A “failed” assay development for the discovery of rescuing small molecules from the radiation damage

Abstract

With improving survival rate for the cancer patients, the side effects of radiation therapy, especially for pediatric or more sensitive adult patients, have raised interests for preventive or rescuing treatment to overcome the detrimental effects of the efficient radiation therapies. For the discovery of rescuing small molecules for the radiation damage for the endothelium, we have been developing a 96 well microplate-based in vitro assay for the high-throughput compatible measurement of the radiation induced the cell damage and its rescue by phenotypic high-content imaging. In contrast to the traditional radiation assays with detached cells for the clonogenic formation, we observed the cells with live-cell imaging in two different kinds of endothelial cells, up to three different cell densities, two Gamma-IR radiation dose rates, more than four different radiation doses, acute (within 24 hours with 1–2 intervals) and chronical responses (up to 7 days) by phenotypic changes (Digital phase contrast) and functional assays (nuclear, live-cell, and dead cell staining) at the end of the assay. Multiple potential small molecules, which have been reported for rescuing the radiation damage, have been tested as assay controls with dose responses. At the end, we didn’t move ahead with the pilot screening. The lessons learned from this “failed” assay development are shared.

Introduction

Assay development, especially the assay development for high throughput screening, is a tough business¹. Usually the challenges, known as the “valley of death” of the drug discovery process^2,3 are attributed to the non-perfect in vitro assays for the early drug discovery. Robustness, biology/physiology relevance, and automation compatibility (micro-plate based) of the assays are to be optimized to generate hits that ultimately can be used/developed for in vivo and clinical application⁴. These three factors also greatly limit the available in-vitro models^5,6 and the windows of the signals^7,8 in the assay development, despite the facts that plethora of target-based and phenotypic assays have been developed for early drug discovery and systems biology research^9,10. Here we present a “failed” assay development for the discovery of small molecules that rescue or mitigate radiation damage in non-cancerous tissue. This example in contrast demonstrates the complications and difficulties that can arise during assay development.

Radiation therapy for cancer¹¹ has been part of standard treatment for 50% of for cancer patients, in addition to surgery, chemotherapy, and more recently immuno-therapy¹². Because of the continuingly improving survival rate of cancer patients, the long-term consequences of this primary treatment have emerged as a significant risk factor¹³, especially on the cardiovascular related side effects^14,15,16. Small molecule drugs have been explored for rescuing the radiation damage therapies for their transient, efficient, and biologically available characteristics. Recilisib sodium¹⁷, i.e. Ex-Rad, has been in Phase I trial as the only known oral radioprotectant. There are majorly five categories of reported irradiation rescuing effects: 1) Antioxidants or anti-inflammatory (e.g., Auranofin¹⁸, GC1149¹⁹, or MitoTempo²⁰); 2) 8-Oxoguanine glycosylase (OGG) activators (e.g., V028–5832²¹, or Melatonin²²); 3) GPX4 activators through ferroptosis (Y600–0815²³ or (±)-α-Tocopherol Acetate²⁴ as a Vitamin E analog); 4) Hits/leads from whole animal irradiation experiments (e.g., recilisib sodium¹⁷ also as Ex-Rad, and γ-Tocotrienol²⁵); 5) others, e.g. HDAC inhibitors (Trichostatin A²⁶). However, there is no systematic studies on the rescuing molecule, especially on specific tissue/organs, e.g., cardiac endothelium or brain microvascular endothelium. A high throughput, plate-based, in vitro assay would greatly facilitate the discovery of rescuing small molecules from radiation damage. The above-mentioned small molecule compounds will be utilized as positive controls for the assay development.

Currently most of the assays for the discovery of rescuing small molecules from the radiation damage are: 1) clonogenic formation assays of radiated detached cells; and 2) in vivo (mostly through mouse) experiment. Both are limited by the throughput to do systematic unbiased study on the radiation effects. A high throughput plate-based in vitro assay would attest not only hypothesis-based (e.g., 8-Oxoguanine glycosylase (OGG) activators) but also unbiased phenotypic cell-based (e.g., synergistic effects of two or more targets) discovery of small molecules for rescuing the radiation damage.

