PI3Kα-regulated gelsolin activity is a critical determinant of cardiac cytoskeletal remodeling and heart disease.

Biomechanical stress and cytoskeletal remodeling are key determinants of cellular homeostasis and tissue responses to mechanical stimuli and injury. Here we document the increased activity of gelsolin, an actin filament severing and capping protein, in failing human hearts. Deletion of gelsolin prevents biomechanical stress-induced adverse cytoskeletal remodeling and heart failure in mice. We show that phosphatidylinositol (3,4,5)-triphosphate (PIP3) lipid suppresses gelsolin actin-severing and capping activities. Accordingly, loss of PI3Kα, the key PIP3-producing enzyme in the heart, increases gelsolin-mediated actin-severing activities in the myocardium in vivo, resulting in dilated cardiomyopathy in response to pressure-overload. Mechanical stretching of adult PI3Kα-deficient cardiomyocytes disrupts the actin cytoskeleton, which is prevented by reconstituting cells with PIP3. The actin severing and capping activities of recombinant gelsolin are effectively suppressed by PIP3. Our data identify the role of gelsolin-driven cytoskeletal remodeling in heart failure in which PI3Kα/PIP3 act as negative regulators of gelsolin activity.


Supplementary Methods
Transverse aortic constriction. 8 -8.5 weeks old male WT, PI3KαDN, PI3Kα flox/flox , PI3Kα flox/flox/Cre , Gsn -/-(GSNKO) and PI3KαDN/GSNKO mice weighing 20-25 g were used in these experiments as previously described [1][2][3][4][5] . Mice were anesthetized with isoflurane (#CP0406V2, Fresenius Kabi Canada). The skin was cleaned with Germex and Betadine. One dose of penicillinstreptomycin (10 mg kg -1 , 0.1 mL i.p.; #LS15140122, Fisher Scientific) was administered prior to start of surgery. Mice were placed supine and body temperature was maintained at 37°C with a heating pad. A horizontal skin incision of 1 cm in length was made at the level of second intercostal space, once the animal was in surgical plane of anesthesia (lack of reflex or response to toepinching). A 6-0 silk suture was passed under the aortic arch, a bent 27-gauge (27G) needle was then placed next to the aortic arch and the suture was snugly tied around the needle and aorta between the left carotid artery and the brachiocephalic trunk. The needle was quickly removed allowing the suture to constrict the aorta. The incision was closed in layers and the mice were allowed to recover on a warming pad until they were fully awake. Immediately after the surgery, mice received one dose of buprenorphine (#258-396-8, Millipore Sigma). Sham animals underwent the same procedure without the aortic banding. Measures were taken to avoid any infection to the mice.
Echocardiographic imaging and pressure-volume loop analysis. Transthoracic M-mode and Doppler echocardiographic examination at 2 weeks-and 9 weeks post-TAC were performed using the Vevo 770 high-resolution imaging system equipped with a 30-MHz transducer (#RMV-707B; VisualSonics, Toronto, Canada) as previously described [6][7][8] . Mice were placed on a heating pad and a nose cone with 0.75-1% isoflurane in 100% oxygen was applied. The temperature was maintained at 36.5 to 37.5°C. Ultrasound gel was placed on the chest of the anesthetized mouse. The ultrasound probe was placed in contact with the ultrasound gel and scanning was performed over 30 min. The temperature and heart rate (HR) were constantly monitored during the scanning. M-mode images were obtained for measurements of left ventricular (LV) wall thickness (LVWT), LVEDD, and LVESD. M-mode images were used to measure LV chamber size and wall thickness. LVEF, LVFS and VCFc were also calculated.
To measure LV pressure-volume relationship, 1.2F admittance catheter (Scisense Inc.) was used as previously described 9,10 . Briefly, mice were anesthetized with isoflurane, and an incision was made in the right carotid artery. A catheter was inserted into the incision and was advanced to the LV through the ascending aorta and aortic valve. The position of the catheter was monitored by pressure along with the magnitude and phase using ADvantage pressurevolume system (Scisense Inc.) and iworx data acquisition system (iWorx Systems Inc.) connected to the catheter. Once, the desired range for magnitude and phase was achieved, a baseline scan was performed to derive volume using Baan's equation, and the pressure-volume loop was obtained using the LabScribe2 software (iWorx Systems Inc., version 2.347000). The inferior vena cava was briefly occluded to obtain alterations in venous returns to derive end-diastolic pressurevolume relationships. The pressure-volume loops were analyzed using the LabScribe2 software.
