The phosphatases of regenerating liver (PRLs) are highly over-expressed in metastatic cancers yet their mechanism of action is poorly understood. The PRL family belongs to the protein tyrosine phosphatase superfamily and is comprised of closely related proteins PRL1, PRL2, and PRL3. They all contain a C-terminal prenylation site and a single catalytic domain of roughly 170 amino acids. Like other phosphatases, their activity occurs through a two-step catalytic cycle involving the transient phosphorylation of a catalytic cysteine residue. In PRLs, this intermediate is extremely long-lived leading to the accumulation of a cysteine-phosphorylated form of the enzyme [1,2]. Recently, CNNM proteins, a family of membrane proteins involved in magnesium homeostasis, were identified as PRL-binding partners [3]. Disruption of the PRL-CNNM interaction promotes tumor formation and invasiveness in animal and cellular models, strongly suggesting that the physiological function of PRLs is to regulate CNNM magnesium transport.
We determined crystal structures of PRLs bound to the CBS-pair domain of CNNM3. The CBS-pair domain is present as a dimer in the head-to-head arrangement that is typical for CBS-pair domains. The CNNM3 CBS-pair domain contains a long loop interacting with PRL catalytic site. The side chain of CNNM3 Asp426 sits in the catalytic site and likely mimics the phosphate of a bound substrate. Mutagenesis showed that Asp426 is required for high affinity binding, suggesting that the CBS-pair domain might act as a pseudo-substrate. The structures reveal why disulfide formation and catalytic cysteine phosphorylation dramatically decrease binding affinity. Addition of the CBS-pair domain inhibited phosphatase activity and CNNM3 binding was blocked by phosphorylation of the PRL active site cysteine. We used isothermal titration calorimetry (ITC) and mutagenesis to probe the importance of PRL residues for CNNM binding [4]. Comparison of binding activity and in vitro phosphatase activity shows that they are strongly correlated with the notable exception of the PRL3 R138E mutant, which showed weak CNNM3 binding but normal phosphatase activity. We also show that anticancer targets thienopyridone and iminothienopyridinedione inhibit PRLs through oxidation of catalytic cysteine [5]. Our results support the hypothesis that PRLs function as pseudophosphatases in regulating the action of CNNM proteins in cancer.