Development of Targeted Therapy of NRF2^high Esophageal Squamous Cell Carcinoma

Abstract

Esophageal squamous cell carcinoma (ESCC) is a deadly disease and one of the most aggressive cancers of the gastrointestinal tract. As a master transcription factor regulating the stress response, NRF2 is often mutated and becomes hyperactive, and thus causes chemo-radioresistance and poor survival in human ESCC. There is a great need to develop NRF2 inhibitors for targeted therapy of NRF2^high ESCC. In this review, we mainly focus on three aspects, NRF2 inhibitors and their mechanisms of action, screening novel drug targets, and evaluation of NRF2 activity in the esophagus. A research strategy has been proposed to develop NRF2 inhibitors using human ESCC cells and mouse models.

Introduction

Esophageal cancer is expected to affect 19,260 adults and cause 15,530 deaths in the US in 2021 ¹. In the world, it is the seventh most prevalent cancer and the sixth leading cause of cancer-related death, with more than 604,100 new cases and 544,076 deaths in 2020². Two main histological types of esophageal cancer exist, esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma. Human ESCC develops from precancerous lesions, and its histopathology follows a step-wise pattern of hyperplasia, dysplasia, and squamous cell carcinoma (SCC). The 5-year survival rate for ESCC is ~18%, a number that reflects late diagnosis, the aggressiveness of the disease, and a lack of effective treatment strategies^{3, 4}. Esophageal adenocarcinoma has become the predominant type of esophageal cancer in Western countries, and gastroesophageal reflux disease and obesity are the main risk factors. Barrett’s esophagus is a precancerous lesion that may further develop into dysplasia and adenocarcinoma⁵. Similar to ESCC, adenocarcinoma has a very poor prognosis. Thus, there is a great need to further elucidate the molecular mechanisms and develop more effective treatment strategies for esophageal cancer.

Nuclear factor (erythroid-derived 2)-like 2 (NRF2 or NFE2L2) mutations are commonly seen in ESCC, with frequencies between 5% and 30% ⁶. As a major cellular defense mechanism, the NRF2 signaling pathway regulates the expression of enzymes involved in detoxification and anti-oxidative stress response. NRF2 forms heterodimers with small MAF proteins and binds to the antioxidant response elements (ARE) of target genes when cells are exposed to oxidative stress or xenobiotics. Kelch-like ECH-associated protein 1 (KEAP1) inhibits the function of NRF2 by retaining NRF2 in the cytoplasm and facilitating its ubiquitination-dependent degradation under normal physiological conditions. Oxidative stress and electrophilic modification of KEAP1 results in NRF2 nuclear translocation and ARE-dependent gene transcription (Figure 1) ⁷. In addition to the KEAP1-CUL3 E3 ubiquitin ligase complex, NRF2 can also be ubiquitinylated by β-TrCP-CUL1 or HRD1 for proteasomal degradation ^{8, 9}.

Figure 1. NRF2 signaling pathway.

Under basal conditions, NRF2 is bound to KEAP1, which is a CUL3-based E3 ubiquitin ligase adapter that regulates NRF2 ubiquitination and proteasomal degradation. When cells are exposed to oxidative or electrophilic stress, KEAP1 changes its conformation and releases NRF2. Then NRF2 translocates into the nucleus, forms a heterodimer with its obligatory partner (small MAF proteins), binds to the ARE, and activates the transcription of downstream genes, such as antioxidative genes and phase II detoxification enzymes.

