Tnk1/kos1 Knockout Mice Develop Spontaneous Tumors

Tnk1/Kos1 is a non–receptor protein tyrosine kinase implicated in negatively regulating cell growth in a mechanism requiring its intrinsic catalytic activity. Tnk1/Kos1 null mice were created by homologous recombination by deleting the catalytic domain. Both Tnk1 +/À and Tnk1 À/À mice develop spontaneous tumors, including lymphomas and carcinomas, at high rates [27% (14 of 52) and 43% (12 of 28), respectively]. Tnk1/Kos1 expression is silenced in tumors that develop in Tnk1 +/À mice but not in adjacent uninvolved tissue, and silencing occurs in association with Tnk1 promoter hyper-methylation. Tissues and murine embryonic fibroblasts derived from Tnk1/Kos1-null mice exhibit proportionally higher levels of basal and epidermal growth factor–stimulated Ras activation that results from increased Ras-guanine exchange factor (GEF) activity. Mechanistically, Tnk1/Kos1 can directly tyrosine phosphorylate growth factor receptor binding protein 2 (Grb2), which promotes disruption of the Grb2-Sos1 complex that mediates growth factor–induced Ras activation, providing dynamic regulation of Ras GEF activity with suppression of Ras. Thus, Tnk1/Kos1 is a tumor suppressor that functions to down-regulate Ras activity.


Introduction
Tnk1 is a 72-kDa non-receptor protein tyrosine kinase (NRPTK) located on human chromosome 17p13.1 and has been implicated in the regulation of apoptosis, cell growth, nuclear factor-nB, and Ras (refs.1, 2; Supplementary Fig. S1).In mice, two gene products (Tnk1a and Tnk1b) are predicted to be produced by alternative splicing and use of distinct polyadenylation signals from the murine Tnk1 locus in chromosome 11 (that corresponds to human chromosome 17p13.1; ref. 3).Deduced amino acid sequences from the Tnk1a transcript predict a 72-kDa protein, and when splicing from exons 8 to 9 does not occur, an alternative splicing gene product (Tnk1b) is produced (Supplementary Fig. S2).Tnk1b consists of eight exons encoding a 47-kDa protein that we initially discovered in differentiating murine embryonic stem cells and named it Kos1 (kinase of embryonic stem cells; ref. 3).Kos1 shares the NH 2 terminal 380 amino acids containing the entire kinase domain with Tnk1a and contains a unique 54 amino acid sequence at its COOH terminal.Tnk1/Kos1 belongs to the Ack family of NRPTKs.However, unlike Ack1 and Ack2 (4, 5), Tnk1/Kos1 lacks a CRIB motif that can bind the GTP-bound form of activated cdc42.The antisera raised against a peptide from the NH 2 terminal region of Tnk1/Kos1 detected only a 47-kDa product in cell lines (including human) and other tissues that were analyzed, indicating that the 47-kDa Kos1 might be the predominant protein expressed in mice (3).Tnk1/Kos1 expression is developmentally regulated in mouse embryos, as well as in differentiating embryonic stem cells in vitro (3).Forced expression of Tnk1/Kos1 in various cell lines inhibits cell growth by a mechanism requiring its intrinsic tyrosine kinase activity (2,3).Importantly, auto-tyrosine phosphorylation of Tnk1/Kos1 results in the suppression of Ras activity.Conversely, expression of a Kos1 kinase-inactive mutant (K148A) transforms NIH3T3 cells, as evidenced by the anchorage-independent growth in soft agar, associated with increased Ras activity (Supplementary Fig. S3).Furthermore, ectopic expression of Tnk1/Kos1 in either human lung or breast cancer cells can partially revert to their anchorageindependent phenotype, suggesting that Tnk1/Kos1 may be a tumor suppressor (3).
The Ras-Raf-MEK-ERK signaling pathway occupies a central role in regulating cell proliferation and differentiation (6,7).Interestingly, cell growth arrest mediated by Kos1 results from its indirect suppression of Ras activity (3).In response to growth factors, such as epidermal growth factor (EGF) or those in serum, Ras becomes GTP loaded and activated (8).Activated Ras promotes downstream growth signaling by sequentially activating Raf1 and the Map kinases (6)(7)(8)(9)(10).Dynamic termination of growth signaling occurs when Ras, through its intrinsic GTPase activity, cycles back to its GDP-bound, inactive state.Intrinsic Ras GTPase activity is stimulated by GTPase-activating proteins (GAP), such as NF1 (11,12).Ras can also be activated by membrane-associated guanine exchange factors (GEF), like Sos1, which shift the equilibrium from inactive Ras-GDP to the active Ras-GTP form (13,14). Importantly, the growth factor receptor binding protein 2 (Grb2), through its SH3 motif, binds and transports Sos1 to the plasma membrane to initiate Ras activation (13,14).Point mutation of Ras at amino acids G12, G13, and Q61 is transforming due to constitutive activation of Ras, which occurs in about one third of all human malignancies, underscoring the importance of Ras dysregulation in tumorigenesis (15)(16)(17)(18).Ras is reversibly regulated as part of dynamic growth factor-mediated signaling.For example, the mammalian Sprouty proteins inhibit Ras signaling by competing with Sos1 for Grb2 binding (19)(20)(21).Mutational inactivation of Ras inhibitors, including NF1, can lead to the spontaneous development of human tumors as a result of the up-regulation of Ras activity (22,23).NF1 knockout mice also develop spontaneous tumors due to hyperactivation of Ras, indicating that NF1 is a tumor suppressor (24,25).Another such negative regulator of Ras, the Ras association domain family 1a (Rassf1a), also functions as a tumor suppressor because both Rassf1a homozygous and heterozygous null mice develop spontaneous tumors (26,27).These reports underscore the importance of the negative regulators of Ras in tumorigenesis.
