Dynamic combinatorial chemistry (DCC) has proven its potential in drug discovery speeding the identification of modulators of biological targets. However, the exchange chemistries typically take place under specific reaction conditions, with limited tools capable of operating under physiological parameters. Here we report a catalyzed protein-directed DCC working at low temperatures that allows the calcium sensor NCS-1 to find the best ligands in situ. Ultrafast NMR identifies the reaction intermediates of the acylhydrazone exchange, tracing the molecular assemblies and getting a real-time insight into the essence of DCC processes at physiological pH. Additionally, NMR, X-ray crystallography and computational methods are employed to elucidate structural and mechanistic aspects of the molecular recognition event. The DCC approach leads us to the identification of a compound stabilizing the NCS-1/Ric8a complex and whose therapeutic potential is proven in a Drosophila model of disease with synaptic alterations.
Sanders and Lehn groups reported the concept of Dynamic Combinatorial Chemistry (DCC) in the mid-1990s1,2. By using reversible chemical reactions, DCC establishes molecular networks under thermodynamic control that respond to external stimuli3,4,5.
DCC systems that employ a protein to direct assemblies of small molecules at dynamic equilibrium are highly interesting. Huc and Lehn reported the use of carbonic anhydrase as a template proving its inhibition by a dynamic combinatorial library (DCL) of imines created in situ6. Since then, successful applications discovering novel enzyme inhibitors have been reported7. On protein-directed DCC experiments, one designs the system rather than the molecule allowing the protein to find its best ligand in situ8,9,10.
The Neuronal Calcium Sensor 1 (NCS-1) is a high-affinity Ca2+-binding protein predominantly expressed in neurons11,12. NCS-1 is a highly conserved protein13 that regulates synaptogenesis, synaptic transmission and is critical for learning and memory11,14,15,16. The Drosophila Neuronal Calcium Sensor 1 (dNCS-1 or Frequenin-2) displays a large, concave hydrophobic crevice onto which the guanine exchange factor Ric8a binds15,17. The interaction between NCS-1 and Ric8a regulates synapse number and probability of neurotransmitter release, thus constituting a pharmacological target for synaptopathies15,18. NCS-1 contains a C-terminal dynamic helix (called H10) that works as a built-in competitive inhibitor and inserts into the crevice to prevent Ric8a binding (Fig. 1). In fact, inhibitors of this protein-protein interaction (PPI) target the NCS-1 crevice and stabilize the orientation that presents the helix H10 inside the crevice. This topology in turn, decreases synapse number and enhances associative learning in a Fragile X syndrome animal model18,19. Following the same reasoning, it would be tempting to hypothesize that an stabilizer of this PPI would permit to enhance synapse number and therefore constitute a pharmacological target of neurodegenerative diseases, where synapse number is abnormally low (Fig. 1)20,21. The low number of ligands reported for dNCS-118,19 makes the DCC approach attractive as a genuine discovery tool for modulators able to unveil the mechanism to control neurotransmission22 and synaptogenesis.
NMR spectroscopy has been reported as a particularly useful technique to analyze protein–ligand interactions (e.g., STD-NMR, tr-NOESY) and to understand reaction mechanisms using Ultrafast NMR (UF-NMR)23,24.
Herein, we apply the DCC approach targeting dNCS-1 at low temperatures and physiological pH with an efficient catalyst to accelerate the DCL equilibration. UF-NMR technique is used to monitor in real-time the details of the acylhydrazone exchange process. Next, ligand-based NMR methods (STD-NMR, tr-NOESY and DOSY) in the presence of dNCS-1 are performed to get further insights into the interaction aspects of the chemical process. Moreover, the affinity of the amplified molecules is measured using fluorescence techniques and the modulation of the NCS-1/Ric8a interaction is tested in a protein-protein binding assay. These methodologies together with blood–brain barrier penetration assays, cell toxicity studies and ADME predictions permit to identify compound 3b as the most effective molecule able to stabilize the NCS-1/Ric8a complex. Furthermore, the structure of the homologous hNCS-1 bound to 3b is solved by X-ray diffraction to understand at the atomic level the basis of its ability to modulate the protein-protein interaction. Importantly, the therapeutic potential of compound 3b is also assessed in vivo, showing that 3b mediates the recovery of normal synapse number and improves the locomotor activity in a Drosophila model for Alzheimer´s disease.
Acylhydrazone exchange catalyst at low temperature
Most of the protein-directed DCC approaches reported to date have been tested under room temperature conditions7,8. However, in our case, besides the standard requirements such as compatibility with the biological target and short equilibration time, the reaction must occur at neutral pH and low temperatures to increase the stability time of dNCS-1.
To conduct acylhydrazone exchange at neutral pH, Greaney and coworkers used high concentrations of aniline as a nucleophilic catalyst in a protein-directed DCL25. The aniline catalyzed the acylhydrazone exchange through a Schiff-base intermediate. We started our DCL by reacting aldehyde 1 (Fig. 2a), with an excess of five acylhydrazides (2a–2e) at 4 °C in the absence and presence of the protein. Unfortunately, the required high concentrations of aniline interfered with the techniques employed for the analysis of protein–ligand interactions. Therefore, we studied different p-substituted aniline bases such as p-aminophenol, p-anisidine and p-phenylenediamine to compare their efficiency as nucleophilic catalysts, given the capacity of electron donating groups on p-position of the ring for increasing the basic character of the corresponding Schiff-bases (Fig. 2b)26.