The NRF2 signaling pathway is a double-edged sword in the context of carcinogenesis. Chemical or genetic activation of NRF2 induces cytoprotective enzymes conferring protection against chemical carcinogenesis in multiple models. Nrf2^−/− mice are more susceptible to chemical carcinogenesis than wild-type counterparts^{10, 11}. On the other hand, cancer cells can hijack the NRF2 signaling pathway for their survival through mechanisms that lead to constitutive activation of NRF2 signalings, such as somatic mutations of KEAP1/NRF2/CUL3, accumulation of disruptor proteins, skipping of NRF2 exon 2, KEAP1 succinylation, KEAP1 hypermethylation, increased NRF2 expression, and electrophilic attack of KEAP1 by oncometabolites. NRF2 hyperactivation promotes cell proliferation and metabolic reprogramming, accelerates distant metastases, and confers chemo- and radio-resistance^{12, 13}. NRF2 and KEAP1 are classified as high-confidence cancer driver genes¹⁴. Although not yet realized, the NRF2 signaling pathway is regarded as a tractable molecular target for cancer therapy¹⁵. We have previously reviewed the functional role of NRF2 in ESCC ¹⁶.

Using gene microarray data of human ESCC (GEO23400; n=53)¹⁷, we performed clustering analysis with a list of esophageal NRF2 target genes¹⁸ and identified two subtypes, NRF2^high cases (n=17) and NRF2^low cases (n=36)¹⁹. Similarly, human ESCC can be clustered into NRF2^Mut and NRF2^WT cases according to DNA mutations. NRF2^Mut ESCC was associated with a significantly worse prognosis than NRF2^WT ESCC²⁰. These data suggest subtyping human ESCC according to the NRF2 status is essential for targeted therapy, and potent NRF2 inhibitors are highly desirable.

The esophagus is a unique organ site for studies on NRF2 hyperactivation in cancer. Genetic activation of NRF2 in Keap1^−/− mice results in robust phenotypes in the esophagus, esophageal hyperplasia and hyperkeratosis²¹. To date, all transcriptionally impacted genes downstream of Keap1 knockout are NRF2 responsive, although it remains possible that other KEAP1 substrates could alter transcription by an NRF2-independent mechanism. Overexpression of KEAP1 substrates (e.g., WTX, PALB2, SQSTM1, DPP3, CDK20) containing competitive binding motifs (e.g., ETGE, DLG) results in NRF2 displacement and subsequent NRF2 activation^22–29. However, in the esophagus, NRF2 is the most important KEAP1 substrate as the esophageal phenotype of Keap1^−/− mice can be completely rescued in Nrf2^−/−;Keap1^−/− and K5Cre;Nrf2^fl/fl;Keap1^−/− mice^{21, 30}.

It should be noted that in addition to ESCC, genetic mutations and copy-number alterations of the NRF2 signaling pathway are commonly seen in human cancers, most notably non-small cell lung cancer, uterine carcinoma, head and neck cancer, and bladder carcinoma³¹. Thus, mechanistic studies on and therapeutic development for NRF2^high ESCC will also likely benefit patients of these cancers.

NRF2 inhibitors and their mechanisms of action

At least 5 strategies have been proposed to target the NRF2 signaling pathway for cancer therapy: (1) transcriptional downregulation of NRF2; (2) increased degradation of NRF2 mRNA or decreased translation; (3) enhancement of NRF2 degradation through up-regulation/activation of E3 ubiquitin ligase complexes for NRF2; (4) blocking the dimerization of NRF2 with small MAF proteins; and (5) blocking the NRF2-sMAF DNA-binding domain⁷. Many small-molecule compounds with NRF2-inhibitory activities have been reported in the literature (Table 1). However, some NRF2-inhibitory compounds, i.e., 4-methoxychalcone, apigenin, ascorbic acid, BET bromodomain inhibitor JQ1, chrysin, cryptotanshinone, epigallocatechin 3-gallate, luteolin, metformin, N-N-dimethylformamide, trichostatin A, triptolide, valproic acid, and wogonin, in fact, activate NRF2 expression and/or activity in some other experimental settings. These compounds have been well-reviewed in the literature and may have limited potential for further drug development^{15, 32, 33}. For example, apigenin significantly sensitized doxorubicin-resistant BEL-7402 cells to doxorubicin by dramatically reducing NRF2 expression at both the mRNA and protein levels through downregulation of PI3K/Akt pathway³⁴. On the contrary, apigenin activated NRF2 nuclear translocation, nuclear NRF2-ARE binding activity, ARE-dependent luciferase activity, and expression NRF2 target genes in rat primary hepatocytes and human hepatoma HepG2 cells^{35, 36}.

