View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Cell, Vol. 112, 673–684, March 7, 2003, Copyright 2003 by Cell Press , a GTPase Activator, Is Disrupted during Tumorigenesis

Vijay Yajnik1,2 Charles Paulding,1 coincidence of two different allelic deletions, with the Raffaella Sordella,1 Andrea I. McClatchey,1 size of the overlapping homozygous deletion restricted Mako Saito,1 Doke C.R. Wahrer,1 by the presence of essential genes. In 1993, Lisitsyn Paul Reynolds,1 Daphne W. Bell,1 and Wigler first proposed representational difference Robert Lake,1 Sander van den Heuvel,1 analysis (RDA) as a screening method for identifying Jeff Settleman,1 and Daniel A. Haber1,* homozygous deletions in genomes of tumor cells. The 1Massachusetts General Hospital Cancer Center technique combines subtractive hybridization with dif- 2 Gastrointestinal Unit ferential PCR-amplification to isolate DNA fragments Harvard Medical School absent in tumor cells compared with matched normal Charlestown, Massachusetts 02129 tissues (Lisitsyn et al., 1993, 1994; Lisitsyn 1995; Lisitsyn and Wigler, 1995). The application of RDA to cancer gene discovery was instrumental in the initial identification of the BRCA2 and PTEN tumor suppressor genes; both of Summary which were found to reside within homozygous genomic deletions in human tumor specimens (Schutte et al., We used representational difference analysis to iden- 1995; Li et al., 1997). tify homozygous genomic deletions selected during A limitation to the broad application of RDA in human tumor progression in the mouse NF2 and TP53 tumor cancer gene discovery appears to be the relatively high model. We describe a deletion targeting DOCK4,a prevalence of silent polymorphic microdeletions in the member of the CDM gene family encoding regulators . Incidental heterozygous germline dele- of small GTPases. DOCK4 specifically activates Rap tions may become homozygous in tumor specimens GTPase, enhancing the formation of adherens junc- based on their fortuitous location at sites of somatic tions. DOCK4 mutations are present in a subset of allelic loss (LOH), leading to their isolation in an RDA human cancer cell lines; a recurrent missense mutant screen (C.P. and D.A.H., unpublished data). To circum- identified in human prostate and ovarian cancers en- vent this problem, we explored the application of RDA codes a that is defective in Rap1 activation. to syngeneic mouse tumor models. We reasoned that The engulfment defect of C. elegans mutants lacking the absence of polymorphisms in such mice would the CDM gene ced-5 is rescued by wild-type DOCK4, greatly increase the identification of biologically signifi- but not by the mutant allele. Expression of wild-type, cant homozygous deletions, and that genes targeted but not mutant, DOCK4 in mouse osteosarcoma cells during cancer progression in mouse models would likely with a deletion of the endogenous gene suppresses have similar roles in human tumorigenesis (McClatchey growth in soft agar and tumor invasion in vivo. DOCK4 and Jacks, 1998). Following an initiating genetic lesion, therefore encodes a CDM family member that regu- malignant transformation in mouse models follows a lates intercellular junctions and is disrupted during series of stages that are comparable to those observed tumorigenesis. in human cancers. For instance, in the well characterized Rip-TAG model, expression of SV40T antigen initiates a hyperproliferative state which progresses to frank ma- Introduction lignancy, associated with loss of E cadherin expression, and induction of insulin-like growth factor 2 and matrix Allelic losses resulting in inactivation of tumor suppres- metalloproteinases (Hanahan and Weinberg, 2000). sor genes are thought to contribute to both the initiation A particularly attractive model in which to study tumor and progression of human cancer (Cavenee et al., 1983; progression is the TP53ϩ/Ϫ NF2ϩ/Ϫmouse, which repro- Fearon, 1997). Tumor suppressors implicated in tumor ducibly gives rise to tumors with high metastatic poten- initiation have been most readily identified by analysis tial and from which established cell lines are readily of cancer pedigrees, in which inheritance of a mutant generated (McClatchey et al., 1998). The use of tumor allele segregates with cancer predisposition. To date, cell lines, free of contaminating normal cells, greatly few genes have been isolated by virtue of their inactiva- enhances the efficiency of genomic subtraction in RDA. tion during cancer progression. Tumor loss of heterozy- The most common tumor cell types in this mouse model gosity at polymorphic markers (LOH) typically results are fibrosarcomas and osteosarcomas, virtually all of from somatic chromosomal deletion or non-disjunction which demonstrate homozygous inactivation of TP53 events that render a cancer cell homozygous for a mu- and NF2 genes, residing on mouse 11. tant allele. However, the relatively large size of hemizy- Loss of TP53, the most common genetic event in human gous allelic losses in cancer cells complicates efficient cancer, leads to genomic instability and may increase gene discovery strategies; homozygous genomic dele- the likelihood of subsequent genetic alterations includ- tions are rare, but their smaller size facilitates identifica- ing specific chromosomal deletions (Jain et al., 2001). tion of the targeted tumor suppressor gene (Dryja et al., NF2, the gene responsible for Neurofibromatosis type 1986; Call et al., 1990; Kamb et al., 1994; Li et al., 1997; II, encodes a member of the Moesin, Ezrin, and Radixin Steck et al., 1997). Such deletions may result from the (ERM) family of structural and is commonly mutated in human mesotheliomas (Sekido et al., 1995). *Correspondence: [email protected] While inactivation of these two genes initiates malignant Cell 674

proliferation in this mouse tumor model, it is likely that the outside limits of the homozygous deletion. Remark- additional genetic events, including homozygous dele- ably, this is an extremely “gene-poor” locus, containing tions, may contribute to the highly metastatic tumor only one transcription unit identified by gene prediction phenotype. programs in both mouse and humans. Here, we identify a gene, DOCK4, targeted by a homo- A single human transcript comprised of 53 exons was zygous deletion in an osteosarcoma cell line derived predicted, spanning a 500 kb locus between ZNF277 from the TP53-NF2 mouse model. DOCK4 encodes a and NLRR3, with the most 5Ј exon present in human member of the CDM gene family, implicated in the regu- BAC AC 003077 and the most 3Ј exon and untranslated lation of Rac GTPase signaling and defined by its found- region within BAC AC 005047. This putative transcript ing members, C. elegans ced-5, vertebrate , was confirmed by analysis of the human EST database, and Drosophila Myoblast City (Hasegawa et al., 1996; which revealed several matches from multiple tissue Nolan et al. 1998; Wu and Horvitz, 1998). DOCK4 acti- types. Reverse transcription (RT)-PCR was used to am- vates Rap GTPase, regulating the formation of intercel- plify the predicted exons in overlapping fragments, con- lular adherens junctions. Reconstitution of DOCK4 ex- firming their presence within a single transcript. We used pression in mouse osteosarcoma cells with a deletion 5Ј-RACE-PCR to generate multiple clones from a pla- of the endogenous gene results in reestablishment of cental cDNA library, identifying the transcriptional start cell-cell junctions and suppression of soft agar colony within an upstream BAC AC 004111.1. The large geno- formation and tumor invasion in vivo. Transgenic ex- mic locus spanned by this gene is in part due to intron pression of DOCK4 in C. elegans ced-5 null mutants 1, covering an entire BAC (AC 004001.1). Of note, marker corrects their defect in the engulfment of apoptotic cell D7S523, which defines a common site of LOH in human bodies. A subset of human cancer cell lines harbor muta- prostate cancers, is within this BAC, raising the possibil- tions in DOCK4; a Pro1718Leu mutation found in both ity that this gene may be a target for allelic losses in prostate and ovarian cancers abrogates Rap1 GTPase human cancers (Takahashi et al., 1995; Koike et al., activation, its correction of the Ced-5 phenotype in C. 1997). Taken together, these studies allowed assembly elegans, and its tumor suppressor properties in mouse of the full-length coding region for a previously unchar- osteosarcoma cells. acterized gene, encoding a protein of 1966 amino acids with extensive homology to DOCK180, a regulator of Results the small GTPase Rac (Hasegawa et al., 1996, Brugnera et al., 2002). In vertebrates, DOCK gene family members Isolation of a Homozygous Deletion in 3081 include DOCK180, its -specific homolog Osteosarcoma Cells (Nishihara et al., 1999), and an EST sequence We used RDA to screen sarcomas derived from six (KIAA0299) initially deposited in the database as TP53ϩ/Ϫ NF2ϩ/Ϫ mice for the presence of homozygous DOCK3. DOCK3 has recently been identified as a pre- deletions. Genomic DNA from tumor-derived cell lines senilin binding protein (Kashiwa et al., 2000) and as a was PCR-amplified as the “driver” and used in excess “modifier of cell adhesion” (MOCA) in neuronal cells to subtract shared sequences from the matched normal (Chen et al., 2002). We have therefore named the gene DNA “tester” (Lisitsyn and Wigler, 1995). Differentially DOCK4. DOCK4 is expressed in multiple tissue types, amplified genomic sequences were obtained in 3/6 tu- with highest levels in skeletal muscle, prostate, and mors, all of which were confirmed as being homozy- ovary (Figure 1B). gously deleted by Southern blotting in the tumor cell line of origin. Fibrosarcoma 3452 was found to have a DOCK4, a CDM Family Member deletion of the p16INK4a/p19ARF locus, a common site of The overall structure of DOCK4 is highly similar to other chromosomal deletions in both mouse and human can- CDM family members, with an N-terminal SH3 domain, cers. Fibrosarcoma 3872 contained multiple large dele- a region of extended homology from amino acids 100 tions of the Y chromosome, an observation that is also to 1700 (35% amino acid identity with DOCK180, 39% frequently observed in cancers derived from males (Lis- with DOCK2, and 54% with DOCK3), and a C-terminal itsyn and Wigler, 1995). Osteosarcoma 3081 produced proline-rich region, which appears to be unique to each two homozygously deleted genomic fragments, mapped family member (Figure 2). The CDM protein family has using radiation hybrid screening near marker D12Mit148 recently been divided into two classes based on se- on mouse chromosome 12, a genomic locus that is syn- quence alignment programs (Meller et al., 2002; Reif and tenic to human chromosome 7q31 (Figure 1A). The RDA Cyster, 2002; Cote and Vuori, 2002). Class A includes products were further mapped to two adjacent bacterial mammalian DOCK180 and DOCK2, and Drosophila Myo- artificial chromosome (BAC) clones, AC 079370 and AC blast City, while class B includes mammalian DOCK3 079369. While the mouse genomic locus was unordered, and DOCK4, and an unnamed Drosophila protein, the syntenic human locus was completed and showed CG31048. DOCK class A proteins appear to function as highly conserved coding sequences that were used to upstream regulators of Rac GTPase signaling, and their define the extent of the murine deletion. Flanking the inactivation is associated with dramatic defects in cell telomeric side of D12Mit148 is the zinc finger gene, migration and developmental patterning (Hasegawa et ZNF277 (Liang et al., 2000), which is not deleted in tumor al., 1996; Nolan et al., 1998; Wu and Horvitz, 1998; Fukui 3081 (Figure 1A). On the centromeric side, the Neuronal- et al., 2001). The signaling pathway regulated by class specific leucine-rich repeat gene, NLRR3, is also present B proteins has not been defined. Interestingly, only a in tumor 3081. These two genes are separated by 900 single CDM family member, ced-5, exists in C. elegans. kb in the human genomic sequence and hence define ced-5 was initially identified by its role in both the en- GTPases Regulator DOCK4 Disrupted in Tumorigenesis 675

