Biochemical investigations of the mechanism of action of small molecules ZL006 and IC87201 as potential inhibitors of the nNOS-PDZ/PSD-95-PDZ interactions

ZL006 and IC87201 have been presented as efficient inhibitors of the nNOS/PSD-95 protein-protein interaction and shown great promise in cellular experiments and animal models of ischemic stroke and pain. Here, we investigate the proposed mechanism of action of ZL006 and IC87201 using biochemical and biophysical methods, such as fluorescence polarization (FP), isothermal titration calorimetry (ITC), and 1H-15N HSQC NMR. Our data show that under the applied in vitro conditions, ZL006 and IC87201 do not interact with the PDZ domains of nNOS or PSD-95, nor inhibit the nNOS-PDZ/PSD-95-PDZ interface by interacting with the β-finger of nNOS-PDZ. Our findings have implications for further medicinal chemistry efforts of ZL006, IC87201 and analogues, and challenge the general and widespread view on their mechanism of action.

potentially relevant for future drug discovery efforts, we decided to examine it further with techniques not previously applied for either ZL006 or IC87201, namely fluorescence polarization (FP), isothermal titration calorimetry (ITC) and 1 H-15 N HSQC NMR. These methods are known to be very reliable for investigating ligand-protein interactions. We hoped to provide evidence for the reported inhibitory properties of the nNOS-PDZ/PSD-95-PDZ interaction via interaction with the β -finger, enabling us to explore these compounds further using medicinal chemistry approaches. However, our results robustly demonstrate that neither ZL006 nor IC87201 directly interacts with extended nNOS-PDZ or any of the PSD-95 PDZ domains in vitro. Neither do they inhibit the formation of the nNOS-PDZ/PSD-95-PDZ complexes. Therefore, we believe the correct mechanism of action of ZL006 and IC87201 has yet to be revealed, but hopefully our findings will stimulate further investigations into these biological active and efficient compounds.

Results
Fluorescence Polarization Assay. ZL006 and IC87201 were expected to bind nNOS-PDZ at its β -finger. However, to ascertain that the published biochemical results could not be explained by inhibition of canonical PDZ domain-mediated interactions, we checked the potential affinities of ZL006 and IC87201 towards PDZ1, PDZ2, PDZ3 of PSD-95 and nNOS-PDZ in our well-established FP assay 16,32,33 . Fluorescent probes were made by attaching fluorophores (Cy5) to C-terminal 10-or 11-mer (C10 or C11) peptide ligands representing canonical and natural occurring protein partners (Supplementary Table S1). Saturation curves were then created by applying increasing concentrations of PDZ (typically 0-150 μ M) to a low (5-50 nM) and fixed concentration of probe, and K d values between probe and PDZ were determined 16,32,33 . The affinities for both ZL006 and IC87201 towards the PDZ domains were investigated in an inhibition FP assay using constant concentrations of probe and PDZ and increasing concentrations of test compound or inhibitory control peptide (Supplementary Table S1). The results clearly showed that ZL006 and IC87201 at concentrations reaching 500-1800 μ M did not inhibit any of the probe-PDZ interactions involving PDZ1, PDZ2, PDZ3 of PSD-95 or nNOS-PDZ, while all control peptides demonstrated expected potencies 16,32 (Fig. 2). Thus ZL006 and IC87201 do not bind the canonical PDZ ligand binding sites.
Next, we designed a novel 'direct' FP assay in order to measure if ZL006 and IC87201 could inhibit the interaction between nNOS-PDZ and PSD-95 PDZ domains (PDZ1, PDZ2, PDZ1-2). Inhibitory effects in this assay would support the proposed mechanism of ZL006 and IC87201 binding to the β -finger of nNOS-PDZ and allosterically inhibit the nNOS-PDZ/PSD-95-PDZ interactions. The assay was established by creating a fluorescent version of nNOS-PDZ by TAMRA-labelling nNOS-PDZ via an introduced cysteine residue (V36C). K d values for PDZ1 and PDZ2 towards this TAMRA-nNOS were measured by saturation curves to be 2.4 μ M and 1.1 μ M, respectively (Fig. 3a), which correlate with the literature 20,34 . PDZ1-2 was found to interact with TAMRA-nNOS with high affinity (K d = 0.15 ± 0.03 μ M) (Fig. 3a); while TAMRA-nNOS and PDZ3 did not bind each other (data not shown), as also reported by others 34 . We then tested ZL006 and IC87201 for their ability to inhibit the TAMRA-nNOS/PSD-95-PDZ interactions. ZL006 revealed no activity in this assay involving any of the PSD-95 PDZ domains when tested at up to 1200 μ M ( Fig. 3b-d). This demonstrates that ZL006 does not inhibit the nNOS-PDZ/ PSD-95-PDZ interaction, and indicates that ZL006 does not perturb the nNOS β -finger. Control compounds, N-cyclohexylethyl-ETAV 16,32 and Tat-N-dimer 11 , clearly inhibited the interaction as expected, as they bind the canonical PDZ pockets of PDZ1 and PDZ2 and displace TAMRA-nNOS ( Fig. 3b-d).
