Discovery and functional characterization of a neomorphic PTEN mutation

Paired-end Illumina 100-base pair (bp) reads were mapped to the reference genome, and, after quality filtering, realignment, and recalibration, the average fold-coverage of tumor and normal tissue was 98× and 47×, respectively (SI Appendix, Table S1). We developed a somatic variant-calling pipeline to comprehensively characterize single nucleotide variants (SNVs), insertions/deletions (INDELs), structural variation (SV), and copy number variants (CNVs) present uniquely within the tumor genome (SI Appendix, Fig. S2). This approach allows for detection of novel driver lesions within a patient that are overlooked by targeted SNV panels focusing on known mutations of a few dozen genes. We identified a total of 6,050 tumor-specific variants consisting of 4,823 SNVs, 356 INDELs, 867 copy number variants, and four structural variants (Table 1). We note that the allelic fraction distribution of detected SNVs had an upper quartile value of 20%, which is in line with pathological estimates of ∼15% tumor purity (SI Appendix, Fig. S1 A and B). We manually curated all structural variants and any copy number variants within coding or regulatory regions. In parallel, to prioritize potentially pathogenic SNV and INDEL mutations, we focused on rare (≤2% 1,000 Genomes global allele frequency) variants that were predicted to be deleterious coding or regulatory mutations (Table 1). After applying this stringent set of criteria, we were left with a total of 12 candidate driver mutations to further investigate (SI Appendix, Table S2).

Manual curation of the 12 variants identified a nonsynonymous somatic substitution predicted to convert alanine to glycine at amino acid 126 (p.A126G) of the PTEN protein. Given the well-characterized role of PTEN as an important modulator of prostate cancer growth (9) and the location of the mutation in the active site of the PTEN protein, we prioritized this genetic lesion as the top candidate driver mutation. Interestingly, this mutation has been detected in an invasive breast carcinoma case within The Cancer Genome Atlas database (case ID: TCGA-LL_A5YP) (2) but has remained uncharacterized or targeted in extant somatic variant detection platforms (Illumina TruSeq Amplicon v1 or Ion AmpliSeq v2). PTEN is a commonly mutated tumor suppressor that, by dephosphorylating PI(3,4,5)P₃ (PIP₃) and PI(3,4)P₂ to PI(4,5)P₂ and PI(4)P, respectively, functions as a phosphoinositide (PI) 3-phosphatase, inhibiting the PI3K/Akt pathway controlling cellular proliferation, metabolism, and growth (10, 11). Studies have shown that decreased activity or loss of PTEN can elicit cancer phenotypes so it is traditionally classified as a tumor suppressor gene (i.e., a gene where loss-of-function mutations lead to oncogenesis) (12). However, recent work suggests a more expansive role for PTEN in cancer biology. For example, mutant PTEN alleles have been shown to inhibit WT PTEN activity by interacting in a dominant negative manner (13). Amino acid 126 (Fig. 1A, green) is within the highly conserved catalytic “P-loop” site of PTEN and is two amino acids away from the catalytic cysteine at residue 124 (purple) (Fig. 1 A and B). Curiously, the p.A126G mutation produces a binding pocket amino acid sequence identical to VSP family proteins (14–16), which are PTEN-homologous phosphoinositide 5-phosphatases (17) (Fig. 1C). Although the mutation does not clearly disrupt the P-loop (Fig. 1D), mutant homology with VSPs led us to hypothesize a change in substrate specificity for the mutant form of PTEN. To this end, we used the Ins(1,3,4,5)P₄ molecule to model PI(3,4,5)P₃, calculating binding free energies of 3- and 5-oriented substrate in the PTEN WT and mutant binding pockets.

Using the available crystal structure of PTEN bound to L(+)-tartrate (18), we obtained docking coordinates for the Ins(1,3,4,5)P₄ substrate. Ins(1,3,4,5)P₄ was superimposed with the tartrate molecule to orient it with either the 5- or 3-phosphate positioned in the catalytic site (SI Appendix, Fig. S3 A and B). Our calculations, based on these simulations of binding free energy, show that the WT active site thermodynamically favors a 3-oriented substrate whereas the mutant p.A126G allele prefers a 5-oriented substrate (Fig. 1E and SI Appendix, Fig. S3C). Further, we repeated these analyses on other reported residue 126 somatic mutations (COSMIC v64) (1) and found that none were predicted to produce the gain-of-function activity identified in the A126G PTEN mutant (Fig. 1E and SI Appendix, Fig. S3C). These results suggest that p.A126G shifts binding preference toward 5- rather than 3-oriented phosphoinositides, causing a shift in PTEN enzymatic specificity.

