Keap1 and Nrf2 mutations predict shorter overall survival in patients with advanced HNSCC
Keap1 mutations and the resulting Nrf2 activations have been reported in many cancers 5,21. As an approach to exploring the tumor-associated Keap1 alterations, resulting in Nrf2 activation and chemotherapeutic resistance through CSC induction, we first investigated the presence of genomic alterations of Keap1 in a large panel of 21 distinctive cancers sequenced by The Cancer Genomic Atlas consortium (TCGA) and recently developed mutations significance method (MutSigCV), which provides a statistical metric to identify driver candidates in cancer with respect to the gene nucleotide length and the background mutations rate of each type of cancer analyzed 5,22. This analysis revealed that the Keap1 mutations occurred in several cancers, including head and neck cancer (Fig. 1A; Suppl. Figure 1A). Memorial-Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) is a platform for archiving a hybridization capture-based next-generation sequencing panel that detects protein-coding mutations and copy number alterations (CNAs) and selects promoter mutations and structural rearrangements in more than 410 cancer-associated genes 23,24. We explored a cohort of 186 sequentially profiled HNSCC patients for tumor-specific somatic mutation in Keap1 only (n = 1), Nrf2 only (n = 7), or both (n = 1) (Fig. 1B). Additionally, we included a third mutation in our analysis, TERT (n = 40), since TERT mutations often co-occur with Nrf2 (n = 1), but not with Keap1 (n = 0). In this cohort, we observed a marked increase in the hazard ratio (HR 4.28, p < 0.001) and a significant decrease in median survival from 29.17 months in patients with Keap1 and or Nrf2 wild-type (WT) to 15.1, 10.2, and 6.47-months patients harboring either Keap1 and or Nrf2 alone or double mutations (Fig. 1C, D). In a multivariate analysis, Keap1 and Nrf2 double mutations significantly (p < 0.001) predicted overall poor survival (Fig. 1D). In the TCGA dataset, Keap1 gene alterations are mostly missense mutations that occur in Kelch or BTB domains of Keap1 (Suppl. Figure 1B), thereby impeding Keap1 protein interaction with Nrf2. On the other hand, Nrf2 mutations are mostly missense and occur within the first 100 amino acids that contain the Neh2 domain. That includes the two degrons bound by the Keap1 and thus likely impede Keap1-mediated Nrf2 degradation 25 (Suppl. Figure 1C).
To explore the role of Keap1 mutations in HNSCC pathogenesis, and given the size of the available clinical samples, we begin by examining the Keap1 mutations in our samples (n = 24). We amplified and sequenced all five protein-coding exons of the Keap1 from 24 HNSCC surgical samples. We found 4 (17%) Keap1 mutations and 2 out of 4 mutations had novel pathogenic somatic Keap1 mutations (c.403C > T and c.1129G > A), while the remaining two mutations (c.1112G > A and c.1766A > G) had likely pathogenic and benign germ line in nature (Fig. 1E). Intriguingly, tumors with Keap1 mutations showed positive Nrf2 expression (Fig. 1E; the table below). These mutations reside in the functionally important Keap1 protein domain, such as BTB, IVR, and KR regions, governing Nrf2 ubiquitination, redox sensing, and Nrf2 binding sites (Fig. 1F). The significance of mutations was further analyzed by in silico predictors (Suppl. Table 3). We also detected five separate synonymous germline variants in various frequencies (Suppl. Table 4). The most notable variant is Keap1 c.1815G > A and is highly enriched in the HNSCC population compared to previously reported global healthy population frequencies. Prognostic analysis of patient tumors carrying Keap1 mutations revealed a significant correlation with poor DFS (p < 0.0001 by Log-rank analysis; Fig. 1G). Due to frequent alteration of Keap1 and TP53 26 in HNSCC, we were interested in whether Keap1 alterations showed association with TP53 molecular alterations. As expected, frequent TP53 overexpression (12/24, 50.0%) was detected in our cohort. Interestingly, Keap1 alterations were detected exclusively in the TP53-overexpressed HNSCC tumors (Suppl. Table 5).