HPLC-MS was used to screen the proposed catalysts for the formation of the acylhydrazone 3b. The reaction was performed at 4 °C in the presence and in the absence of the p-substituted aniline derivatives. The reactions were initiated by the addition of the aldehyde 1 and the formation of 3b (Fig. 2c) was monitored over time. The resulting data were fit to a pseudo-second-order rate equation (see Supplementary Figs. 1, 2 and Supplementary Methods) and the kinetic parameters are summarized in Fig. 2d.
Under our experimental conditions, p-phenylendiamine and p-anisidine showed superior catalytic activity compared to aniline. Fig. 2c illustrates the time course of the reaction. In the absence of the catalyst (Fig. 2d), the half-time (t1/2) of the reaction is 303 min (Kobs = 0.61 ± 0.02 M−1 s−1). As expected, aniline enhanced the rate of acylhydrazone formation reducing the t1/2 from 303 to 101 min (Kobs = 1.6 ± 0.1 M−1 s−1). However, the t1/2 for p-phenylendiamine and p-anisidine was highly reduced from 303 min to 53 min (p-phenylendiamine, Kobs = 3.1 ± 0.8 M−1 s−1) and 55.5 min (p-anisidine, Kobs = 3.0 ± 0.2 M−1 s−1), respectively. The reaction is completed after 2.5 h. Due to solubility reasons we decided to use p-anisidine instead of p-phenylendiamine as catalyst.
Real-time acylhydrazone exchange mechanism
NMR experiments were then conducted to monitor the dynamic acylhydrazone exchange in real-time and to confirm the proposed mechanisms in the absence (path i) and in the presence (path ii) of the catalyst at physiological pH (Fig. 3).
Initially, the products and intermediates participating in the reaction between aldehyde 1 and acylhydrazide 2b were identified in the absence of the catalyst (Fig. 3a) using standard 1D-1H-NMR spectra acquired in a sequential manner (Fig. 3b). Intermediate I (green) was identified by analyzing their 1H-NMR signals (the signal at δ5.25 ppm corresponds to the H on the carbinolamine carbon while that at δ7.79 ppm, represents the aromatic H ortho to the nitro group). These signals disappear as the final product is being formed. In fact, the formation of 3b can be followed by the increasing presence of the imine-type 1H-NMR signal at δ 8.13 ppm (purple). Although the mentioned NMR signals of intermediate I and 3b are already present in the initial recorded NMR spectrum, the aldehyde signals completely disappeared after 24 h. As expected, at physiological pH, the acylhydrazone formation is rate limited by the dehydration step.
Similar sequential 1D-1H-NMR spectra were recorded to study the catalytic pathway adding 2b to the mixture of the catalyst (p-anisidine) and 1. However, in this case, the intermediates could not be identified due to the higher speed of this process and to the severe overlapping of the NMR signals arising from the mixture. Therefore, we considered the use of 2D-NMR to get better signal dispersion.
The so-called ultrafast 2D-NMR (UF-NMR) method was employed, since it has been demonstrated that it may be used to monitor chemical reactions in-situ23,24. In particular, 2D-UF-TOCSY experiments were recorded to identify the intermediates formed upon adding 2b, using a fast mixing device, into the NMR tube containing a solution of 1 and p-anisidine. The p-anisidine concentration was 0.5 equivalents with respect to 1 to be able to detect the three different p-anisidine states (free state, Schiff-base, Intermediate II).
Figure 3c shows a selection of different UF-TOCSY experiments recorded at different times (see video in the Supplementary Movie for the sequence of the five hundred UF-TOCSY recorded spectra and Supplementary Fig. 12). Initially, the NMR signals for the imine-type proton, the aromatic protons of the aldehyde and the p-anisidine, both taking part in the Schiff-base (see Supplementary Fig. 11) were readily identified (red). At 7 min, the signal of the imine proton of 3b was already observable (purple), while the NMR signals of the protons at p-position from the released p-anisidine catalyst (blue) were evident. The process finished after 40 min, but it was not possible to confirm the presence of the Intermediate II. Furthermore, a small amount of Intermediate I (11%) was observed in the UF experiments as result of the coupling between aldehyde 1 and acylhydrazide 2b (Supplementary Figs. 13, 14 and Supplementary Table 4). Experiments with other acylhydrazides were also performed confirming these results.
dNCS-1 dynamic combinatorial library
After establishing p-anisidine as catalyst, the DCL approach was attempted by mixing aldehyde 1 (Fig. 4a) with five acylhydrazides (2a–2e) in the presence of dNCS-1. The DCL control was also performed in the absence of the protein. The selection of the aldehyde and the 5 acylhydrazides was based on previous DCL experiments in which the building blocks reactivity and their concentrations were carefully assessed to ensure the full solubility of the different components. The stability of the protein under the experimental conditions (DMSO tolerance and stability over time) was also tested using fluorescence and NMR techniques (Supplementary Figs. 15, 16 and Supplementary Methods). The equilibration was completed after 5 h (Fig. 4a) and the acylhydrazones were identified by HPLC-MS (Supplementary Figs. 4–8).