In this report, we describe how we developed a 96 well microplate-based in vitro assay for the high-throughput measurement of the radiation induced the cell damage and its rescue by phenotypic high-content imaging. In contrast to the traditional radiation assays with detached cells for the clonogenic formation, we observed the cells with live-cell imaging in two different kinds of endothelial cells, with up to three different cell densities, at two Gamma-IR radiation dose rates, and at more than four different radiation doses, for acute (within 24 hours with 1–2 intervals) and chronical responses (up to 7 days) by phenotypic changes (digital phase contrast, DPC) and functional assays (nuclear, live-cell, and dead cell staining) at the end of the assay. 11 reported small molecules have been tested with dose responses as controls. In the end, we didn’t move ahead with the pilot screening because we cannot identify any potential radiation rescuer controls. The lessons learned from this “failed” assay development are discussed.

Materials and Methods

Chemicals and reagents

Dactinomycin and piplartine were purchased from Selleck Chemicals (Houston TX); Melatonin²², Trichostatin A²⁶, γ-Tocotrienol²⁵, Auranofin¹⁸, Genistein²⁷, and Mito-Tempo²⁰ from Cayman chemicals (Ann Arbor, MI); GC1149¹⁹ from Galera Therapeutics (Malvern, PA); Recilisib sodium (also named as Ex-Rad)¹⁷ from MedKoo Biosciences (Morrisville, NC; V028–5832 (also named as compound C)²¹ and Y600–0815(also named as PKUMDL-LC-101)²³ from Enamine, (Monmouth Junction, NJ). The stock solutions of all these chemicals at 10 mM in DMSO were prepared, unless specified otherwise.

Cell-permeable Hoechst 33342 dye, and Calcein-AM were from Thermo Fisher Scientific (Waltham, MA). 1 mM and 2 mM (1000x) stock solutions were prepared according to the vender’s instructions, respectively.

Cells and cell culture

All cells were incubated at 37°C in a humidified atmosphere of 5% CO2. An hCMEC/D3 BBB cell line from Millipore Sigma (Burlington, MA) were cultured in DMEM:F12 (Dulbecco’s Modified Eagle Medium nutrient mixture Ham’s F-121 in a 1:1 mixture) supplemented with 10% filtered fetal bovine serum (FBS) (10:100 media), 100 U/mL penicillin/100 μg/mL streptomycin (1:100 media), sodium pyruvate (1:100 media), L-glutamine (2:100 media, unless media already contains L-glut, then add 1:100), and non-essential amino acids (2:100 media). HUVEC cells from American Type Culture Collection (ATCC, Rockville, MD) were cultured in endothelial cell medium-basal (ECM-b) from Sciencell Research Laboratories (Carlsbad, CA). In the 500 mL of basal media added 25 mL tube of Bovine Serum, Endothelial Cell Growth Supplement (ECGS), and 5 mL of penicillin/streptomycin (P/S). HEC-50 cells from Dr. Kim Leslie and Dr. Xiangbing Meng were maintained in DMEM supplemented with 10% FBS and 1% P/S (all from Gibco BRL, Carlsbad, CA).

Radiation and drug treatment

Before radiation exposure, the cells were cultured in T75 flasks in less than 90% confluency before the cell seeding with routine cell detachment/splitting and counting for each cell lines. In most case unless otherwise specified, the cells were seeded onto 96 well-microplates (Perkin-Elmer CellCarrier-96 Ultra microplates, tissue culture treated, black, 96-well with lid, Waltham, MA) at a cell density of either 2500, 10,000 cells/well and incubated for 48 hours to ensure that the cells formed a high density, confluent monolayer, but still separated at a low cell density.

The control compounds were prepared in a 96-well drug plate (Greiner, Microplate, 96-well, polypropylene, U-bottom, natural) for the serial dilution (1:3 for 8 wells or 1:2 for 12 wells) of each control compounds. The drugs were added 24 hours, 4 hours, or immediately before the irradiation treatment (unless otherwise specified) with a change of culture media using the Hamilton MicroLab Star liquid handling system (Reno, NV).

