Phosphoproteomics-Mediated Identification of Fer Kinase as a Target of Mutant Shp2 in Noonan and LEOPARD Syndrome

Noonan syndrome (NS) and LEOPARD syndrome (LS) cause congenital afflictions such as short stature, hypertelorism and heart defects. More than 50% of NS and almost all of LS cases are caused by activating and inactivating mutations of the phosphatase Shp2, respectively. How these biochemically opposing mutations lead to similar clinical outcomes is not clear. Using zebrafish models of NS and LS and mass spectrometry-based phosphotyrosine proteomics, we identified a down-regulated peptide of Fer kinase in both NS and LS. Further investigation showed a role for Fer during development, where morpholino-based knockdown caused craniofacial defects, heart edema and short stature. During gastrulation, loss of Fer caused convergence and extension defects without affecting cell fate. Moreover, Fer knockdown cooperated with NS and LS, but not wild type Shp2 to induce developmental defects, suggesting a role for Fer in the pathogenesis of both NS and LS.


Introduction
Noonan syndrome (NS) (OMIM 163950) is a congenital disorder that manifests itself in heart defects, short stature, webbed neck, hypertelorism and an increase in the occurrence of juvenile myelomonocytic leukemia (JMML) and other malignancies. The most common causes for NS are mutations in PTPN11 encoding for Src-homology domain 2 (SH2) containing phosphatase 2 (Shp2) [1] A similar syndrome is also caused by mutations in PTPN11 and patients display similar symptoms as NS. An acronym of the symptoms, Lentigines, Electrocardiographic conduction anomalies, Ocular hypertelorism, Pulmonary stenosis, Abnormal genitalia, Retarded growth and Deafness gave this syndrome its name, LEOPARD syndrome (LS)(OMIM 151100) [1]. Both NS and LS are part of a group of congenital syndromes caused by mutations in the RAS mitogen activated protein kinase (MAPK) pathway called RASopathies. Despite the similarities in the clinical manifestations of NS and LS, NS mutations lead to an 'active' form of the Shp2 phosphatase, while LS is thought to result from 'inactivating' mutations [2,3]. However, a gain-of-function for LS has also been described in Drosophila [4].
The phosphatase Shp2 consists of two N-terminal SH2 domains that are able to bind to tyrosine phosphorylated targets of Shp2, a protein tyrosine phosphatase (PTP) domain, and a C-terminal tail with a proline rich domain and two tyrosines that are able to bind Grb2 when phosphorylated. Under basal conditions, the N-SH2 domain blocks access of the PTP domain to its substrates, resulting in an inactive conformation of Shp2 [5]. In NS, mutations cause a disruption of the interaction between the N-SH2 domain and the PTP domain, resulting in a hyperactive form of Shp2 [1,2]. In LS however, mutations are mostly present in the PTP domain at the interface of the PTP and SH2 domain, resulting in a loss of catalytic activity. How both activating and inactivating mutations of Shp2 lead to similar developmental defects is largely unknown.
Homozygous mice defective for Shp2 die in utero and have gastrulation defects, showing malformations of the notochord and posterior truncations [6]. Embryos lacking Shp2 also develop failure of neural tube closure resulting in spina bifida and secondary neural tubes resulting from gastrulation defects [7]. Shp2 is essential for limb formation since chimeric mice with defective Shp2 expressed in the mesenchyme of the progress zone showed limb bud defects [8]. More recent results have shown cell migration defects during gastrulation in NS and LS Shp2 expressing zebrafish embryos as well [9]. These convergence and extension (C&E) cell movements mediate the anterior-posterior lengthening and lateral narrowing of the developing embryo. C&E cell movements are under strict spatiotemporal control of various signaling pathways [10]. Being a protein proximal to many RTKs, Shp2 acts upstream of multiple signaling pathways [2].
As NS and LS Shp2 are thought to act biochemically opposite, yet give rise to similar clinical symptoms, we sought to identify potential common targets of NS and LS Shp2 in a zebrafish model. Pinpointing disease associated Shp2 signaling shared by NS and LS, may contribute to the understanding of the underlying mechanisms of NS and LS pathogenesis and the development of therapeutic strategies for both NS and LS. Using a comparative phosphoproteomics approach, phosphotyrosine (pTyr)-containing peptides of NS-and LS-Shp2 expressing zebrafish embryos were isolated, identified and compared to peptides of WT embryos. A peptide corresponding to the autophosphorylation site of Fer kinase was identified as one of the most down-regulated peptides in developing NS and LS zebrafish. Further investigation revealed a role for Fer in C&E cell movements during gastrulation. Downregulation of Fer cooperated with NS and LS to induce developmental defects, suggesting a genetic interaction between Fer and the NS and LS variants of Shp2.

Ethics statement
Only embryos up to 4 days post fertilization (dpf) were used for these experiments, which do not require approval of the animal experiments committee according to national and European law.

