The Antitumor Drug LB-100 is a Catalytic Inhibitor of Protein Phosphatase 2A (PPP2CA) and 5 (PPP5C) Coordinating with the Active Site Catalytic Metals in PPP5C

Abstract

LB-100 is an experimental cancer-therapeutic with cytotoxic activity against cancer cells in culture and antitumor activity in animals. The first Phase I trial (NCT01837667) evaluating LB-100 recently concluded that safety and efficacy parameters are favorable for further clinical testing. Although LB-100 is widely reported as a specific inhibitor of serine/threonine phosphatase 2A (PP2AC/PPP2CA:PPP2CB), we could find no experimental evidence in the published literature demonstrating the specific engagement of LB-100 with PP2A in vitro, in cultured cells, or in animals. Rather, the premise for LB-100 targeting PP2AC is derived from studies that measure phosphate released from a phosphopeptide (K-R-pT-I-R-R) or inferred from the ability of LB-100 to mimic activity previously reported to result from the inhibition of PP2AC by other means. PP2AC and PPP5C share a common catalytic mechanism. Here we demonstrate that the phosphopeptide used to ascribe LB-100 specificity for PP2A is also a substrate for PPP5C. Inhibition assays using purified enzymes demonstrate that LB-100 is a catalytic inhibitor of both PP2AC and PPP5C. The structure of PPP5C co-crystallized with LB-100 was solved to a resolution of 1.65Å, revealing that the 7-oxabicyclo[2.2.1]heptane-2,3-dicarbonyl moiety coordinates with the metal ions and key residues that are conserved in both PP2AC and PPP5C. Cell-based studies revealed some known actions of LB-100 are mimicked by the genetic disruption of PPP5C. These data demonstrate that LB-100 is a catalytic inhibitor of both PP2AC and PPP5C and suggest that the observed antitumor activity might be due to an additive effect achieved by suppressing both PP2A and PPP5C.

Introduction

The reversible phosphorylation of proteins regulates many aspects of cell growth and metabolic homeostasis. Therefore, compounds that specifically manipulate key phosphorylation-regulated processes should be useful in the medical management of human disease. Indeed, over 25 FDA-approved drugs used to manage human cancers target protein kinases (1,2). In contrast, the development of drugs targeting protein phosphatases received little attention until recently, when it became clear that, like their kinase counterparts, protein phosphatases are responsive and highly regulated enzymes (3).

To date, the only FDA-approved drugs targeting ser/thr phosphatases are cyclosporine, the macrolide tacrolimus (FK506) and related compounds, which are immunosuppressive agents that specifically inhibit PP2B/calcineurin (encoded by PPP3C) (4). However, several natural compounds that target PPP-family phosphatases (i.e. fostriecin, cytostatin A, cantharidin, and tautomycetin) have impressive antitumor activity in cell culture and animal models (5–11). Fostriecin and cytostatin A are structurally related natural compounds produced by Streptomyces (12). Fostriecin entered clinical development due to its potent and efficacious antitumor activity nearly a decade before it was shown to act as a potent catalytic inhibitor of PP2AC (13). More rigorous analysis later revealed that fostriecin also acts as a potent inhibitor of PPP4C and as a weaker inhibitor of PPP1C and PPP5C (6,11,14). In the largest Phase I trial conducted with fostriecin, disease stability was observed in 16 (of 46) solid tumor patients without reaching dose-limiting toxicity (15). Unfortunately, the study was closed before establishing the maximum tolerated dose due to problems with the supply of fostriecin (15).

Cantharidin, a natural compound made by many species of blister beetles, is the active constituent in a traditional Chinese medicine used for more than two millennia for a variety of purposes (5). Cantharidin was used briefly as an internal medicine in Europe, and in China it was reported to have anticancer activity in patients with gastrointestinal cancer when used as a beetle extract (Mylabris) (16). Cantharidin was also first reported to act as an inhibitor of PP2A in 1992 (17). Again, more rigorous testing revealed that cantharidin acts as a catalytic inhibitor of ser/thr phosphatases 1, 2A, 5, and 6 (PPP1C/PPP1C, PPP2AC/PPP2CA/B, PPP5C/PPP5C, PPP6C/PPP6C) with submicromolar to low micromolar affinity (5,9,18). In non-prescription formulations cantharidin is currently used to treat a variety of skin lesions. However, cantharidin is considered by many as too toxic for mainstream oncology (5) and modern clinical trials have not been attempted. Microcystin (19) and nodularin (20) are highly potent (IC₅₀ < 2 nM) but non-selective inhibitors of the cantharidin sensitive phosphatases (i.e. PPP1C/PPP2AC/PPP5C/ PPP6C). Both demonstrate marked systemic toxicity, notably hepato- and nephro-toxicity, arguing that potent non-selective inhibitors cannot be developed into useful medications.