We have determined the function of Tnk1/Kos1 in vivo by gene targeting to create a Tnk1/Kos1 knockout mouse.Results indicate that both homozygous and heterozygous Tnk1/Kos1 mice develop spontaneous tumors at a high frequency in association with hyperactivation of Ras.These data indicate that Tnk1/Kos1 functions as a tumor suppressor.Whereas any direct interaction between Tnk1/ Kos1 and Ras could not be observed, we discovered that Grb2 can be directly tyrosine phosphorylated by Kos1 to result in the dissociation of the Grb2-Sos1 complex with indirect inhibition of Ras.

Materials and Methods
Generation of Tnk1/Kos1 null mice.The targeting vector was constructed using the pKO scrambler vector (Stratagene) that contains a Pgk-Neo (Neo).A 2.6-kb Sca1-Kpn1 genomic fragment containing the first three exons and a portion of the 5 ¶ sequence of exon 4 was subcloned into the 5 ¶ arm, whereas an EcoR1-Spe1 genomic fragment containing exons 8 to 13 (3.2 kb) was inserted into the 3 ¶ arm of the targeting vector (Fig. 1A).Embryonic stem cells (RW4) were electroporated with 10 Ag of linearized targeting vector and selected with G418 (Sigma-Aldrich, Inc.).Three hundred G418-resistant clones were further screened for homologous recombination by digesting their genomic DNAs isolated from the clones with BamH1 and hybridizing with the 3 ¶ probe (Fig. 1A).Correct homologous recombination of the two positive clones was further confirmed by the 5 ¶ probe on Xba1 digested samples (Fig. 1A).One of the two heterozygous clones was microinjected into C57BL/6 (B6) mouse blastocysts, followed by surgical implantation into the uteri of pseudopregnant foster mother.Chimeras were intercrossed with wild-type B6 mice, and Tnk1 heterozygous mice were maintained on 129Sv/B6 hybrid background.Heterozygous male and female mice were interbred to generate Tnk1 À/À mice.
Immunoprecipitation and immunoblotting.Cells or frozen tissues were homogenized and lysed in radioimmunoprecipitation assay lysis buffer (3).Immunoprecipitation was performed on a 500-Ag cell lysate using a-GFP (MBL), a-Grb2, and a-Sos1 (Santa Cruz Biotechnology) antibodies.Immunoblotting was performed with a-Kos1 antiserum (3), a-Grb2, a-Sos1, a-phosphorylated Erk, a-Erk2 (Santa Cruz Biotechnology), and a-Ras (Upstate Biotechnology).Protein was estimated using the Bradford reagent (Bio-Rad Laboratories).In the targeting construct, the entire catalytic domain (exons 4-8) was deleted and replaced with the neomycin-resistant cassette (Neo ).The thymidine kinase cassette (TK ) was used for negative selection.Restriction enzyme sites for BamH1 (B ), EcoRI (E), HindIII (H ), KpnI (K), ScaI (S ), SpeI (Sp ), and XbaI (X ) are indicated.Note that Xba1 and BamH1 sites were created in the mutant Kos1 allele by the insertion of the Neo cassette.These sites were used for distinguishing the knockout allele from the wild-type allele by Southern hybridization using the 5 ¶ probe or 3 ¶ probe.B, genotyping of Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À mice by Southern blot analysis.Genomic DNAs isolated from tails of Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À mice were digested by BamH1.Hybridization of the digested DNA with the 3 ¶ probe detected a 7.8-kb wild-type band and a 4.2-kb mutant band.C and D, expression of Kos1 in mice liver and MEFs.Liver lysates prepared from Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À littermates (C ) or cell lysates prepared from Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À MEFs (D ) were resolved on a 10% SDS-PAGE, and Kos1 expression was analyzed by Western blotting with anti-Kos1 antiserum.Erk2 expression level was used as an internal control for equal loading in each lane in both cases.
Ras activation assay.The in vitro Ras activation assay was carried out on MEFs or tissues including tumors (500 Ag of protein) using a nonradioactive kit (Upstate Biotechnology; ref. 3).To determine total Ras, 50 Ag of lysate in each case was run on 14% SDS-PAGE gel, and Western analysis was performed using a-Ras antibody (3).
Ras GEF activity assay.In vitro Ras GEF activity was performed as described (28).Ras-agarose (0.25 Ag) was incubated in 50 AL of reaction buffer [25 mmol/L Tris-HCl (pH 7.5), 5 mmol/L MgCl 2 , 1 mmol/L DTT, and 100 Ag of bovine serum albumin] with 100 Amol/L unlabeled GDP at 37jC for 30 min.Cell extract (3 Ag) was added to the unlabeled GDP Ras-agarose complex in the presence of 10 ACi of [ 3 H]GDP (Amersham Biosciences) and 18 mmol/L MgCl 2 at 37jC for 10 min.The Ras-agarose complex was pulleddown washed thrice with the reaction buffer, and the amount of bound Ras-[ 3 H]GDP was determined by Beckman Coulter LS6500 scintillation counter.Addition of 2 mmol/L GTP-g-S to the mix of [ 3 H]GDP, cell extract, and unlabeled GDP Ras-agarose complex abolished the exchange reaction.
Histologic analysis.Tissue samples fixed in 10% neutral buffered formalin overnight at room temperature, followed by transfer in 70% ethanol, were paraffin embedded.Staining was performed with H&E.Immunohistochemistry of the deparaffinized tissue sections were performed using primary antibodies [CD45R (B220) and CD3, BD PharMingen] and appropriate biotinylated secondary antibody.Slides were stained with 3,3 ¶-diaminobenzidine for visualization and counterstained with hematoxylin.
Methylation analysis.Genomic DNA (500 ng) isolated from wild-type liver and tumor mass, using a kit (Quagen, Inc.) following the manufacturer's instruction, was modified chemically by sodium bisulfite (Active Motif).The methylation status of the promoter region was analyzed by methylation-specific PCR using the following primers: forward 5 ¶-TATTAGGGAATTAATTGGTTTTCGGA-3 ¶, reverse 5 ¶-GAAAACGAAAAAAA-CAACTACGAA-3 ¶.The 136-bp amplified product obtained from each tumor was subcloned in pCRII vector (Invitrogen), and individual clones were sequenced to map the methylation sites.

