In vitro and in silico assessment of the effect of WWOX expression on invasiveness pathways associated with AP-2 transcription factors in bladder cancer

Background WW Domain Containing Oxidoreductase (WWOX) belongs to the unusual tumor suppressors, whose molecular function is not fully understood in bladder cancer, especially regarding interaction with Activator Protein 2 (AP-2) α/γ transcription factors. Thus, using lentiviral systems we created an in vitro model overexpressing or downregulating WWOX in CAL-29 cell line to assess invasiveness pathways. Surprisingly, while WWOX overexpression was accompanied with increased expression of both AP-2 factors, its downregulation only affected AP-2α level but not AP-2γ which remained high. Methods Using cellular models and unpaired t-test or Wilcoxon test, we investigated significant changes in biological processes: clonogenicity, extracellular matrix adhesion, metalloproteinases activity, 3D culture growth, proliferation, mitochondrial redox potential and invasiveness. Relative gene expression acquired through Real-Time qPCR has been analyzed by Welch's t-test. Additionally, using oncoprint analysis we distinguished groups for bioinformatics analyzes in order to perform a follow-up of in vitro experiments. Results Downregulation of WWOX in bladder cancer cell line intensified ability of single cell to grow into colony, mitochondrial redox potential and proliferation rate. Moreover, these cells shown elevated pro-MMP-2/9 activity but reduced adhesion to collagen I or laminin I, as well as distinct 3D culture growth. Through global in silico profiling we determined that WWOX alters disease-free survival of bladder cancer patients and modulates vital processes through AP-2 downstream effectors. Conclusions Our research indicates that WWOX possesses tumor suppressor properties in bladder cancer but consecutive examination is required to entirely understand the contribution of AP-2γ or AP-2α. Supplementary Information The online version contains supplementary material available at 10.1186/s12894-021-00806-7.

Several subtypes of BLCA are recognized, and these are distinguished on the basis of their molecular profile. The University of North Carolina (UNC) classification assumes two subtypes (luminal and basal bladder cancer), while the MD Anderson Cancer Center (MDA) recognizes three (basal, luminal and TP53-like) and The Cancer Genome Atlas (TCGA) database four [3]. BLCA is a heterogeneous disease in terms of its molecular and clinicopathological status; it is associated with disruptions of pathways, with the best known being FGFR3/ RAS, PI3K/AKT/mTOR or TP53/RB1 [4].
One of many tumor suppressor genes whose function remains unclear in bladder cancer (especially at the molecular level) is WWOX. The gene is located in a common fragile site FRA16D (16q23.1-q23.2) and belongs to the group of tumor suppressor genes inactivated through haploinsufficiency. Mutations inactivating both WWOX alleles have not been found in cancer; however, the loss of one allele is sufficient to affect its normal biological function [5]. One of the mechanisms regulating WWOX expression in bladder cancer is methylation of both the promoter region and the first exon of the gene [6,7]. In addition, our previous research proved a frequent (47.5%) loss of heterozygosity (LOH) located in intron 1 (microsatellite marker D16S518) of the WWOX gene, together with positive correlation between the levels of WWOX and CCND1 mRNA in bladder cancer patients [7]. Furthermore, in vitro study revealed that cigarette smoke extract (CSE) treatment could inhibit the expression of WWOX in a time-dependent manner, through the induction of CpG islands methylation near to the transcription start site (TSS) [8]. Until now, only one publication reported that WWOX may regulate cell cycle and apoptosis but also reduces tumorigenicity in mice; research was done on 5637 bladder cancer cell line (grade 2) [9]. Studies on bladder cancer patients indicate a relationship between progressive loss of WWOX protein expression and higher tumor grade, more advanced stage, larger tumor size and shorter progression-free survival [10]. WWOX is known as a global modulator of gene expression and cell metabolism. It encodes a protein containing two N-terminal WW domains and a centrally-located short chain dehydrogenase domain (SDR), suggesting that it acts as a steroid dehydrogenase and participate in the regulation of steroid hormone receptor interactions [11][12][13]. The first WW domain interacts with a number of partner proteins, including ERBB4 [14,15], transcription factor AP-2γ [16,17], YAP [14], c-Jun [18], HIF1α [19,20], p73 [20,21], DVL2 [22], RUNX2 [23], SMAD3 [24] and GLI [25].
A recent study, by Abu-Odeh et al. identified a set of 240 proteins that may be able to bind with WWOX, either classically (via PPxY or LPxY motifs) or through non-canonical proline-rich motifs. These new partners are believed to be involved in the transcription process, RNA processing and splicing, chromatin remodeling, metabolism, and signaling pathways [20].
One of WWOX partners is AP-2γ which belongs to the AP-2 transcription factor family, a crucial group of TFs that regulate gene expression during early development but also in carcinogenesis. Another member of the AP-2 family, commonly interacting with WWOX, is AP-2α. Of the two, AP-2γ has been described in more detail, which is presumably due to greater PPxY motif affinity demonstrated by this protein ( 56 PPPYFPPPY 64 ) in contrast to proline-rich motif found in AP-2α ( 59 PPPY 62 ) [16]. Their transcriptional transactivation function is suppressed by the interaction with WWOX through their sequestration in the cytoplasm [16]. Both of the AP-2 factors possess ambiguous functions [26]. Despite the discrepancies observed in the literature, the prevailing opinion is that AP-2α is more likely to be downregulated during tumor progression [27,28]; however, AP-2γ is believed to be upregulated during progression (but not initiation [16]), suggesting that it has an oncogenic character. This pattern was confirmed in breast carcinoma, where AP-2γ nuclear translocation resulting from WWOX inactivation was observed during tumor progression [16].
The aim of this study was to determine the impact of WWOX on invasiveness pathways in the grade 4 bladder cancer cell line CAL-29. To provide clearer results, two in vitro models were developed: one with low WWOX expression and the other with overexpression. The differential WWOX gene expression was found to be associated with various levels of TFAP2A (encoding AP-2α) and TFAP2C (encoding AP-2γ) expression. Therefore, an in silico analysis was also performed on clinical data obtained from TCGA database since the tumor profile is changing globally in terms of WWOX and TFAP2C levels which corresponded to our research model.