It is assumed that the development of highly selective inhibitors of PPP-family phosphatases will generate compounds that have less systemic toxicity, and it is hoped that some selective inhibitors will retain antitumor activity. The simple structure (Fig. 1) and ease of chemical synthesis makes the cantharidin/endothall scaffold attractive for the development of specific inhibitors (9,21). Indeed, derivatives that do not inhibit PP4C have been already been developed (9). LB-100 (3-(4-methylpiperazine-1-carbonyl)-7-oxabicyclo[2.2.1]heptane-2-carboxylic acid) represents the first endothall-based derivative to enter human clinical trials (22,23). LB-100 is membrane permeable and was well tolerated in animal studies. Preclinical studies revealed that when used to supplement existing anti-tumor agents (e.g. cisplatin, doxorubicin, temozolomide, focal X-rays) against pancreatic (24), nasopharyngeal (25), hepatocellular carcinoma (26), fibrosarcoma (27), and glioblastoma xenografts (23), LB-100 increased potency without limiting toxicity (22,24,28). Recently, the first Phase I clinical trial had a favorable outcome (22), and a second trial is in progress.

Figure 1. Cantharidin family of inhibitors and structural derivatives.

Structures of (1) cantharidin, (2) cantharidic acid, (3) norcantharidin, (4) norcantharidic acid (endothall), (5) LB-100 and (6) LB-102. All members of the cantharidin-like family of inhibitors share a common 7-oxabicyclo[2.2.1]heptane-2,3-dicabonyl moiety.

PPP5C (commonly called PP5 or PP5C) shares a common catalytic mechanism with PP2AC, and the ten amino acids that coordinate the catalytic metals as well as position and orient the substrate phosphoryl moiety for nucleophilic attack at the catalytic site are identical in both enzymes (29). Because inactivating alterations of PP2A subunits have been observed in human cancers and the suppression of the same PP2A components contribute to cell transformation (for review see (30)), many view PP2A as a tumor suppressor. In contrast, PPP5C is found in high levels in several types of human cancers (31–36), and many lines of evidence indicate that inhibitors of PPP5C may prove useful for the medical management of several types of human cancers (31,33,37–44). Therefore, when several carboxamide derivatives containing the 7-oxabicyclo[2.2.1]heptane-2-carboxylic acid moiety shared with LB-100 were identified as novel inhibitors of PPP5C in an ultra-high-throughput screening campaign (45), we were prompted to test LB-100 for inhibitory activity against PPP5C. The data reported here represents the most comprehensive assessment of LB-100 activity against PP2AC reported to date, confirming the inhibitory activity of LB-100 against PP2AC. In addition, our studies identify PPP5C as a second LB-100 sensitive phosphatase. Structural studies using X-ray crystallography revealed that the 7-oxabicyclo[2.2.1]heptane-2,3-dicarbonyl moiety of LB-100 is positioned over the catalytic metals of PPP5C, blocking substrate access.

Materials and Methods

Materials.

LB-100 (3-(4-methylpiperazine-1-carbonyl)-7-oxabicyclo[2.2.1]heptane-2-carboxylic acid) was purchased from MedChem Express (HY-18597). Additional materials are provided in supplemental material.

Cell culture.

U-937 monocytes were obtained from ATCC (CRL-1593.2) and cultured in RPMI-1640 media (Gibco) supplemented with 1x non-essential amino acids, 100 units/mL penicillin, 100 μg streptomycin, and 10% fetal bovine serum. HEK-293 cells (ATCC; CRL-1573) were cultured in Dulbecco’s Modified Eagles Medium (DMEM) supplemented with 1x L-glutamine, pyruvate, non-essential amino acids, 100 units/mL penicillin, 100 μg streptomycin (Gibco, Life Technologies), and 10% fetal bovine serum. Mycoplasma testing was performed using PCR (eMYCO Plus kit, iNtRON Biotechnology). Upon receipt from ATCC cell lines are routinely expanded and frozen. Following thawing cell lines are generally discarded after 10–15 passages. Because the passage log for the U-937 cells in this study was not complete, the U-937 cell line was authenticated by Genetica Cell Line Testing (Burlington, NC) using short tandem repeat (STR) DNA profiling and confirmed as authentic using both the percent identity algorithm and percent match algorithm (ANSI/ATCC ASN-0002–2011). All cells were grown at 37 °C with 5% CO₂ in a humidified incubator. The PPP5C-KO cells will be provided for research purposes upon request.

Western Analysis.

