Triazolothiadiazine derivative positively modulates CXCR4 signaling and improves diabetic wound healing

Abstract

Development of specific therapies that target and accelerate diabetic wound repair is an urgent need to alleviate pain and suffering and the huge socioeconomic burden of this debilitating disease. C-X-C Motif Chemokine Ligand 12 (CXCL12) also know an stromal cell-derived factor 1α (SDF-1α) is a chemokine that binds the CXC chemokine receptor type 4 (CXCR4) and activates downstream signaling resulting in recruitment of hematopoietic cells to locations of tissue injury and promotes tissue repair. In diabetes, low expression of CXCL12 correlates with impaired wound healing. Activation of CXCR4 receptor signaling with agonists or positive allosteric modulators (PAMs) provides a potential for small molecule therapeutic discovery and development. We recently reported high throughput screening and identification of the CXCR4 partial agonist UCUF-728, characterization of in vitro activity and reduced wound closure time in diabetic mice at 100 μM as a proof-of-concept study. We report here, the discovery of a second chemical scaffold demonstrating increased agonist potency and represented by thiadiazine derivative, UCUF-965. UCUF-965 is a potent partial agonist of β-arrestin recruitment in CXCR4 receptor overexpressing cell line. Furthermore, UCUF-965 potentiates the CXCL12 maximal response in cAMP signaling pathway, activates CXCL12 stimulated migration in lymphoblast cells and modulates the levels of specific microRNA involved in the complex wound repair process, specifically in mouse fibroblasts. Our results indicate that UCUF-965 acts as a PAM agonist of the CXCR4 receptor. Furthermore, UCUF-965 enhanced angiogenesis markers and reduced wound healing time by 36% at 10.0 μM in diabetic mice models compared to untreated control.

1. Introduction

Impaired wound healing, especially in diabetes, results in significant morbidity, hospitalization, surgery, pain, suffering and socioeconomic impact with annual expenditure in wound healing and associated products exceeding US $3 billion [1]. Due to a dearth of specific therapies, there is intense interest in developing treatments that are clinically effective for diabetic wound repair and can have far-reaching consequences on patient outcomes and healthcare expenditure [2]. The G-protein coupled receptor (GPCR) CXC motif chemokine (CXCR4) and its endogenous ligand CXCL12 are known to play a crucial role in several physiological processes and are dysregulated in diseases such as inflammation, cancer, HIV and diabetes [3]. CXCR4 and CXCL12 are upregulated during injury, hypoxia, stress and vascular tissue damage [4]. Wound healing is a systematic and synchronized process involving epithelialization, angiogenesis, granulation tissue formation, and wound contraction. It requires the interaction of many cell types, growth factors, extracellular matrix (ECM) proteins, and enzymes [5]. The CXCL12/CXCR4 axis is essential during wound healing for activating chemotaxis to damaged tissues, cell proliferation for repair and collagen deposition for tissue remodeling [4]. Moreover, diabetic wounds are known to have significantly lower expression (mRNA and protein) of CXCL12 [6]. Overexpression of CXCL12 results in greater granulation tissue, a smaller epithelial gap, and wound size, whereas expression of a mutant form of CXCL12 competitively inhibits activation of CXCR4 by endogenous CXCL12 and significantly increases inflammation, decreases angiogenesis and impairs the rate of wound healing in the diabetic mouse [6-8].

AMD3100 (Plerixafor), a FDA approved small molecule antagonist of CXCR4/CXCL12 signaling has been reported to accelerate wound healing as a single topical application and when delivered on an acellular dermal matrix [9-10]. However, this action has been attributed to enhanced expression of CXCL12 and an increase in CXCR4⁺ cells at the wound site as an outcome of AMD3100 antagonism [10]. AMD3100 has also been reported to be an antagonist for inhibiting G-protein signaling but an agonist for stimulation of β-arrestin recruitment and receptor internalization [11]. AMD11070 a full antagonist does not exhibit hematopoietic stem cell (HSC) mobilization [11]. Moreover, in a pilot, phase IIa, double-blind, randomized, placebo controlled trial, AMD3100 failed in the primary endpoint of diabetic wound healing and was inferior to placebo control despite successful hematopoietic stem/progenitor cells mobilization in patients [12]. ATL-234, a Pepducin peptide derived from the intracellular loop of CXCR4 was reported as an agonist of G-protein-dependent signaling, receptor internalization and chemotaxis in CXCR4-expressing cells [13]. ATL-234 treatment led to dose-dependent recruitment of neutrophils when delivered intraperitoneally attributed to agonist mode of action. However, when delivered as an i.v. bolus, ATL-234 acted as a functional antagonist and led to release of neutrophils from bone marrow niches in mice and monkeys. There are no reports on stability of these peptides in vivo or their use in diabetic wound healing. CXCL12 fits inside a large binding site (2049 Å) making it a challenge to develop orthosteric small molecule agonists that are selective, have high potency and binding response [14]. NUCC-390 is one such orthosteric agonist that was found as an agonist of CXCR4 signaling without any added CXCL12 [14]. NUCC-390 was obtained via structure based virtual screening of a commercial database and the same data set resulted in agonists and antagonists based on subtle changes around the molecule. NUCC-390 (10 μM) induces calcium-flux, receptor internalization, ERK activation and induced comparable chemotaxis in C8161 cells vs CXCL12, however no lead development and in vivo studies in wound healing models are reported to the best of our knowledge.

Hence, small molecule orthosteric or allosteric agonists and activators of endogenous CXCL12 binding and CXCR4 signaling hold tremendous therapeutic promise in diabetic wound healing, especially if developed as topical therapy without associated systemic activation [15]. Small molecule therapeutics are also desirable due to the proteolytic nature of wounds throughout the healing process which limit application of peptide therapeutics and other biomolecules [16].

To identify small molecule CXCR4 activators for pathologic wound healing in diabetes, we conducted high-throughput screening (HTS) of the NIH Molecular Library Small Molecule Repository (MLSMR) using a Florescence Resonance Energy Transfer (FRET) based β-arrestin recruitment assay. We recently reported the results of the HTS screen and subsequent identification of a small molecule hit UCUF-728 [17]. UCUF-728 confirmed activation of β-arrestin recruitment (EC₅₀ = 0.5 μM and E_max = 30%) and equipotent activation in lymphoblast migration, normalized to treatment with CXCL12. Importantly, the CXCL12-mediated response was potentiated in the presence of saturating concentrations of UCUF-728, suggesting that UCUF-728 is a potential positive allosteric activator. UCUF-728 reduced overexpression of miRNA-15b and miRNA-29a, negative regulators of angiogenesis and type I collagen production respectively in diabetic fibroblasts. In a proof-of-concept experiment, UCUF-728 (100 μM) reduced the wound closure time in vivo by 36% and increased the evidence of angiogenesis in diabetic mice [17]. However, UCUF-728 chemical scaffold did not result in tractable structure–activity relationship (SAR). The lead compound’s weak potency and potential for off-target activity at high concentrations needed to elicit wound healing response in mice limit its utility. Herein, we report a SAR study focused on a distinct triazolothiadiazine scaffold also identified in our HTS screen. Optimization of this scaffold resulted in the lead compound UCUF-965, which activated β-arrestin recruitment, albeit with reduced potency and efficacy compared to CXCL12, activated migration in the sub micromolar range at 10-fold higher potency than UCUF-728 and exhibited agonist activity in CXCR4-mediated cAMP and calcium signaling. We present the synthesis, medicinal chemistry, SAR and in vitro and in vivo biological profile of UCUF-965, a promising positive allosteric modulator of CXCR4 signaling to treat diabetic would healing.

