Dual role of GRHL3 in bladder carcinogenesis depending on histological subtypes

The effect of grainyhead‐like transcription factor 3 (GRHL3) on cancer development depends on the cancer subtypes as shown in tumor entities such as colorectal or oral squamous cell carcinomas. Here, we analyzed the subtype‐specific role of GRHL3 in bladder carcinogenesis, comparing common urothelial carcinoma (UC) with squamous bladder cancer (sq‐BLCA). We examined GRHL3 mRNA and protein expression in cohorts of patient samples, its prognostic role and its functional impact on tumorigeneses in different molecular and histopathological subtypes of bladder cancer. We showed for GRHL3 a reverse expression in squamous and urothelial bladder cancer subtypes. Stably GRHL3‐overexpressing EJ28, J82, and SCaBER in vitro models revealed a tumor‐suppressive function in squamous and an oncogenic role in the urothelial cancer cells affecting cell and colony growth, and migratory and invasive capacities. Transcriptomic profiling demonstrated highly subtype‐specific GRHL3‐regulated expression networks coined by the enrichment of genes involved in integrin‐mediated pathways. In SCaBER, loss of ras homolog family member A (RHOA) GTPase activity was demonstrated to be associated with co‐regulation of eukaryotic translation initiation factor 4E family member 3 (EIF4E3), a potential tumor suppressor gene. Thus, our data provide for the first time a detailed insight into the role of the transcription factor GRHL3 in different histopathological subtypes of bladder cancer.

The effect of grainyhead-like transcription factor 3 (GRHL3) on cancer development depends on the cancer subtypes as shown in tumor entities such as colorectal or oral squamous cell carcinomas.Here, we analyzed the subtype-specific role of GRHL3 in bladder carcinogenesis, comparing common urothelial carcinoma (UC) with squamous bladder cancer (sq-BLCA).We examined GRHL3 mRNA and protein expression in cohorts of patient samples, its prognostic role and its functional impact on tumorigeneses in different molecular and histopathological subtypes of bladder cancer.We showed for GRHL3 a reverse expression in squamous and urothelial bladder cancer subtypes.Stably GRHL3-overexpressing EJ28, J82, and SCa-BER in vitro models revealed a tumor-suppressive function in squamous and an oncogenic role in the urothelial cancer cells affecting cell and colony growth, and migratory and invasive capacities.Transcriptomic profiling demonstrated highly subtype-specific GRHL3-regulated expression networks coined by the enrichment of genes involved in integrin-mediated pathways.In SCaBER, loss of ras homolog family member A (RHOA) GTPase activity was demonstrated to be associated with co-regulation of eukaryotic translation initiation factor 4E family member 3 (EIF4E3), a potential tumor suppressor gene.Thus, our data provide for the first time Abbreviations ARHGEF19, rho guanine nucleotide exchange factor 19; BASQ, basal/squamous-like; BLCA, bladder cancer; BP, biological processes; CB, cytoskeleton buffer; CC, cellular compartments; DEG, differentially expressed gene; EIF4E3, eukaryotic translation initiation factor 4E family member 3; EMT, epithelial-to-mesenchymal transition; FA, focal adhesion; FC, fold change; GO, gene ontology; GRHL, Grainyhead-like transcription factor; IRS, immunoreactive score; MF, molecular functions; MIBC, muscle-invasive bladder cancer; MIX, urothelial carcinoma with squamous differentiation; NAC, neoadjuvant chemotherapy; NMIBC, non-muscle-invasive bladder cancer; NU, normal urothelium; OS, overall survival; RFS, relapse-free survival; RHOA, ras homolog family member A; RND3, rho family GTPase 3; RT, room temperature; SCC, pure squamous cell carcinoma; sq-BLCA, pure squamous cell carcinomas and urothelial carcinoma with squamous differentiation; sq-Cis, squamous cell carcinoma in situ; sq-Meta, squamous metaplasia; TCGA, The Cancer Genome Atlas; TMA, tissue microarrays; UC, urothelial carcinoma.

