Cellular Nrf2 Levels Determine Cell Fate during Chemical Carcinogenesis in Esophageal Epithelium

Nrf2 is essential for cytoprotection against carcinogens, and through systemic Nrf2 knockout mice, Nrf2-deficient cells were shown to be susceptible to chemical carcinogens and prone to developing cancers. However, the oncogenic potential of Nrf2-deficient epithelial cells surrounded by normal cells in the esophagus could not be assessed by previous models, and the fate of Nrf2-deficient cells in such situations remains elusive.

ABSTRACT

Nrf2 is essential for cytoprotection against carcinogens, and through systemic Nrf2 knockout mice, Nrf2-deficient cells were shown to be susceptible to chemical carcinogens and prone to developing cancers. However, the oncogenic potential of Nrf2-deficient epithelial cells surrounded by normal cells in the esophagus could not be assessed by previous models, and the fate of Nrf2-deficient cells in such situations remains elusive. In this study, therefore, we generated mice that harbor almost equal levels of cells with Nrf2 deleted and those with Nrf2 intact in the basal layer of the esophageal epithelium, utilizing inducible Cre-mediated recombination of Nrf2 alleles in adults through moderate use of tamoxifen. In this mouse model, epithelial cells with Nrf2 deleted were maintained with no obvious decrease or phenotypic changes for 12 weeks under unstressed conditions. Upon exposure to the carcinogen 4-nitroquinoline-1-oxide (4NQO), the cells with Nrf2 deleted accumulated DNA damage and selectively disappeared from the epithelium, so almost all 4NQO-induced tumors originated from cells with Nrf2 intact and not from those with Nrf2 deleted. We propose that cells with Nrf2 deleted do not undergo carcinogenesis due to selective elimination upon exposure to 4NQO, indicating that cellular Nrf2 abundance and the epithelial environment determine the cell fate or oncogenic potential of esophageal epithelial cells in 4NQO-induced carcinogenesis.

INTRODUCTION

Our bodies are continuously exposed to various environmental stresses, which occasionally provoke either activation of oncogenes or inactivation of tumor suppressors, leading to carcinogenesis (1). During the developmental course of preneoplasia, cellular changes at the early and intermediate stages have been considered essential adaptations to the surrounding environment (2). This concept introduces clonal adaptation as a fundamental response to many genotoxic carcinogens. Upon exposure to carcinogens, cells show various types of adaptive responses to their microenvironment. Therefore, cancer development needs to be considered in light of how tumor-initiating cells acquire autonomous cell growth and influence neighboring microenvironments through cellular interactions.

Esophageal carcinoma is one of the most aggressive tumors and the sixth most common cause of cancer-related death worldwide (3). There are no molecularly targeted drugs for esophageal carcinoma, and patient prognosis remains poor despite the development of multimodal therapies (4). To establish new therapies for esophageal carcinoma, researchers must conduct multifaceted studies to clarify the pathogenesis and elucidate the molecular biology underlying cancer development. In particular, the esophagus tends to be exposed to various external materials. Oxidative and electrophilic stresses from foods, tobacco smoking, and alcohol overconsumption have been proven to be critical mediators in the development of esophageal carcinoma (5).

The transcription factor Nrf2 plays pivotal roles in the protective response against oxidative and electrophilic insults (6–8). Under unstressed conditions, Nrf2 is arrested by Keap1 and subjected to protein degradation through the ubiquitin-proteasome pathway (9, 10). Upon exposure to oxidative and electrophilic stresses, cysteine residues of Keap1 are modified, and Nrf2 ubiquitination is halted (11). Nrf2 accumulates in the nucleus and regulates the expression of a set of cytoprotective target genes, such as those coding for NAD(P)H:quinone oxidoreductase 1 (Nqo1) and glutamate-cysteine ligase catalytic subunit (Gclc). These Nrf2 target gene products contribute to cellular protection against oxidative and electrophilic stresses.

Nrf2 knockout mice are sensitive to various carcinogens due to the lack of cytoprotective function (12–15). Nrf2 has also been shown to activate immunocytes near epithelial cells to improve the tumor microenvironment (16–18). We previously found that Nrf2 knockout mice were more susceptible than wild-type mice to 4-nitroquinoline-1-oxide (4NQO), a carcinogen in the upper aerodigestive tract, including the tongue and esophagus (19). In contrast, Keap1 knockdown, Nrf2-constitutively activated mice (20) were resistant to 4NQO-induced carcinogenesis. However, it remains unclear whether chemoprevention in Keap1 knockdown mice is attributable to the induction of Nrf2-dependent cytoprotective enzymes in the esophageal epithelium or modification of the tumor microenvironment by Nrf2 activation.

It has been suggested that both stromal cells and epithelial cells near cancer cells contribute to the formation of tumor microenvironments (21–23). Recent progress has revealed that somatic mutations are found in nontumor regions within the human normal esophagus and that there are several cell clones in the esophageal epithelium that form the esophageal tissue structure as a patchwork (24). These cell clones compete with each other in epithelial tissue and maintain tissue homeostasis (25). Based on these observations, we surmise that esophageal tumors may arise from tumor-initiating cells harboring driver mutations, which are surrounded by clones of noncancer cells with various somatic mutations. We also surmise that the Keap1-Nrf2 pathway is an important candidate that affects the nature of microenvironmental cells and stimulates clones of microenvironmental cells. However, little is known about the effects of loss of Nrf2 function in vivo on the selection of various microenvironmental as well as epithelial cell populations in the process of carcinogenesis or adaptation to chemical carcinogens.

To address these questions, we generated Nrf2 deletion at a moderate level in the esophagus of adult mice. We exploited tamoxifen (Tam)-inducible Nrf2 deletion in the esophagus of Nrf2^flox/flox mice. For this purpose, Keratin5-CreERT2::Nrf2^flox/flox (K5CreERT2-Nrf2^F/F) mice were generated, and these mice were treated with a moderate amount of Tam that could induce recombination in approximately half of the esophageal epithelial cells. This experimental system allowed us to examine clonal interactions of epithelial and microenvironmental cells. To examine the contributions of the Nrf2-deficient epithelial clones in esophageal carcinogenesis, we adopted a 4NQO-induced chemical carcinogenesis model with K5CreERT2-Nrf2^F/F mice. Our results demonstrated that Nrf2-deficient cells (i.e., cells with Nrf2 deleted) in the esophageal epithelium of K5CreERT2-Nrf2^F/F mice were subjected to cell selection during 4NQO carcinogenesis in the mouse esophagus, and cells with Nrf2 intact neighboring cells with Nrf2 deleted developed 4NQO-induced cancers.

