Yet another “active” pseudokinase, Erb3

Considerable attention has focused recently on the role of pseudokinases that were assumed to be “dead” protein kinases owing to their lack of one or more key catalytic residues. Erb3 was thought to be one of those kinases that lacked catalytic activity. Shi et al. (1) now demonstrate that Erb3 does, in fact, have residual kinase activity under some conditions. They also show that the activity seems to require what is thought to be an inactive kinase conformation. Erb3 thus provides yet another example of the creativity of the kinase fold in being able to adapt in unique ways to provide either highly selective phosphorylation of a specific protein using a unique mechanism as in calcium/calmodulin-dependent serine protein kinase (2), or recruiting another residue to substitute for the missing one as in Wnk (3), or by simply filling the adenine pocket with a hydrophobic side chain, thereby locking the kinase into a solid active-like conformation that serves as a docking scaffold that is independent of ATP as in VRK3 (4).

The structure of Erb3 solved by Shi et al. is virtually identical to the structure that was solved recently by Jura et al. (5), and both groups show that the Erb3 construct that contains either the intact intracellular kinase domain (ICD) or the kinase domain only (TKD) was incapable of autophosphorylation in solution or of phosphorylating substrate peptides. Shi et al., however, go on to demonstrate that when the histidine-tagged Erb3–ICD construct was associated with vesicles containing lipids that contain NTA-Ni head groups, it was now capable of autophosphorylation but still unable to phosphorylate substrate peptides. The activity was approximately 1,000-fold less than the active EGF receptor (EGFR), and the physiologic relevance of this activity remains to be established. However, using the Erb3–TKD to create a heterodimer with Erb3–ICD, the authors demonstrate that the activity represents trans phosphorylation of the heterodimer.

To determine whether the conformation seen in the crystal structure corresponds to the structure that mediates catalysis, the authors did several mutations. To appreciate these mutations, however, and to also appreciate some of the other specific residues that are unique to Erb3, one needs to look more closely at the unique features of this structure compared with a typical active kinase (Fig. 1). Two regions in particular assume a markedly different conformation, the αC Helix and the Activation Segment. Instead of β-strand 9 in the Activation Segment, which is found in active kinases, there is a 3₁₀ helix that corresponds to an inactive state. In addition, the N terminus of the αC Helix in the N-lobe is “melted” to create a loop that is referred to as the β3–αC loop. Because the 3₁₀ helix at the beginning of the Activation Segment was found in other inactive kinases, such as CDK2 (6), NEK2 (7), Src (8), and EGFR (9), it may actually represent a common inactive conformation. Several mutations were made to determine whether the Erb3-specific residues are important for the catalytic activity.

Fig. 1.

Erb3-specific residues contribute to a conformation that is different from a typical active kinase. There are two structural motifs that are novel to Erb3, in contrast to an active kinase. One is the 3₁₀ helix in the activation segment of the C-lobe (dark red). The other, in the N-lobe, is the truncated αC Helix (teal) and the resulting linker (bright red) that joins it to β strand 3. Phe⁷³⁴, His⁷⁴⁰, and Asp⁸³⁸ are all residues that are unique to Erb3, and each contributes to this conformation in specific ways. Also shown is how the N-lobe is anchored through the β3–αC loop to the N-terminal linker of a symmetry-related molecule.

Phe⁷³⁴ is located in the β3–αC loop and seems to help nucleate a unique hydrophobic interface with the 3₁₀ helix in the Activation Segment. Mutation of Phe⁷³⁴ results in loss of the residual autophosphorylation activity, which is consistent with the interpretation that Phe⁷³⁴ is important for stabilizing the active conformation. Two other residues in the 3₁₀ helix, Val⁸³⁶ and Leu⁸³⁹, also contribute to this hydrophobic interface. Mutation of these residues also abolishes the activity. However, one cannot assume that the observed conformation of the activation loop and the αC Helix is stable in solution. In fact, if one looks at the crystal structure, it is clear that the position of the Activation Loop and the β3–αC loop is stabilized not only by the hydrophobic residues but also by the N-terminal linker of the symmetry-related molecule (Fig. 1). This crystal packing creates a symmetry-related head-to-head dimmer, whereas crystal packing to another monomer creates an extended polymer of these head-to-head dimers. Obviously it is not known whether this conformation is stabilized in the same way when the protein is anchored to the membrane by its N-terminal His-tag. So far these are the only conditions whereby activity is observed.

