Redirecting tyrosine kinase signaling to an apoptotic caspase pathway through chimeric adaptor proteins

Abstract

Signal transduction pathways are typically controlled by protein–protein interactions, which are mediated by specific modular domains. One hypothetical use of such interaction domains is to generate new signaling pathways and networks during eukaryotic evolution, through the joining of distinct binding modules in novel combinations. In this manner, new polypeptides may be formed that make innovative connections among preexisting proteins. Adaptor proteins are specialized signaling molecules composed exclusively of interaction domains, that frequently link activated cell surface receptors to their intracellular targets. Receptor tyrosine kinases (RTKs) recruit adaptors, such as Grb2 and ShcA, that activate signaling pathways involved in growth and survival, whereas death receptors bind adaptors, such as Fadd, that promote apoptosis. To test the ability of interaction domains to create new signaling pathways, we have fused the phosphotyrosine recognition domains of Grb2 (Scr homology 2 domain) or ShcA (phosphotyrosine-binding domain) to the death effector domain of Fadd. We find that these chimeric adaptors can reroute mitogenic or transforming RTK signals to induce caspase activation and cell death. These hybrid adaptors can be used to selectively kill oncogenic cells in which RTK activity is deregulated.

Receptor tyrosine kinases (RTKs) typically transmit signals that promote cell growth and survival and can be constitutively activated by mutations that induce malignant cell transformation (1). Specific autophosphorylated tyrosine motifs on activated RTKs serve as docking sites for the Src homology 2 (SH2) and phosphotyrosine-binding (PTB) domains of cytoplasmic adaptors, such as Grb2 and ShcA, and these modules can therefore target specific complexes to activated RTKs (2, 3). In signaling from normal or oncogenic tyrosine kinases, the nature of the signal activated by these phosphotyrosine (pTyr) recognition domains depends on the sequences to which they are linked. The SH2 domain of the Grb2 adaptor, for example, is flanked by two Src homology 3 (SH3) domains that bind proteins, such as the Sos guanine nucleotide exchange factor and the Gab1 docking protein, which are involved in activation of the Ras and phosphatidylinositol 3′ kinase pathways, respectively (4, 5). Grb2 therefore couples pTyr-X-Asn motifs, recognized selectively by the SH2 domain, to signaling pathways that are recruited by the SH3 domains, and promote cell proliferation, growth, and survival. A variation on this theme is provided by mammalian docking proteins, such as Shc, FRS2, and IRS-1 family members. These proteins all possess a PTB domain that binds phosphorylated NPXY motifs on activated RTKs, and are phosphorylated on tyrosine on recruitment to the receptor. Their phosphorylation creates binding sites for the SH2 domains of cytoplasmic signaling proteins, including Grb2, and thereby potentiates the activation of specific biochemical pathways that stimulate growth and survival (6).

The activation of signaling pathways through adaptor proteins comprised of modular interaction domains is not limited to RTK signaling, but is a common mechanism used by diverse cell-surface receptors. For example, members of the tumor necrosis factor-receptor superfamily contain cytoplasmic domains and motifs that interact with corresponding domains on adaptor proteins. This occurrence is typified by the Fas receptor, which contains a death domain (DD) in its C-terminal tail (7, 8). Trimerization of the receptor leads to binding of the Fas DD to the DD of an adaptor, Fadd, which also possesses a death effector domain (DED) (9). The DED of Fadd associates with the DEDs of procaspase 8/10 (10). Fadd therefore bridges the Fas receptor to procaspases. The assembly of this multiprotein complex leads to caspase dimerization and activation, followed by caspase autocleavage and the stimulation of pathways that elicit apoptosis (11–13). Thus, both RTKs and death receptors are activated by oligomerization, and recruit modular adaptors to couple activated receptors to cytoplasmic signaling pathways.

