ΔNp63 regulates a common landscape of enhancer associated genes in non-small cell lung cancer

Distinct lung stem cells give rise to lung adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC). ΔNp63, the p53 family member and p63 isoform, guides the maturation of these stem cells through the regulation of their self-renewal and terminal differentiation; however, the underlying mechanistic role regulated by ∆Np63 in lung cancer development has remained elusive. By utilizing a ΔNp63-specific conditional knockout mouse model and xenograft models of LUAD and LUSC, we found that ∆Np63 promotes non-small cell lung cancer by maintaining the lung stem cells necessary for lung cancer cell initiation and progression in quiescence. ChIP-seq analysis of lung basal cells, alveolar type 2 (AT2) cells, and LUAD reveals robust ∆Np63 regulation of a common landscape of enhancers of cell identity genes. Importantly, one of these genes, BCL9L, is among the enhancer associated genes regulated by ∆Np63 in Kras-driven LUAD and mediates the oncogenic effects of ∆Np63 in both LUAD and LUSC. Accordingly, high BCL9L levels correlate with poor prognosis in LUAD patients. Taken together, our findings provide a unifying oncogenic role for ∆Np63 in both LUAD and LUSC through the regulation of a common landscape of enhancer associated genes.

analysis of typical enhancers need to be supported by data.
There is very trivial difference of the signal intensity between control and ΔNp63 in Fig.6D and E. The change in Fig.5H is also moderate. Do these changes meet both the statistical test (e.g., certain q value?) as well as certain fold change?
Also, I do not understand the motif bar in Fig.5G. ΔNp63 motif is just around 20-30 bp, and why the motif bar is as long as 5kb?
The authors focused strongly on KRT5 and BCL9 genes, but both of them ranked quite low in the super-enhancer list. I understand that they were discovered initially from Fig.7A, but again, the result of Fig.7A needs to be re-visited. Moreover, as mentioned above, the IGV tracks do no show evident change of peak intensity of BCL9 in either cell type.
Reviewer #2: Remarks to the Author: In the present manuscript the authors find that ΔNp63 plays a pivotal role in the maintenance of stemness in the normal lung and in driving an oncogenic transcriptional program. This takes place through the formation of specific chromatin structures (super-enhancers) involving specific basal cell identity-genes. The authors find that this mechanism is maintained both in mouse KRASdriven LUAD as well as in lung AT2 and basal cells. These are progenitor cells with stem-like properties, necessary for normal lung regeneration and thought to be cells-of-origin of LUSC and LUAD, respectively. The authors also translate their findings in human datasets of LUAD and LUSC, and in human cell lines, finding KRT5, ETV5 and BCL9L genes as important stemness genes regulated by ΔNp63. In particular, in vitro they find that BCL9L and KRT5 are ΔNp63 targets critical for the maintenance of LUSC, while BCL9L and ETV5 are ΔNp63 targets critical for the maintenance of LUAD. The authors further demonstrate that BCL9L depletion both in LUAD and in LUSC cell lines reduces tumorigenesis in vivo in mice xenografts. Finally, the authors find a prognostic role for BCL9L expression both in LUAD and LUSC human casuistries. The work is rather novel, the experiments are well performed and the related results are of high quality. The listed below specific comments need to be addressed for further evaluation in Nature Communications: 1. In figure 1A and 1B, the authors need to indicate how many mice were used for the analysis of tumor lesions. Same comment for Figure 1G, H and in general for all the figures in which the number of samples is not indicated. 2. In figure 1I, authors only describe colony assay for H1944 and H2009. They should mention also H358. 3. For a better evaluation of colony assay experiment in the different cell lines (Fig. 1I, J, K), would it be possible to indicate the percentage of colonies respect to relative control instead of absolute number of colonies? 4. In describing Figure 3, the authors mention Ac-tubulin and Mucin5 as markers of ciliated and goblet cells, respectively, and they also mention other differentiation markers. However, in the figure 3 panels there are no other differentiation markers analyzed. Authors need to solve this issue. 