Figure 1. Identification of DOCK4 within a Homozygous Genomic Deletion in Osteosarcoma 3081 (A) Schematic representation of the homozygous deletion on mouse chromosome 12 containing RDA products clone 11 and clone 13. This region is a representation of the linear sequence of mouse chromosome 12 from Ensembl database (nucleotides 12.345 ϫ 105 to 12.354 ϫ 106 ). Southern blot hybridization of these deleted RDA products, as well as markers from the flanking non-deleted genes ZNF277 and NLRR3, are shown for tumor 3081 and two unrelated controls 3442 and 3978. The gene targeted by the deletion, termed DOCK4, is comprised of 53 exons, spanning 500 kb of genomic DNA (Ensembl database 12.346 ϫ 105 to 12.351 ϫ 106 ). No other identifiable transcription unit is present between ZNF277 and NLRR3 in either mouse or human sequence. Human marker D7S523, frequently targeted by allelic loss affecting chromosome 7q31 in human cancers, maps to the first intron of DOCK4. (B) Northern blot demonstrating expression of the 8.5 kb DOCK4 transcript in multiple human tissues: lanes 1, heart; 2, brain; 3, placenta; 4, lung; 5, liver; 6, skeletal muscle; 7, kidney; 8, pancreas; 9, spleen; 10, ; 11, prostate; 12, testis; 13, ovary; 14, small intestine; 15, colon; and 16, peripheral blood leukocytes. GAPDH, loading control. gulfment of apoptotic cells and the migration of the While Zizimin1 is not a CDM family member, two discrete distal tip cell of the gonad. These processes require domains of 250 and 550 amino acids, respectively, show sequence identity with DOCK180 and have been %20ف cytoskeletal reorganization and may involve distinct GTPase-dependent signals (Wu and Horvitz, 1998; Red- termed either the DOCK homology regions (DHR1 and dien and Horvitz, 2000). DHR2 or “docker”) or the conserved Zizimin homology The functional properties of CDM proteins have been (CZH1 and CZH2) (Cote and Vuori, 2002; Meller et al., most clearly defined for DOCK180, which is activated 2002). DHR2/CZH2 binds to nucleotide-free GTPase, by integrin signaling, leading to an interaction between while the function of DHR1/CZH1 is unclear. DOCK4 its C-terminal proline-rich domain and the SH3 domain contains both of these conserved domains, raising the of the adaptor protein CrkII (Albert et al., 2000). The SH2 possibility that it may also be an unconventional nucleo- domain of CrkII in turn binds to the scaffold protein p130 tide exchange factor (Figure 2). In addition to these do- CAS, which is phosphorylated by Src and focal adhesion mains, two other motifs are present only in DOCK4 and kinase (FAK), recruiting the exchange factor Rac-GEF CED-5: a consensus binding site for the SH3 domain of and triggering the cytoskeletal changes characteristic Src (SSB) and a second C-terminal proline-rich motif of Rac activation (Kiyokawa et al., 1998a and 1998b; (Figure 2). These motifs raise the possibility that some Gu et al. 2001). The proline-rich C-terminal domain of functional properties of CED-5, the unique CDM family DOCK4 may therefore play a similar role in Crk II binding member in C. elegans, may be shared with DOCK4, but and GTPase signaling (Figure 2; see below). An alterna- not DOCK180. In summary, analysis of the conserved tive pathway for GTPase activation by DOCK180 has domains of DOCK4 suggests that it may play a role in recently been uncovered with the observation that it can the regulation of small GTPases implicated in cellular bind directly to nucleotide-free Rac and interact with motility, cell adhesion, and invasion. ELMO to mediate nucleotide exchange, despite lacking the characteristic Dbl homology domain implicated in DOCK4 Mutations in Human Cancer Cell Lines guanine nucleotide exchange (Brugnera et al., 2002). Marker D7S523, at chromosomal locus 7q31, demon- DOCK180 has been called an “unconventional nucleo- strates a high frequency of allelic loss (LOH) in human tide exchange factor”, along with other proteins such breast, ovarian, and prostate cancers, and in gliomas as the CDC42 activator Zizimin1 (Meller et al., 2002). (Takahashi et al., 1995; Koike et al., 1997). A candidate Cell 676

Figure 2. Schematic Representation of Domains Conserved between DOCK4 and other CDM Family Members and Targeted by Mutations in Human Cancers Alignment of DOCK4 and the founding members of the CDM protein family: C. elegans CED-5, mammalian DOCK180, and Drosophila Myoblast City (MBC). All four proteins share an N-terminal SH3 domain. The CDM domain of extended homology between amino acids 100 to 1700 in DOCK4 contains the Lys1059Thr mutation, a residue conserved in all four family members. Within the C-terminal proline-rich domain shared by all CDM family members are three conserved regions: C-terminal motif-1, shared only by DOCK4 and CED-5, is disrupted by the Pro1718Leu mutation; a highly conserved consensus binding site for the SH3 domain of Src is also shared only by DOCK4 and CED-5; and C-terminal motif-2, shared by DOCK4 and DOCK180, which is targeted by the Val1884Met mutation. DHR1/CZH1 and DHR2/CZH2 domains are present in unconventional vertebrate GEFs and have been implicated in nucleotide free GTPase binding. tumor suppressor, ST7, has been identified at that locus, other missense change, Val1884Met, is encoded by a but recent studies have not supported the initial report of heterozygous mutation in two independent glioma cell frequent intragenic mutations, indicating that additional lines and maps to another conserved domain shared tumor suppressor genes are likely to reside at this locus with DOCK180 (C-terminal motif-2; Figure 2). A third (Zenklusen et al., 2001; Thomas et al., 2001). As an initial heterozygous mutation present in a colorectal cancer screen for DOCK4 mutations in human tumors, we se- cell line, encoding Lys1059Thr, affects a residue within quenced the 6 kb coding sequence from 44 cancer cell the extended homology domain that is conserved in all lines, representing a broad range of tumor types. RT- CDM family members (Figure 2). We selected Pro1718- PCR products were subjected to direct sequencing, and Leu as a human cancer-derived mutation in DOCK4 and any sequence variants identified by cDNA analysis were compared its functional properties with those of the confirmed by amplification of genomic DNA. DNA speci- wild-type protein. mens from 200 healthy individuals (400 ) were used to exclude polymorphisms that might be Rescue of the Engulfment Defect in C. elegans present in the population. Analysis of the DOCK4 coding ced-5 Mutants by Wild-Type sequence revealed five missense changes that were but Not Mutant DOCK4 absent in controls (Table 1). Among these, Pro1718Leu The high degree of evolutionary conservation in CDM was found as a homozygous mutation in two indepen- proteins made it possible to use a well-characterized dent cell lines, one derived from ovarian cancer and functional assay in C. elegans to test the properties of the other from prostate cancer. The targeted proline DOCK4 and the tumor-associated mutant Pro1718Leu. residue, within the C-terminal proline-rich domain, is While distinct CDM proteins may mediate activation of conserved in CED-5 (C-terminal motif-1; Figure 2). An- different GTPases in vertebrates, in C. elegans these