Zebrafish maintenance and in situ hybridization
Wild type and Tg(24.9sox10:EGFP) ba2 zebrafish were kept under standard conditions. Embryos were staged as described by Westerfield [11]. In situ hybridizations were done as described before [9], using probes specific for bmp2b, chd, cyc, gsc, krox20, myod, ntl, pax2 and six3. For the fer probe, Fer-pBSK-was digested with NotI for the antisense probe and with ApaI for the sense probe, respectively. T7 and T3 RNA polymerase were used to generate digoxigenin labeled RNA (Roche). Embryos were fixed at the indicated stages with 4% paraformaldehyde (PFA). Embryos were treated from 24 hpf onwards with 0.2 mM 1-phenyl 2-thiourea (Sigma) in E3-medium to block pigmentation.

Morpholinos, RNAs and injections
Fer splicing morpholinos (MOs) were designed for several intron-exon boundaries and ordered from GeneTools (Philomath, OR, USA). The p53 MO (59-GCGCCATTGCTTTGCAA-GAATTG-39) was described before and was used as recommended by the manufacturer [12]. Synthetic mRNA was synthesized using mMessage mMachine kit (Ambion). pCS2+ plasmid was digested with NotI and RNA was synthesized with SP6 RNA polymerase. The NS and LS constructs were previously described [9]. MOs and mRNA was injected at one cell stage and the working concentration was titrated for each MO and mRNA. Fer MO i5e6 (59-GGCTTCTGGTGATCTGTTGAAATA-39) was used at 1.0 ng/ml (high concentration) or 0.5 ng/ml (low concentration). Fer MO e9i9 (59-GGGATTACACTTACTGAC-GAAGAGC-39) was used at 2.5 ng/ml (high concentration) or 1.0 ng/ml (low concentration). For co-injection, suboptimal concentrations of both MOs were combined.

RNA isolation
Total RNA of 3 dpf Nacre MO and Fer MO injected embryos was isolated. Approximately 30 embryos were dissolved in 1.0 ml TriZol (Invitrogen). The homogenate was centrifuged (12.000 g, 4 degrees, 10 min). 0.2 ml Chloroform was added to the supernatant, vortexed for 15 sec, incubated for 3 min at room temperature (RT) and afterwards centrifuged (12.000 g, 4 degrees, 15 min). 0.5 ml isopropanol was added to the upper phase, incubated for 10 min at RT and centrifuged (12.000 g, 4 degrees, 15 min). The pellet was washed with 75% ethanol and after vortexing centrifuged (10.000 g, 4 degrees, 5 min). The pellet was dried for 5 min and dissolved in 100 ml deionized water. Subsequently, the solution was incubated for 10 min at 37uC, precipitated in 250 ml 100% ethanol and 10 ml 3 M sodium acetate and stored at 280uC.

Mass spectrometry
Digest Preparation. Embryos were injected at the 1-cell stage with 150 pg wild type Shp2, 150 pg Shp2-D61G or 50 pg Shp2-A462T RNA, respectively. Shp2 RNA injected embryos were co-injected with enhanced green fluorescent protein (GFP) RNA used for screening of injection efficiency. Embryos expressing GFP were selected, manually dechorionated and collected at 26 hpf to 28 hpf. Embryos were then deyolked and washed using deyolking buffer (1/2 Ginzburg Fish Ringer) without calcium (2610 ml/embryo) snap frozen and stored at 280uC for further usage [11,13]. Embryo pellets corresponding to a total of approximately 2000 embryos per condition were resuspended and lysed in 8 M Urea, 50 mM ammoniumbicarbonate, 1 PhosSTOP tablet per 10 ml plus 1 mM vanadate and ethylenediaminetetraacetic acid-free protease inhibitor cocktail (Sigma). Embryo lysates were sonicated for 4630 seconds and then centrifuged at 14 500 rpm for 30 minutes at 4uC.
Lysate supernatants equivalent to 3 to 4 mg total protein per condition were generated in principle as described previously [14,15]. Briefly, the lysates were reduced with dithiothreitol (10 mM) for 1 hour at 56uC and alkylated with iodoacetamide (55 mM) at room temperature, in the dark for 45 minutes and digested for four hours with the protease Lys-C (1:100 enzyme/ substrate) at 37uC. Samples were then diluted 4 fold with 50 mM Ammoniumbicarbonate to 2 M Urea and trypsin (1:100) was added and digestion was performed overnight at 37uC. The peptide mixtures were acidified to pH 3 by adding acetic acid to a final concentration of 2.5%. The individual peptide solutions were then desalted, dimethyl labelled on-column as described previously [16]. Wild type Shp2 injected embryos were labeled ''light'' whereas D61G (Noonan) Shp2 and A462T (LEOPARD) Shp2 injected embryos were labeled ''intermediate'' and ''heavy'', respectively.
