Somatic mutation of the protein phosphatase 2A (PP2A) Aα-subunit gene PPP2R1A is highly prevalent in high-grade endometrial carcinoma. The structural, molecular, and biological basis by which the most recurrent endometrial carcinoma–specific mutation site P179 facilitates features of endometrial carcinoma malignancy has yet to be fully determined. Here, we used a series of structural, biochemical, and biological approaches to investigate the impact of the P179R missense mutation on PP2A function. Enhanced sampling molecular dynamics simulations showed that arginine-to-proline substitution at the P179 residue changes the protein's stable conformation profile. A crystal structure of the tumor-derived PP2A mutant revealed marked changes in A-subunit conformation. Binding to the PP2A catalytic subunit was significantly impaired, disrupting holoenzyme formation and enzymatic activity. Cancer cells were dependent on PP2A disruption for sustained tumorigenic potential, and restoration of wild-type Aα in a patient-derived P179R-mutant cell line restored enzyme function and significantly attenuated tumorigenesis and metastasis in vivo. Furthermore, small molecule–mediated therapeutic reactivation of PP2A significantly inhibited tumorigenicity in vivo. These outcomes implicate PP2A functional inactivation as a critical component of high-grade endometrial carcinoma disease pathogenesis. Moreover, they highlight PP2A reactivation as a potential therapeutic strategy for patients who harbor P179R PPP2R1A mutations.
This study characterizes a highly recurrent, disease-specific PP2A PPP2R1A mutation as a driver of endometrial carcinoma and a target for novel therapeutic development.
See related commentary by Haines and Huang, p. 4009
Uterine cancer is the most common gynecologic malignancy in the United States with approximately 60,000 women diagnosed each year (1). Although most cases of uterine endometrial carcinoma have favorable outcomes with recurrence-free long-term survival, outcomes for the high-grade, treatment-refractory histologic subtypes remain an important clinical problem (2, 3). Uterine serous endometrial carcinoma (USC) accounts for only 10% of uterine cancer cases but a disproportionate 39% of deaths, with a 5-year survival of 55% (2). Uterine carcinosarcoma (UCS), a more aggressive mixed histologic subtype, is rare, <5%, but accounts for >15% of deaths, with a 5-year survival around 35% (4–6). Importantly, despite an overall decline in cancer-related deaths in the United States, mortality rates for uterine cancers continue to rise (1). A lack of established disease driving mechanisms for high-grade subtypes has limited the use of targeted treatment strategies, an approach that has had some success in the treatment of other cancers. New insight and validation of biological drivers of disease progression are therefore needed to improve disease management. Large-scale genomic profiling efforts have made substantial progress in identifying alterations characteristic for USC and UCS (7–10). Foremost, both are typified by TP53 mutation, which is present in 80% to 90% of cases. In addition to TP53, PPP2R1A mutation was also more frequent in high-grade subtypes, occurring in approximately 30% of USC or UCS patients versus approximately 5% in patients with the endometrioid subtype (UEC). PPP2R1A encodes the scaffold subunit of the protein phosphatase 2A (PP2A) tumor-suppressive phosphatase.
PP2A is a serine/threonine phosphatase that is involved in the physiologic regulation of diverse signaling pathways. PP2A is also a tumor suppressor whose diminished activity contributes to transformation and tumor development (11–13). Although a number of mechanisms can underlie diminished PP2A activity in cancer, notable is the occurrence of “hotspot” mutations to PPP2R1A, the gene encoding the PP2A Aα-subunit. The PP2A holoenzyme is a heterotrimer composed of a scaffolding A-subunit (PP2A-A), catalytic C-subunit (PP2A-C), and regulatory B-subunit (PP2A-B). Each is a subunit family composed of multiple isoforms; for PP2A-A, this includes α and β isoforms.
The biogenesis of canonical PP2A heterotrimers is sequential: first a C-subunit binds to an A-subunit and is followed by incorporation of one of 15 identified B-subunits (14, 15). Several activation steps must also take place, facilitated by PP2A regulatory proteins, to convert the C-subunit into its catalytically active form and to modulate the binding of B-subunits (15). Critically, throughout this process, the A-subunit serves as a highly flexible scaffolding protein to facilitate protein interactions and trimerization. Cancer-derived hotspot mutations of PPP2R1A cluster at the structural interface between A- and B-subunits, and have been shown to disrupt subunit binding to varying degrees (11, 16–18). A recent report characterizing several hotspot mutations found that these mutant proteins have increased binding to the PP2A inhibitor TIPRL, resulting in a dominant-negative phenotype (18). Our data suggest this mechanism may be true for some hotspot mutations but may not hold true for others that are most specific and recurrent in high-grade endometrial carcinoma. Specifically, although PPP2R1A mutations are found in cancers of multiple origins, they are striking for USC and UCS not only for their high prevalence, but also for the nearly exclusive occurrence of mutation at two hotspot sites: P179 and S256.
Our work focuses on P179R, the most common of these recurrent, endometrial carcinoma–enriched mutations, for which we have developed crystallography, modeling, and biochemical data to describe its impact on PP2A function. The P179R missense mutation induces global changes in A-subunit protein structure and its preferred conformational dynamics, which results in loss of interaction with the catalytic subunit and diminished holoenzyme stability. When wild-type (WT) protein is restored in the P179R-mutant context, tumorigenesis is significantly inhibited. Moreover, pharmacologic activation of PP2A with a small-molecule activator of PP2A (SMAP) phenocopied this outcome of tumor growth inhibition (19). Altogether, these works demonstrate that PP2A is a key tumor suppressor in endometrial carcinoma and is functionally disrupted by recurrent P179R PPP2R1A mutation to drive tumorigenesis.
Materials and Methods
Analysis of primary human endometrial and ovarian cancer samples
Deidentified tissue specimens were obtained from the Case Comprehensive Cancer Center (Case CCC) Gynecologic Tumor Biobank (Dr. Analisa DiFeo). DNA and RNA were isolated from tissue pieces using the AllPrep DNA/RNA Mini Kit (Qiagen) in combination with a QIAshredder (Qiagen) for initial tissue homogenization. The DNA region of interest was amplified by PCR (PCR Master Mix, Promega) and submitted for Sanger sequencing. Results were viewed using the software program 4Peaks. DNA and RNA were stored at −20°C and −80°C, respectively.