HEK293 wild type or HEK293 PPP5C-KO cells (2 ×10⁵) were plated in 60 mm dishes and allowed to grow until the culture reached a confluent monolayer. The cells were then treated with the indicated concentration of LB-100 dissolved in H₂O. After 3 hours, the cells were washed with PBS and lysed by scraping in near boiling 2x SDS-PAGE sample buffer (62.5 mM Tris-HCl, pH 6.8, 20% glycerol, 4% SDS, 0.0025% bromophenol blue, and 0.02% ß-mercaptoethanol). Cell debris was sedimented by centrifugation at 16,900 x g for 15 minutes. The supernatant was transferred to a new tube, protein concentrations were estimated using RC DC™ Protein Assay (Bio-Rad #5000121) and 50 μg of each protein sample were separated by SDS-PAGE (10% polyacrylamide gels). Proteins were then transferred onto an Immobilon- P Membrane (Millipore). The membranes were blocked at room temperature for 1 hour in Odyssey blocking buffer (LiCor) and then incubated over night with 1:2000 dilution of primary antibodies P-S6 (S235/236) (Cell Signaling #2211) or ß-actin (Sigma #A2228) in 5% BSA at 4 °C. The next day, membranes were incubated with either anti-mouse (Sigma #A5278) or anti-rabbit (GE healthcare #NA934V) secondary HRP-conjugated antibodies (diluted at 1: 10,000) for 1 hour at room temperature. Bands were visualization using Clarity™ Western ECL Substrate (Bio-Rad #170–5060) using a Fuji-LAS-100 imaging system. Image J software was used for quantification of band density.

Cell viability assays.

U-937 cell viability was assessed using a resazurin-based fluorescence assay (46). For IC₅₀ determination, U-937 cells were seeded (1.5 ×10⁴ cells/well) in costar black 96-well clear bottom plates with LB-100 serially diluted with or without 500 nM dexamethasone to generate the indicated final concentrations. Following 48 hours, the resazurin solution (for details see supplemental material) was added to reach a final concentration of 120 μM and incubated for 24 hours. At 72 hours, the relative fluorescence was measured at 540 ± 20 nm excitation and 620 ± 20 nm emission on a Tecan Infinite M1000 Pro multimode plate reader. All assays were conducted in quadruplicate and the mean ± SD was derived from the cumulative data generated from four separate experiments (n=16).

Phosphatases, substrates, and phosphatase assays.

Human PPP5C was purified as previously described (6,47,48). Briefly, recombinant human PPP5C was expressed in E. coli as a maltose binding protein (MBP) fusion and purified by immobilized metal affinity chromatography. Following protease treatment, PPP5C was separated from MBP and other contaminants by anion exchange chromatography using previously described methods (47). The amino acid sequences of human and bovine PP2AC share 100% identity, making bovine PP2AC a surrogate for the human PP2AC. PP2AC was purified from bovine blood using ammonium sulfate fractionation, affinity (HiTrap heparin), and ion-exchange (HiTrap Q) chromatography according to established procedures (6). Details for all methods are provided as supplemental material.

[³²P]-phosphohistone was prepared by the phosphorylation of bovine brain histone with cAMP-dependent protein kinase (PKA) in the presence of cAMP and [³²P]-ATP using established methods (47,48). DiFMUP (6,8-Difluoro-4-methylumbelliferyl phosphate; from Invitrogen) based inhibition assays were conducted as described (47,48) in a 96 well format using 100 μM DiFMUP (final assay concentration). Phosphopeptide (KRpTIRR)/malachite green based inhibition assays were conducted using the EMD Millipore PP2A Phosphatase Assay Kit (catalogue number 17–313) using the methods, buffers, and reagents provided. In some assays, PP2AC was replaced with PPP5C as indicated. Phosphohistone phosphatase assays were performed as described (47,48). Briefly, LB-100, at indicated concentrations, or vehicle control (H₂O) were added to enzyme/buffer aliquots ~10 minutes prior to starting assays by the addition of [³²P]-phosphohistone substrate (to a final assay concentration of 300 nM incorporated phosphate). Phosphatase activity was measured by the quantification of [³²P]-labeled orthophosphate liberated from the substrate using established protocols (48).

Crystallization and Data Collection.

PPP5C was combined in a 1:2.5 molar ratio with LB-100. Crystals of the PPP5C-LB-100 complex were obtained using sitting-drop vapor diffusion experiments at 16 °C using 10 mM Tris-HCl pH 8.0, 35% MPD, and 10% PEG-MME 5000 as the crystallization reagent combined in a 1:1 ratio with the concentrated protein-inhibitor complex. Crystals were obtained after ~3 days and grew to ~100 × 50 μm. The crystals were mounted onto MiTeGen micro loops and flash cooled by plunging into liquid nitrogen. The data were collected in-house using an IμS 3.0 microfocus source (INCOATEC) and PHOTON II CPAD detector (Bruker, Wisconsin). Data (60 sec per exposure) were collected to 1.65 Å with a crystal-to-detector distance of 70 mm and an image width of 0.5^o per frame. Data were processed and reduced using the PROTEUM III program suite (Bruker, AXS). Data collection statistics are summarized in Table 2.

Table 2.

