Identifying radioresistant and radiosensitive rectal cancer patient-derived organoids
Previously, we screened rectal cancer-derived organoids for differential in vitro radiosensitivity using a Cell Titer Glo-based assay [12]. We selected from this study two of the most radioresistant organoids (HUB005 and HUB183) and two of the most radiosensitive organoids (HUB106 and HUB062) [12]. Tumour characteristics, clinical info, and mutation status are detailed in Supplementary Table S1. We performed two validation experiments to confirm the differential radiosensitivity of these organoids. First, the organoids were exposed to 5 Gy of radiation and apoptotic cells (with sub-G1 DNA content) were quantified by flow cytometry. This dose was chosen because it is well within the range that has been shown to correlate with clinical radiation responses [8, 9]. Radioresistant organoids (HUB005 and HUB183) showed a marginal increase in apoptotic cells (fold change 1.3 ± 0.5 and 2.9 ± 0.05, respectively), while radiosensitive organoids (HUB062 and HUB106) exhibited a significantly greater increase (fold change 5.9 ± 0.9 and 11.8 ± 0.7, respectively). (Fig. 1a, P-value = 2.3 ×10−2, Student’s t test, Supplementary Fig. S1a–c). Secondly, clonogenic survival assays showed similar differential sensitivities, with HUB005 and HUB183 retaining more clonogenic capacity than HUB062 and HUB106 (Fig. 1b–e, Supplementary Fig. S1d).
a Fold change of the percentage of sub-G1 cells from 0 to 5 Gy. Barplots represent the mean ± sd of two independent experiments. Clonogenic assay survival curves for HUB005 (b), HUB183 (c), HUB062 (d), and HUB0106 (e). Dots are mean ± sem. One independent experiment was performed.
Radioresistant organoids display increased transcriptional adaptability and DNA damage response following irradiation
To test whether radioresistant and radiosensitive organoids responded differently to irradiation, three-day-old organoids were exposed to 0 or 10 Gy of radiation. RNA was isolated from all organoids and bulk RNA sequencing was performed. (Methods and Supplementary Fig. S2a). Gene set variation analysis was performed for gene sets belonging to the KEGG [22], Reactome [23], Biocarta, and Hallmark [24] databases. In this analysis, 469 gene sets were significantly up-or downregulated in radioresistant organoids in response to irradiation, whereas 145 gene sets were significantly up- or downregulated in radiosensitive organoids (Supplementary Fig. S2b, Supplementary Tables S5 and S6). Subsequently, all the differentially expressed gene sets were plotted on a ‘gene set-gene set’ graph using a force-directed graph algorithm [25]. This algorithm is employed to visualise gene sets as nodes and positions gene sets that share many genes closer together, thus facilitating the identification of gene sets with similar functions.
This analysis revealed that radioresistant organoids significantly upregulated 43 DNA repair pathways (out of 469 significantly differentially expressed gene sets in total) (Supplementary Fig. S2c) [31]. The upregulated DNA repair pathways included those associated with double-strand break repair, such as non-homologous end joining and homologous recombination (Supplementary Fig. S2d). Radioresistant organoids also upregulated gene sets involved in cell cycle regulation (87 out of 469) such as M to G1 and G1 to S transition (upregulated), as well as mitosis-associated gene sets such as the resolution of sister chromatid cohesion and regulation of mitotic cell cycle (both upregulated). Furthermore, radioresistant organoids upregulated gene sets involved in the regulation of transcription (19 out of 469, e.g., gene sets involved in RNA polymerase I transcription termination and mRNA capping).
In contrast, radiosensitive organoids exhibited fewer upregulated DNA repair pathways in response to irradiation (16 out of 145), as well as cell cycle pathways (19 out of 145). Moreover, they upregulated pathways associated with programmed cell death (11 out of 145, Supplementary Fig. S2e,f, Supplementary Table S6). Finally, quantification of the transcriptional responses (Methods) revealed that the DNA repair pathways were significantly more strongly upregulated in radioresistant than in radiosensitive organoids (Supplementary Fig. S2g, P = 5.4 ×10−6).
These results link the differential sensitivities to irradiation of the organoids with the enrichment of gene sets known to be associated with radiation sensitivity [5] and indicate that radioresistant organoids have superior transcriptional adaptability in response to irradiation.
