Structural organization of RDGB (retinal degeneration B), a multi-domain lipid transfer protein: a molecular modelling and simulation based approach

Abstract Lipid transfer proteins (LTPs) that shuttle lipids at membrane contact sites (MCS) play an important role in maintaining cellular homeostasis. One such important LTP is the Retinal Degeneration B (RDGB) protein. RDGB is localized at the MCS formed between the endoplasmic reticulum (ER) and the apical plasma membrane (PM) in Drosophila photoreceptors where it transfers phosphatidylinositol (PI) during G-protein coupled phospholipase C signalling. Previously, the C-terminal domains of RDGB have been shown to be essential for its function and accurate localization. In this study, using in-silico integrative modelling we predict the structure of entire RDGB protein in complex with the ER membrane protein VAP. The structure of RDGB has then been used to decipher the structural features of the protein important for its orientation at the contact site. Using this structure, we identify two lysine residues in the C-terminal helix of the LNS2 domain important for interaction with the PM. Using molecular docking, we also identify an unstructured region USR1, immediately c-terminal to the PITP domain that is important for the interaction of RDGB with VAP. Overall the 10.06 nm length of the predicted RDGB-VAP complex spans the distance between the PM and ER and is consistent with the cytoplasmic gap between the ER and PM measured by transmission electron microscopy in photoreceptors. Overall our model explains the topology of the RDGB-VAP complex at this ER-PM contact site and paves the way for analysis of lipid transfer function in this setting. Communicated by Ramaswamy H. Sarma


Introduction
Eukaryotic cells consist of sub-cellular compartments, enclosed by a membrane bilayer, each with a unique composition of proteins and lipids that is critical to their function (Van Meer et al., 2008).The unique lipid composition of organelles has to be maintained in the face of ongoing intracellular vesicular transport, cell signalling and physiology.Given that lipids are hydrophobic in nature, establishing the lipid composition of organelle membranes requires their exchange with other organelle membranes, especially the endoplasmic reticulum (ER), the principal site of lipid synthesis in cells (Balla et al., 2019).Lipid transfer between organelles occurs by multiple mechanisms; one of these is the function of lipid transfer proteins (LTPs) that are able to transfer lipids between membranes (Cockcroft & Raghu, 2018).It has also become apparent in recent years that LTPs are enriched at specific membrane regions called membrane contact sites (MCS), locations in cells where two organelle membranes are closely apposed (< 30 nm) but without undergoing fusion (Levine & Patel, 2016) (Prinz et al., 2020).Most LTPs contain a lipid transfer domain within which lipid binding and transfer activity is contained.In addition, many LTPs also contain evolutionarily conserved protein domains that play a role in their localization and regulation (Levine & Patel, 2016).However, the mechanism by which the multiple domains of an LTP function in a co-ordinated manner remains unclear.One reason for this is that there have been relatively few LTPs for which the molecular structure of the protein has been determined.Insights into this problem could therefore help better understand the regulation of LTP function at MCS.
Among the many classes of LTPs are phosphatidylinositol transfer proteins (PITPs).These proteins are able to transfer phosphatidylinositol (PI) between membranes in vitro.In vivo, several studies have demonstrated their importance in supporting phosphoinositide signalling (Cockcroft & Garner, 2013).Phosphoinositide signalling is a widely used method of cellular communication where cell surface receptor CONTACT Padinjat Raghu praghu@ncbs.res.inSupplemental data for this article can be accessed online at https://doi.org/10.1080/07391102.2023.2179545.
activation is transduced into biological activity through the activation of phosphoinositide specific phospholipase C that hydrolyses the membrane lipid phosphatidylinositol 4,5 bisphosphate (PIP 2 ) to generate inositol 1,4,5 trisphosphate (IP 3 ) and diacylglycerol (DAG) at the plasma membrane.During this process, PIP 2 , a quantitatively minor membrane lipid is rapidly consumed and has to be replenished.This is achieved through phosphorylation of phosphatidylinositol (PI); since PI is synthesized at the ER and PIP 2 is generated at the PM, PITPs are thought to mediate the transfer of PI from the ER to the PM where lipid kinases generate PIP 2 .Through this mechanism, PITPs likely contribute to the numerous cell biological and physiological functions where they have been proposed to function (Routt & Bankaitis, 2004).
Among the known classes of PITPs two broad categories can be recognised (i) Small proteins which consist of only a single PITP domain (ii) Proteins in which a PITP domain is present in cis with multiple other domains within the same polypeptide (Hsuan & Cockcroft, 2001).This latter class are referred to as the multidomain PITPs and the function of the non-PITP domains in regulating the function of the PITPd has been a matter of much interest.The founding member of multi-domain PITP family, RDGB, (Lev, 2004) was identified in Drosophila melanogaster (Dm) (Harris & Stark, 1977) as the protein encoded by the rdgB gene (Milligan et al., 1997) (Vihtelic et al., 1991).In addition to the N-terminal PITP domain, Dm-RDGB contains an FFAT (two phenylalanine residues in an acidic tract) motif followed by two C-terminal domains, the DDHD domain (lipids and heavy metal binding domain with three aspartate (D) and one histidine (H) residue that are functionally important) (IPR004177) (Lev, 2004) and an LNS2 domain (belongs to the HAD superfamily) (Csaki et al., 2013;(Han et al., 2006) (IPR013209).Dm-RDGB is enriched in Drosophila photoreceptors and in these cells it is localized to the ER-PM MCS between the ER and the microvillar plasma membrane (Vihtelic et al., 1991).Loss of function mutants in rdgB show a reduced electroretinogram amplitude upon light stimulation, light-dependent retinal degeneration (Harris & Stark, 1977) and altered kinetics of PIP 2 resynthesis following PLC activation (Yadav et al., 2015).The PITP domain of RDGB has been biochemically characterized and is known to transfer PI and phosphatidic acid (PA) in vitro, and this PI binding and transfer activity is required to support RDGB function in vivo (Milligan et al., 1997) (Yadav et al., 2015).A number of recent studies have addressed the function of the additional domains in RDGB.The interaction of the FFAT motif with the ER integral protein dVAP-A is required for the localization of RDGB (Yadav et al., 2018).A recent study has shown that the LNS2 and DDHD domains are also required for correct localization to the ER-PM MCS and normal function (Basak et al., 2021).It has also been suggested that inter-domain interactions in RDGB may be important for function (Basak et al., 2021).However, in the absence of a structure for the full-length protein, the organization of the individual domains and their contribution to orienting the protein for normal function at the ER-PM MCS remains unclear.In order to address this question, we have used in silico integrative modelling to generate a structure of full-length RDGB and using simulation based approaches uncovered the mechanism by which RDGB is anchored to the ER and PM side of the ER-PM MCS in photoreceptors.Our findings topologically position RDGB at the ER-PM MCS and offer an insight into the importance of the non-PITP domains in orienting the protein for function.