To validate the prediction that p.A126G shifts PTEN from a 3- to 5-phosphatase, we performed in vivo enzymatic assays using PTEN phosphoinositide substrates to assess the activity of WT, catalytically dead (p.C124S), and mutant (p.A126G) PTEN proteins. We performed these experiments by creating a voltage-dependent VSP-PTEN chimera (19, 20) that we can selectively activate and monitor for changes of phosphoinositide levels in living cells (Fig. 2A). When activated by depolarization, p.A126G PTEN (Fig. 2B, in green) shows the ability to reduce PI(3,4,5)P₃ levels, albeit at a reduced rate relative to WT (in black) (Fig. 2B). However, p.A126G PTEN produces PI(3,4)P₂ whereas WT depletes it (Fig. 2C). Unlike VSP phosphatases, PI(4,5)P₂ levels remain unchanged during activation in our engineered p.A126G PTEN, suggesting that the mutation had no detectable activity toward PI(4,5)P₂ (Fig. 2D). To demonstrate the specificity of the signals, a phosphatase-dead PTEN mutant (p.C124S) is shown (Fig. 2 B and C, purple). Further, we repeated these experiments on other reported residue 126 somatic mutations (COSMIC v64) (1) and found that none produced the gain-of-function activity identified in the p.A126G PTEN mutant (Fig. 2 E and F). Additionally, we performed ELISAs measuring cellular PI(3,4,5)P₃ and PI(3,4)P₂ levels in PTEN-null prostate cancer cells (PC-3) transiently expressing canonical WT, catalytic dead (p.C124S), and mutant (p.A126G) PTEN (Fig. 3 A and B). We found that p.A126G PTEN significantly decreases PI(3,4,5)P₃ (Fig. 3A) (P ≤ 0.05) and increases PI(3,4)P₂ (Fig. 3B) (P ≤ 0.05) relative to the catalytic dead p.C124S allele, consistent with our initial findings. These results suggest that the p.A126G mutation shifts PTEN enzymatic specificity from a 3- to a 5- phosphatase.

To assess the sole impact of the p.A126G PTEN enzymatic specificity change on the regulation of PI3K/Akt cellular signaling, we transiently expressed WT, catalytic dead (p.C124S), and mutant (p.A126G) PTEN in PTEN-null PC-3 cells. We found that p.A126G PTEN significantly increases phospho-Akt and phospho-S6 levels in the cell relative to WT (Fig. 3 C–E). These results demonstrate hyperactivation of the PI3K/Akt pathway, which is associated with tumorigenic phenotypes, including increased cell proliferation and reduced cellular apoptosis; they are consistent with our findings that p.A126G exhibits a decreased ability to reduce PI(3,4,5)P₃ relative to WT (Fig. 2B). Importantly, we note that treatment with isoform-specific inhibitors of PI3K, predicted to antagonize loss of PTEN-mediated dephosphorylation of PI(3,4,5)P₃, are capable of restoring Akt and S6 phosphorylation to near WT levels (21). Further, to ensure no bias in characterizing the effect of the p.A126G mutation on global phosphorylation levels, we performed mass spectrometry on purified phosphopeptides of PC-3 cells transiently expressing catalytic dead (p.C124S) and mutant (p.A126G) PTEN (SI Appendix, Fig. S4 A and B). We then performed differential phosphoprotein abundance analysis on p.A126G-expressing cells, using catalytic dead (p.C124S)-expressing cells as the background. We identified a total of 601 enriched and 497 depleted (at least twofold) phosphopeptides relative to the catalytic dead background (SI Appendix, Fig. S4C). We note that, of the two PH-domain–containing PIP₃-regulated proteins expressed in PC-3 cells (Akt and PDK1), only phospho-Akt is seen above twofold enrichment. This finding is consistent with the above Akt Western blots demonstrating aberrant phospho-Akt expression in mutant cells. SI Appendix, Table S7 lists PH-domain–containing phosphopeptides enriched or depleted greater than twofold relative to p.C124S. To identify common processes and pathways that are enriched within the total pool of identified phosphopeptides, we performed GO-enrichment annotation clustering (SI Appendix, Tables S5 and S6). SI Appendix, Fig. S4D provides a selected subset of the most enriched statistically significant (P ≤ 0.05) annotation clusters in p.A126G-expressing cells. We found enrichment of PTEN-regulated downstream pathways and processes, including the cancer-associated mTOR pathway and groupings associated with PI3K/Akt pathway hyperactivation, such as transcriptional and translational activation. Intriguingly, we also found enrichment of processes affiliated with polarized cell migration and projection.

