Coding variants identified in patients with diabetes alter PICK1 BAR domain function in insulin granule biogenesis

Bin/amphiphysin/Rvs (BAR) domains are positively charged crescent-shaped modules that mediate curvature of negatively charged lipid membranes during remodeling processes. The BAR domain proteins PICK1, ICA69, and the arfaptins have recently been demonstrated to coordinate the budding and formation of immature secretory granules (ISGs) at the trans-Golgi network. Here, we identify 4 coding variants in the PICK1 gene from a whole-exome screening of Danish patients with diabetes that each involve a change in positively charged residues in the PICK1 BAR domain. All 4 coding variants failed to rescue insulin content in INS-1E cells upon knock down of endogenous PICK1. Moreover, 2 variants showed dominant-negative properties. In vitro assays addressing BAR domain function suggested that the coding variants compromised BAR domain function but increased the capacity to cause fission of liposomes. Live confocal microscopy and super-resolution microscopy further revealed that PICK1 resides transiently on ISGs before egress via vesicular budding events. Interestingly, this egress of PICK1 was accelerated in the coding variants. We propose that PICK1 assists in or complements the removal of excess membrane and generic membrane trafficking proteins, and possibly also insulin, from ISGs during the maturation process; and that the coding variants may cause premature budding, possibly explaining their dominant-negative function.


Immunostaining
Cells were washed in ice-cold PBS and fixed with 4% paraformaldehyde, 10 min on ice and 10 min at RT. COS7 cells were washed with PBS and milliQ, and mounted on a glass slide with Prolong® Gold antifade mounting reagent. INS-1E cells were washed in PBS and permeabilized for 30 min in PBS containing 0.2% saponin and 5% goat serum (GS). Subsequently, cells were incubated with primary antibodies (diluted in PBS with 5% GS) for 1 h at RT or overnight at 4°C and washed in PBS prior to incubation with secondary antibodies (diluted in PBS with 5% GS) for 30 min. Cells were washed in PBS and milliQ before mounted on a glass slide using Prolong® Gold antifade mounting reagent. Glass slides were stored in the dark at 4°C. b-cells were fixated and immunostained following the same protocol as for INS-1E cells but mounted on a glass slide using DAPI Fluoromount-G® (SouthernBiotech).

Cell lysates
For immunoblotting and ELISA, cells were washed in ice-cold PBS prior to lysis in 1xRIPA buffer , 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 1mM EGTA, 0.5% NP-40) containing 1 mM PMSF or cOmplete protease inhibitor cocktail tablets and phosphatase inhibitor cocktail 3 (Sigma-Aldrich). Cells were scraped off, centrifuged at 11,000 g for 40 min at 4°C and the supernatant was stored at -20°C.
The protein content was determined using the bicinchoninic acid assay kit (Pierce TM , Thermo Fisher Scientific) to adjust protein concentration for immunoblotting and to calculate relative proinsulin and insulin content.

Immunoblotting
Protein lysates were mixed with 5xsodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer, boiled at 95°C for 5 min before loaded into an Any-kD or 4-15% precast gel (Mini-ProTEAN® TGX TM , BIO-RAD) and run at 100V for 1-2 hours depending on the protein of interest. The size-separated proteins were transferred to polyvinylidene difluoride membranes (BIO-RAD) for 2 hours at 18 V or for 10 min at 25 V on a trans-blot turbo transfer system (Bio-Rad) and blocked for 1 h in 5% milk in wash buffer . Membranes were incubated with primary antibodies for 1 h at RT or overnight at 4°C, washed 3 times and incubated with HRP-conjugated secondary antibodies for 30 min. Membranes were washed 3 times before developed using either the SuperSignal ELISA Femto Substrate (Thermo Fisher Scientific) or ELC Prime Western Blotting system (Sigma-Aldrich) and captured with a cooled CCD camera.

