Abstract
Transmembrane signaling by receptor tyrosine kinases typically involves a dynamic receptor monomer-dimer equilibrium in which ligand binding to soluble extracellular domains triggers receptor dimerization and subsequent signaling events. Although the role in signal transduction of the single transmembrane helices of individual receptors, which connect the extracellular with the intracellular protein domains, is not understood in detail, we show here that the single transmembrane domains of all 58 human receptor tyrosine kinases alone have an intrinsic propensity to form stable dimeric structures within a membrane. Thus, defined interactions of the transmembrane domains are most likely generally involved in signaling by all human receptor tyrosine kinases.
Introduction
Receptor tyrosine kinases (RTKs) are integral membrane proteins with a large extracellular ligand-binding domain and an intracellular tyrosine kinase domain, which are connected by a single transmembrane (TM) helix (1). With a few exceptions, individual RTKs are believed to exist in a dynamic monomer-dimer equilibrium within the membrane, and ligand binding to the extracellular domains induces dimerization of individual RTK extracellular domains and, subsequently, conformational rearrangement and transphosphorylation of the intracellular kinase domains (1). The disulfide-linked insulin receptor, as well as the members of the ErbB RTK family, preexists as a dimer within the membrane even without a bound ligand, and ligand binding to the extracellular domains most likely results in a structural rearrangement of the preformed dimers and in transphosphorylation of the intracellular kinase domains. Transphosphorylation leads to activation of different signaling cascades, which are involved in regulation of diverse cellular processes such as growth, migration, adhesion, differentiation, chemotaxis, angiogenesis, as well as apoptosis (2). Consequently, malfunction of individual RTKs can result in carcinogenesis (3).
Studies with the soluble extracellular domains of human RTKs have shown that ligand binding to these domains is critical for dimerization, a key step in RTK signaling (2). However, certain types of cancer are caused by malfunction of human RTKs due to mutations of single amino acids in the TM domains, which results, for example, in continuous activation of the mutated RTK (4, 5). These observations indicate that interactions of the single TM helices could also be important for RTK activation and signaling. Indeed, for members of the human epidermal growth factor receptor (EGFR, also known as ErbB) RTK family, specific interactions of the TM domains have been implicated in protein functioning (6–10). Biophysical studies have indicated that the isolated TM domain of the fibroblast growth factor receptor 3 (FGFR3) and all human ErbB family members interact in detergent (11–13), whereas the single TM domains of all human ErbBs only weakly dimerize in micelles (14) and the TM domain of the human cholecystokinin receptor 4 (CCK4) does not self-interact in a detergent environment (15). However, studies in detergent are complicated because micelles are a poor mimic of a membrane environment (16, 17). TM domain interactions occurred in model membranes for human ErbB, EphA1, and FGFR3 (18–27). Thus, although there is evidence both for and against the propensity of human RTK TM domains to self-interact, there has not been a comprehensive analysis in a single system of all of the TM domains of the human RTKs.
Escherichia coli genetic systems have been developed to measure the interaction propensity of selected TM helices within the inner membrane of the engineered bacteria (28–31). In these systems, a TM domain of interest is fused to a bacterial DNA binding domain. Oligomerization of the fusion protein, which is mediated by a TM helix-helix interaction, brings (at least) two DNA binding domains together. Because only the dimeric DNA binding domains can properly bind to a promoter or operator region, which controls expression of a reporter gene, this provides a method for quantifying TM domain interactions by monitoring the induction or repression of the reporter gene as changes in the activity of the encoded product. The propensities of the TM domains of the human ErbB and the insulin receptor to self-associate within the E. coli inner membrane were tested with a genetic system based on the ToxR DNA binding domain, and the isolated TM domains exhibited substantial self-interaction (32–34). Notably, in eukaryotic cells, the insulin receptor is a disulfide-linked dimer. Thus, this RTK has a different mode of activation from that of ligand-induced dimerization, and ligand binding could induce a structural rearrangement of the preformed dimer. For the ErbB TM domains, motifs of two small residues separated by four residues (small-xxx-small) are involved in homo- as well as heterodimerization of the individual TM helices (10). The first small-xxx-small motif described was the GxxxG-motif, which was identified in studies of the human glycophorin A (GpA) TM domain (35–37). The GxxxG motif appears to frequently mediate and stabilize TM helix-helix interactions in membranes (38). Furthermore, motifs involving other small residues (such as Ala, Ser, or Thr) can also mediate and stabilize defined TM helix-helix interactions (30, 32, 39–41). Small residues are believed to allow two helices to come into close contact so that other forces, such as van der Waals interactions and hydrogen bonding, can stabilize a given TM helix bundle (42–45). In some cases, even a single glycine residue appears to be sufficient as a framework for TM helix-helix interactions (46).
