Salivary gland tissue chip screening identifies candidate radioprotective drugs

Abstract

Background

Most head and neck cancer patients treated with ionizing radiation loose salivary gland function. Patients with decreased saliva have trouble eating, speaking and are predisposed to oral infections and tooth decay. Amifostine is the only FDA approved drug to prevent radiation-induced hyposalivation. However, it has intolerable side-effects that limit its use, motivating the discovery of alternative therapeutics.

Methods

We leveraged our salivary gland tissue chip platform for high-content drug discovery that we developed using submandibular gland tissue from female SKH1 hairless mice, backcrossed 6 generations with C57BL/6 J mice. We developed in-chip assays to quantify reduced glutathione and cellular senescence, which are accepted biomarkers of radiation damage. We validated radioprotection using WR-1065, the active form of Amifostine and tested other reported radioprotective drugs including Edaravone, Tempol, N-acetylcysteine, Rapamycin, Ex-Rad, and Palifermin. Next, a Selleck Chemicals library of FDA-approved drugs was screened for radioprotection. Lead hits were tested in mouse models.

Results

We identify 25 candidate compounds and down-select them using EC50 values and published pharmacologic data. This lead us to test Phenylbutazone (an anti-inflammatory), Enoxacin (a fluoroquinolone antibiotic), and Doripenem (a carbapenem antibiotic) for in vivo radioprotection in mice. Results confirm that these three drugs exhibit radioprotection equivalent to Amifostine but with superior EC50 values, ranging from 140 to 6900-fold lower values.

Conclusions

This body of work demonstrates the development and validation of assays using a tissue chip platform for high-content drug screening and the successful discovery and in vivo validation of candidate radioprotective drugs with non-antioxidant primary modes of action. These results point to possible unknown mechanisms of radioprotection. These drugs can be developed to improve radioprotection efficacy and clinical administration without adverse side-effects.

Plain Language Summary

Salivary glands are often damaged in patients undergoing radiation therapy to treat head and neck cancers. Lack of saliva results in a poor quality of life for these patients. Currently there are no good drugs that patients can take to protect their salivary glands from radiation damage. We used a system that models the salivary gland called a tissue chip to test a large panel of drugs to see if they can prevent damage by radiation. We identified several drug candidates that could protect mouse salivary glands from radiation. These drugs should now be further investigated as they could potentially be used to protect the salivary glands of patients undergoing radiation therapy.

Piraino, Chen, Mereness et al use a salivary gland tissue chip for high-content radioprotective drug discovery. Three drugs are identified that exhibit higher potency and equivalent radioprotection to Amifostine, the only FDA approved drug that is rarely used clinically due to severe side effects and lack of efficacy.

Introduction

During head and neck cancer treatment, ionizing radiation damage to the salivary glands often causes a permanent loss of secretory function and reduced salivary flow. Due to decreased saliva production, patients have trouble eating, speaking, and swallowing^1,2. Additionally, patients are at an increased risk of oral infections and tooth decay and suffer a reduced quality of life^2,3. Current treatment options, including sialogogues, mouthwashes, and chewing gum, only provide temporary relief, and there is no cure². Several strategies have been proposed to alleviate this damage, including cell transplantation^4–7 and gene therapy^8–11. Despite promising results, these methods have remained experimental and are targeted toward patients already experiencing xerostomia. Hence, there is an unmet need to provide current and future head and neck cancer patients with preventative therapies to protect salivary gland function.

Intensity-modulated radiation therapy (IMRT) and the radioprotective drug Amifostine are used clinically to prevent salivary gland damage. IMRT involves using 3D imaging to target the radiation beams at the tumor and away from sensitive organs such as the salivary gland¹². While this method can be beneficial in some cases, there are mixed results with patient-reported outcomes of dry mouth. Furthermore, IMRT is sometimes impossible to use due to tumor location^12,13. The antioxidant Amifostine is the only FDA-approved drug to prevent radiation-induced xerostomia. However, its use is often discontinued during fractionated radiation regimens due to severe side effects, including nausea, vomiting, and hypotension^14,15.

Additionally, Amifostine must be administered in a narrow time window (15–30 min) before each radiation dose because of its short half-life¹⁶, constraining its clinical use¹⁷. Furthermore, its efficacy is limited, reducing grade ≥ 2 acute xerostomia from 78% to 51% (p < 0.0001) and chronic xerostomia grade ≥ 2 from 57% to 34% (p = 0.002)¹⁸. These drawbacks highlight the critical need to discover new radioprotective drugs to prevent xerostomia.

To address this need, we developed a salivary gland tissue chip¹⁹. The tissue chip is comprised of an array of microbubble (MB) wells, each containing isolated salivary gland tissue mimetics (SGm) entrapped within an engineered matrix metalloproteinase degradable poly(ethylene glycol) (PEG)-based hydrogel extracellular matrix (eECM) that maintains secretory behavior for over 2 weeks¹⁹. The spherical architecture of the MB combined with the cell-degradable eECM creates a distinct niche that promotes cell viability and maintenance of the secretory acinar cells based on gene and protein expression as well as secretory function based on tissue architecture, gene expression, and secretory agonist-responsive calcium signaling¹⁹. This platform was validated for use in radioprotection studies using immunohistochemical staining of individual SGm to quantitate foci of DNA damage markers γH2AX and 53BP1 after irradiation^19,20. Analysis of control chips versus chips treated with WR-1065, the active form of Amifostine, showed a reduction in DNA damage with drug treatment¹⁹.

Array-based assays that can interrogate each SGm on the chip ( ~ 280 MBs/cm²) are necessary to enable high-content drug screening. Several assays commonly used to report radiation-induced cellular damage were tested in the tissue chip format. Assays tested included for example, reactive oxygen species (ROS) generation, mucin secretion, apoptosis, and cytotoxicity at various time points post-irradiation (Supplementary Table 1). Based on signal-to-noise ratio and reproducibility, the glutathione²¹ and cellular senescence²² assays were selected for further development for high-content screening of radioprotective drugs.

Assays are tested with 0 Gray (Gy) irradiation, 15 Gy, or 15 Gy + 4 mM WR-1065 to validate the detection of radiation damage and radioprotection of individual SGm within the MB-hydrogel tissue chip (40–50 MBs per chip). The assays are further validated with additional drugs previously reported to be radioprotective, including Tempol^23,24, N-acetylcysteine²⁵, Edaravone²⁶, Rapamycin²⁷, Ex-Rad^28,29, and Palifermin^30,31. A library of 438 FDA-approved drugs is then screened for radioprotection using both assays. From this screen, 25 double hits are identified and further investigated for therapeutic potential based on pharmacologic data from the PubChem database and experimentally determined dose-response relationships. Three of these hits, Phenylbutazone (anti-inflammatory), Enoxacin (fluoroquinolone antibiotic), and Doripenem (carbapenems antibiotic), are selected for in vivo radioprotection testing of the mouse submandibular gland using retroductal injection prior to irradiation. IHC staining for γH2AX and measurement of saliva secretion show that these three drugs provide radioprotection equivalent to Amifostine.