Data collection, phasing, and refinement statistics for PP5C co-crystallized with LB-100

PDB Code	5WG8
Beamline	Home source (D8 Quest, Bruker)
Wavelength (Å)	1.5418
Space group	P 21 21 21
Unit-cell parameters (Å, °)	a = 40.656 b = 91.15 c = 95.494
	α = β = γ = 90
Molecules per asymmetric unit	1
Data collection statistics
Resolution range (Å)	30.34 – 1.65 (1.709 – 1.65)
Unique reflections	43438 (4165)
Redundancy	8 (3.6)
Rmerge	0.142 (0.738)
Rmeas	0.151 (0.854)
Rpim	0.050 (0.420)
CC1/2	0.990 (0.671)
Overall I/σ	23.6 (3.5)
Completeness (%)	99.7 (97.5)
MR phasing statistics
Top LLG	3125.995
Top TFZ	44.4
Refinement statistics
Reflections used in refinement	43429 (4160)
Reflections used for R-free	4300 (436)
Rwork (%)	0.1587 (0.3047)
Rfree (%)	0.1961 (0.3472)
Number of non-hydrogen atoms	2890
Macromolecules	2599
Ligands	29
Solvent	262
r.m.s.d values
Bond length (Å)	0.009
Bond angles (°)	0.96
B-factor (Å2)
Wilson B	9.72
Protein	8.68
Ligand	15.04
Water	19.47
Ramachandran plot
Ramachandran favored (%)	96.54
Ramachandran allowed (%)	3.46
Ramachandran outliers (%)	0
Clashscore	1.93

R_merge = Σ|I−〈I〉|/ΣI, where〈I〉is the average intensity from multiple observations of symmetry-related reflections. R_work and R_free = Σ‖F_o| − |F_c‖/Σ|F_o|, where F_o and F_c are the observed and calculated structure factor amplitudes, respectively. R_free was calculated with 10% of the reflections not used in refinement. MR, molecular replacement. Values for the highest resolution shell are shown in parentheses.

Structure Determination and Refinement.

Orthorhombic crystals of the PPP5C-LB-100 protein complex belonging to the P2₁2₁2₁ space group contained 1 molecule per asymmetric unit. The structure of the PPP5C-LB-100 complex was solved using molecular replacement with Phaser-MR (49) using the crystal structure of PPP5C (PDB: 1S95) as a starting model (29). Water molecules were updated during refinement and manually checked. Refinement was performed using Phenix.Refine (50). Model building and map fitting were performed in COOT (51,52). The final model for the structure was found to exhibit good geometry, as determined using MolProbity (PHENIX;(53)). Refinement statistics are shown in Table 2. All structure figures were prepared using PyMOL (The PyMOL Molecular Graphics System, Version 1.8 Schrödinger, LLC).

Protein Data Bank Accession Code.

Atomic coordinates and structure factor amplitudes have been deposited with the Protein data bank (http://www.pdb.org) and are accessible under accession code 5WG8.

Results

Inhibition of phosphatase activity by LB-100.

Although LB-100 is widely reported to act as a specific inhibitor PP2AC (23), we could not find a report testing the inhibitory activity of LB-100 against PPP5C, which has an identical catalytic mechanism (29) and shares several common substrates with PP2AC (54–56). Further scrutiny of the literature revealed that the ability of LB-100 to inhibit the catalytic activity of PP2A is based on studies using a commercial assay (Millipore ser/thr phosphatase assay kit). In this kit, PP2A activity is measured using a small phosphopeptide (K-R-pT-I-R-R) substrate (23,24). When we tested the ability of PPP5C to also utilize the peptide substrate provided in the PP2A-assay kit, we observed that K-R-pT-I-R-R is also an excellent substrate for PPP5C (10,670 ± 90 nMole/min/mg). We next tested the ability of LB-100 to inhibit PP2AC and PPP5C in head-to-head assays using the malachite green based assay provided in the Millipore kit, substituting PPP5C for PP2AC as indicated. These studies revealed that LB-100 inhibits the catalytic activity of both PP2AC and PPP5C in a dose dependent manner (Fig. 2A; Table 1).

Figure 2. Inhibitory effect of LB-100 on the activity of PP2AC and PPP5C.

Inhibition of PP2AC (circles) and PPP5C (triangles) activity by LB-100 assayed using (A) a phosphopeptide (K-R-pT-I-R-R), (B) DiFMUP or (C) [³²P]-labeled histone as a substrate. Assays were conducted as described in methods. All compounds were mixed with the enzymes for ≥10 minutes at 23 °C prior to the initiation of the reaction with the addition of substrate. Each point represents the mean +/− SD (n=3–8). IC₅₀ values calculated from a 9, 10, or 11-point concentration/dose-response curves by a 4-parameter logistic fit of the data are provided in Table 1.

Table 1.

Inhibitory activity of LB-100 against PP2AC and PPP5C^*

Substrate	Phosphatase; IC50 mean ± SE (μM)**
DiFMUP [32P]-labeled phosphohistone (K-R-pT-I-R-R)	PP2AC 0.39 ± 0.013 0.64 ± 0.037 1.27 ± 0.13	PP5C 1.82 ± 0.093 4.9 ± 0.29 6.64 ± 0.24

^*Phosphatase activity was measured as described in methods.