Unfortunately, IC87201 showed high degree of fluorescence-based artefactual signal when using TAMRA-nNOS as probe. This is a common problem for fluorescence-based assays in relation to small molecules 35,36 , and it prevented reliable testing of IC87201 above 7.3 μ M. However, up to 7.3 μ M no inhibition of TAMRA-nNOS/PSD-95-PDZ interactions could be measured (data not shown).
In a final attempt to provide evidence for the expected mechanism of ZL006 and IC87201 using FP, we designed an 'indirect FP assay' measuring the abilities of the compounds to inhibit the nNOS-PDZ/ PSD-95-PDZ interaction in the presence of a PSD-95-PDZ-binding probe. First we tested a series of 23 fluorescent-labelled C-terminal peptide ligands (Supplementary Table S1) derived from PDZ-interacting proteins in order to find a probe with good activity towards the PSD-95 PDZ domains and no or little activity against nNOS-PDZ ( Supplementary Fig. S1). The Cnskr2 probe fulfilled these selectivity criteria and was used to generate saturation curves with PDZ2 and PDZ1-2 of PSD-95 (Fig. 4a). The Cnskr2/PDZ interactions were subsequently inhibited with unlabelled nNOS-PDZ or the control peptide GluN2B-C5 representing the C-terminal pentapeptide of the NMDA receptor subunit GluN2B, and it was seen that nNOS-PDZ inhibited the Cnskr2/PDZ interactions potently demonstrating its high affinity to both PDZ2 and PDZ1-2 of PSD-95 (Fig. 4b). Next, fixed concentrations of TAMRA-Cnskr2, PSD-95-PDZ2/1-2, and nNOS-PDZ were applied, corresponding to the conditions at IC 50 of the nNOS-PDZ inhibition curve (Fig. 4b) where sensitivity of the assay is greatest. If the compounds inhibited the nNOS-PDZ β -finger-mediated interaction to PDZ2 or PDZ1-2 of PSD-95, they would liberate PSD-95 PDZ domains from nNOS-PDZ thereby enabling PSD-95 PDZ domains to interact with the peptide probe (TAMRA-Cnskr2) resulting in a rise in FP signal. However, when testing ZL006 under these conditions we observed no rise in FP signal; only a flat line was generated (Fig. 4c). This indicates, again, that ZL006 is not able to inhibit nNOS-PDZ/PSD-95-PDZ interactions. Finally, we established the assay using Cy5-probes instead of TAMRA-probes, in order to diminish artefactual FP signal when testing IC87201.
In addition, we investigated if the compounds inhibited the α 1-Syntrophin-PDZ/nNOS-PDZ interaction, also mediated by the nNOS β -finger. Here, we used the Sapk3-peptide as it has the desired high affinity to α 1-Syntrophin-PDZ and low affinity to nNOS-PDZ ( Supplementary Fig. S1). However, also under these conditions we observed no activity of ZL006 or IC87201 when tested up to 500-600 μ M (Fig. 4d). Overall, our results indicate that neither ZL006 nor IC87201 inhibit the nNOS-PDZ β -finger-mediated interactions to PDZ domains of PSD-95 and α -Syntrophin.