Methods

Materials

Detailed information on the drugs used for assay development and validation, including, Tempol, N-acetylcysteine, Edaravone, WR-1065, Rapamycin, Ex-Rad, and Palifermin is listed in Supplementary Table 2. Drug screening was completed using a 438 Selleck Chemicals library of FDA-approved drugs (Supplementary Data 1). The top 25 drug hits (Supplementary Table 3) were purchased from Selleck Chemicals for EC₅₀ dose-response studies. The drugs were prepared and stored per the manufacturer’s instructions.

Animals

Female SKH1 hairless mice, backcrossed 6 generations with C57BL/6 J mice, aged 6–12 weeks were used in this study to prepare MB tissue chips for assay development and drug discovery. Female C57BL/6 J mice aged 6–8 weeks were used for in vivo validation studies. Only female mice were used due to known sex differences in rodent salivary glands, with female glands more accurately emulating human salivary gland structure and function^32,33. Animals were maintained on a 12 h light/dark cycle and group-housed with food and water available ad libitum. All procedures were approved and conducted per the University Committee on Animal Resources at the University of Rochester Medical Center (UCAR #2010-024E, UCAR-2008-016E).

Microbubble (MB) array fabrication

Microbubble (MB) arrays were fabricated in poly(dimethyl) siloxane (PDMS) using gas expansion molding as previously described^19,34,35. PDMS (Dow Corning Sylgard 184) was mixed in a 10:1 base-to-curing agent ratio and poured over a silicon wafer template consisting of deep cylindrical pits with a 200 µm diameter, spaced 600 µm apart on a square lattice. The PDMS was cured at 100 °C for 2 h before peeling off the template, resulting in an array of spherical cavities with 200 µm opening and ~350 µm diameter. Circular chips with 0.7 cm diameter (48 well plates) or 0.5 cm diameter (96 well plates) were punched from the PDMS cast and glued into well plates using a 5:1 ratio of PDMS cured at 60 °C for 8 h. The MB arrays were primed in a desktop vacuum chamber with 70% ethanol to facilitate air removal from the MBs and replacement with fluid. Ethanol was exchanged for PBS and arrays were incubated overnight before cell seeding.

Synthesis and characterization of hydrogel precursors

4-Arm PEG-Norbornene was synthesized by functionalizing 4-arm 20 kDa PEG-NH₂ with norbornene using N, N′-dicyclohexylcarbodiimide (DCC) coupling, as previously described^4,5,36. The structure and percent functionalization ( > 90%) of lyophilized macromer was determined by ¹H-NMR: 4-arm PEG-amine norbornene (¹H NMR (CDCl₃): d  =  6.0–6.3 ppm (norbornene vinyl protons, 8H, multiplet), 3.5–3.9 ppm (PEG ether protons, 1817H, multiplet)).

The peptide crosslinker, GKKCGPQG ↓ IWGQCKKG was synthesized using standard solid-phase peptide synthesis on FMOC-Gly-Wang resin (EMD) on a Liberty 1 Microwave-Assisted Peptide Synthesizer (CEM) with UV monitoring, as previously described^4,5,37. The photoinitiator, lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP), was synthesized in a two-step process and validated using ¹H-NMR, as previously described in ref.³⁶. To test the degree of peptide incorporation into the hydrogels, 40 μl hydrogels were formed for all compositions. The concentrations of unreacted thiol in the supernatant (unincorporated) and in-gel (partially-incorporated) were quantified by Ellman’s assay³⁸, revealing that >95% of peptide crosslinkers and epitopes were incorporated into the hydrogels of all compositions tested, as previously described in ref. ³⁹.

Cell isolation

Mice were euthanized and the submandibular glands were removed and chopped with a razor blade for 5 min. Typically, the submandibular glands from two mice were combined for the preparation. The tissue was then incubated in Hank’s buffered salt solution (HBSS) containing 15 mM HEPES, 50 U/mL collagenase type II (Thermo Fisher 17101015), and 100 U/mL hyaluronidase (Sigma Aldrich H3506) at 37 °C for 30 min. Cells were centrifuged, resuspended in HBSS with 15 mM HEPES, and sequentially passed through 100 µm and 20 μm mesh filters to isolate cell clusters between 20 and 100 µm. The digestion protocol produces cell cluster sizes evenly distributed between 20 to 100 µm⁴⁰. As described below, the isolated clusters were combined with hydrogel precursor solution and seeded into the primed MB array chips. Typically, 40 microbubble array chips (0.5 cm diameter for 96 well plates), with n = 40–50 MB per chip, could be prepared from one mouse.

MB-hydrogel encapsulation of salivary gland cells

Isolated submandibular gland cell clusters (20–100 µm) were encapsulated within matrix-metalloproteinase (MMP) degradable poly(ethylene glycol) (PEG) hydrogels in the MB arrays (Fig. 1), synthesized and characterized as previously described^4,5,19. Briefly, the cells were resuspended in hydrogel precursor solution containing 2 mM norbornene-functionalized 4-arm 20 kDa PEG-amine macromers, 4 mM of the dicysteine functionalized MMP degradable peptide (GKKCGPQG ↓ IWGQCKKG), 0.05 wt% of the photoinitiator lithium phenyl-2,4,6- trimethylbenzoylphosphinate (LAP)⁴¹, and 0.1 mg/mL laminin in PBS^4,5,19,42. The cell/gel precursor solution (25 µL for 48 wells; 20 µL for 96 wells) was pipetted onto the MB chips and incubated for 30 min, pipetting every 10 min to redisperse cells that had settled onto the surface of the chip. The hydrogels were polymerized in situ using a Hand-Foot 1000 A broad spectrum UV light (UVA: 5 mW/cm²; UVB: 0.4 mW/cm²) with a UVC filter for 1.5 min and cultured with media (0.5 mL for 48 well; 150 µL for 96 well) that was changed every 2 days. Culture medium consisted of Dulbecco’s Modified Eagle Medium (DMEM): Ham’s F-12 Nutrient Mixture (1:1) supplemented with 100 U/mL Penicillin and 100 μg/mL Streptomycin, 2 mM Glutamine, 0.5x N2 supplement, 2.6 ng/mL insulin, 2 nM dexamethasone, 20 ng/mL epidermal growth factor (EGF), and 20 ng/mL basic fibroblast growth factor (bFGF).

Fig. 1. Overview of the salivary gland tissue chip.

A Isolated primary salivary gland cell clusters 20–100 µm. Scale bar is 100 µm. B Schematic representation of the hydrogel encapsulation of salivary gland cell clusters within an MB chip showing norbornene-functionalized 4-arm PEG-amine macromers plus the dicysteine MMP degradable peptide combined with cells and LAP photoinitiator. Image created using Biorender.com (license #AD25LJLLID). C MB arraywith inset image showing the cross-sectional view of a MB. Scale bara is 500 µm. Inset scale bar is 100 µm. D Over time clusters aggregate and grow within the MBs, example shows a SGm at day 7. Scale bar is 100 µm.

