Lock and chop: A novel method for the generation of a PICK1 PDZ domain and piperidine‐based inhibitor co‐crystal structure

Abstract The membrane protein interacting with kinase C1 (PICK1) plays a trafficking role in the internalization of neuron receptors such as the amino‐3‐hydroxyl‐5‐methyl‐4‐isoxazole‐propionate (AMPA) receptor. Reduction of surface AMPA type receptors on neurons reduces synaptic communication leading to cognitive impairment in progressive neurodegenerative diseases such as Alzheimer disease. The internalization of AMPA receptors is mediated by the PDZ domain of PICK1 which binds to the GluA2 subunit of AMPA receptors and targets the receptor for internalization through endocytosis, reducing synaptic communication. We planned to block the PICK1‐GluA2 protein–protein interaction with a small molecule inhibitor to stabilize surface AMPA receptors as a therapeutic possibility for neurodegenerative diseases. Using a fluorescence polarization assay, we identified compound BIO124 as a modest inhibitor of the PICK1‐GluA2 interaction. We further tried to improve the binding affinity of BIO124 using structure‐aided drug design but were unsuccessful in producing a co‐crystal structure using previously reported crystallography methods for PICK1. Here, we present a novel method through which we generated a co‐crystal structure of the PDZ domain of PICK1 bound to BIO124.


Introduction
Protein interacting with kinase C1 (PICK1) is a membrane protein composed of a N-terminal PDZ domain and a C-terminal BAR domain, and is expressed highly in the brain and peripheral tissues. 1,2 It functions as an intracellular trafficking protein involved in the internalization of amino-3hydroxyl-5-methyl-4-isoxazole-propionate (AMPA) type receptors which are localized on the surface of neurons and mediate synaptic communication. 3,4 Down regulation of surface AMPA receptors on neurons reduces synaptic communication leading to cognitive impairment in progressive neurodegenerative diseases such as Alzheimer disease. 5 GluA2 is a subunit of AMPA type receptors and is composed of an extracellular N-terminal ligand binding domain, a transmembrane domain, and an intracellular C-terminal domain. The PDZ domain of PICK1 recognizes and binds to the intracellular PDZ binding motif on the c-terminus of GluA2. The PICK1-GluA2 protein-protein interaction leads to the trafficking of AMPA type receptor away from the neuron surface through endocytosis, thereby reducing synaptic communication. 6,7 We proposed to develop an inhibitor targeting the PICK1 PDZ domain-GluA2 interaction using structure-aided drug design for the treatment of neurodegenerative diseases through the stabilization of surface AMPA type receptors on neurons. PDZ domains provide a somewhat selective recognition function that serves to assemble protein complexes by binding PDZ binding motifs on target proteins. PDZ binding motifs are short, generally four amino acid stretches at the c-terminus of a target protein. The nomenclature for PDZ peptide binding motifs classifies the last residue of the cterminus as the 0-position and the residues nterminal to the 0-position as negative positions.
Specificity between PDZ domains and the PDZ binding motifs are imparted by the 22 position of the motif, and the nature of the residue at this position is what classifies PDZ binding motifs into class I, class II, or class III motifs. 8,9 Class I sequence motifs have S/T-X-U-COOH; class II sequence motifs have U -X-U -COOH, and class III sequence motifs have D/E-X-U-COOH; where U is a hydrophobic residue and X is any amino acid. This selectivity for PDZ binding motif classes has been attributed to the interaction between the 22 position of the PDZ binding motif with the first amino acid side chain on the aB helix (aB1) of the PDZ domain. 10 PICK1's PDZ domain is unique in its ability to bind both class I and II motifs, however, it binds class II motifs (i.e., GluA2 (SVKI)) with a 10-foldhigher affinity over class I motifs (i.e., PKCa (QSAV)). [9][10][11] Our goal was to disrupt the proteinprotein interaction between PICK1 and the higher affinity class II GluA2 PDZ binding motif. We used a fluorescence polarization assay for our high throughput screen (HTS) efforts to identify small molecule inhibitors that displaced a labeled GluA2 peptide from PICK1. We identified a novel class of piperidine-based inhibitors that led us to BIO124 [ Fig. 1(A)] a sub-lM molar inhibitor of the PICK1-GluA2 interaction.
In this study, we wished to understand the binding mode of the PICK1-GluA2 inhibitor BIO124 by crystallography and NMR. Since the apo-PICK1 PDZ domain was sparingly soluble, we were unable to form a complex with BIO124 suitable for structural studies. Here, we report a novel "lock and chop" method for producing a co-crystal structure of PICK1's PDZ domain with the small molecule inhibitor BIO124 which enabled our understanding of its binding mode.

