A High Content Clonogenic Survival Drug Screen Identifies MEK Inhibitors as Potent Radiation Sensitizers for KRAS Mutant Non-Small Cell Lung Cancer

Abstract

Introduction

Traditional clonogenic survival and high throughput colorimetric assays are inadequate as drug screens to identify novel radiation sensitizers. We developed a method which we call the High Content Clonogenic Survival Assay (HCSA) that will allow screening of drug libraries to identify candidate radiation sensitizers.

Methods

Drug screen using HCSA was done in 96 well plates. After drug treatment, irradiation, and incubation, colonies were stained with crystal violet and imaged on the INCell 6000® (GE Health). Colonies achieving 50 or more cells were enumerated using the INCell Developer image analysis software. A proof-of-principle screen was done on the KRAS mutant lung cancer cell line H460 and a Custom Clinical Collection (146 compounds).

Results

Multiple drugs of the same class were found to be radiation sensitizers and levels of potency seemed to reflect the clinical relevance of these drugs. For instance, several PARP inhibitors were identified as good radiation sensitizers in the HCSA screen. However there were also a few PARP inhibitors not found to be sensitizing that have either not made it into clinical development, or in the case of BSI-201, was proven to not even be a PARP inhibitor. We discovered that inhibitors of pathways downstream of activated mutant KRAS (PI3K, AKT, mTOR, and MEK1/2) sensitized H460 cells to radiation. Furthermore, the potent MEK1/2 inhibitor tramenitib selectively enhanced radiation effects in KRAS mutant but not wild type lung cancer cells.

Conclusions

Drug screening for novel radiation sensitizers is feasible using the HCSA approach. This is an enabling technology that will help accelerate the discovery of novel radiosensitizers for clinical testing.

Introduction

Radiation plays an important role in the treatment of cancer of all types. For a number of diseases, adding chemotherapy to radiation as a sensitizer has improved survival outcomes by improving locoregional disease control compared to radiation alone, but the improvement has only been modest¹. Further advancements in the field require accurate strategies to identify novel agents that could enhance radiation responses. One potential approach is to screen for drugs based on synthetic lethality, a well-described phenomenon in genetics where lethality to the cell is induced only if two or more genes are inactivated, but not so when individual genes are inactivated². This mechanism is seen in the susceptibility of BRCA1 or BRCA2 mutant breast or ovarian cancers to PARP inhibition^3–6, and for sensitivity to cell cycle inhibitors (chk1 and chk2, wee1, polo-like kinase, and aurora-kinase inhibitors) of TP53 mutant cancers treated with DNA damaging agents such as radiation and/or chemotherapy^7–9. Synthetic lethality screens have been employed to identify interacting genes using shRNA libraries^{10, 11} or with drug libraries for combination drug therapies¹², but have not been done with radiation treatment. While radiation sensitization with drugs is not technically defined as synthetic lethality, in that it is not a radiation enhancement in the face of genetic susceptibility, the output could be similar in that drugs can block pathways or molecules that mimic a genetic “hit”, and in that setting, radiation stress could render the cells more susceptible to cytotoxic injury. This could be the basis of sensitizer screens, identifying compounds which have little to no effects on the cancer cells themselves, but have significant synergy with radiation. However, current approaches for testing sensitizers are difficult to perform simultaneous screens of numerous compounds. Current gold standard approach for testing radiation sensitizers is the clonogenic survival assay (CSA). It is a robust and reproducible technique but is low throughput and impractical for drug screening. Various methods have been used to screen for radiation sensitizers, such as cell proliferation colorimetric assay¹³, colorimetric sulforhodamine B assay¹⁴, or γH2AX foci formation assay¹⁵, but such approaches do not appropriately identify compounds that inhibit low cell density clonogenic survival and therefore may not appropriate for radiation screening of compounds¹⁶. We sought to develop a method that would facilitate drug screen with radiation, capitalizing on the power of the traditional clonogenic survival assay in a higher throughput, less cumbersome format.

