Scaffold Ranking and Positional Scanning Identify Novel Neurite Outgrowth Promoters with Nanomolar Potency

Abstract

Central nervous system (CNS) neurons typically fail to regrow their axons after injury. Injuries or neuropathies that damage CNS axons and disrupt neuronal circuitry often result in permanent functional deficits. Axon regeneration is therefore an intensely pursued therapeutic strategy for numerous CNS disorders. Phenotypic screens utilizing primary neurons have proven successful at identifying agents that promote axon regeneration in vivo. Here, we report the screening of mixture-based combinatorial small molecule libraries in a phenotypic assay utilizing primary CNS neurons and the discovery of neurite outgrowth promoters with low nanomolar potency.

The failure of central nervous system (CNS) neurons to regenerate damaged axons poses a barrier for functional recovery.^1,2 Traumatic brain injury, for example, can produce lifelong psychological and cognitive deficits,³ and spinal cord injury can result in permanent paralysis.⁴ In addition to mechanical injury, CNS axons can be damaged by degenerative neuropathies including glaucoma,⁵ multiple sclerosis,⁶ Alzheimer’s disease, and Parkinson’s disease. Promoting axon regeneration is therefore a highly pursued therapeutic strategy with clinical applicability in a wide range of indications.⁷⁻¹⁰

Suppression of axon regeneration is mechanistically complex,^9,11 making it difficult to tackle using target-based drug discovery approaches. In recent years, phenotypic screening has emerged as an alternative approach for discovering safe and efficacious drugs.¹²⁻¹⁴ This approach tests compounds directly on cells or tissues, thereby circumventing challenges related to identifying effective drug targets.¹⁵ We have previously developed and optimized a phenotypic assay using primary neurons to screen for agents that can promote neurite outgrowth.^16,17 The assay allowed us to identify genes¹⁸ as well as small molecules^11,16,19 that promote axon regeneration in vivo.

Although phenotypic screening provides a powerful approach for identifying biologically active compounds, the added burden of handling, imaging, and data processing typically imposes a lower throughput compared to target-based biochemical screening.¹⁷ The majority of neurite outgrowth screens reported in the literature were carried out with relatively small compound libraries comprising <2000 compounds.¹⁹⁻²² Recently, a neurite outgrowth assay was scaled up to screen a compound library comprised of ∼50,000 compounds.²³ In the current proof-of-concept study, we applied a mixture-based and scaffold ranking approach to a neurite outgrowth screen, which allowed us to interrogate hundreds of millions of unique compound structures and identify a novel group of hit compounds suitable for lead development (Figure 1). Hits were confirmed by resynthesis and validation in the phenotypic assay. One compound, 2539-14, promoted neurite outgrowth by 2–3-fold relative to control with remarkable potency (13 nM). The success of our approach demonstrates the ability of mixture-based combinatorial libraries to identify novel compounds with nanomolar activity in cell-based assays.

Figure 1.

Approach used to identify active compounds. Starting with just 79 samples comprised of >400 million compounds in the Scaffold Ranking Library. The specific Positional Scanning Library (PSL) 2275 was selected based on its activity. PSL 2275 contains 116 samples, and the results from screening this library were used to select 27 compounds that were then synthesized (a process known as deconvolution). The screening results of the 27 individual crude samples led to the purification and additional testing of nine of the compounds.

