key: cord-0694382-k01v1yqz
authors: Knatko, Elena V.; Castro, Cecilia; Higgins, Maureen; Zhang, Ying; Honda, Tadashi; Henderson, Colin J.; Wolf, C. Roland; Griffin, Julian L.; Dinkova-Kostova, Albena T.
title: Nrf2 activation does not affect adenoma development in a mouse model of colorectal cancer
date: 2021-09-15
journal: Commun Biol
DOI: 10.1038/s42003-021-02552-w
sha: 61c8d37290a37d3cad0d7fd5c900d72401d269ce
doc_id: 694382
cord_uid: k01v1yqz

Transcription factor nuclear factor erythroid 2 p45-related factor 2 (Nrf2) and its main negative regulator, Kelch-like ECH associated protein 1 (Keap1), are at the interface between redox and intermediary metabolism. Nrf2 activation is protective in models of human disease and has benefits in clinical trials. Consequently, the Keap1/Nrf2 protein complex is a drug target. However, in cancer Nrf2 plays a dual role, raising concerns that Nrf2 activators may promote growth of early neoplasms. To address this concern, we examined the role of Nrf2 in development of colorectal adenomas by employing genetic, pharmacological, and metabolomic approaches. We found that colorectal adenomas that form in Gstp(−/−): Apc(Min/+) mice are characterized by altered one-carbon metabolism and that genetic activation, but not disruption of Nrf2, enhances these metabolic alterations. However, this enhancement is modest compared to the magnitude of metabolic differences between tumor and peri-tumoral tissues, suggesting that the metabolic changes conferred by Nrf2 activation may have little contribution to the early stages of carcinogenesis. Indeed, neither genetic (by Keap1 knockdown) nor pharmacological Nrf2 activation, nor its disruption, affected colorectal adenoma formation in this model. We conclude that pharmacological Nrf2 activation is unlikely to impact the early stages of development of colorectal cancer.

T o adapt to conditions of oxidative, electrophilic, and inflammatory stress, cells have evolved networks of cytoprotective proteins, the gene expression of which is controlled by transcription factor nuclear factor erythroid 2 p45related factor 2 (Nrf2) and its main negative regulator the Kelchlike ECH associated protein 1 (Keap1) 1 . Activation of Nrf2 has shown protective effects in numerous animal models of chronic disease 2 and has beneficial effects in human clinical trials 3 . Nrf2 is considered a drug target, and several small molecule Nrf2 inducers are currently at various stages of drug development 4 . Among them, the naturally occurring isothiocyanate sulforaphane, the semi-synthetic pentacyclic cyanoenone triterpenoids, and their tricyclic derivatives are the most potent Nrf2 inducers known to date [5] [6] [7] [8] . The cyanoenones activate Nrf2 at low nanomolar concentrations by modifying C151 in Keap1 9, 10 . The tricyclic cyanoenone TBE-31 is highly bioavailable 11 , has a covalent and reversible mode of action, and is suitable for chronic oral administration 12 . Furthermore, TBE-31 has demonstrated protective effects in mouse models of cutaneous squamous cell carcinoma 13 , and non-alcoholic steatohepatitis 14 and in a rat model of aflatoxin-induced hepatocarcinogenesis 6 .

Although Nrf2 is primarily regulated at the protein stability level, the expression of NFE2L2, the gene encoding Nrf2, is another important determinant of cellular Nrf2 levels, which has been shown to affect susceptibility to disease. Thus, a single nucleotide polymorphism (SNP) in the human NFE2L2 promoter lowers its gene expression and increases the risk for lung cancer 15 . However, persistent activation of Nrf2 is frequently exploited by cancer cells, where it promotes their survival and resistance to chemotherapy and radiation therapy 16, 17 , and contributes to metabolic adaptation 18 . Mutations in Keap1 or Nrf2, which abrogate formation of the Keap1/Nrf2 protein complex or prevent Nrf2 ubiquitination, leading to its constitutive activation, occur in several types of human cancer, and are particularly prominent in squamous cell carcinomas of the lung, contributing to tumor growth and resistance to chemoand radiation therapy 19, 20 . Importantly, it was recently shown in the context of non-small cell lung cancer that constitutively active Nrf2 generates enhancers at gene loci that are not normally regulated following transient activation of Nrf2 under physiological conditions 21 . Nrf2 is also upregulated in human colorectal tumors, where high levels of nuclear Nrf2 correlate with poor patient prognosis 22 . However, whether Nrf2 affects the early stages of adenoma development in the colon is unclear.

It is well established that in human colorectal cancers, the adenomatous polyposis coli (APC) gene is frequently mutated 23 or inactivated by promoter hypermethylation 24 , and its germline mutations cause familial adenomatous polyposis (FAP), an autosomal dominant inherited condition in which numerous adenomas form in the epithelium of the large intestine 25 . Similar to humans, Apc Min/+ mice that have a mutant allele encoding a nonsense mutation at codon 850 of the murine Apc gene, are predisposed to intestinal adenoma formation, but unlike in humans, the tumors in mice form predominantly in the small intestine 26 . Interestingly, Apc Min/ + mice that are also deficient for glutathione transferase Pi (Gstp −/− : Apc Min/+ ) have a 6-fold increase in colon adenoma incidence, and a 50-fold increase in colorectal adenoma multiplicity, in comparison with their Gstp wild-type (Gstp +/+ : Apc Min/+ ) counterparts 27 . Using the Gstp −/− : Apc Min/+ mouse model, in this study we examined the role of Nrf2 in the development of colorectal adenomas by employing genetic and pharmacological approaches. We then used metabolomics to characterize tumorous and non-tumorous tissue from mice with different levels of Nrf2.

The expression of classical Nrf2 targets and pro-inflammatory genes is not affected by deletion of Gstp. Because Gstp is involved in xenobiotic metabolism 28 as well as cell signaling, by sequestering c-Jun N-terminal kinase (JNK) 29 , we first asked whether deletion of Gstp affects the Nrf2-mediated transcription using the classical Nrf2-target NAD(P)H:quinone oxidoreductase 1 (Nqo1) as a marker, which is expressed in an Nrf2-dependent manner in the mouse colon 30 . The mRNA levels for Nqo1 in the colon tissues did not differ between Gstp +/+ and Gstp −/− animals (Fig. 1a) . As expected, these levels were~60% lower in mice with disrupted Nrf2 (Gstp −/− : Nrf2 −/− ) and~4-fold higher in mice with high Nrf2 levels due to Keap1 downregulation (Gstp −/− : Keap1 flox/flox ) (Fig. 1a) .