Aldehyde 1 could not be detected, indicating that it was continuously being sequestered as an acylhydrazone component. The reversibility of the DCL was evident, since an identical equilibrium distribution to that shown in Fig. 4a was obtained when two different starting points were employed. The observed degree of amplification was 3b > 3e ≥ 3d > 3a. The precise composition of the DCL (with and without dNCS-1), was assessed by measuring the relative peak area (RPA). Indeed, the normalized change of RPA was used to quantify the protein influence in the final outcome (Supplementary Figs. 9, 10 and Supplementary Tables 1, 2 and 3)28. The presence of acylhydrazone 3c was clearly reduced in presence of dNCS-1 indicating a lack of significant affinity for dNCS-1. Note that compounds 3a–3e can exist as E/Z isomers of the C = N bond. Quantum mechanics calculations of the geometries for the E/Z stereoisomers of 3a–3e revealed that isomer E is preferred; both in vacuum and in water (Supplementary Table 6). Interestingly, the calculated pKa for the acylhydrazone NH (8.0 and 8.5 for Z and E isomers, see Supplementary Figs. 19 and 20) of compound 3b shows the acidic nature of this NH proton, which strongly suggests that the isomerization from the Z to the most stable E isomer may easily occur in the reaction medium at pH 8.
DOSY-NMR and tr-NOESY-NMR experiments were also recorded to follow the exchange process. The obtained DOSY spectra in the presence of dNCS-1 revealed that the formed products displayed larger diffusion coefficients than the initial components, in agreement with their larger size increase (Fig. 4b). Moreover, the presence of protein-bound products was further assessed by the presence of negative cross-peaks for the acylhydrazones in the tr-NOESY experiments, while the reactants and the catalyst only displayed positive and zero-quantum cross-peaks (Fig. 4c).
Additional probe of the existence of a protein template effect was extracted from the NMR analysis of the evolution of the mixture of 3c (extremely weak or non-binder) and 2b with dNCS-1 (Fig. 4d). The 1H-NMR spectra revealed the presence of signals at the aromatic region assigned to 3b as well as a triplet at δ 2.2 ppm. corresponding to 2c, the acylhydrazide precursor of 3c. Thus, dNCS-1 induces the synthesis of 3b, a dNCS-1 ligand, at the expense of 3c, which is not bound to the protein.
NMR binding studies and compounds epitope mapping
The analysis of the STD-NMR spectra29 further identified four acylhydrazones as dNCS-1 binders (Fig. 5), while 3c was not recognized. Compound 3b displayed the largest STD intensities. Moreover, the STD analysis permitted to map its binding epitope, revealing structural details of the dNCS-1/3b binding mode.
NCS-1 affinity and NCS-1/Ric8a complex modulation
Fluorescence-based experiments with 3a–3e were carried out to estimate their affinity to dNCS-1. As shown in Fig. 6a, the binding of 3a, 3b and 3d quenches the fluorescence of tryptophans W30 and W103 located in the dNCS-1 hydrophobic cavity17. A similar effect was observed when Chlorpromazine (CPZ), an antipsychotic drug and a well-known dNCS-1 binder18, was used as control. The apparent Kd for 3a (Kd = 32 ± 2 μM), 3b (Kd = 43 ± 6 μM) and 3d (Kd = 61 ± 9 μM) were slightly larger than that of CPZ (Kd = 12 ± 2 μM), suggesting the existence of a similar binding affinity of all these molecules to dNCS-1. Compound 3c did not show affinity for dNCS-1 while the limited solubility of 3e under the experimental conditions precluded the acquisition of the data. Therefore, 3c and 3e were discarded for further studies.
Binding assays with NCS-1 and Ric8a in co-transfected HEK cells were carried out to study the modulation effect of compounds 3a, 3b and 3d in the protein-protein interaction (Fig. 6b). CPZ, a reported mild inhibitor of the NCS-1/Ric8a interaction18, was also included for comparison purposes. Interestingly, our data showed that compounds 3b and 3d promote the stabilization of the NCS-1/Ric8a interaction whereas 3a is an inhibitor similar to CPZ.
In vitro permeability
Taking into account that one of the main difficulties in treating central nervous system diseases is the drug’s capacity to cross the blood–brain barrier (BBB), the ability of compounds 3 (a, b, and d) to enter into the brain by passive diffusion was evaluated in a Parallel Artificial Membrane Permeation Assay (PAMPA methodology, Supplementary Fig. 18 and Supplementary Methods)30. The PAMPA methodology is a high-throughput technique to predict passive permeability through biological membranes that employs a brain lipid porcine as membrane. The in vitro permeabilities (Pe) of 3a, 3b and 3d and ten commercial drugs were then determined. Compounds with Pe > 4.47 × 10−6 cm s−1 are able to cross the BBB by passive diffusion. As a result, compound 3b can be classified as CNS + with a permeability of 12.9 ± 0.8 × 10−6 cm s−1. In contrast, 3a and 3d did not show good permeability values (Fig. 6c, Supplementary Table 5).