Radioresistant organoids have upregulated metabolism associated with ROS detoxification
While the previous analysis focused on how organoids respond to irradiation, distinctions between radioresistant and radiosensitive organoids may also exist inherently before radiation. These differences may persist or change upon irradiation. These inherent differences may be more easily translated into actionable treatments, as they can be identified before administering radiation therapy. Therefore, a comparison of expression profiles between unirradiated radioresistant and unirradiated radiosensitive organoids was made. Hierarchical clustering (Fig. 2a) and principal component analysis (Fig. 2b) of genome-wide expression data separated radioresistant and radiosensitive organoids into distinct clades and groups, respectively. The analysis of differential gene expression between these two groups identified 4,093 differentially expressed genes (Supplementary Table S7). Notably, the top upregulated genes in radioresistant organoids included several genes associated with detoxification, such as GSTT1 (involved in the conjugation of reactive oxygen species with glutathione), CYP1A1 (participating in detoxification of xenobiotic compounds), UGT1A6 (associated with glucuronidation), and ALDH3A1 (involved in acetaldehyde detoxification) (Supplementary Table S7). Gene set variation analysis and gene set-gene-set graph analysis indicated that radioresistant organoids exhibited increased expression of gene sets related to the detoxification of reactive oxygen species (ROS) and xenobiotics, including pathways such as glutathione metabolism, nicotinate, nicotinamide, and ascorbate metabolism (Fig. 2c, Supplementary Table S8). Furthermore, these distinctions between radioresistant and radiosensitive organoids were even more pronounced after irradiation (Fig. 2d, Supplementary Table S9), with ‘Glucuronidation’ and ‘Glutathione synthesis and recycling’ being the most differentially expressed gene sets, primarily upregulated in radioresistant organoids.
Principal component analysis (a) and unsupervised clustering (b) of radioresistant and radiosensitive organoids based on bulk RNA expression data. c Gene set-gene set graph of differentially expressed gene sets (false discovery rate controlled at 1%, using Benjamini–Hochberg) between unirradiated radioresistant and unirradiated radiosensitive organoids. Each dot indicates an MSigDB gene set. Gene sets that share many genes will cluster together and thus have similar functions. Gene sets are together by vertices, whose thickness is proportional to the number of genes shared between gene sets. Opaque dots indicate upregulated gene sets in radioresistant organoids, while translucent dots indicate downregulated gene sets. d Gene set-gene set graph of differentially expressed gene sets between irradiated radioresistant and irradiated radiosensitive organoids. e Gene expression of GCLC in unirradiated radioresistant organoids and unirradiated radiosensitive organoids. f Western blot of indicated proteins in unirradiated radioresistant and radiosensitive organoids. g Luminescence measurements of total glutathione levels of unirradiated radioresistant vs. radiosensitive organoids. Three independent experiments were performed.
We next focused on glutathione metabolism, as glutathione is one of the most important antioxidants in cells. GCLC, the rate-limiting enzyme in glutathione (GSH) synthesis, was found to be upregulated in unirradiated radioresistant organoids compared to unirradiated radiosensitive organoids (Fig. 2e, Benjamini–Hochberg (BH)-adjusted P-value = 2.7 ×10−3). Moreover, western blot analyses revealed strong protein expression of GCLC in radioresistant organoids HUB005 and HUB183, but not in radiosensitive organoids HUB106 and HUB062 (Fig. 2f). Finally, radioresistant organoids had higher levels of GSH, although this difference was not statistically different (Fig. 2g, Student’s t test, P = 0.059).
Identification of a synergistic combination therapy for radioresistant organoids
The above results suggest that radioresistant organoids may have a high capacity to resolve oxidative stress. An exciting new drug, RRx-001, a first-in-class dinitroazetidine (Fig. 3a), is a potent inducer of oxidative stress in cancer cells by forming covalent adducts with thiol sulfurs. As such, it depletes cellular stores of thioredoxin, cysteine, and glutathione [16, 32, 33]. RRx-001 has antitumour activity in several cancer types and is currently evaluated in multiple clinical trials [13,14,15]. As radioresistant organoids had upregulated detoxification metabolism, it was hypothesised that combining RRx-001 with irradiation could overwhelm the cells’ capacity to cope with oxidative stress. However, pretreatment of radioresistant organoids with RRx-001 failed to synergise with irradiation in killing radioresistant organoids (Fig. 3b).
a Chemical formula of RRx-001. b Viability of radioresistant organoids HUB005 and HUB183 at indicated RRx-001 concentrations and radiation doses. The conditions are normalised to DMSO-treated conditions (0 μM RRx-001; 0 Gy). YRx2Gy(E) / YRx4Gy(E): Bliss’ expected viability combining 0.5 μM and 2 or 4 Gy, respectively (Methods). c Schematic illustrating the mode of action of the screened drugs. d Dot plot showing synergy scores of the tested compounds with 20 μM BSO. Size reflects the concentrations of the compounds (Supplementary Table S4). Cys Cysteine, GCLC/GCLM Glutamate-cysteine ligase catalytic/modifier subunit, GSS Glutathione synthetase, GR Glutathione reductase, GSSG Glutathione disulfide, GSH Glutathione, GLS Glutaminase, GPx Glutathione peroxidase, G6P(D) Glucose-6-phosphate (dehydrogenase), IDH1/2 Isocitrate dehydrogenase 1/2, NADPH Nicotinamide adenine dinucleotide phosphate, NRF2 nuclear factor erythroid 2–related factor, NAMPT Nicotinamide phosphoribosyl transferase, Trxn Thioredoxin reductase, ME1/2/3 Malic enzyme 1/2/3. At least three independent experiments were performed for each drug combination and each organoid.