Irradiation was performed on above-mentioned 96 well microplates with the full culture media, with or without control compound treatment in serial dilutions. Cells were irradiated from 0 up to 440 Gy X-rays in the Radiation and Free Radical Research Core, either delivered with a PANTAK HF-320 ortho volt X-ray unit or an Xstrahl small animal radiation research platform (SARRP) unit with different dose rates and doses (Gy) as specified. Cells in complete culture medium in microplates were exposed to irradiation at 22 °C, usually within half an hour. Two approaches of irradiation were applied on the micro-plate irradiation: one was covered with Lead metal plates so that one single SBS standard 96 well plate can be divided into 3–4 blocks with different doses by the controls of the irradiation time. The other micro-plate was set up using one dose per plate.

The PerkinElmer Operetta High Content Imaging System (Waltham, MA) was used to monitor micro-plates using DPC imaging at 1-hour intervals during the first 12 hours, followed by either 2- or 8-hour intervals over 24 hours, unless otherwise specified. Usually, each plate was imaged before irradiation as well (denoted as −1 or −2, negative here indicates hour before irradiation). After 24 hours, plates were imaged every 24 hours, after the daily media change, up to 7 days. At the end of the experiment, all the plates were stained with Hoechst for nuclear staining, Calcein AM for live-cell staining, and/or ethidium homodimer for dead-cell staining. Three additional channels of excitation (360–400 nm) and emission (500–550 nm) for Hoechst, excitation (460–490 nm) and emission (500–550 nm) for Calcein AM, and excitation (520–550 nm) and emission (580–650 nm) for ethidium homodimer were imaged.

Image analysis were done by instrument accompanying Harmony software for single DPC channel or four (DPC, Hoechst, Calcein-AM and Ethidium Homodimer) channels. Nuclei and cell were selected for both intensity of all channels and morphology studies. The responses of radiation doses, control compounds doses, and time course responses were analyzed by individual wells and/or individual cells within the wells using Spotfire (TIBCO and Perkin-Elmer, Waltham. MA) software for the visualization of the datasets. In cases where there was a dose response, additional analysis was performed with GraphPad Prism (Prism 9, San Diego, CA).

Z’ factor and S/N ration are calculated according to the following formular:

$$Z^{\prime }=1-3*\frac{SD1+SD2}{Abs\left(Av1-Av2\right)}$$ Formular I

$$\frac{S}{N}=\frac{Abs\left(Av1-Av2\right)}{\surd \left(SD1*SD1+SD2*SD2\right)}$$ Formular II

Results & Discussion

Assay development scheme: how and why?

The objective of this assay development was to develop a microplate-based, in vitro, high-throughput assay that can be used for the discovery of and screening for small molecule radiation protectors and radiation mitigators in the endothelium upon radiation damages.

Current epithelial cell-based models for radiation damage detection use hCMEC/D3 and HUVEC cells, and in some cases, bovine aortic endothelial cells (BAOEC) are also used to evaluate radiation damage. In this report, hCMEC/D3 and HUVEC cells were used as cell-based in vitro models; the hCMEC/D3 cell line was used for its availability and ease of use, and primary HUVEC cells were used for their human relevance.

The phenotypic assays intentionally were picked for major assay development based on the earlier observation of the transient cell size changes when irradiated. It also reflects the uncertainty of molecular mechanism of actions, especially the dose-dependent and time-related mechanisms and responses. There have been reports on target-based assays for radiation responses, e.g., DNA damage-based; mitotic catastrophe and mitotic death; apoptosis, necrosis and senescence; necroptosis and ferroptosis; and bystander effect, e.g., immunogenic cell death. However, no dominant factor was identified that correlated yet well with the radiation cellular effects. Thus, for this project, the phenotypic assay development was the major direction we chose.