The GSTs are a large family of detoxification enzymes, and we next asked whether deletion of Gstp affected the expression of other members of the GST family. The mRNA levels in the colon for Gsta1/2 (Fig. 1b) , Gsta3 (Fig. 1c) , and Gstm (Fig. 1d ) did not differ significantly between Gstp −/− : Nrf2 +/+ : Keap1 +/+ and Gstp +/+ : Nrf2 +/+ : Keap1 +/+ mice. As expected based on published work comparing their expression in the intestine of wild-type and Nrf2-knockout mice 31 , the expression levels of Gsta3 and Gstm were lower in Nrf2 −/− and higher in Keap1 flox/ flox mice, in comparison with their Nrf2 +/+ and Keap1 +/+ counterparts (Fig. 1c, d) . By contrast, the mRNA levels for the pro-inflammatory cytokines interleukin-6 (Il6) (Fig. 1e) and Il1β (Fig. 1f) were lower in Keap1 flox/flox mice. The basal levels of expression of prostaglandin-endoperoxide synthase 2 (Ptgs2), more commonly known as cyclooxygenase 2 (Cox2), were not affected by Gstp deletion, Nrf2 disruption or Keap1 knockdown (Fig. 1g) .

Genetic activation or disruption of Nrf2 does not affect colorectal adenoma formation in Gstp −/− : Apc Min/+ mice. To address whether genetic activation or disruption of Nrf2 affects the development of colorectal tumors, we generated Gstp −/− : Apc Min/+ : Nrf2 −/− : Keap1 +/+ and Gstp −/− : Apc Min/+ : Nrf2 +/+ : Keap1 flox/flox mice by crossing Gstp −/− : Apc Min/+ mice with Nrf2 −/− or Keap1 flox/flox mice, respectively, all on the C57BL/6 genetic background. For simplicity, throughout the rest of the manuscript, we refer to these genotypes of mice as follows: Gstp −/− : Apc Min/+ : Nrf2 +/+ : Keap1 +/+ = wild-type (WT); Gstp −/− : Apc Min/+ : Nrf2 −/− : Keap1 +/+ = Nrf2knockout (Nrf2-KO); and Gstp −/− : Apc Min/+ : Nrf2 +/+ : Keap1 flox/flox mice = Keap1-knockdown (Keap1-KD). The number and volume of colon neoplasms (>1 mm height) were assessed at 20 weeks (when due to welfare concerns, many animals had to be euthanized) in order to establish whether genetic activation or disruption of Nrf2 affects the development of colorectal adenoma. We found that the multiplicity of tumors was not affected by either disruption of Nrf2 or its genetic upregulation by Keap1 knockdown (Fig. 2a, b) . Interestingly, the multiplicity of large (>2.5 mm diameter) tumors tended to be lower in female mice than in male mice, again with no significant differences among the genotypes (Fig. 2c, d) . Overall tumor burden was also similar among the genotypes, and greater in male than female animals (Fig. 2e, f) .

Nrf2 activation does not affect the expression of proinflammatory cytokines in tumor or peri-tumoral tissue. In a subset of mice, we determined the expression of the proinflammatory cytokines Il6 and Il1β, and the pro-inflammatory enzyme Ptgs2. The mRNA levels for both cytokines were higher in tumor compared to peri-tumoral tissue, had high inter-tumor variability, and did not differ significantly among the genotypes (Fig. 3a, b) . The expression of Ptgs2 was increased in the tumors (Fig. 3c) , although this increase was smaller in tumors from Keap1-KD mice in comparison with WT or Nrf2-KO, consistent with our previous observations of a reciprocal relation between pharmacological Nrf2 activation and interferon-γ (IFNγ)-stimulated Ptgs2 transcription 5 . The expression of the antiinflammatory cytokine Il10 in tumor tissue was not affected by either Nrf2 disruption or genetic activation (Fig. 3d) . Notably, the mRNA levels for Nqo1 in tumor and peri-tumoral tissues in WT (i.e. Gstp −/− : Apc Min/+ : Nrf2 +/+ : Keap1 +/+ ) mice were highly variable among 12 randomly selected individual animals (Fig. 3e) , suggesting no consistent changes in Nrf2-transcriptional activity during adenoma development. Although lower than in WT, as expected based on its regulation by Nrf2, the Nqo1 expression did not differ between tumor and peri-tumoral tissues of Nrf2-KO mice. Similarly, the mRNA levels for Gclc and Gclm, two other Nrf2-transcriptional targets encoding the catalytic and the modifier subunits, respectively, of glutamate-cysteine ligase, the enzyme catalyzing the rate-limiting step in the biosynthesis of glutathione (GSH), were similar in tumor compared to peritumoral tissue, although both were higher in tissues from Keap1-KD compared to WT mice (Fig. 3f, g) , as expected 31 . Interestingly, however, the mRNA (Fig. 3h ) and the protein (Fig. 3i ) levels for heme oxygenase 1 (Hmox1), which is also partly regulated by Nrf2, were profoundly increased in tumor vs. peritumoral tissues. The protein levels of Hmox1 were lowest in Nrf2-KO and highest in Keap1-KD peri-tumoral tissues, as expected. By contrast, the increase in Hmox1 levels in the tumor tissues was higher in Nrf2-KO than in WT mice, indicating that transcription factors other than Nrf2 were responsible for it. Considering that the expression of Hmox1 is also regulated by the proinflammatory transcription factor nuclear factor κB (NFκB) and activator protein 1 (AP1) 32 , and that Hmox1 is known to be upregulated in inflamed colonic tissues 33 , we speculate that the enhanced Hmox1 expression in the tumors is at least in part due to the markedly enhanced inflammation in tumor compared to peri-tumoral tissue (Fig. 3a-c) .