Keap1 mRNA expression and concurrent Keap1 and Nrf2 mutations in Nrf2 immunopositive HNSCC tumors
Keap1 is an essential regulator of Nrf2 functions, and the role of Keap1 in regulating Nrf2 signaling in cancers has been reported previously 27. To evaluate Keap1 expression and concurrent mutations of Keap1 and Nrf2 in Nrf2 immunopositive tumors, we first analyzed Keap1 mRNA expression by qRT-PCR, followed by Nrf2 sequence analysis in Nrf2 immunopositive tumors (n = 24). As shown in Supplementary Table S6, 4 tumors with positive nuclear Nrf2 staining had absent Keap1 transcript expression, with the remaining 19 tumors being positive for Keap1 transcript. Notably, although, one tumor had absent Keap1 transcript expression (HNSCC-17, oral cavity, Suppl. Table 6) but harbored no Keap1 mutations and was also positive for Nrf2 protein expression. We then sequenced the Nrf2 gene from all tumor samples that had positive Nrf2 staining (Fig. 2A). Somatic Nrf2 mutations were found only in 2 tumors (c.145G > A and c.241G > C) including a novel Nrf2 mutation (c.145G > A) in the Neh2 domain where Keap1 binds and with high cytoplasmic Keap1expression (Fig. 2A; Suppl. Table 3). In our samples, no tumors harbored both Keap1 and Nrf2 mutations concurrently, confirming the MISK-IMPACT results in Fig. 1B. Given the smaller size of the available clinical samples, however, prognostic analysis of HNSCC carrying a Nrf2 mutation revealed a significant correlation with poor DFS (approximately ten months) (p < 0.0001 by Log-rank analysis; Fig. 2B, Suppl. Figure 2).
The biological effect of Keap1 mutations and effects of glutaminase inhibitor CB-839 in chemosensitizing Keap1 mutant cells
To evaluate the loss of Keap1 and its effects on Nrf2 overexpression and cellular localization in primary HNSCC tumors, we immunoassayed Nrf2 expression using the anti-Nrf2 antibody in HNSCC primary tumor tissues. Strong nuclear and cytoplasmic and nuclear Nrf2 expression was detected in primary tumor tissues harboring Keap1 mutations (Fig. 3A; part a). Five tumor tissues with wild-type Keap1 demonstrated decreased nuclear Nrf2 and weak cytoplasmic Nrf2 expression (Fig. 3A; parts b, c), while the normal tissue did not show Nrf2 expression (Fig. 3A, part d). We also detected nuclear Nrf2 localization in parts of Keap1-wild-type tumor tissues but to a lesser extent than Keap1- mutant tissue (Fig. 3A, part c). Aiming to study the levels of known Nrf2 target genes in tumor and normal samples, we measured the total GSH levels and enzymatic activity of SOD1, NQO1, and GST levels in eleven primary tumors and adjacent normal tissues. Among all tissues, 4 tumors (4/11; 36%) harbored Keap1 mutations, 2 (2/11; 18%) Nrf2 mutations, and remaining samples (5/11; 45%) from patients with Keap1 wild-type status (Suppl. Figure 3A). We examined the GSH, SOD1, and NQO1 enzyme activity and GST levels in tumor and normal cells and found that these enzymes were at relatively higher levels in tumor tissues than in their corresponding adjacent normal tissues. Importantly, patients carrying Keap1 and Nrf2 mutations had higher levels of GSH, SOD1, NQO1, and GST compared to wild-type (Suppl. Figure 3A).
To determine the nuclear accumulation of Nrf2, we immunostained the Nrf2 protein in Keap1 mutant SSC9 cells. The results showed nuclear accumulation of Nrf2 protein in Keap1- mutated SSC9 cells (Suppl. Figure 3B). To further examine the nuclear accumulation of Nrf2, we immunoassayed Nrf2 in Keap1 mutant and WT patients’ tumor samples and in two established HNSCC cell lines. Tumor cells with Keap1 mutations (SSC9 and Keap1-mutated patient's tumor cells) demonstrated increased nuclear localization of Nrf2 in comparison to normal and Keap1 wild-type (Cal33 and Keap1 wild-type patients’ tumor cells) cells (Fig. 3B). Since Nrf2 controls the key components of the glutathione (GSH) and tightly regulates GSH levels by directly controlling glutamate-cysteine ligase complex (GCLC) and GCLM as well as glutathione S-transferase (GST) 28,29 and thus cells acquire chemotherapeutic resistance, we investigated Nrf2-regulated target genes. Real-time RT-PCR analysis revealed that the majority of the Nrf2-regulated target genes were highly modulated in the cancer cells (Suppl. Figure 3C). In addition, drug resistance markers MDR1 and ABCG2 were also highly upregulated in the cancer cells but not in the normal cells (Suppl. Figure 3C). Importantly, as expected, the Keap1 transcripts and proteins were downregulated in Keap1 mutated cells (Suppl. Figure 3C).