Although the authors focus on the hydrophobic interactions, it is interesting to examine more comprehensively the sequence changes that are unique to Erb3 compared with the other active versions of the EGF receptor (EGFR, Erb2, and Erb4). There are numerous such changes, and all seem to reinforce in different ways this conformation that is seen in the crystal structure. Activation of a kinase requires optimization of the C-lobe, and this is usually achieved by phosphorylation of the activation loop. However, the N-lobe also needs to be reorganized so that it cannot only bind ATP but also transfer the γ-phosphate to a protein substrate. Key residues that mediate these changes in both the N- and C-lobes are also missing in Erb3. The primary mechanism for organizing the N-lobe for catalysis is to couple the αC Helix to β strand 3 and the glycine-rich loop through a conserved lysine in β3 and a conserved glutamate in αC. The other requirement is that the N terminus of the C helix interacts with the activation segment in a way that allows the cleft to open and close. Lys⁷²³ in β3 is conserved in Erb3, but the Glu in Erb3 is changed to His so that His⁷⁴⁰ is actually repelled away from β3. This can be thought of as a “loss-of-function” mutation. Equally interesting, however, is to consider what the consequence is in terms of a “gain of function” for Erb3. Like Phe³⁷⁴, His⁷⁴⁰ now is also used to stabilize this unique conformation by interacting with the carbonyl of glycine from the conserved Asp-Phe-Gly (DFG) motif at the beginning of the 3₁₀ helix. Another key difference in the Erb3 sequence is Asp⁸³⁸, which is in the P+5 position relative to the DFG aspartate. In conventional kinases, including EGFR, this Asp is either an Arg or a Lys, and it binds to the phosphorylated residue in the activation loop. This interaction is critical for stabilizing the active conformation of the C-lobe. Because in Erb3 the essential Tyr that gets phosphorylated is missing, the basic residue is no longer needed. In this structure, however, Asp⁸³⁸ gains a critical new function—binding to the arginine in the conserved His-Arg-Asp (HRD) motif (Arg⁸¹⁴). Thus, two Erb3-specific residues, His⁷⁴⁰ and Asp⁸³⁸, stabilize in a unique way the 3₁₀ helix conformation that we see here in Erb3.

Another interesting feature of this Erb3 structure is the way that a nonhydrolyzable ATP analog, adenylyl imidodiphosphate (AMP-PNP) and Mg²⁺ are bound. This conformation is different from what is thought to be optimal for catalysis but similar to what is seen in several other inactive tyrosine kinases, such as Src and EGFR (10, 11). There is only one metal ion bound, and it is assumed to be Mg because that is that was included in the crystallization buffer. Jura et al. (5) observed that the density in their structure was too much to be accounted for Mg, but those authors were not able to identify the ion. This is also probably true for the Shi et al. structure, because their density also seems to be too strong for Mg. Using mant-ATP, a fluorescent ATP analog, Shi et al. also show that Erb3 binds ATP quite well. In fact, the K_d is 1 μM. This is at least 10-fold lower than what is seen for EGFR and suggests that ATP may be a required cofactor for stabilizing this structure. Certainly it is required for catalysis, but this structure of Erb3 may also be less stable in the absence of ATP. To further explore the ATP binding site, the authors mutated the conserved Lys⁷²³, and this abolished both binding of ATP and catalysis. This also reflects on the novelty of this structure because mutation of this Lys does not typically abolish ATP binding in other kinases such as Erk2 and protein kinase A; it simply interferes with the ability of the N-lobe to position the γ-phosphate of ATP for transfer (12, 13). Another Erb3-specific residue is Asn⁸¹⁵, which is usually an Asp. Restoring the Asp does not, however, affect the autophosphorylation activity. This is also consistent with previous proposed mechanisms whereby the conserved carboxylate is not actually thought to be a strong catalytic base but rather serves as a simple “proton trap” (14), a function that the Asn could fulfill as well.

Shi et al. (1) also use quantum mechanics/molecular mechanics simulations in an effort to explore what mechanism might be used by this unusual kinase conformation. Given, however, that we do not have a good mechanistic understanding of most protein kinases, let alone the EGFR, it is unclear whether discussion of associative or dissociative mechanisms is relevant. The ATP binding site needs to be studied more carefully, especially with regard to the metal ion; however, it is clear that the associative vs. dissociative studies of other metabolic kinases such as hexokinase or DNA polymerases are different from the mechanism that is used by the protein kinases. In these other kinases or phosphotransferases the phosphate is typically anchored at the base of the cleft where closing of the cleft facilitates transfer of the phosphate as proposed initially by Steitz et al. for hexokinase (15). In the protein kinases, however, it is the base of the nucleotide that links the N-lobe and the C-lobe, creating a catalytic spine that then mediates the transfer of the phosphate (16). This is also true for the eukaryotic-like kinases (ELKs), which use the same catalytic machinery as the eukaryotic protein kinases but lack the complex mechanism of regulation by phosphorylation (17). In this regard Erb3 seems to more closely resemble the ELKs. If this crystal structure of Erb3 does reflect the structure of the protein in solution, then it is clear that a different mechanism would certainly be required. In any case, this structure of Erb3 and the demonstration that it has residual catalytic activity highlights and challenges once again the assumption that kinases lacking key catalytic residues are truly “dead” and points to the importance of carefully examining each of these pseudokinases to determine which of their residual activities are actually still there but modified in unique ways.