Our hypothesis for working with a phenotypic assay (Figure 1) was that the radiation caused acute cellular responses (< 24 hours) as well as chronic responses (from 24 hours to 7days), and the additional effects of rescuers/sensitizers were observed respectively. The responses of radiation were time-dependent, irradiation dose-dependent, and rescuer/sensitizer dose dependent. Briefly, live cell imaging was performed that acute responses were monitored with ~2–4 hour intervals for the first 24 hours (Day 1) and followed by a 24 hour intervals up to 7 days (Day 7), to detect acute and chronic post-radiation responses (i.e. short-term and long-term effects) using label-free DPC imaging. Usually at the endpoints of both cases, the same plate was used to further quantify the phenotypes of radiation using nuclear, live-cell, and dead-cell staining. Both rescuing and sensitizing compounds (control compounds tested) in acute phase (<24hr) and chronical phase (up to 7 days) can be monitored by the transient phenotypes of cell numbers, cell area (μm²), cell roundness and contrast of DPC; and endpoint cell staining results (among other parameters tested). These efforts were to explore and discover the optimal parameters (time and phenotypes) for the rescuing/mitigating compound responses. Both well-based image analysis and single cell-based image analysis were used for the characterization/optimization of the irradiation and drug effects.

Figure 1:

Hypothesis of acute and chronic responses of cells upon irradiation. A: Schematic early time responses (<24 hrs) with cell area (or others). Green: rescuing compounds with increasing dose upward. Red: sensitizing compounds with increasing dose downward. Black: Radiation effect without compounds. B: Schematic long term time responses (<7days) with cell number (or others). Green: rescuing compounds with increasing dose upward. Red: sensitizing compounds with increasing dose downward. Black: Radiation effect without compounds. 20K cells/well of hCMEC/D3 cells was used for the monolayer formation within the experimental timeframe. The radiation dose (e.g.,2.5 Gy) was picked arbitrarily from the experiments.

To validate our hypothesis of the phenotypic assay, we designed the experiment as summarized in Supl_Scheme 1. Multiple parameters of phenotypes in the cellular responses of irradiation were explored for the correlation with the radiation doses. The radiation dose rates, the irradiation doses at the same dose rate, two cell line models (mainly hCMEC/D3 with further validation with HUVEC), and approximately ~10 different control compounds (Figure 2) in serial dilution were systematically interrogated to try to develop the optimal robust assays for the effect of radiation.

Figure 2:

Representative control compounds tested in dose responses in the phenotypic assay development as positive or negative controls to evaluate the assay responses.

The control compounds (Figure 2) used as positive or negative controls tested in dose responses in the phenotypic assays were from the three categories of reported irradiation rescuing effects: 1) Antioxidants or anti-inflammatory drugs (e.g., Auranofin¹⁸, GC1149¹⁹, or MitoTempo²⁰); 2) 8-Oxoguanine glycosylase (OGG) activators (e.g., V028–5832²¹, or Melatonin²²); 3) GPX4 activators through ferroptosis (Y600–0815²³ or (±)-α-Tocopherol Acetate²⁴ as a Vitamin E analog); 4) Hits/leads from whole animal irradiation experiments (e.g., recilisib sodium¹⁷ also as Ex-Rad, and γ-Tocotrienol²⁵); 5) others, e.g. HDAC inhibitors (Trichostatin A²⁶). Dactinomycin and piplartine were used as negative controls or sensitizer controls.

The assay development workflow (Supl_Scheme 2) was designed to achieve the following objectives: 1). Determine the maximum window of signals for the phenotypes the maximal irradiation doses (up to 444 Gy at 22.2 Gy/min). Earlier trials with low doses (<10 Gy) of irradiation have shown only minimal, if any, post-irradiation signal. The maximum windows of signals were found to be time-point dependent. 2) Optimize the radiation doses to observe the dose responses of the phenotypes (enough irradiation signal windows) to interrogate control compound effects, i.e., to generate dose responses of rescuing/mitigating or sensitizing controls. 3) Reduce radiation doses to find optimal conditions (doses and timepoints) for rescuing/mitigating control compounds, preferably with highest Z’ factor or optimal dose responses of the controls to ensure assay robustness. This reduction of radiation doses will also be used to accommodate the different mechanisms of action for low-dose irradiation.

The results from this workflow, experimental design and timeline are discussed separately in the following sections.

