A c-Cbl/Cbl-b antagonist inhibits EGFR ubiquitylation and sustains EGFR phosphorylation to enhance corneal re-epithelialization

Significance

Damage to the cornea is painful and can lead to ocular infections and potentially blindness. Homeostasis and healing of the outermost epithelial layer is governed by growth factor receptors. The epidermal growth factor receptor (EGFR) is sufficient and necessary for maintaining the corneal epithelium, but the clinical use of exogenous EGF is limited by desensitization of the EGF:EGFR complex. Here, we describe a compound that prevents receptor desensitization by antagonizing EGFR:c-Cbl interactions. This decreases ligand-mediated ubiquitylation of the EGFR and slows trafficking to the lysosome for degradation. The resulting increased and sustained EGFR signaling accelerates corneal re-epithelialization.

Abstract

Growth factor receptor signaling is a critical component of tissue growth, homeostasis, and wound healing. However, receptor desensitization limits the use of exogenous growth factors as a restorative agent therapeutically. An example of this is the epidermal growth factor receptor (EGFR) in the corneal epithelium. Despite laboratory data indicating that EGFR activity accelerates corneal re-epithelialization in mice and rabbits, the clinical administration of EGF to damaged corneal epithelium has limited impact due to the attenuated signaling that occurs following sustained growth factor administration. We hypothesized that inhibition of receptor desensitization would prolong receptor activity and enhance tissue homeostasis. Having previously identified the E3 ubiquitin ligase, c-Cbl, as a key negative regulator of EGFR signaling in the corneal epithelium, we have developed a class of small molecule inhibitors of EGFR binding to CBL family proteins using virtual screening with experimental validation. Through multiple rounds of structural optimization, we have identified compound 3-120. This compound was designed to compete with phosphotyrosine 1045 of the EGFR for binding to c-Cbl. Compound 3-120 binds to c-Cbl with an ~10-fold higher affinity than phosphoEGFR, reduces EGFR ubiquitylation by 40%, and increases the magnitude of ligand-stimulated EGFR phosphorylation by 30 to 40%. Ultimately, this compound can enhance the restoration of corneal epithelial debridement wounds. Thus, compound 3-120 is an antagonist that specifically disrupts EGFR ubiquitylation to sustain receptor signaling.

If an epithelial tissue is damaged by trauma (physical, thermal, chemical, etc.) or disease, stimulating growth-promoting receptors with exogenous ligand can often accelerate tissue regeneration (1, 2). However, this tissue restoration can be limited due to receptor desensitization (3, 4). That is, the added ligand stimulates the receptor to produce a robust response that is rapidly attenuated and returns to near basal levels of signaling. We hypothesize that the complementary strategies of exogenous ligand and antagonizing receptor desensitization can prolong receptor signaling and accelerate epithelial regeneration.

The endocytic pathway is the primary mechanism for attenuating cell surface receptor signaling, particularly among receptor tyrosine kinases. Following ligand binding, the ligand:receptor complex enters the endocytic pathway and traffics through the early and late endosomes until it enters the lysosome for degradation (5). For many receptor tyrosine kinases, the posttranslational modification ubiquitylation directs the internalization and lysosomal degradation of the active ligand:receptor complex. Inhibition of receptor ubiquitylation prevents receptor degradation by diverting it from the lysosome (6).

A clear example of this is the epidermal growth factor receptor (EGFR) as a mediator of corneal epithelial homeostasis and restoration. EGFR activity is sufficient to promote the re-epithelialization of mechanically debrided corneas in laboratory studies (7) and patients taking EGFR inhibitors (i.e., erlotinib) have an increased incidence of corneal erosions (8–10). However, the clinical use of EGF is limited by its lack of reliably sustaining signaling (11–16). Previous work has shown that EGFR endocytosis and degradation is regulated by receptor ubiquitylation (3, 17). Using RNAi and CRISPR/CAS9 to attenuate expression of c-Cbl in corneal epithelial cells, we observed a partial decrease in EGFR ubiquitylation and an increase in the magnitude and duration of EGFR phosphorylation (3, 17). The phosphorylated EGFR is able to serve as a docking site for downstream signaling molecules that activate effectors including ERK1/2 MAPK, p38 MAPK, and PI3K which mediate cell migration and proliferation (18, 19).

Based on these findings, we hypothesized that pharmacologic inhibition of ubiquitylation is a viable strategy for sustaining EGFR signaling with the potential for accelerating corneal re-epithelialization (5). A well-formed corneal epithelium is critical for visual clarity as well as preventing the entry of infectious agents (i.e., bacteria, viruses, fungi) and subsequent loss of sight. Further, when the corneal epithelium is not fully formed deleterious growth factors made in the tear fluid, such as transforming growth factor beta (TGFβ), can enter the stroma and lead to scarring and fibrotic tissue that impact visual clarity. In addition, persistent corneal defects are associated with recurrent corneal erosions, epithelial thinning, stromal edema, and can progress to dry eye disease (20, 21).

To develop a small molecule inhibitor of EGFR ubiquitylation, we used the Zinc15 library and computer modeling to source over 24 million molecules, prefiltered for their “drug-likeness,” to identify compounds that could disrupt EGFR:c-Cbl interactions using computer modeling. Compounds with the highest predicted affinity score from docking were biochemically tested for their ability to bind c-Cbl, inhibit ligand-mediated EGFR ubiquitylation, and sustain EGFR phosphorylation. The top compound was structurally modified to enhance binding and maximize efficacy. Here, we describe an inhibitor (compound 3-120) that antagonizes ligand-mediated EGFR ubiquitylation, increases the magnitude and duration of EGFR phosphorylation, and accelerates EGFR-mediated corneal re-epithelialization.