Glutathione assay

When cells experience oxidative stress the levels of glutathione decrease⁴³. Radioprotective drugs will hinder glutathione loss following ionizing radiation exposure. We developed a glutathione assay for in-chip measurements by adapting the Cellular Glutathione Detection Assay Kit (Cell Signaling Technology #13859). The monochlorobimane reagent was prepared by reconstitution in DMSO per manufacturer’s directions. For 96-well plates, 10 µL of prepared reagent (1:50 ratio of monochlorobimane (MCB) to Tris assay buffer, per manufacturer’s instructions) was added to wells containing 100 µL of culture media and incubated for 30 min at 37 °C, washed with PBS, and imaged using an Olympus IX70 microscope with a DAPI filter (Excitation/Emission: 358 nm/461 nm).

Senescence assay

Genotoxic stresses, including radiation-induced DNA damage, induce cellular senescence that contributes to hyopsalivation²². We developed a cellular senescence assay for in-chip measurements by adapting the Cellular Senescence Detection Kit – SPiDER-βGal (Dojindo Molecular Technologies, Inc. SG04). Balifomycin A1 and SPiDER-βGal stock solutions were prepared in DMSO per manufacturer’s directions. The assay was performed by first incubating the chips with Balifomycin A1 (1:1000 dilution in media) for 1 h at 37 °C. The solution was removed, replaced with 30 µL of media containing Balifomycin A1 (1:1000 dilution) and SPiDER-βGal (1:500 dilution), and incubated for 45 min at 37 °C. Chips were washed twice with media and imaged using a fluorescence microscope with a Texas Red filter (Excitation/Emission: 580 nm/604 nm).

Drug mechanism meta-analysis

Drug interaction data was examined using PubChem BioAssays results for the hits discovered in the Selleck Chemicals drug library screen described below^44–46. Pathway analysis was performed using the Drug Set Enrichment Analysis (DSEA) tool to identify pathways with gene expression patterns significantly impacted by treatment with the drug hits (p < 0.05)⁴⁷.

Image quantification and statistical analysis

For the glutathione and senescence assays, images were quantified in ImageJ⁴⁸. Each MB in the array was defined as a region of interest (ROI). All MBs in an array were included in the analysis. On rare occurrences, MBs would be excluded from analysis if cells were not seeded in a MB on Day 0 or if a MB in the array was malformed (absent). ROIs were created by thresholding on the fluorescence signal (localized to the SGm), and each ROI’s mean intensity was measured. Data were graphed and statistical analyses (ANOVA with Tukey’s post-hoc test) were performed using GraphPad Prism 9.

Tissue chip drug treatment and irradiation

For radioprotection experiments, SGm were cultured in MB-hydrogel chips for 4 days, then drugs were added to the chips 30 min before radiation and washed with media 30 min post-irradiation to parallel how Amifostine is used clinically⁴⁹. A dose of 15 Gy ionizing radiation was delivered using a JL Shepherd ¹³⁷Cs irradiator. Drug treatment schemes and radiation doses were established in our previous work¹⁹. The glutathione and senescence assays were performed at 4- and 5-days post-irradiation where maximum signal differential relative to control (0 Gy) was observed. For assay validation experiments and EC₅₀ studies, at least 3 chips (N = 3) were used for each drug, corresponding to > 100 MBs (n > 100); these values are listed in the figure captions for each experiment. Mean, standard deviation, and statistics were calculated based on the number of MBs (n).

The same treatment scheme was used for screening the Selleck Chemicals library, with drugs administered at 100 µM. One MB chip (N = 1) was used per drug, with ~40–50 MBs per chip (n = 40–50); statistics were calculated using the number of MBs (n) and compared to 0 Gy controls. Drugs were first screened using the glutathione assay. Drugs that exhibited statistically insignificant differences compared to the 0 Gy control (hits) with the glutathione assay were then tested with the senescence assay to discover library members that were efficacious in both assays (double hits). The glutathione assay was selected as the first screen because it had higher signal-to-noise ratio, more rapid throughput (30 min versus 2 hrs for completion of the senescent assay), lower assay kit cost, and a higher shelf-life of reagents in the kit after reconstitution.

To discriminate the effectiveness of drug candidates, for both the glutathione and senescence assays, we defined a signal range bounded by the values of the assay result for 0 Gy unirradiated and 15 Gy irradiated controls. Full radioprotection would yield a signal for 15 Gy+drug that was statistically equivalent to the 0 Gy control, or a recovery of 100%.

Immunohistochemical analysis

Submandibular glands and SGm were isolated and fixed in 4% paraformaldehyde overnight at 4 °C. Tissues were paraffin-embedded and then cut into 5 μm sections. Slides were treated with HIER buffer (10 mM sodium citrate, 0.05% Tween-20, pH 6.0) for antigen retrieval in a pressure cooker for 10 min then sections were blocked in CAS-block histochemical reagent (Thermo Fisher Scientific, 008120). Permeabilization was performed with 0.5% Triton X-100 in PBS for 5 min. Immunostaining was performed overnight (at 4 °C) with primary antibody for γH2AX at 2 μg/ml (EMD Millipore, 05-636). Alexa-Fluor 594-conjugated donkey anti-mouse IgG was diluted 1:500 to 4 μg/ml (Invitrogen, A21203) as secondary antibody and applied on sections for 1 h at room temperature. Following a PBS rinse, 10 μg/ml DAPI (Invitrogen, Carlsbad, CA) in PBS was applied to sections for 5 min. Sections were washed thrice in PBS for 5 min and the slides were mounted using Immu-Mount mounting medium (Thermo). Microscopic images were acquired using a Leica TCS SP5 confocal microscope with a 100X oil immersion objective and Argon laser. Analysis of images was performed in ImageJ.