Table 1.

NRF2 inhibitors

Compound	Mechanisms of Action	Note
(E)-3-(3,5-dimethoxyphenyl)-1-(2-methoxyphenyl)prop-2-en-1-one90	Unknown
1-(4-(tert-Butyl)benzyl)-3-(4-chlorophenyl)-N-hydroxy-1H pyrazole-5-carboxamide91	Unknown
3′,4′,5′,5,7-pentamethoxy flavone92	Upregulates KEAP1 expression and inhibits ERK
Ailanthone93, 94	Post-translational mechanisms, i.e., increase of KEAP1 and decrease of UCHL1 deubiquitinase
All-trans-retinoic acid95–97	Inhibits NRF2-ARE binding through a direct interaction between NRF2 and RARα
AEM137	Unknown	High-throughput screen
Bexarotene96, 98	RXRα activation which interacts with the Neh7 domain of NRF2 and antagonizes ARE-dependent mRNA expression	RXRα-specific ligand
Brusatol99–101	Global translation inhibitor
Camptothecin102	May inhibit NRF2 transcription, translation and/or promoting mRNA degradation	DNA topoisomerase I inhibitor
Clobetasol propionate38,103	Binds to glucocorticoid receptor and recruits a corepressor to suppress ARE-dependent transcription; Prevents NRF2 nuclear translocation; Promotes NRF2 degradation by promoting β-TrCP-dependent degradation in a glucocorticoid receptor and GSK3-dependent manner	High-throughput screen
Corvallatoxin (digoxigenin, cymarin)104	Activates GSK3β and thus promotes NRF2 degradation	Na+/K+-ATPase inhibitor
Costunolide105	Inhibits telomerase
Halofuginone106	Inhibits prolyl-tRNA synthetase and thus global translation
HER2 antibodies (Trastuzumab, Pertuzumab)107	NRF2 promoter hypermethylation
Homoharringtomine108	Stabilizes GC-rich sequence in 5’-UTR of NRF2 DNA
HSP90 inhibitors (17-AAG, 17-DMAG, IPI-504)80	Metabolism by NRF2 target genes (NQO1 and TXNRD1) into more potent HSP90 inhibitors
IM3829 (4-(2-Cyclohexylethoxy)aniline)109	Inhibits NRF2 nuclear translocation
Isoniazid110, 111, Ethionamide112	Activates SIRT1; Inhibits NRF2 nuclear translocation by inhibiting ERK1 phosphorylation	Anti-tubercular drug
K67 and its derivatives 113, 114	Inhibits KEAP1-phospho-p62 protein-protein interaction and thus promotes NRF2 degradation
Malabaricone-A 115	Unknown
Mitomycin C 81	Bioactivation by NRF2 target genes (cytochrome P450 reductase, NQO1 and PPP enzymes) causes DNA damage
ML38539	Binds to the Neh1 domain of NRF2 and thus interferes with the DNA binding activity of NRF2-MAFG complex	High-throughput screen
Ochratoxin A116–118	Inhibits NRF2 nuclear translocation, NRF2-ARE binding, and histone acetylation, and increases miR-132
Sorafenib 119	Unknown	Inhibitor of multiple kinases (VEGFR, PDGFR and RAF)
Tetrahydrocarbazoles120	May stabilize KEAP1-NRF2 interaction
Trigonelline121, 122	Inhibits NRF2 nuclear translocation
Vorinostat123	Downregulates c-Myc, increases KEAP1 expression, and inhibits NRF2 nuclear translocation
• PHA-767491 • AZ-628 • SL-32741	Inhibits NRF2 nuclear translocation	• Cdc7/CDK9 inhibitor • RAF inhibitor • MEK inhibitor
• Stattic • Grassypeptolide A • Cardiac glycosides (lantoside C, strophanthidin, peruvoside, proscillaridin, ouabain) • Emetin and anisomycin • Actin-disrupting agents (lyngbyabellin A, dolastatin 12)38, 40	• STAT3 inhibitor • May inhibit DPP8 and DPP4 and thus promote NRF2 degradation • May inhibit Na+,K+-ATPase pump and thus decrease KEAP1 phosphorylation124, or through HIF pathway • Protein synthesis inhibitor • Possibly interferes actin polymerization and NRF2 nuclear translocation	High-throughput screen
• Antimetabolites (e.g., methotrexate)38		High-throughput screen