Introduction
Grainyhead-like transcription factor 3 (GRHL3) belongs to the GRHL transcription factor family, which also includes GRHL1 and GRHL2 [1].GRHL genes were first detected in a Drosophila mutant, which showed incomplete neural tube closure, epithelial barrier disorders, and a granular structure of the head caused by GRHL deficiency and thus became the eponym of Grainyhead gene [2].Beyond that, GRHL3 modulates wound healing [1] and affects the terminal differentiation of urothelial cells [3].In the urothelium of the bladder, GRHL3 is strongly expressed, especially in umbrella cells proposing an important function for maturation of umbrella cells [3].This was supported by identifying target genes of GRHL3 expression in the healthy bladder, such as cell adhesion molecules [4] and uroplakin II, known to be crucial for the urinary bladder barrier but downregulated in GRHL3-deficient mice [3,5].
In carcinogenesis, the GRHL transcription factor family has been reported to exert both tumorsuppressive and oncogenic functions, dependent on the cancer type and the tissue of origin [6].In squamous cell carcinomas (SCC) of the skin, in oral SCC, head and neck SCC as well as breast cancer, GRHL3 has been demonstrated to mediate tumor-suppressive effects [6][7][8][9][10].The tumor-suppressive function of GRHL3 in the skin has been linked with the transcriptional regulation of the direct GRHL3 target gene PTEN to control the PI3K/mTOR pathway [11].GRHL2 is known to play a protumorigenic role, for example, in oral SCC [12].However, GRHL3 can also function as tumor promoter, particularly in colorectal carcinoma [13] and diffuse large B-cell lymphomas [14].
In bladder cancer, the role of GRHL3 is less well known, especially in the context of molecular and histological subtypes.Like most carcinomas, urinary bladder cancer is a heterogeneous disease comprising the most frequent urothelial cancers but also rare subtypes like squamous-differentiated bladder cancers [15] known to show different biological and clinical aspects [16,17].SCC occurs in approximately 3% of all bladder cancers, with a squamous differentiation (MIX) in 15%.Pure SCC is associated with a more aggressive phenotype compared with urothelial carcinoma (UC) [18][19][20][21].Recently, Wezel and colleagues studied GRHL3 in urothelial cancer development showing an impaired invasion and migration potential of GRHL3 overexpressing urothelial cancer cells [22].Apart from that, no further studies of GRHL3 in bladder cancer have been described so far, and the role of GRHL3 in sq-BLCA remains unknown.Since opposing functions of GRHL3 depending on cancer type and/or differentiation status have been demonstrated in previous studies of other cancer entities [6,8], we aimed to characterize for the first time GRHL3 expression, function and associated molecular downstream targets and pathways in urothelial versus squamousdifferentiated bladder cancer.

Patient samples
Overall, our retrospective study cohort comprised n = 687 patient samples (archive of the Institute of Pathology and the RWTH centralized Biomaterial Bank (RWTH cBMB) or collected by the German Study Group of Bladder Cancer (DFBK e.V.)) diagnosed with squamous metaplasia (sq-Meta; n = 107), squamous carcinoma in situ (sq-Cis; n = 9), pure squamous cell carcinoma (SCC; n = 160), urothelial carcinoma (UC; n = 103), and urothelial carcinoma with squamous differentiation (MIX; n = 98).Normal urothelium of cystectomies (NU; n = 103) served as reference tissue.For comparing non-muscle-invasive bladder cancer (NMIBC) and its possible progression to muscle-invasive bladder cancer (MIBC), a previously published, meticulously histologically characterized cohort of NMIBC (n = 107) with patients treated at the RWTH Aachen University and the LMU Munich University Hospitals between 2012 and 2018 was also included [23,24].Experiments were in accordance with the regulations of the centralized Biomaterial Bank (RWTH cBMB) and the Institutional Review Board (IRB)-approved study and protocols of the Medical Faculty of RWTH Aachen University (EK 286/11, EK 206/09) as well as the Declaration of Helsinki.The experiments were undertaken with the understanding and written consent of each subject.Patients with sq-Meta within the urethra or trigonum were excluded since metaplasia can occur physiologically and inflammation-independent in these parts of the bladder [25].Tissue microarrays (TMA) of patient samples (n = 396) were constructed with TMArrayer (Pathology Devices, Westminster, MD, USA) as described previously [26].The characteristics of patient samples are summarized in Table S1 and for the NMIBC cohort in Table S2.

Cell lines and reagents
For in vitro studies, human bladder cell lines EJ28 (urothelial cell carcinoma; RRID:CVCL_5983), J82 (urothelial cell carcinoma; RRID:CVCL_0359), and SCaBER (squamous cell carcinoma; RRID: CVCL_3599) were used.J82 were originally obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA).EJ28 were a gift from Dr Alexander Buchner (LMU M ünchen, Germany), and SCa-BER was kindly provided by Prof Wolfgang Schulz (D üsseldorf University Hospital, Germany).EJ28 cells were cultured with RPMI 1640 medium, J82 and SCa-BER cells in DMEM high glucose medium (Thermo Fisher Scientific, Waltham, MA, USA), both supplemented with 10% fetal bovine serum (Capricorn Scientific, Ebsdorfergrund, Germany) and LGPS (200 mM L-glutamine, 50 UÁmL À1 penicillin, 50 mgÁL À1 streptomycin; Thermo Fisher Scientific), at 37 °C in 5% CO 2 .Test for potential mycoplasma infections was done on a regular basis.All cell lines have been authenticated based on single-nucleotide polymorphism (SNP) typing within the last 3 years and were already available at the Institute of Pathology RWTH Aachen University.

Nucleic acid extraction
Prior to RNA extraction, tissue was manually microdissected.RNA isolation of FFPE bladder tissues (n = 82) was performed using the Maxwell ® 16 LEV RNA FFPE kit (Promega, Mannheim, Germany) and of cryo bladder tissues (n = 102) using Maxwell ® 16 LEV simplyRNA Tissue Kit (Promega) according to the manufacture's protocols (ANP).cDNA synthesis was accomplished with the reverse transcription system (Promega Kit A3500), using 1 μg of total RNA.