Retroductal injection and irradiation

Retroductal injection delivery of drugs to the murine submandibular gland has been described in detail^50,51. This is an effective means to deliver drug directly to the salivary gland but it is an invasive procedure. For these initial efficacy studies, we elected to test drugs with one dose of ionizing radiation rather than conducting fractionated radiation doses to minimize tissue damage and animal stress. Briefly, 6–10 week old female C57BL/6 J mice were anesthetized by intraperitoneal injection of sterile saline solution of 100 mg/kg ketamine and 10 mg/kg xylazine. Maxillary incisors were secured over a metal beam, while an elastic band provided tension from behind the mandibular incisors. The mouth was widened using a custom steel retractor to apply pressure to the buccal mucosa, the tongue was retracted, and cotton was placed in the oral cavity. The wire inset of a 32 G intracranial catheter was cut at 45° to create a bevel. The beveled wire created a shallow puncture in the left salivary papilla. A beveled catheter section containing the wire insert for support was gently inserted into the puncture site. The catheter was removed, and 1 mg/kg atropine was administered by intraperitoneal injection. After 10 min, the needle of a Hamilton syringe, loaded with vehicle or drug solution, was inserted into the catheter, and the catheter was inserted into the orifice produced in the papilla. Drug solutions were injected by hand at 10 µl/min at a volume of 1 µl/g of body weight. Following injection, the pressure was maintained on the syringe for 1 min to ensure material retention before removal of the catheter. The cotton and retractor were removed from the oral cavity, and the elastic band and metal beam were released from the incisors. The known radioprotectant, WR-1065 (50 mg/kg, saline) was used as a control for radioprotection. Phenylbutazone, Enoxacin, and Doripenem hydrate were dosed at 50, 0.3, and 26 mg/kg, with N = 4–6 per group, which is sufficiently powered based on prior studies^50–52. Drug doses were chosen based on solubility and published in vivo studies^53–55. Saline was used as the vehicle for all compounds, except Phenylbutazone, which required corn oil due to poor aqueous solubility.

Mice injected with vehicle controls (saline or corn oil) or drug were treated with 0 Gy or 15 Gy within 15–30 min of injection. Mouse submandibular glands were exposed to a single dose of 15 Gy gamma radiation delivered by a ¹³⁷Cs radiation source, using a brain-slit collimator to target radiation to the glands as described previously^51,52,56. This single dose in mice recapitulates human sequalae of salivary gland radiation damage^19,57. Animals were allowed to recover for 48 h to measure persistent long-lived DNA damage by IHC⁵⁸, after which, the submandibular glands were harvested, fixed, sectioned, and analyzed using immunohistochemistry. In prior studies we quantified γH2AX foci per nucleus as a function of radiation dose and corresponding stimulated saliva secretion data at 12 weeks post-irradiation which is the gold standard test for radioprotection⁵¹. These data show a statistically significant inverse correlation (Supplementary Fig. 1) lending credibility to using the number of y-H2AX foci per cell as a reliable measure of salivary gland function. For the studies reported here, measurements were made with n = 4–6 mice per group (4–6 glands per group). After IHC, three separate sections were imaged per gland with 747 ± 142 nuclei counted per gland. For the 12 week studies the mice were monitored weekly to observe their activity. They were weighed to monitor health effects. All groups were given soft chow to aid in eating if experiencing hyposalivation.

12-week saliva collection

To validate radioprotection, saliva collections were performed one time per mouse as previously described⁵⁰. Briefly, mice were anesthetized with ketamine/xylazine and do not suffer pain from this terminal procedure. Tracheotomy was performed to ensure a patent airway. An IP injection of muscarinic agonist pilocarpine hydrochloride (100 mg/kg) (P6503, Sigma Aldrich, St. Louis, MO, USA) was given to stimulate saliva secretion. Mice were placed prone and inverted on an angled surface to put the head near the table surface. Total saliva was collected by fluid aspiration for 10 min using a capillary tube. Following euthanasia, SMGs were excised and weighed. Gland weights were not significantly impacted following irradiation and drug treatments, though the weights did show greater variability than non-irradiated controls (Supplementary Fig. 2). To account for this variability and possible differences in secretion due to gland size, saliva secretion measurements are reported as saliva weight/gland weight as previously reported⁵¹, assuming a saliva density of 1.0 g/mL⁵⁶. Saliva collection was performed by 2 blinded, independent researchers on up to 5 randomly selected mice at a time. Potential confounders, including the animal cage location, were  not controlled for.

Scientific rigor/statistical analysis

All experiments use power-based sample sizes. Sample sizes were determined based on a power analysis using α = 0.05, β = 0.2, and effect sizes of 20% based on our experience with murine salivary gland cell hydrogel encapsulations and in vivo irradiation studies. All experiments included appropriate positive and negative controls. All experiments were subjected to statistical analyses to identify significant differences between groups. As appropriate, ANOVA and post hoc multiple comparisons or Student t-tests were performed as appropriate, with the significance threshold set at p < 0.05. For each experimental group there were no exclusions of animals or data points.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results and Discussion

Salivary gland tissue chip

Our previously reported salivary gland tissue chip¹⁹ was leveraged for high-content drug screening to identify novel radioprotective compounds. The chip platform is comprised of primary salivary gland cell clusters of 20–100 µm (Fig. 1A) photoencapsulated within MMP-degradable poly(ethylene glycol) (PEG) hydrogels (Fig. 1B) in arrays of near-spherical microbubble (MB) cavities (Fig. 1C) formed in poly(dimethylsiloxane) (PDMS). Each chip, containing 40–50 MBs, is affixed within wells of a 96-well plate. In situ polymerization of the hydrogel was achieved using long-wave, low intensity UV light. Over time, the cell clusters aggregate and proliferate to form SGm (Fig. 1D).

High-throughput methods to assess drug radioprotection

In prior studies¹⁹, immunohistochemical (IHC) staining was used to quantify the number of γH2AX and 53BP1 puncta within nuclei, which are markers for double-stranded DNA breaks²⁰ used to measure radiation damage. IHC staining is, however, a laborious time and resource-consuming process requiring retrieval of the SGm from the MB array chip, tissue sectioning, staining, and imaging. While IHC enables sensitive analysis of radiation damage, it completely abrogates the goal of in situ high-content screening for which the tissue chip was developed. As mentioned above, several assays commonly used to assess radiation-induced cellular damage were tested in the tissue chip format at various time points post-irradiation (Supplementary Table 1). Based on signal-to-noise ratio and reproducibility, the glutathione²¹ and senescence²² assays were selected for further development. The glutathione assay measures reduced glutathione levels in the cells at various time points post-irradiation to determine the optimal time for detecting differences between 0 Gy and 15 Gy. Based on the data, the greatest signal separation was measured at 4 days post-irradiation (Supplementary Fig. 3). This time point matches previous reports of decreased in glutathione post-irradiation²¹ and was used for all experiments moving forward.

Next, WR-1065 (0-4 mM) was tested to identify effective concentration(s) that prevented changes in glutathione levels post-irradiation. A SGm tissue chip was cultured for 4 days to allow spheres to form¹⁹. The chip was then treated with WR-1065 30 min before and during radiation, followed by drug wash out with media 30 min post-irradiation (Fig. 2A). This dosing scheme is consistent with the clinical use of Amifostine ¹⁴ and with our previous work¹⁹. The glutathione assay was performed 4 days post-irradiation. Example images show high glutathione levels at 0 Gy, which are decreased by treatments with 15 Gy radiation and restored in the presence of 4 mM WR-1065 (Fig. 2B–D). Quantification shows that 0.1 mM and 0.4 mM WR-1065 were ineffective at preventing radiation damage. In contrast, 1 mM and 4 mM WR-1065 provided significant protection compared to the 0 Gy control (Fig. 2E). The 4 mM dose corresponds with our previous work using DNA damage markers γH2AX and 53BP1¹⁹ and values from literature⁵⁹ and establishes the range of effective concentrations for WR-1065 treatment in vitro. Moreover, it is clinically relevant as 15–30 min before irradiation, patients are administered Amifostine intravenously at 200 mg/m^2 15,49. Assuming an average adult body surface area of 17,000 cm² and a blood volume of 1.35 gallons (5.1 liters)^60,61, Amifostine is clinically administered at 0.3 mM which is similar to our in vitro dose studies.