High-throughput screening of chemical compound libraries containing natural products, synthetic compounds, and FDA-approved drugs is a popular approach to identify candidate compounds or compound classes with NRF2-inhibitory activities^37–41. Apart from small molecule inhibitors, siRNA is also promising for targeting NRF2^{42, 43}. However, siRNA formulation and delivery into esophageal epithelial cells in vivo are major hurdles in drug development. Following the screen, compounds are subject to validation of their NRF2-inhibitory activities in cultured cells and animal models. However, in vitro screen is subject to limitations. Variations in culture conditions (e.g., seeding density, cell cycle, growth status, medium, confluence, cell line divergence from the original sources), affect drug response in the cells⁴⁴.

While the NRF2-inhibitory activity of many compounds can be validated using ESCC cell lines in vitro, they have rarely been validated in proper animal models. Organ-specific animal models are needed to determine whether these compounds are effective at nontoxic doses to suppress ESCC. In order to develop NRF2 inhibitors for NRF2^high ESCC, CRISPR-Cas9 was used to create a novel mouse carrying the Nrf2^E79Q mutation within the endogenous Nfe2l2 locus, which is the most commonly observed activating mutation in human cancer and is known to activate NRF2⁴⁵. This mouse line allows conditional esophagus-specific activation of NRF2 when crossed with K14Cre line or Sox2CreER line⁴⁶. Four weeks after tamoxifen exposure, Sox2CreER;LSL-Nrf2^E79Q/+ mice developed strong NRF2^high-driven esophageal phenotype in the esophagus, similar to Keap1^−/− mice (data not shown). Combination of Nrf2 mutation and esophagus-specific carcinogen exposure (e.g., 4-nitroquinoline 1-oxide, N-nitrosomethylbenzylamine) is expected to generate NRF2^high ESCC in mice for testing the cancer therapeutic efficacy of NRF2 inhibitors.

To our best knowledge, none of the NRF2 inhibitors have entered clinical trials for targeted therapy. It remains necessary to identify more novel NRF2 inhibitors and clarify their mechanisms of action. Such compounds must be potent in the target cells with acceptable toxicity profiles and act through proper mechanisms. Halofuginone and brusatol inhibit NRF2 through inhibition of global protein synthesis, thus decreasing enthusiasm for further clinical development.

Drug targets in NRF2^high ESCC

Historically transcription factors are viewed as “undruggable”. This is mainly because of the challenges associated with targeting either the protein–DNA or protein–protein interactions that mediate their functions, as opposed to more tractable active sites of enzymes or receptors ⁴⁷. While multiple approaches have been developed to target transcription factors, some other genes may represent viable therapeutic targets if they prove functionally critical. For example, NR0B1 was identified as a selectively expressed protein in NRF2^high lung cancer and small molecules that disrupted NR0B1 protein complexes inhibited NRF2-dependent lung cancer growth⁴⁸. To find additional drug targets for NRF2^high ESCC, we downloaded the publicly available omics data of human cancer cell lines among which 22 human ESCC cell lines were included (Table 2; Supplementary Material). Using the RNAseq data and a list of differentially expressed genes in NRF2^high ESCC¹⁹, we performed clustering analysis and principal component analysis to subtype ESCC cell lines into two clusters, NRF2^high and NRF2^low (Figure 2A, B). Among the NRF2^high ESCC cell lines, KYSE70 (NFE2L2^W24C), KYSE180 (NFE2L2^D77V;KEAP1^P278Q), KYSE520 (NFE2L2^T80I), OE21 (NFE2L2^G81S/D318H), TE6 (NFE2L2^F71_D77del), and TE11 (NFE2L2^D29G) are NRF2^Mut cells, and KYSE510, TE9, and TT are NRF2^WT cells. KYSE510 has an NRF2^high status probably due to PIK3CA^E545K mutation. TE9 has been reported to express a high level of NRF2⁴⁹.

