Mitogenic signaling by Ret/ptc2 requires association with enigma via a LIM domain.

The ret/ptc2 papillary thyroid cancer oncogene, an oncogenic form of the c-Ret receptor tyrosine kinase, is the product of a somatic crossover event fusing the dimerization domain of the type Ialpha regulatory subunit of cyclic AMP-dependent protein kinase (RI) with the tyrosine kinase domain of c-Ret. Mitogenic activity of Ret/ptc2 required dimerization via the N terminus of RI and a tyrosine residue located C-terminal to the kinase core of Ret, Tyr-586 (Durick, K., Yao, V. J., Borrello, M. G., Bongarzone, I., Pierotti, M. A. and Taylor, S. S. (1995) J. Biol. Chem. 270, 24642-24645). Using the yeast two-hybrid system, Ret/ptc2 binding proteins were identified, and the sites of interaction with Ret/ptc2 were mapped. The SH2 domains of phospholipase Cgamma and Grb10 were both identified, and binding depended on phosphorylation of Tyr-539 and Tyr-429, respectively. These interactions, however, were not required for mitogenic signaling. The second of the three LIM domains in Enigma (Wu, R. Y., and Gill, G. N. (1994) J. Biol. Chem. 269, 25085-25090) was also identified as a Ret/ptc2 binding domain. Enigma, a 455-residue protein, was discovered based on its interaction with the insulin receptor through the C-terminal LIM domain. Although the association with Enigma required Tyr-586 of Ret/ptc2, the interaction was phosphorylation-independent. In contrast to the SH2 interactions, disruption of the interaction with Enigma abolished Ret/ptc2 mitogenic signaling, suggesting that LIM domain recognition of an unphosphorylated tyrosine-based motif is required for Ret signal transduction.

The c-ret proto-oncogene encodes a receptor tyrosine kinase with a cadherin-like extracellular domain (1). Mutations of c-ret are responsible for two distinct classes of genetic disease. Germline loss of function mutations in c-ret result in the de-velopmental disorder Hirschsprung's disease (2,3), while activating mutations result in the multiple endocrine neoplasia family of inherited cancers (4 -7). In addition to the germline alterations, somatic mutational events lead to constitutively active forms of c-ret, and these are found in nearly half of all papillary type thyroid carcinomas (8).
Chromosome translocations or inversions in papillary thyroid carcinoma (ptc) 1 give rise to various fusion proteins where the C-terminal tyrosine kinase domain of c-Ret is fused to an N-terminal portion of another gene product. One of the resultant transforming proteins, observed in multiple independent cases of ptc, was the product of a crossover between the type I␣ regulatory subunit of cyclic AMP-dependent protein kinase (RI) gene with c-ret (9). This protein, Ret/ptc2, is 596 residues in length and contains the N-terminal two-thirds of RI followed by the entire tyrosine kinase domain of c-Ret (10). Using a microinjection-based assay for mitogenic activity, we previously showed that the N-terminal dimerization domain of RI was essential for constitutive activation of Ret/ptc2 (11).
In the absence of a known ligand, studies of Ret signaling have been done using activated forms of the Ret tyrosine kinase, like Ret/ptc2, or a chimeric epidermal growth factor/Ret kinase receptor. It was suggested from work with the EGFR/ Ret chimera that Ret couples to a novel mitogenic signaling pathway because, while growth stimulatory effects were as strong as those of the EGF or platelet-derived growth factor receptors, stimulation of mitogen-activated protein kinases and PLC␥ by Ret was, in comparison, very weak (12). Work with Ret/ptc2 indicated that both intrinsic protein tyrosine kinase activity and a tyrosine residue (Tyr-586) located outside the kinase core were absolutely required for Ret/ptc2-induced mitogenesis (11).
A yeast two-hybrid screen was used to identify the presumed Src homology 2 (SH2) or phosphotyrosine binding domain that interacts with Tyr-586 of Ret/ptc2. The SH2 domains of PLC␥ and Grb10 were both found to interact with Ret, but neither bound at Tyr-586. Mutations in Ret/ptc2, which interfered with these SH2 interactions, had no effect on the mitogenic activity. A protein that interacted with Tyr-586 was identified as Enigma (13). This interaction was found to be highly specific, mediated by the second of the three LIM domains of Enigma and independent of Ret phosphorylation or activity. Disruption of the interaction of Ret/ptc2 with Enigma, either by mutation of Tyr-586 in Ret/ptc2 or by co-expression with a dominant negative form of Enigma, abolished the mitogenic activity of Ret/ptc2.