Materials and Methods

Cell Culture

The non-small cell lung cancer cell lines H460, A549, H661, H1299, H2030, EKVX were acquired courtesy of Dr. John D. Minna (UT Southwestern, Dallas, TX) and were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 2 mM L-glutamine (Life Technologies, Grand Island, NY). U251, DU145, MiaPaca2 and PC3 were obtained from the NCI DCTD cell repository and grown in RPMI-1640 supplemented with 5% FBS. Cells were grown at 37ºC under 5% CO₂ atmosphere in a humidified incubator.

Traditional Clonogenic Suvival Assay (tCSA)

Single cell suspension were created by trypsinizing <90% confluent monolayer of cells with 0.1% trypsin/EDTA, and seeded at various cell numbers depending on the dose in triplicates in 6 well plates at 3 mL media per well. After overnight incubation for <24 hours, drugs were added to the plates at IC₃₀ concentrations, and 6 hours later, the plates were irradiated at specified doses in 2 Gy increments, media exchanged at 72 hours, and incubated for an additional 10–14 days. The colonies were stained with 0.5% crystal violet in 50/50 methanol/water for 10 minutes, washed, dried and counted manually. All in vitro irradiations were done using a Co-60 gamma irradiator from a decommissioned clinical gamma-irradiator using a 10 cm × 10 cm field size at isocenter.

High Content Clonogenic Survival Assay

Single cell suspension were seeded at varying densities (1–400 cells per well depending on the amount of radiation used to treat the plates; for 2 Gy-50 cells, for 4 Gy-100 cells, for 6 Gy-200 cells) into standard 96 well tissue culture plates (BD Biosciences, San Jose, California), in a total volume of 100 microliters per well. After seeding the plates were placed in a humidified tissue culture incubator at 37 ºC and 5% CO₂ overnight but < 24 hours. Drugs at 11-fold the final concentration or diluent were added at 10 microliter per well (see details below), and incubated for another 6 hours prior to sham or irradiation. Plates were kept in the tissue culture incubator for an additional 4 days without disturbance, and colonies were stained with crystal violet as done for tCSA, washed, dried, and colonies enumerated by the IN Cell Analyzer 6000® (GE Healthcare, Pittsburgh, PA).

Automated high throughput microscopy imaging analysis (Fig. S2)

After crystal violet staining, the dried plates were placed into an automated plate loader (Peak Analysis and Automation, Inc., Colorado Springs, Colorado) and loaded automatically into the IN Cell Analyzer 6000® equipped with a 5.5 Mp large field-of-view sCMOS camera. Images from four overlapping fields per well (5% overlap between fields) were collected from each well using a 4x/0.20NA Nikon lens with 642nm excitation and 706nm emission. Four independent images were stitched together into a single composite image for further analysis by the GE IN Cell Developer software (version1.9). Image analysis began by identifying the individual cell colonies. Once the colonies were identified, the number of individual cell nuclei within each colony were identified and counted using masks generated from intensity and object filers. The analysis algorithm enumerated the total number of colonies per plate and linked the total number of cells with each identified colony. Colonies with 50 or more cells were reported. There were situations where colonies overlapped. In these cases, the colonies and the cells contained within were counted as a single colony.

Drug screening with HCSA method

One-hundred forty-six drugs representing targeted agents to various pathways (either first generation tool compounds to drugs in clinical testing or FDA approved) were arrayed in two 96 well plates (see supplementary Fig. 3B). The compounds were screened using the quantitative HTS method as described by Inglese et al.¹⁷ where each assay plate became a point on the concentration response curve thus all of the compounds could be screened on the plate at once. Ten microliters of drugs were dissolved and diluted to 0.1% DMSO in growth media ranging from 10 pM to 11 μM by full log dilutions and were added to four assay plates. Two of the four replicate plates were irradiated at 2 Gy while the other two plates were brought to the irradiator but not irradiated (sham treatment). This process was repeated for each drug concentration. Drug-induced radiation sensitization was determined by comparing the colony count of each well with a drug added in the sham treated plates to the corresponding well on a similarly treated plate that was irradiated. A drug was considered a radiosensitizer if there was more than a 25% reduction in the number of colonies containing 50 or more cells at a particular dose with radiation treatment than the corresponding well without radiation treatment.

