Novel Anticancer Agents Based on Targeting the Trimer Interface of the PRL Phosphatase

Reversible and coordinated protein tyrosine phosphorylation is central to diverse signal pathways regulating cell growth, migration, and survival. Disturbance of the normal pattern of tyrosine phosphorylation, due to perturbed balance between the activities of protein tyrosine kinases (PTK) and protein tyrosine phosphatases (PTP), causes abnormal cell signaling and has been linked to the etiology of many human diseases including cancer (1). Thus, there is vast interest in targeting dysfunctional pathways driven by aberrant tyrosine phosphorylation for therapeutic interventions. Notable success has been achieved by targeting the PTKs, as shown by the more than two-dozen small molecule inhibitors already in the clinic (2). However, resistance to kinase inhibitor treatments prevents durable responses. Therefore, there is heightened interest to modulate disease progression at the level of PTPs.

The phosphatase of regenerating liver (PRL) phosphatases constitute a unique group of PTPs, with three closely related members (PRL1, 2, and 3; refs. 3–6). Unlike other PTPs, the PRLs function as positive signal transducers capable of activating both ERK1/2 (7–11) and Akt (12–15), two of the major pathways that are aberrantly upregulated in cancer (16, 17). PRL1 was initially identified as an immediate early gene induced during liver regeneration upon partial hepatectomy (18). Subsequent studies found that exogenous expression of PRLs accelerates cell proliferation and anchorage-independent growth (7, 18–21). Constitutive PRL expression also promotes cell migration and invasion (7, 8, 11, 22–25). Moreover, PRL overexpressing cells form tumors with high metastatic potential when injected into mice (9, 22, 23, 26), whereas PRL knockdown reduces cell proliferation and migration as well as tumorigenesis in vivo (9, 11, 25, 27–30). Most significantly, PRL level is elevated in human cancers of colon (31, 32), liver (23, 33), ovarian (27, 34), prostate (35), gastric (36, 37), pancreatic (13), and breast (9, 38), as well as in melanoma (20, 39), multiple myeloma (40) and acute myeloid leukemia (41, 42), and PRL overexpression strongly correlates with late-stage metastasis and poor clinical outcomes. Taken together, the data implicate PRLs as novel molecular markers and therapeutic targets for metastatic cancers. Consequently, PRLs have garnered considerable interest for drug discovery (6). Unfortunately, the rather flat PRL active site and its structural similarity to other members of the PTP family present significant challenge for PRL inhibitor design. Indeed, reported active site directed PRL inhibitors are neither sufficiently potent nor selective, and so are not suitable for in vivo pharmacologic study and therapeutic development (6).

We describe a novel approach to inhibit PRL function by targeting a unique structural and regulatory property of the PRLs. One of the most striking features of PRL1 is that it exists as a trimer in the crystalline state and has a high propensity to form trimer in solution and inside the cell (8, 10, 43, 44). Moreover, trimer formation is essential for PRL1-mediated cell growth and migration, suggesting that small molecules targeting the trimeric interface of PRLs could potentially have therapeutic value (8). To capitalize on these findings, we used a computer-based virtual screen to search the available chemical databases for compounds capable of disrupting PRL trimerization. Biochemical and structural analyses demonstrate that Cmpd-43 and its close analogs bind the PRL1 trimer interface and block PRL1 trimerization. Cmpd-43 also specifically abrogates the PRL1-induced cell proliferation and migration through attenuation of both ERK1/2 and Akt activity. Importantly, Cmpd-43 exhibits excellent anticancer activity both in vitro and in a xenograft mouse model of melanoma. The study provides pharmacological validation that trimerization is important for PRL1 function and targeting PRL trimerization is a viable approach for therapeutic development.