RESULTS

Generation of a transgenic mouse line expressing inducible Cre recombinase in basal layers of esophageal epithelium.

To introduce Nrf2-deficient cells into the esophageal epithelium, we generated a transgenic mouse line expressing inducible Cre recombinase fused to a mutated estrogen receptor (CreERT2) under the regulation of the human Keratin5 promoter (K5CreERT2 mice) independent of a mouse line previously reported (26) (Fig. 1A).

FIG 1.

Generation of a transgenic mouse line expressing inducible Cre recombinase in basal layers of esophageal epithelium. (A) Generation of the K5CreERT2 plasmid construct. (B) Construction of K5CreERT2. A transgenic mouse expressing inducible CreERT2 under the regulation of the human K5 promoter (K5CreERT2 mice) was generated and crossed with Rosa26-tdTomato reporter mice. (C) TdTomato expression in squamous epithelial tissues 1 week or 24 weeks after Tam in K5CreERT2::Rosa26-tdTomato mice. A moderate amount of Tam (100 μg/g body weight) was used in this analysis. Tam or corn oil was injected intraperitoneally for 3 consecutive days into K5CreERT2::Rosa26-tdTomato mice at the age of 7 to 9 weeks. Arrowheads point to the junction between the forestomach and hindstomach.

To assess the efficiency and specificity of Cre-mediated recombination by using the new K5CreERT2 mouse line, we crossed the mice with Rosa26-tdTomato reporter mice (Fig. 1B). Tam or vehicle was intraperitoneally injected into the K5CreERT2::Rosa26-tdTomato mice for 3 consecutive days (Tam×3), and tdTomato fluorescence was examined 1 week after the last Tam injection. We detected fluorescence specifically in squamous epithelial cells in the mouse esophagus, tongue, forestomach, and skin (Fig. 1C). These results indicated that in our present system, K5CreERT2-mediated recombination occurs specifically in squamous epithelial cells. To examine whether Cre is expressed in the stem cell compartment of squamous epithelium, we analyzed tdTomato fluorescence 24 weeks after Tam injection. We observed a similar tdTomato fluorescence pattern 24 weeks after Tam injection, indicating that Cre is expressed in the stem cells of squamous epithelium and that the recombinant cells are maintained for at least 24 weeks.

Nrf2-deficient cells were maintained stably for weeks.

To investigate the nature and stability of the Nrf2-deficient cells in the adult esophageal epithelium, we generated inducible and epidermis-specific Nrf2 knockout (K5CreERT2-Nrf2^F/F) mice. We administered Tam once a day for 3 days to the K5CreERT2-Nrf2^F/F and Nrf2^F/F mice at 7 to 9 weeks of age to induce deletion of Nrf2 alleles in the esophageal epithelia and analyzed the mice at 1 and 12 weeks after Tam administration (Fig. 2A). We selected a moderate Tam treatment condition (100 μg/g body weight ×3) to attain Nrf2 recombination in almost half of the esophageal epithelial cells.

FIG 2.

Cells with Nrf2 deleted are maintained stably in the esophageal epithelium for weeks. (A) Experimental schedule after Tam administration. (B) Recombination rate of the Nrf2 gene in the esophageal epithelium. The remaining Nrf2 DNA was quantified after deletion by Tam (n = 4 to 6). (C) mRNA expression levels of Nrf2 and the target gene, Nqo1 (n = 3 to 6). Data represent the mean ± SD. These data were analyzed using Student's t test. *, P < 0.05, and **, P < 0.01, compared with the Nrf2^F/F mice at each time point.

To validate the recombination efficiency of Nrf2 alleles, we analyzed the presence of the loxP locus by quantitative PCR (qPCR) analysis of the genomic DNA in the esophageal epithelium. As expected, the recombination efficiency in K5CreERT2-Nrf2^F/F epithelium was almost 50% 1 week after the last Tam administration, indicating that the cells with Nrf2 deleted and those with Nrf2 intact coexisted almost equally in the epithelium after Tam administration (Fig. 2B). Showing very good agreement with the recombination efficiency of the Nrf2 gene, the mRNA levels of Nrf2 and the Nrf2 target gene Nqo1 were decreased to almost half at 1 week after Tam administration (Fig. 2C).

To examine whether the Nrf2-deficient cells are stably maintained in the esophageal epithelium, we analyzed the recombination efficiency of the Nrf2 gene and the mRNA levels of Nrf2 and Nqo1 at 12 weeks after Tam administration. We found that the recombination efficiency of the Nrf2 gene and the mRNA levels of Nrf2 and Nqo1 at 12 weeks were comparable to those at 1 week, indicating that the Nrf2-deficient cells were maintained stably through cell generation until 12 weeks after Tam administration (Fig. 2B and C).

No disturbance of esophageal tissue formation by the Nrf2-deficient cells was observed under unstressed conditions.