Cell culture
The

Survival analysis
Disease-free survival (DFS) analysis was performed regarding the significance of WWOX, TFAP2A and TFAP2C expression alterations combined together. BLCA patients were divided into eight subgroups according to the expression level of WWOX, TFAP2A and TFAP2C determined through oncoprint analysis at cBioPortal based on mRNA expression z-scores (RNAseq V2, RSEM normalized, z-score threshold ± 0) [29,30]. Table 1 shows detailed specification of groups with expression levels of each gene and number of patients assigned. The survival curves for the groups were fitted with survfit() (survival R package) using days to last followup as time variable, person neoplasm cancer status as event indicator (dummy coding) and group assignment as strata. Kaplan-Meier plots presenting the expected duration of time until recurrence of a disease for all groups together and chosen pairwise comparisons were generated via survminer R package. Significance of DFS between specific pairs of survival curves was tested with log-rank test. Hazard ratios (HR) with 95% confidence intervals (95% CI) and p values were computed using univariate Cox proportional hazards model accompanied by test of proportional hazards with Schoenfeld residuals (coxph() and coxzph() functions).
Differences in expression of AP-2γ and AP-2α downstream targets were determined using ExpressCluster software (http://cbdm.hms.harva rd.edu/). The analysis was performed separately for each transcription factor targets. Primary input was filtered with minimum of three-fold-change in expression followed by applying K-means clustering algorithm with z-norm signal transformation method, rank correlation as distance metric, 1000 iterations and clusters (K) = 18. Finally, for the most interesting clusters the gene ontology analysis in terms of biological processes was performed using the Protein ANalysis THrough Evolutionary Relationships (PAN-THER) Classification System (Annotation Data Set: GO-Slim Biological Process) [33].