Materials and Methods

Cell Culture.

Telomerase reverse transcriptase-immortalized human corneal epithelial [hTCEpi, (22)] cells were obtained from Evercyte (Vienna, Austria). Cells were grown in Keratinocyte Basal Medium (KBM-2) (#CC-3103, Lonza, MD) with growth supplements (hydrocortisone, transferrin, epinephrine, BPE, hEGF, and insulin, #CC-4152, Lonza) and Pen Strep (#15140-122, Thermo-Fisher, MA) at 37 °C with 5% CO₂. Cells were propagated twice a week to maintain normal growth and never allowed to grow more than 90% confluent to prevent differentiation and quiescence of cells. c-Cbl, Cbl-b, and c-Cbl/Cbl-b (DKO) double knockout cells were generated as described in (23). Primary human corneal epithelial cells were isolated from human corneas that could not be used for transplantation from the Kentucky Eye Bank (4).

Cell Viability Assays.

Parental (CAS9) and c-Cbl/Cbl-b knockout (DKO) hTCEpi cell lines were measured for viability following incubation with varying concentrations of EGF. Following a 72-h incubation, alamarBlue^TM (ThermoFisher Scientific, Waltham, MA) was added and the number of viable cells were measured in accordance with manufacturer’s recommendation.

Radioligand Trafficking.

Was performed as described previously (24). Thirty-five (3) mm dishes of hTCEpi cells were incubated with 1 ng/mL of ¹²⁵I-EGF (catalog number NEX160; PerkinElmer Life Sciences; specific activity 150 to 200 μCi/μg) for 2 h on ice to achieve steady-state binding. Following the removal of unbound radioligand with four washes of ice-cold binding buffer, cells were incubated with 37 °C binding buffer and incubated at 37 °C. At the indicated times, cells were removed from the incubator. Media were collected. Cell surface ¹²⁵I-EGF was collected by pooling two 8-min washes of the cells with 0.5 M acetic acid/0.5 M NaCl. Internalized ¹²⁵I-EGF was collected by solubilizing the remaining cells with 0.1 M NaOH/0.1% SDS. Each fraction was counted in a Wizard gamma counter (PerkinElmer). Rate constants were calculated as described (25).

Protein Purification.

GST-c-Cbl was a kind gift of Dr. Stan Lipkowitz (NIH) (26). GST-Cbl-b was obtained from the Medical Research Council Protein Phosphorylation and Ubiquitylation Unit (pGEX6P-2-CBLB, University of Dundee, Scotland). Full-length GST-c-Cbl and GST-Cbl-b were purified using glutathione agarose beads according to the manufacturer’s instructions (GE Healthcare, Chicago, IL).

Compound Screen.

Initial compounds were identified using an in silico screen of compounds as previously described (27). Briefly, 24,877,119 compounds from the ZINC15 library were screened at the binding site for the phosphorylated EGFR peptide (amino acids 1063–1078 DSFLQRpYSSDPTG corresponding EGFR amino acids 1063–1078) using the program Surflex-Dock v2.601 (BioPharmycs). The 71 compounds with the lowest theoretical free energy for binding at the pEGFR peptide binding site were purchased and tested for binding to c-Cbl experimentally. Of the six compounds that significantly bound c-Cbl from preliminary biophysical screens, two were found to sustain EGFR signaling and inhibit ligand-mediated EGFR ubiquitylation in cells.

Chemistry.

Compound AF (3-(1H-indol-3-yl)-N-(4-methyl-2,5-dioxo-4-phenethylimidazolidin-1-yl) propenamide) was purchased from MolPort SIA (Riga, Latvia) and was used without further purification. Compound 3-120 and other derivatives were prepared by a modified literature procedure (28). See SI Appendix, Supporting Text for 3-120 experimentals. Details of AF deratives can be found at Open Science Framework (https://osf.io/yspr5/overview?view_only=07141c5df2ba4081a5e6a7b047fb9ad2) (29).

Differential Scanning Fluorimetry (DSF) Assays.

DSF assays were performed using an Applied Biosystems StepOne Plus real-time PCR system using the adapted protocol (30). Briefly, recombinant purified c-Cbl protein was diluted to 20 µM in PBS buffer, mixed with 1X Sypro Orange dye, and its melt transition midpoint (T_m) was measured in a 96-well PCR plate (20 µL final volume) in the presence of either DMSO (control) or experimental compounds from virtual screening.

Microscale Thermophoresis (MST).

MST experiments were performed on a Monolith Nanotemper instrument (Nanotemper, München, Germany) using its MO.Control v2.0.4 software. The purified protein was exchanged into PBS buffer with 1 mM TCEP prior to labeling with the Nanotemper RED-NHS 2nd generation lysine red kit protocol. In all cases, labeling efficiency was greater than 50%. Titration solutions consisted of 40 nM Lysine Red-labeled c-Cbl protein mixed in 12- or 16-point serial dilution series of compounds. The concentration series ranged from 100 μM to 0.31 nM of ligand or peptide with a matched background of 8% DMSO. Dilution reactions were made in 384-well plates at a volume of 20 μL and incubated in the dark for at least 10 min before loading into Nanotemper premium capillaries and measuring thermophoresis. Measurements were performed at 25 °C. Data analysis was performed in the program PALMIST using the T-jump mode and fitting to a 1-site binding model (31). MST experiments for the highest affinity binding compounds were repeated on independent days and with independent batches of recombinant protein showing good reproducibility of binding.