^**IC₅₀ values are calculated from a 9, 10 or 11-point concentration/dose response curve by a 4-parameter logistic fit of the data, using 3–8 replicates per concentration.

Because LB-100 is a low molecular weight inhibitor (MW=268.31), it is predicted to disrupt limited but key substrate-enzyme contacts. Therefore, we next tested the inhibitory activity of LB-100 using DiFMUP, a low molecular weight (MW=292.13) fluorogenic substrate expected to interact with phosphatase active sites primarily via limited contacts with its phosphoryl moiety. As observed with the malachite green based assay, when DiFMUP is used as a substrate, LB-100 inhibits the activity of both PP2AC and PP5C (Fig. 2B).

Although LB-100 was found to be a potent inhibitor of both PP2AC and PPP5C with small non-physiological substrates, interaction between protein phosphatases and their substrates can occur over relatively large domains that produce multiple contacts. Therefore, we next tested the ability of LB-100 to inhibit the activity of PP2AC and PPP5C using protein kinase A phosphorylated [³²P]-histone, which is an established substrate for both PP2AC and PPP5C (6,29,45,47,48). As observed using DiFMUP or K-R-pT-I-R-R as a substrate, the dephosphorylation of [³²P]-radiolabeled phosphohistone by PP2AC is potently inhibited by LB-100. When tested against PPP5C, we found that LB-100 again also inhibits PPP5C activity in a concentration dependent manner (Fig. 2C; Table 1). With all three substrates, LB-100 demonstrated modest (~4–7 fold) selectivity for PP2AC versus PPP5C. Together, these studies confirm that LB-100 is a catalytic inhibitor of PP2AC and identified PPP5C as a second LB-100 sensitive phosphatase.

Sensitivity of the U-937 cells to LB-100.

Thorough review of the literature revealed that the premise that LB-100 targets PP2A in cells or tissues is derived from studies in which cells or animals were treated with LB-100 and the lysates from untreated and treated cells/tissues were compared (23,57–65). This was done by measuring either: 1) the ability of the LB-100 treated lysates to inhibit the activity of PP2A when only PP2AC was added to the assay, 2) measuring the release of phosphate from a phosphopeptide (K-R-pT-I-R-R) added to the lysates under the assumption the this represents PP2AC activity; or 3) by assessing LB-100 induced changes in the phosphorylation of proteins that were previously reported to occur when PP2A was inhibited by other means (i.e. studies with okadaic acid, which inhibits both PP2A and PPP5C, or with siRNA/antisense oligonucleotides that specifically suppress the expression of PP2AC or PPP2R1A (a regulatory/scaffold protein needed to properly assemble the PP2A holoenzyme) (66).

Having demonstrated that the phosphopeptide used to measure PP2A activity is readily dephosphorylated by both PP2AC and PPP5C (Fig. 2A) and that LB-100 inhibits both PP2AC and PPP5C at similar concentrations in vitro (Fig. 2), we next conducted cell based studies to determine if LB-100 can elicit actions altered by the inhibition of PPP5C but not by the inhibition of PP2A. Although PP2A and PPP5C act on several common substrates, the ability of arachidonic acid (AA) to activate PPP5C but not PP2A has been established for nearly two decades (67,68). In response to oxidative stress the activation of PPP5C leads to the suppression of both ASK1 induced apoptosis (69–72) and dexamethasone induced glucocorticoid receptor (GR) action (73–75). In cells belonging to the monocyte/macrophage lineage (e.g. U-937 cells), production of arachidonic acid provides a survival advantage during periods of oxidative stress that can be suppressed by dexamethasone (Fig. 3A) (76–78). Therefore, if LB-100 acts to suppress the actions of PPP5C, the cytotoxic activity of LB-100 should be augmented by treatment with dexamethasone. To test this, we conducted dose-response studies in the presence or absence of 500 nM dexamethasone using a metabolic capacity assay (resazurin-based fluorescence) as a surrogate to assess toxicity (Fig. 3B). These studies revealed that the cytotoxic activity of LB-100 is indeed modestly increased in the presence of 500 nM dexamethasone (Fig. 3B), consistent with the notion that LB-100 is inhibiting PPP5C in U-937 cells.

Figure 3. The impact of LB-100 treatment on PPP5C in human cells.