In silico physicochemical parameters and neuron viability
In addition, in silico evaluation of Absorption, Distribution, Metabolism and Excretion (ADME) descriptors such as log Po/w (pH-independent partition coefficient) and log D (pH-dependent partition coefficient) were predicted for 3b, obtaining 2.58 and 2.04, respectively at pH = 8 (Supplementary Table 7). This is in agreement with optimal log Po/w values (as an indicator of brain-blood partitioning) of 1.5–2.5 for drugs targeting CNS31. Furthermore, the aqueous solubility (log S) of 3b was also calculated yielding values of −4.183/−4.675, similar to those obtained for other CNS drugs (see Supplementary Table 8).
Finally, cell toxicity in neurons was quantified as percentage of picnotic bodies for 3b (Fig. 6d). There were no significant differences in 3b treated cells with those obtained with the same amount of the drug vehicle, DMSO. Our results suggest a physiological effect of compound 3b without affecting neuron viability.
In light of these results, compound 3b was chosen as candidate to understand the binding properties and to study the in vivo effect as a promising hit compound.
The crystal structure of hNCS-1 bound to 3b
To understand the activity of 3b as an stabilizer of the NCS-1/Ric8a interaction, the crystal structure of the Ca2+ bound hNCS-1/3b was solved at 1.78 Å resolution (PDB code 6QI4, Table 1). Crystals belonged to the monoclinic P21 space group. The asymmetric unit (AU) contained two hNCS-1 molecules with an RMSD for all atoms of 1.28 Å. The feature-enhanced and the 2Fo − Fc electron density maps, together with different map calculations (see Fig. 7a and Supplementary Fig. 17) allowed the unambiguous modelling of 3b bound to the hydrophobic crevice of one of the two independent hNCS-1 molecules of the AU, while the second hNCS-1 molecule only showed a PEG molecule at the 3b equivalent position (Fig. 7). Interestingly, 3b targets the same region as other inhibitors (Fig. 8)18,19, displaying a contact area of 303.4 Å232.
The amino acids participating in 3b recognition are: W30 and D37 (helix H2), defining the upper wall of the cavity (Fig. 7c). The base of the cavity is formed by F72 and V68 (helix H4), F48 (helix H3) and W103 (helix H6). As lateral walls: F85, L89 and T92 (helix H5) and opposite to it, I51, F55 and Y52 (helix H3). The indole group of 3b is stabilized with π-π interactions with W30 and F85, weak water-mediated H-bonds with D37, and hydrophobic interactions with L89 and I51. The acylhydrazone oxygen of 3b is forming a strong H-bond with a water molecule in the upper part of the cavity. Nitrogen atom N2 is stabilized with van der Waals contacts with F48 and nitrogen N3 with Y52 and F55, being the latter mediated by a water molecule (w158). Furthermore, hydrophobic interactions are observed between C11 and F48 and F72. Interestingly, the electron density map showed that the NCS-1 bound 3b molecule only displays the E geometry, the QM-predicted and most stable isomer (Supplementary Table 6). Nevertheless, since 3b is present in solution as a mixture of Z/E isomers, the molecular recognition event takes place with a conformational selection process. In addition, the 2-hydroxy-3-nitrophenyl ring perpendicular to the surface interacted with V68 and Y52 and F72, W103 and T92. The 3b most implicated atoms in these interactions are C13, C14 and C17. It is important to note that the interactions observed in the hNCS-1/3b crystal structure match the STD-NMR epitope mapping (Fig. 5).
When comparing the 3b-free and bound hNCS-1 structures found in the asymmetric unit, a rearrangement takes place to allow ligand recognition: helix H3 shifts and D37 carboxyl reorients to establish weak H-bonds between a group of water molecules and 3b indole group (Fig. 7d). Indeed, I52 side chain changes to permit the positioning and interaction of water w155. Furthermore, T92 side chain, that shows double conformations in the absence of 3b, fixes its conformation in the presence of 3b through a H-bond with a water molecule, which permit to establish hydrophobic contacts with the 2-hydroxy-3-nitrophenyl ring.
The comparison of hNCS-1/3b structure with the reported structures of dNCS-1 bound to strong (FD44, IGS-1.76)18,19 and mild (CPZ) inhibitors (Fig. 8) shows that 3b indole group is placed upper in the cavity enabling D37 participation. Moreover, 3b does not form strong H-bonds with T92 and Y52 or contact the helix H10, as inhibitors do (Fig. 1 and Fig. 8b). Particularly, the binding of PEG molecules to the C-terminal part of the crevice stabilizes the helix H10 outside and parallel to it (Figs. 7b, 8a). While the strong inhibitors (Fig. 8a–b, d) use apolar rings to properly contact L182 and L184 and stabilize the helix H10 inside the crevice, compound 3b locates at L182 and L184 interaction region the nitro and hydroxyl groups, hindering the stabilization of the helix H10 inside. Therefore, our data suggest that 3b stabilizes the NCS-1/Ric8a complex keeping the helix H10 out of the crevice, and promoting the entrance and recognition of Ric8a.