Protein extraction and western blotting
Cells were lysed using RIPA buffer with protease and phosphatase inhibitor cocktail and phenylmethylsulfonyl fluoride (PMSF). Protein concentration was determined with the Bradford method (Bio-Rad Laboratories). SDS-PAGE was followed by semi-dry transfer (45 min, 250 mA, Fast Blot, Biometra) of the proteins to a PVDF membrane (Sigma-Aldrich). The process was supported by Whatman ® Lens Cleaning Tissue (Sigma Aldrich), the use of which requires pre-soaking in TB 10 × transfer buffer. Protein transfer was confirmed by transient staining with Ponceau red. Precluding of non-specific binding of the antibodies was accomplished via membrane blocking with non-fat milk (5% solution) in TBST buffer 1 × . Anti-WWOX/AP-2γ/AP-2α (Thermo-Fisher; catalogue numbers: PA5-29701, PA5-49862 and MA5-14856, respectively) were added as primary antibodies at a dilution of 1:1000 to 1% non-fat milk in TBST buffer solution. Overnight incubation was followed by three times washing using TBST buffer. Following this, the PVDF membranes were incubated with secondary anti-rabbit antibodies conjugated with alkaline phosphatase (1:30,000) and the analyzed protein were visualized with Novex ® AP Chromogenic Substrate (Invitrogen).
Anti-GAPDH antibody was used as reference (sc-59540, Santa Cruz Biotechnology Inc.). The relative amount of protein was determined by densitometric analysis with ImageJ software.

RNA isolation and cDNA synthesis
RNA isolation was performed using TRIzol ® reagent (Invitrogen). The ImProm-II ™ Reverse Transcription System (Promega) was used for cDNA synthesis. The samples (10 µg of total RNA) were mixed with 6 µL Random Hexamers and 5 µL oligo(dT) 15 (both at concentration 0.05 µg/µL) and incubated at 70 °C for 10 min. Following this, 1 × buffer M-MLV, dNTPs, and reverse transcriptase ImProm RT-II ™ were added according to the manufacturer's protocol. cDNA synthesis was performed at the following conditions: annealing at 25 °C for five minutes, elongation of the products at 42 °C for 60 min, completion at 70 °C for 15 min. 50 μL of water was added to each sample and the samples were stored at − 20 °C.

Clonogenic assay
All variants of the CAL-29 cell line were seeded (5 × 10 2 / well) onto 6-well plates in full culture medium and incubated (10 days, 37 °C, 5% CO 2 ). The cells were fed with  full medium every four days. The cells were then fixed with 4% paraformaldehyde in PBS solution. Staining was conducted with 0.005% crystal violet (15 min, RT) and cell colonies were counted using ImageJ software. The assay was performed in triplicate for each cell variant as described previously [34].

Cell adhesion assay
The adhesiveness of CAL-29 cells (with overexpressed or downregulated WWOX gene) to four extracellular matrix proteins: collagen I, collagen IV, laminin I and fibronectin was evaluated using Corning ® BioCoat ™ plates. The cells were seeded (3 × 10 5 /well) onto 24-well adhesion plate and allowed to adhere via incubation (4 h, 37 °C, 5% CO 2 ). Adherent cells were stained (0.1% crystal violet, 10 min) and their cellular adhesion was measured spectrophotometrically at 570 nm (BioTek) after extraction in 10% acetic acid.

Zymography
Cells were seeded in 6-well plates (1.5 × 10 6 /well) and cultured for 80% confluency. Medium was changed into starving medium and the cells were cultured for 48 h (7 °C, 5% CO 2 ). The medium was then collected to measure protein concentration using Qubit Protein Assay on the Qubit 2.0 Fluorometer (Life Technologies). A gelatin (2 mg/mL) was embedded in the resolving gel during preparation of the acrylamide gel (10%). Briefly, 8 µg of protein was added to each lane and separated during electrophoresis, then SDS was removed from the gel by incubation in 2.5% Triton X-100, followed by incubation in buffer (0.5 M Tris-HCl, 2 M NaCl, 50 mM CaCl 2 , pH 7.5), overnight at 37 °C. The resulting zymogram was subsequently stained with Coomassie Brilliant Blue R-250, the gel was then washed with bleaching solution (methanol:acetic acid:water, 3:1:6). The gel areas where gelatin was enzymatically degraded appeared as clear bands against a stained dark blue background. A protein marker was used to measure the molecular weights of the MMPs; their activity was measured using ImageJ software. The assay was performed in duplicate for each cell variant.

3D culture assay
To estimate spatial growth potential, 15 × 10 3 cells with different WWOX levels were seeded on 96-well plate on a thick (2 mm) solidified layer of ECM matrix (Geltrex, Gibco) and incubated for 14 days. Cells were fed every four days by 2% solution of ECM matrix. After this period, the cells were observed under the microscope and photographed. The assay was performed in triplicate for each cell variant.