To determine whether Nrf2 deletion gives rise to histological abnormalities in the esophageal epithelium, we examined esophageal sections stained with hematoxylin-eosin (HE). Microscopic inspection of esophageal structures revealed that there was no obvious change between the Nrf2^F/F and K5CreERT2-Nrf2^F/F mice at either 1 or 12 weeks (Fig. 3A). Thicknesses of the keratinous layer (Fig. 3D) and cell layer (Fig. 3E) were comparable between the control Nrf2^F/F and K5CreERT2-Nrf2^F/F mice. As Nrf2 protein is degraded quite rapidly within cells (10, 27), Nrf2 was not detected clearly in either the control or K5CreERT2-Nrf2^F/F mouse esophagus in the immunohistochemical analyses (Fig. 3B). As Nrf2 is degraded quite rapidly by the ubiquitin-proteasome system, it is hard to detect Nrf2 protein at the basal level. For this reason, we examined the Nqo1 protein level as a surrogate marker of Nrf2. In the case of Nqo1 staining, positively stained cells were uniformly observed in the basal cell layer of the control Nrf2^F/F mouse esophagi at both 1 and 12 weeks. In contrast, both Nqo1-positive cell clusters and Nqo1-negative cell clusters were observed in the K5CreERT2-Nrf2^F/F mouse esophagi at 1 week (Fig. 3C). This patchy pattern of Nqo1 expression was maintained at 12 weeks after Tam administration. Altogether, these results indicate that partial loss of Nrf2 did not disturb the tissue formation of the esophagus in unstressed conditions.

FIG 3.

No disturbance of esophageal tissue formation by cells with Nrf2 deleted under unstressed conditions. (A) HE staining of the esophageal sections. These panels show cross-sections of the esophagus. The yellow arrow shows the epithelial cell layer, while the black arrow shows the keratinous layer. (B) Immunohistochemistry of Nrf2. (C) Immunohistochemistry of Nqo1. (D) Thickness of keratinous layers in panel A (n = 5 to 9). (E) Thickness of the cell layers in panel A (n = 5 to 9). The length at 3 points per sample was measured. Data represent the mean ± SD. These data were analyzed using Student's t test. n.s., not significant compared with the Nrf2^F/F mice at each time point.

Nrf2-deficient cells were selectively eliminated after 4NQO treatment.

We next examined whether Nrf2-deficient cells show significant changes in number and phenotype during esophagus-oriented carcinogen 4NQO treatment. We employed a 4NQO-triggered esophageal carcinogenesis model following the standard procedure (28). From 1 week after the Tam-induced Nrf2 deletion, we administered 4NQO through drinking water for 12 weeks (84 days) to the K5CreERT2-Nrf2^F/F and control Nrf2^F/F mice (Fig. 4A). We analyzed the recombination efficiency of the Nrf2 gene in the esophageal epithelium at day 0 (i.e., 1 week after Tam administration [Fig. 4A]) and found that Nrf2 alleles were deleted in 54% of the esophageal epithelial cells of the K5CreERT2-Nrf2^F/F mice (Fig. 4B). To our surprise, we found that upon exposure to 4NQO, the Nrf2 deletion level returned to the control level 84 days after 4NQO administration (Fig. 4B). This finding is in stark contrast to the condition without 4NQO treatment. In fact, in Fig. 2B, we observed that the ratio of the remaining Nrf2 DNA was maintained stably 12 weeks after Tam administration.

FIG 4.

Cells with Nrf2 deleted are selectively eliminated after 4NQO treatment. (A) Administration schedule of Tam and 4NQO. Note that the start day of 4NQO treatment was set as day 0, and the last day of the treatment was day 84. (B) Recombination rate of the Nrf2 gene in the esophageal epithelium. The remaining Nrf2 DNA was quantified after deletion by Tam (n = 4 to 6). (C) Immunohistochemistry of Nqo1. These panels show a cross-section of the esophagus. Insets are enlargements of the box regions. (D) The rate of Nqo1-positive area in esophageal epithelium (n = 5 to 8). Data represent the mean ± SD. These data were analyzed using Student's t test. n.s., not significant, and **, P < 0.01, compared with the Nrf2^F/F mice.

In immunohistochemical analyses at day 0 (i.e., before the 4NQO treatment and 1 week after the Tam×3 treatment), there were Nqo1-positive and Nqo1-negative cell clusters in the K5CreERT2-Nrf2^F/F mouse esophagi (Fig. 4C). Similarly, Nqo1 was very weakly positive in the basal cell layer of the control Nrf2^F/F mouse esophagus.

Notably, Nqo1 expression was induced by 4NQO treatment. At day 10, Nqo1 was clearly expressed in a widespread manner in the cell layers of the control Nrf2^F/F mouse esophagi (Fig. 4C, second panel, top right). This finding might be due to the 4NQO activation of Nrf2 through oxidative stress (29). In contrast, whereas the Nqo1-expressing cells were also induced in the K5CreERT2-Nrf2^F/F mouse esophageal epithelium at day 10, the cells formed clusters and were observed in a patchy manner (Fig. 4C, second panel, top left). Since Nqo1 is tightly regulated by Nrf2 (7), it seems reasonable to assume that these Nqo1-positive cells in the clusters correspond to cells with an intact Nrf2 gene.

Importantly, the cells Nqo1 positive and with Nrf2 intact were expanded, but the cells Nqo1 negative and with Nrf2 deleted decreased at 20 and 50 days after the start of 4NQO treatment (Fig. 4C, lower two panels, and Fig. 4D). In this analysis, we measured the length of the Nqo1-positive cell clusters and compared it with the entire length of the esophageal epithelium, which was set to 100%. This observation was consistent with the DNA content analysis. These results thus strongly support the notion that the Nrf2-deficient cells in the esophagus were eliminated during 4NQO treatment, while in contrast, 4NQO induced Nrf2 and Nqo1 in the cells with Nrf2 intact, and these Nrf2-positive cells gradually occupied the entire epithelial layers.

High-grade structural atypia and DNA damage accumulation in the Nrf2-deficient cells during 4NQO treatment.

A closer examination of the 4NQO-treated esophagus showed that 4NQO treatment gave rise to significant changes in the cell population of the esophageal epithelium of the K5CreERT2-Nrf2^F/F mice. To further address the mechanism by which the Nrf2-deficient cells were initially eliminated during 4NQO treatment, we tried to evaluate the morphology of basal cells in the K5CreERT2-Nrf2^F/F and control mouse esophagi at day 10 of 4NQO administration.