Cell lines and culture conditions
The UT42, UT89, UT150, and UT185 cell lines were generated by Dr. Analisa DiFeo. These stable lines were derived from primary human endometrial cancer tissue following protocols described previously (20). HEK293T cells were purchased from the ATCC. All lines were cultured in DMEM (Corning, 10-013) supplemented with 10% FBS (GE Healthcare, SH30070.03) and 1% penicillin–streptomycin (GE Healthcare, SV30010). Cells were maintained in humidified sterile incubator conditions with 5% CO2 at 37°C. Cells are passaged once 70% to 90% confluency is reached, and are maintained in culture for no more than 20 passages. Freshly thawed cells are passaged 2 to 3 times before use in experiments. Mycoplasma testing of cell lines is performed once annually using a detection kit (Lonza, LT07-710), and positive cell lines are discarded. Authentication of cell lines was performed via short tandem repeat profiling after the cell line was established (typically after two passages in culture), and compared with that of the primary tumor.
Generation of stable cell lines
Standard methodologies were used; details are provided in the Supplementary Methods.
Standard methodologies were used; details and the primary antibodies used are provided in the Supplementary Methods.
Standard methodologies were used; details and primer sequences are provided in the Supplementary Methods.
Coimmunoprecipitation and phosphatase activity assays
UT89 and UT42 isogenic cell lines were plated to achieve 70% confluency at 24 (UT89) or 48 (UT42) hours. Cells were harvested, and coimmunoprecipitation (IP) of V5-tagged protein was performed according to the Dynabeads Co-Immunoprecipitation Kit (Invitrogen) protocol. Note that 1.5 mg Dynabeads per IP reaction were coupled to V5 antibody (Abcam ab27671) at a concentration of 8 μg Ab/mg beads. Cells were lysed in a 1:9 ratio of mg cells to mL lysis buffer, and equal volumes of this pre-IP lysate were incubated with antibody-conjugated beads. Pre-IP lysates, and IP isolates obtained after elution from beads, were stored at −20°C for immunoblotting.
The phosphatase activity assay was performed using the IP protocol with modifications: cells were lysed in a phosphatase assay buffer (25 mmol/L HEPES, 1 mmol/L MgCl, 0.1 mmol/L MnCl, 1% Triton-X, and protease inhibitor); 5 mg of protein lysate was incubated with beads; beads were triple-washed with a wash buffer (25 mmol/L HEPES, 1 mmol/L MgCl, and 0.1 mmol/L MnCl); and aliquots were portioned out for immunoblotting or activity assay. For assessment of phosphatase activity, beads with bound proteins were suspended in phosphatase assay buffer supplemented with 1 mmol/L DTT. Samples were incubated with increasing concentrations of DiFMUP (ThermoFisher, D6567) for 15 minutes on a shaker at 37°C, at which point fluorescence was measured at 365/455 nm. As a control for specificity of activity, 50 nmol/L of okadaic acid (OA) was also added to select wells of V5-WT IP isolate. Measurements from V5-EGFP IP isolates were considered background and subtracted out. Phosphatase activity rate was then calculated for V5-WT and V5-P179R as fluorescent units per minute incubation time. Final values were normalized to total C-subunit protein in IP isolates as determined by immunoblotting.
Cell-free GST pull-down assay
Approximately 0.6 μmol/L of purified GST-B56α, 6 μmol/L of His8-Cα, and 6 μmol/L of Aα WT or mutant protein were incubated with 50 μL of GSH Sepharose 4B resin (GE Healthcare) for 1 hour at 4°C in binding buffer (20 mmol/L Tris, pH 8.0, 150 mmol/L NaCl, 5% glycerol, 0.05% Tween 20, 2 mmol/L DTT; total volume of 250 μL). The beads were then washed with binding buffer 3 times. The bound proteins were resolved on SDS-PAGE and stained with Coomassie Brilliant Blue. The intensities of the bands were quantified by Image Lab Software (Bio-Rad), and values were normalized to the GST-B56α of WT. Each experiment was repeated 3 times.
Proteasome inhibition and half-life studies
Cycloheximide solution was obtained from Sigma-Aldrich as a stock solution of 100 mg/mL in DMSO and used at a concentration of 100 μg/mL for experiments. Bortezomib (Velcade) was obtained from Selleck Chemicals; a stock solution was prepared in DMSO and diluted to 1 μmol/L for experiments. Treatments were added 24 hours after cell plating. At the reported timepoints, cells were harvested and lysed for protein immunoblotting. To calculate protein half-life, cycloheximide protein quantification data were Ln-transformed and fit with a regression line using GraphPad Prism 7 software. Details of the statistical analysis of these data are presented in the Supplementary Methods. The slope of the best-fit line was used to calculate half-life with the formula T(1/2) = ln(2)/slope.
Standard methodologies were used; details are provided in the Supplementary Methods.
SMAP dose–response assay
Evaluation of SMAP treatment–induced growth inhibition was performed to determine the IC50 of four endometrial carcinoma cell lines. Cells were plated into 96-well plates at optimized cell densities such that approximately 90% confluency is achieved at 72 hours. Twenty-four hours after plating, wells received media containing DMSO or SMAP-061, with triplicate wells per treatment condition. The IncuCyte live-cell imaging system (Essen Biosciences) was utilized to monitor cell number in each well for 72 hours. The fold change in cell number was determined for DMSO or SMAP, at each dose, and used to construct the dose–response curves. In GraphPad Prism, data were log-transformed and fit with a nonlinear curve.
Xenograft tumor assays
All animal work was conducted in accordance with Case Western Reserve University Animal Resource Center–approved protocols and ethical guidelines. For the subcutaneous xenograft studies, 1 × 106 (UT89) or 5 × 106 (UT42) cells suspended in a 1:1 mixture of matrigel, and cell media were injected s.c. into the right flank of 6- to 8-week-old female Balb/c nu/nu mice. Tumor growth was assessed by caliper measurement, and volume (V) was calculated using the formula V = 0.5 × (L × W2), where L = length and W = width. At study end, tumor tissue was harvested, formalin-fixed, and paraffin-embedded for histologic evaluation, or snap-frozen in liquid nitrogen for protein isolation and immunoblotting.