Based on these observations, we considered the possibility that the joining of interaction domains in nonphysiological combinations might be sufficient to reengineer cellular behavior. We therefore sought to rewire RTK signaling from proliferation and survival pathways to apoptosis by exploiting the modular nature of the pathway components downstream of tyrosine kinase and death receptors. We hypothesized that chimeric adaptor proteins composed of interaction domains from both RTK and Fas signaling pathways could potentially convert a mitogenic tyrosine kinase signal into an apoptotic response by recruiting apical components of the caspase pathway to activated RTKs.

Materials and Methods

Recombinant Adenoviruses. Plasmids containing SH2, PTB, DEDR86K, DED-R175Q, DED-SH2, and DED-PTB were cloned into the KpnI/NotI sites of the AdTrack cytomegalovirus (CMV) vector (14). This vector is bicistronic, containing both the construct of interest, and a GFP reporter, under control of the CMV promoter. These plasmids were used to make recombinant, replication-deficient adenoviruses, according to the method of He et al. (14). Adenoviruses were amplified and purified as described (15). Adenoviral titer was determined by monitoring cytopathic effect in an endpoint dilution assay (Q-Biogene, Carlsbad, CA). Titers were confirmed by Western blotting to ensure similar protein expression levels of chimeric adaptors, and by GFP fluorescence of the different viruses.

Cell Culture. RAT-2 fibroblasts and TWO3 nasopharyngeal carcinoma cells were propagated at 37°C in DMEM, supplemented with 10% FBS, 100 units/ml^–1 penicillin, and 100 μg/ml^–1 streptomycin. TWO3 cells have lost the Epstein–Barr virus episome through serial passaging, and are no longer overtly transformed (16). To derive NTR2 cells, RAT-2 cells were transfected with ErbB2/NeuNT, and cells were maintained in media (DMEM plus 5% FBS) until foci developed (2–3 weeks). Foci were picked and expanded. Expression of ErbB2/NeuNT was confirmed by Western blot analysis.

Survival and Caspase Assays. To determine cell survival, 5 × 10⁵ cells were plated onto 24-well dishes and grown overnight. The next day, cells were infected with adenoviruses bearing GFP, SH2, PTB, DED-R86K, DED-R175Q, DED-SH2, or DED-PTB at the multiplicities of infection (mois) indicated. Caspase 8 inhibitor was used at a concentration of 20 μM where indicated. To determine the effect of epidermal growth factor (EGF) stimulation, infected TWO3 cells were serum-starved overnight and stimulated with EGF (0–100 ng/ml). The next day, cell survival relative to GFP was determined according to the method of Serrano et al. (17). The experiments were performed in triplicate and were repeated at least three times.

Colony and Tumorigenicity Assays. NTR2 cells were plated onto 10-cm plates and grown overnight. The next day, cells were counted, infected at an moi of 200, and incubated for 1 h. Growth in soft agar was assessed by plating 5 × 10⁵ infected NTR2 cells (in triplicate) in 0.25% agarose, supplemented with DMEM (10% FBS), on top of 0.5% agarose in DMEM (10% FBS) on 60-mm plates. Cells were maintained for 21 days (at 37°C in 5% CO₂), and then subsequently stained overnight with MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide). Colonies were counted and averaged.

For xenograft experiments, NTR2 cells were plated on 10-cm plates and grown overnight. The next day, cells were infected at an moi of 200 and incubated for 4.5 h. Cells were counted and resuspended in PBS at a density of 10⁶ cells per ml. Female severe combined immunodeficient (SCID) mice (6–8 weeks old) were anesthetized, and 10⁵ cells (100 μl) were injected into the left gastrocnemius muscle. Three mice were injected per construct. Each experiment was performed three times. Tumor progression was monitored by measuring leg plus tumor diameter. The mice were killed once this diameter reached 15 mm, or when the tumor was ≈1.7 g. All experiments were conducted in accordance with the guidelines of the Animal Care Committee, Ontario Cancer Institute.