We classified the morphology of the 4NQO-exposed esophagus into 3 grades based on the histological structure of basal cells (Fig. 5A). Grade 0 is the normal epithelium in which basal epithelial cells in the esophagus align in an orderly manner on the submucosal layer (Fig. 5A and B, left panel). Grade 1 corresponds to structural atypia in which the nucleolus becomes apparent in basal epithelial cells (Fig. 5A and B, middle panels). Grade 2 corresponds to structural and alignment atypia in which the nucleus swells and some basal epithelial cells separate from the basement membrane (Fig. 5A and B, right panels). In human histological classification, both grades 1 and 2 correspond to intraepithelial neoplasia. We examined the appearance of these dysplasia grades in the K5CreERT2-Nrf2^F/F and control Nrf2^F/F mouse esophagi by measuring the length of the basement membrane neighboring the dysplasia using imaging software, and the corresponding lengths for each grade were calculated. Notably, these two types of high-grade structural atypia were markedly increased in the K5CreERT2-Nrf2^F/F mouse esophagus at day 10 of 4NQO treatment (Fig. 5C). Grade 2 atypia accounted for as much as 72% of the K5CreERT2-Nrf2^F/F mouse esophagus. These results indicate that the Nrf2 deletion in almost half of the esophageal epithelial cells gives rise to high-grade structural atypia during 4NQO treatment.

FIG 5.

High-grade structural atypia and DNA damage accumulation in cells with Nrf2 deleted during 4NQO treatment. (A) Schematic of histological classification. Based on the histological structure of basal cells of Nrf2^F/F mouse esophagus 10 days after 4NQO administration, 4NQO-exposed esophagus was classified into 3 grades. Grade 0 is the normal epithelium in which basal epithelial cells in the esophagus align in an orderly manner on the submucosal layer. Grade 1 corresponds to structural atypia in which the nucleolus becomes apparent in basal epithelial cells. Grade 2 corresponds to structural and alignment atypia in which the nucleus swells and some basal epithelial cells separate from the basement membrane. (B) HE staining. (C) Ratio of histological grades judged by HE staining. (D) Immunohistochemistry of γH2A.X. (E) The number of γH2A.X-positive cells per the length of the basal epithelial layer (n = 5). Note that γH2A.X-positive cells are detached from the basement membrane in grade 2 epithelium. (F) Immunohistochemistry of cleaved caspase-3 (Casp3). (G) The number of cleaved Casp3-positive cells per the length of the basal epithelial layer (n = 5). (H) Immunohistochemistry of Sox2. (I) Expression of Sox2. Data in panels C and I were analyzed using the chi-square test. Data in panels E and G represent the mean ± SD. Data in panels E and G were analyzed using Student's t test. n.s., not significant, *, P < 0.05, and **, P < 0.01, compared with the Nrf2^F/F mice at each time point.

It has been shown that 4NQO provokes DNA damage (19). Therefore, we analyzed phosphorylated histone H2A.X at Ser139 (γH2A.X), a marker of DNA double-strand breaks in the 4NQO-treated epithelium. While γH2A.X-positive cells could be observed at the basal layer in grade 0 esophagus, the cells increased significantly in grade 1 and 2 atypia (Fig. 5D). γH2A.X-positive cells detached from the basal layer and started migration toward the lumen side. Importantly, these γH2A.X-positive cells were more abundant in the K5CreERT2-Nrf2^F/F mouse esophagus than in the control mouse esophagus, showing very good agreement with the increase in atypia grade (Fig. 5E).

To address the reason why the Nrf2-deficient cells were eliminated during 4NQO treatment, we next examined whether the Nrf2-deficient cells in the K5CreERT2-Nrf2^F/F mouse esophagus were vulnerable to apoptosis. We examined the expression of cleaved caspase-3, a marker of apoptosis, and found that the expression of cleaved caspase-3 was very low in all three grades (Fig. 5F). We counted the numbers of cleaved caspase-3-positive cells, and the numbers were comparable between the K5CreERT2-Nrf2^F/F and control Nrf2^F/F mouse esophagi (Fig. 5G). These results suggest that the elimination of Nrf2-deficient cells during 4NQO treatment might be caused by mechanisms other than apoptosis.

Next, we addressed the question of whether the self-renewal ability and stemness of esophageal epithelial cells were maintained during 4NQO treatment. For this purpose, we examined the expression of Sox2, which is essential for maintaining self-renewal and stemness (30). We found that the expression of Sox2 markedly decreased in grade 2 atypia (Fig. 5H). We measured the Sox2 expression level independently from the dysplasia grades. While most of the basal cells expressed high levels of Sox2 in the control Nrf2^F/F mouse esophageal epithelia, approximately 45% of basal cells in the K5CreERT2-Nrf2^F/F mouse esophageal epithelia expressed low levels of Sox2 (Fig. 5I), suggesting that the cells with low levels of Sox2 are eliminated through the turnover of the epithelium. These results thus indicate that cells with Nrf2 deleted develop structural atypia and turn over. We surmise that this mechanism underlies, at least in part, the selective elimination of Nrf2-deficient cells.

Low-frequency contribution of the Nrf2-deficient cells to 4NQO-induced tumors.

To examine 4NQO-induced tumor formation in conditionally Nrf2-deficient mice, we treated the Tam-induced Nrf2-deficient mice with 4NQO for 12 weeks following a water-drinking chase period of 12 weeks to grow tumors to detectable levels (Fig. 6A). It should be noted that the amount of Tam employed in this study could elicit the Nrf2 gene deletion in approximately half of the esophageal epithelial cells, so that the epithelium was a mixture of cells with Nrf2 deleted and those with Nrf2 intact, at least in the first week of the 4NQO treatment. Most Nrf2-deficient cells were then eliminated from the epithelium in the later phase of 4NQO treatment. There are two possibilities: tumors are developed from cells with Nrf2 intact or developed from the remaining cells with Nrf2 deleted.

FIG 6.

Inducible Nrf2 deletion does not enhance 4NQO esophageal carcinogenesis. (A) Administration schedule of Tam and 4NQO. (B) Body weight changes during the experiment. Body weights at day 0 of 4NQO administration were defined as 100% (n = 11 to 15). (C) Representative images of the esophagus 24 weeks after 4NQO administration. Arrows indicate tumors. (D) The number of tumors >1 mm in length (n = 11 to 15). (E) Maximum length of tumors per mouse (n = 11 to 15). (F) HE staining of tumor sections. The panels show cross-sections of the esophagus that contain tumors. (G) The number of invasive cancers (n = 8 to 12). Data represent the mean ± SD. These data were analyzed using Student's t test. n.s., not significant compared with the Nrf2^F/F mice.