Structure prediction of RDGB
In order to obtain a 3-D structure of the protein, homologybased modelling (Sali & Blundell, 1993), fold prediction (PHYRE 2) (Kelley et al., 2015) and ab-initio structure prediction (I-Tasser) (Yang et al., 2015) methods were combined (Table 1).The regions which have high homology to known structures available in PDB database were modelled using the homology modelling protocol in MODELLER.The regions that did not have template predicted with high identity but had closely related protein FOLD (classified as per SCOP classification) structure were modelled using fold prediction method while those regions of the model which had neither of the two above mentioned templates were modelled ab-initio.The table lists all templates that were obtained by either homology or fold prediction, while regions where template coverage was low were modelled using ab-initio methods.The final fulllength model was energy minimized for 500 steps of conjugate gradient cycles followed by steepest descent gradient for 500 steps.The initial steps of minimization were stringent to avoid large structural deviations in the model while minimization.Once the model reached a negative potential energy state with allowed Ramachandran plot values, relaxed energy minimization steps were performed to further reach an energy minimum.The final model has negative potential energy and is stable.The Ramachandran plot distribution of amino acids in the model was checked using the PROCHECK analysis tool (Laskowski et al., 1993).ProSA was used to check whether the model built satisfied the correctness of protein structure as predicted by the ProSA Z-scores (Wiederstein & Sippl, 2007).To understand the importance of each domain in the protein, structures with C-terminal domain deletions were generated from the full length model as follows-RDGB (USR1-LNS2)D , RDGB (DDHD-LNS2)D and RDGB LNS2D .

Normal mode analysis (NMA) to understand interdomain movements
Normal modes to understand inter-domain movements in the RDGB protein were calculated using the ANM 2.1 server.ANM (Anisotropic network model) is a simple NMA tool for analysis of vibrational motions in molecular systems (Eyal et al., 2015).Elastic Network methodology was used and it helped to represent the system at the residue level.The macromolecule is represented as a network of atoms.In the model each protein node is the C a atom of a residue and the overall potential is the sum of harmonic potentials between interacting nodes.The network included all interactions within a cut-off distance (distance cut-off of 15 Å).This was the predetermined parameter in the model.Information about the orientation of each interaction with respect to the global coordinates system was considered along the force constant.The force constant was described by Hessian matrix.Each element of the matrix is interaction between two nodes i and j (two C-alpha atoms of two amino-acids).The distance between two nodes was added as a weight at each element of the matrix.The Eigen vectors of the matrix describe the vibrational direction and the relative amplitude in the different modes.The mean square fluctuations of individual residues were obtained by summing the fluctuations in the individual modes.20 normal modes were calculated using the method described above for RDGB protein.Informative modes with the lowest frequency possible and modes responsible for the conformational changes were analysed further for domain movements.

Molecular docking of RDGB, RDGB (USR1-LNS2)D and RDGB DDHD-LNS2D to dVAP-a
Molecular docking was performed for each of the RDGB models with that of dVAP-A [obtained by homology modelling using dVAP-A X-ray crystal structure, PDB ID: 1Z9O (from Rattus novergicus)].Since the template structure was more than 90% identical, molecular modelling was straight forward based on homology modelling.GRAMM-X protocol (Tovchigrechko & Vakser, 2006) was used to carry out the docking studies.GRAMM-X follows the Fast Fourier Transformation (FFT) methodology.It uses a smoothed Lennard-Jones potential on a fine grid during the global search FFT stage, followed by the refinement optimization in continuous coordinates and then performs re-scoring of the complexes using the knowledge-based potential terms for protein-protein interactions.20 models were generated for every set of protein-protein interactions.The energy values were calculated for all the 60 models using the PPcheck algorithm (Sukhwal & Sowdhamini, 2015).The models with low potential energy values were selected to understand the interaction of RDGB with dVAP-A protein.Each of the low energy complex was manually visualized in PYMOL (The PyMOL Molecular Graphics System, Version 2.0 Schr€ odinger, LLC.) for interactions between FFAT motif of the RDGB protein and MSP domain of dVAP-A.The models which had the relevant known interactions between the two proteins were selected for further analysis.The protein-protein complex was analysed for stable interactions based on PPcheck evaluation and models with lowest overall energy and very low unfavourable interaction was selected as the best model.Each of the best models selected was analysed further using molecular dynamics simulation for stability of the complex in solution.Desmond module of molecular dynamics was used to perform 100 ns simulation for each protein (in replicates).OPLS_2005 force field with standard NPT conditions was used.The protein complexes were solvated in an orthorhombic box with periodic boundary conditions by adding TIP3P water molecules.The initial equilibration was carried out using default protocol of restrained minimization followed by molecular dynamics simulations for 100 ns.

Molecular dynamics (MD) simulation of RDGB interactions at a membrane
MDS were done with full length RDGB protein using Desmond MDS and the parameters such as stable secondary structure, RMSD and RMSF were analysed.MDS were also The table represents the details on templates used for modelling the RDGB protein using an integrative modelling approach.The first column represents the different regions of RDGB protein modelled individually using different templates.The second and third column represents the available template and the protein name of the template.The fourth and fifth column represent the confidence (in percentage) with which the template is selected for modelling and percent identity of the template with the RDGB protein sequence respectively.The last column indicates the method used for modelling.The well-structured domains of the RDGB protein were modelled using the available templates, either by homology modelling or fold prediction.For regions with low confidence or low sequence identity with the template, ab-initio modelling protocol was used.
carried out for RDGB protein in the presence of a DPPC membrane using Gromacs to identify the domains and residues important for interaction of the protein with the membrane.CHARMM-GUI membrane builder module (http://www.charmm-gui.org/?doc=input/membrane.bilayer) was used to generate the membrane input file for Gromacs MDS (Berendsen et al., 1995).A dipalmitoylphosphatidylcholine (DPPC) membrane with 600 molecules of lipids each on upper and lower leaflet of the membrane was built using the CHARMM-GUI membrane builder.Three independent membrane-protein systems were generated, RDGB protein (system 1), RDGB KK/AA protein (system 2) and the RDGB LNS2D protein (system 3).The protein was placed at a distance of 10 ( � Å) along the Z-axis from the membrane at the start of the simulation.The protein and the membrane system were solvated in an orthogonal box with TIP3P water and neutralized with ions to get a final system with net charge being zero.The simulations were performed with a 2 fs step size and the nearest neighbour list was recorded every 20 ps.NVT and NPT equilibration was performed as mentioned in the protocol (https://charmm-gui.org/?doc=input/membrane. bilayer).The temperature and pressure were maintained at 300K and 1 bar À 1 respectively.The system was energy minimized until convergence using Gromacs charmm36m force field for 50 ns followed by six steps of equilibration each for 50 ns.This helps in stabilization of the large protein-membrane complex.A minimum of six equilibration steps is recommended by the protocol.The final MD run was carried out for 100 ns (with two replicates for each system with different initial velocities) using charmm36m force-field.All the downstream analysis was carried out using different modules of GROMACS and PYMOL (The PyMOL Molecular Graphics System, Version 2.0 Schr€ odinger, LLC).