Focus Assay. RAT-2 cells were transfected with 0.5 μg ErbB2/NeuNT cDNA and maintained in culture until foci developed. Cells were then incubated with adenoviruses bearing SH2, DED-R86K, DED-SH2, or DED-PTB, and individual foci were marked and monitored over a 48-h period.

Caspase 8 Activity. Caspase 8 activity was measured by using a caspase 8 assay kit according to the manufacturer's instructions (Calbiochem). Briefly, 2 × 10⁶ NTR2 cells were infected with adenoviruses bearing (at an moi of 200) SH2, DED-R86K, DED-R175Q DED-SH2, or DED-PTB, and grown at 37°C. At 2-h intervals, the cells were scraped, pelleted, and washed with PBS. The cell lysates were incubated with synthetic substrate (IETD-pNA) at 37°C and monitored for absorbance at 405 nm every 30 min. For TWO3, the cells were infected (at an moi of 100; DED-PTB, at an moi of 100 or 10 as indicated) and grown overnight in low serum media (DMEM plus 0.5% FBS). The next day, the media was replaced with fresh low-serum media, with or without EGF (100 ng/ml), and then incubated at 37°C. Cells were restimulated with EGF at 6 h incubation. Caspase activity was determined as above.

Results

Chimeric Adaptors Stimulate Apoptosis in ErbB2/NeuNT-Transformed Cells. To link RTK signaling to an apoptotic pathway, we constructed hybrid adaptor proteins containing the N-terminal DED from Fadd, attached to either the Grb2 SH2 domain (DED-SH2) or the ShcA PTB domain at the C terminus (DED-PTB). These chimeric adaptors have the potential to bind procaspase 8 through their DED, and to target the resulting complex to pTyr motifs on activated RTKs through their SH2 or PTB domains. The constructs were C-terminally tagged with an Myc-His epitope and incorporated into recombinant adenoviruses (see Fig. 7, which is published as supporting information on the PNAS web site, www.pnas.org). We also tested the effects of SH2 and PTB domains expressed on their own. In addition, DED domains can spontaneously oligomerize and initiate apoptosis when overexpressed in fibroblast cell lines (9, 18). As a control for apoptosis induced by DED overexpression, we constructed a DED-SH2 adaptor in which Arg 86 of the Grb2 SH2 domain, which is critical for pTyr binding, was mutated to Lys (DEDR86K). We also constructed a DED-PTB adaptor in which Arg 175 of the ShcA PTB domain is mutated to glutamine (DEDR175Q). This substitution abrogates the ability of ShcA PTB domain to bind pTyr, but not phospholipid (19).

To test the rewiring strategy, we derived a cell line (NTR2) of RAT-2 fibroblasts transformed by a constitutively active variant of the ErbB2 RTK (NeuNT), which has binding sites for both the Grb2 SH2 and ShcA PTB domains (20). Immunoblotting of lysates of NTR2 cells, which had been incubated overnight with DED-SH2, DED-PTB, DED-R86K, or SH2-bearing adenoviruses (at an moi of 200), detected equivalent levels of the different adenovirus-encoded adaptor proteins (Fig. 1A). To determine the effect of the chimeric adaptors on cell survival, NTR2 cells were incubated with an increasing amount of adenoviruses, and cell viability was quantified after 24 h (Fig. 1B). The expression of DED-PTB led to a rapid decrease in cell survival, even at an moi as low as 10. DED-SH2 also diminished cell survival, but required mois >100 to reduce cell survival to <20%. DED-R86K and DED-R175Q had only a modest effect on cell survival, and required very high expression levels (at an moi of 200; Fig. 1B). The SH2 and PTB (data not shown) domains alone did not affect cell survival. Expression of the DED-PTB and DED-SH2 chimeric adaptors induced DNA laddering and Poly (ADP-ribosyl) polymerase cleavage, which is characteristic of apoptotic cell death (Fig. 1 C and D). Thus, expression of chimeric DED-SH2 and DED-PTB adaptors in ErbB2/NeuNT-transformed fibroblast appears sufficient to initiate apoptosis.