Co-immunoprecipitation
15x10 6 transduced INS-1E cells were seeded out in tissue culture flasks three days prior to lysis. Cell lysis were performed as described above, using lysis buffer , 150 mM NaCl, 10 mM MgCl and 0.5% NP-40) containing cOmplete protease inhibitor cocktail tablets and phosphatase inhibitor cocktail 3. 50 µg protein lysate per condition was saved for input and 500-1000 µg protein per condition was used for immunoprecipitation analysis. The protocol from the manufacturer was followed. In short, the protein lysate was transferred to prewashed (in lysis buffer) GFP-Trap® magnetic particles (M-270, ChromoTek) and incubated on a rotator for 1 h at 4°C. Particles were separated from the supernatant by using a magnetic DynaMag rack (Invitrogen, Thermo Fisher Scientific). The particles were washed three times in wash buffer (10 mM Tris/Cl (pH 7.4), 150 mM NaCl, 0.05% NP-40, 0.5 mM EDTA) before mixed with 2xSDS-PAGE buffer and boiled for 5 min at 95°C. Immunoprecipitated protein and input were assessed by immunoblotting. The images were analyzed using the ImageJ software program (Rasband W. S., ImageJ, U.S. National Institutes of Health, Bethesda, MD, USA).

Bio-ID
48 10x10 6 INS-1E cells were seeded out in tissue culture flasks a day prior to transient transfection. Transfection was performed as mentioned above, 5 µg DNA (in optiMEM) was used per flask and 5 hours after substituted with F10-DMEM media (Thermo Fisher Scientific). hours post-transfection cells were washed with PBS and scraped off. The cell suspension was centrifuged at 1000 g for 5 min at 4°C, supernatant was aspirated and the pellet was snap frozen using liquid nitrogen. The pellet was resuspended in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate, pH 7.4), rotated for 1 h at 4°C and centrifuged at 13,000 g for 15 min at 4°C. Subsequently the supernatant was transferred to prewashed (in lysis buffer) sepharose 4B beads (Sigma-Aldrich). The sepharose 4B bead suspension was incubated on a rotator for 1 h at 4°C and centrifuged at 1000 g for 5 min at 4°C.

Protein expression and purification
Bacteria were pre-cultured in 25 mL lysogeny (LB) medium (supplemented with kanamycin and chloramphenicol) and incubated on rotating shakers at 37 o C overnight. Bacteria pre-cultures were transferred to 500 mL LB medium and grown at 30°C to an OD of 0.6. Protein expression was induced with 250 µL 1 M Isopropyl β-D-1-thiogalactopyranoside and bacteria were incubated overnight at 20°C. Bacteria were harvested by centrifugation and resuspended in 15 µL lysis buffer (50 mM Tris, 125 mM NaCl, 2 mM dithiothreitol (DTT) (Sigma-Aldrich), 1% Triton X-100 (Sigma-Aldrich), 30 µg/ml DNAse 1 (BioNordika) and one tablet of cOmplete protease inhibitor cocktail (Roche) per 100mL buffer). The resuspended pellet was frozen to -80°C. The bacterial suspension was thawed at 4°C and cleared by centrifugation (18.000 x g for 30 min at 4°C).
The supernatant was collected and incubated with 400µL glutathione-sepharose 4B beads (GEHealthcare) for 1.5 hour at 4°C on a rotator. The beads were spun at 3.000 x g for 5 min at 4°C, the supernatant was discarded, and the beads were resuspended in 15 mL wash buffer (50 mM Tris, 125 mM NaCl, 2 mM DTT and 0.01% Triton X-100).
The washing step was repeated three times. The washed beads were transferred to Bio-Spin® chromatography columns (Bio-Rad Laboratories, Inc. cat. #7326008) and washed in 1 mL wash buffer. Bead solution on the column was incubated with 2 µL 0.075U Thrombin (EMD Millipore, Novagen®) in 250 µL wash buffer overnight at 4 o C on a rotator. PICK1 was eluded on ice and absorption at 280 nm was measured on a TECAN plate reader or a Thermo Scientific TM NanoDrop 2000c. The protein concentration was calculated using lambert beers law (εA280PICK1=32320(cm*mol/L)-1).
Alexa Fluor 488 C5 maleimide (Invitrogen TM ) labelling of GST-PICK1 WT and the coding variants was prepared as above, but without addition DTT using 2mM tris (2-carboxyethyl) phosphine instead. Alexa Fluor 488 C5 maleimide was added to the Bio-Spin® chromatography columns prior to Thrombin restriction. The columns were sealed and incubated on a rotator for 4-16 hours at 4 o C. The bead solution was washed 5x with wash buffer removing unconjugated dye. Thrombin was added and protein was eluded as described.