Signaling analysis

Protein lysates were produced by the addition of RIPA buffer with protease inhibitors to monolayers of cells irradiated or not irradiated, with or without the addition of drug, scraped, resuspended by pipetting, centrifuged, and resuspended in SDS-page loading dye, boiled for 5 minutes, and 50 mg protein were loaded per lane of a 10–15% SDS-PAGE gel. Immunoblotting was performed using antibodies against total ERK, phospho-ERK, and pAKT. Beta-actin was used to check for protein loading (all antibodies were acquired from Cell Signaling Technology, Danvers, MA)

Gamma H2AX staining

Immunofluorescence for gamma-H2AX was done at various time increments up to 24 hours after gamma-irradiation. Cells were fixed with 4% paraformaldehyde, washed, permeabilized, and antibody against gamma-H2AX (Cell Signaling Technology, Danvers, MA) (1:100) were added to the cells and incubated at 4ºC overnight. Cells were then washed, and incubated with FITC-conjugated secondary antibody for 1 hour at room temperature, washed, mounted and visualized with a fluorescence microscope.

Cell cycle analysis

Cells were pretreated with 30 nM trametinib (Selleckchem.com, Houston, TX) for 6 hours and gamma-irradiated. The cells were then placed back in the incubator, and at various times after irradiation, the cells were washed in PBS, trypsinized, fixed and stained with propidium iodide + RNase A. Flow cytometry was conducted using a Gallios Flow Cytometer (Beckman Coulter, Indianapolis, IN).

Tumor growth delay

Animals were maintained in an Association for the Assessment and Accreditation of Laboratory Animal Care approved facility in accordance with current regulations of the U.S. Department of Agriculture and Department of Health and Human Services. Experimental methods were approved by and in accordance with institutional guidelines established by the Institutional Animal Care and Use Committee. Tumor xenografts were produced in the leg of male NCr nu/nu mice by intramuscular inoculation of 1 × 10⁶ H460 cells in 10 μL. Irradiation and trametinib treatment were started when tumors reached 8-mm diameter. Trametinib (2 mg/kg) as a daily oral gavage in the morning was given either alone or with irradiation 6 hours after the drug dose. Drug treatment lasted a total of 14 days. Drug treatment was done concurrently with daily irradiation with 2 Gy in the first 5 days of treatment. Tumor growth was determined using thrice weekly caliper measurements. Mice were euthanized by CO₂ inhalation when tumors reached 15 mm in diameter.

Data Analysis

Plating efficiency (PE) for both tCSA and HCSA methods was estimated by dividing the number of colonies by the number of cells seeded in an otherwise untreated plate. This number was used for normalization in calculating the survival fraction, by the equation: SF = number of colonies/(number of cells seeded) (PE). Survival curves were generated with the use of Sigma Plot.

The growth-delay effect of trametinib with radiation (enhancement factor EF) was quantified as the ratio of differences between times required to reach 12 mm diameter:

$$EF=\frac{\text{Time}\text{to}12\text{mm}(\text{Drug}+\text{radiation})-\text{Time}\text{to}12\text{mm}(\text{Vehicle}+\text{radiation})}{\text{Time}\text{to}12\text{mm}(\text{Drug})-\text{Time}\text{to}12\text{mm}(\text{Vehicle})}$$

The times to reach 12 mm were estimated from the parameters of linear regressions of the four growth curves, with fits limited to the segments of the curves that were linear (growth was linear for times longer than 4 days for Vehicle, longer than 6 days for Vehicle + radiation, longer than 12 days for Drug, and longer than 16 days for Drug+radiation). 95% confidence intervals for EF were estimated from nonparametric bootstrapping of the above ratio of differences in times calculated from linear regressions of the growth curves.