For the UT42 intrauterine xenograft study, procedures were modeled after a published study (21). In brief, 6- to 8-week-old female athymic NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) were put under general anesthesia, a small abdominal incision was made, and the left uterine horn was identified and ligated at either end using 4-0 Vicryl suture (Ethicon). Note that 5 × 106 cells suspended in 50 μL sterile PBS were then injected into the left uterine horn; uterine enlargement and absence of visible leakage were confirmed. Injections were performed across 3 days for an equal number of animals per experimental group. The study was ended at 8 weeks from the date of injection. Animals were euthanized, and the left and right uterine horns were isolated, evaluated for tumor presence, and weighed intact for all animals. Metastatic nodules throughout the abdominal cavity were collected, counted, and weighed separately from the uterus as the “metastatic tumor burden.” Collected tissue specimens were formalin-fixed and paraffin-embedded for hematoxylin and eosin (H&E) staining. Upon microscopic evaluation for tumor presence, the largest linear dimension (LLD) was calculated as the sum of the lengths (in mm) of all tumor cell foci observed in a single section. Metastases were also confirmed microscopically.
The UT42 SMAP treatment study was performed using patient-derived xenograft (PDX). Tumor pieces were surgically implanted into the flank of female NSG mice. Successful outgrowth was monitored, and upon reaching a tumor volume of 100 mm3, animals were enrolled and randomly assigned to a treatment group. Treatment was delivered by oral gavage once daily: vehicle control (N,N-Dimethylacetamide, DMA), 50 mg/kg SMAP-1154, or 100 mg/kg SMAP-1154. Animal body weight and tumor volume were measured every 2 to 3 days for the study duration. Animals were observed for signs of toxicity (i.e., debilitating diarrhea, abdominal stiffness, jaundice, hunched posture, and rapid weight loss). Animals were euthanized once morbidity criteria were met as established in the protocol; in all instances, this was due to a tumor size that accounted for >10% of body weight. Terminal sacrifice was considered the survival endpoint for construction of Kaplan–Meier curves. Tumor tissue was harvested 2 hours after a final treatment dose and was formalin-fixed for IHC evaluation or snap-frozen in liquid nitrogen for immunoblotting.
H&E staining was performed following standard protocols. H&E-stained sections were imaged using a Leica SCN400 slide scanner at 20x magnification. Images were exported through the SCN400 Digital Imaging Hub and cropped where appropriate for publication (no other image adjustments were performed).
Protein expression, purification, crystallization, and structure determination
A mouse PP2A Aα P179R–mutant construct was cloned using site-directed mutagenesis based on a mouse WT PP2A Aα pGEX-4T1 construct with an N-terminal GST tag and a tobacco etch virus (TEV) protease cleavage site in between (22). The GST fusion protein was overexpressed in Escherichia coli BL21 (DE3) cells grown in Luria broth media. Bacteria cell pellets were lysed by sonication. The GST fusion proteins were eluted from Glutathione Sepharose 4B beads. GST tag was removed by TEV at 4°C overnight. Then the proteins were further purified by an anion exchange column, and finally purified by gel filtration on a Superdex 200 10/300 GL column (GE Healthcare). The peak fractions were pooled and concentrated to approximately 5 mg/mL in a buffer containing 10 mmol/L Tris HCl, pH 80, 100 mmol/L NaCl, and 2 mmol/L DTT.
The hanging-drop vapor diffusion method for crystallization was used to prepare crystals of mouse PP2A Aα P179R. To obtain protein crystals for structural study, 1 μL of protein sample (5 mg/mL) was mixed with 1 μL of well solution containing 0.2 mol/L sodium malonate, pH 7.0, and 20% w/v PEG 3,350. The crystals were flash frozen by liquid nitrogen in a cryogenic solution containing 0.2 mol/L sodium malonate, pH 7.0, 10% glycerol, and 20% w/v PEG 3,350.
Screening and data collection were performed at the Advanced Light Source, beamline 8.2.1 at wavelength 1.00 Å. Details of the data collection and refinement are provided in the Supplementary Methods. All diffraction data were processed by HKL2000 (23). The initial phase was determined by molecular replacement in Phaser with model from PDB: 2IAE (22, 24). All models were improved using iterative cycles of manual rebuilding with the program COOT and refinement with Refmac5 of the CCP4 6.1.2 program suite (25, 26).
Molecular dynamics simulations
Structural coordinates for PP2A Aα were retrieved from the Protein Data Bank (PDB: 2IAE; ref. 22). The mutants were constructed in silico using the ICM mutagenesis program (27). The Amber ff14SB force field was used to describe the systems, with explicit TIP3P water molecules (28–30). PROPKA 3.1 and PDB2PQR 2.1, as implemented in the proteinPrepare() functionality of HTMD, were used to determine the protonation state of the residues (31). Note that 1,000 steps of steepest descent were used to minimize all systems. Energy minimization and equilibrium for volume relaxation were carried out under NPT conditions at 1 atm, with initial velocities sampled from the Boltzmann distribution at 300 K. The temperature was kept at 300 K by a Langevin thermostat. The PME algorithm was used for electrostatic interactions with a cutoff of 1.2 nm. A 1.05 μs production run was carried out for all the systems in the NVT ensemble with hydrogen mass repartitioning that allowed for a time step of 4 fs. All simulations were performed using the ACEMD program using an identical protocol (continued in the Supplementary Methods; refs. 32–38).
Results are displayed as mean ± SD, unless otherwise indicated. In each figure legend, n indicates the number of independent experimental replicates completed or, when in reference to an in vivo study, the number of animals per group. These were distinct samples whose data points were averaged to obtain group means. The statistical tests used are identified in the figure legends and/or text. All Student t tests performed were two-sided. GraphPad Prism 7 software was utilized for graph construction and statistical analysis.
The atomic coordinate and structure factor of mouse PP2A Aα P179R is deposited in the PDB with the accession code 6EF4. Other data that support the findings of this study are available from the corresponding author upon reasonable request.
Details regarding the application and use of previously published algorithms were available upon request.