Fig. 2. Radiation-induced decrease in glutathione is restored to unirradiated control levels by WR-1065 and other reported radioprotective compounds.

Timeline of drug treatment and assay for glutathione at 4 days post-radiation created using Biorender.com, license# TW28JHFK9V. A Representative images of the glutathione assay, for (B) 0 Gy, (C) 15 Gy, (D) 15 Gy + 4 mM WR1065 (WR). Scale bar is 600 µm. E Quantification of the fluorescence intensity of individual MBs normalized to 0 Gy. n = # of MBs; 0 Gy n = 610, 15 Gy n = 559, 15 Gy+0.1 mM WR n = 156, 15 Gy+0.4 mM WR n = 251, 15 Gy+1 mM WR n = 302, 15 Gy+4 mM WR n = 691. Brackets with asterisks compared to 15 Gy: ns = nonsignificant p = 0.9988 (15 Gy vs. 15 Gy+0.1 mM WR), **** = p < 0.0001; Money sign compared to 0 Gy: = p < 0.0001; WR = WR-1065. Drugs are used to treat SGm on day 4 for 30 min prior to and 30 min after radiation and senescence is analyzed on day 9 for (F) Tempol, (G) Edaravone, (H) N-acetylcysteine (NAC), (I) 50 µM Rapamycin, (J) 100 ng/mL Palifermin, and (K) 50 µM Ex-Rad. n = # of MBs; Tempol: 0 Gy n = 752, 15 Gy n = 750, 15 Gy + 1 mM n = 193, 15 Gy + 4 mM n = 354; Edaravone: 0 Gy n = 762, 15 Gy n = 750, 15 Gy + 1 mM n = 202, 15 Gy + 4 mM n = 192; NAC: 0 Gy n = 802, 15 Gy n = 750, 15 Gy + 1 mM n = 205, 15 Gy + 4 mM n = 185; Rapamycin: 0 Gy n = 762, 15 Gy n = 750, 15 Gy + 50 µM n = 256; Palifermin: 0 Gy n = 762, 15 Gy n = 750, 15 Gy + 100 ng/mL n = 274, Ex-Rad: 0 Gy n = 762, 15 Gy n = 750, 15 Gy + 50 µM n = 185. Brackets with asterisks compared to 15 Gy: ns = nonsignificant p = 0.9288 (15 Gy vs. 15 Gy+ 1 mM NAC), 0.7043 (15 Gy vs. 15 Gy+ 4 mM NAC), and 0.9991 (15 Gy vs. 15 Gy + 50 µM Ex-Rad), **** = p < 0.0001; Money sign compared to 0 Gy: = p < 0.0001.

Since WR-1065 is an antioxidant and mediates radioprotective effects through free radical scavenging and induction of superoxide dismutase expression⁶², we tested other antioxidants implicated as radioprotective (Tempol, Edaravone, N-acetylcysteine). Tempol exhibited excellent radioprotection at both 1 and 4 mM (Fig. 2F) which is consistent with prior studies^23,57,63. Edaravone showed complete radioprotection at 1 mM but not at 4 mM (Fig. 2G). Edaravone maintained ~43% of glutathione levels at 0.1 mM (Supplementary Fig. 4A), suggesting that the optimal concentration range for Edaravone might be lower than WR-1065, which is consistent with literature indicating radioprotective dose ranges are 0.1–1 mM^26,64,65. Higher Edaravone doses are likely cytotoxic. For N-acetylcysteine (NAC), no improvement in glutathione levels was observed at 1 or 4 mM compared to untreated SGm (Fig. 2H); however, glutathione levels were rescued by 56% when treated with 10 mM NAC (Supplementary Fig. 4B), also consistent with previous literature²⁵.

Drugs reported with non-antioxidant radioprotective mechanisms were also tested using the glutathione assay. Rapamycin is an mTOR inhibitor that has been reported to restore salivary flow rate post-irradiation in swine²⁷. Ex-Rad reduces p53-dependent and p53-independent apoptosis²⁸. Palifermin (keratinocyte growth factor) has been reported to stimulate salivary gland stem/progenitor cell expansion post-irradiation³⁰. Using drug concentrations based on the literature^29,31, our data shows some protection against radiation-induced glutathione level changes resulting from radiation with 50 µM rapamycin (58%, Fig. 2I) and 100 ng/mL Palifermin (47%, Fig. 2J) but no protection using 50 µM Ex-Rad (Fig. 2K).

For senescence, a protocol similar to the glutathione assay was followed to determine the optimal time point post-irradiation for detecting a change in senescence following 0 Gy and 15 Gy irradiation, by measuring senescence-associated β-galactosidase activity. A time point of 5 days post-irradiation was optimal (Supplementary Fig. 5), similar to a previous study²². WR-1065 radioprotection was tested by adding the compound drug to the chips 30 min before irradiation, followed by washing out 30 min post-irradiation (Fig. 3A). An expected increase in senescence was detected for SGm exposed to 15 Gy compared to 0 Gy, which was restored to levels equivalent to 0 Gy with the addition of WR-1065 (Fig. 3B–D). Quantification shows that both 1 mM and 4 mM WR-1065 treatment resulted in complete radioprotection (Fig. 3E), consistent with the glutathione assay results.

Fig. 3. Radiation-induced increase in senescence is restored to unirradiated control levels by WR-1065 and other reported radioprotective compounds.