Fig. 1.

(A) Western blot analysis (α-Myc) of NTR2 (RAT-2 cells transformed with ErbB2/NeuNT) cells infected with SH2-, DED-R86K-, DED-SH2-, and DEDPTB-bearing adenoviruses. (B) Survival of NTR2 cells (24 h) after treatment with increasing mois of adenoviruses. (C) Agarose gel electrophoresis of DNA isolated from NTR2 cells after treatment with the chimeric adaptors. (D) Western blot analysis [anti-Poly(ADP-ribose) polymerase (PARP)] after treatment with the chimeric adenoviruses.

Chimeric Adaptors Stimulate Caspase Activity and Form a Disc-Like Complex in ErbB2-Transformed Cells. To investigate whether apoptosis of ErbB2-transformed NTR2 cells induced by the chimeric adaptors depends on caspase 8 activity, we measured cell survival in the presence or absence of the caspase 8 inhibitor Z-IETD-FMK. Cotreatment of NTR2 cells with caspase 8 inhibitor restored cell viability to DED-SH2 cells (at an moi of 200), and partially restored viability to DED-PTB-expressing cells. At low levels of DED-PTB (i.e., at an moi of 10), cell viability was fully restored with treatment of Z-IETD-FMK, arguing that DED-SH2/DED-PTB-induced cell death depends on caspase 8 activation (Fig. 2A). To test whether the expression of DED-SH2 and DED-PTB leads to caspase 8 activation, caspase 8 activity in lysates from NTR2 cells was measured. As shown in Fig. 2B, caspase 8 activity was markedly elevated in DED-PTB-expressing cells and reached a maximum level within 10 h after infection. Similarly, caspase 8 activity in DED-SH2 lysates increased, peaking at 12 h. Caspase 8 activity in DEDR86K-expressing cells was not detectable until 16–18 h postinfection, 6 h after DED-PTB and DED-SH2 lysates had reached maximum activity. The activity of DED-R86K is therefore likely due to the spontaneous oligomerization of the DED, which depends on accumulation of DED-R86K protein over time. In contrast, caspase 8 activation by DED-PTB/DED-SH2 occurs within a few hours of viral addition, even before GFP expression is detectable, which is most likely due to the recruitment of the wild-type PTB or SH2 domains in the chimeric adaptors to the activated ErbB2 RTK.

Fig. 2.

(A) NTR2 cells were incubated with DED-PTB, DED-SH2, DED-R86K, PTB, or SH2 encoding adenoviruses (at an moi of 200), and survival (24 h) in the presence or absence of caspase 8 inhibitor (Z-IETD-FMK) was measured by crystal violet staining. The data represent the average ± SE for the experiment. (B) Caspase 8 activity in lysates from SH2-, DED-R86K-, DED-SH2-, or DED-PTB-expressing NTR2 cells was determined by measuring cleavage of a synthetic IETD-pNA caspase 8 substrate over time.

We determined whether the observed cell death was correlated with the recruitment of caspase 8 to ErbB2/NeuNT by DED-SH2 and DED-PTB. Expression of either DED-SH2 or DED-PTB, but not DED-R86K, resulted in a decrease in ErbB2/NeuNT expression (Fig. 3A). The rapid down-regulation of ErbB2 induced by the chimeric adaptors suggests that recruitment of caspase 8 to activated ErbB2 by DED-SH2 and DEDPTB may lead to ErbB2 proteolytic cleavage by caspase 8. Indeed ErbB2 contains potential caspase 8 sites (amino acids 1013–1016 and 1127–1130, Rattus norvegicus XP_218925.1) that may be cleaved when caspase 8 is artificially tethered to ErbB2/ NeuNT by the chimeric adaptors. This loss of the receptor complicated the detection of any complex between ErbB2, DED-SH2/DED-PTB, and caspase 8. To circumvent this problem, we cultured the cells with caspase 8 inhibitor (Z-IETDFMK), immunoprecipitated adaptor molecules from cell lysates with anti-Myc antibody, and blotted the resulting immune complexes with ErbB2 antibody (Fig. 3B). Under these circumstances, the ErbB2/NeuNT RTK was stabilized, and was coprecipitated with the SH2, DED-SH2, PTB, and DED-PTB proteins; in contrast, the DED-R86K mutant with an inactive SH2 domain was not significantly associated with ErbB2/NeuNT. These results argue that wild-type SH2 and PTB domains retain their ability to bind the activated ErbB2 RTK when they are coupled to the Fadd DED.