In Situ Ubiquitylation Assay.

EGFR ubiquitylation was monitored using a modification of a protocol by Visser Smit et al. (32). Serum starved cells were pretreated with the indicated concentrations of compound 3-120 for 30 min followed by stimulation with 50 ng/mL EGF in Keratinocyte Serum Free Media (K-SFM) with no additions for the indicated times, and harvested in 4 °C EGFR-UB lysis buffer (0.5% Triton x-100/50 mM Tris pH 7.5/150 mM NaCl/1 mM EDTA/1 mM sodium orthovanadate/10 mM sodium fluoride) supplemented with 2 mM PMSF (Calbiochem, Billerica, MA)/16 µM G5 Ubiquitin isopeptidase inhibitor I (Santa Cruz Biotechnology). Cell lysates were prepared and immunoprecipitated with 1 µg EGFR antibody (mouse monoclonal, clone 528), (Ab-1, EMD Millipore, Burlington, MA), incubated at 4 °C overnight followed by another 2-h incubation at 4 °C with protein A/G Agarose (Santa Cruz Biotechnology). Immunoprecipitates were washed thrice in chilled EGFR-UB lysis buffer. Proteins were eluted with 6XSDS sample buffer and separated by 7.5% SDS-PAGE and immunoblotted for EGFR (Cell Signaling, #4267), Ub (Santa Cruz Biotechnology, sc-8017), or phosphorylated EGFR (pY1068 or pY1045, Cell Signaling Technology, #2234 or #2237, respectively) as indicated. Immunoblots were quantified using NIH ImageJ software, taking care to make sure the exposures were in the linear range.

EGF Time Courses.

Prior to experimentation, cells were serum starved with two PBS washes and 2 h in K-SFM. Cells were treated with either vehicle (0.05% DMSO in K-SFM) or compound 3-120 at the indicated concentrations for 30 min. After that time, cells were treated with the indicated concentrations of ligand (EGF) and harvested at 0, 15, 60, or 120 min. Cell lysates were prepared as described previously (3) and the protein concentration was assessed by BCA assay (ThermoFisher, Waltham, MA).

Cell Lysate Preparation and Immunoblotting.

Equivalent amounts of protein were resolved by 7.5% SDS-PAGE, transferred to nitrocellulose and the upper portion of the nitrocellulose (>75 kDa) was immunoblotted with antibodies against total EGFR (Cell Signaling, #4267), phosphorylated EGFR (pY1068 Cell Signaling, #2234 or pY1045 Cell Signaling, #2237), or α-tubulin (Sigma, T6199). Immunoblots were exposed in the linear range and quantified using NIH ImageJ (NIH, Bethesda, MD). Data are plotted as the average (±SD) fold change in immunoreactivity (n = 3).

Statistical Analysis.

Statistical analysis was performed using GraphPad Prism (La Jolla, CA). Statistical tests are indicated in the figure legends.

In Vitro Wound Healing.

In vitro wound healing was performed as previously published (33). Cells were kept in SFM or treated with vehicle alone (0.01% DMSO), vehicle with 3 ng/mL of EGF, 5 µM compound 3-120, or 5 µM compound 3-120 with 3 ng/mL EGF and imaged using a BZ-X800 Keyence All-in-One fluorescent microscope at 4× objective. “Wounds” were visualized for 24 h by taking brightfield images every 15 min. All wounds were analyzed by finding the area (μm²) every 2 to 4 h, graphing, and finding the area under the curve (AUC). The time it took to reach 50% confluency was found by analysis of images. GraphPad/Prism was used for statistical analysis and generating graphs.

In Vivo Wound Healing.

Mechanical debridement of corneas is a well-established model for monitoring re-epithelialization of the cornea (17, 34, 35) (University of Louisville IACUC approved protocol # 23335). The epithelial layer from 1.5 mm circular wound is removed using an Algerbrush II (Precision Vision, Woodstock, IL) from the right eye of an anesthetized 8-wk-old, female, C57Bl/6 mouse (Jackson Laboratory). The size of the initial wound was measured by fluorescein staining and imaged with a fluorescent microscope (Nikon SMZ1000 stereomicroscope). After imaging (t = 0), the mouse was treated at the time of wounding as indicated [PBS, vehicle (0.01 DMSO) + 10 ng/mL EGF, or 10 µM 3-120+ 10 ng/mL EGF]. After 16 or 24 h, the mouse was anesthetized and wounded eyes were fluorescein stained and reimaged.

Filamentous Actin Staining.

hTCEpi cells were grown on coverslips until ~80 to 90% confluent. Cells were serum starved for 2 h, followed by treatment with either vehicle (0.05% DMSO) or compound 3-120 (10 µM) in K-SFM for 30 min. After that time, cells were treated with EGF (10 ng/mL) for 0, 120, or 240 min and fixed in 4% p-formaldehyde at room temperature for 30 min. After 3 × 5-min washes in PBS/250 mM MgCl₂/250 mM CaCl₂, cells were permeabilized in 0.1% triton-X-100/ PBS/1% fetal bovine serum/250 mM MgCl₂/250 mM CaCl₂, for 5 min. Cells were washed as described above and incubated for 60 min in 0.1% triton-X-100/ PBS/1% fetal bovine serum/250 mM MgCl₂/250 mM CaCl₂, containing a 1:100 dilution of Fluorescent Dye 488-I Phalloidin (Abnova, Taipei City, Taiwan). Cells were rewashed and mounted on a microscope slide using Prolong with DAPI (ThermoFisher Scientific). Images were collected on a Nikon Ti-E inverted fluorescent microscope using 60X objective.