Radiation doses and cellular responses

For the development of our in-vitro plate-based assay, the maximum radiation dose up to 444GY at 22.2 Gy/min was applied to the hCMEC/D3 cells. Representative live cell imaging of the time course responses upon irradiation was shown in Figure 3. The label-free DPC imaging of the plate provided the time dependent phenotypical changes in cell numbers, cell area, cell roundness and cell contrast of DPC intensity.

Figure 3:

Images (digital phase contrast, DPC) for the time-course of hCMEC/D3 cells with 444Gy dose of irradiation at 22.2 Gy/min. Seeding in 20K/well and monitored for 2 days post-irradiation.

The radiation dose effects on the cells were further studies at 4 different doses at same dose rate of 22.2 Gy/min at seeding density of 20K/well (monolayer formation). The post-irradiation cell response over 72 hours is summarized in Figure 4. The cell number responses post-irradiation were shown in Figure 4A, as the increasing irradiation from 0 Gy to 111Gy, the cell number decreased accordingly with increasing irradiation doses alone the time up to 72 hr. The cell area (μm²), cell roundness and cell DPC intensity contrast upon the irradiation at 0, 25, 55, and 111 Gy were shown in Figure 4B, 4C and 4D, respectively. All of them showed radiation dose dependent changes.

Figure 4:

Radiation dose responses of the endothelial cells in 72 hr time course. A: Cell numbers in doses of 0 (green), 25 (Red), 55 (Yellow), and 111 (purple) Gy at 22.2 Gy/Min. B: Cell area (μm2) in doses of 0 (green), 25 (Red), 55 (Yellow), and 111 (purple) Gy at 22.2 Gy/Min. C: Cell roundness in doses of 0 (green), 25 (Red), 55 (Yellow), and 111 (purple) Gy at 22.2 Gy/Min. D: Cell Digital Phase Contrast (DPC) intensity contrast in doses of 0 (green), 25 (Red), 55 (Yellow), and 111 (purple) Gy at 22.2 Gy/Min. hCMED/D3 cells in 20K/well, monitored for 2 days after irradiation.

Radiation doses and control drug dose responses

The effects of six control compounds (Auranofin, Dactinomycin, GC1149, Mito-Tempo, Recilisib sodium, and Trichostatin A) in serial dilutions were tested at four different radiation doses at the same dose rate of 22.2 Gy/min (Figure 5). At 0 Gy irradiation, auranofin, dactinomycin, and trichostatin A already had the dose responses on inducing cell death. With increasing radiation doses, these three compounds had similar potency (IC50s) but increasing efficacy of inducing cell death. They can be considered as irradiation sensitizers (Figure 5A). GC1149, Mito-Tempo, and Recilisib sodium had no effect on cells in 0 Gy. No dose responses from these three compounds were observed in increasing radiation doses, merely only irradiation effect itself with the time course responses. Marginal protection was observed in low concentration range for these control compounds (<0.1 uM), although this result was not reproducible.

Figure 5:

Compound effects (dose responses) in different radiation doses. A: Cell numbers in different timepoints −2 hr (Blue), 24 hr (Purple), 25 hr (Green), 48 hr (Red) and 72 hr (Yellow). Top: Irradiation doses at 0, 25, 55, 111 Gy at 22.2 Gy/Min. Right vertical: compound names: Auranofin, Dactinomycin, GC1149, Mito-Tempo, Recilisib sodium, and Trichostatin A. B: Cell area (μm²) in 0 (green), 25 (Red), 55 (Yellow), and 111 (purple) Gy at 22.2 Gy/Min. C: Cell roundness in 0 (green), 25 (Red), 55 (Yellow), and 111 (purple) Gy at 22.2 Gy/Min. D: Cell Digital Phase Contrast (DPC) intensity contrast in 0 (green), 25 (Red), 55 (Yellow), and 111 (purple) Gy at 22.2 Gy/Min. hCMED/D3 cells in 20K/well, monitored for 2 days after irradiation. hCMED/D3 cells in 20K/well, monitored for 2 days after irradiation.