Thus, many studies have indicated that small residues in TM helices can have an important role in mediating and stabilizing defined helix-helix interactions. Interestingly, the single TM helices of all human RTKs are rich in small residues, and the small-xxx-small motif is common in these TM domains (47). The initial observations that TM helix interactions could be involved in RTK signaling and that the RTK TM domains are enriched in small amino acid residues and often contain GxxxG-like motifs raise the question of whether the single TM helices of all human RTKs have an intrinsic propensity to mediate or stabilize, or both, the interactions of these RTKs, and thus may be generally involved in RTK activation and signaling. To address this question, we have analyzed the interaction propensities of all 58 human RTK TM domains within a membrane by the E. coli TOXCAT assay (28). We show that all human RTK TM domains can form oligomeric structures within a membrane, which suggests that the interaction of the TM domains is a general property of RTKs and contributes to activation and signaling by human RTKs.
Results
To determine whether all human RTK TM domains have the ability to self-interact, we globally analyzed the interaction propensity of the human RTK TM domains within a biological membrane. On the basis of their sequence and ligand-binding properties, human RTKs are grouped into different families (Figs. 1 and 2). The TM domain of each of the 58 members of the 20 human RTK families (48) (Table 1) was expressed as a chimeric protein targeted to the inner membrane of E. coli, and the propensity to form homooligomers was measured with the TOXCAT assay (28), a genetic system developed to study TM helix-helix interaction within the E. coli inner membrane. The N termini of the single TM domains of the individual human RTKs were fused to the ToxR DNA binding domain, and the C termini were fused to the MalE protein from E. coli. For the TOXCAT assay, the hydrophobicity of the TM domain functions as a membrane insertion signal, placing the ToxR domain in the E. coli cytoplasm and the MalE domain in the periplasm of E. coli. If the TM domains interact, the ToxR cytoplasmic domains are in close proximity, forming a functional dimer that can bind to the operator region, inducing the expression of the reporter gene encoding chloramphenicol acetyl transferase (CAT). Chimeric proteins with noninteracting TMs fail to stimulate the production of CAT, and the bacteria have no CAT activity.
Homooligomerization of the human RTK TM domains measured with the TOXCAT assay. (A) Families with two members. (B) Families with three members. The top graphs show the interaction propensities relative to the interaction of the human GpA wild-type (WT) TM domain; the weak interaction propensity of the GpA G83I mutant is also shown. Below the graphs are Western blots of total E. coli cellular extracts (w) and the supernatant (s) containing the soluble proteins and the pellet (m) containing the integral membrane proteins after NaOH extraction.
Homooligomerization of the human RTK TM domains measured with the TOXCAT assay. (A) Families with four members. (B) The PDGFR (platelet-derived growth factor receptor) family. (C) Families with single members. (D) The Eph family. The data are presented as in Fig. 1. All expressed proteins are found almost exclusively in the membrane protein fraction. In some cases, a small quantity of the chimeric protein could be partially alkali extracted. This fraction is thought not to interfere with the assay.
Interaction propensities of the human RTK TM domains. Sequences of the human RTK TMs are given in bold letters. Non-bold letters indicate amino acids added by the expression plasmid. The 58 human RTKs are classified as described in (48). The propensity of the individual TM domains to interact was measured with the TOXCAT assay (see Materials and Methods for details). Alternate names for the proteins are separated by a slash.