Immunoprecipitation (IP). IP was performed as described [17]. In principle, labeled peptide solutions were mixed in equal concentrations, vacuum dried and resuspended in IP Buffer (50 mM Tris, pH 7.4, 150 mM sodium cloride, 1% n-octyl-b-Dglucopyranoside and a Complete Mini protease inhibitor tablet per 10 ml IP buffer (Roche Diagnostics)). PY99-agarose beads (Santa Cruz Biotechnology) were washed five times with IP buffer prior to IP. The labeled peptide mixture was mixed with the PY99-agarose beads, and incubation was performed overnight at 4uC under constant rotation. After peptide immunoprecipitation, the beads were washed three times with IP buffer and two times with ultrapure water. Peptides were eluted by adding two times 50 ml of 0.15% trifluoroacetic acid for 20 min at room temperature. Eluted peptides were then desalted and concentrated on C18 tips and resuspended in 10% formic acid prior to MS analysis.
On-line Nanoflow Liquid Chromatography. As described in [17]: Nanoflow LC-MS/MS was performed on an Linear Trap Quadrupole-Orbitrap XL mass spectrometer (Thermo Electron, Bremen, Germany) coupled to an Agilent 1100 HPLC system (Agilent Technologies, Waldbronn, Germany). Dried peptides were trapped at 5 ml/min in 100% solvent A (0.1 M acetic acid in water). Subsequently, peptides were transferred to an analytical column (ReproSil-Pur C 18 -AQ, 3 mm (Dr. Maisch GmbH, Ammerbuch, Germany); 40 cm * 50 mm inner diameter, packed in house) at ,100 nl/min and eluted using a 3-h gradient from 0 to 40% solvent B (0.1 M acetic acid in 8:2 (v/v) acetonitrile/ water). The eluent was sprayed via distal coated emitter tips (New Objective) butt-connected to the analytical column. The LTQ Orbitrap XL was operated in data-dependent mode, automatically switching between MS and MS/MS. Full-scan MS spectra (from m/z 350 to 1500) were acquired in the Orbitrap with a resolution of 60,000 at m/z 400 after accumulation to target value of 500,000. The three most intense ions at a threshold above 500 were selected for collision-induced fragmentation in the linear ion trap at a normalized collision energy of 35% after accumulation to a target value of 10,000.
Data Analysis. All MS 2 spectra were processed and quantified with Proteome Discoverer (version 1.3.0.399). Runs were searched against a concatenated forward-decoy version of the Zebrafish Uniprot database (version 20130916, 81920 sequences) with Mascot (version 2.3.02, Matrix Science). The database search was performed with the following parameters: a mass tolerance of 650 ppm for precursor masses and 60.6 Da for CID fragment ions, allowing two missed cleavages, cysteine carbamidomethyla-tion as fixed modification. Light, intermediate and heavy dimethylation of peptide N-termini and lysine residues; methionine oxidation; phosphorylation on serine, threonine and tyrosine were set as variable modifications. The enzyme was specified as trypsin. The phosphorylation site localization of the identified phosphopeptides was performed by the phosphoRS algorithm 3.1 implemented in Proteome Discoverer using a 75% cut-off. For phosphopeptides that did not have a phosphorylation site with a pRS score above 75%, the site was counted as ''ambiguous''. The dimethyl-based quantitation method was chosen in Proteome Discoverer, with mass precision requirement of 2 ppm for consecutive precursor mass measurements. A 0.5 min retention time tolerance was applied for isotope pattern multiplets and allowed spectra with maximum 1 missing channels to be quantified. Mascot results were further filtered with the following criteria: (i) mass deviations of 610 ppm; (ii) Mascot ion score of at least 20; (iii) a minimum of 6 amino-acid residues per peptide; and (iv) position rank 1, which results in a peptide FDR,1%. Phosphopeptides that were found to be differentially phosphorylated were manually validated.
Quantification. Quantified phosphopeptides were normalized against the median of all non-phosphopeptides. For peptides with phosphorylation sites with a pRS score above 75%, only these phosphopeptides were used for quantification. For peptides with only pRS scores below 75% all phosphopeptides were used for quantification, but the phosphorylation site was counted as ''ambiguous''. Log2 ratios of heavy/light and medium/light of more than 1 or less than 21 were accepted as significant changes in phosphorylation levels.
Annotation. Phosphorylation sites of each peptide were compared to the total protein sequence of the corresponding zebrafish protein to identify the phosphorylation site. These sites were then compared to human protein in Phosphosite.org. For ambiguous phosphorylation sites (pRS,75) the sequence was compared to Phosphosite.org to identify the most commonly identified site. Non-annotated peptides were BLASTed against the zebrafish proteome to identify the protein.

Mass spectrometric analysis identifies Fer kinase as a hypotyrosyl-phosphorylated protein in NS and LS zebrafish embryos
To identify novel downstream targets of disease associated Shp2 that are affected similarly in NS and LS, we performed a mass spectrometric analysis on zebrafish embryos as outlined in Figure 1A. In short, zebrafish embryos were injected with synthetic mRNA encoding wild type Shp2, NS-Shp2 (D61G) or LS-Shp2 (A462T) which induced developmental defects at 24 hpf [9] ( Figure 1B). Embryos were co-injected with mRNA encoding GFP, which facilitated selection of embryos that were injected properly. Approximately 2000 embryos per condition were lysed and combined prior to digestion with LysC and Trypsin. Zebrafish lysates were isotopically labeled with normal formaldehyle (CH 2 O, light) for the WT embryos, deuteroformaldehyde (CD 2 O, medium) for the NS embryos and 13C deuteroformaldehyde ( 13 CD 2 O, heavy), for LS embryos. Labeled peptides were then combined and the mixture was processed further. The labeled peptide mixtures were enriched for pTyr by immunoprecipitation, using PY99 agarose beads [17,18] ( Figure 1A).