Several recent studies have reported that loss of Keap1 alters cellular metabolic requirements and confers sensitivity to glutamine metabolism inhibitors 30–32. These reports suggest that glutathione (GSH) production is increased by glutamine metabolism. In lung cancer cells, it was shown that glutaminase inhibition can sensitize the radiation-induced treatment resistance in the Keap1 and Nrf2 mutant cells 30. We therefore aimed to determine whether loss of Keap1 preferentially chemosensitizes by targeting glutaminase metabolism. First, we analyzed publicly available RNA-sequence data set GSE112026 and found significant overexpression of genes involved in glutamine metabolism (Fig. 3C). Next, we explored the possibility if targeting glutamine metabolism can chemosensitize Keap1 mutant cells. We used a combination of chemotherapeutic agent cisplatin and a small-molecule glutamine inhibitor CB-839, which is currently under investigation in phase 1 and 2 clinical trials. Our results showed that although treatment of cells with CB-839 alone did not show significant sensitivity to CB-839 in Keap1 mutant SSC9 and Keap1 wild-type Cal33 cells. However, the combination of cisplatin and CB-839 significantly increased the sensitivity to combination treatment and killed a substantial number of cells in Keap1 mutant SSC9 cells (Fig. 3D). In addition, the combination treatment significantly abolished the sphere growth efficiency in Keap1 mutant SSC9 cells suggesting that the combination treatment may exhibit the potential to inhibit the self-renewal capacity of Keap1 mutant cells (Fig. 3E). Furthermore, we noticed substantial inhibition of sphere growth effect in Keap1 wild-type Cal33 cells after silencing by Keap1-siRNA (Fig. 3F). To identify by which mechanisms CB-839 preferentially chemosensitize the cells, we assessed the intracellular ROS levels after treatment with CB-839 and cisplatin. The baseline ROS levels in Keap1 mutant SSC9 cells showed lower than in wild-type cells (Fig. 3G). However, unlike Keap1 wild-type cells, a combination of cisplatin and CB-839 treatment significantly increased the ROS levels in Keap1 mutant SSC9 cells compared to cisplatin alone treatment (Fig. 3G). Additionally, the CB-839 treatment significantly reduced the GSH activity in Keap1 mutant SSC9 cells (Fig. 3H). These results suggest that CB-839 treatment preferentially follows the inhibition of free radical scavenging capacity in Keap1 mutant cells compared with Keap1 wild-type counterpart. We then tested the hypothesis that if the addition of exogenous free radical scavenger preferentially rescues the capacity of Keap1 mutant cells from CB-839-mediated chemosensitization. Our results showed that treatment of Keap1 mutant SSC9 cells with a ROS scavenger NAC did not show significant effects on cell survival by NAC alone or either 10 µM cisplatin or CB-839 alone (Fig. 3I). Importantly, treatment of cells by NAC significantly rescued the increased cell death which was caused by the combination treatment of CB-839 and 10 µM of cisplatin (Fig. 3I).
Loss of Keap1 increases the Nrf2 transcriptional activity, cancer stem cells characteristics in HNSCC
To get more insight into the Keap1 mutations and resulting chemoresistance through Nrf2 activation, we assessed the effects of siRNA knockdown of Keap1 on the sensitivity of Cal33 tumor cells to cisplatin. Keap1 siRNA was transfected into Cal33 cells under the treatment of 10 µM cisplatin. On days 1 and 2, siRNA against Keap1 steadily reduced Keap1 mRNA and maintained at this level for two days and again increased but stayed below baseline levels on days 3 and 4, while untreated control and scrambled-siRNA treated cells retained higher levels or on the baseline level (Fig. 4A). To identify if the knockdown of Keap1 also activates the transcriptional activity of Nrf2 target genes, we assessed SOD1 expression following Keap1 siRNA knockdown. Keap1 knockdown substantially increased the expression of SOD1 transcript by 6.2 and 5.1-fold on days 2 and 3 (Fig. 4B). In contrast, no significant change in SOD1 transcript was observed in control and scrambled siRNA-treated cells (Fig. 4B).