Proliferation and cell viability assays with an evaluation of mitochondrial redox activity
Proliferation rate was evaluated using 5-bromo-2′deoxyuridine (BrdU). This assay was performed on fixed cells, incubated (one hour, 37 °C, 5% CO 2 ) with europium labelled anti-BrdU antibody and visualized at wavelength OD 340 nm using VICTOR ™ X4 Multilabel Plate Reader (Perkin Elmer). The mitochondrial redox activity was estimated with Presto reagent (Invitrogen). Cell viability was measured spectrophotometrically at a wavelength of 570 nm every 10 min on a VICTOR ™ X4 Multilabel Plate Reader (Perkin Elmer).

Invasion assay
The Corning ® BioCoat ™ Matrigel ® Invasion Chambers with 8.0 μm PET membrane was used to assess the role of the WWOX gene in CAL-29 invasiveness. Cells were suspended in starving medium and then seeded (2 × 10 5 /well) onto the inner compartment of each insert, after adding full culture medium to the wells. The plate was incubated (48 h, 37 °C, 5% CO 2 ). Following this, the cells attached to the outer side of the membrane were stained with 0.1% crystal violet. The inserts were transferred to 200µL of extraction solution (10% acetic acid) and incubated on an orbital shaker. The intensity of stain extract from the cells was measured spectrophotometrically (OD 560 nm, BioTek). The assay was performed in triplicate for each cell variant.

Statistical analysis
Levene's test was used to test the homogeneity of variance. The normality of distribution was determined by the Shapiro-Wilk test; depending on the result, either the unpaired t-test or Wilcoxon tests was used for further analysis. The differences in relative gene expression between variants were determined by Welch's t-test. Results with a p value less than 0.05 were considered as statistically significant.

Construction of an in vitro CAL-29 cell line model
To investigate the relationship between WWOX, AP-2α and AP-2γ in bladder cancer, WWOX gene expression was downregulated or upregulated in the CAL-29 cell line. Overexpression of the WWOX gene was confirmed at the protein level by Western Blot analysis. A higher level of WWOX was observed (p = 0.0002) in the WWOX↑ variant (mean protein expression level 4.56 ± 0.50) than in WWOX↑ contr (mean protein expression level 1.12 ± 0.15). Silencing of the WWOX gene was also confirmed (p = 0.0172) using Western Blot assay (mean protein expression level 0.29 ± 0.08 for sgWWOX vs 1.39 ± 0.19 for sgWWOX contr variants).
The next stage examined whether differences in WWOX expression influence AP-2α and AP-2γ level. It was found that level of AP-2γ expression was higher (p = 0.0169) in the WWOX↑ variant (mean protein expression level 0.15 ± 0.009) than in WWOX↑ contr (mean protein expression level 0.09 ± 0.007). The same tendency was noticed for AP-2α − expression level was higher (p = 0.0002) after WWOX gene overexpression (mean protein expression level 0.20 ± 0.017) compared to control (mean protein expression level 0.05 ± 0.01).
All biological experiments were performed on CAL-29 cell line variants with either overexpressed or downregulated WWOX gene expression. While the level of AP-2γ was consistently high for all variants, compared to their respective controls, the levels of AP-2α varied.
Numerous significant changes in biological behavior of the investigated cell variants were observed with regard to WWOX expression level. These are thoroughly described in the following sections.

The effect of WWOX on bladder cancer cell proliferation, clonogenic survival and mitochondrial activity
Reduced proliferation activity was noticed in the WWOX↑ variant (p < 0.0001) compared to its control whereas inverse tendency concerned sgWWOX and its reference (Fig. 2a). These observations were consistent with the analysis of the ability of a single cell to grow into a colony. The CAL-29 variant overexpressing WWOX demonstrated a 2.6-fold decrease in the number of colonies compared to WWOX↑ contr (p = 0.0036) while variants representing downregulation shown significant (p = 0.0342) 1.4-fold opposite trend (Fig. 2b). Lastly, sgWWOX variant was significantly associated with increased mitochondrial redox activity compared to its control (Fig. 2c).

