Introduction
The ErbB family of receptor tyrosine kinases (RTKs) or subclass I RTKs comprises four members: epidermal growth factor receptor (EGFR)/ErbB‐1 (
Ullrich et al., 1984), ErbB‐2 (
Yamamoto et al., 1986), ErbB‐3 (
Kraus et al., 1989;
Plowman et al., 1990) and ErbB‐4 (
Plowman et al., 1993a). The four proteins are widely expressed in epithelial, mesenchymal and neuronal tissues and play fundamental roles during development (
Gassmann et al., 1995;
Lee et al., 1995;
Sibilia and Wagner, 1995;
Threadgill et al., 1995). Interest in the ErbB family of RTKs is high due also to the involvement of some of its members in human cancers (
Hynes and Stern, 1994;
Salomon et al., 1995).
Subclass I RTKs have an extracellular domain (ECD) which bears two cysteine‐rich clusters and is responsible for interaction with polypeptide ligands. A direct consequence of ligand binding to the ECD is the formation of receptor dimers and stimulation of the intrinsic kinase activity, which leads to the phosphorylation of tyrosine residues in the intracellular domain of the receptors (
van der Geer et al., 1994). These serve as docking sites for a number of SH2‐ and PTB‐domain containing proteins (
Kavanaugh and Williams, 1994;
Cohen,G.B. et al., 1995) including the adaptor proteins SHC (
Pelicci et al., 1992) and Grb2 (
Lowenstein et al., 1992) and the p85 subunit of phosphatidylinositol (PtdIns) 3‐kinase (
Fedi et al., 1994;
Prigent and Gullick, 1994), which link RTKs to intracellular signaling pathways such as the mitogen‐activated protein kinase (MAPK) pathway (
Egan and Weinberg, 1993) or the S6 kinase cascade (
Ming et al., 1994)
Regulation of ErbB receptor function is complex, since a large number of ligands, the EGF‐related peptides, have been described. ErbB ligands can be classified into three groups and include: EGF and heparin binding EGF‐like growth factor (HB‐EGF), which bind ErbB‐1 (
Savage et al., 1972;
Higashiyama et al., 1991); betacellulin (BTC), which is a ligand of ErbB‐1 and ErbB‐4 (
Shing et al., 1993;
Beerli and Hynes, 1996;
Riese et al. 1996); neu differentiation factors (NDFs)/heregulins (
Peles and Yarden, 1993), which are ligands of ErbB‐3 and ErbB‐4 (
Plowman et al., 1993b;
Carraway et al.; 1994), the respective low and high affinity receptors (
Tzahar et al., 1994).
By binding to the ECD of their respective receptors, EGF‐related peptides induce not only receptor homodimers but also heterodimers. Consequently, although none of these peptides directly bind ErbB‐2, all of them induce its tyrosine phosphorylation by triggering heterodimerization and cross‐phosphorylation (
King et al., 1988;
Plowman et al., 1993b;
Sliwkowski et al., 1994;
Beerli and Hynes, 1996). Cooperation of ErbB‐2 with the other ErbB receptors has been reported (
Alimandi et al., 1995;
Wallasch et al., 1995;
Pinkas‐Kramarski et al., 1996;
Zhang et al., 1996). Moreover, EGF and NDF receptors have been shown to compete for dimerization with ErbB‐2 (
Karunagaran et al., 1995;
Chen et al., 1996). By means of intracellular expression of an endoplasmic reticulum (ER)‐targeted single chain antibody (scFv) that leads to the specific and stable loss of cell surface ErbB‐2 (
Beerli et al., 1994), we have previously shown that: (i) ErbB‐2 enhances EGF‐induced tyrosine phosphorylation of ErbB‐1 and NDF‐induced tyrosine phosphorylation of ErbB‐3 and ErbB‐4 (
Beerli et al. 1995;
Graus‐Porta et al., 1995); (ii) ErbB‐2 potentiates and prolongs the signal transduction pathways elicited by EGF and NDF (
Beerli et al., 1995;
Graus‐Porta et al.; 1995;
Karunagaran et al., 1996). In addition, we and others have shown that ErbB‐2 increases the affinity of both EGF and NDF for their receptors (
Wada et al., 1990;
Sliwkosky et al., 1994;
Karunagaran et al., 1996). Together, these results suggest that ErbB‐2 acts as a common receptor subunit of all the other ErbB proteins and that the physiological receptors for the EGF‐related peptides are ErbB‐2‐containing heterodimers.