Analytical Ultracentrifugation.

Sedimentation velocity experiments were performed in a Beckman Coulter ProteomeLab XL-A analytical ultracentrifuge (Beckman Coulter Inc., USA) at 20 °C and 40,000 rpm, using standard two-sector cells (36). Samples contained recombinant c-Cbl, unlabeled pEGFR peptide (DSFLQRpYSSDPTG), fluorescently labeled pEGFR peptide (EGFR-FAM), and compound 3-120, as indicated. Sedimentation was monitored at A₅₀₈ or A₂₉₆ every 5 min. Buffer density (1.00422 g/cm³) was determined with a Mettler/Paar Calculating Density Meter DMA 55A (Anton Paar, Graz, Austria), and viscosity (1.0619 mPa·s) was measured using an Anton Paar Lovis 2000 M viscometer (Anton Paar), with both measurements performed at 20 °C. Data were analyzed with SEDFIT using the continuous c(s) distribution model. The partial specific volume of c-Cbl (0.7366) was calculated with SEDNTERP 3 software. Experimental sedimentation coefficients were corrected to s_20,w using the measured density and viscosity.

Results

c-Cbl and Cbl-b Are the Principal EGFR E3 Ubiquitin Ligases in Corneal Epithelial Cells.

Previous work in immortalized corneal epithelial cells demonstrated that ubiquitylation of the EGFR is a negative regulator of receptor signaling (3, 17). Knockdown or knockout of c-Cbl resulted in an incomplete inhibition of ligand-mediated EGFR ubiquitylation indicating a role for additional E3 ligases. The structurally related E3 ligase Cbl-b emerged as a candidate. It is also expressed in the corneal epithelium and ubiquitylates the EGFR (37–39).

Immortalized human corneal epithelial (hTCEpi) cells were bioengineered to knockout c-Cbl and Cbl-b [referred to as “Double knockout” or DKO cells (33)] using CRISPR/CAS9 (Fig. 1A). The double knockout cells were treated with EGF for 0 to 10 min, cell lysates were prepared, and the EGFR was immunoprecipitated. Immunoprecipitates were resolved by SDS-PAGE and immunoblotted for the presence of ubiquitin, phosphorylated EGFR, and total EGFR (Fig. 1B). In the absence of both E3 ligases, there was a substantial decrease in EGFR ubiquitylation. Quantification of multiple experiments reveals that there is a ~85% decrease in the EGFR ubiquitylation in DKO cells (Fig. 1C).

Fig. 1.

Knockout of c-Cbl and Cbl-b in corneal epithelial cells increases EGFR signaling. (A) Lysates (30 μg) were prepared from hTCEpi Cas9 (CAS9), hTCEpi c-Cbl knockout (c-Cbl KO), hTCEpi Cbl-b knockout (Cbl-b KO), and hTCEpi -c-Cbl/-Cbl-b knockout (DKO) cells, resolved by SDS-PAGE, and immunoblotted using antibodies against c-Cbl, Cbl-b, and α-tubulin as a loading control. Shown are representative blots from an experiment repeated three times. (B) Serum-starved CAS9 and DKO cells were treated with 10 ng/mL EGF for 0, 2, or 10 min and cell lysates were prepared as described in Materials and Methods. EGFRs were immunoprecipitated using an anti-EGFR antibody (mAB-1, Merck Millipore). Immunoprecipitates were divided into thirds, resolved by SDS-PAGE, and immunoblotted using antibodies against ubiquitin, phosphorylated EGFR (pY1068), and total EGFR. Shown is a representative blot repeated three times. (C) Immunoblots from (B) were quantified using NIH ImageJ and the relative Ubiquitylation (Ubiquitin/phosphorylated EGFR) was plotted as the average ± S.D. (n = 3). (D) Serum-starved CAS9 and DKO cells were treated with 10 ng/mL EGF for 0 to 2 h. Cell lysates were prepared, resolved by SDS-PAGE and immunoblotted using antibodies against phosphorylated EGFR (pY1068), total EGFR, and α-tubulin as a loading control. (E) Immunoblots from (D) were quantified using NIH ImageJ and the relative EGFR phosphorylation (pY1068/α-tubulin) was plotted as the average ± SD (n = 3). Data from each of the three replicate time courses in E, the AUC was calculated. Plotted are the average ± SD (n = 3). Symbols represent individual replicates. (G) The rate of ¹²⁵I-EGF internalization was measured in parental (Cas9) and DKO cells. Plotted are the average ratio of internalized/surface ¹²⁵I-EGF at each time point ± SD (n = 3). Inset is the average ± S.D. rate of internalization (K_e). Symbols represent individual replicates. (H) EGF-mediated rates of cell growth. Parental (CAS9) and DKO cells were treated with the indicated concentrations of EGF for 24 h. The number of viable cells were quantified using alamarBlue. Data are plotted as the average fold increase in cell number ± SD (n = 3). Data were analyzed by a paired t test. *P < 0.05; **P < 0.01; ***P < 0.001.

Loss of EGFR Ubiquitylation Results in Enhanced and Sustained EGFR Phosphorylation.

