An unexpected INAD PDZ tandem-mediated plcβ binding in Drosophila photo receptors

INAD assembles key enzymes of the Drosophila compound eye photo-transduction pathway into a supramolecular complex, supporting efficient and fast light signaling. However, the molecular mechanism that governs the interaction between INAD and NORPA (phospholipase Cβ, PLCβ), a key step for the fast kinetics of the light signaling, is not known. Here, we show that the NORPA C-terminal coiled-coil domain and PDZ-binding motif (CC-PBM) synergistically bind to INAD PDZ45 tandem with an unexpected mode and unprecedented high affinity. Guided by the structure of the INAD–NORPA complex, we discover that INADL is probably a mammalian counterpart of INAD. The INADL PDZ89 tandem specifically binds to PLCβ4 with a mode that is strikingly similar to that of the INAD–NORPA complex, as revealed by the structure of the INADL PDZ89–PLCβ4 CC-PBM complex. Therefore, our study suggests that the highly specific PDZ tandem – PLCβ interactions are an evolutionarily conserved mechanism in PLCβ signaling in the animal kingdom.


Introduction
Scaffold proteins can serve as platforms for the assembly of signaling components into macromolecular complexes, targeting them to specific cellular localizations, as well as actively modulating signaling processes (Bhattacharyya et al., 2006;Pawson and Nash, 2003;Zhang and Wang, 2003). They therefore support the occurrence of signaling events in precise locations and at specific time points in different tissues. The Drosophila compound eye rhodopsin-mediated photo-transduction signaling process is highly elaborate. It is also one of the best-studied model systems showing how light signals, via INAD scaffold-organized signaling complexes, can be transduced at a very large dynamic range with extremely rapid kinetics and intricate regulatory mechanisms (Huber, 2001;Li and Montell, 2000;Liu et al., 2011;Mishra et al., 2007;Tsunoda et al., 1997;Tsunoda and Zuker, 1999).
At the inner surface of fly photoreceptor rhabdomeric membranes, the master scaffold protein INAD (encoded by inaD for inactivation no after potential D and contains 5 PDZ domains arranged in tandem; Figure 1A) forms stoichiometric multi-molecular complexes with phospholipase Cb (NORPA), Ca 2+ -permeable transient receptor potential (TRP) channel and eye-specific protein kinase C (eye-PKC) (Adamski et al., 1998a;Chevesich et al., 1997;Huber et al., 1996;Kimple et al., 2001;Liu et al., 2007;Montell, 2005;Peng et al., 2008;Shieh and Zhu, 1996;Tsunoda et al., 1997;van Huizen et al., 1998;Ye et al., 2016). Genetic, cell biology, biochemistry, and structural biology studies have revealed that INAD PDZ2 is required for binding to eye-PKC, INAD PDZ3 for interaction with TRP channel, and INAD PDZ5 for engaging NORPA (Adamski et al., 1998b;Chevesich et al., 1997;Tsunoda et al., 1997;Ye et al., 2016). Mutations of inaD that lead to disruption of each of these interactions invariably impair fly photo signal transduction (Scott and Zuker, 1998). Fly photoreceptorsare capable of responding, via the INAD-mediated assembly of the signaling complex, to light signals with extremely fast response time and termination kinetics (Henderson et al., 2000;Ranganathan et al., 1995). One critical step in fast light signaling in fly photoreceptors is the efficient coupling of NORPA, one of the fastest enzymes catalyzing PtdIns(4,5) P 2 hydrolysis, to the other INAD-organized signaling components (Minke and Parnas, 2006;Shieh et al., 1997). Previous genetic studies have revealed that a mutation within PDZ5 of INAD (inaD 2 ) or a single point mutation within the PBM of NORPA could selectively impair the INAD-NORPA interaction, indicating that the fifth PDZ domain of INAD and the PBM of NORPA are required for INAD-NORPA interaction (Cook et al., 2000;Shieh et al., 1997;  1997). The direct interaction between NORPA and PDZ5 is, however, extremely weak (K d~5 60 mM) (Liu et al., 2011) and is unlikely to be capable of supporting the specific interaction between INAD and NORPA. It is possible that the INAD-NORPA interaction follows a mode that is entirely different from all known PDZ domain-mediated target-recognition modes.
In this study, we show that the PDZ45 tandem of INAD functions as a supramodule binding to the entire C-terminal coiled-coil domain and PDZ-binding motif of NORPA (CC-PBM) with an unexpectedly high affinity. The crystal structure of INAD PDZ45 in complex with NORPA CC-PBM uncovers a highly unusual PDZ domain -target binding mode and explains the high binding affinity and specificity between INAD and NORPA. Guided by the INAD-NORPA complex structure, we discover that in the vertebrate system, INADL but not MUPP1, a close paralogue of INADL, specifically binds to PLCb4. Our biochemical and structural studies demonstrate that INADL PDZ89 forms a supramodule and binds to the entire C-terminal coiled-coil domain of PLCb4 in a manner that is strikingly similar to that of the INAD-NORPA complex. The striking similarity between the PDZ scaffold and PLCb interactions at the molecular level might point to an evolutionarily conserved molecular adaption in PLCb signaling in the animal kingdom.