Results

Development of a high content drug screen for radiation sensitizers

In the traditional CSA (tCSA), the two most cumbersome and time consuming steps are initial plating of cells and final colony quantification (Fig. 1A). The initial plating step in the tCSA requires handling of large amounts of growth media and dishes or plates. The final step requires long incubation periods to allow single cell clones to reach colony sizes of ≥50 cells. While automated colony counters have greatly streamlined the process from the manual one in the past, the throughput is still not feasible for screening multiple drugs at different concentrations. Adapting tCSA for high content drug screening requires streamlining these steps by seeding cells in a 96 well plate format and automating colony inclusion and enumeration. To determine if it is feasible to miniaturize the tCSA, which we will henceforth call the High Content Clonogenic Survival Assay (HCSA), the lung cancer cell line H460 was seeded at different cell numbers, and placed in culture to assess clonogenicity. We found that single cells were able to grow discretely as single colonies, reaching colony sizes of ≥50 cells by day 5, and reflect the plating efficiency (PE; or number of colonies formed divided by number of cells seeded) expected of this cell line using traditional approaches (~50–70%) (Fig. 1B, Fig. S1A). A similar experiment was done with another KRAS mutant lung cancer line A549 and despite differences in growth characteristic (more spread out and mesenchymal-like), similar results of single colony generation were obtained (Fig. 1C). We found the colony counts and plating efficiency of replicate wells between plates and at different times were quite reproducible, generating around the same number of colonies depending on the numbers of cells seeded on day zero (Fig. S1B). There was an absolute 20% reduction in PE with 2 Gy radiation, and the levels remain fairly constant between 10 and 200 cells, after which PE dropped off significantly when colonies became too crowded to allow for accurate automated enumeration (Fig. S1C). Flow activated cell sorting (FACS) could be used for cell seeding for adaptation for robotic cell culture workstations; however PE was reduced by ~40% using FACS as compared to manual multichannel pipetting, which is probably gentler on the cells and has less impact on clonogenic survival (Fig. S1D). We next determined if HCSA could reproduce the same clonogenic survival results as obtained from tCSA. This was done for the H460 and several other cancer lines of various origins (PC3, DU145: prostate; Miapaca2: pancreas; U251: GBM). We found HCSA recapitulated the same results in terms of the survival curve and produced similar surviving fraction at 2 Gy (SF2) for most of the cancer cell types tested (Figs. 1D–E).

Figure 1. The HCSA recapitulates results from traditional clonogenic survival assay and allows the identification of radiation sensitizers.

(A). Comparison in techniques between traditional and HCSA methods. The workflow on the left panel depicts the typical workflow of the traditional assay, whereas the right panel depicts workflow for the HCSA. The length of arrows corresponds to the relative length of time it takes to execute each of the steps. (B). Left panel: crystal violet stain and low magnification images of the triplicate wells from a 96 well plate with cells seeded at various densities per well. Right panel: INCell6000 images of the corresponding wells. (C) A comparison of the colonies formed by two lung cancer cell lines with different morphologies, one the H460 which forms tighter aggregates whereas the A549 expresses mesenchymal characteristics and grows colonies in patterns that are more spread out. The image software was able to enumerate colonies of both types, as depicted by the image analysis. (D) Comparison in survival curves between the traditional and HCSA methods. HCSA recapitulates a similar survival curve as produced by the traditional method. (E) The survival fraction at 2 Gy (SF2) comparing various cell types: glioblastoma (U251), prostate cancer (DU145, PC3), and pancreatic cancer (Miapaca2).

To determine if HCSA could be used to identify radiosensitizing compounds, we tested vorinostat (SAHA), a HDAC inhibitor well known to radiosensitize^{18, 19}, in both tCSA and the HCSA. As expected, 1 μM vorinostat for 72 hours caused a significant radiation sensitizing effect using tCSA, with a dose enhancement ratio of 2.78 (Fig. 2A). Using HCSA, we also found that 200 and 400 nM vorinostat dose dependently enhanced radiation effect, with enhancement ratios of 1.49 and 2.27, respectively (Figs. 2B–C). We performed a similar comparison using another known radiation sensitizer, DNA-PK/p110α inhibitor PI-103²⁰. Because of the basal potency of this drug, the radiation enhancement effect was not as profound as seen in the tCSA, but the effect was still dose-dependent (Fig. 2D–F).