Results
Tnk1/Kos1 gene targeting resulted in a null allele.To investigate the role of Tnk1/Kos1 in the development and regulation of Ras activity in vivo, we generated a strain of mice deficient in both Tnk1a and Kos1 gene products using conventional gene targeting technology.In the knockout vector, a Tnk1 genomic region that includes exons 4 to 8 was replaced with a neomycinresistant cassette (Fig. 1A).Because exons 4 to 8 encode the entire catalytic domain of Tnk1/Kos1, which is indispensable for its biological function (2,3), we expected that the resulting targeted allele will be a null allele.One of two correctly targeted embryonic stem cell clones was used for the generation of chimeric mice.The germline chimeras were crossed with wild-type C57BL/6 (B6) mice, and the Tnk1 heterozygous (Tnk1 +/À ) mouse line was established and maintained on a mixed 129/B6 hybrid background.Adult homozygous mice (Tnk1 À/À ) were obtained from intercrosses of heterozygotes at the expected Mendelian ratio (Fig. 1B), indicating that Tnk1a and Kos1 are dispensable for development and postnatal growth.Western blot analysis using anti-Kos1 antibodies on protein extract from the liver (Fig. 1C) and MEFs (Fig. 1D) showed an undetectable level of Kos1 expression in Tnk1 À/À mice, indicating that the Tnk1 knockout allele is null.Tissue from Tnk1 +/À mice and Tnk1 +/À MEFs showed reduced (f50%) Kos1 expression compared with wild-type (Tnk1 +/+ ) littermates.These data indicate a gene dosage effect for endogenous Kos1 expression.
Spontaneous tumors develop in Tnk1/Kos1 null and deficient mice.Tnk1/Kos1 null littermates were up to 24 months of age, during which time they were analyzed for spontaneous tumor formation.Whereas no tumors developed in 22 Tnk1 +/+ mice, 12 of 28 Tnk1 À/À mice (43%) developed spontaneous neoplasms with an average latency of 17 months (Table 1A).The majority of tumors occurred in the gastrointestinal tract, wherein seven mice developed malignant lymphomas of the small intestine and mesenteric lymph nodes.Lymphomas also developed in the spleen, thymus, salivary gland, and pancreas.Histologically, lymphomas resemble lymphoblastic lymphomas with immunohistochemistry, confirming that the tumors are predominantly of B-cell origin.An example of one such lymphoma invading the muscularis mucosa of the small intestine displays dense brown B cells rapidly proliferating in the wall layers (Table 1A, case 1; Fig. 2A and B).Another example of an aggressive lymphoma from a lymph node shows invasion of the adjacent fatty connective tissue (Table 1A, case 7; Fig. 2C).In the salivary gland, a histiocytic lymphoma was observed infiltrating the gland (Table 1A, case 10; Fig. 2D ).In addition, Tnk1 À/À mice can develop splenic hemangiosarcomas (Table 1A, case 2; Fig. 2F), lymphoma of the spleen (Table 1A, case 3; Fig. 2G), thymic lymphoma (Table 1A, case 12; Fig. 2H), epithelial tumors of the skin (Table 1A, case 4; Fig. 2E), adenocarcinoma of the lung (Table 1A, case 5; Fig. 2J  and K), and lymphoma of the pancreas (Table 1A, case 12; Fig. 2I).Interestingly, the Tnk1 +/À mice also develop spontaneous tumors, albeit at a lower frequency (Table 1B).Of 52 Tnk1 +/À mice necropsied, 14 (27%) developed tumors, the majority of which were  A-K) and heterozygous mice (L-P ).A, lymphoma, small intestine (400Â).H&E sections of a highly proliferating and invasive lymphoid tissue that has completely effaced the muscularis layers of small intestine.Inset shows the affected tissue.B, immunostaining of lymphoma using B220 (brown staining, inset ) and CD3 (blue staining, inset ).The B cells, intensely stained with B220 antibody, are pleomorphic with a significant level of necrosis.The small population of T cells that showed staining with CD3 antibody seems normal.C, lymphoma, mesenteric lymph node (400Â).The large and highly proliferating nodules in the mesenteric lymph node completely disrupted the normal cortex and medullary architecture of the lymph node, infiltrated the subcapsular and medullary sinuses, and have invaded the surrounding fatty connective tissue.The section shows large areas of necrosis, and the large blast cells that stain for B220 (not shown) is predominant in most areas.Inset shows the lymphoma of the affected tissue.D, histiocytic lymphoma in salivary gland (400Â).Neoplastic population of lymphocytes from the adjacent submandibular lymph node (not shown) infiltrated the salivary gland.In many areas, multinucleated giant cells are scattered among the pleomorphic population of lymphocytes.Inset shows the lymphoma of the affected tissue.E, trichoepithelioma (400Â).The mass within the dermis and subcutaneous tissue contains well-differentiated epithelial cells forming several components of the follicle, including hair bulb, hair, and follicular wall.F, spleen hemangiosarcoma (400Â).Neoplastic endothelial cells form variably sized blood spaces that occupy the red pulp and irregularly infiltrated the adjacent areas.G, lymphoma, spleen (400Â).Enlarged discrete nodules sometimes coalesce and largely disrupt the white pulp architecture.A mixed pleomorphic population of lymphocytes composed of follicular center cells, centrocytes, and small lymphocytes form the nodules.H, thymic lymphoma (400Â).Malignant lymphocytes have completely destroyed the architecture of the thymus.Inset shows tumor in the thymus.I, carcinoma of the pancreas.J and K, bronchio alveolar adenocarcinoma, lung (400Â).The tumor composed of a mildly pleomorphic population of neoplastic cells infiltrated the parenchyma.Inset shows tumor in the lung.L and M, adenocarcinoma, small intestine (400Â).Neoplastic cells frequently forming glands invade the layers of intestine, including the muscularis.Dense schirrous reaction surrounds isolated glands within the intestinal wall.Inset shows the tumor.N and O, hepatocellular carcinoma (400Â).A large portion of the liver is invaded by the moderately anaplastic malignant hepatocytes showing significant pleomorphism.Neoplastic cells have eosinophilic, often mildly vacuolated cytoplasm.Inset shows the tumor.P, pancreatic carcinoma (400Â).Inset shows the tumor.The scale bar shown in A is applicable to all.A, C, D, E, F, G, H, I, K, L, N and P are at 50 AM, whereas B, J, M and O are at 500 AM.