The interaction propensity of the individual TM domains was determined relative to the well-characterized strong interaction of the wild-type TM domain of human GpA, as well as relative to the much weaker, but still physiologically relevant, interaction of the GpA G83I variant (49).
The TOXCAT assay can be influenced by the abundance of a chimeric protein in the bacterial inner membrane, as well as by the TM orientation of the chimeric protein within the inner membrane. In some cases, we observed that the abundance of the membrane-integrated chimeric proteins was altered compared to that of the chimera containing the GpA wild-type TM domain (Figs. 1 and 2). Therefore, the determined interaction propensities were corrected for the difference in the expression level per cell [as described in detail in (50)].
We also verified that the chimeric proteins were integrated into the membrane, as well as the topology of the chimeric proteins. Membrane localization of the chimeric proteins was shown by NaOH extraction, which separates membrane proteins from cytoplasmic and periplasmic proteins. After rigorous extraction of the membranes with NaOH, the expressed chimeric proteins were essentially exclusively found in the membrane protein fraction with only a small quantity of the chimeric protein partially extracted in the case of the EphB1 and CCK4 TM domains (Fig. 2). This partially extractable form is thought not to interfere with the assay because of its low abundance and because it most likely represents the protein extracted from the membrane rather than a fraction of soluble chimeric protein. Thus, effectively, all of the chimeric proteins are bound to the bacterial inner membrane and are not localized in the cytoplasm or periplasm.
We examined the topology of the chimeric proteins by performing complementation assays on M9 minimal medium (Fig. 3). Because the E. coli strain NT326 lacks endogenous MalE, the cells are unable to transport maltose into the cytoplasm. If the chimeric proteins are inserted into the inner membrane with the MalE portion localized in the periplasm, the MalE domain will compensate for this deficiency and the cells can grow on minimal medium with maltose as the only carbohydrate source. With the exception of chimeric proteins with the DDR1 or DDR2 TM domains, most chimeric proteins complemented the endogenous MalE deficiency, which shows correct insertion of the protein into the E. coli cytoplasmic membrane with the MalE domain located in the periplasm (Fig. 3C). Furthermore, because the TOXCAT assay shows reporter gene activation for all of the chimeric proteins, the ToxR domains are located in the cytoplasm (Fig. 3A).
The TOXCAT assay and test for membrane insertion and orientation. (A) A diagram of the TOXCAT assay and topology of the chimeric proteins. Homodimerization of a TM helix results in dimerization of the cytosolic ToxR DNA binding domain, which subsequently results in activation of the cat reporter gene. The MalE domain of the chimeric proteins is located in the bacterial periplasm; thus, assaying the periplasmic MalE activity can be used to show an appropriate TM topology. (B) The various cloned TM domains are listed with respect to their position on the assay plates shown in (C) and (D). The ToxRstop plasmid codes for a protein where only a truncated ToxR domain is expressed (no fusion to any TM domain), and pccKan (Kan) is the empty expression plasmid. As controls, the nontransformed NT326 cells were plated, as well as cells expressing the soluble MalE protein into the periplasm (pMalp2, positive control) or the cytoplasm (pMalc2, negative control) of E. coli. (C) Results of the malE complementation assay to test for the TM orientation (see Materials and Methods for details). E. coli cells were grown on maltose as the carbohydrate source. (D) Viability of the E. coli NT326 cells transformed with the plasmids on medium containing glucose as the carbohydrate source.
For the cells expressing the DDR1 and DDR2 TM domains, which both belong to the same RTK family, the cells did not grow on maltose minimal medium; thus, we could not establish the TM topology by the maltose complementation assay. However, all cells expressing the chimeric proteins grew on minimal medium with glucose as the only carbohydrate source (Fig. 3D), which shows that the viability of the cells was not impaired. To compensate for the MalE deficiency of the E. coli NT326 cells, the MalE domain of the chimeric proteins must not only be localized within the E. coli periplasm but must also diffuse freely, and, most likely, the fused MalE domains must contact the maltose transporter in a defined way (51). If diffusion is limited or the structure of the TM dimer does not allow proper contact with the transporter, the cells cannot grow on maltose minimal plates. Because the chimeric proteins with the DDR TM domains were localized solely within the E. coli inner membrane (Fig. 1A), and because the expression of the ToxR-responsive reporter gene was activated (Table 1 and Fig. 1A), the ToxR domain must be localized within the E. coli cytoplasm. Together, the complementation assay and activation of the ToxR-dependent reporter gene assay establish that all 58 of the chimeric proteins are expressed and properly integrated in the bacterial inner membrane in the correct orientation.