To assess the consequence of decreased EGFR ubiquitylation, DKO cells were treated with EGF for 0 to 2 h. At various points of this time course, cells were assayed by immunoblot for their EGFR phosphorylation (Fig. 1D). In the absence of c-Cbl and Cbl-b (DKO cells), there was an increase in EGFR phosphorylation after 15 min of treatment that was sustained for 2 h as compared to the parental (CAS9) cells (Fig. 1E). An analysis of total phosphorylation as measured by the AUC indicates there is more than a 2.5-fold increase in receptor phosphorylation in DKO cells (Fig. 1F). Further, DKO cells exhibited slowed ¹²⁵I-EGF endocytosis (Fig. 1G) and enhanced cell proliferation in response to EGF (Fig. 1H) consistent with previous reports of CBL proteins negatively regulating EGFR function (3, 40)

Scheme for Identifying c-Cbl Antagonist.

Based on our finding that knockout of c-Cbl and Cbl-b enhanced EGFR signaling, we hypothesized that antagonizing c-Cbl and Cbl-b would be a feasible strategy to enhance EGFR signaling. Previous work by Ng et al. used X-ray crystallography to identify the site of interaction between c-Cbl and the EGFR (41). Cbl-b has a closely related crystal structure to c-Cbl (42). There is 100% identity between the amino acids in c-Cbl (Genebank: NP_005179.1) and Cbl-b (Genebank: AAH32851.1) that bind the phospho-EGFR peptide (41). Thus, we predicted that compounds that displace phospho-EGFR peptide from c-Cbl would also be able to displace phospho-EGFR peptide binding from Cbl-b. Therefore, using the (2.5 Å) crystal structure of this interaction, we performed an in silico screen of 24,877,119 million compounds from the ZINC15 library (43) (Fig. 2A). Compounds were ranked based on the lowest theoretical ΔG_bind when docked into the EGFR-binding site on c-Cbl. The top 71 compounds from that list were obtained from commercial sources.

Fig. 2.

Schematic for drug discovery. (A) Schema for identifying new compounds that disrupt EGFR:c-Cbl binding. (B–D) Differential scanning fluorimetry was used to measure c-Cbl (B and D) and Cbl-b (C) protein stability. Melting temperatures (Tm) are indicated for each sample. Recombinant protein and Sypro Orange were combined with increasing concentrations of (B and C) phosphoEGFR peptide (DSFLQRpYSSDPTG) or (D) 50 mM of candidate compound (AA and AF). (E) Serum-starved hTCEpi cells were pretreated with serum-free media containing either 0.1% DMSO,100 µM AF, or 50 µM AF for 30 min. Cells were treated without or with 50 ng/mL EGF for 10 min. Cell lysates were prepared and immunoprecipitated with an anti-EGFR antibody. Immunoprecipitates were divided into thirds, resolved by 7.5% SDS-PAGE, and immunoblotted for ubiquitin (Ub), tyrosine phosphorylation (PY), or EGFR. (F) Serum-starved hTCEpi cells were pretreated with serum-free media containing either 0.1% DMSO (vehicle) or 200 µM AF for 30 min, followed by treatment with 50 ng/mL EGF for the indicated amounts of time. Cell lysates were prepared and immunoblotted for phosphorylated EGFR (pY1068) and α-tubulin.

Top Candidate Compounds Derived from In Silico Screen.

Compound binding to c-Cbl was assessed using differential scanning fluorimetry (DSF), a high-throughput assay that measures protein unfolding as a function of increasing temperatures (Fig. 2 B–D). Molecules that stably bound to the recombinant c-Cbl (Fig. 2B) and Cbl-b (Fig. 2C) stabilize the protein in its folded state at higher temperatures. A phosphoEGFR peptide that binds to c-Cbl (DSFLQRpYSSDPTG corresponding to EGFR amino acids 1063–1078, published in ref. 41), was used as a control. There is a dose-dependent stabilization in c-Cbl (Fig. 2B) and Cbl-b (Fig. 2C) in the presence of increasing concentrations of phosphoEGFR peptide.

When the top-ranked compounds were tested, most compounds had no effect on the melting temperature of c-Cbl (Fig. 2D —compare c-Cbl alone and c-Cbl with representative compound AA). However, six compounds were identified that enhance c-Cbl stability (SI Appendix, Fig. S1), one of which was compound AF (Fig. 2D—c-Cbl vs. compound AF).

These six compounds were subjected to functional analysis. Specifically, they were tested for their ability to inhibit ligand-dependent EGFR ubiquitylation (Fig. 2E and SI Appendix, Fig. S1). Subsequently, compound AF was tested by immunoblot for the ability to sustain phosphorylation of EGFR (pY1068) over the course of 4 h (Fig. 2F). Compound AF was able to mimic the effects of c-Cbl/Cbl-b knockout cells in blocking ligand-mediated receptor ubiquitylation and sustaining EGFR phosphorylation. Compound AF was used as the backbone for further structural optimization.

Identification of Higher Affinity c-Cbl Antagonist.

The structure–activity relationship for compound AF binding to c-Cbl was optimized by systematically modifying the structure of the parent AF compound (Schematic in Fig. 3A). One hundred and twenty (120) derivative compounds were synthesized that were subject to two levels of compound screening. The first assay examined whether pretreatment with compound (200 µM) could sustain ligand-stimulated EGFR phosphorylation. The affinities of the 45 compounds that sustained EGFR phosphorylation were subsequently measured using Microscale Thermophoresis (MST) titrations (Fig. 3). MST is a medium-throughput, immobilization-free assay that measures the diffusion of fluorescently labeled proteins in a thermal gradient as a function of ligand concentration. The thermogram data are then transformed and plotted to be fit with a one-site binding model that yields the dissociation constant (K_d). As a control, we used MST to measure the K_d of the phosphorylated EGFR peptide (pEGFR) binding to c-Cbl. We note that the measured K_ds for pEGFR binding to c-Cbl and Cbl-b were 10- and 22-fold lower affinity than what was published using Isothermal Titration Calorimetry (ITC) (41). We speculate this may be due to the covalent labeling that is required for MST measurements or the addition of 8% DMSO to make comparisons with the compounds (44). Selected compounds were assessed for binding to Cbl-b (Fig. 3C). Interestingly, all compounds had a higher affinity for c-Cbl than Cbl-b.