Isothermal Titration Calorimetry. Unlike the FP experiments above, which rely on inhibition of interaction partners, the ITC method is a direct measure of binding, regardless of where it takes place on the protein. We, therefore, titrated ZL006 and IC87201 into solutions of either nNOS-PDZ or PSD-95-PDZ2 to detect heat changes in order to follow potential specific ligand-protein interactions. The data conclusively showed that IC87201 does not bind either the extended nNOS-PDZ or PSD-95-PDZ2 ( Fig. 5a-b). Titration of ZL006 into these proteins gave some heat change but the binding was too weak to produce a binding isotherm that would yield a reliable K d value ( Fig. 5c-d). An experiment performed for the binding between extended nNOS-PDZ and PSD-95-PDZ2 showed that these two proteins bind with K d ~ 0.67 μ M (Fig. 5e). In the presence of 1 mM IC87201, the affinity between nNOS-PDZ and PSD-95-PDZ2 appeared to be affected although the compound also caused significant interference with the data, reducing its reliability (Fig. 5f). Similar experiments to investigate if ZL006 decreases the affinity between the two PDZ domains also suffered from interference and, hence, deemed unreliable for further interpretation. Overall, ITC indicated that ZL006 and IC87201 do not interact directly or specifically with nNOS-PDZ or PSD-95-PDZ2, nor do they drastically disrupt the complexes once they are formed.
Nuclear Magnetic Resonance. 1 H-15 N HSQC NMR is another direct method for detecting ligand-protein interactions. However, when we tested IC87201 for binding to the extended nNOS-PDZ , none of which appear to be from residues Asp62, Leu107, Phe111 and Arg121, which are residues proposed to bind or to be affected by ligand-binding 24,25 . Typically, chemical shift perturbations of >0.1 ppm and up to 0.3 ppm are observed on binding even very weak ligands with K d ~ 300-500 μ M 37 . Here, the maximum shift changes observed were less than 0.05 and are, therefore, considered non-specific. The data for ZL006 were not conclusive since ZL006 caused extensive non-specific shift changes and line-broadening to the 1 H-15 N HSQC NMR spectrum of nNOS-PDZ, possibly due to aggregation effects of ZL006. Interestingly, a 20-fold excess of either IC87201 or ZL006 does not appear to disrupt the preformed nNOS-PDZ/PSD95-PDZ2 complex as no shifts changes were observed in the 1 H-15 N HSQC spectrum of the complex upon addition of either compound (Fig. 6d,e). Thus, these NMR data are consistent with the ITC experiments and provide further evidence that neither IC87201 nor ZL006 interacts specifically with nNOS-PDZ.

Discussion
ZL006 and IC87201 have been suggested to target the extended PDZ domain of nNOS and thereby serve as inhibitors of the nNOS/PSD-95 interaction. These compounds would be of great value as pharmacological tool compounds and could represent novel drug leads if they specifically inhibit the nNOS/ PSD-95 interaction without affecting other PSD-95-mediated interactions. Indeed ZL006 and IC87201 seem very convincing in neuronal and animal models of ischemic stroke, pain and depression. However, our data -based on sensitive, informative and direct methods, such as FP, ITC, and NMR -demonstrate that ZL006 and IC87201 do not bind the extended nNOS-PDZ domain (or the PSD-95-PDZ domains) and are not able to inhibit nNOS-PDZ/PSD-95-PDZ interactions.
In this study, we have focused on the PDZ domain-mediated interactions of nNOS and PSD-95, while the reported inhibition by ZL006 and IC87201 were based on studies involving full-length or larger protein constructs 24,26,30 . Our data, therefore, strongly indicate that the true mechanism of action of ZL006 and IC87201 is not via direct binding to the extended nNOS-PDZ domain as originally suggested for ZL006 24,25 and as widely believed for both ZL006 and IC87201 12,[27][28][29]31 . Given their reported potency, we speculate that ZL006 and IC87201 could bind to other parts of the large and complex 321 kDa homodimer nNOS protein 38 , or perhaps even to other proteins affecting the nNOS/PSD-95 system. In addition, it must be borne in mind that small molecules can affect and disturb biochemical assays in a number of subversive and unforeseeable ways 39 . Specifically, compounds comprising a 2-hydroxybenzylamine moiety, as seen in ZL006 and IC87201 and which is essential for ZL006 activity 24 , are known as phenolic Mannich bases. Such compounds have been identified as 'frequent hitters' in AlphaScreening campaigns and can cause artefacts by chelating metal ions or forming quinone methide intermediates that react covalently with protein 40,41 . The assays used in literature for demonstrating nNOS/PSD-95 inhibition were based on either co-immunoprecipitation (ZL006) 24 or capturing assays (IC87201) 26 . These multi-component methods involving antibodies, recognition-tags, complex buffers and surface-immobilization may be prone to pick up compounds with undesirable off-target effects.