Table 2.

Multi-omics data of human ESCC cell lines

Data	Spreadsheet in Supplementary Material	Weblink	Reference
RNAseq	CCLE_RNAseq-genes-rpkm_ESCC	https://depmap.org/portal/download/	125
Metabolomics	Metabolomics data	https://portals.broadinstitute.org/ccle/data	126
CRISPR gene dependency screen	Achilles CRISPR Dependency	https://depmap.org/portal/download/	127
RNAi gene dependency screen	Combined RNAi Dependency	https://depmap.org/portal/download/	51
PRISM drug sensitivity screen	PRISM Drug Sensitivity	https://depmap.org/portal/download/	71
CTRP V2 drug sensitivity screen	CTRP V2	https://depmap.org/portal/download/	69
GDSC1 drug sensitivity screen	GDSC1	https://depmap.org/portal/download/	68,70
GDSC2 drug sensitivity screen	GDSC2	https://depmap.org/portal/download/	68,70
Protein-Protein Interaction and drug sensitivity	Original data not available	https://mutanome.lerner.ccf.org	79

Figure 2. Subtyping 22 ESCC cells into two subtypes, NRF2^high and NRF2^low, using RNAseq data.

(A) Clustering analysis; (B) Principal component analysis; (C) Significant dependency on the NFE2L2 gene by NRF2^high ESCC cells as compared to NRF2^low ESCC cells. Student’s t test is used for statistical evaluation of NFE2L2 gene dependency score using the original data from the combined RNAi screen.

To validate the clustering analysis, we compared the metabolomics data of NRF2^high ESCC cells and NRF2^low ESCC cells. Several metabolites (e.g., glutathione oxidized, glutathione reduced, NADP) are significantly higher, and several others (e.g., 6-phosphogluoconate, glycine, aspartate, glutamate) are significantly lower in NRF2^high ESCC cells than NRF2^low ESCC cells. These observations are in agreement with several studies on the functional role of hyperactive NRF2 in cellular metabolomics^{19, 49, 50}. Using the combined RNAi screen data of 20 ESCC cell lines (9 NRF2^high and 11 NRF2^low), we also found that NRF2^high ESCC cells were more dependent on the NFE2L2 gene than NRF2^low ESCC cells (Figure 2C). These data support the clustering result of NRF2^high and NRF2^low ESCC cells.

We then compared gene dependency scores of NRF2^high ESCC cells with those of NRF2^low ESCC cells using data from two screening techniques, RNAi and CRISPR. It is generally believed that these two techniques are complementary especially when an improved bioinformatics tool is used to eliminate the off-target effect of RNAi⁵¹. The precision of the two libraries in detecting essential genes is similar and combining data from both screens improves performance, even though results from these two screens show little correlation, which can be partially explained by the identification of distinct essential biological processes with each technology^{52, 53}. In total, 9 genes (ABL1, ALDH3A1, C1QTNF9B, CASC3, DDX4, EXOSC3, GPAA1, SNAP47, SYT2) were found to have higher dependency (i.e., lower dependency scores) in NRF2^high ESCC cells by both screens. For example, ABL1, as a non-receptor tyrosine kinase, is known to be essential and actionable for lung cancer cell survival⁵⁴. ABL1 promotes NRF2 nuclear localization in kidney cancer cells⁵⁵. ALDH3A1 oxidizes various aldehydes to the corresponding acids. NRF2 is known to regulate ALDH3A1 expression in pancreatic cancer cells⁵⁶. ALDH3A1 is strongly expressed in human ESCC tissue, but barely detectable in the non-malignant esophageal epithelium. Knockdown of ALDH3A1 in ESCC cells suppresses cell viability and clonogenic capacity as well as tumor growth in vivo ⁵⁷. These data suggest that ABL1 and ALDH3A1 are potential drug targets for NRF2^high ESCC.