2. Materials and methods

2.1. General chemisty considerations

Small molecule derivatives were synthesized and purified at the College of Pharmacy, University of Florida, Gainesville, Florida, The Chemical Genomics Center at Sanford Burnham Medical Discovery Institute, Orlando, Florida, USA and Bio-Duro-Sundia, China. Building blocks, reagents and starting compounds were purchased from Millipore-Sigma, St. Louis, MO, USA and Enamine Monmouth Junction, New Jersey, USA. Biotage initiator + was used for microwave accelerated synthesis. All reactions were monitored by thin-layer chromatography (TLC) on silica gel (Merck, 60 Å F-254) TLC plates using UV light as the visualizing agent. Flash column chromatography were performed on silica gel (Merck, 60 Å 0.049–0.063 mm). Biotage Selekt flash chromatography system adaptable to normal and reverse-phase separation and purification, Biotage V-10 Touch and a Buchi RII rotovap for microscale and larger scale evaporation of solvents and isolation of compounds. NMR analysis was done on AVANCE NEO 600 MHz NMR spectrometer equipped with PRODIGY probe and N2-Cryo-Platform. ¹H and ¹³C NMR spectra were recorded on Bruker 500 MHz or 600 MHz spectrometer. Data for ¹H NMR (CDCl3 referenced at δ 8.07) and ¹³C NMR (CDCl3 referenced at δ 77.16) reported as following: multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, hept = heptet, dd = doublet of doublets, and m = multiplet), integration, coupling constant (Hz), and chemical shift (ppm). Low resolution mass spectrometry (LRMS) data were recorded by electrospray ionization (ESI) on a Waters Acquity UPLC I-class Core RPHPLC-MS system equipped with PDA and QDA detectors for reaction monitoring and assessing purity of small molecules. All synthesized compounds and commercially purchased compounds that were used in biological screening were ≥ 95% in purity as established by RPHPLC-MS.

2.2. Commercially sourced analogs

Compounds 1–10 and 13–22 were procured from commercial sources including ChemDiv, San Diego, USA and Princeton Biomolecular, Princeton, New Jersey, USA. Compounds 11–12, 23–29 and 39 were synthesized using the representative procedure in section 2.3 at BioDuro-Sundia, Shanghai, China. Purity was established to be > 95% by LCMS and compound powders were used as is for biological screening.

2.3. Representative procedure for Synthesis of 3-(2-ethoxyphenyl)-6-(4-(pyrrolidin-1-yl)phenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4]thiadiazine, compound 36 (UCUF-965) and analogs

A microwave reaction tube was charged with 4-amino-5-(2-ethoxyphenyl)-4H-1,2,4-triazole-3-thiol (0.25 g, 1.06 μmol) and 2-bromo-1-(4-(pyrrolidin-1-yl)phenyl)ethan-1-one (0.284 g, 1.06 μmol). To the mixture, 3 mL acetonitrile and the tube was sealed. The reaction was heated at 120 °C for 1 h. A yellow brown precipitate was observed. The seal was opened and reaction mixture was cooled to room temperature. 2–3 mL ethyl acetate was added to precipitate more product. The product was filtered and washed extensively with ethyl acetate (5 mL, 5x) and air-dried to yield compound 36, 3-(2-ethoxyphenyl)-6-(4-(pyrrolidin-1-yl)phenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4] thiadiazine (UCUF-965) as a bright yellow solid (0.362 g, 84%). ¹H NMR (600 MHz, DMSO) δ 7.76 – 7.67 (m, 2H), 7.57 – 7.47 (m, 2H), 7.17 (d, J = 8.3 Hz, 1H), 7.09 (td, J = 7.4, 0.9 Hz, 1H), 6.61 – 6.56 (m, 2H),4.27 (s, 2H), 3.98 (q, J = 6.9 Hz, 2H), 3.31 – 3.26 (m, 4H), 3.16 (s, 1H), 1.99 – 1.91 (m, 4H), 1.03 (t, J = 6.9 Hz, 3H).13C NMR (151 MHz, DMSO) δ 156.83, 155.38, 150.02, 142.61, 132.67, 131.26, 129.02, 120.33, 118.32, 114.28, 112.61, 111.60, 63.73, 48.59, 47.25, 24.92, 22.56, 14.29. LRMS, ESI(+ve) calculated for C₂₂H₂₃N₅OS, [M + H] = 406.17, observed [M + H] = 406.15.

2.4. Synthesis of analogs

i). Synthesis of compounds 30, 31, 32, 33–38, 40–42.

Starting from 4-amino-5-(2-ethoxyphenyl)-4H-1,2,4-triazole-3-thiol (0.400 g, 1.69 μmol) and tert-butyl (4-(2-chloroacetyl)phenyl)carbamate (0.457 g, 1.69 μmol) using the representative procedure 2.3 for the synthesis of compound 36 to obtain 4-(3-(2-ethoxyphenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4]thiadiazin-6-yl)aniline hydrobromide, 33 as an orange-brown solid (200 g, 78%). In a glass vial, compound 33, was weighed in followed by addition of 1:4 of acetic acid:THF and 5–10 equivalents of the corresponding carbonyl compound, acetal or ketal. After stirring for 10–15 min, excess sodium acetoxyborohydride (10 equivalents) was added and mixture stirred overnight. The reaction mixture was extracted with ethyl acetate and organic layer was dried and evaporated and residue was purified by preparative TLC or flash chromatography.

Compound 30:

Starting from 33 (15 mg, 43 umol) and cyclobutanone (10 uL, 25 equiv.), the product N-cyclobutyl-4-(3-(2-ethoxyphenyl)-7H-[1–2,4]triazolo[3,4-b] [1,3–4]thiadiazin-6-yl)aniline, 30 (5.5 mg) was obtained in 32% yield. ¹H NMR (600 MHz, CDCl3) δ 7.71 – 7.66 (m, 1H), 7.67 – 7.63 (m, 2H), 7.48 (ddd, J = 8.4, 7.4, 1.7 Hz, 1H), 7.09 (td, J = 7.5, 1.1 Hz, 1H), 6.98 (dd, J = 8.4, 1.1 Hz, 1H), 6.55 – 6.51 (m, 2H), 4.02 – 3.93 (m, 3H), 3.88 (s, 2H), 2.51 – 2.42 (m, 2H), 1.94 – 1.78 (m, 4H), 1.13 (td, J = 6.9, 2.2 Hz, 3H). ESI-MS: Calculated for C₂₂H₂₃N₅OS, [M + H] = 406.17 and observed [M + H] = 406.17.

Compound 31:

Starting from 33 (22 mg, 63 umol) and cyclobutanone (28 uL, 5 equiv.), the product N-cyclopentyl-4-(3-(2-ethoxyphenyl)-7H-[1–2,4]triazolo[3,4-b] [1,3–4]thiadiazin-6-yl)aniline, 31 (15.3 mg) was obtained in 58% yield. ¹H NMR (600 MHz, CDC13) δ 7.69 (dd, J = 7.5, 1.8 Hz, 1H), 7.65 (d, J = 8.8 Hz, 2H), 7.48 (ddd, J = 8.3, 7.5, 1.8 Hz, 1H), 7.09 (td, J = 7.5, 1.0 Hz, 1H), 6.98 (dd, J = 8.4, 1.0 Hz, 1H), 6.58 (d, J = 8.8 Hz, 2H), 3.98 (q, J = 7.0 Hz, 2H), 3.88 (s, 2H), 3.85 (s, 1H), 2.12 – 2.01 (m, 2H), 1.79 – 1.71 (m, 2H), 1.71 – 1.63 (m, 2H), 1.54 – 1.46 (m, 2H), 1.13 (t, J = 6.9 Hz, 3H). ESI-MS: Calculated for C₂₃H₂₅N₅OS, [M + H] = 420.19 and observed [M + H] = 420.20.

Compound 32:

Starting from 33 (22 mg, 63 umol) and cyclohexanone (30 uL, 5 equiv.), the product N-cyclohexyl-4-(3-(2-ethoxyphenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4]thiadiazin-6-yl)aniline, 32 (13.6 mg) was obtained in 55% yield. 1H NMR (600 MHz, CDCl3) δ 7.69 (dd, J = 7.6, 1.8 Hz, 1H), 7.64 (d, J = 8.9 Hz, 1H), 7.48 (ddd, J = 8.3, 7.5, 1.8 Hz, 1H), 7.09 (td, J = 7.5, 1.0 Hz, 1H), 7.01 – 6.96 (m, 1H), 6.57 (d, J = 8.9 Hz, 1H), 4.06 (s, 1H), 3.98 (q, J = 7.0 Hz, 2H), 3.88 (s, 2H), 3.33 (s, 1H), 2.06 (d, J = 13.1 Hz, 3H), 1.80 (dt, J = 13.6, 3.9 Hz, 2H), 1.72 – 1.64 (m, 1H), 1.13 (t, J = 7.0 Hz, 3H), 0.90 (t, J = 7.0 Hz, 2H). ESI-MS: Calculated for C₂₄H₂₇N₅OS, [M + H] = 434.20 and observed [M + H] = 434.20.