Colorectal adenomas that form in Gstp −/− : Apc Min/+ mice are characterized by altered one-carbon metabolism. By use of metabolomics, combining nuclear magnetic resonance (NMR) spectroscopy and two complementary liquid chromatography mass spectrometry (LC-MS) approaches, we examined if we could identify a metabolic signature induced by the tumor, irrespectively of the genotype. Multivariate analysis of both the NMR spectroscopy ( Fig. 4a, b ) and the LC-MS data ( Fig. 5a-d) showed a clear separation between tumors and peri-tumoral tissues, indicating distinct metabolic states, with no apparent subgrouping related to the genotype. The NMR spectroscopy analysis showed significantly higher levels of betaine, AMP, succinate, alanine, glutamate, glutamine, glycine, aspartate, and malate ( Fig. 6a -c and Supplementary Fig. 1 ) in the tumors compared to peri-tumoral tissues, and lower levels of myo-inositol, creatine and inosine (Fig. 6d , e). The increase in betaine ( Fig. 6a ) was particularly intriguing, due to its role in methylation. The LC-MS analyses confirmed the higher AMP levels in tumors, and further showed increases in the ribonucleotides GMP, CMP and UMP (Supplementary Fig. 1c and Supplementary Fig. 2 ). This analysis also confirmed the increased glutamate and glycine levels, and additionally showed that the levels of proline, hydoxyproline and CDP-choline were also increased ( Supplementary Fig. 1d, e) . Furthermore, in addition to confirming higher levels of betaine ( Supplementary Fig. 1f ), this analysis revealed increases in choline, S-adenosyl methionine (SAM), S-adenosyl homocysteine (SAH) ( Supplementary Fig. 1f , g), and cystathionine (Supplementary Fig. 1e ). Together, these changes indicate that onecarbon metabolism is altered in the colorectal adenomas from Gstp −/− : Apc Min/+ mice.

Knockdown of Keap1, but not disruption of Nrf2, enhances the metabolic alterations in colorectal adenomas from Gstp −/− : Apc Min/+ mice. We next focused our analysis on differences among the genotypes within the same type of tissue. Both the NMR spectroscopy and the LC-MS analyses showed that the levels of glutathione were higher in both tumors and peri-tumoral tissues from Keap1-KD mice, in comparison with their WT and Nrf2-KO counterparts (Fig. 6g, h) , in close agreement with the higher expression levels of Gclc and Gclm (Fig. 3f, g) . Glutathione levels tend towards an increase, but this increase is not statistically significant between peri-tumor and tumor samples (p = 0.059), if the genotype factor is ignored. However, plotting its levels in all three genotypes in both types of tissue shows the different behavior in each case. Significantly different amounts of glutathione were found in both the peri-tumoral tissues (ANOVA p = 1.37E−8) and the tumors (ANOVA p = 5.42E−7), with a particular increase in the Keap1-KD genotype in both cases. This plot also shows that Nrf2-KO is the only genotype where glutathione increases in the tumors.

In general, the differences in metabolites between WT and Keap1-KD were greater than the differences between WT and Nrf2-KO genotypes in both tumor and peri-tumoral tissues. Thus, compared to WT, the levels of myo-inositol (Fig. 6d) , betaine (Fig. 6a) , AMP (Fig. 6b) , succinate (Fig. 6c) , creatine (Fig. 6e) and NAD (Fig. 7b) were higher in tumor tissue of cystathione were increased in the tumors, and although still increased in tumor compared to peri-tumoral tissue, were lower in tumors from Keap1-KD and higher in tumors from Nrf2-KO mice (Fig. 7a) . Finally, symmetric dimethylarginine (SDMA) (Fig. 7g) , asymmetric dimethylarginine (ADMA) (Fig. 7h) and phenylalanine ( Fig. 7i) were markedly increased in tumor compared to peri-tumoral tissue, but these increases were much less pronounced in tumors from either Keap1-KD or Nrf2-KO mice.

Overall, most differences in metabolites between tumor and peri-tumoral tissues were enhanced in Keap1-KD mice; however, the magnitude of this enhancement was much more modest in comparison with the magnitude of the differences between tumor and peri-tumoral tissues. This observation is further supported by the fact that the mRNA levels for glucose-6-phosphate dehydrogenase (G6pdx and G6pd2) (Fig. 8a , b) and the pentose phosphate pathway enzyme phosphogluconate dehydrogenase (Pgd) (Fig. 8c) , which increase upon Nrf2 activation in rapidly 4 Tumor and peri-tumoral tissues have distinct metabolic states as revealed by NMR spectroscopy analysis. a Principal component analysis (PCA) score plot from NMR data for total fatty acids (R 2 = 75.9% and Q 2 = 41.7%). At least seven independent biological replicates of colon tissue from peritumorous (red) and tumorous (black) tissues were included. b Plot showing the individual metabolites driving the separation among the tissue types (loadings), with the variables (i.e., metabolites) represented in green circles (Asp = aspartate and Asn = asparagine).

proliferating cells 18 , were not different among the Nrf2/Keap1 genotypes. Interestingly, the expression of G6pdx was increased in tumor in comparison with peri-tumoral tissues, whereas that of G6pd2 was decreased, and the mRNA levels for Pgd were unchanged. Together, these findings suggest that in this model, the enhanced metabolic changes conferred by the Keap1 knockdown may have negligible contribution to the carcinogenesis process, a conclusion also supported by the fact that neither Keap1 knockdown nor disruption of Nrf2 affected colorectal adenoma formation in Gstp −/− : Apc Min/+ mice (Fig. 2) .

Chronic pharmacological activation of Nrf2 by TBE-31 does not affect colorectal adenoma formation in Gstp −/− : Apc Min/+ mice. We used the potent tricyclic cyanoenone TBE-31 ( Fig. 9a) to activate Nrf2 pharmacologically. The mRNA levels for Nqo1 (Fig. 9b) , were increased by oral administration of TBE-31, confirming Nrf2 activation in the colons of Gstp −/− : Apc Min/+ mice.

In full agreement with the previous experiment (Fig. 2) , there was no significant difference in tumor multiplicity (Fig. 9c, d) or burden (Fig. 9e ) between WT and Nrf2-KO mice, confirming that the absence of functional Nrf2 neither accelerates nor inhibits tumor development in this model. Consistent with the lack of effect of genetic Nrf2 activation by Keap1 knockdown, chronic intervention with TBE-31 (5 nmol/g body weight, orally, 3 days per week for 12 weeks, starting at week 8 of age) did not affect tumor development in Gstp +−/+− : Apc Min/+ mice.