The post-irradiation effects of these six control compounds on other phenotypes, including cell area (μm²), cell roundness, and cell DPC intensity contrast, are shown in Figure 5B, 5C and 5D, respectively. Similar results were also observed for auranofin, dactinomycin, and trichostatin A, along with GC1149, Mito-TEMPO, and recilisib sodium. No significant protection was detected with the potential rescuing/mitigating molecules.

The effect of the radiation doses on the cell numbers for control compounds (Figure 6) demonstrates an radiation dose-dependent relationship. At 10 Gy, most of the compounds reached >80% cell death at 72 hours. Hence, 10 Gy was used for later experiments with the rescuing/mitigating molecule assay.

Figure 6.

Radiation dose effects with different drug controls and time course responses. Top: control drug concentrations in μM. Right vertical: compound names: Auranofin, Dactinomycin, GC1149, Mito-Tempo, Recilisib sodium, and Trichostatin A. hCMED/D3 cells in 20K/well, monitored for 2 days after irradiation. hCMED/D3 cells in 20K/well, monitored for 2 days after irradiation.

The timing of the application of the small molecule rescuers/mitigators in all experiments is intentionally set at 24 hr before the irradiation. This timing could maximize the drug effect (protection or sensitization) in the cells in combination with the serial doses of the drugs. This could minimize the effect of timing by covering all the possible scenario with diverse concentrations instead of single concentration. No drug in the cell anymore after irradiation. All the effects are due to the downstream effects of the drugs before irradiation. This is also intentionally to correlated with potential clinical application to eliminate radiation side effects before the treatment. Application after irradiation was not explored at this time because the different drugs might take different time to be effective and might ly induce further compound effects for the result interpretation.

Non-linear cell responses and irradiation doses:

It has been established that the post-irradiation noxious effects to cells and tissue are non-linear at low and high doses²⁸, arbitrarily separated at around 0.1 Gy, based on epidemiology studies (Supl_Scheme 3A)^29,30. Several non-extrapolatable effects have been reported, including a bystander effect³¹, linear extrapolation, adaptive responses³² and hormesis³³, with different kinds of mechanisms of actions at low doses. Especially important for this project is the discovery for the radiation rescue/mitigation, which usually in the loosely defined low dose range. For this consideration, the radiation dose has to be intentionally set at less than 10 Gy to have enough possibility to be rescued/mitigated by the small molecule compounds. Otherwise, the rescuing/mitigating effects of the small molecules (in others as well) could not be detected due to higher radiation damages. Alternatively, too low irradiation doses would limit the window of the signal and potentially the robustness of the assay. The non-linearity of the radiation dose effects eliminates the possibility of using higher radiation doses for larger signal windows, with extrapolation back to low radiation doses.

Furthermore, the bystander effect (Supl_Scheme 3B)³⁴ is also an important mechanism for the low irradiation effect for cell-to-cell interactions. The direct-hit cells and non-direct-hit cells (hence “bystander”) form two distinct populations of cells that that not only have unique responses to radiation, but also introduce downstream cell signaling responses through cell-cell tight junctions and intercellular signaling. By controlling cell density, the direct-hit cell population can be identified using low cell density, whereas both direct-hit and non-direct-hit cell populations can be detected using a monolayer of cells. In this report, we used a monoculture of epithelial cells, but multi-culture cellular models are available to observe more complicated bystander effects.

The post-irradiation effects of the cell density and different cell monocultures (hCMEC/D3 and HEC-50) were tested (Figure 7). hCMEC/D3 cells, an epithelial cell lines, did show cell density-dependent responses. This is displayed by the DPC response (increasing DPC cell number), DPC intensity (increasing contrast), hoechst nuclei staining (increasing nuclei number), ethidium homodimer for dead cells (increasing dead cells), and calcein-AM for live cells (not many live cells) at 444 Gy irradiation dose. In contrast, HEC 50, an endometrial cancer cell line, has shown the DPC response (increasing DPC cell number), DPC intensity (increasing contrast), hoechst nuclei staining (increasing nuclei number), ethidium homodimer for dead cells (decreasing dead cells), and calcein-AM for live cells (increasing live cells) at an irradiation dose of 444 Gy.

Figure 7.