Tnk1/Kos1, a NRPTK, Is a Tumor Suppressor www.aacrjournals.orgepithelial in origin (72%) in contrast to the higher incidence of lymphoid tumors in Tnk1 À/À mice (75%).Among 14 Tnk1 +/À tumor-bearing mice, nine formed lung or intestinal adenomas.An adenocarcinoma that originated in the small intestine shows a highly invasive pattern of tumor cells arranged in irregular glandular structures (Table 1B, case 4; Fig. 2M and N).A hepatocellular carcinoma was observed to invade three lobes of the liver that was composed of tumor cells demonstrating highlevel pleomorphism (Table 1B, case 6; Fig. 2O and P).We have observed only a single pancreatic islet tumor (Table 1B, case14; Fig. 2L), but f30% of the Tnk1 +/À mice did develop islet cell hyperplasia.Although malignant lymphomas occur at a lower frequency (28%), they did develop in the spleen, pancreas, small intestine, and mesenteric lymph nodes.In addition, the Tnk1 +/À mice formed benign skin tumors.Interestingly, 50% of Tnk1 À/À and 28% of Tnk1 +/À tumor-bearing mice synchronously develop more than one primary neoplasm.These findings indicate a strong predisposing influence for the loss of Kos1 expression on spontaneous tumor development in various tissues, consistent with the ubiquitous expression pattern of Kos1.Therefore, we can conclude that Tnk1/Kos1 has tumor suppressor activity.
Tnk1/Kos1 null and deficient mice display hyperactivated Ras.Tnk1/Kos1 was initially discovered to function as a negative regulator of cell growth (2,3).We reported that Tnk1/Kos1 could inhibit cell growth by indirectly suppressing Ras activity with consequent down-regulation of the Ras-Raf-MAPK pathway (3).Importantly, we discovered that wild-type human Tnk1 also suppresses Ras activity (Supplementary Fig. S1).However, the mechanism was not clear.We assayed the GTP-Ras level (a measure of activated Ras) in liver lysates from Tnk1 À/À , Tnk1 +/À , and Tnk1 +/+ littermates using a Raf1-RBD agarose strategy to selectively bind and pull-down activated Ras.The basal GTP-Ras level (i.e., unstimulated) was found to be significantly higher in the cell lysates from either the heterozygous (f2.5-fold) or homozygous null mice (f5-fold) compared with wild-type littermates (Fig. 3A).These data confirm our initial biochemical findings in cell lines (3) and support the hypothesis that Tnk1/Kos1 functions to suppress Ras activity in vivo.
Forced expression of exogenous wild-type Kos1 in NIH3T3 cells was previously shown to suppress Ras activation in response to EGF (Supplementary Fig. S3; ref. 3).Therefore, we tested the effect of EGF stimulation on Ras activation in MEFs derived from Tnk1 À/À , Tnk1 +/À , and Tnk1 +/+ embryos.MEFs were grown to confluence in the presence of serum, washed, and placed in a serum-free medium for 24 hours, after which time the cells were either treated with or without 100 ng/mL EGF for 5 minutes.The cells were then harvested, and GTP-Ras levels were measured in the cell lysates.Lysates from untreated Tnk1 +/À and Tnk1 À/À cells show a f2-fold and f3-fold increase, respectively, in Ras activity compared with wild-type cells (Fig. 3B).In response to EGF stimulation, both Tnk1 +/À and Tnk1 À/À cells exhibit an additional f3-fold increase in Ras activation compared with the Tnk1 +/+ MEF cells (right).Thus, the reduction or loss of Tnk1/Kos1 expression in both Tnk1 +/À and Tnk1 À/À MEFs, respectively, results in a substantial up-regulation of Ras-GTP with elevated mitogenactivated protein kinase activity compared with the Tnk1 +/+ MEFs (Supplementary Fig. S4).These data indicate that Tnk1/Kos1 functions as a negative regulator of cell growth by dynamically reversing Ras activation.
Tnk1/Kos1 indirectly inhibits Ras by uncoupling the Grb2-Sos1 complex.A direct interaction between Tnk1/Kos1 and Ras Figure 3. Endogenous Ras and Ras GEF activities.A, hyperactivation of Ras in Tnk1/Kos1 knockout mice.GTP bound Ras, quantitatively pulled down from liver lysates prepared from Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À mice using Raf1-RBD agarose, was resolved on SDS-PAGE gel and immunoblotted with Ras antibody.The Ras activity (fold) in Tnk1/Kos1 knockout mice compared with the wild-type littermate was determined by densitometry, and the graph plotted represents an average of data obtained from three sets of experiments using different wild-type and mutant-paired littermates.Total Ras in each case was determined by running 50 Ag of tissue lysate on a separate gel, as described in Materials and Methods.B, hyperactivation of Ras in response to EGF in the absence of Kos1 expression.Embryonic fibroblasts prepared from Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À embryos (E15) were grown to 80% confluence, serum starved for 24 h, and treated with 100 ng/mL EGF for 5 min.GTP bound Ras was quantitatively pulled down from serum-starved cells (control, without EGF; left ) and cells exposed to EGF (with EGF; right ) and analyzed as above to obtain the fold activation in response to EGF treatment.The graph represents an average obtained from three experiments.Total Ras in 50-Ag cell lysate serve as an internal control for equal loading.C, Ras GEF activity in Tnk1/Kos1 knockout mice.The release and exchange of unlabeled GDP for [ 3 H]GDP on Ras-Agarose (0.25 Ag) was measured in the presence of clarified cell or liver extract (3 Ag) from the Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À littermates.Black columns, [ 3 H]GDP bound Ras; gray columns, annulment of the exchange reaction in the presence of 2 mmol/L GTP-g-S.The graph plotted represents an average of data obtained from three sets of experiments using different wild-type and mutant-paired littermates.