The influence of WWOX on the invasiveness of bladder cancer cells (metalloproteinases activity, adhesion and migratory potential)
Downregulation of the WWOX gene resulted in more than 10-fold increasement of pro-MMP-2 and 2.7-fold of pro-MMP-9 activities (p = 0.0052 and p = 0.0481, respectively) (Fig. 3a). Nevertheless, it did not appear to have any significant effect on cell motility through a basal membrane (p = 0.0921) (Fig. 3b). Moreover, sgWWOX variant demonstrated a greater ability to adhere to collagen IV (p = 0.0008) but not to collagen I, for which it presented inverse character (p = 0.0277) (Fig. 3c).

The influence of WWOX on 3D culture growth
The CAL-29 variants varied according to their ability to organize cells into colonies and form spatial structures, as well as their cell sizes and dissemination in 3D culture growth. The cells with high WWOX level eagerly organized themselves into larger aggregates compared to its control (without incrementing the number of separate colonies), while sgWWOX cells not only formed assemblages but they were scattered throughout the substratum on a larger scale than that of sgWWOX contr (Fig. 4).

Correlation between WWOX level and expression changes of AP-2γ target genes
The study also aimed to investigate whether WWOX can affect the expression of AP-2γ target genes. In the case of WWOX overexpression, the TP63 or SMAD4 genes respectively decreased by 2-fold (p = 0.0347) or increased by 2-fold (p = 0.0037) compared to the appropriate control. Analogously, the same tendencies were observed during WWOX silencing (3-fold decrease with p = 0.0038 for TP63; 5-fold increase with p = 0.0183 for SMAD4). Furthermore, for variants representing WWOX downregulation, in the sgW-WOX variant there was a 4-fold increase in IKBKB and 5-fold in CDKN1A gene expression (p = 0.0068 and p = 0.0101, respectively) as well as a 1.5-fold decrease in AP1M2 expression (p = 0.0212) compared to sgW-WOX contr. Lastly, variants representing WWOX overexpression shown a statistically significant (p = 0.0487) 1.8-fold increasement in the level of expression for the CRABP2 gene in the WWOX↑ variant compared to the control (Fig. 5).

Global profiling of BLCA patients according to WWOX, TFAP2A and TFAP2C expression: in silico analyses
It has been shown that the interplay of WWOX and TFAP2A/TFAP2C affects the biology of the tumor thus corresponding bioinformatics analyses involving BLCA cohort were performed as a follow-up of in vitro experiments.

WWOX modulation of TFAP2C/TFAP2A significantly alters disease-free survival of BLCA patients
DFS analysis was applied in groups of patients that were classified according to expression levels of WWOX, TFAP2A and TFAP2C based on oncoprint analysis (Fig. 6a). Eight groups that were considered showed differential survival time (Fig. 6b), among which A vs D (HR = 0.34; Fig. 6c) and D vs F (HR = 2.9, Fig. 6d) were the most relevant indicating essential role of signaling through TFAP2C modulated by WWOX. More specifically, combination of low WWOX and high TFAP2C/TFAP2A (group A) was less favorable on DFS than high WWOX and low TFAP2C/TFAP2A (group Fig. 2 Results of proliferation, colony formation and mitochondrial activity assays for CAL-29 cell line variants. a Differences in proliferation potential between CAL-29 variants. p < 0.0001 (****). b Bar plot presenting number of colonies in all variants of WWOX. p < 0.05 (*), p < 0.01 (**). The first column of biological assay images shows sgWWOX variant in triplicate. The second row refers to the sgWWOX contr, the third for WWOX↑ and the fourth for WWOX↑ contr. c The dependence of mitochondrial activity on WWOX downregulated variants. p < 0.05 (*), p < 0.001 (***), p < 0.0001 (****). d The dependence of mitochondrial activity on WWOX overexpressed variants D), whereas high WWOX with high TFAP2C and low TFAP2A (group F) was unfavorable in comparison with high WWOX and low both TFAP2C/TFAP2A (group D). Table 4 presents detailed statistics of DFS analysis.