Mutation of PPP2R1A at P179 and S256 is specific to high-grade subtypes of endometrial carcinoma
Although three amino acid residues are notable PPP2R1A hotspot mutation sites in cancer (Fig. 1A), two of these residues, P179 and S256, display a striking specificity in occurrence that is in contrast to the third, most frequently mutated site, R183 (Fig. 1B). Specifically, P179 and S256 mutations are highly enriched in endometrial carcinoma, occurring nearly exclusively in USC and UCS, and also accounting for nearly all PPP2R1A mutations in these endometrial carcinoma subtypes (Fig. 1B and C). This is in stark contrast to the R183 mutation, which was identified in cancers of >20 different organ systems, with colorectal carcinoma being the most common. Although R183 mutations do occur in endometrial carcinoma, they are much more prevalent in the less aggressive endometrioid carcinoma (UEC) subtype. Critically, this differential pattern of occurrence may indicate differing mechanistic roles through which these mutations contribute to cancer. In addition, mutation at these hotspot sites is also largely mutually exclusive. Across cancer, only one tumor was found to harbor two PPP2R1A hotspot mutations: a mixed ductal and lobular breast carcinoma was identified with both R183W and S256F mutations. In endometrial carcinoma, occasional tumors were found to harbor more than one PPP2R1A mutation but in none of these cases were there two co-occurring hotspot site mutations (Supplementary Fig. S1A).
The P179 and S256 mutation sites reside in HEAT domains 5 and 7, respectively, of the PP2A Aα-subunit. The P179 mutation is the more common of the two and is substituted with arginine (P179R, 48 cases), lysine (P179L, 2 cases), or threonine (P179T, 1 case). The S256 is mutated to phenylalanine (S256F, 15 cases) or tyrosine (S256Y, 6 cases). Importantly, large-scale sequencing studies have revealed that 28% to 32% of USC patients and 8% to 12% of UCS patients harbor one of these mutations (The Cancer Genome Atlas, MSK-IMPACT; Supplementary Fig. S1A). These patient cases display features characteristic of the high-grade USC and UCS subtypes in which they occur, and which are distinct from, and often in contrast to, the characteristic features of UEC tumors. First, P179-mutant tumors are frequently high-stage, with greater than 50% of patients diagnosed with stage III or IV disease (Supplementary Fig. S1B). A stage III or IV diagnosis indicates cancer progression involving extrauterine dissemination. This advanced disease is typical of USC and UCS, whereas 80% or more of UEC tumors are low-stage. Correspondingly, P179 tumors, similar to all USC tumors, display poor disease-free survival (Supplementary Fig. S1C). At the genome level, P179 tumors present with other defining features of USC and UCS, such as a low overall mutational burden, high copy-number alteration, TP53 mutant, and MYC, CCNE1, or ERBB2 amplification (Supplementary Fig. S1D–S1F). Hallmarks that distinguish UEC, including microsatellite instability (MSI) or PTEN, ARID1A, or CTNNB1 mutation, are rare in P179 tumors.
We sought to identify patient tumor specimens harboring recurrent PPP2R1A mutations in the Case CCC Gynecologic Tumor Biobank. Seventy-four endometrial carcinoma tumors and 18 ovarian carcinoma tumors underwent targeted Sanger sequencing at the HEAT 5 and 7 hotspot foci. Sixteen percent (15 tumors) were found to harbor PPP2R1A mutations, and all were at residues P179, R183, or S256 (Supplementary Fig. S2A and S2B). In addition to the previously reported P179R and P179L variants, we also identified a new variant, P179H, in a mixed serous-/endometrioid-subtype tumor (Supplementary Fig. S2B and S2C). In this cohort, 38% of USC tumors were found to harbor P179/S256 site mutations (Supplementary Fig. S2A). Consistent with previous studies, our Sanger sequencing results demonstrated that these tumors are frequently heterozygous for mutation, with a WT allele remaining (Supplementary Fig. S2C; refs. 39–41). However, for a subset of tumors, there was no WT allele detected, which suggests that mutations can be either homozygous or co-occur with allelic deletion.
P179R-Aα disrupts PP2A holoenzyme assembly
Studies have previously reported altered interactome and holoenzyme complex formation with cancer-associated mutant isoforms of the PP2A Aα-subunit (16–18). However, these studies were not performed in USC or UCS endometrial carcinoma cells, which is the most relevant disease context for studying the P179 and S256 mutations. Thus, we generated a primary serous endometrial carcinoma cell line, UT89, to characterize the interactome of P179R, which represents the most common missense mutation in high-grade endometrial carcinoma. For comparison, we also evaluated S256F and R183W. Cell lines with stable expression of V5-tagged WT or mutant Aα-subunit were generated using a lentivirus approach, and tagged Aα protein was coimmunoprecipitated with its binding partners from whole-cell lysates. Binding to PP2A family members was evaluated by Western blotting of coimmunoprecipitants. Overall, the P179R mutation induced significant disruption of PP2A holoenzyme formation, presenting with a significant reduction in binding to members of each B-subunit family except the Striatins, for which binding remained intact, as well as a significant loss of catalytic C-subunit binding (Fig. 2A–C). By comparison, S256F and R183W mutants also demonstrated losses in B-subunit interactions, but in many cases, these losses were less severe. Interestingly, P179R and S256F shared a near-complete loss of binding to the B55α-subunit, whereas R183W retained some binding. Importantly, P179R was the most severely impaired in its interaction with the PP2A catalytic C-subunit. Finally, a published report identifies increased TIPRL binding to mutant Aα as a dominant-negative mechanism through which PP2A is inactivated in cancer (18). We therefore also assessed mutant protein binding to TIPRL. From our results, the P179R and S256F mutants do not display increased TIPRL binding in a serous endometrial carcinoma cell line (Fig. 2A). Meanwhile, R183W did display this gain of binding.
To complement the co-IP approach, we additionally employed a cell-free binding assay to evaluate the impact of P179 mutation on holoenzyme assembly (Supplementary Fig. S3A and S3B). GST-tagged B56α-subunit was incubated with purified C-subunit and purified WT or mutant Aα-subunit proteins, and then isolated and assessed for binding. When in the presence of P179 mutant Aα isoforms, assembly of the complete Aα-B56α-C holoenzyme heterotrimer was impaired. This is notable, as even the sole loss of B56α holoenzymes has been identified as sufficient for cell transformation when substituted into an established transformation model involving PP2A inactivation (13).
Altogether, these data support the previous literature describing loss of PP2A subunit interactions upon Aα mutation (16–18). Moreover, it highlights a near-complete loss of catalytic subunit binding to P179R-Aα, which ultimately suggests a marked change in protein structure that influences C-subunit contact points distant from the site of mutation.
The P179R-Aα protein displays altered conformational dynamics
To understand how the overall protein structure may be altered by the P179R mutation, we determined the crystal structure of the P179R-mutant protein at 3.4Å resolution (Fig. 2D and E). Comparison of this PP2A P179R-Aα structure with the WT-Aα crystal structure (PDB: 1B3U) revealed an obvious conformational difference between the mutant and WT proteins (Fig. 2D and E, overlays; ref. 42). Although we cannot rule out a contribution from crystal packing, our crystal structure indicates a change of the conformation-energy landscape of the PP2A P179R-Aα protein.