Quantitative real-time reverse transcription PCR
mRNA expression of GRHL3 was quantified by realtime qPCR (RT-qPCR) on Thermal-Cycler C1000 Touch (Bio-Rad, Munich, Germany), using the iQ™ SYBR ® Green Supermix (Bio-Rad).Samples were measured in triplicate.Relative expression was quantified using the help of 2 ÀΔΔCt method with GAPDH serving as reference gene.Primer sequences and PCR conditions are described in Table S3.

GRHL3 immunohistochemistry
Immunohistochemical staining of GRHL3 protein was performed with Dako EnVision FLEX system (K8000; Agilent Dako, Waldbronn, Germany) as previously described [31] with slight modifications: FFPE sections (4 μm) were dried overnight (37 °C).Dewaxing and heat-induced antigen retrieval were performed in 10 mM citrate buffer (pH 6.0), using Dako PT Link, at 95 °C for 20 min.Endogenous peroxidase was blocked by incubation with EnVision FLEX peroxidase blocking reagent.The tissue sections were incubated at room temperature for 1 h with a primary rabbit monoclonal anti-human GRHL3 antibody (1 : 1000, ab221058; Abcam, Cambridge, UK).The primary antibody binds a polymer strand including secondary antibody and a horseradish peroxidase (HRP) enzyme.Adding the HRP substrate diaminobenzidine (DAB) chromogen solution visualizes the binding of the primary antibody via oxidation of DAB.
GRHL3 protein expression was analyzed by applying the semi-quantitative immunoreactive score (IRS) according to Remmele and Stegner [32].

Generating GRHL3 overexpressing singlecell clones
For generating stable GRHL3 single-cell clones, the cell lines EJ28 and SCaBER were transfected with the GRHL3-pCMV6-Entry vector or empty vector (Ori-Gene Technologies).72 h after transfection cells were separated by limiting dilutions, and individual cell clones were selected by incubation in geneticin (G418; EJ28: 1.00 mgÁmL À1 ; SCaBER: 1.45 mgÁmL À1 ; Thermo Fisher Scientific).Isolated single-cell clones were verified according to GRHL3 overexpression by real-time PCR and western blot analyses compared to mock control clones.RNA was isolated from adherent cells using Nucleospin RNA Plus Kit (ANP) (Macherey-Nagel, D üren, Germany) according to manufacturer's conditions.A second in vitro model reflecting urothelial carcinoma was generated by overexpressing GRHL3 (GRHL3-pCMV6-Entry vector) or empty vector (OriGene Technologies) in invasive J82 cells as stable pools (G418: 0.75 mgÁmL À1 ).

Cell growth assay
Cell count assays were used to determine cell growth.2 × 10 4 cells per well were seeded into six-well plates, and living cell number was measured every 24 h for a total of 96 h at 37 °C in 5% CO 2 , using the CASY cell counter (OLS OMNI Life Science, Bremen, Germany).

XTT proliferation assay
Cell growth and proliferation rate of stable GRHL3 and mock clones were analyzed with the XTT cell proliferation kit II (Hoffmann-La Roche, Basel, Switzerland) according to manufacturer's conditions.The proliferation rate was measured every 24 h for a total of 96 h with ELISA reader Infinite M200 (Bio-Rad) at absorbance λ = 492 nm.

Migration assays
Migration was analyzed by performing a wound healing and Boyden chamber assay.For wound healing, cells were treated with mitomycin C (2.5 μgÁmL À1 ; Sigma-Aldrich, St. Louis, MO, USA) to inhibit cell proliferation [36].A wound (500 μm) was set with the help of a pipette tip.Cell-free areas were documented every 8 h for a total of 96 h.Images were captured with an Axiovert 100 TV microscope (Carl Zeiss, Oberkochen, Germany) using CELLSENS DIMENSION software 2.3 (Olympus, Tokyo, Japan), and the cell-free areas were quantified with IMAGEJ software.
Boyden chamber assays were performed as previously described [37] with slight modifications using 24transwell plates with chamber inserts (6.5 mm; Costar, Corning, New York, NY, USA) including a polyethylene terephthalate (PET) membrane (pore size 8.0 μm).Cells were starved for 16 h in serum-free medium and seeded into the upper chamber of the insert (15 × 10 4 cells per chamber).Upon chemotactic stimulus with medium containing 10% fetal calf serum (FCS) in the lower chamber, migrated cells were stained in 0.05% crystal violet solution (10% formaldehyde, 80% methanol) and quantified with IMAGEJ software.

Matrigel invasion assay
To assess cell invasion, 24-transwell plates with chamber inserts (6.4 mm; Corning BioCoat Matrigel Invasion Chamber), including a PET membrane (pore size 8.0 μm), and coated with Matrigel were used.Cells were starved for 16 h with serum-free medium, seeded into the upper chamber of the inserts (50 × 10 4 cells per well), and incubated for 22 h (37 °C, 5% CO 2 ).Due to the chemotactic stimulus of medium containing 10% FCS in the lower chamber, cells invade through the porous membrane.The invasion was measured after crystal violet staining (0.05%; 10% formaldehyde, 80% methanol) and quantified with IMAGEJ software.