(A) Schematic of known cytoprotective roles for PPP5C. Dexamethasone stimulated expression of lipocortin-1 suppresses oxidative stress induced release of arachidonic acid (AA) by phospholipase A₂ (PLA₂). PPP5C is activated by AA, and activated PPP5C suppresses apoptosis signaling kinase 1 (ASK1) mediated cytotoxicity. (B) The cytotoxicity of LB-100 in U-937 cells is augmented by the addition of dexamethasone. Representative dose-response curves plotted as the relative fluorescent units (RFU) generated using the resazurin-based fluorescent assay versus the final concentration of LB-100. Square: LB-100 alone, EC₅₀=11.7 +/−0.49 μM; circle LB-100 and 500 nM dexamethasone, EC₅₀= 5.48 +/− 0.073 μM. Fluorescent measurements were made 72 hours after LB-100 was added to the cultures. All assays were conducted in quadruplicate and the mean ± SD shown was derived from the cumulative data generated from four separate experiments (n=16). (C) Representative western blot comparing S6 phosphorylation in cell extracts generated from wild-type (WT) HEK-293 cells treated with the indicated concentration of LB-100, and untreated HEK-293 cells in which the expression of PPP5C was disrupted (PPP5KO) using CRISPR-cas9 based methods. HEK-293 cells were treated with the indicated concentration of LB-100 (0–2 μM) for 3 hours and phosphorylation of ribosomal S6 at S235 and S236 was detected using a phospho-specific antibody as described in methods. (D) Quantitation of western blot normalized to β-actin; ns= not significant, * p<0.05, ** p<0.01, ***p<0.001

Because most compounds used to inhibit PP2A activity in cells or animals (i.e. okadaic acid, calyculin A, cantharidin, endothall, microcystin, nodularin, tautomycetin) also inhibit PPP5C at similar concentrations (79), we used a CRISPR-Cas9 based approach to generate stable clonally derived HEK-293 cell lines in which the expression of PPP5C was disrupted by inserting a single base leading to a frame shift induced stop codon in exon one (80). In the PPP5C-deficient HEK-293 cells, we found that phosphorylation of ribosomal S6 (Ser235/236) is constitutively elevated. To determine whether LB-100 targeted the activity of PPP5C in a cellular context, we conducted dose-response studies by treating wild-type cells with LB-100 and assessed S6 (Ser235/236) phosphorylation status. We found that phosphorylation of ribosomal protein S6 (Ser235/236) was induced in the wild-type HEK-293 cells by LB-100 in a dose-dependent manner (Fig. 3C), again mimicking the activity achieved by the specific suppression of PPP5C (Fig. 3D).

Crystal structure of PPP5C in complex with LB-100.

To further understand the inhibitory activity of LB-100 against PPP5C, the structure of PPP5C co-crystallized with LB-100 was solved to a resolution of 1.65 Å (Table 2). The active site of PPP5C (identified by two closely apposed metal ions) is located at the base of a shallow depression on the surface formed by residues from four loops connecting secondary structural elements in the protein: β4-αD, αG-αH, β10-β11, and β12-β13 (29). The structure reveals clear density for the 7-oxabicyclo[2.2.1]heptane-2,3-dicarbonyl moiety (Fig. 4A). However, there was insufficient density to position the 3-[4-methyl-1-piperazine] ring. Thus, we modeled the minimal 7-oxabicyclo[2.2.1]heptane-2,3-dicarbonyl moiety of LB-100 into the observed density. To corroborate this observation, we also examined our data using isomorphous difference methods with the PP5C-4TE complex (PDB ID: 4ZX2). The difference map obtained indicated that the C5 methyl group in 4TE was not present and no apparent density was revealed for the methylpiperazine ring of LB-100 (Supplementary Fig. S1). Because there was ambiguity associated with placing the 3-[4-methyl-1-piperazine] ring in our model, the ring and apposing oxygen atoms of the 2,3-carbonyl were not included and assigned as positions X and Y for demonstrative purposes (Fig. 4).

Figure 4. Representative contacts of the 7-oxabicyclo[2.2.1]heptane-2,3-carbonyl moiety of LB-100 within the active site of PP5C.

(A) Composite omit map generated by Autobuild (PHENIX) contoured at 1σ showing electron density for LB-100, active-site metal ions, and active-site residues. (B) Stick representation showing the coordination sphere of each metal ion in the LB-100 complex. Carbonyl dipole-dipole stacking interactions and hydrogen bonds are shown as yellow dotted lines and metal-ligand bonds are shown as solid lines. Hydrophobic contacts between the inhibitor and the side-chains of Val429, Phe446, and Tyr451 are represented as blue dotted lines. (C) Comparison showing superposition of amino acid residues near the active site in the substrate analogue complex (PDB code 1S95; (29) green) and the LB-100 complex (cyan). The metal-coordinated waters in the phosphate complex are shown as magenta spheres (W1 and W2). Metal ions (M1 and M2) are shown as purple spheres and the ambiguous positions X and Y are colored bronze.