Cancer Research
Cancer Res 2008; 68: (21).November 1, 2008 has not been observed.Therefore, we tested whether Kos1 may function to regulate Ras by interacting with an upstream Ras activator.In response to growth factor-stimulated signaling, the Grb2-Sos1 complex interacts with the phosphotyrosine residues on activated receptor protein tyrosine kinases, like EGF receptor (EGFR), through its SH2 domain.Docking of the Grb2-Sos1 complex is required for Sos1, a Ras GEF, to directly interact with and activate Ras at the plasma membrane (13,14).To test whether Tnk1/Kos1 may regulate Ras GEF activity, Ras GEF activity was measured in liver lysates prepared from Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À littermate animals (Fig. 3C and Supplementary Fig. S5).Lysates from Tnk1 +/À and Tnk1 À/À mice show a f1.5-fold and f3-fold increase, respectively, in Ras GEF activity compared with those from Tnk1 +/+ lysates.These data indicate that Tnk1/Kos1 can suppress Ras GEF activity in vivo.
Tyrosine phosphorylation of Grb2 uncouples the Ras activation complex by disrupting the Grb2-Sos1 complex in vivo (29).Therefore, we tested whether Kos1 can directly phosphorylate Grb2, and if so, whether Grb2 phosphorylation may affect the stability of the Grb2-Sos1 complex.Cos7 cells were transfected with Sos1, and lysates were subjected to immunoprecipitation using anti-Sos1 agarose beads.The Grb2-Sos1 complex contained on immunobeads was used as substrate for either purified Flag-Kos1 or the catalytically inactive Flag-Kos1(CN) mutant in an in vitro protein tyrosine kinase assay (3).Our data indicate that both Flag-Kos1 and Flag-Kos1(CN) associate with the Grb2-Sos1 complex, but only Flag-Kos1 will tyrosine phosphorylate Grb2 (Fig. 4A and Supplementary Fig. S6).Flag-Kos1, but not Flag-Kos1(CN), is able to auto-tyrosine phosphorylate, indicating that only Flag-Kos1 is enzymatically active.Thus, we can conclude that Grb2 is a Kos1 substrate.Importantly, tyrosine phosphorylation of Grb2 by Kos1 is associated with a decrease in Grb2 bound to Sos1, indicating that Kos1 mediates the disruption of the Grb2-Sos1 complex.Consistent with these findings, endogenous tyrosine-phosphorylated Grb2 can be pulled down from Flag-Kos1 expressing Cos7 cell lysates using an anti-Sos1 antibody, but not from lysates from cells expressing Flag-Kos1(CN) or Flag-only vector.These data show that Grb2 is in vivo tyrosine-phosphorylated by Kos1 (Fig. 4B).Furthermore, a 2-fold to 3-fold decrease in the interaction between Grb2 and Sos1 occurs in the presence of Flag-Kos1, suggesting a mechanism by which Kos1 may dynamically reverse Ras activation (Fig. 4A and B).Next, to determine the phosphorylation status of Grb2 in Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À mice, liver lysates were subjected to immunoprecipitation using an anti-Grb2 or anti-Sos1 antibody.The resolved proteins transferred to a nitrocellulose membrane were analyzed by Western blotting using Py20, anti-Grb2, and anti-Sos1 antibodies.Grb2 immunoprecipitates from Tnk1 +/+ mice show a f10-fold and f4-fold increase, respectively, in Grb2 tyrosine phosphorylation over that in Tnk1 À/À and Tnk1 +/À lysates.Therefore, increased Grb2 phosphorylation by Tnk1/Kos1 is Figure 4. Grb2 is a physiologic substrate of Kos1.A, in vitro protein tyrosine kinase assay performed with 1.5 Ag of purified Flag-Kos1 or kinase inactive mutant Flag-Kos1(CN) using immunoprecipitated Grb2-Sos1 complex as substrate in the presence of cold ATP.The pellets (P) were separated from the supernatant, washed thrice with 20 AL of kinase buffer, and resolved by SDS-PAGE.Western analysis indicates that Grb2 is tyrosine phosphorylated by Flag-Kos1 but not by Flag-Kos1(CN) or Flag.The pellet-associated Flag-Kos1 is auto-tyrosine phosphorylated but not Kos1(CN).In the presence of Flag-Kos1, the Sos1 bound Grb2 is 2-fold to 3-fold less than in the presence of Flag or Flag-Kos1(CN).B, Grb2 is tyrosine phosphorylated in vivo in the presence of exogenously expressed Flag-Kos1, but not in the presence of Flag-Kos1(CN) or Flag in Cos7 cells.C, the liver of the Tnk1 +/+ mice show f10-fold and f4-fold increase in tyrosine phosphorylated Grb2 over the Tnk1 À/À and Tnk1 +/À mice, respectively, with a substantial reduced Grb2-Sos1 complex when immunoprecipitated with anti-Grb2.D, the Sos1 immunoprecipitates from liver lysates of the Tnk1 À/À and Tnk1 +/À mice show an increase in Grb2-Sos1 complex compared with lysates from the Tnk1 +/+ mice.Protein bands were subjected to densitometric analysis, and the graph plotted represents an average of data obtained from three sets of experiment using different wild-type and mutant-paired littermates.