Protein-membrane interaction
The g_mindist command was used to calculate the minimum distance between the protein and the bilayer throughout the simulations.

Interacting residues
The distance of the residues that interact with membrane was calculated using the n_index and g_minddist module of Gromacs.

Video-files for molecular dynamics simulation
The video files were generated using the PYMOL tool (https://pymol.org/2/).

Generation of the mutant RDGB protein, RDGB KK/AA structure in-silico
RDGB model with K1186 and K1187 residues mutated to alanine was obtained using FOLDX (Schymkowitz et al., 2005).
The repair model and build model module of FOLDX were used along with rotamer stabilization to obtain the structure of the mutant protein.

Molecular biology
BDGP gold clone 09970 containing the RDGB-RA transcript was used as the parent vector for making various constructs of RDGB used for the experiments (Basak et al., 2021).In the full length RDGB clone, site directed mutagenesis was used to mutate lysine residues at position 1186 and 1187 to alanine.For cloning of LNS2 KK/AA , site directed mutagenesis was done in the pJFRC-LNS2:: GFP clone.Lysine residues corresponding to 1186 and 1187 positions in RDGB were mutated to alanine in pJFRC-LNS2::GFP by introducing the mutations directly in the primers used for amplification.

Cell culture, transfection and immunofluorescence
S2R þ cells were cultured in Schneider's insect medium (HiMedia) supplemented with 10% Fetal Bovine Serum and with antibiotics Penicillin and Streptomycin.Cells were transfected using Effectene (Qiagen) as per manufacturer's protocol.Post 24 hours of transfection, cells were fixed with 4% paraformaldehyde (Electron Microscopy Sciences) and imaged to observe for GFP fluorescence using a 60 � 1.4 NA objective, in Olympus FV 3000 microscope.

Co-immunoprecipitation
Snap-frozen Drosophila heads were lysed in ice-cold Protein Lysis Buffer (50 mM TrisCl, 1 mM EDTA, 1 mM EGTA, 50 mM NaF, 0.27 M Sucrose, 1% Triton X-100 and 0.1% b-Mercaptoethanol).10% of the head lysate was taken out separately to be used as input in western blotting.The remaining head lysate was then split equally into two equal parts.To one part, 2 ml of dVAP-A antibody sera was added, and to other part, 2 ml of control IgG was added.The vials were then kept on a vertical rotor overnight at 4 � C. On the following day, Protein-G sepharose (GE Healthcare) beads were spun at 13000X g for 1 minute, and then washed using Tris-buffered saline (TBS) for two times.The beads were blocked using 5% Bovine Serum Albumin (BSA) (HiMedia) in TBS containing 0.1%Tween-20(Sigma Aldrich) (TBST) for 2 hrs at 4 � C. Post blocking, equal amount of beads were then added in the experimental and IgG control samples.The samples were then incubated for another 2 hours at 4 � C. Post incubation, the samples were washed and centrifuged at 13000Xg twice with TBST containing 0.1 M EGTA and 0.1% b-Mercaptoethanol for 5 minutes each.The supernatant was taken out, and the precipitated beads were mixed with 2X Laemmli sample buffer and boiled at 95 � C for 5 minutes, for further use in western blotting.

Immunocytochemsitry
For immunohistochemistry, retinae of one-day old flies (w-; LNS2(RdgB)GFP/TM3Sb and w-; LNS2 KK/AA (RDGB) GFP/TM3Sb) were dissected under bright light in PBS.The samples were then fixed using 4% paraformaldehyde (Electron Microscopy Sciences) in PBS with 1 mg/ml saponin (Sigma Aldrich) for 30 minutes at RT. Post fixation, samples were washed thrice with PBS having 0.3% Triton X-100 (PBTX) and blocked using 5% Fetal Bovine Serum (ThermoFisher Scientific) in PBTX for 2 hrs at RT.The samples were then incubated overnight with anti-GFP [antibody (1:5000) ab13970 (Abcam Cambridge, UK)] in blocking solution at 4 � C. Samples were then washed thrice with PBTX and incubated with the secondary antibody [anti-chick (A21103), IgG (Molecular Probes)] at 1:300 dilutions for 4 hrs at RT.For staining of the F-actin, Alexa Fluor 568-Phalloidin (Invitrogen, A12380) at 1:300 dilutions were added during incubation with the secondary antibody.Samples were then washed in PBTX thrice and mounted with 70% glycerol in PBS.The whole-mounted preparations were imaged under 60 � 1.4 NA objective, in Olympus FV 3000 microscope.

Optical neutralization
The flies were immobilised by cooling them on ice, then decapitated and fixed on the microscope slide with a drop of colourless nail varnish.A drop of immersion oil was used to neutralise the cornea's refractive index, which was then viewed and imaged using the Olympus BX43 microscope's 40X oil immersion objective.The digital image acquisition and documentation was done by using CellSens software.

Scoring retinal degeneration
A total of 50 ommatidia from 5 individual flies of each genotype were analysed for each time point in order to produce a quantitative indicator of degeneration.The reference photoreceptor was a single central UV sensitive photoreceptor that did not demonstrate any light-dependent retinal degeneration, and the rest of the photoreceptors were scored.Each rhabdomere that appeared to be wild type received a score of one.As a result, the control photoreceptors will receive a score of 7, whereas the mutant photoreceptors that are degenerating will receive a score ranging from 1 to 7. GraphPad Prism software was used to analyse the data and plot the graph.

Electroretinogram
Anesthetized flies were placed in truncated 200 ml disposable pipette tips, with the head protruding through the narrow opening.The fly's head was immobilised with colourless nail varnish.The voltage variations were measured by inserting the experimental electrode on the surface of the eye and the reference electrode on the thorax, using two glass microelectrodes (640786, Harvard Apparatus, Massachusetts, USA) filled with 0.8 percent NaCl solution.The flies were dark adapted for 5 minutes preceding the ERG recordings, then given a 2 second green light stimulus, with 10 stimuli (flashes) per recording and 15 seconds of recovery time between each flash.An LED light source was used to emit a green light stimulus to within 5 mm of the fly's eyes through a fibre optic guide.Voltage changes were recorded using pCLAMP 10.7 and amplified using DAM50 amplifier (SYS-DAM50, WPI, Florida, USA).Data analysis was done using Clampfit 10.7 (Molecular Devices, California, USA).Graphs were plotted using GraphPad Prism software.