Our structure superposes well with other PPP5C structures (r.m.s.d. values reported in Supplemental Table S1). From the superposition of our current model with the PPP5C-phosphate complex (PDB ID: 1S95), it is apparent that the two Mn²⁺ ions (M1, M2) occupy the same positions (Fig. 4C). However, the coordination geometry between the active site residues and the metal ions differs from 1S95 due to the presence of the inhibitor (Fig. 4B, C). In the PPP5C-phosphate complex, both active site metals exhibit a slightly distorted octahedral geometry, interacting with six PPP5C active site residues and two coordinated waters (one being the attacking nucleophile in the phosphomonoester hydrolysis reaction) (29). With LB-100 bound, M2 maintains an octahedral sphere where coordination is mediated by PPP5C residues Asp242, His244, and Asp271 as well as three oxygen atoms of the inhibitor (the O7 bridging oxygen, and the 2,3-carbonyl groups) (Fig. 4B). On the other hand, M1 displays trigonal bipyramidal coordination geometry, interacting with four PPP5C residues (Asp271, Asn303, His352, and His427) and the 2-carbonyl of LB-100. When comparing the PPP5C-phosphate substrate analog complex to the PPP5C-LB-100 complex, we observe that the bridging water/hydroxide is replaced by the 2-carbonyl of LB-100, which coordinates both M1 and M2. The water coordinating with M2 is replaced by the O7 bridging oxygen, while the phosphate oxygen in 1S95 coordinating M2 is substituted by the 3-carbonyl group of LB-100 (Fig. 4C). Thus, in the PPP5C-LB-100 complex, the inhibitor occupies the coordination sites for the substrate phosphoryl moiety and the two active site waters. Additionally, in order to accommodate the inhibitor, the side chains of Arg275 and Tyr451 adopt conformations slightly farther away (2.0 Å for CZ of Arg275 and 1.6 Å for the hydroxyl of Tyr451) from the active site metals (Fig. 4B).

The substrate-binding pocket harbors several conserved active site residues (i.e. Arg275, His304, Arg400 and His427) that participate in a complex network of hydrogen bonding interactions. Arg275 and His304 are stabilized by H-bonds with Tyr451 and Asp274, respectively, while Arg400 donates a hydrogen bond to the carbonyl oxygen of His427. Further stabilization of the complex is provided by hydrophobic contacts between the inhibitor and the side chains of Val429, Phe446 and Tyr451 as well as an anti-parallel dipole-dipole stacking interaction between the 3-carbonyl of LB-100 and the backbone carbonyl of His427 ((C^…O) = 3 Å and 3.3 Å (Fig. 4B)).

Discussion

In pre-clinical studies, LB-100 was shown to inhibit proliferation and induce apoptosis in a variety of cultured cancer cells. LB-100 also acts as an effective chemo- and radio-sensitizer for the treatment of various cancers, and LB-100 exhibited potent in vivo anti-neoplastic activity in combination with cisplatin or doxorubicin in mouse xenograft models (22,24,26,28,81,82). The first in-human Phase I trial (NCT01837667) recently concluded that the safety, tolerability, and preliminary evidence of anti-tumor activity support the continued development of LB-100 as a novel antitumor agent, either alone or in combination with existing therapies (22).

In nearly all reports describing LB-100 activity, the target of LB-100 is ascribed to the specific inhibition of ser/thr protein phosphatase type 2A (PP2A) (22,24,26,28,81,82). However, the most cited reference to support selectivity for PP2A (23) was corrected in 2009 revealing that the compound with >10,000 fold selectivity for PP2A was another endothall derivative (LB-102; 4-(3-carboxy-7-oxa-bicyclo [2.2.1] heptane-2-carbonyl)piperazine-1-carboxylic acid tert- butyl ester (Figure 1)). After a thorough search of the published literature, we could find no experimental evidence documenting the selective engagement of LB-100 with PP2A in vitro, in cells or in animals, nor are there reports that tested the ability of LB-100 to inhibit PPP5C activity. The studies described here tested the inhibitory activity of LB-100 against both PP2AC and PPP5C using three different substrates, and, therefore, represent the most comprehensive assessment of the inhibitory activity of LB-100 against PP2A published to date. Our data confirm that LB-100 does indeed act as a catalytic inhibitor of PP2AC. In addition, LB-100 also inhibits the catalytic activity of PPP5C, demonstrating only modest selectivity towards PP2AC (Fig. 2: Table 1). Because LB-100 only demonstrates modest selectivity for PP2AC versus PPP5C and there is no experimental evidence in the published literature for the selective engagement of LB-100 with PP2A in vitro, in cells or in animals, we conclude that the observed effects of LB-100 in vivo may not be due to the specific inhibition of PP2AC alone.