Epigenetic silencing of Tnk1/Kos1 occurs in tumors that develop in Tnk1 +/À mice.Interestingly, Tnk1 +/À mice also develop malignant tumors with aging, albeit at a lower frequency than in Tnk1 À/À mice (Table 1).Results indicate that the Ras-GTP levels in the lymphomas that develop in the Tnk1 +/À mice are significantly higher than that in adjacent, uninvolved tissue or from tissue of Tnk1 +/+ mice (Fig. 5A, lanes 4, 2, and 1 and Supplementary Fig. S4).Furthermore, lymphomas that arise in the Tnk1 +/À mice express similar high levels of activated Ras to that found in lymphomas arising from Tnk1 À/À mice (Fig. 5A, lanes 4 and 3).These data indicate that the nontargeted wild-type allele in Tnk1 +/À mice may be inactivated in these tumors.Because tumor suppressors can be epigenetically silenced by promoter methylation in tumors (30)(31)(32), we tested whether the wild-type Tnk1/Kos1 allele is altered in the Tnk1 +/À tumors.The Tnk1/Kos1 proximal promoter is GC-rich and contains several putative CpG sites that may be susceptible to methylation (S.Hoare, K. Hoare, and W.S. May, data not shown).Genomic DNA was isolated from lymphomas and carcinomas that developed in the Tnk1 +/À mice and from the liver of Tnk1 +/+ littermates.After modifying the genomic DNA with sodium bisulfite, PCR was performed using methylation-specific primers (Fig. 5B).A PCR product was obtained from all tumor tissues isolated from an intestinal adenocarcinoma (Fig. 5B, lane 1), a hepatocellular carcinoma (Fig. 5B, lane 2), a hepatoma (Fig. 5B, lane 3), and lymphomas (Fig. 5B, lanes 4, 5, and 6).However, liver DNA isolated from the Tnk1 +/+ mice did not have a methylated promoter fragment (Fig. 5B, lane 7).To map the potential promoter methylation sites, the PCR product from each tumor or the Tnk1 +/+ tissue was subcloned and sequenced.Results reveal that all six of the CpG sites in the Tnk1/Kos1 proximal promoter were methylated, but none of them were methylated in the Tnk1 +/+ tissue.These data indicate that promoter methylation may be one mechanism by which transcriptional gene silencing of Tnk1/Kos1 expression may occur in the spontaneously arising tumors of the Tnk1 +/À mice.Therefore, we determined the level of Kos1 expression in intestinal lymphomas that develop in Tnk1 +/À mice compared with uninvolved liver from the same Tnk1 +/À mouse or from Tnk1 +/+ or Tnk1 À/À littermates.Results indicate that Kos1 expression is not detectable in tumor tissue from Tnk1 +/À mice but is easily detectable in the uninvolved tissue from the same Tnk1 +/À or Tnk1 +/+ mice (Fig. 5C, lanes 3, 4, and 5).We conclude that a functional loss of heterozygosity through Tnk1/Kos1 promoter methylation may account for the spontaneous tumorigenesis observed in Tnk1 +/À mice whose tumors do not express Tnk1/Kos1.

Discussion
Tnk1/Kos1 produces alternative splice products, which contain the functionally indispensable kinase domain (Supplementary Fig. S2; refs.2, 3).Whereas the 47-kDa Kos1 is ubiquitously expressed in mouse tissues, as well as human fetal blood and leukemia cell lines, the predicted 72-kDa Tnk1 can be detected in human fetal blood, but not in mouse cells, by Western blotting (3,33).A point mutation in Tnk1/Kos1 that inactivates the catalytic activity has been shown to abolish its growth inhibitory properties, indicating that the tyrosine kinase activity is required for Tnk1/Kos1 function (2,3).Interestingly, in support of this notion, a point mutation in the catalytic domain of Tnk1 (R339K) has recently been identified in a human lung adenocarcinoma that may potentially affect its tyrosine kinase activity (34).
We previously discovered that both Kos1 and Tnk1 can inhibit Ras activation (ref.3; Supplementary Fig. S1).However, the mechanism was not clear.Therefore, we chose to delete the kinase domain in the mouse gene to effectively eliminate the kinase activity of either splice form produced, making the gene knockout results relevant in both the murine and human context (Fig. 1A).Our results show that the loss of Tnk1/Kos1 expression in mice predisposes animals to develop spontaneous tumors upon aging.Interestingly, both the Tnk1 À/À and Tnk1 +/À mice develop tumors at a high frequency (i.e., 43% and 27%, respectively).Furthermore, tumors arise in several organs, including lymphoid and epithelial tissues, consistent with the ubiquitous expression pattern of Kos1 (3).Among the neoplasms Figure 5. Aberrant Ras activation, epigenetic silencing, and loss of Kos1 expression in tumors developed in Tnk1 +/À mice.A, hyperactivation of Ras in lymphomas developed in Tnk1 +/À and Tnk1 À/À mice.Tissue lysates were prepared from lymphoma developed in the small intestine of Tnk1 À/À (lane 3 ; Table 1A, case 11) and Tnk1 +/À (lane 4; Table 1B, case 8) along with normal liver tissue derived from Tnk1 +/+ and Tnk1 +/À littermates (lanes 1 and 2 , respectively) as a control, and GTP-Ras level was determined in each case as described earlier.B, analysis of Tnk1/Kos1 promoter methylation in malignant tumors developed in heterozygous mice.Genomic DNA treated with sodium bisulfite was analyzed by methylation-specific PCR.PCR product derived from methylated DNA is 136 bp.Lanes 5, 6, and 7 (cases 8, 10, and 13 in Table 1B, respectively) are intestinal lymphomas, and lane 1 is an adenocarcinoma developed in small intestine (case 4), lane 2 is a hepatocellular carcinoma (case 6), and lane 3 represents a hepatoma (case 7).Lane 7 is a control DNA isolated from wild-type mouse liver.C, absence of Kos1 expression in intestinal lymphomas collected from heterozygous mice.Tissue lysates were prepared from lymphomas collected from heterozygous mice (lanes 3, 4 , and 5 corresponding to cases 8, 10, and 14, respectively in Table 1B), a homozygous mice (lane 6; Table 1A, case 11), and normal liver tissue from Tnk1 +/À (lane 1 ), Tnk1 +/+ (lane 2 ), and Tnk1 À/À (lane 7) littermates were on the same SDS-PAGE gel for comparison of Kos1 expression.Protein (50 Ag) from each lysate was resolved on a 10% SDS-PAGE gel, and the blot was probed with Kos1 antibody.Erk2 level determined in each case indicates equal loadings.