Next, we tested the cisplatin sensitivity in Keap1 siRNA-transfected cells. Untreated control and scrambled-siRNA-treated cells showed sensitivity to cisplatin. Although, Keap1 siRNA treated cells showed resistance at a lower dose but showed sensitivity to cisplatin at higher doses (Fig. 4C), and a higher EC50 value was recorded in Keap1-siRNA (the EC50 to cisplatin was 32.3 µM for Keap1 siRNA-transfected cells and 7.2 µM and 8.7 µM for control and scrambled-siRNA respectively. (p < 0.05 displays the differences between Keap1 siRNA and control and scrambled groups).
Next, we sought to examine if the alterations of Keap1 show any associations with chemotherapy resistance in HNSCC cells. To establish the functional differences in sensitivity to chemotherapy in Keap1-WT and Keap1-mutant cells, we examined whether loss of Keap1 showed any effect on cell survival. Treatment of cells with cisplatin showed that Keap1-mutant SSC9 and patients’ primary tumor cells had resistance to cisplatin compared to Keap1 wild-type cells (Fig. 4D). To explore whether the restoration of Keap1 in Keap1-mutant cells affects cancer cell growth, we established a clone of SSC9 cells that stably express Keap1 cDNA. The results showed the expression of Keap1 mRNA in control and Keap1 clone cells while the absence of Keap1 mRNA in parental SSC9 cells (Fig. 4E). We further assessed changes in the expression of Nrf2 target genes SOD1 and NQO1 in the Keap1-clone cells. We found that restoration of Keap1 expression eliminated the SOD1 and NQO1 gene expression (Fig. 4F). In addition, we compared the cell proliferation activity in the Keap1-expressing clone. We found that Keap1 clone cells grew comparatively slower than the parental and mock-transfected control cells (Fig. 4G). Next, we examined the cisplatin sensitivity by reintroducing the Keap1 clone in parental SSC9 cells. Keap1-expressing clone demonstrated poorer cell survival after cisplatin treatment for 72 hours than parental and control cells (Fig. 4H). Furthermore, loss of Keap1 showed extraordinary self-renewal capacity in Keap1 mutant cells (Fig. 4I). Consistent with baseline differences in cell growth, Keap1 expressing clones showed poorer tumorsphere formation after cisplatin treatment with clear contrast in parental and control cells (Fig. 4J), suggesting additional evidence of therapeutic resistance in HNSCC through self-renewal of the tumor cells.
Next, we examined the level of a well-known cancer stem cell (CSC) marker CD44 mRNA by qRT-PCR in chemoresistant (n = 13) and sensitive (n = 11) HNSCC patients’ samples. Twelve chemoresistant (54.17%) out of twenty-four patients treated with chemotherapy showed higher expression of CD44 (> 50%) compared with the corresponding chemosensitive group (Fig. 4K). In addition, mutation analysis (review Figs. 1 and 2) led us to find Keap1 mutations in four cases (17%) and Nrf2 in 2 cases (8%). On the basis of a positive aberration score for the mutations of Keap1 and Nrf2 and/or CD44 expression, we assigned 24 cases to two groups: 13 chemoresistant cases with 2 to 3 aberration scores to a “high group” and 11 chemosensitive cases with 0 and 1 aberration score to a “low score group” (Fig. 4L). Interestingly, the high score group (chemoresistant group) had worse DFS (Fig. 4M; Log-rank P < 0.0001). These results suggest that a fraction of patients treated with chemotherapeutic agents experience resistance to treatment, enhancing the CSC marker CD44 expression in addition to Keap1 mutations and Nrf2 activation in HNSCC and impacting the patients’ overall treatment outcome.
Knockdown of Nrf2 in Keap1 defective cells leads to activation of ROS-mediated stress pathway and enhances the chemosensitivity
Next, we assessed the role of Nrf2 activation and chemoresistance in Keap1-mutant SSC9 cells. First, we silenced Nrf2 expression by siRNA for Nrf2 and assessed chemosensitivity. Silencing Nrf2 by siRNA significantly reduced the endogenous Nrf2 expression and activity in SSC9 cells (Fig. 5A). Cells treated with Nrf2 siRNA showed increased sensitivity to cisplatin treatment in comparison with control siRNA-treated cells (Fig. 5B), concomitant with the decrease in cell proliferation in Keap1-mutant SSC9 cells (Fig. 5C; p < 0.001). In addition, we tested cisplatin sensitivity in Cal33 cells (Keap1 WT) and observed high sensitivity to cisplatin in Nrf2 knockdown cells (Fig. 5D; p < 0.05).