In conclusion, the potential of artefactual effects together with our experimental findings using sensitive in vitro assays highlight the critical need for further mechanistic studies of ZL006 and IC87201. This is essential to clarify the true promise of these two compounds as pharmacological tools and drug leads and to aid further medicinal chemistry efforts in this important area of neuroscience.
nNOS-PDZ (12-130, human) and α -Syntrophin-PDZ (82-200, human), both comprising an N-terminal His-tag/FXa-site sequence (MHHHHHHHGGIEGRKL), were likewise expressed in E. coli and purified by HisTrap chromatography followed by gel-filtration as described in details previously 19 . To make the TAMRA-labelled nNOS-PDZ, mutant nNOS-PDZ (V36C) protein was generated by standard site-directed mutagenesis and characterized as described previously 19 . Here, dithiothreitol was added to the pooled protein sample after HisTrap purification to a final concentration of 10 mM to eliminate unwanted disulfide bonds, and a final gel-filtration (50 mM NaPi buffer pH 7.5) was performed to remove impurities and salts. Next, purified nNOS-PDZ (V36C) was desalted by dialysis into 50 mM (NH 4 ) 2 CO 3 buffer (SnakeSkin dialysis tubing 3.5 MWCO, Pierce, Rockford, IL, USA) and subsequently lyophilized. nNOS-PDZ (V36C) was redissolved in 500 μ L buffer (6 M GnHCl, 200 mM NaPi) and solid TAMRA-maleimide (5-tetramethylrhodamine-C2-maleimide; Anaspec, Fremont, CA, USA) was added in excess (cf. ESI-LC-MS). The coupling reaction was incubated at room temperature for 16 hours and TAMRA-labelled nNOS-PDZ was purified by preparative HPLC, and refolded in assay buffer. Functional validation was performed by FP 19 , and all final proteins were characterized for purity and identity by SDS-PAGE and ESI-LC-MS.
For isothermal titration calorimetry (ITC) and NMR, PSD-95-PDZ2 (157-249, human) was expressed and purified as described previously 37 . The sequence for the extended Mus musculus nNOS-PDZ (residues 12-134, Uniprot Q9JJ30) was sub-cloned into a pETM11 vector using cloning sites Nco1 and Kpn1, and the protein was expressed and purified following a similar protocol to that used for the production of PSD-95-PDZ2.
Fluorescence Polarization Assay. FP was generally performed as previously described 11,16,32,33 , but with the following specifications: Saturation binding experiments were performed for measuring binding affinity (K d ) between TAMRA-or Cy5-labelled fluorescent peptide or protein (i.e. the probe) and the PDZ domain protein by applying an increasing amount of PDZ domain protein (typically 0-150 μ M) to a fixed and low concentration of probe (5 or 50 nM). Incubation time was 10-15 minutes (room temperature), and the assay was performed in a 1 × TBS buffer (150 mM NaCl, 10 mM Tris, pH 7.4) using black flat-bottom 384-well plates (Corning Life Sciences, NY) and a Safire2 plate-reader (Tecan, Männedorf, Switzerland). G-factor was adjusted so that probe alone (i.e. without protein interacting-partner) would give an FP value of 20 mP. This background value was subsequently subtracted from the raw data before plotting and analyzing the data. TAMRA-and Cy5-probes were measured at excitation/emission values of 530/585 nm (bandwidth = 20 nm) and 635/670 nm (bandwidth = 15 nm), respectively. The FP values were fitted to the equation Y = B max × X/(K d + X), with B max being the maximal FP value, X is the PDZ concentration, and Y is the experimental FP values. As long as the concentration of labelled peptide is well below the true K d during the assay, the K d can be directly derived from this saturation curve as being equal to the PDZ concentration where the curve is half-saturated (at these conditions EC 50 ∫ [total PDZ] half-saturation = [free PDZ] half-saturation ∫ K d ) 42 .