LC-MS was performed and MS1 and MS2 spectra were obtained. Ratios of light/medium and light/heavy peptides were determined with Proteome Discoverer 1.3.0.399 using the MS 1 spectra and normalized to the total ratio of all (non-phosphorylated) peptides as input. The input corresponding to the LEOPARD Shp2 expressing embryos (heavy labeled) was 2.35 fold higher than wild type (light labeled), whereas Noonan (medium labeled), was 1.38 fold higher than wild type. Whereas many peptides were identified, only a small number of phosphopeptides were identified and of suitable quality for quantification (see Table S1). In total, we identified and quantified 69 phosphorylation sites, of which 44 were pTyr sites with a pRS score of .75%. The other phosphorylation sites were predicted to be on serine or threonine, or the pTyr site could not be localized. The low abundance of tyrosine phosphorylation is likely the cause of the low number of identified and quantified pTyr peptides. In addition, many peptides that were identified corresponded to highly abundant proteins like actin, keratin and vitellogenin, a zebrafish yolk protein. These abundant proteins may have interfered with binding of the antibody to peptides of interest, leading to reduced specificity/efficiency for the enrichment of pTyr containing peptides.
The normalized Log2 ratios of the quantified phosphopeptides indicate that the majority of phosphopeptides was unaltered upon expression of NS or LS Shp2 compared to WT Shp2 (Table 1, Figure 1C). For clarity, the human phosphorylation sites are used in the text. A minor fraction of phosphopeptides was up-or downregulated in NS and LS Shp2 expressing zebrafish embryos, compared to WT Shp2 expressing embryos. This fraction of peptides may represent downstream factors that contribute to the etiology of both NS and LS. As an internal control for the expression of NS and LS Shp2 compared to wild type, we confirmed similar levels of Shp2 Y63 phosphopeptide in the WT compared to LS. Indeed, the 'WT' Shp2 pY63 peptide was not observed in the medium labeled (NS) sample, as this peptide has a D61G substitution and was thus not identified (See Table 1, Figure 1B, peptides ''m'' and ''n'').
Direct targets of Shp2 PTP activity were expected to have decreased pTyr in NS and an increase in LS. Surprisingly, these peptides were not abundant in our results (see Figure 1C upperleft quadrant). However, we identified pTyr peptide Y612 of IRS1B (peptide ''k''), a known Shp2 substrate, which was decreased (Log2: 21.18 fold) in NS but unaltered in LS compared to WT. Peptides that were increased in LS included a putative phoshorylation site of cell division kinase 15 and/or 16 (CDK15/ CDK16)(peptide ''i'')(Log2: 1.92 fold). However, these sites could not be localized with a precision score of .75%. In addition, S39 of eukaryotic translation initiation factor 3 subunit C (EIF3C) (peptide ''h'') was increased (Log2: 2.24 fold) in LS.
Alternatively, proteins that are phosphorylated downstream of Shp2 in a phosphatase-dependent manner would be expected in the lower right quadrant with increased phosphorylation in response to NS Shp2 and decreased phosphorylation in response to LS Shp2. We identified phosphopeptides that were increased in NS, including Y43 of Histone 2B (H2B) (peptide ''c'') (Log2: 1.50 fold), Y185 (Log2: 1.16 fold) (peptide ''f'') and T183 (Log2: 1.18 Figure 1. Comparative pTyr mass spectrometry on 1 day old zebrafish embryos expressing wild type, NS (D61G) or LS (A462T) Shp2. A. Work flow depicting the mass spectrometry approach. Approximately 2000 zebrafish embryos per condition were injected at the 1-cell stage and sorted for GFP expression. Embryos were lysed, trypsinised and labeled using the dimethyl labeling method. WT, NS and LS samples were combined and immunoprecipitated using pTyr specific antibodies. Immunoprecipitate was subjected to MS and peptides were identified and quantified based on MS2 and MS1 spectra, respectively. B. 1 dpf zebrafish embryos expressing WT, D61G and A462T Shp2. Body axis length, craniofacial defects and heart edema in D61G and A462T Shp2 expressing zebrafish are indicated with arrowheads. C. Normalized plot of quantified phosphopeptides Log2 ratios. Peptide ratios with Log2 ratios .21 and ,1 are indicated in black. Peptides with more that 16 Log2 difference are annotated with a-n (see Table 1    Additionally, we identified a phosphopeptide that was decreased in LS including Y666 of receptor tyrosine kinase-like orphan receptor 2 (ROR2) (peptide ''j'') (Log2: 21.15 fold). Multiple peptides that were affected in a similar way (upregulation or downregulation in both NS and LS) were identified (lowerleft and upper-right quadrant). To gain further understanding in how activating and inactivating mutations in NS and LS, respectively lead to similar outcomes, we chose to focus on this group of peptides. Two of the main hyperphosphorylated peptides in both NS and LS corresponded to PZR, an earlier identified target of Shp2 [19,20]. Peptides corresponding to Y241 (peptide ''a'') and Y263 (peptide ''b'') were increased (Log2: 2.95 fold and Log2: 2.24 fold, respectively in NS and Log2: 1.77 and Log2: 1.63 fold, respectively in LS). A further analysis of the role of PZR in NS and LS is described elsewhere [21]. In addition, we identified a pTyr peptide corresponding to Y989 in human Insulin receptor substrate 1B (IRS1B) to be highly abundant in NS (Log2: 1.39 fold) and LS (1.59 fold) (peptide ''d''). A phosphopeptide corresponding to Y792 of EphB3 (peptide ''g'') was also upregulated in NS and LS (Log2: 1.04 fold in NS and Log2: 2.90 in LS).