Fluorescent pseudopupil
The PIP 2 biosensor PH-PLC coupled to GFP driven by the transient receptor (trp) promoter of flies was used to monitor changes in PIP 2 levels at the microvillar PM in live flies.Using the same protocol as in ERG recordings, the flies were rendered insentient and immobilized.The 10X objective of the Olympus IX71 microscope was used to focus and image the pseudopupil created by the combined fluorescence of roughly 20-40 neighbouring ommatidia.The images of the pseudopupil were acquired using the Micromanager software by collecting fluorescence released from the eye when GFP was stimulated by a 90 ms pulse of blue light.The flies were dark acclimated for 6 minutes prior to the recordings, during which the probe binds to PIP 2 and localises to the microvillar PM.The PLC activity is triggered by a flash of blue light, which causes PIP 2 to hydrolyse, causing the probe to be displaced from the PM and the pseudopupil to be lost.The probe is retained at the apical PM in the central UV sensitive photoreceptor as it is resistive to blue light.Mean fluorescence intensity indicating the basal PIP 2 pools were calculated using ImageJ from NIH (Bethesda, MD, USA).Quantification of DPP fluorescence intensity was done by measuring the intensity values per area of the pseudopupil.

Lipid binding assay
S2 cells were transfected as mentioned previously.48 hours post transfection cells were lysed in fat blot buffer (5 mM Tris-Cl and 15 mM NaCl).The amount of protein was quantified using Bradford assay and protein expression was checked using western blot (GFP antibody for LNS2::GFP and LNS2 KK/AA:: GFP, mCherry antibody for mCherry::SPO20).Strips made using nitrocellulose membranes [Hybond-C Extra; (GE Healthcare, Buckinghamshire, UK)] were spotted with increasing picomoles of PA (1,2-dioleoyl-sn-glycero-3-phosphate (sodium salt), 840875 Avanti Polar Lipids).After spotting, membrane was dried and then blocked using 5% BSA (HiMedia) in fat blot buffer (5 mM Tris and 15 mM NaCl) for 2 h at RT.The strips were then incubated overnight at 4 � C with the cell lysate.The membranes were washed next day three times with 0.1% TBST and then incubated with anti-GFP antibody (1:2000) or anti-mCherry antibody (1:3000) at RT for 3 hrs.The membranes were then probed with the appropriate HRP-conjugated secondary antibody (1:10,000) and binding was detected using ECL (Bio-Rad) in an Image quant LAS 4000 instrument.

Electron microscopic determination of distance between PM-EM in photoreceptors
The fly heads of mentioned genotypes were cut and immersed in 2% osmium tetroxide, kept at 4 � C for 1 h followed by incubation at 40 � C for 4 days.Specimens were washed with distilled water, stained enbloc with uranyl acetate (0.5% in distilled water) for 3 h.After washing with distilled water, specimens were subjected to dehydration step and embedded in epon.Ultrathin sections of 60 nm were cut, and grids were subjected to post-staining with 2% uranyl acetate (in 70% ethanol) and Reynold's lead solution.Sections were imaged at 120 KV on a Tecnai G2 Spirit Bio-TWIN (FEI) electron microscope.PM-SMC distance quantification was done using ImageJ.Line was drawn from the base of the rhabdomere to the outer layer of the SMC and distance was calculated in nm range.30 photoreceptor cells were taken/biological replicate (wild type, Red Oregon R flies).

Construction and validation of a full-length in silico structural model of RDGB
A full-length structural model of RDGB was built using an integrative modelling.RDGB is a multi-domain protein and contains the following domains-PITPd (1-280 aa), FFAT-motif (396-416 aa), DDHD domain (711-900 aa) and LNS2 domain (1059-1200 aa).There is an unstructured region-USR1, i.e. a region with no secondary structure prediction (Raghu et al., 2021) between the FFAT-motif and DDHD domain (425-675 aa) (Figure 1A).Secondary structure prediction shows that the PITPd contains alpha helices and beta sheets, the DDHD domain is mostly alpha helical and the LNS2 domain also contains both alpha helices and beta sheets.A large stretch of residues between the FFAT motif and DDHD domain showed mainly presence of coils and no similarity to any known protein domain secondary structure.This region was predicted to be mainly composed of disordered segments using three independent algorithms (DisEMBL, PrDOS and IUPred) (Linding et al., 2003;Ishida & Kinoshita, 2007;M� esz� aros et al., 2018;Raghu & Krishnan, 2021).
The PDB structure of rat and mouse PITPa was available (1t27a and 1kcma).The sequence identity between the PITPd of RDGB and the templates were 44% and 52% respectively and functional regions were well conserved.Thus, homology modelling method was used to obtain a structural model of the PITPd.However, for the DDHD domain, no closely related template structure was available.Thus, based on fold prediction, a hydrolase domain (IPR003712) and a transferase domain (IPR010758) was used for modelling the structure of the DDHD domain.The LNS2 domain belongs to the superfamily of HAD-like phosphatase proteins (Csaki et al., 2013).Crystal structures of proteins from the HAD superfamily show that they have a conserved alpha/beta-domain classified into hydrolase fold.Thus, based on homology, the most closely related HAD-like phosphatase proteins were chosen as templates (PDB ID: 1ltqa and 1xpja).The PDB structures 1itq and 1xpg which had 99% similarity (to what) were used to model the structure of the LNS2 domain.
The full-length RDGB model was built in parts (as described above) and then joined sequentially using MODELLER.The individual domains were modelled separately based on the availability of templates.The regions connecting the domains were modelled as mentioned in the Table 1.Using MODELLER, the models of individual domains were combined starting from N-terminus PITPd upto the C-terminal LNS2 domain sequentially.At every stage of model building, Ramachandran plot values were checked to verify that there are minimum structural deviations from the allowed phi-psi angles in protein structure (Figure 1C).Ramachandran plot values show that most of the model has more than 85% residues in allowed regions and 10% in the additionally allowed regions of the Ramachandran plot.The ProSA validation tool was used to further verify for errors in protein structure obtained by any structural biology method used (Figure 1D).ProSA analysis shows that the Z-score for all the parts of the model and the final model is less than À 5.0 indicating the structures have minimum error.Thus, the full-length model built satisfies required structural criteria and also remains stable during MDS.We measured the root mean square fluctuation (RMSF) of each residue throughout the 100 ns molecular dynamics simulation.A plot of RMSF as a function of residue number shows that most of the regions within structured domains show lower than 4 � Å RMSF.The USR1 region which is predicted to have disordered regions shows the highest RMSF (Figure 1B).This indicates that the known structured domain regions are stable in the model.The radius of gyration (Rg), solvent accessible surface area (SASA), secondary structure of the protein has been calculated during the 100 ns simulation to indicate their deviations during the simulations (Supplementary Figure S1A, S1B and S1C).The PCA analysis on the simulation trajectory reveals the conformational differences arising from the axes of maximal variance from the distribution of the structures.3-5 such dimensions are enough to depict the total variation.The PCA plots for the trajectory and the residue-wise representations of the same has been calculated using Bio3D package (Grant et al., 2006) (Supplementary Figure S1D and  1E).The model was then subjected to normal mode analysis (NMA) using the NMA protocol of ANM2.1.The domain movements were observed from the lowest frequency modes.The DDHD domain shows the largest movements towards the LNS2 domain (Figure 1E, Video1).In this NMA, PITP domain remains stable and does not interact with the other domains.The model is available in ModelArchive at https://www.modelarchive.org/doi/10.5452/ma-x536e.