Results and Discussion
Developing a method to co-crystallize PICK1 PDZ domain with BIO124 To enable a structure-aided drug design approach to improve the potency of BIO124, we wished to understand the interaction of the inhibitor to the PDZ domain of PICK1 using x-ray crystallography. We designed the PDZ domain of PICK1 for crystallization using the self-binding extension method described previously for PICK1 and other PDZ domains. 12 We fused the class 1 PDZ binding motif QSAV from PKCa to the C-termini of the PICK1 PDZ domain to enhance solubility and aid crystallization. The purified PICK1 PDZ-QSAV protein eluted as a dimer on size exclusion chromatography under reducing conditions indicating that we successfully designed an intramolecular interaction between the PDZ domains using the fused QSAV extension. Since the construct contained two reactive cysteines, we confirmed that the dimer could be "locked" by disulfide bridging when oxidized. 10 The reaction was 90% complete after addition of H 2 O 2 as assessed by nonreducing SDS-PAGE and comparing the intensity of the monomer to dimer bands. The disulfide-bridged PICK1 PDZ-QSAV protein could be concentrated to 10 mg/mL and produced diffraction quality crystals in 2 days. This "locked" version of the PICK1 PDZ-QSAV dimer crystallized much more readily than the unlocked reduced dimer. We hypothesize that the oxidized dimer is more stable to aggregation as seen from its increased thermal stability compared to the unlocked reduced dimer (Supporting Information Fig. S1).
Once we established that we could produce a crystallization-competent PICK1 PDZ-QSAV dimer, this protein was used for co-crystallization with BIO124. The structure revealed electron density that confirmed that the dimer was locked by two disulfide bounds between C44(A)-C46(B) and C44(B)-C46(A) between the two PDZ domain chains of PICK1 (Supporting Information Fig. S2). However, because electron density for the fused QSAV tail from the dimeric partner peptide was visible in the peptide groove, this method failed in allowing BIO124 to displace the QSAV tail within the crystal lattice.
Comparison of class I (QSAV) and class II (SVKI) peptide binding to PICK1 The crystal structure that was obtained in the attempt to co-crystallize with BIO124 represents the first PICK1 structure bound to a class I PDZ binding motif (QSAV from PKCa). We wished to better understand the differences in the 10-fold higher binding affinities for class II vs. class I motifs to PICK1 by comparing our PICK1 PDZ domain class I motif structure with structures of PICK1 PDZ domain bound to class II PDZ motifs from the PDB. 10,[12][13][14] The superposition of class II peptide bound PICK1 PDZ-3C-VYGIESVKI crystal structure (PDB code 3HPK) with our class I peptide bound PICK1 PDZ-QSAV demonstrated few differences in the sidechain positions of the residues surrounding the peptide binding groove [ Fig. 2(A)]. However, there were two hydrophobic "hotspots" that differ between the two peptides and could explain the differences in binding affinities [ Fig. 2(B,C)]. The lysine on the GluA2 class II SVKI peptide buries into the S -1 subpocket where it is involved in lipophilic interactions with Leu32 and Phe53. The class 1 PKCa QSAV peptide has an alanine at the 21 position and is unable to make this interaction. The second difference is between the polar and nonpolar sidechains at the 22 position. The polar serine of the class I QSAV does not make an interaction to K83 in the aB1 position in PICK1. In the class II SVKI, the hydrophobic valine sidechain at 22 position forms a lipophilic interaction with the aliphatic segment of the Lys83 sidechain in the S -2 subpocket. These two additional lipophilic interactions support the observations that class II peptides demonstrate higher binding affinities for PICK1 over class I peptides.
This figure also includes an iMolecules 3D interactive version that can be accessed via the link at the bottom of this figure's caption. To determine if BIO124 could reasonably compete with the neighboring QSAV tail in solution, we used two solution-based methods to confirm that the compound was likely binding in the peptide binding groove for this construct. The first method used a protein-based 2D NMR approach to confirm that the compound can compete out the QSAV for the peptide binding pocket. Lysine methylation was used to probe the effects of compound binding. Seven lysines are available for methylation in the PICK1 PDZ domain, of which three are near (less than 6Å ) the binding pocket [ Fig. 3(A)]. Upon binding to the GluA2 peptide, three peaks in the 13 C-HSQC showed chemical shift changes [ Fig. 3(B)] which we reasoned to be the three closest lysines (K27, K83, and K88 to the peptide binding pocket of PICK1 [ Fig. 3(A)]). Upon addition of BIO124 the same lysines' exhibit chemical-shift changes [ Fig. 3(C)]. The peaks that shift upon titration of the GluA2 peptide and BIO124 are the same, suggesting that the GluA2 peptide and BIO124 bind to the same peptide binding groove, confirming that this construct is capable of binding the compound and displacing the peptide extension in solution.
The second method that we used to confirm displacement of the neighboring QSAV peptide from the peptide binding pocket of PICK1 was through partial proteolysis. BIO124 was added at a 4.5-fold molar excess to the "locked" PICK1 PDZ-QSAV dimer to make the peptide extension accessible to bovine carboxypeptidase A (bCPA). Mass spectrometry showed that a 16-hour digestion with catalytic amounts of bCPA resulted into >95% removal of the 0 and 21 position residues of the QSAV tail [Supporting Information Fig. S3(A,B)]. In contrast, in the same experiment in the absence of the compound, no proteolysis took place [Supporting Information Fig. S3(C)]. This suggests that the compound is in fact displacing the peptide in solution, thereby making it accessible to bCPA cleavage.
We hypothesized that if we could exploit the proteolysis method to irreversibly prevent the QSAV peptide from rebinding to the carboxylate binding loop, we could shift the binding equilibrium from the peptide bound form to the BIO124 bound form for crystallization trials. We hypothesized that if we used PICK1 PDZ-QSAV crystals as seeds, we could grow diffraction-quality crystals of the proteolyzed PICK1-PDZ-QS BIO124 complex. Interestingly, if the dimer was not locked by oxidation, no crystals were observed, suggesting that the dimer packing needed to be maintained to obtain quality crystals.
The crystal structure of PICK1-PDZ-QS/BIO124 was solved to 1.75 Å and is similar to the previously solved PICK1-QSAV dimer structure except for the peptide binding region. The electron density begins at Val19 and ends at Leu104, and stops just prior to the C-terminal extension. The structure has one dimer molecule per asymmetric unit as described in the QSAV-structure [ Fig. 4(A)]. The F o -F c electron density maps clearly showed that BIO124 was present in molecule A and B [ Fig. 4(B)], and thus a "lock and chop" method through crosslinking and proteolysis successfully produced a compound-bound PDZ domain co-crystal structure for PICK1.
BIO124 binds in an extended conformation at the surface of the peptide binding pocket As expected, BIO124 is anchored into the peptide binding pocket through several hydrogen bonds [ Fig.  4(C)]. Its carboxylic acid is involved in an intricate network of hydrogen bonds to the backbone of three amino acids (Ile33, Gly34, and Ile35) of the carboxylate binding loop. There is also a water-mediated interaction to the backbone carbonyl of Asn31 and the side chain amine group of Lys27. In addition, the peptide linker of BIO124 hydrogen bonds to the top and bottom of the pocket through backbone interactions with the amide of Gln91 and carbonyl of Ile35. The cyclopropyl group makes lipophilic interactions with Ile37, Ala87, and Ile90 within the S 0 sub-pocket [ Fig. 4(D)]. The piperidine core occupies the 21 position and provides two vectors for additional lipophilic interactions. The pyrolidine is positioned to stack between Leu32 and Phe53 in the S -1 sub-pocket and the bromophenyl moiety travels from the 22 position toward the aliphatic segment of K83 in S -2 sub-pocket. The piperidine group of Bio124 is important to make multiple contacts with PICK1, the lowest energy conformer places the bromophenyl in the equatorial position which forces the amide into the axial position and allows the piperidine to make hydrophobic contacts underneath the carboxylate binding loop as it reaches into the 21 position sub-pocket. This extended binding mode shows that additional interactions can be made outside the peptide binding pocket which could be optimized to improve potency and/or selectivity for this class of inhibitors.
On comparing the binding mode of BIO124 to PICK1 with the published small molecule inhibitors bound to their PDZ structures including AF6  [15][16][17] In all three cases, the inhibitors form hydrogen bond interactions with the carboxylate binding motif and fill the S 0 sub-pocket making the lipophilic interactions which anchor them into the peptide binding pocket. Yet, all three compounds from these known small molecule/PDZ structures exhibit modest inhibition potencies at 50 lM, 10.7 lM, 17.2 lM for AF6, Dvl, and Shank3, respectively [ Fig. 5(D)]. [15][16][17] Of the three, Sulindac bound to Dvl is the only inhibitor that binds into the S -2 sub-pocket, by making a hydrogen bond to Arg332 of its aB helix. None of the molecules extend very far into their respective S -1 sub-pocket. In contrast, BIO124 extends into the S -1 sub-pocket and uses its pyrolidine moiety to form lipophilic interactions with Leu32 of the carboxylate binding loop and This figure also includes an iMolecules 3D interactive version that can be accessed via the link at the bottom of this figure's caption. Phe53. This leads us to suggest that the binding affinities of these ligands for their respective PDZ domains could be improved by making additional interaction at the surface of the peptide binding pocket either by extending into the S -1 sub-pocket or toward the aB1 position as seen for BIO124.