Table 1. hDOCK4 Mutations in Human Cancer Cell Lines Cancer Cell Line Nucleotide Change and Position Effect on Protein Allele Frequency OV1063 (Ovarian) C to T at nt 5153 (Homozygous) Missense (Pro1718Leu) 0/380 DU145 (Prostate) G to A at nt 5650 (Heterozygous) Missense (Val1884Met) 0/374 U251 (CNS) C to T at nt 260 (Heterozygous) Missense (THr87Ile) 0/380 SNB19 (CNS) A to C at nt 3176 (Heterozygous) Missense (Lys1059Thr) 0/374 HCT15 (Colorectal) T to C at nt 5263 (Heterozygous) Missense (Ser1755Pro) 0/380

Polymorphisms C to G at nt 5197 Missense (Pro1733Ala) 2/268 C to T at nt 5777 Missense (Ser1926Leu) 4/126 C to T at nt 5750 Missense (Pro1917Leu) 2/126 G to A at nt 5740 Missense (Val1914Ile) 8/88 G to C at nt 1816 Missense (Glu606Gln) 7/76 GTPases Regulator DOCK4 Disrupted in Tumorigenesis 677

Activation of Rap GTPase by DOCK4 We generated CMV-driven, flag-tagged expression con- structs encoding wild-type DOCK4 or the Pro1718Leu mutant. Transient transfection into 293T cells confirmed expression of the predicted 225 kDa protein, while ex- pression in Caco-2 cells clearly demonstrated localiza- tion to the (Figures 4A and 4B). We there- fore tested the functional properties of wild-type DOCK4 without the need for membrane targeting using a syn- thetic CAAX box (Hasegawa et al., 1996). Given the ho- mology between the proline-rich C-terminal domains of DOCK4 and DOCK180, we first tested whether DOCK4 also interacts with the N-terminal SH3 domain of the adaptor protein CrkII. Indeed, transfected DOCK4 pro- tein is readily coprecipitated from 293T cell lysates with the GST fused N-terminal SH3 domain of CrkII, whereas no binding is observed with other SH3 domains, includ- ing the C terminus of Crk II or the SH3 domain of Abl (Figure 4B). Of note, the Pro1718Leu DOCK4 mutant demonstrates significantly reduced binding to CrkII (6ϫ lysate was used to demonstrate reduced binding by the Figure 3. Rescue of Apoptotic Body Engulfment Defect in ced-5 mutant protein). While the mutation does not directly Mutants by Expression of Wild-Type but not Mutant DOCK4 affect the binding site for CrkII, we presume that loss Nomarski photomicrographs showing the reflexed region of (A) wild- of the conserved proline residue leads to altered protein type adult gonad, and (B) the same loop region in a ced-5(n1812) folding. mutant. Arrows indicate multiple cell corpses. The association between DOCK4 and CrkII is consis- (C) Expression of heat shock-driven ced-5 or DOCK4 rescued the tent with its function within a signaling complex analo- cell corpse engulfment defect in the germline of ced-5(n1812) mu- gous to that of DOCK180. The DOCK180 protein com- tants. The number of cell corpses observed within a single morpho- plex has been implicated in Rac signaling, and we logically normal gonad arm were counted in heat-shocked trans- genic animals. Animals expressing Pro1718Leu-DOCK4 or a control therefore screened a panel of small GTPases to identify transgene (GFP) did not show rescue. The DTC migration defect the most relevant physiological target for DOCK4. 293T of ced-5(n1812) mutants was corrected by ced-5 but not DOCK4 cells were transiently transfected with constructs en- expression. The percent of animals displaying a migration defect of coding either DOCK4 or the Pro1718Leu mutation, along either DTC was determined in adult transgenic animals. with tagged constructs encoding either Rac, Rho, CDC42, or Rap. Cellular lysates were quantified for GTP bound forms by binding to affinity matrices, followed by properties are encoded by a single gene, ced-5. Inacti- Western blotting using antibody against the transfected vation of ced-5 is associated with two primary defects: GTPases. Dramatic activation of Rap GTPase (Ͼ 5-fold) failure of neighboring cells to engulf apoptotic cells lead- is evident in cells transfected with wild-type DOCK4, ing to the persistence of cell corpses and abnormal but not the Pro1718Leu mutant (Figures 4C and 4D). For migration of the gonadal distal tip cells (DTCs) (Wu and comparison, 293T cells were also transfected with either Horvitz, 1998). In a dramatic illustration of the conserva- wild-type Rap or the constitutively active form, Rap63E tion of this signaling pathway, expression of a heat (Figure 4D). Transient expression of DOCK4 had no ef- shock-driven DOCK180 construct has been shown to fect on other GTPases, including Rho, Rac, and CDC42 rescue the DTC migration defect in ced-5 mutants (Wu (Figure 4C). Remarkably however, the Pro1718Leu mu- and Horvitz, 1998). However, the mechanism underlying tant of DOCK4 shows significant activation of both Rac engulfment of apoptotic corpses has not been defined. and CDC42. This tumor-associated mutation, therefore We therefore tested whether heat shock-driven expres- exhibits both loss-of-function as well as gain-of-function sion of DOCK4 could rescue this component of the properties with respect to the activation different small Ced-5 phenotype. Remarkably, expression of wild-type GTPases. Given the failure of Pro1718Leu-DOCK4 to DOCK4 in C. elegans corrects the engulfment defect in bind CrkII (Figure 4B), its ability to activate Rac and ced-5 mutants, while expression of Pro1718Leu-DOCK4 CDC42-GTPases may result from its association with has no effect (Figure 3). This rescue was most readily another adaptor protein, from altered interactions in- demonstrated in the germline, while apoptotic corpses volving its DHR/CZH domains, or from indirect effects in somatic lineages were less affected by ectopic on small GTPase pathways. DOCK4 expression (Figure 3 and data not shown). Ex- pression of DOCK4 has no effect on DTC migration. Restoration of Adherens Junctions Taken together with the results of Wu and Horvitz (1998), by DOCK4 and Rap1 our results suggest that DOCK4 and DOCK180 together The original isolation of DOCK4 from a mouse osteosar- reconstitute the functional properties of CED-5. More- coma cell line with a deletion of the endogenous gene over, the different results obtained with wild-type and made it possible to test the effects of reconstituting Pro1718Leu mutant forms of DOCK4 indicate that the its expression in these cells. We therefore generated missense mutation indeed affects protein function. multiple cell lines stably expressing DOCK4 or Pro1718- Cell 678

Figure 4. Activation of Rap GTPase by DOCK4 Expression of constructs encoding either wild-type DOCK4 or the Pro1718Leu mutation in transient (A–D) and stable (E–F) transfection assays. (A) Cell surface localization of epitope-tagged DOCK4 following transfection into Caco-2 cells. (B) Binding of wild-type DOCK4 to the N-terminal SH3 domain of CrkII (GST-Crk N), but not to the C-terminal SH3 domain of CrkII (GST-Crk C) or to the SH3 domain of Abl (GST-Abl SH3). The Pro1718Leu mutant exhibits reduced binding to GST-Crk N, which is only evident with addition of 6 ϫ cellular lysate. Comparable expression of wild-type and mutant DOCK4 constructs is shown by Western blotting (lysate), along with equivalent expression of the bacterially synthesized GST-SH3 domains. (C) Activation of small GTPases following transient transfection of wild-type DOCK4 and the Pro1718Leu mutant into 293T cells. These expression plasmids were cotransfected along with constructs encoding Rap, HA-Rac, HA-CDC42 or HA-Rho, and active GTP bound forms were isolated from cellular lysates by incubation with appropriate GST-RBD constructs, followed by Western blotting using antibody to either Rap or the HA epitope tag (GTP bound). Western blotting of cellular lysates (total) shows comparable expression of the small GTPases. Activation of Rap GTPase is observed with expression of wild-type DOCK4, but not Pro1718Leu-DOCK4. In contrast, expression of Pro1718Leu- DOCK4, but not wild-type DOCK4, is associated with activation of Rac and CDC42. Rho GTP levels are unaffected by either variants of DOCK4. (D) Quantitative analysis of Rap activation by wild-type DOCK4, compared with the Pro1718Leu mutant. For comparison, Rap GTPase levels are shown for cells transiently transfected with either wild-type Rap or the constitutively active mutant Rap63E. (E) Western blot analysis demonstrating expression of epitope-tagged DOCK4 or the Pro1718Leu mutant in two independent, stably transfected clones derived from 3081 mouse osteosarcoma cells with a homozygous deletion of the endogenous gene. (F) Phalloidin staining to demonstrate cellular actin patterns of 3081 cells stably reconstituted with either wild-type DOCK4 or the Pro1718Leu mutant. Compared with vector-transfected cells, DOCK4 expressing cells have increased actin stress fibers, while cells expressing the Pro1718Leu mutant demonstrate presence of filopodia, a characteristic of CDC42 activation. All cells were in mid log phase of growth.