Timeline of drug treatment and assay for senescence at 5 days post-radiation created using Biorender.com, license# TW28JHFK9V. A Representative images of the senescence assay for (B) 0 Gy, (C) 15 Gy, (D) 15 Gy + 4 mM WR-1065, and (E) quantification of the fluorescence intensity of individual MBs normalized to 0 Gy. n =  # of MBs; 0 Gy n = 1344, 15 Gy n = 1111, 15 Gy+1 mM WR n = 534, 15 Gy+4 mM WR n = 504. Brackets with asterisks compared to 15 Gy: **** = p < 0.0001; Money signs compared to 0 Gy: = p < 0.0001, ns = nonsignificant p = 0.9996 (0 Gy vs. 15 Gy +1 mM WR) and 0.9981 (0 Gy vs. 15 Gy +4 mM WR); WR = WR-1065. Drugs are used to treat SGm on day 4 for 30 min prior to and 30 min after radiation and glutathione is analyzed on day 8 for (F) Tempol, (G) Edaravone, (H) N-acetylcysteine (NAC), (I) 50 µM Rapamycin, (J) 100 ng/mL Palifermin, and (K) 50 µM Ex-Rad. n = # of MBs; Tempol: 0 Gy n = 1344, 15 Gy n = 1111, 15 Gy + 1 mM n = 400, 15 Gy + 4 mM n = 406; Edaravone: 0 Gy n = 1344, 15 Gy n = 1111, 15 Gy + 1 mM n = 390, 15 Gy + 4 mM n = 360; NAC: 0 Gy n = 1344, 15 Gy n = 1111, 15 Gy + 1 mM n = 346, 15 Gy + 4 mM n = 384; Rapamycin: 0 Gy n = 1344, 15 Gy n = 1111, 15 Gy + 50 µM n = 354; Palifermin: 0 Gy n = 1344, 15 Gy n = 1111, 15 Gy + 100 ng/mL n = 363, Ex-Rad: 0 Gy n = 1344, 15 Gy n = 1111, 15 Gy + 50 µM n = 471. Brackets with asterisks compared to 15 Gy: ns nonsignificant, **** = p < 0.0001; Money sign compared to 0 Gy: = p < 0.0001, $$$ = p = 0.0002 (0 Gy vs. 15 Gy + 4 mM Edaravone), $$= p = 0.0058 (0 Gy vs. 15 Gy + 4 mM Tempol), ns nonsignificant p = 0.3395 (0 Gy vs. 15 Gy + 1 mM Edaravone), p = 0.9842 (0 Gy vs. 15 Gy + 50 µM Rapamycin).

Next, the level of senescence was measured using the other reported radioprotective drugs. Tempol (Fig. 3F) and Edaravone (Fig. 3G) reduced senescence at both 1 mM and 4 mM by 75% and 91% for Tempol, 94% and 113% for Edaravone, respectively, versus untreated, irradiated controls. Edaravone reduced senescence by 96% at 0.1 mM versus untreated, irradiated controls (Supplementary Fig. 6A). For NAC, 83% and 57% reduction in senescence was observed with 1 mM and 4 mM (Fig. 3H) and 82% at 10 mM (Supplementary Fig. 6B) versus untreated, irradiated controls. For drugs with non-antioxidant mechanisms, rapamycin (50 µM) showed complete protection (Fig. 3I), whereas Palifermin (100 ng/mL) provided 64% protection (Fig. 3J) and Ex-Rad (50 µM) conferred only 45% radioprotection (Fig. 3K).

A summary of results from the glutathione and senescence assays shows similar trends for radioprotection (Supplementary Table 4). The few differences may result from different mechanisms of action and/or assay targets. Logically, the glutathione assay may be more sensitive to antioxidant function (NAC, Tempol, Edaravone), while the senescence assay may be more appropriate for drugs such as rapamycin, which has anti-senescence properties⁶⁶. These results highlight the trade-offs in developing screening assays and point to the benefit of screening with two assays. Although the assays developed are indirect measures of radiation-induced DNA damage, they nonetheless were validated to detect radiation-induced cell-damage and drug radioprotection. More importantly, these assays can be used for high-content drug screening with multiple replicates (40–50) per test and enhanced throughput compared to IHC staining for γH2AX.

Drug library screening identified several promising radioprotective compounds

The glutathione and senescence assays were used to screen a library of FDA-approved drugs (Selleck Chemicals) at 100 µM which is 10x lower than the concentration of where WR1065 that was radioprotective (1 mM). Drugs were first screened using the glutathione assay according to the timeline in Fig. 2A. Any compound resulting in statistically equivalent glutathione levels compared to the 0 Gy control was considered a hit (Supplementary Fig. 7, orange circles). Hits with the glutathione assay were then tested using the senescence assay and considered a “double hit” if senescence levels were statistically equivalent to levels at 0 Gy (Supplementary Fig. 7, blue circles), providing two independent assays to measure radioprotective efficacy. A list of the 438 drugs screened and relevant statistics are shown in Supplementary Data 1. Overall, 438 drugs from the library were tested with a hit rate of 5.7%, for a total of 25 double hits (Supplementary Fig. 7) listed in Supplementary Table 3. This hit rate is higher than many other drug screening reports (0.1–0.3%)^67,68 and may be due to the high statistical rigor afforded by the tissue chip format and/or the high drug concentration (100 µM) used for screening, which was intentional to avoid missing compounds tested on 3D tissues rather than cell monolayers. Additionally, phenotypic screens generally have higher hit rates than target-based screens and they may be a more successful strategy for discovery of first-in-class medicines^69–71. The 25 double hits span a wide range of log P values from 5 to −3.8 (Supplementary Table 5), with most drugs exhibiting hydrophobic properties (positive logP), lessening concerns that hydrophobic drug absorption to PDMS hinders ‘hit’ identification^72,73.

Identification of potential radioprotective drug mechanisms

Of the 25 potential radioprotective compounds, 20 have known interactions with proteins involved in calcium signaling identified within the BioAssay database in PubChem (Supplementary Table 6). These compounds may impact secretory signaling in the salivary gland, which is reported to be radioprotective^74,75. While degranulation may not be key to radioprotection, secretory stimulation may play a role in proliferation and survival of the secretory cells⁷⁴. Similarly, using the Drug Set Enrichment Analysis (DSEA) tool to identify pathways, the Reactome analysis related to secretion appears to be upregulated by many of the identified drugs, supporting this potential mechanism (Fig. 4, and Supplementary Fig. 8). Interestingly, only 9 of the 25 compounds have known antioxidant properties, and 12 are anti-inflammatory. These are critical data indicating that alternate mechanisms of radioprotection may be achievable and represented in the identified hits. Pathways related to cell-cell, cell-matrix interactions and neuronal survival and differentiation were identified and represent potential areas of mechanistic exploration. Multiple studies have identified changes in integrin expression, cell adhesion and matrix interactions following radiation exposure, which may be linked to cell response^1,76–79. Studies report that the preservation of nerve activity in the salivary gland is radioprotective⁸⁰. Manipulation of these mechanisms in the salivary gland may reveal molecular targets of radioprotection.

Fig. 4. Potential mechanisms of action on salivary gland and similarities among identified radioprotective drugs.

A Drug hit activity related to secretion (calcium signaling), prostaglandin production, and anti-inflammatory or antioxidant properties. B Reactome pathways were significantly associated with gene expression patterns affected by treatment with 12 of the identified potential radioprotective compounds (p < 0.05). E-scores (−0.5 to 0.5) associated with each pathway show predicted activity related to treatment with the identified drugs. C Heat map showing ranks for drug-pathway interactions from most-upregulated (orange) to most-downregulated (blue) drugs for each pathway.

