High NRF2 Levels Correlate with Poor Prognosis in Colorectal Cancer Patients and with Sensitivity to the Kinase Inhibitor AT9283 In Vitro

Abstract

Aberrant hyperactivation of nuclear factor erythroid 2 (NF-E2) p45-related factor 2 (NRF2) is a common event in many tumour types and associates with resistance to therapy and poor patient prognosis; however, its relevance in colorectal tumours is not well-established. Measuring the expression of surrogate genes for NRF2 activity in silico, in combination with validation in patients' samples, we show that the NRF2 pathway is upregulated in colorectal tumours and that high levels of nuclear NRF2 correlate with a poor patient prognosis. These results highlight the need to overcome the protection provided by NRF2 and present an opportunity to selectively kill cancer cells with hyperactive NRF2. Exploiting the CRISPR/Cas9 technology, we generated colorectal cancer cell lines with hyperactive NRF2 and used them to perform a drug screen. We identified AT9283, an Aurora kinase inhibitor, for its selectivity towards killing cancer cells with hyperactive NRF2 as a consequence to either genetic or pharmacological activation. Our results show that hyperactivation of NRF2 in colorectal cancer cells might present a vulnerability that could potentially be therapeutically exploited by using the Aurora kinase inhibitor AT9283.

Keywords: NRF2; colorectal cancer; drug screening.

Figure 1

Relevance of the nuclear factor erythroid 2 (NF-E2) p45-related factor 2 (NRF2) pathway in colorectal tumours in vivo. (A) Bioinformatic analysis of the expression of various NRF2 target genes in normal and tumour tissues. The activity of NRF2 in colon (upper row) and rectal (lower row) cancers (T) and normal (N) tissues of the TCGA project was assessed through the expression of the NRF2 target genes NQO1 (left) and through the combined score from the expression of NQO1, GPX2, TXNRD1, GCLC and GCLM (right). p: p-value of the Welch’s t-test. (B) NQO1 expression in matched normal and tumour colorectal tissue quantified using real-time PCR. Figures show the combined data (upper panel) or individual patients (lower panel). The data were normalised using β-actin as an internal control. Data represent means ± SD (n = 9). The differences between the tumour and the normal tissue for NQO1 were statistically significant (paired t-test analysis) (* p = 0.0401). (C) Kaplan-Meier plots showing differential cancer-specific survival for colorectal patients with high or low nuclear NRF2 levels (p = 0.041).

Figure 2

Validation of a new colorectal cancer NRF2-GOF (gain-of-function) in vitro model. (A) Schematic representation of the NRF2 domains and the sequencing on the NRF2-GOF clones. (B) Isogenic NRF2-WT (wild-type) and NRF2-GOF DLD1 cell lines were treated with either DMSO (−) or with 5 µM of sulforaphane (SFN) (+) for 3 h, and the protein levels of NRF2 were compared. (C) The mRNA levels for NQO1 in the indicated cell lines treated with dimethyl sulfoxide (DMSO) or with 5 µM of SFN for 16 h were quantified using real-time PCR. The data were normalised using β-actin as an internal control. Data represent means ± SD (n = 3) and are expressed relative to the WT DMSO cells. ** p ≤ 0.01. (D) Representation of the differential expression of oxidative stress-related genes in NRF2 WT versus NRF2-GOF. Highlighted either in red (upregulated) or in green (downregulated) are genes with more than 2-fold changes; only those with p-values < 0.05 were labelled (n = 3).

Figure 3

Synthetic lethality drug screening. (A) Scatter plots showing the drug sensitivity score (DSS) values of the Aurora kinase inhibitor compounds in the screening panel in DLD1 WT and GOF cells for (upper panel) cell viability (CellTiter-Glo, CTG) and for (lower panel) cell toxicity (CellTox Green, CTX) readouts. Labels show only Aurora kinase inhibitors. (B) Key drug response parameters of AT9283 in both WT and NRF2-GOF DLD1 cells are shown. TC₅₀ (half-maximal toxic concentration), EC₅₀ (half-maximal effective concentration), AUC (area under the curve) and DSS (drug sensitivity score).

Figure 4

Validation of the selectivity of AT9283 against active NRF2. (A) NRF2-WT and GOF DLD1 cells were exposed to increasing concentrations of AT9283, as indicated. After three days, cell viability was measured using Alamar blue. Data represent means ± SD (n = 4) and are expressed relative to the DMSO control, which was set as 100%. (B) Upper panel: DLD1 cells were pretreated with the indicated concentrations of NRF2 inducers or with DMSO, as indicated. After 24 h, fresh media with either DMSO or 80-nM or 200-nM AT9283 was added, and NRF2 inducers were replenished. On the next day, the NRF2 inducers were added again, and 24 h later, the cell viability was measured using Alamar blue (n = 3). The viability of all control cells (without AT9283) were set as 100%. Lower panel: The cell viability of control cells (without AT9283) treated with the different NRF2 inducers is shown. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001.