Leu-DOCK4 in 3081 osteosarcoma cells (Figure 4E). For such intercellular contact points has been linked with all subsequent experiments, multiple cell lines were cancer progression (Christofori and Semb, 1999; Hajra used to ensure against the effects of clonal selection. and Fearon, 2002). We therefore tested 3081 parental The morphology of cells expressing these constructs cells and their reconstituted derivatives for expression differs remarkably: compared with parental 3081 cells, of ␤-catenin, a marker for adherens junctions. Osteosar- those expressing wild-type DOCK4 have a flattened coma cell line 3081 cells do not display contact inhibition morphology and prominent actin stress fibers, while and fail to form adherens junctions, as demonstrated those transfected with Pro1718Leu-DOCK4 have filo- by a diffuse staining pattern for ␤-catenin (Figure 5A). podia, a structural characteristic of CDC42 activation In marked contrast, 3081 cells expressing wild-type (Figure 4F). DOCK4 grow to a lower cell density at confluence (Figure While Rap1 was initially identified as a Ras antagonist 6A), demonstrating marked contact inhibition. ␤-catenin thought to sequester Ras effectors within an inactive staining shows characteristic localization to discrete re- complex, more recent data have raised the possibility gions of the membrane representing adherens junctions that Rap may itself mediate distinct signaling pathways (Figure 5B). No such effect is seen in 3081 cells express- (Bos et al., 2001). In the developing Drosophila embryo, ing Pro1718Leu-DOCK4 (Figure 5C). Rap1 plays an important role in formation of cell-cell To test whether Rap GTPase activation is required adherens junctions (Knox and Brown, 2002). Loss of for the restoration of adherens junctions by wild-type GTPases Regulator DOCK4 Disrupted in Tumorigenesis 679

Figure 5. Regulation of Cellular Adherens Junctions by Expression of DOCK4 and Rap GTPase Staining of confluent cells 3081 osteosar- coma cells (A–E) and primary mouse osteo- blasts (G–I), using antibody to ␤-catenin, a constituent of cellular adherens junctions. (A) Adherens junctions are not visualized in the parental DOCK4 null 3081 osteosarcoma cells. (B) Reconstitution of wild-type DOCK4 ex- pression in these cells is associated with ap- pearance of intercellular adherens junctions. (C) No such effect is seen in 3081 cells ex- pressing the Pro1718Leu-DOCK4 mutation. (D) Coexpression of a dominant-negative variant, RapN17, abrogates the induction of adherens junctions by wild-type DOCK4 in 3081 cells. (E) Transfection of parental 3081 cells with the constitutively active Rap variant, Rap63E, is sufficient to restore formation of adherens junctions. (F) siRNA to the DOCK4 mRNA is effective in suppressing the endogenous transcript in primary mouse osteoblasts, as demonstrated by RT-PCR (35 cycles). No effect is seen with a nonspecific siRNA (targeting NY-REN-60 mRNA), and an irrelevant transcript (p16) is unaffected by siRNA treatment. (G) Normal primary mouse osteoblasts dem- onstrate presence of adherens junctions. (H) Treatment of primary osteoblasts with siRNA targeting endogenous DOCK4 for 48 hr leads to a reduction in cellular adherens junction formation. (I) siRNA targeting the nonspecific gene has no effect on adherens junctions.

DOCK4, 3081 cells expressing wild-type DOCK4 were definitive cadherens junctions (Figure 5G). Treatment of stably cotransfected with a dominant-negative Rap1 cells with DOCK4 siRNAi for 48 hr effectively reduces of %5ف construct (RapN17). Disruption of Rap1 signaling in mul- the levels of endogenous DOCK4 mRNA to tiple clones abrogated the effect of DOCK4 on ␤-catenin baseline (Figure 5F). Remarkably, this treatment is asso- expression (Figure 5D). To determine if activation of Rap ciated with marked suppression of adherens junctions itself is sufficient to enhance formation of adherens junc- (Figure 5H). No such effect was observed in primary tions in parental 3081 cells, these cells were transfected osteoblasts treated with control oligonucleotides (Fig- with a constitutively active variant of Rap1, Rap63E. In ure 5I). The effect of DOCK4 on cell junctions therefore multiple stable cell lines, expression of Rap63E in the appears to be independent of NF2 function. absence of DOCK4 was sufficient to induce the forma- tion of adherens junctions (Figure 5E). Thus, reconstitu- Tumor Suppressor Properties of DOCK4 tion of DOCK4 expression in 3081 cells appears to re- To explore the physiological significance of DOCK4- store cell-cell contacts through its activation of Rap mediated signaling, we examined the growth properties GTPase. We note that the low levels of endogenous of 3081 osteosarcoma cells expressing either wild-type Rap1 in 3081 cells prevented direct measurement of or Pro718Leu-DOCK4. The restoration of contact inhibi- altered Rap-GTP levels in cells reconstituted with wild- tion in 3081 cells expressing wild-type DOCK4 was asso- type DOCK4. However, the potent activation of Rap1 by ciated with a reduction in maximal cell density, without transient transfection of DOCK4 in 293T cells (Figure affecting generation time in culture (Figure 6A). Colony 4C), coupled with the effect of Rap mutants in stably formation in soft agar, an in vitro correlate of tumorigene- transfected 3081 cells (Figures 5D and 5E), suggest that sis was markedly reduced for DOCK4-reconstituted the effect of DOCK4 is likely to be mediated through 3081 cells, whereas expression of Pro1718Leu-DOCK4 Rap-GTP signaling. had little effect (Figure 6B). To further examine the po- In addition to lacking DOCK4 expression, 3081 cells tential tumor suppressor effect of wild-type DOCK4, re- are null for TP53 and NF2. NF2 itself encodes a cytoskel- constituted 3081 cells were inoculated subcutaneously eton-associated protein implicated in GTPase signaling into nude mice. At three weeks, large tumors (Ͼ 2cm (Shaw et al., 2001), raising the possibility that the effects diameter) were evident in 4/4 mice injected with vector- of DOCK4 in these cells may depend on the absence transfected 3081 cells (Figure 6C). Histological analysis of functional NF2. To address this possibility, we used showed frank invasion of subcutaneous tissue, fat, and siRNA to reduce DOCK4 expression in normal mouse muscle (Figure 6D). In contrast, 3081 cells expressing osteoblasts. These primary cells comprise the pre- wild-type DOCK4 produced much smaller nodules at sumed cell of origin of osteosarcomas and demonstrate the site of injection (3–4 mm diameter) in 4/4 mice. Histo- Cell 680