5. Page 10, lines 197-200 are not clear. What does it mean that the proliferation of the tracheal epithelium at 1 month after ΔNp63 depletion is not due to compensatory de-differentiation of ciliated and goblet cells? From the immunofluorescence experiment shown in fig. 3B-M, after epithelial damage there is loss of differentiated cells in ΔNp63 depleted epithelium However, the sentence "the proliferation of ΔNp63Δ/Δ;RosaΔ/Δ tracheal epithelium at the 1-month time point post adenoviral cre injection (see Fig. 2j) is primarily due to basal cell proliferation and not a compensatory dedifferentiation of the ciliated or goblet cells." is not clear. 6. For Figure 4A-D, authors should provide some representative images, even in Supplementary material. In the legend relative to figure 4E, authors should explain how do they mark the tracheosphere (EdU)? Same comment for figure 4T: what the red signal stands for? 7. In figure 5G and 5H, it seems unrealistic that a ΔNp63 motif would be 1-7kb long. Maybe the scale of the picture is wrong? Moreover, a different acetylation pattern between control cells and ΔNp63Δ cells seems striking only for KRT5 and BCL9L genes in tracheal cells (Fig. 5G, H), while in distal lung cells it is difficult to appreciate a difference in the acetylation patterns of ETV5 and BCL9L genes with or without ΔNp63. This needs to be mentioned and experimentally clarified. 8. The authors need to analyze the presence of a positive correlation between ΔNp63 and BCL9L in LUAD and LUSC cohorts. 9. Rescue experiments in vitro showing that BCL9L overexpression in ΔNp63 cells may rescue oncogenic properties (stemness, anchorage-independent growth etc) are required.
Reviewer #3: Remarks to the Author: Summary In this manuscript, Napoli et al. dissect the role of DNp63 in NSCLC, including adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC). The conditional allele that enables specific ablation of DNp63 in vivo and in vitro is a major strength of the study. Overall, the data associated with LUSC and its cell(s) of origin is convincing and either confirms or extends previous observations about the role of DNp63 in squamous cell carcinoma arising in different tissues. In particular, the analysis of DNp63 in tracheal basal cells is quite intriguing. However, the data on LUAD do not fully support the authors' conclusions and need to be expanded to adequately delineate the role of DNp63 in this disease.
Major points 1. The data in Figure 1A clearly support a role for DNp63 at some stage of Kras-driven lung tumorigenesis. However, there are concerns with other panels in this figure as outlined below.
2. Figure 1C. The IHC for DNp63 shown does not support the statement "Tumors from KrasG12D/+ mice stained positive for ΔNp63". The signal to noise is quite low in this image and it is difficult to believe that all tumor cells are positive. If a subset of cells are positive, it is hard to distinguish them. Additional characterization and precise description of DNp63 expression in control tumors is needed. Options could include: using a different antibody for IHC (such as the monoclonal antibody BC28, which has been used in multiple publications on human and mouse FFPE tissue); including positive control normal cells (such as basal cells) from the same slide to illustrate relative staining levels; and evaluation of recent single cell analyses of K and KP tumors (DOI: 10.1016/j.ccell.2020.06.012) to quantitate expression of DNp63 in those datasets. Since most human LUAD are thought to be DNp63-negative (with rare exceptions), characterization of DNp63 expression in this model needs to be very clear, particularly if the conclusion is that DNp63 expression in the mouse model of LUAD differs substantially from human LUAD.
3. Figures 1I/J-M, V: siRNA and shRNA experiments are difficult to interpret in the absence of information (immunoblotting) on baseline expression of DNp63 and the reduction in protein levels. (Specific nuclear positivity is difficult to identify in the IHC in 1P/T.) The use of a single shRNA against DNp63 also makes the data less robust. At a minimum, it would be nice to know if this shRNA has no effect in cell lines lacking DNp63. Figures 2U-B' do not seem to be described in the manuscript. 5. Figure 4. Loss of DNp63 expression after Ad-Cre is not documented anywhere in the figure. This is particularly important for AT2/BASC studies, as these cells are not anticipated to have high levels of DNp63 at baseline. Relative levels of DNp63 in the different cell types are important to document.