Finally, the structural comparison of the PPI modulators indicate that these compounds can be divided in two parts: i) an aromatic region that targets the molecules to an aromatic-enriched area of the NCS-1 crevice and confers affinity (highlighted in Fig. 8d), and ii) a variable region that confers function: inhibition or stabilization. Inhibitors need long-enough hydrophobic moieties that reach the helix H10 for interaction, and stabilizers need polar groups that hinder the helix insertion (Fig. 8d).
Compound 3b in a Drosophila model of Alzheimer’s disease
As we have established that compound 3b stabilizes the NCS-1/Ric8a interaction and given the reported effects of this interaction on regulating synapse number and synapse function15,16,18, we assayed 3b on an in vivo model of synaptopathy, where synaptic loss is a primary hallmark of disease20,21. The expression of synaptotoxic amyloid aβ42arc in motor neurons leads to a reduction in the number of synapses with respect to normal age-matched neuromuscular junctions33. Moreover, the expression of amyloid peptides in Drosophila neurons displays various symptoms reminiscent of Alzheimer’s disease including defective locomotion, memory loss or reduced longevity34.
Aβ42arc overexpressing flies and the corresponding control (LacZ expression) were fed with 3b or the solvent, DMSO, throughout all life cycle (Fig. 9 and see Methods section). The data confirmed that synapse counting was reduced in aβ42arc33, but this pathological phenotype was largely suppressed in the 3b feed flies. By contrast, 3b and its solvent DMSO showed no effect on the control flies (Fig. 9a).
To measure the physiological impact of this synaptic recovery, we evaluated fly locomotor activity. As described previously, overexpression of human aβ42arc peptide leads to severe locomotor dysfunction starting at day 15–20 post-eclosion35. Remarkably, the locomotor deficit was recovered by 3b feeding (Fig. 9b). Furthermore, the statistical analysis of the data does not reveal a significant difference in the locomotor activity of control flies fed with 3b vs. DMSO.
Adaptability is the essence for evolution and guides the emergence of diverse chemical structures. Modulators of protein-protein interactions are relatively rare. We designed a dynamic reversible system from one aldehyde and five acylhydrazides able to uncover an unexpected protein-protein interaction stabilizer.
The reversible chemistry chosen, acylhydrazone exchange, was prepared to work at low temperatures and neutral pH using p-anisidine as a catalyst broadening its range of application to other biological targets. Moreover, ultrafast NMR experiments have allowed the detection of the carbinolamines (hemiaminals) intermediates and could successfully be applied to determine the mechanism of C = N double bonds formation of pyrazoles37 and isoxazoles38.
The calcium sensor protein NCS-1 has been proved to be an excellent DCL template directing the library to the synthesis of compound 3b, the protein-protein interaction enhancer of the NCS-1/Ric8a complex ever reported, still with a moderate binding affinity. Nevertheless, detailed NMR and X-ray studies have shed light on the structural and chemical requirements to stabilize the NCS-1/Ric8a complex.
We had previously shown how the interaction of NCS-1 and Ric8a emerged as a potential therapeutic target for diseases affecting synapses, due to its role in regulating synapse number and neurotransmitter release11,15,16,18. In this context, the in vivo results show that 3b-mediated stabilization of the NCS-1/Ric8a complex, indeed increases the number of synapses to normal levels, exclusively in the presence of a synaptic pathology, which is an essential requirement for any treatment directed to synapses. Therefore, compound 3b constitutes a promising prototypic probe for further research in the treatment of neurodegenerative disorders such as Alzheimer’s, Huntington’s or Parkinson’s diseases characterized by a decrease in the number and efficacy of synapses that precedes neuronal death.
Catalysis of reversible acylhydrazone formation
The kinetic experiment was performed adding aldehyde 1(1.2 μL, 50 mM, DMSO), acylhydrazide 2b (3.6 μL, 1.5 M, DMSO), the catalyst (p-anisidine, p-phenylendiamine, aniline) or control DMSO (1.0 μL) in buffer Tris (pH 7.4, 640 μL, 20 mM, 0.5 M NaCl, 1 mM CaCl2, 1 mM DTT) at 4 °C at 15 mM and 50 mM catalyst concentration. 5% of DMSO was present in the final mixture. The absorbance data of 1 and 3b were measured for seven hours by HPLC. The data were collected and treated by using a least squares algorithm to fit the equation for pseudo-second-order (Supplementary Figs. 1, 2 and Supplementary Methods)
A solution of aldehyde 1 (125 μL, 0.2 M, DMSO-d6) and of p-anisidine (2 μL, 6 M, DMSO-d6) were added to a mixture of DMSO-d6 (273 μL) and Tris buffer D2O (100 μL), in a 5 mm NMR tube. Inside of a NMR tube was assembled a fast mixing device for adding the acylhydrazide 2b (50 μL, 0.5 M, DMSO-d6) (Supplementary Methods).