Therefore, we next performed classical and enhanced sampling molecular dynamics simulations of WT-Aα and P179R-Aα to identify stable protein conformations. Free energy landscape (FES) plots were generated as a function of ϕ and ψ dihedral angles of the P179 and mutant R179 residues, for both apo Aα and Aα in complex with the C-subunit (Fig. 3A and B; Supplementary Fig. S4A–S4P). These landscapes provide evidence that the R179 residue of the Aα mutant is capable of exploring an altered FES to adopt multiple metastable conformations of the Aα-subunit protein, which may alter interactions with other PP2A subunits and proteins. We further extracted and analyzed the most stable conformations from the free energy minima of the WT (basin A) and mutant (basin C) FES (Fig. 3A and B). Importantly, when we extrapolate the obtained P179R-Aα crystal structure on the FES of the simulated P179R-Aα, we found that the conformation adopted by the crystal structure is indeed one of the stable conformations and can be found in the largest minimum of the R179 apo Aα FES (Fig. 3B, indicated by star).
Due to a cyclized side chain, P179 does not make interresidue interactions in the WT crystalline conformation or throughout the course of any of our molecular dynamics simulations. By contrast, the R179 residue of the mutant protein makes direct interactions with nearby residues. In the resolved crystal structure, the guanidinium side chain of R179 orients toward the solvent, but still interacts with the side chain of Q217 of the adjacent HEAT repeat (Fig. 3C, white structure). Comparably, extracting the most stable, representative conformation from the largest basin on the FES identifies an ion pair interaction between R179 and E216 (Fig. 3C, cyan structure). In both, the interaction of R179 is facilitated by an additional ion pair interaction between R182 and D215.
The cyclic side chain of proline is effective in imparting conformational rigidity to the secondary structure of proteins (43). Proline is also exceptional in the cis-trans isomerization of the peptidyl-prolyl backbone dihedral angle (ω; defined as Cα-C-N-Cα) preceding itself (0° for cis and ±180° for trans). The cis-trans configuration is isoenergetic, with the difference being approximately 1 kcal/mol between the two states (44–46). This isomerization is an important structural mechanism through which proteins achieve large conformational changes and reach various macrostates of multidomain proteins without affecting the covalent structures. Therefore, mutation of proline can have an influence not only on the conformation of the protein but also on these biological processes (45–48). To study the structural influence of P179R mutation on isomerization of the Aα P179 residue, including its dynamics in binding to the catalytic C-subunit, we carried out additional simulations of the WT and mutant residues. First, we analyzed cis-trans state as a function of ω dihedral of P179 (Cα178-C178-N179-Cα179) in both Aα/C complex and apo states of Aα. The simulations identified that the ω dihedral of the WT P179 residue exists predominantly in the trans configuration (Supplementary Fig. S4Q and S4R). This is more pronounced in the apo state when compared with the complex. By contrast, in the P179R mutant, the ω dihedral of R179 preferentially adopts a cis configuration in apo and complex states (Supplementary Fig. S4S and S4T). In this cis conformation, R179 makes a strong ion pair with D215, whereas R182 interacts with N211 (Fig. 3D).
Next, we analyzed the global conformational changes and the effect of mutation on C-subunit binding to the Aα-subunit. The FES plots indicated a complete change in conformational landscape between the WT and mutant complex (Fig. 3A and B). We extracted the representative conformations from the most populated minimum of the Aα/C complex FES (basin A in P179 WT and C in R179 mutant). The WT-Aα-subunit adopts a “clam-like” conformation, with the catalytic subunit sandwiched between the two ends of the Aα-subunit (Fig. 3E). The catalytic subunit is deeply embedded within the ends, thereby maximizing its interactions with the Aα-subunit. In the P179R mutant, the Aα-subunit prefers to adopt a near-closed “donut-shaped” conformation (Fig. 3F). The result is that the C-subunit appears to be squeezed out of the complex and now makes peripheral interactions with the Aα-subunit, which is indicative of a destabilized complex. We calculated the binding energy for the C-subunit to either the WT or P179R mutant Aα-subunit, and determined them to be 236 and 347 kJ/mol, respectively. The greater binding energy between P179R-Aα and C, versus WT-Aα and C, is indicative of a greater C-subunit binding affinity for the WT isoform than the mutant isoform. This is again consistent with a destabilized complex and in support of observations from the extracted structures. Altogether, these data show that a P179R mutation of the Aα protein leads to conformational disruption and a destabilized interaction with the C-subunit, which in turn lends context to the altered P179R-mutant interactome described previously.
Impaired binding to P179R-Aα results in C-subunit destabilization
Expression of the P179R-mutant isoform in UT89 cells was found to reduce total protein levels of C- and B-subunits to a degree that reflected their respective loss of binding in co-IP experiments (Fig. 4A). Indeed, the loss of catalytic C-subunit total protein, as well as the regulatory B55α- and B56α-subunit proteins, was significant for the P179R mutant, which had demonstrated marked impairment of binding and conformation-related destabilization of the Aα/C complex (Fig. 4B). This led us to hypothesize that impaired subunit incorporation into the trimeric holoenzyme results in increased degradation of the free subunit monomers. Real-time PCR verified that there was no change in PPP2CA (C-subunit), PPP2R2A (B55α), or PPP2R5A (B56α) mRNA, confirming that the decreased protein resulted from a posttranslational mechanism (Fig. 4C). We then asked whether this protein loss could be reversed upon treatment with the proteasome inhibitor Velcade. After a 24-hour treatment, total C-subunit protein was restored to a level comparable with that of EGFP control lines (Fig. 4D and E), implicating proteasome-mediated degradation of C-subunit in P179R-expressing lines. Evaluation of a B-subunit isoform, B55α, revealed that its abundance was also increased with proteasome blockade in P179R-expressing cells, as well as in the EGFP and WT lines (Fig. 4D and E). This could suggest that at least some B-subunits are present in excess and continuously turned over. To further verify decreased subunit stability in the presence of P179R-Aα, we investigated protein half-life using the translation inhibitor cycloheximide. P179R-Aα–expressing cells demonstrated an increased rate of C-subunit degradation, represented by the slope of protein reduction over time, which was significantly different from both EGFP and WT-Aα–transduced cells (Fig. 4F and G). Although EGFP and WT demonstrated modest C-subunit degradation within the experiment timeframe (∼20%–30% reduced relative to time = 0 hour), the higher rate of reduction in P179R allowed for calculation of a half-life of 36 hours, at which point C-subunit total protein was 50% reduced. The protein degradation rate of B55α was also significantly increased in P179R-mutant cells relative to EGFP or WT cells (Fig. 4F and G). Finally, we evaluated a panel of PPP2R1A WT and P179R-mutant patient tumor specimens from the Case CCC tumor biobank for C- and B-subunit protein levels; WT samples representative of both serous and endometrial carcinoma tumors were included (Supplementary Fig. S5A). We calculated the ratio of C:A (C-subunit to A-subunit) and B:A (B55α-subunit to A-subunit) to account for variation in the basal level of A-subunit protein present in individual tumor specimens. Overall, P179R-mutant tumors had lower levels of C- and B-subunit protein when compared with WT serous or endometrioid tumors (Supplementary Fig. S5A–S5C). The differences were statistically significant for B55α (Mut vs. Serous P = 0.0012; Mut vs. Endometrioid P = 0.0028), and trended toward significant for the C-subunit (Mut vs. Endometrioid P = 0.0688; Mut vs. Serous P = 0.1838).