EXPERIMENTAL PROCEDURES
Two-hybrid Screen-A yeast two-hybrid screen was performed by the methods of Vojtek et al. (14), with reagents from Stan Hollenberg. Ret/ptc2 cDNA was subcloned into the LexA-fusion vector pBTM116 and coexpressed in the L40 strain of Saccharomyces cerevisiae with an embryonic mouse random primed cDNA library. From approximately two million co-transformants, seven interacted specifically with the Ret portion of Ret/ptc2. The cDNA inserts of these were sequenced by the dideoxy method (15), and sequences obtained were compared with the contents of GenBank using the BLAST program through the NIH/NCBI * This research was supported in part by United States Army Grant AIBS1762 (to S. S. T.) and National Institutes of Health Grant DK13149 (to G. N. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ server on the World Wide Web. Three library vectors encoded the following mouse sequences: the C-terminal 155 residues of Grb10 (16); 156 residues that share 97% identity with residues 537-693 of rat PLC␥2 (17); and 131 residues with 95% identity to the C-terminal 131 residues of human Enigma containing all of LIM2 and LIM3 (13).
Microinjection Mitogenic Activity Assay-The development of this assay is described in detail elsewhere (11). Briefly, mouse 10T1/2 fibroblasts were plated on glass coverslips and grown to 70% confluence in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. The coverslips were then transferred to Dulbecco's modified Eagle's medium containing 0.05% calf serum. After 24 h of serum starvation, the cells were injected into their nuclei with solutions of injection buffer (20 mM Tris, pH 7.2, 2 mM MgCl 2 , 0.1 mM EDTA, 20 mM NaCl) containing 100 g/ml Ret/ptc2 expression plasmid DNA and 8 mg/ml rabbit IgG (Sigma). For co-injection experiments, 200 g/ml of a second expression plasmid was also present. All microinjection experiments were performed using an automatic micromanipulator (Eppendorf, Fremont, CA), with glass needles pulled on a vertical pipette puller (Kopf, Tujunga, CA). Entry into S-phase was assessed through incorporation of the thymidine analog 5-bromodeoxyuridine and its subsequent detection by immunostaining. Injected cells were identified by immunostaining of the rabbit IgG injection marker.

RESULTS
Yeast Two-hybrid Interactions-To search for proteins that interact with the Ret/ptc2 oncogenic protein, a mouse randomprimed cDNA library was screened using a yeast two-hybrid system (14). Three sequences isolated from the library by interaction with Ret/ptc2 matched the SH2 domain of Grb10, the first SH2 domain of PLC␥, and a C-terminal fragment of Enigma that contained LIM domains 2 and 3 (LIM2/3). The interaction of Ret with PLC␥ and Grb10 has been observed previously (12,19). Using the two-hybrid system, the interactions of these proteins with Ret were characterized.
Mutants of Ret/ptc2 were prepared and interactions were quantitated by ␤-galactosidase activity (Fig. 1b). The SH2 domains failed to interact with a kinase-inactive mutant of Ret/ ptc2 (K282R), indicating that these interactions depended on autophosphorylated tyrosine residues. In contrast to the SH2 domains, binding of the LIM2/3 was not diminished in the kinase-inactive mutant. By testing for interaction between the SH2 domains and various Tyr to Phe mutants of Ret/ptc2, Ret/Enigma Mitogenic Signaling 12692 residues required for Grb10 and PLC␥ binding were identified as Tyr-429 and Tyr-539, respectively. Both the Grb10 and PLC␥ SH2 domains bound to the Y586F mutant, but binding to the Enigma LIM domains was eliminated by this mutation and by a mutation in which the C-terminal 23 residues of Ret were deleted (CЈ574).
In Vitro Binding to Ret-To verify the two-hybrid results, affinity precipitation experiments were performed. The binding domains from PLC␥, Grb10, and Enigma were expressed in Escherichia coli as GST-fusion proteins. The three GST-fusion proteins were incubated with lysates of NIH3T3 cells expressing the EGFR/Ret chimeric receptor (Fig. 1a), where EGF-dependent activation of EGFR/Ret has been characterized (12). In each case binding to EGFR/Ret from lysates of EGF-treated cells was observed (Fig. 1c). Results shown were using a GSTfusion protein of only LIM2 from Enigma, because Enigma binding to Ret was determined to be mediated by LIM2. 2 GST-fusion proteins with the SH2 domains of Grb2 and v-Src were also expressed and tested for in vitro binding. Neither GST alone nor GST-Grb2 bound to the EGFR/Ret chimera. The SH2 domain of v-Src, however, did interact with EGFR/Ret, and interaction with all three GST-SH2 domains required EGFstimulated receptor autophosphorylation. In contrast, interaction with GST-LIM2 of Enigma did not require receptor autophosphorylation. Grb2 served as a negative control because Ret has two splice isoforms (10). The long form binds to Grb2 (20) while the short form, used in all of the constructs described here, does not contain the Grb2 consensus site. Both isoforms of Ret are mitogenic (12).