Genetic activation or disruption of Nrf2 does not affect colorectal adenoma formation in Gstp −/− : Apc Min/+ mice fed high-fat diet. The combination of high-fat diet and Apc truncation has been shown to alter the bile acid profile, antagonizing the function of intestinal farnesoid X receptor (FXR) to drive malignant transformation in cancer stem cells 34 . Thus, we asked if genetic alterations in Nrf2 may affect tumor development in mice fed highfat diet, which is increasingly more common worldwide. To this end, starting at week 6 of age, the animals were fed diet delivering 60% of calories from fat for 9 weeks, and tumor development was assessed at week 15. Although Keap1 is the main negative regulator of Nrf2, the activity of the transcription factor in cancer is also affected by several other proteins 35 . Similarly, in addition to Nrf2, Keap1 regulates the function of other cellular proteins 36 . Thus, in addition to the three genotypes (i.e. WT, Nrf2-KO and Keap1-KD), for this experiment we also generated a fourth category of mice combining Nrf2 deficiency and Keap1 downregulation, namely Gstp −/− : Apc Min/+ : Nrf2 −/− : Keap1 flox/flox , to which we refer as Nrf2-KO/Keap1-KD double-transgenic mice. Consistent with the carcinogenesis experiments described above, all mice were Gstp −/− : Apc Min/+ . Once again, we found that tumor multiplicity tended to be higher in male than in female mice, and that there were no significant differences in the multiplicity (Fig. 10a-d) or burden (Fig. 10e, f) of colorectal tumors among any of the four genotypes, in close agreement with the previous experiments in mice fed standard diet (Figs. 2, 9c-e) . This experiment further confirmed the conclusion that disruption of Nrf2 or its constitutive upregulation by either genetic or pharmacological means, does not affect Fig. 5 Tumor and peri-tumoral tissues have distinct metabolic states as revealed by LC-MS analysis. a, b Partial least square-discriminant analysis (PLS-DA) score plot from LC-MS data for aqueous metabolites, analyzed by: a C18pfp-based method (R 2 X = 71.9%, R 2 Y = 94% and Q 2 = 89.3%) (a); and a behamide-based method (R 2 X = 55.8%, R 2 Y = 80.1% and Q 2 = 74.7%) (b). In both cases, at least seven independent biological replicates of colon tissue from peri-tumorous (red) and tumorous (black) tissues were included. c, d Plots showing the individual metabolites driving the separation among the tissue types (loadings), for the C18pfp-based method (c) and the behamide-based method (d), with the variables (i.e., metabolites) represented in green circles, while the center for each genotype (peri-tumorous-DA(1), tumorous-DA (2)) is in blue.

colorectal adenoma development in this mouse model even under conditions of high-fat diet.

Notably, it was necessary to terminate the high-fat diet experiment earlier than the experiments on standard diet, when the animals were 15-rather than 20-week old, and we had to enroll many more mice as a number of them had to be sacrificed before the 15-week endpoint. This necessity was related to concerns for animal welfare, the close monitoring of which showed that many of the mice on high-fat diet developed anemia earlier than those on standard diet. Indeed, it has been shown that C57BL/6 mice fed high-fat diet for 8 weeks develop iron deficiency due to diminished intestinal iron uptake 37 . Curiously, we found that anemia was more prevalent in Keap1-KD animals than in any of the other three genotypes (p < 0.001 for Keap1-KD vs. WT; p = 0.012 for Keap1-KD vs. Nrf2-KO; p = 0.005 for Keap1-KD vs. Nrf2-KO/Keap1-KD). Thus, the percentage of anemic mice within each genotype was: WT: 4.8% (3/62; 3 males and 0 females), Nrf2-KO: 11.7% (7/60; 5 males and 2 females), Keap1-KD: 27.4% (17/62; 9 males and 8 females), and Nrf2-KO/ Keap1-KD: 9.8% (6/61; 4 males and 2 females). We hypothesized that the increased anemia in the Keap1-KD mice was at least in part caused by the documented Nrf2-dependent transcription of bone morphogenetic protein 6 (Bmp6) in liver sinusoidal endothelial cells 38 . This is because Bmp6 is known to regulate the synthesis of the liver hormone hepcidin, reducing serum iron levels 39 . Therefore, we measured the hepatic mRNA levels for Bmp6, and found them to be higher in livers of Keap1-KD mice compared to their WT or Nrf2-KO counterparts (Fig. 10g) . Consistent with Nrf2 dependence, the hepatic expression levels of Bmp6 did not differ from WT in Keap1-KD mice with disrupted Nrf2 (i.e. Nrf2-KO/Keap1-KD double transgenic) mice. We attribute the small magnitude of hepatic Bmp6 increase in Keap1-KD mice to the fact that the liver sinusoidal endothelial cells represent~15% of the total liver cell population, and only 3% of the liver volume 40 . Alternatively, it is possible that other transcription factors attenuate the Nrf2-mediated increase in Bmp6 transcription, such as HIF1α, which is upregulated by iron deficiency 41 , and has been shown to downregulate the expression 

Over the years since its discovery, Nrf2 has been attributed both tumor-suppressing and tumor-promoting effects, and has been linked to the hallmarks of cancer 44 . Activation of Nrf2 is commonly observed in established human tumors 45 , raising concerns over the current drug development and clinical use of pharmacological Nrf2 activators. In this study, we addressed this concern by use of a mouse model of colorectal cancer initiated by a truncation of the Apc gene. Inactivating mutations of both alleles of the tumor suppressor APC represent one of the earliest events in the colorectal carcinogenesis in humans 46 . Our results show that neither genetic (by Keap1 knockdown) nor pharmacological (by a potent activator, the cyanoenone TBE-31) Nrf2 activation affects adenoma development at the early stages. However, Nrf2 is upregulated in human colorectal tumors and correlates with poor patient prognosis 22 , and thus it is probable that Nrf2 activation occurs later in the carcinogenesis process, such as during transition from adenoma to adenocarcinoma, supporting tumor growth. Further studies are needed to examine this possibility, which would require the use of a different animal model. One restriction of the model used in our study is that, in keeping with animal welfare regulations, it is not possible to maintain the experimental mice beyond the adenoma stage of the tumors, which limits its clinical relevance to the early stages of the carcinogenesis process.

TBE-31 is a cyanoenone, which shares the same electrophilic centers with bardoxolone methyl (CDDO-Me) and RTA-408, two pentacyclic cyanoenone triterpenoids, which are currently in advanced clinical trials for a number of disease indications, including Alport syndrome, chronic kidney disease associated with type 2 diabetes mellitus, liver disease, Friedreich's ataxia 4, 47 , and in patients hospitalized with confirmed COVID-19 (NCT04494646). The concern that pharmacological Nrf2 activation may promote the development of early neoplastic lesions is particularly relevant to colorectal cancer, because it is the third most common cancer globally in both men and women, with high mortality rates 48 , and because orally-administered pharmacological Nrf2 activators are in direct contact with the gastrointestinal tract epithelium. The findings from our study suggest that it is unlikely that pharmacological Nrf2 activators that are being developed for clinical use will promote the initiation of colorectal cancer.

