Interplay between Notch1 and Notch3 promotes EMT and tumor initiation in squamous cell carcinoma

Notch1 transactivates Notch3 to drive terminal differentiation in stratified squamous epithelia. Notch1 and other Notch receptor paralogs cooperate to act as a tumor suppressor in squamous cell carcinomas (SCCs). However, Notch1 can be stochastically activated to promote carcinogenesis in murine models of SCC. Activated form of Notch1 promotes xenograft tumor growth when expressed ectopically. Here, we demonstrate that Notch1 activation and epithelial–mesenchymal transition (EMT) are coupled to promote SCC tumor initiation in concert with transforming growth factor (TGF)-β present in the tumor microenvironment. We find that TGFβ activates the transcription factor ZEB1 to repress Notch3, thereby limiting terminal differentiation. Concurrently, TGFβ drives Notch1-mediated EMT to generate tumor initiating cells characterized by high CD44 expression. Moreover, Notch1 is activated in a small subset of SCC cells at the invasive tumor front and predicts for poor prognosis of esophageal SCC, shedding light upon the tumor promoting oncogenic aspect of Notch1 in SCC.

Brief Summary: The authors present interesting data about the understanding of how tumor initiation and heterogeneity occurs in squamous cell carcinoma.The authors report a mechanism where Notch1 and Notch 3 may have an opposing function to regulate these processes both in vitro and in vivo. They purport that Notch1 activation and EMT are coupled to promote tumor initiation and cancer cell heterogeneity while the repression of Notch3 by transcription factor ZEB1 limits ICN1-induced differentiation permitting ICN1-mediated EMT. Although the manuscript is scientifically sound and the data is very nice, it appears these mechanisms have already been reported separately, and hence there may be a significant overlap between these data and recently published papers. Which might impact its relevance for NComm verses another venue. The manuscript is well written, it has important clinical message since disease progression via TGFβ-mediated EMT could offer the TGFβ-ZEB1-Notch1 axis as a potential therapeutic target in SCC and should be of great interest to the field and beyond. Overlap concern derived from the following manuscripts: Major points to address: 1. If EpCAM is overexpressed in the majority of human epithelial cancers including SCC and plays an important role in EMT. Why do you select tdTomatopos, EpCAMneg instead of EpCAMpos to evaluate the mesenchymal characteristics? If EpCAM participate in TGF-β1-induced EMT. 2. In the figure 3, the authors explain that tdTomatopos, EpCAMneg cells have mesenchymal characteristics, and then they explain (for the same figure) that complete loss of epithelial characteristics was uncommon in 4NQO-induced SCC. It is not clear which cells have mesenchymal characteristics. 3. In the figure 1g the authors explain that when p53 was deleted prior to 4NQO administration, p53 loss did not prevent ICN1 expression in 4NQO-induced lesions; it could be important highlight, differences in the expression of ICN1 between cells (p53+/+ and p53-/-) with 4NQO induction and without 4NQO. 4. In the figure 2e, it could be interesting to corroborate the mesenchymal characteristics in EN60 and TE11 cell lines with high expression of CD44 (H and L) to make a more accurate comparison. 5. In the figure 2e the authors explain that CD44H cells do not require Notch for tumor initiation, but it could add that the tumor initiation is independent of Notch because there are not differences of tumor formation with DOX or without DOX. 6. Why the authors select just TE11 cell line to evaluate if TGFβ influences Notch1-mediated tumor promotion? One cell line is not sufficient to generalize claim. 7. It could be interesting to evaluate the levels of Notch expression using anti-TGFβ blocking antibody. 8. Extension of results to clinically relevant samples such as PDX models would certainly increase enthusiasm.
Overall this is a very nice study and if the authors can address the major reservations of the manuscript the reviewer thinks it would be acceptable for publication.

Reviewer #2 (Remarks to the Author):
The immunostaining data with the chemically induced oral cancer does not does not add much new, and most of the new results are derived subsequently from a variety of cancer cell lines. Using these lines, the authors nicely demonstrate importance of the notch intracellular domain. Although they show ZEB1 repression of Notch3, the linkage between TGF-beta, ZEB1, Notch3 and notch1 is not as clearly demonstrated as shown in the final model. Given the previous extensive study of the molecules examined here, it would have been nice to examine roles for at least some of the molecules in vivo in the chemically induced model, as opposed to cell lines.
The following is a point-by-point response to the helpful and insightful comments of the Reviewers:

Reviewer #1:
"The authors present interesting data about the understanding of how tumor initiation and heterogeneity occurs in squamous cell carcinoma"…"The authors report a mechanism…both in vitro and in vivo..."…"the manuscript is scientifically sound and the data is very nice…"it has important clinical message…and should be of great interest to the field and beyond"…"Overall this is a very nice study and if the authors can address the major reservations of the manuscript the reviewer thinks it would be acceptable for publication…there may be a significant overlap between these data and recently published papers….Overlap concern derived from the following: We appreciate the positive comments from this Reviewer. With regard to the Reviewer's concern involving potential overlap with published findings, the current manuscript represents a logical extension of our earlier work (Natsuizaka et al., a meeting abstract for the AACR annual meeting and Ohashi et al., an original article in Cancer Research) which demonstrated that (1) expression of transcription factor ZEB1 is associated with EMT in esophageal squamous cell carcinoma (ESCC); and (2) Notch3 inhibition may promote EMT. The current manuscript provides new functional and mechanistic insights into the role of Notch1 in carcinogenesis. Specifically, this manuscript (1) defines a novel mechanism through which ZEB1 represses Notch3 directly to suppress squamous cell differentiation while Notch1 and TGFβ cooperate to promote EMT (de-differentiation); and (2) elucidates how the activated form of Notch1 induces CD44H cells with enhanced malignant potential; and (3) utilizes the single cell-derived 3D neoplastic organoid system as well as cell-lineage tracing experiments to document EMT, for the first time, in a mouse model of ESCC. Additionally, while Wang et al. used a single ESCC cell line to demonstrate that Notch1 influences expression of Snail and other EMT markers in the context of cell migration/invasion in vitro, this study failed to establish a mechanistic link between Notch1 and EMT, a knowledge gap that is addressed by the current manuscript. Finally, the minireview article by Li et al. does not cite any previous studies diminishing the novelty of our current manuscript. We have updated the introduction and discussion sections of the revised manuscript to better describe the studies noted by this Reviewer as they relate to the current manuscript.
Major points to address:

"If EpCAM is overexpressed in the majority of human epithelial cancers including SCC and plays an important role in EMT. Why do you select tdTomato-pos, EpCAM-neg instead of EpCAM-pos to evaluate the mesenchymal characteristics? If EpCAM participate in TGF-β1-induced EMT."
Although epithelial cancer cells express EpCAM, they may lose EpCAM expression as a result of EMT mediated by TGFβ present in the tumor microenvironment. Since tumor tissues contain non-epithelial cells (e.g. fibroblasts), EpCAM-negative cells are not necessarily cancer cells. In cell-lineage tracing experiments, epithelial cells are labeled with tdTomato prior to chemically-induced carcinogenesis. By looking at tdTomatopositive tumor cells, we can document EMT in cancer cells even if they no longer express EpCAM. We have clarified this experimental design in the revised text (Page 5, lines 8-22).

"…the authors explain that tdTomato-pos, EpCAM-neg cells have mesenchymal characteristics, and then they explain (for the same figure) that complete loss of epithelial characteristics was uncommon in 4NQOinduced SCC. It is not clear which cells have mesenchymal characteristics."
The reviewer refers to Figure 1c about the data from 4NQO-induced SCC. We defined SCC cell mesenchymal characteristics when two criteria were met: (1) cells are negative for cell surface EpCAM expression and (2) cells exhibit upregulation of mesenchymal markers including Zeb1 and Cdh2 (Figure 1c). Figure 1c). tdTomato expression assures epithelial-cell origin of SCC cells as answered for above question #1 by this reviewer. We postulated that complete loss of epithelial characteristics was uncommon in 4NQOinduced SCC because intratumoral SCC cells display heterogeneity as corroborated by continuous distribution of EpCAM expression (Supplementary Figure 1c). Mesenchymal cell markers are detectable in cells with low EpCAM expression. Such findings may be consistent with the epithelial-mesenchymal hybrid phenotype in cancer cells that has been postulated by others [Sci Signal. 2014

"In the figure 1g the authors explain that when p53 was deleted prior to 4NQO administration, p53 loss did not prevent ICN1 expression in 4NQOinduced lesions; it could be important highlight, differences in the expression of ICN1 between cells (p53+/+ and p53-/-) with 4NQO induction and without 4NQO."
We agree. In the original submission, experienced pathologists (AJK and SN) scored independently ICN1 expression in 4NQO-induced tumors from mice with or without targeted p53-deletion in the esophagus. The ICN1 labeling index (Figure 1g) was determined based on the frequency of ICN1-positive SCC cells within each tumor examined. We examined also ICN1 expression in mice with or without esophagus targeted p53deletion in the absence of 4NQO treatment; however, there was no difference noted in ICN1 expression level between p53 +/+ and p53 -/cells and that ANOVA with Tukey's post-hoc test failed to detect a significant difference (p<0.05) in the ICN1 labeling index between groups (three animals per group) when p53 +/+ and p53 -/cells were compared in normal esophageal mucosa without 4NQO treatment. As these data represent a negative result, we did not include in the original manuscript. For the revised manuscript, we now report these findings (Page 6, lines 16-20) along with a reference [Mol Cell Biol. 2007 May;27(10):3732-42. PMID: 17353266] reporting a consistent finding that p53 deficiency alone does not impair basal Notch1 expression as well as Notch1-mediated squamous-cell differentiation in mice.

