔEx7 mice. Dissected thymi were finely chopped and the obtained fragments digested in a mix of Collagenases I and 3 (Worthington; 3 mg ml −1 ), dispase II (Roche; 7 mg ml −1 ), DNaseI (Invitrogen; 2 U ml −1 ) at 37°C with gentle agitation for 20 min. The dissociated cells were centrifuged at 300×g for 5 min and re-suspended in culture medium. The total thymic cell population was divided in two aliquots and seeded onto 2 × 10 6 NIH-Venus and NIH-Kit cells, respectively, in 10 cm tissue culture dishes. Cells were co-cultured for 24 h before collecting for cell surface staining followed by intracellular Ki67 staining. For drug treatment total thymic cells were seeded in 3 aliquots onto 1 × 10 6 NIH-Venus or NIH-Kit cells in 6 cm tissue culture dishes. Drug or vehicle (DMSO) was added for additional 6 h. Cells were cultured in IMDM with 10% FBS throughout. Mice were maintained in accordance with protocols ethically approved by the Oxford University Clinical Medicine Animal Welfare and Ethics Review Board and the UK Home Office.
Flow cytometry. Cell lines were collected by scraping or light trypsinization and re-suspended in PBS, 5% FBS. The antibodies used are listed in Supplementary  Table 2. Cells were stained on ice for 20 min in the dark. Data were acquired using LSR Fortessa X-20 (BD Biosciences) and analysed using FlowJo software (TreeStar). Thymocyte and thymic VECs were quantified by flow cytometry using the antibodies described in Supplementary Table 4.
Phospho-proteomic analysis. Cells were stimulated with Kit-Fc or IgG-Fc in 3 up-scaled (3×) replicates each as described above. The obtained lysates were reduced with 200 mM TCEP for 1 h at 55°C and alkylated with 375 mM iodoacetamide for 30 min at room temperature in the dark. Proteins were precipitated with chloroform and methanol 33 twice to remove detergents, before re-suspension in 50 mM TEAB and digest with 30 µg of trypsin per sample overnight at 37°C. Samples were dried and re-suspended in 200 µl 50 mM TEAB for TMT labeling. Peptide concentrations were quantified using the Pierce Quantitative Colorimetric peptide assay following manufacturer instructions (Thermo Fisher). Equal amounts of each sample (465 μg) were labeled with TMT6 reagents (Thermo Fisher) for 1 h according to the manufacturer's instructions before the labeling reaction was quenched with 5% hydroxylamine. The six samples were pooled and desalted using a C18 Sep-Pak cartridges (Waters): The cartridge was conditioned with 5 ml buffer B (0.1% TFA in 65% CH 3 CN), equilibrated with 10 ml of buffer A (0.1% TFA in 2% CH 3 CN). The tryptic peptides were added to bind to the cartridge and then washed with 10 ml buffer A. The peptides were eluted in 1.2 ml of buffer B and dried in a vacuum concentrator. Peptides were solubilised in 120 µl of 80% CH 3 CN/0.1% TFA solution. Phospho-peptide enrichment was performed based on a refinement 34 of the 2step enrichment procedure published by McNulty et al. 35 . Fractionation was performed on a Dionex HPLC using a TSKgel Amide-80 column (5 µm, 4.6 mm ID × 25 cm, TOSOH) for hydrophilic interaction. The peptides were eluted at a 500 µl/minute flow rate using the following gradient: 10 min with 80% buffer B (98% CH 3 CN with 0.1% TFA), then from 10 to 55 min 80 to 60% of buffer B. The column was then washed for 10 min with 98% buffer A (98% water with 0.1% TFA) and the column was re-equilibrated for another 10 min with 80% buffer B. The collection was performed from 10 to 70 min each fraction was collected for 4 min. A total of 15 fractions were collected. Neighboring fractions were combined into 6 pools and 100 µl of PHOS-Select™ Iron affinity beads (Sigma) were added to each pool. The samples were incubated 1 h at 4°C and filtered at 8200 g for 1 min over a 0.22 μm nylon Spin-X cartridge (Sigma) to remove un-phosphorylated peptides. The beads were successively washed with 500 µl of 250 mM of acetic acid with 30% CH 3 CN and 500 µl of water and centrifuged for 2 min at 2000 rpm after each wash. Phospho-peptides were eluted by incubation of the beads in 500 µl of 1% ammonium hydroxide for 10 min, followed by a centrifugation at 8200×g for 2 min. Samples were dried down and re-suspended in 10 µl of 1% CH 3 CN 0.1% TFA prior to mass spectrometry analysis.
LC-MS/MS analysis. Samples were analysed in an Orbitrap fusion Lumos coupled to a UPLC ultimate 3000 RSLCnano System. The samples were loaded in 1% CH 3 CN 0.1% TFA buffer and eluted using a gradient of 2-35% CH 3 CN in 0.1% formic acid and 5% DMSO in 120 min at a flow rate of 250 nl/min. The survey scan was acquired within a 380-1500 m/z window at a resolution of 120,000 and an AGC target of 2E5. Selected precursor ions were isolated in the quadrupole with a mass isolation window of 1.6Th and analysed in the Ion trap following CID fragmentation and detection in rapid mode. Peptide quantitation was based on synchronous precursor selection for the 10 most abundant MS2 fragments for a MS3 scan at a resolution of 60,000 in the Orbitrap detector with a normalized collision energy of 65% (HCD) and a scan range of 120-500 m/z. The total cycle time was fixed at 4 s with a dynamic precursor exclusion of 7 s.
Proteomics data analysis. Raw files were imported into Proteome Discoverer (2.1) and analysed using the standard workflow for TMT-6plex and synchronous precursor selection: MS2 spectra were searched with SequestHT using 10 ppm as precursor and 0.5 Da as fragment error tolerance against the Uniprot/ Swissprot database of Mus musculus (retrieved 17.07.2015). Variable modifications were methionine oxidation, phosphorylation on serine, threonine or tyrosine and N-terminal protein acetylation. Static modifications were cabamidomethylation on cysteine and TMT-6plex modification of the peptide Nterminus and lysine. Identifications were re-processed with Percolator to match a FDR of 1%. Quantitative data were analysed and visualized with Perseus 36 . Canonical Pathway Analysis was performed using Ingenuity Pathway Analysis (Qiagen).
Yeast 2-hybrid screening. The murine KitL intracellular domain was cloned into GBT9 and used to screen a normalized human universal cDNA library (Clontech) 37 . Positive preys were sequenced and those known to be promiscuous based on previous screens, as well as those encoded only by 3′-UTR sequence, discarded. Hits are shown in Supplementary Table 1.
Statistical analysis. Unpaired, two-tailed Student's t-test was used throughout for data analysis

Data availability
The mass spectrometry proteomics data have been deposited to the Proteo-meXchange Consortium via the PRIDE partner repository [http://proteomecentral. proteomexchange.org] with the dataset identifier [PXD011061]. All other primary data are available from the authors upon request.