Rethinking Pseudokinases

Abstract

Pseudokinases lack conservation of one or more of the catalytic residues in the kinase core and as a consequence are typically thought to be catalytically inactive. New work by Mukherjee et al. (2008) challenges this assumption. They show that the pseudokinase domain of CASK (Ca²⁺/calmodulin activated serine-threonine kinase) adopts an active conformation and displays catalytic activity in vivo.

Protein kinases are critical for the regulation of most biological events. In eukaryotes, the protein kinase superfamily (the kinome) typically accounts for approximately 2% of all genes in a species (Manning et al., 2002). Embedded within the core of each kinase is a set of conserved residues that are essential for catalysis. However, many kinases (nearly 10% of the human kinome) lack one or more of these residues and are therefore classified as pseudokinases. Pseudokinases are considered to be catalytically inactive—vestigial remnants of active kinases that are relegated to mere scaffolds. This view, however, is challenged by Mukherjee et al. (2008) in this issue of Cell. They show that despite lacking key catalytic residues, the pseudokinase domain of CASK (Ca²⁺/calmodulin-activated serine-threonine kinase) displays catalytic activity under physiological conditions. This study raises fundamental questions regarding the adaptability of the phosphoryl transfer mechanism, the regulatory role of Mg²⁺ ions, and the functional roles of psuedokinases.

CASK is a multidomain pseudokinase that is targeted to synapses of neurons by the anchoring of its PDZ domain to membrane-associated proteins such as neurexin. CASK also has SH3 and guanylate kinase domains. Mukherjee et al. determined the crystal structure of the pseudokinase domain of CASK and show that it has a typical protein kinase fold that assumes an active conformation. The authors also show that CASK can bind to ATP despite lacking two conserved residues that coordinate Mg²⁺ ions in typical kinases (Figure 1). In contrast to other kinases, CASK binds to free ATP, and Mg²⁺ ions are inhibitory. A low concentration of Mg²⁺ ions at the synapse could therefore optimize CASK’S catalytic activity. The authors demonstrate that CASK undergoes autophosphorylation and that the C terminus of neurexin is also a CASK substrate, both in vivo and in vitro. However, the functional consequences of these phosphorylation events remain unknown.

Figure 1. Catalytic Versatility of the Protein Kinase Core.

(Left) The active site of protein kinase A (PKA) is shown in a conformation that resembles a transition state (Madhusudan et al., 2002). The γ phosphate is trapped between the glycine-rich loop and Lys168 in the catalytic loop. The two Mg²⁺ ions are anchored by Asn171 in the catalytic loop and Asp185 in the DFG motif. (Right) The active site of CASK is shown bound to the ATP analog AMP-PNP; only the α phosphate is visible. Lys72 is conserved to anchor the α /β phosphates and the catalytic site (Lys168, Asp166) is poised to accept a substrate hydroxyl moiety. The critical backbone amide of Ser53 in PKA is replaced in CASK with a Pro, which provides a rigid cap but not a hydrogen bond. A histidine residue in the catalytic loop points toward the phosphates and can help to neutralize the charge in the ATP.

The protein kinase core is a remarkable catalyst in which the N- and C-lobes converge to bind ATP and then transfer the γ phosphate to a protein substrate (Figure 1). By docking to a hydrophobic pocket at the base of the active site cleft, the adenine ring of ATP brings the two lobes together and nucleates a highly dynamic allosteric network. The two lobes must then trap the γ phosphate of ATP and position it for transfer to the protein substrate. Typically, the γ phosphate is trapped between the glycine-rich loop in the N-lobe and a lysine in the catalytic loop in the C-lobe (Madhusudan et al., 2002). The phosphates are positioned and neutralized by two Mg²⁺ ions, which are bound by an asparagine residue in the catalytic loop and an aspartic acid in the DFG loop (Figure 1). Both residues are missing in CASK. Thus, although the adenine pocket is conserved, CASK must use a different mechanism to position the phosphates, neutralize their negative charge, and transfer the γ phosphate.