that develop in the Tnk1 À/À mice, the most frequent are lymphomas, predominantly of the small intestine and mesentery.The lymphomas that develop are in large volumes (1 cm or larger in diameter) and occur with a mean latency of 17 months.The Tnk1 À/À mice also develop carcinomas, including hepatocellular and adenocarcinomas of the lung and other tissues (Table 1).Interestingly, Tnk1 +/À mice also develop both lymphomas and carcinomas, albeit at a lower frequency but with similar average latency as the Tnk1 À/À mice.Interestingly, tumors arising in the Tnk1 +/À mice are predominantly adenomas and adenocarcinomas of the small intestine and lung (Table 1).Importantly, no tumors were observed in the Tnk1 +/+ littermates at ages of up to 24 months.These data indicate that expression of Tnk1/Kos1 is required to block the development of spontaneous tumors and for maintenance of homeostatic growth in these mice.Conversely, the development of malignant tumors of the lung, liver, small intestine, and pancreas that occur in the null mice indicate that Tnk1/Kos1 deficiency is sufficient to predispose mice to spontaneous tumor development.Thus, we can conclude that Tnk1/Kos1 is a tumor suppressor.
Cell growth and development is regulated by the Ras-Raf1-MAPK pathway (6,7).We reported that expression of Tnk1/Kos1 can negatively regulate Ras activity, leading to growth inhibition and cell death (3).Now we observe a significant increase in Ras activity in liver extracts obtained from Tnk1 +/À and Tnk1 À/À mice compared with the Tnk1 +/+ littermates (Fig. 3A), confirming that Tnk1/Kos1 can suppress Ras.Because of Ras is a tumorigenic mechanism that occurs in up to 30% of human tumors, these findings indicate that Kos1 and Tnk1 are necessary for Ras regulation in normal cell growth.When growth factors bind to their surface receptors, negative regulators of the downstream signaling pathways also become activated to dynamically regulate and prevent excessive/prolonged mitogenic signaling that may lead to dysregulated growth.In the case of Tnk1/Kos1, our data indicate that MEFs derived from Tnk1 +/À or Tnk1 À/À mice stimulated with EGF show a rapid up-regulation of GTP-Ras (Fig. 3B), indicating that Tnk1/Kos1 functions as a negative inhibitor of Ras activation in response to growth factor signaling.Furthermore, the direct interaction of Kos1 with the Grb2-Sos1 complex results in Kos1-mediated tyrosine phosphorylation of Grb2 with disruption of the complex (Fig. 4 and Supplementary Fig. S6) and inhibition of Ras GEF activity (Fig. 3C and Supplementary Fig. S5).These data suggest a plausible mechanism by which Tnk1/Kos1 can indirectly suppress Ras.Grb2 is a signal transducer and regulator of Ras that can be activated by several human oncogenes, including oncogenically mutated EGFR and the Bcr/Abl fusion (8).Grb2 contains an SH2 domain flanked by SH3 domains on the amino and carboxyl termini that form h barrels (35).The two SH3 domains bind the proline-rich motif of Sos1 to form the active Grb2-Sos1 complex, which stimulates GTP loading onto Ras (12,13).Activation of Ras is abrogated when the Grb2-Sos1 complex is blocked, either by using SH3 domain-mimetic peptides of Grb2 (36,37) or by expression of Grb2 mutants lacking the amino or carboxyl terminal SH3 domain (38).Importantly, disruption of the Grb2-Sos1 complex is associated with the reversal of the oncogenic Ras phenotype (38), similar to what we previously reported when Tnk1/Kos1 is expressed in Ras transformed cells (3).Disruption of the Grb2-Sos1 complex occurs as a result of Grb2 phosphorylation at Y 209 when Bcr/Abl or EGFR is overexpressed (29).In addition, Grb2 contains other tyrosine residues that may be phosphorylated.Interestingly, Tnk1/Kos1 can phosphorylate Grb2 on more than one site because the Grb2 209 (YÀ F) mutant is phosphorylated by Tnk1/Kos1 (data not shown).Thus, we conclude that Tnk1/Kos1-mediated tyrosine phosphorylation of Grb2 may promote the functional disruption of the Grb2-Sos1 complex and thereby lead to indirect suppression of Ras activation.
It is now well known that hyperactivation of Ras results in spontaneous tumor development (39,40).The development of spontaneous tumors in the Tnk1 +/À mice is associated with the hyperactivation of Ras.Furthermore, lymphomas that develop in Tnk1 +/À or Tnk1 À/À mice show similar hyperactivated levels of Ras (Fig. 5A).Whereas tumors from Tnk1 +/À mice do not express Kos1, uninvolved tissues from the same mouse continue to express Kos1 (Fig. 5C).Silencing of gene expression in heterozygote tumors can potentially be explained by a functional loss of heterozygosity in the Tnk1 +/À mice through promoter methylation.Whereas the Tnk1/Kos1 promoter does not contain a classical CpG ''island'' it does contain multiple CpG sites, and our findings indicate that all of the CpG sites located in the proximal promoter are methylated in the tumor, but not the adjacent, uninvolved tissue (Fig. 5B).Whereas it is formally possible that a mechanism other than promoter methylation may account for silencing, including deletion of the nontargeted locus, we have not addressed this issue.Additional studies will be required to test this possibility.Therefore, we propose that epigenetic silencing of the Tnk1/ Kos1 promoter by hypermethylation may be a mechanism that accounts for Tnk1/Kos1 inactivation (30)(31)(32).In support of this possibility in human cancer, expression of Tnk1 is reported to be down-regulated/silenced in human leukemia and prostate cancer-derived cell lines, indicating a potential role for the loss of Tnk1/Kos1 activity in human cancer (1,2,33).It will now be important to determine the Tnk1 promoter methylation status in these cell lines.