At the other end of the spectrum, the most decreased phosphopeptide in NS and LS corresponded to Fer kinase (peptide ''l'', NS: Log2 21.91 fold, LS: Log2 21.47 fold). The MS 2 spectrum of the Fer phosphopeptide is given in Figure S1. Unfortunately, a reference peptide for total Fer levels was not identified. Thus, the reduced levels of Fer phosphopeptide may represent reduced Fer phosphorylation, reduced total Fer levels or a combination of both. While phosphorylation of serine instead of tyrosine in this peptide could not be excluded, the tyrosine in this peptide is found frequently in other mass spectrometry based experiments (http://www.phosphosite.org). In addition, the peptide was immunoprecipitated with anti-pTyr antibodies, which indicates that this tyrosine is likely phosphorylated. Interestingly, this tyrosine is the autophosphorylation site of Fer [21]. Since Fer had not been studied in zebrafish nor had Fer been implicated in NS and LS, we further investigated the role of Fer in these syndromes.

Fer kinase expression in zebrafish embryos and Fer MO induced splicing defects
The Fujinami poultry sarcoma (fps)/Feline sarcoma (fes) related kinase (Fer) is a widely expressed, 94 kDa non-transmembrane protein tyrosine kinase (PTK) that is associated with cell migration, tumor growth and was first described in 1986 [22][23][24]. Fer is involved in focal adhesion and cadherin signaling, as well as ERK, Akt and Stat3 signaling [22]. Fer and Shp2 share many signaling pathways and a requirement for both proteins has been described in synapse formation [25]. It is thus likely that signaling between Fer and Shp2 plays an important role in cell-cell adhesion, migration and morphogenesis.
The identified peptide corresponded to tyrosine 714 (Y716 in zebrafish) which is the Fer autophosphorylation site, phosphorylation of which results in activation of Fer kinase activity [26]. Sequence alignment of the Fer kinase primary amino acid structure showed conservation of FER protein among vertebrates, from Homo sapiens to Danio rerio (Figure 2A), all coding for the conserved autophosphorylation site tyrosine. We cloned zebrafish fer from cDNA obtained from 1 dpf embryos. Sequence analysis confirmed that the fer transcript that was cloned was identical to ENSDART00000050957 or zebrafish fer-201 encoding fulllength Fer protein. Thus, Fer and its autophosphorylation site are evolutionarily conserved in vertebrates, including zebrafish.
As a first step to investigate the role of Fer kinase during development, in situ hybridization (ISH) was performed on zebrafish embryos, using a fer-specific probe ( Figure 2B). As a negative control sense fer probe was used ( Figure 2C). Ubiquitous expression of Fer is already apparent at the 8-cell stage, prior to zygotic transcription, indicating maternal contribution of fer mRNA. Fer remained ubiquitously expressed until 10 somite stage. At later stages, fer expression was enriched anteriorly and in the pectoral fins. In conclusion, fer expression was ubiquitous during gastrulation and more restricted during later stages of development.
To investigate the role of Fer in vivo, morpholinos against fer were injected into zebrafish embryos at the 1-cell stage. As a negative control, non-related Nacre MO was injected that does not induce developmental defects, as we have observed before [27]. We assessed Fer protein knockdown by immunoblotting using commercially available Fer antibodies, but unfortunately, these reagents did not allow detection of endogenous Fer in zebrafish embryos. Instead, knockdown of fer was verified by RT-PCR of knockdown embryos ( Figure 2D). Injection of fer e5i6 MO blocked splicing of fer intron 5, resulting in the incorporation of intron 5 and introducing multiple stop codons in the aberrantly spliced fer mRNA (Figure 2Da and Db). Injection of fer e9i9 MO caused aberrant splicing, thereby excluding exon 9 of fer (Figure 2Dd and De). These effects of MO injections were verified by sequencing (Figure 2Dc and Df, respectively).