Given the PP2A Aα mutant–driven disruption of PP2A holoenzyme assembly and reduction of catalytic subunit protein, and that PP2A has prominent tumor-suppressive functions, we investigated whether expression of P179R-Aα in the PPP2R1A WT UT89 cell line led to an alteration of in vitro or in vivo tumor cell growth. No significant difference in clonogenesis or tumorigenesis was observed by colony formation assay or subcutaneous xenograft assay, respectively, when P179R-Aα was introduced into a complete WT background (Supplementary Fig. S6A and S6B).
Restoration of WT PP2A function suppresses malignant features of P179R-Aα–mutant cells
We derived another primary patient cell line, UT42, from a patient tumor that harbors the P179R mutation endogenously, in order to directly characterize mutant biology (Supplementary Fig. S2). This model system provides the environmental, cellular, and genetic context native to a cell that acquired the P179R mutation as it underwent transformation. The UT42 cell line can be stably maintained in cell culture for greater than 30 passages, and was verified to retain the P179R mutation by Sanger sequencing. UT42 cells were transduced with V5-tagged WT-Aα or P179R-Aα to evaluate the impact of reconstitution with WT Aα protein on tumorigenic potential (Fig. 5A). Expression of WT-Aα protein increased C- and B-subunit protein, whereas expression of additional P179R-Aα did not, suggesting that these subunits are stabilized when in the presence of the binding-competent WT protein (Fig. 5B and C). In this model system, co-IP of the V5-tagged P179R-Aα displayed an 80% reduction in C-subunit binding, which is consistent with data from UT89 (Fig. 5C and D). To assess the catalytic activity of reconstituted WT-Aα and compare with that of mutant P179R-Aα protein, a phosphatase activity assay was performed on the coimmunoprecipitants of WT- or P179R-Aα V5-tagged protein from the UT42 isogenic lines. Robust dephosphorylation of the substrate DiFMUP was observed for extracted WT-Aα protein bound to C- and B-subunits (Fig. 5E); this activity could be reversed by treatment with the serine/threonine phosphatase inhibitor OA (49). By contrast, significantly less dephosphorylation activity was recorded for extracted P179R-Aα and was comparable with that seen with OA-mediated PP2A inhibition, suggesting a near-complete loss of catalytic activity with this recurrent, patient-derived Aα-mutant protein. Phosphatase activity has been normalized to the relative amount of C-subunit in the pull-down isolate, confirmed by Western blotting, to account for the decreased catalytic C-subunit binding to P179R-Aα. Finally, expression of WT-Aα led to dephosphorylation of the established PP2A substrates β-catenin and GSK3β at the specific residues against which the PP2A-B55α and PP2A-B56δ holoenzymes, respectively, have previously demonstrated phosphatase activity (Fig. 5F; refs. 50–52). Consistent with the literature, dephosphorylation of β-catenin led to increased total protein as this event protects it from degradation.
A colony formation assay was used to assess the impact of WT-Aα expression on in vitro clonogenic growth. UT42 cells reconstituted with WT-Aα protein were significantly impaired in their ability to form colonies, whereas expression of additional mutant P179R-Aα protein did not affect clonogenesis compared with control (Fig. 5G).
Using both genetic and pharmacologic approaches, we sought to determine the sensitivity of UT42 tumor growth to restoration of PP2A activity. First, UT42 cells that were reconstituted with catalytically active, WT Aα protein were injected s.c. in the flank of female mice. Tumor growth was compared with that of control EGFP-expressing cells (Fig. 6A). Expression of WT-Aα protein significantly inhibited subcutaneous tumor growth.
We next employed an orthotopic xenograft model with tumor cell injection into the uterus (Fig. 6B). In a pilot study with parental UT42 cells, all animals developed primary tumors within the uterus by 8 weeks, as well as intraperitoneal metastatic nodules; no lymph node enlargement was discerned. This model thus allows for assessment of both primary tumor growth within the native environment of the uterus corpus, as well as cell metastatic potential. We carried out the full experiment under the same experimental conditions, injecting UT42 EGFP or WT-Aα cells into the left uterine horn of female NSG mice. At study end, tumor burden was evaluated macroscopically through gross inspection of the isolated gynecologic tract, as well as inspection of the pelvic cavity, abdominal cavity organs, and lungs for metastatic nodules. Tissue was collected for further microscopic evaluation and verification of tumor cell presence. At 8 weeks, 4 of 7 EGFP animals had tumor filling the entirety of the injected uterine horn, and the remaining 3 animals had partial tumor presence visible macroscopically as a mass along the uterine wall (Fig. 6B; Supplementary Fig. S6C and S6D). By comparison, no WT animals had tumor filling the uterine cavity; four had partial tumor presence along the uterine wall, and the remaining three had no macroscopically discernable tumor presence. Reduced tumor formation resulted in a significantly reduced uterus weight, despite apparent hydrometra (Fig. 6C). WT animals also demonstrated a significant decrease in metastatic burden, as determined by metastasis nodule weight and nodule count (Fig. 6C and D). In both groups, metastatic nodules were predominantly peritoneal and surrounded by subperitoneal fat. No organ-invasive metastases were identified in WT animals. By contrast, 2 EGFP animals displayed metastatic nodules that were invasive into the liver parenchyma, and 1 animal had a retroperitoneal metastasis attached to the kidney capsule (Supplementary Fig. S6D and S6E). It should be noted that it is not possible to separate the decreased occurrence of metastasis from the decreased primary tumor burden of WT animals as an independent phenotype in this study. We did however perform a scratch wound assay to assess whether the WT-Aα cells may be impaired in migratory function and observed no significant difference from EGFP cells (Supplementary Fig. S6F).