Fig. 3.

(A) Lysates of NTR2 cells expressing Myc-tagged SH2, DED-R86K, DED-SH2, or DED-PTB were immunoblotted with the indicated antibodies. The SH2 polypeptide is not shown because it migrated off the bottom of the gel. (B) NTR2 cells cultured with the Z-IETD-FMK caspase 8 inhibitor, immunoprecipitated with Myc (9E11) antibody, and blotted with ErbB2/Neu (AB3) antibody. (C) NTR2 cells were cultured with (+) or without (–) the caspase 8 inhibitor. Myc immunoprecipitates were probed with anti-caspase 8. Full-length procaspase 8 (+ Z-IETD-FMK), or the p26 cleavage product, coimmunoprecipitate with adaptors containing DED. (D) Flag immunoprecipitation of caspase 8 (Left) or ErbB2/NeuNT immunoprecipitation (Right) from transfected 5637 cells (with caspase 8 inhibitor) probed with either α-Flag or α-ErbB2/NeuNT. The ± indicates whether or not a construct was transfected/infected. Arrows indicate position of caspase 8 and ErbB2/NeuNT. *, a non-specific band in Flag Ip.

To explore the ability of the DED of the chimeric adaptors to recruit caspase 8, anti-Myc immunoprecipitates of the various adaptors were blotted for endogenous caspase 8 in the presence or absence of caspase 8 inhibitor. Procaspase 8 is cleaved on activation to produce 43- and 10-kDa products, of which the 43-kDa protein is further processed into 26- and 18-kDa polypeptides (21). In the presence of caspase 8 inhibitor, full-length procaspase 8 was coimmunoprecipitated with the DEDSH2, DED-R86K, and DED-PTB proteins. In the absence of caspase 8 inhibitor, the 26-kDa (DED-containing) cleavage product was detected in immunoprecipitates of DED-SH2 and DED-PTB, but was poorly associated with DED-R86K. These results indicate that the DED-SH2/DED-PTB adaptors interact with both ErbB2/NeuNT and caspase 8, stimulate caspase 8 autocleavage, and induce apoptosis. We also tested whether, in the presence of the DED-SH2 and DED-PTB adaptors (plus inhibitor), caspase 8 could coprecipitate with ErbB2/NeuNT, and vice versa. Immunoprecipitation of ErbB2/NeuNT from 5637 cells transfected with ErbB2/NeuNT and Flag-caspase 8, and blotting with anti-Flag showed that Flag-caspase 8 coimmunoprecipitated with ErbB2/NeuNT only in the presence of the chimeric adaptors (Fig. 3D) Similarly, ErbB2/NeuNT coprecipitated with Flag-caspase 8 only in the presence of DEDSH2 or DED-PTB, suggesting the formation of a ternary complex between the chimeric ErbB2/NeuNT, the chimeric adaptors, and caspase 8.

