Roles of Keap1–Nrf2 System in Upper Aerodigestive Tract Carcinogenesis

Cancers in the upper aerodigestive tract, including cancers in the oral cavity, oropharynx, hypopharynx, larynx, and esophagus, are the third leading cause of cancer-related deaths in the world (1). Tongue and esophageal cancers are the most frequent among them. Because both the tongue and the esophagus are covered by stratified squamous epithelium (2), squamous cell carcinoma is the most common cancer type in these regions. Important risk factors for tongue and esophageal cancers include tobacco smoking and alcohol use, and the cancer risk attributable to both factors is estimated to be more than 60% (3). It has been reported that carcinogenesis induced by tobacco and alcohol is mediated, at least in part, by oxidative stress (4). However, while the importance of oxidative stress in oral and esophageal carcinogenesis is well recognized, little is known about the mechanisms mediating how stress provokes carcinogenesis or how cancer prevention is attained through targeting oxidative stresses.

The transcription factor Nrf2 plays a pivotal role in cellular defense against toxic electrophiles and oxidative stresses (5). Nrf2 regulates both basal and inducible expression of antioxidative and detoxifying enzymes (6), including NAD(P)H quinone oxidoreductase 1 (NQO1), glutathione S-transferases (GST), glutamate–cysteine ligase catalytic subunit (GCLC), and hemoxygenase-1 (HO-1). Due to a lack of activation of these genes, Nrf2-knockout (Nrf2-KO) mice are susceptible to various chemicals, including carcinogens for stomach (7), bladder (8), skin (9), liver (10), colon (11), and breast (12).

We have been studying the molecular basis of Nrf2 activation in response to environmental insults (13). We found that Keap1 regulates Nrf2 stability and subsequent activity (14). Keap1 is a subunit of the ubiquitin E3 ligase complex (15). Under nonstressful conditions, Keap1 constitutively and efficiently ubiquitinates Nrf2 and promotes its rapid degradation via the proteasome pathway. Under stressful conditions, Keap1 is inactivated, and Nrf2 is stabilized and activated. Reactive cysteine residues of Keap1 are modified by electrophiles or reactive oxygen species (ROS), which leads to Nrf2 stabilization and accumulation in the nucleus and subsequently activates Nrf2 (15, 16). While some antioxidant chemicals have cancer preventive effects in an adequate range of doses, there are concerns that an overdose of these chemicals could have some degree of toxicity due to electrophilic insults (17, 18). Another hurdle for the Nrf2-inducing chemicals is that the pharmaceutical Nrf2 activation is only transient (19). Therefore, 1 of the important strategies for cancer prevention is promoting constitutive Nrf2 activation by Keap1 suppression using nonelectrophilic molecules.

Recent studies have found that there are somatic mutations of KEAP1 in various human cancer cells that cause constitutive NRF2 activation and promote cancer malignancy (20, 21). Another intriguing observation is that Keap1-knockout mice die at weaning due to malnutrition caused by severe hyperkeratosis of the upper digestive tract (22). Considering these effects of Keap1 loss-of-function, it remains to be seen whether suppression of Keap1 function is beneficial for cancer prevention. Recently, we generated Keap1-knockdown (Keap1-KD) mice that are viable with constitutive Nrf2 activation (23). These mice are resistant to some types of stresses (24, 25) and provide an excellent model system to study whether Keap1-KD mice are resistant to carcinogens.

To examine the cancer preventive effect of Keap1-KD in the tongue and the esophagus, we used the water-soluble carcinogen 4-nitroquinoline-1-oxide (4NQO) for experimental carcinogenesis (26). We selected this system because the animal model of oral carcinogenesis with 4NQO allows for the reproducible isolation of all stages of tongue and esophageal carcinogenesis, and the tumors have histologic and molecular changes that are similar to human cancer (27). 4NQO generates oxidative stress, and its metabolite covalently binds to DNA. In this study, we found that Nrf2-KO mice are susceptible to tongue and esophageal carcinogenesis induced by 4NQO, showing the importance of Nrf2 for the prevention of tongue and esophageal cancer. We also showed that Keap1-KD mice are resistant to tongue and esophageal carcinogenesis. No growth advantage of 4NQO-induced cancer cells in Keap1-KD mice was observed, showing that the suppression of Keap1 activity has potential for cancer prevention without significant adverse impacts.