The advantages of scaffold ranking libraries and the methods by which they are constructed have been previously described.²⁴ Briefly, scaffold ranking libraries are arranged so that each sample contains close structural analogues based on a single core scaffold. In our study, the most complex library (library 2388) consisted of over 5 million compounds, whereas the least diverse (library 2058) consisted of just over 2000. We used our previously established phenotypic assay with primary mammalian neurons to screen for compounds that promote neurite outgrowth.¹⁶ In vivo differentiated neurons best recapitulate the gene and protein expression profiles of actual neuronal subpopulations^17,25 and are therefore better suited than cell lines or in vitro-differentiated cells for purposes of drug discovery and target validation. The assay typically has a Z-factor ≥ 0.7 and is quite sensitive for identifying promoters of neurite outgrowth. Seventy nine different mixtures, representing over 400 million unique structures, were screened at six doses ranging from 10 ug/mL to 3.2 ng/mL. Scaffold ranking data were acquired, and the most active scaffolds (Figure 2) were considered for subsequent testing of the corresponding positional scan library.²⁴ Several scaffold samples that contained an imidazoline branched to a heterocycle (1276 and 2275 branched to a cyclic thiourea, 1319 to a cyclicurea, and 2137 to a diketopiperazine) produced large neurite total length relative to control (%NTL) across a wide concentration range; 1276 and 2275 are two different synthesis lots of the same scaffold mixture, providing clear evidence for the reproducibility of the chemistry and screening assay.

Figure 2.

Prioritization of active scaffolds. Scaffold ranking data at the 0.16 ug/mL treatment concentration compared to positive control (ML-7). Values are neurite total length relative to DMSO control (%NTL).

On the basis of the results of the scaffold ranking screen (Figure 2), mixture 2275 was selected for positional scanning to identify individual compounds that promote neurite outgrowth. Positional scanning samples from library 2275 contain either 1258 or 1665 dihydroimidazolyl-butyl-cyclic thiourea compounds systematically arranged into 116 subset mixtures (Table S1). Samples were tested in the same manner as the scaffold ranking step at six doses ranging from 10 ug/mL to 0.003 ug/mL (Table S1). The higher doses of 10 and 2 ug/mL for the majority of samples were toxic. Samples with broad activity profiles and low apparent toxicity were selected for retesting. These samples were tested at three doses ranging from 500 to 0.2 ng/mL (Table S2). The data show a range of dose-dependent activities at all variable R positions. This differentiation in activity is vital for successful deconvolution and structure–activity relationship (SAR) studies. On the basis of these data, three different functionalities were selected from each of the diversity positions. First, we prioritized the samples that produced a significant increase in %NTL over a broad concentration range. Then, we prioritized compounds based on chemical diversity. For example, in the R3 position, cyclohexyl-ethyl (2275-107), cyclohexyl-methyl (2275-105), cyclopentyl-methyl (2275-112), and cyclohexyl-propyl (2275-114) have chemically similar functionalities, and they all displayed similar activity profiles. Thus, rather than choosing all four functionalities, we selected cyclopentyl-methyl to represent this class. By selecting three functionalities at each position, we prioritized 27 individual compounds (3 functionalities × 3 positions) for synthesis.

Compounds with R-methyl in the R1 position (compounds #1–9, Table 1) had good activity at the highest dose (500 ng/mL) but did not maintain activity at lower doses compared to compounds with other R1 substitutions (compounds #10–27). Some preference for the cyclopently methyl in R3 is observed within this series and holds for all compounds (compounds 5 and 8). There is an obvious drop in activity for the R-isobutyl series (compounds 10–18) when an aromatic group (butylbenzene) is in the R3 position (compounds 12, 15, and 18). The substitution of an aromatic group (R-phenyl) did not decrease activity, however, when introduced in the R1 position (compounds 19–27). The R2 substitution of 4-methyl-1-cyclohexylmethyl within this series yielded the most potent compounds. The R-phenyl series maintained the preference for the cyclopently methyl substitution in the R3 position with the compound having these two substitutions (compound 26) showing the highest potency. Nine individual compounds from the set of 27 were purified and retested in the neurite outgrowth assay. All nine compounds induced outgrowth by at least 200% relative to controls (Table 2). Five compounds (compounds 14, 16, 23, 25, and 26) were retested over an extended dose range and verified to have nanomolar potency (Figure 3).

Table 1. Average %NTL Data for 27 Individual Compounds at Six Doses (ng/mL).