Fer knockdown phenocopies Shp2 knockdown, NS and LS Shp2 expression in zebrafish
Injection of both Fer morpholinos (i5e6 and i9e9) induced developmental defects including reduced body axis length, heart edema and craniofacial defects at 3 dpf ( Figure 3A). These defects were similar to the phenotypes observed in zebrafish expressing NS and LS Shp2 that were analyzed in parallel ( Figure 3A). MOs are known to induce non-specific p53 activation and apoptosis [12]. To rule out that the observed phenotype is the mere result of p53 activation, co-injection of p53 MO was performed, which is an accepted control to assess specificity of MOs. Knockdown of p53 did not affect the fer knockdown phenotype, indicating that the phenotype was independent of p53 ( Figure 3B). Taken together, two independent Fer MOs were used that blocked normal splicing of fer and induced developmental defects.
The craniofacial defects we observed in Fer knockdown embryos were reminiscent of the defects observed in Shp2 knockdown embryos as well as embryos expressing NS or LS Shp2 [9,28]. To verify these craniofacial defects, we first used the Tg(24.9sox10:EGFP) ba2 transgenic line, which expresses GFP in neural crest-derived cartilage in the head [29]. We observed heart edema (Figure 3Ca, Cd and Cg) an increased distance between the eyes (Figure 3Cb, Ce and Ch), an increased angle of the ceratohyal (Figure 3Cc, Cf and Ci, red asterisk), a less protruded Meckel's cartilage (white asterisk), confirming the craniofacial defects. To verify the specificity of the Fer MOs again, we coinjected low doses of the two Fer-MOs that did not induce developmental defects on their own ( Figure 3D). Co-injection of suboptimal doses of both Fer MOs together induced a drastic increase in the Fer knockdown phenotype with reduced body axis extension, heart edema (orange) and craniofacial defects (yellow) as prominent features (see Figure 3D lower panel). This was further verified by alcian blue stainings on Fer knockdown embryos at 4 dpf ( Figure 3E). The angle of Meckel's cartilage ( Figure 3F) and the ratio of the distance between the eyes and the back of the head to the ceratohyal ( Figure 3G) were used to quantify these defects. Whereas NIC embryos showed an angle of 8561 degrees (n = 40), Quantification of the craniofacial defects of these embryos showed an angle of 12366, n = 70 (p,0.001) and a ratio of 0.7960.01 (p,0.001) ( Figure 3E-G).
We tried to rescue the Fer MO-induced developmental defects by expression of mRNA encoding full length Fer. Unfortunately, despite many attempts, we failed to rescue the Fer MO induced developmental defects ( Figure S2A). Expression of Fer by itself induced reduced body axis extension, cardiac edema and craniofacial defects ( Figure S2B), which impaired assessment of the rescues. It is not uncommon that up-and down-regulation of signalling proteins, including Shp2 and RhoA, induce similar developmental defects [9]. In conclusion, knockdown of Fer was specific and resulted in craniofacial defects, heart edema and reduced body axis extension, which is reminiscent of the is depicted. Embryos were fixed and stained with Alcian blue at 4 dpf and imaged laterally and ventrally. For quantification, the angle of the ceratohyal (F) and the ratio of the distance from the back of the head to Meckel's cartilage and the width of the head was determined (G) (* indicates significance, Student's t-test p,0.005). doi:10.1371/journal.pone.0106682.g003 Figure 4. Fer knockdown resulted in C&E defects but not in changes in cell fate. A. Krox20/myoD in situ hybridization as a method to quantify C&E defects. Krox20 (red) staining for rhombomeres 3 and 5 was used to measure the width, which correlates with convergence cell movements in the embryo. MyoD (light blue) staining for the somites was used to measure the length, i.e. extension of the embryo. The ratio of Krox20/myoD correlates directly with convergence & extension cell movements during gastrulation. B. Flatmounts of krox20/myoD stained NIC (n = 70), control MO (n = 44) and Fer MO (n = 40) embryos. C. The ratio of the width of a krox20-positive rhombomere and the length of 8 somites was determined (* indicates significance, Student's t-test p,0.005). D-X. Embryos were injected at the 1-cell stage with control MO or Fer e9i9 MO and subjected to ISH for various markers of cell fate determination. Note that the staining of the gsc, pax2, six3 and cyc probes show broader and shorter expression in Fer knockdown embryos than in controls. doi:10.1371/journal.pone.0106682.g004 developmental defects in Shp2 knockdown embryos as well as NS and LS Shp2 expressing embryos.