Anti-HA, anti-tubulin, and anti-GAPDH antibodies were purchased from Santa Cruz Biotechnology. Anti-ERK1/2, anti-pERK1/2 (Thr202/Tyr204), anti-Akt, anti-pAkt (Ser473), and anti-LSD1 antibodies were obtained from Cell Signaling Technology. DMEM, FBS, penicillin, and streptomycin were from Invitrogen. HEK293, MeWo, and MCF7 cell lines were purchased directly from ATCC between 2008 and 2015. The ATCC cell lines were characterized by short tandem repeat (STR) DNA profiling. MCF10A cell was received as a gift from Dr. Mircea Ivan's lab in Indiana University School of Medicine (Indianapolis, Indiana), and was authenticated by morphology. All cell lines were passaged for fewer than 6 months after resuscitation.

Asinex and ChemBridge subsets in ZINC (45) database were downloaded from ZINC website (http://zinc.docking.org) and used for virtual screening. The monomer B in PRL1 trimer structure (PDBID: 1ZCK; ref. 44) was used as receptor, and the coordinates were retrieved from the Protein Data Bank. The DOCK6.2 program (46) was used for rigid docking to generate a potential subset of molecules binding at PRL1 trimer interface, and then AutoDock4.01 software (47) was used for flexible docking to get the most potent hits.

In the first-stage docking, the structure of monomer B was processed using the “Dock Prep” module in UCSF CHIMERA, then the docking region was defined through a standard pipeline of running dms, sphgen, sphere_selector, and showbox program, and the energy scoring values were calculated by grid program. About 560,000 small molecules (downloaded in 28 mol2 files, ∼20,000 molecules in each file) were submitted to the dock6.mpi program to perform docking calculations simultaneously. During each docking, the small molecule was positioned with 1,000 orientations, the lowest interaction energy and corresponding conformation was recorded. All ligands in each mol2 file were ranked according to their lowest interaction energy, and the top 2,000 were kept for the second-stage docking, thus 56,000 (2,000 × 28) molecules were picked out for next-stage screening.

In the second-stage docking, the structure of monomer B was processed in AutoDockTools1.4 software, Gasteiger charge was added and nonpolar hydrogens were merged. The docking area was designated around the BA or BC interface, and the energy grids of 51 × 51 × 73 points with 0.375 Å spacing on each axis were calculated for 17 atom types (H, HD, HS, C, A, N, NA, NS, OA, OS, F, P, SA, S, Cl, Br, and I), as well as the electrostatic and desolvation potential using autogrid4 program. On the contrary, each ligand structure was used to generate pdbqt and dpf files using prepare_ligand4.py and prepare_dpf4.py scripts. Based on these prepared files, molecular docking was carried out in autodock4 program as follows: 10 separate docking runs were performed for each ligand. In each docking run, the optimal binding conformation was achieved by Lamarckian Genetic Algorithm with Local Search (LGALS) method. After all ligands were docked, the lowest binding free energy of each ligand was extracted and ranked, and hit molecules were picked out through binding free energy comparisons, structure similarity analyses and binding mode inspections.

HEK293, MeWo, and MCF7 cells were grown in DMEM supplemented with 10% FBS, penicillin (50 units/mL), and streptomycin (50 μg/mL) in a 37°C incubator containing 5% CO₂. MCF10A cells were grown in Mammary Epithelial Basal Medium supplemented with MEGM Single Quots and 100 ng/mL cholera toxin (Lonza, Basel, Switzerland). HEK293 cells were seeded at 40% confluence in antibiotic-free medium and grown overnight. Transfection was performed using Lipofectamine 2000 from Invitrogen according to the manufacturer's recommendations.

Cells were grown to 90% confluence in a 12-well plate at 37°C in an atmosphere of 5% CO₂. A wound was created by scratching cells with a sterile 200 μL pipette tip. Cells were washed with PBS to remove the floating cells, and then treated with fresh medium containing 20 μmol/L compound or DMSO. The wounds were photographed at 0 and 24 hours under ×10 magnitude microscope. Wound healing magnitude was quantified by measuring the relative wound closure compared with control cells at 24 hours.