4NQO was purchased from Sigma-Aldrich for experimental carcinogenesis and from Wako for experiments in cultured cells. Diethyl maleate (DEM) was purchased from Wako.

The human oral squamous cell carcinoma cell line HSC3 was kindly provided in 2011 by Dr. Saiki (Tohoku University). Cells were regularly tested for mycoplasma contamination using the e-Myco Mycoplasma PCR Detection Kit (iNtRON Biotechnology). No cell authentication was done by the authors. Cells were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% FBS and supplemented with antibiotics.

Six- to 7-week-old female wild-type (WT), Keap1-KD, and Nrf2-KO mice on the C57BL/6J genetic background were used for this study. The mice were maintained under specific pathogen-free conditions in the animal facility at Tohoku University. All of the animal experiments were executed with the approval of the Tohoku University Animal Care Committee.

Cultured cells were collected and lysed in SDS sample buffer and stored at −80°C (16). The samples were subjected to immunoblot analysis using anti-Nrf2 (clone 103; ref. 28), anti-phosphorylated histone H2AX (Millipore), anti-histone H2AX (Calbiochem), and anti-tubulin (Sigma) antibodies.

Total RNA was isolated from either cultured cells or tongues and esophagi from mice using the Isogen RNA extraction kit (Nippon Gene). cDNA was synthesized from 1 μg total RNA using Superscript III (Invitrogen). Quantitative real-time PCR (qRT-PCR) analysis was conducted using the ABI7300 system (Applied Biosystems). The primers and probes for detecting human NQO1, GCLC, GSTP1, GPX2, and mouse Nqo1, Gclc, Gstp1, Gpx2, G6pd, Pgd, Tkt, Taldo1, Me1, and 18S ribosomal RNA are described in Supplementary Table S1.

4NQO was dissolved in dimethylsulfoxide (DMSO) at 10 mg/mL and diluted into the drinking water to 100 μg/mL. WT, Keap1-KD, and Nrf2-KO mice were subdivided into 3 groups. Five mice per genotype received tap water for 24 weeks as the control groups. For the experimental groups treated with 4NQO, at least 20 mice per genotype were treated first with 100 μg/mL 4NQO in the drinking water for either 8 (short term) or 16 weeks (long term) and subsequently with tap water for either 16 (short term) or 8 weeks (long term). All of the mice were allowed to drink 4NQO-containing water or tap water ad libitum in all periods of the experiments. All of the mice were weighed and checked once a week. 4NQO-containing water and tap water were changed fresh every week. At the end of the experiments, the mice were sacrificed after anesthetization, and tongues and esophagi were dissected. Whole esophagi were opened longitudinally, and tongue and esophageal tumors larger than 1 mm diameter were counted.

For the analysis of blood counts, WT and Nrf2-KO mice were treated with 100 μg/mL 4NQO in the drinking water for 8 or 12 weeks, and peripheral blood taken from the retro-orbital plexus was analyzed by a hemocytometer (Nihon Kohden).

Tissue from the mice was fixed with Mildform 10N (Wako), embedded into paraffin, and cut into 5-μm thick sections. After staining with hematoxylin and eosin (H&E), a histopathological diagnosis of the tongues and the esophagi was carried out by a pathologist without knowledge of the genotypes and the carcinogen experiment. The sections were graded as normal, dysplasia, or squamous cell carcinoma using established criteria (29). For immunohistochemical analysis, an anti-Ki-67 antibody (DAKO) was used. Ki-67-positive cells in tumors or progenitor cells in the basal layer of the tongues were counted using the Image J software for the analysis of cell growth.

Significant differences were determined using the Student's t test or the Mann–Whitney test. The values are presented as the means ± SEM.

To explore the roles the Keap1–Nrf2 system plays in upper aerodigestive tract carcinogenesis, we used a 4NQO experimental carcinogenesis model. While Nrf2 is activated in response to various environmental stresses (30), the relationship between 4NQO and Nrf2 activation has not been elucidated. Therefore, to delineate whether 4NQO directly activates NRF2, we examined the response of HSC3 human oral cancer cells to 4NQO by measuring NRF2 protein expression. While DEM, an NRF2 inducer, nicely increased NRF2 protein expression, 4NQO did not. When we tested 1 and 3 μmol/L 4NQO, 3 μmol/L 4NQO treatment was sufficient to promote the phosphorylation of histone H2AX, a DNA damage marker (Fig. 1A). Consistent with the observation that 4NQO did not change NRF2 expression, 4NQO also did not increase the mRNA expression of either NQO1 or GCLC, while DEM strongly increased the expression of both genes (Fig. 1B).