However, several recent observations apparently argue against this model. First, it has been suggested that ErbB‐1–ErbB‐3 heterodimers occur in cell lines overexpressing ErbB‐1, where EGF can efficiently elevate tyrosine phosphorylation of ErbB‐3 (
Kim et al., 1994;
Soltoff et al., 1994). Second, in cells that express moderate levels of the four proteins, EGF, HB‐EGF and BTC not only activate their respective receptors and ErbB‐2, but also ErbB‐3 (
Beerli and Hynes, 1996). Third, cooperative signaling of ErbB‐3 and ErbB‐4 with not only ErbB‐2 but also ErbB‐1 has been demonstrated when the receptors were expressed in cells devoid of ErbB proteins (
Cohen,B.D. et al., 1996;
Pinkas‐Kramarski et al., 1996;
Zhang et al., 1996). Finally, expression of different combinations of ErbB receptors in Ba/F3 hematopoietic cells has revealed that all heterodimers can be formed in response to the appropriate ligand (
Riese et al., 1995,
1996).
Expression of recombinant ErbB receptors in pairwise combinations is a viable approach to study their function. However, the four ErbB proteins are often co‐expressed and this approach does not allow delineation of which ErbB receptor interactions are biologically relevant. Different heterodimers have been shown to elicit very different biological responses (
Riese et al., 1995,
1996), implying the recruitment and activation of distinct signaling molecules. Thus, it is crucial to elucidate which heterodimers are formed in a natural cellular context and to characterize their signaling properties.
To understand in more detail ligand‐induced ErbB receptor heterodimerization and transactivation, as well as the consequent diversification of intracellular cascades, we have down‐regulated the cell surface expression of ErbB‐1 and ErbB‐2 in a number of cell lines that co‐express various amounts of the four ErbB proteins. To do so, ErbB‐2‐specific scFv‐5R (
Beerli et al., 1994) has been expressed in T47D (
Graus‐Porta et al., 1995) and A431 cells, leading to a complete loss of cell surface ErbB‐2 and to its functional inactivation. ErbB‐1‐specific scFv‐R1R has been expressed in T47D cells (
Jannot et al., 1996), leading to a dramatic reduction in cell surface ErbB‐1. This approach has enabled us to analyze in more detail EGF‐, HB‐EGF‐, BTC‐ and NDF‐induced signaling and to unravel a hierarchy guiding the ligand‐induced coordinated action of ErbB receptors.
Discussion
In this study we have analyzed EGF‐, HB‐EGF‐, BTC‐ and NDF‐induced signaling in several human epithelial cell lines co‐expressing all four presently known ErbB receptors. Single chain antibody‐mediated down‐regulation of cell surface ErbB‐1 and ErbB‐2 has allowed us to discover novel aspects of ligand‐induced ErbB receptor interplay. First, ErbB receptor interactions induced by various EGF‐related growth factors are less diverse than previously suggested (
Cohen,B.D. et al., 1996;
Riese et al., 1995,
1996;
Zhang et al., 1996) and follow a strict hierarchy, with ErbB‐2 being the preferred heterodimerization partner for all other ErbB proteins. Second, ErbB‐2 is involved in horizontal signaling, mediating the lateral transmission of signals between ErbB receptors. Third, ErbB‐2 enhances and prolongs the MAPK signaling cascade in response not only to EGF and NDF (
Graus‐Porta et al., 1995) but also in response to BTC. Fourth, the data presented in this paper demonstrate that a given ErbB receptor can acquire different signaling properties depending on its dimerization partner. Finally, our results suggest that ErbB‐1 is not the only receptor for HB‐EGF.
NDF binds and activates ErbB‐3 and ErbB‐4, triggers ErbB‐2‐containing heterodimers and induces ErbB‐2 tyrosine phosphorylation (
Carraway et al., 1994). Moreover, the intracellular retention of ErbB‐2 results in impaired NDF‐induced activation of both receptors (
Beerli et al., 1995;
Graus‐Porta et al., 1995). In contrast, we never observed an increase in ErbB‐1 tyrosine phosphorylation in response to NDF, suggesting that no ErbB‐1‐containing heterodimers are formed. Moreover, in T47D cells depleted of ErbB‐1, NDF activated ErbB‐2, ErbB‐3 and ErbB‐4 to the same extent as in the control cells, indicating that ErbB‐1 does not participate in NDF signaling. Other reports have presented evidence for NDF‐induced formation of ErbB‐3–ErbB‐1 and ErbB‐4–ErbB‐1 heterodimers when these pairs of receptors were co‐expressed in host cell lines that do not bear endogenous ErbB proteins (
Riese et al., 1995;
Cohen,B.D. et al., 1996;
Pinkas‐Kramarski et al., 1996;
Zhang et al., 1996). In accordance with these findings, NDF readily induced ErbB‐1 tyrosine phosphorylation and association with the adaptor protein SHC in T47D cells devoid of cell surface ErbB‐2. These results directly show, for the first time, that ligand‐induced ErbB receptor heterodimerization follows a strict hierarchy. If no ErbB‐2 is available, ErbB‐3 and/or ErbB‐4 are able to heterodimerize with ErbB‐1 in response to NDF. However, if the four ErbB proteins are present on the plasma membrane, the NDF receptors preferentially dimerize with ErbB‐2.