Active RHO detection
Cells were cultured at a density of 1 × 10 5 cellsÁcm À2 and allowed to attach overnight.After harvest with 1 mM PMSF, lysis buffer cells were pelleted (16 000 g, 15 min, 4 °C).Protein concentrations of cell lysates prepared from supernatants were quantified using a Pierce™ BCA Protein Assay kit (23 225, Thermo Fisher Scientific).RHO kinase activation was measured by using Active RHO Detection Kit (#8820, Cell Signaling) according to the manufacturer's instructions.Provided spin cups were inserted into collection tubes and treated with 100 μL glutathione resin and 400 μg GST-Rhotekin-RBP.500 μg total protein of cell lysis supernatants was added to treated spin cups.GTPγS and GDP were used to provide positive and negative controls appropriately.Reaction mixtures were incubated at 4 °C for 1 h with gentle rocking.After centrifugation (6000 g, 30 s, RT) and three times washing with washing buffer proteins were eluted in 50 μL 2× SDS sample buffer (6000 g, 2 min, RT).The eluted samples were heated for 5 min at 95 °C and were used for SDS electrophoresis.
Sample treatment with fluorescently coupled secondary antibodies and phalloidin (all Invitrogen, Waltham, MA, USA) was equivalent to primary antibody treatment for 45 min at RT.
Samples were analyzed with the Airy scan detector of an LSM880 (Carl Zeiss) using the 'Resolution vs. Sensitivity' mode equipped with a Plan-Apochromat 63×/1.4Oil DIC M27 objective (Carl Zeiss).Raw Airy scan image data were processed with the ZEN BLACK software (Carl Zeiss) in 2D mode.All images of vinculin were acquired focusing on the cell-glass interface, while ZO-1 and desmoplakin were acquired focusing on the centered cell plane.For statistical analysis of focal adhesions, migrating EJ28 cells contacting less than two cells were imaged.EJ28 cells migrating into the scratch were imaged and analyzed.

Focal adhesion detection
To detect focal adhesions (FA), an in-house developed PYTHON (version 3.8) program was used.In the first step, a cell mask was created to separate the cell from the background.Therefore, the actin and vinculin image channels were summed up into one image, and a threshold was calculated using Li's iterative minimum cross-entropy method [39][40][41].The threshold was multiplied by 0.7, and all pixels above this threshold were defined as cells.To remove artifacts, only the biggest label within this mask was considered further.
To generate the FA mask, the local z-score [42] (51 × 51 pixel environment, z-score threshold = 1) was calculated from the Gaussian smoothed (sigma = 1) vinculin channel.FAs were only considered within the previously created cell mask.Subsequently, FAs with a size of less than 150 pixels were rejected as well as FAs where the mean intensity of the FA was less than the mean intensity of its surrounding pixels.

Next-generation sequencing and bioinformatic analyses
Total RNA was isolated from cells from different clones as described above.The quantity of RNA was analyzed with the NanoDrop 1000 (Thermo Scientific NanoDrop Technologies).RNA quality control was performed with the Eukaryote Total RNA Nano Series assay using the Agilent 2100 Bioanalyzer system (Agilent Technologies, Inc., Santa Clara, CA, USA).An RNA Integrity Number (RIN) of at least 8 verified the quality of the RNA.Illumina libraries were generated from 1 μg of total RNA using the NEBNext Ultra II Directional mRNA Library Prep Kit for Illumina as described by the manufacturer (New England Biolabs, Frankfurt am Main, Germany).The libraries were run on an Illumina NextSeq500 platform using the High Output v2.5 kit with 150 cycles (Illumina, San Diego, CA, USA).FASTQ files were generated using bcl2fastq (Illumina).Sequencing data were processed using the nf-core/RNA-seq pipeline 3.4 [43] with the minimal command.In brief, reads were trimmed by TRIM GALORE 0.6.7 [44] and aligned to the Gencode human genome (GRCh38) v28 using STAR 2.7.9a [45].Quality assessment was done by PCA plots generated with VST (variance stabilizing transformation) and normalized for all genes but excluded for genes with count lower than 5 in all samples (see Fig. S2).Genelevel assignment was done using FEATURECOUNTS 1.6.457[46].Transcript-level quantification was performed by SALMON v1.5.2 [47].The downstream analysis was done using custom scripts in R version 4.1.1.Differential expression analysis was done with custom script with DESEQ2 package v1.32.0 [48].Genes with adjusted P-value < 0.05 between the analyzed sample groups were identified as differentially expressed genes (DEG).The NGS data were deposited in the genomics data repository Gene Expression Omnibus (GEO accession: GSE241298).Genome-wide gene expression profile data were applied in gene set analysis using R package piano_2.8.0 [49].Curated gene sets for hallmark genes, pathways, and Gene Ontology (Biological Process, Cellular Component, and Molecular Function) in Human MSIGDB (v2023.1)were gene set collections used in this analysis.The top enriched gene sets were chosen by non-adjusted P-value with nondirectional option.String-db [50] was used to visualize putative interaction networks of enriched pathways and biological processes.