Although PI3K/Akt pathway hyperactivation is also seen in PTEN loss-of-function variants, the up-regulation of PI(3,4)P₂ that we observed for PTEN p.A126G (Fig. 2C and SI Appendix, Fig. S4D) is unique to this mutation. Interestingly, increased PI(3,4)P₂ levels are thought to promote increased cell migration (22–24). Given these findings and our mass spectrometry data showing enrichment of cell migration processes, we hypothesized that the increased PI(3,4)P₂ levels in p.A126G would cause increased cell migration relative to PTEN loss-of-function mutations. To test this hypothesis, we performed scratch wound healing assays on confluent monolayers of cells solely expressing the various PTEN proteins (Fig. 3G). We demonstrate that, compared with cells expressing catalytic dead (p.C124S) PTEN, p.A126G PTEN-expressing cells show increased cell migration after 24 h and sensitivity to intervention with PI3K inhibitors (Fig. 3F) (P ≤ 0.05). To rule out differential cell proliferation rates confounding these results, we performed cell viability, proliferation, and clonogenic cell survival assays across WT, catalytic dead (p.C124S), and mutant (p.A126G) PTEN. We found no statistically significant difference in proliferation between the alleles up to 96 h posttransfection (SI Appendix, Figs. S5–S8) although we did note a trend of higher total cell number and colony formation capacity in the mutant (p.A126G) PTEN-expressing cells. These results suggest that p.A126G PTEN expression in cells causes an oncogenic phenotype that sensitizes the cells to PI3K inhibition.

We extracted germ-line DNA from saliva and somatic DNA from a fresh frozen prostate tumor biopsy using a Qiagen DNeasy Blood and Tissue DNA Purification kit. Each sample was sequenced on an individual Illumina HiSEq. 2000 lane producing 100-bp paired end reads. We assessed each sequencing run for multiple quality metrics, including number of reads, read quality, and reads mapping to the human genome (SI Appendix, Table S1). Sequencing reads were mapped to the human reference genome (hg19) using the BWA mapping algorithm (v 0.6.2) (27). Reads were dynamically trimmed using the “-q” flag to eliminate any nucleotides with Phred quality scores less than 20. Mapped reads were filtered to include only properly paired reads. Tumor and normal mapped pair reads were duplicate-marked via Picard tools (v 1.96) (broadinstitute.github.io/picard/) and underwent joint INDEL realignment and base quality score recalibration using the Genome Analysis Tool Kit (v 2.7–2-g6bda569) (28).

Germ-line SNV and INDEL variant calling was performed using “Unified Genotyper” via the Genome Analysis Tool Kit (v 2.7–2-g6bda569) (28). Somatic SNVs were called using MuTect (v 1.1.4) (29). Somatic INDELs were called using Indelocator (v IndelGenotyper.36.3336), which is within the Genome Analysis Tool Kit (28). Somatic SVs were called using Breakdancer (v 1.2) (30). Somatic CNV calling was done using VarScan (v 2.3.2) (31). All tools were run using the best practices outlined in their respective manuals. Germ-line and somatic variants were annotated for a series of functional predictions, conservation scores, in addition to publically available database annotations using a combination of perl scripts and ANNOVAR (32).

PC-3 cells were purchased from the American Type Culture Collection and maintained in RPMI 1640 with 10% (vol/vol) FBS, and CHO cells were grown as previously described (17). PI3K-p110α inhibitor BYL719 (Selleckchem) and PI3K-p110β inhibitor AZD6428 (Selleckchem) were dissolved in DMSO. PTEN WT (Addgene plasmid 10750), p.C124S catalytic dead (Addgene plasmid 10744), and p.A126G mutant expression vectors were used for transient transfections into the PC-3 cells. The PTEN p.A126G vector was created by site-directed mutagenesis using the PTEN WT plasmid and the following primers: forward primer, 5′-CTGTAAAGGTGGAAAGGGACGAACTG-3′; reverse primer, 5′-CCTTTCCACCTTTACAGTGAATTGCTG-3′. Construction of PTEN_CiV chimera (previously named Ci-VSPTEN16) WT was described earlier (19). p.C124S catalytic dead, p.A126G, p.A126P, p.A126V, and p.A126S mutant plasmids were created by site-directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies). All sequences were verified before use (Seqlab Sequence Laboratories).