Cells were seeded in a 96-multiwell plate (3,000 cells/well) containing DMEM, 10% FBS at 37°C in an atmosphere of 5% CO₂ overnight. Cells were then treated with various concentrations of compounds or DMSO for 24 and 48 hours. Cell proliferation was then determined by MTT assay as described previously (8) using a multiwall spectrophotometer. Data are presented as relative proliferation rate compared with control cells.

Cell migration was determined as described previously (10) with some modifications. The assay was performed with Transwells (6.5 mm diameter; 8 μm pore size polycarbonate membrane) obtained from Corning Costar. Cells (3.75 × 10⁵) in 1.5 mL of serum-free medium were placed in the upper chamber, whereas the lower chamber was loaded with 2.5 mL of medium containing 10% FBS. Cells were then treated with 10 μmol/L of different compounds as indicated. After 24-hour incubation (37°C, 5% CO₂), the total number of cells that had migrated into the lower chamber was counted with a hemacytometer. Data are presented as relative migration rate compared with control cells. Cell motility of MeWo cells was also measured using live-cell imaging. 5 × 10³/well of MeWo cells were seeded in a 96-well plate for overnight and then treated with 5 μmol/L of Cmpd-43 for 4 hours. One μg/mL of Hoechst 33342 was used to label the nuclei, and Thermo Scientific ArrayScan XTI Live High Content Platform was then used for live-cell tracking to measure the motility in the presence of Cmpd-43. Motility of the cells was assessed over 6 hours and image data were collected every 30 minutes.

NSG (NOD/scid/IL2Rg^null) mice were purchased from In Vivo Therapeutics Core at Indiana University Simon Cancer Center. Experiments on mice were carried out in accordance with the regulations of The Institutional Animal Care and Use Committees at Indiana University. All mice were housed under pathogen-free conditions in the animal facility and received autoclaved water and food. Ten to twelve weeks old NSG mice were used in the study. MeWo cells were suspended in PBS at 8 × 10⁷ cells/mL. A total of 8 × 10⁶ cells (100 μL) were subcutaneously implanted into both left and right flank (n = 24) using a 27-gauge needle. Once the tumor volume reaches 200 mm³, daily intraperitoneal injection of either control or 30 mg/kg Cmpd-43 was performed, and the tumor growth was monitored for 3 weeks. Tumor volume was calculated using the formula V = (W² × L)/2 for caliper measurements. Mice were sacrificed after injection for 21 days, and organs were collected for immunohistological and biochemical analysis.

For cell-based proliferation, migration, and wound healing assays, the Student t test was used to measure the significance. For MeWo cell xenograft tumors, tumor volumes at different time and final tumor weights were compared using the Student t test. In comparing the mRNA level of PRLs in human normal skin and melanoma samples, Student t test was used to assess the significance of differences between groups. Survival analysis was performed according to the Kaplan–Meier method and the log-rank test, a P-value of less than 0.05 was considered statistically significant.

Given the functional requirement of PRL trimerization, disruption of PRL trimerization was proposed as a potential therapeutic approach for PRL-based drug discovery (8), but this strategy has not been validated with pharmacologic approaches. As revealed by the homotrimeric PRL1 crystal structure (44), each PRL1 monomer (e.g., monomer B) has two dimer interfaces, namely the BA (residues from 125 to 150) and BC interfaces (residues from 11 to 18, 36 to 41, and 92 to 98), which are 18 and 19 Å away from the active site Cys104 in the catalytic P-loop C₁₀₄VAGLGR₁₁₀ (Fig. 1A). To discover small molecules capable of blocking PRL1 trimerization, we used structure-based virtual screening to identify compounds that bind to the dimer interfaces in each PRL1 monomer. We employed a sequential screening strategy, starting with rigid docking in DOCK6.2 (46) to sample a total of 560,000 compounds (Asinex and ChemBridge subsets in the ZINC database; ref. 45) to each dimer interface and score protein–ligand complexes based on the calculated interaction energies, which was followed by flexible docking in AutoDock4.01 (47) to analyze the top 10% hits obtained from rigid docking (Fig. 1B). This process led to the selection of 100 top-ranked compounds for each interface. Upon further binding mode verification and structural similarity analyses, 56 structurally diverse compounds were purchased for further experimental evaluation.