Hit down-selection using drug promiscuity data and EC₅₀ values

A systematic approach was used to down-select the 25 double hits for further testing. Since drugs within the library are FDA-approved, considerable information on their pharmacology in mice and humans is readily available through resources such as PubChem. Within PubChem, the BioAssay database was created by the National Institute of Health (NIH) as an open repository containing small molecule screening data results⁴⁶. We used the BioAssay data to analyze drug promiscuity, which refers to the ability of a drug to bind multiple molecular targets with distinct pharmacological outcomes, often causing unwanted side effects⁸¹. The drugs exhibiting bioactivity in many assays were deprioritized. Data for each double hit was obtained from the database, and promiscuity was calculated as the percentage of assays reported as “active” (Supplementary Table 6). Drugs with high activity ( > 10%) were excluded from further testing. Additionally, Etidronate, Melatonin, and Albendazole were excluded due to poor bioavailability^82–84, and Eplerenone was excluded due to solubility concerns⁸⁵.

The remaining 13 drugs were tested for glutathione levels post-irradiation in dose-response experiments (1–100 µM) to determine their effective concentration more accurately. Many drugs were only effective at 100 µM and were were deprioritized due to their limited potency. For the remaining seven drugs (Phenylbutazone, Meropenem, Diethylstilbestrol, Prazosin, Enoxacin, Glipizide, and Doripenem), dose responses and heat maps summarizing the dose-response results for the glutathione and senescence assays are presented in Fig. 5. Dose-response curves for the glutathione assay show that Phenylbutazone (Fig. 5B) and Meropenem (Fig. 5C) exhibited radioprotection equivalent to 0 Gy over a concentration range of 100 nM to 100 µM and Diethylstilbestrol (Fig. 5D) was radioprotective at 10–100 µM. Prazosin (Fig. 5E), Enoxacin (Fig. 5F), Glipizide (Fig. 5G), and Doripenem (Fig. 5H) showed protection at higher concentrations ( > 10 µM).

Fig. 5. Dose-response data for lead compounds from the drug screen analyzed for glutathione and senescence post-irradiation.

A Heat map of combined dose-response results for intracellular glutathione levels, with 0 Gy (1.0) at left and 15 Gy (0.3) only on right. Dose-response results for (B) Phenylbutazone, (C) Meropenem, (D) Diethylstilbestrol, (E) Prazosin HCl, (F) Enoxacin, (G) Glipizide, and (H) Doripenem hydrate in the glutathione assay. I Heat map of combined dose-response results for cellular senescence, with 0 Gy (1.0) at left and 15 Gy (1.3) only on right. Results for (J) Phenylbutazone, (K) Meropenem, (L) Diethylstilbestrol, (M) Prazosin HCl, (N) Enoxacin, (O) Glipizide, and (P) Doripenem hydrate. Green and red lines represent 0 Gy and 15 Gy averages, respectively. Statistics were calculated using ANOVA with Dunnett’s post-hoc test. n = # of MBs. ns nonsignificant; **p < 0.01, ***p < 0.001, ****p < 0.0001. See Supplement Fig. X for the exact n and p values for each group.

The radioprotection trends based on the senescence assay differed somewhat, with Phenylbutazone (Fig. 5J), Meropenem (Fig. 5K), and Enoxacin (Fig. 5N) showing only partial protection. Diethylstilbestrol (Fig. 5L) exhibited radioprotection equivalent to 0 Gy between 50–100 µM concentrations whereas Prazosin (Fig. 5M) showed complete protection between 0.1–100 µM. Glipizide (Fig. 5O) and Doripenem (Fig. 5P) also showed variable protection.

EC₅₀ values extrapolated from dose-response curves for the top 7 double hits are shown in Table 1. Based on the glutathione assay these drugs exhibit 140-to-6900-fold lower EC₅₀ values than WR-1065, some in the nanomolar range. Phenylbutazone showed the most promising results, with low EC₅₀ values for both the glutathione (80 nM) and senescence (50 nM) assays. Phenylbutazone is a non-steroidal anti-inflammatory drug (NSAID) that inhibits cyclooxygenases (COX-1 and COX-2), enzymes that produce prostaglandins⁸⁶. Prostaglandins, specifically PGE₂, have been shown to increase in irradiated salivary glands, and mitigation of salivary gland damage was achieved with the anti-inflammatory drug Indomethacin^1,87. Indomethacin also showed radioprotection in our drug screen but was ineffective at concentrations lower than 100 µM. Indomethacin only blocks COX-1, underscoring the greater efficacy of Phenylbutazone. Phenylbutazone was originally developed for chronic pain for conditions such as arthritis but has since been restricted to treating ankylosing spondylitis due to induction of rare but severe blood disorders, including anemia and leukopenia⁸⁶. However, doses ranged from 300 to 1000 mg, generating a plasma concentration of 30–50 µg/mL⁸⁶. In contrast, the EC₅₀ of 0.08 µM for protecting against radiation-induced glutathione changes established in this study equates to a 26 ng/mL concentration. Thus, the risk of severe adverse effects may be greatly diminished for doses necessary for radioprotection. Additionally, Phenylbutazone has excellent bioavailability (up to 90%)^86,88 and long half-life (50-105 hrs)⁸⁶, which may enable dose de-escalation, further decreasing risks.

Table 1.

Calculated EC₅₀ values for the top radioprotective drugs. Values were estimated using a nonlinear fit in Prism

Compound	Estimated EC50 (µM)	Senescence
	Glutathione
WR-1065	573	undetermined
Phenylbutazone	0.08	0.05
Meropenem	0.1	undetermined
Diethylstilbestrol	0.7	37.7
Prazosin HCl	1	0.04
Enoxacin	2.4	0.5
Glipizide	4.0	undetermined
Doripenem hydrate	8.6	16.4

“Undetermined” indicates that the software could not accurately fit a curve to the data

Enoxacin is a fluoroquinolone antibacterial agent used for treating urinary tract infections⁸⁹ that was previously identified as radioprotective⁹⁰. Using a high-throughput screening method with the viability of lymphocytes as the primary readout, two classes of antibiotics (tetracyclines and fluoroquinolones) were identified as robust radioprotectors, including Enoxacin⁹⁰. This observation corroborates our drug screening results, in which several antibiotics were identified as double hits, including Enoxacin, Meropenem, Doripenem hydrate, Rifampin, and Rifabutin. The Enoxacin EC₅₀ of 2.4 µM for the glutathione assay is close to the 13 µM EC₅₀ reported for viability of lymphocyte cells⁹⁰. Notably, eight other fluoroquinolones (Levofloxacin, Gatifloxacin, Ofloxacin, Moxifloxacin, Marbofloxacin, Sparfloxacin, Perfloxacin and Norfloxacin), some of which are reported as radioprotectors⁹⁰, were not hits in our drug screen (Supplementary Data 1). Exoxacin is the only fluoroquinolone in this series containing a naphthyridine group at the 8^th position suggesting the importance of this functional group in conferring radioprotection (Supplementary Fig. 9).