Chimeric Adaptors Suppress the Growth of ErbB2/NeuNT-Transformed Cells in Soft Agar and SCID Mice. To pursue the effects of chimeric adaptors on mitogenesis, we tested the ability of NTR2 cells treated with adenoviruses encoding DED-SH2, DED-PTB, and control proteins to grow in semisolid agar. Infection of NTR2 cells with DED-SH2 or DED-PTB adenoviruses reduced the ability of these cells to form colonies (Fig. 4A). There was a 50- to 100-fold reduction in the ability of DED-SH2 or DED-PTB cells to form colonies when compared with cells expressing DED-R86K, or the SH2 or PTB domains alone (data not shown). We also tested the effects of the DED-SH2 and DEDPTB adaptors on tumor formation in an in vivo xenograft transplant model. NTR2 cells infected with DED-PTB, DEDSH2, SH2, or DED-R86K adenoviruses (at an moi of 200), or uninfected cells, were injected into the gastrocnemius muscle of SCID mice (Fig. 4B). Muscle plus tumor diameter was determined over a 3-week period. In control mice, tumors were detected within 7 days postinjection and progressed very rapidly. However, treatment of NTR2 cells with DED-PTB, and, to a lesser extent, DED-SH2, delayed tumor formation by ≈4 days (DED-SH2), or 12 days (DED-PTB), respectively. Eventually, tumors did arise in both DED-SH2- and DED-PTB-treated mice, possibly because of cells that had acquired resistance to apoptosis induced by this mechanism, or to cells that escaped adenovirus infection.

Fig. 4.

(A) Colony formation in soft agar of NTR2 cells incubated with SH2, DED-R86K, DED-SH2, or DED-PTB (at an moi of 200). (B) NTR2 cells were either left untreated (uninfected control) or infected with SH2, DED-R86K, DED-SH2, or DED-PTB (at an moi of 200) adenoviruses, and then injected into the leg muscle of SCID mice. Leg plus tumor diameter was measured over a 3-week period. The plotted data represent the mean ± SE for the experiments.

Receptor Tyrosine Kinase Activation Stimulates Apoptosis in Cells Expressing Chimeric Adaptors. We investigated whether the chimeric adaptors selectively kill cells in which basal tyrosine kinase activity is elevated, either through mutation that renders an RTK constitutively active, or through growth factor stimulation. We first determined whether the DED-SH2/DED-PTB adaptors specifically kill ErbB2-transformed cells, as compared with normal RAT-2 cells. To this end, RAT-2 cells were transfected with ErbB2/NeuNT to generate foci of transformed cells in a background of normal, contact-inhibited cells. The cells were then incubated with equal amounts of adenoviruses encoding DED-SH2, DED-PTB, or DED-R86K (SH2 and PTB not shown), and individual foci were examined by phase contrast microscopy. NeuNT-transformed foci infected with DED-SH2 and DED-PTB displayed significant apoptosis 24–48 h after infection, and often detached from the substrate (Fig. 5). In contrast, foci from plates infected with DED-R86K were not obviously affected. Significantly, the surrounding monolayer of nontransformed RAT-2 cells survived treatment with DEDSH2, and with DED-PTB, indicating that the nontransformed cells are less sensitive than their ErbB2/NeuNT-transformed counterparts to apoptosis induced by the chimeric adaptors. GFP fluorescence, which provides a marker for adenovirus infection, was detected in both the monolayer and foci (data not shown). These data suggest that the apoptosis induced by the DEDSH2/DED-PTB adaptors depends on the deregulated pTyr signal of the oncogenic ErbB2/NeuNT RTK, and that residual tyrosine kinase signaling in quiescent cells (contact inhibited for 2 weeks) is inefficient at inducing apoptosis. We have confirmed these results by using fusion proteins of DED-SH2, DED-R86K, and SH2, in which the adaptors are fused to a membrane permeable sequence from the HIV TAT protein, which allows fusion proteins to transfer across biological membranes (22). The results were similar to adenovirus-mediated delivery and rule out spurious contribution of adenoviral proteins to DED-SH2-mediated apoptosis (data not shown). To quantify the sensitivities of ErbB2/NeuNT-transformed and parental cells to DEDSH2 or DED-PTB treatment, we treated serum-starved parental RAT-2 and NTR2 cells with identical amounts of DED-SH2 and DED-PTB adenoviruses. Whereas DED-SH2 or DED-PTB (at an moi of 200) reduced the viability of NTR2 cells to <20% within 24 h, parental RAT-2 cells remained viable (70–80%) at the same dose. Thus, we conclude that expression of activated ErbB2 leads to an ≈4-fold increase in killing by chimeric adaptors.