Histological analyses, wheat germ agglutinin and F-and G-actin staining. Hearts were arrested in diastole with 1M KCl, fixed with 10% buffered formalin, and embedded in paraffin. Tenmicrometer thick sections were stained with picro-sirius red (PSR) or Masson trichrome to assess myocardial fibrosis and were visualized using fluorescence microscopy (Olympus IX81) and light microscopy (DM4000 B, Leica), respectively, as we have previously described 7,11 . Myocardial collagen contents were measured from the picrosirius red staining images using Metamorph Basic (Molecular Devices Inc., version 7.7.0.0) software.
Five-micrometer thick OCT embedded cryosections were also stained with phalloidin and WGA to visualize cytoskeletal actin filaments (F-actin). Briefly, the cryosections were fixed in 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 in DPBS and incubated with mixture of Alexa Fluor 488 conjugated phallodin (#A12379, ThermoFisher) and tetramethylrhodamine-WGA (#W849, ThermoFisher) in 1.5% BSA for 20 minutes at room temperature. The sections were then mounted using Prolong gold antifade mounting medium with DAPI to counterstain nuclei and visualized under the fluorescence microscope (Olympus IX81). The sections were washed three times with DPBS in between each step. Separately, the cryosections were double-stained with Alexa Fluor 488 conjugated phalloidin to stain F-actin and Alexa Fluor 594 conjugated DNAse I (#D12372, ThermoFisher) to stain the monomeric globular actin (G-actin) and visualized and imaged using fluorescence microscope (Olympus IX81). The images were morphometrically quantified using Metamorph Basic (Version 7.7.0.0) and F-actin to G-actin ratio was represented as an index of actin polymerization and compared between each group.
Immunoprecipitation and immunoblot analysis of Gelsolin and p110α. Human and murine heart tissues (400 mg) were lysed in 400 μl of CelLytic™ M lysis buffer (#C2978, Millipore Sigma) containing 1x cOmplete Protease (#11697498001, Sigma-Aldrich) and PhosSTOP Phosphatase (#4906845001, Millipore Sigma) inhibitor cocktails. Protein concentration of the whole tissue lysates was measured using Bradford protein assay kit I (#5000001, Bio-Rad Canada). A total of 50 μl of the tissue lysates was saved. The same amount of total protein was applied to DynaMag TM Trial Magnet beads (ThermoFisher) pre-incubated with a Goat polyclonal anti-Gelsolin antibody (#sc-6406, Santa Cruz) or a Rabbit monoclonal anti-P110α antibody (#MA5-14870, ThermoFisher) to immunoprecipitate Gelsolin and P110α, respectively. Immunoprecipitated samples were washed three times with PBS buffer containing 0.01% Triton X-100. The immunoprecipitated proteins were then eluted from the beads by addition of 4× SDS-PAGE sample buffer (#1610747, Bio-Rad Canada) containing 10% β-mercaptoethanol. The eluted samples and the same amount of the whole tissue lysates were subjected to SDS-PAGE and transferred to immobilon @ P PVDF membranes (#IPVH00010, Millipore Sigma) by electroblotting. Immunoblotting was performed using rabbit anti-Gelsolin polyclonal antibody (#sc-48769, Santa Cruz), and goat anti-P110α polyclonal antibody (#sc-1331, Santa Cruz) to detect immunoprecipitated Gelsolin and P110α, respectively. Antibody binding was detected using antigoat or anti-rabbit IgG (HRP).
TaqMan Real-time PCR. RNA expression levels of various genes were determined by TaqMan Real-time PCR as previously described 1,5,6 . Total RNA was extracted from flash-frozen tissue using TRIzol extraction protocol, and cDNA was synthesized from 1 μg RNA by using random hexamers. For each gene, a standard curve was generated using known concentrations of cDNA (0.625, 1.25, 2.5, 5, 10 and 20 μg) as a function of cycle threshold (CT). Expression analysis of the reported genes was performed by TaqMan Real-time PCR using ABI 7700 Sequence Detection System (Conquer Scientific). The SDS2.2 software (integral to ABI7700 real-time machine) fits the CT values for the experimental samples and generates values for cDNA levels. All samples were run in triplicates in 384 well plates. 18S rRNA was used as an endogenous control.