The body weight changes of the K5CreERT2-Nrf2^F/F mice were comparable with those of the control mice by 24 weeks during the 4NQO treatment (Fig. 6B). We found comparable levels of tumor formation in both genotypes: the number of tumors with a diameter of more than 1 mm in the K5CreERT2-Nrf2^F/F mice was similar to that in the control Nrf2^F/F mice (Fig. 6C and D). Sizes of individual tumors were also comparable in the K5CreERT2-Nrf2^F/F and control Nrf2^F/F mouse esophagi (Fig. 6E).

In the histological analyses, the tumors in the K5CreERT2-Nrf2^F/F and control Nrf2^F/F mice appeared similar and did not show apparent differences (Fig. 6F). We observed one difference between the two genotypes: the number of mice harboring invasive cancers decreased in the K5CreERT2-Nrf2^F/F mice compared with the control Nrf2^F/F mice following 4NQO treatment (Fig. 6G). These results indicate that the moderate depletion of the Nrf2-positive cells did not strongly affect 4NQO-induced tumor formation.

We next asked whether the tumors were composed of cells with Nrf2 deleted or those with Nrf2 intact. To this end, we conducted immunohistochemical staining of Nqo1 in the 4NQO-induced tumors at 24 weeks after the start of 4NQO treatment. We found that the majority of the tumors in both the Tam-treated control Nrf2^F/F mice and the K5CreERT2-Nrf2^F/F mice expressed Nqo1 (Fig. 7A, left two panels). However, there was a small number of Nqo1-negative tumors (Fig. 7A, right panel). When we counted the Nqo1-positive and Nqo1-negative tumors, all 15 tumors examined in the control Nrf2^F/F mice showed Nqo1-positive staining. Similarly, 21 out of 24 tumors (88%) in the esophagus of the K5CreERT2-Nrf2^F/F mice were positive for Nqo1 (Fig. 7B).

FIG 7.

Low-frequency contribution of cells with Nrf2 deleted to 4NQO-induced tumors. (A) A representative photo of Nqo1 immunohistochemistry in the 4NQO-induced tumors of the Nrf2^F/F and K5CreERT2-Nrf2^F/F mice. (B) Nqo1 protein expression levels in tumors in the Nrf2^F/F and K5CreERT2-Nrf2^F/F mice 24 weeks after the start of 4NQO administration. (C) Immunohistochemistry of Nrf2 in the Nrf2^F/F and K5CreERT2-Nrf2^F/F mice 24 weeks after the start of 4NQO administration. Note that these tumors express Nrf2.

We selected some Nqo1-positive tumors formed in both genotypes and immunostained Nrf2. Consistent with the Nqo1 immunostaining, we observed nuclear accumulation of Nrf2 in the Nqo1-positive tumors in both the control Nrf2^F/F and K5CreERT2-Nrf2^F/F mouse esophagi, as Nrf2 was induced by 4NQO in the esophageal epithelial cells (Fig. 7C). These results thus demonstrate that all tumors in the control Nrf2^F/F mice and most tumors in the K5CreERT2-Nrf2^F/F mice were composed of cells with Nrf2 intact, indicating that such cells contribute to 4NQO-induced tumor formation. These observations further suggest that the low-frequency contribution of Nrf2-deficient cells to tumor formation may be due to the elimination and disappearance of these cells from the esophageal epithelium during 4NQO treatment.

Nrf2 deletion after 4NQO exposure did not influence tumor formation.

It is unclear whether cells with Nrf2 deleted could contribute to 4NQO-induced tumor formation if they did not disappear during 4NQO treatment. To address this question, we exposed the esophageal epithelial cells after the completion of 12-week 4NQO treatment to Nrf2 deletion by Tam. We introduced the Nrf2 deletion in the K5CreERT2-Nrf2^F/F mice 12 weeks after 4NQO administration (i.e., post-4NQO protocol [Fig. 8A]). In this protocol, the Nrf2-deficient cells were not directly exposed to 4NQO.

FIG 8.

Nrf2 deletion after 4NQO exposure does not influence tumor formation. (A) Administration schedule of Tam and 4NQO in this experimental set. (B) Body weight changes during the experiment. Body weights at day 0 of 4NQO administration were defined as 100% (n = 13 or 14). (C) Representative images of the esophagus 24 weeks after the start of 4NQO administration. Arrows indicate tumors. (D) The number of tumors >1 mm in length (n = 13 or 14). (E) Maximum length of tumors per mouse (n = 13 or 14). (F) HE staining of the tumors. These panels show cross-sections of the esophagus harboring tumors. (G) The number of invasive tumors (n = 13 or 14). (H) Recombination rate of the Nrf2 gene in whole esophageal epithelium. The remaining Nrf2 DNA was quantified after deletion by Tam (n = 4 or 5). (I) Recombination rate of the Nrf2 gene in the tumor (n = 3 or 4). Note that the Nrf2 gene deletion occurred similarly in the tumors from the control Nrf2^F/F and K5CreERT2-Nrf2^F/F mouse esophagi in this protocol. Data represent the mean ± SD. These data were analyzed using Student's t test. n.s., not significant, *, P < 0.05, and **, P < 0.01, compared with the Nrf2^F/F mice.

Throughout the experimental period, body weight changes of the K5CreERT2-Nrf2^F/F mice were comparable to those of the control Nrf2^F/F mice (Fig. 8B). In this protocol, all mice in both genotypes were treated with 4NQO and developed macroscopic tumors (Fig. 8C). There were no significant differences in the numbers and sizes of the tumors between the control Nrf2^F/F and K5CreERT2-Nrf2^F/F mice (Fig. 8D and E). The numbers of invasive cancers in the K5CreERT2-Nrf2^F/F mice were also comparable to those in the control Nrf2^F/F mice (Fig. 8F and G).