Figure 2. Validation of the HCSA method to identify radiation sensitizers using vorinostat and PI-103.

(A) One micromolar vorinostat was used for the tCSA method, but at such doses no clonogens were produced in the HCSA, and therefore lower doses (200 and 400 nM) were used (B–C). HCSA had very similar enhancement effects compared to tCSA, with a dose-dependent radiation enhancement effect. DER = Dose Enhancement Ratio as a radiation dose of control over drug at survival fraction 0.5 (SF0.5). (D) 500 nM PI103 was added to seeded cells for 6 hours prior to irradiation at indicated doses. Media was changed at 72 hours of total drug exposure. (E) PI103 at 25 nM or 50 nM (F) was added to seeded cells for 6 hours prior to irradiation at corresponding doses. Cells were not perturbed until crystal violet staining on day 5

HCSA identifies classes of sensitizers directed at Kras mutant lung cancer

Supplementary Figure 3A shows the schema for HCSA for screening of drug libraries. We used the clinically relevant radiation dose of 2 Gy and seeded 50 cells per well. After cell seeding and overnight incubation, serially diluted drugs from 10 pM to 1 μM were added to the cells in duplicates and incubated for 6 hours prior to irradiation. We screened two custom Clinical Collections (CC) 1 and 2 (146 compounds) that contained many small molecules in clinical testing with a few that were FDA approved (Fig. S3B). Each drug plate was tested in duplicates and the colonies in replicate wells were averaged (Fig. 3A). We identified several compounds that were cytotoxic to the cells alone (inhibitors to Chk, Aurora Kinase, PLK, PKC, proteasome) and some very potent compounds such as HSP90 inhibitors that killed cells at low nM concentrations. As expected, many of the HDAC inhibitors had some sensitizing effect to radiation (a leftward shift in the IC50 curve by ~10 fold). However, many of the drugs exhibited significant radiation sensitizing effect with some or no activity by themselves, such as inhibitors to PARP, SRC, DNA-PK, mTOR and MEK1/2. A summary of radiation enhancers found from the HCSA screen is listed in Fig. 3B. We found not all drugs of the same class were radiation sensitizers. For instance, out of the 6 PARP inhibitors, only 3 were found to be radiosensitizing with little cytotoxicity by themselves, with the order of potency observed being AG014699 > AZD2281 ≫ ABT888. No activity was seen for BSI-201, DR2313, and NU1025 (Fig. 3C). Similarly, the HDAC inhibitors Trichostatin, PXD101, PCI-24781 exhibited cytotoxicity with little radiation enhancement effects, while drugs like NVP LAQ824, JNJ-264815, LBH589, MGCD0103, MS-275, and vorinostat showed varying degrees of cytotoxic potencies and were enhanced by radiation by about 10 fold. In addition, we also found that suppression of clonogenic survival by many inhibitors of the downstream effectors of activated mutant KRAS (PI3K, AKT, mTOR, MEK) was enhanced by radiation (Fig. 3B). This demonstrates that for this Kras mutant NSCLC cell line H460, in addition to known sensitizers such as HDAC and PARP inhibitors, drugs that block downstream of Kras signaling also sensitizes these cells to the effects of radiation.

Figure 3. HCSA enables the identification of classes of radiation sensitizers through drug screening.

(A) Representative screening results of plate one of the Custom Clinical Collection (CC1) and plate two of the collection (CC2) on the H460 cells, with the identity of the drugs on the right side of each row of cells. The columns represent one drug concentration or sham control, and each row depicts one particular drug. #N/A are blanks with 0.1% DMSO or RPMI. A green block represents single agent activity without apparent radiation enhancement effect, a pink block represents drugs that are weakly enhancing (<1 log10 shifted) and red block represents drugs that are strongly enhancing (≥1 log10 shifted). (B) List of drugs that are identified in the screen to have significant enhancement effects. Drugs within the same class are listed together. The inset figure illustrates with asterisks a few of the proteins along the signaling cascade of activated KRAS that radiosensitize when blocked. (C) Examples of three classes of radiosensitizing drugs identified in the screen, namely the HDAC inhibitors, PARP inhibitors and MEK inhibitors.