Isolated cardiomyocyte contractility. For measurement of contractility, cardiomyocytes were isolated as described previously 5 . After isolation myocytes were kept in perfusion buffer solution (pH 7.4). An aliquot of isolated cardiac myocytes were transferred in a glass-bottomed recording chamber on top of inverted microscope (Olympus IX71) and allowed to settle for 5-6 min. Cells were superfused at a rate of 1.5-2 mL min -1 with modified Tyrode's solution (containing in mmol L -1 : 135 NaCl, 5.4 KCl, 1.2 CaCl 2 , 1 MgCl 2 , 1 NaH 2 PO 4 , 10 Taurine, 10 HEPES, 10 glucose; pH 7.4 with NaOH; all purchased from Millipore Sigma). The perfusion solutions were heated to inbath temperature of 35-36°C using in-line heater (SH-27B, Harvard Apparatus) controlled by automatic temperature controller (TC-324B, Harvard Apparatus). Quiescent rod-shaped myocytes with clear striations were selected for study. Platinum-wire electrodes were placed near the cell just outside of the microscope view at 400X magnification. Myocytes were paced at 1 Hz with voltage of 3-4 V (50% above threshold) and pulse duration of 2.5 ms using S48 stimulator (Grass Technology). Sarcomere length was estimated in real time from images captured at a rate 200 frame s -1 via 40X objective (UAPO 40X3/340, Olympus) using high-speed camera (IMPERX IPX-VGA-210, Aurora Scientific). Sarcomere length was calculated by HVSL software v. 1.75 (Aurora Scientific) using auto-correlation function (ACF/Sine-fit) algorithm and stored in a file for analysis at a later time. Myocytes were paced for at least 2 min. Only recordings of myocytes that produced stable contractions of similar amplitude and kinetics at a steady-state (past 2 min) were selected for analysis. At about 2 min of stimulation time, 10 consecutive contractions were selected and averaged to reduce noise and make calculations of derivatives more precise. Averaged contraction was used to calculate fractional shortening (FS), relaxation times (t50 and t90), and ±dL dt -1 . Calculations were performed in Origin 8.5 (OriginLab) using custom-made script of built-in LabTalk language.

Measurement of L-type Ca 2+ current (I Ca,L ).
Aliquots of solution containing murine cardiomyocytes were placed in a bath mounted on top of an inverted microscope (Olympus IX71, Olympus, Canada), and rod-shaped quiescent myocytes were selected for study. The myocytes were superfused with K + -free modified Tyrode's solution (containing in mmol L -1 : NaCl, 135; CaCl 2 , 1.2; MgCl 2 , 1; NaH 2 PO 4 , 1; HEPES, 10; taurine, 10; glucose, 10; all purchased from Millipore Sigma) at 35-36°C. Pipettes with resistance of 1.5-2.5 MOhm were filled with Cs + pipette solution (containing in mmol L -1 : CsOH, 110; CsCl, 30; L-Aspartic acid, 110; MgATP, 5; EGTA, 5; HEPES, 10; all purchased from Millipore Sigma). Currents were recorded from whole cell in the ruptured patch configuration using Multiclamp 700B amplifier (Molecular Devices, USA) in voltage-clamp mode. Signal was digitized by 16-bit analog-digital board DigiData 1440A (Molecular Devices, USA) under control of pClamp 10 software (Molecular Devices, USA) and stored for offline analysis. Ca 2+ current (I Ca,L ) was elicited by step depolarizations from holding potential of -40 mV to the test potential 0 mV for 200 ms. Peak I Ca,L was defined as the difference between peak inward current and the steady state current at the end of the test pulse.

In Silico Modeling of gelsolin and its individual N-and C-terminus domain.
Human gelsolin has a solved X-ray crystal structure (PDB ID: 3FFN) 12 , but several residues (1-49, 260-264, 756-782) were missing in the crystal structure. Comparative homology modeling was used to construct the full length structure of human gelsolin using this crystal structure (PDB ID: 3FFN) as a template. The crystal structure of calcium activated human gelsolin N-terminus complexed with actin (PDB ID: 3FFK) 12 also contains missing residues. We have used this structure as a reference to build the full length model of calcium activated human gelsolin N-terminus bound with actin. Human gelsolin C-terminus has its both calcium bound (PDB ID: 1P8X) 13 and actin bound structure (PDB ID: 1H1V) 14 with several missing residues. So, we have built full length model for all of them. Models were built using Modeller 9.14 (http://salilab.org/modeller/) 15,16 .

Docking of PIP2 and PIP3 with N-and C-terminus domain.