Figure 6. Suppression of Tumorigenicity by Wild-Type DOCK4, but not Pro1718Leu-DOCK4 Neoplastic properties of 3081 cell lines stably reconstituted with either wild-type or mutant DOCK4. For each construct, two independent cell lines were tested in multiple experiments. (A) Quantitation of viable cells following seeding onto plastic substrate. The doubling rate of all cell lines is comparable, but the final confluent cell density of cells expressing wild-type DOCK4 is considerably reduced. At each time point, standard deviation was within 10% of average cell numbers. (B) Anchorage-independent growth of 3081 cells expressing vector, wild-type DOCK4 or the Pro1718Leu-DOCK4 mutation, following plating in soft agar. (C) Tumor formation by reconstituted 3081 osteosarcoma cells, following subcutaneous inoculation into nude mice. For each construct, two independent stably transfected cell lines were tested in duplicate. All injected cell lines gave rise to tumors, but the size of those expressing wild-type DOCK4 was considerably reduced. (D) Hematoxylin and eosin-stained sections of 3081 tumors arising in nude mice (2.5ϫ and 10ϫ magnification). Vector- and Pro1718Leu- DOCK4-expressing cells give rise to very large and invasive tumors, while DOCK4-expressing cells produce small circumscribed tumors. Tumor cells (T) are shown invading surrounding subcutaneous fat (F), as well as metastatic to lymph node (LN). Invasion of surrounding muscle (M) is also evident in Pro1718Leu-DOCK4 expressing tumors. In contrast, DOCK4-reconstituted 3081 cells give rise to noninvasive tumors with well-demarcated margins (black arrows). logically, these tumors appeared well-circumscribed . In contrast, DOCK4 modulates signaling and pseudo-encapsulated, and they failed to show any through the Rap GTPase, enhancing the formation of invasion of surrounding tissues (Figures 6C and 6D). cellular adherens junctions. We show that reconstitution Osteosarcoma cell line 3081 cells expressing Pro1718- of DOCK4 expression in mouse osteosarcoma cells with Leu-DOCK4 produced large invasive tumors, compara- a deletion of the endogenous gene restores the forma- ble to the parental DOCK4 null cell lines. These data tion of adherens junctions, an effect that is suppressed suggest that mutations of DOCK4 in human cancer cells by coexpression of a dominant-negative Rap construct. may be associated with loss of cell junctions and a more Expression of constitutively active Rap1 in these cells invasive phenotype. is sufficient to restore adherens junctions, supporting the link between Rap signaling and formation of intercel- Discussion lular contacts, which has previously been documented in Drosophila (Knox and Brown, 2002). These functional We have used RDA to screen a mouse tumor model for properties of DOCK4 are not restricted to cells derived homozygous genomic deletions, leading to the isolation from p53- and NF2-driven mouse tumor models, since of DOCK4, a gene that is also targeted by intragenic treatment of normal mouse osteoblasts with siRNA tar- mutations in human prostate, ovarian, and potentially geting DOCK4 also demonstrates a reduction in ad- other cancers. DOCK4 encodes a member of the CDM herens junctions. Consistent with the importance of gene family, characterized by their activation of Rac these intercellular contacts in suppressing tumorigenic- GTPase and their essential roles in morphogenesis and ity, DOCK4-reconstituted mouse osteosarcoma cells GTPases Regulator DOCK4 Disrupted in Tumorigenesis 681

show contact inhibition, suppression of growth in soft lian cells and suggests that DOCK4 functions, at least agar, and a reduction in tumor invasiveness in vivo. in part, by maintaining active Rap signaling. Finally, we show that ectopic expression of DOCK4 in Intercellular adherens junctions are formed by homo- C. elegans ced-5 mutants rescues their defect in en- philic interactions between the extracellular domains of gulfment of apoptotic bodies, pointing to a highly con- E-cadherin, linked by their intracellular tail to catenins served evolutionary pathway that mediates intercellular and to the actin (Jamora and Fuchs, 2002). contact. Taken together, these studies illustrate a gen- AF6, a PDZ domain-containing protein present in the eral strategy using mouse tumor models to identify adherens junction, binds to both Rap and the actin cy- genes inactivated during tumor progression and point toskeleton, and may enhance the link between structural to a potentially important role for Rap signaling in sup- components of the adherens junction and small GTPase pressing human cancer. signaling (Boettner et al., 2000). Several lines of evidence The CDM gene family encodes proteins that are criti- support the role of adherens junctions in the invasive cal integrators of extracellular signals leading to small properties of cancer cells (Perl et al., 1998; Perego et GTPase activation. DOCK180, for instance, is recruited al. 2002). Germline mutations in E-cadherin underlie a to the cell membrane and activated following integrin subset of familial gastric cancer with a particularly inva- signaling, leading to formation of a complex including sive phenotype (Guilford et al., 1998). Somatic loss of the adaptor protein CrkII, the scaffold protein P130CAS E-cadherin expression, either through mutations or pro- (Kiyokawa et al., 1998a, 1998b). DOCK2, a CDM family moter methylation, has been linked to progression of member whose expression is restricted to , tumorigenesis, with restoration of expression leading to also appears to regulate Rac signaling and modulate tumor suppression (Christofori and Semb, 1999). An- cellular migration (Fukui et al., 2001). Both DOCK180 other component of the adherens junction, ␤-catenin, and DOCK2 share extensive homology with DOCK4, al- serves as a useful histological marker for adherens junc- though they diverge at the C-terminal 250 amino acids, tions. While it is frequently mutated in human cancer, a domain that may mediate binding to distinct adaptor these mutations are thought to function directly in the proteins and lead to the activation of different subsets , rather than mediating alterations of small GTPases. The high degree of evolutionary con- in adherens junctions. To date, unlike Ras, mutations servation of CDM family members also provides impor- have not been reported in Rap1. Altered expression of tant insight into their respective functions. Only one fam- the GTPase RhoC has recently been linked to metastasis ily member, ced-5, is present in C. elegans, and its loss in a melanoma model (Clark et al., 2000). Taken together, of function involves apparent defects in both cellular mutational and functional studies suggest that loss of migration (distal tip cell migration) and intercellular ad- adherens junction components are likely to contribute hesion (engulfment of apoptotic corpses by neighboring to invasive properties of tumor cells. Support for the cells). Remarkably, CED-5 and DOCK4 share a C-ter- role of small GTPases in this process is derived from minal domain that is absent in DOCK180 and DOCK2, the presence of DOCK4 mutations in human cancers. including a conserved proline residue that is targeted Analysis of human cancer cell lines identified five dis- by the Pro1718Leu mutation, leading to loss of CrkII tinct missense mutations in DOCK4 that were not detect- binding and Rap activation. Based on its sequence con- able in nearly 400 control chromosomes. Of particular servation, CED-5 may therefore integrate signaling path- interest is the Pro1718Leu mutation that was detected ways, which in vertebrates, are regulated by distinct in both a prostate and an ovarian cancer cell line. The CDM genes. fact that this mutation arose independently in two can- cer cell lines raises the possibility of a specific gain-of- Rap was first identified in a screen for suppressors function mechanism. Similarly, two independent brain of K-Ras-mediated cellular transformation (Kitayama et tumors were found to have the Val1884Met mutation, al., 1989). Given the close homology between the ef- which we have not characterized in detail. Functional fector binding domains of Ras and Rap, its apparent analysis of Pro1718Leu-DOCK4 demonstrated failure of tumor suppressor properties were initially postulated the mutant protein to bind to CrkII, activate Rap GTPase to result from the sequestration of Raf and other Ras signaling, restore cellular adherens junctions in 3081 cells, effectors within an inactive complex (Vossler et al., 1997; or affect their contact inhibition, soft agar colony forma- Bos et al. 2001). However, more recent studies have tion, or in vivo invasiveness. While this tumor-derived suggested that Rap may in fact mediate a distinct cellu- mutation thus appears to be non-functional in these lar signaling pathway. In Drosophila, a gain-of-function assays, it is of interest that expression of Pro1718Leu- mutation in Rap leads to the “roughened” phenotype, DOCK4 activates CDC42 and Rac GTPase activity, prop- which is not affected by altered expression of Ras (Hari- erties that are not observed with the wild-type protein. haran et al., 1991). Similarly, Rap loss-of-function muta- The physiological significance of this effect is supported tions in flies lead to severe morphogenesis defects that by the observation of characteristic CDC42-associated are not modulated by Ras (Asha et al., 1999). Specific filopodia in cells expressing ectopic Pro1718Leu- Rap effectors have not been identified, but a potential DOCK4. Activation of CDC42 and Rac GTPases may be link to the regulation of intercellular contact has been a direct property of the mutant protein or, alternatively, suggested by observations on Drosophila Rap1 null may be the indirect result of dysfunctional protein inter- cells, which demonstrate loss of circumferential ad- actions affecting interconnected GTPase signaling herens junctions (Knox and Brown, 2002). Our observa- pathways. The observation that Pro1718Leu-DOCK4 tions linking Rap activation to adherens junctions in shows reduced binding to the adaptor protein CrkII ar- both mouse osteosarcoma cells and in normal mouse gues against direct activation of these GTPases through osteoblasts supports this effector pathway in mamma- this protein complex. However, DOCK4 may also medi- Cell 682