Compound 37:

Starting from 33 (25 mg, 71 umol) and 1,1,3,3-tetraethoxypropane (68 uL, 4 equiv.), the product 6-(4-(azetidin-1-yl)phenyl)-3-(2-ethoxyphenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4]thiadiazine, 37 (12.2 mg) was obtained in 44% yield. ¹H NMR (600 MHz, CDCl3) δ 7.72 – 7.64 (m, 3H), 7.48 (ddt, J = 10.0, 7.4, 2.0 Hz, 1H), 7.09 (tdd, J = 7.4, 2.1, 1.0 Hz, 1H), 6.98 (dt, J = 8.5, 1.5 Hz, 1H), 6.68 – 6.64 (m, 1H), 6.61 – 6.57 (m, 1H), 4.02 – 3.94 (m, 2H), 3.88 (s, 2H), 3.75 – 3.69 (m, 2H), 3.44 (q, J = 7.1 Hz, 1H), 3.40 – 3.34 (m, 1H), 3.23 (t, J = 6.8 Hz, 1H), 1.79 – 1.74 (m, 1H), 1.13 (dt, J = 10.4, 7.0 Hz, 3H). ESI-MS: Calculated for C₂₁H₂₁N₅OS, [M + H] = 392.15 and observed [M + H] = 392.18.

ii). Synthesis of Compound 34:

A 8 mL glass vial was charged with 33 (25 mg, 71 umol), 4-dimethylaminopyridine (26 mg, 213 umol) and 2 mL acetonitrile. Acetyl chloride (10uL, 2 equiv.) was added and mixture stirred overnight at room temperature. The reaction mixture was diluted with ethyl acetate (2 mL) and washed with 0.1 M HCl 5–6 times and organic layer collected and evaporated. The residue was purified by preparative TLC (50% ethyl acetate/hexanes) to obtain the product N-(4-(3-(2-ethoxyphenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4]thiadiazin-6-yl)phenyl)acetamide, 34 (20 mg, 72%). ¹H NMR (600 MHz, CDC13) δ 7.81 – 7.75 (m, 3H), 7.71 – 7.65 (m, 3H), 7.50 (ddd, J = 8.3, 7.5, 1.8 Hz, 1H), 7.10 (td, J = 7.5, 1.0 Hz, 1H), 6.99 (dd, J = 8.4, 1.0 Hz, 1H), 4.01 – 3.92 (m, 4H), 2.23 (s, 3H), 1.11 (t, J = 7.0 Hz, 3H). ESI-MS: Calculated for C₂₀H₁₉N₅O₂S, [M + H] = 394.13 and observed [M + H] = 394.14.

iii). Synthesis of compound 35:

A 8 mL glass vial was charged with 33 (15 mg, 43 μmol), 4-dimethylaminopyridine (10.4 mg, 85 μmol) and 2 mL acetonitrile. Isobutyryl chloride (7uL, 1.5 equiv.) was added and mixture stirred overnight at room temperature. The reaction mixture was diluted with ethyl acetate (2 mL) and washed with 0.1 M HCl 5–6 times and organic layer collected and evaporated. The residue was purified by preparative TLC (60% ethyl acetate/hexanes) to obtain the product N-(4-(3-(2-ethoxyphenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4] thiadiazin-6-yl)phenyl)acetamide, 35 (11 mg, 64%). ¹H NMR (600 MHz, CDCl3) δ 7.83 – 7.74 (m, 3H), 7.73 – 7.64 (m, 3H), 7.52 – 7.46 (m, 1H), 7.10 (tt, J = 7.5, 1.1 Hz, 1H), 6.98 (d, J = 8.3 Hz, 1H), 3.95 (d, J = 8.3 Hz, 4H), 2.58 (pd, J = 6.8, 1.0 Hz, 1H), 1.27 (dd, J = 6.9, 1.9 Hz, 7H), 1.10 (td, J = 7.0, 1.1 Hz, 3H). ESI-MS: Calculated for C₂₂H₂₃N₅O₂S, [M + H] = 422.17 and observed [M + H] = 422.18.

iv). Synthesis of compound 38:

Starting from 4-amino-5-(2-ethoxyphenyl)-4H-1,2,4-triazole-3-thiol (25 mg, 106 umol) and 2-bromo-1-(4-(piperidin-1-yl)phenyl)ethan-1-one (30 mg, 106 μmol) using the representative procedure 2.3 to obtain 3-(2-ethoxyphenyl)-6-(4-(piperidin-1-yl)phenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4]thiadiazine, 38 (21 mg, 47%). ¹H NMR (600 MHz, CDCl3) δ 7.69 (tt, J = 5.5, 1.8 Hz, 3H), 7.48 (ddd, J = 8.8, 5.3, 1.6 Hz, 1H), 7.09 (tdd, J = 7.6, 2.0, 1.0 Hz, 1H), 6.98 (d, J = 8.3 Hz, 1H), 6.89 (dd, J = 9.2, 2.6 Hz, 2H), 3.98 (qd, J = 7.0, 1.7 Hz, 2H), 3.90 (d, J = 1.7 Hz, 2H), 3.34 (t, J = 5.4 Hz, 4H), 1.75 – 1.62 (m, 6H), 1.13 (td, J = 6.9, 2.0 Hz, 3H). ESI-MS: Calculated for C₂₃H₂₅N₅OS, [M + H] = 420.19 and observed [M + H] = 420.21.

v). Synthesis of compound 40:

An 8.0 mL vial under nitrogen was charged with 33 (50 mg, 142 μmol) and 3 mL dry dichloromethane and Rh₂(esp)₂ catalyst (5.4 mg, 5 mol%). Ethyl diazoacetate (60 mL, 3.5 equiv) was added as a dichloromethane solution in 1 equivalent portion followed by stirring at room temperature for 5–10 min till no further consumption of starting material was detected. The solvent was evaporated and residue was purified by preparative TLC to obtain ethyl (4-(3-(2-ethoxyphenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4] thiadiazin-6-yl) phenyl)glycinate, 40 (34.5 mg, 55%). ¹H NMR (600 MHz, CDCl3) δ 7.74 – 7.61 (m, 3H), 7.51 – 7.45 (m, 1H), 7.09 (tdd, J = 7.5, 3.1, 1.0 Hz, 1H), 6.98 (ddd, J = 8.3, 4.1, 1.0 Hz, 1H), 6.71 – 6.58 (m, 2H), 4.80 (s, 1H), 4.26 (dq, J = 25.0, 7.2 Hz, 2H), 4.18 (s, 1H), 4.00 – 3.93 (m, 4H), 3.89 (d, J = 3.1 Hz, 2H), 1.31 (dt, J = 16.2, 7.1 Hz, 3H), 1.12 (t, J = 7.0 Hz, 3H). ESI-MS: Calculated for C₂₂H₂₃N₅O₃S, [M + H] = 438.16 and observed [M + H] = 438.13.

vi). Synthesis of compound 41:

A 8.0 mL glass vial was charged with 40 (14 mg, 32 μmol) and Lithium hydroxide monohydrate (4 mg, 96 μmol) followed by addition of 0.5 mL ethanol and 2 mL water. The mixture was stirred overnight and diluted with 2 mL ethyl acetate and 1 mL water. The organic layer was separated, dried and evaporated to obtain the product (4-(3-(2-ethoxyphenyl)-7H- [1–2,4]triazolo[3,4-b] [1,3–4]thiadiazin-6-yl)phenyl)glycine, 41 (11 mg, 84%). 1H NMR (600 MHz, MeOD) δ 7.75 – 7.69 (m, 2H), 7.58 – 7.51 (m, 2H), 7.17 – 7.07 (m, 2H), 6.70 – 6.57 (m, 2H), 4.24 (s, 1H), 4.20 – 4.15 (m, 2H), 4.01 (q, J = 7.0 Hz, 2H), 3.94 (s, 1H), 1.12 (td, J = 7.0, 2.2 Hz, 3H). ESI-MS: Calculated for C₂₀H₁₉N₅O₃S, [M + H] = 410.13 and observed [M + H] = 410.14.

vii). Synthesis of compound 42:

Starting from 4-amino-5-(2-ethoxyphenyl)-4H-1,2,4-triazole-3-thiol (50 mg, 212 μmol) and 1-(4-(2-bromoacetyl)phenyl)pyrrolidin-2-one (60 mg, 212 μmol) using the representative procedure 2.3 to obtain 1-(4-(3-(2-ethoxyphenyl)-7H-[1–2,4]triazolo[3,4-b] [1,3–4]thiadiazin-6-yl)phenyl)pyrrolidin-2-one, 42 (74 mg, 83%). ¹H NMR (600 MHz, CDCl3) δ 7.83 – 7.79 (m, 2H), 7.78 – 7.73 (m, 2H), 7.68 (dd, J = 7.5, 1.8 Hz, 1H), 7.48 (ddd, J = 8.2, 7.4, 1.8 Hz, 1H), 7.09 (td, J = 7.5, 1.0 Hz, 1H), 6.97 (dd, J = 8.4, 1.0 Hz, 1H), 3.98 – 3.92 (m, 5H), 3.89 (t, J = 7.1 Hz, 2H), 2.65 (t, J = 8.1 Hz, 2H), 2.24 – 2.16 (m, 2H), 1.09 (t, J = 7.0 Hz, 3H). ESI-MS: Calculated for C₂₂H₂₁N₅O₂S, [M + H] = 420.15 and observed [M + H] = 420.20.