Fer knockdown induced convergence and extension defects
Fer knockdown phenocopied Shp2 knockdown and NS and LS expression at later developmental stages. Loss of Shp2 and expression of NS and LS variants of Shp2 induce C&E defects in developing zebrafish embryos [9]. We investigated whether Fer knockdown caused developmental defects during gastrulation as well. To investigate C&E defects upon Fer knockdown, embryos were subjected to in situ hybridization (ISH) using probes for myoD, staining the somites and krox20 staining rhombomeres 3 and 5, a verified method to quantify C&E defects in developing zebrafish embryos ( Figure 4A) [30,31]. Indeed, Fer morphant embryos showed C&E defects compared to control morphants (Student's t-test, p,0.005) ( Figure 4B and 4C). In addition to C&E defects, cell specification may be affected by the loss of Fer function. To investigate this, we subjected developing embryos to ISH using panel of cell fate markers, including bmp2b, cyc, chd, ntl, six3, gsc and pax2 ( Figure 4D-X). Expression levels of these markers were unchanged in Fer MO and control MO injected embryos compared to non-injected controls (NIC). However, expression of gsc, pax2, six3 and cyc showed broader and shorter signal in Fer knockdown embryos, compared to NIC and control MO injected embryos, likely resulting from C&E defects ( Figure 4 J-L, P-R, S-U and V-X, respectively). Thus, loss of Fer leads to C&E defects similar to Shp2 knockdown in developing zebrafish, and does not cause alterations in early cell fate determination.

Fer knockdown cooperates with NS and LS, but not WT Shp2 expression
Mass spectrometric analysis showed reduced Fer Y714 phosphopeptide. In developing zebrafish, we found that loss of Fer phenocopies Shp2 knockdown and expression of NS and LS Shp2, and leads to C&E defects. To investigate if downregulation of Fer contributes to the pathogenesis of NS and LS, we performed a genetic epistasis experiment, where we induced suboptimal knockdown of Fer in combination with injection of low amounts of NS and LS Shp2 mRNA that do not induce defects by themselves. Control injections only marginally affected craniofacial development. The ceratohyal angle in control injected embryos was 9361 degrees, which was a slight increase compared to the ceratohyal angle of control injected embryos, 8861 degrees (p,0.05). Whereas Fer knockdown caused developmental defects ( Figure 5A, B), partial knockdown of Fer using a low dose of Fer MO e9i9 did not induce developmental defects (ceratohyal angle: 8961 degrees, n.s). Likewise, injection of NS and LS, but not WT Shp2 mRNA at normal doses induced heart edema, reduced body axis extension and craniofacial defects (NS: ceratohyal angle: 11063 degrees, p,0.005 and LS: angle: 9662 degrees, p,0.001). At lower doses, NS and LS Shp2 also did not cause major developmental defects or an increase in the angle of the ceratohyal compared to control injections (Low NS: 9162 degrees, n.s and low LS: 8462 degrees, p,0.05). When combined with low doses of Fer MO however, NS and LS showed heart edema and major increases in craniofacial defects (Low NS + low Fer MO: 10363 degrees, p,0.001 and Low LS + low Fer MO: 9862 degrees, p, 0.001) ( Figure 5A, 5C and 5D). This indicates that a decrease in Fer cooperates with NS and LS to induce developmental defects seen in NS and LS.

Discussion
Using a comparative phosphoproteomics approach focused on pTyr-containing proteins, we identified a phosphopeptide corresponding to Fer kinase as the main decreased phosphopeptide in zebrafish embryos expressing NS or LS mutant Shp2 compared to WT. Fer knockdown induced developmental defects in zebrafish embryos that are reminiscent of defects induced by NS and LS Shp2. Epistatic interaction analyses suggested a genetic interaction between Fer kinase and mutant Shp2.
Previously, we used comparative phosphoproteomics in zebrafish to identify differences between wild type and Fyn/Yes knockdown embryos [18]. At that time, TiO2 columns were used to enrich for phosphopeptides, and only highly abundant phosphoserine-and phosphothreonine-containing phosphopeptides were identified, excluding low abundant pTyr-containing peptides [18]. Improved methods using immunoprecipitation of pTyr peptides allow for the identification of endogenous tyrosine phosphorylated peptides [27]. Stable isotope dimethyl labelling allows for the quantitative comparison of three different samples in an economical fashion, compared to other peptide labelling methods [17]. Other groups have successfully analysed the proteome of zebrafish under different conditions [32][33][34], however these experiments often focussed on the whole proteome in a specific organ. Given the crucial role of tyrosine phosphorylation in signalling during development, we hypothesized that specific analysis of the tyrosyl phosphoproteome would provide insights into development and disease. Particularly in case of NS and LS, two syndromes that are associated with expression of mutant Shp2, a protein-tyrosine phosphatase. Recently, a MS based approach was used to identify proteins that were affected by the altered binding properties of NS and leukemia associated Shp2 [35]. Although highly informative, these experiments were performed using the tandem SH2 of Shp2 expressed in cells. As full length NS and LS Shp2 exhibit different dynamic properties than WT [36,37], using the full length protein under physiological conditions in vivo gives more relevant insights into the etiology of the disease. Therefore, we performed for the first time a comparative phosphoproteomics experiment in zebrafish using pTyr immunoprecipitation and stable isotope dimethyl labelling.