4.
6. "BCL9L is critical for maintenance of LUSC and LUAD" is a very broad claim to make when a single cell line of each tumor type has been evaluated. More cell lines and/or mining of public data (like DepMap) are needed. Were these experiments performed in the other LUAD cell lines from figure 1, all of which appear sensitive to DNp63 knockdown? Description of cell line genetics is warranted (how does driver mutation status compare with the GEMM?). As in figure 1, the use of a single shRNA and the absence of immunoblotting for target proteins is an issue. There is also no data directly demonstrating that BCL9L mediates the effects of DNp63 (rescue of shDNp63 by exogenous BCL9L, for example).
Minor points 1. "we assessed a greater than 90% recombination in the trachea and distal lung assessed by GFP expression from the ROSA allele". This is not a very precise description…90% of all cells? Or of all epithelial cells? It should be described in more detail and documented if essential to the manuscript. Or it could potentially be removed.
2. Figure 1C. I have no doubt that there were tumors in the KrasG12D model, but the H&E image here contains a cluster of lymphocytes. A different image of a tumor should be taken.
3. Figure 1D. The cells here appear to mainly include a junction between an airway and alveoli. Ideally, Figures 1C-F would compare tumors of the same grade (grade 1?) from each genotype so that there is an apples to apples comparison. Depicting AAH from each genotype would also be helpful in understanding which stage of tumorigenesis DNp63 expression is most readily detectable.
4. "…suggesting that ΔNp63 serves to maintain the proliferation of distal lung stem cell populations (Fig. 1g,h)." This conclusion is not really supported by the observation. Figure 5G/H -these seem to depict H3K27ac signal mainly at promoters -are there known enhancers for these genes, either in the depicted region or elsewhere.

5.
6. Although it may be beyond the scope of a revision, if the authors have done any experiments with Cre driven by cell type-specific promoters (SPC-Cre, CCSP-Cre), this would be informative.

Response to Reviewers' Comments:
We would like to thank the reviewers for their constructive comments and suggestions. We have addressed all the points raised by the reviewers and believe that the manuscript is of high significance to the cancer stem cell and super-enhancer fields. We now provide additional evidence of the oncogenic activities of BCL9L as a ΔNp63 regulated super-enhancer associated gene. In this resubmission, we include data demonstrating the regulation of BCL9L expression by ΔNp63 in both human lung adenocarcinoma and lung squamous cell carcinoma cell lines. More importantly, we also unveiled the crucial role of BCL9L as a mediator of the pro-tumour functions of ΔNp63 in non-small cell lung cancer. Indeed, overexpression of BCL9L is sufficient to rescue the colony formation caused by the downregulation of ΔNp63 in a panel of human lung cancer cells. Our specific point-by-point response to each comment is below in boldface type.

Reviewer #1 (Remarks to the Author):
In this manuscript, the authors tested the function of ΔNp63 by knocking it out in a Kras-driven GEMM LUAD model as well as a xenograft model of LUSC. They found decreased tumor development and growth in both models. They went on and investigated ΔNp63 function in normal lung basal cells and AT2 cells, which are proposed cell-of-origins for LUSC and LUAD, respectively. They also studied the epigenetic changes after deleting of ΔNp63 in these two cell types. Overall, the study was systematic and revealed new insights into the role of ΔNp63 in both lung cancer types as well as the cell-of-origin.
We would like to thank the reviewer for deeming our study systematic and appreciating the new roles for DNp63 in lung cancers and their cells of origin that we unveiled.
However, careful review found a number of weaknesses and major concerns from experimental design, data interpretation, as well as scientific rigor, as outlined below:   As pointed out by the reviewer, we stated in the methods that CMV-cre adenovirus was administered intratracheally. Given the nature of the promoter used, multiple cell types may undergo recombination beyond the epithelial cells, including stromal and immune system cells. It is important to note that our in vitro data using isolated basal cells, AT2, and BASC cells clearly demonstrate that ΔNp63 is required for self-renewal and terminal differentiation of these lung cell populations in a cell autonomous manner (see Figure 4); we cannot completely exclude that any additional non-autonomous effects may take place in vivo, which will be the focus of future studies.
2) Relatedly, the authors took a GEMM to study ΔNp63 for LUAD, but only investigated the LUSC subtype using a xenograft model, which is inadequate and hard to compare. It would add more enthusiasm and value to this work if a similar GEMM is utilized for LUSC.