Fig. 3.

Analysis of AF derivatives. (A) Schema for making and analyzing derivatives of AF. (B) Derivatives of the AF compound were assessed for binding to c-Cbl using microscale thermophoresis (MST) as described in Materials and Methods. Shown is a graph of those K_ds (in µM) on the y-axis and the compounds (by internal reference number) on the x-axis. (C) K_d’s of selected compounds for binding to c-Cbl vs. Cbl-b.

Compound 3-120 Decreases Ligand-Mediated EGFR Ubiquitylation.

We performed a more detailed characterization on one of these compounds, 3-120 (Fig. 4A; see SI Appendix, Supporting Text for experimental and details of the chemical synthesis). We first confirmed that 3-120 could displace EGFR binding to c-Cbl using analytical ultracentrifugation (SI Appendix, Fig. S2). Relative to the control (c-Cbl with fluorescently labeled pEGFR), the compound 3-120 was able to reduce bound fluorescently labeled pEGFR, indicative of binding site competition. Next, we wanted to determine if 3-120 could antagonize the E3 activity of c-Cbl (Fig. 4). We measured ligand-dependent EGFR ubiquitylation in hTCEpi cells in the presence of increasing concentrations of 3-120. Serum starved cells were preincubated with 0-100 µM for 30 min to allow the compound to passively diffuse into the cell. Next, cells were incubated without (−) or with (+) EGF for 10 min. Cell lysates were prepared, and the EGFR was immunoprecipitated. Immunoprecipitates were resolved by SDS-PAGE and immunoblotted for Ubiquitin (Ub), tyrosine phosphorylated EGFR (pY1045), and total EGFR. Treatment with EGF caused robust increase in EGFR phosphorylation and ubiquitylation. Pretreatment with 3−120 for 30 min before the addition of EGF, we observe a 40% reduction in EGFR ubiquitylation with no change in EGFR phosphorylation (Fig. 4 B and C).

Fig. 4.

Compound 3−120 sustains EGFR signaling in corneal epithelial cells. (A) Structure of compound 3−120. (B) Serum-starved hTCEpi cells were treated with the indicated concentrations of compound 3−120 for 30 min followed by treatment with or without EGF (50 ng/mL) as indicated. Cell lysates were prepared and immunoprecipitated for the EGFR using an anti-EGFR antibody (mAB-1, Merck Millipore). Immunoprecipitates divided into thirds, resolved by SDS-PAGE, and immunoblotted using antibodies against ubiquitin, phosphorylated EGFR (pY1045), and total EGFR. Shown is a representative blot repeated five times. (C) Immunoblots from (B) were quantified using NIH ImageJ and the relative Ubiquitylation (Ubiquitin/phosphorylated EGFR) was plotted as the average ± SD (n = 5). Symbols represent individual replicates. An ANOVA was used to compare each 3−120 concentration. Statistical significance was only observed at 100 µM 3−120. (D) Serum-starved hTCEpi and (E) primary human corneal epithelial cells were treated with compound 3−120 (50 µM) for 30 min followed by treatment with EGF (10 ng/mL) for the times indicated. Cell lysates were prepared, and equal amounts (30 μg) were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblotted using antibodies against phosphorylated EGFR (pY1068 and pY1045), and α-tubulin as a loading control. (F−H) Immunoblots from (D) and (E) were quantified using NIH ImageJ and the relative EGFR phosphorylation was plotted as the average ± SD (n = 3). (I−K) From each of the three replicate time courses in E – (G), the AUC was calculated. Shown are the average ± SD (n = 3) for each phosphorylation site. Symbols represent individual replicates. Data were analyzed by a paired t test. *P < 0.05; **P < 0.01. (L) hTCEpi cells were treated with vehicle (0.05% DMSO) or 10 µM 3-120 for 30 min followed by EGF treatment for 0, 2, or 4 h. Fixed cells were stained with Phalloidin (green), nucleic acids were stained with DAPI (blue). Shown are representative images collected with a 60X objective. (Scale bar, 10 µm.)

Treatment with 3-120 Enhances EGFR Signaling.

We wanted to assess whether these decreases in receptor ubiquitylation were sufficient to enhance EGFR signaling in magnitude and duration. hTCEpi cells (Fig. 4D) and primary human corneal epithelial cells (Fig. 4E) were pretreated with 50 µM 3-120 for 30 min, followed by EGF treatment (10 ng/mL) for 0 to 120 min. Cell lysates were prepared and immunoblotted for phosphorylated EGFR at two different tyrosines (pY1068 and pY1045) to measure receptor activity. Densitometric analysis of multiple immunoblots indicated both immortalized and primary corneal epithelial cells exhibited an increase in EGFR phosphorylation at all time points with 3-120 treatment (Fig. 4 F–H). Both cell lines had a greater than 30% increase in the AUC (Fig. 4 I–K).

During the re-epithelialization process, cells undergo a temporary epithelial to mesenchymal transition (EMT) which makes the cells more migratory (45, 46). To monitor this, we stained cells with Phalloidin to monitor filamentous actin assembly, a marker of cell migration (Fig. 4L). Treatment with compound 3-120 alone (no growth factor) resulted in an increase in directional actin assembly as compared to vehicle control. With additional treatment of EGF for 2 h, there was increased actin alignment in both control and 3-120 treated cells. After 4 h, the actin alignment was sustained in 3-120 treated cells as compared to vehicle control cells. These data are consistent with the enhanced EGFR signaling sustaining the transient EMT.