The TOXCAT assay revealed that all 58 members of the different human RTK families exhibited self-interaction and that, with the exception of FGFR1, the interaction propensities were greater than or equal to that of the chimera containing the GpA G83I variant TM domain (Figs. 1 and 2 and Table 1). All interaction propensities were determined relative to the interaction of the GpA wild-type TM domain. When the interaction propensity of the RTK TM domains was compared to that of the GpA wild-type TM domain, most RTK TM domains exhibited a weaker interaction propensity, and only few proteins showed an interaction propensity stronger than that of wild-type GpA (Table 1). The median interaction propensity was about 40% of the wild-type GpA dimerization propensity. More than 80% of the RTK TM domains interacted only 0.006 to 0.6 times as strongly as the interaction mediated by the GpA wild-type TM domain. Only the TM domains of both DDR family members exhibited an interaction propensity greater than that of the wild-type GpA TM domain. For all other families, there was greater variability, with the TM domains of some family members exhibiting a stronger interaction than that of the GpA TM domain (LTK and TIE1), some exhibiting an interaction propensity similar to that of the GpA wild-type TM domain (HER2/ErbB2, EphB6, and IGFR1), and some exhibiting an interaction propensity similar to that of the weakly interacting GpA G83I mutant (FGFR1, FLT1/VEGFR1, EphA4, FGFR3, MET, and RON).
Discussion
Activation of RTKs is believed to involve a ligand-induced dimerization or a conformational change of preformed dimeric structures. Several studies have indicated that interactions of the RTK TM domains could be involved in RTK signaling in eukaryotes; thus, the individual TM domains may have an intrinsic tendency to form oligomers. Here, we show that all 58 TM domains of the human RTKs indeed have a self-interaction propensity; thus, interactions of all the human RTK TM domains are likely to contribute to receptor dimerization and activation.
The interaction of the single TM domain of wild-type GpA, which was used as an internal control, results in formation of a structurally rigid TM helix dimer (35). For the activation of RTKs, dynamic TM domain interactions and promiscuous formation of less stable dimers are feasible (42, 52). The contribution of the TM domain to RTK signaling, which involves a structural rearrangement of the TM domains, would be precluded if the TM helices form an excessively strong TM dimer and become locked in a certain TM configuration. Nearly all RTK TM domains interacted at least as strongly as the GpA G83I variant, with the exception of the human FGFR1 TM domain, which interacted slightly less well. The GpA G83I TM helix forms stable dimers in detergent (49); thus, we conclude that all human RTK TM domains have the propensity to form stable TM dimers in a membrane environment. Because the interactions were measured within the E. coli inner membrane rather than in the eukaryotic plasma membrane, the observed strength of the RTK TM interactions has to be taken with caution and can be different from the interactions that occur in a different membrane environment. Furthermore, only one orientation of the TM domain relative to the ToxR DNA binding domain was analyzed, and the relative orientation of these domains can also influence the measured interaction propensity.
Although a few of the RTK TM domains exhibited an interaction propensity as strong as or stronger than that of the wild-type GpA TM domain (Fig. 4), most of the RTK TM domains exhibited an interaction propensity consistent with a more dynamic interaction than that predicted for the wild-type GpA TM domain. One might speculate that those RTK TM domains that exhibited an interaction stronger than or equal to that of the wild-type GpA TM domain in the TOXCAT assay formed highly stable TM dimers; however, it is difficult to extrapolate this to the in vivo setting of a human cellular membrane. Indeed, the formation of highly stable TM dimers would not permit the dynamic equilibrium between monomeric and dimeric RTKs that is believed to regulate the activity of many of these proteins.