Future direction in generating PDZ small molecule co-crystal structures using the lock and chop method
Elkins et al. demonstrated that the addition of a selfbinding C-terminal PDZ binding motif to PDZ domains can enhance solubility and promote crystallization. 12 This methodology has proven useful in generating novel PDZ domain crystal structures and mapping interactions with PDZ binding motifs. In this study, we used this method to generate a PICK1 PDZ domain structure in complex with the class I PDZ binding motif peptide from PKCa. In addition, by using a Cterminal proteolysis step to liberate the C-terminal QSAV extension from the peptide binding site, we achieved the co-crystallization of the PICK1 PDZ domain with BIO124. The crosslinking of the dimer through cysteine oxidation was critical for locking and maintaining the dimeric state of the protein for crystallization. We hypothesize that this method could be used to enable other PDZ domain small molecule inhibitor co-crystal structures. Here, after a selfbinding C-terminal PDZ binding motif extension has been identified for producing a crystal structure, site directed mutagenesis could be used to introduce interactions to lock the PDZ domain into its crystallization competent form. Once locked, small molecule inhibitors targeting the peptide binding site could be introduced to liberate the C-terminal peptide extension for proteolytic digestion to produce a PDZ small molecule inhibitor complex to enter crystallization trials.
In summary, we report a novel method for producing co-crystal structures of the PDZ domain of PICK1 with a small molecule inhibitor. Our PICK1 PDZ BIO124 co-crystal structure demonstrates a new extended compound binding mode in which lipophilic interactions are made with Lys83 at the aB1 position and in the 21 position sub-pocket spanning the peptide binding site of PICK1.