Cisplatin induces intrinsic apoptosis by producing mitochondrial ROS 33 leading to the induction of apoptosis. Furthermore, various antioxidant enzymes are induced by Nrf2 activation and reduce the intracellular ROS level, therefore, resulting the cells becoming more resistant to chemotherapies 19. Moreover, Nrf2 directly affects the homeostasis of ROS by regulating the antioxidant defense system 34. Nrf2-mediated chemotherapeutic resistance is likely to occur due to the reduction of drug-induced ROS. Considering these, we hypothesized that Nrf2-induced chemotherapy resistance might be partially due to the decrease in drug-induced ROS generation. To address this question, we treated mock and Keap1-expressing SSC9 clones with cisplatin and assessed the mitochondrial ROS production using a fluorescent indicator. At the same time, we analyzed the ROS level in Nrf2-siRNA-treated SSC9 cells. We observed that both the Keap1-expressing clone and Nrf2 knockdown cells showed higher ROS generation under the treatment of cisplatin (Fig. 5E, F), suggesting that increased proliferation (Fig. 5C) was seen in Keap1-mutated cells is largely mediated by Nrf2. To further demonstrate that Nrf2 activation contributes to the increased expression of anti-oxidants, and xenobiotic metabolism enzymes, we challenged cells with Nrf2 siRNA in SSC9 cells. Transfection of Nrf2siRNA in cells decreased the Nrf2 mRNA by 70–75% with the reduction of Nrf2 target genes (Fig. 5G). Conversely, inhibition of Keap1 expression by siRNA increased the Nrf2 target genes in Cal33 cells (Fig. 5H). In addition, in vitro sphere formation assay revealed that Keap1-expressing cells decreased the growth of spheres by 1.5-fold compared with parental SSC9 cells under the cisplatin treatment condition (Fig. 5I). The growth of the sphere continued to decrease further in the secondary sphere culture approximately by 1.5-2.0-fold (Fig. 5I). In addition, Nrf2 knockdown impaired tumorsphere growth in these cells confirmed that overexpression of Nrf2 contributes to a stem-like phenotype in HNSCC cells (Fig. 5J). These data indicate that loss of Keap1 in HNSCCs leads to increased cell proliferation, and expression of Nrf2, as well as sphere growth efficiency, suggesting that Nrf2 activation and decrease of ROS and chemotherapeutic resistance is the vital mechanistic mediator observed in loss of Keap1.
Keap1 mutations and Nrf2 overexpression regulates Notch signaling in HNSCC cells
Our clinical results indicate that Keap1 mutations are strongly associated with chemo-radio resistance. However, patients with Keap1 mutations developed tumor regrowth in the lung. Furthermore, loss of Keap1 and Nrf2 activation has previously been reported to confer resistance to chemotherapy 36,37. Recent studies have reported that the Notch target genes show direct downstream transcriptional mediators of Nrf2 signaling 38–41. Moreover, previous studies have reported that activation of the Notch signaling enhances self-renewal of oral squamous cell carcinoma cancer cells, while its loss impairs the maintenance of self-renewal 42. Therefore, we hypothesized that the Keap1-Nrf2 pathway likely modulates activation of the Notch pathway. To explore this, we first measured the expression of Notch1 and Notch target genes in Keap1-expressing clone cells. Notch1 and Notch target genes prominently decreased in Keap1-expressing cells (Fig. 6A), while their expression significantly increased in Keap1 mutant cells. Next, we tested if the Nrf2 overexpression and Keap1 mutations may have any effects on Notch activity. The expression levels of Notch1 and Hes1 were decreased in Keap1 expressing cells compared to Keap1 mutant cells (Fig. 6B). On the other hand, Nrf2 knockdown cells expressed significantly lower levels of Notch1 and Hes1 mRNA (Fig. 6C) and protein compared to controls (Fig. 6C, D). These results suggest that Keap1-Nrf2 regulates Notch signaling in HNSCC. We obtained tumor tissues from Keap1 mutant, wild-type, and Nrf2 mutant patients’ samples and immunostained them for the expression of Notch1, Hes1, Ki67, and Nrf2. Immunostaining results confirmed the absence of Nrf2 in Nrf2 mutant tumor tissues and high levels of Keap1 mutant tumors (Fig. 6E). Ki-67, a cell proliferation marker, was highly expressed in Keap1-mutant tumors compared with those in Nrf2 mutant tumors (Fig. 6E). Expression of Notch1 and Hes1 were significantly highly expressed in Keap1 mutant tumors as compared with those of Nrf2-mutant tumors (Fig. 6E). These results indicated the functional role of the Keap1-Nrf2 pathway in regulating the cell proliferation and active involvement of the Notch signaling HNSCC tumors.