To measure inhibition of probe and protein by test substance (controls compounds, ZL006, IC87201, or unlabelled protein or peptide), heterologous competition binding assays were performed by adding increasing concentrations of test substance to a fixed concentration of probe and protein under the same conditions as in the saturation binding experiments. FP values were fitted to the general equation: Y = Bottom + (Top-Bottom)/[1 + (10 (X-logIC50)*HillSlope )], where X is the logarithmic value of peptide/compound concentration, and Y is the experimental FP values. Competitive inhibition constants, K i values, were calculated from the IC 50 values to quantify affinities between test substance and protein 42 .
To measure the expected ability of ZL006 and IC87201 to inhibit the PSD-95-PDZ/nNOS-PDZ interaction using unlabelled protein, an indirect FP experiment was established using a probe (Cy5/ TAMRA-Cnksr) that selectively binds PSD-95-PDZ2 (or PDZ1-2) but not nNOS-PDZ (Fig. S1, Fig. 4a). Increasing concentrations of ZL006 or IC87201 were added to a fixed concentration of PSD-95-PDZ, nNOS-PDZ, and probe ( Fig. 4c-d). If IC87201 and ZL006 bind nNOS-PDZ and inhibit the PSD-95-PDZ/ nNOS-PDZ interaction, PSD-95-PDZ would get liberated from the interaction and instead bind to the probe and thus give rise to an increase in FP signal in accordance with the inhibition curve of nNOS vs PDZ/Cnskr (Fig. 4b). Similar, this assay was established with Syntrophin-PDZ and Sapk3 as probe (Fig. 4d).
ZL006 and IC87201 stocks were prepared in 100% DMSO and dilution curves were generated using 1 × TBS buffer and, if necessary to get full solubilisation, extra DMSO. Final DMSO concentrations in the FP assay did not exceed 10% for IC87201 and 20% for ZL006 at their highest test concentrations; and in test concentrations ≤ 600 μ M, DMSO concentrations did not exceed 5% and dropped proportionally with compound concentrations. In all cases, the DMSO effect on the assay was tested and found negligible.

Isothermal Titration Calorimetry. All ITC experiments were carried out at 25 °C on an iTC200
Microcalorimeter (GE Healthcare) with a 200 μ L cell capacity and 40 μ L syringe volume. The buffer used for all solutions was 20 mM phosphate buffer at pH 6.3; each experiment consisted of an initial injection of 0.5 μ L, followed by fifteen 2.39 μ L injections before a final injection of 1.89 μ L. Control experiments were performed whereby IC87201 or ZL006 was titrated into buffer and buffer titrated into nNOS or PSD-95-PDZ2. In both cases no heat exchange was detected, confirming that there was appropriate match of buffer conditions with no indication of dilution effects. The ITC titration experiments for the interactions between IC87201 or ZL006 and each PDZ domain were performed in duplicates, with 100 μ M nNOS-PDZ or PSD-95-PDZ2 in the cell and 750 μ M IC87201 or ZL006 in the syringe, with 1% DMSO present in all the solutions. For the nNOS-PDZ/PSD-95-PDZ2 interaction, 100 μ M PDZ2 in the cell was titrated with 750 μ M nNOS-PDZ. To investigate whether IC87201 or ZL006 was able to inhibit the nNOS-PDZ/PSD-95-PDZ2 interaction, the ITC titration was performed as in the nNOS-PDZ/PSD-95-PDZ2 interaction but with 1 mM IC87201 or ZL006 (i.e. PSD-95 PDZ2:IC87201 or ZL006 = 1:10) present in both the PSD-95-PDZ2 and the nNOS-PDZ samples. All data was analysed using the Origin ® 7 software programme. 1 H-15 N HSQC NMR. The NMR spectra were acquired at 25 °C on Bruker AVANCE II 600 MHz or 800 MHz spectrometers, equipped with triple resonance cryoprobes. The Bruker TopSpin programme version 3.1 was used to process the resultant NMR spectra and the Collaborative Computational Project for NMR (CCPN) Analysis software programme was used for interactive spectral analysis. 2-D 1 H- 15 N HSQC NMR experiments 43 were performed using 0.05-0.25 mM 15 N-labelled nNOS-PDZ or PSD-95-PDZ2 in 20 mM phosphate buffered to pH 6.3. When required, IC87201 or ZL006 in DMSO-d 6 was added to a 10-20-fold excess relative to protein.