Our co-crystal structure also defines the binding site of LB-100 in PPP5C, revealing coordination with the catalytic metals in the active site. The ability of LB-100 to inhibit both PP2AC and PPP5C is similar to the actions of cantharidin and endothall, which both show modest selectivity for PP2AC vs. PPP5C. The hydrolyzed diacid forms of both cantharidin and endothall share a 7-oxabicyclo[2.2.1]heptane-2,3-dicarbonyl moiety with LB-100, and the crystal structures of PPP5C in complex with cantharidin, endothall, and two novel cantharidin derivatives have been reported (9). For all five compounds, contacts between the shared core scaffold and catalytic site residues and metal ions are conserved across these complexes. In our PPP5C-LB-100 structure, there was insufficient density to assign a position for the methylpiperazine ring. This prompted us to re-evaluate the reported stability of LB-100 in order to assess the possibility that the ring was hydrolyzed during storage. MS analysis of the same batch of LB-100, stored for >6 months as a powder and dissolved three months prior to generating the stock solution used to make the co-crystals retained the correct molecular weight even after lyophilization and re-suspension (Supplemental Fig. S2). These observations are consistent with previous reports indicating that LB-100 is stable during prolonged storage (22). Therefore, we concluded that the ring does not likely contribute to the retention of LB-100 at the active site of PPP5C.

To date there are no reports of compounds that act as specific inhibitors of PPP5C. However, in numerous studies of PPP5C function, multiple lines of evidence have revealed a positive correlation between PPP5C overexpression and human cancers, including invasive ductal carcinoma of the breast (39), hepatocarcinogenesis (41), lymphoma (83), glioma (84,85), colorectal (86), and prostate cancers (87). In cell culture models, PPP5C over-expression aids cancer-cell proliferation and survival during hypoxia (39,88,89), and in mouse xenograft models of tumor development, constitutive PPP5C over expression markedly aids tumor growth (39,90). Similar to observations made with LB-100 in cell culture, studies conducted with siRNA or antisense oligonucleotides targeting PPP5C-mRNA indicate that the suppression of PPP5C prolongs stress-induced signaling cascades that favor apoptosis and sensitize cancer cells to stress- or drug-induced apoptosis (39,88,91–93). Indeed many actions reported in response to treatment with LB-100 (e.g. Chk1-phosphorylation, AKT-1 phosphorylation, increased sensitivity to doxorubicin) are mimicked by the action of siRNA or antisense oligonucleotides specifically targeting PPP5C (5,6,29,39,47,83–85,87–90,94). In addition, the suppression of PPP5C is known to augment many actions elicited by GR-activation (89,92,95) and dexamethasone increased the sensitivity of U-937 cells to LB-100 mediated cytotoxicity (Fig. 3B). We also observed a concentration dependent increase in the phosphorylation of ribosomal protein S6 after treatment with LB-100, which is also observed when the expression of PPP5C is disrupted using CRISPR-Cas9 based methods (Fig. 3C, D). Together, these studies suggest that many of the actions of LB-100 that were assumed to be due to the specific inhibition of PP2A could result from the inhibition of PP2A, PPP5C, or both PP2A and PPP5C.

Nonetheless, there are also actions of LB-100 that can be mimicked by the specific suppression of PP2A expression. For example, the ability of LB-100 to induce G2/M arrest or the appearance of aberrant mitotic spindles (22,23,25,26) is mimicked by antisense oligonucleotides targeting PP2AC (7) but is not observed in cells treated with siRNA/antisense oligonucleotides targeting PPP5C (39,91,92). Further, the genetic disruption of PPP5C in mice only produces a mild metabolic phenotypic change (94) and HEK-293 cells that do not express PPP5C protein are also viable. In addition, in the absence of genomic stress, the suppression of PPP5C expression with siRNA or antisense oligonucleotides does not induce apoptosis in many types of cells grown in culture (39,88,90). These observations argue that the specific inhibition of PPP5 in the absence of hypoxia, or treatment with drugs that induced genomic stress/damage may not kill cancer cells unless PP2A is also inhibited. Therefore, future studies should explore the possibility that the observed antitumor activity of LB-100 might be due to an additive affect achieved by suppressing both PP2A and PPP5C.

Finally, it is important to remember that, unlike PPP5C, which encompasses catalytic, targeting and regulatory function within a single polypeptide chain encoded by a single gene, PP2AC is a catalytic subunit that is shared by many PP2A holoenzyme complexes. A vast array of functional heterotrimeric PP2A-holoenzymes can be assembled combinatorially from a set of two isoforms of the 65 kDa scaffolding subunit, two isoforms of the 36 kDa catalytic subunit, and 18 different regulatory subunits that confer substrate specificity, affect subcellular localization, and/or alter catalytic activity. Therefore, further studies are also needed to determine if the strength of LB-100 binding is altered when PP2AC is assembled into particular holoenzyme complexes or when PPP5C is in complex with known binding partners (96). Future investigations are also needed to assess the ability of LB-100 to inhibit the activity other PPP-family members that share the same catalytic mechanism, notably PPP6C which is highly sensitive to cantharidin and other endothall derived derivatives (9).

Abbreviations:

LB-100

3-(4-methylpiperazine-1-carbonyl)-7-oxabicyclo[2.2.1]heptane-2-carboxylic acid

PP2A:

serine/threonine protein phosphatase type 2A/PPP2CA

PPP5C

serine/threonine protein phosphatase type 5/PPP5C