Using this technique, we identified several phosphopeptides that were increased or decreased in NS and LS Shp2 embryos. One of these peptides corresponded to the Fer autophosphorylation site and was the most strongly decreased phosphopeptide in the zebrafish disease model for NS and LS. Follow-up experiments showed that Fer mRNA is maternally contributed and is expressed ubiquitously during gastrulation. Later, Fer expression is restricted anteriorly and in the pectoral fins of developing zebrafish embryos ( Figure 2B). Fer is conserved and fps/fes/fer homologues exist in sponges (Fes/FER_SR), D. melanogaster (dfps85D) and C. elegans (FRK-1) [38,39]. In chick, Fer expression is ubiquitous, with higher expression levels during development than at later stages [24]. Also in mammalian cells, Fer expression is widely distributed [23]. Fer expression is vital for dorsal closure of C.elegans, yet its kinase activity seems to be redundant in development [39][40][41]. We demonstrate that in fish, a decrease in total Fer levels by MOinduced knockdown resulted in developmental defects that were reminiscent of Shp2 knockdown, and NS and LS expression in zebrafish ( Figure 3A) [9]. Moreover, loss of Fer resulted in C&E defects during gastrulation, which were also observed in NS and LS zebrafish (Figure 4) [9,31]. Indeed, also at later stages craniofacial malformations and heart edema were observed in Fer knockdown embryos, similar to NS and LS Shp2 expressing embryos ( Figure 3) [9,28,31,42]. We show a genetic interaction between fer and ptpn11, since loss of Fer contributes to the NS and LS phenotype ( Figure 5).
The mechanism underlying the role of Fer kinase in development remains to be determined definitively. Fer has been reported to be involved in various signaling pathways. Expression of full length fer is essential for Drosophila gastrulation via Src42A, a Drosophila Src homologue [40] However, it is unclear if kinase activity is essential for this function of Fer. In C.elegans, FRK-1 is essential for embryonic closure and morphogenesis but this is not dependent on FRK-1 kinase activity [39]. Depletion of Fer kinase activity leads to a reduction in p38 MAPK but not ERK activity in activated mast cells [43] Fer is able to sustain ERK activation under hypoxic conditions but kinase activity is not essential [44]. Interestingly, Fer kinase activity is essential for b-catenin-cadherin complex formation. Expression of kinase dead Fer attenuates formation of this complex. EGF treatment rescues this defect, indicating a compensatory mechanism for the loss of Fer kinase activity. Interestingly, PTP1B, another member of the PTP family, associates with Cadherin in a Fer kinase-dependent manner, but whether PTP1B dephosphorylates Fer remains to be determined [45]. How NS and LS Shp2 cause a downregulation of Fer protein tyrosine phosphorylation remains to be determined. Since both activated NS Shp2 and catalytically impaired LS Shp2 have similar effects on Fer, it is unlikely that Fer is a direct substrate of Shp2. Disease associated Shp2 was found to bind to hyperphosphorylated substrates, both in NS and LS and in experimental conditions [20,46]. Fer and Shp2 share many interacting partners, including PDGFR, integrins, PECAM-1 and others [25,[46][47][48]. In addition, Fer and Src (an important Shp2 interacting protein) share many substrates, including cortactin [41,49]. Disease associated Shp2 exhibits altered dynamics and binding properties [35][36][37]. Mutant Shp2 may quench binding proteins that are normally required for full activation of Fer kinase, resulting in reduced tyrosine phosphorylation of Fer upon expression of NS and LS Shp2.
Alternatively, Fer activity might be downregulated through enhanced growth factor signaling that occurs in NS and LS [36,45]. While many studies have shown a loss of phosphatase activity in LS Shp2, the ability of LS Shp2 to enhance ERK activation remains subject of debate [4,36]. In zebrafish, we observe enhanced ERK activation in both NS and LS [50] and hence, reduced Fer phosphorylation may result from a negative feedback mechanism, resulting from enhanced growth factor signaling. In conclusion, we identify Fer kinase as a likely downstream target of NS and LS Shp2 and we provide evidence that deregulation of Fer enhanced the NS and LS phenotype in vivo. Table S1 Raw MS data. Tab1 ''PD_Output'' shows raw, nonnormalized data from Proteome Discoverer, which is used for normalization on all nonphosphopeptides. ''numberOFphospho'': number of phosphoresidues present in the peptide. ''numberO-Flocalized'': number of localized phosphoresidue (pRS.75).

Supporting Information
''peptideSite'': localized site. Tab2 ''Aggregation'' shows data averaged per rows based on the following values (protein group accession, sequences, description, numberofphospho, numer of localized, peptide sites). ''SITES'' tab shows all normalized, quantified, localized, protein assigned peptides. ''SITES with lacking protein'' tab shows all normalized, quantified, localized, peptides, including non-protein assigned peptides. ''quantified only pRS,75'' shows all normalized quantified peptides for which no phosphosite could be assigned. ''all quant phosphopepts combined'' shows a combination of ''SITES with lacking protein'' and ''quantified only pRS,75'' with duplicate peptides for which one site is assigned and the other is not removed. '' Table 1'' shows ''all quant phosphopepts combined'' where non-assigned peptides are BLASTed against the zebrafish proteome to identify the protein, phosporylation site in zebrafish and the phosphosite is compared to human homologs from PhosphoSite.org to identify the site. Sites without a human homolog were removed from the analysis. (XLSX)