Multi-model functionalization of disease-associated PTEN missense mutations identifies multiple molecular mechanisms underlying protein dysfunction

Abstract

Functional variomics provides the foundation for personalized medicine by linking genetic variation to disease expression, outcome and treatment, yet its utility is dependent on appropriate assays to evaluate mutation impact on protein function. To fully assess the effects of 106 missense and nonsense variants of PTEN associated with autism spectrum disorder, somatic cancer and PTEN hamartoma syndrome (PHTS), we take a deep phenotypic profiling approach using 18 assays in 5 model systems spanning diverse cellular environments ranging from molecular function to neuronal morphogenesis and behavior. Variants inducing instability occur across the protein, resulting in partial-to-complete loss-of-function (LoF), which is well correlated across models. However, assays are selectively sensitive to variants located in substrate binding and catalytic domains, which exhibit complete LoF or dominant negativity independent of effects on stability. Our results indicate that full characterization of variant impact requires assays sensitive to instability and a range of protein functions.

Fig. 1. Functional assessment of PTEN variants for genetic interactions in yeast.

a Schematic of 106 MS and NS variants indicating their positions across the functional domains of PTEN with each variant represented by a circle colored following the coding table at right. b Venn diagram showing the overlap of PTEN variants used in this study identified in individuals with different disorders. The ASD category includes DD; and Somatic Cancer includes variants with 8 or more reports in COSMIC. c Overexpressing WT-PTEN in a library of 4699 yeast strains, each lacking one nonessential gene, identified 8 sentinel strains showing significant genetic interaction (five of which function in phospholipid metabolism; three in vacuole fusion). WT PTEN = gray, E.V. = empty vector, green. Data are expressed as means ± SEM (n = 16 for all sentinels). Two-tailed Student’s t test for comparison between WT-PTEN and E.V. found p < 0.05 for all eight sentinel strains. d Schematic of sentinel functions in PIP3 metabolism. e 99 non-tagged PTEN variants were expressed in the sentinel yeast strain Δvac14 and assayed for colony size. Data are expressed as means ± SEM, and variants are colored based on their categories following the same colr table in a. * indicates nominal p < 0.05 by two-tailed Satterthwaite approximation for the contrast between WT and variant in the mixed-effects model (see Methods). Individual variant Means, n and nominal p-values for all plots are provided in the Source Data file.

Fig. 2. Functional assessment of PTEN variants in Drosophila development.

a Schematic of Drosophila life cycle highlighting eclosion transition form pupa to adult. b Transgenic Drosophila overexpressing WT-PTEN (black triangles) exhibited delayed development, indicated by increased time to reach eclosion compared to empty vector (EV, green squares), and the catalytically inactive variant C124S (blue circles). Data are expressed as means ± SEM. *p < 0.05 for WT vs. C124S, #p < 0.05 for WT vs. EV by two-tailed Student’s t test. c Effect on time to eclosion for 88 strains of transgenic flies each expressing a different non-tagged PTEN variant, with variants colored by category following the color table below. Data are expressed as means ± SEM. *p < 0.05 for the contrast between WT and variant, and #p < 0.05 for the contrast between EV and variant by two-tailed Satterthwaite approximation in the mixed-effects model (Methods section). d Correlation of genetic interaction values from yeast sentinel Δvac14 and developmental delay in Drosophila for 84 PTEN variants. Variants are colored following the same color table used in c. Data are expressed as means ± SEM, Pearson r  = 0.69, p < 0.0001. Individual variant means, error, n and nominal p-values for all plots are provided in the Source Data file.

Fig. 3. PTEN variant impact on rat neural growth and sensorimotor behavior in C. elegans.

a Representative immunofluorescence images of proximal dendrites from primary cultured hippocampal pyramidal neurons co-stained for the postsynaptic markers of excitatory (PSD-95) and inhibitory (Gephyrin) synapses. Similar results were seen in 128 replicates of WT PTEN, 109 replicates for GFP Control, and 75 replicates for C124S. Scale bar = 5 μm. b Representative images of cultured hippocampal and DRG neurons expressing GFP alone, or with WT-PTEN and PTEN-C124S (scale bar = 100 μm for Hippocampus, and 60 μm for DRG). Similar results were seen in replicates for WT PTEN (132 hippocampus, 463 DRG), GFP Control (125 hippocampus, 488 DRG), and C124S (84 hippocampus, 444 DRG). Effect of overexpression of 20 3XHA-tagged PTEN variants on excitatory synapses (c), soma size (d), total dendritic arbor length (e) in hippocampal neurons, and axonal length in DRG neurons (f). Data are expressed as means ± SEM. *p < 0.05 by two-tailed Satterthwaite approximation as in Fig. 1 (Methods section). g Wild type C. elegans (cePTEN) exhibit chemotaxis towards a NaCl source, while worms harboring a NS mutation in the worm homolog daf18 (cePTEN(rf)) exhibit chemotaxis deficits, which is rescued by overexpression of human PTEN (cePTEN(rf)+hPTEN). h Impact of 20 non-tagged PTEN variants on NaCl chemotaxis in C. elegans is shown by scoring salt preference as (A–B)/(A+B). Data are expressed as means ± SEM and are normalized to cePTEN(rf) + hPTEN = 1 and cePTEN(rf) = 0. Variant data represented in all histograms are color-coded by variant category depicted in the color table. *p < 0.05 compared to WT, #p < 0.05 compared to cePTEN(rf) by two-tailed Satterthwaite approximation. Individual variant means, error, n and nominal p-values for all plots are provided in the Source Data file.