The MEK inhibitor GSK1120212 (trametinib) selectively sensitizes Kras mutant lung cancer cells to radiation treatment

To determine if inhibition to downstream signaling of Kras specifically sensitized KRAS mutant lung cancer cells, we focused on the MEK1/2 inhibitors, given the importance of the recent clinical development of this drug in KRAS mutant lung cancer²³. We found many of the compounds in the class of MEK1/2 inhibitors were radiation sensitizers with variable potency between drugs. The first generation compounds CI-1040 and PD98059, which are low potency drugs not in clinical development, had relatively no independent effect. However several of the later generation MEK inhibitors such as PD0325901, AZD6244, and RDEA119 exhibited some single agent activity but whose effect were greatly accentuated in combination with radiation (Fig. 3C). Given the more advanced clinical development, potency, and bioavailability profile of the MEK1/2 inhibitor GSK1120212 (trametinib)²⁴, we tested this compound in greater detail in additional cell lines. We found trametinib used at IC30 concentration was a potent radiation sensitizer in KRAS mutant lung cancer cell lines but not in the 3 wild type (WT) cell lines tested (Fig. 4A). In both the H460 and A549 cells, radiation significantly enhanced pERK expression, which was not apparent in the Kras WT cell lines H661 and H1299 (Fig. 4B, Fig. S4). This pERK expression was abolished with trametinib treatment, and caused a paradoxical increase in pAKT activity, likely due to feedback inhibition on EGFR/PI3K/AKT activation by MEK/ERK²⁵ (Fig. 4B). We tested whether AKT inhibition would synergize with radiation in addition to trametinib, and found that there were no additional effects with this drug combination (Fig. S5). This suggests that feedback upregulation of PI3K/AKT from MEK inhibition does not attenuate the radiation enhancement response. We explored the mechanism by which trametinib sensitized H460 cells to radiation, and found that trametinib and radiation did not enhance DNA damage (as assessed by γH2AX foci) (Fig. S6A–B) or induce apoptosis (by PARP cleavage, Fig. S6C), but instead induced prolonged cell cycle arrest at G2/M (Fig. S7) and significantly reduced the S-fraction over 96 hours (Fig. 4C), leading to increased senescence over radiation alone, but not significantly greater than MEK inhibition alone (data not shown). Using the H460 cells in a xenograft model, trametinib in combination with 2 Gy × 5 fractions of radiation significantly prolonged tumor growth delay (Enhancement Factor=1.86 at 12 mm (95%CI=1.23–3.05)) compared to either agent alone (Fig. 4D).

Figure 4. The MEK inhibitor trametinib is a selective radiation sensitizer in KRAS mutant lung cancer cells.

(A) Traditional CSA was used to determine the relative potency of trametinib (MEKi) in various lung cancer cell lines, with or without KRAS mutations. This effect seems to be selective in cells lines with KRAS mutations, also the potency differ between the KRAS mutant lines. No effect was apparent in the wild type cells. (B) Radiation induced pERK activation that was seen in the H460 but not in the WT H661 cell line. This activation was fully blocked by treating cells with trametinib. (C) Cell cycle analysis in two KRAS mutant and two KRAS WT cells. Significant S phase prolongation and diminution of G1/S cells was apparent in the KRAS mutant cells treated with trametinib and radiation, but not seen with either agent alone or in the WT cell lines. (D) Xenograft model of H460 cells to determine the amount of growth delay with combined treatments versus single agent alone. Drug treatment lasted 14 days, the first 5 were combined with radiation at 2 Gy per day. Drugs was given by oral gavage at 2 mg/kg 4 hours prior to radiation administration.