There is a strong rationale to target kinases in combination with NRF2 inhibition for NRF2^high ESCC. Multiple signaling pathways, for example, PI3K/AKT/mTOR can activate NRF2⁵⁸. AKT can also increase the stability of NRF2 by activating p21 which disrupts the NRF2-KEAP1 interaction, and by inhibiting GSK3β which leads to reduced NRF2 phosphorylation, prevents its nuclear translocation, and promotes its ubiquitination and degradation^{23, 59}. On the other hand, NRF2 overexpression is known to activate PI3K/AKT signaling in melanocytes and hepatocytes^{60, 61}. NRF2 directly regulates mTOR transcription when the PI3K pathway is intact, whereas this function is lost when PI3K is activated⁶². It should be noted that keratinocyte-specific deficiency of Pten resulted in AKT activation and esophageal hyperplasia and hyperkeratosis⁶³, similar to the NRF2^high-driven esophageal phenotype. Recently, the NRF2 and PI3K pathways have been shown to synergize in driving non-small-cell lung cancer which was associated with metabolic reprogramming and altered immune microenvironment⁶⁴. These data suggest that the NRF2 and PI3K pathways play a synergistic role in promoting carcinogenesis through mutually reciprocal positive reinforcement. In fact, in a pan-cancer analysis of TCGA datasets, the strongest co-occurrence of affected pathways was between genomic alterations of the NRF2 and PI3K pathways. These co-occurring alterations appeared most frequently in lung cancer, ESCC, head and neck SCC, and uterine carcinoma⁶⁵.

Synthetic lethality is another potential approach to identify additional drug targets for NRF2^high ESCC^{66, 67}. Using isogenic ESCC cells with varying NRF2 status and high-throughput screening technologies, this approach may open up novel therapeutic opportunities.

Drug candidates for NRF2^high ESCC

In recent years, several large-scale drug sensitivity databases have become publically available. These databases provide a huge information base for further data mining that may lead to discovery of new drug candidates or drug repurposing opportunites^68–71. Using drug sensitivity data from three databases (PRISM, CTRP V2, and GDSC), we compared drug sensitivity scores of NRF2^high ESCC cells in comparison to NRF2^low ESCC cells (Supplementary Material). From the PRISM database, we found NRF2^high ESCC cells are significantly more sensitive to 62 drugs than NRF2^low ESCC cells. These 62 drugs mainly fall into two classes, microtubule inhibitors and kinase inhibitors (especially Aurora A kinase inhibitors). For example, alisertib (an Aurora A kinase inhibitor) is known to induce oxidative stress and inhibit the expression of NRF2 in osteosarcoma cell lines ⁷². Crizotinib (an inhibitor of ALK and c-Met) induces hepatotoxicity to reduce cancer cell viability, by activating oxidative stress responses, stimulating mitochondrial apoptosis and necrosis, accumulating reactive oxygen species, and inhibiting NRF2 signaling⁷³. Apatinib (a VEGFR2 tyrosine kinase inhibitor) promotes oxidative stress-dependent apoptosis, by suppressing glutathione and NRF2 signaling in ovarian cancer cells ⁷⁴.