Aldehyde 1 (1.2 μL, 50 mM, DMSO), the five acylhydrazides 2a–2e (5 × 3.6 μL, 50 mM, DMSO), p-anisidine (1 μL, 12 M, DMSO) and buffer 20 mM Tris, 0.5 M NaCl and 1 mM CaCl2, 1 mM DTT at pH 7.4 (750 μL) in 3.3 % DMSO. The mixture was stabilized in 5 h at 4 °C. HPLC analysis was performed.
Aldehyde 1 (1.2 μL, 50 mM, DMSO), the five acylhydrazides 2a–2e (5 × 3.6 μL, 50 mM, DMSO), p-anisidine (1 μL, 12 mM, DMSO) and dNCS-1 in buffer 20 mM Tris, 0.5 M NaCl and 1 mM CaCl2, 1 mM DTT at pH 7.4 (66.7 μM, 750 μL, 1 eq.). The experiment is in 3.3% DMSO. The mixture was stabilized for 5 h at 4 °C. Then, dNCS-1 was removed by ultrafiltration using an Amicon Ultra filter (0.5–10 KDa). HPLC analysis was performed and the traces were compared with the blank composition.
Synthesis of acylhydrazones 3a–3e
See Supplementary Fig. 3, Supplementary Methods and Supplementary Discussion.
The experiments were performed using a deuterated Tris (pH 7.9) buffered solution with an aliquot of DMSO-d6. A sample containing 10 μM of dNCS-1 and 1 mM of DCL species (molar ratio 1:100) was prepared. The STD experiments were acquired at 281 K on a Bruker Avance 600 MHz spectrometer equipped with a cryoprobe. The saturation frequency was set at δ −0.5 ppm (aliphatic region) and the saturation time was 2 s. A spin lock filter was applied to minimize signals from dNCS-1. The same conditions were used for the acquisition of the STD-NMR spectra of the individual products with dNCS-1.
tr-NOESY and DOSY experiments
The experiments were performed on the same samples and spectrometer, at 281 K. The tr-NOESY mixing time was fixed at 200 ms. The DOSY experiments were acquired with 16 gradient increments to a final intensity decay of 90%.
Quantum mechanics calculations
See Supplementary Methods.
Co-Immunoprecipitation assays and western blotting
hNCS-1 and V5-tagged hRic8a were co-transfected into HEK293 cells (Dharmacon) using Lipofectamine 2000 (Invitrogen)13. After 24 h, after transfection, DMSO alone or the compounds dissolved in DMSO (20 μM) were added to the culture cells. Then, 24 h after transfection, cells were lysed with Lysis buffer (150 mM NaCl, 1.0% Nonidet P-40, 50 mM Tris, pH 8.0) in the presence of the compounds (20 μM), whose concentration was maintained throughout the immunoprecipitation assay. Precleared lysates were incubated overnight (12 h) at 4 °C with mouse anti-NCS-1 (1:500; Cell Signaling). Samples were subsequently incubated overnight with Protein-G-Sepharose (Sigma-Aldrich), washed and eluted from the Sepharose. Samples were analyzed by Western blot following standard procedures. The amount of V5-tagged hRic8a bound to hNCS-1 was revealed by V5 antibody (1:1000, Invitrogen). 1/10 cell lysates before IP, were run in a Western blot and the NCS-1 input (anti NCS-1, Cell Signalling 1:2000) and Ric8a input (anti V5, Invitrogen, 1:2000) were then revealed. Original representative western blots are found in Source Data file.
Toxicity in primary cultured neurons
Cortical neurons were obtained from E14 wt mice (C57BL6J). Mice were maintained in accordance to European law and following Hospital Ramón y Cajal´s animal guidelines. Cells were obtained using the neuron isolation kit with papain (Thermofisher) and then maintained 7 days in neurobasal medium and treated the last 24 h with 0.2, 2, 10 and 20 µM of CPZ or 3b or with the same volume of the vehicle DMSO. Three independent experiments counting cells from three different wells per concentration were performed. To analyzed cell death, neurons were fixed and stained with DAPI.
Parallel artificial membrane permeability assays
Methodology and data on the permeability in the PAMPA-BBB assay of 10 commercial drugs and compound 3b, linear correlation between experimental and reported permeability of commercial drugs (see Supplementary Fig. 18, Supplementary Table 5 and Supplementary Methods).
Protein expression and purification
See Supplementary Methods.
Crystallization and diffraction data collection
Fly locomotor activity assay
Fifteen-day-old males were placed individually in locomotor activity monitor tubes (DAM2, TriKinetics Inc.) The DAM2 system automatically counts the number of beam breaks for flies walking in a horizontal tube over a specific period of time. This setup allowed for characterization of the locomotor and behavior rhythms of Drosophila. The tubes contained fly food with 100 µM compound 3b or same volume of DMSO. Flies were raised at 25 °C in 12-h light/12-h dark. In the first 2 days flies get habituated and the next 2 days the locomotor activity was quantified.
Synapse number quantification
The Drosophila neuromuscular junction (NMJ) allows the accurate quantitative determination of the in vivo effects of drug application on a single glutamatergic synapse. Each presynaptic motor neuron and postsynaptic muscle fiber can be easily identified and has a stereotypical morphology with minimum inter-individual variability.