Cell density effects of hCMEC/D3 and HEC50 cells upon irradiation by 444 Gy at 22 Gy/min at 72hr time point. The cells at different cell densities were stained with Hoechst for nuclei, Calcein-Am staining for live cells and Ethidium homodimer as dead cell staining. DPC (Digital phase contrast) as label-free cell staining.

Irradiation rescuing/mitigating molecule assays

Based on the above-mentioned optimization of radiation doses, cell density, and time course responses from acute (<24 hour) and chronic (up to 7 days) phases, we further tested irradiation doses of less than 10 Gy with reported radiation rescuing control compounds and live-cell, time-course imaging.

As for the acute radiation effect (<24 hour) at 2.5 Gy, there were no significant effects on the cell number (Supl_Figure 8) for all tested control compounds (dactinomycin, MitoTempo, V028–5832, Y600–0815, melatonin, (±)-α-tocopherol acetate in serial dilutions), except the dactinomycin, a known cell killing drug (top row). Other parameters of post-radiation effects, i.e., cell area (μm²), cell roundness and cell DPC intensity contrast, for these six control compounds were similar (Supl_Figure 9). Cell density effects were also observed, with dactinomycin having more potency/efficacy in lower cell seeding density (5000 cells/well) than in higher (monolayer forming, 20000 cells/well).

As for the chronical radiation effect (up to 7 days) with the same platemap with 0, 2.5, 5 and 10 Gy/plate of irradiation, only dactinomycin demonstrated dose responses with increasing efficacy in cell number changes with increasing irradiation doses (Supl_Figure 10). Other parameters of post-radiation effects, i.e., cell area (μm²), cell roundness and cell DPC intensity contrast, for these six control compounds were similar (Supl_Figure 11).

No dose responses were observed in both acute and chronical irradiation effects with five reported rescuing control compounds in serial dilution and four different doses of “low-dose” irradiation.

Based on these results we decide not to proceed with the pilot screening for the following considerations: 1) With the current phenotypic assay approach, we could not generate a dose response with the control compounds, even though some reported rescuing molecules have already been tested in vivo and in some clinical trial cases³⁵. 2) Multiple mechanisms of action, especially in the low-dose irradiation condition, might not generate enough of a signal window to produce the necessary cellular phenotypes for irradiation rescuing/mitigating compounds. 3) For low-dose irradiation, the signal window is too small to be considered for physiological relevance. Despite our efforts, we were not able to determine the optimal radiation dose. Radiation doses that were too low resulted in a small signal, and those that were too high (or even at a medium high level) raised concerns about the relevance by extrapolation and difficulty of the rescuing.

We are reporting this “failed” assay to demonstrate that even with much effort and a strong rationale supporting the different conditions tested, the signal window can still be limited to the extent that it does not generate a robust assay for potential screening effects. These lessons learned fall into the category of the proverbial “valley of death” often encountered in assay development, which can also be similar to the challenges experienced during clinical trials: 1) phenotypic assays are not exactly physiologically relevance, and 2) target-based assays are not necessarily correlated to phenotypic observations. The different kinds of targets discussed above can only be considered as related, but not correlated, especially when the kinetic time-course measurement is used. To overcome these “valleys of death” in assay development, more thorough systematic approaches are needed to bridge the gap between assays and phenotypes, similar to the efforts in this “failed” assay. In addition, there is a need to target the responses required to facilitate screening efforts that allow for the discovery of in vivo, and preferably clinical-trial-relevant hits, which can eventually lead to therapeutics.

Despite the “failed” development of phenotypic assays for irradiation rescuing compounds in this report, we are still working on alternative assays: 1) Target-based assays based on the low dose mechanisms of actions, e.g., Irradiation induced direct DNA damage level and recovery, mitochondrial DNA damage and recovery, ROS and NOS as cellular signaling pathways for both direct hit cells and bystander cells. 2) Better in-vitro cellular models for irradiation effects, such as the transwell systems for irradiation induced the leaking of epithelial monolayers, wound-healing functional models for epithelial cells, and 3D Organ-on-a-Chip models for epithelium (e.g., MIMETAS system). 3) Innate immune-responses upon the irradiation as well as co-culture immune-responses for the irradiation effects.