The ability of the 56 compounds to disrupt PRL1 trimerization was initially assessed by in vitro cross-linking experiments using recombinant PRL1 protein (44). Ten of the 56 compounds exhibited significant activity in blocking PRL1 trimerization. To further confirm the efficiency of these compounds to disrupt PRL1 trimer formation, we also evaluated them in an in vivo cell-based cross-linking assay. HA-tagged PRL1 expressing HEK293 cells were treated with the compounds, fixed with 1% formaldehyde, and the HA-tagged PRL1 was immunoprecipitated with HA antibodies, analyzed by SDS-PAGE, and visualized by immunoblotting with anti-HA antibodies. As shown in Fig. 1C, the top three compounds, Cmpd-3, Cmpd-26, and Cmpd-43, significantly decreased PRL1 trimer formation inside the cell, with Cmpd-43 being the most potent PRL1 detrimerizer (Fig. 1D). Importantly, Cmpd-43 at 20 μmol/L did not inhibit the phosphatase activity of PRL1 as well as a large panel of PTPs including receptor-like PTPs, PTPμ, PTPϵ, LAR, PTPσ and PTPγ, cytosolic PTPs, PTP1B, Lyp, SHP1, PTPH1, HePTP, STEP, and PEZ, the dual specificity phosphatase VHR, VHZ, MKP5, CDC14A, and the low molecular weight PTP.

Given that PRL1 trimerization is essential for the PRL1-mediated cell proliferation and migration (8), Cmpd-43 is expected to suppress both cellular processes if it disrupts PRL1 trimerization inside the cell. To test this hypothesis, we determined the effect of Cmpd-43 on cell proliferation and migration in PRL1-expressing HEK293 cells. As expected, Cmpd-43 inhibited PRL1 induced cell proliferation in a dose-dependent manner (Fig. 1E). In addition, Cmpd-43 also markedly delayed the wound closure induced by PRL1 overexpression (Fig. 1F). To delineate the structural features of Cmpd-43 important for inhibiting PRL1-mediated cellular processes, a series of Cmpd-43 derivatives were either purchased (Analogs 1 to 4) or synthesized (Analogs 5 to 7; Fig. 2A). As shown in Fig. 2B and C, Analog-3 displayed similar efficacy as Cmpd-43 in attenuating PRL1 induced cell proliferation and migration, whereas Analogs 2 and 4–7 appeared slightly less effective than Cmpd-43. Interestingly, Analog-1 exerted no effect on either cell proliferation or migration, suggesting that the iminomethyl-aromatic moiety, which is missing in Analog-1, is critical for the inhibitory activity of Cmpd-43 and the other 6 analogs. Collectively, through a two-stage virtual screening strategy, biochemical and cell-based evaluation, and a limited structure and activity analysis, we identified Cmpd-43 and several analogs as potential disruptors of PRL1 trimerization. We also found a structurally related but inactive Analog 1, which could serve as a negative control in mechanistic studies.