Discussion

In the present study we show that HCSA is a novel platform to help discover drugs that can be combined with radiation for cancer therapy. The technique is sensitive enough to simultaneously detect small radiation enhancement effects by multiple drugs. We identified several drugs belonging to distinct classes of inhibitors that blocked specific pathways. This revealed apparent cellular processes that were synthetically lethal with radiation. One of the well-known sensitizer class of drugs are the PARP inhibitors. Of the 6 that were in our library, 3 came out to be radiation sensitizers, with a potency order of AG014699 > AZD2281 ≫ ABT888. All three of these drugs are in various phases of clinical testing combined with chemotherapy, and ABT888 is currently being tested with chemoradiation in locally advanced NSCLC in a phase I/II trial (SWOG 1206). The three other drugs, BSI-201, DR2313, and NU1025, had no activity in the HCSA screen. Interestingly, only one of these drugs, BSI-201, has gone into advanced clinical testing in phase III trials but was recently demonstrated to lack activity as a PARP inhibitor in vitro^{21, 22}

Interestingly, a number of these pathways were linked to signal transduction downstream of activated mutant KRAS, which is present in the particular cell line employed for this screening. We validated that inhibition of one of these pathways, MEK/ERK, with the best-in-class drug trametinib, caused prolonged cell cycle arrest and significant growth delay in combination with radiation in a xenograft model.

Many drugs that have promising systemic activity in the stage 4 disease setting are increasingly moving forward in phase II–III testing. While many of these drugs may be FDA approved for indications in advanced disease, the treatments remain non-curative, since most of the cancers develop resistance to the drug therapy. As a better understanding of the acquired resistance becomes known, more drugs are being developed that could bypass the resistance mechanism(s) or to potentially prevent the development of these cellular resistance strategies. These efforts by academia and industry are leading to even more specific and potent drugs that inhibit pathways specific to the cancer cells and minimize cross reactivity with normal tissues. The effectiveness of these treatment strategies may be enhanced if they are brought to more curative setting in stage III disease where both systemic and local therapies are critically important to render cures for patients. Coupled with improved staging methods to identify high risk patients with minimal residual or metastatic disease, drugs that can enhance radiotherapy as well as either having single agent activity or synergistic effects with chemotherapy are ideally suited to the curative setting. Methods that can elucidate these synergistic activities with radiation may accelerate the translation of these drugs to clinical testing. HCSA is a new strategy that has the potential to fulfill this unmet need. The power of this assay is the ability to predict the specific pathways that enhance radiation effects when multiple compounds in the same class independently demonstrate radiation enhancement effects within the same screen. However, the technique in its current form does have some limitations, namely the inability to test doses of radiation larger than about 6 Gy since the high number of cells that are needed to overcome the effects with radiation alone is too great to make any sensitive measurement of drug effect. However it has sufficient dynamic range between 2 and 4 Gy to make it amendable for screening drugs using these doses. Given the sensitivity of the assay, it was also difficult to ascertain the radiosensitive nature of compounds if they exert significant single agent activity, such as the HSP90 inhibitors, which are known sensitizers^{26, 27}. It is possible that at doses below 10 pM, which was the lower limit used in these experiments, we will be able uncover the radiation enhancement effects. Thus, drugs with these higher potencies may need to be assessed separately at lower doses. Lastly, while HCSA appeared capable of recapitulating similar SF2 values as compared to tCSA in a number of cell lines, there were some (e.g. DU145) that had different radiation sensitivity profiles or which couldn’t form adequate colonies in the HCSA. The problem of low plating efficiency or loosely associated colonies that impacts tCSA become a bigger issue in the small well format. Future studies will optimize conditions which may promote better colony formation for these difficult-to-use cells.

As drugs become more specific, these compounds act as “probes” for specific genes or pathways, and very quickly one can identify genetic susceptibilities for each cell line. By integrating the drug sensitivity profile for each cell line using the HCSA along with genomic, transcriptomic, and proteomic information, we should in the future be able to identify the molecular contexts that explain the synthetic lethal effects of these drugs on cells treated with radiation. This information, once extensively validated, could help personalize targeted therapies in combination with radiation in the clinic²⁸.

Supplementary Material

SDC 1. Supplementary Figure 1. HCSA optimization.