Statistical analysis
Statistical analyses were performed using IBM SPSS 27.0.1.0(SPSS, Chicago, IL, USA) and GRAPHPAD PRISM 9.0 (GraphPad, San Diego, CA, USA).For comparison of two groups, non-parametric Mann-Whitney U tests were performed.More than two groups were compared with Kruskal-Wallis and Dunn's multiple comparison tests.Spearman correlation coefficients were determined between clinicopathological parameters and GRHL3 mRNA and protein expression levels.Overall survival (OS) and relapse-free survival (RFS) rates were analyzed with Kaplan-Meier curves and log-rank tests.For all tests, two-sided P-values less than 5% were considered statistically significant.

Prognostic impact of GRHL3 expression associated with bladder cancer subtypes
Next, associations of GRHL3 RNA and protein expression with clinicopathological characteristics including tumor stage, histological grading, and nodal status were determined by applying the Fisher exact test in the context of subtype-specific cohorts.We observed a significant association between increased GRHL3 mRNA expression and a higher tumor stage in urothelial but not in squamous bladder cancers (Tables S4 and S5).No significant association was shown for GRHL3 protein expression in UC samples to any of the analyzed clinicopathological parameters (Table S6).In turn, for sq-BLCA we revealed a strong inverse correlation between GRHL3 expression and tumor stage (P = 0.036, r = À0.258;Table 1).Pretreatment with neoadjuvant chemotherapy (NAC) showed no association with GRHL3 expression (Table 1 and Tables S4-S6).Correlating clinicopathological parameters and nuclear GRHL3 protein expression in NMIBC, a strong positive association with grading (P = 0.008, r = 0.268) and gender (P = 0.004, r = 0.320) was found (Table S7).This correlation of grading and nuclear and cytoplasmic GRHL3 protein expression was proven over all UCs (NMIBC and MIBC; HG/LG: P = 0.021, r = 0.238; Table S8).
Additionally, we used TCGA data sets to determine patient outcome by calculating univariate Kaplan-Meier survival curves to classify into histological and molecular subtypes.Since GRHL3 is weakly expressed in basal/squamous (BASQ) tumors, we dichotomized expression levels into low and moderate to strong GRHL3 expression by considering the lower 25% quartile to define the cut-off.Based on that, relapsefree survival (RFS) rate of patients diagnosed with BASQ UC tended to a slight longer RFS in the setting of increased GRHL3 expression (Fig. 2) but without significance.Significance was also missed for luminal UC; however, none of the luminal bladder tumors characterized by low GRHL3 mRNA expression presented an event.Classifying this data set by histological parameters, we demonstrated for patients with sq-BLCA that RFS was significantly (P = 0.007) longer with medium to strong GRHL3 expression (mean RFS: 2191.4 days AE568.8,95% CI 1076.6-3306.3)compared to tumors with low GRHL3 expression (mean RFS: 368.9 days AE70.8, 95% CI 230.1-507.7)(n = 28; Fig. 2).

GRHL3 mediates subtype-specific growth of bladder cancer cells and colony forming
Since a reverse GRHL3 expression was revealed for squamous and urothelial bladder cancers, we aimed to analyze the functional role of GRHL3 in different bladder cancer subtypes in vitro.We stably overexpressed GRHL3 in urothelial EJ28 and the basal/squamous SCaBER cell lines and extended with a second model for invasive urothelial carcinomas by overexpressing GRHL3 in urothelial stable J82 pools.Overexpression of GRHL3 in EJ28 cells verified by RT-qPCR and western blot analyses (Fig. 3A) caused a significantly (P < 0.001) increased cell growth for 96 h when compared to mock controls (median growth: +32%) (Fig. 3B,C).These findings were confirmed by independent XTT assays demonstrating increased cell proliferation of GRHL3 overexpressing clones (median cell growth: +26.2%, P < 0.05) compared to controls (Fig. 3D).In turn, GRHL3 overexpression (Fig. 3E) caused impaired cell growth (median growth: À52.6%, P < 0.001) in SCaBERderived clones compared to mock controls (Fig. 3F,  G).Blocked cell growth by GRHL3 was confirmed by XTT assays (median growth: À9.6%, P < 0.05; Fig. 3H).

GRHL3 affects motility and invasion of bladder cancer cells depending on the cancer subtype
Performing a wound healing assay for both models, we further demonstrated an effect on cell motility of GRHL3 SCaBER clones compared to mock controls (Fig. 4A): The wound size at 24 h of SCaBER GRHL3 clones closed 45% in mean of the original scratch size, while mock clones were already able to repopulate 71% of the wounded area (P < 0.05; Fig. 4B,C).For EJ28 cells, we did not observe any significant impact of GRHL3 on wound closure (data not shown).However, based on transwell assays allowing a specific documentation of cell migration without any unspecific cell growth effects, an increased migration capacity of GRHL3 overexpressing EJ28 clones was shown compared to mock controls (median cell motility: +196%, P < 0.01; Fig. 4D).Contrary to that, GRHL3 overexpression in SCaBER cells was characterized by reduced motility (median cell motility: À52%, P < 0.05) as compared to controls (Fig. 4E) confirming the wound healing data.
Finally, we analyzed the impact of GRHL3 overexpressing cells to invade through a simulated basal membrane using Matrigel-coated transwell assays.In EJ28, GRHL3 expression fostered the invasion potential (median invading cells: +39%; Fig. 4F), while causing a lower ability of cell invasion (median invading cells: À59%) in SCaBER cells compared to mock clones (Fig. 4G).