Because radioresistant organoids had increased glutathione metabolism compared to radiosensitive organoids, we evaluated whether RRx-001 could synergistically induce cell death when combined with inhibitors of GSH or GSH-related metabolism such as buthionine sulfoximine (BSO), erastin and telaglenastat. Other inhibitors of redox metabolism and inhibitors of pathways involved in the production of NADPH, such as polydatin and malic enzyme inhibitor (MEi), were also included, as well as agents affecting purine and pyrimidine metabolism, and the apoptotic activator venetoclax. In total, 16 drug combinations were tested for their effect on both radioresistant organoids (HUB005 (TP53, APC mutant) and HUB183 (TP53 mutant), Fig. 3c, Supplementary Table S1). For each combination in each organoid, a ‘Synergy Score’ was calculated to reflect the effectiveness of the combination compared to the expected cell death if the agents were to act independently (Methods). A synergy score of two, indicating that the combination killed twice as many cells as expected, was defined as synergistic, while a score of more than five was considered strongly synergistic. Most RRx-001-based drug combinations did not exhibit a synergistic effect (Fig. 3d). This observation aligns with recent extensive anti-cancer drug combination screens showing that synergy is rare [34,35,36]. However, the combination of RRx-001 with BSO, a selective inhibitor of Glutamate—cysteine ligase catalytic subunit (GCLC), displayed an extremely strong synergistic effect in both HUB005 and HUB183 (synergy score 113 [193–23.6], Fig. 3d).
Next, it was tested whether BSO or RRx-001 affected the cell viability at the concentrations used in the screen. While BSO strongly decreased GSH levels in radioresistant organoids HUB005 and HUB183 (Supplementary Fig. S3a), it was ineffective in inhibiting cell viability alone, even at doses as high as 100 μM (Supplementary Fig. S3b). RRx-001 alone had no apparent effect on cell viability for HUB005, while it reduced viability by approximately 50% in HUB183 (Fig. 4a, b). However, combining 20 μM BSO with 2 μM RRx-001 induced strong synergistic cell death in both HUB005 and HUB183, causing a near-complete eradication of these radioresistant organoids (Fig. 4a, b). The addition of N-acetyl cysteine (NAC) to the medium, an antioxidant widely used to counteract oxidative stress [37], effectively attenuated the synergistic effect of BSO and RRx-001 (Fig. 4c). Next, a CRISPR-Ecas9-engineered GCLC knockout (KO) variant of the colorectal cancer-derived organoid Tor10 (GCLC 24.1, Figure S3c) was used and compared to control Tor10 organoids that only expressed Ecas9 [27]. As previously described, GCLC KO alone did not interfere with cell viability (Supplementary Fig. S3d, e) [27]. While Tor10 Ecas9 control was insensitive to 1 μM of RRx-001 and retained viability even at 5 μM, GCLC KO organoids were nearly eradicated by treatment with 1 μM and 5 μM RRx-001 (Fig. 4d, e). The addition of NAC to the medium attenuated the effect of RRx-001 in GCLC KO organoids (Fig. 4f).
a Representative images of radioresistant organoids HUB005 and HUB183 after treatment with the indicated concentrations and drugs. Viability of radioresistant organoids at indicated RRx-001 concentrations, normalised to DMSO-treated conditions (0 μM RRx-001; 0 μM BSO) without (b) or with (c) the presence of N-acetyl cysteine (NAC). YRxBSO(E): Bliss’ expected viability for RRx-001 and BSO combined. d Representative images of Tor10 Ecas9 control and Tor10 GCLC knockout organoids. Organoids were treated with DMSO or with the indicated concentrations of RRx-001. Luminescence measurements of ATP levels (using CellTiter-Glo 3D) at indicated RRx-001 concentrations, normalised to DMSO-treated conditions (0 μM RRx-001) without (e) or with (f) the presence of N-acetyl cysteine (NAC). Three independent experiments were performed.
Combining buthionine sulfoximine with RRx-001 is synergistic in a set of colon cancer-derived organoids
To test whether the drug combination was synergistic in other aggressive colorectal cancer subtypes, we evaluated it on a panel of organoids derived from liver, peritoneal, and lymph node metastases; pretreated vs. chemoradiation-naïve tumours; microsatellite unstable tumours; cancers of consensus molecular subtype 4 (CMS4); and KRAS- and BRAF-mutated tumours (Fig. 5a). The combination therapy was found to be synergistic or strongly synergistic in three out of eight organoids (CRM1, HUB006, and p19b, Fig. 5b). Notably, two of these organoids carried the BRAF V600E mutation (HUB006 and p19b).
a Heatmap showing the characteristics of the organoids for which the BSO RRx-001 combination was tested. P19b has no APC mutation but is mutated in the wnt-related gene AXIN1. b Dot plot indicating synergy score for each tested organoid using 20 μM BSO and 2 μM RRx-001. A synergy score of x indicates the drug combination killed x times more cells than expected (Methods). Translucent dots indicate experiments in which the drug combination was not synergistic (synergy score <2). Each dot is an independent experiment.
link