EGF has been found to activate ErbB‐3 in cell lines overexpressing ErbB‐1 (
Kim et al., 1994;
Soltoff et al., 1994). Moreover, co‐expression of ErbB‐1 with either ErbB‐3 or ErbB‐4 in cell types lacking endogeneous ErbB proteins allows EGF to regulate tyrosine phosphorylation of these receptors (
Riese et al., 1995;
Cohen,B.D. et al., 1996;
Zhang et al., 1996). These results imply that EGF may induce ErbB‐1–ErbB‐3 and ErbB‐1–ErbB‐4 heterodimers. However, our experiments demonstrate that the absence of functional ErbB‐2 dramatically impairs EGF‐triggered ErbB‐3 and ErbB‐4 tyrosine phosphorylation, as well as association of ErbB‐3 with the p85 subunit of PtdIns 3‐kinase. Significantly, efficient EGF‐induced activation of ErbB‐3 is dependent on ErbB‐2 not only in T47D cells, which have low levels of ErbB‐1 (7000 molecules/cell), but also in ErbB‐1 overexpressing A431 cells (1×10
6 molecules/cell). Thus, formation of ErbB‐1–ErbB‐3 and ErbB‐1–ErbB‐4 heterodimers cannot account for activation of NDF receptors by EGF. Instead, it appears that EGF activates ErbB‐3 and ErbB‐4 via ErbB‐2. Indeed, in SKBR3 cells an EGF‐induced direct ErbB‐2–ErbB‐3 interaction was detected. It is known that dimerization of receptors is not a static, but rather a dynamic and reversible process (
Yarden and Schlessinger, 1987). Thus, ErbB‐2 may first dimerize with ErbB‐1 in response to EGF and then, in its phosphorylated and activated state, be released and dimerize with and phosphorylate ErbB‐3 or ErbB‐4. Alternatively, the formation of ErbB receptor oligomers may occur (
Lax et al., 1991). Therefore, EGF‐induced activation of ErbB‐3 and ErbB‐4 could be the result of receptor transphosphorylation occuring only within ErbB‐1–ErbB‐2–ErbB‐3 or ErbB‐1–ErbB‐2–ErbB‐4 trimers. In both models, ErbB‐2 is required to activate NDF receptors in response to EGF. However, future studies will be required to delineate the precise mechanism of lateral transmission of signals.
BTC is a ligand for ErbB‐1 and ErbB‐4 (
Beerli and Hynes, 1996;
Riese et al., 1996) and BTC‐induced ErbB‐1–ErbB‐2 and ErbB‐4–ErbB‐2 heterodimers have been reported (
Riese et al., 1996). Consistent with this, down‐regulation of ErbB‐1 results in diminished ErbB‐2 phosphorylation and the absence of cell surface ErbB‐2 results in impaired BTC‐induced ErbB‐1 and ErbB‐4 activation, suggesting that BTC also signals through ErbB‐2‐containing heterodimers. Although BTC could presumably induce ErbB‐4–ErbB‐1 heterodimers, down‐regulation of ErbB‐1 in T47D cells does not affect ErbB‐4 activation, suggesting that ErbB‐4–ErbB‐2 heterodimers, and possibly ErbB‐4 homodimers, are sufficient for BTC to activate ErbB‐4. BTC was also able to elevate tyrosine phosphorylation of ErbB‐3 when co‐expressed with ErbB‐1 in Ba/F3 cells, but not when co‐expressed with ErbB‐4, implying that at least ErbB‐1–ErbB‐3 heterodimers are possible (
Riese et al., 1996). However, in the human epithelial cell lines T47D and A431, BTC‐induced tyrosine phosphorylation of ErbB‐3 was dramatically impaired in the absence of ErbB‐2, indicating that, by analogy with the ErbB‐1 ligand EGF, BTC activation of ErbB‐3 involves the interplay of three receptors, ErbB‐1–ErbB‐2–ErbB‐3 and ErbB‐4–ErbB‐2–ErbB‐3.
We show that down‐regulation of cell surface ErbB‐2 has dramatic consequences on downstream signaling events elicited by BTC. In particular, intracellular retention of ErbB‐2 results in transient ligand‐induced MAPK activity. We previously made a similar observation with EGF and NDF (
Graus‐Porta et al., 1995), which correlated with a decreased ligand affinity due to an accelerated off‐rate (
Karunagaran et al., 1996). Thus, it is reasonable to suspect that ErbB‐2‐containing heterodimers are not only the high affinity receptors for EGF and NDF, but for all EGF‐related growth factors.