In vivo validation of hits using retroductal injection

Based upon drug down selection data and measured EC₅₀ values, Phenylbutazone, Enoxacin, and Doripenem hydrate were selected for in vivo validation in mice using γH2AX foci per nucleus IHC staining as an outcome measure consistent with prior work showing correlation with the development of xerostomia (Supplementary Fig. 1)^19,50–52. Vehicle-treated SGm exposed to 15 Gy irradiation exhibited a 3.3-fold increase in the number of γH2AX foci per nucleus, indicating a significant increase in double-stranded DNA breaks due to radiation exposure (Fig. 6A–D, I). Treatment with WR-1065 resulted in a 0.5-fold reduction in γH2AX foci per nucleus compared to 15 Gy controls (Fig. 6D, E, I). No significant differences existed between 0 Gy controls and WR-1065 treated SGm exposed to 15 Gy (Fig. 6C, E, I). These results are similar to prior studies utilizing WR-1065 in vitro and in vivo via retrograde ductal injection^19,51. Treatment with the test compounds, Phenylbutazone, and Enoxacin resulted in 0.4- and 0.5-fold reduction in γH2AX foci per nucleus compared to 15 Gy controls, respectively (Fig. 6E–G, I). Results observed after Phenylbutazone and Enoxacin treatment were not significantly different from 0 Gy controls, suggesting they are protective (Fig. 6C, F, G, I). Treatment with Doripenem hydrate did not appear to reduce γH2AX foci per nucleus relative to 15 Gy controls and showed a 2.7-fold increase compared to 0 Gy controls (Fig. 6C, D, H, I). To examine whether retroductal injection of these hits preserves functional secretion, stimulated saliva secretion was evaluated 12 wks post-irradiation. Salivary funtion was significantly decreased in vehicle-treated SGm exposed to 15 Gy radiation compared to non-irradiated controls (Fig. 6J). However, treatment with WR-1065, Enoxacin, and Doripenem restored secretion to control levels (Fig. 6J). Notably, Doripenem fully protected saliva secretion despite the predicted lack of protection based on 48 hr γH2AX staining (Fig. 6H, I). This discrepancy is likely due to the different evaluation times (48 hrs for IHC vs, 12 wks for saliva). In contrast, Phenylbutazone showed radioprotective effects in γH2AX but failed to restore saliva secretion, possibly due to the long-term toxicity effects of corn oil used as a solvent (Fig. 6I, J).

Fig. 6. In vivo testing of radioprotective drugs identified via SGm screening.

A Representative image of retroductal drug administration. Injection of trypan blue dye into a single submandibular gland is shown. B Schematic of drug treatment, irradiation, and harvest protocol. Representative images of γH2AX staining in mouse salivary gland tissue 48 h post-treatment with (C) vehicle and (D) 15 Gy radiation exposure, as well treatment with (E) control drug WR-1065 and lead drug hits, (F) Phenylbutazone (PB), (G) Enoxacin (EN), and (H) Doripenem hydrate (DH) followed by 15 Gy radiation exposure. Scale bars: 20 μm. I Quantification of γH2AX foci per nucleus from all treatment groups. J Measurement of stimulated saliva secretion normalized by gland weight at 12 wk after 15 Gy radiation exposure (N = 3–6). ns nonsignificant p = 0.9997 (15 Gy vs. 15 Gy + PB), *p = 0.0134 (15 Gy vs. 15 Gy + WR-1065), **p = 0.004 (0 Gy vs. 15 Gy), **p = 0.0024 (15 Gy vs. 15 Gy + DH), **p = 0.001 (15 Gy vs. 15 Gy + EN)

Conclusions

To overcome long-standing challenges associated with off-target radiation damage resulting in life-long dry mouth and poor quality of life, we sought to leverage our salivary gland tissue chip technology to identify candidate radioprotective drugs¹⁹. Assays that detect biomarkers of radiation damage (Supplementary Table 1), glutathione, and β-galactosidase (cell senescence), were developed for high-content screening due to favorable sensitivity and reliability. We validated the assays with known radioprotective drugs, including WR-1065, the active metabolite of Amifostine, the only FDA-approved preventative therapy for radiation-induced xerostomia^14,15,59. We next tested other drugs reported to be radioprotective, including Tempol^23,24,57, Edaravone^26,64,65, N-acetylcysteine²⁵, Palifermin^30,31, Ex-Rad^28,29, and Rapamycin²⁷. Our results found that Tempol and Edaravone showed complete radioprotection similar to WR-1065 in both assays, whereas the other drugs showed partial radioprotection. Partial protection likely reflects differences in tissue types, or the assay protocols used in the other studies, but our results confirm that all the drugs tested exhibit radioprotection at some concentration, in one or both assays (Supplementary Table 4). A library of FDA-approved drugs was screened, enabling the de novo identification of 25 drugs with radioprotective activity (Supplementary Table 6). Of the 25 hits, 20 drugs have known interactions with proteins involved in calcium signaling (Fig. 4A), which is critical for saliva secretion⁹¹. Only 9 of the 25 compounds have known antioxidant properties, and 12 are anti-inflammatory. This suggests alternate mechanisms of radioprotection and is an area of future exploration. A DSEA pathways analyses suggests these drugs upregulate secretion and activate pathways related to cell-cell, cell-matrix interactions, neuronal survival and differentiation suggesting possible new mechanistic targets for designing novel radioprotective drugs. Using the PubChem database and experimentally determined EC₅₀ values (Fig. 5 and Table 1), the list of 25 double hits was narrowed down to 3 candidates (Phenylbutazone, Enoxacin, and Doripenem) for testing using an in vivo radiation damage mouse model. Results demonstrate that these three drugs have equivalent radioprotection to WR-1065 (Fig. 6) and may be progenitors of new chemical classes of radioprotectant drugs.

In conclusion, this body of work validates the SGm tissue chip as a high-content drug discovery tool. We validated glutathione and β-galactosidase (cell senescence) as biomarkers of ionizing radiation damage. Candidate drugs were discovered that confer radioprotection by mechanisms other than antioxidant activity. Ongoing studies seek to investigate chemical analogs and mechanisms of action to identify lead drug candidates that can be developed for clinical translation. Additionally, we are developing a more sophisticated innervated salivary gland tissue (iSGm) chip model to more faithfully recapitulate the native gland which is highly innervated. This will enable testing drugs for off target effects on the salivary gland and for discovery of radioprotective candidates that act on nerves. While studies herein utilize mouse tissue chips, we developed human salivary gland tissue chips that preformed equivalently¹⁹. Nonetheless studies to validate our lead candidates in human chip are planned.

Author contributions

L.R.P, C.Y.C. and J.A.M. established methods, designed and conducted experiments, analyzed data, wrote and reviewed the manuscript. P.M.D provided critical consulting expertise and reviewed the manuscript. L.A.D., C.E.O. and D.S.W.B. conceived and led the project, designed studies, interpreted data, wrote and revised the manuscript.

Peer review

Peer review information

Communications Medicine thanks Kuo-Chan Weng and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Data are available in the main text, supplementary materials, or from the corresponding author upon request. Raw source data for the figures are available on Figshare⁹².

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43856-025-01136-7.