Characterization of the Ret-Enigma Interaction-Because the LIM2 domain of Enigma bound at a site crucial for the mitogenic activity of Ret, this interaction was investigated further. Using an inducible two-hybrid system (21,22), where higher expression levels were achieved, it was possible to observe the phosphorylation state of Ret using anti-phosphoty- 2

FIG. 2. Requirements for the association between Ret/ptc2 and
Enigma. a, mapping of binding determinants in Ret/ptc2. Various fragments of Ret/ptc2, shown schematically, were used in the yeast two-hybrid system to measure interaction with the product of a plasmid expressing full-length Enigma. ␤-Galactosidase activity of transformants was measured by solution assay, and values shown are averages of duplicate assays. The same pattern was observed in results from three separate experiments with full-length Enigma, the C-terminal half containing the LIM domains, or LIM2 alone. b, tyrosine autophosphorylation of Ret/ptc2 mutants. Lysates of yeast transformed with plasmids expressing fragments shown in panel a were run on 10% SDS-polyacrylamide gel electrophoresis, blotted to nitrocellulose, and probed as in Fig. 1c. Anti-PY, anti-phosphotyrosine. c, specificity of the LIM domains of Enigma for Ret/ptc2. pJG4 -5 plasmids coding for full-length CRP, the three LIM domains of Zyxin (residues 339 -542), and the three LIM domains of Enigma (residues 275-455) were cotransformed into yeast with either Ret tk , the pEG202 plasmid expressing the Ret kinase with an intact C terminus, or EGFR tk , a pEG202 construct containing the intracellular domain of EGFR. ␤-Galactosidase activity was assessed by streaking transformants on 5-bromo-4chloro-3-indoyl ␤-D-galactoside plates. ϩϩϩ, denotes dark blue after 12 h of incubation compared with white (Ϫ Ϫ) that was equivalent to background after 24 h.

FIG. 3. Effect of mutations in Ret that disrupt association with
Enigma and of co-expression with the LIM domains of Enigma on mitogenic activity of RET/ptc2. a, effects of mutations in Ret/ ptc2. Serum-starved mouse fibroblasts (10T1/2) were microinjected with plasmids expressing either wild-type Ret/ptc2 or various mutants and then assessed for entry into S-phase by immunofluorescent detection of 5-bromodeoxyuridine (BrdU) incorporation. The fraction of injected cells positive for 5-bromodeoxyuridine incorporation is shown with error bars displaying the 95% confidence interval calculated using the standard error of proportion. The numbers in parentheses are the total number of injected cells. Plasmids were injected at a concentration of 100 g/ml. Asterisk denotes a highly significant difference between cells injected with Ret/ptc2 or Y586F and Ret/ptc2 or CЈ574 (p Ͻ 0.001). b, effects of co-expression of Enigma and its fragments. Serum-starved fibroblasts were microinjected with a mixture of two of the following expression plasmids: Ret/ptc2 or RI/EGFR, a construct analogous to Ret/ptc2 with the EGFR intracellular domain in place of the Ret kinase, plus either con, control empty plasmid; C275, the C-terminal 275 residues of Enigma containing LIM domains 1, 2, and 3; FL, full-length Enigma; N279, N-terminal 279 residues of Enigma lacking LIM domains; or Zyxin, LIM domains from Zyxin, residues 339 -452. In each case Ret/ptc2 and RI/EGFR constructs were injected at 100 g/ml, while the other constructs were present at 200 g/ml. Double asterisk denotes a highly significant difference between cells injected with Ret/ptc2ϩcon or Ret/ptc2ϩC275 (p Ͻ 0.001).