Fig. 4. PTEN variants destabilize PTEN.

a The abundance of 97 PTEN variants overexpressed in yeast was assayed by western blot. Variants are color-coded by category depicted in the color table at the bottom right. Quantified values by band densitometry were normalized to WT-PTEN = 1 and empty vector = 0. Data is expressed as mean ± SEM. *p < 0.05 compared to WT by two-tailed Satterthwaite approximation (Methods section). b Representative scatter plot of single-cell fluorescence intensities from flow cytometry for overexpressed sfGFP- fusions of PTEN WT (light purple), D326N (pink), and H93Q (dark purple) versus their level of transfection as visualized with mTagRFP-t. c Histogram showing the relative frequency of sfGFP/mTagRFP-t ratio for the same variants in b. d Protein stability of 105 PTEN variants assayed in HEK293 cells by flow cytometry calculated as median of sfGFP/mTagRFPt and expressed as normalized to WT = 1. Data is expressed as mean of well replicates ±SEM, and variants are color-coded by category depicted in the color table at the bottom right. *p < 0.05 compared to WT by two-tailed Student’s t test. e Plot of yeast abundance and HEK293 PTEN stability variant data sets with color coding as in a and d. Data are expressed as means ± SEM, Pearson r = 0.55, p < 0.0001. Individual variant means, error, n and nominal p-values for all plots are provided in the Source Data file.

Fig. 5. PTEN variant impact on pAKT/AKT reveals LoF and dominant negative activity.

a, b Flow cytometry single-cell rolling median of pAKT/AKT ratios versus expression levels of sfGFP-PTEN in HEK293 cells for WT (light purple), C124S (pink) and 4A (dark purple) variants and the corresponding frequency histogram in b. Vertical bars indicate median values. c Functional impact of 105 sfGFP-tagged PTEN MS variants on pAKT/AKT levels. Data plotted are median differences between transfected and in-well untransfected cells for each variant, with variants color-coded by category as depicted in the color table at top right. Data are expressed as mean of well replicates ±SEM. Values are normalized to WT = 1 and no difference to untransfected = 0. *p < 0.05 compared to WT, #p < 0.05 compared to 0 by two-tailed Student’s t test. d Relative pAKT/AKT changes are shown for variants exhibiting dominant negative effects on pAKT/AKT, as well as the constitutively active 4A in both parental (black) and a PTEN-KO (green) HEK293 cell line. Data are expressed as mean of well replicates ±SEM. Values are normalized to WT = 1 and no difference to untransfected = 0. *p < 0.05 comparing each variant in parental and a PTEN-KO HEK293 cell line by two-tailed Student’s t test. e, f Plots of PTEN variant function by pAKT/AKT changes in HEK293 cells versus variant function of genetic interaction with ΔVac14 in yeast (Pearson r = 0.79, p < 0.0001), and rate of eclosion in Drosophila (Pearson r = 0.78, p < 0.0001). Data are expressed as means ± SEM with variants color-coded as in c. Individual variant means, error, n and nominal p-values for all plots are provided in the Source Data file.

Fig. 6. Functional and stability data identifies distinct mechanisms of molecular dysfunction.

a A normalized plot displaying stability values subtracted from function values for each assay with variants displayed according to their amino acid position below a schematic of PTEN structural domains. Normalized function – normalized stability scores are depicted as a heat map in which a score of 0 (white) indicates variants whose function matches their stability. A positive score (orange) indicates higher function than predicted from instability, while negative scores (magenta) indicates greater dysfunctional than instability (color scale at top right). b Variant function-stability scores for pAKT/AKT assay in HEK293 cells plotted against amino acid position with variants separated into N-terminus, WPD-loop and P-loop domains (red), and variants outside these domains (blue). Data are expressed as means ± SEM. c Frequency distribution of normalized function - stability scores for pAKT/AKT assay in HEK293 assay showing two distinct populations, with red and blue bars denote variants as in b. ***p < 0.0001 by two-tailed Student’s t test. d HEK293 pAKT/AKT assay data plotting PTEN variant function vs. stability with data separated by variants as in b: functional dysfunction of variants that is explained by stability (blue), and not (red). e High correlations are found for variant impact on multiple assays for variants in which dysfunction is associated with instability (blue variants in b–d). Color scale at right. f Weaker cross-assay correlations are seen for variants in catalytic domains, which exhibit greater dysfunction then explained by instability (red variants in b–d). Color scale at right. Individual variant means, error, n and nominal p-values for plot are provided in the Source Data file.

Fig. 7. Severity of variant dysfunction correlate with occurrence in cancer.

Distribution of normalized PTEN variant function by its frequency within the COSMIC database, with variants with a frequency of 0 in purple, 1–7 reports in green, and 8 or more reports in tan. Mean PTEN function is plotted as bars ±SEM. *p < 0.05, **p < 0.005, ***p < 0.0005 by two-way ANOVA. Individual variant means, error, n and nominal p-values for plot are provided in the Source Data file.

Fig. 8. PTEN variant classification based on 9 assays of variant functions.

a The results of 9 functional assays for 106 PTEN variants are summarized according to the following criteria: WT-like variants exhibit no differences to WT, GoF exhibit function values significantly greater than WT, Partial LoF exhibit function significantly greater than null but significantly less than WT, Complete LoF exhibit no differences to null, dominant negative exhibit function significantly less than both null and WT (Below LoF for fly and yeast assays). Classification are based on statistically significant differences of at least nominal p < 0.05 compared to the WT-PTEN, null or both using the statistical method of each individual assay. White cells indicate that the variant was not tested in that assay. Color-coding defined in color table at right. b PTEN variant predicted impact based on a binary notation of either LoF (<50% of WT effect) or WT-like (≥50% of WT effect) and classified accordingly based on their frequency of LoF (pathogenic) or WT (likely benign) in 9 assays. Color-coding defined in color table at right. Source data are provided as a Source Data file.