HB‐EGF, although to a lesser extent than NDF and BTC, is also able to elevate tyrosine phosphorylation of ErbB‐4 (
Beerli and Hynes, 1996;
Figure 1B). Binding of an EGF agonist to ErbB‐1 could theoretically trigger ErbB‐1–ErbB‐4 dimers and subsequently activate ErbB‐4. Nevertheless, in T47D cells depleted of ErbB‐1, HB‐EGF, but not EGF, still activated ErbB‐4 to the same extent as in the control cells, which implies the existence of an additional receptor for HB‐EGF. This result is reminiscent of BTC, making it possible that HB‐EGF also binds ErbB‐1 and ErbB‐4. Alternatively, HB‐EGF could bind ErbB‐3, thereby triggering ErbB‐3–ErbB‐4 dimers. Indeed, HB‐EGF activates ErbB‐3 not only in T47D/puro but also in T47D/R1R cells (our unpublished data). In both cases, a requirement for ErbB‐2 for HB‐EGF activation of ErbB‐4 becomes evident in T47D/5R cells. Further studies will be required to elucidate the mechanism by which HB‐EGF activates ErbB‐4.
It is likely that the diversity of the biological responses triggered by the EGF‐related ligands is due to the ability of each different ErbB family member to couple with distinct and specific intracellular signaling pathways (
Di Fiore et al., 1990;
Taverna et al., 1991;
Riese et al., 1995.,
Beerli and Hynes, 1996;
Pinkas‐Kramarski et al., 1996). For instance, ErbB‐3 differs from the other receptors in its ability to directly interact with p85 but not with phospholipase Cγ or GTPase activating protein (
Fedi et al., 1994). On the other hand, ErbB‐1 seems to be the only ErbB receptor able to interact with and phosphorylate the Cbl proto‐oncogene product (
Levkovitz et al., 1996). Moreover, the signal elicited by a receptor heterodimer is not simply defined by addition of the signaling properties of the individual dimerization partners (
Alimandi et al., 1995). Indeed, our results demonstrate that the ability of a specific ErbB protein to recruit downstream signaling molecules is dependent upon its dimerization partner. In particular, all EGF‐related peptides induce ErbB‐1 to interact with the adaptor protein SHC. However, only ErbB‐1 ligands, which activate the receptor mainly by inducing ErbB‐1 homodimers and ErbB‐1–ErbB‐2 heterodimers, but not NDF, which can only activate ErbB‐1 by promoting heterodimers with ErbB‐3 and ErbB‐4, allow for detectable Cbl tyrosine phosphorylation and complex formation with ErbB‐1. This result raises the possibility that ErbB‐1 activated by NDF bears a pattern of phosphorylated sites or has a conformation distinct from ErbB‐1 activated by EGF agonists.
In summary, the results presented in this paper allow us to propose a model of ErbB receptor function (
Figure 7). According to this model, ligand‐induced ErbB receptor dimerization is governed by a strict hierarchy, with ErbB‐2 being the preferred heterodimerization partner of all other ErbB proteins. Thus, NDF receptors readily dimerize with ErbB‐2 but not with ErbB‐1 if all four receptors are present (
Figure 7A, left). Interactions between NDF receptors and ErbB‐1 are only favored in cells engineered to lack cell surface ErbB‐2 (
Figure 7A, right). Similarly, the receptors for EGF and BTC also preferentially interact with ErbB‐2. However, in contrast to NDF receptors, ligand‐activated EGF and BTC receptors have the capability to transactivate ErbB proteins distinct from ErbB‐2. This could at least in part be due to direct crosstalk, as evidenced in cells lacking cell surface ErbB‐2 (
Figure 7B and
C, right). However, the presence of ErbB‐2 dramatically enhances transactivation of NDF receptors by ErbB‐1, demonstrating that ErbB‐2 is involved in lateral transmission of signals (
Figure 7B and
C, left). It is noteworthy that this signaling occurs in a directed manner, since NDF receptors did not transactivate ErbB‐1 in the presence of ErbB‐2 (
Figure 4A). This suggests that NDF receptors form more stable complexes with ErbB‐2 than does ErbB‐1, as already previously suggested (
Chen et al., 1996).
In conclusion, the proposed model of ErbB receptor function provides a possible explanation for the high oncogenic potential of ErbB‐2. Overexpression of ErbB‐2 leading to constitutive activation of its kinase is observed in many human tumors and frequently correlates with more malignant disease (
Hynes and Stern, 1994). Moreover, ErbB‐2 is important not only for growth of ErbB‐2 but also of ErbB‐1 overexpressing tumor cells (
Jannot et al., 1996). Our results suggest that the remarkable transforming potency of ErbB‐2 is due not only to its ability to heterodimerize with the other ErbB receptors, but also to its involvement in lateral signaling.