2.5. Reagents and drug treatment protocol

Test compounds were prepared as 10 mM DMSO stocks and stored at −80 °C. Test compounds were dispensed using the TECAN D300e digital dispenser to assay plates to determine concentration–response effects in the respective cell-based assays. DMSO (Sigma Aldrich) concentration was backfilled in all wells to a constant concentration and did not exceed 0.5% in any cell-based assay. All cell-lines used were from human origin and purchased from commercial vendors as indicated. Cells were expanded and frozen as aliqouts between passage 2–4. After thaw, cell passage did not exceed 10 passages and in most cases did not exceed 6 passages.

2.6. β-Arrestin-2 recruitment

Test compounds and DMSO were dispensed using the TECAN D300e digital dispenser into 384-well black wall clear-bottom microplates (Corning; 3764) as previously described [17]. In brief, Tango^™ CXCR4-bla U2OS Cells (ThermoFisher Scientific Cat# K1779) were resuspended in Freestyle 293 assay media at a density of 3.0 × 10⁵ cells/mL and dispensed at 30 μL per well into the assay plate. Test compounds (10 mM stock in DMSO) were dispensed using the TECAN D300e digital dispenser to assay plates to determine concentration–response effects. Cells and test compounds were incubated overnight in a cell culture incubator. DMSO (Sigma Aldrich) concentration was backfilled in all wells to 0.5%. CXCL12 ligand, prepared in assay media, (PeproTech; 30028A) was dispensed manually the next day and cells were incubated for additional 5 hr in a cell culture incubator. After incubation, Live-Blazer FRET B/G Loading Kit (Invitrogen; K1095) working reagent was prepared according to supplier guidelines and 8 μL of solution was dispensed into each well. Assay plate was covered and incubated in the dark for 2 hr and fluorescence intensity was measured using the BMG LABTECH CLARIOstar Plus (BMG LABTECH Inc., Cary, NC) according to the FRET Loading Kit excitation and emission measurement guidelines.

2.7. Migration assay

CEM-CCRF cells (ATCC CCL-1119) were suspended in RPMI 1640 at 5.0 × 10⁵ cells/mL and incubated for 1 hr at 37°C as previously described [17]. In brief, agonist activity was determined for CXCL12 ligand (PeproTech; 300-28A) or test agonist in the presence of 0.5% DMSO by dispensing agonist into the receiver tray of a Multi-Screen 96-well assay plate (Millipore Sigma; MAMIC3S10) in concentration response using the TECAN D300e digital dispenser. Following incubation, 150 μL of RPMI 1640 supplemented with 2% Fetal Bovine Serum was added to each well of the receiver tray and the cell suspension was added to each well of the top filter plate. CEM cells were added to top portion of the filter plate as previously described¹¹. Plates were reassembled and placed in the cell culture incubator for 3 hr. After incubation, the Multi-Screen plate was disassembled and viable cells was quantified by ATPLite-1step reagent (Perkin Elmer; 6016731) luminescence measured using the BMG LABTECH CLARIOstar Plus (BMG LABTECH Inc., Cary, NC).

2.8. Adenylate cyclase assay

Compounds and DMSO normalization to 0.5% were dispensed using the TECAN D300e digital dispenser into the wells of a 384-well white wall clear-bottom assay plate (Griener 781983). CHO-CXCR4-cAMP (DiscoverX, 95-0081C2) or CHO-CXCR6-cAMP (PerkinElmer, ES-720C) cells maintained in monolayer culture were prepared in DMEM/F12 (Corning; 10–091-CV) supplemented with 2% Fetal Bovine Serum (Corning; 35–011-CV) at a density of 0.15 × 10⁶ cells/mL. This cell suspension was dispensed at 20 μL per well into the assay plate. Test compounds (10 mM stock in DMSO) were dispensed using the TECAN D300e digital dispenser to assay plates to determine concentration–response effects. Cells and test compounds were incubated overnight in a cell culture incubator. DMSO (Sigma Aldrich) concentration was backfilled in all wells to 0.5%. Following incubation, DMEM/F12 media was gently removed, then the Lance Ultra cAMP Kit (Perkin Elmer; TRF0263) was performed according to the supplier guidelines. Forskolin was dispensed at 300 nM to elicit cAMP production and CXCR4 natural ligand CXCL12 (Peprotech; 300-28A) was added in titration to determine the maximal cAMP inhibition. For the CXCR6 receptor assay, forskolin was dispensed at 500 nM to elicit cAMP production and CXCR6 natural ligand CXCL16 (Peprotech;300–55) was added in titration to determine the maximal cAMP inhibition. Time-resolved fluorescence was measured on the BMG LABTECH CLARIOstar Plus (BMG LABTECH Inc., Cary, NC) using the Lance Ultra cAMP kit guidelines for excitation and emission.

2.9. Calcium flux activity

Fluo-4 Direct Calcium Assay Kit (Invitrogen; F10472) reagents were prepared according to the supplier guidelines at 2X concentration and allowed to equilibrate to room temperature in the dark. CCRF-CEM cells (ATCC CCL-1119) in fresh assay media were used to prepare a new 4.5 mL cell suspension at a density of 4.8 × 10⁵ cells/mL. An equal volume of 2X Fluo-4 Direct Calcium Assay Kit reagent was added to this cell suspension and placed in a cell culture incubator at for 1 hr. Following incubation, 25 μL of cell suspension loaded with Fluo-4 Direct Calcium reagent was dispensed into a 384-well black wall clear-bottom plate (Corning; 3764). For screening CXCL12 control ligand and test compounds, stock solutions were prepared in assay buffer consisting of Calcium Direct buffer supplemented with 0.5 mg/ml BSA and injection volumes were limited to 2 μL. For antagonist response measurements, assay wells were pre-treated with titrations of antagonist AMD3100 (TOCRIS; 3299) using the TECAN D300e digital dispenser 10 min prior to agonist injections. Calcium flux responses were measured on the BMG LABTECH CLARIOstar Plus (BMG LABTECH Inc., Cary, NC) using the CLARIOstar Plus supplier excitation and emission settings for Fluo-4 (Calcium Saturated). Measurements were conducted using a bottom-read kinetic well-mode with measurements occurring at 0.5 sec intervals for 150 sec and injection occurring at 10.0 sec. Peak values were determined by area under the curve (AUC) analysis with baseline set as mean of first 21 (10.0 sec.) data points.

2.10. Binding activity

Binding activity was measured using the Cisbio Tag-Lite Chemokine CXCR4 system (Cisbio; C1TT1CXCR4) as previsouly described [17]. In brief, Tag-Lite Chemokine CXCR4 labelled cells, red fluorescent labelled CXCR4 ligand (Cisbio; L0012RED), and 1X Tag-Lite buffer (Cisbio; LABMED) was dispensed into each assay well of 384-well microplate (Greiner Bio; 784075) and incubated for 3 hr in the dark with test compounds or AMD3100. Saturation binding and competition binding experiments were conducted on the same plate (estimated Kd = 12.5 nM for fluorescent ligand). HTRF measurements were quantified using the BMG LABTECH CLARIOstar Plus (BMG LABTECH Inc., Cary, NC).