NORPA CC-PBM synergistically interacts with the INAD PDZ45 tandem with a very high affinity
Before performing detailed biochemical characterization of the INAD-NORPA interaction, we carefully analyzed the NORPA protein sequence and found that the C-terminal region of NORPA contains a coiled coil (CC) (residues 850-1086) immediately followed by a PDZ binding motif (PBM) (residues 1092-1095, amino acid 'EFYA') ( Figure 1B). The predicted coiled-coil and PBM are separated by just six amino acid residues, suggesting the possibility that the coiled coil may cooperate with the PBM for NORPA to bind to INAD. To test this hypothesis, we characterized the INAD-NORPA interaction quantitatively using purified proteins ( Figure 1C and D). Isothermal titration calorimetry (ITC) analysis revealed that the full-length INAD binds to NORPA CC-PBM with a K d~8 nM ( Figure 1D and Figure 1-figure supplement 1B), an affinity that is much stronger than those of all known canonical PDZ-target interactions (usually with K d of a few to a few tens of mMs) (Ye and Zhang, 2013). Further mapping showed that INAD PDZ45 is the minimal and complete region of INAD in binding to NORPA CC-PBM, as the PDZ45 tandem binds to NORPA CC-PBM with an affinity similar to that of the full-length INAD ( Figure 1C1 and D). Deletion of the coiled-coil domain from the NORPA CC-PBM decreased its binding to INAD PDZ45 by~2400 fold, and removal of the three-residue PBM completely eliminated NORPA CC-PBM's binding to INAD PDZ45 ( Figure 1D and Figure 1-figure supplement 1A and C), indicating that the coiled coil and PBM of NORPA function synergistically in binding to INAD PDZ45. We showed earlier that PDZ45 forms a supramodule that is necessary for fly visual signaling (Liu et al., 2011). Consistent with this finding, isolated PDZ5 displayed a weak binding to NORPA CC-PBM ( Figure 1D and Figure 1-figure supplement 1D). Taken together, the above biochemical results revealed that the INAD PDZ45 tandem binds to NORPA CC-PBM with a very high affinity. It is noted that the binding affinity between INAD PDZ45 and NORPA CC-PBM is among the tightest in all known PDZ-target interactions (Ye and Zhang, 2013), and that the involvement of a coiled-coil structure in binding to PDZ domains is highly unusual and previously unknown.