GRHL3 modulates cell-matrix interactions and RHO activity
Next, transcriptomic analyses of independent GRHL3 overexpressing SCaBER (n = 3) and EJ28 (n = 2) clones compared to corresponding independent mock clones (for each model n = 3) were performed.Gene set enrichment analysis was used to gain insights into GRHL3-affected hallmarks, pathways, and gene ontology (GO), that is, biological processes (BP), cellular compartments (CC), and molecular functions (MF) in context of used cell lines.Top 15 enrichment results for each category and cell lines are listed in Table S9.Overall, we observed enrichment of genes associated with epithelial-to-mesenchymal transition (EMT) (hallmarks), integrin complexation (CC), collagen receptor activity (MF), and integrin-mediated cell adhesion (BP).There were also cell line-specific enrichments: In SCaBER, we observed enrichment of sets of genes involved in actin cytoskeleton and the myosin complex (MF and CC).Heatmaps and associated interaction networks for integrin-mediated cell adhesion in EJ28 and actin cytoskeleton in SCaBER are visualized in Fig. S4.In EJ28 cells, chromatin binding and RNA polymerase II transcription regulatory (MF and CC), among others, were enriched upon GRHL3 expression.
Typical characteristics of the subtype-specific cell identity were observed in cell lines mirrored by tight growth patterns with pronounced cell-cell coupling by desmosomes and tight junctions in squamous-like SCa-BER cells.Urothelial EJ28 cancer cells showed only weak intercellular coupling indicated via sparse tight junction and desmosome formation at distinct contact sites (Fig. S6).At the edges of the scratch mesenchymal-like single-EJ28 cells rich in contractile actin stress fibers migrated into the wounded area (Fig. S5).In contrast, the collective cell migration of SCaBER clones was accompanied by reduced stress actin fiber formation.SCaBER cells exhibited cytoskeletons with pronounced cortical actin structures.Both phenotypes were found to be independent of GRHL3 expression.However, calculating the number and size of FA sites demonstrated a clear regulation of FAs in GRHL3-expressing EJ28 clones, which correlate with enlarged cell sizes (Fig. 5B).The significant increase in focal adhesion number, size, and vinculin intensity reflected GRHL3-induced cell-matrix adhesion dynamics.The larger cell size (Fig. 5B) and the actin stress fiber-rich phenotype demonstrated GRHL3-dependent cell spreading and migration during wound healing in EJ28 cancer cells.This finding further suggested that the observed shift in FA dynamics is linked with enhanced contractile cell force in these urothelial cancer cells.So we analyzed RHO GTPase dynamics as primary mediators for actomyosin force-based cell motility using a pulldown assay.Indeed, we observed a nearly complete loss of RHOA GTPase activity in SCaBER clones overexpressing GRHL3 compared to mock control clones, indicating a complete imbalance of RHO-RAC mediated signaling cascade.In contrast, GRHL3-expressing EJ28 clones showed only a slight reduction in RHOA activity compared to controls (Fig. 5C), suggesting a functional and putative dynamic RHO-RAC turnover [51].
Finally, we studied the influence of GRHL3 expression on cell-matrix adhesion capability of our in vitro models.An increased adhesion on Matrigel of GRHL3-expressing EJ28 cells (median + 135%, P < 0.001) was observed compared to mock clones (Fig. 5D).This result is consistent with the previously observed increase in focal adhesion formation (Fig. 5B).In turn, SCaBER cells overexpressing GRHL3 were characterized by impaired cell-matrix adhesion, that is, GRHL3 SCaBER clones showed significantly (P < 0.001) decreased attachment to Matrigel by 29% in median (Fig. 5E).

Transcriptomic profiling reveals GRHL3specific gene patterns and regulation of EIF4E3 in dependence on the cell line model
Based on RNA-seq profiling data, we aimed to decipher subtype-specific gene patterns and putative target genes of GRHL3 in SCaBER and EJ28 tumor cells and identified GRHL3-regulated differential expressed   S10 and S11).Overall, n = 174 genes were inversely regulated, and n = 69 genes were induced in GRHL3-expressing EJ28 clones.In SCaBER clones, n = 37 genes were repressed and n = 49 genes were co-expressed with GRHL3.Principal component analysis of RNA-seq data is shown in Fig. S2.Downregulation of RHOG, a potential target gene of GRHL2 in non-small lung cancer [52], was observed in GRHL3-expressing SCaBER cells.Inverse regulation of GRHL3 and RHOG was confirmed by RT-qPCR using independent GRHL3-expressing SCa-BER single-cell clones compared with mock clones (Fig. S7).Interestingly, only a small overlap of three genes (0.9%: TMEM98, EIF4E3, BMF ) between both cell lines was observed (Fig. 6C).Of these genes, only TMEM98 was similarly expressed in both models, whereas both EIF4E3 and BMF have been identified as potentially reversely regulated (Fig. 6D).EIF4E3 was confirmed to be upregulated in a set of n = 5 independent SCaBER single-cell clones overexpressing GRHL3 as compared to n = 4 independent mock clones, whereas downregulation missed significance in EJ28 GRHL3 clones (Fig. 6D).
A putative binding site for GRHL3 was found in intron 5 (GAAACCAGCCTGACAGGATTG) of the EIF4E3 gene comprising the characteristic GRHL binding motifs, that is, two adjacent repeats of Grainyhead consensus sequences with two tandem core CNNG motifs set apart by five bases [5].Co-regulation of GRHL3 and EIF4E3, a putative tumor suppressor [53], was then verified by Spearman rank correlation in the TCGA BLCA data set (Spearman r = 0.257, P < 0.001, Fig. 6E).Of clinical relevance, increased expression of the GRHL3-EIF4E3 axis tends to predict prolonged overall survival of patients with bladder cancer as compared to those with low expression of GRHL3-EIF4E3, pointing to a pathophysiological impact of this coregulation in bladder cancer (Fig. 6F).