Expression vectors used for transfection of CHO cells were as follows: PLCδ₁-PH (UniProt accession number P51178) or Btk-PH (Q06187) in pEGFP-N1 vector; TAPP1 (Q9HB21) in FUGW vector (contains eGFP); bovine phosphatidylinositol 3-kinase p110α (constitutively active mutant K227E; P32871) (33) in pcDNA3; and PTEN_CiV chimera and mutants in mRFP-C vector.

Twenty-four hours before transfection, PC-3 cells were plated at a density of 1.5 × 10⁵ cells per well in six-well plates. The PC-3 cells were then transfected with 1.5 μg of vector DNA complexed with 5 μL of Lipofectamine 3000 (Life Technologies) in accordance with the manufacturer’s Lipofectamine 3000 protocol. CHO cells were plated onto glass bottom dishes (WillCo Wells B. V.) or coverslips 24–48 h before transfection, which was performed using jetPEI (Polyplus Transfection) in accordance with the manufacturer’s instructions. All experiments on CHO cells were performed 24–48 h after transfection.

For total internal reflection fluorescence (TIRF) experiments, the whole cell configuration of the patch clamp technique was used using an EPC-10 amplifier controlled by PatchMaster software (HEKA Elektronik). CHO cells were clamped to −60 mV, and chimeras were activated by depolarizing the cell to 80 mV for 30 s. In these experiments, series resistance (R_s) typically was below 5 MΩ, and no R_s compensation was applied. Patch pipettes were pulled from borosilicate glass using a P2000 puller (Sutter Instrument Company), and they had an open pipette resistance between 2 and 4 MΩ after back-filling with intracellular solution containing 135 mM KCl, 2.41 mM CaCl₂, 3.5 mM MgCl₂, 5 mM EGTA, 5 mM Hepes, and 2.5 mM Na₂-ATP, pH 7.3 (with KOH), 290–295 mosm/kg. Throughout the recordings, cells were kept in extracellular solution containing 144 mM NaCl, 5.8 mM KCl, 0.9 mM MgCl₂, 1.3 mM CaCl₂, 0.7 mM NaH₂PO₄, 5.6 mM d-glucose, 10 mM Hepes, pH 7.4 (with NaOH), 305–310 mosm/kg. All experiments were performed at room temperature (22–25 °C).

TIRF imaging was done as described before (17). Briefly, a BX51WI upright microscope (Olympus) equipped with a TIRF-condenser (N.A. 1.45; Olympus) and a 488-nm laser (20 mW; Picarro) was used. Fluorescence was imaged through a LUMPlanFI/IR 40×/0.8 N.A. water-immersion objective. Images were acquired with a TILL-Imago QE cooled CCD camera (TILL Photonics) in combination with a Polychrome IV light source (TILL Photonics) controlled by TILLvision software (TILL Photonics). The frame interval was 3 s, and the laser shutter was controlled by the Polychrome IV. The electrophysiology setup was synchronized to the imaging setup. Cells were cotransfected with plasmids coding for an mRFP-tagged phosphatase, a PI sensor domain, and, for experiments involving sensors for PI(3,4)P₂ or PI(3,4,5)P₃, additionally p110α(K227E). Experiments were done 24–48 h posttransfection on cells selected for expression of the mRFP tag. Cells under investigation were whole cell patch clamped as described above. Experiments were carried out at room temperature (22–25 °C). Imaging data were analyzed using TILLvision (Till Photonics) and IgorPro (Wavemetrics). Regions of interest (ROIs) encompassed the footprint of a single cell, excluding cell margins to avoid movement artifacts. F/F₀ traces were calculated from the background-corrected TIRF signal intensity F, normalized pixelwise to the initial intensity F₀, which was calculated as the average over the baseline interval, by averaging over the ROI. F/F₀ traces were corrected for bleaching according to monoexponential fits to the baseline interval as described previously (19).