From the CTRP V2 database, we found 4 drugs are more sensitive for NRF2^high ESCC cells than NRF2^low ESCC cells: fluorouracil, leptomycin B, BRD-K71781559, tanespimycin. From GDSC1 and GDSC2 databases, we found motesanib and navitoclax, respectively. Motesanib (sorafenib) is a multi-kinase inhibitor that selectively inhibits EGFR1, VEGFR2, and VEGFR3, and inhibits tumor growth. A study of sorafenib in the treatment of advanced gastric and gastroesophageal cancers showed that when combined with docetaxel and cisplatin, the regimen was effective in reducing tumor size and increase progression-free survival and overall survival⁷⁵. Lung cancer cells that were sensitive to trametinib plus navitoclax (a BCL-xL/BCL-2 inhibitor) expressed higher levels of NRF2 than did those that were resistant⁷⁶. Navitoclax also induces apoptosis and synergizes with chemotherapy by targeting stemness pathways in esophageal cancer ^{77, 78}.

Furthermore, in a recent study on the pharmacogenomics landscape of protein-protein interaction (PPI)-perturbing mutations, NRF2-KEAP1 PPI was identified as one of the top PPIs among 470 putative PPIs in a pan-cancer analysis of 33 cancer types. Sensitivity to 8 drugs was found higher in cancer cells with mutant NRF2-KEAP1 PPI as compared with those with wild-type NRF2-KEAP1 PPI: 17-AAG, docetaxel, temsirolimus, JNJ-2684165, midostaurin, mitomycin C, ZG-10, and embelin⁷⁹. Interestingly, both 17-AAG and mitomycin C have been reported in the literature to be more potent for NRF2^high cancer cells than NRF2^low cancer cells acting through synthetic lethality^{80, 81}.

Monitoring NRF2 activity in ESCC

When NRF2 inhibitors are used for clinical trials on NRF2^high ESCC, there will be a need to assess the NRF2 activity in the cancer tissue to serve two clinical needs: (1) Diagnosis of NRF2^high ESCC: We currently depend on exome sequencing to detect gene mutations, RNAseq to elucidate mRNA expression profiles, or quantitative multiplex immunohistochemistry to evaluate protein expression, for selecting cancer patients for targeted therapy. Although this tissue-based invasive approach has been widely used and is reasonably successful in clinical practice, they all require invasive biopsy. In addition, more or less they downplay tumor heterogeneity by using one piece of tumor or biopsy tissue to represent the whole tumor. (2) Evaluation of the efficacy of NRF2 inhibitors: The efficacy of cancer-targeted therapy is assessed in the same way as traditional cancer therapy, and clinicians depend on symptoms, biomarkers, and radiology to guide treatment planning. More efficient and non-invasive tools that can reveal molecular changes as well as functional abnormalities are highly needed (Table 3).

Table 3.

Approaches for the evaluation of NRF2 activity in the esophagus

Sample	Assay	Measurement	Pros	Cons
Surgical/biopsy specimens	Whole exome sequencing	NRF2, KEAP1, CUL3 mutations	Accurately reflect the NRF2 status in cancer tissues	Invasive procedure to harvest tissue samples; Potential variations due to tumor heterogeneity; High quality of tissue samples;
	RNAseq or qPCR	NRF2high signature
	Quantitative multiplex IHC	Nuclear localization of NRF2 and overexpression of NRF2 and its target genes
	NanoString	A combination of DNA, mRNA, proteins
Live animal or human patients	PET/CT	Radionuclide avidity	Reflect the functional status of NRF2 in the whole cancer tissues; Non-invasive and convenient for clinical follow-up	Possible lack of specificity for NRF2 status