Structural characterization of the NORPA-INAD complex
Having characterized the detailed interaction between NOROA and INAD, we wanted to understand the molecular mechanism governing this specific binding by trying to solve the atomic structure of the INAD PDZ45 and NORPA CC-PBM complex using X-ray crystallography. Multiple years of attempts to crystallize the complex prepared from the wildtype INAD PDZ45 and NORPA CC-PBM all failed, even though our NMR-based study indicated that the PDZ45-NORPA CC-PBM complex is conformationally homogeneous in solution ( Figure 2-figure supplement 1A). Confronted by this disappointing result, we decided to take a step back and tried to solve the NORPA CC-PBM structure alone first, hoping to be able to get some clues that might lead us to a way to obtain PDZ45-NORPA CC-PBM complex crystals (the structure of INAD PDZ45 was solved in our earlier study (Liu et al., 2011)).
Using lysine reductive-methylated protein, we were able to solve the structure of NORPA CC-PBM at the resolution of 3.0 Å (Supplementary file 1 and Figure 2-figure supplement 2A). Like the coiled-coil structure of mammalian PLCb3 (Lyon et al., 2013), NORPA CC folds into an elongated three helix bundle structure with an N-terminal short helix coupling to the middle region of the helix bundle ( Figure 2-figure supplement 2A). The C-terminal 16 residues (aa 1080-1095) that include the PBM is disordered and invisible in the structure (Figure 2-figure supplement 3C). The packing surface of NORPA CC-PBM crystals involves an extensive surface composed of a1-a4 ( Figure 2-figure supplement 2C). We reasoned that this packing surface might overlap with the INAD PDZ45 binding surface, and therefore prevented the same surface from being used for crystal packing in the PDZ45-NORPA CC-PBM complex (also see Figure 2-figure supplement 2D). We noted that the solvent-exposed face of the a2 helix is lined with eight Lys residues, which are all evolutionarily conserved and required for the membrane binding of NORPA CC (Figure 2-figure supplement 2B, and see below for more details) (Lyon et al., 2013). We substituted the eight Lys residues with Ala residues (designated as CC 8KA -PBM) with the rationale that such substitutions should not impair the helical conformation of a2 but might create a new crystal packing surface for the PDZ45-NORPA CC-PBM complex ( The most important feature of the complex is that both PDZ domains of the PDZ45 tandem and both the coiled coil and PBM of NORPA participate cooperatively to form a highly stable INAD-NORPA complex burying an extensive interaction surface area of~1373 Å 2 ( Figure 2A). The PDZ45-NORPA CC-PBM complex structure also reveals a hitherto uncharacterized interaction mode for PDZ domains. PDZ45 binding stabilizes the conformation of NORPA's C-terminal 16 amino acids by extending the a4 helix for five more turns, immediately followed by a four-residue b-strand corresponding to the PBM of NORPA ( Figure 2B and  Figure 2C) (Liu et al., 2011).
The binding interface of the PDZ45-NORPA CC-PBM complex can be divided into three distinct sites ( Figure 2D): the PDZ5-PBM interaction site (site 1, Figure 2E), the PDZ5/CC packing site (site 2, Figure 2F), and the PDZ4/CC binding site (site 3, Figure 2G). In site 1, NORPA PBM ('EFYA') inserts into the bB'/aB' groove of PDZ5 following the typical PDZ-ligand interaction mode ( Figure 2B and E) (Ye and Zhang, 2013). We note that the last residue of the NORPA PBM is Ala, but Ala at the 0 position is very rare for canonical PBMs (Ye and Zhang, 2013). Structural analysis revealed that the pocket of PDZ5 that accommodates the sidechain of the 0 position residue of the NORPA PBM is shallow and enriched in bulky aromatic residues (F632, F649, F642 and F-2, Figure 2E and  For example, Y-1 from the NORPA PBM interacts with W944 from a3 of the NORPA CC domain; E-3 from the PBM forms hydrogen bonds with Y521 from bB of PDZ4 and with Y1085 from a4 of the NORPA CC; R952 from a3 of NORPA forms salt bridges with D610 from bC' of PDZ5 ( Figure 2F). In site 3, two His residues from INAD PDZ4 form hydrogen bonds with E1081 from a4 of the NORPA CC, and these residues are further stabilized by a hydrogen bond with W956 from a3 of the NORPA CC ( Figure 2G). Mutations of individual residues in these binding sites invariably weakened the NORPA-INAD interaction ( Figure 2H and The NORPA CC-PBM-INAD PDZ45 complex structure provides a clear mechanistic explanation of why the inaD 2 mutant (a missense mutation changing Gly605 in PDZ5 to Glu) led to mislocalization of NORPA in photoreceptor cells and to severe defects in photo-response amplitude and kinetics (Cook et al., 2000;Tsunoda et al., 1997). G605 is in a hydrophobic core that couples PDZ45 into a structural supramodule (Figure 2-figure supplement 5B). The G605E mutation would introduce a negatively charged sidechain to this hydrophobic core, and thus is expected to destabilize the PDZ45 supramodule and impair its binding to NORPA. Consistent with this analysis, the PDZ45 G605E protein showed significantly decreased binding to NORPA (Figure 2-figure supplement 5C), showing that the structural integrity of the INAD PDZ45 supramodule is important for photo signaling in Drosophila photopreceptors. To further confirm this, we generated a T669E mutant of PDZ45, which was previously found to impair conformational coupling between PDZ4 and PDZ5 (Liu et al., 2011). Pulldown analysis showed that the T669E mutant of INAD PDZ45 had significantly weaker binding to NORPA CC-PBM than wildtype INAD PDZ45 (Figure 2-figure supplement 5D), further indicating that the formation of the PDZ45 supramodule is important for the strong binding of INAD to NORPA.