Before Cmpd-43 can be used as a chemical probe to address PRL1's roles in normal physiology and in cancer and serve as a lead for therapeutic development, it is important to establish whether Cmpd-43 exerts its effect inside the cell through disruption of PRL1 trimerization and inhibition of PRL1-mediated signaling. To this end, we first compared the effect of Cmpd-43 and its inactive Analog-1 on PRL1 trimerization. As shown in Fig. 3A, Cmpd-43 effectively blocked PRL1 trimerization in HEK293 cells whereas at the same concentration Analog-1 had no effect, consistent with its lack of inhibition in PRL1-mediated cell proliferation and migration (Fig. 2B and C). To further evaluate the specificity of Cmpd-43, we utilized a trimerization deficient mutant PRL1/G97R, which is incapable of promoting cell growth and migration (8). Thus, Cmpd-43 would not be expected to affect the growth and migration of PRL1/G97R-expressing cells if its main mode of action is disrupting PRL1 trimerization. We first confirmed that PRL1/G97R was defective in trimer formation (Fig. 3B). We also confirmed that although ectopic expression of PRL1 increased both cell proliferation and migration, the rates for the PRL1/G97R cells were similar to those of the vector control cells (Fig. 3C and D), again validating the functional importance of PRL1 trimerization. We then measured the effect of Cmpd-43 and Analog-1 on the proliferation and migration of both wild-type PRL1 and PRL1/G97R expressing cells. As expected, treatment with Cmpd-43, but not Analog-1, attenuated PRL1-induced cell proliferation in a dose-dependent manner, whereas neither Analog-1 nor Cmpd-43 had any effect on PRL1/G97R-expressing cell proliferation (Fig. 3C). Similarly, Cmpd-43, but not Analog-1, was capable of reducing wild-type PRL1-induced cell migration, whereas neither Cmpd-43 nor Analog-1 was able to alter the cell migration behavior of the PRL1/G97R-expressing cells (Fig. 3D). Finally, we evaluated Cmpd-43 in mouse embryo fibroblast (MEF) derived from either wild-type or PRL1-deficient mice. As expected, Cmpd-43 preferentially inhibited wild-type over PRL1^−/− MEF cells (Supplementary Fig. S1), indicating that Cmpd-43 exerted its antiproliferative activity through blocking PRL1 trimerization.

PRL1 promotes cell proliferation and migration through activation of ERK1/2 and Akt pathways (8, 13, 24). To delineate the biochemical mechanism by which Cmpd-43 exerts its inhibitory activity on cell growth and migration, we analyzed the effect of Cmpd-43 and Analog-1 on ERK1/2 and Akt activity in both wild-type PRL1 and the trimerization impaired PRL1/G97R mutant expressing cells. Consistent with the results from the phenotypic assays, expression of PRL1 increased ERK1/2 and Akt activity by 3.4- and 2.5-fold, respectively, whereas the activation status of ERK1/2 and Akt in PRL1/G97R cells was similar to that of the vector control cells (Fig. 3E). As expected, Cmpd-43 effectively abrogated the PRL1-induced ERK1/2 and Akt activation whereas the negative control Analog-1 had no effect on ERK/1/2 and Akt activity. In line with PRL1/G97R being a loss of function mutant, neither Cmpd-43 nor Analog-1 had any effect on ERK1/2 and Akt signaling in PRL1/G97R cells. Collectively, these mechanistic studies provide additional strong evidence that Cmpd-43 inhibits PRL1-mediated cellular signaling as well as cell proliferation and migration by blocking PRL1 trimerization.

To provide direct evidence that Cmpd-43 binds to the PRL1 trimer interfaces and to determine the molecular basis of PRL1 detrimerization by Cmpd-43, we sought to co-crystallize PRL1 with Cmpd-43 as well as Analog-3. We obtained co-crystals of PRL1 bound with Analog-3. The 3D structure of PRL1Analog-3 complex was solved by molecular replacement using monomer A in the PRL1 trimer structure (PDBID: 1ZCK; ref. 44) as a search model and refined to 1.90 Å resolution. The final atomic model encompasses residues 4-160 of PRL1 and the intact Analog-3, which is unambiguously identified by the unbiased F_o − F_c omit density map (Fig. 4A). The details of data collection and structure refinement are summarized in Table 1. The complex structure belongs to the C222₁ space group with one PRL1 molecule per asymmetric unit. Remarkably, although a homotrimeric arrangement was always observed in previous crystal structures of wild-type PRL1 (43, 44), the PRL1/C104S mutant in complex with sulfate in the active site (44), and PRL1 in complex with a peptide ligand (10), the PRL1Analog-3 complex crystallized as a monomer. The overall structure of PRL1Analog-3 is quite similar to the initial search model used for molecular replacement. PRL1 adopts a compact α + β structure comprising a central five-stranded β sheet surrounded by four α helixes on one side and two α helixes on the other side (Fig. 4A). The PTP signature motif (C₁₀₄VAGLGR₁₁₀) forms a loop (P-loop) between β5 and α4 located at the base of the active site pocket. The binding site for Analog-3 is situated at the backside of the PRL1 active site, which is defined by residues within the α5 helix, α4–α5 loop, β1–β2 hairpin, and the C-terminus (Fig. 4A). Interestingly, residues involved in binding Analog-3 come from both the BC- and BA-dimer interfaces in the resolved complex crystal structure, with the majority of the contact area in the BC interface (Fig. 4B).