2.11. Real time quantitative PCR

Primers and probes for mouse miR-15b and miR-29a were obtained from Applied Biosystems TaqMan gene expression assay (Applied Biosystems, Foster City, CA). Quantitative PCR was performed on a BIO-RAD CFX96 as previously described [17]. Results are reported as mean ± SD.

2.12. Animal studies

As previously described [17], age-matched (10-week-old) female (Db/Db) mice and heterozygous, non-diabetic (non-Db) controls from the Jackson Laboratory (Bar Harbor, ME) were approved for use in these experiments by the Institutional Animal Care and Use Committee at the University of Colorado Denver - Anschutz Medical Campus. In brief, 50 mL of UCUF-965 or PBS was injected into a dorsal wound generated in each mouse with a 8-mm punch biopsy. Ten nanoliters were injected intradermally at different time periods at the wound base after which, wounds were dressed. Postoperatively, the mice received a subcutaneous injection of an analgesic. Skin samples, centered on the wound, were harvested 3 and 7 days after surgery.

2.13. Statistical analysis and curve fitting

All concentration response curves were analyzed to determine EC₅₀ and E_max as previsouly described [17]. 100% activity for migration data refers to activity induced by EC₉₀ concentration of CXCL12 alone; 0% activity in this assay refers to the activity of wells left untreated with CXCL12 or compounds. Unless otherwise noted, EC₅₀ and E_max values reported in the tables for structure–activity relationship studies were determined from an average of two non-linear regression analysis determined on separate days and each non-linear regression analysis was an average of two dose–response datasets. UCUF-965 was tested as a control on each plate for quality assurance and to track compound potency over time. Data displayed in Figs. 1 and 5 is the non-linear regression analysis of an average of three separate dose–response curves determined on separate days. The error bars are displayed as the standard deviation. Each test compound was tested in two separate concentration–response mode on the day of the assay and averaged and fit to a non-linear regression analysis. Calcium flux peak values were generated using AUC analysis with baseline set as mean of first 21 rows of data and peaks defined only above baseline. Additional parameters included ignoring peaks less than 50% of the distance from minimum to maximum Y and ignoring peaks defined by less than 10 adjacent points. Following AUC analysis, peak Y for the first generated peak for each experiment was plotted to generate response values for AMD inhibition curve. Plotted data for calcium flux AMD inhibition was analyzed following the same equation used to evaluate dose–response curves in additional experiments.

Fig. 1. UCUF-965 is a partial allosteric activator of CXCR4 cell signaling.

A) UCUF-965 shows dose-dependent partial activation of β-arrestin recruitment in CXCR4-bla U2OS cells (red squares) compared to full activation by CXCL12 (blue circles). AMD3100 at 1 μM completely blocks the CXCL12 β-arrestin recruitment (navy triangles). B) CXCL12 dose-dependent β-arrestin recruitment in the absence (blue circles) and presence of UCUF-965 at 0.1 μM (green squares), 0.5 μ M (purple down triangles), 1 μ M (red triangles), 5 μ M (black squares) and 10 μ M (orange diamonds). C) β-arrestin recruitment: Ratios of half-efficient concentrations EC₅₀/EC′₅₀ extracted expressed as logarithms (blue squares, left y-axis) and observed maximal responses E′_max /E_max (red circles, right y-axis) are plotted against log concentrations of UCUF-965. D) UCUF-965 (red squares) shows dose-dependent partial inhibition of forskolin-stimulated cAMP production in CXCR4-CHO-K1 cells. CXCL12 shows dose-dependent full inhibition of forskolin-stimulated cAMP production (blue circles) and this activity is inhibited in the presence of 1 μM AMD 3100 (navy diamonds). E) CXCL12 dose-dependent inhibition of forskolin-induced cAMP in the absence (blue circles) and presence of UCUF-965 at 0.1 μM (green squares), 1 μ M (red triangles), 5 μ M (black squares) and 10 μ M (orange diamonds). F) cAMP inhibition: Ratios of half-efficient concentrations EC₅₀/EC′₅₀ expressed as logarithms (blue squares, left y-axis) and observed maximal responses E′_max /E_max (red circles, right y-axis) are plotted against log concentrations of UCUF-965.

Fig. 5. UCUF-965 activates endogenous CXCR4 cell signaling in CEM-CCRF lymphoblasts.

A) CXCL12 dose-dependent cell migration in in CEM-CCRF cells shows a bell-shaped curve (closed circles) that is shifted to the right in the presence of 5 μM AMD 3100 (closed squares). UCUF-965 dose-dependent cell migration in CEM-CCRF cells (closed circles) indicates a weak agonist activity. B) 10 nM CXCL12 (black line) and 10 μM UCUF-965 (gray line) induction of CXCR4-mediated calcium flux in CEM-CCRF cells. Dotted lines indicate SD of triplicate biological determinations. C) AMD 3100 dose-dependently inhibits 10 nM CXCL12 (closed circles) or UCUF-965 (open circles) induced calcium flux. Error bars are +/− SD of triplicate determination.

3. Results

3.1. Synthesis and Hit-to-lead SAR studies leading to UCUF-965

Our SAR strategy was based on assessing analogs for activation of β-arrestin recruitment in CXCR4-bla U2OS cells as primary assay and migration in CEM-CCRF lymphoblast cell line as a secondary assay to confirm functional activity. Active analogs were subsequently profiled for their ability to agonize downstream G-protein signaling including inhibition of cAMP production and stimulation of calcium-flux. We accessed the SAR analogs utilizing a combination of commercially available analogs (Table 1) and synthesized analogs. The synthesis of analogs was accomplished by condensation of suitably substituted 4-amino-(2-alkoxyphenyl)-4H-1,2,4-triazole-3-thiols and α-chloro or bromoacetophenones under microwave irradiation at 120 °C for 1 h following previously reported procedures [18]. To understand the importance of R¹ vs R² methoxy groups towards activity, we first stripped the scaffold to the unsubstituted analog which was inactive (Table 1, entry 2). Adding a 2-methoxy group at R¹ position alone did not result in activity, however 4-methoxy substituent at R² position did result in weak activity (entries 3 and 4 vs 1) suggesting that contribution of methoxy group at R² was more significant, however both R¹ and R² substituents were required for activity of primary hit entry 1.

Table 1.

SAR of thiadiazine scaffold

Entry	R1	R2	β-arrestin recruitment
			EC50 (uM)	Emax (%)
1	2-OMe	4-OMe	0.054	45
Primary hit
2	H	H	>80	NA
3	2-OMe	H	>80	NA
4	H	4-OMe	11.2	54
5	2-OMe	4-fluoro	>80	NA
6	2-OMe	4-chloro	4.14	44
7	2-OMe	4-bromo	1.51	37
8	2-OMe	4-nitro	>80	NA
9	2-OMe	4-methyl	0.268	42
10	2-OMe	4-ethyl	0.203	44
11	2-OMe	4-n-Bu	14.1	12
12	2-OMe	4-Ph	8.35	95
13	2-OMe	3-OMe	10.7	40
14	2-OMe	2-OMe	>80	NA
15	2-OMe	2,4-dimethoxy	0.866	23
Primary hit
16	3-OMe	4-OMe	2.51	63
17	4-OMe	4-OMe	6.71	65
18	2-Cl	4-OMe	0.282	43
19	3-Cl	4-OMe	>80	NA
20	4-Cl	4-OMe	7.75	44
21	2-OEt	4-OMe	0.13	60
22	2-OEt	3,4-dimethoxy	0.571	42
Primary hit

We next screened a range of substituents at R² position with 2-OMe group constant at R¹ position and results indicate a preference for small electron-donating alkyl substituents such as methyl and ethyl groups over electron withdrawing groups, larger butyl, or a phenyl substituent (entries 5–8, 11–12 vs 9–10). Moving the 4-methoxy group at R² to 3- and 2- positions led to a 200-fold loss in activity and an inactive analog, respectively (entries 13 and 14). The result agrees with a 16-fold loss in activity for the 2,4-dimethoxy substituent at R² position and suggest a narrow binding space around the R²-phenyl group (entry 15). At the R¹ position, moving the 2-methoxy to 3- and 4- position led to 50 to 100-fold loss, respectively (entries 16 and 17). Replacement of electron-donating methoxy with electron withdrawing Cl- group at R¹ position was detrimental to activity of the primary hit 1 (entries 18–20). A 5-fold improvement was obtained with a larger 2-ethoxy substituent at R¹ position (entry 21). Addition of a 3-methoxy group to analog 21 at R² position led to a 60-fold loss in activity, highlighting the steric restriction around the R² phenyl group.