Interaction propensities (IPs) of human RTKs relative to the IP of the GpA WT TM domain. The WT GpA TM domain was set as 1 and is represented by the line. Expression of the ToxR domain without any fused TM domain (ToxRstop) results in no measurable IP (0 ± 0.006), and the weakly interacting GpA G83I TM domain has an IP of 0.09 ± 0.007. Each bar represents the relative IP (rel. IP) of a single human RTK TM domain. See Table 1 for each RTK TM domain IP value.
This study is limited to an evaluation of the TM domain in the ability of RTKs to self-interact and does not take into account several factors that are likely to contribute to RTK oligomerization and activation in their native environment. For example, this study does not include regulatory functions of the intracellular and extracellular RTK domains. Additionally, interaction propensities of RTKs are potentially also controlled by the lipid composition of the membrane, as well as by the biophysical properties of the bilayer, which are not the same in the bacterial inner membrane and eukaryotic plasma membranes. For example, some RTKs appear to function within lipid domains, which are specifically enriched in cholesterol and other lipids (53, 54).
Furthermore, on the basis of in vivo analysis of the ErbB family of RTKs (55, 56), which suggest that a concentration-dependent shift in the amount of homodimers may control receptor activation, it is likely that the monomer-dimer equilibrium of RTKs is controlled by the actual concentration of certain RTKs within eukaryotic membranes, which is also not considered in this study. The results show that the TM domains of all human RTKs have the intrinsic propensity to form or stabilize (or both) RTK dimers within a membrane environment. This observation agrees with several in vivo observations on the potential function of the RTK TM domains in RTK signaling.
Because of their importance in human disease, changes in expression of ErbB genes is associated with various types of cancer (57, 58); many aspects of RTK structure and function have been analyzed and described for the human ErbB family of RTKs. All human ErbB family members appear to exist as homo- and heterodimers within cellular membranes (55) even in the absence of ligand, and truncation of the extracellular domain of human ErbBs often results in constitutively active protein dimers (34–36). Thus, in the absence of the extracellular domains, the receptors form a signaling competent structure, indicating a critical function of the ErbB TM domains in signal transduction (34–36). Ligand binding induces dimerization of ErbB1 protein fragments in detergent: The induced dimerization was, however, far more efficient when the extracellular domain and the TM domain were used rather than the extracellular domain alone (7). This observation suggests that interactions of the TM domains could be critically involved in ErbB signaling. The isolated TM domain of ErbB receptors homo- and heterodimerize in a bacterial membrane, and it has been suggested that TM helix-helix interactions are involved in ErbB signaling (10, 32). Most likely, a ligand-induced rotation of RTKs is involved in activation of the receptors (56), and the relative orientation of the TM helices and the intracellular kinase domains are thus critical for receptor signaling. Indeed, two different TM structures of the ErbB TM domains have been suggested (10), and the structure of an inactive ErbB2 TM dimer has been solved by nuclear magnetic resonance spectroscopy (59). When a dimerization-inducing Glu-Val-Val-Val-Val-Val-Val-Glu motif was shifted across the TM helix of the rat homolog of the human ErbB2 protein, different conformations of the protein in the absence of ligand were promoted, and the kinase activity of ErbB2 was only activated when two kinase domains are positioned in a specific rotational conformation (60). Thus, activation of the RTK in the absence of ligand required the TM domains to interact in a specific way. This study showed that TM interactions are involved in ErbB2 signaling and that the geometry of the TM helix dimer is critical. A dimerization interface of the ErbB2 TM domains has been mapped with a protein disulfide cross-linking approach (61).
The observation of a required specific TM geometry explains the earlier finding that, despite strong dimerization, replacement of the ErbB2 TM helix by the strongly dimerizing interaction motif of GpA did not activate the receptor function (62). Most likely, dimerization of the two GpA helices did not result in correct positioning of the kinase domains or locked the RTK in an inactive configuration.