To ascertain whether Nrf2 deletion by Cre recombinase occurs normally in the 4NQO-treated cells or Nrf2-deficient cells did not disappear during this postprotocol, we examined the recombination rate of the Nrf2 gene in the esophageal epithelium. The recombination rate in the whole esophageal epithelium was almost 50% at 24 weeks (Fig. 8H). We also collected tumor regions and analyzed the recombination rate. The recombination rates of Nrf2 in tumor regions were also approximately 50% on average, but the rates diverged from 26% to 99% in individual tumors (Fig. 8I). These results indicate that the Cre recombinase-mediated Nrf2 deletion was certainly induced by Cre recombinase and that the Nrf2-deficient cells were maintained in the epithelium throughout this postprotocol. We surmise that the Cre recombinase-mediated Nrf2 deletion occurs rather randomly in the 4NQO-induced tumor-initiated cells and that both Nrf2-positive and Nrf2-negative tumors were developed in this postprotocol.

DISCUSSION

In this study, we analyzed the fate of the epithelial cells with Nrf2 deleted surrounded by cells with Nrf2 intact under 4NQO carcinogen-stressed conditions. For this purpose, we established Cre recombinase-based Nrf2-deletion model mice (K5CreERT2-Nrf2^F/F) in which the number of Nrf2-positive cells in the esophageal epithelium was reduced to approximately half. In this line of mice, the mRNA levels of Nrf2 and Nqo1 in the esophageal epithelium were also reduced to approximately half of those in the control Nrf2^F/F mice. Under unstressed conditions, the number of cells with Nrf2 deleted was maintained at approximately half in the epithelium for at least 12 weeks, and the epithelial cells did not show apparent abnormalities. In contrast, in the 4NQO-induced carcinogenesis experiment, the cells with Nrf2 deleted were susceptible to the carcinogen and accumulated DNA damage, and these Nrf2-deficient cells were selectively eliminated from the esophageal epithelial cell population (Fig. 9). Therefore, our results indicate that these cells with Nrf2 deleted are less likely to develop 4NQO-induced tumors, but cells with Nrf2 intact form tumors under the 4NQO carcinogen-treated conditions.

FIG 9.

Cells with Nrf2 deleted suffer more DNA damage and selectively disappear in the esophageal epithelium. (A) Deletion of Nrf2 before 4NQO exposure. Nrf2-deficient cells remain for weeks in the esophageal epithelium in unstressed conditions. After exposure directly to 4NQO, esophageal epithelial cells accumulate DNA damage. Under this condition, Nrf2-deficient cells are selectively eliminated from the epithelium. As a result, almost all 4NQO-induced tumors are derived from cells with Nrf2 intact, but not from cells with Nrf2 deleted. (B) Deletion of Nrf2 after 4NQO exposure. In this experimental set, cells with Nrf2 deleted are not directly exposed to 4NQO. Therefore, Nrf2-deficient cells persist in the epithelium. As a result, 4NQO-induced tumors are derived from both cells with Nrf2 deleted and cells with Nrf2 intact. The present study presents models in which both cells with Nrf2 deleted and cells with Nrf2 intact coexist in the esophageal epithelium. These results should be interpreted in a distinct context with the oncogenic potential of Nrf2-deficient esophageal epithelial cells, in which Nrf2-deficient cells occupy the entire epithelium.

It has been shown that the tumorigenicity of 4NQO is elicited through two mechanisms: reactive oxygen species (ROS)-mediated oxidative DNA damage (31) and direct DNA adduct formation by the reactive metabolite 4-hydroxyaminoquinoline-1-oxide (4HAQO) (32). The activation of Nrf2 works to eliminate ROS through upregulation of antioxidative enzymes. Moreover, glutathione S-transferase P1 (GSTP1), the representative Nrf2 target gene product, catalyzes the glutathione conjugation of 4NQO and detoxifies the carcinogen, leading to the prevention of carcinogenesis (33). However, it should be noted that a product of another representative Nrf2 target gene, NQO1, acts to convert 4NQO to 4HAQO (32). In this study, we found that epithelial cells with Nrf2 deleted were susceptible to 4NQO treatment and developed DNA damage, resulting in structural atypia. These observations indicate that Nrf2 plays an important role in the prevention of 4NQO-induced DNA damage, which is in very good agreement with previous observations (19). Since Nrf2 has been shown to be important for the detoxification of various chemical carcinogens (7, 34), we surmise that Nrf2-deficient cells are also susceptible to other chemical carcinogens.

The selective elimination of Nrf2-deficient cells may act as a defense mechanism of whole esophageal epithelial tissue against carcinogenesis. In fact, several lines of recent studies using mouse carcinogenesis models demonstrated that precancerous cells are eliminated by cell competition (22, 23, 35). For instance, when constitutively active Ras-transformed cells are surrounded by normal epithelial cells, the transformed cells compete with normal cells, and the transformed cells eventually lose and are eliminated (36). In this regard, it is interesting to note that normal esophageal epithelium harbors cell heterogeneity, which consists of several cell clones with somatic mutations (24). These cell clones compete in epithelial tissues (25). Supporting this notion, it has been shown that irradiated esophageal basal cells accumulate DNA damage and differentiate into suprabasal cells, and thereby, these damaged cell clones are eliminated from the esophageal epithelium (23). Showing very good agreement with this notion, our present results revealed that, whereas Nrf2-deficient cells persist in unstressed esophageal epithelium for at least 12 weeks, the Nrf2-deficient cells are excluded efficiently after challenge with 4NQO because 4NQO induces DNA damage preferentially in Nrf2-deficient cells. These observations suggest that high-risk precancerous cells with DNA damage may be more susceptible to cellular elimination than cells without DNA damage in the mouse esophagus, and this system is likely to act as a defense system for chemical carcinogenesis.