Charged residues in the LNS2 domain stabilize the membrane interaction of RDGB
At the MCS, RDGB has been shown to interact with the microvillar plasma membrane via the LNS2 domain (Basak et al., 2021) and a similar mechanism has also been proposed for its human ortholog PITPNM1/Nir2 (Kim et al., 2013).In the case of PITPNM1, it has been reported that a specific aspartic acid residue at position 1128 mediates the interaction of the LNS2 domain of Nir2 with PA (Kim et al., 2013).Although the LNS2 domain of RDGB also binds PA, mutation of the equivalent D residue to that tested for PITPNM1 (Asp 1164) does not abolish PA binding, RDGB D1164A was still correctly localized and basal PIP 2 levels were not affected (Basak et al., 2021).Thus, the molecular mechanism by which the LNS2 domain of RDGB interacts with the membrane has not been determined.We investigated the membrane binding mechanism of the LNS2 domain by MDS using the protein model generated in this study and a dipalmitoylphosphatidylcholine (DPPC) membrane (system 1-RDGB).The system was subjected to six steps of equilibration at 50 ns per step followed by an MD run for 100 ns (in replicates).We noted the interactions between the LNS2 domain of RDGB and the membrane and the minimum distance between RDGB and the DPPC membrane was measured throughout the 100 ns run.During the MDS, we observed (Video 2) that an alpha helical region of the LNS2 domain containing two charged residues approaches closest to the membrane.These were two lysine residues found at positions 1186 and 1187 of RDGB protein (K1186 and K1187).The distance between K1186 and K1187 in the alpha helical region of the LNS2 domain and the membrane was measured.It was observed that in the MD simulation with system 1-RDGB, the minimum distance between RDGB and the membrane is 2 � Å and remains stable throughout the simulation (Figure 2A (i) and (ii)).The distance between the hydrogen atoms of the residues K1186 (Figure 2A (iii)) and K1187 (Figure 2A (iv)) (closest to the lipid) and the DPPC molecule was less than 2.5 � Å for up to 80% of the duration of the simulation.To test the importance of these two residues approaching the lipid molecules in the membrane, we performed in silico mutations of these two lysine residues to alanine (RDGB KK/AA ) (Figure 1A) using the FOLDX protocol.The mutant RDGB KK/AA protein was as stable as the wild type as measured using FOLDX energy values (RDGB WT protein energy=-1940.68Kcal/moland RDGB KK/AA protein energy¼ À 1939.01Kcal/mol).The RDGB KK/AA protein was then subjected to MDS (referred to as system 2-RDGB KK/AA ) similar to that done for wild type RDGB (system 1).For system 2-RDGB KK/AA , the minimum distance between the protein and the membrane at the start of the simulation was 2 � Å which increased to 8 � Å as the simulation progressed (Figure 2B (i) and (ii)).The distance between the hydrogen atoms (closest to the lipid) of the residues A1186 (Figure 2B (iii)) and A1187 (Figure 2B (iv)) (in system 2-RDGB KK/AA ) and the DPPC molecule remained more than 5 � Å for up to 60% of the simulation after which it deviated to about 30 � Å.The charged residues K1186 and K1187 of LNS2 domain in system 1-RDGB moved at a distance to the membrane such that they could form Van der Waals interactions with each other while the distance between the protein and the membrane in system 2-RDGB KK/AA remained larger throughout the simulation after the system stabilized (Video files 3).The system in which LNS2 domain is deleted (system 3-RDGB LNS2D ) shows no interaction with the membrane throughout the simulation (Figure 2C (i) and (ii) and Video file 4).These findings suggest that presence of the LNS2 domain is important for interaction of RDGB with the membrane.Since the minimum distance between the protein and the membrane is smaller in system 1(RDGB) as compared to system 2 (RDGB KK/AA ), it is clear that this domain and in particular K1186 and K1187 are required for the protein to associate with the membrane.
To experimentally validate the results of our simulations, we mutated residues K1186 and K1187 of RDGB in an LNS2::GFP construct (LNS2 KK/AA ::GFP) and expressed this in Drosophila S2R þ cells.Unlike wild type LNS2 which was specifically localized to the PM and some endomembrane structures, LNS2 KK/AA ::GFP was found to be distributed throughout the cell, with some residual localization at the PM (Figure 3A).This observation is consistent with the predictions of the MD simulation of system 2, where RDGB KK/AA showed minimum interaction with the membrane at the start of the simulation and later deviated away from the membrane due to lack of stable interactions.We also tested the importance of K1186 and K1187 in localizing the LNS2 domain in Drosophila photoreceptors.When we expressed LNS2 KK/AA ::GFP in wild type photoreceptors, as opposed to LNS2::GFP which localized to rhabdomeres (Basak et al., 2021), LNS2 KK/AA ::GFP was found to be diffusely distributed throughout the photoreceptor cell body, implying that K1186 and K1187 residues in the LNS2 domain are essential for targeting this domain to the apical PM (Figure 3B).