Using LC-MS analysis of metabolites in colon tissue extracts, we recently found that the levels of glucose 6-phosphate and fructose 6-phosphate, metabolites involved in the initial steps of glycolysis, were higher in colons of Keap1-KD mice in comparison with their WT counterparts, whereas dihydroxyacetone phosphate and glyceraldehyde 3-phosphate, which are involved in the later steps of glycolysis, were lower 30 . Additionally, gas chromatography-MS (GC-MS) analysis showed that the levels of saturated and mono-unsaturated fatty acids were lower in Keap1-KD mice, whereas the levels of polyunsaturated fatty acids were higher. Together, these findings indicate that knockdown of Keap1, and the consequent activation of Nrf2, affect multiple aspects of intermediary metabolism in the healthy colon. Here, we show that the differences in metabolites between tumor and peritumoral colon tissues of Gstp −/− : Apc Min/+ mice are enhanced by Keap1 knockdown. Nonetheless, the magnitude of this enhancement is modest compared to the magnitude of the difference in the metabolic state between tumor and peri-tumoral tissues. These observations, together with the finding that neither genetic Keap1 knockdown nor its pharmacological inactivation by TBE-31 affects adenoma formation, lead us to conclude that in this model, the enhanced metabolic changes conferred by the Keap1 knockdown do not substantially contribute to the early stages of the colorectal carcinogenesis process.

It is noteworthy that tumor multiplicity, particularly that of large (>2.5 mm diameter) tumors, and burden are consistently higher in male than in female mice in all of our experiments. This finding is in agreement with the increased prevalence of colorectal cancer in males in human populations 49 and the enhanced adenoma multiplicity in the colon of Apc Min/+ mice (that are wildtype for Gstp) and rats carrying a nonsense mutation in the Apc gene, termed Polyposis in the rat colon (Apc Pirc/+ ), as well as in rat models of colorectal carcinogenesis induced by the chemical carcinogens 1,2-dimethylhydrazine 50 or azoxymethane 51 . Detailed studies by Amos-Landgraf et al 51 . have found that neither ovariectomy nor hormone replacement affected the prevalence of adenomas in female Apc Pirc/+ rats, whereas orchidectomy reduced adenoma development in their male counterparts, and supplementation with testosterone reversed this effect. Thus, as in the above discussed colorectal cancer models, the observed sexual dimorphism in multiplicity of colonic tumors in our model is more likely due to tumor promotion by male hormones rather than protection by female hormones 51 .

Cheung et al. 27 have reported increased inflammation and intestinal carcinogenesis in Gstp +/+ : Apc Min/+ : Nrf2 −/− : Keap1 +/+ mice 52 . In our model, unbiased pathway enrichment analysis following mRNA profiling, as well as additional supportive biochemical analyses, revealed that the absence of Gstp promotes a state of enhanced inflammation, accompanied by a 6-fold increase in colon adenoma incidence, and a 50-fold increase in colorectal adenoma multiplicity. The present study confirmed the profound increase in pro-inflammatory markers (Il6, Il1β, and Ptgs2) in tumor compared to peri-tumoral tissue in mice of all genotypes (Fig. 3a-c) . Importantly however, Nrf2 disruption did not increase inflammation further (Fig. 3a-c) . Together, these studies illustrate the critical importance of inflammation in colorectal adenoma formation, consistent with the tumor-suppressive effects of anti-inflammatory drugs, such as celecoxib 53, 54 and aspirin 55, 56 in Apc Min/+ mice, and the epidemiological data showing association of regular aspirin use with reduced incidence of colorectal cancer in humans 57 .

Of note, quantitative proteomic analysis of livers from Gstp −/− mice, which were used to obtain the Gstp −/− : Apc Min/+ mice, showed similar overall protein S-glutathionylation profiles to those of their Gstp +/+ counterparts, indicating that whereas Gstp may be involved in catalyzing S-glutathionylation of some proteins in response to oxidative stress, including Keap1 58 , its role in global S-glutathionylation is limited 59 . Nonetheless, the absence of Gstp is an important aspect of our mouse models because Sglutathionylation of Keap1 may lead to Nrf2 activation 58 .

An interesting corollary of this study is the observation that a number of metabolites involved in one-carbon metabolism (comprising the folate and methionine cycles) are altered in the colorectal adenomas from Gstp −/− : Apc Min/+ mice. The enhanced effect of the Keap1 knockdown is consistent with the importance of Keap1/Nrf2 for mitochondrial health 60, 61 , which in turn supports one-carbon metabolism 62 . One-carbon metabolism and SAM have critical roles in multiple cellular processes, including generation of precursors to nucleotide biosynthesis and one-carbon units for DNA-and histone-methylation reactions, and can drive tumorigenesis 63 . Crucially, because all tumors in this model are adenomas 27 , our findings strongly suggest that alterations in one-carbon metabolism occur early in the colorectal carcinogenesis process, and together with changes in expression/ activity of the enzymes that catalyze their formation, these metabolic alterations allow for the well-documented aberrations in DNA and histone methylation in colorectal tumors from mice and humans 48 . Curiously, Khor et al. 64 have reported promoter hypermethylation of NFE2L2, the gene encoding Nrf2, and a trend for a decrease in the intensity and percentage of Nrf2positive cells in advanced-stage prostate cancer compared to normal prostate tissue. Conversely, Hanada et al. 65 have shown hypermethylation of the Keap1 promoter, leading to decreased expression and increased nuclear Nrf2 and downstream Nrf2target gene expression in colorectal cancer. Thus, it is possible that the observed increase in one-carbon metabolism in our mouse models may additionally affect Keap1 and/or Nrf2 expression via epigenetic mechanisms.