Discussion
In the present study, we show evidence for a dual role of GRHL3 in the context of histological and molecular subtypes of cancers arising from the urothelium.Our comprehensive expression study provides insights into subtype-specific GRHL3 expression associated with different clinicopathological associations and patients' outcomes.Previously, Wezel et al. [22] provided insight into the potential role of GRHL3 in urothelial carcinomas, with a special focus on urothelial growth patterns focusing on a papillary and invasive urothelial model.They also showed that luminal non-invasive RT4 tumor cells seem more aggressive when downregulating GRHL3, while overexpression slightly impaired cell migration and invasion in T24 cells [22], hence postulating a generally tumor-suppressive function of GRHL3 in urothelial carcinomas (UC).However, GRHL3 expression was determined in cell lines only.Here, we determined GRHL3 mRNA and protein expression considering more than 600 human tissue samples and 314 TCGA samples identifying high GRHL3 mRNA and protein levels associated with increased tumor stage and a trend toward shorter recurrence-free survival in UC.We demonstrated that loss of GRHL3 expression was abundant and associated with increased tumor stage and unfavorable prognosis in squamous bladder cancers.The correlation of GRHL3 and tumor stage was previously shown by Yuan et al. [54] in colorectal cancer.In breast cancer, Xu et al. [10] reported subtypespecific expression of GRHL3 with lower levels in triple-negative breast cancer, and persistent GRHL3 expression was generally associated with longer overall survival of patients with lymph node metastases.
Consistent with these findings, our in vitro models provide functional evidence that GRHL3 is closely associated with a distinct molecular background of cells depending on histological differentiation.Overexpression in urothelial EJ28 and J82 cells drives oncogenic features, whereas squamous-differentiated SCaBER cells cause opposite, that is, tumor-suppressive, functions mediated by GRHL3.In fact, GRHL3 expression fostered cell growth, colony formation, cell migration, and invasion in EJ28 clones and increased colony formation in J82 cells.Contrary to that, GRHL3 overexpressing SCaBER clones were characterized by impaired cell growth and effectively suppressed migration, invasion, and the clonogenic potential.So far, opposed roles of grainyhead-like proteins have been shown between various tumor entities [6], and suppressive functions have been often observed in squamouslike cancer types.Loss of GRHL3 expression was first described in SCC of the skin [11,55] associated with tumor-suppressive properties.In head and neck SCC, Saffarzadeh et al. [56] also demonstrated downregulation of GRHL3, whereas oncogenic properties of GRHL3 were described in colorectal cancer [13,54] and diffuse large B-cell lymphoma [14].
A rationale for the divergent role of the GRHL gene family might be its function as a pioneer transcription factor, directly binding to chromatin, especially regulating epithelial gene expression.Klein et al. [57] observed specific binding of GRHL3 at distinct enhancers and promotors depending on the differentiation status of keratinocytes.They demonstrated that the chromatin relocates GRHL3 binding and enhancers to regulate the irreversible commitment of progenitor keratinocytes to differentiation and a reversible transition to migration.Thus, accessible DNA regions in individual cells may coin the downstream consequences of GRHL3 activity as observed in our urothelial EJ28 and squamous SCaBER model.This context-specific regulation of gene sets by GRHL3 is mirrored by a marginal overlap of significant gene expression patterns between our models.Interestingly, affected biological processes showed greater similarity, whereas underlying sets of enriched genes mostly vary, which might be a basis for the reversely functional outcome of GRHL3 expression in the urothelial and squamous in vitro model.In both lines, GO analyses suggest that epithelial-to-mesenchymal transition (EMT) might be affected, which is consistent with previous studies potentially each with a different outcome for cell behavior.During the progression of various tumor entities, GRHL transcription factors appear to mediate the broad spectrum of involved players of the EMT process modulating EMT/MET dynamics [58].A decisive role of GRHL3 in tumor development is also attributed to cell adhesion [59] as demonstrated in our bladder cancer models.Overall, we observed changes in cell-matrix interactions, both functionally and on the expression level of distinct surface receptors such as integrins confirmed by the modulated assembly of FA sites.
In single-migrating EJ28 cells, GRHL3 re-expression increased with the FA size and number.This finding emphasizes an impact of GRHL3 on cell-matrix adhesion-mediated motility, similar to other studies dealing with FA and showed that size predicts migration speed, for instance, in highly invasive fibrosarcoma cells [60].This notion was functionally supported by increased cell migration, cell-matrix adhesion, and invasiveness of GRHL3-expressing EJ28 clones.The impact of GRHL3 on wound healing and cell migration was first described in epidermis, and modulated cell adhesion potential of tumor cells triggered by GRHL3 was demonstrated [61][62][63].ARH-GEF19 (RhoGEF19) was previously identified as a target gene of GRHL2 and GRHL3 affecting the maintenance of epidermal differentiation by activating RHOA and regulating the planar cell polarity signaling pathway in epidermal wound repair [63,64].Consistent with these findings, we revealed modulated RHO GTPase activity in bladder cancer cells.In EJ28 cells, slight effects on RHOA activity could be the outcome of a dynamic turnover between RHO and RAC signaling causative for higher cellular motility, as mesenchymal migration is characterized by RAC activity at the leading edge, whereas RHO is active toward the cell rear, together resulting in a lamellipodium at the leading edge [65].In turn, activity loss of RHOA in SCaBER cells suggests a substantial imbalance of underlying processes.However, squamous SCaBER cells form generally less actin stress fibers associated with collective cell migration capacities.Hence, the impact of retarded RHO activation on cytoskeleton and FA remodeling remains unclear.Since RHO GTPases are known to control various processes like cell-cell adhesion, vesicle trafficking, cell cycle, or collective cell polarization [66], implications of RHO activity loss caused by GRHL3 expression in squamous cancer cells might be multifactorial and should be further addressed in future studies.