We have generated an NRF2^high mRNA signature that is responsive to NRF2 inhibition and an NRF2^high protein signature. Algorithms will be generated for the assessment of NRF2 activity in ESCC tissue samples. In addition, hyperactive NRF2 is known to cause metabolic reprogramming in the esophagus through transcriptional regulation of metabolic genes, for example, glucose transporter/enzymes (GLUT1, HK1, HK2) and acetate enzyme (ACSS2)^{19, 82}. NRF2 also positively regulates the expression of a monocarboxylate transporter (MCT1) in mouse skeletal muscle, colonic epithelial cells, and mouse liver^83,84. It has become feasible to use PET/CT imaging tools with radionuclides to evaluate whether an individual’s cancer is NRF2^high before treatment and whether an NRF2 inhibitor successfully hits NRF2 after treatment. For example, ¹⁸F-FDG, is widely used for clinical staging and follow-up of human cancer including ESCC. A multicenter prospective trial showed that ¹⁸F-FDG PET/CT had a sensitivity of 79% and a specificity of 95% for stage IV human ESCC, while early-stage T1 and T2 tumors tended to have minimal or no FDG uptake⁸⁵. ¹⁸F-FDG is transported into the cells via glucose transporters on the membrane (e.g., GLUT1) and metabolized in the cytosol into ¹⁸F-FDG-6P by hexokinases (e.g., HK1, HK2), which are overexpressed in NRF2 hyperactive tissues⁸⁶. In patients with lung SCC, the maximum standardized uptake value (SUVmax) was significantly higher in NQO1^high tumors than NQO1^low tumors (NQO1 is a bona fide NRF2 target)⁸⁷. GLUT1 expression is significantly correlated with SUVmax of ¹⁸F-FDG in human ESCC tissues⁸⁸. ¹¹C-acetate is transported into the cells via monocarboxylate transporters on the membrane (e.g., MCT1), and converted into ¹¹C-acetyl-CoA by acyl-CoA synthetase short-chain family members (e.g., ACSS2). ¹¹C-acetyl-CoA will then be used for de novo lipogenesis, histone acetylation, and the tricarboxylic acid cycle. ¹¹C-acetate PET/CT is primarily used for detecting prostate cancer due to the poor uptake of ¹⁸F-FDG and the proximity of the prostate to the bladder⁸⁹.

We have conducted a preliminary imaging study on wild-type and Sox2CreER;LSL-Nrf2^E79Q mice using ¹⁸F-FDG/¹¹C-acetate PET imaging in combination with contrast-enhanced CT. Both ¹⁸F-FDG uptake and ¹¹C-acetate uptake were significantly elevated in the NRF2^high esophagus as compared to the control. Ex vivo autoradiography of ¹¹C-acetate showed dramatic accumulation in the NRF2^high esophagus (data not shown). Further research is ongoing to develop dual ¹⁸FDG/¹¹C-acetate PET for assessment of NRF2 activity in the esophagus. Although the idea remains exploratory or speculative, metabolic imaging warrants further studies in consideration of the critical role of NRF2 in metabolism.

Conclusion

There is a pressing need for NRF2 inhibitors for targeted therapy of NRF2^high ESCC and other NRF2^high cancers. Promising inhibitors and novel drug targets have been identified and await further validation, in particular, in animal models. NRF2 activity assays with tissue-based methods and imaging tools are highly needed. Further research is expected to develop the inhibitors and elucidate their mechanisms of action for NRF2^high ESCC.

Supplementary Material

1

Comparison of NRF2^high and NRF2^low ESCC cells. This file contains the RNAseq data (22 cell lines), metabolomics data (22 cell lines), combined RNAi gene dependency screen data (cell viability as the readout, 20 cell lines), Achilles CRISPR gene dependency screen data (cell viability as the readout, 16 cell lines), PRISM drug sensitivity data (viability as the readout, 20 cell lines), CTRP V2 drug sensitivity data (cell viability as the readout, 20 cell lines), GDSC1 drug sensitivity data (cell viabilility IC50 and AUC as readouts, 22 cell lines), and GDSC2 drug sensitivity data (cell viabilility IC₅₀ and AUC as readouts, 20 cell lines).

Abbreviations

ARE

antioxidant response element

ESCC

esophageal squamous cell carcinoma

KEAP1

Kelch-like ECH-associated protein 1

NRF2/NFE2L2

nuclear factor erythroid 2-related factor 2

SCC

squamous cell carcinoma