Another interesting observation from this study is the increased prevalence of anemia in Keap1-KD animals fed high-fat diet than in any of the other three genotypes. The fact that this phenotype is more common in Keap1-KD than in Nrf2-KO/ Keap1-KD double-transgenic mice suggests that it is related to Nrf2 activation rather than any other process that might be influenced by the downregulation of Keap1. We found that the mRNA levels for Bmp6, which controls the synthesis of the liver hormone hepcidin that in turn lowers serum iron levels 39 , are higher in livers of Keap1-KD, but importantly not in Nrf2-KO/ Keap1-KD mice, in comparison with their WT or Nrf2-KO counterparts (Fig. 10g) , consistent with Nrf2-dependent upregulation of hepcidin in mice fed high-fat plus iron 66 or polyphenolrich 67 diets. In the small intestine, hepcidin is the predominant negative regulator of iron absorption by enterocytes 68 . As mentioned in Results, additionally and independently of the levels of hepcidin, high-fat diet feeding causes iron deficiency due to diminished intestinal iron uptake 37 . Thus, the increased hepcidin levels may exacerbate the diminished intestinal iron uptake due to the high-fat diet by limiting dietary iron absorption by enterocytes, ultimately causing anemia. Colorectal carcinogenesis. A scoring system was developed to monitor animal welfare and signs of intestinal neoplasia, including occult blood in faeces, rectal bleeding, stool consistency, anemia (by pale feet), rectal prolapse and weight loss. Mice were euthanized at defined time-points indicated in the text and figures, or when the scoring reached a set value. The large intestine was removed, flushed with PBS, placed on ice, cut longitudinally, the tumors counted, and their height, length, and width measured. Tumor volume was calculated according to a formula for the volume of a sphere, using the average of the three dimensions as the diameter. The tumors and peri-tumor tissue were then excised, snap-frozen in liquid N 2 and stored at −80°C till further analysis.

Real-time quantitative PCR. Total RNA was extracted from mouse colon and tumor tissue using RNeasy Kit (Qiagen Ltd.), and 500 ng of total RNA was reversetranscribed into cDNA using Omniscript RT Kit (Qiagen Ltd.). Real-time PCR was carried out on Applied Biosystems QuantStudio™ 5 Real-Time PCR System. The TaqMan data for the mRNA species were normalized using mouse ribosomal protein lateral stalk subunit P0 (Rplp0), hypoxanthine phosphoribosyltransferase 1 (Hprt1) and actin-beta (Actb) as internal controls. The TaqMan TM Gene Expression Assay IDs (Thermo) used were: Mm00607939_s1 (Actb); Mm01332882_m1 (Bmp6); Mm0051 6005_m1 (Hmox1); Mm03024075_m1 (Hprt1); Mm00658204_s1 (G6pd2); Mm042 60097_m1 (G6pdx); Mm00802655_m1 (Gclc); Mm00514996_m1 (Gclm); Mm0420 7463_m1 (GstA1/2); Mm00494798_m1 (GstA3); Mm00833915_g1 (GstM1); Mm0043 4228_m1 (Il1b); Mm00446190_m1 (Il6); Mm00439614_m1 (Il10); Mm01253561_m1 (Nqo1); Mm00503037_m1 (Pgd); Mm00478374_m1 (Ptgs2); Mm00725448_s1 (Rplp0).

Immunoblotting. Frozen tissues (15-40 µg) were disrupted by coarse grinding in liquid N 2 , followed by homogenization for 40 sec in 10 volumes (µl per mg tissue) ice-cold RIPA buffer (50 mM Tris-HCl, pH 8.0; 150 mM sodium chloride; 1.0% NP-40; 0.5% sodium deoxycholate; 0.1% sodium dodecyl sulfate) supplemented with 1.4 × complete EDTA-free protease inhibitors cocktail (Roche). Homogenates were clarified by centrifugation for 10 min at 17,000 × g at 4°C, protein concentrations in the supernatants were determined by the bicinchoninic acid (BCA) assay (Thermo) and adjusted to the same protein concentration in all samples using RIPA buffer. Samples were then mixed with NuPAGE LDS sample buffer (Thermo) and NuPAGE sample reducing agent (Thermo), heated for 10 min at 70°C, and loaded at 20 µg/lane onto pre-cast gradient (4-12%) Tris-Glycine gels (Thermo). Proteins were resolved by electrophoresis and transferred onto 0.45-µm nitrocellulose membranes (Amersham Biosciences) at 100 V for 45 min at 4°C. Membranes were blocked in 5% milk in PBST (0.1% Tween-20) for 45 min, on a rocker (60-70 rpm) , at room temperature, and then incubated with the primary antibodies at 4°C on a rocker overnight. Dilutions of all primary and secondary antibodies were in 5% milk in PBST (0.1% Tween-20). The following antibodies were used: rabbit polyclonal anti-Hmox1, 1:2,000, Abcam; mouse monoclonal antiβ-actin, 1:5,000. The uncropped/unedited scans of the immunoblots are shown in Supplementary Fig. 3 .

Metabolomics. Metabolites were extracted using the methanol/chloroform/water (2:2:1; v/v) method 70, 71 . Briefly, 50 mg of wet weight tissue were mixed with 600 µl of CH 3 OH/CHCl 3 (2:1; v/v), and the samples were homogenized with a Tissuelyser (Qiagen, UK) for 5 min at a frequency of 20/s and sonicated for 15 min. Water and chloroform (200 μl of each) were added to the samples before centrifugation at 13,300 rpm for 20 min. The resulting aqueous and organic phases were separated from the protein pellets. The extraction procedure was repeated on the remaining protein pellets as part of a double extraction procedure. Both organic and aqueous phases were collected and evaporated to dryness. The dried samples were stored at −80°C until further analysis.

NMR analysis of aqueous extracts. The dried aqueous fractions were rehydrated in 600 μl D 2 O, containing 0.05 mM sodium-3-(tri-methylsilyl)−2,2,3,3-tetradeuteriopropionate (TSP) (Cambridge Isotope Laboratories, MA, USA) as an internal standard. The samples were analyzed using an AVANCE II + NMR spectrometer operating at 500.13 MHz for the 1 H frequency and 125.721 MHz for the 13 C frequency (Bruker, Germany) using a 5 mm TXI probe. The instrument was equipped with TopSpin 3.2. Spectra were collected using a solvent suppression pulse sequence based on a one-dimensional nuclear Overhauser effect spectroscopy (NOESY) pulse sequence to saturate the residual 1 H water signal (relaxation delay = 2 s, t1 increment = 3 µs, mixing time = 150 ms, solvent presaturation applied during the relaxation time and the mixing time). One hundred and twenty-eight transients were collected into 16 K data points over a spectral width of 12 ppm at 27°C. Assignment of the peaks was done with reference to published literature and databases and the Chenomx spectral database contained in Chenomx NMR Suite 7.7 (Chenomx, Alberta, Canada).