To examine the in vivo response to 4NQO, we treated 6-week-old female mice with either tap water or 4NQO (100 μg/mL)-containing water for 24 and 48 hours and examined the expression of Nrf2 target genes in the tongue. The expression of Nqo1, Gclc, Gstp1, and Gpx2 was not increased in the tongues at either 24 or 48 hours after the administration of 4NQO (Fig. 1C). These results show that 4NQO does not activate NRF2 in either human oral cells in culture or mouse tongue in vivo.

It has previously been reported that Nrf2 target genes are highly expressed in the liver and lungs of Keap1-KD mice (23). Consistent with this observation, high levels of Nqo1, Gclc, and Gstp1 were also observed in the tongues and esophagi of Keap1-KD mice (Fig. 1D left and right, respectively). These results suggest that Nrf2 can be stabilized and activated in the tongues and esophagi of Keap1-KD mice.

We also observed that the Nqo1 mRNA expression in the Nrf2-KO mouse tongues was lower than that in the WT mouse tongues, and the expression of Nqo1, Gclc, Gstp1, and Gpx2 mRNA in the esophagi of Nrf2-KO mice was lower than that observed in the WT mouse esophagi (Fig. 1D). These results show that Nrf2 is expressed at a basal level, but the suppression of Keap1 causes a strong activation of Nrf2 in the tongue and esophagus.

To examine the contribution of Nrf2 and Keap1 to tongue and esophageal carcinogenesis, we conducted 2 series of 4NQO carcinogenesis experiments using Nrf2-KO and Keap1-KD mice. In 1 series of experiments, we fed these mutant lines of mice with 4NQO-containing water for 8 weeks. These mice were subsequently fed with tap water for 16 weeks and then sacrificed for analysis. We refer to this experiment as the short-term 4NQO experiment (Fig. 2A). Macroscopically, we found that Nrf2-KO mice had many tongue and esophageal tumors, while Keap1-KD mice showed almost no tongue and esophageal tumors (Fig. 2B). WT mice showed no tongue tumors and a very small number of esophageal tumors.

Statistical analyses revealed that the number of tongue tumors in the Nrf2-KO mice (1.43 ± 0.22) was significantly higher than that in the WT mice (0.29 ± 0.12), while there were no significant differences in the numbers of tongue tumors between Keap1-KD and WT mice (Fig. 2C, left). The number of esophageal tumors in Nrf2-KO mice (5.48 ± 0.45) was significantly higher than that in the WT mice (1.76 ± 0.24), while the number of esophageal tumors in Keap1-KD mice (0.20 ± 0.09) was significantly smaller than that in the WT mice (Fig. 2C, right).

Histologic analyses revealed that there was both dysplasia and squamous cell carcinoma in the tongues of Nrf2-KO, Keap1-KD, and WT mice, which suggests that the tongue tumors from the 4NQO model experiments progressed toward tumorigenesis. Representative histologic sections of the tongues, including dysplasia in WT mice (left) and squamous cell carcinoma in Nrf2-KO mice (right), are shown in Fig. 2D. The analyses revealed that 13 of 21 Nrf2-KO mice had dysplasia (62%) and 5 Nrf2-KO mice had squamous cell carcinoma (24%) in the tongue. These numbers were markedly higher than the numbers observed in WT mice (dysplasia 7/21, 33%; squamous cell carcinoma 1/21, 5%), indicating that tongue tumors from Nrf2-KO mice show more malignant progression than those from WT mice (Table 1A). There was no such difference in the tongues from Keap1-KD and WT mice (Table 1A).

Similarly, there was both dysplasia and squamous cell carcinoma in the esophagi from these mutant mice. Representative histologic sections of the esophagi are shown in Supplementary Fig. S1A (dysplasia in WT mice and squamous cell carcinoma in Nrf2-KO mice). Statistical analyses revealed significantly more advanced esophageal squamous cell carcinomas in Nrf2-KO mice (18/21 were squamous cell carcinoma) compared with the WT mice (1/21 was squamous cell carcinoma, Table 1B). In contrast, the histologic stage of the esophageal carcinomas in Keap1-KD mice was significantly less advanced than those in the WT mice (Table 1B).