Molecular mechanisms to eliminate cells with Nrf2 deleted are unclear at present. We suppose that the elimination is mediated via cell death. In this regard, apoptosis appears to be unlikely as there is not much difference in the cleavage of caspase-3. On the contrary, other types of cell death program or autophagic cell death are possible in this elimination event. An alternative possibility is that the changes in the tumor microenvironment may lead to the selective cell elimination. It has been shown that phagocytosis of damaged cells is mediated by macrophages (37, 38). In the tumor microenvironment, tumor cells tend to convert macrophages into tumor-associated macrophages (39). Tumors with high levels of expression of Nrf2 are reported to suppress tumor immunity and acquire malignant cell growth (40). These wide-ranging observations suggest that the selective cell elimination in the esophagus epithelium involves complex mechanisms, and further studies are necessary to clarify the mechanisms.

The results of this study should be interpreted in a distinct context with the oncogenic potential of Nrf2-deficient esophageal epithelial cells in which Nrf2-deficient cells occupy the entire epithelium. In fact, previous analyses of 4NQO carcinogenesis utilizing systemic Nrf2 knockout mice showed that Nrf2-deficient cells are prone to develop cancers (19). Other models also showed similar results (12, 13). However, we found that cells with Nrf2 deleted surrounded by cells with Nrf2 intact are selectively eliminated during carcinogen treatment. These findings raise the question of whether somatic mutations resulting in Nrf2 deficiency are relevant to cancer risk. Whereas we do not have an immediate answer to this question, the following observation may be pertinent. Currently, somatic mutations related to the Keap1-Nrf2 system in cancer patients are known to cause constitutive activation of Nrf2 but not the loss of function of Nrf2 (41). While this observation does not exclude the possibility that the loss-of-function mutation of Nrf2 retains a certain impact on carcinogenesis, we surmise based on our current observations that in most of the Nrf2 loss-of-function mutations, the cells are susceptible to elimination from the epithelial tissue. Thus, we propose that somatic mutations provoking Nrf2 deficiency may not be a risk factor for cancers in most cases.

In contrast, it has been found that minor A/A homozygotes of single nucleotide polymorphism in the human Nrf2 promoter region (rs6721961) exhibit low levels of Nrf2 mRNA with an increased risk of lung cancer (42). Thus, low-level expression of Nrf2 is suggested to be a cancer risk in the lung. Based on this observation, we envisage that the cancer risk of Nrf2 deficiency is context dependent and different from tissue to tissue or organ to organ. Elimination of Nrf2-deficient cells from the tissue may be an important determinant in this context, but the mechanisms by which Nrf2-deficient cells are eliminated from the esophageal epithelium remain to be clarified.

Similarly, how Nrf2 activation contributes to cell selection has not yet been elucidated. It has been reported that Nrf2-overexpressing cells are eliminated by neighboring cells in a Drosophila wing disc study (43). However, high levels of endogenous antioxidants in Nrf2-activated cells protect the cells from competition with normal cells in the esophageal epithelium (23). It has also been shown that cell competition in the squamous epithelial population is affected by various factors, including cellular proliferation, differentiation, and stress tolerance (35, 44). Outcompeting cells are characterized by the promotion of proliferation, suppression of differentiation, and stress resistance. Activation of Nrf2 promotes differentiation and stress resistance (45), indicating both advantages and disadvantages in cell competition. Thus, proper Nrf2 activation may be important in cell competition.

In this regard, we have examined the expression of Sox2, which is essential for the maintenance of self-renewal or stemness. It has been shown that Nrf2 is required for the Sox2 expression (46, 47). We found in this study that Sox2 expression is decreased in the Nrf2-deficient cells upon treatment with 4NQO. This result further supports the notion that Nrf2 contributes to the Sox2 expression or maintenance of cell stemness, but loss of Nrf2 leads to the elimination of cells.

As most human esophageal cancers at advanced stages exhibit an NRF2-positive nature (41, 48), we surmise that the observations in this study indicate a good mouse model to reflect human carcinogenesis. Our current results clearly demonstrate that early-stage carcinogenesis in the esophageal epithelium is heavily linked to the maintenance of the cell population, and the Keap1-Nrf2 system is a key regulator of cellular functions directing the initial stage of carcinogenesis or adaptation of cells to the epithelial cell population.

MATERIALS AND METHODS

Generation of the K5CreERT2 plasmid construct.

The pKM2L-phK5 plasmid containing the human Keratin5 (K5) promoter was obtained from Riken BioResource Research Center (no. RDB05886). An NheI site prior to the luciferase gene in the pKM2L-phK5 plasmid was converted to a NotI site by PCR mutagenesis (Fig. 1A). The resulting plasmid was digested by NotI to remove the luciferase gene. CreERT2 cDNA was generated by PCR using pCAG-CreERT2 (Addgene) as a template. CreERT2 cDNA was inserted into the NotI site in the pKM2L-phK5 plasmid by homologous recombination using the In-Fusion HD cloning kit (Clontech) to generate the K5CreERT2 plasmid. The K5CreERT2 plasmid was digested by SwaI to remove a vector and used for microinjection.

Mice.

K5CreERT2 transgenic mice were generated by standard pronuclear injection of the K5CreERT2 construct into C57BL/6 × BALB/c-fertilized eggs using Leica AM6000 (Leica). Gt (ROSA) 26Sor^{tm7 (CAG-tdTomato)Nat} (Rosa26-tdTomato) mice (no. 026006) were obtained from the Jackson Laboratory. Nrf2^flox mice were kindly provided by Jingbo Pi (49). These mice were maintained under specific-pathogen-free conditions in the animal facility at Tohoku University. All animal experiments were approved by the Animal Care Committee at Tohoku University.

Gene deletion in the esophagus using K5CreERT2 mice.

K5CreERT2 transgenic mice were crossed with Rosa26-tdTomato mice to generate K5CreERT2::Rosa26-tdTomato mice. In addition, K5CreERT2 mice were crossed with Nrf2^flox mice on a mixed background, and we obtained Nrf2^F/F mice and K5CreERT2-Nrf2^F/F mice. For induction of Cre expression, Tam (Sigma-Aldrich) dissolved in corn oil at 20 mg/ml was intraperitoneally administered to male mice (100 μg/g body weight) at the age of 7 to 9 weeks for 3 consecutive days. In K5CreERT2::Rosa26-tdTomato mice, several tissues were collected, and frozen sections were observed under a DM2500LED microscope (Leica).

Carcinogenesis induced by 4NQO in the esophagus.