Although the previous work strongly indicated that the TM domains of human ErbB receptors are critical for receptor activation and signaling, the results presented here suggest that the single TM domains of all human RTKs encode self-interaction. Thus, interactions of the TM domains are most likely critically involved in signaling by all human RTKs. RTKs may undergo a series of rearrangements, starting with dimerization triggered by ligand binding of the extracellular domain, then interactions mediated by the TM domain position the intracellular catalytic domains into an active conformation allowing receptor activation and signaling (Fig. 5). For RTKs that exist as dimers in the absence of ligand, ligand binding may initiate TM domain interactions or a rearrangement of a preformed TM dimer that finally results in the active conformation of the intracellular domains.
Formation of human RTK dimers and the role of the TM domain in RTK signaling. Specific interactions of the extracellular domains (step 1), as well as dynamic TM helix interactions (step 2), are critical for the formation or stabilization, or both, of RTK dimers. As suggested for human ErbB family members (68), interactions between TM dimers (step 3) could also contribute to the formation of the active conformation of the intracellular domains of the receptors.
The data we present not only expand the current view of the importance of the TM domain in RTK signaling but also suggest that treatments targeting the TM domain interactions, for example, by synthetic peptide drugs (63, 64), may be applicable to any human disorder involving RTK (mal)functions. The interaction of the TM domains and the function of the ErbB2, as well as those of the insulin receptor, could be controlled by the addition of small TM peptides that bind to the respective TM domains and thereby inhibit interactions of the RTK TMs of the whole proteins (33, 34). The design and global use of such “interceptors” could be a promising strategy to target diseases, such as cancer, which are caused by malfunction of human RTKs.
Materials and Methods
TOXCAT assay
In vivo measurements of the homodimerization of the various RTK TM domains were done with the TOXCAT system (28). The construction of the plasmid pccKan and of the chimera of GpA has been described (28). To create the chimeras for the individual RTK TM regions, synthetic oligonucleotide cassettes (Eurofins MWG Operon) encoding parts of the RTK TM domains were ligated into the Nhe I/Bam HI restriction digested plasmid pccKan. The correct insertion of the genes was checked by DNA sequencing. Ligation of a gene to this plasmid results in generation of an open reading frame that encodes for a chimeric protein with an N-terminal fusion to the ToxR DNA binding domain and a C-terminal fusion to the MalE domain of E. coli. Plasmids were transformed into E. coli NT326 (65). Cells were grown in the presence of antibiotics overnight and diluted in the same medium to A600 of 0.1 the next morning. Cells were harvested at A600 of 0.6 and CAT activities were measured as described previously (66). The interaction propensities of the individual TM domains were determined by measuring the CAT activities three times for at least three independent clones. Whole-cell lysates were used to estimate the abundance of the expressed proteins. Proteins were detected by Western blot analysis using polyclonal antibodies that recognize MalE (New England Biolabs). The amount of expressed fusion protein with the various RTK TM domains was determined relative to the amount of expressed GpA fusion protein by comparing the respective Western blot intensities with standardized amounts of cells expressing the GpA fusion protein following procedures described in (50).
Test for membrane insertion and orientation
Each single plasmid used for the TOXCAT measurements was tested for proper integration of the encoded chimeric protein into the E. coli inner membrane. Membrane insertion of the chimeric proteins was tested by NaOH extraction of lysozyme-treated cells as described (30, 67). Only membrane integral proteins remain in the membrane after this treatment.
To test for the orientation of the chimeras in the E. coli membrane, plasmids were transformed into E. coli NT326 cells, which are malE deficient, and grown on M9 medium with 0.4% maltose as the only carbohydrate source. Cells in which the MalE domain of the chimeric proteins is located in the periplasm complement the absence of endogenous MalE, thus establishing that the topology is correct. Viability was established by transforming the plasmids into the E. coli NT326 cells and growing the bacteria on M9 medium containing 0.4% glucose.
Acknowledgments
We thank D. M. Engelman for providing the TOXCAT system. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (SCHN 690/2-3). Financial support from the Ministry of Science, Research, and Arts of Baden-Württemberg is also gratefully acknowledged. C.F. is supported by a Schlieben-Lange fellowship of the Ministry of Science, Research, and Arts of Baden-Württemberg. We thank E. Salfelder for technical support.