The interaction of LNS2 with PA at the membrane is stabilized by K1186 and K1187
A lipid binding assay was used to test LNS2 domain interaction with the lipid PA; as a positive control we used the well-known PA binding protein SPO20 (Horchani et al., 2014).As previously reported (Basak et al., 2021), we found that the LNS2 domain of RDGB binds PA in a concentration dependent manner (Figure 3C); however, the LNS2 domain does not show binding to the lipid PIP 2 which is also a negatively charged lipid found at the PM.This PA binding was lost when the charged residues (K1186 and K1187) were mutated to alanine (Figure 3C).This finding suggests that C-terminal  ii)) The minimum distance between RDGB KK/AA (system 2) and the DPPC membrane during the 100 ns molecular dynamics simulation (averaged over replicates) is represented.The minimum distance between the residues A1186 (B (iii)) and A1187 (B (iv)) of RDGB KK/AA and the DPPC molecule of the membrane is represented (averaged over replicates).(2 C (i) and ( ii)) The minimum distance between RDGB LNS2D (system 3) and the DPPC membrane during the 100 ns molecular dynamics simulation is represented.The graphs were generated using Graphpad PRISM.The X-axis represents simulation time (ns) and Y-axis represents distance between atoms ( � Å).
LNS2 domain of RDGB interacts with PA through K1186 and K1187 at the membrane and this helps the protein to localize at the PM.The expression of the proteins used for lipid binding assays was checked using western blot and both the LNS2-GFP and LNS2 KK/AA ::GFP proteins express well in S2 cells (Figure 3D).

K1186a and K1187A are required for functional RDGB protein at the contact sites
To understand if the membrane interacting residues of LNS2 are important for RDGB protein function we introduced K1186A and K1187A mutations in the full length protein (RDGB-LNS2 KK/AA ) and expressed this protein in rdgB 9 photoreceptors (rdgB 9 ; GMR > rdgB-LNS2 KK/AA ).Unlike, the full length RDGB that is localized to the ER-PM contact site, RDGB-LNS2 KK/AA showed a partial mislocalization away from the base of the rhabdomere suggesting that K1186 and K1187 in the LNS2 domain are essential to localise RDGB to the ER-PM MCS (Figure 4A).
To test the impact of these residues for RDGB function, we studied three phenotypes of RDGB mutants.First we measured the electrical response of the photoreceptors to light and found that RDGB-LNS2 KK/AA was able to partially rescue the reduced amplitude of rdgB 9 (Figure 4B, C).Second, we measured the PIP 2 levels at the plasma membrane of the rhabdomere and found that reconstitution with RDGB-LNS2 KK/AA could partially restore the reduced PIP 2 levels seen in rdgB 9 to that seen in wild type controls (Figure 4D, E).Lastly, RDGB-LNS2 KK/AA was able to partially supress the light-dependent retinal degeneration phenotype of rdgB 9 (Shweta Yadav et al., 2015) (Figure 4F and G).Collectively, these findings suggest that the membrane interacting residues (K1186 and K1187) of LNS2 contribute to the proper localisation of RDGB at the ER-PM interface and in turn are required for its normal function.

An LNS2 domain of RDGB with PA phosphatase activity reduces it function in photoreceptors
A large number of proteins including Lipins have an LNS2 domain with a conserved motif DXDX(T/V) that is important for PA phosphatase activity (Kim et al., 2013) (Supplementary Figure 2).This motif is not conserved in LNS2 domain of RDGB proteins from multiple organisms, with the first aspartic acid residue that is critical for phosphatase activity replaced by serine (Figure 5A); this observation indicates that the LNS2 domain in RDGB is not likely to have a phosphatase activity.We mutated the serine residue within the SXDX(T/V) motif of the LNS2 domain of RDGB (this protein is referred to as RDGB PAP ) and tested its functionality.When expressed in photoreceptors, in contrast to wild type RDGB, RDGB PAP did not localize to the base of the rhabdomere (Figure 5B).When RDGB PAP was reconstituted in rdgB 9 photoreceptors, this protein was able to only partially rescue the reduced electrical response to light in contrast to RDGB (Figure 5C).We also measured the PIP 2 levels at the plasma membrane of the rhabdomere and found that reconstitution of rdgB 9 with RDGB PAP could not rescue the PIP 2 levels back to that seen in wild type cells (Figure 5D and E).Together these findings suggest that endowing the LNS2 domain of RDGB with PA phosphatase activity, reduces the function of RDGB.

An unstructured region of RDGB supports the interaction between FFAT/dVAP-a
In order for the RDGB protein to function effectively at the ER-PM MCS, it is necessary for it to be docked, correctly oriented, to both the ER and PM faces of the ER-PM MCS.To understand the mechanism by which this interaction takes place, we performed protein docking studies between dVAP-A and RDGB in silico.A structure for dVAP-A was obtained by homology modelling using the previously determined crystal structure of VAP-A (Kaiser et al., 2005).We performed MDS for all the three models mentioned above for 100 ns and observed that the structures of the RDGB-VAP complex remains stable through a 100 ns simulation and do not show large deviations (Figure 6B (i), (ii) and (iii)).Each of the RDGB mutants was then used to perform docking studies with the structure of dVAP-A.The energies of the complexes where we observed interactions between FFAT motif and MSP domain were in the order: RDGB/dVAP-A (-276.46Kcal/ mol) < RDGB (DDHD-LNS2) D /dVAP-A (-234.99Kcal/mol) < RDGB (USR1-LNS2) D /dVAP-A (-228.28Kcal/mol) (Figure 6A).This clearly indicates that the full-length RDGB/dVAP-A complex is most stable compared to RDGB (DDHD-LNS2) D /dVAP-A and RDGB (USR1-LNS2) D /dVAP-A complexes.Binding free energy (mmpbsa) was calculated for the protein-protein complexes using PRODIGY algorithm (Vangone & Bonvin, 2015;Xue et al., 2016).The binding affinity (DG, Kcal/mol) for RDGB WT -VAP complex is À 14.2 and the Kd (M) is 3.6 e À 11 .The binding affinities for RDGB (DDHD-LNS2)D -VAP and RDGB (USR1-LNS2)D -VAP complexes are À 11.9 (Kd (M)¼ 1.8 e À 9 ) and À 11.4 (Kd (M)¼4.6eÀ 9 ) respectively.Importantly, the USR1 region (marked in yellow) between the FFAT motif and the DDHD domain (Figure 6A)], was observed to interact with dVAP-A in the RDGB/dVAP-A and RDGB (DDHD-LNS2)D/ dVAP-A complexes.Although, previous studies have shown that the interaction between the FFAT motif of RDGB and dVAP-A is important for the correct localization of RDGB to ER-PM MCS (Yadav et al., 2018), these data suggest the requirement for the USR1 of RDGB in facilitating this interaction.
To test the idea that additional regions C-terminal to the FFAT motif of RDGB are required to support the FFAT/VAP interaction, we made a series of progressively shorter deletions of RDGB starting from the C-terminus (Figure 1A).The interaction of full length RDGB with dVAP-A can be detected by immunoprecipitation [Figure 6C].We removed the C-terminal of the protein from just before the start of the DDHD domain (Figure 1A-RDGB (DDHD-LNS2)D ) leaving the USR1 after the FFAT motif intact.We expressed this protein in fly photoreceptors (rdgB 9 ; GMR>rdgB (DDHD-LNS2)D ) and found that RDGB (DDHD-LNS2)D also coimmunoprecipitates with dVAP-A (Figure 6C).Lastly, we expressed RDGB (USR1-LNS2)D and observed that it cannot interact with dVAP-A as detected in immunoprecipitation assays (Figure 6C).This suggests that the FFAT/dVAP-A interaction is supported by the unstructured region (USR1) present between the FFAT motif and the DDHD domain.