We studied the 20-day old male NMJ from the third abdominal hemisegment. Synapses were visualized under confocal microscopy by the nc82 marker (DSHB Hybridoma Bank), which identifies the Bruch pilot protein, a constituent of the presynaptic active zone. Throughout the text, we refer to nc82-positive spots as mature synapses. Neuronal membranes, delimitating motor neuron terminals were revealed by rabbit anti-HRP antibody (Jackson ImmunoResearch). Serial 1-μm confocal images were acquired in a Leica TSC SP5 Confocal Microscope and quantified by Imaris software (Bitplane). Experimental and control genotypes were run in parallel, and quantifications were done blindly.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
The atomic coordinates and structure factors of the hNCS-1/3b complex have been deposited in the Protein Data Bank under accession code 6QI4. A reporting summary for this Article is available as Supplementary Information file. The source data underlying Figs. 2, 6 and 9 as well as Supplementary Figs. 1, 2, 9, 10, 15, 18 and Supplementary Table 5 are provided as a Source Data file.
Brady, P. A. & Sanders, J. K. M. Thermodynamically-controlled cyclisation and interconversion of oligocholates: metal ion templated ‘living’ macrolactonisation. J. Chem. Soc., Perkin Trans. 1, 3237–3253 (1997).
Hasenknopf, B. et al. Self-assembly of tetra- and hexanuclear circular helicates. J. Am. Chem. Soc. 119, 10956–10962 (1997).
Li, J., Nowak, P. & Otto, S. Dynamic combinatorial libraries: from exploring molecular recognition to systems chemistry. J. Am. Chem. Soc. 135, 9222–9239 (2013).
Lehn, J.-M. Dynamic combinatorial chemistry and virtual combinatorial libraries. Chem. Eur. J. 5, 2455–2463 (1999).
Cougnon, F. B. & Sanders, J. K. M. Evolution of dynamic combinatorial chemistry. Acc. Chem. Res. 45, 2211–2221 (2012).
Huc, I. & Lehn, J.-M. Virtual combinatorial libraries: dynamic generation of molecular and supramolecular diversity by self-assembly. Proc. Natl Acad. Sci. USA 94, 2106–2110 (1997).
Mondal, M. & Hirsch, A. K. H. Dynamic combinatorial chemistry: a tool to facilitate the identification of inhibitors for protein targets. Chem. Soc. Rev. 44, 2455–2488 (2015).
Huang, R. & Leung, I. K. H. Protein-directed dynamic combinatorial chemistry: a guide to protein ligand and inhibitor discovery. Molecules 21, 910–928 (2016).
Frei, P., Hevey, R. & Ernst, B. Dynamic combinatorial chemistry: a new methodology comes of age. Chem. Eur. J. 25, 60–73 (2019).
Atwood, J. L., Gokel, G. W. & Barbour, L. in Comprehensive Supramolecular Chemistry II Ch. 5.20 (Elsevier, Amsterdam, 2017).
Dason, J. S., Romero-Pozuelo, J., Atwood, H. L. & Ferrús., A. Multiple roles for frequenin/NCS-1 in synaptic function and development. Mol. Neurobiol. 45, 388–402 (2012).
Burgoyne, R. D. & Haynes, L. P. Sense and specificity in neuronal calcium signalling. Biochim. Biophys. Acta Mol. Cell Res. 1853, 1921–1932 (2015).
Burgoyne, R. D. & Weiss, J. L. The neuronal calcium sensor family of Ca2+-binding proteins. Biochem. J. 353, 1–12 (2001).
Mun, H. S. et al. Self-directed exploration provides a NCS-1 dependent learning bonus. Sci. Rep. 7, 17697 (2015).
Romero-Pozuelo, J. et al. The guanine exchange factor Ric8a binds to the Ca2+ sensor NCS-1 to regulate synapse number and neurotransmitter release. J. Cell Sci. 127, 4246–4259 (2014).
Dason, J. S. et al. Frequenin/NCS-1 and the Ca2+ channel α1-subunit co-regulate synaptic transmission and nerve terminal growth. J. Cell Sci. 122, 4109–4121 (2009).
Baños-Mateos, S., Chaves-Sanjuan, A., Mansilla, A., Ferrús, A. & Sánchez-Barrena, M. J. Frq2 from Drosophila melanogaster: Cloning, expression, purification, crystallization and preliminary X-ray analysis. Acta Cryst. F70, 530–534 (2014).
Mansilla, A. et al. Interference of the complex between NCS-1 and Ric8a with phenothiazines regulates synaptic function and is an approach for fragile X syndrome. Proc. Natl Acad. Sci. USA 114, E999–E1008 (2017).
Roca, C. et al. Deciphering the inhibition of the neuronal calcium sensor 1 and the guanine exchange factor ric8a with a small phenothiazine molecule for the rational generation of therapeutic synapse function regulators. J. Med. Chem. 61, 5910–5921 (2018).
Selkoe, D. J. Alzheimer’s disease is a synaptic failure. Science 298, 789–791 (2002).
Herms, J. & Dorostkar, M. M. Dendritic spine pathology in neurodegenerative diseases. Annu. Rev. Pathol. 11, 221–250 (2016).