Focusing on the 4-methoxy substituent at R² position, we found that having the unsubstituted phenol or butyl, allyl or methoxyethyl ether at 4-position led to loss in activity compared to primary hit 1 or analog 21 (Table 2, entries 23–26 vs Table 1, entries 1 and 21). Tuning the electronic nature of 4-methoxy with a 4-trifluormethoxy or difluoromethoxy ether did not lead to improvements in potency (Table 2, entries 27–28). Replacement of the 4-methoxy group at R² position with 4-N, N-dimethylamino group at R² position led to complete loss of activity (entry 29). However, secondary amines retained activity (entries 30–32). The results suggest a polar or H-bonding interaction aided by the basic nitrogen atom may be involved at R₂ position. Amides (entries 34 and 35) were inactive suggesting a requirement for a basic amine. Replacement of 4-N, N-dimethylamino group with cyclic tertiary amines, resulted in compound 36, with a five membered 4-pyrrolidinyl group that was the most potent analog. The smaller four membered 4-azetidinyl or larger six membered 4-piperdinyl or 4-mormpholinyl groups led to less potent or inactive analogs, highlighting specific nature of interaction with the target in this region (entries 36 vs 37–39). Further, β-alanine ester derived analog 41 was equipotent to 36 and offers promise for further exploration by replacement of the ester with amide or other polar functional groups and with added diversity. In addition, compound 42, a cyclic amide analog of compound 36 was equipotent and suggesting accommodation for a polar hydrogen bonding group. These diverse analogs (entries 36, 41 and 42) are promising and suggest that further optimization of the scaffold is possible.

Table 2.

SAR of thiadiazine scaffold

Entry	R1	R2	-arrestin recruitment
			EC50 (uM)	Emax (%)
23	2-OEt	4-OH	1.28	96
24	2-OMe	4-On-Bu	0.7	31.3
25	2-OEt	4-OCH2CH = CH2	0.20	95
26	2-OMe	4-OCH2CH2OCH3	3.43	28
27	2-OEt	4-OCF3	1.76	83
28	2-OEt	4-OCHF2	0.3	48
29	2-OMe	4-NMe2	ND	~25
30	2-OEt	4-NH-cyclobutyl	0.33	77
31	2-OEt	4-NH-cyclopentyl	1.4	31
32	2-OEt	4-NH-cyclohexyl	0.77	10
33	2-OEt	4-NH2	>10	N/A
34	2-OEt	4-NHCOCH3	>10	N/A
35	2-OEt	4-NHCOCH(CH3)2	>10	N/A
36	2-OEt	4-pyrrolidinyl	0.02	44
UCUF-965
37	2-OEt	4-azetidinyl	0.10	40
38	2-OEt	4-piperidnyl	>10	N/A
39	2-OMe	4-morpholinyl	7.81	96
40	2-OEt	4-NHCH2CO2Et	>5
41	2-OEt	4-NHCH2CO2H	0.03	43
42	2-OEt	4-(2-oxo-pyrrololidinyl)	0.05	22

Compound 36 or UCUF-965 is the most potent among the small molecule agonists of CXCR4/CXCL12 signaling for β-arrestin-2 recruitment with an EC₅₀ = 0.02 μM (log −7.7 ± 0.1 M, n = 3) and average E_max = 44% normalized to maximal CXCL12 response (Fig. 1A). We used AMD3100 as a control antagonist in the assay and at 1 μM, the CXCR4 antagonist abolishes CXCL12 β-arrestin response. The functional response of CXCL12 was tested in the absence or presence of 0.1, 0.5, 1.0, 5.0, and 10 μM UCUF-965 and the individual curves were fit to nonlinear regression analysis (Fig. 1B). To determine the effect on the coopertivity factors α on CXCL12 potency and β on CXCL12 efficacy, we plotted the log (EC₅₀/EC₅₀′) and the E’_max/E_max ratios, respectively versus the concentration of UCUF-965 (Fig. 1C) [19]. The EC₅₀ and E_max values are the response to CXCL12 in the absence of the modulator and the EC₅₀′ and E_max’ values are the response to CXCL12 in the presence of the modulator. UCUF-965 decreased the E_max’ ratios indicating β value less than 1. At UCUF-965 concentrations ≥1 μM, the log (EC₅₀/EC₅₀′) was equal to zero indiating an α value = 1 and no change in CXCL12 potency in the presence of UCUF-965. The E’_max/E_max ratio decreased to 0.4–0.5 response at 5 and 10 μM. At these concentrations, we start to see toxicitiy in the cells that may account for the decrease in response. These results indicate that UCUF-965 is a partial agonist of CXCR4-mediated β-arrestin activity. Based on the agonist potency in the β-arrestin assay, UCUF-65 was chosen as a lead molecule for further biological characterization and in vivo proof of concept studies in a murine wound healing model.

3.2. UCUF-965 is a potential allosteric activator of CXCR4/CXCL12 signaling

CXCL12 potently inhibited forskolin-stimulated cAMP production by adenylate cyclase in CXCR4 overexpressing CHO cells as a measure of its activity on CXCR4 mediated G-protein coupled signaling (G_i) pathway (IC₅₀ = 1.5 nM, log −8.9 ± 0.1 M) and this response was attenuated in the presence of 1 μM AMD3100 with a 20-fold rightward shift in IC₅₀ and reduction of E_max to 40% (Fig. 1D). We evaluated UCUF-965 in the cAMP assay normalized to the maximal response of CXCL12 and the compound showed dose-dependent partial agonist response with EC₅₀ = 0.05 μM (log −7.3 ± 0.2 M, n = 3) and average E_max = 19% of cAMP inhibition (Fig. 1D). Pretreatment with increasing concentrations of UCUF-965 of 0.1, 1.0, 5.0 and 10 μM potentiated the cAMP maximal inhibitory response (E_max) of CXCL12 by 128, 200, 208 and 195%, respectively but did not significantly affect the EC₅₀ value of CXCL12 (Fig. 1E). UCUF-965 showed saturation between 1 and 10 μM concentration, indicative of insurmountable activation of CXCR4/CXCL12 signaling. CXCL12 in the presence of 10 μM 965 shows varabliity at low concentrations of CXCL12 and high concetnrations of 965. We choose to test at this concentration to match the concentration tested in the in vivo would healing assay. Collectively, the dataset indicates 965 potentiates CXCL12 maximal response in cAMP siganling.

We performed the same analysis as shown for β-arrestin activity to determine UCUF-965′s effect on the coopertivity factors α and β for CXCL12 induced cAMP inhibition. Fig. 1F shows the results for log (EC₅₀/EC₅₀′) and E’_max/E_max ratios plotted versus the concentration of UCUF-965. The E’_max/E_max plot confirms β > 1.0 and significant potentiation of CXCL12 maximal response in the presence of UCUF-965. At 10 μM UCUF-965, the log (EC₅₀/EC₅₀′) value is equal to zero indicating α value = 1. To confirm that UCUF-965 does not potentiate the potency of CXCL12, we tested the compound in the presence of an EC₂₀ concentration of CXCL12 and did not observe an effect (Fig. 2). These results further support UCUF-965 as a partial agonist of CXCR4-mediated signaling and a modulator of CXCL12 efficacy. Conversly to the compound’s effect on the maximal ligand response in the β-arrestin assay, UCUF-965 acts as a positive modulator of CXCL12 maximal response in the cAMP inhibiton assay.

Fig. 2. Characterization of UCUF-965 potentiation by CXCL12.

UCUF-965 is a partial agonist of CXCR4-mediated cAMP inhibition (open circles). UCUF-965 does not show potentiation in the presence of an EC₂₀ concentration of CXCL12 (closed circles).