Table 2. Calculated Concentrations for the Doses Needed to Induce a %NTL of 200% (ED200%) and 300% (ED300%) for the Pure Individual Compounds^a.

cmpd ID	ED200% (nM)	ED300% (nM)
8	130	282
14	4	>543
16	10	53
17	4	25
18	25	461
23	7	91
25	5	54
26	2	>520
27	17	>471

^aValues were calculated by use of linear interpolates of each compound’s dose response curve.

Figure 3.

Neurite total length (%NTL) relative to DMSO in hippocampal neurons treated with pure 2539 compounds at the indicated concentrations for 2 DIV.

Mixture-based synthetic libraries are highly effective tools for generating novel lead compounds in a fraction of the time and cost required to screen the equivalent amount of individual compounds.²⁴ The approach has been successfully applied to a variety of assay types and was recently used to identify binders of the nicotinic acetylcholine receptor.²⁶ Here, we applied it to a phenotypic cell-based assay utilizing primary neurons and identified compounds with remarkable potencies for inducing neurite outgrowth.

Given the high potencies of the compounds discovered here, which are comparable to, or even exceed, the range of potencies for extracellular receptor ligands previously discovered using this screening approach,²⁶ it is tempting to speculate that our hits may be acting through an extracellular receptor. However, previous studies by our group and others have identified promoters of neurite outgrowth that exert their effects via intracellular targets (e.g., lipid metabolizing enzymes and kinases) with submicromolar potency.^19,23 This suggests that the hits reported in this study could also be exerting their effects via intracellular targets. Identification of targets by affinity purification followed by mass spectrometry is likely appropriate here and will be pursued in future studies to identify the putative mechanisms of action of these compounds.

The structure–activity relationship conclusions generated from the scaffold ranking library (i.e., that a branched imidazoline is a favored scaffold) and the positional scanning library (i.e., identifying key R group functionalities) indicate that the hit compounds are amenable to further optimization as they progress into more advanced in vivo and preclinical studies. The scaffold ranking library data indicate that the cyclic urea moiety could be modified as needed based on future efficacy or toxicity results.

In conclusion, we screened over 400 million compounds via the use of mixture-based libraries and identified novel and potent promoters of neurite outgrowth. Characterizing the physicochemical properties of these compounds, such as their membrane permeability, as well as identifying their high affinity binding partners, will be important for elucidating their mechanism(s) of action and potentially revealing novel drug targets for promoting CNS axon regeneration.

Experimental Procedures

Rabbit anti-βIII (T2200) was purchased from Sigma-Aldrich; Alexa Fluor 488 cross-linked Goat anti-Rabbit (A11034) antibodies were purchased from Life Technologies. Poly-d-Lysine (P7886-500MG) and sterile dimethyl sulfoxide (DMSO) (D2650) were purchased from Sigma-Aldrich. Hippocampal tissue was incubated in Hibernate E from BrainBits supplemented with NeuroCult SM1 (05711) from StemCell Technologies. Neurons were cultured in NbActive4 media from BrainBits.

Compound mixtures were screened in our previously described neurite outgrowth assay.^16,17,19 Briefly, dissociated embryonic (E18) rat hippocampal neurons were immediately seeded in 96-well plates (coated overnight with PDL at 0.5 mg mL^–1 in HBSS and washed 5 times with PBS) at 1800 cells per well in 150 ul of NbActive4 and allowed to adhere for 2 h before treatment. Compound mixtures were prediluted in culture media at 4× final concentration and equilibrated in a CO₂ incubator for at least 1 h prior to treatment, after which 50 μL of compound solution was moved onto the cells to a final experiment volume of 200 μL. Cells were incubated for 48 h after which plates were fixed, immunostained, and imaged in a Cellomics ArrayScan VTI robot. Images were automatically traced using the Neuronal Profiling Bioapplication (version 3.5). Mean neurite total length was calculated as a percentage of vehicle controls in each plate (%NTL). Conditions that reduced valid cell counts by more than 50% of control were considered toxic. ML-7 was used as the positive control, and experiments with Z-factor > 0.5 were considered valid. Each compound mixture was tested at least twice using independently prepared neuronal cultures on separate weeks. Dose–responses for individual hits were obtained using a similar assay format.