3-120 Enhances Re-epithelialization.

To examine the consequence of 3-120 on cell function, we performed an in vitro wound healing assay using hTCEpi cells (Fig. 5 A and B). Cell movement into a 2 mm acellular area was monitored with time-lapse imaging with or without EGF (3 ng/mL) with vehicle (0.01% DMSO) or 3 µM 3-120. Images were collected every 15 min for 24 h (Fig. 5A). The percent wound closure was quantified every 4 h and plotted as a function of time (Fig. 5B). From these data, we observed a significant decrease in the time it took for the wound to close (Fig. 5 B, Inset).

Fig. 5.

Compound 3-120 enhances EGF-mediated wound healing. (A) In vitro wound healing was monitored using hTCEpi cells that were plated in a tissue culture dish with 2 mm silicone plugs attached to the bottom of the dish. When cells reached confluency, the cells were serum starved, pretreated with vehicle (0.05% DMSO) or 3 µM compound 3-120 for 30 min, followed by addition of 3 ng/mL of EGF (final) to the dish. The size of the 2 mm acellular area was imaged every 15 min for 24 h. (B) wound closure was calculated as the percentage of the original wound that remained at each time point. Data were plotted as the percentage healed for each time point using Prism Software. Shown are the average ± SD from 3 to 4 experiments. Inset shows calculated time for 50% and complete healing for each condition. (C) A 1.5 mm debridement wound was made in the central cornea of an 8-wk-old female mouse. After staining with fluorescein and imaging, mouse eyes were treated with either PBS, 10 ng/mL EGF with vehicle (0.01% DMSO), or 10 ng/mL EGF with 10 µM compound 3-120. After 16 or 24 h, the cornea was restained with fluorescein and imaged. Shown are representative images of an experiment repeated 5 to 6 times. (D) The remaining debrided area was measured after 16 h using ImageJ software. Plotted are the average percentage of wound healed ± SD. Individual data points are indicated in the bar graph.

Finally, we wanted to determine if 3-120 could accelerate corneal re-epithelialization in a murine model. Briefly, a 1.5 mm circular debridement wound was made in a female, wild type C57Bl/6 mouse (Fig. 5C). After the initial wound was stained with fluorescein and visualized with a fluorescent microscope (T = 0 h), the mice were given a single treatment of 15 µL PBS alone, 10 ng/mL EGF + vehicle (0.01% DMSO), or 10 ng/mL EGF + 10 µM 3-120. Quantification of the wounds shows 3-120 significantly enhances the EGF mediated increase at 16 h (T = 16 h) (Fig. 5D). After 24 h, re-epithelialization was enhanced with EGF and vehicle and EGF and 3-120 as compared to PBS, but the addition of 3-120 did not accelerate healing, likely due to almost complete closure of the wound.

Discussion

A frequently used strategy for enhancing receptor-mediated biology has been to add more ligand to increase the magnitude of response. However, this can often result in ligand concentrations that can saturate and desensitize the receptor, causing a net decrease in signaling. This is observed when exogenous EGF is added to accelerate the healing of damaged corneal epithelium. High levels of endogenous EGF coupled with the exogenous growth factor leads to EGFR desensitization. We propose an alternative approach of inhibiting EGFR desensitization to sustain the duration of receptor signaling.

In this study, we identify that c-Cbl and Cbl-b are the primary E3 ligases that regulate the ubiquitylation of EGFR in the corneal epithelium. Knockout of these proteins results in an 85% decrease in ligand-mediated ubiquitylation and a twofold increase in receptor phosphorylation. Having identified a viable pharmacological target, we used in silico assays to identify lead compounds that could bind and antagonize the E3 ligases. Additional structural modification led to the generation of a moderately high affinity compound 3-120. The addition of the compound 3-120 with EGF decreases ligand-mediated EGFR ubiquitylation and sustains EGFR phosphorylation in both primary and immortalized corneal epithelial cells. At the cellular level, we observe sustained Phalloidin staining with 3-120 and EGF treatment, consistent with the cells adopting a more migratory phenotype. Although the efficacy of compound 3-120 is less than 40% of what is observed with knockout of the target proteins, it can accelerate in vitro wound healing as well as in vivo corneal re-epithelialization. Endogenous EGF in tear fluid may provide sufficient EGFR stimulation and negate the need for additional EGF when used in vivo. In this pharmacodynamic study, we have not explored whether multiple doses of 3-120 would further enhance re-epithelialization. Other structurally related AF derivatives also enhance EGFR phosphorylation and accelerate corneal re-epithelialization.

Compound 3-120 has only been shown to treat the most proximal events in corneal wound restoration. It is likely that accelerating these early stages of healing will prevent development of secondary pathologies [i.e., recurrent corneal erosions, epithelial thinning, stromal edema, dry eye disease, etc. (20, 21)]. During the re-epithelialization process, epithelial cells undergo a temporary epithelial to mesenchymal transition (EMT) (45, 46). This provides migratory properties to the cell so they can cover the injured area. If this process is not properly regulated, too much secretion of growth factors, such as TGFβ, can lead to fibrosis and neovascularization in the stroma. This highlights the importance of balancing epithelial layer regeneration with stromal clarity. As this compound is in early development, we do not know if antagonizing c-Cbl/Cbl-b will have a net positive effect on all aspects of cornea health.