Ret/Enigma Mitogenic Signaling 12693 rosine antibodies. Enigma bound to both Ret/ptc2 and to the Ret tyrosine kinase alone (Ret tk ) but did not bind to either when the C-terminal 23 residues were deleted (CЈ574, Fig. 2a). The interaction was not dependent upon the phosphorylation state of Ret because Ret tk was not phosphorylated on tyrosine in the absence of the dimerization domain of RI, whereas Ret/ ptc2 underwent tyrosine autophosphorylation (Fig. 2b, lanes Ret tk and Ret/ptc2). Based on optical densitometry, the extent of tyrosine phosphorylation of Ret/ptc2 and the C-terminal deletion Ret/ptc2 CЈ574 were equivalent. Because deletion of the C terminus of Ret/ptc2 did not decrease the phosphotyrosine content by a detectable amount, Tyr-586, the only tyrosine in the C terminus of this isoform of Ret, does not appear to be a major site of autophosphorylation. Results using both the EGFR/Ret chimera and Ret/ptc2 thus indicated that interaction with Enigma required the C terminus containing Tyr-586 but was independent of tyrosine autophosphorylation of Ret.
The interaction between LIM2 of Enigma and Ret was specific because LIM domains from other proteins failed to interact with Ret (Fig. 2c). Another tyrosine kinase, EGFR, failed to interact with the LIM domains of Enigma either in the twohybrid system (Fig. 2c) or in GST-fusion binding reactions (data not shown).
Effect of Enigma on Ret/ptc2 Mitogenic Activity-The functional significance of the association between Ret and Enigma was investigated in vivo using microinjection. Various Ret/ptc2 expression constructs were injected into nuclei of serumstarved fibroblasts, and the capacity of these constructs to induce DNA synthesis was assayed by monitoring incorporation of the thymidine analog 5-bromodeoxyuridine. Mutations in Ret/ptc2 that blocked association with Grb10 (Y429F) or PLC␥ (Y539F) had no significant effect on mitogenic activity, while mutation of Tyr-586 to Phe or deletion of the C terminus of Ret/ptc2 completely blocked the ability of Ret/ptc2 to induce DNA synthesis (Fig. 3a).
The strong correlation between mutants that failed to bind Enigma and loss of mitogenic activity suggested that Enigma was either required for the mitogenic signaling of Ret/ptc2 or that some other protein, which was not detected in the twohybrid screen, also binds at Tyr-586. To discriminate between these two possibilities, co-injection experiments were performed to attempt to block the Ret/ptc2 mitogenic signal. Coinjection of Ret/ptc2 with a plasmid that expressed the three LIM domains of Enigma (C275, Fig. 3b) blocked Ret-induced DNA synthesis, while co-injection with full-length Enigma had no effect. These results support the conclusion that Ret/ptc2 requires Enigma for mitogenic signaling. If Enigma was simply competing for binding with some other signaling protein, then both full-length Enigma and the LIM domains alone should block signaling, given that both interact with Ret with an equivalent affinity. 2 The inhibition of mitogenesis was specific to the LIM domains of Enigma because the LIM domains of Zyxin (23) were without effect. It was also specific for Ret because the LIM domains of Enigma did not block the mitogenic activity of the EGFR tyrosine kinase analog of Ret/ptc2 (RI/EGFR) previously shown to have mitogenic activity in this assay (11). DISCUSSION LIM domains contain approximately 50 amino acids, bind two atoms of Zn 2ϩ , and are found in a variety of homeodomain proteins (24), cytoskeleton-associated proteins (23,25), protein kinases, and proteins of unknown function (26). Enigma was originally discovered as a protein that binds to exon 16 of the insulin receptor (13) at a tyrosine-based sequence important for receptor internalization, and that interaction is through the C-terminal LIM domain of Enigma, LIM3. In the present study, Enigma bound to Ret via the LIM2 domain to a sequence required for mitogenic signaling. LIM3 of Enigma is highly specific for the Tyr-based motif in the insulin receptor whereas LIM2 is highly specific for the Tyr-586-based motif in Ret. 2 The N-terminal portion of Enigma is required for mitogenic signaling because co-injection of only the LIM domains with Ret/ptc2 ablated the mitogenic signal, while co-injection with full-length Enigma did not.
As shown here, Enigma is required for Ret/ptc2 mitogenic signaling while previous results established that Ret tyrosine kinase activity was also required (11). Kinase activity, however, is not required for recruitment of Enigma to Ret/ptc2. Enigma binds to a tyrosine-containing sequence in an activation-independent manner, and this is clearly different from SH2 or phosphotyrosine binding domain interactions. Enigma might either become phosphorylated upon Ret activation or serve to localize Ret to a subcellular position required for kinase-mediated signaling. In either case, these results define a novel mechanism for mitogenic signaling.