In H&E-stained sections, tumor growth was found to expand within the endometrial layer of the uterus and through the myometrium, frequently present within parametrial soft tissue and occasionally encasing adnexal structures (Fig. 6E). Growth in the outer uterine wall, subserosa, and parametrial soft tissue was observed in animals from both experimental groups. This growth pattern mirrors typical tumor growth observed in human tumor specimens. One additional WT animal was found to have microscopic tumor foci that were not apparent during gross evaluation. The LLD was calculated for each animal as an aggregate measure of tumor foci size. Expression of WT-Aα led to a marked reduction in tumor foci size relative to the tumors formed by EGFP-expressing cells (Fig. 6F).
Finally, we sought to evaluate the sensitivity of the UT42 P179R mutant tumor to a pharmacologic method of PP2A reactivation. We have previously reported on a novel series of SMAP that demonstrate in vivo suppression of tumor growth via activation of PP2A (19, 53, 54). SMAP-061 and SMAP-1154 are two lead compounds of the SMAP series, which was generated through reverse engineering of tricyclic neuroleptics to remove anti-CNS toxicity and enhance antiproliferative properties (structures published in ref. 19; ref. 55). Through in vitro characterization, the UT42 cell line demonstrated robust growth inhibition with SMAP treatment and displayed an IC50 that was significantly lower than three other PPP2R1A WT endometrial carcinoma lines, including UT89 (Supplementary Fig. S7A). Relative to UT89, the UT42 cell line also demonstrated greater sensitivity and marked growth impairment in the face of SMAP treatment during a 2-week colony formation assay (Supplementary Fig. S7B).
To investigate SMAP response in vivo, PDXs of the UT42 tumor were implanted subcutaneously in the flank of female mice and treated with vehicle (DMA), 50 mg/kg SMAP-1154, or 100 mg/kg SMAP-1154 once daily. Tumor growth was markedly reduced by SMAP treatment at both doses (Fig. 6G). Animals receiving SMAP treatment did not display signs of weight loss (Fig. 6H) or other side effects during monitoring. From the initiation of treatment to study end, vehicle-treated tumors demonstrated robust increases in tumor volume, whereas minimal change in tumor volume was observed with SMAP treatment (Fig. 6I). Critically, animals in the SMAP treatment groups exhibited a greater rate of survival (Fig. 6J), as tumors receiving vehicle treatment grew rapidly to meet morbidity criteria for animal euthanasia. Evaluation of protein isolates collected from vehicle- or SMAP-treated tumors confirmed dephosphorylation of the PP2A substrates Akt, GSK3β, and c-Myc (Supplementary Fig. S7C and S7D). Alteration of these targets is consistent with our previous works investigating the treatment of tumors with SMAPs (19, 53, 54). Overall, the UT42 tumor harboring a P179R mutation was highly responsive to SMAP treatment, which significantly reduced its growth in vivo.
PP2A provides critical regulatory activity that counter-balances kinase-mediated phosphorylation to maintain appropriate cell growth, regulate cell division, and prevent tumor development. A recurrent mutation of the PP2A Aα scaffolding subunit (PPP2R1A) is present in approximately 30% of high-grade subtypes of endometrial carcinoma, with mutation at the P179 site accounting for approximately 20% to 25%. The UT42 patient-derived cell model harbors an endogenous P179R mutation and thus provides a unique system in which to investigate mutant biology as this tumor developed in the immediate context of somatic PPP2R1A mutation. The striking impairment of tumor formation and metastasis that occurred when WT Aα protein was restored provides evidence that functional disruption of PP2A is critical to this tumor and its malignant features. In a similar manner, tumor growth was significantly reduced when treated with a pharmacologic PP2A activator, again highlighting sensitivity of the P179R mutant UT42 tumor to PP2A activity. Overall, PP2A's robust roles as a suppressor of transformation and tumor development have been well-documented and reviewed in the literature, to which our outcomes add further emphasis on its importance as a tumor suppressor in high-grade endometrial carcinoma (56–58).
P179 site mutations are notable for displaying marked disease specificity, occurring almost exclusively in endometrial carcinoma and the serous carcinoma and carcinosarcoma subtypes. This pattern suggests a role for PP2A as a disease driver, in which the biochemical consequence of mutating this residue is specifically advantageous to endometrial tumorigenesis. Consistent with this perspective, a large-scale sequencing study has shown that PPP2R1A mutations are somatic and truncal, and so are presumed to be acquired during the early processes of malignant cell transformation (9). Several research groups have investigated PP2A as a tumor suppressor whose functional loss contributes to cell transformation. Interestingly, a majority of developed models rely on coinactivation of p53 with PP2A for complete transformation; meanwhile, TP53 and PPP2R1A P179 mutations highly co-occur and are both truncal in these otherwise low mutational burden endometrial carcinoma tumors (8–10, 59). P179 mutation may therefore be a mechanism through which models of PP2A in transformation bear out in endometrial carcinoma.
To understand how the P179R mutant alters PP2A function, we began with investigation of PP2A subunit binding. Our findings are in agreement with previous literature that identified disrupted PP2A subunit interactions due to P179R PPP2R1A mutation (17, 18). The consequence for unbound B- and C-subunits, an increased rate of protein degradation, is a novel finding for this cancer-associated mutation but is in support of a standing view that monomeric PP2A B- and C-subunits are less stable than their trimeric counterparts (11, 60). Together, these findings suggest that acquisition of a P179R mutation will impede assembly of holoenzymes containing catalytic and regulatory subunits and thereby induce a corresponding loss of canonical PP2A function within the cell.