CASK belongs to the Ca²⁺/calmodulin-dependent protein kinase group. Thus, one question to ask is what sequence and structural features distinguish CASK from active Ca²⁺/calmodulin-dependent protein kinases, and how might these features compensate for the missing residues and contribute to CASK’S unusual mode of phosphoryl transfer? As Figure 1 shows, the highly conserved hydrogen bond donor in the glycine-rich loop has been replaced strategically with a proline, which is not capable of hydrogen bonding. This unique feature of CASK provides an unusual mechanism for ATP positioning and phosphoryl transfer. In addition, instead of the negatively charged DFG loop, which would repel free ATP, CASK has a very flexible loop, GGFG. Flow the γ phosphate is positioned for transfer to neurexin and how its charge is neutralized remain unanswered questions. CASK could possibly use an arginine from the substrate in a manner similar to the “arginine finger” involved in Ras-GTPase activation in which the arginine that anchors the γ phosphate of GTP comes from another protein (Scheffzek et al., 1997).

Comparative genomics provides a rich source of clues about pseudokinase functions (Kannan et al., 2007). For instance, identifying residues that vary along with catalytic residues can point to alternative modes of ATP binding and phosphoryl transfer in individual pseudokinase families. This has been illustrated nicely in the Wnk family (Min et al., 2004), previously designated as pseudokinases, and the choline and aminoglycoside kinases (CAKs), which are eukaryotic-like kinases found in microbes (Kannan et al., 2007). Although both lack a key ATP-positioning lysine (equivalent to K72 of protein kinase A in Figure 1), other co-conserved residues compensate for the missing lysine and thereby facilitate alternative modes of ATP binding and phosphoryl transfer. Given that many eukaryotic-like kinases lack one or more catalytic residues, characterizing them may shed light on how unusual modes of phosphoryl transfer evolved in eukaryotic pseudokinases.

Mukherjee et al. also point out the potential regulatory role of Mg²⁺ ions. CASK binds to free ATP, and high levels of Mg²⁺ inhibit kinase activity, most likely by reducing the availability of ATP that is not complexed with Mg²⁺. The synapse may be a unique environment with low levels of Mg²⁺, although localized fluxes in Mg²⁺ ions may be found in other intracellular microenvironments. Previous studies certainly have shown that Mg²⁺ ions can serve as a regulator of kinase activity. The catalytic activities of PKA and ERK kinases, for example, are both regulated by a second Mg²⁺ ion; PKA is inhibited (Adams and Taylor, 1993), whereas ERK is activated (Waas and Dalby, 2003). In addition, inhibition of PKA by pseudosubstrate inhibitors such as PKI and Rlα is only achieved when Mg²⁺ ion levels are high thus creating the potential for cAMP-independent activation of PKA (Zimmermann et al., 2008).

Prior to the Mukherjee et al. study under discussion here, many of the pseudokinases were viewed simply as scaffolds. Clearly CASK with its many domains, especially its PDZ domain, does serve as a scaffold protein and plays an important role in protein targeting independent of its kinase activity. Indeed CASK does not phosphorylate neurexin, in vivo or in vitro, without its PDZ domain; recruitment of the substrate is essential. However, all kinase cores are scaffolds as well as catalysts; they bind to proteins, not peptides, and the tethering of substrates/inhibitors is typically mediated by the C-lobe. There is growing evidence that pseudokinase domains can also play important roles as scaffolds by interacting with functional kinases (reviewed in Boudeau et al., 2006). For instance, in the HER3 receptor tyrosine kinase subfamily, the inactive catalytic domain of HER3 activates other members of the EGF receptor family (HERI, HER2, and HER4) by forming heterodimeric complexes. Similarly, the pseudokinase domain of STRAD directly interacts with a functional kinase, LKB1, to increase its catalytic activity by 100-fold. In the JAK kinases, a pseudokinase is fused to a conventional active kinase. But are these pseudokinases truly inactive? Given nature’s ability to innovate new and alternative modes of phosphoryl transfer, each pseudokinase needs to be revisited. It is likely that we will not fully appreciate their functional roles until we look at each kinase in the context of its interacting partners.