(A) Example of a plate seeded with varying cell densities (10–200) and stained with crystal violet. A bright field image of four wells from the plate (top) and a fluorescent image (bottom) from a single well on the same plate are shown to the right. (B) To determine reproducibility of the assay, cells (N=50) were manually seeded in 12 different plates, and 2 Gy radiation was administered to 6 of the plates. Colonies were counted to determine the between-plate reproducibility in the colony count. Each plate was a summary of 12 wells, with the mean and standard deviation representing the reliability of the measure. (C) Plating efficiency as related to the cell numbers seeded. The efficiency was stable between 10–200 cells, after which the efficiency decreases. To plate the different cell densities, replicates of 12 were seeded by FACS. (D) Manual versus FACS seeding, showing the differences in the efficiency in colony formation using these two approaches.

SDC 2. Supplementary Figure 2. Image analysis using INCell6000.

The numbers correspond to the step sequence for the image analysis. 1–2) Fluorescent images are collected from four fields from within each well and then digitally stitched together to create a composite image. 3) A colony mask was applied to identify colonies based on set parameters (green) and cellular debris or artifacts. 4) A cell mask was applied utilizing set parameters for cellular identification. 5–6) Combining these approaches and setting the criteria of colony of ≥50 cells, colonies that met this cutoff were enumerated (red colonies). Colonies not meeting this cutoff were excluded (gray colonies).

SDC 3. Supplementary Figure 3. Drug Screen using HCSA method.

(A) A schematic depicting the workflow of the drug screen. (B) Plate images of Custom Clinical collection plates 1 and 2 (CC1 and CC2) and the representative drugs in these plates, along with blanks that were included in the wells that served as internal controls. (C) Results of a typical screen from custom drug plate 1 (CC1), with the number of colonies as numbers in each well. The colors represent the relative number of colonies, with white being zero and red being the highest colony number in the plate. Each plate with a particular drug concentration was tested in duplicate.

SDC 4. Supplementary Figure 4. Radiation induced pERK activation that was blocked by trametinib.

Assessment of phospho-Erk (p-ERK) staining on western blot comparing 2 KRAS mutant (A549 and H460) and 2 KRAS WT cells (H1299 and H661). In both of the KRAS mutant cells, radiation was able to induce p-ERK staining, and this effect was fully blocked by trametinib (MEKi). This p-ERK activation was not evident in the two KRAS WT cell lines.

SDC 5. Supplementary Figure 5. AKT inhibition does not radioenhance beyond MEK inhibition.

Assessment of the radiation sensitizing or additive effect to MEK inhibition by adding an AKT inhibitor “427” to H460 cells. The 427 compound was added at 50 nM alone or combined with 30 nM trametinib, with or without radiation at 2 or 4 Gy. There was no evidence of any additional effects of AKT inhibition to the radiation sensitizing effects of MEK inhibition.

SDC 6. Supplementary Figure 6. MEK inhibition and radiation do not impact DNA damage repair or apoptosis. (A).

γH2AX staining of A549 cells irradiated at 6 Gy, and fixed at 0.5 hr and 16 hr time points, with (right panel) or without trametinib (MEKi) (left panel). Punctate foci formation was manually enumerated by capturing images at high power field (20X). (B) Quantitation of the foci showed little difference in foci formation at 0.5 or 16 hrs. There was perhaps a decrease in γH2AX formation in the RT+MEKi (trametinib) group at 16 hrs. No difference in γH2AX was also seen for H460 cells. (C) Apoptosis was assessed by PARP cleavage with either treatment alone or in combination in the cell lines indicated. Apoptosis was seen with radiation alone in the H460 cells that was not enhanced with trametinib.

SDC 7. Supplementary Figure 7. MEK inhibition with radiation induced prolonged cell cycle arrest at G2/M in KRAS mutant cells.

Cell cycle analysis of H460 (A) and H1299 (B) with trametinib (MEKi) alone, with radiation at 4 Gy alone or combined with 30 nM trametinib for 24 or 72 hours. The histogram shows the summary data from the FACS analysis of two cell lines from KRAS mutant (A549 and H460) and two KRAS WT cells (H1299 and H661). An increase in G2/M arrest was evident in the H460 cells at 72 hours. There was a trend in A549 cells but was not statistically significant. There was no evidence of enhanced G2/M arrest for either of the KRAS wild type cell lines.