As a counterscreen assay, we performed the same cAMP detection but in CHO-K1 cells overexpressing the CXCR6 receptor. The CXCR6 natural ligand, CXCL16 was used as a positive control and UCUF-965 did not show an agonist response in the CXCR6 receptor-mediated cAMP assay compared to CXCL16 nor did the compound potentiate the CXCL16 dose–response curve in this assay system (Fig. 3). Next, to determine whether UCUF-965 binds to a different site on the CXCR4 receptor than the orthosteric site, we tested UCUF-965 in a competitive binding assay in Tag-Lite Chemokine CXCR4 labelled cells (Fig. 4). UCUF-965 did not displace fluorescently labelled CXCL12 binding at a constant concentration of 12.5 nM, whereas AMD3100 dose-dependently displaced CXCL12 with an IC₅₀ = 110 nM (K_i = 55.5 nM), a potency comparable to literature values [20]. These results confirm that UCUF-965 does not bind orthosterically to the CXCL12 site on the CXCR4 receptor and supports UCUF-965 as a positive allosteric modulator (PAM) agonist of CXCR4 cAMP siganling.

Fig. 3. UCUF-965 selectivity: comparison with CXCR6 receptor expression.

A) CXCL16 dose-dependently inhibited forskolin-induced cAMP in CHO cells overexpressing the CXCR6 receptor (closed circles). UCUF-965 does not exhibit agonist response in these cells (open circles). B) UCUF-965 at 0.1 (open circles), 1.0 (closed squares), and 10 (open squares) μM does not exhibit significant potentiation of CXCL16 dose-dependent cAMP activity (closed circles) in CHO cells overexpressing the CXCR6 receptor.

Fig. 4. UCUF-965 binding is non-orthosteric.

CXCL12 saturation binding in Tag-Lite CXCR4 cells in the presence of varying concentrations of AMD3100 (open circles) or UCUF-965 (closed squares). Data is average of triplicate biological determinations tested in duplicate ± standard deviation.

CXCL12 stimulates chemotaxis in a high percentage of resting and active T lymphocytes, and the CXCR4 receptor is highly expressed in the CEM-CCRF (Leukemia) cell line [21-22]. To confirm functional activity, we evaluated UCUF-965 in a transwell migration assay utilizing CEM-CCRF human lymphoblast cells (Fig. 5A). UCUF-965 induced migration at EC₅₀ = 0.4 μM (log −6.4 ± 0.3 M, n = 3) and average E_max = 31% indicating 10-fold lower potency in the endogenous assay compared to the compound’s activity in the β-arrestin recruitment assay. AMD3100 completely blocks CXCL12 induced migration as previously reported¹⁷. Further, In CEM-CCRF cells, 10 μM UCUF-965 induces CXCR4-mediated calcium flux in comparison to 10 nM CXCL12 (Fig. 5B). Pre-treatment with AMD 3100 resulted in dose-dependent inhibition of CXCL12-induced calcium mobilization at an IC₅₀ = 630 nM (Fig. 5C). In the case of downstream calcium mobilization, AMD 3100 also dose-dependently inhibited UCUF-965 mediated activation at 10-fold higher potency (IC₅₀ = 60 nM, Fig. 5C), reflecting that UCUF-965 is a weaker agonist compared with the endogenous CXCR4 ligand, CXCL12. This lends further proof for mode of action of UCUF-965 as a PAM agonist of CXCR4 signaling.

We selected four active analogs of UCUF-965 (compounds 30, 37, 41 and 42, Table 2) from the SAR campaign for inhibition of forskolin stimulated cAMP production and stimulation of migration of CEM-CCRF human lymphoblast cells to select the best allosteric activator lead molecule for in vivo proof of concept studies of wound healing in diabetic mice. All the four leads were selected based on diverse substructure at the R₂ substituent and showed partial agonist activity in both the cAMP and migration assays (Table 3) like UCUF-965, however with lower potency than UCUF-965. Hence, we advanced UCUF-965 as our most promising lead for further assessment of ability to modulate wound healing specific micro-RNA levels ex vivo in murine diabetic fibroblasts and POC studies in diabetic mice.

Table 3.

Activity profile of select lead compounds

Entry	R1	R2	cAMP EC50 (uM)	Emax (%)	Migration CEM
					EC50 (uM)	Emax (%)
36	2-OEt	4-pyrrolidinyl	0.05	19	0.40	31
30	2-OEt	4-NH-cyclobutyl	0.4	19	0.2	16
37	2-OEt	4-azetidinyl	0.16	22	0.18	12
41	2-OEt	4-NHCH2CO2H	1.6	17	0.32	23
42	2-OEt	4-(pyrrolidin-2-one)	0.08	18	0.22	32

3.3. Ex vivo validation of UCUF-965 in modulating levels of micro-RNA

We measured the expression of specific micro-RNA levels reported to be crucial in the wound-healing process in order to validate and corelate the in vitro activity of UCUF-965 as a selective CXCR4 receptor modulator to micro-RNA expression. Murine diabetic and non-diabetic fibroblasts, which may themselves secrete CXCL12, were cultured with increasing concentrations of the compound. The cells were incubated for 24 h and then total cellular RNA was isolated to examine the ability of the compound to correct the abnormal expression levels of miR-15b, miR-29a, and miR-146a [23-25]. First, we examined the effect of UCUF-965 on expression of miR-15b, which inhibits angiogenesis and wound repair. Our previous studies show that diabetic wounds have increased miR-15b expression at the baseline compared to non-diabetic skin [23]. We found that UCUF-965 treatment decreased the expression of miR-15b in both diabetic and nondiabetic fibroblasts in a dose dependent manner (Fig. 6). In preliminary studies, we have shown that miR-29a and miR-146a, which control collagen production and pro-inflammatory pathways, respectively are significantly dysregulated in diabetic wounds [24-25]. Here, we found that that UCUF-965 resulted in a significant decrease of miR-29a levels and increase of miR-146a levels (Fig. 6). These data suggest that UCUF-965 may potentiate low levels of secreted CXCL12 via the CXCR4/CXCL12 axis.

Fig. 6. UCUF-965 modulates wound healing specific miRNA levels in diabetic murine fibroblasts.

Quantitative polymerase chain reaction (qPCR) evaluation of miR15b, miR29a, and miR146a expression was evaluated in diabetic (gray bars) and non-diabetic (HZ, black bars) murine fibroblasts for non-treated, DMSO treated, and treated with 0.1 μM, 1.0 μM, and 10 μM UCUF-965. Target miRNA expression is normalized to the expression of the housekeeping miRNA U6.

3.4. In vivo activity of UCUF-965 in a mouse model of wound healing

We examined the ability of our lead compound UCUF965 to improve diabetic wound healing in vivo. Full-thickness excisional 8 mm wounds were created in a diabetic murine model and were immediately treated with 10 μM UCUF-965 or PBS control. We monitored wound healing over the course of 22 days. Initial wound size was calculated immediately after wounding, and wound closure was assessed over time as the percentage of initial wound area. Fig. 7A depicts the wound healing effect of UCUF-965 application on the excision wound model for sequential days following injury. By post-injury day 6, diabetic wounds treated with UCUF-965 exhibited a decrease in the wound surface area compared to wounds treated with PBS. The time of full closure was 14 days compared to 22 days in wounds treated with PBS, indicating that UCUF-965 treatment in diabetic wounds enhanced wound healing by 36% at 10 μM over PBS control (Fig. 7B).

Fig. 7. UCUF- 965 enhances diabetic wound healing.

A) Representative images of diabetic murine wounds during the healing process after PBS or UCUF-965 (10 μM) treatments. B) Wounds were significantly smaller starting at 6 days after wounding following treatment with UCUF-965 compared to PBS-treated controls (**p less than 0.01). UCUF-965-treated diabetic wounds closed around 14 days after treatment whereas control PBS-treated wounds healed at day 22.

3.5. UCUF-965 enhances angiogenesis in diabetic wounds

The effect of UCUF-965 treatment on angiogenesis was assessed using immunohistochemistry for the endothelial marker CD31. Representative photos of immunoperoxidase staining for CD31 at 7 days in non-diabetic control wounds from heterozygous mice treated with PBS and diabetic wounds treated with PBS or treated with UCUF-965 are shown in Fig. 8A. Fig. 8B shows the quantification of the staining. Our results support that diabetic wounds treated with PBS have significantly less CD31 positive cells per high power field compared to non-diabetic wounds treated with PBS. In contrast, diabetic wounds treated with UCUF-965 significantly increased the number of vessels compared with PBS-treated diabetic wounds to similar levels as control wounds suggesting that UCUF-965 stimulates angiogenesis in the wound healing model.