LC-MS analysis of aqueous metabolites. Half of the extracted aqueous samples were reconstituted in 0.1 ml 7:3 acetonitrile: 0.1 M aqueous ammonium carbonate containing a mixture of 8 internal standards at the concentration of 10 µM (Proline, Valine D8, Leucine D10, Lysine U13, Glutamic acid C13, Phenyl alanine D5, Succinic acid D3, Serotonin D4) (all from Sigma Aldrich except the glutamic acid from Cambridge Isotope Laboratories, MA, USA). The samples were vortexed then sonicated for 15 min followed by centrifugation at 21,000 × g to pellet any remaining undissolved material. They were analyzed on a Quantiva triple stage quadrupole mass spectrometer coupled to a Vanquish Horizon Ultra High Performance Chromatography (UHPLC) unit (all analytical instrument combinations supplied by Thermo Fisher Scientific), using a bridged ethylene hybrid (BEH) amide hydrophilic interaction liquid chromatography (HILIC) column 72 . The strong mobile phase (A) was 100 mM ammonium carbonate, the weak mobile phase was acetonitrile (B) with 1:1 water:acetonitrile being used for the needle wash. The LC column used was the BEH amide column (150 × 2.1 mm, 1.7 μm, Waters). The following linear gradient was used: 20% A in acetonitrile for 1.5 min followed by an increase to 60% A over 2.5 min with a further 1 min at 60% A after which the column was re-equilibrated for 1.9 min. After each chromatographic run the column was washed with 30 column volumes of 6:4 water:acetonitrile followed by a further 10 column volumes of 95:5 acetonitrile:water for storage. The total run time was 7 min, the flow rate was 0.6 mL/min and the injection volume was 5 μL. In order to resolve pentose phosphates for the identification of ribose-1phosphate a shallower gradient was employed: 30% A in acetonitrile for 2.0 min followed by an increase to 50% A over 3.0 min with re-equilibration for 1.9 min.

The second half of the aqueous fraction was reconstituted in 0.1 mL of a 10 mM ammonium acetate water solution containing the same mixture of 8 internal standards at the concentration of 10 µM. The samples were vortexed then sonicated for 15 min followed by centrifugation at 21,000 × g to pellet any remaining undissolved material. They were analyzed with an ACE Excel 2 C18 PFP (100 A. 150 × 2.1 mm 5 µm) column. The electrospray voltage was set to 3500 V for the positive ionization and to 2500 V for the negative ionization. Nitrogen at 48 mTorr and 420°C was used as a drying gas for solvent evaporation. The column was conditioned at 30°C. The mobile phase consisted of: (A) a 0.1% of formic acid water solution and (B) a 0.1% of formic acid acetonitrile solution. The mobile phase was pumped at a flow rate of 500 µL/min programmed as follows: initially 100% of A for 1.60 min, then subjected to a linear decrease from 100% to 70% of A in 2.4 min and to 10% in 0.5 min then constant for 0.5 min and brought back to initial condition after 0.1 min.

NMR and LC-MS data processing. NMR spectra were automatically processed in TopSpin v 3.1 (Bruker, Germany). The integrals of the different metabolites were obtained using Chenomx, normalizing to TSP and weight of the sample. LC-MS chromatograms were analyzed using Xcalibur, version 2.0 (Thermo Fisher), integrating each peak individually. while LC-MS peaks were normalized to the internal standard and weight of the sample.

Statistics and reproducibility. For multivariate statistical analysis of metabolite profiles, datasets were imported into SIMCA-P 15.0 (Sartorius AG, Gottingen, Germany) for processing using PCA and PLS-DA (a regression extension of PCA used for supervised classification). 1 H NMR data were Pareto scaled, in which each variable was centered and multiplied by 1/(Sk)1/2 where Sk is the standard deviation of the variable. LC-MS data were scaled to unit variance by dividing each variable by 1/(S k ). Univariate statistical analyses were performed using Excel (Microsoft). Values are expressed as mean ± S.D. and the significance level was set at p < 0.05. For three groups comparison, one-way ANOVA and Tukey post-hoc test were used.*0.05 > p > 0.01; **0.01 > p > 0.001; ***p > 0.001. Sample sizes were determined on the basis of published experiments using similar methodologies and are stated in the figure legends. The experimental animals were randomly assigned to treatment groups. For all experiments, the stated replicates are biological replicates. For experimental carcinogenesis outcomes and metabolite analyses, samples were processed in random order and analyzed blinded to genotype, treatment and tissue type.

The KEAP1-NRF2 system: a thiol-based sensor-effector apparatus for maintaining redox homeostasis

Synthetic oleanane triterpenoids: multifunctional drugs with a broad range of applications for prevention and treatment of chronic disease

Rapid and sustainable detoxication of airborne pollutants by broccoli sprout beverage: results of a randomized clinical trial in China

Therapeutic targeting of the NRF2 and KEAP1 partnership in chronic diseases

Extremely potent triterpenoid inducers of the phase 2 response: correlations of protection against oxidant and inflammatory stress

A novel acetylenic tricyclic bis-(cyano enone) potently induces phase 2 cytoprotective pathways and blocks liver carcinogenesis induced by aflatoxin

Tricyclic compounds containing nonenolizable cyano enones. A novel class of highly potent anti-inflammatory and cytoprotective agents

International Natural Product Sciences, T. & Supuran, C. T. Natural products in drug discovery: advances and opportunities

C151 in KEAP1 is the main cysteine sensor for the cyanoenone class of NRF2 activators, irrespective of molecular size or shape

KEAP1, a cysteine-based sensor and a drug target for the prevention and treatment of chronic disease

An exceptionally potent inducer of cytoprotective enzymes: elucidation of the structural features that determine inducer potency and reactivity with Keap1

Pharmacokinetics and pharmacodynamics of orally administered acetylenic tricyclic bis(cyanoenone), a highly potent Nrf2 activator with a reversible covalent mode of action

Nrf2 activation protects against solar-simulated ultraviolet radiation in mice and humans

Experimental nonalcoholic steatohepatitis and liver fibrosis are ameliorated by pharmacologic activation of Nrf2 (NF-E2 p45-Related Factor 2)

Regulatory nexus of synthesis and degradation deciphers cellular Nrf2 expression levels

Involvement of Nrf2 activation in resistance to 5-fluorouracil in human colon cancer HT-29 cells