INAD-mediated membrane micro-domain organization of the fly photosignal transduction complex
A number of studies have suggested that the PLCb CC domain is important for the membrane targeting of the enzyme, an event that is necessary for the full activity of the enzyme in vivo Lyon et al., 2013;Park et al., 1993;Waldo et al., 2010). We next analyzed whether the formation of the tight complex between INAD PDZ45 and the NORPA CC-PBM might influence the membrane binding of the NORPA CC. The evolutionarily conserved, positively charged residues in the a2 helix of the NORPA CC are aligned on a flat and exposed surface of the domain ( Figure 3A and Figure 2-figure supplement 2B). A liposome sedimentation-based assay showed that the NORPA CC-PBM strongly interacted with lipid membranes ( Figure 3B1 and C). By contrast, the NORPA CC 8KA -PBM showed significantly decreased membrane-binding capacity, indicating that the positively charged a2 helix is essential for membrane binding ( Figure 3B2 and C). In the NORPA-INAD complex, INAD PDZ45 binds to one face of the NORPA CC with the two PDZ domains crossing over the a2, a3, and a4 helices ( Figure 2A). Interestingly, one end of PDZ5 (formed by the bA/bB-loop, the bC/aA-loop, and the aB/bF-loop) is positively charged and aligned with the NORPA CC a2 helix forming a flat surface ( Figure 3A), indicating that binding to INAD PDZ45 does not interfere with the membrane binding of the NORPA CC-PBM mediated by its a2 helix. Consistent with this structural analysis, the INAD PDZ45-NORPA CC-PBM complex displayed comparable (or somewhat enhanced) lipid membrane binding when compared with the NORPA CC-PBM alone ( Figure 3B3 and C). Our previous study showed that INAD PDZ3 tethers the TRP channel through specific and strong binding to the TRP channel C-terminal tail (Ye et al., 2016). Therefore, the INAD PDZ345 tandem can simultaneously recruit NORPA and the TRP channel, forming a closely arranged signaling membrane micro-domain ( Figure 3D). It is envisioned that the physical positioning of NORPA right next to the TRP channel allows the channel to sense efficiently the membrane lipid component changes initiated by light-induced phosphatidylinositol 4,5-bisphosphate (PIP 2 ) hydrolysis by NORPA. It is noted that eye protein kinase C (eye-PKC) is also recruited to this membrane micro-domain through binding to the PDZ2 domain of INAD ( Figure 3D).
An important question is how the nano-molar strong interaction between INAD and NORPA is regulated in vivo during the deactivation of the photo-transduction process. Previous studies showed that light-induced acidification that results from the NORPA-mediated hydrolysis of PIP 2 can lead to conformational uncoupling of the INAD PDZ45 tandem (Huang et al., 2010;Liu et al., 2011). Interestingly, we found that decreasing the pH of the buffer from 7.8 to~5.8 dramatically weakened the binding between INAD and NORPA by~6700-fold ( Figure 3E), supporting the notion that the light-induced acidification of the rhabdomeric compartment could dissociate the INAD-NORPA complex.

PLCb4 binding to INADL and NORPA binding to INAD occur through similar modes
Next, we wanted to test whether the unexpected INAD PDZ tandem-mediated NORPA binding mode could also occur for the mammalian PLCb enzymes. No mammalian homologue of INAD has been identified, although mammals do contain several multi-PDZ-domain proteins including one called INAD-like (INADL, also known as PATJ), which contains a total of 10 PDZ domains. A previous study using microarray-based technique found that, upon exposure to light, the expression of the INADL and Slc9a3r1 (encoding NHERF1) genes were upregulated in mice lacking both rods and cones (Peirson et al., 2007). Recent human genetic studies of sleeping-disorder patients showed that INADL might be associated with multiple sleeping disorders and circadian timing variations (Forni et al., 2014;Jones et al., 2016;Lane et al., 2017). Considering that melanopsin is known to be a sleep modulator, and that melanopsin polymorphisms have been associated with circadian dysfunction (Lee et al., 2014;Roecklein et al., 2012), we hypothesized that INADL might function as a scaffold protein in the melanopsin-mediated intrinsically photosensitive retinal ganglion cells (ipRGC) in mammals.
PLCb4, a vertebrate orthologue of NORPA in Drosophila, was previously found to be a key regulator of melanopsin signaling in ipRGC (Xue et al., 2011). Amino-acid sequence analysis revealed that the CC-PBM domains of NORPA and PLCb4 share a sequence identity of 27% ( Figure 4A). More importantly, PLCb4 shares the following features with the NORPA CC-PBM that are important for both INAD PDZ45 and lipid membrane binding: (i) PLCb4 also contains eight Lys residues in the predicted a2 helix with positions identical to those in NORPA (highlighted with green dots in Figure 4A); (ii) PLCb4 also contains a highly conserved PBM with the sequence 'ATVV' (highlighted with blue dots in Figure 4A), the PLCb4 PBM also immediately follows the a4 of the predicted coiled-coil domain ( Figure 4A and B), indicating that the coiled coil and PBM of PLCb4 may also function synergistically in binding to PDZ scaffold proteins; and (iii) a number of residues from NORPA CC domains that directly interact with INAD PDZ45 also exist in the same positions in the PLCb4 CC (highlighted with black dots in Figure 4A). Therefore, we hypothesized that PLCb4 may also bind to certain PDZ scaffold protein(s) in the same way that NORPA does to INAD.
To test this hypothesis, we screened for possible PDZ proteins that may interact with the PLCb4 CC-PBM with high affinity using quantitative ITC-based binding assay coupled with size exclusion chromatography using purified proteins. We focused our efforts on multi-PDZ mammalian scaffold proteins that either share certain similarity to INAD (e.g., have five or more PDZ domains) or have been indicated to be related to visual signaling. This narrowed our targets to the following four proteins: INADL, MUPP1, GRIP1, and NHERF. We used multiple PDZ domains from these proteins (i.e. two or more consecutive PDZ domains) to test their possible binding to the PLCb4 CC-PBM (Figure 4-figure supplement 1). Among the numerous PDZ tandems tested, we found that only the INADL PDZ89 tandem interacted with the PLCb4 CC-PBM with a strong affinity (Kd~0.2 mM, Figure 4B-D). The other purified proteins either did not bind or bound with very low affinities to the PLCb4 CC-PBM (Figure 4-figure supplement 1), indicating that the INADL PDZ89 and PLCb4 interaction is very specific. Further mapping of the interaction revealed that, as in the NORPA CC-PBM-INAD PDZ45 interaction, both the CC and the PBM of PLCb4 are required for its binding to INADL PDZ89 ( Figure 4D and Figure 4-figure supplement 2). Conversely, isolated PDZ8 or PDZ9