WWOX modulates pivotal biological processes through AP-2γ and AP-2α downstream effectors
Downstream targets of AP-2 factors were clustered among three the most significant groups with respect to DFS analysis (A, D and F) to reveal common and distinct profiles of downstream effectors expression followed by identification of biological processes that altered genes are involved in. Regarding targets of AP-2γ, total of 9 clusters of genes were determined with 3 specific patterns related to DFS groups differing A vs F, A vs D and D vs F groups. Specifically, "Cluster 1" reflected overrepresentation of cell differentiation regulators that were downregulated in A group in contrast to F group, while opposite profile was associated with interferon signaling pathway in "Cluster 12" (Fig. 7a, b). Furthermore, cell to cell adhesion was downregulated in the group D ("Cluster 5") compared to A group, in contrast to negative regulation of Wnt signaling pathway genes that were downregulated in the latter ("Cluster 7") ( Fig. 7c, d). Finally, cell communication was upregulated in D group and downregulated in F group ("Cluster 0") in comparison to cell migration presenting opposing expression pattern ("Cluster 3") ( Fig. 7e, f ). Remaining expression heatmaps for both transcription factors are included in Additional file 2. Summary of clustering and gene ontology analysis is shown in Table 5.

Discussion
The WWOX tumor suppressor gene, located in a common fragile site, is involved in the regulation of carcinogenesis in many tissue types [34][35][36][37][38] including bladder cancer [7]. Only a few papers clarified that processes such as promoter methylation or LOH leads to downregulation of WWOX expression in BLCA patients [7,8,39] and mentioned the connection with a more aggressive cancer phenotype [10]. Therefore, the primary aim of this study was to determine the role of the WWOX gene in biological cancer-related processes in the aggressive bladder cancer cell line CAL-29 (derived from grade 4 tumor) and identify its connections with the AP-2 family of transcription factors. Our in silico analysis indicated various pathways that might be modulated by changes in the expression of the WWOX and TFAP2C genes.
Several biological assays were performed to investigate the specific effects of WWOX on tumor cell behavior regarding invasiveness in CAL-29 cell line. Firstly, a cell variant with high WWOX expression was found to demonstrate 2.6-fold reduction of anchorage-independent clonogenicity. Similar results have been noticed during in vitro experiments on other cancer types such as breast cancer [11], endometrial adenocarcinoma [34,40], the endometrial cancer cell line ECC1 [41], non-small cell lung cancer [42] or colon cancer [43]. Such trend is further supported by comparison of downregulated WWOX to its control, where sgWWOX variant exhibited higher clonogenicity than sgWWOX contr.
Moreover, the CAL-29 cell line with WWOX downregulation demonstrated disparate adhesion to collagen I, laminin and collagen IV with the first two decreased while the latter increased. Same observation of cells attachment to collagen I and laminin after WWOX silencing has been noticed in vitro for the THESC cell line [34] and additionally for laminin in MCF10 cell line [24]. This can be further supported by confirmation that collagen I expression in colon cancer HT29 cells significantly increases during WWOX overexpression [43], indicating our results for sgWWOX variant to form logical whole by complementing this statement. However, the results for collagen IV are inverse compared to the available literature [24], but we consider that these differences may reflect various processes of cell remodeling which depends on tissue type or cancer differentiation grade.  Our results also indicated that changes in WWOX expression influence the spatial growth of tumor cells. Meaning, cells having high WWOX level exhibited bigger colonies however no quantitative increase has been noted compared to control. Contrastingly, cells with WWOX silencing induced both colony size and formation of newly separated cell habitats. This is consistent with the effect of WWOX on the 3D growth of T98G glioblastoma cells for which the amount of emerging colonies along with the proliferation rate were examined [44]. Likewise, a promotion in the activity of the pro-MMP-2 and pro-MMP-9 metalloproteinase enzymes was noted in the sgWWOX variant, which may suggest an enlarged ECM-degrading potential [45].
The WWOX↑ variant also demonstrated 43% lower proliferation potential than control, which together with confirmatory trend between sgWWOX vs sgWWOX contr, is in line with results obtained for osteosarcoma and glioblastoma [44,46]. Additionally, those changes in AP-2γ targets expression that can be considered simultaneously between sgWWOX and WWOX↑ vs. controls (i.e. only TP63 and SMAD4) suggest no effect of WWOX on expression of certain AP-2γ targets. However, this is not necessarily the case with all genes, as exemplified by the tendencies visible for CDKN1A and IKBKB, which have references in the literature. In the former, an expression increase of the gene encoding the p21 protein was observed during loss of WWOX expression in breast cancer [47]. For the latter, the only report concerns primary effusion lymphomas (PELs) in which the loss of suppressor genes such as WWOX, fragile histidine triad (FHIT) or glutamate receptor ionotrophic delta 2 (GRID2) is observed. In one of the PEL subsets, IKBKB amplification is observed among others, which is not a direct confirmation of the relationship between WWOX and IKBKB expression, but gives a hint about their potential relationship [48].