In summary, inactivation of Tnk1/Kos1 in mice by gene targeting results in spontaneous tumor formation associated with hyperactivation of Ras.The spectrum of malignant phenotypes observed in both Tnk1 À/À and Tnk1 +/À mice indicates that the signaling pathway in which Tnk1/Kos1 functions is required for many cell types and is consistent with the ubiquitous expression pattern of Kos1.Furthermore, tumor development in Tnk1/Kos1 knockout mice mimics transformation induced by oncogenic H-Ras, K-Ras, and N-Ras that frequently occurs in several human tumor types.This suggests that, in addition to oncogenic mutations that can directly activate Ras, there are other indirect mechanisms that, when dysregulated, can also account for the Ras hyperactivation observed in spontaneously formed tumors.Tnk1/Kos1 is an example of a tyrosine kinase-negative regulator of Ras that has tumor suppressor activity.A more structural understanding of the mechanism by which Tnk1/Kos1 can uncouple and inactivate the Grb2-Sos1 complex and Ras activation may point the way in the development of novel cancer therapies.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.

Figure 1 .
Figure 1.Generation of Tnk1/Kos1 null mice.A, schematic diagram of Tnk1/Kos1 wild-type allele, knockout vector, and the resulting knockout allele.Kos1 consists of eight exons, whereas Tnk1 contains an additional six exons, indicated by the boxes (filled boxes indicate coding exons).In the targeting construct, the entire catalytic domain (exons 4-8) was deleted and replaced with the neomycin-resistant cassette (Neo ).The thymidine kinase cassette (TK ) was used for negative selection.Restriction enzyme sites for BamH1 (B ), EcoRI (E), HindIII (H ), KpnI (K), ScaI (S ), SpeI (Sp ), and XbaI (X ) are indicated.Note that Xba1 and BamH1 sites were created in the mutant Kos1 allele by the insertion of the Neo cassette.These sites were used for distinguishing the knockout allele from the wild-type allele by Southern hybridization using the 5 ¶ probe or 3 ¶ probe.B, genotyping of Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À mice by Southern blot analysis.Genomic DNAs isolated from tails of Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À mice were digested by BamH1.Hybridization of the digested DNA with the 3 ¶ probe detected a 7.8-kb wild-type band and a 4.2-kb mutant band.C and D, expression of Kos1 in mice liver and MEFs.Liver lysates prepared from Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À littermates (C ) or cell lysates prepared from Tnk1 +/+ , Tnk1 +/À , and Tnk1 À/À MEFs (D ) were resolved on a 10% SDS-PAGE, and Kos1 expression was analyzed by Western blotting with anti-Kos1 antiserum.Erk2 expression level was used as an internal control for equal loading in each lane in both cases.

Figure 2 .
Figure2.Histology of spontaneous neoplasms developed in Tnk1/Kos1 homozygous mice (A-K) and heterozygous mice (L-P ).A, lymphoma, small intestine (400Â).H&E sections of a highly proliferating and invasive lymphoid tissue that has completely effaced the muscularis layers of small intestine.Inset shows the affected tissue.B, immunostaining of lymphoma using B220 (brown staining, inset ) and CD3 (blue staining, inset ).The B cells, intensely stained with B220 antibody, are pleomorphic with a significant level of necrosis.The small population of T cells that showed staining with CD3 antibody seems normal.C, lymphoma, mesenteric lymph node (400Â).The large and highly proliferating nodules in the mesenteric lymph node completely disrupted the normal cortex and medullary architecture of the lymph node, infiltrated the subcapsular and medullary sinuses, and have invaded the surrounding fatty connective tissue.The section shows large areas of necrosis, and the large blast cells that stain for B220 (not shown) is predominant in most areas.Inset shows the lymphoma of the affected tissue.D, histiocytic lymphoma in salivary gland (400Â).Neoplastic population of lymphocytes from the adjacent submandibular lymph node (not shown) infiltrated the salivary gland.In many areas, multinucleated giant cells are scattered among the pleomorphic population of lymphocytes.Inset shows the lymphoma of the affected tissue.E, trichoepithelioma (400Â).The mass within the dermis and subcutaneous tissue contains well-differentiated epithelial cells forming several components of the follicle, including hair bulb, hair, and follicular wall.F, spleen hemangiosarcoma (400Â).Neoplastic endothelial cells form variably sized blood spaces that occupy the red pulp and irregularly infiltrated the adjacent areas.G, lymphoma, spleen (400Â).Enlarged discrete nodules sometimes coalesce and largely disrupt the white pulp architecture.A mixed pleomorphic population of lymphocytes composed of follicular center cells, centrocytes, and small lymphocytes form the nodules.H, thymic lymphoma (400Â).Malignant lymphocytes have completely destroyed the architecture of the thymus.Inset shows tumor in the thymus.I, carcinoma of the pancreas.J and K, bronchio alveolar adenocarcinoma, lung (400Â).The tumor composed of a mildly pleomorphic population of neoplastic cells infiltrated the parenchyma.Inset shows tumor in the lung.L and M, adenocarcinoma, small intestine (400Â).Neoplastic cells frequently forming glands invade the layers of intestine, including the muscularis.Dense schirrous reaction surrounds isolated glands within the intestinal wall.Inset shows the tumor.N and O, hepatocellular carcinoma (400Â).A large portion of the liver is invaded by the moderately anaplastic malignant hepatocytes showing significant pleomorphism.Neoplastic cells have eosinophilic, often mildly vacuolated cytoplasm.Inset shows the tumor.P, pancreatic carcinoma (400Â).Inset shows the tumor.The scale bar shown in A is applicable to all.A, C, D, E, F, G, H, I, K, L, N and P are at 50 AM, whereas B, J, M and O are at 500 AM.