EGFR-mediated cell migration, proliferation, and differentiation found in the corneal epithelium is not unique to this tissue. There are several other pathological conditions whose resolution can be accelerated with enhanced EGFR signaling. Most analogous is dermal wound healing and resolution of dermal scar tissue (47, 48). In addition, hair loss can, partially, be reversed with enhanced EGFR signaling (49). These are additional potential uses for compound 3-120.

Compound 3-120 validates our model for enhancing EGFR signaling by preventing receptor desensitization. There have been previous reports of other compounds that target the CBL family of proteins, namely Cbl-b with the goal of enhancing T cell receptor signaling (50, 51). Notably, these compounds target the TKBD-LHR-RING domain of Cbl-b and are designed to inhibit all E3 ubiquitin ligase activity, whereas our compound specifically disrupts EGFR:CBL interactions (Fig. 6). It was not indicated if these proprietary compounds antagonize c-Cbl as well (50, 51). We believe the design of our compound will minimize off-target effects.

Fig. 6.

Modeling of 3-120 binding to c-Cbl and Cbl-b. (A) cCbl, Cbl-b, and Cbl-3 overlaid and aligned with each of the bound EGFR peptides shown in purple (derived from PDB ID; 3OB2, 8GCY, 3OP0, respectively). (B) cCbl with the top ranking docked conformation of 3-120 shown in purple and space-filling representation. (C) Comparison of the docked conformation of 3-120 (purple) and the binding site of compound C7683 (magenta) with Cbl-b.

Despite 100% homology between the predicted binding site on c-Cbl and Cbl-b that interface with the phosphoEGFR, both 3-120 and the phosphoEGFR demonstrated selectivity for c-Cbl. At this point, it is not clear if the ubiquitylation events mediated by c-Cbl and Cbl-b are unique or functionally redundant. In oral cancer cells, inhibition of Cbl-b, but not c-Cbl, decreased EGF-mediated chemotaxis (40). These data suggest c-Cbl and Cbl-b may have different regulatory roles. Additional structural optimization of 3-120 may not only improve efficacy but may also produce a more targeted effect.

The EGFR is not the only substrate for c-Cbl and Cbl-b. These E3 ligases ubiquitylate at least a dozen other receptor tyrosine kinases (5), such as c-Met (33), PDGF (52), and IGF-1R (53). Many of these receptors are expressed in corneal epithelial cells and promote the re-epithelialization of debrided corneas. The crystal structure of some of these receptors has been solved [e.g. c-Met, (41)] and predictions can be made regarding whether 3-120 inhibits receptor ubiquitylation. However, independent of these models, cell biology and biochemical studies are required to determine if compound 3-120 can inhibit the ubiquitylation and desensitization of those receptors, how their signaling is affected, and the impact on corneal re-epithelialization.

Given the role of inactivating c-Cbl mutations in myeloid leukemia (54, 55), there is a theoretical concern that the use of c-Cbl antagonists clinically may promote aberrant cellular responses such as tissue hyperplasia elsewhere in the body. However, the likelihood of that is low when our antagonist is administered topically. Topical administration of the antagonist to the eye would require a maximal volume of 20 µL. If administered at a concentration of 10 µM (~10X the estimated K_d) and 100% of that dose were to enter the systemic circulation, the theoretical plasma concentration (based on 45L) would be ~4.5 pM, well below the drug’s K_d. Additional pharmacokinetic testing would be required to determine if compound 3-120 exhibited tissue specific accumulation or adverse effects to other surrounding ocular tissues. Thus, topically administered 3-120 seems like the most appropriate application initially.

Compound 3-120 may be a useful tool for better understanding the role of ubiquitylation in regulating receptor tyrosine kinase trafficking. For the EGFR, the role of ubiquitylation has been controversial. There is clear evidence that receptors that cannot be ubiquitylated due to mutation of the ubiquitin acceptor sites are not targeted for lysosomal degradation (6). This is due to the receptor’s inability to interact with the ubiquitin binding domains (UBD) on the endosomal sorting complexes required for transport (ESCRT) proteins that guide the receptor into the intraluminal vesicles of the late endosome. However, it is less clear whether receptor ubiquitylation is a prerequisite for internalization. The kinetics of ligand-mediated internalization do not change in ubiquitin defective mutant receptors.

Compounds that inhibit ubiquitylation provide an alternative strategy for blocking receptor endocytosis and assessing the consequence on trafficking. Unlike strategies that knockout the E3 ligase or overexpression strategies that introduce abnormally high levels of a receptor into a foreign cell environment, pharmacologic approaches provide instantaneous inhibition that limit the development of compensatory mechanisms.

Conclusion

Small molecule inhibitors that prevent receptor downregulation have been overlooked as a therapeutic strategy to amplify a receptor response. Combining in silico, biochemical, and cell biological screens, we were able to identify a lead compound that bound c-Cbl, inhibits ligand-mediated EGFR ubiquitylation, and sustains EGFR phosphorylation. Further structure/function analysis was performed to enhance its affinity. After synthesizing and testing 120 compounds, we have identified a c-Cbl antagonist that: 1) binds c-Cbl with 70 nM affinity, 2) decreases EGFR ubiquitylation and 3) sustains the EGFR in corneal epithelial cells. As a specific disruptor of EGFR:CBL interactions, it holds promise as a therapeutic to promote corneal re-epithelialization and homeostasis.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (PDF)

Dataset S02 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information. Information regarding synthesis of AF derivatives is available at Open Science Framework (https://osf.io/yspr5/overview?view_only=07141c5df2ba4081a5e6a7b047fb9ad2) (29).