Peritoneal metastases from colorectal cancer belong to Consensus Molecular Subtype 4 and are sensitised to oxaliplatin by inhibiting reducing capacity

Abstract

Background: Peritoneal metastases (PM) in colorectal cancer (CRC) are associated with therapy resistance and poor survival. Oxaliplatin monotherapy is widely applied in the intraperitoneal treatment of PM, but fails to yield clinical benefit. We aimed to identify the mechanism(s) underlying PM resistance to oxaliplatin and to develop strategies overcoming such resistance.

Experimental design: We generated a biobank consisting of 35 primary tumour regions and 59 paired PM from 12 patients. All samples were analysed by RNA sequencing. We also generated a series of PM-derived organoid (PMDO) cultures and used these to design and test strategies to overcome resistance to oxaliplatin.

Results: PM displayed various hallmarks of aggressive CRC biology. The vast majority of PM and paired primary tumours belonged to the Consensus Molecular Subtype 4 (CMS4). PMDO cultures were resistant to oxaliplatin and expressed high levels of glutamate-cysteine ligase (GCLC) causing detoxification of oxaliplatin through glutathione synthesis. Genetic or pharmacological targeting of GCLC sensitised PMDOs to a 1-h exposure to oxaliplatin, through increased platinum-DNA adduct formation.

Conclusions: These results link oxaliplatin resistance of colorectal PM to their CMS4 status and high reducing capacity. Inhibiting the reducing capacity of PM may be an effective strategy to overcome PM resistance to oxaliplatin.

Fig. 1. Molecular analysis of colorectal peritoneal metastases reveals multiple features of aggressive tumour behaviour.

a RNA sequencing of PM (n = 59) and paired primary tumour regions (n = 35) from 12 individual patients was performed. The t-SNE plot shows the clustering of most samples per patient. b The same t-SNE plot colour coded by site (primary versus PM). c Gene set enrichment analyses of all Hallmark signatures (n = 50) and immune cell gene sets (Bindea and Cell-ID). All gene sets that were significantly enriched in either PM or primary tumour samples are shown and ordered by the significance and direction of enrichment. d Examples of box plots of individual gene sets expressed at significantly different levels in either PM or primary tumour samples.

Fig. 2. PM and the primary tumours from which they arise represent an almost uniform CMS4 entity within CRC.

a Application of the CMS classifier on the RNAseq data reveals strong enrichment of CMS4. b Expression of the genes positively identifying CMS4 in the original CMS random forest classifier (CMS4 (RF); 143 genes) in PM and primary tumour samples. c Differential gene expression analysis identified 366 genes that were significantly higher expressed in PM than in paired primary tumour samples. The scatter plot shows the correlation of meta-gene expression values of this gene set in relation to the CMS4-identifying gene set (CMS4 (RF)). d As in (c), but in the original composite CMS cohort (n = 3232 primary tumours). e k-means clustering of the CMS cohort using the 366 PM-enriched gene set shows a correlation between high expression of this ‘PM-signature’ and a significantly shorter recurrence-free survival. f Expression of CMS4-identifying genes from the RF classifier in the k-means groups generated with the PM signature. g Distribution of CMS subtypes in the k-means groups generated with the PM signature.

Fig. 3. PM-derived organoids display relative resistance to oxaliplatin and express high levels of glutamate-cysteine ligase (GCLC).

a Schematic representation of the experimental setup. b Dose–response curves for five non-PM CRC-derived organoids and threePM-derived organoids. The dose range of oxaliplatin was 1, 100, 200, 300, 400 and 500 μM. Three days after a 1 h exposure to oxaliplatin the fraction of remaining viable cells was measured using Cell Titer Glo and plotted as percentage of untreated control cells. c Boxplot showing area under the curve (AUC) values, calculated from the dose–response curves. d The contribution of cancer-associated fibroblasts (CAFs) and N-acetyl-cysteine (NAC) in glutathione synthesis. GCLC, the rate-limiting enzyme in this pathway, is indicated in red. e GCLC mRNA expression levels in non-PM and PM-derived organoids, measured by RNA sequencing. f, g GCLC protein expression levels in non-PM and PM-derived organoids in relation to actin. The boxplot in (f) shows optical density measurements of the exposed films in (g), normalised to actin.

Fig. 4. Depletion of NAC sensitises PM-derived organoids to oxaliplatin.

a–c Drug response curves for a series of six distinct PM-derived organoids treated with oxaliplatin for 72 h in the presence (a) or in the absence of N-acetyl-cysteine (NAC) (b), or treated with oxaliplatin for 1 h in the absence of NAC (c). The range of clinically applied oxaliplatin concentrations in the HIPEC perfusate is indicated by the grey area (215–439 μM). d Boxplot showing AUC values, calculated from the dose–response curves in (a–c). e Boxplot showing surviving cancer cell fractions calculated from the dose–response curves in (a–c). f Schematic representation of the experimental setup of the long-term regeneration assays. g Dotplot showing a quantification of long-term re-growth potential of five distinct PMDOs after a 1-h or 72-h exposure to oxaliplatin (20 μM). h Boxplot showing the aggregated data from (g).

Fig. 5. GCLC knockout increases sensitivity to oxaliplatin.

a Organoids derived from control (ECAS) and two independent GCLC-knockout clones (KO1 and KO2) were seeded in an NAC-depleted growth medium with or without reduced glutathione (GSH, 2 mM) as indicated (n = 3 per condition). Cells were treated with oxaliplatin for 72 h at the indicated concentrations (μM), and cell viability was then measured using Titer-Glo assay. The bar graphs show the percentage of surviving cell fractions. All values are plotted as % of untreated controls. b As in (a), with 5-FU or irinotecan, as indicated. (n = 3 per condition). c Control and GCLC-knockout organoids were exposed to oxaliplatin (5 h; 200 μM) in the presence or absence of reduced GSH (2 mM). After extensive washing of the cells, genomic DNA was isolated and the platinum (Pt) content was quantified by HPLC. The barplot shows the Pt content (n = 3) in ng per ug total genomic DNA. d Western blot analysis of pRPA, GCLC and actin protein expression levels in control ECAS and GCLC-knockout organoids following exposure to oxaliplatin (200 μM) at the indicated time points.

Fig. 6. Inhibition of glutathione synthesis sensitises PMDOs to short-term oxaliplatin treatment.

a PMDOs were treated for 24 h with BSO (100 μM) or solvent. Reduced glutathione (GSH) levels were then measured as described in the 'Materials and methods'. The barplot shows reduced GSH levels in control- and BSO-treated PMDOs. b, c Two distinct PMDOs were treated for 1 h with oxaliplatin at the indicated concentrations in the presence or absence of BSO (100 μM). The surviving cell fractions were then quantified by Cell Titer Glo. All values are plotted as % of untreated controls. d Bar graphs showing a quantification of Pt adducts in the genomic DNA of PMDOs following treatment with oxaliplatin (200 μM) with or without BSO (100 μM) and in the presence or absence of reduced GSH. e Western blot analysis of pRPA, GCLC and actin protein expression levels in PMDOs following a 5 or 24 h exposure to oxaliplatin (200 μM) in the presence or absence of BSO. f Experimental setup of the long-term regeneration experiments. g, h Two distinct PDOs were treated with oxaliplatin (200 μM) and BSO (100 μM) alone and in combination for 1 h at 42 °C. Three weeks after drug washout the regenerative capacity was assessed in all conditions by Cell Titer Glo measurements and photographs.