Fig. 8. Diabetic wounds treated with UCUF-965 demonstrated increased angiogenesis.

A) Representative images of CD31 + staining in day 7 non-diabetic, diabetic wounds, and diabetic wounds treated with UCUF-965 (10x) from heterozygous mice. B) Diabetic wounds treated with UCUF-965 demonstrated increased staining of CD31 + cells compared to diabetic wounds treated with PBS (P less than 0.05).

4. Discussion

Normal secretion of the chemoattractant CXCL12, activation of its chemokine receptor, CXCR4, [6-8] and the subsequent recruitment of hematopoietic progenitor cells to areas of tissue injury is essential for effective wound healing [26-27]. In diabetic wounds, decreased expression of CXCL12 leads to reduced cellular migration and impaired wound healing. This results in an increased wound closure time, decreased granulation tissue, and a larger epithelial gap compared to non-diabetic wounds [23-25]. In addition, non-diabetic mouse wounds lacking CXCL12 and lymphocyte recruitment recapitulate many features of impaired wound healing, including preferential M1 polarization, increased basal ROS levels, and reduced angiogenesis [28]. Novel therapeutics that can circumvent deficits in CXCL12 levels and/or CXCR4 receptor activation to correct healing impairment have a great potential for clinical use and commercial drug development.

Allosteric modulators targeting GPCRs such as CXCR4 are attractive because they act at sites separate from the orthosteric ligand binding site to modulate endogenous ligand activity. Since they do not compete with the natural ligand, their effect may be saturable upon occupation of all allosteric sites on the target [15]. CXCR4 is widely expressed in various tissues and promotes proliferation. By preserving endogenous receptor-ligand signaling and not overstimulating the receptor, allosteric positive modulators may potentially offer improved selectivity and safety over orthosteric agonists [29]. This is especially critical for receptor sub-types that have a common ligand, share conserved sequences around the ligand binding orthosteric site, form heterodimers with other chemokine receptors and critically regulate normal physiology. In addition, allosteric modulators offer unique modes of action that control only selective functions of a receptor, expanding the range of therapeutic utility.

We have previously developed a high-throughput β-arrestin recruitment assay to screen compounds for CXCR4 receptor activators and have recently reported in vivo proof-of concept studies with UCUF-728 [17]. We have identified a second more potent and tractable scaffold represented by the lead molecule UCUF-965 via the HTS screen and subsequent SAR studies and medicinal chemistry optimization of the hit analogs. UCUF-965, is a partial agonist of CXCR4 receptor activation, determined by β-arrestin recruitment, inhibition of cAMP stimulation and calcium flux measurement. UCUF-965 also activates migration in human CCRF-CEM cell line.

To determine the mode of action of UCUF-965, we tested the compound in the presence of increasing concentrations of CXCL12 in the suite of assays we established for SAR screening. UCUF-965 did exhibit competitive inhibition of CXCL12 in the β-arrestin and cAMP signaling assays typical of a partial agonist binding orthosterically in the presence of a full agonist. UCUF-965 did not competitively displace fluorescently labeled CXCL12 binding to CXCR4 expressing cells, indicating non-orthosteric binding and mode of action. Moreover, we found that pre-treatment with UCUF-965 potentiated CXCL12-mediated inhibition of cAMP production upon stimulation by forskolin (G_i signaling) by doubling the E_max indicating a β coopertivity value > 1 but not significantly affecting the EC₅₀ (α value = 1) compared to treatment by CXCL12 alone. UCUF-965 depressed the maximal efficacy of CXCL12 β-arrestin recruitment at concnetrations ≥1 μM. In both the cAMP and β-arrestin assays, the treatment time for UCUF-965 is 18 hr. It is probable that UCUF-965 at high concentration causes internalization of the receptor in the β-arrestin recruitment assay, although a partial agonist should cause less internalization than a full agonist. In addition, AMD3100 dose-dependently inhibited CXCL12 and UCUF-965 mediated calcium flux with IC₅₀ values of 60 nM and 640 nM, respectively, further supporting that UCUF-965 is a CXCR4 receptor modulator. The data collectively supports that UCUF-965 interacts with the CXCR4 receptor as a PAM agonist of CXCR4/ CXCL12 signaling.

We assessed functional activity of UCUF-965 in a transwell migration assay utilizing CEM-CCRF human lymphoblast cells and found that UCUF-965 induced migration with 10-fold lower potency compared to its activity in the β-arrestin recruitment assay and AMD3100 was unable to inhibit UCUF-965-induced migration consistent with the allosteric mode of action proposed for UCUF-965. Treatment of murine diabetic fibroblasts with UCUF-965 resulted in the induction of miR146a and suppression of miR15b and miR29a expression, with response observed even at the lowest concentration tested (0.1 μM). MiR-15b is a negative modulator of angiogenesis and is upregulated in diabetic wounds during the early phase of healing [23,30]. This results in decreased expression of pro-angiogenic target genes, including vascular endothelial growth factor (VEGF), hypoxia inducible factor (HIF-1), and B-cell lymphoma 2 (BCL2) [23]. MiR-29a is upregulated in diabetic wounds which potentially leads to decreased collagen I content in diabetic wounds and delayed healing [24]. MiR146a inhibits inflammation and is down-regulated in diabetic wounds [25]. UCUF-965 may contribute towards accelerated wound closure by modulation of miR-15b, miR29a and miR146a expression in diabetic wounds among others.

In order to test this hypothesis, we evaluated the efficacy of UCUF-965 in accelerating diabetic wound healing in a mouse model. Murine dorsal wounds, as described above, were allowed to heal completely with pictures taken daily to measure wound size until full closure. Diabetic wounds treated with PBS healed at day 22, while wounds treated with 10 μM UCUF-965 healed 36% faster at day 14. This preliminary study demonstrates that UCUF-965, may act as a PAM agonist in the presence of low circulating levels of CXCL12 signaling to induce wound healing in diabetic mice model and serve as an in vivo lead for further optimization and preclinical studies.

While antagonists at the CXCR4 receptor (agonist on β-arrestin pathway) such as AMD3100 have shown promise in diabetic wound healing, the mode of action has been suggested to be an increase in expression of CXCL12. However, AMD3100 failed in comparison to placebo in a phase II study of diabetic wound healing. Based on current data, our PAM agonist, UCUF-965 potentiates the action of endogenous CXCL12. Further studies are needed that can compare PAM agonist and antagonists directly. In addition, the exact site of binding of UCUF-965 to CXCR4 receptor is not known at this time. In future we aim to conduct molecular modeling studies to identify the most likely site of binding of UCUF-965 to CXCR4 receptor and optimize the scaffold for more potent analogs. We also plan to utilize the current SAR to develop suitable photo-crosslinking probe from UCUF-965 and subsequent proteomics-based identification of the binding site in future.

This preliminary proof of concept study clearly demonstrates that UCUF-965, a PAM agonist of CXCR4/CXCL12 signaling can significantly reduce wound healing in diabetic mice model. The chemical scaffold is synthetically tractable and our study provides the basis for further lead optimization and therapeutic development. We are currently working on improving potency, aqueous solubility and evaluating skin-based formulations for profiling in biomarker assays and ultimately further drug development via animal models of diabetic wound healing.

Abbreviations:

cAMP

cyclic adenosine monophosphate

CXCL12

C-X-C Motif Chemokine Ligand 12

AUC

area under curve

BCL2

B-cell lymphoma 2

BSA

bovine serum albumin

CD31

cluster of differentiation 31

CXCR4

chemokine receptor type 4

DMSO

dimethyl sulfoxide

HZ

heterozygous

EC₅₀

half-maximal effective concentration

Emax

maximal effect 50%

ECM

extracellular matrix

ERK

extracellular signal-regulated kinases

FRET

Florescence Resonance Energy Transfer

GPCR

G-protein coupled receptor

HTS

high-throughput screening

HSC

hematopoietic stem cell

HIV

human immunodeficiency virus

HIF-1

hypoxia inducible factor

i.v

intravenous

IC₅₀

half-maximal inhibitory concentration

MLSMR

Molecular Library Small Molecule Repository

PAM

potential allosteric modulator

ROS

reactive oxygen species

SDF-1α

stromal cell derived factor alpha

VEGF

vascular endothelial growth factor

Data availability

No data was used for the research described in the article.