Despite the low tumor incidence in the esophagus, Keap1-KD mice drank significantly larger amounts of the 4NQO-containing water per body weight than the WT mice did (Supplementary Fig. S2). In contrast, the Nrf2-KO mice consumed a much smaller amount of the 4NQO-water than the WT mice did (Supplementary Fig. S2). This result is particularly interesting if we consider the fact that the high incidence of squamous cell carcinoma was observed in the tongues and esophagi of Nrf2-KO mice. Thus, these short-term 4NQO experiments show that Nrf2 loss-of-function promotes 4NQO-induced tongue and esophageal carcinogenesis, whereas Nrf2 activation by Keap1-KD prevents tumorigenesis in the esophagus.

Because we were unable to fully evaluate carcinogenesis in the tongues of Keap1-KD and WT mice due to a low incidence of tumors in the short-term experiment with 4NQO, we decided to use more stringent conditions in a long-term experiment with 4NQO. In the long-term experiment, the mice were treated with 4NQO for 16 weeks instead of 8 weeks (Fig. 3A). In the long-term experiment, all of the mice lost body weight (Supplementary Fig. S3A–D). While the majority of the WT and the Keap1-KD mice survived until the end of the experiment, most of the Nrf2-KO mice lost more than 20% of their body weight, and the experiment was stopped by euthanizing the animals (Supplementary Fig. S4A). To identify the reason why the Nrf2-KO mice had such a severe decrease in body weight, we examined Nrf2-KO and WT mice after 8- or 12-week-treatments with 4NQO and found that only a few tumors were observed in the Nrf2-KO and WT mice (data not shown). We believe that the stringent conditions of the long-term 4NQO experiment likely produced severe acute toxicity in the Nrf2-KO mice (31). However, we did not find a difference in hematologic indices (Supplementary Fig. S2) or histologic damage of the liver (Supplementary Fig. S4B) in the Nrf2-KO mice in this experimental condition. We surmise that, as we stopped the 4NQO experiments at 20% body-weight-loss point for animal protection reasons, severe toxicity might not be apparent at this timing. Because the focus of this study is on cancer prevention, we did not conduct further examination about the acute toxicity of 4NQO. For these reasons, we excluded the Nrf2-KO mice from this long-term experiment.

We continued the evaluation of tumor incidence in Keap1-KD mice and WT mice. This long-term treatment with 4NQO caused tongue and esophageal tumors in almost all the WT mice but only in half of the Keap1-KD mice (Fig. 3B and C). Statistical analyses revealed that the number of tongue tumors in Keap1-KD mice (0.83 ± 0.62) was significantly smaller than that in the WT mice (1.68 ± 0.95; Fig. 3C, left panel). The numbers of esophageal tumors in Keap1-KD mice (0.94 ± 1.51) were also significantly smaller than that observed in the WT mice (3.68 ± 1.39; Fig. 3C, right). Histologic analysis revealed that squamous cell carcinoma was observed in the tongues of only 1 of 18 Keap1-KD mice, while more than half of WT mice had squamous cell carcinoma (Table 1C and Fig. 3D). It should also be noted that squamous cell carcinoma was not observed in the esophagi of Keap1-KD mice, while 14 of 22 WT mice had squamous cell carcinoma in the esophagus (Table 1D and Supplemental Fig. S1B). These results show that Keap1-KD improves incidence and prevents the progression of 4NQO-induced carcinogenesis.

There still remain concerns that Nrf2 activity may confer a growth advantage to cancer cells (20, 32). To evaluate the growth ability of cancer cells, we examined the density of Ki-67-positive cells (Fig. 4A and B) by using Ki-67 as a proliferation marker (33). We found that the Ki-67-positive cell density was significantly higher in tongue squamous cell carcinoma than in tongue dysplasia in both Nrf2-KO and WT mice (Fig. 4C). Although we could not determine the proliferative ability of squamous cell carcinoma in Keap1-KD mice due to its rare incidence, the Ki-67-positive cell density was not higher in the tongue dysplasia in Keap1-KD mice than that observed in WT mice (Fig. 4C). In contrast, there was no difference in the Ki-67-positive cell density in either tongue dysplasia or squamous cell carcinoma between Nrf2-KO and WT mice (Fig. 4C). These results support the hypothesis that Nrf2 activation or deficiency does not affect the growth of cells in 4NQO-induced tumors.