Orientation of RDGB at the ER-PM MCS
In transmission electron micrographs (TEM) of Drosophila photoreceptors, both the SMC and the microvillar PM can be clearly visualized (Figure 7 A (i), (ii) and (iii)).Using such TEM, we measured the thickness of the cytoplasmic gap between the SMC (ER) and the microvillar membrane (PM).These measurements from a large number of photoreceptors in wild type flies gives an average cytoplasmic gap of ca.12.5 nm at the ER-PM MCS (Figure 7B).We positioned our computed structural model of the RDGB/dVAP-A complex in this gap, constraining its orientation such that the VAP molecule was in contact with the ER membrane and the LNS2 domain was in contact with the microvillar PM through the helix that contains K1186 and K1187 (Figure 7C).This RDGB/dVAP-A model spans 11.9 � Å in its longest dimension as measured by the distance between the two farthest atoms in the protein.However, when we orient the protein   between the membranes with the constraints mentioned above, the distance spanned by the protein between the plasma membrane and the endoplasmic reticulum is 10.06 � Å (Figure 7C).The PITP domain is localized between the two regions that interact with PM and EM.It is connected to the rest of the protein by USR1 region at one end and FFAT motif at the other.

Discussion
In recent years a large number of PITPs have been described, many of them localized to membrane contact sites (Cockcroft & Raghu, 2018).In many cases, these PITPs are multi-domain proteins with the lipid transfer domain present in cis with several other domains [reviewed in (Raghu et al., 2021)].Although many of these domains have been studied individually using biochemical and cell biological approaches, the structure of a full length multi domain PITP protein that would help understand the orientation of the protein and interactions between the domains, leading to correct localization and function at an MCS has not been studied.Attempts to model the structure of PITPNM1, the vertebrate homolog of RDGB using alphafold (Jumper et al., 2021) have not been useful since the algorithm used here predicts a structure of which a large proportion is of low confidence and unstructured.
We built a 3-dimensionl model of RDGB using integrative modelling and constrained its orientation at the ER-PM MCS using experimentally determined ER and PM anchors.When positioned in this manner, the longest dimension of the computed, oriented structure fits snugly along the length of the experimentally measured cytoplasmic gap.In this setting, it is interesting to ask how the full length RDGB is positioned such that the PITPd can contact the ER membrane to extract lipid and deposit it at the PM.
For every protein localized to an ER-PM contact site, there must be a signal that anchors it to the ER side of the contact site and one that anchors it to the inner leaflet of the PM.Although previous experimental work had shown that the LNS2 domain of RDGB is important for PM localization, the molecular mechanism underlying this had not been known (Basak et al., 2021).Using MDS of the protein structure of RDGB, we identified two positively charged residues K1186 and K1187 that appeared to make contact with lipid molecules in the bilayer.When experimentally tested, these residues were necessary for PA binding, normal localization of RDGB to the MCS and function.Further by introducing PA phosphatase activity to the LNS2 domain of RDGB that could still bind PA, we noted that the RDGB protein was now mislocalized and loses function.This observation underscores the importance of PA binding to the LNS2 domain for the localization and function of RDGB.
The interaction between the FFAT motif and VAP has historically been considered as a bivalent reaction (Levine & Patel, 2016).In the case of RDGB too, it has previously been shown that an intact FFAT motif and dVAP-A are required for normal localization and physiological function (Yadav et al., 2018).In this study, we found that despite having an intact FFAT motif and normal levels of dVAP-A, an unstructured region (USR-1) just C-terminal to the FFAT motif and upstream of the DDHD domain was required for normal localization and function of RDGB.Our docking studies of RDGB with dVAP-A revealed that the underlying mechanism was that the USR1 region was required to stabilize this interaction.Proteomic studies have demonstrated that VAP-A interacts with a large number of proteins within cells and it is possible that additional interactions such as that with USR1 may modify the extent and priority of interaction of any given FFAT motif with VAP-A.Consistent with this, covalent modifications such as phosphorylation have recently been shown to modify the FFAT/VAP interaction in mammalian models (Di Mattia et al., 2020).
Historically the PITPd has been ascribed the function of PI transfer and indeed rdgB mutants can be rescued by reconstitution with the PITPd alone.How is the PITPd positioned in the context of the full-length protein to perform its lipid transfer function?In the full length structure elucidated in this study, the PITPd is set aside from the other domains and connected to the rest of the protein by USR1 and did not show any movements in the normal mode analysis.The transfer of lipid by the PITP domain would likely require large movements of the domain at the MCS so that the domain can interact with membrane on both the ER and PM side of the MCS.Such movements, though not captured in the present normal mode analysis might be triggered by signals generated during PLC signalling.One possibility is that USR1 which is just C-terminal to the PITPd acts as a hinge allowing it to flip between the two membranes.USR1 might be well suited to do this as it can undergo substantial conformational change as indicated by the large RMSD values for this domain in our model.A recent study has suggested a similar mechanism for OSBP (Jamecna et al., 2019).This is in contrast to an alternate mechanism of lipid transfer for LTPs that has recently been described for ESyt (Schauder et al., 2014) and VPS13 (Leonzino et al., 2021) where the multidomain LTP forms a channel through which lipids are proposed to move from one membrane to the other.
A recent bioinformatics survey for RDGB orthologs in eukaryotes reveals that PITPd occurs in conjunction with various domains in a range of genomes.The most common combination is the PITPd in conjunction with the DDHD and LNS2 domain.PITPd is also found in combination with the DDHD domain most of the times followed by a large variety of C-terminal domain.The domains might guide the localization and function of PITPd in eukaryotic cells (Raghu et al., 2021).In summary, we have developed a 3D model for RDGB and identified two novel intermolecular interactions that are required for correct localization and function of RDGB protein.The availability of the model will provide a framework for further studies to understand the structural basis of lipid transfer function in a multidomain PITP.
Video file 1: NMA analysis for RDGB protein: PITPd is marked in red, FFAT motif in green, USR1 in yellow, DDHD domain in blue and LNS2 domain in orange.The videos were generated using PYMOL (educational version) at a speed of 5 FS and with complete smoothening of the trajectory.The PITPd, FFAT motif and YW motif remain static throughout the