Structural characterization of the PLCb4 CC-PBM-INADL PDZ89 complex
To elucidate the molecular mechanism governing the interaction between PLCb4 and INADL, we determined the crystal structure of the INADL PDZ89-PLCb4 CC-PBM complex. To facilitate the complex crystal growth, we also substituted the eight Lys residues in the a2 helix of PLCb4 CC with Ala ( Figure 4A). ITC analysis showed that the PLCb4 CC 8KA -PBM binds to INADL PDZ89 with the same affinity as WT PLCb4 CC-PBM ( Figure 4D and Figure 4-figure supplement 2E). Furthermore, we discovered that covalently linking INADL PDZ89 with the PLCb4 CC 8KA -PBM was necessary to obtain diffraction-quality crystals of the complex. The complex structure was determined by the single-wavelength anomalous dispersion method using gold derivatives at a resolution of 2.8 Å (Supplementary file 1). Strikingly, the overall architecture of the PLCb4 CC 8KA -PBM-INADL PDZ89 complex is remarkably similar to that of the NORPA-INAD complex ( Figure 5A-C). In the complex, PDZ89 forms an integral structural unit and extensively binds to the CC-PBM domain of PLCb4, burying a total of~1211 Å 2 surface area ( Figure 5A). Structural analysis showed that the PDZ tandems and the corresponding contact region on PLCb of the two complexes could be nicely aligned ( Figure 5C, dashed box). In the two complexes, INADL PDZ9 can be nicely aligned with INAD PDZ5, and INADL PDZ8 needs to be rotated by~43 degrees clockwise to superimpose with INAD PDZ4 ( Figure 5D-F). The secondary structural elements that are involved in the inter-PDZ domain packing of the two PDZ tandems are quite similar ( Figure 5D vs Figure 5E; Figure 5-figure supplement 3). The conformations of the C-terminal half of a3 and the N-terminal half of a4 of the two CC-PBMs are very different. The three helices (a2, a3, and a4) of the PLCb4 CC form a flat sheet. By contrast, the a3 helix of the NORPA CC has an obvious kink in the center, resulting in a large curvature of the a3/a4 coiled coil ( Figure 5C).
In the complex, the formation of a supramodule between PDZ8 and PDZ9 is mediated by a number of hydrophobic residues from both domains ( Figure 5H). NMR-based analysis showed that, before binding to the PLCb4 CC-PBM, there is minimal directly coupling between PDZ8 and PDZ9 (see Figure 5-figure supplement 1A for details). This result indicates that, unlike the tight coupling between PDZ4 and PDZ5 in INAD (Liu et al., 2011), the direct coupling between PDZ8 and PDZ9 is induced by the binding of PLCb4. Mutations of the residues involved in PDZ89 inter-domain coupling led to decreased binding of the PDZ tandem to PLCb4 CC-PBM ( Figure 5I and Figure 5-figure supplement 2B). In INAD PDZ45, the PDZ5 C-terminal extension tucks into a groove at the rear of PDZ4, facilitating the formation of the PDZ45 supramodule (Liu et al., 2011). In INADL PDZ89, although the C-terminal extension of PDZ9 could not be resolved in the complex crystals, deletion of the C-terminal 10-residue extension of PDZ9 also weakened PDZ89's binding to the PLCb4 CC-PBM (last row of Figure 5I, marked as 'DCT'), suggesting that the PDZ9 C-terminal extension may also participate in the formation of the PDZ89 supramodule when in complex with PLCb4.