Materials and Methods
Cloning, protein expression, and purification PICK1 PDZ-QSAV was constructed as follows: cDNA for human PICK1 encoding residues 1 to 105 were cloned into a pET15b vector in frame with a TEV cleavable N-terminal HIS tag and the Cterminal fused QSAV tail (PICK1 PDZ-QSAV). BL21 (DE3) Escherichia coli cells transformed with PICK1 PDZ-QSAV were grown at 378C in LB media supplemented with ampicillin to an OD600 of 1, at which point the temperature was reduced to 188C and protein expression was induced with 1 mM IPTG. After 16 hours, the cells were harvested and resuspended in lysis buffer (25 mM TRIS-HCl pH 8.0, 250 mM NaCl, 10% (v/v) glycerol, 2.5 mM b-mercaptoethanol (b-ME), 20 mM imidazole, and Roche EDTA-free protease inhibitor cocktail) and subjected to two passes through a microfluidizer. The lysate was clarified by centrifugation at 20,000g for 1 hour at 48C, and PICK1 PDZ-QSAV was captured by batch binding to nickel resin for 16 hours at 48C. The nickel resin was washed with buffer A (25 mM TRIS-HCl pH 8.0, 250 mM NaCl, 10% (v/v) glycerol, 2.5 mM b-ME, and 20 mM imidazole) and then loaded into an XK16/20 column and washed to baseline on an AKTA purifier. PICK1 PDZ-QSAV was eluted from the nickel column using buffer A supplemented to 250 mM imidazole and analyzed by SDS-PAGE. The

Discovery of BIO124
HTS screening was done using a Fluorescent Polarization (FP) assay with the PICK PDZ-BAR construct as described in Alfonso et al. 5 After several rounds of hit optimization, compound BIO124 was identified with an IC50 of 513 nM in the FP assay [ Fig. 1(B)].