Figure 4C shows the detailed interactions between PRL1 and Analog-3. The dimethyl-isoindoline moiety is placed within a hydrophobic pocket defined by Tyr14, Met124, Phe132, and the aliphatic carbon atoms in Asp128. Specifically, the two methyl groups interact with Asp128 and Phe132 respectively, whereas the isoindoline moiety makes contacts with Tyr14, Met124, and Phe132. The adjacent benzohydrazide motif has several van der Waals contacts with residues Thr13, Tyr14, and Lys15, and the oxygen atom provides an additional polar interaction with terminal amine of Lys15. The furan ring extends into a cavity constituted by Lys15, Asn16, and Arg159, making van der Waals interactions with these residues as well as a polar interaction between the oxygen atom and the side chain of Asn16.

To further substantiate the molecular interactions between PRL1 and Analog-3, we mutated Tyr14 and Phe132, which have strong hydrophobic contacts with Analog-3 (Fig. 4C). Analyses of the buried surface area in the dimer interfaces indicated that these two residues make very limited, if any, contribution to PRL1 trimerization. Thus, we predicted that substitutions at Tyr14 and Phe132 would weaken the interaction between PRL1 and Analog-3/Cmpd-43, without interference with PRL1 trimerization. As expected, replacement of Tyr14 and Phe132 by an Ala had no effect on PRL1 trimerization and PRL1-mediated cell migration (Fig. 4D and E). Importantly, the PRL1/Y14A and PRL1/F132A mutants were resistant to Cmpd-43 treatment. Indeed, although Cmpd-43 blocked wild-type PRL1 trimerization and PRL1-mediated cell migration, it failed to inhibit PRL1/Y14A and PRL1/F132A trimer formation and had little effect on PRL1/Y14A- or PRL1/F132A-mediated cell migration (Fig. 4D and E). These results are in complete agreement with the structural observations that residues Tyr14 and Phe132 are involved in binding Analog-3. Taken together, the structural and mutagenesis data provide direct evidence that Analog-3 and Cmpd-43 bind at the PRL1 trimer interface and prevents PRL1 trimerization.

As mentioned in the introduction, PRLs are overexpressed in many tumor cell lines. To investigate the clinical relevance of PRL overexpression and tumor progression, we analyzed the Gene Expression across Normal and Tumor tissue (GENT) database, a publicly available microarray dataset containing more than 34,000 human cancer and normal samples (48). We found that samples from melanoma patients (n = 302) had significantly higher PRL1 and PRL2 mRNA expression compared to normal skin samples (n = 141; Fig. 5A and B). To further evaluate the significance of PRL1 overexpression in predicting survival in patients with melanoma, we analyzed The Cancer Genome Atlas (TCGA) skin cutaneous melanoma dataset. Consistent with the oncogenic role of PRL1 in human melanoma samples, patients with high PRL1 mRNA expression had significantly decreased survival (n = 53, median survival = 72.8 months) compared with those with low PRL1 mRNA expression in the melanoma (n = 57, median survival = 362.3 months), with an HR of 0.46 (95% confidence interval of ratio = 0.24 to 0.88, P = 0.019; Fig. 5C). These clinical data suggest that inhibition of PRL1 could be beneficial for melanoma treatment. Thus, we hypothesized that the PRL1 detrimerizer Cmpd-43 may exhibit antimelanoma activity, possibly via downregulating the activity of both ERK1/2 and Akt pathways. To directly test this hypothesis, we examined whether Cmpd-43 could suppress human melanoma MeWo cell growth and motility. As shown in Fig. 5D, Cmpd-43 dose-dependently decreased MeWo cell proliferation as measured by the MTT assay. Live-cell tracking was used to measure the motility of MeWo cells in the presence of Cmpd-43. As shown in Fig. 5E, the total distance traveled over 6 hours for Cmpd-43 treated MeWo cells was significantly less than that of the control cells. To determine whether Cmpd-43 preferentially inhibits cancer cell growth, we treated both mammary carcinoma cell line MCF7 and normal mammary epithelial cell line MCF10A with Cmpd-43. As observed with MeWo cells, Cmpd-43 dose-dependently suppressed the growth of MCF7 breast cancer cells, but the inhibitory effect of Cmpd-43 was significantly compromised towards nontumorigenic MCF10A cells (Supplementary Fig. S2A). In addition, the antiproliferative activity of Cmpd-43 to MEF cells was also significantly reduced when compared to the MeWo cells (Supplementary Fig. S2B). These data indicate that Cmpd-43 displays significantly lower cell toxicity toward normal cells.