Discussion
In this study, we discover that the INAD PDZ45 supramodule binds to the CC-PBM domain of NORPA with very strong binding affinity. The strong binding between INAD and NORPA requires the direct conformational coupling of PDZ4 and PDZ5 (i.e. formation of the PDZ45 supramodule) of INAD as well as precise spacing between the coiled-coil domain and the PBM of the NORPA. Such conformational arrangement provides an elegant way to synergistically integrate multiple interaction sites, each with relatively weak binding affinity, to form the very strong overall INAD-NORPA interaction. In addition, the stringent conformational requirement of both proteins for the tight binding also means that the interaction between INAD PDZ45 and the NORPA CC-PBM is highly specific. Finally, the sensitivity of the binding to the conformation of both INAD PDZ45 and the NORPA CC-PBM suggests that the interaction between INAD and NORPA may be regulated by altering the conformation of one or both proteins.
Earlier studies showed that, upon light activation, NORPA can rapidly hydrolyze PIP 2 and cause transient and localized acidification of microvilli, an event that can open the TRP channel (Huang et al., 2010) and disrupt the domain coupling between PDZ4 and PDZ5 of INAD PDZ45 (Liu et al., 2011). We demonstrated in this study that a mild acidification of the binding buffer from neutral pH to a mild acidic pH of 5.8 was sufficient to weaken the binding between INAD PDZ45 and NORPA CC-PBM dramatically ( Figure 3E). Therefore, it is likely that the light-induced transient acidification of microvilli can trigger conformational uncoupling of the INAD PDZ45 supramodule, which acts as a negative feedback mechanism for rapid dissociation of NORPA from INAD and subsequent signal termination of each light-induced signaling cycle in Drosophila compound eyes. Therefore, INAD not only functions as a passive scaffold in assembling various components into a large signaling complex, but is also capable of actively modulating light signal transductions by regulating the dynamic organization of the entire light signaling complex.
Combination of the available genetic and biochemical data in the literature and the findings presented here leads us to propose an atomic model depicting the INAD-organized signaling complexbeneath the Drosophila rhabdomeric membranes ( Figure 3D). In the dark, the INAD PDZ345 tandem assembles a high-affinity, stoichiometric complex with the TRP channel and NORPA that is located right beneath the rhabdomeric membranes, thereby positioning NORPA in very close proximity to the TRP channel ( Figure 3D). In addition, eye-PKC is also tethered to NORPA and TRP by binding INAD PDZ2. It is envisioned that, in this condition, the TRP channel is at its greatest sensitivity and efficiently senses the membrane lipid component changes initiated by light-induced PIP 2 hydrolysis by NORPA. In line with this model, disruption of the interaction between INAD and NORPA, either by mutation of INAD PDZ45 (in the inaD 2 mutant flies) or by mutation of NORPA (Y-1 mutation), is known to cause severe functional defects in signal sensitivity and in kinetics in the photon-transduction pathway (Cook et al., 2000;Shieh et al., 1997;Tsunoda et al., 1997). Once the light activates the signaling cascade, the transient local acidification induced by the NORPAmediated hydrolysis of PIP 2 can rapidly promote the structural uncoupling of PDZ45, thereby leading to the dissociation of the NORPA-INAD complex and termination of the light signal.
Different from mammal rod and cone opsin signaling, melanopsin photo-transduction is analogous to the Drosophila photo-transduction cascade. Activated by light, melanopsin can interact with a Gq/G11-type G protein (Graham et al., 2008), which in turn activates PLCb4 (a vertebrate orthologue of NORPA in Drosophila) (Jiang et al., 1994;Lee et al., 1994). PLCb4 hydrolyzes PIP 2 and thereby generates inositol 1,4,5 triphosphate (Ins(1,4,5)P 3 ) and diacylglycerol (DAG), which may ultimately modulate ion channels such as the TRPC6 or TRPC7 ion channels (Xue et al., 2011). Protein kinase C zeta (PKCz), which is analogous to the eye-PKC in flies, may influence TRP channel activity via kinase-dependent phosphorylation(s) (Peirson et al., 2007). However, the molecular components and their action mechanisms in the ipRGCs photo signaling processes are far from certain. Given the sequence similarity between NORPA and PLCb4, we speculate and also provide biochemical evidence to suggest that PLCb4 may use a binding mode similar to that used by NORPA in ipRGC signaling. Biochemical characterizations revealed that PLCb4 can specifically interact with PDZ89 of INADL. The complex structure of PLCb4-INADL demonstrates that PLCb4 interacts with the INADL PDZ89 supramodule with a mode very similar to that of the NORPA-INAD complex. Future studies will be required to test whether the INADL-PLCb4 interaction identified in this study might be functional in ipRGC signaling or in PLCb4 signaling in other tissues.
Finally, as a technical note, we struggled for a long time to obtain optimal crystals of the INAD-NORPA complex for structural analysis. It was a bitter but ultimately sweet lesson to learn that systematic substitution of eight Lys residues on one helix of the coiled coil of NORPA by Ala produced diffraction-quality crystals due to introduction of a new crystal packing surface. The elongated conformation of coiled coils provides fewer favorable surface areas for crystal packing, and thus coiled coils are often more difficult to crystalize than to globular proteins. To assist coiled-coil protein crystallization, protein engineering tricks such as surface entropy reduction, site-directed mutations (Wine et al., 2009), construct optimization to deplete the disordered sequences (Zhou et al., 2017), introducing new packing surface by replacing charged amino acid acids with small hydrophobic and helix favoring Ala (this study) may be useful strategies. Cloning and constructs for recombinant protein expression cDNA encoding PDZ5 (residues 580-665) and PDZ45 (residues 473-674) were PCR amplified from Drosophila melanogaster inaD and cloned into a modified version of the pET32a vector.