Mansilla, A. et al. Molecular mechanisms that change synapse number. J. Neurogenet. 32, 155–170 (2018).
Herrera, A. et al. Monitoring organic reactions by UF-NMR spectroscopy. Magn. Reson. Chem. 53, 952–970 (2015).
Herrera, A. et al. Real-time monitoring of organic reactions with two-dimensional ultrafast TOCSY NMR spectroscopy Angew. Chem. Int. Ed. 48, 6274–6277 (2009).
Bhat, V. T. et al. Nucleophilic catalysis of acylhydrazone equilibration for protein-directed dynamic covalent chemistry. Nat. Chem. 2, 490–497 (2010).
Dirksen, A., Hackeng, T. M. & Dawson, P. E. Nucleophilic catalysis of oxime ligation. Angew. Chem. Int. Ed. 45, 7581–7584 (2006).
Nachod, F. C. Determination of Organic Structures by Physical Methods Ch. 14. (Academic Press, New York, 1955).
Frei, P. et al. Target-directed dynamic combinatorial chemistry: a study on potentials and pitfalls as exemplified on bacterial target Chem. Eur. J. 23, 11570–11577 (2017).
Caraballo, R., Dong, H., Ribeiro, J. P., Jiménez-Barbero, J. & Ramström, O. Direct STD NMR identification of beta-galactosidase inhibitors from a virtual dynamic hemithioacetal system. Angew. Chem. Int. Ed. 49, 589–593 (2010).
Di, L., kerns, E. H., Fan, K., McConnell, O. J. & Carter, G. T. High throughput artificial membrane permeability assay for blood–brain barrier. Eur. J. Med. Chem. 38, (223 (2003).
Misra, A., Ganesh, S., Shahiwala, A. & Shah, S. P. Drug delivery to the central nervous system: a review. J. Pharm. Pharm. Sci. 6, 252–273 (2003).
Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from crystalline state. J. Mol. Biol. 372, 774–797 (2007).
Lopez-Arias, B., Monedero, I., Turiégano, E. & Torroja, L. The Drosophila adult neuromuscular junctions a model for unraveling amyloid peptide influence on synapse dynamics. Neural Regen. Res. 12, 1987–1989 (2017).
Moloney, A., Sattelle, D. B., Lomas, D. A. & Crowther, D. C. Alzheimer’s disease: insights from Drosophila melanogaster models. Trends Biochem Sci. 35, 228–235 (2010).
Iijima, K. et al. Abeta42 mutants with different aggregation profiles induce distinct pathologies in Drosophila. PLoS ONE 3, e1703 (2008).
Pfeiffenberger, C., Lear, B. C. & Keegan, K. P., Allada, R. Locomotor activity level monitoring using the Drosophila activity monitoring (DAM) system. Cold Spring Harb Protoc. 2010, 15518 (2010).
Singh, S. P. et al. The reaction mechanism between hydrazines and β-dicarbonyl compounds: Proposal for a mechanism. Can. J. Chem. 78, 1109–1120 (2000).
Katritzky, A. R., Barczynski, P., Ostercamp, D. L. & Yousaf, T. I. Mechanism of heterocyclic ring formation. A 13C NMR study of the reaction of β-ketoesters with hydroxylamine. J. Org. Chem. 51, 4037–4042 (1988).
R.P.-F., M.J.S.-B. and A.Mansilla thank Prof. J. Elguero (IQM-CSIC), Prof. J. E. Verdasco (UCM) and Dr. L. Infantes (IQFR-CSIC), for helpful discussions on the paper and Prof. A. Ferrús for guidance in the Drosophila experiments. M.J.S.-B. thanks Alba Synchrotron for Xaloc Beamline access and support of the staff. This work was supported by different funding agencies: the Spanish Ministry of Economy and Competitiveness (MINECO) with Grants CTQ2015-69643-R (R.P.F), SAF2015-74507-JIN (A. Mansilla) and BIO2017-89523-R (M.J.S.-B.), and European Cooperation in Science and Technology (COST) action CMI304 Emergence and evolution of complex chemical systems (R.P.-F.), and a 2017 Leonardo Grant for Researchers and Cultural Creators from the BBVA Foundation (M.J.S.-B) and a 2018 CaixaImpulse program from La Caixa Foundation (A.Mansilla). M.J.S.-B. was supported by a Ramón y Cajal Contract RYC-2008-03449 from MINECO.
The Spanish National Research Council and the Biomedicine Research Foundation of Ramon y Cajal Hospital have filed the patent applications (P201830933 and EP19382242.6) with the Spanish Patent Office on the use of the compounds described in the paper as synaptic modulators. R.P.-F., A.C.-M., A.Mansilla and M.J.S.-B. are listed as inventors. The remaining authors declare no competing interests.
Peer review information: Nature Communications thanks Jeremy Derrick and other anonymous reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Canal-Martín, A., Sastre, J., Sánchez-Barrena, M.J. et al. Insights into real-time chemical processes in a calcium sensor protein-directed dynamic library. Nat Commun 10, 2798 (2019). https://doi.org/10.1038/s41467-019-10627-w