In Keap1-KD mice, the cancer preventive effects seem to be due to the high expression of antioxidant and detoxifying enzymes (8, 10, 11). We hypothesized that thickening of the epithelia in the tongue and esophagus might protect epithelial cells from exposure to carcinogens. Keap1-KD mice have been shown to have hyperkeratosis in these tissues (23), and we surmised that this might be an additional mechanism to protect epithelial cells from carcinogens.

To address this hypothesis, we measured the thickness of the tongue epithelia and its keratin layers (34). We found that both the tongue epithelia and its keratin layer in Keap1-KD mice were significantly thicker than those in the WT mice (both P < 0.05; Fig. 5A and B). On the other hand, both the tongue epithelia and its keratin layer in Nrf2-KO mice was thinner than those in the WT mice (both P < 0.05; Fig. 5A and B). It has been reported that the epithelial thickness of stratified squamous epithelium is regulated by the growth of progenitor cells in the basal layer (2). Therefore, we examined Ki-67-positive cells in the tongue. The density of Ki-67-positive cells in the basal layer was significantly higher in Keap1-KD mice than in the WT mice (P < 0.05; Fig. 5C and D). On the contrary, while the Ki-67 density in Nrf2-KO mice was lower than that observed in the WT mice, it was not statistically significant (P, 0.09; Fig. 5C and D).

We recently found that Nrf2 regulates antioxidant and detoxifying enzymes and growth-related genes that provide growth advantages to cells by increased anabolism and reinforcement of metabolic reprogramming (35). Therefore, we examined the expression of glucose-6-phosphate dehydrogenase (G6pd), phosphogluconate dehydrogenase (Pgd), transketolase (Tkt), transaldolase 1 (Taldo1), and malic enzyme 1 (Me1) in the tongue. We found markedly higher expression of G6pd, Pgd, and Taldo1 in the tongues of Keap1-KD mice compared with the WT mice (Fig. 5E). These results support the idea that suppressing Keap1 increases Nrf2, which accelerates progenitor cell growth and increases epithelium thickness and may, in part, contribute to mechanical protection against carcinogens.

In this study, we found that Nrf2-KO mice are more susceptible to 4NQO-induced carcinogenesis in the tongue and esophagus than WT mice. Because 4NQO does not activate Nrf2 effectively, this result suggests that Nrf2 plays a protective role against carcinogenesis, even in uninduced conditions. Keap1 is a repressor of Nrf2 (36), and Keap1-KD or derepression of Nrf2 is found to elevate the expression of detoxifying and antioxidant enzymes in the tongue and esophagus of mice. Because Nrf2 acts as an oncogenic factor in cancer cells (37), there are controversial interpretations with regard to the cancer preventive effects observed in Keap1-KD mice. In this study, we have shown that Keap1 suppression prevents 4NQO-induced carcinogenesis in the tongue and esophagus without showing any adverse effects. The cancer preventive effects observed in the Keap1-KD mice seem to be due to the high expression of Nrf2 and the subsequent expression of detoxifying enzymes specific for 4NQO. Alternatively, thickened stratified squamous epithelium might contribute to the mechanical defense against 4NQO (Supplementary Fig. S5).

The carcinogenicity of 4NQO could be elicited by 2 pathways; 1 is the ROS-mediated oxidative DNA damage (38) mechanism, while the other is the direct DNA-adduct formation by 4-hydroxyaminoquinoline-1-oxide (4HAQO), a carcinogenic metabolite of 4NQO (27). One of the important characteristics of 4NQO is the inability to activate Nrf2, while Nrf2 is often activated by various environmental insults, including carcinogens (30). The incapability of Nrf2 activation suggests that 4NQO is not able to generate ROS sufficient to activate Nrf2, and the carcinogenicity of 4NQO is mainly mediated by the direct DNA-adduct formation by 4HAQO. Observations in the Nrf2-KO mice suggest that the basal expression of Nrf2 is critical for cancer prevention in tongue and esophagus. Indeed, the basal expression of Nqo1, Gclc, Gstp1, and Gpx2 in the esophagi of Nrf2-KO mice is significantly lower than that in WT mice.