Figure 1 .
Figure 1.(A) Domain structure of RDGB and the lists of constructs used in this study.RDGB protein is 1284 amino acid long and contains three domains-PITPd, DDHD and LNS2, a FFAT motif and a large unstructured region (USR1).(B) RMSF plot for 100 ns MD simulation for RDGB wild-type protein homology model.(C)The Ramachandran plot values for model generated at each step.The X-axis denotes the number of residues in the model.The Y-axis represents the % of residues in the Ramachandran plot (allowed regions, additionally allowed regions and disallowed regions).(D) The ProsA Z-score validation for model generated at each step.The X-axis denotes the number of residues in the model.The Y-axis denotes the Z-score (model number 1-. 1-300 residues.2-300-420 residues, 3-420-600 residues, 4-600-700 residues, 5-700 to 800 residues, 6-800-900 residues, 7-900 -1059 residues, 8-1059 -1200 residues, 9 -1200-1284 residues and 10-full length model after energy minimization).(E) Domain movements in RDGB full length protein-The normal mode analysis movement static image to show movement of DDHD domain towards the LNS2 domain in the lowest frequency mode.

Figure 2 .
Figure 2. (A (i) and (ii)) The minimum distance between RDGB protein (system 1) and the DPPC membrane during the 100 ns molecular dynamics simulation (averaged over replicates) is represented.The X-axis represents simulation time (ns) and Y-axis represents distance between the atoms of the protein and membrane ( � Å) The graphs were generated using Graphpad PRISM.The minimum distance between the residues K1186 (A(iii)) and K1187 (A (iv)) of RDGB and the DPPC molecule of the membrane is represented.The X-axis represents simulation time (ns) and Y-axis represents distance between atoms ( � Å). (2B (i) and (ii)) The minimum distance between RDGB KK/AA (system 2) and the DPPC membrane during the 100 ns molecular dynamics simulation (averaged over replicates) is represented.The minimum distance between the residues A1186 (B (iii)) and A1187 (B (iv)) of RDGB KK/AA and the DPPC molecule of the membrane is represented (averaged over replicates).(2 C (i) and (ii)) The minimum distance between RDGB LNS2D (system 3) and the DPPC membrane during the 100 ns molecular dynamics simulation is represented.The graphs were generated using Graphpad PRISM.The X-axis represents simulation time (ns) and Y-axis represents distance between atoms ( � Å).

Figure 4 .
Figure 4. (A) Confocal images showing the localisation of endogenous RDGB protein in the photoreceptors of 1-day-old dark reared flies of the indicated genotypes probed with antibody against RDGB.rdgB 9 photoreceptors show no staining with RDGB antibody.Phalloidin marks F-actin staining and highlights rhabdomeres, R1-R7.Scale bar-5 lm.RDGB KK/AA mutant shows a diffused localization.(B) Representative ERG traces of the indicated genotypes; the duration of the light pulse is shown.X-axis indicates time in msec and Y-axis indicates the average ERG amplitude in mV.(C) Quantification of the electrical response of photoreceptors to light.Average of the peak amplitude of ERG recorded from 1 day old dark reared flies.Each data point represents an individual fly tested (N ¼ 8).(D) Representative deep pseudopupil images of PIP 2 levels in the microvillar membrane of photoreceptors.The fluorescence of the PH-GFP probe is depicted.Genotypes as indicated.(E) Quantification of the mean fluorescence intensity of the PIP 2 probe PH-GFP from the deep pseudopupil formed by one day old flies of the indicated genotypes (N ¼ 10).(F) Quantification of the time course taken for retinal degeneration.10 ommatidia were scored from 5 flies of each genotype and plotted.The quantification shows number of rhabdomeres that are intact at different timepoints post eclosion.(G) Representative optical neutralization (ON) images showing rhabdomere structure of the indicated genotypes.Rearing conditions and the age of the flies are indicated on top.( � CL-Constant Light).Data information: Scatter plots and XY plots with mean ± SD are shown.Statistical tests: (b (ii) and d (i)) Student's unpaired t test.(c (ii)) Two-Way ANOVA Grouped analysis with Bonferroni post-tests to compare replicate means.�� p value < 0.01; ���� p value < 0.0001.

Figure 5 .
Figure 5. (A) Alignment of LNS2 domain containing proteins (LIPIN 1, RDGB (D. melanogaster, RDGB (C.elegans, Nir 2 (H.sapiens), Nir 1 (H.sapiens) and Nir 3 (H.sapiens) to show conservation of the residue important for phosphatase activity.(B) Confocal images showing the localisation of endogenous RDGB protein in the photoreceptors of 1-day-old dark reared flies of the indicated genotypes probed with antibody against RDGB.rdgB 9 photoreceptors show no staining with RDGB antibody.RDGB PAP mutant shows a diffused localization.Phalloidin marks F-actin staining and highlights rhabdomeres, R1-R7.Scale bar-5 lm.(C) Representative ERG traces of the indicated genotypes; the duration of the light pulse is shown.X-axis indicates time in msec and Y-axis indicates the average ERG amplitude in mV (D) Quantification of the electrical response of photoreceptors to light.Average of the peak amplitude of ERG recorded from 1 day old dark reared flies.Each data point represents an individual fly tested.(E) Representative deep pseudopupil images of PIP 2 levels in the microvillar membrane of photoreceptors.The fluorescence of the PH-GFP probe is depicted.Genotypes as indicated.(F) Quantification of the mean fluorescence intensity of the PIP 2 probe PH-GFP from the deep pseudopupil formed by one day old flies of the indicated genotypes (N ¼ 10).One-way Anova with Turkey's multiple comparison test was used to test statistical significance.�� p value < 0.01; ���� p value < 0.0001.

Figure 7 .
Figure 7. (A) TEM of Drosophila photoreceptors (i) Cross section through a single ommatidium showing the cell body (c), nucleus (n) and rhabdomere (r) (ii) Cross section view of a single photoreceptor.The microvillar plasma membrane/rhabdomere (r) and the sub-microvillar cisternae (SMC) stained in black are shown (iii) Magnified view of PM-SMC distance (B) Quantification of the PM-SMC distance from three individual flies is shown.Each point represents independent PM-SMC distance measurement.(C) Model of the RDGB protein depicted in relation to the PM and ER at the contact site.The gap between the ER and PM is the mean of the measurements in B. The protein model is also drawn to the same scale.

Table 1 .
Details on modelling protocol for RDGB protein.