4) It is known that
ΔNp63 is expressed at low levels of LUAD tumors. Thus, for the human LUAD cell line experiments (Fig.1i), western blot validation is required.       Indeed, low-cell number ChIP-seq was performed due to the technical challenge of obtaining primary cells from mice, especially AT2 cells, that express ΔNp63 at low levels. Therefore, the ∆Np63 regulated regions were identified by searching for the ΔNp63 motif in the H3K27ac peaks as shown in Fig. 5g,h

8) In the pathway analysis (Fig.5B), instead of raw p value, multi-test correction is required.
We thank the reviewer for raising this point. We have now clarified in the methods that the "enriched pathways were determined using the hypergeometric distribution, with significance achieved for FDR-adjusted P < 0.05." The list of pathways and their respective

10) Another important question is, what is the rationale to restrict the analysis on super-enhancer
only? ΔNp63 is known to regulate both typical-and super-enhancers. The reason to exclude the analysis of typical enhancers need to be supported by data.
We would like to clarify that we never excluded any typical enhancers from our analyses.
To avoid any confusion, we have now revised the manuscript specifying that we focused on the top 2000 ΔNp63-regulated enhancers in basal cells and AT2 cells. Fig.6D and E. The change in Fig.5H is also moderate. Do these changes meet both the statistical test (e.g., certain q value?) as well as certain fold change?

11) There is very trivial difference of the signal intensity between control and ΔNp63 in
The difference in the intensity of all the H3K27ac peaks considered in the manuscript was always significant (q < 0.05). This information has now been added to the methods. The  Fig.5G. ΔNp63 motif is just around 20-30 bp, and why the motif bar is as long as 5kb?

12) Also, I do not understand the motif bar in
We apologise for the confusion caused by our wording. Yes, a transcription factor motif is only a few base pairs long, and in Fig. 5c we report the ΔNp63 motif that we identified through our ChIP-seq data. In the tracks shown in Fig. 5g,h and Fig. 6d,e we just wanted to highlight which among the H3K27ac peaks also contained this ΔNp63 motif. To clarify that, we have now corrected these figures by labelling these regions as "H3K27ac peaks w/ ΔNp63 motif".
13) The authors focused strongly on KRT5 and BCL9 genes, but both of them ranked quite low in the super-enhancer list. I understand that they were discovered initially from Fig.7A, but again, the result of Fig.7A needs to be re-visited. Moreover, as mentioned above, the IGV tracks do no show evident change of peak intensity of BCL9 in either cell type.
As mentioned in our answer to point 11, the difference in the intensity of these H3K27ac signals had a q value below 0.05 and is statistically significant. More importantly, in our previous version of the manuscript, we showed that ΔNp63 regulates the expression of these genes both in mouse and human cells (see Fig. 5f, Fig. 7d,e) by promoting the H3K27ac levels at the enhancers of these genes (see Fig. 7g,h). Finally, since we identified BCL9L as a novel  Data are mean ± SD, n = 3, **** vs. siNT, P < 0.001, two-tailed Student's t test.

Reviewer #2 (Remarks to the Author):
In the present manuscript the authors find that ΔNp63 plays a pivotal role in the maintenance of stemness in the normal lung and in driving an oncogenic transcriptional program. This takes place through the formation of specific chromatin structures (super-enhancers) involving specific basal cell identity-genes. The authors find that this mechanism is maintained both in mouse KRAS-driven We would like to thank the reviewer for considering our work novel and for the suggestions provided to improve it.

The listed below specific comments need to be addressed for further evaluation in Nature
Communications: 1) In figure 1A and 1B, the authors need to indicate how many mice were used for the analysis of

2) In figure 1I, authors only describe colony assay for H1944 and H2009. They should mention also H358.
We thank the reviewer for pointing it out, and we have now corrected the paragraph relative to this figure (now listed in the revised manuscript as Fig. 1e). (Fig. 1I, J, K fig. 3B-M, after epithelial damage there is loss of differentiated cells in ΔNp63 depleted epithelium However, the sentence "the proliferation of ΔNp63Δ/Δ;RosaΔ/Δ tracheal epithelium at the 1-month time point post adenoviral cre injection (see Fig. 2j) is primarily due to basal cell proliferation and not a compensatory dedifferentiation of the ciliated or goblet cells." is not clear. Fig. 2j (now relabelled as Fig. 2f) Fig. 4e, that were previously pseudocolored in red, are now pseudocolored in green to match their respective endogenous fluorescence. Additionally, we now labelled the panels accordingly, as seen below in Figure 7, which is also included in the manuscript as new Fig. 4e and 4t.