"…it could be interesting to corroborate the mesenchymal characteristics in EN60 and TE11 cell lines with high expression of CD44 (H and L) to make a more accurate comparison".
In revised Figure 3a and Supplementary Figure 2a and 2b, we demonstrate that CD44H cells express mesenchymal characteristics compared to CD44L cells within the tumors for both cell lines. We now describe these observations in the revised text (Page 7, line 24-Page 8, Line 2).

"…the authors explain that CD44H cells do not require Notch for tumor initiation, but it could add that the tumor initiation is independent of Notch because there are not differences of tumor formation with DOX or without DOX".
We agree that tumor initiation by CD44H cells may be independent of Notch because purified CD44H cells gave rise to tumors with or without DOX-induced ICN1 expression or in the presence of DNMAML1, a genetic pan-Notch inhibitor (revised Figure 3c). We now state this explicitly in the revised Results (Page 8, line 15) and discussed (Page 15, lines 11-15).

"Why the authors select just TE11 cell line to evaluate if TGFβ influences Notch1-mediated tumor promotion? One cell line is not sufficient to generalize claim".
In the revised manuscript, we now demonstrate that TGF-β inhibition via the TGFβ neutralizing therapeutic antibody 1D11 suppresses growth of patient-derived xenograft (PDX) tumors from head and neck squamous cell carcinoma (HNSCC) (Figure 4a), indicating that this finding is not specific to TE11 ESCC cells and may be more generalizable to SCCs beyond ESCC.

"It could be interesting to evaluate the levels of Notch expression using anti-TGFβ blocking antibody".
We agree. In the revised manuscript, we have evaluated the effects of the TGFβ neutralizing antibody 1D11 in 4NQO-induced murine esophageal lesions (Figure 4b) where administration of 1D11 diminished significantly the ICN1 level. Additionally, 1D11 decreased the frequency of ESCC and precursor lesions and EMT (Figure 4c).

Extension of results to clinically relevant samples such as PDX models would certainly increase enthusiasm.
As answered for the concern expressed by this reviewer in the above point #6, we have analyzed the effect of the TGFβ neutralizing therapeutic antibody 1D11 on a HNSCC PDX and report impairment of tumor growth (Figure 4a) in the revised manuscript.

Reviewer 2:
"The immunostaining data with the chemically induced oral cancer does not add much new, and most of the new results are derived subsequently from a variety of cancer cell lines." Our findings using immunostaining in 4NQO-induced ESCC coupled with cell-lineage tracing experiments are the first unequivocal documentation of EMT in squamous-cell carcinomas. Additionally, ICN1 IHC data in mice and human SCC patients provides insights into Notch1 activation status as it relates to the natural history of SCC development and progression. Notch1 deletion in the murine skin results in non-cell autonomous SCC development due to inflammation associated with epidermal barrier defects [Cancer Cell. 2009 Jul 7;16(1):55-66. PMID: 19573812]. Such an approach limits the assessment of the cell-autonomous oncogenic role of Notch1 in vivo. We have utilized the three-dimensional (3D) organoid system where a tissuederived single cell suspension gives rise to spherical structures with recapitulation of the histopathology present in the original in situ neoplastic lesions. Generating 3D organoids from Notch1 loxP/loxP mice with 4NQOinduced ESCC and precursor lesions, we now demonstrate functional consequences of Notch1 deletion ex vivo to find that Notch1 plays a role in the initiation of neoplastic organoid formation (Figure 2f) and EMT (Figure 3e). We now highlight the relevance of these findings to SCC biology in the revised discussion.
"Using these lines, the authors nicely demonstrate importance of the notch intracellular domain. Although they show ZEB1 repression of Notch3, the linkage between TGFβ, ZEB1, Notch3 and notch1 is not as clearly demonstrated as shown in the final model. Given the previous extensive study of the molecules examined here, it would have been nice to examine roles for at least some of the molecules in vivo in the chemically induced model, as opposed to cell lines".
We acknowledge the Reviewer's concern regarding evaluation of the roles of TGFβ, Notch1, Notch3 and Zeb1 in vivo. In the revised manuscript, we now evaluate the influence of the TGFβ neutralizing therapeutic antibody 1D11 upon Notch signaling and EMT in the context of 4NQO-mediated carcinogenesis (Figure 4b, c; Supplementary Figure 3e), providing evidence that TGFβ supports ICN1 expression and EMT in vivo. To examine further the functional role of Notch1 in Zeb1 expression and EMT, we have utilized 4NQOinduced neoplastic 3D organoids from animals expressing Notch1 loxP/loxP for ex vivo recombination experiments (Figure 2f and Figure 3e) as described above.
In summary, we believe that the manuscript is much improved and are appreciative of the helpful comments of the Reviewers. We hope you find the revised manuscript suitable for publication.