Ultimately, CAL-29 cells representing WWOX downregulation demonstrated a significant increase in cell viability. This proves that potentiated mitochondrial activity is visible during WWOX depletion, which is consistent with available literature [49]. Additionally, overexpression of WWOX led to a decrease in cell viability of colon cancer cell line SW480 which complement our observations for silenced WWOX variant [43]. The aforementioned results confirmed that WWOX can regulate a wide range of biological processes taking place in the advanced stage of bladder cancer. As the WWOX interacts with many proteins including transcription factors AP-2 α and γ [26], the present study examined relationship between these proteins in bladder cancer hence we also pursued this issue in further bioinformatics analyzes.   A correlation analysis of disease-free survival time with WWOX, TFAP2C and TFAP2A levels revealed that in BLCA patients the most favorable is high WWOX expression with simultaneously low TFAP2C and TFAP2A (Fig. 6b, group D). Our results from the experimental section on CAL-29 cell line are consistent with in silico analysis for the WWOX overexpressing model. Furthermore, we obtained statistically significant results for groups characterized by low WWOX and high TFAP2C/TFAP2A expression (group A), but also high WWOX/TFAP2C and low TFAP2A (group F), in which modifications of TFAP2C and TFAP2A levels worsen DFS in comparison to the group D. Thus, we decided to perform independent clustering of AP-2γ or AP-2α targets to investigate differentially expressed genes among groups A, D, F with subsequent examination of processes in which AP-2 downstream effectors are implicated in. Considering groups of AP-2γ targets, two Clusters ("0" and "3") seemed to explain the nature of this transcription factor most accurately since these are the only clusters in which two groups differing only in TFAP2C expression (D and F, respectively TFAP2C low and high) have inverse pattern of gene expression.
By profound analyzing both Fig. 7 and Table 5, the main noticeable trend between AP-2γ targets in Clusters "0" and "3" and their expression in groups D and F is that the level of AP-2γ negatively correlates with the expression of their targets from "Cluster 0", while it correlates positively with its effectors from "Cluster 3". In "Cluster 0" we observed genes upregulated in group D which further supported its favorable outcome during survival analysis. Examples are COL4A6, WNK2 and STK32B whose roles have been assigned in processes such as prevention of early invasion stages, negative regulation of epidermal growth factor receptor signaling and opposite correlation with tumor size, respectively [50][51][52]. Regarding "Cluster 3" which was elevated in group F, the gene capable to represent the tendency is CASS4 whose overexpression has been associated with promotion of lung cancer invasion by inhibiting E-cadherin expression and activating AKT signaling pathway [53]. This is consistent with gene ontology description indicated by the bioinformatics database, which concerns cell migration for "Cluster 3".
Nevertheless, we consider it appropriate to broaden the analysis of the remaining clusters not only obtained from the AP-2γ targets but also from AP-2α analysis, where processes such as cell motility, migration, positive regulation of T-cell proliferation, cell adhesion or DNA repair were noted. However, due to the lack of proper and significant bioinformatics model that will directly explain TFAP2A participation in specific biological processes, we cannot conclude about its characteristics. Following in vitro assays are required to elaborate the properties of both AP-2γ and AP-2α factors depending on different levels of the WWOX gene.

Conclusions
To conclude, our findings suggest that WWOX deficiency may have an impact on the invasion process in the bladder. This is supported by the fact that during WWOX downregulation, an increase in both mitochondrial redox potential (indicating greater cell viability) or the metalloproteinases' activity (suggesting ECM-degrading characteristics) were observed for sgWWOX CAL-29 cell line variant. Moreover, these cells demonstrated higher proliferation potential and ability of a single cell to grow into a colony. The subsequent analyzes are required to reveal the additional contribution of AP-2γ or AP-2α.
Bioinformatics analyses in bladder cancer patients found high expression of TFAP2C gene to be associated with an enhancement of adverse processes related to cancer progression, probably due to its oncogenic nature. Further research based on the dual modification of the AP-2α and AP-2γ expression is needed to precisely determine the nature of the interaction between WWOX and particular transcription factors in bladder cell line models also in different stage of cell differentiation.