ate direct GEF activity through the DHR/CZH domains, database and corresponds to a 900 kb segment (nucleotides which are unaltered by the Pro1718Leu mutation. While 12.345 ϫ 105 to 12.354 ϫ 106 in the Ensembl database). further work will be required to demonstrate whether this mutant truly exerts gain-of-function properties, it is Mutational Analysis of DOCK4 in Human Cancer-Derived of interest that CDC42 and Rac activities have been Cell Lines The entire coding region of DOCK4 was amplified by RT-PCR from linked with increased cellular migration and may thus a panel of 44 sporadic cancer cell lines representing a variety of enhance the invasive phenotype (Keely et al., 1997; tumor types. Cell lines analyzed were derived from tumors of the Schmitz et al., 2000). breast (MCF7ADR, MDAMB435, T47D, BT483, MDAMB436, Finally, the successful application of RDA to mouse MDAMB468, MDAMB415, MDAMB231, MDAMB157, HS467T, tumor models provides a general strategy for identifying HS496T, UACC893, and BT549), ovary (ES-2, IGROV-1, MDAH2774, genes that are targeted by chromosomal deletions, fol- OV1063, OVCAR3, OVCAR4, OVCAR5, OVCAR8, SKOV3, and SW626), lung (NCIH460, NCIH522, and HOP92), CNS (SF295, SNB19, lowing the initiating genetic lesions that define each and U251), colon (COLO205, HCT116, HCT15, and HT29 SW620), model. The use of syngeneic mouse tumor models cir- kidney (ACHN, CAKI-1, and U031), and prostate (DU145 and PC3), cumvents a major problem of RDA, namely the isolation in addition to melanomas (LOXMVII, SKMEL2, UACC62) and osteo- of silent polymorphic deletions in the human population sarcomas (U20S and SAOS2). A control population was provided that are rendered homozygous in tumor cells by virtue by EBV-immortalized lymphoblastoid cell lines established from 200 of their presence within a common region of LOH. While healthy blood donors. Total cellular RNA was extracted using RNA- STAT60 (Teltest, Friendswood, Texas) according to the manufactur- technically challenging, RDA offers the advantage of a er’s instructions. The DOCK4 coding region was amplified in a series genome-wide screen for homozygous deletions, which of 13 overlapping fragments (A–M) using the following primer combi- are typically considerably smaller than regions of hemi- nations which are listed in the 5Ј-3Ј orientation: fragment A (AAGGG zygous allelic loss measured by traditional LOH map- GCGCGGGGATTACAAAGCC and TGGGGCAATGAACAACTGGGA ping. High throughput approaches to identifying ge- CTG); fragment B (TGGGGCAATGAACAACTGGGACTG and CTG nomic deletions, such as array-based comparative CACAGCCAAAGGGTCGTCGGTACTGG); fragment C (CTGAATA GAAACGGGCTTCCCAAA GC and GGTACTCACTGGCTGGTGGCT genome hybridization (CGH) will greatly enhance this CCCC); fragment D (CTATTGAAAGGGGA GAATTTGAGAAAG and strategy, once the increased density of arrayed markers CGAGAAACTTTACTATCTCTGAG); fragment E (GAAATG GAGAAC allows for high resolution deletion mapping (Hodgson CCACCCAGACAAG and GTGATGTAACACGACAGGCAGAA); frag- et al., 2001). Together with the completion of the mouse ment F (CTGAGCTCTTTCCCTGCCGTGTAC and CATCTCCGGGC genomic sequence, the increasing number of physiolog- GTATCAATATTCGG); fragment G (ACAGATAGACATTATCAACAG ically relevant mouse tumor models makes this ap- CTTC and CTGTTTAAAGTTGCCACT CCGCCTC); fragment H (CAG CCAGATCTTCGGAATGTCATG and GTCATGCCCT CGACACACAA proach attractive to identify genes whose inactivation ACTC); fragment I (CCGGAAGATTGCAGAGCAGTATGAG and AAT may be relevant to human cancer. CAGAGTCTTCAGGTGCTGATTC); fragment J (AGAGTCTCTGGGT GGAGA GAACGTC and CAAACTTCTCATGCACGGCCAAACC); Experimental Procedures fragment K (CAGGTATC A AGAGGCATTCTTTGTC and GATTAAA CAGCATCCTCTGCGAAGGC); fragment L (CTCAAGCTTGAGTTCT Representational Difference Analysis (RDA) ACTCACTCG and CCTCCCCGCCGTAGCTCGGCACGGG); and and Positional Cloning of DOCK4 fragment M (CAACTCCCCTGTCTTGTCGGGCAG and GCATCG For RDA, genomic DNA from six tumor cell lines derived from NF2/ CAGGTACATA GAAAAGTGAC. All PCRs were performed at an an- Њ TP53 double heterozygous mice (driver) was compared with nealing temperature of 58 C with the exception of fragments C, F, Њ Њ Њ matched normal DNA (tester). Mice used were inbred in the 129/ and J for which annealing temperatures of 63 C, 63 C, and 60 C were Sv background to reduce polymorphic differences that complicate used, respectively. Uncloned RT-PCR products were sequenced in RDA. Genomic DNA was digested with BglII, ligated to adapters, both directions using the BigDyeTerminator kit (Applied Biosystems) and subjected to three rounds of subtractive hybridization and PCR and analyzed on the ABI3100 genetic analyzer. Heterozygous posi- amplification as described (Lisitsyn and Wigler, 1995). RDA products tions were marked using Factura and displayed using Sequence comprising differentially amplified PCR products were cloned and Navigator software. Sequence variants detected at the cDNA level hybridized to Southern blots containing the initial driver and tester were subsequently confirmed at the genomic DNA level. The allele amplicons, as well as primary genomic DNA from tumor cell lines frequency of each sequence variant was determined by genotyping and matched normal mouse tissue. RDA products confirmed to the control population. reside within homozygous deletions in mouse tumor cell lines were sequenced and mapped using mouse-hamster radiation hybrid pan- Expression Analysis and Transfection Studies els (Research Genetics) as well as database analysis (NCBI). Posi- Northern blots of multiple human tissues (Clontech blots H1 and tional cloning of DOCK4 was achieved using RDA clones 11 and 13 H2) were carried out using Express Hybridization buffer (Clontech). from tumor cell line 3081. The RDA products initially mapped to Full-length DOCK4 expression constructs were generated encoding two unordered genomic bacterial artificial chromosomes (BACs) on the wild-type and Pro1718Leu mutation in plasmid pCDNA3.1, both mouse chromosome 12, BAC AC079369, and BAC AC079370. The with and without an N-terminal flag epitope. All constructs were syntenic human genomic sequence on (BAC confirmed by sequencing analysis. All cells were grown in DMEM AC003079 and AC003080) is completed and ordered and was there- with 10% fetal calf serum, and transient transfection was carried fore used in gene prediction programs. EST matches, RT-PCR, and out using the lipid-based reagent, Fugene (Roche). Stable cell lines 5Ј-RACE (placental library) were used to confirm predicted coding were prepared by cotransfecting DOCK4 expression constructs with and untranslated sequences. The transcriptional start and the first the selection plasmid pBABE puro, followed by selection using 3 exon of DOCK4 resides within BAC AC004111.1; a large intron 1 ug/ml puromycin. RapN17 and Rap63E expression plasmids were spans the entire BAC AC004001.1; with the second exon and re- cotransfected with plasmid pcDNA4/TO-E (Invitrogen), followed by maining coding sequence spanning BAC AC003080, BAC selection using 100 ug/ml zeocin. Expression of stably transfected AC003079, and BAC AC005047. The following primers were used constructs was confirmed by Western or Northern blot analysis, and to amplify genomic fragments specific for ZnF277 (AGGTTCCTTGG multiple cell lines were tested to ensure against the effects of clonal TGGCAACGATGTGG and AGTTCTGTCGCAGTATGGCTGTGGC) selection. Primary mouse osteoblasts were isolated as described and NLRR3 (ACCCTGCTCTCTACC ATTTCTCCCGGAGAA and (Thomas et al., 1996). RNAi studies were done as described in El- TGCCGTGGTAGAGGGCACTGAGGGCGTTGG). Mouse sequence bashir et al. (2001). SiRNA for DOCK4 was designed to target nucleo- for this segment of chromosome 12 is also available in the Ensembl tides GUGCGACGGCUGGUACAGA, corresponding to N19 after the GTPases Regulator DOCK4 Disrupted in Tumorigenesis 683