Fig. 5.

(A) RAT-2 cells were incubated with adenoviruses encoding DED R86K, DED-SH2, or DED-PTB proteins, and individual foci were examined over a 48-h period. The surrounding monolayer (48 h) of nontransfected RAT-2 cells is less sensitive to apoptosis induced by DED-SH2/DED-PTB. (B) The survival of parental RAT-2 cells versus NTR2 cells (serum-starved) was compared 24 h after adenoviral delivery (at an moi of 200).

We also investigated the effects of growth factor stimulation on cells expressing the chimeric adaptors. For this purpose, we used TWO3 human epithelial cells, which express the EGF receptor (EGFR) and respond to EGF (16). As shown in Fig. 6A, expression of DED-SH2 and DED-PTB in TWO3 cells stimulated with EGF (100 ng/ml) produced DNA laddering, which is characteristic of apoptosis. Unstimulated cells displayed reduced levels of DNA laddering (data not shown). To test the effect of EGF stimulation on the survival of TWO3 cells, cells were treated with SH2, PTB, DED-R86K, DED-SH2, or DED-PTB adenoviruses, serum-starved overnight, and then stimulated with serum free medium (–EGF) or medium containing EGF (100 ng/ml) for 24 h. EGFR activation decreased the survival of DED-SH2-expressing cells, whereas cells expressing the control DED-R86K, SH2, or PTB proteins were not obviously affected by EGFR stimulation (Fig. 6B). At high mois (>100), DED-PTB expression was lethal to TWO3 cells, even in the absence of EGF. However, at reduced mois (an moi of 10), we observed that EGF induced apoptosis in DED-PTB-expressing cells, indicating that this adaptor is more potent in nucleating an apoptotic response. Consistent with these observations, EGF stimulation of TWO3 cells expressing DED-PTB or DED-SH2 caused a rapid rise in caspase 8 activity that reached a maximum at ≈4 h after stimulation (see Fig. 8, which is published as supporting information on the PNAS web site). Control DED-R86K and DEDR175Q adaptors did not induce this strong spike of EGF-induced caspase activity.

Fig. 6.

(A) TWO3 cells were treated with SH2, DED-R86K, DED-SH2, or DED-PTB adenoviruses in the presence of EGF (100 ng/ml), and DNA laddering was analyzed by agarose gel electrophoresis. (B) TWO3 cells were infected with SH2, PTB, DED-R86K, DEDSH2, or DED-PTB adenoviruses at an moi of 100 (an moi of 100 or 10 for DED-PTB), and were serum-starved overnight. Cells were stimulated with serum-free medium (–EGF) or medium containing EGF (100 ng/ml). Quantification of the survival of TWO3 cells after EGF stimulation, 48 h after infection (24 h after EGF stimulation).

Discussion

Protein–protein interactions mediated by modular domains form an important mechanism for controlling signal transduction pathways (6, 23). One possible reason for the prevalent use of interaction domains is to facilitate the formation of new connections between existing proteins during the course of evolution, and thus to create new signaling pathways. Conversely, aberrant protein interactions can disturb cellular phenotype, and thereby contribute to disease. Proteins encoded by pathogenic bacteria and viruses, or chimeric cellular oncoproteins, frequently exert their effects by usurping normal protein–protein interactions, and thus reorganizing cellular behavior. For example, the Tir protein of enteropathogenic Escherichia coli modifies the actin cytoskeleton of the host cell by binding the SH2/SH3 adaptor protein Nck (24), whereas the Bcr-Abl oncoprotein is phosphorylated within its Bcr region at a YVNV motif that binds the Grb2 adaptor (25, 26). The modular nature of signaling pathways thus lends itself to rewiring, in which proteins are assembled into novel complexes with new biological properties.