Fluorescence polarization assay
Fluorescence polarization measurements were carried out as previously described, both for the saturation and competition assay (5,6). In short, for the saturation assay

Liposome preparation
The lipids were dissolved in chloroform, before addition of 2% DiD (w/w) for the liposome deformation assay. Lipids were dried while rotating using N2-gas, creating a thin lipid film, followed by vacuum dehydration for at least 10 hours to remove residual chloroform. Lipids were rehydrated in a sterile 200 nM D-Sorbitol solution (pH 7.4) to a final concentration of 0.5 or 1 mg/mL for the liposome deformation assay or TEM imaging, respectively. The rehydrated lipids were put through minimum 8 "freeze-thaw" cycles; frozen using liquid nitrogen and thawed in a 40-50°C water bath. The liposomes were extruded through a 19 mm polycarbonate Whatman TM Nuclepore TM Track-Etched membrane with a pore size of 1 µm using a LiposoFast liposome extruder (Avestin).

Flow cytometry
Increasing concentrations of purified PICK1 WT and PICK1 coding variants were incubated with a fixed concentration of liposomes (~2.5 µg/mL, 2% DiD (w/w)) and left at room temperature (RT) for 1 h prior to acquisition. Fluorescence intensity for DiD and AF488 were detected by photomultiplier tubes (PMTs) after passing through 660/20 nm and 505LP, 530/30 nm light filters, respectively. DiD was excited by a 633 nm JDS Uniphase HeNe laser (17 mW) and AF488 by a Coherent Sapphire 488 nm, air cooled laser (20 mW). Events were triggered by fluorescence, using the 488 nm laser filtered through a 685 long pass filter, and 695/40 nm bandpass filter. The voltage applied to the photomultiplier tube detecting events was set to 500V with a threshold value of 200, defined as the highest gain allowing a maximum of 1 event/s when running a filtered phosphate buffered saline solution. The voltages applied to the detectors of AF488 and DiD fluorescence were kept constant throughout experiments at 335V and 465V, respectively. We recorded at least 100,000 events per condition for each experiment, with a constant number of events per condition within each experiment.
For fission analysis, we applied gates removing >99.9% of the DiD background noise, acquired when running samples containing protein diluted in PBS. The resulting data were applied in a kernel density estimation with 512 equally sized bins on a log scale ranging from 0 to 5 (A.U.) and normalized to the population obtained from a sample only containing liposomes by subtraction. Data were analyzed using FlowLogic software (Inivai).

Proinsulin and insulin ELISA
High range rat proinsulin and insulin ELISA kit immunoassays (ALPCO TM ) were used to measure the proinsulin and insulin content in INS-1E lysates. Cells were seeded at a density of 250,000 cells/well in 12-well tissue culture plates (TPP®, Sigma-Aldrich) four days prior to cell lysis. Protocol provided by the manufacturer was followed. Briefly, the kit consists of a 96-well microplate pre-coated with a monoclonal antibody for proinsulin or insulin. Lysates and HRP-conjugated secondary antibodies were added. Absorption was measured using an Omega POLARstar plate reader at 450nm and 590nm.