Transfected PC-3 cells were seeded at a density of 2 × 10⁶ cells per 15-cm dish for PI(3,4,5)P₃ and PI(3,4)P₂ extractions, respectively. At 80% confluence, the media was aspirated, and 10 mL of ice cold 0.5 M TCA was immediately added. After a 5-min incubation on ice, cells were scraped, transferred to a 15-mL centrifuge tube, and spun at 300 × g for 5 min. The pellet was resuspended in 3 mL of 5% trichloroacetic acid/1 mM EDTA and centrifuged at 300 × g for 5 min, and the supernatant was discarded. Neutral lipids were extracted by adding 3 mL of methanol:chloroform (2:1), vortexing three times over 10 min at room temperature, centrifuging at 300 × g for 5 min, and discarding the supernatant. Acidic lipids were extracted by adding 2.25 mL of methanol:chloroform:12M hydrochloric acid (80:40:1) to the pellet, resuspending, vortexing four times over 15 min at room temperature, centrifuging at 300 × g for 5 min, and transferring the supernatant to a new 15-mL centrifuge tube. Then 0.75 mL of methanol and 1.35 mL of 0.1 M hydrochloric acid were added to the supernatant, vortexed, centrifuged at 300 × g for 5 min to separate organic and aqueous phases. The organic (lower) phase was collected into a new vial and dried in a vacuum dryer for 1 h. Samples were suspended in 65 mL of PBS-Tween+3% Protein Stabilizer (provided by the Echelon kit), vortexed, and spun down before adding to the ELISA. PI(3,4,5)P₃ and PI(3,4)P₂ phospholipid quantification was performed according to the manufacturer’s instructions respective to each kit (K-2500s and K-3800; Echelon Biosciences).

Whole cell protein lysates were extracted from PC-3 cells 24 h after transfection. Six micrograms of whole cell lysates were loaded into polyacrylamide gels and subjected to SDS/PAGE electrophoresis. Nitrocellulose membranes were incubated with anti-PTEN, anti-S6, anti-Phospho-S6, anti-Akt, or anti-Phospho-Akt (nos. 9559, 2217, 2215, 4691, and 4060, respectively; Cell Signaling Technology). All antibodies were diluted 1:1,200 in Tris-buffered saline and Tween 20 with 5% BSA. Goat anti-rabbit IgG (H+L) labeled with IRDye 680RD conjugate (Licor) was used as a secondary antibody. Bands were detected using a Licor ODYSSEY fluorescent imaging system.

Before seeding with cells, auxiliary lines used as reference points for imaging the wound were drawn on the underside of each well of a six-well plate. Plates were seeded at a density of 1.5 × 10⁵ cells per well and allowed to grow for 48 h. Cells were transfected, and a linear wound was applied to the monolayer using a 200-µL pipette tip 24 h later. Loose cells were removed with three washes of PBS. Drugs, if any, were added to the cells, and images were captured immediately after wounding (T₀) and subsequently in 12-h intervals (T₁₂, and T₂₄). TScratch was used to calculate percentage of wound closure (34).

Cells were transfected with PTEN WT, catalytic dead (p.C124S), or mutant (p.A126G) plasmids. Six-well plates were seeded with PC-3 cells at a density of 5 × 10⁴ cells per well and allowed to grow for 96 h. Presto blue (Life Technologies) was added to each well after the 96-h posttransfection incubation. After a 2-h incubation with presto blue, dye absorbance was measured on a Tecan Infinite 200 PRO plate reader at 570 nm and a reference wavelength of 600 nm. Measured absorbance values were normalized by absorbance at the reference wavelength.

Cells were transfected with PTEN WT, catalytic dead (p.C124S), or mutant (p.A126G) plasmids. Six-well plates were seeded with PC-3 cells at a density of 5 × 10⁴ cells per well and allowed to grow for 96 h. Then 1 mL of 0.05% Trypsin-EDTA was added for 5 min to detach cells, followed by the addition of 3 mL of media to resuspend the detached cells. Total cell number was measured using a Scepter 2.0 Handheld Automated Cell Counter (EMD Millipore).

Cells were transfected with PTEN WT, catalytic dead (p.C124S), or mutant (p.A126G) plasmids. After a 96-h posttransfection incubation period, six-well plates were seeded with PC-3 cells at a density of 200 cells per well and allowed to grow for 21 d. Media was then gently aspirated, 3 mL of methylene blue was added for 30 min, the methylene blue was then gently aspirated, and the six-well plates were allowed to dry at room temperature for 30 min. Individual colonies were then counted using a Fischerbrand Counter Pen (Fischer Scientific).

All statistical analyses were performed in the R programming language. Statistical significance was determined by the Student’s t test for comparisons between groups and one-way ANOVA followed by a post hoc Tukey's honest significant difference test for pairwise comparisons across all groups.