Previous work (27–29) has suggested that signaling proteins can be experimentally modified to form novel interactions. For example, the oncogenic potential of Bcr-Abl can be attenuated by chimeric adaptors comprising the catalytic domain of the tyrosine phosphatase Shp1 fused to the Abl-binding domain of Rin1 (27). Similarly, caspase 8 can be artificially activated through chemically inducible dimerization (CID) domains fused in tandem to Fadd or caspase 8 (30), likely resulting in higher-order oligomerization of the fusion proteins in response to chemical dimerizer. Chimeric scaffold proteins have also been used to reroute signaling between distinct MAP kinase pathways in yeast (28). In the example discussed here, we have linked two entirely different signaling pathways by using chimeric adaptors. The ErbB2 RTK has multiple Grb2 SH2-binding sites (at least one direct and two through ShcA) and one ShcA PTB-binding site. The activated receptor exists as a dimer, and could therefore recruit several caspase 8 molecules through the DED-SH2 adaptor, and two caspase 8 molecules through DED-PTB. Although the activated Fas receptor itself forms a trimer, recent data suggest that dimerization is the crucial step in caspase 8 activation (11, 13). These observations suggest a possible mechanism through which the DED-SH2/DED-PTB chimeric adaptors induce caspase 8 activation in response to RTK signaling, namely that the adaptors recruit caspase 8 to the activated receptor in a fashion that suffices for caspase dimerization and activation. These data support the view that quite distinct receptors and signaling pathways operate through the same general principles of modular protein interactions and proximity effects.

Our results show that chimeric adaptors containing a pTyrrecognition domain (SH2 or PTB) fused to a DED can nucleate a complex in which an activated RTK is coupled to caspase 8, leading to stimulation of caspase activity and consequent cell death. We have demonstrated this rewiring of tyrosine kinase and caspase pathways in cells transformed by an oncogenic variant of ErbB2, and in tumors arising from such cells, as well as in epithelial cells stimulated with EGF. The ShcA PTB domain appears more potent than the Grb2 SH2 domain in stimulating apoptosis when linked to the Fadd DED domain, and, indeed, when a high level of the DED-PTB adenovirus is used, apoptosis is constitutive in TWO3 cells. The reasons for the increased ability of DED-PTB to induce apoptosis are unclear, but they likely reflect innate differences between the binding properties of the ShcA PTB domain and Grb2 SH2 domain, or their roles in signaling from ErbB1/ErbB2. The ShcA PTB domain has a slower off-rate for phosphorylated motifs than do SH2 domains (31), and this rate may enhance the potential for caspase 8 dimerization. Furthermore, the PTB domain also binds PIP₂, which promotes ShcA association with the plasma membrane (19). Although the lipid binding ability of ShcA PTB is likely not sufficient to induce significant apoptosis, because this activity should be retained by the mutant PTB domain in the DED-R175Q adaptor, it may contribute. Another distinguishing feature of the PTB and SH2 domains is that Grb2 contributes to the recruitment of the Cbl E3 protein-ubiquitin ligase to ErbB1, raising the possibility that overexpression of the Grb2 SH2 domain could interfere with this negative regulator of RTK signaling (32). Regardless, the data illustrate the inherent differences between these two classes of pTyr-binding modules. Our results raise the possibility that cells can be engineered to alter their responses to external signals, in ways that might be therapeutically beneficial. With detailed knowledge of pTyr signaling in individual tumors, for example, through SH2 profiling (33), chimeric adaptors could be tailored to specifically target the RTK expression profile of the individual tumor. Alternatively, small molecules that could bridge endogenous components of RTK signaling pathways to apoptotic pathways might also reengineer cellular behavior in a manner that could antagonize the growth of transformed cells. Indeed, drugs such as cyclosporin and rapamycin exert their effect by nucleating novel protein–protein interactions (34, 35). In summary, our data support the idea that new signaling pathways can be engineered by the joining of interaction domains in novel combinations.