two A residues at nucleotides 104 and 105 of DOCK4 transcript. Received: September 24, 2002 Transfections were done with oligofectamine (Invitrogen) as per Revised: January 29, 2003 manufacturer’s instructions. Cells were either fixed for immunofluo- rescence or harvested for RNA isolation, 48 hr after transfection. References For immunofluorescence, cells growing on glass cover slips were fixed with 4% formaldehyde and blocked in TBS-0.5% Triton X-100 Albert, M.L., Kim, J.I., and Birge, R.B. (2000). ␣v␤5 integrin recruits for one hour. A mouse monoclonal antibody was used for ␤-catenin the CrkII-Dock180-rac1 complex for phagocytosis of apoptotic cells. (Transduction Labs; 1:1000 dilution), and for flag epitope the M2 Nat. Cell Biol. 2, 899–905. monoclonal (Sigma), followed by a fluorescein-conjugated second- Asha, H., de Ruiter, N.D., Wang, M.G., and Hariharan, I.K. (1999). ary goat anti-mouse antibody 568 nm (Molecular Probes; 1:250 dilu- The Rap1 GTPase function as a regulator of morphogenesis in vivo. tion). Actin cytoskeleton was stained by FITC conjugated phalloidin EMBO J. 18, 605–615. (Molecular Probes). All photographs were taken on a Zeiss Axio- Boettner, B., Govek, E.E., Cross, J., and Van Aelst, L. (2000). The plan 2 microscope and on an immunofluorescent confocal micro- junctional multidomain protein AF-6 is a binding partner of the scope (Bio-Rad, Hercules, California). Growth curves for 3081 cells Rap1A GTPase and associates with the actin cytoskeletal regulator expressing DOCK4 constructs were performed by seeding 1 ϫ 103 profilin. Proc. Natl. Acad. Sci. USA 97, 9064–9069. cells on 10 cm dishes in triplicate and counting refractile cells. For soft agar colony formation, 1 ϫ 103 cells were plated on a 60 mm Bos, J.L., de Rooij, J., and Reedquist, K.A. (2001). Rap1 signalling: petri dish, within DMEM in 0.33% agarose and over a layer of DMEM adhering to new models. Nat. Rev. Mol. Cell Biol. 2, 369–377. in 0.5% agarose, and after two weeks, colonies with Ͼ 2mmin Brugnera, E., Haney, L., Grimsley, C., Lu, M., Walk, S.F., Tosello- diameter were counted. For mouse tumorigenicity studies, 2 ϫ 106 Trampont, A.-C., Macara, I.G., Madhanis, H., Fink, G.R., and Ravi- cells were inoculated subcutaneously into nude mice, and the size chandran, K.S. (2002). Unconventional Rac-GEF activity is mediated of resulting tumors was monitored at six weeks. Two, independent, through the Dock180-ELMO complex. Nat. Cell Biol. 4, 574–579. stable cell lines were tested for each expression construct, each in Call, K.M., Glaser, T., Ito, C.Y., Buckler, A.J., Pelletier, J., Haber, D.A., duplicate. Tumor histology was examined using hematoxylin and Rose, E.A., Kral, A., Yeger, H., Lewis, W.H., et al. (1990). Isolation and eosin staining. characterization of a zinc finger polypeptide gene at the human cromosome 11 Wilms’ tumor locus. Cell 60, 509–520. Coprecipitation and GTPase Assays Cavenee, W.K., Dryja, T.P., Phillips, R.A., Benedict, W.F., Godbout, Association of DOCK4 with SH3 domains was demonstrated by R., Gallie, B.L., Murphree, A.L., Strong, L.C., and White, R.L. (1983). incubating RIPA lysates (0.5 mg) from 293T cells transfected with Expression of recessive alleles by chromosomal mechanisms in flag-tagged DOCK4 and Pro1718Leu-DOCK4 constructs with bacte- retinoblastoma. Nature 305, 779–784. rially expressed GST fusion proteins immobilized on glutathione Chen, Q., Kimura, H., and Schubert, D. (2002). A novel mechanism agarose (5 ␮g; Pharmacia). Incubations were carried out for 2 hr at for the regulation of amyloid precursor protein metabolism. J. Cell 4ЊC, followed by SDS-PAGE, and Western blotting using anti-flag Biol. 158, 79–89. antibody (Sigma). For GTPase studies, mammalian expression vec- tors encoding either Rap1 or HA-epitope tagged Rho, Rac, and Christofori, G., and Semb, H. (1999). The role of the cell-adhesion CDC42 were cotransfected with either DOCK4 or Pro1718Leu- molecule E-cadherin as a tumour-suppressor gene. Trends Bio- DOCK4 into 293T cells. GTPase assays were done as described by chem. Sci. 24, 73–76. Ren and Schwartz (2000). Cells were lysed in buffer containing 1% Clark, E.A., Golub, T.R., Lander, E.S., and Hynes, R.O. (2000). Geno- Triton and GST binding was assessed by incubation with 5 ug of mic analysis of metastasis reveals an essential role for RhoC. Nature GST-linked Rho binding domain (RBD) of Rhotekin (for Rho-GTP), 406, 532–535. CRIB domain of Pak (for Rac and CDC42-GTP), and Rap binding Cote, J.F., and Vuori, K. (2002). Identification of an evolutionarily domain (for Rap; gift from V. Boussiatis). Total and GTP bound Rap, conserved superfamily of DOCK180-related proteins with guanine Rho, Rac and CDC42 were detected either using anti-Rap1 antibody nucleotide exchange activity. J. Cell Sci. 115, 4901–4913. (Santa Cruz Biotechnology) or antibody to the HA epitope. Dryja, T.P., Rapaport, J.M., and Petersen, R.A. (1986). Molecular detection of deletions involving band q14 of chromosome 13 in Expression of DOCK4 in C. elegans retinoblastomas. Proc. Natl. Acad. Sci. USA 83, 7391–7394. The ced-5(n1812) strain was obtained from the Caenorhabditis Ge- Elbashir, S.M., Harborth, J., Lendeckel, W., Yalcin, A., Weber, K., netics Center. Transgenic animals were created by co-injection of and Tuschl, T. (2001). Duplexes of 21-nucleotide RNAs mediate RNA ␮ 180 g/ml pRF4, which harbors the dominant phenotypic marker interference in cultured mammalian cells. Nature 411, 494–498. rol-6(su1006), along with 50 ␮g/ml of heat shock-driven expression Fearon, E.R. (1997). Human cancer syndromes: clues to the origin constructs created from plasmid pPD49.83 (gift from A. Fire). Results and nature of cancer. Science 278, 1043–1050. for each transgene were obtained from at least two stably transmit- ting transgenic lines. Cell corpses were observed using Nomarski Fukui, Y., Hashimoto, O., Sanui, T., Oono, T., Koga, H., Abe, M., optics. Germline cell corpses were scored following a 1.5 hr 33ЊC Inayoshi, A., Noda, M., Oike, M., Shirai, T., and Sasazuki, T. (2001). heat pulse and 24 hr recovery period. In order to avoid secondary Haematopoietic cell-specific CDM family protein DOCK2 is essential effects of deformed gonads, persistent cell corpses in the adult for lymphocyte migration. Nature 412, 826–831. germline were counted only within gonad arms that were properly Gu, J., Sumida, Y., Sanzen, N., and Sekiguchi, K. (2001). Laminin- reflexed and displayed normal developmental progression of the 10/11 and fibronectin differentially regulate integrin-dependent Rho germ cells. DTC migrations were scored in transgenic animals by and Rac activation via p130(Cas)-CrkII-DOCK180 pathway. J. Biol. exposing synchronously developing populations to three cycles of Chem. 276, 27090–27097. Њ ϫ 33 C 1.5 hr heat pulses, followed by a 10.5 hr recovery period at Guilford, P., Hopkins, J., Harraway, J., McLeod, M., McLead, N., Њ 20 C, as described by Wu and Horvitz (1998). Harawira, P., Taite, H., Scoular, R., Miller, A., and Reeve, A.E. (1998). E-cadherin germline mutations in familial gastric cancer. Nature 392, Acknowledgments 402–405. Hajra, K.M., and Fearon, E.R. (2002). Cadherin and catenin alter- We are grateful to M. Cutro for mouse tumor injection studies; A. ations in human cancer. Genes Chromosomes Cancer 34, 255–268. Yaktine for generation of tumor cell lines; D. Sgroi for mouse tumor Hanahan, D., and Weinberg, R.A. (2000). The hallmarks of cancer. pathology; C. Reinecker for confocal microscopy; and D. Podolsky Cell 100, 57–70. for helpful comments. Rap1 constructs were kindly provided by V. Boussiatis. This work was supported by NIH 5T32DK07191 (V.Y.), a Hariharan, I.K., Carthew, R.W., and Rubin, G.M. (1991). The Drosphila grant from the Avon Foundation (D.W.B.) and the Martell Foundation roughened mutation: activation of a rap homolog disrupts eye devel- (D.A.H.), NIH CA58596 (D.A.H), NIH GM63081 (S.v.d.H.) and the opment and interferes with cell determination. Cell 67, 717–722. AACR-NFCR Professorship in Basic Cancer Research (D.A.H.). Hasegawa, H., Kiyokawa, E., Tanaka, S., Nagashima, K., Gotoh, N., Cell 684