Importantly, Cmpd-43 treatment dose-dependently reduced HGF-induced ERK1/2 and Akt phosphorylation (Fig. 5F). To make certain that the effect of Cmpd-43 in MeWo cells is also mediated by blocking PRL1 trimerization, we overexpressed either wild-type PRL1 or the trimerization deficient mutant PRL1/G97R mutant in MeWo cells. Similar to what we observed in HEK293 cells (Fig. 3E), we found that overexpression of PRL1 but not PRL1/G97R significantly enhanced both ERK1/2 and Akt phosphorylation by about 2-fold (Fig. 5G). More importantly, Cmpd-43 but not Analog-1 suppressed PRL1-induced ERK1/2 and Akt activation. However, neither Cmpd-43 nor Analog-1 were able to reduce pERK1/2 and pAkt levels in PRL1/G97R expressing MeWo cells (Fig. 5G), suggesting that Cmpd-43 inhibits PRL1-induced ERK1/2 and Akt activation in MeWo cells by blocking PRL1 trimerization.

The PRL1 crystal structure revealed that residues at the trimer interface are conserved among all three PRLs (44, Supplementary Fig. S3A). We previously showed that like PRL1, PRL3 could also form trimer in solution and inside the cells (8, 44). Amino acid sequence alignment shows that key residues involved in Analog-3 binding are also highly conserved among all three PRLs (Supplementary Fig. S3A). Therefore, we hypothesized that trimerization is a general property for all PRLs and Cmpd-43 should inhibit trimerization of all PRLs. Indeed, both PRL2 and PRL3 can form trimer, and Cmpd-43 but not Analog-1 significantly reduces PRL2 and PRL3 trimer formation (Supplementary Fig. S3B).

To further validate the specificity of Cmpd-43 for PRLs in MeWo cells, we knocked down both PRL1 and PRL2, which are the major PRL isoforms expressed in MeWo cells. We first demonstrated that knocking down both PRL1 and PRL2 significantly reduced ERK1/2 and Akt activation in MeWo cells (Supplementary Fig. S4). In addition, the dose-dependent inhibition of both pERK1/2 and pAkt by Cmpd-43 in scramble siRNA-treated MeWo cells was significantly compromised in PRL1 and PRL2 knocked down cells (Fig. 5H and I). These data suggest that Cmpd-43 inhibits both ERK1/2 and Akt signaling pathways in MeWo cells specifically through targeting both PRL1 and PRL2. Overall, these results demonstrate that pharmacologic inhibition of PRL trimerization in MeWo cells attenuates both the ERK1/2 and Akt pathway activation and inhibits cell proliferation and motility.