As shown in Fig. 5J and 6D, knockdown of Nrf2 significantly impaired tumorspheres and downregulation of Notch1 and Hes1. Therefore, we tested the possibility of the Notch pathway as a target for directed therapy by exploring its functional consequences in Nrf2-activated HNSCC cells with Keap1 mutations. First, we inhibited Notch1 activity by siRNA in SSC9 cells. Following transfection of cells with Notch1 siRNA, cells showed a significant decrease in cell proliferation (Fig. 6F, G), coupled with a significant reduction in two Notch pathway target genes, Hes1 and Hey1 (Fig. 6H). We then assessed whether inhibition of Hes1 also shows any impact on cell growth. Congruent with the results obtained for Notch1 inhibition in Fig. 6G, inhibition of Hes1 significantly decreased cell proliferation (Fig. 6I, J). Similarly, treatment of cells with a Notch pathway inhibitor DAPT significantly inhibited cell growth (Fig. 6K, L).
Keap1 mutation is a strong predictor of chemotherapeutic outcomes in patients with advanced HNSCC.
Given the fact that Keap1 mutations and Nrf2 overexpression lead to chemotherapeutic resistance, we, therefore, hypothesized that Keap1 mutations might lead to an increased rate of local recurrence in advanced HNSCC patients treated with chemotherapeutic agents. As described in Figs. 1 and 2, patients with Keap1 mutations were predicted to be deleterious and had a cumulative incidence of local treatment failure at ten months which was 80% in patients whose tumors carried Keap1 mutations as compared with less than 12% in patients with wild-type tumors (Log-rank p < 0.0001), while analyzing the patients with higher stages, particularly in patients with stage III-IV had higher rates of treatment failure in patients who harbored Keap1 mutations (Log-rank p < 0.0001) (Fig. 7A-C).
Combination therapy with cetuximab, paclitaxel and cisplatin led to a partial response in a patient with Keap1 mutant advanced stage metastatic HNSCC
The Keap1-Nrf2 pathway has been shown to cancer cell survival and mutations in Keap1 or Nrf2 are clinically relevant predictive biomarkers of chemo-radio resistance 30. The standard chemotherapy for advanced-stage HNSCC patients are cisplatin, 5-fluorouracil, and docetaxel/paclitaxel, and show improved progression-free and overall survival 43. Furthermore, the most common sites of distant metastases were reported to be the lung (70%) followed, by the liver (42%) and bones (15%) 44,45. However, chemotherapy resistance results in poor treatment outcomes, and the reasons for chemotherapy resistance are diverse and multifaceted. To identify the Keap1 mutations and associated chemo-radio resistance, we present two case reports that recently underwent a combination of chemotherapy and related resistance to therapy. Patient #1 is a 59-year-old male visiting the clinic with advanced metastatic squamous cell carcinoma of the laryngeal. The patient received initial radiotherapy followed by two lines of chemotherapy (cetuximab/paclitaxel/cisplatin [CPP]). The patient had rapid disease progression and underwent biopsy and genotyping of lung metastasis that revealed Keap1 mutations (Fig. 7D). Following CPP, the patient was then treated with cetuximab weekly for six months and achieved a partial response (PR) but had rapid disease progression and eventually succumbed as a consequence of the disease. Patient #2 is a 64-year-old male diagnosed with metastatic oral cavity cancer and initially treated with laryngeal-preservation surgery, followed by three cycles of docetaxel, cisplatin, and continuous infusion of fluorouracil (TPF) followed by a chemoradiation with cisplatin. The patient achieved a partial response (PR). Approximately after 9 months, recurrence was observed, and CPP was initiated. Following CPP, the patients had rapid disease progression and underwent biopsy and genotyping of lung metastasis and found Keap1 mutations (Fig. 7E). Further disease progression of oral cavity cancer was observed shortly after the completion of cycle 3. In both cases, sequencing of cells from lung metastatic site showed Keap1 and Shh mutations and strong expression of Notch1 and Hes1 confirmed the activation of the Notch pathway. Thus, among recurrent and metastatic HNSCC patients, Keap1 mutations appear to be the most significant cause of clinical chemo-radio resistance (Fig. 7F) and are coupled with the activation of the Notch signaling pathway.