Light microscopy
LSM 510 settings; Alexa Fluor® 488 and YFP fluorescent signals were detected using a 488 nm argon laser. Alexa Fluor® 568 and Alexa Fluor® 647 fluorescent signals were detected using a 543 nm helium-neon laser and a 633 nm helium-neon laser, respectively. LSM 710 settings; Alexa Fluor® 488, Alexa flour® 568 and Alexa Fluor® 647 fluorescent signals were detected using a 488 nm argon laser, a solid state 561 nm laser and a helium-neon 633nm laser, respectively. .Channels were imaged separately

Co-localization analysis
To examine colocalization between PICK1 and different cellular markers, Van Steensel's cross-correlation function was used on the processed images with the JaCoP plugin in ImageJ (7). The Van Steensel's cross-correlation calculates the Pearson cross-correlation as the signal from channel one (ex. Cyan PICK1) shift relative to the signal from channel two (ex. Magenta insulin) in the x-direction pixel per pixel. The peak at ∆x = 0 indicates a specific colocalization and a value of 1 indicates total overlap. Analysis of colocalization was performed blinded.

dSTORM imaging
Images were acquired at 10,000-30.000 cycles of one frame of 561 nm laser activation followed by one frame of 647 nm laser activation at 16 Hz per cycle. Shutters regulated the light path in order to separate the light between the frames. The 561 nm and 647 nm lasers were held constant at 0.6 and 1.1 kW cm −2 , respectively, while a 405 nm laser was gradually increased to <0.1 kW cm -2 .

Airyscan 2 image analysis
Images detected with the Airyscan 2 detector were processed in the Zen software (Carl Zeiss, Oberkochen, Germany) using airyscan mode to generate a single image, same setting was used for each individual experiment. Due to high background noise in the 488 channel, untransduced b-cells were immunostained for eGFP and the highest intensity value observed was used as threshold to identify eGFP positive and negative cells for each individual experiment. ROIs in the 488-channel combined with the DAPI channel and insulin channel were used to identify cells. Prior to quantification of the immunosignal, an auto threshold (v1.17.2) from the ImageJ software was employed for each channel. These settings were held constant throughout each individual experiment. The images were converted to binary images and multiplied into each ROI of the original image, resulting in a total intensity value. For total immunosignal, the immunosignals from transduced cells were normalized to the mean of untransduced cells from each individual experiment. Quantification was performed blinded.

dSTORM image analysis
Localizations were fitted to the dSTORM movies with the ThunderSTORM plugin for imageJ (8). Images were filtered with the Wavelet filter (B-Spline), and localizations detected with the local maximum detector with a threshold of 2.0*std (Wave.F1). Drift was corrected with cross correlation and localization with uncertainty greater than 20 nm was filtered out. For the colocalization analysis, we used coordinate based colocalization (CBC) (9) as previously described for PICK1 and insulin clusters (1). The CBC was completed in 20 steps of 5nm, and the clusters were identified with the DBSCAN algorithm through the sklearn library in python 3.6, where a localization was classified as clustered if it had 10 localizations or more within a 40nm radius. Each localization within a PICK1 cluster was assigned a CBC value ranging from 1 to -1 based on the distance to clusters of insulin and vice versa. Values of 1 define a perfect overlap between a PICK1 and an insulin cluster, whereas a value of -1 indicates no overlap between the clusters. DBSCAN was performed in 3D. Diameter was determined from the 2D area of the convex hull of the cluster in the X-Y plane by fitting the area to a circle. The software programs ImageJ and MATLAB (MathWorkers, Natick, MA, USA) were used for image visualization.

SIM image analysis
Images were reconstructed using a built-in algorithm in the Zen software. A noise filter was added and held constant for each individual experiment, although always set between -4.2 and -4.5. Bead alignment matrix was used for drift correction. For 3D visualization we used the Amira software 2019.1 (Thermo Fisher Scientific, Waltham, MA, USA).

AlphaFold2 Setup
AlphaFold2 structures were predicted based on multiple sequence alignment results mapped using mmseqs2. 5 models were predicted for each model and ranked based on their respective pTMscores. The structures were fed back into the neural network a maximum of 3 times for refinement. Subsequently, the sidechains of the refined structures were relaxed using Amber-Relax (10 The two dotted lines represent mean ± SEM from KD and KD + WT from (C), n = 13-