Given the promising activity of Cmpd-43 in cell-based assays, we next aimed to establish the therapeutic potential of targeting PRL with Cmpd-43. First, we characterized the pharmacokinetic properties of Cmpd-43 in mice. Cmpd-43 displayed a very respectable pharmacokinetic profile in mouse with a plasma compound exposure C_max = 0.3 μmol/L and a half-life t_1/2 = 15.8 hours at a single 20 mg/kg intraperitoneal dosage. We then assessed the effect of Cmpd-43 on in vivo tumor growth in a mouse xenograft model using MeWo cells subcutaneously implanted into immunodeficient NSG mice. Once the tumor volume reached 200 mm³, we started daily intraperitoneal injection of either vehicle control or 30 mg/kg Cmpd-43 and monitored the tumor growth for 3 weeks. Mice treated with Cmpd-43 displayed reduced tumor growth throughout the experiment compared with mice treated with vehicle control (Fig. 6A). At 21 days posttreatment, we collected the tumors (Supplementary Fig. S5), and observed an approximately 62% shrinkage in tumor volume and approximately 48% reduction in tumor weight (Fig. 6A and B). Dissection and histologic analyses revealed no apparent toxicity in major organs when the mice were treated with Cmpd-43 at 30 mg/kg (Supplementary Fig. S6A and S6B). Furthermore, biochemical studies performed in samples isolated from the melanoma tumors revealed substantial reduction in both ERK1/2 and Akt phosphorylation, upon treatment with Cmpd-43 (Fig. 6C). Immunohistological analyses of tumor tissues revealed significantly reduced proliferation and increased apoptosis in Cmpd-43–treated MeWo tumors (Fig. 6D). Taken together, the reduction in tumor growth correlated with a decrease in ERK1/2 and Akt activity, validating the on-target activities of Cmpd-43.

In summary, recent studies expose an oncogenic role of PRLs in many cancers (3–6), raising the possibility that inhibition of these phosphatases might have broader therapeutic applications in oncology. Interestingly, the oncogenic potential of PRLs is always associated with their overexpression, which should increase the propensity of PRL trimerization inside the cell. Given the functional requirement of PRL trimerization, pharmacologic disruption of PRL trimerization represents an innovative approach for the treatment of human cancers with elevated PRL expression. By targeting the unique, noncatalytic trimerization interfaces that are unrelated to any other member of the PTP family, such PRL detrimerizers would be highly specific to the PRL. Starting from a sequential structure-based virtual screening strategy, we have identified Cmpd-43 and several analogs that are capable of preventing PRL trimerization. Biochemical and structural analyses demonstrate that Cmpd-43 and its close analogs directly bind the PRL1 trimer interface and obstruct PRL1 trimerization. Cmpd-43 also specifically blocks the PRL-induced cell proliferation and migration through attenuation of both ERK1/2 and Akt activity. Importantly, Cmpd-43 exhibits excellent anticancer activity both in vitro and in a xenograft mouse model of melanoma. The results not only further validate the importance of trimerization for PRL function but also support the clinical potential of compounds that inhibit PRL trimerization. Although additional medicinal chemistry optimization is required, these PRL detrimerizers represent valuable tools for elucidating PRL signaling and for developing novel agents for cancer therapy.

In vitro and in vivo cross-linking experiments, synthesis of Cmpd-43 and its analogs, assessment of PTP inhibition by Cmpd-43, PRL1 expression, purification, crystallization and data collection, structural determination and refinement, histological studies, and Supplementary Figs. S1–S6.

No potential conflicts of interest were disclosed.

Conception and design: Y. Bai, Z.-H. Yu, L.-F. Zeng, Z.-Y. Zhang

Development of methodology: Y. Bai, Z.-H. Yu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Bai, Z.-H. Yu, L. Zhang, R.-Y. Zhang, L.-F. Zeng, S. Zhang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Bai, Z.-H. Yu, S. Liu, L. Zhang, L.-F. Zeng

Writing, review, and/or revision of the manuscript: Y. Bai, Z.-H. Yu, S. Liu, L. Zhang, L.-F. Zeng, Z.-Y. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Bai

Study supervision: Z.-Y. Zhang

This work was supported in part by the NIH Grants CA69202 (Z.-Y. Zhang) and P30CA023168 (T. Ratliff).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.