Conclusion
In conclusion, our comprehensive expression data of GRHL3 in different bladder cancer subtypes are associated with distinct prognostic implications.The functional outcomes in vitro argue for a histological subtype-specific impact of GRHL3, that is, a tumor-suppressive effect of GRHL3 in sq-BLCA and, in contrast, a more oncogenic property in urothelial carcinomas with a putative role of the GRHL3-EIF4E3 expression axis involving integrin and actin-associated processes and pathways and factors like ARHGEF19.Since ARHGEF19 (RhoGEF19) is known to activate RHOA [62], while RHOE has been postulated as antagonist of RHOA [67] and affects EIF4E family members by impairing their cap-dependent translational functions [68], a putative interdependent feedback between these is conceivable but remains speculative at this stage.Bearing in mind that Wezel et al. observed tumor-suppressive effects in urothelial T24 cells, the configuration of the molecular context such as chromatin structures might be finally crucial for GRHL3 function.Thus, our study along with prior research efforts gain further insights into role of GRHL3 transcription factor helping to further decipher the clinically important pathways of bladder cancer subtypes.

d
According to WHO 2004 classification.e Neoadjuvant chemotherapy (NAC).

Fig. 5 .
Fig. 5. GRHL3 modulates cell-matrix interactions and intracellular signaling.SCaBER and EJ28 cells adhere to substrate via focal adhesions.(A) Confocal images of representative GRHL3 overexpressing clones (overall n = 2, illustrated clones: #18 (SCaBER) and #2 (EJ28)) and corresponding mock clones (overall n = 2, illustrated clones: #28 (SCaBER) and #43 (EJ28)) show the actin cytoskeleton (phalloidin, white), focal adhesions (FA) (Vinculin, green) of cells adjacent to a scratch wound introduced to a confluent cell layer 16 h prior to fixation.Image contrast was adapted for the accurate representation of FA presence in both cell lines and does not represent relative vinculin signal intensity.Scale bar = 50 μm.(B) Statistical analysis of FA intensity, mean size, and number as well as cell size in EJ28 cells migrating into the scratch wound shown in A before fixation.Scatter plots show the respective values calculated for GRHL3-expressing (n = 77) and mock cells (n = 77) from vinculin signal.Bars show the median (red) and 95% confidence interval (black).**P < 0.01; ***P < 0.001 (Mann-Whitney U tests).(C) Analysis of integrin downstream signaling.Representative western blot results illustrate activated RHOA GTPases in independent GRHL3-expressing EJ28 and SCaBER clones compared to mock clones.Total RHOA served as loading control.(D, E) Cell adhesion assay (n = 4 independent experiments).(D) Box plot analyses of cell-matrix adhesion illustrate a significantly higher relative number of adhesive cells of EJ28 GRHL3 clones (n = 4) as compared to EJ28 mock clones (n = 4).***P < 0.001 (Mann-Whitney U tests).(E) Box plot analyses of cell adhesion of SCaBER GRHL3 clones (n = 4) show significantly reduced cell adhesion compared with SCaBER mock clones (n = 4).***P < 0.001 (Mann-Whitney U tests).For all illustrated box plots: Horizontal lines: Grouped medians.Boxes: 25-75% Quartiles.Vertical lines: Range, maximum and minimum.

Table 1 .
Clinicopathological parameters in relation to GRHL3 expression of sq-BLCA in patient cohort.Significant P-values are marked in boldface.
a Dichotomized at 25% quartile.b Only patients with primary bladder cancer were included.c Fisher's exact test.