Materials and methods
The NORPA CC-PBM was PCR amplified from the cDNA of Drosophila NORPA (from Drosophila Genomics Resource Center, NCBI reference sequence: NP_525069.2, 1095aa) and cloned into a modified version of the pET32a vector containing an N-terminal 6 Â His tag. PLCb4 cDNA encoding residues 912-1175 was PCR amplified from mouse PLCb4 plasmid (Genebank: NM_013829.2). INADL cDNA encoding PDZ89 (residues 1421-1625) was PCR amplified from human INADL (Genebank: NM_176877.3). Various mutations of PLCb4 or INADL89 were generated using the standard PCR-based method and confirmed by DNA-sequencing. MUPP1 PDZ10-11 cDNA encoding residues 1610-1802 was PCR amplified from mouse MUPP1 (NM_001305284.1). INADL PDZ89 was fused to the N-terminus of PLCb4 CC-PBM with a 15aa linker (GGGGSGGGGSGGEGS) to generate a fusion protein used for crystallization.

Protein expression and purification
Recombinant proteins were expressed in Escherichia coli BL21 (DE3) or Rosseta (DE3) host cells at 16˚C. The His-tagged fusion proteins were purified by Ni 2+ -NTA-Sepharose (GE) affinity chromatography followed by size-exclusion chromatography (Superdex 200 column from GE Healthcare) in the final buffer of 50 mM TrisÁHCl, 1 mM DTT, 1 mM EDTA, pH 7.5, and 100 mM NaCl. When needed, the N-terminal His-tagged was cleaved by protease 3C and removed by another step of size-exclusion chromatography. To obtain highly purified PLCb4 CC-PBM, a cation-exchange chromatography was used after the size-exclusion chromatography. Uniformly 15 N-labeled INAD PDZ45 and INADL PDZ89 were prepared by growing bacteria in M9 minimal medium using 15 NH 4 Cl (Cambridge Isotope Laboratories Inc.) as the sole nitrogen source. Se-Met-labeled NORPA CC-PBM was prepared by expressing the protein in B834 (DE3) host cells grown in Se-Met supplemented M9 minimal medium at 37˚C for~13 hr and purified by size-exclusion chromatography coupled with cationexchange chromatography as described above.
Purified isolated NORPA CC-PBM protein (1 mg/ml in a buffer of 50 mM HEPES pH 7.6, 100 mM NaCl, 1 mM DTT and 1 mM EDTA) was reductive methylated for crystallization. Stock solutions of 1 M dimethylamine-borane complex (ABC) and 1 M formaldehyde in water were freshly prepared and kept at 4˚C or on ice (Walter et al., 2006). To start the methylation reaction, 20 mL ABC and 40 ml formaldehyde were added to each ml of NORPA CC-PBM solution, and the reactions were continued for 2 hr at 4˚C before another 20 mL ABC and 40 ml formaldehyde was added to the reaction mixture. The reaction was continued for another 2 hr, and finally an additional 10 ml of ABC was added and the reaction was allowed to continue overnight at 4˚C. To quench the reaction, the reaction mixture was passed through a size-exclusion column equilibrated with a 50 mM Tris buffer (pH 7.5) containing 100 mM NaCl, 1 mM DTT and 1 mM EDTA.
Diffraction data were collected at the Shanghai Synchrotron Radiation Facility (BL17U or BL19U1) at 100 K. Data were initially processed and scaled using HKL2000 or HKL3000 (Otwinowski and Minor, 1997). The data from the PLCb4 CC 8KA -PBM-INADL PDZ89 complex were further corrected for anisotropy using the diffraction anisotropy server (https://services.mbi.ucla.edu/anisoscale/) and truncated to 3.0 Å , 3.1 Å and 2.8 Å along the a, b and c axes, respectively. In the structure determinations processes, different strategies were used for different crystals. The Se-Met NORPA CC-PBM K884E & K898E and PLCb4 CC 8KA -PBM-INADL PDZ89 complex structures were determined using the single-wavelength anomalous dispersion (SAD) method. The Se or gold sites were found by SHELXD (Dall'Antonia et al., 2003). Subsequent site refinement, phase calculation, density modification and initial model building were carried out with Autosol (Terwilliger et al., 2009). Molecular replacement was carried out for the methylated NORPA CC-PBM and the NORPA CC 8KA -PBM-INAD PDZ45 complex, using the N-terminal halves of the NORPA CC-PBM K884E & K898E structure and the INAD PDZ45 structure (PDB ID: 3R0H) as the search models for PHASER (McCoy et al., 2007). Subsequent model building and refinement for all of the four different structures were completed iteratively using COOT (Emsley et al., 2010) and PHENIX (Adams et al., 2010). The final models were validated by MolProbity . The final refinement statistics are summarized in Supplementary file 1. All structure figures were prepared by PyMOL (www.pymol.org). The sequence alignments were prepared and presented using ClustalW (Larkin et al., 2007) and ESPript (Robert and Gouet, 2014), respectively. The structure factors and the coordinates of the structures reported in this work have been deposited to PDB under the accession codes of 6IRB for the Se-Met NORPA CC-PBM with K884E and K898E substitutions; 6IRC for the methylated NORPA CC-PBM; 6IRE for the NORPA CC 8KA -PBM-INAD PDZ45 complex; and 6IRD for the PLCb4 CC 8KA -PBM-INADL PDZ89 complex.