To model the gelsolin N-terminus-PIP2 complexes, we used the knowledge-based Autodock Vina docking method 18 . The grid for docking was developed using structurally aligned Horse N-terminus-ATP bound structure (PDB ID: 2FGH) 19 with human gelsolin N-terminus. Using the built-in Grid map option, we have prepared the axes dimensions and center points for performing the docking of PIP2 with Human gelsolin N-terminus. For modeling the human gelsolin C-terminus bound with PIP3 we have used the active site grid dimension of N-terminus bound PIP2. Structural alignment between the N-terminus and C-terminus domains of human gelsolin revealed that the two domains are homologous to each other. The N-terminus G1-G3 sub domains resembles structurally with the G4-G6 of Cterminus. Keeping this knowledge in mind, we have used the binding grid of PIP2 with N-and Cterminus as same as for PIP3. In each docking experiment, we have used 10 conformations of the ligand. After selecting the conformations of the ligand that best fit with human gelsolin Nterminus and C-terminus domain. We have refined the complex based on the active site residues using COOT 20 . We optimized the substrate residues by using rotamer selection and regularize zone tools.
MD simulation. All molecular dynamics (MD) simulations of the PIP2 and PIP3 bound gelsolin complexes were performed using GROMACS 5.1.2 software package 21 , utilizing the GROMOS96 43a1 force field. The starting structures for the 50 ns simulations of the PIP2 and PIP3 each complexed with both the GSN N-ter and GSN C-ter were obtained from the molecular docking of PIP2 and PIP3 to the catalytic domain of gelsolin, respectively as described above. Dundee PRODRG Server 22 was used for generating the topology and parameter files of PIP2 and PIP3 molecules for MD simulation. Each protein-ligand complex was placed in the center of a cubic box with specific dimensions containing definite simple point charge water molecules for the individual complex as shown Supplementary Table 3. Some water molecules were replaced by ions to neutralize the systems.
All starting structures were subsequently energy minimized with a steepest descent method for 5000 steps. The results of these minimizations produced the initial structures for the position restraining MD simulations. These structures were subjected to 200 ps of position restrained simulations, where the protein and the ligands were kept fixed and water molecules were allowed to settle down around the ligand protein complex. These structures were finally used for the 50 ns of MD simulations in the aqueous solution at 300K temperature. The simulations were performed under two successive equilibration phases. At the first phase of equilibration, constant number of particles (N), volume (V) and temperature (T) i.e. NVT ensemble is maintained for all the complex systems. In the second phase, a constant number of particles (N), pressure (P) and temperature (T) i.e. NPT ensemble were maintained. The SETTLE algorithm 23 was used to constrain the bond length and angle of the water molecules, while the LINCS algorithm 24 was used to constrain the bond length of the peptides. The long range electrostatic interactions were calculated by the Particle Mesh Ewald (PME) method 25 with interpolation order of 4 and a grid spacing of 12 Å. A constant pressure of 1 bar was applied with a coupling constant of 1.0 ps; peptide, ligand, water molecules and ions were coupled separately to a bath at 300 K with a coupling constant of 0.1 ps. The periodic boundary conditions were applied and the equation of motion was integrated at time-steps of 2 fs. All the systems were treated in position restrained environment. The SHAKE algorithm was used to restrain bond lengths involving hydrogen atoms 26 . The van der Waals interactions were treated by using a cutoff of 12 Å. Finally, 50 ns production MD run were performed under the normal temperature and pressure with the coupling time constants of 0.1 ps and 1.0 ps respectively. All the simulations were performed with a time step of 2 fs, and the coordinates were saved every 1 ps. The analyses were carried out using the programs within the GROMACS 5.  [34][35][36] . PCA is a standard tool in statistical mechanics, which allows us to determine the correlated motions of the residues to a set of linearly uncorrelated variables called principal components. This method is based on the construction of the covariance matrix of the coordinate fluctuations of the simulated proteins. The covariance matrix is diagonalized to obtain the eigenvectors and eigenvalues that provide information about correlated motions throughout the protein. We have compared the eigenvalues for first two eigenvectors of principal component (PC1 and PC2) of mutant residues in their corresponding systems. Average structure as a function of each principal component has been investigated to determine the role of specific mutation on protein-ligand binding mode and interaction affinity.
Isolation, culture and stretching of adult cardiomyocytes. Adult murine left LV cardiomyocytes were isolated and cultured as described previously 11,37 . Cardiomyocytes were cyclically stretched at 1 Hz and a maximal elongation of 10% for 6 or 24 hours with Flexcell FX-5000 Tension System (Flexcell Int. Corp.). After the completion of stretching protocol the cells were imaged under light microscope (DM4000 B, Leica) to assess the cell viability. Media were collected to perform the LDH assay to assess the cell death. The cells were then washed with PBS and fixed with 2% paraformaldehyde for 20 minutes followed by 3 washings with PBS. The fixed cardiomyocytes were then used to perform the F-actin and G-actin double staining. The ratio between F-actin and G-actin staining intensities was represented as an index of actin polymerization.          Figures 2n, 5a, and 5b and Supplementary Figures 1a,  10a, 10b, and 10c . 26