Two carcinogenic protocols were performed: Tam-induced deletion of the Nrf2 gene with Tam before or after administration of 4NQO (Sigma-Aldrich), a carcinogen in the esophagus. The detailed schedules are illustrated in each figure. The former protocol (preprotocol) was as follows. Tam (20 mg/ml in corn oil) was intraperitoneally administered to male K5CreERT2-Nrf2^F/F and control Nrf2^F/F mice (100 μg/g body weight) at the age of 7 to 9 weeks for 3 consecutive days. At 1 week after the last administration of Tam, these mice were administered 4NQO for 12 weeks and subsequently were given 4NQO-free drinking water for 12 weeks after 4NQO administration. Then, 4NQO was dissolved in dimethyl sulfoxide (DMSO) at 10 mg/ml and diluted in drinking water to 0.1 mg/ml. All of the mice were allowed to drink water ad libitum in all periods. Once a week, the mice were weighed, and the drinking water was replaced with fresh water. The later protocol (postprotocol) was as follows. Male K5CreERT2-Nrf2^F/F and control Nrf2^F/F mice were administered 4NQO-containing water (0.1 mg/ml) for 12 weeks. After 4NQO administration, Tam (20 mg/ml in corn oil) was intraperitoneally injected into these mice for 3 consecutive days, and they were given 4NQO-free drinking water for another 12 weeks. At 24 weeks, these mice were euthanized with isoflurane, and the esophagus was dissected. The whole esophagus was opened longitudinally and photographed. The maximum diameter of every tumor was measured using ImageJ Fiji software, and the number of tumors larger than 1 mm in diameter was counted.

Quantitative reverse transcription-PCR.

Dissected esophagi were opened longitudinally and treated with I-Trypsin (Nacalai Tesque) for approximately 8 h, and then the epithelial layer was separated from the other layers. Total RNA was isolated from the esophageal epithelium using Sepasol-RNA I Super G (Nacalai Tesque). The RNA concentration was measured using a NanoPhotometer NP80 (Implen). RNA was transcribed into cDNA using ReverTra Ace (Toyobo). The obtained templates were used for quantitative real-time PCR with the KAPA SYBR Fast qPCR master mix (2×) kit (Kapa Biosystems) and the QuantStudio 6 Flex real-time PCR system (Thermo Fisher Scientific). rRNA was used as an internal control. The primers used for quantitative reverse transcription-PCR (qRT-PCR) are listed in Table 1.

TABLE 1.

Primers used for qRT-PCR

Mouse gene	Primer	Primer sequence (5′→3′)
Nrf2	Forward	TGGCTGCTTTTGGTGGTGAG
	Reverse	CAGCGCTGAGCAAACATAAG
Nqo1	Forward	AGCTGGAAGCTGCAGACCTG
	Reverse	CCTTTCAGAATGGCTGGCA
rRNA	Forward	CGGCTACCACATCCAAGGAA
	Reverse	GCTGGAATTACCGCGGCT

DNA recombination rate.

The DNA recombination rate was analyzed by qPCR. Dissected esophagi were opened longitudinally and treated with I-Trypsin (Nacalai Tesque) for approximately 8 h, and then the epithelial layer was separated from the other layers. Tumors detected macroscopically were collected and used for DNA analysis. DNA was isolated from the esophageal epithelium or tumors. The obtained templates were measured using a NanoPhotometer NP80 (Implen), and qPCR was performed as described above. The β-Actin allele was used as an internal control. The primers used for qPCR are listed in Table 2.

TABLE 2.

Primers used for qPCR of DNA recombination

Mouse gene(s)	Primer	Primer sequence (5′→3′)
β-Actin	Forward	CCATAGGCTTCACACCTTCCTG
	Reverse	GCACTAACACTACCTTCCTCAACCG
Nrf2 loxP	Forward	CACAATGGTATGCCTGCTGT
	Reverse	AAGAGGGGGTTGGAAAGAGA

Histological analyses.

Esophagi were opened longitudinally or cross-sectionally and fixed with Mildform 10N (Wako Pure Chemical Industries). The fixed esophagi were embedded in paraffin, cut into 4-μm-thick sections, and stained with hematoxylin and eosin (HE). Histologic images were captured with DM2500LED (Leica) using Leica Application Suite version 4.8 software (Leica). The lengths of the keratinous layer and epithelial cell layer were measured at 3 points per slide using ImageJ Fiji software, and the average of these was evaluated. To quantify the ratios of three dysplasia patterns (grades 0 to 2), the length of the basement membrane neighboring the dysplasia was measured using ImageJ Fiji software, and the corresponding lengths to each grade were calculated. In 6 or 7 cross-sectional slices that included tumors for each mouse, the invasive locus was counted without knowledge of the genotype.

Immunohistochemistry.

Immunohistochemical staining was performed using rat monoclonal anti-Nrf2 antibody (1:400; no. D9J1B, Cell Signaling Technology), goat polyclonal anti-Nqo1 antibody (1:500; no. ab2346, Abcam), and rabbit monoclonal anti-γH2A.X (Ser139) antibody (1:300; no. 9718, Cell Signaling Technology) as a DNA double-strand break marker. Rabbit polyclonal anti-Sox2 antibody (1:1,000; no. ab97959, Abcam) was used as a stemness marker, and rabbit polyclonal cleaved caspase-3 antibody (1:300; no. 9661, Cell Signaling Technology) was used as an apoptotic marker. All antibodies were incubated for 16 h at 4°C. Immunohistochemically, positive cells were detected and counted using ImageJ Fiji software, and the rate of positive cells in the epithelium was calculated. Nqo1-positive areas or areas with a low level of Sox2 were also detected using ImageJ, and we measured the length of basal layer, counted the number of basal cells on the line, and evaluated the Sox2 immunostaining. We determined the immunostaining level of individual cells as either high or low based on the staining intensity under the supervision of pathologists.

Statistical analyses.

The average values were calculated, and the error bars indicate the standard deviations (SD). The differences in continuous data were analyzed using Student's t test. Correlations between two variables were analyzed using the chi-square test. A P value <0.05 was considered statistically significant.