7) In figure 5G and 5H, it seems unrealistic that a ΔNp63 motif would be 1-7kb long. Maybe the scale of the picture is wrong? Moreover, a different acetylation pattern between control cells and
ΔNp63Δ cells seems striking only for KRT5 and BCL9L genes in tracheal cells (Fig. 5G, H), while in distal lung cells it is difficult to appreciate a difference in the acetylation patterns of ETV5 and BCL9L genes with or without ΔNp63. This needs to be mentioned and experimentally clarified.
We apologize for the confusion caused by our wording in describing the presence of the ΔNp63 motif in these tracks. We have now clarified in the text that we are indicating which and BCL9L both in mouse AT2 (see Fig. 6c) and human cancer cells (see Fig. 7c,d) by promoting the H3K27ac levels at the enhancers of these genes (see Fig. 7h,g), thus validating our ChIP-seq data.

8) The authors need to analyze the presence of a positive correlation between ΔNp63 and BCL9L
in LUAD and LUSC cohorts.

Even though it was not possible for us to look for this correlation in LUAD and LUSC
patients given the difficulty to discriminate between ΔNp63 and TAp63 in the TCGA datasets due to the insufficient number of reads in the isoform-specific exons of TP63, we tested for this correlation in vitro. As shown in Figure 6 above, which is now included in the manuscript as new Supplementary Fig. 3a,e, downregulation of ΔNp63 in a panel of human LUAD and LUSC cell lines decreases the expression levels of BCL9L, thus providing additional data demonstrating that BCL9L is a novel ΔNp63 target gene.

9) Rescue experiments in vitro showing that BCL9L overexpression in ΔNp63 cells may rescue oncogenic properties (stemness, anchorage-independent growth etc) are required.
To address the reviewer's request, we performed soft agar assays with 3 LUAD and 2 LUSC cell lines, where we downregulated ΔNp63 via siRNA and overexpressed BCL9L. As shown in Figure 8, which is also included in the manuscript as new Fig. 8c and new Supplementary   Fig. 4a,c, the reduced anchorage-independent growth due to the reduced levels of ΔNp63 was rescued by the overexpression of BCL9L, indicating that BCL9L is sufficient to mediate the oncogenic properties of ΔNp63 in both non-small cell lung cancer subtypes. Notably, when we performed a western blot analysis in the same cells used for the soft agar assays shown in Figure 8, we found that the protein levels of BCL9L decreased in all cell lines when ΔNp63 was downregulated, further supporting our data of the regulation of BCL9L expression by ΔNp63 (see Figure 9, which is also included in the manuscript as new Supplementary Fig. 4d,e).  H358   +  +  ----+  +  +  -+  --+  -+   H1944   +  +  ----+  +  +  -+  --+  -+   H2009   +  +  ----+  +  + - We would like to thank the reviewer for deeming our data as intriguing and for providing constructive suggestions to improve our manuscript.

Major points
1) The data in Figure 1A clearly  We really appreciate the suggestion of using the BC28 antibody to detect ΔNp63 in Krasdriven LUAD via immunohistochemistry. We were able to more clearly show that ΔNp63 is expressed in these lesions and have now replaced the previous version of Fig. 1c with the new data obtained with the BC28 antibody and shown here as Figure 10.  . 1e) and LUSC cells (see Fig. 1f-r). The western blot analysis for ΔNp63 (now included in the manuscript as new Supplementary Fig.   1b,c) is also shown above as Figure 2.
3) Figures 2U-B' do not seem to be described in the manuscript.
We thank the reviewer for pointing that out. We have now removed Fig. 2u As requested by the reviewer, we extended our analysis to a group of 3 LUAD and 2 LUSC cell lines, all of which express Kras G12D/+ as now stated in both the results and the methods of the revised manuscript. As shown in Figure 8, which is also included in the manuscript as new Fig. 8c and new Supplementary Fig. 4a,c, the overexpression of BCL9L is capable of rescuing the reduced growth in soft agar assays due to the reduced levels of ΔNp63, thus indicating that BCL9L is an important oncogene mediating the oncogenic properties of ΔNp63 in both LUAD and LUSC. Additionally, as mentioned in our reply to point #2, we have included the data showing the downregulation of ΔNp63 with 2 independent sequences (see Figure 2 above, also included in the manuscript as new Supplementary Fig. 1b,c), the most effective of which was used in the 5 cell lines used for these rescue experiments and caused a reduction in BCL9L mRNA (see Figure 7, which is also included in the manuscript as new Fig. 8c and new Supplementary Fig. 4a,c) and protein levels (see Figure 9, which is also included in the manuscript as new Supplementary Fig. 4d,e).  2) Figure 1C. I have no doubt that there were tumors in the KrasG12D model, but the H&E image here contains a cluster of lymphocytes. A different image of a tumor should be taken. Figure 1D. We have now replaced the previous panels with a representative image of the IHC for ΔNp63 (BC28 antibody) in lung lesions from Kras G12D/+ mice (see Figure 10 above, which is also included in the manuscript as new Fig. 1c).
3) "…suggesting that ΔNp63 serves to maintain the proliferation of distal lung stem cell populations (Fig. 1g,h)." This conclusion is not really supported by the observation.
We have now clarified in the text that the results are "suggesting that ΔNp63 is required for the maintenance of the distal lung stem cell populations". Figure 5G/H -these seem to depict H3K27ac signal mainly at promoters -are there known enhancers for these genes, either in the depicted region or elsewhere.