Isothermal titration calorimetry assay
ITC measurements for INAD-NORPA interactions were carried out on VP-ITC or ITC200 Microcal calorimeters (Microcal) at 25˚C. For INADL-PLCb4 interactions, ITC measurements were carried out on an ITC200 Microcal calorimeter (Microcal) at 16˚C. The titration buffer contained 50 mM Tris-HCl (pH 7.5), 1 mM DTT, 1 mM EDTA, and 100 mM NaCl. Each titration point was performed by injecting a 10 mL (for VP-ITC) or 2.5 mL (for ITC200) aliquot of a protein sample from a syringe into a protein sample in the cell at a time interval of 120 s to ensure that the titration peak returned to the baseline. The titration data were analyzed by Origin7.0 (Microcal).

NMR spectroscopy
NMR samples contained 0.2 mM of INAD PDZ45 or INADL PDZ89 in 50 mM Tris-HCl (pH 7.2, with 1 mM DTT and 1 mM EDTA) in 10% D 2 O. NMR spectra were acquired at 30˚C on Varian Inova 750-or 800-MHz spectrometers, each equipped with an actively z-gradient shielded triple resonance probe. The backbone resonance assignment of INAD PDZ45 was obtained using the data from our previous study (Liu et al., 2011).

Pull-down assay
HEK293T cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), and 1% of penicillin-streptomycin at 37˚C with 5% CO 2 . Cells were tested negative for mycoplasma contamination by cytoplasmic DAPI staining. The cell line was only used for heterologous protein expression, so no further authentication was performed. After reaching 70-80% confluency, cells were transfected with ViaFect transfection reagent (Promega) using GFP control, GFP-INAD PDZ45 (aa 473-674), or the G605E and T669E mutants of INAD PDZ45. The cells were collected 48 hr after transfection. HEK293T cell lysate expressing GFP control, GFP-INAD PDZ45, or the G605E and T669E mutants of GFP-INAD PDZ45 was incubated with 0.5 nmol of purified His-StrepII-tagged NORPA CC-PBM (aa 863-1095) for 40 mins at 4˚C. Each mixture was incubated with 20 ml StrepTactin Sepharose High Performance slurry beads (GE Healthcare) for another 1 hr at 4˚C before pelleting by centrifugation. After washing twice with buffer (50 mM Tris (pH 7.5), 100 mM NaCl, 1 mM EDTA and 5 mM DTT), the captured proteins in the pellets were eluted by boiling with 2 Â SDS loading buffer, resolved by 12% SDS-PAGE, and detected by immunoblotting with anti-GFP antibody (SANTA CRUZ, GFP-B2: sc-9996).

Fluorescence assay
Fluorescence assays were performed on a PerkinElmer LS-55 fluorimeter at 25˚C. In the assay, FITC (Molecular Probes)-labeled peptide samples were titrated with binding partners in 50 mM Tris buffer (with pH value and DTT concentration specifically indicated) containing 50 mM NaCl and 1 mM EDTA. The titration curves were fitted with the GraphPad Prism five software package.

Analytical gel-filtration chromatography
Protein samples (typically 100 mL at a concentration of 20 mM pre-equilibrated with the column buffer) were injected into an AKTA FPLC system with a Superose 12 10/300 GL column (GE Healthcare) using the column buffer of 50 mM Tris-HCl (pH 7.8), 1 mM DTT, 1 mM EDTA, and 100 mM NaCl.

Liposome-binding assay
The total bovine brain lipid extracts (Folch fraction I, Sigma B1502) were re-suspended by sonication in a buffer containing 50 mM Tris (pH 7.5), 100 mM NaCl, 1 mM EDTA, and 1 mM DTT) at 10 mg/ ml. The sedimentation-based assay followed the method described earlier (Wen et al., 2008). The liposome solution was pre-cleared by centrifugation at 20,817 g, 4˚C for 2 min and the protein samples were also pre-cleared at 200,000 g, 4˚C for 30 min. The protein samples (5 mM) were incubated with different concentrations of liposomes in 40 ml of buffer for 20 mins at room temperature and then centrifuged at 100,000 g, 4˚C for 30 min in a Beckman TLA100.1 rotor. The supernatants were transferred into a new tube to determine the amount of proteins that did not bind to the liposomes. The pellets were washed twice with buffer and re-suspended in a volume of buffer equal to that of the supernatants. Finally, the proteins recovered from the supernatants and pellets were analyzed by SDS-PAGE.