PICK1 PDZ-QSAV inter-disulfide bridge formation
PICK1 PDZ-QSAV concentrated to 3 mg/mL was treated with 2.5 mM H 2 O 2 and incubated for 16 hours at 48C to form intermolecular disulfide bridges which was expected to lock the protein into a dimeric form. Dimer formation was confirmed by nonreducing SDS PAGE. Disulfide-linked dimeric PICK1 PDZ-QSAV was separated from higher order PICK1 PDZ-QSAV species using a Superdex 75 column equilibrated in buffer C: 25 mM TRIS-HCl pH 8.0, 250 mM NaCl, and 5% (v/v) glycerol and concentrated to 3 mg/mL.

PICK1 PDZ-QSAV differential scanning fluorimetry
The thermal denaturation (Tm) curves for the unlocked and locked dimer of PICK1 PDZ-QSAV were determined using a QuantStudio 12K Flex QPCR instrument. The assay was carried out in 25 mM TRIS-HCl pH 8.0, 250 mM NaCl, 5% (v/v) glycerol, and 1x Protein Thermal Shift dye (Life Technologies) with the PICK1 PDZ-QSAV proteins at 1 mM. The proteins were denatured from 20 to 858C with a 18C/minute gradient and fluorescence measured every 0.5 seconds.

Crystallization of dimeric PICK1 PDZ-QSAV in the presence of BIO124
Dimeric PICK1 PDZ-QSAV at 3 mg/mL (220 lM) was incubated with 4.5 molar equivalents of BIO124 for 2 hours on ice and then co-concentrated to 10 mg/mL for crystallization. Crystals of dimeric PICK1 PDZ-QSAV grew from 0.1 M BisTRIS pH 5.5 and 25% PEG3350 (w/v) at 48C in 2 days. Crystals were cryoprotected in the mother liquor supplemented with 20% (v/v) glycerol prior to being frozen in liquid nitrogen.
Data collection and structure determination of the PICK1 PDZ QSAV X-ray diffraction data for crystals of dimeric PICK1 PDZ-QSAV with BIO124 was measured using a Rigaku FRE (Rigaku, The Woodlands, TX) and was processed with HKL2000. 18 The crystals belonged to a P3 2 space group with one disulfide-linked PICK1 PDZ-QSAV dimer per asymmetric unit. The structure was solved with MOLREP using a PICK1 PDZ domain structure (PDB code 2GZV) in which the last 4 residues were removed. 12,19 The initial model was subjected to several rounds of refinement and model building using Refmac5 and Coot. 20 Dimeric bCPA-treated PICK1 PDZ-QS BIO124 cocrystallization bCPA-treated PICK1 PDZ-QS BIO124 complex was concentrated to 10 mg/mL. PICK1 PDZ BIO124 crystals were grown by seeding the drop with PICK1 PDZ-QSAV crystals in 0.1 M BisTRIS pH 5.5 and 25% PEG3350 (w/v) at 48C. Crystals of the PICK1 PDZ BIO124 complex were cryoprotected for data collection by transferring them to 0.1 M BisTRIS pH 5.5 and 25% PEG3350 (w/v) and 20% glycerol prior to being frozen in liquid nitrogen.
Data collection and structure determination of bCPA-treated PICK1 PDZ-QS BIO124 complex X-ray diffraction data for crystals of dimeric bCPAtreated PICK1 PDZ-QS BIO124 was measured at beam line 31ID at the Argonne Photon Source and was processed with HKL2000. 18 The crystals belonged to a P3 2 space group with one disulfidebridged PICK1 PDZ BIO124 dimer per asymmetric unit. The structure was solved with MOLREP using the PICK1 PDZ-QSAV structure with the AV removed as the search model. 19 The structure went through multiple rounds of refinement and model building using Refmac5 and Coot which led to a model with R work of 13.9% and R free of 15.7% to 1.75 Å with good geometry (Table 1). 20,21 The structure has been deposited with the PBDID: 6BJO.