5) Although it may be beyond the scope of a revision, if the authors have done any experiments
with Cre driven by cell type-specific promoters (SPC-Cre, CCSP-Cre), this would be informative.
We agree with the reviewer that it would be informative to perform some of the experiments described in our manuscript with cell type-specific cre-drivers and they will be the focus of future studies. 3) In the response to question #7: "ΔNp63 regulated regions were identified by searching for the ΔNp63 motif in the H3K27ac peaks as shown in Fig. 5g,h and Fig. 6d As stated in our reply to point #2, we followed the reviewer's suggestion and replaced the "H3K27ac peaks w/ ΔNp63 motif" tracks with the tracks of the ΔNp63 ChIP-seq signal.
These data shown above in Figure 1 and also included in the manuscript as updated Fig. 5g,h and Fig. 6e,f indicate that ΔNp63 binds to these chromatin regions in ΔNp63 fl/fl ;Rosa M/M basal and AT2 cells, similarly to what is shown in two distinct human cancer cell lines (Fig. 7c-h). Fig.5b is still showing "p value" instead of "FDRadjusted p-value". It needs to be updated.

4) Regarding the question #8, the revised
We thank the reviewer for pointing that out. We have now corrected the x-axis of Fig. 5b accordingly.

5)
In the top500 enhancers, regardless of the p63 motif, they are all significantly decreased in the KO cells, indicating that many of these changes are likely indirect effect, or even batch effect. This is even more so in Fig.6b, where all top2000 enhancers are considerably reduced, which cannot be explained by the direct effect of p63 regulation. How many replicates were used for the ChIP-Seq? And the robustness of the correlation between the replicates?
The ChIP-seq was performed by pooling either primary basal or AT2 cells collected from at least 5 ΔNp63 fl/fl ;Rosa M/M mice, thus compensating for any possible variability across mice and guaranteeing for the robustness of the signals. All the observed effects, including those shown in Fig.6b (now listed as Fig. 6c), are associated to ΔNp63. Even though we cannot rule out all indirect effects due to the loss of ΔNp63, the targets validated via qRT-PCR (Fig. 5f,   Fig. 6d, and Fig. 7c-e) and those validated via ChIP assays (Fig. 7f-h) are directly regulated by ΔNp63. The ChIP-seq data for both the basal and the AT2 cells have been deposited to NCBI Gene Expression Omnibus (GEO) repository (series GSE131671), and can be accessed by using the following token: qbqpysoqnpsdpyh.

6) In line 294 "Genes associated with the top 2,000 enhancers regulated by ΔNp63 are involved
in epithelialization and cell junction maintenance (Fig. 5e,f).".
How was this conclusion derived? Fig. 5e,f only highlighted a few genes. However, based on the new Fig.7a, these 2000 enhancers are associated with almost 8,000 genes. A more global analysis is needed for this conclusion. Otherwise, the authors should revisit this statement.
We have now revised the sentence as suggested by the reviewer to clarify that "among the genes associated with the top 2,000 enhancers in basal cells, there are genes involved in epithelialization and cell junction maintenance".