Role of KEAP1/NRF2 and TP53 mutations in lung squamous cell carcinoma development and radiation resistance

Nrf2 redirects glucose and glutamine into anabolic pathways in metabolic reprogramming

Proteomic analysis of ubiquitin ligase KEAP1 reveals associated proteins that inhibit NRF2 ubiquitination

A catalogue of somatic NRF2 gain-of-function mutations in cancer

Enhancer remodeling promotes tumor-initiating activity in NRF2-activated non-small cell lung cancers

High NRF2 levels correlate with poor prognosis in colorectal cancer patients and with sensitivity to the kinase inhibitor AT9283 in vitro

Mutational analysis of the APC/beta-catenin/Tcf pathway in colorectal cancer

Analysis of adenomatous polyposis coli promoter hypermethylation in human cancer

Molecular diagnosis of familial adenomatous polyposis

ApcMin: a mouse model for intestinal and mammary tumorigenesis

Markedly enhanced colon tumorigenesis in Apc(Min) mice lacking glutathione S-transferase Pi

Glutathione transferases: substrates, inihibitors and pro-drugs in cancer and neurodegenerative diseases

Regulatory functions of glutathione S-transferase P1-1 unrelated to detoxification

Downregulation of Keap1 confers features of a fasted metabolic state

The Cap'n'Collar basic leucine zipper transcription factor Nrf2 (NF-E2 p45-related factor 2) controls both constitutive and inducible expression of intestinal detoxification and glutathione biosynthetic enzymes

Regulation of inflammation by the antioxidant haem oxygenase 1

Analysis of intestinal haem-oxygenase-1 (HO-1) in clinical and experimental colitis

FXR regulates intestinal cancer stem cell proliferation

NRF2 activation in cancer: from DNA to protein

Beyond repression of Nrf2: An update on Keap1. Free Radic

High-fat diet causes iron deficiency via hepcidinindependent reduction of duodenal iron absorption

Nrf2 controls iron homeostasis in haemochromatosis and thalassaemia via Bmp6 and hepcidin

BMP6 is a key endogenous regulator of hepcidin expression and iron metabolism

Liver sinusoidal endothelial cells (LSECs) function and NAFLD; NO-based therapy targeted to the liver

Iron depletion by deferoxamine up-regulates glucose uptake and insulin signaling in hepatoma cells and in rat liver

HIF-1alpha-mediated BMP6 down-regulation leads to hyperproliferation and abnormal differentiation of keratinocytes in vitro

HIF-1 induction attenuates Nrf2-dependent IL-8 expression in human endothelial cells

NRF2 and the hallmarks of cancer

NRF2 and the ambiguous consequences of its activation during initiation and the subsequent stages of tumourigenesis

APC mutations occur early during colorectal tumorigenesis

Can Activation of NRF2 Be a Strategy COVID-19?

Epigenetics of colorectal cancer: biomarker and therapeutic potential

Ethnic and sex disparities in colorectal neoplasia among Hispanic patients undergoing screening colonoscopy

Influence of gonadal hormones and age on 1,2-dimethylhydrazine-induced colon carcinogenesis

Sex disparity in colonic adenomagenesis involves promotion by male hormones, not protection by female hormones

Nrf2 knockout enhances intestinal tumorigenesis in Apc(min/+) mice due to attenuation of anti-oxidative stress pathway while potentiates inflammation

The cyclooxygenase-2 inhibitor celecoxib is a potent preventive and therapeutic agent in the min mouse model of adenomatous polyposis

Chemoprevention of colon cancer by specific cyclooxygenase-2 inhibitor, celecoxib, administered during different stages of carcinogenesis

Suppression of intestinal and mammary neoplasia by lifetime administration of aspirin in Apc(Min/+) and Apc(Min/+), Msh2(−/−) mice

Aspirin prevents tumors in a murine model of familial adenomatous polyposis

Effects of regular aspirin on long-term cancer incidence and metastasis: a systematic comparison of evidence from observational studies versus randomised trials

S-Glutathionylation of Keap1: a new role for glutathione S-transferase pi in neuronal protection

Proteome-wide identification and quantification of S-glutathionylation targets in mouse liver

The emerging role of Nrf2 in mitochondrial function. Free Radic

Regulation of Nrf2 by mitochondrial reactive oxygen species in physiology and pathology

Mitochondrial dysfunction remodels one-carbon metabolism in human cells

glycine and one-carbon units: cancer metabolism in full circle

Epigenetic DNA methylation of antioxidative stress regulator NRF2 in human prostate cancer

Methylation of the KEAP1 gene promoter region in human colorectal cancer

Dysregulated expression of fatty acid oxidation enzymes and iron-regulatory genes in livers of Nrf2-null mice

Phytoestrogens modulate hepcidin expression by Nrf2: Implications for dietary control of iron absorption. Free Radic

Hepcidin, a key regulator of iron metabolism and mediator of anemia of inflammation

An improved synthesis of a hydroxymethyl tricyclic ketone from cyclohexanone, the key process for the synthesis of a highly potent antiinflammatory and cytoprotective agent

Comprehensive metabolic profiling of age-related mitochondrial dysfunction in the high-fat-fed ob/ob mouse heart

A targeted metabolomics assay for cardiac metabolism and demonstration using a mouse model of dilated cardiomyopathy

FAMIN is a multifunctional purine enzyme enabling the purine nucleotide cycle

Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

The metabolomics data have been deposited in MetaboLights, and are associated with the identifier MTBLS2961. All other data are available from the corresponding author upon reasonable request.

The authors are most grateful to Cancer Research UK (C20953/A18644), Stony Brook Foundation, and Reata Pharmaceuticals for funding.

Author contributions E.K., A.D.K., C.C., C.H., C.W. and J.G. designed the research and analyzed the data. A.D.K., E.K., C.C. and J.G. wrote the manuscript. E.K., M.H. and Y.Z. performed the genotyping, inducer treatment, and carcinogenesis experiments. C.C. and J.G. performed the metabolomics experiments. T.H. synthesized the pharmacological Nrf2 inducer.

The authors declare no competing interests.

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s42003-021-02552-w.Correspondence and requests for materials should be addressed to Albena T. Dinkova-Kostova.Peer review information Communications Biology thanks Federico Pio Fabrizio, and the other, anonymous, reviewers for their contribution to the peer review of this work. Primary Handling Editors: Eve Rogers.Reprints and permission information is available at http://www.nature.com/reprintsPublisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.