The compounds were synthesized as described in Scheme 1 using standard Boc chemistry. The solid-phase synthesis was performed using the “tea-bag” methodology.²⁷ The desired product was cleaved from the solid support resin and extracted using 95% acetic acid. Samples were then repeatedly frozen and lyophilized in 50% acetonitrile and water three times prior to analysis. During purification, the peak corresponding to the desired product with calculated m/z was isolated and concentrated. Compounds were purified using a Shimadzu Prominence preparative HPLC system consisting of LC-8A binary solvent pumps, an SCL-10A system controller, an SIL-10AP auto sampler, an FRC-10A fraction collector, and a Shimadzu SPD-20A UV detector. The wavelength was set to 214 nm. Chromatographic separations were obtained using a Phenomenex Gemini NX- C18 preparative column (5 μm, 150 mm × 21.2 mm i.d.). The column was protected by a Phenomenex C18 column guard (5 μm, 15 mm × 21.2 mm i.d.). Prominence prep software was used for detection and collection parameters. The mobile phases for HPLC purification were HPLC grade obtained from Sigma-Aldrich and Fisher Scientific. The mobile phase A consisted of water with 0.1% trifluoroacetic acid, and mobile phase B consisted of acetonitrile with 0.1% trifluoroacetic acid. The initial setting was set to 2% mobile phase B and was gradually increased over time to achieve ideal separation for each compound. See the Supporting Information for details on the synthesis of compounds 8, 14, 16, 17, 18, 23, 25, 26, and 27.

Scheme 1. Synthetic Route for Library 2275 and Individual Compounds.

¹H NMR spectra were obtained utilizing the Bruker 400’ Ascend (400 MHz). NMR chemical shifts were reported in δ (ppm) using the δ 2.50 signal of DMSO-d₆ as an internal standard. The ACD/NMR Processor Academic Addition Software package was utilized to aid in the integration and assignment of peak tables. Confirmation of the desired product was obtained by reverse-phase LCMS analysis using a Shimadzu 2010 LCMS system consisting of an LC-20AD binary solvent pump, a DGU-20A degasser unit, a CTO-20A column oven, an SIL-20A HT auto sampler, and an SPD-M20A diode array set to scan 190–600 nm. Separation was achieved using a Phenomenex Gemini C18 column (5 μm, 50 mm × 4.6 mm i.d.) protected with a Phenomenex C18 column guard column (5 μm, 4 × 3.0 mm i.d.), mobile (A) water with 0.1% formic acid and mobile phase (B) acetonitrile with 0.1% formic acid. A gradient of 5–95% mobile phase B over 6 min was run. See Supporting Information for LCMS and ¹H NMR spectral data for compounds 8, 14, 16, 17, 18, 23, 25, 26, and 27.

Glossary

Abbreviations

CNS

central nervous system

NTL

neurite total length

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.8b00425.

• Synthesis, LCMS, and ¹H NMR data for compounds 8, 14, 16, 17, 18, 23, 25, 26, and 27 and average %NTL data for the 116 mixture samples that make up library 2275 (PDF)

This work was supported by grants from the Department of Defense (W81XWH-13-1-077) and the National Institutes of Health (HD057632) to J.L.B. and V.P.L., the Wallace H Coulter Center (University of Miami) to J.L.B., V.P.L., and H.A.A., and the Florida Drug Discovery Acceleration Program by the State of Florida, Department of Health to G.D., R.G.S., R.A.H., J.C.D., and M.A.G. V.P.L. holds the Walter G. Ross Distinguished Chair in Developmental Neuroscience.

The authors declare the following competing financial interest(s): H.A.A., V.P.L., J.L.B., and M.A.G. are inventors on a provisional patent application covering the compounds reported in this manuscript.