Shibuya, M., Kurata, T., and Matsuda, M. (1996). DOCK180, a major cell-specific expression of DOCK2, a member of the human CDM- CRK-binding protein, alters cell morphology upon translocation to family proteins. Biochim. Biophys. Acta 1452, 179–187. the cell membrane. Mol. Cell. Biol. 16, 1770–1776. Nolan, K.M., Barrett, K., Lu, Y., Hu, K.Q., Vincent, S., and Settleman, Hodgson, G., Hager, J.H., Volik, S., Hariono, S., Wernick, M., Moore, J. (1998). Myoblast city, the Drosophila homolog of DOCK180/ D., Nowak, N., Albertson, D.G., Pinkel, D., Collins, C., et al. (2001). CED-5, is required in a Rac signaling pathway utilized for multiple Genome scanning with array CGH delineates regional alterations in developmental processes. Genes Dev. 12, 3337–3342. mouse islet carcinomas. Nat. Genet. 29, 459–464. Perego, C., Vanoni, C., Massari, S., Raimondi, A., Pola, S., Cattaneo, Jain, A.N., Chin, K., Borresen-Dale, A.L., Erikstein, B.K., Eynstein M.G., Francolini, M., Vicentini, L.M., and Pietrini, G. (2002). Invasive Lonning, P., Kaaresen, R., and Gray, J.W. (2001). Quantitive analysis behaviour of glioblastoma cell lines is associated with altered organ- of chromosomal CGH in human breast tumors associates copy num- isation of the cadherin-catenin adhesion system. J. Cell Sci. 115, ber abnormalities with p53 status and patient survival. Proc. Natl. 3331–3340. Acad. Sci. USA 98, 7952–7957. Perl, A.K., Wilgenbus, P., Dahl, U., Semb, H., and Christofori, G. Jamora, C., and Fuchs, E. (2002). Intercellular adhesion, signalling (1998). A causal role for E-cadherin in the transition from adenoma and the cytoskeleton. Nat. Cell Biol. 4, E101–E108. to carcinoma. Nature 392, 190–193. Keely, P.J., Westwick, J.K., Whitehead, I.P., Der, C.J., and Parise, Reddien, P.W., and Horvitz, H.R. (2000). CED-2/CrkII and CED-10/ L.V. (1997). Cdc42 and Rac1 induce integrin-mediated cell motility Rac control phagocytosis and cell migration in Caenorhabditis ele- and invasiveness through PI(3)K. Nature 390, 632–636. gans. Nat. Cell Biol. 2, 131–136. Kamb, A., Gruis, N.A., Weaver-Feldhaus, J., Liu, Q., Harshman, K., Reif, K., and Cyster, J. (2002). The CDM protein DOCK2 in lympho- Tavtigian, S.V., Stockert, E., Day, R.S., 3rd, Johnson, B.E., and Skol- cyte migration. Trends Cell Biol. 12, 368–373. nick, M.H. (1994). A cell cycle regulator potentially involved in gene- Ren, X.D., and Schwartz, M.A. (2000). Determination of GTP loading sis of many tumor types. Science 264, 436–440. on Rho. Methods Enzymol. 325, 264–272. Kashiwa, A., Yoshida, H., Lee, S., Paladino, T., Liu, Y., Chen, Q., Schmitz, A.A., Govek, E.E., Bottner, B., and Van Aelst, L. (2000). Rho Dargusch, R., Schubert, D., and Kimura, H. (2000). Isolation and GTPases: signaling, migration, and invasion. Exp. Cell Res. 261, characterization of novel presenilin binding protein. J. Neurochem. 1–12. 75, 109–116. Schutte, M., da Costa, L.T., Hahn, S.A., Moskaluk, C., Hoque, A.T., Kitayama, H., Sugimoto, Y., Matsuzaki, T., Ikawa, Y., and Noda, M. Rozenblum, E., Weinstein, C.L., Bittner, M., Meltzer, P.S., Trent, J.M., (1989). A ras-related gene with transformation suppressor activity. et al. (1995). Identification by representational difference analysis of Cell 56, 77–84. a homozygous deletion in pancreatic carcinoma that lies within the BRCA2 region. Proc. Natl. Acad. Sci. USA 92, 5950–5954. Kiyokawa, E., Hashimoto, Y., Kobayashi, S., Sugimura, H., Kurata, T., and Matsuda, M. (1998a). Activation of Rac1 by a Crk SH3- Sekido, Y., Pass, H.I., Bader, S., Mew, D.J., Christman, M.F., Gazdar, binding protein, DOCK180. Genes Dev. 12, 3331–3336. A.F., and Minna, J.D. (1995). Neurofibromatosis type 2 (NF2) gene is somatically mutated in mesothelioma but not in lung cancer. Kiyokawa, E., Hashimoto, Y., Kurata, T., Sugimura, H., and Matsuda, Cancer Res. 55, 1227–1231. M. (1998b). Evidence that DOCK180 up-regulates signals from the CrkII-p130(Cas) complex. J. Biol. Chem. 273, 24479–24484. Shaw, R.J., Paez, J.G., Curto, M., Yaktime, A., Pruitt, W.M., Saotome, I., O’Bryan, J.P., Gupta, V., Ratner, N., Der, C.J., et al. (2001). The Knox, A.L., and Brown, N.H. (2002). Rap1 GTPase regulation of Nf2 tumor supressor, merlin, functions in Rac-dependent signaling. adherens junction positioning and cell adhesion. Science 295, 1285– Dev. Cell 1, 63–72. 1288. Steck, P.A., Pershouse, M.A., Jasser, S.A., Yung, W.K., Lin, H., Ligon, Koike, M., Takeuchi, S., Yokota, J., Park, S., Hatta, Y., Miller, C.W., A.H., Langford, L.A., Baumgard, M.L., Hattier, T., Davis, T., et al. Tsuruoka, N., and Koeffler, H.P. (1997). Frequent loss of heterozy- (1997). Identification of a candidate tumour suppressor gene, gosity in the region of the D7S523 locus in advanced ovarian cancer. MMAC1, at chromosome 10q23.3 that is mutated in multiple ad- Genes Chromosomes Cancer 19, 1–5. vanced cancers. Nat. Genet. 15, 356–362. Li, J., Yen, C., Liaw, D., Podsypanina, K., Bose, S., Wang, S.I., Puc, Takahashi, S., Shan, A.L., Ritland, S.R., Delacey, K.A., Bostwick, J., Miliaresis, C., Rodgers, L., McCombie, R., et al. (1997). PTEN, a D.G., Lieber, M.M., Thibodeau, S.N., and Jenkins, R.B. (1995). Fre- putative protein tyrosine phosphatase gene mutated in human brain, quent loss of heterozygosity at 7q31.1 in primary prostate cancer breast, and prostate cancer. Science 275, 1943–1947. is associated with tumor aggressiveness and progression. Cancer Liang, H., Guo, W., and Nagarajan, L. (2000). Chromosomal mapping Res. 55, 4114–4119. and genomic organization of an evolutionarily conserved zinc finger Thomas, D.M., Hards, D.K., Rogers, S.D., Ng, K.W., and Best, J.D. gene ZNF277. Genomics 66, 226–228. (1996). Insulin receptor expression in bone. J. Bone Miner. Res. 11, Lisitsyn, N.A. (1995). Representational difference analysis: finding 1312–1320. the differences between genomes. Trends Genet. 11, 303–307. Thomas, N.A., Choong, D.Y., Jokubaitis, V.J., Neville, P.J., and Lisitsyn, N., Lisitsyn, N., and Wigler, M. (1993). Cloning the differ- Campbell, I.G. (2001). Mutation of the ST7 tumor suppressor gene ences between two complex genomes. Science 259, 946–951. on 7q31.1 is rare in breast, ovarian and colorectal cancers. Nat. Genet. 29, 379–380. Lisitsyn, N., and Wigler, M. (1995). Representational difference anal- ysis in detection of genetic lesions in cancer. Methods Enzymol. Vossler, M.R., Yao, H., York, R.D., Pan, M.G., Rim, C.S., and Stork, 254, 291–304. P.J. (1997). cAMP activates MAP kinase and Elk-1 through a B-Raf- and Rap1-dependent pathway. Cell 89, 73–82. Lisitsyn, N.A., Leach, F.S., Vogelstein, B., and Wigler, M.H. (1994). Detection of genetic loss in tumors by representational difference Wu, Y.C., and Horvitz, H.R. (1998). C. elegans phagocytosis and analysis. Cold Spring Harb. Symp. Quant. Biol. 59, 585–587. cell-migration protein CED-5 is similar to human DOCK180. Nature 392, 501–504. McClatchey, A.I., and Jacks, T. (1998). Tumor suppressor mutations in mice: the next generation. Curr. Opin. Genet. Dev. 8, 304–310. Zenklusen, J.C., Conti, C.J., and Green, E.D. (2001). Mutational and functional analyses reveal that ST7 is a highly conserved tumor- McClatchey, A.I., Saotome, I., Mercer, K., Crowley, D., Gusella, J.F., suppressor gene on human chromosome 7q31. Nat. Genet. 27, Bronson, R.T., and Jacks, T. (1998). Mice heterozygous for a muta- 392–398. tion at the Nf2 tumor suppressor locus develop a range of highly metastatic tumors. Genes Dev. 12, 1121–1133. Accession Numbers Meller, N., Irani-Tehrani, M., Kiosses, W.B., Del Pozo, M.A., and Schwartz, M.A. (2002). Zizimin1, a novel Cdc42 activator, reveals a The GenBank accession number for the nucleotide and protein se- new GEF domain for Rho proteins. Nat. Cell Biol. 4, 639–647. quence of DOCK4 reported in this paper is AY233380. Nishihara, H., Kobayashi, S., Hashimoto, Y., Ohba, F., Mochizuki, N., Kurata, T., Nagashima, K., and Matsuda, M. (1999). Non-adherent