The P179 residue resides at the A/B subunit interface, and alteration of its side chain chemistry could reasonably be predicted to directly impair B-subunit binding. On the other hand, significant loss of C-subunit binding suggested to us that this mutation had a larger, global effect on the scaffolding protein's conformation. By utilizing crystallography and molecular dynamics modeling, we provide context to the interactome dataset by detailing the structural alterations induced by the P179R substitution. Simulations of the WT protein reveal proline residue 179 as a site of cis-trans isomerization that can confer dynamic flexibility to the A-subunit scaffolding protein. The substituted arginine changed the dominant isomerization state and introduced new interresidue interactions within the protein tertiary structure. The P179R substitution modifies the A-subunit's most stable, and preferentially adapted, conformation in a manner that is unfavorable to C-subunit binding, as evidenced by Aα/C complex simulations for the mutant isoform as well as its increased binding energy requirement. In addition, we were able to resolve the crystal structure of the mutant protein at a 3.4 Å resolution. Although published crystallographic structures have resolutions ranging from 1 to 4 Å, with an average resolution of approximately 2.2 Å, many structures, especially those with all-helical secondary structures such as the PP2A Aα protein, can be clearly resolved in 3 to 4 Å without ambiguity. This is because alpha-helices have more characteristic electron densities and have well-defined main-chain structural restrictions. P179R-Aα is the first resolved crystal structure for a cancer-derived mutant PP2A protein, and it underscores that mutation is a pathogenic mechanism through which cancer cells can disrupt the PP2A structure.
Importantly, our work suggests that a previously described mechanism of PP2A inactivation by PPP2R1A hotspot mutation may not hold true for P179R (18). Although some PPP2R1A mutations display increased binding to the PP2A inhibitor TIPRL and may be inactivated through a dominant-negative mechanism, P179R did not display an increase in TIPRL binding in our model systems. This included protein co-IP in both a WT serous endometrial carcinoma cell line (Fig. 2A), as well as within a cell line that itself harbors the P179R mutation endogenously (Fig. 5C). It is possible that the conflicting observations in P179R interactome reflect on the context specificity of this mutation; however, a targeted investigation would be needed to lend more conclusive clarity to this discrepancy.
There is a strong clinical need for new therapeutic options in endometrial carcinoma, and in particular for high-grade subtypes like serous endometrial carcinoma that portend a poor prognosis due to tumor recurrence. Recently, immunotherapy has become an exciting prospective therapeutic opportunity for microsatellite instable endometrial carcinoma tumors due to their abundance of neoantigens (61). However, tumors that harbor P179 mutations are unlikely to be amenable to the immunotherapy approach as we show this mutation occurs exclusively in low mutational burden, MSI-negative tumor types. The outcomes of the in vivo work presented here highlight a path forward for use of a targeted therapeutic approach that capitalizes on PP2A reactivation.
Finally, in addition to cancer, instances of germline PPP2R1A mutation have been reported in association with developmental and intellectual disabilities (62, 63). A P179L mutation was one of several missense mutations identified. Whether this confers increased risk for cancer is unknown. From our results, one could reasonably predict these mutations lead to significant disruption of PP2A physiologic activity within neuronal cells.
Overall, this body of work suggests a loss of function of PP2A tumor-suppressive activity due to P179R mutation of the Aα-subunit. “Rescue” through expression of holoenzyme-forming and catalytically competent WT Aα protein, or through pharmacologic PP2A activation, suppressed tumorigenesis. These findings support PPP2R1A P179R mutation as key driver of disease in the 20% to 30% of high-grade endometrial carcinoma tumors that harbor one, and present pharmacologic targeting of PP2A as a potential therapeutic direction for this patient population.
Disclosure of Potential Conflicts of Interest
K. Resnick reports receiving honoraria from the speakers' bureau of Clovis. G. Narla is CSO at RAPPTA Therapeutics, has an ownership interest (including stock, patents, etc.) in RAPPTA Therapeutics, and is a consultant/advisory board member for HERA. The Icahn School of Medicine at Mount Sinai has filed patents covering composition of matter on the small molecules disclosed herein for the treatment of human cancer and other diseases (International Application Numbers: PCT/US15/19770, PCT/US15/19764; and US Patent: US 9,540,358 B2). Mount Sinai is actively seeking commercial partners for the further development of the technology. G. Narla has a financial interest in the commercialization of the technology. No potential conflicts of interest were disclosed by the other authors.
Conception and design: S.E. Taylor, C.M. O'Connor, S. Haider, A. DiFeo, G. Narla
Development of methodology: S.E. Taylor, G. Shen, D. Leonard, W. Xu, S. Haider, A. DiFeo, G. Narla
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.E. Taylor, C.M. O'Connor, Z. Wang, H. Song, D. Leonard, J. Sangodkar, C. LaVasseur, S. Avril, S. Waggoner, K. Zanotti, A.J. Armstrong, C. Nagel, K. Resnick, W. Xu, A. DiFeo
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.E. Taylor, Z. Wang, H. Song, D. Leonard, J. Sangodkar, S. Avril, W. Xu, S. Haider, A. DiFeo, G. Narla
Writing, review, and/or revision of the manuscript: S.E. Taylor, C.M. O'Connor, Z. Wang, D. Leonard, J. Sangodkar, C. LaVasseur, S. Avril, W. Xu, S. Haider, G. Narla
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.E. Taylor, H. Song, W. Xu, S. Haider
Study supervision: S.E. Taylor, M.W. Jackson, W. Xu, S. Haider, G. Narla
The authors would like to thank the Case CCC, which supports the Gynecologic Tumor Biobank, for the invaluable resources provided by this work; Kristen Weber Bonk and the CWRU Athymic Animal and Xenograft Core for their continued support with the in vivo studies; Daniela Schlatzer and the CWRU Proteomics Core for support with proteomics work; and finally, Richard Lee and the CWRU Light Microscopy Imaging Core. The Imaging Core's Leica SCN400 slide scanner utilized for this work is made available through the NIH Office of Research Infrastructure (NIH-ORIP) Shared Instrumentation Grant (S10 RR031845). Funding for this work was provided by grants from the NIH-NCI to G. Narla (R01 CA181654), A. DiFeo (R01 CA197780), and S.E. Taylor (F30 CA224979); the Department of Defense to A. DiFeo (OC150553); and The Young Scientist Foundation to A. DiFeo. S.E. Taylor and D. Leonard are additionally supported by T32 GM007250 (NIH-NIGMS), and C.M. O'Connor by T32 GM008803 (NIH-NIGMS). S. Avril is supported by a Clinical and Translational Science Award KL2 TR0002547 (NIH-NCATS).
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.