The detoxification of 4NQO is mostly mediated by glutathione conjugation (27, 39). Of the Nrf2 target genes, Gclc increases glutathione synthesis and Gstp1 catalyzes formation of 4NQO-glutathione (39). In contrast, Nqo1 is the candidate enzyme that converts 4NQO to 4HAQO (40). Therefore, there was an alternative hypothesis suggesting that Nrf2 acts to stimulate 4NQO carcinogenesis. However, our current results clearly show that Nrf2 prevents carcinogenesis, which suggests that Nrf2 targets gene products other than Nqo1 contribute to the cancer prevention. We surmise that Gclc and Gstp1 are 2 important targets in this regard. Alternatively, Nqo1 may not convert 4NQO to a carcinogen. In human epidemiology, decreased expression of NQO1, GCLC, and GSTP1 by polymorphisms is independently related to an increased incidence of oral or aerodigestive tract cancer (41–43), which suggests that Nqo1, Gclc, and Gstp1 are critical among the Nrf2 target genes for preventing carcinogenesis.

Because loss of Keap1 function in cancer cells by somatic mutations or other mechanisms activates Nrf2 constitutively and promotes malignancy, cancers derived from the Keap1-KD mice appear to be malignant (21, 44). However, our study showed that suppressing Keap1 activity prevented 4NQO-induced carcinogenesis. Constitutive Nrf2 activation does not affect the growth and development of tongue and esophageal tumors in this model. While it has been reported that pharmacologic and transient Nrf2 activation prevents chemical carcinogenesis (7, 8), our study shows, for the first time, that constitutive Nrf2 activation by Keap1 suppression is a cancer prevention mechanism. In contrast, Nrf2 promotes Ras-mediated (45) or urethane-induced lung carcinogenesis (32). These observations show that the influence of Nrf2 activation depends on the context. Accumulated lines of evidence support the notion that suppressing Keap1 activity or inducing Nrf2 activity is effective for preventing the initial stage of carcinogenesis.

One of the most plausible explanations for the mechanism underlying how Nrf2 prevents upper aerodigestive tract carcinogenesis is that Nrf2 induces the expression of antioxidant and detoxifying enzymes and renders cells resistant to carcinogens (8, 10, 11). In addition, the mechanical defense of thickened stratified squamous epithelium may also be an important mechanism to prevent tissue exposure to carcinogens. Nrf2 regulates growth-related genes and increases the growth of progenitor cells in the basal layer, leading to thickened epithelia. In humans, the upper aerodigestive tract is covered by stratified squamous epithelium, and the growth of progenitor cells in the basal layer causes thickening of the epithelia. It is conceivable that the epithelial thickness regulates resistance to chemical and mechanical insults (2, 34). This function may be physiologically critical as the defense mechanism. Another possible mechanism is that Nrf2 may prevent the creation of the cancer microenvironment that supports metastasis in the lung (46).

In conclusion, we have shown that Nrf2-KO mice are susceptible to 4NQO carcinogenesis, while Keap1-KD mice are resistant to 4NQO-induced tongue and esophageal carcinogenesis. Because 4NQO does not activate or induce Nrf2, the importance of the basal Nrf2 activity on cancer prevention can be identified. We have also shown the increase in progenitor cells and the thickened epithelium in the tongues of Keap1-KD mice, which shows the unique protective mechanism of stratified squamous epithelium.

No potential conflicts of interest were disclosed.

Conception and design: A. Ohkoshi, T. Suzuki, T. Kobayashi, M. Yamamoto

Development of methodology: C. Mascaux, J.I. Eckelberger, W.A. Franklin, F.R. Hirsch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Ohkoshi, M. Ono

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Ohkoshi, M. Ono

Writing, review, and/or revision of the manuscript: A. Ohkoshi, T. Suzuki, T. Kobayashi, M. Yamamoto

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Yamamoto

Study supervision: T. Suzuki, M. Yamamoto

The authors thank Ms. Eriko Naganuma for technical assistance.

This work was financially supported in part by Grants-in-Aids for Creative Scientific Research and Scientific Research from JSPS, JST CREST, the Tohoku University Global COE Program for Conquest of Signal Transduction Diseases with “Network Medicine”, the NAITO Foundation, and the Takeda Science Foundation.